Escolar Documentos
Profissional Documentos
Cultura Documentos
Atoms,
Molecules and
Optical Physics 2
Molecules and Photons - Spectroscopy
and Collisions
Graduate Texts in Physics
Series Editors
Professor Richard Needs
Cavendish Laboratory
JJ Thomson Avenue
Cambridge CB3 0HE, UK
rn11@cam.ac.uk
To my Wife Gudrun
CPS
Preface
The two textbooks on Atomic, Molecular and Optical (AMO) physics presented
here aim at providing something like the canonical knowledge of modern atomic and
molecular physics together with a first entry into optical physics and quantum optics.
All of these topics constitute a vital area of active and highly productive research in
physics. And in spite of, or perhaps even because of its remarkable history the field
continues to constitute an indispensable basis for any more profound understanding
of nearly all branches of modern physics, physical chemistry and partially even
biological and material sciences. Specifically the latter appear to become more and
more based on genuine molecular concepts.
We want to address on the one hand advanced students of physics and physical
chemistry, who have to study these topics within their respective curricula. At the
same time, however, these textbooks should be useful to all those who discover
in different contexts that they miss some essential basics from this field and who
seek for suitable means to acquire that knowledge. Of course, we also address quite
specifically Ph.D.-students or young researchers who start for the first time their own
activities in the field – or just want to know more about it. They will find here reliable
knowledge and stimulating challenges for their own work. We thus have tried not
only to provide the essential basics for working with these topics, but whenever
possible also to inform the interested reader about the present state-of-the-art and to
allow her or him a glimpse on today’s cutting edge research.
The general remarks as well as details about formats, notation, units, and typog-
raphy outlined in the preface to Vol. 1 (H ERTEL and S CHULZ 2014) are equally valid
for the present Vol. 2. In the following we just give a guide through the contents, as
the readers will find many topics and details far beyond the standard textbooks and
routine teachings on AMO science.
Chapter 1 resumes the discussion of light (comprising in the broadest sense the
whole electromagnetic spectrum) and photons, which are key themes of these two
volumes. The focus is here on lasers (one of the most important tools of modern
AMO physics), Gaussian beams, polarization and nonlinear processes. In the fol-
lowing Chap. 2 the emphasis is on the properties of photons and coherence. In ad-
dition to discussing some basics of quantum optics and its applications, we also
vii
viii Preface
complete now the theory of photon induced transitions (Chap. 4, Vol. 1) and intro-
duce field quantization which allows us to treat spontaneous emission.
After these preparations we are ready to enter into molecular physics and modern
molecular spectroscopy. We start in Chap. 3 with diatomic molecules, the most sim-
ple prototypes, and enhance our view with ‘real’ polyatomic molecules in Chap. 4,
including a brief excursion into the subject of symmetries. While along this way
we have already encountered various comparatively simple examples of molecular
spectroscopy, Chap. 5 leads us deep into a variety of sophisticated modern methods
of spectroscopy. The field is dominated today by laser based methods, but we also
gain some insights e.g. into the possibilities of photoelectron spectroscopy.
Three quite detailed Chaps. 6–8 are devoted to the present status in the physics of
electronic, atomic, molecular and ionic collisions (including collisional ionization)
– a topic of great practical importance and with demanding intellectual challenges,
both from an experimental and theoretical view point. Chapter 9 gives a down to
earth manual for using the density matrix. It also gives a brief look on the theory
of measurement, including a powerful method to analyze radiation patterns from
anisotropically populated mixtures of excited states. Finally, making use of these
concepts, an introduction of the optical B LOCH equations is given in Chap. 10,
which again addresses many exciting facets of quantum optics with interesting ex-
amples.
A possible extension of the two volumes published now is under consideration.
Such a Vol. 3 would approach modern research even more closely and illuminate
some particularly hot and rapidly developing areas such as ultra-cold matter and
quantum gases, ultrafast dynamics, attosecond physics, cluster spectroscopy and
similar themes.
We wish all our readers an exciting and stimulating reading as well as efficient
understanding and successful learning. In several readings, we have tried to produce
text, mathematical formulas and figures as free from errors as possible. Clearly, this
can only be an approximation process. Thus, we encourage our readers to kindly
communicate any critical comments, errors or simply even typos which they may
discover – and to make suggestions for improvements wherever it appears advisable.
We shall correct such errors at the web-site http://staff.mbi-berlin.de/AMO/book-
homepage/ if and as soon as they become known to us.
As further reading we recommend for comparison or for a deeper look into some
specialties the textbooks listed below.
Berlin Adlershof Ingolf V. Hertel
January 2014 Claus-Peter Schulz
Preface ix
References
ATKINS, P. W. and R. S. F RIEDMAN: 2010. Molecular Quantum Mechanics. Oxford: Oxford Uni-
versity Press, 2nd edn.
B ERGMANN, L. and C. S CHAEFER: 1997. Constituents of Matter – Atoms, Molecules, Nuclei and
Particles. Berlin, New York: de Gruyter, 902 pages.
B LUM, K.: 2012. Density Matrix Theory and Applications. Atomic, Optical, and Plasma Physics
64. Berlin, Heidelberg: Springer, 3rd edn., 343 pages.
B ORN, M. and E. W OLF: 2006. Principles of Optics. Cambridge University Press, 7th (expanded)
edn.
B RANSDEN, B. H. and C. J. J OACHAIN: 2003. The Physics of Atoms and Molecules. Prentice Hall
Professional.
B RINK, D. M. and G. R. S ATCHLER: 1994. Angular Momentum. Oxford: Oxford University Press,
3rd edn., 182 pages.
D EMTRÖDER, W.: 2010. Atoms, Molecules and Photons. Berlin, Heidelberg, New York: Springer,
2nd edn.
D EMTRÖDER, W.: 2008a. Laser Spectroscopy, vol. 1: Basic Principles. Berlin, New York:
Springer, 4th edn., 457 pages.
D EMTRÖDER, W.: 2008b. Laser Spectroscopy, vol. 2: Experimental Techniques. Berlin, New York:
Springer, 4th edn., 697 pages.
D RAKE, G. W. F., ed.: 2006. Handbook of Atomic, Molecular and Optical Physics. Heidelberg,
New York: Springer.
E DMONDS, A. R.: 1996. Angular Momentum in Quantum Mechanics. Princeton: Princeton Uni-
versity Press, 154 pages.
H ERTEL, I. V. and C. P. S CHULZ: 2014. Atoms, Molecules and Optical Physics 1; Atoms and
Spectroscopy, vol. 1 of Springer-Textbook. Berlin, Heidelberg: Springer, 1st edn.
L OUDON, R.: 2000. Quantum Theory of Light. Oxford, New York: Oxford University Press, 3rd
edn.
M UKAMEL, S.: 1999. Principles of Nonlinear Optical Spectroscopy. Oxford: Oxford University
Press, 576 pages.
S TEINFELD, J. I.: 2005. Molecules and Radiation – 2nd Edition, An Introduction to Modern Molec-
ular Spectroscopy. Mineola: Dover Edition.
W EISSBLUTH, M.: 1978. Atoms and Molecules. Student Edition. New York, London, Toronto,
Sydney, San Francisco: Academic Press, 713 pages.
Acknowledgements
Over the past years, many colleagues have encouraged and stimulated us to move
forward with this work, and helped with many critical hints and suggestions. Most
importantly, we have received a lot of helpful material and state of the art data for
inclusion in these textbooks.
We would like to thank all those who have in one or the other way contributed
to close a certain gap in the standard textbook literature in this area – that is at
least what we hope to have achieved. Specifically we mention Robert Bittl, Wolf-
gang Demtröder, Melanie Dornhaus, Kai Godehusen, Uwe Griebner, Hartmut Ho-
top, Marsha Lester, John P. Maier, Reinhardt Morgenstern, Hans-Hermann Ritze,
Horst Schmidt-Böcking, Ernst J. Schumacher, Günter Steinmeyer, Joachim Ullrich,
Marc Vrakking und Roland Wester; their contributions are specifically noted in the
respective lists of references.
Of course, there all other sources are documented which we have used for infor-
mations and which have provided data which we have used to generate the figures
in these books.
One of us (IVH) is particularly grateful to the Max Born Institute for provid-
ing the necessary resources (including computer facilities, library access, and office
space etc.) for continuing the work on this book after official retirement. Special
thanks are expressed to the Wilhelm und Else H ERAEUS foundation for sponsoring
a Senior Professorship at Humboldt Universität zu Berlin.
xi
Contents of Volume 2
xiii
xiv Contents of Volume 2
1 Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Overview, History and Magnitudes . . . . . . . . . . . . . . . . 1
1.1.1 Quantum Nature of Matter . . . . . . . . . . . . . . . . . 2
1.1.2 Orders of Magnitude . . . . . . . . . . . . . . . . . . . . 5
1.2 Special Theory of Relativity in a Nutshell . . . . . . . . . . . . . 10
1.2.1 Kinematics and Dynamics . . . . . . . . . . . . . . . . . 10
1.2.2 Time Dilation and LORENTZ Contraction . . . . . . . . . 13
1.3 Some Elementary Statistics and Applications . . . . . . . . . . . 14
1.3.1 Spontaneous Decay and Mean Lifetime . . . . . . . . . . 15
1.3.2 Absorption, LAMBERT-BEER Law . . . . . . . . . . . . 17
1.3.3 Kinetic Gas Theory . . . . . . . . . . . . . . . . . . . . 18
1.3.4 Classical and Quantum Statistics – Fermions and Bosons 20
1.4 The Photon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
1.4.1 Photoelectric Effect and Quantization of Energy . . . . . 26
1.4.2 COMPTON Effect and Momentum of the Photon . . . . . 28
1.4.3 Pair Production . . . . . . . . . . . . . . . . . . . . . . 30
1.4.4 Angular Momentum and Mass of the Photon . . . . . . . 30
1.4.5 Electromagnetic Spectrum . . . . . . . . . . . . . . . . . 31
1.4.6 PLANCK’s Radiation Law . . . . . . . . . . . . . . . . . 31
1.4.7 Solar Radiation on the Earth . . . . . . . . . . . . . . . . 34
1.4.8 Photometry – Luminous Efficiency and Efficacy . . . . . 37
1.4.9 X-Ray Diffraction and Structural Analysis . . . . . . . . 40
1.5 The Four Fundamental Interactions . . . . . . . . . . . . . . . . 43
1.5.1 COULOMB and Gravitational Interaction . . . . . . . . . 44
1.5.2 The Standard Model of Fundamental Interaction . . . . . 46
1.5.3 Hadrons . . . . . . . . . . . . . . . . . . . . . . . . . . 48
1.5.4 The Electron . . . . . . . . . . . . . . . . . . . . . . . . 49
1.6 Particles in Electric and Magnetic Fields . . . . . . . . . . . . . 51
1.6.1 Charge in an Electric Field . . . . . . . . . . . . . . . . 52
1.6.2 Charge in a Magnetic Field . . . . . . . . . . . . . . . . 53
xxiii
xxiv Contents of Volume 1
xxxv
Lasers, Light Beams and Light Pulses
1
Overview
We still use a classical wave description of light, but start in Sect. 1.1 with
a brief introduction into the physics of lasers – certainly the most important
tools of modern optics and spectroscopy. In Sect. 1.2 Gaussian light beams are
explored, and their manipulation and measurement is illustrated. Section 1.3
gives a precise definition of polarization and describes some experimental
tools for its characterization. Wave-packets are discussed in Sect. 1.4, with
focus on short pulses as interesting examples from current research. Sec-
tion 1.5 introduces “correlation functions” and describes methods for deter-
mining short pulse durations. Finally, in Sect. 1.6 we explore some character-
istics of intense laser fields – a topic of great importance in present research –
and thus transcend classical, linear spectroscopy.
1. Today a broad spectral range is covered by laser systems, from the microwave re-
gion into the X-ray region, in many instances combined with excellent tunability
of wavelengths.
2. The stability and monochromaticity achieved today allows one to maintain some
standard light frequencies to a precision of a few Hz (i.e. with an accuracy on the
order of 1015 ) for very long times.
3. Conversely, phase controlled light pulses in the visible spectral range of a few fs
are provided today on a stable and reliable basis for experiments. The shortest
laser pulses available today (typically in the soft X-ray region) have a durations
of several attoseconds (1 as = 10−18 s).
4. A historic overview on peak intensities accessible in the focus of a pulsed laser
beam is given in Fig. 1.1. To put this into perspective, we recall that a 100 W
incandescent light bulb emits about 120 cd, i.e. 0.176 W sr−1 or ca. 2.1 mW cm−2
in a distance of 1 m.1 Note (right hand scale in Fig. 1.1) the extremely high
ponderomotive potentials of free electrons associated with such laser pulses (see
Sect. 8.5 in Vol. 1). With the new large scale facilities (e.g. the European Laser
Institute, ELI) one hopes to reach 1024 W cm−2 in the coming years, and thus to
enter into ultra relativistic optics.
(1/683) W sr−1 . In contrast, a relatively weak tunable dye laser generates 100 mW CW output
power without problems. Even without focussing, this implies much more than the 0.63 W cm−2
needed to saturate the Na D transition – as we shall show in Sect. 10.5.2. The same laser might,
well focussed, provide already 1.3 × 107 W cm−2 . Presently, large scale high intensity short pulsed
lasers facilities achieve about 1020 to 1021 W cm−2 .
1.1 Lasers – A Brief Introduction 3
ELI
relativistic optics
1020 8.5 × 1018: Up = mc 2 @ λ = 800nm 1 MeV
3.51×1016: EH (in H atom at a0)
strong laser fields
1015 chirped pulse amplification
mode synchronization (mode locking) (CPA) 1 eV
3.69 × 1011: x 0 ~ a0 @ λ = 800nm
1010
Q switching
1960 1970 1980 1990 2000 2010
Fig. 1.1 Progress in generating highest intensities in laser pulses since the realization of the first
laser systems. Marked in red are physical phenomena connected to the respective intensities. Black
labels indicate methodological advances. The trend curve (full, black), adapted from the inventor
of “Chirped pulse amplification, CPA” Mourou (S TRICKLAND and M OUROU 1985) had to be
modified somewhat downward to match reality
The fundamental concept for a laser, schematically illustrated in Fig. 1.2, is essen-
tially the same as that for any generator of high frequency electric or magnetic fields.
The three key elements are:
1. the amplifier which generates the oscillation (out of noise initially) and compen-
sates for losses,
2. the resonator which suppresses all unwanted frequencies, and
Fig. 1.2 Scheme of (a) an RF generator and (b) a laser. Key elements are in both cases amplifier,
resonator and feedback mechanism – as well as of course a power source
4 1 Lasers, Light Beams and Light Pulses
3. the feedback which returns most of the amplified signal to the entrance of the
amplifier.
it as a laser resonator. As indicated Fig. 1.2, in the simplest case a FABRY-P ÉROT
resonator consists of two mirror plates positioned parallel at a distance L. These
mirrors are polished with highest precision. One of them has a reflectivity of R1 =
100 % (as good as possible), while the other has a somewhat smaller R2 < 100 %
in order to transmit the fraction (1 − R2 ) of the intensity stored inside the resonator
for external use. Depending on the specific application, the mirrors may be planar
or curved (for details see Sect. 1.1.3). The two key parameters of the resonator are:
1. The free spectral range (FSR) according to (6.9) in Vol. 1 determines the dis-
tance (in frequency) of two maxima of transmission. It is identical to the inverse
turnaround time Tr :2
νFSR = c/2L = 1/Tr . (1.8)
In wavenumbers this corresponds to ν̄FSR = 1/2L. For whole numbers z, called
order of the interference or index of the mode, light of the frequencies
With (1.9) and (1.10) the Q factor (1.4) of the resonator becomes
ν ν
Q= =F = F × z, (1.11)
νr νFSR
with the order of interference z according to (1.9). By the way, the Q factor is iden-
tical to the resolution (6.17), Vol. 1 of the FPI when it is used as a spectrometer, with
F corresponding to an effective number of interfering beams. Finally, the effective
lifetime (1.3) of a photon in the resonator is given by
1 zF zF FL F
τr = = = = = , (1.12)
ωr ω0 2πν0 πc 2πνFSR
2 More precisely, we would have to use here the respective group velocity v
g (see Sect. 8.4.4, Vol. 1)
in the different media instead of the speed of light c in vacuum. To keep things simple we assume,
however, n ≡ 1 for the index of refraction inside the whole resonator.
6 1 Lasers, Light Beams and Light Pulses
with ωr = 2πνr . To give two typical numerical examples: due to its low am-
plification, a He-Ne laser allows only little output coupling (typically R 99 %,
F 300), while for a pulsed excimer laser the contrary is the case (R = 30 %,
F = 2.5). At a resonator length of 1 m this leads to an effective photon lifetime of
τr = 330 ns and 2.6 ns, respectively.
Of course, several additional types of losses have to be taken into account beyond
output coupling. Absorption in the amplifier medium or in optical elements is one of
them. An absorbing medium of length L1 (absorption coefficient μ, see Eq. (4.21)
in Vol. 1) causes losses per resonator turnaround time Tr which are ∝ exp(−2μL1 )
or, recalculated per time t
I /I0 = exp −2μL1 (t/Tr ) = exp −μ(L1 /L)ct = exp[−t/τa ], (1.13)
Of particular importance are losses due to diffraction. We shall see that finally they
are responsible for the spatial profile of the electromagnetic field in the resonator.
They define the radial mode structure of the light.
Up to now we have treated the FABRY-P ÉROT resonator simply as if there were
no lateral boundaries: the field was described by infinitely extended plane waves.
In reality, the mirrors and the amplifier medium have finite diameters: we speak
about an “open resonator”. Even by geometrical optics one may easily visualize,
how light rays move out of the resonator after a few turnaround passages when
they are not normally incident onto the mirrors. In an early, pioneering publication
KOGELNIK and L I (1966) were able to show that the FABRY-P ÉROT resonator with
plane parallel mirrors does not represent an ideal resonator configuration in this
respect. However, a range of “stable” mirror configurations exists for which even
non-paraxial rays remain inside the resonator after multiple reflections from both
end-mirrors. For these
must hold, with the resonator length L, and the radii r1 and r2 giving the curvature
of the end mirrors. This criterion is illustrated schematically in the stability diagram
Fig. 1.3.
From a wave-optical point of view diffraction occurs at all beam limiting aper-
tures. During each round trip the light field is thus confined twice, at the end mirrors.
Its energy content is reduced accordingly. One may visualize this process rolled out
along the optical z-axis as shown in Fig. 1.4.
Mathematically, the steady state in the resonator is obtained by requiring that the
profile of the field in the ρϕ plane (perpendicular to the optical axis),
end mirror ρ ρ ρ
left S0 right S1 left S2 right S3
z
j=0 1 2 3
L L L
Fig. 1.4 Schematic illustration of the formation of a radial beam profile by iterative diffraction at
the beam limiting apertures (e.g. at the end mirrors of the resonator). Sj indicates the beam limiting
areas
8 1 Lasers, Light Beams and Light Pulses
0.01
0.4 0.8 1.2
TEM01* TEM01 TEM02 F = w2 / λ L
Fig. 1.5 (a) Mode structure in a resonator with cylindrical symmetry for the lowest TEM resonator
modes; shown is the intensity profile (very schematic) in a plane perpendicular to the optical z-axis;
(b) diffraction losses for some of these modes as a function of the F RESNEL number F , in a FAB -
RY-P ÉROT resonator with plane parallel mirrors (pp), dashed lines, and for a confocal resonator
(cf), full lines
linear combination of the TEM01 mode shown in the figure and a degenerate one
turned by 90◦ .
Of key interest are the diffraction losses for the different modes. They are deter-
mined by the F RESNEL number
w2
F= (1.15)
λL
of the resonator, with w being the radius of the beam limiting aperture and L again
the length of the resonator. Some examples are shown in Fig. 1.5(b). As expected,
the diffraction losses decrease rapidly with increasing F RESNEL number. As a typi-
cal order of magnitude we estimate for a He-Ne laser F 0.8–3.2 (resonator lengths
L = 50 cm, beam limiting gas discharge tube w = 0.5–1 mm, λ = 632.8 nm). Fig-
ure 1.5(b) shows that at these conditions diffraction losses for a TEM00 mode are
negligible for a confocal resonator. However, as already mentioned, a plane paral-
lel mirror setup is very unfavourable: it is unstable and subject to high diffraction
losses.
We note in particular, that in any case higher modes are subject to significantly
higher losses as compared to the fundamental mode. It is due to this fact that in
almost all laser systems only the fundamental mode TEM00 is active: for all other
modes amplification usually is not sufficiently high to compensate for the losses.
When modelling a specific laser system quantitatively, all losses have to be ac-
counted for – including diffraction losses (effective photon lifetime due to diffrac-
tion τd ) and absorption (effective photon lifetime due to absorption τa ). In to-
tal, all losses due to the different mechanisms multiply according to the scheme
I = I0 exp(−t/τr ) × exp(−t/τa ) × exp(−t/τd ) . . . . The inverse, effective photon
lifetimes thus add and the resulting overall intensity becomes
ωb , i.e. the linewidth of the laser line (at angular frequency ω) is negligible com-
pared to the width of the absorption profile. Then the total intensity I (t) of the
laser radiation is relevant for the induced transition probability. According to (5.10),
Vol. 1 the transition rate per atom is
I (t) I (t)
Rba = Bba gL (ω) = B̃(ω) .
c c
Let the population density of the upper laser level be Nb , that of the lower laser
level Na . One defines population inversion as
N = Nb − Na . (1.18)
3A more profound rationalization of these rate equations will be presented in Sect. 10.5.
4 We recall: The dimension of I is Enrg T−1 L−2 , that of ω is Enrg, for c it is L T−1 so that Nω
has the dimension (number of photons) L−3 .
10 1 Lasers, Light Beams and Light Pulses
gb λ 2 c N
α(ω) = . (1.21)
ga 2π 1 + (2ω/A)2
Up to now we have tacitly assumed that the linewidth in the upper laser level is
ωb = A, i.e. is determined exclusively by radiative decay. If this is not the case, the
line profile would have to be replaced correspondingly, e.g in the case of collision
broadening by
gb λ2 c A N
α(ω) = = cσba (ω)N, (1.22)
ga 2π ωb 1 + (2ω/ωb )2
with ωb given by (5.18), Vol. 1. We have introduced here a cross section for ab-
sorption σba (ω) which has a maximum value
gb λ 2 A
σba (ωba ) = . (1.23)
ga 2π ωb
For a purely radiative linewidth ωb = A and with gb /ga = 3 (p ← s transition)
this maximum is 0.477λ2 . In the case of inhomogeneous line broadening (i.e. if
one has to average over atoms with different absorption frequencies) the L ORENTZ
profile in (1.22) has to be replaced correspondingly, e.g. by a G AUSS distribution
(5.21), Vol. 1 in the case of D OPPLER broadening.
We now also account for the fact that the light intensity will in general not be
purely monochromatic. Clearly, (1.20) also holds for the intensity distribution I˜(ω)
in (angular) frequency space. Thus, if we just consider the amplifier medium with-
out accounting for losses, I˜(ω) will change when passing through the amplifier
according to
I˜(ω) = I˜0 (ω) exp α(ω)z/c = I˜0 (ω) exp σba (ω)Nz , (1.24)
where I˜0 (ω) is the distribution at the entrance of the amplifier. As gain profile G(ω)
of an amplifier medium in a resonator one defines the relative change after one full
round trip:
I˜(ω)
G(ω) = = exp α(ω)2L/c = exp σba (ω)N 2L . (1.25)
I˜0 (ω)
1.1 Lasers – A Brief Introduction 11
Remarkable about this formula is the fact that a highly amplified intensity pro-
file may look rather different from the line profile (1.22): in the centre of the line
(where amplification is high) the exponential factor leads to particularly high inten-
sity, while in the line wings amplification is very small. This leads to a significant
narrowing of the line, so called gain-narrowing as indicated in Fig. 1.6. Consid-
ering that in an active laser system the gain profile is in addition convoluted with
the resonator profile (which deliberately induces losses for unwanted frequencies)
one understands why the spectral distribution of the radiation generated in a laser
is much narrower than both, the profiles of the amplifying medium or that of the
resonator.
So far we have not mentioned the role of spontaneous emission for the laser
process. It always occurs as a byproduct. In an amplifier medium it will also be
amplified. This may – especially in long stretched geometries with high amplifica-
tion – lead already in one passage (i.e. even without mirrors) to substantial light
intensities. This amplified spontaneous emission (ASE) is often used to provide an
intensive light source with properties similar to that of laser radiation. Prominent
examples are the well known nitrogen-laser, as well as many of the so called X-ray
and free electron lasers (see Sect. 10.6.5 in Vol. 1). ASE may, however, be very
troublesome, e.g. when intense, short laser pulses are to be generated. Typically,
one first tries to build up sufficient population inversion for amplification, before in
a second step the laser action is triggered. During this building-up process ASE may
already lead to premature destruction of population inversion and formation of an
unwanted background signal.
It is now straight forward to derive criteria for the possibility of laser action. To
this end one has to account for amplification, according to (1.20), as well as for the
losses, according to (1.16), and obtains for the intensity in the laser resonator:
dI I I
= α(ω)I − = cσba (ω)NI − . (1.26)
dt τe τe
For simplicity we assume that the amplifier medium fills the whole resonator.
12 1 Lasers, Light Beams and Light Pulses
Amplification after one full turnaround is obtained only if dI /dt ≥ 0. From this
follows the threshold condition for laser action:
1
N = . (1.27)
σba (ω)cτe
We remember that the effective photon lifetime τe may contain several contributions
according to (1.16), and cτe is something like an ‘effective path length’ for a photon.
In the volume σba cτe at least one atom has to be in the upper state for laser action
to occur.
Note that the actual intensity I of the laser field in the resonator has dropped
out of (1.27). Clearly, under active laser operation I will be finite and depend on
the details of the operating conditions as we shall discuss in a moment. However,
for stationary laser operation we must always demand dI /dt = 0. Hence, when the
laser is active, (1.27) will also be satisfied. The important message is thus:
Population inversion under stationary state lasing conditions
is identical to threshold inversion.
Strictly speaking, one also has to consider the increase of the radiation field due
to spontaneous emission. Since that is emitted, however, into the full solid angle
and with the full bandwidth ωb , only a negligible fraction falls into the active
laser mode. Spontaneous emission thus plays a role only in the starting phase of
laser action, and possibly as noise (see, however, the above discussion on ASE).
In a two level system population inversion may only be reached for short times, and
stationary laser operation is not possible with such a scheme. Hence, in efficient
laser media typically three or four levels are involved in the overall process.
Rate equations for a laser process have to account for all losses and gains. They
must describe the temporal evolution of the population densities for all levels in-
volved as well as for the radiation intensity I (t) in the resonator according to (1.26).
A schematic is shown in Fig. 1.7, and the overall balance for upper (b) and lower
(a) laser level reads:
dNb σba (ω)I
= Pb − N − ANb − γb Nb (1.28)
dt ω
dNa σba (ω)I
= Pa + N + ANb − γa Na . (1.29)
dt ω
1.1 Lasers – A Brief Introduction 13
pump
Na
σba I
γa >>
ħω
These rate equations take into account that upper and lower laser levels may in
principle be populated with rates Pb and Pa , respectively, from other levels, and
decay in turn with rates γb and γa into other levels. Note the dimensions are L−3
for Na,b , T−1 L−3 for Pa,b , Enrg L−2 T−1 , L2 for σba , Enrg for ω, and T−1 for γa,b .
Ideally, in the stationary state one would like to have Na = 0 and N = Nb or at
least Na
Nb . For this to happen, one tries to achieve Pa = 0 and γa
σba I /ω,
so that the lower laser level is rapidly depleted during laser action. The ruby laser,
the first laser actually realized by M AIMAN (1960), was a three level laser.
More convenient, and realized in most of today’s laser systems, are four (or more)
level laser schemes which allow to establish nearly ideal conditions. Such a scheme
is sketched in Fig. 1.8. Highly simplified one may assume that γb = 0 and γa is
indeed very large,5 so that Na 0 and N Nb . With A = 1/Tb and Pb = P one
obtains from (1.28) and (1.29):
dN σba (ω)I
=P − N − N/Tb . (1.30)
dt ω
For stationary conditions dN/dt = 0 must hold, and N is given by the threshold
inversion (1.27). Inserting this into (1.30) one obtains for the laser intensity
ω
I = P ωcτe − . (1.31)
σba Tb
Hence, with rising pump rate the intensity in the laser resonator rises – as one would
expect – even though the population inversion remains constant (the excess energy
is essentially converted into useful output). If the resonator is tuned to the maximum
of the line profile, with (1.23) the intensity becomes
4π 2 cωb
I = P ωcτe − , (1.32)
3λ3
where we have assumed that the relaxation time Tb of the upper state is determined
exclusively by spontaneous radiative decay (ATb = 1). Thus, the laser intensity I –
as far as externally controllable – depends on the pump rate P and on the resonator
5 Interestingly, in a He-Ne laser the lower laser level is depopulated by collisions with the walls
surrounding the gas discharge. Hence, the diameter of the plasma tube must not become too large.
14 1 Lasers, Light Beams and Light Pulses
losses (via τe ), while the negative term with the effective linewidth ωb of the
excited level is system specific. We note that the loss term is ∝ λ−3 ∝ ν 3 , and point
out again that this explains why building a laser becomes increasingly difficult as
the wavelength decreases.
For laser operation high above threshold one may neglect the loss term in (1.32),
and the intensity becomes I P ωcτe . It now looks as if one just has to increase
τe to obtain more intensity – i.e. one may think about increasing the reflectivity R
of the mirrors in the resonators. Albeit this holds within the resonator, and is even
used in various spectroscopic applications (see e.g. Sect. 5.5.3), for the laser output
intensity Iout we may have to reconsider such strategy: only a fraction (1 − R)
is coupled out; and it is the very fact that power is coupled out which leads to a
reduction of photon lifetime. With (1.16), (1.12) and (1.10) one estimates:
√
Iout P ωcτe (1 − R) < P ωcτr (1 − R) = P ω RL. (1.33)
The output intensity depends of course on the population rate P for the upper laser
level and on the length L of the resonator (here assumed equal to the length of
the amplifying medium). And interestingly enough, by increasing the reflectivity of
the output mirror one indeed gains some output power – however, this only holds
as long as other losses (such as diffraction or absorption) can be neglected, i.e. if
1/τe 1/τr (see Eq. (1.16)).
Figure 1.9 summarizes what we have learned so far about amplification, line pro-
files and mode structure in active and passive laser resonators, and combines it with
the concept of homogeneous and inhomogeneous line broadening introduced in
Sect. 5.1, Vol. 1. We emphasize the specificities of homogeneous vs. inhomoge-
neous gain profiles in respect of laser operation. We also recall hole burning in a
D OPPLER profile which was introduced in Sect. 6.1.5, Vol. 1.
Figure 1.9(a) shows a typical line profile as a function of angular frequency ω
of an amplifying medium (FWHM = ωb ) – still without laser activity. The dash
dotted red line marks the threshold amplification necessary for laser action. The po-
sitions of the resonance frequencies (vertical dashed grey lines) and the free spectral
range ωFSR of the laser resonator are indicated. Two resonance frequencies are
emphasized in red: at these positions amplification is above threshold and lasing is
possible. Complementary to this, Fig. 1.9(b) shows the transmission of the passive
resonator. The bandwidth ωr of these transmission lines is a function of the finesse
F of the resonator according to (1.12).
Figure 1.9(c) illustrates the situation for the active laser in the case of homoge-
neous line broadening (e.g. natural linewidth, pressure broadening etc.). Since all
atoms or molecules contribute to the amplification with the same spectral profile,
and since the population inversion under operating conditions is equal to the thresh-
old inversion, the whole amplification profile is attenuated such that the operating
1.1 Lasers – A Brief Introduction 15
transmission
resonator modes passive
threshold amplification profil
passive Δ ωr
Δω b
ω
Δω FSR
Δω FSR ω
threshold A
threshold operating point
Δω b
amplification profile
active
Δω FSR ω ω
Δω FSR
I (ω) (d) resonator modes active I (ω) (f ) resonator modes active
ω ω
Fig. 1.9 Amplification profiles ∝ σba (ω)N and longitudinal modes in passive and active res-
onators for homogeneous and inhomogeneous line profiles; for details see text
in the case of inhomogeneous line broadening, typically many resonator modes may
be amplified. But also for homogeneous line profiles, in general several resonator
modes are in principle ready for laser activity and will compete with each other –
a situation which may lead to instable laser operation.
It is now obvious that a simple laser setup as sketched in Fig. 1.2(b) on p. 3
will only in exceptional cases lead to strictly monochromatic radiation. As a rule,
a rather broad mixture of lines from many longitudinal modes will emerge. They
compete for population inversion and may strongly fluctuate. Thus, one needs addi-
tional means to achieve stability and to limit the bandwidth, in order to finally obtain
a monochromatic, coherent and parallel output corresponding to the ideal of laser
radiation. For further details the reader is referred to the specialized literature.
Section summary
• The key elements of a laser are (i) an optical amplifier – based on population
inversion, (ii) a resonator to suppress all unwanted frequencies – typically a
FABRY-P ÉROT, and (iii) feedback which returns a major fraction of the am-
plified signal back to the entrance of the amplifier – which is achieved by the
resonator mirrors.
• Q factor (a large number), angular frequency ω, and average photon lifetime
τr in the resonator are related by τr = Q/ω. The bandwidth of the passive
resonator is ωr = 1/τr = ω/Q. The linewidth of an active laser mode is
typically much narrower due to gain narrowing.
• A FABRY-P ÉROT is characterized (i) by its free spectral range (or the
turnaround time for light Tr )
• Resonator stability depends on the curvature and distance of the mirrors. Sub-
stantial losses for different (transverse) TEMn modes may arise from diffrac-
tion at the mirrors which act as finite apertures of radius w in the resonator.
These losses are smallest for the TEM00 mode and increase with the F RESNEL
number F = w 2 /λL.
• Laser threshold is reached when the gain compensates the losses
−1
N = σba (ω)cτe ,
with the inversion density N and the absorption cross section σba (ω). This
threshold population inversion is also maintained during active laser action:
additional pump energy is converted into laser output.
• Homogeneous and inhomogeneous line broadening leads to very different
longitudinal mode patterns. Of particular interest is the hole burning which
occurs in inhomogeneous amplifier media.
1.2 Gaussian Beams 17
We now have to focus in some more detail onto the spatial properties of the light
which is generated by a laser. Typically, one obtains light which inside the resonator
corresponds more or less to the (desired) TEM00 fundamental mode. When such
light propagates freely in space, it becomes what is known as a Gaussian beam. The
reader may find some of the following text and formulas known from Vol. 1 or else-
where, or even trivial. We need, however, a solid basis to avoid misunderstandings
later on.
We start with the spatial profile of a Gaussian beam, and specify some termi-
nology such as beam radius, beam divergence and complex beam parameter. We
then introduce the so called ABCD matrices, important, simple tools for the whole
of laser physics. We discuss focussing and widening of laser beams – and how to
measure the width. Finally, we shall introduce the so called M 2 factor as a practical
measure of beam quality.
Naively, one may imagine a light beam as a (narrow) cone of small but finite di-
vergence angle θe , which is filled with a quasi-monochromatic plane wave of elec-
tromagnetic radiation at an angular frequency ω and wavelength λ (wave vector
|k| = ω/c = 2π/λ). This would imply a constant amplitude over the whole cross
section, dropping to zero outside the cone. But even if the eye may perceive a laser
beam as something like that, such a sharply limited energy distribution is not com-
patible with the general wave equation:6
1 ∂2
− 2 2 E(x, y, z, t) = 0. (1.34)
c ∂t
A trivial solution is of course the well known plane wave (4.1), Vol. 1. We have
used it throughout Vol. 1 when describing the interaction of electromagnetic radia-
tion with atoms. However, such a wave field extends laterally to infinity, and if one
tries to cut out of it a light beam – e.g. by a circular aperture of radius w0 , the cut out
‘beam’ forms immediately a characteristic diffraction pattern as we have seen it in
the laser resonator (Fig. 1.4) and a divergent bundle of light with smooth boundaries
emerges: Diffraction simply washes out all sharp contrast.
6 For simplicity we only consider wave propagation at the speed of light c, i.e. in vacuum.
18 1 Lasers, Light Beams and Light Pulses
(slowly) on position in space and time, and an arbitrary phase φ0 relative to the
envelope. We thus generalize (4.1), Vol. 1:7
i ∗
E(r, t) = E0 (r, t)ei(kr−ωt+φ0 ) e − E0 (r, t)e−i(kr−ωt+φ0 ) e∗ . (1.35)
2
Somewhat more compact we write
i −
E(r, t) = E (r, t)e − E + (r, t)e∗
2 (1.36)
∗
with E + (r, t) = E0 (r, t)e−i(kr−ωt+φ0 ) = E − (r, t) .
∂ 2 E0 ∂ 2 E0 ∂ 2 E0 ∂E0
+ + −2ik = 0. (1.37)
∂x 2 ∂y 2 ∂z 2
∂z
:= 0 in SVE approximation
so that the overall character of the plane wave does not get lost. Specifically, for
paraxial propagation the change of the amplitude in z direction is assumed to be
particularly small, and one neglects the second derivative in respect of z completely,
as indicated in (1.37). This differential equation can then be solved by the ansatz
x2 + y2
E0 (x, y, z) = E0 exp −iP (z) exp −ik . (1.39)
2q(z)
7 Note that this is identical to (4.1), Vol. 1 for φ0 = 0 and E0 (r, t) = E0 = const. For linearly
polarized light, e = e∗ , we have E(r, t) = eE0 (r, t) sin(kr − ωt).
1.2 Gaussian Beams 19
This is already
written in cylinder symmetrical form, i.e. laterally the field depends
only on ρ = x 2 + y 2 , the distance from the optical axis. After some mathematical
manipulations this leads to rather simple ordinary differential equations for P (z) and
q(z). Without entering into details we communicate the key results: With suitable
boundary conditions one obtains P = i ln(1+z/izR ) and the so called complex beam
parameter
q(z) = z + izR . (1.40)
It characterizes the Gaussian beam by just one real parameter, the R AYLEIGH length
zR , and may be rewritten as
1 1 i 1 λ
= − = −i . (1.41)
q z(1 + (zR /z)2 ) zR (1 + (z/zR )2 ) R(z) πw 2 (z)
This suggests to describe the beam profile locally by two variables, the
radius of curvature R(z) = z 1 + (zR /z)2 , (1.42)
(which gives the surfaces of constant phase in the far field z
zR ) and the
beam radius w(z) = w0 1 + (z/zR )2 . (1.43)
πw02 kw02
zR = = . (1.45)
λ 2
Thus, a Gaussian beam may also be characterized by w0 rather than by zR . With
these definitions we can now rewrite (1.39) and define the field amplitude for a
Gaussian beam:
E0 ρ2 ikρ 2
E0 (ρ, z) = × exp − 2
× exp − × exp iφG (z) ,
1 + (z/zR )2 w(z) 2R(z)
(1.46)
and specifically for z = 0: E0 (ρ) = E0 exp −(ρ/w0 )2 . (1.47)
The maximum field amplitude E0 / 1 + (z/zR )2 on the beam axis obviously
depends on z. Radially the field drops to 1/e (36.8 %) of its maximum value at
a distance ρ = w from the axis (beam radius). At the beam waist (z = 0, w(0) =
w0 ) the radius of curvature is R(0) = ∞ (plane wave). The maximum curvature
(minimum |R|) is found at z = ±zR with 1/|R| = 1/2zR ; the beam radius there is
20 1 Lasers, Light Beams and Light Pulses
√
w(zR ) = 2w0 . For larger |z| the radius of curvature increases again, approaching
R(z) → z.
The first factor in (1.46) ensures conservation of total energy in the beam, the sec-
ond describes the actual radial G AUSS profile. The third factor exp[−ikρ 2 /2R(z)]
contains a phase in the exponent and essentially quantifies the deviation of a Gaus-
sian beam from a plane wave. It is also called F RESNEL factor.8 In the far field it
approaches unity and the Gaussian beam practically becomes a plane wave. The last
factor in (1.46) is also interesting; it contains the so called G OUY phase:
This is a phase which the wave ‘collects’ in comparison to a plane wave due to the
fact that the curvature of the wavefront changes. The G OUY phase changes from
π/2 to −π/2 when z runs from −∞ to ∞ and is defined so that the overall phase
vanishes for z = 0. Exactly 50 % of the overall phase change are collected between
z = ±zR . Recently, the G OUY phase gains increasing importance in the context of
ultrafast laser pulses and has been determined experimentally for the first time by
L INDNER et al. (2004).
Figure 1.10 illustrates this profile graphically. The red line in Fig. 1.10(a) indi-
cates where the intensity I (ρ, z) has dropped to 1/e2 and defines asymptotically the
beam divergence θe . The 3D plot in Fig. 1.10(b), a blow up of the centre part in (a),
8 We may write the exponent also ikρ 2 /2R = iπ × ρ 2 /λR and recognize the F RESNEL number
according to (1.15).
1.2 Gaussian Beams 21
Fig. 1.10 Profile and parameters in a strongly focussed Gaussian beam w0 = 2λ, zR = 2πw0 .
(a) Geometry in the vicinity of the beam waist w0 , (b) 3D plot of the intensity profile around the
focus, (c) radius of the wavefront as a function of z/w0 . Note the different scales for the ρ and
z-axes
shows some kind of a ‘dog bone’ structure. Indicated is also the confocal√param-
eter b = 2zR . Note that with (1.43) at z = ±zR the beam radius is w = 2w0 so
that the beam area πw 2 = 2πw02 has doubled. Thus, the intensity decreases within
the confocal parameter in both directions to 50 % of its maximum. Figure 1.10(c)
shows R, the radius of curvature of the wave front, as a function of z/w0 . Notice
the pronounced minima of |R| = 2zR at z = ±zR .
w0 2|z| λ|z|
lim w(z) = |z| = = . (1.52)
|z|→∞ zR kw0 πw0
with [P (θ )] = W/sr. Thus, the divergence angle of the Gaussian beam is given by
2 w0 λ
θe = = = . (1.55)
kw0 zR πw0
There, the intensity has dropped to 1/e2 of its maximum, the field strength to 1/e.
The corresponding solid angle divergence is
λ2
δΩe = πθe2 = . (1.56)
πw02
at which the intensity has dropped to 1/e, which we shall call G AUSS radius. Its
minimal value at the waist is related to the R AYLEIGH width by
a(0) = λzR /2π = zR /k,
The maximum intensity on the beam axis in terms of the total power Ptot in the
beam is obtained from (1.50):
Ptot 2Ptot Ptot
I0 (z) = = = 0.693 . (1.60)
πa 2 πw 2 π(d1/2 /2)2
This is easy to memorize for the G AUSS radius: the maximum intensity of a Gaus-
sian beam is equal to the intensity in a hypothetical, cylindrical rod with radius a
(assumed as constant).
Often one is interested in the fraction P (ρ) of the total power Ptot which is con-
tained in the centre of the beam between ρ = 0 and ρ. By integrating (1.58) one
obtains the useful formula
I (ρ)
P (ρ) = Ptot 1 − (1.61)
I0
and the average intensity
2
w2 ρ
Iav = I0 1 − exp −2 . (1.62)
2ρ 2 w
1.2 Gaussian Beams 23
I / I0 ln( I / I0)
0
(a) (b ) d1/2
-1
a
d1/2 -2
w
-3
a
-4
w
-5
-2 0 2 -2 0 2
ρ /mm ρ / mm
Fig. 1.11 Radial intensity profile of a continuous dye ring laser. Experimentally determined points
(•) and G AUSS fit ( ) plotted (a) on a linear and (b) on a logarithmic intensity scale. The quantities
FWHM = d1/2 , G AUSS radius a, and beam radius w are marked to scale
dρ z
dφ
diffracting surface S
nar object with a surface S as superposition of spherical waves.9 The field detected
at a point r D at the time t = 0 is
iES+ exp(ikr)
E + (r D , t = 0) = T (ρ, ϕ) ρdρdϕ. (1.63)
2λ S r
The geometry is sketched in Fig. 1.12. We choose cylinder coordinates z, ρ and ϕ
for the computation. The object is described by a transmission function T (ρ, ϕ) –
in the most general case complex – which may modify phase and amplitude ES+ of
the incoming wave. One reads the distance r between a point on S – characterized
by ρ and ϕ – to the detection point D – characterized by s and ψ – from Fig. 1.12:
With the scalar product of the two radial vectors ρ on the diffracting surface, and
s on the detector plane one writes xξ + yη = ρ · s = ρs cos(ψ − ϕ). We replace
ξ 2 + η2 = ρ 2 and expand r for w0
r0
xξ + yη ρ 2 ρs cos(ψ − ϕ) ρ2
r = r0 1 − 2 + r0 − + . (1.64)
r02 r02 r0 2r0
For small diffraction angles θ = s/r0 holds, so that finally the spherical wave in
(1.63) may be written:
exp(ikr) exp(ikr0 ) ρ2
→ exp −ikρθ cos(ψ − ϕ) exp −ik .
r r0 2r0
9 We refrain here from presenting the full derivation according to K IRCHHOFF, and assume for
simplicity the wave to encounter the object perpendicularly.
1.2 Gaussian Beams 25
The last factor is the F RESNEL factor already known from the field amplitude (1.46)
of the Gaussian beam. Again the F RESNEL number F , (1.15), comes into play;
depending on the characteristic dimensions ( w0 ) of the diffracting surface and
the distance (r0 z) to the detector one distinguishes
One has to evaluate this integral for the given geometry and transmission function
T , and obtains the intensity of the diffraction pattern ∝ |E + (r D )|2 . The simplest
case is a circular aperture of radius w0 with T (ρ, ϕ) ≡ 1. One makes use of the
properties of the B ESSEL functions of 1st order:
2π x
1
Jn (u) = e iu cos ϕ inϕ
e dϕ and xJ 1 (x) = uJ0 (u)du. (1.66)
2πin 0 0
i kw 2 J1 (kw0 θ )
= ES+ exp(ikz) 0 .
2 z kw0 θ
The power diffracted per solid angle finally becomes
(kw0 )2 2J1 (kw0 θ ) 2
P (θ ) = Ptot . (1.68)
4π kw0 θ
In Fig. 1.13 this is compared with a Gaussian profile (1.54). Even though the di-
vergence is of similar magnitude in both cases, the Gaussian beam is clearly better
collimated.
A Gaussian beam is the best possible approximation to the hypothetical light
ray in geometrical optics.
Its divergence angle θe is related to its minimal radius w0 (at intensity I = I0 /e2 )
and its wave vector k or wavelength λ by (1.55). This may, somewhat sloppily, be
interpreted as a kind of uncertainty relation: for the transverse photon momentum
26 1 Lasers, Light Beams and Light Pulses
-6 -4 -2 0 2 4 6
kw0 θ
k⊥ = kθe one finds k⊥ w0 ≥ 2. Generally, one defines a so called beam param-
eter product BPP ≡ w0 θe as a measure for the quality of a light beam (specifically
for lasers). For a Gaussian beam we have
while for the aperture limited plane wave according to (1.68) the divergence angle
λ zλ
w0 θe ≥ 3.83/k = 1.22 and sAiry = θe z = 1.22 (1.70)
2 2w0
holds. The characteristic quantity here is the radius sAiry of the central diffraction
pattern (the so called A IRY disc) on the detector screen.10 We shall come back to
these expressions again in Sect. 2.1.7 in the context of lateral coherence of natural
light and laser radiation.
Before we can efficiently treat the manipulation of laser beams (focussing, de-
focussing, deflecting etc.) we have to introduce an important tool of geometrical
optics, applicable also to Gaussian beams: the so called ray transfer matrices or
ABCD matrices. They are used for ray tracing of paraxial light rays (with sin θ ∼
= θ ),
i.e. they allow to describe in a simple manner the propagation of light rays through
an optical arrangement. These ABCD matrices are efficient tools e.g. for designing
laser systems and for computing the properties of laser resonators – typically de-
scribing the propagation of well collimated Gaussian beams through a number of
optical elements such as mirrors, lenses, and prisms.
10 Note that this formula is the basis of the well known R AYLEIGH criterium for the diffraction
limited resolving power of optical instruments. It states that sAiry is the smallest distance between
two objects that can be resolved by an instrument with an effective limiting aperture of diameter
D = 2w0 and focal length f = z.
1.2 Gaussian Beams 27
A paraxial light ray that propagates at a (small) angle θ in a distance ρ from the
optical axis, is described by a ray vector:
ray
ρ
corresponding to the scheme
θ optical axis z
In Fig. 1.14 the change of a ray into ray is illustrated for two important cases:
translation and refraction. Such a change upon transition of the ray through the
optical setup corresponds to a linear transformation of the ray vector,11 described
by an ABCD matrix:
ρ A B ρ
= . (1.71)
θ C D θ
Specifically, for a Gaussian beam the complex beam parameter (1.41) changes under
such transformation from q to q (here without proof; see e.g. KOGELNIK and L I
1966):
Aq + B 1 C + D/q
q = or = . (1.72)
Cq + D q A + B/q
The most important simple cases are collected in Table 1.2. One verifies
ρ 1 d ρ ρ + θd
that case 1 = =
θ 0 1 θ θ
describes a translation,
ρ 1 0 ρ ρ
while case 2 = = nθ
θ 0 n
n θ n
represents the S NELLIUS diffraction law for small angles, n θ = nθ (paraxial rays).
Note that in the latter case ρ = ρ does not change.
The propagation of a ray through several optical devices is described by the prod-
uct of the respective ABCD matrices. For example, the ray vector after translation
and refraction by a thin lens is derived by multiplying matrix 1 with matrix 3 from
11 Ina mathematical sense the ray vector is simply the position vector expressed by ρ, the distance
from the optical axis (z-axis), and θ = ρ/z, the polar angle. The pictographs used in the following
are not always correct in that sense, but quite instructive.
28 1 Lasers, Light Beams and Light Pulses
3 thin lens of focal length f 1 0
− f1 1 f f
4 reflection from a concave mirror 1 0
− R2 1
R
5 refraction at a spherical surface 1 0
n
− nn−n
R
n n′
n R
Table 1.2:
1 0 1 d ρ 1 d ρ ρ + θd
= = . (1.73)
− f1 1 0 1 θ − 1
f 1− d
f θ θ − ρ+θd
f
One verifies e.g. for any ray coming out of the focal point (d = f and ρ = 0), that
after passing the lens the exit angle is always θ = 0, i.e. all such rays leave the lens
parallel – as well known from geometrical optics. Conversely, a ray entering the lens
parallel to the optical axis (θ = 0) always leaves the lens at ρ = ρ with θ = −ρ/f ,
independently of d, i.e. the ray is refracted into the focal point.
We may also describe ray propagation behind the lens to a position d if we know
the ray vector in front of the lens:
1 d 1 0 ρ 1 − df d ρ ρ − ρd
f +d θ
= = . (1.74)
0 1 − f1 1 θ −1 1 θ −ρ +θ
f f
One verifies e.g. that rays which are parallel to the axis before entering the lens
(θ = 0) cross the axis behind the lens at d = f .
Finally, the product of matrix 1 with matrix 3 and again matrix 1 allows one to
describe imaging by a lens. Further multiplication with the matrices for additional
optical pathways and other elements allows one to easily describe even complex
optical setups very efficiently. We shall now apply this technique to the manipulation
of a Gaussian beam with lenses.
1.2 Gaussian Beams 29
1 λ i
= −i 2
=− . (1.75)
q πwL zR
It changes with distance travelled according to (1.72). With the propagation matrix 1
in Table 1.2 we find at a distance z = d
1 C + D/q 1/q
= = ,
q A + B/q 1 + z /q
1 1 i 1 λ
= − = −i .
q 2 2
z (1 + zR /z ) zR (1 + z /zR ) R
2 2 πw 2
Thus we find R = z (1 + zR2 /z2 ) and w = wL 1 + z2 /zR2 , expressions which are
completely equivalent to (1.42) and (1.43). We see that the matrix method repro-
duces indeed the behaviour of a Gaussian beam as it propagates along the optical
axis – including the correct divergence angle θe = λ/πwL corresponding to (1.52).
We now want to focus this laser beam by a collecting lens of focal length f . Let
the lens be positioned at z = 0, and we state now more precisely that the assumed
initial parallelism implies RL
f . In geometrical optics such a “plane wave” is
assumed to be focused in the focal point of the lens, whereby the initial plane wave
front (RL = ∞) is converted into a spherical wave with a radius of curvature R =
(z − f ). In reality, however, due to diffraction the focal point becomes a circular
disc of radius w0 , as sketched in Fig. 1.15.
The complex beam parameter in front of the lens is given by (1.75) as in the
previous example. According to (1.72) we use this time the product matrix (1.74) to
obtain the new complex beam parameter at a distance z behind the lens:
Without going into further details we concentrate onto the focal point of the lens,
z = f , in order to determine the waist w0 = w (f ).12 The result is simply
1 1 πw 2 1 λ
= − i 2L = − i
q f λf R πw02
with the latter equality again according to (1.41). Comparing real and imaginary
parts we obtain the waist radius w0 of the focused beam
fλ
w0 = , (1.77)
πwL
while the radius of curvature of the wave front becomes R (f ) = f . For practical
√
purposes we also note the relation with the 1/e intensity G AUSS radius a = w/ 2
fλ
a0 = , (1.78)
2πaL
and recall that the maximum intensity (1.60) in the beam centre which may be ob-
tained by this kind of focussing is I0 = Ptot /πa02 .
Inserting (1.77) into (1.45) gives for the focused beam the R AYLEIGH length zR
and the confocal parameter (the distance around the focus for which the intensity
on axis is above 50 % of the maximum):
2πw02 2λf 2
b = 2zR = = .
λ πwL2
Table 1.3 presents numerical examples for three different focal lengths f and two
different initial beam diameters wL , all at the 800 nm wavelength of the Titanium-
sapphire laser – the workhorse of ultrafast spectroscopy.
12 A closer look into the algebra shows, however, that the smallest cross section is found at z =
f z02 /(zR2 + f 2 ), i.e. slightly in front of the focal point of the lens. There the radius of curvature
even approaches R → ∞. Since typically we start with a relatively large beam radius wL the
R AYLEIGH length is very large, zR = πwL2 /λ
f , so that the small difference does not play a
practical role.
1.2 Gaussian Beams 31
Conversely, one may parallelize a diverging laser beam by putting a lens of focal
lengths f at a distance f behind its waist. As illustrated in Fig. 1.16 the lens converts
the curved wave front into an essentially plane wave. Starting with a beam waist w0
at z = 0 and a radius of curvature RL = −f the complex beam parameter at the
waist is
1 1 λ
=− −i .
q f πw02
We use again (1.72), this time with the product matrix (1.73). After a brief calcula-
tion we obtain the new complex beam parameter directly behind the lens:
1 C + D/q π
= = −i 2 w02 .
q A + B/q f λ
Beam radius and radius of curvature immediately behind the lens are thus
fλ
w0 = and R = ∞, respectively.
πw0
As expected, this is just the inverse of the focussing process according to (1.77).
We have to be aware that the ‘new’ laser beam thus generated has now its waist
just behind the lens. From there it will propagate in a slightly divergent manner.
According to (1.55) the new divergence angle of the ‘parallelized’ Gaussian beam
w'0 w'(z' )
z=0
θe
Laser beam
z'
close to its waist w0
θ'e
R= - f
θ'e
f f
R' = ∞
Table 1.4 Beam radii of Gaussian beams as a function from the beam waist for different initial
waist radii w0 according to (1.43). The R AYLEIGH length zR and the far field divergence angle
θe are related to w0 by (1.45) and (1.55), respectively. The numbers are given for the Titanium-
sapphire laser wavelength λ = 800 nm
Nr. Beam parameter w/mm Beam radius at 1/e2 Intensity at z
(distance to waist)
w0 /mm zR /mm θe /mrad 1m 2m 20 m 100 m 1000 m
1 10 400 × 103 2.5 × 10−2 10 10 10 10.3 27
2 3 35 × 103 3 × 10−2 3 3 3.4 9 85
3 1 3.9 × 103 0.25 1 1.1 5.2 25 255
4 0.3 353 0.84 0.9 1.7 16 85 850
5 0.1 39 2.5 2.5 5 51 254 2546
6 0.03 3.5 8.5 8.5 16 170 850 8500
7 0.01 0.4 25 25 50 250 2550 25 465
is
λ λπw0 w0
θe = = = . (1.79)
πw0 πf λ f
This divergence angle may even be interpreted geometrically as indicated in
Fig. 1.16: it corresponds to the divergence one would expect from classical ray op-
tics imaging an extended light source with a radius w0 to infinity.
To give a feeling for the divergence of Gaussian beams, in Table 1.4 numerical
examples are collected for different initial beam radii w0 (equivalently R AYLEIGH
lengths zR or divergence angles θe ). The beam radii w at different distances z from
the waist illustrate clearly how the beams diverge – one may often observe this by
laser illuminations in the nightly skies of big cities (thanks to R AYLEIGH scattering
in dry weather, or even better in humid air due to M IE scattering as discussed in
Sect. 8.4.5, Vol. 1).
These numbers are astonishing indeed, at least at first glance. Compare e.g. beam
1, 3, and 5 in Table 1.4. Beam 1 may initially appear as a rather broad ‘brush’
of light. Note, however, that it widens only by a factor of 3 over the distance of
1km, and it would thus be well suited e.g. in telecommunication or metrology. In
contrast, beams 3 or 5 look on the laser table nicely thin and ‘laser-like’. They
expand, however, very rapidly and get a rather large radius in some distance. Beam
5 will hardly be useable in a neighbouring lab at 20 m distance, while beam 6, finally,
is a highly focussed beam generating very high intensity in its focal point, but with
a R AYLEIGH length of 0.4 mm, its on axis intensity decays rapidly to less than 50 %
outside the confocal length of 0.8 mm around the focus.
For transporting beams over wide distances one first has to expand them. To this
end one uses telescope systems as sketched in Fig. 1.17. In the destination where
the beam is to be used, one may place a second telescope system – more or less
identical to the first one – in order to re-compress the beam. Even if one wants to
1.2 Gaussian Beams 33
lens 1
lens 2
generate a particularly tight focus, it is recommendable to first expand the beam: ac-
cording to (1.77) the minimal radius w0 achievable in the focus of a lens is inversely
proportional to the radius wL of the laser beam in front of the focusing lens.
As illustrated in Fig. 1.17 one may construct telescope systems essentially in two
ways: either according to K EPLER with two collecting lenses, arranged con-focally
or according to G ALILEI with one dispersing and one collecting lens, the focus of
the latter matching the backward focus of the dispersing lens. From the geometry
shown in Fig. 1.17 one reads the beam expansion from radius w1 to w2 in both cases
to be
The divergence angle of the expanded beam is again estimated from (1.55):
λ
θ2 = . (1.81)
πw2
Strictly, this only holds if the radius of the phase surfaces at lens 2 exactly matches
f2 – which may always be achieved by fine adjustment of the lens positions.
Both types of telescope systems have their specific merits. Between two collect-
ing lenses a real focal-point exists, into which one may e.g. place a small, circular
aperture (diameter typically 3 to 5 × w0 ) as indicated in Fig. 1.17. With such a
spatial filter one often may improve the lateral profile of not completely perfect
Gaussian beams considerably. The (lateral and longitudinal) adjustment of such an
aperture requires some experimental skills.
On the other hand, for very high laser intensities – as often encountered when
working with femtosecond laser pulses – one must avoid a real focus where ioniza-
tion and electric breakdown in air may occur, and destroy the laser pulse. In such
cases only the G ALILEI type of telescopes is applicable.
34 1 Lasers, Light Beams and Light Pulses
-2 -1 0 1 2
y0 /a = 1 y0 / a
If one already knows that the radial beam profile is Gaussian, or is least described
well enough by (1.58), one may determine the radii a or w according to (1.59) by
the rather simple, practical “knife-edge” method. One mounts a razor blade onto a
precision optical table, which can be moved perpendicularly into the beam, say in y-
direction. One registers the power P (y0 ) which reaches a detector positioned behind
of y. If the beam is covered from y = y0 to y = ∞ this amounts to
it as a function
(with ρ = x 2 + y 2 )
+∞ 2
Ptot y0 x + y2
P (y0 ) = dy dx exp − . (1.82)
πa 2 −∞ −∞ a2
with the error function erf(y0 /a). P (y0 /a)/Ptot is shown in Fig. 1.18, being 0.2398
and 0.7602 for y0 /a = √ ∓0.5, respectively. At these two positions one reads the
G AUSS radius a = w/ 2 as marked in Fig. 1.18.
In practice, of course, laser beams are never completely perfect. More or less pro-
nounced deviations from the ideal Gaussian beam profile are the rule. This is very
important when describing the quality of a laser system. Most concisely this is ex-
pressed by the beam parameter product (BPP) according to (1.69): the larger the
BPP, the poorer the beam quality. The international norm ISO 11146 defines as a
quantitative characteristic of a laser beam with a divergence angle θe and a beam
radius w0 the so called M 2 factor by
λ
θe = M 2 . (1.84)
πw0
1.3 Polarization 35
One may thus write M 2 as the ratio of ideal to real BPP, or real (θe ) to ideal diver-
gence angle (λ/πw0 ):
θ e w0 θe BPP BPP
M2 = = = = .
λ/π λ/(πw0 ) λ/π BPPideal
For a Gaussian beam M 2 = 1, while M 2 1.2 is a typical value for a very good,
real laser system.
Section summary
• Gaussian beams are as close as one can get to mimic a classical “light ray” –
they are, so to say, the diffraction limited version of a plane wave. Their spatial
profile is characterized by a single,√ real parameter, the R AYLEIGH length zR . It
is related to the waist radius w0 = zR λ/π of the beam at which the intensity
has dropped to 1/e2 of its maximum I0 .
• Along the propagation direction z, the intensity decreases to I0 /2 at z = ±zR .
The beam divergence angle in the far field is θe = w0 /z0 = λ/πw0 . This leads
to the so called beam parameter product BPP = θe w0 (= λ/π for a Gaussian).
The divergence of an arbitrary light beam is characterized by the factor M 2 =
BPP/BPPideal ≥ 1.
• A brief excursion into F RAUNHOFER diffraction theory shows that even the
well known A IRY pattern of a plane wave, diffracted by a circular aperture, is
less well focused than the Gaussian beam.
• Beam propagation in an optical setup is readily described with the help of the
complex beam parameter q(z) = z + izR and so called ray transfer matrices
(also ABCD matrices).
• Focussing a (nearly) parallel Gaussian beam with a lens of focal length f from
an initial beam radius w leads to a new waist radius w0 = f λ/πw. Hence, the
wider the beam originally is, the tighter it can be focussed. Telescope systems
are important tools for laser beam manipulation.
• The razor blade method allows an easy determination of beam radii.
1.3 Polarization
The electric field vector (1.35) of a light beam is expressed in terms of amplitude
E0 , polarization vector e and the exponential propagation terms exp[∓i(kz − ωt)].
We emphasize again, that E(r, t) is a real observable, and both exponential terms
are necessary for its description. The field amplitude E0 = E0 (ρ, z, t) may depend
(slowly) on ρ, and on z as we have just seen, as well as on t as we shall discuss in
Sect. 1.4.
The vector character of light is fully described by its polarization – an impor-
tant property of light. It determines e.g. the selection rules for optical transitions as
discussed in Chap. 4, Vol. 1. In the present section we resume and extend the dis-
36 1 Lasers, Light Beams and Light Pulses
cussion on polarization from Sect. 4.1, Vol. 1. We begin with fully polarized light
and discuss some specific temporal aspects of interest for nonlinear processes. We
then present several useful tools and recipes for the preparation and analysis of op-
tical polarization. Finally, we introduce the S TOKES parameters and the degree of
polarization for a realistic description of incompletely polarized light.
Depending on the case, one may use a Carthesian basis (ex ey ez ), or alternatively
the helicity basis (e+1 e−1 ez ) to describe unit polarization vectors (see Sect. 4.1.2 in
Vol. 1). We recall that the general unit vector for elliptically polarized light, propa-
gating parallel to the +z axis, may be written as
The ellipticity angle β describes the degree of ellipticity and the alignment angle δ
gives the direction of the ellipse in respect of ex .
2 ε0 c + 2
I (t, β) = ε0 cE(r, t) = E (r, t)e + E − (r, t)e∗ (1.86)
4
ε0 c −
= E (r, t)E (r, t) + sin(2β) Re E + (r, t)2 ,
+
(1.87)
2
and for a plane wave, the most simple case, at a given position in space (say kr = 0)
we obtain from (1.87)
ε0 c 2
I (t, β) = E0 1 + sin(2β) cos(2ωt − 2φ0 ) . (1.88)
2
Obviously, the intensity oscillates rapidly with twice the light frequency ω – ex-
cept for LHC or RHC polarized light, where β = 0 or = π/2, respectively, and
sin 2β = 0. The oscillations are most pronounced for linearly polarized light, where
β = ±π/4 and sin(2β) = 1.
1.3 Polarization 37
with PN (x) being the L EGENDRE polynomial of N th order. One easily verifies
that I N (t, β) decreases with increasing ellipticity – the higher N the more pro-
nounced. Indeed, recent experiments have shown that in the strong field of intense
femtosecond laser pulses the efficiency of multi-photon ionization decreases rapidly
and correlates in some cases surprisingly well with I N (t, β) (H ERTEL et al. 2009;
S HCHATSININ et al. 2009).
By the way: the relative simplicity of these expressions is essentially due to using
the helicity basis. It allows one to describe linearly, elliptically and circularly po-
larized light in the same coordinate system. In contrast, in the literature one usually
changes the coordinate system and chooses the z-axis to be parallel to the polar-
ization vector for linear polarization, while for circular polarization z is assumed
parallel to the propagation direction. Not only does this choice make the results eas-
ily confusing, it also does not allow to change the polarization continuously between
the two extremes.
Finally, we mention that the more general case of short pulsed laser radiation
may be treated in much the same manner as CW light discussed so far. The first
term in (1.87) will also vary with time (albeit slowly in the framework of the SVE
approximation), while the second term oscillates rapidly – subject to the magnitude
38 1 Lasers, Light Beams and Light Pulses
of the ellipticity angle β. Thus, the relations (1.90) for the cycle averaged N th power
of the intensity will also depend on time.
We shall now familiarize ourselves with some experimental tools for manipulation
of polarized light. We start with the transformation of linearly into circularly po-
larized light. As discussed in Sect. 4.1.2, Vol. 1, circularly polarized light may be
written as superposition of two linear components, oscillating perpendicularly to
each other with a phase difference of ±π/2, manifested in (4.12) and (4.13), Vol. 1.
The standard tool for generating such conditions is a so called λ/4 plate. That is
a thin, very plane ground, birefringent plate (typically made of quartz, magnesium
fluoride, or calcite). It has different indices of refraction nf and ns > nf for two
crystal axes, a so called fast (f) and a slow (s) axis. The phase velocity for light with
field vector parallel to these axes (i.e. light propagating perpendicularly to them)
is vf = c/nf and vs = c/ns , with vf > vs , respectively. The wavelengths for slow
and fast axes are correspondingly λf > λs , and the two orthogonal field components
develop a phase difference while passing through the plate. Specifically, for a λ/4
plate this will be 90◦ , i.e. the thickness of the plate d is given by
λ
d × (ns − nf ) = . (1.91)
4
The transformation of linearly polarized light into circularly polarized light is
sketched schematically in Fig. 1.19: linearly polarized light enters perpendicularly
onto the λ/4 plate, its E vector being aligned at 45◦ between fast and slow axis.
Its components, Ex and Ey , are thus parallel to one of the two crystal axes. The
figure illustrates how at the exit (out) of the λ/4 plate a shift of λ/4 between the
components arises. Together, these two components represent an electric field E
vector which rotates around the propagation axis z, as indicated in the figure. The
illustration shows σ + (LHC) light. Right hand circularly polarized σ − light (RHC)
would be generated if initially the linearly polarized E vector was aligned at −45◦ .
fast axis
-1 slow axis -1
A practical warning appears in order: λ/4 plates must be adjusted very care-
fully, in respect of perpendicular incidence as well as for proper alignment at 45◦ .
A simple check is done by rotating a linear polarizer very fast around the axis of the
circularly polarized beam (at best with a motor). The transmitted light – detected by
a synchronized oscilloscope – must not show any variation of intensity.
A useful rule about the orientation of the light applies: The sense of rotation of
the circularly polarized light behind a λ/4 plate is obtained by imagining to turn
the incident, linearly polarized E vector on the fastest possible way into the slow
axis. This rule is independent of the direction from which the beam is viewed.
For the materials used as λ/4 plates typically ns − nf
1. Hence, a λ/4 is much
thicker than λ/4 – otherwise it would be a very fragile object. We also mention that
the plates are usually not cut exactly parallel, but rather with a very small wedge so
that interferences from reflections are avoided. Equivalent phase shifts may also be
generated by plates with a thickness 5d, 9d, etc. One speaks of λ/4 plates of higher
order, which are much more stable.
When working with ultrafast light pulses one must, however, be very careful
with such plates. On the one hand thicker plates may lead to unwanted nonlinear
effects. On the other hand one has to realize that e.g. a pulse at 800 nm of 7.5 fs
duration (today not unusually short) consists of only a few cycles! A comparison of
the temporal profiles behind a λ/4 plate of first and higher order shown in Fig. 1.20
illustrates the consequences for ultrashort laser pulses very clearly. The absolute
phases are shifted in respect of the envelope of the two field components. Hence,
one expects from a first-order λ/4 plate a circularly polarized pulse of essentially
the same temporal shape as its linearly polarized parent. In contrast, second (and
higher) order plates will lead to serious distortions of the pulse shape. Note, that for
continuous light beams such considerations do not play a role.
Quarter wave plates have the disadvantage that they fulfill the condition (1.91)
only for exactly one wavelength. Alternatively, one may exploit the phase shift due
to total reflection inside a prism. The difference of the phase shift for light polarized
perpendicular and parallel to the plane of incidence is used.13 Specifically, the so
called F RESNEL rhomb sketched in Fig. 1.21 is often used. For an index of reflection
of n = 1.51 (glass) a phase difference of 45◦ is achieved with a rhomb angle of
13 According to B ORN and W OLF (2006) (Eq. (1.61)) the phase difference δ = δ⊥ − δ is given by
tan(δ/2) = cos θi sin2 θi − 1/n2 / sin2 θi , where n is the index of refraction of the medium, and θi
is the angle of incidence.
40 1 Lasers, Light Beams and Light Pulses
54◦ 37 (which is equal to the angle of incidence). The two reflections thus lead to an
overall phase difference of π/2. The figure illustrates also the sense of rotation of
the circularly polarized light obtained for two possible alignments of the incident E
vector. According to the above mentioned rule, the vertical axis of the rhomb is the
“slow axis”. The F RESNEL rhomb has the great advantage that it can be used over a
wide range of wavelengths. A disadvantage of F RESNEL rhombi is the displacement
of the beam.
F RESNEL rhombi are often used as pairs, turned by 180◦ , thus forming a λ/2
plate in which the beam displacement is compensated. In any case, half-wave plates
(F RESNEL rhombi or birefringent crystal plates) are also very useful optical devices.
Instead of (1.91) they satisfy
λ
d(ns − nf ) = , (1.92)
2
and are used, e.g. for rotating the plane of linearly polarized light as illustrated in
Fig. 1.22. We assume the fast axis of the λ/2 plate to be aligned at an angle α
in respect of the E vector of the incoming light beam. As shown in Fig. 1.22(a),
one may again consider the two components of E in respect of the slow and fast
axis. After passing half the plate (i.e. λ/4 path difference) the beam is elliptically
polarized as shown in Fig. 1.22(b). After a full passage through the λ/2 plate a phase
difference of π has been accumulated, i.e. the electric vector along the slow axis has
now the opposite sign as initially. Hence, the full E vector has been rotated by an
angle δ = 2α as shown in Fig. 1.22(c). A full rotation of the λ/2 plate from α = 0 to
2π thus rotates the polarization vector through 4π ; in between the E vector is four
times parallel to its original direction. An alternative application of a λ/2 plate is to
change LHC light into RHC and vice versa.
Fig. 1.22 Rotation of the polarization plane by means of a λ/2 plate. (a) E vector at the entrance
into the plate, (b) at half distance, (c) after full passage
1.3 Polarization 41
t
fas
w
slo
hI
h II
fast
with the path lengths hI and hII though the two plates. One may thus change s
from negative to positive values continuously, typically from −λ/4 up to 2λ. The
devices are built so that this optical path length is constant over a sufficiently large
area of the plate as needed for beams with finite radii.
With even more comfort, electro-optical devices are used (e.g. ADP) with a bire-
fringence that may be varied continuously by applying a high electric field (linear
electro-optical effect). These so called P OCKELS cells are of particular importance
in ultrafast laser technology.
I (0◦ ) − I (90◦ )
P1 = (1.94)
I (0◦ ) + I (90◦ )
I (45◦ ) − I (135◦ )
P2 = (1.95)
I (45◦ ) + I (135◦ )
I (RHC) − I (LHC)
P3 = . (1.96)
I (RHC) + I (LHC)
42 1 Lasers, Light Beams and Light Pulses
They characterize the relative intensity differences I (ep ) of a light beam in respect
of the three pairwise orthogonal polarization vectors defined in Sect. 4.1.2, Vol. 1.
For their measurement one needs in principle six different filters, each of which
transmits exactly one of these different polarizations ep . For 0◦ , 45◦ , 90◦ and 135◦
this is simply a linear polarization filter,14 for the analysis of circular polarization a
combination of a λ/4 plate and a linear polarizer.
The S TOKES parameters may easily be rewritten in terms of the parameters β
and δ, defining the general unit vector eel of polarization for elliptically polarized
light according to (1.85): one projects eel onto the 4 linear polarization vectors ep
according to (4.7) and (4.9), Vol. 1, or onto (e− , e+ ) for linear and circular polariza-
tion, respectively. The S TOKES parameters are then given by the difference of the
absolute squares of these amplitudes:
Degree of Polarization
In all previous considerations we have assumed that the light beams or wave fields
were essentially monochromatic, plane waves – modified if necessary for a spatial
variation of the amplitudes within the framework of the SVE approximation, leading
e.g. to Gaussian beams. In physical reality, however, we often deal with (i) only
quasi-monochromatic and (ii) only partially polarized light beams or wave fields.
We shall approach a full description of these facts in several steps. Presently, we
simply introduce as an easy to measure quantity the degree of polarization of light:
P = + P12 + P22 + P32 with 0 ≤ |P| ≤ 1. (1.100)
For fully polarized light – which may be described by eel according to (1.85) – the
degree of polarization is P = 1, as one verifies easily by inserting (1.97)–(1.99) into
(1.100). For many laser sources P ∼ = 1 is indeed a good description, while natural
light is most often unpolarized, i.e. one finds P = 0. This holds e.g. for diffuse illu-
mination by daylight, for incandescent light bulbs, or generally – expressed in more
scientific terms – for black body radiators (or cavity radiators). The three S TOKES
parameters completely characterize the polarization state of light. One also speaks
of a S TOKES vector P = (P1 , P2 , P3 ) of the light, its magnitude P being given by
(1.100).
14 As linear polarizers one uses specially cut arrangements of prisms, exploiting birefringence and
total reflection, such as the N ICOL or G LAN -T HOMPSON prism. Alternatively, thin film polarizers
are used, exploiting interference effects and special material properties.
1.3 Polarization 43
1
I (pol) = I0 1 + P12 cos 2 δ (anl) − δ . (1.104)
2
By varying the alignment angle δ (anl) of the analyzer one may thus determine P12 as
well as the alignment angle δ, for which the signal becomes largest. P12 < 1 implies
poor polarization – or a circularly polarized background. In both cases the ratio of
minimum (Imin ) to maximum signal (Imax ) is
Imin I (π/2 + δ) 1 − P12
= = . (1.105)
Imax I (δ) 1 + P12
For P12 = 1 and polarization along the analyzer axis, (1.104) is equivalent to the
well known M ALUS’s law I (δ (anl) ) = I0 cos2 δ (anl) .
Section summary
• Polarization is described most flexibly in the helicity basis, with the general
unit vector for elliptic polarization
eel = e−iδ cos βe+1 − eiδ sin βe−1 ,
with the ellipticity angle β and the alignment angle δ.
• On a sub-cycle time scale the light intensity depends crucially on the ellip-
ticity angle β. While the (1st order) cycle averaged intensity is independent
of β, higher order averages depend strongly on β. This is highly relevant for
processes such as MPI.
• λ/4 and λ/2 plates are useful devices for the generation and manipulation
polarized light.
• The S TOKES parameters, compact as S TOKES vector P = (P1 , P2 , P3 ), give
an even more general description of polarization, including partially polarized
light. The signal from a light beam with polarization P passing through an
analyzer described by P (anl) is given by
I (pol)=(I0 /2) 1 + P · P (anl) .
1.4 Wave-Packets 45
1.4 Wave-Packets
1
E + (r, t) = E0 (r, t)e−i(kr−ωt+φ0 )
2
1 + (k)e−i(kr−ωt) d3 k.
⇒ E (1.106)
(2π)3
+ (k)/(2π)3 is the (generally complex) amplitude for the different wave vectors k
E
(and correspondingly different angular frequencies ω = kc).
In Sect. 1.2 we have considered only the dependence of the field on position r.
Now, we take the opposite approach and focus on the variation with time t. To
make things not too complicated, we consider quasi-monochromatic light in vac-
uum, composed of plane waves, propagating in only one direction, with different
angular frequencies ω = kc from a narrow bandwidth δω |ω − ωc | around a car-
rier frequency ωc
δω. The latter condition indicates that we stay well within the
limits of the SVE approximation. We shall return to the dependence on r in the next
subsection.
We assume now that these waves propagate along the z-axis. We substitute
kz = ωz/c for kr and express the amplitudes as a function of ω only, replacing
E(k)d
3 k → E(ω)dω. We further simplify (1.106) by considering the field at one
point in space, say kr = 0. We write the time dependence of the field envelope as
an (inverse) F OURIER transform,15 i.e. we replace
i
E + (r, t) = − E0 (r, t)e−i(kr−ωt+φ0 ) ⇒
2
∞
1 + (ω)eiωt dω
E + (t) = E0 (t)ei(ωc t−φc ) = E (1.107)
2π −∞
∞
−iφc 1 − ωc )eiωt dω.
=e E(ω (1.108)
2π −∞
15 For more details about F OURIER transforms and spectral distributions see Appendix I in Vol. 1.
46 1 Lasers, Light Beams and Light Pulses
-3 -2 -1 1 2 3
-0.5 t/τ
The last identity is the frequency shift relation for F OURIER transforms (I.22),
Vol. 1. The phase shift φc refers here to the maximum of the carrier envelope
∞
1
E0 (t) = E0 h(t) = E(ω)e iωt
dω. (1.109)
2π −∞
The maximum field amplitude is E0 , and h(t) is a real envelope function, here nor-
malized such that h(0) = 1. If it is symmetric in respect of t = 0, E(ω) is also real
and symmetric around ω = 0. The carrier frequency ωc enters via (1.108), which
is to be inserted into (1.35) to describe the full time dependence of the electric
field vector E(t). It is important to note that this construction does not describe a
continuous light beam but rather a wave-packet, i.e. a light pulse of finite exten-
sion in space and time. With today’s ultrafast laser systems one may generate such
F OURIER transform limited light pulses without problems, typically with pulse du-
rations between picoseconds (ps) and some femtoseconds (fs) – with a trend towards
even shorter, attosecond (as) pulses. Experience shows that the temporal profile of
ultrafast laser pulses can often be described very well by a Gaussian, in analogy to
their spatial profile (1.47):
h(t) = exp −(t/τG )2 . (1.110)
As obvious from Fig. 1.24, the relative phase φc of the carrier oscillation becomes
an increasingly significant parameter as the pulse duration gets shorter. We still re-
main, however, within the limits of SVE. The temporal profile of the cycle averaged
intensity (I.19), Vol. 1 is16
ε0 c 2
I (t) = E0 h(t) = I0 exp −2(t/τG )2 , (1.111)
2
√
with a FWHM t1/2 = 2 ln 2τG = 1.177τG . (1.112)
16 We follow here the convention of the laser community with τG denoting the time at which the
intensity has decreased to 1/e2 .
1.4 Wave-Packets 47
10-2
0 10-3
-2 -1 0 1 2 -2 -1 0 1 2
t / Δ t 1/2
Fig. 1.25 Comparison of a Gaussian (red) and a sech2 (grey) intensity distribution on (a) linear
and (b) logarithmic scale. The FWHM t1/2 are in both cases identical, the time scale is measured
in units of this FWHM
Alternatively one often uses the squared hyperbolic secant for the temporal inten-
sity distribution, albeit mathematically less convenient (see Appendix I.4.2, Vol. 1):
2
ε0 cE02 2
I (t) = I0 sech (t/τs ) =
2
(1.113)
2 et/τs + e−t/τs
√
with a FWHM a t1/2 = τs 2 ln( 2 + 1) = 1.763τs . (1.114)
To obtain the same FWHM as in the Gaussian profile (1.111), one thus has to set
τs = 0.668τG . In Fig. 1.25 both time profiles are compared on a linear as well as on
a logarithmic scale. The main difference is in the wings: for large |t| the Gaussian
distribution decays significantly faster.
The spectral properties of these pulse shapes are discussed extensively in Ap-
pendix I.3, Vol. 1. One key result is the relation between the F OURIER transform
E(ω) of the field envelope and the spectral intensity profile:
ε0 c + 2 ε0 c 2
I˜(ω) = E (ω) = E(ω − ωc ) . (1.115)
4π 4π
which is centred at the carrier frequency ωc . For F OURIER transform limited pulses
we note the general rule,
2 ln 2
t1/2 ν1/2 = = 0.441. (1.118)
π
ν̄1/2 t1/2
= 14710, and (1.119)
cm−1 fs
(λ/nm)2
λ1/2 /nm = 1.471 × 10−3 . (1.120)
t1/2 /fs
A typical short, but not too short laser pulse may have t1/2 = 100 fs, a bandwidth
of ν̄1/2 150 cm−1 (or λ1/2 10 nm at 800 nm).
For the hyperbolic secant (1.113) the spectral intensity distribution (I.47), Vol. 1
is
˜ ε0 c E02 2 ω − ωc
I (ω) = sech , (1.121)
π ωs2 ωs
2 1.1224
where ωs = and the FWHM ω1/2 = 1.763ωs = . (1.122)
πτs τs
to be compared with (1.118) for the Gaussian. In Fig. 1.26 both spectral distributions
are compared, illustrating the narrower time-bandwidth product of sech.
1.4 Wave-Packets 49
The full position and temporal intensity distribution of a Gaussian light pulse is
obtained from the stationary expression (1.51) by replacing the maximum intensity
I0 there with I (t) according to (1.111):
2
2
I0 ρ t
I (ρ, z, t) = exp −2 exp −2 (1.124)
1 + ζ2 w τG
with w 2 = w02 1 + ζ 2 , w02 = zR λ/π and ζ = z/zR .
The overall maximum intensity I0 can be related to the total energy Wtot in the
pulse by integration over time and space. First, integration over the time gives the
so called fluence, in units [F ] = J/cm2 :
∞
I0 π
F (ρ, z) = I (ρ, t)dt = τ
2 G 2
exp −2(ρ/w)2 . (1.125)
−∞ 1 + ζ
Integration over the whole cross section of the beam as in (1.50) gives the pulse
energy Wtot , and the maximum intensity becomes
3/2
2 Wtot Wtot
I0 = = 0.83 . (1.126)
π τG w02 2
t1/2 d1/2
Wtot can readily be measured, as well as the beam width w0 (see e.g. Sect. 1.2.5).
With Z0 = (ε0 c)−1 = 376.7 one may then also derive the maximum field ampli-
tude (4.2), Vol. 1
E0 (0, 0) = 2I0 Z0 .
For a temporal dependence according to sech2 instead of (1.126) the relation
between pulse energy Wtot and intensity becomes with (I.44), Vol. 1
Wtot Wtot
I0 = 2
= 0.78 2
. (1.127)
πτs w0 t1/2 d1/2
With the N OBEL prize for H ALL and H ÄNSCH (2005) a broad scientific community
has become aware of frequency combs as tools for high precision calibration of
optical frequencies. In the present context they are an interesting, special kind of
wave-packets and just one more example.
In our treatment of Gaussian light pulses, we have so far described exactly one
isolated pulse with a duration t of typically a few fs. This situation may indeed be
realized in the laboratory without problems, typically with a repetition rate of some
Hz to several kHz as needed for any specific experiment. However, the preparation
50 1 Lasers, Light Beams and Light Pulses
~
E(ω) ωc= mcωr +ω0
ωr
ω0
Δωb
Fig. 1.27 Spectrum of a frequency comb with a carrier frequency ωc , a free spectral range
(turnaround time) ωr and an ‘offset’ ω0 (after U DEM et al. 2002)
ϕ 2ϕ
Δt1/2 ~ Tr / 5 field envelope
field amplitude
ϕ 2ϕ
Δt1/2 ~ Tr / 10 field envelope
field amplitude
Fig. 1.28 Example of frequency combs with two different bandwidths ωb of the laser amplifier
(ωb for the lower trace is twice that of the upper one). For practical reasons we have only summed
over a couple of modes in this model calculation, hence the remaining wiggles around zero signal
which disappear in reality
The electric field of such a mode structure is given by the superposition of all
modes. Depending on the amplification profile of the laser, the modes will have
different intensities (for simplicity we assume a Gaussian intensity profile ωb as
indicted in Fig. 1.27). Thus, the field amplitude is
∞
E(t) ∝ Re exp i (z + zc )ωr + ω0 t) (1.129)
z=−∞
× exp −4 ln 2[zωr /ωb ]2 .
Contributions to this sum come only from those modes which are amplified – still
a large number, typically on the order of 105 . This F OURIER series with many dis-
crete modes obviously replaces in the present case the F OURIER integral (1.107).
Figure 1.28 shows two model frequency combs.
As already discussed in the context of Fig. 1.24, for very short pulses the relative
phase of the carrier wave in respect of the envelope may play an important role. Here
we have now such a case. Since the carrier wave propagates with the phase velocity,
the envelope however with the group velocity, a little extra phase shift φ = ω0 /Tr is
accumulated for each full turnaround in the resonator, as illustrated in Fig. 1.28.
When working with frequency combs and “passion for precision”, as cultivated
by N OBEL laureate Ted H ÄNSCH (2005), one important issue is to measure this
offset or to make the frequency comb stable enough so that the offset can be com-
pensated completely. It was one of the crucial observation to find out that the short
laser pulses generated from these frequency combs may contain several octaves in
their frequency spectrum – and nevertheless remain coherent in phase. This opens
52 1 Lasers, Light Beams and Light Pulses
Section summary
• We construct wave-packets for short pulses as F OURIER transforms of plane,
monochromatic waves. Contributions come from a relatively narrow band-
width ωb around a carrier frequency ωc
ωb . The most used temporal
intensity profiles are Gaussian and sech2 .
• The spectrum of the pulse is proportional to the squared F OURIER transform
of the field amplitude
ε0 c 2
I˜(ω) = E(ω) .
4π
• Temporal width and the width of the spectrum of short pulses are inversely
proportional. The time-bandwidth products of F OURIER limited pulses are:
t1/2 ν1/2 = 0.441 (Gaussian) and ν1/2 t1/2 = 0.315 sech2 .
• Frequency combs are a fascinating and important tool for modern high preci-
sion spectroscopy. We have given here only a brief introduction.
1.5.1 Principle
We now give a brief introduction into the measurement of ultrashort laser pulses.
The pico-, femto-, or even attosecond time scales, studied in research today, are
much too short for direct time resolved recording with electronic techniques. One
has to resort to optical methods, essentially comparing optical pathways of the light.
In Fig. 1.29 the principle of such a measurement is sketched, exemplified by the
scheme for recording an autocorrelation function (the pulse is compared with a
copy of itself). First, the pulse is split into two equal parts, e.g. by a semi-transparent
mirror. Both parts then propagate along different pathways, with a variable time de-
lay δ built into one of them. In practice this is achieved by an optical delay line
which simply involves a longer optical path, e.g. with the help of a M ICHELSON or
M ACH -Z EHNDER interferometer. Both beams are then superposed again and finally
1.5 Measuring Durations of Short Laser Pulses 53
detected, exploiting some suitable linear or, more often, nonlinear optical effect. As
we shall see, one typically determines correlation (or autocorrelation) functions. The
signal may e.g. depend on the square of the total field or of the total intensity. As we
shall see in a moment, this effectively leads to a multiplication of the fields or inten-
sities, as indicated by [×] in Fig. 1.29, and finally one integrates over several cycles,
as symbolized by the boxed integral [ ]. Practical examples will be described at the
end of this section.
We briefly summarize here what has been communicated about correlation func-
tions in Appendix G, Vol. 1. The determination of pulse profiles (in the most simple
case just of the FWHM) is an important application. Ideally, one correlates the pulse
shape f1 (t) to be measured, with a second pulse shape f2 (t) which is well known.
The functions f1 and f2 may be the field amplitudes or the intensity or other charac-
teristic observables of the pulse. In an actual measurement one delays the pulse to be
measured in respect of the reference pulse by a well defined, variable time δ, multi-
plies both and then integrates over a sufficiently long time t. The signal detected as
a function of the delay time δ is thus given by17
∞
G(δ) = (f1 f2 )(δ) = f1∗ (t)f2 (t + δ)dt. (1.130)
−∞
The delay 2
time at which the cross-correlation function reaches 1/e of its maximum
value is τ22 + τ12 .
17 For later use, we generalize this expression to include also complex functions.
54 1 Lasers, Light Beams and Light Pulses
Often the pulse to be determined is also used as reference pulse (as indicated
in the schematic Fig. 1.29) – which of course requires a detailed knowledge of the
pulse shape if one wants to extract quantitative information. In that case one speaks
of an autocorrelation function. The pulse ‘inquires’ about itself, so to say, at a later
time how far it still remembers its own history. Specifically for a Gaussian intensity
pulse, with a FWHM t1/2 , the autocorrelation function is again a Gaussian with a
√
FWHM of auto
t1/2 = 2t1/2 . (1.132)
For a pulse whose intensity is described by sech2 (t/τs ) the situation is somewhat
more complicated, as outlined in Appendix G.4, Vol. 1. The squared hyperbolic
secant does not have the nice property of reproducing itself as autocorrelation func-
tion. One finds for the autocorrelation function a
FWHM of auto
t1/2 = 1.542t1/2 . (1.133)
In the schematic setup, introduced in Fig. 1.29, the multiplier is a key element for
measuring pulse durations. It indicates superposition of the pulse to be measured
with the reference pulse. Naturally, interference effects play an important role at
this point, though in practice one often tries to avoid them. Reliable detection of
interferences requires high spatial and temporal stability of the beam guiding optics.
In standard measurements one thus tries to average over an extended area so that
interferences cancel. If one actually is interested in detecting such interferences,
special care must be taken for the stability of the source and the setup. Also, one
has to superpose both beams as parallel as possible in order to be able to localize
the inference patterns well enough. Such an experiment is called an “interferometric
setup”.
In the following we shall start with the discussion of such a setup and only with
hindsight consider averaging processes in time and space. Typically one uses multi-
photon processes for detection which occur efficiently in the field of intense, band-
width limited laser pulses. Harmonic generation from the fundamental oscillation
frequency of the wave is most commonly exploited, especially second harmonic
generation (SHG). During the passage of intensive laser pulses (angular frequency
ωc ) through certain (nonlinear) optical crystals a fraction of the light is converted
into light with angular frequency 2ωc . This is just a consequence of the nonlinear
response of the crystal to electromagnetic radiation: the polarization and the index
of refraction depend on intensity, so that the pulse shape is distorted and contains
higher harmonics. Alternatively, one may resort to multi-photon excitation or ion-
ization of atoms and molecules.
Quite generally, N -photon processes occur to a good approximation with a prob-
ability proportional to the N th power of the light intensity (see Sect. 5.3 in Vol. 1).
As indicated in the schematic Fig. 1.29, the electric field at the detector originates
from superposition (interference) of the fields in the two beam parts – displaced in
time by δ. The signal S(δ) at the detector is then the cycle averaged N th power of
the intensity (1.90). For simplicity we assume circularly polarized light (β = 0 or
π/2) so that
N
S(δ) = I (δ)
ε0 c N − N
= E (t) + E − (t + δ) E + (t) + E + (t + δ) (1.134)
2
ε0 c N + 2 2 N
= E (t) + 2 Re E − (t)E + (t + δ) + E + (t + δ)
2
I0N Tav /2 2 N
= h (t) + 2h(t)h(t + δ) cos(ωc δ) + h2 (t + δ) dt, (1.135)
Tav −Tav /2
where h(t) is the envelope function of the field amplitude (1.109). For short pulses
of duration τ the angle brackets . . . refer to averaging over a sufficiently long time
±Tav /2 with Tav
(τ + δ). In addition, the experiment averages over variations of
the phase differences ωc δ, which may occur statistically from pulse to pulse due to
experimental instabilities, or due to spatial averaging over an extended detector area
etc.19
We may account for this averaging by one further integration over delay times
equivalent to one period of the carrier frequency Tc = 2π/ωc :
Tc /2
1
S(δ) = S(δ)dδ.
Tc −Tc /2
This will now be specialized for some examples as illustrated in Fig. 1.30.
19 Note, here we do not refer to the absolute stability of the “carrier envelope phase” φc according
to Fig. 1.24, which does not enter into the cycle averaged expression (1.135). The signal is only
influenced by fluctuations of the relative phase φ = ωc δ between the two time delayed pulses.
56 1 Lasers, Light Beams and Light Pulses
Fig. 1.30 Autocorrelation functions S(δ) of order N as function of the delay time δ. Calculated
from (1.135) for a Gaussian pulse; red S(δ): interferometric stability, black S(δ): averaged over
phase fluctuations. The signal is normalized at large delay times to S(δ → ∞) = 1
N =1
Let us start by assuming that we have a detector (photodiode, electron multiplier,
thermopile) which simply registers the total intensity linearly, i.e. only one photon is
involved in the detection process. Then (1.135) essentially describes YOUNG’s clas-
sical double slit interference experiment. The crucial interference term in (1.135) is
determined by the autocorrelation function of h(δ) according to (1.130). Normalized
to the signal at δ/τG
1 one finds
∞
S(δ)/S(∞) = 1 + cos(ωc δ) h(t)h(t + δ)dt. (1.136)
−∞
For a Gaussian envelope (1.110) this gives according to (1.130) and (1.131)
1 δ 2
S(δ)/S(∞) = 1 + exp − (cos ωc δ). (1.137)
2 τG
This expression is illustrated in Fig. 1.30, panel N = 1: YOUNG’s double slit ex-
periment thus measures the autocorrelation function of the electric field. In prin-
ciple, one could use such an experiment to determine the pulse duration – if the
phase φ = ωc δ was sufficiently stable (any fluctuations would have to be small
∂(φ)
π ). This is, however, in the usual simple setup not the case, and the exper-
iments averages over δ for a number of periods. Consequently, cos(ωc δ) → 0 and
S(δ)/S(∞) → 1, i.e. the measured signal becomes completely structureless (black
line in Fig. 1.30, panel N = 1).
N =2
One avoids this destruction of the autocorrelation function in the phase averaged
signal by a nonlinear detection scheme. A popular method is to exploit SHG which
originates from two photons of frequency ωc . The signal depends quadratically
(N = 2) on the laser intensity. In the case of a Gaussian envelope one may again
1.5 Measuring Durations of Short Laser Pulses 57
integrate (1.135) in closed form, and obtains as autocorrelation function (2nd order
in the field amplitude)
2
S(δ) 3δ 2 δ
= 1 + 4 exp − 2 cos ωc δ + exp − 2 1 + 2 cos2 ωc δ . (1.138)
S(∞) 4τG τG
This is sketched in Fig. 1.30, panel N = 2. The maxima are now massively en-
hanced due to the squared dependence of the signal on intensity, equivalent to
|E(t)|4 (red trace). Without specific provision for an interferometric measurement,
the phase fluctuations will, here too, wash out the interference structures. Averaging
over at least one period makes the first cos(ωc δ) term disappear, while the second,
cos2 ωc δ averages to 1/2. Hence
S(δ)/S(∞) = 1 + 2e−(δ/τG ) .
2
(1.139)
Clearly, in this case the averaged signal remains a function of the time delay δ (black
line in Fig. 1.30, panel N = 2). With such a measurement the autocorrelation func-
tion (1.131) of the laser intensity may be determined even if the phase ωc δ fluctuates
slightly from pulse to pulse. To be specific: the averaged signal (1.139) corresponds
to the autocorrelation function 2nd order in the field amplitude, and the exponential
term is just the autocorrelation function 1st order of the intensity. Independent of
the line profile (1.110), the phase averaged signal for the case N = 2 is given by:
∞ 2
S(δ) h (t)h2 (t + δ)dt
= 1 + 2 −∞ ∞ 4 . (1.140)
S(∞) −∞ h (t)dt
Equation (1.140) is the basis for standard analysis20 of short pulses by SHG detec-
tion.
N =4
The interference signal expected in a four photon process as a function of the delay
time δ is shown in Fig. 1.30, panel N = 4, again for two coherent Gaussian laser
beams. It represents essentially the 4th order autocorrelation function of the laser
field. Again, the full red line predicts the interference signal if the detector effec-
tively averages only over phase fluctuations δ
τ . The black bell shaped curve
represents the average in case of strong phase fluctuations. Note that the maximum
of the signal (also called the four photon coherence signal) is 50 times higher than
the background at δ
τ .
In Table 1.5 the results discussed above and some more are summarized. Note
the strong enhancement of the maximum as the order N increases, while at the same
time the overall width t1/2 of the phase averaged signal decreases.
20 However, in a real experiment the two time delayed beam parts typically intersect at a very small
angle, and the SHG signal is detected at the angle bisector. For this geometry we have to add the
specific phase matching conditions to (1.134). Due to this arrangement the measured signal is then
free of background (S(δ → ∞) = 0).
58 1 Lasers, Light Beams and Light Pulses
Generalization
In our examples we have assumed so far Gaussian envelopes of the field amplitude.
Important limits of (1.135) may, however, be formulated quite generally. For exam-
ple, at large delay times one may neglect all terms containing products of functions
taken at different times t and t + δ, respectively. Thus, for large δ only the N th
power of the first and the last term contribute and we obtain
for δ
τ simply S(δ → ∞) = 2 × (I0 )N .
The other limiting case is δ = 0, which leads according to (1.135) to the maximum
signal S(δ = 0) = (4 × I0 )N , so that
S(δ = 0) 4N
= (1.141)
S(δ → ∞) 2
independent of the pulse shape. This relation describes the maximum signal (at
δ = 0) in the case of interferometric stability, i.e. one requests fluctuations ∂(ωc δ)
π/2 – in the observation volume and over the total observation time.
Conversely, if the phase fluctuation is large (i.e. ∂(ωc δ)
π ) – and that is indeed
the case for most experimental setups which do not take special care for stabiliza-
tion – the averaging may be modelled independently from the line profile. We still
assume that the pulse duration is long compared to the period of oscillation, i.e. that
ωc
1/τ . We may then assume h(t) h(t + δ) over the phase averaging time and
do the averaging prior to the integration over time t. Independent of the pulse shape,
one obtains then for
N
δ
τ always S(0) = 2N I0N 1 + cos(φ) and hence (1.142)
π
S(0) 1 φ (2N )!
= 22N −1 cos2N dφ = . (1.143)
S(∞) 2π −π 2 2(N !)2
Note that this expression reproduces the values explicitly derived for Gaussians as
shown in Table 1.5.
1.5 Measuring Durations of Short Laser Pulses 59
1
I(λ) (a)) 1 (b))
(b 1.0 ( )
(c)
spectrum
c
~
Δλ ≈ sech2
22 4
22.4nm FWHM
0.5 0
0.5 0.1
118 fs
Gaussian
0
1000 1050 1100 - 200 0 200 - 100 0 100
wavelength λ / nm Δν1/2 Δt1/2 = 0.45 delay time δ / fs
Fig. 1.31 Experimentally determined (a) spectral and (b, c) temporal intensity distribution of
a nearly F OURIER limited laser pulse, with ν1/2 = 6.25 THz and t1/2 = 74 fs according to
S CHMIDT et al. (2010)
Figure 1.31 shows a nice experimental example for the determination of a short
pulse duration. A mode locked laser pulse generated in a diode pumped crystal made
out of a new material, Yb:LuScO3 , was investigated by S CHMIDT et al. (2010).21
The experimentally determined amplification profile I˜(λ) in Fig. 1.31(a) may be fit-
ted surprisingly well by a Gaussian or a sech2 profile and has a bandwidth (FWHM)
of ca. 22.4 nm (corresponding to ν1/2 = 6.25 THz). The autocorrelation function
Fig. 1.31(b) has been measured in a pump-probe scheme as just discussed (N = 2).
From the fit of the experimental data with a sech2 (t/τs ) distribution one finds for
the autocorrelation function a FWHM ≈ 118 fs, which according to (1.133) cor-
responds to a pulse width of t1/2 = 74 fs. The logarithmic display Fig. 1.31(c)
allows an instructive comparison of the two pulse shapes described in Sect. 1.4.1.
In the present case excellent agreement is found with a suitably fitted G AUSS pro-
file. The time-bandwidth product measured corresponds to ca. 0.45, to be compared
with the theoretical value 0.315 according to (1.123). One may consider this pulse
as nearly F OURIER limited, in particularly so as the choice of the profile for fitting
the experimental data is not actually compelling (see also footnote 18).
Figure 1.32 shows the realization of a typical interferometric measurement.22
The experimental setup used here consists of a very small, stable M ICHELSON in-
terferometer and detection by SHG (N = 2) in a thin BBO crystal. To compensate
dispersion effects two beam splitters are used in this special setup. This makes the
setup applicable also for very short laser pulses (<10 fs). The experimental measure-
ment shown here has been obtained with a pulse from a T I :S APPH laser (800 nm)
having a pulse duration of ca. 19.5 fs. One sees a nice interference pattern, sym-
metric in respect of δ = 0, and rather close to the expected ratio S(0):S(∞) = 1:8
21 We thank Uwe Griebner for providing us with the original, measured data.
22 We thank Günter S TEINMEYER (2010) for kindly letting us have his experimental data and
sketch of the setup as well as for helpful discussions.
60 1 Lasers, Light Beams and Light Pulses
be
am
provided by S TEINMEYER input
8
(2010)
sp
lit
te
rs
delay
1
0
-150 -100 -50 0 50 100 150
delay time δ / fs
(cf. Fig. 1.30, panel N = 2). The side maxima originate from satellite pulses due to
incomplete compensation of mode turnaround times in the laser resonator.
Section summary
• The duration (FWHM) t1/2 of ultrashort laser pulses can be measured
by optical time delay δ. This is achieved by different optical path lengths
s1 − s2 = cδ for two equal fractions of the laser beam. Effectively one mul-
tiplies their electric fields (or intensities) f1 (t) and f2 (t + δ), and integrates
over a time
t1/2 .
• Mathematically such a measurement is expressed as first-order autocorrela-
tion function:
∞
G(δ) = f1 (δ) f2 (δ) = f1∗ (t)f2 (t + δ)dt.
−∞
As long as processes are studied which depend linearly on intensity, averaging over
the temporal and spatial profile of a laser beam leads to signals which are simply
proportional to the averaged intensity: the cross section or absorption coefficient
does not depend on intensity. However, in the case of nonlinear processes the sit-
uation is more complicated as just illustrated, and some basics about multi-photon
processes have been discussed in Sects. 5.3 and 8.5, Vol. 1. Since the interaction of
free atoms, molecules and cluster with intense laser fields is an important theme in
modern laser based science, we now take a closer look at multi-photon ionization to
illustrate the consequences of nonlinearity.
(N )
Let N be the particle density of the target, σba the relevant cross section for N
photon ionization, and Φ = I /ω the photon flux (dimension L−2 T−1 ). According
to (5.43), Vol. 1 the rate (dimension T−1 ) for the MPI process is:
(N ) (N )
Rba = σba Φ N = sN I N (ρ, z, t) with sN = σ (N ) /(ω)N (1.144)
(N )
dimension of σba : L2N TN −1 ,
Such processes may be investigated efficiently with short laser pulses. Let us con-
sider a Gaussian beam whose intensity depends on time and position according to
(1.124). To compute the expected experimental signal one has to integrate over time
and detection volume. Also, one must account for the fact that the initial target
density N0 may change significantly during the pulse, since ionization depletes the
initial state. Hence,
at which the initial number of target species in the ground state has decreased to
1/e. We can now rewrite (1.145):
N
N(ρ, z) = N0 exp − I (ρ, z)/Is . (1.148)
πN0 w02 zR
dS(u) = 1 + ζ 2 γ + ln uN + E1 uN dζ, (1.151)
2N
∞
with E1 (x) = − Ei(−x) = 1 (e−xt /t)dt (see e.g. W EISSTEIN 2004, Ei(x) being
the so called exponential integral) and γ = 0.5772157 E ULER’s constant. Alterna-
tively one may expand the integrand in (1.150) into powers of ρ and then integrate:
∞
πN0 w02 zR uj N
dS(u) = − 1 + ζ2 (−1)j dζ (1.152)
2N j j!
j =1
∞
πN0 w02 zR (I0 /Is )j N 1−j N
= (−1)j −1 1 + ζ2 dζ.
2N j j!
j =1
For sufficiently low intensities, typically I0 /Is < 0.5, the first term of the series
dominates, and independent of the geometry the signal becomes ∝ (I0 /Is )N as
expected. The situation is more complicated if the intensity becomes comparable
to saturation intensity. The series expansion converges then only very slowly and
further considerations are required which will be explicated in the following. Two
experimental geometries are distinguished.
1.6 Nonlinear Processes in Gaussian Laser Beams 63
S /S (1)
100
1
I0 / Is
0.1 10 100
( I0 / Is ) 5 0.01 d
z
10-4 ρ
(I0 / Is ) 8 10-6
tar
ge
t
10-8
Fig. 1.33 Multi-photon ionization signal S as a function of intensity I , measured in units of the
saturation intensity Is in a log − log plot. The strictly cylindrical geometry is sketched in the inset.
The full lines are computed according to (1.153) with S ∝ I 5 and ∝ I 8 . The experimental points
are data for the ionization of C60 → C+ ++
60 (red) and C60 (grey) obtained with laser pulses at 800 nm
and a pulse duration of 27 fs (S HCHATSININ et al. 2006)
Most simple is the strictly cylindrical 2D geometry where the target is a thin sheet
of thickness d
zR which is traversed perpendicularly by the focussed laser beam
(R AYLEIGH length zR ). This may be realized by a molecular beam whose angular
divergence is limited by a slit. The geometry is illustrated in the inset of Fig. 1.33.
In this case the ζ dependence in (1.151) can be ignored and the signal is
N
πN0 w02 d I0 I0
S(u) = γ + N ln + E1 . (1.153)
2N Is Is
We recall that for low intensities this expression is ∝ (I0 /Is )N . Conversely, for suf-
ficiently high intensity, say I0 /Is > 3, the E1 term may be neglected and the ln term
dominates. Figure 1.33 illustrates this signal for a 5 and an 8 photon ionization pro-
cess. The experimental example chosen is MPI of C60 by focussed 800 nm (1.55 eV)
laser pulses of 27 fs duration (FWHM) according to S HCHATSININ et al. (2006).
The double logarithmic display documents the power law S ∝ I N and allows one
to directly extract the exponent N as slope for low intensities. The generation of
C+ ++
60 (red) obviously requires 5 photons (S ∝ I ), the double ionization (C60 grey)
5
potentials of C60 (7.56 eV 5ω) and of C+ 60 (11.8 eV 8ω), respectively. For in-
tensities I0 > Is one clearly recognizes the saturation like behaviour. It sets in when
in the centre of the laser beam most of the neutral target molecules are ionized. The
23 Figure 1.33 is scaled dimensionless, i.e. the intensities have been normalized to the saturation
intensity Is , while the ion signal S is normalized to the signal S(I0 /Is = 1).
64 1 Lasers, Light Beams and Light Pulses
4
C602+
S / arb. un.
2 Is
C60+
0
0.2 1 2 10
I 0 / 1014 Wcm-2
Fig. 1.34 Multi-photon ionization signal S as a function of the intensity I as in Fig. 1.33 – now,
however in a lin − log plot. The experimental data (signal S in arb. un. vs. maximum I0 in the
focus in W cm−2 ) are not rescaled here. Slightly different saturation intensities are found for C+
60
and C++
60 (red and grey arrow, respectively)
volume (here a circular disk) within which I (ρ, z) ≥ Is grows with increasing max-
imum intensity I0 . Thus, the increase in signal above saturation in the centre results
from an increase in effective volume.
This simple geometry can be evaluated easily. Alternatively to the log − log dis-
play, one may plot the signal S linearly versus log(I0 ) and extrapolate the data at
high intensity to zero signal, as suggested by H ANKIN et al. (2001). With (1.153)
one reads the saturation intensity Is as intersection of this line with the intensity
axis (apart from a small shift due to γ and the influence of the E1 term, negligible
at high intensities). This is illustrated in Fig. 1.34 for the example just discussed.
The experimental data are not scaled. Clearly, in this special example the satura-
tion intensities for C+ ++
60 and C60 differ only little from each other. This is a some-
what puzzling result if one assumes sequential ionization according to the scheme
C60 + N1 ω → C+ − +
60 + e as a first, and C60 + N2 ω as a second step (with N1 = 5
and N2 = 8): obviously the physics is somewhat more complex! From the saturation
intensities thus determined one may in principle determine the MPI cross section by
simply inverting (1.147). With (1.146) and (1.144) one obtains
−1 −N
sN = τN Is and σ (N ) = (ω)N τN
−1 −N
Is . (1.154)
This is, of course, only valid if the power law holds up to saturation. Even if this is
not strictly correct, the relation may serve as a first-order approximation for obtain-
ing an estimate for an ‘equivalent’ MPI cross section.
For the record we note another important fact: the measured saturation intensities
depend on the pulse width. This is of course not surprising since saturation implies
depletion of the initial target state which clearly is a function of time. Providing
(1.144) is strictly valid, we obtain from (1.147) for a given cross section at two
different pulse widths τ1 and τ2 the ratio of the respective saturation intensities:
To achieve the highest possible intensities one has to focus the laser beam tightly.
Typically in such experiments, the extension d of the target volume in z-direction
‘seen’ by the detector gets larger than the R AYLEIGH length, d
zR . The simple
cylindrical geometry sketched in Fig. 1.33 is thus no longer applicable. Instead, one
has to integrate (S PEISER and J ORTNER 1976) over the full ‘dog bone’ geometry of
the focus illustrated in Fig. 1.10(c).
Let us first get a rough estimate of the volume Vs within which the intensity
is larger than the saturation intensity Is : At sufficiently high I0 the boundary of
Vs is located in the far field region so that the intensity (1.124) on the beam axis
is, reasonably accurate, I √ I0 /ζ 2 . The extension of the ‘saturated dog bone’ in
z-direction is thus zs = zR I0 /Is , from which we obtain
zs πw 2 w2 zs πw02 zR I0 3/2
Vs = dV dz = π 20 z2 dz = .
−zs 2 zR 0 3 Is
The thus derived dependence (of the observed signal) on I 3/2 is characteristic for
saturated processes in 3D geometry. A clean evaluation has to account also properly
for the radial expansion – as it turns out, this does not change the overall result.
Strictly speaking one also has to account for the regions outside the saturation vol-
ume Vs . This implies integration of (1.151) or (1.152) over all z from −∞ to +∞.
One finds that the integral exists if N > 3/2. But it is not completely trivial to eval-
uate it, since the series (1.152) converges rather slowly. It has been evaluated for the
first time by C ERVENAN and I SENOR (1975).
One chooses a desired precision by the parameter L (typically L ≡ 3) and in-
tegrates for small (I0 /Is )N ≤ L the series, for large (I0 /Is )N > L the expression
(1.151) were then the exponential integral may be neglected. With u ≡ I0 /Is the
signal for uN > L is
∞
S(u) u3/2
1/N −j
1/N
1/N
= a(j ) u/L + G u/L + H u/L .
S(1) V (N )
j =0
(−1)m Γ (1/2)π
V (m) =
Γ (m − 1)Γ (5/2 − m)
∞
Γ (1/2) (−L)n
a(j ) = (−1)j
Γ (j + 1)Γ (1/2 − j ) (2(N n + j ) − 3)nn!
n=1
66 1 Lasers, Light Beams and Light Pulses
S 1000 3D geometry
( I / I s) 3/2
100 ln ( I / I s)
1 2D geometry
0.1 10 100
0.01 I / Is
( I / I s) 5
10-4
( I / I s) 8
10-6
Fig. 1.35 Multi-photon ionization signal S as a function of the intensity I measured in units of the
saturation intensity Is in a log − log plot – as in Fig. 1.33. In contrast, however, here this simulation
of the signal represents an integration over the full volume of the laser beam (3D geometry). For
comparison, also shown are the corresponding (I0 /Is )3/2 traces (dashed lines), and the ln(I0 /Is )
fits (thin lines) reproduced from Fig. 1.33
1
G(u) = (γ + ln L)(1 + 2/u)(1 − 1/u)1/2
3
2
H (u) = N (1 − 1/u)3/2 + 6 1/u − 1 − arcsin(1 − 1/u)1/2 .
9
In Fig. 1.35 the behaviour of this function for the full 3D geometry is sketched. One
recognizes that this leads to a faster rise of the signal in the saturation region in
comparison to the 2D geometry.
In practice the observed signals are even more complicated. As already men-
tioned, sequential multi-photon ionization is observed in many cases, but may be
complicated by more delicate processes, involving e.g. non-sequential processes
where several electrons are involved. In addition, in molecules fragmentation pro-
cesses may occur, further complicating the observed signals. In clusters interaction
of the field with a micro-plasma induced within the cluster can lead to a wealth of
processes which evolve independently of each other. Due to their specific intensity
dependence these processes may in turn impress characteristic structures onto the
temporal and spatial distribution of the ionization yield. In any case, the interaction
of atoms, molecules and nano-particles with intense laser pulses is an exciting field
of modern research which develops at very fast pace.
Time and position dependent structures of these processes have been studied in a
nice, still relatively simple experiment by S TROHABER and U ITERWAAL (2008).
It is summarized in Fig. 1.36. MPI of Xe atoms is investigated with a short laser
pulse (800 nm, 50 fs). (a) A combination of a narrow slit (y-coordinate) with energy
analysis of the ions by time of flight (TOF, x-coordinate) leads to a 3D image of the
1.6 Nonlinear Processes in Gaussian Laser Beams 67
lens
(a) MCP
laser
beam
(d)
slit
ion trajectory d repeller
V3 V2 V1 = 0 V= 0 VR
V0 = (x0 /d )VR
x0 y
x
optical z
axis
(b) (c) Xe+
TOF/μs
Fig. 1.36 Experiment with spatially resolved detection of multi-photon ionization for Xe atoms in
the focus of a Gaussian beam according to S TROHABER and U ITERWAAL (2008). (a) Experimental
setup, (b) dependence of the TOF on charge and origin x of ions, (c) Xeq+ signal for different
charges q as a function of x (signal traces) and (d) as a 3D plot of the signal strength in respect of
the xy plane
different charge states observed. Figure 1.36(b) shows the TOF as a function of the
positions x where the ions have been created, for Xeq+ ions of charge state q = 1–4.
The saturation intensity for these MPI processes is highest for q = 4 and obviously
chosen in the present experiment such that the all other charge states are already
bleached out in the centre of the laser focus. The smaller q, the lower the saturation
intensity – which leads to detection of these ions from the outer zones of the laser
beam: the smaller q the farther away from the focus the respective maximum signal
is seen. Figures 1.36(c) and (d) give a 2D and 3D overview, respectively, of the ob-
served ion signals. Ultimately (d) is something like a direct, nonlinear image of the
laser intensity. One may directly compare it to the ‘dog bone’ intensity distribution
in Fig. 1.10.
Section summary
• While in linear spectroscopy the spatial and temporal profile of the laser beam
is usually irrelevant, it plays a crucial role when nonlinear processes, such as
N photon ionization, are studied.
• At low intensities, one expects a signal ∝ I N (except for more complex pro-
cesses, such as non-sequential MPI). Characteristic in high intensity Gaussian
beams is a depletion of the initial target state and hence saturation of the ob-
68 1 Lasers, Light Beams and Light Pulses
served signal, if the maximum intensity I0 in the beam is higher than a process
dependent saturation value Is . For I0 > Is it is geometry rather than physics
that determines the signal.
• Assuming N th power dependence of the rate, the saturation intensity
−1/N
I s ∝ τ N σ (N )
References
B LOEMBERGEN , N. and A. L. S HAWLOW: 1981. ‘The N OBEL prize in physics “for their contribu-
tion to the development of laser spectroscopy” ’, Stockholm. http://nobelprize.org/nobel_prizes/
physics/laureates/1981/.
B ORN , M. and E. W OLF: 2006. Principles of Optics. Cambridge: Cambridge University Press,
7th (expanded) edn.
C ERVENAN , M. R. and N. R. I SENOR: 1975. ‘Multi-photon ionization yield curves for Gaussian
laser-beams’. Opt. Commun., 13, 175–178.
E INSTEIN , A.: 1916. ‘Strahlungs-Emission und -Absorption nach der Quantentheorie’. Verh.
Dtsch. Phys. Ges., 18, 318–323.
G ORDON , J. P., H. J. Z EIGER and C. H. T OWNES: 1955. ‘Maser – new type of microwave am-
plifier, frequency standard, and spectrometer’. Phys. Rev., 99, 1264–1274.
H ALL , J. L. and T. W. H ÄNSCH: 2005. ‘The N OBEL prize in physics: for their contributions to
the development of laser-based precision spectroscopy, including the optical frequency comb
technique’, Stockholm. http://nobelprize.org/nobel_prizes/physics/laureates/2005/.
H ANKIN , S. M., D. M. V ILLENEUVE, P. B. C ORKUM and D. M. R AYNER: 2001. ‘Intense-field
laser ionization rates in atoms and molecules’. Phys. Rev. A, 6401, 013405.
H ÄNSCH , T. W.: 2005. ‘N OBEL lecture: Passion for precision’, Stockholm. http://nobelprize.org/
nobel_prizes/physics/laureates/2005/hansch-lecture.html.
H ERTEL , I. V. I. S HCHATSININ, T. L AARMANN, N. Z HAVORONKOV, H.-H. R ITZE and C. P.
S CHULZ: 2009. ‘Fragmentation and ionization dynamics of C60 in elliptically polarized fem-
tosecond laser fields’. Phys. Rev. Lett., 102, 023003.
H ODGSON , N. and H. W EBER: 2005. Laser Resonators and Beam Propagation, vol. 108 of
Springer Series in Optical Sciences. Berlin: Springer, 2nd edn., 824 pages.
ISO 21348: 2007. ‘Space environment (natural and artificial) – Process for determining solar irra-
diances’. International Organization for Standardization, Geneva, Switzerland.
JAVAN , A., W. R. B ENNETT and D. R. H ERRIOTT: 1961. ‘Population inversion and continuous
optical maser oscillation in a gas discharge containing a He-Ne mixture’. Phys. Rev. Lett., 6,
106–110.
KOGELNIK , H. and T. L I: 1966. ‘Laser beams and resonators’. Appl. Opt., 5, 1550–1567.
L INDNER , F., G. G. PAULUS, H. WALTHER, A. BALTUSKA, E. G OULIELMAKIS, M. L EZIUS and
F. K RAUSZ: 2004. ‘Gouy phase shift for few-cycle laser pulses’. Phys. Rev. Lett., 92, 113001.
M AIMAN , T. H.: 1960. ‘Optical and microwave-optical experiments in ruby’. Phys. Rev. Lett., 4,
564–566.
M ILLONI , P. W. and J. H. E BERLY: 2010. Laser Physics. Hoboken: Wiley, 832 pages.
S CHÄFER , F. P., W. S CHMIDT and J. VOLZE: 1966. ‘Organic dye solution laser’. Appl. Phys.
Lett., 9, 306–309.
S CHAWLOW , A. L. and C. H. T OWNES: 1958. ‘Infrared and optical masers’. Phys. Rev., 112,
1940–1949.
70 1 Lasers, Light Beams and Light Pulses
S CHMIDT , A. et al.: 2010. ‘Diode-pumped mode-locked Yb:LuScO3 single crystal laser with 74 fs
pulse duration’. Opt. Lett., 35, 511–513.
S HCHATSININ , I., H.-H. R ITZE, C. P. S CHULZ and I. V. H ERTEL: 2009. ‘Multi-photon excitation
and ionization by elliptically polarized, intense short laser pulses: Recognizing multi-electron
dynamics and doorway states in C60 vs. Xe’. Phys. Rev. A, 79, 053414.
S HCHATSININ , I., T. L AARMANN, G. S TIBENZ, G. S TEINMEYER, A. S TALMASHONAK, N.
Z HAVORONKOV, C. P. S CHULZ and I. V. H ERTEL: 2006. ‘C60 in intense short pulse laser
fields down to 9 fs: excitation on time scales below e-e and e-phonon coupling’. J. Chem. Phys.,
125, 194320.
S IEGMAN , A. E.: 1986. Lasers. Sausalito: University Science Books, 1283 pages.
S OROKIN , P. P. and J. R. L ANKARD: 1966. ‘Stimulated emission observed from an organic dye,
chloro-aluminum phthalocyanine’. IBM J. Res. Dev., 10, 162–163.
S PEISER , S. and J. J ORTNER: 1976. ‘3/2 power law for high-order multi-photon processes’.
Chem. Phys. Lett., 44, 399–403.
S TEINMEYER , G.: 2010. ‘Interferometric determination of the autocorrelation function of a sub
20 fs laser pulse’. Private communication.
S TRICKLAND , D. and G. M OUROU: 1985. ‘Compression of amplified chirped optical pulses’.
Opt. Commun., 56, 219–221.
S TROHABER , J. and C. J. G. J. U ITERWAAL: 2008. ‘In situ measurement of three-dimensional
ion densities in focused femtosecond pulses’. Phys. Rev. Lett., 100, 023002.
U DEM , T., R. H OLZWARTH and T. W. H ÄNSCH: 2002. ‘Optical frequency metrology’. Nature,
416, 233–237.
W EISSTEIN , E. W.: 2004. ‘En-function’, Wolfram Research, Inc., Champaign, IL, USA. http://
mathworld.wolfram.com/En-Function.html, accessed: 9 Jan 2014.
Coherence and Photons
2
Overview
After the previous extensive exploration into the wave character of light, the
present chapter focuses on its particle properties and on the statistical prop-
erties of photons. In Sect. 2.1 concepts such as “quasi-monochromatic” and
“partially coherent” light will be defined and exemplified by simple models
for a laser and a classical light source. We shall familiarize ourselves with
the fundamental experiments, beginning with the famous “Hanbury B ROWN -
T WISS experiment”. In Sect. 2.2 we shall try to find a pragmatic approach to
the quantum mechanical description of photon states – giving an introduction
for “pedestrians” so to say. Finally, we shall in Sect. 2.3 apply the new tools
to the theory of absorption and emission of light – this time with explicit con-
sideration of the quantum nature of photons. This will allow us for the first
time to derive the basic formulas for spontaneous emission – as opposed to
the previous, hand waving introduction of this inherently quantum mechanical
phenomenon.
2.1.1 Introduction
Ground breaking work on the quantum statistics of light has been carried out in 1954
and the following years. Of fundamental importance are the experiments of R. Han-
bury B ROWN1 and R.Q. T WISS (1954, 1956a, 1956b, 1958). Roy J. G LAUBER was
one of the pioneers of theoretical quantum statistics (see e.g. G LAUBER 1963) and
received the N OBEL prize for his work 2005 – together with John H ALL and Ted
H ÄNSCH as already noted in the context of precision spectroscopy and frequency
combs. The work of G LAUBER provides much of the essential theoretical back-
ground for the present chapter.
We start by describing a continuous light source, be it a laser beam whose light
is not strictly monochromatic, be it a completely chaotic light source such as an
incandescent bulb, our sun, or a fluorescent lamp. It has a finite bandwidth δωc
around a central frequency ωc , and is called quasi-monochromatic, if
δωc
ωc . (2.1)
Correlation functions have already been introduced in Sect. 1.5.2, and more details
are found in Appendix G.2, Vol. 1. Now we shall use these correlation functions
to characterize the coherence properties of electromagnetic radiation. For the field
amplitude E + (t) = (E − (t))∗ as defined by (1.36) we write2
1 The
experiment is usually referred to as “Hanbury B ROWN -T WISS Experiment”, but one should
know that “Hanbury” is a first name, and the second author’s name is “Twiss”.
2 Notethat this definition differs slightly from L OUDON (2000) (δ → −δ) who uses a somewhat
unconventional definition of the F OURIER transform.
2.1 Some Basics for Quantum Optics 73
G(1) (δ) = E − (t)E + (t + δ) (2.2)
∞
= E − (t)E + (t + δ)dt for a pulse, and (2.3)
−∞
Tav /2
1
= E − (t)E + (t + δ)dt for a CW source. (2.4)
Tav −Tav /2
The mode of averaging . . . depends on the specific case. Note that the averaging
time Tav for the CW case has to be sufficiently long, so that G(1) (δ) does no longer
change when Tav is extended. As just defined, the dimension of G(1) (δ) depends
on the case (pulse or CW). Thus it is advantageous to introduce the dimensionless
first-order degree of temporal coherence:
In general g (1) (δ) is complex and |g (1) (δ)| gives a quantitative measure of coher-
ence. It determines how far the field E + (t) and its displaced image E + (t + δ) may
be separated in time and still have a memory of each other. If they fully overlap
g (1) (0) = 1 and the light is said to be fully coherent, if they are far apart g (1) (∞) = 0
the light is incoherent.
In the case of a wave-packet (1.107), with an envelope E0 h(t) according to
(1.109), the degree of coherence with (2.2) becomes simply
∞ ! ∞
g (δ) = e
(1) iωc δ
h(t)h(t + δ)dt h2 (t)dt. (2.6)
−∞ −∞
with I0 = ε0 cE02 /2. Inserting this into (2.7) and applying the definition (2.5) of
g (1) (δ) we obtain
∞
F
I˜(ω) = g (1) (δ)e−iωδ dδ. (2.9)
2π −∞
74 2 Coherence and Photons
Thus, the intensity spectrum is given by the inverse F OURIER transform of the first-
order degree of temporal coherence of a light source. ∞
It is also useful to recall the units [I˜(ω)] = J s m−2 , while −∞ I˜(ω)dω = F
(which is easily verified from Eq. (2.9)) has indeed the unit [F ] = J m−2 .
Specifically for a Gaussian pulse with the field amplitude (1.110), one finds from
the convolution (2.6)
− 12 ( τδ )2
g (1) (δ) = eiωc δ e G . (2.10)
With this and (2.9) the intensity spectrum of a Gaussian pulse follows (see also
Appendix I.4.1 in Vol. 1):
˜ F ω − ωc 2 I0 ω − ωc 2
I (ω) = √ exp − = 2 exp − (2.11)
ωG π ωG ωG ωG
√
with F = π/2I0 τG , and ωG = 2/τG .
In contrast, for CW light the evaluation may not be that trivial, since any realistic
model will have to describe a stationary light source as a random ensemble of wave-
packets. It turns out to be more convenient to average in frequency space. To this
end, we adapt the spectral intensity distribution (2.7) appropriately, and replace the
integrals (2.3) by averages (2.4):
−Tav /2
˘ ε0 c ∞ −iωδ 1
I (ω) = e dδ E − (t)E + (t + δ)dt
4π −∞ Tav −Tav /2
∞
I
= g (1) (δ)e−iωδ dδ (2.12)
2π −∞
Tav /2
ε0 c − 1
with I = E (t)E + (t) = I (t)dt. (2.13)
2 Tav −Tav /2
We have used here the symbol I˘(ω) (unit [I˘(ω)] = W s m−2 ) for the intensity spec-
trum
∞ of the CW light, in order to indicate its different definition. From this follows
˘
−∞ I (ω)dω = I , which is now the average intensity of the stationary source, mea-
sured in units [I ] = W m−2 (rather than the fluence as in the case of a pulse).
We finally invert (2.9) and (2.12) and find the useful relations by which thefirst-
order degree of coherence can be derived in both cases (properly normalized) as
inverse F OURIER transform of the spectrum. For pulsed and CW sources we obtain
1 ∞ ˜
g (1) (δ) = I (ω)eiωδ dω and (2.14)
F −∞
1 ∞ ˘
= I (ω)eiωδ dω, respectively. (2.15)
I −∞
The next two subsections are devoted to quasi-monochromatic light beams, with
their degree of coherence g (1) (δ) and their interference properties.
2.1 Some Basics for Quantum Optics 75
We recall that the laser pulses which we have discussed in the previous chapter
have been introduced in Sect. 1.4.1 as coherent superposition of plane waves from a
limited frequency range of a FWHM ω1/2 . Such a light pulse has a finite duration
τ ∝ 1/ω1/2 . Alternatively we have described in Sect. 1.4.3 periodic pulse trains as
a F OURIER series. Obviously, neither of these two descriptions can lead to a realistic
model of a quasi-monochromatic and continuous laser beam, since such a CW laser
radiates effectively from t = −∞ to t = +∞ without obvious intermission (at least
for a couple of hours or days). With some effort and good electronics the frequency
may be stabilized for a long time to a few Hz. Still it cannot be modelled by a plane
(or Gaussian) continuous wave – nor by any kind of a wave-packet.
Such a CW light beam has to be modelled with “stationary and ergodic statistical
properties, so that ensemble averages over the probability distribution are equivalent
to long-time averages over the beams in a single experiment” (L OUDON 2000). Let
us imagine – as a simple model3 – a laser beam to be composed of a large number
of rectangular wave trains (see Appendix I.4.3 in Vol. 1) of constant amplitude but
different, finite durations. One such wave-packet is illustrated in Fig. 2.1(a) as a
function of time at a fixed position in space (without loss of generality we choose
again r = 0 ⇒ kz = 0). Thus, in our standard notation (1.36) we have
+ E0 ei(ωc t−ωc ti −φi ) for ti < t < ti + τi ,
Ei (t) = (2.16)
0 else,
and the intensity is as usual I0 = ε0 cE02 /2 in the wave-packet and zero outside.
The pulse begins at t = ti , it has a duration τi and its relative phase φi is statistically
distributed. To make things not too complicated we assume, however, that the period
Tc = 2π/ωc (or its wavelength λc ) is constant. Such a wave-packet may typically
contain 108 to 1011 periods. The spectral intensity distribution of this pulse is given
3 Similarly one has to treat any chaotic light, with large phase and intensity fluctuations, as e.g.
emitted from a collision or D OPPLER broadened gas discharge, an incandescent bulb or an en-
semble of excited atoms – even if the spectrum may be different, the bandwidth larger and the
coherence time to be introduced here correspondingly shorter.
76 2 Coherence and Photons
by (I.53), Vol. 1:
I0 τi2 τi (ω − ωc ) sin x
I˜i (ω) = sinc2 with sinc x = . (2.17)
2π 2 x
A real quasi-monochromatic light beam, which extends over large times and dis-
tances, can now be modelled by many such pulse trains as indicated in Fig. 2.1(b).
They may, of course, also overlap each other. The frequency bandwidth in a CW
laser is usually determined by mechanical and thermal instabilities of the experi-
mental setup, such as vibrations of the mirrors, collision processes in the amplifier
medium, dust particles accidentally passing the laser beam etc. These processes oc-
cur completely statistically, let us assume with a constant average rate 1/τc . We
further assume that such events after the times τ1 , τ2 , . . . , τi just change the phase
φ1 , φ2 , . . . , φi . The amplitude is kept constant. The probability that such a wave-
packet has a duration between τi and τi + dτi , is described by an exponential distri-
bution as outlined in our elementary introduction to statistics, Sect. 1.3.1 in Vol. 1:
1 −τi /τc
w(τi )dτi = e dτi . (2.18)
τc
The average time between the phase changes is τc . We call it coherence time. The
corresponding length of the wave-packet sketched in Fig. 2.1(b) is the so called
coherence length
c = τc c. (2.19)
The whole light beam is described by this statistical distribution of individual
wave-packets. Each of them is characterized by a spectral distribution according to
(2.17) and an arbitrary statistical phase φi . We emphasize again, that this continu-
ous light beam cannot be described by any kind of coherent, linear superposition of
waves. Its overall spectral distribution is found as the statistical average of the in-
dividual spectral distributions for all possible durations τi of the wave-packets and
re-normalization according to (I.32), Vol. 1. With (2.17) and (2.18) the integration
can be carried out in closed form:
∞
I0 τi (ω − ωc )
I˘(ω) = I˜i (ω) = w(τi )τi2 sinc2 dτi
2πτc 0 2
τc I I ω1/2
= = , (2.20)
π 1 + (ω − ωc ) (τc )
2 2 2π (ω − ωc ) + (ω1/2 /2)2
2
Thus, one finds a L ORENTZ profile. It is normalized here so that the integration over
all frequencies gives the local average intensity I = I0 = ε0 cE02 /2 of the laser beam
(assumed to be independent of time). The profile is characterized by the coherence
time τc . The maximum of the spectral intensity distribution (intensity per angular
2.1 Some Basics for Quantum Optics 77
Table 2.1 Coherence time τc and first-order degree of coherence g (1) (δ) for different spectral
distributions with FWHM ω1/2
Spectrum ω1/2 g (1) (δ) × e−iωc δ g (1) (τc )
Lorentziana [τc2 (ω − ωc )2 + 1]−1 2/τc exp[−|δ|/τc ] 1/e
√
Gaussianb exp[−τc2 (ω − ωc )2 ] 2 ln 2/τc exp[−(δ/τc )2 ] 1/e
Rectangle 1 for − τπc ≤ ω − ωc ≤ τπc 2π/τc sinc(πδ/τc ) 0
a Notethat this definition for the L ORENTZ profile differs slightly from (5.8), Vol. 1, used there for
spontaneous emission with a FWHM of ω1/2 = 1/τnat
b Coherence time and the usual Gaussian time are related by τ = 2τ
√
c G
frequency) at ω = ωc is
2I
I˘(ωc ) = . (2.21)
πω1/2
For the first-order degree of temporal coherence (2.5) one obtains
for the L ORENTZ profile (2.20), as can easily be verified with (2.12).
By way of example, a continuous dye laser (often used in spectroscopy) may
provide an intensity I 1 W cm−2 with a typical bandwidth of ν1/2 1 MHz.
Coherence time and coherence length are then τc 320 ns and c 100 m, respec-
tively. The peak spectral intensity is I˘(ωc ) 5 × 10−8 W cm−2 s. We may compare
this to the spectral intensity of the sun at 555 nm which according to (1.85), Vol. 1
is I˘(ωc ) 3.5 × 10−12 W s cm−2 (at the surface of the sun!).
The above description of a quasi-monochromatic light beam is just one possible
model. In principle, one has to start from a detailed analysis of a given experimen-
tal situation. A variety of wave-packets differing from those shown in Fig. 2.1 are
conceivable. In any case, g (1) (δ) and the spectral distribution I˘(ω) are related by
(2.12)–(2.15). If e.g., the radiation source is mainly D OPPLER broadened, it will
be characterized by a distribution of frequencies corresponding to a Gaussian with
statistically distributed phases. The corresponding degree of coherence will be the
same√as that derived for the Gaussian pulse (2.10) and the coherence time is then
τc = 2τG .
The definition of a coherence time τc (or the coherence length c = cτc ) must,
inevitably, be somewhat arbitrary. We shall use the time for which g (1) (τc ) = 1/e,
unless it passes through g (1) (τc ) = 0 at a finite delay time, in which case that time
is taken.
In Table 2.1 we summarize the spectra and first-order coherence properties for
three important cases of quasi-monochromatic light. Their first-order degree of co-
herence is plotted in Fig. 2.2 as a function of delay time δ. They are compared with
strictly monochromatic light (I˜(ω) ∝ δ(ω − ωc )) which – in contrast to the three
statistical light sources – shows no fluctuations at all, i.e. |g (1) (δ)| ≡ 1 holds inde-
pendent of δ.
78 2 Coherence and Photons
r2 , k light
s2
detector
parallel ray 2
f light rays
optical delay
Fig. 2.3 Very schematic layout of an interference experiment with two parallel light rays originat-
ing from a point like light source; the brace on the right just indicates that the two rays are made to
interfere – it does not sketch a light path
2.1 Some Basics for Quantum Optics 79
The prefactors a1 and a2 account for the fact the ray A and B are only a fraction of
the beam and may even be further reduced before reaching the detector. If only ray
A or only ray B were present, the signals would be
ε0 c 2 + 2
I1 = a1 E (r 1 , t1 ) = a12 I or
2
(2.23)
ε0 c 2 + 2
I2 = a2 E (r 2 , t2 ) = a22 I, respectively,
2
with I being the averaged total intensity of the original beam. For a first-order pro-
cess (N = 1) we write the averaged time dependent intensity (1.134) at the detector
as
ε0 c −
I (r, t) = E (r, t)E + (r, t) = I1 + I2 + I12 . (2.24)
2
The interference term in which we are mostly interested, is given by
I12 = C 2 E − (r 1 , t1 )E + (r 2 , t2 ) + E + (r 1 , t1 )E − (r 2 , t2 )
ε0 c
= 2C 2 Re E − (r 1 , t1 )E + (r 2 , t2 ) with C 2 = a1 a2 , (2.25)
2
s1 s2 s1 − s2
t1 = t − , t2 = t − = t1 + δ, and δ = .
c c c
The expected pattern is a function of the relative phase ωδ between rays A and B.
Since partially coherent light with a coherence time τc has a bandwidth ω 1/τc
of different frequencies, one expects the interference structure to smear out when
ωδ ≥ π , i.e. if δ ≥ πτc . The detector usually integrates over times
τc .
For a quantitative evaluation we have to keep in mind, that in any model of quasi-
monochromatic light the electric field will be described as a statistical ensemble of
many individual wave-packets Ei+ (r, t), e.g. as described by (2.16). Thus, we have
to average the interference term in (2.24) temporally – or to find the ensemble aver-
age. It turns out that this is done most conveniently in frequency space: we rewrite
the interference term (2.25) by using the (inverse) F OURIER transform (1.106), with
+ (k) being independent of the direction of k (parallel light with k = ω/c):
E
2 " #
C + ∗ i(k·r −ωt ) + −i(k ·r −ω t )
I12 = 2 dω (ω) e
dω E 1 ω e
1 E 2 2 . (2.26)
i j
2π
Obviously, only E − (ω)E + (ω ) is affected by the statistical averaging over wave-
i j
packets. We also observe that each of the wave-packets i and j carries its own
statistical phase φi or φj , respectively – as exemplified by (I.51), Vol. 1. Thus, these
complex quantities are distributed at random on a circle in the complex plane – and
hence they average out over the whole ensemble. Only those terms which are caused
by the same wave-packet i = j contribute to (2.26). Somewhat laxly one says:
We also note that E − (ω)E + (ω ) according to (I.51), Vol. 1 contains the function
i i
exp[i(ω − ω )ti ]. Averaging over the statistically distributed starting times ti (i.e.
integrating over all times ti ) lets all terms with ω = ω disappear. Thus
+ ∗ + + 2
(ω) E
E ω = 2πδij δ ω − ω E (ω) . (2.27)
i j i
I˘(ω) = I˜i (ω) is the ensemble averaged intensity spectrum. In the last step, using
(2.15), we have identified the resulting integral as the first-order degree of (longitu-
dinal) coherence and use the abbreviations (2.23).
For quasi-monochromatic light with a carrier frequency ωc the first-order degree
of coherence always assumes the form ±|g (1) (δ)| exp(iωc δ) (see the examples given
in Table 2.1). Thus, inserting (2.29) into (2.24) we obtain the interference signal
(first-order) as a function of the delay time δ:
I (δ) = I1 + I2 ± 2 I1 I2 g (1) (δ) cos ωc δ. (2.30)
We emphasize that the above derivation is characteristic for any kind of quasi-
monochromatic light composed of an ensemble of wave-packets with statistically
distributed phases. The |g (1) (δ)| is the quantitative measure for temporal coherence
we have been looking for. One calls this property temporal coherence or longitu-
dinal coherence, since coherence time and coherence length are directly related by
c = cτc : c gives the distance by which a wave-packet may be displaced from its
image so that the degree of coherence decreases to 1/e.
Some Examples In practice, interference fringes often show less contrast than
expected from (2.30) where Re[g (1) (0)] = 1. Instrumental imperfections or spatial
incoherences can be responsible, as we shall discuss Sect. 2.1.7. To quantify this
reduction of contrast one introduces a parameter
Imax − Imin
visibility, V= , (2.31)
Imax + Imin
2.1 Some Basics for Quantum Optics 81
^
S(δ ) =2π /ω c
S(∞) 1+2V √I1 I2 / (I1 + I2 )
1.2
1.0
0.8
1- 2V √I1I2 / (I1+ I2 )
0.6
-4 -3 -2 -1 0 1 2 3 π∆ ν δ
Fig. 2.4 Interference pattern for two rays with a rectangular spectrum according to (2.32). The
dotted line represents suitably processed experimental data, measured at the CHARA high resolu-
tion stellar interferometer array, extracted from TEN B RUMMELAAR et al. (2005). The full black
line is proportional to sinc(πνδ)
by which the interference term in (2.30) has to be multiplied. Visibility can be mea-
sured by registering I (δ) in a delay scan. It may contain valuable information about
the light source as we shall discuss in Sects. 2.1.7–2.1.8. In the following we as-
sume for an ideal interferometric measurement I1 = I2 and normalize the signal to
the uncorrelated limit I (∞). Finally, the expressions given in Table 2.1 have to be
inserted.
First we consider a rather broad band CW light source which is passed through
a narrow-band spectral filter, as done e.g. in stellar interferometry. If the filter has
a rectangular profile with a bandwidth ω as described in Appendix I.4.4, Vol. 1,
we obtain from (2.14) and (I.56), Vol. 1 Re[g (1) (δ)] = sinc(ωδ/2) cos ωc δ. The
normalized interference signal is thus
I (δ)
= 1 + 2V sinc(ωδ/2) cos ωc δ. (2.32)
I (∞)
Figure 2.4 shows such an interference pattern. For reference we note that the fringes
vanishes for the first time at ωδ/2 = πνδ, with ν given in frequency units. The
fringes are caused by the cos ωc δ term in (2.32) and depend on the phase difference
ωc δ = k(s1 − s2 ) between rays 1 and 2. The contrast clearly changes with delay
time and is given by V × |g (1) (δ)|. It has its maximum for δ = 0 where |g (1) (δ)| = 1,
while it disappears for long delay times. To compare the above theoretical derivation
with some real experiment, we show in Fig. 2.4 a “fringe scan” extracted from one
of the first publications of the CHARA optical/infrared interferometric array located
on Mount Wilson, CA (TEN B RUMMELAAR et al. 2005). The spectra were taken in
the K band at 2.133 µm with one of their 15 very long baseline interferometers.
The agreement with (2.32) is impressive, albeit – as expected with such simple
modelling – not perfect. We shall come back to these experiments in Sect. 2.1.8.
For the quasi-monochromatic light model introduced in the previous subsection,
the spectrum I˘(ω) is a L ORENTZ distribution (2.20). Its first-order degree of coher-
82 2 Coherence and Photons
ence g (1) (δ) is given by (2.22). Thus, the interference pattern (2.30) becomes
I (δ)
= 1 + 2V e−|δ|/τc cos ωc δ. (2.33)
I (∞)
We recall: the coherence time τc corresponds here to the average duration of the
wave-packets which define the temporal properties of the quasi-monochromatic
light.
If the light originates from a D OPPLER broadened (Gaussian) source, with (2.10)
the interference pattern (2.30) becomes
I (δ) − 1 ( δ )2
= 1 + 2V e 2 τG cos ωc δ. (2.34)
I (∞)
To extend the concept “degree of coherence” introduced in Sect. 2.1.2 one defines a
general degree of coherence of N th order as
g (N ) (r 1 , t1 , . . . r N , tN , . . . r 2N , t2N ) (2.35)
E − (r
1 , t1 ) . . . E − (r
N , tN )E + (r
N +1 , tN +1 ) . . . E + (r
2N , t2N )
= ,
[|E (r 1 , t1 )| . . . |E (r N , tN )| . . . |E (r 2N , t2N )|2 ]−1/2
+ 2 + 2 +
with |E + (r N , tN )|2 ≡ E − (r N , tN )E + (r N , tN ) ∝ I (r N , tN ).
For details the interested reader is referred to the specialized literature (see e.g.
L OUDON 2000; G LAUBER 2006). In the following we refer again to the dependence
on time t only – which may be replaced by t − rk/ω if the r is explicitly needed –
and discuss some basic aspects of the particularly important second-order degree of
temporal coherence
where the brackets . . . again imply the same kind of averaging as in (2.2)–(2.4).
The symmetry relation is now somewhat simpler than (2.5):
However, while for the first-order degree of coherence the limits 0 ≤ g (1) (δ) ≤ 1
hold, no general upper limit exists for g (2) (δ). Still, one may show that
The latter follows from the C AUCHY-S CHWARZ inequality,4 which leads to I 2 =
I (t)2 ≤ I (t)2 .
Physically g (2) (δ) represents the correlation function of the light intensity, i.e. it
answers the question whether fluctuation in the light intensity is completely random
or whether there is some kind of memory effect. In a quantum picture the photon
flux, I (t)/ω ∝ w(t) is proportional to probability for a photon to arrive (at time t)
per unit of time, and g (2) (δ) gives the probability w(t)w(t + δ) for two photons
to arrive with a specific time delay: do the photons arrive completely at random or
perhaps with an enhanced probability to come in pairs? At first thought this appears
a strange question. Why should it be more probable to find two photons at once than
at random – if the light is otherwise completely chaotic?
We cannot go into details of the statistics of chaotic light sources, but let us glance
over the key arguments. We recall the model of a chaotic light source presented in
Sect. 2.1.3 and assume the light to originate from many atoms. They all contribute
with their individual electric field Ej+ (t), each characterized by its own statistical
phase. The overall field is thus given by
E + (t) = Ei+ (t).
i
Using this expression one derives the second-order correlation function (i.e. the
nominator in Eq. (2.36))
G(2) (δ) = I (t)I (t + δ) = E − (t)E − (t + δ)E + (t)E + (t + δ) . (2.38)
Since uncompensated phases cancel out statistically in the averaging process, again
only those terms are kept, where the field from each atom is multiplied by its own
conjugate complex (at time t or t + δ). However, since now the products of four
amplitudes are involved, and a large number of atoms participates, the remaining,
dominant terms are those which arise from two pairs of atoms i = j :
−
Ei (t)Ei+ (t)Ej− (t + δ)Ej+ (t + δ) = Ii (t) Ij (t + δ) and (2.39)
−
Ei (t)Ei+ (t + δ)Ej− (t + δ)Ej+ (t) = Ei− (t)Ei+ (t + δ) Ej− (t + δ)Ej+ (t) . (2.40)
As the average intensity I¯ in a stationary source is independent of time, the first line
is simply = I¯2 . In the second line we recognize the first-order correlation function
(2.5) and its complex conjugate. Thus, in its normalized form (2.36) becomes
2
g (2) (δ) = 1 + g (1) (δ) . (2.41)
4 The C AUCHY-S CHWARZ inequality may be written in an easy to comprehend relation between
N dimensional vectors: |a · b|2 ≤ |a|2 · |b|2 . If one chooses the intensities I (tj ) as components of
a vector a and 1 as components of b, the latter relation follows immediately.
84 2 Coherence and Photons
LORENTZ
1
classical wave
(fully coherent)
-2 -1 0 1 2
δ/τc
Fig. 2.5 Second order degree of temporal for chaotic light. Compared are light sources with a
spectral distribution of L ORENTZ, G AUSS and rectangular type. All are assumed to have equal
coherence time τc , i.e. the delay time δ is given in units of the coherence time. They are contrasted
with a classical source of radiation such as a CW laser of very large coherence length or an RF
generator
Specifically, for light with a Lorentzian or Gaussian type of spectrum the second-
order degree of temporal coherence (2.36) becomes
g (2) (δ) = 1 + exp −2|δ|/τc , and (2.42)
g (2) (δ) = 1 + exp −2δ 2 /τc2 = 1 + exp −δ 2 /τG2 , respectively. (2.43)
These functions are illustrated schematically in Fig. 2.5. We recall that the correla-
tion times τc are related to the respective spectral distributions by Table 2.1.
An important limiting case is the classical continuous, constant and coherent
wave, e.g. a highly stabilized RF generator or an ideal CW laser. In that case
I (t)I (t + δ) ≡ I¯2 , there are no intensity fluctuations and
g (2) (δ) = 1.
It may sound somewhat surprising at first sight, but in a fully coherent radiation
source, such as an ideal laser, the photons are distributed as randomly as possible!
Quite generally, for long delay times δ there are no correlations in the statistical
intensity fluctuations and thus
The proposal of Hanbury B ROWN and T WISS (1954) for “A new type of interferom-
eter for use in radio astronomy” marks the beginning of quantum optics (G LAUBER
2006). In their pioneering investigations, correlations in the intensity of an extended
light source were measured for the first time – both in a table top laboratory exper-
2.1 Some Basics for Quantum Optics 85
P2 beam
splitter
P1
pulse
amplifiers 0
THC
paper-tape
-6 -4 -2 0 2 4 6
PHA computer
time delay δ / ns
Fig. 2.6 Photon bunching experiment according to P HILLIPS et al. (1967). (a) Schematic of the
experimental setup with two photo-multipliers P1 and P2, time to height converter (THC) and
pulse height analyzer (PHA). (b) Observed true two photon coincidence rate (normalized) as a
function of time delay between the two photons; light filtered with a 3 cm FPI; the maximum is
ρ(0) ∼ 17.3 %
iment (1956a, 1958) and for light from a star (1956b). In such an HBT experiment
the intensity of chaotic light is registered by two spatially separated detectors whose
signal is then correlated – in contrast to YOUNG’s double slit experiment where the
electric field amplitudes of the light are superposed. However, before we can dis-
cuss HBT type experiments, we shall have to introduce spatial or lateral coherence
in Sect. 2.1.7.
Conceptually somewhat more straight forward are so called photon “bunching”
experiments – a number of which were performed in the years following the original
HBT experiment. One with particular nice data by P HILLIPS et al. (1967) is sketched
schematically in Fig. 2.6(a). Quasi-monochromatic light from a mercury spectral
lamp passes a narrow band filter and then an FPI to select the 435.8 nm line and
reduce the bandwidth, i.e. to increase the coherence time. The light beam is then
strongly collimated by pin holes with diameters of 0.3 mm and 2 mm, separated by
1.5 m before it reaches a beam splitter which provides two branches of equal light
intensity. Two separate photo-multipliers P1 and P2 are setup to detect individual,
single photons, which are recorded after amplification and clipping as pulses with
a rise time of less than 2 ns. The time delay between these pulses is registered by
the combination of a time to pulse height analyzer (THC) and a multichannel pulse
height analyzer (PHA). Data storage and communication with a computer was at
that time still done by punched paper-tape.
The experiment thus determines coincidence rates for counting a photon in
branch 2 after a time δ when a photon has been registered in branch 1 (or vice
versa). If the individual count rate at P1 is R1 and at P2 it is R2 (in this experiment
some 104 counts/ s). The coincidence rate is R1 (t) × R2 (t + δ) × δ, in this exper-
86 2 Coherence and Photons
iment <10 s−1 with δ (here some ns) being the time resolution of the electronics.
One subtracts the statistical coincidence rate R1 × R2 × δ (corresponding to the
coincidence rate for δ → ∞). In summary, one records the true coincidence rate,
which properly normalized is
R1 (t) × R2 (t + δ) 2
ρ(δ) = − 1 ∝ g (2) (δ) − 1 = g (1) (δ) .
R1 (t) × R2 (t)
A typical result is shown in Fig. 2.6(b). The frequency bandwidth of the FABRY-
P ÉROT filter was in this case ca. ν1/2 160 MHz, in fair agreement with 208 MHz
gleaned from the correlation measurement. The maximum of the normalized true
coincidence was found to be ρ(0) ∼ 17.3 % – the authors attribute the fact that
it is not = 1 to the finite temporal resolution of the electronics, but also to finite
lateral coherence (a compromise had to be found between a reasonable count rate
and low angular divergence of the beam). But clearly, the experiment shows beyond
any doubt, that the probability to register two photons at the same time is signifi-
cantly higher than expected by purely random coincidences (observed at long delay
times) – in full agreement with the considerations outlined in Sect. 2.1.5.
Fig. 2.8 On spatial coherence: very schematic diagram of an interference experiment with slightly
diverging light rays 1 and 2 from an extended source (uniform disc angular diameter ϑd )
ϑc = ϑmin , the maximum from S2 would fully coincide with the first minimum from
S1 : hence, the fringe patterns would disappear: interference structures can only be
discerned if
λ
ϑc ≤ . (2.44)
2B
The light is said to be spatially or laterally coherent if ϑc is smaller than this limit.
Even though the assumed limit is somewhat arbitrary, clearly this spatial coherence
or incoherence will influences the fringe visibility discussed in Sect. 2.1.4 for tem-
poral (or longitudinal) coherence.
To obtain a quantitative understanding we now consider an extended, station-
ary light source of diameter D0 (= 2w0 ) which is collimated by a lens with a focal
length f and a (useable) diameter D(= 2w), as sketched in Fig. 2.8. The initial
divergence of this “beam” is given by ϑ0 ≈ D0 /f (angular diameter), quite anal-
ogous to the situation for a Gaussian beam according to Fig. 1.16, if we identify
the disc radius w0 with the beam waist and the Gaussian divergence angle θe with
ϑ0 /2. For not too large aperture angles ϑ0 , the (full) angular divergence ϑd after
collimation is
wϑd w0 ϑ0 , (2.45)
if the lens is used up to a diameter 2w. With this more realistic description of a light
beam we have to modify our treatment of the interference experiment presented in
Sect. 2.1.4.
The following derivation is completely independent of the origin of the two
slightly divergent light rays. The source-collimator arrangement (grey shaded area
in Fig. 2.8) may e.g. be replaced by a distant star that emits light with a small diver-
gence angle ϑd (“uniform disc angular diameter” equivalent to its diameter divided
by its distance). The light may be collected by two different mirrors or telescopes
placed at a distance B(= 2w). In the context of astronomical interferometry this
distance is called baseline. We shall come back to this context in Sect. 2.1.8.
88 2 Coherence and Photons
Comparison of Figs. 2.3 and 2.8 shows that we now have to treat interferences
of plane waves with wave vectors k i around the mean wave vector k. In analogy
to the averaging over frequencies, we now have to sum in addition over the contri-
butions from all k i . As before the contributions from superpositions belonging to
different k i and k j statistically average out: as before “each photon interferes only
with itself”.
The key question is now whether, and to what extent, the interference patterns
from different k i disturb each other. We start again from (1.106) and write the elec-
tric fields propagating from r 1 and being detected at time t1 = t − s1 /c as
Ei+ (r 1 , t1 ) = Ei+ (ω)ei(kr 1 −ωt1 +k i ·r 1 ) dω. (2.46)
The propagation vectors k i of the individual wave-packets is now written with ref-
erence to the central wave vector k
k i = k + k i . (2.47)
The interference term I12 ∝ Ei+ (r 1 , t1 )Ej− (r 2 , t2 ) – after summation over differ-
ent wave-packets and exploiting the “one photon interferes only with itself” rule –
becomes in analogy to (2.28)
C2 2
I12 = 2 Re dω Ei (ω) ei[k·(r 1 −r 2 )−ω(t1 −t2 )+k i ·(r 1 −r 2 )] . (2.48)
2π
The averaging . . . must include the angular divergence reflected in k i · (r 1 − r 2 ).
We write the distance vector r 1 − r 2 = r, with |r| = B, and account for the fact
that r is per definition perpendicular to k, hence k · r = 0 holds. The delay time
is again given by δ = t1 − t2 = (s1 − s2 )/c. Thus, (2.48) becomes
I12 = 2a1 a2 Re dωeiωδ I˜i (ω)eik i ·r . (2.49)
representing the polar angle θ at which the light from the source enters, while ϕ
is the azimuthal angle of k in respect of r = B in a plane perpendicular to k.
We recall that we consider a light source with a small angular diameter ϑd , e.g. a
collimated disc or a distant star, with an intensity distribution I (θ, ϕ) = I (u/kB, ϕ).
2.1 Some Basics for Quantum Optics 89
The averaging over the angular part in (2.49) (essentially over a cone with 0 ≤ θ
ϑd /2 or 0 ≤ u kBϑd /2), properly normalized, may be written
2π
ik i ·r
udu 0 I (u/kB, ϕ)eiu cos ϕ dϕ
g (1s)
(x) = Re e = Re 2π
0 I (u/kB, ϕ)ududϕ
(2.51)
I (ξ, η)eik(pξ +qη) dξ dη
= Re .
dξ dηI (ξ, η)
In the second line, the integrals are just rewritten from cylindrical coordinates u, ϕ
into a Cartesian ξ η plane perpendicular to the average wave vector k (for details see
B ORN and W OLF 2006, Chap. 10.4). In analogy to (2.29), g (1s) (x) is called degree
of spatial coherence (or spatial correlation function), and (2.51) represents the VAN
C ITTERT-Z ERNICKE theorem according to which the degree of spatial coherence
is equal to the normalized F OURIER transform of the intensity distribution of the
source.
We specialize now to an “uniform disc” model for the light source, with constant
emission I (u/kB, ϕ) for 0 ≤ θ ≤ ϑd /2, independent of ϕ and obtain
1 x 2π
g (1s) (x) = eik i ·r = udu eiu cos ϕ dϕ. (2.52)
πu2d 0 0
The prefactor 1/πu2d ensures proper normalization. The double integral here is the
same as that encountered in Sect. 1.2.2 where we have derived the diffraction pattern
from a uniform circular aperture.5 With (1.66) for n = 0 it can be expressed by the
first-order B ESSEL function J1 (x),
2J1 (x)
g (1s) (x) = with x = kBϑd /2 = πϑd B/λ (2.53)
x
as illustrated in Fig. 2.9. For x = 3.83 (2.53) reaches zero and interference structures
disappear: one says that the light is laterally coherent if
πϑd B 1.22λ
< 3.83 or ϑd < . (2.54)
λ B
This may be compared to our initial, crude estimate (2.44) – giving the right order of
magnitude. We recognize the second inequality as the famous R AYLEIGH criterium
for the angular resolution of optical instruments, if we interpret B as diameter of
the objective lens. One may also convert this into lateral resolution by setting ϑd =
w0 /f , where w0 is the smallest object that can be resolved and f the focal length
5 Note, however, that here the diameter B of the entrance pupil replaces the radius of the aper-
ture w0 there: essentially, (2.52) describes how the diffraction pattern from the source affects the
interference patterns in the experimental scheme Fig. 2.8.
90 2 Coherence and Photons
|g (1s) (x)|
coherence (2.53) as a function
of x = πBϑd /λ, with the
baseline B, the angular 0.5
diameter of the light source
3.83
ϑd and wavelength λ; note
that 3.83/π = 1.22 so that the
first diffraction minimum is
found at ϑd = 1.22λ/B 0 2 4 6 8 10
x= π B ϑd / λ
of the objective. Note that for larger opening angles ϑd , as often encountered in
optical instruments, the
The final evaluation proceeds as in Sect. 2.1.4. Thus, the overall interference
patterns given by
I (δ) = I1 + I2 + 2 I1 I2 Re g (1) (δ) g (1s) (πϑd B/λc ) cos ωc δ, (2.55)
replacing (2.30). The visibility of the interference fringes (2.31) is thus determined
by |g (1s) |. For I1 = I2 we obtain at δ = 0:
Imax − Imin (1s) 2J1 (x)
V= = g (x) = with x = πϑd B/λc . (2.56)
Imax + Imin x
We recall that our derivation is for a circular disc light source with a uniform angular
diameter ϑd , such as a distant star. A systematic measurement of the visibility V as
a function of baseline B at well defined wavelengths λc thus allows one to extract
ϑd as will be discussed in the next subsection.
To summarize: the interference structure is lost not only for long delay times
|δ|
τc – as a consequence of an optical path difference larger than the coherence
length, |s1 − s2 |
c . It also disappears for a light beam with too large lateral
extension or to large angular divergence (2.54). This can be rewritten for a light
beam of a half divergence angle θe = ϑd /2, a radius w = B/2 (see Fig. 2.8) and
with 3.83 4:
λ 2
w< := wcoh . (2.57)
πθe θe k
Light is considered coherent if the left inequality holds. We have defined here (some-
what arbitrary) wcoh , a spatial (lateral) coherence radius of a light source. Corre-
spondingly, for a source of radius w we call θe = λ/πw the coherence angle.
This description implies that all wave-packets originating from a cross section
πwcoh2 are considered coherent: their respective interference patterns do not dis-
2.1 Some Basics for Quantum Optics 91
λ2 λ2
Acoh = πwcoh
2
= = (2.58)
πθe2 δΩe
is the coherence area of a light source, where δΩe = πθe2 is the solid divergence
angle of the beam. Correspondingly, for a given width w of a source, we call
δΩe = λ2 /πw 2 the coherence (solid) angle. We finally combine the lateral coher-
ence area (2.58) with the longitudinal coherence length according to (2.20) – slightly
arbitrarily and for a Lorentzian frequency distribution – and define a coherence vol-
ume
4cλ2
Vcoh = Acoh 2c = . (2.59)
ω1/2 δΩe
Photons are considered as coherent if they originate from a cylindrical volume ex-
tending from +c to −c in k direction around the center of the beam with of radius
wcoh (beam waist) with a solid divergence angle δΩe .
For a “beam” of light derived from a chaotic (or natural) source, these consid-
erations just imply that phase fluctuations within the so defined coherence volume
are small enough so that interference structures are not disturbed significantly. For
a freely propagating, stationary laser beam the definition of a coherence volume
comes even more naturally: Let the radial profile of the beam be Gaussian, and
the frequency profile Lorentzian. The lateral coherence radius is identified as the
beam waist w0 (we recall that according to Table 1.1 86 % of the total power flows
through the corresponding cross section). On the other hand, the relation of w0 to
the divergence angle (1.55) is identical to that of wcoh according to (2.57). And the
(longitudinal) coherence lengths Table 2.1 is the same as just assumed. In summary,
expressions (2.57)–(2.59) also hold for a Gaussian laser beam with a Lorentzian
spectral profile.
A direct application of the concept of spatial coherence just developed, is the lateral
or angular characterization of extended light sources emitting at far distances: this
is exploited by astronomical interferometry which dates back to M ICHELSON and
P EASE (1921) who mounted a steel beam of initially 6 m length with four mirrors on
top of a 2.5 m diameter telescope on mount Wilson, California, in order to determine
the lateral degree of coherence of stellar light.
The scheme is sketched in Fig. 2.10. If one changes the distance B (the so called
baseline) between the two light receiving mirrors M1 and M2 , according to (2.54)
interference is only observable for ϑd ≤ 1.22λ/B. This allows one to determine the
angular diameter ϑd at which the object studied is seen (e.g. a disc like star, double
stars). If the distance of the star is known, one may thus determine its diameter.
The resolving power of such an astronomical interferometer depends on the
fringe spatial frequency B/λ: the larger it is, the smaller divergence angles can
92 2 Coherence and Photons
M
2
telescope mirror
so rom
ce
with delay
from telescope 1
from telescope 2
from
ur
f
source
tele-
scope 1
pr a s e B
l
l a rica
oj l i
b
ec n e
so rom
ce
d e et
te
y
om
ur
f
d
ge
single beam
tele- tele- pixel splitter
scope 1 scope 2 detectors
baseline b +-
amplifier
δ
beam combination variable detector array
unit delay line delay scan
Fig. 2.11 Schematic of modern astronomical interferometers adapted from M ONNIER (2003).
(a) Overall layout with the telescopes, the beam guiding optics, delay line and beam combina-
tion. Two types of beam combination schemes are shown: (b) image plane interference (similar
to YOUNG’s double slit setup), and (c) pupil plane where the collimated beams are brought to
interference by a beam splitter
1.0
0.8
visibility
0.6 HD146233
0.4
0.2
HD4614
0.0
80 100 120 140 160
B λ-1 / 10 6 rad -1
Fig. 2.12 Examples of visibilities for two stellar objects as determined by B OYAJIAN et al. (2012)
at the CHARA interferometric array. The (red) data points were taken at a series of values (base-
line/wavelength) = Bλ−1 . Note that the full black curves are not straight lines; rather they are
fits by functions essentially of the type (2.56), from which the disc diameter angle is derived:
ϑLD = 1.623 ± 0.004 mas for HD4614 and ϑLD = 0.780 ± 0.017 mas for HD146233
angular diameters of the stars, here exemplified for ϑLD = 1.623 ± 0.004 mas and
ϑLD = 0.780 ± 0.017 mas.7
We cannot close this topic without at least mentioning radio-frequency interfer-
ometry. Radio astronomy is a very powerful and highly developed area of modern
astronomy, with hundreds of modern facilities worldwide, operating at wavelengths
between 1.3 mm and several metres. Baselines of radio-frequency interferometers
must be much larger than optical or infrared interferometers to allow detection of the
same angular diameters ϑd λ/B. However, radio-frequencies have the great ad-
vantage that amplitudes and phases can be recorded directly, while at optical wave-
lengths typically only cycle averaged intensities can be detected. Hence, amplitudes
have to be superposed locally in a beam combining unit to record interference pat-
terns.
In contrast, radio frequency interferometry correlates the amplitudes electroni-
cally, and no local superposition on the detector is needed. The signals (amplitudes
and phases) may be collected anywhere in the world and be brought to “interfere”
later on by a mathematical algorithm in a powerful computer.
This concept is realized in very long baseline (radio) interferometry networks,
e.g. in the global mm-VLBI array in which several dozens of the most powerful
radio telescopes of the world co-operate – including very large single dishes such
as the 100 m diameter telescope at Effelsberg (Germany) and the 305 m telescope
at Arecibo (Puerto Rico), as well as a number of large radio telescope arrays. All
what needs to be done is to record simultaneously the electric field of a particular
frequency and direction received from space, store it on a tape, and provide that
with an accurate time marker – based essentially on synchronized atomic clocks
(or masers). With baselines of more than 10000 km an angular resolutions of about
7 While (2.56) is exact for a uniform disc (UD), astronomical models also account for limb-
One may determine the degree of lateral coherence also by measuring the second-
order correlation function, i.e. by recording the intensity correlations and exploiting
(2.41). Hanbury B ROWN and T WISS have suggested for the first time such an ex-
periment in 1954. They tested it in a laboratory setup with a spectral lamp (1956a,
1958) and performed the first successful astronomical measurement determining the
angular diameter of Sirius (1956b) based on a measurement of intensity correlation.
In principle, such kind of measurement is much more flexible than the interfer-
ometry just described – one simply has to record intensities at two detectors, sep-
arated by a baseline B, and to determine the correlation g (2) (B) ∝ I (r 1 )I (r 2 ) =
I (r 1 )I (r 1 + B) between these signals according to (2.36). With (2.41) one derives
the first-order degree of coherence g (1) (πϑd B/λ), the same quantity as measured by
an interferometer. But with such technique there is no need for a highly stable setup,
precisely adjusted to a fraction of a wavelength over long distances, and even the
telescopes do not require high quality as long as one can resolve the object stud-
ied. The baseline B can easily be varied and may, in principle, be chosen very long
as the signal can be registered at widely separated locations. The setup originally
used by B ROWN and T WISS (1956b) is shown in Fig. 2.13(a). They actually used
two standard search light mirrors of 1.56 m diameter as telescopes. The normalized
second-order correlation function is recorded as a function of the projected base-
line distance B. One expects a signal corresponding to Fig. 2.5, convoluted with
the experimental resolution. As an example, in Fig. 2.13(b) the normalized signal
g (2) (B) − 1 is plotted for the star Sirius, which was the test object of B ROWN and
T WISS (1956b). From the fit shown in the graph an angular diameter of 63 mas was
determined.
Hanbury B ROWN continued a successful carrier as a radio astronomer, but still
made several contributions to measuring stellar diameters based on his method in
the optical spectral region. However, according to DAVIS and L OVELL (2003), “with
rapid improvements in the technology of the phase-correlation interferometer, Han-
bury’s intensity interferometer did not survive as a technique for the measurement
of the angular sizes of radio sources. As Hanbury later remarked, he had spent two
years ‘building a steamroller to crack a nut’.”
96 2 Coherence and Photons
0.5
b
telescope (1) telescope (2)
× δ
0
amplifier (1) amplifier (2) 0 5 10
integrator
& detector projected baseline B / m
Fig. 2.13 Hanbury B ROWN -T WISS stellar interferometer. (a) Experimental setup according to
B ROWN and T WISS (1956b). The two “telescopes” where standard searchlight mirrors of 1.56 m
diameter and the baseline was varied up to B 10 m. The detector also acts as integrator. (b) Com-
parison of experiment and theory for a measurement of the angular diameter of Sirius determining
the angular diameter to be 63 mas; adapted from B ROWN and T WISS (1956b)
P2
ϑd /2
beam
spliter
variable
delay δ
P1 P2 P1
Fig. 2.14 Two varieties of the basic concept for Hanbury B ROWN -T WISS type experiments. In
each case the signals from the two detectors P1 and P2 are correlated (see text). (a) Radiation
(photons or other particles) originate from a point like source with limited temporal coherence; the
total flux is split into two equal branches, one of which can be delayed by a variable time δ. (b) The
two branches originate from an extended source with limited lateral coherence, corresponding to a
phase difference πϑd B/λ
To fully appreciate the impact that the HBT concept and its realization had, as
a starting point of quantum optics, let us look again at its essential ingredients.
Figure 2.14(a) gives a highly simplified schematic of the photon bunching exper-
iment (temporal/longitudinal coherence) introduced in Sect. 2.1.6. The incoming
2.1 Some Basics for Quantum Optics 97
light, highly collimated from a point like source, is split into two equal branches and
is detected by two photo-multipliers. One records the probability for detecting one
photon at detector (P1) and another photon at the other detector (P2) – with some
time-delay δ, corresponding to a phase difference ωδ. This time delay between pairs
of photons is measured electronically (see e.g. Fig. 2.6). Figure 2.14(b) shows the
scheme of the original HBT experiment with an extended light source. It differs
from (a) by the fact that now the lateral extension of the source (angular diameter
ϑd ) creates a phase difference πϑd B/λ between the two detectors as explicated in
Sect. 2.1.7. In this case, the baseline B is varied.
In both cases one measures the second-order correlation function (2.38). And in
either case one finds (in an ideal experiment) for statistical light sources that the
correlated signal at δ = 0 (or at b = 0) is twice that for δ → ∞ (or for b → ∞,
respectively) – provided the detectors are sufficiently fast, i.e. their response time is
much shorter that the temporal coherence time of the source (for that purpose the
light is passed through a narrow band pass filter prior to detection). In contrast, a
fully classical source, such as an ideal, intense CW laser beam, shows no enhanced
second-order correlation at any time – the photons are distributed completely at
random as we shall discuss in Sect. 2.2.
As recently pointed out by K LEPPNER (2008), the Hanbury B ROWN -T WISS ef-
fect is one of those rare occasions where a classical explanation is quite straight
forward, while at first sight it appears to contradict intuition from a quantum point
of view: As we have seen in Sect. 2.1.7 the HBT effect arises essentially from the
statistical fluctuations of the amplitudes of the radiation.8 Hanbury B ROWN actually
started as a radio engineer and was quite familiar with noisy signals. The mathemati-
cian T WISS helped him to work out the theory for his experiment on a fully classical
basis.
However, from a quantum point of view the experiment was completely puzzling
and started a vivid and controversial discussion: a photon is either here or there.
Why should the probability of finding one simultaneously at each of the detectors
(g (2) (δ) = 2) be higher than the statistical probability for random coincidence? But
the experiment shows, even at low count rates, that if a photon is registered at (P1)
the probability to register at the same time a photon at (P2) is twice that (g (2) (δ) = 2)
for purely random arrival of completely uncorrelated beams. The answer to this
puzzle is quite simple: photons are bosons, so they may occupy the same phase
space – and in a chaotic sources they have indeed a clear tendency to bunch, rather
then to occupy all modes equally. Consequently, for electrons and other fermions
one may expect the opposite: anti-bunching as we shall see in the next subsection.
8 We also recall that the second-order degree of coherence (autocorrelation function of the intensity)
is efficiently used for measuring the duration of femtosecond pulses (Sect. 1.5). There, nonlinear
processes such as SHG are applied to detect a signal which is proportional to the square of the
intensity (compare Table 1.5, for N = 2 with (2.43)). Note, however, that the results differ in the
prefactor of the exponential (2 vs. 1), owing to the fully coherent nature of the laser pulse vs. the
chaotic light source assumed here.
98 2 Coherence and Photons
G LAUBER developed the quantum theory of light which also explains the HBT ef-
fect, consistent with the particle nature of photons (a summary of his work is avail-
able as G LAUBER 2007). But intensity interferometry has in the mean time suc-
cessfully been adapted for other particles, exploiting the advances with fast imaging
detector arrays. In high energy heavy ion and particle collisions, two particle corre-
lations between protons, pions, or even photons again, are studied to obtain infor-
mation on the “space-time geometry” of such collisions (BAYM 1998). For fermions
one expects and observes anti-bunching (H ENNY et al. 1999; H ASSELBACH 2010).
Fermions cannot occupy the same phase space, they avoid each other and this can
indeed be observed experimentally.
In this context it is appropriate to mention that even in photon correlations one
may encounter situations where anti-bunching is observed (K IMBLE et al. 1977): if
a single atom fluoresces while being excited by a not too intense radiation field, this
atom will have zero probability for emitting a second photon immediately after it
has just decayed from its excited state into its ground state. Even more decisive is
the experiment of G RANGIER et al. (1986). They prepared a genuine single photon
source: photons emerging from an atomic cascade are detected only when triggered
by the first photon in the cascade. As expected, they observe strong anti-correlation
between the triggered detection on both sides of a beam splitter. We shall come back
to further experiments of this type in Chap. 10.
A relatively new field for fascinating applications of the HBT effect appears
to be – quite unexpectedly – the physics of ultracold quantum gases, where com-
plex phases and structures are revealed by such experiments. We cannot go into de-
tails here. We show, however, one particularly neat experiment on ultracold helium,
which bears out the difference between fermions and bosons.
J ELTES et al. (2007) prepared in a magnetic trap ultracold, metastable 3 He∗ or
alternatively 4 He∗ at 0.5 µK. The trapped samples were approximately Gaussian
ellipsoids of 110 × 12 × 12 µm3 size. The atoms are released from the trap by turning
off the magnetic field – the atoms fall under the influence of gravity and the cloud
expands. They are detected by a position-sensitive detector (micro-channel plate and
delay-line anode) that detects single atoms. The single atom signal simply reflects
the overall shape of the (expanded) cloud.
However, the two particle coincidences allow in principle to determine a full 3D
second-order correlation function g (2) (x, y, z) of the particle positions – analogous
to the (one dimensional) schematic shown in Fig. 2.14(b) for photons. The best
resolution is obtained in z-direction (determined by the arrival time of the atoms at
the detector). The results for g (2) (0, 0, z) shown in Fig. 2.15 give a very clear and
impressive picture of anti-bunching in 3 He∗ (fermions) and of bunching for 4 He∗
(bosons).
Section summary
• The intensity spectrum of a light pulse is given by the (inverse) F OURIER
transform of the first-order degree of temporal coherence of the field ampli-
tude.
2.1 Some Basics for Quantum Optics 99
g (2)(∆z)
atoms originate from the
same position in space
(z = 0), very clear 1.00
correlation
anti-bunching or bunching is
observed for fermions and
3He*
bosons, respectively. Adapted
from J ELTES et al. (2007) 0.95
0 1 2 3
atom separation ∆z / mm
associated with the particle photon. The photon travels (in vacuum) with the speed of
light c and has no rest mass. We may attribute to it a relativistic mass mph = ω/c2 .
The momentum of the photon, also known from experiment (C OMPTON effect) is
h
p ph = k with pph = = hν̄. (2.61)
λ
Finally, photons have an intrinsic angular momentum, the photon spin S, with a spin
quantum number S = 1. This too is based on experimental evidence (B ETH 1936),
as reported in Sect. 4.1.4, Vol. 1.
Photon states |k, e may be characterized by the photon’s propagation vector
k and its polarization e according to Sect. 1.3. One may introduce a photon spin
operator S and its components, in particularly Sz and express the photon states in
the helicity basis |eq , where the z-axis is chosen parallel to k. The states |e± refer
to circularly polarized light, and the usual angular momentum algebra applies:
2
S |eq = 2 S(S + 1)|eq with S = 1, (2.62)
9 Actually, a similar scheme can be applied to the electronic states of atoms. But that scheme is
much simpler since these states can only be occupied or not be occupied.
102 2 Coherence and Photons
Sz |eq = q|eq with q = ±1 and (2.63)
eq |eq = δqq . (2.64)
Sz2 |ex = 2 |ex and
Sz2 |ey = 2 |ey (2.66)
We emphasize that (2.63) includes only two states, with q = ±1, i.e. angular mo-
mentum components ± in z-direction. Even though in conventional angular mo-
mentum algebra (2.62) and (2.63) would formally define three substates (with q = 0,
±1), the particle “photon” exists only with the two angular projections, q = ±1.
This somewhat unusual behaviour reflects the transverse nature of the polarization
of light and the fact that photons do not have a rest mass, i.e. always propagate with
the speed of light. Classically we had associated the spherical basis vectors with
three oscillators: two of them oscillating in the xy plane (q = ±1) while the third
oscillates along the z-axis. The corresponding radiation characteristics are described
in Chap. 4, Vol. 1. Somewhat loosely one might say that the photon state with q = 0
does not propagate along the z-axis.
The photon states |k, eq discussed above correspond to a single photon with polar-
ization eq , wave vector k, and frequency ω = k/c. A realistic light beam consists
of many photons and a range of wave vectors. As a complete basis set for con-
structing any “photon wave function” one could e.g. take all plane waves with all
possible values of k – as we have already shown in the previous Sect. 2.1. A quasi-
monochromatic, stationary light beam with a mean wave vector k c would contain a
narrow range of angular frequencies δω around ωc (or δk = δω/c around the mag-
nitude of the wave vector kc ) and have an angular distribution of wave vectors in an
angular range δθ (or solid angle δΩ = πδθ 2 , respectively).
Of course, photons are in principle unbound particles, so that we would have to
deal with an infinite number of basis states and an infinite number of energies. To
avoid these complications, one usually switches to a very large but finite normal-
ization volume (in real position space), say a cubic cavity of edge length L, and
2.2 Photons, Photon States, and Radiation Modes 103
perfectly conducting walls where the electric field must vanish. Mathematically this
is equivalent to introducing periodic boundary conditions10
2π 2π 2π
kx = mx , ky = my , kz = mz (2.68)
L L L
with mx , my , mz = 0, 1, 2, 3, . . . .
Thus, a countable number of bound states emerges, called modes of the radiation
field, to each of which we attribute a photon state |k, e.
The whole k space is thus divided into very small but finite cells as sketched in
Fig. 2.16. The size of the cells is determined by kx = ky = kz = 2π/L, so that
3
2π
k = kx ky kz =
3
. (2.69)
L
10 The treatment given here is quite analogue to that for electrons in a 3D box, presented in
Sect. 2.4.2, Vol. 1 – except that there fermions (spin 1/2) were described and each cell in k space
was only filled by at most two electrons (of opposite spin).
104 2 Coherence and Photons
It is important to point out here that in the preceding paragraph one key assump-
tion has been made which is crucial (albeit plausible) for the following considera-
tions: periodic boundary conditions in k space (2.68) for the electromagnetic field. –
One may, of course, also turn the arguments around and define (2.71) as the funda-
mental theorem: the minimal cell size in phase space is h3 . This, together with the
well defined energy ω of a photon, may be seen as the key paradigm beyond the
quantization of the electromagnetic wave field. It will turn out to have decisive con-
sequences, e.g. in the context of spontaneous emission.
We note that the size of the box, L3 , which is our reference volume, does not
necessarily refer to a real physical situation. Usually it is just a mathematical con-
struct introduced to avoid an infinite number of photon states with which one would
otherwise have to deal, and one simply has to choose L just large enough so that the
grid is sufficiently fine for describing the properties of the radiation field applied.
On the other hand, there are situations where the normalization volume really
refers to a genuine physical geometry, e.g. to a laser resonator or any type of optical
cavity in which light may be confined. The genuine modes of this cavity will have
to be used if one wants to describe an experimental situation quantitatively. Laser
theory is one such application. Another field is the so called “cavity QED” which
we shall touch briefly in Sect. 2.3.7. If the size of the cavity becomes comparable to
the wavelength of the radiation studied one finds that even spontaneous emission is
substantially modified.
Later on we shall need an expression for the number of modes in a specified range
of k vectors with a given polarization. This can now easily be derived. The number
of modes dmke between k = (kx , ky , kz ) and k + dk = (kx + dkx , ky + dky , kz + dkz )
is obtained by dividing the volume element in k space dkx dky dkz by the size of the
unit cell (2.71):
L3
dmωe = ω2 dωdΩ = ρ(ω, e)dωdΩ. (2.73)
(2πc)3
The values dmke and dmωe give the number of modes with polarization e propagat-
ing into a solid angle dΩk and with wave vectors between k and k + dk, or angular
frequencies between ω and ω + dω, respectively. The expressions
dmke k2
ρ(k, e) = = L3 , (2.74)
dkdΩ (2π)3
dmωe ω2
ρ(ω, e) = = L3 , and (2.75)
dωdΩ (2πc)3
dmνe ν2
ρ(ν, e) = = L3 3 (2.76)
dνdΩ c
2.2 Photons, Photon States, and Radiation Modes 105
are called mode density. The mode density obviously depends on the square of the
wavenumber or frequency.
After having identified the radiation field as a discrete and countable set of
modes, as a final step one extends the normalization volume L3 to values so large
(essentially to infinite) that the usual continuous spectrum is effectively recovered,
i.e. L is chosen large enough to obtain a sufficiently fine mesh kx,y,z = 2π/L in k
space to describe the problem at hand to any degree of accuracy needed. This allows
one to finally replace all necessary summations over spectral modes by an integra-
tion over the solid angle Ω and k (or ω). With the mode densities (2.74) and (2.75)
just derived we may thus write symbolically
L3 L3
... → . . . k 2 dkdΩ = . . . ω2 dωdΩ. (2.77)
(2π)3 k,Ω (2πc)3 k,Ω
k
As far as the radiation field is spatially isotropic one may carry out the angular
integration and obtains
L3 L3
... → . . . k 2 dk = . . . ω2 dω (2.78)
2π 2 k 2π 2 c3 ω
k
We take here a little detour back to black body radiation. Dividing ρ(ν, e) given
in (2.74) by L3 , and multiplying it by 8π (integration over the full solid angle and
summation over the two polarization directions) leads to the density of states (per
volume) as introduced in Sect. 1.3.4, Vol. 1. For photons one usually refers to fre-
quency space:
8π 2
g(ν)dν = ν dν. (2.79)
c3
Inserting this into the B OSE -E INSTEIN distribution (1.63), Vol. 1 for a black body
radiator in thermal equilibrium, we obtain the spectral photon density:
8πν 2 dν
N(ν)dν = 3 . (2.80)
c exp(hν/kB T ) − 1
106 2 Coherence and Photons
The chemical-potential of the massless particle photon has been set here m̄e = 0 –
it takes no energy to split or unite photons in statistical interactions (e.g. with the
surrounding walls) as long as the total energy remains constant. Note that the ν 2
factor in the nominator prevents divergence for hν → 0 (i.e. avoids the so called in-
frared catastrophe). Integration of (2.80) over all (positive) frequencies gives a finite
value for the photon density in the black body, N = 16(kB T )3 π ζ (3)/(hc)3 , with
the R IEMANN function ζ (x). N amounts to about 20 photons/cm3 at 1 K. We recall
now that P LANCK’s law describes the spectral energy density of the photons, i.e. it
is obtained from (2.80) by multiplication with the photon energy hν. Comparison
with (1.81), Vol. 1 shows that we have indeed derived P LANCK’s law.
L3 I L3
Ne = u = . (2.81)
ω c ω
More specific, in an interval ω to ω + dω of angular frequencies we find
L3 I˜(ω) L3
Ne (ω)dω = ũ(ω) dω = dω (2.82)
ω c ω
photons, with ũ(ω) = I˜(ω)/c being the spectral radiation density and I˜(ω) intensity
spectrum (per unit angular frequency). Considering the finite divergence angle δΩ
of a light beam, the number of photons with polarization e in a frequency interval
dω per solid angle dΩ is
I˜(ω) L3
N (ω, Ω; e)dωdΩ = dωdΩ. (2.83)
δΩ cω
Finally, we recall dmωe , the number of modes (2.73) in a range dωdΩ of frequencies
and solid angles. With this we obtain the number of photons per mode:
We thus have worked out a relation between the quantum mechanically relevant
number of photons per mode and the measurable intensity per solid angle and an-
gular frequency. To be even more specific: with (2.21) for the maximum I˜(ωc ) of a
Lorentzian spectral distribution (FWHM = ω1/2 = 2/τc ) we obtain
2I λ3 I
Nke = = 2τc λ2 . (2.85)
ω1/2 δΩ πc ωδΩ
I
Nke = Vcoh . (2.86)
cω
As I /(cω) is the photon number density, this relation can be read as: the number
Nke of photons per mode is equivalent the number of photons in the coherence vol-
ume of the beam. We recall: for a laser beam Vcoh is simply its geometrical waist
cross section πw 2 (at 1/e2 width) multiplied by 2c = 2cτc .
A slightly different situation is encountered for a chaotic radiation source. Let it
have a small but finite diameter d = 2w and radiate with a total power P isotrop-
ically into the full solid angle δΩ = 4π . Coherent emission from its effective area
πw 2 occurs into a solid angle δΩe = πθe2 = λ2 /πw 2 , with an intensity I = P /πw 2 .
Thus, (2.85) may be written
P 4λ2 P δΩe
Nke = = 2τc .
πw 2 ω ω1/2 4π ω 4π
The second equality states that the number of photons per mode is equivalent to the
number of photons emitted coherently (i.e. into a solid angle δΩe ) during twice the
coherence time.
Table 2.2 summarizes characteristic parameters for some typical radiation
sources: total power, lateral extension, wavelength, bandwidth and relative band-
width. From these one calculates coherence (half) angle, coherence time and coher-
ence lengths according to Table 2.1 as well as the rate of coherent photon emission
Pcoh /ω and the number of photons per mode Nke according to (2.85).
Among the sources compared are two essentially chaotic ones (spectral lamp and
atoms at rest) and three quasi-monochromatic sources with rather long coherence
times. The “spectral lamp” could be a typical, commercially available device, here
108 2 Coherence and Photons
Table 2.2 Characteristic parameters of five typical light sources: Total power of emitted light P ,
beam waist or source radius w, wavelength λ, FWHM of the spectral distribution δλ1/2 and δν1/2 ,
coherence (half) angle δθe , coherence time τc , coherence length c , coherently emitted photon rate
Pcoh /ω, number of photons per mode Nke
Source P/W w/ mm λ λ/ nm ν1/2 ν1/2 /ν
light (total) = λ/λ
spectral lamp 0.5 5 590 nm 10−3 860 MHz 1.7 × 10−6
atoms at rest 10−6 0.05 780 nm 1.2 × 10−5 5.9 MHz 1.5 × 10−8
CW dye laser 1 0.5 590 nm 1.2 × 10−6 1 MHz 2 × 10−9
TiSa laser pulse 2.0 × 1010 0.1 800 nm 19 8.8 × 103 GHz 2.3 × 10−2
microwave oscillator 103 100 3 cm 0.3 100 Hz 1 × 10−8
derived from the above parameters:
Source δθe / rad τc / s−1 c / m Pcoh /ω s−1 Nke
= λ/πw = 1/πν1/2 = c × τc Photons/ s
spectral lamp 3.8 × 10−5 3.7 × 10−10 0.11 5.3 × 108 0.4
atoms at rest 5 × 10−3 5.4 × 10−8 16 2.3 × 107 2.5
CW dye laser 3.8 × 10−4 3.8 × 10−7 95 3 × 1018 1.9 × 1012
TiSa laser pulse 2.6 × 10−3 3.6 × 10−14 1 × 10−5 8 × 1028 5.8 × 1015
microwave oscillator 0.1 3.2 × 10−3 106 1.5 × 1026 1024
emitting at the Na wavelength. The “atoms at rest” might e.g. be a B OSE -E INSTEIN
condensate, assuming 105 excited 87 Rb atoms to emit at the 780 nm, with the natu-
ral width of this resonance line. These two sources are assumed to emit isotropically
into the full solid angle 4π . The (half) angle θe indicates the maximum angle within
which the light can be considered as coherent. The other sources are highly direc-
tional and are assumed spatially coherent over their full cross section. The dye laser
is operating CW in the yellow spectral range, with reasonable stabilization and a
bandwidth as often used in spectroscopy. The pulsed source (ca. 1 mJ with a tempo-
ral FWHM of 50 fs at 800 nm) represents a standard femtosecond Titanium-sapphire
laser setup, with a beam focused moderately to w = 100 µm. We assume the whole
pulse to represent one mode of radiation – due to the short pulse duration with a
rather broad bandwidth. Finally, we also compare with a classical radiation source,
a microwave oscillator.
The characteristic quantity Nke derived here, the number of photons per mode,
gives of course an average value if many modes are needed to describe the spectrum
and the angular profile of a source. Note that these sources represent rather different
types of radiation: For the spectral lamp Nke is very small so that most modes do not
contain any photon at all; the atomic source shows already a significant probability
to find one or even more photons per mode; the highly coherent laser sources as well
as the microwave source contain a very large number of photons per mode, and the
field can be considered as essentially classic.
2.2 Photons, Photon States, and Radiation Modes 109
We may now explicitly write down the field variables of a multi-mode electromag-
netic radiation field, i.e. its vector potential A(r, t), and its electric E(r, t) and mag-
netic field B(r, t). They are related to each other as described in Appendix H.1.1,
Vol. 1. We focus again on the electric field vector. Following (1.35) we write now
i − +
E(r, t) = eq Ekq (t)eikr − e∗q Ekq (t)e−ikr (2.87)
2
kq
− − −iωt + + iωk t
with Ekq (t) = Ekq e and Ekq (t) = Ekq e . (2.88)
The summation has to be carried out over all occupied field modes. In a classical
description Ekq is the field amplitude in each mode, with wave vector k and po-
larization q. As discussed previously, random phase fluctuations will have to be
included if one wants to describe a stationary light beam.
We now have to make the translation to quantum mechanics, i.e. we are looking
for the field operator. A good starting point is the total energy W stored in the
electromagnetic field. By inserting (2.87) into (1.86) we obtain the intensity I and
the energy density u = I /c. Integration over the whole normalization volume L3
eventually leads to
2
W = L3 ε0 cE(r, t) (2.89)
ε0 − + ε0 − +
= L3 Ekq (t)Ekq (t) = L3 Ekq Ekq . (2.90)
2 2
kq kq
Section summary
• In this section several conceptual steps were taken to familiarize ourselves
with the notion of photon states, and to prepare the quantization of the elec-
tromagnetic field.
• After recalling the quantum properties of photons, we introduced modes of an
electromagnetic field by assuming a large but finite normalization volume L3
and demanding periodic boundary conditions. As a consequence we find that
3 pL3 = h3 is the smallest size of a phase space cell.
110 2 Coherence and Photons
So far we have not really quantized the field yet. In order to do so, one needs some
quantum mechanical tools: the matrix formulation of the harmonic oscillator and
second quantization. The latter is a clever method of book keeping for the population
of states with particles – here of photon states with photons.
We start with the total energy (2.89) of the electromagnetic field, and consider
one single mode k, q populated. The field energy in this mode is
ε0 − + ε0 − + + + iωk t
Wkq = L3 E (t)Ekq (t) = L3 Ekq Ekq with Ekq (t) = Ekq e . (2.91)
2 kq 2
In the following we drop the indices kq for simplicity of writing, and introduce new
variables:
L3 ε0 − +
L3 ε0 −
Q= E (t) + E (t) and P = −i E (t) − E + (t) . (2.92)
2ω 2
The inverse relations are
− 1 + 1
E (t) = 3
(ωQ + iP ) and E (t) = (ωQ − iP ). (2.93)
L ε0 L3 ε0
1 2
W= P + ω2 Q2 . (2.94)
2
2.3 Field Quantization and Optical Transitions 111
px2 mω2 2
W= + x .
2m 2
The key idea is now, to identify the oscillations of the electromagnetic radiation
field with the harmonic oscillator, and use the rules sketched in Chap. 2, Vol. 1
to translate the field modes into quantum mechanics. For this, Q and P must be
canonical conjugate coordinates. With (2.92) we find
Ṗ = −ω2 Q and Q̇ = P ,
∂W ∂W
= ω2 Q = −Ṗ and = P = Q̇.
∂Q ∂P
This set of equations are the classical H AMILTON equations in one dimension with
Q and P being indeed canonical conjugates. Thus, W := H represents the Hamil-
tonian of the electromagnetic field!
What follows is the decisive step in the quantization process: canonical conju-
gates are replaced by operators which obey the commutation rule
P
[Q, ] = i. (2.95)
We
now rewrite the relations (2.93) in dimensionless form by multiplying them with
L3 ε0 /(2ω):
1 + i P) 1 − iP
).
â = √ (ωQ and â + = √ (ωQ (2.96)
2ω 2ω
With (2.95) one verifies that these operators obey the simple commutation rule
â, â + = â â + − â + â = 1, (2.97)
+ 1.
â â + = â + â + 1 = N (2.98)
= â + â.
N (2.99)
The Hamiltonian of the electromagnetic field (2.94) takes now the form
112 2 Coherence and Photons
F = ω â + â + â â + = ω â + â + 1
H (2.100)
2 2
= ω N + 1 . (2.101)
2
The derivation of the algebra for the operators N, â and â + is straight forward,
based on the commutation rule (2.97). It can be found in all quantum mechanics
text books. We thus only summarize here the results. The number operator N is
+
Hermitian (while â and â are not). It has eigenstates |N with integer numbers N
as eigenvalues:
|N = N |N
N (2.102)
with N = 0, 1, 2, . . . and N | N = δN N . (2.103)
F |N = WN |N
H (2.104)
where WN = (N + 1/2)ω for N = 0, 1, 2, . . . . (2.105)
We recognize the well known eigenenergies of the harmonic oscillator. One inter-
prets N as the number of photons present in the particular resonator mode under
consideration. Since photons are bosons, the mode may be populated with any num-
ber N photons.
This is illustrated in the energy diagram Fig. 2.17. The levels of the harmonic
oscillator are equally spaced, and the lowest energy is given by ω/2, the “zero point
energy”. Excitation of the N th harmonic of the classical oscillator corresponds to a
state occupied by N photons.
The eigenstates |N of the number operator (and the harmonic oscillator) may
be generated from the vacuum state |0 by repetitive application of the operator â +
for which
√
â + |N = N + 1|N + 1. (2.106)
2.3 Field Quantization and Optical Transitions 113
Hence, â + is called creation operator. In contrast, the operator â reduces the photon
number by one, when applied to a number state
√
â|N = N |N −1, (2.107)
and hence â is called annihilation operator. N such operations lead to the vacuum
state |0 for which the relation
â|0 ≡ 0 (2.108)
must hold, since a nonexisting photon cannot be destructed any further. The inverse
scheme starts with√ the vacuum state, from √ which one can√ generate any number state
+ N
by (â ) |0) = N !|N . The factors N + 1 and N in (2.106) and (2.107),
respectively, make sure that the number states are correctly normalized as stated by
(2.103).
Obviously, the number operator (2.99) counts the number of photons N in the
mode under consideration. This number is increased or decreased by one when the
operator â + and â, respectively, acts on the photon states. Writing the Hamiltonian
in the form (2.101) implies simply counting the occupation number. This proce-
dure is called second quantization and may be applied to other quantum objects as
well.
The evolution of the photon states with time |ψN (t) is obtained from the trivial
time dependent S CHRÖDINGER equation
F ψN (t) = i δ ψN (t) ,
H (2.109)
δt
ψN (t) = e−iWN t/ |N = e−i(N + 12 )ωk t |N . (2.110)
Note that in all this discussion the S CHRÖDINGER picture is used. All time depen-
dence of the radiation field is now cast into the time dependence of the photon states.
The operators â and â + are not time dependent.
Finally, we come back to the key question: how to quantize the electromagnetic
field? Comparing the definition (2.96) for annihilation and creation operators with
the classical field quantities (2.93) leads us immediately to operators for the electric
114 2 Coherence and Photons
Again, in the S CHRÖDINGER picture these operators are independent of time. They
have to be inserted into (2.87) in place of their classical counterparts. Thus, the
electric field operator may be written as
i 2ωk
t) =
E(r, âk ukq (r) − âk+ u∗kq (r) (2.112)
2 ε0
kq
Since creation and annihilation operators are defined dimensionless, one easily ver-
= V m−1 . We point
ifies that the unit of the electric field operator is indeed [E]
t) is a Hermitian operator (while its constituents âk and â + are not).
out that E(r, k
We also mention here, that (2.112) is sufficiently flexible to adapt the quantization
formalism for any specific experimental situation by an appropriate change of the
modes (2.113) – e.g. for application to quantum optics in a cavity.
To obtain the field energy one has to insert (2.112) into (2.89) and evaluate it
in full analogy to the classical considerations. However, now the commutation rule
(2.97) must be observed. This finally leads to a sum of Hamiltonians (2.101) for all
modes:
HF = 1 +
ωk âkq âkq +
+ âkq âkq = +
ωk âkq
1
âkq + . (2.114)
2 2
kq kq
It is important to realize that the photon number states introduced above do not
represent coherent light. Rather, coherent light must be described by a linear super-
position of many number states as shown for the first time by G LAUBER (1963). For
a single mode, these so called G LAUBER states (also coherent photon states) are
given by
1 αN
|α = exp − |α|2 |N (2.115)
2 (N !)1/2
N
11 We mention that G LAUBER (1963) uses time dependent field operators (H EISENBERG picture)
and a slightly different notation. He writes (in esu) “the positive frequency part of the electric field
operator”
E (+) (r, t) = i ω/2ak uk (r)e−iωk t .
k
2.3 Field Quantization and Optical Transitions 115
1 α ∗N
and α| = exp − |α|2 N |.
2 (N !)1/2
N
α ∗N α N
α|α = exp −|α|2 = 1, (2.116)
N!
N
as the sum corresponds to the exponential function exp(|α|2 ). They are, however,
not orthonormal, rather we have
1 2 1 2 α ∗N β N 1 2 1 2 ∗
α|β = exp − |α| − |β| exp − |α| − |β| − α β ,
2 2 N! 2 2
N
and for the absolute squared of this scalar product one obtains
α|β2 = exp −|α − β|2 . (2.117)
This set of coherent states is thus overcomplete, i.e. there are more coherent states
than number states |N . From (2.117) we see, however, that two G LAUBER states
get nearly orthogonal, if |α − β|
1. Applying the photon annihilation operator
onto (2.115), we obtain with (2.107)
1 αN
â|α = exp − |α|2 N 1/2 |N − 1 (2.118)
2 (N !)1/2
N
1 α N −1
= α exp − |α|2 |N − 1 = α|α.
2 ((N − 1)!)1/2
N
G LAUBER states are thus eigenstates of the photon annihilation operator â. One
may extract photons of a G LAUBER state without changing that state. Conversely,
|α is not an eigenstate of the photon creation operator. It is important to note, that
with (2.111) single mode G LAUBER states are also eigenstates of the field operator
− (but not of its conjugate counter part). Measuring electromagnetic fields usually
E
implies that photons are registered, i.e. a photon is extracted from the radiation field.
If the field can be described by a G LAUBER state |α, the detectable probability
amplitude will thus be proportional to α E − |α ∝ α. This is in essence what makes
G LAUBER states coherent. Since the characteristic parameter α can also be complex,
α may be seen to represent phase and amplitude of the electromagnetic field.
The expectation values of the annihilation and creation operators in a G LAUBER
state follow from (2.118) and by applying (2.106) onto (2.115), respectively:
Finally we have to obtain a relation between a G LAUBER state and the intensity
of the radiation. Let us first note that the population of photon number states |N in
a G LAUBER state is given by a P OISSON distribution
2 |α|2N
pN = N |α = exp −|α|2 . (2.120)
N!
The expectation value of the photon number operator (2.99) in a state |α, i.e. the
mean photon number N in a G LAUBER state, is
α ∗N α N
|α = exp −|α|2
N = α|N N = |α|2 . (2.121)
N!
N
In this context we recall a well known property of the P OISSON distribution: its
standard deviation is given by
2 |α − α|N
α|N |α2
N = = |α| = N . (2.122)
|α
α|N
This is actually very good news for all of our following discussion. A look at Ta-
ble 2.2 shows that for lasers – the light source typically used todayin spectroscopy
–
the number of photons per mode is extremely large. And since N /N = 1/ N ,
the relative width of the distribution of photon numbers is very small (e.g. for the
dye laser mentioned in Table 2.2 on the order of 10−6 ). Thus, for all intents and
purposes in spectroscopy, we may represent the ideal, coherent G LAUBER state by
a pure number state |N , where N represents the average number of photons N per
mode according to (2.84). With this – extremely good – approximation the derivation
and application of optical transition probabilities given below can be accomplished
without any mathematical difficulties.
Although the G LAUBER states considered here refer to a single occupied
mode only, they provide a good description for a sufficiently intense, quasi-
monochromatic and well collimated radiation field, such as a laser beam (one may
even adapt the modes (2.113) suitably).
Quantitatively, we derive the expectation value of the electric field operator
(2.112) with the help of (2.119)
ωk
α|E|α = iCk αe · e − α e · e
ikr ∗ ∗ −ikr
with Ck = . (2.123)
2L3 ε0
we can relate the amplitude E0 = 2Ck α of the classical, single mode field to the
number of photons in the mode:
√ √ |E0 | I0 3
N N + 1 |α| = = L . (2.126)
2Ck cω
In the last step we have used the standard relation I0 = ε0 c|E0 |2 /2 between intensity
and field amplitude. We mention that this relation is equivalent to (2.86), with N
being the number of photons in the (presently thus defined) coherence volume L3 .
In the general case of radiation with a finite bandwidth and divergence angle, N has
to be identified with (2.84).
As discussed above the average number of photon per mode in a typical laser
beam may be very high and a classical, coherent radiation field will be de-
scribed by G LAUBER states. However, in principle the photon number states
|0, |1, . . . , |N . . . can be found also with quite different populations – and thus
represent different coherence properties of the radiation field. This is a key theme of
modern quantum optics.
Here we just add a few remarks relevant to multi-mode states as needed to de-
scribe any realistic radiation field in some detail. These states are typically written
as products of single particle states.12 As the most simple case we discuss here only
products of pure number states which by the arguments given in the preceding sub-
section can be a valid description of a quasi-monochromatic laser beam:
{Nkq } = |N1 |N2 . . . |Ni | . . . = |N1 N2 . . . Ni . . .. (2.127)
Since each of these operators acts only onto one of the modes, the commutation rule
(2.97) is now generalized by
+ +
âkq , âk+ q = δkk δqq while [âkq , âk q ] = âkq , âk q ≡ 0. (2.129)
12 A different situation is encountered with so called entangled states – an interesting subject but
N1 N2 . . . Ni . . . |N1 N2 . . . Ni . . . = δN1 N1 δN2 N2 . . . δNi Ni . . . . (2.130)
13 Onecould also use the H EISENBERG picture with a time dependent field operator and time
independent states. The final result would be the same.
2.3 Field Quantization and Optical Transitions 119
|a >
analogy to (4.55), Vol. 1 by14
ωk
(r) = er · E
U = −D · E
=i er · eâk − e∗ âk+
3
2L ε0
k
ωk
=i eCk D† âk+ ,
Dâk − with Ck = (2.132)
2L3 ε0
k
for absorption and emission of a photon with polarization e, respectively. For con-
venience of writing we have here again pulled the elementary charge e out from the
electron dipole moment D = −er.15
As expected, in contrast to (4.55), Vol. 1 the interaction Hamiltonian (2.132) is
now time independent and documents energy conservation: in this fully quantized
picture energy is simply exchanged between atomic and photonic states.
Before we derive the matrix elements of the interaction Hamiltonian U , we point
out that each relevant mode k, e in the sum (2.132) contains two parts: the first part
(with âk ) destroys a photon (in a mode with the wave vector k and the polarization
vector e) and corresponds to absorption, while the second part (with âk+ ) generates
a photon (in a corresponding mode) and describes emission. We recall that in the
semiclassical description of the electromagnetic field (2.87) these two terms corre-
spond to the positive and negative frequency part, respectively.
Without interaction between field and atomic system the eigenstates of the total
system may be written as product states |ψ; {Ni } of atomic states |ψ and photon
states |N according to (2.127).
If the spectral intensity distribution of the radiation is close to a resonance of
the atomic system – as sketched in Fig. 2.18 – a two level system is a usually a
good approximation. The eigenfunction of |ψ then corresponds to either upper or
14 As in the semiclassical description we apply the dipole length approximation. In dipole velocity
With exact eigenfunctions both approximations lead to the same transition probabilities.
15 If
more than one active electrons are involved, one has to replace the position vector r by a sum
over all r i for the active electrons.
120 2 Coherence and Photons
lower state, |b and |a, respectively, while N1 , N2 . . . Ni . . . defines the number of
photons in the modes k 1 , k 2 . . . k i . . . The matrix elements of the interaction Hamil-
tonian (2.132) are given by:
b; Ni U a; {Ni } = b; N1 N2 . . . Ni . . . |er · E|a; N1 N2 . . . Ni . . .
=i eCk Dba N1 . . . Ni . . . |âk |N1 . . . Ni . . .
(2.133)
k
Dba N1 . . . Ni . . . |âk+ |N1 . . . Ni . . . .
−†
The latter rearrangement is possible since r acts only onto the atomic part, while
act only onto the photon part of the system. The matrix
âk and âk+ (and thus E)
elements of the dipole transition operators for absorption and emission are the same
as (4.57), Vol. 1, elaborated for the semiclassical treatment in Sect. 4.3.4, Vol. 1:
and Dab = r ab · e∗ = r ∗ba · e∗ =
D∗ba
†
Dba = r ba · e (2.134)
with r ba = b|r|a = ψb∗ (r)rψa (r)d3 r = r ∗ab .
Since r has odd parity, |b and |a must have different parity. According to (2.128),
the operators âk+ and âk create or annihilate a photon of one specific mode and
polarization. And the photon states are orthogonal according to (2.130). The matrix
element of the U is thus zero unless N = Ni ± 1 holds for one of the photon states,
i
while all others are the same before and after the transition. For one single occupied
mode, i.e. for Nke photons with momentum k and polarization e, the nonvanishing
matrix elements (2.133) may be written in compact form:
|aNke = b|
bNke − 1|U D|aNke − 1|iâk |Nke
= i
Dba eCk Nke (2.135a)
|bNke = a|
aNke − 1|U D|bNke − 1|iâk |Nke
= i
Dab eCk Nke (2.135b)
|bNke = a|
aNke + 1|U D† |bNke + 1| − iâk+ |Nke
= −i †
Dab eCk Nke + 1 (2.135c)
|aNke = b|
bNke + 1|U D† |aNke + 1| − iâk+ |Nke
= −i †
Dba eCk Nke + 1. (2.135d)
Here Ck is the field normalization constant used in (2.132), originating from proper
calibration of the total field energy (2.114). When deriving (2.135a)–(2.135d) from
(2.133) we have used (2.106) and (2.107). The somewhat abstract number Nke of
photons per mode can be related by (2.84) to the (measurable) spectral intensity
distribution I˜(ωk ).
2.3 Field Quantization and Optical Transitions 121
absorption â |b 〉 emission â+ |b 〉
^ ^
〈b –1|U |a 〉 〈a +1|U |b 〉
ħω ħω
(a) |a 〉 (c) |a 〉
â |b 〉 â+ |b 〉
^ ^
〈a –1|U |b 〉 〈b +1|U |a 〉
ħω ħω
(b) |a 〉 (d) |a 〉
Fig. 2.19 Interaction matrix elements between atom and field, schematically; (a)–(d) refer to
equations (2.135a)–(2.135d), respectively
The other two nonvanishing matrix elements correspond to so called “virtual de-
excitation” (Fig. 2.19(b)) and “virtual excitation” (Fig. 2.19(d)) processes by ab-
sorption and emission of a photon, respectively. These processes are not energy
conserving and do not play a role in 1st order perturbation theory, as we shall see
in a moment. However, they are of crucial importance in the description of higher
order processes, such as multi-photon excitation or ionization in strong fields, as
well as for R AMAN scattering and other nonlinear processes.
We finally note that – if necessary –
Dba may be modified appropriately as in the
semiclassical description to describe other types of transitions, such as E2 and M1.
We shall use again 1st order perturbation theory to describe E1 transitions. Even
though this approach has its limitations, to be discussed at the end of this section,
we shall be able now to derive a rate for spontaneous emission, which was not
possible with the semiclassical approach. In any case, the following, fully quantized
treatment of radiation induced transitions will form the basis for later, more rigorous
treatments, e.g. in Chap. 10.
We consider an effective two level system with the atomic states |a and |b
being nearly in resonance with the radiation as indicated in Fig. 2.18. Let H A be
122 2 Coherence and Photons
∂|ψ(t)
i = (H )ψ(t) = (H
0 + U A + H )ψ(t)
F + U (2.138)
∂t
Here cj N is the probability amplitude for finding N photons in the field while
the atom is found in state |j . For simplicity of writing we have dropped again the
indices k and e for the photon states N and for the angular frequency ω of the
field. We insert (2.139) into (2.138), multiply from the left with bN | or aN |,
and obtain two sets of differential equations for cbN and ca N , respectively:
We have exploited the fact that only matrix elements between different atomic states
are non-zero. We insert now the matrix elements (2.135a)–(2.135d). Since only the
terms with N = N ± 1 are non-zero, two types of exponential factors appear in
(2.140):
2.3 Field Quantization and Optical Transitions 123
|ω| = |ω − ωba |
ωba , (2.141)
but allow nevertheless for small detuning ω, as indicated in Fig. 2.18. In typi-
cal spectroscopic applications we shall have to account for detuning on the order
of 108 s−1 , which are to be compared with transition frequencies on the order of
1015 s−1 . Thus, to a very good approximation the non-resonant terms 1/(ω + ωba )
can be neglected. This approximation is called rotating wave approximation (RWA),
since the terms exp[±i(ω − ωba )t] imply, so to say, that the system follows the field
in phase, while the others rotate in the opposite sense and thus average out with
time.16
Consequently, this simplifies (2.140) substantially. Inserting (2.135a)–(2.135d)
for the matrix elements this leads for the two level system to a simple set of two
coupled equations
eCk √
ċbN = Dba ca N +1 ei(ωba −ω)t
N + 1 (2.142)
eCk √
ċa N +1 = − N + 1D∗ba cbN e−i(ωba −ω)t , (2.143)
√
with Ck ∝ ω being the field normalizing constant in (2.132). To derive the absorp-
tion probability we now assume, as in the semiclassical case, that at time t = 0 all
atoms are in the lower state |a. We also assume that the photons in the mode k with
polarization e are represented sufficiently well by a photon number state |N . Thus,
our initial conditions are
In 1st order perturbation theory one assumes in addition that ca N 1 remains con-
stant. Thus, (2.142) may be integrated directly to obtain the probability amplitude
cbN −1 (t) for finding |b N − 1, i.e. for a transition of the system into the excited
state |b by annihilation of a photon. During this process one of the originally N
photons in mode k is absorbed, so that in complete analogy to the classical case
(4.58), Vol. 1 we have
√
eCk N ei(ωba −ω)t − 1
cbN −1 (t) =
Dba . (2.145)
i(ωba − ω)
16 Originally this terminology was coined by microwave and radio frequency spectroscopy (EPR
and NMR), where this phase reflects indeed a real physical rotation of the spin, induced by the
exciting field.
124 2 Coherence and Photons
The transition probability per unit of time is again |cbN −1 (t)|2 /t. This leads to a
transition rate
(N ) 2π 2 2 πωk e2
dRba k = e Ck |
Dba |2 Nke g(ωk ) = 3 |
Dba |2 Nke g(ωk ) (2.146)
2 L ε0
induced by the Nke photons in the mode. We have now reintroduced the indices for
polarization e and wave vector k of the radiation. With g(ωk ) we identify again the
line profile as introduced in Sect. 4.3.5, Vol. 1. Integrated over all frequencies it is
normalized to unity.
Completely equivalent one assumes for the de-excitation process |b → |a ini-
tial conditions
By integration of (2.143) one derives the probability amplitude for finding the sys-
tem in a state |aN + 1. From this the transition rate for a ← b by emission of a
photon into the mode k, e is obtained:
(N ) πωk e2
dRab k = |
Dab |2 (Nke + 1)g(ωk ). (2.148)
L3 ε0
So far the derivation was completely analogous to the semiclassical approxima-
tion. Now we have to recall, however, that there are dmωe modes per frequency
interval dωk and solid angle element dΩ, with dmωe given by (2.73). We thus find
the absorption or emission probability into given solid angle dΩ element by integra-
tion over all available angular frequencies of the electromagnetic radiation inducing
the transition. This leads to a rate
+∞
(N k ) πe2 L3 ω2
dRba = dRba dmωk e = 3 2 dΩ dωk |
D |2 Nke ωk g(ωk )
3 ba
L ε0 −∞ (2πc)
2
πe2 ωba
= dΩ |
Dba |2 Nke ωba . (2.149)
ε0 2 (2πc)3
In the last step we have assumed that the line profile of the transitions is very narrow,
g(ωk ) = δ(ωk − ωba ), compared to the spectral bandwidth of the radiation which
induces the transition. We shall present in Chap. 10 a simple recipe to modify the
result for narrow band radiation.
We see now, that in the final step the normalization volume L3 has happily
dropped out, since normalization of the field operator cancels versus mode den-
sity. We have written (2.149) in a manner to show the essential ingredients: apart
from the numerical prefactor we recognize the mode density per angle and volume
(2.75), the dipole transition moment projected on the polarization (2.134), Dba , and
the total photon energy Nke ωba , with Nke being the number of photons in the
mode k, e prior to absorption or emission with an angular frequency corresponding
to the transition frequency ωba .
2.3 Field Quantization and Optical Transitions 125
˜
πe2 2 I (ωba )
= dΩ|
D ba | .
ε0 c2 beam δΩ
We point out that the mode density and the single photon energy cancel out. The
spectral intensity I˜(ωba ) is a measurable source parameter. E.g., its value is given
by (2.21) for a laser tuned into resonance ωba – if its overall bandwidth is much
larger than the linewidth of the transition.
We now recall that | Dba |2 = |r ba · e|2 depends on the propagation direction of
the light. For any reasonable laser beam, well collimated to δΩ
1, we may con-
sider |r ba · e|2 to be constant for all populated wave vectors k in the beam. Under
such conditions the angular dependence of I˜(ωba )/δΩ may be considered a delta
function (beam) and the integration of (2.149) over all solid angles yields the total
absorption probability
πe2 ˜
2 I (ωba ) 4π 2 α I˜(ωba )
Rba = |
D ba | = Dba |2 I˜(ωba ) = Bba
| (2.151)
ε0 2 c c
with the fine structure constant α = e2 /4πε0 c and Bba = 4π 2 αc| Dba |2 /, the
E INSTEIN coefficient for the specific sub-transition b ← a induced with polarization
e. We note that this expression is completely identical to (4.63), Vol. 1, derived in
our previous semiclassical treatment. The rate Rba (dimension T−1 ) is – as already
mentioned earlier – by a factor of 3 larger than usually given in textbooks, since we
have derived the expression for a laser beam, rather than for isotropic radiation.
Of particular interest is now the emission of a photon in the transition |b → |a.
The factor (Nke + 1) in (2.150) suggests to distinguish between induced and spon-
taneous emission: the induced emission probability is taken proportional to Nke ,
126 2 Coherence and Photons
that is to the number of photons in the relevant modes prior to the emission process.
A comparison of (2.149) and (2.150) shows, that this probability is identical to the
absorption probability. Thus, for one well defined upper and one well defined lower
state |b and |a
We note in passing, that the photons generated by induced emission appear by def-
inition exactly in the mode by which they are created – as we have partitioned the
whole radiation field (in k space) into well defined, discrete modes and treated them
independently prior to integration. This confirms what in the semiclassical treatment
has simply be assumed: radiation due to induced emission agrees in frequency and
direction exactly with the inducing field.
The factor (Nke + 1) in (2.150) implies that emission may occur even if there
is initially no field present, or more precisely, if initially the field is describe by
the vacuum state with Nke = 0 (the initial state being |b 0). There is an additional
finite, albeit small probability for the de-excitation process |b → |a, involving the
emission of a photon. This transition may be seen as induced by the vacuum field.
This notion may appear somewhat difficult to accept and we shall come back to it
in the next subsection.
In any case, according to (2.150) the resulting spontaneous transition probability
for emission of a photon into a mode k with a frequency ωba and polarization e into
a solid angle dΩ is:
2 3
πe2 ωba αωba
(spont)
dRab = |
D ab |2
ω ba dΩ = |
Dab |2 dΩ. (2.153)
ε0 2 (2πc)3 2πc2
We recall that we had “gleaned” this expression as (4.67), Vol. 1 for spontaneous
emission earlier on. Above derivation supplements the proof.
However, we cannot confine the derivation to just one mode: all empty modes
of frequency ωba do indeed contribute to the process. The angular distribution of
this radiation and its polarization is described by | Dba |2 = |r ba · e|2 (for details we
refer to Sect. 4.5 in Vol. 1). While our earlier treatment of spontaneous emission was
essentially guesswork, it is now firmly based on the quantized radiation field. The
integration over all solid angles gives again the characteristic factor 8π/3|r ba |2 , so
that the total spontaneous emission rate becomes
(spont)
(spont) 4α 1
Rab = dRab = 2 |r ba |2 ωba
3
= Aab = . (2.154)
e 4π 3c τ ab
Aab refers here to spontaneous decay of a specific excited sub-state |b into the
specific lower sub-state |a and fully confirms (4.109), Vol. 1, onto which we have
based up to now all discussions and applications of the A coefficients. To obtain the
overall natural lifetime τnat of a level, one has to sum this expression also over all
final states. Detailed evaluation of the A and B coefficients for specific transitions
2.3 Field Quantization and Optical Transitions 127
and the generalization to degenerate levels has already been presented in Sect. 4.4,
Vol. 1.
We thus have achieved not less than an ab initio derivation of spontaneous emis-
sion and of the relation between the E INSTEIN coefficients. We can now trace the
origin of the well known ω3 dependence of spontaneous emission: It arises from
the mode density (2.75), which is ω2 /(2πc3 ) per unit volume and per angular fre-
quency – and from the photon energy ω. In the treatment of induced probabilities
both terms cancel against the number of photons per mode and the spectral intensity
per frequency interval. Thus, the induced rate (2.151) does not depend directly on
ω – except through the resonance condition g(ωk ) = δ(ωk − ωba ) in (2.149).
Finally, we also have to realize the limitations of the present treatment for emis-
sion and absorption of electromagnetic radiation: we only have used 1st order per-
turbation theory. For induced processes we shall correct this to some extend in
Chap. 10. We shall show there that the set of coupled equations (2.142) and (2.143)
may be solved exactly, as long as the RWA holds, and spontaneous emission can be
neglected. For intense (but not too intense) laser fields and times short compared to
the natural lifetime this is indeed an excellent approximation. For very high inten-
sities – as available today with state-of-the-art ultra fast, high power laser systems,
the rotating wave approximation breaks down, and similarly perturbative approaches
are of limited value only. Hence, special strong field approximations or brute force
numerical methods must be applied.
As for spontaneous emission, it is of fundamental importance for any more rig-
orous treatment of radiative problems, and warrants further efforts beyond 1st order
perturbation theory. The problem is, that the initial conditions (2.147) are, at a closer
look, not strictly valid if spontaneous emission is to be included: All empty modes
are present at time zero. It is important to realize that these empty modes are not
nothing, but represent the vacuum field which is (almost) always present. Indeed,
the vacuum field associated with these many unoccupied modes close to resonance
leads to a broadening of the excited states which we know as natural linewidth.
A quantitative treatment can be achieved in 2nd order perturbation theory – but is
somewhat involved, and we refrain from presenting it here.
The result is, as already assumed in Chap. 5, Vol. 1, that g(ω) in (2.146) and
(2.148) can no longer be treated as a δ-function. Rather, it has to be described by
a L ORENTZ profile with a linewidth (FWHM) ωnat = Aab = 1/τnat . With today’s
techniques the bandwidth of lasers used in spectroscopy may easily be kept much
below that value. This implies that also our assumptions for deriving the relevant
expression (2.151) for induced processes do no longer hold. However, as we shall
show in Chap. 10 this particular problem may be cured by a small modification.
In order to obtain a quantized form of the radiation field we have introduced a very
large but finite normalization volume. This has led us to discrete radiation modes –
still infinitely many, but countable. And all these modes ‘own’ a characteristic vac-
128 2 Coherence and Photons
uum field. It is this very vacuum field that we hold responsible for spontaneous
emission.
Many other physical phenomena are also caused or influenced by the vacuum
field (one speaks of radiative corrections). We recall the L AMB shift or the g − 2
anomaly of the electron magnetic moment treated in Chap. 6, Vol. 1, for which
radiative corrections are held responsible. Nevertheless, the vacuum state is not a
really trivial concept. It is by no means empty space, its eigenenergy being hωk /2
according to (2.114) – for each mode k. Quantum electrodynamics deals with the
problem of this infinite energy by its specific recipe for “re-normalization”. But of
course, one might pose the question: How real is this vacuum field? Is it perhaps
just a mathematical construct made to give the right answers for spontaneous emis-
sion?
What may happen, if one forcefully chooses experimental conditions so that the
normalization volume cannot be made infinitively large? What if we confine any
potential radiation to a small volume and let it interact there with an excited atom?
What does “vacuum state” mean in such a case? The idea of such an experiment has
been around for some time (see e.g. P URCELL 1946; K LEPPNER 1981). However, it
became feasible only by modern lasers and sophisticated experimental techniques.
The first experiment of this type was performed by G OY et al. (1983) – and finally
led, loosely speaking, to the N OBEL prize for H AROCHE and W INELAND (2012).
They studied Na RYDBERG atoms, prepared by two photon resonant excitation in
the 23s state, and investigated its spontaneous decay to the 22p state in a microwave
cavity. The dipole transition moment between such high n levels is very high, while
at the same time the spontaneous transition probability in free space is very small,
being proportional to ∝ ν 3 . In the case discussed here the transition frequency is
ν = 341 GHz (as one easily verifies with the quantum defects of Na given in Ta-
ble 3.4, Vol. 1). The spontaneous decay rate between the 23s and 22p levels is only
τ22p23s = 150 s−1 . At this transition wavelengths (λ = 0.88 mm) super-conducting
microwave resonators with extremely high finesse can be built, through which an
atomic beam can pass without problems.
Figure 2.20(a) shows a very schematic summary of the experimental setup. A low
density sodium beam passes through the microwave cavity where the atoms are ex-
cited in a two photon process by 2 collinear, pulsed (5 ns) dye laser beams, entering
the resonator perpendicular to the Na-beam and to the resonator mode. The RYD -
BERG atoms are detected by field ionization in a parallel plate capacitor to which
(after the laser pulse) a ramp voltage is applied as indicated in Fig. 2.20(b). The
ionization process is monitored very efficiently by detecting the ejected electrons
with an electron multiplier. Atoms with the lower ionization potential WI (23s) are
ionized at a lower field strength (i.e. earlier) than those with higher ionization po-
tential WI (22p). In the detected ionization signal, Fig. 2.20(c), atoms in the 23s
state are recorded first, 22p atoms appear later. The black signal trace is taken with
the cavity out of resonance for the 22p ← 23s transition. The cavity may be tuned
into resonance mechanically, but fine adjustment is done by a small electric field in
the cavity which can S TARK shift the 22p levels slightly. The actual experiment is
carried out with only a few (1–3) excited atoms in the cavity, so that they do not
influence each other. The red line shows the signal with the cavity tuned into reso-
2.3 Field Quantization and Optical Transitions 129
WI
ramp voltage
(a) (b)
tuning 22p
245cm-1
ramp
LHe-cooled voltage 23s
234cm-1
(5.7K)
Nb microwave 1 2 3 t ramp / μs
cavity electric field
cavity in
(c)
Na beam e– resonance
pulsed
laser beams, electron
perpendicular multiplier cavity off
resonance
23 s 22p
nance: a surprisingly intense signal originates from atoms in the 22p state – which
is attributed to spontaneous emission in the cavity.
This is fascinating result! To fully understand it we first have a closer look at the
cavity. It is made of very precisely machined and highly polished, niobium spheres
with 20 mm diameter and 26 mm radius of curvature, arranged in nearly confocal
configuration at L = 25 mm distance. The Gaussian mode sustained by the cavity
at λ = 0.88 mm has a waist w = 1.9 mm and a total volume of Vcav = Lπw 2 /4 =
70 mm3 . Nb becomes super-conducting at 9.2 K and the cavity is liquid He cooled
to ca. 5.7 K. This ensures that the surface of the cavity is highly conducting and
together with very good polishing this leads to a very high quality factor Q of the
cavity (see Eq. (1.11)) on the order of 106 . In addition, the cooling ensures that
black body background radiation cannot lead to induced transitions: according to
(1.63), Vol. 1, B OSE -E INSTEIN statistics gives a population of the N = 1 mode of
[exp(ω/kB T ) − 1]−1 0.06 relative to the vacuum state N = 0 so that radiation
induced processes can be neglected compared to spontaneous transitions which are
caused by the vacuum field. We emphasize the fact that no external or background
microwave field is involved in this experiment.
The atoms spent only about 2 µs in the resonant cavity mode. In the free field case
this would lead to a maximum of 150 s−1 ×2 µs 3 × 10−4 transitions. How then
can we understand the observed enhancement of spontaneous emission by about at
least 5 orders of magnitude?
From our derivation of spontaneous emission in the free field case we recall now,
that one crucial parameter was the mode density per unit volume, according to (2.75)
ρfree (ω) = ωba
2 /(2πc)3 . This has entered directly into the spontaneous emission rate
(2.153). In the final step, integration over all angles leads to multiplication by a
130 2 Coherence and Photons
factor of 8π/3, and another factor of 2 for the two possible polarization modes.
We now compare this to the situation in the cavity. The mode density per angular
frequency, polarization and volume is ρcav (ω) = 1 mode/(Vcav ωr ), where Vcav is
the cavity volume and ωr = ωr /Q with the Q factor of he cavity according to
(1.11). Integration over solid angles is obsolete in this case, since the cavity sustains
only two modes (of different polarization) which are resonant with the transition. In
summary, one has to replace 2 × (8π/3)ρfree (ω) in free space by 2 × ρcav (ω) in the
resonant cavity case. Thus, an enhancement of the spontaneous radiation probability
is expected. In the present case this enhancement factor is on the order of 103 , so that
the transition probability is changed from 150 s−1 to 3 × 105 s−1 so that during
the passage time of 2 µs a substantial fraction of the Na atoms in the initial 23s state
decays by spontaneous transitions into any of the 22p substates – as shown by the
experimental result Fig. 2.20(c).
In conclusion, this experiment documents that the vacuum field is not just a the-
oretical construct, it is real and can be manipulated in a finite cavity – leading to an
observable modification of the spontaneous transition probability – also reflected in
the respective “natural” lifetime.
A number of additional question arise from these findings: e.g. what happens
to the emitted photon? Can spontaneous emission also be suppressed? What about
other effects caused by the vacuum field in a cavity? As it turns out, in the experi-
ment described here the Q factor of the cavity is not high enough to store the emitted
photon long enough for subsequent reabsorption.
In the mean time, experiments have been reported in which oscillatory energy
between the cavity and a single atom in it has been observed. And, yes, spontaneous
emission can also be quenched in a cavity of the right dimensions if the vacuum
field is not in resonance. Also changes of atomic energy levels and modifications
of the L AMB shift have been observed. Cavity quantum electrodynamics has be-
come a very active and productive topic of modern research as e.g. summarized in
a nice review by WALTHER et al. (2006). We also mention that such effects play an
important role in nano-optics, another area of cutting edge research.
Section summary
• We have quantized the electromagnetic field, based on the preceding intro-
duction of discrete, countable modes of the electromagnetic radiation field in
a large, but finite normalization volume L3 .
• To this end we have introduced in (2.92) new variables P and Q as linear
combinations of the components E − and E + of the electric field. The mode
energy was then recognized as formally equivalent to the harmonic oscillator,
with P and Q being canonic conjugate coordinates.
2.3 Field Quantization and Optical Transitions 131
References
BAYM , G.: 1998. ‘The physics of Hanbury Brown-Twiss intensity interferometry: From stars to
nuclear collisions’. Acta Phys. Pol. B, 29, 1839–1884.
B ETH , R. A.: 1936. ‘Mechanical detection and measurement of the angular momentum of light’.
Phys. Rev., 50, 115–125.
B ORN , M. and E. W OLF: 2006. Principles of Optics. Cambridge: Cambridge University Press,
7th (expanded) edn.
B OYAJIAN , T. S. et al.: 2012. ‘Stellar diameters and temperatures. I. Main-sequence A, F, and G
stars’. Astrophys. J., 746, 101 (26 pages).
B ROWN , R. H. and R. Q. T WISS: 1954. ‘A new type of interferometer for use in radio astronomy’.
Philos. Mag., 45, 663–682.
B ROWN , R. H. and R. Q. T WISS: 1956a. ‘Correlation between photons in 2 coherent beams of
light’. Nature, 177, 27–29.
B ROWN , R. H. and R. Q. T WISS: 1956b. ‘A test of a new type of stellar interferometer on Sirius’.
Nature, 178, 1046–1048.
B ROWN , R. H. and R. Q. T WISS: 1958. ‘Interferometry of the intensity fluctuations in light II.
An experimental test of the theory for partially coherent light’. Proc. R. Soc. A 243, 291–319.
TEN B RUMMELAAR , T. A. et al.: 2005. ‘First results from the Chara array. II. A description of the
instrument’. Astrophys. J., 628, 453–465.
DAVIS , J. and B. L OVELL: 2003. ‘Robert Hanbury Brown, 1916–2002’, Australian Academy of
Science. http://www.science.org.au/fellows/memoirs/brown.html, accessed: 9 Jan 2014.
G LAUBER , R. J.: 1963. ‘Coherent and incoherent states of radiation field’. Phys. Rev., 131, 2766–
2788.
G LAUBER , R. J.: 2005. ‘The N OBEL prize in physics: for his contribution to the quantum theory
of optical coherence’, Stockholm. http://nobelprize.org/nobel_prizes/physics/laureates/2005/.
G LAUBER , R. J.: 2006. ‘N OBEL lecture: 100 years of light quanta’. Rev. Mod. Phys., 78, 1267–
1278.
G LAUBER , R. J.: 2007. Quantum Theory of Optical Coherence, Selected Papers and Lectures.
New York: Wiley-VCH Verlag, 643 pages.
G OY , P., J. M. R AIMOND, M. G ROSS and S. H AROCHE: 1983. ‘Observation of cavity-enhanced
single-atom spontaneous emission’. Phys. Rev. Lett., 50, 1903–1906.
G RANGIER , P., G. ROGER and A. A SPECT: 1986. ‘Experimental-evidence for a photon anticor-
relation effect on a beam splitter – a new light on single-photon interferences’. Europhys. Lett.,
1, 173–179.
G RYNBERG , G., A. A SPECT and C. FABRE: 2010. Introduction to Quantum Optics: From the
Semi-classical Approach to Quantized Light. Cambridge: Cambridge University Press, 665
pages.
H AROCHE , S. and D. J. W INELAND: 2012. ‘The N OBEL prize in physics: for ground-breaking
experimental methods that enable measuring and manipulation of individual quantum systems’,
Stockholm. http://www.nobelprize.org/nobel_prizes/physics/laureates/2012/.
H ASSELBACH , F.: 2010. ‘Progress in electron- and ion-interferometry’. Rep. Prog. Phys., 73.
H ENNY , M., S. O BERHOLZER, C. S TRUNK, T. H EINZEL, K. E NSSLIN, M. H OLLAND and
C. S CHONENBERGER: 1999. ‘The fermionic Hanbury Brown and Twiss experiment’. Science,
284, 296–298.
ISO 21348: 2007. ‘Space environment (natural and artificial) – Process for determining solar irra-
diances’. International Organization for Standardization, Geneva, Switzerland.
J ELTES , T. et al.: 2007. ‘Comparison of the Hanbury Brown-Twiss effect for bosons and fermions’.
Nature, 445, 402–405.
134 2 Coherence and Photons
K IMBLE , H. J., M. DAGENAIS and L. M ANDEL: 1977. ‘Photon anti-bunching in resonance fluo-
rescence’. Phys. Rev. Lett., 39, 691–695.
K LEPPNER , D.: 1981. ‘Inhibited spontaneous emission’. Phys. Rev. Lett., 47, 233–236.
K LEPPNER , D.: 2008. ‘Hanbury Brown’s steamroller’. Phys. Today, 61, 8–9.
L AMBROPOULOS , P. and D. P ETROSYAN: 2007. Fundamentals of Quantum Optics and Quantum
Information. Berlin, Heidelberg, New York: Springer Verlag, 325 pages.
L OUDON , R.: 2000. Quantum Theory of Light. Oxford, New York: Oxford University Press, 3rd
edn.
M ANDEL , L. and E. W OLF: 1995. Optical Coherence and Quantum Optics. Cambridge: Cam-
bridge University Press.
M ICHELSON , A. A. and F. G. P EASE: 1921. ‘Measurement of the diameter of a orionis with the
interferometer’. Astrophys. J., 53, 249–259.
M ILLONI , P. W. and J. H. E BERLY: 2010. Laser Physics. Hoboken: Wiley, 832 pages.
M ONNIER , J. D.: 2003. ‘Optical interferometry in astronomy’. Rep. Prog. Phys., 66, 789–857.
M UKAMEL , S.: 1999. Principles of Nonlinear Optical Spectroscopy. Oxford: Oxford University
Press, 576 pages.
P HILLIPS , D. T., H. K LEIMAN and S. P. DAVIS: 1967. ‘Intensity-correlation linewidth measure-
ment’. Phys. Rev., 153, 113–115.
P URCELL , E. M.: 1946. ‘Spontaneous emission probabilities at radio frequencies’. Phys. Rev., 69,
681, Note B10.
WALTHER , H., B. T. H. VARCOE, B. G. E NGLERT and T. B ECKER: 2006. ‘Cavity quantum
electrodynamics’. Rep. Prog. Phys., 69, 1325–1382.
W EISSBLUTH , M.: 1978. Atoms and Molecules. New York, London, Toronto, Sydney, San Fran-
cisco: Academic Press, Student Edition, 713 pages.
W EISSBLUTH , M.: 1989. Photon-Atom Interactions. New York, London, Toronto, Sydney, San
Francisco: Academic Press, 407 pages.
Diatomic Molecules
3
Overview
This chapter outlines the basic concepts of molecular physics as exempli-
fied for diatomic molecules. We begin with some energetic considerations in
Sect. 3.1 and introduce in Sect. 3.2 the B ORN -O PPENHEIMER approxima-
tion – the basis of all molecular physics. Molecular rotation and vibration
are treated in Sect. 3.3 followed by an elaboration on dipole transitions in
Sect. 3.4. Elements of the molecular orbital concept are presented in Sect. 3.5
while Sect. 3.6 focusses on angular momentum coupling and the famous
H UND’s cases. While all the previous discussion was focussed on homonu-
clear molecules, the chapter ends by introducing the specificities of heteronu-
clear diatomic molecules in Sect. 3.7. The content of this chapter is essential
for understanding most of the following ones – it is one of the keystones
within these textbooks. The reader should thus familiarize him- or herself
very thoroughly with all topics discussed here.
experiments and more recently also ultrafast methods have revealed important struc-
tural information, but they also allow in addition to study the dynamics of molecular
systems, i.e. the evolution of processes within or between molecules when interact-
ing with each other or with photons. We shall treat these themes in some detail in
Chaps. 6–8, and some special aspects will be mentioned in Chap. 10. In the present
chapter we shall develop the key concepts for understanding molecules as exempli-
fied for diatomic molecules, the simplest molecular systems. Based on these con-
cepts we shall explain the most important experimental findings for homonuclear
(A2 ) and heteronuclear (AB) diatomic molecules.
C
H
H
1 Notwithstanding this fundamental fact, there is a notable, brave quest to observe influences of
length”. The value given here for CH4 is based on state-of-the-art quantum chemical calcula-
tions and comparison with precision infrared and R AMAN spectra of different isotopologues of
CH4−x Dx according to S TANTON (1999).
3.1 Characteristic Energies 137
MB x
3.1.1 Hamiltonian
In this chapter we concentrate on diatomic molecules and start with deriving the
Hamiltonian. We use atomic units (a.u.), and relative coordinates in respect of the
centre of mass (O) of the system as illustrated in Fig. 3.2. Translational motion of
the molecule as a whole does not play a role in the following discussion. It simply
reflects the thermal motion in an ensemble. The kinetic energy of the two nuclei and
the N electrons is3
N
me 2 1 2
Tn = − ∇R and Tr = − ∇ ri , (3.1)
2M̄ 2
i=1
where now r is considered to represent the entirety of the coordinates of all elec-
trons. With this follows the S CHRÖDINGER equation
Tn (R) + Tr (r) + V (r, R) Ψ (r, R) = W Ψ (r, R). (3.5)
3 Weignore here the kinematic shift (which is anyhow problematic for multi-electron systems) and
assume (with reasonable accuracy) m̄e = me M̄/(me + M̄) me .
138 3 Diatomic Molecules
In this form it is also valid for polyatomic molecules, if R is taken to represent the
entirety of all nuclear coordinates.
Let us try to obtain an estimate for the electronic energy of such a system. For clarity
we switch back to SI units. We identify the extension of the electron orbitals r with
the bond length R0 of the molecule. For molecular hydrogen (H2 ) as an example
one finds r R0 0.074 nm. The uncertainty relation gives us an estimate for
the momentum of the electrons p = /r, and an average kinetic energy Tr =
p 2 /(2me ), which we may insert into the well known Virial theorem 2Tr = −Ve
for −1/r potentials. Thus, the binding energy of the electrons can be estimated
roughly as
We = Tr + Ve = −Tr −p 2 /(2me ) = −2 / 2me r2 7 eV.
Typically, We is found in an energy range of several eV, and the electronic spectra
are expected in the ultraviolet and visible range of the electromagnetic spectrum –
quite comparable to electronic transitions within atoms. More interesting are the
nuclear degrees of freedom. We distinguish vibrations of the atomic nuclei within
a molecule, relative to each other, and rotation of the whole nuclear structure –
considered as fixed in a 0th order approximation.
Since atoms and electrons are bound to each other, the force on the nuclei must be
of the same order of magnitude as that on the electrons. Let us assume this force
i.e. F = −kR. The (angular) vibrational frequency of the nuclei is
to be harmonic,
then ωv = k/M̄, with the reduced mass M̄ on the order of the nuclear masses.
The frequency of the electron is given by a force constant of similar magnitude so
√
that ωe = k/me . The corresponding
energies of vibrational motion and electronic
√
energy are Wv = ωv = k/M̄ and We = ωe = k/me , respectively. Thus,
for
the ratio of vibrational
nuclear motion and electronic energy we find Wv /We
me /M̄, and since me /M̄ 10−2 vibrational energies will be on the order of
magnitude
Wv me /M̄We 0.1 eV. (3.6)
Transitions are found in the infrared spectral range, e.g. for HCl at ν̄ = 1/λ = ν/c
3000 cm−1 = 0.37 eV or λ 3.33 µm.
3.2 B ORN O PPENHEIMER Approximation 139
O
R0 ~
1Å
N
2
2 me
WN = 2
2
= We
2M̄R M̄r M̄
10−4 –10−3 We 1 meV to 10 meV. (3.7)
The corresponding transitions are in the far infrared and microwave spectral range
(ν̄ = 1 cm−1 to 10 cm−1 ).
Section summary
• We have introduced the molecular Hamiltonian (3.4) and derived some rough
estimates for typical molecular energies, being several eV for the electronic
part, 0.1 eV for the molecular vibration and in the 1 meV –10 meV region
for rotational motion.
The big difference between electronic and nuclear energies suggests to separate
electronic and nuclear motion. This was first proposed by B ORN and O PPEN -
HEIMER (1927)who expanded the contributions of the nuclei to the Hamiltonian
4
into a series of me /M̄. They found that nuclear vibrations correspond to 2nd, ro-
tation to 4th order, while 1st and 3rd order disappear. The B ORN -O PPENHEIMER
(BO) approximation turns out to be an excellent approximation and forms the basis
of all molecular structure theory.
140 3 Diatomic Molecules
The first step is a product ansatz composed by the wave functions of electrons
φ(r 1 , r 2 , . . . , r N ) and those of atomic nuclei ψ(R),
Ψ (r 1 , r 2 , . . . , r N , R) = φ(r 1 , r 2 , r 3 , . . . , r N )ψ(R) ≡ φ(r)ψ(R), (3.8)
where r refers to the entirety of electronic and R to all nuclear coordinates.
The key idea is to consider – on a timescale relevant for the rapid motion of the
electrons – the nuclei as fixed in space. The electronic part of the Hamiltonian (3.4)
at a fixed value of R is
H(el) = H − Tn (R) = Tr + V (r; R) , (3.9)
and the corresponding S CHRÖDINGER equation is now written as
(el) φγ (r; R) = Tr + V (r; R) φγ (r; R) = Vγ (R)φγ (r; R).
H (3.10)
For a start, BO approximation considers R simply as a ‘parameter’. The semicolon
in the electronic wave function φγ (r; R) and in the potential V (r; R) is meant to
emphasize this assumption. But in the back of our minds we have to remember
that electronic and nuclear coordinates are, strictly speaking, nonseparable: in the
potential V (r, R) given by (3.3), the r and R coordinates are nicely intertwined –
except for the C OULOMB repulsion terms of which ZA ZB /R can be treated as a
simple additive constant to the electronic energy.
In the framework of the BO approximation, (3.10) is solved for each, fixed value
of R independently. This leads to a set of electronic quantum numbers γ . The elec-
tronic energy Vγ (R) for each set of γ is a continuous function of R, called molec-
ular potential (in the diatomic case), and potential hypersurface (in the general,
multidimensional case).
As the velocities of the electrons are orders of magnitude higher than those of the
nuclei, the electrons orbit many times around the nuclei before these have moved
significantly. Considering R as fixed is thus a very reasonable approximation for
solving (3.10).
All preceding discussion holds for any number of atoms. In the following we
shall, however, concentrate for simplicity on the diatomic case, the generalization
being straight forward. Then R stands for the relative coordinate R = R A − R B
of the two atomic nuclei, and the electronic energy will depend exclusively on the
internuclear distance R = |R|. By solving (3.10) for a range of R, one thus obtains
for each γ a molecular potential Vγ (R), along with the corresponding electronic
wave function φγ (r; R).4
In analogy to atoms, the indices α, β, γ define a set of quantum numbers which
characterize the electronic charge cloud. The electronic wave functions φγ form
again a complete, orthonormal set
φγ |φγ = φγ∗ (r; R)φγ (r; R)d3 r = δγ γ . (3.11)
4 Depending on the required accuracy, r refers usually only to the coordinates r i of the most im-
portant electrons, e.g. to the valence electrons, and one ignores the inner shells.
3.2 B ORN O PPENHEIMER Approximation 141
De
R0
united equilibrium separated
atoms distance atoms
Electronic energies, for diatomic molecules called potentials Vγ (R), may be ob-
tained from (3.10) for any set of electronic quantum numbers as a continuous func-
tion of R. We postpone to Sect. 3.5 a more detailed discussion about how exactly to
compute them. For the moment we just give a qualitative overview. With (3.3) we
may write the potentials derived from the electronic S CHRÖDINGER equation (3.10)
ZA ZB
Vγ (R) = Uγ (R) + , (3.12)
R
now again in a.u. The C OULOMB repulsion of the nuclei dominates at short inter-
nuclear distances R while the term Uγ (R) is due to the electrons and is dominantly
attractive if the molecular electron configuration is bonding.
A typical molecular potential has several characteristic regions as illustrated in
Fig. 3.4, for a bonding and for an antibonding diatomic potential. At large distances
the polarizability of the atomic electron clouds leads to an attractive (albeit usually
very weak) VAN DER WAALS potential ∝ −R −6 , as discussed in Sect. 8.3, Vol. 1.
The repulsive C OULOMB interaction plays an important role in a systematic treat-
ment of molecular orbitals – even though the limit of united atoms (more precisely
atomic nuclei) is never reached. We shall come back to this in Sect. 3.5.4.
In between these limiting cases the interaction of the electron shells is particu-
larly strong. Its nature determines whether two atoms can form a molecule or not:
as we shall proof in Sect. 3.5 a strong overlap of the electronic charge clouds leads
to bonding, little or no overlap to antibonding potentials. As sketched in Fig. 3.4
the bond energy (or dissociation energy) De and the equilibrium distance R0 corre-
spond to the potential minimum, zero energy referring here to completely separated
atoms, Vγ (∞) = 0.
142 3 Diatomic Molecules
= − 1 ∇ 2R + H
H (el) . (3.13)
2M̄
The S CHRÖDINGER equation (3.5) for the molecule as a whole is then
with Wγ vN = W being the total energy of the molecule. The indices vN refer to
the nuclear motion as we shall see in a moment. Now let us have a detailed look
at the components of this S CHRÖDINGER equation. The electronic term is indeed
given by (3.10), here explicitly as H(el) φγ ψvN = Vγ (R)φγ ψvN , since H(el) acts
as differential operator only on the electronic coordinates. However, the kinetic
energy operator for the nuclei, −∇R2 /(2M̄), acts on both factors of the product
φγ (r i ; R)ψvN (R)φγ ψvN . The S CHRÖDINGER equation (3.14) for the whole sys-
tem thus becomes:
me 2
H φγ (r i ; R)ψvN (R) = φγ − ∇ R ψvN + Vγ (R)φγ ψvN (3.15)
2M̄
me 2 me
− ψvN ∇ R φγ + 2(∇ R φγ )(∇R ψvN ) (3.16)
2M̄ 2M̄
= Wγ vN φγ (r i ; R)ψvN (R).
2 2 2 2 p2 me pe2 me
∇ R φγ ∇ r i φγ e = Vγ ,
2M̄ 2M̄ 2M̄ M̄ 2m e M̄
3.2 B ORN O PPENHEIMER Approximation 143
i.e. we neglect an energy on the order of 10−3 . . . 10−5 Vγ . As a matter of fact, one
finds that the B ORN -O PPENHEIMER approximation is a surprisingly excellent ap-
proximation, far beyond what one might expect from this estimate!
Thus, we have to solve for the nuclear motion (now back to a.u.)
1 2
φγ − ∇ R ψvN + Vγ (R)φγ ψvN = Wγ vN φγ ψvN .
2M̄
By multiplying this from the left with φγ∗ , integrating over r, and using the or-
thogonality of the electronic wave functions (3.11), we obtain the S CHRÖDINGER
equation for the nuclear wave function ψvN (R) and its eigenvalues Wγ vN :
me 2
− ∇ R ψvN (R) + Vγ (R)ψvN (R) = Wγ vN ψvN (R). (3.17)
2M̄
The Hamiltonian for the nuclear motion is thus
Our next goal is to understand the nuclear motion and describe it in quantum me-
chanical terms. Traditionally, several steps of approximation are taken: again one
tries to separate different types of motion. A first step is to expand the potential into
a TAILOR series and see how far the lowest order approximation leads. To study
the oscillations of the molecule, the potential is expanded around its equilibrium
distance R0 :
dVγ 1 2
2 d Vγ
Vγ (R) = Vγ (R0 ) + (R − R0 ) + (R − R0 ) + ··· .
dR R=R0 2 dR R=R0
2
At the potential minimum dVγ /dR|R=R0 = 0, and, neglecting higher terms, we ob-
tain a harmonic potential
1
Vγ (R) = Vγ (R0 ) + k(R − R0 )2 (3.19)
2
with the force constant k = d2 Vγ /dR 2 R=R .
0
-4 -2 0 2 (R - R0 ) / l
The harmonic oscillator is probably one of the best treated objects on all levels of
physics education. We just summarize the essential results. In quantum mechanics
one obtains the wave functions Rv (R) and the energy eigenvalues Wv from the
one-dimensional S CHRÖDINGER equation:
2 d 2 M̄ω02
− + (R − R 0 )2
Rv (R) = Wv Rv (R). (3.21)
2M̄ dR 2 2
Introducing a characteristic length l (typically between 0.1 and 0.25a0 )
l = /(M̄ω0 ) and setting x = (R − R0 )/ l, (3.22)
√
1 √ 2 −x 2 /2
4 π xe
√ 1√ e−x
2 /2
2 (2x 2 − 1)
2 π
√ 1√ xe−x
2 /2
3 (2x 2 − 3)
3 π
√1√ e−x
2 /2
4 (4x 4 − 12x 2 + 3)
2 6 π
3.2 B ORN O PPENHEIMER Approximation 145
(−1)v dv
hv (x) = √ exp x 2 /2 v
exp −x 2 . (3.24)
v
2 v! π dx
The wave functions for the four lowest levels are sketched in Fig. 3.5 and sum-
marized in Table 3.1. Note that they alternate between being symmetric and anti-
symmetric in respect of the equilibrium distance R0 . We also emphasize the finite
extension of the ground state wave functions (v = 0, zero point oscillation), which
is a pure G AUSS function. The corresponding zero point energy of the ground state
is Wmin = ω0 /2.
Figure 3.6 shows an example for a rather high vibrational quantum number v =
20. One sees that in this case the probability density at the boundaries increases
substantially – corresponding fully to the longer time a classical harmonic oscillator
spends at the classical turning point.
mation to reality. Often the potential introduced by P.M. M ORSE (1929) is used:
VM (R) = De e−2a(R−R0 ) − 2e−a(R−R0 ) . (3.26)
This M ORSE potential offers already some flexibility with three parameters De
(bond energy), R0 (equilibrium distance) and a (some kind of stiffness of the po-
tential). It reproduces the limits correctly and is rather simple to handle in compu-
tations. For CO the fit to the experimental data by a M ORSE potentials (full red line
in Fig. 3.7) looks surprisingly good.5
As indicated in Fig. 3.7 one usually defines zero energy for the separated atoms in
the electronic ground state, i.e. Vγ (∞) = 0, and obtains for the potential minimum
Vγ (R0 ) = −De . Alternatively one also uses the form
2
VM (R) = De 1 − e−a(R−R0 ) ,
which differs from (3.26) just by a shift of zero energy to −De . The true ground
state bond energy required to dissociate the molecule (from its vibrational ground
state v = 0 as indicated in Fig. 3.7) is in any case
V(R) / eV R / nm
v=0
Fig. 3.7 Potential of the CO ground state. Crosses give the experimentally determined points of
the RKR potential according to F LEMING and R AO (1972), M ANTZ et al. (1971). A M ORSE po-
tential has been fitted to it (full, red line) with a bond energy De = 11.108 eV and an equilibrium
distance R0 = 0.11282 nm. This potential may be compared to (dashed red lines) (a) a harmonic
potential of equal vibrational frequency in the ground state or (b) a 12, 6 L ENNARD -J ONES poten-
tial
5 However, experimental precision is so far advanced that the M ORSE potential does not suffice
state-of-the-art requirements. The accuracy of the data points given in the literature is in some
instances as good as 1 ppm.
3.2 B ORN O PPENHEIMER Approximation 147
Table 3.2 Potential parameters for some characteristic diatomic molecules: Reduced mass M̄,
equilibrium distance R0 , dissociation energy D00 = De − ω0 /2 in respect of the ground vibrational
state, its vibrational energy ω0 , and its dipole momenta Dγ v
Molecule M̄/ u R0 / nm D00 / eV ω0 / eV Dγ v /10−30 C m
H+
2 0.504 0.1052 2.651 0.2714
H2 0.504 0.07414 4.478 0.5156
D2 1.0071 0.07415 4.556 0.37095
7 Li1 H 0.8812 0.15957 2.4287 0.16853 19.6256
1 H35 Cl 0.9796 0.12746 4.433 0.3577 3.6979
N2 7.0015 0.109768 9.759 0.28888
14 N16 O 7.466 0.115077 6.497 0.23260 0.52943
O2 7.997 0.120752 5.115 0.19295
12 C16 O 6.8562 0.112832 11.09 0.26573 0.3662
Na2 11.4949 0.30788 0.720 0.01955
23 Na35 Cl 13.870 0.23608 4.23 0.0448b 30.025
Cl2 17.4844 0.1987 2.479 0.0687
As far as not otherwise mentioned, according to H UBER and H ERZBERG (1979)
a L OVAS et al. (2005)
b R AM et al. (1997)
For general orientation and further reference, Table 3.2 summarizes the relevant
potential parameters for some characteristic diatomic molecules.
The harmonic approximation to the M ORSE potential is
VM (R) = −De 1 − a 2 (R − R0 )2 + · · ·
and with (3.19) and (3.20) the stiffness a is related to the spring constant k close to
the potential minimum and to the fundamental frequency ω0 by
k = 2De a 2 and ω0 = k/M̄ = a 2De /M̄. (3.27)
This is also the physical cause for thermal expansion of solids: with increasing
temperature the average vibrational energy increases, hence the average value of v
increases as well, and so does R.
148 3 Diatomic Molecules
Note, however, that the number of bound states in such a “potential well” is
always finite – even though the energy spacing between neighbouring levels gets
very small close to the dissociation limit. We recall that, in contrast, the number of
bound electronic states in an atom is infinite. This is entirely due to the particular
nature of the 1/r C OULOMB potential. All other (bonding) potentials support at
most only a finite number of bound states.
While the parameter b (the so called co-volume) reflects the finite extension of
atoms and the repulsive part of the intermolecular potentials and decreases the avail-
able volume, the term a/V 2 (the so called cohesive pressure) is attributed to VAN
DER WAALS forces at large distances which effectively increase the external pres-
sure p.
The vdW interaction is about a factor 102 to 103 smaller than typical bond ener-
gies. For example, for the non-bonding rare gas system Ar· · · Ar it amounts to about
0.08 eV in the minimum, while Cl2 (in the periodic table directly next to Ar) is
chemically bound, albeit very weakly, with De 2.48 eV. VAN DER WAALS inter-
action is typically of the same magnitude as thermal energies at room temperature
203.6 cm−1 .
kB × 293 K 0.025 eV = (3.29)
Hence, to study it in detail one has to work at very low temperatures. Helium, for
instance, may be liquified at low temperatures (4.3 K) in spite of its closed 1s 2
shell. At sufficiently low temperature He gas forms clusters or little droplets. At any
rate, rare gases have a strong tendency to form atomic clusters at low temperatures.
A popular method to generate such objects is adiabatic expansion of the rare gases
in dense atomic beams.
3.2 B ORN O PPENHEIMER Approximation 149
1.5
Tang-Toennies
1.0 Lennard-Jones (12,6)
0.5
0.0
De = 0.348
- 0.5
4 8 10 12 R / a0
R 0 = 5.62
Fig. 3.8 Interaction potential between two He atoms. We compare the nearly exact TANG -T OEN -
NIES potential (TANG et al. 1995, red line) with a L ENNARD -J ONES 12, 6 potential (3.30) fitted to
it (black line). Note that only two free parameters De = 0.34784 × 10−4 Eh ( = 0.9465 meV) and
R0 = 5.62a0 determine the L ENNARD -J ONES potential. Both potentials are – on the scale used
here – nearly indistinguishable
ultracold He atomic beams the existence of the He2 molecule has been established
beyond any doubt. According to G RISENTI et al. (2000) the presently most accu-
+0.3
rate value for the bond energy of its one existing state is D00 = (1.1−0.2 ) mK, i.e.
−7
10 eV(!), its average size is R = (5.2 ± 0.4) nm(!). Clearly, He2 is a quite
pathological case of a molecule!
By a combination of VAN DER WAALS contact distances between different part-
ner atoms one estimates the so called VAN DER WAALS radii of individual atoms.
Table 3.3 presents examples for some important elements. A graphical overview has
already been communicated in Fig. 3.3, Vol. 1.
Section summary
• We have introduced the B ORN -O PPENHEIMER approximation as fundamental
for our quantitative understanding of molecules. It is based on the fact that
electrons move much faster than nuclei. In a classical picture, electrons circle
on their orbits many times before the nuclei have significantly changed their
position R.
• Hence, the electronic S CHRÖDINGER equation (3.10) can, to a very good
approximation, be solved for the electronic coordinates only, with R being
treated as a freely variable parameter. The electronic energy Vγ (R) for a set
of electronic quantum numbers γ is a function R, called molecular potential
(hypersurface for polyatomic molecules).
• Characteristic shapes of Vγ (R) are shown for the diatomic case in Fig. 3.4.
Nuclear motions occurs on these potentials.
• The harmonic potential and the harmonic oscillator were found to be only
a very crude approximation, describing the near equilibrium region and the
lowest vibrational levels of a diatomic molecule reasonably well.
• The M ORSE potential (3.26) was shown to be a useful analytic form for a
chemically bound diatomic molecule, characterized by dissociation energy
De , equilibrium distance R0 and molecular stiffness a related to the funda-
mental frequency ω0 by (3.27).
• Even if there is no chemical bond, neutral atoms experience a long range at-
tractive force caused by mutual polarization, the so called VAN DER WAALS
interaction ∝ −1/R −6 . Based on this and a short range repulsion term ∝
1/R 12 the L ENNARD -J ONES (12, 6) potential is used to describe non-bonding
3.3 Nuclear Motion: Rotation and Vibration 151
Z Z M
Θ Θ R = RA – R B
MA ζ
R A = (M B / M ) R
O
Y O Y
Φ
X
X Φ
MB R B = (MA / M ) R
Fig. 3.9 The ‘reduced nuclear particle’: one transforms the two body problem by introduction of
a reduced molecular mass M̄ = MA MB /M (with M = MA + MB ) into an effective one particle
problem whose motion is described by the relative nuclear position coordinate R
interactions, e.g. between rare gas atoms. He2 was seen to be a very special
case of a weakly bound molecule with De 10−7 eV.
With our present understanding of molecular potentials it is possible to study the nu-
clear motion in some more detail. To avoid confusions, we switch back to SI units
in the following discussion. The reduced mass M̄ = MA MB /M moves on a spher-
ical symmetric electronic potential Vγ (R). By introducing the relative coordinate
R = R A − R B instead of the individual nuclear coordinates R A = R MB /M and
R B = R MA /M (with M = MA + MB ) one replaces the vibrating nuclear mass, so
to say by one ‘reduced nuclear particle’ as illustrated in Fig. 3.9.6 This leads again to
a one particle S CHRÖDINGER equation very similar to that for the hydrogen atom.
Merely the potential 1/r has to be replaced by Vγ (R), the electronic eigenenergy
which now depends on the distance R between the nuclei. In complete analogy to
(2.108) in Vol. 1 we write (3.17) now:
n ψγ vN (R) = H
H R + H rot + Vγ (R) ψγ vN (R) = Wγ vN ψγ vN (R) (3.31)
2
R = − 1 ∂ R 2 ∂ and H
2
with H rot = N .
2M̄ R 2 ∂R ∂R 2M̄R 2
The quantum number v characterizes the radial motion (vibration). The angular
momentum operator N of the reduced nuclear particle is a conserved quantity –
just as commutes according to the usual angular
L for the electron of the H atom. N
6 The electron coordinates too are transformed and referred to the molecular axis (in Fig. 3.9 the ζ
axis).
152 3 Diatomic Molecules
2 N(N + 1)
Veff (R) = + Vγ (R). (3.35)
2M̄R 2
The reduced mass M̄ moves in this effective potential. In principle, one may solve
(3.34) for each N numerically – as usual for reasonable physical boundary con-
ditions. One can thus obtain rotational-vibrational energies Wγ vN and the radial
eigenfunctions Rγ vN (R). It is, however, useful to simplify the problem even fur-
ther.
One exploits again the different time and energy scales: The molecule oscillates
rapidly in comparison to the rotation, as we have seen. Thus, as a good first ap-
proximation, one may assume the nuclear distance R to be fixed. As sketched in
Fig. 3.10 the molecule may assume any arbitrary polar Θ and azimuthal angle Φ
in space. One may identify this fixed value of R with the equilibrium distance in
3.3 Nuclear Motion: Rotation and Vibration 153
ω
y
MB Φ
x
the lowest vibrational state. The S CHRÖDINGER equation (3.31) may then indeed
be fully separated, and we obtain the eigenvalue equation for the rigid rotor:
N2
rot YN MN (Θ, Φ) =
H YN MN (Θ, Φ)
2M̄R 2
= WN YN MN (Θ, Φ). (3.36)
2
WN = N(N + 1) = hcBN(N + 1) (3.37)
2M̄R 2
with the rotational constant (unit cm−1 )
2 1 2 1
B= = , (3.38)
2M̄R 2 hc 2I hc
I = M̄R 2 being the molecular moment of inertia. The resulting energy scheme is
sketched in Fig. 3.11. In wavenumbers the rotational energies are simply
The rotational constant B is (with the reduced mass known) a very sensitive measure
for the bond length R0 of the molecule. Table 3.4 summarizes typical values of
important molecules. The subscript in I0 , B0 and ω0 indicates that the constants
given here refer to the lowest vibrational state. We see that spectra from the infrared
to the microwave spectral region are expected.
We must emphasize at this point that the image of a diatomic molecule as a rotat-
ing dumbbell has now to be replaced by the quantum mechanical picture: The spher-
Table 3.4 Rotational and vibrational constants and corresponding absorption wavelengths for sev-
eral characteristic diatomic molecules. Note that only molecules with a finite dipole moment are
infrared active
Molecule 2 (2I0 )−1 / eV B0 / cm−1 λrot ω0 / eV λvib / µm IR active
H+
2 3.641 × 10−3 29.4 170 µm 0.2714 4.568 no
H2 7.356 × 10−3 59.32 84 µm 0.5156 2.405 no
D2 3.708 × 10−3 29.90 167 µm 0.37095 3.342 no
LiH 9.184 × 10−4 7.4065 675 µm 0.16853 7.357 yes
HCl 1.2945 × 10−3 10.440 479 µm 0.3577 3.466 yes
N2 2.467 × 10−4 1.9896 2.51 mm 0.2888 4.292 no
O2 1.7827 × 10−4 1.4377 3.48 mm 0.19295 6.426 no
CO 2.384 × 10−4 1.9225 2.60 mm 0.26573 4.666 yes
NO 2.103 × 10−4 1.696 1 2.95 mm 0.23260 5.330 yes
Na2 1.913 × 10−5 0.15427 3.24 cm 0.01955 63.4 no
NaCla 2.6938 × 10−5 0.21725 2.30 cm 0.0448 27.68 yes
Cl2 3.02 × 10−5 0.243 2.05 cm 0.0687 18.0 no
As far as not otherwise mentioned according to H UBER and H ERZBERG (1979)
a For the isotopologue 23 Na 35 Cl according to R AM et al. (1997)
N= 0 N = 1 py N = 1 pz N = 1 px
N = 3 MN = 3
real states
Fig. 3.12 The rigid rotor in space: the probability for finding the molecular axis aligned at Θ and
Φ, is given by the absolute square of the spherical harmonics (or linear combinations of these);
here shown for N = 0 and N = 1, the latter having been excited by linearly polarized light; also
shown is the distribution for N = 3, MN = 3 as an example where the molecule has been excited
by circularly polarized light
ical harmonics YN MN (Θ, Φ) describe the probability amplitudes for the angles Θ
and Φ, and |YN MN (Θ, Φ)|2 sin ΘdΘdΦ is the probability to find the molecular axis
aligned between (Θ, Φ) and (Θ + dΘ, Φ + dΦ). We have discussed this already
in detail in Sect. 2.5.3, Vol. 1. As |YN MN |2 does not depend on Φ this probability
is rotational symmetric around the Z-axis. Alternatively the real basis YN |MN | (Θ)
may represent the excitation probabilities more conveniently – if linearly polarized
light has been used for excitation into the NMN states. In Fig. 3.12 the correspond-
ing probabilities are shown for N = 0 and N = 1. The image of a rotating molecule
3.3 Nuclear Motion: Rotation and Vibration 155
2
kB Trot = = Bhc. (3.40)
2I
This rotational temperature plays an important role when determining the specific
heat capacity as we shall see in a moment. Typical values are given in Table 3.5.
In analogy to n levels in the atomic case ( being the orbital angular momen-
tum), the degeneracy of rotational levels is gN = 2N + 1. The population probability
of a rotational level N is given by a B OLTZMANN distribution
gN WN (2N + 1) N (N + 1)Trot
w(N, T ) = exp − = exp − , (3.41)
ZN kB T ZN T
156 3 Diatomic Molecules
w(N) w(N)
ortho H2
CO 0.6
0.06
0.04
(a) 0.4 (b)
0.2 para H2
0.02
0.00 N 0.0 N
0 10 20 30 0 1 2 3 4 5
Fig. 3.13 Relative population of rotational levels N in diatomic molecules at room temperature
(293 K) in thermodynamic equilibrium: (a) CO – nuclear spin statistics is irrelevant in this case
since the two nuclei are distinguishable particles; (b) for H2 (nuclear spin I (1 H) = 1/2) – here
states with even and odd N are populated differently (red: para H2 ↑↓, grey: ortho H2 ↑↑), see
text
w(T )
0.5
para H2 (↑↓)
0.25
0.0
0 100 200 T /K
is symmetric in respect of reflection at a plane through the nuclear axis but also in
respect of exchange of the nuclear coordinates. The nuclear spin function may be a
singlet or a triplet (analogous to the electron states in excited atomic He, see Chap. 7,
Vol. 1). To obtain an overall antisymmetric wave function for even rotational states
(N = 0, 2, 4 . . . ) the nuclear spin function must be antisymmetric, i.e. be a singlet.
One speaks of para hydrogen (p-H2 ). In contrast, ortho hydrogen (o-H2 ) is char-
acterized by rotational states with N = 1, 3, 5, . . . belong to nuclear spin triplets.
The population density (3.41) has to be multiplied then by the spin state degeneracy
gS = (2S + 1), i.e. with gS = 3 and = 1 for triplet and singlet states, respectively.
The resulting populations for p-H2 and o-H2 at room temperature are depicted in
Fig. 3.13(b).
Transitions between p-H2 and o-H2 are very improbable under normal conditions
in a gas. One may almost speak of two different species which are stable over days,
once generated. In thermodynamic equilibrium, the respective probabilities w(T )
to find either species is shown in Fig. 3.14 as a function of temperature. For very
low temperatures only para H2 is found (rotational ground state) while at higher
temperatures the ratio between p-H2 and o-H2 approaches the value 1:3 reflecting
the ratio of the spin state degeneracy gS .
One can generate para H2 from normal H2 gas by cooling it below 20 K and
bringing it into contact with a catalyzer, e.g. Fe III containing substances, which
support collision induced singlet-triplet transitions (magnetic interaction), and thus
achieving thermodynamical equilibrium. As shown in Fig. 3.14 for low tempera-
tures this equilibrium corresponds to pure para-H2 . After this process one heats the
gas carefully up again without catalyzer(!). Only states of the para system will then
be populated thermally, e.g. states with N = 0, 2, 4 . . . . It takes many hours, even
days before the thus prepared p-H2 (without catalyzer) returns into thermodynamic
equilibrium.
As already mentioned, nuclear spin statistic plays an important role in the inter-
pretation of rotationally resolved electronic or R AMAN spectra. We shall come back
to this in Sect. 5.6.6.
158 3 Diatomic Molecules
where gi is the degeneracy of level i with the energy Wi . The summation is over all
energy levels, if necessary, it has to be replaced by an integration and gi becomes
the density of states. The average internal energy of a system with N molecules
may be expressed in terms of the partition function as
∞
N W
− i N kB T 2 ∂Z ∂ ln Z
U = Wi gi e kB T = N = kB T 2 .
Z Z ∂T ∂T
i=1
From this one derives the molar heat capacity (specific heat capacity per mol at
constant volume). Setting N kB = R (general gas constant) it is given by
∂U ∂ 2 ∂ ln Z ∂ RT 2 ∂Z
CV = = RT = . (3.44)
∂T V =const ∂T ∂T ∂T Z ∂T
From this expression we first derive the specific heat of an ideal gas due to trans-
lational motion. The partition function of a freely moving atomic particle without
internal energy is given by the M AXWELL -B OLTZMANN distribution:
∞
mv 2 π kB T 3/2
Z(T ) ∝ v 2 exp − dv = .
0 2kB T 2 m
When inserting this into (3.44) we obtain the well known contribution from
CV 3
translational motion, = , (3.45)
R 2
reflecting the equipartition law with an average energy kB T /2 per molecule for each
of the three translational degrees of freedom.
3.3 Nuclear Motion: Rotation and Vibration 159
3.5 Cv / R 3.5
H2
CO vi ration
3.0 thermal 3.0
T/K
eq ili ri m (c)
2.5 2.5 2.5
CO rotation 0 2000 4000 6000
(a)
2.0 H2 2.0
( )
3:1 ortho-para
mixt re
1.5 1.5
0 200 400 0 10 20
Fig. 3.15 Molar heat capacity CV for diatomic molecules (in units of the general gas constant R).
At very low temperature only the translational degrees of freedom are active (CV = 1.5R). The
rotational contribution rises up to R; (a) for H2 thermal equilibrium conditions are compared with
a mixture of ortho to para H2 of 1:3 as found at higher temperatures; (b) shows the rotational
contribution for CO and (c) illustrates how CV rises again at higher temperatures due to vibrational
excitation, in the limit again by R
∞
N (N + 1)Trot
ZN (T ) = gS (N )(2N + 1) exp − .
T
N =0
and with (3.44) the contribution of vibrational excitation to the molar heat capacity
is
CV Tvib 2 exp(−Tvib /T ) ω0
= , with Tvib = . (3.47)
R T [1 − exp(−Tvib /T )]2 kB
In the limit of high temperatures T
Tvib the denominator becomes (Tvib /T )2 and
again CV → R (we remember the two degrees of freedom for vibration mentioned
in Sect. 1.3.3, Vol. 1, i.e. per molecule 2 × kB T /2 energy).
For multi-atomic molecules with several vibrational degrees of freedom, (3.47)
is the starting point for evaluating the specific heat capacity. One has to sum (3.47)
over all vibrational frequencies. If the molecule possesses a quasi-continuum of
vibrational levels (e.g. a large organic molecule) one has to multiply (3.47) with the
density of states and to integrate over all frequencies. For solid state materials the
lattice contribution to the specific heat is derived in just the same manner (theories
of E INSTEIN and D EBYE).
3.3.5 Vibration
Vibrational energies are typically two to three orders of magnitude larger than ro-
tational energies (see Table 3.4), so that one may separate the two forms of motion
in a 1st order approximation as indicated above. Typically, the anharmonicity of the
potentials is not too dramatic, so that for low vibrational excitation one may indeed
assume that the average bonding lengths is constant, R = R0 , and the rotational
energy derived for the rigid rotor can just be inserted into the radial S CHRÖDINGER
equation (3.34) for the nuclear motion. One replaces
2 2
N (N + 1) in (3.35) by WN = N (N + 1) = B0 hcN(N + 1).
2M̄R 2 2M̄R02
The total energy Wγ vN of the rotating oscillator may thus be understood as the sum
of this vibrational energy Wγ v in the potential Vγ (R) and the rotational energy WN .
One has to solve the radial wave equation without centrifugal potential:
2 d 2
− + Vγ (R) Rv (R) = Wγ v Rv (R). (3.48)
2M̄ dR 2
More explicitly, the total energy of a ro-vibrational state with vibrational and
rotational quantum numbers v and N , respectively, is given by
V0(R) R0(1)
W0υN R
R0(0)
v =1
N= 4
3
2
v =0 ħω0 /2 1
0
potential minimum
Vγ (R0 ) refers here to the minimum of the potential in the electronic state γ , while
in the most simple approximation (harmonic oscillator + rigid rotor) the energy of
the nuclear motion is given by
1
WvN = Wv + WN = ω0 v + + B0 hcN(N + 1). (3.50)
2
Qualitatively we obtain the picture shown in Fig. 3.16: for bound electronic states
(ground state and several excited states), γ = 0, 1, 2 . . . , the molecular potentials
(γ )
Vγ (R) are sketched with their equilibrium distances R0 . The nuclear motion oc-
curs in these potentials. The energies of vibration and rotation WvN = Wv + WN are
just added to the respective electronic energies at equilibrium distance.
On an enlarged scale Fig. 3.17 illustrates for one electronic state the vibrational
and rotational energies WvN = Wv + WN according to (3.50). Real values for vibra-
tional ω0 and rotational B0 hc = 2 /(2I0 ) energies have been reported in Table 3.4
for a selection of molecules.
162 3 Diatomic Molecules
Table 3.6 Nonlinearity parameters for our set of diatomic reference molecules: ωe xe (Anhar-
monicity), αe (change of B with vibrational state) and centrifugal term De in comparison to the
harmonic vibrational frequency ωe
Molecule ωe / cm−1 ωe xe / cm−1 Be / MHz αe / MHz De / kHzb
H+
2 2321.7 66.2 905400 50370
H2 4401.21 121.33 1824330 91800
D2 3115.50 61.82 912660 32336
LiH 1405.65 23.20 225258 6491
HCl 2990.946 52.8186 317582 9209
N2 2358.57 14.324 59906 519
O2 1580.19 11.98 43100 477
CO 2169.756 13.288 57908 524.8 184
NO 1904.2 14.075 50121 534 34
Na2 159.124 0.7254 4638.0 26.19
NaCla 364.684 1.776 6537.37 48.709 9.3506
Cl2 559.7 2.67 7319.5 45.5
If not mentioned otherwise according to H UBER and H ERZBERG (1979), L OVAS et al. (2005)
a For the isotopologue 23 Na35 Cl according to R AM et al. (1997)
b Not to be confused with the minimum potential energy De
Of course, the simple model of a harmonic oscillator with a rigid rotor is only a
beginning of a realistic description. For a given potential Vγ (R) one may (still as-
suming a rigid rotor) solve the radial S CHRÖDINGER equation (3.34) without prob-
lems numerically – using the standard recipes for stable solutions as for the bound
electron in atoms. Solutions Wv found in that manner may be expanded for N = 0
and any anharmonic potential into a series of the type
1 1 2
G(v) = Wv / hc = ωe v + − ω e xe v + + ··· . (3.51)
2 2
For spectroscopic reasons one typically gives the term positions in wavenumbers,
with the harmonic oscillation frequency (in wavenumbers)
Rc Fr =−k(Rc−R0)
For a M ORSE potential according to Sect. 3.2.5 one finds (here without proof)
2 ω02 2 hc
hcωe xe = = a2 or ωe xe = ωe2 . (3.53)
4De 2M̄ 4De
We recall again, that in any potential well the number nmax of bound states is fi-
nite. Specifically for a M ORSE potential one finds that nmax ≤ 2De /ω0 − 1. For
instance, in the case of CO the number of bound vibrational states is 41.
So far we have used the rigid rotor model. The radial S CHRÖDINGER equation (3.34)
was solved for fixed nuclear distance R ≡ R0 , with the centrifugal term 2 N (N +
1)/(2M̄R 2 ) considered to be constant. This assumption allows one to treat rotation
and vibration as separate, uncoupled forms of motion. Rotation is just an additive
constituent of the total energy Wγ νN . For a real molecule one has to account for the
effective potential (3.35)
2 N(N + 1)
Veff (R) = + Vγ (R)
2M̄R 2
as a function of R and to solve the full radial S CHRÖDINGER equation (3.34). This
leads to further modifications of the energy.
One characteristic effect is centrifugal distortion by rotation, which leads to a
coupling of rotation and vibration. Without explicitly solving the radial equation we
try to estimate this effect. Rotational energy (3.37) leads to a centrifugal force
dWN 2 N(N + 1)
Fc = − = .
dR M̄R 3
As sketched in Fig. 3.18 it is balanced by the restoring force of the harmonic poten-
tial,
with the force constant k = M̄ω02 according to (3.20). At the new equilibrium posi-
tion Rc we must have Fc (Rc ) + Fr (Rc ) = 0, hence
2 N(N + 1)
M̄ω02 (Rc − R0 ) = and
M̄Rc3
2 N(N + 1) 2 N (N + 1)
Rc ≈ R0 1 + > R 0 with
1.
M̄ 2 ω02 R04 M̄ 2 ω02 R04
The molecule stretches with higher rotational energy and the rotational constant
becomes smaller. The overall rotational energy of the stretched molecule is the sum
of rotational energy and potential (stretch) energy:
2 N(N + 1) M̄ω02
WN = + (Rc − R0 )2 .
2M̄Rc2 2
2 4
WN = N(N + 1) − N 2 (N + 1)2 + · · ·
2M̄R02 2M̄ 3 ω02 R06
Bhc N (N + 1) − De hc N 2 (N + 1)2 . (3.54)
1 4 1 4(Bhc)3 4B 3 4B 3
De = = = = .
hc 2ω02 (M̄R02 )2 hc 2 ω02 (Wv / hc)2 ωe2
Comparison for CO, NO and NaCl on the basis Table 3.6 shows rather good agree-
ment with the spectroscopically determined values of De .
A precise treatment of the S CHRÖDINGER equation (3.34) requires at least one
more term. Typically one writes:
The spectroscopic accuracy today is often so high that expansions up to the 6th
or even 10th power of (v + 12 ) become meaningful (see e.g. M ANTZ et al. 1971;
L E ROY 1970).
This formulation has first been used by D UNHAM (1932) already in the early
years of quantum mechanics. D UNHAM coefficients Yik – not to be confused with
the spherical harmonics – have become more and more accepted during the past
decades. For comparing different spectroscopic literature it is useful to note the fol-
lowing equivalences and relations:
Although these are approximations, they are sufficient for most comparisons. The
exact expressions are given in the original work of D UNHAM (1932). We only note
here that due to anharmonicity, the vibrational terms are slightly displaced in respect
of the potential minimum (M ANTZ et al. 1971) by
G(v) + Y00
with Y00 = Be /4 + αe ωe /12Be + (αe ωe )2 /144Be3 − ωe xe /4.
166 3 Diatomic Molecules
Important is also the dependence on the reduced mass M̄, which allows one to com-
pare spectra for different isotopologues.
Section summary
• We have developed an understanding of nuclear motion. In summary, energy
levels of a diatomic molecule may be expressed as a sum of electronic, vi-
(γ )
brational and rotational energy, Wγ vN / hc = Vγ [R0 ]/ hc + ωe (v + 12 ) +
B0 N (N + 1) to 1st order approximation (rigid rotor with rotational constant
B0 and harmonic oscillator with eigenfrequency ωe , both given in wavenum-
bers).
• We have discussed higher order approximations, obtaining reasonable es-
timates for the changes due to centrifugal stretch and anharmonicity in a
M ORSE potential. Even more general one expands the total energy (3.59) as
a series in powers of [v + 1/2]i [N(N + 1)]k with the so called D UNHAM
coefficients.
• We also have made a brief excursion into statistical thermodynamics in
Sect. 3.3.3, acquainting ourselves with the population of rotational energy
levels at low temperatures. Based on this information we have derived in
Sect. 3.3.4 specific heat capacities CV (T ) for diatomic molecules as a func-
tion of temperature T . At very low temperature only the kinetic degrees of
freedom can be excited and CV = 3/2R, while at temperatures T Trot =
hcB/kB the specific heat capacity rises to 5/2R. At still higher temperatures,
around T Tvib = ω0 /kB vibration can be excited and the specific heat ap-
proaches its limiting value CV = 7/2R.
where again r represents all electronic, R all Nnu nuclear coordinates, and e the
unit polarization vector of the radiation. In principle this expression holds for any
number of nuclei. Specifically for a diatomic molecule with Nnu = 2, the electric
field acts on a distance weighted, average charge (see Fig. 3.9)
= (ZA RA + ZB RB )/(RA + RB ).
Z (3.62)
3.4 Dipole Transitions 167
The transition probability into a state |a = |γ vN M from a state |b = |γ v N MN
is proportional to the squared dipole transition matrix element
Dba = γ v N MN |D|γ vN M = Ψb∗ (r, R)
D(R, r)Ψa (r, R)d3 Rd3 r.
The contributions for the other two components ξi and ηi disappear in this symmetry
when averaging over all r i . Obviously
of the molecule in the state |γ v (see e.g. Appendix F.2 in Vol. 1, specifically
Eq. (F.10)). With this abbreviation, the angular part of the integration in the dipole
transition matrix element is given by
DN N = Dγ v YN∗ M (Θ, Φ)eR · eYN MN (Θ, Φ) sin ΘdΘdΦ. (3.65)
N
The important message from this expression is: pure rotational spectra can only
be observed for molecules with a permanent dipole moment, but not for homonu-
clear molecules such as H2 or N2 . Table 3.4 explicitly emphasizes the infrared active
168 3 Diatomic Molecules
heteronuclear molecules. The selection rules are derived in analogy to those for
and m in the case of atoms, described in detail in Sect. 4.4, Vol. 1. We just have
to replace the electronic dipole transition operator D = e · r there, by
D = Dγ v e R · e
here. The permanent dipole moment of the molecule D = −eDγ v eR is parallel to
the molecular axis eR . As in the atomic case, the three components of the molecular
unit operator eR are now expressed in terms of the renormalized spherical harmon-
ics eR = {C1−1 (Θ, Φ), C10 (Θ, Φ), C11 (Θ, Φ)}. We can now evaluate (3.65), which
leads to selection rules in full analogy those for atomic transitions between different
m states:
N = ±1 and MN = 0, ±1. (3.66)
For absorption, N = N + 1 ← N , the transition energy in wavenumbers is
WN +1 − WN
ν̄rot = = B (N + 1)(N + 2) − N(N + 1) = 2B(N + 1). (3.67)
hc
For induced emission, N − 1 ← N , one finds correspondingly
ν̄rot = (WN − WN −1 )/ hc = 2BN . (3.68)
The spectral lines for a rigid rotor all have the same distance 2B in wavenumbers.
The measurement of such a spectrum is, however, not trivial since it involves a broad
spectral range from microwave to the FIR spectral range. Tabulated values usually
refer to a number of different measurements.
In Fig. 3.19 we show an artificially synthesized spectrum of rotational lines for
CO in the vibrational ground state v = 0, as derived from transition frequencies
published by L OVAS et al. (2005). It is quite instructive to consider the intensities
in these line spectra. The absorption probability RN MN N MN for specific transition
|N MN ← |N MN between the orientation states is proportional to the square of
the rotational dipole matrix element (3.65), which according to (4.79) and (C.28) in
Vol. 1 is (for a polarization vector e = eq )
2
RN MN N M ∝ |Dγ v |2 N MN |C1q |NMN (3.69)
2 2
N 1 N N 1 N
= |Dγ v | 2N + 1 (2N + 1) ×
2
.
0 0 0 MN q MN
The C1q (Θ, Φ) are the renormalized spherical harmonics characterizing the polar-
ization (q = 0, ±1) of the incident light, and N = N ± 1 is the rotational quantum
number of the upper state. The 3j symbols (: : :) have been introduced in Ap-
pendix B, Vol. 1.
In view of the fact that many rotational states are populated, as discussed in
Sect. 3.3.3, we also have to include induced emission in our evaluation of the line
spectrum. The population probabilities for the specific rotational substates |N MN
and |N MN are given by the B OLTZMANN factor7 exp(−N (N + 1)Trot /T ), where
relative transmission
data given by L OVAS et al.
(2005). The lines have a
distance of 2B = 3.84 to
3.76 cm−1 corresponding to
the rotational constant
B0 = 1.9225 cm−1 according
to Table 3.4. The absorption
has been derived from (3.71)
0 20 40 60 80 100
ν rot / cm-1
T is the temperature of the sample and Trot the rotational temperature of the
molecule studied (see Table 3.5).
Now, the probability RN MN N MN for the absorption process |N MN ← |N MN
is exactly equal to the induced transition probability RN MN N MN for |N MN →
|N MN . On the other hand, the B OLTZMANN factor for the upper rotational level
is somewhat smaller than for the lower states, so that for each individual rotational
state the relative net absorption for the incident radiation I is
I (N MN N MN ) 2
∝ |Dγ v |2 N MN |C1q |NMN (3.70)
I
× exp −N(N + 1)Trot /T − exp −N N + 1 Trot /T .
For the overall absorption we have to sum over all upper and lower orientation
states. With (3.69) we obtain the relative absorption signal for a single rotational
line N ← N :
I (N N ) I (N M NMN )
= N
I
I
MN MN
∝ |Dγ v |2 exp −N(N + 1)βr − exp −N N + 1 βr
2 2
N 1 N N 1 N
× 2N + 1 (2N + 1) × .
0 0 0
M N q MN
MN MN
Exploiting the orthogonality relation (B.42), Vol. 1, and with (B.53), Vol. 1 for eval-
uation of the remaining 3j symbol we finally obtain a simple expression for the
overall absorption probability:
I (N N ) N + 1 −N (N+1)Trot /T
∝ |Dγ v |2 e − e−(N +2)(N +1)Trot /T . (3.71)
I 3
This is presented in Fig. 3.19 for the vibrational ground state of CO.
170 3 Diatomic Molecules
/ cm-1
distortion in CO. Shown is a
quadratic fit to the 3.84
CO
/ (N +1)
spectroscopic data from v=0
L OVAS et al. (2005). The 3.83
ordinate intercept gives 2B,
(WN+1 - WN )
while the parameters De and 3.82
H in (3.72) are derived from
the slope and the slight 3.81
curvature of the line
connecting the data points 0 400 800 1200 (N+1)2
In a more precise analysis of the data one notices, however, that with increasing
N the distance between neighbouring lines decreases slightly. This documents the
centrifugal distortion discussed in Sect. 3.3.6. For a quantitative evaluation of the
experimental material we write the rotational energy
WN / hc = BN (N + 1) − De N 2 (N + 1)2 + HN 3 (N + 1)3 .
WN +1 − WN
= 2B − (4De − 2H)(N + 1)2 + 6H(N + 1)4 . (3.72)
hc(N + 1)
Figure 3.20 shows this expression, plotted as a function of (N + 1)2 , for CO in the
v = 0 state. To a first approximation this gives a straight line with the ordinate inter-
cept 2B while the slope is −4De . The parameters derived from this fit correspond
to the values given in Table 3.6. The curvature of the fit even allows one to derive
the parameter H = 0.1715(4) Hz.
Today rotational lines are determined with highest precision by F OURIER trans-
formation IR spectroscopy (FTIR). An example is shown in Fig. 3.21. We shall
describe FTIR spectroscopy in Sect. 5.3.2 in detail.
We are now prepared to treat the S TARK effect in polar molecules, i.e. to evaluate the
energies of rotational levels in an external static electric field. In Sect. 8.2, Vol. 1 we
have discussed the S TARK effect in atoms at some lengths. We have found there that
it is a weak effect for low lying electronic terms, while for highly excited RYDBERG
states it may become substantial due to interaction with other, near neighbouring
RYDBERG states. For polar molecules we expect the largest S TARK effect for the
lowest rotational levels, since the energies (3.37) rise quadratically with the rota-
tional quantum number, and hence neighbouring levels interact less and less as N
increases.
Let the external electric field E be parallel to the molecular z-axis (E eR ). With
Vel (R) = eE · eR Dγ v and (3.69) the interaction matrix elements (8.53), Vol. 1 are
Using the same formulas (8.58) and (8.59), Vol. 1 as in the atomic case we have for
each rotational N > 0 two nonvanishing matrix elements
(0)
N MN |C10 |N MN 2
WN MN = WN MN − WN MN = |eE Dγ v |2 . (3.75)
WN − WN
N
|eE Dγ v |2
WN MN = f (N, MN ), (3.76)
Bhc
1 N(N + 1) − 3MN2
with f (N, MN ) =
2 N(2N + 3)(2N − 1)(N + 1)
and = −1/6 for N = 0.
103 f(N,M )
N=4 2
0
3
-5
-10 4
To obtain a feeling for the order of magnitude of the S TARK effect, we consider
again CO as an example. In its ground state it has a permanent dipole moment
|D| = eDγ v = 0.3662 × 10−30 C m and its rotational constant is B0 = 1.9225 cm−1 .
For an electric field strength of 1 kV / cm, which is still conveniently achievable in
the laboratory, one obtains a relative S TARK shift
WN MN |eE Dγ v |2
= 2 f (N, MN ) 10−6 f (N, MN ). (3.77)
WN MN B0 h2 c2
To evaluate the dipole transition matrix elements (3.63) for vibrational excitation
(or de-excitation) we follow Sect. 3.4.1 – still within one electronic state. We have
to replace the permanent dipole moment Dγ v = −eDγ v by the transition dipole
moment and (3.65) now becomes
Dγ v ←v = R∗γ v (R)Dγ (R)Rγ v (R)dR
(3.78)
N
2 3
−
with Dγ (R) = ZR ζi φγ (r; R) d r i . (3.79)
i=1
Once more, we have exploited the symmetry of the diatomic molecule and chosen
electronic coordinates with ζi eR : the dipole moment, −eD, and its derivatives are
parallel to the molecular axis. We may thus expand Dγ (R) around the equilibrium
3.4 Dipole Transitions 173
distance R0 :
∂ Dγ
Dγ (R) = Dγ (R0 ) + (R − R0 ) + · · · . (3.80)
∂R R0
This has to be inserted into (3.78) for a transition v ← v. The first (constant) term
disappears due to the orthogonality of the radial wave functions Rv (R). Thus, the
linear term dominates for dipole induced (E1) vibrational transitions:
∂ Dγ
Dv ←v = R∗v (R)(R − R0 )Rv (R)dR + · · · . (3.81)
∂R R0
When the demand for precision is higher, one has to account for the anharmonic-
ity of the potential and possibly also has to include higher terms in the series ex-
pansion (3.80). In consequence the selection rule v = ±1 does no longer strictly
apply, transitions with v = ±2, ±3, . . . become weakly allowed, and the separa-
tion of the energy levels changes with v.
The population of vibrational states in thermodynamic equilibrium is determined
by the B OLTZMANN distribution:
g(v) Wv Tvib ω0
Nv = exp − exp − with Tvib = . (3.82)
Zv kB T T kB
Here Zv is the partition function, for the harmonic oscillator given by (3.46). Nv
decreases monotonically with v, since in contrast to rotation the vibrational states
of a diatomic molecule are non-degenerate, i.e. g(v) = 1 holds. We note that at
room temperature for small diatomic molecules ω0
kB T 0.025 eV holds (cf.
Table 3.4 for some examples). Hence, at normal conditions and thermodynamic
equilibrium essentially only the ground vibrational state (v = 0) is populated.
8 One readily verifies this for the lowest levels by inserting the H ERMITE functions from Table 3.1
into (3.81).
174 3 Diatomic Molecules
P branch R branch
band origin
19
N'' 20
transmission
20
N' 19
1 0
1← 0 2← 0
0 1
×100 CO
2000 4000 v' = 1 ← v'' = 0
Fig. 3.23 Vibration-rotation band for CO, in the electronic ground state for the v = 1 ← v = 0
transition, at a temperature T = 293 K, simulated with HITRAN (ROTHMAN et al. 2009) data for
the R and P branch; the inset shows also (weak) second harmonics lines, v = 2 ← v = 0
However, pure vibrational transition are not allowed! Simultaneously (3.65) has to
be applied (with Dγ v replaced now by Dγ v ←v ). And as no E1 transitions occur with-
out change of the rotational quantum number N (parity conservation). In summary,
we have now three selection rules:
According to H ERZBERG (1989), the upper levels are designated with v N , the
lower ones with v N . As an introductory example, Fig. 3.23 shows the infrared
absorption spectrum of CO in the ground state (v = 1 ← v = 0). One recognizes a
typical band structure with many lines. The inset (not rotationally resolved) gives a
feeling for the importance of higher harmonics, showing the v = 2 ← v = 0 band
which amounts to less than 1 % of the total absorption. The spectrum (so called
sticks spectrum) shown has been generated synthetically from data of the HITRAN
data bank. In this impressive collection of molecular spectra more one finds nearly
three million spectral lines of presently 47 molecules with 120 isotopologues – a real
treasure for analysts and spectroscopists, who want to search for traces of molecules,
e.g. in the earth atmosphere.
3.4 Dipole Transitions 175
Note that due to anharmonicity the band origin is not exactly at ωe , as one recog-
nizes in Fig. 3.23 for the CO spectrum: ωe = 2169.756 cm−1 according to Table 3.6,
however, the origin (between P and R branch) is located at 2143.24 cm−1 , corre-
sponding to (3.86).
Figure 3.24 illustrates schematically how the band structure arises due to rotation.
In respect of a (hypothetical) pure vibration spectrum (Q branch) with Q(N) = ωe
the lines in the R branch have higher (R(N ) > ωe ), those in the P branch lower
energies (P (N ) < ωe ). The Q branch with N = 0 is missing due to the overall
parity conservation. Somewhat more precisely: this is due to the dipole moment
Dγ = −eD(R)eR being parallel to the molecular axis. Its changes are, according
to (3.78)–(3.81), responsible for vibrational excitation – one speaks about “par-
allel” transitions. They are the only ones possible in a diatomic molecule. How-
ever, already for a triatomic linear molecule this may be different as we shall see in
Sect. 4.2.3.9
The intensity distribution in the vibration-rotation absorption bands may be de-
rived in complete analogy to our considerations for pure rotational transitions in
Sect. 3.4.1. It arises again essentially from the thermal population. However, in-
duced emission usually does not play a role since initially the final vibration state is
practically unpopulated.
ΔN = 0
N'' = 5
4
v'' = 0
3
1 2
Q branch 0
P branch R branch
_
ν
P branch R branch
2626
2652
3073
3059
2678
3045
2703
band origin
3030
2728
3015
2752
2866
transmission
2906
2998
2776
2844
2799
2981
2926
2822
Fig. 3.25 Vibration-rotation spectrum of HCl, simulated according to HITRAN (ROTHMAN et al.
2009) for absorption at room temperature (sticks spectrum)
3.4 Dipole Transitions 177
the spectra become quite complicated, since the term distances are no longer equal.
When evaluating experimental data one uses a nice trick to separate the rotational
constants of the upper and lower vibrational states: in suitably chosen differences
of spectral lines, one or the other of these constant drops out. One finds from (3.56)
and (3.58):
R(N − 1) − P (N + 1) = 4B − 6De (N + 1/2) − 8De (N + 1/2)3 (3.88)
R(N ) − P (N) = 4B − 6De (N + 1/2) − 8De (N + 1/2)3 . (3.89)
By plotting the data correspondingly, one obtains by suitable fits the four parameters
B and De for the upper, and B and De for the lower state – in the same manner as
discussed in Sect. 3.3.6 for pure rotational spectra.
By measuring several vibrational states one can obtain in a similar manner the
anharmonicity. In a so called B IRGE -S PONER plot the differences between experi-
mentally determined energies of adjacent vibrational lines (band origin) are plotted
as a function of (v + 1). With (3.58) up to second order
G v + 1 − G v = ωe − 2ωe xe v + 1 + · · · , (3.90)
holds – which is of course identical to (3.86). The slope of this curve gives the anhar-
monicity ωe xe , the intercept with the ordinate yields the ground state frequency ωe .
Specifically for a M ORSE potential one may estimate the bond energy with the thus
determined parameters from (3.53)
ωe2
De = hc .
4ωe xe
As for the pure rotational lines, different isotopologues have different vibration-
rotation frequencies. In addition to the change
of the moment of inertia I0 , now the
shift of the vibrational frequency ω0 = k/M̄ is significant. As a rule, the (abso-
lute) isotope shift of the vibrational transitions is larger than that for the rotational
absorption line.
Before turning to the ab initio (i.e. the quantum mechanical) computation of molec-
ular potentials we briefly want to introduce a standard method to determine poten-
tials directly from the experimentally measured spectra. The method developed by
RYDBERG, K LEIN and R EES already in the early years of quantum mechanics, the
so called RKR method, uses measured vibration-rotation spectra in a semiclassical
ansatz. It tries, so to say, to invert the solutions of the S CHRÖDINGER equation for
energies and nuclear wave functions. One has to know the vibrational term ener-
gies G(v) and the rotational constants B(v) as good as possible, for as many v as
possible. They are then considered as continuous functions of v and one determines
178 3 Diatomic Molecules
from these the classical turning points Rmax (v, N = 0) and Rmin (v, N = 0). From
these one may then construct the potential as illustrated in Fig. 3.7. One may show
rigorously that
Rmax = f 2 − f/g + f and Rmin = f 2 − f/g − f (3.91)
Evaluation of these integrals is not completely trivial. The interested reader is re-
ferred to the literature (e.g. M ANTZ et al. 1971; F LEMING and R AO 1972, where
one also finds references to the original work).
Section summary
• E1 transitions between vibrational and rotational states within one electronic
state are studied by absorption in the infrared and microwave region, respec-
tively. Spontaneous emission is not observed at these frequencies as a conse-
quence of the ν 3 factor for spontaneous emission.
• They require a permanent molecular dipole moment (i.e. do not occur in
homonuclear molecules). Vibrational transitions depend on the change of the
dipole moment with distance.
• Pure rotation spectra in polar molecules obey selection rules for the angular
momentum, N = ±1 and its projection MN = 0, ±1. To 1st order the lines
are equally spaced by 2B, with B being the rotational constant. Centrifugal
distortion adds small terms ∝ (N + 1)2 and more.
• A quadratic S TARK effect in polar molecules is weak but observable due to
the high accuracy of microwave spectra. In intense laser field it becomes im-
portant.
• In harmonic approximation, only v = ±1 transitions are allowed. Some
weak higher harmonic lines may be observed due to anharmonicity.
• Vibration-rotation bands have, according to (3.85)–(3.87) in principle three
branches, P , Q, and R for N = −1, 0, and 1, respectively (in absorption).
The Q branch is not allowed in diatomic molecules within one electronic state.
• Relatively simple formulas (3.88)–(3.90) can be used for extracting anhar-
monicities from observed spectra. The RYDBERG -K LEIN -R EES method in-
verts in principle the S CHRÖDINGER equation and derives the molecular po-
tential from measured vibration-rotation spectra.
3.5 Molecular Orbitals 179
Let us now discuss the electronic part of the S CHRÖDINGER equation (3.10). It is
most instructive to explain the basic concepts used in typical quantum chemical
methods for the example of H+ 2 , the simplest of all molecules. The extension to
multi-electron systems is relatively straight forward, even though it may become
rather elaborate. With the coordinates defined in Fig. 3.26 the Hamiltonian (again in
a.u.) becomes:
el = − 1 ∇ 2r − 1 − 1 + 1 .
H (3.94)
2 rA rB R
The variational principle has been used already in Sect. 7.2.5, Vol. 1. Even in its sim-
plest form, it provides a good first approximation for the molecular potentials, i.e. for
the electronic energies as a function of R, and for the wave functions. One chooses
a suitably parameterized “trial function” φ(r i ; R) and rewrites the S CHRÖDINGER
equation Hφ = W φ as
∗
φd3 r
φ H
W = min ∗ 3 . (3.95)
φ φd r
In this formulation φ(r; R) doesn’t even need to be normalized. The best eigenvalue
W is obtained by variation of φ (i.e. by changing the parameters which define φ)
such, that W becomes a minimum. Formally, the “functional” W (φ) is minimized
!
by searching for ∂W (φ)/∂φ = 0.
2
φ(r; R) = ci Φi (r j ). (3.96)
i j =1
r
A B
O R
180 3 Diatomic Molecules
such that the coordinates refer to the same origin O as indicated in Fig. 3.26. For
different masses MA and MB , the origin O is shifted towards the heavier mass. With
M = MA + MB the electronic coordinates are
r B = r − (MA /M)R and r A = r + (MA /M)R. (3.97)
(In the case of several electrons r, r A and r B represent all electron coordinates.)
With the LCAO “trial” wave function (3.96) the trial energy becomes
( ∗ ∗ (
c Φ H c k Φk d 3 r
" = (i i ∗ i ∗ ( k 3
> Wtrue ,
i c i Φi k c k Φk d r
where Wtrue is the true eigenvalue. Since the coefficients ci are just (complex) num-
bers, we may interchange summation and integration and obtain
( ( ∗
c ck Hik
" = (i (k i∗ with the (3.98)
i k ci ck Sik
molecular Hamiltonian in the atomic basis Hik = Φi∗ H Φk d3 r, (3.99)
and the so called overlap integral Sik = Φi∗ Φk d3 r. (3.100)
Note, that for normalized AOs, Φi , we have Sii = 1, but Sik vanishes only for or-
bitals i = k which are localized on identical atoms. We have to find now the mini-
mum of ", which is as close as possible to the true energy eigenvalue W within the
given class of functions. At this minimum,
∂" ∂"
= ∗ =0
∂ci ∂ci
must hold for all i. We rewrite (3.98) as
ci∗ ck Hik = " ci∗ ck Sik ,
i k i k
after some reordering for the optimal set {ck }. This homogeneous system of linear
equations has nontrivial solutions only if
det(Hik − "Sik ) = 0. (3.102)
This is called the characteristic equation for ". The solutions "γ (i.e. the roots of
a polynomial) are the sought-after energies of the electronic system. To obtain the
MOs one inserts a specific solution "γ (for a given state γ ) into the system of linear
γ
equations (3.101). The solutions are the coefficients {ck } to be inserted into (3.96).
3.5 Molecular Orbitals 181
The most simple approach to the lowest MOs is to superpose just two 1s atomic
hydrogen orbitals, each centred at one of the two protons:
1/2
1 A B
on proton A: ΦA = Φ1s (rA ) = e−rA /a0
a0 π
1/2
1 A B
on proton B: ΦB = Φ1s (rB ) = e−rB /a0
a0 π
Since ΦA and ΦB are normalized, SAA = SBB = 1 holds and we just have to com-
pute the overlap integral
∗
S(R) = SAB (R) = Φ1s (rA )Φ1s (rB )d3 r (3.103)
indicated in Fig. 3.27. This two centre integral has limiting values S = 0 and
R→∞
S = 1.
R→0
To compute Hik we exploit the fact that ΦA and ΦB are eigenfunctions of the H
atoms. We indicate this in the Hamiltonian (3.94):
A
=H
1
el = − e − 1 1 1
H 2
− + .
2 r rB R
A
B
=H
The so called C OULOMB integral C(R) is positive and −C(R) represents simply
the interaction of the electron charge distributed around nucleus A with the positive
charge of nucleus B – it disappears for large R and compensates for small R the
C OULOMB repulsion of the nuclei. The matrix element HAB = HBA is called the
resonance integral:
∗ el Φ1s (rB )d3 r ∗ 1 1
HAB = Φ1s (rA )H = Φ1s (rA ) − + HB Φ1s (rB )d3 r
R rA
W1s Φ1s (rB )
1 ∗ 1
= W1s + S(R) − Φ1s (rA ) Φ1s (rB )d3 r
R rA
1
HAB = W1s + S(R) − K(R) = HBA . (3.105)
R
K(R) is a kind of exchange integral. For not too small R one finds HAB < 0, since
S(R) decreases with increasing R and the term −K(R) dominates.
In summary, with Sii = 1 we obtain for the characteristic equation (3.102) the
determinant
HAA − " HAB − "S
HAB − "S HAA − " = 0. (3.106)
HAA − HAB
"u = and (3.107)
1−S
HAA + HAB
"g = . (3.108)
1+S
With HAB < 0 this leads to the energy scheme sketched in Fig. 3.28.
Inserting these values for " into the system of Eqs. (3.101) allows one to compute
the coefficients cA and cB :
H2+
We thus find two different (lowest) LCAO-MOs for H+ 2 : one which is symmetric in
respect of inversion r → −r, the so called even or “gerade” state (g) and one which
changes its sign upon inversion, the odd or “ungerade” state (u):
1
φg = √ Φ1s (rA ) + Φ1s (rB ) (3.110)
2(1 + S(R))
1
φu = √ Φ1s (rA ) − Φ1s (rB ) . (3.111)
2(1 − S(R))
Figure 3.29 illustrates these two MOs (more about the g − u symmetry in
Sect. 3.5.4). The electron charge density becomes
−e ∗ ∗
−e|φg,u |2 = Φ1s (rA ) ± Φ1s (rB ) Φ1s (rA ) ± Φ1s (rB ) (3.112)
2(1 ± S)
−e 2 2
= Φ1s (rA ) + Φ1s (rB )
2(1 ± S)
e ∗ ∗
∓ Φ1s (rA )Φ1s (rB ) + Φ1s (rB )Φ1s (rA ) ,
2(1 ± S)
overlap of the atomic orbitals
where the sign ± refers to the (g) and (u) state, respectively. These two wave func-
tions (MOs) have cylinder symmetry around the molecular axis. We summarize their
key properties:
• The φg = 1σg orbital with the energy "g (R) has even inversion symmetry, i.e. the
sign of the wave function does not change upon inversion r → −r. It describes
the electronic ground state of H+ 2 and is a bonding orbital.
• The φu = 1σu∗ orbital with the energy "u (R) has odd inversion symmetry, i.e. the
sign of the wave function changes upon inversion r → −r. It is an antibonding
orbital, indicated by labelling it *.
184 3 Diatomic Molecules
The negative sign of HAB is responsible for the bonding of the symmetric orbital,
hence with (3.105) eventually the exchange integral is
∗ 1
K(R) = Φ1s (rA ) Φ1s (rB )d3 r. (3.113)
rA
Crucial for the bonding of an orbital is the overlap of Φ1s (rA ) and Φ1s (rB ), i.e. the
second term in the charge density (3.112). This is illustrated in Fig. 3.29.
For H+2 the energies "g and "u according to (3.107) and (3.108) may explicitly be
computed as a function of R. This leads to a first approximation for the potentials,
the equilibrium distance R0 , and the bond energy De . If we insert the integrals
(3.104) and (3.105) we find:
W (R)/Eh 2Σ +
W (R )/eV
u exact
3
0.10
1σu* LCAO
2
0.05 1σg LCAO
1
0.00
1 2 3 4 5 R / a0
-1
- 0.05
R0 -2
- 0.10 De 2Σ + exact
g
-3
Fig. 3.30 Potentials for the H+ 2 molecule: dashed red lines give LCAO orbital energies in the sim-
plest form. The full red lines show for comparison the exact potentials according to S HARP (1971).
For the bonding ground state (1σu and 2 Σg+ ) LCAO gives only a rough first order approximation.
For the repulsive state, exact (2 Σu+ ) potential and approximation (1σu∗ ) are practically identic. The
energy zero has been fixed for the dissociated atoms in their ground state
to the fact that the molecular states may have even or odd symmetry. In both cases
the probability to find the electron at one or the other proton is equal. Quantum me-
chanically the two positions can, strictly speaking, not be distinguished. However,
if the nuclei are separated by a collision or a photoinduced dissociation process, at
some point during the separation process the electron has to ‘decide’ at which of the
nuclei it wants to remain for good.
This is illustrated in Fig. 3.31 by ‘snapshots’ as seen in the centre of mass sys-
tem. After the collision one detects the charge exchange by detecting the newly
formed, fast H atoms, which have been scattered into a specific (small) angle θ . In
this particular experiment θ = 3◦ . In the experiment one first selects all particles
scattered at the angle θ by an aperture, thus discriminating scattered from unscat-
tered particles. The respective signal, S(a) + S(b) , is detected by a particle multiplier
(see Appendix B.1). Then one deflects from the scattered particle beam all protons
with the help of an electric field, and detects only the fast H atoms (signal S(b) ). In
186 3 Diatomic Molecules
this manner one determines that a close encounter has happened. The probability
for charge exchange during such an interactions is thus we = S(b) /(S(a) + S(b) ).
At the time of the close encounter one can, in principle, not distinguish whether
the system H+ +H moves on the potential curve belonging to φg (1σg ) or to φu (1σu ).
We have a situation very similar to a “double slit interference experiment”. The
probability amplitudes for both possibilities have to be added and this leads to typi-
cal interference structures.
Let us have a closer look at this process. Prior to the collision (t → −∞, or
R → ∞) the electron is localized at proton A, as sketched in Fig. 3.31(a). With
(3.110) and (3.111) we may express this in the molecular H+ 2 picture:
1
Φ1s (rA ) ≡ √ (φg + φu ). (3.117)
2
If we want to describe the temporal evolution of the system, we have to account for
the different time dependence of the two states involved:
"g (R) "u (R)
φg (R, t) = φg exp −i t and φu (R, t) = φu exp −i t .
The energies "g,u (R) for g and u states (i.e. the potentials shown in Fig. 3.30) split
as the two protons approach each other. Correspondingly, the initial wave function
(3.117) evolves with time:
1
φ(t) √ φg exp(−i"g t/) + φu exp(−i"u t/) .
2
With Wgu = "g (R) − "u (R) and W = ("g (R) + "u (R))/2 we rewrite this:
1 Wgu t Wgu t iW t
φ(t) √ (φg + φu ) cos + i(φg − φu ) sin exp − . (3.118)
2 2 2
3.5 Molecular Orbitals 187
The phase factors sin(Wgu t/2) and cos(Wgu t/2) appear and disappear in opposi-
tion and φ(t) oscillates between
1 1
√ (φg + φu ) Φ1s (rA ) and √ (φg − φu ) Φ1s (rB ).
2 2
Correspondingly, during the close encounter, the electron ‘oscillates’ between pro-
ton A and proton B back and forth.
Strictly speaking, the ansatz (3.118) is valid only if the energy splitting Wgu
between the g and u state is constant. But for a rough guess we may replace Wgu t
by an average value
The probability to find the electron after the collision at A or B is given by the square
of the respective amplitudes. The probability for electron exchange is thus
aWgu 1
we = sin2 π . (3.121)
h v
In the experiment one reads between the maxima or between the minima a differ-
ence of (1/v) 6.6 × 10−7 m−1 s on the inverse velocity scale (with only a slight
variation). This corresponds to aWgu = h/(1/v) 6.27 eV nm. From the poten-
tial energy diagram Fig. 3.30 one estimates an average distance between 1σg and
1σu∗ potential of about Wgu 10 eV and an effective interaction length of about
12a0 = 6.3 nm – in plausible agreement with the experiment. We note, however, that
the first maximum is observed at a phase 2.4 and not at π/2 1.57 as predicted
by (3.122) – which shows the limitations of such a simple model.
(a) (b) -5 0 5 10 15
10 0.5
laser pulse
8
0
amplitude (d)
6
4 0
t / fs
2
0 - 0.2
signal 0 1 2 3 4 ϕc / π - 0.5 A
(c) A
B (e)
laser pulse B
shape charge asymmetry (schematic)
Fig. 3.33 Experiment and model calculation for the dissociation of D+ 2 by phase stabilized, ultra-
short laser pulses (5 fs) according to K LING et al. (2006) (for details see text)
is shown. The light is linearly polarized. In the experiment one selects preferentially
such dissociation processes for which the molecular axis is aligned parallel to the
laser field. Let us think these molecules to be align parallel to the vertical axis in the
paper plane. All three pulses have the same carrier frequency ωc and the same pulse
duration characterized by a Gaussian as discussed in Sect. 1.4.1. They differ only
in respect of the carrier envelope phase. While for φc = 0, 2π etc. the field points
at the maximum of the carrier ‘up’ in respect of the molecule, its direction points
‘down’ for φc = π, 3π etc. For φc = π/2, 3π/2 etc. the highest field strengths are
of equal magnitudes in up and down direction.
The key issue is now, to observe whether the phase has an influence on the dis-
sociation process. The results shown in Fig. 3.33(b) give clear evidence that it does!
Plotted is the asymmetry of the measured ion signal (SA − SB )/(SA + SB ) for ions
which leave the D+ 2 in the direction ‘up’ (SA ) or ‘down’ (SB ), respectively. This
asymmetry (colour code) is plotted as a function of both, the kinetic energy (verti-
cal axis) and the phase φc (horizontal axis). The influence of the phase is amazingly
clear for kinetic energies between ca. 2 eV and 8 eV, while for lower energies no
asymmetry is observed.
These processes too may be understood on the basis of the potential energy di-
agram Fig. 3.30 for the H+ 2 molecule. We cannot analyze here the quite complex
reaction processes which lead to dissociation and generate the rather substantial rel-
ative kinetic energy of the two nuclei. But obviously, the laser pulse produces a
range of wave-packets which describe the two dissociating atomic nuclei with dif-
ferent initial kinetic energies. Low kinetic energies may be attributed to processes
where only the repulsive 1σu potential of the D+ 2 is populated. One may describe this
fully within the B ORN -O PPENHEIMER approximation and speaks about an “adia-
batic” process: the two atomic components of the molecule just separate on the 1σu
potential. Since in this orbital (3.111) the electron has equal probability to be local-
ized at atom A or B, no asymmetry is observed. This is different for higher kinetic
190 3 Diatomic Molecules
energies. Under the influence of the laser field the 1σu and 1σg are strongly coupled
and the negative charge oscillates between the two states.
Very similar to the charge exchange collisions discussed above, this implies os-
cillation of the electron density between the two atoms A and B, indicated in the
(very schematic) sketch Fig. 3.33(e) as a function of time for the separating protons.
The result of corresponding model calculations is shown in Fig. 3.33(d) and verifies
these considerations quantitatively. In the example shown (φc = 0), for large times
the electron charge is found preferentially at atom B (down). The detector then de-
tects preferentially D+ ions ejected up (ion A). This depends strongly on the carrier
envelope phase φc as shown in the 2D plot Fig. 3.33(b). In the model calculation
too, for φc = π the directions A (up) and B (down) a just reversed (not shown here)
and one expects preferentially to observe the D+ ion in direction B. This is verified
in the experiment.
In the following we treat the building-up principles for molecular orbitals for di-
atomic molecules. This concept corresponds to the filling of the n shells in the
atomic case (Sect. 3.1 in Vol. 1), and leads to rules which may be considered a
periodic system for molecules.
(±λ)
√
φel (ρ, z, ϕ) = ∓φγ λ (ρ, z) exp(±iλϕ)/ 2π (3.123)
z φ
with L
(±λ) (±λ)
(ρ, z, ϕ) = ±λφel (ρ, z, ϕ).
el
We use the standard phase convention and assume that φγ λ (ρ, z) is normalized to
unity. The new quantum number introduced here is λ = |m |, since the sign of m
has no influence on the energy (just as in the case of the S TARK effect). The angular
part of the electronic wave functions may thus be constructed as introduced by (D.6),
Vol. 1 for the real representations of the spherical harmonics.
The quantum number λ is used to characterize one electron wave functions
(MOs) for diatomic molecules. It thus replaces, so to say, the quantum number
for atoms. Following the atomic designation s, p, d etc. for orbitals the following
notation is used:
λ 0 1 2 3 ... (3.124)
σ π δ φ
Correspondingly, to characterize the total molecular state with several electrons one
uses capital Greek letters. The quantum number for the projection of the total angu-
lar momentum on the molecular axis is called Λ (Lambda), its values are designated
as follows:
Λ 0 1 2 3 ... (3.125)
Σ Π Φ
(
Since angular momenta( are added vectorially, Λ ≤ λi holds, just corresponding
to the atomic case L ≤ i .
Another allowed symmetry operation with homonuclear molecules is (as we have
already discussed above) inversion at the centre of mass, r → −r. In cylinder co-
ordinates this corresponds to z → −z and ϕ → ϕ + π . The states are thus distin-
guished according to their parity “g” (even) and “u” (odd):
Even though LCAO-MOs are not the very best approximation for estimating R0 and
De , they provide a good 1st order guide for the construction of molecular orbitals.
Important is in this context that the atomic orbitals involved in constructing a par-
ticular MO must all have the same symmetry in respect of the molecular axis. Only
in this case the overlap integral does not disappear Sik = 0.
For the construction of MOs from atomic s and p electrons the following possi-
bilities exist (for compact writing we use somewhat loosely |s, |px , |py , |pz for
AOs and |σg,u and |πg,u etc. for MOs; antibonding orbitals are designated with an
192 3 Diatomic Molecules
y
(a) σg (ns) y x (b) x
+ + πu (np)
+ py ± p y + +
+ z – –
+ z + z
s±s +
– z y
– σ*u (ns) – x
+ +
π*g (np)
– –
–
z –
+ z
y x
(c) σg (np)
pz ± pz – y
– x
– + – σ*u (np)
+ z –
– + –
+ +
+ z
Fig. 3.34 Construction of molecular orbitals from |s and |p atomic orbitals: (a) |s ± |s,
(b) |py ± |py (y ⊥ plane), (c) |pz ± |pz . Red shaded are the positive, dark-grey the negative
areas of the wave function. In each case the upper MOs are bonding, the lower ones antibonding
(denoted by *)
asterisk *):
|s + |s → |σg and |s − |s → σu∗
|pz + |pz → σu∗ and |pz − |pz → |σg
|py + |py → |πu and |py − |py → πg∗ .
The πu and πg∗ orbitals are each twofold degenerate, as they may also be obtained
from |px + |px and |px − |px , respectively. Figure 3.34 illustrates this orbital
construction scheme. Antibonding orbitals always have a nodal plane perpendicular
to the molecular axis (indicated in Fig. 3.34 by dashed lines). It is also important to
note, that the bonding σ orbitals have even symmetry, while the bonding π orbitals
have odd symmetry.
It should be noted at this point that such schemes of molecular or atomic orbitals
as sketched in Fig. 3.34 may be somewhat misleading at times. They typically rep-
resent the magnitude of the wave function (or its absolute square) at given angles θ
and ϕ in space seen from the origin, for a fixed value of the radial distance. There-
fore, they do not necessarily reflect the spatial extension of the orbitals which would
require at least a contour plot such as shown in Fig. 3.29.
Correlation Diagrams
To obtain an overview of the relative position of the energy levels one may consider
the two limits: “united atom” on the one hand, where the two nuclei are arbitrarily
close to each other with a united charge and “separated atoms” on the other hand,
with infinite distance of the two constituents and no interaction. In between, the
3.5 Molecular Orbitals 193
s 2s σ ↔ σg 2p p-p
s 1s σ ↔ σg 1s s+s
molecular orbitals are formed as just discussed. We summarize the two perspectives
in Fig. 3.35.
The energetic association of the MOs to united and separated atoms leads to so
called correlation diagrams, which allow a semi-quantitative discussion of the ener-
getics for the different MOs. Figure 3.36 shows schematically the potential energy
194 3 Diatomic Molecules
as a function of the nuclear distance for the general homonuclear, diatomic case. On
the left and on the right one marks the energies of the corresponding atomic states
(ignoring the C OULOMB repulsion of the two nuclei). The connections between the
respective energies for united atom and separated atoms leads to a prediction about
the trend for potential energies of the molecular orbitals. A few rules have to be
observed:
The non-crossing rule 3 has already been derived in Sect. 8.1.6, Vol. 1.
From the correlation diagram Fig. 3.36 one may read the energies for different
MOs as a function of internuclear distance. All MOs correlated to the 1s, 2s and
2p AOs in the limit of separated atoms are shown for the general case. For infinite
distance they correspond to the three respective atomic energies, in the case of the
united atom to eight. Also indicated in Fig. 3.36, by vertical dashed lines, are the
orbital energies for some important diatomic molecules from the first row of the
periodic system of elements.
If one is interested in the details for a specific molecule one has, of course, to
draw such a correlation diagram starting with the correct numerical data for this
particular system. One finds e.g., that for light atoms up to N2 the 1πu orbital is
energetically lower than the 3σg orbital, while for O2 and larger molecules this
reverses as summarized in Fig. 3.37(a) and (b), respectively.
molecules emerges. Slightly different notations are used in the literature: e.g. the
σu∗ 2px MO is also referred to as 2px σu∗ or simply just as 3σu∗ . The highest occupied
molecular orbital in any given molecule is called HOMO, the lowest unoccupied
molecular orbital is called LUMO.
Figure 3.38 illustrates schematically how the lowest orbitals are filled for the
molecules H+ 2 up to C2 . As in the periodic system of elements, there are some
irregularities, here e.g. for B2 .
The electron configuration of some further molecules is given in Table 3.7 along
with a summary of bonding properties. The 1σg orbital is bonding and the 1σu∗
orbital is antibonding. Thus, H+ +
2 , H2 and He2 are stable molecules, since they own
more bonding than antibonding electrons in their ground state. Among these, H2
has most bonding electrons, and consequently the smallest equilibrium distance R0
and the highest bond energy De .
More general, the strength of a bond may be estimated from the so called bond
order, which is defined as the difference between the number of bonding and anti-
bonding electrons divided by 2, i.e. (n − n∗ )/2. H+ +
2 and He2 have bond order 1/2,
for H2 it is 1, and He2 with the bond order 0 should not form a stable molecule at
all – according to this simple rule. As we have already discussed in Sect. 3.2.6 one
observes indeed only a very weakly bonded VAN DER WAALS molecule.
The same scheme can be used to derive the periodic system of molecules for
other homonuclear diatomic molecules. As mentioned above, the energetic ordering
of the MOs changes between N2 and O2 (see correlation diagram Fig. 3.36 and term
scheme Fig. 3.37). As documented in Table 3.7 the bond lengths beyond Li2 are all
larger than for H+ +
2 , H2 and He2 due to the more extended valence orbitals (n = 2).
One particularly interesting point is the fact that O2 has two unpaired electrons,
a consequence of H UND’s rules which we have introduced already for atoms (see in
particular Chap. 7 in Vol. 1). It says, that among otherwise identical electron config-
urations, states with the highest total spin have the lowest energies: electrons are first
filled into all energetically degenerate orbitals before one of the orbitals is doubly
occupied. The electrons of the singly occupied orbitals have the same orientation. In
the case of oxygen they form a triplet (S = 1). Thus, O2 is paramagnetic, in contrast
e.g. to N2 .
For an approximative treatment of the potentials for multi-electron molecules one
usually may confine the efforts to a few outer electrons, the valence electrons, i.e.
to the highest occupied orbitals. The other electrons are typically strongly localized
196 3 Diatomic Molecules
Table 3.7 Filling the MO shells in the most simple homonuclear, diatomic molecules (electron
configuration in the ground state). The bond order is (n − n∗ )/2, R0 is the bond distance. Disso-
ciation energies De and term energies Te refer to the potential minimum; usually Te = 0 for the
neutral ground state, see (3.55)
n−n∗
Molecule Electron config. 2 State De / eV R0 / nm Te / cm−1
H+
2 (σg 1s)1 1/2 2Σ +
g 2.65 0.1052 125 443
H2 (σg 1s)2 1 1Σ + 4.48 0.074 0
g
He+
2 (σg 1s)2 (σu∗ 1s)1 1/2 2Σ +
u 2.47 1.08 178 400
He2 (σg 1s)2 (σu∗ 1s)2 0 1Σ +
g 0.00095a 0.297a 0
Li2 He2 (σg 2s)2 1 1Σ + 1.07 0.267 0
g
Be2 He2 (σg 2s)2 (σu∗ 2s)2 0 1Σ +
g not observed
B2 He2 (σg 2s)2 (σu∗ 2s) . . . (πu 2px )2 (σg 2pz ) 2 3Σ −
g 3.0 0.159 0
C2 Be2 (πu 2px )2 (πu 2py )2 2 1Σ + 6.32 0.1243 0
g
N+
2 Be2 (πu 2px )2 (πu 2py )2 . . . (σg 2pz ) 2 1/2 2Σ +
g 8.85 0.112 125 744
N2 Be2 (πu 2px )2 (πu 2py )2 . . . (σg 2pz )2 3 1Σ + 9.90 0.1098 0
g
O2 N2 (πg∗ 2px )(πg∗ 2py ) 2 3Σ −
g 5.21 0.121 0
F2 N2 (πg∗ 2px )2 (πg∗ 2py )2 1 1Σ +
g 1.66 0.141 0
Ne2 N2 (πg∗ 2px )2 (πg∗ 2py )2 . . . (σu∗ 2pz )2 0 not observed
a See Sect. 3.2.6
to the atomic nuclei and contribute only little to the molecular bonding. Of course,
here as with atomic structure calculations the general rule holds, that in a rigorous
computation the results become the better the more MOs (which are or possibly
might be occupied) are accounted for.
Section summary
• We have learned how molecular orbitals can be constructed as linear super-
positions of atomic orbitals (MO from LCAO). Using the variational method
to minimize the energy, one has to determine the Hamiltonian matrix (3.99)
and the overlap integrals (3.100) in the atomic basis. This leads to the char-
acteristic equation (3.102) from which the eigenenergies of the orbitals are
derived.
• We have detailed this for the simplest molecule, H+ 2 . By including only the
two 1s AOs localized on either of the protons, the two energetically low-
est MOs are found; they have even (1sσg ) and odd (1sσu∗ ) inversion symme-
try (r → −r). The 1sσg is bonding, the 1sσu∗ antibonding (repulsive poten-
tial). A surprisingly good first guess for the molecular potential is obtained as
shown in Fig. 3.30.
• These two molecular potentials give rise to interesting interference phenom-
ena in charge exchange collisions and in ultrafast photo-dissociation.
• The systematic construction of MOs for a range of homonuclear, diatomic
molecules starts by writing the wave function (3.123) in cylinder coordinates,
the key quantum number λ being the projection of the angular momentum
3.6 Construction of Total Angular Momentum States 197
onto the internuclear axis. We have constructed a few characteristic MOs from
s and p atomic orbitals (Fig. 3.34).
• MOs must have a finite electron density in between the atoms to be bonding.
• Correlation diagrams for MOs give an overview of the energetics; they con-
nect the energies of the united atom with those for separated atoms (Fig. 3.36).
Based on this one may derive some kind of periodic system for homonu-
clear, diatomic molecules (Table 3.7), which allows to make predictions about
the strength, symmetries and spin properties of the molecular bonding. As a
prominent example, the electronic ground state of molecular oxygen is found
to be a triplet, hence O2 is paramagnetic.
For the total orbital angular momentum too, only the z-component L z is a conserved
quantity. It is obtained as the sum of the z-components of the individual MOs:
z =
L zi with eigenvalues ML =
L mi . (3.126)
i i
The designation Λ = |ML | with the term notation Σ, Π , etc. for Λ = 0, 1, 2 has
already been introduced above. We note here in addition, that molecular orbitals
have a well defined reflection symmetry in respect of planes through the molecular
axis. This reflection symmetry must remain conserved when composing the total
wave function. When combining several MOs to an overall molecular state positive
and negative mi may contribute in different combinations, so that the states are
somewhat more complex than suggested by (3.126).
3.6.2 Spin
As in the atomic case the total spin is the sum of the individual spins of the electrons
S=
S i with | S| = S(S + 1)
i
Σ = S, S − 1, S − 2, . . . , −S.
Caution: do not mistake this spin quantum number Σ for a Σ state (total angular
moment projection Λ = 0)!
198 3 Diatomic Molecules
In analogy to the term designation for atoms, molecular states are denoted
2S+1
Λg,u . (3.127)
We recall briefly and symbolically the construction of spin states which has been
discussed extensively in Vol. 1 (here for one and two electron systems):
Of course, as in the atomic case, the PAULI principle has to be observed, i.e. the
total wave function must be antisymmetric. For a system with two active electrons –
e.g. for the H2 molecule – this implies – as for the He atom: for singlet states (two
antiparallel spins with one antisymmetric spin function, 2nd column in Eq. (3.128))
the spatial electron wave function must be symmetric; for triplet states (two parallel
spins, three symmetric spin functions, 3rd column in Eq. (3.128)) the spatial wave
function must be antisymmetric.
3∆ Ω
3 Λ Σ =1
3∆ Ω
2 Λ Σ=0
Ω Σ = −1
3∆
1 Λ
Fig. 3.39 Electronic angular momentum coupling in diatomic molecules for the example of Λ = 2
( state) with S = 1 (triplet): only the components of the angular momenta in the direction of the
molecular axis are good quantum numbers
H UND’s Case (a) As illustrated in Fig. 3.40(a) one assumes the coupling between
molecular rotation and electronic orbital angular momentum L or spin
S to be very
weak. In contrast, the coupling of the orbital angular momentum to the internuclear
axis is strong and the spin couples to the thus generated internal magnetic field
parallel to the molecular axis. The situation corresponds to the electric analogue to
the PASCHEN -BACK effect. As for the not rotating molecule we have Ω = |Λ + Σ|,
and Ω together with the nuclear rotation N forms the total angular momentum J.
Values of J < Ω are not possible and we have:
J = Ω, Ω + 1, Ω + 2, . . . . (3.130)
10 The finite number of letters in the alphabet limits the choices, unfortunately. H OUGEN (2001)
e.g. use the letters R instead of N , and N instead of K.
200 3 Diatomic Molecules
K N N
Ω Ω
Λ
Λ Σ Je
L S L L S
Fig. 3.40 H UND’s cases (a), (b), (c) and (d) for coupling angular momenta in diatomic molecules
Because of (3.129) in principle J may also be half integer (if the spin is half integer).
In this coupling case the total energy (3.49) for a 2S+1 ΛΩ state is found to be (here
without proof):
Wγ vN / hc = Te + G(v) + F (J ) with F (J ) = Bv J (J + 1) − Ω 2 . (3.131)
The electronic term energy Te may in addition be split by fine structure (cf.
Eq. (6.36), Vol. 1):
Te = T0 + aΛΣ. (3.132)
Apart from the lowest state and the missing values J < Ω there is no difference
between (3.131) and the rigid rotor (3.37), if one replaces N by J . For each fine
structure level of an electronic state a specific ladder of rotational states exists.
H UND’s Case (b) If the orbital angular momentum is zero, or if the electron spin
is coupled only very weakly to the orbit, there is no coupling of the spin to the
nuclear axis. As illustrated in Fig. 3.40(b) we have in that case practically just a
rigid rotor, whose angular momentum N couples with the projection of the orbital
angular momentum The quantum number K replaces now the rotational
Lz to K.
quantum number. For Σ states it is even identical with N . Only at the very end
one has to account for fine structure coupling. With the usual rules the total angular
momentum becomes:
J = K + S, K + S − 1, . . . , |K − S|. (3.133)
The energy of the rotational states is determined essentially by the nuclear rotation,
possibly showing a small (2S + 1) fold FS splitting. For the particularly simple case
of a 2 Σ state the rotational energy becomes
γ 1
F (K) = B(v)K(K + 1) + K for J = K +
2 2
γ 1
= B(v)K(K + 1) − (K + 1) for J = K − , (3.134)
2 2
where the small constant γ is due to FS interaction.
3.6 Construction of Total Angular Momentum States 201
H UND’s Case (c) If the coupling of the orbital angular momentum to the internu-
clear axis is weak, the spin-orbit coupling of the electron cloud is not broken. As
shown in Fig. 3.40(c) an electronic total angular momentum J e is formed, which as
a whole couples to the internuclear axis. The projection onto z is again called Ω.
This component of the electronic angular momentum finally couples again with the
to the overall total angular momentum
rotation N J . In its final result this case
is hardly different from H UND’s case (a) and the energy terms and possible angu-
lar momentum quantum numbers J are indeed described by (3.131) and (3.130),
respectively.
H UND’s Case (d) Figure 3.40(d) shows the rather simple vector diagram for the
case that the coupling of the orbital angular momentum L with the internuclear axis
is very weak (e.g. for highly excited electrons). On the other hand one assumes that
the coupling of the molecular rotation N to
L is strong. One then observes primarily
the energy spectrum of the rigid rotor,
with
where, however, the coupling of N L leads to a (2L + 1) fold splitting of each
rotational level.
H UND’s Case (e) Finally, it is also possible that L and S couple strongly with each
other and form an electronic total angular momentum J e . However, the interaction
with the internuclear axis is weak. This leads to a situation very similar to case (d).
Except that the spitting of the rotational levels is now (2Je + 1) fold. In H ERZBERG
(1989) one still reads about this case that – albeit thinkable – it has never been
observed. This has changed by now, e.g. it has clearly been detected in the case of
moderately high lying RYDBERG states of diatomic molecules such as O2 . Also,
when discussing low energy collision processes between atoms one has to account
for such a possibility, e.g. if collision induced fine structure transitions are observed.
In the spectroscopic practice there are, of course, all kinds of transitions between
the five pure cases introduced here. Higher precision requires, finally, also the in-
clusion of hyperfine interaction – if the atomic constituents have a nuclear spin. In
Fig. 3.21 we have already seen an example of that. The analysis of such interactions
follows essentially that for atoms (see Chap. 9 in Vol. 1). H UND’s cases have to be
extended correspondingly.
For a range of questions it is important to know the reflection symmetry of the elec-
tronic states with respect of a plane through the molecular axis. For single orbitals
this is rather clear and one can extend the considerations presented in Appendix D,
Vol. 1 – replacing the quantum numbers q used in Appendix D.4, Vol. 1, or M
202 3 Diatomic Molecules
ı̂ : z → −z and ϕ → ϕ + π. (3.136)
For the simplest MOs these symmetries can be gleaned from Fig. 3.34.
It is instructive, however, to elucidate this on the basis of wave functions for the
orbitals. Typically, one uses real combinations of (3.123) as introduced with (D.6)
in Vol. 1. For σ orbitals (λ = 0) we write (somewhat laxly in kets)
φσ (ρ, z)
|σ = φel (ρ, z, ϕ) = √ , (3.137)
2π
φπ (ρ, z) cos ϕ
|πx = √ exp(+iϕ) + exp(−iϕ) = φπ (ρ, z) √ (3.138a)
2 π π
φπ (ρ, z) sin ϕ
|πy = √ exp(+iϕ) − exp(−iϕ) = φπ (ρ, z) √ . (3.138b)
2 πi π
The angular part is normalized to unity in all cases. By applying (3.135a) and
(3.135b), we see that σ orbitals have always positive reflection symmetry (no de-
pendence on ϕ):
Example H2 Molecule
In the most simple case we only have to fill one orbital with two electrons, as e.g.
the 1sσ orbital for the ground state of the H2 molecule. With the notations (3.127)
and those given in Fig. 3.35 (close to the united atom), the total electronic wave
function is written:
1 + 0
Σ = χ φ1sσ (ρ1 , z1 ) · φ1sσ (ρ2 , z2 ) . (3.139)
g 0
Since the spatial part is obviously symmetric in respect of exchange of the electrons,
the spin function |χSMS must be antisymmetric, i.e. be a singlet. As for both elec-
trons λ = 0 holds, also the total orbital angular momentum remains Λ = 0, i.e. we
describe a Σ state. Obviously this function is also symmetric in respect of inver-
sion, and finally, positive reflection symmetry for both 1sσ orbitals leads to overall
positive reflection symmetry. This is indicated by the superscript + sign. The bound
ground state of H2 is a singlet X 1 Σg+ state (the letter X is just used for numbering
the states, and one start with X for the ground state, the letters A, B, . . . follow for
excited states).
Next we keep one electron in the 1sσ orbital and bring the other one in the next
higher 2pσ ∗ orbital (σu∗ 1s in MO notation). The configuration |1sσ 2pσ ∗ may be
combined to two different spatial functions:
√
φ1sσ (ρ1 , z1 )φ2pσ (ρ2 , z2 ) ∓ φ1sσ (ρ2 , z2 )φ2pσ (ρ1 , z1 ) / 2. (3.140)
Both wave functions have again Σ + reflection symmetry (they inherit this property
from the constituents), but both change their sign upon inversion (due to the involve-
ment of a 2pσ ∗ orbital). We also see that one of the functions is antisymmetric, the
other symmetric in respect of exchange of the two electrons. The spin function thus
has to be triplet or singlet, respectively. Closer inspection shows that the resulting
|3 Σu+ state is repulsive (one bonding and one antibonding orbital). The other state
does not lead at all to a low lying molecular state.
Next we have a look at the configuration |1sσ 2pπ (two bonding orbitals with
λ1 = 0 and λ2 = 1). We may construct from this e.g. an odd (u), bonding state,
symmetric in respect of electron exchange. With Λ = λ1 + λ2 = 1 we obtain from
(3.138a):
1 0 √
Πu = χ |1sσ 2pπx + |2pπx 1sσ / 2 (3.141)
0
0 iϕ
= χ0 φ1σg (z1 ρ1 )φ1πu (z2 ρ2 ) e 2 + e−iϕ2
√
+ φ1σg (z2 ρ2 )φ1πu (z1 ρ1 ) eiϕ1 + e−iϕ1 /(2 π ).
With (3.136) we verify that we have indeed constructed a (u) state: for both electrons
φ(zρ) = φ(−zρ) and replacing both azimuthal angles ϕ → ϕ + π brings an overall
minus sign.
Reflection symmetry in this case is positive in respect of the xz plane (πx like –
just replace both ϕ → −ϕ). In contrast, rotation of the molecule around the z-axis
by π/2 generates a state whose reflection symmetry is πy like, i.e. negative. The
electronic problem is, however, completely symmetric in respect of the z-axis: both
204 3 Diatomic Molecules
states |1 Πu+ and |1 Πu− are degenerate. Thus, one usually omits the symmetry des-
ignation ± for molecules with Λ = 0. We shall, however, discuss the limitations of
this view in Sect. 3.6.6.
Example O2 Molecule
What is, however, the situation for Σ states with Λ = 0? They may e.g. be con-
structed from two π orbitals with opposite orientation. According to Table 3.7 the
ground state of O2 is an example. The system is well bonded by the electrons in the
completely filled N2 of the inner electrons. In addition, two πg∗ orbitals have to be
filled (configuration (πg∗ 2p)2 with λ1 = λ2 = 1). We first note that the overall spa-
tial wave function is in any case even (g) in respect of inversion. We start again with
the complex orbitals (3.123) and use now the MO notation |φel = |πg∗ 2p±1 .
(±λ)
The terms e±i(ϕ2 −ϕ1 ) make sure that the angular momenta compensate each other,
so that
z1 + L
(L z2 )3 Σg− ≡ 0
and Λ = λ1 − λ2 = 0. For inversion symmetry we note that again for both electrons
φ(zρ) = φ(−zρ), but when replacing both azimuthal angles ϕ → ϕ + π the π is
compensated and the sign does not change. We have indeed constructed a Σg state.
Reflection in respect of the xz plane (i.e. ϕ1 → −ϕ1 and ϕ2 → −ϕ2 ) changes the
sign, hence, the overall reflections symmetry of the state is negative. It is important
to note that in this case rotation around the z-axis by an arbitrary angle δ (i.e. ϕ1 →
ϕ1 − δ and ϕ2 → ϕ2 − δ) does not change anything about the eigenfunction! This
is obviously due to the fact that the two orbital angular momenta just compensate
each other. Thus, for Σ states reflection symmetry is a good quantum number which
characterizes the state. This statement also holds for the singlet state complimentary
to (3.142):
1 + 0 √
Σ = χ φ1π ∗ (z1 ρ1 )φ1π ∗ (z2 ρ2 ) cos(ϕ2 − ϕ1 ) /(2 π). (3.143)
g 0 g g
It has positive (g) inversion symmetry and positive reflection symmetry, again in-
dependent of the alignment of the reflection plane through the molecular axis. The
exchange symmetry for the two position coordinates (1 and 2) is now positive (cos
function). Hence, the spin function must be antisymmetric and we have a singlet
3.6 Construction of Total Angular Momentum States 205
state – as it turns out its electronic energy is significantly higher than that of the
|3 Σg− state.
In contrast, the Π , . . . etc. states are degenerate in respect of the ± symmetry.
Still with the same electron configuration (πg∗ 2p)2 we construct the remaining two
singlet states, with spatial wave functions which are again symmetric in respect of
electron exchange:
1 0 ∗ √
g = χ π 2p−1 π ∗ 2p1 + π ∗ 2p1 π ∗ 2p−1 / 2
0 g g g g
#
0 1 i(ϕ2 +ϕ1 )
= χ0 φ1πg (z1 ρ1 )φ1πg (z2 ρ2 ) √
∗ ∗ e +e −i(ϕ2 +ϕ1 )
2 πi
√
= χ00 φ1πg∗ (z1 ρ1 )φ1πg∗ (z2 ρ2 ) sin(ϕ2 + ϕ1 ) /(2 π). (3.144)
that the py orbital (dumbbell, rotationally symmetric around the y-axis) is hardly
influenced by this rotation. In contrast, the electron in a px orbital (dumbbell, per-
pendicular to the y- and z-axis) is exposed quite evidently to a centrifugal force.
This leads to a splitting which is called Λ-type doubling. Even if this is an ener-
getically very small contribution, it is well observable with modern spectroscopic
precision. The effect may become quite significant in dissociation processes where
e.g. diatomic molecules are ejected from a larger complex. The rotational distribu-
tion of these fragments may then have a pronounced asymmetry in respect of the
two molecular rotation axes.
where both electrons are in antibonding orbitals. In the separated atom limit, both
combinations do not correlate with the atoms in their 1s ground state and only con-
tribute to higher lying bound and unbound states.
In a similar manner the higher lying molecular states are constructed from more
complex MOs. The principles are those treated in Sect. 3.6.5. An overview of the
H2 potentials computed with quantum chemical methods gives Fig. 3.41. It also in-
cludes its ions H− +
2 and H2 . As already mentioned, the states are usually ‘labelled’
with capital letters, the ground state with X, the next higher with A, B, C etc. In
a number of cases, due to the historical development, this is not followed conse-
quently. Often one uses the lower case alphabet for states which have been found
later. As H2 is of particular importance (and its potentials form a rather complex
manifold) a section of the energy diagram is shown once again in Fig. 3.42 on an
enlarged scale.
Note that no stable negative ion (anion) H− −
2 exists. The lowest energy of H2 lies
−
above the bound, neutral ground state of H2 . Thus, H2 can decay spontaneously into
H2 + e− – if it is at all ever formed in the first place. One observes such states as
resonances in the scattering of low energetic electrons by H2 . The observed feature
looks very similar to autoionization which we have treated in Chap. 7, Vol. 1. We
shall come back to this in Chap. 8.
One more, closely related, process should be mentioned at this point, the so called
predissociation. We note the quite remarkable potential maxima for a number of H2
states, as seen in Fig. 3.42 e.g. for the states designated with I , i and h. They are
a consequence of avoided crossings – which generate local maxima in the poten-
tials. These maxima make it possible to form vibrational states which lie above the
dissociation limit. Although these states are strictly speaking not stable states, they
still live for some time since the potential barrier prevents immediate decay. The
probability for dissociation of such molecular states by “tunnelling” depends on the
height and width of the barrier.
A few words on how to calculate such potentials are in order at this point. For the
H2 molecule one may in principle follow the same procedure as in the H+ 2 case, and
use suitably chosen trial functions as approximate solutions of the S CHRÖDINGER
equation. The electronic Hamiltonian is now (in a.u.)
1 1
H = H1 + H2 + + (3.149)
r12 R
i = − 1 ∇ 2i − 1 − 1 .
with H
2 rAi rBi
‘New’ as compared to H+ 2 is in particular the repulsive term with r12 , the dis-
tance of the two electrons, while rAi and rBi are the distances of electron i from
atomic nucleus A and B, respectively, and R is again the distance between the nu-
208 3 Diatomic Molecules
24 24
+ D' 1 Π u 4p 22
22
H2 V Π u 4f
1
2Σ+ +
20 D'' 1 Π u 5p u 2pσu B'' 1Σu 4pσ 20
+
+ g 3Σg 3dσ
B' 1Σu3pσ
+ H++H(1s)
18 X 2Σg 1sσg
+ 5ℓ 4ℓ
e 3Σu2pσ
d 3Π u 3p H(1s)+H(3ℓ)
+
16 m 3Σu 4fσ 3 16
3Σ+
i Π g 3d
h g 3sσ H(1s)+H(2ℓ)
14 14
-
H(2s)+H (1s 2)
D 1Π u 3p and J,j 1,3∆ g 3d δ
potential energy V(R ) / eV
12 I Π g 3d 12
1
B Σu 2pσ
+
C 1 Π g 2p m 3Σu 4fσ
+ +
C +e- 2Σg a 3Σg 2sσ
10 10
+
H 1Σg 3sσ
+
c 3Π u 2p
+ +
c +e- 2Σg E,F 1Σg 2sσ + 2pσ2
8 H2 8
+
6 b 3Σu 2pσ 6
+ H(1s)+H(1s)
b +e- 2Σg
4
4
- 10
H2 -
X + e - 2Σ g
+ H(1s)+H
-
(1s2)
2 5 2
H2
+
0
X 1Σg 1sσ 0
Fig. 3.41 Potentials for the most important states of the H2 molecule (for comparison also for the
anion H− +
2 and the cation H2 ) after S HARP (1971). The equilibrium distance of H2 in its X Σg
1 +
electronic ground state is R0 = 0.07416 nm, the dissociation energy D0 = 4.476 eV. The b 3 Σu+
state is repulsive. For brevity we omit the description 1sσ . Note that the ion pair p + H− (1s 2 ) is
formed at 17.5 eV
clei. The molecular potential for a state φγ is found by minimizing (3.95). In the
R ITZ variational method one may construct φγ as linear combination of MOs φg
and φu . The MOs needed for the ansatz (3.147) may e.g. be found by first solving
3.6 Construction of Total Angular Momentum States 209
17 17
V 1 Π u 4f
H(1s)
+H(3ℓ)
H2+ +
m 3Σu 4fσ
+
16 g 3Σg 3dσ 16
D1 Π u 3p
X 2Σ+ d 3Π u 3p
g
i 3Π g 3d
g 3d δ
J,j 1,3∆
I Π g 3d
D'' 1 Π u 5p D' 1 Π u 4p
+
15 h 3Σg 3sσ 15
B'' 1Σ + 4pσ
u + H(1s)
potential energy V(R ) / eV
B' 1Σu3pσ
+H(2ℓ)
14 +
+ 14
H 1Σg 3sσ
C 1 Π g 2p
3
a Σg 2sσ
13 e 3Σ+ 2pσ 13
u
+
E,F 1Σg 2sσ + 2pσ2
12
c 3Π u 2p
H2 12
+
B 1Σu 2pσ
+
b 3Σu 2pσ
11 11
0.05 0.1 0.15 0.2 0.25 0.3 0.35
nuclear distance R / nm
Fig. 3.42 Enlarged part of the potential energy diagram for H2 Fig. 3.41 after S HARP (1971).
Note the interesting states designated E, F and H with double minima, and the “predissociating”
states with double minimum potential marked I , i and h
210 3 Diatomic Molecules
i φ(r i ) = W φ(r i ).
H
However, it is also possible and quite instructive to start with LCAO-MOs as a first
approximation, in the most simple case with (3.110). Explicitly for the ground state
(without normalization) this leads to:
φX(S) (1, 2) ∝ Φ1s (rA1 )Φ1s (rB2 ) + Φ1s (rA2 )Φ1s (rB1 ) (3.150)
+ Φ1s (rA1 )Φ1s (rA2 ) + Φ1s (rB1 )Φ1s (rB2 ) . (3.151)
The two parts (3.150) and (3.151) of this wave function may be identified with a
covalent and an ionic contribution, respectively. The covalent part represents a rel-
atively uniform distribution of the two electrons at both atoms. For large distances
between the atoms, the covalent wave function eventually merges with the descrip-
tion of two separated atoms H + H. In contrast, the ionic part localizes both electron
either at atom A, i.e. Φ1s (rA1 )Φ1s (rA2 ), or at atom B, i.e. Φ1s (rB1 )Φ1s (rB2 ). This
corresponds thus to the situation H− + H+ and H+ + H− , respectively. Asymptot-
ically the thus described state is found for H2 (in its vibrational ground state) at an
energy WI + D00 − WEA 17.32 eV, where WI is the ionization potential of the H
atom, with D00 being the dissociation energy of H2 in its vibrational ground state,
and WEA stands for electron affinity of the H atom. Looking at Fig. 3.41 tells us that
this is far above the H + H asymptote. Interestingly, the ionic state contributes to
the excited B 1 Σg+ state which is much weaker bound than the X 1 Σg+ ground state
of H2 .
To finally evaluate the energy according to (3.95) with the simple trial function
(3.150)–(3.151) one again introduces the coordinate transformation (3.97) in order
to compute the integrals which evolve. We refrain from presenting here the details
of the calculation – in particular because the results of this procedure is not really
convincing: on obtains with this ansatz R0 = 0.08 nm and De = 2.68 eV – which has
to be compared with the precisely calculated and experimentally determined values
R0 = 0.07414 nm and D00 + ω0 /2 = 4.747 eV (see Table 3.2).
Instead of starting from MOs, one may simply use a covalent charge distribution as
trial function. One chooses, e.g. for the H2 ground state (not normalized)
φX(S) (1, 2) = Φ1s (rA1 )Φ1s (rB2 ) + Φ1s (rA2 )Φ1s (rB1 ) × χ00 (1, 2) (3.152)
|φX(S)
φX(S) |H
WX(S) = .
φX(S) |φX(S)
3.6 Construction of Total Angular Momentum States 211
11 Whichisn’t really surprising here, as we have seen that the ionic component of the MOs leads to
completely wrong asymptotic states.
212 3 Diatomic Molecules
28
N(2Do) + N+(3P)
2
u
26
N(4So) + N+(1D)
2Σ –
C 2Σ + u
u
2∆
24 u 4
N(4So) + N+(3P)
4Σ –
u
u
22 B 2Σ u+ 4∆
u
4Σ + D2 g
u
20
+
N2 A 2 Πu
–
18 + +N
N
16 X 2 Σ g+
C' 3 N(2Do) + N(2Do)
u
14 +
b' 1Σu N(4So) + N (2P o)
+
potential energy V(R ) / eV
E 3Σ
g N(4So) + N(2Do)
12 3
C u
5 –
u +
7Σ N(4So) + N (3P)
10 a1 g 5Σ
+ u
w 1∆u g
1Σ –
N(4So) + N(4So)
a' u 3∆
– u
8 B' 3Σu
B 3Π g
6 +
A 3Σu
–
4 N2 N2
X2 g
2
X 1Σg+
0
0.04 0.08 0.12 0.16 0.20 0.24 0.28 0.32 0.36
nuclear distance R / nm
Fig. 3.43 Potentials for the nitrogen molecule and its ions according to G ILMORE (1965). The
potential of the unstable N−2 anion has been adjusted to fit the experimental observations from
electron scattering resonances (see Sect. 8.1.2)
3.6 Construction of Total Angular Momentum States 213
24
O(3P) + O+(2Do)
22
2Φ
2Σ
– u O(1D) + O+(4So)
g
2Δ 6Σ +, 6Σ + , 6 6
20 g g u u, g
4Σ
+
4
– g g O(3P) + O+(4So)
2Σ
b 4Σ g g
18 2Σ
+ +
A2 u g and 4Σu
+
16 O2
a4 u
X2 g
potential energy V(R ) / eV
14
1Σ
+
12 u O(1D) + O(1S)
10 O(3P) + O(1S)
1Δ
u
O2 O(1D) + O(1D)
8
O(3P) + O(1D)
3Σ
–
1 B u 5 5Σ + 5Δ 5 +
g u g g and Σg
6 3 3
1
u
– +
5Σ , 3Σ and 1Σ
+
g u u u g
+
A 3Σ u O(3P) + O(3P)
A' 3Δ
4 u -
c 1Σ u 2
– o
u
O(3P) + O (2P )
+
b 1Σ
g –
2
a 1Δg
O2
–
0 X 3Σ g X2 g
Fig. 3.44 Potentials for the oxygen molecule and its ions according to G ILMORE (1965); minima
of the A, A and c states have been modified according to J ENOUVRIER et al. (1999)
214 3 Diatomic Molecules
have already discussed that the ground state X 3 Σg has highest multiplicity accord-
ing to H UND’s rules. The next higher states (built from the same atomic electron
configurations) are the a 1 g and b 1 Σg+ states.
As already mentioned, oxygen gas O2 in its ground state X 3 Σg is paramagnetic,
due to its two electrons with parallel spin, S = 1. We also note that a stable anion
exists, O− − −
2 in the ground state (in contrast to H2 and N2 ). Nevertheless, the anion
−
decays when vibrationally excited: O2 → O2 + e; conversely, it may also be gen-
erated in a vibrationally excited state, again as a resonance in electron scattering
by O2 .
The overviews on the potentials of N2 and O2 presented here (as well as that
on NO in Fig. 3.53) are based on the very systematic work of G ILMORE (1965).
Spectroscopic data have been evaluated there in the spirit of the RKR method (see
Sect. 3.4.6), but also quantum chemical computations have been included. Even
though since then many new measurements and calculations have been published,
we have added only a few modifications to this data compilation, since a comparable
critical overall evaluation of all available data has – up to now – not been reported.
Section summary
• The electronic states of diatomic molecules are characterized by projections
of total angular momenta (sums over all active electrons) onto the internuclear
axis.
• In the simplest cases (e.g. for H UND’s case a) the projection of the total orbital
angular momentum, Λ, and total spin, S, are good quantum numbers. States
with Λ = 0, 1, 2 etc. are designated by Σ, Π, etc., while the spin quantum
numbers S = 0, 1/2, 1 etc. lead to singlets, doublets, triplets etc. In analogy
to atoms, typically the notation is 2S+1 Λg,u is used.
• The five H UND’s cases differ by the strength of the interaction between in-
ternuclear field, orbital, spin and rotational angular momentum. The ensuing
different angular momentum coupling schemes lead to spectroscopically rel-
evant differences in the energy spectra.
• In addition to angular momenta, molecular states are characterized by their
symmetries. For homonuclear, diatomic molecules, the relevant symmetry op-
erations are inversion in respect of the origin, and reflection at planes through
the molecular axis.
• For multi-electron systems (exemplified by H2 and O2 ) positive or negative
reflection symmetry lead to different energies for Σ states – and only for
these. For other values of Λ, states with ± reflection symmetry are degenerate
if the molecule does not rotate. However, high rotational levels show a small
splitting, the so called Λ-type doubling.
• Potential energy diagrams for H2 , O2 , and N2 show many interesting struc-
tures. In all three cases negative ions (anions) may be observed as short-lived
resonances in electron scattering, while only the O− 2 anion has a stable vibra-
tional ground state.
3.7 Heteronuclear Molecules 215
• N2 is the most strongly bound diatomic molecule (De 9.9 eV) and has also
a particularly high ionization potential. H2 and N2 have rather isolated elec-
tronic ground states, while O2 has three lowest states separated by less than
1.7 eV and ordered according to H UND’s rules: the X 3 Σg− ground state (para-
magnetic), the a 1 g and the b 1 Σg+ state.
For the sake of simplicity we have set the overlap integral S = 0 – which will
change the overall trends only when both atoms come very close. For heteronuclear
molecules the secular equation (3.106) thus is
For many heteronuclear molecules |HAA − HBB |
|HAB | holds (in particular for
those with very different nuclear charges). Then the two eigenvalues are approxi-
mately
|HAB |2 |HAB |2
"+ HAA + and "− HBB − . (3.153)
HAA − HBB HAA − HBB
Figure 3.45 illustrates the raising and lowering of the atomic orbital energies. We
have assumed here that HAA > HBB .
Instead of the even and odd functions (3.110) and (3.111), respectively, we obtain
now for the orbitals
(±) (±)
φ± = cA ΦA + cB ΦB
216 3 Diatomic Molecules
4σ *
2s
2s 3σ
2σ *
1s
1s 1σ
A B
(+)
cA HAB HAA − HBB
with =
1 and
(+)
cB "+ − HAA HAB
(−)
cA HAB HAB
− − =
1.
cB(−) "− − HAA HAA − HBB
The molecular orbitals are thus rather similar to the AOs, i.e. φ+ ΦA and φ−
−ΦB . The lower level correlates as expected with ΦB .
3d δ 3π
3d 3d π 7σ* σ*3sA
3dσ* 3sA
6σ* π 2pB
3p 3p π 2π σ2pB 2pB
3pσ 5σ* π 2pA 2pA
3s σ2pA
3sσ* 4σ* σ*2sB 2sB
2p π 1π 2sA
2p 3σ σ2sA
2s 2pσ
2sσ* 2σ* σ*1sB
1sB
1sA
σ1sA
1s 1sσ 1σ
united atom molecule separated atoms ZA > ZB
(R = 0) R (R → ∞)
energies. We have discussed this in the context of the S TARK effect in Sect. 8.2.7,
Vol. 1. For large internuclear distances the energetic ordering is thus σ < π < δ.
However, due to the necessary correlation with the united atom scheme this trend
may be inverted in the MO region. In Fig. 3.47 this is particularly evident for the
1π , 5σ , 2π and 6σ states.
In Table 3.9 the electron configurations, equilibrium distances R0 and the bond
energies De of the ground state for a number of heteronuclear diatomic molecules
are summarized.
218 3 Diatomic Molecules
It should be noted that the scheme communicated here gives only a first overview
of potential MOs. In practice one has to investigate the energetics and each atomic
configuration in detail to find a suitable MO basis set for describing a particular
molecule. As already mentioned for homonuclear molecules, today a number of
sophisticated methods and efficient computer programmes are available for this task.
To get somewhat more specific, we shall in the following discuss a few interesting
examples in more detail.
3.7.3 Lithiumhydrid
LiH is quite a remarkable molecule – albeit not available in a gas bottle. A number
of specialities may be studied with it. In recent years it has even attracted interest
in the context of B OSE -E INSTEIN condensation of ultracold atoms (see e.g. C ÔTE
et al. 2000; J UARROS et al. 2006), not least because of its large dipole moment
which makes one expect specific long range effects. One may e.g. apply two-photon
processes to generate such molecules by photo-association from ultracold atoms.
The constituents of the LiH molecule have the configuration 1s 2 2s (Li) and
1s (H), whereas the closed 1s 2 does not participate in the bonding. These elec-
trons fill the 1σ MO localized at the Li nucleus. A comparison of the first excited
atomic states 1s 2 2p (Li) and 2s, p (H) with 1.848 eV and 10.2 eV excitation energy,
respectively, shows that a correlation diagram would not lead us very far in this case.
According to the above considerations, the active MOs will be determined by the
easy to excite AOs of Li. The ground state is essentially given by the 2s electron
of Li which interacts with the 1s electron of the hydrogen atom: the atomic orbitals
2s (Li) and 1s (H) form a 2σ MO. In Fig. 3.48(a) the potentials for some states of
LiH are collected from several sources.
Most clear is the situation for the triplet states 3 Σ, where the PAULI principle
provokes a spatial repulsion of the σ MOs from n − 1s. A closer inspection shows
(not recognizable in Fig. 3.48) that here too long range dispersion interaction leads
to a weak attraction. For the a 3 Σ + state this is of importance in a distance of 1 −
2 nm and is of significant relevance for the interaction among ultracold atoms.
More complicated are the conditions for the singlet states for which in addition
to the symmetric, neutral AO configuration Li0 H0 also the ion pair Li+ H− must be
accounted for. The hypothetical energy of the ion pair is indicated in Fig. 3.48(a)
by the dashed red line. As seen, the asymptotic energy of this ion pair lies below
the ionization energy of Li (5.39 eV) by just the electron affinity of the H atoms,
WEA = 0.756 eV. A suitable MO which accounts for this, will be characterized by a
significantly asymmetric charge distribution.
In the most simple approach one realizes this by a linear combination of a 2s and
a 2pz AO of lithium (a |σ 2s and a |σ 2p AO, in respect of the internuclear axis z).
This is sketched in Fig. 3.49.12 Such a combination of atomic orbitals with different
12 We remind the reader that such schematic polar plots of orbitals just display the magnitude of
the wave function under a given polar angle θ at fixed radial distance. They do not give a realistic
image of the spatial extension of the orbital.
3.7 Heteronuclear Molecules 219
V R) / eV
V( Li++H
dipole moment / ea0
–
5 (a) Li++ H WEA (b)
4 well depth R 4
ca. 34.5 meV A state
3 Σ+ X state
3 3
2 B 1Π +
2
Li(2p 2P) + H(1s 2S)
A 1Σ +
1 1
Li+H– Li(2s 2S) + H(1s 2S)
0 a3Σ+ 0
B state
−1 -1
X 1Σ +
−2 -2
0 0.2 0.4 0.6 0.8 1.0 0.2 0.4 0.6 0.8 1.0
R / nm
Fig. 3.48 Potentials (a) and dipole moments (b) of the LiH molecule. Ground state and excited
state A 1 Σ + are strongly influenced by the ionic Li+ H− configuration. Triplet states and B state
are essentially covalent (adapted from PARTRIDGE and L ANGHOFF 1981; Y IANNOPOULOU et al.
1999; C ÔTE et al. 2000)
1
_ |σ 2s〉 √3
√
|σ 2p〉
+
2 2
hybrid: = |2spσ 〉
+ z
–
angular momenta is called a hybrid orbital. According to PAULING (1931), who first
introduced this important concept of hybridization, the bond will be strongest for a
wave function which has the largest value in the direction of the bond. In the√present
case one easily verifies that this implies superposition coefficients 1/2 and 3/2 for
the |σ 2s and |σ 2p orbital, respectively, obviously resulting in a quite asymmetric
shape of the orbital. We recall that such type of orbital was already encountered in
Sect. 8.2.8, Vol. 1 in the context of the linear S TARK effect for the 2p, 2s states of
atomic hydrogen. When a molecule is formed, the very strong electric field of the
two atomic nuclei acts correspondingly. The phenomenon always occurs if ns and
np states have similar energies. A more detailed treatment of hybridization will be
given in Sect. 4.4.2.
220 3 Diatomic Molecules
The alkali halides (i.e. LiF, NaCl, NaI, KBr etc.) are prototypical examples for ionic
molecular bonding. As crystal solids some of them are also of considerable practical
importance. Their preparation as free molecules for spectroscopic purposes requires
13 Inprinciple, one may let a Li+ cation collide with an H− anion in a scattering experiment. For
large distances, the interaction potential of such a system is indeed given by the red dashed line
(∝ −1/r) and remains – for sufficiently high kinetic energies – a good first approximation also at
smaller distances. Transitions at the (avoided) crossings may indeed be induced in such collision
processes.
3.7 Heteronuclear Molecules 221
potential energy / eV
for a hypothetical, purely 1
A1Σ+
ionic system Na+ + Cl− 0
which converges for large R
– Na(3s) + Cl(3p)5
towards 1.522 eV. The -1 Na+Cl
ground state is dominated by
-2
the ionic bonding. At about
X1 Σ +
1 nm one sees the classical -3
avoided crossing with the RH
excited A 1 Σ + state which D e=
becomes covalent for small 4.23eV 0.2 0.4 0.6 0.8 1.0 1.2 1.4
internuclear distances R0 = 0.236079 nm R / nm
A + BC → A+ + (BC)− → A+ B− + C, (3.154)
in principle already by a diatomic molecule, as shown by Fig. 3.50 for the example
NaCl.
The configuration of AOs in this system is 3s (Na) and 3s 2 3p 5 (Cl). The 3s
valence electron of the Na atom is only weakly bound and is given off easily to
complete the 3p shell in Cl. The ionization potential of Na is WI = 5.1391 eV, the
electron affinity of Cl is WEA = 3.617 eV. At infinite distance one thus needs an
energy
to form the ion pair Na+ and Cl− . However, at shorter distances energy is gained by
the C OULOMB attraction and the ionic potential is expected to be
e2
Vγ (R) = Vγ (∞) − ,
4πε0 R
14 For small distances and around the potential minimum we have used the potential derived an-
alytically from spectroscopic data reported by R AM et al. (1997). For larger distances the semi-
empirical valence bond calculations of C OOPER et al. (1987) appeared more plausible: experimen-
tal data were available up to v = 8 only (ca. 350 meV). It does not seem reasonable to assume
that the extrapolation of such data to about 4 eV excitation energy gives reliable information in the
vicinity of RH .
3.7 Heteronuclear Molecules 223
σ*
3s σ* antibonding
π π 4 "lone pairs"
5p
σ bonding
σ
Na NaI I
Fig. 3.51 MOs from AOs for the alkali halogenide molecules. Due to their very different orbital
energies the two “lone pairs” of π electrons are localized at the I atom and do not participate in
bonding
orbitals are localized in great distance from each other, at the Na and I atom, respec-
tively. The next states, A1 Σ + and 3 Σ + states, which are formed with the . . . σ π 4 σ ∗
MOs, lie somewhat higher, as the excitation of the σ orbital requires more energy
than exciting a “lone pair” of electrons out of the π orbitals. At this stage of the com-
putation one obtains a picture very similar to that in the case of NaCl in Fig. 3.50:
with pronounced curve crossings as a consequence of the ionic influence.
However, one now has account for spin-orbit interaction and diagonalize the
whole Hamiltonian again with it included. This leads now to a range of Ω ± states
which arise from Λ + Σ as we have discussed formally already in Sect. 3.6.3. Fig-
ure 3.52 shows only three of these states, which can be accessed by optical excitation
from the ground state.
According to A LEKSEYEV et al. (2000), the above mentioned Λ states with
S = 0 and 1 are responsible for the X0+ ground state and five more Ω states (2(I ),
1(I ), 1(II), 0− (I ) and A0+ ) which correlate asymptotically with Na(3 2 S1/2 ) + I
(5 2 P3/2 ). For symmetry reasons, of those only the A0+ state interacts with the
ionic configuration and leads to an avoided crossing. In addition, there are the states
(1(III), 0− (II) and B0+ ) which correlate for R → ∞ with Na(3 2 S1/2 ) + I(5 2 P1/2 ).
Of those, again only the B0+ state interacts with the ionic configuration and avoids
the crossing. Those two avoided crossings give rise to two excited, quasi-bound
states, A0+ and B0+ , as clearly seen in Fig. 3.52. For completeness we also have
224 3 Diatomic Molecules
drawn another state, 0+ (IV) according to C OOPER et al. (1987), which correlates
with Na(3 2 P1/2 ) + I(5 2 P3/2 ) which is, however, not bonding since the ionic poten-
tial does not cross this state. The diagonalization of the spin-orbit interaction hardly
influences the ground state. It is, however, responsible for a lowering or raising
of the asymptotic potentials by −1/3 and +2/3 of the atomic spin-orbit splitting
(0.9426 eV). Thus, the overall bonding energy of the X0+ ground state becomes
De = 3.02631 eV. The bond-length is R0 = 0.2836 nm.
NaI has played a key role in the development of femto-chemistry. With ultrashort
laser pulses one may generate wave-packets on the repulsive part of the excited A0+
state potential (just above the minimum of the X0+ at R0 , see Fig. 3.52). These
wave-packets may oscillate between the left and right border of the A0+ potential,
just as a classical oscillator. However, these vibrational states are not completely
stable. Speaking figuratively: at the avoided crossing the potential well has a leak,
through which during each oscillations a certain fraction of the wave-packet ‘flows
out’. With femtosecond pump-probe techniques one may follow this process. For his
seminal work studying such transition states in chemical reactions with femtosecond
spectroscopy Ahmed Z EWAIL (1999) received the N OBEL prize in chemistry.
1Δ 5 Σ+
20
1Σ - N(4So)+O+(4So)
A1
18 3 Σ-
w 3Δ
16 b3
NO+
14
potential energy V(R ) / eV
a 3Σ+
12
X 1Σ + 4 Σ- o
N(4S )+O(1S)
o
N(2P )+O(3P)
10 S 2Σ+, M 2Σ+, K 2 (from top to bottom)
F 2Δ H' 2 and H 2Σ o
2 Σ+
N(2D )+O(3P)
E 2Σ +
2Φ o
8 i 4 4Δ N(4S )+O(1D)
D 2Σ + G 2Σ
-
i
4Σ+ and 6Σ+
6
C2 B' 2Δ i o
N(4S )+O(3P)
6 -
5 and 5 Σ
A 2Σ + 3
B2 r
- o - o
b 4Σ N(4S )+O (2P )
4 a4 i
NO
2 NO-
-
X 3Σ
0
X2 r (1 )2 (2 )2 (3 )2 (4 )2 (1 )4 (5 )2 (2 )1
-2
0.04 0.08 0.12 0.16 0.20 0.24 0.28 0.32 0.36
nuclear distance R / nm
Fig. 3.53 Potentials for nitrogen monoxide and its ions according to G ILMORE (1965)
226 3 Diatomic Molecules
bonded than the ground state (smaller R0 , more rigid potential). In this manner a
whole series of potentials is generated, A 2 Σ + , C 2 Π , D 2 Σ + , E 2 Σ + , with nearly
identical shape of their potentials. As seen in Fig. 3.53, these states converge to-
wards the ground state X 1 Π of the NO+ cations and reflect already its potential:
the outer electron is far away from the ion core of the molecule, and does not par-
ticipate in the bonding. It behaves like an atomic RYDBERG electron, hence one
speaks about molecular RYDBERG states. In contrast, the states a 4 Π , B 2 Π , b 4 Σ −
show neither a great similarity to the neutral nor to the ionic ground state. They are
significantly weaker bound than either of those: they are generated by exciting elec-
trons from lower lying, bonding MOs with subsequent rearrangement of the whole
electron configuration and a softening of the bonding. One speaks of valence states.
As states of equal symmetry must not cross, a whole series of avoided crossings
arises, both for RYDBERG and valence states. These are clearly seen in Fig. 3.53 at
energies between 7 eV and 9 eV.
As also indicated in Fig. 3.53, here too an anion exists. NO− is, however, not very
stable (electron affinity WEA = 24 meV). It may be observed readily as resonance in
low energy electron scattering. We shall come back to this briefly in Sect. 7.2.6.
Section summary
• MOs for heteronuclear molecules are constructed essentially by the same prin-
ciples as those for homonuclear molecules. Of course, the even (g) vs. odd (u)
symmetry is no longer relevant and correlation diagrams become somewhat
more complicated.
• The scheme for filling the orbitals, shown in Table 3.9, is consequently less
transparent and less rigorous than for homonuclear molecules.
• In principle, for all heteronuclear diatomic molecules ionic bonding is impor-
tant – the strength of its influence depending on the specific system. Caused
by this ionic character of the bonding, all have a permanent dipole moment,
i.e. they are microwave and infrared active.
• We have looked into the potential energy diagrams of several specific exam-
ples of heteronuclear, diatomic molecules, LiH, the alkali halides, NaCl and
NaI and finally NO, the ‘Drosophila’ of molecular physics. A number of inter-
esting properties have been discussed. We recall specifically the “harpooning”
mechanism in molecular reactions involving an intermediate ionic complex.
DFT: ‘Density functional theory’, today one of the standard methods for computing
atomic and molecular electron densities and energies (see Sect. 10.3 in Vol. 1).
E1: ‘Electric dipole’, transitions induced by the interaction of an electric dipole
with the electric field component of electromagnetic radiation.
EPR: ‘Electron paramagnetic resonance’, spectroscopy, also called electron spin
resonance ESR (see Sect. 9.5.2 in Vol. 1).
FIR: ‘Far infrared’, spectral range of electromagnetic radiation. Wavelengths be-
tween 3 µm and 1 mm according to ISO 21348 (2007).
FS: ‘Fine structure’, splitting of atomic and molecular energy levels due to spin
orbit interaction and other relativistic effects (Chap. 6 in Vol. 1).
FTIR: ‘F OURIER transform infrared spectroscopy’, (see Chap. 5, p. 298ff.).
FWHM: ‘Full width at half maximum’.
good quantum number: ‘Quantum number for eigenvalues of such observables that
may be measured simultaneously with the H AMILTON operator (see Sect. 2.6.5
in Vol. 1)’.
HF: ‘H ARTREE -F OCK’, method (approximation) for solving a multi-electron
S CHRÖDINGER equation, including exchange interaction.
HITRAN: ‘High-resolution transmission molecular absorption database’, http://
www.cfa.harvard.edu/hitran (ROTHMAN et al. 2009).
HOMO: ‘Highest occupied molecular orbital’.
IR: ‘Infrared’, spectral range of electromagnetic radiation. Wavelengths between
760 nm and 1 mm according to ISO 21348 (2007).
isotopologue: ‘Molecules that differ only in their isotopic composition’, http://en.
wikipedia.org/wiki/Isotopologue.
LCAO: ‘Linear combination of atomic orbitals’, linear superposition of atomic, sin-
gle electron wave functions to form a molecular orbital (MO).
LUMO: ‘Lowest unoccupied molecular orbital’.
MO: ‘Molecular orbital’, single electron wave function in a molecule; typically the
basis for a rigorous molecular structure calculation.
NIR: ‘Near infrared’, spectral range of electromagnetic radiation. Wavelengths be-
tween 760 nm and 1.4 µm according to ISO 21348 (2007).
NIST: ‘National institute of standards and technology’, located at Gaithersburg
(MD) and Boulder (CO), USA. http://www.nist.gov/index.html.
NMR: ‘Nuclear magnetic resonance’, spectroscopy, a rather universal spectro-
scopic method for identifying molecules (see Sect. 9.5.3 in Vol. 1).
RKR: ‘RYDBERG -K LEIN -R EES’, precise method for determining molecular po-
tentials from spectroscopic data.
SCF: ‘Self-consistent field’, method for solving coupled integro-differential equa-
tions iteratively.
SI: ‘Système international d’unités’, international system of units (m, kg, s, A, K,
mol, cd), for details see the website of the Bureau International des Poids et Mé-
sure http://www.bipm.org/en/si/ or NIST http://physics.nist.gov/cuu/Units/index.
html.
vdW: ‘VAN DER WAALS’, interaction between atoms or molecules, ∝ R −6 .
228 3 Diatomic Molecules
VMI: ‘Velocity map imaging’, experimental method for registration (and visual-
ization) of particle velocities as a function of their angular distribution (see Ap-
pendix B).
References
A LEKSEYEV , A. B., H. P. L IEBERMANN, R. J. B UENKER, N. BALAKRISHNAN, H. R. S ADEGH -
POUR , S. T. C ORNETT and M. J. C AVAGNERO : 2000. ‘Spin-orbit effects in photodissociation
of sodium iodide’. J. Chem. Phys., 113, 1514–1523.
B ORN , M. and R. O PPENHEIMER: 1927. ‘Zur Quantentheorie der Molekeln’. Ann. Phys. Berlin,
84, 0457–0484.
C OOPER , D. L., S. B IENSTOCK and A. DALGARNO: 1987. ‘Mutual neutralization and chemiion-
ization in collisions of alkali-metal and halogen atoms’. J. Chem. Phys., 86, 3845–3851.
C ÔTE , R., M. J. JAMIESON, Z. C. YAN, N. G EUM, G. H. J EUNG and A. DALGARNO: 2000.
‘Enhanced cooling of hydrogen atoms by lithium atoms’. Phys. Rev. Lett., 84, 2806–2809.
D ULIEU , O. and C. G ABBANINI: 2009. ‘The formation and interactions of cold and ultracold
molecules: new challenges for interdisciplinary physics’. Rep. Prog. Phys., 72, 086401.
D UNHAM , J. L.: 1932. ‘The energy levels of a rotating vibrator’. Phys. Rev., 41, 721–731.
F LEMING , H. E. and K. N. R AO: 1972. ‘A simple numerical evaluation of Rydberg-Klein-Rees
integrals – application to X1 Σ + state of 12 C16 O’. J. Mol. Spectrosc., 44, 189–193.
GAMESS: 2010. ‘General atomic and molecular electronic structure system’, Gordon research
group at Iowa State University, USA. http://www.msg.chem.iastate.edu/gamess/, accessed:
9 Jan 2014.
G AUSSIAN: 2013. ‘Gaussian 09 rev. D’, Gaussian, Inc., Wallingford, CT, USA. http://www.
gaussian.com/, accessed: 9 Jan 2014.
G ILMORE , F. R.: 1965. ‘Potential energy curves for N2 , NO, O2 and corresponding ions’. J. Quant.
Spectrosc. Radiat. Transf., 5, 369–389.
G RISENTI , R. E., W. S CHÖLLKOPF, J. P. T OENNIES, G. C. H EGERFELDT, T. KÖHLER and
M. S TOLL: 2000. ‘Determination of the bond length and binding energy of the helium dimer
by diffraction from a transmission grating’. Phys. Rev. Lett., 85, 2284–2287.
H ERSCHBACH , D. R., Y. T. L EE and J. C. P OLANYI: 1986. ‘The N OBEL prize in chemistry:
for their contributions concerning the dynamics of chemical elementary processes’, Stockholm.
http://nobelprize.org/nobel_prizes/chemistry/laureates/1986/.
H ERZBERG , G.: 1989. Molecular Spectra and Molecular Structure, vol. I. Diatomic Molecules.
Malabar, Florida: Krieger Publishing Company, 660 pages.
H OUGEN , J. T.: 2001. ‘The calculation of rotational energy levels and rotational line intensities
in diatomic molecules (version 1.1)’, NIST. http://physics.nist.gov/DiatomicCalculations, ac-
cessed: 9 Jan 2014.
H UBER , K.-P. and G. H ERZBERG: 1979. Constants of Diatomic Molecules. New York: Van Nos-
trand Reinhold.
ISO 21348: 2007. ‘Space environment (natural and artificial) – Process for determining solar irra-
diances’. International Organization for Standardization, Geneva, Switzerland.
J ENOUVRIER , A., M. F. M ERIENNE, B. C OQUART, M. C ARLEER, S. FALLY, A. C. VANDAELE,
C. H ERMANS and R. C OLIN: 1999. ‘Fourier transform spectroscopy of the O2 Herzberg
bands – I. Rotational analysis’. J. Mol. Spectrosc., 198, 136–162.
J UARROS , E., K. K IRBY and R. C ÔTE: 2006. ‘Laser-assisted ultracold lithium-hydride molecule
formation: stimulated versus spontaneous emission’. J. Phys. B, At. Mol. Phys., 39, S965–S979.
K LAUS , T., S. P. B ELOV and G. W INNEWISSER: 1998. ‘Precise measurement of the pure rota-
tional submillimeter-wave spectrum of HCl and DCl in their υ = 0, 1 states’. J. Mol. Spectrosc.,
187, 109–117.
K LING , M. F. et al.: 2006. ‘Sub-femtosecond control of electron localization in molecular disso-
ciation’. Science, 312, 246–248.
References 229
Overview
Section 4.1 presents the arbitrarily shaped, rigid rotor, and Sect. 4.2. in-
troduces normal coordinates used to treat vibrations. Section 4.3 addresses
the symmetries of point groups as a key principle to classify polyatomic
molecules. The specialties of the electronic structure of polyatomic molecules
are introduced in Sect. 4.4 by way of example, starting with H2 O. We then in-
troduce the important concept of orbital hybridization by discussing sp 3 MOs
for CH4 and NH3 (Sect. 4.4.2). Finally, in Sect. 4.5 conjugated hydrocarbon
systems are introduced. A simple, quantitative description is provided by the
H ÜCKEL approximation (HMO).
4.1.1 General
We first recall some classical mechanics and focus on the case of an arbitrarily
shaped, rigid rotor. We assume that the Nnu atomic nuclei of the molecule with
masses m1 , m2 , . . . , mk , . . . , mN are positioned in 3D space at R 1 , R 2 , . . . , R k ,
. . . , R Nnu which are constant in respect of their distance from the centre of mass
and their relative orientation within the molecule. The moments of inertia of such a
molecule are described by a second rank tensor, the inertia tensor I with the 3 × 3
components
Nnu
Iij = mk R 2k δij − Rk,i Rk,j , (4.1)
k=1
where Rk,1 = Xk , Rk,2 = Yk and Rk,3 = Zk .
If the molecule rotates around an arbitrary axis with an angular velocity ω, its
angular momentum (denoted by N as in Chap. 3) is
N = Iω.
1
3
IR = e $
R IeR = Ri Iij Rj . (4.2)
R2
ij =1
There is always one body fixed coordinate system, let us call its axes abc, in which
the inertia tensor becomes diagonal. By convention, the thus defined principle mo-
ments of inertia are ordered:
Ia ≤ I b ≤ I c . (4.3)
In respect of an axis e$
R = (a1 b1 c1 ) in the body fixed coordinate system, with
a1 + b1 + c1 = 1, the moment of inertia (4.2) becomes
2 2 2
1 = Ia a 2 + Ib b 2 + Ic c 2 . (4.5)
K
b
γ
X α Y',Y''
α
X' β Y
X'' γ a
We have to expand the notation of Sect. 3.3.2 for the components of the angular mo-
mentum operator in respect of the space fixed system, N X , N
Y and N Z , as well as in
respect of the three body fixed, orthogonal principle axes of the ellipsoid of inertial,
a , N
N b and Nc , which relate to the square of the angular momentum operator by
2 = N
N X2 + N
Y2 + N
Z2 = N
a2 + N
b2 + N
c2 . (4.6)
2 and its components obey the usual angular momentum algebra and commutation
N
rules in the space fixed as well as in the body fixed coordinate system. One may find
eigenstates |N MK which are simultaneously eigenstates of N 2, NZ and Nc :
In analogy to the spherical components of the angular momenta J± = ∓(Jx ± iJy )/
√
2 used for atoms in Vol. 1, one may compose the corresponding components of
X and N
N Y or of Na and Nb :1
√
N a ± iN
± = ∓(N b )/ 2. (4.8)
A close analysis shows, however, that the matrix elements in the body fixed coor-
dinate system correspond to the conjugate complex operators of those in the space
fixed system (see e.g. VAN V LECK 1951). Thus, with (B.15) in Vol. 1 we have:
±
N (M ± 1)K N
|N MK = ∓ N(N + 1) − M(M ± 1) /2 (4.9)
±
N M(K ∓ 1)N |N MK = ∓ N(N + 1) − K(K ∓ 1) /2. (4.10)
a ±iN
1 Note that we use, here too, the orthonormalized operators in contrast to the combinations N b
often used in the literature.
234 4 Polyatomic Molecules
2 2 2
A= , B= and C = , (4.12)
2Ia hc 2Ib hc 2Ic hc
sorted according to (4.3) as A ≥ B ≥ C.
The simplest case is clearly the spherical rotor (Ia = Ib = Ic = I ) describing sym-
metric molecules such as CH4 , SF6 and the like. The S CHRÖDINGER equation be-
comes
2 N(N + 1)
WN = = Bhc N (N + 1) (4.14)
2I
quite analogous to (3.37). However, we have now a (2N + 1)2 fold degeneracy – in
contrast to the linear molecule which the degeneracy is only (2N + 1).
Still relatively clear is the rigid symmetric rotor (short: symmetric top), characterized
by a well defined, at least threefold symmetry axis, with two identical principle
moments of inertia.
The prolate top (cigar shaped) symmetric rotor Ia < Ib = Ic and thus A >
B = C. Examples are methyl-chloride Cl−CH3 , chloroform CHCl3 or propyne
CH3 C≡CH. Alternatively, the oblate top (pancake like) is characterized by Ia =
Ib < Ic and thus A = B > C. Examples are ammonia NH3 as well as all planar
molecules such as benzene C6 H6 . The respective ellipsoids of inertia are shown in
Fig. 4.2.
4.1 Rotation of Polyatomic Molecules 235
c c
(a) (b) K
N
N
b
a K
a b
Fig. 4.2 (a) Prolate rotational ellipsoid with angular momentum N and projection K onto the
symmetry axis, here a. (b) Oblate rotational ellipsoid with symmetry axis c
To solve the S CHRÖDINGER equation one rewrites the Hamiltonian (4.11) suit-
ably, e.g. for the prolate symmetric top as
2 2 2
rot = 1 N
H c2 + Na = N + 1 − 1 N
b + N a2 , (4.15)
2Ib 2Ia 2Ib 2Ia 2Ib
so that the S CHRÖDINGER equation can be solved in closed form. Using (4.7) we
obtain
2 2 2 1
Hrot |N MK = N(N + 1) + K −
1
|N MK. (4.16)
2Ib 2 Ia Ib
The rotational energy of the prolate top is thus given by
The rotational energy obviously depends now also on the projection |K| = 0, 1,
. . . , N of the angular momentum N onto the figure axis. The resulting energy terms
are illustrated in Fig. 4.3 as a function of N and K. According to (4.17) and (4.18),
respectively, the prefactor of K 2 is positive for the prolate rotor (A > B = C), and
236 4 Polyatomic Molecules
0 0 0 0
000
negative for the oblate rotor (A = B > C). This may be visualized from Fig. 4.2
by recalling from (4.12) that for a given value of N the axis with highest rotational
energy corresponds to the lowest moment of inertia. For the prolate top the rota-
tional energy is largest if the angular momentum N is as parallel as possible to the
symmetry axis (here a), i.e. for K = N ; in contrast, for the oblate top the energy is
is perpendicular to the symmetry axis (here c), i.e. for K = 0 when it
highest if N
N lies within the ab plane. In both limiting cases, (4.18) becomes identical to the
corresponding expression (3.37) for the linear rigid rotor, with I = Ib .
may assume different orientations in space, expressed in (4.16) by the
Finally, N
quantum number M. Terms with K = 0 are thus 2N + 1 fold degenerate, those with
|K| > 0 however 2(2N + 1) fold since K may be positive or negative.
We finally note (without proof) that a planar molecule with at least one three-
fold symmetry axis is an oblate top for which Ic = 2Ia = 2Ib , i.e. A = B = 2C
holds. In contrast, nonlinear triatomic species as e.g. water, H2 O, are asymmetric
top molecules which we shall discuss next.
In the general case with Ia = Ib = Ic , the asymmetric rigid rotor (short: asymmetric
top), simply rewriting the Hamiltonian as done in the last subsection does no longer
help: The energies and eigenfunctions of the asymmetric top cannot be given in
closed form – not even in the case of a rigid rotor. Nevertheless, many important
molecules belong to this class, e.g. those with a twofold symmetry axis such as
H2 O.
One may, however, obtain a qualitative estimate for this situation by simply in-
terpolating the energy terms between prolate and oblate top. This is illustrated in
Fig. 4.4.
The asymmetric top is characterized by three rotational constants A > B > C
which are associated to the three principle axes abc of the ellipsoid of inertia
4.1 Rotation of Polyatomic Molecules 237
(Ia < Ib < Ic ). The prolate top (A > B = C) and the oblate top (A = B > C) are
limiting cases in this notation. The twofold degeneracy for K > 0 in the case of a
symmetric top is now removed. One speaks about Ktype doubling in analogy to Λ
type doubling to which we have been introduced in the context of electronic states
of diatomic molecules Sect. 3.6.6: it follows directly from breaking the symmetry
and is characterized here by two quantum numbers Ka and Kb . In the limit of the
prolate rotor Ka becomes what we have called K, the projection of the angular mo-
mentum onto the axis of the smallest moment of inertia; in the case of the oblate top
Kc becomes the projection onto the axis with the largest moment of inertia. Clearly,
each of the |N Ka Kc states is still 2N + 1 fold degenerate, since here too, the an-
gular momentum N may have 2N + 1 orientations in space (again described by the
quantum number M).
The exact computation of the eigenstates and eigenenergies is somewhat elabo-
rate. In principle the Hamiltonian (4.11) is written as
+ 2 − 2
rot = α N
H 2 + βN
c2 + γ N
+ N , (4.19)
where N+ and N − are the operators defined in (4.8). The constants α, β and γ are
different linear combinations of A, B and C, chosen according to the ratios A : B : C
in such a way that the constant γ describing the deviations from the symmetric top
becomes as small as possible. The eigenstates |N MΓ of the Hamiltonian may then
2 and N
be written as linear combination of those for the symmetric top, i.e. of N c :
N
|NMΓ = fN K |N MK.
K=−N
With this ansatz one may diagonalize the Hamiltonian (4.19) for each N , using the
matrix elements (4.9) and (4.10). For larger values of N this leads to increasingly
complicated expressions. Moreover, the rigid rotor is only a first approximation.
In order to meet spectroscopic accuracy one has to include centrifugal distortion,
vibration-rotation coupling and coupling with the electronic angular momentum in
the spirit of H UND’s cases, possibly also hyperfine interaction. Also, vibronic cou-
plings may play an important role, as we shall discuss in connection with the JAHN -
T ELLER effect in Sect. 4.3.4.
Numerical methods and approximations have been developed for this purpose
and extended reviews and monographs have been written on this subject. Today
one interprets the experimental spectra by direct numerical comparison of a suitably
parameterized ansatz and with the help of sophisticated simulation programmes.
We cannot go into details here, but sketch as an example the situation for the H2 O
molecule. Pure rotational spectra are found in the sub-millimetre range, but only few
experimental data are available. An overview of the rotational level scheme for 0 ≤
N ≤ 3 is shown in Fig. 4.5. It has been obtained (within the highly simplified model
of the rigid rotor) from the convenient computer programme PGOPHER (2013) in
the online version.
238 4 Polyatomic Molecules
WN K K (hc) -1 / cm -1
left to right according to 321 322
increasing Kc and decreasing 200
Ka derived with the help of 312
PGOPHER (2013) 313 303
220 221
a c
100 211
212 202
110 111
101
0 000
H H
b
The parameters used in this computation are derived from the H2 O geome-
try, shown in Fig. 4.6 and the known atomic masses. The three moments of in-
ertia are Ia = 0.00632 u nm2 , Ib = 0.01154 u nm2 , and Ic = 0.01786 u nm2 ; due
to the light H atom, the corresponding rotational constants are rather large: A0 =
27.880591 cm−1 , B0 = 14.5216246 cm−1 , and C0 = 9.27774594 cm−1 (B ERNATH
2002b).
Section summary
• Polyatomic molecules have in general 3 rotational degrees of freedom. From
the inertia tensor (4.1) one obtains the moments of inertia for a given axis in
the direction eR by IR = e$ R IeR . In diagonalized form I is expressed by the
three principle moments of inertia, Ia ≤ Ib ≤ Ic . They may be visualized √ by
the√ellipsoid
√ of inertia (4.5) whose three principle axes are given by 1/ Ia ,
1/ Ib , 1/ Ic , respectively. The rotational constants (4.12) are A ≥ B ≥ C,
correspondingly.
• The angular momentum N of an arbitrarily shaped rotor and its components
N b , N
a , N c in respect of the principle axes obey the usual commutation rules
and eigenvalue equations for angular momenta. The total energy of the rotor
(4.11) is given by the sum of the energies of the three components.
• For the spherical (or symmetrical) rotor with three identical moments of in-
ertia (molecules such as CH4 ) the energy is simply WN = Bhc N (N + 1),
just as in the case of a diatomic molecule. The degeneracy is now, however,
(2N + 1)2 fold.
4.2 Vibrational Modes of Polyatomic Molecules 239
• The symmetric top too can be solved in closed form, leading to energies
(4.17)–(4.18) for the prolate (A > B = C) and oblate top (A = B > C), re-
spectively. They depend on the total angular momentum N and on its projec-
tion onto the figure axis, |K| = 0, 1, . . . , N . Terms with K = 0 are 2N + 1
fold, with |K| > 0 they are 2(2N + 1) fold degenerate.
• The asymmetric top is more complicated but may be understood as general
case between prolate and oblate top.
We note at this point that with present state of computation and computerized vi-
sualization techniques a host of sources in the internet provide useful, illustrative
information on molecular vibrations. Specifically, for small polyatomic molecules
we mention I MMEL (2012), who relates the normal modes of many common small
molecules to their infrared spectra. A host of useful links are collected at J MOL
(2011) to web-pages using Jmol: an open-source Java viewer for chemical struc-
tures in 3D (which appears to become the standard for 3D molecular visualization).
A molecule consisting of Nnu atoms has 3Nnu degrees of freedom since each atom
can move into three spatial directions. Of these 3Nnu degrees of freedom three de-
scribe the overall motion of the molecule (translation of the centre of mass). Ro-
tation of the molecule (in the general case) is described by three more degrees of
freedom (in the case of a linear molecule by two degrees of freedom). This leaves
us with
3Nnu − 6
degrees of freedom (3Nnu − 5 for linear molecules) to describe the internal motion,
i.e. the vibrations of the molecules.
Each atomic nucleus may oscillate around its equilibrium distance. We designate
the (3Nnu − 6) relative displacement coordinates – measured in a body fixed sys-
tem – as ξi . They describe all internal motion of the Nnu atoms in the molecule. For
small oscillations around the equilibrium we may, just as in the case of a diatomic
molecule, expand the potential into a series:
3N nu −6 3Nnu −6
∂V 1 ∂ 2 V
V (ξ1 . . . ξNnu ) = V0 + ξ + ξi ξj . (4.20)
∂ξi ξi =0 ∂ξi ∂ξj 0
i
2
i i,j
We set V0 = 0, i.e. we choose the absolute minimum of the potential as zero en-
ergy. Since we have expanded around the equilibrium the first partial derivatives
are ∂V /∂ξi |ξi =0 = 0, and the total energy is written as the sum of kinetic (T ) and
240 4 Polyatomic Molecules
potential energy (V ):
3Nnu −6 3Nnu −6
1 1 ∂ 2 V
W =T +V = mi ξ̇i +
2
ξi ξj . (4.21)
2 2 ∂ξi ∂ξj 0
i i,j
with:
and the so called Hessian matrix V
∂ 2 V
Vij = . (4.23)
∂qi ∂qj 0
Die Hessian matrix is real, symmetric (Vij = Vj i ) and positive definite, since the
potential has a minimum for qi = 0. With this the energy (4.21) becomes
3Nnu −6 3Nnu −6
1 1
W= q̇i2 + Vij qi qj , (4.24)
2 2
i i,j
or more compact with (3Nnu − 6) dimensional row and column vectors, q $ and q,
respectively:
1 1
W = q̇ $ · q̇ + q $ V q. (4.25)
2 2
In general Vij = 0, and the cross terms qi qj in the sum (4.24) do not vanish: in
general, the vibrations are coupled.
One thus searches for a new coordinate system Qi , in which this coupling is
removed. Since the matrix V is symmetric and real, an orthogonal matrix2 A
exists
which diagonalizes it:
A A
−1 V = Ω.
(4.26)
It is found as usual by solving the characteristic equation det(V − λ 1) = 0. The
elements of the diagonal matrix Ω are the eigenvalues λi of V . Since V is positive
definite, all its eigenvalues are also positive, thus we write λi = ωi2 . The eigenvectors
of V are then found by solving the system of linear equations
V − ωi2
1 ai = 0
for each eigenvalue ωi2 . The transformation matrix is the matrix of the eigenvectors:
= a1
A a2 ... ai ... a Nnu .
q = AQ or Q=A $ q.
−1 q = A (4.27)
Inserting this into (4.25), one obtains with a few lines of algebra
1 $ 1 1 2
=
W = Q̇ Q̇ + Q$ ΩQ Q̇i + ωi2 Q2i . (4.28)
2 2 2
i
Hence, the vibrational energy can be written by a single sum over i. That is equiv-
alent to saying that motions in these coordinates are decoupled. The Qi describe
3Nnu − 6 independent, harmonic oscillators.
We identify (4.28) as the classical H AMILTON function, with the canonical pairs
of position and momentum coordinates, Qi and Pi = ∂Ti /∂ Q̇i = Q̇i , respectively:
1 2
H= (Ti + Vi ) = Pi + ωi2 Q2i . (4.29)
2
i i
∂H ∂H
Q̇i = and Ṗi = − = −ωi2 Qi = Q̈i
∂Pi ∂Qi
=⇒ Q̈i + ωi2 Qi = 0,
the different coordinates are completely decoupled, and the individual solutions are
Qi (t) = Qi (0)e±iωi t .
The motions in the coordinates Qi are thus simple, harmonic vibrations of fre-
quency ωi . They are called normal modes of the molecule. Back transformation
into the original (mass weighted) molecular coordinates is done by superposing all
normal modes
q = AQ or qj = Aj k Qk . (4.30)
k
Such a normal mode describes in the general case complex (harmonic) motions
of all atoms in a molecule. If only one normal mode Qi is excited, we find qj =
Aj i Qi , i.e. all atoms j oscillate with equal frequency ωj and in phase (if Aj i = 0):
a normal mode is delocalized over the whole molecule. Conversely, so called local
modes for which dominantly one bond vibrates, may be constructed by judiciously
superposing several normal modes.
242 4 Polyatomic Molecules
2 d 2 1
=
H i
H i = −
with H + ωi2 Q2i . (4.31)
2 dQi 2 2
i
with 3Nnu − 6 (or −5) factors for all Qi . For each coordinate one vibrational quan-
tum number vi describes the energy and the total energy becomes
W= (vi + 1/2)ωi . (4.33)
i
For large molecules this leads to substantial internal energies already at ther-
mal excitation. Even the zero point energy may be high: The famous “football”
molecule, C60 , e.g. has 174(!) normal modes (the lowest mode energy corresponds
to ν̄ 500 cm−1 ). The calculation leads to a zero point energy on the order of
Wmin 5.4 eV (we mention the N OBEL prize 1996 honouring the discovery of C60
by C URL, K ROTO, and S MALLEY).
The selection rules for dipole transitions are derived in analogy to the considera-
tions in Sect. 3.4. In harmonic approximation
The permanent dipole moment of the molecule −eDγ (Q), in an electronic state γ
(3.79), now depends on all normal coordinates Q$ = (Q1 Q2 . . . Qi . . . ). A series
expansion around Qi = 0 gives in analogy to (3.80)
∂ Dγ
Dγ (Q) = Dγ (Q0 ) + Qi + · · · . (4.34)
∂Qi 0
i
As discussed in Sect. 3.4.4, here too a normal mode Qi is infrared active (with
vi = vi ± 1), if and only if ∂ Dγ /∂Qi |0 = 0. All other vibrational modes remain
unchanged in such a one-photon transition.
4.2 Vibrational Modes of Polyatomic Molecules 243
q4 q5 q6
Efficient programmes are available today for computing normal modes of poly-
atomic molecules by diagonalization of the Hessian matrix. Important simplifica-
tions are achieved by considering the molecular symmetry. For the various point
groups according to which molecules may be classified (see Sect. 4.3) group theory
provides suitable tools when determining the normal modes. We refrain from treat-
ing these procedures in detail and just summarize in the following a few results for
instructive examples. It must be pointed out, however, that the harmonic approxima-
tion just discussed, is again only a first order approximation. Today’s spectroscopic
accuracy affords much higher quality of the theory – and correspondingly allows
highly precise structural determinations for even rather complex molecules.
The coordinates qi and masses mA,B of a linear AB2 molecule are specified in
Fig. 4.7. It has 3Nnu − 5 = 4 internal degrees of freedom, and thus four normal
modes: two vibrations along the molecular axis and two bending modes perpendic-
ular to it (one in the paper plane and one perpendicular to it). The normal coordinates
are plausible by realizing that the centre of mass of the molecule must remain at rest
during normal motions:
3 Inthe older literature (e.g. H ERZBERG 1991) one finds for ν̄1 = 1 388.3 cm−1 and for 2ν̄2 =
1 285.5 cm−1 . This leads to an interchange of the states (100 0) and (020 0).
4.2 Vibrational Modes of Polyatomic Molecules 245
(011 0)
(0000)
0
Something similar happens between the states (111 0) and (033 0). The two F ERMI
resonance pairs are marked red in Fig. 4.8. F ERMI resonances are a rather common
phenomenon in atomic and molecular physics.
Also indicated in Fig. 4.8 are some infrared transitions (grey arrows). Of specific
importance are those at 9.4 µm and 10.4 µm wavelengths: these are the most im-
portant transitions in the CO2 laser which still is a work horse in material process-
ing technologies (welding, cutting, drilling of massive material). Due to the closely
spaced rotational lines the CO2 laser (exploiting several isotopes) is tuneable over a
spectral range from 9.2 µm to above 11 µm – quasi continuously.
The infrared spectroscopic data for CO2 are excellently documented (e.g. in
the HITRAN data bank by ROTHMAN et al. 2009). As an example, we show in
Fig. 4.9 a simulated absorption spectrum for the (weak) vibration-rotation transition
(11102) ← (00001) in CO2 .4
The Q branch makes this spectrum particularly interesting. It is observed here in
addition to the P and R branch which we know already from diatomic molecules.
For the diatomic molecules (see Sect. 3.4.5) we had excluded N = 0 transitions
for parity reasons. Albeit CO2 is also a linear molecule, for this bending vibration
the change of the dipole moment is perpendicular to the molecular axis. And it is this
change which according to (4.34) is responsible for transitions. Hence, its average
over YN∗ M (Θ, Φ)YN M (Θ, Φ) does not disappear in this case, and the transition is
allowed (so called “perpendicular” transition).
4 HITRAN notation. See Sect. 5.2.4, Vol. 1 how to convert spectroscopic line strengths into spectra.
246 4 Polyatomic Molecules
1.0
0.9
transmission
0.8
0.7
1880 1900 1920 _ 1940 1960 1980
ν / cm-1
Fig. 4.9 Simulated transmission spectrum for the infrared bands of the (weak) vibrational transi-
tion (11101) ← (00001) in 12 C16 O2 , obtained from HITRAN-W EB (2012), using the HITRAN
databank (296 K, 1 atm, opt. path 1 m, app. resolution 0.1 cm−1 ). Note that for this perpendicular
transition one observes – in addition to the P and R branch – also a Q branch
Table 4.1 Normal modes of ν̄1 / cm−1 ν̄2 / cm−1 ν̄3 / cm−1
the water molecule (ground 16 O
state) for the three most H2 3657.053 1594.746 3755.929
important isotopologues HD 16 O 2723.68 1403.48 3707.47
D2 16 O 2669.40 1178.38 2787.92
positive. The three vibrational modes are sketched in Fig. 4.10. We refrain here
from a mathematical description. For all three modes ∂ Dγ /∂Qi |0 = 0 holds, they
are infrared active. However, as a detailed calculations show, their absorption cross
sections are rather different, with σ (ν1 ):σ (ν̄2 ):σ (ν̄3 ) 0.07:1.47:1.00. A look at
Fig. 4.10 makes it plausible that the change of dipole moment is smallest for exci-
tation of the symmetric stretch mode, ν̄1 . The eigenfrequencies for the three most
important isotopologues of H2 O are summarized in Table 4.1.
The term energies of the lowest vibrational levels are shown in Fig. 4.11. It gives
already a feeling of the high complexity in the overall vibrational spectrum of the
water molecule: the validity of the harmonic approximation is rather limited and nu-
merous higher harmonic bands are observed. As in the case of diatomic molecules,
superposed onto the vibrational excitation is the rotational structure. It is evident
from the discussion in Sect. 4.1.4 that these bands are much more complicated
than for diatomic molecules. The coupling of vibration and rotation, centrifugal
4.2 Vibrational Modes of Polyatomic Molecules 247
(000)
0
0.10
absorption spectrum of
vibration-rotation bands in
the H2 16 O molecule in a 0.08
small section of the visible
spectrum according to
C ARLEER et al. (1999). One 0.06
sees overtones of the
fundamental vibrational
modes with the corresponding 0.04
rotational structure 16870 16880 16890 16900
wavenumber / cm-1
Q2 (a) (b) ν1 ν2
N N
N
0 nm H H H H H H
1 24 106.67o
10
0. H ν3 ν4
- Re N N
H
H H H H H H
H
Fig. 4.13 NH3 : (a) Coordinate system: x is the distance of the N atom from the (H-H-H) plane,
Re = 0.0383 nm is the equilibrium distance of the inversion mode; (b) shows the normal modes,
of which ν3 and ν4 are each twofold degenerate
only slightly from the ideal tetrahedral angle 109.47◦ .5 The equilibrium distance
of the inversion coordinate x is
α 1 2 α
Re = R0 cos2 − sin = 0.0381 nm.
2 3 2
M̄ = 3mH mN /(3mH + mN )
5 One
also finds slightly different values for R0 and α in the literature. For a recent discussion see
YACHMENEV et al. (2010).
4.2 Vibrational Modes of Polyatomic Molecules 249
H V(Q 2) / cm-1 H
(a) (b)
H N N H
H H
- v2 = 4
597 4.5ν 2
+ 4000
v2 = 3
-
511.4
3.5ν 2
3000
+
Vb v2 = 2
35.7 284.7
-
2.5ν 2
+
2000
1.5ν 2 v2 = 1
+ -
1000
ν 2/2
0.793
v 2 =0
+ -
- Re Re
Q2
-0.8 -0.4 0.0 0.4 -0.8
Fig. 4.14 Cut through the NH3 potential hypersurface in the direction of the inversion vibra-
tion (red lines) and the corresponding harmonic approximation (grey). The term positions for the
ν2 mode are indicated by horizontal lines – (a) exact values accounting for tunnelling splitting
(in cm−1 ), (b) harmonic approximation
in a potential with a barrier. According to DAMBURG and P ROPIN (1972) one may
use a model potential6
2
V (x) = k Re2 − x 2 / 8Re2 (4.35)
sketched in Fig. 4.14(a) as dashed, red line. The force constant k = M̄ω22 and the
two equilibrium distances x = ±Re are derived from the experimental data. V (x) is
a good approximation in the region of the vibrational ground state of the ν2 mode.
Close to the barrier height it is, however, no longer correct. In Fig. 4.14(a) we thus
have drawn V (x) slightly corrected by eye (full red line).
In the classical picture a particle of energy W < Vb cannot cross the barrier. The
quantum mechanical treatment of the tunnelling process is done without difficul-
ties by numerical integration of the one-dimensional S CHRÖDINGER equation for
the mass M̄ moving along the inversion coordinate x on a potential hypersurface as
good as possible. The term positions shown in Fig. 4.14(a) correspond the experi-
mentally determined values which can be fully reproduced by theory.
As a 1st order approximation V (x) according to (4.35) may be used. But a good
physical understanding of the tunnelling process and the ensuing level splitting is
obtained already from considering eigenfunctions of harmonic oscillators around
6 Note that the position x of N in respect of the H3 plane is here not mass scaled. The corresponding
√
normal coordinate is given by Q2 = M̄x.
250 4 Polyatomic Molecules
the two equilibrium positions. In the vibrational ground state where the two minima
are separated by a high barrier, we first assume that the oscillation is localized fully
in the left or right potential minimum. Harmonic potentials according to Fig. 4.14(b)
are fitted to each minimum as indicated by grey lines in Fig. 4.14(a). In this 0th
order approximation eigenfunctions correspond to Table 3.1. The respective ener-
gies are plotted in Fig. 4.14(b) and characterized by vibrational quantum numbers
v2 = 0, 1, . . . . Each term is twofold degenerate, corresponding to the two possible
positions.
Since the two positions are physically completely equivalent, we cannot distin-
guish them at all. Thus we have to combine them to symmetric and antisymmetric
states – quite analogue to the electronic eigenfunctions of the H+ 2 molecule (see
Sect. 3.5.2). In other words, the potential V (x) has even symmetry, i.e. the parity
operator P and the Hamiltonian H of the system commute: [H = 0 and we
, P]
define common eigenstates of P and H by symmetric and antisymmetric linear
combinations:
+ 1 (r) (l)
v = √1 v (r) + v (l) and v2− = √ v2 − v2 . (4.36)
2 2 2
2 2
Specifically, for the vibrational ground state the respective wave functions R+ (x)
and R− (x) are constructed from
depending on whether the molecule is close to the left or the right minimum, as
sketched in Fig. 4.15.
In perturbation theory this approach leads to the removal of degeneracy. The per-
turbation potential is the difference between inversion potential V (x) and the two
isolated, harmonic oscillator potentials (red and gray lines in Fig. 4.14(a), respec-
tively). In the barrier region it is negative, and we note (insets in Fig. 4.15) that the
antisymmetric wave function R− (x) has there a lower probability than the sym-
metric eigenfunction R+ (x). Thus, for the corresponding eigenenergies one finds
− +
Wv2 > Wv2 , i.e. the energies of the antisymmetric states lie higher than those for the
symmetric states – as illustrated in Fig. 4.14(a).
Since the probability in the barrier region is the higher the closer the eigenenergy
of the states Wv2 comes to the barrier, higher lying states will split more. This trend
obviously continues above the barrier. Experimentally best determined is the split-
ting in the ground state where the two modes are separated by the so called inversion
frequency:
ω0 W0− − W0+
ν0 = = = 23.870 GHz. (4.37)
2π h
4.2 Vibrational Modes of Polyatomic Molecules 251
As the wave functions R+ and R− have different parity and the molecule has a
permanent dipole moment along the x-coordinate, E1 transitions may be induced
between these states. For the ground state the transition frequency in ammonia lies
in the range of millimetre waves with λ0 = 12.56 mm or ν̄0 = 0.796 cm−1 .
other. Hence ν0 is called inversion frequency and corresponds exactly to the en-
ergy difference between the two slightly split ground states |v2+ = 0 and |v2− = 0.
T OWNES and co-worker (G ORDON et al. 1955) were first to show that for the
example ammonia amplification is possible by stimulated emission. Their maser
(Microwave Amplification by Stimulated Emission of Radiation) was the prede-
cessor of the laser. As explained in Sect. 1.1 amplification by stimulated emis-
sion requires a population inversion among the levels of the transition of interest.
This is not possible in thermodynamic equilibrium. In the present case we have
ω0
kT at room temperature, i.e. nearly equal population of both states.
However, in a molecular beam both states may be separated spatially very ef-
ficiently using an inhomogeneous electric fields (S TARK effect) as schematically
indicated in Fig. 4.16. The process is based on the fact that both states have an
opposite electric permanent dipole moment. They may thus be separated in an in-
homogeneous electric field (analogous to the S TERN -G ERLACH experiment treated
in Sect. 1.9, Vol. 1 where the two electron spin orientations with opposite magnetic
dipole moment were separated in an inhomogeneous magnetic field). In this man-
ner, one deflects the molecules in the energetically higher lying state |v2− from the
beam into a microwave resonator. The latter is tuned onto the inversion frequency
and allows one to detect the amplified microwaves.
Section summary
• There are 3Nnu − 6 degrees of freedom for the vibrational motions in poly-
atomic molecules. Normal modes Qi are such linear combinations of the co-
√
ordinates qi (conveniently mass scaled by mi ) that diagonalize the Hamil-
tonian as shown by (4.31).
• The total vibrational energy is the sum of the energy in all normal modes, the
nuclear motion is described by products of the respective wave functions.
• Optical transitions (typically in the infrared) can be induced in such modes
that change the permanent dipole moment of the molecule.
• As examples of triatomic molecules we have discussed the CO2 (linear) and
H2 O. In both cases the normal modes are: the symmetric stretch, the bending
mode (lowest frequency) and the asymmetric stretch (highest frequency). In
the case of CO2 the bending mode is two fold degenerate and allows for a type
of rotational motion, compensating for the fact that as a linear molecule CO2
has only two degrees of rotational motion.
• A quite particular molecule is the ammonia, NH3 . The symmetry of its nearly
ideal tetrahedral shape leads to two indistinguishable minimum positions,
4.3 Symmetries 253
with the N atom on top or below the H3 plane. Consequently, the six nor-
mal vibrational modes include the so called “umbrella” mode in which the N
atom may tunnel through this plane. This mode is split into a symmetric and
an antisymmetric state, of which the symmetric one has the lowest energy. The
splitting in the ground state is 23.870 GHz and is identical to the tunnelling
frequency.
• The ammonia maser uses this inversion transition. It exploits the fact that
symmetric and antisymmetric ground states have opposite permanent electric
dipole moments. They may be separated in a molecular beam, passing through
an inhomogeneous electric field so that population inversion can be achieved.
4.3 Symmetries
E Identity (no change) or rotation through 360◦ (E originating from the German
word Einheit = unit or unity).
Cn Rotation in respect of a symmetry axis through an angle 2π/n with n =
2, 3, . . . . Applying this operation k times is denoted by Cnk . With these defi-
nitions the identity becomes E = Cnn .
σ Reflection at a plane – for which σ σ = σ 2 = E holds.
7 Note: Quantum mechanical operators associated with these symmetry elements will be designated
n for the operator rotating a wave function
by a caret () on top of the element symbol, e.g. C
through an angle 2π/n.
254 4 Polyatomic Molecules
Rotation Groups
Cn Only rotations though an angle 2π/n around an axis (Cn1 , Cn2 , . . . , Cnk , . . . ,
Cnn = E), i.e. n elements, including E; e.g. hydrogen peroxide (H2 O2 ) has
C2 symmetry.
Cnh Rotation though 2π/n and reflection, σh , in a plane perpendicular to the rota-
tional axis and in addition inversion (for n = 2) or n − 1 improper rotations (for
n > 2), i.e. 2n elements; e.g. boric acid (B(OH)3 ) belongs to the C3h group.
Cnv Rotation though 2π/n as well as n reflections, σv , in planes parallel to the rota-
tional axis, i.e. 2n elements; e.g. the water molecule (H2 O) has C2v , ammonia
(NH3 ) has C3v symmetry.
Dihedral Groups
Dn One Cn axis (principle axis) and n C2 axes perpendicular to the principle axis.
Dnh As Dn , with an additional reflection plane, σh , perpendicular to the principle
axis; e.g ethene (C2 H4 ) belongs to the group D2h and benzene, the simplest
aromatic ring (C6 H6 ) has D6h symmetry.
Dnd As Dn , and in addition n reflection planes σd parallel to the principle axis; e.g.
ethane (C2 H6 ) has D3d symmetry.
4.3 Symmetries 255
Molecule
y n
linear ?
two
y n y or more n
D ∞h i? C∞v
Cn with
n >2 ?
y n y
i? Td C n?
n
C2 ┴
n y
to Cn with Cs σ?
largest
n?
y n
y y y n
Th C 3? Dnh σh ? Ci i? C1
n n y
σh? Cnh
y y
Oh 3C4? Dnd nσd?
n
n n y
nσυ ? Cnv
Ih Dn
n
n y
Cn S2n ? S2n
Fig. 4.17 Decision tree for determining the molecular symmetry in the notation of S CHÖNFLIES
Tetrahedral Groups
T The group of genuine tetrahedral rotations: E, 4C3 , 4C32 , (axes from corners to
the middle of opposite faces), 3C2 (axes from the middle of the edges to the
middle of the opposite edges), in total 12 symmetry elements.
Td As T and in addition the 6σd reflections and the respective 6S4 operations, in
total 24 symmetry elements; e.g. methane (CH4 ).
Th As T and in addition all operations that arise from multiplication the former
with inversion ı, in total again 24 symmetry elements; e.g. C60 Br24 .
Octahedral Group
O The group of true cube rotations: E, 8C3 (four 3-fold axes through the corners
of the cube, with the three elements C31 , C32 and C33 = E, i.e. two new elements
for each axis), 6C4 and 3C2 elements (3 coinciding C2 and C4 axes from the
middle of the faces to the opposite faces, with the elements C21 and C41 , C42 =
C21 , C42 , and C44 = E, i.e. one C2 element and two C4 elements for each axis),
and finally 6C2 (axes from the middle of the edges to the middle of the opposite
edges), in total 24 symmetry elements.
Oh Complete octahedral group: O and in addition the operations due to multipli-
cation with inversion ı (improper rotations) lead to 48 symmetry elements; e.g.
SF6 .
Icosahedral Group
The icosahedral group contains
From atomic physics we are familiar with the full three dimensional rotation group
(called O(3) or SO(3), respectively, depending on whether inversion is included
or the group is redistricted to pure rotation). This symmetry group determines the
behaviour of angular momenta of atoms. Obviously, for polyatomic molecules this
full freedom of rotation has to be replaced by the symmetry operations of the point
groups. Each point group may be represented by a set of irreducible representations
Γi . And the number of irreducible representations in point group is equal to its
dimension (i.e. to the number of different classes of symmetry operations).
The irreducible representations Γi replace, so to say, the angular momentum
states with quantum numbers LM for states (or m for orbitals) in atomic physics –
which are the irreducible representations of the O(3) group. All point groups de-
scribed above are sub groups of O(3). We have already obtained a taste of the en-
suing complications in the context of diatomic molecules. However, there it was
possible to derive the electronic states in a rather straight forward manner by pro-
jecting the spherical harmonics onto the symmetry axis. In the case of nonlinear
tri- and polyatomic molecules the situation is even more complex. The symmetry
groups help with book keeping.
As mentioned above, applying a certain class of symmetry operations onto an ir-
reducible representation (a molecular state, an orbital, a normal mode) corresponds
to a linear transformation in space. It turns out, that the entirety of characters for all
symmetry operations in a group onto one irreducible representation Γi fully char-
acterizes the symmetry properties of this Γi . Hence, the characters of all irreducible
representations of a point group are summarized in a so called character table. For
the simplest groups the symmetry operations have only the characters −1 and +1,
indicating whether upon this particular operation, the irreducible representation Γi
(i.e. the state, orbital, normal mode) does or does not change its sign, respectively.
For degenerate states one does not recognize this result immediately, since in this
case linear combinations of the different degenerate states are transformed (by a
2 × 2, 3 × 3 etc. matrix).
258 4 Polyatomic Molecules
Table 4.2 Notation of irreducible representations of the 3D point groups according to M ULLIKEN
(1955). Cn stands for rotation around the principle axis, C2 and C2 for rotation around one or two
axes, respectively, perpendicular to it; note that the subscripts for B symbols depend on the special
point group and include in some cases also 3
Symbol Degeneracy Symmetry Indices Symmetry in respect of
A 1 + in respect of Cn ı σh C2 C2 σv
B 1 − in respect of Cn sub g, u +, −
E 2 sup , +, −
T 3 sub 1, 2 +, − or +, −
G, H 4, 5 in I and Ih
Since the thirties of the past century one characterizes the irreducible represen-
tations (i.e. the molecular states) by a notation in letters and indices according to
M ULLIKEN (M ULLIKEN 1966), summarized in Table 4.2.
As in atomic physics, electronic orbitals are written in lower case letters, total
states in capital letters. Vibrations are usually (but not always) characterized by
lower case letters, while the letter t is replaced by f . For the point groups C∞v
and D∞h the notation Σ, Π, , . . . is maintained as in Chap. 3. The degeneracy of
these molecular states and vibrations describes the number of possibilities to obtain
the same symmetry properties for different arrangements in space. It corresponds to
the 2L + 1 fold degeneracy of the |LM angular momentum states with different
projections M in respect of the z-axis in the O(3) group. The letters 2S+1 A, 2S+1 B
etc. with their superscripts (sup) and indices (sub) replace the denotation 2S+1 L of
atomic physics (2 S, 1 P etc.), while the multiplicity 2S + 1 of states with well defined
electron spin S is still included in the established form. In detail the rules given
in Table 4.2 for denoting nonlinear molecules are not always unambiguous. Thus,
already M ULLIKEN (1955) recommends “that in general every author, in referring
to Bi or Bip species8 shall define these species clearly in each paper in terms of the
specific geometry of the molecules he is discussing and in so doing, that he shall
make an effort to follow previous usage if such exists and there is no strong reason
to change”.
This is all best explained with some examples at hand and we begin with the
rather simple point group C2v , to which e.g. the water molecule H2 O belongs. Its
electronic structure will be discussed in Sect. 4.4.1.
Its symmetry operations are illustrated in Fig. 4.18: a rotation around the twofold
symmetry axis z, characterized by C2 , and reflection in respect of the xz and the
yz planes, characterized by σv (xz) and σv (yz), respectively. Table 4.3 represents
the so called character tables, of the relevant point groups, here C2v for the full
symmetry shown in Fig. 4.18, and Cs for the case that one of the bonds is stretched,
and one is compressed as in the asymmetric stretch vibration. In the top left cell the
point group is specified, here C2v and Cs , respectively. Below that, the character
tables as such are in the present case 4 × 4 and 2 × 2 matrices. The first column
8 With i = 1, 2, or 3 and p = u or g.
4.3 Symmetries 259
z x
C2
y
O σ'v
H
RO – O
+ θ + H H
H H
σv
D
Fig. 4.18 Geometry and symmetry of the H2 O molecule. On the left the definition of the molec-
ular parameters is shown, on the right the coordinates and symmetry operations of the point group
C2v are indicated (following M ULLIKEN 1955; different authors occasionally use different la-
belling of the axes, e.g. C HAPLIN 2013). D indicates the permanent dipole moment of H2 O along
the z-axis due to excess electron charge at the O atom (−0.8e, |D| = 2.35 D)
Table 4.3 Character tables of the point groups C2v and Cs with basis functions
C2v E C2 σv (xz) σv (yz) Cs E σh
A1 1 1 1 1 z x 2 , y 2 , z2 A 1 1 x 2 , y 2 , z2 , xy
A2 1 1 −1 −1 Rz xy A 1 −1 yz, xz
B1 1 −1 1 −1 x, Ry xz
B2 1 −1 −1 1 y, Rx yz
gives the M ULLIKEN symbols for the irreducible representations Γi of the group
(i.e. the possible total states, orbitals or normal modes), following the scheme given
in Table 4.2. In the first row the possible symmetry operations, including the unit
operation E, are listed. The main content of the table are the so called characters of
the respective symmetry operations for irreducible representations given in column
one. The C2v and Cs are among the simplest cases, where all representations are
singly degenerate.
In this terminology 3 A1 (second row in C2v Table 4.3) characterizes a fully sym-
metric triplet state. An electron configuration, e.g. {. . . (nb2 )2 . . . } (fifth row in C2v
Table 4.3), characterizes a system which contains two electrons (spin up and spin
down) in a (spatially non-degenerate) nb2 orbital which changes its sign upon 180◦
rotation around the z-axis as well as upon reflection on the xz plane.
The character table also contains information about how the Carthesian basis vec-
tors xyz, rotations around them Rx , Ry , Rz , and certain quadratic combinations – all
shown in the two last columns of the C2v Table 4.3 – transform under these symme-
try operations: e.g. polar as well as axial vectors do not change under the respective
rotations, while x, y, z (representing polar vectors) change their sign under reflec-
tion through a plane perpendicular to them, and Rx , Ry , Rz change their sign for
each reflection at a plane which contains them.
260 4 Polyatomic Molecules
H y H
z
Table 4.4 Character table of the D2h point group, e.g. ethylene: C C . Coordinates xyz
aligned as suggested by M ULLIKEN (1955) x
H H
D2h E C2 (z) C2 (y) C2 (x) ı σv (xy) σv (xz) σv (yz)
Ag 1 1 1 1 1 1 1 1 x 2 , y 2 , z2 x 2 , y 2 , z2
B1g 1 1 −1 −1 1 1 −1 −1 Rz xy
B2g 1 −1 1 −1 1 −1 1 −1 Ry xz
B3g 1 −1 −1 1 1 −1 −1 1 Rx yz
Au 1 1 1 1 −1 −1 −1 −1 xyz
B1u 1 1 −1 −1 −1 −1 1 1 z zx 2 , z3 , y 2 z
B2u 1 −1 1 −1 −1 1 −1 1 y yz2 , y 3 , x 2 y
B3u 1 −1 −1 1 −1 1 1 −1 x xz2 , x 3 , y 2 x
Oh
x
An example with significantly more variety is the point group D2h whose char-
acter table is shown in Table 4.4. Here eight symmetry operations are relevant, in
addition to rotation around the C2 (z) axis, σv (xz), and σv (yz) reflections, there
are two more C2 axes (x and y), one more reflection σv (xy) as well as inversion ı
(leading to additional subscripts g and u). The symmetry properties of the states (or-
bitals, normal modes) are now characterized as shown in the first column: A states
are again always symmetric in respect of rotation (to all rotations as it turns out),
and a 3 Ag state would stand for a totally symmetric triplet state. In contrast, the
different symmetries of the B states in respect of rotation are now characterized by
subscripts 1, 2, and 3. The symmetry properties of basis vectors, rotations, quadratic
and cubic combinations is here indicated by writing the respective expressions into
the last three columns. We see now e.g. that inversion changes the sign of the polar
basis vectors (x, y, z) but not that of axial vectors, i.e. rotations (Rx , Ry , Rz ).
As a last example of character tables we shall now discuss the full octahedral
group Oh , a somewhat more complex and general case. It is relevant not only for
molecular physics but also describes the structure of mixed crystals with cubic lat-
tice. Its geometry is sketched in Fig. 4.19: consider a cation positioned in the centre
of an octahedron, its ligands located at the 8 vertices, i.e. in the centres of the cube
4.3 Symmetries 261
And they continue: “Thus if we know of a polyatomic molecule that the nuclei
in the equilibrium configuration do not all lie on a straight line, then we know at
the same time that its ground electronic state does not possess orbital degeneracy.”
Later reformulations of these original statements such as nonlinear molecules with
degenerate electronic states will undergo distortion and assume a lower symmetry
and energy, thus removing the degeneracy are somewhat misleading from present
perspectives. The key point is, that an electronic degeneracy of potential surfaces
determined in the framework of the B ORN -O PPENHEIMER approximation leads to
a coupling between nuclear and electronic coordinates (so called vibronic coupling)
which calls for a special treatment. The literature on this subject is quite extensive
(see e.g. B ERSUKER 2001, and references given there), and we can give here only a
brief introduction.
4.3 Symmetries 263
The JTE is of rather fundamental importance, especially as it bears out the limits
of the BO approximation quite clearly. Already in 1934 had R. R ENNER investi-
gated the situation for linear, triatomic molecules and found, that degenerate elec-
tronic states split when the bending vibration is excited: today this is called the
R ENNER -T ELLER effect. We have already seen a related behaviour for diatomic
molecules, called there avoided crossing of two potential curves with equal symme-
try. However, if the molecule has more than one relevant nuclear degree of freedom
such crossings of reduced dimensionality are possible. For example, in the case of
two degrees of vibrational freedom the corresponding potential surfaces may cross
in one point. This leads to a so called conical intersection, which we shall meet
later again. Conical intersections represent, however, not potential minima of the
adiabatic potential surfaces. Rather, they are instable in respect of the nuclear co-
ordinates as stated by the JT theorem. Such vibronic couplings play a role also if
the electronic states are not degenerate but lie very close together. This situation is
called pseudo-JAHN -T ELLER effect (PJTE). All three effects are characterized by
vibronic couplings which require an extension of the usual BO approximation.
The JTE is best explained by way of example. Molecular complexes with equi-
lateral sided octahedral geometry Oh (see Sect. 4.3.3), such as compounds of metal
atoms with several d electrons, may be subject to stretching or compression. Let
us discuss the doubly charged copper cation (Cu2+ , electron configuration 3d 9 )
which forms complexes of the Cu(H2 O)2+ 6 type. We first recall the angular de-
pendence of the d orbitals (Sect. 2.5.3 in Vol. 1). In atomic physics one nor-
mally uses the complex representation of these orbitals (eigenfunctions of orbital
angular momentum with quantum numbers L and M). More appropriate for the
present situation are real orbitals, i.e. linear combinations (Table D.2, Vol. 1) des-
ignated by dz2 , dxz , dyz , dx 2 −y 2 , dxy which have good quantum numbers L = 2 and
|M| = 0, 1, 1, 2, 2, respectively:
t2g :
√ √ √
3 3 3
dxz = 2 xz dxy = xy dyz = 2 yz (4.38)
r r2 r
√
1 3
dz2 = 2 3z2 − r 2 dx 2 −y 2 = 2 x2 − y2 .
2r 2r
Alternatively and symmetry adapted to the octahedral case, one may combine the
latter two as:
√ √
eg : (2dz2 + dx 2 −y 2 )/ 5 and (2dz2 − dx 2 −y 2 )/ 5. (4.39)
As a result we have three t2g and two eg orbitals, as sketched in Fig. 4.20.
Electrons in eg orbitals are thus clover shaped and aligned towards the 6 ligands
at the vertices of the octahedron, while the t2g orbitals (also clover shaped) point
towards the 8 edges, i.e. exactly in between the ligands. If one assumes that the lig-
ands are anions, one expects a stronger repulsion for the eg electrons in comparison
to the t2g electrons. Hence, the five d levels, which are degenerate in isotropic O(3)
264 4 Polyatomic Molecules
x x x
space, are now replaced by two higher lying eg and tree lower lying t2g levels as il-
lustrated in Fig. 4.21 on the left. One may easily imagine how these atomic orbitals
form bonding MOs – e.g. with the p electrons of the O atoms from the six water
ligands in the Cu(H2 O)2+6 complex.
Figure 4.21 illustrates in addition how in the special case of a Cu2+ cation in
the centre, the nine 3d electrons are to be distributed to the orbitals. For the higher,
twofold degenerate eg orbital only three more electrons are available, leading to an
electron hole. The three electrons may thus arrange their spin orientation in two en-
ergetically completely equivalent ways, ↑↓ + ↑ and ↑ + ↑↓, respectively. Accord-
ing to perturbation theory such states split in the presence of a suitable interaction
potential (here provided by the molecular environment). In the present case the Oh
4.3 Symmetries 265
Section summary
• Symmetries play an important role in classifying and characterizing the elec-
tronic and geometric structure of polyatomic molecules. The point groups in-
troduced here are based on several elementary symmetry operations, i.e. in-
version, rotations, reflections, and combinations thereof. Classes of symmetry
operations include powers of the relevant rotations.
• The number of these classes is identical to the number of different irreducible
representations in each symmetry group. Essentially, they correspond to the
angular momentum quantum numbers in atomic physics and characterize the
symmetry of electronic orbitals, states, and vibrational modes.
• The general rules for denomination of these irreducible representations is ex-
plained in Table 4.2, and their properties are summarized in the character table
of each group, some of which have been communicated and discussed above.
For non-degenerate states, the characters directly reflect their behaviour if the
symmetry operations act on them.
266 4 Polyatomic Molecules
• The JAHN -T ELLER (JT) theorem states that all nonlinear nuclear configu-
rations are unstable for an orbitally degenerate electronic state. In such a
situation, the molecule assumes a lower symmetry, thus avoiding the degener-
acy. This is called JT effect. In modern language it refers to vibronic coupling
which cannot be fully treated in BO approximation. As an example we have
discussed compounds of Cu2+ which has an electron configuration 3d 9 .
For diatomic molecules we have learned that molecular bonding occurs if the atomic
orbitals of the atoms involved have sufficient overlap, i.e. if there is a significant
amount of electron density in the region between the nuclei. This condition also
holds for polyatomic molecules and determines their geometry. We may thus for-
mulate as general
We start with this extremely important molecule from which a lot of fundamental
aspects about polyatomic molecules can be learned – even though H2 O may appear
a rather simple example at first sight. Rotational and vibrational spectra of H2 O
have already been introduced in Sects. 4.1.4 and 4.2.4, and were found to be quite
complex. Consequently, in condensed phase, water is characterized by 68 anomalies
of its physical properties – ultimately a consequence of its molecular structure and
dynamics. For a comprehensive collection of facts on water in all its phases we
refer to the well kept web-page of C HAPLIN (2013). A quote from it may stimulate
further interest: “Water is the main absorber of the sunlight in the atmosphere. The
13 million tons of water in the atmosphere (∼0.33 % by weight) are responsible for
about 70 % of all atmospheric absorption of radiation, mainly in the infrared region
where water shows strong absorption. It contributes significantly to the greenhouse
effect ensuring a warm habitable planet, but operates a negative feedback effect,
due to cloud formation reflecting the sunlight away, to attenuate global warming.”
The greenhouse effect, obviously, does not just have negative consequences as one
might think when reading the daily newspapers. The issue is simply to maintain the
optimal balance which has been reached by nature over many millions of years, and
which should not be disturbed by mankind in a serious manner!
4.4 Electronic States of Some Polyatomic Molecules 267
that the two unpaired electrons in orthogonal orbitals are the key to the bonding
and structure of the water molecule, each forming an MO with a 1s electron of one
of the H atom. A first guess of the bond angle would thus be 90◦ , reflecting the
perpendicular nature of the real as px , py and pz AO. This has to be compared with
an experimentally observed H−O−H angle of 104.474◦ . Indeed, even though these
unpaired p orbitals are essential in forming the OH bonds of water, the reality is
somewhat more complex.
The correlation diagram between the MOs of the H2 O molecule (C2v symmetry)
and the AOs of its constituents O and 2H is shown in Fig. 4.22. The H atom has
an IP (= −electron binding energy) of WI = 13.606 eV, the first IP of O for the
process O2s 2 2p 4 3 P → O2s 2 2p 3 4 S + e− is almost identical (see K RAMIDA et al.
2013), WI = 13.62 eV. Obviously it corresponds to the removal of one of the paired
electrons in 2p AO. This “lone pair” of electrons (see also Sect. 3.7.4) does not
contribute to the bonding. In C2v symmetry (Table 4.3) the respective MO it is
denoted as 1b1 . As indicated in Fig. 4.22 on the right, it reflects the shape of a px
AO (with reference to the coordinate system introduced in Fig. 4.18). It is filled with
two electrons of opposite spin and is the HOMO of neutral H2 O in its ground state.
The second IP of O allows an estimate of the 2py and 2pz energies and the excitation
energy O+ 2s 2 2p 3 → 2s2p 4 positions the 2s orbital energy. The construction of the
3a1 , 1b2 and 2a1 MOs is evident from Fig. 4.22. They are all characterized by charge
density along the OH bonds, and hence, all are bonding. In contrast, the closed 1s
shell electrons (i.e. the 1a1 MO) do not participate notably to the chemical bonding.9
It is interesting to note that electron binding energies in liquid water are essentially
the same (their absolute value just being 1.5 to 2 eV smaller due to the solvation
shell formed around each molecule (see e.g. W INTER et al. 2004, 2007).
9 The electron binding energies of the water MOs given in Fig. 4.22 are derived from gas phase pho-
toionization energies (see e.g. W INTER et al. 2004): WI = 539.9 eV (1a1 ), 32.6 eV (2a1 ), 18.8 eV
(1b2 ), 14.8 eV (3a1 ), and 12.6 eV (1b1 ). According to KOOPMAN’s theorem (see Sect. 10.2.4 in
Vol. 1) these values should correspond to the orbital energies W = −WI . The energies of the un-
occupied MOs are taken from C HAPLIN (2013) 6 eV (4a1 ), 8 eV (2b2 ) and 28 eV (3b2 ) (RHF
approximation).
268 4 Polyatomic Molecules
W / eV z
H 2O
y
3b2
20 y O
unoccupied orbitals
H H
2b2 y
LUMO
4a1
0
→ O+ 4S)
- WI (O 3P
HOMO x O
O 2px 1b1 H 1s H
H 1s
3a1 y
- 20 O 2py O 2pz 1b2
O 2s
y
2a1
≈
occupied orbitals y
- 530 O 1s
1a1 y O
- 540
H H
Fig. 4.22 Lowest MOs of the water molecule (C2v geometry); left: correlation with AOs for O and
H and electron binding energies; right: schematic 3D contour plots of the these MOs derived from
C HAPLIN (2013) and other web-sources. Note: for the 1b1 orbital the paper plane corresponds the
xz molecular plane, for all others to the yz plane
W I = 12.6 eV ~
D'' 1A 2 ~ 1 ~ ~
D A1 E' 1A' D'' 1A''
~ ~ ~1 ~
12 F 1A1 C 1B1 F A' C 1A''
~ ~ ~
F E' 1B2 D 1A'
11 ~ ~ ~
D D'' E'
10 2 1A1
~ ~
C 2 1A'
energiy / eV
B ~
9 ~1 B 1A''
B A2
8 ~ ~
A A 1A''
7
~
A 1 B1 asymmetric (Cs)
6
symmetric (C2v)
5
0 1 2 3 2 3 4 2 3 4 5
σ / 10 -17cm-1 R(O-H1,2) / a 0 R(O-H1) / a0
Fig. 4.23 Left: photo-absorption spectrum of the H2 O molecule. Here the photon energy is given
as ordinate, the absorption cross section σ as abscissa, in order to directly compare with potential
energy diagrams. Middle: cut through the potential surface in C2v symmetry (both OH bonds are
stretched). Right: Cs symmetry (only one OH bond is stretched). The potentials are sketched after
the computation of VAN H ARREVELT and VAN H EMERT (2000). However, we have scaled the data
published there such that the energetically observed spectra match the potentials
the case of diatomic molecules – just a spectroscopic abbreviation for the ground
state (the tilde is set to distinguish the states from diatomic molecules).
To understand the electronic structure of such polyatomic molecules, especially
the excited states, one has to be aware that the electronic energies Wγ (R) now
depend on more than one distance parameter R. Potential curves are replaced by
Potential hypersurfaces on which the nuclear motion proceeds. Usually one illus-
trates these potential surfaces by cuts along the relevant coordinates or – alterna-
tively – as contour plots in two dimensions. For the free water molecule we show
in Fig. 4.23 and absorptions spectrum in the UV and VUV spectral range, from
the X 1 A1 ground state to the excited states A, B etc. On the right of these spectra,
and directly comparable, Fig. 4.23 shows two examples of cuts through the H2 O
hyper-surface, demonstrating the complexity of the problem: in both cases, the po-
tential is shown along the ROH coordinate. In the middle panel the C2v symmetry is
conserved, i.e. both OH bonds are stretched symmetrically. In contrast, in the right
panel only one bond is stretched and the symmetry is broken: what remains is only
reflection symmetry σ̂v (zy) in respect of the zy plane.
We learn again a little piece of group theory: the point group Cs has only two
irreducible elements: A if the sign under reflection in respect of the molecular
plane is conserved, and A if it changes. A comparison of the C2v and Cs char-
acter Tables 4.3 shows, that A1 and B2 become A due to the symmetry breaking,
while A2 and B1 become A .
We see the dramatic difference between symmetric and antisymmetric stretch:
while in the first case the A 1 B1 state appears to be bound, the second geometry
270 4 Polyatomic Molecules
reveals that in actual fact it may dissociate very rapidly since the in Cs symmetry
1 A is strongly repulsive. This is reflected directly in the absorption spectrum
A
(left panel in Fig. 4.23): the absorption bands for the transition from X to A
state
is broad and unstructured. Because of the short lifetime of the excited A state the
energy is not sharply defined and vibrations cannot be associated with the spectra.
In contrast, of the higher excited states some are obviously stable when deformed
(they have a well defined energy minimum). Consequently the absorption spectrum
is increasingly more structured. Due to the large number of closely spaced rotational
and vibrational spectra, however, as energy increases one observes again only more
or less broad bands as energy increases.
2s
y
2py
x 2px
expected bonding angle between the p orbitals would be 90◦ – and not 109.47◦ as
observed experimentally in the case of an ideal tetrahedron such as methane.
Mathematically, one linearly superposes the familiar 2s and 2p AOs:
1
|1 = √ |2s − |2px + |2py + |2pz
4
1
|2 = √ |2s − |2px − |2py − |2pz
4
(4.40)
1
|3 = √ |2s + |2px − |2py + |2pz
4
1
|4 = √ |2s + |2px + |2py − |2pz .
4
These four new, hybrid sp 3 AOs are illustrated in Fig. 4.26. They are fully equiva-
lent: in each case all four original atomic orbitals contribute with equal weight. One
readily verifies that they are orthonormal if their constituents are orthonormal. √
The p components of the sp 3 hybrids may be written |j p = |j − |2s/ 4
with j = 1, 2, 3, 4; they determine the directions into which the orbitals are aligned.
As the thus defined |j p are real, they may be viewed as vectors pointing into the
direction of the bond. The angle θ between two of them, say between |1p and |2p ,
can thus be derived from the scalar product by
1|2p 1
cos θ = =− =⇒ |θ | = 109.47◦ . (4.41)
1|1p 2|2p 3
272 4 Polyatomic Molecules
4 3 z
x
y
109.5o
109.5o
(
10 If
one would simply add the charge densities of the four hybrid orbitals ( |ψj |2 ) the result
would be a fully spherical charge distribution.
4.4 Electronic States of Some Polyatomic Molecules 273
Table 4.6 Character table of the Td point group to which e.g. methane belongs
Td E 8C3 3C2 6σd 6S4
A1 1 1 1 1 1 x 2 + y 2 + z2 xyz
A2 1 1 1 −1 −1
E 2 −1 2 0 0 (x 2 − y 2 ), 3z2 − r 2
T1 3 0 −1 −1 1 (Rx , Ry , Rz ) [x(z2 − y 2 ), y(z2 − x 2 ),
z(x 2 − y 2 )]
T2 3 0 −1 1 −1 (x, y, z) (xy, yz, zx) (x 3 , y 3 , z3 ); [x(z2 + y 2 ),
y(z2 + x 2 ), z(x 2 + y 2 )]
σ Bonding
To describe the simplest hydrocarbon molecule methane, CH4 , one has to combine
each of the four sp 3 hybrid AOs of the C atom with a 1s AO from one of the four H
atoms. Thus, one constructs the following four MOs:
|σ1 = C a|1s1 + b|1
|σ2 = C a|1s2 + b|2
(4.42)
|σ3 = C a|1s3 + b|3
|σ4 = C a|1s4 + b|4 .
√
Here |1sj is the 1s orbital of the j th H atom and C = 1/ a 2 + 2abS + b2 is a
normalization constant with the overlap integral S (see Sect. 3.5). The shape of the
resulting MOs (4.42) is quite similar to σ orbitals in diatomic molecules. Hence, this
type of bonding is called σ bond. Again, these σ MOs may each be filled with two
electrons. In the present case the four σ bonds are filled with four electrons from
the C atom and with four electrons from the four H atoms. We obtain in this way
the tetrahedrally shaped methane molecule as sketched schematically in Fig. 4.28.
Note that the “sausage” type schemes of orbitals presented in Fig. 4.28 are typically
used in chemistry. They are rather useful to emphasize the bond characters, but do
not reflect the probability densities of electrons in a real molecule. As shown for
methane in Fig. 4.29 the electron charge distribution is typically much smoother
and indicates the directionality of the bonds only vaguely (if at all).
For reference, in Table 4.6 we also communicate the character table of the Td
point group to which methane belongs. As shown, it has five classes of symmetry
operations and consequently five different irreducible representations, two of which
(A1 and A2 ) are singly degenerate, one (E) is twofold and two (T2 and T3 ) are
threefold degenerate. The ground state, illustrated in Figs. 4.28–4.29, is a totally
symmetric A1 state.
Sigma bonding also plays the key role in other organic molecules, wherever sin-
gle bonds are involved – as illustrated in Fig. 4.30 for the example of C2 H6 (ethane).
One may visualize the carbon atoms as being kept together by the overlap of two
hybrid sp 3 wave functions (AOs). They jointly form σ MOs – again sketched in a
very schematic, simplified picture as “sausage” type symbols.
274 4 Polyatomic Molecules
H H
N
H
H
106.7o
H
106.7o
Fig. 4.31 Geometry and schematic of molecular orbitals for the electronic ground state of NH3 .
One recognizes the origin of the three bonding molecular orbitals from the hybridized 2spx,y,z
orbitals of N and one 1s electron of atomic H. The orbital pointing upwards does not participate in
the bonding and is occupied by a lone pair of electrons (indicated by two black dots)
4.4 Electronic States of Some Polyatomic Molecules 275
z
(a) ( )
12
sp
90 0°
2 ı3
°
sp2ı3 y
x x
sp2ı1
sp2ı1
sp2ı2 pʌ pʌ
2
ı
sp 2
y
Fig. 4.32 Schematic of the three carbon sp 2 hybrid σ AOs defined by (4.43) and the (hatched) pz
(or pπ ) orbital, forming a double bond; (a) side view, (b) view from below
less than the ideal tetrahedral angle of ca. 109.5◦ , obviously as a consequence of
electron repulsion by the lone pair. A schematic representation of the MOs involved
in NH3 is shown Fig. 4.31. Ammonia belongs to the C3v symmetry group with the
1
elements A1 , A2 and E, the latter being two fold degenerate. The ground state, XA
sketched in Fig. 4.31 has again total symmetry.
Planar molecules with a C=C double bond are explained by sp 2 hybrid orbitals
which are formed by the 2s and two of the 2p AOs. Let the x-axis point into the
direction of the double bond, the z-axis perpendicular to it. Then the three sp 2 AOs
are:
2 √
sp σ1 = (1/ 3)|2s + 2/3|2px
2 √ √
sp σ2 = (1/ 3)|2s − 1/6|2px + (1/ 2)|2py (4.43)
2 √ √
sp σ3 = (1/ 3)|2s − 1/6|2px − (1/ 2)|2py .
These sp 2 AOs too, have σ character. The bond angle between them is derived as
in the case of sp 2 σ AOs. In analogy to (4.41) one obtains from the p components
of the orbitals |sp 2 σ1 and |sp 2 σ2 for cos θ = −1/2, i.e. a bond angle |θ | = 120◦ .
The three |sp 2 σj orbitals thus span an equilateral triangle in the xy plane. The
fourth orbital is perpendicular to that plane and parallel to the z-axis as illustrated
in Fig. 4.32. It is called π orbital and has negative reflection symmetry in respect of
the xy plane.
In C2 H4 (ethene or ethylene) the three sp 2 hybrid orbitals of each carbon atom
lead to three σ bonds (one for each of the two H atoms and one for the other C
atom) as illustrated in Fig. 4.33 – again very schematically. The remaining two pπ
electrons (one from each C atom) combine to an additional bond with an electron
charge distribution perpendicular to the σ bond and the xy plane. Considering the
276 4 Polyatomic Molecules
σ bonds
similarity in the corresponding diatomic case, see e.g. Fig. 3.34(a), one speaks of
a π bond. The geometry of these orbitals sketched in Fig. 4.32 suggests that the
double bond leads to D2h symmetry (Table 4.4) of the molecule. All four H atoms
lie in the xy plane. In a serious MO calculation, five σ type valence orbitals (with
ag , b3u , b2u , ag , b1g symmetry) are constructed from the 2s, 2px , 2py carbon AOs
and the four 1s hydrogen orbitals, while the (bonding) π HOMO is derived from
the remaining 2pz carbon AOs and has b1u symmetry.
Finally, a third type of hybridization is possible with carbon orbitals: the sp hybrid
which is needed for all triple bonds. The most simple example is C2 H2 (ethyne or
acetylene), a linear molecule with D∞h symmetry. The sp hybrid orbital is formed
by from the 2s and the 2px AOs of carbon:
√
|spσ1 = |2s + |2px / 2
√ (4.44)
|spσ2 = |2s − |2px / 2.
The two orbitals point into +z- or −z-direction, respectively. Together with corre-
sponding AOs of neighbouring atoms one bonding σ orbital is available per neigh-
bour (as well as an antibonding σ ∗ orbital contributing to higher excited states).
The two remaining electrons in the 2px and 2py orbitals form – jointly with a
neighbouring C atom – two additional π bonds. In summary, a triple bond is formed
between two carbon atoms, as sketched in Fig. 4.34 for the example acetylene.
The σ π double and the σ π 2 triple bonds are characterized by a certain rigidity
which makes a rotation around the C=C or C≡C axis difficult. In contrast, the
barrier for rotation around a σ single bond is low. This difference is important for
understanding many properties of organic molecules, e.g. in the context of cis-trans
isomerization.
Section summary
• To describe the electronic structure of polyatomic molecules, symmetry
adapted orbitals are required, from which the total state of the molecule is
constructed. We have discussed some important examples.
• Water belongs to C2v symmetry and has a bond angle of ca. 104.5◦ in its
(overall totally symmetric) ground state 1 A1 . Its electron configuration is
(1a1 )2 (2a1 )2 (1b2 )2 (3a1 )2 (1b1 )2 , of which the valence MOs (2a1 )2 (1b2 )2
(3a1 )2 are bonding while the (1b1 )2 is filled with a lone pair of electrons,
which does not contribute to the bond.
• sp 3 , sp 2 , and sp hybridization of AOs is an important concept. Specifically
for C atomic orbitals it forms the basis for carbon chemistry. The sp 3 AOs
(Sect. 4.4.2) lead to tetrahedrally bonded structures, as in the case of methane
CH4 (with an ideal tetrahedral bond angle of 109.47◦ ), and essentially also in
all linear hydrocarbon chains.
• The structure of NH3 (Sect. 4.4.3) is also very close to tetrahedral, with a bond
angle of 106.7◦ (slightly less than ideal due to one lone pair of electrons in
the valence shell).
• On the other hand, sp 2 hybrid AOs are the basis for all double bonds in organic
compounds, leading to a bond structure with three symmetric σ bonds in a
plane (bonding angle 120◦ , Sect. 4.4.4) and one perpendicular π bond.
• Finally, sp hybrids (Sect. 4.4.5) provide the concept for triple bonds in or-
ganic chemistry: they consist, per carbon atom, of two σ bonds (in opposite
direction) and two perpendicular π bonds.
The class of conjugated organic molecules consists of a chain of carbon atoms which
are bound together by σ bonds (lying in one plane) based on sp 2 hybrid AOs and π
bonds from the 2p orbital perpendicular to the former (in the following denoted by
pz ). Examples are butadien, C4 H6 , or benzene, C6 H6 , a cyclic molecule. The basic
structure of these molecules is determined by the properties of the sp 2 hybrid or-
bitals (such as the bonding angle of 120◦ , see Fig. 4.32). The additional π bonds are
278 4 Polyatomic Molecules
as shown here: .
Figure 4.35 gives an image of a still rather naive LCAO approach: molecular
orbitals are seen (a) as an addition of the carbon σ 2sp 2 hybrid orbitals with the
hydrogen 1s orbitals, and (b) as addition of all π2pz orbitals. Strictly speaking,
each of the thus constructed MOs can only be filled with two electrons (assuming
all to be spin saturated). However, benzene has 4 × 6 + 6 = 30 valence orbitals
(ignoring the 1s 2 carbon core electrons) which have to be filled into suitable MOs.
To classify these, we recall that benzene belongs to the symmetry group D6h . The
main C6 symmetry axis z is perpendicular to the plane of the ring. Some of the
C2 and C2 axes are also indicated in Fig. 4.35. As documented in the character
Table 4.7, the D6h group has 12 symmetry classes (including the unit operator) and
hence 12 irreducible characters (types of states or orbitals).
The σ orbitals shown in Fig. 4.35(a) together form a totally symmetric 2a1g wave
function. And the π electron configuration in benzene shown in Fig. 4.35(b) has a2u
character: no sign changes for rotation around the symmetry axes (C6 , C3 , C2 ), nei-
ther upon reflections through the planes containing the principle z-axis (σ̂d , σ̂v ); but
all operations which somehow interchange top and bottom of the molecule change
4.5 Conjugated Molecules and the H ÜCKEL Method 279
Table 4.7 Character table for the point group D6h and basis functions
D6h E 2C6 2C3 C2 3C2 3C2 ı 2S3 2S6 σ̂h 3σ̂d 3σ̂v
A1g 1 1 1 1 1 1 1 1 1 1 1 1 x 2 + y 2 , z2
A2g 1 1 1 1 −1 −1 1 1 1 1 −1 −1 Rz
B1g 1 −1 1 −1 1 −1 1 −1 1 −1 1 −1
B2g 1 −1 1 −1 −1 1 1 −1 1 −1 −1 1
E1g 2 1 −1 −2 0 0 2 1 −1 −2 0 0 (Rx , Ry ) (yz, zx)
E2g 2 −1 −1 2 0 0 2 −1 −1 2 0 0 (x 2 − y 2 , xy)
A1u 1 1 1 1 1 1 −1 −1 −1 −1 −1 −1 z
A2u 1 1 1 1 −1 −1 −1 −1 −1 −1 1 1
B1u 1 −1 1 −1 1 −1 −1 1 −1 1 −1 1
B2u 1 −1 1 −1 −1 1 −1 1 −1 1 1 −1
E1u 2 1 −1 −2 0 0 −2 −1 1 2 0 0 (x, y)
E2u 2 −1 −1 2 0 0 −2 1 1 −2 0 0
the sign of the wave function. As we shall see in a moment, this orbital corresponds
to the energetically lowest π level in benzene.
To compute the binding energies and spatial distributions of the 2pz electrons in
conjugated molecules one often uses as a first approximation a method developed
already in 1931 by H ÜCKEL. In the end it is simply an application of the variational
method to obtain optimal LCAO orbitals as already introduced in the context of H+ 2
(see Sect. 3.5.1).
We illustrate the H ÜCKEL method by way of example for benzene and concen-
trate on the π orbitals of the six carbon atoms. Albeit very simple, this approxima-
tion leads to quite reasonable results for an estimate of the electronic structure and
energies. As we have just discussed, the symmetry of the σ bonds in this planar
molecule is different from that of the π bonds. One may thus separate their Hamil-
tonians (they have no joint non-diagonal elements) and derive the solutions for both
types of orbitals independently. The σ bonds are substantially stronger than the π
bonds. As it turns out, the energies of the π and π ∗ orbitals fall (essentially) into
the band gap between σ and σ ∗ orbitals. Thus, it is mainly the π and π ∗ orbitals
which determine the spectroscopic behaviour of conjugated molecules in respect of
absorption and fluorescence in the VIS and UV spectral region.
The approximations of the H ÜCKEL method may be summarized as follows:
Here |pj stands for the 2pz orbital at C atom j . The values of α (C OULOMB inte-
gral) and β (resonance integral) are both negative and may be treated as parameters
280 4 Polyatomic Molecules
Rj
x
Since in benzene all C atoms are equivalent, |cj |2 must be identical for all j . The
most simple procedure is now to write down the H AMILTON matrix and to diago-
nalize it according to the standard rules. Applying (3.102) to the 6 carbon atoms,
the problem is described now by a 6 × 6 determinant for which we have to find the
roots. With the rules 1.–3. just specified we obtain:
α − ε β 0 0 0 β
β α−ε β 0 0 0
0 β α − ε β 0 0
0 0 β α − ε β 0 = 0. (4.46)
0 0 0 β α−ε β
β 0 0 0 β α −ε
One readily verifies that the solutions are:
Two of these energy levels (ε±1 and ε±2 ) are doubly degenerate.
Alternatively one may derive a general solution for ring molecules with n carbon
atoms and identical conjugated bonds from solving the system of linear equations
(3.101) by exploiting symmetry properties of the system. As a starting point one
defines a wave function Φ(r) for a carbon 2pz orbital, which is localized in the
centre of the benzene ring as sketched in Fig. 4.36. The six atomic orbitals Φj are
then given by
Φj (r) = Φ(r − R j )
now be the effective Hamil-
where R j is the position vector of the C atom j . Let H
tonian for the π electrons. For symmetry reasons it is invariant in respect of rotation
of the ring through 360◦ /n. Mathematically this implies that H and C n commute:
, C
[H n ] = 0.
and C
Thus, H n have joint eigenfunctions. The eigenvalues of H (i.e. the energies ε)
may thus be derived from the eigenvalues of C n . Without going into details of that
consideration we obtain the solution (4.47) for the n = 6 eigenvalues, written in the
4.5 Conjugated Molecules and the H ÜCKEL Method 281
εk
k =3 α - 2β
1b2g
k =- 2 k =2
LUMO α + 2 β cos(4π / 6) = α - β
1e2u 1e2u
α
k =- 1 k =1
HOMO α + 2 β cos(2π / 6) = α + β
1e1g 1e1g
k=0
α + 2β
1a2u
-3 -2 -1 0 1 2 3 k
Fig. 4.37 Orbital energies of the π electrons in benzene according to the H ÜCKEL model. The
red dotted curve represents (4.48), the red dots give the orbital energies for k = −2 to +3, corre-
sponding to the energy terms (full, horizontal red lines). The three lowest orbital are each occupied
with two π electrons of different spin orientation (black arrows)
This is illustrated in Fig. 4.37. These molecular orbitals are filled with the six π elec-
trons in the benzene ring – according to the PAULI principle – indicated in Fig. 4.37
by double arrows. The doubly degenerate 1e1g MOs (k = ±1) are obviously the
HOMOs. If we fill instead higher lying levels, we obtain excited states of benzene.
We note that the total contribution of the π orbitals to the ground state energy may
be estimated from the occupied orbitals as
εloc = 6α + 6β.
localized bonds we had (3.108) εg = (HAA + HAB )/(1 + S). Identifying HAA = α and
11 For
HAB = β gives for one atom εg = α + β when the overlap integral S is ignored.
282 4 Polyatomic Molecules
+ -
b2g
- +
α-2β
+ -
- + e2u
+ -
- α-β
+
- +
-
- e1g
+
+ α+β
a2u
α+2β
Fig. 4.38 MO orbitals of the 6 benzene π valence orbitals; left: symmetry properties of the or-
bitals (D6h , Table 4.7); middle, black the designation of the orbitals follow the irreducible repre-
sentations of D6h and middle, red their Hückel binding energies corresponding to (4.47); right the
3D contour plots of the MO orbitals have been gleaned schematically from I MMEL (2012) and
represent the volume within which 90 % of the electron charge is confined
from which, trivially, cj +6 = cj emerges. The solutions of the coefficients {cj } are
thus given by
(k) 2 (k) 2
for which cj = c0
(k) 2πi (k)
cj = e 6 k·j c0 ,
holds, i.e. for all MOs |k the probability for finding an electron at a particular atom
j is independent of j : the π electrons are symmetrically delocalized. The symmetry
adapted wave functions of a single electron thus reads
6
2πi
φk (r) = cj(k) Φ(r − R j ) = c0(k) e 6 k j
Φ(r − R j ), (4.49)
j j =1
√
for k = 0, ±1, ±2, 3, with the normalizing constant c0(k) = 1/ 6. And explicitly, the
eigenfunctions (4.49) are given by
φ3 (1b2g ) ∝ −Φ1 + Φ2 − Φ3 + Φ4 − Φ5 + Φ6
2π 4π 8π 10π
φ±2 (1e2u ) ∝ e± 3 i
Φ1 + e ± 3 i
Φ2 + e±2πi Φ3 + e± 3 i
Φ4 + e ± 3 i
Φ5 + e±4πi Φ6
2π 4π 5π
φ±1 (1e1g ) ∝ e± 3 i Φ1 + e± Φ2 + e±πi Φ3 + e± Φ4 + e ± Φ5 + e±2πi Φ6
π
3 i 3 i 3 i
φ0 (1a2u ) ∝ Φ1 + Φ2 + Φ3 + Φ4 + Φ5 + Φ6 ,
ordered here according to their energy. In Fig. 4.38 on the left their symmetry prop-
erties are schematically sketched, corresponding to the irreducible representations
of the point group D6h (cf. Table 4.7). A three dimensional impression of these six
π orbitals is given in Fig. 4.38 on the right. Full 3D images of these orbitals may
be viewed on several rather instructive web-sites, e.g. NASH (2004), or somewhat
4.5 Conjugated Molecules and the H ÜCKEL Method 283
more detailed and universal I MMEL (2012), where these images can be viewed and
manipulated in 3D space with Chime and Jmol,12 respectively.
The energies of the σ orbitals cannot be derived by a similarly simple treatment.
We thus refer here only to the literature without going into much detail. Table 4.8
gives a summary of calculations and measurements on orbital symmetries and ener-
gies for all occupied MOs in benzene, together with a few of the excited ones. Most
of this data has been taken from relatively recent work by D ELEUZE et al. (2001).
They performed state-of-the-art ab initio calculations for ionization bands of several
aromatic hydrocarbons and compared them with experimental data.
The symmetries of the orbitals in the first column of Table 4.8 refer to the D6h
point group (see Table 4.7). The next column indicates whether the orbital refers to
a σ or π bond and the third column gives the occupation number of each orbital (all
e orbitals are doubly degenerate). The valence orbitals are completely filled – up to
the HOMO 1e1g orbital with spin saturated electrons.
12 Jmol is an open source Java viewer for molecular structures which appears now to be the gener-
MO energy εj / eV
HF Hückel
10
1b2g
1a2u
Fig. 4.39 Energy of molecular π type orbitals (left) and electronic states (right) in benzene. The
energies of the HF molecular orbitals (heavy black lines) are those given in Table 4.8; they are
compared with the (scaled) H ÜCKEL energies (thin red lines) according to (4.48). The state ener-
gies are the ab initio values of H ASHIMOTO et al. (1998) (experiment in brackets); singlets: 1 B2u
4.70 (4.90), 1 B1u 6.21 (6.20), 1 E1u 6.93 (6.94), 1 E2g 7.82 (7.80) eV, triplets: 3 B1u 3.89 (3.95),
3E 3 3
1u 4.53 (4.76), B2u 5.54 (5.60), E2g 7.02 (6.83) eV. The ionization energy of benzene in the
1 A ground state (gas phase) is W = 9.24378 ± 0.00007 eV (NIST 2011)
1g I
Since the π MOs are the most loosely bound orbitals they determine the ground
state and the lowest excited states. Hence, the ground state electron configuration is
described by (1a2u )2 (1e1g )4 1 A1g (as in the atomic case, in a closed shell system
the ground state is a singlet state).
To obtain the first excited states of benzene, one electron is raised from a 1e1g
HOMO into an empty 1e2u LUMO, leading to a (1a2u )2 (1e1g )3 (1e2u ) electron con-
figuration. Thus, we have to combine a 1e1g and a 1e2u orbital, each filled with an
unpaired electron, in such a manner that the PAULI principle is conserved and at
the same time a irreducible representation of the D6h point group emerges. Since
both orbitals are two-fold degenerate, there are four different possibilities to comply
with these conditions. In mathematical terms one writes these electronic states as
direct product E1g ⊗ E2u = B1u + B2u + E1u . We thus have three states: B1u , B2u
and the doubly degenerate E1u state. Singlet as well as triplet states can be formed.
According to H UND’s rules, the latter is always lower in energy than the respective
singlet state.
Figure 4.39 summarizes the energetics of benzene. On the left, the orbital ener-
gies are sketched. The Hückel energies have been scaled such that the energy differ-
ence between 1b2g and 1a2u orbital is the same as that for the ab initio calculated
HF values. Obviously, the Hückel method does rather well in predicting the relative
distances between the four π levels. The energies of the overall states have been
calculated by H ASHIMOTO et al. (1998) with multi-reference M ØLLER -P LESSET
perturbation theory (MRMP, see Sect. 10.1.1, Vol. 1) with complete active space
self-consistent field (CASSCF) reference – a rather rigorous ab initio approach. It
Acronyms and Terminology 285
compares extremely well with experiment (values given in the figure caption in
brackets). Note that direct optical transitions from the 1 A1g ground state are al-
lowed to the 1 E1u state. They are observed experimentally as a strong absorption
band while the other transitions are symmetry forbidden, but may be observed as
weak transitions.
As we have seen here, the H ÜCKEL method as presented above gives a valuable
first order estimate for energies and structures. It is very simple and flexible, and
gives reasonable estimates even for larger molecules and clusters, typically up to
the HOMO. One may even generalize what has been exemplified for benzene to
any kind of Nnu regularly ordered, identical atoms – and derive thus e.g. a one
dimensional solid state model. The eigenfunctions then become B LOCH functions
and instead of discrete eigenvalues one obtains energy bands. The method is used
quite often in solid state physics, usually called extended H ÜCKEL tight-binding
approach (the latter essentially being another name for LCAO).
Section summary
• Delocalized double bonds in organic molecules can be treated to a good ap-
proximation by the H ÜCKEL method. This particularly simple approach for
obtaining quantitative information on the structure and energy of bonding π
MOs uses an LCAO approach which parameterizes the energies in terms of
only two parameters. We have applied the method to benzene, C6 H6 , the sim-
plest aromatic ring. The spectroscopy of benzene is essentially determined by
the six valence π electrons and their orbitals. We have familiarized ourselves
with the dihedral symmetry group D6h to which benzene belongs.
AO: ‘Atomic orbital’, single electron wave function in an atom; typically the basis
for a rigorous structure calculation.
APES: ‘Adiabatic potential energy surfaces’, potential energy hyper-surface deter-
mined in B ORN -O PPENHEIMER approximation.
BO: ‘B ORN O PPENHEIMER’, approximation, the basis when solving the S CHRÖ -
DINGER equation for molecules (see Sect. 3.2).
E1: ‘Electric dipole’, transitions induced by the interaction of an electric dipole
with the electric field component of electromagnetic radiation.
FT: ‘F OURIER transform’, see Appendix I in Vol. 1.
good quantum number: ‘Quantum number for eigenvalues of such observables that
may be measured simultaneously with the H AMILTON operator (see Sect. 2.6.5
in Vol. 1)’.
HF: ‘H ARTREE -F OCK’, method (approximation) for solving a multi-electron
S CHRÖDINGER equation, including exchange interaction.
HITRAN: ‘High-resolution transmission molecular absorption database’, http://
www.cfa.harvard.edu/hitran (ROTHMAN et al. 2009).
286 4 Polyatomic Molecules
References
B ERNATH , P. F.: 2002a. ‘Laser chemistry – water vapor gets excited’. Science, 297, 943–944.
B ERNATH , P. F.: 2002b. ‘The spectroscopy of water vapour: Experiment, theory and applications’.
Phys. Chem. Chem. Phys., 4, 1501–1509.
B ERSUKER , I. B.: 2001. ‘Modern aspects of the Jahn-Teller effect theory and applications to
molecular problems’. Chem. Rev., 101, 1067–1114.
B UNKER , P. R. and P. J ENSEN: 2006. Molecular Symmetry and Spectroscopy. Ottawa: NRC Re-
search Press, 2nd edn., 747 pages.
C ARLEER , M. et al.: 1999. ‘The near infrared, visible, and near ultraviolet overtone spectrum of
water’. J. Chem. Phys., 111, 2444–2450.
C HAPLIN , M.: 2013. ‘Water structure and science’, London South Bank University. http://www.
lsbu.ac.uk/water/, accessed: 9 Jan 2014.
DAMBURG , R. J. and R. K. P ROPIN: 1972. ‘Rotational structure of the inversion spectrum of
ammonia’. J. Phys. B, At. Mol. Phys., 5, 1861–1867.
References 287
Overview
After a brief introduction in Sect. 5.1 we shall expand our knowledge about
rotational (microwave) and vibrational (infrared) spectroscopy in Sects. 5.2
and 5.3, respectively, and supplement it with short excursions into infrared
F OURIER transform spectroscopy (FTIR) and IR action spectroscopy. In
Sect. 5.4 we turn to the spectroscopy of electronic transitions (VIS, UV and
VUV) and present a few state-of-the-art methods of modern molecular spec-
troscopy. In Sect. 5.6 basics of R AMAN spectroscopy will be developed –
a very important spectroscopic art, which may be said to reside in between
electronic and vibrational spectroscopy. In Sect. 5.5.4 we illustrate the aston-
ishing capabilities of today’s high resolution spectroscopy with sophisticated
methods, as applied to larger, even biologically relevant molecules. Finally, in
Sect. 5.8 we introduce the important field of photoelectron spectroscopy.
5.1 Overview
Let us briefly recall some of the most important facts. In the framework of B ORN -
O PPENHEIMER approximation (i.e. almost always except for some notable excep-
tions), the total wave function Ψγ vN (r, R) of a molecular system may be written as
product of electronic and nuclear wave functions, φγ (r; R) and ψγ vN (R), respec-
tively:
Ψγ vN (r, R) = φγ (r; R) × ψγ vN (R).
Here r represents the entirety of all electronic, R of all nuclear coordinates. The
nuclear wave function may usually again be factorized into vibration and rotation.
In the case of a diatomic molecule one finds
Rγ vN (R)
ψγ vN (R) = × YN MN (Θ, Φ),
R
with R being the internuclear distance and Θ, Φ describing the polar and azimuthal
angle of the internuclear axis. The indices characterize the electronic state (γ ), the
nuclear vibration (v) and the angular momentum of nuclear rotation (N ). Accord-
ing to (3.49) and (3.50) the total energy of the system may be written as a sum of
electronic, rotational and vibrational energy:
γ
Wγ vN = Vγ R0 + Wv + WN . (5.1)
WvN Vγ(R )
higher
electronic rotational levels WN'
state V'(R )
v' = 3
lower vibrational
v' = 2 levels, Wv'
electronic
v' = 1
state
v' = 0
R 0'
V''(R )
rotational levels WN''
rot
V'(R 0' )- V ''(R''0 )
-
el-vib
vib-rot
v''
v'' = 3 vibrational
v'' = 2 levels, Wv''
rot
v'' = 1
v'' = 0
R''0 R
Fig. 5.1 Composing the total energy of a diatomic molecule from rotation, vibration, and elec-
tronic energy. The heavy black double arrows indicate three different types of molecular spectra:
electronic bands with rotational and vibrational transitions (el-vib-rot), vibration-rotation spectra
(vib-rot) and pure rotation spectra (rot); traditionally (according to H ERZBERG 1989) upper levels
are denoted by a single dash , lower levels by two dashes
Fig. 5.2 The spectrum of electromagnetic radiation seen from the perspective of molecular spec-
troscopy. MW: microwave, FIR: far infrared, MIR: middle and NIR: near infrared (IR), VIS visible,
UV: ultraviolet, VUV: vacuum ultraviolet, EUV: extreme ultraviolet, XUV: soft X-ray
gions, and hence, a broad variety of techniques and methods is applied for molecular
spectroscopy, mostly exploiting E1 type transitions. Very important exceptions, not
shown in Fig. 5.2, are M1 transitions as the basis of NMR and EPR spectroscopy
(which we have already introduced in the context of hyperfine structure in Chap. 9,
Vol. 1) and R AMAN spectroscopy (to be explained in Sect. 5.6).
In the following sections we shall present a selection of powerful spectroscopic
methods for the different spectral regions and categories of spectra. We shall illus-
trate these, as far as possible and instructive, with examples from recent literature.
292 5 Molecular Spectroscopy
Section summary
Molecular energy levels are essentially given by the sum of electronic, vibra-
tional and rotational energy. Correspondingly, spectroscopy in the different spec-
tral regions of electromagnetic radiation probes different aspects of molecular
structure:
• Radio frequencies (kHz to some 100 MHz) induce nuclear spin resonance
transitions. NMR spectroscopy exploits these with highly sophisticated meth-
ods for structural determination of large molecules.
• Frequencies from about 1–100 GHz (λ 30 cm –3 mm) are called microwaves
MW. Electron spin resonances EPR are induced by such frequencies. At least
equally important are applications to rotational transitions in larger molecule
as we have seen already in Chap. 3.
• The infrared spectral region extends from microwaves to the visible spectral
range at around 760 nm. In the far infrared (λ = 0.1 mm to 1 mm) rotational
spectra of many molecules are found. Vibration-rotation spectra are typically
associated with the centre of the IR range but extend often even down to the
far red (λ = 700 nm to 1.4 µm).
• The visible, ultraviolet and vacuum ultraviolet spectral regions are associated
with electronic transitions and their band spectra.
• Beyond the VUV region extend the XUV, X-ray and γ -ray regions. With high
photon energies, inner shell electrons are studied. Photoelectron spectroscopy
is probably the method most often applied (UPS: with VUV light; XPS: with
X-ray radiation). But also X-ray absorptions spectroscopy (XAS, XANES,
NEXAFS) provides useful information. As these methods address specific,
localized atoms (e.g. in large organic molecules) such methods are very selec-
tive in respect of the geometry of the molecule studied.
We may keep this short as the key aspects have already been discussed in Sect. 3.4.1.
Rotational transitions are studied almost exclusively in absorption, since the proba-
bility for spontaneous emission is extremely low, due to the low transition frequency
and the ω3 factor for the E INSTEIN Aab coefficients (see Eq. (2.154)). A typical
microwave spectrometer consists of a microwave source, a waveguide structure in
which the gaseous molecular target to be studied is positioned, and a detector. As
sources one uses magnetrons, reflection klystrons or travelling wave tubes. Long
absorption paths (metres) are needed since the absorption cross sections are small
and the target density has to be kept low (to avoid collisional line broadening). Al-
ternatively one uses high quality (Q) resonators and determines the damping as a
function of frequency – which is equivalent to a long absorption path. To improve
the signal to noise ratio one applies modulation techniques combined with phase
sensitive detection. When studying rotational lines, modulation may be achieved by
5.2 Microwave Spectroscopy 293
mixer
microphone. The MB carrier motor&
frequency ν is modulated in microphone
various mixers and again ν
mixer
demodulated in the detection 157.5
system. The signal is finally MHz
recorded by an analogue to
digital converter (A/D) and A/D
registered in a computer,
which is also used for MW com-
controlling the experiment source puter
signal (linear)
0+404-303A 8-6
0+404-303E H H
1011 - N
0 404-303A
1010 -
0 404-303E
H
109
H C H
108
107
Fig. 5.4 Parts of the microwave absorption spectrum for p toluidine according to H ELLWEG
(2008), obtained with an advanced version of the MW-MWFT shown in Fig. 5.3 (G RABOW et
al. 1996, operating in a frequency range from 3 GHz to 25 GHz). (a) Overview (b) blow up of the
rotational spectrum with the transitions 404 –303 of the methyl torsion mode A from the vinv = 0+
vibrational state. Shown are three hyperfine transitions 2F → 2F , 10–8, 6–4 and 8–6. The rect-
angular brackets mark the D OPPLER splitting of these lines
of-the-art of ab initio quantum chemical calculations – allows also for the determi-
nation of distortions and modifications of the equilibrium distances due to rotation,
vibrational excitation, inner rotation of sub groups in the molecule, and hyperfine in-
teraction. We illustrate this by a relatively recent measurement of H ELLWEG (2008)
who studied p toluidine (C7 NH9 ). As shown in Fig. 5.4, it consists of a benzene
ring with one amino and one methyl group. The rotational levels of this asymmetric
rotor are characterized by NKa Kc (as introduced in Sect. 4.1.4 the rotational quan-
tum number is N , its projection onto the principle axes of the ellipsoid of inertia
are denoted by Ka and Kc ). This particular molecule has two extra specialties: the
methyl group may rotate, leading to two series of A and E symmetry. In addition,
the two H atoms of the amino group, which point out of the plane defined by the
benzene ring, may undergo inversion vibrations. Figure 5.4(a) gives an overview of
the MW spectrum for a frequency range of about 60 MHz. As in the case of NH3
(see Sect. 4.2.5) the structure leads to two split vibrational ground states vinv = 0+
and 0− . Figure 5.4(a) documents that the latter is populated only weakly, owing to
the cooling in the supersonic jet beam. The impressive resolution is illustrated in
the blow up Fig. 5.4(b) for part of this spectrum over less than 1 MHz with fully re-
solved hyperfine structure: nuclear spin I (I = 1 for 14 N) and rotational momentum
N add to a total angular momentum F with quantum numbers F = N or N ± 1.
One recognizes in Fig. 5.4(b) that all lines appear in pairs due to D OPPLER split-
ting. Albeit very small in the microwave region due to ν = νv/c, the very high res-
olution allows one to resolve and assign it: note that each line is observed as clearly
split by ca. 100 kHz in contrast to the usual D OPPLER broadening. As the molecular
beam enters the resonator coaxially (see Fig. 5.3), the microwave excites different
velocity components of the beam depending on the MW travelling from left to right
5.2 Microwave Spectroscopy 295
0.9
(a) 111 72
Å
(d)
201
90.04o
/ GHz
Å
50 (c)
88
101
1.6
0 (b) 111.02o
1.
225
NK K = 000
a c Å
(b) (c)
F=0-1
F=1-1
18708.5 18709.5
/ MHz 80154 80158
Fig. 5.5 Microwave spectroscopy of the HO3 radical according to S UMA et al. (2005).
(a) Term scheme for rotational levels, (b) microwave spectrum for one rotational transition
(NKa Kc = 101 –000 , J = 1.5–0.5, F = 2–1), (c) double resonance (NKa Kc = 111 –000 , J = 0.5–0.5)
with hyperfine splitting due to the proton (F = 1–1 and F = 0–1), observed by reduction of the
maximum signal for transition (b) when tuning the mm wave through transition (c) as indicated in
the term scheme (a); (d) geometry of HOOO determined by this experiment
or after reflection from right to left. With the internal translational temperature of
the molecules in the supersonic beam being very low, both components are well
resolved and can be assigned as indicated in Fig. 5.4(b). Supported by advanced
quantum theory evaluation of these spectra allowed H ELLWEG (2008) to determine
44 well identified molecular parameters with high precision – thus demonstrating
the high potential of microwave spectroscopy for structural assignments also in the
case of larger molecular systems.
Often microwave spectra are rather complex and difficult to analyze. Then it
may be helpful to induce transitions by two (or even more) frequencies simultane-
ously (so called double resonance spectroscopy). In Fig. 5.5 we illustrate this impor-
tant and highly selective method for the example of the HO3 radical. In Fig. 5.5(a)
a part of the rotational level scheme of HOOO is sketched schematically (with-
out hyperfine splitting). A ‘simple’ absorption line observed with cm waves is
shown in Fig. 5.5(b). It arises from the rotational excitation NKa Kc = 101 ← 000
at 18709.1 MHz as indicated by the red arrow marked (b) in the term scheme
Fig. 5.5(a). Keeping this frequency fixed, one records the double resonance sig-
nal shown in Fig. 5.5(c) by tuning a mm wave over a range from 80153 MHz to
80161 MHz as indicated by the grey arrow marked (c) in the term scheme. As
shown in Fig. 5.5(c), two hyperfine components of the 111 ← 000 transition are
detected by reduction of the single photon absorption line Fig. 5.5(b) at its maxi-
mum: the ground state is depleted if one hits the (second) resonance. (Note that the
scanning range in Fig. 5.5(c) is much smaller than the levels linewidths indicated in
Fig. 5.5(a).)
296 5 Molecular Spectroscopy
Using such precision spectroscopy, S UMA et al. (2005) were able to determine
the structure of this molecule with high accuracy. The key results are summarized
in Fig. 5.5(d). The HO3 radical is very important for atmospheric photo-chemistry
and we shall come back to it again later.
Section summary
• Microwave radiation induces rotational transitions in small and large mole-
cules (E1). Due to the ν 3 factor associated with spontaneous emission, only
absorption spectra are observed.
• Using various sophisticated methods such as up and down frequency conver-
sion, frequency synthesizing, heterodyne detection and employing FT tech-
niques or molecular beam targets, it has been developed today to an extreme
precision, which allows one to determine molecular constants with highest
accuracy.
• In addition to the principle moments of inertia from which overall dimensions
of a molecule may be extracted, also internal motions such as torsion or in-
version vibrations may be gleaned from such spectra.
5.3.1 General
With infrared spectroscopy (partially also in the visible) one studies vibration-
rotation bands, typically in absorption. To give an impression of the complexity that
some of these spectra may involve, we show in Fig. 5.6 the line strengths of CO2
and H2 O over a wide spectral range from 400 nm to 17 µm. These sticks-spectra
have been derived from GATS (2012), using the HITRAN spectral database.1 We
cannot go into details, but we mention that parts of these spectra have already been
discussed in Chap. 4. For CO2 this is indicated by the blow up in the inset, which
corresponds to the (111 0) ← (000 0) transition for which a simulated absorption
spectrum has been shown in Fig. 4.9 – there as a function of wavenumbers.
For the water molecule, most of the bands seen in Fig. 5.6, in particular those at
shorter wavelengths, are overtone and combination bands of the three normal vibra-
tional modes of H2 O. We may compare the F OURIER transform spectrum Fig. 4.12,
briefly discussed in Sect. 4.2.4, which corresponds to a tiny section of the visible re-
gion presented here (0.5913 µm to 0.5928 µm). This just emphasizes – specifically
for the water molecule – the amazing richness of its visible and infrared spectrum.
So much for the complexity of vibrational spectra. Nevertheless, for the vast
majority of molecules IR spectra show very characteristic features and vibration-
rotation spectroscopy provides an excellent tool for chemical analysis. Convention-
1 Note that these stick-spectra just communicates molecular line strengths. They do not yet repre-
ally one still uses so called glow bars as radiation source. These are graphite or SiC
rods are heated by an electric current up to 1000 °C to 1500 °C and emit continuous
IR radiation corresponding to P LANCK’s radiation law. The IR radiation is fed by
mirrors through a typically rather long absorption cell. Usually, a reference beam
passes through an identical, but empty absorption cell in order to register fluctu-
ations in the IR source. Behind the cell the radiation is spectrally dispersed by a
grating spectrometer. Detection is achieved by thermic devices (bolometers) or by
IR sensitive photodiodes. Typically, the IR beam is modulated to distinguish the
signal from the background of the apparatus (thermal radiation).
But laser sources or laser diodes conquer more and more of the IR spectral re-
gions and allow of course for better resolution and higher sensitivity. In particular
for various applications in analytical sciences, absorption spectroscopy with tunable
laser diodes appears to become a method of choice. Synchrotron radiation too is a
frequently used source, in particularly so if broad tunability is needed. In addition,
several free electron laser systems are available worldwide, designed specifically
for the IR spectral region. They have proven their usefulness as an efficient tool in
the spectroscopy of molecules and clusters.
Generally speaking, infrared spectroscopy is a very important method in chem-
ical and physical analytics. Numerous textbooks on the subject exist which treat
the subject extensively (often together with R AMAN spectroscopy). Thus, we just
mention here that IR spectroscopy is highly selective for specific chemical bonds –
based on a wealth of data available in the literature. Various molecular groups in
larger (specifically organic) molecules oscillate with very characteristic frequen-
cies: one finds e.g. the stretch vibrations of the CH group around ν̄ 3000 cm−1 ,
NH at 3400 cm−1 and OH at 3600 cm−1 . In contrast, bending vibrations are usually
found at wavenumbers below 1000 cm−1 . From characteristic vibrational spectra of
a system the expert may already ‘see’ its essential structural features – even without
298 5 Molecular Spectroscopy
detector
Standard spectrometers have the big disadvantage that the spectra are measured se-
quentially, i.e. the spectrometer is tuned through a range of wavelengths and the
detector registers the absorption of the radiation – effectively as a function of time.
Since absorption lines are typically narrow, most of the time no absorption signal
is recorded during a measurement. Thus, the average signal is usually very low and
requires long integration times (or many iterative measurements) to obtain a reason-
able signal to noise ratio.
F OURIER transformation IR spectrometer (FTIR) overcome this problem by an-
alyzing and recording the whole spectrum at once with the help of a M ICHELSON
interferometer. Thus, F OURIER transform spectroscopy belongs today to the most
efficient methods of optical spectroscopy (in various frequency regions). We have
already mentioned some applications previously, e.g. in Sects. 3.4.1 and 5.2. To un-
derstand FTIR spectrometry we may resort to what has already been discussed in
Chaps. 1 and 2 (and Appendix I in Vol. 1) about F OURIER transform, interference
and coherence.
Experimental Setup
Sketched in Fig. 5.7 is the schematic of a typical setup with a M ICHELSON inter-
ferometer (see also Fig. 6.4 in Vol. 1). One uses a broad band light source (‘white
light’) which after collimation to a parallel light beam hits a beam splitter (half trans-
parent mirror plates) where it is divided into two essentially identical beams. Each
of these beams travels on a different optical path. One of the mirrors is moveable
over a distance s/2 and allows one to introduce an optical path difference s between
5.3 Infrared Spectroscopy 299
δ = s/c (5.2)
(we assume here rays which are parallel to the optical axes). After the two beams
have acquired this delay, their wave fronts are merged again through the beam split-
ter and are then focussed onto the probe of molecules to be studied. There, the
spectrum of the probe is impressed onto the optical field. Finally, the thus modified
white light is focussed onto the detector to interfere. By continuous variation of the
spatial (respectively temporal) delay one obtains an interferogram which contains
the whole spectral information about the probe.
Method: FT of a Spectrum
To see this, we first note that the FTIR scheme corresponds essentially to the general
setup of an interference experiment as discussed in Sect. 2.1.4. According to (2.24)
the intensity registered at the detector consists of a constant background and an
interference term (5.4) which depends on the optical delay time δ and the spectral
intensity distribution I˘(ω) of the light source. The initial intensity I0 at the entrance
slit of the interferometer may be written
I0 = I˘(ω)dω. (5.3)
√
It is split into two equal beams (a1 = a2 = 1/ 2) and with (2.29) the signal without
probe, detected by the FTIR spectrometer, is
I (δ) = Re 1 + eiωδ I˘(ω)dω. (5.4)
When passing the probe the intensity distribution I˘(ω) in frequency space is mul-
tiplied with the absorption profile of the probe. Thus, I˘(ω) becomes S̆(ω), which
is nothing else but the absorption spectrum of the probe as it would be obtained by
a dispersive monochromator when tuning its transmission frequency (for the same
light source being absorbed by the same probe). Corresponding to (5.3) the detector
now integrates over the spectrum S̆(ω):
I (δ) = Re 1 + eiωδ S̆(ω)dω. (5.5)
Thus, we obtain an experimental signal I (δ) which depends on the temporal delay δ.
The first term of this integral gives a constant background that is practically identical
to I0 (since from the whole spectrum of the source usually only a small fraction is
absorbed). However, the second term depends explicitly on δ:
Re S̆(ω)eiωδ dω = I (δ) − I0 . (5.6)
300 5 Molecular Spectroscopy
S(ω;t1) 1.0
0.5
w
wt1=2
wt1→ ∞
-6 -4 -2 0 2 4 6
(ω- ω0)/ w
Fig. 5.8 Comparison of a genuine L ORENTZ line profile (full, red line) with the truncated
F OURIER back-transformed shape according to (5.8). The back-transformation has been constraint
alternatively to two different time intervals: wt1 = 2πt1 ν1/2 = 1 (grey) and = 2 (red dashed)
The integral is nothing else but 2π times the real part of the inverse F OURIER trans-
form (I.3), Vol. 1 of the spectrum S̆(ω). F OURIER back transformation gives2
1
S̆(ω) = I (δ)e−iωδ dδ. (5.7)
π
The big advantage of the FTIR method is that no separation of the frequency
components by a dispersive monochromator is necessary. All spectral components
are ‘multiplexed’, i.e. they are, so to say, registered simultaneously. Very important
too is the fact that one works with an extended light source and a collimated beam.
It turns out that in this case the lateral coherence (see Sect. 2.1.7) of the source is
much less critical than in a conventional, dispersive spectrometer.
can only be carried out up to a maximum delay time t1 = s/c as given by the max-
imum optical path difference s available by the spectrometer. Figure 5.8 shows the
observed profiles for two different values of wt1 = 2πt1 ν1/2 in comparison with
the original L ORENTZ profile – as one would see it by full back transformation with
2 Strictly mathematical the following equation is not completely correct: since we back-transform
only the real part of the inverse FT, negative frequencies arise . . . which have to be ignored.
5.3 Infrared Spectroscopy 301
t1 → ∞. The line profiles shown in Fig. 5.8 are indeed very similar to those ob-
served experimentally, as we have seen e.g. in Fig. 3.21. For wt1 ≥ π the signal
S̆(ω; t1 ) is already very similar to the original.
Thus, with ν1/2 1/(2t1 ) the resolving power of this spectrometer is
λ ν s s
2t1 ν = 2 ν = 2s ν̄ = 2 = N × z (5.9)
λ ν1/2 c λ
– apart from a somewhat arbitrary prefactor which characterizes the separability of
neighbouring spectral lines. This corresponds again to the key formula (6.4) already
derived in Vol. 1 for the resolving power of all spectrometers based on interference.
Here we have identified N = 2 as the number of interfering beams and z = s/λ as
the order of interference.
Some Technicalities
Today, the necessary F OURIER back transformation, i.e. the integration (5.7) over
the measured signal I (δ), can be carried out more or less ‘on-line’ with a PC, using
“fast F OURIER transform algorithms” and special processors built into the electron-
ics of the experiment.
The side maxima, illustrated in Fig. 5.8, caused by this evaluation step, may
still pose a serious problem. To overcome this influence of finite optical delay, the
measured interferograms are often apodized, i.e. they are convoluted (smoothed)
with a function p(x) which has no or only weak side maxima, e.g. a triangular
function. Albeit this may reduce the resolution, the overall contrast in the back-
converted spectra can be improved in this manner.
The step width when scanning s has also to be chosen with care. The so called
N YQUIST sampling theorem demands that per period at least two points have to be
known in order to reproduce a sine function. Finally, one has to account for the finite
divergence angle of the beams, since (5.2) strictly holds only for coaxial rays.
Modern, commercially available FTIR spectrometers have resolving powers
from 1000 up to 1 000 000. To separate e.g. two lines at ν̄ 3300 cm−1 with a
difference in wavenumbers of ν̄ = 0.1 cm−1 the mirror has to be translated by
at least 5 cm, according to (5.9). Typical optical delay lines in FTIR spectrometers
range from 10 cm up to 2 m.
F OURIER transform spectrometers are particularly popular for investigations in
the IR spectral range, and we have already mentioned applications in the microwave
range. The mirrors in FTIR spectrometers have to be moved over rather long dis-
tances with a precision of distance and alignment to only fractions of a wavelength.
Clearly, this is easier to realize for infrared and a fortiori for microwave radiation
than in the visible. More and more the technique is, however, also used in the visible,
ultraviolet and even VUV spectral region, employing highly sophisticated methods
for the stabilization and beam control. As broad band light source one may use with
advantage synchrotron radiation. Also for the spectroscopy of astronomical objects,
for solar studies as well as atmospherical investigations, F OURIER spectrometry is
a method of choice since these are all broad band radiations sources whose fine
structure one wants to investigate.
302 5 Molecular Spectroscopy
If only few particles are available for a spectroscopic investigation, as e.g. in stud-
ies of dynamical processes in molecular beams, or IR spectroscopy of molecular
clusters in a supersonic jet, the kind of absorption spectroscopy described above
soon reaches its limits due to small absorption coefficients, combined with finite
sensitivity and dynamical range of the detectors. Absorption spectroscopy, in gen-
eral, has the disadvantage that signals are recorded as potentially small reduction
of a high background (the incoming light). Fortunately, a number of methods have
been devised to outwit this problem by direct detection of the products generated
in the absorption process. Such methods may be very sensitive and even allow the
detection of single particles. We shall further explore this in the following sections.
Here we present the so called infrared action spectroscopy (IAS). We illustrate
this method by experiments with HO3 – a radical already known to us from Sect. 5.2.
It plays an important role in the photochemistry of the earths atmosphere as an
intermediate species of critical reactions in ozone chemistry of the type
H + O3 → HOOO → OH(v ≤ 9) + O2
OH(v) + O2 → HOOO → OH(v − 1) + O2 (5.10)
O + HO2 → HOOO → OH + O2 .
34.0
OH A2Σ+
32.0
UV probe, fixed
8.0 O2 a1Δg
energy / 103cm-1
vibrational
_ pre- N=
6.0 2 νOH dissociation
9
5.0 8 v= 1
IR pump 0
_ (tuned) LIF
2.0 νOH
9
8 v= 0
0.0 0
OH X 2ΠΩ
-
-2.0 +O2 X 3Σg
HOOO X 2A''
Fig. 5.9 IR pump, UV probe scheme for the detection of OH vibrations in the HOOO radical
according to D ERRO et al. (2007). After IR excitation of the OH stretch vibration νOH in the v = 1
(alternatively v = 2) state HOOO decays into the end products OH(v − 1)X 2 ΠΩ + O2 X 3 Σg− ,
thereby loosing one ωOH quantum. One detects the IR absorption by laser induced fluorescence
(LIF) in the dissociated OH after UV excitation. The energy scale refers to the dissociation channel
OH + O2 (note the scale change between 8000 and 32000 cm−1 )
This double resonance experiment illustrates quite impressively the high standard
of the methods in modern molecular spectroscopy. As IR source (2.8 µm or 1.4 µm
for νOH and 2νOH , respectively) optical parametric oscillators are used, pumped
by an injection pumped Nd:YAG laser. For exciting the product OH radicals a fre-
quency doubled dye laser is used, again pumped by a Nd:YAG laser. Thus, detection
is done state selectively. Even the rotational distribution of the OH may be deter-
mined in this way. Scattered radiation has to be screened very well and the detection
efficiency of the fluorescence detectors has to be very high.
Figure 5.10 shows two typical IR vibration-rotation spectra, detected by action
upon absorption as just explained. The rotational structure of this asymmetric top
rotor is only partially resolved. Due to the large moment of inertia it is much nar-
rower as we have seen in Sect. 3.4.5 for diatomic molecules. Note that P , R and Q
branches are observed – we have already discussed in Sect. 4.2.3 the appearance of
the Q branch in polyatomic molecules. The simulation of the spectra by D ERRO et
al. (2007) shown in Fig. 5.10 is based on high quality quantum mechanical compu-
tations (MRCI), using the rotational constants determined by S UMA et al. (2005).
It turns out that inclusion of additional parameters such as centrifugal distortions,
spin-rotation and magnetic dipole coupling does not improve the overall excellent
agreement any further.
Quite remarkable is the unstructured, broad background in the spectra shown
in Fig. 5.10, corresponding to the red dashed contribution of the simulations. It is
attributed to the cis-configuration of HOOO (H atom and final O atom are on the
same side). Obviously it is only very short-lived, so that the rotational structure is
washed out.
304 5 Molecular Spectroscopy
_
(a) 2 νOH
intensity
(b) _
νOH
-10 -5 _ 0 5 10
IR Δν / cm-1
Fig. 5.10 IR action spectra for the HOOO radical in (a) the νOH and (b) 2νOH spectral region
according to D ERRO et al. (2007). The experimental detection (black lines) was done at fixed
probe laser wavelength on the P1 (4) transition of the OH A 2 Σ + ← X 2 Π (v = 1, 0) and (1, 1)
bands, respectively. The simulation (grey, nearly indistinguishable from the experiment) accounts
for trans-HOOO (full red) and cis-HOOO (dashed red). The relative wavenumbers refer to the band
origins of the trans-HOOO species at ν̄OH = 3569.30 ± 0.05 and 2ν̄OH = 6974.18 ± 0.05 cm−1 ,
respectively
ν4 ν5 ν6
OOO bend central OO stretch torsion
In the mean time, the IR spectra of HO3 and DO3 have been determined in even
more detail (see e.g. D ERRO et al. 2008; M C C ARTHY et al. 2012). The normal
mode structure shown in Fig. 5.11 may illustrate what kind of details modern IR
spectroscopy is able to reveal.
Section summary
• By infrared (absorption) spectroscopy vibration-rotation transitions (E1) of
molecules are studied. They are a rich source for obtaining detailed informa-
tion on molecular structure with characteristic absorption bands that may be
attributed to specific molecular bonds.
5.4 Electronic Spectra 305
which depends on the nuclear coordinate R – quite similar as discussed in Sect. 3.4.4
for the case of pure vibrational transitions. The transition probability (and the ab-
sorption cross section) between two vibrational states v ← v thus become propor-
tional to
2
|Dγ v ←γ v | = Rγ v (R)Dγ ←γ (R)Rγ v (R)dR .
2 ∗ (5.12)
Dγ ←γ (R) varies only slowly with R, and the vibrational wave functions are lo-
calized to the vicinity of R0 . Since, however, these radial wave functions between
different electronic states are not orthogonal, expansion around R0 as in (3.80) does
not lead us any further. Rather, we average Dγ ←γ (R) over R and use to 1st order
its mean value Dγ ←γ (R):
2
2
| Dγ ←γ (R) × R∗γ v (R)Rγ v (R)dR .
Dγ v ←γ v |2 = (5.13)
It is the square of the overlap integral between the vibrational wave functions in
the potentials Vγ (R) and Vγ (R) and determines the relative intensities of vi-
brational lines within one electronic transition γ ← γ . The FC factors may be
computed explicitly in the harmonic approximation since the radial wave functions
Rγ v (R) = hv (x) are then simply the Hermitian functions (3.24) with x given by
(3.22). One may then show rather easily that for absorption the sum rule
FC γ v ← γ v = 1 (5.15)
v
leads to more or less strict selection rules due to angular momentum conservation,
just as in atomic physics. We shall come back to these in the next subsection.
However, which vibrational levels are populated in absorption or emission de-
pends very specifically on the FC factors (5.14) of each molecule. The rules emerg-
ing, the so called F RANCK -C ONDON principle, are no strict selection rules. Rather
the FC factors lead to some kind of propensity rules which are essentially a conse-
quence of the shape and relative position of the two potential wells involved in the
electronic transition. We distinguish the two cases most often encountered:
T e'
D''e
X
D''0
v'' = 0
R
Fig. 5.13 F RANCK -C ONDON principle for an optical excitation of a molecule from its ground
state X into an excited state A. In polyatomic molecules or in a relaxing medium (liquid) the
initial vibrational excitation in the A state is rapidly converted (dissipated) and the molecule emits
from the vibrational ground state of A. This leads to a characteristic red shift in respect of the
absorption profile (S TOKES shift), as indicated in the inset for the example of rhodamine 6G in
ethanol (excitation at 480 nm, equivalent to 20.8 × 1000 cm−1 )
The relative magnitude of these F RANCK -C ONDON factors for emission is esti-
mated by the same arguments as for absorption: The most intensive lines are found
close to a vertical transition (γ v ← γ 0) with high v , as indicated in Fig. 5.13
(wide, pink arrow). A direct transition to v = 0 is not very probable. The emis-
sion is thus shifted towards longer wavelengths, as indicated in the inset for the well
known dye molecule rhodamine 6G in ethanol as solvent – as used in tuneable dye
lasers. This red shift is called S TOKES shift. Absorption and emission spectra are
approximately mirror images of each other.
4 4 4
3 3 3
2 2 2
1 1 1
v' =0 v' =0 v' = 0
Fig. 5.14 Absorption spectra (schematic) as a function of equilibrium distances R0 : (a) R0 does
not change during excitation (case 2); (b) R0 increases (case 1); (c) the equilibrium distance R0 in
the excited state is shifted so much that excitation occurs dominantly into the dissociative contin-
uum (also case 1)
Summary
Figure 5.14 summarizes and specializes the FC principle of vertical transitions for
different relative equilibrium positions of ground and excited states. The different
forms of an electronic band spectrum allow already to glean important information
about the change in equilibrium distance R0 during electronic excitation.
The F RANCK -C ONDON factors identify the most probable vibrational transitions
between different electronic states. In addition, owing to conservation of angular
momenta and parity, more or less rigid selections rules also hold for molecules –
similar to those in atomic physics. We summarize some key rules here briefly. As
for electric dipole transitions in atoms, quite generally the following selection rules
for the total angular momentum J must hold:
J = 0, ±1 but J = 0 J = 0 (5.16)
and MJ = 0, ±1.
In addition, for small molecules (RUSSEL -S AUNDERS coupling) the total electron
spin S is conserved in E1 transition, i.e. S = 0. As already discussed for atoms,
E1 transitions between singlet and triplet system (intercombination lines) are for-
bidden.
310 5 Molecular Spectroscopy
Table 5.1 Allowed electronic E1 transitions for diatomic molecules (C∞v symmetry)
Selection rule For coupling case
total angular momentum J = 0, ±1 all
J = 0 J = 0
electron spin S = 0 all
Σ = 0 only (a)
orbital angular momentum Λ = 0, ±1 (a) and (b)
Σ + ↔ Σ +, Σ − ↔ Σ −, Σ + Σ −
electron Ω = 0, ±1 only (a)
total angular momentum J = 0 forbidden if Ω = 0 → Ω = 0
total angular momentum K = 0, ±1 only (b)
without spin K = 0 forbidden for Σ–Σ
Generally speaking, the situation is more complex than for atoms. For each spe-
cific case the selection rules depend on the symmetry of the system, and on the cou-
pling scheme between the angular momenta (orbital angular momenta of electrons,
electron spins, nuclear rotation, nuclear spin). The simplest case is again C∞v sym-
metry, encountered in diatomic and other linear molecules. We summarize the most
important rules for H UND’s coupling cases (a) and (b) in Table 5.1. These rules will
be illustrated for some examples later on.
Much more complex is the situation for polyatomic, nonlinear molecules. Tran-
sition probabilities are determined here by the molecular symmetry. Clearly, the
dipole transition matrix element (5.11) will depend on the direction of the electric
vector in respect of the molecular axes as well as on the symmetries of initial and
final state. With the help of the group theoretical instruments addressed in Sect. 4.3
one may derive selection rules in a relatively simple manner, if the symmetry group
of initial and final state are the same.4 An electric dipole transition is allowed if the
direct product of the irreducible representation of initial electronic state (ΓGS ), final
electronic state (ΓES ) and of the dipole operator (ΓD ) contains the totally symmetric
irreducible representation (ΓSym ), i.e. if
This rule follows directly from (5.11) by application of group theory. It is intu-
itively evident since only if the product under the integral (5.11) contains at least
some totally symmetric parts, the integral over all space will not vanish. Depending
on the symmetry group the totally symmetric irreducible representation is desig-
nated by A, A1 , A1g , Ag or A, as indicated in the character tables of the respec-
tive symmetry groups. The irreducible group of the dipole operator – in essence
the vector (x, y, z) – can also be found in the character table. For example, in the
Oh octahedral group ΓD = T1u (according to Table 4.5), and for the C2v group
4 Obviously, that is not necessarily the case, since a change in electronic structure may also change
v'' = 0
representative nuclear coordinate
312 5 Molecular Spectroscopy
vibrational states in the S0 ground state and the F RANCK -C ONDON region of the
Sn states above it (as a rule into higher lying vibrational states). In the example
shown in Fig. 5.15 only the ground state S0 , the first excited singlet state S1 and the
lowest triplet state T1 are considered (as in atomic physics the T1 state minimum is
somewhat lower than that of the S1 state and a T0 state does not exist).
After the optical excitation, IVR processes (also radiationless) may redistribute
energy among different vibrational degrees of freedom5 in the S1 state (we have
already encountered IVR in Sect. 5.3.3). In addition, ISC transitions may occur from
S1 → T1 – competing with spontaneous reemission from S1 → S0 (fluorescence) –
typically on a time scale of several ns, depending on the strength of the spin-orbit
interaction and the magnitude of the FC factors between the vibrational levels of S1
and T1 states. In the T1 state the initially high vibrational energy may again relax by
further IVR processes, so that finally the system is found in the lowest vibrational
states of T1 (in Fig. 5.15 indicated by v = 0). Back reactions into the S1 state are
then no longer possible.
In this manner excitation energy may be stored for a long time in the lowest
triplet state T1 . Since E1 intercombination transitions are to 1st order forbidden, the
transition probability into the S0 ground state is typically very small. If the excited
state finally decays by radiation one speaks about phosphorescence. This is a very
weak, long lasting radiation as mean lifetimes of T1 triplet states may be in the ms
to min range.
For completeness we mention other types of radiationless transitions which may
occur, in addition to ISC, among different electronic states with the same multi-
plicity, e.g. Sn (v ) → Sn (v ). Such transition are called internal conversion (IC).
They may occur by “surface hopping” close to avoided crossings (see Sect. 8.1.6 in
Vol. 1) – as far as energetic position and FC factors permit. We speak about non-
adiabatic transitions. We shall address further important processes in the context of
inelastic collisions in Sect. 7.4. A specific kind of such transitions occur at conical
intersections (see Sect. 7.6).
5 One has to keep in mind that during the IVR process in an isolated molecule vibrational energy
does of course not get lost – as one might infer from the display of the energy terms in Fig. 5.15.
Energy (except in optical emission) is just redistributed among the many other vibrational degrees
of freedom within the molecule. A flow of energy back into the ‘representative nuclear coordinate’
is – for statistical reasons – the less probable the larger the molecule.
5.4 Electronic Spectra 313
where Te , G and F represent the differences of the electronic, vibrational and
rotational term energies, respectively (more precisely: the wavenumbers in cm−1 ).
Neglecting centrifugal distortion one obtains for the latter according to (3.57):6
F = Bv N N + 1 − Bv N N + 1 .
Note that the rotational constants Bv and Bv belong to different electronic states
and may thus differ substantially. As selection rule
N = 0, ±1
holds. In contrast to pure rotational spectra which we have treated in Sect. 3.4.5, now
N = 0 is allowed in principle, since angular momentum may also be transferred
to the electronic system. Thus, for electronic transitions three branches exist with
different contributions from rotational transitions, F = P (N), Q(N) and R(N ),
respectively:
P branch N = N − 1: P N = −2Bv N − Bv − Bv N N − 1
Q branch N = N : Q N = − Bv − Bv N N + 1
(5.18)
R branch N = N + 1: R N = 2Bv N + 1
− Bv − Bv N + 1 N + 2 .
As in the case of pure vibration-rotation spectra (see Sect. 3.4.5) we expect the
highest transition wavenumbers for the R branch, the lowest for the P branch (at
least for small rotational quantum numbers N ). Since the difference (Bv − Bv )
may also assume significant positive as well as negative values, we expect large
deviations from equidistance (2Bv ) between the lines – which we have seen as a
first approximation in contrast to pure rotation and vibration-rotation spectra.
The absorption spectrum for an electronic transition will thus look, as a rule,
quite complicated. One observes vibrational bands with a characteristic rotational
structure on top. If one plots the rotational quantum numbers N (for a specific
vibrational transition) as a function of the transition wavenumber ν̄ one obtains
parabolas, so called F ORTRAT diagrams.
ν̄ = o + bN + cN 2 . (5.19)
Their shape reflects the expressions R(N ), Q(N ) and P (N ). They are sketched
in Fig. 5.16. From the form of these parabolas one may in principle glean the dif-
ferences (Bv − Bv ) and thus the change of equilibrium distance from electronic
ground to excited state. Today one typically simulates the measured spectra by
multi-parametric fit-functions – ideally they are rotationally resolved by high res-
olution spectroscopy. One needs to know from theory the relative line strengths for
rotational transitions (for linear molecules these are the so called H ÖNL -L ONDON
factors) which have to be weighted with the thermal population of rotational states
prior to the transition (see examples in Sects. 3.3.3 and 5.5.4).
6 In the literature one often finds J for the rotational quantum number, instead of N .
314 5 Molecular Spectroscopy
_ _ _
ν ν ν
Fig. 5.16 F ORTRAT diagrams for various relative magnitudes of the rotational constants. Dashed
vertical lines indicate the positions of rotational band heads in the spectra from electronic transi-
tions
After these general considerations it is obvious that emission and absorption spectra
from electronic transitions will usually be very complex – even in simple molecules.
Absorption spectra are typically expected in the UV and VUV spectral range, emis-
sion spectra may also be found in the VIS region. Particularly in emission one ex-
pects superposition of many vibration-rotation bands with different v and v , each
of them with typical structures from P , Q and R branches as just discussed. Addi-
tional complications arise from the superposition of many different electronic tran-
sitions, as we recognize already by looking at typical potential energy diagrams for
diatomic molecules (see e.g. Chap. 3, p. 212 ff.).
In hindsight, one can only be full of admiration for past generations of spectro-
scopist who identified and parameterized the majority of simple molecules by their
spectra – long before the invention of lasers. Molecular spectroscopy has been and
still is an active field of research for more than a century, and most of the ground
breaking discoveries have been made and understood long before 1960. Instrumen-
tal for the accumulation and interpretation of a huge wealth of data has been the
life’s work of the physicist Gerhard H ERZBERG – honoured in 1971 with the N O -
BEL prize in chemistry.
The intricate detective work of whole generations of spectroscopists may be il-
lustrated somewhat by Fig. 5.17. There, the emission spectrum of the diatomic PN
molecule is shown. It has been generated in a gas discharge lamp and was analyzed
in classical manner by a highly resolving grating spectrometer and recorded on a
photo plate. The original paper published in the then leading journal was still writ-
ten in German language – before H ERZBERG had to leave Germany and found a
new home in Canada. The interpretation reproduced from his N OBEL prize lecture
(H ERZBERG 1971) in the lower part of the figure is more or less self explaining.
One obtains from it a glimpse of the experimental skill, spectroscopic intuition and
combinatoric insight which were essential in those early days prior to the laser, to
acquire the wealth of knowledge about molecules which today we just take for guar-
anteed.
5.4 Electronic Spectra 315
P Q R
rotation-resolved 0 - 0 band
0 -2 0 -1 0 -0 v' = 0
1 -3 1 -2 1 -1 1 -0 v' = 1
2 -4 2 -3 2 -1 2 -0 v' = 2
3 -5 3 -4 3 -2 3 -1 v' = 3
4 -6 4 -4 4 -3 4 -2 v' = 4
5 -7 5 -5 5 -4 5 -3 v' = 5
∆υ = -2 ∆v = -1 ∆v = 0 ∆v = 1 ∆v = 2
Fig. 5.17 Emission spectrum from the A 1 Π → X 1 Σ + transition in the PN molecule according
to C URRY et al. (1933), recorded in a high resolution 3m UV grating monochromator (for cali-
bration atomic lines in the P I and As I spectra have been used). Middle: photographic record of
the spectrum (in first order diffraction) displaying several vibrational bands. Top: (partially) rota-
tion-resolved 0–0 band (second order diffraction, enlarged section) indicating the position of the
P , Q and R band head. Bottom: scheme for interpretation of the bands as presented in the N OBEL
prize lecture of H ERZBERG (1971)
Such high resolution emission and absorption spectra have remained to be very
useful tools in molecular spectroscopy even after the invention of the laser. Of course
the experimental techniques of registration have been improved substantially. Spec-
tral photometers, CCD cameras and image amplifiers have replaced in the second
half of the past century the direct recording by photographic plates. The case of PN
(Fig. 5.17) may be considered a still surprisingly clear spectrum without overlapping
bands, albeit of a somewhat exotic molecule. Figure 5.18 illustrates how complex
it can get – in spite of micro densitometers – to evaluate the emission spectrum
for one of the most important homonuclear diatomic molecules such as O2 , when
several electronic transitions overlap each other. The bands shown in Fig. 5.18 are
sections (VIS) of the so called H ERZBERG bands, forbidden for optical dipole tran-
sitions, which are important for atmospheric physics. We recall (Fig. 3.44) that O2
is built from two triplet 3 P atoms, from which a large variety of molecular states
may be formed. The spectra shown here are observed in the flowing afterglow of a
gas discharge in O2 -He. They are due to transitions from the closely spaced excited
states c 1 Σu− , A 3 u and A 3 Σu+ (each in its vibrational ground state) into higher vi-
brational states of the X 3 Σu− ground state – or in one case into the first excited state
316 5 Molecular Spectroscopy
+ -
c 1Σu→X 3Σg 0- 12 0- 11 0- 10 0- 9 0- 8 0- 7 0- 6
b 1Σ + →X 3Σ - -
g g A' 3Δu( Ω =2) →X 3Σg 0 -11 0 -10 0- 9 0- 8
0-0
∞
100 + -
50 OH MEINEL band heads A 3Σu → X 3Σg 0-10 0-9 0-8
30
20
15
intensity
10
Fig. 5.18 Section of the H ERZBERG bands of O2 in the visible spectral range. This emission
spectrum from the afterglow of an O2 -He discharge was recorded by S LANGER (1978) behind a
high resolution Echelle spectrometer, still using a photographic film and image amplifiers. It has
been evaluated with a micro densitometer (note the essentially logarithmic sensitivity of the film,
which complicates the quantitative evaluation)
o 1 g which lies only 1 eV above the X state. The readers may follow the interpre-
tations given in Fig. 5.18 without difficulty by comparing them with Fig. 3.44. Here
again, one has to admire the intellectual achievements of a generation of spectro-
scopists, which have solved a multitude of such delicate puzzles (using the relevant
theory), and finally have transformed these data into flocks of molecular potentials
of the type shown in Sects. 3.6 and 3.7.
Such classical spectrometers are still used today and compete often remarkably
well with much more expensive methods from laser spectroscopy – in particular so
for analytical purposes. Detection and recording is done today usually with optical
multi channel analyzers (OMA) that are read out and controlled by a computer. In
addition FTIR spectroscopy is used more and more also in the VIS and UV spectral
range. However, without further selection techniques the interpretation of unknown
spectra of this type in absorption or emission, is even today still a challenging and
complicated undertaking.
Section summary
• Spectra of electronic transitions in molecules typically have a rather com-
plex band structure which is determined by F RANCK -C ONDON factors
|γ v |γ v |2 . These describe the overlap between the vibrational states
|γ v and |γ v in the electronic states characterized by the quantum num-
bers γ and γ , respectively. In a classical picture they correspond to “verti-
cal” transitions between the two potentials involved.
5.5 Laser Spectroscopy 317
At this point, laser spectroscopy comes to help. The high monochromaticity of laser
light, ideally, makes it possible to induce transitions between exactly one initial and
one final vibration-rotation level in the initial and final electronic states, respec-
tively. For any subsequent process this reduces the large number of intermediate
states which otherwise might participate in a second step absorption or emission
process. Thus, the complexity of the spectra is dramatically reduced and precision
is substantially improved. As a rule, the price to pay for these advantages is consid-
erably higher technical effort.
Since the invention of the laser in 1960, and in particular since flexible, tuneable
laser systems for a wide spectral range are available, numerous methods of laser
spectroscopy have been devised – more or less sophisticated and efficient. They
differ in the methods of preparing the species to be investigated and in the detection
schemes for the photo-absorption processes exploited. The amount of data gained in
this way is enormous and we cannot attempt here any kind of summary (we mention,
however, the N OBEL prize to S IEGBAHN 1981). We shall simply give a brief survey
on the most important methods and present a few, particularly interesting examples.
Laser induced fluorescence (LIF) is probably the most often used method in laser
spectroscopy. One excites with a narrow band laser and detects the emitted fluores-
cence. The most straight forward procedure is to tune the exciting laser and to detect
318 5 Molecular Spectroscopy
all emitted fluorescence. Whenever the exciting laser hits a resonance, fluorescence
increases correspondingly. The result corresponds basically the classic absorption
spectroscopy. The two obvious advantages of this procedure are (i) a substantially
enhanced resolution due to the monochromatic laser radiation, (ii) the detection ef-
ficiency can be orders of magnitude higher than in classical absorption spectroscopy
since the fluorescence is now detected on a negligible background. LIF is, roughly
speaking, a flop in method. In contrast, in usual absorption spectroscopy the trans-
mitted signal of the intense, incident radiation is just slightly reduced due to the
resonance one is looking for (flop out experiment). Thus, LIF may also be com-
bined very conveniently with D OPPLER free methods as we have discussed them in
Vol. 1 on several occasions. The high sensitivity of the method allows one e.g. to
perform such measurements on cold, supersonic molecular beams where the num-
ber of thermally populated vibration-rotation levels in the initial electronic state is
kept low. This leads to very clear, easy to interpret spectra.
Alternatively one may also set the excitation wavelength to a fixed value and ana-
lyze the spectral distributions of the emitted fluorescence. In this manner one obtains
emission spectra from well defined vibration-rotation levels in an electronically ex-
cited state. The combination of both schemes, finally, gives very comprehensive and
unambiguous sets of data (partially redundant if desired), which allow for a precise
and unique determination of the electronic and nuclear structures of the molecules
studied, even in the case of rather complex species.
As a special, particularly simple and clear example we discuss the laser induced
fluorescence of isolated iodine molecules. I2 is a molecule that has been studied
with extremely high spectroscopic precision. Owing to its high molecular mass of
253.8 (very low D OPPLER broadening) and its negligible isotope abundance (spec-
troscopically unique), it is often used in spectroscopy as an excellent secondary
wavelengths standard. Many thousands of absorption and emission lines of I2 have
been tabulated with high precision.
As shown in Fig. 5.19(a), spin-orbit splitting is very large.7 The X state dissoci-
ates (R → ∞) into two ground state atoms I(2 P3/2 ), the excited B state into one I
atom in the 2 P3/2 and one in the excited 2 P1/2 state.
The example shown here may even be presented as a demonstration experiment
in the classroom for undergraduate students.8 As illustrated in Fig. 5.19, one ob-
serves the fluorescence after excitation of the B0+ 3
u (or Πg ) state from the ground
+ 1
state X0g (or Σg ).+
7 Thus the coupling of angular momenta in I2 is most appropriately described as H UND’s case (c),
see Sect. 3.6.4. The often used classification by singlet and triplet looses its validity due to the
strong spin-orbit splitting. Otherwise the transitions studied here would all be forbidden intercom-
bination lines.
8 As done by Hartmut H OTOP (2008) who made these data available for us.
5.5 Laser Spectroscopy 319
W / eV _
I2 ν / 1000 cm-1 cell
(a) laser I2
+ 514.5 nm molecules
laser excitation 514.5nm
B 0u ( 3 Π g ) 2P
3/2 +
2P
1/2 12 40
v' = 43
N' = 12;16 mono-
2 30
chromator
14 _
2P
3/2 +
2P
3/2 ν
20 (b) det.
+ +
X 0 g (1Σ g)
16
1 v'' = 40
30 10
20 (c)
18
10
v'' = 4
0
0 v'' = 0
0.2 0.3 0.4 0.5 R / nm
R 0 = 0.266 nm 20 fluorescence intensity
Fig. 5.19 Laser induced fluorescence from I2 after excitation of the B(v = 43, N = 12; 16) ←
X(v = 0, N = 13; 15) transition; (a) potential energy diagram (essentially according to DE J ONG
et al. 1997); the black arrow illustrates monochromatic, vertical absorption at λ = 514.5 nm (ar-
gon-ion laser), the two pink arrows indicate emission; (b) schematic of the experimental setup with
an iodine cell from which the B(v = 43) → X(v ≥ 0) fluorescence spectrum (c) was obtained
according to H OTOP (2008)
B(v = 43, N = 16) are excited. In the reemission process (two pink arrows in
Fig. 5.19(a)), only transitions B(v = 43) → X(v ≥ 0) occur, with N = 0, ±1.
The small rotational broadening is not resolved in the present experiment. Other
electronic states (grey dashed in Fig. 5.19(a)) do not play any role in this exper-
iment. The fluorescence spectrum Fig. 5.19(c) shows very clearly all vibrational
levels of the electronic ground state with 40 v ≥ 0 (indicated in the figure up
to v = 4 by black horizontal lines, spaced by 214 cm−1 ). The F RANCK -C ONDON
factors |B(v = 43)|X(v )|2 decrease rapidly with increasing v , corresponding
to the positions of the potentials. The alternating intensity of the lines too, may eas-
ily be understood, considering the alternating, approximately even or odd symmetry
of the vibrational wave functions R0v (R) in respect of R0 (see Fig. 3.5): in one
case the maxima in the B state (v = 43) are just above the vibrational maxima of
the X(v ) states, in the other case maxima coincide with zero transits so that certain
regions of the wave functions average out.
We finally mention that LIF provides also a very efficient detection method for
molecules or radicals with well known spectra. We have seen this already in the
case of OH in Sect. 5.3.3: one tunes the exciting laser in resonance with a known
transition of the species to be detected and registers their presence or formation via
fluorescence. The method is not only very sensitive and allows the detection of low
320 5 Molecular Spectroscopy
(a) W
(b) W e-
e- e- AB++C e- AB++C
ABC+ hνpr2
ABC+
Fig. 5.20 Scheme for (a) resonantly enhanced two-photon ionization spectroscopy (RTPI) and
(b) depopulation spectroscopy by ionization or fragment detection
concentrations, it is also state selective. Hence, it is used with great success e.g. for
the analysis of reactive collision processes or photoinduced reactions.
Next to LIF, the resonantly enhanced multi-photon ionization (REMPI) is one of the
most often used, particularly sensitive methods of modern molecular spectroscopy
with lasers. It comes in several versions, such as resonant two-photon ionization
(RTPI, also R2PI). We have already encountered a non-resonant variety in connec-
tion with D OPPLER-free methods in Vol. 1, Sect. 6.1.8. The basic idea of REMPI
is quite similar to LIF. Detection is now, however, achieved by ionization of the
molecule with the help of one or more additional laser sources. This leads to an
even higher detection probability since ions may be collected with nearly 100 %
efficiency. By using time of flight or quadrupole mass spectrometers one may in
addition exploit mass selectivity. This can be very useful e.g. when studying iso-
tope mixtures or photoinduced fragmentation processes. Of course such methods
are limited to investigation of molecules or clusters in the gas phase or at surfaces in
the vacuum, since one wants to transport the ions or electrons generated with little
losses to the detector.
The scheme of RTPI is sketched in Fig. 5.20(a) for a triatomic molecule ABC.
With the tunable “pump” photon hνpu different vibration-rotation levels may be
populated within the FC region of the potential for electronically excited ABC∗ .
Starting from this level, a “probe” photon hνpr (usually at a fixed frequency) ion-
izes the system. Mass selection may be necessary, if the ABC+ ion fragments or
if the initial target is composed of several constituents. One finally detects the ions
with high efficiency by a secondary electron multiplier (see Appendix B.1). In this
manner one obtains an electronic molecular spectrums which is fully equivalent to
the classical absorptions spectrum (including the typical rotational bands with P , Q
and R branches) – except with much higher sensitivity. Owing to the latter, super-
sonic molecular beams may be used to cool the initial population of rotational and
5.5 Laser Spectroscopy 321
vibrational states dramatically, so that the spectra become much more transparent
than those obtained from a gas cell or even a liquid.
Alternatively one may also detect the photo-electrons – in more sophisticated
experiments with analysis of their kinetic energy or even in coincidence with the
ion (see also Sect. 5.8).
The detection scheme has to be even more specific if the excited state decays
during the laser pulse duration by unimolecular dissociation ABC∗ → AB + C,
or internal conversion (IC) or other processes. As indicated in Fig. 5.20(b) there
are several options to overcome this problem and gain additional information about
the decay mechanism. One may, e.g. try to detect the reaction products directly
by ionizing them with a suitable probe photon hνpr2 (we have met an example in
Sect. 5.3.3). Alternatively one may study the depletion of the initial state of the
ABC molecule by ionization with a probe photon hνpr1 of sufficiently high energy.
This ion signal is then detected as a function of the pump photon energy hνpu . At
each resonance the signal decreases, hence one speaks about depletion spectroscopy.
One finds a wealth of beautiful examples in the literature for all these varieties
of REMPI. They range from diatomic molecules, with NO being some kind of
‘Drosophila’ of molecular RTPI studies, up to small biomolecules with many atoms,
for which we shall present some examples in Sect. 5.5.4. At this point we want to
discuss in some detail the spectra of Na3 and Li3 which are already quite complex.
They will allow us to familiarize ourselves with a number of interesting aspects in
molecular spectroscopy.
These triatomic molecules (owing to their relatively weak bonding they often are
also considered to be metal clusters) are generated in supersonic molecular beams.
A rare gas at high pressure (usually argon at 2–20 bar) flows through a temperature
resistant cartridge (oven) with the hot metal atoms (vapour pressure 10 to 100 mbar)
and expands together with this metal vapour through a narrow nozzle into vacuum
(so called ‘seeded’ beam). This expansion leads to adiabatic cooling and partial
condensation. The thus generated clusters have internal vibrational, rotational and
translational temperatures of only a few degrees K – however a high overall veloc-
ity. By collimation through a set of “skimmers” (usually of conical shape) a well
defined cluster beam is formed. In a second, differentially pumped chamber it is
then crossed with the laser beams. From the interaction volume one extracts the
ions and/or electrons which are then detected.
Pioneering experiments with Na3 have already been performed in 1986 with
pulsed, tunable dye lasers (D ELACRÉTAZ et al. 1986; B ROYER et al. 1986, 1987).
Figure 5.20 shows a small selection from these data, illustrating the rich kind of
spectroscopic information that demands adequate interpretation. D ELACRÉTAZ et
al. (1986) have tried this, using half integer quantum numbers which they attributed
to so called pseudorotation: The basic arrangement of the three Na atoms is an
equilateral triangle, which is, however, JAHN -T ELLER distorted. Later, quantum
mechanical calculations have shown that the Na3 system is even more complex as
initially assumed; the interpretation given in Fig. 5.21(b) was not fully confirmed.
It turns out that the JT features are superposed by a PJTE effect (see Sect. 4.3.4).
322 5 Molecular Spectroscopy
(a)
(c)
Qs
500 550 600 650 nm
9/ 7/ 5/ 3/ 1/
2 2 2 2 2
9/ 7/ 5/ 3/ 1/
(b) 2 2 2 2 2 Qb
3/ 1/
2 2 9/ 7/ 5/ 3 / 1/
2 2 2 2 2
Qa
Fig. 5.21 RTPI spectra of Na3 , recorded with two pulsed dye lasers. (a) Survey spectrum in the
visible spectral region. (b) Blow up for one electronic transition, with interpretation of the bands
by half integer quantum numbers for the pseudorotation (according to D ELACRÉTAZ et al. 1986).
(c) The three normal modes of trimer molecules such as Na3 and Li3 in D3h geometry
Since then quite a few theoretical and experimental investigations have been de-
voted to these problems. Highly resolved measurements and state-of-the-art quan-
tum chemical calculations for the significantly simpler Li3 (K RÄMER et al. 1999;
K EIL et al. 2000) have led to a complete understanding of the ro-vibronic struc-
ture of such systems and their interesting spectroscopy (see also B ERSUKER 2001).
Without entering into full depth of the problems, we give a brief introduction to the
not completely trivial phenomena connected with these symmetry issues.
The starting point is an equilateral triangle in D3h symmetry. The character Ta-
ble 5.3 lists the irreducible representations of this point group. The electronic ground
state of Li3 belongs (as that of Na3 ) to a two fold degenerate E representation. As
discussed in Sect. 4.3.4, this degeneracy is not allowed in such a molecule and is
avoided by a JT distortion, lowering the energy minimum.
The potential surface thus forms two sheets which conically intersect. In the
vicinity of this intersection the state can no longer simply be written as a (B ORN -
O PPENHEIMER) product of electronic and nuclear wave function. An exact descrip-
5.5 Laser Spectroscopy 323
obtuse acute
Qb
triangle triangle
© = 60o
C2v © = 240 C2v
WJT
acute obtuse
triangle © = 120o triangle
© =180o
C2v C2v
Qa
Fig. 5.22 Adiabatic potential surfaces with conical intersection for a trimer (schematic). When
linear and quadratic vibronic JAHN -T ELLER coupling terms are included one obtains for the lower
potential the somewhat deformed ‘Mexican hat’ with three minima, separated by corresponding
walls. The upper part of the potential just forms a cylinder symmetric cone. WJ T designates the JT
reduction in respect of the position of the conical intersection. The minima correspond to an obtuse,
the maxima to an acute triangle – the saddle points occur if the molecule forms an equal-sided
triangle
tion requires the inclusion of ro-vibronic coupling and the joint treatment of all
degrees of freedom. As evident from Fig. 5.21(c) the D3h symmetry remains con-
served only by the symmetric stretch vibration Qs . However, bending vibration
Qb and asymmetric stretch vibration Qa break this symmetry. The combination
Qb sin φ ± Qa cos φ describes a so called pseudorotation in C2v geometry, as illus-
trated in Fig. 5.22. As a function φ the molecule changes its isosceles triangular
shape between obtuse at the minima and acute at the saddle points of the potential.
The whole motion constitutes a coordinated motion of all three atoms which feigns
a kind of rotation of the whole molecule around its centre (some sort of ‘hula-hoop’
motion which is, however, not a true rotation).9
To obtain a feeling for the orders of magnitude for the case of Li3 : according
to the potential calculation (very close to the experimental fits) in the potential
minimum of D3h symmetry (equilateral triangle) the atomic distances are 5.428a0
in the X 2 E ground state, 5.551a0 in the excited A 2 E . The JT distortion leads
◦ ◦
to acute isosceles triangles with 71.6 and 77.3 , respectively, for the potential
minima in C2v symmetry for the X and A state, respectively (the length of the
equal sides are 5.225a0 and 5.551a0 , respectively). The potential minima there are
WJT / hc = −501.8 cm−1 and −787.2 cm−1 , respectively, measured in respect of the
conical intersection. Finally, the barrier height (saddle point) between the three min-
ima is 72 cm−1 and 156 cm−1 in the X and A states, respectively.
9 We note that the pseudorotation does not occur on a perfect circle. This would only be the case if
the potential minima and the saddlepoints would both lie on the same circle.
324 5 Molecular Spectroscopy
W / cm-1
the pseudorotation angle φ.
The vibrational states
0, 0) in the electronic 80 A 1''
X(0,
0, 0) in
ground state and A(0, E''
the excited state show a
tunnelling splitting 0
(representation in D3h ~
80 X state A '2
symmetry). Shown as dashed
lines are the squared wave E'
functions of the
pseudorotation states (after 0
0 60 120 180 240 300 360
K EIL et al. 2000)
pseudorotational angle ©/ o
The barriers between the minima shown in Fig. 5.22 hinder of course the pseu-
dorotation in the vibronic ground state. The low barrier height leads to tunnelling
splitting quite similar as we have seen it in Sect. 4.2.5 for the inversion vibration
in NH3 . For the specific example of Li3 Fig. 5.23 shows a cut along the pseudoro-
tation coordinate φ through the potential hypersurface Fig. 5.22. It also shows the
energetic positions of the thus resulting pseudorotation states with E and A2 vi-
bronic symmetry in the ground state X(0, 0, 0) as well as in the electronically ex-
0, 0) state. The quantum numbers in brackets represent (here and in the
cited A(0,
following) the vibrational quantum numbers (vs , vb , va ) – see Fig. 5.21(c).
Let us now have a look at some of the results from the impressive studies by
K RÄMER et al. (1999) and K EIL et al. (2000) for Li3 . Figure 5.24 shows the spec-
2 E ← X
tra for the A 2 E transition, recorded according to the scheme RTPI (a).
Here the mass selected isotopologue (7 Li)3 , briefly 21 Li3 , has been detected. The
overview spectrum Fig. 5.24(a) showing several vibrational bands A(v s , vb , va ) ←
0, 0) has been recorded with a pulsed dye laser. The heavily magnified, rota-
X(0,
0, 0) ← X(0,
tionally resolved section in Fig. 5.24(b) (upper part) for the A(0, 0, 0)
band was measured using a tunable, CW single mode dye laser for the excitation
step, together with an Ar-ion laser (also CW) for the ionization step. By using a
quadrupole mass filter, different isotopologues were separated. The modelling of
the RTPI spectra has been achieved with the help of reliable quantum chemical
computations (lower part of Fig. 5.24(b)). The excellent agreement with the ex-
perimental data documents convincingly, that today – by a combination of modern
laser spectroscopic methods, advanced laser systems, and state-of-the-art quantum
chemistry – such a system is fully under control.
In such a demanding evaluation one has, of course, to include that the system
may also rotate as a whole, and that normal rotation and pseudorotation have to
be combined to ro-vibronic states in D3h symmetry. Helpful in this context is the
fact that the system is to 1st order a symmetric top which can be described by the
quantum numbers N and Kc . Nevertheless, a unique identification of the lines is
anything else but trivial. K EIL et al. (2000) thus have used an additional trick, a
further step of spectroscopic sophistication: the optical-optical double resonance
5.5 Laser Spectroscopy 325
(0,0,0) (7Li)3
(a)
(0,1,0)
simulated spectrum
(OODR). The principle of OODR is sketched in Fig. 5.25(a). One uses here two
tunable CW dye lasers with different photon energies. With the pump photon hνpu
a specific initial rotational state is marked by exciting from this level a well defined
but otherwise uninvolved ro-vibronic level of the upper electronic state (here in the
0, 0) vibrational state). When scanning the pump photon one obtains the RTPI
A(1,
spectrum shown in Fig. 5.25(b). The line selected which will be kept fixed during the
OODR experiment is marked by a black arrow. The pump photon beam (having the
selected photon energy hνpu ) is now modulated mechanically at a rate f1 , so that the
population of the thus marked lower rotational level is also modulated, accordingly.
The probe photon, hνpr , too is modulated, albeit with a different frequency f2 . Again
one detects the absorption processes by ionization of the excited state levels, using
a third photon at a fixed energy from an Ar-ion laser. When tuning the probe photon
one may now distinguish the ‘simple’ RTPI spectrum from the OODR spectrum:
the former is modulated with the frequency f2 , the latter with f1 + f2 . It is easy to
separate the two signals by a special electronic device, a so called lock-in-amplifier.
A section of the rotationally resolved A(0, 0, 0) ← X(0, 0, 0) band of 21 Li3 –
recorded as a function of hνpr – is shown for both methods of detection in
326 5 Molecular Spectroscopy
hνpu
Q(3,2,E' )
(a) (c)
R (2,0,E' )
OODR signal
Q (1,1, A''2)
detected at f 1+ f 2 RTPI (b)
Q (1,1,E'' )
R (3,3,E'' )
R (3,2,E' )
Q (6,6,E' )
detecting
R (3,2,A1' )
14835.5 14837.0
laser
P (3,2,E' )
Q (8,8,A'2 )
~
P (3,1, A''1 )
Li3( A ) (Ar-ion)
hνpu RTPI
signal
with f 1 hνpr
marked with f 2
OODR
initial ~
Li3(X )
state hνpr / cm-1
14572.25 14572.30 14575.25 14575.30 14579.00 14579.05
Fig. 5.25 Optical-optical double resonance at 21 Li3 according to according to K EIL et al. (2000).
(a) The experimental scheme. Pump- and probe photon, hνpu and hνpr are modulated with the
frequency f1 and f2 respectively. (b) Low resolution RTPI spectrum for A(1, 0, 0) ← X(0, 0, 0),
recorded by tuning hνpu ; the frequency marked hνpu is kept fixed for the OODR spectrum, shown
0, 0) ← X(0,
in (c) (lower trace) for which hνpr has been tuned through the A(0, 0, 0) transitions.
0, 0) ← X(0,
The RTPI also shown (c) (upper trace) is also recorded by tuning through A(0, 0, 0)
but with hνpu off (see text). The change of the rotational quantum numbers is marked as usual
by P , Q and R, the quantum numbers in round brackets refer to (N , Kc , Γ ) refer to the initial
rotational quantum numbers and to the initial ro-vibronic symmetry Γ . All lines in the RTPI
spectrum are split by HFS. Note the dramatic simplification of the spectrum by OODR
Fig. 5.25(c): as RTPI and as OODR spectrum. The RTPI spectrum still shows the
full complexity of ro-vibronic transitions – even though the Li3 beam is well cooled:
it still contains numerous excited rotational levels. In contrast the OODR spectrum
is dramatically more simple. One only sees now the isolated P , Q and R lines
(corresponding N = −1, 0 and 1, respectively). They all start from a single rota-
tional level, here the N = 3, Kc = 2, Γ = E state. The representation Γ refers
here to the ro-vibronic symmetry of the initial state (to be distinguished from the
electronic as well as from the vibrational symmetry for which the same letters are
used). Instead of using the somewhat problematic half integer quantum numbers for
pseudorotation, K EIL et al. (2000) use solutions of the full Hamiltonian for the JT
problem.
We cannot go into further details of this rather complex problem (and refer to
the detailed presentation in the papers of K RÄMER et al. 1999; K EIL et al. 2000).
We mention, however, that the potentials have been determined in the demanding
MRCI approximation, and the full JT Hamiltonian has been solved in hyperspherical
coordinates, including the interactions from pseudorotation and rotation (C ORIOLIS
coupling).
Finally we mention hyperfine interaction which at that lever of precision be-
comes relevant. The splitting of the rotational lines in the RTPI probe spectrum
Fig. 5.25(c) will not have escaped the attentive reader. 7 Li has a nuclear spin of 3/2,
5.5 Laser Spectroscopy 327
piezo
generator data acquisition
high voltage
Jet III
pump laser L SM SM resonator resonances
on
II
CW ring AOM PD
dye laser off piezo ramp
Fig. 5.26 Scheme of a cavity ring down (CRD) experiment after B IRZA et al. (2002). (a) Core
of the experiment is the FABRY-P ÉROT resonator between two high reflecting, piezo controlled
spherical mirrors (SM). It encloses the plasma (jet) ejected from a pulsed slit nozzle which provides
the molecular ions or radicals to be investigated. (b) Pulse sequences to control the experiment (see
text)
which leads – beyond F ERMI contact interaction (see Sect. 9.2.4, Vol. 1) – with one
unpaired electron spin and three valence electrons in 21 Li3 to a clearly measurable
HFS splitting. In the OODR spectrum it is of course not observed since the pump
photon marks just one specific initial hyperfine level.
by a pulsed supersonic molecular beam with a slit nozzle (3 cm × 200 µm) with
a built in plasma discharge. In the supersonic expansion they cool down to a rota-
tional temperature of 20 K to 40 K. The centrepiece of the setup is the FPI-resonator,
consisting of highly reflecting spherical mirrors (SS) at a distance of 32 cm, with a
reflectivity R = 99.995 %. This resonator encloses the plasmajet in the vacuum. The
photon lifetime in the empty resonator is τe = 27 µs, implying that the light pulse
interact 25 000 times with the target. Together with the slit nozzle one obtains a
remarkable effective total absorption length of ca. 760 m.
Absorptions spectra are recorded with a continuous, tunable dye laser. As pho-
tons may be filled into the resonator only if the photon frequency is in resonance
with the resonator, one has to couple the latter with the resonator of the dye laser
(so called passive mode locking). At a fixed wavelength, the laser beam is then fed
into the FPI by switching the AOM into the ‘on’ status, as indicated in Fig. 5.26(a).
Then the length of the measuring resonator is tuned with a piezo crystal, driven by
a triangular ramp voltage from a piezo generator, marked as curve I in Fig. 5.26(b).
The scanning range of the piezo crystal corresponds to two free spectral ranges of
the FPI, so that for each back and forth movement of the piezo, the resonator be-
comes four times resonant (curve II). Only one of these resonances is used for a real
measurement (marked red in curve II). The resonances are detected in the transmis-
sion signal at the photodiode (PD). At a certain signal height (threshold detector) the
laser beam is coupled out by the AOM (status ‘off’), and one now follows the decay
of the transmitted signal. This procedure is repeated several times. Only during each
second decay cycle the plasma discharge is ignited to measure τe with the absorb-
ing molecule. It is then compared with the resonator lifetime τe (without plasma),
see Fig. 5.26(b) curve III and IV). After 15 measurements each, with and without
plasma, the laser wavelength is changed (autoscan).
M AIER and his group have studied a large number of molecular ions with this
method. Such measurements are of special astrophysical relevance, in particular
those of unsaturated hydrocarbons. On the basis of these laboratory experiments,
a number of spectra observed from interstellar matter have been identified as be-
ing due such molecules. The analysis is not trivial and requires good spectroscopic
knowledge of this type of molecules. Often isotope substitution turns out to be a
helpful tool in such studies.
As a typical example the C6 H+ 4 cation may serve. Its structure is shown in
Fig. 5.27(a) while (b) and (c) show parts of absorption spectra obtained by CRD in
the vicinity of 604 nm (with different resolution). The simulations (dark grey) show
convincing agreement between experiment and theory. C6 H+ 4 is an asymmetric ro-
tor as indicated in Fig. 5.27(b) by displaying Ka and N . It turns out to be a nearly
stretched, symmetric rotor. Only for Ka = 1 one recognizes a slight asymmetry. For
further details we refer to the original papers.
Among the most impressive achievements of modern laser spectroscopy is the struc-
tural elucidation of isolated amino acids, peptides and similar molecules of biolog-
5.5 Laser Spectroscopy 329
H Ka = 4
(a) +C H Ka = 3 (c)
C 6H 4 + Ka = 2
H C C C C C 246
C6H4+
H Ka = 1
1 357
R branch
(b) Q branch Ka = 0
0 2 46
P branch
×
× ×
×
Fig. 5.27 (a) Structure of the C6 H+4 ion which is studied with the cavity ring down (CRD) method
at 604 nm. (b, c) Absorption spectra from the origin band 2 A ← X 2 A . Experiment (red) and
simulation (dark grey). (b) According to B IRZA et al. (2002) with an experimental resolution of
0.15 cm−1 ; peaks marked with × do not originate from C6 H+ 4 ; the simulation agrees best for a
temperature of 40 K. (c) Section of the spectrum (R branch) with improved resolution (0.01 cm−1 )
according to K HOROSHEV et al. (2004); the rotational temperature is now 20 K
ical relevance and a multitude of their compounds and complexes. The interest in
such spectroscopy arises from the possibility to study the intrinsic properties of these
‘building blocks of life’ free of the complex biological environment. This allows –
at least in principle – a reductionistic approach to the construction principles of bi-
ological entities. In such an approach it is e.g. possible to add one by one additional
components of a true biological environment such as single water molecules, and
study the spectroscopic consequences.
The big challenges in this field of research are, for one, to bring these big
molecules without destruction into the gas phase. The second difficulty is to iden-
tify the true structures out of complex variety of possibilities by a rigorous, theory
based evaluation of equally complex spectra. Astonishing progress has been made
in the past two decades, uncovering a wealth of information (summarized e.g. in
R IZZO et al. 2009; DE V RIES and H OBZA 2007). Recent advances may illustrate
the possibilities and problems ahead (see e.g. B ISWAL et al. 2012; WASSERMANN
et al. 2012).
Typically, we are talking about molecules with more than a dozen atoms, which
in addition may form a rather flexible structure: they do not only occur as different
isomers, but also as different conformers. Thus, the observed spectra are usually
very complicated and consist of superpositions of several components which mass
330 5 Molecular Spectroscopy
S0
hνUV hνIR
(probe) (probe)
spectroscopy cannot attribute to specific molecular structures. Smart and often time
consuming detective work is required to obtain a clear, unambiguous picture of the
structure and dynamics of such systems. In addition to elaborate, high resolution
studies with as much selectivity as possible, one also studies different isotopomers
or molecules with different substitutes. In addition, one always needs strong support
by advanced quantum chemical calculations.
For preparation of such species one typically uses ‘seeded’ supersonic molec-
ular beams in which the species to be studied are added to a carrier gas (neon or
argon) and thus cooled down to a few K. More and more MALDI (Matrix assisted
laser desorption ionization TANAKA 2002) and ESI (Electro spray ionization F ENN
2002) based methods are applied, allowing to study also such biomolecules which
cannot easily be vaporized. Detection is done again by LIF and REMPI – usually
supplemented by different double resonance techniques.
Three varieties are schematically illustrated in Fig. 5.28. The basic idea of
optical-optical double resonance spectroscopy has already been introduced in
Sect. 5.5.2 (see Fig. 5.25): there, certain rotational levels in the electronic ground
state were marked. Here one marks specific molecules by fixing one laser (grey
arrow) to a specific excitation energy hνUV1 and thus detects only one particular
geometry and/or isomer either by RTPI with a second photon hνUV2 or by fluo-
rescence hνLIF (black arrows). One then tunes a third laser, the probe photon (red
arrow). Resonance occurs only for transitions within the selected molecular species,
and the ion signal (or the fluorescence) is reduced whenever resonance conditions
are met by the tunable laser (“dip” or “hole burning” spectroscopy).
The transition from the electronic S0 ground state to the first excited state S1
is typically found in the UV range. The scheme according to Fig. 5.28(a) employs
UV-UV double resonance spectroscopy for selection and detection, with photons
hνUV1 and hνUV2 (in suitable cases one may even use the same frequency). The
key of the method is to fix the pump photon hνUV1 (grey arrow) onto a well defined
absorption resonance of the S1 ← S0 transition for one specific isomer or conformer.
5.5 Laser Spectroscopy 331
GG1
UV- UV
GG2
UV- UV
GG
UV- RTPI
Fig. 5.29 UV-UV RTPI hole burning spectra for the base pair guanine-guanine according to
A BO -R IZIQ et al. (2005). The base pair is generated in a supersonic molecular beam and de-
tected by TOF mass spectroscopy. The simple RTPI spectrum obtained by just two UV photons
is marked as GG. The two other spectra, GG1 and GG2, show hole burning spectra for the two
dominant geometrical arrangement of the guanine molecules
The (usually) pulsed probe laser is timed to interact prior to the detection process
with the molecule.
Pioneering work on the spectroscopy of base pairs, the ‘letters’ of DNA, has
been performed by DE V RIES and collaborators. As an example in Fig. 5.29 UV-
UV RTPI spectra are shown for a base pairs consisting of two guanine molecules.
These may assume (among others) two different, nearly planar structures – as one
concludes from the spectra and accompanying quantum chemical calculations. The
comparison of the standard RTPI spectrum in Fig. 5.29 (GG) with the UV-UV hole
burning spectra (GG1 and GG2), determined at different photon energies hνUV1 of
the marking laser, show that both species have different, characteristic UV absorp-
tion spectra. Even if the two double resonance spectra are somewhat more noisy, it
is rather plausible that both added together essentially reproduce the RTPI spectrum
(red, GG).
Complementary to UV-UV double resonance spectroscopy one may also apply
RIDIRS as sketched in Fig. 5.28(b) to this system. Detection is identical to UV-
UV hole burning. However, tuning now an IR probe photon hνIR records the IR
absorption spectrum in the S0 ground state – instead of the electronic absorption
bands for the S1 ← S0 transition in the previous case.
Results are summarized in Fig. 5.30. The recorded spectra of the two species
GG1 and GG2 (the same species as those in Fig. 5.29) show very characteristic
differences. As indicated by arrows in the sketches of the two structures, theory is
able to attribute the bands arising to the different vibrational modes of the two base
pairs (arrows, red, black, dark and light grey), and thus to identify the structure of
these highly complicated, anharmonic and flexible entities. We cannot go any deeper
into the details.
332 5 Molecular Spectroscopy
GG2
GG1
Fig. 5.30 Resonant ion dip infrared spectroscopy for two modifications of the guanine-guanine
base pair according to A BO -R IZIQ et al. (2005). With the help of quantum chemical computations
characteristic vibrations have been identified (sticks spectrum), to which the two IR absorption
spectra can be assigned. The insets show the respective geometry of the two guanine-guanine base
pairs. Full red circles correspond to O atoms, open, black circles indicate N atoms
Fig. 5.31 Rotationally resolved absorption spectra of the S1 ← S0 bands in tryptamine above
the origin (0, 0) according to B ÖHM et al. (2009). The experimental spectra (a), (c), and (e) are
compared with simulations, (b), (d), and (f), respectively. The latter is combined of two species,
(g) and (h), and illustrates the complexity of the problem
S CHMITT 2006). Figure 5.31 gives an impressive example for the potential of such
modern methods. It accounts convincingly for consistency between experiment and
theoretical model for such a large and complex molecule (23 atoms!). Here too, the
interested reader has to be referred to the original sources for more detail.
A variety of other techniques are also employed in laser spectroscopy. Different de-
grees of sophistication are available for the study of electronic, photoinduced transi-
tions in smaller or larger molecules. We just mention photo-fragment spectroscopy
with cations and anions in electromagnetic traps (P ENNING trap, PAUL trap) as well
as matrix isolation spectroscopy. Both methods are particularly well suited for rare
or difficult to generate species. In matrix isolation spectroscopy one prepares the
molecule, radical or cluster in a suitable fashion and deposits it into a rare gas ma-
trix at low temperatures (typically a Ne matrix at 6 K). One may e.g. select a specific
molecular or cluster size by a mass spectrometer and enrich the species in the matrix
by deposition over a sufficiently long time (the thus deposited ions are usually neu-
tralized in the matrix). One may use such a matrix to record more or less standard
absorption spectra. The method thus features a kind of solid state spectroscopy with
a variety of potential complications. Since, however, the interaction of the species
studied with the rare gas matrix is usually very weak, one may obtain in this manner
a good first order impression of unknown absorption spectra for interesting, but rare
classes of molecules. Usually these spectra are only slightly shifted as compared to
the free molecule. For a first ‘screening’ of new types of molecules this method has
proven to be very valuable (s.z.B J OCHNOWITZ and M AIER 2008). Based on such
knowledge one may obtain more detail by LIF or RTPI.
334 5 Molecular Spectroscopy
Section summary
• Laser spectroscopy is a key tool for unravelling molecular structure. A broad
variety of methods is used today. They all exploit the selectivity, high resolu-
tion and sensitivity of modern, (mostly tunable) laser sources.
• The most simple scheme is laser induced fluorescence as described in
Sect. 5.5.1 – where the whole (integrated) fluorescence spectrum is observed
to monitor absorptions spectra as a function of the exciting laser wavelength.
In a more sophisticated version one also analyzes the wavelengths emitted
(DF).
• Multi-photon processes, in particular REMPI or more specific RTPI, are
among the most frequently used and highly selective methods to analyze the
electronic bands of small and larger molecules, as illustrated in Sect. 5.5.2 by
high resolution spectroscopy for the interesting case of triatomic metal clus-
ters. They are subject to JTE, the detailed understanding of which requires
very sophisticated double resonance methods and a high level of quantum
chemical theory.
• CRD, discussed in Sect. 5.5.3, exploits the damping of laser radiation in a high
Q resonator due to absorption within the molecule studied. The method is
highly sensitive and suitable for very low concentrations of target molecules.
• Many variations and sophistications of these and related methods have been
devised for various spectroscopic applications. Perhaps most remarkable is
today’s capacity in resolving and understanding the structure of band spectra
for rather complex molecules, such as amino acids and even small peptides,
the building blocks of life (Sect. 5.5.4).
5.6.1 Introduction
IR absorption
νba = νb – νa = – νab
b
} a
Fig. 5.32 Level scheme explaining the origin of a R AMAN spectrum (JABLONSKY diagram). Grey
arrows correspond to the incident radiation frequency ν0 , red arrows illustrate the frequencies of
the scattered light: ν0 − νj i S TOKES, ν0 R AYLEIGH and ν0 + νj i anti-S TOKES line. The horizontal
red, dashed lines are occasionally considered to represent so called virtual intermediate states.
However, real are only the black energy levels, denoted by i, j and k. For comparison, infrared
absorption (black arrow up) and standard fluorescence (red arrow down) are also indicated
N = N − N = 0, ±2, (5.20)
again with indicating the lower, and the upper levels. In analogy to P , Q and
R branches in vibration-rotation spectra discussed in Sect. 3.4.5 (as well as in elec-
336 5 Molecular Spectroscopy
νba ν- νba
The origin of these three branches is illustrated in Fig. 5.33 for the S TOKES lines.
The differences and analogies to infrared absorption spectroscopy and electronic ab-
sorption and emission bands are evident: while these imply absorption or emission
of a single photon, R AMAN spectroscopy is based on the absorption and emission
of one photon in each step.
We can now discuss a typical R AMAN spectrum for a diatomic molecule, very
schematically sketched in Fig. 5.34. The characteristic vibration-rotation bands with
their branches can be observed red shifted (S TOKES) or blue shifted (anti-S TOKES)
in respect of the incident frequency. They may be compared to the pure IR ab-
sorption spectra discussed in Sect. 3.4.5. For the O branch the rotational quantum
number in the final state is reduced by 2 with regard to the initial state. By consid-
ering the scheme shown in Fig. 5.32, one easily realizes that the S TOKES shift (in
RAYLEIGH
STOKES STOKES anti STOKES
2. harm.
Q ~
~
4B
S O S S Q
O S
Fig. 5.34 R AMAN spectra of a diatomic molecule (schematic): S TOKES and anti-S TOKES vibra-
tional R AMAN, each with S, Q and O branch; the pure rotational R AMAN spectrum left and right
of the (elastic) R AYLEIGH line has only an S branch; on the very left (weaker) a S TOKES spectrum
of the second harmonics (2νba ) is indicated
5.6 R AMAN Spectroscopy 337
respect of the incident line) is smaller for the O branch than for the Q and S branch.
This also holds for the anti-S TOKES spectrum (N in both cases refers to the lower
level); however, the ordering of the branches is just opposite.
Note that (for diatomic molecules) pure rotational R AMAN spectra (Fig. 5.34
middle) have only S branches: for the S TOKES lines (ω < ω) one starts with a
lower lying rotational level N and ends up in N = N + 2, corresponding to an
S transition. The anti-S TOKES lines (ω > ω) arise when the process starts with a
higher lying rotational level N and ends on lower level N = N − 2. According to
(5.21) this transition corresponds again to the S branch.
The popular classical interpretation considers the molecule time dependent in the
oscillatory electric field (ω = 2πν) of the incident light E(t) = E 0 cos(ωt). The
field polarizes the molecule and hence an oscillating electric dipole is induced that
oscillates with the frequency of the light ν = ω/2π :
D el = αE E = αE E 0 cos(ωt). (5.22)
Without the field the molecule oscillates harmonic with angular eigenfrequency
ωba = 2πνba around R0 :
dαE
= αE (R0 )E0 cos(ωt) + R1 E0 cos (ω + ωba )t + cos (ω − ωba )t .
dR R0
Thus, the field induced dipole oscillates at three frequencies, i.e. the wave with
incident frequency ω is modulated by two side bands. This obviously corresponds
just to the three types of scattered lines which we have discussed above:
The first approaches towards a quantum mechanical treatment of the R AMAN ef-
fect go back to Maria G ÖPPERT-M AYER (1931), then a student of Max B ORN.
One of the most commonly quoted references is A LBRECHT (1961), who refers to
B EHRINGER and B RANDMÜLLER (1956), who start with a more general formula
without reference.
Fortunately, the differential R AMAN cross section may readily be derived in 2nd
order perturbation theory: two photons are involved, the incident and the scattered
photon. We briefly sketch the key steps for such a calculation. It essentially follows
that for multi-photon processes as outlined in Chap. 5, Vol. 1; explicitly we have ex-
pounded there two-photon excitation processes, and we expect an expression similar
to (5.46), Vol. 1. However, in the present case of the (spontaneous) R AMAN effect,
only the absorption of the incident photon ω is an induced process while the scat-
tered photon ω is emitted spontaneously. The prefactors will thus be different.
Since we want to understand spontaneous R AMAN scattering, we have to use
now the quantized form of the interaction as presented in Sect. 2.3.6. Basically, we
start with equations (2.140). But we can no longer restrict the treatment to quasi
resonant excitation, so that the full set of coupled ODEs
dck Nk (t) i |iNi ei[(Nk −Ni )ω+ωki ]t
=− ci Ni kNk |U (5.26)
dt
i Ni
has to be solved, with |iNi being a state where the molecule under investigation is
described by quantum numbers i, and Ni gives the number of photons in a mode
of interest. The transition matrix elements are given by (2.135a)–(2.135d). They
contain, quite familiar to us by now, the dipole transition operators D = r · e and
D = r · e for the exciting and the emitted photon of frequency ω and ω , respec-
tively. In principle we also have to account for different modes. But this ansatz is suf-
ficient for the first step (1st order interaction with the incident quasi-monochromatic,
well collimated photon beam of frequency ω). To obtain the 1st order solution, we
assume that on the right hand side of (5.26) only the amplitudes ci Ni for one molec-
ular state |i are finite, all others vanish.
Both processes, absorption and induced emission from this state into any other
atomic state |k have to be considered. Conveniently the matrix elements vanish
unless Nk − Ni = ∓1, respectively. Thus, making explicit use of (2.135a)–(2.135d)
we obtain
eC
Nk + 1 Dki ci Nk +1 ei(ωki −ω)t − Nk
Dki ci Nk −1 ei(ωki +ω)t ,
†
ċk Nk =
5.6 R AMAN Spectroscopy 339
with C = ω/2L3 ε0 being the field normalizing constant in (2.132). Here Nk
1
characterizes the intense laser field, so that Nk Nk ± 1 = N . We thus may drop
the index Nk completely and assume on the right hand side ci = ci Nk ±1 = 1 in 0th
order. After integration the excitation amplitudes in 1st order thus become
i(ωki +ω)t − 1
eC √ ei(ωki −ω)t − 1 † e
ck (t) = N
Dki −
Dki . (5.27)
i(ωki − ω) i(ωki + ω)
In the next step we address spontaneous emission from the thus prepared inter-
mediate state. We have to solve again the ODE (5.26), now for the final state am-
plitudes cf Nf (t) in 2nd order (replacing |kNk → |f Nf and |iNi → |kNk ) by
inserting the 1st order amplitude (5.27) for all intermediate molecular states |k and
all radiation modes. We may treat each radiation mode separately since each emis-
sion process changes the photon number in one mode only. Almost all of the many
intermediate radiation modes are empty (except for the laser mode which we ignore
in this step) so that for all initial modes Nk = 0 and for all final modes Nf = 1
(marked by dashed quantities if necessary). We then have to solve
|k0 = −ieC †
with f 1|U Df k according to (2.135a)–(2.135d). To integrate this we
apply the usual procedure as detailed in Sect. 4.3.5, Vol. 1. For long times, only
stationary exponential terms contribute to the integral. In the present case such situ-
ation arises only for the first exponential term in the square bracket of (5.28), which
in combined form is
†
D Dki
ei(ωki −ω+ω +ωf k )t .
fk
∝ (5.29)
ωki − ω
When integrated it contributes for t → ∞ if and only if ω + ωf k + ωki − ω =
!
ω − ω + ωf i = 0, i.e. for
ω = (ω − ωf i ). (5.30)
This corresponds indeed to the R AMAN condition, with the energy difference
Wf − Wi = ωf i between the final (f ) and initial (i) state of the molecule.10 All
other terms in (5.28) do not contribute to the integral at reasonable experimental
conditions. However, at this point we have to extend our considerations slightly.
10 We emphasize that (5.30) holds for S TOKES and anti-S TOKES lines, since i and f in ωf i refer
here to the initial and final molecular states, respectively. In contrast, in the JABLONSKY diagram
Fig. 5.32 the indices b and a refer to upper and lower R AMAN levels, respectively.
340 5 Molecular Spectroscopy
Characteristic for perturbation theory in 2nd order is the summation over all pos-
sible intermediate states |k, in principle including the continuum. One may visual-
ize this in terms of F EYNMAN type graphs as sketched in Fig. 5.35. Graph (a) cor-
responds precisely to (5.29) as just discussed: an incoming photon ω is absorbed
so that the state |i changes into |k. Somewhat later the photon ω is emitted and
|k becomes |f . Graph (b) corresponds to an additional term which still has to be
added to (5.29): first the photon ω is emitted spontaneously and |i changes into
|k while only in the second step ω is absorbed by |k which is transferred into |f .
Further evaluation in the usual manner leads to the spontaneous R AMAN proba-
bility per mode,
2
2 1
πωe2
dRRAMAN = N . . . δ ωf i − ω + ω πω e , (5.31)
L3 ε0 L3 ε0 2π
k
where we have abbreviated the action of the two terms just mentioned by
D
†
f k Dki †
Df k D ki
= + .
ωki − ω ωki + ω
k k
We may compare (5.31) with (2.146) and (2.148) for absorption and emission, re-
spectively. The terms in square brackets are identical, with N ∝ I (the incident
laser intensity) and considering that for spontaneous emission Nk ≡ 0. The addi-
tional factor 1/2π arises, since only one line shape function is involved here, in the
combined form δ(ωf i − ω + ω ).
In the final evaluation we have to account for the mode densities, normaliza-
tion etc. as detailed with Eqs. (2.149)–(2.153). We then have to integrate over the
laser line (which removes the delta function) to obtain the probability dR̄RAMAN
for scattering into a given solid angle dΩ. This probability is linear in respect of
the intensity of the incident photon source. Finally, the scattered R AMAN intensity
dI = ω dR̄RAMAN is obtained, from which the differential cross section for sponta-
neous R AMAN scattering follows:
†
dσf i dI e4 ω4 D f k Dki † 2
Df k D
= = + ki . (5.32)
dΩ I (4πε0 )2 c4 ωki − ω ωki + ω
k
It has the standard dimension L2 . If one wishes, one may replace the prefactor by
re2 m2e ω4 , where re = α 2 a0 is the classical electron radius and me the electron mass.
We emphasize the proportionality ∝ ω3 characteristic for spontaneous processes
5.6 R AMAN Spectroscopy 341
while the additional factor ω arises from the fact that the cross section is propor-
tional to the scattered intensity dI . Note that this expression for dσf i /dΩ refers to
well defined polarizations e and e for the exciting laser as well as for the sponta-
neously emitted photons, respectively. All quantities are written in SI units.
In the theoretical literature (especially in early work such as A LBRECHT 1961)
one often finds the total scattering intensity averaged over all polarizations and inte-
grated over all angles which implies multiplication of our expression by 8π/9; also
esu are used which implies multiplication by (4πε0 /e2 )2 .
For quantitative evaluation of the dipole matrix elements in (5.32) one follows es-
sentially the procedure described in Sect. 5.4.1. In addition to strict selection rules
of the type given in (5.20), F RANCK -C ONDON factors determine again the scatter-
ing intensities – here between the initial state |i and the intermediate states |k as
well as those between the latter and the final states |f .
At this point, we note another important aspect about Chap. 8 in Vol. 1: the sum
is nearly identical to the polarizability αE (for linear polarization e = e parallel to
the z-axis) according to (8.94) as derived in Chap. 8, Vol. 1. This establishes the
connection to the classical interpretation. Actually, one often starts the treatment
of R AMAN scattering with the polarization tensor and may expand it around the
equilibrium position R0 just as discussed in the semiclassical case, Sect. 3.4.4. This
leads to a quantum mechanical justification of the rule presented there, which reads
in a more general form:
For more details we refer to the literature (e.g. A LBRECHT 1961; C HAN -
DRASEKHARAN and S ILVI 1981, and references there). In practice, explicit eval-
uation of (5.32) or total cross sections can be very elaborate. One usually tries to
find suitable approximations in order to reduce the number of intermediate states
necessary for reasonable estimates of the cross sections.
However, as outlined above, the R AMAN shift ωf i according to (5.30) as such
depends only on the term energies of the initial and final vibration-rotation levels of
the molecule studied. Explicit expressions are particularly transparent for diatomic
molecules. The S TOKES shifts in the three branches are simply the differences of
rotational and vibrational terms F (N) and G(v), according to (3.57) and (3.58),
respectively, (in wavenumbers):
Here B1 and B0 are the rotational constants in the excited and in the initial vi-
brational level, D1 and D0 are the corresponding rotational stretch corrections and
ν̄10 = G(1 ← 0) is the vibrational frequency of the transition according to (3.86).
With these expressions one obtains a good estimate of the structure of the spectra.
Specifically, it becomes clear that the Q branch leads only to very small shifts of in
the rotational R AMAN spectra, since in this case only the (small) differences of the
rotational constants B and of the correction terms D in the two vibrational states in-
volved enter in the spectral position. For larger N , however, one expects a quadratic
growth of the line distances, while in the O and S branches the distances between
neighbouring rotational lines are expected to be constant to 1st order.
Cross sections for R AMAN scattering are notoriously very small. Since for all in-
termediate states |k as a rule ωka = ω holds – for the so called normal R AMAN
spectroscopy one even has ωka
ω – there are no typical resonance denomi-
nators in (5.32). In contrast, these resonance denominators make one-photon spec-
troscopy in absorption as well as in emission so efficient, as described in Chaps. 4
and 5, Vol. 1. In principle, in special cases it may happen that ωka ω (one then
has to introduce a damping term iΓe into the respective denominator in (5.32)). One
occasionally calls such processes “resonance R AMAN scattering”. However, such a
situation may actually be better described as a special variety of optically induced
fluorescence, specifically as laser induced dispersed fluorescence (DF).
While the classical absorption and emission spectroscopy of molecules dates back
to the 19th century and celebrated its big achievements already in the first half of
the 20th century, the extremely successful history of R AMAN spectroscopy – explor-
ing molecular structure and becoming one of the most important tools in analytical
science – is closely related to the discovery of the laser. A number of important
publications soon after the discovery of the effect in 1928 and the N OBEL prize for
R AMAN in 1930 document that the extraordinary potential of R AMAN scattering has
been realized already very early. However, the cross sections are extremely small,
so that efficient highly intensive, quasi-monochromatic and well collimated laser
beams were necessary before efficient use could be made of it. Fortunately, several
fixed frequencies of the lasers in the visible and near UV spectral range were quite
sufficient – already available in the 60th of the past century, e.g. from argon ion
lasers. A typical setup for a Raman spectrometer shows Fig. 5.36 – essentially self
explaining. While the principle of the measurement is quite straight forward, the
big experimental challenge of R AMAN spectroscopy is, (i) to collect the scattered
radiation efficiently, (ii) to separate it carefully from the (elastic) R AYLEIGH scat-
tering which is orders of magnitude more intense, as well as from stray light of the
incident beam which may be reflected at various parts of the spectrometer and (iii)
finally in providing highest spectral resolution. A variety of R AMAN spectrometers
are commercially available today, both for gas phase analysis as well as for solid
state or surface targets – in the latter case even spatially resolved on a µm and sub
5.6 R AMAN Spectroscopy 343
FPI CM
Fig. 5.36 Schematic of a gas phase R AMAN spectrometer. The laser beam from the Ar ion laser
is reflected several times within the target cell (containing the gas to be studied), thus allowing a
sufficiently long interaction region. Very important is the design of the collecting mirrors (CM) for
the R AMAN scattered light and direct it onto the spectrometer entrance slit
µm scale – which is of great practical relevance it the time of nano-, bio-, and similar
technologies. In special cases, R AMAN scattering from solid surfaces may even be
enhanced substantially. This is exploited in the so called “surface enhanced R AMAN
spectroscopy” (SERS).
Modern R AMAN spectrometers with highest resolution use interferometric ana-
lyzing methods (similar to IR spectrometers), in particular with F OURIER transform
spectroscopy (see e.g. C HASE and R ABOLT 1994) and with highly stabilized single
mode argon ion lasers, typically at 488 nm. A resolution on the order of 0.01 cm−1
can be reached – which requires, however, that the collection angle is reduced so
that D OPPLER broadening can be avoided. For details of the various spectrometer
types we refer the readers to specialized original literature and a number of relevant
monographes.
Q branch
N2 6 8
4 10
12
2 14
5 7
9
3 11
13
1
0
Fig. 5.37 Vibration-rotation R AMAN bands for N2 for excitation of the fundamental vibration.
The left spectrum, originally recorded by BARRETT and A DAMS (1968) with the 488 nm line of
an Ar+ ion laser, has been shifted and stretched slightly in order to match recent high resolution
data of B ENDTSEN and R ASMUSSEN (2000). The fully resolved Q branch (shown on the right) is
taken from the latter work. The lower scales give the S TOKES shift of the lines with respect to the
incident wavenumber. Also noted are the initial rotational quantum numbers N of the spectra for
the O, Q and S branches. Due to nuclear spin statistics the intensity for odd N is only half that
of even numbered lines (see Sect. 5.6.6)
We briefly discuss also the vibration-rotation Raman spectrum for the O2 mo-
lecule shown in Fig. 5.38. Here the values from Table 3.6 give ν̄10 = 1556.23 cm−1
corresponding exactly to the limit of the Q branch marked as Q(0) in Fig. 5.38.
5.6 R AMAN Spectroscopy 345
S(N) 70 60 50 40 30 20 10 10 20 30 40 50 60 70 S(N)
R(N) 75 55 35 15 15 35 55 75 R(N)
H
N C
HC N
N C
H
70 60 50 40 30 20 10 0 10 20 30 40 50 60 70
anti-STOKES shift / cm-1 STOKES shift / cm-1
Fig. 5.39 Highly resolved pure rotational R AMAN spectrum of s-triazine, recorded with the
488 nm Ar+ ion laser line according to W EBER (1979)
More recent measurements are available with even higher precision, however, only
in tabulated form. Thus we communicate here only the first R AMAN spectrum for
O2 reported in the literature (BARRETT and A DAMS 1968). As a new specialty,
we note that in the electronic ground state of O2 obviously only rotational states
with odd quantum numbers N are populated – again a consequence of nuclear spin
statistics, to be discussed in the following subsection.
Finally, Fig. 5.39 shows a highly resolved, pure rotation R AMAN spectrum
(v = 0) for the polyatomic molecule s-triazine. As mentioned at the end of
Sect. 5.6.1, pure rotational R AMAN spectra of diatomic molecules show only an
S branch, both on the (N = 2) S TOKES as well as on the anti-S TOKES side. This
is similar for any symmetric and asymmetric rotor. However, due to the possible
change of the rotational projection quantum number K, now in addition on both
sides of the R AYLEIGH line appears also an R branch (N = 1). However, we
cannot indulge her into details.
We have already discussed in Sect. 3.3.3 the influence of nuclear spin on the popu-
lation of rotational levels in diatomic molecules. There, we have become acquainted
with ortho and para hydrogen. The R AMAN spectra of N2 and O2 just discussed
require further attention.
The intensity change of 2:1 between even and odd rotational quantum numbers in
the N2 spectra for even and odd rotational quantum numbers N , as well as the miss-
ing lines for even N in the case of O2 have the same origin. Such intensity changes
may occur in molecules with two identical nuclei and are a consequence of the
PAULI exclusion principle according to which the total wave function of fermions
must be anti-symmetric with respect to exchange of these particles. For bosons, in
contrast, it has to be symmetric.
346 5 Molecular Spectroscopy
One may write the total wave function of a molecule as product of spatial and
spin function with respect to electronic and nuclear coordinates r and R. For the
symmetries discussed here, only the components relating to angular momenta (in-
cluding the spin) need to be considered. We write:
Here φel (r) describes the electronic wave function, (see Sect. 3.6) and ψrot (R) the
nuclear wave function. We now focus on the angular components of these wave
functions. Since we consider only exchange of the nuclei, the electron spin is only
relevant as it defines the electronic wave function and thus the symmetry of the local
environment of the nuclei: χsp (I ) thus refers exclusively to the nuclear spin.
In the following we consider a sequence of symmetry operations which in total
are equivalent to the exchange of the two atomic nuclei. In Fig. 5.40 we indicate
the two nuclei by A and B (in reality they are of course indistinguishable). The elec-
tronic state φel (r̂) is characterized by the ellipse, its symmetry behaviour is indicated
by the red marks. For the complete exchange of the nuclei (exchange operator P A )
A Ψ = ±Ψ
P
1. Ĉ2 : rotation of the whole molecule with regard to the z-axis (perpendicular to the
molecular axis) through 180◦ . We obtain:
The symmetry of the nuclear spin function χsp (I ) = |IIIM follows the usual
rules of angular momentum coupling. We refer to the individual nuclear spins by I ,
to the total nuclear spin by I and its projection is M. Since 0 ≤ I ≤ 2I with
(2I + 1) substates in total (2I + 1)(2I + 1) different nuclear spin states exist, which
are either symmetric or anti-symmetric – as one easily derives for each case.
As illustrated in Fig. 5.40, the total sequence of symmetry operations 1–4 is fully
equivalent to exchange of the two nuclei. Thus, in total:
A Ψ = σ̂h ı̂φel (−1)N ψrot P
P S χsp = ±Ψ.
11 One easily verifies this by looking up the respective C LEBSCH -G ORDAN coefficients.
348 5 Molecular Spectroscopy
16 O Molecule
2
In this case the electronic wave function in the ground state is 3 Σg− for which
σ̂h ı̂φel = −φel . Again we have bosons as the nuclear spin of 16 O2 is I = 0. The
total wave function has to be symmetric. Now, from two nuclear spins I = 0 we
can construct only one symmetric singlet total nuclear spin state with I = 0 can be
constructed. The positive total symmetry can thus only be realized with odd rota-
tional quantum numbers N which compensate the negative symmetry of φel . Hence,
there are only R AMAN lines in the spectrum Fig. 5.38 arising from odd N : in the
electronic ground state of 16 O2 there are simply no other nuclear states providing
the correct symmetry. This nuclear spin statistics has, of course, also consequences
on other properties of O2 , e.g. on the temperature dependence of its heat capacity.
Quite generally one may show (easy to verify for the three cases discussed here),
that the degeneracies gs of symmetric states to ga for anti-symmetric states are
related by
gs I +1
= . (5.41)
ga I
Section summary
• R AMAN spectroscopy of molecular vibration and rotation belongs today to
the most important analytical methods.
• R AMAN scattering is a two-photon process: one photon (energy ω) is ab-
sorbed (induced process), one photon (ω ) is spontaneously emitted. Their
energy difference is the difference of energies between initial and final
vibrational-rotational level (typically within one electronic state).
• Characteristic of a R AMAN spectrum are the S TOKES lines (ν < ν), the anti-
S TOKES lines (ν > ν) and the elastic R AYLEIGH line.
• Strict selection rules hold for the rotational quantum number. In diatomic
molecules it can change by N = 0, ±2, corresponding to Q, S and O
branches, in polyatomic molecules also R and P branches occur. For vibration
propensity rules hold, determined again by F RANCK -C ONDON type consid-
erations.
• A normal mode of a molecule is R AMAN active, if and only if that motion
changes the electronic polarizability.
• In homonuclear, diatomic molecules, nuclear spin statistics has a significant
influence on R AMAN spectra – as discussed for H2 , N2 and O2 (see R AMAN
spectra of the latter two, Figs. 5.37 and 5.38).
12 With the exception of pure emission spectroscopy where no incident light is involved.
5.7 Nonlinear Spectroscopy 349
LIF as well as for the just discussed, spontaneous (also called incoherent) R AMAN
scattering. However, at higher laser intensities, easily reached with today’s laser
sources (see Sect. 8.5 in Vol. 1), one may use nonlinear processes with advantage
for certain applications in molecular spectroscopy.
Generally speaking, nonlinear optics NLO and nonlinear spectroscopy are impor-
tant areas of modern research. We cannot even give a rudimental introduction to it
and refer to various reviews (e.g. W RIGHT et al. 1991; K NIGHT et al. 1990; D RUET
and TARAN 1981) and monographs (e.g. B OYD 2008; S HEN 2003). Here we only
present a few basics and indicate the potential of nonlinear methods in molecular
spectroscopy by way of example.
Nonlinear processes arise from changes of the material under investigation due to
its interaction with the electromagnetic radiation. In turn, the latter is influenced
by these changes. In a perturbative approach, i.e. for not too high intensities, one
describes this by a field induced polarization P:
P = P(1) + PN L = ε0 χ (1) E + χ (2) EE + χ (3) EEE + · · · . (5.42)
This expression thus replaces (8.79) in Vol. 1, where we have introduced the
linear term P = P(1) . In analogy to χ (1) , the (linear) susceptibility, χ (2) and χ (3)
are denoted as nonlinear susceptibilities of 2nd and 3rd order, respectively. The
corresponding terms PN L are called nonlinear polarization. Note, however, that in
the general case (i.e. for a non-isotropic medium) the susceptibilities χ (N ) of N th
order are tensors of rank N + 1.13 Just as in the linear case described by (8.81),
Vol. 1, the absolute magnitude of the polarization is proportional to the particle
density N (i) in the initial state of the molecule studied (note that N (i) must not be
confused with the rotational quantum number Ni of that state).
According to (8.99), Vol. 1, the index of refraction n and the linear susceptibility
are related by χ (1) = n2 − 1 (in isotropic media), with χ (1) being typically on the
order of 1 in solid state materials. At low light intensities, the terms of higher order
in (5.42) are comparatively very small. Unfortunately, there is no general rule for
light intensities at which nonlinear processes might be expected. Each case needs to
be considered individually (see also Chap. 10).
13 They are measured in units [χ (k) ] = mk−1 V−k+1 , i.e. only χ (1) is dimensionless. Note that
(5.42) is an abbreviation. Explicitly, the components of the polarization vector are
(2) (3)
P i = ε0 χij(1) Ej + χij k Ej Ek + χij k Ej Ek E + · · ·
j jk j k
However, to obtain a feeling for the order of magnitude of these quantities B OYD
(1999) used an old argument which we follow.14 One assumes that linear and non-
linear terms in (5.42) should become of equal order of magnitude for electric field
strengths on the order of the atomic field strength (8.141), Vol. 1, EH = e/(4πε0 o02 ),
i.e. at intensities 3 × 1016 W cm−2 . Thus we obtain
Using the polarizability αE for atomic hydrogen according to (8.77), Vol. 1 the
linear susceptibility per atom according to (8.81), Vol. 1 is χa(1) 4π4a03 7.4 ×
10−30 m3 . To determine the particle density N (i) we use the hydrogen VAN DER
WAALS radius (see Fig. 3.3 in Vol. 1) with rH 0.120 nm and assume close-packed
(1)
spheres (74 % filling). Then the linear susceptibility becomes χ (1) = N (i) χa
0.76. Inserting this and EH into (5.43) we obtain χ (2) = 1.5 × 10−12 m V−1 and
χ (3) = 2.9 × 10−24 m2 V−2 . When using the corresponding data for sodium with
(8.82), Vol. 1, all estimates increase by a factor of about six.
Including both, linear and nonlinear polarization, the general wave equation
(1.34) has to be replaced by:
n2 ∂ 2 1 ∂PN L
− 2 2 E(x, y, z, t) = 2 . (5.44)
c ∂t c ε0 ∂t 2
This wave equation describes the propagation of electromagnetic radiation in mat-
ter, and thus must also be applied to devise and interpret spectroscopic experiments.
Due to the terms of higher order in (5.42) one may in principle expect the forma-
tion of sums and differences of all frequency components contained in the incident
electromagnetic wave(s). Hence, (5.44) forms together with (5.42) the basis of all
nonlinear spectroscopy.
The evaluation of these equations for the general case is not trivial. Strictly speak-
ing, the susceptibilities χ (1) , χ (2) and χ (3) are tensors of rank two, three and four
(see footnote 13). In practice, symmetry considerations usually help to simplify the
problem dramatically. For example, χ (2) is different from zero only if the system has
no centre symmetry – optically active crystals show indeed very typical and strong
anisotropies which are used e.g. for generating the second and higher harmonics of
laser light. However, if the target has centrosymmetry (as e.g. a molecular beam or
an isotropic liquid), the χ (2) term drops out completely, and the χ (3) terms usually
become quite transparent.
The nonlinear polarization is in this case PN L = ε0 χ (3) · EEE and describes
so called coherent four wave mixing (CFWM, short FWM) processes: three elec-
tromagnetic waves of different or equal frequencies (ω1 , ω2 , ω3 ) and polariza-
tion (e1 , e2 , e3 ) are incident and generate a time dependent polarization PN L ,
which in turn produces an electromagnetic field of frequency ω4 and polarization
14 Unfortunately he still used the old Gaussian esu, strangely in combination with the unit V. We
e4 – the desired signal which is emitted coherently with the incident radiation. In
the most simple case all incident light waves have the same polarization. Then,
χ (3) (−ω4 , ω1 , ω3 , −ω2 ) is simply a scalar function of the incident frequencies (pos-
itive sign indicates excitation, negative sign denotes stimulated emission), and the
nonlinear polarization becomes PN L = ε0 χ (3) [E(t, z)]3 , with E(t, z) being the sum
of the electric field vectors of the incident radiation.
An ab initio computation of χ (3) contains expressions in analogy to (5.32), how-
ever, in this case in perturbation theory of 3rd order: one has to deal with triple
sums over products of four dipole matrix elements, each between initial and fi-
nal state and the various intermediate states. Each of them is multiplied by three
G REEN’s propagators (resonance denominators) of the type (ω − ωkj + iΓkj )−1 ,
with ω = ω3 ∓ ω2 ± ω1 or ω3 ± ω2 ± ω1 where ωkj is the energy difference of the
intermediate states and Γkj the average value of the natural lifetime of the levels j
and k involved. The resulting sums may contain up to 48 terms for each level in-
volved. Their evaluation is a rather demanding task, for which one uses F EYNMAN
type diagrams according to Y EE et al. (1977), also called B ORDÉ (1983) diagrams
(we have already used them to illustrate spontaneous R AMAN scattering, Fig. 5.35).
For details in the context of the processes of interest here we refer e.g. to W ILLIAMS
et al. (1994, 1995, 1997) and D I T EODORO and M C C ORMACK (1999). A variety of
such FWM processes exist and may in principle be used for spectroscopy.
In Fig. 5.41 a small selection is sketched, along with their respective energy level
schemes. We follow the usual convention with ω1 and ω3 describing absorption,
ω2 and ω4 emission processes. The incident frequencies are ω1 , ω2 and ω3 , while
ω4 refers to the coherent signal generated in the process. Coherent anti-S TOKES
R AMAN scattering (CARS) and coherent S TOKES R AMAN scattering (CSRS) are
the coherent version of R AMAN scattering. They allow for a wealth of interesting,
vibrationally selective applications in spectroscopy, particularly so in surface ana-
lytics. Degenerate four wave mixing (DFWM), i.e. the resonant variety of FWM
where three equal frequencies are mixed to a signal of again the same frequency
may – as absorption or LIF – be used for the study of electronic transitions. Two
colour resonant four wave mixing (TC-RFWM) is in the ‘UP’ version a special kind
of double resonance spectroscopy which we have met already several times in its
linear version. The other diagrams in Fig. 5.41 are more or less self explaining, so
that we refrain here from further discussion.
Generally speaking, nonlinear methods are often useful in situations where linear
methods fail. A typical application is suppression of a diffuse background, e.g. when
studying molecular ions or radicals generated in a plasma, which in itself generates
a lot of stray light.
It is important to note that in all FWM processes sketched in Fig. 5.41, both
energy and momentum conservation for the photons must hold:
ω2 ω4 ω1 ω2
ωSTOKES
pump probe ω3
ω0 ω1 ω3 ω4
e e' e e' e
ω2 ω4 dump
ω4 = ω3
f
pump pump probe
ω1 ω3 ω 1 = ω2 ω 1 = ω2
f ω4 = ω3
i i' i i i'
DFWM TC-RFWM, SEP TC-RFWM, UP
Fig. 5.41 Comparison of COORS (common ordinary old R AMAN scattering) or – in the res-
onant case – LIF, with different CFWM processes. Full red arrows correspond to the incident
radiation, dashed red arrows mark the signal. CARS: ω1 = ω3 (pump and probe, respectively),
ω2 = ω1 − ωvib = S TOKES and ω4 = ω1 + ωvib = anti-S TOKES signal; CSRS: ω1 = ω3 (pump
and probe), ω2 = ω1 + ωvib = anti-S TOKES and ω4 = ω1 − ωvib = S TOKES signal; DFWM: four
equal frequencies, starting from two degenerate states, i and i , ending in two degenerate excited
states e and e ; TC-RFWM with stimulated emission pumping (SEP): ω1 = ω2 (pump), ω3 = ω4
(dump); TC-RFWM with double resonance from the ground state (UP): ω1 = ω2 (pump), ω3 = ω4
(probe)
The last relation with the wave vectors k j (kj = ωj /c = 2π/λj ) ensures the so
called phase matching of the four waves involved. Phase matching is essential for
coherent interaction of the waves to be mixed. It is the key to any nonlinear process
and requires very careful alignment of the four light beams, i.e. a characteristic
geometry as we shall illustrate in the next subsection. A detailed analysis shows
(W ILLIAMS et al. 1997) that the signal intensities observed are given by expressions
of the type
2
I4 ∝ N (i) × L2 × I1 I2 I3 (5.47)
× [Sie ]2 [Sef ]2 × L(ω1 , ω3 ) × G(e4 , e1 , e2 , e3 ; Ni , Ne , Nf ) .
2 2
(This specific expression holds for TC-RFWM.) Here N (i) is the target density in
the initial state, L the length of the interaction region, I1 , I2 and I3 are the three
incident intensities and Sj k the line strengths of the transitions between the states
j and k. As detailed in Appendix H.2, Vol. 1 these line strengths are proportional
to the square of the reduced matrix elements of the dipole operator (H.32), Vol. 1
for the respective transition, Sj k ∝ |j Dk|2 . In the present case they depend
on the vibrational and rotational quantum numbers, Ni , Ne , Nf , involved. The line
5.7 Nonlinear Spectroscopy 353
I1 I2
(a) (b) R
j
jet M dye laser
pump SF
k3 I3 L P BS Nd:YAG
M jet
k4 R
zle R
n oz R pulse
slit R generator
M k1
valve
k2 HV probe RF
PM Nd:YAG
oscillograph computer
PD dye laser
R
Fig. 5.42 Experimental scheme for four wave mixing spectroscopy according to M AZZOTTI et
al. (2008). (a) BOXCARS setup for achieving phase matching at the plasma slit source emerging
from a supersonic molecular beam, with masks (M) for defining the beam positions. (b) Overall
setup with two pulsed, tunable dye laser systems (in DFWM only one is used), a 1000 mm lens,
beam splitters (BS), high reflectivity mirrors (R), a prism (P) and spatial filters (SF), jet setup (with
high voltage, HV) for plasma generation), photo-detector for triggering (PD) of the experiment,
and signal detecting photo-multiplier (PM), as well as the typical electronics for controlling and
detection
profile L(ω1 , ω3 ) accounts for the velocity distribution of the target molecules (i.e.
for D OPPLER broadening) as well as for the natural and collision induced linewidths
of the transitions. Finally, G(e4 , e1 , e2 , e3 ; Ni , Ne , Nf ) describes the influence of
the polarization of the incident and the detected radiation and depends of course
on the angular momenta of the states involved, first and foremost on the rotational
quantum numbers.
Note that the signal – the result of a coherently amplified process – depends on
the square of the target density N (i) and the interaction length L. Observation of
FWM (or higher order) processes always requires the coherent interaction of many
particles. In a single atom or molecule such processes are not relevant (even though
the cross section may depend on a higher power of the radiation intensity). As a
consequence, the signal is emitted in a highly collimated beam of radiation, which
may readily be distinguished form a fluorescent background emitted into a large
solid angle.
5.7.2 An Example
Figure 5.42 shows a typical experimental setup as used by M AZZOTTI et al. (2008),
with a so called BOXCARS arrangement (a) by which the phase matching can be
realized. The slit nozzle plasma discharge is already known to us from Sect. 5.5.3,
where we have discussed its use together with the CRD method. Here one uses
DFWM and TC-RFWM for electronic spectroscopy of radicals. Characteristic is
the beam positioning with the help of a matrix M, which defines the entrance and
354 5 Molecular Spectroscopy
exit angles, as well as the divergence, by apertures – also for the signal to be detected
(red arrow in Fig. 5.42(a)). In this particular setup the crossing angle of the beams
is 1.7◦ , the beam diameter prior to the collection lens (L) ca. 2 mm, which allows
for an overlap length of L 30 mm in the target (L enters the signal quadratically
as mentioned above). The photomultiplier for detecting the signal is positioned at
4 m distance from the jet. Several spatial filters on the way (not shown in Fig. 5.42)
allow a very good separation of signal and stray light from the plasma source.
Figure 5.43 shows for the example of the d 3 Πg − o3 Πu transition in C2 , some
typical spectra thus obtained with DFWM (all four waves have the same frequency
and are jointly tuned).
Without entering into details of the spectroscopy of the C2 molecule, we note
the excellent signal to noise ratio. The rotational progressions in the R, Q, and P
branches are characterized by a large variety of lines. For high rotational quantum
numbers N one recognizes in the R branch a clear triplet structure due to the spin
of the electronic states.
Section summary
• Nonlinear processes in the interaction of matter with radiation fields arise
from changes of the properties of the material studied, due to the radiation
field. The resulting nonlinear polarization depends on powers E 2 , E 3 etc. of
the incident electric field strength, and the re-emitted radiation thus mixes all
incident frequencies, so that one may (at sufficiently high laser intensities) e.g.
generate higher harmonics or various difference frequencies from the incident
radiation.
• For gas phase spectroscopy this may be exploited in sophisticated pump-
probe schemes, e.g. to suppress undirected stray background light. In such
centrosymmetric systems (gas phase) the second order nonlinearity χ (2) van-
ishes and the third order χ (3) leads to a polarization which depends on three
5.8 Photoelectron Spectroscopy 355
monochromator: ħω
(XPS)
liquid jet Wkin (b)
surface
UV (UPS)
ħω
X+ A+B+
synchroton Wkin(max)
AB+
FC region
electrons
electon lens
D'0
ionization
A+B
electron detector X
AB WI WV
D''0
WX v'' N''
v'' = 0
electron spectrometer R
Fig. 5.44 Photoelectron spectroscopy of molecules. (a) Experimental realization, schematic, with
different sources for excitation, various target preparations, electron beam guidance, energy anal-
ysis (spherical capacitor) and electron detection. (b) Energy scheme with adiabatic ionization po-
tential WI , and vertical ionization potential WV . From the measured kinetic energies of the emitted
electrons Wkin and the photon energy ω one derives the energy levels of the ionic states according
to (5.49)
As targets for PES one may use molecules in the gas phase (today preferen-
tially prepared in a cold supersonic beam), but liquids beams may also be used (see
Sect. 5.8.2). The broadest application of UPS and XPS today is, however, in surface
physics.
Photoelectrons are emitted – as discussed in the context of atomic ionization
in Sect. 5.5, Vol. 1 – with a characteristic angular distribution. In the experiment
one typically selects a small fraction of this distribution and may thus determine
(if required) the anisotropy parameter β. We recall the discussion in Vol. 1 on the
angular distribution (5.80) of photoelectrons.15 With γ being the angle between
15 Equation (5.80), Vol. 1 holds for pure, linearly polarized light. If the light is not fully polar-
ized one has to correct for the finite degree of linear polarization P12 of the source according to
(1.101). |P12 | ≤ 1 is usually calibrated by the well known angular distributions from rare gases.
The observed electron angular distribution is then
I (γ ) ∝ 1 + β P12 3 cos2 γ − 1 /2 , (5.48)
(linear) polarization vector and ejected electron, one thus derives the anisotropy
parameter β, usually as a function of photon energy.
Note: If one is only interested in average cross sections, it is advisable to detect
the electron yield at the so called magic angle γ = 54.76◦ – where the β dependent
term in (5.48) vanishes.
In Appendix B some details on the experimental methods are summarized. To-
day electrostatic setups are mostly used. The hemispherical analyzer sketched in
Fig. 5.44(a) and explained in more detail in Appendix B.3 is particularly popular.
As outlined in Appendix B.4, alternatively, time of flight electron spectrometers
are also used. They are of particular advantage when working with pulsed laser
sources. Finally, the electrons have to be detected, which in general is achieved by
secondary electron multipliers (SEM) described in Appendix B.1. Today, often so
called ‘imaging’ methods are used by which behind the electrostatic monochroma-
tors a broad section of the electron energy spectrum may be registered simultane-
ously (VMI). In more sophisticated schemes, angular and energy distributions are
registered together in such imaging device. We have mentioned an example already
in Sect. 5.5.5, Vol. 1.
Figure 5.44(b) illustrates for the most simple case of a diatomic molecule the
characteristic energetics of a UPS process. Energy conservation requires
ω + AB γ v N → AB+ γ v N + e− (Wkin ), (5.49)
where the quantum numbers γ , v and N refer as usual to the electronic state, to
vibration and rotation, respectively. We assume here that γ and γ correspond to
the neutral electronic ground state X and to its cation ground state X + , respectively.
The kinetic energy of the electrons Wkin (to be determined experimentally) results
from the ionizing photon energy ω, the adiabatic ionization potential WI of the
molecule with respect to the lowest vibration-rotation level of γ , the total energy
Wγ v N of the initial state16 and the excitation energy Wγ v N , in which the ion
remains after the process:
Wkin = ω − WI + Wγ v N − Wγ v N . (5.50)
The schematic energy diagram Fig. 5.44(b) assumes v = 0 in the electronic ground
state γ = X (broad arrow upward), typically with several occupied rotational levels
N . The ion in this scheme is also generated in its electronic ground state γ = X + .
For the relative position of the potentials as indicated, ionization leads to a distribu-
tion of several vibration-rotation states. They are populated in a similar fashion as
in electronic excitation processes by dipole transitions (see Sect. 5.4.1). Again we
expect that F RANCK -C ONDON (FC) factors (between initial ground state and final
ionic state) play a key role for the respective ionization cross sections. Often the
which obviously differs from the true ionization potential WI . Its negative value,
WBV = −WV , is often called vertical binding energy. These quantities are, however,
defined only somewhat loosely as one recognizes from Fig. 5.44(b), simply because
the FC region has no sharp limit.
Over all, photoelectron spectra for ionization, correspond essentially to absorp-
tion spectra for excitation – except that now the photon energy ω does not need
to be tuned: the energy balance is automatically taken care of by the kinetic energy
Wkin of the electron emitted according to (5.50) which also contains the spectro-
scopic information.
Of course it is possible at high photon energies, that ionization occurs into differ-
ent final electronic states of the ion. Just as in absorption spectroscopy, the spectra
observed may thus become quite complicated. In addition we have to remember that
in PES different electrons of the target may be ionized. In UPS studies different va-
lence electrons with different binding energy may be ejected – each of them may
lead to its own vibration-rotation bands. In XPS one usually focusses on ionization
of electrons from inner shells: these may show characteristic energy shifts according
to their different chemical environment. This has evolved to be a very valuable ana-
lytical tool. As exemplified below, this so called chemical shift can be exploited for
chemical analysis of this very environment. According to S IEGBAHN this method is
called electron spectroscopy for chemical analysis (ESCA).
5.8.2 Examples
We want to discuss here only a few, characteristic examples which illustrate the
potential but also the limitations of PES. We start with the H2 O molecule with which
we are already well familiar. The photoelectron spectrum of H2 O in the gas phase
has been studied with synchrotron radiation for the first time by T RUESDALE et al.
(1982), and somewhat later by BANNA et al. (1986) over a larger energy interval,
30 eV ≤ ω ≤ 100 eV.
As discussed in Sect. 4.4.1 the electron configuration of the electronic ground
state of H2 O is (1a1 )2 (2a1 )2 (1b2 )2 (3a1 )2 (1b1 )2 . With photon energies up to 100 eV
only electrons from the four valence orbitals may be ionized, as one reads from
Fig. 4.22. One thus detects according to the scheme in Fig. 5.44(b) ionic states (dou-
blets) – each with one hole in one of the valence shells. The respective ionization
processes may be described as:
H2 O + ω → H2 O+ 1b1−1 2 B1 , v + e− (Wkin ) (5.52)
5.8 Photoelectron Spectroscopy 359
1b1 1b1
(b) ×5 (a)
3a1 gas phase gas phase
1b2
electron signal / arb. un.
1b1
(c)
1b2 3a1
liquid phase 2a1
Fig. 5.45 Photoelectron spectrum of the valence electrons of water. (a) H2 O in the gas phase at
ω = 100 eV (b) with improved resolution according to G ODEHUSEN (2004); (c) measured with
a liquid jet, i.e. at the surface of the liquid, at 60 eV according to W INTER et al. (2004). Clearly
visible are the shifts (indicated by dashed lines) of the absorption maxima due to solvation energy
in the liquid
H2 O + ω → H2 O+ 3a1−1 2 A1 , v + e− (Wkin ) (5.53)
H2 O + ω → H2 O+ 1b2−1 2 B1 , v + e− (Wkin ) (5.54)
H2 O + ω → H2 O+ 2a1−1 2 B2 , v + e− (Wkin ). (5.55)
Figure 5.45(a) (and Fig. 5.45(b) with high resolution) shows characteristic photo-
electron spectra from a recent measurement of G ODEHUSEN (2004) at ω = 100 eV
with high resolution. The four main structures correspond without doubt to the pro-
cesses (5.52)–(5.55) (for simplicity we simply use the initial orbitals for charac-
terization). One usually plots the signal as a function of “binding energy” WB =
Wkin −ω. One has to be aware, however, that in the ion too vibrational states v may
be excited, the real binding energy should thus be determined according to (5.50)
from −Wγ v N . In the present case such vibrational excitation obviously occurs: one
clearly recognizes, specifically in the high resolution spectrum Fig. 5.45(b), clear vi-
brational structures within the main peaks 1b1 , 1a2 and 1b2 (we assume here that
H2 O initially was prepared mainly in the vibrational ground state). Figure 5.45(a)
and (b) illustrate also the limits of photoelectron spectroscopy. The typical band-
width of the electron monochromators is a few meV at best, which allows usually
for an analysis of vibrational structures, but excludes in most cases to resolve rota-
tional structures.
For comparison, in addition to the isolated H2 O molecule, Fig. 5.45(c) also shows
photoelectron spectra for water at the liquid surface according to W INTER et al.
(2004). One may record such spectra today with a very thin water beam (some
µm diameter) with well focussed synchrotron radiation. In the liquid each water
molecule may be thought to have its own solvation shell. The high relative permit-
360 5 Molecular Spectroscopy
17 We have to use the dynamic relative permittivity (dielectric constant) here, which is much smaller
experiment
(a)
sticks spectrum:
NH2 OVGF/6-311++G**
N
O 12 11
H N
7
10 8
9 5 1 = 21a(π5 )
2
6 4 31 2 = 20a(π4 )
3 = 19a(σN )
4 = 18a(σN )
photo electron simulation
theory: ADC(3)/6-31G
5 = 17a(π3 )
(b) 6 = 16a(π2 )
7 = 15a(σO)
8 = 14a(σ)
9 = 13a(π1 )
7 10 = 12a(σ)
11 8
14 11 = 11a(σ)
15 12 10 12 = 10a(σ)
+ 9 13 = 9a(σ)
16 13 5 2
3 1 14 = 8a(σ)
6 4
15 = 7a(σ)
-22 -20 -18 -16 -14 -12 -10 -8 16 = 6a(σ)
electron binding energy / eV
Fig. 5.46 Photoelectron spectrum of cytosine (2b) in the gas phase according to T ROFIMOV et al.
(2006). (a) Experiment and OVGS sticks spectrum, (b) ADC(3) sticks spectrum and convolution
with experimental line profile
T RAVNIKOVA et al. (2012) have recently had a fresh look at the C 1s elec-
tron spectrum of the ESCA molecule in the gas phase, using a photon energy of
340 eV at the new French third generation synchrotron SOLEIL, with high reso-
lution (ca. 50 meV FWHM: photon source 36 meV, electron monochromator ca.
20 meV). The photoelectron yield is plotted in Fig. 5.47 as a function the chemical
shift Wchem = WB − WB (CH3 ) with respect to the binding energy WB (CH3 ) of
the 1s electron at the CH3 group. Surprisingly, the four clearly resolved peaks are
still significantly broader than the 50 meV FWHM of the apparatus. They also show
a pronounced asymmetry – but no structure is resolved. T RAVNIKOVA et al. (2012)
have carried out state-of-the-art structure calculations (using G AUSSIAN) for the
four peaks exploring two conformers of the ESCA molecule: the anti-anti structure
shown in Fig. 5.47 (which has Cs symmetry) and the anti-gauche structure (C1 , i.e.
no symmetry) which is obtained from the latter by changing the dihedral angle of the
COCC planes from 180◦ to 80◦ (not shown here). The ratio of the two conformers
was assumed to be 44:56, based on prior ab initio calculations. Including initial and
final state effects, a proper F RANCK -C ONDON analysis of the ionization process,
and accounting for the finite lifetime of the core holes, they obtain a nearly perfect
fit to the data in Fig. 5.47. The asymmetry of the peaks is attributed to post-collision
effects (i.e. to electronic and nuclear rearrangement in the ion, including dissocia-
tion). It should also be noted that the ratio of the (integrated) four peaks is about
1:0.91:0.90:0.78 (and not 1:1:1:1 as one would expect from their stoichiometric re-
lation). This effect is attributed to multiple scattering processes of the ejected pho-
toelectron. We mention at this point (without going into details) that such effects
are exploited with advantage for structural analysis of solid state targets in X-ray
absorption spectroscopy by NEXAFS and EXAFS.
XPS (alias ESCA) in a variety of forms with different acronyms has been devel-
oped over the past decades to a very efficient and robust method for the chemical
analysis of surfaces, coatings, and thin films (a nice survey has been given by R EIN -
ERT and H ÜFNER 2005). It may also be used in position sensitive versions, e.g. in
connection with X-ray microscopy, based on synchrotron radiation. In addition to
C 1s edge spectroscopy (first and foremost relevant for organic materials) K edges
of O, N, S and other characteristic atom may be used to analyze their chemical
environment. Even K and occasionally L edges of metals are often used for a quan-
titative analysis of surfaces and depth profiles of thin layers (in the latter case in
connection with sputtering methods for surface etching). State-of-the-art equipment
for such analysis are today available on a commercial basis in various designs (see
e.g. K ELLY 2004).
If one compares the photoelectron spectra shown above with optical spectra which
we have seen in previous sections of this chapter, one notices the significantly poorer
energy resolution of PES – essentially a consequence of the fundamentally different
properties of electrons and photons: no electron monochromator can ever reach the
5.8 Photoelectron Spectroscopy 363
electron detectuib
(channel plates)
field free cted) transmission = 100 %
ħω - ot dete
fast e (n
signal
detection cone
1 2 3
123
TOF
μ metal shielding slow (near ZEKE)
V<0 pulsed
photoelectrons (low transmission)
extraction field
Fig. 5.48 Principle of detecting ZEKE photoelectrons and discriminating against nearly ZEKE
electrons according to M ÜLLER -D ETHLEFS and S CHLAG (1991). In the example shown an ex-
traction field is applied to the originally field free interaction region after 1 µs. Electrons with an
energy of only 0.1 meV (v = 6 mm/µs) have then travelled already on a sphere of 6 mm diameter
(the centre of this sphere being the ionization volume); a few of them fly paraxially and are de-
tected in the TOF spectrum on positions 1 and 3 and may thus be discriminated easily against true
ZEKE electrons detected at 2. Non-paraxial nearly ZEKE electrons cannot reach the detector at
all, since their perpendicular velocity component is too large (steradiancy discrimination)
fragment ions after dissociative ionization of the Na2 as a function of photon en-
ergy ω.
As mass analyzed threshold ionization (MATI) spectroscopy this has also been
proposed as an alternative to ZEKE photoelectron spectroscopy (see Z HU and
J OHNSON 1991; L EMBACH and B RUTSCHY 1996). Figure 5.49 compares for the
example of a two-photon ionization process of pyrazine (a) the total ion signal from
direct photoionization with (b) the corresponding MATI and (c) ZEKE spectra.
Detailed experimental studies and theoretical considerations have shown that the
astonishingly high detection sensitivity of ZEKE and MATI methods are based on
the fact that photo-absorption does not directly lead into the continuum, but rather
to long-lived RYDBERG states very close to the ionization threshold. These are then
ionized by the pulsed electric field. ZEKE spectroscopy has been developed into an
extraordinary powerful method (M ÜLLER -D ETHLEFS and S CHLAG 1998), which is
used in a variety of version for many applications, and which today is even available
on a commercial basis.
So far we have only treated ionization of neutral, isolated molecules, radicals and
clusters for which PES provides direct access to the electron binding energies, and
allows at the same time to derive spectroscopic information on electronic and nu-
clear structure of the ions created. Of course, this method may also be used for neg-
ative ions. Typically, anions can only be prepared with very low concentration and
their spectroscopy often relies exclusively on PES. Usually the weakly bound outer
electron may readily be detached with typical laser sources. The reaction (5.49) has
5.8 Photoelectron Spectroscopy 365
now to be written as
ω + AB− γ v N → AB γ v N + e− (Wkin ), (5.56)
and the ionization potential WI in (5.51) has to be replaced by the electron affin-
ity WEA of the neutral molecule or cluster studied. PES with anions (electron
photo-detachment spectroscopy) is an active field of research for now at least 40
years and literature about it is extensive. The field has essentially been shaped by
W.C. L INEBERGER and his many students and collaborators (see e.g. L INEBERGER
and W OODWARD 1970; H OTOP and L INEBERGER 1985; E RVIN and L INEBERGER
1992; N EUMARK 2001, 2002; R IENSTRA -K IRACOFE et al. 2002; E LLIOTT et al.
2008; S HEPS et al. 2009) and a few other groups worldwide (e.g. C HA et al. 1992;
YANG et al. 1987; L EE et al. 1991; TAYLOR et al. 1992; M ARKOVICH et al. 1994;
C ASTLEMAN and B OWEN 1996; W RIGGE et al. 2003, and further references there).
The spectroscopic methods differ only little from those for neutral atoms, molecules
and clusters.18 Imaging detection methods (VMI) too are very popular these days in
18 There is, however, one key difference: anions do not come out of the bottle and have to be specif-
ically prepared, and – as they carry a charge – may also be mass selected prior to the interaction
with photons. Specifically for the study of clusters this is an essential advantage compared to neu-
tral cluster beams. They usually have a broad distribution of cluster sizes, and mass selective de-
tection after the interaction process does not really help, since usually very difficult to discriminate
against fragments from larger clusters.
366 5 Molecular Spectroscopy
anion spectroscopy (e.g. E LLIOTT et al. 2008) and a number of acronyms have been
coined to characterize various techniques.
For all the above discussed methods of PES one tacitly makes two essential assump-
tions: 1. The molecule (or cluster) studied remains intact – apart from the electron
loss due to the ionization process (5.49) – or potential dissociation processes can
be identified without doubt as being correlated with the observed electron. 2. One
deals with one well known species of target molecules or may distinguish different
species by different positions of the absorption bands (the kind of complications
that arise when this is not the case have been indicated for the “ESCA” molecule on
p. 361 f.).
If one of these assumptions is not satisfied the electrons detected cannot be cor-
related to a specific ion, and consequently the photoelectron spectrum observed can
provide only limited insight. At sufficiently high photon energy ω, even for the
most simple case – a diatomic molecule – either a stable (vibrationally) excited
parent-ion is generated as indicated in Fig. 5.44(b).
Alternatively, the molecule might fragment during ionization if one assumes the
potential minimum position for the AB+ ion to be shifted to a slightly larger R so
that the FC region would extend above the dissociation limit (WI + D0 ) of the AB+
parent-ion, as sketched in Fig. 5.44. This so called dissociative ionization is a rather
common process for larger molecules. Already for a triatomic molecule one expects
at higher ω several open channels for such processes:
ω + ABC → ABC+ γ v N + e− (Welkin ) (5.57)
→ AB + C+ (Wionkin ) + e− (Welkin )
→ A + BC+ (Wionkin ) + e− (Welkin )
→ AB+ + C(Wionkin ) + e− (Welkin ), . . . etc.
Figure 5.50 illustrates the first and the second channel, here as a cut along one nu-
clear coordinate through the potential hypersurface of the triatomic ABC+ ion. We
have to distinguish between the kinetic energy of the electron Welkin and the kinetic
energy Wionkin of the fragments. The latter refers of course to the total relative en-
ergy in the centre of mass system AB + C+ of the dissociating molecular ion. The
balance of energy (5.50) has to be extended slightly for such a process:
Welkin + Wionkin = ω − (WI − Wγ v N ) − Wγ v N . (5.58)
(
In the case of polyatomic molecules or clusters, Wγ v N is the sum of all internal
energies of all molecular fragments (neutral and/or ionic) – in the example (5.57)
these would be the vibration-rotation energies in ABC+ , or in AB or BC+ etc.,
respectively, depending on the channel studied. An unambiguous spectroscopy of
5.8 Photoelectron Spectroscopy 367
W R)
W(
e-
ħω
Welkin(AB+C+)
AB+C+
Welkin (ABC+) +Wionkin
FC region
v'N'
ABC+ D0'
~
X+
RAB-C
Fig. 5.50 Dissociative ionization of a triatomic molecule. The F RANCK -C ONDON (FC) region
(heavy, black arrow) is defined by the potential minimum of the ground state (not shown here) and
extends in the present case over bound states of the ABC+ ions up to the dissociative limit D0
(electron energy Welkin (ABC+ )). But it is also sufficiently high to reach into the dissociation con-
tinuum of the open channel AB + C+ ; in that case, the C+ ion and the neutral fragment AB separate
with a relative kinetic energy Wionkin , while the electron carries a kinetic energy Welkin (AB + C+ )
such a process is possible only if the ejected electron e− can be associated uniquely
with a molecular ion ABC+ or fragment C+ or BC+ etc., respectively. If several
channels are open, this can be achieved by coincident detection of ions and elec-
trons – if possible with a full analysis of all energies and momenta (directions) of
fragments and electrons. This is, of course, a big challenge which only in a few
special cases has been achieved completely. Over the years, however, great progress
has been made towards this goal.
Since more than 40 years, energy selected ions are prepared and analyzed by
photoelectron-photoion coincidence spectroscopy, PEPICO (see e.g. B REHM and
VON P UTTKAMER 1967; E LAND 1972; DANBY and E LAND 1972; W ERNER and
BAER 1975; JARVIS et al. 1999; S ZTARAY and BAER 2003; BAER et al. 2005;
E LAND 2009). The basic concept, sketched in Fig. 5.51(a), has remained essen-
tially the same as first used in the 1970s. However, modern time and position sen-
sitive detection methods with fast electronics, sophisticated electron and ion optics
and imaging methods as indicated in Fig. 5.51(b) have brought us rather close to
the ideal of a state selective analysis of photoionization and fragmentation dynam-
ics for some selected model systems. To understand the principle, one has to realize
that these methods have been devised originally for rather weak light sources (gas
discharge lamps with monochromators), and individual dissociative processes were
detected with a very low count rate. Thus, each photoelectron detected can in prin-
ciple be correlated to one specific ion. Due to its much lower velocity the latter is
observed with a time delay in respect of the electron. The electron triggers, as in-
368 5 Molecular Spectroscopy
(a) ħω acceleration
ion energy
Ionen Masse
e- energy and/or mass
bzw. Energie MCP for ions
e- MCP analyser analyser
Analsator
extraction
detecting field detecting
electronics electronics
ion TAC or
pulser TDC stop
start
variable PC
delay line (MCA, dig. osci.)
optional
(b) ħω 40 V/cm deceleration
Roentdek hot e- ions
MCP for ions
DLD40
ZEKE e-
split plate
Fig. 5.51 (a) Principle of PEPICO or TPEPICO experiments – in the spirit of E LAND (1972)
and BAER (1979). Today, electrons and ions are usually detected by multi channel plates (MCP).
The electron signal triggers the ion extraction and the timing (start). In earlier experiments, the
difference between the flight times of the electron and the ion (stop) have measured by time–
to-pulse-height conversion (TAC) with subsequent multichannel analyzers (MCA); today one dig-
italizes this time delay directly (TDC) and records it in a PC. (b) Modern version of electron (left)
and ion optics (right) according to B ODI et al. (2009). The trajectories for ‘hot’ (pink) and threshold
electrons (ZEKE, red) are separated by spatially resolved detection (RoentDek DLD40)
dicated in Fig. 5.51(a), a coincidence electronics which measures the time until the
detection of an ion.
In the most simple case one lets the electrons drift out of the interaction region
with the photon ω in a very low electric field and thus limits the acceptance angle
as already discussed in Sect. 5.8.3. Such a “steradiancy” analyzer selects preferen-
tially very slow electrons (TPES) – emitted when the photon energy is just sufficient
to excite a specific state of the ion. The electron signal then triggers a pulsed elec-
tric extraction field for the ions which were essentially at rest until this moment.
The delay until detection identifies the mass of the ion which has been generated in
this particular ionization event – with suitable ion optics one may even record the
kinetic energy of the ion. Each event is registered in the memory of the PC accord-
ing to its delay time. Subsequent events are added in the storage corresponding to
their time delay until sufficiently good statistics has been reached. The whole proce-
dure is repeated for each photon energy of interest, preferentially in rapid iteration
to compensate for experimental fluctuations. In this manner a so called threshold
photoelectron-photoion coincidence (TPEPICO) spectrum is built up. A wealth of
5.8 Photoelectron Spectroscopy 369
results documents the efficiency of the TPEPICO method (see e.g. the review of
BAER et al. 2005, and further references there).
We mention also the fact that TPEPICO has also been used successfully in cases
where – albeit no fragmentation was expected – the target contained several species
which all could be ionized at a given photon energy. Specifically this is the case
for neutral cluster beams where typically a broad mass distribution is created. In
that manner during the 1980s and 1990s numerous ionization potentials of neutral
atomic (Arn , Krn , K AMKE et al. 1989) and molecular clusters ((NH3 )n , (N2 O)n ,
K AMKE et al. 1988; G REER et al. 1990) have been determined using TPEPICO, to
mention just a few examples.
Over the years the methods have been improved systematically. While initially
the time of flight was measured by time to amplitude converters (TAC) with sub-
sequent recording in multi channel analyzers,19 in the mean time one uses direct,
digital recording of the time difference between electron and ion signal in TDCs
(time to digital converters) or digital oscillographs. Modern devices also register
several subsequent hits (multi hit) or may start again during one measuring pro-
cess (multi start) – an overall significant improvement in flexibility and reduction
of dead-time losses (B ODI et al. 2007). This is particularly important for investi-
gations with intense light sources, such a synchrotron radiation and laser systems,
which allow high count rates.
As indicated in Fig. 5.51(b), recent developments combine PEPICO methods
with state-of-the-art imaging techniques (the acronym iPEPICO has been coined).
In this setup the electron extraction optics is designed such that the slow threshold
electrons (ZEKE) hit the position sensitive detector in the middle, while faster elec-
trons are detected exclusively on rings around the axis and are thus energy analyzed.
The high resolution and the ability to discriminate different masses is documented
in Fig. 5.52 for the example of the Ar2 dimer molecule (TPES). The resolution is
19 In these storage devices, originally designed for nuclear physics experiments, events are added
on the order of a few meV. The grey line in the upper half of the graph (signal
without coincidence, i.e. from all Ar+ n cluster ions in the molecular beam) differs
in this case only very little from the black line, the pure electron signal correlated
with Ar+ 2 . Obviously the ionization cross section for higher clusters e.g. Ar3 , is very
small in this energy range. Generally speaking, the possibility to correlate electron
spectra to different fragment masses is a key feature of this kind of setup – but ob-
viously very demanding and time consuming, considering the overall output from
these sophisticated sources making full use of their potential.
In contrast to the determination of ionization thresholds with TPEPICO, the more
general PEPICO method requires a true electron energy analyzer for arbitrary en-
ergies. In earlier years, electrostatic fields were used for this purpose. However,
the imaging techniques just described, with sophisticated electron optics, provide
an interesting alternative (see also the summary on the reaction microscope in Ap-
pendix B.4).
On the other hand, today one often uses pulsed light sources which are a pri-
ori ideally suited for time of flight methods to determine the electron energy. One
must, however, keep the number of random coincidences sufficiently low. Hence,
for a long time it appeared nearly impossible to perform PEPICO experiments with
pulsed laser sources: as a rule, each laser pulse generates many ions and the assign-
ment of the electrons detected to specific ions appeared impossible. For the first time
S TERT et al. (1997, 1999) have overcome this problem by working at extremely low
ion yields (
1 per laser pulse). They used a two colour pump-probe scheme with
strongly attenuated, high repetition femtosecond laser pulses to ionize molecules
and clusters via a resonant intermediate state. This femtosecond time resolved elec-
tron ion coincidence (FEICO) method exploits a magnetic bottle (see Sect. B.4) by
which electrons may be collected from a nearly 4π emission angle, while their en-
ergy is measured by electron TOF. Here too the detected electrons trigger a fast high
voltage pulse to extract the ions. The method has opened for the first time the pos-
sibility to study dynamical process on a femtosecond time scale in photo-excited,
neutral molecules and clusters by TOF in a mass selective manner. As an example
we mention ammonia clusters (FARMANARA et al. 1999), which show an interest-
ing fragmentation scheme. In a mass spectrum they are predominantly observed as
protonated species:
ωpu (NH3 )n (Ã) ωpr
(NH3 )n (X̃) −→ (NH3 )n (Ã) → −→ · · · (5.59)
(NH3 )n−1 H(Ã) + NH2
or
ωpr (NH3 )+
(NH3 )n (Ã) −→ e− (Welkin ) + (NH3 )+
n →
n (5.60)
(NH3 )n−1 H+ + NH2 .
We cannot enter here into the details of the observed dynamics. However, Fig. 5.53
gives a brief survey for one fixed delay time (t = 0 fs) between pump (ωpu ) and
probe pulse (ωpu ). Shown here are the raw data and some key results. They docu-
ment that one can indeed observe mass selective photoelectron spectra from neutral
molecular clusters using femtosecond pulsed lasers.
5.8 Photoelectron Spectroscopy 371
(NH3)2NH4+
(NH3)3NH4+
NH3NH4+
NH3NH4+
(a)
(NH3)2NH4+
(NH3)+
electron time of flight / μs
4 raw data
NH4+
0
Welkin(max)
2
1
AP/eV 9.54
NH4+
2
0
10 20 30 40 normalized e- signal
Welkin / eV
ion time of flight / μs
(c)
(b) electron spectra
mass spectrum
ion signal
n=2 3 4 5 6 7
n=
1 2 3 4 5
M/u
20 40 60 80 100
Fig. 5.53 Electron ion coincidence (PEPICO) for the resonant two-photon, two colour ionization
of (NH3 )n clusters in a FEICO experiment. According to (5.59) or (5.60) protonated (red) and un-
protonated (pink) cluster ions may be generated. (a) Raw data: each registered delayed coincidence
is represented by one point as a function of electron and ion time of flight. In vertical direction the
electron energy varies, in horizontal direction the observed mass M. (b) Mass spectra are obtained
by projection of all these data onto the TOF axis of the ions. The values of n refer to the size of
the respective neutral parent cluster. (c) Electron spectra associated with a given ion are obtained
by projection of the raw data for a particular ion onto the electron TOF axis. Of course, finally the
TOFs and signal heights have to be converted, in the former case onto a mass scale M/u, in the
latter case onto a the electron energy scale Welkin . The grey, horizontal arrows indicate in each case
Welkin (max), the maximum electron kinetic energy observed. It corresponds the grey numbers in
called appearance potential, AP = ω − Welkin (max)
Section summary
• The kinetic energy of photoelectrons Wkin depends on the energy of the
ionizing photon and on the internal energies of the initial neutral and final
ionic state according to (5.50). This relation forms the basis of photoelectron
372 5 Molecular Spectroscopy
References
B ERSUKER , I. B.: 2001. ‘Modern aspects of the Jahn-Teller effect theory and applications to
molecular problems’. Chem. Rev., 101, 1067–1114.
B IRZA , P., T. M OTYLEWSKI, D. K HOROSHEV, A. C HIROKOLAVA H. L INNARTZ and J. P.
M AIER: 2002. ‘CW cavity ring down spectroscopy in a pulsed planar plasma expansion’. Chem.
Phys., 283, 119–124.
B ISWAL , H. S., E. G LOAGUEN, Y. L OQUAIS, B. TARDIVEL and M. M ONS: 2012. ‘Strength
of (NHS)-S-. . . hydrogen bonds in methionine residues revealed by gas-phase IR/UV spec-
troscopy’. J. Phys. Chem. Lett., 3, 755–759.
B LOEMBERGEN , N. and A. L. S HAWLOW: 1981. ‘The N OBEL prize in physics “for their con-
tribution to the development of laser spectroscopy” ’, Stockholm. http://nobelprize.org/nobel_
prizes/physics/laureates/1981/.
B ODI , A., B. S ZTARAY, T. BAER, M. J OHNSON and T. G ERBER: 2007. ‘Data acquisition schemes
for continuous two-particle time-of-flight coincidence experiments’. Rev. Sci. Instrum., 78,
084102.
B ODI , A., M. J OHNSON, T. G ERBER, Z. G ENGELICZKI, B. S ZTARAY and T. BAER: 2009. ‘Imag-
ing photoelectron photoion coincidence spectroscopy with velocity focusing electron optics’.
Rev. Sci. Instrum., 80, 034101.
B ÖHM , M., J. TATCHEN, D. K RÜGLER, K. K LEINERMANNS, M. G. D. N IX, T. A. L E G REVE,
T. S. Z WIER and M. S CHMITT: 2009. ‘High-resolution and dispersed fluorescence examina-
tion of vibronic bands of tryptamine: Spectroscopic signatures for La /Lb mixing near a conical
intersection’. J. Phys. Chem. A, 113, 2456–2466.
B ORDÉ , C. J.: 1983. ‘Matrix equations and diagrams for laser spectroscopy’. In: F. T. A RECCHI
et al., eds., ‘Advances in Laser Spectroscopy’, 1. New York: Plenum Press.
B OYD , R. W.: 1999. ‘Order-of-magnitude estimates of the nonlinear optical susceptibility’. J.
Mod. Opt., 46, 367–378.
B OYD , R. W.: 2008. Nonlinear Optics. Burlington, San Diego, London: Academic Press, 3 edn.,
640 pages.
B REHM , B. and E. VON P UTTKAMER: 1967. ‘Koinzidenzmessungen von Photoionen und Pho-
toelektronen bei Methan’. Z. Naturforschg., A22, 8.
B ROYER , M., G. D ELACRÉTAZ, P. L ABASTIE, J. P. W OLF and L. W ÖSTE: 1987. ‘Spectroscopy
of vibrational ground-state levels of Na3 ’. J. Phys. Chem., 91, 2626–2630.
B ROYER , M., G. D ELACRÉTAZ, P. L ABASTIE, R. L. W HETTEN, J. P. W OLF and L. W ÖSTE:
1986. ‘Spectroscopy of Na3 ’. Z. Phys. D, 3, 131–136.
C ASTLEMAN , A. W. and K. H. B OWEN: 1996. ‘Clusters: Structure, energetics, and dynamics of
intermediate states of matter’. J. Phys. Chem., 100, 12 911–12 944.
C HA , C. Y., G. G ANTEFÖR and W. E BERHARDT: 1992. ‘New experimental setup for
photoelectron-spectroscopy on cluster anions’. Rev. Sci. Instrum., 63, 5661–5666.
C HANDRASEKHARAN , V. and B. S ILVI: 1981. ‘Transition polarizabilities and Raman intensities
of hydrogenic systems’. J. Phys. B, At. Mol. Phys., 14, 4327–4333.
C HANTLER , C. T., K. O LSEN, R. A. D RAGOSET, J. C HANG, A. R. K ISHORE, S. A. KO -
TOCHIGOVA and D. S. Z UCKER : 2005. ‘X-ray form factor, attenuation, and scattering tables
(version 2.1)’, NIST. http://physics.nist.gov/ffast, accessed: 7 Jan 2014.
C HASE , D. B. and J. F. R ABOLT: 1994. Fourier Transform Raman Spectroscopy: From Concept
to Experiment. New York: Academic Press.
C ONDON , E. U.: 1928. ‘Nuclear motions associated with electron transitions in diatomic
molecules’. Phys. Rev., 32, 0858–0872.
C OUTO , H., A. M OCELLIN, C. D. M OREIRA, M. P. G OMES, A. N. DE B RITO and M. C. A.
L OPES: 2006. ‘Threshold photoelectron spectroscopy of ozone’. J. Chem. Phys., 124, 204311.
C URRY , J., L. H ERZBERG and G. H ERZBERG: 1933. ‘Spektroskopischer Nachweis und Struktur
des PN-Moleküls’. Z. Phys., 86, 348–366.
C VEJANOV , S. and F. H. R EAD: 1974. ‘Studies of threshold electron-impact ionization of helium’.
J. Phys. B, At. Mol. Phys., 7, 1841–1852.
378 5 Molecular Spectroscopy
S ZTARAY , B. and T. BAER: 2003. ‘Suppression of hot electrons in threshold photoelectron pho-
toion coincidence spectroscopy using velocity focusing optics’. Rev. Sci. Instrum., 74, 3763–
3768.
TANAKA , K.: 2002. ‘N OBEL lecture: The origin of macromolecule ionization by laser irradiation’,
Stockholm. http://nobelprize.org/nobel_prizes/chemistry/laureates/2002/tanaka-lecture.html.
TAYLOR , K. J., C. L. P ETTIETTE -H ALL, O. C HESHNOVSKY and R. E. S MALLEY: 1992. ‘Ultra-
violet photoelectron-spectra of coinage metal-clusters’. J. Chem. Phys., 96, 3319–3329.
T RAVNIKOVA , O., K. J. B ØRVEB, M. PATANENA, J. S ÖDERSTRÖM, M IRON C ATALIN, L. J.
S ÆTHRE, N. M ARTENSSON and S. S VENSSON: 2012. ‘The ESCA molecule – historical re-
marks and new results’. J. Electron Spectrosc., 185, 191–197.
T ROFIMOV , A. B., J. S CHIRMER, V. B. KOBYCHEV, A. W. P OTTS, D. M. P. H OLLAND and L.
K ARLSSON: 2006. ‘Photoelectron spectra of the nucleobases cytosine, thymine and adenine’.
J. Phys. B, At. Mol. Phys., 39, 305–329.
T RUESDALE , C. M., S. S OUTHWORTH, P. H. KOBRIN, D. W. L INDLE, G. T HORNTON and D.
A. S HIRLEY: 1982. ‘Photo-electron angular-distributions of H2 O’. J. Chem. Phys., 76, 860–
865.
DE V RIES , M. S. and P. H OBZA : 2007. ‘Gas-phase spectroscopy of biomolecular building blocks’.
Annu. Rev. Phys. Chem., 58, 585–612.
WASSERMANN , T. N., O. V. B OYARKIN, B. PAIZS and T. R. R IZZO: 2012. ‘Conformation-
specific spectroscopy of peptide fragment ions in a low-temperature ion trap’. J. Am. Soc. Mass
Spectrom., 23, 1029–1045.
W EBER , A., ed.: 1979. Raman Spectroscopy in Gases and Liquids, vol. 11 of Topics in Current
Physics. Berlin, Heidelberg, New York: Springer.
W ERNER , A. S. and T. BAER: 1975. ‘Absolute unimolecular decay-rates of energy selected C4 H+ 6
metastable ions’. J. Chem. Phys., 62, 2900–2910.
W ILLIAMS , S., E. A. ROHLFING, L. A. R AHN and R. N. Z ARE: 1997. ‘Two-color resonant four-
wave mixing: Analytical expressions for signal intensity’. J. Chem. Phys., 106, 3090–3102.
W ILLIAMS , S., J. D. T OBIASON, J. R. D UNLOP and E. A. ROHLFING: 1995. ‘Stimulated-
emission pumping spectroscopy via 2-color resonant 4-wave-mixing’. J. Chem. Phys., 102,
8342–8358.
W ILLIAMS , S., R. N. Z ARE and L. A. R AHN: 1994. ‘Reduction of degenerate 4-wave-mixing
spectra to relative populations .1. Weak-field limit’. J. Chem. Phys., 101, 1072–1092.
W INTER , B., R. W EBER, W. W IDDRA, M. D ITTMAR, M. FAUBEL and I. V. H ERTEL: 2004.
‘Full valence band photoemission from liquid water using EUV synchrotron radiation’. J. Phys.
Chem. A, 108, 2625–2632.
W RIGGE , G., M. A. H OFFMANN, B. VON I SSENDORFF and H. H ABERLAND: 2003. ‘Ultraviolet
photoelectron spectroscopy of Nb− −
4 to Nb200 ’. Eur. Phys. J. D, 24, 23–26.
W RIGHT , J. C., R. J. C ARLSON, G. B. H URST, J. K. S TEEHLER, M. T. R IEBE, B. B. P RICE,
D. C. N GUYEN and S. H. L EE: 1991. ‘Molecular, multiresonant coherent 4-wave-mixing spec-
troscopy’. Int. Rev. Phys. Chem., 10, 349–390.
YANG , S. H., C. L. P ETTIETTE, J. C ONCEICAO, O. C HESHNOVSKY and R. E. S MALLEY: 1987.
‘Ups of buckminsterfullerene and other large clusters of carbon’. Chem. Phys. Lett., 139, 233–
238.
Y EE , S. Y., T. K. G USTAFSON, S. A. J. D RUET and J. P. E. TARAN: 1977. ‘Diagrammatic
evaluation of density operator for nonlinear optical calculations’. Opt. Commun., 23, 1–7.
Z HU , L. C. and P. J OHNSON: 1991. ‘Mass analyzed threshold ionization spectroscopy’. J. Chem.
Phys., 94, 5769–5771.
Z WIER , T. S.: 2001. ‘Laser spectroscopy of jet-cooled biomolecules and their water-containing
clusters: Water bridges and molecular conformation’. J. Phys. Chem. A, 105, 8827–8839.
Basics of Atomic Collision Physics: Elastic
Processes 6
Overview
The subject of this and the following two chapters is collisions between elec-
trons, atoms, ions and molecules. We mostly refer here to examples from the
particularly productive pioneering period between 1965 and 1990. However,
when appropriate, we mention already state-of-the-art research. In Sects. 6.1
and 6.2 we familiarize ourselves with cross sections, and how they are mea-
sured, with collision kinematics and its applications. As far as scattering the-
ory is concerned we shall refrain from rigid derivations and prefer easy to
understand models. In Sect. 6.3 we introduce elastic scattering and its clas-
sical theory while Sect. 6.4 outlines the quantum theory of elastic scattering.
A first glimpse on resonances is given in Sects. 6.5 and 6.6 introduces B ORN
approximation for elastic scattering. The two following chapters will go into
more depth and treat in particular also inelastic processes. We thus recom-
mend to the reader to study the present chapter with particular care.
6.1 Introduction
1 We shall use the terms “collision” and “scattering” physics essentially as synonyms – perhaps
spectroscopy:
W< 0
bound particles
discrete energies
structure
transition frequencies
here electron scattering (e− + atom, ion, molecule) as well as the so called “heavy
particle” scattering, i.e. processes where atoms, ions and molecules interact with
each other (an excellent, profound introduction into the latter area gives L EVINE
2005, in the following we shall refer again to the original literature and to reviews
for more details). Figure 6.1 illustrates – somewhat schematically – the relation of
energy domains to scattering physics (W > 0) and spectroscopy (W < 0) for the
interaction of two particles. Clearly, the limits are not sharp and the most interesting
phenomena and insights of modern atomic and molecular physics are often found
where the fields overlap, as e.g. for autoionization and resonance scattering, with ul-
tra cold atoms and molecules or in ultrafast physics and femtochemistry – all cutting
edge research themes which to some extend bridge the two fields as far as atomic,
molecular and cluster physics is concerned.
Specifically in the development of new methods, today spectroscopy and scatter-
ing physics profit of each other in a very direct manner. It may suffice at this point
to mention the application of state-of-the-art laser techniques and molecular beams,
or multidimensional measuring and detection methods with sophisticated imaging
devices – methods which both fields rely on since several years.
Nevertheless, one should bear in mind the different methods, concepts and goals
if one wants to assess the status of modern research in both fields. The most obvi-
ous distinction are the objects of the respective investigations: in spectroscopy one
is usually interested in determining the positions of energy levels. This is possible
today with a precision of which collision physics does not even dare to dream. The
latter usually takes the energetic position of atomic or molecular levels as given and
measures cross sections for specific dynamical processes. As a consequence of the
different nature of such measurements for collision processes, measurements with
an accuracy of a few percent have to be rated as great achievements. Nevertheless,
during the history of physics, this kind of experiments has led time and again to
key insights and breakthroughs. We recall the historic experiments of RUTHER -
FORD (atomic structure) and H OFSTÄDTER (structure of atomic nuclei, Nobel prize
1961), as well as the whole high energy physics which finally lives of collision pro-
6.1 Introduction 385
cesses. With Ugo FANO one may say that the “final goal of all low energy physics
is elucidation of physico-chemical elementary processes in terms of wave mechan-
ical concepts”. Beyond this somewhat puristic aspect, we emphasize the enormous
practical significance which the precise knowledge of scattering and reaction cross
sections has for practically all areas of physics and physical chemistry, in particular
for plasma physics (and thus in particular for fusion research), for astrophysics, for
atmospheric research or for radiation physics and chemistry.
To specify a few examples for these types of collisions for illustration, we mention:
ionization by electron impact, e.g. e− + H → 2e− + H+ , a so called (e, 2e) process,
dissociative attachment e− + AB → A− + B, charge transfer A + B+ → A+ + B
or chemical reactions A + BC → AB + C. The variety of processes is gigantic and
many of them are of considerable practical importance. Today, a multitude of me-
thodical developments open unforeseen possibilities for studying such processes,
many of them based on the continuous progress of laser physics and refined detec-
tion techniques for atomic and molecular particles.
The key quantity to be determined is the so called cross section, which we specify
as σel , σinel , σion and σreact for elastic, inelastic, ionizing, and reactive cross sections.
At this point we do not yet distinguish according to the angle under which the col-
liding particles are scattered. This aspect will be introduced in Sect. 6.2. The so
called integral cross section has been integrated over all scattering angles. The sum
of all the different integral cross sections for a given pair of interacting particles is
usually called the total cross section
σtot = σel + σinel + σion + σreact , (6.5)
even though there is occasionally some confusion in the literature between integral
and total cross section. A total cross section is determined by an absorption exper-
iment as sketched in Fig. 6.2. A beam of particles A (think e.g. of atoms, ions or
electrons) with a flux Φ0 ([Φ0 ] = particles s−1 m−2 ) passes through a static target
in a collision chamber. We recall from Sect. 1.3.2, Vol. 1 that the loss of flux for
386 6 Basics of Atomic Collision Physics: Elastic Processes
where NB is the density of the target gas ([NB ] = particles m−3 ). The thus intro-
duced total cross section σtot for all the reactions A + B according to (6.1)–(6.4) has
the dimension L2 . It may be thought to represent an interaction area πRint 2 if A and
B interact over a distance Rint . In this simple form (6.6) is of course only valid if
σtot NB x
1. For an extended or more dense interaction volume of length d the
integration leads to the well know L AMBERT-B EER law
being the mean free path length of particle A in the target gas B.
To compare the terminology of collision physics with that of chemical reaction
kinetics one writes (6.6) in symmetric form:
ΦA
− = ΦA σ NB = σ vNA NB . (6.8)
x
We suppress here the index for σ , since the relation holds as well for all the partial
cross sections, e.g. for elastic or reactive processes. The particle flux ΦA = vNA
is derived from velocity v and density NA of the particle beam A. Its decrease per
length, i.e. (minus) its derivative −dΦA /dx corresponds to the number of scattering
events ṄAB per time and volume. Thus, (6.8) is also valid for a gas mixture in a
cell – providing the velocity v is replaced by the relative velocity vrel of the two
particles involved. Hence, we may rewrite (6.8) as a rate equation
with the rate constant2 kAB = vrel σ which is measured in units [kAB ] = m3 s−1 .
Chemical reactions are usually studied in gas cells or in the condensed phase, in
each case in an environment of many particles with a distribution of velocities. The
latter can often be described by a thermal (M AXWELL -B OLTZMANN) distribution
fA (vA , MA ) and fB (vB , MB ), respectively, with velocities vA and vB and the re-
spective masses MA and MB . A similar situation is encountered e.g. for collision
2 One has to distinguish the rate constant kAB defined here (unit m3 s−1 ) of the rate Rba (unit s−1 )
which we have introduced in Chap. 4, Vol. 1, in the context of light induced processes, which is
defined per target atom.
6.1 Introduction 387
If the cross section depends only weakly on the relative velocity on may pull
σ (vrel ) = σ in front of the integral, so that:
kAB vrel σ with vrel = vA 2 + vB 2 . (6.11)
At this point we derive a useful relation between cross sections for collision pro-
cesses which may be considered the inverse of each other. Let us consider as an
example the inelastic process (6.2) and its inverse and let us assume these processes
to occur in a gas in thermodynamic equilibrium. For simplicity we take particle B
to be very heavy3 as compared to A, the latter being e.g. an electron which carries
essentially all available kinetic energy T prior to the (inelastic) excitation process
σba (T )
A(T ) + B(a) −→ A(T − Wba ) + B(b), (6.12)
while T − Wba is its kinetic energy after the process, and Wba is the excitation
energy for the transition |b ← |a. The de-excitation process (also called “super-
elastic” process) is exactly the inverse: one starts with a kinetic energy T − Wba and
the state |b as initial state:
σab (T −Wba )
A(T − Wba ) + B(b) −→ A(T ) + B(a). (6.13)
3 One may drop this restriction without changing the result – the derivation becomes, however,
The fact, that these two, complementary processes always exist is called micro re-
versibility.
Let us now consider a gas mixture consisting of particles A and B with densi-
ties NA and NB , respectively. Let the distribution of kinetic energies of particles
A be given by f (T ), hence dNA = NA f (T )dT particles A per unit volume have
an energy between T and T + dT . The ensemble of all particles A looses energy
by inelastic processes (6.12), it gains energy by de-excitation processes (6.13), as
schematically illustrated in Fig. 6.3. In an energy range between T and T + dT
according to (6.9)
excitation processes occur per unit time and volume. They fill the energy distribution
between (T − Wba ) and (T − Wba ) + dT and depopulate the energy range from T
to T + dT . The corresponding number of de-excitation processes is
dṄab = vkin σab (T − Wba )NA f (T − Wba )NB(b) dT . (6.15)
They fill the energy distribution between T and T + dT and depopulate it between
(T − Wba ) and (T − Wba ) + dT . The velocity of the particles A before and after
, respectively. The particle densities N
the excitation process is vkin and vkin B(a) and
NB(b) refer to particles B in state |a and |b, respectively.
Now, in thermodynamic equilibrium the so called principle of detailed balance
must hold: each individual process is kept in equilibrium by its inverse process. The
energy loss by reaction (6.14) is kept in balance (for the sum of all processes) by the
energy gain according to (6.15). Hence
vkin σba (T )f (T )NB(a) = vkin σab (T − Wba )f (T − Wba )NB(b) . (6.16)
At the same time for the densities of particle B the B OLTZMANN statistics must hold
NB(b) gb − Wba
= e kB Θ (6.18)
NB(a) ga
which connects the cross sections for excitation and de-excitation of an atomic or
molecular collision induced transition process. We may regard this relation as the
‘collisional analogue’ to the equivalence of induced emission and absorption, i.e. as
corresponding to the E INSTEIN relation for the B coefficients according to (4.38) in
Vol. 1. In contrast to optical processes, no energy resonance is needed for collisional
excitation or de-excitation: the projectile looses or gains as much energy as needed
for the excitation or de-excitation process. Of course, the respective cross sections
are a function of energy, and excitation cross sections clearly disappear for kinetic
energies below the threshold T < Wba . Another important difference between (6.19)
and the optical equivalent (4.38), Vol. 1 is expressed by the energetic prefactors in
the case of collision processes. They are a consequence of the changing particle
velocity due to inelastic processes and ensure that the flux of particles A remains
conserved during a collision.
One word about the orders of magnitude of atomic and molecular cross sections.
They may vary over many decades. Typical gas kinetic (elastic) cross sections are
on the order of 10−19 m2 (see next section). But values from some 10−17 m2 or even
larger, down to 10−25 m2 and less are observed – depending on the type of process
and the kinetic energy during the interaction. In Chap. 7 we shall get to know some
general trends for inelastic processes.
At sufficiently low energies and for non-reactive collision partners only elastic
scattering occurs during which the kinetic energy T in the centre of mass sys-
tem of the collision partners is conserved. In this case the order of magnitude of
the cross section observed may be visualized intuitively: if one thinks of two bil-
liard balls of radius R1 and R2 , the classical prediction for the elastic cross sec-
tion is σel = π(R1 + R2 )2 . This is the gas kinetic cross section. If we consider
one of the most simple scattering systems H + He, one estimates from the cova-
lent radii for H and He according to Fig. 3.4 in Vol. 1 for the elastic cross section
σel 0.4 nm2 = 4 × 10−19 m2 . A comparison with real experimental data, as pre-
sented in Fig. 6.4, shows that the order of magnitude is right – the details suggest,
however, that further consideration is indicated!
390 6 Basics of Atomic Collision Physics: Elastic Processes
Atoms simply are not hard spheres; they have, so to say, a soft cover. Thus
it becomes understandable that the cross section decreases with increasing veloc-
ity (or energy) in the supra-thermal energy range:4 A collision at higher energies
probes the repulsive part of the interaction potential (see e.g. Fig. 6.1) at lower dis-
tances R. For the H + He system the interaction potential V (R) is shown in Fig. 6.5,
and is compared with an idealized hard sphere. The classical turning point Rc de-
pends strongly on the total energy T (i.e. on the initial kinetic energy in the centre
of mass system): Rc is obtained from T = V (Rc ). At higher energies the elastic
cross section decreases explicitly with the relative velocity u – and not with en-
ergy – as documented by the comparison of H + He and D + He in Fig. 6.4. This
will be explained in Sect. 6.3.3. In contrast, for hard spheres (dashed red line in
Fig. 6.5) one would expect a constant Rc (and thus a cross section independent of u
and T ).
4 Wenote, that the root mean square relative velocity in a hypothetical gas of H + He (or D + He)
atoms at room temperature corresponds to u 1750 m /s (or 1350 m /s).
6.1 Introduction 391
∆σ
(b)
___el
σel
2%
0%
- 2%
1000 2000
u / ms-1
The fact that for still lower velocities the cross section decreases again is of
quantum mechanical origin. We shall discuss this below in some detail. Here we
just point out that for the system H + He (reduced mass M̄ = 4/5 u) the DE
B ROGLIE wavelength is about λdB = h/M̄u 0.38 nm at u 1300 m /s (where
the maximum cross section is reached). Thus, we find the interesting ‘coincidence’
πλ2dB = 0.46 nm2 σel (max). For the system D-He at the same velocity λdB is
smaller by a factor 3/5, and the cross section still increases with decreasing relative
velocity, down to u (700 to 800) m s−1 .
We show a few more examples of elastic cross sections, which illustrate that such
scattering processes may lead to a variety of interesting structures. First the system
Na + Hg in Fig. 6.6, where we recognize a series of oscillations as a function of
relative velocity, the so called glory oscillations, which we shall explain later. The
general trend – the cross section decreases with velocity – is explained in the same
manner as in the previous example by a softness of the interaction potential.
A different behaviour may be observed for the scattering of slow electrons e.g.
by rare gases. In Fig. 6.7 the situation is shown for the system e− + He. This still
looks very similar to in the case of H + He scattering which we have just discussed.
One may glean from the data that the cross section first goes through a maximum
before joining the general trend to decrease with increasing energy.
However, as shown in Fig. 6.8, for e− + Ar, e− + Kr and e− + Xe one observes
distinctively different structures as a function of energy. The cross section starts at
finite values and passes through a minimum to climb a rather pronounced maximum
before finally decaying again with energy. The pronounced minimum which is seen
here, the so called R AMSAUER minimum is today well understood and often encoun-
tered in low energy electron scattering; in some rarer cases it may also be observed
in heavy particle scattering. These observations can, however, not be understood in
terms of classical models – we shall come back to this in Sect. 6.4.5. Also, when
392 6 Basics of Atomic Collision Physics: Elastic Processes
20 Ar
10
He
Ne
0
1 10 100
electron kinetic energy T / eV
using the concept of a potential one should be very careful in the context of interac-
tions between electrons and atoms: atoms themselves consist of electrons and may
become transparent – as the initial rapid decrease of the cross section shows. Hence,
the theoretical treatment of electron scattering has to be fully quantum mechani-
cally, in contrast to the theory of heavy particle collisions which usually are treated
in a semiclassical manner. We shall develop these theories in some detail later in
this and the two following chapters.
Section summary
• When studying elastic or inelastic cross section one observes a variety of in-
teresting structures as a function of velocity or energy. We have shown some
examples in Figs. 6.4–6.8. One may extract from such observations specific,
microscopic information on the dynamics of a collision process.
• Obviously, experiments in a gas cell where only rate constants are measured,
such fine details are averaged out according to (6.10). The price for additional
6.2 Differential Cross Sections and Kinematics 393
Atomic and molecular collision processes have been studied over the past decades
in ever finer details, aiming at a precise microscopic understanding of the interaction
processes. On this basis a quantitative comparison with sophisticated quantum me-
chanical models can be achieved. Beyond the measurement of total cross sections,
as discussed in the previous section, the next logical step is the measurement of the
angular distribution of scattering intensities for the different processes. In Fig. 6.9 a
scheme for such a differential scattering experiment is presented.
The number of particles Ṅ which are scattered per unit time from a scattering
volume VS = AP d into the solid angle Ω is given by
As in Sect. 6.1.1 the flux of particles A in the projectile beam is ΦA = vNA , and
NB is the particle density in the target gas B. The area of the projectile beam is AP ,
and d is the effective length from which particles are registered at the detector. The
angular distribution I (θ ) characterizes the physics of the collision process, and is
called differential cross section (often just DCS):
dσ dσ
I (θ ) = = . (6.21)
dΩ dϕ sin θ dθ
The DCS has the dimension L2 per solid angle, [I (θ )] = m2 sr−1 . In the most gen-
eral case it depends on the (polar) scattering angle θ as well as on the azimuthal
angle ϕ. For statistically aligned projectiles and targets, however, I (θ ) is only a
394 6 Basics of Atomic Collision Physics: Elastic Processes
θ
the detector. The dashed red volume
e
rtur
lines indicate the finite ape
angular resolution. Δθ
b detector for
(b) Simplified scheme to
illustrate the necessary sin θ scattered
particles A
correction of the scattering
volume (b) d = b / sinθ
A beam
θ
b
function of θ . Integration over all scattering angles leads again to the integral cross
section (ICS) introduced in Sect. 6.1:
π
dσ
σ= dΩ = 2π I (θ ) sin θ dθ. (6.22)
4π dΩ 0
The differential cross section may refer to elastic, inelastic, ionizing or reactive
processes, just as the ICS. We emphasize again that integral cross sections σ should
not be called total cross section – as sometimes done. The total cross section is
defined according to (6.5) as the sum of all possible integral cross sections. Ex-
perimentally one may keep the solid angle seen by the detector constant without
problems by using angle limiting apertures. If the detector area is AD and its dis-
tance from the scattering volume is R, then Ω = AD /R 2 is the solid angle ‘seen’
by the detector.5
In contrast to Ω, the effective length d of the scattering volume in the beam-gas
experiment Fig. 6.9(a) also depends on the scattering angle, as sketched schemati-
cally in Fig. 6.9(b). Hence, in such an experiment the observed scattering intensity
has to be divided by sin θ to obtain the true angular distribution I (θ ). One immedi-
ately sees that this may lead to substantial uncertainties for small scattering angles.
Experiments with crossed beams may overcome this problem – if they are setup
properly. Figure 6.10 illustrates schematically such an experiment. Ideally, two well
collimated, energy selected and (as far as possible state selected) particle beams
cross – usually at right angles. The detector has to be constructed such that one
of the collision partners A or B is detected, if necessary after suitable energy and
state analysis. Important for such an experiment is, that the whole scattering volume
is seen by the detector completely for each arbitrary scattering angle. For this, the
5 Note: this solid angle of detection must not be confused with the finite angular resolution of the
B beam
scattering volume
energy and/or state
analyzer
ure
energy and/
ert
or state
ap
selector B
detector for
source scattered
B particles
Fig. 6.10 Scheme of a crossed beam scattering experiment. Note that the detector always sees
the whole scattering volume, independent of the scattering angle, so that no sin θ correction is
necessary
detection cone must be large enough, at the same time, however, provide the neces-
sary angular resolution. In practice it is difficult to meet all these demands and often
special corrections still have to be applied.
For completeness we mention here that for electron scattering often the so called
momentum transfer cross section is given for practical applications:
dσ
σemom = (1 − cos θ ) dΩ
4π dΩ
π
= 2π (1 − cos θ )I (θ ) sin θ dθ. (6.23)
0
From Fig. 6.11 one reads the magnitude of the momentum transfer (p e → p e ) of an
electron e− during an elastic collision: pe = 2pe sin(θ/2). This momentum dif-
ference is transferred to the target B (atom, ion, molecule), which then has the mo-
mentum p B = −pe . Strictly speaking, one has to compute the energy balance in
the centre of mass system (see next subsection). One finds that the target receives a
kinetic energy (pe )2 /2MB = 4p 2 sin2 (θ/2)/2MB = 2T (M̄/MB )(1 − cos θ ). Here
p is the relative momentum and T the initial kinetic energy of the electron in the
centre of mass system, M̄ its reduced mass, and MB the target mass. The momentum
transfer cross section, defined in (6.23), thus gives a measure for the average energy
which is transferred per elastic collision onto the target. This energy may e.g. lead
to the heating of a plasma.
The assumption, particle B had infinite mass and could be assumed to be at rest be-
fore and after collision is of course not generally valid. We drop it now and present
the correct treatment which considers the motion of both particles. According to
classical mechanics the momentum of the centre of mass of two particles is inde-
pendent of their interaction. All definitions remain valid if we refer to the relative
motion of the collision partners. Only the relative coordinates of A and B are used
to describe the dynamics. The necessary kinematic transformations are, however, of
key importance for the correct interpretation of all heavy particle collisions. Only
for collisions of electrons with atoms and molecules they may be disregarded.
Figure 6.12 shows trajectories of the particles A and B. In laboratory coordi-
nates (origin O), briefly lab-frame, they are described by the coordinates R A (t) and
R B (t). The centre of mass (CM) moves on a straight line (dot-dashed). We recall the
transformation from lab-frame to CM-frame as used when treating the H2 molecule.
The Hamiltonian for the system of particles A and B with momenta p A and pB
(masses MA and MB ), respectively, and the interaction potential V (R A − R B ) is
given by
p2A p2
HAB = + B + V (R A − R B ). (6.24)
2MA 2MB
It is transformed into the CM-frame by introducing the relative coordinate R and
the relative velocity u
R = RA − RB and u = Ṙ A − Ṙ B = v A − v B (6.25)
according to Fig. 6.12. With this, after a brief calculation (6.24) becomes
P2 p2
HAB = + + V (R) (6.26)
2M 2M̄
P = MV = MA v A + MB v B = const, (6.27)
The constant kinetic CM energy P 2 /2M may be subtracted from the total energy
HAB (separation of the overall translation and relative motion) and we obtain (clas-
sically as well as quantum mechanically) for the energy of the relative motion in the
CM system the H AMILTON operator:
p2
HCM = + V (R). (6.30)
2M̄
The whole relevant dynamics is thus described by the relative canonical coordinates
R and p. So far all this is in principle known already from the H atom or the H2
molecule.
In the following we have to transform the possible changes of the relative mo-
mentum p, as a consequence of the collision, into experimentally observables in
the laboratory system. In the context of these kinematic considerations we are
only interested in the asymptotic behaviour before and after the collision (i.e. for
R → ∞, V (R) → 0), which we mark by un-dashed and dashed quantities, respec-
tively. Energy conservation then implies for the relative kinetic energy in the CM
system:
1 1
TCM = M̄ u2 = M̄u2 ∓ W = TCM ∓ W. (6.31)
2 2
This includes the possibility of inelastic and super-elastic processes, which due to
changes of internal energy ±W of the participating particles A and/or B may lead
to a corresponding change of the relative velocity. For reactive processes even the
reduced mass M̄ → M̄ may change. In the lab-frame the total kinetic energy is
then
1 1 1
TLab = MA vA 2
+ MB vB2 = MV 2 + TCM . (6.32)
2 2 2
By replacing in this expressions all quantities by dashed ones (v → v ), one obtains
the lab energies after the collision process. With (6.32) it is clear that of the total
kinetic energy which the collision partners have in the laboratory system, only a
fraction is available for the collision process, that is TCM .
Very clear and helpful for visualizing the kinematics is the so called N EWTON
diagram of the scattering process, which we introduce in Fig. 6.13. It represents the
398 6 Basics of Atomic Collision Physics: Elastic Processes
laboratory velocities v A and v B together with the relative velocity u and the velocity
of the centre of mass V . We rewrite the definition (6.27) for the momentum of the
centre of mass as
M A v A + M B v A − MB v A + MB v B
V= ,
M
MB MA
so that with uA = u and uB = − u,
M M
MB MA
V = vA − u = v A − uA = v B + u = v B − uB or
M M
finally: v A = V + uA and v B = V + uB . (6.33)
The N EWTON diagram Fig. 6.13 represents the last equation graphically. The centre
of mass, CM, is thus represented by a point which divides the relative velocity u =
v A − v B proportional to the mass MB and MA , respectively.
One big advantage of this diagram is that one may use it equally well after the
collision process, to transform the kinematics from the CM system back into the
lab-frame. Since the centre of mass velocity V remains constant while the relative
velocity u after the collision may be different from that before, u, one obtains all
potentially possible laboratory velocities after the collision from the relations
For the graphical construction one simply has to draw circles around the CM
with a radius u MA /M and u MB /M for particle B and A, respectively. On these
circles all kinematically allowed laboratory velocity vectors v A and v B will be found
(their origin being at O). The magnitude of u is derived from the energy balance
(6.31) for the process studied. In the case of reactive processes the masses may also
change, i.e. MA and MB may differ from MA and MB . Figure 6.14 illustrates the sit-
uation for elastic, inelastic and super-elastic scattering processes without reactions.
The examples shown assume in all three cases a centre of mass scattering angle
θCM = 45◦ .
The physical quantity one wants to determine is the differential cross section
I (θCM ) in the CM system, i.e. the probability to observe (for a specific collision
process studied) a certain CM scattering angle θCM . This is the quantity which re-
veals the details about the interaction dynamics. In contrast to the latter, N EWTON
6.2 Differential Cross Sections and Kinematics 399
(c) uA
vA
uA' CM u B'
V vB'
vA'
θ CM
θ Blab
uB
O vB
diagrams describe what one calls kinematics, i.e. the transformation from the CM
system into the laboratory system and vice versa.
We have to point out that the conversion of data measured in the lab-frame into
the CM-frame is not always unambiguously possible – if only the scattering angle
is determined. If the circles with radius uA or uB , respectively, do not fully enclose
the CM velocity V , then there exist laboratory scattering angles which belong to
two different CM scattering angles, as seen in Fig. 6.14(a) and (b). If one knows
the type of process studied, a velocity analysis may help. However, the situation
becomes very complicated if all three process types illustrated in Fig. 6.14(a)–(c)
occur simultaneously.
Another complication arises when transforming the scattering rates measured in
the lab into differential cross sections dσCM /dΩCM in the CM system: the detector
acceptance angle changes from dΩlab → dΩCM , and one has to apply an appropri-
ate transformation factor. Corresponding considerations hold for the energy scale.
These transformations are achieved by applying the corresponding Jacobian deter-
minant. Finally, at low and possibly thermal energies one has to account for the fact
that different initial velocities lead to uncertainties in determining the CM system.
The de-convolution of all these distributions can be rather complicated (PAULY and
T OENNIES 1965). Thus, in the following we just present for illustration a partic-
400 6 Basics of Atomic Collision Physics: Elastic Processes
ularly elegant example which shows how one may even exploit the cumbersome
kinematic relations with advantage to solve specific problems.
The investigation of atomic and molecular clusters – i.e. of more or less loosely
bound systems of many atoms and/or molecules – is an important field of mod-
ern research. It has developed essentially out of atomic and molecular physics and
is closely related with today’s very fashionable ‘nano’ physics. To generate such
clusters one may use supersonic, cold molecular beams. By adiabatic expansion of
gases at high pressure (typically 1–100 bar) through a thin nozzle into vacuum, the
atoms or molecules are efficiently cooled over a short distance of flight, and par-
tially condense, thus forming clusters. Their size N may range from 2 to many ten
thousands – usually in a more or less broad distribution. When studying them it is
of course desirable to know their size. If one treats ionic clusters, N may be fixed
by mass spectrometric methods. But neutral clusters have to be ionized prior to de-
tection, and one usually has to accept that the ionization process may change the
cluster distribution.
B UCK and M EYER (1984) have developed an elegant, albeit rather elaborate,
method to select even neutral clusters according to their size: by using the very
same collision kinematics which we have just discussed. The idea is to deflect the
different cluster sizes N by collisions, and use the kinematics to select them. Fig-
ure 6.15(a) explains the principle using again the N EWTON diagram, here for the
scattering of ArN clusters by He atoms – as applied in the first, pioneering exper-
iment. Since the mass of the ArN clusters is large compared to the light collision
partner He, the relevant part of the N EWTON diagram is localized around the ini-
tial velocity of the ArN clusters. This region is sketched enlarged in Fig. 6.15(a).
The relative velocity u (red, intermitted arrow) prior to the collision defines the CM
system. The red circles mark the possible positions of velocity vectors after a col-
lision for clusters with 2 ≤ N ≤ 5. As seen, for each laboratory angle θlab there
are two scattering angles in the centre of mass system which lead to different lab
velocities (black arrows). One may distinguish these by measuring the velocity of
the scattered clusters. The experimental realization is illustrated schematically in
Fig. 6.15(b). ArN cluster beam and He beam cross each other at right angle. The
scattered clusters obtain a temporal structure by passing the chopper (C). They are
collimated and detected behind a quadrupole mass spectrometer. Their time of flight
allows to determine their velocity. The measured signal is shown in Fig. 6.15(c) for
three different lab scattering angles θlab . As predicted by the N EWTON diagram (a)
one finds at θlab = 14◦ only two peaks, corresponding to Ar2 clusters at the two dif-
ferent possible scattering angles in the CM system. At θlab = 10◦ one already detects
also Ar3 , while at θlab = 8◦ in addition two more peaks from Ar4 are seen. Overall,
these results show a rather impressive experimental confirmation of the kinematic
considerations explained in the previous subsection.
One may now use the thus documented angular and velocity distribution of the
scattered clusters for spectroscopic studies with mass selected neutral clusters. One
6.2 Differential Cross Sections and Kinematics 401
ArN velocity vA 5 3
2
He beam s 0
(b) rt ure 10 N = 2
ape r θ lab=14°
e cto
C det
θ lab 5
2
skimmer ArN beam
skimmer
ArN
source 0
He 0.6 0.8 1.0 1.2
source time of flight / ms
Fig. 6.15 Cluster size selection by elastic scattering according to B UCK and M EYER (1984).
(a) N EWTON diagram for scattering ArN by He atoms at velocities vA = 570 m /s and
vB = 1790 m /s. The red circles give the positions of all possible final velocity vectors for elasti-
cally scattered clusters (note the breaks in the velocity arrows). (b) Experimental setup schematic,
featuring a pseudo-statistical chopper C for the time of flight analysis of the scattered clusters.
(c) Measured time of flight spectra after ArN He collisions at three different laboratory angles,
corresponding to the N EWTON diagram (a)
may, e.g. localize a laser beam at a fixed position in space behind the scattering
centre, irradiating the cluster perpendicularly to the scattering plane of the clusters
indicated here. Buck and collaborators have developed and used this method as a
very powerful tool for spectroscopy and for studying fragmentation dynamics of
atomic and molecular clusters (see e.g. P OTERYA et al. 2009; B ONHOMMEAU et al.
2007; S TEINBACH et al. 2006; FARNIK et al. 2004).
Section summary
• Differential cross sections dσ/dΩ (DCS) describe the angular distribution of
scattered particles and give deep insights into the interaction dynamics. They
may be measured in well defined beam-gas or in crossed beam experiments.
Both have their specific problems and advantages.
• Collision kinematics – as opposed to collision dynamics – is merely defined
by momentum and energy conservation. The so called N EWTON diagrams
(see examples Figs. 6.13–6.15) provide very useful tools for visualization and
quantitative evaluation of the kinematic constraints.
402 6 Basics of Atomic Collision Physics: Elastic Processes
We treat now the elastic scattering of particle A in the isotropic potential V (R)
of particle B and interpret the differential cross section (6.21) geometrically as
sketched in Fig. 6.16: Let A approach B with a range of impact parameters b to
b + db. After the interaction A is scattered into an angle between θ and θ + dθ . Tak-
ing advantage of the axial symmetry, the elastic scattering rate for the whole conical
ring with dΩ = 2π sin θ dθ becomes proportional to dσ = 2πbdb. Particle A must
approach B within this cross section in order to be scattered into scattering angles θ
impact parameter
with radius R for the sun
wavelengths λ
R, α β
R
schematic. The classical p=1
b
b α
rainbow arises for trajectories α
β
of type p = 2. Further Θ2
reflections are possible and
lead to (weaker) secondary α β
rainbows (not shown here)
p=2
The dynamical problem is now simply, to determine the classical deflection function
Θ = Θ(b). Note: we distinguish here between scattering angle θ , which is always
taken as positive, and the deflection function Θ, which may assume positive as well
as negative values, depending on whether the overall interaction acts as attractive
or repulsive, respectively (the deflection sketched in Fig. 6.16 is negative). It is im-
portant to realize that this classical cross section diverges if the deflection function
goes through a maximum or a minimum, i.e. when |dΘ/db| = 0 holds. This leads
to the so called classical rainbow. This interesting phenomenon is indeed closely
related to the well known and popular optical rainbow, which arises from scattering
of sunlight by small water droplets.
So let us make a little detour into the physics of optical rainbows, which will allow
us to understand quite intuitively how the phenomena arise which we have just indi-
cated. Figure 6.17 shows a light beam arriving from the sun, at the interface between
air and water droplet, where it is refracted and/or reflected. Light may be reflected
at the surface (p = 0) or after entering the droplet and refraction may undergo sev-
eral additional reflections before finally leaving the droplet after back-refraction
(p = 1, 2, 3, . . . ).6 The famous optical rainbow is formed for p = 2, possibly fol-
lowed by one or more secondary rainbows p > 2.
6 As long as for the radius of the droplet R λ holds (with λ being the wavelength of the light) –
and this is usually the case for standard rainbows – we may treat the nearly parallel light arriving
from the sun as a bundle of light beams and apply geometric optics. Light scattering at smaller
objects shows very pronounced interference structures and is described by the M IE theory – an
application of classical electrodynamics. It is treated in detail in standard textbooks on optics (see
e.g. B ORN and W OLF 2006, p. 759ff).
404 6 Basics of Atomic Collision Physics: Elastic Processes
scattering angle θ
p=1 p=2
0˚
-150˚
-90˚
p=2 -160˚
-180˚
-170˚
-270˚ p=3
p=4
-360˚ -180˚
0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0
x = b /R x = b/R
Fig. 6.18 Classical deflection function Θ(b) for the scattering of light at a water droplet (radius
R) as a function of the ratio x = b/R of impact parameter to radius. (a) Different orders p for
an average value of the index of refraction. (b) Enlarged scale for p = 2 for red (red line) and
blue wavelength (grey line). The clearly visible maximum at deflection angles −137◦ and −139◦ ,
respectively, lead to the rainbow phenomenon
Hence, the deflection function Θp (b) depends only on the ratio of impact parameter
to droplet radius x = b/R and on the index of refraction n. Notably it does not
depend on the size of the droplet as such! For water, the index of refraction for
blue light is 1.3427 and for red light 1.3282. With these values (6.37) leads to the
deflection function as a function of b/R shown in Fig. 6.18.
The deflection function for p = 2 has a maximum at about −137.2◦ and −139.3◦
for red and blue light, respectively. This implies that many light rays with different
b lead to nearly the same deflection angle and thus contribute to a maximum of
intensity at these angles. In (6.36) this is expressed as a singularity. For different
wavelength this rainbow angle is slightly different – that is what we experience as
6.3 Elastic Scattering and Classical Theory 405
r
erve
obs p = 2 : main rainbow
the wonderful colours of an optical rainbow – red light appearing to come from a
slightly higher origin (or spanning a larger circle) than blue light.
Sometimes a so called secondary rainbow (p = 3) may be observed in addition
to the main rainbow (p = 2). It arises from one more reflection within the droplet.
Since the intensity of this scattered light is of course much weaker the secondary
rainbow is much more pale. As illustrated in Fig. 6.19 the observer sees the main
rainbow at an inclination 180◦ − |Θ| (relative to the direction of the incoming sun-
light) i.e. at ca. 41◦ and 43◦ for blue and red, respectively. For the secondary rainbow
one finds according to (6.37) an inclination of 49.6◦ and 54.4◦ for red and blue, re-
spectively. The order of the colours is thus inverted in the secondary rainbow as
compared to the main rainbow. According to Fig. 6.19 this is an immediate conse-
quence of the reflection and refraction geometry.
Let us get back to particle scattering! We consider a classical trajectory in the centre
of mass system as illustrated in Fig. 6.20(a).
During elastic scattering the magnitude of linearmomentum p = M̄u as well as
the angular momentum || = pb is conserved (u = 2T /M̄ is the magnitude of the
relative velocity). Hence, the impact parameter too is a constant of motion. With the
(a) b
R Rc
b Á Θ = π – Á(t = ∞)
(b) ψ
ψ b
R 2
= d/
b Á Rc Θ = π – 2ψ
Fig. 6.20 Classical trajectories in cylinder coordinates R, φ with the classical turning point Rc
(distance of closest approach). The deflection angle Θ is a function of the impact parameter b.
(a) General case, (b) elastic collision of point like particle with a solid hard sphere of diameter d
406 6 Basics of Atomic Collision Physics: Elastic Processes
moment of inertia M̄R 2 of the relative motion one may express || by the angular
velocity φ̇:
|| = M̄ub = M̄R 2 φ̇ = const. (6.38)
We have used these relations already in Chap. 2, Vol. 1 when formulating the
H AMILTON operator for bound states. With the effective potential
||2 b2
Veff (R) = V (R) + = V (R) + T (6.39)
2M̄R 2 R2
and the total energy T the H AMILTON function of the relative motion is written as:
M̄ 2
H= Ṙ + Veff (R) = T . (6.40)
2
With (6.38)–(6.40) one finds
dφ φ̇ b
= = , (6.41)
dR Ṙ R 2 1 − Veff (R)
T
Rc is the classical turning point (distance of closest approach of the particles) in the
effective potential Veff .
I (θ ) = d 2 /16, (6.44)
θ qA qB qA qB θ
= arctan and b = cot .
2 2T b 2T 2
Inserted into (6.36) this leads to the well known RUTHERFORD cross section:
2
dσ qA qB 1
I (θ, T ) = = θ
. (6.45)
dΩ T 16 sin4 2
Here we have used atomic units, i.e. T is given in units of Eh , the differential cross
section is measured in units a02 sr−1 .
For the scattering of two neutral atoms in their isotropic ground state (L ENNARD -
J ONES potential) with s = 6 the prefactor becomes (s − 1)f (s) = 15π/16. For the
differential cross section (6.36) in general one obtains
1 (s − 1)f (s)C 2/s −2(s+1)/s
I (θ ) = θ and specifically (6.47)
s T
1/3
C
= 0.2389 θ −7/3 for s = 6. (6.48)
T
τ (b, T ) = θ T , (6.49)
and instead of the differential cross section (6.36) the reduced cross section
τ db2
ρ(b, T ) = θ sin θ I (θ, T ) = . (6.50)
2 dτ
408 6 Basics of Atomic Collision Physics: Elastic Processes
holds. Hence, to 1st order, τ depends only on the impact parameter b and not on the
kinetic energy T . More generally, one may show (see e.g. S MITH et al. 1966), that
τ can be expanded in powers of T −n :
θ
τ (b, T ) = τ0 (b) + T −1 τ1 (b) + · · · = τ0 (b) + τ1 (b) + · · · . (6.52)
τ
Correspondingly, the reduced cross section may be written as
θ
ρ(b, T ) = ρ0 (b) + T −1 ρ1 (b) + · · · = ρ0 (b) + ρ1 (b) + · · · . (6.53)
τ
The approximation holds for high energies and arbitrary scattering angles – or alter-
natively for arbitrary energies and small scattering angles (large impact parameters).
For collisions with hard spheres the situation is unambiguous: for b > d no scatter-
ing occurs, hence the maximum impact parameter is bmax = d/2. Alternatively one
may also insert (6.44) into the left integral and obtain the same result, σ = πd 2 /4,
corresponding to the geometric visualization.
However, if we study a potential with infinite interaction range, on first sight
it looks as if the integral elastic cross section diverges: however large one chooses
bmax , classically there remains always some deflection, albeit very small. Hence, the
question arises, which is the maximum impact parameter bmax that leads to a still
measurable deflection. This question can only be answered quantum mechanically,
i.e. in last consequence by the uncertainty relation. From this perspective only such
scattering angles Θ(b, T ) may be considered reasonable for which the product of
lateral momentum Θ(b, T )M̄u and impact parameter b is larger than h. Hence, there
exists always a maximum impact parameter bmax for which this is just fulfilled:
For large b the interaction potential is given by V (b) = −C/R s and the classical
deflection function is given by (6.51). Hence, with (6.55)
1/(s−1)
h C C
Θ ∝ s =⇒ bmax ∝ .
M̄ubmax bmax M̄u2 hu
6.3 Elastic Scattering and Classical Theory 409
bg
Θ
b1' π b1 b2 b 3 b→∞
(c)
0
-θr
br
sinθ × I(θ)
(d)
0˚ θr 180˚
the differential cross section at the rainbow angle θr . Just as in the case of the optical
rainbow, the phenomenon arises since a whole range of impact parameters around
br leads to the same scattering angle θr – just there where Θ(b) goes through a
minimum. Of course, nature will smooth out such singularities. In a quantum me-
chanical or semiclassical treatment instead of the sum over scattering cross sections
(6.57), the square of the sum of corresponding scattering amplitudes f (θ ) will have
to be used:
2
I (θ, T ) = fj (θ ) . (6.58)
j
θ 7/3 × I(θ)
determined differential cross (b)
section for the system Cs-Hg 1000
according to B UCK et al.
(1972). (a) I (θ) weighted
with sin θ , (b) blow up of
I (θ) up to the rainbow 100
angle θr , weighted with θ 7/3 0° 10° 20° 30° 40°
(see Eq. (6.48))
10 θr
(a)
1
0° 60° 120° 180°
scattering angle θ
0
0° 10° 20° 30° 40°
scattering angle θ
b3
supernumerary orbiting
(a) rainbows b2 (d) resonance
b 2, b 3
shadow
rapid scattering
(b) oscillations (e) (diffraction)
b1'
θ ~ λ dB / a
bg
(c) glory (f ) symmetry-
oscillations oscillations
Fig. 6.24 Schematic illustration of interfering trajectories (see Fig. 6.21) as origin of different
types of oscillations in the low energy elastic atom-atom scattering
Figure 6.24 summarizes schematically the type of trajectories responsible for the
different anomalies. Figures 6.24(a) and (b) illustrate the origin of the just docu-
mented slow (supernumerary rainbows) and rapid oscillations (in Fig. 6.23 recog-
nizable on the rainbow and on the supernumerary rainbows). Related are the so
called glory oscillations according to Fig. 6.24(c). They arise by interference be-
tween trajectories with very large and small impact parameters bg , both leading to
forward scattering – the latter by compensation of attractive and repulsive parts of
the potential (see also Fig. 6.21(c)). Experimentally one observes them most clearly
in the integral cross section as already shown in Fig. 6.6.
The phenomenon indicated in Fig. 6.24(d) is called an orbiting resonance. For
a short time a trajectory is – so to speak – resonantly captured into a quasi-bound
state of the interaction potential (collision complex). A particularly nice example
according to T OENNIES (2007) gives the elastic, integral cross section for H + Xe
scattering shown in Fig. 6.25. Quantum mechanically such short-lived resonances
may e.g. be formed behind a rotational barrier which arises by the superposition
of an attractive potential and the repulsive centrifugal potential, as schematically
illustrated in the inset of Fig. 6.25. This may lead to resonances for very low kinetic
energies. Experimentally the observation is a big challenge. In the example shown
here the very low kinetic energies have been achieved by crossing two well cooled
and velocity selected beams at an angle of 45◦ (instead at the usual 90◦ ).
We note in passing, that similar resonances are also known in molecular
spectroscopy, there called predissociation, which we have already mentioned in
Sect. 3.6.7. They are formed e.g. when a molecule is excited into a quasi-bound
state, shortly above the nominal dissociation energy, just as sketched in Fig. 6.25
(so called shape resonances). Another type of resonance are the so called F ES -
HBACH resonances which occur shortly below bound states, e.g. by rotational or
vibrational excitation of molecules or in heavy particle collisions or in electronic
excitation of atoms by electron impact. We shall come back to these in Sect. 6.5.
6.3 Elastic Scattering and Classical Theory 413
500
Veff ℓ =7
ℓ =6
ℓ =5
R
ℓ =7
ℓ =6
100 ℓ =5 H + Xe
50
200 500 1000 2000 5000 10000
relative velocity u / m s-1
Fig. 6.25 Orbiting resonances in the integral elastic cross section for H-Xe scattering from two
different experiments (experimental points with error bars) in comparison with quantum mechan-
ical model calculations (lines) according to T OENNIES (2007). The inset (top right) illustrates
schematically the origin of these resonances in the effective potential Veff for angular momenta
> 0
Here T (b) is a transmission function – in general complex – for the collision pro-
cess studied. It also accounts for the range of the interaction potential. For a hard
sphere of radius a by which a point like projectile is scattered, the integral leads to
the same result (1.68) as in the optical case, with an opening angle of the central
diffraction disc on the order of θ1 = 0.61λ/a. In the past, experimental observations
of this phenomenon have occasionally been reported, mostly, however, accompa-
nied by some controversial discussion (see e.g. G EIGER and M ORÓN -L EÓN 1979;
B ONHAM 1985). The experimental requirements on angular resolution are just ex-
treme.
414 6 Basics of Atomic Collision Physics: Elastic Processes
y 2.5
x Li + (b) 0
z be - 2.5
am
-2.5 0 2.5 -2.5 0 2.5
MOT coils pz / a.u.
u = 0.15 a.u.
0
0.135 a.u.
0 0 0.40 a.u.
0.0
04 0.04
0.04
0 0 0 0
θz / de
eg θy / deg 0.28 a.u.
-0.04 -0.04
Fig. 6.26 F RAUNHOFER diffraction patterns in very small angle scattering for the charge ex-
change process 6 Li+ + 23 Na(3s) → Li(3s) + Na+ according to VAN DER P OEL et al. (2002).
(a) Experimental setup with MOT (the heavy red arrows mark the laser beams that cool the Na
atoms and position them in the trap), (b) experimental raw data, (c) reconstructed differential cross
sections for small scattering angles θ for different velocities in a.u. (energies 2.4 keV to 40 keV)
For the charge exchange 6 Li+ + 23 Na(3s) → Li(3s) + Na+ such F RAUNHOFER
diffraction rings have been reported by VAN DER P OEL et al. (2002) for collision
energies from 2.4 keV to 40 keV. If the typical interaction radius is assumed to be
a 5 to 10a0 , diffraction angles θ1 in the range 0.01◦ –0.005◦ are expected. To be
able to resolve such structures cutting edge measurement technology is required.
Figure 6.26(a) shows the scheme of the experimental setup used. With the help of
a so called reaction microscope a complete momentum analysis has been achieved,
detecting both collision partners in coincidence with position and temporal analysis
after the collision.7
To achieve this analysis with sufficient precision, one has to use a well collimated
Li+ projectile beam (the angular resolution was <0.3◦ ), and also the Na target gas
has to be extremely cold, so that the initial state is overall very well defined. In this
experiment a so called magneto optical trap (MOT) has been used, which achieved
7 For details about this extremely powerful detection method, originally introduced as COLTRIMS
(Cold Target Recoil Ion Momentum Spectroscopy), we refer the interested reader to the fundamen-
tal review by U LLRICH et al. (2003). See also Appendix B.4.
6.3 Elastic Scattering and Classical Theory 415
temperatures <1 mK. The combination of all these techniques finally provided the
necessary momentum and angular resolution. In Fig. 6.26(b) we show the raw data
for two velocities. A three dimensional reconstruction of the angular distribution on
the basis of the experimental data is shown in Fig. 6.26(c). The impressive, charac-
teristic diffraction patterns are indeed very similar to optical diffraction. They allow
very critical tests of the relevant quantum mechanical computations of the charge
exchange process.
As an important final topic we want to properly recognize the significance
of the so called symmetry oscillations. Schematically their origin is illustrated in
Fig. 6.24(f). They occur due to the superposition of forward and backward scatter-
ing for systems with two identical particles which cannot be distinguished quantum
mechanically. Since only the relative velocity of the two particles can be measured
and both particles leave the collision in opposite directions, we do not know which
particle goes forward and which backwards. In fact, the magnitude of the scattering
amplitude for forward and backward scattering is identical and there is no way to
distinguish which particle is scattered forward and which backward. Hence, corre-
sponding to (6.58) one expects that the differential cross section becomes
2
I (θ, T ) = fj (θ ) ± f (π − θ ) . (6.60)
According to the general symmetry rules, the positive sign holds for bosons, the
negative sign for fermions. Since the phase changes with the scattering angle this
gives again rise to characteristic oscillations which of course may also be observed
in the integral elastic cross section. The pioneering experiment on this fascinating
phenomenon has been carried out by F ELTGEN et al. (1982). Unfortunately it has,
over the years, remained without much attention by main stream atomic physics.
Figure 6.27 illustrates the experimental setup to study the integrated elastic cross
section σel for He + He scattering.
As one easily recognizes: to carry out such experiment requires heroic efforts. It
requires among other sophisticated technology, two low temperature cryostats and
affords highest professional skills and frustration tolerance of the experimenters.
The target gas should be as cold as possible, and good velocity selection of the
projectile beam is required (as indicated by M, mechanical selectors are used, based
on the concept used by F IZEAU to measure the velocity of light: the molecular beam
is chopped with fast rotating slit discs). High angular resolution must be achieved
to limit the kinematic uncertainties and to allow for a clear observation of the cross
section as a function of relative velocity8 u.
The experimental results are summarized in Fig. 6.28. 4 He as well as 3 He have
been studied in the various possible combinations. The symmetry oscillations are
clearly visible as a function of the velocity for the symmetric systems 4 He + 4 He
8 Actually, F ELTGEN et al. (1982) give the lab velocity of the primary beam, which is essentially
identical to the relative velocity u since the target is very cold, i.e. at rest.
416 6 Basics of Atomic Collision Physics: Elastic Processes
(a) GB CB
GA CA C1
RS
RS V
C2 C3
BB
D
( (
((( (((
SC RS SEM
M BC
S C1 SC C2 C 3
(b)
Ø6 Ø3 Ø5 4.5×4.5
4×3 384 178 12 656 265
Fig. 6.27 Experiment for low energy elastic He–He scattering designed by F ELTGEN et al. (1982)
as a characteristic example for glory oscillations according to the scheme Fig. 6.24(f). (a) Details
of the apparatus GA , GB : gas inlet for A and B, respectively; CA , CB : cryostat; S: primary source
for the projectile A; M: mechanical velocity selector (see text); SC: scattering chamber; D: detector
(ionization and deflectors); SEM: secondary electron multiplier; RS: radiation shield; BC: beam
chopper; BB: beam blocker; C1 –C3 : collimating apertures; V: isolating valve. (b) Dimensioning
of the apparatus in mm
η 0= 0 −N=
1 2 3 4 5 6
σel / arb. un.
Fig. 6.28 Integral elastic cross section σel for He–He scattering according to F ELTGEN et al.
(1982) as a function of the primary beam velocity (i.e. in this case the initial relative velocity u
of the colliding particles). The so called g − u oscillations are very pronounced for the symmetric
pairs 4 He + 4 He and 3 He + 3 He, but completely disappear for 3 He + 4 He – in spite of completely
identical interaction potential in all three cases
terference between primary particles and secondary particles which are scattered
backward by the repulsive part of the potential.” To fully appreciate this astonish-
ing result, we remind the readers that the atomic charge clouds of the interacting
particles – which are finally responsible for the interaction potential – are identi-
cal in all three cases. The decisive difference is, that 4 He is a boson (nuclear spin
0) and 3 He a fermion (nuclear spin 1/2), i.e. 4 He2 and 3 He2 obey pure B OSE -
E INSTEIN or F ERMI -D IRAC statistics, respectively, and the scattering amplitudes
have to be added according to (6.60) with a plus or minus sign, respectively. In con-
trast, 4 He+ 3 He are distinguishable particles and squared amplitudes must be added.
For more detail, such as the velocities for which s and p wave phase shifts (η0 and
η1 ), respectively, go through zero, or for the meaning of the oscillation index −N ,
the readers are referred to the original publication.
Undoubtedly these experimental observations touch upon the most subtle and
basic aspects of quantum mechanics: none of the known interactions communicates
among the interacting particles their bosonic or fermionic character: apart from be-
ing bosons or fermions, the ingredients of the interaction potential are identical in
all three cases. And nevertheless the collision partners ‘know’ from the very begin-
ning of the collision (i.e. essentially at infinite distance) how they have to arrange
their wave function: symmetric or antisymmetric or not at all. Of course the exper-
imental observations may be predicted precisely according to the rules of quantum
mechanics. One just needs to apply the proper recipe: that for bosons or fermions
or distinguishable particles. But why the particles do, what they do, is not a ques-
tion to be asked. Clearly, the physics at work here is fundamentally the same as that
needed to construct the periodic system of elements: only the fact that electrons are
fermions makes atomic matter possible in the form as we know it. However, the
present experiment may even be seen somewhat more crucial. Here we do not dis-
cuss bound states (i.e. particles which are essentially closely spaced at all times).
We discuss collision processes, i.e. continuum states which extend essentially over
infinite space.
Section summary
• Classical scattering theory provides a good first approach towards understand-
ing many important phenomena in elastic, low energy heavy particle collision
physics. The differential elastic cross section is derived in a straight forward
manner through (6.36), simply by evaluating (6.42) for the classical trajectory
Θ(b, T ) as a function of kinetic energy T and impact parameter b. Interesting
singularities (rainbow scattering) are predicted (and experimentally observed)
for |dΘ/db| = 0.
• In the limit of small scattering angles the classical differential elastic cross
section I (θ, T ) may be evaluated explicitly to give (6.47). It turns out to be
useful to introduce a reduced scattering angle τ (b, T ) = θ T and a reduced
cross section ρ(b, T ) = θ sin θ I (θ, T ) which (at high enough T ) both depend
only on the impact parameter b.
418 6 Basics of Atomic Collision Physics: Elastic Processes
As we have just seen, classical theory does not suffice to describe all scattering
processes and the broad variety of experimentally observed phenomena fully. For
electron scattering this is already evident according to (6.35) as a consequence of
the large DE B ROGLIE wavelengths λdB . In heavy particle scattering interference
effects force us to use a quantum mechanical interpretation. In principle, quantum
mechanics has to be used when λdB = 2π/k = h/(M̄u) is comparable with charac-
teristic dimensions of the interaction. In this section we shall continue to focus on
elastic scattering and introduce the basic quantum mechanical concepts of scatter-
ing amplitudes and partial wave analysis, and discuss the behaviour for scattering
phases in some characteristic cases.
Here k = p/ is the wave vector of the relative particle motion, T = 2 k 2 /2M̄ the
initial kinetic energy in the CM system. In this stationary picture one describes the
projectile beam (more precisely the relative motion of the particles with reduced
mass M̄) as a plane wave propagating into direction k i , and the CM is the origin of
an outgoing spherical wave. We thus want to find asymptotic solutions for (6.61) of
the type
eikf R
ψ(R) −→ ψ (ki ) (R) + f (θ, ϕ) with ψ (k) (R) = eik·R . (6.62)
R→∞ R
6.4 Quantum Theory of Elastic Scattering 419
This normalization of ψ (k) (R) follows the most common practice in scattering
physics (see e.g. B URKE 2006).9 As well known and easy to verify, ψ (k) is a solu-
tion of the (homogeneous) free particle wave equation
2M̄
∇ 2 ψ(R) + k 2 ψ(R) = V (R)ψ(R). (6.63)
2
To find a special solution one uses the corresponding G REEN’s function G(R)
eikR
G(R) = − , a solution of ∇ 2 G(R) + k 2 G(R) = δ(R), (6.64)
4πR
and obtains as a formal, general solution for the scattering problem:
2M̄
G R − R V R ψ R d3 R .
(k i )
ψ(R) = ψ0 (R) + 2 (6.65)
This is the so called L IPPMANN -S CHWINGER equation in coordinate representa-
tion. By inserting G(R), the full solution ψ (+) is then derived from
ikf |R−R |
1 M̄ e
ψ (+) (R) = eiki ·R − V R ψ (+) R d3 R . (6.66)
2π 2 |R − R |
The scattering signal is detected at very large distances R, for which
R − R −→ R 1 − eR · R + · · · ,
R→∞
9 For plane wave normalization see also Appendix J.2 in Vol. 1. We note that the present, standard
scattering theory normalization of the plane wave differs by a factor (2π)3/2 from normalization
in k scale. For determining cross sections this is irrelevant, since particle fluxes are finally normal-
ized to each other. However, normalization in k scale is also used quite often in the literature (see
e.g. B RAUNER et al. 1989) and leads to slightly different prefactors in the expressions for scatter-
ing amplitudes, T-matrix elements and cross sections. We shall occasionally point this out where
appropriate.
420 6 Basics of Atomic Collision Physics: Elastic Processes
With 2 /me = Eh a02 and ψ (+) defined dimensionless, while the interaction potential
V has the dimension Enrg, one finds that the scattering amplitude is of dimension L.
Note that (6.68) for the scattering amplitude is still exact if ψ (+) is an exact solution
of the S CHRÖDINGER equation. All high energy approximations start at this point
by substituting reasonable approximations for ψ (+) (R).
M̄ M̄
f (θ, ϕ) = − k f |V |ψ (+) = − k f |T|k i . (6.69)
2π2 2π2
were |k i represents the initial plane wave and |ψ (+) the full solution of the scat-
tering problem. The T-matrix is a useful tool in formal scattering theory, as we shall
see in Sect. 6.4.6. Its elements have the dimension Enrg × L3 .
The second term in (6.62) represents the scattered particle. The flux which
reaches the detector is10
k f |f (θ, ϕ)|2
Φf = 2
, so that (6.71)
M̄ Rdet
Ṅ = Φ f · Adet = |Φf |Rdet
2
Ω (6.72)
scattered particles Ṅ per unit of time and per scattering atom reach the detector
in a distance Rdet from the scattering centre. Here Adet = Rdet2 Ω is the detector
area for an acceptance angle Ω. We finally normalize to the initial particle flux
Φ i = k i /M̄ and obtain from the definitions (6.20) and (6.21) the DCS:
dσ Ṅ kf 2
I (θ, ϕ) = = = f (θ, ϕ) .
dΩ Φi Ω ki
incident
particle beam p = ħk
b R z (col)
y (col) ℓ
+Θ
e i kz classical
x (col) trajectory
Fig. 6.29 Standard collision frame, with the x (col) z(col) scattering plane defined by the position
coordinate R and the momentum p = k prior to collision; note that y (col) points here perpen-
dicularly into the scattering plane. With the impact parameter b the magnitude of the angular
momentum is || = = bp
For general reference, we have included the indices i and f referring to initial and
final states of the whole interacting system, respectively.11 Specifically, for elastic
scattering, the magnitudes of the wave vectors are kf = ki = k.
11 With plane wave normalization in k scale the right side of (6.73) is written in a.u.
= R × p = R × k (6.75)
and defines also the collision plane to which it is perpendicular. For spatially
isotropic potentials one usually chooses the direction of the incident particle beam
as quantization axis z(col) and to be perpendicular to the x (col) z(col) collision plane,
i.e. parallel to y (col) . Angular momentum conservation leads to an important conclu-
sion: elastic collisions occur exclusively in the plane defined by linear momentum
and scattering centre. The final momentum also lies in this plane. This still holds for
inelastic collisions – if the interaction potential is isotropic prior to collision. The
scattering plane may change, however, if angular momentum is transferred onto the
collision partner.
For the magnitude of the angular momentum
|| = = bp = bk = b 2M̄T holds, (6.76)
and thus b = /k. (6.77)
Hence, collisions with impact parameters between 0 ≤ b < 1/2k may be viewed as
dominated by s waves, for 1/2k ≤ b < 3/2k by p waves and so on. Adapting this to
quantum mechanics, one sets
bk = ( + 1) ¯ = ( + 1/2) (6.78)
One thus solves the S CHRÖDINGER equation (6.61) by the usual separation ansatz
u (R)
ψ(R) = Ym (θ, ϕ) (6.79)
R
with the spherical harmonics Ym , where and m are as usual the quantum num-
bers for magnitude and orientation of the angular momentum in space.12 As in the
case of bound states, the radial equation is written as
2
d 2M̄ ( + 1)
+ k − 2 V (R) −
2
u (R) = 0. (6.80)
dR 2 R2
12 Here and in the following we use the notation for the angular momentum of the relative nuclear
motion and call its quantum number – as usual done in this context – and we have to accept
the little inconsistency with respect to denoting the molecular angular momentum by N and its
quantum number by N , according to Chap. 3.
6.4 Quantum Theory of Elastic Scattering 423
R R
potential well potential barrier
(attractive) (repulsive)
Fig. 6.30 Illustration of the scattering phase η for a (hypothetical) one dimensional model prob-
lem: (a) attractive and (b) repulsive potential (heavy black lines). The full red lines represent the
scattered waves, the dashed red lines the unperturbed waves
We have written the solutions of (6.80) for vanishing potential V (R) → 0 as spher-
ical B ESSEL functions j (kR) = u0 (kR)/kR.
424 6 Basics of Atomic Collision Physics: Elastic Processes
For the evaluation of the scattering processes we are simply interested in the
asymptotic behaviour for R → ∞ (see Appendix J, Vol. 1). For vanishing potential
the radial wave function is given by:
(0) π
u (R) ∝ sin kR − . (6.83)
R→∞ 2
Among the general solutions u (R) of (6.80) we now have to select those which
describe the scattering function (6.62) correctly as a partial wave expansion (6.81).
Asymptotically we expect
π
u (R) ∝ sin kR − + η (6.84)
R→∞ 2
π π
∝ sin kR − + tan η cos kR − .
R→∞ 2 2
This differs from the free, plane wave (6.83) simply by the scattering phase shift
η , which depend of course on . The partial wave expansion of the scattered wave
is obtained from the definition (6.62) by subtracting the plane wave (6.82) from the
full wave function (6.81):
eikR
ψ (+) (R) − ψ (ki ) (R) → − f (θ ). (6.85)
R
With the asymptotic behaviour (6.84), (6.83) and with some algebra we obtain fi-
nally the elastic scattering amplitude for an isotropic potential
∞
1
f (θ ) = (2 + 1) e2iη − 1 P (cos θ ). (6.86)
2ik
=0
With (6.73) follows from this the differential and with (6.74) the integral cross sec-
tion. We have thus derived a direct relation between measurable observables and the
S CHRÖDINGER equation (6.61).
Summarizing, the task of the quantum theory of elastic scattering consists in
determining the phase shifts η between the partial waves of angular momentum
with and without potential. The square of the scattering amplitude f (θ ) according
to (6.86) provides the differential cross section. Integration over all scattering angles
leads us to the integral elastic cross section:
∞
4π
σel = (2 + 1) sin2 η . (6.87)
k2
=0
Finally, we note that for θ = 0 and P (1) = 1 the scattering amplitude (6.86) be-
comes
∞
1
f (0) = (2 + 1) sin η eiη . (6.88)
k
=0
6.4 Quantum Theory of Elastic Scattering 425
4π
σel = Im f (θ = 0). (6.89)
k
As just discussed, to determine the elastic cross sections we have to solve the radial
equation (6.80) and to determine the scattering phases η from the asymptotic solu-
tions (6.84). For elastic scattering of electrons with a few eV kinetic energy this is
not too difficult (if the scattering potential is sufficiently well known or can be cal-
culated to some reasonable approximation). According to (6.76) only a few partial
waves will be relevant.
However, if the angular momenta can be large – and for heavy particle scatter-
ing this almost always the case – then a fully quantum mechanical computation
of all scattering phases becomes an extremely cumbersome task. Thus, appropriate
approximations are required. Semiclassical procedures are the method of choice if
the DE B ROGLIE wavelength λdB = 2π/k is small compared to typical dimensions
over which the potential changes and at the same time
dλdB
1 (6.90)
dR
holds during the collision process. Then, classical theory which computes for each
impact parameter (6.77) b = /k a trajectory is a good first approximation. To in-
clude interference effects, in the so called eikonal approximation one computes the
phases along all these trajectories in the effective potential
2 ( + 1) b2 2 ( + 1/2)2
Veff = V (R) + = V (R) + T 2 V (R) + , (6.91)
2M̄R 2 R 2M̄R 2
where we have applied (6.78). The effective kinetic energy Teff = T − Veff (R)
changes slowly along the trajectory, and the wavenumbers for the free spherical
wave
k and the partial wave k in the scattering potential V (R) are
( + 1/2)2
k (R) = k2 − and (6.92)
R2
2M̄ ( + 1/2)2
k (R) = k2 − V (R) − , respectively. (6.93)
2 R2
426 6 Basics of Atomic Collision Physics: Elastic Processes
with the classical turning points R = Rc given by the root of (6.92) and by R̃ =
b = ( + 1/2)/k. In practice, due to the singularities at the classical turning point,
the computation of the JWKB phases is not trivial (see e.g C OHEN 1978, for efficient
computational methods).
For high kinetic energies T
V (R) one may expand the first term of (6.95) in
powers of V (R)/T and obtains the J EFFREYS -B ORN phase
∞
k V (R)
η (T , b) = −
JB
dR. (6.96)
2T b 1 − b2 /R 2
To see the connection between quantum mechanics and classical trajectory one
exploits the asymptotic expansion of the L EGENDRE polynomials,
→∞ 2 1 π
P (cos θ ) −→ cos + θ−
π sin θ 2 4
for large , and obtains the scattering amplitude (6.86) conveniently by integration,
rather than by summation:
∞√
1
f (θ ) = √ eiΦ+ () + eiΦ− () d (6.97)
ik 2π sin θ 0
π
with Φ± () = 2η ± θ ∓ . (6.98)
4
If one expands the phase Φ± () around a well defined value of = 0
(0) dη
Φ± () = Φ± + 2 ± θ ( − 0 ) + · · · ,
d 0
one sees that the integral in (6.97) usually averages out to zero, due to fast oscilla-
tions as a function of . Only for such 0 for which
dΦ± dη
=2 ±θ ≡0 (6.99)
d d 0
phase shift.
6.4 Quantum Theory of Elastic Scattering 427
π Θ
Θ ηℓ bg
br
θ
0
-θ b, ℓ
- θr
ηℓ
b'1 b 2 b3
Φ+ Φ- Φ+
0
b, ℓ
Φ-
Fig. 6.31 Classical deflection function Θ(b) (red) and scattering phase η (black) as a function
of impact parameter b and angular momentum = bk, respectively. Below, the corresponding
phases Φ+ () (dotted) and Φ− () (dashed) according to (6.98) are shown. Marked are the impact
parameters and scattering angles for glory scattering bg and for the rainbow br at the maximum
and turning point of the scattering phase ηl , respectively. Also indicated are the three points of
stationary phase at b1 , b2 and b3 which contribute to the scattering amplitude for a given scattering
angle θ . They can be related directly to the corresponding classical trajectories (schematically)
shown in Fig. 6.21
holds, the integral in (6.97) remains finite. These values of Φ± (0 ) are called sta-
tionary phases.
This remarkable result allows us, in one further step, to draw a direct connection
between partial wave expansion (6.97) and classical trajectory: we evaluate the slope
of the phase shift explicitly in JWKB approximation (6.94) (with ¯ = + 1/2):
∞ ¯ ∞ ¯
dη dR dR
= − (6.100)
d ¯2 ¯2
b R 2 k 2 − R 2 Rc R 2 k 2 − 2M̄ V (R) −
2 R2
∞
π dR Θ(b)
= −b √
2 1 − V (R)/T
= . (6.101)
2 Rc R eff 2
The last step follows directly from comparison with the classical deflection func-
tion Θ(b) according to (6.42). Inserting this into (6.99) one recognizes: stationary
phase conditions are realized exactly when the scattering angle θ is identical to the
classical deflection function ±Θ(b). This is a rather nice result: only such partial
waves (or impact parameters) contribute to the scattering amplitude f (θ ) according
to (6.97) which are close to the corresponding classical trajectory.
Schematically this is illustrated in Fig. 6.31 for one scattering angle θ – arbitrarily
chosen here close to the rainbow angle θr . The figure is so to say the semiclassical
equivalent to Fig. 6.21. The thin, vertical, red lines indicate the impact parameters
428 6 Basics of Atomic Collision Physics: Elastic Processes
for which the phases Φ+ or Φ− have a maximum or minimum (i.e. are “stationary”)
and hence contribute to the scattering amplitude.
It should be pointed out that these semiclassical methods are very powerful and
allow one to describe elastic heavy particle collisions (as discussed in Sect. 6.2), as
accurately as demanded by the experiments.
14 When evaluating (6.102) one has to be careful with the arctan function which is multiple valued.
6.4 Quantum Theory of Elastic Scattering 429
0 ka π ηs
-─
0 5 10 5
tering phase shifts are in principle only meaningful modulo 2π one defines:
η −→ 0. (6.103)
k→∞
Figure 6.32 illustrates some general trends of the scattering phases ηs , ηp and ηd
( = 0, 1, 2) at low energies for different attractive (left) and repulsive (right) poten-
tial well depths and barrier heights, respectively. As discussed already in Sect. 6.4.3
and illustrated in Fig. 6.30 the phase shift is positive (negative) in an attractive (re-
pulsive) potential; we have to add that this holds at moderate energies and not too
high .
The absolute value of the phase shift grows usually rather rapidly with k (or with
energy), reaches a maximum and finally falls to zero as defined by (6.103). There
are, however, a few remarkable exceptions in the case of an attractive potential if
that is sufficiently deep: while for small well depths (k0 a = 1.5) all phases start at
η (ka = 0) = 0, for k0 a = 4.4 the s phase as well as the p phase has the limit π for
vanishing k. And for the deepest potential well shown here (k0 a = 6) the s phase
shift at energy zero corresponds already to a full wavelengths (ηs = 2π ). The shifts
for p and d waves start in this case at ηp (0) = ηd (0) = π . In contrast, the repulsive
potential (right in Fig. 6.32) shows no such peculiarities.
rem holds:
η (k) −→ N π. (6.104)
k→0
Hence, Fig. 6.32 shows us that one bound s and one bound p state can exist in a
potential well with k0 a = 4.4. For k0 a = 6 even two bound s states as well as one
p and one d state exist. Very interesting is also the behaviour of the d phase shift in
the middle panel left in Fig. 6.32: for k0 a = 4.4 obviously a bound d state does just
not yet exist – in the limit η2 (0) = 0. However, the d phase shift rises very rapidly
at about ka 0.7 from 0 to nearly π . One interprets this as due to a quasi-bound
state in the continuum. We have seen such resonance phenomena already, e.g. in
Fig. 6.25 as orbiting resonances in the heavy particle scattering, and we shall treat
scattering resonances in more detail in a moment. Here we simply note that such
rapid changes of a scattering phase (by nearly π ) must lead according to (6.86) and
(6.87) to pronounced structures in the DCS and ICS.
The dominance of low values at low energies, as illustrated in Fig. 6.32 for
the one dimensional case, suggests that a corresponding series expansion might be
useful. Such expansions are known as effective range expansion valid under certain
conditions in the general case (see e.g. OM ALLEY et al. 1961, and further references
therein):15
1
k 2+1 cot η = − + k 2 r + O k 4 . (6.105)
a
Here, a and r are constants. For the s phase shift they are called scattering length
(as ) and effective range (rs ), respectively, and the scattering length is positive for
repulsive and negative for (not too strongly) attractive potentials.
In the limit of vanishing kinetic energy T → 0 one may combine (6.104) and
(6.105) to:
η −→ −k 2+1 a + N π. (6.106)
k→0
Thus, for very low energies (e.g for collisions of ultra-cold atoms at temperatures
below µK, i.e. for kinetic energies T < 10−11 eV!) practically only s scattering is
relevant. Then ηs → −kas and the scattering amplitude (6.86) becomes
1 −2iηs ηs
f (θ, k) −→ e −1 −as so that I (θ ) → as2 . (6.107)
k→0 2ik k
Hence, elastic DCS and ICS according to (6.73) become
dσel
−→ a 2 and σel −→ 4πas2 . (6.108)
dθ k→0 s k→0
15 Today,owing to the ever increasing experimental precision and large data collections improved
concepts for effective range expansions exist (see e.g. G ULLEY et al. 1994; R AKITYANSKY and
E LANDER 2009).
6.4 Quantum Theory of Elastic Scattering 431
ηs
as<0 RAMSAUER minimum
π
as > 0 attractive
I
as<0 III II
0
IV k
repulsive
as > 0
Fig. 6.33 Four different possibilities for the behaviour of the s scattering phase shift ηs at low
energies. According to the effective range expansion (6.105) the scattering lengths as is derived
from the slope of ηs (k) as a function of k in the limit (6.106) for very small k
The integral cross section at these low kinetic energies is thus finite! Interestingly,
we recall that for elastic scattering of a point-like particle by a hard sphere of di-
ameter d the scattering length is just as = d/2 (the phase shift of a spherical wave
originating from the centre of the sphere is −η = kd/2 at the surface of the sphere).
Hence, the effective elastic ICS of a sphere is at low energies according to (6.108)
σel = πd 2 , i.e. four times its geometrical cross section (6.54) which is predicted by
classical theory. By the way, this also holds for light waves diffracted by a small
disc. In both cases, interference effects are responsible.
With increasing energy the phases change. For s phases Fig. 6.33 sketches
schematically four different types of possible behaviour for ηs (k) at low energies.
Here terms up to k 3 are considered. Different magnitudes and signs of as and rs and
a different number of bound states is assumed (one for I and II each, none for III
and IV). In all cases the integral elastic cross section according to (6.108) is finite
for k → 0. One particularly interesting case is displayed in curve I: at a particu-
lar value of k the s phase shift ηs passes through π , so that sin ηs = 0. Hence, the
integral elastic cross section will assume there a minimum, it may even disappear
nearly completely since all other phases in (6.87) may still be very small due to the
k 2+1 behaviour. This effect has first been discovered by R AMSAUER in 1921 and
we have seen impressive examples for it already in the introduction to the present
chapter (see Fig. 6.8). We can now understand this quite remarkable phenomenon:
as a transition of the scattering phase shift through N × π (usually the s phase).
Fig. 6.34 Partial wave analysis for e− − He scattering according to A NDRICK and B ITSCH (1975)
and e− − Ne according to G ULLEY et al. (1994) and A DIBZADEH and T HEODOSIOU (2005).
(a)–(d): measured differential cross sections I (θ) (for He the error bars are on the order of the
linewidths); (e)–(i) phase shifts η derived from a corresponding partial wave analysis for the s, p,
and d waves (for Ne also f )
DCS one set of phase shifts may be fitted. For e− − He scattering one finds that
in the energy range from 0 to about 15 eV excellent agreement is obtained by just
including the s, p and d phase shifts as documented by Fig. 6.34(e) and (f), while
for e− − Ne scattering, f waves have to be included at even lower energies.
As a side remark we note: the s wave phase shift for e− − He of π at zero
energy shown in (e) does not allow the conclusion that a He anion exists! While the
L EVINSON theorem (6.104) in principle supports an additional bound s state (with
n = 1), helium already contains two bound (1s)2 electrons, i.e. the 1s shell is fully
occupied, and according to the PAULI principle no additional (low energy) electron
can be attached. A similar argument holds for Ne. The limiting s and p phases are
2π and π (Fig. 6.34(g) and (h)), respectively, but due to the PAULI principle no
extra electron can be added to the full Ne (2s)2 or (2p)6 shells: thus no stable neon
anion exists.
We shall now introduce a somewhat formal, but very important concept of scatter-
ing theory. In the spirit of this book we shall try to keep the mathematics simple and
approach the subject somewhat heuristically by discussing Fig. 6.35 which schemat-
ically illustrates the key elements of a scattering experiment.
6.4 Quantum Theory of Elastic Scattering 433
M̄ M̄/me
ff i (θ, ϕ) = − k f |T|k i = − k f |T|k i (6.110)
2π2 2πEh a02
is simply the probability amplitude for finding |k f in the scattered wave T|k i , i.e.
one projects former onto the latter. The differential cross section is then given by
(6.73), with ki = kf = k in the elastic case. We emphasize again, that the above T-
matrix element in (6.110) has the dimension L2 Enrg so that ff i has the dimension L.
A partial wave expansion of the T-matrix is found by expanding the plane waves
before and after the scattering process. Exploiting the orthogonality relations of the
spherical harmonics Ym one obtains the scattering amplitude:16
2π ∗
ff i (θ, ϕ) = i i− T m m Y m (θ, ϕ)Ym
(θi , ϕi ). (6.111)
kf ki
m m
T m ,m (k) are the matrix elements of T, while θ, ϕ and θi , ϕi characterize the
direction of the scattered plane wave |k f (as seen by the detector) and the incoming
wave |k i , respectively. The latter is usually chosen to be parallel to the z-axis, so
that θi = ϕi = 0 and m = 0 holds. It is instructive to have a look at the dimensions in
16 With plane wave normalization in k scale, the T-matrix elements have to be multiplied by a factor
−2πi.
434 6 Basics of Atomic Collision Physics: Elastic Processes
(6.111). Since the wave vectors lead to a dimension L which also is the dimension of
the scattering amplitude, the T-matrix elements obviously must be dimensionless –
in contrast to the one in (6.110) (the dimension of k f |T|k i is Enrg L3 ). This is due
to the fact that these matrix elements are obtained from radial wave functions which
are normalized in energy scale according to (J.11) in Vol. 1. Integration over dR and
the dimension of T (which is Enrg) cancel the dimensions of this normalization.
For elastic scattering by an isotropic potential we may compare this with the
partial wave expansion (6.86) which is diagonal in . Hence
T = e2iη − 1. (6.113)
In deriving this relation we have used the addition theorem of the spherical harmon-
ics (C.22), Vol. 1.
Alternative to the T-matrix one may also use the S-matrix, for which holds in the
isotropic, elastic case
S = e2iη = 1 + T . (6.114)
So far,
S and T matrices simply offer a trivial possibility to rewrite the partial
wave expansion (6.86) for the scattering amplitude:
1
f (θ ) = (2 + 1)P (cos θ ) S (k) − 1 (6.115)
2ik
1
= (2 + 1)P (cos θ )T (k). (6.116)
2ik
However, for the general case of non-isotropic potentials and/or inelastic scattering
the angular momenta of partial waves are coupled to the eigenstates of projectile and
target. Hence, T and
S are genuine matrices with nonvanishing off-diagonal terms.
They allow a transparent representation of the scattering process by (6.111). In that
case one defines17
S =
1 + T with
SS† =
1. (6.118)
The scattering matrix S is thus defined as a unitary operator. In view of (6.109) this
unitary relation expresses nothing else but the conservation of particle flux during
the scattering process.
The S and T operators allow to formulate the symmetry of a scattering process
with respect to incoming and outgoing plane waves in a clear manner: we may
invert the scattering process by simply inverting the beam directions. In Fig. 6.35
the source would become detector and vice versa. We then have to replace (6.109)
by the time inverse equation:
|k f → |k f + T† |k f =
S † |k f . (6.119)
M̄
fi←f (θ, ϕ) = k i |T† |k f (6.120)
2π2
M̄
= k f |T|k i ∗ = ff∗←i (θ, ϕ).
2π2
For later use we finally note an alternative representation of the asymptotic solutions
for the partial waves (6.84) as superposition of in and outgoing waves:
π
u (R) → sin kR − + η = a∗ e−ikR + a e+ikR
2
(6.121)
−ikR+i π2 +ikR−i π2 π 1 ikR−i π2
∝e − S e ∝ sin kR − − T e
2 2
1 π a
with a = exp −i + iη and S = − ∗ eiπ = e2iη . (6.122)
2i 2 a
Section summary
• Quantum mechanical scattering theory has to find solutions of the S CHRÖDIN -
GER equation which asymptotically are a superposition (6.62) of an incom-
ing plane wave and outgoing spherical wave. The amplitude f (θ, ϕ) of the
latter is called scattering amplitude. The DCS is (kf /ki )|f (θ, ϕ)|2 – with
(kf /ki ) = 1 for elastic scattering.
• Solutions for the scattering problem can be obtained either from the L IPP -
MANN -S CHWINGER equation (6.65) or by a partial wave expansion into a
series of spherical harmonics Ym (θ, ϕ). Angular momenta may become
large. The asymptotic phase shifts η between the partial waves in the scat-
tering potential and the free outgoing spherical wave determine the scattering
amplitude (6.86).
• Classical and quantum picture of a collision process are related by impact pa-
rameter
√ b, wavenumber k and angular momentum quantum number through
bk = ( + 1) ( + 1/2).
436 6 Basics of Atomic Collision Physics: Elastic Processes
6.5 Resonances
6.5.1 Types and Phenomena
This is illustrated schematically in Fig. 6.36 for two most common resonance
types. Figure 6.36(a) shows the characteristic potential for the formation of a so
called shape resonance: for angular momenta > 0 the centrifugal barrier may be
high enough to allow for one or more quasi-bound states in the effective potential
Veff (R) – above the asymptotic zero energy. Clearly, these states are not stable: their
finite lifetime τ = /Γ is determined by the tunnelling probability Γ through the
centrifugal barrier. Nevertheless, they may influence the scattered wave significantly
if the relative kinetic energy of the system is nearly resonant with the energy Wres
of the quasi bound state. The centrifugal barrier may be tunnelled (into either direc-
tion), so that the quasi-bound state may be occupied temporarily. When the kinetic
energy passes such a resonance the scattering phase jumps by nearly π – as we have
already discussed in Sect. 6.4.5 for a potential well. And we have already identified
there in Fig. 6.32 such a phase jump, e.g. for k0 a = 4.4 in the d partial wave.
A somewhat different situation is sketched in Fig. 6.36(b). This so called F ESH -
BACH resonance manifests itself as a quasi-bound state in the potential of an excited
state which is embedded into the dissociation or ionization continuum of the initial
state. The resonance would thus form a stable, bound state of the excited scattering
system – if there were no coupling between this state and the collision continuum.
In this situation too a phase jump is encountered by π within a small range of kinetic
energies. It leads to pronounced interference structures in the DCS as well as in the
ICS.
A little warning at this point appears in order when discussing such concepts for
electron scattering: The two ‘potential images’ shown in Fig. 6.36 have to be used
with care, since the potential which an electron experiences when scattered (e.g.
by an atom) involves also all electrons of the target. In such a situation – due to
the similar velocities of projectile electron and the electrons in the target atom –
no such thing as a B ORN -O PPENHEIMER approximation is strictly valid. At best,
one may resort to the independent particle model which has proved very valuable
when computing the wave functions of complex atoms (see Sect. 10.1, Vol. 1) – i.e.
one might average the interaction potential over the wave functions of all atomic
electrons. Such pseudopotentials are indeed used in the theory of electron scattering
at intermediate energies and often turn out to be quite successful. In the general
case, however, especially for excitation processes, one has to treat the problem with
more rigorous methods which we shall outline in Sect. 7.3.
438 6 Basics of Atomic Collision Physics: Elastic Processes
6.5.2 Formalism
Both resonance types just discussed are characterized by an end state (i.e. a scattered
wave) of well defined energy, which may be reached by a direct process as well as by
a temporary capture into a quasi-bound, intermediate states. We shall now describe
these two processes by two scattering amplitudes fdir and fres . We have discussed
a very similar situation already in Sect. 7.6, Vol. 1 for autoionization. There, direct
ionization and double excitation were interfering. The following considerations are
adapted to the scattering problem. Here too, both channels
ABres
fres ,τ
A+B A+B
fdir
dσ
= I (θ ) = |fdir + fres |2 . (6.123)
dΩ
where Wres would be the energy of a bound state without decay, and Γ = /τ gives
its width due to the decay. With the concepts developed in Sect. 6.4.6 we may now
characterize resonance scattering as follows: a resonance occurs in a particular par-
tial wave, say for = ζ . It has purely outgoing character, hence, the incoming partial
wave must disappear for the complex energy W res . This implies that the correspond-
ing partial wave coefficient according to (6.121) must become aζ∗ (W res ) → 0 at res-
onance. For scattering energies T Wres we may expand aζ∗ around the resonance
energy:
∗
∗ iΓ daζ
aζ = T − Wres + . (6.124)
2 dT Wres
6.5 Resonances 439
With this the scattering matrix (6.122) for the partial wave ζ becomes
aζ iζ π T − Wres − iΓ
daζ /dT |Wres
Sζ = − e = 2
eiζ π .
aζ∗ T − Wres + iΓ
2
daζ∗ /dT |Wres
Of course |Sζ | ≡ 1 holds (unitary matrix). Also the magnitude of the last fraction in
this expression is ≡ 1, so that we may abbreviate
daζ /dT |Wres 0
eiζ π ∗
= e2iηζ .
(daζ /dT |Wres )
is the resonance phase.18 In contrast, within the present approximation ηζ0 repre-
sents a constant background phase shift. Hence, we see that the resonant part of the
scattering phase changes rapidly by π when the energy passes the resonance energy
and assumes an odd multiple of π/2 for Wres − T .
For the special case that all scattering phases except the resonance phase vanish,
the integral cross section (6.90) becomes
4π
σ= (2ζ + 1) sin2 ηζres ,
k2
which leads with (6.126) to a L ORENTZ profile of the resonance19
4π (Γ /2)2
σ= (2ζ + 1) , (6.127)
k2 (Wres − W )2 + (Γ /2)2
which we know from optical excitation of atomic levels. In electronic and atomic
scattering, however, usually many partial waves contribute. They generate a non-
resonant, slowly varying background, while the rapid changes due to a resonance
usually occur only in one partial wave. We have seen this already in Fig. 6.32 (left,
middle) for the d wave scattering phase in a potential well: there a consequence of
a corresponding shape resonance.
18 One easily verifies this from (6.125) using the trigonometric relation tan 2η = 2 tan η/(1 −
tan2 η).
19 Often this is called B REIT-W IGNER distribution since B REIT and W IGNER (1936) have applied
this formula for the first time onto resonance scattering of neutrons.
440 6 Basics of Atomic Collision Physics: Elastic Processes
We are now able to derive expression (6.123), heuristically introduced, from the
partial wave expansion (6.86), simply by inserting there (6.125). For a resonance in
the partial wave = ζ we obtain:
* ∞ +
1 2iη0 2iηζ0 +2iηζres
f (θ ) = (2 + 1)e P (cos θ ) + (2ζ + 1)e Pζ (cos θ )
2ik
=0
=ζ
*∞
1 0
= (2 + 1)e2iη P (cos θ )
2ik =0
+
2iηζ0 2iηζres
+ (2ζ + 1)e e − 1 Pζ (cos θ ) . (6.128)
of which the direct scattering amplitude fdir (θ ) is obtained from the usual partial
wave expansion with phase shifts η0 , slowly varying with energy. The resonant am-
plitude in partial wave = ζ is given by
1 0 res
fres (θ, ") = (2ζ + 1)e2iηζ e2iηζ − 1 Pζ (cos θ ). (6.130)
2ik
It depends in the resonance region with ηζres (") according to (6.126) strongly on the
kinetic energy. Alternatively, the decisive energy depending factor
2iηres
e ζ − 1 ∝ fres (") may also be written (6.131)
2i res 2"i 2
=√ eiηζ = 2 − 2 . (6.132)
" +1
2 " + 1 " +1
(" + q)2
σ (T ) = σres + σdir (6.133)
1 + "2
with the asymmetry parameter q as we have done it for autoionization. The partial
wave analysis which we have presented here is, however, much more powerful for
analyzing the differential scattering cross section, since (6.128) allows in principle
a consistent description for all scattering angles with very few free parameters.
One finds scattering resonances in heavy particle collisions (at very low energies,
see Fig. 6.25), as well as and rather frequently in electron scattering by atoms and
molecules. They occur in a wide energy range and have been studied over the past
decades very thoroughly and comprehensively. A good survey on the state-of-the-
art of theory and experiment for e− atom scattering gives B UCKMAN and C LARK
(1994), and the review of H OTOP et al. (2003) introduces into a broad variety of
phenomena observed in the low energy scattering of electrons by molecules and
clusters.
A ‘bench mark’ case is the He− (1s2s 2 2 S1/2 ) F ESHBACH resonance – a reso-
nance in s wave scattering. Historically it gave the first experimental evidence of
resonance structures in electron atom scattering of the type discussed here, and
has been discovered by S CHULZ (1963) in the total e− − He cross section. The
first angularly resolved measurements of this and similar resonances have been re-
ported by A NDRICK and E HRHARDT (1966). The energy resolution was at that time
about 50 meV,20 which also illuminates the limits of spectroscopic investigations in
the scattering continuum: the price to be paid for the high depth of information
by energy and angular dependence, when studying collision dynamics, is a signif-
icantly reduced energy resolution as compared to optical spectroscopy. Since these
first experiments many groups have worked hard on an improvement of this situa-
tion. The so to say ‘ultimate’ experiment, developed by H OTOP and collaborators
(see e.g. G OPALAN et al. 2003), has achieved an energy resolution down to 4 meV
FWHM (B OMMELS et al. 2005). It has been employed for a number detailed, very
informative investigations of resonances and the threshold behaviour in the elastic
and inelastic electron atom and electron molecule scattering. Figure 6.37 illustrates
schematically the most important aspects of this sophisticated experimental setup.
It is fair to say that here the preceding 40 years of methodical and experimental
development of scattering physics culminate.
20 In a traditional experiment one generates the electron beams by thermal emission from a cathode.
From its rather broad energy distribution one selects a more or less broad fraction by an electrostatic
monochromator, limits the emission angle and accelerates it to the desired energy T . After the
collision, one usually has to analyze the scattered electrons correspondingly.
442 6 Basics of Atomic Collision Physics: Elastic Processes
z towards detector
(a) potassium beam target beam for metastables
x
electron extraction
SC
SV electron beam
PIEQ
electron lenses
FARADAY
and deflection
cup
155 mm
continuum
Kontinuum
T0
(b) (c)
423< λ2< 478 nm
electron beam ≈ F MHz
42P3/2 3 21.0
e- beam 39K(4p)
2
FARADAY 1 0 9.3
focussing 1730.4
SC cup 2 3.2
GHz
42P1/2 1 57.7
≈ ≈
λ1= 766.7 nm
y 2
39K(4s) 42S1/2
x retarding field 461.7
target beam
detector(s)
1
Fig. 6.37 ‘Ultimate’ electron scattering experiment with ultra high energy resolution according
to G OPALAN et al. (2003). Experimental setup: cuts along the (a) zx, and (b) yx plane; the most
important components are the photoionization electron source (PIES) with the source volume (SV)
and the scattering chamber (SC). (c) Resonant two-photon ionization scheme for the generation of
the scattering electrons with an initial kinetic energy of T0 < 1 meV
The heart of the apparatus is the photoionization electron source (in Fig. 6.37(a)
marked as PIEQ), in which a potassium beam (−z-direction) is ionized by two nar-
row band laser beams (z-direction). By this device electrons with a well defined ini-
tial energy are created in a very small start volume (SV). Crucial for this concept is
that the laser defined initial kinetic energy T0 of the photoelectrons is not broadened
significantly by the space charge of the K+ ions from which they are emerging. As
detailed simulations show, one has to keep the ionization energy just above the ion-
ization potential and restrict the extracted current to some 10 pA to 100 pA. Hence,
the intricate, resonant two-photon ionization scheme via the K(4 2 P3/2 ) intermedi-
ate state, which is excited by a CW T I :S APPH laser (λ1 = 766.7 nm) (about optical
pumping see also Appendix D). The two hyperfine levels F = 2 and 3 are excited si-
6.5 Resonances 443
Section summary
• Scattering resonances are interesting and often encountered phenomena. They
occur due to quasi-bound states in the collision continuum which open two
indistinguishable channels. These interfere quantum mechanically and lead to
characteristic structures in the cross sections.
444 6 Basics of Atomic Collision Physics: Elastic Processes
I(θ) / a02 sr -1
present a consistent partial
wave fit with an energy 1.0
2
resolution of 7.4(5) meV
45° 0.5
FWHM and a resonance
width of Γ = 11.2(5) meV. 0 0.0
The black, dashed lines 0.8
(hardly distinguishable from
0.8
the red lines) represent the
results of the RMPS theory, 90° 0.4
convoluted with the energy 0.4
resolution of 7.4 meV; the
vertical, dash-dotted line 0 0.0
marks the resonance energy 19.30 19.40
Wres = 19.365 eV electron energy T / eV
on average. The main advantage of the B ORN approximation is its simplicity, in par-
ticular at high energies where high values of limit the applicability of a rigorous
partial wave expansion. At high energies we have k 2
(2M̄/2 )V (R) so that the
right side of (6.63) may be treated as a small perturbation, so that first order B ORN
approximation (FBA) is derived from
2M̄
∇ 2 ψ (1) + k 2 ψ (1) = V (R)ψ (0) . (6.134)
2
The plane wave ψ (k) (R) = exp(ik · R) = ψ (0) is taken as 0th order solution. Second
order B ORN approximation (SBA) follows from
2M̄
∇ 2 ψ (2) + k 2 ψ (2) = V (R)ψ (1) (6.135)
2
and so on. SBA and higher terms of this B ORN series are actually used successfully
in modern research for a number of problems, e.g. for rather accurate solutions
of electron impact ionization. Presently, we treat first order B ORN approximation
(FBA) applied to elastic scattering.
Starting from the general expression for the scattering amplitude (6.68), all we have
to do is to substitute the incoming plane wave ψ (ki ) in place of the exact solution
ψ (+) to obtain the elastic scattering amplitude in FBA:
M̄
f (FBA) (θ, ϕ) = − ψ (k f )∗ R V R ψ (ki ) R d3 R (6.136)
2π2
M̄
=− 2
k f |V R |k i . (6.137)
2π
And the T-matrix (6.69) in FBA becomes
(FBA)
Tf i = k f |V R |k i = e−ik f R V R eiki R d3 R (6.138)
2π2 (FBA)
= eiKR V R d3 R = − f (θ, ϕ), (6.139)
M̄ f i
√
with |k i | = |k f | = k = 2M̄T / in the elastic case, and the momentum trans-
fer K = k i − k f . Obviously the B ORN scattering amplitude is proportional to
the F OURIER transform of the potential (see Appendix I.5 in Vol. 1). The integral
(6.139) is rather easy to handle. We have already encountered it in similar form in
the context of photoionization in Sect. 5.5, Vol. 1. For H atoms it may be evaluated
explicitly as documented by (5.75), Vol. 1.
446 6 Basics of Atomic Collision Physics: Elastic Processes
2M̄ qA qB a0
= . (6.143)
me (a0 /R0 )2 + (Ka0 )2
21 The attentive reader maynotice by trial and error that direct insertion of a potential ∝ 1/R leads
to problems which are related to the long range of the C OULOMB potential.
6.6 B ORN Approximation 447
For a pure C OULOMB potential, i.e. for RUTHERFORD scattering, R0 → ∞ and the
DCS becomes
2
dσ M̄ 4(qA qB )2 2 M̄ qA qB 2 a02
= a 0 = . (6.145)
dΩ me (Ka0 )4 me T /Eh 16 sin4 (θ/2)
Here we have inserted (6.142) and used the identity (ka0 )2 = 2T /Eh .
Interestingly, in this particular case FBA leads to the same result as the classical
trajectory approximation (6.45). We also mention here – without proof – that the
exact quantum mechanical treatment again leads to the same result. This is of course
a lucky mathematical incident – owing to the special properties of the C OULOMB
potential.
Often the DCS is given with respect to the quantity WK = (K)2 /2me =
Eh (Ka0 )2 /2, somewhat loosely called energy transfer.22 With the momentum trans-
fer K according to (6.142), dΩ = 2π sin θ dθ = πdK 2 /k 2 , and the kinetic energy
T = Eh (ka0 )2 /2 we have to replace dΩ = πdWK /T , and obtain
2 2
dσ M̄ 4π(qA qB )2 2 M̄ π(qA qB )2 2
= a = a . (6.146)
dWK me T (Ka0 )4 0 me T (WK /Eh )2 0
22 As a matter of fact, the energy transfer from projectile A onto a target B at rest is (p)2 /2MB =
(K)2 /2M B = 2T (M̄/MB )(1 − cos θ) (see Fig. 6.11 and the text associated with it) – which for
electron impact on heavy particles is very small.
23 The readers may verify this by applying the following approximation for large :
2 1 −1/2
k 2 R 2 j (kR) → 1 − ( + 1/2)2 /k 2 R 2 .
2
448 6 Basics of Atomic Collision Physics: Elastic Processes
(6.147) or J EFFREYS -B ORN (6.96) phase shifts – instead of the more difficult to
compute WKB phase shift (6.94) – may be helpful even when the B ORN approxi-
mation as such cannot be applied: B ORN phases are always a good approximation
if the interaction remains weak on average, i.e. for large angular momenta (impact
parameters). For many problems one thus obtains very reasonable results by solving
the exact radial equation (6.80) for a few, low values of and using (6.147) for all
higher angular momenta .
For inverse power potentials of the type V (R) = ∓CR −ν , that is essentially for
all potentials and large (or impact parameters) at sufficiently high energies, one
may integrate (6.147) in closed form and finds:
This expression is something like a high angular momentum counterpart to the ef-
fective range formula (6.106), which is applicable only to low kinetic energies.
Section summary
• B ORN approximation is the most simple quantum mechanical approach to-
wards the scattering problem. Historically important, it is still used today for
its straight forward derivation of scattering amplitudes, as introduced here for
elastic scattering.
• B ORN approximation uses a perturbation approach, identifying the transition
operator with the interaction potential. From its matrix elements between in-
coming and outgoing plane wave the scattering amplitude is obtained in a
straight forward manner. Typically, FBA results are rather accurate at high
energies and small scattering angles, while the low energy cross sections are
usually overestimated.
• For C OULOMB (RUTHERFORD) scattering amplitude derived in FBA is iden-
tical to the classical and quantum result.
• The partial wave expansion of FBA allows one to derive B ORN (6.147) and
J EFFREYS -B ORN (6.96) phase shifts. These may be used with advantage for
large in combination with exact quantum calculations for a few low angular
momentum phase shifts.
References
A DIBZADEH , M. and C. E. T HEODOSIOU: 2005. ‘Elastic electron scattering from inert-gas
atoms’. At. Data Nucl. Data Tables, 91, 8–76.
A NDRICK , D. and A. B ITSCH: 1975. ‘Experimental investigation and phase-shift analysis of low-
energy electron-helium scattering’. J. Phys. B, At. Mol. Phys., 8, 393–410.
A NDRICK , D. and H. E HRHARDT: 1966. ‘Die Winkelabhängigkeit der Resonanzstreuung
niederenergetischer Elektronen an He, Ne, Ar, und N2 ’. Z. Phys., 192, 99–106.
BAEK , W. Y. and B. G ROSSWENDT: 2003. ‘Total electron scattering cross sections of He, Ne and
Ar, in the energy range 4 eV–2 keV’. J. Phys. B, At. Mol. Phys., 36, 731–753.
BARTSCHAT , K.: 1998. ‘The R-matrix with pseudo-states method: Theory and applications to
electron scattering and photoionization’. Comput. Phys. Commun., 114, 168–182.
B OMMELS , J., K. F RANZ, T. H. H OFFMANN, A. G OPALAN, O. Z ATSARINNY, K. BARTSCHAT,
M. W. RUF and H. H OTOP: 2005. ‘Low-lying resonances in electron-neon scattering: Measure-
ments at 4-meV resolution and comparison with theory’. Phys. Rev. A, 71.
B ONHAM , R. A.: 1985. ‘Electron-atom elastic shadow scattering’. Phys. Rev. A, 31, 2706–2708.
B ONHOMMEAU , D., N. H ALBERSTADT and U. B UCK: 2007. ‘Fragmentation of rare-gas clusters
ionized by electron impact: new theoretical developments and comparison with experiments’.
Int. Rev. Phys. Chem., 26, 353–390.
B ORN , M.: 1926a. ‘Quantenmechanik der Stoßvorgänge’. Z. Phys., 38, 803–840.
B ORN , M.: 1926b. ‘Zur Quantenmechanik der Stoßvorgänge’. Z. Phys., 37, 863–867.
B ORN , M. and E. W OLF: 2006. Principles of Optics. Cambridge: Cambridge University Press,
7th (expanded) edn.
B RANSDEN , B. H. and C. J. J OACHAIN: 2003. The Physics of Atoms and Molecules. New York:
Prentice Hall Professional.
B RAUNER , M., J. S. B RIGGS and H. K LAR: 1989. ‘Triply-differential cross-sections for ioniza-
tion of hydrogen-atoms by electrons and positrons’. J. Phys. B, At. Mol. Phys., 22, 2265–2287.
B REIT , G. and E. W IGNER: 1936. ‘Capture of slow neutrons’. Phys. Rev., 49, 0519–0531.
B UCK , U., H. O. H OPPE, F. H UISKEN and H. PAULY: 1974. ‘Intermolecular potentials by inver-
sion of molecular-beam scattering data. 4. Differential cross-sections and potential for LiHg’.
J. Chem. Phys., 60, 4925–4929.
B UCK , U., M. K ICK and H. PAULY: 1972. ‘Determination of intermolecular potentials by in-
version of molecular-beam scattering data. 3. High-resolution measurements and potentials for
K-Hg and Cs-Hg’. J. Chem. Phys., 56, 3391–3397.
450 6 Basics of Atomic Collision Physics: Elastic Processes
B UCK , U., K. A. KOHLER and H. PAULY: 1971. ‘Measurements of glory scattering of Na-Hg’.
Z. Phys., 244, 180.
B UCK , U. and H. M EYER: 1984. ‘Scattering analysis of cluster beams – formation and fragmen-
tation of small Arn clusters’. Phys. Rev. Lett., 52, 109–112.
B UCKMAN , S. J. and C. W. C LARK: 1994. ‘Atomic negative-ion resonances’. Rev. Mod. Phys.,
66, 539–655.
B URKE , P.: 2006. ‘Electron-atom, electron-ion and electron-molecule collisions’. In: G. W. F.
D RAKE, ed., ‘Handbook of Atomic, Molecular and Optical Physics’, 705–729. Heidelberg, New
York: Springer.
C OHEN , J. S.: 1978. ‘Rapid accurate calculation of JWKB phase-shifts’. J. Chem. Phys., 68, 1841–
1843.
FARNIK , M., C. S TEINBACH, M. W EIMANN, U. B UCK, N. B ORHO and M. A. S UHM: 2004.
‘Size-selective vibrational spectroscopy of methyl glycolate clusters: comparison with ragout-
jet FTIR spectroscopy’. Phys. Chem. Chem. Phys., 6, 4614–4620.
F ELTGEN , R., H. K IRST, K. A. KOHLER, H. PAULY and F. T ORELLO: 1982. ‘Unique determi-
nation of the He2 ground-state potential from experiment by use of a reliable potential model’.
J. Chem. Phys., 76, 2360–2378.
F LUENDY , M. A. D., R. M. M ARTIN, E. E. M USCHLITZ J R . and D. R. H ERSCHBACH: 1967.
‘Hydrogen atom scattering – velocity dependence of total cross section for scattering from rare
gases hydrogen and hydrocarbons’. J. Chem. Phys., 46, 2172–2181.
G EIGER , J. and D. M ORÓN -L EÓN: 1979. ‘Electron-atom shadow scattering’. Phys. Rev. Lett., 42,
1336–1339.
G ENGENBACH , R., C. H AHN and J. P. T OENNIES: 1973. ‘Determination of H-He potential from
molecular-beam experiments’. Phys. Rev. A, 7, 98–103.
G OPALAN , A., J. B OMMELS, S. G OTTE, A. L ANDWEHR, K. F RANZ, M. W. RUF, H. H OTOP and
K. BARTSCHAT: 2003. ‘A novel electron scattering apparatus combining a laser photoelectron
source and a triply differentially pumped supersonic beam target: characterization and results
for the He− (1s 2s2 ) resonance’. Eur. Phys. J. D, 22, 17–29.
G ULLEY , R. J., D. T. A LLE, M. J. B RENNAN, M. J. B RUNGER and S. J. B UCKMAN: 1994.
‘Differential and total electron-scattering from neon at low incident energies’. J. Phys. B, At.
Mol. Phys., 27, 2593–2611.
H OFSTÄDTER , R.: 1961. ‘The N OBEL prize in physics: for his pioneering studies of electron
scattering in atomic nuclei and for his thereby achieved discoveries concerning the structure of
the nucleons’, Stockholm. http://nobelprize.org/nobel_prizes/physics/laureates/1961/.
H OTOP , H., M. W. RUF, M. A LLAN and I. FABRIKANT: 2003. ‘Resonance and threshold phe-
nomena in low-energy electron collisions with molecules and clusters’. In: ‘Advances in Atomic
Molecular, and Optical Physics’, vol. 49, 85–216. Amsterdam: Elsevier, Academic Press.
L EVINE , R. D.: 2005. Molecular Reaction Dynamics. Cambridge: Cambridge University Press,
554 pages.
OM ALLEY , T. F., L. S PRUCH and L. ROSENBERG: 1961. ‘Modification of effective range theory
in presence of a long-range (r−4 ) potential’. J. Math. Phys., 2, 491–498.
OVCHINNIKOV , S. Y., G. N. O GURTSOV, J. H. M ACEK and Y. S. G ORDEEV: 2004. ‘Dynamics
of ionization in atomic collisions’. Phys. Rep., 389, 119–159.
PAULY , H. and J. P. T OENNIES: 1965. ‘The study of intermolecular potentials with molecular
beams at thermal energies’. In: ‘Adv. Atom. Mol. Phys.’, vol. 1, 195–344. New York: Academic
Press.
VAN DER P OEL , M., C. V. N IELSEN , M. RYBALTOVER , S. E. N IELSEN , M. M ACHHOLM and
N. A NDERSEN: 2002. ‘Atomic scattering in the diffraction limit: electron transfer in keV Li+ -
Na(3s, 3p) collisions’. J. Phys. B, At. Mol. Phys., 35, 4491–4505.
P OTERYA , V., M. FARNIK, U. B UCK, D. B ONHOMMEAU and N. H ALBERSTADT: 2009. ‘Frag-
mentation of size-selected Xe clusters: Why does the monomer ion channel dominate the Xen
and Krn ionization?’ Int. J. Mass Spectrom., 280, 78–84.
R AKITYANSKY , S. A. and N. E LANDER: 2009. ‘Generalized effective-range expansion’. J. Phys.
A, Math. Gen., 42, 225302.
References 451
S CHULZ , G. J.: 1963. ‘Resonance in elastic scattering of electrons in helium’. Phys. Rev. Lett., 10,
104–105.
S MITH , F. T., R. P. M ARCHI and K. G. D EDRICK: 1966. ‘Impact expansions in classical and
semiclassical scattering’. Phys. Rev., 150, 79–92.
S TEINBACH , C., M. FARNIK, U. B UCK, C. A. B RINDLE and K. C. JANDA: 2006. ‘Electron
impact fragmentation of size-selected krypton clusters’. J. Phys. Chem. A, 110, 9108–9115.
S ZMYTKOWSKI , C., K. M ACIAG and G. K ARWASZ: 1996. ‘Absolute electron-scattering total
cross section measurements for noble gas atoms and diatomic molecules’. Phys. Scr., 54, 271–
280.
T OENNIES , J. P.: 2007. ‘Molecular low energy collisions: past, present and future’. Phys. Scr., 76,
C15–C20.
T OENNIES , J. P., W. W ELZ and G. W OLF: 1976. ‘Determination of H-He potential well depth
from low-energy elastic-scattering’. Chem. Phys. Lett., 44, 5–7.
U LLRICH , J., R. M OSHAMMER, A. D ORN, R. D ÖRNER, L. P. H. S CHMIDT and H. S CHMIDT-
B ÖCKING: 2003. ‘Recoil-ion and electron momentum spectroscopy: reaction-microscopes’.
Rep. Prog. Phys., 66, 1463–1545.
Inelastic Collisions – A First Overview
7
Overview
We introduce some characteristic questions about inelastic and reactive colli-
sions and approaches to answer them for several important examples. We start
in Sect. 7.1 with very simple models. The general trends for excitation pro-
cesses as a function of the relative kinetic energy are presented in Sect. 7.2.
Specifically, in Sect. 7.2.7 we focus on the threshold region. In Sect. 7.3 we
introduce multichannel theory, and discuss the alternative adiabatic and di-
abatic viewpoints. In Sect. 7.4 we extend the semiclassical methods already
employed in the elastic case. In Sect. 7.5 we make a short excursions into
the world of collision processes with highly charged ions. Finally, we address
reactive scattering processes in Sect. 7.6.
Starting from the effective potential Veff (R) according to (6.39) one may already
make some predictions on the order of magnitude of cross sections. Let us first
have a look at exothermic reactions without threshold energy and let us assume the
V(R) Rm
interaction potential V (R) is purely attractive. Typical examples are ion-atom and
ion-molecule reactions according to the scheme
In the most simple case this may represent a charge exchange process. Since we con-
sider an exothermic reaction it will always happen if the approaching particles with
their initial kinetic energy T can overcome the centrifugal barrier. This is illustrated
in Fig. 7.1 where the interaction potential V (R) and the effective potential Veff (R)
according to (6.39) are shown. The small inset (bottom right) shows characteristic
trajectories as a function of the impact parameter b.
A trajectory for a given kinetic energy T always leads to a reaction if the impact
parameter b is small enough to reach the critical radius Rm , i.e. if at the maximum
of the potential barrier
(max) b2
Veff = V (Rm ) + T 2
≤T (7.2)
Rm
bm = (2αq 2 /T )1/4 which still leads to a reaction – only for b ≤ bm the reaction
may occur. From this consideration the so called L ANGEVIN cross section for ion-
molecule reactions is derived:
σL = πbm 2
= πq 2α/T . (7.3)
W th kinetic energy T
For endothermic reactions, which occur only above an energetic threshold Wth , the
initial kinetic energy T changes to T − Wth during the process, and we have to
modify our considerations with respect to the critical distance Rth and the maxi-
mum impact parameter bm . Figure 7.2 illustrates the relations for such an inelastic
reaction process in the absorbing sphere model.
The reaction may only occur if the trajectories reach the crossing of the potential
curve of the ground state (black) with the excited state (red), that is to say if the
energy at the crossing radius Rth is sufficient to overcome the threshold energy Wth .
In this case we have to replace (7.2) by
b2
Veff (Rth ) = Wth + T 2
≤ T, (7.4)
Rth
from which with Veff (Rth ) = T again the maximum impact parameter for the reac-
2 = R 2 (1 − W /T ) and the reaction cross
tion is obtained. Here it is given by bm th th
section becomes
σr = πbm
2
= πRth
2
(1 − Wth /T ) for T > Wth . (7.5)
4
v=4
3
2 absorbing sphere
v=1
LANGEVIN
1
0
0 1 2 3 4 5
Wth kinetic energy T / eV
Fig. 7.4 Reaction cross sections for the formation of NeH+ and NeD+ ions (full and open experi-
mental points, respectively) formed in the reaction HD+ + Ne as a function of initial kinetic energy
T (in the CM system) according to D RESSLER et al. (2006). Both reactions are exothermic if the
initial vibrational quantum number is v = 4, and both are endothermic if v = 1; they are compared
with the respective simplified models according to Fig. 7.3
Reality, however, is usually somewhat more complicated. We illustrate this for the
example of charge exchange in a still fairly simple ion molecule reaction
for which experiments with vibrationally selected HD+ molecular ions have been
performed by D RESSLER et al. (2006). Figure 7.4 shows the cross sections for two
initial vibrational states of HD+ . In the case of v = 1 the reactions (7.6) and (7.7)
are endothermic, with threshold energies of Wth = 0.29 and 0.25 eV, respectively,
in the case of v = 4 both are exothermic.
One recognizes in Fig. 7.4 the typical behaviour, essentially predicted by the
models discussed above: exothermic reactions show reaction cross sections which
monotonously decrease with energy, while the endothermic cross sections first rise
rapidly beyond the threshold, reach a maximum and fall again slowly with further
increasing energy (that the signals at very low energies are nearly constant for v = 4
is an experimental artifact due to the finite width of the energy distribution – the
same holds for the small residual signal below threshold Wth in the v = 1 case).
The dashed lines in Fig. 7.4 indicate our attempts to fit the experimental data
by the models discussed above – by a L ANGEVIN cross section for the exothermic
case v = 4 and by the absorbing sphere model for the endothermic case v = 1. As
documented, this is only partially successful. While the decay of the cross section
for v = 4 may be predicted, and the threshold behaviour for v = 1 is verified, the
strong decay with increasing energy is not reproduced by these simple models. The
7.1 Simple Models 457
most obvious explanation is the purely geometric nature of these models. They im-
ply that the reaction probability is 100 % if only the classical trajectory reaches the
critical radius Rm and Rth , respectively. In reality this is clearly not the case: we
expect a reduction of the reaction probability with decreasing interaction time, i.e.
with increasing kinetic energy – as observed in the experiment.
with a kinetic energy T of the collision partner before and T − Wba after the colli-
sion (both with respect to the CM system). In this process, B is initially in a state |a
of energy Wa , and is excited by the collision into the state |b with energy Wb > Wa .
During the collision, the relative kinetic energy is reduced by the excitation energy
Wba = Wb − Wa . The interaction time of this process may roughly be estimated
from a characteristic range a of the potential and the relative velocity u = 2T /M̄:
a M̄
tcol = = a. (7.9)
u 2T
During this short collision time tcol , the internal energy of the system is defined only
within the limits of the uncertainty relation:
W tcol ≥ . (7.10)
The shorter the collision time, the larger the uncertainty W of the internal energy.
For electron atom (or molecule) impact tcol is on the same order of magnitude as the
electronic transition time te = /Wba for the inelastic process. More generally, with
(7.9) we may rewrite (7.10) as
T M̄ Wba a 2
≥ ,
Wba me 2Eh a02
with the atomic units Eh = 27.2 eV, a0 0.053 nm and the electron mass me (2 =
me Eh a02 ). In words: if the excitation energy Wba is on the order of Eh (as typical
for electronically excited states), and the range of the potential is on the order of a0 ,
we expect maximum excitation cross sections for kinetic energies T /Wba ≥ M̄/me .
For electrons (M̄/me 1) this holds just above threshold, while for heavy particle
collisions (M̄/me
1) kinetic energies substantially above threshold are required
for optimal excitation!
458 7 Inelastic Collisions – A First Overview
-1/R 4
b R(t) 2 2
e- t / 2t col
z,t
-1/R
V(R)
t=0 0 1 2 3
1/ ξ ∝ u
√
Fig. 7.5 Illustrating the M ASSEY parameter ξ : (a) Trajectory R(t) = z2 + b2 with z = ut ,
(b) time dependent potentials V (t), which lead to the (c) transition probability |cab (∞)|2 which
depends on the relative velocity u ∝ 1/ξ
In the spirit of the B ORN -O PPENHEIMER approximation (Sect. 3.2) one expects
a strictly non-adiabatic behaviour of the electronic wave function during e− − atom
interactions. In contrast, in heavy particle collisions at kinetic energies on the or-
der of Wba , the collision time is tcol
te . As a consequence, the electronic wave
function of the atoms may follow the interaction potentials adiabatically through-
out the whole scattering process. Thus, in atom-atom (molecule) collisions, only at
very high kinetic energies inelastic processes are expected (when tcol and te become
comparable).
For a quantitative interpretation, we consider the interaction potential V (R)
as time dependent along a classical trajectory R(t). As indicated in Fig. 7.5(a),
for simplicity we assume the relative motion along a straight line trajectory, par-
allel to √the z-axis. With relative velocity u and impact parameter b we obtain
R(t) = b + z = b2 + (ut)2 . For sufficiently large b, the potential is attractive,
2 2
−C/R −s , and thus may be written as:
−s/2
V (t) = −C 1 + (ut/b)2 .
Figure 7.5(b) shows the characteristic time dependence for such a potential for the
pure C OULOMB case ∝ −1/R and for a polarization potential ∝ −1/R 4 (for this
display we have chosen b = utcol ).
As we are interested here only in general trends, we represent these potentials by
a G AUSS function
√
V (t) −U0 exp −t 2 /( 2tcol )2 , (7.11)
as shown in Fig. 7.5(b). With its finite range, it keeps the following computation
simple and the results clear. We identify the collision time according to (7.9) through
a characteristic average range a of the potential.
We now recall from Chap. 4 in Vol. 1, that time dependent perturbation theory
turned out to be very successful for treating optically induced transitions. It should
7.1 Simple Models 459
also work here as long as the interaction potentials remain weak, i.e. for U0 /Wba
1. With (4.47), Vol. 1 we had derived in the optical case that the probability
amplitude cba (∞) for the transition from state |a into state |b is nothing but the
F OURIER transform of the perturbation potential at frequency ωba of the atomic
transition. In the present case we thus have to compute:
i ∞ −iωba t i ∞ √
U0 e−t /( 2tcol ) −iωba t dt.
2 2
cba (∞) = − V (t)e dt
−∞ −∞
This integration can be done in closed form and leads to a probability for finding the
target B after the collision in the excited state |b:
2
U0 U0 2
cba (∞)2 = 2πe−tcol 2 ω2
ab tcol =
2
2πξ 2 e−ξ .
2
(7.12)
Wba
Here we have introduced the so called M ASSEY parameter
|Wba |a |Wba |a M̄
ξ = ωba tcol = = (7.13)
u 2T
(also called “adiabaticity parameter”). We specifically emphasize that it depends
on the reduced mass M̄ of the system and on the relative kinetic energy T . Fig-
ure 7.5(c) illustrates the dependence of the excitation probability (7.12) on the rela-
tive collision velocity u ∝ 1/ξ . Obviously the excitation probability reaches indeed
its maximum for ξ 1 (so called M ASSEY criterium). This corresponds to the equal
sign in (7.10).
Thus, the M ASSEY parameter characterizes an adiabatic energy range (ξ
1),
for which inelastic processes are without relevance, and a diabatic range (ξ ≤ 1),
where transitions in atoms or molecules may be induced by collisions. It under-
lines again the fundamental difference between electron scattering and heavy par-
ticle scattering: for the former the diabatic region ξ ≤ 1 typically begins at kinetic
energies above threshold T > |Wba |.
In contrast, for all heavy particle collision processes which we have discussed
in Chap. 6 the M ASSEY parameter is on the order of ξ ≈ 102 to 103 , thus inelas-
tic processes do not occur. They usually begin to play a role only for initial kinetic
energies in the range of some keV to MeV where ξ 1. Only if the energy dif-
ference Wba (which has to be surmounted during collision) is very small, e.g. in
rotational or vibrational excitation of molecules, thermal or hyperthermal kinetic
energies may be effective in achieving the transition. Clearly, (7.12) gives only a
qualitative description of the velocity dependence for inelastic processes. And the
choice of a ‘characteristic average range’ a is always somewhat arbitrary. However,
practice shows that the observations are usually quite well described by (7.12). In
particular, the M ASSEY criterium ξ 1 gives a rather good indication for the max-
ima of excitation functions1 for many processes – as we shall document in the next
section.
1 Excitation cross sections as a function of collision energy are called excitation functions.
460 7 Inelastic Collisions – A First Overview
Section summary
• For exothermic reactions occurring in an attractive potential the cross sections
are essentially determined by the trajectory (relative kinetic energy T ) which
must lead over the centrifugal barrier. Specifically for charge exchange be-
tween an ion (charge qe) √ and an atom (polarizability α) the L ANGEVIN cross
section (in a.u.) is πq 2α/T .
• For endothermic reactions with a threshold Wth = Wba which is reached at
a critical radius Rth the absorbing sphere model assumes that all trajectories
which reach the sphere of radius Rth contribute, so that the cross section be-
comes πRth 2 (1 − W /T ). Both models ignore the general trend that cross
th
sections decrease at higher energies.
• A semi-quantitative dynamical model is derived by a simple time dependent
approximation. The crucial parameter is the collision time, tcol = a/u, where
a is an effective interaction range and u the relative velocity. The M ASSEY
parameter, ξ = (|Wba |/)(a/u), distinguishes the adiabatic (ξ
1) from di-
abatic energy regime (ξ ≤ 1) where inelastic processes can occur. An aston-
ishing good first order guess for the maxima of inelastic processes gives ξ = 1.
We illustrate this first for the excitation of He atoms from the ground state. Fig-
ure 7.6 shows the excitation functions for the processes
(a) e− + He 1s 2 1 S0 + T → e− + He 1s2p 1 Po1 + T − 21.22 eV
(b) p+ + He 1s 2 1 S0 + T → p+ + He 1s2p 1 Po1 + T − 21.22 eV
as a function of the relative velocity u in the CM system. They are compared with
each other and with several theoretical approximations (CC represents “close cou-
pling”, CCC “convergent close coupling” – approximations which we shall get ac-
quainted with in Sect. 8.1). First B ORN approximation, FBA, gives at low energies
obviously only a very rough guess, but it is astonishingly realistic at higher energies
(dashed lines in Fig. 7.6).
If we assume that the “effective, average interaction range” a of the potential
is given by the VAN DER WAALS radius for He (0.14 nm), the velocity u in the
CM system in Fig. 7.6 may be expressed by the M ASSEY parameter ξ according to
(7.13), u = Wba /(aξ ).
The experimentally observed trend and the theoretical modelling with ambitious
methods confirm the simple considerations of the preceding section. The relevant
literature documents for a multitude of cases that the positions of the maxima of
7.2 Excitation Functions 461
10 (b)
5
ξ =1
0
0 uth 2 4 6 8
relative velocity u / a.u.
the excitation functions are estimated rather well by the M ASSEY criterium ξ 1.
This holds, at least qualitatively, in electron, atom and ion collisions for so called
‘direct’ impact excitation. Clearly, the model cannot make any predictions about
many important and interesting details. And for heavy particle scattering, we shall
encounter in Sects. 7.3 and 7.4 a variety of processes which occur in the vicinity
of potential curve crossings – for which the M ASSEY criterium hast to be modified
dramatically.
We shall now give an introduction into the typical behaviour of excitation func-
tions for electron impact with atoms, by way of example. Rare gases are excellent
prototypes. Probably the best studied case is e− + He(11 S0 ) excitation into various
He(n 2S+1 LJ ) states. Numerous experiments have been compared with theoretical
approaches. Today, elaborate computational methods and programmes allow to de-
termine cross sections with high reliability even when experiments are not available.
For the summary of excitation functions shown in Fig. 7.7 we have made use of the
comprehensive data base of NIFS and ORNL (2007). Such data banks represent
the cumulated ‘know-how’ of several decades. In detail, it may be somewhat deli-
cate to extract exact cross sections for specific processes. But trends and orders of
magnitude for many systems are certainly represented correctly.
In addition to electron impact excitation of He(1s 2 1 S0 ) into the short-lived
1s2p 1 P1 state, a transition which we have already discussed above and which is
462 7 Inelastic Collisions – A First Overview
40
10
He(21P1) He(23P) He+ + e-
2
5 20
σ inel 1 σ inel σ ion
0 0 0
σ / 10-18 cm2
1 2
σ inel σ inel σ tot
0 0 0
10 102 103 104 10 102 10 102 103 104
electron energy T / eV
Fig. 7.7 Excitation function for several final states in e− + He(1s 2 1 S0 ) collisions. Summarized
are data derived from NIFS and ORNL (2007), using also the results of G OPALAN et al. (2003) for
the 23 S1 state. For comparison, also shown is the total cross section (including elastic scattering)
according to V INODKUMAR et al. (2007). Note the different scaling of σ for different processes
also allowed for optical E1 processes, Fig. 7.7 shows the cross sections σinel for
several optically forbidden transitions into metastable states, 1s2s 1 S0 , 1s2p 3 P and
1s2s 3 S1 . Immediately evident is the much sharper rise of these cross sections above
threshold and their much faster decay at higher energies – in comparison to the op-
tically allowed 2 1 P1 excitation. In summary, the excitation function for optically
forbidden transitions are significantly narrower on the energy scale, and the maxi-
mum of their cross section is typically a factor of five lower than the corresponding
optically allowed transition.
Looking back to our preliminary considerations in Sect. 7.1.4 it is not surprising
that optically allowed as well as optically forbidden transitions occur in collisions
over a rather broad energy range: the electron passing the atom rapidly generates a
time dependent interaction potential with a broad spectrum of frequencies, which
can in principle induce these transitions as soon as they are energetically possible
(T ≥ Wba ). And since in and outgoing electron waves may contain in principle
all angular momenta , there are no selection rules which would prohibit certain
processes – in contrast to optically induced E1 transitions, where one photon of
angular momentum is absorbed.
The particular rapid rise of the cross section above threshold for the 2S excita-
tions is obviously due to a dominance of the s scattering wave – as we have seen it al-
ready in elastic e− + He(1s 2 1 S) scattering (see e.g. Fig. 6.34). Clearly, at threshold,
7.2 Excitation Functions 463
where the kinetic energy T − Wba of the outgoing electron disappears, the s wave
dominates in the outgoing channel (see also Eq. (6.106)). In the 1s 2 S → 1s2sS ex-
citation process for the scattered waves = 0 must hold, i.e. the incoming s wave
is responsible for threshold excitation. In contrast, for the 1s 2 S → 1s2pP transition
= −1 describes the partial waves involved and at threshold only the incoming p
wave can induce the transition. And since its phase shift ηp is significantly smaller
the cross section rises much slower than in the former case. However, with increas-
ing kinetic energy T of the electrons the forbidden processes obviously decrease
much more rapidly than those optically allowed. We thus note (and we shall justify
this statement on strict theoretical grounds in Sect. 8.2):
With respect to excitation of atoms and molecules, fast electrons behave es-
sentially like white light!
Also characteristic is the excitation function for the 2 3 S and 2 3 P triplet states:
they are even more localized to a narrow energy range above threshold – as com-
pared to the excitation of the (also optically forbidden) singlet state 2 1 S0 . In the
case of the triplet states the total spin of the He changes from S = 0 to S = 1. As
we have learned in Vol. 1, spin orbit coupling is very weak for the He atom – as for
all light atoms. Thus, even during a collision the spin of an atomic electron will not
flip! The only possibility to induce a singlet-triplet transition by electron impact is
exchange of the projectile electron with one of the atomic electrons: in such an ex-
change process with excitation the chance is 50 % that the total atomic spin changes
according to the scheme
e(↑) + He 1s ↑ 1s ↓ 1 S0 → He 1s ↑ n ↑ 3 LJ + e(↓). (7.14)
However, such exchange processes are only probable at very low relative velocity.
This explains why the cross sections for exciting 2 3 P and 2 3 S1 are of a similar order
of magnitude as for 2 1 S0 and 2 1 P excitation, and why they decrease rapidly with
increasing energy.
Higher excited states of the He atom may also be excited by electron impact and
show similar trends, albeit with significantly smaller cross sections as n increases.
As for electromagnetically induced transitions this is explained by the lesser overlap
between the respective wave functions and that of the ground state.
For comparison we also show in Fig. 7.7 the integral cross section σion for elec-
tron impact ionization over a broad energy range. We shall come back to this in
Sect. 8.4. The cross section shows clear similarities with the 21 P excitation func-
tion, however, it is about a factor of four larger. We note, that this is completely
different to photoinduced processes where excitation is typically orders of magni-
tude larger than ionization. Finally, Fig. 7.7 shows the total electron cross section
σtot for He, which is dominated by σion and the elastic cross section σel for which
the low energy behaviour has already been reported in Fig. 6.7. The contribution of
excitation processes to σtot is relatively small.
464 7 Inelastic Collisions – A First Overview
4 2 3 S1 1.5
CCC
21S0 1.0 CCC
2 20.616 RMPS
eV 2 3P 0.5
0 0.0
20 21 20 30 40 0 100 200
electron energy T / eV
Fig. 7.8 Excitation function for e− + He(1s 2 1 S0 ) collisions in the metastable 2S and 2P states.
(a) Threshold region according to G OPALAN et al. (2003) (taken with the apparatus shown in
Fig. 6.37); on top: highly resolved part between 20.55 and 20.70 eV (2 1 S0 threshold at 20.616 eV).
Calculations: total cross section —, individual processes · · · (state-of-the-art R-matrix theory with
pseudo-states, RMPS); (b) — similar to RMPS according to BARTSCHAT (1998); (c) higher ener-
gies: CCC theory according to F URSA and B RAY (1995) - - -; older measurements ,
We now want to have a look at some selected examples to familiarize ourselves also
with some finer details of the cross sections. First, as a relatively straight forward
case study, we discuss the excitation of the metastable states in He. G OPALAN et al.
(2003) have studied the processes
e + He 1s 2 1 S0 → e + He 1s2s 3 S1 − 19.81 eV
(7.15)
and → e + He 1s2s 1 S0 − 20.62 eV
just above the excitation threshold, both experimentally and theoretically with very
high electron energy resolution. The metastable, excited atoms are registered with
a special L ANGMUIR -TAYLOR detector (see Sect. 1.9.3 in Vol. 1). Figure 7.8(a)
shows the total cross section for excitation of all metastable states as a function of
electron energy. Just above the 2 3 S1 excitation threshold (at ca. 19.8 eV) one ob-
serves on this fine energy scale not simply a smooth rise of the integral cross section
but rather pronounced structures which are reproduced by theory with astonishing
precision. They are, obviously, related with the threshold (20.616 eV) for the 2 1 S0
excitation. An even better resolved measurement in the vicinity of this threshold
(blow up on top of Fig. 7.8(a)) is not completely matched by theory which only
shows a sharp indent (a so called “cusp” due to the unitarity of the wave function
also at this threshold). The physics behind this feature is probably a very sharp res-
onance.
For the pure 2 3 S1 excitation, theory predicts in the intermediate energy range
between 22 and 25 eV some more, sharp resonances as shown in Fig. 7.8(b). Un-
fortunately, only in the higher energetic region some experimental data points are
available in this case. The decrease of the cross section extends up to 200 eV as
documented in Fig. 7.8(c).
7.2 Excitation Functions 465
0 0
4
30
20
2 Ne
20
10
0 Kr Xe
0 1000 2000 10
0 0
electron energy 0 1000 2000 0 1000 2000
T / eV
Fig. 7.9 Integral cross sections for electron scattering by different rare gases according to V IN -
ODKUMAR et al. (2007): σel elastic, σion ionization, σtot total
To illustrate that such investigations are not restricted to the most simple atoms
we now briefly mention other rare gases which are experimentally accessible with
relative ease. Quantitative computations can rely for comparison on sufficient ex-
perimental data. Again we first give a general overview on trends and orders of
magnitude. Elastic scattering has already been discussed briefly in Sect. 6.1.3. We
continue here and show in Fig. 7.9 the integral elastic cross sections σel , ionization
cross sections σion as well as total cross sections σtot for the rare gases He to Xe
over a broad energy range.
The cross sections shown are based on an detailed evaluation of many sets of
experimental data and a careful comparison with most recent ab initio calculations
by V INODKUMAR et al. (2007). Only the theoretical computational results are pre-
sented here. They set a bench mark for electron scattering of rare gases over a broad
energy range. The behaviour of these cross sections as such is quite unspectacular
and shows the expected rise and decay with energy, as we already know it from
helium. The elastic cross section provides in all cases the major contribution. The
ionization cross section is the next important ingredient for σtot while inelastic pro-
cesses play only a minor role – just as in the He case. The rare gas atoms differ
essentially by the order of magnitude of the cross sections which increases from
He to Xe by about a factor of 10 to 15. This reflects very clearly the corresponding
polarizabilities of the rare gas atoms, which dominate the interaction potential with
the scattered electron at large distances: they are α 1.4, 2.6, 11, 17 and 27 a.u. for
He, Ne, Ar, Kr and Xe, respectively.
The fact that such cross sections can be computed for a large variety of targets
is of great fundamental, but also of high practical importance. Modern computa-
tional methods for electron scattering – in Sect. 8.1.2 we shall get to know some
466 7 Inelastic Collisions – A First Overview
4 e- + Ne(2p 6) →
e- + Ne(2p 5 3s)
3
σ / 10-18cm 2
2
0
17 18 19
T / eV
Fig. 7.10 Excitation function for e− + Ne(2p 6 ) → e− + Ne(2p 5 3s) collisions close to the excita-
tion threshold. Compared are experiments with high energy resolution by B OMMELS et al. (2005)
and B UCKMAN et al. (1983) with R-matrix calculations of Z ATSARINNY and BARTSCHAT
(2004) . The experimental data are not calibrated but scaled to the theoretical data
of them in more detail – are very powerful and are improved continuously. Reli-
able experimental data, preferably with much structure, are crucial for testing and
advancing these methods. As we shall discuss later, special, highly selective param-
eters, depending on polarization and/or spin of the collision system, may provide
such crucial tests for theory. Similarly, pronounced structures, such as scattering
resonances in the low energetic region, may serve the same purpose – apart from
being interesting phenomena as such. A special example, according to B UCKMAN
and S ULLIVAN (2006) a “bench mark I”, is low energy electron excitation of neon
atoms. Figure 7.10 shows a small energy range above threshold. We see a series
of sharp resonances, corresponding to short-lived negative ions. Without going into
details we emphasize the excellent agreement between theory and experiment, doc-
umenting the capabilities of modern scattering theory very nicely. More recently,
such studies have been extended successfully also to krypton (H OFFMANN et al.
2010).
We finally mention just two very recent developments. The high energy resolu-
tion achievable with the laser photoelectron source used in these experiments is now
also applied to study electron collisions with rare gases at very low energies (see e.g.
K ITAJIMA et al. 2012). New perspectives for such studies may arise from recently
proposed monochromatic high flux electron sources based on field ionization of
RYDBERG states excited in laser-cooled atom beams (K IME et al. 2013).
We now address electron impact excitation of atomic mercury, Hg, as another, al-
ready rather complex example. The system is not only of practical importance, e.g.
for understanding the plasma in Hg discharge lamps. It also has a remarkable histor-
ical perspective: the famous experiment of F RANCK and H ERTZ (1914), an essential
7.2 Excitation Functions 467
4 6 3P1 (4.89eV)
6 3P2 (5.46eV)
σ / 10-16cm2
2
1
6 1P1 (6.68eV)
0
5 6 7 8 9
6 3P0 (4.67eV) T / eV
Fig. 7.11 Excitation function for e− + Hg(6 1 S0 ) → e− + Hg(6 3,1 P) collision, i.e. for the cross
sections populating the lowest 4 excited states of Hg just above threshold; red metastable states,
black short-lived states. The data have been sketched (somewhat schematically) after S IGENEGER
et al. (2003) and H ANNE (1988), also accounting for KOCH et al. (1984) and N EWMAN et al.
(1985). Note the sharp resonances just above the thresholds for 6 3 P0 and 6 3 P1
proof for the quantization of atomic states in the early development of quantum me-
chanics, was based on electron impact excitation of Hg.
As it turns out, the study of mercury, for this purpose, was a particularly happy
choice. Would the electron impact excitation functions in Hg for the first excited
states have the ‘typical’, rather broad energy width as we have seen them e.g. for
e− + He(1s 2 1 S) → e− + He(1s2p 1 P) in Fig. 7.6(a) this classical experiment would
not have shown any significant structures – and who knows what consequences this
might have had on the further evolution of quantum physics.
In reality, the excitation functions look roughly as sketched (schematically) in
Fig. 7.11. The available experimental data for this case are, even today, still not
fully satisfying. Most significant and well confirmed are, however, the extremely
sharp resonances with high cross sections shortly above the thresholds for 6 3 P0 and
6 3 P1 excitation. It is the latter resonance at 4.89 eV to which the F RANCK -H ERTZ
experiment owes its historical dimension (N OBEL prize in 1925). This very reso-
nance is responsible for a significant jump of the excitation probability whenever the
nominal electron kinetic energy T would be a multiple of 4.9 eV. This in turn leads
to an energy loss of the electrons – which is finally recognized by a reduction of the
current reaching the anode. A detailed modelling of the F RANCK -H ERTZ was and
still is today a substantial intellectual challenge (R APIOR et al. 2006; S IGENEGER
et al. 2003; H ANNE 1988).
Today excitation functions for metastable states are known over a large energy
range. They document the variety of possible phenomena quite impressively. Fig-
ure 7.12 shows a survey for the sum of all metastable processes of the type
e− + Hg 5d 10 6s 2 1 S0 → e + Hg 5d 10 6s6p 3 Po0,2
and → e + Hg 5d 10 6s7p 3 Po2 etc.
468 7 Inelastic Collisions – A First Overview
6d 3D
5 6p 3Po1
×30
σ / arb. un.
6p 1Po1
0
4 6 8 10 12 14 16
6p 3Po0 6p 3Po2 7p 3Po2 11d 3D T / eV
Fig. 7.12 Excitation function for e− + Hg(6 1 S0 ) into metastable states. Vertical lines —— (grey)
mark the thresholds, - - - (grey) indicate potential parent states for resonances. Summarized from
data by N EWMAN et al. (1985) and KOCH et al. (1984)
from threshold for the first excited state up to and beyond the thresholds for
HgII(2 S1/2 ), HgII(2 D3/2 ), and HgII(2 Do5/2 ) ionization. The data presented here are
collected from several different experiments, and the relative calibration is some-
what uncertain. Note also, that the sum of cross sections for excitation of all
metastable states has been measured: one may assume that the energetic region
above 5.46 eV is dominated by excitation of the 6 3 Po2 state, and above 8.83 eV
by the 7 3 Po2 excitation. These excitation functions are obviously characterized by
a number of resonances, i.e. by temporarily formed Hg− anions, for which N EW-
MAN et al. (1985) give plausible configurations. Present theory appears to be just
on the verge of a quantitative interpretation of these interesting data. Clearly, in this
case spin-orbit coupling cannot be treated as small perturbation, but recent relativis-
tic convergent close-coupling calculations (RCCC) reproduce already several of the
resonances shown here (B OSTOCK et al. 2010; B OSTOCK 2011). They even allow
a convincing comparison with experimental data for differential cross sections (not
shown here).
The excitation of molecules by electron impact (as well as purely elastic scattering)
is a wide field of intensive research in which during the last decades a host of excit-
ing material has been collected. We can give here only a slight taste. Everything that
makes collisions of electrons with atoms interesting, and sometimes complicated, is
also found when studying molecules. However, the electronic structure of molecules
is significantly more complex, as we have seen in Chaps. 3 and 4. In addition, vi-
brational and rotational structure of molecules comes on top – and with it a further
degree of difficulty arises. We know this already from spectroscopy and may sum-
marize it by the keyword “F RANCK -C ONDON (FC) factors”. Dissociatative pro-
cesses, including dissociative attachment of electrons, or for larger molecules the
7.2 Excitation Functions 469
/ 10-16 cm2
the authors own experiments 20
(red)
10
40
e- + CO
10
8
6 e- + O2
4
0.5 1 5 10 50 100
electron energy T / eV
Tacitly, one key question has already been addressed in much of the above dis-
cussion: how do integral cross sections behave just above threshold as a function
of kinetic energy? Ever since the early days of collision physics this has been a
challenging theme, and – albeit in principle solved (but just in principle) – remains
of interest until today. The fundamental theoretical concepts have been derived by
W IGNER (1948), and a summary on the present state-of-the-art is found e.g. in
S ADEGHPOUR et al. (2000) or H OTOP et al. (2003). The threshold laws are de-
rived from phase space arguments and the behaviour of the wave function of the
outgoing particle(s).
We recall: a photon (angular frequency ω) can excite an atom or molecule only if
the transition is resonant, i.e. if ω = Wba holds. The excitation function in this case
is thus a delta function (apart from the natural linewidth which may be neglected in
the present context). In contrast, an electron can excite whenever its kinetic energy
T is larger than the threshold energy Wth of the process studied (for atoms, usually
the threshold energy and the excitation energy are identical: Wth = Wba ). In the case
of excitation of a neutral atom or molecule by an electron according to W IGNER the
excitation cross section above threshold is
for an outgoing electron wave with wavenumber kf , energy T − Wba and angular
momentum . Since the √ latter usually is = 0 (s wave), this typically leads to an
energy dependence σ ∝ T − Wba .
With respect to ionization too, photons and electrons behave differently. Accord-
ing to W IGNER, for photoionization where one electron leaves the system in an
attractive C OULOMB potential, at threshold
σ ∝1 (7.17)
holds, i.e. the photoionization cross section starts at threshold (ω = WI ) with a
finite value. We have seen characteristic examples for this behaviour already in
Sect. 5.5, Vol. 1. For electron impact ionization the basic concepts have been de-
veloped by WANNIER (1953) and R AU (1971). The situation here is significantly
more complex, since in this case two electrons leave the atomic ion, and the excess
energy T − WI = WA + WB is shared among these two. For generating a Z fold
charged ion by electron impact, the threshold behaviour of the energy and angle
integrated cross section is according to WANNIER and R AU
1 100Z − 9
σ (ion)
∝ (T − WI )μ/2−1/4
with μ =
2 4Z − 1 (7.18)
hence for Z = 1 : σ (ion) ∝ (T − WI )1.127 .
cross section
an electron (c, d). The cross
section depends on the excess Wba WI
energy W = T − Wth , where e - +A → A*+ e- e - +A → A++2e -
the threshold energy Wth
corresponds to the excitation W 0.5 W 1.27
energy Wba or the ionization
potential WI , respectively.
(c) (d)
The range of validity of such
Wba WI
threshold laws may, however,
be very limited energy T of the exciting particle
Figure 7.14 summarizes the above discussion. Of course, threshold laws are valid
only for energies just above threshold. The key question as to the range of validity
of such behaviour cannot be answered universally – and has kept generations of
experimentalists and theoreticians intensively busy. All cross sections for excitation,
reaction and ionization pass through a maximum and decrease at high energies more
or less rapidly to zero.
As a general rule, the range of validity for threshold laws is very narrow, typi-
cally some tenth or even only a hundredth of an eV. Figure 7.15 illustrates this for a
process already well known to us: for electron impact excitation of helium into the
2 3 S1 state. This threshold study has again been performed with the very high res-
olution apparatus of G OPALAN et al. (2003) described in Sect. 6.5.3. The excellent
agreement between experiment and state-of-the-art R-matrix theory is truly remark-
able. But as one clearly sees, W IGNER’s square root threshold law holds only for a
very small energy range. For somewhat larger energies experiment and theory may
excitation function / arb. un.
e - + He(11S0) → e - + He(23S1)
2
W 0.5 W 0.391
Fig. 7.15 Excitation function for the 2 3 S1 state in e− + He(2 1 S0 ) collisions according to
G OPALAN et al. (2003). The data points are compared with R-matrix theory, which has √ been
convoluted
√ with the experimental energy width . W IGNER’s threshold law σ ∝ W =
T − Wth holds only for less then 10 meV above the threshold for excitation at Wth = Wba =
19.820 eV
472 7 Inelastic Collisions – A First Overview
Section summary
• Characteristic examples of excitation functions for ion and electron impact
have been discussed. For electron and proton excitation of He(1s2p 1 P)
the maximum of the excitation cross section is predicted very well by the
M ASSEY criterium.
• If finer details are overlooked, cross sections for electron induced (optically
allowed) excitation and ionization rise smoothly from threshold up to the
M ASSEY maximum, and then decrease more or less rapidly to zero with in-
creasing energy.
• For optically forbidden transitions, the rise of the cross section is much steeper
and the overall energetic width of the excitation function is much narrower –
particularly narrow for spin exchange processes (e.g. triplet excitation in He).
• State-of-the-art experiments with high resolution and recent results from scat-
tering theory reveal a wealth of finer structures in the excitation cross sections
due to resonances and cusps.
• Particularly structured are electron impact excitation functions for atomic
mercury. The famous F RANCK -H ERTZ experiment owes its historic dimen-
sion to one of these extremely sharp resonances.
• Electron molecule scattering shows even richer structures due to the complex-
ity of molecules and the formation of short-lived, excited states of molecular
anions.
• Threshold laws for photons and electrons differ. For excitation, with photons
strict resonance is required (delta
√ function) while with electrons the cross sec-
tion is predicted to follow σ ∝ W (with W = T − Wba ). For ionization with
photons, the cross section starts with a finite value at the ionization thresh-
old, while for electrons the WANNIER law with σ ∝ W 1.127 should hold. The
range of validity for these threshold laws is typically very narrow (less than
tenth of an eV).
After the discussion above, the relevance of a flexible and efficient, fully quantum
mechanical scattering theory including inelastic processes is evident. Today, a va-
riety of very powerful computer codes are available fulfilling these requirements.
The underlying concepts can be explained most clearly for inelastic heavy particle
collisions. The basic terminology, however, is also applicable to electron scattering
theory. A major part of Chap. 8 is devoted to this topic.
7.3 Scattering Theory for the Multichannel Problem 473
y (col) x (col)
prior to collision ℓy
|a〉
B ri(B)
ri(A) R ΘCM z (col)
A P
z (mol)
y (col) x (col)
after collision ℓy'
A' P' z' (mol)
B' R'
ri(B) '
ΘCM
ri(A) z (col)
|b〉
Fig. 7.16 Scheme of an inelastic scattering process A + B → A + B with excitation |a → |b.
The internuclear distance is R, the relative momentum of the nuclei P . The inner coordinates
r i of the system A–B are indexed here in addition by (A) and (B) to emphasize that these inner
coordinates refer either to collision parter A or B. In the standard collision frame the coordinates
x (col) , y (col) , z(col) are space fixed. However, the molecular axis z(mol) rotates during the collision.
To keep these schematic simple, we have positioned the centre of mass in this picture into the
centre of B (as if the mass of B was very large compared to that of A)
Here we shall focus for simplicity onto collisions of the type (7.8), involving two
atoms A and B as illustrated in Fig. 7.16 (B may also be an atomic ion). Both col-
lision partners may also be internally excited. Let the total state of the system A–B
prior to collision be characterized by |a, after the collision by |b. It is important to
note, that in such an inelastic collision the angular momentum of the nuclear motion
y , perpendicular to the scattering plane, may change – while for elastic collisions
it was a conserved quantity.
The relative motion of the colliding particles A and B is described by the in-
ternuclear distance R, the inner degrees of freedom by r i representing the posi-
tion coordinates of i = 1 . . . N electrons (mass me ) of the colliding partners. The
S CHRÖDINGER equation (6.61) may then be written
− W )Ψ (r 1 . . . r i . . . r N , R) = 0,
(H (7.19)
2 Often the computation can be dramatically simplified by distinguishing between active valence
electrons and the passive core electrons and by using correspondingly pseudopotentials rather than
summing over all N . The generalization to molecules as targets is for this ansatz without problems,
however, when solving the problem in detail, much more complicated.
474 7 Inelastic Collisions – A First Overview
operators for the kinetic energy of the relative motion A–B and those of all electrons
are
2 N
2 2 1
T(R) = − ∇R = P · û and T(r) = − ∇ 2r n , (7.21)
2M̄ 2 n
2m e
respectively. Here M̄ is the reduced mass of the system A–B, while its operators
for the relative momentum and relative velocity are P = /M̄, respec-
k and û = P
tively. We may write the total energy of the system as
2 ka2 2 kb2
W = T + Wa = + Wa = + Wb = T + Wb , (7.22)
2M̄ 2M̄
with the relative kinetic energies T and T , the corresponding wave vectors k a and
k b , the internal energies Wa and Wb of the collision system, before and after the col-
lision, respectively. The various possible excited states |b of the system are called
channels. A channel b is called open, if the corresponding excited state |b may in
principle be excited during the collision, i.e. if T ≥ Wb − Wa .
In the spirit of the B ORN -O PPENHEIMER approximation (Sect. 3.2) a suitable
general ansatz for the solution of (7.19) is
Ψ (r 1 . . . r i . . . r N , R) = φj (r 1 . . . r i . . . r N ; R)ψj (R). (7.23)
j
φj (r; R) −→ φj (r).
R→±∞
Finally, one has to ensure orthonormality φj |φj = δjj . Asymptotically, φj may
be described as a product wave function of the isolated particles A and B. The index
j represents all relevant quantum numbers.
The wave function ψj (R) describes the relative motion of A and B; it contains
the scattering dynamics for excitation of the states |j . For the elastic channel one
looks for a solution with incoming plane wave and outgoing spherical wave in the
form of (6.62). For the inelastic case with a = b, only outgoing spherical waves are
expected. Hence, asymptotically the full scattering wave function is described by
1
Ψ (r, R) −→ eik a R φa (r) + fba (θ, ϕ)eikb R φb (r). (7.24)
R→∞ R
b
7.3 Scattering Theory for the Multichannel Problem 475
To be precise: the summations in (7.23) and (7.24) have to be carried out in principle
over a complete set of basis states, including the respective continua. The fine art
of scattering theory consists mainly in finding appropriate basis sets: these must de-
scribe the scattering process of interest as completely as possible, while on the other
hand rapid convergence of the series expansions is desired. From such a solution
one obtains a whole set of scattering amplitudes fba (θ, ϕ) for the transitions from
an initial state |a into the final states |b. In the most general case they depend on
the polar as well as on the azimuthal scattering angle, and of course on the relative
initial kinetic energy T in the CM system.
For a fully quantum mechanical solution of the problem one will again expand
the scattering wave functions into partial waves. The asymptotic solutions of the
corresponding radial equations – complementary to the elastic case (6.84) – may be
written as:
π π
uΓba (R) ∝ δba sin kb R − b + KbaΓ
cos kb R − b (7.25)
R→∞ 2 2
for open channels and
uΓba (R) = 0 for closed channels.
R→∞
The K-matrix introduced here is real and symmetric, and Γ represents a set of quan-
tum numbers which are conserved during the collision. The S- and T -matrices al-
ready introduced in Sect. 6.4.6 are related to the K-matrix:
Γ
1 + iK Γ
2iK
SΓ = SΓ −
and TΓ = 1= . (7.26)
Γ
1 − iK Γ
Ê − iK
Per definition (6.118) the S-matrix is unitary,
∗
Sij Sj†k = Sij Skj = δik and specifically |Sij |2 = 1, (7.27)
j j j
holds. In contrast to purely elastic scattering the S-matrix elements may no longer
be expressed by just one real scattering phase according to (6.114). Instead, one may
represent inelasticity of the process by a complex scattering phase ηba + iμba :
Quite formally, the scattering process may now be written in analogy to (6.109)
|k a a →
S|k a a = |k a a + T|k a a. (7.30)
Just as in Sect. 6.4.1 the scattering amplitude is found by projection onto the final
state, in complete analogy to (6.110):
476 7 Inelastic Collisions – A First Overview
M̄/me
fba (θ, ϕ) = − k b b|T|k a a
2πEh a02
M̄/me
=− k a a|T† |k b b∗ = fab
∗
(θ, ϕ). (7.31)
2πEh a02
The second line describes the time inverse process: de-excitation of the atom from
the excited state |b into the ground state |a. This simply follows from the fact that
T is a Hermitian operator.
In terms of the partial wave expansion the scattering amplitude may now be writ-
ten in analogy to (6.115)
1 (Γ )
fba (θ ) = √ (2 + 1)P (cos θ )Tba (k), (7.32)
2i ka kb
assuming for the moment cylindrical symmetry of the problem. We shall see in
Sect. 8.1, that it may be necessary to extend the summation if different possibilities
exist to realize the set Γ of conserved quantum numbers. The differential elastic
(a = b) and inelastic (a = b) cross sections (DCS) for scattering into the solid angle
dΩ at θ, ϕ and into a specific channel b is, according to (6.73):
dσba kb 2
Iba (θ, ϕ) = = fba (θ, ϕ) . (7.33)
dΩ ka
The respective integral cross sections are found by integrating over all scattering
angles; the total cross section is obtained by summation (and possibly integration)
over all open channels. These two different quantities are unfortunately not always
clearly distinguished in the literature.
With (7.33) and (7.31), the relation between the differential cross sections for
excitation and de-excitation may be written
ka2 Iba (θ, ϕ) = kb2 Iab (θ, ϕ) or T Iba (θ, ϕ) = T Iab (θ, ϕ), (7.34)
with T = (2 /2M̄)ka2 and T = (2 /2M̄)kb2 = T − Wba . For the corresponding in-
tegral cross sections we have
ka2 σba ka2 = kb2 σab kb2 or T σba (T ) = T σab T . (7.35)
The ansatz (7.23) is still fairly general,3 and we shall now specialize it for heavy
particle scattering. Corresponding to B ORN -O PPENHEIMER approximation the full
Hamiltonian (7.20) will be split into
H (el)
= T(R) + H (7.36)
(el) = T(r) + V (AB) (r 1 , r 2 . . . r N , R).
with H (7.37)
The choice of an appropriate basis for the φj (r; R) will be a key theme in the
following sections.
In any case, a set of coupled equations for the relative motion in R is obtained
by insertion of (7.23) into (7.19), multiplication from the left with φj (r, R)| and
integration over all inner coordinates r. One obtains
(R) 2 2
T (R)
+ Tjj + Ujj − kj ψj (R) (7.38)
2M̄
=− P jj · û + Tjj(R)
+ Ujj ψj (R).
j =j
and the so called non-adiabatic coupling elements are determined according to the
definition (7.21) by
jj · û = − φj |∇ R |φj ∇ R
2
P and (7.40)
M̄
2
Tjj = −
(R)
φj |∇ 2R |φj . (7.41)
2M̄
The potential matrix Ujj as well as the matrix elements of the momentum term
jj · û and of the kinetic energy Tjj are functions of R and depend on the choice
P
of the basis. A key point here is the skillful choice of this basis. In the following
3 The antisymmetrization necessary in the case of electron scattering is, however, not yet included
we shall discuss two alternative representations and their implications, as first in-
vestigated systematically by S MITH (1969) with respect to the treatment of inelastic
heavy particle collisions.
The most obvious approach towards a theoretical treatment of inelastic heavy par-
ticle collisions is to follow the spirit of the B ORN -O PPENHEIMER approximation
outlined in Sect. 3.2 – which is particularly well suited for not too high kinetic ener-
gies. One uses a basis set φj (r; R), for which the potential matrix (7.39), i.e. H(el)
becomes diagonal (Ujj = Ujj δjj ). In the case of atom-atom or atom-ion scattering
the thus defined wave functions φj correspond exactly the electronic eigenfunctions
of the molecule AB,
(el) φj (r; R) = Ujj (R)φj (r; R)
H (7.42)
with φj |φj = δjj for all R,
Table 7.1 Order of magnitude of energies and coupling matrix elements for molecular bonding
and inelastic collisions as relevant in B ORN -O PPENHEIMER approximation
Application Bound molecule Inelastic scattering T ≥ Ujj
T
Ujj Eh /2
Dominant in BO T m e
M̄
Ujj Ujj
jj ·û
P
jj ·û
P
Neglected in BO T 4 m e
M̄ Ujj me T
M̄ Ujj
= uR
ur
Tjj
(R) Tjj
(R)
me
me
Order of magnitude T M̄ Ujj
M̄
The brackets indicate here integration over all R, i.e. over the whole collision
process. We now have to distinguish two cases if one wants to compare the terms in
the S CHRÖDINGER equation (7.38):
1. Bound molecular states: the kinetic energy in this case is identified with the vi-
brational energy of the molecule AB, T = Wv . The diagonal potential matrix el-
ement Ujj = We is its electronic energy. According to (3.6) we had estimated:
T ∼ me /M̄Ujj .
collisions, since the ratio of electron mass me to (reduced) heavy particle mass M̄
is always very small.
The relative magnitude of the non-adiabatic coupling P jj · û is determined by
the ratio of heavy particle to electron velocity, uR to ur , respectively. As long as
uR /ur is small, the (adiabatic) B ORN -O PPENHEIMER approximation holds and in-
elastic processes have a very small probability only. This obviously corresponds to
the M ASSEY criterium treated in Sect. 7.1.4. Inelastic processes may occur either
at much higher energies, or if the energy difference Ubb (R) − Uaa (R) to overcome
gets small, e.g. in the vicinity of potential curve crossings.
With these assumptions, the following set of linear ODEs emerges from (7.38),
describing the inelastic heavy particle scattering problem in the adiabatic represen-
480 7 Inelastic Collisions – A First Overview
tation:
2 2 2 kj2
− ∇ + Ujj (R) − ψj (R) = − jj · ûψj (R)
P (7.43)
2M̄ 2M̄ j
2
or more explicitly = φj |∇ R |φj ∇ R ψj (R).
j
M̄
These ODEs thus replace (6.61), valid for elastic scattering, and the asymptotic
behaviour of their solutions defines the scattering amplitudes fba (θ, ϕ) according to
(7.24). With (7.22) the wavenumbers kj for each channel are derived from the initial
kinetic energy T and the respective asymptotic inner energies Ujj (±∞) = Wj (in
particular Wa and Wb ).
We emphasize that the non-adiabatic coupling terms on the right hand side of
(7.43) are scalar products of two vectors. One may view this in full analogy to
the situation with optically induced dipole transitions (Sect. 4.3, Vol. 1), where the
scalar product of the dipole transition matrix element and electric field or (more
specifically) its polarization vector is responsible for the transitions and determines
the selection rules. In the present case P jj · û determines which transitions are pos-
sible and which cannot be induced by collisions. Since û = −i∇ R /M̄ and acts on
R as well as on ΘCM and ΦCM , one distinguishes radial and rotational coupling. As
we shall explain in detail in Sect. 7.4.2, radial coupling can only induce a change in
the main quantum number (without change of angular momentum), while rotational
coupling can be expressed as a linear combination of the orbital angular momentum
components (see e.g. S MITH 1969). Hence, changes of the atomic angular momenta
are possible as a consequence of collisions.
While the adiabatic procedure just outlined can be used fairly general for inelastic
heavy particle scattering, it does not necessarily converge rapidly, in particularly not
for high kinetic energies. But even at low kinetic energies, as an alternative to the
adiabatic basis a so called diabatic basis may be convenient. One tries to choose
this basis in such a manner, that the matrix elements P jj (R) and T(R)
jj of the non-
adiabatic coupling disappear, i.e.
!
φj (r; R)∇ R φj (r; R) ≡ 0 (7.44)
should hold. It is not always possible to achieve this in an unambiguous manner, and
the price to be paid for it is in any case a non-diagonal potential matrix Ujj (R). We
shall not go into the details of how to find such a diabatic representation. Rather, we
shall later on discuss some characteristic examples. In Sect. 7.4.5 we shall illustrate,
e.g., that the characterization of a basis as adiabatic or diabatic may even depend on
the type of coupling.
7.3 Scattering Theory for the Multichannel Problem 481
adiabatic
Ubb trajectory
Uaa
uPab
diabatic
Fa trajectory
Wba(Rx) Ubb
(D)
Fb
4a
(D)
(D)
Uaa
4×Uba
Rx R
(D)
Fig. 7.17 Adiabatic Ujj and diabatic Ujj representation of the potentials (upper and lower panel,
respectively), describing inelastic processes in the vicinity of an (avoided) crossing according to
(D) (D)
S MITH (1969). In the diabatic case, one may expand Ujj (R) = Ujj (Rc ) + Fj (R − Rc ) linearly
(· · · · · ·). In the adiabatic representation, transitions between states |a and |b are induced by the
non-adiabatic coupling element uP ab (- - -), while in the diabatic representation the non-diagonal
(D)
potential matrix element Uab (- - -) is responsible for the transition
Figure 7.17 compares the two different perspectives for the important case of a
localized (avoided) crossing of two potentials at an internuclear distance Rx . We
have already encountered such crossings previously, e.g. for the alkali halides in
Sect. 3.7.4. They are found rather often in diatomic systems A–B, in particular for
molecules with ionic bonding.
In the adiabatic representation, upper panel of Fig. 7.17, we denote the poten-
tials with Uaa and Ubb and the corresponding non-adiabatic coupling element with
Pab (here we show very schematically only the radial component). Note, that the
character of the adiabatic states which belong to these adiabatic potentials changes
usually very rapidly in the vicinity of the crossing, as indicated in Fig. 7.17 by the
line colour. At the crossing point, the energetic distance Wba (Rx ) of the potentials
is of course much smaller than its asymptotic value Wba (∞) = Ubb (∞) − Uaa (∞).
Hence, transitions between the two states |a and |b may occur already at rather
moderate kinetic impact energies T (at some tenth eV to a few keV), rather than at
much higher energies expected according to the M ASSEY criterium ξ ≤ 1 for Wba
(see Sect. 7.1.4). In collisions of A with B these avoided crossings are typically
reached by trajectories from a rather narrow range of impact parameters b Rx .
They may, however, be identified easily via the reduced scattering angle T θ = τ (b).
Thus, when treating such transitions it is usually sufficient to consider a two state
model.
482 7 Inelastic Collisions – A First Overview
where the plus and minus sign holds for Uaa and Ubb , respectively. At the (diabatic)
(D) (D)
crossing Uaa (Rx ) = Ubb (Rx ), the splitting of the adiabatic potentials is thus given
by
(D)
Wba (Rx ) = Uaa (Rx ) − Ubb (Rx ) = 2Uab (Rx ). (7.46)
Equivalently, if the splitting is known, the diabatic coupling potential at the crossing
point is obtained from:
(D) 1
Uba (Rx ) = Uaa (Rx ) − Ubb (Rx ) . (7.47)
2
Here, r A and r B stand for the respective inner coordinates of A and B, jA and jB
for their quantum numbers. Since φj (r, R) is now independent of R, per definition
(7.44) holds. Also, all Tjj disappear. The components of the Hamiltonian (7.20)
(R)
7.3 Scattering Theory for the Multichannel Problem 483
(A)
H (B)
H
= T(R) + V (R, r) + T(r A ) + V (A) (r A ) + T(r B ) + V (B) (r B ) .
H
(el)
H
Now, the non-diagonal terms of the potential matrix Ujj (R) according to (7.39) do
not vanish, and may induce transitions.
The further calculation becomes more transparent if the inner structure of one
of the collision partners does not play a role. For example, A may be an ion with
a closed rare gas shell which cannot be excited; or A may be an electron at high
kinetic energy so that electron exchange with the target B can be excluded. In this
case (7.48) simplifies to φj (r; R) = φjB (r), i.e. our AB basis is now simply given
by the eigenfunctions of the target. Instead of (7.36) and (7.37) the Hamiltonian is
now
= T(R) + V (R, r) + H
H (B) . (7.49)
V (R, r) is the interaction of the N electrons and the nucleus (charge Ze) of the
(B) is the N electron Hamiltonian of the target
target B with the projectile A, and H
atom B. Its eigenvalues Wj and wave functions φj (r) ≡ φj (r 1 , r 2 . . . r N ) are com-
puted as we have discussed in detail in Sect. 10.1, Vol. 1. In electron atom scattering
the potential is
N
e2 Ze2
V (R, r) = − , (7.50)
4πε0 |R − r n | 4πε0 R
n=1
Obviously, the structure of this system of coupled ODEs is independent of the spe-
cific nature of the interaction potential V (R, r) between projectile A and target B.
They are exact as long as the basis φj used is complete and electron exchange does
not play an essential role. In practice, however, the calculation will always be limited
to a finite basis and exchange can only be neglected for truly high kinetic energies.
The implications have to be judged in each case individually (see e.g. K IMURA and
L ANE 1989). We shall come back to this aspect in Chap. 8.
484 7 Inelastic Collisions – A First Overview
Section summary
• The quantum mechanical treatment of multichannel heavy particle scattering
starts with an ansatz (7.19)–(7.23) in the spirit of the B ORN -O PPENHEIMER
approximation – trying to separate the internal coordinates (r of the colliding
partners A and B from their relative motion R by a sum j φj (r; R)ψj (R)
over all relevant channels j .
• Generalizing the formalism used for elastic collisions, the asymptotic solu-
tion (7.24) for the inelastic case is composed of (i) the product of an incident
plane wave with the initial internal wave function φa (r; R), and (ii) the sum
over products of outgoing spherical waves with φj (r; R) for all relevant final
channels j . The respective amplitudes are the scattering amplitudes fj a (θ, ϕ)
from which the inelastic DCS (7.33) are obtained.
• The dynamics of inelastic collisions is described by the wave functions ψj (R)
for each channel j . They may be derived from a set of coupled ODEs (7.38).
In principle, inelastic processes can be induced by non-adiabatic coupling el-
ements, dominantly by P jj · û, or by off-diagonal elements of the interaction
potential Ujj .
• Two representations of the internal wave functions φj (r; R) are discussed: in
the adiabatic basis (standard BO approximation) the potential matrix is diag-
onal, Ujj = δjj Ujj , and transitions are induced by P jj · û. In the diabatic
basis the adiabatic coupling vanishes and the off-diagonal potential matrix el-
ements induce transitions. While both approaches may differ with respect to
rapid convergence, they are in principle equivalent as discussed in some detail
for an (avoided) curve crossing (Fig. 7.17).
As we have seen, inelastic heavy particle collisions typically require high kinetic
energies. Partial wave expansion – for electron scattering the method of choice –
will thus converge extremely slowly since high angular momenta are involved –
even at moderate kinetic energies. On the other hand, even at relatively low energies
the DE B ROGLIE wavelengths in heavy particle collisions are small compared to
typical distances over which the interaction potential changes significantly. Hence,
semiclassical methods are ideally suited for these scattering problems. In Chap. 6, in
particular in Sect. 6.4.4, we have already documented and discussed their efficiency
for elastic scattering. For inelastic heavy particle collisions semiclassical theory is
beyond any doubt the most commonly used approximation. With appropriate mod-
ifications it is used whenever transitions from one potential surface onto another
are to be described, if they are induced by nuclear motion. Specifically, this also
holds for the dynamics in isolated, polyatomic molecules, where electronic tran-
sitions, internal rearrangement or dissociation may occur as a consequence of an
initial photoexcitation process.
7.4 Semiclassical Approximation 485
The basic idea is to compute a classical trajectory R(t) for the relative motion
of the collision partners – if appropriate a trajectory for all the constituent atoms or
ions if molecules are involved. This trajectory is used to derive time dependent in-
teraction potentials Ujj (R(t)) and/or coupling elements P jj (R(t)). For complex
systems at moderate kinetic energies present state-of-the-art computations integrate
the classical motion of all heavy particles involved in the problem as completely
as possible. For such molecular dynamics (MD) calculations powerful commercial
programmes exist. They even allow one to predict the relevant potentials and forces
in the vicinity of the classical trajectory, ‘on the fly’ is the standard term, by means
of suitable quantum chemical structure programmes. When treating polyatomic sys-
tems, this allows one to limit the computational efforts of complex potential hyper-
surfaces to those geometrical regions which are really required for the dynamical
study.
For more simple atom-atom or atom-ion collisions it is usually sufficient to deter-
mine trajectories on an potential which has been suitably averaged between initial
and final state. Even the assumption of a straight line trajectory R(t) = ut + b often
leads to satisfactory results – specifically so at very high energies. The electronic
transitions are then determined from the time dependent interaction potentials or
non-adiabatic coupling elements, depending on whether a diabatic or an adiabatic
basis is used.
The time dependent S CHRÖDINGER equation
∂
H − i
(el)
Ψ (t, r) = 0 (7.53)
∂t
for the electronic wave function, with H (el) according to (7.37), together with
the classical trajectory R(t), now replaces the stationary S CHRÖDINGER equation
(7.19) for the whole scattering system. Instead of the ‘ansatz’ (7.23) a time depen-
dent series is now used,
Ψ (t, r) = cj (t)e−iϕj (t) φj r; R(t) , (7.54)
j
4 One finds slightly different notations in the literature, which differ by i or i/ in the definition of
As discussed in Sects. 7.3.3 and 7.3.4, it depends on the individual case whether
one uses an adiabatic or diabatic representations for a practical computation of the
transition probabilities according to (7.56). There are no general rules and often both
approaches are tested.
If one chooses the adiabatic basis according to (7.42), the non-adiabatic coupling
elements for j = j are now time dependent and can be derived from (7.53):
∂
Gjj (t) = φj r; R(t) φj r; R(t) . (7.57)
∂t
As illustrated in Fig. 7.16, the classical trajectory is given by the internuclear dis-
tance R(t) and the angle ΘCM (t) in the centre of mass system.5 The collision co-
ordinates are usually defined such that the z(col) -axis is parallel to the relative mo-
mentum P and to the relative velocity u of the colliding particles, prior to collision.
The x (col) -axis lies in the collision plane defined by P and P and points into the
direction into which the particle is scattered. With these definitions, the angular ve-
locity Θ̇CM has a positive or negative sign depending at which side the particles
pass each other: positive and negative sign of ΘCM (∞) correspond to an effectively
attractive or repulsive potential (for the trajectory shown in Fig. 7.16 Θ̇CM > 0
holds and ΘCM (∞) > 0). The nuclear angular momentum6 in y (col) -directions is
y = R × P and its magnitude |y | = = M̄bu = M̄R 2 Θ̇CM is obtained from im-
pact parameter b and relative velocity u. With the reduced mass M̄ and the time
dependent internuclear distance R(t) the angular velocity of the relative motion be-
comes:
bu
Θ̇CM = ± =± . (7.58)
R2 M̄R 2
With this we may write the time derivative of the wave function, which enters into
the coupling matrix element (7.57), explicitly as
Here we have inserted the radial velocity uR = dR/dt as well as the electronic an-
gular momentum of the atom L y = −i∂/∂ΘCM . With (7.59) the coupling element
5 As in the elastic case, the classical deflection angle Θ has a well defined sign, in contrast to the
scattering angle θ .
6 Werecall that we use (a vector) and (a number) for the nuclear angular momentum and its
quantum number, respectively (to be distinguished from the electronic orbital angular momen-
tum L).
7.4 Semiclassical Approximation 487
∂ y |φj .
Gjj (t) = uR φj | |φj ± φj |iL (7.60)
∂R M̄R 2
radial coupling rotational coupling
The two components of Gjj (t) represent radial and rotational coupling, respec-
tively – which we have already mentioned above. The radial coupling element arises
from changes of the internal wave function of the collision system A–B with inter-
nuclear distance R. It can only couple states with equal angular symmetry, i.e. Σ
with Σ and Π with Π states.
In contrast, rotational coupling changes the symmetry of the molecular states.
The operator L y changes the electronic angular momentum with respect to the
y (col) -axis by ±, so that it can induce Σ → Π or Π → Σ or Π → transi-
tions: we have to note here that the symmetry of the molecular states refers to the
internuclear axis z(mol) between A and B (see Fig. 7.16). The latter rotates during the
collision around y (col) with the angular velocity Θ̇CM . The electronic charge cloud
would stay space fixed parallel to z(col) – except for the molecular interaction which
tries to keep the orbitals aligned along z(mol) . It is this very fact that the nuclear an-
gular velocity translates directly into electronic angular momentum. Typically one
y |φj is nearly independent of R, so that overall the rotational cou-
finds that φj |L
pling (7.60) which is ∝ 1/R 2 dominates for small internuclear distances.
Alternatively one may of course choose a diabatic basis according to (7.44). In
Sect. 8.1 we shall come back to this again in the context of electron scattering. For
(D)
heavy particle scattering one will try to make the diagonal matrix elements Ujj of
the potential as equal as possible to the adiabatic energies of the system – except
in the vicinity of crossings. In any case, the diabatic coupling element (7.57) is
obtained from
(el)
Gjj (t) = Ujj R(t) = φj r; R(t) H φj r; R(t) . (7.61)
If one finally knows all potential matrix elements (7.60) or coupling terms (7.61) as
well as the trajectories, the coupled ODEs (7.56) can be integrated in principle by
standard methods. One has to evaluate the excitation amplitudes cb (t) for t → ∞.
Specifically, for exciting state |b from state |a the initial conditions ca (−∞) = 1
and cb (−∞) = 0 for j = a have to be applied. The transition probability (for one
trajectory at a given impact parameter) is then |cb (∞)|2 .
More generally, the transition amplitudes cj (t) are brought into the form (7.25)
by relating the impact parameter b = /kj to the corresponding angular momen-
tum and the outgoing wavenumber kj . Then one applies (7.26) to obtain the S-
and T-matrix elements, from which the scattering amplitudes (7.32) are derived, and
finally the differential cross section (7.33). Alternatively, for sufficiently high ener-
488 7 Inelastic Collisions – A First Overview
gies one may – as for elastic scattering according to Sect. 6.4.4 – also use the eikonal
method. In the limit of small scattering angles and high energies one obtains (see
e.g. D UBOIS et al. 1993) the scattering amplitude
M̄u
fba (θ, ϕ) = (−i)1+|Mb −Ma | e−i(Mb −Ma )ϕ (7.62)
∞
× bdbJ|Mb −Ma | (Kb) cba (b, ∞) − δba
0
(cf. Eq. (6.59) for shadow scattering). This scattering amplitude describes the tran-
sition between states |aMa and |bMb with projection quantum numbers Ma and
Mb of the electronic angular momentum of the target with respect to the z-axis. The
latter is taken here parallel to the relative velocity before collision. J|Mb −Ma | (x) is a
B ESSEL function and K = 2(M̄u/) sin(θ/2) the change of the wave vector during
the collision (∝ to the momentum transfer).
A few general characteristic of the solution may be discussed. The special sig-
nificance of the exponential phase factors exp[i(ϕj − ϕj )] is already recognized in
1st order perturbation theory, where for a transition b ← a
2
+∞
cba (∞)2 = 1 Gba (t)e i(ϕb (t)−ϕa (t))
dt . (7.63)
2 −∞
One sees that – independent of the strength of the coupling Gba – excitation is
only possible in those regions of the trajectory where exp[i(ϕb (t) − ϕa (t))] does not
vary too fast – otherwise the oscillations would lead to cancellation of positive and
negative contributions. Two cases can be distinguished:
1. For direct excitation the potentials of the two states a and b are separated over
the whole trajectory and their distance is approximately given by the asymptotic
conditions. With (7.55) the phase difference is then approximately
(Wb − Wa ) R (Wb − Wa )
ϕ(t) = ϕb (t) − ϕa (t) ∼ t, (7.64)
uR
and the exponential factor in the integrand of (7.63) oscillates with time. De-
pending on whether this oscillation is slow (high velocity uR ) or fast (small uR )
along a typical interaction range, the transition probability will be significant or
disappear. This consideration confirms quantitatively and generally M ASSEY’s
adiabaticity criterium: the phase difference ϕ(tcol ) for the total interaction time
tcol is according to (7.64) indeed identical with the M ASSEY parameter (7.13).
2. Complementary to this, non-adiabatic transitions may also occur at a crossing
of the potential curves for state |a and |b, as illustrated in Fig. 7.17. This
may happen already at low velocities and between channels with asymptoti-
cally very different internal energies. In this case, the phase difference ϕ(t)
remains sufficiently small in the vicinity of the crossing at Rx , at least as long
as (R − Rx ) ≤ uR /(Ubb − Uaa ) uR /Wab (Rx ) holds, so that the coupling
7.4 Semiclassical Approximation 489
can induce the transition. In this case one may focus onto the two relevant states,
and has to solve simply a system of two coupled, linear differential equations
according to (7.56).
In addition to the direct numerical solution of the coupled ODEs (so to say by
‘brute force’) a number of more or less elegant approximations exist which also lead
to the desired result. In the following we shall discuss the important method devel-
oped by L ANDAU (1932) and Z ENER (1932), which even today is still often used as
a simple but reasonable approach for a quantitative understanding of transitions at
(avoided) crossings.
We consider the transition from state |a to |b through a crossing localized at Rx ,
following the diabatic picture as represented in the lower panel of Fig. 7.17. We
calculate the transition probability between the diabatic states |a (D) and |b(D) .
According to (7.56) with (7.61) one has to solve the following coupled differential
equations for the initial conditions caa (−∞) = 1 and cba (−∞) = 0 (the second
index refers to the initial state):
i (D)
ċaa (t) = − Uab (t)eiϕab cba (t) (7.65)
i (D)
ċba (t) = − Uba (t)e−iϕab caa (t) (7.66)
1 t (D) (D)
with ϕab (t) = Uaa t − Ubb t dt . (7.67)
(D) (D)
In the vicinity of the crossing at Rx , where Uaa (Rx ) = Ubb (Rx ), one expands the
diabatic potentials for small distances R = R − Rx : 7
The constant Fab is the difference of the slopes of both potentials at the crossing
(i.e. the difference of the respective forces onto the trajectory), and R = uR t, if
(D)
time zero is set when the trajectory passes the crossing. The coupling element Uab
is assumed to be constant over the crossing region:
(D)
U (R) = U (D) (R) = U (D) . (7.70)
ab ba ab
7 Forsimplicity of the derivation we consider here only radial coupling, for which the L ANDAU -
Z ENER model is typically used. A LLAN and KORSCH (1985) have shown, however, that the for-
malism may also be applied to rotational coupling.
490 7 Inelastic Collisions – A First Overview
π Fab uR
ϕab (t) = α t 2 with α = , (7.71)
2 π
and the coupled ODEs (7.65) and (7.66) simply become:
i (D) π 2
ċaa (t) = − Uab eiα 2 t cba (t)
(7.72)
i (D)
ċba (t) = − Uba e−iα 2 t caa (t).
π 2
It is instructive to first have a look at the solution in 1st order perturbation theory.
We thus insert on the right of (7.72) caa (−∞) = 1 and cba (t) = cba (−∞) = 0,
√
substitute x = αt, integrate from −∞ to ∞, and obtain:
∞
i signum(α) − i (D)
e−i 2 x dx =
(D) π 2
cba (∞) = − √ Uab √ Uab .
α −∞ |α|
For exp(−iπx 2 /2)dx = cos(−πx 2 /2)dx + i sin(−πx 2 /2)dx we have inserted
here the well known limits of the F RESNEL integrals. Thus, in 1st order perturbation
theory the probability for the transition |a (D) → |b(D) during one transit of the
crossing is:
(D) (D)
(D) 2|Uab |2 2π|Uab |2
wba = = 2πξ. (7.73)
2 |α| |Fab |uR
Of course this approximation only holds for 2πξ
1. As already shown by Z ENER
(1932), the system (7.72) of coupled ODEs can be solved exactly. An attractive
modern derivation is found e.g. in W ITTIG (2005). Accordingly, the probability for
a transition between the adiabatic states |a → |b is
which is identical to the probability for remaining on one of the diabatic potentials.
Inversely, the probability for a transition from one to the other diabatic state
|a (D) → |b(D) (i.e. to stay on the adiabatic potential) is
wba = 1 − e−2πξ ,
(D)
(7.75)
(D)
|Uab |2 ∂ (D)
ξ= with Fab = U − Uaa
(D)
. (7.76)
uR |Fab | ∂R bb
7.4 Semiclassical Approximation 491
x (col)
(D)
For pure radial coupling, according to (7.46) the coupling potential Uab = Wab /2
is related to the splitting Wab of the adiabatic potentials. Hence,
1. For a trajectory, which has achieved the jump |b ← |a on the way in with a
probability wba , on the way out the probability to remain in this state |b is given
by (1 − wba ). The total transition probability from one to the other adiabatic state
is thus wba (1 − wba ).
2. On the other hand, a trajectory which has not made the jump on the way in
(probability 1 − wba ), may make it on the way out with a probability wba . Again,
the overall probability for a jump is (1 − wba )wba .
Both trajectories contribute to transitions, and the total probability for the transition
|b ← |a at a given impact parameter is8
the probability to remain in the initial state during the overall process is wba wba + (1 −
8 Inversely,
tot
wba = 2wba (1 − wba ) with wba = exp(−2πξ ). (7.78)
This is the L ANDAU -Z ENER probability for a non-adiabatic transition at a
localized crossing, ξ being the modified M ASSEY parameter (7.77).
The dependence of wba tot on the adiabaticity parameter ξ is similar to that sketched
in Fig. 7.5(c) on page 458. It has, however, at ξ 0.11 a maximum of wba tot = 1/2:
As a still rather straight forward example we discuss the collision of an ion with
an excited atom, during which two pronounced, non-elastic processes can occur: a
“super-elastic” (exothermic) de-excitation and an excitation process:
*
+
2 → Na+ + Na(3s 2 S) + (T + 2.10 eV)
Na + Na 3p P3/2 + T (7.79)
→ Na+ + Na(3d 2 D) + (T − 1.48 eV).
Figure 7.19 shows the relevant interaction potentials of the system Na+ 2 . Marked
by circles are three crossings of which we consider here only (C) for the transition
|3p → |3s and (B) for the transition |3p → |3s.9 In both cases we have a cross-
ing between molecular |nΣu and |n Πu states. Their potentials are obtained as
solutions of the electronic S CHRÖDINGER equation (7.42) for the system Na+ 2.
In the usual static picture, used in Chap. 3 for the presentation of molecular po-
tentials as a function of R, these are genuine crossings. Even though the radial
parts of the wave function depend on R, due to the different symmetries of their
angular parts the transition matrix element disappears nΣu |∂/∂R|n Πu ≡ 0,
and the states do not split along the R coordinate. On the other hand, according
to (7.60) the rotational coupling (/M̄R 2 )nΣu |iLy |n Πu does not disappear
and may induce transitions. Quantitatively one finds that the coupling matrix ele-
ments 3pΠu |L y |3sΣu and 3pΣu |L̂y |3dΠu are about (A LLAN and KORSCH
1985). The terminology “diabatic” and “adiabatic” is here somewhat confusing,
rather one should speak about crossing and non-crossing potentials, respectively.
When discussing the L ANDAU -Z ENER model in Sect. 7.4.4, we have denoted
(D) (D)
the potentials of the states which cross as a function of R by Uaa and Ubb . In the
present case (with strong rotational coupling) these potential are the adiabatic eigen-
9 Crossing (D) is important for the dependence of the process on polarization – which we cannot
describe here.
7.4 Semiclassical Approximation 493
≈
potential energy / eV
and (C) are responsible. In
3 3p 2 Σ g
the insets these crossings are C 3p 2 Σ u
magnified. The other D Na(3p 2P)+ Na+
potentials have no influence 2
on the processes under 3p 2 Σ g
3p 2u
discussion 1
3s 2 Σ u Na(3s 2S1/2)+ Na+
0
3s 2 Σ g
-1
0 10 20 30 40
R / a0
10 Thisis a consequence of the rotation of the internuclear axis during the collision (rotational
coupling). We note here in passing, that this splitting corresponds directly to lambda-type doubling
in molecular spectroscopy (Sect. 3.6.6), a splitting of energy levels into Λ+ and Λ− states for
higher rotational quantum numbers due to the coupling of the electronic angular momentum Λ
with the nuclear rotation N .
494 7 Inelastic Collisions – A First Overview
|S12|2 b / a0
0 1 2 3 4 5 6
1.0
0.8
0.6
LZ
0.4
0.2
AKo85
0.0
0 400 800 1200 1600
ℓ
Fig. 7.20 Calculated transition probability for the transitions |3p → |3s in collisions of Na(3p)
with Na+ at an initial kinetic energy 47.5 eV – here as a function of the relative nuclear angular
momentum or of the impact parameter b, respectively. Shown are the semiclassically exact values
of |S12 |2 according to A LLAN and KORSCH (1985) ( AKo85), compared with the simple
L ANDAU -Z ENER (LZ) transition probability 4 exp(−2πξ )[1 − exp(−2πξ )] ( )
as JWKB phases according to (6.94). Without entering into details (see e.g. A L -
LAN and K ORSCH 1985; BANDRAUK and C HILD 1970), we also communicate the
resulting S-matrix:
()
Saa = e−2πξ + 1 − e−2πξ e−2iδ e2iηa
Sbb = e−2πξ + 1 − e−2πξ e2iδ e2iηb
()
(7.81)
Sab = Sba = 2ie−πξ 1 − e−2πξ sin δei(ηa +ηb ) .
() ()
(a) fluorescence ( )
laser beam
detector ion
source
3 3 d ← 3p
dσ / dΩ CM / 100 a20 sr -1
Na-oven
3s ←3p
2
atom bea
m
θ Lab
shutter
1
beam expansion
λ /4 plate
ion detecto
r polarizer
0
dye laser 0º 2º 4º 6º 8º 10º
θ CM
Fig. 7.21 Na+ + Na(3p) → Na+ + Na(3s, 3d) scattering experiment: (a) experimental setup,
schematic. (b) Differential cross section as a function of the scattering angle θCM in the CM system
at a kinetic energy TCM = 47.5 eV: measured data •, with line to guide the eye - - - according to
BÄHRING et al. (1984); semiclassical computation according to A LLAN and KORSCH (1985)
(slightly rescaled angle, see text)
in these calculations, we have rescaled the scattering angle ΘCM ∝ 1/Rx slightly
in each case. This is reasonable, since according to (6.46) the scattering angle is
approximately ΘCM ∝ 1/b (in the relevant range of distances we deal with a pure
repulsive C OULOMB potential). This leads to a better agreement between theory
and experiment which is now surprisingly good, considering the simplicity of the
theoretical model. Even the relative cross sections of both processes are computed
correctly (theory was used for absolute calibration of the cross sections).
Finally, in Fig. 7.22 the measurements at different collision energies are sum-
marized. Obviously, the reduced scattering τ = T θ for which the differential cross
section reaches its maximum is indeed nearly constant, as expected in first order
approximation according to (6.52).
We have already mentioned above, that different phases may be accumulated on dif-
ferent trajectories in a collisional excitation process via curve crossing (see Fig. 7.18
where the excitation may occur either at the crossing marked in or out). Thus, the
()
scattering matrix elements Sab (7.81) for inelastic processes may lead to oscilla-
tions as a function of (or b) and consequently as a function of scattering angle (as
well as of kinetic energy). The phenomenon has already been treated by S TÜCKEL -
BERG (1932) in the heydays of early quantum mechanics. The calculation shown in
Fig. 7.20 for the process just discussed exemplifies such behaviour. Unfortunately,
after conversion into differential cross sections, these oscillations are not significant
enough to be observed experimentally in this case as documented by Fig. 7.21(b).
There are, however, numerous examples where such interferences can be observed.
We discuss here charge exchange between the quadruply charged C4+ ion and a
neutral He atom. This example is interesting for several reasons and quite popular
in scattering physics. Experimentally, the system is easy to access (highly charged
C ions are obtained from standard ion sources), and a whole series of exothermic
electron capture processes into doubly or triply ionized Cq+ ions may be studied.
These can easily be identified by their different energy gain W for different final
7.4 Semiclassical Approximation 497
-20
C2+(2s2p)+He2+ [4] 2 4 6 8 10 12
40 (c) CHe 4+
-30 C3+(2p)+He+ [3]
20
C2+(2s 2)+He2+ [2] 0 [9]
[2]
C3+(2s)+He+ [1] - 20 [3]
[1]
-40 R / a0
≈
0 20 40 60 80 ∞ 0 2 4 6
Fig. 7.23 Potentials and energetics for the system CHe4+ according to P ICHL et al. (2006). State
[9] represents the entrance channel, dashed grey lines - - - indicate 2e− capture, full grey lines
−
—– 1e capture. (a) Overview; (b) critical crossing region; (c) simplified four state model accord-
ing to BARAT et al. (1990) with · · · · · · indicating a diabatic basis. Red circles mark the relevant
crossings
channels:
⎧
⎪ +
⎨C (1s n L) + He (1s) + W
3+ 2 2
C4+ 1s 2 1 S + He 1s 2 1 S → C2+ (1s 2 2sn 1 L) + He2+ + W (7.82)
⎪
⎩ 2+ 2
C (1s 2pn 1 L) + He2+ + W.
From a theoretical point of view, this is a case of pure radial coupling, in contrast
to the previous example, and both collision partners have initially identical elec-
tronic structures (1s 2 1 S0 ). In Fig. 7.23 we show the potential curves for all exother-
mic processes according to P ICHL et al. (2006) (for simplicity, configurations are
given without the 1s electrons). The potentials are dominated by C OULOMB repul-
sion in the exit channels, the initial channel corresponds to a nearly horizontal line,
interrupted by avoided crossings at which the charge exchange (7.82) occurs. The
enlarged presentation (b) shows these avoided crossings more clearly (red circle).
Here too, a classical trajectory can take (several) different paths towards the classi-
cal turning point on the way in and out – just as we have discussed it in Sect. 7.4.4.
In this case, the resulting phase differences lead to pronounced oscillations in the
differential cross sections which have been studied by BARAT et al. (1990) at high
kinetic energies. Figure 7.24 gives some examples for the one and two electron
498 7 Inelastic Collisions – A First Overview
dσ
__ 2
500
/ a 0 rad -1 2e- capture 1e- capture
dθ
1600 9.6 keV 9.6 keV
400
1200
300
800 200
400 100
0 0
0 2 4 6 8 10 12 0 2 4 6 8 10 12 τ / keV deg
Fig. 7.24 Differential cross section for C4+ + He as a function of reduced scattering angle τ
according to BARAT et al. (1990) at 9.6 keV. The measured data points are compared with semi-
classical calculations for the four state model shown in Fig. 7.23(c)
capture process at 9.6 keV. Shown is the differential cross section summed over
different final states as a function of the reduced scattering angle τ = T θ . The ex-
perimental data are compared with computations from a semiclassical calculation
using the models discussed above. BARAT et al. (1990) have used the diabatic ver-
sion of a simplified four state model with the potentials shown in Fig. 7.23(c). The
agreement between theory and experiment is astonishingly good, considering the
simplicity of the model.
Note that for single electron capture (potential curves [1] and [3] in Fig. 7.23(c))
the relevant crossings are localized at very small internuclear distances. Due to the
dominant C OULOMB repulsion these can be reached only at rather high kinetic en-
ergies.
More recently, two electron capture has been studied by H OSHINO et al. (2007)
at significantly lower energies. Figure 7.25 shows a 2D representation of the double-
differential cross section as a function of the energy gain (W = T − T ) and the
laboratory scattering angle θLab . The dominant process in this case is obviously
the 2e− capture into the ground state of C2+ (1s 2 2s 2 ). Here too, S TÜCKELBERG
oscillations are very clearly resolved as a function of the scattering angle.
Section summary
• Semiclassical methods for inelastic heavy particle collisions follow the gen-
eral concepts developed for elastic processes (see Sect. 6.4.4) and combine
these with time dependent coupled differential equations for multichannel
transitions, summarized by (7.56).
• For inelastic processes which occur at localized potential curve crossings, two
types of coupling are distinguished according to (7.60): Radial coupling and
rotational coupling. The latter is closely related to lambda type doubling of
molecular electronic states, and may be rationalized due to rotation of the
7.5 Collision Processes with Highly Charged Ions (HCI) 499
W / eV
smooth and the dashed curves 20
represent W for the C2+
10
ion as expected from the
kinematics of two double 0
electron capture processes -10
with nominal energy gain 30
Wab = 33.4 and
Wab = 20.7 eV, 20
respectively 10
C2+(1s 22s 2 1S)
0
C2+(1s 22s 2p 1P)
-10
-8˚ -4˚ 0˚ 4˚ 8˚
θ Lab
molecular symmetry axis of the collision system with respect to the space
fixed laboratory system.
• In the semiclassical picture, at these localized curve crossings transitions oc-
cur along a trajectory with a probability wba on the way in (towards the clas-
sical turning point) as well as on the way out. The L ANDAU -Z ENER model
tot = 2w (1 − w ) for a non-adiabatic tran-
predicts an overall probability wba ba ba
sition, with wba = exp(−2πξ ) where ξ is essentially the M ASSEY parame-
ter – modified for a localized crossing according to (7.76) and (7.77).
• As an example, we have discussed excitation and de-excitation in Na+ +
Na(3p 2 P3/2 ) collisions. The differential cross sections are found to peak
strongly, with reduced scattering angles τ = ΘCM T nearly independent of
kinetic energies – corresponding to a narrow range of impact parameters.
• S TÜCKELBERG oscillations may occur in such inelastic heavy particle colli-
sions as a consequence of different phases accumulated by different trajecto-
ries leading to the same scattering angle. An impressive example is electron
capture in C4+ + He collisions.
In the context of L AMB shift we have already mentioned in Sect. 6.5, Vol. 1 that to-
day quite fundamental experiments are carried out with highly charged ions (HCI).
For example, H like ions of heavy elements (up to 91 times ionized uranium with
a nuclear charge Z = 92) allow to perform very sensitive tests of quantum electro-
500 7 Inelastic Collisions – A First Overview
Wpot (q) / eV
AUMAYR (1999). In principle
104
Wpot (q) can be released when
q electrons are captured 103
102
10
0 20 40 60 80 100
q
11 W
pot is the sum of all ionization potentials WI (q ) for q ≤ q.
7.5 Collision Processes with Highly Charged Ions (HCI) 501
laboratories worldwide. They are all based essentially on the same physical princi-
ples, but are quite different in their specific construction. In all three cases HCI are
generated by multiple collisions with electrons and stored in an electric field, while
strong magnetic fields (in some cases generated by superconducting magnets) keep
the electrons focussed together. In the EBIS design, electrons play an important role
in ion storage. Addition of several low Z gases allows an efficient cooling by small
angle scattering (so called evaporative cooling). For the study of collision processes,
the HCI are then extracted from the source, accelerated, selected according to mass
and energy and finally focussed onto a target or inserted into a storage ring.
In HCI collisions a manifold of processes may occur, so that efficient detection
and measuring methods for the various reaction products is crucial for the success
of such experiments. Typically, time and position sensitive methods are used, often
with coincident determination of the momentum components for several reaction
products (e.g. by COLTRIMS, which is briefly explained in Appendix B.4).
Let the binding energy (<0) of electron transferred be Wb after the collision and Wb
prior to it. For high charge q, the energy defect of the process
Q = Wb − Wb (7.84)
will typically be negative, since the captured electron is bound much stronger in
A(q−1)+ than before in B. Thus, (7.83) describes exothermic processes which – as
discussed in Sect. 7.1.1 – may have high cross sections. Due to its high potential
energy, the HCI (Aq+ ) acts in a collision with the neutral target (B), so to say,
as a ‘vacuum cleaner’ which sucks up the electrons from the target as soon as it
comes close enough. As sketched in Fig. 7.27, the potential barrier between projec-
tile and target is reduced during this process dramatically. This classical, so called
“over-the-barrier” model has already been developed during the late 80s of the past
century (see N IEHAUS 1986, and further work quoted there), but still enjoys great
popularity for qualitative discussions and makes predictions of astonishing accuracy.
We discuss here the basic ideas. When the projectile Aq+ approaches the target B,
the valence electrons (coordinate r) of B ‘see’ an effective potential V (r; R) which
changes as sketched in the scenarios Fig. 7.27(a–h) with internuclear distances R.
At a critical distance R = Rth the barrier decreases below the local binding energy
of the target electrons as indicated in Fig. 7.27(c). Projectile and target then form –
for a very short time – a quasi-molecule and the HOMO of the target B and high
lying, unoccupied n states of Aq+ come very close, energetically.
502 7 Inelastic Collisions – A First Overview
- 20 0 20 - 20 0 20
≈
0 0
R→ ∞ R→ ∞
-2 (a) t (h) -2
n=8
-4 -4
B Aq+ B+ A(q-1)+
≈
0 0
-2 (b) (g) -2
n=8
V(r;R ) / E h
-4 -4
R = 50 R = 50
0 0
(c) n=9 (f ) n=9
-2 -2
8 8
-4 7 7 -4
R th = 12.4a 0 R th = 12.4 a 0
0 0
(d ) (e)
-2 -2
-4 -4
R= 8 R= 8
- 20 0 20 40 60 - 20 0 20 40 60
electron coordinate r / a 0
Fig. 7.27 Classical over-the-barrier model for a single electron capture process for the example
He + Ar18+ (we follow here a presentation of M ORGENSTERN 2009). Plotted are the potentials
V (r; R) which the electrons (coordinate r) experience in the system for different internuclear
distances R. The heavy red arrows indicate the time sequence. Red, horizontal lines correspond to
energy levels in the isolated collision partners, and for R ≤ Rth in the quasi molecule, respectively
(full line: occupied, dashed: unoccupied). For energetic consistency the two He electrons are drawn
on different levels – even if they are of course indistinguishable in the first ionization step
Thus, the valence electron with the smallest binding energy may change its place
and will occupy these n states with high probability. It will stay there on the way
out, as indicated in Fig. 7.27(f): the probability for recapture from the n states of
A(q−1)+ to the ionized target B+ is very low, since phase space (degeneracy) is much
higher in these high lying states than in the typically light target. Mostly, processes
with high kinetic energies are studied, so that – in contrast to Sect. 7.1 – one may
assume approximately straight line trajectories. The critical distance Rth may thus
be identified with the maximum impact parameter bm for a charge exchange process
to happen. Thus, the resulting capture cross sections σ πRth 2 may become rather
large.
Quantitatively, one assumes a simplified potential
t q
V (r; R) = − − .
r R−r
Here and in the following we use again atomic units (a.u.), Eh (energy) and a0
(length). The screening of the nuclear charge of B by its inner electrons is modelled
by an effective charge t – quite similar as done in Sect. 7.1.2 (Vol. 1) for bound state
7.5 Collision Processes with Highly Charged Ions (HCI) 503
We write the binding energy of the HOMO electron at infinite distance Wb (∞,
HOMO) = −WIB (with the ionization potential WIB ). As the highly charged ion
approaches, this will be lowered. In the above-barrier model one assumes that at the
critical distance Rth this decrease is given simply by the potential of the HCI:
(B)
Wb (Rth , HOMO) = −WI − q/Rth . (7.86)
so that it becomes
√
t + 2 qt
Rth = (B)
. (7.87)
WI
And in analogy to (7.86) the energies of the n final states of the projectile ion
are lowered at Rth :
Electrons captured from B will preferentially populate such levels n for which the
energy at Rth matches best with the energy of the target electron according to (7.86):
q t
−WIB − Wb (∞, n) − ,
Rth Rth
√
q −t q + 2 qt
Wb (∞, n) −WIB − = −WIB √ , (7.89)
Rth t + 2 qt
where the latter result is obtain by using (7.87). Hence, the energy defect (7.84)
becomes:
(q − t)
Q = Wb (∞, n) − Wb (∞, HOMO) = −WI
(B)
√ . (7.90)
t + 2 qt
If we now write the binding energy of the captured electron in the projectile ion
Wb (∞, n) = −q 2 /(2n∗2 ), we finally find with (7.89) the effective quantum number
of the state predominantly populated:
√
(B) −1/2 t + 2 qt 1/2
n∗ q 2WI √ . (7.91)
q + 2 qt
504 7 Inelastic Collisions – A First Overview
12 Thescenarios sketched in Fig. 7.27 correspond to just this reaction for q = 18 (i.e. for the naked
Ar nucleus). Not completely correct but illustrative we have shown in these schematics also the
reduction of the states at R > Rth – according to (7.86) and (7.88).
7.5 Collision Processes with Highly Charged Ions (HCI) 505
800
8 Ar18+ 7 Ar15+
1500 (a) (b)
ion yield / counts
7 600 n =6
1000 400
9 8
500 200
n =6 10
0 0
-120 -80 -40 0 -120 -80 -40 0
Q value / eV Q value / eV
10 10 Ar15+
Ar18+ n =6
dσ/dθ /10-12cm2
n =7
1 n =8 1
(c) (d)
0.1 0.1
0.01 0.01
0.0 0.1 0.2 0.3 0.4 0.0 0.1 0.2 0.3 0.4
θ / mrad θ / mrad
Fig. 7.28 One electron capture according to K NOOP et al. (2008) in the reactions He − Ar18+ →
He+ − Ar17+ (1sn) (a, c) and in He + Ar15+ (1s 2 2s) → He+ + Ar14+ (1s 2 2sn) (b, d); the mea-
sured Q value spectra (a, b) are fitted (red lines) assuming occupation of the n states 6 to 10;
in (c) and (d) the experimentally determined differential cross sections as a function of scattering
angle θ are compared with theory
two terms (see e.g. U LLRICH et al. 2003; D EPAOLA et al. 2008, where, however,
the signs are defined differently). For high energies (in the experiment discussed
here >200 keV) the third term may indeed be neglected too, as one easily verifies.
Also, me v 2 /2 is usually much smaller than |p v|. Hence, negative Q (exothermic
reaction) leads to p < 0, i.e. to backscattering of the B+ .
In Fig. 7.28 some of the results from K NOOP et al. (2008) are shown. They show
very clearly that electron capture into the projectile ion occurs indeed very selec-
tively to a few states n as illustrated in Fig. 7.28(a, b). For q = 18 preferentially
n = 8 and 7 are populated, for q = 15 we see n = 7 and 6 to dominate. A look
onto Fig. 7.27(c) shows that this essentially agrees with the predictions of the over-
the-barrier model. Quantitatively according to (7.91) the values n∗ = 8.5 and 7.3,
respectively, are expected – slightly above the experimental observations – but in
view of the simplifications of the model this is an extremely surprising agreement.
The observed differential cross sections shown in Fig. 7.28(c, d) are strongly
peaked for small scattering angles, as expected. The agreement with theory is im-
pressive, considering the complexity of the processes. The computed integral cross
section by which the experiment has been calibrated are very large – owing to the
large values of the critical distance Rth . Summed over all occupied states n∗ one
finds σ (q) to be between 24 and 26.9 × 10−16 cm2 for q = 15 to 18.
506 7 Inelastic Collisions – A First Overview
The attentive reader will have noticed that the over-the-barrier model used here is
very similar to the concepts which we have used in Sect. 8.5, Vol. 1 for describ-
ing the interaction of atoms and molecules with intense femtosecond laser pulses.
Indeed, the electric field strengths which are active in the interaction with highly
charged ions are very similar to those in an intensive laser field. Thus, it is not
surprising that the ionization and fragmentation patterns observed in the two cases
often show astonishing similarities. Also the interaction time in such HCI collisions
may be extremely short. By increasing the interaction energy or by focussing onto
smaller regions of interaction this time may even be reduced by an order of magni-
tude or so. Thus, we are discussing here attosecond physics – also a very ‘hot topic’
in state-of-the-art laser matter interaction. However, the interaction pulses induced
by HCI are strictly monodirectional – a property of which laser physicists often
dream. Nevertheless, one has to state in all fairness that collision physics with HCI
does (in contrast to attosecond laser science) not know how to do pump-probe ex-
periments – which are a basic requirement for real time observations of dynamical
processes in physics and chemistry.
Section summary
• Collision experiments with highly charged ions are an area of particular cur-
rent interest. The high potential energy inherent in highly charge ions leads to
a rich variety of interesting interaction processes.
• Electron capture processes are visualized and semi-quantitatively modelled
by the classical above-barrier model. In the course of the collision, the strong
C OULOMB field of the HCI modifies the potential which the valence electrons
the neutral target ‘see’. At a critical interaction distance the C OULOMB barrier
is suppressed such that for a short time a quasi-molecule is formed. The n
states of the HCI ion which are energetically resonant at that moment to the
target states will be occupied.
• The physics of these processes closely resembles the dynamics of atoms and
molecules in intense, ultrafast laser pulses.
So far, our introduction into inelastic heavy particle scattering was restricted to rel-
atively small colliding systems (electrons, atoms, ions, small molecules), i.e. to a
very small section of relevant bimolecular interactions. Today one is, however, able
to investigate by experimental and theoretical methods also reactions of complex,
polyatomic systems. Instead of the potential curve crossings and L ANDAU -Z ENER
type transitions discussed above, one has to deal in this case with multidimensional
potential hypersurfaces and one speaks about “conical intersections” and “surface
7.6 Surface Hopping, Conical Intersections and Reactions 507
e-- gun
imaging
ion optics
pressure
gauge
extra
xtra ction
xtrac
e-
repeller
Fig. 7.29 Imaging spectrometer according to M IKOSCH et al. (2006). The key components are
a pulsed projectile beam (red dashed), an also pulsed target gas beam (white dashed) and the
imaging ion lens system which images the reaction products (pink line) onto a multi channel plate.
A special pressure gauge and an electron gun are used to characterize the gas beam and to calibrate
the detector
which the ions hit the MCP are registered, position sensitive, via a phosphorescence
screen by a computer controlled CCD camera. The time of flight is determined by a
pulsed voltage switch, triggering the detector system. From the position at which the
ions hit the detector and their time of flight one can derive the velocity components
of the scattered ions and reaction products unambiguously.
With this apparatus one may investigate, e.g. vibrational and rotational excitation
of small molecules in low energy collisions, or reactions of small molecules of the
type
X− + RY → XR + Y− . (7.96)
In this case, one ion is exchanged by another ion. One measures the angular and
energy resolved differential cross section by the imaging method.
As an example for a characteristic result Fig. 7.30(a) shows a 2D plot of reac-
tion probabilities for a so called nucleophilic substitution reaction CH3 I + Cl− →
CH3 Cl + I− as experimentally determined by M IKOSCH et al. (2008). The colour
shades (red: very high, white: average, black: vanishing) represent the scattering
intensity as a function of the x- and y-components of the relative velocity of the
scattered I− ions after the reaction in the CM system. At the kinetic energy used
here (1.9 eV) the reaction obviously is most efficient when the product ion I− is
emitted into the direction of the incoming Cl− ion, which is equivalent to saying
that the reaction product CH3 Cl is scattered into the backward direction: reaction
(7.96) is found to be a direct process. This may be explained in a suggestive manner
by a reaction mechanism as illustrated in Fig. 7.30(b): the Cl− ion attacks the target
molecule CH3 I on the side opposite to the I atom. Shown is the potential (red line),
calculated ab initio (on the MP2 level), along the so called reaction coordinate. The
reaction coordinate characterizes the path from the educts (initial molecules) to the
7.6 Surface Hopping, Conical Intersections and Reactions 509
0
1000 CH I I-
3 θ Cl - Cl -
I -4.0
I-
kcal mol -1
500 CH3
-0.4
Cl
-10
-11.6 CH3
uy / m s -1
V / eV
0 -12.2
-0.8
-20
-21.6
-500
-1000 -10 0 10
-1000 -500 0 500 1000 R C-I - R C-Cl / 10 -8 cm
(a) ux / m s -1 (b)
Fig. 7.30 Nucleophilic reaction CH3 I + Cl− → CH3 Cl + I− according to M IKOSCH et al. (2008).
(a) ‘Velocity map’ for TCM = 1.9 eV measured as a function of the relative velocity in x- and
y-direction in the CM system. The white N EWTON rings mark positions of equal energy in the CM
system. The largest circle corresponds the maximum kinetic energy after the collision. (b) ab initio
calculation of the potential along the ‘reaction coordinate’ RC−I − RC−Cl ; characteristic maxima
and minima are given in kcal /mol
final products along a trajectory with locally minimal changes of potential energy
(the minimum energy path). One may visualize this concept by a sledge ride which
leads you downhill as much as possible, occasionally also uphill but always with
minimal energetic effort. In the present case, the system has all together 12 internal
degrees of freedom – and the reaction coordinate is determined by the difference
between the distances RC−I − RC−Cl .
As the potential curve in Fig. 7.30(b) shows, the reaction proceeds through a
number of local maxima and minima on the potential hypersurface (so called tran-
sition states). Clearly, the geometry sketched here is only one out of many possibil-
ities: different collision parameters may also lead to a successful reaction, and the
orientation of CH3 I relative to the approaching Cl− ion is statistically distributed.
The preference for I− emission along the (positive) CM axis with essentially max-
imum kinetic energy possible – as experimentally observed by M IKOSCH et al.
(2008) at 1.9 eV – simply says that the orientation indicated in Fig. 7.30(b) is the
most efficient geometry for a reaction to take place. We cannot enter here into fur-
ther details, but mention that this behaviour depends strongly on the initial kinetic
energy.
Section summary
• In this short section we have discussed the nucleophilic substitution reaction
CH3 I + Cl− → CH3 Cl + I− as an example for a detailed study of a reac-
tion process. Advanced experimental methods, with miniaturized molecular
beams, VMI methods, and state-of-the-art ab initio calculations of the inter-
action potential along the reaction path can give deep insights into complex
reaction mechanisms.
510 7 Inelastic Collisions – A First Overview
References
A LLAN , R. J. and H. J. KORSCH: 1985. ‘2-state curve crossing processes involving rotational
coupling in the Na+ 2 molecular ion’. Z. Phys. A, 320, 191–205.
BÄHRING , A., I. V. H ERTEL, E. M EYER, W. M EYER, N. S PIES and H. S CHMIDT: 1984. ‘Excita-
tion of laser state-prepared Na*(3p) to Na*(3d) in low-energy collisions with Na+ : Experiment
and calculations of the potential curves of Na+ 2 ’. J. Phys. B, At. Mol. Phys., 17, 2859–2873.
BANDRAUK , A. D. and M. S. C HILD: 1970. ‘Analytic predissociation linewidths from scattering
theory’. Mol. Phys., 19, 95–111.
BARAT , M., P. RONCIN, L. G UILLEMOT, M. N. G ABORIAUD and H. L AURENT: 1990. ‘Single
and double electron-capture by C4+ ions colliding with helium target’. J. Phys. B, At. Mol.
Phys., 23, 2811–2818.
BARTSCHAT , K.: 1998. ‘Electron-impact excitation of helium from the 1 1 S and 2 3 S states’. J.
Phys. B, At. Mol. Phys., 31, L469–L476.
B OMMELS , J., K. F RANZ, T. H. H OFFMANN, A. G OPALAN, O. Z ATSARINNY, K. BARTSCHAT,
M. W. RUF and H. H OTOP: 2005. ‘Low-lying resonances in electron-neon scattering: Measure-
ments at 4-meV resolution and comparison with theory’. Phys. Rev. A, 71.
B OSTOCK , C. J.: 2011. ‘The fully relativistic implementation of the convergent close-coupling
method’. J. Phys. B, At. Mol. Phys., 44, 083001.
B OSTOCK , C. J., D. V. F URSA and I. B RAY: 2010. ‘Relativistic convergent close-coupling
method applied to electron scattering from mercury’. Phys. Rev. A, 82, 022713.
B UCKMAN , S. J., P. H AMMOND, G. C. K ING and F. H. R EAD: 1983. ‘High-resolution electron-
impact excitation-functions of metastable states of neon, argon, krypton and xenon’. J. Phys. B,
At. Mol. Phys., 16, 4219–4236.
B UCKMAN , S. J. and J. P. S ULLIVAN: 2006. ‘Benchmark measurements and theory for
electron(positron)-molecule(atom) scattering’. Nucl. Instrum. Methods B, 247, 5–12.
C AMPBELL , E. E. B., H. S CHMIDT and I. V. H ERTEL: 1988. ‘Symmetry and angular momentum
in collisions with laser excited polarized atoms’. Adv. Chem. Phys., 72, 37.
D EPAOLA , B. D., R. M ORGENSTERN and N. A NDERSEN: 2008. ‘Motrims: magneto-optical trap
recoil ion momentum spectroscopy’. In: ‘Advances in Atomic, Molecular, and Optical Physic’,
vol. 55, 139–189. Amsterdam: Elsevier.
D RESSLER , R. A. et al.: 2006. ‘The study of state-selected ion-molecule reactions using the vac-
uum ultraviolet pulsed field ionization-photoion technique’. J. Chem. Phys., 125, 132306.
D UBOIS , A., S. E. N IELSEN and J. P. H ANSEN: 1993. ‘State selectivity in H+ -Na(3s/3p) col-
lisions – Differential cross-sections, alignment and orientation effects for electron-capture’. J.
Phys. B, At. Mol. Phys., 26, 705–721.
F RANCK , J. and G. H ERTZ: 1914. ‘Über Zusammenstöße zwischen Elektronen und Molekülen
des Quecksilberdampfes und die Ionisierungsspannung desselben.’ Verh. Dtsch. Phys. Ges., 16,
457–467.
F RANCK , J. and G. L. H ERTZ: 1925. ‘The N OBEL prize in physics: for their discovery of the laws
governing the impact of an electron upon an atom’, Stockholm. http://nobelprize.org/nobel_
prizes/physics/laureates/1925/.
F URSA , D. V. and I. B RAY: 1995. ‘Calculation of electron-helium scattering’. Phys. Rev. A, 52,
1279–1297.
512 7 Inelastic Collisions – A First Overview
Overview
In Sect. 8.1.1 we develop – on a somewhat abstract level – the general formal-
ism of close-coupling theory (CC). We then return in Sect. 8.2 once again to
B ORN approximation as the most simple theoretical approach to electron im-
pact excitation. We present – complementary to optical excitation – the con-
cept of the generalized oscillator strength for e− -atom collisions. Section 8.4
treats electron impact ionization, beginning with integral cross sections which
are of particular importance for practical applications. Singly and doubly dif-
ferential cross sections follow, while finally, triply differential cross sections
contain the maximum information about any (e, 2e) process. This is further
elaborated in Sect. 8.4.6 with a brief excursion into (e, 2e) spectroscopy,
which may be understood as complementary to photoelectron spectroscopy
(see Sect. 5.8) in the VUV and XUV spectral region. Finally, in Sect. 8.5 we
discuss an example for electron ion recombination – the inverse process to
photoionization.
Readers who are not interested in formal details may well skip this, somewhat am-
bitious section without concern for understanding the following text.
was first introduced by P ERCIVAL and S EATON (1957) (rather comprehensive re-
views on the present state-of-the-art are B URKE 2006; B URKE et al. 2007). Elec-
tron impact differs fundamentally from heavy particle collisions in that the scattered
electron has velocities which are comparable to those of the electrons inside the tar-
get. In contrast, the adiabatic potential approach (used throughout the last chapter) is
based on the assumption that the collision proceeds slowly in respect of the internal
electron dynamics. Thus, for electrons it may only be used at very low kinetic ener-
gies T (if at all). As a general rule the diabatic representation (7.48) is the method
of choice for the multichannel electron scattering problem (7.19). This leads to a
system of ODEs of the type (7.52) which has to be solved.
An additional complication arises from the fact that the scattered electron, let its
position coordinate be R, has an electron spin. Thus, in full analogy to the com-
putation of bound atomic states in Sect. 10.2, Vol. 1, the total wave function (7.23)
ms
must be anti-symmetrized. With the spin function χ1/2 the asymptotic wave function
(7.24) is then written as:
msi
eikf R msf
Ψ (R, r) eik i R χ1/2 φi (r) + ff i (θ, ϕ) χ φf (r). (8.2)
R 1/2
f
In the framework of RUSSEL -S AUNDERS coupling (i.e. for light atoms) a set of
(good) quantum numbers
Γ ≡ LM L SM S Ptot (8.3)
for the total system scattered electron and target remains conserved through the col-
lision. Here L and S are the quantum numbers for total orbital angular momentum
and total spin of the system, respectively, while ML and MS are their components
with regard to a given z-axis, and Ptot is the parity of the system. The states |Γ
of the whole collision system are constructed from the states |Lj MLj Sj MSj of
the target atom1 and the scattered electron |j mj s = 12 msj – in the usual notation
for the total orbital angular momenta of the target atom and of the scattered elec-
tron, respectively – the additional index j refers to the different channels which are
considered. The standard rules for angular momentum coupling apply (using the
relevant C LEBSCH -G ORDAN coefficients).
The partial wave expansion is now somewhat more complicated than (6.82). In-
stead of the radial equation (6.80), due to the necessary antisymmetrization of the
overall electron wave function one obtains a system of coupled integro-differential
1 This implies already recoupling from the standard (LS)J M coupling scheme for light atoms; see
J
also footnote 3.
8.1 Formal Scattering Theory and Applications 517
equations for the radial wave functions uΓf i (r) of the scattered electron:
d2 f (f + 1)
− + k 2 Γ
f uf i (R) (8.4)
dR 2 R2
∞
2μ Γ
= 2 Vfj (R)uΓji (R) + Γ
Kfj R, R uΓji R dR .
0
j
for a transition from the initial state |i = |γi Li MLi Si MSi msi into the final state
|f = |γf Lf MLf Sf MSf msf . The projectile electron is characterized by i mi and
f mf (orbital angular momenta) and the spin orientation quantum numbers msi
and msf , each prior and after the collision, respectively.
Even if (8.5) looks somewhat complicated, it just introduces into (7.32) the cou-
pling of the orbital angular momenta (Li,f and i,f ) as well as for the spins (Si,f and
1/2) of atomic and scattered electron respectively. Each again prior and after the col-
lision, i and f , respectively. They form the overall total orbital angular momentum
L and the overall total spin S. Both are conserved during the collision. The T-matrix
elements TΓf i = SΓf i − δf i describe the transition amplitudes between different
partial waves i and f of the scattered electron for each set of conserved quan-
tum numbers Γ . In a partial wave expansion for the elastic case (without electron
exchange), as we have discussed it in Sect. 6.4.6, each orbital angular momentum
corresponded to just a single value of Γ = , and TΓf i became diagonal according
to (6.112).
518 8 Electron Impact Excitation and Ionization
The direction of the k i vector of the incoming electron in this notation is given
by θi , ϕi with respect to the z-axis. In the standard collision frame, as sketched in
Fig. 7.16, the reference axis z points into the direction k i of the incoming electron,
while x lies in the scattering plane and points to the side into which the electron is
scattered.2 In this coordinate frame θi = ϕi = ϕ ≡ 0 and the projectile electron is
characterized prior √ to collision by mi = 0 so that (8.5) is further simplified with
Y∗i mi (θi , ϕi ) → (2i + 1)/(4π). Consequently, in this frame ML = MLi so that
mf = MLi − MLf and Yf −mf (θ, 0) = (−1)mf Yf mf (θ, 0). Thus, with angular
momentum and parity conservation we obtain from (8.5) an important symmetry
relation for the scattering amplitude:
The differential cross sections are obtained again from (7.33). Note, however,
that now the azimuthal angle ϕ enters explicitly into the scattering amplitude, and
thus conservation of the scattering plane is no longer compulsive, in contrast to what
has been found in the classical treatment of elastic scattering. Whenever the target
is prepared or detected in a non-isotropic state, the azimuthal dependence of the
cross section is relevant. To measure it requires, however, additional selectivity of
the experiment.
The individual spin quantum numbers of atom and scattered electron (Sj MSj and
2 sj , with j = i, f ) may well change during the collision. If spin orbit coupling
1
m
may be neglected – which is a good approximation for all light atoms – the overall
total spin is conserved during collision – with respect to magnitude and direction
(S and MS ). However, the components MSi,f and msi,f may well change due to
exchange – as we have seen in Sect. 7.2 for excitation of a triplet state out of a
singlet, or when the outer electron is exchanged during electron impact with a 2 S1/2
atom.3
One may also study changes of spin direction of the scattered electron directly –
in experiments with spin polarized electrons. Such experiments are an interesting
and active field of atomic collision physics (the interested reader is referred to
K ESSLER 1985; A NDERSEN et al. 1997; A NDERSEN and BARTSCHAT 2003).
2 Alternatively, many good reasons speak for a z-axis perpendicular to the scattering plane, as we
Here we can only sketch some basics, exemplified for the still relatively simple
case of electron scattering by a (quasi) one electron system with one active elec-
tron (Si,f ≡ 1/2). We focus on the relation between the experimentally accessible
exchange cross section Iex (θ ) and the scattering amplitudes (8.5). In this case, the
C LEBSCH -G ORDAN coefficients describing the spin coupling of atomic and scat-
tered electron, may be pulled out of the summation, and (8.5) is split into a singlet
(S = 0) and a triplet amplitude (S = 1), f 0 and f 1 , respectively:
" #" #
0 1 1 1 1
msf msi
ff i = ff i MSi msf 00 MSi msi 00 (8.7)
2 2 2 2
" #" #
1
1 1 1 1
+ ff1i MSf msf 1MS MSi msi 1MS .
2 2 2 2
MS =−1
This expression allows one to compare the results of scattering calculations with
quantities that can be measured experimentally. One may, e.g. determine the prob-
ability for changing the spin orientation (spin flip) of the scattered electron in a
collision with an unpolarized target. Alternatively to the singlet (f 0 ) and triplet
(f 1 ) scattering amplitudes one often defines a direct scattering amplitude (f ) and
an exchange amplitude (g). For a quasi one electron system the different possible
processes and the respective amplitudes are summarized in Table 8.1. These re-
lations follow directly by entering the respective C LEBSCH -G ORDAN coefficients
into (8.7).
As long as the spin may only change due to electron exchange, these relations
are still rather straight forward. However, for electron scattering by heavy atoms the
situation gets much more complex: We have seen this in Sect. 7.2 already for the
example of mercury. In the radial equation (8.4) one has to include then spin-orbit
interaction terms, i.e. the interaction potential now explicitly depends on MS and
ML , and also L and S are no longer independent good quantum numbers. The spin
of the electron may now change by flipping (and not only by electron exchange).
Correspondingly the T-matrix will depend on further parameters and the number
of independent scattering amplitudes emerging from (8.5) increases dramatically.
Scattering amplitudes may in principle obtain a left-right asymmetry, i.e. depending
on spin orientation the amplitudes can differ for ϕ = 0 and ϕ = π .
520 8 Electron Impact Excitation and Ionization
(a) (b)
exact target spectrum approximated spectrum
continuum states
pseudo states
ionization
threshold
all bound states
finite nuber of
bound states
Fig. 8.1 (a) Exact spectrum of an atom or ion and (b) its representation by pseudo-states according
to B URKE (2006)
equations (8.4) in a suitable approximation for the boundary conditions (8.2). To-
wards this aim one writes the wave function (7.23) of the system scattered electron
with atom, molecule or ion typically (see e.g. BARTSCHAT 1998) as
ΨkΓ (1 . . . N , N + 1) (8.8)
= Â aijΓ k ψjΓ (1 . . . N )R −1 uij (R) + bjΓk χjΓ (1 . . . N + 1),
ij j
electron which are not much larger than the excitation energy of the channels en-
closed. More sophisticated calculations include in additions the other states by so
called pseudo-states, whose characters are supposed to represent as many bound
and unbound states as possible, as Fig. 8.1 suggests. By choosing the pseudo-states
wisely one may describe, in addition to inelastic processes, also collisional ioniza-
tion and photoionization quite well – for not too high energies. Depending on the
method used to select the pseudo-states and on the method of integration, one speaks
of pseudo-state method, of convergent close-coupling calculations (CCC) or of R-
matrix theory. The latter enjoys continuously increasing popularity during the past
years as a very efficient tool for the successful computation of a variety of processes.
In this approach one divides the whole space into two (more precisely into three)
regions. The inner region rn , R ≤ a is selected such that only here electron ex-
change must be accounted for. At this ‘boundary’ the target wave functions have
typically decreased to about 0.1 % of their maximum value. The radial equations of
the projectile are solved such that they form a basis which is (as good as possible)
orthogonal to the target states. These solutions must obey the R-matrix boundary
conditions:
a duij
uij (0) = 0 and = b. (8.9)
u (a) dR
ij R=a
Thus, one uses a constant b for the logarithmic derivation of the radial function
on the boundary surface which can be freely chosen (e.g. =0). Inside this boundary
the computation has to be performed only once. From this follows the R-matrix,
which connects the projectile function inside and outside. Even though outside this
boundary one has to solve the radial equations for each( energy of interest, one has to
deal here only with long range potentials of the type ∝ Cs R −s ; electron exchange
is no longer relevant. Finally, from these solutions one obtains the asymptotic be-
haviour (third region) uij (R)|R→∞ which leads to the K-, T- and S-matrix.
Another approximative method to account for the continuum – applicable in par-
ticular for high collision energies – is the introduction of local or non-local polar-
ization potentials. Finally, the so called distorted-wave (DW) approximation derives
the T-matrix from the close-coupling equations by using a simple first order approx-
imations for the scattering wave. This may be seen as intelligent expansion (DWB)
of the first order B ORN approximation, applied to inelastic electron scattering. The
latter will be described in the next section.
Several examples for the ability of these modern methods of scattering theory
have already been presented at the beginning of Chap. 7. A wealth of data are found
in the literature (see e.g. B UCKMAN and S ULLIVAN 2006). We want to end this
excursion with two further ‘benchmark’ type systems.
A still relatively simple case with a quasi one electron target is electron impact
excitation of Na atoms as shown in Fig. 8.2. The resonance doublet 3p 2 P1/2,3/2 is
excited out of the 3s 2 S1/2 ground state. For several kinetic energies T of the excit-
ing electron and for a wide range of scattering angles (0◦ to 140◦ ) we may compare
here the experimental data for the DCS from different sources directly with theory.
The grey dotted line is a CC calculation, into which only the 3 2 S, 3 2 P and 3 2 D
8.1 Formal Scattering Theory and Applications 523
I (θ) / Å2 sr -1
according to B RAY et al. 1
(1991). For several initial
kinetic energies T 10 -2 3CC 3CCO
experimental data from 10 4
different authors ( ) are T = 10 eV T = 100 eV
10 2
DCS
compared with
close-coupling calculations 1
(3CC · · · ), also including a
non-local polarization 10 -2
potential (3CCO, ) 0° 45° 90° 135° 0° 45° 90° 135° 180°
θ
states enter as H ARTREE -F OCK target orbitals. The red line accounts generally also
for all other states (including the continuum) by a non-local polarization potential:
one recognizes the improvement. Nevertheless significant differences from the ex-
perimental data remain. Considering the fact, that the differential cross sections are
shown here on a logarithmic scale, it becomes obvious that, even for this seemingly
rather simple case of e− + Na scattering, significant improvement is still desirable.
In view of the different experimental results (the error bars are typically given much
smaller than the differences between different authors) we tend here to trust the
theoretical results.
Figure 8.3 documents that this kind of calculations can also provide reliable re-
sults for electron molecule collisions, here for the example of elastic and inelastic
e− + N2 scattering. Shown is a blow up for a small, particularly interesting section
of the excitation function which we have already introduced in Fig. 7.13. We had
already mentioned in Sect. 3.6.9 that a short-lived anion N− 2 exists (see Fig. 3.43).
Figure 8.3(a) and (b) provides the proof for this by a very impressive resonance
structure: each time when the electron energy corresponds to the energy of a vibra-
tional state of this instable anion, one observes pronounced, albeit broad maxima.
The FWHM of these structures give an indication of the lifetimes of these states,
which amounts to some fs. The scattering process may schematically be described
as follows:
2 1 +
e− + N2 X 1 Σg+ v = 0 → N− −
2 X Πg v → e + N 2 X Σ g v .
(8.10)
1
0.8
0
1 2 3 4 0° 45° 90° 135° 180°
DCS
1.0
(b) (d)
0.20 v=0→1 0.8
0.6 v=0→1
0.10 0.4
0.2
0.00 0
1 2 3 4 0° 45° 90° 135° 180°
electron kinetic energy T / eV scattering angle θ
Fig. 8.3 DCS for the elastic (a, c) and inelastic (b, d) electron scattering by N2 in the energetic
region of the N−2 resonance according to S UN et al. (1995). (a, b) energy dependence of the cross
sections at θ = 60◦ , (c, d) I (θ) as a function of the scattering angle indicated at the energy indicated
in (a, b) by arrows. Experimental data ( ) are compared with CCC calculations — including 15
vibrational channels
Section summary
• The theoretical treatment of low energy elastic and inelastic electron scatter-
ing requires a partial wave expansion of the radial wave functions in terms of a
suitable set of good quantum number (8.3) for the total system (scattered elec-
4 More recent experimental and theoretical data (see e.g. T ELEGA and G IANTURCO 2006, and
references given there) show an even slightly better agreement. For the sake of clarity we omit
here a comparison.
8.2 B ORN Approximation for Inelastic Collisions 525
tron plus target atom or molecule). This leads to a system of coupled integro-
differential equations (8.4). A judicious choice of states or pseudo-states to be
included in these so called closed coupling (CC) equations is the key to their
successful solution, predicting differential cross sections and various, even
more specific observables.
• We have illustrated what today may be achieved with such calculations by
comparing experimental and theoretical differential cross sections and their
energy dependence for the examples e− + Na and N2 . The agreement is im-
pressive, including in the latter case pronounced low energy scattering reso-
nances.
In Sect. 6.6 we have introduced first B ORN approximation (FBA) for elastic elec-
tron atom scattering. It can be expanded in straight forward manner to inelastic
channels.5 FBA gives a quick, first overview onto the orders of magnitude of in-
elastic cross sections, while for higher energies it often provides even rather reliable
data: In a classical picture high energies correspond to a short interaction time dur-
ing which the state of the target cannot change very much. In that case one may also
neglect exchange processes.
Starting point is the fully diabatic expansion of the S CHRÖDINGER equation
(7.52) – the close-coupling equations without exchange – which we write
2 2M̄
∇ + kj2 ψj (R) = 2 Vjj (R) + Wj δjj ψj (R), (8.11)
j
B ORN approximation assumes as 0th order a plane wave in the initial channel |i
and zero for all other channels |j :
(k)
ψi (R) = ψi (R) = eiki R and ψj (R) ≡ 0 for all j = 0. (8.12)
Inserting this into the right hand side of (8.11) only one term j = j remains and
one obtains for each final, inelastic channel |f
2 2μ
∇ + kf2 ψj (R) = 2 Vf i (R)eik i R . (8.13)
5 For a detailed description we refer to the excellent review by I NOKUTI (1971), which even today
after more than 40 years has not lost any of its validity and generality.
526 8 Electron Impact Excitation and Ionization
Using G REEN’s function one follows essentially the same procedure as outlined for
the elastic case by (6.64)–(6.68). Substituting in (6.68) the plane wave exp(ik i · R)
(+)
for ψi (R) we obtain, in analogy to (6.136), the inelastic scattering amplitude in
FBA for exciting the state |f out of the state |i
with the reduced mass M̄ of the system and the transition matrix element
Tf i = k f |Vf i R |k i
(FBA)
= e−ik f R Vf i R eiki R d3 R = eiKR Vf i R d3 R . (8.15)
As in the elastic case, B ORN approximation has again axial symmetry – in contrast
to the general solution, for which the wave vectors of the electron before and after
collision, k i and k f , respectively, define a symmetry plane. Thus, the inelastic scat-
tering amplitude in FBA is again characterized by the momentum transfer K from
the scattered electron to the target, which depends on the scattering angle θ :
K = ki − kf with K = kf2 + ki2 − 2kf ki cos θ , (8.16)
2μ 2μ
and kf2 − ki2 = (Wf − Wi ) = 2 Wf i . (8.17)
2
Obviously, (8.15) and (8.16) are only slightly more complicated than (6.140) and
(6.142) for the elastic case. The potential matrix element Vf i (R) replaces now the
classical potential V (R).
We may further simplify the inelastic T-matrix element (8.15) by inserting (7.51)
with (7.50) for the interaction potential:
e2
Tf(FBA)
i = eiK·R (8.18)
4πε0
N
1 Z
× φf (r 1 . . . r N ) − φi (r 1 . . . r N ) d3 R .
|R − r n | R
n=1
Due to the orthogonality of the atomic wave functions φj the interaction Z/R with
the atomic nucleus does not contribute. For the remaining double integral we invert
according to B ETHE the sequence of integration over scattering coordinate R and
internal coordinates r i and use the so called B ETHE integral (here without proof)
eiK·r 3 4π
d r = 2. (8.19)
r K
8.2 B ORN Approximation for Inelastic Collisions 527
Inserting atomic units, e2 /(4πε0 ) = Eh a0 , into (8.18) we finally obtain the FBA
T-matrix for the transition from |i to |f as a function of momentum transfer6
This matrix element Ff i (K) is called inelastic atomic form factor. For f = i it
is identical to the atomic form factor (1.97) in Vol. 1 which we have encountered
there in the context of X-ray diffraction. Ff i (K) is dimensionless, while Tf i
(FBA)
(FBA)
and ff i have the dimension Enrg × L3 and L, respectively. As we shall see in
Sect. 8.3.2, Ff i (K)/K is directly comparable to what we have called transition
matrix element in a radiation field, specified in Appendix H.1.6, Vol. 1: its dimen-
sion is also L, and we shall see that eK = K/K effectively assumes the role of the
unit linear polarization vector elin .
(FBA)
dσf i (θ, ϕ) 1 kf (FBA) 2 4kf 1
= 2
Tf i (k f , k i ) = 4
Ff i (K)2 . (8.22)
dΩR (2π) ki ki K
6 For larger atoms and molecules with closed shells, the summation in the matrix element
(
φf | n eiK·r n /K|φi needs to be carried out only over the active electrons. The interaction with
the core potential can be summarized by an effective core potential V (core) (R), which in FBA drops
out due to the orthogonality of the φj .
528 8 Electron Impact Excitation and Ionization
For an isotropic target the differential cross section does not depend on ϕ and the
scattering angle θ is determined exclusively by the magnitude of the momentum
transfer K according to (8.16).
The integral inelastic cross sections σf i is obtained as usual by integration
. . . d(cos θ ). As in the elastic case (Sect. 6.6) we change variables using (8.16);
we note that dK 2 = 2KdK = ki kf (2π sin θ dθ )/π = ki kf dΩR /π and write
dσf(FBA) (θ ) 1 (FBA) 2 4π 1 2
Tf i (K) = 2 4 Ff i (K)
i
= or (8.23)
dK 2 2
4πki ki K
(FBA)
dσf i (θ ) π 1 2
= Ff i (K) . (8.24)
dW T W2
We have inserted T = ki2 /2 and the so called energy transfer WK = K 2 /2 (see foot-
note 22 in Chap. 6). When integrating over dW the integration limits change to
2
Wf i 1 Wf i Wf i
WKmax 4T 1 − and WKmin 1+ , (8.25)
2T 4 T 2T
We add here a note on a peculiarity of the differential cross section in B ORN approx-
imation according to (8.22) and (8.24). These relations are obviously the product of
(i) a pure C OULOMB part according to (6.145) and (6.146), respectively, which sim-
ply describes RUTHERFORD scattering of the projectile electron by a free electron
(q1 = q2 = 1), and (ii) the square of the absolute value of an inelastic atomic form
factor Ff 0 (K), as one realizes by comparing (8.20) with (1.97) in Vol. 1. This in-
elastic atomic form factor thus modifies the pure C OULOMB repulsion of two free
electrons corresponding to the electronic density distribution in the initial and final
state of the target. Without this complication by initial and final state one might
8.2 B ORN Approximation for Inelastic Collisions 529
BE scaled
30
CCC
20
10 R matrix
3 2D←3 2S
0
1 W th 10 100 1000
T / eV
Fig. 8.4 Integral cross section for electron impact excitation e + Na(3 2 S) → e + Na(3 2 P) and
→ e + Na(3 2 D) according to L IN and B OFFARD (2005). Compared are as a function of collision
energy T experimental data from different sources (◦, •) with CCC ( ), R-matrix theory ( ),
B ORN approximation (– – –) and BE scaled (see text) B ORN approximation (—). At high energies
the B ETHE formula (pink) is also displayed – nearly identical with FBA
even think of a method for determining the momentum distribution of the electrons
in the target. We shall, however, come back to this idea in the context of collisional
ionization in Sect. 8.4.6.
8.2.4 An Example
The B ETHE formula (8.27) and suitable modifications are used very successfully,
e.g. in radiation chemistry to determine the stopping power of electrons, protons,
and alpha particles (see e.g. B ERGER et al. 2005). Quite generally one may say
that the B ORN approximation – albeit strictly applicable only for high energies
T
Wf i = (Wf − Wi ) – often gives a reasonable idea about the general behaviour
of excitation functions, and allows for simple estimates of processes without ex-
change. If one knows the atomic or molecular states sufficiently well (today many
reliable programme packages provide such data), one may evaluate (8.22) with mod-
erate effort. For high energies, integral excitation cross sections found in this way
approach the B ETHE formula.
To obtain an impression of the effectiveness of FBA, we have again a look at
electron scattering by atomic sodium. Figure 8.4 shows the experimentally deter-
mined integral cross section for 3P ← 3S and 3D ← 3S excitation. The experimen-
tal data (somewhat older and not in full agreement with each other) are compared
with B ORN approximation, convergent close-coupling (CCC) and R-matrix calcu-
lations. As seen, FBA reflects the general trends and agrees with experimental data
for high kinetic energies where the more sophisticated theories can no longer be ap-
plied. Interestingly, the optically forbidden 3D excitation – which is about a factor
of 10 weaker – is reasonably well reproduced by FBA even close to the excitation
maximum (probably due to the fact that the perturbation is much smaller than for
530 8 Electron Impact Excitation and Ionization
3P excitation). We note that the B ETHE formula (8.27) agrees in this case very well
with FBA over the whole high energy region.
According to K IM (2007) there are good reasons to re-scale the B ORN approxi-
mation for optically allowed transitions by
(BE) T (FBA)
σf i = σ . (8.28)
T + Wf i + WI f i
Here WI is the ionization energy (= −binding energy) of the active electrons. This
so called BE scaled approximation reduces the overshooting maximum of the B ORN
approximation without changing the high energy behaviour; drawn in Fig. 8.4 as
full black line. It matches the experimental data astonishingly well. This finding has
been confirmed for a number of cases.
Section summary
• B ORN approximation (more precisely FBA, see Sect. 6.6) for elastic scat-
tering is readily extended to electron impact excitation (inelastic collisions).
Using again a plane wave as 0th order approximation for the free electron,
one obtains from the diabatic representation of the S CHRÖDINGER equation
(without exchange) the rather plausible expression (8.15) for the transition
matrix elements.
• The differential excitation cross section (8.22) for a transition f ← i in FBA
turns out to depend (apart
from a flux factor) only on the momentum transfer
K = |k f − k i | = kf2 + ki2 − 2kf ki cos θ .
• The integral cross sections in FBA represent the trends with initial kinetic en-
ergy reasonably well: they rise from threshold to a maximum at a few times
the excitation energy, and then decay slowly, following essentially the B ETHE
formula (8.27), a simplified version of FBA. While the maxima of the cross
sections are typically over-estimated for optically allowed excitations, one ob-
tains rather good agreement with experiment for weak, optically forbidden
transitions.
• Further improvement is obtained by judiciously rescaling of the B ORN cross
sections.
tion (‘white light’). This is born out in mathematical terms by defining the gener-
alized oscillator strength (GOS) – in full analogy to the optical oscillator strength
(5.27), Vol. 1. It was first introduced by B ETHE (1930) (for details see e.g. I NOKUTI
1971):
2me Wf i 2 2Wf i Ff i (K) 2
ff(GOS) (K) = Ff i (K) = (8.29)
i
2 K 2 Eh Ka0
2Wf i K 2 (FBA) 2
or = T f i (K) in a.u. (8.30)
(4π)2
– in close analogy to the T HOMAS -R EICHE -K UHN sum rule (5.28), Vol. 1.
tor eK = K/K obviously plays now the role of the polarization vector e in optical
excitation. However, in contrast to optical excitation where the higher order terms of
the expansion (H.24), Vol. 1 can be neglected as a consequence of the large optical
wave length (k · r
1), the magnitude of the momentum transfer in electron impact
excitation is of the same order of magnitude as atomic dimensions and |K · r| cannot
be neglected a priori in the series expansion (8.32).
To evaluate the matrix elements f |(K · r)n |i, the quantization axis z is con-
veniently chosen parallel to K, so that K · r = Kz = Kr cos γ .7 This leads to a
selection rule M = 0 for electron impact excitation in FBA – in full analogy to
the corresponding optical transitions with linearly polarized light. We may rewrite
(8.32):
2 2
= 2Wf i f |z|i + K 2 f |z2 |i + 2f |z|if |z3 |i · · ·
(GOS)
ff i
(opt)
= ff i − f2 K 2 + f4 K 4 + O K 6 . (8.33)
Obviously this is some kind of multipole expansion into a power series of K 2 , with
(opt)
the first term being the optical dipole oscillator strength ff i for linearly polar-
ized light with polarization vector e = eK . The second term contains a contribution
from quadrupole transition moments and a product of dipole and octupole moments,
while the forth term is mainly determined by the octupole transition moment. We
thus find a nearly perfect analogy to the electromagnetically induced excitations.
The key difference is the broad band character of the excitation already mentioned
and the fact that the momentum transfer vector K, which replaces here the wave
vector of the photon cannot be neglected. Rather, according to (8.16) it may become
substantial. Hence, with electron impact one may observe nearly arbitrary types of
transitions which may be strictly forbidden for optical excitation, such as e.g. oc-
tupole transitions (see e.g. H ERTEL and ROSS 1969).
For high collision energies T differential cross sections typically are strongly
forward peaked, as one reads from their dependence on 1/K 2 according to (8.31).
Thus, in many practical situations very few terms of the power series (8.33) are
sufficient to describe small angle scattering rather well. In practice on determines
usually the effective GOS from the measured differential cross sections by inverting
(8.31). One may extrapolate it then towards vanishing momentum transfer K and
compare with the usually well known optical oscillator strengths. Note, however,
that according to (8.25) the limiting value K = 0 is experimentally not accessible
for inelastic collision.
Electron energy loss spectroscopy (EELS) is a powerful technique to study
atoms, molecules and solid-state systems. As mentioned above, a monochromatic
electron beam may spectroscopically be used as a white light source. Electron beams
with sufficiently narrow energy spread and careful energy analysis of the scattered
electron allow one to determine the energy loss Wf i in (8.33), and hence to obtain
7γ is the polar angle of the position vector r of the target electron with respect to K – not to be
confused with the scattering angle θ contained in (8.16) defining K.
8.3 Generalized Oscillator Strength 533
(GOS)
ffi e- + Na(3s) → e- + Na(3p)
100eV 40eV
1.0
20eV
0.8 (opt)
ffi - f2K 2 +f4K 4
0.6 10eV
0.4 FBA
0.2
0
10-3 0.01 0.1 1 10
(Ka 0) 2
Fig. 8.5 Generalized oscillator strength for the e− + Na(3s) → e− + Na(3p) excitation as a func-
tion of the square of the momentum transfer, K 2 . The 3CCO differential cross sections (Fig. 8.2)
of B RAY et al. (1991) have been transformed into generalized oscillator strengths ( ) and are
compared with FBA — and the series expansion of FBA for very small K 2 - - -
spectra which are essentially equivalent to optical absorption spectra – their quan-
titative analysis being based on (8.1). The foundations for EELS have been laid by
L ASSETTRE et al. (1968) and others, and a recent overview on the present status of
experiment and theory is found in TAIOLI et al. (2010).
Even though optical spectroscopy is superior in the VIS and near UV spectral
range with respect to spectral resolution, electron spectroscopy (including EELS,
PES and AUGER electron spectroscopy) becomes compatible in the high energy
regime. In particular in the VUV, XUV and X-ray region where laser sources are
not readily available, EELS offers a relatively inexpensive alternative to synchrotron
radiation. One interesting feature is its sensitivity for optically forbidden transitions:
while their cross sections disappear for K 2 → 0, they may become quite significant
for larger momentum transfer. From a careful study of the generalized oscillator
strengths as a function of K 2 one may directly infer the type of transition stud-
ied.
3CCO calculations really are for very small K 2 . It may well be that in the limit
K → 0 the numerical uncertainties of the Na wave functions used distort the results
somewhat.)
The computation of integral inelastic cross section too, may now be pushed one step
further. If one inserts into (8.26) the generalized oscillator strengths according to
(8.29), one obtains
(FBA) πa02 Kmax a0 1 (GOS)
σf i = f d(Ka0 ). (8.34)
(Wf i /Eh )(T /Eh ) Kmin a0 Ka0 f i
(GOS)
Expanding ff 0 for small momentum transfer with (8.33) and inserting the limits
of integration according to (8.25), one obtains a specialized form of the B ETHE
formula (8.27) for practical applications:
(Bethe) πa02 (opt) T
σf i = f ln + B(T ) . (8.35)
(Wf i /Eh )(T /Eh ) f i Eh /2
Section summary
• One introduces a generalized oscillator strength (8.29) for electron impact
excitation. This concepts allows a direct comparison with optically induced
transitions.
• For not too large momentum transfer K, the GOS may be expanded into
a series of powers K 2n , the prefactors being essentially the squared dipole,
quadrupole, octupole etc. moments. Thus, for larger K i.e. scattering angles,
optically forbidden transitions may be accessed.
• Based on this concept, EELS has been developed to a powerful spectroscopic
method for atoms, molecules, clusters and solid-state materials.
One of the most challenging problems of AMO collision physics is impact ioniza-
tion of an atomic or molecular target Tg by electron impact, short (e, 2e), more
specific Tg(e, 2e)Tg+ process,8 and in some more detail:
8 We focus again on electron impact only, for clarity and also because the richest experimental and
theoretical data are available for these processes. We note, however, that fast ion collisions may be
treated in a very similar manner, as already mentioned for excitation processes. The kinematics, of
course, is quite a bit more complicated.
8.4 Electron Impact Ionization 535
(a) e- (b) kB
WA z
x - φB qI = k 0 - k B - k A
y
e- e- T θB θK
k0
θA
kA
-k
A
=k
ħω 0
Wba WI K
Fig. 8.6 (a) Energetics and (b) kinematics of the (e, 2e) process: energies and wave vectors (mo-
menta) prior (T and k i ) and after ionization for the primary (WA and k A ) as well as the secondary
electron (" and k B ) according to (8.37) and (8.39); for comparison in (a) on the left, the energetics
for a photoionization process is also indicated (photon energy ω)
comparability with photoionization, in the following we shall mostly use the letter " ≡ WB to
9 For
partitioning is one of the key parameters determined by the dynamics of the ioniza-
tion process. The sign in (8.37) indicates that we have neglected the recoil energy
which is transferred onto the target ion Tg+ – which is finite, albeit very small in
comparison to that of the electrons due to the large mass ratio MTg /me
1 between
target and electron. Still the momentum balance (8.39) can always be satisfied by
the recoil momentum q I = k i − k A − k B which is transferred to the Tg+ ion.
The total momentum transfer K = k i − k A from the projectile to the target atom
is shared between the secondary electron e− +
B and the remaining ion Tg . As shown
in Fig. 8.6, for each scattering angle θA of the electron eA the secondary electron e−
−
B
may exit the interaction region in principle at arbitrary polar and azimuthal angles
θB , ϕB .
The problem poses also a substantial challenge for the experiment. Not only the
integral ionization cross section has to be measured – which for applications is very
important – one also has to determine the probability of final angles and kinetic
energies for both electrons. Depending on the degree of detail detected, one distin-
guishes the single-differential ionization cross section (SDCS), double-differential
(DDCS) and triple (TDCS) differential cross section,
dσ d2 σ
SDCS = , DDCS = , (8.40)
dW dW dΩ
d3 σ
and TDCS = , (8.41)
dWA dΩA dΩB
The SDCS only specifies W , the energy of one scattered electron after the ion-
ization process; the missing index indicates that we do not know which of the two
electrons (with energies WA or WB = ") is detected. The DDCS specifies in addition
the scattering angle Ω of the detected electron – again without recording anything
about the second electron. Finally, for determining the TDCS one has to register the
scattered electron e− −
A in coincidence with the ejected electron eB , thereby record-
ing both scattering angles. In addition, the energy WA of one of these electrons is
determined (and thus also that of the other free electron WB = T − WI − WA ).
A full theoretical description of this complex process has to follow the trajecto-
ries of both electrons over large distances – if one takes a classical point of view.
Quantum mechanically the asymptotic wave functions of both receding electrons
must be computed in the long range C OULOMB field of the finally remaining ion.
The impact ionization problem was already attacked by theory in the first half
of the past century, notably by B ETHE (1930). But only in the its last three decades
a comprehensive picture has emerged, based on concentrated efforts with highly
8.4 Electron Impact Ionization 537
In a first step we want to obtain an overview about integral cross sections for impact
ionization – integrated over all scattering angles and energy distributions of the pri-
(ion)
mary and secondary electron. They are called total, σtot , if they are summarized
also over all electrons of the target (which is usually the case if no further selection
occurs). For modelling plasmas of various kinds, these very total cross sections as a
function of initial kinetic energy are of utmost importance. It is important to know
them for many atomic and ionic species. Well kept and documented data collections
are found e.g. at NIFS and ORNL (2007). To use them efficiently (e.g. for plasma
modelling) in the past years various quasi-empirical relations have been developed.
As a rule, starting point is the B ETHE formula (8.35) for the Nj electrons of all
atomic or ionic subshells j of a target with the respective ionization potentials WIj ,
which replace the excitation energies Wf i in (8.35):
*
(ion) =0 for T < WIj
σj ln(T /W ) (8.43)
∝ WIj TIj for T ≥ WIj .
We recall here that the B ETHE formula, which in turn is based on the B ORN approx-
imation,is distinctively a high energy approximation. If one wants to improve it for
low energies T one has to add further empirical parameters fitted onto the available
experimental material. A generally used standard is still the formula, developed by
L OTZ (1967, 1968, 1970):
(ion)
N
ln(T /WIj )
σtot = a j Nj 1 − bi exp −ci (T /WIj − 1) . (8.44)
T WIj
j =1
Corresponding to (8.43) each term j contributes only for T ≥ WIj . One easily veri-
fies that the L OTZ formula reproduces a linear behaviour ∝ (T − WIj ) in the thresh-
old region (T WIj ), while at higher energies it approaches the B ETHE formula
(8.43). The parameters aj are typically between 2.6 and 4.5 × 10−14 cm2 eV2 and
are for each atom and each shell individual constants (the same holds for bj and cj ).
The parameter values, tabulated by L OTZ for many atoms and ions, are however
538 8 Electron Impact Excitation and Ionization
threshold region
1s: a = 3.9 3s: 2a = 7.2
5 20 b = 0.69
b = 0.56
4 c = 0.45 c = 0.0
σ tot / 10-17 cm2
15
WI =13.6 eV WI = 29.2 eV
medium WT
3 3p: 6a = 28
10
b = 0.61
low WT
2
5 c = 0.16
1 high
(ion)
WI =15.76 eV
WT > 10 WI
0 0
10 100 1000 10 100 1000
13.6 eV 15.8 eV
electron energy T / eV
Fig. 8.7 Total ionization cross section as a function of kinetic energy T . (a) e− + H: experimental
data ( ) from S HAH et al. (1987) and CCC calculations ( ) according to BARTSCHAT and
B RAY (1996). (b) e− + Ar: experimental data from K RISHNAKUMAR and S RIVASTAVA (1988)
( ) and S OROKIN et al. (2000) ( ), respectively, as well as SCOP calculations ( ) according to
V INODKUMAR et al. (2007). The experimental data are approximated by the L OTZ formula (—–)
with fit parameters a, b, c for the contributing shells as noted in the legend
based on older experiments and have to be refitted if necessary to the most recent
experimental data. We show in Fig. 8.7 as examples the total ionization cross sec-
tions for atomic hydrogen and argon together with more recent measurements and
calculations which we have fitted by the L OTZ formula.11
The typical dependence of ionization cross sections on kinetic energy T is rather
similar to that for electron impact excitation, except for the fact that in the thresh-
old region it depends nearly linearly on energy, cf. (7.16) and (7.18). Beyond the
maximum (at typically 4 to 5 times threshold energy) the cross sections fall essen-
tially ∝ ln(T /WIj )/T as predicted by the B ETHE formula. The agreement between
theory, experiment and L OTZ formula is very good for the H atom.
Helium is considered a benchmark case for (e, 2e) processes. With WI = 24.6 eV
the integral ionization cross section peaks at ca. 120 eV (not shown here). Very
accurate CCC(469) calculations (supported by several sets of experimental data)
are believed to be correct to within a few percent (B RAY and F URSA 2011). We
shall refer to He again in the context of SDCS, DDCS, and TDCS.
For argon some further clarification is still needed: while measurements and cal-
culations agree rather well above ca. 200 eV, at low energies the experimental data
are higher than the optimal fit by the L OTZ formula, while a quite recent theory of
V INODKUMAR et al. (2007) with a spherical, complex optical potential, SCOP, are
too low. Obviously the ionization probability for the weaker bound 3p 6 electrons
(WI = 15.8 eV) is underestimated.
11 These experimental data do, however, not allow to test the theoretical prediction for the threshold
law ∝ (T − WI )1.127 , which anyhow holds only in a very narrow energy range above threshold (see
Sect. 8.4.3).
8.4 Electron Impact Ionization 539
The next question concerns the energy balance of the ionization process (8.36): how
is the excess energy T − WI = WA + WB shared among primary and secondary
electron? One electron is detected after ionization and its kinetic energy W is mea-
sured – which is either WA or WB = ". According to M OTT (see e.g. RUDD 1991;
K IM 1975a), a first understanding may be gleaned from the RUTHERFORD cross
section (6.146) for the interaction of projectile and ejected electron (qA = qB = 1).
What is there called energy transfer WK corresponds here to W + WI . Thus, the
SDCS will be ∝ (W + WI )−2 . However, since both free electrons are in principle
indistinguishable this relation has to be symmetrized and one derives with (8.38)
from (6.146) the so called M OTT cross section:
dσ (M) πa02 Eh2 1 1 1
= + − . (8.45)
dW T (W + WI )2 (T − W )2 (W + WI )(T − W )
The last term accounts for interference between direct (RUTHERFORD) and ex-
change amplitude. It turns out that already this somewhat hand waving guess ex-
plains the experimentally observed trends qualitatively rather well.
RUDD (1991) has modified the M OTT formula and parameterized it in a semi-
empirical manner, so that the behaviour of the (angle integrated) single-differential
cross sections (SDCS) for electron impact ionization of H and He is well described
at higher energies:
dσ (R) S
= F (t)f1 (w, t) (8.46)
dw WI / eV
2
2 Eh A1 ln t + A2 + A3 /t
with S = qI π(a0 ) , F (t) = and
WI t
1 1 1
f1 (w, t) = + − .
(w + 1)n (t − w)n (w + 1)n/2 (t − w)n/2
Reduced energies w = W/WI and t = T /WI have been introduced here. For the
e− +He →2e− +He+ RUDD (1991) uses the values n = 2.4, A1 = 0.85, A2 = 0.36,
and A3 = −0.1. As documented by Fig. 8.8(d) this formula fits nicely the somewhat
older data of O DA (1975) for the SDCS at 500 eV. Using the same parametrization,
(8.46) agrees remarkably well with the more recent experimental data from S CHOW
et al. (2005) close to the threshold region and above. The CCC calculations show
significantly more variation than the simple M OTT formula (8.45); nevertheless, the
experimental errors do not permit a clear decision about their validity.
Characteristic for these energy distributions is the symmetry with respect to W =
(T − WI )/2. For very low excess energies (T − WI )
WI the distribution is nearly
constant, while at very high energies the distributions become essentially bimodal:
for T
WI one may indeed speak about a primary electron (with nearly maximum
energy) and one (very low) energy secondary electron. More recent experimental
540 8 Electron Impact Excitation and Ionization
5×10-2 5×10-2
(a) 26.3 eV (c) 40.7 eV
1×10-2 1×10-2
0.0 0.02 0.04 0.06 0.0 0.2 0.4 0.6
SDCS / Å2 eV -1
0.0691 0.6545
5×10-2 (b) 10-1 (d)
28.3 eV 500 eV
10-2
10-3
1×10-2 10-4
0.0 0.05 0.10 0.15 0 5 10 15
19.32
W / WI
Fig. 8.8 dσ/dw (SDCS) for the He(e, 2e)He+ process (log-lin scales) as a function of reduced
electron energy w = W/WI after ionization. Experimental data (◦) and CCC calculations ( )
from S CHOW et al. (2005) at low primary kinetic energy and at 500 eV from O DA (1975) are
compared with the modified M OTT /RUDD formula (8.46) ( )
data confirm this behaviour also for intermediate initial kinetic energies (see e.g.
B RAY et al. 2003).
Recent measurements for other atoms at intermediate energies indicate a some-
what more complex behaviour (YATES and K HAKOO 2011). Advanced theoreti-
cal efforts by Z ATSARINNY and BARTSCHAT (2012) show even for He significant
structures and good agreement with experimental data as illustrated in Fig. 8.9. The
RUDD formula is still surprisingly accurate – but of course it does not reproduce the
wiggly structure of the sophisticated BSR theory.
Hyperspherical Coordinates
The Hamiltonian for the fundamental three body C OULOMB problem, i.e. for
H(e, 2e)H+ but also for H− and H+
2 , is given in a.u. by
= − 1 − 2 − 1 − 1 + 1 .
H (8.47)
2 2 r1 r2 r12
8.4 Electron Impact Ionization 541
W / eV
0 10 20 30 40 50 60 70
3
experiment
SDCS / 10-18 cm 2 eV - 1
BSR theory
2 Rudd formula
0
0 0.5 1 1.5 2 2.5 3
W / WI
Fig. 8.9 dσ/dw (SDCS) for the He(e, 2e)He+ process (linear scale) as a function of reduced elec-
tron energy w = W/WI after ionization at 100 eV; experimental data () from M ÜLLER -F IEDLER
et al. (1986) and state-of-the-art BSR calculations (—) from Z ATSARINNY and BARTSCHAT
(2012), Fig. 2. Comparison with the RUDD formula ( ) shows a surprising agreement – only
the finer details (wiggles) are missing
According to M ACEK (1967) one may express this with advantage in hyperspherical
coordinates, R, α and θ12 :
R = r12 + r22 with 0 ≤ R ≤ +∞
cos α = r1 /R, sin α = r2 /R with 0 ≤ α ≤ π/2 (8.48)
r1 · r2
cos θ12 = with 0 ≤ θ12 ≤ π.
r1 r2
In these coordinates the S CHRÖDINGER equation is written as
1 d2 Λ2 + 15/4 ζ (α, θ12 ) 5/2
2
− 2
− + W R Ψ = 0, (8.49)
2 dR 2R R
with the so called C ASIMIR operator:
2 2
1 d ˆ 1 ˆ 2
Λ =− 2
2 2 2
sin α cos α + + . (8.50)
sin α cos2 α dα cos2 α sin2 α
Here, ˆ 1 and ˆ 2 are the usual orbital angular momentum operators for the two elec-
trons, and the potential ζ (α, θ12 )/R is given by:
R R R 1 1 1
ζ (α, θ12 ) = − − + =− − +√ . (8.51)
r1 r1 r12 cos α sin α 1 − sin 2α cos θ12
90˚ 1
60˚ 0
α 30˚
0˚ -1 cosθ12
face is instructive and leads to some kind of ‘visual’ understanding of the correlated
dynamics between the two electrons.
Figure 8.10 gives a 3D representation of ζ (α, θ12 ), i.e. the potential energy is
plotted as a function of the two correlation angles α and θ12 which describe the
position of the two electrons with respect to each other. In one further dimension
one has to imagine the decrease of the potential ∝ 1/R with the hyper-radius.
Intuitively, one visualizes quite easily that in the region of the ridge of this poten-
tial surface (α = 45◦ ), the so called WANNIER ridge (black dash-dotted in Fig. 8.10)
stable forms of motion at low energies are possible. Now, α = 45◦ implies r1 = r2
even for large R – which corresponds to ionization. This motion will tend to lead to-
wards cos θ12 = −1, to the minimum of the ridge, where both electrons recede from
the ion with θ12 = 180◦ in opposite direction. The more α deviates from α = 45◦ ,
while the three particles are still interacting strongly, the higher will be the chance
that the trajectory ‘roles’ into the valley, i.e. towards α → 90◦ or → 0◦ (in particular
this is expected for cos θ12 > 0 where the electrons are close). According to (8.48)
α = 90◦ or α = 0◦ implies that one of the two electrons does not leave the atom,
i.e. that no ionization occurs. In Sect. 8.4.5 we shall discuss that our conjectures are
actually not too bad when compared to reality.
He(1s 4 ℓ )
1s5ℓ
low energy electron signal
1s6ℓ
He+(1s) + e-
∝ SDCS at W =0
7ℓ
∝ ( T - WI ) 0.127
Fig. 8.11 Yield of slow electrons from the ionization process He(e, 2e)He+ in the threshold re-
gion (ionization potential WI ) as function of the initial kinetic energy of the ionizing electron T
as determined by C VEJANOV and R EAD (1974b). The full red line is a best fit through the
experimental data according to (8.52)
section as a function of collision energy T , the SDCS for threshold electrons was
studied. For detection they used a method which is sensitive only for electrons of
nearly vanishing energy W (emitted into all scattering angles) – one may see this as
early precursor of ZEKE spectroscopy (see Sect. 5.8.3) for electron collisions.
As we have seen in Sect. 8.4.2, for a given primary energy T close to threshold,
the SDCS, dσ/dW (integrated over all scattering angles) is practically independent
of the energy W of the outgoing electrons and depends only on T (we recall: 0 ≤
W ≤ T − WI ). Hence, the integral ionization cross section (8.42) is approximately
T −WI
dσ dσ
σ (ion)
= dW (T − WI ).
0 dW dW
One thus expects the SDCS to rise as
dσ σ (ion)
∝ (T − WI )0.127 (8.52)
dW T − WI
if the threshold law (7.18) of WANNIER and R AU holds. By an experimental deter-
mination of dσ/dW at W 0 one may thus test the deviation of the threshold law
from a slope of 1 with high sensitivity. Figure 8.11 shows for e− + He → 2e− + He+
the yield of zero kinetic energy electrons as measured by C VEJANOV and R EAD
(1974b) close to threshold, plotted as a function of the kinetic energy T of the pro-
jectile electron. The dependence ∝ (T − WI )0.127 for T above the ionization thresh-
old WI is clearly documented by the red line, fitted to the data. In this case, the
threshold law (7.18) appears to be valid in an energy range of at least 1.5 eV above
threshold.
544 8 Electron Impact Excitation and Ionization
It is interesting to note that also below threshold a similar excitation law seems
to hold (a mirror image) for high RYDBERG states. The high lying RYDBERG states
were detected in this experiment by field ionization – by a similar method as in the
(much later) ZEKE experiments (see Sect. 5.8.5). In view of the potential hypersur-
face Fig. 8.10 this trend is not too surprising: the genuine electron dynamics should
be similar slightly below and above threshold. From this viewpoint, RYDBERG ex-
citation corresponds to trajectories which just ‘do not make it’ into the continuum.
Rather, they ‘role into the valley’ at α = 0◦ or 90◦ .
The threshold region (WANNIER region) up 1.5 eV above threshold was and still
is an object of sophisticated experimental and theoretical work. Of specific interest
is the angle θ12 between both electrons. An educated glance at Fig. 8.10 suggests,
that at low energies the two electrons will preferentially separate at θ12 = 180◦ – if
ionization occurs at all. This has indeed been confirmed experimentally (not shown
here, see C VEJANOV and R EAD 1974b, Fig. 2) as well as by state-of-the-art quan-
tum mechanical calculations (see e.g. BARTLETT and S TELBOVICS 2004b). Of
course one observes a distribution of angles, the width (FWHM) of which increases
with excess energy above ionization threshold. For the system e− + H → 2e− + H+
one finds (π − θ12 )FWHM 3(T − WI )1/4 . Recent results will be discussed in
Sect. 8.4.5 (see in particular Fig. 8.21).
Wf i → " + WI = T − W (8.53)
8.4 Electron Impact Ionization 545
and obtain the DDCS per energy interval d" = dW and solid angle dΩ into which
the detected electron is scattered after ionization:
(FBA)
dσ"i(FBA) (θ, ϕ) 4a02 W 2 a02 1
W T"i (k f , k i ) 2
= F (K) =
Eh (2π)2 T
"i
d"dΩ (Ka0 )4 T 1/2
Eh a03
√
with (Ka0 )2 = 2(T + W − 2 T W cos θ )/Eh . (8.54)
√
We have expressed the flux factor kf /ki = W/T in terms of the energy W of the
detected electron and the initial kinetic energy T .
The matrix element F"i (K) is formally still given by (8.21). But it describes
now a transition of the (ejected) target electron between a bound, discrete initial
state |i and a free, continuum final state "|. The latter is normalized in energy
scale (dimension L−3/2 Enrg−1/2 ) as described in Appendix J.1 of Vol. 1, so that
F"i (K) and T"i have now the dimension Enrg−1/2 and Enrg1/2 L3 , respectively,
(FBA)
and the DDCS has the dimension L2 Enrg−1 per solid angle.
We may compare the step from bound-bound to bound-free transitions in elec-
tron impact with that from photoexcitation (Sect. 5.2.2 in Vol. 1) to photoioniza-
tion (Sect. 5.5 in Vol. 1). The energy balance (5.59), Vol. 1 for photoionization
was ω = " + WI and the OOSD df"i /d" (dimension Enrg−1 ) was defined by
(opt)
(5.62), Vol. 1. Correspondingly one extends the concept of the (generalized) oscil-
lator strength to the continuum (in the framework of FBA). With the substitution
(8.53) for Wf i we write in analogy to (8.31) for the DDCS
(FBA) (GOS)
d2 σ"i (θ, ϕ) 2a02 Eh W 1 df"i (K)
= . (8.55)
d"dΩ T − W T (Ka0 ) 2 d"
By comparison with (8.54) the generalized oscillator strength density, GOSD, in-
troduced here may be computed (in a.u.) from
(GOS)
df"i (K) Ff i (K) 2 (opt)
df"i
= 2(T − W ) −→ (8.56)
d" K K→0 d"
with the matrix element (8.21). In the limit K → 0 the GOSD approaches the op-
tical oscillator strength density – just as for the excitation process according to
(8.33). The GOSD has the dimension Enrg−1 , and as in the case of photoionization
it strongly depends on the excitation energy " + WI = T − W of the electron raised
into the continuum.
For the H atom the GOSD may be evaluated in closed form, as already done
by B ETHE. Following I NOKUTI (1971) this somewhat complicated formula, the so
called B ETHE surface, is plotted in Fig. 8.12 as a function of the excitation energy
" + WI and the logarithm of K 2 , again in a.u. In the optical limit12 K → 0 GOSD
and OOSD decay very rapidly with " + WI . Quite remarkable is the occurrence of a
12 Ina collision process the limit K → 0 may be approached arbitrarily well, but can never be
reached completely, as obvious from (8.54).
546 8 Electron Impact Excitation and Ionization
maximum for the GOSD at higher K and the shift of this maximum as the excitation
energy " + WI rises. As a closer analysis shows (see e.g. C OPLAN et al. 1994), in
the region of this so called B ETHE ridge the ionization process is dominated by
binary interaction between the scattered and the ejected electron. In this situation
the remaining ion participates in the process more or less as a spectator only.
We now illustrate these considerations for the reaction He(e, 2e)He+ – a bench-
mark system for electron impact ionization. Of course no analytical formula for the
GOSD is available here. However, a number of experimental and theoretical studies
have been performed, also for somewhat higher initial kinetic energies. Figure 8.13
shows for T = 100 eV and 500 eV the DDCS, i.e. the angular distribution of elec-
trons detected at a number of selected energies W . For He the ionization energy
is WI = 24.6 eV, thus, in Fig. 8.13, the energy " of the other, not detected elec-
tron ranges from practically 0 eV up to 20 eV (and up to 40 eV at T = 500 eV).
The strongly forward peaked distribution is quite remarkable (most dramatically at
500 eV and low energy " of the ejected electron). With increasing excitation energy
" + WI = T − W the cross section decreases (as already expected from Fig. 8.8(d),
and the angular distribution gets flatter.
The experimental data of M ÜLLER -F IEDLER et al. (1986) are compared to sev-
eral state-of-the-art computational results, including very recent BSR calculations
of Z ATSARINNY and BARTSCHAT (2012). The theoretical models reproduce the
experimentally observed trends quite well. As a comparison between the data of
M ÜLLER -F IEDLER et al. (1986) and AVALDI et al. (1987) at 500 eV shows, there
are some uncertainties also among the experiments.13
13 We must note here, that without the re-calibration (suggested quite convincingly by S AENZ et al.
1996) the agreement with the BSR calculations at 100 eV would even be better.
8.4 Electron Impact Ionization 547
DDCS / 10 -23 m 2 eV -1 sr -1
(a) T = 100eV (b) T = 500eV
10 2 10 2
W = 73.4 eV
101 W = 71.4 eV 101 W = 471.4 eV
W = 55.4 eV W = 435.4 eV
Avaldi 435.4
100 100
BSR DWB
DWB GA
10-1 GA 10-1
0° 20° 40° 60° 80° 0° 20° 40°
scattering angle θ
Fig. 8.13 Double-differential cross section (DDCS) for electron impact ionization of He at a pri-
mary energy T = 100 and 500 eV for several energies of the “scattered” electron W as a func-
tion of scattering angle. Experimental data from M ÜLLER -F IEDLER et al. (1986), one data set
from AVALDI et al. (1987). The data were re-calibrated by S AENZ et al. (1996). Theory (lines):
G LAUBER approximation, GA (R AY et al. 1991), distorted wave B ORN, DWB (M C C ARTHY and
Z HANG 1989), and BSR (Z ATSARINNY and BARTSCHAT 2012)
Even though, from a fundamental point of view, one cannot really distinguish the
scattered electron from the (secondary) ejected electron, the pronounced forward
scattering for small secondary energies " indicates that there is little exchange at
these relatively high initial kinetic energies (the energy distribution discussed in
Sect. 8.4.2 also supports this finding).
Thus, the intuitive model of a scattered electron (with little energy loss, deflected
only little) and an ejected electron (at rather low energy) is quite reasonable in this
range of primary energies. This becomes even more evident, when the angular dis-
tribution is studied for electrons with low energies as shown for a few examples in
Fig. 8.14: identifying this low energy electron as ‘ejected’ (i.e. originally bound) it
is plausible that it has no pronounced forward peak. On the contrary, for the low-
est energies of 2 eV and 4 eV we even recognize a distinctive backward tendency,
complementary to the sharp forward peak of the high energy counterpart at 75.4 eV
and 73.5 eV in Fig. 8.13(a): this appears to correspond rather plausibly with a tra-
jectory around cos θ12 = −1 on the WANNIER surface Fig. 8.10. As the detected
energy gets larger, the distribution becomes more isotropic – albeit still structured.
The agreement with theoretical models is quite satisfactory, especially so for CCC
from B RAY and F URSA (1996).
It is instructive to express the measured DDCS for the (primary) scattered elec-
tron in terms of the generalized oscillator strength density (GOSD). This is dis-
played for the He(e, 2e)He+ process in Fig. 8.15. Using (8.55) and (8.56), the ex-
perimental cross section data from M ÜLLER -F IEDLER et al. (1986) (re-calibrated
according to S AENZ et al. 1996) are thus compiled into one single plot, for six dif-
ferent initial kinetic energies T and three different values of the excitation energy
" + WI . The result is quite convincing: while there are – as expected – some de-
viations in the GOSD at larger momentum transfer Ka0 , the overall convergence
548 8 Electron Impact Excitation and Ionization
4 4 BSR
CCC
DDCS / 10 -23 m 2 eV -1 sr -1
W = 2eV W = 4eV
2 2
0 0
2 W = 10eV 2 W = 20eV
1 1
0 0
0 90º 180º 0 90º 180º
scattering angle θ
Fig. 8.14 Double-differential cross section (DDCS) for electron impact ionization of He at 100 eV
for different energies W of the detected (“secondary”) electron as a function of its scattering an-
gle θ . Note the different scales for the DDCS in the upper and lower panels. Experimental data
from M ÜLLER -F IEDLER et al. (1986) are compared with recent BSR calculations (Z ATSARINNY
and BARTSCHAT 2012) and CCC (B RAY and F URSA 1996)
one may envisage easily a more pronounced “B ETHE ridge” for higher excitation
energy, similar as in the H(e, 2e)H+ case according to Fig. 8.12.
A more differentiated picture emerges from recent experimental and theoreti-
cal work at lower kinetic energies, in particular in the threshold region for ioniza-
tion. The observations may often be explained by using the hyperspherical potential
Fig. 8.10 in a quite intuitive manner. For details the readers are referred to the orig-
inal literature (see e.g. S CHOW et al. 2005; B RAY et al. 2003).
multiplier 2
5 cm
collector B
235º
repeller
lenses 5
electron gun
lenses 2 70º
field free multiplier 1
190º
energy
selector e-
atomic
collimation beam
┴ plane
lenses 4
energy
lenses 1 cathode 125º analysor
lenses 3
collector A
Fig. 8.16 Pioneering experimental setup by E HRHARDT et al. (1969) for the determination of
triple differential cross sections (TDCS) for the He(e, 2e)He+ process with coincident detection of
the scattered and ejected electron
Both cases have been studied worldwide intensively – mostly in close collabo-
rations between theoretical and experimental groups. Today the agreement between
experiment and theory is quite impressive. Relatively clear and easy to understand
intuitively is the situation for intermediate and high energies (beyond the M ASSEY
maximum of integral cross sections) while at initial kinetic energies below that max-
imum and in particular in the threshold region a very rich and complex dynamics is
observed.
1 kA (FBA) 2
(FBA)
dσ"i (θ, ϕ) ,
= T "i (k f , k i )
d3 k B dΩA (2π)2 ki
14 Ifthe plane waves used to determine T"i(FBA) are normalized in k scale (see Appendix J.2.2 in
Vol. 1), as e.g. in OVCHINNIKOV et al. (2004), the factor (2π)−2 has to be replaced by (2π)4 .
8.4 Electron Impact Ionization 551
1 kB kA (FBA) 2 4kB kA
(FBA)
dσ"i (θ, ϕ)
= 2
T"i (K) = 4
F"i (K)2 , (8.57)
dΩA dΩB d" (2π) ki ki K
where F"i (K) is formally still given by (8.21). However, while there the integra-
tion is assumed to be over all N electrons of the target, it is now carried out only
over N − 1 electrons, while the momentum k B of the ejected electron is kept as
a parameter which determines energy and angles of the ejected electron. As in im-
pact excitation, in FBA the form factor F"i (K) – and hence T"i
(FBA)
(K) as well as
the triple differential cross – are symmetric around the momentum transfer K. Ex-
plicit expressions (available for atomic H) contain characteristic terms of the type
K · k B = K · kB cos θK where θK is the angle between the ejected electron and the
momentum transfer K. According to (8.56), in the limit of really very small mo-
mentum transfer the GOSD approaches the OOSD, and one even expects ultimately
an angular distribution of the ejected photoelectron as in the case of photoioniza-
tion:
df"
(opt)
d3 σ (θ ) 1 WA 1
−→ 1 + βP2 (cos θB ) . (8.58)
dΩA dΩB d" K→0 2π T K (WI + ") d"
2
15 In view of this definition, in the theoretical literature the TDCS is often referred to as “FDCS”
since, including the azimuthal angles, strictly speaking it refers to five variables in the denominator.
16 As illustrated in the kinematics, Fig. 8.6, we use an xz scattering plane defined by k i and k A ,
the scattering angle θA being positive by definition, and this also holds for the momentum transfer
angle 0 ≤ θK ≤ 360◦ . In the experiment discussed here the secondary electron is detected in the
scattering plane (ϕ = 0), its emission angle being 0 ≤ θB ≤ 360◦ . In the literature one finds a
variety of slightly different notations.
552 8 Electron Impact Excitation and Ionization
0° (b) 0°
(a) k
B kA k B
recoil
0°
(c) k
B
45° kA 315°
0.5
0.25
90° 270°
135° 225°
A = 14° 180°
B = 10 eV, = 204°, = 1.05 0-1
Fig. 8.17 TDCS for e− + He → 2e− + He+ at T = 250 eV. Polar diagram as a function of the
emission angle θB of the ejected electron (energy WB ), at fixed scattering angle θA of the projectile
electron and fixed momentum transfer K (its length is not drawn to scale here). The magnitude of
the TDCS is proportional the distance from origin (values at the circles are in 10−22 m2 sr−2 eV−1 ).
Experimental data from S CHLEMMER et al. (1991) and as communicated by B RAY and F URSA
(1996) who also performed the CCC(101) calculations ( ); for (a) at small momentum transfer
also FBA ( ) from B YRON et al. (1986); the experimental data in (b) where recalibrated to match
the binary peak of FBA
and we refrain from discussing a large variety of further theoretical approaches, all
reproducing the experimental results rather satisfactorily.
For the smallest momentum transfer (Fig. 8.17(a)) we may compare the data
with FBA (dashed red line), and see that even at K = 0.38/a0 we are far from the
8.4 Electron Impact Ionization 553
expected dumbbell shape in the limit K → 0. However, experiment, FBA, and the
nearly exact CCC(101) theory display a characteristic double lobe structure. The
forward lobe, pointing more or less into the direction of the momentum transfer
+K, is called binary peak and is understood to arise essentially from a binary in-
teraction between the projectile electron and the ejected atomic electron – with the
atomic nucleus remaining a spectator. In contrast, the recoil peak, emitted more or
less in the opposite direction, is considered to arise from backscattered electrons
undergoing multiple interactions with the nucleus.
It is interesting to note – and quite typical – that the FBA cross section overes-
timates the binary peak and underestimates the recoil peak. Closer inspection also
shows, that the symmetry predicted by FBA does not completely match the obser-
vations – even at very low K. Rather, the maximum of the binary peak is slightly
shifted towards smaller emission angles (typically 270◦ ≤ θB < θK ) while the re-
coil maximum is found at slightly larger angles (θK − 180◦ < θB < 180◦ ). In con-
trast to FBA, the CCC(101) results reproduce the experimental data at the smallest
scattering angle θA = 4◦ almost perfectly (Fig. 8.17(a)). At larger scattering an-
gles and higher momentum transfer, displayed in Fig. 8.17(b) and (c), the overall
cross section decreases as expected, and K still provides an approximate symmetry
axis. However, while the binary peak dominates now the process, the recoil peak
tends almost to vanish. This is a general observation at higher kinetic energies and
momentum transfer approaching that of the B ETHE ridge where ionization occurs
essentially by binary encounter of projectile and target electron.
Over the years, a variety of theoretical approaches has been tested for repro-
ducing these and other experimental data, among others the B ORN series, modified
B ORN approximations e.g. with outgoing C OULOMB waves and various other “dis-
torted wave” methods as well as the so called G LAUBER approximation (an eikonal
approximation similar to that sketched in Sect. 7.4 as semiclassical approach to
heavy particle scattering). Nearly all these approaches predict similar overall trends,
even B ORN approximation (except that in FBA K is a strict symmetry axis, in all
other theories it is not). However, it is not trivial to make exact predictions, espe-
cially for the ratio between binary and recoil peak. Only CCC(101) and similarly
expensive calculations match the presently available experimental data almost per-
fectly, as illustrated by the few examples shown in Fig. 8.17.17
Most of the earlier experiments were restricted to two particle coincidence in the
scattering plane – so some important information was missing. Recently, R EN et
al. (2011) performed an impressive 3D benchmark experiment using a state-of-the-
art reaction microscope, testing CCC and TDCC theory (at relatively low kinetic
energy, T = 70.6 eV, for several scattering angles θA and different energy sharing
between the two electrons). The momenta of all three charged particles were de-
termined and full 3D images were generated of the probability distributions for the
ejected second electron.
17 The
experimental data points in Fig. 8.17(b) have been rescaled to match the binary peak max-
imum from CCC. This appears justified as the experiment was calibrated by extrapolation to the
OOSD according to (8.58) – which is questionable at these high K values.
554 8 Electron Impact Excitation and Ionization
II recoil
Fig. 8.18 TDCS for the He(e, 2e)He+ process at T = 76 eV according to R EN et al. (2011). Plot-
ted is the cross section for a fixed scattering angle θA = 20◦ as a function of the emission angle θB ,
ϕB of an electron with energy WB = 5 eV (a) experiment with an advanced reaction microscope,
(b) TDCS calculated with state-of-the-art CCC; note that here the length of K (K = 0.878a0−1 )
is to scale with respect to k i and k A ; the maxima in experiment and CCC theory are found at
θB 286◦ (binary) and θB 151◦ (recoil) – for comparison θK = 317◦ and θK − 180◦ = 137◦
One sample from these beautiful studies is shown in Fig. 8.18.18 At this partic-
ular emission angle and energy partitioning the observed bimodal lobe still resem-
bles the p shaped dumbbell. However, kinked downward at both ends, it is far from
being symmetric with respect to the momentum transfer K. We specifically empha-
size the pronounced, non-zero waist in the perpendicular plane – in contrast to a p
dumbbell – and recall that total momentum conservation between k i and any two
finally detected electrons (with energy conserving k A , k B ) may be ensured by recoil
q I to the target ion, as sketched in Fig. 8.6. A detailed comparison of experiment
and theory by R EN et al. (2011) documents excellent agreement between CCC and
the experimental results (with very small error bars) for all geometries studied (not
shown here), while TDCC tends to overestimate the recoil peak. In general, at these
relatively low energies the recoil peak (i.e. electron interaction with the atomic core)
dominates at small momentum transfer, while the binary peak (i.e. electron-electron
interaction) dominates for larger momentum transfer. For equal energy partitioning
between the two outgoing electrons and larger momentum transfer, the recoil peak
disappears almost completely.
The e− + H System
From the above discussion, the solution of the tree body problem appears to require
‘brute force’ computational methods to obtain good agreement between theory and
experiment, in particularly so since the e− − He system is not a true three body
18 We are grateful to the authors for providing the specially adapted version of their 3D images
shown here.
8.4 Electron Impact Ionization 555
θ A= 3°
12 (a) T=250eV
W B= 5 eV
FBA K = 0.38a 0-1
8
SBA θ K = 321°
TBW
4
0.05
0
0° 90° 180° 270° 360°
θB
Fig. 8.19 TDCS for the H(e, 2e)H+ process, comparison of experiment and theory for detection
of the second electron in the scattering plane. Plotted is the cross section as a function of the emis-
sion angle θB of an electron with well defined energy WB . (a) High energy regime (T = 250 eV),
the scattered electron is detected at a fixed angle θA = 3◦ ; the experimental data from E HRHARDT
et al. (1986) (recalibrated ×0.88) are compared with FBA and SBA from DAL C APPELLO et al.
(2011), and (almost identical to SBA) with TBW from B RAUNER et al. (1989); (b) initial kinetic
energy very close to threshold (T = 17.6 eV), the excess energy of 4 eV is equally shared between
the two electrons, the angle between them is now kept fixed at θAB = 180◦ ; the experimental data
RÖDER et al. (2003) are compared with ECS (BAERTSCHY et al. 2001), CCC (B RAY 2002), and –
added here – TDCC (C OLGAN and P INDZOLA 2006) calculations; (c) same as (b) but now the
angle between the two electrons is kept at θAB = 90◦
problem: the influence of the second atomic electron certainly adds even more com-
plications. Irrespective of the excellent understanding which has been achieved for
the He(e, 2e)He+ system, the genuine three body problem, H(e, 2e)p+ , remains a
unique challenge for theory and experiment alike. Figure 8.19 compares experiment
and theory for some selected geometries at two different energies. Figure 8.19(a)
documents the characteristic quantitative failure of FBA. We note, however, that al-
ready the second term in the B ORN series, SBA, in the special variety reported by
556 8 Electron Impact Excitation and Ionization
position
sensitive exit
detectors lenses
DAL C APPELLO et al. (2011) gives excellent agreement with experiment at an inter-
mediate energy (T = 250 eV) and competes well with the calculations of B RAUNER
et al. (1989) based on an asymptotically exact form of the three-body C OULOMB
wave function.
Figure 8.19(b) and (c) document the situation for very low initial kinetic en-
ergy – just 4 eV above the ionization threshold. In this case the detection angle θAB
between the two electrons (equal final energy) has been kept constant. Obviously,
two electrons emission into opposite directions (b), θAB = 180◦ , is by a factor of
20 more probable than emission at right angles (c). This nicely verifies the classi-
cal intuition of preferred motion on the WANNIER ridge as discussed in Sect. 8.4.3.
For this geometry it is obviously most probable that one electron just follows the
direction of the projectile while the other electron is ejected in the opposite direc-
tion as displayed in Fig. 8.19(b). The (less probable) θAB = 90◦ geometry obviously
involves more complex dynamics.
For obvious reasons, there is much less experimental material available then for
He (it is just not trivial to generate a stable, well controlled beam of atomic hydro-
gen). Thus, efficient data collection is very important. An interesting alternative to
the reaction microscope is the toroidal energy analyzer shown in Fig. 8.20, devised
by VAN B OEYEN and W ILLIAMS (2005). The two final electrons are simultaneously
selected according to their energy and emission angle and detected in coincidence.
So called wedge and strips anodes are used, which allow simultaneous temporal
and spatial registration of particles (at the bottom of Fig. 8.20 shaded in light red).
In contrast to the traditional planar geometry used in the early experiments by the
E HRHARDT group in Kaiserslautern, here the perpendicular geometry is used where
scattered and ejected electron are both detected perpendicular to the electron beam.
In our terminology displayed in Fig. 8.6, this implies θA = θB = 90◦ . Both electrons
are detected in coincidence at azimuthal angles 0◦ ≤ ϕB ≤ 360◦ – just by simulta-
neous registration of the respective positions on the wedge and strips anodes behind
8.4 Electron Impact Ionization 557
20
TDCS / 10 -20 cm 2 sr -2 eV -1
T-WI = 0.5eV T -WI = 6.8eV
W B /WA=1:10 W B /WA=1:10
15 W B =WA W B =WA
10
×5
5
0
0 90° 180° 270° 360°
φB
the toroidal analyzer. Even though the TDCS is very small in this configuration, it
is particularly sensitive for critical tests.
Figure 8.21 shows the results of W ILLIAMS et al. (2006) for e− + H in the en-
ergy range just above threshold, T WI . For excess energies 0.5 eV and 6.8 eV
each, two types of energy sharing have been studied: one very asymmetric and a
fully symmetric version with WA = " = (T − WI )/2. It is interesting to note that
the TDCS changes only little with energy partitioning. However, the ϕB dependence
varies dramatically with the primary energy T . While at T − WI = 0.5 eV the be-
haviour predicted by the WANNIER model – the electrons emerge at opposite direc-
tion (ϕB = 180◦ ) is very pronounced, it has disappeared already at 6.8 eV above the
ionization threshold and yields a structured angular distribution.
The full lines in Fig. 8.21 correspond to exact quantum mechanical calcula-
tions with the so called propagating exterior complex scaling method of BARTLETT
and S TELBOVICS (2004a): on a grid of the overall size Rmax = 180a0 the
S CHRÖDINGER equation is solved numerically for partial waves up to = 5! The
fantastic agreement between this theory and experiment documents nicely the state-
of-the-art for this fundamental three body problem.
the efficient imaging techniques described above. The experiments show clearly,
that the simple double lobe structure seen in the latter case is no longer observed
for Ar(e, 2e)Ar+ . The 3D images observed experimentally and computed by BSR
theory show a rather complex structure – and do agree very well.
As yet another direction of present and future research in this field, molecules
come into view (see e.g. A L -H AGAN et al. 2010; S ENFTLEBEN et al. 2012, and
further references there). The Heidelberg group (R EN et al. 2012b) reported again
a new type of electron impact ionization study for H2 where the TDCS is even cor-
related with a specific alignment of the molecular axis – determined by registering
the momentum of a proton emitted by dissociative ionization.
Before ending this section an important spectroscopic aspect of the (e, 2e) experi-
ments has to be mentioned. It has been developed and exploited very successfully
by several groups over the past decades. While all our above discussion focussed
on the dynamics of the ionization process, we shall now show that the TDCS de-
pends directly on the momentum distribution of the target electrons. Under certain
conditions, the coincidence technique may thus be used to determine the latter for
arbitrary atomic or molecular targets.
The basic idea of this electron momentum spectroscopy (EMS) starts by assuming
that the momentum k B of the ejected (secondary) electron is determined by the sum
of the momentum transfer K = k i − k A and its own momentum q at the time of the
collision, i.e.
k B = K + q. (8.59)
The angular distribution of the ejected electron, that is the idea, should then allow
direct conclusions about the momentum density distribution of the ejected electron
prior to the collision. And the latter is the F OURIER transform of the electron density
distribution in position space. In this manner one may thus obtain a direct image of
the atomic or molecular wave function.
Let us have a look again at the kinematics of the (e, 2e) process according to
Fig. 8.6. We assume the target atom or molecule to be initially at rest (the slight
thermal motion may usually be neglected). The target as a whole has initially no
net momentum. One may recast this trivial fact into the statement, that the initial
momentum of the target electron (averaged over the whole momentum density dis-
tribution) is zero. After the ionizing collision, the target ion has in general a non-
vanishing momentum
q I = ki − kB − kA = K − kB,
so that momentum conservation is satisfied. With (8.59) this leads to q I = −q: the
electron momentum q prior to collision is thus equal in magnitude to the ion recoil
momentum q I and directed opposite to it. From this point of view the momentum
8.4 Electron Impact Ionization 559
k i = k A + k B − q. (8.60)
In contrast, the respective recoil energy q 2I /2Mion of the ion is very small and may
be neglected.
Since the measurement of the TDCS implies according to (8.39) a determination
of q I one thus determines – according to this concept – also q = −q I . However,
to really obtain a genuine momentum distribution of the target electron one has to
make sure that the process can be described as a truly binary interaction between
scattered electron and ejected target electron. Effectively, neither the scattered pro-
jectile electron nor the ejected electron must directly interact with the remaining
target ion.
At high energies, where T as well as WA and " are large compared to the binding
energy WIj of the electron studied, one may hope that this assumption is justified
under certain conditions. One uses the so called plane-wave impulse approximation
(PWIA), by representing the projectile electron before and after as a plane wave,
which leads to (8.61). In addition, one now further assumes that also the ejected
secondary electron is approximated by a plane wave. This further simplifies the
atomic (or molecular) form factor defined in (8.21) and leads to (in a.u.)
They effectively represent the electron hole created in the cationic wave function
and can be computed by standard quantum chemical programmes. The DYSON or-
bital gγ (r B ) reduces to the initial MO of the electron studied if the most simple
560 8 Electron Impact Excitation and Ionization
approximation is used: assuming that target and ionic wave functions may be fac-
torized in MOs, and that the ionic core remains unchanged when the secondary
electron emerges from the target (so called frozen-core or sudden approximation).
Using DYSON orbitals the TDCS (8.61) is in any case written simply as
d3 σ (θ ) 4 kA kB 2
= 4 gγ B (p) , (8.63)
dΩA dΩB d" K ki
where gγ B (p) is the F OURIER transform of the DYSON orbital. Thus, the TDCS is
indeed proportional to the density in momentum space of the target electron studied.
Here . . . is the overlap integral of the spectator orbitals in the ion and target,
respectively, and the last integral is just the F OURIER transform of the target orbitals
φB (q), i.e. the wave function in momentum space.
At which experimental conditions these approximations are valid and how reli-
able the density maps obtained are, has been subject to detailed experimental and
theoretical studies in the 1990ies. We cannot discuss the details and refer the inter-
ested reader again to the relevant literature (e.g. W EIGOLD and M C C ARTHY 1999;
C OPLAN et al. 1994; M C C ARTHY and W EIGOLD 1991). In the meantime (e, 2e)
EMS has become a well established, powerful method in molecular spectroscopy –
often called ‘synchrotron of the poor’ as one may obtain similar (typically more ex-
pensive) information from PES using XUV and X-ray radiation from synchrotron
sources.
High electron kinetic energies (T > 1 keV) are required for a clean interpreta-
tion of the data. One typically does not use the coplanar setup, which has provided
most information about the dynamics. Rather, one often uses symmetric geometries
(θA = θB and WA = "), keeps the momentum transfer K constant, and varies in-
stead the azimuthal angle ϕB and thus q. Since the experimental signals are low
in this geometry, one uses today parallelized measuring techniques with position
sensitive detectors for registering the electron coincidence, and electron energy an-
alyzers of the type shown in Fig. 8.20. This allows to collect data with sufficiently
good statistics in acceptable measuring time. In the past years impressive data have
been collected, determining experimentally the momentum density distributions for
a variety of atoms and molecules.
We illustrate this by a more recent experiment with the H2 O molecule from N ING
et al. (2008). The kinetic energy was T = 1200 eV, with good energy and angular
resolution (W = 0.68 eV, ϕB = ±0.84◦ , θ = ±0.53◦ ) under fully symmetric
conditions with WA = " = (T − WIj )/2 and θA = θB = 45◦ . Here WIj is again the
ionization potential (= −binding energy) of the orbitals studied. The coincidence
rate, that is the TDCS, is measured as a function of the azimuthal angle ϕB with
a position sensitive detector behind a double toroidal analyzer through which both
electrons pass. From the energy balance (8.36) the detected signal can be attributed
uniquely to a certain binding energy, −WIj , which allows unique identification of
the orbital from which the detected electron originates. In this geometry the recoil
momentum q I = −q is a unique function of the azimuthal angle ϕB . Thus, assuming
the PWIA to be valid (with frozen ion core), a direct determination of the momen-
tum density distribution prior to the ionization process is possible for all orbitals
8.4 Electron Impact Ionization 561
studied. To obtain a visual image of this momentum distribution (and thus of the
wave function in the different orbitals) one plots the TDCS in a 2D presentation as
a function of the binding energy and the azimuthal angle.
For the valence electrons of H2 O this is displayed in Fig. 8.22(a). The projection
of these density distributions onto the axis of the binding energies Fig. 8.22(b) repro-
duces nicely the spectra known from photoelectron spectroscopy (see Fig. 5.45(a)).
One recognizes very clearly in Fig. 8.22(a) the different symmetries of the or-
bitals which are directly reflected in the momentum density distributions. We recall
the discussion of the H2 O orbitals in Chap. 4. As schematically sketched in Fig. 4.22
we verify that 1b1 , 3a1 and 1b2 each have one nodal plane, while 2a1 does not have
one. This is directly mimicked in the momentum density distribution which is the
F OURIER transform of the spatial distribution. Nodes in the wave functions are re-
flected as minima in Fig. 8.22(a). In contrast, the 2a1 distribution shows no such
minimum.
When discussing these images one has, however, to bear in mind that the spa-
tial orientation of gaseous free water molecules (as studied here) is statistically
distributed. Thus, we cannot expect any information about this orientation from
the momentum density distributions. A quantitative comparison between experi-
ment and modern quantum chemical methods for the computation of these orbitals
thus requires averaging over the orientation angle as well as a convolution over ex-
perimental resolutions. Carefully analyzed, EMS experiments may thus provide an
important, critical test to molecular orbital calculations beyond the standard spec-
troscopy – which usually ‘only’ measures energies, albeit with high precision. EMS,
in addition, contains information on the spatial structure of the wave functions which
otherwise is not trivial to obtain.
562 8 Electron Impact Excitation and Ionization
Section summary
• Electron impact ionization of atoms and molecules is both, of fundamen-
tal physical interest and of great practical relevance. The simplest process,
H(e, 2e)H+ , represents a three body C OULOMB break up which cannot be
solved exactly in analytic form – in spite of its apparent simplicity. In prin-
ciple, the probability distributions of all final momenta can be determined,
leading to information on four levels of detail:
• An overview is obtained from the integral cross section as a function of initial
kinetic energy T above ionization threshold WI . For small excess energies
(few eV) the WANNIER /R AU threshold law σ (ion) ∝ (T − WI )1.127 is now
well established. Beyond that, σ (ion) rises more or less linearly and reaches
a maximum at about three to six times threshold. It then decreases slowly,
essentially ∝ ln(T )/T as predicted by the B ORN -B ETHE approximation. Ex-
cellent semi-empirical fits are obtained with the L OTZ formula (8.44), very
useful for practical applications.
• The SDCS, dσ/dW , describes the energy partitioning between electron A
and B in the final channel, with T − WI = WA + WB , and W the energy
of the detected electron – both being in principle indistinguishable. The
M OTT /RUDD formula (8.46) describes dσ/dW quite well, with symmetry be-
tween T − WI ≥ W ≥ 0 and a minimum at W = (T − WI )/2.
• The DDCS dσ/dW dΩ specifies in addition also the emission angle of the
detected electron. At high kinetic energies T and low momentum transfer
K = k i − k A the B ORN -B ETHE approximation agrees reasonably well with
experiment. It tacitely implies a model – valid for high T and negligible ex-
change – which distinguishes between the forward scattered electron (high
kinetic energy) and the ejected electron (essentially scattered backward with
low kinetic energy). For the former the GOSD is introduced by (8.55), in anal-
ogy to photoionization, which for very small K approaches the optical limit.
• Finally, the TDCS, dσ/d"dΩA dΩB , also refers to the emission angle of the
second electron. It contains maximum information on the ionization dynam-
ics experimentally accessible. It depends in principle on four observables, the
energy W = T − " − WI of one electron, its scattering angle θA (" being
the energy of the “other” electron), and the polar and azimuthal scattering
angles θB and ϕB of the latter. Characteristic for s states is a bimodal distribu-
tion of electron B, as beautifully imaged in Fig. 8.18: the binary peak arises
essentially from interaction of the two electrons, the ionic core being just a
spectator, while the recoil peak involves multiple interactions of the ejected
electron with the latter.
• Today, one may consider the (e, 2e) process as well understood for H and
He atoms, based on sophisticated experiments and advanced theory. For more
complex atoms and molecules it is still subject to current research.
• EMS is an efficient application of the (e, 2e) process, providing detailed in-
sights into molecular structure – complementary to standard spectroscopy.
8.5 Recombination 563
8.5 Recombination
where the photon is emitted immediately during the transition of the captured elec-
tron from the continuum into a bound state |f of the target. The cross sections are
usually rather small – essentially due to the limited phase space of the final state, i.e.
the density of states is much larger in the continuum than in bound states. The cross
sections do, however, increase with decreasing electron energy and may become
significant.
An altogether different situation is encountered with the so called dielectronic
recombination, which is kind of inverse AUGER process (see Sect. 10.5.1, Vol. 1).
Schematically one may view this process as being composed of three phases,
sketched in Fig. 8.23.
The critical first step (a) is only possible if energy is conserved, i.e. if the kinetic
energy of the electron T (n) together with the (negative) binding energy Wn of the
state |n just corresponds to the excitation energy Wf i = Wf − Wi (one may call
this exchange of a virtual photon):
T (n) − Wn = Wf i . (8.65)
The short-lived state |f looses its energy by spontaneous emission of a photon hν.
However, the electron is typically captured into long-lived, high lying RYDBERG
states. Thus, the state |n remains occupied while the inner electron is typically
found in its initial state.
We note that dielectronic recombination – in contrast to direct radiative recombi-
nation – is a resonant process which occurs only at well defined energies of the free
electron to be captured:
2
Eh qeff
T (n) = Wf i − . (8.66)
2 n2
We have here rewritten (8.65) and assumed that only highly excited states |n of the
ion A(q−1)+ capture an electron. For these we may apply the RYDBERG formula for
hydrogen like atoms, if necessary considering incomplete core screening by qeff .
Electron-ion collisions (see M ÜLLER 2008) at low kinetic energies are studied with
advantage in so called merged-beams. They are used to investigate elastic scattering,
impact excitation and multiple ionization of ions, and of course electron-ion recom-
bination. Figure 8.24 shows (very schematically) a typical setup, this one has been
used by B ÖHM et al. (2002) at the Stockholm heavy ion storage ring CRYRING.
The electron beam originates from a thermionic cathode, is accelerated to some
100 eV, focussed and guided by toroidal magnetic fields to merge with the target
Aq+ ion beam (3.3 MeV to 9.4 MeV). After interaction the two beams are separated
in the same manner. Recombination is detected with the help of a dipole magnet
which separates the parent ion Aq+ and products A(q−1)+ . One finally measures
the A(q−1)+ ion yield as a function of electron energy (and thus as a function of
available relative kinetic energy T in the CM system).
The experiment is by no means trivial. For important details we refer to the orig-
inal publication of B ÖHM et al. (2002). Of general interest are the advantages of
the parallel trajectories of ions and electrons in this merged-beams setup. Not only
does it allow for a sufficiently long interaction region, the adjustment of the relative
energy in this setup is also advantageous. For kinetic energies (masses) Te (me ) and
TI (mI ) of electron and ion, respectively, the relative kinetic energy is given by19
M̄ 2 M̄
T= v = (ve − vI )2 (8.67)
2 rel 2
with the velocities ve,I = 2Te,I /me,I of electron or ion, and the reduced mass is
M̄ = me mI /(me + mI ). At a given ion energy, say of 7 MeV for oxygen, the re-
quired relative kinetic energies of 2 eV ≤ T ≤ 15 eV may be obtained by laboratory
electron energies between 200 eV and 375 eV, which are much easier to generate
and control than in a genuine, low energy electron beam. At the same time the en-
ergy width of the beam is reduced by a factor between 10 and about 5. Thus, the
large energy width of electrons from a thermionic cathode is significantly decreased
without dedicated energy selection.
With a transition energy W2p←2s = 11.95 and 12.02 eV for the 2 P1/2 and 2 P3/2 lev-
els of O4+ , respectively, and with qeff = 5 according to (8.66), the lowest accessible
RYDBERG state that can be populated will have a principle quantum number n = 6,
corresponding to T (n) 2.5 eV. The experimental data shown in Fig. 8.25 even al-
lows to recognize 6 states to different orbital angular momenta . The results are
published as rate constants σ vrel , i.e. as product of cross section and relative ve-
locity, averaged over the energy distribution of electrons and ions. The population of
higher, unresolved RYDBERG states up to the continuum can be recognized clearly.
A significant fraction of the total recombination strengths is accumulated there. The
fact that the signal decreases strongly at the ionization threshold is attributed to
quenching of high lying RYDBERG states by the electric and magnetic fields used in
the optics for guiding ions and electrons, i.e. it is considered an experimental artifact
(B ÖHM et al. 2002).
Profound understanding of the underlying physics has been achieved in this field
of research, not the least owing to sophisticated experimental methods and ambi-
19 At the velocities used here of less than 0.05c one does not yet need to compute these numbers
relativistically – even though this is possible without problems (see Eq. (45) in M ÜLLER 2008).
The results are identical within the experimental energy resolution.
566 8 Electron Impact Excitation and Ionization
tious theory. Beyond the intellectual interest a variety of applications are based on
this knowledge, in particular in plasma physics.
Section summary
• Recombination of a free electron with an ion may occur in a direct, non-
resonant process where all excess energy is released by emission of a photon.
Alternatively and more efficiently the so called dielectronic recombination
occurs resonantly: the electron populates a particular RYDBERG state and the
excess energy is used to excite another electron.
• Such processes may be studied in merged-beams experiments which allow
to efficiently control very small relative kinetic energies between ions and
electrons with good energy resolution.
• Dielectronic recombination is particularly efficient with HCIs. Results for
e− + O5+ are discussed as an illustrative example.
DW: ‘Distorted wave’, method for approximate solution of the close coupling equa-
tions in electron scattering (see Sect. 8.1.2).
DWB: ‘Distorted wave B ORN approximation’, method for approximate solution of
the close coupling equations in electron scattering (see Sect. 8.1.2).
ECS: ‘Exterior complex scaling’, method for solving the close-coupling equations
for scattering problems.
EELS: ‘Electron energy loss spectroscopy’, see Sect. 8.3.2.
EMS: ‘Electron momentum spectroscopy’, method to determine the momentum
distribution of electrons in atoms and molecules exploiting e, 2e processes (see
e.g. M C C ARTHY and W EIGOLD 1991).
FBA: ‘First order B ORN approximation’, approximation describing continuum
wave functions by plane waves; used in collision theory and photoionization (see
Sects. 6.6 and 5.5.2, Vol. 1, respectively).
FWHM: ‘Full width at half maximum’.
GA: ‘G LAUBER approximation’, method for approximate solution of the close cou-
pling equations in electron scattering (see Sect. 8.4.5).
good quantum number ‘Quantum number for eigenvalues of such observables that
may be measured simultaneously with the H AMILTON operator (see Sect. 2.6.5
in Vol. 1)’.
GOS: ‘Generalized oscillator strength’, characterizes the strength of electron im-
pact excitation in analogy to the optical oscillator strength see Sect. 8.3.2.
GOSD: ‘Generalized oscillator strength density’, characterizes the strength of elec-
tron impact ionization per energy interval in analogy to the optical oscillator
strength density (see Sect. 8.4.4).
HCI: ‘Highly charged ions’, see Sect. 7.5.
MO: ‘Molecular orbital’, single electron wave function in a molecule; typically the
basis for a rigorous molecular structure calculation.
ODE: ‘Ordinary differential equation’.
OOSD: ‘Optical oscillator strength density’, characterizes the strength of photoion-
ization per energy interval (see Sect. 5.5.1 in Vol. 1).
PES: ‘Photoelectron spectroscopy’, see Sect. 5.8.
PWIA: ‘Plane wave impulse approximation’, basic assumption for EMS (see
Sect. 8.4.6).
SBA: ‘Second order B ORN approximation’, second order term in the B ORN series
(see Sect. 6.6).
SDCS: ‘Single-differential cross section’, in e, 2e ionization processes (see
Sect. 8.4).
TBW: ‘Three body approximate scattering wave function’, accurate method for cal-
culating triple-differential cross section for ionization (B RAUNER et al. 1989).
TDCC: ‘Time dependent close-coupling calculations’, a method, in principle accu-
rate, for solving the S CHRÖDINGER equation for scattering problems (C OLGAN
and P INDZOLA 2006).
TDCS: ‘Triple-differential cross section’, in e, 2e ionization processes (see
Sect. 8.4).
UV: ‘Ultraviolet’, spectral range of electromagnetic radiation. Wavelengths be-
tween 100 nm and 400 nm according to ISO 21348 (2007).
568 8 Electron Impact Excitation and Ionization
References
A L -H AGAN O., A. J. M URRAY, C. K AISER, J. C OLGAN and D. H. M ADISON: 2010. ‘Electron-
impact-ionization cross sections of H2 for low outgoing electron energies from 1 to 10 eV’.
Phys. Rev. A 81, 030701.
A MALDI , U., A. E GIDI, R. M ARCONER and G. P IZZELLA: 1969. ‘Use of a 2 channeltron coin-
cidence in a new line of research in atomic physics’. Rev. Sci. Instrum., 40, 1001–1004.
A NDERSEN , N. and K. BARTSCHAT: 2003. Polarization, Alignment and Orientation in Atomic
Collisions. Berlin, Heidelberg: Springer.
A NDERSEN , N., K. BARTSCHAT, J. T. B ROAD and I. V. H ERTEL: 1997. ‘Collisional alignment
and orientation of atomic outer shells: 3. Spin-resolved excitation’. Phys. Rep., 279, 252–396.
A NDERSEN , N., J. W. G ALLAGHER and I. V. H ERTEL: 1988. ‘Collisional alignment and orienta-
tion of atomic outer shells: 1. Direct excitation by electron and atom impact’. Phys. Rep., 165,
1–188.
AVALDI , L., R. C AMILLONI, E. FAINELLI and G. S TEFANI: 1987. ‘Absolute double-differential
ionization cross-section for electron-impact: He’. Nuovo Cimento D, 9, 97–113.
BAERTSCHY , M., T. N. R ESCIGNO and C. W. M C C URDY: 2001. ‘Accurate amplitudes for
electron-impact ionization’. Phys. Rev. A, 64, 022709.
BARTLETT , P. L. and A. T. S TELBOVICS: 2004a. ‘Differential ionization cross-section calcula-
tions for hydrogenic targets with Z ≤ 4 using a propagating exterior complex scaling method’.
Phys. Rev. A, 69, 040701.
BARTLETT , P. L. and A. T. S TELBOVICS: 2004b. ‘Threshold behavior of e-H ionizing collisions’.
Phys. Rev. Lett., 93, 233201.
BARTSCHAT , K.: 1998. ‘The R-matrix with pseudo-states method: Theory and applications to
electron scattering and photoionization’. Comput. Phys. Commun., 114, 168–182.
BARTSCHAT , K. and I. B RAY: 1996. ‘Electron-impact ionization of atomic hydrogen from the 1s
and 2s states’. J. Phys. B, At. Mol. Phys., 29, L577–L583.
B ERGER , M. J., J. S. C OURSEY, M. A. Z UCKER and J. C HANG: 2005. ‘ESTAR, PSTAR, and
ASTAR: Computer programs for calculating stopping-power and range tables for electrons,
protons, and helium ions (version 1.2.3)’, Physical Measurement Laboratory. http://physics.nist.
gov/Star, accessed: 9 Jan 2014.
B ETHE , H.: 1930. ‘Zur Theorie des Durchgangs schneller Korpuskularstrahlen durch Materie’.
Anal. Phys., 397, 325–400.
VAN B OEYEN , R. W. and J. F. W ILLIAMS : 2005. ‘Multidetection (e, 2e) electron spectrometer’.
Rev. Sci. Instrum., 76.
B ÖHM , S. et al.: 2002. ‘Measurement of the field-induced dielectronic-recombination-rate en-
hancement of O5+ ions differential in the Rydberg quantum number n’. Phys. Rev. A, 65,
052728.
References 569
Overview
Those who want to benefit from reading the next chapter and other advanced
text in AMO physics and are not yet familiar with the density matrix, should
study the summary given in this chapter with some care. After some introduc-
tory remarks we shall start in Sect. 9.1.1 by defining the concept of pure and
mixed states. In Sect. 9.1.2 we formally introduce the density operator and in
Sect. 9.1.3 its matrix representation which will be illustrated by some simple
examples. A general formalism for describing a physical measurement is de-
rived in Sect. 9.2 and applied in Sect. 9.3 to two typical examples. Finally, in
Sect. 9.4 we present an introduction to the general theory of radiation from
quantum systems in non-isotropic and oriented excited states, based on the
formalism developed by FANO and M ACEK (1973).
Up to now, we have usually assumed that the quantum systems discussed can
be characterized by states or wave functions with a set of well defined quantum
numbers – implying that it is completely known prior to an interaction to be dis-
cussed, i.e. at time t → −∞. The evolution under the influence of a H AMILTON
operator describing the whole system occurs then according to the time depen-
dent S CHRÖDINGER equation. At a later time t → +∞ the whole system is still
fully characterized by a wave function, typically a coherent superposition of several
states. This is strictly true for any completely isolated quantum system, as long as
the whole system is considered.
In the real world, however, the basic assumption of pure states is hardly ever a
realistic description of the situation in an actual experiment. We often do not know
the initial state completely, but even if we do the final outcome of an experiment
must not necessarily be a pure state either. Consider e.g. the thermal population of
rotational and vibrational states of a molecule and their alignment and orientation
is space. It may only be described by a probability distribution. Also, we often deal
with quite complex quantum systems which frequently consist of several subsys-
tems (photons, electrons, atoms, molecules) whose individual fate cannot be fol-
lowed in details – and may not be of interest. Thus, at the end of an experiment as
a rule only some characteristic observables of a quantum system are measured. In
the language of quantum mechanics: we project the state of the whole system onto
a certain subset of states which are experimentally accessible. Very often the system
of interest interacts with a dissipative environment – a so called “bath” which is a
very large quantum system whose state cannot be described in quantum mechanical
detail. There, our nice, coherent states are flushed away, so to say.
Hence, real experiments always average over the unobserved quantum numbers,
and this has to be accounted for when comparing with theoretical predictions. The
density matrix provides a convenient method of book-keeping for this purpose.
We mention that there are indeed approaches to completely avoid the concept of
the density matrix or density operator. One may instead use expectation values of
certain tensor operators (multipole moments) which are constructed from angular
momentum operators (see Appendix C in Vol. 1). As it turns out, these quanti-
ties are equivalent to irreducible components of the density matrix. – But even if
one finds the density matrix “sometimes frightening” (Z ARE 1988) it remains dubi-
ous whether these irreducible tensor operators are more comforting to the ordinary
reader. On the contrary, we hope to show that the density matrix offers in many
cases a much more direct access to standard physical intuition.
Nevertheless – as we shall briefly outline in Sect. 9.4 – the representation of
certain experiments in terms of such multipole moments allows a very convinc-
ing disentanglement of dynamical parameters (characterizing a physical interaction
process) from purely geometric factors (which characterize a specific experimental
setup). Also, for irreducible tensor moments it is often much simpler to perform
rotations in space and recoupling of angular momenta (if necessary) – than for the
density matrix itself. Hence, this elegant ansatz has been used by FANO and M ACEK
(1973) to describe the radiation characteristics and the polarization of light emitted
by atoms after collisional excitation. M ACEK and H ERTEL (1974) extended this
concept for collisions with laser excited atoms, and G REENE and Z ARE (1983)
further developed it for the analysis of laser induced fluorescence from aligned
molecules.
The density matrix and its irreducible moments are treated in several textbooks
quite comprehensively (A NDERSEN and BARTSCHAT 2003; M UKAMEL 1999;
K LEIMAN et al. 1998; B LUM 2012; B RINK and S ATCHLER 1994; W EISSBLUTH
1989; Z ARE 1988). We just try here to give a compact version of some important
9.1 Some Terminology 575
state-of-the-art tools for the conception and interpretation of many advanced experi-
ments in laser spectroscopy and collision physics. In this presentations – so to say for
pedestrians – we shall refrain (as elsewhere in these books) from presenting lengthy
mathematical excursions and detailed proofs and built on the willingness of the read-
ers to simply accept certain results and concepts – and find out that they are useful.
To have something concrete in mind, Fig. 9.1 shows – very schematically – three
typical experimental arrangements for studying atomic and molecular interaction
processes with state selection and/or analysis.
Figure 9.1(a) represents what may be called an ‘ideal’ interaction experiment
between two subsystems (A, B) – we may think of a scattering experiment or a
chemical reaction – where the quantum states |γA |γB of both partners before the
interaction process are assumed to be well characterized. One may achieve this by
suitable state selectors (sel A, B), the physical details of which is of no interest to
us at this point. The total system prior to interaction may be described by a wave
function of the type
with r A and r B being the coordinates, γA and γB the quantum numbers of all interior
degrees of freedom for A and B, respectively (including spin variables). The relative
momentum1 between A and B is characterized by k, while R describes their relative
coordinates. The fact that the total wave function can be factorized expresses the
independence of the subsystems prior to the interaction.
During the interaction process the state of the system changes. A formal de-
scription for a scattering process has been presented in Sects. 6.4.6 and 8.1, using
the transition operator T (T-matrix) of the process. It transfers |γA γB k i asymptot-
ically (t → ∞) into T|γA γB k i . In principle, both subsystems (A, B) may undergo
changes, they may e.g. be excited, ionized, rearranged internally or even form new
products (C, D).
The final state T|γA γB k i and the respective wave function may thus describe a
rather complex situation. In principle one can no longer assume that final state may
still be written as a product state φC (r C )φD (r D ) of the subsystems (C, D) after the
process. To use a fashionable word: as a rule one deals with entangled states of the
subsystems after the interaction. (See also Appendix E.3 in Vol. 1.) The constituents
‘know’ of each other even at time t → ∞.
In an ‘ideal’ interaction experiment one will try to obtain as much information
about the final states of the subsystems after the interaction, by using suitable state
1 We use here again, somewhat loosely, wave vector (k) synonym for momentum (k).
576 9 The Density Matrix – A First Approach
(a) (b)
A sel
anl
A C C
sel anl
A
A C
D anl
sel
B D anl
B
D
(c)
A
bath B
Fig. 9.1 Schematics of prototypical interaction experiments: one tries to characterize the quantum
state of the system prior and after the interaction (T) as good a possible by using state selectors
(sel) and analyzers (anl), respectively; (a) reaction or collision experiment between partners A
and B; (b) photoexcitation process in an isolated system; (c) photoexcitation of A in a “bath” B
analyzers (anl C, D) to determine the quantum states |γC and |γD of the sepa-
rated reaction products C and D. And this should be done in a correlated fashion,
i.e. by detecting C and D in (if necessary delayed) coincidence. Using quantum me-
chanical terms, this detection process projects the final state T|γA γB k i of the total
system onto product states |γC γD k f of the subsystems, with k f representing the
magnitude and direction of the final relative momentum of C and D. Thus one ob-
tains a scattering amplitude γC γD k f |T|γA γB k i and a cross section corresponding
to (7.31) and (7.33). This kind of measurement implies automatically a significant
reduction of the full reality – which is described by T|γA γB k i . In addition, only in
exceptional cases it will be possible to determine all relevant quantum numbers of
all subsystems after the interaction process.
In a quite similar manner the experiment indicated in Fig. 9.1(b) may be de-
scribed: one or more photons interact with a well prepared atom or molecule A. In
this process A may e.g. be excited, subsequently undergo internal rearrangement,
re-emit photons and finally may even fragment. Here too, new end products will be
created (C, D). And the T operator is in this case the interaction operator for the
atom in an electromagnetic field which we have encountered already in Chap. 4,
Vol. 1 – as far as necessary supplemented by the description of further reaction
steps.
Still somewhat more complicated is the experiment sketched in Fig. 9.1(c). For-
mally, however, it will be described in very similar terms. In this case, the T operator
has to describe also the interaction with the bath B, which in the individual situation
may be rather complicated – even if one assumes that prior to photo-absorption A
was very well characterized and did not interact with the bath.
9.1 Some Terminology 577
For one particularly simple, special case we shall now specify this general de-
scription in order to illustrate the concept. Be A an atom which initially is found in
a γ 2 S1/2 state. It may assume two sets of quantum numbers: {γ 2 S1/2 , m = +1/2}
or {γ 2 S1/2 , m = −1/2}, i.e. the electron spin may point up or down. These two
states (let us call them |+ and |− for briefness) correspond then to the initial state
|γA γB k i in (9.1).
Figure 9.2 shows a schematic of state selection prior to an interaction experi-
ment. Let us have a somewhat closer look at it. After collimation (by apertures) and
velocity selection, the atomic beam may be described approximately as a “plane
wave”. It passes an m state selector (e.g. a S TERN -G ERLACH magnet or a circularly
polarized laser beam employed for optical pumping as described in Appendix D)
which prepares one of the states |+ or |−.
We emphasize that such a selection characterized by + or − always refers to a
well defined z(sel) -axis of the m selector, i.e. to the direction of the magnetic field B
in a S TERN -G ERLACH magnet or to the direction of light propagation in the case
of optical pumping. Note that the interaction experiment is often described with
advantage in a different coordinate system, e.g. in the standard collision frame as
introduced in Sects. 6.4.2 and 7.3.1, denoted by x (col) y (col) z(col) .
of the angular momentum operators J and Jz . In general, an ideal state selector
2
(0, π/2, π) so that the z(col) axis points into z(sel) direction. One obtains (E.20),
Vol. 1 which we write here
∓1
|φ0 (sel) = |± = √ |+1/2 ± |−1/2 . (9.3)
2
The states |±1/2 are characterized by the projection quantum numbers m = ±1/2
with respect to the z(col) -axis.2
We emphasize that (9.3) is a linear (or coherent) superposition of eigenstates.
More generally, any so called pure state may be written as a linear superposition of
basis states |γ
|α = aγ(α) |γ (9.4)
γ
(α)
with the probability amplitudes aγ .
One has to sum over all quantum numbers
which characterize the system studied. With the eigenfunctions φγ (r) which belong
to these quantum numbers γ we may write |α also as a wave function:
φ0 (r) = r | α = aγ(α) φγ (r). (9.5)
γ
Thus, reality confronts us with the situation that the initial state can no longer be
described by a pure state (or a wave function) – not even by any kind of linear
combination of pure states. Such a state of the system is called a mixed state or an
incoherent superposition of pure states. Of course the probabilities p± are normal-
ized such that still
p+ + p − = 1 (9.7)
holds. The quality of the selector (or of the selected atomic beam) is described by
its polarization
2 InEq. (E.20), Vol. 1 these spin states |±1/2 are called |α and |β, respectively, not to be con-
fused with the state designation in the following text.
9.1 Some Terminology 579
I+ − I−
P = p+ − p − = , (9.8)
I+ + I−
where I + and I − stands for the measurable beam intensities of atoms with spin up
or spin down, respectively.
Coherent or Incoherent?
At this point a warning is appropriate against too naive usage of the terms coherent
and incoherent. For example, the question about the probability to find the states
|+1/2 or |−1/2 with respect to the collision coordinates (col) in case (b) would
correctly be answered with p+1/2 = p−1/2 = 1/2. But clearly, these probabilities do
not fully characterize the pure states (9.3): we deal here with a coherent superposi-
tion of two states. According to (9.2) the same two states written in selector frame
(sel) are |+ and |−. And the respective probabilities in this coordinate system are
{p+ = 1, p− = 0} and {p+ = 0, p− = 1}, respectively, depending on the choice of
selector transmission.
It is evident that the probability to find a specific basis state depends on the
coordinate system. With reference to Fig. 9.2: only when considering the state of the
atoms A before the beam enters the selector we find a completely incoherent mixture
of states {p+ = p− = 1/2} which does not change upon changing the coordinate
system {p+1/2 = p−1/2 = 1/2}.
This may be generalized to the case of N possible basis states |γ = |1, |2 . . .
|N with probabilities
p1 , p2 . . . pN . (9.9)
Again, in general such probability distribution refers to a particular coordinate sys-
tem and is normalized:
N
pγ = 1. (9.10)
γ =1
Thus, if we do not know more than this probability distribution we cannot distin-
guish whether we deal with a pure or mixed state.
in a fully incoherent state if and only if each of the basis states |γ is found
with equal probability.
As a consequence of (9.10) this implies pγ = 1/N for each basis states |γ with
respect to any coordinate system.
To check this quality in our case (in a “Gedanken experiment”), one would have
to analyze the fully or partially selected beam emerging from the selector with a
580 9 The Density Matrix – A First Approach
second (if possible ideal) selector, the reference axis of which could be aligned into
an arbitrary direction in space. If the first selector would have prepared a pure state,
then the second selector would transmit the beam to 100 % only if aligned into one
specific direction in space. If the state prepared by the first selector is a mixed state,
always less than 100 % will pass the second selector irrespective of its alignment.
−1
p+ for |+ = √ |1/2 + |−1/2 and
2
(9.11)
1
p− for |− = √ |1/2 − |−1/2 .
2
Again, there is no possibility to transform this description by a linear transformation
of basis states into a single pure state. This is not surprising since we still discuss the
same incoherent superposition of states as in (9.6), only described in a different co-
ordinate system. We see, however, that the coordinate transformation may allow to
see a certain degree of coherence even in the mixed state – it describes the presence
of two states with different probabilities which as such are coherent superpositions
of basis states |1/2 and |−1/2. Thus, one has to be very careful when using the
terms coherence and incoherence.
Hence, it is desirable to obtain a quantitative description of terms such as “mixed
state”, “degree of coherence” and “degree of polarization”. This becomes even more
evident in the general case of mixed states which are described by an incoherent
superposition of pure states |α – i.e. by a set of probabilities with respect to a
reference coordinate system:
p1 for state |1 = aγ(1) |γ
γ
p2 for state |2 = aγ(2) |γ (9.12)
γ
..
.
pN for state |N = aγ(N ) |γ
γ
N
again with pα = 1.
α=1
9.1 Some Terminology 581
Obviously, this kind of description of an initial state (not to speak of the final
state after an interaction) is rather clumsy – in particular in comparison with pure
eigenstates |γ , which we have assumed so far tacitly as starting point for our con-
siderations (also in Vol. 1). When including the detection process into the descrip-
tion, the situation becomes increasingly complex (see Fig. 9.1), since the analyzers
too are usually far from perfect.
We thus need an efficient and clear method of bookkeeping. For this purpose one
defines a density matrix of the states to be described. In the most trivial case of a
pure state |α according to (9.4) its elements are given by
(α)
In the case of mixed states according to (9.12), with amplitudes aγ of individual
states α = 1 . . . N , the density operator it is simply given as the weighted sum of its
components:
N
N
pα aγ aγ∗(α) |γ γ |.
(α)
ρ̂ = pα ρ̂ (α) = pα |αα| = (9.15)
α=1 α=1 γ γ α
N
N
ργ γ = γ |ρ̂|γ = pα aγ aγ∗(α) = aγ aγ∗ ,
(α) (α)
pα ρ γ γ = (9.17)
α=1 α=1
aγ aγ∗ being an abbreviation for the averaging. One verifies easily
In the following, the trace of the density operator will play an important role, also
in combination with other operators. For normalized, pure states
Tr ρ̂ (α) = a (α) 2 =α | α= 1
γ =
ργ(α) γ (9.19)
γ γ
holds, and the same is true for the trace of any density operator:
Tr ρ̂ = ργ γ = γ =
pα ργ(α) pα = 1. (9.20)
γ αγ α
The density matrix3 is thus a convenient abbreviation for lengthy expressions re-
quired for a realistic description of an averaging processes which typically occurs
in a measurement.
Let us work out the expectation value O of an arbitrary observable O in a
mixed state according to (9.12). We first recall from (2.42), Vol. 1 that for a pure
state
∗(α)
= α|O|α
O = aγ(α) .
aγ γ |O|γ (9.21)
γ γ
For a mixed state, in addition one has to average over all pure states |α involved:
N
= = ∗(α)
O pα α|O|α γ |O|γ pα aγ aγ(α) . (9.22)
α γ γ α=1
where Oγ γ are the matrix elements of the observable O in the basis {|γ }. We shall
meet such kind of relations in several contexts as useful abbreviations for measur-
able quantities.
All these definitions may appear still somewhat abstract, even artificial. They
turn out, however, to be very useful if one tries to formulate the theory of measure-
ment for specific examples in some detail. Before doing so we need to familiarizes
ourselves somewhat better with the density matrix.
3 One uses the terms density operator and density matrix more or less synonymous, the latter em-
beam with γ 2 S1/2 atoms which were (partially) selected according to their quantum
numbers.
We recognize now the characteristic off-diagonal terms ρ−1/2 1/2 , which imply co-
herence among the substates |+1/2 and |−1/2. They ensure that these density ma-
trices represent pure states, while the diagonal matrix elements |a1/2 |2 = |a−1/2 |2 =
1/2 just give the probability to find states |+1/2 and |−1/2, respectively, with
reference to the collision coordinates (col). This may, e.g., be verified with a second
selector which is aligned along the z(col) -axis.
In the last equality we have used the polarization P as defined in (9.8). A completely
unpolarized, mixed state (P = 0) is represented by
1 1 0 1
ρ̂ = = 1. (9.29)
2 0 1 2
584 9 The Density Matrix – A First Approach
This matrix would describe e.g. the initial atomic beam prior to entering the selec-
tor. Note that the zero off-diagonal matrix elements in (9.28) and (9.29) indicate a
complete lack of coherence between the states |+ and |−.
Now the density matrix has a nonvanishing off-diagonal matrix element which docu-
ments (partial) coherence between the basis states |1/2 and |−1/2. We thus clearly
recognize that any quantitative expression of coherence among basis states depends
on the choice of the basis system: (9.28) and (9.30) describe exactly the same mixed
state but have different off-diagonal matrix elements. Mere rotation of the coordi-
nate system may thus ‘generate’ or ‘destroy’ coherence terms (here P)! This should
be born in mind if one reads or writes about the observation of coherence in an
experiment.
Only for a completely unpolarized state the density matrix is diagonal indepen-
dent of the coordinate system – as one verifies by inserting P = 0 into (9.28) and
(9.30).
while for a pure state the averaging ai ak∗ over the amplitudes is suppressed.
This general form of the density matrix may look somewhat complicated, and in
general one does not recognize immediately what kind of state is described by it –
a pure or a mixed state.
All we can say is that the probability to find the states |1, |2 . . . |N is given by
ρ11 , ρ22 . . . ρN N and that a certain coherence between the state |i and |k exists
if ρik = 0. The picture becomes much clearer by diagonalizing (9.31). This may
be achieved using the rules of linear algebra by a suitable unitary transformation
9.1 Some Terminology 585
{|γ } = {|γ (d) } of the basis states |γ . One thus obtains a density matrix in diag-
U
onal form:
⎛ (d) ⎞
ρ11 0 0 ··· 0
⎜ (d) ⎟
⎜ ρ22 0 ··· 0 ⎟
⎜ .. ⎟
ρ̂ (d) = U † = ⎜
ρ̂ U ⎜
..
.
⎟
. ⎟. (9.32)
⎜ .. ⎟
⎜ .. ⎟
⎝ . . ⎠
(d)
0 ρN N
Such a transformation removes the coherence terms which depend on the choice
of basis states. It allows one to express the density matrix exclusively by the proba-
bilities ρ (d) for finding the states |γ (d) in this particular choice of basis states. The
magnitude of these probabilities gives a clear measure for the degree of purity or
mixture of states under consideration. There are two extreme cases: for a pure state
one and only one diagonal element (=1) remains, all others disappear:
⎛ ⎞
0 0
⎜ 0 ⎟
⎜ ⎟
(d) ⎜
ρ̂pure = ⎜ 1 ⎟. (9.33)
⎟
⎝ 0 ⎠
0 0
In contrast, the extreme opposite is a mixed (unpolarized) state where all basis states
are found with equal probability p = 1/N :
⎛ ⎞
1/N 0 0 ··· ··· 0
⎜ .. ⎟
⎜ 1/N 0 . ⎟
⎜ ⎟
(d) ⎜
ρ̂unpol = ⎜ .. ⎟ = 1
⎜ 1/N . ⎟ ⎟ N
. (9.34)
⎜ .. ⎟
⎝ . ⎠
1/N
Note that the matrix representation of this particular state is completely inde-
pendent of the choice of the basis: any unitary transformation of the identity matrix
contained in (9.34) reproduces simply the identity matrix. We have seen this already
for the 2 × 2 matrices according to case (c) and (d), if P = 0 is inserted into (9.28)
and (9.30), respectively.
We shall now try to formulate the above discussion quantitatively without recur-
ring to the sketched diagonalization procedure, which can be rather tiresome. What
would be desirable is a single, easy to compute parameter which allows us to quan-
tify the degree of mixture or purity in between the two extremes just defined.
586 9 The Density Matrix – A First Approach
We first recall that the trace of a density matrix is always Tr ρ̂ = 1 per defini-
tion (9.20) – for a pure as well as for any mixed state. In contrast, the trace of the
square of a density matrix is unity Tr ρ̂ 2 = 1 for a pure state, as obvious from (9.33).
However, for a completely unpolarized state Tr ρ̂ 2 = N × 1/N 2 = 1/N holds, as
evident from the diagonal form (9.34). For any other mixed state one finds values in
between the two extremes. This is again evident for the two diagonal representations
of ρ̂. And since the trace of any matrix is independent of any unitary transformation,
the general relation also holds for any density matrix:
1 ≥ Tr ρ̂ 2 ≥ 1/N . (9.35)
The limits 1 and 1/N refer to a pure and a completely unpolarized state,
respectively.
Let us familiarize ourselves with these terms for the example of a 2 × 2 density
matrix
ρ11 ρ12
ρ̂ = . (9.36)
ρ21 ρ22
The diagonal matrix elements are real (they represent probabilities) and Tr ρ̂ =
ρ11 + ρ22 = 1, while the off-diagonal matrix elements may be complex with
∗ . Hence, three real parameters suffice to fully characterize the most general
ρ21 = ρ12
2 × 2 density matrix, let us say ρ11 , |ρ12 | and arg ρ12 . Thus
2
ρ11 + ρ12 ρ21 ρ11 ρ12 + ρ12 ρ22
ρ̂ 2 = ,
ρ11 ρ21 + ρ21 ρ22 2 +ρ ρ
ρ22 12 21
and with (ρ11 + ρ22 )2 = 1 and ρ12 ρ21 = |ρ12 |2 we obtain for (9.35)
1 ≥ Tr ρ̂ 2 = 1 + 2 |ρ12 |2 − ρ11 ρ22 ≥ 1/2. (9.37)
which leads to two eigenvalues for the diagonalized density matrix ρ̂ (d)
12 3
ρ (d) = 1± 1 + 4 |ρ12 |2 − ρ11 ρ22 . (9.41)
2
9.1 Some Terminology 587
With (9.37) one may write the difference between these eigenvalues as
(d) (d)
ρ11 − ρ22 = 1 + 4 |ρ12 |2 − ρ11 ρ22 = 2 Tr ρ̂ 2 − 1. (9.42)
For the pure state (Tr ρ 2 = 1) this gives ±1, for the unpolarized state (Tr ρ 2 = 1/2)
(d) (d)
we obtain ρ11 − ρ22 = 0.
(d) (d)
This difference ρ11 − ρ22 corresponds to the polarization P of the two state sys-
tem defined by (9.8), as we see by direct comparison with the diagonal form (9.28).
Since Tr ρ̂ 2 is independent of the choice of the coordinate system, (9.42) offers a
good starting point for the sought-after definition of a degree of polarization. For
the two state system we define it (omitting the sign in order to be truly independent
of the coordinate system) as:
(d)
(d)
|P| = ρ11 − ρ22 = + 2 Tr ρ̂ 2 − 1 (9.43)
The left and right limits correspond to the pure and unpolarized states, respectively.
The thus defined degree of polarization is more tangible than the somewhat abstract
expression Tr ρ̂ 2 .
We generalize (9.43) for an N × N density matrix ρ̂ and define the
While (9.45) corresponds to the geometric mean value of all degrees of polarization
of any two states of the diagonalized density matrix ρ̂ (d) , the definition (9.46) is
independent of the coordinate system. The identity of the two expressions may be
verified with a little bit of algebra,
√ using the normalization according to (9.20). The
normalization constant ∝ 1/ N − 1 is chosen such that
1 ≥ P (N ) ≥ 0, (9.47)
again with the left and right limits representing fully coherent and unpolarized (fully
incoherent) states, respectively.
One may also be interested in the coherence (or purity) of a subset of N basis
states. The degree of polarization of such a subset is obtained in full analogy to
(9.46), where instead of ρ̂ the corresponding N × N submatrix ρ̂ and instead of
N → N is to be used.
588 9 The Density Matrix – A First Approach
Section summary
• The density operator ρ̂ defined by (9.16) and the density matrix ργ γ accord-
ing to (9.17) has been introduced as a useful and flexible tool for book keeping
when describing experiments with mixed (not fully coherent) states of quan-
tum systems.
• Pure states are represented by a linear superposition of basis states. In contrast,
mixed states cannot be described in that way. Rather they are characterized by
the probabilities for finding different pure states.
• Expectation values of observables are obtained from the trace of the their
product with the density matrix according to (9.22).
• We have familiarized ourselves with the concept by simple 2 × 2 matrices,
describing pure as well as mixed states of a system with two basis states. It
is very important to realize that the density matrix depends crucially on the
choice of the coordinate system – and so does the observation of coherence
among states.
• Each density matrix may be brought into diagonal form by suitable unitary
transformation. Pure states are characterized by a single unit element in this
diagonal form, while the fully unpolarized (incoherent) state has equal proba-
bility 1/N for all its component states.
• A unique measure for the degree of purity of a state is obtained from 1 ≥
Tr ρ̂ 2 ≥ 1/N . Based on this we define a general degree of polarization (9.45)
for which 1 ≥ P (N ) ≥ 0. The upper and lower limits refer to pure and com-
pletely unpolarized (mixed) states, respectively.
We shall now develop the tools described above into a handy formalism to describe
the signal detected in real experiments. With experiments in mind as sketched in
Fig. 9.1 we have to account for the fact that neither the state selectors nor the ana-
lyzers are perfect.
Up to now we have used the density matrix (or the density operator) to describe
a quantum system in a mixed state of several pure states |α. We now change the
perspective slightly and use ρ̂(α) = ρ̂ (sel) to describe a state selector acting as a
filter which transmits (from an initially completely unpolarized state) a mixed state
described by ρ̂ (sel) – which is constructed according to (9.31).
In complete analogy we use the density matrix ρ̂(β) = ρ̂ (anl) to describe an an-
alyzer. Again, it filters the state described by ρ̂ (anl) out of a completely unpolarized
ensemble.
We may now describe a preliminary experiment, sketched in Fig. 9.3: the key pa-
rameter is the transmission probability p(β, α) through a state analyzer – character-
ized by ρ̂ (anl) – for a quantum ensemble prepared by a state selector – characterized
by ρ̂ (sel) .
9.2 Theory of Measurement 589
source
state
selector
^
ρ(α)
state
>
{|q }
^
analyzer
^
ρ(β)
ρ(α)
^
ρ(β)
towards
detector detector
Fig. 9.3 Schematic of a preliminary experiment: quantum states prepared by a state selector are
detected after passing an analyzer. Indicated by the large circle on the bottom left is the total set
of all basis states {|q}. From these, the selector would transmit the subset denoted by ρ̂(α), the
analyzer would transmit ρ̂(β). The intersection of these subsets of states reaches the detector after
passing through both, selector and analyzer – as indicated by the pink coloured area
Let us first consider an ideal state selector which prepares a pure state |α, i.e.
we assume its density operator ρ̂(α) = |αα| to project just this state out of the
initially unpolarized ensemble. And similarly, we assume the state analyzer to be
characterized by a density operator ρ̂(β) = |ββ| which projects only the state |β
out of whatever enters into it (in our case the state |α). Thus, the state passing
through the analyzer is given by
and the squared absolute value of the probability amplitude β|α is just the proba-
bility to detect a signal (in state |β) after passing the analyzer:
2
p(β, α) = β|α . (9.49)
With β|α∗ = α|β, using (9.4) we may rewrite this with respect to the basis states
{|γ } as
∗(β) (α)
p(β, α) = β|αα|β = aγ aγ aγ∗(α) aγ(β) . (9.50)
γ γ
Using the definition (9.13) of the density matrix elements for the pure states |α and
|β the signal expected in the experiment will be proportional to
(β) (α)
p(β, α) = ργ γ ργ γ = Tr ρ̂(β)ρ̂(α) . (9.51)
γ γ
The generalization is obvious: let us now consider an imperfect state selector which
prepares a mixture of states |α, each with a probability pα , and an imperfect ana-
lyzer which transmits several states |β with the probabilities pβ . We thus have to
sum in (9.50) over all prepared and transmitted signals, weighted with the proba-
bilities pα and pβ characterizing selector (sel) and analyzer (anl), respectively. The
590 9 The Density Matrix – A First Approach
By reorganization this sum we recover the density matrix elements (9.17) for state
selector and analyzer and obtain the measured signal:
(sel) (anl)
S(sel, anl) = ργ γ ργ γ = Tr ρ̂ (sel) ρ̂ (anl) = Tr ρ̂ (anl) ρ̂ (sel) . (9.53)
γ γ
This is the basis for all further considerations about measurable signals with
state selection and/or analysis.
We thus emphasize again the significance of the operators for selector and analyzer,
ρ̂ (sel) and ρ̂ (anl) , respectively: they describe the mixed state which the selector and
analyzer would filter out from a completely unpolarized source.
We are now prepared to exploit the book keeping function of the density matrix for
the evaluation of experiments with (partial) state selection prior to an interaction
experiment and (partial) state analysis after it – i.e. we use the density matrix for
a realistic description of real experiments. We have to derive an expression for the
signal detected after an interaction process with an experimental scheme as sketched
in Fig. 9.1. For compact writing we abbreviate the quantum numbers of the system
prior to the collision with {γ } = {γA , γB }, including if relevant also k i . The basis
states are correspondingly |γ . The operator of the state selector is then
ργ γ |γ γ |.
(sel)
ρ̂ (sel) = (9.54)
γ γ
into
T
ργ γ T|γ γ |T† = Tρ̂ (sel) T† .
(sel)
ρ̂ (sel) −→ (9.56)
γ γ
The strength of the signal detected in the interaction experiment is obtained by re-
placing ρ̂ (sel) in (9.53) with (9.56):4
S(sel, anl) = Tr Tρ̂ (sel) T† ρ̂ (anl) = Tr ρ̂ (sel) T† ρ̂ (anl) T . (9.57)
Note that state selector and analyzer enter into this expression in fully equivalent
manner. We may thus use the same theoretical tools to describe the time inverse
experiment, by just interchanging source and detector in Fig. 9.3: one obtains the
same signal if the selector is described by ρ̂ (anl) and the analyzer by ρ̂ (sel) .
We point out that the important relation (9.57) may also be written in terms of
matrix elements for the transition operator T and density matrix elements. One sim-
ply has to insert the quantum mechanical unit operators (2.43), Vol. 1 in between
the operators in (9.56). We do not want to discuss the resulting, somewhat lengthy
expressions for the general case, but focus on two specialized examples for illustra-
tion.
4 Strictly speaking one has to multiply the following expression with normalizing factors which
account for the fact that selector and analyzer operators are not necessarily normalized to 1. For
clarity of writing we suppress this detail here, knowing that in each experiment anyhow several
further transmission and detection factors have to be accounted for when evaluation a measured
signal.
5 In a next step of sophistication pγ̃ could e.g. follow a B OLTZMANN distribution.
592 9 The Density Matrix – A First Approach
after the process. Using the left equality of (9.57) the measured signal is
S(sel, anl) = dk f pγ ε k f |T|γ k i γ k i |T† |εk f εk f |ρ̂ (anl) |ε k f
ε γ ε
∗ (anl)
= pγ Tε γ (k f , k i )Tεγ (k f , k i ) ρεε . (9.60)
εε γ
Obviously, the term in square brackets [. . . ] describes the total system directly af-
ter the collision, it defines what may be called the density matrix of the collision
process:
−
→
ρε ε = CT−1 ∗
(col)
pγ Tε γ (k f , k i )Tεγ (k f , k i ). (9.61)
γ
The summation over γ may be rather elaborate and comprises in principle a full
partial wave expansion as given in (8.5). Alternatively we may exploit the propor-
tionality between the scattering amplitude f and the T-matrix according to (7.31)
and write
−
→
ρε ε = Cf−1 ∗
(col)
pγ fε γ (θ, ϕ)fεγ (θ, ϕ). (9.62)
γ
With this notation we may rewrite (9.60) in the general form (9.53). We obtain the
differential cross section for such an experiment by inserting (9.62) into (9.60) and
accounting for the flux factors (7.31):
(anl)
dσ −→
I (θ, ϕ) = (k f , k i ) = I0 (θ ) Tr ρ̂ (col) ρ̂ (anl) . (9.64)
dΩ
The prefactor I0 denotes the standard inelastic differential cross section – not state
specific, averaged over all initial and summed over all final states:
(av)
dσ kf kf 2
I0 (θ ) = (k f , k i ) = Cf = pγ fεγ (θ, ϕ) . (9.65)
dΩ ki ki εγ
Equation (9.64), with the definitions (9.62) and (9.65), is the basis for the evaluation
of all scattering experiments with state analysis after collision.6 In full analogy one
6 This includes of course the experiment without any state analysis. In that case ρ̂ (anl) =
1/Nβ
and Tr(ρ̂ (col) ) = 1. Thus I (θ, ϕ) = I0 (θ) which is given by the standard formula (9.65) for the
differential cross section.
9.2 Theory of Measurement 593
may evaluate optical excitation and detection, such as the experiments sketched
in Fig. 9.1(b, c), with T representing in that case the electromagnetic interaction
operator D.
−
→
At this point it is important to ensure that ρ̂ (col) and ρ̂ (anl) used in (9.64) refer to
the same coordinate system. As we have seen in Sect. 9.1.1, in practice this is often
−
→
not the case. Typically, it is convenient to describe ρ̂ (col) in a particular coordinate
system – e.g. in the standard collision frame (col), while ρ̂ (anl) is best described with
respect to other reference axes (anl) as dictated by the experimental setup. Hence,
one of these coordinate systems has to be transferred into the other reference frame.
This implies rotation by D(αβγ ) through the E ULER angles (αβγ ) as described in
Appendix E, Vol. 1. Introducing such rotation into (9.64) leads to
−→ † (−→ −→
Tr ρ̂ (col) ρ̂ (anl) → Tr D ρ̂ (anl) = Tr ρ̂ (col) D
ρ̂ col) D ρ̂ (anl) D
† . (9.66)
Explicitly one has to insert unit operators in between all these operators, so that this
formalism obviously results in somewhat clumsy expressions.
Alternatively one may rewrite it in terms of expectation values of the irreducible
tensor operators into which the density operator may be expanded. This allows a
much more elegant formulation of transformation problem. This procedure has long
been established in nuclear physics in the context of so called “perturbed angular
correlations”. In atomic physics it was introduced by FANO and M ACEK (1973) to
describe the angular distribution and polarization of light emitted from atoms ex-
cited in atomic collisions obtained in coincidence experiments. The essentials are
derived in Sect. 9.4. This theory of measurement may well be extended to other
types of state selection or analysis, e.g. to experiments with electron spin polariza-
tion (BARTSCHAT et al. 1981) or to collisions with laser excited atoms (M ACEK and
H ERTEL 1974).
(Often one has pγ 1 for a particular group of states, and = 0 for all others, which
may be suppressed e.g. by the energy analyzer.)7 Note the slightly different use of
7 This shows again the necessity to introduce a normalization constant for detector and state ana-
lyzer (which we have omitted here for simplicity). If a quantitative signal evaluation is aimed for
one always has to ensure that density matrices are normalized to Tr ρ̂ = 1.
594 9 The Density Matrix – A First Approach
the indices (for reasons which will become obvious at the end of this section): we
still denote the group of not analyzed states by γ and γ , the selected group by ε
and ε (as in the previous example). However, the former now refer to the states
after the interaction process, the latter to those before. We now evaluate the right
side of the detection equation (9.57) in full analogy to the preceding example (9.60).
Introducing unit operators and (9.68) gives
S(sel, anl) = dk i ε k i |ρ̂ (sel) |εk i εk i |T† |γ k f pγ γ k f |T|ε k i
εε γ
pγ Tγ ε (k f , k i )Tγ∗ε (k f , k i ) ,
(sel)
= ρε ε (9.69)
εε γ
where in the second line we have used (9.67). With the notation
←
−
(col)
ρεε = CT−1 pγ Tγ∗ε (k f , k i )Tγ ε (k f , k i ) (9.70)
γ
= Cf−1 pγ fγ∗ε (θ, ϕ)fγ ε (θ, ϕ) (9.71)
γ
←
−
and Tr ρ̂ (col) = 1 (9.72)
←− ←−
I (θ, ϕ) = I0 (θ ) Tr ρ̂ (sel) ρ̂ (col) = I0 Tr ρ̂ (col) ρ̂ (sel) (9.73)
kf kf 2
with I0 = Cf = pγ fγ ε (θ, ϕ) . (9.74)
ki ki εγ
Since now we perform state selection prior to the collision, and since the
T-matrix now describes a scattering process from the states |ε to |γ (while be-
fore T was referring to transitions from |γ to |ε) a subtle difference between both
examples becomes apparent: the density matrix of the scattering process (9.70) is
constructed such that one does not only replace ε by γ (and vice versa); one also
has to apply the complex conjugate definition to (9.61). In addition, we have so
sum over all detection probabilities pγ for the individual states, while in example
1 the quantity pγ refers to the probability of finding the state |γ ) prior to the col-
lision process. A somewhat more symmetric representation of the present result is
obtained when we view example 1 and example 2 as being the time inverse of each
other, i.e. as processes |γ → |ε and |ε → |γ , respectively.
In this scheme we also have to replace k i by −k f and k f by −k i so that
∗
Tγ ε (k f , k i ) = Tεγ (−k i , −k f ). (9.75)
9.2 Theory of Measurement 595
(b)
B collisio
B
where Cf is the flux factor according to (9.65), and the angles θ̃ , ϕ̃ indicate that
the scattering amplitudes have now to be described in a coordinate system which is
defined by −k i and −k f . Expressions (9.76)–(9.77) are fully equivalent to (9.61)–
(9.62) in the previous example – except that we have replaced k i by −k f and k f
by −k i . Thus, a scattering experiment inducing transitions ε → γ with state selected
particles and no final state analysis, provides identical information as an experiment
for the inverse process γ → ε without initial state selection but with final state
analysis. All measurable quantities are contained in the respective collision density
matrix ρ̂ (col) .
Figure 9.4 illustrates the two time inverse experiments schematically. One exam-
ple for such experiments has already been treated in Sect. 7.4.5.
We emphasize, however, that the formalism described here is by no means lim-
ited to scattering processes. We have presented here a rather general approach to the
theory of measurement with state analysis. In completely similar manner one may
e.g. describe an optical excitation process which is analyzed by a suitable detection
scheme – e.g. by measuring the polarization of the fluorescence. And analogue con-
siderations are to be made e.g. for pump-probe studies of the temporal evolution in
molecules, clusters or liquids. A comprehensive description is found in the standard
textbook by M UKAMEL (1999) on this particular subject.
596 9 The Density Matrix – A First Approach
Section summary
• We have made use of the density matrix to introduce a general theory of mea-
surement involving mixed (or partially polarized) states. To do so we have
extended the concept of the density matrix to describe also state selectors and
analyzers. These are represented by a matrix, equivalent to the density matrix
of the state which they filter out from a completely unpolarized state.
• The signal in such experiments is described essentially by S ∝ Tr[ρ̂ (anl) ρ̂ (sel) ]
= Tr[ρ̂ (sel) ρ̂ (anl) ], where ρ̂ (sel) denotes the state prepared either by a selector
or in a collision process or by any other dynamics, while ρ̂ (anl) refers to a state
analyzer. A scattering process with an unpolarized projectile and an unpolar-
ized target is characterized by ρ̂ (sel) = ρ̂ (col) = TT† , where T is the standard
transition matrix (T-matrix) defined in collision physics.
Partially polarized light and the S TOKES parameters P1 , P2 and P3 have already
been introduced in Sect. 1.3.3 – somewhat heuristically. The S TOKES parameters
allow us to fully characterize light which is not necessarily fully coherent – i.e. par-
tially polarized light. The following, more rigorous derivation of these parameters
and their properties is used as a practical example for exploring the use of density
matrices.
We refer here again to the standard, spherical unit basis vectors e+1 and e−1 , and
denote the corresponding states for left and right hand polarized light with |1 and
|−1, respectively, assuming that the light propagates in the +z-direction (helicity
basis). Due to the transversality of electromagnetic radiation the amplitude for the
|0 component is zero. We also recall: the alignment of the polarization ellipse with
9.3 Selected Examples of the Density Matrix 597
Table 9.1 Pure polarization states of light (apart from an overall phase factor) and the correspond-
ing density matrices in the helicity basis; we may consider these expressions as ρ̂ (anl) to describe
the respective polarization filter; these are actually all 3 × 3 matrices, but for compact writing we
have omitted all zero components
Row Polarization β Basis state |eel Density matrix ρ̂el
4 5
1 left circular (LHC) π
2 |1 ρ̂LHC = 10 00
4 5
2 right circular (RHC) 0 |−1 ρ̂RHC = 00 01
4 5
√1 [e−iδ |1 − eiδ |−1] 1 −e−2iδ
3 linear at δ π
4 ρ̂δ = 12
2 −e 2iδ 1
4 5
1 −1
4 δ = 0◦ π
4
√1 [|1) − |−1)] ρ̂0◦ = 2 −1 1
1
2
4 5
−i
5 δ = 90◦ π
4
√ [|1) + |−1)] ρ̂90◦ = 12 1 1
2 11
4 5
δ = 45◦
1 i
2 [(1 − i)|1 − (1 + i)|−1] ρ̂45◦ = 2 −i 1
π 1 1
6 4
4 5
7 δ = 135◦ π
4 − 12 [(1 + i)|1 − (1 − i)|−1] ρ̂135◦ = 12 1 −i
i 1
One easily verifies for this effective 2 × 2 matrix that Tr ρ̂pol2 = 1 (full coherence),
Of special interest are the three pairs (1, 2), (4, 5) and (6, 7) of photon states
according to Table 9.1. Their differences may be written as:
0 −1
ρ̂ − ρ̂ =
0◦ 90◦ with Tr(ρ̂0◦ − ρ̂90◦ ) = 0 (9.80)
−1 0
0 i
ρ̂45◦ − ρ̂135◦ = with Tr(ρ̂45◦ − ρ̂135◦ ) = 0 (9.81)
−i 0
1 0
ρ̂LHC − ρ̂RHC = with Tr(ρ̂LHC − ρ̂RHC ) = 0. (9.82)
0 −1
Interestingly, these expressions are identical with the PAULI spin matrices according
to (2.101) in Vol. 1 (apart from an overall phase factor −1 for the first two). They
are unitary and the following ‘orthogonality’ relations hold:
(9.89)
Of course for each one of the wave trains |ep (α) in the mixed ensemble the basis
vectors are correlated with a fixed phase difference 2δα and an ellipticity angle βα :
each of them represents pure elliptic polarization over a time scale τ0 . However,
averaged over a whole ensemble this may nevertheless lead to an unpolarized state:
e.g. in the case that the phases δα are distributed statistically, and hence the off-
diagonal matrix elements ρ+− average out (i.e. the coherence term disappears).
In the general case the light described by (9.89) is neither fully polarized nor
completely unpolarized. The matrix, i.e. the polarization state of the light, may be
described by three real parameters, since ρ++ + ρ− − = 1 and ρ+− = ρ−+ ∗ is a
complex quantity. The relations (9.85)–(9.87) suggest to attribute these three pa-
rameters to the three differences of polarization matrices (9.80)–(9.82) and to add
an unpolarized background (identity matrix):
1
ρ̂ (pol) = 1 + P1 (ρ̂0◦ − ρ̂90◦ ) + P2 (ρ̂45◦ − ρ̂135◦ ) + P3 (ρ̂RHC − ρ̂LHC ) . (9.90)
2
This defines a convenient parametrization of the most general polarization matrix
(9.89) for light travelling into z-direction. Written as 2 × 2 matrix it is
ρ11 ρ1−1 1 1 − P3 −P1 + iP2
ρ̂ (pol)
= = . (9.91)
ρ−1+1 ρ−1−1 2 −P1 − iP2 1 + P3
Up to now, P1 , P2 and P3 are still free parameters. We shall derive their physical
significance in the following and identify them as S TOKES parameters which we
have introduced in Sect. 1.3.3.
For reference we also communicate the polarization matrix in the {|ex , |ey }
basis which is often used alternatively in the literature:
ρxx ρxy 1 1 + P1 P2 + iP3
ρ̂ (pol) (x, y) = ∗ = . (9.92)
ρyx ρyy 2 P2 − iP3 1 − P1
This is readily derived from (9.91) using the relations (4.4)–(4.6), Vol. 1 between
the basis states.
We finally mention that for linearly polarized light it is sometimes convenient to
use an alternative coordinate frame, with its z-axis parallel to the polarization vector
600 9 The Density Matrix – A First Approach
i.e. perpendicular to the wave vector, z ⊥ k. In this coordinate system the polariza-
tion vector is simply elin = e0 while all other components vanish. The respective
polarization matrix is given by
⎛ ⎞
0 0 0
ρ̂ (pol) = ⎝ 0 1 0 ⎠ . (9.93)
0 0 0
where the normalization constant I0 turns out to be the total intensity in the beam;
this is verified by inserting (9.91) and the analyzer matrices from Table 9.1, and
exploiting (9.83):
And for the relative intensity differences one finds from (9.92) with the ‘orthogo-
nality’ relations (9.85)–(9.88)
I (0◦ ) − I (90◦ ) 1
= Tr ρ̂ (pol) (ρ̂0◦ − ρ̂90◦ ) = P1 Tr(ρ̂0◦ − ρ̂90◦ )2
I (0◦ ) + I (90◦ ) 2
= P1 , (9.96)
I (45◦ ) − I (135◦ )
= P2 and (9.97)
I (45◦ ) + I (135◦ )
I (RHC) − I (LHC)
= P3 . (9.98)
I (RHC) + I (LHC)
We thus have expressed the parameters used in (9.90) and (9.91) in terms of quan-
tities which are directly accessible to the experiment. And these turn out to be in-
deed the S TOKES parameters defined in Sect. 1.3.3. A suggestive abbreviation is
the S TOKES vector of the light beam:
P = (P1 , P2 , P3 ). (9.99)
9.3 Selected Examples of the Density Matrix 601
1
ρ1−1 = ρxy ≡ 0 and ρ11 = ρ−1−1 = ρxx = ρyy ≡ . (9.102)
2
This may suggest the definition of a so called degree of coherence:
√
|μ| = |ρxy |/ ρxx ρyy . (9.103)
However, while this quantity is used in the literature quite often, its definition is
somewhat unfortunate as it depends on the choice of the basis and the coordinate
system used: this is immediately seen by (9.92), where P2 depends of course on the
choice of the x-axis, while P3 is independent of this selection. Much more robust
is the degree of polarization as defined by (9.43). If one inserts ρ̂ (pol) according to
(9.91) or alternatively according to (9.92), it is determined by the S TOKES parame-
ters completely independent of the coordinate system
|P| = + P12 + P22 + P32 with 1 ≥ |P| ≥ 0. (9.104)
602 9 The Density Matrix – A First Approach
The limiting cases indicate completely polarized and unpolarized light, respectively.
In summary, we note that the three S TOKES parameters are easy to determine ex-
perimentally, and they contain the complete information on the polarization state of
the light beam.
For the experimental characterization of a light beam, described by (9.91) in
the helicity basis, by means of a linear polarizer (as ideal as possible) one has to
apply (9.94). With the analyzer matrix ρ̂δ from Table 9.1 (row 3) one obtains the
transmitted signal as a function of the analyzer angle δ:
I (δ) = I0 Tr ρ̂ (pol) ρ̂δ .
We leave it to the reader as an easy exercise to show that this leads to the formulas
already communicated in (1.103)–(1.105).
General Discussion
We continue our exploration of the concept and use of density matrices by a look at
specific atomic states. We want to describe the simple case of an atom in a p state
with its three substates |m = ±1, |m = 0, pertaining to important atomic model
systems. For the sake of simplicity we ignore for the moment the electron spin and
the nuclear spin and simply focus on the electronic orbitals, assuming a 1 P np state
(e.g. in He). We further assume that the p level considered is energetically well
separated from other energy levels of the system. The general 3 × 3 density matrix
(for any system with angular momentum j = 1),
⎛ ⎞
ρ11 ρ10 ρ1−1
ρ̂ = ⎝ c.c. ρ00 ρ0−1 ⎠ , (9.105)
c.c. c.c. ρ−1−1
The angular component (solid angle ) of the wave function for this pure state is
given by
|p = b1 Y11 (θ, ϕ) + b0 Y10 (θ, ϕ) + b−1 Y1−1 (θ, ϕ). (9.107)
9.3 Selected Examples of the Density Matrix 603
In the general case of mixed states, several different contributions of similar type
constitute the charge cloud: the amplitude products bm bm ∗ in (9.108) must then
∗
be replaced by density matrix elements ρm m = bm bm . To simplify the situation
somewhat we restrict the discussion in the following to a situation (often found in
experiments) where at least one symmetry plane exists with respect to which the
charge density – proportional to (9.108) – is symmetric.
For convenient writing we choose the xy plane to be this symmetry plane, and
call this reference system the atomic coordinate system,8 x (at) y (at) z(at) . The last
terms in (9.108) contain products of the type sin θ cos θ which in the upper and
lower hemisphere, z > 0 and z < 0, respectively, have different sign and are not
mirror symmetric with respect to the x (at) y (at) plane. We drop them completely in
(9.108) so that the corresponding off-diagonal terms disappear: ρ0±1 = ρ±10 ≡ 0.
The most general distribution of an electron charge cloud for a p state with xy
reflection symmetry in the “atomic” coordinate system is thus
2 3 1
I (θ, ϕ) = |p = ρ00 cos2 θ + ρ11 + ρ−1−1
4π 2
+ 2|ρ1−1 | cos 2(ϕ − γ ) sin2 θ , (9.109)
8 In the literature related to collisional alignment and orientation studies this is usually called the
As the structure of this density matrix shows, a fully coherent state implies that
one of the two sub-matrices disappears: only then (9.110) can be transformed into
the form (9.33) by diagonalization. If ρ00 = 0 one determines from (9.38) whether
ρ̂ + represents a pure or a mixed state.
We also note that this special choice of the coordinate system allows a direct
physical interpretation of the density matrix elements of ρ̂ (at) (which otherwise are
somewhat abstract quantities). We first recall that the diagonal matrix elements ρ11 ,
ρ00 , ρ−1−1 give the relative probability for finding the substates m = 1, 0 and −1.
From this a particular important observable is deduced: the expectation value of the
angular momentum10 in z-direction:
L z ) = ρ11 − ρ−1−1 .
z = Tr(ρ̂ L (9.113)
9 Note that this xy reflection symmetry must not be confused with the xz reflection symmetry used
to define the real angular momentum states and real tensors introduced in Appendix D.3, Vol. 1.
For a graphical illustration see Fig. D.1 in Vol. 1.
10 Forsimplicity of writing we use here again atomic units, i.e. angular momenta are measured in
z |m = m|m.
units and L
9.3 Selected Examples of the Density Matrix 605
Even the off-diagonal matrix element, the so called coherence term, obtains in these
coordinates a direct physical meaning which may be read directly from the charge
distribution (9.109). A graphic illustration is given in Fig. 9.5.
We first note that for |ρ1−1 | ≡ 0 the charge distribution becomes a rotational
ellipsoid with z being the symmetry axis. For ρ00 = 1/3 this degenerates to a com-
pletely isotropic distribution with ρ11 + ρ−1−1 = 2/3. But even in this case the
angular momentum (9.113) may still remain finite: a fully spherical charge distribu-
tion may still contain an inherent asymmetry. In general, ρ00 = 1/3 and |ρ1−1 | will
have a finite value so that the charge cloud has three symmetry axes as illustrated in
Fig. 9.5: the z-axis and axes of maximal and minimal probability in the xy plane.
The latter point into the direction ϕ = γ and ϕ = γ + π/2, respectively. The align-
ment angle γ is obtained according to (9.109) directly from the phase arg(ρ1−1 ) of
the off-diagonal matrix element:
1
γ = − arg(ρ1−1 ) ≥ π/2. (9.114)
2
For symmetry reasons γ is defined only modulo π . The absolute value |ρ1−1 | char-
acterizes the polarization of the ρ̂ + submatrix, which we shall call linear polariza-
tion of the charge cloud:11
11 Note that the corresponding, explicit expressions (9.8) and (9.43) refer to the diagonal form of
a 2 × 2 matrix, which in the present case can be obtained by rotation through −γ around the
z(at) -axis.
606 9 The Density Matrix – A First Approach
Let us assume that the initial states of A and B are fully known and can be described
(A) (B)
by wave functions φi (r A ) and φi (r B ), respectively. In the following we shall
analyze two situations of different complexity.
and following the derivations in Sects. 7.3 and 7.3.1, we write the asymptotic wave
function of the whole quantum system as
1
Ψ (R, r) = exp(ik i R)φi (r) + ff i (θ, ϕ) exp(ikf R)φf (r). (9.122)
R
f
The wave vectors k i and k f characterize again the relative motion before and af-
ter the collision process, respectively, and ff i (θ, ϕ) is the corresponding scattering
amplitude. The wave function of the total system prior to the interaction (i.e. for
t → −∞, R → ∞) can thus be written as a product function:
(A) (B)
Ψ (R, r) = exp(ik i R)φi (r A )φi (r B ). (9.123)
Note that this expression represents a pure state, and the question is: how can a
collision process ever lead to the observation of incoherent ensembles of states?
The sum in (9.122) must in principle be carried out over all possible final states
φf (r) of the total system. Now, let the scattering experiment be designed such that
after the interaction process only one specific set of states of the excited atom A is
selected by a state analyzer – let us say it detects all substates |npm in one level.
This may e.g. be achieved by measuring the relative kinetic energy after the collision
9.3 Selected Examples of the Density Matrix 607
or by a spectral analysis of the fluorescence from atom A after the collision. The thus
prepared state of the total system is then described by
+1
1 (B)
Ψanl (R, r) = exp(iknp R)φi (r B ) (A)
fnpm φnpm (r A ), (9.124)
R
m=−1
with the abbreviation fnpm = ff i (θ, ϕ). Note that the sum in this expression de-
scribes the wave function of the atom excited to the np level by the collision. We
emphasize that the total wave function Ψanl (R, r) still represents a pure state: being
written as product of the wave functions of the collision partners A and B and their
relative momentum. Thus, the coherence of the quantum system is conserved dur-
ing and after this particular interaction process – and the selection of the specified
states. This still holds if we ask the question in which state particle A is found after
the collision process: the simple answer is that its wave function after the interaction
( (A)
and after analysis describes a pure, coherent state: fnpm φnpm (r A ).
Before discussing the next case, we also must recall that the scattering ampli-
tudes obey certain symmetry relations, such as (8.6), which refer there to scattering
amplitudes in the standard collision frame.12 One finds (here without proof) that
these symmetry relations translate into the atomic frame as: symmetry with respect
to the x (at) y (at) plane (now the scattering plane) is conserved in the collision pro-
cess. Thus, if we start with an xy positive state, e.g. in an n1 s → n2 p transition, only
the m = ±1 states are excited. In an n3 d → n2 p transition md = ±1 → mp = 0 are
possible, as well as md = ±2 → mp = ±1 and md = 0 → mp = ±1.
Clearly, the total quantum system, consisting of particle A and particle B, is still in
a coherent state as given by the two particle wave function in the square brackets.
However, if we now ask how to describe atom A as an isolated particle after the
interaction, it is no longer possible to factor out the wave function of particle B –
12 Tocompare the standard collision frame introduced in Fig. 7.16 with the “atomic frame” (at)
used here (also called the “natural frame”) the latter has to be rotated through the E ULER angles
(−π/2, −π/2, 0).
608 9 The Density Matrix – A First Approach
which was the critical step applied to (9.124) in the previous situation: No longer
can we describe A after the collision by one well defined wave function – and the
same holds for particle B! The total system consisting of particle A and particle B
has to be described by an entangled state.
If we insist on looking only at the state of particle A after the collision – and that
is what a standard scattering experiment usually does – A has to be characterized
(j )
now by a density matrix. With the scattering amplitudes fm , which are the proba-
bility amplitudes for simultaneously finding B in state |j and A in state |npm, we
obtain
⎛ (1) (1)∗ ⎞
f1 f1 0 f1(1) f−1(1)∗
⎜ ⎟
ρ̂ (at) = p (1) ⎝ 0
(1) (1)∗
f0 f0 0 ⎠
(1) (1)∗
c.c. 0 f−1 f−1
⎛ (2) (2)∗ (2) (2)∗
⎞
f1 f1 0 f1 f−1
⎜ ⎟
+ p (2) ⎝ 0
(2) (2)∗
f0 f0 0 ⎠. (9.125)
(2) (2)∗
c.c. 0 f−1 f−1
This description (in the atomic frame) allows also to describe processes which
change the reflection symmetry in respect to the x (at) y (at) plane. The general shape
of the resulting p state charge cloud is illustrated in Fig. 9.5. The probabilities p (j )
are obtained from
(B) 2 3 (
|φj | d r B (j ) 2
m |fm |
p = (B)
(j )
(B)
× ( (1) 2 (2) 2
.
(|φ1 |2 + |φ2 |2 )d3 r B m (|f m | + |f m | )
e− + H(1s) → e− + H(2p).
The total spin of the system S = SH + S e remains constant during the collision,
since spin-orbit interaction can be neglected for this very light atom. Two possible
total spin quantum numbers are possible, S = 0 or 1, i.e. the total system is found
to be either in a singlet state (S) with the statistical weight p (S) = 1/4, or in a triplet
state (T) with the weight p T = 3/4. Both spin states may have rather different scat-
tering amplitudes. Thus, the density matrix elements of the excited hydrogen state
in its 2p state are given by ρmm = C1 ( 34 fmT fmT∗ + 14 fmS fmS∗ ), with the normalization
(2p)
constant being
1
3 T 2 1 S 2
C= fm + fm . (9.126)
4 4
m=−1
One easily verifies that in general this does not lead to a coherent state. Trivially,
a mixed state is obtained when both wave functions participate, those with positive
reflection symmetry (m = ±1) as well as antisymmetric ones (m = 0). In the present
9.3 Selected Examples of the Density Matrix 609
case, the latter processes would require spin flip processes – expected only for atoms
with large Z. However, even if only states with positive reflection symmetry are
excited one has to distinguish by (9.38) whether a pure or a mixed state is observed.
Only a strictly linear relation between singlet and triplet amplitudes would allow a
simplification of (9.125) such that coherence is maintained. As a rule this will not
be the case and the excited state will be incoherent.
Optical Excitation
Alternatively we discuss optical excitation of an atom into an np state. For simplic-
ity we restrict ourselves here to a situation where optical pumping can be neglected
(see however Appendix D). We thus consider a 1 P1 state |b without hyperfine struc-
ture, excited by single photon absorption from an initial 1 S0 state |a. As discussed
in Chap. 4, Vol. 1 the density formalism is not needed as long as the excitation is
achieved with fully polarized, coherent light: in this case the excited atom is de-
scribed by a wave function as in (9.107). To warm up to our task we nevertheless
write down the density matrix for this case, recalling the results from Sect. 4.7 in
Vol. 1.
As far as possible the polarization states are described in a coordinate system with
its +z-axis parallel to the direction of light propagation (wave vector kz). The well
known selection rules m = +1 and −1 hold for left (σ + , LHC) and right (σ − ,
RHC) circularly polarized light, respectively. This light thus excites the |1 Pm = 1
and |1 Pm = −1, respectively, the excitation amplitudes are b1 = 1 and b−1 = 0 (al-
ternatively b1 = 0 and b−1 = 1). In contrast, if one uses linearly polarized light prop-
agating also into kz-direction (its electric field vector E(r, t) lying in the xy plane
aligned at an angle γ with respect to the x-axis, so called σ light), then a linear com-
bination of√both states is excited as described
√ in (4.136), Vol. 1. The amplitudes are
b1 = +(1/ 2)e−iγ and b−1 = −(1/ 2)e+iγ . The respective density matrices are
⎛ ⎞
1 0 0
for σ + light with kz: ρ̂LHC = ⎝ 0 0 0 ⎠ , (9.127)
0 0 0
⎛ ⎞
0 0 0
for σ − light with kz: ρ̂RHC = ⎝ 0 0 0 ⎠ and (9.128)
0 0 1
⎛ ⎞
1 0 −e−2iδ
1⎝
for σ light with E ⊥ z: ρ̂lin = 0 0 ⎠. (9.129)
2
−e 2iδ 1
1. Of all the states, potentially excited, according to (9.110), only those with posi-
tive reflection symmetry can be excited: as long as the exciting light propagates
into z-direction and the E vector lies in the xy plane, we have ρ00 = 0.
2. According to (9.127)–(9.129), the sub-matrices of ρ̂ + are completely identical to
those of the exciting light (see Table 9.1).
610 9 The Density Matrix – A First Approach
In practice a light beam has always some divergence (not all components of it
propagate exactly into one direction) and we have to average over the contributing
directions of incidence. If the laser beam is slightly divergent and its central axis
is assumed to be parallel to the z-axis, not all components with negative reflection
symmetry average out. The density matrix of a 1 P1 state excited with such a light
beam will contain a small contribution ρ00 :
⎛ ⎞
1 − P3 0 −P1 + iP2
1 ⎝ ⎠.
ρ̂ = 0 ρ00 0 (9.132)
2 + ρ00 −P − iP 0 1+P
1 2 3
Again, four real parameters P1 , P2 , P3 , ρ00 , describe such a 1 P1 state. They may
again be related to the physical quantities discussed above: the angular momentum
in z-direction according to (9.113) is
practice one has to be somewhat cautions when automatically extracting γ from P1 and P2 .
13 In
Since the standard function arctan is not univalued one has always to keep the physical geometry
in mind.
9.4 Angular Distribution and Polarization of Radiation 611
The linear polarization of the charge cloud (9.115) is equal to the linear polarization
of the exciting light:
1/2
Plin = 2|ρ1−1 | = P12 + P22 . (9.136)
Finally, we may introduce a fourth S TOKES parameter of the charge cloud which
measures the relative difference between its length in x-direction and its height.
With (9.116), (9.119) and −2 Re ρ1−1 = 2|ρ1−1 | cos(2γ ) = Plin cos(2γ ) this quan-
tity becomes
I ( π2 , 0) − Iz λ − ρ00 1 − 3ρ00 − 2 Re ρ1−1
P4 = = = . (9.137)
I ( π2 , 0) + Iz λ + ρ00 1 + ρ00 − 2 Re ρ1−1
Section summary
• Partially polarized light may be described with advantage by a density matrix
ρ (pol) according to (9.91). Convenient parametrization is provided by the three
S TOKES parameters, summarized as S TOKES vector P = (P1 , P2 , P3 ).
• A polarization analyzer can also be described by such a density matrix ρ (anl)
or by a S TOKES vector P (anl) . The signal from a partially polarized source
which passes through this analyzer is given by I0 Tr(ρ (pol) ρ (anl) ) = (I0 /2)(1 +
P · P (anl) ).
• The density matrix of an isolated 1 P state has a quite similar 3 × 3 structure.
It becomes most transparent if an “atomic” coordinate system is chosen with
reflection symmetry of the charge distribution in respect to its x (at) y (at) plane.
The density matrix then decomposes into a 2 × 2 submatrix corresponding
to wave functions with positive reflection symmetry and a single element ρ00
with negative wave function symmetry. The state is fully characterized by 4
parameters: in addition to ρ00 which reflects the height of the charge cloud, we
have its alignment angle γ = arg(ρ1−1 ), its linear polarization Plin = 2|ρ1−1 |
and its inherent angular momentum L⊥ = ρ11 − ρ−1−1 .
• Incoherence in an excited atom A∗ can be generated in a collision processes
with a partner B which may change its quantum state: After such a process
the state of the whole system is entangled, and the isolated atom A can no
longer be described as a coherent wave function – it requires a density matrix
description.
information can be gleaned about the excited system by a judicious choice of the
experimental observation geometry?
In the terminology used above: How do we have to design the state analyzer in
order to disentangle mere geometry of the experimental setup from the informa-
tion about the state of the system studied? How can we measure the density matrix
of an excited atom or molecule that has been prepared e.g. in a binary collision,
by photoexcitation, photo-dissociation or internal conversion? Such knowledge can
provide much deeper insight into the preceding dynamics than average cross sec-
tions or probabilities.
The subject has different names in different areas of physics. In nuclear physics
one speaks about perturbed angular correlations, in atomic collision physics about
coherence and correlation studies. Elegant theoretical concepts have been intro-
duced to atomic physics, based on a key publication by FANO and M ACEK (1973).14
However, this original work is somewhat difficult to read for the unexperienced
reader. We shall therefore try to draw out the key considerations, following essen-
tially the step by step derivation presented in the review of A NDERSEN et al. (1988),
Appendix C.
We have to describe the radiation from a set of excited states |b which decay
into a set of final states |a. According to (4.53) and (4.56), Vol. 1 the relevant
dipole transition operator is
1
1
e∗ = aq∗ e∗q while e = aq e q . (9.139)
q=−1 q=−1
For the moment we assume that atom and photon are described in the same co-
ordinate system, characterized by the spherical unit basis vectors eq introduced in
Sect. 4.1, Vol. 1. The dipole transition operator
D contains all relevant information
on the geometry of the experiment – including the detected polarization determined
by the polarization amplitudes aq . The emission probability (4.67), Vol. 1 from one
14 For more details and illustrative examples we refer the interested reader to the specialized liter-
ature (see e.g. A NDERSEN and BARTSCHAT 2003; A NDERSEN et al. 1997a,b, 1988; B LUM 2012;
Z ARE 1988; H ERTEL and S TOLL 1978; M ACEK and H ERTEL 1974).
15 The minus sign can be pulled out of the sum as only terms with q = ±1 participate.
9.4 Angular Distribution and Polarization of Radiation 613
initially excited state |b = |γj m to one final state |a = |γ̄ j¯m̄ is simply propor-
tional to the squared matrix element | †
Dab |2 .
With (9.138) the dipole transition matrix elements for emission are
1
γ̄ j¯m̄|
D† |γj m = r ab · e∗ = −γ̄ |r|γ j¯m̄|C1−q |j maq∗ . (9.140)
q=−1
Note that the emission rate dR (spont) /dΩ = Re (θ, ϕ), short Re , is given here for
specified polarization e and emission angles θ, ϕ (see e.g. in Fig. 4.3, Vol. 1).
Re (θ, ϕ) is a probability16 per atom, per solid angle and per unit of time and has
the dimension T−1 . We can easily generalize (9.141) for an incoherent superposi-
( (β)
tion of initial states bj m |j m, each present with a probability pβ :
Re = C bj m bj∗m j¯m̄| D|j¯m̄.
D† |j mj m | (9.142)
j m j m m̄
as elements of the density matrix of the excited atom. Hence (9.142) may be written
as
† (det)
Re = C Tr ρ̂D D = C Tr ρ̂ ρ̂ = C Tr
Dρ̂
D† . (9.144)
16 Integration over the excited state lifetime and 4π solid angle should give 1 (if Tr ρ = 1): each
excited atom eventually emits one photon (its average lifetime being ∝ 1/C). In the literature, the
4 /(2πc3 ) – which would reflect the corresponding photon energy
constant C is often given as e2 ωba
emitted, the electron charge e being measured in esu!
614 9 The Density Matrix – A First Approach
ρ̂ (det) =
D†
D, (9.145)
which in the present experiment acts in the same manner as the analyzer matrix in-
troduced in Sect. 9.2. In principle, (9.144) may be evaluated by inserting (9.143) and
(9.140). By suitably chosen polarization, represented by the amplitudes aq , one can
hope to extract the density matrix elements ρ̂j m ,j m of the atom under investigation.
The atom may have been prepared e.g. by optical excitation, internal rearrangement,
or collision – all of these processes may be studied in some detail by analyzing the
polarization of the emitted light.
The key question is, however, which physical information can be extracted by
which specific experiment and how can the geometry of the experiment be disen-
tangled from the dynamics of the preceding preparation process. This requires a
transparent formulation of (9.144). Also, we have to account for the fact that the
density matrix ρ̂ of the system and that for the emitted fluorescence, ρ̂ (det) , are of-
ten described conveniently in different coordinate systems. Thus, we have to rotate
either ρ̂ or ρ̂ (det) according to (9.66). In terms of matrix elements, the detected signal
becomes
(j )∗ (j ) (det)
Re = C Dmm̃ ρ̂m̃m̃ Dm̃ m ρ̂m m , (9.146)
mm̃m̃ m
(j )
with the rotation matrix elements Dmm̃ (0, θ, ϕ) as detailed in Appendix E, Vol. 1.
This is obviously a rather clumsy expression, with the additional complication that
the matrix elements of D† D describing the polarized light emission are also a non-
trivial construct. Hence, this approach is not suited for a direct disentanglement of
geometrical and dynamical parameters.
Consequently – like it or not – we switch to an irreducible representation of the
density and detector matrix in terms of state multipoles. Details are explicated in
Appendix C. For the present discussion we just need to know that the density matrix
can be expanded according to (C.3)
†
ρ̂ = tˆ j j KQ tˆ j j KQ (9.147)
KQ
Thus we insert (9.147) and (9.140) into (9.144) and apply the summations nec-
essary according to (9.141):
†
Re = C Tr tˆ j j KQ
Dtˆ j j KQ
D†
KQ
†
1
=C tˆ j j KQ γ |r|γ̄ γ̄ |r|γ × aq aq∗ (9.149)
KQ q,q =−1
× j¯m̄|C1−q |j m j m |tˆ j j KQ |j mj m|C1−q
∗
|j¯m̄.
j¯j j m̄mm
Here we have made one more generalization: as indicated by aq aq∗ we average
over the products of amplitudes according to (9.139) which describe the detected
polarization. As explained in Sect. 9.3.1, the polarization amplitudes define the po-
larization matrix
ρ̂q q = aq aq∗
(pol)
(9.150)
of the light transmitted by the polarization analyzer. Note that this polarization ma-
trix ρ (pol) has to be distinguished from the detector matrix ρ (det) (9.145) introduced
above. Their multipole moments differ, however, only by recoupling factors, as we
shall see in a moment.
A perfect polarization analyzer would be represented by (9.79), a realistic one
by (9.89). We use again the convenient parametrization (9.91) in the helicity basis,
(det) (det) (det)
with the three S TOKES parameters P1 , P2 , P3 .
Now we insert the matrix elements (C.5) of the statistical tensor operator into
(9.149), replace C1−q ∗ = (−1)−q C1q , and apply the W IGNER -E CKART theorem
(C.8), Vol. 1 to the matrix elements of the renormalized spherical harmonics Ckq .
The reduced matrix elements,
can be pulled out. One obtains a sum over a triple product of 3j symbols to which
the contraction formula (B.69), Vol. 1 can be applied. Careful evaluation of all phase
factors17 finally leads to (see also B LUM 2012, Chap. 5)
C̃
¯ 1 1 K †
Re = (−1)1+j +j (−1)K+Q ¯ tˆ j j KQ ŝ(11)†KQ (9.151)
3 j j j
j¯j j KQ
1
1−q
√ 1 1 K (pol)
with ŝ(11)†KQ = (−1) 2K + 1 ρ . (9.152)
q −q −Q q q
q,q =−1
¯
17 Note also the factor (−1)j −j according to (C.52), Vol. 1 when inverting one of the reduced
matrix elements.
616 9 The Density Matrix – A First Approach
Equation (9.151) is the key relation for the analysis of radiation from anisotropically
populated excited atoms.18
C̃ contains all numerical factors such as C and the product of the reduced matrix
elements, i.e. |γ̄ j¯rγj |2 for j = j . By comparison with (9.148) we identify
the parameters ŝ(11)†KQ as state multipoles of the polarization analyzer matrix
ρ̂q q , while the state multipoles tˆ(j j )†KQ characterize the excited atom. In this
(pol)
terminology, the state multipoles ŝ(11)†KQ of the polarization matrix ρ̂ (pol) differ
only by the recoupling factors from those of the detector matrix ρ̂ (det) . Compared to
(9.146) we have clearly achieved our goal of disentangling the characteristic of the
atom studied from the geometry of the experiment.
After this detailed derivation, we may now appreciate the more general per-
spective given by FANO and M ACEK (1973): the operators D and D† are of rank
k1 = k2 = 1. Their matrix elements are j m |r|j¯m̄ and j¯m̄|r|j m, displayed
in the original equation (9.142). They couple to tensors of rank K. This coupling
scheme of four angular momenta, as read from the matrix elements, may schemati-
cally be indicated by [(j j¯)1(j¯j )1](K) . However, we want to obtain the irreducible
representation of the detector matrix ρ̂ (det) defined by its matrix elements between
j m | and |j m. This corresponds to a coupling scheme [(j j )K(j¯j¯)0](K) (rank 0
since we sum incoherently over the final states j¯m̄). The recoupling of the former
scheme into the latter is achieved by a 9j symbol (recoupling coefficient for four
angular momenta). Fortunately, it simplifies to a 6j symbol since one of the j ’s is 0.
From (B.78), Vol. 1 our result can be retrieved. Alternatively, we may apply (C.50),
Vol. 1 which gives the reduced matrix elements for products of two tensor operators
acting on the same system.
Before illustrating this important result for specific experimental situations, we have
to realize and discuss a few essential points.
18 Notethat in the helicity basis Q drops out of the recoupling factors (−1)K {. . . } since it is even,
as we shall see below.
9.4 Angular Distribution and Polarization of Radiation 617
¯
Hence, the phase factor (−1)1+j +j is cancelled and the isotropic part of the sum
(9.151) is indeed positive as required.
3. On the other hand, the atomic system is characterized by angular moments j, j
from which multipole moments up to rank K = j + j may be constructed. They
all may, in principle, be excited e.g. in atomic collisions. However, if j + j > 2, a
full analysis of all anisotropy parameters is not possible with single photon detec-
tion. This problem can be overcome by experiments with laser excited atoms. In-
stead of analyzing the emitted radiation, in a time inverse experiment the atom is
prepared by optical pumping as sketched in Fig. 9.4 (see also Appendix D). Since
the optical pumping process involves many photons and occurs in the hyperfine
coupling scheme with angular momentum F , state multipoles up to K = 2F are
prepared by this “state selector”. The same formalism can be used to evaluate
such experiments. One simply has to replace (−1)K+Q {6j }ŝ(11)†KQ in (9.151)
by ŝ(F F )†KQ . Details are described in H ERTEL and S TOLL (1978) and A N -
DERSEN et al. (1988), Appendix D.
4. The angular momenta j and j of the excited atom studied may represent either
the hyperfine quantum number F , or the total electronic angular momentum J ,
or even the orbital angular momentum L – depending on the preparation pro-
cess. According to the so called hypothesis of P ERCIVAL and S EATON (1958)
the relevant coupling scheme depends essentially on the interaction time tint dur-
ing which the atom is excited. If tint
/FS (with FS being the fine structure
splitting) spin and orbit are decoupled during the process and the orbital angular
momentum L = j is the relevant quantity. This is typically the case for elec-
tron or fast ion impact excitation. If on the other hand /FS
tint
/HFS ,
where HFS is the hyperfine splitting, one has to describe the process in terms
of the total electronic angular momentum J = j . Typical cases are fine structure
transitions induced in thermal atom-atom collisions. Finally, if /HFS
tint as
in the optical pumping case just described, the hyperfine quantum number F = j
is relevant for describing the atom.
5. This brings us to the temporal behaviour of the emission process, which so far
we have completely neglected. However, if the excitation time is experimentally
recorded, e.g. in a coincidence experiment between exciting laser pulse (or im-
pacting particle) and emitted photon one has to account for the temporal evo-
lution of the excited state. If different excitation energies are involved the time
dependence of the emission amplitudes may lead to quantum beats as already in-
troduced in Sect. 4.7.2, Vol. 1. Even in the absence of electric or magnetic fields,
excited state levels j and j may be split by Ej − Ej = ωj j due to fine and/or
hyperfine structure. This may be accounted for by factors (see e.g. B LUM 2012,
618 9 The Density Matrix – A First Approach
Chap. 5)
1 − exp[−iωj j − (Aj + Aj )/2]
∝
−iωj j − (Aj + Aj )/2
In the relevant literature different authors use different representations of the multi-
pole moments. In the tradition of FANO and M ACEK (1973) we prefer the real mul-
tipole moments constructed from angular momenta as detailed in Appendix C.2 –
partially because the numerical factors involved look somewhat simpler, while e.g.
B LUM (2012) and A NDERSEN and BARTSCHAT (2003) use the state multipoles.
State multipoles and multipole moments are proportional to each other as described
by (C.16)–(C.20). For compact writing we denote the real multipole moments of the
atom by TKQp and those describing the polarization analyzer PKQp .
We rewrite (9.151) in this notation for j = j after pulling the 6j symbol (9.153)
for K = 0 out the sum. This makes the result quite transparent, noting that T00+ =
P00+ = 1 and T00− = P00− = 0 (see also Table C.1):
C
Re (θ, ϕ) = gK (j, j¯)GK TKQp PKQp (9.154)
3
KQp
1+j +j¯+K
1 1 K v(K, j )v(K, 1)
¯
with gK (j, j ) = (−1) 3(2j + 1)
j j j¯ v(0, j )v(0, 1)
9.4 Angular Distribution and Polarization of Radiation 619
with v(K, j ) given by (C.16). With this definition g0 = 1 holds. The depolarization
factors GK (with G0 = 1) can be derived using the recoupling concepts sketched
in Appendix C.3 and appropriate time averaging over the decay process. Explicit
expressions for GK are documented in A NDERSEN et al. (1988), together with a
compact treatment of the closely related quantum beats. Also given there is an in-
troduction to the theory for the time inverse experiment – scattering from laser ex-
cited atoms. The reader may also be interested in the main part of that review and its
companions A NDERSEN et al. (1997a,b), which give a comprehensive and unified
summary of pioneering experiments on alignment and orientation in atomic colli-
sions.
g0 (j, j¯) = 1s
3 3
g1 (1, 0) = g1 (2, 1) = (9.155)
2 4
1 1
g2 (1, 0) = g2 (2, 1) = .
2 12
Let us study an aligned and/or oriented excited atom as illustrated in Fig. 9.5 (the
geometry is the “atomic frame”, called “natural frame” in collision experiments).
We first assume that the experiment detects fluorescent light propagating into
+z(at) -direction. The analyzer is characterized in the helicity basis by the polariza-
tion matrix (9.91). The values of the real multipole moments PKQp for the detector
are derived from this polarization matrix ρ̂ (pol) as outlined in Appendix C.2. With
(C.10) and (C.17)–(C.20), or from the explicit values read from Table C.1 for J = 1,
we obtain:
P00+ = 1
(pol)
P10+ = −P3
(9.156)
P20+ = 1
√ (pol) √ (pol)
P22+ = − 3P1 P22− = − 3P2 .
All other multipole moments of the analyzer are zero. Note in particular, that all
PKQ± are zero if Q is odd – this is a consequence of the special structure of ρ̂ (pol) ,
620 9 The Density Matrix – A First Approach
expressing the transversality of light (see also footnote 18). Inserting these parame-
ters into (9.154) and abbreviating gK = gK (j, j¯) gives
C √ (pol) √ (pol)
Re = 1 + g2 T20+ − g2 T22+ 3P1 − g2 T22− 3P2
3
(pol)
− g1 T10+ P3 . (9.157)
We have assumed here that no hyperfine splitting is involved and all GK = 1. This
restriction can easily be removed by replacing gK → gK GK . Clearly, Re (θ, ϕ) ≥ 0
as one may verify for specific situation with the explicit expressions for the multi-
pole moments given in Table C.1.
As a consequence, the extremely simple structure of (9.157) for the signal allows
a straight forward evaluation: Specific multipole moments may be measured by a
specific choice of the analyzer S TOKES parameters. The total intensity emitted into
+z(at) = z(ph) -direction is formally obtained by adding the signal for two orthogonal
we obtain:
polarizer settings. Omitting the overall scaling factor C,
(pol) (pol) 2
Rz = Re Pi = 1 + Re P i = −1 = {1 + g2 T20+ } (9.158)
3
(pol)
(Pi may be any of the three S TOKES parameters). The S TOKES parameters Pi
of the detected light are derived from the corresponding differences of the signal:
(pol) (pol) 2
Rz P1 = Re P1 = 1 − Re P1 = −1 = − √ g2 T22+ (9.159)
3
(pol) (pol) 2
Rz P2 = Re P2 = 1 − Re P2 = −1 = − √ g2 T22− (9.160)
3
(pol) (pol) 2
Rz P3 = Re P3 = 1 − Re P3 = −1 = − g1 T10+ . (9.161)
3
Alternatively, when observing linearly polarized light emitted in the x (at) y (at)
plane, with linear polarization perpendicular to it, one may with advantage use a
“photon frame” (ph) with k ⊥ z elin , i.e. the z-axis is this time perpendicular to the
wave vector k and parallel to the polarization. The polarization matrix (9.93) in this
case has only one non-zero component, ρ00 . From Table C.1 one finds the detector
multipole moments:
⊥ ⊥
P00+ = 1 and P20+ = −2. (9.162)
All others disappear and the signal detected with such polarization – parallel to the
z(at) , propagating in the x (at) y (at) plane – is thus
1 1
R⊥ = ⊥
1 + g2 T20+ P20+ = [1 − 2g2 T20+ ]. (9.163)
3 3
9.4 Angular Distribution and Polarization of Radiation 621
Angular Distribution
It is important to note at this point that in all above discussion the polarization
(pol)
parameters Pq as well as the atomic multipole moments TKQ± were referring to
the same coordinate frame. In the above discussion we have called it “atomic frame”
(at) x (at) y (at) z(at) which was identical to the photon frame. Detection always occurs
in the photon frame. Thus, we now mark the multipole moments in (9.157) by the
(ph)
superscript TKQ± to remind us of this fact. In the most general case the detector
points in an arbitrary direction θk , ϕk , thus defining the x (ph) y (ph) z(ph) photon frame
(with the wave vector k z(ph) ). We now allow this frame to be different from the
atomic frame (for an illustration see Fig. 4.3 in Vol. 1). Thus, to eventually extract
(at)
multipole moments TKQ+ which describe the atom in a most convenient “at” frame,
we have to rotate the coordinate system by (C.13).
For the total emission rate this is relatively easy to achieve, since (9.158) in the
(ph)
photon frame contains only the zero component TK0+ of the atomic alignment ten-
sor. And that relates to the multipole moments of interest (in the atomic frame)
corresponding to (C.14):
(ph)
K
(at)
TK0+ = TKqp CKq± (θk , ϕk ).
q=0,p=±
Inserting this into (9.158) and using the real spherical harmonics CKq± as tabulated
in Table D.1, Vol. 1, we obtain
6 7
2 (ph) 2
2
(at)
R(θk , ϕk ) = 1 + g2 T20+ = 1 + g2 C2q± (θk , ϕk )T2qp
3 3
q=0,p=±
2 1 (at)
= + g2 3 cos2 θ − 1 T20+ (9.164)
3 3
√ (at) (at)
+ 3 sin2 θ cos 2ϕT22+ + sin 2ϕT22−
√ (at) (at)
+ 2 3 sin θ cos θ cos ϕT21+ + sin ϕT21− .
is independent of the polar angle θk as long as it is emitted with k ⊥ x (at) and po-
larization elin x (at) . Thus, emission of this polarization within the x (at) y (at) plane
(pol)
will also be described by (9.157) with P1 = 1:19
1 √
R =
(at) (at)
1 + g2 T20+ − g2 T22+ 3 . (9.165)
3
With (9.163) and (9.165) one may define an extra S TOKES parameter for the linear
polarization detected in the x (at) y (at) plane (light propagating in the y (at) -direction).
(at) (at) √
R − R⊥ 3g2 T20+ + g2 T22+ 3
P4 = = √ . (9.166)
R + R ⊥ 2 − g2 T (at) − g2 T (at) 3
20+ 22+
(at) 1 2P4 + P4 P1 − P1
T20+ = . (9.167)
g2 3 + P 1 + P 4 − P 1 P 4
With this and Rz from (9.158) we can derive from (9.159) and (9.160) similar ex-
pressions for the other two alignment parameters:
(at) P3 (at)
T10+ = − 1 + g2 T20+ . (9.169)
g1
As one easily sees from the tabulated values Table C.1 of the multipole moments,
(at)
T10+ is simply the expectation value of the atomic angular momentum with respect
(at)
to the z(at) -axis. With reference to Fig. 9.5, in contrast T20+ (which for a pure L = 1
state is 1 − 3ρ00 ) reflects essentially the height of the atomic charge cloud, the latter
(at) (at)
being ∝ ρ00 . Finally, from the alignment parameters T22+ and T22− one recovers
the width and the alignment angle of the charge cloud. We leave it to the interested
reader as a simple exercise to derive the parameters of the density matrix of the atom
19 Of course this can be verified explicitly with some careful effort by rotating (E.25)–(E.28), Vol. 1
through the E ULER angles α = ϕk = −π/2 and βk = π/2, using the expressions given in (E.13),
2
Vol. 1 for the d±20 2
and d±2±2 matrices.
Acronyms and Terminology 623
(at)
(e.g. for a pure L = 1 state) by comparing the TKQ± via Table C.1 with (9.132).
Note, that the S TOKES parameters used there to describe the atom relate to the
directly measurable quantities Pi discussed here by Pi /(gK GK ).
Section summary
• Based on the concepts of FANO and M ACEK (1973), we have derived the
general theory for the angular and polarization distributions Re (θ, ϕ) of light
emitted from anisotropically and incoherently populated excited states.
• Starting from the dipole transition operators D† = r · e∗ for emission of a
photon with polarization e we define a detector matrix ρ̂ (det) = D†
D. The
(det)
expected signal is then given by Re (θ, ϕ) = Tr(ρ̂ ρ̂ ), where ρ̂ describes
the excited system.
• In order to disentangle geometrical and dynamical parameters in such an ex-
periment both density matrices are expanded into series of state multipoles of
rank K, with 0 ≤ K ≤ 2 for ρ̂ (det) in single photon detection.
• The detector matrix ρ̂ (det) , originally described in a [(j j¯)1(j¯j )1](K) cou-
pling scheme has to be recoupled into a more transparent and convenient
[(j j )K(j¯j¯)0](K) coupling scheme for the polarization analyzer ρ̂ (pol) .
• The final result (9.154) for Re (θ, ϕ) is expressed in terms of multipole mo-
ments (constructed from angular momenta) for both, the atom and the polar-
ization analyzer.
• Detailed evaluation in terms of S TOKES parameters and angular distribution
of the total intensity leads to a clear disentanglement of experimental geome-
try from the atomic alignment and orientation parameters to be determined.
References
A NDERSEN , N. and K. BARTSCHAT: 2003. Polarization, Alignment and Orientation in Atomic
Collisions. Berlin, Heidelberg: Springer.
624 9 The Density Matrix – A First Approach
Overview
To set the stage, in Sect. 10.1 we take a look onto an experiment from mod-
ern quantum optics. In Sect. 10.2 the important “dressed state” model is in-
troduced to analyze the two level system in a quasi-monochromatic light.
Section 10.3 presents several characteristic experiments which may be ex-
plained effortless with this model. Section 10.4 derives the theoretical frame-
work for treating such systems quantitatively, starting with the fundamen-
tal L IOUVILLE - VON -N EUMANN equation from which the “Optical B LOCH
equations” are derived. In Sect. 10.5 we apply these to a number of important
questions which in earlier chapters could only be discussed by hand waving
arguments. On these grounds, Sect. 10.6 develops some basics of short pulse
spectroscopy. Finally, Sect. 10.7 introduces a somewhat more complex appli-
cation, the STIRAP method, today of increasing interest also in the context of
“quantum information”.
count rate / ms
to vacuum pump 4
photon
Rb-dispenser
2
objective
dichroic
(NA = 0.38) 0
mirror
10 20 30 40
measuring time / s
Si avalanche
diode
laser
beams
MOT for cooling
coils
from dipole trap laser single mode glas fibre
Fig. 10.1 Experimental setup according to W EBER et al. (2006) for measuring the 2-photon cor-
relation function for a single atom in a MOT trap
RL
F =1 5 2S1/2 F = 2
0.0
- 1.0 - 0.5 0.0 0.5 1.0
Δν / MHz
signal shown in Fig. 10.3(b) gives clear evidence of anti-bunching: if one photon
has been emitted, the stored atom is found in the ground state and it takes some time
until it has a high probability for being found in the excited state again. Obviously,
this probability oscillates between a maximum and a minimum. Later in this chap-
ter, we shall develop a quantitative understanding of these damped oscillations, the
so called R ABI oscillations.
In Fig. 10.4 the corresponding experimental setup (top) for the spectral analysis
of the fluorescence (bottom) is shown. One uses an FPI, scanned by a piezo. For
comparison the laser linewidth is studied simultaneously (red line). Both profiles
are much narrower than the natural linewidth of the atom (6 MHz). At the very
low intensity used in this particular experiment one simply observes R AYLEIGH
scattering of the incident light. As only one atom is captured in the trap at a given
time, many individual measurements have to be carried out to obtain these nice data.
The thus calibrated atoms, albeit very cold, still have a finite temperature which
leads to a measurable D OPPLER broadening of the fluorescence line. The authors
derive from the linewidth an atomic temperature of 105 µK.
In view of experiments discussed below, we have to point out that the extremely
narrow line profile shown in Fig. 10.4 is only observed at extremely low laser inten-
sities. As soon as the level shift due to the dynamic S TARK effect (see Sect. 8.4.1 in
Vol. 1) gets into the order of the natural linewidth, the picture changes completely –
as we shall see below.
Section summary
• Perturbation theory can only describe optical excitation at low intensities. It
fails, if the excited state probability becomes substantial. Also, it cannot ad-
dress questions about the spectral distribution of light emitted from a system
excited with narrow band lasers.
10.2 Two Level System in Quasi-Monochromatic Light 629
The two level system is an extremely simple and useful approximation for
understanding the temporal evolution of quantum systems in intense, quasi-
monochromatic, nearly resonant radiation fields. We have already used this concept
in Sect. 2.3.5 (see Fig. 2.18). Here we shall demonstrate that it provides qualitative
and quantitative insight into the key processes. The following questions will have to
be addressed – without recurrence to perturbation theory: How does the population
and coherence of a quantum system with two levels develop in time if it is exposed
to a strong, nearly resonant electromagnetic radiation field? How does it depend on
the detuning from resonance? And which radiation is re-emitted?
From Sect. 2.3.5 we recall some basics about the fully quantized description of tran-
sitions induced by a single mode electromagnetic field. We distinguish the H AMIL -
TON operator HA of the atom, H
F of the field and the interaction U
. With the energy
ωγ of the unperturbed atomic state |γ and the energy N ω of a photon state |N
for N photons in the mode,
A |γ = ωγ |γ and H
H F |N = N ω|N
0 |γ N = Wγ N |γ N
H
0 = H
with H A + H
F and Wγ N = ωγ + N ω.
Neglecting for a moment the interaction of the field with the atom, we have for a
system with two levels γ = a, b
H0 b(N − 1) = ωb + (N − 1)ω b(N − 1)
0 |aN = [ωa + N ω]|aN .
H
In Fig. 10.5 the energy levels with different photon numbers N are sketched for
nearly resonant excitation ω
ω, the detuning being defined as
ħω
|b, N +1〉
ħΔω
|a, N+2〉 ~~ ~
~
ħωba
ħω
|b, N 〉
ħΔω
|a, N+1〉
~
~ ~
~
ħωba
ħω
|b, N-1〉
ħΔω
|a, N 〉
The states are grouped in pairs, slightly split as a consequence of the detuning.
In principle this ladder may be extended below and above the levels shown. For
clarity and without loss of generality we choose the energy of the state |aN as zero
energy. But as we shall see, only two such pairs of levels need to be considered for
describing the two level system in a nearly resonant, oscillating field. With Fig. 10.5
in mind, the term “dressed states” has been coined – evoking the image of a quantum
system dressed by photons.
|aN = i e
bN − 1|U Dba E0 (10.2a)
2
|bN = i e
aN − 1|U Dab E0 (10.2b)
2
|bN = − i e
aN + 1|U D∗ E 0 (10.2c)
2 ba
|aN = − i e
bN + 1|U D∗ E 0 . (10.2d)
2 ab
10.2 Two Level System in Quasi-Monochromatic Light 631
In a genuine two level system, only one lower substate |a = |γa ja ma and one
upper substate |b = |γb jb mb must be involved. The transitions are thus induced
by radiation with a basis polarization vector e = eq with q = mb − ma . The relevant
transition matrix element is then r ba · e = rq (ma mb ) as defined in (4.77), Vol. 1,
which in this particular case (two level system) is also identical to |r ba | according
to (4.79), Vol. 1.1
We now introduce the (resonant) R ABI frequency (we mention in this context the
N OBEL prize for R ABI 1944)
e
Dba E0 e|r ba |E0
ΩR = = , (10.3)
proportional to the dipole transition matrix element |r ba | between the two levels and
to the electric field strength E0 . For quantitative comparisons it is useful to relate
|r ba | to the E INSTEIN coefficients Aab and Bba , (2.151):
4α 4h 1
Aab = 2
|r ba |2 ω3 = 3 Bba = . (10.4)
3c 3λ T1
Here we call the natural lifetime of the excited state T1 . We must emphasize that
these relations hold in this special form only for a genuine two level system. R ABI
frequency and intensity I of the laser field are related to each other by
For the two level system we shall now apply the rotating wave approximation
(RWA) which we have already introduced in Sect. 2.3. It simplifies the problem
significantly: only the near energy resonant processes (10.2a) and (10.2c) are taken
into account, i.e. induced emission and absorption of a photon. As we have seen in
the framework of perturbation theory this neglects the fast oscillating terms which
average out more or less perfectly when the time dependent S CHRÖDINGER equa-
tion is integrated.2 While deriving in the following a non-perturbative solution, it is
1 Typically, a two level system may be realized for jb = ja + 1 with maximum or minimum pro-
jection quantum numbers by the states |γa ja ±ja and |γb jb ±jb with circularly polarized light,
q = ±1, for excitation. A classical example is the 3 2 S1/2 F = 2 MF = 2 ↔ 3 2 P3/2 F = 3 MF = 3
transition, populated by optical pumping (see Appendix D).
2 However, when studying processes of higher order (e.g. multi-photon processes), the application
useful to keep in mind that inserting the electric field strength into the matrix ele-
ments – as done in (10.2a)–(10.2d) – implies a strictly single mode, monochromatic
radiation field (in contrast to the previous treatment in Sect. 2.3.6 where we have
assumed a broad-band, multi-mode radiation field). This is certainly permissible if
the bandwidth ω1/2 of the radiation field used is small enough, i.e. much narrower
than the natural linewidth 1/T1
ω1/2 .
One may approach the remaining problem in two different ways:
With this ansatz one obtains, in analogy to (2.142) and (2.143), a set of differen-
tial equations (independent of N )
1
ċb = + ΩR ca eiωt
2
(10.7)
1
ċa = − ΩR cb e−iωt
2
where ω is the detuning according to (10.1). The solutions ca (t) and cb (t)
are the time dependent probability amplitudes of the upper, |b(N − 1), and
lower, |aN states, respectively. One finds that they are exponential functions
which contain the fast temporal oscillations corresponding to ωb + (N − 1)ω
and ωa + N ω, respectively, and in addition some side bands.
2. Somewhat more compact one takes the “dressed state” picture, Fig. 10.5, literally
and writes the Hamiltonian H =H A + HF + U with the interaction in matrix
form. The state |aN defines the energy zero:
0 −iΩR
H= . (10.8)
2 iΩR 2ω
With this not only the lowest pair of states is described. Rather, for all other pairs
of states the same energy operator acts, raised by ω for each additional level
pair. In between these states no coupling exists if only single photon processes are
considered.3 The time dependent S CHRÖDINGER equation reads in the rotating
wave approximation
dc i
=− H c, (10.9)
dt
3 This will be different as soon as another photon is involved which is absorbed or emitted.
10.2 Two Level System in Quasi-Monochromatic Light 633
where c(t) stands for a vector constructed from the time dependent probability
amplitudes c1 (t), c2 (t) . . . . In this representation the fast oscillations drop out
and only the physically relevant frequency differences are displayed. The differ-
ential equations are now
1
ċb = + ΩR ca − iωcb
2
1
ċa = − ΩR cb . (10.10)
2
They are – apart from the fast oscillations of which only iωcb reminds us –
fully equivalent to (10.7).
Before entering in Sects. 10.4 and 10.5 into details of the temporal dynamics, us-
ing the so called optical B LOCH equations, we want to have a somewhat closer
look at the energetics. The H AMILTON operator (10.8) with the interaction term
represented by the R ABI frequency ΩR , may readily be diagonalized to explore the
energetic shift due to the interaction with the electromagnetic field. As usual, the
secular determinant must disappear:
−W − iΩ R
det(H 1) = iΩR
− W 2 = 0. (10.11)
2 ω − W
W± = (ω ± Ω ) (10.12)
2
The thus identified energy shifts hold for each of the level pairs sketched in Fig. 10.5.
In Fig. 10.6 this is illustrated for the level pairs |bN , |aN + 1 and |bN − 1, |aN .
Thus, both level pairs split in the field (beyond ω), the total slitting being
Ω . One occasionally calls this splitting in the field R ABI splitting, which is pro-
portional to the field for resonant excitation (ω = 0). The additional splitting is a
consequence of the coherent superposition of the two initial states in each pair |bN ,
|aN + 1 and |bN − 1, |aN , respectively. The new basis states in the diagonalized
634 10 Optical B LOCH Equations
|+'〉
|b, N 〉 (ΩΔ+Δω)/2
Δω ΩΔ
|a, N +1〉 (Ω Δ- Δω)/2
~ ~
~ ~ |-'〉
~
~ ~
~ ~
~
ω ~
~
ω
ωba ω
ω −ΩΔ ω+ΩΔ
|+ 〉
|b, N-1〉 (ΩΔ+Δω)/2
Δω ΩΔ
|a, N 〉 (Ω Δ- Δω)/2
|-〉
Fig. 10.6 “Dressed states” – for illustration of the four energy levels arising due to R ABI split-
ting from the unperturbed states |a N , |b(N − 1), |a(N + 1) and |bN of the two level system.
Excitation occurs with a frequency ω which is detuned by ω = ωba − ω in respect to the reso-
nance frequency ωba . The resonant R ABI frequency ΩR and detuning ω add geometrically to the
non-resonant R ABI frequency Ω = ω2 + ΩR2 which determines the splitting. The emitted ra-
diation forms a so called M OLLOW triplet, as indicated by red arrows, with the angular frequencies
ω − Ω , ω and ω + Ω
system are obtained from (10.8) with (10.12) by Hc = W± c and are4
|+ = −i sin Θ|aN + cos Θ b(N − 1) (10.14)
|− = cos Θ|aN − i sin Θ b(N − 1) (10.15)
with
cos Θ ΩR
= . (10.16)
sin Θ Ω − ω
In the limit of vanishing laser field the amplitudes are cos Θ → 1 and sin Θ → 0,
while for very high fields (high√ R ABI frequencies) or exact resonance the limiting
values are cos Θ = sin Θ = 1/ 2, i.e. both states contribute with equal amplitude to
the new basis states. The original “ground states” |aN or |a(N + 1) then become
|− and |− , respectively, while |b(N − 1) and |bN are transferred into the linear
combinations |+ and |+ , respectively, as indicated in Fig. 10.6.
From this energy splitting, illustrated in Fig. 10.6, one may already understand
the characteristic spectral profiles which such a “dressed atom” emits. This is in-
dicated in the figure by red marked double arrows and the dashed single arrows
(side bands): one expects from this picture that the fluorescence of a near resonantly
4 The phase factors −i can be traced back to our choice of the phase in the definition (2.87) of the
electric field.
10.3 Experiments 635
Section summary
• The “dressed state” model Fig. 10.5 pictures the energies of a two level quan-
tum system in a quasi-monochromatic, nearly resonant electromagnetic radi-
ation field as a ladder of levels separated by the photon energy ω. Each level
is split into a doublet which – in the limit of low intensity – is split by the
detuning ω = ωba − ω of the radiation from resonance ωba .
• The interaction (10.2a)–(10.2d) of the field with the atom is characterized
by the resonant R ABI frequency ΩR = e|r ba |E0 /. Eigenvalues (10.12) and
eigenstates (10.14)–(10.15) in the dressed state model are derived in the rotat-
ing wave approximation RWA.
• This leads to quantitative predictions about the emitted fluorescence radiation.
One expects a so called M OLLOW triplet as illustrated in Fig. 10.6, arising
the ground and excited state doublets, each with a splitting due to Ω =
from
ω2 + ΩR2 , the non-resonant R ABI frequency.
10.3 Experiments
Pioneering experiments have been performed by Herbert WALTHER and his collabo-
rates 1974 and 1976, followed by G ROVE et al. (1977). They studied the prototypical
3 2 P3/2 F = 3 ↔ 3 2 S1/2 F = 2 transition in Na atoms, which may be considered a
genuine quasi two level system. Its preparation is described in Appendix D in some
detail. A typical experimental setup is illustrated schematically in Fig. 10.7. Key
laser from
back
photo-
multiplier
636 10 Optical B LOCH Equations
Fig. 10.8 M OLLOW triplet in resonant excitation at different power P of the exciting laser accord-
ing to H ARTIG et al. (1976). (a) Experimentally observed spectra (base line are shifted vertically
for each measurement); (b) splitting Ω of the triplet in units of the natural linewidths γ as function
of P
elements are a well collimated Na atom beam, a stable, narrow band CW dye laser,
and an FPI for analyzing the spectrum of the fluorescence.
Figure 10.8(a) shows the observed spectrum of the resonance fluorescence for ex-
actly resonant irradiation (ω = 0) at different laser intensities. M OLLOW (1969)
has predicted this triplet, named after him, and provided a quantitative theory –
first semiclassically, later on fully quantum mechanically. Both calculations essen-
tially lead to the same result (M OLLOW 1975). Quantitatively the splitting Ω as a
function of the laser power P is shown in Fig. 10.8(b). From our above discussion
one expects that Ω equals the R ABI frequency in the resonant limit and should be
Fig. 10.9 M OLLOW triplet according to G ROVE et al. (1977) for three different values of detuning
ν – the intensity was I = 640 mW cm−2 in all three cases. (a) Theory with pronounced elastic
(R AYLEIGH) scattering in the case of significant detuning. (b) Comparison of experiment (black)
and the theory (red), the latter has been convoluted with the experimental profile
10.3 Experiments 637
√ √
proportional to I ∝ P as stated (10.5). And this is exactly what is observed
experimentally as documented in Fig. 10.8(b).
H ARTIG et al. (1976) have also studied already the non-resonant case. While the
theory predicts full symmetry of the fluorescence around the irradiating laser fre-
quency, the first experiments showed some asymmetric profiles. However, G ROVE
et al. (1977) were able to show that a very clean preparation of the two level sys-
tem and accounting for geometrical effects does lead indeed to fully symmetric
M OLLOW triplets, both in the resonant as well as in the non-resonant case. In these
experiments the laser frequency has been stabilized with the Na resonance fluores-
cence as reference. As documented in Fig. 10.9 one observes excellent agreement
between the measured spectrum of the fluorescence and the theory – which has been
properly convoluted with the experimental profiles.
The observation of the M OLLOW triplet confirms nicely the predictions about the
resonance fluorescence of a two level system for quasi resonant excitation which we
have gleaned from Fig. 10.6. Nevertheless, this is not yet a direct proof of a splitting
of the originally non-degenerate levels into two terms each. However, the so called
AUTLER -T OWNES effect gives direct evidence for it, by using two lasers and three
levels. The concept is sketched in Fig. 10.10. An intense pump laser “A” irradiates
and atom with a fixed frequency which is nearly in resonance with the transition
between states |a and |b. It thus generates the level splitting (e.g. between the two
hyperfine levels in the case of the 3 2 S1/2 and 3 2 P3/2 levels in Na discussed above).
The weaker probe laser “B” is tuned such that it may induce a transition from the
upper state |b to a third, higher lying level |c. The population of the intermediate
levels |b is determined by the fluorescence detected from level |c (in the present
example the F = 4 level in the 3 2 D5/2 state of Na).
G RAY and S TROUD (1978) have indeed measured this splitting of the |b state
as documented in Fig. 10.11. Figure 10.11(a) shows the measured excitation prob-
ability |b of the intermediate level as a function of the detuning of the probe lases
with respect to the transition |c ← |b in a weak laser field “A”. One sees a clear
splitting of the state |b, which grows with the intensity of the pump laser “A”.
(probe) σ+
fluorescence
3 2P3/2
F' = 3, M' = 3
|b >
σ+ laser A
(pump, splitting)
3 2S1/2 F = 2, M' = 2
|a >
638 10 Optical B LOCH Equations
excitation probability
IA = 5.3 (a) (b)
60
IA = 470 20
0
-50 0 50 0 5 10 15 20 25
Δν B / MHz IA1/2 / (mW cm-2)1/2
Fig. 10.11 Experimentally observed AUTLER -T OWNES effect (see scheme Fig. 10.10) accord-
ing to G RAY and S TROUD (1978). (a) Fluorescence spectra with resonant irradiation at differ-
ent intensities IA /(mW cm−2 ) as a function of the detuning of√the probe laser νB (intensity
IB = 15 mW cm−2 ); (b) measured line distance as a function of IA
ΔνB / MHz: - 60 0 60
Section summary
• Experimental observations of the resonance fluorescence in an intense, nearly
resonant, quasi-monochromatic laser field confirm the predictions of the
dressed state model – both for resonant and non-resonant excitation.
10.4 Quantum Systems in Strong Electromagnetic Fields 639
• A direct experimental proof of the level splitting into a doubled – due to in-
teraction with the electromagnetic radiation field – is obtained in a three level
scheme, with a pump and a probe laser as illustrated in Fig. 10.10. When tun-
ing the probe laser two resonances are observed in the fluorescence from the
third level: this observation is called AUTLER -T OWNES effect.
∗ = H
Accounting for the Hermiticity of the H AMILTON operator H † we write the
time dependent S CHRÖDINGER equation
∂|ψ(t) ∂ψ(t)|
i ψ(t)
=H or − i = ψ(t)H. (10.17)
∂t ∂t
With this, we derive the time dependence of the density operator (9.15) as
∂ ρ̂ ∂
i = i pα α(t) α(t)
∂t ∂t
∂|α(t) ∂α(t)|
= i pα α(t) + α(t)
α
∂t ∂t
= pα Hα(t) α(t) − α(t) α(t)H =Hρ̂ − ρ̂ H
, (10.18)
α
∂ ρ̂ , ρ̂].
i = [H (10.19)
∂t
640 10 Optical B LOCH Equations
We apply now again the RWA. For a two level system we use the R ABI frequency
ΩR according to (10.3) to characterize the strengths of the coupling. We just need
to insert the Hamiltonian (10.8) into the L IOUVILLE - VON -N EUMANN equation
(10.19) to obtain without problems the optical B LOCH equations for the two level
system with a detuning ω = ωba − ω, for the moment still without relaxation:
ΩR
ρ̇aa = − (ρba + ρab ) (10.20)
2
ΩR
ρ̇bb = (ρab + ρba ) (10.21)
2
ΩR
ρ̇ba = − (ρbb − ρaa ) − iωρba . (10.22)
2
Only the last of the three equations is complex. Normalization of the density matrix,
ρbb + ρaa = 1, makes one of these equations redundant. With ρab = ρba ∗ this finally
1. Often the electric field amplitude is not constant, e.g. when the system is ex-
cited by a laser pulse such as (1.110). However, as long as the amplitude varies
slowly with time (we recall the SVE approximation) one still may use the
equations just derived and simply replace E0 → E0 h(t) and correspondingly
ΩR → ΩR (t) = ΩR h(t). Note that the envelope function h(t) is dimensionless,
usually with h(0) = 1.
2. We have to account for relaxation processes of various origins. For example, if
spontaneous emission is the cause of relaxation we would have to sum inco-
herently over all modes of the vacuum field, since all empty modes contribute
statistically to spontaneous emission. Both, the excitation probability ρbb as well
as the coherence term ρba are affected.
With quite some effort this may be done to derive a clean theory of relaxation.
For simplicity we choose a much simpler, phenomenological approach. We first
remember that the excited state decays with a natural lifetime say T1 = 1/Aab =
1/Γab . Without external electric field the probability to find the excited state thus
decays as
(0)
ρbb (t) = ρbb exp(−Γab t) (10.23)
(0)
from its original population ρbb at t = 0. In differential form this reads ρ̇bb =
−Γab ρbb . Thus, to account for this decay in our density matrix equations one adds
−Γab ρbb to (10.20) and (10.21). In analogy one adds −Γbb ρba to (10.22) for a phe-
nomenological description of phase relaxation. With these additions we write the
Optical B LOCH equations with relaxation:
10.4 Quantum Systems in Strong Electromagnetic Fields 641
For purely optical decay, the relation between Aab = Γab and Γbb may be gleaned
from the field free limit at very low population density in the excited state. We recall
the origin of the density matrix elements from amplitudes. In the coherent case we
had ρik = ci ck∗ , and for weak excitation ca 1 holds, so that ρba = cb ca∗ cb . Thus
dcb
ρ̇ba = = −Γbb ρba = −Γbb cb =⇒ cb (t) = cb0 exp(−Γbb t)
dt
2
=⇒ ρbb (t) = cb (t) = |cb0 |2 exp(−2Γbb t).
642 10 Optical B LOCH Equations
Note that this only holds, if other relaxation processes can be neglected. In the
general case other mechanisms may be important, such as collision processes within
the system studied, internal relaxation or interaction with an environment (so called
bath). In such a case T2 and T1 may well be independent of each other. For example,
collision processes in the gas phase may lead to shorter T2 (so called dephasing) but
not necessarily to a decay of the excitation. On the other hand, loss processes (e.g.
into a third state not coupled to the field) may dominantly lead to a shortening of T1
without substantial dephasing of the population remaining in states a and b.
Finally, we end these general considerations by noting for later use that the RWA
Hamiltonian (10.8) of the two level system in a near resonant electromagnetic field
with relaxation reads now
(t) = 0 −iΩR (t)/2 + iΓab
H . (10.33)
iΩR (t)/2 + iΓab iΓbb + ω
This is easily verified by inserting H into the L IOUVILLE equation (10.19) and
comparing the result with the optical B LOCH equations (10.29a)–(10.29c).
Section summary
• In the general case of mixed states, the temporal evolution of the population
and coherence is described by the L IOUVILLE equation (10.19) for the density
matrix ρ̂ of a quantum system. Diagonal terms ρii describe the population, the
off-diagonal terms ρij represent the coherence among the states.
• Both can change under the influence of a near resonant electromagnetic ra-
diation field. Inserting the coupling strength (described by the R ABI fre-
quency ΩR ) into the L IOUVILLE equation leads – after some manipulations
and after introducing relaxation terms – to the optical B LOCH equations
(10.24)–(10.26) for a two level system.
• In the case of pure optical decay, the relaxation times for population decay,
T1 , and for decay of coherence, T2 , are related by T2 = 2T1 .
We now consider several special examples. For some of them, the optical B LOCH
equations may be solved in closed form, while for others a numerical treatment is
required. In the present section we set h(t) ≡ 1 for t ≥ 0, and thus study excitation
with continuous light, which is switched on at time t = 0, just as in the semiclassical
and quantized perturbational treatment.
10.5 Excitation with Continuous Wave (cw) Light 643
If the damping is finite, we just need to wait sufficiently long after switching on
the field (t
T1 , T2 ), and all time derivatives disappear, so that (10.29a)–(10.29c)
becomes
1 ΩR2
ρbb = . (10.35)
2 Γab Γbb + (Γab /Γbb )ω2 + ΩR2
In the limit of very high intensities (R ABI frequency ΩR
ω, Γab , Γbb ) this
leads to ρbb → 1/2, i.e. to equal population of ground and excited state – as ex-
pected. This is called saturation as already predicted by phenomenological semi-
classical considerations. We mention that in this case the off-diagonal matrix ele-
ment disappears, since it becomes ρba = ρR + iρI ∝ 1/ΩR . We also note that for
large detuning ω
ΩR the excitation probability decreases as ρbb ∝ 1/ω2 , the
typical Lorentzian behaviour.
For purely radiative decay into the ground state with Γbb = Γab /2 = A/2 the popu-
lation density (10.35) becomes
ΩR2 /4 ΩR2 /4
ρbb = = . (10.36)
(A/2)2 + ΩR2 /2 + ω2 Ωs2 /4 + ω2
Thus, we have derived an expression for the power broadening (or saturation broad-
ening) of spectral lines. Obviously (10.36) describes a L ORENTZ profile with a
FWHM
Ωs = A2 + 2ΩR2 . (10.37)
Figure 10.14 illustrates the saturation profile for various values of ΩR (and thus
intensities).
We recall (10.5), ΩR2 = (3λ3 A)I /(2πhc), the relation between R ABI frequency
ΩR and intensity I , and introduce a saturation intensity (see e.g. A SHKIN 1978)
2πhcA
Is = . (10.38)
λ3
644 10 Optical B LOCH Equations
0.00
0 1 2 3 4 5 I / Is
√
For I = Is the R ABI frequency is 3 times the natural linewidth A. By writing
ΩR2 = 3(I /Is )A2 the average population density (10.36) of the upper state becomes:
1 6(I /Is )
ρbb = . (10.39)
2 1 + 6(I /Is ) + (2ω/A)2
-20 -10 0 10 20
Δω / A
10.5 Excitation with Continuous Wave (cw) Light 645
1. Excitation with a very broad band source (ω1/2
A) whose maximum is tuned
into resonance ωc = ωba . One obtains from (10.40) and (2.20) for the population
density:
3λ3ab I 3λ3 I˜(ωba )
ρbb = = ab . (10.41)
2πh πcω1/2 4h c
Here I˜(ωba )/c = ũ(ωba ) = 2I /(πcω1/2 ) is the spectral radiation density of
the light source at ωba . Note that only the wavelength is specific for the transi-
tion studied and no other atomic properties enter into the relative population of
the excited state. This may appear surprising at first sight. However, it is simply
a consequence of the fact that at stationary conditions spontaneous emission and
absorption balance each other. While the time to reach the stationary state de-
pends on |r ba | – as we shall see in Sect. 10.5.6 – the stationary state itself does
not.
2. The other extreme is a very narrow band laser ω1/2
A. Again we excite in
the line centre (ω = 0) and find in that case:
3λ3ab 2I
ρbb = . (10.42)
2πh πcA
The excitation probability under stationary conditions is obviously inversely pro-
portional to the natural linewidth A of the excited state – and of course propor-
tional to the intensity. One may rationalize this by considering that excitation oc-
curs only in the centre of the line, spontaneous emission, however, which counter
646 10 Optical B LOCH Equations
acts excitation, occurs over the whole natural linewidth A. One could also say that
the irradiation into a spectral region smaller than the natural linewidth is equiva-
lent to a source which has a bandwidth equal to A, such radiation corresponds in
this respect to a spectral radiation density ũ(ωba ) = I˜(ωba )/c = 2I /(πcA).
So far we have completely ignored the temporal evolution of the density matrix
elements – assuming that the quantum system studied was exposed to the radiation
field already for a long time (t → −∞) so that transient responses have already
settled.
A reasonable, first approach to attack the time dependence of the population
densities is to assume that coherences have soon reached their steady state (e.g.
when exciting with a broadband, low intensity source, or when collisions play a
major role). We then set in (10.26) simply ρ̇ba = 0, switch on the light source at
t = 0 and let it have constant intensity thereafter, i.e. h(t) ≡ 1. We then obtain from
(10.26):
−ΩR Γbb
ρba + ρab = (ρbb − ρaa ).
ω2 + Γbb
2
Inserting this into (10.25) and expressing again ΩR2 according to (10.5) by the
linewidth A and the laser intensity I one finally finds:
I
ρ̇bb = −B̃(ω) (ρbb − ρaa ) − Aρbb (10.43)
c
I
ρ̇aa = B̃(ω) (ρbb − ρaa ) + Aρbb (10.44)
c
3λ3 A Γbb 3λ3 A2 /4
with B̃(ω) = = . (10.45)
4πhc Γbb + ω
2 2 2πh A /4 + ω2
2
In the last step we have assumed purely radiative processes with Γbb = Γab /2 =
A/2.5 Often it is appropriate to have an equation for half the inversion probability
ρD = (ρbb − ρaa )/2:
Thus, we have now derived the well known rate equations, which have been
introduced for the first time heuristically by E INSTEIN in the context of deriving the
P LANCK radiation law (see Sect. 4.2.5 in Vol. 1). Expression (10.45) is, however,
5 If the intensity is sufficiently low, so that optical pumping can be neglected, these expressions may
also be applied to transitions with several sublevels involved. One just has to replace the factor 3
in (10.45) by gb /ga (see footnote 2 in Chap. 5, Vol. 1).
10.5 Excitation with Continuous Wave (cw) Light 647
more detailed in as far as it includes now the dependence on ω, the detuning with
respect to resonance for finite absorption linewidth. We could say that we have truly
derived the frequency dependence of the E INSTEIN B coefficient. In Sect. 1.1.4
we had already used these relations, based essentially on guess work. For direct
comparison of (10.43) and (10.44) with the classical rate equation – where a broad
band source is assumed – we have to integrate (10.45) over all frequencies and
replace
I 3λ3 A I˜(ωba )
B̃(ω) → = B ũ(ωba ).
c 4h c
As mentioned on several occasions the relation between B and A differs from others,
often found in the literature, by a factor 3/2π , since we discuses light beams instead
of isotropic sources (factor 3) and b) relate the spectral radiation intensity ũ(ω) to
the angular frequency and not to the frequency (factor 1/2π ).
Such rate equations are often used and describe the behaviour of quantum sys-
tems in the electromagnetic radiation field usually rather well – the basic assumption
being that rapidly oscillating coherence terms have reached their equilibrium. They
even allow to introduce additional terms beyond the two state system, and we have
made use of this possibility in previous chapters occasionally. Clearly, rate equa-
tions cannot be used to describe coherence effects which are observed for narrow
band excitation and are of importance when studying ultrafast processes.
We now take a step into the opposite direction and assume that there is no relaxation
= Γbb = 0). With the resonant and non-resonant R ABI frequencies, ΩR
at all (Γab
and Ω = ΩR2 + ω2 , respectively, the optical B LOCH equations (10.24)–(10.26)
may be solved for initial conditions6 ρbb = 0 and ρab = 0 in analytical form (we use
SWP 5.5 2005):
ΩR2 Ω t
ρbb = 2
sin2 with ρaa = 1 − ρbb and (10.47)
Ω 2
2 2
ΩR Ω Ω t Ω t
ρR = − 2
sin cos
Ω 2 2
2
ΩR ω 2 Ω t
ρI = − 2
sin .
Ω 2
6 These initial conditions imply again that the radiation field is switched on at t = 0.
648 10 Optical B LOCH Equations
0 π 2π 3π
ΩRt
In Fig. 10.15 we have plotted the population density ρbb (t) of the excited state
as a function of time (in units of 1/ΩR ) for three different values of the detuning
ω/ΩR . One sees that the population oscillates between ground and excited state.
These are the so called R ABI oscillations.
For exact resonance ω/ΩR = 0 the whole population which was originally in
the ground state is after a time t = π/ΩR found in the excited state – and returns at
t = 2π/ΩR again completely back into the ground state. If one uses a square wave
pulse, switching the laser field on at t = 0 and off at t = π/ΩR , the system remains
to 100 % in the excited state (as long as we can neglect relaxation, as assumed here).
Such a laser pulse with ΩR t = π is called a π pulse. Correspondingly, a π/2 pulse
generates equal population in the ground and excited state, while a 2π pulse brings
the hole population completely back into the ground state. Of course, the same may
be achieved by two π pulses following each other with some arbitrary time delay.
With increased detuning the R ABI oscillations get faster, as determined by
(10.37) while the maximum excitation probability is reduced.
To obtain a feeling for the time tπ which such a π pulse takes, we choose
again the excitation of Na atoms as an example. Let us assume an intensity of
I 630 mW cm−2 (see Sect. 10.5.2). In this case ΩR 7 × 2π × 10 MHz and hence
tπ = 7 ns. However, since the natural lifetime of the excited state is only 16 ns, we
cannot really describe the situation well without accounting for relaxation. Never-
theless, equations (10.47) may be used as a realistic description for very short times
t
T1 , T2 . At very high intensities as obtained with state-of-the-art short pulse
lasers these equations without relaxation are quite adequate to describe the R ABI
oscillations.
In the general case one usually has to solve the optical B LOCH equations numer-
ically. Only in the special case of exact resonance, ω = 0, the solutions can be
10.6 B LOCH Equations and Short Pulse Spectroscopy 649
given in closed form, again for the initial conditions ρbb = 0 and ρab = 0:
1 ΩR2 3A 3tA
ρbb = 2 /2 + Ω 2
1 − cos Ω x t + sin Ω x t exp −
2 Γab R
4Ωx 4
A2
with Ωx = ΩR2 + .
8
In Fig. 10.16 this solution is shown for ΩR /A = 2 (strong damping) and =10 (weak
damping). It is possible to detect such oscillations experimentally, e.g. by determin-
ing the second order degree of coherence as documented for a single atom in a trap
quite impressively at the beginning of the present chapter in Fig. 10.3.
Section summary
• Using the stationary limit of the optical B LOCH equations (excitation with
a narrow band CW laser) we have derived power (or saturation) broadening
of the natural linewidth A with increasing
radiation intensity I . We find a
√
L ORENTZ profile with a linewidth Ωs = A + 2ΩR , where ΩR ∝ I is the
2 2
R ABI frequency.
• For the temporal evolution of the excited state density after switching on a
narrow band CW laser we find an oscillatory behaviour. Without relaxation,
ground and excited state population oscillate with half the non-resonant R ABI
(angular) frequency, Ω /2 between 0 and 1.
• On resonance, a square wave pulse of duration t = π/ΩR (so called π pulse)
leads to complete population inversion.
When exciting an atom or molecule with an ultrashort laser pulse usually the optical
B LOCH equations cannot be solved in closed form. However, numerical integration
is straight forward. We start the discussion with some preliminary remarks.
650 10 Optical B LOCH Equations
This relation is quite remarkable: it says that for quasi-resonant excitation with short
laser pulses the laser intensity is not the key parameter; rather, it is the product of
fluence (i.e. laser pulse energy per area) and pulse duration which determines the
process.
To obtain some feeling for numbers, we compute ΩR τG for the excitation of
the Na resonance at λ = 589 nm with a laser pulse of 50 fs. A typical pulse energy
might be 0.3 mJ. Focused on a (1/e) diameter of 100 µm this gives a fluence of
F0 4 J cm−2 (at a maximum intensity I0 = 6.3 × 1013 W cm−2 ). With A = 2π ×
10 MHz we obtain a phase angle ΩR τG = 222. This is a large multiple of π , and
we have to question the whole approximation at such intensity. This becomes quite
evident when computing the R ABI frequency ΩR = 222/(50 fs) = 4.44 × 1015 s−1 ,
which is already larger than the angular frequency of the transition studied, which
is 2πc/λ 3.2 × 1015 s−1 . Thus, strictly speaking, we should not use the RWA
and even the SVE approximation becomes questionable – it only holds if the field
amplitude changes little during one optical cycle.
With today’s short pulse lasers one easily enters into an intensity region where
even the optical B LOCH equations may no longer be a good approximation (note,
however, that the example discussed here has an exceptionally high oscillator
strength). In the following considerations we shall therefore assume a significantly
smaller fluence. As long as only radiative relaxation plays a role (nano second time
scale), one may neglect relaxation terms altogether when working with femtosecond
laser pulses. We concentrate for the moment on the nearly resonant case τG ω
1.
Thus, we need only consider the first term on the right hand side of (10.30a) and
(10.30b), and have to solve the following coupled equations:
dρbb
= h(ϑ)ΩR τG ρR (10.49)
dϑ
dρR
= −h(ϑ)ΩR τG (ρbb − 1/2). (10.50)
dϑ
10.6 B LOCH Equations and Short Pulse Spectroscopy 651
ρbb
Á ba = π
∞
1.0
GAUSS
0.8 I(t)
0.4
Á∞ba = 2π
0.2
0.0
-2 -1 0 1 2 3 t /τ
Fig. 10.17 Population of the excited state in a two level system for resonant excitation with a
short laser pulse. The numerical solution of the optical B LOCH equations has been carried out for
∞ . The time scale is calibrated in
different laser pulse fluence, characterized by the phase angle φba
units of the FWHM of the exciting G AUSS pulse (dashed red) and set to zero at its maximum
whole cycle of excitation and de-excitation (2π pulse). Excitation with such pulses
or pulse sequences plays a key role when studying so called photon echoes. They
are standard tools in NMR and EPR spectroscopy.
Typically, experiments in short pulse spectroscopy are performed by pump-probe
methods at sufficiently low laser intensities to avoid such effects: what is needed is
the preparation (pump) of an excited state density in atoms, molecules or condensed
matter which is high enough to be probed conveniently with a second (probe) laser.
Usually one operates far away from saturation, ρbb
1/2. In Fig. 10.18 the popula-
tion is plotted at very large times after the pump pulse (relaxation is neglected) as a
∞ accumulated during the pulse. Note that the pop-
function of the effective phase φba
ulation in the excited state increases for small phase angles, but decreases again for
652 10 Optical B LOCH Equations
ρbb (t→ ∞)
1.0
0.5
0.0
0 0.5 π π Á∞
ba
Fig. 10.18 Population ρbb of the excited state in a two level scheme for nearly resonant ex-
citation with a short laser pulse at negligible relaxation. Shown is the population density at
times t long compared to √ the laser pulse duration as a function of the effective phase angle
∞ = Ω τ ∞ h(ϑ)dϑ ∝ F τ
φba R −∞
∞ ≥ π after the system has undergone half a R ABI cycle. For significantly higher
φba
fluence several maxima and minima can be reached.
For free atoms and molecules with isolated excited states the dissipation constants
Γab = A and Γbb = A/2 describe the (spontaneous) radiative decay completely,
typically on the ns time range. Additional relaxation mechanisms may arise from
the interaction with a bath (gas, surface, liquid). In larger molecules and clusters a
multitude of additional, internal relaxation mechanisms (electronic or nuclear) may
occur. It is a key topic in short pulse spectroscopy to follow and understand this kind
of dynamics. Femtosecond laser pulses are particular useful to study these processes
in pump-probe experiments. One creates two synchronized laser pulses which can
be delayed with respect to each other by using an optical delay line (typically an in-
terferometer of the M ICHELSON or M ACH -Z ENDER type) as sketched in Sect. 1.5.
The first pulse excites the system (pump) the second induces emission or leads to
ionization which allows to probe the process as a function of the variable delay
time t.
The optical B LOCH equations turn out to be very useful when analyzing the
experimentally observed dynamics of quantum systems in the field of short laser
pulses. In such pump-probe experiments one uses the parameters T1 = 1/Γab and
T2 = 1/Γbb to model an experimentally determined transient signal of the system –
with the hope that these parameters allow one to glean information on the character-
istic temporal behaviour of the inherent processes in the system, i.e. non-adiabatic
or adiabatic transitions, dissociation processes, internal ro-vibrational redistribution
etc. Even though, as a rule, such phenomenological numbers on lifetimes may be
a somewhat course approach towards the many underlying processes, the informa-
tion is very valuable. In the following we want to discuss a few special cases, with
emphasis on the use of the optical B LOCH equations and finally present and discuss
some experimental results. For a more detailed introduction into the methods and
10.6 B LOCH Equations and Short Pulse Spectroscopy 653
results of short pulse spectroscopy in the gas phase we refer the interested readers
to a review by H ERTEL and R ADLOFF (2006).
Quite generally, an extension of the two level scheme for which the optical B LOCH
equations in the form (10.24)–(10.26) are originally designed is in principle possible
without problems. If several sublevels (or even other electronic states) are involved,
one just has to add for each interacting pair of sublevels an extra set of equations –
and one must account for all the ‘relaxation channels’ which can empty or fill such
sublevels individually. This is even more straight forward if only rate equations are
treated, as we shall show in the following.
Typical applications of short pulse spectroscopy focus on the fs or ps time domain
so that one may safely ignore radiative decay which occurs on the ns time scale.
Often the dephasing, i.e. the loss of coherence, is much faster than population or
depopulation of the excited state, i.e. Γbb
Γab = 1/T1 . One also has to be aware
that the processes studied are often rather complex and proceed in several steps,
which cannot be resolved individually. Nevertheless, one may account for these
sub-processes phenomenologically by additional rate equations which supplement
the optical B LOCH equations for the two level system. These extra rate equations
describe the population and depopulation of different ‘channels’ (see e.g. L IPPERT
et al. 2003; F REUDENBERG et al. 1996). In addition to the equations describing the
excitation in the sense of (10.30a)–(10.30c) one adds equations of the type
ρ̇kk = ρii Γki − ρkk Γj k ,
j
(
which have to be solved simultaneously. Note that j Γj( k = 1/τk is then the total
decay rate of a channel (k) with the lifetime τk and Γik / j Γj k characterizes the
branching ratio for a transition from channel k to channel i. The observed signals
(ions, electrons, fluorescence) are considered to be proportional to the population
densities of the states involved, ρbb (t), ρii (t), ρkk (t), as they evolve with time. Fi-
nally, the thus modelled signals have to be weighted with the temporal profile of the
probe pulse and must be added (possibly with different specific weights) to the total
signal which is finally observed. These weights, decay rates and branching ratios
are then used to obtain a fit to the experimental data which provides optimal agree-
ment. Even if such models cannot serve to obtain an accurate picture of all processes
which occur in a molecule, they allow nevertheless a good comparison among dif-
ferent species and processes, they allow to determine appropriate time scales for the
dynamics within the system, and they provide measurements which can possibly be
compared to theoretical predictions derived from more detailed models.
Limiting Cases
For a genuine pump-probe experiment one wants to avoid saturation phenomena
∞ ∼ Ω τ
1, i.e. close to zero
and thus works with low laser intensity, i.e. with φba R
654 10 Optical B LOCH Equations
in Fig. 10.18. In this case one may even integrate the optical B LOCH equations
(10.30a)–(10.30c) by a perturbative approach in closed form. One obtains for the
depopulation of the ground state (again with ϑ = t/τ )
ϑ
τ 2 ΩR2 t/τ
1 − ρaa (t) = dϑ h ϑ e−τ Γbb ϑ +iτ ω·ϑ dϑ h ϑ e+τ Γbb ϑ −iτ ω·ϑ
4 −∞ −∞
+ c.c., (10.53)
which for a long-lived excited state is identical to ρbb (ϑ). Clearly, contributions to
this integral arise only during the time of the pulse duration τ . For exact resonant
excitation ω = 0 one may evaluate the double integral (10.53) even further by par-
tial integration (F REUDENBERG et al. 1996) and can now distinguish two limiting
cases:
• τ Γbb
1: the dephasing time is much larger than the pulse duration. The expo-
nential expression in (10.53) becomes then 1 and we have as coherent limit
τ 2 ΩR2 t/τ 2
1 − ρaa (t) = dϑ h ϑ (10.54)
4 −∞
1 √ 2 t→∞ π
= πτ 2 ΩR2 erf t/(τ 2) + 1 −→ τ 2 ΩR2 ,
8 2
√
where the error integral erf(t/(τ 2)) determines the temporal behaviour of the
signal.
• τ Γbb
1: Conversely, if the dephasing time is much shorter than the pulse du-
ration, (10.53) leads to the incoherent limit
τ ΩR2 t/τ 2
1 − ρaa (t) = dϑ h ϑ (10.55)
2Γbb −∞
√
1 √ τ 2 ΩR2 t→∞ 1 π 2 2
= π erf(t/τ ) + 1 −→ τ ΩR ,
4 τ Γbb τ Γbb 2
which differs significantly from the coherent case.
• For large detuning τ ω ≥ 1 the situation is more complicated. Numerical simu-
lations show that detuning in combination with a fast decay of the excited state
(Γab
1/τ ) makes the process incoherent and in the limit τ ω
1 leads to
maximum excitation at t = 0 (i.e. at the pulse maximum) which is not so in the
fully coherent case.
Excitation of a Quasi-Continuum
Up to now we always have assumed that the excited state |b which is prepared
by the pump pulse is completely isolated as indicated in Fig. 10.19(a). However,
as a rule in a large molecule or cluster one has to account for a multitude of ex-
cited vibrational and rotational states, especially so in higher electronic levels. To a
10.6 B LOCH Equations and Short Pulse Spectroscopy 655
first approximation one may model this by an infinite number of states with differ-
ent detuning, as sketched in Fig. 10.19(b). With some mathematical effort (see e.g.
H ERTEL and R ADLOFF 2006) one may apply the above formalism just derived also
for the situation sketched in Fig. 10.19(b). One obtains an expression equivalent to
(10.55) and may thus draw the following important conclusion: the coherent excita-
tion of a quasi-continuum of states may be treated in the same way as the incoherent
limit of a pure two level system. This is a very practical result for pump-probe exper-
iments in very large molecular systems. It justifies indeed the rate equations which
typically are used to evaluate such experiments.
Experimentally the cases may well be distinguished if the zero of the delay time
is well known (this is, however, not a trivial requirement). The density of the exited
state is plotted in Fig. 10.20 according to (10.54) and (10.55) for the case of a long-
lived excited state |b with Γab → 0 and ρbb (t) = 1 − ρaa (t) for a pump pulse with
a temporal Gaussian profile. Note that in the incoherent case the half maximum of
the signal is reached at time delay zero (with respect to the pulse maximum). In
contrast, in the coherent case this point is reached for 0.385τG = t1/2 , with t1/2
being again the FWHM of the probe pulse. In life experiments this offset is crucial
when one wants to determine relaxation times T1 below the width of the laser pulse.
An illustrative example are experiments with ammonia molecules and clusters
reported by F REUDENBERG et al. (1996). Figure 10.21 shows the transient ion sig-
nals after excitation of the fast decaying Ã(v = 4) state in NH3 . The pump pulse
Bz+ 204 nm
NH3+ non (NH3)3NH4+ Bz+ 200nm
res. NH3+
non resonant (200 nm)
Fig. 10.21 (a) Calibration of the delay in a pump probe experiment with the help of resonant
(204 nm) and non-resonant excitation (200 nm) of NH3 in the fast decaying Ã(v = 4) state. Note
the offset towards positive delay times in the resonant case. The offset is even somewhat larger
for the longer living ND3 molecule as well as for ammonia clusters. The fit curves have been
obtained by numerical integration of the optical B LOCH equations (10.34a)–(10.34c). For these
fits the experimentally determined laser pulse width of 160 fs FWHM was used. Optimal fits are
obtained for a lifetime of 40 fs for NH3 and 180 fs for ND3 . Shown for comparison are cluster ion
signals, resonantly excited by two different wavelengths (204 nm and 200 nm) (data according to
F REUDENBERG et al. 1996). (b) Comparison with benzene (Bz)
has been tuned alternatively to exact resonance (∼204 nm) or in between two reso-
nances at (∼200 nm). The following ionization in this experiment has been achieved
with a probe pulse of 267 nm. The pulse duration in this experiment was ∼150 fs to
170 fs for both, pump and probe pulses. Shown are the experimental ion signals.
The fit curves shown have been derived numerically from the optical B LOCH equa-
tions (10.34a)–(10.34b). With a decay time of 40 fs, a laser pulse duration 160 fs
and a detuning of 1.5 nm in the non-resonant case one obtains excellent agreement
for the NH+ 3 signal. This may be used to calibrate the time zero, if one assumes that
this experimental signal is essentially symmetric around time zero, as predicted by
the calculations. In the case of resonant excitation the signal maximum is shifted
by 70 fs, which is excellently reproduced by the B LOCH fit. Simultaneously fitting
the resonant and non-resonant transients allows to determine the lifetime of excited
NH3 Ã(v = 4) with an accuracy of about ±10 fs, even though the half width of the
laser pulse is much larger. The result agrees well with estimates derived from the
measured linewidth in NH3 Ã (Z IEGLER 1985). For ND3 , which has been excited
with the same wavelength, an effective lifetime of ∼180 fs has been determined.
It should be pointed out that such extremely short excited lifetimes can of course
not be explained by radiative processes. Rather, they must be attributed to fast in-
ternal conversion and dissociation of the system. For comparison we also show in
Fig. 10.21 the transients in benzene and ammonia clusters. The fit curves correspond
to the solution of the B LOCH equations for a resonantly excited two level system.
Obviously the density of excited states is in this case high enough to allow resonant
excitation, but not high enough to generate incoherence according to (10.55).
10.7 STIRAP 657
Section summary
• After CW excitation treated in the preceding section, we have now applied the
optical B LOCH equations to excitation with short pulses. It turns out that the
key parameter for the strength of the excitation is the time integrated phase
√ essentially the product of R ABI frequency and pulse duration
(10.52), i.e.
ΩR τG ∝ |r ba |F0 τG . Note that here the product of fluence (∝ total laser pulse
energy) and pulse duration enters, rather than the intensity!
• The population in the excited state increases for small phase angles, but de-
∞ ≥ π after the system has undergone half a R ABI cycle
creases again for φba
as shown in Fig. 10.18.
• If coherence decays very fast one may supplement or even fully replace the
optical B LOCH equations by rate equations.
• As illustrated in Fig. 10.20 incoherent excitation is slightly faster then co-
herent excitation. For a long-lived excited state, half maximum population is
reached (a) for incoherent excitation simultaneously with the laser pulse max-
imum, (b) for coherent excitation at a time delay of 0.327t1/2 (the latter
being the FWHM pulse duration).
10.7 STIRAP
Up to now we have discussed exclusively two level systems in one nearly resonant
laser field – even though we have included loss channels by rate equations. In the
following we show how this limitation can be overcome. The B LOCH equations
have shown us that a nearly resonant laser field generates a coherent superposition
of lower state |a and excited state |b. This implies a periodic population change
between the two levels as illustrated in Fig. 10.15. For exactly resonant radiation
the system may temporally be found to nearly 100 % in the excited state. Relax-
ation processes (e.g. spontaneous emission) destroy coherence and damp out the
population oscillations as illustrated in Fig. 10.16. For still longer times finally the
rate equations as proposed by E INSTEIN lead to the well known equilibrium popu-
lations of lower and upper states with ρaa ≥ ρbb – which in the limit of very high
intensity can lead at most to equal population.
It is now interesting to consider for a multilevel system, how one may transfer
the initially very high population of the excited state into a third level – by a judi-
cious sequence of several laser pulses. One possible scheme is the so called lambda
configuration (Λ) as sketched in Fig. 10.22. (Alternatively, in a so called ladder
configuration, the third state lies above the second.)
A typical example would be the population transfer between vibration-rotation
states of a molecule. Assume |1 to be a low lying and |3 a higher lying vibration-
rotation state in the electronic ground, while |2 is an electronically excited state.
The goal is now to shift as much population as possible into level |3, using level
658 10 Optical B LOCH Equations
The result is very plausible, so we refrain from deriving it explicitly. For clarity we
have not introduced here any damping terms. The R ABI frequencies for the pump
and the S TOKES laser are now
e|r 21 |E0 e|r 32 |E0
ΩP (t) = hP (t) and ΩS (t) = hS (t) (10.57)
7 Note
that due to our definition (1.35) for the electric field of the radiation the off diagonal matrix
elements differ from those of B ERGMANN et al. (1998) by an (insignificant) phase factor ±i.
10.7 STIRAP 659
with the envelopes hP,S (t) of the respective field amplitudes. Note that for the
STIRAP process the time dependence of the R ABI frequencies plays a key role.
In Λ configuration one defines the detuning
If damping plays a decisive role – and this is often the case in practice – one has
to add the corresponding damping terms to the Hamiltonian (10.56), in analogy to
(10.33). Using the L IOUVILLE equations (10.19), one easily derives the coupled
linear differential equations for the temporal evolution of the three level system.
It allows one to describe the dynamics as well as the losses during the population
transfer. We discuss here only the case without damping. It is easier to analyze by
solving the coupled equations (10.9). For details we refer to F EWELL et al. (1997).
Here we only discuss the most important results and one example.
In analogy to the two level system, one has to treat now a coherent superposi-
tion of the three levels |1, |2 and |3. If one follows B ERGMANN et al. (1998) and
assumes exact two-photon resonance (i.e. ωP = ωS ) one may verify by diago-
nalization of the Hamiltonian (10.56) that the (time dependent) eigenvalues of the
three dressed states are given by
ω+ = ωP + ωP2 + ΩP2 + ΩS2 , ω0 = 0 and (10.59)
ω− = ωP − ωP2 + ΩP2 + ΩS2 .
|a-〉 – |3〉
|a+ 〉
Θ
|2〉
| ψ〉
|a0 〉
|1〉 losses
Fig. 10.23 Schematic representation of the states for the three level system in H ILBERT space:
|1, |2 and |3 correspond to the basis vectors in the unperturbed system, |a + , |a 0 and |a −
those in the laser fields; |ψ is the state vector of the total system. By variation of the mixing angle
Θ one may in principle transfer |1 → |a 0 and |a 0 → |3 without involving state |2 which is
subject to losses
and Φ being known function of detuning and R ABI frequencies, here without further
interest.
The three new basis states |a + , |a 0 and |a + of the system with laser field are –
as the original basis – again orthogonal. Of particular interest is the state |a 0 which
obviously may be changed continuously from |1 to |3 by choosing the appropriate
mixing angle. The latter can be adjusted to values between Θ = 0 and Θ = π/2
depending on the intensity of the two laser fields and their temporal evolution. This
is illustrated schematically in Fig. 10.23. The optimal temporal sequence of the laser
pulses for this population transfer and the corresponding temporal evolution of the
mixing angle (10.61) is sketched in Fig. 10.24(a) and (b), respectively. The temporal
evolution of the energies of the three basis states in the field (for one and two-photon
resonance) is shown in Fig. 10.24(c).
As already mentioned, amazingly the realization of state |a 0 requires that at
the beginning of the process (region I) first the S TOKES pulse must rise and conse-
quently also ΩS (t) – it couples states |2 and |3! Conversely, at a later time (region
III) the S TOKES pulse must decrease before the pump pules. During the time in be-
tween (region II) where the three states are clearly split due to the laser interaction
both pulses must overlap.
In short: the interaction in the system must start with the S TOKES laser shortly
before the pump intensity rises, however, it must end before the pump laser. Albeit
completely counterintuitive, this scheme is highly effective. The reason is, that for
these conditions state |a 0 consists initially to 100 % of |1 and at the end to 100 %
of |3, as indicated in Fig. 10.24(d). So far, this is just a conceptual idea. Whether in
reality the physical state of the system (in Fig. 10.23 indicated as |ψ(t)) is identical
with |a 0 (more precisely: whether the |ψ(t) follows the basis vector |a 0 adiabat-
ically) requires a detailed, time dependent calculation and, of course, experimental
proof.
10.7 STIRAP 661
energy
|1〉 |3〉
|a 0 as a function of time ħω 0
ħω -
1
|1〉 |3〉
population
(d)
0
I II III
time
Let us first have a look at the experiment. The temporal variation of the interaction
may be detected by two different methods: B ERGMANN and collaborators typically
have used a molecular beam setup as sketched in Fig. 10.25. Here a well collimated
target beam passes two slightly displaced CW laser beams with a Gaussian spatial
profile. The target atoms or molecules ‘see’ effectively a temporal sequence of two
pulses G AUSS pulses. Focussing the laser beams to about 100 µm up to 3 mm (typi-
cal laser powers are some mW to W) one generates an effective “pulse width” in a
D1 D2
preparation if required
signal
target beam
pump laser
Fig. 10.25 Scheme of a STIRAP experiment with a molecular beam and spatially slightly dis-
placed, continuous S TOKES and pump laser beams. The displacement realizes the temporal pulse
sequence. The STIRAP process occurs in the region of overlap; detector D1 registers the fluores-
cence emitted there. Further downstream the final state is excited with a probe laser and detected
with D2 (again by fluorescence)
662 10 Optical B LOCH Equations
588
denoted by |1, |2 and |3
nm
nm
nm
(P indicates the pump and S 145
losses
616
633
the S TOKES laser); state |4 is
used for probing the 140 S
P probe
population transfer
135 2p 5 3s
|1〉 |3〉
VUV
~ ~
~ ~
~
~
1S 1s2 2s2 2p6
0 0
range of 100 ns to µs. The tunable dye lasers used are highly stabilized and thus have
a very narrow bandwidth (1 MHz) – and correspondingly high coherence. The in-
teraction is monitored via the fluorescence of the short-lived intermediate state |2.
The resulting population transfer into the (long-lived) final state |3 is probed down
stream with a further laser tuned to the optically allowed transition between |3 and
a further, short-lived state |4.
Alternatively one may use pulsed ns lasers (in this case fully overlapping) with
pulse energies in the Joule region. The demand for stability is in this case partic-
ularly critical and the pulse duration must be F OURIER limited. Experiments with
ps lasers are in principle also possible, while ultrashort pulses are not appropriate,
in particular due to their broad spectral bandwidth. Over all, experience shows that
the conditions required for the adiabatic passage between the states involved is most
conveniently achieved with CW lasers as described above.
A nice example (B ERGMANN et al. 1998) is STIRAP with excited, metastable
Ne atoms. The term scheme is sketched in Fig. 10.26 (see also Sect. 10.4.2 in Vol. 1,
specifically Fig. 10.9). We remember: the ground state of Ne, the completely closed
rare gas shell, is characterized by 1s 2 2s 2 2p 6 1 S0 , the first excited states have the
configuration 1s 2 2s 2 2p 5 3s and 3p and have an excitation energy of 16.6 to 19 eV.
They correspond to transitions in the VUV. The total angular momentum of the
core j couples to the angular momentum l of the excited electron, which leads
to an angular momentum K, which in turn couples with the electron spin to J =
K ± 1/2. They are thus characterized by (2S+1 Lj )nl 2S+1 [K]J . For optical dipole
transitions we have the usual selection rule J = 0, ±1, with 0 0 forbidden.
Two of the 2p 5 3s states are metastable and are used for the STIRAP process as
initial and final states:8 |1 = |(2 P◦1/2 )3s 2 [1/2]◦0 and |3 = |(2 P◦3/2 )3s 2 [3/2]◦2 . One
generates the metastable Ne in a gas discharge and depopulates state |3 by optical
pumping, i.e. by excitation of |4 = |(2 P◦3/2 )3p 2 [5/2]◦2 with 633 nm, followed by
decay into the two lowest lying J = 1 states, which in turn decay into the ground
8 Weuse here, different from B ERGMANN et al. (1998), the standard terminology according to
NIST.
10.7 STIRAP 663
state. The thus prepared atomic Ne beam in state |1 enters after collimation (to limit
D OPPLER broadening) the STIRAP setup. As intermediate state one chooses |2 =
|(2 P◦ 1/2 )3p 2 [1/2]◦ 1 which decays spontaneously (J = 1) into the two lowest 3s
states with J = 1. In principle, this opens an efficient loss channel. The interaction
region is observed with a channeltron which is a simple and efficient detector for
the VUV emission into the ground state. Downstream, long after the interaction,
one finally probes by laser induced fluorescence the population of the final state |3.
Again the transition |3 → |4 at 633 nm is used and the following VUV emission
from the two lowest lying J = 1 states into the 1 S0 ground state.
In Fig. 10.27 the experimentally observed VUV signal from both detection re-
gions is shown as a function of the detuning of the pump laser. The S TOKES
laser frequency, slightly off resonance, remains fixed in this measurement. Fig-
ure 10.27(a) shows the fluorescence which monitors the population of the inter-
mediate level |2: as long as the two-photon resonance is not realized exactly one
observes large fluorescence losses. The spectral width of the laser induced fluo-
rescence is larger than the D OPPLER width and reflects the power broadening of
the transition |1 → |2. However, if two-photon resonance conditions are reached
(ωP = ωS ) the sharp and pronounced minimum in the loss signal documents
that exclusively the ‘dark’ state |a 0 is populated. In the experiment less then 0.5 %
of the signal without S TOKES laser is observed! This disappearance is all the more
remarkable as the atoms spent more than 20 times the natural lifetime of state |2 in
the interaction region.
Under these conditions the final state |3 is populated by adiabatic passage as
documented impressively by the signal shown in Fig. 10.27(b). The broad back-
ground observed corresponds to a certain fraction which decays spontaneously from
|2 to |3 while the intermediate state is populated (ωP = ωS ).
A quantitative determination of the population transfer is reproduced in
Fig. 10.28. The measurement shows convincingly that nearly complete population
transfer occurs when the S TOKES laser is active distinctively before the probe laser,
664 10 Optical B LOCH Equations
transfer efficiency / %
between S TOKES (S) an
pump laser beam (P) as
reported by B ERGMANN et 50
al. (1998). The upper part of
the figure illustrates
schematically the offset
between the two lasers. 0
Optimal transfer is obtained -1000 -500 0 500
only if (S) interacts before (P)
beam displacement / μm
with the target atom
but overlaps with it. The transfer efficiency decreases to about 25 % if both beams
fully overlap.
The experiments thus proof the effectiveness of the stimulated R AMAN process
and the realization of the adiabatic passage. In view of Fig. 10.24(c) and Fig. 10.23
one thus finds that the state vector |ψ(t) of the system (with |ψ(−∞) = |1),
originally parallel to state |a 0 , follows this state indeed during the whole process
so that at the end of the interaction the population has been transferred completely
and coherently from |1 → |3.
This is not trivial. One may well imagine that during the process transitions can
be induced into the STIRAP states |a + and |a − according to (10.60) and thus state
|2 would be excited – as seen for the case ωP = ωS . The necessary conditions
for true adiabatic passage have been intensively analyzed both experimentally as
well as theoretically. Good overlap between the two pulses as well as sufficiently
strong coupling appears mandatory. B ERGMANN et al. (1998) give as a general
criterium
Ωeff τ > 10, (10.62)
where Ωeff = ΩP2 + ΩS2 is an effective mean R ABI frequency and τ the overlap
time between the two pulses. Only if this condition is fulfilled one may assume that
during the whole process |ψ(t) ∝ |a 0 and that transitions to the other states with
losses can be neglected.
The STIRAP method has been successfully exploited for a variety of atomic and
molecular processes. More recently it has been used even in solid state physics and –
as already mentioned – appears to be a promising road to create ultra cold molecules.
After this brief introduction we shall leave it to the reader to explore this fascinating
subject further.
Section summary
• STIRAP is a very efficient scheme to transfer population from one state into
another in a three level system, using two nearly resonant laser fields, a pump
and a S TOKES laser as illustrated in Fig. 10.22.
Acronyms and Terminology 665
• Contrary to intuition, for a most efficient population transfer the quantum sys-
tem must interact first with the S TOKES laser, while the pump laser follows
and overlaps with it.
• The scheme may quantitatively be understood by a generalization of the op-
tical B LOCH equations. The properties of the three dressed states (10.60) are
schematically illustrated in Fig. 10.24.
• The key to efficient population transfer by “adiabatic passage” is to keep the
system during the whole process in state |a 0 .
References
A LLEN , L. and J. H. E BERLY: 1975. Optical Resonance and Two-Level Atoms. New York: Dover,
2nd edn., 233 pages.
A SHKIN , A.: 1978. ‘Trapping of atoms by resonance radiation pressure’. Phys. Rev. Lett., 40,
729–732.
B ERGMANN , K., H. T HEUER and B. W. S HORE: 1998. ‘Coherent population transfer among
quantum states of atoms and molecules’. Rev. Mod. Phys., 70, 1003–1025.
F EWELL , M. P., B. W. S HORE and K. B ERGMANN: 1997. ‘Coherent population transfer among
three states: Full algebraic solutions and the relevance of non adiabatic processes to transfer by
delayed pulses’. Aust. J. Phys., 50, 281–308.
F REUDENBERG , T., W. R ADLOFF, H. H. R ITZE, V. S TERT, K. W EYERS, F. N OACK and I. V.
H ERTEL: 1996. ‘Ultrafast fragmentation and ionisation dynamics of ammonia clusters’. Z. Phys.
D, 36, 349–364.
G RAY , H. R. and C. R. S TROUD: 1978. ‘Autler-Townes effect in double optical resonance’. Opt.
Commun., 25, 359–362.
G ROVE , R. E., F. Y. W U and S. E ZEKIEL: 1977. ‘Measurement of spectrum of resonance fluo-
rescence from a 2-level atom in an intense monochromatic-field’. Phys. Rev. A, 15, 227–233.
H ARTIG , W., W. R ASMUSSEN, R. S CHIEDER and H. WALTHER: 1976. ‘Study of frequency-
distribution of fluorescent light-induced by monochromatic radiation’. Z. Phys. A, 278, 205–
210.
H ERTEL , I. V. and W. R ADLOFF: 2006. ‘Ultrafast dynamics in isolated molecules and molecular
clusters’. Rep. Prog. Phys., 69, 1897–2003.
ISO 21348: 2007. ‘Space environment (natural and artificial) – Process for determining solar irra-
diances’. International Organization for Standardization, Geneva, Switzerland.
L IPPERT , H., V. S TERT, L. H ESSE, C. P. S CHULZ, I. V. H ERTEL and W. R ADLOFF: 2003. ‘Anal-
ysis of hydrogen atom transfer in photoexcited indole(NH3 )n clusters by femtosecond time-
resolved photoelectron spectroscopy’. J. Phys. Chem. A, 107, 8239–8250.
M OLLOW , B. R.: 1975. ‘Pure-state analysis of resonant light-scattering – radiative damping, sat-
uration, and multi-photon effects’. Phys. Rev. A, 12, 1919–1943.
M OLLOW , R. B.: 1969. ‘Power spectrum of light scattered by two-level systems’. Phys. Rev., 188,
1969–1975.
P IRO , N. et al.: 2011. ‘Heralded single-photon absorption by a single atom’. Nat. Phys., 7, 17–20.
R ABI , I. I.: 1944. ‘The N OBEL prize in physics: for his resonance method for recording the
magnetic properties of atomic nuclei’, Stockholm. http://nobelprize.org/nobel_prizes/physics/
laureates/1944/.
S CHUDA , F., C. R. S TROUD and M. H ERCHER: 1974. ‘Observation of resonant stark effect at
optical frequencies’. J. Phys. B, At. Mol. Phys., 7, L198–L202.
S TELLMER , S., B. PASQUIOU, R. G RIMM and F. S CHRECK: 2012. ‘Creation of ultracold Sr2
molecules in the electronic ground state’. Phys. Rev. Lett., 109, 115302.
SWP 5.5: 2005. ‘Scientific work place’, Poulsbo, WA 98370-7370, USA: MacKichan Software,
Inc. http://www.mackichan.com/, accessed: 9 Jan 2014.
T ICHY , M. C., F. M INTERT and A. B UCHLEITNER: 2011. ‘Essential entanglement for atomic and
molecular physics’. J. Phys. B, At. Mol. Phys., 44, 192001.
VOLZ , J., M. W EBER, D. S CHLENK, W. ROSENFELD, C. K URTSIEFER and H. W EINFURTER:
2007. ‘An atom and a photon’. Laser Phys., 17, 1007–1016.
W EBER , M., J. VOLZ, K. S AUCKE, C. K URTSIEFER and H. W EINFURTER: 2006. ‘Analysis of a
single-atom dipole trap’. Phys. Rev. A, 73, 043406.
W EISSBLUTH , M.: 1989. Photon-Atom Interactions. New York, London, Toronto, Sydney, San
Francisco: Academic Press, 407 pages.
W INKLER , K., F. L ANG, G. T HALHAMMER, P. VON DER S TRATEN, R. G RIMM and J. H. D EN -
SCHLAG : 2007. ‘Coherent optical transfer of Feshbach molecules to a lower vibrational state’.
Phys. Rev. Lett., 98, 043201.
Z IEGLER , L. D.: 1985. ‘Rovibronic absorption analysis of the à ← X̃ transition of ammonia’.
J. Chem. Phys., 82, 664–669.
Appendices
Overview
These appendices contain supplementing material for those readers who want
to obtain some more in depth information on a few selected topics. Except for
Appendix C they do not contain essential tools for the main text.
Appendix A Gives an example for the explicit evaluation of the first B ORN ap-
proximation in the case of inelastic electron scattering.
Appendix B acquaints the reader with important devices for the detection, ma-
nipulation and energy selection of electrons and ions which are used in many exper-
iments discussed throughout the main text.
Appendix C introduces the so called state multipole moments as irreducible rep-
resentation of the density matrix, a concept which is intensely used in Chap. 9, it
compares different types of multipole expansions and introduces some useful tools
for working with them.
Appendix D, finally, illustrates the concept of optical pumping by way of exam-
ple for the very often used hyperfine pumping of Na atoms.
First B ORN Approximation
for e + Na(3s) → e + Na(3p) A
as defined in (8.21). The integration has to be carried out over all active elec-
trons with the coordinates r n .
The key quantity in this approximation is the momentum transfer vector
K = k i − k f with K = kf2 + ki2 − 2kf ki cos θ , (A.2)
× dϕB sin θB dθB Y0mi (θB , ϕB )eiKz Y1mf (θB , ϕB ) .
The most reasonable coordinate system has its z-axis parallel to K, and z = r cos θB .
Here θB is the polar angle of the position vector r of the target electron with respect
to K – not to be confused with the scattering angle θ , which according to (A.2) is
built into K.
As a consequence of the symmetry with respect to this z-axis only transitions
with m = 0 may occur. Thus, the only nonvanishing matrix element to be evalu-
ated is
3p0|e iK·r
|3s0 = 2π dr u3s (r)u3p (r)
× Y10 (θB , ϕB )eiKz Y00 (θB , ϕB ) sin θB dθB .
√ √
If we insert Y10 (cos θB ) = 3/4π cos θB and Y00 (cos θB ) = 1/ 4π the angular in-
tegration can be carried out in closed form:
√ √ π
3 3 iKr cos θB
cos θB e iKr cos θB
sin θB dθB = e (iKr cos θ − 1) .
2 2
2K r 2 B
0
This leads to a power series for the generalized oscillator strength. Explicitly one
finds in the present case for the GOS up to the 6th power in K:
(GOS) 2 1 rr 5 r 3 2
f3p3s = W3p3s r2 − K 2 r r 3 + K 4 +
3 5 140 100
rr r r
7 3 5
− K6 + + O K8 . (A.6)
7560 1400
We see here, that the step from the series expansion (8.33) in the main text to com-
putable radial matrix elements is not completely trivial. Alternatively and valid for
any single electron system one may obtain this result also by expressing the powers
of z which appear in (8.33) in terms of spherical harmonics Y0 . Thus one obtains a
genuine multipole expansion
... 2
(GOS) 2Wf i
ff i = K 2 s γf Jf Mf |r Y0 |γi Ji Mi (A.7)
gi
Ji Mi Jf Mf =0
...
= f K 2 ,
=0
which one can extend as far as the experimental accuracy requires it. The matrix
elements may be rewritten with the help of the W IGNER -E CKART theorem into re-
duced matrix elements while the M dependence averages out (as long as the initial
state is populated isotropically). One finally obtains expressions of the type (A.6)
with coefficients f which can be determined experimentally or theoretically. The
advantage here is that this procedure can be used in principle for arbitrary transi-
tions and coupling schemes. The coefficients may be compared to the corresponding
expressions for optical transitions (dipole, quadrupole etc. transitions). For a com-
parison with theory one has of course to compute the necessary matrix elements of
powers of r by integration over the wave functions. In any case, it is recommended
to check in each individual case whether a complete integration of (A.1) is possibly
more convenient. The results presented it in Sect. 8.3.3 for the p ← s transition in
Na originate from such full integration.
In our case the evaluation of the matrix elements gives r = −4.2687, so that the
first, constant term in the series (A.6) assumes the value 0.939. This limiting value
should for K → 0 be identical to the optical oscillator strength. The experimentally
determined literature value is f (opt) = 0.960. It gives us a feeling for the quality
or deficits of the wave functions used. According to K IM (2007) it is advisable to
672 A First B ORN Approximation for e + Na(3s) → e + Na(3p)
rescale the thus computed oscillator strengths as well as the differential cross section
as
References
K IM, Y. K.: 2007. ‘Scaled Born cross sections for excitations of H2 by electron impact’. J. Chem.
Phys., 126.
S CHUMACHER, E.: 2011. ‘FDAlin programme, computation of atomic orbitals (Windows and
Linux)’, Chemsoft, Bern. http://www.chemsoft.ch/qc/fda.htm, accessed: 5 Jan 2014.
Guiding, Detecting and Energy Analysis
of Electrons and Ions B
In the main text we refer on various occasions to methods for the manipulation,
detection or energy analysis of electron and ion beams which today are standard ex-
perimental tools far beyond modern atomic, molecular and optical physics. Electron
and ion optics and the detailed layout of energy selectors are nowadays designed
with the help of commercial programmes (see e.g. SIMION 2012). Nevertheless it
is useful to have some elementary knowledge about principles and typical compo-
nents. One essential basis of all these methods are the forces which act on these par-
ticles in electric or magnetic fields which lead to characteristic deflections of particle
beams as described already in Chap. 1, Vol. 1. The following short introduction fo-
cusses onto low energy particle beams for which today usually electrostatic guiding
and focussing is used. In the last section we shall, however, also show an important
application of magnetic fields. We do not discuss here space charge effects since
they usually do not play any role for the detection and analysis of particles (very
low currents).
We start with a summary on secondary electron multipliers (SEM) and related de-
vices which are most commonly used in low energy atomic and molecular physics
for direct detection of ions, electrons, VUV and XUV photons. By repeated sec-
ondary electron emission they provide a typical amplification of ca. 108 . Each single
particle thus generates a pulse of some ns duration with currents in the range of mA.
These may then be further amplified with conventional electronics, be discriminated
against noise and are finally counted. This allows to count single electrons, ions or
photons with high efficiency.
The classical standard arrangement of an SEM is sketched in Fig. B.1(a). The
electron to be detected hits the first dynode and ejects there one, two or even more
secondary electrons of low kinetic energy. These are accelerated towards the next
dynode where the secondary electron emission is repeated for each impacting elec-
tron. The whole process is repeated many times so that an avalanche of electrons is
pulse
prototype SEM scheme with to
individual dynodes, amplifier
(b) principle of a channeltron, e- or
ion
(c) typical realization of a
C
channeltron
_ high voltage
(HV) +
electron exit
e- or glass wall
ion _
HV +
ca.15mm
built up which is finally captured on an anode. From there it is coupled into an elec-
tronic circuit for further processing. For particle detection one typically uses 14–18
dynodes, for photomultipliers usually less. An overall voltage (HV in Fig. B.1(a)) of
2–4 kV is applied and distributed to the dynodes via a voltage divider chain as shown
in Fig. B.1(a). In front of each dynode the electrons have a kinetic energies of 100 eV
to 200 eV, which allows for efficient secondary electron emission. For the detection
of photons (also in the visible) standard photomultipliers (PM) are still built accord-
ing to this classical scheme. In this case a photocathode is mounted in front of the
SEM where the photons are efficiently converted by photoelectron emission into
electrons. The whole assembly – photocathode and SEM – are mounted in a com-
pact, evacuated tube, the voltage for the dynodes is fed into this tube by connector
pins in the bottom of the tube. This is today a very well established technology, and
a broad range of very efficient devices for various applications is commercially of-
fered: low dark currents and high sensitivity cover today a broad spectral range for
single photon counting.
For the detection of electrons, ions and fast neutral particles one uses today al-
most exclusively channel electron multiplier (short channeltrons or CEM). Their
principle is illustrated in Fig. B.1(b). Instead of discrete dynodes in an SEM
(Fig. B.1(a)) a channeltron uses the whole inner wall of a thin (some mm diam-
eter) glass tube continuously as a kind of extended dynode for secondary electron
emission. This inner wall is coated with a thin semiconducting layer of some 100 nm
thickness with a high resistance (in the 100 M region). On top of it a several 10 µm
thick layer of SiO2 is applied for passivation and improving the secondary electron
emission. At typical voltages of 3000 V (current ca. 30 µA) the potential rises lin-
early over the total length of the tube (some cm), so that along the whole length
secondary electrons are ejected from the walls and accelerated continuously. A va-
riety of specific shapes for channeltrons are used. Figure B.1(c) shows an example
B.1 SEM, Channeltron, Microchannel Plate 675
-
electical e or ion
contact entrance
channel wall
material
_
secondar- _
elektronen HV
+ ca. 1000 V
+
glass structure
ca. 103-104 e- microchannels
e- exit
on exit
primary channels
(d) 2 microchannel plates particle _
in chevron configuration
MCP1 +_
MCP2
+
electron anode
avalanche pulse
on exit exit
Fig. B.2 Microchannel plate (MCP) schematic: (a) function of a single channel, (b) cut through
a single MCP, (c) two MCP’s in chevron configuration to enhance amplification with typical cou-
pling to the electronics
with an entrance funnel (important for “hitting” the channeltron) and an amplifying
channel in spiral shape. This particular shape is supposed to improve the emission
geometry and to minimize echoes from ions moving into the back direction which
may be ejected from the walls. Alternatively, compact channeltrons embedded into
ceramics in sinusoidal shape (Ceratron) and other shapes are offered. Amplifica-
tions of more then 108 are possible, which decreases, however, rapidly at higher
count rates above 104 s−1 (saturation).
Microchannel plates (MCP) may be seen as a consequent further development
of the channeltron, schematically illustrated in Fig. B.2. An MCP consists of many
microchannels as shown in Fig. B.2(a) with a diameter of 6 µm to 10 µm, each of
them acting just like a channeltron. They are combined to plates of some tenth mm
up to 2 mm thickness as seen in Fig. B.2(b). Typically the channels are tilted by
a small angle (ca. 8◦ ) with respect to normal in order to avoid direct transmission
of the particles to be detected and to reduce the ion back passages. The amplifica-
tion of a single MCP is, however, moderate, only about 103 –104 . Thus, as indicated
in Fig. B.2(c), one uses two plates in tandem, either in the so called “Chevron” or
in a V configuration (one plate is turned around the normal by 180◦ ). With such
a setup one reaches a gain of 106 to 107 , which as a rule is sufficient for particle
counting; if necessary one may even post three plates behind each other (Z config-
uration). The main advantage of microchannel plates is their large area: diameters
676 B Guiding, Detecting and Energy Analysis of Electrons and Ions
detection probability P / %
secondary emission as a
function of their kinetic 80
energy T at the entrance into
the channeltron 60
40
20
0
10 100 1000 104
T / eV
γeN
Pe (N ) = exp(−γe ). (B.1)
N!
The quantity γe is called secondary electron emission coefficient. The probability
that no electron is emitted is Pe (0) = exp(−γe ). Thus, the detection probability,
which we look for, is given by the probability to emit one or more electrons:
P = 1 − exp(−γe ). (B.2)
In principle γe and thus P depends on the velocity v as well as on the mass M of the
ions studied. In general one expect the detection efficiency to rise with the velocity,
and for the same velocity probably also with mass. During the past decades there
have been a number of attempts to determine γe experimentally, in particular for
larger masses, unfortunately with somewhat uncertain outcome (see the discussion
by F RASER 2002). We communicate two relatively recent results for MCPs. In both
cases a calibration was attempted with alternative ion detection methods, assumed
to be quantitative. Different masses with molar weights of some 100 u up to some
1000 u were investigated. W ESTMACOTT et al. (2000) find by comparison with su-
perconducting tunnel contacts an empirical formula for the “reduced” secondary
electron emission coefficient (per unit of mass):
γe /u = Av B . (B.3)
The authors report a value B = 4.3 ± 0.4, while A = 5.6748 × 10−24 can be gleaned
from their data. Here v is measured in m s−1 . In practice, usually of interest is the
detection probability as a function of kinetic energy T (in keV) and mass M (in u)
of the detected ion. We find:
B
P (M, T ) = 1 − exp −M × A × 4.4 × 105 T /M (B.4)
= 1 − exp −10.5M −1.15 T 2.15 .
Both results are compared in Fig. B.4 with each other for two different masses.
As seen, for typical extraction voltages 3–4 kV the detection probability for ions of
678 B Guiding, Detecting and Energy Analysis of Electrons and Ions
P/ %
MCP. They grey lines show 60
relation (B.4) derived from
W ESTMACOTT et al. (2000), 40
while the red lines illustrate
relation (B.6) after 20
T WERENBOLD et al. (2001).
Full lines refer to mass 0
M = 720 u (C60 ), dashed 0 5 10 15 20
T / keV
lines to M = 120 u
the higher mass is below 10 %. Special efforts are needed to obtain maximum de-
tection efficiency for large masses (e.g. for protein analysis). Either one has to place
the MCP’s at very high negative potential – far above the operation voltage for the
plates. Alternatively, one may generate the ions at a very high positive potential –
which is technically not easy to realize since usually one wants to have the source on
ground potential. Unfortunately, the calibration data reported above from the litera-
ture do not agree. We would tend to use the data of T WERENBOLD et al. (2001), i.e.
relation (B.6) illustrated in Fig. B.4 by red lines; their calibration by cryodetectors
appears relatively straight forward (see, however, F RASER 2002).
The deflection and focussing of charged particle beams may be treated in a very sim-
ilar manner as done for light beams in geometrical optics, hence the field is called
electron and ion optics. Refraction of a particle beam occurs when it passes through
different electric potentials. In contrast to light optics there usually are no sharp
boundaries, rather the change of direction occurs continuously corresponding to the
respective local kinetic energy, as sketched in Fig. B.5. Here we show the example
of a homogeneous electrostatic field between two transparent plane, parallel metal
grids in a distance d at the potential U1 and U2 , respectively.
When the particle beam (particle mass m, charge qe) passes through such a field
its kinetic energy changes from T√ 1 = qeU1 to T2 = √qeU2 . The magnitude of the
velocity thus changes from v1 = 2T1 /m to v2 = 2T2 /m. Perpendicular to the
optical axis (z-axis, normal to the surface) the velocity component vy remains con-
B.2 Index of Refraction, Lenses and Directional Intensity 679
stant. For entrance and exit angles (with reference to the z-axis) θ1 and θ2 , respec-
tively, we have vy = v1 sin θ1 and vy = v2 sin θ2 . From this follows immediately the
refraction law for particle beams
sin θ1 v2 T2 n 2
= = = , (B.7)
sin θ2 v1 T1 n 1
in full analogy to S NELLIUS’ law in geometrical optics. The index of refraction for
particle
√ beams is thus proportional to the square root of the local kinetic energy
n ∝ T . Note that this relation does not depend on the magnitude of the angle and
for the geometry sketched in Fig. B.5 is not limited to small entrance or exit angles.
Electron and ion lenses may thus be constructed in analogy to light optics. Some
characteristic examples are summarized in Fig. B.6. The red lines in (a) and (c)
indicate typical electron trajectories in collecting lenses. One distinguishes einzel
lenses (a, b) and immersion lenses (c, d). The former consist of three elements
(typically apertures or cylinders), of which the two outer ones are fixed to the same
potential, so that the energy of the particles is conserved. In contrast, immersion
lenses are made of two elements at different potential. In addition to focussing or
defocussing the charged particles, they also change their energy. The focal lengths
and other imaging properties cannot be written in terms of simple formulas as in
light optics. Rather, a detailed calculation or measurement is needed (a survey on the
classical literature is given by M ULVEY and WALLINGTON 1973). Today one uses
commercially available programmes, already mentioned above, to simulate particle
trajectories. These programmes allows one to design much more sophisticated and
better optimized geometries than implied by the simple lens patterns introduced
above.
The so called H ELMHOLTZ -L AGRANGE relation (also known from light optics
as A BBE sine condition), is an other important relation for the propagation of parti-
cle beams:
It describes the relation between lateral magnification β = y2 /y1 and the respective
divergence angles θ1 and θ1 for a particle beam which images an object of the
size y1 onto y2 as sketched in Fig. B.7. It is valid only for near-axial rays (small
680 B Guiding, Detecting and Energy Analysis of Electrons and Ions
divergence angles) – in contrast to the S NELLIUS law of refraction (B.7), with which
it must not be confused. In two dimensional perspective, i.e. with reference to the
differential areas dA1 and dA2 of a particle beam at two different positions along the
beam path one obtains the usually discussed form of the H ELMHOLTZ -L AGRANGE
relation
T1 dA1 dΩ1 = T2 dA2 dΩ2 , (B.9)
where dΩ1 and dΩ2 are the respective differential solid angles.
One defines now the so called directional intensity1 of a particle beam:
dI
R= , (B.10)
dAdΩ
with the current dI which passes through the area dA along the beam axis. The
directional intensity thus characterizes the current flux dI /dA of the beam with
respect to its divergence angle dΩ. With (B.9) for the directional intensity the fol-
lowing conservation law holds for two different positions in one beam:
R1 dI1 dI2 R2
= = = . (B.11)
T1 T1 dA1 dΩ1 T2 dA2 dΩ2 T2
For each well collimated particle beam the ratio of directional intensity to kinetic
energy, R/T , is a conserved quantity – of course only if no particles are lost and if
no energy dispersive elements are active on the beam path.
This relation is of quite fundamental importance for the design of electron and
ion optics. If one wants to obtain as much as possible current through a small cross
section – which often is the standard requirement – one has to make sure that al-
ready at the beginning, i.e. when the ions or electrons are generated, the directional
intensity is as high as possible (which means we have to design the cathode of an
electron source well, or we have to focus the light source properly if we are talking
about photoelectrons etc.). High energies are also helpful, they may, however, not
be what is otherwise required for the experiment.
For energy analysis in photoionization spectroscopy, today one typically uses elec-
trostatic methods if continuous radiation sources are involved (CW lasers, syn-
1 Also denoted as “brightness of the beam”, or by the original German term “Richtstrahlwert”.
B.3 Hemispherical Energy Selector 681
R1
R0
focussing optics w max
R2
detector B
MCP's
x2
chrotron sources), while for pulsed sources (short pulse lasers, isolated SR pulses)
time of flight methods are appropriate (they will briefly be discussed in the next sec-
tion). In atomic, molecular and cluster science electrons and/or ions are detected.
And in photoemission spectroscopy from solid surfaces electrons are detected.
The basic concept of all electrostatic energy selectors is a combination of (i)
spatial separation of particles with different kinetic energy by deflection in the elec-
tric field with (ii) geometric focussing of different angles at the entrance slit of
the selector. In this way one tries to collect as much signal as possible to pass
through the dispersive element. This concept succeeds with different perfection de-
pending on the geometry used (cylindrical, hemispherical, toroidal, trochoidal etc.).
Such electrostatic energy monochromators can be used for both, energy analysis
of charged particles as well as for the generation of monoenergetic electron and
ion beams.
Here we discuss as an example the hemispherical analyzer, the advantages of
which are well proven in practice. A number of commercial realizations are mar-
keted with great success. This analyzer consists of two concentric hemispheres as
sketched in Fig. B.8. This setup has first been described by P URCELL (1938) and
was introduced to low energy photoelectron spectroscopy by K UYATT and S IMP -
SON (1967). The following description is based on these papers and includes mod-
ern types of realization. We assume an idealized field distribution which is not dis-
turbed by boundary effects. One may achieve this according to H ERZOG (1935) by
suitable limiting apertures which we consider to be realized in Fig. B.8 by an opti-
mally positioned entrance slit and the first detector plate at the exit of the analyzer,
respectively.
The electrostatic potential between the hemispheres in the geometry of Fig. B.8
is according to the laws of electrostatics C1 /R + C2 . The magnitude of the electric
field thus becomes E(R) = U R1 R2 /[(R2 − R1 )R 2 ], with U being the overall poten-
682 B Guiding, Detecting and Energy Analysis of Electrons and Ions
tial difference between the hemispheres. The field is directed radially and we expect
circular (more generally elliptical) orbits just as for the K EPLER problem. Let us first
discuss electrons which start at point A on the radius R0 = (R2 + R1 )/2 (i.e. exactly
in the middle between the spheres) and enter perpendicular to the connecting axis
between points AB (dash dotted line). Their orbit is exactly a circle if the centrifu-
gal force and the electric field compensate each other, i.e. if qeE(R0 ) = −mv02 /R0 .
Hence, to guide an electron of kinetic energy T0 = eU0 = mv02 /2 on the nominal
circular orbit with R0 , the voltage difference between inner and outer semi-sphere
must be
The fact that the angular divergence only enters in quadratic form means that the
spherical analyzer does focus different angles to first order in the sectional plane
shown in Fig. B.8. To derive the energy resolution of such a system one has to
account for the width of the entrance and exit apertures. In case of slits with equal
width w and neglecting the α 2 term, one expects a triangular transmission function
with a FWHM
w
T1/2 = T0 . (B.14)
2R0
B.4 Magnetic Bottle and Other Time of Flight Methods 683
For best energy resolution T1/2 one thus works advantageously at low transmis-
sion energies which can be obtained by suitable particle optics (immersion lenses) –
also for high initial kinetic energies. However, according to (B.11) this can only be
achieved at the expense of a larger divergence angle. The most useful geometry and
electron optics has thus to be assessed specifically for each individual experiment.
In the setup sketched in Fig. B.8 the entrance aperture is considered to be a
vertical slit. At the exit a multichannel plate with 2D resolution is used – the exit
slit is thus realized by the spatial resolution of the plate. In the situation shown
here an extended area of a target is illuminated with photons. This is imaged onto
the entrance slit of the analyzer. In the sectional plane shown, one thus collects
all electrons which are emitted from a finite part of the target in this plane, emitted
perpendicular to the target. Correspondingly electrons which are emitted with angles
differing from normal to the target plane are imaged onto the entrance slit below or
above A. The analyzer images these below or above the sectional plane onto the
detector MCP – again selected for energies in the x2 direction. In summary, this
allows a two dimensional recording of the photoelectron emission: according to
energy (horizontal direction) and emission angle (vertical direction).
As mentioned above, if the light source is pulsed and thus marks sharply the time
of the photoionization process, time of flight (TOF) methods are appropriate for
energy analysis. An often used setup is the so called magnetic bottle, whose main
merit is a high collection efficiency, specifically for low energy electrons. It has
been used by K RUIT and R EAD (1983) for low energy electron spectroscopy for the
first time. For many applications it is still the method of choice. If one is interested
in the kinetic energy of electrons from weak sources it helps if one can register –
if possible – all electrons emitted into one half full solid angle 2π . With simple
time of flight methods which allow the electrons to drift without guiding field, one
typically looses a substantial part of the signal which is emitted with lateral velocity
components.
If, on the other hand, one first extracts the electrons into the direction of the drift
tube, different electron emission angles lead to different lifetimes for the same initial
kinetic energy: without special precautions one just measures the velocity compo-
nent parallel to the extraction and detection region. The magnetic bottle is the ideal
remedy to this problem: one first parallelizes the momenta of the emitted electrons
during an initial, short time of flight. This is done in a strongly inhomogeneous
magnetic field without changing the total kinetic energy.
The essence of this method is schematically illustrated by Fig. B.9. A strong
magnetic field B i (some Tesla, typically generated by permanent magnets) is main-
tained at the interaction centre, where the photoelectrons are created. In cylindrical
symmetry the field is smoothly changed along the z-axis into a weak field B f that
684 B Guiding, Detecting and Energy Analysis of Electrons and Ions
extends along the actual, electric field free, drift distance of the TOF. Typically B f
is some mT, generated by a solenoid which in turn is shielded by µ-metal from
external magnetic fields.
Let an electron (mass me , charge −e) originally be generated in the lower part
of the setup with a velocity v i and be emitted at an angle θi with reference to the
z-axis. In the strong field B i it will spiral with the cyclotron frequency
eBi
ωi = . (B.15)
me
The initial radius of the orbit ri is defined by the radial component vi sin θi of the
velocity:
vi sin θi me vi sin θi
ri = = . (B.16)
ωi eBi
This circular movement has an angular momentum
2
vi sin θi vi2 sin2 θi m2 v 2 sin2 θi
i = Θωi = me ri2 ωi = me ωi = me = e i . (B.17)
ωi ωi eBi
Thus, the transverse component of the velocity decreases strongly while the longi-
tudinal component in z-direction which determines the time of flight grows from
initially v cos θi to
vzf = v cos θf = v 1 − sin θf = v 1 − (Bf /Bi ) sin2 θi .
2
Acronyms and Terminology 685
The trajectories thus become indeed parallel. For a magnetic field ratio of Bf /Bi =
1:1000 the z-component of the velocity vzf in the TOF region differs from the over-
all magnitude v only by 0.5 h.
With (B.16) one may write (B.18) as
1/2
ri Bf
= . (B.19)
rf Bi
This means that the magnetic flux Bπr 2 through one circular orbit of the electron
is a constant of motion.
Aside from this particular variety of a TOF electron spectrometer, optimized for
maximum collection efficiency, during the past years a variety of powerful imag-
ing techniques has been developed for photoelectrons and photoions. In addition
to energy measurement these allow a complete determination of the particle mo-
mentum by direction and magnitude (velocity map imaging VMI, also energy imag-
ing, momentum mapping etc.) which we have mentioned already several times (e.g.
in Sect. 5.5.4, Vol. 1). With the help of modern electronics one may detect with
such methods several particles simultaneously (electrons and one or more fragment
ions). These methods make a whole new class of experiments accessible for modern
atomic and molecular physics in collision processes or laser matter interaction –
with hitherto unknown high efficiency and detail. Particularly successful in this
context is the so called reaction microscope, originally introduced by S CHMIDT-
B ÖCKING and collaborators as COLTRIMS (cold target recoil ion momentum spec-
troscopy). Recent developments of it are MOTRIMS (magneto-optical trap recoil
ion momentum spectroscopy). They all are extremely powerful but by no means
trivial to handle and to evaluate. Among other ingredients the evaluation requires
a detailed analysis of the particle trajectories in the magnetic guiding fields which
are used in these devices, as well as sophisticated position and time resolved coin-
cidence detection techniques. In the MOTRIMS variety one needs in addition ex-
pertise with cooling and storing of particles by “state-of-the-art” laser traps. A good
entrance for the interested readers give the reviews of U LLRICH et al. (2003) and
D EPAOLA et al. (2008) (while optical traps are reviewed by G RIMM et al. 2000).
CCD: ‘Charge coupled device’, semiconductor device typically used for digital
imaging (e.g. in electronic cameras).
CEM: ‘Channel electron multiplier’, see Appendix B.1.
COLTRIMS: ‘Cold target recoil ion momentum spectroscopy’, see Appendix B.4.
CW: ‘Continuous wave’, (as opposed to pulsed) light beam, laser radiation etc.
FWHM: ‘Full width at half maximum’.
MCP: ‘Multi channel plate’, electron multiplier with many amplifying elements.
MOTRIMS: ‘magneto-optical trap recoil ion momentum spectroscopy’, see Ap-
pendix B.4.
686 B Guiding, Detecting and Energy Analysis of Electrons and Ions
References
D EPAOLA, B. D., R. M ORGENSTERN and N. A NDERSEN: 2008. ‘Motrims: magneto-optical trap
recoil ion momentum spectroscopy’. In: ‘Adv. At. Mol. Opt. Phys.’, vol. 55, 139–189. Amster-
dam: Elsevier.
F RASER, G. W.: 2002. ‘The ion detection efficiency of microchannel plates (MCPs)’. Int. J. Mass
Spectrom., 215, 13–30.
G RIMM, R., M. W EIDEMÜLLER and Y. B. OVCHINNIKOV: 2000. ‘Optical dipole traps for neutral
atoms’. Adv. At. Mol. Opt. Phys., 42, 95–170.
H ERZOG, R.: 1935. ‘Berechnung des Streufeldes eines Kondensators, dessen Feld durch eine
Blende begrenzt ist’. Arch. Elektrotech., 29, 790–802.
ISO 21348: 2007. ‘Space environment (natural and artificial) – Process for determining solar
irradiances’. International Organization for Standardization, Geneva, Switzerland.
K RUIT, P. and F. H. R EAD: 1983. ‘Magnetic-field parallelizer for 2π electronspectrometer and
electron-image magnifier’. J. Phys., E J. Sci. Instrum., 16, 313–324.
K UYATT, C. E. and J. A. S IMPSON: 1967. ‘Electron monochromator design’. Rev. Sci. Instrum.,
38, 103–111.
M ULVEY, T. and M. J. WALLINGTON: 1973. ‘Electron lenses’. Rep. Prog. Phys., 36, 347–421.
P URCELL, E. M.: 1938. ‘The focusing of charged particles by a spherical condenser’. Phys. Rev.,
54, 818–826.
SIMION: 2012. ‘Industry standard charged particle optics simulation software’, Scientific Instru-
ment Services, Inc., Ringoes, NJ, USA. http://simion.com/, accessed: 5 Jan 2014.
T WERENBOLD, D., D. G ERBER, D. G RITTI, Y. G ONIN, A. N ETUSCHILL, F. ROSSEL, D. S CHEN -
KER and J. L. V UILLEUMIER : 2001. ‘Single molecule detector for mass spectrometry with mass
independent detection efficiency’. Proteomics, 1, 66–69.
U LLRICH, J., R. M OSHAMMER, A. D ORN, R. D ÖRNER, L. P. H. S CHMIDT and H. S CHMIDT-
B ÖCKING: 2003. ‘Recoil-ion and electron momentum spectroscopy: reaction-microscopes’.
Rep. Prog. Phys., 66, 1463–1545.
W ESTMACOTT, G., M. F RANK, S. E. L ABOV and W. H. B ENNER: 2000. ‘Using a superconducting
tunnel junction detector to measure the secondary electron emission efficiency for a microchan-
nel plate detector bombarded by large molecular ions’. Rapid Commun. Mass Spectrom., 14,
1854–1861.
Statistical Tensor and State Multipoles
C
As explicated in Chap. 9, the density matrix describes mixed states in concise form.
It is the key tool for evaluating experiments with imperfect state selection prior to
and/or incomplete state analysis after an interaction process – i.e. essentially for any
experiment in the real world. In practice, the final expressions involving the density
matrix may become rather clumsy, in particular when state selection, interaction and
analysis are best described in different coordinate systems and in different angular
momentum coupling schemes. In this context, a multipole expansion of the density
matrix often facilitates such evaluations and allows to disentangle the dynamics of
the processes studied from the geometry of a specific experiment.
We give here a short introduction to the concept of state multipoles as first intro-
duced by FANO (1953). They provide an irreducible representation of the density
matrix. Alternatively, the multipole moments introduced in Appendix F.3.2, Vol. 1
may be used. As we shall see below, these different representations are related to
each other by simple numerical factors derived from the W IGNER -E CKART theo-
rem Appendix C.1.2, Vol. 1. Slightly different definitions are used in the literature.
We essentially follow the notation of B LUM (2012), who gives a rigorous deriva-
tion of the concept. In A NDERSEN et al. (1988) both notations – state multipoles
and multipole moments are used. The reader interested in details finds there also
extensive tabulation of the relevant conversion relations and parameters.
We start with the density operator (9.15)–(9.18) and specify the states |γ of the
system as standard angular momentum states |j m:
ρ̂ = ρj m ,j m |j m j m|. (C.1)
j m j m
ρj m ,j m = ρj∗m,j m (C.2)
and characterizes a quantum system (or a specified subset) completely. One may
rewrite (C.1) as a multipole expansion of the type
†
ρ̂ = tˆ j j KQ tˆ j j KQ , (C.3)
KQ
with tˆ(j j )KQ representing appropriate irreducible tensor operators. At this point
the quantities tˆ(j j )†KQ are introduced as suitable expansion coefficients for a
given quantum system. (We use here the lower case letter tˆKQ to avoid confusion
with the multipole moment operators Tkq introduced in Appendix F.3.2, Vol. 1.)
In view of (C.1) the so called statistical tensor operators are constructed by cou-
pling angular momentum states |j m and j m | in essentially the same manner
as one couples angular momenta in a vector coupling scheme J1 + J2 = K (see
Eq. (B.31), Vol. 1), except that here the outer product |j m j m| of these states has
to be used:
tˆ j j KQ = (−1)j −m j m j − m|KQ|j m j m| (C.4)
m m
j −m 1/2 j j K
= (−1) (2K + 1) |j m j m|
m −m −Q
mm
(C.5)
with j − j ≤ K ≤ j + j and −K ≤ Q ≤ K.
The latter equality is derived from the W IGNER -E CKART theorem (C.8), Vol. 1.
Hence, the reduced matrix elements of the statistical tensor operators are2
2K + 1
j tK j = . (C.7)
2j + 1
1 Interestingly, this very relation was used by FANO (1953) to define the “state multipoles”.
2 in
the notation of B RINK and S ATCHLER (1994), while according to E DMONDS (1996), A NDER -
√
SEN et al. (1988), B LUM (2012) j tK j (Ed) = 2K + 1.
C.1 Multipole Expansion of the Density Matrix 689
Exploiting Hermiticity of the operator and the symmetry of the C LEBSCH -G ORDAN
coefficients
†
j m|tˆ j j KQ |j m = j m |tˆ j j KQ |j m∗
= (−1)j −m j m j − m|KQ
(C.8)
= (−1)j −j +Q (−1)j −m j mj − m |K − Q
= (−1)j −j +Q j m|tˆ jj K−Q |j m ,
Their expectation values – the so called state multipoles – are according to (9.23)
† † †
tˆ j j KQ = Tr ρ̂ tˆ j j KQ = ρj m ,j m j m|tˆ j j KQ |j m
m m
= ρj m ,j m j m |tˆ j j KQ |j m or
m m
† j j K
tˆ j j KQ = (−1)j −m (2K + 1)1/2 ρ .
m −m −Q j m ,j m
mm
(C.10)
Inserting this and the statistical tensor operator (C.5) into (C.3) and exploiting the
orthogonality relation (B.40), Vol. 1 for 3j symbols, we indeed recover (C.1). Sim-
ilarly, (C.10) may be inverted by multiplying both sides with (−1)j −m (2K +
1)1/2 × C LEBSCH -G ORDAN coefficient and summing over K and Q. One obtains
the density matrix elements:
j j K †
ρj m ,j m = (−)j −m (2K + 1)1/2 tˆ j j KQ . (C.11)
m −m −Q
KQ
If all tˆ(j j )†KQ are known, the density matrix elements may be derived from them.
With (C.3), (C.4), (C.10), and (C.11) we have thus fully achieved our initial goal: to
reformulate the density matrix as a multipole expansion.
Note that state multipoles up to rank K = j + j may be constructed from the
angular momenta j and j of the set of states to be described.
The simplest situation is the completely incoherent, equal population of all sub-
levels with ρm ,m = δm m /gj with the degeneracy gj = 2j + 1. The density matrix
3 The star indicating complex conjugate is actually redundant in this case since the matrix elements
are real.
690 C Statistical Tensor and State Multipoles
is then
1
ρ̂ = ,
2j + 1
in which case only the zero rank state multipole moment is nonvanishing:
1
tˆ(jj )†00 = √ .
2j + 1
The state multipoles are as such also irreducible tensor operators of rank K
and provide an alternative possibility to characterize a given quantum system. In
general, they are complex quantities, and we record here some of their properties –
easily derived from the above definitions, exploiting the Hermiticity of the density
matrix (C.2):
† ∗ †
tˆ j j KQ = tˆ j j KQ = (−1)j −j +Q tˆ jj K−Q , and
† ∗ (C.12)
tˆ j j KQ = tˆ j j KQ = (−1)j −j +Q tˆ jj K−Q .
This is in agreement with the relations (D.5), Vol. 1 communicated earlier for irre-
ducible tensor operators.
Under rotation through E ULER angles (ϕk θk 0) – see e.g. Fig. 4.3 in Vol. 1 – the
state multipoles transform as described by (E.15), Vol. 1:
† # † k
tˆ j j KQ = tˆ j j Kq DqQ (ϕk θk 0)∗ . (C.13)
q
† #
K
†
tˆ j j K0 = tˆ j j Kq CKq (θk , ϕk ). (C.14)
q=−K
Note that these expressions are dramatically more simple and transparent than
(9.146) which describes rotation of the density matrix itself. This is one of the key
advantages for using the state multipoles.
j TK j
j m |TKQ |j m = j m |tˆ j j KQ |j m, (C.15)
j tK j
and also holds for the respective expectation values. It allows us by comparison
with (F.32), Vol. 1 (see the examples given in Table F.1, Vol. 1) to associate the
better ‘physical intuition’ with the state multipoles. We point out here, however,
that for evaluation of experiments it is irrelevant which kind of multipole moments
are used, as long as their definition is clearly stated.4
Of special interest is the case with ‘sharp’ angular momentum j = j . The ratio of
the reduced density matrix elements is then given by (see e.g. M ACEK and H ERTEL
1974, and references quoted there):
1/2
j tK j 2K (2K + 1)1/2 (2j − K)!
v(K, j ) = = . (C.16)
j TK j K! (2j + K + 1)!
In standard literature both types of multipole expansions are used, e.g. for de-
scribing the angular distribution and polarization of radiation from anisotropically
populated excited atoms. Often it is convenient to rewrite them as real parameters
which directly express important symmetry properties as outlined in Appendix D.3,
Vol. 1. According to (D.5), Vol. 1 and using (C.15) and (C.16) one constructs them
from their complex counterparts for 0 < Q ≤ K by
√
T(j )KQ+ = (−1)Q 2 Re T(j )KQ and (C.18)
√
T(j )KQ− = −(−1)Q 2 Im T(j )KQ , (C.19)
while for Q = 0 T (j )K0+ = T (j )K0 and TK0− = 0, (C.20)
4 Atomic anisotropies and orientation may uniquely be represented by a set of expectation values
of tensor operators. However, as illustrated in Sect. 9.3, we find the density matrix usually a more
instructive representation – provided an appropriate coordinate system is chosen. In this spirit, mul-
tipole moments of any kind are seen mainly as an intermediate step for designing and evaluating
experiments.
692 C Statistical Tensor and State Multipoles
Table C.1 Complex multipole moments T(J )KQ i.e. expectation values of multipole tensor
operators, up to rank K = 2 for sharp angular momentum values up to J = 2, as a function of the
respective density matrix elements; real multipole moments are obtained from (C.18)–(C.20) and
state multipoles by multiplication with v(K, Q)
J K Q T(J )KQ v(K, Q)
any 0 0 1 (J + 1)−1/2
√
1/2 1 0∗ (ρ 1 , 1 − ρ− 1 ,− 1 )/2 2
2 2 2 2
√ √
1 1 −ρ 1 ,− 1 / 2 2
2 2
√
1 1 0 ρ1,1 − ρ−1,−1 1/ 2
√
1 1 −(ρ0,−1 + ρ1,0 ) 1/ 2
√
2 0a 1 − 3ρ0,0 1/ 6
√ √
2 1 3(ρ0,−1 + ρ1,0 ) 1/ 6
√ √
2 2 6ρ1,−1 1/ 6
√
3/2 1 0 3
+ 12 ρ 1 , 1 − 12 ρ− 1 ,− 1 − 32 ρ− 3 ,− 3
2 ρ 32 , 32 1/ 5
2 2 2 2 2 2
√ √ √ √
1 1 −( 3/2ρ 3 , 1 + 2ρ 1 ,− 1 + 3/2ρ− 1 ,− 3 ) 1/ 5
2 2 2 2 2 2
2 0 3(ρ 3 , 3 − ρ 1 , 1 − ρ− 1 ,− 1 + ρ− 3 ,− 3 ) 1/6
2 2 2 2 2 2 2 2
√
2 1 3 2(−ρ 3 , 1 + ρ− 1 ,− 3 ) 1/6
2 2 2 2
√
2 2 3 2(ρ 3 ,− 1 + ρ 1 ,− 3 ) 1/6
2 2 2 2
√
2 1 0 2ρ2,2 + ρ1,1 − ρ−1,−1 − 2ρ−2,−2 1/ 10
√ √ √ √ √
1 1 − 2ρ2,1 − 3ρ1,0 − 3ρ0,−1 − 2ρ−1,−2 1/ 10
√
2 0a 6[2(ρ2,2 + ρ−2,−2 ) + 12 (ρ1,1 + ρ−1,−1 ) − 1] 1/(3 14)
√ √ √
2 1 3[ 6ρ−1,−2 + ρ0,−1 − ρ1,0 − 6ρ2,1 ] 1/(3 14)
√ √
2 2 6[ρ2,0 + 3/2ρ1,−1 + ρ0,−2 ] 1/(3 14)
(
a Here we have made use of ρM,M = 1
For some examples of T(j )KQ , constructed from angular momenta with quan-
tum numbers 0 ≤ j ≤ 2, explicit expressions are given in Table C.1 (see also Table 3
in A NDERSEN et al. 1988). Multipole moments up to rank K = 2j are possible. We
communicate these here up to rank K = 2 only, since higher rank multipoles cannot
be determined in experiments with single photon emission (see Sect. 9.4).
C.3 Recoupling
When evaluation experiments which are sensitive to anisotropies, often an addi-
tional problem arises since the density matrix ρ̂ of the system studied and that of
the state analyzer σ̂ (sometimes also state selector may be involved) must be de-
scribed in different angular momentum coupling schemes. This requires recoupling
C.3 Recoupling 693
of the respective angular momentum states. Again, this is best done in by using an
irreducible representation of the density matrix.
To have something concrete in mind, let us consider the coupling scheme (J I )F
of electronic angular momentum J and nuclear spin I to the total angular momen-
tum F in an atomic system. According to (C.19), Vol. 1 one may construct tensor
operators of rank K from direct products of state multipole operators tˆ(J J )kJ qJ
and tˆ(I I )kI qI , describing the electronic and nuclear component of the system, re-
spectively:
t J J k ⊗ t I I k KQ = tˆ J J k tˆ I I k kJ qJ kI qI |KQ.
J I J qJ I qI
qJ qI
Note that tˆ(J J )kJ qJ is generated from the coupling scheme (J J )kJ while
tˆ(I I )kI qI corresponds to the scheme (I I )kI . The above direct product thus cor-
responds to an overall coupling scheme |(J J )kJ , (I I )kI ; K. In contrast, in
the statistical tensor operator tˆ(F F )KQ in the coupled system (J I )F must be
constructed according to (C.4) from |F mF states, i.e. in a coupling scheme
|(J I )F, (J I )F ; K. Thus, to rewrite tˆ(F F )KQ in terms of multipole moments of
its subsystems we have to recouple the four angular momenta involved correspond-
ingly:
tˆ F F KQ = (J I )F, J I F ; K J J kJ , I I kI ; K
J J I I kJ kI
× t J J k ⊗ t I I k KQ .
J I
And since the state multipoles (i.e. the tensor operators averaged over the popula-
tion densities of the states described) are also irreducible tensor operators, the same
relation holds also for these. Using above relations and (B.77), Vol. 1 we may relate
the state multipoles in the coupled and uncoupled scheme by
† 1/2
tˆ F F KQ = (2F + 1) 2F + 1 (2kJ + 1)(2kI + 1) (C.21)
J J I I kJ kI
⎧ ⎫
⎨J I F ⎬ † †
× J I F tˆ J J k q tˆ I I k q kJ qJ kI qI |KQ.
⎩ ⎭ J J I I
kJ kI K qJ qI
Now, one is often confronted with the situation that one subsystem is completely
isotropically populated while the other is not. For example, in an atomic collision
processes the nuclear spin states are typically without relevance, and remain statis-
tically populated and the density matrix elements of the nuclear spin subsystem are
simply ρMI ,MI = δMI MI /(2I + 1). With (C.10) only state multipoles of rank zero
exist, hence, kI = 0 so that K = kJ and the whole summation reduces to a single
term, while the 9j symbol reduces to a 6j symbol according to (B.78), Vol. 1. With
694 C Statistical Tensor and State Multipoles
√
tˆ(I )†00 = 1/ 2I + 1 one finally obtains:
† †
tˆ F F KQ = V (K, F ) × tˆ J J KQ (C.22)
[(2F + 1)(2F + 1)]1/2 F F K
with V (K, F ) = .
2I + 1 J J I
In the case that the electron spin S too is statistically distributed and all information
on an atomic anisotropy is contained in the orbital part L of the density matrix,
we may just repeat the above derivation for the coupling scheme (LS)J . The state
multipole moment in the hyperfine coupling scheme are then related to that in the
orbital angular momentum subsystem by
† †
tˆ F F KQ = V (K, F )V (K, J ) × tˆ LL KQ (C.23)
[(2J + 1)(2J + 1)]1/2 J J K
with V (K, J ) = .
2S + 1 L L S
The same recoupling relations also hold for the TKQ± . Thus, if recoupling is
necessary, in the transformation relations (C.18)–(C.20) the factor 1/v(K, F ) has
to be replaced by V (K, F )V (K, J )/v(K, F ). For further details we refer the inter-
ested reader to the original literature, specifically to B LUM (2012), A NDERSEN et
al. (1988), Appendices A and B, and to H ERTEL and S TOLL (1978), Chap. IV.
References
A NDERSEN, N., J. W. G ALLAGHER and I. V. H ERTEL: 1988. ‘Collisional alignment and orienta-
tion of atomic outer shells: 1. Direct excitation by electron and atom impact’. Phys. Rep., 165,
1–188.
B LUM, K.: 2012. Density Matrix Theory and Applications. Atomic, Optical, and Plasma Physics
64. Berlin, Heidelberg: Springer, 3rd edn., 343 pages.
B RINK, D. M. and G. R. S ATCHLER: 1994. Angular Momentum. Oxford: Oxford University Press,
3rd edn., 182 pages.
E DMONDS, A. R.: 1996. Angular Momentum in Quantum Mechanics. Princeton: Princeton Uni-
versity Press, 154 pages.
FANO, U.: 1953. ‘Geometrical characterization of nuclear states and the theory of angular correla-
tions’. Phys. Rev., 90, 577–579.
H ERTEL, I. V. and W. S TOLL: 1978. ‘Collision experiments with laser excited atoms in crossed
beams’. In: ‘Adv. Atom. Mol. Phys.’, vol. 13, 113–228. New York: Academic Press.
M ACEK, J. and I. V. H ERTEL: 1974. ‘Theory of electron-scattering from laser-excited atoms’. J.
Phys., B At. Mol. Phys., 7, 2173–2188.
Optical Pumping
D
The concept of optical pumping implies the repeated, nearly resonant absorption
and reemission of photons, due to which the state population of a quantum sys-
tem is modified. Dramatic changes of the thermal equilibrium population of atoms,
molecules or solids may thus be achieved. The origin of optical pumping goes back
into the thirties of the past century – long before the laser was invented – and
are connected to names like H ANLE, B ERNHEIM, B ROSSEL and B LOOM. The re-
lated atomic spectroscopy reached a first summit with the N OBEL prize for Alfred
K ASTLER (1966).
It entered into a second, very fruitful period in the beginning of the seventies
with the development of tunable dye lasers. They allowed first the first time to ex-
ploit the full potential of this method. As an example, our group succeeded in 1973
by laser optical pumping to study electron scattering processes from exited states in
an atomic beam (H ERTEL and S TOLL 1974a). Today optical pumping belongs, so
to say, to the standard tools of working with light, in laser physics, spectroscopy,
collision physics or in creating ultracold atoms and molecules. Also in the present
textbook several experiments have been discussed where optical pumping is of im-
portance to prepare atoms and molecules with a non-thermal population. We thus
want to briefly illustrate the method by way of one, very often used example.
As discussed in Chap. 10 one often wants to prepare a pure two level system. In
nature, such a system exists only in very special cases, e.g. as a pure spin 1/2 system.
Thus, one tries to populate two states which may be transferred by absorption or
emission of radiation only into each other. Such a nearly ideal two state situation
can be achieved for a well defined hyperfine transition (HFS) of the Na D2 Line by
optical pumping. Figure D.1 illustrates two alternative optical pumping processes in
this system. The HFS term scheme Fig. D.1(a) and the allowed optical transitions
have already been explicated in Fig. 9.5, Vol. 1.
Fig. D.1 Hyperfine optical pumping with quasi-monochromatic laser light (red double arrows)
of Na between the 32 S1/2 (F = 2) and 32 P3/2 (F = 3) levels. (a) Energy levels (not to scale) with
HFS: F, F and FS: 3 2 P3/2,1/2 ; (b) MF = +1 excitation with σ + light for preparation of a pure
two level system; (c) MF = 0 excitation with π light. In both cases spontaneous emissions with
MF = ±1, 0 lead to the redistribution of population
If the laser is tuned into resonance with the 3 2 S1/2 F = 2 ↔ 3 2 P3/2 F = 3 tran-
sition, spontaneous emissions occur only between 3 2 P3/2 F = 3 → 3 2 S1/2 F = 2,
owing to the selection rule F = 0, ±1. These two HFS levels consist of 7 and 5
projection levels, with MF and MF , respectively. If one uses, however, circularly
polarized σ + light (alternatively σ − light) as indicated in Fig. D.1(b) for excitation
one enforces for the excitation process the selection rule MF = +1 (respectively
MF = −1) with reference to z k, where k is the wave vector of the light. In con-
trast, for spontaneous emission all transitions with MF = ±1, 0 are allowed. In
summary this leads to a positive transfer of angular momentum onto the ensemble
of atoms. For each repetition of the process, again positive angular momentum is
transferred on average. With a lifetime of 16 ns in the excited state the atoms un-
dergo typically several 100 pump cycles when passing through the laser beam (at a
velocity of ca. 1000 m s−1 in an atomic beam). Already after a few cycles one finds
practically all atoms only in one ground state |3 2 S1/2 F = 2MF = 2, and – cou-
pled by the laser light – in the corresponding excited state |3 2 P3/2 F = 3MF = 3.
Transitions into other hyperfine sublevels MF are no longer possible. Thus, one has
prepared a pure two state system.
The situation is somewhat different in the case of linearly polarized π light as
indicated in Fig. D.1(b). With reference to a coordinate system with z E, the
electric field vector of the pumping light, the selection rule is now MF = 0 and
the electric field couples the states F = 2, MF in the electronic ground state 3 2 S1/2
only with F = 3, MF in the excited 3 2 P3/2 state. However, spontaneous emission
nevertheless leads to changes in the population density among the MF states. This
is summarized in Fig. D.2.
To model this quantitatively, we first note that the coordinate systems chosen
above lead to a diagonal density matrix (9.17), here a (7 + 5) × (7 + 5) matrix. Thus,
we simply have to evaluate the population probabilities for ground and excited state
sublevels, w(MF , t) and w(MF , t). To follow the optical pumping process with
D.1 A Standard Case: Na(3 2 S1/2 ↔ 3 2 P3/2 ) 697
(a) during the first pumping (c) during the first pumping
cycle cycle
MF' = - 3 -2 -1 0 1 2 3 MF' =- 3 -2 -1 0 1 2 3
Fig. D.2 Population of the sublevels −3 ≤ MF ≤ 3 during HFS pumping of Na between
3 2 S1/2 (F = 2) and 3 2 P3/2 (F = 3) level (see e.g. H ERTEL and S TOLL 1974b). (a, b) linearly,
(c, d) circularly polarized pumping light; (a, c) during the first pumping cycle, (b, d) in the station-
ary limit
time t in detail one has to setup 12 − 1 rate equations and to solve them. Following
(10.43)–(10.44) they can be read from the term schemes Fig. D.1(b) and (c) for π
and σ + pumping, respectively. The required transition probabilities are according
to (4.113) and (4.123) in Vol. 1:
2
F 1 F
A F MF , F MF , q ∝ ωF3 F 2F + 1 F C1 F 2
MF q MF
3λ3
B F MF , F MF , q = B F MF , F MF , q = ba A F MF ; F MF .
4h
Since C1 acts only onto the spatial part of the state one may recouple the reduced
matrix element with (C.46) in Vol. 1 and obtains for the hyperfine transitions within
a fine structure transition:
2
F 1 F
B F MF , F MF , q ∝ 2F + 1 (2F + 1) . (D.1)
−MF q MF
In our case (F = 3, F = 2) this gives for the initial values of MF = −2, −1, 0, 1,
2 when exciting with circularly polarized light: B(MF , q = 1) = 1/3, 1, 2, 10/3, 5
and for linear polarization B(MF , q = 0) = 5/3, 8/3, 3, 8/3, 5/3. The rate equa-
tions are solved numerically. Two limiting cases are easily evaluated: during the
first excitation cycle the probability of finding excited states will be proportional to
the transition probabilities, we thus expect w(MF , t = +0) ∝ B(MF + q, q). This
is displayed in Fig. D.2(a) and (c). With little more effort one may derive the sta-
tionary limit of induced processes and spontaneous emission which is illustrated in
Fig. D.2(b) and (d) for the two pumping conditions.
Optical pumping has been developed to great perfection in modern laboratories.
A survey about the early developments of this method in atomic collision physics is
found in H ERTEL and S TOLL (1978).
698 D Optical Pumping
Tkq± (F ) = ρM M̃ F̃ M̃F |Tkq± (F )|F MF . (D.4)
F F
MF M̃F
+F
;
+F
2
a0 (F ) = T20 (F ) =
3MF − F F + 1 w(MF ) w(MF ).
MF =−F M=−F
(D.6)
The orientation parameter is simply the expectation value of the angular mo-
mentum with respect to the z-axis, −F ≤ o0 (F ) ≤ +F . For the alignment pa-
rameter one verifies easily that it disappears for equal population of the M states,
w(M) = 1/(2F + 1), and must take a value between the minimum and maximum
−F (F + 1) ≤ a0 (F ) ≤ F (2F − 1). In the present case with F = 3 we thus have
−12 ≤ a0 (F ) ≤ +15. These limiting values are, however, only reached for circular
polarized pumping in the stationary limit – as one can guess from Fig. D.2.
We have introduced the multipole moments Tkq± (F ) as expectation values of
certain combinations of the angular momentum operator J , here of the total angular
momentum F in the HFS coupling scheme (J I )F . Often this coupling scheme is,
D.2 Multipole Moments and Their Experimental Detection 699
however, without relevance for the physical process studied, e.g. since the interac-
tion with the nuclear spin is negligible – the nuclear spin behaves, so to say just as
a spectator. But also the angular momentum J is a composite quantity, in RUSSEL -
S AUNDERS coupling (LS)J , again composed of the orbital angular momentum L
and the electron spin S. The latter often is also a spectator only. The key interac-
tion – e.g. in a collision process – may just be determined by the electron charge
distribution and its orbital angular momentum, in summary it is specific to L. Often
one is thus only interested in multipole moments which are defined by the compo-
nents of
L. As described in Appendix C.2, one may derive these from the Tkq (F )
by reduction with the help of the W IGNER -E CKART theorem. With the respective
reduced matrix elements one obtains:
LTk L
Tkq± (L) = Tkq± (F ) . (D.7)
F Tk F
We also may write the parameters of interest here as:
z |F M = L M̃ |L
o0 (L) = F M̃F |L
F L z |L ML (D.8)
= T10 (L) = T10 (F ) /3 (D.9)
2z −
a0 (L) = F M̃F |3L
2 2z −
L |F MF = L M̃L |3L
2
L |L ML (D.10)
= T20 (L) = T20 (F ) /15. (D.11)
The second equality in both, (D.8) and (D.10), holds since the choice of the ba-
sis cannot influence the result of the averaging procedure. The numerical factors
in (D.9) and (D.11) are derived from detailed evaluation of the reduced matrix ele-
ments (D.7) as outlined in Appendix C.3 (see also F ISCHER and H ERTEL 1982).
As described in Sect. 9.4 one may experimentally determine the a0 (F ) and o0 (F )
parameters by measuring the fluorescence emitted from the excited Na(3 2 P3/2 ) – in
several geometries. Figure D.3 shows the results of such a measurement from an
Na atomic beam excited by the pump process as characterized in Figs. D.1 and D.2
(F = 3). The parameters a0 (L) and o0 (L) (renormalized for the L basis) are shown
together with the excitation density Nex as a function of the pump laser intensity.
The figure documents that indeed, by optical hyperfine pumping one may achieve
considerable electronic alignment and orientation. Excitation, alignment and orien-
tation is obtained with particularly high efficiency when pumping with circularly
polarized light: one obtains at sufficiently high intensity nearly the maximum val-
ues of o0 (L) = 1 and a0 (L) = 1 – the latter corresponds to the oblate shaped charge
distribution, which we have already mentioned often here, symmetric around the
wave vector of the σ + light (note the footnote 4 in Appendix F, Vol. 1). Even at low
pump intensity substantial values are reached (the red numbers at the scales indicate
the values expected for o0 (L) and a0 (L) during the first pumping cycle as illustrated
in Fig. D.2(d).
Clearly less efficient is pumping with linearly polarized light as documented by
Fig. D.2(b): the minimal value (at L = 1) of a0 (L) = −2 cannot be reached by HFS
700 D Optical Pumping
max
1.0
10 σ+ pump
(a) (c)
σ+ pump norm. 0.9
1.0
Nex / ar.un.
o1(L)
π pump 0.8
0.1
0.7
0.667 max
- 0.4 1.0 σ+ pump
π pump
- 0.480 0.8
- 0.5
(b) (d)
a0(L)
a0(L)
0.6
- 0.6 0.4
Fig. D.3 Experimentally determined data points for the 3 2 P3/2 F = 3 ↔ 3 2 S1/2 F = 2 pump
process in Na atoms. (a) Excitation probability, (b, d) alignment a0 (L), and (c) orientation pa-
rameter o1 (L) as a function of the laser intensity; grey: linearly polarized light (π pumping), red:
circularly polarized light (σ + pumping). The data are taken from F ISCHER and H ERTEL (1982),
the calculation (dashed red line) corresponds to the saturation profile according to (10.39) with
Is = 0.038 W cm−2
pumping due to the coupling of nuclear and electron spin. With the MF distribu-
tions according to Fig. D.2(b) one expects a minimum of a0 (L) = −0.667, which
is approached at sufficiently high intensity. Even the excitation density of 50 %
cannot be reached with linearly polarized light (albeit this cannot be seen directly
in Fig. D.2(a) since all experimental data have been normalized to the model at one
point as indicated). Clearly, there are deviations from the pure two level model when
pumping with linearly polarized light, which leads, as we shall see in a moment, to
substantial losses: D OPPLER broadening and a partial overlap of the F = 3 levels
with F ≤ 2 leads to population of the F = 1 ground state which cannot be excited
again by the one pump laser used in this experiment.
Obviously the initially assumed model of coupling the two preferential hyperfine
levels 3 2 P3/2 F = 3 ↔ 3 2 S1/2 F = 2 with just one laser frequency does not cor-
respond completely to reality. The HFS levels have a finite linewidth and – most
importantly – power broadening at higher intensities leads to population of other
D.3 Optical Pumping with Two Frequencies 701
hν1 hν1
1mm 1mm
Fig. D.4 Population density in a Na beam optically pumped with (a) one frequency on the
F = 2 → 3 transition and (b) with an additional laser tuned to F = 1 → 2 transition. The fluores-
cence signal from the excitation region has been recorded with a CCD camera and is proportional
to the excitation density. The data have been adapted from C AMPBELL et al. (1990)
excited HFS levels with F < 3. Hence, optical pumping via spontaneous emission
into the 3 2 S1/2 F = 1 ground state leads to substantial losses. The evident solu-
tion – we have mentioned it already in the main text (e.g. in Sect. 6.5.3) – is to
use two laser frequencies hν1 and hν2 for pumping, to empty just this particular
sink 3 2 S1/2 F = 1 again. Figure D.4 show the dramatic differences in both cases.
Plotted is the spatially (in 2D) resolved fluorescence from the excited Na(3 2 P3/2 )
state. The signal images the excitation density for excitation (a) with one and (b)
with two laser frequencies. The pump scheme is sketched in the respective insets.
In Fig. D.4(a) one observes a slightly asymmetric, so to say sluggish excitation pro-
file. The asymmetry is mainly due to the D OPPLER broadening in the wings of the
atomic beam due to which the ‘false’ HFS levels are excited, which relax immedi-
ately into the 3 2 S1/2 F = 1 ground state. Irradiation by the second laser frequency,
here propagating into opposite direction, leads to a massive enhancement of the ex-
citation density, which in addition is quite symmetric. Obviously the second laser
field manages efficiently to empty the sink in the ground state, thus supporting a
high excitation density.
Clearly, two separately tunable, well stabilized dye lasers imply a significant
experimental effort, which one certainly would like to reduce. C AMPBELL et al.
(1990) thus have developed a two mode laser specifically for this type of excitation.
It exploits neatly the spatial variation of the gain profile in the liquid jet of a dye
laser: the gain profile explained in Sect. 1.1.7 shows of course also a spatial variation
along the z-axis since laser modes are standing waves. In nodes of these standing
waves a second mode can be built up. By suitable dimensions of the laser resonator
and positioning of the gain medium one may indeed generate simultaneously two
laser modes which have a frequency distance corresponding to the HFS splitting of
the ground state in Na.
702 D Optical Pumping
With such a setup and one additional laser it was possible to excite Na in suf-
ficient density even in the 4D state and to study charge exchange processes with
K+ ions. Alternatively one may modulate the frequency of a single mode laser with
tunable, acousto-optical modulators, which oscillate at the frequency of the HFS
splitting of the ground state: the thus induced light modulation leads to correspond-
ing side bands impressed onto the carrier frequency. This again leads to effectively
two stable optical frequencies which can be tuned jointly as well as with respect to
each other. This method is used today preferentially since it is robust and stable. We
have quoted an example in Sect. 6.5.3.
CCD: ‘Charge coupled device’, semiconductor device typically used for digital
imaging (e.g. in electronic cameras).
FS: ‘Fine structure’, splitting of atomic and molecular energy levels due to spin
orbit interaction and other relativistic effects (Chap. 6 in Vol. 1).
HFS: ‘Hyperfine structure’, splitting of atomic and molecular energy levels due to
interactions of the active electron with the atomic nucleus (Chap. 9 in Vol. 1).
References
C AMPBELL, E. E. B., H. H ÜLSER, R. W ITTE and I. V. H ERTEL: 1990. ‘Near resonant charge trans-
fer in Na(4D) + K+ → Na+ + K∗ : optical pumping of the Na(4D) state and energy dependence
of rank 4 alignment’. Z. Phys. D, 16, 21–33.
F ISCHER, A. and I. V. H ERTEL: 1982. ‘Alignment and orientation of the hyperfine levels for laser
excited Na-atom beam I. The 3 2 S1/2 F = 2 ↔ 3 2 P3/2 F = 3 transition’. Z. Phys. A, 304, 103–
117.
H ERTEL, I. V. and W. S TOLL: 1974a. ‘A crossed beam experiment for the inelastic scattering slow
electrons by excited sodium atoms’. J. Phys. B, At. Mol. Phys., 7, 583–592.
H ERTEL, I. V. and W. S TOLL: 1974b. ‘Principles and theoretical interpretation of electron-
scattering by laser-excited atoms’. J. Phys. B, At. Mol. Phys., 7, 570–582.
H ERTEL, I. V. and W. S TOLL: 1978. ‘Collision experiments with laser excited atoms in crossed
beams’. In: ‘Adv. Atom. Mol. Phys.’, vol. 13, 113–228. New York: Academic Press.
K ASTLER, A.: 1966. ‘The N OBEL prize in physics: for the discovery and development of optical
methods for studying Hertzian resonances in atoms’, Stockholm. http://nobelprize.org/nobel_
prizes/physics/laureates/1966/.
Index of Volume 2
Symbols Benzene
12, 6 potential, 149 D6h point group, 262
H ÜCKEL orbitals, 277–285
A β (anisotropy) parameter, 356, 360, 551, 557
A BBE sine law, 679 B ETHE formula, 528–530, 534, 537, 538, 546
Adiabatic representation, 478–480 ionization, 537
A IRY diffraction pattern, 25 B ETHE integral, 526
A IRY disc, 26 B ETHE ridge, 546, 548, 549, 553
Alignment, 596, 604, 616, 619, 698–700 B ETHE surface, 545
angle, 36, 44, 605, 610, 611 Binary peak in (e, 2e) process, 553
parameter, 622, 623, 698 Birefringence, 38, 40, 41
Alkali halide potentials, 220–224 B IRGE -S PONER plot, 177
Ammonia maser, 1, 252 Black body radiation, 105
Amplification profile, 15 B OLTZMANN
Amplified spontaneous emission, 11 distribution, 155, 173, 335, 389, 591
Anharmonicity constant, 162 statistics, 158
Anti-S TOKES lines, 334 Bond order, 195
Anti-symmetrization, 521 Bonding orbital, 183
Antibonding orbital, 183 B ORN approximation
Ar2 TPES spectrum, 369 first order, elastic, 444–448
Arn clusters, 400 generalized oscillator strength (GOS),
Atomic form factor 530–534
inelastic, 527 inelastic collisions, 460, 461, 522, 525–530
AUGER electron, 533 integral inelastic cross sections, 534
AUTLER -T OWNES effect, 637–639 B ORN phase shift, 447
Autocorrelation function, 53–60 B ORN series, 553
Avoided crossing, 220–224, 263 B ORN -O PPENHEIMER approximation,
139–151
B collision processes, 458, 474
Baseline in interferometry, 87, 91 BPP, see Beam parameter product
Beam divergence, 21, 22 Branching ratio, 653
Beam parameter, 17–21
product, 26, 34, 35 C
Beam radius, 17–35, 650 C6 H +
4 absorption spectra, 328
Beam waist, 17–21 C2 DFWN spectrum, 354
and coherence volume, 87 Carrier envelope, 46
and lateral coherence radius, 91 phase, 46, 55, 188
theorem for F OURIER transform, 645 Density of states, 21–24, 238, 240
VOIGT profile, 628, 629 particle in a box, 105, 106
C OOPER minimum, 260, 529 Detailed balance
Coordinate rotation, 575, 605–611 principle of, 184
exercise, 207 Detuning, 229
Coordinate system Diamagnetism, 396–398
atomic vs. photon, 174 Dielectric function, 57
cartesic and polar, 108 Diffraction
electron vs. nucleus, 295 D EBYE-S CHERRER, 58, 59
Core electron, 148, 151 experiment, 60
Core potential, 149 He scattering, 60
alkali atoms, 165 image, 42
Correlation function LEED, 59
1st order, 625 Dimensional analysis, 119, 192, 554, 556, 661
Correspondence principle, 127 Dipol vector, 183
C OULOMB gauge, 631 Dipole approximation, 189, 631–640
C OULOMB integral, 356, 359, 365, 506 electric, 635
C OULOMB law, 50, 59 magnetic, 250–254
C OULOMB potential, 119, 121, 123 Dipole excitation, linear combination of states,
H atom, 121 217–225
screened, 148, 149 Dipole matrix element, 193
C OULOMB wave, 262 length approximation, 193, 636
Coupling
velocity approximation, 193, 636
jj , 301, 354, 360
Dipole operator, 594
LS, 301, 307, 354, 360, 361
magnetic field, 386
LS vs. jj , 512–514
multi-electron system, 264, 363
RUSSEL -S AUNDERS, 301, 354, 360, 361,
Dipole oscillator, classical, 182
508
Dipole radiation
break down, 362
angular characteristic, 203–212
spin-orbit, see Spin-orbit
Dipole transition
Crystal lattice, 40
C URIE constant, 396 amplitude, 193
C URIE’s law, 396 E1 transitions in the H atom, 201–203
Cyclotron frequency, 54, 332 in He, 362–365
electron, 79 length approximation, 636
Cylindrical capacitor, 52 matrix element, 190, 635
operator, 190, 193, 263
D selection rules for E1 transitions,
Damping constant, 229 196–203
DARWIN term, 305 selection rules for E1-transition, 202
D E B ROGLIE, 57, 89 velocity approximation, 636
wavelength, 57–59 D IRAC delta function, 644
Decay constant, 15 D IRAC equation, 79, 93, 296, 303, 333
Decay rate, 15 Direct product, 578
Degeneracy, 21, 22 Dispersion, 421
, 123 anomalous, 422
m, 111 close to resonance, 232, 233
, removal of, 137–146 normal, 420, 422
m, removal of, 130, 131 Dispersion relation, 131–134
Degree of coherence matter waves, 94
1st order, 625 Displaced terms
Degrees of freedom, 19 alkaline earth term schemes, 372
Density distribution, 125 C atom, 515
Density functional theory, 510–512 Ne atom, 518
718 Index of Volume 1
319, 342, 347, 349, 351, 371, 373, Nuclear quadrupole moment, 447, 449, 616
431, 514, 522, 525, 529, 551, 629 oblate or prolate, 478
NMR spectroscopy, 487–491 Nuclear radius, 29, 121
apparatus, 488 liquid drop model, 476
CW spectrum of ethanol, 488 Nuclear spin, 449
magic angle spinning (MAS), 490 eigenvalue equations, 449
occupation probability of levels, 489 Nuclear spin resonance, see NMR
N OBEL prize in chemistry Nucleons, 48
Richard R. E RNST (1991), 490
KOHN and P OPLE (1998), 510 O
F ENN, TANAKA, W ÜTHRICH (2002), 490 O atom, G ROTRIAN diagram, 516
N OBEL prize in physics Oblate, 477, 478, 619
Wilhelm C. RÖNTGEN (1901), 530 nuclear shape, 478
L ORENTZ and Z EEMAN (1902), 377 Observable, 97–99, 118
Joseph J. T HOMSON (1906), 50 commuting, 101
Albert A. M ICHELSON (1907), 278 non-commuting, 101
Max K. E. L. P LANCK (1918), 31 simultaneous measurement, 100, 101
Johannes S TARK (1919), 399 One electron cyclotron oscillator, 334
Albert E INSTEIN (1921), 27, 31 One particle problem, 117–134
Niels H. D. B OHR (1922), 64 One sided exponential distribution, 653
C ORNELL, K ETTERLE, W IEMAN (1925), One-loop QED effects, 323, 324
25 Operator, 96, 97, 100
Arthur H. C OMPTON (1927), 28 energy, 101
Louis DE B ROGLIE (1929), 57 momentum, 101
Werner K. H EISENBERG (1932), 100 position in space, 101
S CHRÖDINGER and D IRAC (1933), 90 simultaneous measurement, 101
Otto S TERN (1943), 70 Optical path difference, 40
Isidor I. R ABI (1944), 482 Orbital angular momentum, 71, 295, 299,
Wolfgang PAULI (1945), 22, 138 302
Max B ORN (1954), 88, 89 components, 110, 111
L AMB and K USCH (1955), 317 eigenfunctions, 109–113
T OMONAGA, S CHWINGER, F EYNMAN square, 111, 112
(1965), 79, 324, 534 Orbital energies, 509
R AMSEY, D EHMELT, PAUL (1989), 55, Orders of magnitude, 5–9
225, 288, 332, 482 energy scales, 7
G LAUBER, H ALL, H ÄNSCH (2005), 248, length scales, 5
319 time scales, 7
M ATHER, S MOOT (2006), 6, 7 Orthonormality relation, 96
E NGLERT and H IGGS (2013), 46 Oscillator strength, 238, 239, 636–640
N OBEL prize in physiology or medicine density, 256
L AUTERBUR, M ANSFIELD (2003), 490 sum rule, 239, 639, 640
Non-crossing rule, 391, 394
Non-local potential, 507 P
Non-stationary problems Pair production, 526
dipole excitation (E1), 169–225 Paramagnetism, 394–396
light matter interaction, 227–270 Parity, 593, 594
Non-stationary states, 186, 187 conservation in E1 transitions, 202
‘normal’ Z EEMAN effect, 128–131, 382–386 multi-electron systems, 594–603
Nuclear gN factor, 447, 449 Parity violation, 249
Nuclear magnetic moment, 447, 449, 485, 487, Particle detection, 75
488 Particle diffraction, 58–61
Nuclear magnetic resonance, see NMR C60 , 60, 61
Nuclear mass, 119 He atoms, 59, 60
energy correction, 68, 69 Particles and waves, 57, 64
724 Index of Volume 1