A. H. Kitai (Auth.), A. H. Kitai (Eds.) - Solid State Luminescence - Theory, Materials and Devices-Springer Netherlands (1993)

Solid State Luminescence
Solid State Luminescence

Theory, materials and devices
Edited by
A.R. Kitai
Department of Materials Science and Engineering and Engineering Physics
McMaster University, Ontario, Canada
I~nl SPR1NGER-SCIENCE+BUSINESS MEDIA, B.V.

PubIished by Cbapman & HalI, 2-6 Boundary Row, London SEI 8HN
Chapman & Hali, 2-6 Boundary Row, London SEI 8HN, UK

Blackie Academic & Professional, Wester Cleddens Road, Bishopbriggs,
Glasgow G64 2NZ, UK
Chapman & Hali Inc., 29 West 35th Street, New York NY1000l, USA
Chapman & Hali Japan, Thomson Publishing Japan, Hirakawacho Nemoto
Building,6F, 1-7-11 Hirakawa-cho, Chiyoda-ku, Tokyo 102, Japan
Chapman & Hali Australia, Thomas Nelson Australia, 102 Dodds Street,
South Melboume, Victoria 3205, Australia
Chapman & Hali India, R. Seshadri, 32 Second Main Road, CIT East,
Madras 600 035, India
First edition 1993

1993 Springer Science+Business Media Dordrecht
Originally published by Chapman & Hali in 1993
Softcover reprint of the hardcover 1st edition 1993
Typeset in 10/12 Times by Interprint Limited, Malta
ISBN 978-94-010-4664-0 ISBN 978-94-011-1522-3 (eBook)
DOI 10.1007/978-94-011-1522-3
Apart from any fair dealing for the purposes of research or private study, or
criticism or review, as permitted under the UK Copyright Designs and
Patents Act, 1988, this publication may not be reproduced, stored, or
transmitted, in any form or by any means, without the prior permission in
writing of the publishers, or in the case of reprographic reproduction only in
accordance with the terms of the Iicences issued by the Copyright Licensing
Agency in the UK, or in accordance with the terms of Iicences issued by
the appropriate Reproduction Rights Organization outside the UK.
Enquiries conceming reproduction outside the terms stated here should be
sent to the publishers at the London address printed on this page.
The publisher makes no representation, express or implied, with regard to
the accuracy of the information contained in this book and cannot accept
any legal responsibility or Iiability for any errors or omissions that may be
made.
A catalogue record for this book is available from the British Library
Library of Congress Cataloging-in-Publication data available.
@! Printed on permanent acid-free text paper, manufactured in accordance

with the proposed ANSIjNISO Z 39,48-199X and AN SI Z 39.48-1984
Contents
List of contributors Vll

Preface ix
Principles of luminescence 1
A.H. Kitai
2 Luminescent centres in insulators 21
G. Blasse
3 Luminescence spectroscopy 53
U. W. Pohl and H.-E. Gumlich
4 One-photon rare earth optical transitions: recent theoretical
developments 97
G. W. Burdick and M.e. Downer
5 Thin film electroluminescence 133
G.O. Muller
6 Powder electro luminescence 159
S.S. Chadha
7 Thin film electroluminescence devices 229
R Mach
8 Light emitting diodes: materials growth and properties 263
S.P. DenBaars
9 Atomic layer epitaxy of phosphor thin films 293
B.W Sanders
10 Lamp phosphors 313
T.E. Peters, RG. Pappalardo and RB. Hunt, Jr
11 Phosphors for other applications 349
G. Blasse
Index 373
v
Contributors
G. Blasse
Debye Research Institute
University of Utrecht
Utrecht
The Netherlands
G.W. Burdick
The University of Texas at Austin
Texas
USA
S.S. Chadha
School of Biological and Chemical Sciences
University of Greenwich
London
UK
S.P. DenBaars
Materials Department
University of California, Santa Barbara
California
USA
M.C. Downer
The University of Texas at Austin
Texas
USA
H.-E. Gumlich
Technical University of Berlin
Berlin
Germany
Vll
Vlll Contributors
R.B. Hunt, Jr
GTE Laboratories Inc., Danvers
Massach usetts
USA
A.K. Kitai
Departments of Materials Science and Engineering Physics
McMaster University, Hamilton
Ontario
Canada
R. Mach
Central Institute for Electron Physics
Berlin
Germany
G.O. Muller
Central Institute for Electron Physics
Berlin
Germany
R.G. Pappalardo
GTE Laboratories Inc., Danvers
Massachusetts
USA
T.E. Peters
GTE Laboratories Inc., Chelmsford
Massachusetts
USA
U.W. Pohl
Technical University of Berlin
Berlin
Germany
B.W. Sanders
Institute for Environmental Chemistry
National Research Council Canada, Ottawa
Ontario
Canada
Preface
Historically, black body radiation in the tungsten filament lamp was our
primary industrial means for producing 'artificial' light, as it replaced gas
lamps. Solid state luminescent devices for applications ranging from lamps
to displays have proliferated since then, particularly owing to the develop-
ment of semiconductors and phosphors. Our lighting products are now
mostly phosphor based and this 'cold light' is replacing an increasing
fraction of tungsten filament lamps. Even light emitting diodes now chal-
lenge such lamps for automotive brake lights.
In the area of information displays, cathode ray tube phosphors have
proved themselves to be outstandingly efficient light emitters with excellent
colour capability. The current push for flat panel displays is quite intense,
and much confusion exists as to where development and commercialization
will occur most rapidly, but with the need for colour, it is now apparent that
solid state luminescence will play a primary role, as gas phase plasma
displays do not conveniently permit colour at the high resolution needed
today. The long term challenge to develop electroluminescent displays
continues, and high performance fluorescent lamps currently illuminate
liquid crystal monochrome and colour displays. The development of tri-
component rare earth phosphors is of particular importance.
This book begins with a chapter on the physics of luminescence, covering
the classical and quantum mechanical theory of radiation in atoms and
solids. Chapter 2 focuses on phosphors, describing the fundamental pro-
cesses and models that are useful to sort out the rather complex electronic
and vibrational interactions. Chapter 3 deals with state of the art experimen-
tal methods and gives examples of fundamental luminescence processes in
solids. Chapter 4 presents a current understanding of rare earth ions in
crystalline fields.
The next section of the book focuses on different device and material
types: thin film electroluminescence is described from a physics perspective
in Chapter 5 and Chapter 6 traces the rather painful but important
development of powder phosphors for electroluminescence.
IX
x Preface
Chapter 7 presents the current state of thin film electroluminescence, and

Chapter 8 is devoted to both the theory and technology involved in light
emitting diodes which are now reaching new levels of efficiency.
Chapter 9 discusses, in some detail, the concept of kinetically limited
growth for thin films which allows large area, uniform coverage of both
phosphor and semiconductor materials.
Finally Chapters 10 and 11 present a very comprehensive account of
phosphors for fluorescent lamps and special applications in for example,
medicine.
A.H. Kitai
Ontario
Canada
1
Principles of luminescence
A.H. Kitai
1.1 INTRODUCTION
Technologically important forms of luminescence may be broken into
several categories, as shown in Table 1.1. Although the means by which the
luminescence is excited varies, all luminescence is generated by means of
accelerating charges. The portion of the electromagnetic spectrum visible to
the human eye has wavelengths from 400 to 700 nm. The evolution of the
relatively narrow sensitivity range of the human eye is a complex subject,
but is intimately related to the solar spectrum, the absorbing behaviour of
the terrestrial atmosphere, and the reflecting properties of organic materials,
green being the dominant colour in nature and, not surprisingly, the
wavelength at which the human eye is most sensitive. In this chapter, we
cover the physical basis for radiation and radiation sources in solids that
produce visible light.
Table 1.1 Luminescence types, applications and typical efficiencies (visible output
power/electrical input power)
Luminescence type Typical application Luminous efficiency
Blackbody radiation Tungsten filament lamp -5%

Photoluminescence Fluorescent lamp -20%
Cathodoluminescence Television screen -10%
Electrol uminescence Light-emitting diode, 0.1-50%
flat panel display
1.2 RADIATION THEORY

A stationary point charge has an associated electric field E as shown in Fig.
1.1. A charge moving with uniform velocity relative to the observer gives
rise to a magnetic field as shown in Fig. 1.2.
1
2 Principles of luminescence
Fig. 1.1 The lines of electric field E due to a point charge q.
Fig. 1.2 The lines of magnetic field B due to a point charge q moving into the page
with uniform velocity.
Both electric and magnetic fields store energy; the total energy density is
given by
It is important to note that the energy density moves with the charge so long
as the charge is either stationary or undergoing uniform motion; this is
evident since a new reference frame may be constructed in which the
observer is stationary with respect to the charge.
For an accelerated charge, however, energy continuously leaves the
charge to compensate exactly the work done in causing the charge to
accelerate. Consider the charge q in Fig. 1.3. Initially at rest in position A,
it accelerates to position B and then stops there. The electric field lines now
emanate from position B, but would, further out, have emanated from
position A, since the field lines cannot convey information about the
location of the charge at speeds greater than the velocity of light c. This
results in kinks in the lines of electric field which propagate away from q
with velocity c. Each time q accelerates, a new series of propagating kinks
is generated. Each kink is made up of a component of E that is transverse
to the direction of expansion, which we call E 1-' If the velocity of the charge
during its acceleration does not exceed a small fraction of c, then for r large,
qa . ()
E1- = 2 sm
4n8oc r
Radiation theory 3
-
a
Fig. 1.3 Lines of electric field emanating from an accelerating charge. (After
Eisberg and Resnick [1].)
Here, a is acceleration, and r is the distance between the charge and the
position where the electric field is evaluated. The strongest transverse field
occurs in directions normal to the direction of acceleration, as suggested by
Fig. 1.3.
Likewise, a transverse magnetic field B 1- is generated during the acceler-
ation of the charge as shown in Fig. 1.4, given by
Iloqa .
B1-=--sm8
4ncr
Fig. 1.4 Lines of magnetic field B emanating from an accelerating charge. B is

perpendicular to the page.
The two transverse fields propagate outward with velocity c each time q
undergoes an acceleration, giving rise to the electromagnetic radiation
whose frequency matches the frequency with which q acclerates. Note that E.l
and B.l are perpendicular to each other. The energy density of the radiation is
.1. 2 1 2
Iff = 21::0E.l + -2 B.l
f.lo
The Poynting vector or energy flow per unit area (radiation intensity) is
1
S=-E.l xB.l
f.lo
2 2
-q-=---a-=--=-sin 2 0 f
16nl::oc 3 r 2
where f is a unit radial vector.

Maximum energy is emitted in a ring perpendicular to the direction of
acceleration, and none is emitted along the line of motion. To obtain the
total radiated energy per unit time or power P leaving q due to its
acceleration, we integrate S over a sphere surrounding q to obtain
f
P= S(O) dA= f: S(0)2nr2 sin 0 dO
since dA is a ring of area 2nr 2 sin 0 dO.

Substituting for S(O), we obtain
1.3 SIMPLE HARMONIC RADIATOR

If a charge q moves about the origin of the x-axis with position x = A sin wt
then we can easily calculate the average power radiated away from the
oscillating charge. Note that
and
2q2 A 2W4 sin 2wt
P=----,----:;,---
4nl::03c 3
Now, average power P is the root-mean-square power which gives
_ q2 A2w4
P=-=-----;;- (1.1)
4nl::03c 3
Quantum description 5
If we now consider that an equal and opposite stationary charge - q is
located at x=o then we have a dipole radiator with electric dipole moment
of amplitude p = qA. Now we may write
_ p2W 4
P=---:o
12n80c3
Non-oscillatory radiation does exist also; the synchroton radiation source is
an example of a radiator that relies on the constant centripetal acceleration of
an orbiting charge. Quadrupole and higher-order poles may exist even in the
absence of a dipole moment; however, they have lower rates of energy release.
1.4 QUANTUM DESCRIPTION

A charge q (quite possibly an electron) does not exhibit energy loss or radiation
when in a stationary state or eigenstate of a potential energy field. This
requires that no net acceleration of the charge occurs, in spite of its uncertainty
in position and momentum dictated by the Heisenberg uncertainty principle.
Experience tells us, however, that radiation may be produced when a charge
moves from one stationary state to another; it will be the purpose of this
section to show that radiation may only be produced if an oscillating dipole
results from a charge moving from one stationary state to another.
Consider a charge q initially in stationary state I/In and eventually in state I/In'.
During the transition, a superposition state is created which we shall call 1/1.:
where a and b are time-dependent coefficients. Initially, a = 1, b = and

finally, a=O, b= 1.

Quantum mechanics allows us to calculate the expected value of the
position (r) of a particle in a quantum state. For example, for stationary
state I/In,
provided I/In is normalized, and V represents all space. Since, by definition

II/Inl 2 is not a function of time because I/In is a stationary state, the answer to
this integral is always time independent and may be written as ro, Note that
the time dependence of a stationary state is given by le(iE/h)tI2=
e(iE/h)te(-iE/h)t = 1. If we now calculate the expectation value of the position of
q for the superposition state 1/1., we obtain
(r). = (al/l n+ bl/l n,Irl al/l n+ bl/l n,)
= Ia 12 (1/1 nIr 11/1 n) + b2 <1/1 n' Ir 11/1n' ) + a* b< 1/1 nIr 11/1 n' )
+ b*a( 1/1n' Ir II/In)
We let
where n is the spatially dependent part of "'n' Hence
since the position must be a real number. This may be written as

<r(t)s = 21 a*b< n Irln') 1cos(wnn' t + b)
=2Irnn'l cos(wnn't+b) (1.2)
Note that we have introduced the relationship E = hw that defines the
energy of one photon generated by the charge q as it moves from "'n to
"'n" Note also that <r(t) is oscillating with frequency W nn ' = (En - En,)/h
such that the required number of oscillations at the required frequency
releases one photon having energy E = hw nn , from the oscillating charge.
The term r nn' also varies with time, but does so slowly compared with
the cosine term. Consider that an electron oscillates about x = 0 with
amplitude A = 1 A to produce a photon with A= 550 nm. From equation
(1.1 ),
_ (16 X 10- 19 )2 x (10- 10 )2 X (2n)4 x (3 x 108 )
p= . =4xlO- 12 w
12n(8.85 x 10- 12 )(5.5 x 10- 7 )4
since
2nc
w=T
One photon of this wavelength has energy E = hel A= 3 x 10 - 19 1. Hence, the

approximate length of time taken to release the photon is (3 x 10- 19 J)j
(4 x 10 - 12 J s - 1) = 7.7 x 10 - 8 s. Since the period of electromagnetic oscilla-
tion is T= Alc = 1.8 x 10- 15 s, approximately 10 7 oscillations take place. We
have assumed 1r nn' 1 to be a constant which will be shown not to be the case
in a later section.
We may define a photon emission rate Rnn , of a continuously oscillating
charge. We use equations (1.1) and (1.2) and E=hw to obtain
P q2w 3
R nn ,= -h = 3 3hlrnn'12 photonss- 1
W neoe
Selection rules 7
1.5 SELECTION RULES

A particle cannot change quantum states without conserving energy. When
energy is released as electromagnetic radiation, we can determine whether or
not a particular transition is allowed by calculating the term Ir nn' I, and
seeing whether it is zero or non-zero. The results over a variety of possible
transitions give selection rules that name allowed and forbidden transitions.
The transitions involved in the hydrogen atom are of particular import-
ance. We will now derive the well-known selection rules for the electron in
hydrogen states, or more generally in one-electron atomic states. We use
polar coordinates and begin by calculating r nn'.
rnn,=(nlrln')=
J
r
all space
I/I:rl/ln,dV
Note that since we are working in three dimensions, we must consider r in

vector form, and let I/I(r, (), )=Rn(r)8 Im (())tP m(). Now,
r nn ,= tXl Rn,(r)r Rn(r)dr[t

3
2
1[ f: 8I'm,(())8 Im (()) sin ()rtP!()tPm,()d()d ]
The term in brackets may be broken up into orthogonal components of unit

vector r = sin () cos x + sin () sin y + cos () z to obtain three terms:
(1.3)
since tPm()=e im </>, the three integrals in may be written
11 = f 0
21[
cos ei(m-m')</> d
12 = f 0
21[
sin ei(m -m')</> d
13= f 0
21[
ei(m-m')</> d
13 is zero unless m'=m.

11 may be written:
which is zero unless m' = m 1. 12 gives the same result.

Now consider the integrals in () which multiply 11 ,12, and 13 , We shall
name them J 1, J 2, and J 3' If 13 is non-zero, then m' = m. Hence we obtain
The integral
is zero unless l' = 1, a property of the associated Legendre polynomials

which, being eigenfunctions, are orthogonal to each other [2]. Since cos () is
an odd function over the range 0 ~ () ~ n, the parity is reversed in J 3 and
hence J 3 =0 unless l' = 1 1.
If 11 is non-zero, then m' = m 1. Hence, we obtain
Using the properties of associated Legendre polynomials once again, we

note that it is always possible to write B'm(())=aBI-1.m+1(())+bB,+1.m+1(()),
where a and b are constants. Choosing m' = m + 1, we obtain
For a non-zero result, l' = 1 1 using the orthogonality property. The

same conclusion obtains from the case m' = m - 1 and from the J 2 integral.
We have therefore shown that the selection rules for a one-electron atom are
L1m=O, 1 and Al= 1
Note that we have neglected spin-orbit coupling here. Its inclusion would
give
L11= 1 and L1j=O, 1
Selection rules do not absolutely prohibit transitions that violate them,
but they are far less likely to occur. Transitions may take place from
oscillating magnetic dipole moments, or higher-order electric pole moments.
Einstein coefficients 9
These alternatives are easily distinguished from allowed transitions since
they occur much more slowly, resulting in photon release times of mil-
liseconds to seconds rather than nanoseconds as calculated earlier. It is
important to realize that practical phosphors having atomic luminescent
centres often release photons via 'forbidden' transitions. The surrounding
atoms in a crystal may lift the restrictions of ideal selections rules because
they lower the symmetry of atomic states.
1.6 EINSTEIN COEFFICIENTS

Consider that an ensemble of atoms has electrons in quantum states k of
energy Ek which may make transitions to states 1 of energy E, with the
release of photons (see Fig. 1.5).
many atoms
many atoms
Fig. 1.5 The decay of an electron from state k to state I results in the release of a
photon.
In order to begin making such transitions, something is needed to perturb

the electrons in states k, otherwise they would not initiate the transitions,
and would not populate superposition states I/Is. The study of quantum
electrodynamics shows that there is always some electromagnetic field
present in the vicinity of an atom at whatever frequency is required to
induce the charge oscillations, and to initiate the radiation process. This is
because electromagnetic fields are quantized and hence a zero-point energy
exists in the field. We call this process spontaneous emission.
Alternatively, the transition may be initiated by applied photons (an
applied electromagnetic field) which gives rise to stimulated emission. It is
also possible to excite electrons in state 1to state k using photons of suitable
energy.
These ideas may be summarized as follows. The rate at which atoms in the
Ek state decay is Wkl . This is proportional to the number of photons of
frequency (J) supplied by the radiation field, which is proportional to photon
energy density u(v) and to the number of atoms in the Ek state. The
spontaneous process occurs without supplying radiation, and hence its rate
is determined simply by the number of atoms in the Ek state, N k We may
write
(1.4 a)
The proportionality constants A and B are called the Einstein A and B

coefficients, and OJkl is the rate on a per atom basis.
Atoms in the El state may not spontaneously become excited to the Ek
state; however, photons of energy Ek-E 1 may be absorbed. Hence,
(1.4 b)
At this point, the idea of stimulated emission needs to be developed in
order to explain why transition rates are proportional to u(v). It is, however,
clear that Akl is simply another name for R nn " the photon emission rate, in
the case of dipole radiation.
1.7 HARMONIC PERTURBATION

Consider an atom possessing electron levels k and I that experiences a weak
electromagnetic field. By 'weak' we require that the potential energy experi-
enced by the electrons due to this field is small compared with the Coulomb
potential from the nucleus and other electrons. The total Hamiltonian is given
by the sum of the atomic term Ho(r) and the perturbation term H'(r, t):
H(r, t)=Ho(r)+H'(r, t) with H'(r, t)=H'(r)f(t)
If the field is turned on at t = 0 with frequency OJ then
'( ) {O
H r,t = t<O
2H'(r) cOSOJt t~O
Time-dependent perturbation theory [2J may be used to determine the

wavefunction that results from the perturbation which is harmonic in this
case. Assume the electron is initially in eigenstate t/llr, t). In general, if
t/lk(r, t) are all eigenstates of Ho(r) then the wavefunctions after the perturba-
tion term H'(r, t) is added will be of the form
t/I(r, t)= L Ck(t)t/lk(r, t)
k
where
and
t/I(r, t) = cPk(r )e iwkt
The probability of a transition from the initial eigenstate t/ll(r, t) to a new
eigenstate t/lk(r, t) is given simply by ICk(tW. We write
Pl_k(t)=ICk(t)12=(~~ly 1 t eiwkI''j(t')dt' 12
Harmonic perturbation 11
Because of the weak electromagnetic field, f(t)=2 cos wt, and therefore
= 2iHkl {e i (o)kl- W )t / 2 sin[(wkl-w)t/2] + ei (Wkl+ W )t / 2 Sin[(W k1 +W)t/2]}

h Wkl-W Wkl+W
Resonance occurs when Wkl = w. The two signs signify either a sti-
mulated absorption process (Wkl = w) or a stimulated emission process
(Wkl= -w) since energy is then released (Ekl negative). If Wkl~W,
(1.5)
The probability of the transition (stimulated emission or absorption) is

always proportional to 1Hkzl 2. P 1k is shown in Fig. 1.6, which should be
thought of as a graph that grows rapidly in height with time t. Note,
however, that being taller to begin with, the central peak grows faster than
the others with time, and the function resembles a delta function for long
time evolution. This is consistent with the uncertainty relationship
I1E M ~ h/2 since, as time increases, the uncertainty in energy approaches
zero.
The term IHkzl2 may be expressed in terms of the electric field E of the
electromagnetic perturbation. If p is the dipole moment of the electron as it
undergoes the lk transition, then H' = 1- P E 1ex 1E I.
-61t -41t -21t 21t 41t 61t

-t- t -t- -t- -t-
t
Fig. 1.6 Dependence of transition probability on Wk1-W as a result of harmonic
perturbation.
Since energy density u( v) is proportional to 1E 12, it is clear that

IHkI12OCU(v) and hence we have shown that the Einstein B coefficients must
be multiplied by u(v), as in equations (1.4 a,b).
When we wish to describe the time evolution of the rate of emission for an
ensemble of N atoms undergoing stimulated emission, we may use
. . -1 N1P 1k (transitions)
Wzk ( tranSItIons s ) = t (s)
Thus it is evident that when P 1k OCt 2, then the transition rate increases
linearly with time. This situation obtains for small t, since from equation
(1.5) we see that

P Ik OC 11m sin 2[!-(Wkl-W)tJ IHkti 2 1. 2
= 2 4t
n-+oo (Wkl-W)
Of course, for long times, W1k becomes constant as equilibrium is reached.
Note that Blk = Bkl since P1k=Pkl .
1.8 BLACKBODY RADIA nON

In an ensemble of electron states, in equilibrium, W1k = Wk!. However, the
spontaneous emission process may only take place in one direction, and we
can write
Therefore
NI Ak,+Bk1u(v)
Nk Bk1u(v)
and
Since the populations of atoms having excited states of certain energies

will obey Boltzmann statistics,
it follows that
AlB (1.6)
u(v)= ehrojkT -1
where subscripts have been dropped.

Blackbody radiation 13
Consider a cavity with metallic walls uniformly heated to temperature T.

If we could observe the cavity through a small hole through a cavity wall, we
would detect electromagnetic radiation due to the thermally agitated elec-
trons in the cavity walls.
For analysis, suppose the cavity is cubic with edge length a, and principal
axes x, y and z, as shown in Fig. 1.7. Since the cavity walls are electrically
conductive, the electric field in the radiation field must be zero at the cavity
walls, and, because of electromagnetic reflections at metallic surfaces,
standing waves only will exist in equilibrium. Hence, the E field for waves
travelling in the x-direction will be given by
E(x, t)=Eo Sine~x) sin(2;rrvt) where v=l
To satisfy boundary conditions, E(a, t)=O and therefore
Note that the frequencies are quantized and may be counted using integers
nx Similar expressions may be written for Ey and Ez Consider an artificial
z
>-~~--------~~y
x
Fig. 1.7 Cavity of cubic shape with edge length a. (After Solymar and Walsh, [3].)
space having axes (n x , ny, nz ). Such a space consists of a lattice of points,

each of which uniquely describes a particular three-dimensional radiation
pattern or mode. It is easy to show that all points (n x , ny, nz ) at a given
distance r = 2av/c from the origin represent standing waves of the same
frequency v, but along different directions within the cavity [1]. We can
then count the number of cavity modes between spheres of radii 2av/c
and 2a(v + dv)/c as shown in Fig. 1.8. Since each point occupies a unit
'volume', the number of points in the spherical shell is shell volume
4;rrrw 2 dr=4;rr(2a/c)3v2 dv. Since we wish only to consider positive values of
nz
/C-----h4-----1~ ny
nx
Fig. 1.8 Spherical shell enclosing points in (nx, ny, nz ) space lattice that represent
standing waves that range in frequency from v to v+dv. (After Solymar and Walsh
[3].)
n, we divide by 8 to count only one octant of the shell, and multiply by 2

because each standing wave has two possible polarizations. Hence, the
number of modes over frequency range dv is
N(v)= 8na 3 v2 dv= 8nV v2 dv

c3 c3
Because each mode has a degree of freedom, namely the choice of electric
field amplitude, on average, each mode will have the same energy E which,
from classical kinetic theory, is E = kT. Should one mode gain in E, it would
lose it owing to collisions of electrons in the cavity walls which would
transfer it to other modes. Therefore the energy per unit cavity volume over
the frequency interval dv may be expressed in terms of the energy density
u(v) as
8nv 2
u(v)dv=-3- kTdv (1.7)
c
This expression clearly differs from equation (1.6). This is because our
classical wave theory assumes that the energy of each cavity mode is
continuously variable as just stated, even though the allowed cavity modes have
discrete frequencies v. In our treatment leading to equation (1.6), however, we
treated the energy levels giving rise to modes at frequency v as discrete, such
that hv = 11E. Starting with lowest frequency mode, for example along the
x-direction, nx = 1 and V1 = c/2a. If nx = 2, V2 = cia. This implies a pair of
discrete energy levels, E1 =hV1 and E2=hv2 with difference I1E = hc/2a. So
long as I1E ~ kT, there is no real problem with the classical treatment;
however, for higher-order modes, or for lower temperatures, the energy
spacing between modes may by far exceed kT and it becomes essential to
Blackbody radiation 15
take the energy of each mode as discrete. Since equation (1.7) gives the
correct result for
() I' 8nv 23kT

Ilmuv=lm (1.8)
V-a) v-co C
we can now evaluate AlB in equation (1.6) by requiring that
I. AlB AlB 8nv 2 kT

1m hv/kT
v--+ooe -1 hvlkT c3
Therefore
and the final result, valid over all v and T, is Planck's famous blackbody
radiation energy density function
8nv 3
U(V)=-3-
c ehv/kT - 1
This is shown in Fig. 1.9 for three different temperatures.
12,0000 K
--.....
\r\
I \ / "
,
/ ........ 3,OoooK
I 'I
I
I
" ........
........
I "
I
I
~ ~ ~ ~ 1~ 1~ 1~ 1~ 1~ ~
Wavelength Injlm
Fig. 1.9 Blackbody radiation spectrum showing spectral power density for sources
at temperatures of 3000 K, 6000 K and 12000 K. Note that the 6000 K curve
matches the visible range best, and is similar to the solar spectrum. The three curves
are artificially normalized to appear identical in height.
For visible light sources, tungsten filament lamps which are blackbody
radiators are limited in filament temperature to somewhat below the melting
point of tungsten, or '" 3000 K. As is clear from Fig. 1.9, only a small
fraction of the area under the curve corresponding to this temperature is in
the visible range: the physical basis for the low efficiency of such lamps. A
considerable attenuation of short-wavelength (blue-violet) compared with
long-wavelength (red) visible is also evident. A lamp operating at 6000 K
would approximately match the sun's surface temperature (5700 K) and be
far more efficient. The tungsten halogen lamp allows for a modest gain in
performance over a regular tungsten lamp by chemically stabilizing the
tungsten filament, allowing for higher filament temperature.
1.9 DIPOLE-DIPOLE ENERGY TRANSFER

We can now explain how energy may be transferred from one atom to
another without the actual release of a photon.
Consider an excited atom, S, and a nearby unexcited, but otherwise
identical atom, A (see Fig. 1.10). As S radiates, it generates an oscillating
RSA ------I
Fig. 1.10 Excited atom S a distance RSA from atom A.
electric field E due to its oscillating dipole. This field falls off as l/r 3 [4]
and, provided that its energy does not have time to escape as a photon, it
will directly stimulate a transition in A by means of the same process
described by equation (1.4). Since E falls off as l/r 3 , then the energy density
in the electric field
U(V)=16 0 E 2
falls off as l/r6. From equation (1.4), therefore, the rate (or the probability)
of energy transfer from S to A depends on R;1. Energy transfer is discussed
further in Chapter 2.
1.10 ENERGY LEVELS IN ATOMS

It is always possible to formulate Schrodinger's equation to give the
energy levels of electron states for an atom. Consider an optically active
Crystal field splitting 17
atom in a crystal surrounded by a space lattice of atoms. The total

Hamiltonian is:
H tot = H isolated + H electrostatic lattice + H dynamic lattice
involves a Coulomb potential due to the atom nucleus and appro-

Hisolated
priate screening effects of inner shell electrons (Ho). The optically active
electrons are now affected by spin-orbit coupling (Hso), and LS coupling or
exchange energy (H c):
Hisolated = H 0 + Hso + H c
Ho is spherically symmetric and yields the one-electron atom states having
quantum numbers n, I, m, s. Spectroscopists use notation to describe the
shell according to n levels s, p, d, f, g that correspond to 1=0, 1, 2, 3, 4
respectively, which represent subshells. For example, a subs hell containing
five electrons with n = 3 and 1=2 would be written 3d 5. Such a subshell
exists in manganese.
Spin-orbit coupling requires the introduction of a new quantum number
j, and is caused by the magnetic moments due to electron orbit and electron
spin. In fact, j = I 1/2, which gives rise to a splitting of each energy level
into two levels unless 1=0 (no orbital magnetic moment).
In LS coupling, the effects of more than one electron within unfilled
subshells are considered. A Coulombic electron--electron interaction energy
term exists. The spin angular momenta of individual electrons add together,
as do orbital angular momenta, giving rise to a total spin s' and a total
orbital angular momentum I'. For example, consider an atom with configur-
ation 3d 14pl. There are, because of both Hso and H c , 12 levels in this case
[1] which are labelled 3D 3, 1F 3,1 P b etc. The superscript is 2s' + 1; the letter
designates I' (the same scheme as for I, but now using capital letters to
acknowledge addition of orbital angular momenta) and the subscript is j'
which is formed by adding s' and I' as vectors. Energy splittings occur since
the average separation and therefore Coulombic energy between electrons
depends upon the way in which angular momenta are added.
Additional complication arises if Hso and Hc are similar in strength. The
level splittings become more complex, and the LS coupling exclusion
principle forbids the existence of certain states.
1.11 CRYSTAL FIELD SPLITTING

When an atom S is placed in a crystal, it experiences the crystal field, or the
electric field due to surrounding atoms. We will assume that the electrons
involved in the energy levels of interest for luminescence are not involved in
bonding. This is generally true in practice. Because of crystal symmetry,
atoms surrounding S will give rise to an electric field with some symmetry.
For example, a tetrahedral crystal site will have tetrahedral symmetry; an
octahedral site will have octahedral symmetry. The electric potential of an

electron in atom S due to the crystalline environment may always be
expressed in the form
y'(8, )= L aim yr(8, )
I.m
where yr are the spherical harmonics. This is analogous to a Fourier series

expansion, but is specifically applicable to an atom or sphere surrounded
by a field that has angular dependence. The crystal field may now be
taken into account, using time-independent perturbation theory if Vs is
small. If we know the eigenstates of H o and add perturbation H', then, to
first order,
and
En=E~O) + <~O)IH'I~O
Note that the new eigenstates n and eigenenergies En are based on the
eigenstates ~O) and eigenenergies E~O) of the unperturbed system. To
account for the crystal field, we simply substitute H'(8, ) = Vs(8, ) and
determine the new states due to the crystal field.
If the symmetry of the crystal field is different from eigenstates involved,
which is usually the case, then degenerate states will probably split because
of the crystal field. This is known as crystal field splitting. In this case,
E ~O) - E lO) = 0 for some n, i, n 01= i. In order to determine the n and En, we
must first make the matrix element <lO) IH' I ~O zero whenever ElO) = E~O)
such that these singular terms vanish. This may be accomplished by
diagonalizing the submatrix of H;n which contains the degenerate states. The
energy splittings are then obtained directly.
For transition metal ions placed in a crystal field, the three-dimensional
states may be more influenced by the crystal field than by He. This is known
as the strong field scheme [5]. For such an ion in an octahedral crystal field,
for example, the five-fold degenerate three-dimensional orbitals split into a
two-fold degenerate eg state and a three-fold degenerate t 2g state. The energy
separation between the two states is called lODq where Dq is a parameter
that is determined by the crystal field strength. Now, He may be accounted
for in a manner analogous to free-ion LS coupling.
In Cr3+, for example, there is a 3d 3 configuration. If placed in an octa-
hedral crystal field, the splitting depends on the term Dq/B which is a
measure of the crystal field influence. A theoretical treatment [6] gives the
splittings. Results for Cr3 + are shown in Fig. 2.16.
Transitions that were forbidden in the free ion may become dipole
transitions with a crystal field. The lower symmetry allows a dipole moment
References 19
to exist, and these new transitions can take place, although with small rates.
Radiative lifetimes of 10- 3 s are not unusual for transition metal ions such
as Mn2+ in a tetrahedral crystal field.
REFERENCES
1. Eisberg, R and Resnick, R (1985) Quantum Physics of Atoms, Molecules, Solids,
Nuclei and Particles, 2nd edn, Wiley, New York.
2. Liboff, RL. (1980) Introductory Quantum Mechanics, Holden-Day.
3. Solymar, L. and Walsh, D. (1985) Lectures on the Electrical Properties of
Materials, 3rd edn, Oxford University Press.
4. Kip, A.F. (1969) Fundamentals of Electricity and Magnetism, 2nd edn, McGraw
Hill, New York.
5. Yen, W.M. and Selzer, P.M. (1986) Laser Spectroscopy of Solids, Vol. 49, 2nd edn,
Springer, Berlin.
6. Tanabe, Y. and Sugano, S. (1954) J. Phys. Soc. Japan, 9, 766.
2
Luminescent centres in insulators
G. Blasse
2.1 INTRODUCTION
The existence of optically active centres in solids, liquids and molecules is
now well known. Their properties have been studied intensively, not only
for fundamental reasons, but also in view of their potential applications.
Energy-saving fluorescent lamps, X-ray photography, and television display
tubes are well-known examples [1].
This can be hardly better illustrated than by the case of ruby
(AI 2 0 3 :Cr3+). The optical properties of ruby have been studied for over a
century starting with the work of Becquerel in 1867 [2] who excited ruby
with sunlight. He claimed that the properties of ruby were intrinsic, but it
soon became clear that they are due to the Cr 3 + ion that is an optically
active centre in the nonabsorbing Al 2 0 3 host. Later crystal field theory was
able to explain the spectroscopy of Cr 3 + in Al 2 0 3 in detail. The application
followed rather soon: the first solid state laser was based on a ruby crystal.
In the development of tunable infrared lasers the Cr 3 + ion plays an
important role [3,4].
The purpose of this chapter is to present the theories which are at present
in use to describe the luminescent properties of optically active centres in
solids. This will be done in such a way that even the unexperienced reader
can apply the theoretical results to practical cases. In order to do so, section
2.2 treats the interaction between an optical centre and its immediate
surroundings by presenting the configurational coordinate diagram. Radi-
ative as well as nonradiative transitions will be dealt with. In Section 2.3 we
will consider the mutual interactions between optical centres, and especially
those leading to energy transfer and energy migration. In section 2.4 we will
illustrate how theory applies to specific cases by discussing some carefully
selected case studies relating to important optically active centres.
For those who are not familiar with these types of phenomena, this
introduction gives now a schematic, first picture of the physical properties
of optically active centres. Figure 2.1 shows an optical centre (an ion or a
21
22 Luminescent centres in insulators
x M
q H
Fig. 2.1 Luminescence processes in a centre A in a solid: X, excitation; M, emission;

H, heat (nonradiative return to the ground state).
complex ion) in a solid or a liquid. The centre is irradiated. For simplicity

we assume that the surroundings do not absorb the irradiating light. The
centre shows optical absorption, so that it makes a transition from the
ground state to the excited state. If the irradiation is with visible light, the
sample is coloured. The excited state will ultimately return to the ground
state. This may occur by a nonradiative or a radiative process.
In the former case the energy of the excited state is used to excite the
vibrations of the surroundings (generation of heat). The latter case is known
as luminescence (see Fig. 2.1). In this field of research the irradiation is called
excitation. Usually the emission is situated at longer wavelengths than the
excitation. The energy difference between these two is called the Stokes shift.
The quantum efficiency (q) of the luminescence is the ratio of the number
of photons emitted and the number of photons absorbed. If there are no
competing nonradiative transitions, q = 1; if the non-radiative transitions are
dominating, q~O, and there is practically no emission.
A more complicated situation occurs if two (equal or unequal) centres are
close together (Fig. 2.2). The excited centre may transfer its excitation energy
to the neighbouring centre that is still in the ground state:
centre S * + centre A--+centre S + centre A * (2.1)
where the excited state is marked by an asterisk. This process may be
followed by emission from A or by a nonradiative decay on A. In the former
case we speak of sensitized emission (A is sensitized by S); in the latter case
A is called a quenching centre.
Before starting the presentation of the physical models in use to explain
the properties of optically active centres, it seems appropriate to mention a
few excellent literature reviews on our topic of discussion. A recent, clear
and rather elaborate discussion of the configurational coordinate diagram
is given in the book by Henderson and Imbusch [5]. This book deals also
with energy transfer. However, the report on the relevant Erice meeting [6]
The configurational coordinate diagram 23
x M
Fig. 2.2 Energy transfer between two centres in a solid. The excitation X excites
centre S which transfers its excitation energy to A (T). Finally A shows emission (M).
can also be used. A more chemical approach to these topics was presented
by ourselves [7,8].
2.2 THE CONFIGURATIONAL COORDINATE DIAGRAM

Let us consider a dopant ion in a host lattice and assume that it shows
luminescence on illumination. What we will have to discuss is the interac-
tion of the dopant ion with the vibrations of the lattice. The environment of
the dopant ion is not static: the surrounding ions vibrate about some
average positions, so that the crystalline field varies. The simplest model to
account for the interaction between the dopant ion and the vibrating lattice
is the single configurational coordinate model.
In this model we consider only one vibrational mode, i.e. the so-called
breathing mode in which the surrounding lattice pulsates in and out around
the dopant ion (symmetrical stretching mode). This mode is assumed to be
described by the harmonic oscillator model. The configurational coordinate
(Q) describes the vibration. In our approximation it represents the distance
between the dopant ion and the surrounding ions. In ruby this Q would be
the Cr 3 + _0 2 - distance.
If we plot energy versus Q we obtain for the electronic states parabolae
(harmonic approximation). This is presented in Fig. 2.3 for the electronic
ground state u and one electronic excited state v. Further, Qo represents the
equilibrium distance in the ground state, and Qb that in the excited state.
Note that in general these will be different. The u parabola is given by
(2.2)
where k is the force constant. Within the parabolae the (equidistant) vibra-
tional energy levels have been drawn. They are numbered by n=O, 1,2, ...
for the ground state parabola u, and m = 0, 1,2, .. , for the excited state
Q'
o
Fig. 2.3 A configurational coordinate diagram. The potential energy E is plotted

versus the configurational coordinate Q for the ground state u and an excited state
v. The equilibrium positions are Qo and Q~ respectively. Absorption (lines pointing
upwards) is at higher energy than emission (lines pointing downwards). The
absorption and emission band maxima correspond to the full lines. The thin
horizontal lines indicate the vibrational levels in the states u and v.
parabola v. The excited state parabola is drawn in such a way that the force
constant is weaker than in the ground state. Since the excited state is usually
more weakly bound than the ground state, this is a representative situation.
Optical absorption corresponds to a transition from the u to the v state
under absorption of electromagnetic radiation. Emission is the reverse
transition. Let us now consider how these transitions have to be described
in the configurational coordinate model. It is essential to remember that the
wavefunction of the lowest vibrational state (i.e. n=O or m=O) is Gaussian;
that is, the most likely value of Q is Qo (or Qo in the excited state). For the
higher vibrational states, however, the most likely value is at the edges of
the parabola (i.e. at the turning points, as in the classical pendulum).
The most probable transition in absorption at low temperatures is from
the n = 0 level in u, starting at the value Qo. Optical absorption corresponds
to a vertical transition, because the transition u--+v on the dopant ion occurs
so rapidly that the surrounding lattice does not change during the transition
(Born-Oppenheimer approximation). Our transition will end on the edge of
parabola v, since it is there that the vibrational states have their highest
amplitude. This transition, drawn as a solid line in Fig. 2.3, corresponds to
the maximum in the absorption band. However, we may also start at Q
values different from Qo, although the probability is lower. This leads to the
width of the absorption band, indicated in Fig. 2.3 by broken lines. It can
be shown that the probability of the optical transition between the n=O
vibrational level of the ground state and the mth vibrational level of the
excited state is proportional to
(2.3)
where r represents the electric dipole operator and Vm and Uo the vibrational
wavefunctions. The first term, the electronic matrix element, is independent
of the vibrational levels; the second term gives the vibrational overlap. The
transition from n = 0 to m = 0 does not involve the vibrations. It is called the
zero-vibrational transition (or no-phonon transition). Equation (2.3) shows
that the effect of the vibrations is mainly to change the shape of the
absorption line (or band), but not the strength of the transition (which is
given by the electronic matrix element).
What happens after the absorption transition? First we return to the
lowest vibrational level of the excited state; that is, the excited state v relaxes
to its equilibrium position, giving up the excess energy as heat to the lattice.
The system of dopant ion and surroundings is then in the relaxed excited
state. The emission transition can be described in exactly the same way as
the absorption transition. This is indicated in Fig. 2.3 in the same way as
for the absorption transition. Finally the system relaxes within the u
parabola to the lowest vibrational level.
If the temperature is not low, higher vibrational levels may be occupied
thermally, so that we start the process not only from n = 0, but also from
n = 1, and possibly from even higher levels. This leads to a further broaden-
ing of the absorption and emission bands, but does not change our
arguments essentially.
The emission transition will usually be situated at lower energy than the
absorption transition. This phenomenon is known as the Stokes shift. Only
the zero-vibrational transition is expected to occur at the same energy in the
absorption and emission spectra. The Stokes shift is a direct consequence of
the relaxation processes that occur after the optical transitions. It is obvious
that the larger Qo - Qo is, the larger the Stokes shift will be. If the two
parabolae have the same shape and vibrational frequency, it is possible to
define a parameter S (the so-called Huang-Rhys parameter) as follows:
tk(Qo - QO)2 = Shw (2.4)
where hw is the energy difference between the vibrational levels. The Stokes
shift is then given by
!lE. = k(Qo - Q)2 - hw = 2Shw (2.5)
340 310nm
(a)
500 600nm
(b)
ss
x M
300 500nm
(e)
The parameter S measures the interaction between the dopant ion and the
vibrating lattice. Equation (2.5) shows that, if S is large, the Stokes shift is
also large. Equation (2.4) shows that S is immediately related to the offset
of the parabolae in the configurational coordinate diagram (Fig. 2.3). This
offset, LlQ = Qo - Qo, may vary considerably as a function of the dopant ion
and as a function of the vibrating lattice, as we will see below.
It can be shown that the relative intensity of the zero-vibrational
transition (n=O~m=O) is exp( -S) [5]. We can now divide our lumines-
cent centres into three classes:
1. those with weak coupling (i.e. S < 1), so that the zero-vibrational transi-
tion dominates the spectrum;
2. those with intermediate coupling (i.e. 1 < S < 5), so that the zero-vibra-
tional transition is observable, but not the strongest line in the absorp-
tion or emission band;
3. those with strong coupling (i.e. S> 5), so that the zero-vibrational
transition is so weak that it is not observable in the spectra. (This case
is also characterized by large Stokes shifts.)
Figure 2.4 shows three emission spectra that are representative of the
three cases. Characteristic examples of case 1 are the trivalent rare earth
ions. The value of S is so small for these ions that the spectra consist in good
approximation of the zero-vibrational transitions only. Figure 2.4(a) gives as
an example the emission spectrum of the Gd3+ ion in LaB 3 0 6 It consists
of one strong electronic line at about 310 nm, whereas the weak repetition
at about 325 nm is a vibronic transition. Actually the energy difference
between these two lines corresponds to the vibrational stretching frequency
of the borate group in LaB 3 0 6 .
A characteristic example of case 2 is the uranyl ion (UO~ +). The m = 0--+
n = 2 line dominates in the spectrum (Fig. 2.4(b)). The tungstate ion (WOi -)
is a good example of case 3. The very broad emission spectrum (see Fig.
2.4(c)) does not show any vibrational structure at all, the Stokes shift is very
large (,...., 16 000 cm -1) and the zero-vibrational transition is not observable,
not even at the lowest possible temperatures, or for the highest possible
resolving powers.
Fig. 2.4 (a) An example of an emission spectrum in the weak-coupling case. The
spectrum consists of a line which corresponds to the zero-phonon transition. The
weak line at about 325 nm is a vibronic transition. This spectrum is the Gd 3 +
emission spectrum of LaB 3 0 6 :Gd 3 +. (b) An example of an emission spectrum in the
medium-coupling case. There is a progression of vibronic lines. The line on the
left-hand side is the zero-phonon line. This spectrum is the uranyl (UO~ +) emission
spectrum. (c) An example of an emission spectrum (M) in the strong-coupling case.
The spectrum consists of a broad, structureless band which shows a large Stokes shift
(SS) relative to the absorption (excitation) band (X). The spectra relate to the
tungstate luminescence of CaW0 4 .
Finally we draw attention to the fact that the single configurational

coordinate diagram is only an approximation. In practice there is more than
one vibrational mode involved and the system is not harmonic. Therefore
the value of S is not so easy to determine as suggested above. However, for
a general understanding the simple model is extremely useful, as we will see
below.
If we measure an absorption or emission spectrum, the following proper-
ties of the bands or lines are of importance:
1. their spectral position, that is, the energy at which the transition
occurs;
2. their shape, that is, sharp line, structured narrow band, or structureless
broad band;
3. their intensity.
For the spectral position the reader will be referred to the literature,
except for details of importance. The shape of the bands was discussed
above (see Figs 2.3 and 2.4). The intensity is contained in the electronic
matrix element
<v(Q)lrlu(Q) (2.6)
in equation (2.3). The intensity can be very low if selection rules apply. Here
we mention a few, well-known examples.
For electric dipole transitions the parities of the initial and final states
should be different (parity selection rule). This implies that transitions within
one and the same shell, for example 3d or 4f, are forbidden. This selection
rule may be relaxed by the admixture of opposite-parity states due to the
crystal field, or by vibrations of suitable symmetry.
Optical transitions are forbidden between states of different spin multi-
plicity (spin selection rule). This selection rule may be relaxed by spin-orbit
coupling. Since the latter increases strongly with the atomic number, the
value of this selection rule decreases if we proceed from top to bottom
through the periodic table. Many other selection rules of a more specialized
nature are known.
If we consider dopant ions in a solid, their spectral features will show
inhomogeneous broadening, even if their mutual interaction is neglected.
The reason for this is that the crystal field at the dopant ion varies slightly
from ion to ion owing to the presence of defects, such as impurities,
vacancies, dislocations, or the surface [9]. As a matter of fact the in-
homogeneous broadening will be more pronounced for line spectra than for
broadband spectra. Its magnitude is also much larger in disordered solids
(glasses) than in ordered solids [10].
An exceptional application of the configurational coordinate diagram has
been given by Giidel [11] in order to explain the nature of light-induced
metastable states in nitroprussides. Irradiation with visible radiation can
convert up to about 50% of sodium nitroprusside (Na2[Fe(CN)sNO]'
2H 20) into a metastable excited state with a very long lifetime at low
temperatures (> 107 s at 0 K). This is of interest for optical information
storage. An explanation of this phenomenon has been proposed in terms of
a configurational coordinate diagram where the excited (metastable) state
lies far outside the ground state parabola (Fig. 2.5). Return from the lowest
vibrational level of the excited parabola to the ground state is now very
difficult, since the vibrational overlap (unlvo> is very small. This return can
be achieved by thermal activation (> 165 K) or by irradiation with red light.
The explanation requires a large offset between the two parabolae. However,
since the proposed electronic transition is a two-electron transition, this may
well be possible.
Fig. 2.5 A configurational coordinate diagram in which the excited state v has its
equilibrium position outside the u ground state parabola. Return from the v
minimum to the ground state u is difficult and therefore slow. See also text.
After this extreme example we now tum to a very difficult problem in

luminescence, i.e. the nonradiative transitions [12, 13].
Up to this point it was assumed that the return from the excited state to
the ground state is radiative. In other words, the quantum efficiency (q),
which gives the ratio of the numbers of emitted and absorbed quanta, was
assumed to be 100%. This is usually not the case. Actually there are many
centres which do not luminesce at all. We will try to describe here the
present situation of our knowledge of nonradiative transitions that is
satisfactory only for the weak-coupling case.
Let us consider the configurational coordinate diagrams of Fig. 2.6 in
order to understand in a qualitative and rough way the relevant physical
processes. Figure 2.6(a) presents essentially the same information as Fig. 2.3.
Absorption and emission transitions are quite possible and are Stokes-
shifted relative to each other. The relaxed-excited state may, however, reach
the crossing of the two parabolae if the temperature is high enough. Via the
E E
Q
(a) (b)
E
Q
(c)
Fig. 2.6 Nonradiative transitions in the configurational coordinate diagram: (a)
strong coupling; (b) weak coupling; (c) combination of both.
crossing, it is possible to return to the ground state in a nonradiative
manner. The excitation energy is then completely given up as heat to the
lattice. This model accounts for the thermal quenching of luminescence.
In Fig. 2.6(b) the parabolae are parallel (S=O) and will never cross. It is
impossible to reach the ground state in the way described for Fig. 2.6(a).
However, nonradiative return to the ground state is possible if certain
conditions are fulfilled; that is, the energy difference ~E should be equal to
or less than 4-5 times the highest vibrational frequency of the surroundings.
In that case this amount of energy can excite simultaneously a few high
energy vibrations and is then lost for the radiative process. Usually this
nonradiative process is called multi phonon emission.
In Fig. 2.6(c) both processes are possible in a three-parabola diagram. The

parallel parabolae will belong to the same configuration, so that they are
connected by forbidden optical transitions only. The third one originates
from a different configuration and is probably connected to the ground state
by an allowed transition. This situation occurs often. Excitation (absorption)
occurs now from the ground state to the highest parabola in the allowed
transition. From here the system relaxes to the relaxed-excited state of the
second parabola. Figure 2.6(c) shows that the nonradiative transition
between the two upper parabolae is easy. Emission occurs now from the
second parabola (line emission). This situation is found for AI 20 3 :Cr3+
(4Ar-+4T2 excitation, 4Tr-+2E relaxation, 2E--+4A2 emission), Eu3+
CF --+charge transfer state excitation, charge transfer state to 5D relaxation,
5D--+ 7F emission), and Tb3+ CF --+4f?Sd excitation, 4f?Sd--+ 5D relaxation,
5D--+ 7F emission).
In general the temperature dependence of the nonradiative processes
is reasonably well understood. However, the magnitude of the non-
radiative rate is not, and cannot be calculated with any accuracy except
in the weak-coupling case. The reason for this is that the temperature
dependence stems from the phonon statistics which are known. How-
ever, the physical processes are not accurately known. In particular, the
deviation from parabolic behaviour in the configurational coordinate dia-
gram (anharmonicity) may influence the nonradiative rate by many orders
of magnitude.
Certain aspects of our qualitative considerations can be put on a more
quantitative basis as shown by Struck and Fonger [14, 15]. These authors
have given a unified model of the temperature quenching of narrow-line
and broadband emission using a quantum mechanical single configurational
coordinate model (see also Fig. 2.7). Calculations were made possible by
evaluating the vibrational overlap integrals exactly using the Manneback
recursion formulae. In addition to the parameters used above we introduce
auv , the parabolae offset, defined by a;v = 2(Su + Sv)' The relaxation energies
after emission and absorption are Suhwu and Svhwv respectively. Further, we
assume thermal equilibrium for the initial vibrational levels. Then the
radiative rate between Vm and Un is given by
Rnm=Ruv(1-rv)r::,<unlv m)2 (2.7)
and the nonradiative rate by
N nm = N uv(1-rv)r::,<u nlv m)2 (2.8)
Here Ruv and N uv are the electronic parts of the transition integral and
rv=exp( -hwv/kT) is the Boltzmann factor. The expression (l-rv)r::' gives
the thermal weight.
The electronic factors are considered to be constants. However, this
has been criticized, for example by Englman and Barnett [16]. Coupling
12
Q
oL---~~~~------~---
Fig. 2.7 Configurational coordinate diagram showing nonradiative transitions in

the Struck and Fonger approach. The excited state v shows an offset relative to the
excited state u and the ground state g. Full horizontal lines indicate the lowest
vibrational state in the parabolae, broken horizontal lines those which are at the
same energy in a lower parabola. The latter have their wavefunction maxima on the
parabolic curves. Therefore a nonradiative transition from u to g has a low
probability and that from v to u has a high probability, whereas that from v to g has
an intermediate probability.
between the electronic states U and v cannot be neglected. For Cr 3 + this has
been shown theoretically as well as experimentally [16, 17].
However, let us take the approximation that the electronic parts of
equations (2.7) and (2.8) can be considered to be constants. The total
radiative rate is obtained by summing over all initial and final states Un and
Vm' For nonradiative transitions the rate is obtained by summing over all
nearly resonant Vm, Un states.
Figure 2.8 shows the results of a model calculation for a red-emitting
phosphor [18]. Ruv and N uv are taken to be 104 s- 1 and 10 14 s- 1 respect-
ively. Further, hw=500cm- 1 in the U and v states, and auv =7.746 (large
offset). The distance between the parabolae minima is 25000 cm -1. Such
a phosphor has a thermal quenching temperature T1/2 of 450 K (at T1/2
the emission intensity at 0 K has dropped to 50%). The results show how
a variation of one of the parameters influences the thermal quenching,
i.e. the nonradiative rate. Shifting the excited parabola to lower energy
Fig. 2.8 Calculations with the Struck-Fonger approach on a model phosphor. The
quantum efficiency q is calculated as function of temperature. In each of the three
pictures only one parameter is varied: in the picture on top the energy of the zero-
phonon transition (E zp ), in the centre the vibrational frequency (hw.) and in the lower
one the parabolae offset a. v The latter two parameters in particular show a drastic
influence on the temperature dependence of q. (After Bleijenberg and Blasse [18].)
decreases T l/2 Increasing hw increases the nonradiative rate drastically;

even at 0 K there is a considerable influence of the nonradiative processes. A
larger offset implies also a faster nonradiative rate.
These considerations can be used to predict the occurrence of lumines-
cence with high thermal quenching temperature. A full account has been
given elsewhere [7,8,12]. As examples we mention here the use of very stiff
host lattices [7,8], the application of surroundings around the luminescent
centre which can hardly expand, and the use of well-fitting organic cages
around a luminescent ion [8,19].
Nonradiative transitions between parallel parabolae (Fig. 2.6(b, of special
importance for rare earth ions, are theoretically more easy to handle. The ex-
perimental and theoretical situation at the moment is quite satisfactory [12,20].
For transitions between 4f n levels, the temperature dependence of the
nonradiative rate is given by
N(T)=N(O)(l +nY (2.9)
where N(T) is the rate at temperature T, p=I1E/hw, I1E is the energy

difference between the levels involved, and
n= [exp(hw/kT)-lr 1 (2.10)
N(T) is large for low p, that is for smalll1E or high vibrational frequencies.
Further,
N=/3 exp[ -(I1E-2hw max )a] (2.11)
with a and /3 constants and W max the highest available vibrational frequency
of the surroundings of the rare earth ion. This is the energy gap law in the
revised form of van Dijk and Schuurmans [20] that makes it possible to
calculate N with an accuracy of one order of magnitude.
In solids recently a new mechanism for loss of excitation energy gained
interest, although in principle this was already known for molecular species
in solution. Let us start with the latter where this mechanism is known as
quenching by electron transfer [21]. Figure 2.9 shows schematically the
essentials of this process. Parabolae a, band c represent energy levels of a
species A, parabola d that of a state A + B -. This state is obtained by
electron transfer from species A to a nearby species B. The offset of parabola
d relative to the other parabolae is large, and its position is at not too high
an energy. It is clear that from parabola c we can reach the ground state
/
/ d
/
/
/
/
Fig. 2.9 Schematic representation of luminescence quenching by electron transfer.

The state d is a strongly shifted electron transfer state via which nonradiative return
to the ground state is possible.
parabola a via parabola d in a nonradiative way. This process is expected to
be of importance if the energy difference between the electron transfer state
A +B - is not too high above the ground state A + B. The requirement of a
large offset is usually satisfied for electron transfer states.
This quenching mechanism has been observed for ion pairs in solution
[22]. As an example we mention here [Eu3+ c 2.2.1], where c stands for
encapsulation, and M(CN)~ - (M2 + = Fe2+, Os2+ or Ru 2+) [23]. The
electron transfer state involved can be written schematically as Eu 2+ _M3 +.
Molecular complexes where this process is of importance are the lanthanide
decatungstates [RE3+ W 10036]9 - [8]. For RE = Eu efficient emission is
observed, while for RE = Tb only quenching occurs because of a low-lying
Tb 4 + -W5+ electron transfer state.
In nonmolecular solids the examples are abundant, but often not recog-
nized. We mention the absence of luminescence in YV0 4 :Tb due to a
Tb 4 + _V 4 + electron transfer state and the quenching of Ce3+ and Eu3+
luminescence in Ce3+ -Eu3+ pairs owing to an electron transfer state
Ce 4 + -Eu2+ [7, 8].
These examples all relate to ion pairs. However, it has recently been
shown that photoionization can also quench luminescence [24, 25]. This
process is very similar to quenching by electron transfer in an ion pair
[26]. As a well-studied example we take Y3AI 5 0 12 :Ce3+ [24,27]. Figure
2.10 shows the energy levels of the Ce3+ ion relative to the valence
and conduction band of the host lattice. The ground state of Ce 3+ is 4f 1,
and the excited state consists of the crystal-field levels of the 5d 1 configur-
ation. As long as we excite in the lowest crystal-field level, band emission
from Ce 3+, Stokes-shifted but with high quantum efficiency, is observed.
However, on excitation into higher levels the quantum efficiency of the
Ce3+ emission drops considerably, and simultaneously photoconductivity is
----5d
CB
----5d
----4f
VB
Fig. 2.10 Schematic representation of the Ce3+ energy levels in Y3Als012' VB and
CB indicate the top of the valence band and the bottom of the conduction band of
Y3Als012 respectively. 4f gives the ground state of the Ce 3+ ion. The lowest excited
5d s-tate of Ce3+ is situated in the forbidden zone, but the lowest-but-one state is in
the conduction band making photoionization of Ce 3+ possible when the ion is
promoted into that state.
observed. Because these higher levels are situated in the conduction band,
photoionization of the Ce 3 + ion can take place. In analogy to the electron
transfer in ion pairs described above, the electron-transfer state involved can
be indicated as Ce 4 + --(HL)-, where HL denotes the host lattice.
Pedrini et al. have shown that this is an important mechanism for other
ions too, for example Eu 2 +, Sm 2 + and Yb 2 + [24,25]. Not always does this
mechanism prevent the occurrence of emission. In several cases they observe
what is called impurity-bound exciton emission. A nice example is Yb 2 + in
the fluorides CaF 2, SrF 2 and BaF 2. The former two yield impurity-bound
exciton emission, but in BaF 2 there is no emission at all [25]. The emission
is ascribed to exciton recombination: the hole resides on the (photoionized)
Yb 2 + ion, the electron on the surroundings. In Fig. 2.9 this emission
transition would be described by a transition from parabola d to the ground
state.
The processes described in the final part of this section may be of
importance for every optically active centre which can easily change its
valency, as will be clear from the examples presented.
Let us now turn to phenomena which are due to interaction between
centres.
2.3 ENERGY TRANSFER

If luminescent centres come closer together, they may show interaction with
each other that results in new phenomena. Consider two centres, S and A,
with a certain interaction. The relaxed excited state of S may transfer its
energy to A. This energy transfer has been treated by Forster and Dexter
and is now well understood [5-7].
Dexter, following the classic work by Forster, considered energy transfer
between a donor (or a sensitizer) S and an acceptor (or activator) A in a
solid. This process occurs if the energy differences between the ground and
excited states of S and A are equal (resonance condition) and if a suitable
interaction between both systems exists. The interaction may be either an
exchange interaction (if there is wavefunction overlap) or an electric or
magnetic multipolar interaction. In practice the resonance condition can be
tested by considering the spectral overlap of the S emission and the A
absorption spectra.
Figure 2.11 shows the energy level scheme and parameters involved. The
emission transition S*--+S and the absorption transition A--+A* have
normalized line shape functions gs(E) and gA(E). The initial state is IS*, A),
and the final state (after energy transfer) 1S, A*). The transfer is brought
about by an interaction H SA ' The resulting transfer probability is
(2.12)
Energy transfer 37
o R
- --0
---r--S* ----,r--A*
H
E----------- ~
-~--S A
------+ E
Fig. 2.11 Energy transfer from S to A. R is the SA separation, H is the SA
interaction. The hatched area in the lower picture presents the spectral overlap. See
also text.
Here the integral represents the spectral overlap. The distance dependence of
PSA depends on the interaction mechanism.
If the interaction is of the exchange type, PSA decreases exponentially with
the S-A distance (RSA)' because the wavefunctions do so too. If the
interaction is of the multipolar type, PSA decreases as Rs; where n depends
on the type of interaction. For electric dipole-dipole interaction, for
example, n = 6 (see Chapter 1).
It should be realized that a high value of PSA does not imply automati-
cally that transfer will occur. The excited state of S, i.e. S*, has other ways to
decay, i.e. radiatively (P r ) and/or nonradiatively (Por)' Transfer occurs if
PSA>Pr+Por' The distance for which the transfer rate equals the internal
decay rate is called the critical distance Re. For exchange-mediated transfer
Re is not much larger than 7-8 A, determined by wavefunction overlap. For
transfer by multipolar interactions Re can be much larger, i.e. 50--100 A if
favourable spectral overlap of allowed S*-+S and A-+A* transitions occurs.
An example of energy transfer can be found in the classic lamp phosphor
Ca5(p04h(F, CI):Sb3+, Mn2+ [28]. Short wavelength ultraviolet excitation
excites only the Sb 3 + ion which yields blue emission. Some of these ions,
however, transfer their excitation energy to Mn 2 + which yields yellow
emission. In this way white emission occurs. The emission colour can be
varied by varying the Sb3+:Mn 2 + ratio. The Sb3+ -Mn2+ transfer occurs by
exchange; its critical radius is 12 A.
Not always is all of the excitation energy transferred. If only part of it is

transferred, this is called cross-relaxation. Let us consider an example. The
higher energy level emissions of Tb3+ and Eu3+ (Fig. 2.12) can be quenched
if the concentration is high. The following cross-relaxations may occur:
Tb3+eD 3 )+ Tb3+CF 6 )--+Tb 3 +eD 4 )+ Tb3+CF o) (2.13)
Eu3+eD 1 )+ Eu3+CF o)--+Eu 3 +eDoH Eu3+CF 3) (2.14)
The higher energy level emission is quenched in favour of the lower energy
level emission.
If we consider now transfer between two identical ions, for example
between 8 and 8, the same considerations can be used. If transfer between
two 8 ions occurs with a high rate, what will happen in a lattice of 8 ions, for
example in a compound of 8? There is no reason why the transfer should be
restricted to one step, so that we expect that the first transfer step is followed
by many others. This can bring the excitation energy far from the site where
the absorption took place: energy migration. If in this way the excitation
energy reaches a site where it is lost nonradiatively (a killer or quenching
20 _ _ _ _ 50
,
----50
o
10
Fig. 2.12 Cross-relaxation for two Eu 3 + ions (left-hand side) and for two Tb 3 +
ions (right-hand side). The arrow indicates the amount of energy which is trans-
ferred from one ion to the other. As a consequence the higher-level emission is
quenched.
Energy transfer 39
site), the luminescence efficiency of that composition will be low. This
phenomenon is called concentration quenching. This type of quenching will
not occur at low concentrations, because then the average distance between
the S ions is so large that the migration is hampered and the killers are not
reached.
Energy transfer is often studied by measuring the time dependence of the
S emission intensity after a short excitation pulse. We will discuss now a
couple of expressions often used in the literature. Simultaneously they
illustrate the several regimes of energy migration which have been observed.
Here energy migration is defined as a large number of subsequent energy
transfer steps between identical centres.
Consider first a crystal with donors and acceptors and assume that only
one-step energy transfer from S to A is possible. The S and A ions will be
distributed at random. An excited donor can interact with all unexcited
acceptors and it is necessary to account for the distribution in SA separa-
tions. This problem has been treated by Inokuti and Hirayama [29].
They obtained the following expression for the decay of S in the presence
of A:
(2.15)
Here 1:0 is the decay constant of S in the absence of A, CA is the

concentration of A, Co is the critical activator concentration and n = 6, 8 or
10 depending on the electric multipole interaction. For exchange interaction
their result reads
(2.16)
where y = 2RjL with Rc the critical distance and L an effective Bohr radius.
Note that I(t) is not an exponential in the presence of A. In this treatment
only SA transfer is considered and SS transfer is assumed not to occur. In
the absence of A, the S species decays exponentially according to
I(t)=I(O) ex p ( - :J (2.17)
Until now we have considered only the microscopic, single-step donor-

acceptor energy transfer. It is possible, however, that donor-donor transfer
also plays a role. Excitation energy may migrate among the donor species
before being transferred to an acceptor. A good survey of earlier and recent
work on this problem has been given by Huber [30]. From this survey we
derive the following for our purpose.
Consider the time evolution of Pit), the probability that species n is

excited and all other atoms are in their ground state:
dPn(t)/dt= -(YR+Xn+ n~n Wnn)Pn(t) + n~n Wn'nPn,(t) (2.18)
The first term on the right-hand side corresponds to processes which bring
the species n back to the ground state: YR is the radiative probability, Xn is
the transfer rate to acceptors and Ln' Wnn , gives the transfer rate from species
n to other donor species n'. The second term describes the reverse process.
For simplicity back-transfer from the acceptors is neglected.
The energy difference between ground and excited state, En" will vary
from donor to donor owing to perturbations from impurities, strains,
etc. This yields the inhomogeneous line broadening observed under broad-
band excitation. There are two techniques to follow the excitation energy
migration in the donor system, i.e. fluorescence line narrowing (FLN)
and the time evolution of the donor luminescence in the presence of
acceptors.
In FLN a pulsed, narrow band light source (a laser) excites those donors
whose resonance frequencies span a small part of the inhomogeneous line.
After the pulse, the luminescence evolves as shown schematically in Fig.
2.13. Broadband luminescence arises due to energy transfer to donors which
were not excited directly. The decay of the narrow component yields
information on the microscopic transfer process.
A well-known example is the case of Lao.sPro.2F 3 [31]. In Fig. 2.14 we
give the time evolution of the emission of the 3P o--+ep 6)1 transition on the
Pr 3 + ion. Excitation is at 12 cm -1 higher energy than the line centre. Note
Fig. 2.13 Schematic representation of the time development of the luminescence in

FLN: (a) t=O; (b) t>O. The broad area in (b) corresponds to ions which were not
initially excited. Excitation occurred only for the ions emitting in the narrow line.
Energy transfer 41
that the line decreases in time, whereas the background luminescence
increases. This shows that energy transfer occurs within the Pr 3 + subsystem
and that the temperature is high enough (14 K) to make the transfer process
independent of the energy mismatch. From these experiments we can find
the ratio R(t):
R(t)= narrow b~nd in~ensity.at time t

total mtensIty at hme t
Theoretical expressions for R(t) have been derived in the literature. In this
way it becomes possible to derive transfer characteristics from a comparison
between experiment and theory. For the case of Fig. 2.14, for example, it has
been found that electric dipole-dipole transfer is dominant in the system
Lao.sPro.2F 3 and that the nearest-neighbour transfer rate is 0.4 x 106 S-l
(14 K).
(a)--=-----=------
(b) _-"'---_ _ _ _ __
(c) _-=-_ _ _ _ _ _....::....
+20 +10 o -10
6v(cm-')
Fig. 2.14 Time-resolved emission spectra for the 3PO-.eH6h luminescence in
La o.S Pr O.2 F 3. Excitation is 12 cm -1 above the line centre. T= 14 K. (a) Immediately
after the excitation pulse, (b) 0.811S after the pulse, and (c) 311S after the pulse.
(Constructed after data by Huber et al. [31].)
The time evolution of the donor fluorescence on broadband excitation

is an old problem in luminescence. By measuring the time dependence of
the donor fluorescence it is possible to obtain information about the
donor-donor and donor-acceptor transfers by analysing the decay curve.
Since the integrated intensity at a time t is proportional to the number of
excited donors at that time, No(t), the decay can be described by
(2.19)
Here N 0(0) is the number of excited donors at the time the pulse is turned
off and f(t) is the fraction of excited donors if the radiative lifetime (YR" 1)
were infinite. The function f(t) depends on time as described above for Pn(t),
if YR =0.
Exact solution is possible for two extreme cases, i.e. no donor-donor
transfer at all (see above) and very rapid donor-donor transfer. The
behaviour of f(t) between these two cases is extremely complicated.
In the limit of no donor-donor transfer at all we obtain
(2.20)
This is a generalization of the results obtained by Inokuti and Hirayama

[29]. CA gives the acceptor concentration and X 01 the transfer rate from
a donor at site 0 to an acceptor at site l. The value of 1- CA gives the
probability to find no acceptor on site l. If site 1 is occupied by an ac-
ceptor, it contributes a factor exp( -XOlt) to exp( -Xnt). Here Xn is the
total donor-acceptor transfer rate for the nth donor. Our equation
represents therefore an average of exp( - X nt) over all configurations of
acceptors.
In the case of rapid transfer (or fast diffusion) the donor-donor transfer
takes place so rapidly that for t>O all donors have equal probability to be
excited. f(t) has now a very simple form, i.e.
f(t)=ex p ( -CA~ XOlt) (2.21)
In Fig. 2.15 we present some schematic plots of f(t) versus t. In general

f(t) is initially nonexponential, but becomes exponential after a certain time.
In the rapid transfer case (curve b in Fig. 2.15) No(t) is exponential in the
whole time region. In the absence of donor-donor transfer N o(t) becomes
exponential after long times only, with a slope equal to the radiative decay
time. Also in the intermediate case N o(t) is initially nonexponential, but
becomes exponential in the limit t-HX!. The slope, however, is steeper than
in the absence of donor-donor transfer (curve c in Fig. 2.15). In order to
describe this case several theories have been presented in the literature, for
example a hopping model [32] and a diffusion model [33]. The latter
solution in particular has become popular if the diffusion is not fast enough
Energy transfer 43
fIt)
0.1
0.01
0.001 "-_"'--"'---'--_....L-_........_---'-_ _
o 4 8 t
Fig. 2.15 Schematic plot of f(t) versus t: curve a, donor-donor transfer absent
(equations (2.15) and (2.16; curve b, rapid transfer (equation (2.21; curve c,
intermediate between curves a and b (e.g. equation (2.22.
to maintain the initial distribution of excitation (diffusion-limited transfer).

The following expression was found:
ND(t) = ND(O) exp( -yRt)
exp [ -3 n
~ 3/2
CA
(C )1/2
t
(1 + 10.87x+ 15.50X 2)3/4]
1 +8.743x (2.22)
Here C is the interaction parameter for donor-acceptor transfer, and

x = DC - 1/3 t 2/3 where D is the diffusion constant. For t -+ 00 an exponential
time dependence is predicted with decay rate L01 = 11.404CAC1/4D3/4. Here
the diffusion is assumed to be isotropic. For one- and two-dimensional
diffusion, however, f(t) has the asymptotic limits [4n(C A/a)2Dtr 1/2 and
(4nC Aa- 2Dt)-1 respectively, where a is the lattice constant [30]. A more
fundamental theory of energy transfer has been given by Huber [30].
Note finally that back-transfer from acceptor to donor has been neglected
until now. It is possible, however, to incorporate back-transfer in the
existing theories [30].
Let us close this section with an important example of energy migration

and its application. The luminescent material GdMgB 5 0 1o :Ce,Tb [1] is the
green-emitting phosphor in energy-saving luminescent lamps. Excitation is
into the Ce 3 + ion at 254 nm. The excited Ce 3 + ion transfers its energy to
one of the neighbouring Gd 3 + ions. The transfer rate is much higher than
the Ce3+ radiative rate, so that Ce 3 + emission is practically absent.
Subsequently the energy migrates over the Gd3+ sublattice (fast diffusion).
Every transfer step has a rate of 106 -10 7 S-1, whereas the radiative rate is
only some 102 s -1. Consequently the excitation energy would, in an un-
doped compound, make some 104-10 5 steps before ending its life radiatively.
In the phosphor the migrating excitation energy is trapped by the Tb 3 +
ions, from where emission occurs. The quantum efficiency of the overall
process is close to 1. This makes gadolinium compounds one of the most
promising hosts for photoluminescent materials.
2.4 SOME CASE STUDIES
2.4.1 The Cr3 + (3,P) ion

As mentioned above the luminescence of Cr 3 + has been studied extensively
because of its intriguing luminescence properties. Figure 2.16 gives its energy
level diagram as far as necessary to understand the luminescence. Depending
on the strength of the crystal field ..1, the emission can be the spin-forbidden
2E-+ 4A2 transition (strong field), or the spin-allowed 4T 2-+ 4A2 transition
(weak field). The former occurs within the octahedral t~ subconfiguration,
Fig. 2.16 Schematic representation of the lower energy levels of Cr3+ (d 3 ) as a

function of the crystal field ..1. To the left of the broken line the emission is from 4T 2
to 4Az (broad band, infrared, spin allowed); to the right of the broken line the
emission is from 2E to 4A2 (narrow line, red, spin forbidden).
Some case studies 45
i.e. between parabolae without offset, and consists mainly of a sharp line.
The latter consists of a broad band because it occurs between the two
subconfigurations t~e and t~. This illustrates nicely how the parabolae offset
determines the band shape of the optical transition. For the isoelectronic ion
Mn4+ (in Cs 2GeF 6) the relevant parameters have been carefully deter-
mined: for 4A2-+2E the expansion is only 0.003 A, for 4A2-+4T 2 it is 0.053 A
(8=3) [34].
Ruby (Ah03:Cr3+) is the most famous case of 2E line emission. The
quenching temperature of this emission is very high (800 K); quenching
occurs via thermal population of the higher 4T 2 level.
Broadband 4T 2 emission has been reported for SC203: Cr3+, ScB0 3:Cr3+
and Cr 3+-doped glasses. In these lattices the crystal field is much smaller. In
the crystalline host the quantum efficiency of this emission may be high at
room temperature; however, in the glass modifications it is always low with
values of about 20% as a maximum. This phenomenon has prompted much
research. The explanation is in principle simple and derives directly from a
consideration given above, i.e. the value of AQ, the parabolae offset, needs to
be small for high quantum efficiency. Glasses with their loose structure do
not favour small AQ values [7,8]. This can be seen directly from the Stokes
shift of the 4T 2 emission of Cr3+: in oxide crystals it varies from 2400 to
3500 cm -1, in oxide glasses from 4500 to 5300 em - 1 [35, 36].
In conclusion, the luminescence of Cr 3+ illustrates our theoretical con-
siderations of the influence of parabolae offset on emission band shape and
the nonradiative rate.
2.4.2 The rare earth ions Eu3+ (4f 6 ), Gd3+ (4f7) and Tb3+ (4f 8 )
Absorption spectra of the rare earth ions consist of sharp and weak lines
and, at higher energy, of broad band(s). Figure 2.17 gives an example. The
sharp lines are parity-forbidden transitions within the 4f n configuration, and
the broad bands are allowed transitions from the 4f n to excited configur-
ations. These are either 4fn -15d or 4f n + 1L -1 (charge transfer, L = ligand).
This situation bears an analogy with the Cr 3+ ion, and shows again the
influence of parabolae offset on spectral band shape, since AQ ~O for
intraconfigurational and AQ =/: 0 for interconfigurational transitions.
The emission of the three ions under discussion consists usually of
sharp lines due to transitions from the excited level to the ground levels:
5Do- 7F J for Eu3+ in the red, 6P7/2-+8S for Gd3+ in the ultraviolet, and
5D4 -+ 7FJ for Tb 3 + in the green. However, under special conditions other
emissions appear, for example 5D 2- 7 F J for Eu3+ in the blue, 5D3-7FJ for
Tb3+ in the blue, and for Gd3+ even vacuum ultraviolet (187 nm). Higher
level emission is only possible if (i) the vibrational frequencies in the lattice
are not too high, and (ii) cross-relaxation is absent owing to a low
concentration.
20 30 40
10 cm- 1
3
Fig. 2.17 Schematic absorption spectrum of the Eu 3 + (4f 6 ) ion in an oxide. The
broad band on the high-energy side is due to the charge-transfer transition. The lines
are due to transitions within the 4f 6 configuration.
The latter point is immediately clear from our discussion above. The
former will become clear from equation (2.11). The higher CO max is, the larger
is N. Indeed, the higher level emissions mentioned here do not occur at all in
hosts such as borates and phosphates.
If the vibrational frequencies become exceptionally high, as in aqueous
solutions (co max ~ 3500 cm -1), the quantum efficiency of even the lowest level
emission decreases. Only in Gd3+ (AE ~ 32 000 cm -1) does the radiative
rate still exceed the nonradiative rate by many orders of magnitude, but in
Tb3+ (AE ~ 15 000 cm - 1) and Eu 3+ (AE ~ 12 000 cm - 1) a drastic decrease
of the quantum efficiency occurs, whereas the other rare earth ions practi-
cally do not emit in aqueous solution. If, however, heavy water is used
(D 20, Vmax ~ 2200 cm -1) the efficiencies go up again. For solids these effects
can be studied in NaLa(S04h H 20 where the rare earth site is coordinated
to one H 20 molecule only. The q values are 100% for Gd3+, 70% for Tb3+,
10% for Eu3+ and 1% for Sm3+ and Dy3+ [7,8].
In many compounds of Eu3+, Gd3+ and Tb3+ energy migration has been
observed [6-8]. An example was given above for Gd3+. The Gd3+ ion
transfers efficiently to Eu3+ and Tb3+, and Tb 3+ to Eu3+ which has the
lowest emitting level of the three. The latter transfer explains why some
Tb 3+ compounds emit in the red: the Tb 3+ ions are excited, and the
excitation energy migrates over the Tb 3+ sublattice until it is trapped by a
Eu 3+ impurity from where emission occurs.
Finally we want to discuss what happens on excitation into the broad
absorption band. As an example we consider the charge transfer CT band of
Eu3+ [37]. In the commercially applied lamp phosphor Y203:Eu3+ excita-
tion into the CT state results in relaxation to the minimum of the CT
parabola, followed by a transition to the higher parabolae of the 4f 6
configuration. Finally emission from these parabolae occurs with high

quantum efficiency (see Fig. 2.6(c)).
However, this is not always the case. For Eu3+ in calcium compounds the
resulting q is always relatively low, and for Eu3+ in glasses the same has
been observed. This has been explained by a somewhat larger expansion of
the CT state, i.e. a larger parabolae offset. If this is the case, the CT state
feeds directly the ground state (Fig. 2.18). In section 2.2 it was shown that a
large offset implies a high nonradiative rate. This is observed here for Eu3+
in glasses, which is directly related to similar observations for Cr 3 + in
glasses. The case of Eu3+ in calcium compounds is less obvious. However,
the excess charge on Eu3+ relative to Ca2+ results in an extra contraction of
the ground state, which does not occur in the CT state. In this way AQ(CT)
is relatively large. Although these nonradiative processes are of a compli-
cated nature, the approach outlined here informs us about the essentials of
these processes.
CT CT
\ \
\ \ E
\ \
\ E
\.
"- /
5D 5D
(a) (b)
Fig. 2.18 Configuration coordinate diagram for the Eu 3 + ion in (a) an yttrium host
lattice and (b) a calcium host lattice. In the latter the charge transfer state (CT) feeds
the 7F ground levels directly which quenches the luminescene due to emission
transitions between the 5D and 7F levels.
2.4.3 The tungstate group

The tungstate group has a long history in the field of luminescence [38, 39].
A well-known luminescent compound is CaW0 4 . The luminescent centre
is the WOi- group where the central metal ion is W 6 + (5dO). The optical
transition involved in excitation and emission is a tungsten-oxygen charge
transfer transition, which brings about a large reorganization of the
chemical bonding resulting in AQ ~ 0.20 A. Because of this large AQ
value the absorption and emission spectra are very broad (half-width
'" 1 eV) and structureless (see Fig. 2.4{c)), and the Stokes shift is very large
(1.5-2.0 eV).
It is interesting to note that in CaW0 4 there is no concentration
quenching although the tungstate groups are neighbours in the lattice.
However, the large reorganization of the excited state, as observed from the
large Stokes shift, brings the excited state out of resonance with the ground
state, i.e. the spectral overlap between emission and absorption vanishes, so
that tungstate-tungstate transfer becomes impossible: although there is
probably enough interaction, there is no resonance (compare equation
(2.12)).
This is not necessarily always the case. In YV0 4 :Eu3+ and Ba2MgW06:
U 6 + the dO complex ions (VOl- and WO~- respectively) have a relatively
small Stokes shift ('" 1 eV) and at room temperature (but not at low
temperatures) the thermal activation is high enough to allow energy migra-
tion over the sublattice of the complex ions. Finally this energy is trapped by
the activator (Eu3+ and U 6 + respectively) from where emission occurs. The
composition YV0 4 :Eu 3 + has been for some time a cathode ray and a lamp
phosphor [40]. At room temperature the undoped host lattices YV0 4 and
Ba2MgW06 are for the greater part concentration quenched: the migrating
excitation energy is trapped by quenching centres.
Let us return to CaW0 4 Why are isomorphous SrW0 4 and BaW0 4
considerably less luminescent than Ca W0 4 ? The answer to this problem is
in principle easily found. The ionic radius of the alkaline earth ion increases
from calcium to barium. Therefore expansion of the excited state is easier in
the softer BaW0 4 than in CaW0 4 . Again small changes in AQ have drastic
influences on the luminescence efficiency (i.e. on the rates of the nonradiative
processes).
It would be incorrect to think that the exact nature of the optical
transitions in such classic materials is known. The reason for this is that the
excited state, with an electron in the d orbital of the central ion, and a hole
spread over the ligands, gives rise to a large number of excited levels. Two
facts seems to be clear nowadays. First, the emission originates from a triplet
state [39], so that it is spin forbidden and slow (the excited state of the
vanadate group, for example, lives about 1 ms). Second, the excited state is
strongly Jahn-Teller distorted as shown recently by van der Waals' group
(see for example ref. 41). This implies that the simple configurational
coordinate used so far is certainly incorrect. Other coordinates have to be
taken into account too.
2.4.4 The Bj3+ (6s2) ion

The ground state of ions with S2 configuration consists of the lSO level. The
states arising from the excited sp configuration are 3p 0, 3p 1, 3p 2 and 1 P b in
sequence of increasing energy. At low temperatures the emission is usually
due to the 3PO-+1So transition which is highly forbidden; at higher tempera-
tures the less forbidden 3p 1-+ 1So transition appears [42].
One of the surprising issues of BP + luminescence is the strong depend-
ence of the spectral characteristics on the nature of the host lattice. Let us
concentrate on three cases, i.e. Cs 2NaYCI 6:Bi3+, Bi4Ge3012 and
Cs3Bi2Br9, which each show a strikingly different luminescence [43]. This is
immediately clear from the Stokes shift, which amounts to 1000 cm -1,
18000 cm -1, and ",0 cm -1 respectively.
Consider first Cs 2NaYCI 6:BP+. The emission consists of a narrow band
with a strong vibrational structure. Obviously AQ is small, i.e. the coupling
between the electronic Bi 3+ states and the lattice is small. As a consequence
Cs 2NaBiCl 6 does not luminesce. The system Cs 2Na Y1-xBixCl6 shows
pronounced concentration quenching which starts already at about x = 0.01.
In Bi4Ge3012 the situation is completely different. Spectrally the situation
reminds one very much of the tungstate luminescence. The large Stokes shift
points to a large value of AQ which is related to the special type of bonding
of S2 ions (pseudo Jahn-Teller effect). The large Stokes shift prevents energy
transfer among the Bi 3+ ions, and therefore also concentration quenching of
the luminescence.
Now we turn to Cs3Bi2Br9' Its behaviour is completely different, and is
formally outside the scope of this chapter. However, it is interesting to see
where the boundaries of our models lie. The compound Cs3Bi2Br9 is a
semiconductor owing to wavefunction overlap of the Bi3+ orbitals. Energy
band broadening leads to a low absorption edge: the compound is yellow,
whereas Bi4Ge3012 absorbs only in the short-wavelength ultraviolet. Ex-
citation of Cs3Bi2Br9 leads to the creation of electrons and holes, and
pronounced photoconductivity has been observed. This compound is one of
the rare examples in which free exciton recombination has been observed: it
is a sharp line at the position of the band gap (for this reason the Stokes shift
was formally given as 0 above). In addition there is also bound-exciton
recombination and deep-centre recombination due to the presence of
defects. Energy migration occurs here by free charge carriers, a process very
different from that described in section 2.3.
Schematically the three possibilities are indicated in Fig. 2.19. The top of
the triangle presents the free Bi 3+ ion, and Cs 2Na YCI 6:Bi3+ is not far from
that point. Along the right-hand leg localization in the solid increases:
electron-lattice interaction and relaxation in the excited state become more
and more important. Therefore Bi4Ge3012 is at the end of this leg. Along
the left-hand leg, delocalization by wavefunction overlap increases. There-
fore semiconductors such as Cs3Bi2Br9 and Bi 20 3 appear at the end of this
leg. Along the base of the triangle we move from delocalized (left) to
localized (right). The Bi 3+ ion is a very peculiar ion because it is one of the
ions whose optical properties span the whole triangle. This chapter dis-
cussed only phenomena occurring along the right-hand leg. The triangle
FREE ION
Cs 3 Bi2Brg Bi 4 Ge 3 0'2
Fig. 2.19 Schematic representation of luminescent centres with different character-
istics: Elf' electron-lattice relaxation; J, band broadening. See also text.
shows that there are still more, but these phenomena are discussed in other
chapters in this book.
REFERENCES
1. Blasse, G. (1989) Chern. Mater., 1, 294.
2. Becquerel, E. (1867) La Lurniniere - ses Causes et ses Effets, Didot, Paris.
3. Imbusch, G.F. (1988) In Lasers, Spectroscopy and New Ideas (eds W.M. Yen and
M.D. Levenson, Springer, Berlin, p. 248.
4. DiBartolo, B. (ed.) (1987) Spectroscopy of Solid-state Laser-type Materials,
Plenum, New York.
5. Henderson, B. and Imbusch, G.F. (1989) Optical Spectroscopy of Inorganic
Solids, Clarendon, Oxford.
6. DiBartolo, B. (ed.) (1984) Energy Transfer Processes in Condensed Matter,
Plenum, New York.
7. Blasse, G. (1988) Prog. Solid State Chern., 18, 79.
8. Blasse, G. (1990) Adv. Inorg. Chern., 35, 319.
9. Holstein, T., Lyo, S.K. and Orbach, R. (1981) In Laser Spectroscopy of Solids (eds
W.M. Yen and P.M. Selzer), Springer, Berlin, p. 39.
to. Weber, M.J. (1981) Laser Spectroscopy of Solids (eds W.M. Yen and P.M. Salzer),
Springer, Berlin, p. 189.
11. Glidel, H.U. (1990) Chern. Phys. Lett., 175,262.
12. DiBartolo, B. (ed.) (1980) Radiationless Processes, Plenum, New York.
13. DiBartolo, B. (ed.) (1991) Advances in Nonradiative Processes in Solids, Plenum,
New York.
14. Struck, C.W. and Fonger, W.H. (1975) J. Lurnin., 10, 1.
15. Struck, C.W. (1991) In Advances in Nonradiative Processes in Solids (ed. B.
DiBartolo), Plenum, New York.
16. Englman, R. and Barnett, B. (1970) J. Lurnin., 3, 37, 55.
17. Donnelly, CJ., Healy S.M., Glynn, TJ., Imbusch, G.F. and Morgan, G.P. (1988)
J. Lurnin., 42, 119.
18. Bleijenberg, K.c. and Blasse, G. (1979) J. Solid State Chern., 28, 303.
19. Blasse, G., Dirksen, G.J., Sabbatini, N. and Perathoner, S. (1987) Inorg. Chirn.
Acta, 133, 167.
References 51
20. van Dijk, 1M.F. and Schuurmans, M.F.H. (1983) J. Chem. Phys., 78,5317.
21. Balzani, v., Bolletta, F., Gandolfi, M.T. and Maestri, M. (1978) Top. Curro Chem.,
75, 1.
22. Balzani, V., Sabbatini, N. and Scandola, F. (1986) Chem. Rev., 86, 319.
23. Sabbatini, N. and Balzani, V. (1985) J. Less-Common Met., 112, 381.
24. Pedrini, e., Rogemond, F. and McClure, D.S. (1986) J. Appl. Phys., 59, 1196.
25. Moine, B., Courtois, B. and Pedrini, e. (1989) J. Phys. (Paris), 50, 2105.
26. Biasse, G., Schipper, W.J. and Hamelink, 1.1. (1992) Inorg. Chim. Acta., 159, 77.
27. Hamilton, D.S., Gayen, S.K., Pogatshnik, G.J. and Ghen, R.D. (1989) Phys. Rev.
B,39,8807.
28. Soules, T.F., Bateman, R.L., Hewes, R.A. and Kreidler, E.R. (1973) Phys. Rev. B,
7, 1657.
29. Inokuti, M. and Hirayama, F. (1965) J. Chem. Phys., 43, 1978.
30. Huber, D.L. (1981) In Laser Spectroscopy of Solids (eds W.M. Yen and P.M.
Selzer), Springer, Berlin, p. 83.
31. Huber, D.L., Hamilton, D.S. and Barnett, B. (1977) Phys. Rev. B, 16, 4642.
32. Burshtein, A.I. (1972) Sov. Phys. J ETP, 35, 882.
33. Yokota, M. and Tanimoto, I. (1967) J. Phys. Soc. Jpn., 22, 779.
34. Chien, R.L., Berg, 1M., McClure, D.S., Rabinowitz, P. and Perry, B.N. (1986) J.
Chem. Phys., 84,4168.
35. Imbusch, G.F., Glynn, T.J. and Morgan G.P. (1990) J. Lumin., 45, 63.
36. Biasse, G. (1990) Mater. Chem. Phys., 25, 393.
38. Kroger, F.A. (1948) Some Aspects of the Luminescence of Solids, Elsevier,
Amsterdam.
39. Biasse, G. (1980) Struct. Bonding, 42, 1.
40. Levine, A.K. and Palilla, F.e. (1964) Appl. Phys. Lett., 5, 118.
41. Barendswaard, W. and van der Waals, J.H. (1986) Mol. Phys., 59, 337.
42. Ranfagni, A., Mugnai, D., Bacci, M., Viliani, G. and Fontana, M.P. (1983) Adv.
Phys., 32, 823.
3
Luminescence spectroscopy
U. w. Pohl and H.-E. Gumlich
3.1 INTRODUCTION
Energy levels within the band gap of semiconductors play an important role
in both the physics and the technology of semiconductors. They control
mobility and lifetime of free carriers and give rise to both transitions
connected with luminescence emission and radiationless transitions. While
luminescence results from radiating relaxation, a considerable amount of the
relaxation usually occurs within nonradiative processes.
Luminescence spectroscopy is a very helpful tool to unravel the some-
times complicated electronic structure of localized centres and the pro-
cesses of energy transfer among different centres in the semiconductors.
Concerning visible luminescence, the II-VI compound semiconductors
are of special interest because their spectra cover the whole visible and
infrared range, while the 111-V cover a part of the visible and infrared
spectrum and the elementary semiconductors such as silicon the infra-
red range only.
Luminescence emission can be excited in various ways. The photo-
luminescence (PL) is excited by photons, cathodoluminescence (CL) by
cathode rays, electroluminescence (EL) by electric fields and thermolumines-
cence (TL) by thermal release of energy stored in previously excited solids.
This chapter predominantly refers to photoluminescence spectroscopy. A
brief survey of different luminescence processes in semiconductors (section
3.2) is followed by a description of the experimental set-up which is
necessary for their excitation and detection (section 3.3). Finally some case
studies demonstrate applications from luminescence and excitation spectros-
copy in current research (section 3.4).
53
54 Luminescence spectroscopy
3.2 INFORMA nON DEDUCED FROM LUMINESCENCE SPECTRA
3.2.1 Excitons
Excitons are localized excited electronic states in a nonmetallic crystalline
lattice. The theory of excitons developed in the 1930s is well established by
now [1]. Two models are applied depending on the degree oflocalization of
the excitation energy. The Frenkel exciton [2] occurs if the energy is
essentially localized on one ion within the crystal as realized in the alkali
halides. The opposite extreme of the Wannier-Mott exciton [3] is found in
materials with a considerable degree of covalent bonding as in II-VI, III-V
or element semiconductors. In this case the excitation energy is extended
over several unit cells and can be regarded as an interacting electron-hole
pair with a small binding energy. The effective mass theory gives a good
description of this model in analogy to the hydrogen atom. The properties of
the crystal are represented by effective masses of the electron m:
and the
hole mt and by the dielectric constant e. The model leads to hydrogen-like
energy series of the exciton with quantum number n:
f.1*e 4 1 fj 2 k2
E =E - - - -+---..,-
n 9 2e 2 h 2 n2 2(m: + m:)
with the reduced mass 1/f.1* = 11m: + 11m:. The ionization energy is repre-
sented by the gap energy E g As the valence band of a semiconductor can be
split by spin-orbit interaction or electric fields originating from crystal fields
or strains several gap energy values E~, E: can appear and thus also
different corresponding exciton series. In addition direct and indirect transi-
tions have to be considered. The last term in the formula represents the
kinetic energy which leads to a broadening of the free exciton lines.
Excitons may form complexes by a bond to defects as donors, acceptors
or dislocations [4, 5]. The exciton transition energy in these cases is lowered
by the binding energy which increases approximately linearly with increas-
ing ionization energy of the defect [6]. For an additional offset factor called
the central-cell correction see ref. 7. As the kinetic energy of the bound
exciton complex is zero the optical transitions lead to sharp lines.
Under intensive excitation of the crystal, excitons start to interact with
each other by scattering processes and by forming biexcitons. Formation
and decay of these excitonic molecules give rise to various nonlinear optical
phenomena [8, 9]. At an extreme increase of the excitation intensity and
thus of the exciton density a threshold value called the Mott density is
reached where the Coulomb bond between individual electrons and holes is
broken and a plasma state is formed.
The general features of excitonic systems described above provide much
information on crystals by the study of exciton transitions. By recording the
free exciton luminescence the gap energy E g , its temperature dependence,
Information deduced from luminescence spectra 55
and a possible splitting of the valence band can be determined. In hetero-

epitaxial films the magnitude of the strain-induced splitting is a measure of
the degree of lattice relaxation. Using the energy-dependent absorption of
the exciting radiation the strain profile in the depth of the layer can be
probed. As the spatially extended Wannier-Mott excitons are very sensitive
to perturbations of the periodic potential, a broadening of the lines indicates
fluctuations originating from crystalline imperfections or impurities.
Further details can be read from the bound exciton transitions. Good
crystalline samples show sharp lines which allow the identification of donors
and acceptors; and often also their charge state and an estimate of their
concentration. While the complexes of a neutral donor D with an exciton X
labelled (D, X) and the corresponding acceptor complexes (A 0, X) are stable
(see for example, ref. 10), the complexes of charged impurities exist only
within a restricted range of m: /m:.
The intensity ratio of the lines resulting from free and bound exciton
recombinations can be taken as an additional measure of quality and
doping. Some examples of information deduced from excitonic features are
reported in section 3.4.1.
3.2.2. Donor-acceptor pairs

The existence of pairs of oppositely charged defects in semiconductors as
mentioned above provides efficient radiative recombination centres for
excitons [11]. While the ground state of the system is given by the ionized
donor and acceptor with interim purity distance R, photo-excitation leads to
a neutral pair in an excited state. This state can be described as an exciton
bound to the ionized donor-acceptor core or as a neutral donor and a
neutral acceptor perturbed by the interaction with each other [12]. A simple
approach based on the effective mass theory leads to the recombination
transition energy
hv= Eg-(E o + E A )+ e 2 /eR
EA and ED being the ionization energy of the donor and acceptor respective-
ly; additional van der Waals and s-p exchange corrections are discussed in
ref. 11. As donors and acceptors occupy definite sites within the crystal
lattice the distribution of different pairs with distances R j is determined by
lattice statistics. The validity of this basic approach was demonstrated first
by highly resolved spectra of zero-phonon lines due to donor-acceptor pairs
in GaP where an unambiguous assignment of pair lines to the impurity
distances R j has been given [13].
Even if the zero-phonon lines are not resolved the series of pronounced
phonon-assisted transitions dominated by LO phonons can be detected. The
ratio of the phonon peaks with numbers n of one series is given by a Poisson
distribution In= Ion!/N n with the number N of emitted phonons of the order
of unity. Different series originating from different donor-acceptor pairs

can be distinguished if EA + ED differs or R j belongs to different sublattices
in compound semiconductors. In type I pairs both donor and acceptor
occupy either an anion or a cation site; in type II pairs mixed occupation is
realized.
3.2.3 Deep level centres

Deep level centres often form localized states in the fundamental band gap
of semiconductors. Four types of information can be extracted from the
experiments for the understanding of deep level structures
1. the chemical nature and the local geometrical arrangement of the ions
which participate in the defect structure;
2. the electronic structure of the defects, which means the free or bound
electronic states, their origin and their properties;
3. the vibronic states connected with the electronic states, which implies the
phonon coupling and local lattice distortions resulting from Jahn-
Teller interactions;
4. the energy transfer between different deep level systems within the
semiconductors.
Deep level centres originate from intrinsic defects (vacancies, antisites,
dislocations), impurities (substitutional, interstitial) or complex structures
(combinations). As deep level centres in low concentration compete very
often with numerous other defects in the crystals [14] the first problem
seems to be the most difficult one.
The most prominent deep level luminescence centres are rare earth ions
(f-elements) and transition metals (d-elements) in solids when the concentra-
tion is so low that the centres do not interact with each other [15]. The
electronic states originate from excited levels of the partially unfilled inner
shell of f or d-electrons and can be described by both the quantum
mechanical theory of free ions and by the basic approach of crystal field
theory [16-18].
In the case of 3d ions, the system of the resulting energy levels can be
labelled in terms of the transformation properties of the electron eigenfunc-
tions with respect to the point symmetry of the crystal field. For the rare
earth ions the crystal field potential is weak compared with spin-orbit and
electron-electron interaction because it is shielded by the outer 5s and 5p
shells of electrons. Details are given in chapter 4 of this book. In contrast to
the rare earth ions, the 3d electrons are on the outside of the ions and
therefore very sensitive to the effects of the crystal field as exemplified in
section 3.4.2.
Both rare earth ions and transition metals in crystals show zero-phonon
lines and phonon satellites. A study of the zero-phonon line fine structure
Information deduced from luminescence spectra 57
and of the shape of the phonon-assisted transitions determines the vibration

properties of the local ion environment. The usually observed broadening of
the zero-phonon lines is due to different effects, entitled inhomogeneous and
homogeneous broadening.
The inhomogeneous broadening arises because the sites occupied by
the optically active ions are not identical, as a result of strains, disloca-
tions and fluctuations of the local potential due to neighbouring, often
unintentionally introduced impurities. Broad band excitation (or detection)
therefore gives rise to inhomogeneously broadened lines in emission (or
excitation) spectra. This is the field of site selecting spectroscopy with
narrow band tunable lasers for selecting special sites of the ions within the
crystals (section 3.4.2).
Homogeneous line broadening, on the other hand, is due to the finite
lifetime of both the initial and final states of the transition. As the
Heisenberg uncertainty principle states, there is an uncertainty in energy of
a state controlled by its lifetime. When this lifetime is a function of
different relaxation processes, the homogeneous line broadening is a
sum of all relaxation rates. Among the relaxation processes, the direct,
the Orbach and the Raman processes have to be taken into account
[15].
So far the deep levels are considered as free ions in a static environment
which shifts levels and lifts the degeneracy of some of them (internal
transitions). While some of the impurities causing deep levels are stable
under irradiation in the visible part of the spectrum (e.g. Mn in II-VI
compounds), other transition metals change the valence state (charge
transfer processes) when the exciting light leads to transitions from the
valence band to deep levels or from deep levels to the conduction band (e.g.
V, Fe and Co in II-VI compounds). In luminescence excitation spectroscopy
these transitions give rise to charge transfer bands from which the energies
needed for the liberation of charges from transition metals within crystals
can be determined.
The situation is much more complicated when considering complex
defects [14] which typically consist of more than one defect site. In almost
all semiconductors, such complex defects are found, influencing more or less
strongly the carrier recombination, the lifetimes and the degradation behav-
iour. Among these complex centres several types of chemical structures are
known. One of them is the impurity cluster occurring at higher doping rates
[19]. Since decay times and energy of the luminescence emission often
deviate strongly from the properties of isolated centres time-resolved and
site selection spectroscopy (sections 3.4.2 and 3.4.3) are helpful tools to
identify the nature of the complexes. The most famous example of the II-VI
compounds is the combination of cation vacancies with impurities such as
chlorine or aluminium leading to the characteristic 'self-activated emission'
in ZnS and ZnSe [20].
3.2.4 Energy transfer

Energy transfer phenomena are concerned with the general question of how
the energy which initially is localized at an excited centre called an 'energy
donor' is transferred to another centre labelled an 'energy acceptor'. If donor
and acceptor have the same term level structure (energy match) the transfer
is called energy migration and is essentially described by a diffusion process.
If, however, an energy mismatch exists between donor and acceptor, the
transfer is referred to as energy transport.
The parameters determining an energy transfer are the distance (and thus
the concentration) of the involved centres, the kind and coupling strength of
their mutual interaction and their energy match.
Energy transfer processes lead to a temperature and concentration
dependence of the radiative transition intensities of (energy-) donor and
acceptor centres and can thus also account for quenching phenomena.
Information on the transfer probabilities and the underlying interactions
can be deduced from time-resolved and transient spectroscopy of the
luminescent centres. In the study of the dynamic processes the lumines-
cence of a donor and an acceptor centre shows specific features. The
processes are similar to the nuclear decay scheme of a mother and a
daughter nucleus. If the relaxation probabilities of the donor and acceptor
centre are of the same order of magnitude an initial increase of the
acceptor relaxation after pulse excitation of the donor centre is observed.
Often this phenomenon is a first hint of energy transfer. If, however, the
relaxation probabilities of donor and acceptor are significantly different the
observed lifetime of an acceptor centre is equal to the longer lifetime of
either the excited donor or the acceptor state. If the energy transfer occurs
by emission and reabsorption of photons the dynamics of the donor
luminescence gives no evidence of this transfer process. The relaxation is
unchanged compared with the case of an isolated, noninteracting centre.
More often, however, a radiationless energy transfer is realized by multi-
pole or exchange interaction between donor and acceptor states. As this
type of transfer depopulates excited donor states earlier than in the case of
an isolated system, the decay of the donor luminescence becomes
faster compared with that of an isolated system. As the transfer probabili-
ties are distributed statistically the decay of the donor state is no longer
exponential.
Models for different transfer mechanisms and possible energy migra-
tion within the donor system have been developed in order to account
for the observed dynamic processes. The basic ideas referring to dipole-
dipole interaction [21] and exchange interaction [22] are described in
Chapter 2. The study of the energy transport helps to identify the inter-
acting centres, the mechanism of their interaction and their electronic
structure.
Experimental equipment 59
3.3 EXPERIMENTAL EQUIPMENT
3.3.1 Principles of operation

This section is intended to give a brief survey of the instrumentation which is
required for the optical measurements mentioned in the previous section (for
a more basic introduction see, for example, refs 23 and 24). Usually the first
step in the investigation of an unknown system is the measurement of the
optical absorption ranging from the band gap energy of the crystal down to
the infrared region. The peaks (dips) and bands of the absorption spectrum
are related to transitions from the ground state of the system to excited
energy levels. The radiating part of the relaxation following the excitation is
the main field of luminescence spectroscopy. The basic set-up for lumines-
cence spectroscopy is outlined in Fig. 3.1.
Fig. 3.1 Basic set-up for luminescence spectroscopy.
The radiation from the excitation source (laser or lamp, continuous or

pulsed mode) may be spectrally selected by filters (bandpass or cut-oft) and
polarizers (linear or circular) before exciting the studied sample. The samples
are usually kept at low temperatures and may be exposed to external
magnetic or electric fields. The luminescence emitted from the samples might
be polarized by the anisotropy of centres and can thus be analysed as a
function of the degree of polarization before being spectrally selected by a
monochromator. Only the spectrally selected light is recorded by the
detector which generates an electrical signal for the recording unit.
Three kinds of spectra are generally measured, which provide comple-
mentary information.
1. Luminescence or emission spectra are recorded by scanning the mono-
chromator which analyses the energy emitted from the sample under
fixed excitation conditions. In general the emitted photon energy is lower
than the excitation energy (Stokes' law).
2. If the excited states leading to a specific emission energy are examined,
excitation spectra are recorded. In this case the position of the emission
monochromator is fixed at the selected emission energy and the energy
of excitation is tuned by means of a dye laser or a broad band lamp
combined with a scanning monochromator. As a result the luminescence
output is recorded as a function of the excitation photon energy.
3. If the time behaviour (transient) of the emission is to be studied, energies
of both the excitation and the emission are fixed and the time-dependent
luminescence intensity after the end of the pulse excitation is recorded.
Emission and excitation spectra are presented as graphs of intensity

versus (emission or excitation) photon energy or versus wavelength. Gen-
erally the intensity is presented in terms of arbitrary units instead of
radiometric absolute values. Abscissa units for photon energy are usually
electronvolts or wavenumbers v (cm -1) (reciprocal of wavelength in units of
centimetres) and nanometres in the case of wavelength.
In the following sections the components depicted in Fig. 3.1 are de-
scribed in more detail.
3.3.2 Excitation sources

The type of the required light source essentially depends on the kind of
spectra to be recorded. Generally the following parameters of the light
source have to be considered: spectral distribution, time characteristic,
polarization and the spatial radiation characteristic. Any efficient excitation
of a studied luminescence centre requires light sources whose radiant flux is
concentrated in the spectral range where the centre can be excited. Hence
line sources such as lasers are superior to broad band sources because the
entire output can be contained in a narrow line. Although lasers usually
have a poor conversion efficiency of input pumping energy to light output
energy (typically 10 - 3) the coherence of the laser light leads to an extraordi-
narily high excitation efficiency. Additional advantages of lasers are the
extreme collimation of the beam which enables the utilization of the entire
available emission, low noise and the possibility of extremely short pulse
width generation for dynamic studies.
The energy range of tunable lasers, however, is still restricted. Thus the
classical broad band radiation sources such as lamps are still indispensable.
We first will refer to lasers commonly used in optical spectroscopy and
subsequently describe suitable lamps for spectroscopic studies in a wide
energy range.
Lasers are generally used for recording emission spectra if the lumines-
cence can be excited by band-band transitions or if the (fixed) laser photon
energy coincides with an excitation resonance of the studied centres or at
least of a centre which can transfer energy to the studied systems. In
addition excitation spectra within a restricted range can be obtained with
tunable lasers, e.g. for the study of excitation zero-phonon lines.
The characteristics of laser radiation depend on the amplifying active
medium, the construction of the optical resonator and the pumping mode.
Since there are many basic texts on lasers (see, for example, ref. 25 or 26)
only a brief survey of some lasers is given; see Table 3.1.
The amplifying media of lasers consist of gases, liquids or solids contain-
ing the optically active atoms or molecules whose energy term levels allow
an inversion of the population of an upper excited state with respect to a
lower state. This population inversion is achieved by different techniques of
Table 3.1 Typical data for different lasers*
Laser medium Active ion Pumping Mode Main line(s) (nm), Pre.(W) Ptyp(W) Notes
He-Cd gas Cd+ Electrical CW 325.0 (0.2)/441.6 (1.0) 10- 2 CW UV laser
He-Ne gas Ne Electrical CW 543.5 (0.2)/611.8 (0.1)/632.8 (1.0) 10- 3 Several lines near 1152 run
Ar gas Ar+ Electrical CW 351.1 (0.2), 457.9 (0.2), 476.5 (0.3) 101 351 nm has severallin,es,
488.0 (0.8), 496.5 (0.3), 501.7 (0.2) UV special optics
514.5 (1.0), 528.7 (0.2)
Kr gas Kr+ Electrical CW 337.5 (0.6), 406.7 (0.3), 413.1 (0.5) 10 1 338 run has several lines,
468.0 (0.1), 530.9 (0.4), 568.2 (0.3) UV special optics
647.1 (1.0), 676.4 (0.3), 752.5 (0.3)
N2 gas N2 Electrical Pulse (5 ns) 337.1 10- 1
Excimer gas XeCI* Electrical Pulse (20 ns) 308 10 1
KrF* Electrical Pulse (20 ns) 248 10 1
YAG:Nd garnet Nd 3+ Optical Pulse (1 ms) 1064 (1335/1319) 10 1 Q-switch pulse 10 ns
YLF:Nd garnet Nd 3+ Optical Pulse (1 ms) 1053 (1047/1321) 101 Q-switch pulse 10 ns
Ah03:Cr ruby Cr 3+ Optical Pulse (0.5 ms) 694.3 Q-switch pulse 20 ns
AI 20 3:Ti Ti 3+ Laser CW 700-1100 Tunable
sapphire
Dye in solution Rh6G Laser Pulse-CW 560-630 Tunable
Other dyes Laser Pulse-CW Near UV to near IR Tunable
Laser medium with active ion, pumping (electrically, usually by gas discharge; optically, usually by flash lamps), mode of operation (continuous wave
(CW) or pulse mode with order of magnitude of typical pulse duration time), main lines (lines which require a special resonator arrangement are separated
by /) and relative power with respect to the strongest line, typical total average output power (order of magnitude).
'pumping'. In gas lasers the excitation of the active ions is performed by an
electric gas discharge while most solid state lasers are optically pumped by
(flash or gas discharge) lamps. The lasing activity consists of a stimulated
emission of the excited ions in an optical resonator. Tunable laser operation
is achieved with broad band emitting active media. In those lasers normally
one of the two mirrors of the resonator is replaced by a reflection grating
which enables the selection of a narrow emission line. By tilting this
dispersive element the laser wavelength can be tuned within the emission
band. Tunable lasers are usually pumped by fixed frequency lasers. The laser
radiation is usually linearly polarized.
Some laser media allow laser operation on various lines. Often these
systems are designed for operation on one of these lines only. Some systems,
however, enable a selection of one line either by tuning a dispersive element
(as a prism) or by using appropriate (exchangeable) dielectric mirrors.
The typical laser linewidth (t12j2= 10- 6 -10- 9 ) can be further decreased
by use of a mUltiple interference Fabry-Perot etalon which in tunable lasers
has to be tuned synchronously.
Laser radiation occurring in the near infrared can be transformed into the
visible (or near ultraviolet) spectral range by frequency doubling or tripling.
As the nonlinear effect of (second) harmonic generation with birefringent
crystals requires intense fields this technique is normally used with pulsed
lasers.
The time characteristic of the laser output power is determined by the
pumping mode and the resonator design. Steady state pumping enables
continuous wave (cw) operation as well as pulse generation by mode-locking
techniques. On the other hand, pulsed pumping principally leads to pulsed
laser activity. Because of the gain threshold of the active medium the laser
pulse is shorter than the pumping pulse. Additional optical switching
elements in the resontator (Q-switches, nonlinear absorbers) allow a further
pulse shortening and a simultaneous increase of the laser power during the
pUlse.
Lamps are generally used as excitation sources whenever a continuous
spectrum over a broad spectral range is required. This is especially the case
in many absorption and excitation spectra. Lamps represent point sources
which emit over a solid angle of 41t. Consideration of the usable aperture in
the experiments and the corresponding light collecting systems (lenses or
mirrors) are thus of basic importance. The small amount of used light
(typically a few per cent) is drastically further reduced by the spectral
selection of the desired excitation energy from the broad band spectrum of
the lamp.
In the visible spectral range the black body radiation of incandescent
tungsten-halogen lamps can be used as a smooth radiation source. As the
main part and the maximum of die spectral radiant flux occur only in the
infrared the radiation has to pass IR blocking filters (and water) before being
focused on the slits of a monochromator and on the sample. In the blue and
near ultraviolet spectral range arc light sources have a higher irradiance
than tungsten lamps. Usually high pressure xenon and mercury arc lamps
are employed operating at pressures up to 100 bar and at temperatures up
to 6000C. Under these conditions the atomic line spectra are broadened,
yielding structured continua. The variations of the wavelength-dependent
intensity have to be taken into account by a suitable compensation in the
recorded spectra; see section 3.3.4. Enhanced ultraviolet radiation can be
provided by deuterium or helium lamps. Figure 3.2 shows the spectra of
some lamps commonly used in luminescence spectroscopy. The ordinate
values refer to a distance of 0.5 m from the lamp.
--
'o.!r.~~~~~~~~~~~~~~~~~~~~~~~~
~ t== - - - +-~ ~_ -=- ~_
~ .200 watt mercury_lamp - - --
_ .-
. m-_'-
g
oo
t;;
"'E-
:1--I----lI--.r.--~--H-_I__-H-I---I---~ r-- -l-
l~.O
:.~_
.
tg --
.0
3 .0
2.
1111 --
1'-
\:
--
-b~ ~ .:::
- -->:) -'0
---=f----:.~:j:t:t=t~
150 watt xenon lamp
-- -- m-~
-'- -.--+~
.--
!-
U- j
.
~.
--
--l&'--M - - IV . \ - - ~;:;;7
I-
p--\,. - -..... 4:-
ug. ~:: ~=-~Ii~~'~~~~~~/.:i~

:i!
oo
!!!
~'.~
0 ii~~~ii~~
f--=-- ~i-_~J~ri~~~--i'~
~--~_-~-~-'i--~"i- I---- I-- _- - _ -
'Ii .3 1- - r- -I--- -h...... 100 watt quartz halogen (filament) lamp --
E 02 1-- f- - -r-7 -
~ .0:: ~._ f - -f---- / - 1---+---+----1- -- --- - -.-I-- --l------l
1: ~::
~ 0.05
~ ~:: ~-L ~h~t.~~~~~~~~~~:~~-~~=~~.~..~.-~-~-~-~~~~~~
'" I--- ~50 watt deuterium lamp
e
~ .02
O.OIS
. !!._o.~ _
::; 006
0.006
_ _ ~
0.005
00< II\.
000 l'
O.O(n$
0.002 ~=:j==!:=t=:t==~=+:s::*~~=f4=4=+=+=+=~
1_
0.00' L200-...L-300..L-L-.,.,l,oo:--..-L-----;500..!:--L--600:;!;;-L-.LJ----;;!,00~---L..--..1600~--L---;;""';;;;'
Wavelength (nm)
Fig. 3.2 Spectra of various lamps used as broadband radiation sources (Catalog
Oriel).
3.3.3 Temperature controlling and application of external fields

Optical measurements are usually performed at low sample temperatures
(down to 1 K). Generally the appropriate ambient temperature is provided
by placing the specimen in a cryostat. In bath cryostats and continuous flow
cry os tats the cooling is provided by evaporation of cold liquid gases (helium,
nitrogen). The general assembly consists of a concentric construction with
an inner sample chamber.
Continuous flow cryostats comprise a heat exchanger, attached to a cold
finger, or surrounding the sample chamber and containing exchange gas
(helium), a radiation shield and a vacuum case (p< 10- 4 mbar). The cold
liquefied gas from a separate storage vessel is drawn through a vacuum-
insulated transfer tube and is circulated through the heat exchanger by a gas
flow pump which operates at room temperature. The exhaust gas cools the
shield by heat transfer. The coarse temperature control is obtained by a
flow-adjusting needle valve, and the fine control by a temperature sensor
and a heater fitted to the heat exchanger. Thus a temperature ranging from
4 K to 300 K (or higher) can be fixed.
Bath cryostats are especially suitable in the low temperature range. The
samples are mounted on a sample holder and immersed into the cryogenic
liquid. The thermally isolated sample chamber is usually surrounded by a
concentric chamber filled with liquid nitrogen venting into air which serves
as a radiation shield. The shield is surrounded by a further vacuum case. The
liquid gases which cool the sample are transferred from mobile storage
vessels into the sample chamber which subsequently is kept under either
normal or lowered pressure. The pressure-dependent boiling points of the
liquids thus enable the variation of the temperature within a restricted range
(helium 4He, 4.2-1 K; nitrogen, 77-63 K (solidification); hydrogen, 20.4-
13.9 K). The lower limit is given by the exponential decrease of the vapour
pressure and the pump throughput. In the case of helium the superfluid
phase below 2.18 K (at 51 mbar) provides highly efficient cooling owing to
the high thermal conductivity. Data of some commonly used cryogenic
liquids are summarized in Table 3.2.
Refrigerators with a closed cycle of the cooling gas are used if no helium
recovery is installed in the laboratory. The cooling is obtained by expanding
a highly compressed gas. With helium as a working gas a temperature as
low as 6 K can be achieved by this principle. The system comprises a
compressor with oil separation equipment (oil vapour solidifies at low
temperatures) and a refrigerator unit with an expansion chamber and a heat
Table 3.2 Data for cryogenic liquids at normal pressure (1013 mbar): boiling point
T boll , liquid density Q, heat capacity cp at T boll volume ratio of gas (at 273 K) and
liquid (at T boil ) and latent heat of vaporization Iv
3He 4He H2 Ne N2 Ar
TboiJ (K) 3.2 4.2 20.4 27.1 77.3 87.3

Q (kgm- 3 ) 59 125 71 1024 804 1399
cp at Tboll (kJ kg- 1 K -1) 4.4 9.3 1.8 2.0 1.1
V...Jl'tiquid 438 699 788 1338 644 785
Iv (kJkg- 1 ) 7 21 448 87 199 163
exchanger for precooling. The heat exchange can be designed as a moving

piston ('displacer') with several stages for subsequent temperature lowering.
All optical cryostats must be equipped with air-tight sealed windows trans-
parent for the spectral range under study. Additional application of external
fields requires a sufficiently large sample chamber or a special assembly.
The study of the effect of electric and magnetic fields on the optical
properties of solids helps in understanding the nature of electronic and
vibronic states. The application of hydrostatic pressure provides an efficient
tool to introduce high electric fields preserving the symmetry of the crystal.
Uniaxial pressure, on the other hand, introduces an electric field with a
unique axis which reduces the crystal symmetry. The same is true for the
application of magnetic fields. In any case, the application of an external
constraint changes the electronic and vibronic properties of the crystal and
of localized centres in the samples.
Hydrostatic pressure essentially changes the chemical bond and conse-
quently band structure, phonon modes and, at sufficient strength, even the
crystal structure. Thus zincblende structured A-B compounds (such as the
Zn and Cd chalcogenes) are known to undergo a phase transition into an
NaCl structure in a medium pressure regime and a metallic (mostly
orthorhombic) structure at high pressure [27]. These changes are closely
connected to the ionicity of the chemical bond.
Uniaxial stress (Stark effect) and a magnetic field (Zeeman effect) lift
partially or fully the degeneracy of electronic levels. In particular, in the
study of zero-phonon lines related to localized defect centres the symmetry
of the involved electronic states may be deduced from the resulting splitting
pattern. The parameters to be determined are the number of components for
the external fields along some particular crystalline axis (in cubic crystals
along [100], [110] and [111]), the energy shift of each component, their
relative intensity and their polarization (n or 0'). The comparison of these
experimental findings with calculations based on group theory allows the
assignments of electronic states to energy term levels of well-defined symme-
try [18]. In the following some experimental details of pressure and Zeeman
set-ups are considered.
For the application of high hydrostatic pressure the gasketed diamond
anvil cell has proved to be the most suitable technique [28]. It can be used for
optical measurements ranging from the ultraviolet to the infrared (apart from
a range with diamond absorption between 4 Jlm and 6 Jlm) with pressures up
to 1.7 X 1011 Pa (1.7 Mbar) within a cell size of typically 0.1 mm height and
0.2 mm diameter. The cell is composed of two opposed transparent diamond
anvils with parallel faces. The anvils are pressed against a metal platelet with a
centred hole of the cell diameter (Inconel gasket). This pressure chamber is
filled with a fluid acting as a pressure-transmitting agent, the sample and some
tiny pieces of ruby [29]. The ruby is used as a pressure gauge by recording the
pressure-induced line shift of the Cr 3 + R1 luminescence (0.365 nm GPa -1,

linear at least up to 30 GPa). As an additional feature the line broadening
indicates nonhydrostatic spatial pressure gradients. As a convenient pressure-
transmitting fluid a mixture of methanol:ethanol (4: 1) is used, remaining
hydrostatic up to 10 GPa at room temperature.
Uniaxial stresses are generally applied by two metal anvils being pressed
with a known force directly on the sample. The pressure is controlled by the
force applied on the sample faces with known size. The force can be created
by a screw construction or a pressure chamber with a piston and is
transferred by a rod to the sample holder in the cryostat. Near the anvils the
uniaxial character of the stress is disturbed by shear forces. The mechanical
properties of the crystals therefore limit the maximal pressure to about
10 kbar (1 GPa). In addition only the volume in the middle of the sample
should be measured. Careful sample preparation with parallel faces has to
be carried out to avoid mechanical destruction.
The effect of an external static magnetic field on the optical transitions of
a studied centre is referred to as the Zeeman effect. In contrast to the effect
of uniaxial pressure the observed splitting of zero phonon lines is connected
to the removal of the spin degeneracy of the involved electronic levels and
provides additional information.
High magnetic fields are usually generated by superconducting coils.
Conventional type II superconductors are employed to reach a high critical
magnetic induction Be. Below lOT wires with Nb-Ti alloy (Te=9K)
filaments on a copper billet are convenient. At higher fields (up to 30 T)
brittle intermetallic compounds such as Nb 3 Sn (Te = 18 K, Be = 23 T) are
used. The combination with a low resistance normal conductor provides a
protection if the superconductor becomes resistive owing to magnetic flux
jumping or cryogenic failure.
For the choice of a suitable magnet the required field strength and
homogeneity and the necessity of radial access to the high field region has to
be considered. Magneto-optical studies often require a geometry with B
being perpendicular to the propagation vector k of the light (Voigt configur-
ation) in addition to the usual parallel set-up (Faraday configuration). For
this purpose a split coil geometry, the Helmholtz pair, is used despite its
inefficient high field production. The Helmholtz pair produces a homogene-
ous central field region with a rapid fall-off at the edges.
In the cryostat the sample chamber is usually separated from the helium
bath containing the magnet. This allows one to change the sample while the
magnet rests at low temperatures and to vary the sample temperature.
3.3.4 Filters, spectrometers and polarizers

Filters and spectrometers are used for the spectral selection of radiation.
While spectrometers are more versatile with respect to wavelength and
bandpass, filters are more convenient and highly efficient despite their
simplicity. As often a combination of both means is appropriate for
spectroscopy, we give a brief description of filters as well as spectrometers.
Filters may be selected to have a (narrow or large) bandpass or to provide
a wavelength cut-off (usually long pass). Most commonly used types are
absorption filters and interference or birefringent filters. They usually have
high peak transmission (up to 90%) and a large area.
Absorption filters consist of glasses doped with heavy metal or rare earth
ions, by gelatine films or by liquids containing solved complexes. The
absorbed radiation leads to a temperature increase in the filter and may
additionally lead to a luminescence at long wavelength. In particular, in the
case of prolonged UV irradiation (A < 330 nm) a deterioration of the trans-
mission and a (small) shift of the cut-off edge to longer wavelength may
result from photoinduced processes.
Interference filters are essentially Fabry-Perot etalons built by a dielec-
tric-filled spacing between two reflecting surfaces. A (single) small spacing
leads to a large spectral range of the transmission within one order but
inversely also to a small order separation. A desired characteristic can be
designed by a combination of different etalons (realized by multilayer
dielectric stackings). The ratio of transmitted light within the bandpass to
that at all other wavelengths determines the quality of the filter and may be
of the order of 1: 10- 4 . The peak transmission of interference filters can be
tuned either by inclination of the filter with respect to the beam direction
(tuning to longer wavelength) or with interference wedge filters by moving a
slit parallel to the apex.
Birefringent filters such as the Lyot-filter have similar characteristics. In
particular, the Lyot filter can be tuned as well (by rotation).
Spectrometers use a dispersive element to separate light of different

wavelengths spatially: a grating, a prism or an interferometer. Comparing
spectrometers for a given spectral resolution, a grating instrument is found
to have highest light throughput. We therefore focus on the commonly used
grating spectrometer although prism instruments might have lower stray
light and are free of overlapping orders and interferometers have higher
resolving power at sufficiently strong input signals.
Usually the elements of a grating spectrometer are arranged in a symmet-
rical Czerny-Turner mount. The light passes an entrance slit and is
collimated by a concave mirror to a parallel beam which is directed to an
opposed reflection grating. The dispersed light is focused by a second
concave mirror on the exit slit which selects a transmitted spectral range.
Turning the grating on its axis by a sine drive allows one to scan spectra
with an equidistant wavelength scale. Normally the reflection grating is
'blazed' at a particular wavelength Ab in order to shift the intensity
maximum from the zero order to the order where Ab should be detected.
This is achieved by forming the grooves with a steplike shape and thus
fulfilling the reflection condition in addition to the diffraction condition for
Ab' The FWHM of the grating efficiency then roughly ranges from (2/3)Ab to
(4/3)Ab'
The spatial separation of different wavelengths in the plane of the exit slit
(or usually its inverse in units of nanometres per millimetre) is given by the
reciprocal linear dispersion dA/dl = (1/f) dA/d<P with the focal length f of the
concave mirrors and the angular dispersion dA/d<P of the grating which is
well approximated by the ratio of order n to the groove density d. (In practice
the wavelength dependence has to be taken into account if measurements
cover a wide spectral range.) The resulting bandpass of a monochromator is
given by the product of the slit width (optimized as being equal at the
entrance and the exit) and the reciprocal linear dispersion. It should be
noted that a lower limit of the chromatic resolving power A/dA is not
determined by the slit width but rather by the product of angular dispersion
and grating width.
The light flux through the spectrometer is given by the product of the
solid angle of the cone of acceptance (angular aperture) and the size of the
slits. Therefore condensing lenses should carefully match the spectrometer
aperture to the radiation source (thus ensuring an exact illumination of the
first mirror) and should uniformly illuminate the entrance slit. Overfilling
the aperture increases the stray light ratio, and underfilling reduces the
resolving power. Provided that a good alignment exists the flux throughput
increases with the square of the slit width.
Stray light, e.g. light emerging from the exit slit with wavelength outside
the selected range, may worsen the signal-to-noise ratio. Different sources
include ruling errors of the grating and reflections within the instrument
from the matt black surfaces. By using holographic gratings a strong
reduction of stray light can be obtained and 'ghosts' resulting from ruling
errors can be avoided (the efficiency, however, is usually smaller than that of
ruled gratings). The main contribution to stray light originates from
wavelengths close to the nominal setting. This part may be effectively
reduced using a double spectrometer. In this case the stray light ratio is
given by the square of that of a single instrument of the same type. Additive
or subtractive dispersion can be arranged, the latter being especially useful
in ultrashort-time spectroscopy.
As to the wavelength dependence of the transmission characteristic of a
grating spectrometer, the Wood grating anomaly should be mentioned. In
contrast to the slowly varying grating efficiency this effect leads to more or
less sharp cusps especially for the light perpendicular to the grating grooves.
The anomalies occur if the diffraction angle of a particular wavelength in
any order reaches 90. According to Rayleigh the irregularities then orig-
inate from a redistribution of the energy contained in the disappearing order
among the remaining orders.
Polarizers are used to produce or analyse linearly or circularly polarized
light. Linear polarizers employ dichroism, Brewster angle reflection or
birefringence. Most popular are Polaroid film polarizers built by stretched and
coloured long-chain polymers. The extinction ratio of absorbed to transmitted
component is of the order of 10- 4 ; however, films with a good ratio show poor
transmittance. High transmittance is obtained with Glan-Thompson prism
polarizers which use a birefringent material (calcite) cut into two prisms. While
the extraordinary ray passes through the polarizer without deflection the
ordinary wave experiences a total reflection at the cut and is removed by an
absorber.
Circularly polarized light can be achieved by a combination of a linear
polarizer and a subsequent quarter-wave plate (mica or quartz) which
produces a quarter-wave phase difference between the ordinary and extraordi-
nary wave (propagating in the same direction but with orthogonal electric
fields). By setting different orientations of the polarizer with respect to the
optical axis of the quarter-wave plate, left or right circularly polarized light can
be obtained.
3.3.5 Detectors and signal processing

For the conversion of the incident luminescence light into an electrical signal
several physical effects can be used. In the case of photoemission or photo-
excitation, random fluctuations known as noise normally are superimposed
on the signal. In luminescence spectroscopy various properties of a detector
playa significant role [30].
1. The sensitivity, usually given in amperes of the electric signal current per
watt of the incident light power. The detector may also be characterized by
its quantum efficiency, defined as the ratio of signal electrons to incident
photons.
2. The range of the wavelengths in which the detector is sensitive.
3. The signal-to-noise ratio (S + N)/N. Usually the noise of the detector is
characterized by an equivalent of the incident light which would cause the
same noise level as the detector does (noise-equivalent power (NEP)). The
reciprocal of the NEP is designated as the detectivity D.
4. The upper limit ofthe frequency of the detector which is determined by the
lowest build-up time of the signal triggered by an ideal rectangular light
pulse.
5. The linear range of the detector where the signal is proportional to the
incident radiation intensity.
In the following a brief characterization of commonly used photodetectors

will be given. A more detailed consideration of basic principles is given by
Keyes [31].
In photocells the external photoeffect is used for the registration of

incident light. The electrons are liberated by the light from light-sensitive
thin films (alkali compounds or GaAs) on a cathode and accelerated to an
anode, giving rise to a photocurrent which can be amplified. The spectral
responses of some materials usually taken for photocathodes are drawn in
Fig. 3.3. The build-up times r of photocells are short owing to the small
capacity of the anode (r < 10- 10 s). In short-time spectroscopy photocells
allow the resolution of subnanosecond pulses. The practical use is limited by
fluctuations of the electron transit times between cathode and anode.
100
10
0.4 0.6 0.8
Wavelength (IJm)
Fig. 3.3 Spectral response and quantum efficiency of several photocathode ma-
terials [31].
The photomultiplier tube is the most sensitive of all optical detectors in

the visible and neighbouring spectral ranges. Its main parts are the photo-
cathode, the dynodes and the anode. At the photocathode the photoelectric
effect takes place as previously described for the photocell. The released
electrons are accelerated in a field between the cathode and an electrode
called the first dynode where more electrons are emitted. These electrons are
attracted by the next dynode, being connected to a more positive potential,
where again the number of electrons is increased. The typical gain at each
dynode is given by a factor of 4. After typically ten dynodes the photo-
current is enhanced by a factor of 4 10 ~ 10 7 . Apart from the dynode
configuration the time resolution is given by the load resistance and the
capacitance of the anode and is typically of the order of 10- 9 s.
In the case of extremely weak light intensities photomultipliers are used in
the photon counting mode. In this case voltage pulses due to the registration
of single quanta are measured instead of an integrated photocurrent. In a
discriminator photoelectron pulses may be separated from smaller pulses
created by thermionic emission and from larger pulses originating from
cosmic radiation. Compared with the analogue registration the method of
photon counting has several advantages.
1. Fluctuations due to the statistical amplification by the dynodes do not
play any role because photoelectron pulses are selected by discrimina-
tion. Also, pulses triggered by cosmic or radioactive radiation near the
tube can nearly be eliminated.
2. The leakage current at the photomultiplier tube will not be counted.
3. The counting rate can be digitally treated, facilitating the computing of
the spectra.
The upper limit of the counting rate depends essentially on both the dead
time of the tube and the time resolution of the discriminator and the counter.
As a consequence of this limit, which is of the order of 10- 8 s, sequences of
statistically distributed pulses occurring at a rate of 10 MHz can be handled.
Since the dark current of photomultiplier tubes decreases to a certain
extent with decreasing temperature the signal-to-noise ratio can be en-
hanced by cooling. In most cases, cooling down to - 40 DC is sufficient,
which can easily be achieved by Peltier elements.
A drastically improved real-time response in luminesence spectroscopy is
attained by optical multichannel analysers (OMAs). These detectors can be
divided into two classes depending on the method by which the stored
spectrum is read out [32]. In the vidicon TV camera technology the charge
pattern on the target of the vidicon is read out by an electron beam. The
other group uses a solid-state array of photodiodes and shift registers for the
read-out procedure. OMA systems are particularly useful for studying
short-time optical phenomena because the whole spectrum can be stored
after a single laser pulse regardless of its pulse duration. To reduce noise
many spectra can be integrated before read-out. OMA systems are thus
equivalent to up to 104 photomultiplier tubes working in parallel.
Often OMA systems are equipped with image intensifiers such as the
micro channel plate (MCP). The MCP consists of a photocathode coated
on a fibre optic face plate which is mounted in close proximity to a
mosaic of electron channel multipliers. These multipliers are hollow glass
fibres of 10-40 Ilm diameter with an internal resistive coating. The fibres are
arranged in a rectangular array and electrically connected in parallel by

metal electrodes on each end of the tubes. The front face of the photo-
cathode is positive. As in photomultiplier tubes, electrons emerging from the
photocathode are accelerated in the channels in which a quasi-uniform
voltage gradient is maintained. The accelerated electrons striking the walls
of the channels cause secondary emission of electrons which creates an
avalanche of electrons resulting in a gain of 103 -10 4 . The transit time from
one end of the tube to the other is about 1 ns. The electrons strike a
phosphor screen and produce cathodoluminescene.
Other image intensifiers are the photodiode-like single-stage electrostatic
intensifier with a phosphor-coated anode or comparable constructions using
the sensitive area of an OMA system as the anode.
Semiconductor detectors are widely used in addition to photomultiplier
tubes. The detection process is based on the creation of electron-hole pairs
by incident photons (internal photoeffect). Therefore the wavelength re-
sponse is determined primarily by the bandgap of the semiconductor
material. The most interesting devices are the PIN diode and the avalanche
photodiode (APD).
In the PIN diode three differently doped regions can be distinguished: a
p-type, an intrinsic (compensated) and an n-type range. The diode is
operated with a reverse bias voltage reinforcing the internal electric field and
therefore expanding the intrinsic depletion layer. Electron-hole pairs created
in the intrinsic region are separated by the field and create a signal current.
Because of the intrinsic range and the comparable high bias voltage the
signal rise time is quite fast 1 ns). For Si photodiodes the maximum
sensivity is in the 800-900 nm range.
The avalanche photodiode is similar to the PIN diode, but has an
additional p-type multiplying region between the nand i regions. The
electrons generated in the i region pass through the multiplying p-region
and generate further electron-hole pairs by collision processes. This effect
delivers a typical gain factor of 10-100.
There exists a large variety of signal processing methods which tend to
achieve a good signal-to-noise ratio. We briefly focus on only two important
techniques, phase-sensitive detection and single photon counting.
In phase-sensitive detection (see, for example, ref. 24) the signal of the
radiation source is periodically chopped (switched on and off). In excita-
tion or luminescence spectroscopy the luminescence intensity is thus
modulated at the same frequency. This allows a narrow band amplification
of the detected signal with frequencies centred at the chopping reference
frequency and thus suppressing all other noise components. In addition to
the frequency dependence the signal recovery system is also phase selective.
This allows the discrimination between luminescence centres with different
radiative decay times and thus the separation of overlapping luminescence
bands originating from different centres.
The single photon counting technique is especially useful for recording the
transient of a very weak luminescence in the nanosecond timescale. The
principle of this method is based on the fact that the transient of lumines-
cence after pulse excitation reflects the number of emitted photons which is
proportional to the probability of detecting photons. The luminescence must
be so weak (or attenuated so strongly) that not more than one single photon
can be detected after a single pulse excitation. For many laser shots the
delay time between the pulse excitation and the registration of a single
photon is measured and subsequently an abundance of different delay times
is counted. In the electronic equipment this is achieved by a time-to-
amplitude converter (TAC) creating a pulse height that is proportional to a
delay time between a 'start' and a 'stop' signal (during this time a capacitor
is charged by a constant current supply). While the start pulse is provided by
the laser during the laser shot the stop pulse is created from the detection of
the luminescence photon. The voltage of the T AC is read out and collected
in a multichannel analyser operating in a pulse height analysis mode. Thus
the collected pulse height distribution which is delivered after a large
number of laser shots scans the decay of the weak luminescence signal. It
should be emphasized that the measured transient is falsified by 'pile-up
errors' if more than one single photon is allowed to reach the detector after
one laser pulse. While this technique is well known with photomultiplier
detection its use has also been described in the near infrared using gated
avalanche photodiodes by Levine et ai. [33].
3.4 SOME CASE STUDIES
3.4.1 Near band edge emission and doping

The luminescence near the bandgap provides much information on the
structural and electronic properties of a semiconductor. The so-called near
band edge emission includes radiative relaxations of free and bound excitons
and donor-acceptor pairs. Especially for the characterization of epitaxially
grown semiconducting thin films, apart from X-ray diffraction analysis,
photoluminescence spectroscopy at the band edge region has become the
most frequently applied method. We therefore focus our interest on the
luminescence spectra of epilayers.
A basic requirement for good epilayers is the choice of a suitable substrate
which should provide a good lattice match and a thermal expansion
coefficient close to that of the layer. In addition the semiconducting epilayer
should enable efficient conductivity control by p and n doping.
The investigation of the near band edge emission allows a nondestructive
probing of the strain situation in the layer and verification of the doping
without the need of ohmic contacts. This will be demonstrated with ZnSe
epilayers on GaAs substrates which currently constitute an intensively

studied heterosystem [34]. We first refer to the problem oflattice match and
then report on recent progress in doping.
The lattice mismatch of ZnSe/GaAs structures of 0.3% at RT is quite
small. Up to a critical layer thickness of de~0.15l!m X-ray diffraction
analysis reveals a coherent pseudomorphic growth of ZnSe [35]. Because of
the slightly smaller lattice constant of GaAs the ZnSe layer is strained
accepting the in-plane lattice constant of GaAs and a larger lattice constant
perpendicular to the interface. Above de the increasing strain gives rise to
dislocations which can relax the lattice. At a thickness of about ll!m the
layers are relaxed and exhibit the bulk lattice constant of ZnSe crystals. The
compressive in-plane strain resulting from the lattice mismatch is superim-
posed by an additional differently acting strain due to the larger thermal
expansion coefficient of ZnSe with respect to GaAs. Cooling down the
ZnSe/GaAs structure from the growth temperature Tg to room temperature
(or measuring temperature) thus causes a tensile in-plane strain whose
magnitude depends on T g
The simultaneous effect of both strain components leads to a thickness-
dependent strain which effects a splitting of the valence band into an upper
light hole and a lower-lying heavy hole subband. As a consequence the free
and bound exciton lines exhibit a twofold strain-dependent splitting. Figure
3.4 shows the luminescence (spectrum a), excitation (spectrum b) and
reflection (spectrum c) spectra in the exciton energy region of a ZnSe/GaAs
epilayer [36]. The emission lines denoted X 1h , X hh , 12 and 12 result from the
recombinations of the light hole excitons, the heavy hole excitons and from
complexes of these excitons bound to neutral donors, e.g. (DO, X 1h) and
(DO,X hh) respectively. Spectrum b shows the excitation lines 12i (i=a ... e) of
the 12 emission. These lines represent transitions into excited states of the
donor-bound exciton (DO, X) with different quantum numbers of the excited
hole In, I). The line shifts of the 12 emission and the 12i excitation lines with
respect to the energy observed in ZnSe bulk crystals are plotted in Fig. 3.5
for various samples grown at different temperatures Tg (one column of data
points corresponds to one sample). All transition energies increase with
increasing tensile strain. From the different strain sensivity of the 12(hh) and
12(lh) emission lines which is also observed in the strain-dependent energies
of X hh and X 1h reflection loops (not shown in the figure) the assignments to
the (DO, X hh ) and (DO, X 1h) complexes could be established.
In addition the spacing of the 12i resonances could be calculated by a
three-particle model yielding a donor Rydberg constant which corresponds
to Al or Cl. The comparison of the 12-1 2 crossing point (Fig. 3.5) to the 12
energies of differently donor-doped bulk ZnSe confirmed that, in the
non-intentionally doped layers, Al or Cl have to be assigned to the donor in
the (DO, X) complexes instead of a Ga donor expected from outdiffusion of
the GaAs substrate. In some cases the origin of the unintentional dopants
Wavelength (nm)
444 443 442
12 ZnSe/GaAs
T=1.6K
I; X1h
I I
0
a
~
li.I;. I;. 12c 12d 12
Ii)
."!:
c:
::J
b
.c
~ 0
->
-
'in
c:
Q)
c:
c
O~-r ____~____~____~__~
2.790 2.795 2.800 2.805
Energy (eV)
Fig. 3.4 Luminescence under band-band excitation (a), excitation of the 12 line (b)
and reflection (c) of a ZnSe/GaAs epilayer in the excitonic energy range [36].
could be clarified (CI resulting from the substrate cleaning procedure before
growth, Al from residual impurities of the source materials).
A similar picture of strain probing as obtained from the measurements of
different samples can be obtained with a single sample by resonant Raman
scattering if the laser energy is tuned [37]. The variation of the laser energy
versus the energy of the Stokes lines resulting from a resonant enhancement
at the 12i transitions shows an identical dependence to that for the excitation
line shifts versus the 12 detection energy shown in Fig. 3.5. This phenomenon
has been explained by a resonant amplification of the Raman signal only by
those (DO, X) complexes which have suitable 12i energy levels with respect to
the irradiated laser energy owing to their local strain. While the lumines-
cence and excitation spectra provide superimposed signals reporting the
average strain of a broad region in a layer the resonant Raman scattering
therefore allows one to scan the depth profile of the strain in one layer by
tuning the laser.
b-E(meV)
compressive I tensile strain

+5
I-
I
I
I
.:=
:.c: 0
(/)
Q)
c
:.J
ZnSe
-5
E(eV)
2.800 2.795" 2.790

12 detection position
Fig. 3.5 Line positions of the 12i excitation resonances of the 12 emission line in
differently strained ZnSe/GaAs layers [36].
The results of these investigations are summarized in Fig. 3.6. which

illustrates the separate influence of lattice misfit and thermally induced
strain (dashed lines) and the resulting light hole and heavy hole band shifts
(solid lines) as a function of layer thickness [38]. Below the critical thickness
de the ZnSe grows with the in-plane lattice constant of GaAs and the valence
band splitting remains constant. Above de the strain is reduced owing to the
generation of dislocations until a strain-free lattice is reached at do
(~O.85Ilm). Above do the thermal strain which acts in the opposite manner
to the lattice misfit strain starts to split the light hole and heavy hole valence
band in an inverse order of energy. The shaded areas denote the layer
regions from which the response of the free exciton reflection spectra (FE)
and the bound exciton luminescence spectra (BE) originate.
As to the intentional p and n-doping of II-VI compound semiconductors
there are obvious limits given by equilibrium thermodynamics [39]. Whether
tlE(meV)
thermally indo strain
~:I----: FE
- 0
::
:c
'"
'">
"c: 0c;;_
~
~
"'c:" Co
E
lattice
~'"
o misfit strain
"
-10
Ih
-15 I
I ZnSe/GaAs
__
Ih JI TG'OWlh = 593 K
ThP = 1.6K
0.1 d, do 10 d(~m)
Layer thickness
Fig. 3.6 Splitting and energy shifts of the heavy hole (hh) and light hole (lh) valence
bands in ZnSe/GaAs epilayers in dependence of the ZnSe layer thickness. The
dashed lines represent the contributions of lattice-misfit and thermally induced strain
[38].
these limits are marked by their inherent tendency for self-compensation or by

other mechanisms such as solubility laws is still under discussion.
The basic concept of self-compensation as worked out by Mandel and
co-workers [40, 41] assumes a creation of compensating carriers by the
formation of electrically active intrinsic defects as a reaction to intentional
doping. The compensation effect results from a balance of two competing
actions. On the one hand intrinsic compensating defect centres (such as
vacancies) are created which requires a certain amount of energy. This
energy is related to the cohesive energy Ecoh per gram atom which is half the
standard molar enthalpy change for the reaction MXsolid -+ Mgas + Xgas. On
the other hand the crystal gains energy by trapping the free carriers at the
defect centres. This energy correlates with the bandgap Eg The ratio of free
carriers to ionized dopant atoms is then determined by a mass action
formalism. For single ionizable defect centres a ratio Eg/E coh > 1, i.e. a
resulting energy gain, was found to evoke a relevant compensation (com-
plete for alkali halides) while only little self-compensation should take place
for a ratio less than 0.5 (see GaAs). The II-VI compounds occupy an
intermediate position with EJ Eooh ~ 1. In this case strong self-compensation
may result from the additional action of doubly ionizable defects which in
II-VI compounds are known to originate from vacancies on cation sites
(building acceptors; in the ionic picture of the chemical bond a cation
vacancy leads to a lack of electrons at a neighbouring anion) or on anion
sites (building donors; it has been shown by Mandel [40] that self-compen-
sation by interstitial host atoms is essentially equivalent).
Mandel et al. [41] considered a simple model of an F centre to study the
second ionization level of a vacancy and found a pronounced effect of both
the cavity size and the ratio of cation M to anion X radii on the electric
properties. A small cavity leads to a small second ionization energy and thus
to a strong tendency to compensate. Independently a small ionization
energy also results from large next neighbour atoms surrounding the cavity
because of their good polarizability. If, as in ZnSe, the cations (Zn) are large
compared with the anions (Se) the cation vacancies (acceptors) form large
cavities with high second ionization energies and n-type conductivity will be
obtainable. On the other hand, the anion vacancies (donors) are surrounded
by large cations and thus have low second ionization energies. Therefore a
strong self-compensation of p-type conductivity will result.
In fact Mandel et al. gave a semi-empirical relationship between the ratio
of covalent radii RM/Rx and the intrinsic type of conductivity which
describes well the gross electrical behaviour of the II-V compounds. If this
ratio roughly exceeds 1.15 n-type conductivity is accessible (ZnSe: 1.15),
while a ratio below 1.06 leads to p-type conductivity.
In addition to the classic theory of self-compensation solubility limits
have been claimed for the usual dopants of ZnSe by statistical considera-
tions and a comparison of isocoric compounds [42].
Recently Laks and co-workers [43] drew a more complex overall picture
emphasizing that the concentration of native defects created at thermal
equilibrium is much too low to account for self-compensation. The authors
used first-principles total-energy calculations of supercells based on density
functional theory and a local density approximation. As a result a much
richer defect structure than considered in previous work was found to occur.
Thus p-type ZnSe should be compensated by native Znr +, V~n and Se~:
donors and n-type ZnSe by V~';- and Zns. acceptors. In both cases, however,
the concentrations of the compensating charges were found to be several
orders of magnitude smaller than a given dopant concentration.
According to Laks et al. [43] a similar refined picture is reported for ZnSe
slightly deviating from perfect stoichiometry. The native defects introduced
by the lack of any component may compensate n-type doping as well as
p-type doping. Thus p-type ZnSe should be compensated by the dominating
double donors Znr + in Zn-rich and by Se~: in Se-rich material. On the
other hand, n-type ZnSe should be compensated by the intrinsic acceptors
Zns. in Zn-rich and V~: in Se-rich ambient. The latter prediction is
evidenced by the SA centres known to be a prominent defect centre in n-type
Se-rich crystals.
The amphoteric behaviour found for intrinsic defects also occurs for the
dopants at different lattice sites (substitutional, anti site or interstitial) or
even for a dopant at a definite site. The latter has recently been demon-
strated by optical studies of substitutional vanadium on Zn sites in ZnSe
which forms both a deep donor and a deep acceptor [44].
In conclusion, the theory of self-compensation under thermal equilibrium
is still incomplete. Nevertheless, in practice the nonequilibrium growth
techniques such as MBE or MOVPE are supposed to overcome this old
problem, at least in the case of layer growth.
Intentional p-type doping of ZnSe can be performed either by replacing

the Zn (group II) by a group 1 element such as Li, Na or Cu or by
substituting the Se (group VI) by a group V element such as N, P or As. In
the same way n-type doping might result from group III elements such as AI,
Ga or In on Zn sites or by group VII elements, e.g. F, CI, I, on Se sites.
In the following the more difficult acceptor doping of ZnSe is considered
[34, 45]. The most studied dopants are Li, Na, Nand P. Although Li and
Na substituting Zn are known to produce p-type conductivity the samples
are not stable owing to the amphoteric character of the alkali elements
which tend to diffuse on interstitial sites where they act as donors. The
group V elements therefore seem to be more suitable. P and As, however,
both form deep acceptor levels. Therefore N has turned out to be a very
promising candidate for both MOVPE growth [46] and MBE growth [47]
of p-doped ZnSe epilayers.
The acceptor-related emission lines, referred to as 11 lines, appear at lower
energies with respect to the donor-related 12 lines mentioned above. Usually
two lines (or line groups) have to be distinguished. The 11 emission is
assigned to acceptor-bound exciton complexes (A 0, X) with the shallow
acceptors Li, Na, N or P, while the assignment to the deeper-lying emission
line Wep (or I~) has not yet been finally clarified. This line is supposed to
originate from an exciton bound to either a deep Cu acceptor or a complex
(Cuz n- VZn) or to a complex of VZn with another acceptor.
Figure 3.7 shows the near band edge emission which appears as a result
of N doping [46]. From Hayne's rule (section 3.2.1) which assumes a
constant ratio of the (experimentally determined) exciton binding energy to
the ionization energy of the impurity of approximately 0.1 an activation
energy of about 100 meV was estimated. The bound exciton emission is
accompanied by a donor-acceptor pair emission band the peaks of which
are labelled Qi' The estimated acceptor activation energy could be con-
firmed at 112 meV by the application of the donor-acceptor pair energy
formula given in section 3.2.2 using a typical depth of a donor (26 me V).
The first effective and stable p-doping and subsequent fabrication of a p-n
Tm=15K 00 I~
ZnSe:N
12
?:
.,
iii
c:
a,
E.
~
~
.E
...J
D.. I~
4400
~
4450
I~-LO
O2
4000
t 03
04
5000 6000 7000
Wavelength (A)
Fig. 3.7 Photoluminescence spectrum of a nitrogen-doped ZnSejGaAs film exhi-

biting bound-exciton lines and donor-acceptor pair lines [46].
junction laser with ZnSe has been performed with an N-doped MBE
structure [47].
The luminescence intensity provides additional information. An effective
doping can be read from a high ratio of the intensity owing to bound exciton
lines to the deep level centre luminescence. However, a reduced ratio might
be observed at high doping, indicating the formation of associates. On the
other hand, from a high intensity ratio of the free exciton lines versus the
bound exciton lines a high purity of the sample can be concluded.
Strong bound exciton lines indicating an effective doping have especially
been found in lattice-matched ZnSSe/GaAs layers. It should be noted that
this feature was observed with p doping [48] as well as with n doping [49].
In both cases the strong ratio of the bound exciton lines with respect to the
deep 'self-activated' luminescence observed in the optical spectra was
accompanied by low electrical resistivities.
3.4.2 Site-selectioD spectroscopy

This section refers to the optical spectroscopy of deep level impurity centres.
It will be shown how the occurrence of zero-phonon lines allows the
application of the powerful tool of site selection spectroscopy for the
discrimination of weakly luminescent centres and the study of energy
transfer processes.
In contrast to most centres emitting in the near band edge region deep
level luminescence usually originates from strongly localized centres. The
luminescence of these centres thus essentially contains information on the
electronic nature of the centre and on its local environment. Any impurity
centre with a known electronic structure can therefore act as a probe for the
crystalline environment given by its host material. This probing is quite
effective if zero-phonon lines are detected which are quite sharp in an
unperturbed crystal.
Crystalline imperfections lead to variations of the local potential of
impurity ions at different sites and result in an inhomogeneous broadening
of the zero-phonon lines. It has been proved that the broadening originates
from the superposition of numerous sharp zero-phonon lines with slightly
different transition energies. The effect is especially pronounced in glasses as
host material of impurities (see, for example, Ca(P03h glass:Eu3+ [50], but
the phenomenon can also be observed in crystalline semiconductors (see, for
example, the study of the excitation of the Rl emission in ruby AI 2 0 3:Cr3+
[51]).
The inhomogeneous broadening of zero-phonon lines can be detected in
luminescence and excitation spectra as well as in absorption spectra. The
selection of an optically active centre in a specific environment, however, is
not possible in absorption studies. The technique of site selection is based on
small band excitation and detection. By this means the registration is
confined to centres with specific excitation and emission energies. The
excitation usually is spectrally separated from the recorded emission. This
means the centres are excited into an electronic level which is above the
luminescent level. Site selection can be used if both the energy of the
excitation level and the energy of the radiating level are different, depending
on variations in the neighbourhood of the emitting centre.
In order to demonstrate the principle of site selection spectroscopy two
centres with slightly different energies of both their excitation and emission
energies are considered; see Fig. 3.8. A small band registration at an energy
coinciding with the emission energy of centre 1 and a scanned small band
excitation will essentially produce an excitation spectrum of this centre.
Centre 2 can be completely suppressed if its emission energy is lower than
the selected energy of centre 1 as assumed in Fig. 3.8. If the emission energy
of centre 2 is higher than the selected energy of centre 1 the discrimination is
less effective owing to the phonon-assisted (Stokes) emission of centre 2 in
the selected spectral range. The same considerations are valid for emission
spectra resulting from small band excitation. In this case centre 2 can be
completely suppressed if its excitation energy is higher than the selected
excitation energy of centre 1. Otherwise phonon-assisted excitation will
deteriorate the discrimination. This technique of selecting centres can lead to
an effective selection of very weak luminescence centres which under
nonselective conditions are superimposed by a strong luminescence of other
centres. As an example of effective site selection the intraionic transitions of
manganese Mn 2 + centres (3d 5) in a ZnS host crystal will be discussed in
more detail.
ell:citation
emission
_ _ _ _ _ _______ L..-_ _ _ _ _ _ _ _ _ _,.. I

centre 1 centre 2
Fig. 3.8 Principle of site selection applied on two centres with different energies of
their excitation and emission levels.
Mn is well known to create an efficient luminescence centre in many wide

gap compound semiconductors [16]. In ZnS low level doping with Mn leads
to five absorption bands which simultaneously are excitation bands of the
yellow Mn emission [52, 53]. Each band consists of sharp zero-phonon lines
accompanied by a broad range of phonon-assisted transitions. In the early
studies of the zero-phonon lines their assignment to spin-orbit split crystal
field states was complicated by the fact that several lines appeared with
varying relative intensities in different samples. The intensities of these lines
could be related to a hexagonality coefficient (X which denotes the deviation
of the crystal structure from the cubic zincblende structure [54]. These
structural imperfections are related to the polymorphic properties of ZnS
which can grow in different modifications. All modifications consist of
sequences of hexagonal Zn-S double layers being stacked along the direc-
tion of growth. There exist three equivalent layers A, Band C (mutually
rotated by 2n/3) which determine the crystal structure by the stacking
sequence. The cubic zincblende structure results from a regular ABC
sequence while the hexagonal wurtzite structure is formed by an AB
sequence; see Fig. 3.9 [55].
An Mn2+ ion substitutes isoelectronically a Zn2+ ion and is thus located
Zincblende (AN) Axial 2 (PN) Axial 1 (AS) Wurtzite (PS)

<111 )g <111 )g <111 )g <001)
i i i i
::I:tJT.L~:: ::~}.[~::
c ~
I
~
I
C B CI / CI / B
Fig. 3.9 Four different lattice sites of Mn2+ in polymorphic Zn8 differing in the
stacking layers above and below the impurity. Points and circles represent Zn2+ and
8 2 - ions, respectively, small circles denote the Mn2+ impurity ion [55].
in a tetrahedrally coordinated S2 - environment which belongs to two Zn-S

double layers. Different stacking orders of these layers which are placed
above and below the layer containing the Mn 2+ can therefore build four
inequivalent lattice sites of the impurity. The different crystal fields of these
sites evoke different energies of the electronic levels of the unfilled d shell of
Mn 2+. As a result different transition energies of the zero-phonon lines of
Mn 2+ ions at different lattice sites are observed. The nonselective excitation
spectrum in the zero-phonon range of the excited 4E(G) level of the Mn 2+
d S electrons (Fig. 3.10, spectrum a) exhibits three triplets [56]. The spectrum
has been recorded in the maximum of the phonon-assisted emission and
resembles the absorption spectrum [52]. By small band detection within the
three corresponding emission zero-phonon lines only the excitation spec-
trum of Mn2+ at a specific lattice site is recorded (see Fig. 3.10, spectra b-d).
The remaining threefold structure of the zero-phonon lines is related to
spin-orbit splitting [57]. The sample examined in this work had a pre-
dominantly cubic structure but contained many stacking faults which created
the two axial environments shown in Fig. 3.9. As the sample hardly
contained hexagonal sites, the fourth spectrum referring to this site could
not be recorded with this sample.
In more recent investigations also the effect of the sequence of a further
stacking layer above and below the range previously considered has been
demonstrated [58]. The extension of the considered lattice environment by
two additional layers leads to a fourfold distinction of each of the four
'main' lattice sites shown in Fig. 3.9. The effect of differences in the
second-nearest layer with respect to the impurity is quite small. The
zero-phonon lines of the impurities whose environments differ only in these
layers are thus only slightly shifted with respect to each other. In a contour
line plot of combined emission and excitation spectra the impurities at dif-
ferent sites appear as different groups of coherent zero-phonon lines appear-
ing in the vicinity of each of the 'main' sites; see Fig. 3.11. Horizontal
a total
(")
(")
N
N
en
T N
'So N co
(")
N
N
co
ot
N
N
ot
ot
N
N (")
Ltl
N
C axial 1 N
d axial :2
21200 21220 21240 21260 21280 21300 cm-' 21340

excitation energy (cm- 1 )
Fig. 3.10 Site-selected excitation spectra recorded with a single sample of polymor-
phic ZnS:Mn at different emission energies: (a) non-selective at the maximum of the
phonon-assisted Mn2+ emission band, (b--d) in the zero-phonon emission line due to
Mn2+ ions in cubic (b) and two different axial distorted sites (c and d) [56].
cuts of this diagram correspond to excitation spectra recorded at a fixed

emission energy, and vertical cuts to emission spectra of a fixed excita-
tion energy. 12 of the 16 possible impurity sites could be distinguished by
site selection and an assignment of the spectra to well-defined crystallo-
graphic sites could be given. In addition, a general superposition model
was deduced from the energy shift pattern which has been found to be
valid also for other transition metal impurities in ZnS such as Fe, Co or
Ni.
rec.:18006 cm-'
.E
11'1
0
~ co
>.
CI
"-
III
C
III
C
0
'iii
UI 11'1
'f 0
-
0
III co
21245 21250 21255

excitation energy [cm-')
Fig. 3.11 Contour-line diagram of isointensities of combined emission and excita-
tion spectra of axial 1 centers in stacking-faulted ZnS:Mn, Maxima of coherent
groups of zero-phonon lines are connected by dashed lines, The separate excitation
spectrum shown at the top is recorded at a fixed emission energy [58].
Another kind of lattice imperfection which is also observed in single

crystals is remarkable at medium levels of impurity doping. As the ionic
radii of the substitutional impurities and the chemical bonding differ from
that of the substituted host ion, the doping introduces local strains in the
vicinity of any impurity which can influence a neighbouring impurity. This is
the origin of inhomogeneous line broadening (for a study of ZnS:Mn see ref.
59). At a higher doping level the discrete nature of lattice sites and thus of
the mutual interimpurity distances becomes evident. The impurities start to
interact with each other, forming small impurity clusters such as pairs or
triples. The concentration required for impurity clustering depends on the
spatial range of interaction.
In cubic-structured ZnS crystals with a medium Mn doping level
(XMn::::: 10- 3-10- 2) a number of zero-phonon lines in both emission and
excitation spectra appear in addition to the transitions of 'isolated' Mn 2 +
ions observed at low doping levels [60]. The energy of the additional
emission lines appears in the form of a series, the limit of which is represented
by the single Mn2+ emission zero-phonon line. A convincing illustration of
site selection can be seen in the excitation spectra recorded in these emission
lines labelled Pj, P j and P k in Fig. 3.12. Because of spin-orbit coupling two
ZnS:Mn
6 A ,(S)-+4 T,(G)
Rec.: P,
x3
Rec.: Pk
::J
x4
'"
~
'iii
c:
Ql
C
Rec.: P,
Rec.: Sem
19660 19700
Wavenumber (cm ')
Fig. 3.12 Zero-phonon excitation spectra recorded at the 'single' Mn2+ emission
line (Sem) and at the additional emission lines Pj' Pk and Pl' Arrows denote positions
of excitation lines connected to the selected emission line [60].
sharp lines and one broad line are expected in the Mn 2 + zero-phonon
excitation of the 4T 2(G} level. Despite the strong superposition of the lines
from the dominating Mn 2 + 'singles', selective registration allows the
detection of the much less abundant centres. The relative intensities of their
emission lines with respect to the single Mn 2 + emission increase with Mn
doping up to a limit of about xMn~3 X 10- 3 This finding is in accordance
with calculations of the probabilities of Mn 2 + pairs with different mutual
interimpurity distances. For a given interaction range the probability for
each kind of small cluster can be expressed by a polynomial whose powers
and coefficients have been calculated for various cubic-structured lattices
[19]. The application to substitutional Mn 2 + ions on a Zn f.c.c. sublattice of
a zincblende crystal with an interaction reaching the seventh-nearest neigh-
bour confirmed a maximum in the doping-dependent probability below
X~ 10- 2 followed by a decrease at stronger doping (this effect is related to
the formation of larger clusters).
From the series behaviour and the cluster statistics an assignment of the
Mn2+ pair clusters emitting the lines Pi (not shown in Fig. 3.12), Pi' P k and
PI to pair distances ranging from the third- to the sixth-nearest lattice site
has been inferred. At smaller distances the fine structure changes drastically
owing to the formation of the d lo electron system reported earlier [61].
3.4.3 Time-resolved spectroscopy

Differences in the dynamics of the luminescent relaxation of centres provide
a method of discrimination which can be used in addition to the spectral site
selection described in the previous section. Both techniques complement one
another, because the time-resolved spectroscopy can even distinguish centres
which coincide in both emission and excitation energy.
For purposes of illustration the Mn 2 + pairs in ZnS with nearest interim-
purity distances are considered. As mentioned in section 3.4.2 the first and
second nearest-neighbour pairs have to be regarded rather as d lo electron
systems than as mutually perturbed d 5 electron systems. In contrast to the
internal relaxation within the d 5 configuration the d lo energy level scheme
enables spin-allowed transitions with short radiative lifetimes. At an in-
creased doping rate of XMn ~ 0.01 in Zn I _ x MnxS the transient of the
phonon-assisted luminescence of Mn 2 + is found to be no longer purely
exponential. In a short time range after laser pulse excitation a fast decay
component appears [61]. The zero-phonon range of the luminescene is
shown in Fig. 3.13. Spectra 1 and 2 differ only in the setting of the
registration time gate after pulse excitation which has been set to 0.2-5 ms
and 0-0.2 ms respectively. Spectrum 1 with the zero-phonon line Sem thus
represents the 'slow' component of the luminescence whose lifetime corres-
ponds to that of the single Mn2+ ion emission. This assignment is confirmed
by the spectral characteristics which - apart from a concentration-induced
Wavelength
+---
568 564 560 nm
TO
ZnS:Mn
T=5K
-
17600 17700 17800 cm- 1
Energy
Fig. 3.13 Gated emission spectra of ZnO.99Mno.OlS, Spectra 1 and 2 differ only by
the time gate set to 0.2-5.0 ms and 0-0.2 ms, respectively. Spectra 2 and 3 appear at
different excitation energies [61].
inhomogeneous broadening - resemble the luminescence observed at very

low Mn doping. The 'fast' spectrum 2, however, is dominated by another
zero-phonon line labelled Pa and shows an additional structure labelled Ph'
The Ph emission which also appears only in the fast spectrum is connected
to other excitation resonances than P a and can thus be amplified by
site-selection; see spectrum 3. Because of the concentration dependence of
the Ph and P a lines an assignment to the first and second neighbour pairs of
Mn 2 + ions respectively has been given which has been confirmed by the
study of Mn pairs with larger mutual distances as mentioned in section 3.4.2.
The nonexponential transient of the Mn luminescence as mentioned
above should not be confused with the similar feature of the transient of an
energy donor connected to other centres by an energy transfer. The
nonexponential behaviour of the luminescence of the Mn just occurs by
a superposition of different luminescence centres with different lifetimes
radiating in the same energy range. The nonexponential transient of an
energy donor, however, appears as an inherent feature of the radiative decay
of one centre and cannot be decomposed by spectral or temporal means.
3.4.4 Energy transfer

The general ideas of energy transfer processes have been described in section
3.2.4 and were also discussed in Chapter 1 and Chapter 2. We therefore confine
ourselves to a brief description of energy transfer phenomena observed with
the Mn2+ centre in ZnS which has already been introduced above.
Whenever the Mn content in large band gap II-VI compounds exceeds
1% of the substituted cations, usually three emission bands occur in
addition to the well-known luminescence due to the tetrahedrally coor-
dinated Mn2+ single-ion centre [62]. The origin of these additional bands is
not yet settled in a definite way, but it is quite likely that clustering of Mn
impurities and local phase transitions of the zincblende to the rock salt
structure play an important role. In heavily doped Znl-xMnxS four
luminescence emission bands peaking in the yellow, orange, red and near
infrared spectral range are observed. The excitation spectrum of the yellow
emission band shows the structures characteristic for the single Mn 2 +
centre. The additional emission bands can also be excited by these bands but
exhibit further special excitation structures. This behaviour is a hint of an
energy transfer from the single Mn2+ centre acting as a (energy) donor to
the centres occurring at high Mn doping levels acting as acceptors. This is
proved by time-resolved emission spectra in the microsecond timescale
after a nanosecond laser pulse excitation of the Mn2+ single-ion centre; see
Fig 3.14 [63]. Shortly after the end of the exciting laser pulse (spectrum
~ IllS) only the directly excited system emits light. It decays by the direct
channel via the yellow (donor) luminescence and by energy transfer to the
competing centres emitting in the orange, red and infrared range. Conse-
quently the intensity of the low energy emission bands initially increases
during the decay of the yellow emission band of Mn2+ and is most
pronounced with time gates between 1 IlS and 3.3 IlS.
The energy transfer process can be studied in more detail by an analysis of
the emission transients [62]. While the yellow emission shows a nonex-
ponential donor-like decay without initial increase, all additional emission
bands exhibit a pronounced temperature dependent initial increase of the
luminescence intensity after the donor laser pulse excitation. The transients
of the emission bands can be fitted by nonexponential decay curves
according to Forster's law. The analysis of the decay curves leads to an
energy transfer rate which increases with increasing temperature and to a
small energy migration due to an energy diffusion among the donor centres.
This reflects the large number of different energy acceptors fed by only one
single energy donor system.
T=100 K
x=0.41
80.00 v.x =21650cm- 1
(i)
.~
c:
::J 60.00
.ci
-
~
>
40.00
-
iii
c:
Q)
c:
20.00
10000 12000 14000 16000 18000

Wavenumber (cm-')
Fig. 3.14 Time resolved emission spectra of Znl-xMnxS (x=O.4) at T= 100 K for
different time gates (ttft2). Excitation occurs into the 4E(G} level of the single Mn2+
ion. The curves are normalized.
3.4.5 Zeeman and pressure measurements

Introduction of a luminescent sample into an external magnetic or electric
field enables a safe assignment of the electronic nature of the studied
transition. In particular, the application of a uniaxial field reduces the site
symmetry of the impurity and leads to a splitting of zero-phonon transitions.
This lifting of the degeneracy can be deduced theoretically from symmetry
considerations by group theory and can provide unambiguous assignments
of the involved electronic levels by a comparison with the experimental
findings.
The d 5 electron systems Mn2+ or Fe3+ provide a good illustration
subject for the effect of an externally applied field. The ground state of this
system is realized according to Hund's rule by a parallel orientation of all
spins in the half-filled d shell yielding a 6S state of the free ion. In a
crystalline tetrahedral environment (Td) the spin remains a good quantum
number and the ground state level is given by 6A 1 (S). The introduction of a
magnetic field with its axis parallel to a tetrahedral axis reduces the site
symmetry of the impurity from tetrahedral to trigonal (C 3v) in which all six
Ms states have different energies. As a consequence the zero-phonon line of
the d 5 impurity exhibits a sixfold splitting in the Zeeman spectra. This
behaviour can be regarded as a fingerprint of d 5 centres. In Fig. 3.15 the
Zeeman splitting of the Mn2+ emission (Sem) in ZnS at a magnetic field of
ZnS:Mn
BII [111] B= 5T
17840 17850 17860

Wavenumber (cm-')
Fig. 3.15 Zeeman emission spectra of the single Mn 2 + emission Sem and the
selectively excited lines P j and L 5. Energy scale refers to the L5 spectrum, the spectra
of P j and Sem are shifted by - 23 em -1 and -40 cm -1, respectively. Marked
postions denote transitions to Ms states of the 6A 1(S) ground state [64].
5 T is shown [64]. The spectra have been recorded at low temperatures

(1.8 K) to avoid thermalized emissions originating from the field splitting of
the radiating excited level. Five of the six lines which are labelled according
their Ms final state can be recognized. The spectrum above Sem has been
recorded under selective excitation of a resonance leading to the P j emission
line described in section 3.4.2. The emission spectrum has been drawn with a
shift of - 23 em - 1 for the sake of comparison. For the same reason the
upper spectrum which refers to a selected weak emission line labelled Ls has
been drawn shifted by -40 cm -1. The P j emission clearly exhibits the same
splitting behaviour as the Sem emission. Comparable features can be found
in the Ls splitting. From these experiments a firm assignment of the three
centres to Mn2+ in different local environments could be given. For P j the
origin of the perturbation is a neighbouring Mn 2 + ion while for Ls a
complex centre (Mnzn-V zn ) has been proposed.
In the near infrared emission of iron-doped ZnS the characteristic d S
splitting helped to identify the Fe3+ emission; see Fig. 3.16 [65]. The six
components are labelled AN,-AN v,. Because of stacking faults within the
samples (see Section 3.4.2, Fig. 3.9) additional lines have to be distinguished
(e.g. AN1, in the figure).
The selection rules provide additional information necessary for a safe
assignment of the Zeeman levels. The excitation spectrum of the Mn 2 +
emission into the 4E(G) level exhibits three zero-phonon lines (see section
3.4.2). Two of them are split under the influence of a magnetic field parallel
Wavelength (nm)
1225 1223 1221
ANv, AN v AN III ANII
I I 1570 lJev l
ZnS
B=5T
B11[111]
~ T=1.8K
.iii
...c:c:
Q)
1.012 1.014
Energy (eV)
Fig. 3.16 Zeeman emission spectrum of Fe3+ in ZnS. Transitions into different Ms
states are labelled AN(-AN v( [65].
to a tetrahedral axis into two levels and one into four levels. A configuration
of the detection system parallel to the magnetic field (Faraday configuration)
leads to polarization selection rules with (j +, (j - and 'It polarized lines from
which the symmetry of the initial and final state can be deduced [66]. The
results of these experiments are plotted in Fig. 3.17. The great number of
transitions in the magnetic field originates from excitation lines starting
from various thermally occupied levels of the split 6 Al (S) ground state. The
numbers of the notation of the Zeeman levels in Fig. 3.17 refer to these Ms
levels, the lowest-lying level being denoted by O. Lower-case letters denote
excited states. As transitions starting from all six ground state levels have
been found the application of the selection rules enabled the assignment of
not only the Zeeman levels of the excited 4E(G) state but also those of the
6 A 1 (S) ground state. The experimental assignments have been confirmed by
theoretical perturbation calculations [67].
The action of an external electric field is introduced by the application of
either hydrostatic or uniaxial pressure on the studied sample; see section
3.3.3. The application of uniaxial stress involves a change of the site
symmetry of a studied impurity. The reduction of the symmetry differs,
however, from the magnetic case because the electric field does not affect
directly the spin of the studied centre. Both hydrostatic and uniaxial
pressure experiments can thus be used to complement the results of Zeeman
measurements.
00
21255
21250 eO
dO
21245 e11
'i
E
.e rS
> 21240
Ol
Q)
C o d2
ra
Q)
o 0 91
21235
~~ ____+-------h1
r1 o
h2
92
09 2
o h2
21230 o 93
o h3
94
Icp
21225
o h5 rcp
linear
21220+0----~1----~2-----3r----4~---+5~
magnetic field B (T)

Fig. 3.17 Zeeman splitting of the excited 4E(G) level of Mn 2 + in ZnS deduced from
excitation spectra. Different line types denote different polarizations (left circular,
right circular and linear) of zero-phonon lines [66].
As an example we refer to the Mn 2 + centre in ZnS with its transition to

the excited 4E(G) level as introduced above. From crystal field calculations
two degenerated levels 4E(G) and 4A 1(G) are expected in the corresponding
energy region which should result in four zero-phonon transitions. The
experimentally observed structure appears to be composed of only three
lines; see Fig. 3.10. Symmetry considerations show that these lines can be
associated with the three spin-orbit states r 6, r 7 and r 8 of the 4E(G) level.
The validity of this attribution could be proved by applying pressure parallel

to the [110J axis of the ZnS:Mn crystal [57]. The authors obtained evidence
that the degeneracy of one line (the transition to the r 8 level) can be lifted
into two levels by the pressure-induced rhombic distortion and excluded any
assignment to the 4A 1(G) level. This first assignment was supported by more
recent calculations of the oscillator strengths of transitions to 4A 1 (G), these
being three orders of magnitude smaller than transitions to 4E(G) [68]. The
final confirmation has now been given by the Zeeman studies described
above.
REFERENCES
1. Cho, K. (ed.) (1979) Excitons, Topics in Current Physics, Vol. 14, Springer, New
York.
2. Frenkel, 1.1. (1936) Phys. Z. Sowjet Union, 9, 158.
3. Wannier, G.H. (1937) Phys. Rev., 52, 191.
4. Lampert, M.A. (1958) Phys. Rev. Lett., 1, 450.
5. Dean, PJ. and Herbert, D.C. (1979) In Excitons (ed. K. Cho), Topics in Current
Physics, Vol. 14, Springer, New York, p. 55.
6. Haynes, 1.R. (1960) Phys. Rev. Lett., 4, 361.
7. Halsted, R.E. (1967) In Physics and Chemistry of II-VI Compounds (eds M. Aven
and 1.S. Prener) North-Holland, Amsterdam.
8. Haug, H. (ed.) (1988) Optical Nonlinearities and Instabilities in Semiconductors,
Academic Press, New York.
9. Klingshirn, C. (1990) Semicond. Sci. Technol., 5, 457.
10. Ungier, W. Suffczynski, M. and Adamowski, 1. (1981) Phys. Rev. B, 24, 2109.
11. Dean, P.l. (1973) In Progress in Solid State Chemistry (eds lO. McCalvin and G.
Somorjai), Vol. 8, Pergamon, New York, p. 1.
12. Williams, F. (1968) Phys. Status Solidi 25, 493.
13. Thomas, D.G., Gershenzon, M. and Trumbore, F.A. (1964) Phys. Rev., 133,
A269.
14. Monemar, B. (1990) In Impurities, Defects and Diffusion in Semiconductors: Bulk
and Layered Structures (eds D.l. Wolford, 1. BernhoIc and E.E. Haller), MRS
Symposium Proceedings, Vol. 163, Materials Research Society, Pittsburgh, PA.
15. Imbusch, G.F. and Kopelman, R. (1981) In Laser Spectroscopy of Solids (eds
W.M. Yen and P.M. Selzer), Topics in Applied Physics, Vol. 49, Springer, New
York.
16. Curie, D. (1963) Luminescence in Crystals, Wiley, New York.
17. Sugano, S., Tanabe, Y. and Kamimura, H. (1970) Multiplets of Metal Ions in
Solids, Academic Press, New York.
18. Griffith, 1.S. (1971) The Theory of Transition Metal Ions, Cambridge University
Press, Cambridge.
19. Pohl, U.W. and Busse, W. (1989) J. Chern. Phys., 90, 6877.
20. Curie, D. and Prener, 1.S. (1967) In Physics and Chemistry of II-VI Compounds
(eds M. Aven and lS. Prener), North-Holland, Amsterdam.
21. Forster, T. (1949) Z. Naturforsch. A, 4, 321.
22. Dexter, D.L. (1953) J. Chern. Phys., 21, 836.
23. Hamilton, T.D.S., Munro, I.H. and Walker, G. (1978) In Luminescence Spectro-
scopy (ed. M.D. Lumb), Academic Press, New York.
References 95
24. Henderson, B. and Imbusch, G.F. (1989) Optical Spectroscopy of Inorganic
Solids, Oxford University Press, New York.
25. Shimoda, K. (1984) Introduction to Laser Physics, Springer Series in Optical
Sciences, Vol. 44, Springer, New York.
26. Mollenauer, L.F. and White, J.e. (1987) Tunable Lasers, Topics in Applied
Physics, Vol. 59, Springer, New York.
27. Phillips, J.e. (1971) Phys. Rev. Lett., 27, 1197.
28. Martinez, G. (1980) In Handbook of Semiconductors (ed. M. Balkanski), Vol. 2,
North-Holland, Amsterdam, p. 184.
29. Piermarini, G.J., Block, S. and Barnett, J.D. (1973) J. Appl. Phys., 44, 5377.
30. Demtroder, W. (1977) Grundlagen und Techniken der Laserspektroskopie, Spring-
er, Berlin.
31. Keyes, R.J. (ed.) (1977) Optical and Infrared Detectors, Topics in Applied Physics,
Vol. 19, Springer, New York.
32. Chang, R.K., Long, M.B., Cardona, M. and Giintherodt, G. (1982) Light
Scattering in Solids II, Topics in Applied Physics, Vol. 50, Springer, New
York.
33. Levine, B.F. and Bethea, e.G. (1984) Appl. Phys. Lett., 44, 558, 581, 649.
34. Gutowski, J., Kudlek, G. and Presser, N. (1990) Phys. Status Solidi A, 120, 11.
35. Kleiman, J., Park, R.M. and Quadri, S.B. (1987) J. Appl. Phys., 61, 2067.
36. Kudlek, G., Presser, N., Pohl, U.W., Gutowski, 1., Lilja, J., Kuusisto, E., Imai, K.,
Pessa, M., Hingerl, K. and Sitter, H. (1992) J. Cryst. Growth, 117, 309.
37. Kudlek, G., Presser, N. and Gutowski, J. (1991) Semicond. Sci. Technol., 6, A83.
38. Kudlek, G. and Gutowski, 1. (1992) J. Lumin., 52, 55.
39. Gumlich, H.-E. (1988) In Growth and Optical Properties of Wide-Gap II-VI
Low-Dimensional Semiconductors (eds T.e. McGill, e.M. Sotomayor-Torres and
W. Gebhardt), NATO ASI Series, Vol. 200, Plenum, New York.
40. Mandel, G. (1964) Phys. Rev., 134, A1073.
41. Mandel, G., Morehead, F.F. and Wagner, P.R. (1964) Phys. Rev., 136, A826.
42. Neumark, G.F. (1989) Phys. Rev. Lett., 62, 1800.
43. Laks, D.B., Van de Walle, e.G., Neumark, G.F. and Pantelides, S.T. (1991) Phys.
Rev. Lett., 66, 648.
44. Goetz, G., Po hi, U.W. and Schulz, H.-1. (1992) J. Phys. C, 4, 8253.
45. Kukimoto, H. (1988) In Growth and Optical Properties of Wide-Gap II-VI
Low-Dimensional Semiconductors (eds T.e. McGill, e.M. Sotomayor-Torres and
W. Gebhardt), NATO ASI Series, Vol. 200, Plenum, New York.
46. Yoshikawa, A., Muto, S., Yamaga, S. and Kasai, H. (1988) Jpn. J. Appl. Phys., 27,
992.
47. DePuydt, J.M., Haase, M.A., Qui, J. and Cheng, H. (1992) J. Cryst. Growth, 117,
1078.
48. Suemune, I., Yamada, K., Masato, H., Kanda, T., Kan, Y. and Yamanishi, M.
(1988) Jpn. J. Appl. Phys., 27, L2195.
49. Fujita, S., Terada, K., Sakamoto, S. and Fujita, S. (1989) J. Cryst. Growth, 94,102.
50. Motegi, N. and Shionoya, S. (1973) J. Lumin., 8, 1.
51. Szabo, A. (1970) Phys. Rev. Lett., 25, 924.
52. Langer, D. and Ibuki, S. (1965) Phys. Rev., 138, A809.
53. Neumann, E. (1971) Thesis D83, Technical University, Berlin.
54. Lambert, B., Buch, T. and Geoffroy, A. (1973) Phys. Rev. B, 8, 863.
55. Pohl, U.W., Gumlich, H.-E. and Busse, W. (1984) Phys. Status Solidi B, 125, 773.
56. Busse, W., Gumlich, H.-E., Geoffroy, A. and Parrot, R. (1979) Phys. Status Solidi
B,93,591.
57. Parrot, R. and Blanchard C. (1972) Phys. Rev. B, 6, 3992.
58. Pohl, U.W., Ostermeier, A., Busse, W. and Gumlich, H.-E. (1990) Phys. Rev. B,
42,5751.
59. Busse, W., Gumlich, H.-E., Knaak, W. and Schulze, J. (1980) J. Phys. Soc. Jpn.,
49 (Suppl. A), 581.
60. Pohl, V.W. and Gumlich, H.-E. (1989) Phys. Rev. B, 40, 1194.
61. Benecke, C., Busse, W., Gumlich, H.-E. and Pohl, V.W. (1985) Phys. Status Solidi
B, 128, 701.
62. Benecke, C. and Gumlich, H.-E. (1992) In The Luminesence of Wide Band Gap
JI-Mn-VI Semimagnetic Semiconductors in Diluted Magnetic Systems (ed. M.
Jain), World Scientific, London.
63. Benecke, C., Busse, W., Gumlich, H.-E. and Moros, H.-1. (1987) Phys. Status
Solidi B, 142, 301.
64. Pohl, V.W. and Gumlich, H.-E. (1990) J. Cryst. Growth, 101, 512.
65. Hoffman, A., Heitz, R. and Broser, I. (1990) Phys. Rev. B, 41, 5806.
66. Litzenburger, B., Mahrt, 1., Pohl, U.W. and Gumlich, H.-E. (1993) To be
published.
67. Parrot, R., Diarra, M. and Boulanger, D. (1991) Personal communication.
68. Parrot, R., Naud, C., Gendron, F., Park, C. and Lemercier, M. (1986) J. Chem.
Phys., 85, 4932.
4
One-photon rare earth optical
transitions: recent theoretical
developments
G.W. Burdick and M.e. Downer
4.1 GENERAL THEORY OF ONE-PHOTON ABSORPTION

'The puzzle of rare earth spectra in solids', a term first coined by van Vleck
[1], was solved 30 years ago by Judd [2] and Ofelt [3]. The rich variety of
sharp optical absorption and emission lines characteristic of 4fN -+4fN
lanthanide transitions are parity forbidden in the free ions. However, in the
quantitative Judd-Ofelt theory, observation of these transitions is explained
through simple configuration mixing caused by a noncentrosymmetric
crystal field, allowing opposite-parity excited configurations to become
slightly mixed into 4fN.
Significant additions to this theory have been required in order to account
for anomalous transitions which acquire a major fraction of their intensity
from other mechanisms. A prominent example is the 'dynamic coupling'
contribution [4], where the hypersensitivities of some transitions to their
environment are explained through the coupling of the 4f electrons to
transient dipoles induced in the ligands by the radiation field, leading to
amplification of normally negligible electric quadrupole transition ampli-
tudes. The Judd-Ofelt theory, however, remains the standard method for
analysing observed rare earth transition intensities, and successfully ac-
counts for the intensities of up to 20 or 30 transitions per rare earth ion with
an agreement usually better than 15% of observed oscillator strengths.
Why then is it important to re-examine this 'solution?' The most signi-
ficant motivation has come from recent results in two-photon rare earth
absorption spectroscopy [5], where the standard Judd-Ofelt theory, as
applied to two-photon spectroscopy by Axe [6], could not explain experi-
mental observations for Gd 3 + transitions [7]. In order to explain these
97
98 One-photon rare earth optical transitions
experiments, higher-order perturbative terms involving spin-orbit interac-

tions [8] and static crystal field interactions [9-11] were found to be
necessary. The similarity between the Axe theory of two-photon spectros-
copy and the Judd-Ofelt theory of one-photon spectroscopy implies that
analogous terms, particularly those involving spin-orbit interactions, are
also important in one-photon absorption.
4.1.1 The Judd-Ofelt theory

As described in Chapter 1 of this book [12], the general theory oflanthanide
transitions can be described by using standard perturbation theory. In the
electric dipole approximation, second-order time-dependent perturbation of
the interaction Hamiltonian directly yields the Goppert-Mayer [13] for-
mula describing the probability of direct two-photon transitions,
1
_ L [(i IE 1 Dln)(nIE2 ' Dlf) + (iIE2 ' Dln)(nIE1 'Dlf)] 12 (4.1)
n h(w n - wd h(w n - (2)
where E j ' D represent electric dipole operators, h(w n -Wj) denote the
energies of intermediate states n above the single photon energy hWj, and
the sum is taken over all intermediate states n of opposite parity to the
ground configuration 4fN.
In the one-photon theory of Judd [2] and Ofelt [3], an expression
analogous to formula (4.1) is used, called the 'forced' electric dipole, where
the second-order perturbation of the interaction Hamiltonian is replaced by
a time-independent perturbation involving the noncentrosymmetric part of
the crystal field, H CF ' That is, admixtures of opposite parity configurations
into the initial and final state wavefunctions allow the odd-parity electric
dipole to connect two states of the same parity. Thus, we obtain the
corresponding one-photon transition probability,
1
_ L [(iIHcFln)(nIE' Dlf) + (iIE' Dln)(nIHcFlf)] 12 (4.2)
n hWni hWn!
where hWni and hWn! denote the energies of intermediate states n above
initial state i and final state f respectively.
Actual computation of (4.2) for one-photon and (4.1) for two-photon
transitions is impractical because of the large number of intermediate levels
which must be summed over. For example, the lowest opposite-parity
excited configuration of Tb3+ (4f 7 5d) has 3006 SLJ levels [14]. In order to
get around this problem, the 'closure approximation' was independently
introduced by Judd [2] and Ofelt [3], in which each excited configuration
is regarded as being completely degenerate. This closure approximation is
reasonable for trivalent lanthanide ions owing to the large energy gap
between excited configuration levels and the ground and final state 4fN levels.
General theory of one-photon absorption 99
For example, in Pr3+, observed 4f2 levels lie within 23000 cm -1 of the
ground state [15] while energy levels of the lowest excited configuration,
4f5d, range from 61000cm- 1 to about nOOOcm- 1 [14,16]. Thus, the
energy denominator h(Wni -w) in formula (4.1) for two-photon excitation
ranges from 50000 to 70000 cm -1 . For one-photon excitation, the calcula-
tion is somewhat more complicated, since there are two different energy
denominators hWni and hWnJ. However, the closure approximation assumes
that the matrix elements in the two numerators of formula (4.2) are identical,
to within a phase factor, and we can replace the two denominators by a
single denominator using the identity
-1+ - =
1 2( 1 1 -w-2 -)
) ( 1+ (4.3)
hWni hWnJ hWni - hw/2 4 WniWnJ
where the resulting energy denominator (first parenthesized term on the

right-hand side) is numerically identical to that specified for two-photon
excitation above.
Although the closure approximation introduces some inaccuracies into
the calculations, it does greatly simplify the mathematics, making the
problem tractable. The dipole operator, E D, and the crystal field oper-
ator, H CF , can now be united into a single operator acting solely between
levels of fN, dependent on the excited configuration n'l'. Details of the
necessary manipulations are given in the papers of Judd [2] and Ofelt [3],
where operators are expressed in spherical tensor form and application is
made of the Wigner-Eckart theorem and various sum rules and operator
techniques.
The correction factor of (1/4)w2/wniwnJ in equation (4.3) is typically
quite small, and is thus customarily neglected. In practice, a further
simplification is made by assuming that all energy levels within the ground
fN configuration are also degenerate, allowing the use of a single energy
denominator for every fN ~fN transition calculated. Although these further
approximations are not absolutely necessary in order to calculate transi-
tion line strengths, they do simplify the calculation. In section 4.4 we will
examine the significance of removing these two assumptions from the
theory.
Performing the necessary manipulations on expression (4.2), we obtain
the following result for single-photon transition intensities:
SOPA=I L L VAk).(iI(E(1)B(k))().).V().)1f)12 (4.4)

kodd ),=2.4,6
where v is the energy hw of the transition. The coefficients A k ). contain radial

integrals (4flrln'l') and (4fl,-i'ln'l'), the average energy denominator hWI'J
and other numerical factors. The V().) are sums of single-particle unit tensor
operators :u~).) with the property (f II u ().) II f) = 1. Also, the B(k) are kth -rank
tensors representing coefficients of the spherical harmonics in the expansion

of the odd-rank terms of the crystal field operator,
(4.5)
n= 1 kodd q=-k
which are coupled to the electric field by

(E(1) B(k)g.) = L (lpkq 11kAQ)E?) BJk)
P.q
Since the M J components are generally mixed by the crystal field and
often not resolved, it is customary to define the total probability, or line
strength, of a transition as the summed intensity over all components M J
and Mj of the ground and final states. The line strength is anisotropic
owing to the coupling of the electric field to the crystal field. Since the
magnitudes of the coefficients BJk) for k odd are not known with any
degree of accuracy, this polarization dependence is difficult to predict.
Therefore, in order to simplify the comparison of theory and experiment,
this complicated line strength anisotropy is averaged out by using un-
polarized light or samples with randomly oriented rare earth ions, such as
an aqueous solution or powder. Thus, we sum over all polarizations q, and
equation (4.4) reduces to
p= L vT~(iIIU())lIf)2 (4.6)
),=2.4.6
which is equation (16) of ref. 2. In order to facilitate comparison of the T).

parameters across the lanthanide series, a common factor of (2J + 1)-1 can
be factored out of the T~, yielding
p = _v_ L T).(i II U().) II f)2 (4.7)

2J+l ).=2.4.6
This expression is the major result of the Judd-Ofelt theory, and is the usual
form used for analysis of data.
Reduced matrix elements U().) have been tabulated between all levels
within the fN configuration for LS coupling by Nielson and Koster [17], and
between the ground state and low-lying excited fN states for intermediate
coupling by Carnall et al. [18]. The T). however, are not determined from
first principles because of the difficulty of computing odd-parity crystal field
coefficients and radial integrals. Instead, these 'Judd-Ofelt parameters' are
left as phenomenological parameters, determined by fitting experimental
oscillator strengths. In so far as the closure approximation and other
simplifications made by the Judd-Ofelt theory are valid, the T). are constant
for all transitions of a particular rare earth ion in a given host crystal or
solution, allowing all observed transitions to be fitted by a single set of
parameters.
This fitting procedure has been successfully used to explain observed

oscillator strengths for numerous transitions throughout the trivalent lan-
thanide series. The most complete work is that of Carnall et al. [15, 19] for
all trivalent lanthanide ions in aqueous solution. Peacock [4] has sum-
marized many other analyses of lanthanide oscillator strengths in numerous
crystals and complexes. In most of these cases, this phenomenological
treatment successfully fits these transitions with a fitting error of better
than 15%.
4.1.2 Two-photon absorption

Because of the successes of the second-order Judd-Ofelt theory in fitting
lanthanide transitions, the possible role of terms arising in the third and
higher orders were not considered. It has only been since 1980, with the
examination of experimental two-photon absorption, that third- and higher-
order terms in the perturbation series have been considered important. In
order to understand this, it is important to consider how the Axe theory for
two-photon absorption differs from the standard one-photon Judd-Ofelt
theory.
The method employed by Judd and Ofelt can be applied to the electric
dipole approximation for two-photon absorption in an analogous manner
to the application for the one-photon forced electric dipole. Since a second
electric dipole operator replaces the noncentrosymmetric part of the crystal
field, there is no explicit reference to the environment of the ion by the
operators in the second-order description of two-photon absorption. This
eliminates the need for a sum over k. Summing over all components MJ, M J
of the initial and final states, we have the two-photon absorption line
strength, analogous to equation (4.6):
(4.8)
Since only a single term is necessary to describe absorption of two
identical photons, the intensity ratio of two transitions can be found as the
ratio of the squares of corresponding reduced matrix elements, which rely
only on the angular parts of the wave function and are known to great
precision. Thus, equation (4.8) gives a rigorous theoretical prediction of
relative two-photon absorption line strengths with no phenomenological
parameters. No sum over polarizations is necessary either. Thus, experimen-
tal measurements of line strength ratios and polarization dependences
provide a much stronger test of the Axe theory for two-photon absorption
than in the Judd-Ofelt theory for single photon absorption.
The first major test of this theory in two-photon absorption came with the
examination of Gd H by Dagenais et al. [7]. In this analysis, it was found
that the second-order theory could not explain observed polarization
dependences and relative intensities of the 8S 7 / 2 -+6PJ (J =7/2, 5/2, 3/2)
transItIons in Gd3+. This failure of the second-order theory motivated

examination of higher-order contributions in the perturbation series. Writ-
ing out the higher-order contributions explicitly, formula (4.1) is replaced by
_ L (iIE' Dln)(nIE' Din

1
n h(Wni- W )
_L (iIE' ~1l)(lIHllm)(mIH2In)(nIE' Dlf) + ... 2

1 (4.9)
l,m.n h (WU-W)(Wmi-W)(Wni-W)
where the first sum represents the standard second-order term and the
second and third sums represent new third- and fourth-order terms respect-
ively. Two-photon excitation from a single beam has been assumed for
simplicity.
Setting H equal to the spin-orbit interaction Hso, Judd and Pooler [8]
showed that the third-order term explained experimental observations of
the 8S 7 / 2 --.6PJ transitions in Gd 3 +. Downer and Bivas [10] explained
observed two-photon absorption of the 8S7/2 --.6IJ transitions in Gd3+ by
use of third-order terms with H equal to the crystal-field interaction H eF ,
and fourth-order terms including both spin-orbit and crystal-field interac-
tions. These higher-order terms seem to be particularly important for
transitions which violate the selection rules of the lower-order theory.

Thus, the spin-orbit interaction is particularly important for transitions
which violate the I1S = selection rule, and the crystal-field interaction is
important when the I1L, 11)::;; 2 selection rule is violated. A complete
review of the importance of these higher-order terms is given by Downer
[5].
Other third-order contributions involving electron correlation have been
introduced [20, 21] in two-photon absorption calculations. In these calcu-
lations, only those terms which scale directly with the second-order V(2)
term have been calculated [22], since Smentek-Mielczarek and Hess [23]
have concluded that those third-order terms which are dependent on the
particular transition are small compared with the other terms examined,
and thus can be ignored. Since third-order terms involving electron corre-
lation do not affect any of the second-order selection rules, we would not
expect these terms to have the large selectively important effects found in
the spin-orbit and crystal-field interaction terms. Third-order electron
correlation has been shown to have a significant impact on absolute line
strengths, but its impact on relative transition probabilities is considered to
be negligible.
4.1.3 Third-order contributions to one-photon absorption
Each of the third-order contributions discussed above for two-photon
absorption has an analogous contribution in one-photon absorption.
Explicitly adding the third-order contribution, we rewrite expression (4.2)
as
_ I [(iIHcFln)(nIE.DIf) + (iIE.Dln)(nIHcFlf)]
1
n hWni hWnf
+ I [(iIHcFlm)(~IHln)(nIE.DIf) + (iIE.Dlm)~mIHln)(nIHCFlf)] _ ... 12

m.n h WmiWni h WmfW nf
(4.10)
About two-thirds of the observed one-photon lanthanide transitions are

spin forbidden, where fl.S = 1. Thus, third-order spin-orbit terms, where
H = Hso, can be expected to have an importance analogous to the spin-
orbit contribution for two-photon transitions.
Since the second-order forced electric dipole term includes contributions
from the odd-parity crystal field, selection rules for one-photon transitions
are fl.L, fl.] ~ 6 rather than the fl.L, fl.] ~ 2 selection rule for two-photon
absorption. Thus, third-order crystal-field terms are not necessary for a
direct connection between states with fl.L>2 and fl.]>2. For this reason,
we would expect third-order crystal-field terms to have a lesser importance
in one-photon transitions than for analogous two-photon transitions.
Extensive work has been done by Smentek-Mielczarek and Hess [24,
25] on the subject of third-order electron correlation effects in one-photon
lanthanide transitions. While not desiring to discount the importance of
third-order electron correlation effects on absolute one-photon transition
line strengths, we do not contest the assertion by Smentek-Mielczarek and
Hess [23] that electron correlation terms which are dependent on particu-
lar transitions are relatively negligible.
We therefore conclude that the most significant static third-order modi-
fications to relative one-photon line strength intensities will come from
contributions involving the spin-orbit interaction.
4.1.4 Dynamic coupling

All of the coupling models presented so far have been limited to pertur-
bations of the electric dipole model which are essentially atomic in nature,
assuming only a static connection between the central rare earth ion and
the surrounding ligands. However, there is another known contribution to
transition line strengths which involves a dynamic coupling between the
rare earth ion and surrounding ligands.
The model of pseudoquadrupole transitions was first introduced by

J0rgensen and Judd [26], where an inhomogeneous dielectric in the ligands
causes enhancement in the normally negligible electric quadrupole term,
allowing line strength enhancements from this term of as much as five orders
of magnitude. Alternatively, Mason et al. [27] considered that the electric
quadrupole induced on the rare earth ion can polarize the ligands, causing
induced dipoles on the ligands which combine to yield a net nonvanishing
dipole moment. This second mechanism, called 'dynamic coupling' or
'ligand polarization', has become the standard explanation [4] of transitions
which are hypersensitive to the environment. Judd [28] later showed that
these two mechanisms are physically identical, although different with
regard to formal details. Reid and coworkers [29-31] developed this idea
further to show that ligand polarization provides a significant contribution
in all lanthanide transitions.
Other mechanisms, such as overlap and covalency effects [32] and charge
transfer [33, 34] have been considered, but the significance of their contribu-
tions to lanthanide line strengths has not been adequately determined. More
recently, Reid and Ng [35] included these mechanisms along with ligand
polarization effects into their more complete second-order analysis of rare
earth transitions.
4.2 THEORETICAL FORMALISM: DERIVATION OF THIRD-

ORDER TERMS AND CALCULATIONAL METHODOLOGY
4.2.1 Derivation
The line strength for a one-photon intra-IN transition from initial state i to
final state f is
M~Mj 1(i'M'f)
- L{LI';/([i]IEDln)(nIHcFIU])
n
+ LI,ti 1 ([i] IHcFln)(nIE DI [fJ)}

+L {LI.;-jLl,;;}(iIEDln)(nIHlm)(mIHcFIf)
n,m
f)
where (i IMI is the standard magnetic dipole term, the first term in braces
gives the standard forced electric dipole contribution, and the last term in
Derivation and calculation of third-order terms 105
braces gives the new electric dipole contribution. The LJ kj denote energies of
intermediate state k above the initial state i or final state f Standard forced
electric dipole terms arise in the first order of time-dependent perturbation
theory. While the even-parity magnetic dipole operator directly links pure [N
states, the electric dipole term requires a first-order static perturbation of
initial and final states by the non centro symmetric crystal field. Consequent-
ly, this term is designated as the 'second-order' contribution. For spin-
forbidden transitions, static perturbations by the spin-orbit interaction must
also be included. This point has been emphasized by placing [i] and [f] in
brackets to denote that intermediate coupled wavefunctions must be used
for AS> 0 transitions.
The final term in this expression arises in the second order of time-
dependent perturbation theory, and introduces an extra energy denomina-
tor along with the new interaction H. In general, H represents the com-
plete Hamiltonian, but in this case we restrict our attention to the part
H = Hso, which is the most important contribution for spin-forbidden
transitions. The first-order odd-parity static crystal-field perturbation of
initial and final states is required in this term as well, and is shown
explicitly. Consequently, this term is designated as the 'third-order' contri-
bution. When H = Hso, initial and final states which differ in spin by one
unit can be linked directly even in the Russell-Saunders limit. In order to
emphasize this point, i and f are written without brackets to denote LS
coupled wavefunctions. Because of this direct linkage, it is possible for
third-order terms to rival or exceed lower-order terms in magnitude des-
pite the extra energy denominator.
In order to evaluate the second- and third-order terms for specific
transitions, we employ the same approximations as in the original theory [2,
3]. Since the intermediate states belong to opposite-parity configurations
which lie far above both i and f in energy, we follow the conventional
approximation of Judd and Ofelt that the LJ kj are constant for all k within a
particular configuration, allowing sums over each excited configuration to
be evaluated by the closure approximation. Individual operators then unite
into compound operators acting solely between levels of configuration [N.
After these compound operators are recoupled into a convenient spherical
tensor form, their matrix elements can be computed. Details of this operator
recoupling are given in the appendix of ref. 36 using the powerful techniques
of second quantization [37]. Resulting operators, acting between initial state
i and final state f of [N, with intermediate states belonging to [N - 11', are the
second-order term
2[1] [1'] ([
o
1 ~) Ei? L (~
0 kodd
k 1')
0 0
L [t]
teven
1/2 {1
[
~ ~} (E(1) B(k)(t) U(t)
(4.12)
and the third-order term
2[/][1'](1
o
1 I')E,-?
0 0
L
kodd 0
(I k
1')(/{/}1/2
0 0 teven
L {1I k
It'}
+2[/][1'](1
o
1 I')E/-:?
0 0 kodd
L (I k
1') (dl'}1/2
0 0 0
(4.13)
where the notation {/}=/(/+ 1)(2/+ 1)/2, [/]=21+1, has been used for
brevity. The (I are spin-orbit coupling coefficients for nl electrons. Spheri-
cal Racah tensors [38] U(t) and W(lA)t result from the recoupling of
the electric dipole, crystal-field and spin-orbit operators. E(1) denotes the
electric field vector and B(k) is a tensor, the qth component of which
equals (-l)q B':q, where the B: are the ordinary crystal-field para-
meters.
For lanthanides, 1=3, and the only possible values of l' are 2 and 4,
corresponding to d- and g-type excited configurations respectively. The
k = 3, 5 components provide the only important contributions to the
noncentrosymmetric crystal field; the special case of k = 1 will be discussed
later. Substituting these values into expressions (4.12) and (4.13) and
evaluating [39] the required 3-j, 6-j and 9-j symbols, we obtain the following
explicit expressions for the second-order term,
(4.14)
for d orbital excited configurations, and

_~(5)1/2 Eg"f1(E(1) B(3(2)U(2)
_~(?d1/2 E;i 1 [(E(1) B(3 (4) _ 2(H)1/2(E(1) B(5 (4)] U(4)
+tH~r)1/2 E;i1 (E(1) B(5(6) U(6) (4.15)

for g orbital excited configurations. Likewise, the third-order term becomes

[ - 8(fs)I/2(r U(2) W(l1)O -1!((r-(d)W(11)2 -1s(3)1/2 (r W(12)2
+/s(3)1/2((r-i-(d)W(13)2]Ed?(E(I) B(3)(2) + [ _4()1/2 (r U(4) W(II)O
- nm 1/2((r- g(d)W(13)4 -ll (55)1/2(r W(14)4

+ n(1)1/2((r - (d)W(15)4] x Ed? [(E(1) B(3) (4) -rdH)I/2(E(I) B(5)(4)]
+ [ -ll (+W/ 2(d W(13)4 _n-(~)1/2(d W(15)4]E d?(E(I) B(5)(4)
+ EiY (39)1/2 (r U(6) W(11)O + 151 (~)1/2((r _~(d)W(15)6
+n-(6)1/2(r W(16)6]E dr 2(E(1) B(5)(6) (4.16)
for d orbital excited configurations, and
[ - 4(3f) 1/2 (r U(2) W(II)O - ~((r - (g)W(II)2 _~(3)1/2 (r W(12)2
+ 4(3)1/2((r-4(g) W(13)2]E;r2(E(I) B(3) (2) + [ -12(';"-)1/2 (r U(4) W(11)O
_~ml/2((r_~(g)W(13)4 -~(ll)I/2(r W(14)4 +~m)I/2((r _ (g)W(15)4]
x E;r2 [(E(I) B(3)(4) _ 2(H)I/2(E(I) B(5) (4)] + [20W/2 (g W(13)4
+ ?1 (~)1/2 (g W(15)4]E;r2(E(I) B(5)(4) + [i--(n) 1/2 (r U(6) W(II)O
+-rh(~)1/2((r-1f(g)W(15)6 +-rh-(6)1/2 (r W(16)6]E;r2(E(I) B(5)(6) (4.17)
for g orbital excited configurations.
4.2.2 Tensors
As illustrated by expressions (4.14H4.l7), in the standard second-order
theory, intra-fN transitions can be connected only through the second-,
fourth- and sixth-rank tensors U(2), U(4) and U(6). When third-order
terms are included, other even-rank U(/) W(l1)O and W(1.l.)1 terms are
added. Reduced matrix elements ([S L J ] II U(/) II [S' Lj,]) for intermediate
coupled ions throughout the trivalent lanthanide series have been com-
piled by Carnall et al. for ions in aqueous solution [40-43] and in LaF 3
[18]. Relevant third-order eLJl1 W(l1)OIIS'Lj,) and eLJl1 W(1.l.)/IIS'Lj,)
reduced matrix elements have been extensively tabulated by Burdick
[36].
Keeping track of appropriate multipliers, we can add third-order tensors
of the same rank to each other to form single second-, fourth-, and
sixth-rank operators Of), O~4), and O~6). When these composite third-
order terms are added to the corresponding second-order tensors and the
sums squared, we obtain combined second- and third-order operators which
are acted on by phenomenologically fitted parameters T(2), T(4), and T(6) to
yield total line strength predictions. Summing over M J and Mj, we obtain
the simplified final equation for the electric dipole contribution to the line
strength P:
p=_V_ L T().)I(iIIU(A)+O~A)llfW (4.18)

2J + 1 A=2,4,6
Since cross-terms between second- and third-order matrix elements are
non-zero, total third-order contributions may be negative as well as positive, It
should be emphasized that the relative importance of new third-order
contributions with respect to the standard second-order terms can be deter-
mined rigorously since the phenomenologically fitted T(A) act identically on
second- and third-order terms and no new fitted parameters are introduced.
4.2.3 Important special cases

Before concluding this section, it is important to comment on the possible
importance of three special cases arising from expressions (4.12) and (4.13),
but which are conventionally neglected.
(a) k= 1 terms
The k= 1 terms arising in expressions (4.12) and (4.13) are possible only if
linear crystal-field terms B(I) are allowed. Although it can be argued that
linear terms imply nonequilibrium between the rare earth ion and its
complex, Judd [44] has pointed out that electrons of the rare earth ion may
produce an electric field at the nucleus which exactly cancels that coming
from the ligands. Consequently, the nucleus remains stationary while the
electrons develop an asymmetric charge distribution which reflects the
presence of ql) terms in the crystal-field Hamiltonian. Thus, in crystalline
sites possessing symmetries which allow a linear crystal-field term (C n , Cm
and Cs), these k = 1 terms may playa significant role.
Allowing k= 1 in expressions (4.12) and (4.13), we get the following new
second-order terms,
+ 6(fs-)1/2 E;;/(E(I) B(I)(2) U(2)
(4.19)
+i(~)1/2 E;I(E(I) B(I)(2) U(2)
for d and g orbital excited configurations respectively. The zero-rank term

has been neglected since U(O) = N(7) -1/2 is just a number and cannot
connect orthogonal eigenstates. New third-order terms are found to be
(4.20)
for respectively d and g orbital excited configurations. Notice that, except
for an extra factor of 2, these terms are mathematically equivalent to those
used for two-photon absorption. That is, if we replace the B(1) in expressions
(4.19) and (4.20) with E(1) and divide by 2, we obtain the second-order and
third-order spin-orbit operators necessary for two-photon transition calcu-
lations [5].
An examination of the 7Fo+-+500 transition of Eu 3+ and Sm2+ provides
the best indication of the importance of these linear crystal-field terms in
general, and the W(11)O term in particular. Since neither U(O) nor any other
second-order operator can provide a connection between these two levels,
this transition is allowed in second order only through J-J mixing, known
to be very weak for lanthanide ions. The third-order W(l1)O term, in
contrast, provides a direct linkage between two states with J = O. In most
environments, this transition is absent or extremely weak, which can be
adequately explained by means of J-J mixing. However, strong 7FO+-+500
transition lines have been observed in Eu 3+ - and Sm2+ -doped crystals with
ionic substitution sites possessing symmetries allowing a linear crystal-
field term. To give a few examples, Kiss and Weakliem [45] and Alam and
di Bartolo [46] observed that the 50 0-+ 7FO emission line in Sm2+ : BaCIF
(site symmetry C 4v ) was comparable in intensity with the neighbouring
allowed 500-+ 7F 2 line. Blasse and Bril [47] and Nieuwpoort and Blasse
[48] studied the intensity of this emission line in Eu3+ in a variety of
crystals. They found that, in Eu3+ :Sr2 Ti0 4 (C 4v site symmetry) and
Eu 3+ :Ba3Gd2 WO g (C 3v site symmetry), 50 0-+ 7FO had 50% of the intensity
of 5DO -+ 7F 2, and showed strong intensities in several other crystals with
C2v , C 4v and C. substitution site symmetries. In their phenomenological
analysis of energy levels in Sm2+: SrCIF, Grenet et al. [49] invoked a
dominant D[(ll)OA1] term, analogous to the W(l1)O term discussed above,
which was not significantly altered by inclusion of J-J mixing. A more
recent study of the fluorescence of Eu3+ -doped oxyapatites [50] found the
500-+ 7F 0 transition to dominate. In these selected cases, J-J mixing
appears to be inadequate to explain either the magnitude of the transition
intensity or hypersensitivity to the crystalline environment. They seem to
provide strong evidence for the existence and importance of linear crystal-
field terms. Preliminary calculations [51] have been done in an attempt to
quantify magnitudes of the linear crystal field necessary to explain these
observations.
Contributions from a linear crystal field may also help to explain

observed IlJ = 2 hypersensitive transitions. Current theory explains this
hypersensitivity solely in terms of ligand polarizability [35]. However, in a
phenomenological study of Eu 3 + -doped lanthanum oxychloride, H6lsa and
Porcher [52J found that a large B(210) term, corresponding to our
(E(1) B(1)(2), was necessary to explain the hypersensitive 5DO -+ 7F 2 fluor-
escence.
(b) g Orbital contributions

Conventional analysis considers only d orbital contributions, since the
4fN- 1 5d configuration is by far the lowest energy excited configuration in
the free ions [14,16]. However, as pointed out by Becker et al. [53J, linear
combinations of ligand orbitals can transform like components of g electron
wavefunctions. Low-lying states of g character can be formed from these
ligand orbitals, and the energy denominators for d and g orbital excited
configurations may be similar. The importance of g orbital electrons in
lanthanide transitions has long been suspected. In their early attempts to
calculate the Tv.) parameters, both Judd [2J and Krupke [54J concluded
that 4fN -lg configurations make an important contribution to electric
dipole transition line strengths. However, neither they nor subsequent
investigators included these configurations into a quantitative Judd-Ofelt
analysis of lanthanide transitions.
The successes of the Judd-Ofelt formalism in predicting lanthanide
transition line strengths have been a major contributing factor in the neglect
of these terms. In this formalism, crystal-field terms and radial integrals are
included in the empirically fitted Judd-Ofelt parameters. Since second-order
UP.) matrix elements act solely between initial and final states within the 4fN
ground configuration, inclusion of g orbital excited configurations does not
change individual second-order matrix elements, but merely changes their
common multipliers. Thus, appropriate scaling of the fitted parameters will
exactly cancel any second-order g orbital contribution. However, recent
observations of an anomalous absence of asymmetry in certain electronic
Raman transitions in lanthanide phosphates [53, 55J not only have pro-
vided strong evidence that intermediate states of 4fN- 1 g character are
comparable in importance with 4fN -1 d states in lanthanide radiative
processes, but also have given us a quantitative way to establish the
magnitude of the g orbital contribution.
The influence of 4f N- 1 g states enters in second order as well as third
order. Third-order g orbital revisions are significant, and will be considered
in the next section. However, the importance of these g orbital states can be
demonstrated by considering only second-order revisions. Comparing ex-
pressions (4.14) and (4.15), including the radial integrals which multiply these
expressions, we can derive the ratios of g to d orbital contributions. The
second- and sixth-rank terms are
U(2) = ~ Edc (4flrln'g)(4flr3In'g) U(2)

(4.21)
g 2 Egc (4flrln'd)(4flr3In'd) d
U(6) = ~ Edc (4flrln'g)(4flr5In'g) U(6)

(4.22)
g 13 Egc (4flrln'd)(4flr5In'd) d
where U~).) are the second-order g orbital terms, scaled so that U().) in
expression (4.14) can be replaced by the sum U~).) + U~).). Calculation is more
complicated for fourth-rank terms, since the ratio between crystal-field
terms (E(1) B(3)(4) and (E(1) B(5)r) must be known. Considering three cases,
(a) the first term dominates, (b) the two terms are equal, and (c) the second
term dominates, we obtain respectively
3 (4flr3In'g)
11 (4flr3In'd)
U(4) = Edc (4flrln'g) (4)X 15 (4flr 3In'g)+(4flr 5In'g) (4.23)
g Egc (4flrln'd) U d ~ 14 (4flr 3In'd)+(4flr 5In'd)
(4f Ir 5 1n'g)
6 (4f1 r5In'd)
Since the magnitudes of these crystal-field terms will not change for different
transitions from the same ion site, the effect of these g orbital contributions
can be completely masked by an appropriate scaling of the T().).
Radial integrals (4f11In'l') are included in the T().) parameters and thus
normally are not calculated. However, from the Raman asymmetry data of
Becker [55], the value of (Edr/Egc)(4flrln'g)/(4flrln'd) can be determined
rigorously, and is found to be 0.91 for ErP0 4 and 1.09 for TmP0 4 . Since
these ratios are close to unity, it is reasonable to make the further
assumption that (4fllln'd)=(4fllln'g) for k=3,5. Thus, setting all ratios of
energy denominators and radial integrals in expressions (4.21~(4.23) equal
to unity, we find that inclusion of these g orbital terms necessitates a 92%
decrease in the parameter 1(2), anywhere from a 38% to 98% decrease in the
parameter 1(4), and a 14% decrease in the parameter 1(6)'
Parameter revisions of these dimensions will strongly affect any calcula-
tions which attempt to relate the magnitude of the Judd-Ofelt parameters to
the underlying structure of the crystal field, such as those of Krupke [54]
and Alfrey et al. [56]. Recent calculations of Smentek-Mielczarek [57] tend
to discount the importance of g orbital contributions in explaining Raman
asymmetry data. However, as the above calculations show, even small g
orbital contributions will yield significant modification of the Judd-Ofelt
parameters.
( c) Rank 3 and 5 terms

Two other third-order operators arise from the last term of expression (4.13),
but were neglected in the expansions of (4.16) and (4.17). Written out
explicitly, the d orbital contributions take the forms
+ ~(t)1/2 E';/'d(E(1)B(3)(3)W(13)3
(4.24)
If these odd-rank tensors are important, we would have to add two other
parameters T(3) and 1(5) into our calculations, increasing the number of
fitted parameters from three to five. We have found, however, that these
odd-rank terms are always smaller than the corresponding even-rank terms,
typically by several orders of magnitude. Thus, for these odd-rank terms to
be at all important, magnitudes of the fitted parameters T(3) and T(5) would
have to be very much larger than of the parameters T(2), T(4) and T(6). That
is, crystal-field terms in (E(1) B(3)(3) or (E(1) B(5)<5) would have to be more
than a hundred times the magnitude of the corresponding (E(1)B(3)(2),
(E(1) B(3)(4), (E(1) B(5)(4) and (E(1) B(5)(6) terms, which is extremely unlikely.
Thus, we conclude that these odd-rank tensors do not provide a significant

contribution to lanthanide line strengths, and therefore can be ignored.
4.2.4 Conclusions
These third-order computations can determine unambiguously just how
important the third-order terms are relative to conventional second-order
terms. Since the ratios of third- to second-order contributions do not depend
on the radial integrals or fitted parameters 1().), their relative importance or
unimportance for specific transitions is an unambiguous output of the
calculation.
Through appropriate modification of fitted parameters T().), line strength
calculations which include third-order terms can be refitted to experiment,
allowing us to determine whether inclusion of third-order terms can better
explain the experimental data than can second-order terms alone.
4.3 ANALYSIS: LANTHANIDES IN AQUEOUS SOLUTION

In order to determine the true significance of third-order forced electric
dipole terms, Downer et al. [58] and Burdick et al. [59] compared results
calculated by the methods of section 4.2 with the very complete experimen-
tal results and second-order calculations of Carnall et al. [15, 19] for
trivalent lanthanides in aqueous solution. Since the experimental results
were taken from aqueous solution, the symmetry axes of ionic sites were not
oriented with respect to each other, so there is no transition dependence on
linear polarization direction. Also, individual Stark levels were not dis-
Analysis: lanthanides in aqueous solution 113
criminated. Thus, for each individual transition, only a single, total experi-
mental oscillator strength is given for excitation from unpolarized light.
The results of these extensive calculations for all trivalent lanthanides
have been presented graphically by Burdick et al. [59] and in a more
complete tabulated form by Burdick [36]. In order to illustrate the overall
trends along with some of the more interesting details, we present here the
results for four ofthese ions (Nd3+, Sm3+, Eu 3 + and Gd3+). Although these
results are presented considering only d orbital terms, we will close this
section with a quantitative discussion on the importance of third-order g
orbital terms.
4.3.1 Trivalent lanthanides
(a) Nd3+
Table 4.1 gives the second- and third-order line strength contributions for
the various transitions from the ground state 419/2 for Nd3+ in units of
P x 108 , where P is the probability of transition. The first column specifies
the particular transition. The second and third columns give second- and
third-order contributions respectively, and the fourth column gives the
percentage ratio between the third-order contribution,
and the second-order contribution,
(4.26)
where the 1(;.) are given in the bottom row of this table for the fitting when
both second- and third-order terms are included, i is the ground state 419/2,
and the final state f is listed in the first column. The last three columns of
this table compare total experimental line strengths, in the fifth column, with
line strengths calculated using only second-order terms and including
third-order terms, in the sixth and seventh columns respectively. The
seventh column is a summation of the values presented in the second and
third columns. Transitions which are not experimentally resolved have been
grouped together in the last three columns.
Since the third-order contribution (4.25) includes cross-terms between
second- and third-order matrix elements, as well as squares of third-order
matrix elements, the third-order contribution can actually be negative, as is
the case for a number of transitions listed in this table. Third-order
contributions for spin-allowed quartet to quartet transitions presented in
this table have systematically small positive values ranging from 5% to 13%
Table 4.1 Calculated second- and third-order contributions together with experi-
mental and calculated line strengths for Nd3+ transitions from ground state 41 9/2 ,
given in P x 108
Transition Line strength components Total line strengths
2nd 3rd Ratio Experimental 2nd With 3rd

41 1S12 27.0 1.4 5.0% 49 28 28
4F3/2 186.4 15.7 8.5% 256 234 202
4FSI2 635.2 47.9 7.6% 884 862 847
2H9/2 152.8 10.5 6.9%
4F7/2 594.4 43.6 7.3% 890 950 979
4S 3/2 318.7 22.8 7.2%
4F9/2 68.2 7.0 10.3% 65 72 75
2Hl1/2 18.6 -1.5 -8.3% 15 19 17
4G SI2 668.1 62.0 9.3% 976 988 990
2G 7/2 228.1 31.5 13.8%
2KI3/2 61.6 22.2 36.0% 701 535 512
4G 7/2 265.0 8.7 3.3%
4G 9/2 136.5 17.7 13.0%
2K ISI2 35.1 10.9 31.1% 233 139 132
2G 9/2 44.4 -3.2 -7.2%
2D,P 3/2 19.2 0.3 1.8%
4G II/2 22.6 2.3 10.1%
2p 1/2 39.9 7.8 19.5% 410 55 48
2DSI2 5.2 -1.8 -35.2% 9 5 3
2P,D 3/2 3.8 -0.4 -11.7% 3 4 3
4D3/2 309.3 26.6 8.6% 942 1029 1004
4DSI2 154.3 17.6 11.4%
2111/2 31.9 11.4 35.6%
4DI/2 351.6 25.1 7.1%
2L ISI2 62.5 14.0 22.3%
21 1312 7.0 3.0 42.9% 270 45 47
4D7/2 29.7 1.0 3.4%
2LI7/2 5.1 1.2 22.6%
2H9/2 13.1 4.8 36.9% 13 17 18
2D3/2 21.7 -1.8 -8.5% 33 29 21
2Hl1/2 0.9 0.5 50.4%
2DsI2 13.1 -1.8 -13.9% 28 13 11
2FSI2 7.4 6.5 87.9% 18 9 14
T(2) T(4) T(6)

1.20 6.44 10.20 (x 10- 9 cm) 2nd order
1.81 4.73 10.03 (x 10- 9 em) 2nd and 3rd order
of the corresponding second order terms. In contrast, third-order contribu-
tions for spin-forbidden quartet to doublet transitions take on all values
between -35% and +88% of the corresponding second-order terms. This
trend holds true for all trivalent lanthanides. That is, third-order contribu-
tions to spin-allowed transitions have systematically small positive values
while third-order contributions for spin-forbidden transitions have large and
greatly varying contributions.
The largest relative third-order contribution comes from the 419/2 -+2F S/2
transition, where the third-order term is 88% of the magnitude of the
second-order term. As given in the last line of this table, second-order
calculations underestimate this transition by a factor of 2. This large
third-order term greatly improves the calculated fit to experiment, with a
resulting underestimation for this transition of only 22%. The third-order fit
for this transition would actually be closer, except for the reductions to the
Judd-Ofelt parameters, shown at the bottom of this table, necessary to fit
the other transitions.
(b) Sm 3 +
Second- and third-order electric dipole and magnetic dipole contributions
for transitions of Sm 3+ are tabulated in Table 4.2. The first column gives the
transition from the ground state 6H s/2. The second column of this table
presents the calculated magnetic dipole contribution to the oscillator
strengths in units of P x 108 As shown in this table, only the 6Hs/2-+4GS/2,
4F S/2 and 4H7/2 transitions have a significant magnetic dipole contribution,
with a zero or negligible contribution in all other observed transitions. The
third and fourth columns present the second- and third-order electric dipole
contributions to the transition oscillator strengths respectively, also in units
of P x 108 , and the fifth column gives the ratio of the third- to second-order
electric dipole contributions.
Third-order contributions to the spin-allowed sextet to sextet transitions
are all consistently between 5% and 12% of the corresponding second-order
contributions. In contrast, third-order contributions to the spin-forbidden
sextet to quartet transitions have some of the largest relative third-order
contributions of any trivalent lanthanide, often being greater than 50% of
the corresponding second-order contributions. For three transitions,
6Hs/2-+4F3/2, 4H9/2 and 4Hll/2' the third-order contribution is actually
greater than the second-order contribution, with respective contributions of
127%, 160% and 121 %. Experimental observations of the latter two
transitions are hidden by the close-by dominant 6H s/2-+ 4D7(2 transition.
However, the 6H s/2-+ 4F 3/2 transition is clearly resolved experimentally. The
large third-order contribution results in significant improvement in the
fitting of this transition, as illustrated in the last three columns of Table 4.2.
The experimental oscillator strength for this transition is 0.90 x 10- 8 , while
Table 4.2 Calculated magnetic dipole and second- and third-order electric dipole contributions
together with experimental and calculated line strengths for Sm 3 + transitions from ground state
6H s/2 , given in P x 10 8

Magnetic 2nd 3rd Ratio Experimental 2nd With 3rd
dipole
contribution
6F 1 / 2 22 3 11.5% 26 26 26
6H 1S / 2 1.5 0.1 9.0%
6F 3 / 2 89 8 9.3% 97 99 98
6F s/2 166 15 8.9% 170 184 181
6F 7 / 2 280 21 7.5% 274 300 301
6F 9 / 2 186 12 6.3% 197 198 198
6F 11/2 30 2 5.6% 31 32 32
4G S / 2 2.3 1.1 0.5 50.4% 3.0 3.5 3.9
4F3/2 0.10 0.13 126.8% 0.90 0.10 0.23
4G 7 / 2 5.8 -1.1 -19.5% 5.2 6.2 4.7
41 9/2 3.2 1.3 42.0% 153 54 67
4M 1S / 2 35 9 24.7%
4111/2 13 6 50.7%
41 13 / 2 32 11 34.5% 68 34 44
4FS/2 0.6 0.5 -0.3 -56.7% 3.0 1.6 0.8
4M17/2 6.5 1.8 27.0% 39 13 13
4G 9/2 5.4 -1.1 -19.8%
41 1S12 0.3 0.1 38.7%
6. 4P S12 51 6 11.5% 56 57
4L 13/2 29 5 17.4% 43 31 34
4F7/2 2.9 -0.7 -23.8% 384 380 381
6P 3/2 339 34 10.1%
4Kll/2 4.6 1.0 21.0%
4L 1S12 8.5 2.7 32.2% 42 11 13
4G l1/2 1.6 0.3 18.3%
4Dl/2 0.05 -0.05 -96.7% 106 126 126
6P 7/2 114 8 6.9%
4L17/2 0.3 0.1 26.2%
4K 13/2 2.7 0.1 3.2%
4F9/2 0.9 0.1 14.6%
4D3/2 56 -2 -4.5% 123 103 90
(4D,6P)SI2 38 -1 -3.6%
4H7/2 0.9 2.0 1.3 63.2% 6.0 2.9 4.2
4H9/2 1.5 2.4 159.9% 80 67 68
4D7/2 61 0 0.0%
4Hll/2 1.0 1.3 121.3%
T(2) T(4) T(6)
1.17 5.32 3.47 (x 10- 9 em) 2nd order

1.06 4.84 3.31 (x 10- 9 em) 2nd and 3rd order
standard second-order fitting underestimates this strength by a factor of 9 at

0.10 x 10- 8 . However, although the predicted oscillator strength is more
than doubled with the inclusion of third-order terms, to 0.23 x 10- 8 , this
transition is still underestimated by a factor of 4.
Another significant modification occurs in the 6Hs{2--+4H7{2 transition,
where second-order fitting underestimates this transition by more than a
factor of 2. The large 63% third-order contribution modifies the fitting
upward from 2.9 x 10- 8 to 4.2 X 10- 8 , much closer to the experimental value
of 6 x 10- 8. In contrast, the third-order contribution cancels out 97% of the
second-order term for the 6H s{2 --+ 4D 1{2 transition. However, the large
close-lying 6H s{2--+ 6P 7{2 transition hides any experimental confirmation of
this extremely large cancellation.
(c) Eu3+
Second- and third-order electric dipole contributions to transitions from the
ground state 7F 0 of Eu 3 + are given in Table 4.3 in the same format as Table
4.1. As shown in Table 4.3, third-order terms in ground state transitions of
Table 4.3 Calculated second- and third-order contributions together with experi-
mental and calculated line strengths for Eu 3 + transitions from ground state 7FO'
given in P x 10 8

2nd 3rd Ratio Experimental 2nd With 3rd
5D 2 0.45 -0.24 -53.7% 0.21 0.21 0.21
5L6 14.49 3.22 22.2% 17.7 17.8 17.7
5G 2 0.42 0.04 8.9% 0.19 0.45
5G4 1.00 0.78 77.3% 4.3 5.6 7.8
5G6 3.75 2.30 61.4%
5D4 1.63 0.07 4.2% 1.7 1.7 1.7
5H4 2.18 0.12 5.3% 2.3 2.3
5H6 6.50 0.87 13.4% 7.3 8.0 7.4
5F 2 0.35 0.14 40.3% 0.16 0.46
5F4 6.13 3.27 53.3% 6.4 9.4
514 1.11 -0.43 -39.1% 1.1 0.7
51 6 2.19 0.68 31.1% 2.7 2.9
5K6 1.52 0.45 29.5% 2.2 1.9 2.0
T(2) T(4) T(6)
1.88 8.59 6.96 (x 10- 9 cm) 2nd order
4.08 8.33 5.71 (x 10- 9 cm) 2nd and 3rd order
Eu3+ are quite significant, with average contributions greater than one-third
the magnitude, either positive or negative, of the corresponding second-
order terms. Large negative third-order contributions cancel out 54% and
39% of corresponding second order terms for the 7FO-+5D2 and 514
transitions respectively. Also, large third-order contributions add 77%, 61 %
and 53% to respective second-order terms for the 7FO-+5G4, 5G 6 and 5F4
transitions.
The first transition given in Table 4.3 is the 7F 0 -+ 5D 2 transition. Since
this is the only transition used in fitting the empirical parameter T(2), an
exact fit between calculation and experiment can be guaranteed. Third-order
results for this transition are interesting because the third-order term is more
than 50% of the magnitude of the second-order term and of opposite sign.
Thus, the Judd-Ofelt parameter 1(2) must be more than doubled from its
value of 1.88 x 10 - 9 cm to 4.08 x 10 - 9 cm, as shown at the bottom of this
table.
(d) Gd 3 +
Second- and third-order electric dipole and magnetic dipole contributions
for ground state transitions of Gd3+ are tabulated in Table 4.4 with the
same format as for Table 4.2. The first three transitions given in this table
are for the 8S 7/2-+6PJ transitions. For the 8S 7/2-+6P 7/2, 6P 5/2 transitions,
third-order terms are 60% and 113 % of the magnitude of the corresponding
second-order terms. However, third-order contributions have little effect on
predicted intensities for these two lines since the magnetic dipole term
dominates the line strengths. The 8S7/2 -+ 6p 3/2 transition, by contrast, has
no magnetic dipole term, so the large third-order term dominates. The
oscillator strength measured by Carnall et al. [19] for this transition is an
order of magnitude greater than predicted by second-order theory. How-
ever, inclusion of the large third-order term, 948% of the magnitude of the
second-order contribution, accounts for the strength of this previously
unexplained transition.
Relative to the corresponding second-order contributions, overall third-
order contributions for Gd3+ transitions are the greatest of any lanthanide
ion, with contribution percentages ranging from 60% to 948% for
8S 7/2-+6PJ transitions, as discussed above. This fact could have been
expected for two reasons. First, all observed transitions for Gd3+, along
with Eu3+ and Tb3+, are spin forbidden in LS coupling, a fact which
accounts for the reduction of the second-order matrix elements for these
three ions compared to the other trivalent lanthanides. The second reason
comes from a unique characteristic of half-filled shells [60], namely that all
e
diagonal reduced matrix elements s + 1 LII U(k) 11 2S + 1 L) vanish. This latter
fact tends to reduce the value of intermediate coupled reduced matrix
e
elements ([2S + 1 L] II U(k) II s ' + 1 L']) connecting states of differing spin, since
Table 4.4 Calculated magnetic dipole and second- and third-order electric dipole contributions
together with experimental and calculated line strengths for Gd3+ transitions from ground state
BS 7/2, given in P x lOB
Magnetic 2nd 3rd Ratio Experimental 2nd With 3rd
6P7/2 5.5 0.8 0.5 60.2% 7.3 6.8 6.7

6P S / 2 3.10 0.33 0.37 112.7% 4.1 3.6 3.8
6P 3/2 0.012 0.10 948.2% 0.14 0.02 0.11
61 7/2 8.6 2.0 23.5% 12.1 11.2 10.6
61 9 / 2 22.5 5.3 23.6% 84.5 87.7 85.9
61 17/2 47.1 11.0 23.4%
61 11 / 2 39.5 9.4 23.7% 191.4 191.8 191.0
61 13 /2 53.7 12.8 23.9%
61 1S /2 61.3 14.4 23.4%
6D 9 / 2 5.4 1.0 18.6% 7.8 9.3 6.4
6D 7/2 0.5 4.1 0.6 13.9% 8.2 13.8 8.7
6D 3/2 0.8 0.0 4.7%
6D s/2 0.3 2.3 0.2 8.5%
6G 7/2 13.1 4.5 34.4% 12.1 14.9 17.6
6G 9 / 2 30.5 9.4 30.8% 90.3 76.5 96.1
6G 11 /2 36.8 10.4 28.2%
6G S / 2 6.3 2.6 41.1%
6G 3/2 1.7 0.9 56.7% 1.8 2.6
6G 13 / 2 42.5 8.7 20.5% 63.9 45.9 51.2
T(2) T(4) T(6)
3.30 6.06 6.10 (x 10- 9 cm) 2nd order
1.76 5.67 4.83 (x 10- 9 cm) 2nd and 3rd order
the initial and final states will share common LS coupled admixtures,
making second-order contributions unusually small.
4.3.2 Lanthanide fitting

Judd-Ofelt parameters 1(2), 1(4) and T(6) used in fitting calculations with
experiment throughout the lanthanide series are given in Table 4.5 along
with the corresponding standard deviations of the fits to experiment. The
first column in this table gives the trivalent ion, while the second, third and
fourth columns give Judd-Ofelt parameters T(2), T(4) and T(6) respectively in
units of 10- 9 cm for the standard fitting when only second-order terms are
included. The fifth column presents the standard deviation of line strengths
fitted by the least-squares method with experimental values, in units of
P x 108 , where P is the probability of transition. The final four columns
present the corresponding modified Judd-Ofelt parameters and standard
deviations of these fits when third-order terms are included, using the same
format as for the second to fifth columns.
Empirical parameters shown in this table are typically modified by about
20% when third-order contributions are included. In selected cases, how-
ever, the modification is much larger. The most significant parameter change
illustrated in this table occurs for Dy3 +, where the parameter T(2) is
increased by more than a factor of 4 when third-order terms are included,
from 1.93 x 10- 9 cm to 8.50 x 10- 9 cm. Other significant modifications of
this parameter occur in Pr3 + with a factor of 2 decrease, and Eu 3 + with a
factor of 2 increase. The most significant modification of the parameter 1(4)
occurs in Tb 3 +, with a factor of 2.5 decrease, from 9.26 x 10- 9 em to
3.78 x 10- 9 em. In contrast, no parameter modification greater than 20% is
needed for T(6) when third-order terms are included.
With these parameter modifications, the overall fitting of calculated
oscillator strengths to available experimental data from aqueous solution is
somewhat improved. Inclusion of third-order terms yields a significant
improvement in the fitting for Pr 3 +, Nd 3 + and Dy 3 +, ions having the worst
second-order fit. However, the fit between calculation and experiment is
actually slightly worse when third-order terms are included in the calcula-
tion of Tb 3 +, Ho3+, Er3+ and Tm 3+ line strengths. This lack of significant
improvement for some ions may be partially accounted for by the fact that
many important predictions of large third-order contributions could not be
tested with available data, owing to a lack of spectral resolution. Hence
these important cases could not be included in the fit.
The most important predictions of these third-order calculations which
could not be spectrally resolved in aqueous solution include the large third-
order cancellations of the corresponding second-order terms in the
6Hs/2-+4Dl/2 transition of Sm3+, the 7F 6-+ sD 3 ,2,l,O transitions of Tb3+
and the 6H 1S / 2-+ 4(M,Ihs/2, 4F S/2 transitions of Dy3+. Other unresolved
Table 4.5 Judd-Ofe1t parameters used in fitting lanthanide line strengths along with standard
deviations of fits of experiment
Ion 2nd-order fit 2nd- and 3rd-order fit
(J (J
T(2) T(4) T(6) T(2) T(4) T(6)
(x JO- 9 em) (P x J08) (x JO- 9 em) (P x J08)
Pr 3 + 42.00 7.40 41.20 130.2 22.17 7.10 38.30 103.6

Nd3+ 1.20 6.44 10.20 100.0 1.81 4.73 10.03 86.9
Pm 3 + 3.61 3.25 5.42 23.1 3.16 3.13 5.06 22.9
Sm 3 + 1.17 5.32 3.47 31.3 1.06 4.84 3.31 28.7
Eu 3 + 1.88 8.59 6.96 5.4 4.08 8.33 5.71 4.1
Gd 3+ 3.30 6.06 6.10 8.2 1.76 5.67 4.83 5.4
Tb 3 + 0.005 9.26 4.45 6.8 0.20 3.78 3.94 7.9
D y 3+ 1.93 4.44 4.46 58.5 8.50 4.45 5.20 41.4
Ho3+ 0.47 4.05 3.96 28.4 0.23 3.73 3.46 31.3
Er 3 + 2.05 2.51 2.45 15.9 2.06 2.20 2.25 21.4
Tm 3 + 1.03 2.68 2.40 23.5 0.93 2.60 2.37 35.7
transitions with dominant third-order terms are the 6H5/2--+4H9/2.11/2

transItIons of Sm3+, the 6H15/2--+4M17/2 transition of Dy3+, and the
51s--+317 transition of Ho3+. Hopefully, one effect of these calculations will
be to stimulate systematic measurements of many of these weaker, spin-
forbidden levels in hosts which permit adequate spectral resolution. Even in
these aqueous solution data, however, significant improvement has been
achieved in understanding selected individual transitions, such as the
419/2--+2F 5/2 transition in Nd3+, the 6H5/2--+4H7/2 transition of Sm3+, and
the SS7/2--+6P3/2 transition of Gd3+.
4.3.3 g Orbital terms

In order to examine the possible effect of third-order g orbital terms on the
lanthanide fitting calculations, we present numerical results calculated
including both d and g orbital terms for Pr 3 + transitions in Table 4.6,
compiled from the data of Burdick [36]. Numerical results for g orbital
terms have been calculated using the assumption, discussed in section 4.2.3,
that energy denominators and radial integrals for transition terms using g
orbital excited configurations are approximately equal to the corresponding
energy denominators and radial integrals for terms using d orbital excited
configurations. We have also made the assumption that (g is approximately
equal to (d' While these assumptions may not be entirely valid, they do
allow us to determine a rough estimate of what effect inclusion of third-
order g orbital terms will have on fitted line strengths.
The first column of Table 4.6 gives the transition from the ground state
3H4 to the other intra-4f2 levels. Each transition is represented in four rows
of this table for four different calculations. The first row for each transition
gives the results of the standard second-order calculation while the second
row presents the fitting when second-order g orbital terms are included. The
third and fourth rows for each transition present results for d orbital terms
only and for d and g orbital terms respectively, when both second- and
third-order terms are included.
Numerical columns in this table present oscillator strength contributions
given in units of P x lOs. The first numerical column gives the standard
second-order contribution. This number varies between the four rows for
each transition owing to the scaling of fitted parameters T(2), 1(4) and 1(6),
given at the bottom of this table, where parameter values given in each of
the four rows correspond to fittings of the corresponding rows of the main
table. The second and third numerical columns give partial contributions
from the corresponding second-order g orbital and third-order d orbital
terms, where interfering cross-terms with the standard second-order d
orbital term are included. The fourth numerical column gives the remaining
contribution from the inclusion of third-order g orbital terms, including
interference terms between the second-order g orbital and third-order d
Table 4.6 Calculated line strength contributions and experimental line strengths for Pr 3+ transitions
from ground state 3H 4, given in P x 10 8
Transition Term d:2nd g:2nd d:3rd g:3rd Total Experimental
3F3,3F4 2nd (d) 4112 4112 1279
2nd (d and g) 3294 818 4112
2nd and 3rd (d) 3726 122 3848
2nd and 3rd (d and g) 2957 734 88 25 3804
10 4 2nd (d) 132 132 32
2nd (d and g) 108 24 132
2nd and 3rd (d) 120 -23 97
2nd and 3rd (d and g) 98 22 -19 -6 95
ID2 2nd (d) 445 445 308
2nd (d and g) 359 86 445
2nd and 3rd (d) 406 17 423
2nd and 3rd (d and g) 326 78 11 6 421
3P O 2nd (d) 299 299 254
2nd (d and g) 37 262 299
2nd and 3rd (d) 282 16 298
2nd and 3rd (d and g) 36 257 2 16 311
3PI/I6 2nd (d) 727 727 763
2nd (d and g) 272 455 727
2nd and 3rd (d) 643 86 729
2nd and 3rd (d and g) 239 400 41 39 719
3P 2 2nd (d) 1464 1464 1506
2nd (d and g) 1209 255 1464
2nd and 3rd (d) 1364 113 1477
2nd and 3rd (d and g) 1131 238 92 18 1479
I So 2nd (d) 27 27
2nd (d and g) 3 24 27
2nd and 3rd (d) 26 10 36
2nd and 3rd (d and g) 3 24 1 11 39
T(2) T(4) T(6)
42.00 7.40 41.20 (x 10- 9 em) 2nd (d)

3.43 1.72 35.52 (x 10- 9 em) 2nd (d and g)
22.17 7.10 38.30 (x 10- 9 em) 2nd and 3rd (d)
1.13 0.894 33.11 (x 10- 9 em) 2nd and 3rd (d and g)
orbital terms as well as interference terms between the third-order g orbital

term and all other included terms. The last two columns give total absolute
oscillator strengths calculated by these four methods along with experimen-
tal strengths.
As was stated earlier, inclusion of second-order g orbital terms will not
affect the fit to experiment, since the effects of these terms are proportional
to the second-order d orbital terms, and thus can be negated by appropriate
scaling of the fitted parameters. This is shown explicitly in the first two rows
for each transition, where total predicted oscillator strengths are identical.
However, as is shown by the last two rows for each transition, inclusion of
third-order g orbital terms induces some modification of calculated oscil-
lator strengths, although for only two transitions is this change greater than
2%. Thus, the only significant modifications to calculated oscillator
strengths caused by inclusion of third-order g orbital terms are an 8%
increase in predicted intensity for the unobservable 3H4 -+ lSO transition and
a 4% increase for the 3H4 -+3P Otransition.
The most significant change when g orbital terms are included is a major
revision of the fitted parameters. For example, T(l) is decreased by 95%,
from 22.17 x 10 - 9 cm in the third-order d orbital fitting to 1.13 x 10 - 9 cm
when g orbital terms are included. However, the vast majority of this drastic
change is the result of g orbital terms which scale with respect to the
standard second-order term, necessitating a parameter decrease of 92%, to
1.81 x 10- 9 cm. Inclusion of third-order g orbital terms does not significant-
ly affect calculated fitting errors either, with a resultant standard deviation
from experiment which is 2% greater than that for the fit considering only d
orbital third-order terms.
Thus, from a close examination of this table, we find that inclusion of g
orbital terms modifies the fitted parameters very significantly, but does not
have a significant impact on the actual fitting of oscillator strengths. This
same trend holds true throughout the lanthanide series.
4.4 CONCLUSIONS: FUTURE DIRECTIONS

The results of this work demonstrate the importance of third-order contri-
butions involving the spin-orbit interaction for transitions throughout the
trivalent lanthanide series. These third-order terms contribute substantially,
either constructively or destructively, to a majority of spin-forbidden lan-
thanide transitions, with magnitudes typically greater than 20%, and
sometimes greater than 100%, of the second-order electric dipole contribu-
tion to which conventional Judd-Ofelt analysis is confined.
Although inclusion of these third-order terms does not produce a strong
systematic improvement of fitted line strengths, new insight is gained into
selected transitions with a dominant third-order contribution. In those
lanthanides where states of differing spin lie close together, larger spin-
Conclusions: future directions 127
allowed transItIOns tend to dominate the experimental data, and thus
hide many predictions of large third-order contributions. Hopefully, these
predictions will motivate further systematic higher resolution intensity
data.
We conclude that these third-order terms involving the spin-orbit interac-
tion are one important part of a comprehensive intensity analysis of one-
and two-photon lanthanide transition intensities. Other parts include the
second-order calculations of Reid and Ng [35] along with other third-
order terms involving electron correlation [22, 24] and the crystal-field
interaction [10]. At this time, the importance of these other third-order
terms involving electron correlation and the crystal-field interaction for
one-photon lanthanide transition intensities have not been quantitatively
examined. Although the overall effects from these terms on the fitting of
lanthanide transition intensities are likely to be less than those of the
spin-orbit terms, they still may provide important explanations for selected
transitions which cannot otherwise be explained.
Before concluding this chapter, we wish to comment on two other
directions for future study which may help clarify some of the remaining
unanswered questions.
4.4.1 Polarization dependence and symmetry adaptation

Polarization-dependent studies of, for example, the hypersensitive 7F 0-+
5D o,2 transitions of Eu 3 + in oriented single crystals will be able to shed light
on the true importance of the linear crystal field in hypersensitive lanthanide
transitions. The 7F 0 -+ 5Do transition in particular is unique. As discussed in
section 4.2.3, this transition is allowed in third-order theory only through
J-J mixing or if a linear crystal-field term is invoked. Since these two
mechanisms will have differing polarization dependences, intensity compari-
sons for differing polarizations will be able to determine the relative
importance of these two mechanisms for this transition. This has already
been done in two-photon absorption for Sm2+ doped BaCIF by Gacon et ai.
[61], where the magnitude of the linear term, represented by their A, was
found to be 1.57 times the magnitude of the J-J mixing term, represented
by their Jl.
In order to analyse the complicated polarization anisotropy of these
transitions, a site-symmetry adaptation formalism has been established for
both one- and two-photon transitions by Kibler and Gacon [62] and Kibler
[63,64]. These symmetry adaptation techniques have successfully explained
polarization dependent experimental results for two-photon transitions in
both transition metals [65] and lanthanides [61, 66, 67]. However, except
for the qualitative examination of Burdick and Downer [51], these powerful
symmetry adaptation techniques have yet to be applied to single-photon
transitions.
4.4.2 Closure approximation

A common criticism levelled against the Judd-Ofelt theory is with regard to
the closure approximation. In this approximation, various energy denomina-
tors which differ by as much as a factor of 2 are considered totally
degenerate, necessarily introducing some inaccuracy into the calculations. As
more and more complete calculations are done, which include all relevant
second- and higher-order terms, the limits on calculational accuracy may be
determined by the accuracy of the closure approximation. However, we do
not know the magnitude of error introduced by this approximation.
The numbers of levels in the ground and excited configurations of Pr3 +,
Nd 3 + and Tm3+ are low enough that calculation without invoking the
closure approximation might be feasible. However, for the other, often more
interesting trivalent lanthanides, literally millions of individual matrix el-
ements would have to be calculated, connecting hundreds of states in the
ground configuration with thousands of states in excited configurations.
Thus, except for a few selected cases, it is not possible to achieve quantitative
results without using the closure approximation.
However, as discussed in section 4.1.1, the approximation of closure
within the ground configuration need not be invoked in order to perform
the calculation. Considering only the excited configuration to be degenerate
requires the explicit calculation of energy denominators for each transition,
which introduces a frequency dependent factor C(v) into the calculations.
Odd-rank second- and third-order tensors also become formally allowed,
but they are not expected to provide significant line strength contributions.
The line strength equation (4.18) is replaced with,
where separate energy denominators Elli and En' are calculated explicitly in
C(v). Since this scaling factor reduces the range of energy denominators
which must be considered degenerate, it should decrease the amount of error
closure introduces into the calculation.
For observed transitions across the lanthanide series, C(v) is usually
within 10% of unity, and almost always within 20%. One significant
exception is the Nd3+ ion, where scaling factors greater than 40% are found
for selected observed transitions. When this scaling factor is introduced into
the fitting calculations for Nd3+, the Judd-Ofelt parameters must be
significantly revised in the second-order fitting as well as the fit with
third-order terms included. However, with this revision, standard deviations
of fits to experiment are found to be 4% worse than the corresponding
fittings done without scaling factors [36].
We conclude that, for most trivalent ions, the approximation of closure
over the ground configuration does not introduce significant error into the
References 129
calculations. In the case of Nd 3 +, where the closure approximation would
seem to introduce the greatest error, removing closure over the ground
configuration actually increases the error between fit and experiment,
suggesting that there might be other, as-yet unrecognized physics which is
included in the closure approximation.
Since removal of closure in the ground configuration does not have a
significant effect on fitted line strengths, we can suppose that closure of the
excited configurations will not produce a significant calculational error
either. However, explicit line strength calculations which do not invoke
closure of either the ground or the excited configurations will be necessary
before this truly can be known.
ACKNOWLEDGEMENTS
The authors wish to thank J.e. Gacon and M. Kibler for helpful discussion
during the writing of this chapter. Financial support by the Robert A. Welch
Foundation (Grant F-I038) and the National Science Foundation (Grant
DMR-8858388) is gratefully acknowledged. We also acknowledge the support
of the Ministere de l'Education Nationale (France) for one author (G.W.B.).
REFERENCES
1. van Vleck, J.H. (1937) J. Phys. Chem., 41, 67-80.
2. Judd, B.R. (1962) Phys. Rev., 127, 750-761.
3. Ofelt, G.S. (1962) J. Chem. Phys., 37, 511-520.
4. Peacock, R.D. (1975) Struct. Bonding, 22, 83-122.
5. Downer, M.e. (1989) In Laser Spectroscopy of Solids II (ed. W.M. Yen),
Springer, Berlin, pp. 29-75.
6. Axe, J.D., Jr. (1964) Phys. Rev., 136, A42-A45.
7. Dagenais, M., Downer, M., Neumann, R. and Bloembergen, N. (1981) Phys. Rev.
Lett., 46, 561-565.
8. Judd, B.R. and Pooler, D.R. (1982) J. Phys. C, 15, 591-598.
9. Downer, M.C., Bivas, A. and Bloembergen, N. (1982) Opt. Commun., 41,335-340.
10. Downer, M.e. and Bivas, A. (1983) Phys. Rev. B, 28, 3677-3696.
11. Downer, M.e., Cordero-Montalvo, e.D. and Crosswhite, H. (1983) Phys. Rev. B,
28, 4931-4943.
12. Kitai, A.H. (1992) In Visible Luminescence - Solid State Materials & Applications
(ed. A.H. Kitai), Chapman & Hall, London, Chapter 1.
13. Goppert-Mayer, M. (1931) Ann. Phyz. (Leipzig), 9, 273-294.
14. Dieke, G.H. and Crosswhite, H.M. (1963) Appl. Opt., 2, 675-686.
15. Carnall, W.T., Fields, P.R. and Wybourne, B.G. (1965) J. Chem. Phys., 42,
3797-3806.
16. Brewer, L. (1971) J. Opt. Soc. Am., 61, 1666-1682.
17. Nielson, e.W. and Koster, G.F. (1964) Spectroscopic Coefficients for the pN, dN
and fN Configurations, MIT Press, Cambridge, MA.
18. Camall, W.T., Crosswhite, H. and Crosswhite, H.M. (1977) Energy Level
Structure and Transition Probabilities of Trivalent Lanthanides in LaF 3, Argonne
National Laboratory Report.
19. Carnall, W.T., Fields, P.R and Rajnak, K. (1968) J. Chern. Phys., 49, 4412-4423.
20. Jankowski, K. and Smentek-Mielczarek, L. (1987) Mol. Phys., 60,1211-1219.
21. Smentek-Mielczarek, L. and Hess, B.A., Jr. (1987) Phys. Rev. B,36, 1811-1817.
22. Smentek-Mielczarek, L. and Hess, B.A., Jr. (1989) J. Chern. Phys., 90, 1390-1395.
23. Smentek-Mielczarek, L. and Hess, B.A., Jr. (1987) J. Chern. Phys., 87, 3509-3512.
24. Smentek-Mielczarek, L. and Hess, B.A., Jr. (1988) J. Chern. Phys., 89,703-710.
25. Smentek-Mielczarek, L. and Hess, B.A., Jr. (1989) Phys. Rev. B, 39, 517-522.
26. Jorgensen, e.K. and Judd, B.R. (1964) Mol. Phys., 8, 281-290.
27. Mason, S.F., Peacock, R.D. and Stewart, B. (1975) Mol. Phys., 30, 1829-1841.
28. Judd, B.R (1979) J. Chern. Phys., 70, 4830-4833.
29. Reid, M.F. and Richardson, F.S. (1983) J. Chern. Phys., 79, 5735-5742.
30. Reid, M.F., Dallara, J.J. and Richardson, F.S. (1983) J. Chern. Phys., 79,
5743-5751.
31. Reid, M.F. (1987) J. Chern. Phys., 87, 6388-6392.
32. Poon, U.M. and Newman, DJ. (1984) J. Phys. C., 17,4319-4325.
33. Henrie, D.E., Fellows, R.L. and Choppin, G.R. (1976) Coord. Chern. Rev., 18,
199-224.
34. Peacock, RD. (1977) Mol. Phys., 33, 1239-1246.
35. Reid, M.F. and Ng, B. (1989) Mol. Phys., 67, 407-415.
36. Burdick, G.W. (1991) Ph.D. Thesis, University of Texas, Austin.
37. Judd, B.R (1967) Second Quantization and Atornic Spectroscopy, Johns Hopkins,
Baltimore, MD.
38. Judd, B.R (1960) Operator Techniques in Atornic Spectroscopy, McGraw-Hill,
New York.
39. Rotenberg, M., Bivins, R, Metropolis, N. and Wooton, J.K., Jr. (1959) The 3-j
and 6-j syrnbols, Technology Press, Cambridge, MA.
41. Carnall, W.T., Fields, P.R. and Rajnak, K. (1968) J. Chern. Phys., 49, 4443-4446.
42. Carnall, W.T., Fields, P.R and Rajnak, K. (1968) J. Chern. Phys. 49, 4447-4449.
44. Judd, B.R. (1966) J. Chern. Phys., 44, 839-840.
45. Kiss, ZJ. and Weakliem, H.A. (1965) Phys. Rev. Lett., 15, 457-460.
46. Alam, A.S.M.M. and di Bartolo, B. (1967) J. Chern. Phys., 47, 3790-3801.
47. Blasse, G. and Bri!, A. (1966) Philips Res. Rep., 21, 368-378.
48. Nieuwpoort, W.e. and Blasse, G. (1966) Solid State Cornrnun., 4, 227-229.
49. Grenet, G., Kibler, M., Gros, A., Souillat, J.e. and Gacon, 1.e. (1980) Phys. Rev.
B, 22, 5052-5067.
50. Piriou, B., Fahmi, D., Dexpert-Ghys, 1., Taitai, A. and Lacout, 1.L. (1987) J.
Lurnin., 39, 97-103.
51. Burdick, G.W. and Downer, M.e. (1991) Eur. J. Solid State Inorg. Chern., 28,
217-220.
52. Holsa,1. and Porcher, P. (1985) J. Less-Cornrnon Met., 112, 121-125.
53. Becker, P.e., Edelstein, N., Judd, B.R, Leavitt, Re. and Lister, G.M.S. (1985) J.
Phys. C, 18, Ll063-Ll066.
54. Krupke, W.F. (1966) Phys. Rev., 145, 325-337.
55. Becker, P.e. (1986) Ph.D. Thesis, Lawrence Berkeley Laboratory, University of
California, Berkeley.
56. Alfrey, A.J., Stafsudd, O.M., Dunn, B. and Yang, D.L. (1988) J. Chern. Phys., 88,
707-716.
57. Smentek-Mielczarek, L. (1989) Phys. Rev. B, 40, 6499-6504.
58. Downer, M.e., Burdick, G.W. and Sardar, D.K. (1988) J. Chern. Phys., 89,
1787-1797.
References 131
59. Burdick, G.W., Downer, M.e. and Sardar, D.K. (1989) J. Chem. Phys., 91,
1511-1520.
60. Judd, B.R. (1988) In Handbook on the Physics and Chemistry of Rare Earths, Vol.
11, (eds K.A. Gschneidner, Jr, and L. Eyring), North-Holland, New York, pp.
81-195.
61. Gacon, le., Jacquier, B., Marcerou, J.F., Bouazaoui, M. and Kibler, M. (1990) J.
Lumin., 45, 162-164.
62. Kibler, M. and Gacon, J.e. (1989) Croat. Chem. Acta, 62, 783-797.
63. Kibler, M. (1991) In Symmetry and Structural Properties of Condensed Matter
(eds W. Florek, T. Lulek and M. Mucha), World Scientific, Singapore, pp.
237-250.
64. Kibler, M. (1992) In Excited States of Transition Elements (eds W. Strek, W.
Ryba-Romanowski and B. Jezowska-Trzebiatowska), World Scientific, Sin-
gapore, pp. 139-148.
65. Sztucki, J., Daoud, M. and Kibler, M. (1992) Phys. Rev. B, 45, 2023-2028.
66. Gacon, le., Marcerou, J.F., Bouazaoui, M., Jacquier, B. and Kibler, M. (1989)
Phys. Rev. B, 40, 2070-2075.
67. Gacon, le., Bouazaoui, M., Jacquier, B., Kibler, M., Boatner, L.A. and Ab-
raham, M.M. (1991) Eur. J. Solid State Inorg. Chem., 28, 113-116.
5
Thin film electroluminescence
G.O. Muller
5.1 THE NEW ELECTROLUMINESCENCE CONCEPT -

THIN FILMS
5.1.1 History
As easily imaginable, phosphorescence or afterglow was the first observation
of luminescence. People used to observe light from objects in nature as some
kind of re-emission or reflection in daylight or from incandescent or high
temperature sources, emitting thermal radiation. A cold object emitting in
the dark was completely different from this usual experience and of course
found great attention as early as 1603, when Vincenzo Cascariolo detected
the Bologna stones [1].
It was not until Destriau [2J reported on the influences of electric fields
on a luminescing material that electroluminescence (EL) was born. Later,
powder electroluminescence was something very common as a research and
development object over the years. The advent of semiconductor physics,
comprising single crystal growth, doping and p--n junctions brought along
the search for injection (electro-) luminescence. The concept of p--n junc-
tions, emitting minority carriers into the adjacent region of a crystal with a
different type of conductivity, where they recombine radiatively (in part), can
be modified. It has to be changed for materials which are hard to dope
ambivalently. Rather sophisticated structures have been invented and inves-
tigated to overcome these difficulties. The blue diode and the blue diode
laser, however, emerged from concentrated efforts to produce p-type con-
ductivity in ZnSe [3].
Thin film electroluminescence, after becoming successful, exhibited many
roots. Clearly identifiable is the effort at Bell Laboratories (see for instance
ref. 4, and papers cited therein) which was focused on phosphors deposited
in thin film form on anodized Ta. It had been anodized to prevent electric
breakdown when driven with voltages which would give rise to it otherwise.
133
134 Thin film electro luminescence
The paper by Chen and Krupka [4] marked the end of these activities at
that time, as the work was strongly believed not to be useful. The first
activities on thin film electroluminescence by VIa senko and Jaremko [5]
should also be mentioned. Some years later, Inoguchi and coworkers [6, 7]
gave thin film electroluminescence, now referred to in this way, a comeback.
They obtained excellent properties, and, what was the most impressive
feature, very long lifetimes. It was some years until the first results published
by the IBM group [8] convinced a broader scientific and engineering
community that thin film EL could be a solution for high field electro-
luminescence in general.
Less than 10 years after Inoguchi's first paper, his company, Sharp, went
on to industrial production of monochrome, yellow EL displays of more
than postcard size and a complexity of 256 by 512 picture elements (pixels).
In 1991 both at Sharp (Japan) and Planar (USA) large-scale production of
up to 19 inch diagonal displays with 1024 x 768 pixels of high contrast and
'crisp' aesthetic appearance was established, and the race for multicolour
and even full-colour displays continues.
5.1.2 Applicational aspects

In clear contrast to devices based on injection electroluminescence, which
conceptually are tied to the necessity to use single crystals or epitaxial
layers, thin film electroluminescence (TFEL) aims at large area devices for
the display of information.
An all-solid-state replacement of the cathode ray tube (CRT) was in mind
when people started to think about the use of thin films deposited on cheap
(transparent) substrates for electro luminescence. Transparent electrodes had
to be used, at least for one of the contacts to each picture element. The
addressing should be in a time-multiplexed form, as connections to each
pixel would have limited the complexity to some hundred pixels at most. A
'matrix' organization of the pixels in the form of crossings of two electrode
stripe systems, perpendicular to each other, on lower and upper sides of the
active films was the most common and versatile solution (Fig. 5.1). Address-
ing then should be a line at a time: a single pulse at each time selects the
line, which should activate the light information sent to all columns
simultaneously at that time. If the response of each pixel to the applied
voltage is steep enough, or, in other words, if the brightness dependence on
voltage shows threshold behaviour, a pulse slightly beneath threshold will
avoid cross-talk completely.
Thin films from the viewpoint of applications have the tremendous benefit
of being transparent, without the diffuse reflectivity of powders, so that the
contrast between lit and unlit pixels will be very high, even in light ambients.
Specular reflections from the glass can be overcome in conventional ways.
Of course brightness should be high, and efficiency too, as otherwise the
The new electro luminescence concept 135
Metal-_ _ _w::'!~
Insulator
Fig. 5.1 Schematic of the physical structure of a thin film EL display.
heat production required to achieve enough brightness could become

excessive.
Of course the period of useful life (maintenance) should be as high as
possible, but certainly not below some ten thousand hours.
Summing up, from the beginning a thin film electroluminescence device
in the form of a matrix-organized, line-at-a-time addressed display was
wanted, with the following properties:
1. a threshold in the brightness-voltage, B(V), characteristic;
2. high enough efficiency, to avoid safely any need for cooling;
3. long life.
Multi- or even full colour in the beginning was not an issue. Video speed as
well was not discussed, when the search for practical solutions started, but
it was guaranteed from the very nature of the materials.
5.1.3 Limitations from physics

Depositing a thin film of a luminescing material on a glass substrate, which
was covered with a transparent electrode, and putting a second contact on
top of the luminescent film, would result in the simplest TFELD structure
imaginable. Light emission under an applied d.c. voltage, however, would
not last very long, even if it occurred at all. More sophisticated arrange-
ments will be necessary as we shall see. However, first of all, some rather
general considerations will be provided.
Either the recombination of previously excited electron-hole pairs or the
radiative decay of a previously excited electron at a dopant (generally an
inner shell transition) gives rise to the emission. Clearly, the applica-
tion of voltage to any electroluminescent device must cause just this,
either generating electron-hole pairs (if one kind of carrier is present,
only minorities are needed), or exciting luminescence centres. As we
discuss EL, the excitation energy comes from the applied field and dis-
sipative current must flow. Excluding in this chapter the 'trick' of injec-
tion of minority carriers, the carriers forming the dissipative current
must be energetic enough to excite either the e-h pair or the centre. This
excitation is the Coulombic interaction of the energetic carrier with the
one to be excited, which we shall call 'Coulomb impact' or simply
'impact'.
The usual electrical conduction in semiconductors (or metals) does not
supply any kinetic energy comparable with the thermal energy kT to the
carriers - they stay 'cool'. In high fields, however, when the energy gained
from the field between two interactions with cooling media - phonons,
defects or other elementary excitations, or in general the 'lattice' - exceeds
the energy dissipated in this interaction, the carrier becomes 'hot'. Obviously
only hot carriers can excite (noninjection) EL.
To be more specific, as we are referring to visible emission mainly, the
excitation energy (in general at least as high as the energy of the emitted
photon) is of the order of 2.5 eV. So, in order to observe EL, there must be
carriers with about 2.5 eV of kinetic energy. That is far beyond the regime
usually addressed, if in semiconductor physics 'hot electrons' are identified.
Nobody until now has mentioned the possibility of electron energies of
25000 K in a material at room temperature.
However, as the balance between energy gain from the field and energy
loss to the lattice is a statistical process, it might happen that some of the
electrons are 'lucky' in gaining much higher momenta in the direction of the
field than the average, and the energy distribution is no longer a thermal
one, not even to a temperature higher than the lattice temperature. Run-
away electrons occur.
To rely on runaways, occurring eventually with a concentration of 10- 6 ,
would limit the efficiency to less than 10- 6 , as all the other cool carriers add
to the energy input the same amount, but nothing to the output.
Situations in which the fraction of hot electrons dominates are known,
however; for example, the avalanche breakdown regime. There electrons (or
holes) reach high enough energies to multiply, usually more than sufficient
to excite the dopants. So one possibility for efficient EL is operation under
avalanche conditions. Long life, however, does not seem to be guaranteed
for a device breaking down electrically.
One of the above-mentioned important properties, the threshold be-
haviour, however, would be realized in this avalanching device. As avalanch-
ing itself is a threshold process, a steep B(V) characteristic would be
assured.
Basics of thin film electro luminescence 137
5.2 BASICS OF THIN FILM ELECTROLUMINESCENCE
5.2.1 The mechanism
As outlined above, but understood only after the first devices were develop-
ed and waiting for production, noninjection or high field EL only has the
chance to be efficient if the field is high enough to generate plenty of hot
electrons.
Even an avalanching structure, however, does not necessarily luminesce
efficiently; suitable dopings have to be found which can readily be incorpor-
ated in such a way as to conserve their radiative efficiency.
The following section will therefore be split into these two ways of looking
at electroluminescence: electroluminescence and electro luminescence.
(a) The mechanism of electroluminescence

As is well known, avalanching can be stabilized or kept reversible by a series
resistance to limit the current. This concept usually is followed in powder
EL. It is not a resistor in this case, but a distributed resistance, formed by a
layer of semiconducting material, e.g. of MnO. The same concept could have
been followed in TFEL. Much more reliable, however, is a concept adopted
as early as 1967 [9]. Instead of controlling the current, and thereby the
power, a series capacitance can control the charge, and thereby the energy
dissipated in the device. The additional advantage is the fact that, in the case
of a local breakthrough, a resistive film exhibits the 'spreading resistance'
phenomenon, whereas a dielectric does not. The price to be paid is the
necessity to drive the device with bipolar voltages, as it becomes mainly
capacitive.
A schematic of the device structure and its band structure model are given
in Fig. 5.2. The generally used materials are also indicated. For the function
described above, one dielectric film (I) would do; we shall understand later
why the additional voltage requirement of a symmetric structure should be
tolerated. As mentioned, the device has to be operated with bipolar voltages.
In the case of matrices, pulses are used to select a row, and information
pulses are added in all columns the crossings of which with the said row are
to be lit. Sometime before the row is selected in the next frame, a pulse of
opposite polarity has to be applied; this is usually done in the interval
between frames.
The behaviour of the brightness and dissipative current under the action
of the voltage waveform shown, applied to an isolated pixel, is given in Fig.
5.3.
If, as indicated, a voltage above threshold is applied, electrons begin to be
tunnel emitted out of interface states between the near-cathode dielectric
and the active (or semiconductor) film as soon as the threshold field, Et , is
138 Thin film electroluminescence
. - - - - - - - -------{ u-r l) - - - - - - - - --,
5
ZnS
Insul. Semiconductor I Insul
S'S
CaS
Tunnel
em ission
excita tion
Mn 2
Tb3 +
Ce3+
Eu 2
Fig. 5.2 The 'simple' model of a TFEL device, showing the basic processes.
ZnS:: Mn
.,
OJ
~
o
H.
>
Time . 100 IJs, dlv
Fig. 5.3 Time behaviour of dissipative 'current' and 'brightness' under the action of
the applied 'voltage'.
reached. Tunnel emission is one of the most threshold-like processes in

physics. If, as wanted, the field Et is higher than the field necessary for
loss-free acceleration or avalanching, Ea , the emitted electron begins to
avalanche down to the momentary anode. Many electrons for each primary
will reach the near-anode interface, where they are trapped in the interface
states, and from which they are emitted after polarity reversal at Et (which is
not necessarily the same for each interface).
Below threshold the whole device is a loss-free capacitor, as in the
optimum case the S-film is highly resistive (> 10 1 0 Q cm). The applied
voltage is capacitively divided between the film capacitors. Above threshold,
as soon as current flow starts, the voltage partition changes; it increases over
the insulator films, and stays almost constant over the S film. The behaviour
can be modelled in a first approximation by the 'back-to-back Zener diode'
circuit shown in Fig. 5.4. The dissipative current ceases to flow because of
this redistribution of the voltage.
The time integral of the current density, the transferred charge density Q,
is a function of the capacitances of the films and the applied voltage
amplitude U only:
(5.1)
The current through the S film, and so the transferred charge in the film Q.,
is given by
To give a last important formula, the energy dissipated in the S film in one
pulse is related to the (outer) threshold voltage U th by
This is important for the determination of the efficiency [10].

The Zener model is a good first approximation. Going beyond, one has to
take into account the holes produced during multiplication of the electrons.
In the wide-gap II-VI compounds, used as material for the S films mostly,
these holes are efficiently trapped near their birthplace. Therefore they form
I
s
~M
Fig. 5.4 The so-called back-to-back Zener model of the TFELD.
a positive space charge, distorting the inner field, assumed to be constant, in

the S film. This distortion raises the cathode field at the expense of the
near-anode field. Thus tunnel emission will increase in spite of the voltage
across the S film's staying constant, or in other words a hysteresis in the
current-voltage characteristic can occur.
It is this effect that is measured. Figure 5.5 shows simultaneously measured
waveforms of the applied voltage V(t) and the dissipative current density D(t).
Besides them a curve F(t) [10, 11] is given, which was computed from the first
two, and the values C and Cs and the S film thickness ds = Csi PoP., with P
being the permittivity. The capacitance values were obtained by formula (5.1)
from a measurement of the transferred charge-voltage characteristic Q(U).
The equation used to compute F(t), which is the average field in the S film, is
obvious from the physical structure of the samples:
F(t) = [C;/(C j + Cs)] V(t) - (l/C;) f~ i(t') dt' + Vo (5.2)
(a) Current dens. 0, rnA cm- 2 voltage V,V
160 400
v
80 200
(b) i Field F, mV/cm
1.6 ..
F
0.8
o Time, f.lS
Fig. 5.5 (a) Simultaneously measured waveforms of applied voltage V and dissi-
pative current density D; (b) voltage across the active film, expressed in normalized
form as average field F.
Plotting D versus F, D(F), one has the current-voltage characteristic, in

reduced units, of the S film. It is shown for the same data as in Fig. 5.5
and in Fig. 5.6. The effect of the space charge is obvious and limits of
course the meaningfulness of an average value of the field, as eventually
strong deviations from the mean occur, and even vary with time. The latter
arises as the space charge from the previous half-period is destroyed at the
trailing edge of the current pulse near the momentary cathode, and later
on built up near the anode. If it is kept in mind that deviations from the
mean value F occur time dependently, plots such as the one in Fig. 5.6
give nevertheless the possibility of classifying samples, as will be shown
later on.
Current dens. 0, mA cm- 2
160
.A
80
a 1
Field, mV/cm 2
Fig. 5.6 A plot of the dissipative current D (from Fig. 5.5(a)) versus the average
field F gives the (normalized) current-voltage characteristic of the active film D(F)
[10,11].
Until now in this section luminescence or light emission has not been
mentioned. As it is the goal of the whole chapter, we shall concentrate on
this question separately (see section 5.2.l(c)). As outlined above, excitation of
centres can be by Coulomb impact or by recombination of the e-h pairs
generated during avalanching. The latter could proceed either directly, or
via excitons (free or bound), via trapping at suitable defects or impurities;
also, in any of those processes a photon could be generated or the energy
quantum-mechanically transferred to another transition, which decays radi-
atively or nonradiatively.
(b) Ballistic transport

In the preceding section we described a transport mode resulting in
avalanching, which was by no means well established when it was first
proposed, to account for the high EL efficiency of ZnS:Mn (see refs 12, 13
and, in more detail, 14).
A hand-waving argument for the 'loss-free acceleration' was given by

Mach [12]. As the main energy loss of accelerated electrons in materials
such as ZnS is to longitudinal optical (LO) phonons, the energy gain from
the field in the time between two LO emissions has to be compared with the
LO energy of about 30 me V. Assuming an effective mass of 0.3, this is about
1 eV, if a field of 1 MV cm -1 is assumed. It is a very rough approximation,
but a full Monte Carlo simulation, using full energy and angle dependence
and the best-known parameters of the Froehlich coupling [15], leads to the
same result. One point against this calculation, however, was very import-
ant: it neglected the band structure. Taking this into account Brennan [16]
came to the conclusion that ZnS is extremely unsuitable for EL, as almost
all electrons should stay below 2 eV, even at F = 2 MV cm -1.
An old argument against heating to more than about 2 eV [17] is the sign
reversal of the electron effective mass at about this energy, and another is
strong intervalley scattering between the X valleys as soon as they are
reached.
Analysing the efficiency quantitatively, and investigating photo-transport
under identical conditions as in TFELD, however, Miiller and Mach [18]
came again to the conclusion that the heating in fields above 1 MV cm -1 is
without any appreciable loss to phonons, and in the case of not too highly
doped ZnS proceeds until multiplication balances it at an energy above
7eV.
Within the model proposed by Mach, each of the electrons that tunnels
from the interface states into the ZnS conduction band at 106 V cm -1 gains
1 eV per 10 nm travelled, until it reaches the (steep) threshold for multiplica-
tion by excitation from the valence band, determined by energy and
momentum conservation (EM C). It will lose the energy (and momentum)
necessary to excite an electron from the valence band, remain hot by the
same EMC arguments, and start to accelerate once more. Again by EMC it
is not possible that the electron 'multiplied' from the valence band is cool.
As soon as it is in the conduction band, it will accelerate too, without loss.
The stationary electron energy distribution (EED) developing in this way
evidently contains a negligible fraction of cold carriers, i.e. it will be
completely athermal.
Recently, first experimental indications of the EEDs developed under
typical TFEL conditions were obtained by measuring EEDs of electrons
emitted through thin gold films into vacuum by using conventional retard-
ing grid analysers [19-21]. From well-established experimental data of
numerous investigators the gold film must be regarded as a high energy
cut-off filter, so very small attenuations in the low energy region are to be
expected between the outer and the inner EED (except below the difference
between the ZnS electron affinity and the Au work function (we denote this
difference by AW)), if the gold thickness is less than 5 nm. The inner EED
should therefore closely resemble the outer one displayed in Fig. 5.7, with
4
I
,,
N(E), a.u.
J\'I
ZnSe
ZnS
/\\
2 ~
~
\
'. \
~
a 0
./- I -R
/
/
I
~
I
a 2 4 6 8
E, eV
10 12 14
Fig. 5.7 Energy distributions of the electrons emitted into vacuum through the
thinnest gold film which could be made at present; the inner EEDs evolve by adding
the differences between the work function of gold and the electron affinity of the
active material ZnS or ZnSe. Note that the a.u. scales are not the same for the two
materials. (From ref. 20.)
the AWadded to the horizontal scale, shifting the distribution to even

higher energies.
This result answers one of the decisive questions concerning any dis-
cussion of EL efficiency: in a well-designed TFELD 'all electrons are hot'
enough to excite any imaginable luminescent centre.
TFEL research, on the other hand, has prompted the discovery of a new
high field transport mode in wide-gap semiconductors with medium Froeh-
lich coupling: in fields around 106 V cm -1, stationary, completely athermal
EEDs exist; they are spike like, and centred around the threshold of electron
multiplication from the valence band. The latter has been estimated from the
experimental results on vacuum emission to be about 12 eV for ZnS and
about 8 eV for ZnSe (Fig. 5.7).
(c) The mechanism of electroluminescence

Knowing now something about the electron transport in TFEL structures,
one should consider means to produce the luminescence, which the structure
was invented for.
The easiest way would be to dope the S film, which consists for example of
ZnS, with the well-known 'activators', e.g. Ag for a blue emission. If properly
done, one would observe a marvellous blue emission under UV or cathodo-
excitation, but only a faint white emission under application of a voltage,
which definitely causes avalanching. The reason is field quenching. This effect
is well-known from powder EL (e.g. see [22]). The mechanism is simply the
field ionization of the activators, which gives rise to another chance for
nonradiative recombination channels to catch the minority carrier.
So in all doping considerations for EL one has to be careful with the 'field
stability' of the wanted transition.
To begin with a very new, negative experience, we shall discuss one recent
attempt to use the well-behaving ZnS, if doped with Mn, for blue emission
by doping Te into it. It has been known for a long time that ZnS:Te emits in
the UV if lightly doped, in the blue if the Te concentration increases, and
eventually shifts to the green, if the concentration is raised further. Te,
iso-electronic to S, binds by short-range forces to excitons, which decay
highly radiatively, especially if bound to a Te pair by about 600 meV (the
blue line). As an exciton is a neutral entity, the risk of field ionization from a
centre binding it at 0.6 eV should be much less than, for example, for the
hole on the Ag acceptor.
After overcoming the problem of incorporating Te into ZnS properly in a
low temperature preparation (film deposition) process, the result was devas-
tating: well-photoluminescing samples showed no EL in the blue.
Also, looking for photoluminescence under applied voltages (photo-
electroluminescence (PEL)), even below the threshold for tunnel emission,
quenching became obvious (Fig. 5.8). Whether quenched directly by the field
or by impact of energetic carriers would have to be determined, but the
result is certain: no blue EL from ZnS:Te [23].
15000
photo- (PL)/electroluminescence (EL)
1-PL
10000
:J 2-EL
co 3,4-PEL
i-
'iii
c
Q)
E 5000
350 450 550 650

Wavelength, nm
Fig. 5.8 Photoluminescence (curve 1), electroluminescence (2), and photo-
electroluminescence (3,4) spectra of ZnS:Te; curve 4 was obtained with the threshold
voltage of EL applied, curve 3 at about one-half of that. (From ref. 23.)
Basics of thin film electroluminescence 145
The highest stability against field quenching can be expected from
dopants with inner shell transitions, as transition metals (TM) or rare
earths (Ln). The valence electron 3d 5 of Mn 2 + in ZnS remains at its ion in
the highest fields encountered, without shifting spectrally or changing its
transition probabilities. The same is true for the 4f electrons of the Ln ions
in ZnS or similar environments. So they appear to be the ideal centres for
EL.
As shown for Mn in ZnSe as early as 1978 [24], excitation of Mn occurs
by direct Coulomb impact. Whether there is transfer from the recombina-
tion of e-h pairs via for example the self-activated (SA) centre is an open
question. Direct impact can also be assumed for ZnS: Mn to be the
dominant excitation.
Surely the question arises as to whether a 5d -+4f transition of a rare earth
ion is an inner one in this sense. Rather serious doubts about the high field
stability of the Eu 2 + and Ce 3 + excitations in such alkaline-earth sulphides
(AESs) as CaS and SrS respectively have been raised lately by Ando and
Ono [25] and earlier by Yoshiyama et al. [26]. The idea is a field ionization
of the excited electron into the conduction band, where it would be
separated from the centre, drifting to the near-anode interface. Of course,
such a process could render these systems, which by now are the great hopes
for colour EL, inefficient.
The best-known and investigated example of TFEL in general is ZnS:Mn,
which we already mentioned many times. In the context of EL excitation we
refer to Fig. 5.9, which shows the time behaviour of the emission, or
brightness B(t) in relation to the dissipative current D(t) and the mean field
F(t). At a glance B(t) looks like the time integral of D(t), and so it is in a first
approximation, as the decay time of the Mn 2 + centre is of the order of
milliseconds, long compared with the exciting pulse. The excitation by the
Field F, mV/cm D, rnA cm- 2 B, a.u.
ZnS:Mn
1.6
O.B
o Time,l..Is 100
Fig. 5.9 Brightness B is roughly the time integral of the dissipative current D in a
low doped ZnS:Mn sample, as the decay time of luminescence is long compared
with the exciting current flow.
impacting electrons therefore is 'stored' in the centres, only few decaying

during the excitation pulse.
A completely different behaviour is seen in Fig. 5.10, which displays the
Ce 3 + emission in SrS and the Eu 2 +, evidently following closely the current
D(t). This is expected, as the lifetime of the excited Ce 3 + is of the order of
nanoseconds; on the time scale of the current pulse, the emission follows the
excitation instantaneously.
D, Acm-' D, Acm-' B, a.u.
0.12 0.3.
- - CaS: Eu
0.04
~t i;"
::
---SrS: Ce
. o. 1
1/
.':
....... ::. __ ..
a Time, IJS 100
Fig. 5.10 In contrast to ZnS:Mn, shown in Fig. 5.9, in the AES doped with 5d-+4f
lanthanides, exhibiting short decays, brightness B follows closely the exciting
current D.
From a user's point of view the time-averaged behaviour in general is more

interesting. Applying a bipolar voltage pulse of amplitude Vpeak to a sample,
the brightness-voltage characteristic B(V) will be obtained. In Fig. 5.11
brightness B is plotted on a (left-hand) logarithmic scale to show the threshold
behaviour more clearly. On the right-hand scale the same data can be read
from a linear scale, showing that grey scale is possible, simply by choosing the
proper information pulse amplitude, and adding it, as described in section
5.1.2 to the selected pulse, which could be 200 V in this case. Figure 5.11
reveals also the order of magnitude of the brightness to be expected at a drive
frequency of 50 Hz. In a very crude approximation brightness is proportional
to frequency, and this could be used to control brightness coarsely.
We can sum up our picture of thin film electroluminescence in its two
aspects as follows.
1. EL is nothing more than luminescence excited by the impact of energetic
carriers which were accelerated by the electric field. In special cases
energy transfer from the recombination of the simultaneously created
e-h pairs to the dopants might excite luminescence.
2. The presence of high fields and energetic carriers might cause 'field
quenching'; the choice of dopants (activators) therefore is limited. Inner
shell transitions are preferable.
Brightness, cd/m 2
3 r-------~-----------------,
10
ZnS:Mn 150
50Hz
100
50
230 260
Vpeak' V
Fig. 5.11 Brightness versus the peak voltage Vpeak of a bipolar pulse voltage of
50 Hz, on logarithmic and linear scales.
3. Thin film EL, because of its threshold-like brightness-voltage character-

istic, is ideally suited for high multiplex ratios or high row numbers in
matrix-organized displays.
5.2.2 Materials considerations

Discussing 'Basics .. .', this section is not meant to go into detailed consider-
ations about the merits and drawbacks of a possible materials list. However,
of course some general remarks about the selection of materials appear to be
useful.
(a) Host materials

The host material for the dopants, which will be discussed separately, must
provide the electrical properties necessary for the implementation of the
working principle. It must be capable of loss-free acceleration of carriers,
probably electrons, at fields the effects of which can be withstood by the
dielectric films. The last half of the preceding sentence means, by continuity
of the displacement the electric field in the insulator film is related to the one
in the S film by the ratio of the permittivities. This, or for safety at least twice
this, field must be withstood by the dielectric. This limits the choice of the
hosts severely. As yet, only the following representatives of this class of

materials are known: ZnS, ZnSe, CdS, SrS, CaS, and some of the mixtures of
the former [14]. All of them are chalcogenides. Any attempt to find loss-free
acceleration in oxygen-dominated compounds (up to 2 MV cm - 1) failed. It
is, however, by no means certain, that only chalcogenides are feasible.
Materials research is needed here. In fact, EL has very recently been
reported in Zn2Si04: Mn [27]. The big difference between this and the
chalcogenides, however, is the very low threshold, which may prevent
loss-free acceleration and may not allow high multiplex ratios.
Of course the host should be very resistive, as carrier concentrations of
10 12 cm- 3 already can reduce the efficiency.
A last (but not the least) consideration is the manufacturability of thin films
from the host material on cheap substrates, preferably glass. The deposition of
complex materials in low temperature processes is not always easy, as many of
them decompose in plasma-vacuum deposition environments.
(b) Dopants
The selection of dopants has been discussed before in section 5.2.l(c). It is
mainly the field quenching problem that governs decision making. However,
the solubility of the dopant in the host is of some importance too.
Consider the case of rare earths in ZnS. The solubility of most of them is in
the parts per million range. In low temperature deposition processes - electron
beam evaporation and d. sputtering have been mostly used - it is of course
possible to incorporate in some way Ln3+ ions and/or LnCl 3 molecules into
the film of ZnS, although whether into the lattice of the growing crystal grains
or to what other sites is questionable. However, even if we assume that the
incorporation on cation sites at least to some degree is achieved, the question
arises as to on what time scale the system, which is not in thermodynamic
equilibrium (TDE), travels under the action of energy input towards it. Also,
TDE is a phase mixture of ZnS doped to the solubility limit with the rare earth
and other phases containing more Ln, e.g. LnZZnS4.
The very big problem for TFEL, aiming at other or all colours, is the rather
limited choice of dopants able to dissolve in one of the known suitable hosts
and emitting a wanted colour.
The alkaline earth chalcogenides came into the play just from these
considerations - their solubility for lanthanides is orders of magnitude higher.
5.3 VALUES AND MARGINS
5.3.1 Efficiency
A check of the understanding of the basic physics - and the most important
issue for any application - is the modelling of efficiency. Efficiency will
Values and margins 149
usually be given in technical units (1m W- 1 ), as an energy efficiency (%), or

as a quantum efficiency (photons electron -1 em - 1), where in the quantum
efficiency the distance means the thickness of the S film, i.e. distance
travelled by the electron.
In an approximation which is good enough for practical purposes we
neglect temporal changes within an excitation pulse, and therefore do not
relate the emission or excitation to the momentary current or input power.
Instead we speak of the time-averaged brightness and the transferred charge
per pulse.
The brightness B gives us a light flux of cnB 1m (where c is a number
between 1 and 2 accounting for the angular distribution of the emission), a
threshold voltage Uth and a transferred charge (outer) Q (A s cm - 2) results
in the input energy per pulse of QU th. So efficiency n is given by
n=2nB/2fQU th
This efficiency should be modelled from our understanding of the physics.

The processes described above should be split into those which can be
regarded as independent of each other, to allow us to multiply their
probabilities in order to end up with an overall probability. This can be
done in the following way [13]:
1. nhot designates the fraction of the transferred charge consisting of
electrons hot enough to excite;
2. nexc is the probability of an electron hot enough to excite a centre;
3. nlum is the probability of an excited centre to decay radiatively;
4. n out is the probability of a generated photon to escape from the sample.
As at least in a very good approximation these processes are not influencing
each other,
As was shown above, in a well-designed sample nhot = 1. Under the same

precondition there is no limitation of excitation from the side of the carrier,
and the excitation probability is given by the chance to meet a centre,
nexc = sN, with N being the concentration of the centres and s their
cross-section for excitation.
The outcoupling efficiency n out of a photon from a transparent material
with the refractive index ns is given by classical optics to be
n out = [1-(1- n,;-2)].
In the case of the most useful materials the refractive index is about ns = 2.3,
and n out =0.1. Light guiding in the highly refracting material is responsible
for this dramatic loss of usefulness. (Interference by the necessary (insulator)
or additional films may change the angular distribution (differently for
different wavelengths), but cannot improve the net outcoupling.) With these
results inserted, we have
n=O.lsNnlum
To measure nl um for ZnS:Mn is no problem. The radiative lifetime trad is
well known [28]. Also, from a comparison of the decay of brightness after a
sufficiently short excitation, b(t), with the purely radiative decay time trad ,
the efficiency nl um follows directly [29] and without any need for absolute
light measurements, which are tedious:
nl um = I
Loo b(t) dt Loo exp( - t/t rad ) dt
= f" b(t) dt/trad
The only assumption made is that the starting conditions of the real decay
and the fictitious purely radiative decay are the same; this is true if the
excitation period is short compared with the time in which a decay of for
example 5% (to 95%) of the initial brightness occurs.
N has to be determined by a depth profiling, for instance SIMS, analysis,
to be sure about homogeneous doping. Also, by measuring n for a sample
for which one is sure that the outcoupling efficiency nOU! = 0.1, the cross-
section s can be determined, and holds for all samples with just this dopant.
This investigation has been conducted for a variety of preparation
methods, and the results are summarized in Fig. 5.12 [30,31]. Figure 5.l2(a)
shows the measured values of n and Fig. 5.l2(b) the nl um values, measured in
the way described; Fig. 5.12(c) resulted from the multiplication of the nl um
values by sN, with s known from a previous determination, to give an inner
efficiency nin!. Dividing the experimental values n by nin! of Fig. 5.12(c), we
obtain the nOU! values displayed in Fig. 5.12(d). Consistency is seen from the
fact that 0.1 is a lower margin for nOU! (except two samples at high dopant
concentrations, in which case the uncertainties are at a maximum).
Higher values of nOU! than 0.1 have to be explained, however. The classic
optics formula holds for a situation of a perfectly plane surface of the
emitting material. In fact the surface of the polycrystalline film in general
will consist of growth faces inclined against the average surface plane.
Therefore scattering will occur in addition to the refraction and total
reflection, adding to nOU! = 0.1. In fact, on probing surface roughness for
instance by surface plasmon coupling [32], a qualitative correlation is found
between nOU! > 0.1 and the expected dip in plasmon reflection [33]. A more
quantitative comparison of the values from Fig. 5.l2(d) with independently
determined values from brightness profiles [30] did not go beyond qualita-
tive agreement.
n.lm/W
lD'~~-----------------r--------~
I: a '".~
r- x
+
+
.. +
+ +
+
EBE1,HT
+ ElSE 1, L T
THE1
a THE:!
~ ALII:
X RFS
D.l 1 lD
(a) Mn-eoneentration.1020 em-'
nl um
1
+
~ X
D.l~ ..... . ..................................
+ .
I- +
+
(b)
... x
lD ~ .................. ~ ... .d otlI~" O'.;l< ...... :
+ +
+
1 ~.
(e)
nout
1 I:
~ ........................................... .
I- a ~ ~
JC +
D.l ~ ..... . + X. +
'\0 :.. '.' :
I-
D.l 1 lD
(d) Mn-coneentration.10Z"cm-
Fig. 5.12 (a) The EL efficiency n as a function of the Mn concentration measured
on samples made by different deposition methods: EBE, electron beam evaporation;
HT, LT, high, low substrate temperature; THE, thermal evaporation; RFS, radio
frequency sputtering. The efficiency has been decomposed into (b) nl um , (c) ninl and (d)
noul' The symbols are the same in all parts ofthe figure, showing clearly that there are
improvements possible for all of them [30].
More difficult to analyse certainly are these partial efficiencies in AESs,
especially the luminescence efficiency nlum , as not only are the radiative
decay times of the prominent dopants not known yet, but the decays are
much faster (in the nanosecond range), so that measuring is much more
complicated [34].
The physics behind the measured nlum(N) is 'concentration quenching'.
Assuming some defects (impurities or native defects) to act as killers of
luminescence by giving rise to nonradiative transitions, there will be more
chance for an excitation to reach these killers if the centres are highly
concentrated, so that by transfer from the excited via unexcited ones the
killer can be reached during one radiative lifetime. Also, as the probability
for quantum-mechanical transfer goes steeply up with decreasing distance R,
e.g. for dipole-dipole interaction as R- 6 , this means that there exists a
certain concentration where quenching starts suddenly. Of course it will
depend on the killer concentration too, but if the latter is low, the dopant
concentration dominates the behaviour. As known from the classical papers
by Dexter [35], the coefficient in the transfer probability can be approxi-
mated by the spectral overlap between absorption and emission spectra of
the transferring centres, with of course the oscillator strengths determining
the absolute value. So transfer should be more likely for the 5d -+4f dopants
Eu2+ and Ce3+, used mainly in the AESs. However, the radiative lifetimes
scale down in the same way, so that predictions about the concentration
quenching are not simple from hand-waving arguments.
Even if the concentration dependence of nlum is known, the optimum
dopant concentration N opl depends on the absolute value of the excitation
cross-section s, as is evident from the inner efficiency ninl of Fig. 5.12(c), and
manifested for convenience schematically in Fig. 5.13 again.
Concluding, the maximum efficiency attainable in electroluminescence
critically depends on the concentration quenching behaviour of the chosen
dopant in a given host, and its excitation cross-section for the energetic
electrons in that host lattice. This is true for the inner efficiency ninl. A bonus
pertaining to the real outer efficiency can be achieved by introducing
scattering into the optical system, e.g. by surface roughness.
To give a numerical example, which is not far from reality (ZnS:Mn), we
assume, that at a dopant concentration N = 2 X 1020 em - 3 the luminescence
efficiency nlum =0.4; the cross-section ofthis dopant we set to 4 x 10- 16 em 2
This gives us an inner efficiency ninl = 0.4 x 4 x 2 X 104 em -1 = 3.2 ~m -1.
This means that, on its way through the active film of say 1 ~m thickness,
each electron gives rise to the generation of 3.2 photons.
The energy efficiency we obtain by accounting for the photon energy of
2.1 eV, and the fact that the electron travels in a field of 1.5 x 106 V em -1,
which results in 3.2 x 2.1 eV/(e x 1.5 x 106 x 10- 4 V)=4.48 X 10- 2 or 4.5%.
Knowing that the luminous equivalent in the ZnS:Mn emission region is
590 1m W- 1 , the inner luminous efficiency is 4.5% of that, or 26.5 1m W- 1 ,
Values and margins 153
.1 20
m 2
Fig. 5.13 Schematic of the factorization of the efficiency: the shaded area of n)um
indicates the variations obtained experimentally using the different technologies; the
limiting curves of ninl reflect this variation.
or, what is really meaningful for the user, there is an (outer) efficiency of
10.5 1m W - 1, if nOUI = 0.4, which is not unrealistic.
Another way of speaking is the following. A factor of 0.4 is lost be-
cause of the luminescence efficiency; at the assumed concentration and
cross-section, the average impact length is (SN)-1 = 12.5 X 10- 6 cm- 1,
but the electron gains enough energy to excite (2.1 eV) on la =
2.1/1.5 X 10 6 V cm -1 = 1.4 X 10- 6 cm; this accounts for a loss factor of
1.4/12.5=0.112, and these two multiply to 0.112 x 0.4=0.0448. In words, at
the optimum dopant concentration the luminescence efficiency is relatively
high (40%), but this concentration is so small that only 11 % of the hot
electrons are used for excitation, and the others miss their targets.
So what is wanted is a dopant which has a higher cross-section, or shows
concentration quenching not earlier than N'=I/s1 a =1.8xl0 21 cm- 3 . As
there is no general expression available for the concentration quenching
nlum(N), no optimum condition can be generally formulated.
In practice, however, things are even more complicated. An example again
is ZnS:Mn. In this case a rather strong nonlinear interaction in the dopant
system occurs. Incidentally, Mn shows cross-relaxation in the sense of a
transfer of excitation energy from an excited Mn to an excited neighbouring
Mn. The last one is excited to the second excited state, which is degenerate
with the interband continuum of the host, and so two excitations are lost by
reabsorption of this 4.3 eV photon. The transfer probability of this process is
some orders of magnitude larger than the one for transfer from an excited to
an unexcited Mn 2 + (the transition from the first to the second excited state
is dipole allowed), so it is a rather effective nonlinear loss, which shows up
already at moderate excitation levels. As is obvious from the description, it
will depend on the dopant concentration itself, and is less pronounced far
from the optimum doping [36].
5.3.2 Brightness
From a user's point of view the efficiency of a display may be only relevant if
it is so low that heating becomes a limitation. If this is not the case, mainly
the brightness (and contrast) will be an essential specification. As is evident
from the understanding of the mechanism, brightness is roughly propor-
tional to the drive frequency, and to the transferred charge Q. The latter
depends on the drive voltage, or more accurately on the voltage above
threshold. It has already been mentioned that driving a matrix display
usually means that a select pulse is applied to a row at a time, and
simultaneously information pulses of opposite polarity (in order to add at
the pixel) are supplied to all columns that should light when they cross the
selected row. As the latter voltage is limited by the capability of the column
driver, very often in the literature the brightness at some voltage above
threshold (mostly at 30 or 40 V above) is specified. This is important for the
designer of drive circuitry, but almost irrelevant for the physics. In section
5.2.1(a) the charge-voltage characteristic was discussed, and from that it is
very obvious that the brightness B(U th +40V)=nQ(Uth +40V) depends not
only on the efficiency n but strongly on the capacitance ratio C;/( C j + C.),
and on C j itself too. So the capacitances are important design criteria.
Talking about values and margins, for display purposes at mains fre-
quency (50 or 60Hz) about lOOcdm- 2 are wanted, but 60cdm- 2 can be
sold, if the contrast is good [37]. So we have to discuss the limits imposed
by physics on the charge and/or the voltage (independent of the availability
of driver ICs).
The avalanching of the active film clamps the voltage across it, and all
amplitude above threshold stresses the insulator films. So ultimately their
breakdown field limits the brightness. Looking closer, it turns out [38] that
the storable charge on the dielectric films - EoEjF b - is the margin of the
transferrable charge (the E are the permittivities of vacuum (subscript 0) and
insulator (subscript i), and Fb is its breakdown field strength). Maximum
safe values of 6 x 10 - 6 A s cm - 2 have been reported.
So, if there is a principal limitation in Q, the brightness can be easily
increased by raising the drive frequency. 30000 cd m - 2 can be reached safely
by devices driven at 8 kHz [37]. This is not possible, of course, if a complex
display has to be addressed, as even only 200 rows would decrease the select
pulse width to 0.6 J,1S, and with the high capacitive load of a 200-line display
(of reasonable size) this is impossible to achieve. However, applications of
electroluminescence are imaginable for which those brightness levels are
mandatory.
Acknowledgement 155
5.4 OUTLOOK
Thin film electroluminescence is still far from having reached a textbook
level of comprehension. The mechanism appears to be understood on the
basis of a linearized concept, not taking into account, for instance, the
consequences of avalanching for inhomogeneities in carrier concentrations,
space charge distributions and recombination destroying inhomogeneous
space charge in every half-period of the drive voltage. Progress in this
direction might be expected from the evaluation of time-dependent electron
energy distributions, measured in vacuum emission.
Almost nothing is known about the energetic location of the dopant
ground and excited states with respect to the energy bands of the host
crystals. Strong correlation effects have to be expected. It is this feature
which makes any consideration of interactions between a dopant's electrons
and free carriers in the bands very complicated. Especially in the case of the
alkaline earth cha1cogenides, the liberation of electrons from the excited
state of Ce 3 + and Eu 2 + has been discussed [25, 26] which could give rise to
more complicated temperature dependences of (effective) impact excitation
cross-sections of these dopants.
One of the most challenging deficiencies of relevant knowledge concerns
the palette of materials available for EL from the viewpoint of loss-free
acceleration. This special high field transport mode up to now has only been
established in lIB and IIA cha1cogenides, but it would be surprising if
additional materials do not join this class.
Concerning the luminescence processes, the concentration quenching,
discussed mainly in connection with laser materials, enters the limelight
again. A thorough understanding, by simulation for instance, of the complex
interplay of various transfer processes, linear and nonlinear, is urgently
needed. It is not unlikely that, similarly to lamp phosphors, sensitizers could
bring about improvements of the use of available electronic impact power.
As mentioned above, the balance between excitation and quenching by field
and/or impact has in no instance been thoroughly investigated. Also, the
partition of relevant processes into laminates of the active film [39, 40] either
differently doped or composed of different materials has only been touched on.
Summing up, one can state that the first steps in the understanding of thin
film electroluminescence have been taken, but the whole field is not far
beyond infancy - and this in view of a tremendous application potential,
ranking from very simple indicators, displays for computers, monochrome
and colour, up to high definition television.
ACKNOWLEDGEMENT
Thorough and stimulating discussions with Dr R. Mach are gratefully
acknowledged.
REFERENCES
1. Queisser, H.J. (1981) J. Lumin., 24, 3-10.
2. Destriau, G. (1936) J. Chim. Phys., 33, 587-625.
3. DePuydt, 1.M., Haase, M.A., Qiu, 1. and Cheng, H. (1992) In Proc. 5th Int. II-VI
Con!, Tamano, 1991, in J. Cryst. Growth, 117, 1087.
4. Chen, Y.S. and Krupka, D.C. (1972) J. Appl. Phys., 43, 4089.
5. Vlasenko, N.A. and Jaremko, A.M. (1965) Opt. Spektrosk., 8, 39.
6. Inoguchi, T., Takeda, M., Kakihara,Y., Nakata, Y. and Yoshida, M. (1974) In
SID 74 Digest, pp. 86-87.
7. Inoguchi, T. and Mito, S. (1977) In Topics in Applied Physics, 17, Springer,
Heidelberg, p. 222.
8. Riihle, W., Marello, V. and Onton, A. (1979) J. Lumin., 18--19, 729-738.
9. Russ, MJ. and Kennedy, D.I. (1967) J. Electrochem. Soc., 114, 1066.
10. Miiller, G.O., Mach, R, Selle, B. and Schulz, G. (1988) Phys. Status Solidi A, 110,
657-669.
11. Mach, R. and Miiller, G.O. (1991) Semicond. Sci. Techno/., 6, 305-323.
12. Mach, Rand Miiller, G.O. (1982) Phys. Status Solidi A, 69, 11-66.
13. Mach, R. and Miiller, G.O. (1984) Phys. Status Solidi A, 81, 609.
14. Mach, Rand Miiller, G.O. (1989) In Proc. 4th Int. Con! on II-VI Compounds,
1989, in J. Cryst. Growth, 91, 967.
15. Fitting, HJ. and Czarnovski, A. (1986) Phys. Status Solidi A, 93, 385.
16. Brennan, K. (1988) J. Appl. Phys., 64(8), 4024-4030.
17. Allen, J. (1984) In Proc. ICL 1984, in J. Lumin., 31-32, 665-670.
18. Miiller, G.O. and Mach, R. (1986) In Proc. 20th ICPS, 1986, pp. 1587-1590.
19. Miiller, G.O., Mach, R., Halden, E. and Fitting, H.-J. (1990) Proc. 20th ICPS,
World Scientific Press, Singapore, pp. 2510-2513.
20. Miiller, G.O., Mach, R., Reinsperger, G.-v. and Schulz, G. (1991) In Proc. IDRC
91, San Diego, CA, IEEE, Piscataway, New Jersey, p. 17.
21. Miiller, G.O., Mach, R, Reinsperger, G.v., Halden, E. and Schulz, G. (1992) In
Proc. 5th II-VI Con!, Tamano, 1991, in J. Cryst. Growth, 117,948.
22. Gumlich, H.J. (1970) Energy transport in Electroluminescence and Electro-
photoluminescence of II-VI Devices, Sammlung Vieweg.
23. Mach, R., Reinsperger, G.-V., Miiller, G.O., Selle, B. and Matzkeit, G. (1992) In
Proc. 5th Int. II-VI Co/if., Tamano, 1991, in J. Cryst. Growth, 117, 1002.
24. Mach, R, Gericke, W., Treptow, H. and Ludwig, W. (1978) Phys. Status Solidi A,
49,667.
25. Ando, M. and Ono, Y.A. (1991) J. App/. Phys., 69(10), 7225-7230.
26. Yoshiyama, H., Sohn, S.H., Tanaka, S. and Kobayashi, H. (1989) In Electro-
luminescence, Springer Proceedings in Physics, Vol. 38, Springer, Berlin, pp.
48-55.
27. Minami, T., Miyata, T., Takata, S. and Fukuda, I. (1992) In SID Int. Symp.,
Boston, MA, May 17-22, 1992.
28. Busse, W., Gumlich, H.E., Meissner, B. and Theis, D. (1976) J. Lumin., 12-13,
693.
29. Benalloul, P., Benoit, 1., Mach, R., Miiller, G.O. and Reinsperger, G.-v. (1990) J.
Cryst. Growth, 101, 989.
30. Mach, R, Miiller, G.O., Schrottke, L., Benalloul, P. and Benoit, 1. (1990) Acta
Poly tech. Scand., Appl. Phys. Ser., 170, 161.
31. Mach, R, Miiller, G.O., Reinsperger, G.-v., Benalloul, P., Benoit, 1. and
Ohnishi, H. (1990) In SID 90 Digest, pp. 238-241.
32. Harbeke, G. (1985) In Polycrystalline Semiconductors, Springer Series on Solid
State Sciences, Vol. 57, Springer, Berlin, pp. 156-169.
References 157
33. Mach, R., Schrottke, L., Milller, G.O., Reetz, R., Krause, E. and Hildisch, H.
(1988) J. Lumin., 40-41, 779-781.
34. Hilttl, B., Rinck, W., Milller, G.O. and Mach, R. (1982) Proc. Workshop on EL
1992, El Paso, Cinco Puntos Press, EI Paso, Texas, USA (1990) p. 123.
35. Dexter, D.L. (1953) J. Chern. Phys., 21, 836.
36. Milller, G.O., Neugebauer, J., Mach, R. and Reinsperger, G.-v. (1988) J. Cryst.
Growth 86, 890-894.
37. Mach, R. (1993) In Solid State Luminescence (ed. A.H. Kitai), Chapman & Hall,
London, Chapter 7.
38. Howard, W.E. (1977) IEEE Trans. Electron Devices, 24, 903-908.
39. Suyama, T., Okamoto, K. and Hamakawa, Y. (1982) Appl. Phys. Lett., 41, 462.
40. Herrmann, R., Milller, G.O. and Selle, B. (1990) Acta Poly tech. Scand., Appl.
Phys. Ser., 170, 171-179.
6
Powder electroluminescence
Surjit S. Chadha
6.1 INTRODUCTION
This chapter will survey the phenomenon of electroluminescence (EL) in
powder phosphors. Electroluminescence is luminescence produced by the
application of an electric field to a dielectric phosphor; it is also termed
electroftuorescence. Reviews of the AC powder electroluminescence (AC-
PEL) phenomenon as discovered by Destriau [1] in 1936 and of the DC
powder electroluminescence (DCPEL) as promoted by Vecht [2] will be
presented. As far as ACPEL is concerned, the period of 1936-1963 was
reviewed by Ivey [3] and Henisch [4]. Therefore, in this chapter, apart from
historical references and comparisons with the DCPEL devices, the review
of the ACPEL cells will be limited to any progress since 1963 noting major
investigations such as Fischer's [5] and a few others.
The organization of this chapter is in 11 sections. Sections 6.2 and 6.3
consist of historical perspectives of both AC and DCPEL structures re-
spectively. Section 6.4 describes the powder phosphors, their preparation
techniques, the role of the luminescent centres in zinc sulphide (ZnS) with
special reference to copper (Cu) and manganese (Mn) and the Cu coating
process that is essential for DCPEL devices. The possible presence of
copper sulphide phases on the Cu-coated ZnS phosphor is also discussed
in section 6.4. Section 6.5 covers the fabrication of ACPEL and DCPEL
devices. Section 6.6 discusses the forming process that is unique to DCPEL
devices. The possible excitation and injection mechanisms occurring in
ZnS phosphors are covered in section 6.7 with the optoelectrical responses
of both AC and DC excited powder EL devices in section 6.8. The driving
schemes employed for the AC and DC structures are described in section
6.9 and the maintenance characteristics in section 6.10. Finally section 6.11
illustrates some of the typical usage of both AC and DC powder EL
devices.
We begin with a brief history of the ACPEL cells.
159
160 Powder electroluminescence
6.2 ALTERNATING CURRENT ELECTROLUMINESCENCE
IN POWDER PHOSPORS: HISTORICAL
As stated above, Destriau [1] discovered EL in Zn:SCu powder phosphors
in 1936. The structure of his original EL cell is shown in Fig. 6.1.
Figure 6.2 shows a modified version of the original Destriau cell. Al-
though very high voltage (15 kV) AC excitation was used, the luminance was
extremely poor. Thus, even by 1950, Leverenz [6] questioned whether the
Oil
mercury
ZnS
Fig. 6.1 The original Destriau cell.
Salt water
Mica
Phosphor in castor oil
Metal
Fig. 6.2 A schematic diagram of a modified Destriau cell (no transparent conduct-
ing electrodes available then).
AC electro luminescence in powder phosphors 161
light emission process was in fact due to the excitation of the ZnS:Cu
powder phosphor by AC fields. According to him, the phosphor may in fact
be excited by the UV light emitted by the electrical breakdown of the gases
in the fairly porous powder cells. Consequently the effect may be merely
photoluminescence (PL).
The real impetus for EL displays was provided by the arrival of the
transparent conducting electrodes (TCE) based on tin oxide (Sn02) in the
late 1940s. All devices until then had been rather fragile and as a result
precluded industrial interest.
The 1950s saw massive academic and industrial research effort being
devoted to the 'source de lumiere du futur' as Destriau had claimed. The
many attractive features claimed for this new possible light source included
flatness, low power dissipation, multicolour emission and the lack of
catastrophic failure. Among possibilities suggested. for these devices, a TV
screen was prominent. Unlike the bulky and cumbersome TV screens based
on the cathode ray tube (CRT), these slim, lightweight units could have been
hung on the wall, much like a painting.
The cells developed in the early 1950s were made by using a suitably
doped ZnS powder (5-20Ilm grains) suspended in a dielectric (binder),
sandwiched between two electrodes, one of which was transparent, and
supported by a substrate. The construction is essentially that of a flat-plate
capacitor in which the phosphor is the dielectric medium. The structure has
changed very little over this time (Figs 6.3 and 6.4). The substrate may be
glass or flexible plastic or may be metallic, in which case the top electrode
must be transparent. The dielectric binder may be organic, e.g. cyanoethyl
cellulose with a plastic or glass substrate or a low melting glass with a metal
substrate. Usually a secondary white dielectric layer is used as extra
protection against catastrophic breakdown and to reflect the light emitted
through the transparent layer.
Protective glaze
Tin oxide
ZnS:Cu in enamel
Reflector layer
Grip coat
Metal electrode
Fig. 6.3 Structure of a typical ACPEL ceramic cell (rigid).

Plastic
Indium tin oxide
ZsS:Cu in high
dielectric constant
binder
Reflector layer (BaTi0 3

in binder)
Aluminium foil
Fig. 6.4 Structure of typical ACPEL plastic cell (flexible).
The investigations into the phenomenon of ACPEL were conducted

worldwide, with major efforts being devoted in the USA, USSR, UK,
France, Germany, Italy and Japan. The names of Thornton [7-10], Leh-
mann [11], Larach and Shrader [12], Zalm et al., [13, 14], Vlasenko and
Popkov [15], Bowtell and Bate [16], Homer et al. [17], Fischer [5], Busca
et al. [18] and Yoshiyama et al. [19] are all associated with the intensive
studies of the time, some of them investigating materials, others engaged in
empirical and fundamental work and some making fairly complex devices.
Some of the most important fundamental work was done by Fischer [5].
However, despite many improvements in the technology originally intended
for use as room lighting (at least according to the popular press of the time), the
fact remained that, at high brightness levels, the ACPEL cells exhibited poor
life. Mter 1963, the research into ACPEL was severely curtailed. As mentioned
in section 6.1, this period of research on EL has been reviewed by Henisch [3]
and Ivey [4]. Both these books contain extensive bibliographies.
Nonetheless, the promises of the original concept still made the cells an
attractive proposition. These include the facts that such light sources are not
prone to catastrophic failure; they can be cheaply produced from readily
prepared ZnS powders; they are very compact, they have low power
requirements and so do not dissipate large quantities of heat.
Various other modifications of the basic structure have been built though
cells incorporating CdS thin film transistors (TFTs) were probably the most
technically advanced (Fig. 6.5). The reasons for this structure will be
described in section 6.8.1.
6.3 DIRECT CURRENT ELECTROLUMINESCENCE IN

ZnS POWDER PHOSPHORS: HISTORICAL
Zalm et al. [14] reported the earliest observations of direct current electro-
luminescence (DCPEL) in binderless powder phosphors in 1954. This was
DC electro luminescence in powder phosphors 163
black inslating
resin
black back electrode

(carbon) covering
high-dielectric resin,
optionally dyed with
fluorescent
pigment
EL-monograin layer
fl.l.l .l.lX
on thermoplastic
.. ~.~
. : ~., resin, optionally
dyed with
W fluorescent
pigment
vacuum-deposited
AI 20 3 insulator
5000 A thick
glass substrate with

etched-out In203
conducting
transparent numeral
and TFT -addressing
Fig. 6.5 Use of TFT drive for ACPEL cells as proposed by Fischer [5].
followed in 1955 by the observations of EL in DC excited CRT phosphors

by Nicoll and Kazan [20] and by Bowtell and Bate [16]. The earlier work
of Homer et al. [17] had identified that EL by DC excitation would require
a very conductive surface layer, possibly copper sulphide, CU2 -xS. This
layer would provide the medium for current flow and for injection into the
phosphor particles.
Short-lived cells were investigated by Kazankin et al. [21] in 1959 using
ZnS:Mn,Cu powder phosphor suspended in castor oil. Normally the fired-in
concentration of copper, [Cu], was much more than its solubility limit in
164 Powder electro luminescence
ZnS. As a result, a considerable amount of Cu was precipitated on the
surface during the cooling stage. For ACPEL applications, this excess Cu
was removed using a cyanide wash. However, for DCPEL applications, this
surface Cu provided the necessary conducting CU2-xS phase. Favorin et al.
[22] extended this work in 1959. They reported results from demountable
cells in which castor oil was the medium holding the phosphor particles in
suspension. These cells exhibited half-lives of only a few minutes. Under DC
excitation, only the Mn ions emitted efficiently.
In 1961, Favorin and Kozina [23] reported results on a device structure
not too dissimilar from those being made now. Thus the phosphor was
dispersed in a resin mixture and the back electrode was thermally evapor-
ated AI. The types of decay modes (later to be suggested by Alder et al. [24]),
the excitation mechanisms, etc., were discussed, but no quantitative lumi-
nance data were presented.
However, the systematic and sustained study of DCPEL in ZnS:Mn,Cu
phosphors had to await the efforts of Vecht and coworkers [25-58]
including all those who worked with Vecht either at Thames Polytechnic or
at Phosphor Products Company, etc., beginning in 1966. Not surprisingly,
following the commercial failure of Destriau's AC powder cells, the early
reception to DC powder devices was not very enthusiastic. In spite of these
early misgivings, these efforts eventually led to the production of devices
consisting of 640 x 256 individually addressable pixels [56] emitting 25-30
foot-Iamberts (ft-L) and which can display 2000 characters. Figure 6.6 shows
a schematic diagram of a DCPEL device capable of displaying alpha-
numeric and graphical information.
Kirton's group at the Royal Signal and Radar Establishment (RSRE),
now Defence Research Agency, undertook some very useful fundamental
Phosphor
Layer
Glass
",-~---
Substrate
Fig. 6.6 Schematic diagram of a DCPEL phosphor device capable of showing

alphanumeric and graphical information.
Powder phosphors 165
studies [59-68] in the period 1970--1982 in conjunction with Vecht's group

at Thames Polytechnic. These included early studies of the forming process
(section 6.6), time-resolved spectroscopy of the Mn 2 + luminescent centre
(section 6.4.3(d, the early drive schemes (section 6.9.2) and the initial
description of the failure modes (section 6.10.2).
In the UK, apart from Phosphor Products Company (PPC), research into
and pilot plant productions of DCPEL devices were also attempted by
General Electric Company (GEC) [69] and Smiths Industries (SI) [70, 71].
Whilst GEC was involved with the fabrication of alphanumeric displays, SI
was interested in manufacturing displays for the automotive market, i.e. car
dashboards.
Limited research into DCPEL devices was conducted in USA in the early
1970s by Fickiet and Plumb [72] and Hanak and Yocom [73]. In the last 6
years, Cherry Electrical [74, 75] has invested heavily in the DCPEL
technology and succeeded in making panels with 640 x 480 pixels and with
16 levels of grey scale.
In Japan, Matsushita Electric Industrial Co. conducted comprehensive
studies starting in 1970 [76] but appeared to have given up in 1973 [77, 78]
(section 6.9.2). They succeeded in manufacturing a prototype display with
50 176 elements, capable of reproducing TV images. However, the
luminance levels were low at 5-10 ft-L and the maintenance was poor. The
only other Japanese firm that appeared to have shown any interest in
DCPEL was Sony [79].
The Chinese [80--84] attempted to stabilize DCPEL devices by AC
driving techniques. They reported the manufacture of a 1 m 2 mosaic device
for displaying Chinese characters.
The only other place where a passing interest was shown in DCPEL
displays was in France at the Centre National d'Etudes des Telecom-
munications (CNET) by Abdalla's team [85-88]. They studied the initial
forming process and the degradation phenomenon.
In summary, DCPEL in ZnS:Mn, Cu phosphors has been studied in many
countries, with a majority of the academic research being carried out in the
UK. However, because of limited efforts (both manpower and financial), only a
few commercially useful devices have been produced (sections 6.6 and 6.10.2).
In the subsequent pages, the preparation of the powder EL phosphors, the
luminescent centres, the fabrication of the EL devices, the optoelectrical
characteristics, the luminescent mechanisms, the drive circuitry and the
maintenance problem will be reviewed. For DCPEL, the forming stage
necessary for the light emission process will be reviewed in depth.
6.4 POWDER PHOSPHORS

Most powder EL research has been centred around the II-VI compounds
and by far the most important EL lattice is zinc sulphide (ZnS). This exists
in two main structural modifications, the low temperature form which is

cubic (sphalerite) and the high temperature form, which is hexagonal
(wurtzite). The two modifications are shown in Fig. 6.7.
(a) Sphalerite (b) Wurtzite
Fig. 6.7 (a) Cubic (sphalerite) and (b) hexagonal (wurtzite) modifications of zinc
sulphide.
Table 6.1 Some of the powder phosphors known to exhibit EL

Phosphor Excitation Colour Reference
ZnS:Cu,CI(Br,I) AC Blue [89,90]
ZnS:Cu,CI(Br,I) AC Green [89,90]
ZnS:Mn,CI AC Yellow [91]
ZnS:Mn,Cu,CI AC and DC Yellow [8]
ZnSe:Cu,CI AC and DC Yellow [8]
ZnSSe:Cu,CI AC and DC Yellow [8]
ZnCdS:Mn,CI AC Yellow [89]
(Cu)
ZnCdS:Ag,CI AC Blue [89]
(Au)
ZnS:Cu,AI AC Blue [92]
Table 6.1 shows some of the binary and ternary systems that have been
investigated over the years together with the activator(s) used for ACPEL.
Some alkaline earth based phosphors that have been found to be useful for
DCPEL are given in Table 6.2. The maximum luminance under continuous
and pulse DC condition is also given together with the power efficiency
(section 6.4.3(f).
6.4.1 Preparation of pure II-VI compounds

The techniques used for the preparation of AC and DC powder phosphors
are described below. The recipes originally developed for the preparation of
Table 6.2 DC EL characteristics of some alkaline earth sulphide phosphors [45]
Phosphor Continuous Pulsed
system DC DC
(emission colour)
Luminance ft-L '1 (%) Luminance ft-L '1 (%)
CaS:Mn (green) 150 0.032 32 0.03

CaS:Eu,Cl (red) 5.9 0.001 2.l 0.001
Srs:Ce,Cl (blue-green) 100 0.Q2 26 0.011
SrS:Ce (green) 73 0.015 14 0.024
the CRT phosphors were modified for the manufacture of the ACPEL
phosphors. For DCPEL, with its inherent requirement that particles should
be small and of uniform size, newer preparative techniques had to be
developed. Most luminescent and optoelectronic II-VI materials are pre-
pared by the addition of controlled amounts of specific impurities (dopants),
called 'activators' and 'co-activators', to a high purity II-VI compound.
In order to prepare high purity starting materials, one of the following
methods could be used.
1. Direct reaction of the group II and VI elements.
2. Precipitation from aqueous group II salt solutions by gaseous hydrogen
sulphide, selenide or telluride. Another method uses the reaction between
group II salt in aqueous medium with, for example, thiourea or
thioacetamide that provides the sulphur. The reaction yields high purity
ZnS.
3. Thermal decomposition of suitable organometallic compounds such as
zinc or cadmium diethyldithiocarbamates.
Although totally feasible, the direct reaction of pure elemental Sand Zn,
for example, is not recommended, since the reaction is difficult to control
and is explosive especially if the reagents are in the form of finely divided
powders. The thermal decomposition of organometallic compounds leads to
the production of toxic organic residues and traces of carbon in the product.
As a consequence, this method is also not used for the production of powder
II-VI compounds (however, this method has been used successfully for the
deposition of thin films of ZnS [93]). Therefore the precipitation methods
are the most widely used for the preparation of powder II-VI phosphors.
Some of these precipitation methods will now be described in more detail.
6.4.2 Precipitation methods for II-VI compounds

(a) ACPEL phosphors
Leverenz [6] described suitable methods for the preparation of ZnS contain-
ing less than 10 - 5 wt. % of metallic impurity as even small concentrations
of heavy metals can greatly influence the luminescence characteristics of these

phosphors. ZnS can be precipitated using a purified solution of a Zn salt and
hydrogen or ammonium sulphide in alkaline or acidic conditions. The phos-
phor made using this technique contains only a few parts per million of heavy
metal impurity. The firing of the dried precipitate in a nitrogen atmosphere
reduces the non-metal impurity level to about 500 ppm. The firing also leads to
an increase in the particle size and an improved crystallinity of the lattice.
Brown [94] studied the particle size and morphology of ZnS precipitated
from Zn 2 + salt solutions by hydrogen sulphide. He found that the pH of the
solution was the most important factor in determining the final particle size.
A wide range of particle sizes resulted from precipitation at pH < 1, when
many larger particles were formed (spread 0.1-5 J.Illl) whereas, for pH> 1, the
majority of the particles were small, (spread 0.Q1-5 J.lm). Typical morphol-
ogy and particle size distribution of the phosphors formed by these methods
are shown in Fig. 6.8.
(b) DCPEL phosphors

As mentioned earlier, two basic requirements that must be met by a DCPEL
phosphor are that the particles should be small and of uniform size. The
Fig. 6.8 Typical size and morphology of powder ACPEL phosphor.

techniques described above for the preparation of phosphors for ACPEL
produce large variations from particle to particle, both in the activator
concentration and in the particle size (typically 2-25 J.lm). Such variations in
phosphor particle size are unsuitable for DCPEL application since the
current density will vary from particle to particle, resulting in short circuits.
The device may burn out in some regions because of this and so the
luminous output will fall. Although milling (grinding) is often used for the
reduction of the particle size of a number of CRT-type phosphors, this
option is not available for the sulphide phosphors as is clearly demonstrated
in Fig. 6.9. Hence ACPEL phosphors cannot be milled to produce DCPEL
usable phosphors.
~~-----------------.-----------------.
I
50 ................................................................................1.................................................................................. ..
~
!
40 ................................................................................1...................................................................................
g
3. ------------ -----+--------------- i
.~ 20 ................................................................................ !. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.3 10 ...................................................................................;I...............................................................................
.
!
O+-____~~~~~~~i----~~~~~~~
!
10 100 1000
Log,o Milling time (mins)
Fig. 6.9 The reduction in ACPEL output at identical drive conditions for ZnS:Cu
phosphors milled over a range of time.
Fortunately, both these requirements are achievable by the method of

homogeneous precipitation first reported by Willard [95]. This results in a
small spread of particle size since reactants are placed together in solution
under such conditions that no precipitation takes place. One component is
then generated by a change in the pH or the temperature and nucleation
takes place throughout the body of the solution. Owing to an absence of a
concentration gradient, the particle size of the precipitate formed under
these conditions is uniform, and can be controlled by the rate of nucleation
and reaction.
In 1967 P.A. Jackson [96] precipitated ZnS using organic thio reagents.
Ethyl dithiocarbamate, thioacetamide and thiourea led to successful precipi-
tation of ZnS. These organic reagents are easily purified by simple physical
methods such as distillation and recrystallization to remove any inorganic
contaminants. Typical morphology and particle size distribution of ZnS

prepared using this method is shown in the SEM micrograph in Fig. 6.10. It
is obvious that the particles are much smaller and spherical compared with
those formed by the reaction of group II salt with gaseous hydrogen
sulphide. The particle size distribution is also shown in a histogram (Fig.
6.11) which shows a very small spread.
Fig. 6.10 SEM micrograph of a typical DCPEL phosphor.
6.4.3 Activators (dopants)

Pure, stoichiometric ZnS does not luminesce. Luminescence can be caused
by the introduction of lattice defects, by deviation from the stoichiometric
Zn-S ratio, and by the addition of certain atoms to the system. Often, when
impurity atoms are introduced, defects and deviations from stoichiometry
are simultaneously produced, either by the rearrangement of the crystal
lattice or by the physical and chemical processing conditions under which
the impurity is added. Impurity atoms giving rise to luminescent response
are called 'activators'. Typical examples of activators are Mn and Cu. There
are other impurities which can lead to the reduction of the EL phenomenon
and these are termed 'killers' or 'quenchers'. Thus, for example, Co, Fe and
Ni, are well-known quenchers in CRT-type phosphors.
% 50
0.1 1 10 100
Particle size ().1m)
Fig. 6.11 The particle size distribution of DCPEL phosphor showing a very
narrow spread.
(a) The doping of ACPEL phosphors

In normal ACPEL phosphor manufacture, the incorporation of the impur-
ity atoms is carried out by adding the dopants either in a slurry form or by
milling (dry or wet). If the addition is carried out by slurrying or wet milling,
the mixture has to be dried out completely before firing at elevated
temperatures to assist the diffusion of the dopants into the particles. For
ACPEL phosphors, fluxes such as MgO, BaCI 2 , and NaCI are also added to
lower the activation temperature. Elements from group IB or VB (ac-
tivators) are normally introduced with group IIIB or group VIIB impurities
(co-activators). The prime reason for the addition of two or more types of
impurities simultaneously is related to charge compensation in the II-VI
phosphor lattice. In AC powder phosphors, Cu from group IB is the most
important activator giving rise to both green and blue EL. Halide co-
activators are normally employed, with CI- being preferred for high
luminance (section 6.4.3(b)).
(b) The doping of DCPEL phosphors

For uniform DCPEL response, each phosphor particle should contain the
same relative concentrations of Cu and Mn. This is difficult to achieve using
conventional slurrying techniques mentioned above. This is mainly because,
although the Cu + ions diffuse readily into the ZnS, the diffusion rate of the
Mn 2 + ions is much lower. To achieve a more uniform distribution of Cu
and Mn, a technique called 'simultaneous activation' is used. In this
technique, the Cu and Mn are added to the Zn 2 + solution just before the
precipitation of the pure ZnS, thus achieving an even dopant distribution

from particle to particle.
For maximum efficiency and brightness, the concentration of manganese
[MnJ incorporated should be greater than 0.3% by weight [27J. Low [CuJ
0.05% by weight) gives rise to high quantum efficiencies, but low
brightness levels, and high [CuJ gives rise to brighter phosphors, but with
reduced efficiencies. Since Cl- ions act as donors in ZnS, high concen-
trations result in higher current flow, leading to an increased luminance
level. However, the efficiency decreases exponentially with CI- content
greater than 70 ppm. Furthermore, since the amount of non-metallic impur-
ity retained in the fired material is unpredictable, incorporation of CI- ions
can lead to phosphor performance variations. As a result, it is normally
recommended that CI- should not be used in any DCPEL phosphor.
Figures 6.12 and 6.13 show the effect of Cl- on the luminance at constant
/
voltage and constant current respectively.
7
1
- 005:
0.3% Mn ./
,
:- 10.1
-
:::J
c.
:::J
o
/
Q)
g 10.2
co
c
E
:::J
...J
10.3 L..-1...-.L.-J'---1---L---L--'---'---'
o 40 80 120 160
CI content (p.p.m.)
Fig. 6.12 The effect of Cion EL output at constant voltage [29].
In ZnS, the most important luminescent centres are due to Mn and Cu. In
DC PEL, Mn emits a bright yellow-orange peak centred around 585 nm.
The Cu centre is only efficient under AC excitation, emitting either green or
blue, the colour being dependent on [CuJ and the preparative techniques
used. Both of these luminescent centres are discussed below.
(c) Copper in ACPEL phosphors

The earliest EL phosphor was ZnS doped with Cu. To this day, this remains
the most important ACPEL phosphor. Of course, significant progress has
been made in improving luminance, efficiency and stability over these years.
Both blue and green colours are available from the Cu centre. The
10
0.3% Mn
0.05% Cu
--
()JA cm- 2 )
- ..............
-:';:~,
: :I
C.
: :I
00.1
Q)
()
C
<1l
C
E
::I
...J
.,
o
o 40
CI content (p.p.m.)
80 120
""
160
Fig. 6.13 The effect of CIon EL output at constant current [29].
preparation conditions and the dopant concentration are obviously different

in these phosphors. Although the solubility limit of Cu in single crystal ZnS
is low [97], for the observation of any ACPEL [CU]lat had to be well in
excess of its solubility limit. Early experiments showed that the solubility of
Cu could be enhanced by the addition of halides before the firing process.
These materials therefore became known as co-activators (section 6.4.3(a)).
Other cations such as AI3+, In3+ and Ga3+ also helped to increase the
solubility of Cu + in the ZnS lattice. The effect of triply charged AI3+, In3+
and Ga 3+, etc. has been interpreted on the basis of their acting as
compensating donors for the singly charged Cu + ions. The atomic radii of
both Al (0.143 nm) and Cu (0.1275 nm) are not too different from that of Zn
(0.138 nm), so the lattice strains introduced by substitution of these atoms
for Zn are not excessive. Thus one Cu + and one AP + ion, can for example,
occupy two Zn2+ substitutional sites. The Cu + ion can also be present at
the interstitial sites. According to Aven and Parodi [98], [Cu]lat>0.05% by
weight promotes the growth of the cubic form of ZnS. They used a small
quantity of Mn, insufficient to influence the phase transition, to study the
change in phase by measuring its electron paramagnetic resonance (EPR)
spectrum. They found that the 30 lines normally observed in hexagonal ZnS
decreased to 6 equally spaced lines, indicating a phase transition to cubic.
Aven and Parodi [98] postulated that the energy released during the
precipitation of excess CU2-xS at concentrations in excess of 0.05% by
weight was sufficient to initiate the transformation. Nickerson et al. [99],
and later Gushurov and Banks [100], found that cuprous sulphide initiated
a phase transformation when in contact with hexagonal ZnS. This trans-
formation was rapid above temperatures of 800 0c.
As noted above, Cu emits in both the green and the blue part of the visible
spectrum. The colours can also be varied by using organic dyes to absorb
the emitted radiation and to re-emit at a lower wavelength. Some colour
variation can be achieved by altering the excitation frequency. The spectra
of Cu blue and Cu green phosphors, together with the change in the
spectrum, are shown in Figs 6.14 and 6.15 respectively.
100r---------------~----------------------~
so
o~----~----~----~------~----~----~~
200 300 400 SOO 600 700 800
Wavelength (nm)
Fig. 6.14 Typical emission spectra of the blue emitting ZnS:Cu ACPEL phosphor
(conditions: 60 Hz, 200 V, and 400 Hz, 200 V).
loor-------------------~~------------------~
so
o~----~----~~--~
200 300 400 Soo

____ __ ____
~~
600
~
700
~~
800
Wavelength (nm)
Fig. 6.15 Typical emission spectra of the green emitting ZnS:Cu ACPEL phosphor
(conditions: 60 Hz, 200 V, and 400 Hz, 200 V).
(d) The role of the manganese

Mn present as a substitutional impurity is responsible for emission in several
EL materials. In ZnS, Mn 2 + substitutes directly for Zn 2 +. This impurity
system possesses deep-lying levels that can be described approximately by
the states of the free divalent ion. Figure 6.16 shows the approximate energy
level diagram for Mn2+ in ZnS.
Regardless of the ionic or the covalent character of the binding crystal, the
transition responsible for emission in these materials involves states with the
3d 5 electronic configuration. The ground state is the 6S state; the excited
state may be approximated by the 4G state, although coupling of the
angular momentum of the free ion with the crystal field reduces the
significance of the angular momentum quantum number of the free ion. In
many phosphors the Mn activator is effectively excited by resonance
o - Vacuum level
Conduction band
5
Valence band
20
25
Fig. 6.16 Energy levels of the Mn2+ ion.
interaction with other previously excited impurity systems. Since the excited
state responsible for emission is only 2.1 eV above the ground state, and the
band gap of ZnS is 3.7 eV, the activator system in ZnS:Mn is stable against
field ionization in applied fields as great as 10 7 V cm - 1.
In ZnS, because of its 'characteristic' nature, Mn2+ produces only one
emission band with its peak centred at "" 585 nm (Fig. 6.17). (Characteristic
activators luminesce by means of electronic transitions within closed in-
terelectronic shells. Ionization is not involved. This should be contrasted
with fundamental activators such as Cu. These give rise to new energy levels
between the conduction band and the valence band and are capable of being
ionized.) This is very close to the maximum sensitivity of the human eye. The
modes of excitation or the presence of any co-activator has little effect on the
emission spectrum of the Mn2+ centres. In fact, as Fig. 6.18 shows, changes
in the host lattice, CaS:Mn [101], produce only minor shifts in the emission
peak wavelength.
Mn is very soluble in ZnS. Nitsche [102] published a phase diagram for
the ZnS: MnS system. Three points of interest are as follows.
1. The phase transition,
zinc blende (sphalerite) Wurtzite

Cubic 1020C Hexagonal
is affected by the presence of Mn. With increasing [Mn] the transition

temperature is lowered.
2. Mn forms a continuous range of solid solutions with ZnS-Znl_xMnxS.
x can be up to 0.4 according to Juze et al [103].
16
12
-<
~
r:::
o
o
-as .
....
.......
r:::
.gjo
/
ew 4
o /"
,,
0.50 0.55 0.60 0.65

Wavelength (!lm)
Fig. 6.17 Mn emission spectra [29].
CaS:Mn,Sc,Mg
E ::::I
>-E
.~ Ox
-
en co
5ir::: _E 50
.- 0
Q)
.~
co u
5i
-
QiQ;
0::0.
o ~~ ______ ~~~ __ ~ ____ ~~ ______ ~ __ ~
400 500 600 700 800

Wavelength (nm)
Fig. 6.18 Mn emission spectra in CaS:Mn [101].
3. Beyond x = 0.4 MnS separates out as an individual phase. However,

these concentrations are far above those used in the preparation of
luminescent materials.
Although it is very soluble in ZnS, Mn 2 + does not change the electri-
cal properties of the host lattice significantly [29J, even at several per
cent concentration. This is because both the Zn and the Mn ions are
isovalent.
Figures 6.19 and 6.20 show the effect of [Mn] on luminance at constant
current and constant voltage respectively.
-::J
C.
~ 10.1
(I)
u
c
ca
c
'E
::J
....I
0.2 0.4 0.6 0.8

Mn (%)
Fig. 6.19 The effect of [Mn] on the luminance characteristics at constant current
[29].
1
0.045% eu
....--.-
400 v
-::J
C.
S 10.1
o
~
c
ca
c
'E
/'/-
/.
/' 300 v
::J
....I
0.2 0.4 0.6 0.8

Mn (%)
Fig. 6.20 The effect of [Mn] on the luminance characteristics at constant voltage
[29].
As shown, the quantum efficiency increases with increasing [Mn], and

then levels off at about 0.8 %. With [Mn] > 1%, the luminance decreases
owing to concentration quenching. This occurs because, at high [Mn] levels,
the luminescent centres are in close proximity with one another, leading to
the absorption of the emission from de-exciting centres by those ions in the
ground state.
The subject of luminance saturation observed in Mn-doped phosphor has
also been commented on by other authors [104, 105]. The effect has been
observed in powders and thin films under both AC and DC excitation. In
particular, Yamamoto and Yamazoe [77] and Skolnick [67] studied the
effect of [Mn] on the luminance decay time of DCPEL phosphors. Accord-
ing to Yamamoto and Yamazoe, the decay time was nearly proportional to
the concentration of Mn, i.e.
roc [Mn2+r 1/3 (6.1)
where r is the decay time constant. This is shown in Fig. 6.21.
10
-E
......., 5
Q)
E
..;::;
>-
ctI
~ 2
o
0.05 0.1 0.2 0.5 1.0

Fig. 6.21 Effect of Mn concentration on the decay time [77].
Skolnick [67] reported that the decay times at [Mn] > 0.01 % were
non-exponential and became increasingly so for higher currents. In other
words, the decay curve was composed of two or possibly more components.
The luminance intensity, I(r), could be best described by the equation
(6.2)
where tl and t2 are the decay time constants of the isolated Mn2+
luminescent centres and Mn 2 + pairs respectively. The averaged dominant
decay times of Mn 2 + are given in Table 6.3.
In summary, Mn 2 + is the most efficient EL activator in ZnS with a spectrum
output that is very close to the maximum sensitivity of the human eye and is
almost unaffected by changes in the host lattice. However, despite its high
efficiency, it suffers from saturation effects due to quenching by Mn 2 + pairs.
This quenching places limits on achieving extremely high brightness devices.
(e) The role of the copper in DCP EL

Apart from being a dopant that can emit green and blue EL under AC
excitation [89], the Cu centre has also been indicated as taking part in an
Table 6.3 Effect of Mn concentration on the
decay time characteristics [67J
Mn concentration '["
(%) (ms)
0.01 1.8 0.1

0.05 1.50.1
0.1 1.50.1
0.4 1.0 0.1
1 0.40.05
2 0.19 0.03
energy transfer mechanism to Mn2+ [106] (sensitizer) and as providing the

surface conduction necessary for DCPEL [107] via some CU2-xS phase(s).
Being p type, CU2-xS is also said to form a p-njunction at the CU2-xS-ZnS
interface and thus possibly to inject electrons into the formed region,
resulting in the emission oflight from the Mn2+ ions.
In this review, the role of Cu will be limited to ZnS powders excited by
DC voltages only. This will include the effect of Cu in enhancing the
DCPEL emission from the Mn 2+ centres, the Cu coating process that is
absolutely necessary in providing the surface conduction for DCPEL and
the possible CU2 _xS phases.
( f) Copper as a dopant in DCP EL

Cu is not a very efficient luminescent centre in DCPEL. In fact, blue EL is
only visible when pulsed rather than continuous DC excitation is used.
However, the presence of Cu appears to increase the device maintenance.
The effect of [CU]lat on DCPEL characteristics has been given by Vecht et
al. [29] (Figs 6.22 and 6.23).
As shown in Fig. 6.22, when [CU]lat was increased from 0.05% to 0.25%
(i.e. from its solubility limit and above, where the excess Cu must reside in
grain boundaries, stacking faults, surface, etc.) the quantum efficiency, at a
constant current of 10 mA cm - 2, decreased from 17% to only 0.03 %.
However, the luminance achieved at a given voltage rose markedly with the
increase in [CU]lat. Thus, at 300 V, the luminance increased by just over
three orders of magnitude.
The effect of [CU]lat on the power efficiency '1 (%) was determined [108]
using the formula [29]
100BA
(6.3)
'1 370VID
where B is the brightness (ft-L), A the area of a pixel (cm2), V the applied
voltage, I the current (mA), D the duty cycle for pulsed displays and 370 a
constant (370 ft-L= 1 mW cm- 2).
10r-------------~
0.25% Mn
:::J 1
t...
:::J
Q.
\.
~ 10-1
Q)
(,)
c:
co
c:
e 10-2
:::J
...J
10-3 '--......L.._'--......L.._'--......
o 0.1 0.2
Cu (%)
Fig. 6.22 The effect of [Cu].a. on the luminance characteristics at constant current
[29].
10
...
:::J
Q.
~ 10-1
Q)
(,)
c:
co
c:
e 10-2
:::J
...J
o 0.1 0.2 0.3

Cu (%)
Fig. 6.23 The effect of [Cu] on the luminance at constant voltage [29].
It appears that the power efficiency is optimum for a [eu].a! of around

0.032% by weight (Fig. 6.24). The increase in '1 up to [eU]la! 0.032% is
probably because of the resonance transfer of energy from the eu centres to
the Mn centres. The decrease above 0.032% may be due to the leakage paths
introduced by the eu precipitating out in the grain boundaries, defects,
stacking faults and dislocations, etc.
0.4 -r--------------------,

>
u
c:
Q)
.(3
~ 0.2
0.1 -+-----r-----.------,-----,.-----,r---~
0.00 0.01 0.02 0.03 0.04 0.05 0.06
Lattice copper concentration %[w/w]
Fig. 6.24 The effect of [CU]lat on the power efficiency of DCPEL devices [108].
(g) The role of the copper in surface conduction in DCPEL cells

Homer et al. [17] realized that phosphors for DCPEL would require an
electrically conducting phase. Furthermore, the phosphor particles must be
in actual physical contact with one another. Thus Zalm [14] observed DC
emission from binderless layers and produced a separate conducting phase
using lithium sulphide, Li 2 S. Most other workers employed standard AC
powder phosphors with excess fired-in Cu, some of which usually precipi-
tates on the ZnS surface (as well as in the grain boundaries, defects, stacking
faults and dislocations, etc.), providing fairly conductive paths. For AC
applications, this surface layer is removed by washing in either alkaline
sodium or potassium cyanide solution. Similarly Thornton [109] reported a
DC response from conventional AC powders in plastic dispersions with very
high phosphor-binder ratios so that interparticle contact was assured.
Nicoll and Kazan [20] observed DC emission from CRT phosphors excited
by very high fields. Bowtell and Bate [16] also observed DC emission from
cathodoluminescent phosphors that had been treated with a solution of an
ionizable salt such as potassium silicate.
However, none of these was satisfactory for providing high brightness at
low enough voltages. To provide surface high conductivity, a Cu coating
process was developed. This involved the replacement of the Zn on the ZnS
surface by the Cu + ions. The reaction is very rapid and the colour of the
ZnS powder changes from almost white to grey-green. A similar reaction is
also used in the production of photovoltaic cells based on II-VI compounds
such as CdS and Cd1-xZnxS.
The replacement reaction of Zn by Cu on single-crystal ZnS was studied

using radioactive tracers [110]. The replacement reaction occurs since Cu is
lower in the electrochemical series than Zn. The following reaction was given:
ZnS+Cu 2+.-. CuS+Zn 2+ (6.4)
It is interesting that the formation of CU2S was not considered, even
though it was more probable. Figure 6.25 shows the exchange reaction
between the Cu2+ ions in the solution and the ZnS surface, leading to the
deposition of a very conductive CU2-xS layer.
Cu++ SOURCE
Zn++ Zn++
2Cu+
+
~ ,-. --~~ -1---' --------- ~l~~ -~.- ----
ZnS BASE
Fig. 6.25 The exchange reaction between the ZnS surface and the Cu2+ ions in the
solution.
For ZnS powder phosphors, however, Smith [111] found that the
replacement reaction was only possible when excess surface Zn was present.
Thus phosphors fired in a nitrogen-sulphur atmosphere were very difficult
to coat with Cu whereas phosphors fired in excess Zn, were coated easily
with Cu. ZnS powder phosphor is normally fired in air and, owing to the
greater volatility of S and its high reactivity with oxygen, the surface is
expected to be Zn rich (because of the presence of only a limited amount of
oxygen in a closed crucible). Some ZnO is also formed and this is
subsequently removed by a dilute acetic acid wash. Smith gave the following
reaction equations after discovering that the amount of Cu taken up by the
phosphor almost equalled the amount of Zn found in the solution:
ZnO+2Cu 2+.-. Zn 2+ +2Cu+ (6.5)
The cuprous ions then react with the ZnS to give cuprous sulphide:
ZnS+2Cu+.-. Zn 2+ +CU2S (6.6)
Overall, the reaction can be written as:
Zno + ZnS + 2Cu 2+ .-. 2Zn2 + + CU2S (6.7)
Smith also calculated the dimensions of a uniformly distributed layer of

CU2 _ x S on top of the ZnS surface assuming the powder particles to be 1 11m
in diameter and using an experimentally determined Cu uptake figure of
3.175x10- 3 gCu(gZnS)-1. He found that the thickness of the CU2-xS
layer would be 0.5 nm. This low thickness would, of course, make the
existence of any CU2-xS phase(s) implausible. However, Smith realized that
the main error was in the surface area calculation because some particles
tend to form aggregates leading to no Cu uptake on the 'hidden' surfaces.
He therefore went on to state that as a result, at certain regions, the
thickness may be 'several hundred angstroms'. Therefore the possibility of
CU2 _xS phases on the ZnS surface remains.
The Cu uptake is also a function of the surface area and the coating
solution temperature [108] as shown in Fig. 6.26. The Cu uptake increased
by an order of magnitude between 20C and 100 dc.
126.9C 12.5C
2.0,."...-----~--__,_--------____,
E
c.
c.
Q)
c5 1.5
C.
::::>
CD
c.
c.
<31.0
o
Cl
o
......J
O.5+----+----+----+-----i------if...-_--I
0.0025 0.0030 0.0035
Temperature (11K)
Fig. 6.26 Copper uptake variation as a result of the coating temperature [108].
(h) CU2-xS phases

One of the most controversial aspects of the powder DCPEL device has
been the possibility of CU2 _xS phases formed as a result of the above Cu
coating process. The existence of such phases would make the formation of a
p-n junction device more probable since ZnS is weakly n type. As mentioned
above, if the Cu coating process led to a uniform coverage, then the
possibility of the creation of any CU2-xS phase is extremely unlikely.
However, if the coating occurred non-uniformly, then the formation of the
CU2-xS phases was more probable. The existence of such phases has been
proven in photovoltaic devices based on CU2-xS and CdS. However, here
the thickness of the CU2-xS layer is nearly 0.2 J.l.m. Hence characterization
using most modern surface and bulk analytical techniques does not pose a
problem.
Some studies have been carried out to establish the nature of the CU2 - xS-
coated surface layer. Kelly [112] found distinct peaks in the resistance plots
of Cu-coated devices as a function of temperature (Fig. 6.27). These also
coincided with the well-known CU2-xS phase transition temperatures.
However, no discernible peaks were obtained using differential thermogra-
vometric analysis (DT A), indicating an absence of bulk phase transforma-
tions, since only the phase change transformations occurring within a
CU2-xS precipitate were involved. The amount of material involved would
be too small to be measurable by DT A but the electrical conductivity was
affected appreciably.
Zn48 Phosphor 1/2 hr oven Zn49

o Zn48
+Zn49
30 300
Resistance Resistance
103 ohms 103 ohms
20 200
10
J
II 100
20 40 60 80 100 120
Heat Treat Temperature cC
Fig. 6.27 Resistance changes as a function of temperature for C-coated ZnS:Mn,Cu
powder layers [112].
May [52] reported that, by measuring pulsed device conductance or

luminance as function of temperature, he found evidence for the existence of
CU2 _xS phases, as shown in Fig. 6.28.
The two peaks at "" 65C and 85 C were said to represent the digenite
polymorph and the analite-digenite transformations respectively. In other
220
~ l-ta+".
32
~
.,I<~
~ 200 30
!'J~
....
~
.
180 28
:::J
~
Conductance J.A ~
..
~ 160 26
1- -1--) ~ "' : ~
0 -
82 C ~ Q)
C)
E 140 24 <:
M '"c
Q)
'0
Q) 120 ~ rr~ 22 E
85 C
~
C) ~
'"
U
<:
100 ~ I
1 20 c
It)
I
~~ r-- Q)
::J
"0 64 C Luminance ~
<: 80 ~ 18
V ~~
0
--
U
60 ....-"'i 16
"
40
. ~ 14
10 20 30 40 50 60 70 80 90 100 110 120
Temperat ure (0C)

Fig. 6.28 Conductance-temperature plots used by May [52] to indicate the
presence of CU2 - xS phases.
1.2c-4, - -- - - - -- -- - - - -- -- - - - ---,
U)e-4
~ 8.0e5 SSC38(1/R)
<:
It) " SSC39 (1/ R)
U o SSC145(1 R)
-5 6.0<:.5
<:
o
U
4.0e5
2.0e.5 ~~~~~~---"'~-~-r-~~~~~....,...-~~
20 40 60 80 100 120 140
Temperature (0C)
Fig. 6.29 The conductance- temperature characteristics of the excess Cu-coated
devices (SSC38 and SSC 39) compared with an un aged device (SSC 145) [108].
studies (Fig. 6.29), the possible presence of CU2-xS phases was indicated in
devices that were fresh and for those where the Cu coating was carried out
at elevated temperatures [108]. The conductivity versus temperature plot
shows a decrease in conductance above 90 C in all cases. This may reflect
some Cu phase change or merely loss of adsorbed water. The role of the
adsorbed species will be mentioned later. Devices fabricated using Cu-
coated ZnS phosphor at standard conditions revealed no peaks after they

had been life tested for 1000 h. Therefore the question of CU2 -xS phases is
not resolved. However, as will be will be shown later, a p type surface was
confirmed by some authors.
6.5 FABRICATION OF POWDER EL STRUCTURES
6.5.1 Fabrication of ACPEL cells

The structure of ACPEL cells has already been given in Figs 6.3 and 6.4.
The substrate may be glass, plastic or metal. If metal substrates are used,
then the top electrode must be either indium tin oxide (ITO) or Sn02' The
ZnS:Cu phosphor is intimately mixed with a high dielectric binder and
applied to the substrate by spraying or spreading. A highly reflective
material, such as barium titanate (BaTi0 3 ), is used to increase the forward
light efficiency. To increase the stability, a dielectric layer may additionally
be incorporated. A plastic film or a protective glaze is normally used as the
final layer to inhibit moisture ingression.
6.5.2 Fabrication of DCPEL devices

Some fabrication stages of the DCPEL devices are shown in Figs 6.30-6.33.
Initially an ITO-coated glass substrate is etched in the requisite manner. The
phosphor that has been Cu coated as described above is mixed with an
organic binder (composed mainly of nitrocellulose). This phosphor-binder
suspension is then deposited onto the etched ITO substrate. The deposition
of the phosphor may be by spraying, chromatographic spreading or doctor
blading techniques. After drying the phosphor layer, the Al cathode is
thermally evaporated. This is followed by scribing to leave the desired
pattern. This scribing is important since any conductive phosphor material
between the row and column lines can lead to cross-talk. Finally, the
the AI rear electrode
ITO conductor lines

(0.5 IJm thick)
Fig. 6.30 Photolithographically etched ITO- (or SnOr) coated glass substrate.
Fabrication of powder EL structures 187
Phosphor layer
( ~ 25)Jm thick)
Fig. 6.31 Deposition of the Cu-coated ZnS:Mn,Cu powder layer on to the

TCE/glass substrate.
=~~_ AI evaporated
back electrode
Fig. 6.32 Deposition of the Al back electrode on to the ZnS:Mn,CujTCE/glass

substrate.
Scribed phosphor/rear
electrode layer
Fig. 6.33 Scribing of the excess AI/ZnS:Mn,Cu layers to yield electrically separated
emitting pixels.
connections are made between the Al cathode and the contact pad lead-ins
using electrically conductive paint (Ag dag).
Whereas ACPEL cells emit light on the first application of an electric
field, DCPEL devices require a forming process. This leads to the creation of
a very narrow ('" 1 /-lm), highly resistive, layer adjacent to the TCE (anode).
The light emission occurs in this formed region. In the next section, the
forming process, which is peculiar to the DCPEL devices only, will be
described.
6.6 THE FORMING PROCESS

Cu-coated ZnS:Mn,Cu powder phosphors require a forming process before
they can emit light. The reason for this is obvious. Initially after the Cu
coating process and device manufacture, all phosphor particles are virtually
identical with a very highly conductive surface. This is shown schematically
=
in Fig. 6.34.
T=
AI cathode
1
-------AI,Ov film
25 m I
I
I
1
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
Phosphor
I'Y"
ITO anode
Glass substrate
Fig. 6.34 Schematic diagram of an unformed DCEL powder device.
For light emission by an electrical energization process, some part of the

phosphor layer must be highly resistive so that electron heating leading to
efficient impact excitation of the Mn 2 + centres can take place as shown in
Fig. 6.35. Fortunately, it was found that the application of an electric field
with the transparent electrode positive created a thin layer of highly resistive
phosphor material which emitted light at very low applied voltages ( < 10 V).
This phenomenon is known as the 'forming process'. As shown in Fig. 6.36,
initially the current is limited, probably by a thin aluminium oxide film that
is formed under the Al cathode possibly as a result of the reaction of the
Orrich organic binder with the hot evaporated atoms of Al as they
condense to form the cathode. Devices made with either Au cathodes or
binders that are not oxidizable, e.g. silicone oils, do not show the presence of
this resistive film. After the applied voltage reaches around 4 V, the device
begins to conduct almost ohmically. The current rises linearly and then, at a
certain point, begins to drop. The rate of this drop is determined by many
factors, chief amongst which are device temperature, the ambient moisture
content, the ambient atmosphere and the type of binder used. The binder is
not inert as mentioned in the review by Hartmann et al. [113]. In fact, some
The forming process 189
~~BaCklayer
Cu coated
ZnS:Mn,Cu
powder phosphor
- - Formed layer
'c--- Light emission
Fig. 6.35 Schematic representation of a formed ZnS: Mn,Cu powder phosphor

layer.
50
Threshold voltage
40 (A1 2 0 3 breakdown)
~ Luminance
E 30 (ft - L)
+-'
c
....
OJ
.... 20
u 10
:::J Forming
voltage (V f)
0
0 10 20 30
Applied voltage (V)
Fig. 6.36 A typical forming trace for a DC PEL cell (0.2 cm 2 area).
binders can lead to devices that are very stable but dim and others lead to
devices that are initially very bright but exhibit poor maintenance. Normally
DCPEL devices are encapsulated with a desiccant placed in the backcap.
The cell structure after forming is shown in Fig. 6.37.
The first description of the forming process observed in Cu-coated
ZnS:Mn,Cu DCPEL devices was given by Vecht et al. [29] in 1969. They
stated that, 'when a voltage is applied to a Cu coated phosphor, during the first
minute or so (depending on the initial voltage) a marked drop in current is
observed associated with the onset of light emission'. The forming process was
said to be power dependent requiring about 1-2 W cm- 2. It was suggested
that the conduction in the DCPEL devices was via a non-stoichiometric
phase of CU2 - xS.
- - - A I cathode
T~ ____
25 IJm I I I I I I I I I I I I I
BaCklayer
1 ~o.:lllXx:tOo.:l~;:.:;:;- Formed layer

ITO anode
//_'///1 Glass substrate
Fig. 6.37 Schematic diagram showing a completely formed-up powder DCEL

device.
Following this various studies were conducted to elucidate the forming

process mechanism. In one study, an undoped but Cu-coated ZnS phosphor
device was formed and found to show green EL under AC excitation [29].
This is consistent with a [CUJlat value in the phosphor of at least 10- 4 % g of
Cu per g of Zn within the formed layer. In another study, the [CUJlat of the
ZnS particles (initially undoped) in the formed regions was found to have
increased to 100 ppm compared with < 10 ppm for the control phosphor.
Both these experiments point towards the diffusion of Cu into the ZnS
particles because of the applied electrical field. As shown in Fig. 6.38, the
potential drop profile of a formed 'gap cell' was measured and found to be
very steep at the emitting line. (The gap cells were fabricated by spreading a
phosphor layer between two metal electrodes and applying a voltage to
cause forming. This is different from a 'through cell' where a TCE forms one
electrode and, normally, Al the other.)
The potential profile shown in Fig. 6.38 indicates that the majority of the
applied voltage is dropped across the formed region. Thus the forming
process yields a thin, '" 1 J.1m thick layer, which is highly insulating [114].
The net result is that the field becomes intensified in the formed region, with
the concurrent advantage that high luminance is possible at low applied
voltages.
Beatty [115J considered that the forming process may be due to the
diffusion of say, Cu + ions, towards the cathode, leaving behind a less
conducting CU2-"S phase. This process may be assisted by a localized Joule
heating of the phosphor layer adjacent to the Sn02 electrode. Beatty
tentatively suggested that heating alone could cause a phase change in
CU2 _"S, regardless of any Cu + ion migration. This forms the basis of the
second theory for the forming process. In evidence, Beatty conducted some
temperature studies, showing that the resistance of the device decreased up
to '" 100 DC - as would be expected from a semiconductor. However,
Emitting
Cathode region Anode
+
After
forming ---t
E 101 (200 v) Before
E forming
1 (10 v)
~
u i
10-1 f \
! 1
co
0'0
~
!!l 10-
2
f-
,,1
.0
\ '~
"0 O~j/
> I"CT
10-3
o 4 8 12
Distance (mm)
Fig. 6.38 Potential profile in a gap cell [29].
prolonged heating above 100C led to increased resistance. This, he sug-

gested, may be due either to a phase change or to a diffusion process. Beatty
also questioned the forming model of Vecht et al. [29], involving Cu
migration into the ZnS particles for cells made from the standard phosphor
(0.4% Mn-O.05% Cu by weight). Since Cu solubility is low in ZnS, the
normal DC phosphor is already saturated with respect to this dopant [111].
In 1970, Vecht et at. [33] showed that the light emission and therefore the
forming process were localized at the positive electrode. If the polarity was
reversed, then the forming process occurred near the new anode, i.e. under
the aluminium. This result further reinforced the theory that forming
involved a migration process as the positively charged Cu + ions were
naturally expected to drift away from the anode.
Surface analytical techniques consisting of a scanning electron microscope
with an X-ray microanalysis attachment were used to elucidate the forming
process [59]. Unfortunately, since the electron probe diameter was large in
comparison with the size of the phosphor particles, no change in [Cu] was
observed before, during and after the forming process. However, by employ-
ing a heavily Cu-doped ZnS:Mn thin film, they showed that Cu depletion
had occurred near the positive electrode. This change in [Cu] occurred only
when the current started to fall during the forming process. Although they
were unable to detect any Cu ion movement in the powder phosphor
devices, the thin film result suggested that the movement of Cu also occurs
in the powder phosphor but on a subparticle scale.
Yamamoto and Yamazoe [77] observed both rectifying and photovoltaic
characteristics in DC PEL after the forming process. They found that the
surface of the Cu-coated ZnS powder was p type by measurement of the
electromotive force. Thus, using electronic device structure notation, the
formed device could be represented by Sn02-n-ZnS:Mn,Cu-p-Cu2-xS-AI.
As the TCE, i.e. Sn02 or ITO, is normally positive, the device can be
thought of as a reversed biased structure. Using rectifier notation, light was
emitted only when the voltage was applied in the reverse direction.
This is the only known reference to the observation of the rectification
behaviour for the light emission process in DCPEL devices. It is inconsistent
with the observations of Ellis et al. [34] that, on polarity reversal, light was still
emitted from the first formed region, although a second region eventually
formed under the new anode. However, the importance of the research of
Yamamoto and Yamazoe was the discovery that the response of the
short-circuit current of a DCPEL device with wavelength exhibited a peak at
0.33 ~ and a shoulder at 0.42 ~m. As shown in Fig. 6.39, the former
corresponds to the band gap energy of ZnS and the shoulder was thought to be
due to Cu atoms incorporated in the ZnS lattice during the forming process.
,.... 103 ~~I--l---I

.ri
!
c
~ 1021---+-~t--T--+---t
u
.t::
:::l
I:!
u
5 1011-_-+_ _+-~~---t
.c
rn
10
200 300 400 500
Wavelength (mil)
Fig. 6.39 Short-circuit current versus wavelength [77].
According to Kawarada and Ohshima [78], the combined effect of heat

and the high local electric field promotes the diffusion of Cu ions either into
the ZnS particles or along the surface of the particles. The latter forms the
basis of a third mechanism for the forming process. However, no evidence
was given by the authors for this proposed mechanism although migration
over a surface must be a considerably easier process than migration into the
ZnS.
As noted earlier, Smith [111] suggested a mechanism to account for the

possible formation of CUz _xS phases. Smith also recognized that, since
forming is a temperature-dependent process, it could be caused by a
rearrangement of the highly disordered, low chalcocite phase to hexagonal
chalcocite at 105C just before the actual forming process and via Cu ion
diffusion to high digenite. The high digenite phase has the least lattice
mismatch with the cubic ZnS sphalerite structure and thus the junction
between ZnS and Cuz-xS may be almost 'pseudo-epitaxial' in nature.
Unfortunately, because of the submicron dimensions of this junction,
acquiring direct evidence was not possible at the time.
In 1979, Abdalla et al. [88] measured the PL spectra with an argon laser
beam (50mW) operating in the UV range (351.1 nm-363.8 nm). The laser
beam was made to impinge on the glass side of the device so that the
phosphor layer next to the transparent electrode could be investigated.
Figure 6.40 shows the PL spectra recorded for an unformed device (spec-
trum a), a device aged for 100 h (spectrum b) and for 700 h (spectrum c).
lOO~ __________________ ~~ __________-,

x 0.25-+
o
4000 4500 5000 5500 6000 6500 7000
Wavelength (A)
Fig. 6.40 PL spectra of unaged device (spectrum a), formed device (spectrum b) and
aged device for 700 h (spectrum c) [89].
As shown, the unformed device has two peaks (curve a), centred at
460.0 nm and 516.3 nm, which are associated with Cu (blue and green)
centres and one at 584.0 nm, which is the well-known Mn peak [29]. Curve
b shows that the peak due to the green Cu centres is greatly reduced whilst
the blue centre peak is almost absent. Curve c shows that the Cu green band
has been now substantially reduced. Interestingly, the Mn peak is also
reduced and some other, unidentified, peaks have appeared. However,
Abdalla et al. did not consider the possibility that the luminescence from the
Cu centres may be reduced by some other mechanism such as formation of
non-radiative centres, e.g. Cu! ions. The Cu ion pairs may loose their
energy non-radiatively in a similar fashion to the Mn-Mn pair centres. The
oxidation state of the eu ions also may change from eu + to eu 2 +

depending on the availability of moisture.
A major investigation of the forming process and the failure modes
(section 6.10.2) by Alder et al. [24] was reported in 1980. They found that
the resistivity of the eu-coated phosphor was '" 2 !l m which meant that the
resistance of the 0.2 cm 2 phosphor/binder layer with a thickness of 30 ~m
was '" 3!l. The resistance of the electrical lead-ins was measured at '" 200 !l.
However, the resistance determined for the unformed device was much
greater (",4 k!l). As Fig. 6.41 shows the I-V characteristics determined by
single-shot pulse measurements were non-ohmic.
Voltage
(V)
Fig. 6.41 Uncorrected I-V characteristics for a standard test device of area
0.2 cm 2 Curve a was taken prior to forming. The remaining curves were taken after
forming to the following: b, 20 V; c, 30 V; d, 40 V; e, 50 V; f, 60 V; g, 70 V; h, 80 V;
i,90V.
Obviously, the I-V characteristics of the unformed device were dominated

by device components other than the phosphor layer. Small signal equiva-
lent circuit analysis of devices made with different electrode materials
showed that an oxide layer, e.g. A1 2 0 3 , played a major role in determining
the general behaviour of unformed devices.
When the devices were subjected to a forming process, the I-V character-
istics remained essentially parallel to those of the unformed devices. This
implies that the electrical characteristics of the various components external
to the phosphor layer and of the eu-coated phosphor particles are un-
changed. The unformed I-V characteristics thus could be subtracted from
those of the device at the various forming voltages to obtain the characteris-
tics of the formed layer. This is shown in Fig. 6.42.
The importance of this result is that, within the experimental limits, at the
high currents normally required for significant light emission, the voltage
current (A)
I m
100 abc de f h i
80
60
40
20
-50 -25
Voltage
-125 -100 -75 ,Ij 25 50 75 100 125
(V)
20
40
60
80
h gfedcba 100
Fig. 6.42 I-V curves obtained after subtraction of the curve obtained prior to
forming. The curves are derived from those taken at the following voltages: a, 20 V;
~~~~~~~~~~w~Lm~&w~~~~
across the formed layer is constant and is independent of the current passed.
Indeed, the characteristics deduced for higher forming voltages indicated
that the voltage across the layer decreased as the current increased. This
indicates that there is the possibility of negative differential resistance (NOR)
in the ZnS:Mn,Cu powder devices under normal drive conditions.
Alder et al. [64] also found that the reciprocal capacitance was a linear
function of the formed voltage for both the normal (i.e. transparent electrode
as positive) and the inverse voltages. This is shown in Fig. 6.43.
The results of their forming studies led Alder et al. [63] to conclude that,
according to their idealized model (Cu migration), there was 'no reason for
the forming process to stop at any particular value of applied voltage'.
Using surface analytical techniques, Alexander et al. [116, 117] elucidated
that initially (before forming) the Cu on the surface was present as Cu + ions
associated with sulphur, probably as CU2S. The storage of Cu-coated
phosphor in air had the effect of oxidizing some sulphide to sulphate. Auger
spectroscopy revealed that Cu depletion had occurred on the surface of ZnS
after the forming process. Electron probe microanalysis (EPMA) indicated
that the total [Cu] had not altered in the formed region. This was confirmed
in both gap cells and in through cells. The results further support the work
of Vecht et al. [29].
Alexander et al. also found that certain gaseous ambients affected the
forming and the degradation rates of the powder devices. Devices were said
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
o 10 20 30 40 50 60 70 80 90
Breakdown voltage (volts)
Fig. 6.43 Reciprocal device capacitance as a function of the breakdown voltage

[24].
to form rapidly in oxygen-moisture but very slowly in other gases such as

argon, carbon monoxide and methane. Devices that formed at very slow
rate ( < 0.02 V h -1) were said to be 'stuck'. These results lead to a different
conclusion from that of Alder et al. [63] in 1980 who stated that forming
would continue as long as the applied voltage was increased.
Some work on the role of the surface environment had already been
conducted by Beatty in 1970 [115] and 1974 [118]. To a large degree,
Alexander et al. suggested (following Beatty in 1974) that the forming
process may be due to an oxidation of the phosphor surface layer by
extending the Mott-Cabrera [119] theory to insulators such as ZnS.
Originally this theory was devised for explaining the oxidation of metal
surfaces. In this study, the question of whether it was oxygen or moisture
that was responsible for the enhanced forming rate was not resolved.
In an attempt to resolve this issue of oxygen versus water vapour, a series
of experiments were carried out in dry air, wet air and in vacuum [57]. The
results showed that it was the moisture that was most active in accelerating
the forming process. Figure 6.44 shows the forming rates of the two devices
exposed to dry air (oxygen) and wet air (moisture) in a vacuum chamber.
The device exposed to wet air continued to form whereas the device exposed
to dry air slowed down the forming to its previous rate when under high
vacuum.
Further evidence for the role of moisture in the forming process is given in
Table 6.4. This shows the time taken by identical devices when exposed to
various ambient atmospheres.
Thus oxidation by moisture forms the basis of the fourth model for the
forming process.
In summary, four models for the forming process have been proposed
over the years. Two have involved Cu ion migration, either into or over the
1005 10005
-1.8,.-------,-------,------:7'"-,
Dry Air (8 x 10- 4 mBar)

~ -2.0 I X :
Q)
Cl
I
co
~ -2.2
>
Cl
c
'e -2.4
-o
o
odi -2.6
...J
0
-2.8 +-...,--.--.--.-..........-+----.--.---.---.--+..........---.--.--.-...--i
2.0 2.5 3.0 3.5
Log,o time (5)
Fig. 6.44 Changes in the forming rates as a result of device exposure to wet and dry
air [57].
Table 6.4 Forming times of devices from 30 to 70 V

using a constant power supply (1.25 W em - 2) under
various conditions [108]
Forming conditions Forming times from
30 V to 70 V at
constant power
Unencapsulated (1 atm) -2-3 min
Encapsulated (1 atm) -30-60 min
In vacuum (10- 6 mbar) -17200min
surface of the powder particles, and two have involved chemical changes of
the surface layer of the Cll, namely a CU2-xS phase change and an
oxidation model involving formation of the oxides and then sulphates of Cu
and Zn respectively. The possibility exists, of course, that forming involves
combination of two or more of these basic models. As noted earlier,
evidence has only been presented for the migration of Cu into the ZnS
lattice. No evidence exists for the migration of Cu over the surface of the
phosphor particle although it may be a considerably easier and therefore a
faster process and Beatty's [115] reservations about the likelihood of Cu
migration into an already saturated ZnS lattice. No evidence exists for
migration of Cu towards the cathode and as the results of Alexander et al.
indicate, the total [Cu] (surface+lattice per particle) is the same in the
unformed and formed regions. Evidence for any phase change has not been
conclusive.
It appears from the available data that the forming process involves an
initial migration of the Cu ions into the ZnS particles followed by an
oxidation process involving surface moisture (",0.3% by weight). This
adsorbed moisture is extremely difficult to remove.
After the DCPEL devices have been formed up to 70 V, they are given a
long-life treatment which involves the replacement of the dry air in the
backcap by an inert gas such as Ar or N 2 . Normally this treatment is carried
out at 105 DC as recommended by Werring [114]. The high temperature
appears to enhance the useful life of the DCPEL devices, probably by
removing some the adsorbed moisture.
The excitation mechanisms responsible for the luminescence process in
powder EL devices are discussed next.
6.7 INJECTION AND EXCITATION MECHANISMS IN

ZnS PHOSPHORS
It should be clear from the preceding review of the device structures and
properties that EL can be neither a simple nor a single phenomenon. For the
same material, EL may proceed by quite different mechanisms depending on
the energy levels and the local field conditions produced by impurities,
defects and electrode contacts in the material.
ZnS phosphor is particularly complex in this respect. It is a II-VI
compound that has both ionic and covalent properties. The ionic character
is approximately 62%. Because of the self-compensation mechanism, a large
number of impurities can be incorporated into the lattice, leading to a
complex energy level structure. Early experiments on the phosphors paid
very little attention to the problem of materials and led to many conflicting
results.
In this section a brief discussion of the more plausible excitation mechan-
isms will be given.
6.7.1 Injection and excitation mechanisms

Piper and Williams [120J considered that the only acceptable excitation
mechanism in ACPEL is the impact excitation of the luminescent centres
by electrons accelerated under high electric field. This mechanism is also
widely postulated to occur in DCPEL devices. High electric fields in excess
of 105 V cm -1 are required for this mechanism - being localized in a
barrier region produced by carrier exhaustion, e.g. at a metal-phosphor
contact or a reverse-biased P-Cu2_xS-n-ZnS junction. The free carriers
(predominantly electrons in ZnS) are supplied by either thermal genera-
tion from traps in the phosphor material tunnelling through the barrier
or Schottky emission over the barrier. These possibilities are discussed
below.
Injection and excitation mechanisms in ZnS phosphors 199
( a) Field emission
Electrons from localized levels in the band gap are released by the local high
field. This process is very similar to the Zener effect of a reversed biased
diode where a very high field is produced in the depletion region. The
emission current (1) is related to the applied voltage (V) by
Ioc V 2 exp( -constant/V) (6.8)
Piper and Williams [120] and Taylor [121] calculated that the luminance
B obtained is related to the applied voltage, V, by
B=aV" exp( -bVO. 5 ) (6.9)
where a and b are constants and n=O, 2.
(b) Thermal emission

The electron traps are shallow and the captured electrons are thermally
released into the conduction band. The released electrons are accelerated by
a field to a high energy before causing impact excitation of the activator
centres. A detailed analysis by Taylor [121] of the voltage dependence of the
brightness B yielded the following expression:
B=Bo exp( -bV-O. 5 ) (6.10)
where band B o are constants.
( c) Schottky emission and tunnelling

A P-Cu2-xS-n-ZnS heterojunction is assumed in this model. The barrier
height of the junction is lowered by the applied field, leading to the
thermionic emission of the electrons from the CU2-xS phase into ZnS. At
high fields tunnelling of the electrons through the barrier is also possible.
The voltage and temperature (T) dependence of the current are respectively:
(6.11)
and
1= PT 2 exp( - constant/T) (6.12)
where 0(, Pand')' are constants.
The electrons are then accelerated in the high field until they gain
sufficient energy to impact ionize the lattice or a filled recombination centre,
subsequently recombining with the minority carriers. Under the high field
conditions, electrons will be swept away from the ionized recombination
centres and under direct current excitation the rate of recombination will be
low and the EL inefficient. Under AC excitation, however, some electrons
will return to the region ionized in the previous half-cycle and recombine
under low field conditions.
6.7.2 Minority carrier injection

Another possible mechanism involving the injection of holes has been
suggested by Goldberg and Nickerson [107]. This mechanism relies on the
formation of a Cu-rich phase at the anode leading to a heterojunction with
the ZnS. Holes can be injected into the phosphor from this CU2-xS phase.
The injected holes are eventually trapped in the activator centres. Since no
impact ionization is involved, the need for a high field is avoided. The I-V
characteristic for such a junction is given by
I =Io[exp(eV/kT)-lJ (6.13)
where k is Boltzmann's constant.
The large band gap of 3.7 eV in ZnS leads to negligible production of
thermally generated carriers at room temperature (Le. lois very small).
In both cases the excitation of the luminescent centres may occur by two
possible mechanisms. Firstly the Cu + centre is excited by the impact of
electrons or trapping of a hole and then the luminescent Mn 2+ centre is
excited by a resonance energy transfer mechanism through the excited Cu +
centre [107]. Secondly, the luminescent Mn 2+ is directly excited by the
impact of electron or trapping of a hole. In either case the characteristic
emission of an Mn2+ centre is obtained.
6.7.3 Excitation of tbe Mn centre-direct excitation

Zallen et al. [122J have shown that only the bound states of the Mn 2+
centre are involved in the EL process and ionization does not occur. Hence
the excitation mechanism involving a resonance energy transfer between
Cu + and Mn2+ is more generally accepted to occur in the ZnS:Mn,Cu
phosphor system.
However, it must be stressed that direct impact excitation of the lumines-
cent centres in other phosphor systems is not impossible. Bright DCEL has
been demonstrated in Cu-free ZnS:Mn thin films [123J. Furthermore,
Waite and Thomas [124J have studied the post field recombination process
in the ZnS:Mn powder phosphor system under pulsed DC excitation and
raised doubts on the model involving energy transfer to the Mn 2+ centres.
6.7.4 Double injection (Fiscber's model)

The presence of conductive precipitates of CU2-xS within the phosphor
particles is assumed. The band gap of CU2-xS is around 1.25-1.7 eV.
Carriers are thermally generated in the precipitates and injected into the
Optoelectrical response 201
phosphor by a field emission mechanism. This model was originally pro-

posed for ACPEL in powder phosphor by Fischer [S]. In the case of
DCPEL two opposite precipitates of CU2 _ xS are assumed; the electrons and
holes are injected into the phosphor in the opposite directions.
Experiments on thin films of ZnS materials have illustrated that a number
of injection mechanisms are plausible depending on the activators in the
materials and the electrodes used. Although in AC and DC thin films impact
excitation of the Mn 2+ is accepted as the most likely mechanism, in powders
it is more likely than not that multiple injection mechanisms occur, although
only one or two of them are dominant in any phosphor system. It is essential
to remember this when comparing results obtained in different phosphor
systems.
6.8 OPTOELECTRICAL RESPONSE

6.8.1 ACPEL cells
The AC-driven panel is capaCItIve in nature and obeys the following
equation for the light emission process as discovered by Zalm [14]:
(6.14)
where B and V are brightness and applied voltage respectively, and Bo and
Vo are parameters that depend on the temperature, the frequency of the
alternating voltage and other details of the construction of the test cell (e.g.
particle size, phosphor:binder ratio, layer thickness, dielectric constant of
binder. The presence of the square root of the voltage is normally explained
by an electron-accelerating mechanism, together with the fact that maxi-
mum field strength in a Mott-Schottky barrier varies as Vl/2. Departures
from this relationship were often observed and modifications introduced to
fit the results. Figure 6.45 shows a typical B-V curve for the ACPEL cells.
As can be seen, the B-V relationship is non-linear. This leads to a poor
discrimination ratio (DR) which is defined as
d. . . . . brightness at voltage V
(6.15)
IscnmmatIon ratIO = b' h '/2
ng tness at tj
With the DR being only around (3-5):1, these displays are not easy to
multiplex and normally useful alphanumeric and graphical displays would
not be feasible. The only way to overcome this problem is to incorporate a
TFT drive into the AC powder panel structure (Fig. 6.S) as described by
Fischer [S] and attempted by Brody et al. [125]. Unfortunately, this was a
very expensive solution at the time owing to the very low TFT yields and
never fully pursued. In addition, the stable AC thin film structures reported
by Inoguchi et al. [126] from Sharp in 1974 also further diminished interest
in AC powder for alphanumeric and graphical display purposes.
600.-__________________~----------------_,
+ ............................................................................................................................................,.r............................
::J
,
:S 400 -t ......................................................................................!..................................:.-/-..............................................
:;
C.
:; ; ....................................................................................; .............. ,1)(..................................................................
o
Q)
u
c:
co
c: 200
'E
::J
....J
.............................................................:;;'.""'.............+................................................................................. . 1
o
o 100 200
Applied voltage (V)
Fig. 6.45 Typical B- V curve for an ACPEL cell.
1000~-----------,------------~-----------.
~ 100 ....................................................t....................................................1".....................................................
....::J
C.
:;
o
Q)
gco 10 ....................................................( ....................................................j.......................................................

c:
E
::J
....J
1
100 1000 10000 100000
Frequency (Hz)
Fig. 6.46 Typical luminance-frequency curve for ACPEL cells.
The brightness increases with frequency. This is shown in Fig. 6.46.

Although luminance in excess of 600 ft-L is obtainable at 20 kHz, the
maintenance would be extremely poor at this frequency. Most commercial
ACPEL cells are operated at 400-500 Hz and 110-120 V or 60 Hz and 240 V.
The luminance is also determined by the thickness of the powder
phosphor layer. This is shown in Fig. 6.47. It is obvious that bright cells
Optoelectrical response 203
10
1\
:;
1\
c.
:;
o ~
\ -
Q)
(.)
c
co
c
._"\
E 1
::J
...J
.1
\ ~
o 10 20 30 40 50 60 70 80 90 1 00
ZnS:Cu phosphor layer thickness (~m)
Fig. 6.47 The effect of phosphor layer thickness on the luminance of a typical
ACPEL cell.
could be obtained, if the phosphor layer thickness is minimal. However,

as shown in Fig. 6.8, the ACPEL particles are between 5 and 30 11m thick
and, as shown in Fig. 6.9, reducing particle size by milling has a detrimental
effect. Hence fabrication of a very thin ACPEL cell is not very practical. Of
course, fine grains of the phosphor could be separated from a good
phosphor by sieving or sedimentation. However, this has the effect of
making the phosphor very expensive and has been shown by Faria [127] to
affect the maintenance adversely.
6.8.2 DCPEL devices

The brightness of the DCPEL panels shows the following dependence on
current
(6.16)
where B is the brightness, K is a constant, I is the applied current and n is a
number that lies between 0.5 and 2.
The DCPEL panels are easily addressable because of their steep B-V
characteristics and therefore exhibit a high DR, being about 500:1. (Fig.
6.48). Furthermore, since the light emission is only from the narrow formed
region, the thickness of the layer has no effect on the luminance. As such
these displays offer the possibility for use in x-y address situations. The
largest alphanumeric and graphical DCPEL panel manufactured consists of
640 by 480 pixels, i.e. 307 200 pixels.
'"'"c: 101
Q)
.E
OJ
.;::
CD
1~L---------~~~____~~
m ro ~ ~ ~ 00 W M 00 ~
Volts
Fig. 6.48 Typical B-V characteristics of a DCPEL device.
6.9 DRIVE CIRCUITRY
6.9.1 ACPEL cells

As has been mentioned, normal ACPEL cells do not lend themselves for
applications in the alphanumeric and graphical display mode. The ACPEL
cells are utilized for the backlighting of LCDs and plasma panels as well as
for fixed legends (messages or symbols). As such they do not require
complicated drive circuitry and can be powered by either mains or via an
inverter if used in portable equipment, e.g. backlighting of LCDs.
The first and the only prototype ACPEL panel to be assessed for TV
applications was described by Yoshiyama et al. in 1970 [19]. This display
had 52900 pixels on a 13 in diagonal substrate. A block diagram of the drive
circuit is shown in Fig. 6.49. To scan the display, the selecting pulse was first
applied to the row to be scanned with a simultaneous multichannel video
signal to the column electrode that controlled the brightness. There were
over 8600 discrete components in the driving circuitry with very few ICs.
The depth of this panel was 20 mm and it consumed over 100 Wand emitted
only 1 ft-L.
Apart from the work of Brody et al. [125] (Fig. 6.5), this represents the
most complicated ACPEL structure.
6.9.2 DCPEL devices

Similar to the ACPEL cells, the DCPEL panels can be employed in the fixed
legend mode and require only continuous or pulsed DC power supply. A
Drive circuitry 205
r
IY
X~-X;---------x-'
~1
r------------l
H~r~zont~1
I --IT
I - - --------- - - --- -- - - m__ H,PG el,ectrode
I J driving
I
CirCUit
t~H,PG"
Y, -- - - - - - - - - - - - - - - - - - - H,PG
Y, - - - ------------ ------ --,
II x, !I
--+i~
Counter
I,,I
circuits
I
I
I,
, EL matrix II Vert.
5ynch.
5i9
! Ym -- -+1 H"PG
r- - -- ---- ---
i
i
Ir--L-'---I.....---...L...-----L..,
i Delay Line f4t-----~_ ___'
L___~:a~~~~~c..u_it _ _ _ .:
Speaker
Fig. 6.49 The block diagram for the drive circuitry of a ACPEL panel for TV
applications [19].
few papers describing the use of AC drive for the running of DC PEL panels
have also been published with emphasis being placed on apparently en-
hanced stability as a consequence.
Panels with true alphanumeric and graphical display capabilities require
fairly high voltage drivers. The rest of this section will be devoted to a
description of driving circuitry for the DCPEL panels starting with a brief
description of the methods of addressing x-y matrix panels. This will be
followed by the first reported EL TV from Matsushita Electric Industrial
Company to the present displays available from Phosphor Products Com-
pany and Cherry Electrical.
All flat panels employ an x-y matrix structure as shown in Fig. 6.50.
The two sets of electrodes are placed orthogonally so that any single pixel
may be selected by sending pulses along the column and row lines. It
should be recalled that the standard pixel-by-pixel scanning methods used
in conventional CRT receivers is not a viable addressing technique for
matrix panels of any type as this method would result in very low duty
cycle and hence very low display brightness. The preferred method is
therefore sequential line at a time. This method not only simplifies the
system configuration by eliminating the need for horizontal scanning but
also increases the brightness of the display. Even then a display with 640 x
200 pixels driven at 240 Hz only allows 12.5 Jls on time per pixel per frame.
Selected Pixel (LIT)
Row
Vo ltage
Fig. 6.50 Schematic diagram showing the selection of a pixel in any x-y matrix
panel.
DCPEL was also investigated for TV purposes by the workers from the
Matsushita Electric Company [78]. The manufacture of a 13 in diagonal
display with 50 176 pixels was attempted. The display had 150 TV lines and
emitted 10 ft-L. It appears that a contrast ratio of 20:1 and a grey scale
capability of 16 steps were achieved but the maintenance was poor.
However, this display consumed 150 W compared with the 100 W consumed
by the AC variety.
The drive systems for a range of display sizes ranging from 32 to 1250
characters were reported by Mears et al. [61] in 1973 (Fig. 6.51). In a
sense, they were similar to those reported by Yoshiyama et al [19] since
only discrete components were used (although integrated circuit drivers
with the necessary high voltage output had been designed and these
included the necessary control logic). With the available logic controls, the
maximum input rate was only 300 Hz. This would be unacceptable since to
sequentially assemble a line of information (300 bits) and then to write to
the display would reduce the duty cycle to a very low value. The display
would then be very dim and show flicker. The problem could be overcome
by the inclusion of a memory-mapped RAM with a byte representing one
character.
Very similar to the ACPEL cells, the DCPEL devices also suffer from a
gradual increase in the resistance of the light emitting formed region (section
6.10). As a consequence, when driven by a constant voltage drive, typical
large area multiplexed panels decay by 30% over 10000 h. This is still more
than the average lifetime required for most laptop and portable computer
displays (it is worth remembering that for a computer used for 8 h per day,
i.e. nearly 2000 h per annum, a 10 000 h display still has a useful life of 5
years).
Maintenance of powder EL devices 207
Line
counter
Display
Control
Unit
Keyboard Input ROM Character

Generator
store
Fig. 6.51 An early drive system for the DCPEL devices [61].
For even greater stability, an obvious solution is to use a constant

current drive. The device luminance therefore remains steady. Cherry
Electrical have done this for their recent range of panels. A constant current
drive is shown in Fig. 6.52. This circuit coupled with the fact that the drivers
are able to deliver up to 180 V essentially means that a very long life panel
results.
6.10 MAINTENANCE OF POWDER EL DEVICES
6.10.1 Maintenance of ACPEL cells

The luminance of ACPEL cells decreases gradually with operating time
under steady excitation (Fig. 6.53). This deterioration was present even
when attempts were made to exclude moisture effects. Moisture combined
with an electric field leads to the electrolytic generation of free Zn metal.
The decay is worse at higher frequencies. Various failure models have
10 Data
2 Clocks
VS DCEL
HS
DATA Interface IC
CLOCK--~~-'r-~~~~
Display
PGM current
PINS
Ref ..
voltage
639
Level 20 Current
Column
Shifter Lines (REF)
L..--T-~1-~ driver
201C'S
Row 0 Row 199
,--......L.......L.----.
DATA
Row
CLOCK Driver
+ 12V & IC'S
VS ------- Vertical SYNC

HS ------- Horizontal SYNC
AVPC ---- Automatic voltage & pulse control
Fig. 6.52 The constant current drive circuit used by Cherry Electrical [74].
100
....:::I
....0.
g~
Q) I 50
u
c:"O
E
ctlU
c: ~
'E
:::I
.-..J
(b)
5000 10000
Time (hrs)
Fig. 6.53 Maintenance curves of ACPEL cells with (curve a) and without (curve b)
an inverter.
Maintenance of powder ELdevices 209
been proposed for these devices. Roberts [128] invoked ionic diffusion in
the electric field as a possible cause though mentioned that moisture
caused discoloration of the phosphor. Further evidence for the ionic
diffusion model came from the work of Jaffe [90] who showed that by
replacing the halide co-activator in ZnS:Cu,X phosphors from Cl- to Br-
decreased the rate of decay. He thus believed that the slower migration of
the Br- compared with Cl- was responsible for the improvements ob-
served.
The most important work regarding the maintenance of ACPEL cells
was carried out by Fischer [5]. He showed that the decay in brightness
was related to the blunting of microscopic tips of CU2-xS inclusions in the
ZnS phosphor (Fig. 6.54).
CU 2 ,S Lines ZnS Particles
Fig. 6.54 CU2-xS inclusions shaped like a double-comet in the ZnS:Cu,Cl powder
phosphor [5].
These CU2 _xS tips decorated imperfection lines in the ZnS and could be
blunted by the outward migration of the Cu ions amongst others. In a
well-executed experiment, Fischer showed that initially EL was observed in
a double-comet line shape. After some time, the double-comet lines changed
into just two light-emitting points, just at the positions where the comet
heads had been before.
Some deterioration was believed to be photon assisted because blue
emitting phosphors deteriorated faster than orange emitting ones - the host
lattice being identical. Lehmann [11] believed that it was the S vacancies
(Vs) in the ZnS phosphors which were a prime cause of the deterioration.
Formation of Vs is expected since these phosphors are normally fired at
elevated temperatures making loss of S highly probable. Lehmann went on
to make the 'hypermaintenance phosphor' which consisted of phosphor fired
in an S atmosphere.
6.10.2 Maintenance of DCPEL panels
Earlier on it was noted that the initial creation of the formed region in
DCPEL devices probably occurred as a result of Cu migration, although the
gaseous ambient also may be involved. The above section on drive circuitry
also mentioned the luminance drop that is always observed in DC devices
operated at a constant voltage. The initial rate of decay is high, followed by
a more gradual decrease. The decay is not exponential, but it has become
conventional for the maintenance of a device to be defined as the time for
the luminance at a given voltage to reach half the initial value. The current is
also found to decrease simultaneously. In this section, the pertinent litera-
ture on the maintenance characteristics will be presented.
Probably the earliest mention of the degradation phenomenon in DC PEL
devices was by Kazankin et al. [21] in 1959. Their devices had very poor
maintenance, lasting a few tens of minutes. Kazankin et al. [21] attributed
the decay in brightness and current to polarization effects since they
assumed that their cells were condensers. In support of their hypothesis, the
authors reported that under the application of the electric field, the crystals
(particles?) actually align in the direction of the applied field and formed
little bridges between the plates of the condenser.
Vecht et al. [32] found that the luminance decay had been significantly
slowed by the encapsulation of the EL devices. In fact, the devices had not
reached their half-life after hundreds of hours. This result clearly points
towards the role of some constituent of the ambient atmosphere in the
luminesence deterioration process. No mention was made of desiccant to
preserve a dry atmosphere in this paper. However, a subsequent paper
reported the use of dry nitrogen and a desiccant [33]. The current flowing
through the device also decreased. Thus essentially the efficiency of the
devices was found to remain stable. It was suggested that the maintenance of
the device could be improved considerably by using a series resistor and by
further improvements to the encapsulation technique. These therefore for-
med the first two methods for increasing the maintenance of the powder
devices.
Although the half-life was improved from a few minutes to over 900h,
Vecht et al. [31] had recognized that devices must last for 5000-10000 h to
be useful.
Figure 6.55 shows the maintenance curves of a typical device containing
dry air and of a device containing a desiccant (magnesium perchlorate) and
an inert gas.
The first electrical measurements on life-tested devices indicated that the
impedance of devices was increasing with operating time [33]. Following
this, the use of constant power to achieve a pseudo-stability in luminance
was suggested by Ellis et al. [34]. It was shown that running a device at
constant power resulted in luminance remaining essentially unchanged for
Maintenance of powder ELdevices 211
100
:::J,
...
~
:; \. Improved drying
...
I
Q.
--
::J
0 10
CD
tJ
c:
(1)
c:
E
::J
....J
1
20 40 60 80 100 120 140 160
-Hours~
Fig. 6.55 Life curve, showing the effect of dry nitrogen atmosphere. 'Through'
luminescence cell at 100 V [34].
several hundreds of hours. As shown in Fig. 6.56, by improving the

encapsulation technique to prevent ingress of moisture, a 1000 h half-life was
achieved. This represents a remarkable improvement in both the luminance
and the stability characteristics of powder DCPEL devices.
0.2
:::J
t 0.1
Efficiency >-
tJ
c:
CD
CD
tJ
50 u
c:
(1)
if
CD
c: 20
E 0
Ci
::J 0
....J ....J
0
Ci
0
....J o 0 1 2 3
Time (hrs) x 1000
Fig. 6.56 Brightness and efficiency versus operating time, at 100 V [36].
Mears et al. [61] reported that between 30 and 100 ft-L were achieved
with devices pulsed at 0.5% duty cycle with good operating performance.
Figure 6.57 shows the maintenance curve of these pulsed devices.
According to Abdalla et al. [85], the continued migration of Cu was the
culprit for the loss of brightness. In addition, it was felt that the Cu-Mn
resonance energy transfer process may become less efficient on degradation,
again because of continued Cu migration.
In a subsequent publication Abdalla et al. [86] suggested the use of an AC
drive to stabilize DC formed devices. The results of a device formed and run
50
::::J
~
:;
B-
:J
o
OJ
u b
c:
co
c:
"E
:J
...J
10
10
Hours of operation
Fig. 6.57 Stability of a 100 V device driven at 185 V, 3 IlS pulse width, 0.7% duty
cycle (curve a) and 160 V, 121ls, 0.5% duty cycle (curve b) [61].
at 70 V DC were given. Initially, this device emitted 9.8 ft-L and the
luminance dropped to less than 0.1 ft-L after 200 h. Replacement of the 70 V
DC drive with 81 V (rms, 800 Hz) AC drive increased the luminance to
9.8 ft-L again which then dropped to 8.4 ft-L after another 450 h.
Until 1980, most of the authors had referred to device degradation in
terms of either loss of luminance with time or time to half-life or having to
increase the voltage to keep constant brightness level. It was left to Alder et
al. [24] to rectify this situation and to treat EL cells in the same way as Si
devices. They measured pulsed I-V characteristics before, during and after
forming to certain voltages (section 6.6). At each stage, the capacitance of the
devices was also measured. In addition, devices were characterized during
aging. Three possible types of device failure modes were diagnosed.
1. Further forming. As shown in Fig. 6.58, further forming is a parallel shift
in the I-V characteristics normally accompanied by an increasing
efficiency. This failure mode is ascribed to the widening of the formed
120
<100
oS
c: 80
~
:; 60
"
:J
~ 40
.,
il: 20
o 20 40 60 80 100 120 140 160

Applied voltage (V)
Fig. 6.58 Parallel shift in I-V curves indicates further forming (VF= Vo at To and
VF = VI at TI)'
layer that normally results in more Mn 2 + being excited. Thus although
the brightness may be expected to decrease because of lowered current,
the increase in the quantum efficiency often partially offsets this loss (Fig.
6.59). The sharp threshold in the I-V characteristics led to the panels
that failed by further forming being referred to as 'hard' devices.
2. Softening. As shown in Fig. 6.60, a significant leakage of current prior to
the 'threshold voltage' leading to inefficient devices was termed softening.
This failure mode was the result of parallel leakage paths, probably on
the surface of phosphor particles in the formed region. The B-1 curve is
therefore depressed (Fig. 6.61) and the quantum efficiency is reduced.
T=T,
...
:::J
Co
'5
o Significant increase in
Q)
u luminance efficiency exhibited by
c:
III
c: panels which further form
"E
:::J
...J
Pulse current (rnA)

Fig. 6.59 Further forming is normally accompanied by an increase in power
efficiency" (0/0) as shown above.
120
;{100
.s
1: 80
~
B60
:::J
~4O
a:::
20
o
o 20 40 60 80 100 120 140 160
Applied voltage (V)
Fig. 6.60 Softening in the I-V characteristics leads to significant current passage
before the threshold voltage point (VF= VO at To and VF = Vl at Td.
Luminance output decreases
Significant flow of current

prior to emission of light
20
o
o 20 40 60 80 100 120 140
Pulse current (mA)
Fig. 6.61 Softening leads to reduced power efficiency 1J (%).
3. Load line instability. This represents an increase in the resistance of the

backlayer. This leads to decreased luminance although the quantum
efficiency is not altered. Of the three failure modes cited by Alder et al.
[24], this was the only one for which the cause was known - basically
overheating of the device during manufacturing, forming or running,
damaging the CU2-xS layer. Figure 6.62 shows this failure mode.
Further forming was thought to be a continuation of the initial forming
process, i.e. by Cu migration leading to a wider depletion region. This leads
to an increased resistance of the formed region, necessitating ever-increasing
120
;?100
E
~
80
E
~
::; 60
u
:;,
<1l 40
~
<1l
a::: 20
o
o 20 40 60 80 100 120 140 160
Applied voltage (V)
Fig. 6.62 Large increase in the series resistance is termed load line instability
(where formed voltage VF = Vo at To and VI at Td.
voltage to maintain constant luminance. According to Alder et al. [64J, this
failure mode was successfully suppressed by a heat treatment process at
105C [114]. The cure for load line change requires limiting high processing
or running temperatures. The only failure mode for which the cause was not
known was softening. However, it was stated that, by resubjecting the 'soft'
devices to continuous DC (i.e. as during initial forming), they could be
restored to 'hard', i.e. efficient, devices.
Vecht et al. [53J reported the use of thin film (TF) interlayers between the
TCE and the ZnS:Mn,Cu. These TF interlayers of ZnS, SiO, Y203 and
polymeric materials were found not only to reduce the initial forming
powders but also to eliminate softening. The obvious implication of this
result is that ionic migration from the glass may be responsible for soft
devices. The effects of the gas ambient were discussed and the choice of the
correct molecular sieve to eliminate the residual gaseous impurities was said
to be a dominant factor in obtaining improved maintenance. This was one
of the first publications to note the possible ~ole of a negatively charged
species (probably Vs) in the maintenance characteristics.
According to Fujii and Hoshina [79J, DCPEL devices emitting 30--60 ft-L
(lOllS pulse width and 0.5% duty cycle and 120 V) for over 3000 h were
possible. This was achieved by using an encapsulation process that was
completely different from all others in the literature. The devices were heated
to 105C under a vacuum of 10- 3 Torr for 3 h. The devices were then kept
under a vacuum of 10- 6 Torr for 24 h to remove any adsorbed gases and
water. Finally they were sealed in a glass capsule with molecular sieve and
an atmosphere of dry argon.
The optoelectrical characteristics were reinterpreted by Alexander et al.
[116, 117]. They examined the standard technique of using linear plots of
both the I-V and the B-1 characteristics initiated by Alder et al. [24J in
1980. By plotting log I versus V and log I versus log B, they found another
failure mode. Figure 6.63 reproduces the plots of Alexander et al. This new
failure mode pointed towards an increasing barrier height for injection.
According to the authors, there were three important features of the
physics of the DC PEL powder device:
1. injection of electrons into the ZnS;

2. acceleration and multiplication of electrons in the formed layer;
3. conversion of current into light by impact excitation.
They considered various possible mechanisms for the electron injection
process, including
1. field-assisted activation over a Schottky barrier,

2. thermally-assisted tunnelling through a CU2 _xSjZnS heterojunction bar-
rier, and
3. Poole-Frenkel emission from defect sites in the ZnS.
,;7
,, '"
"
" co
Cl
.2 / " Cl
.2
" /
" /
I
/
(a)
V log I
'"
---
Cl /
/ '" '" CO
Cl
.2 / .2
f
'/
~
~
(b)
V log I
CO
Cl
.2 Cl
o
'I
I
f
f
(c)
V log I
".-
"...
,-
I I"
f ,/
f
,
.r
Cl
I co ,,-
.2 Cl /
f
f .2 I
I
I
/
I"
(d)
V log I
Fig. 6.63 Schematic changes in injection and efficiency plots for different degrada-
tion modes: (a) further forming; (b) load line change; (c) softening; (d) effective barrier
height change or decrease in injection site density.
Although no evidence was found for any of these mechanisms actually
operating in the DCPEL devices, all involved transferring an electron across
an energy barrier by field-assisted thermal activation or tunnelling. The
following relationship was obtained for current (I) and field (E):
I =Io(E) exp{ -e[H -F(E)]/kT} (6.17)
where H is the energy barrier for electron injection and F(E) is the
field-dependent reduction in barrier height. This is a variant of the Schottky
thermionic emission expression. The authors chose to put F(E) = A V/d
where A was a constant and d the formed layer thickness. Thus a log I
versus V plot (injection plot) was linear with the slope dependent on the
thickness of the formed layer and the intercept related to the barrier height,
H. Equation (6.16) was used for the luminance, B, i.e.
B=KIn
The slope of log B versus log I (efficiency plot) was said to show the photon
production mechanism and the intercept a measure of the quantum efficiency.
The object was to prove that there was no threshold in both the I-V and the
B-1 characteristics as advocated by Alder et al. [24]. According to Alexander
et al., the same mechanism was operating at both low and high electrical fields
in both the current and the luminance characteristics. The existence of a
threshold voltage has been argued for a long time, beginning with Destriau in
1947 [129J, Thornton in 1960 [10J, Vlasenko and Yaremko in 1965 [130J and
so on. In addition, just as any increase in the formed layer thickness would
have been shown up by a change in the slope of the characteristics, any
parallel shift in the I-V characteristics would have indicated a change in the
barrier height for electron injection (H in equation (6.17)) (Fig. 6.64).
2 40V
55 V
70 V
- 85 V
...
c:
~
::J
U
go O+-~~~'---+---T-~~-+--
....J
250
Applied voltage (V)
-1
Fig. 6.64 Parallel changes in the I-V characteristics are indicative of changes in the
barrier height for electron injection [116].
Since Alexander et al. found that, during forming, the I-V characteristics
exhibited a parallel shift, they concluded that, after the initial forming had
occurred, the thickness of the formed region does not increase. Thus either
the barrier height for electron injection was increasing or there was a
reduction in the injection site density (see Fig. 6.65).
Finally, the use of constant current drive by Cherry Electricals (section
69.2) has led to the appearance of very stable devices. The maintenance
curve of their displays is shown in Fig. 6.66.
The lack of stoichiometry leading to poor maintenance has been inves-
tigated recently [57]. This research produced very stable devices (6%
degradation in light output in 10 000 h). The devices were treated with H 2 S
gas. Although a similar model to that of Lehmann [11] for the ACPEL cells
can be proposed by invoking the reduction of Vs, the actual mechanism is
probably more subtle in the DCPEL case. We have proposed that Vs
vacancies lead to trapping centres and that the trap density affects the
conductivity of the ZnS. For example, O 2 trapped at an S site would reduce
the surface conductivity since it is an acceptor centre. Similarly, H2
adsorbed at the same site would lead to enhanced conductivity since it will
act as a donor centre. The H 2 S gas treated devices are inefficient and would
be traditionally regarded as 'soft'. This more conductive surface may provide
for a more stable injection interface by a reduction in the barrier height to
injection. The improved maintenance may also reflect the greater stability of
the conductive CU2-xS surface layer. Excess S on the surface may inhibit Cu
migration. In a recent paper, it was stated that a decrease in the number of S
vacancies leads to lower clamping field Ez , which results in lower efficiency
in ZnS thin films [131]. The situation regarding reduced power efficiency in
DCPEL panels appears to be the same although it is slightly more difficult
to be precise in this case. This is because the thickness of the formed region,
::J 2
~
Q)
u
C
<II
c
"E
..:1
0
Ci
0
-.J
-1 2
Log,o current
-1
Fig. 6.65 Injection and efficiency plots for a DCPEL device formed in air in the
range 40-85 V [116].
Application areas 219
Cherry display products corporation

panel # E042401
2
:r-l:
120.-----~.----~.----~.----~----~----~-----r
t I : 1.8
1.6 ....
:I-:::-:---:-~[-:--:-=:-~~~!~~;~.-:- :-:- ~1-;-___":__"_ ~_
....
co
1 --=-=-r_:
___ ..
_ 1.4 ~
....
IJ)
Q;
"0 1.2 ~
-
> IJ)
.J ~~~~~_------~---~~--~---~--i---------- c
Q)
E
Qi
x ! ! 0.8 ::J
ii: 40----------i...--------i__ -----------i- - ----------- ---- -----.----- en

0.6 E
~
0.4
0.2
0 I i 0
0 2 4 6 8 12 14
Hours
(Thousands)
1--- LUMINANCE -+- CURRENT - - ROW HV -e- EFFICIENCY I

Fig. 6.66 Maintenance characteristics of a typical Cherry Electrical's DCPEL
display using a constant current drive.
calculation of the dielectric constant and so on are much more difficult to

quantify.
In summary, serious characterization of DCPEL devices was not initiated
until 1980 and very little progress has been made since then apart from the
contributions of Alexander et al. [116, 117]. Of all the failure modes, softening
is the one least understood and for which no satisfactory mechanism has been
found. However, as indicated above, it could be related to the nature of the
adsorbed species on the surface of the ZnS. Effects due to further forming have
largely been overcome by the use of constant current drive.
In the last section, some of the application areas for both ACPEL cells
and DCPEL devices will be described.
6.11 APPLICATION AREAS
6.11.1 ACPEL cells

As mentioned in section 6.2, typical application areas for ACPEL cells are in
the backlighting of LCDs and plasma panels. Other uses are in fixed legends
and instrumentation lighting as in aircraft. The great versatility of the plastic
ACPEL cells is their ability to be cut into any desired shape or size. The
range of colours available is fairly wide with the use of appropriate

phosphors or by a colour shift caused by organic dyes. Figure 6.67 shows
some shapes and sizes of ACPEL lamps.
Fig. 6.67 A photograph showing typical shapes and sizes of ACPEL lamps.
6.11.2 DCPEL devices

Unlike the ACPEL cells, DCPEL panels are used for displaying alpha-
numeric and graphical information. Therefore, these have been developed
for such applications as car dashboards (Fig. 6.68), laptop computers (Fig.
6.69), dealer room applications with touch switch arrays (Fig. 6.70) and
medical scanners (Fig. 6.71) amongst others.
6.12 CONCLUSIONS
The history and developments of both AC and DC powder EL structures
have been reviewed. The bulk of the effort on ACPEL cells was in the period
1950-1963 and most of this was reviewed by Ivey [3] and Henisch [4]. The
situation has not changed much since then for the ACPEL cells, though they
are more efficient and stable than the lamps of the early 1960s. Obviously
the manufacturing processes have improved considerably since then as well.
However, their poor DR has ensured that they will not be considered for
alphanumeric and graphical applications. As a result, most of this review has
been on DCPEL devices although, where applicable, comparative reviews of
Fig. 6.68 A prototype car dashboard DCPEL display.
Fig. 6.69 A DCPEL display for alphanumeric and graphical applications, e.g.
laptop computers. (Courtesy PPC.)
Fig. 6.70 A DCPEL display with touch switch capability for dealer room applica-
tions. (Courtesy PPc.)
Fig. 6.71 A DCPEL display for medical scanner instrumentation. (Courtesy

Cherry Electricals.)
References 223
the ACPEL cells have been made. The peculiarities of the DC system, e.g.
the need for very small phosphor particles, the Cu coating process, and the
forming process as well as the injection, excitation and degradation mechan-
isms, have been covered. It has been shown that the DC system is ideally
suited for alphanumeric and graphical displays and that apparent stability
can be achieved by correct choice of the electrical drive system and the
gaseous ambient. The latter has not received the attention it deserves
considering that the surface area of the ZnS powders employed in the
manufacture of DCPEL is over 4 m 2 g-l. The stability of both the AC and
the DC powder devices is extremely dependent on the control of this
ambient.
ACKNOWLEDGMENTS
The author is indebted to many friends and colleagues in the displays
industry. Particular thanks for supplying photographs and data go to Mr Jo
Mayo and Mr Eric Olsen (Cherry Electrical), Mr Ray Ellis (Phosphor
Products Company), Mr Fred Tagg (Craft Data) and Dr Paul Beatty
(formerly Smiths Industries). He would like to express his sincere thanks to
Mr David Smith (University of Greenwich) for his helpful criticism. Finally,
he is indebted to Professor Aron Vecht for allowing him to 'dabble' in EL
and other display systems during the last 12 years.
REFERENCES
1. Destriau, G. (1936) J. Chim. Phys., 33, 589.
2. Vecht, A. (1984) SID Semin. Notes, vol. 1, section 1, p. 2.
3. Ivey, H.F. (1966) Advances in Electronics and Electron Physics, Suppl. 1, Electro-
luminescence and Related Effects, Academic Press, New York.
4. Henisch, H.K. (1962) Electroluminescence, Pergamon, Oxford.
5. Fischer, A.G. (1971) J. Electrochem. Soc., 118, 139c.
6. Leverenz, H.W. (1950) An Introduction to Luminescence of Solids, Wiley, New
York, p. 61.
7. Thornton, W.A. (1957) J. Appl. Phys., 28(2), 262.
8. Thornton, W.A. (1958) Bull. Am. Phys. Soc., 3, 233.
9. Thornton, W.A. (1959) Phys. Rev., 113, 1187.
10. Thornton, W.A. (1960) J. Electrochem. Soc., 107, 895.
11. Lehmann, W. (1966) J. Electrochem. Soc., 113, 40.
12. Larach, S. and Shrader, R.E. (1957) J. Phys. Chem. Solids, 3, 159.
13. Zalm, P., Diemer, G. and Klasens, H.A. (1954) Philips Res. Rep., 9, 81.
14. Zalm, P. (1956) Philips Res. Rep., 11,353-399,417-451.
15. Vlasenko, N.A. and Popkov, I.A. (1960) Opt. Spectrosc., 8, 39.
16. Bowtell, J.N. and Bate, H.C. (1956) Proc. IRE, 44, 697.
17. Homer, H.M., Rulon, R.M. and Butler, K.H. (1953) J. Electrochem. Soc., 100,
566.
18. Busca, G. and Cortese, C. (1971) Lett. Nuovo Cimento, 2(5), 261.
19. Yoshiyama, M., Kawarada, H. and Sato, T. (1970) Proc. IEEE Int. Computer
Group Conj, pp. 261-269.
20. Nicoll, F.H. and Kazan, B. (1955) Proc. IRE, 43, 1012.
21. Kazankin, D.N., Pekerman, F.M. and Petoshina, L.M. (1959) Opt. Spectrosc., 7,
458-459.
22. Favorin, V.N., Kozina, G.S. and Tikhonova, L.K. (1959) Opt. Spectrosc., 7,
420-421.
23. Favorin, Y.N. and Kozina, G.S. (1961) Opt. Spectrosc., 10, 43.
24. Alder, c.J. Cattell, A.F., Dexter, K., Kirton, J. and Skolnick, M.S. (1980) SID
Biennial Display Research Conf. Digest, Cherry Hill Meeting, p. 1.
25. Vecht, A and Ellis, R (1966) Nature (London), 210, 1251.
26. Vecht, A and Ellis, R (1966) Proc. Int. Conf. on Luminescence, 1966, pp.
1986-1991.
27. Vecht, A, Werring, N.J. and Smith, P.J.F. (1968) Br. J. Appl. Phys., Ser. 2, 1, 134.
28. Vecht, A. (1968) Proc. 4th Oxford Lecture Course on Electronic Components,
October 1968, vol. 9, p. 1153.
29. Vecht, A, Werring, N.J. Ellis, R and Smith, P.J.F. (1969) Br. J. Appl. Phys., Ser.
2,2,953.
30. Vecht A. (1969) Spectrum, 7, 66.
31. Vecht, A., Werring, N.J., Ellis, R. and Smith, P.J.F. (1970) J. Electrochem. Soc.,
117(1), 134.
32. Vecht, A. and Werring, N.J. (1970) J. Phys.: Appl. Phys., 3(2), 105.
33. Vecht, A., Werring, N.J., Ellis, R. and Smith, P.J.F. (1970) J. Phys. D, 3,165-166.
34. Ellis, R, Werring, N.J., Vecht, A., Smith, P.J.F. and Williamson, J.H. (1971)
Conf. on Displays, Loughborough, September 1971, pp. 205-208.
35. Waite, M.S. and Vecht, A. (1971) Appl. Phys. Lett., 19(11),471.
36. Werring, N.J., Ellis, R, Smith, P.J.F., Vecht, A. and Mears, A (1972) SID Symp.
1972, p. 170.
37. Vecht, A (1973) J. Vac. Sci. Technol., 10(5), 795.
38. Vecht, A. (1973) J. Lumin., 7, 213.
39. Vecht, A (1974) J. Phys. (Paris), Colloq. C3, Suppl. 4, 35, 173.
40. Waite, M.S. and Vecht, A. (1974) J. Electrochem. Soc., 121, 109.
41. May, D. and Vecht, A. (1975) J. Phys. C, 8, LS05-L509.
42. Vecht, A. (1976) Electrochem. Soc. Extended Abstr., 76(2), 563.
43. May, D. and Vecht, A (1976) Thin Solid Films, 36, 137-140.
44. Vecht, A., Higton, M.H. and Nicholas, D.M. (1976) Electrochem. Soc. Extended
Abstr., 76(2), 574.
45. Vecht, A., Mayo, J. and Higton, M.H. (1977) SID Int. Symp. Digest 1977, p. 88.
46. Higton, M., Vecht, A. and Mayo, J. (1978) SID Int. Symp. Digest 1978, pp.
136-137.
47. Werring, N.J. (1978) Electron, pp. 1337-8.
48. Vecht, A., Higton, M., Mayo, J. and Blackmore, J. (1980) SID Int. Symp. Digest
1980, p. 110.
49. Vecht, A (1981) SID Int. Symp. Digest 1981, p. 32.
50. Vecht, A, Waite, M., Higton, M.H. and Ellis, R (1981) J. Lumin., 24-25,
917-920.
51. Smith, P.J.F. and Werring, N.J. (1981) Proc 1st European Displays Conf., VDE
Verlag, Berlin, pp. 149-151.
52. May, D. (1981) Electronic Displays '81, Conf. Proc., London, Vol. 1,1, pp. 82-92.
53. Vecht, A, Chadha, S., Hayes, R. and Ellis, R (1982) SID Int. Symp. Digest 1982,
p.16.
54. Vecht, A. (1982) J. Cryst. Growth, 59, 81.
References 225
55. Chadha, S.S. Haynes, C.V. and Vecht, A. (1988) SID Int. Symp. Digest 1988, pp.
35-38.
56. Mayo, 1., Hutchinson, P., Hayes, R and Ellis, R. (1984) SID Int. Symp. Digest
1984, p. 22.
57. Chadha, S.S. and Vecht, A. (1988) Electroluminescence, Springer Proceedings in
Physics, Vol. 38, Springer, Berlin, pp. 337-341.
58. Mayo, J.W., Hutchinson, P., Hinsley, J.L., Alexander, P.W. and Davies, M.
(1986) SID Int. Symp. Digest 1986, Paper 17.4, pp. 313-315.
59. Kirton, 1., Lloyd, K., Sarginson, R, Smith, P. and Freer, W. (1971) 5th Annual
Conf. on Solid State Devices, University of Lancaster, September 14-16, 1971,
Paper 5.8, p. 60.
60. Mears, A.L., Parker, P. and Sarginson, RW. (1972) Solid State Devices 1972,
Paper C5.8, p. 224.
61. Mears, A.L., Parker, 1., Sarginson, R.W. and Ellis, R. (1973) SID Int. Symp.
Digest 1973, p. 30.
62. Mears, A.L. and Sarginson, RW. (1976) SID Int. Symp. Digest 1976, p. 10.
63. Alder, C.J., Cattell, A.F., Dexter, K., Dixon, M. and Kirton, J. (1980) Electron.
Lett., 16, 571.
64.Alder, c.J., Cattell, A.F., Dexter, K., Dixon, M., Kirton, 1. and Skolnick, M.S.
(1981) IEEE Trans. Electron Devices, 28(6), 680-688.
65. Kirton, 1., (1981) Proc. 1st European Displays Conf., VDE Verlag, Berlin, p. 144.
66. Kirton, J. (1981) In Handbook on Semiconductors (ed. C. Hilsum), Vol. 4,
North-Holland, Amsterdam, p. 665.
67. Skolnick, M.S. (1981) J. Phys. D, 14, 301.
68. Dean, P.1. (1981) J. Lumin., 23, 17-53.
69. Wickenden, D.K. (1980) Phys. Technol., 11,211-217.
70. Shepherd, B. (1981) Displays, 2, 338.
71. May, D. and Chadha, S.S. (1982) Conf. on II-VI Compounds, University of
Bradford, February 1982.
72. Fickiet, J.M. and Plumb, J.L. (1973) J. Electrochem. Soc., Solid-state Science and
Technol., 120(9), 1238.
73. Hanak, J.1. and Yocom, P.N. (1973) Final Report Contract No. DAAB07-
69-c-0290 for us Army Electronics Command AD-766741.
74. Glaser, D. and Kupsky G.A. (1986) SID Int. Symp. Digest 1987, Paper 16.2, p.
292.
75. Mayo, 1. and Channing, D. (1990) Inform. Disp., 6(2) 6-8.
76. Yoshiyama, M., Kawarada, H. and Sato, T. (1971) 17, 670-676.
77. Yamamoto, Rand Yamazoe, H. (1973) J. Appl. Phys., 44(7), 3191-3192.
78. Kawarada, H. and Ohshina, N. (1973) Proc. IEEE, 61(7), 907-915.
79. Fujii, H. and Hoshima, T. (1983) Japan Displays '83, Paper 4.6, p. 96.
80. Luo, X. and Zhou, L.X. (1981) Extended Abstracts from the May 1981 Elec-
trochem. Soc. Meeting, Abstract No. 173, pp. 431-433.
81. Xu, S., Chou, L. and Luo, X. (1982) SID Int. Symp. Digest 1982, Paper 4.2, pp.
18-19.
82. Chou, L., Lou, X. and Xu, S. (1982) Extended Abstracts, Electrochem. Meeting
Soc. 1982, Abstract No. 519, p. 847.
83. Wu, B., Chern, 1., Hong, X. and Liu, F. (1984) J. Lumin., 31-32, 951-953.
84. Zhongyi, Y., Weidong, H., Jingpu, L., Nanggang, Z., Wei, Z. and Zhiquan, L.
(1983) Japan Displays '83, Paper 3.6, p. 586.
85. Abdalla, M.I., Godin, A. and Noblanc, 1.P. (1979) J. Lumin., 18-19, 743-748.
86. Abdalla, M.I., Brenac, A. and Noblanc, 1.P. (1980) SID Biennial Display Research
Conf. Digest, Cherry Hill Meeting, pp. 174-176.
87. Pecile, D., Brunei, c., Thomas, l and Godin, A. (1982) Extended Abstracts,
Electrochem. Soc. Meeting 1982, Abstract No. 172, pp. 428-430.
88. Abdalla, M.I., Godin, A., Brenac, A. and Noblanc, lP. (1981) IEEE Trans.
Electron Devices, 28(6), 689-693.
89. Destriau, G. (1937) J. Chim. Phys., 34, 327.
90. Jaffe, P.M. (1961) J. Electrochem. Soc., 108, 711.
91. Gobrecht, H. and Gumlich, H.E. (1956) J. Phys. Radium, 17, 754.
92. Gobrecht, H., Gumlich, H.E., Nelkowaki, H. and Langar, D.Z. (1957) Z. Phys.,
149,504.
93. Vecht, A., Smith, D.W., Chadha, S.S., Morton, D.C. and Miller, M.R. (1990)
Electrochem. Soc. Extended Abstracts, October 1990, No. 625, pp. 896-897.
94. Brown, R.A. (1968) Electrochem. Technol., 6(7-8), 246-251.
95. Willard, H.H. (1956) Anal. Chem., 22(11), 1372-1374.
96. Jackson, P.A. (1967) British Patent 1,071,062 (Cl. C09K) (June).
97. Cusano, D.A. (1962) In Luminescence of Organic and Inorganic Materials (eds
H.P. Kallman and G.M. Spruch), Wiley, New York.
98. Aven, M. and Parodi, J.A. (1960) J. Phys. Chem. Solids, 13, p. 56.
99. Nickerson, J., Goldberg, P. and Baird, D.H. (1963) J. Electrochem. Soc., 110(12),
1228-1230.
100. Gushurov, G. and Banks, E. (1967) J. Electrochem. Soc., 114(11), 1143-1148.
101. Kawamura, T., Hironori, T. and Kazuhiko, T. (1989) J. Electrochem. Soc.,
136(4), 1229-1234.
102. Nitsche, R. (1971) J. Cryst. Growth, 9, 241.
103. Juze, R., Rabrean, A. and Paschier, G. (1956) Z. Anorg. Allg. Chem., 285, 61.
104. Tornqvist, R. (1983) J. Appl. Phys., 54, 4110.
105. Kitai, A.H. (1988) J. Lumin., 39, 227.
106. Miyata, T., Nakagawa, S., Hirayama, H. and Hasegawa H. (1970) Jpn. J. Appl.
Phys., 9, 615-624.
107. Goldberg, P. and Nickerson, J.W. (1963) J. Appl. Phys., 34, 1601-1608.
108. Chadha, S.S. (1981) Unpublished results.
109. Thornton, W.A. (1959) J. Appl. Phys., 30, 123-124.
110. Gaudin, A.M., Fuershenan, D.W. and Mao, G.W. (1959) Min. Eng., (April), 11,
431-436.
111. Smith, P.J.F. (1973) M.SC. Thesis, Thames Polytechnic.
112. Kelly, J. (1979) M.Sc. Thesis, Thames Polytechnic.
113. Hartmann, H., Mach, R. and Selle, B. (1982) Curro Top. Mater. Sci., 9, 224.
114. Werring, N.J. (1973) Ph.D. Thesis, Thames Polytechnic.
115. Beatty, P.H.l (1970) M.Sc. Thesis, Thames Polytechnic.
116. Alexander, P.W., Sherhod, C. and Stowell, M.J. (1988) J. Phys. D, 21, 1627-
1634.
117. Alexander, P.W., Sherhod, C. and Stowell, M.l (1988) J. Phys. D, 21, 1635-
1641.
118. Beatty, P.H.J. (1974) Ph.D. Thesis, BruneI University.
119. Cabrera, N. and Mott N.F. (1949) Rep. Prog. Phys., 12, 163.
120. Piper, W.W. and Williams, F.E. (1955) Br. J. Appl. Phys. (Suppl.), 6, 39.
121. Taylor, J.B. (1954) Br. J. Appl. Phys., 45.
122. Zallen, R., Eriksen, W.T. and Ahlburg, H. (1960) J. Electrochem. Soc., 107,
288-95.
123. Chadha, S.S., Smith D.W., Vecht, A. and Williams, lL. (1991) SID Technical
Digest, Paper 6.6, pp. 80-83.
124. Waite, M.S. and Thomas, A. (1980) J. Electrochem. Soc., 127(2), 509.
125. Brody, T.P. and Yu, K.K. (1975) IEEE Trans. Electron Devices, 22,739.
References 227
126. Inoguchi, T., Takeda, M., Kakihara, Y., Nakata, Y. and Yoshida, M. (1974) SID
Int. Symp. Digest 1974, p. 84.
127. Faria, S. (1988) J. Electrochem. Soc., 135(10), 2627.
128. Roberts, S. (1957) J. App/. Phys., 28(2), 245.
129. Destriau, G. (1947) Phi/os. Mag., 38, 700.
130. Vlasenko, N.A. and Yaremko, A.M. (1965) Opt. Spectrosc., 15, 263~266.
131. Hirabayashi, K., Shibata, T. and Kozawaguchi, H. (1988) IDRC 1988, San
Diego, CA, p. 42.
7
Thin film electroluminescence
devices
R. Mach
7.1 INTRODUCTION
Thin film electroluminescence devices (TFELDs) are discussed in the litera-
ture at present mostly under the heading 'flat panels'. It will be shown that
this is by no means comprehensive; other applications are feasible and
upcoming. Nevertheless, the TFEL Display is on the marketplace, bright
enough for us to predict a bright future for it.
We are going to discuss in what follows the device engineering aspects.
Concerning the underlying physics, the reader is referred to other chapters
of this book (Chs 1, 2 and 5).
The most exciting feature of any TFELD is that it is all-solid-state (Fig.
7.1). The structure is simple. One substrate - glass, in general - is covered
by a stack of thin films of dielectric and semiconducting or phosphor
materials, which are sandwiched between electrodes, one of which is
Top electrode (meta l) ---~.
Insulator -...../-:Jr::='-J"",,,,'-'
Semicondu rTnr --,-
Glass
Fig. 7.1 Schematic of the physical structure of the thin film EL display.
229
230 Thin film electroluminescence devices
transparent. The films are polycrystalline or in the case of the insulators

mostly amorphous. They are optically homogeneous and transparent,
except for the metal electrodes. The substrate is not required to be flat to
stringent tolerances, it can be curved, concave or convex, or of even more
complex shape.
Polycrystallinity means that cheap production - vacuum or chemical
vapour deposition - techniques may be used, leading to cost-effective
devices, even if large areas are wanted.
As the name says, the TFELD emits light under the action of an applied
electric field; in contrast to injection electroluminescence these fields have to
be much higher - of the order of some megavolts per centimetre - and the
underlying process is impact excitation of inner shell transitions [1].
This chapter is organized as follows. In section 7.2 an overview of existing
devices and device concepts will be given, going beyond the display. As it is
not possible to arrange the information in a one-dimensional array, section
7.3 contains concepts for multi- and full colour, besides the knowledge
gathered on single-colour materials. The next section is devoted to the
discussion of possible deposition techniques, their merits and drawbacks, as
far as these can be judged, before large-scale industrial production has gone
beyond ZnS:Mn devices. Since for each commercial product aging is an
important issue, section 7.5 tries to review the few papers published on this
subject, stressing the knowledge obtained on degradation mechanisms.
Finally, in section 7.6, conclusions will be drawn.
It should be noted from the beginning that a chapter devoted to devices
is highly time dependent; it represents the information available at the end
of 1991, and especially concerning industrial production it might be out-
dated quickly. Therefore the concepts of the various devices will be stressed
more than actual parameters.
At present there is large-scale industrial production of monochrome
ZnS:Mn-based diplays, with samples of multicolour displays demonstrated
to the public. When the book comes to the reader, multicolour VGA TFEL
displays may be available.
7.2 DEVICES
7.2.1 Displays
TFEL displays are matrix organized, in a manner shown by Fig. 7.1, and
addressed as shown in Fig. 7.2. The thin film stack of an insulator -
semiconductor (phosphor) - insulator (lSI) sequence is deposited onto a first
system of (bottom) electrode stripes, running parallel to each other. On top
of it a second set of electrode stripes, perpendicular to the first one, is
deposited. A picture element (pixel) is defined by the crossing of top and
bottom electrodes. If voltages of different polarities with respect to a common
Devices 231
f!
ltn . . .
r~
n
012
time
.
Fig. 7.2 The addressing scheme, a line at a time, mostly used in flat panels.
(ground) level are applied to the electrodes, they add across the pixel. This
is the basis of addressing or driving. The most common way to drive any of
the matrix-organized flat panels is a line at a time [1]. In contrast to cathode
ray tubes (CRTs), in which the addressing electron beam scans the pixels in
a row sequentially (and the rows in a frame scan), in a TFEL display all
pixels in a row are addressed simultaneously. Consequently the dwell time
per pixel is independent of the number of columns; it is only determined by
the number of lines, m, as the frame time divided by m.
As is well known from liquid crystal (LC) displays, problems might arise
with such a simple concept of addressing if the response of the active
material is not prompt, and no real threshold behaviour is encountered. If
for instance the response is to the effective voltage, as for LCs, and is rather
soft, almost linear, the multiplex rate, or number of addressable lines, is low.
In the TFEL case, however, the response is instantaneous and steep, so that
1000 multiplexable lines are easily reached. If the frame time is 20 ms, a 20 f.lS
dwell time for each pixel imposes no problems on the luminescence,
although possibly on the driver circuitry.
The steepness of the response of TFEL devices to an applied voltage is
shown in Fig. 7.3. The brightness-voltage characteristic B(V) is shown with
a suppressed zero of the voltage scale, to make it meaningful. Four orders
of magnitude in brightness are bridged by 20 V, or 10% of the applied
voltage. Fortunately enough, in spite of this threshold behaviour, no
bistability is encountered. Grey scale, or the selection of defined brightness,
is achievable by amplitude selection. To evidence this, the B(V) curve of Fig.
7.3 is redrawn in Fig. 7.4 on a linear brightness scale: putting a select pulse
of 200 Von the row, information pulse amplitudes between 10 and 50 V can
give grey scaling.
232 Thin film electro luminescence devices
ZnS:Mn
100 Hz
250
Fig. 7.3 Brightness-voltage characteristic of a typical ZnS:Mn TFEL sample under
100Hz bipolar pulse drive, showing the extreme steepness of response.
600
ZnS:Mn
300 100 Hz
250
Fig. 7.4 The brightness-voltage characteristic of Fig. 7.3, redrawn on a linear
brightness scale, to demonstrate the easy grey scaling by voltage modulation.
From an applicational point of view it is the combination of steepness,

avoiding cross-talk, and almost linear amplitude response above threshold,
allowing for grey scale, which makes the TFELD so attractive.
There is however one very important second device issue, which gives the
TFEL display an advantage compared with other flat panel technologies: the
simplicity of its structure. It is a stack of three laterally unpatterned,
homogeneous films, between two one-dimensionally extended systems of
electrode stripes. No patterning process, except for the electrodes, is needed.
As these electrodes are one dimensional, not even a step-and-repeat process is
needed in photolithography (or screen printing). This holds for monochrome
devices; whether it stays true for colour will be discussed in that section.
The big drawback of TFELDs is the rather high capacitances encoun-
tered. The capacitive load of the drivers is much higher than in the case of
Devices 233
other flat panels. However, drivers are available at about 5 and 10 cents for
one column and row respectively, an appreciable cost item. There has been
much discussion about reducing the voltage needed for operation. The
feeling behind it was the belief '300 V transistors are expensive'. Certainly
such a transistor uses more area on an Si chip, which counts in large-scale
production. There are, however, some trade-offs which need to be consider-
ed. The first is that for a given efficiency the light output can be achieved by
a certain electric power consumed; this can be supplied either by a high
voltage at low current or vice versa. The high current version, however,
would impose insurmountable problems with the transparent electrodes.
They are at their limit even at 200 V with diagonal dimensions of 14 in. On
the other hand, reducing the voltage requirement means that the capacitive
load is increased. This is quite obvious: the field needed in the active film is
set by physics; reducing the thickness of the active film reduces the voltage
but also the brightness for a given transferred charge. Either the voltage
must be increased to increase the transferred charge, or the frequency;
increasing the use of the applied voltage by improving the capacitive
division ratio between dielectrics and active films increases the total
capacitance. In both cases the capacitive currents increase and so do the
losses on the transparent electrode resistance. Inhomogeneities of brightness
arise from these voltage drops.
To give some figures, a typical, well-designed structure has a total
capacitance of about 10nFcm- 2 . The capacity seen by a row or column
driver depends on the details of the driving scheme; for one obvious solution
the calculations have been done by Miller and Kelley [2]. One result,
surprising at first sight, is that the column load is at a maximum when half
of the pixels are lit.
Little has been written in connection with Fig. 7.1 about the location of
the transparent electrodes. They are shown to sit on the substrate glass. This
is the conventional design. Recently in the discussion of colour the advan-
tage of this placing has been questioned. At this point, however, the
discussion of the insufficient conductivity of indium tin oxide (ITO) or
ZnO:Alleads us to mention the tendency to invert the classical structure by
depositing the transparent electrode on top (Fig. 7.5). Below the lSI stack a
thick ITO stripe brings about appreciable 'climbing' of the next films over
the edges of these stripes, giving rise to inhomogeneities and field concen-
trations on the edges, etc. The metal electrodes on glass with only 50 nm
thickness are more convenient. On top of the MISI, on the other hand,
thicker ITO does no harm. Also, even reinforcements by very narrow
metallic lines are more easily made on top.
Available from large-scale industrial production at present is a large
variety of monochrome TFELDs, based on ZnS:Mn as the active material.
Table 7.1 lists the main dimensions, pixel numbers and arrangements, and
the brightnesses of the commercial products. The largest TFELD available
Top Electrode (ITO) - - ! I 1 . _ ._ _ ~~~:::

Insulator-
Semiconductor -
Insulator-
Bottom Electrode (metal)-'!:'~~~~~~~~::::(I"
Glass
Fig. 7.5 Inverted structure of a TFELD; the transparent electrode (e.g. ITO) on top
of the stack has many advantages.
Table 7.1 Some commercial display products (end 1991)

Planar Sharp Planar Sharp
Matrix 640 x 480 640 x 480 1024 x 864 1024 x 800
Active area (mm 2 ) 206 x 154 191 x 143 345 x 292 255 x 200
Brightness (ft-L) 35 30 18 30
(pixel)
Colour peak 585 585 585 585
wavelength (nm)
Power (W) 12 22 85 22
Operating o to 55 -5 to +55 10 to 40 o to 55
temperature (0C)
Mass (g) 708 620 7700 1000
Grey scale 16
Viewing angle (deg) 160 160 160 160
and used in some quanttttes by Digital Equipment Corporation for a

workstation is 19 in diagonal and has a resolution of 1024 x 864 pixels (Fig.
7.6(a)). It is of excellent contrast and appearance, and completely free of
geometric distortions, of course.
That grey shades are no future feature, but reality, is shown in Fig. 7.6(b),
exhibiting a display manufactured by Sharp.
There is a strong tendency of the manufacturers to build more electronics
around the display; complete TFEL monitors are evolving, and the avail-
ability of cheap microelectronics from a multitude of producers, much
greater than the number of TFELD manufacturers, might lead to 'elec-
tronics around a display' instead of 'a display for some electronics'.
Also, there is one strong technical argument; the TFEL display, in contrast
to the analogue CRT, is rather digital. So a perfect compatibility, or 1:1
Devices 235
(a)
(b)
Fig. 7.6 Photographs of the top TFEL products of 1991: (a) a 19 in diagonal
1024 x 864 pixel display of Planar, USA; (b) a 640 x 400 pixel grey scale display made
by Sharp, Japan.
exchangeability with a CRT, which was one goal earlier, may be a poor way
of viewing TFEL.
The extreme simplicity of the TFELD with only one glass (or other
substrate) and no filling as for plasma panels or LCDs avoids the need of
sealing seams around the active area, and lends itself to 'tiling' of a large
display from smaller modules. Of course, a problem with the electric
connections arises if one tries to go beyond four quarters which is simple, as
Fig. 7.7 illustrates. The difficulties arising from the optical discontinuity of
the meeting edges can be overcome, as shown on small samples [3].
~ ..
~
RGB-modul.l I
,:--
RGB-modul.3
/J
L-
~~ .. r::-'""7' .
select =
=
register =
=
3
1
- ~ I-- V
select :::
:::: 4
register
2
=
:-
_V V
... --0 o~J
i###
I RGB-modul.2 I RGB-modul.4
1--
Fig. 7.7 Tiling of large displays from four quarters is easy, as only two edges are
needed for contacts, and no sealing seams are necessary [3].
For completeness it should be mentioned that TFEL displays can be

made using two transparent electrode systems. They are transparent, and
can be used, for instance, as overlays over maps or other objects, to which
they should add temporary or changing information. They could be used on
eye glasses or as helmet displays, as pixel dimensions of 10 J.l.m x 10 J.l.m can
be realized. Even 3D displays based on this appear feasible [4J, as left-eye
pictures can be displayed on the left glass and right-eye on the right one, a
direct view being nevertheless maintained.
A factor of roughly 2 in the brightness is lost, by losing the reflectivity of
the metal back electrode, but contrast can still be high.
7.2.2 Plane, structured, pulsed light sources

Much effort is usually required to make, from conventional light sources,
necessarily not point like, secondary light sources extending homogeneously
over a well-defined area. Very often they are needed for sensing and/or
complex illumination purposes in high level applications such as object
recognition or distance or angle coding. Often these areal lights should be
pulsed or modulated in time.
TFEL obviously is a plane light source and, thinking of one pixel, it can
Devices 237
be arbitrarily shaped by photolithography of the electrode(s). Rather com-
plex geometries are easily achievable by using patterns on both electrodes.
As mentioned in section 7.2.1 completely transparent elements can be made
using two transparent contacts.
In those applications often rather high brightness is wanted. Higher drive
frequencies allow for this. Using ZnS:Mn this leads to little ripple as the
decay time can be as long as milliseconds, if the doping level is kept low. As
explained in ref. 1, however, this material shows cross-relaxation, which
reduces the efficiency on the one hand and on the other shortens decay and
produces some ripple. Careful design according to the special specifications
is therefore needed.
To give an example of what is possible in matching device requirements of
high speed, high brightness and low ripple (Fig. 7.8), the following data [5]
are for an area light emitter that emits light pulses:
brightness 30000 cd m 2
leading edge (10% to 90%) 0.1 ms
trailing edge (90% to 10%) 0.3 ms
ripple <7%
spatial brightness variation < 5%
ZnS:Mn
8 kHz
B,
10' cd m-
time, 500 ~s/div
Fig. 7.8 Rather extreme brightness values can be reached if the driving frequency is
increased and design is carefully tailored to the specifications [5].
Addressing was with 8 kHz pulse bursts between 4 and 10 ms, but the
repetition frequency was not higher than 10 Hz. Some thousand hours of
useful life were achieved under these conditions, which can be termed 'stress
aging'. The doping concentration had to be carefully chosen, and a special
design of the film stack was necessary.
Certainly in future the potentialities of this TFEL application will be
much more exploited.
7.2.3 Edge emitter

The handicap of the TFEL display, the only 10% outcoupling of the light
produced in the high refractive index material of the active film [1] can be
converted into an advantage when using TFEL for edge emitters [6]. The
edge emitter (EE) is the active device for an optical print head which is likely
to replace laser print heads in what is now a laser printer (Fig. 7.9). Either
proximity exposure or optical imaging of the emitter edge onto the print
drum might be used.
Dielectrics
Fig. 7.9 The light guided in the thin film stack along the substrate plane can be
emitted from an edge, addressing photo printers a line at a time [6].
The main idea in this concept is the fact that light guiding along the plane
of the film stack gives access to a much deeper (thicker) zone of the active
film, increasing the light output relative to that from the plane of a display
by factors of some thousands. While the emitting depth of the display is the
thickness of the active film (about 1 Jlm), the length of electrode stripes is the
relevant dimension in the case of the EE. However, severe problems were
encountered when attempts were made to achieve this. It turned out that, in
the usual ZnS:Mn-based devices, so successful for displays, the attenuation
length for the light travelling along the film plane is in the order of some
hundred microns only. Also, this is not because of the scattering out of the
plane, mentioned in ref. 1, but because of absorption. The reason for the
absorption is not really understood, but it seems to be tied to the manganese
centre itself.
As well as the principal advantages of no moving parts, greater expected
lifetimes, solid state design and ease of adjustment, the edge emitter offers
higher resolution (> 1000 lines in - 1), grey shading, and rather high speed,
with the achievable values for the latter not yet assessed or published. On
first glance the single elements (stripes) seem to need a driver each. An
Devices 239
elegant method of multiplexing, however, was proposed by Leksell et al. [7],

reducing their number by an order of magnitude.
7.2.4 Hot electron cold cathode

The hot electron cold cathode (HECC) is the youngest application of the
thin film stack and the materials characteristic for TFEL. It dates back to a
first publication [8] on the direct measurement of the electron energy
distributions in TFEL structures. This was done by emission of the electrons
under investigation into vacuum through a thin gold electrode, and using a
conventional retarding grid analyser. This is the concept of HECC, depicted
in Fig. 7.10. The result of this and further investigations [9, 10] can be
summarized as follows (see also ref. 1). Loss-free acceleration of electrons in
the materials useful for TFEL together with the avalanche multiplication
leads to an inner electron energy distribution (EED) around 12 eV in ZnS,
and around 8 eV in ZnSe (Fig. 7 in ref. 1). These electrons are able to
surmount the work function of a metal electrode (e.g. gold) easily, and
penetrate the metal according to a transmission law with strongly energy-
dependent transmission coefficients, acting as a high energy cut-off filter.
The cut-off energy depends strongly on the thickness of the metal electrode.
According to the working principle of TFEL [1], the emission is pulsed in
nature. The time dependence of the flux of electrons with the energies to be
read from the retarding voltage is shown in Fig. 7.11.
The application potential stems from the obvious possibility of using the
HECC as a plane, matrix-organized multiple cathode for a flat CRT with
of electrons
a: collected electrons
o
5 I
t
::::l
en
Z Au ...... ..
" ............... - -
. . ' ...... . g
Fig. 7.10 Schematic of the hot electron cold cathode, and the arrangement to
measure EEDs [9, 10].
40mV, ....... j .. .
i(t}
I'
n Ivac
U!
20mV
/div .
~
!not!
trig'd
~
-160mV--~--~__~~__~__~__~__~~__~
o time, !-Is 100
Fig. 7.11 Comparing the time behaviours of the dissipative current i(t) and the
vacuum current I vac with the retarding grid tuned to the voltage value given near the
curve.
nothing more than a d.c. voltage, of say 20 kV, between the thin electrode of
the HECC and the top A1 coating of a conventional CRT (colour) screen.
The pixels of the HECC should have a 1:1 correspondence to the pixels of
the screen. Usual a-line-at-a-time addressing of the HECC gives the screen
phosphor in each pixel one line time to be excited, or at least the HECC
emission time (Fig. 7.11), which is much longer than the time the electron
beam 'rests' on each pixel during its line scan. The price which has to be paid
is that of the driving circuitry, more expensive than the deflection coils and
(hot) cathodes of the CRT. First estimates, however, indicate that the
brightness levels could be raised [10].
The application of the HECC was addressed in two other papers almost
at the same time [11, 12].
Thinking of simpler applications first, one very attractive feature of the
HECC is its ability to be of arbitrary geometry: a line focus, which is quite
difficult to achieve from the usual electron optics, is no problem at all.
Miniature electron guns become possible, consisting of nothing more than
the patterned HECC and something to be bombarded a suitable distance
from it. Two lines perpendicular to each other, exciting a commercial colour
TV screen, are shown in Fig. 7.12. It is evident from this first trial that such a
resolution is no problem to achieve.
Monochrome and colour 241
Fig. 7.12 A conventional colour TV screen excited by electrons emitted from two
HECC lines, perpendicular to each other, to demonstrate the application potential.
7.3 MONOCHROME AND COLOUR

All the devices discussed in section 7.2 can be based on various active
materials, giving rise to different emission colours (or EEDs in the case of the
HECC, which we shall not discuss in this section). Active material in this
context is understood to be the combination of the host material and the
specific dopant(s}.
7.3.1 Monochrome
(a) ZnS:Mn
As mentioned earlier, ZnS:Mn is the material first used for TFEL and on
which the commercial production of TFELD at the end of 1991 was solely
based. The reason is historical, but were this material not as forgiving as it
is, TFEL probably would not have reached this stage of development.
ZnS is easily vacuum deposited without too much decomposition. It is not
able to deviate appreciably from stoichiometry, and it is completely miscible
with MnS. The latter means that no clustering or precipitation of the dopant
is to be expected and was never reported. ZnS:Mn crystallizes at low
temperatures even more readily than undoped ZnS. There are two crystallo-
graphic phases of ZnS known, the hexagonal wurtzite and the cubic
zincblende. Both incorporate Mn 2+ on cation sites. The first excited stage of
the dopant, about 2.4 eV above the ground state, is crystal field split and
highly metastable, as the radiative transition to the ground state is dipole
forbidden. It has a radiative lifetime depending on the crystal structure [13]
of about 1.5 ms. The emitted broad band is amber or yellow-orange and the
spectrum is given in Fig. 7.13. Concentration quenching sets in at about
10 20 cm - 3, which is due to increasing excitation travelling in the dopant
system by quantum-mechanical transfer, reaching eventually a killer centre.
This concentration quenching is obvious from the faster decay, which is no
longer exponential [14,15].
intensity, a.u.
100r--------------..------------.
50
500 700nm
Fig. 7.13 Emission spectrum of a ZnS:Mn TFEL structure.
The dopant concentration for maximum efficiency depends on the de-

sired brightness, as there is a nonlinear quenching process present in the
Mn system [1]. At the brightness needed for display operation
(lOOcdm-2) it is about 2 x 102cm- 2, and the efficiency achieved is about
71m W- 1 [16].
It is worth mentioning that there is no standardization of the technologies

even among the industrial producers with respect to the deposition technique:
electron beam evaporation, thermal evaporation and atomic layer epitaxy
(ALE) are used. The insulator films on the other hand are mostly d.-sputtered
Si(O, N) films. Only in one case is ALE (AI, Ti) oxide used.
(b) ZnS:Ln
Ln stands for the lanthanides or rare earths. They have been tested for quite
some time, as they are the only possibility of making ZnS emit at other
colours than the yellow of Mn 2 +. Other transition metals are not known to
emit in the visible, if doped into ZnS. The big risk with all the rare earths in
ZnS is their small solubility, said to be some parts per million. Segregation
or clustering is likely to occur, even if it is possible to incorporate
concentrations above this limit in deposition processes far from the ther-
modynamical equilibrium (TDE).
Nevertheless, good efficiency and brightness values have been reported for
ZnS:Tb3+ or, to indicate how it was prepared, usually ZnS:TbF 3. Its green
emission in narrow lines of the (forbidden) 4f-+4f transitions (Fig. 7.14) is
attractive for display applications.
intensity, a.u.
100.-----------~--------------~
ZnS:Tb
50 -
500 700nm
Fig. 7.14 Emission spectrum of a ZnS:Tb TFEL structure.
The problems with the solubility seem to be reflected by a strong

sensitivity of the reported results to details in the preparation cycle,
especially a strong sensitivity to annealing processes, which do not always
act to improve the brightness or efficiency [17, 18]. It seems to be
characteristic of this system that degradation is faster for higher initial

efficiencies. One recent achievement in this material is the inhomogeneous
doping in an ALE process, reported by Hiirkonen et al. [19]. Stable
operation of these devices at efficiencies slightly above 0.2 1m W - 1 have been
claimed, and the start of industrial production has been announced.
Thulium is the only known dopant of ZnS emitting in the blue in TFEL.
However, this emission is so weak that efficiencies could not be measured, or
at least were not reported. As there are solutions for red on a ZnS:Mn basis,
and for green with ZnS:Tb, the blue of ZnS:Tm would be highly desirable.
Recently efficient photoluminescence of ZnS:Tm films was reported, but on
the same films only some infrared emission of Tm3+ was recorded in
electroluminescence [20]. The explanation given takes into consideration
some speculative EED, which eventually could excite the infrared emission
but not the blue lines.
Concerning the last paper, it could be a counterpart to similar experi-
ments on ZnS:Te. Te 2 - is well known as an isoelectronic impurity on S sites,
binding excitons by short-range forces. At higher Te concentrations especial-
ly, the exciton bound by about 0.6 eV to a pair of Te ions gives rise to a nice
blue emission, studied in single crystals recently by Heimbrodt and Goede
[21]. Using a special method of vacuum deposition it became possible to
obtain good blue photoluminescence from ZnS:Te films. However, all
attempts to obtain electro luminescence failed. From experiments on photo-
luminescence under the action of applied pulsed fields it became evident that
field quenching was the reason [22]; see also ref. 1.
As in injection electroluminescence, but partly for other reasons, blue
always has been a problem. The best candidates at present are SrS:Ce, or,
for even better colour coordinates, SrSe:Ce. While there is an absolute
necessity to filter the green away from SrS:Ce, which with increasing Ce
concentration becomes more pronounced, SrSe:Ce might be regarded as
'almost blue'. The most concise examination of Ce 3 + emission, shifting with
the decreasing crystal field in the hosts CaS, SrS, SrSe to higher energies, has
been made by Oseto et al. [23]; photoluminescence spectra obtained from
their thin films, used also for electroluminescence, are shown in Fig. 7.15.
The best reported efficiency for SrS:Ce is still 1.3 1m W- 1 [24], obtained
on d.-sputtered films of Ohnishi and Okuda [25]. The best reported
brightness values (60 Hz) of 65 cd m - 2 decreased to 6 cd m - 2 when appro-
priately filtered to blue [26].
It was tempting to use thulium as a dopant in SrS, but no breakthrough
can be reported; supposedly the same problems as with ZnS:Tm arise [27].
7.3.2 Multicolour, full colour

Combining different colours separately addressable in a TFEL device gives
a multicolour or full colour device, depending on the possibility of compos-
CilS : Ce thin film

Monitor
Excitation
at 570nm
at 450 nm
<= (2 )
=>
300K
~
~
::J SrS : Ce thin film ::J
-
tV
'-' cti
'-'
u
>- 420nm >-
c:
-
CD 'iii
c:
'u
:e
CD
CD
c:
c:
=>
c:
-
0 0
'Z:; 'iii
tV Ul
'ux 300K 'E

w
w
SrSe : Ce thin film
420nm
300K
400 450 500 600

Wavelength [nm]
Fig. 7.15 Emission and excitation spectra of Ce 3 + in different hosts, demonstrating
the crystal field shifts, and the possibilities of SrSe for a 'good' blue [23].
ing a limited range of colours or virtually all colours respectively, The

limitation on multicolour may arise from
1. not all basic colours - red, green, blue - being available, or

2. only a very limited number of grey scales being available,
The term 'full colour' is used if there are at least 64 grey shades possible in
each of the basic colours, In what follows we shall use the term 'colour'
display or devices as meaning either multi- or full colour.
Two different concepts are followed in achieving colour TFEL devices:
I, patterned phosphor, or
2, colour by filtering.
The first, 'patterned phosphor', is the concept used in colour CRTs. As

illustrated in Fig. 7.16, different materials for the active film are used, usually
columnwise patterned photolithographically from sequentially homogene-
ously deposited active films. To obtain full colour three different materials
have to be used, emitting in the basic colours.
Electrodes
--=. .=',.,=. ========::::;::-"'" . . '
"""-.. _.----=::=====::::::;'......
,~,7
Insulator
Patterned
phosphor
Fig. 7.16 Exploded view of a patterned phosphor colour TFELD.
The first demonstration of this concept in form of a 'Color Exerciser' was

given by Planar [28]. Dry etched stripes of ZnS:Sm for red, ZnS:Tb for green,
and SrS:Ce for blue were used, and very low luminance and efficiency values
were reached, but feasibility was assessed. Certainly valuable experience with
the driving electronics was obtained. The subsequent version [29] with better
technology and more suitable materials exhibited substantial progress. In
1991 another approach to full colour was demonstrated by Oki [30].
At the same conference NEC demonstrated a patterned phosphor multi-
colour approach [31], using sputter etching to make stripes of ZnS:Mn for
yellow-orange, and ZnS:Tb for green. Using half-luminance ('grey scale') of
the two 'basic' colours, six colours could be expressed.
Figure 7.17 shows the much simpler concept 'colour by filtering': a
laterally homogenous broad band emitting active film is column wise del-
egated to different colours by filters deposited in alignment with that
column. This illustrates the full-colour version, if enough grey shades are
provided by the addressing system.
Several positions have been proposed for the filters. The first paper [32]
proposed organic filters on top of the glass, naturally introducing colour
errors by parallax. Another possibility [33] was inorganic interference filters
beneath ITO and glass, giving rise to angle-dependent colours.
......
"" -===============;:"",.:--
,--="':::"========::::::.;:::- ",
..... .. ......
~---:======:::.
,~,7
1,4I'~:Ph"
,~,7
Patterned
Colour
Filters
Fig. 7.17 Exploded view of a colour TFELD using the colour by white or
patterned filter concept.
The inverted structure with ITO (or other transparent) electrodes on top
of the lSI stack, patterned with the filters used as the etch mask for the ITO
(self-aligning, parallax-free system), was proposed recently. It is illustrated in
Fig. 7.18. Testing of the device can be done before the filters are deposited;
screen printing in a flow process of organic filters with high colour
saturation may be used before wet etching.
Very recently multicolour displays were demonstrated by Planar [34] and
by Sharp [35]. Both used colour filtering at least partially. The Planar
demonstrator is shown in Fig. 7.19. It uses a further method of filter
arrangement. The red part of the Mn emission, which is red shifted
compared with the spectra shown in Fig. 7.13 if the Mn concentration is
high, was used for a standard red using an inorganic Cd(S,Se) film below the
ITO on glass, acting as a short wave cut-off. Green emisson, however, was
obtained from ZnS:Tb 3 + made as mentioned previously by inhomogeneous
doping in ALE. So this in fact is a patterned phosphor with filters.
The Sharp example uses filtering from the broad emission spectra of
ZnS:Mn of green and red in an inverted structure, with organic filters (Fig.
7.20) on a second glass, which is put face to face on the inverted structure's
ITO stripes. The visual impression of the example is very good, mainly
because of very good brightness, which seems to be the result of a novel
CVD process [36]; black stripes along the edges of the various colour stripes
increased the contrast. A filtered white phosphor device was introduced by
R G B
"',, '/'~ '~ITO,'. . ' . . ." "
Glass
Fig. 7.18 Inverted structure with parallax-free filters on top, which can be used as
an etch mask for the transparent electrode stripes. The white or broadband emitter is
shown as composed of two differently doped layers in one host.
Aluminum row
Top insulator
Illill~~11 ~Iu~ii~i~= ZnS: M n Phosphor

W ZnS -Tb Phosphor
1 st insulator
ITO Column
ALE Barrier
Red filter
Fig. 7.19 A recently demonstrated patterned phosphor, filtered-emission display

structure of Planar [35].
Tanaka et al. [32]. The white emission was achieved by double doping of
SrS with Ce 3 + (blue-green) and Eu 2 + (red). The separated colours are
shown together with the 'white' in Fig. 7.21. Rather poor efficiency and
brightness were achieved as transfer from Ce 3 + to Eu 2 + limited the
concentrations of the dopants to a rather low use of the hot electrons [1].
Presumably it is much better to separate the two dopants into layers of the
same film, as shown already in Fig. 7.18. This was tested by Tanaka too, but
dismissed because of colour changes with drive.
Summarizing the section on concepts for colour displays, it appears to be
of primary importance to find out whether the efficiency and/or brightness
of a broad band emission can be made so high that filtering becomes
feasible. The subtractive process of filtering spoils the effective efficiency of
10 0 (a)
........
0~
50
c:
0
iii
e
f/)
f/)
c: 0
...
~ (b)
........
::J
~
...>
.iii
....c:
CIl
~
...J
W
500 600 700

Wavelength (nm)
Fig. 7.20 Green and red are filtered from the rather broad emission of ZnS:Mn in a
new concept of a multicolour display of Sharp [36].
for example mixed white by a factor of about 2.5; on the other hand, at least
11m W- 1 should be maintained to avoid excessive heating. There is no
question that the colour-by-filtering concept guarantees much higher pro-
duction yields and lower costs and is most preferable for larger sizes.
7.3.3 White (broad band)

The first acceptable white emitter was introduced by Tanaka et al. [32]. It
consisted of a conventional structure, with SrS as the active material. The
SrS was doubly doped by Eu 2+ for red and by Ce 3 + ,K + for green-blue.
They reported 36 cd m - 2 at 60 Hz of a rather good white emission. The
limitation on higher doping and thereby on higher brightness and efficiency
was according to them the rapid transfer of energy from excited Ce 3 + to
unexcited Eu 2+, as the emission and absorption lines overlap appreciably
(Fig. 7.22).
The last step in this direction is two recent reports [26, 30] both achieving
efficiencies of 0.4 1m W - 1. Tanaka et al. [26] obtained 65 cd m - 2 by post-
annealing in an Ar-S atmosphere at 600 C after EBE deposition, while
Mita et al. [30] achieved 85 cd m - 2 by EBE in an H2 atmosphere. The very
White SrS:Ce,K,Eu
U) Blue
.~
c:
~
J:i
.!
...>-
iii Green
...c:
c:
Q)
Red
400 500 600 700

Wavelength (nm)
Fig. 7.21 Demonstrated firstly by Tanaka et al. [32J, the three basic colours can be
filtered from the 'white' of SrS:Ce,K,Eu.
convincing argument of Mita et al. is that vacancies are avoided with this
active atmosphere.
It was mentioned above, but should be stressed again, that the transfer
process mentioned first by Tanaka et al. in the 1988 SID paper can be
avoided by a physical separation of the two dopants into two or more layers
of the SrS emitter film. A change of the colour coordinates of the white with
voltage is an argument against the application as a white display, not
against colour by filtering. In the latter case different nonlinear brightness-
voltage characteristics result for each colour, but this would be the case
anyway.
7.4 TECHNOLOGY AND MATERIALS

The most experience, of course, has been obtained with the material for the
yellow displays on the basis of ZnS:Mn, now in industrial production. The
two producers, however, use three different technologies for the deposition
of the active films: the oldest is EBE at Sharp, Japan; Planar, USA, uses
thermal evaporation, and the Planar section in Finland, Planar Interna-
tional, uses atomic layer epitaxy, a special CVD method [37]. Judging from
this situation it appears likely that at least no order of magnitude differences
in performance between these technologies are to be expected.
Technology and materials 251
(a) SrS:Ce,K,Eu (0.1,0.1, Cmol%)

Tsub=400C
with S Eu conc.
(mol%)
C=0.01
0.03
....>-
.iii
c:
....Q)c:
(b)
SrS:Eu(abs.) SrS:Ce,K SrS:Eu
,--!'''
: ",
~---~:-::.--~~
, ,
,, '
,,
,
!
,,
,
-'
400 500 600 700

Wavelength (nm)
Fig. 7.22 (a) The total emission is shifted strongly to the red with increasing dopant
concentration as a result of energy transfer from the Ce 3 + ion in SrS to the Eu 2 +,
because of (b) strong overlap in their emission and absorption bands [32].
In an effort to analyse this more carefully, the efficiency of TFEL samples

made by very different methods in a number of laboratories worldwide was
described in some detail in ref. 1. The analysis was carried out under
assessment of the Mn concentration [Mn], as this is the most pronounced
dependence. Only samples exhibiting fairly constant SIMS profiles of the
Mn content were investigated. In fact a very general behaviour of samples
from all makes as a function of [Mn] was obtained (Fig. 7.23), which
concerns the total efficiency. It shows a maximum around 2 x 10 20 em - 3.
The value of the maximum efficiency, however, scatters within almost an
order of magnitude. Looking in more detail at the partial efficiencies (ref. 1,
Fig. 5.12), it turns out that some samples with almost equal total efficiencies
differ in the partial efficiency positively or negatively. On the other hand, it
n,lmjW
c
'"
..- ~
x'
.
x
+
+ +
+ +
EBE1, HT +
+EBE1,LT
lIE THE1
0.1 ~
c THE2
~ ALE
x RFS
0.1 10
Mn-concentration, 1020 cm- J
Fig. 7.23 Dependence of the efficiency of ZnS:Mn TFEL on the Mn concentration
for different deposition techniques [16].
turns out that the best in the total efficiency is not the best in all partials,
which means that it could be improved [16].
However, even more important appears the conclusion that there is no
technology in principle inferior to others. Each deposition process, EBE,
thermal evaporation, d. sputtering and CVD, can be optimized and used to
make highly efficient TFELD. One of the most important parameters in the
process is the substrate temperature, influencing the crystallinity of the
active film. The surface roughness caused by the well-developed grain
surfaces 'penetrating' a geometrical mean surface plane, increases the out-
coupling by scattering, but obviously nlum is increased as well. It is not clear
whether this is caused by a segregation of heavy metals into the grain
boundaries, as tentatively considered by Benalloul et al. [15].
An overview of the various techniques of preparation was given by Ono
[38] at the Helsinki Workshop on EL.
Very recently Mikami et al. [36] proposed a modified CVD process for
the deposition of ZnS:Mn and ZnS:Tb by a so-called HCI bypass flow
process. While Mikami et al. claimed 4lm W - 1 for the yellow emission, in
an obvious application of this method to the Sharp multicolour display [35]
surprising brightnesses were achieved (at 60 Hz): in the filtered green,
80 cd m - 2; in the filtered red, 30 cd m - 2.
Some more standardization appears in the material for the insulator films.
Mostly d.-sputtered Si(O,N) is used; only the Planar division in Finland
uses ALE-grown (AI,Ti) oxide. As will be discussed in section 7.5, the main
influence of the dielectrics is in degradation. Also, the crucial point of the
TFEL structure is the dielectric-semiconductor interface. A constant doping
assumed, any asymmetry in the lightwave emitted under low frequency (e.g.
60 Hz) can be traced to the inadequacy of (at least) one interface. Emission of
the electrons at too Iowan electric field causes a fraction of the charge to be
transferred at energies at which no excitation of luminescent centres is
possible.
As well as efficiency and/or brightness, one further feature is important for
a display product, and this is price, or as a basis for the price, the production
costs. Decisive in this respect is throughput or manufacturing time, automa-
tion of the processes, and production yield. All of these and other factors will
be very dependent on the choice of technology and/or materials. Some of the
materials are much more easy to handle than others; some processes are
better to automate.
In the case where the necessary specifications can be met by virtually all
deposition processes, as for ZnS:Mn [16], costs will become decisive for any
judgement made on new production facilities in the future. The first period
of decision making on availability aspects of production equipment and
training aspects has now ended for ZnS:Mn. It will start again for colour
TFEL.
Materials and technology aspects for colour TFEL at present are rather
diverse. As outlined above, the materials favoured are the good hosts for
rare earths, which embody the loss-free acceleration of electrons at fields
which can be withstood by common dielectrics. These are the alkaline earth
sulphides and selenides.
Numerous techniques have been tried. The main problem is to obtain
stoichiometric, oxygen-free material from SrS and related compounds,
which are difficult to handle. This is all the more true for attempts to
prepare thin films on low temperature substrates. Good crystallinity and
high luminescence efficiency are still to be reported.
Generally speaking, the deposition technique used in most cases for the
alkaline earth sulphides or selenides (AESs) is EBE with coevaporation of
sulphur and deposition at the highest possible temperatures withstood by
the substrate glass, usually 500 0c. It has been developed to some degree of
sophistication at the Tottori University in Japan by Tanaka [39] and
Kobayashi and the latest results [40] have been mentioned before. Further
Japanese groups successful in the field are those of Abe et al. [41] and
Okamoto et al. [27].
CVD and AESs has some tradition. The RCA group published the first
results as early as 1987 [42, 43]. One of the outstanding results of these
efforts was the statement of appreciably different mechanisms of electro-
luminescence in ZnS:Mn and AES.
Following the way prescribed by Suntola [37] for ZnS:Mn, atomic layer
deposition (ALE) of AES was developed in Finland, mainly by Niinisto [44]
and his group at the Helsinki University of Technology. One of the main
concerns in this, as in other CVD processes, is the oxygen content of the
starting and/or doping materials. It appears detrimental to incorporate even
traces of oxygen. On the other hand, most of the usual starting organo
254 Thin .film electroluminescence devices
compounds contain oxygen, if they have enough vapour pressure to be

easily fed into the reactors.
Multisource evaporation methods, from the elements mainly, have been
tried by several groups [45-47]. They certainly have several advantages, as
the evaporation rates of the main components can be separately adjusted,
but the general difficulties with large molecules of Sx and Sex remain.
Reheating of the vapours could improve the situation [46].
Multisource evaporation is certainly a very versatile laboratory deposi-
tion method which allows for testing materials and varying rather freely
dopant concentrations.
Surprisingly limited attention has been paid by the TFEL community to
very successful attempts of Ohnishi and Okuda [25] to deposit AES films by
r.f. sputtering. The highest efficiency values reported were obtained on their
samples [24].
A special sputter regime has been used by them, admixing He to the
normal Ar sputter gas. A different velocity distribution of the impinging ions
could improve the sputter results and change the kinetic energies imparted
to the deposited species [25].
A rather impressive efficiency value of 1.3 1m W -1 in the blue-green
emission band of SrS:Ce has been measured on their samples, which were
codoped with CuBr2' The cod oping possibly was the reason for a rather
rapid degradation but, with feasibility having been shown, more stable
co do pings should evolve from an intensive search [48].
As well as these positive results concerning the most important specifica-
tion of TFEL devices, Ohnishi and Okuda's samples exhibited very ZnS-like
behaviour (Fig. 7.24) [48, 51]. As mentioned above (as well as in ref. 1),
different mechanisms were suspected to prevail in AES-based TFEL struc-
tures, compared with the classic one of ZnS:Mn. A very important feature in
this respect is the trailing edge currents and emissons in AESs.
Early papers by Crandall [42, 43], Tanaka et al. [26] and Ando and Ono
[52, 53] were most involved in the discussion of special features, such as
slow build-up of brightness, field quenching, etc. To our knowledge, it has
never been considered whether the described phenomena are intrinsic
properties of the new materials or features closely connected with nonperfect
AES films. The fact that they are missing or much less pronounced in
r.f.-sputtered samples could be an indication of real structure effects.
Certainly one has to expect a much stronger tendency of the AESs towards
nonstoichiometry and native defects, apart from higher reactivity with
oxygen. Only in a very recent paper of Ando and Ono [53] have attempts
been made to search for properties changing with the crystal perfection of
the active film and the insulating film composition.
The high probability of nonstoichiometry and high concentrations of
native defects have been indicated in recent papers on annealing effects [26],
and on Sr2 + vacancies, causing degradation of SrS:Ce samples [54]. This
Current density, mA/cm 2
60
'j
ZnS-:i',
. :{' ~ .
" +
I I
I ,
, I
80 I~' :
[,
r
J. :
o
o 1 2
(a) Field, MV/cm
mV~--~~~~~~~~--~~
o 100
(b) Time, ~s
Fig. 7.24 (a) Current density versus average field in the active film, exhibiting
striking similarities of ZnS, SrS, and CaS, even quantitatively [49]. (b) In samples of
CaS prepared by Ohnishi et al. [50] by r.f. sputtering, trailing edge emission is
marginal, !,!specially at higher transferred charges [51].
last paper introduces a new codoping process by phosphorus, which leads to

appreciable improvements of the aging behaviour of SrS:Ce samples. The
idea expressed is simple, but rather convincing: normal preparation is done
under sulphur excess, inducing Sr2 + vacancies, which in turn give rise to
ionic conductivity, destroying the lattice. Ce 3 + doping might increase the
vacancy concentration, as it can be charge compensated by formation of
another vacancy for every two Ce 3 + ions. If the sulphur surplus is high
enough to induce polysulphur - S~ - - or to oxidize Ce3+ to Ce 4 +, the
vacancy concentration increases drastically. The proposed and tested p3-
codoping improves stability and overall behaviour substantially. Unfortu-
nately the paper contains no information about the special AES features, or
mechanism considerations.
Important properties of the AES are not known at present. The lumines-
cence efficiency n)um, which is considered in some detail in ref. 1, is unknown
even on an order of magi tude scale for most of the reports, or at least not
given. The concentration of Eu and Ce at which concentration quenching sets
in is not even known for the phosphor powders with some accuracy or for the
films investigated in electroluminescence. There is only one determination of
the radiative lifetimes of the excited states ofCe and Eu in CaS [55]. They are
in the nanosecond range, and therefore more difficult to measure, but of
course accessible. Even for the 4f-+4f transitions of the Ln 3+ dopants in SrS
and CaS these data are missing. Concerning native defects, Sr2 + vacancies as
well as S2 - ones appear possible. A rather broad range of nonstoichiometry
seems to exist, much broader than in ZnS. The lack of single crystals of AES
explains the missing data about electronic transport, on shallow donors and
acceptors, and consequently on deep impurities, which have been investigated
in semiconductors mainly by capacitance spectroscopy of space charge layers.
Some preliminary data can be found in Rennie's papers [56, 57].
Summing up on materials, there is rather good knowledge available on
the materials properties of ZnS:Mn as this substance has been researched in
the form of powders and single crystals intensively over the last four
decades. Much less is known about the alkaline earth sulphides and
selenides, as there are practically no single crystals available, and low
stability and stoichiometry have been thought for years to inhibit research
on powders. Materials research is urgently needed in this class of substances.
Concerning technologies, all film deposition techniques can be success-
fully applied to ZnS:Mn; cost considerations are of primary importance
here. For AES TFEL work, d. sputtering has up to now resulted in the best
efficiencies and in most classic behaviour. It can of course at present not be
excluded that a situation similar to ZnS:Mn may evolve from intensive work
in all directions mentioned above.
The question about other materials, in addition to the II-VI compounds,
can only be speculated about. It is not likely that the known materials are
the only ones showing loss-free acceleration at fields low enough for use in
TFEL. A literature search for relevant parameters is not possible as long as
the theory of this process and knowledge of which fundamental properties
are relevant are not established. Experimental searches have the presence of
good enough thin films as a prerequisite: it might be costly to prepare them.
Luminiscence centres other than those known in phosphors are not likely
to exist for electroluminescence; one can rely on the knowledge gathered
from powder work.
7.5 DEGRADATION
For each practical device, degradation is an important aspect. The lifetime
of CRTs has been raised over the last decade to values well above 5000h,
Degradation 257
and certainly all displays will be measured by comparing them with this
value. Since, to our knowledge, no reports on degradation by the manufac-
turers or independent organizations have been published, we will concen-
trate on degradation mechanisms studied on test samples, in general without
any complex matrix organization of a multitude of pixels.
Again the ZnS:Mn system is the best known, and all other colour
materials are much less investigated.
7.5.1 ZnS:Mn
The most elaborate investigation [58] started from the concept of develop-
ing methods which allowed, on the basis of aging or forced aging of an
ensemble of 'identical' samples, one to follow the changes of some important
parameters. These chosen parameters were
1. the voltage division ratio below threshold Cd( C i + Cs),
2. the threshold voltage, and
3. the low light level value of the outer efficiency no.
The total pixel capacitance C was measured and used together with the
voltage division ratio to determine C i and Cs.
'Identical' samples were those cut from the same large substrate or several
substrates manufactured in the same run, i.e. made under identical condi-
tions. Most of the samples were made in a sequence of EBE depositions
without breaking the vacuum, and consisted of one or two layers of Al2 0 3
and Zr02, stabilized by Ca or Y, for each of the insulator films. The whole
lSI stack was deposited on ITO and contacted from the top by AI. ITO and
Al were wet etched after photolithography. Measurements of the parameters
mentioned previously were made (without destruction of the samples)
intermittently while driving the samples at constant transferred charge Q per
pulse at different frequencies between 50 Hz and 5 kHz.
Up to about 2 kHz, depending slightly on Q, the data scaled nicely with
the sum of the transferred charge, a first scaling law for forced aging. About
this frequency or at very high Q, the temperature in the samples obviously
increased, leading to additional effects, which were not followed.
One first surprising result was the changing of Ci mainly in the down-
wards direction, as illustrated by Fig. 7.25 [58]. This meant that either the
permittivity of the dielectric decreased or its thickness increased. It was
found that this behaviour was tied to non stoichiometric oxides containing a
surplus of oxygen. In extreme cases a segregation of oxygen into micro-
bubbles of gas could be observed during operation, as well as the oxidation
of very thin Al films up to the stage of visual vanishing. The last process can
really account for the growing thickness, and surely the dielectric constant
should decrease when ionic conductivity is destroyed. Careful preparation
of stoichiometric insulator films prevented this degradation mechanism.
CIA pF mm- 2
';'0

-50
I

~.
C;/A pF mm- 2
-120
~100

I
~.6 a
~.
r-0.5 1
~.2 E" 106 Vcm '

r-1.8
I

0 200 400 Time, h 600
Fig. 7.25 Changes of the total capacitance C/A, the insulator capacitance Cd A, the
voltage division ratio ot = C;/( C j + Cs), and the threshold field E, with the time under
forced aging at 5kHz, 10- 6 Ascm- 2 [58,59].
Constant C and C j were regarded as prerequisites to look for further

mechanisms. If this was achieved, in almost all cases constant Cs could be
assessed. This being checked, one could conclude that all changes in the
threshold voltage were due to changes in the threshold field for tunnel
emission of electrons from the interface states. (By more sophisticated
measurements [48J the two interfaces could be differentiated).
The second degradation mechanism identified in this way, and really the
dominant one, is the changing interface state spectrum. It depends on the
chemical nature of the interface and the processing conditions. It can be kept
small, but ultimately it governs the life of the device. The third parameter,
the efficiency, very seldom changed. In the very few cases in which changes
were observed, they could be traced back to changes in the threshold field
Et , crossing the margin for loss-free acceleration [48]. So one very import-
ant conclusion is the stability of the luminescence conditions in this ZnS:Mn
system.
Conclusions 259
7.5.2 ZnS:Tb
Some pecularities exist for Ln3+ -doped ZnS. It was said above that there is
a risk of driving an overdoped host towards a two-phase mixture of ZnS:Tb,
doped to the solubility limit, and phases which contain more of the dopant,
and possibly a pure dopant phase. No detailed investigations of situations
like this have been reported. However, in doing similar studies as the ones
described for ZnS:Mn, an additional degradation of the efficiency was
detected. As the decay of the photoluminescence had not changed, it was
concluded that the excitation probability must have altered. A check of the
electrical characteristics revealed no changes there, so that the conclusion
was a change in the product of dopant concentration and its impact
cross-section [17]. This can easily be reconciled with the idea of segregation
of new phases etc.
7.5.3 Alkaline earth chalcogenides

Of course the aging of the dielectrics may occur in every combination with
active film materials, and it does for AES samples. Interface aging is even
more pronounced in these materials if oxides are used as dielectric films. The
high reactivity of the AESs with oxygen is enhanced by the presence of free
electrons under operating conditions so that rapid changes occur. The most
stable interface reportedly is the one with Si 3 N 4 , which did withstand
extreme loads [60]. There is, however, no report about the stability of the
efficiency in these materials. On the contrary, there are some indications of
real aging of the conversion yield of electrical input into the active film to
photon emission [48]. If this could be confirmed by further experiments, the
most likely reason is a change in the concentration of excitable centres, just
as in the case of ZnS:Tb. One possibility, at least in the case of Ce 3 + doping,
could be the conversion into Ce 4 +.
Summarizing on degradation, stability over about 50000 h of normal

display operation appears possible, judging from accelerated aging tests of
ZnS:Mn TFELD. There is not enough work done or reported on aging of
AES-based device samples to make any predictions about them. It is very
likely that additional mechanisms exist in the less stable materials, attacking
directly the luminescence centres.
7.6 CONCLUSIONS
Thin film electroluminescent devices are well established with respect to
technology, materials, and understanding of the physics (at least in a
linearized model) as far as ZnS:Mn is concerned. Some $10 8 market share in
the tough competition of the LCD-dominated flat panel market of 1991
proves that. Production yields of much better than 75% have been reported
[61]. Multicolour displays based on ZnS:Mn are likely to go into large-scale
production in 1992, and the first steps toward full-colour displays have been
taken in several laboratories. As well as the excellent appearance with
respect to viewing angle, contrast and brightness, the simplicity of produc-
tion in flow-line equipment without the need of costly handling processes
provides a challenge to the older flat panel technologies.
A bright future can be forecast for bright and soon full-colour TFEL
devices.
ACKNOWLEDGEMENT
Gratefully acknowledged are numerous discussions, and the coordination of
the contents of the chapters in this volume with G.O. Miiller.
REFERENCES
1. Miiller, G.O. (1993) In Solid State Luminescence (ed. A.H. Kitai), Chapman &
Hall, London, Chapter 5.
2. Miller, M.R. and Kelley, T.G. (1989) In Electroluminescence, Springer Proceed-
ings in Physics, Vol. 38, Springer, Heidelberg, pp. 259-263.
3. Miiller, G.O., Mach, R., Reinsperger, G.-v. and Schrottke, L. (1991) SID 91
Tech. Digest, p. 567.
4. Uchiike, H. (1989) Electroluminescence, Springer Proceedings in Physics, Vol. 38,
Springer, Heidelberg, pp. 238-245.
5. Miiller, G.O. and Mach, R (1993) Phys. Stat. Sol. Submitted.
6. Kun, Z.K., Leksell, D. and Brandt, G.B. (1992) Proc. 5th Int. Conf on II-VI
Compounds. Tamano. 1991, J. Cryst. Growth, 117, p. 987.
7. LekseIl, D., Kun, Z.K., Asars, J.A., Phillips, N.J., Brandt, G.B., Stringer, J.T. and
Matty, T.e. (1990) 1st Laser Printer Conf.. Colorado Springs. 1990.
8. Miiller, G.O., Mach, R, Halden, E. and Fitting, H.-J. (1990) Proc. XX ICPS,
Thessaloniki. 1990, World Scientific, Singapore, p. 2510.
9. Miiller, G.O., Mach, R., Reinsperger, G.-U. and Schulz, G. (1991) Proc. Int.
Display Research Conf. 1991. San Diego. CA, p. 17.
10. Miiller, G.O., Mach, R, Reinsperger, G.-v., Halden, E. and Schulz, G. (1992)
Proc. 5th Int. II-VI Conf. Tamano.1991, J. Cryst. Growth, 117, p. 948.
11. Kitai, A., Dalacu, N. and Zhizheng, H. (1991) SID 91 Technical Digest, p. 440.
12. Okamoto, S., Nakazawa, E. and Suzuki, T. (1991) Jpn. J. Appl. Phys., 30 (7B),
L1321-L1323.
13. Busse, W., Gumlich, H.-E., Geoffroy, A. and Parrot, R. (1979) Phys. Status Solidi
B, 93, 591-596.
14. Hurd, J.M. and King, e.N. (1979) J. Electron. Mater., 8, 879.
15. Benalloul, P., Benoit, J., Geoffroy, A., Mach, R, Miiller, G.O. and Reinsperger,
G.-v. (1990) J. Cryst. Growth, 101, 989.
16. Mach, R, Miiller, G.O., Reinsperger, G.-v., Benalloul, P., Benoit, J. and
Ohnishi, H. (1990) SID 90 Digest, pp. 238-241.
17. Khomshenko, V.S., Konenec, YF., Vlasenko, N.A., Mach, R, Reinsperger,
G.-v., Selle, B. and Reetz, R (1990) J. Cryst. Growth, 101, 994-998.
References 261
18. Ohnishi, H., Yamasaki, Y. and Iwase, R. (1987) Proc. Soc. In! Disp. 28(4),
345-350.
19. Hiirkonen, G., Hiirkonen, K. and Tornqvist, R. (1990) SID 90 Digest, pp.
232-235.
20. Tanaka, S., Morimoto, S., Samada, K., Kobayashi, H., Zhang, Z. and Jiang, X.
(1992) In Proc. 5th Int. II-VI Con!. Tamano. 1991, J. Cryst. Growth, 117,
997.
21. Heimbrodt, W. and Goede, O. (1986) Phys. Status Solidi B, 135, 795.
22. Mach, R., Reinsperger, G.-D., Miiller, G.O., Selle, B. and Matzkeit, G. (1992),
Proc. 5th Int. II-VI Con!. Tamano. 1991, J. Cryst. Growth, 117, p. 1002.
23. Oseto, S., Kageyama, Y., Takahashi, M., Deguchi, H., Kameyama, K. and
Fujimura, I. (1989) In Electroluminescence, Springer Proceedings in Physics, Vol.
38, Springer, Heidelberg, pp. 191-194.
24. Miiller, G.O. Mach, R. and Ohnishi, H. (1990) In Eurodisplay 90. Proc. 10th Int.
Display Res. Conf.. Amsterdam, pp. 88-91.
25. Ohnishi, H. and Okuda, T. (1989) SID 89 Digest, pp. 317-320.
26. Tanaka, S., Morita, H., Yamada, K. and Kobayashi, H. (1991) IDRC 91, pp.
137-140.
27. Okamoto, S., Nakazawa, E. and Tsuchiya, Y. (1989) In Electroluminescence,
Springer Proceedings in Physics, Vol. 38, Springer, Heidelberg, pp. 195-202.
28. Barrow, W.A., Coovert, R.E., Krug, CN. and Zruchkovski, MJ. (1988) SID 88
Digest, pp. 284-286.
29. Tuenge, R.T. (1989) Electroluminescence, Springer Proceedings in Physics, Vol.
38, Springer, Heidelberg, pp. 132-138.
30. Mita, J., Gao, Q.Z., Kobayashi, M. and Kawamura, K. (1991) SID 91 Digest, pp.
290-293.
31. Konishi, T., Fujita, Y., Minamoto, M., Yamamoto, S. and Tanaka, Y. (1991) SID
91 Digest, pp. 271-274.
32. Tanaka, S., Yoshiyama, H., Nishiuva, 1. and Ohshio, S. (1988) SID 88 Digest, pp.
293-206.
33. Tanaka, S., Kawakami, H., Nakamura, K. and Kobayashi, H. (1989) SID 89
Digest, pp. 321-324.
34. Laakso, C, Khormaei, R. and King, C. (1991) IDRC 91. San Diego 1991, pp.
43-44.
35. Okibayashi, K., Ogura, T., Terada, K .. Taniguchi, K., Samashita, T., Yosheda,
M. and Nakajima, S. (1991) SID 91 Digest, pp. 275-278.
36. Mikami, A., Terada, K., Tanaka, K., Taniguchi, K., Yoshida, M. and Nakajima,
S. (1989) SID 89 Digest, pp. 309-312.
37. Suntola, T. (1990) Proc. 1st Int. Symp. on Atomic layer Epitaxy, Helsinki, in Acta
Poly tech. Scand., Chem. Technol. Metall. Ser., 195, 93-106.
38. Ono, Y.A. (1990) Acta Poly tech. Scand. Appl. Phys. Ser., 170,41-48.
39. Tanaka, S. (1990) Proc. 4th Int. Conf. on II-VI Compounds 1989, in J. Cryst.
Growth, 101, 958.
40. Tanaka, S., Yoshiyama, H., Nakamura, K., Wada, S. Morita, H. and Kobayashi,
H. (1991) Jpn. J. Appl. Phys., 3O(6A), Ll021-Ll024.
41. Abe, Y., Onisawa, K., Tamura, K., Nakayama, T., Honazono, M. and Ono, Y.A.
(1989) Springer Proceedings in Physics, 38, p. 199. Proc. 4th Int. Workshop on
Electroluminescence, Tottori, Japan, 1988. Springer, Berlin.
42. Crandall, R.s. (1987) Appl. Phys. Lett., SO, 641.
43. Crandall, R.S. (1987) Appl. Phys. Lett., 50(10), 551-553.
44. Leppanen, M., Leskelii, M., Niinisto, L., Nykiinen, E., Soininen, P. and Tiitta, M.
(1991) SID 91 Digest, pp. 282-285.
45. Nire, T., Watanabe, T., Tsurumaki, T., Miyakoshi, A. and Tanda, S. (1989),
Springer Proceedings in Physics; 38, p. 218. Proc. 4th Int. Workshop on
Electroluminescence, Tottori, Japan, 1988. Springer, Berlin pp. 218-223.
46. Nakanishi, Y., Suzuki, T., Tatsuoka, H., Kuwabara, H., Fujiyasu, H., Fukuda, Y
and Shimaoka, G. (1990) In Proc. 4th II-VI Conj., Berlin, 1989.
47. Mauch, R.H., Velthaus, K.O., Bilger, G. and Schwek, H.W. (1992) Proc. 5th Int.
II-VI Conj.. Tamano. 1991, l. Cryst. Growth, 117, p. 964.
48. Mach, R. (1990) Acta Poly tech. Scand. Appl. Phys. Ser., 170, 85-98.
49. Muller, G.O., Mach, R., Selle, B. and Ohnishi, H. (1990) l. Cryst. Growth, 101,
999.
50. Ohnishi, H., Iwase, R. and Yamasaki, Y. (1988) SID 88 Digest, pp. 289-292.
51. Mach, R., Ohnishi, H. and Muller, G.O. (1989) In Electroluminescence, Springer
Proceedings in Physics, Vol. 38, Springer, Heidelberg, pp. 176-179.
52. Ando, M, and Ono, Y.A. (1990) l. Appl. Phys., 68(7), 3578-3583.
53. Ando, M. and Ono, YA. (1991) l. Appl. Phys., 69(10), 7225-7232.
54. Takahashi, K., Shibuya, K. and Kondo, K. (1991) Proc. 5th Int. II-VI Con!,
Tamano, 1991, l. Cryst. Growth, 117,979.
55. Hoshina, T. (1980) l. Phys. Soc. lpn., 48(4), 1261-1268.
56. Rennie, 1., Nakazawa, E. and Koda, T. (1990) Jpn. J. Appl. Phys., 29 509.
57. Rennie, 1., Takenchi, K., Kaneko, Y and Koda, T. (1991) J. Lumin., 48--49,
787-791.
58. Mach, R. and Muller, G.O. (1989) In Electroluminescence, Springer Proceedings
in Physics, Vol. 38, Springer, Heidelberg, pp. 264-272.
59. Muller, G.O., Mach, R., Reetz, R. and Reinsperger, G.-v. (1988) SID 88 Digest,
pp.23-26.
60. Tanaka, K., Mikami, A., Ogura, T., Taniguchi, K., Yoshida, M. and Nakajima, S.
(1986) Appl. Phys. Lett., 48(25),1730-1732.
61. King (1990)
62. Onisawa, K.O., Abe, Y., Nakayama, T. and Hanazono, M. (1991) Jpn. l. Appl.
Phys., 30, 314-319.
8
Light emitting diodes:
materials growth and properties
S.P. DenBaars
8.1 INTRODUCTION
The major subject of this book is 'high field' electroluminescence: however,
an equally important class of luminescence is 'low field' injection electro-
luminescence found in light emitting diodes (LEDs). In contrast to the
localized luminescence centres and impact excitation of free carriers in 'high
field' luminescence, excitation in 'low field' injection luminescence occurs by
the injection of minority carriers in semiconductor materials. Current
research interests focus on the development of bright LED sources for full
colour displays. Recently, significant progress in the epitaxial growth of
compound semiconductor materials has been made. This has created new
applications for high brightness LEDs in the outdoor display market. The
purpose of this chapter is to review the physics of LEDs and materials
selection, and to describe the epitaxial growth technologies employed in the
manufacturing of LEDs.
8.2 PHYSICS OF LEDs: INJECTION LUMINESCENCE

Light emitting diodes are a class of diodes that emit spontaneous radiation
under suitable forward bias conditions. LEDs are fabricated from semicon-
ductor materials in which a p-n junction is fabricated. Injection elec-
troluminescence is the most important mechanism for exciting the semicon-
ductor materials. Under forward bias conditions, electrons are injected into
the p-type semiconductor and holes are injected into the n-type material.
Recombination of these minority carriers with the majority carriers at the
p-n junction results in near bandgap radiation. A schematic representation
of a radiative recombination in a forward bias p-n junction is shown in Fig.
8.1. Materials emitting radiation covering the ultraviolet through infrared
263
264 Light emitting diodes
n-TYPE p-TYPE
Acceptor
Level
Junction
Region
Fig. 8.1 Schematic representation of a p-n junction in a light emitting diode under
forward bias.
spectrum exist. The solid state nature of the device makes it a high reliability
product choice. Average lifetime of an LED is of the order of tens of years
[1]. Some of the commercial applications of LEDs include visible displays,
plastic fibreoptic communications, opto"couplers, motion control, bar-code
scanners, LED printers and even automotive brake lights.
8.2.1 Direct and indirect bandgap material

Semiconductor LED materials can be divided into two different classes of
band structures, exhibiting either direct or indirect bandgap behaviour. In a
direct bandgap semiconductor, electrons in conduction band minima and
holes in the valence and minima have the same momentum as shown in Fig.
8.2. The probability of a band-to-band radiative transition for the direct
bandgap semiconductor is high since momentum will be conserved during
electron-hole recombination. In contrast, in an indirect bandgap semicon-
ductor the electrons in the conduction band minima have different momen-
tum from the holes in the valence band minima, and therefore momentum is
not conserved. In an indirect semiconductor an electron would have to
change its momentum before recombining with a hole. A third component, a
phonon, is required to participate in the recombination even for the electron
to change its momentum and a transition to occur in the indirect gap
semiconductor. This results in a significantly lower recombination probabil-
ity for a band-to-band transition to occur in the indirect gap semiconductor.
However, the addition of impurities that cause iso-electronic traps can be
used to increase the radiative efficiency of indirect gap semiconductors.
Iso-electronic trapping centres allow the momentum of the electron to be
conserved by tightly binding the electron in real space. This results in a
spreading out of the electron momentum in momentum space according to
the quantum mechanical prediction of the Heisenberg uncertainty principle.
As a result, the momentum conservation rules are satisfied and efficient
radiative recombination can occur.
Physics of LEDs: injection luminescence 265
>.
~
Q)
c:
W
c:
e
~
jjj
Momentum
Direct
Fig. 8.2 Energy band diagram of direct and indirect bandgap semiconductors.
Electron energy is plotted versus momentum, k. In the direct semiconductor
radiative transitions are allowed because of conservation of momentum in which
only vertical transitions are allowed. In an indirect semiconductor the electrons in
the conduction band minimum possess a different momentum from the holes and
diagonal transitions are forbidden [1].
Nitrogen and zinc-oxygen pairs are the two most important types of
iso-electronic impurities in indirect bandgap GaP. The energy levels of the
iso-electronic traps lie below the conduction band and therefore the emitted
photon will have an energy approximately equal to the bandgap energy
minus the binding energy of the centre.
8.2.2 Radiative recombination

Electron-hole recombination can be either radiative or non-radiative.
Light output from an LED can be increased by maximizing the probability
for radiative recombination and decreasing the probability for non-radi-
ative recombination paths. Several mechanisms cause non-radiative recom-
bination in semiconductors; among these are Auger recombination, recom-
bination at defect sites, and multiphonon emission at deep impurity sites
[2]. Radiative transitions in semiconductors occur primarily by interband
transitions, excitonic recombination, or recombination through impurity
centres.
Under continuous injection of carriers an excess carrier density will exist
in the semiconductor. For charge neutrality to exist the excess electrons L\n
and excess holes L\p must be equal after injection ceases, and the excess
carrier density has been found to decrease exponentially with time with a
characteristic lifetime.
L\n = L\no exp( -t/T r )
where .1no is the initial excess carrier concentration, and !r is the recombina-
tion lifetime. The radiative recombination rate is defined with the excess
carrier density and recombination lifetime as [3]
.1n
Rr = -
!r
Likewise the non-radiative recombination rate is
The internal radiative quantum efficiency is defined as the ratio of the rate of
radiative recombination '1i = Rr/(Rr+ Rnr) to the total recombination rate.
Therefore, the internal quantum efficiency can be expressed in terms of the
radiative and non-radiative lifetimes:
1/!r
'1. = ---'---
I 1/!r + 1/!nr 1 + !r/!nr
By keeping !r/!nr small a maximum internal quantum efficiency can be
obtained. This is done by making !nr long compared to !r. Growing
materials with low defect densities and eliminating impurities that cause
non-radiative deep levels are certain material improvements that increase
the quantum efficiency. Using the principle of detailed balance [4] in which
at equilibrium the rate of optical generation of electron-hole pairs is equal
to their rate of radiative recombination, the spontaneous recombination rate
for band-to-band emission can be expressed as
Rsp=Brnp
where Br (cm 3 S-1) is the radiative recombination coefficient. Under non-
equilibrium conditions additional carriers are introduced into the material
and the spontaneous recombination rate becomes
Rsp = Br(no + .1n)(po + .1p)
where no and Po are the equilibrium electron and hole concentrations
respectively. Charge neutrality requires
.1p=.1n
and the spontaneous recombination rate becomes
Rsp=Br[nopo +.1n (Po + no) + .1n 2 ]
Therefore the spontaneous recombination rate is a combination of the
equilibrium recombination rate R~p and the recombination rate caused by
the excess injected carriers R~p:
Rsp = R~p + R~p
The equilibrium recombination rate can be expressed 10 terms of the
intrinsic material doping nj:
R~p = Br(nj)2 = BrnoPo
This yields the rate due to excess injected carriers:

R~p=Br An (Po +no+An)
By definition the excess carrier pair density, An, divided by the average
radiative carrier lifetime, Tn will yield the excess carrier recombination
rate:
At high injection levels, the excess carrier density will substantially exceed
the background concentration:
An>po+no
so that the average carrier lifetime can be expressed as
1
Tr~ Br An
which occurs in the 'bimolecular' recombination region. In the case of

bimolecular recombination, the lifetime value is continually changing so Tr
represents the average carrier lifetime. The other region for the lifetime
occurs at low injection levels in which the excess carriers decay exponen-
tially with time. In this region the lifetime can be determined with the
background carrier concentration:
1
For n-type material, no~po:
Likewise, for p-type material,

1
't r ;;;: - -
Brpo
For intrinsic material at low injection levels the longest possible lifetime is
found
Interface recombination is another factor that affects the effective lifetime

and is taken into account by dividing the average non-radiative interface
recombination velocity S by the active layer thickness d. The effective
lifetime is then a combination of all the competing process lifetimes:
1 1 2S
- = - + -d +Br(no+PO)
reft rnr
Since interfaces occur on each side of the active region the interface
recombination velocity is multiplied by 2.
8.2.3 External quantum efficiency

The best measure of LED efficiency is how much light is generated for the
injected current. Fundamentally, the external quantum efficiency can be
expressed as the ratio of photons out per injected electron:
'1ext = (photons out)/(electrons in)
The external quantum efficiency is the product of material and optical
coupling efficiency:
where '1inj is the injection efficiency, '11 is the light generation efficiency, and
'10 is the optical or light extraction efficiency. Since light generation is
predominantly generated on the p side of the diode for GaP [5] and AIGaAs
[6] a high electron injection efficiency is typically desired. This dictates that a
high injection efficiency will be achieved if the electron current dominates
over the hole current. Injection efficiency is defined as the ratio of the
electron current J e to the total forward current J f and can be expressed as
Je Je
'1inj= J +J = J
e h f
From p-n junction theory [4] we can define the injection efficiency in terms
of relevant materials properties:
DnnLp
'1' . = -------'---
In) DnnLp + DppLn
where Dn and Dp are the diffusion coefficients for electrons and holes,
respectively. Lp and Ln are the minority carrier diffusion lengths, and nand p
are the net electron and hole concentrations on the two sides of the junction.
At thermal equilibrium the net electron and hole concentration can be
expressed with the doping level of donors N D and acceptors N A:
n=nf/N A
and
p=nflND
Therefore to obtain a high injection efficiency we want to adjust the doping
such that N D > N A. However, to optimize the optical efficiency the material
must be kept transparent to light generated at the junction. Therefore free
carrier absorption must be minimized by lowering the doping level in the
heavily doped n-type material to below 10 18 cm -3 [6].
8.2.4 Heterojunction versus homojunction

Both homojunction and heterojunction semiconductor structures can be
employed for visible LEDs. In homojunction LEDs, the p and n materials
are composed of the same energy gap semiconductor, while in heterojunc-
tion LEDs two materials of different bandgaps are utilized. A typical GaAsP
homojunction LED with a diffused p-n junction is shown in Fig. 8.3. Since
the GaAsP is grown on an absorbing GaAs substrate only light emitted
upwards from the junction escapes. By combining wide bandgap semicon-
ductors with narrow bandgap semiconductors desired properties unattain-
able in homostructures become possible. This so-called 'bandgap engineer-
ing' [8] allows the LED designer to achieve higher luminous efficiencies
than with conventional homostructures. The two major benefits of hetero-
structure LEDs are the increased minority carrier injection efficiency and
the ability to use wide-gap material. The wide-bandgap layers are transpar-
ent to the light generated in the narrow-gap active layer and therefore do
not contribute to photon reabsorption. The radiative efficiency in hetero-
structure LEDs is higher than in homo structure LEDs because of the
increased current injection efficiency at the heterojunction. In particular,
since the radiative efficiency of the p side is usually much higher than in the
AI Contact
~ffUSed p-Type Region
./
,~------~~ ""'II'- n-Type GaAS. 60 P.40

Epitaxial Layer
n-Type GaAs
....- - Substrate
Au Contact
Fig. 8.3 Construction of a typical homojunction GaAsP LED with a diffused
junction. The p-n junction is formed by a Zn diffusion into the n-type GaAsP
epitaxiallayer. (After Steranka et al. [7].)
n side, higher efficiency devices are made by eliminating the minority carrier
injection into the n-type material. By creating a heterostructure at the p-n
interface a valence band discontinuity is formed which provides a hole
potential barrier. By adding an additional heterostructure, confinement at the
other side of the active layer increases because the second heterointerface
prevents the injected electrons from diffusing out of the narrow-gap layer.
Figure 8.4 illustrates the energy band diagram for a p-N heterojunction under
no bias and forward bias. (Capital 'N' signifies the larger energy gap material.)
The conduction band discontinuity A.Ec and the valence band discontinuity
A.Ev produce a barrier to electron and hole flow respectively. The performance
advantages of the double-heterostructure absorbing substrate (DHAS) red
AIGaAs LEDs in comparison with single-heterostructure absorbing substrate
(SHAS) AIGaAs and GaAsP:N LEDs is shown in Fig. 8.5.
No Bias
Forward- 5 ~6Ec {Se2 - E

Biased hI 1
uui:~l
.......
rev
t......t. c
---Ev
p n
Fig. 8.4 Energy band diagram for a p-N heterojunction under no bias and forward
bias. The conduction band discontinuity flEe and the valence band discontinuity flEv
produce a barrier to electron and hole flow respectively. (After KresseI [3].)
For heterojunctions the bandgap discontinuity affords a significant im-

provement in the injection efficiency. For a p-N heterojunction the ratio of
electron to hole currents can be expressed as [9]
Je DnnpoLp
Jh DpPnoLn
where Lp and Ln are the minority carrier diffusion lengths in the p and N
material respectively. Dn and Dp are the electron and hole diffusion coeffi-
cients respectively.
The electron minority carrier concentration on the p side is npo and the
minority hole carrier concentration on the N side is PnO; at thermal
equilibrium these are given by
10.0
;;C
---E 8.0
>-
u
c::
-
Q)
'(3
6.0 DH-AS
;;::::
Q)
-
.g
Q)
E
4.0
~ _____=-SH
... - - - - GaAsP:N
-
0
0 2.0
..c::
a..
0.0
0.1 1.0 10.0 100.0
Forward current (rnA)
Fig. 8.5 Performance (1m A-I) of double-heterostructure absorbing substrate red
AlGaAs LEDs in comparison with that of single-heterostructure AlGaAs and
GaAsP:N LEDs. (After Steranka et al. [7].)
and
PnO =nr /no
The electron-to-hole injection ratio now becomes
Je DnLp nr no
Jh DpLn nr Po
From Fermi-Dirac statistics the electron-hole density product is a function
of the materials energy gap:
np=nr =NcNy exp( -Eg/kT)
where N c and Ny are the conduction and valence band effective density of
states respectively. Therefore the electron-to-hole injection ratio now be-
comes a function of the energy gap difference:
- - - -no-N~N~
J e-_ -DnLp expu (AE /k T )
J h DpLn Po N~N~ g
For most compound semiconductor heterostructures, the exponential term

in the above equation dominates and a high electron-to-hole injection
efficiency can be achieved as long as the barrier height is sufficient. This
nearly ideal single-sided injection of electrons into the p-type active layer
makes the double-heterostructure AIGaAs LEDs highly efficient. The energy
gap differences also minimize the effect of doping level on the injection
efficiency, which dominated the homojunction LED efficiency. Selection of

the wide gap materal must take lattice matching into consideration to avoid
forming dislocations near the interface that would act as non-radiative
recombination centers and therefore significantly reduce the radiative effi-
ciency.
8.3 MATERIALS SELECTION

For a material to be an efficient visible light emitter it must always meet
three important criteria, namely (1) possess an efficient radiative pathway,
(2) have an energy gap of appropriate width, (3) be controllably doped both
p and n type with low resistivities. The most important materials system that
has had great technological success in meeting these demands is the III-V
semiconductors and their alloys. An emerging area of materials research is
also occurring in the wide-bandgap II-VI materials system. A plot of the
energy gap of compound semiconductors is shown in Fig. 8.6. From this
plot we can see that the entire visible spectrum should be achievable with
compound semiconductors. In the late 1980s remarkable progress for high
brightness LEOs for displays occurred. This evolution in LED performance
for the commercially important compound semiconductors materials system
is illustrated in Fig. 8.7. Notable progress is the development of high
efficiency AIGaAs and AIGaInP alloys. This was caused primarily by the
higher efficiency afforded by heterojunction-based LEOs. Outdoor displays
and even brake lights become viable applications for LEOs as performance
4
ZnSo
3 r-
>-
~ AlP 0
ZnSe o
o CdS ZnTe
go GaP 0
AIAso 0
1>0
;... 2 I-
CdSe
foil 00 CdTe
~ GaAso A1Sb 0
~ 0
Si 0 InP
1 r-
Ge o oGaSb
InAs o
InSb 8,. Sn
I I I I I
0
2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9
Bond length (A)
Fig. 8.6 Bandgap energy versus bond length for various semiconductors. Note the
general increase in bandgap as the bond length decreases.
Materials selection 273
100
--
~
~
I
I Red, Yellow,
co
--
and
~
( J) DH AIGaAs/AIGaAs I Green AlinGaP
c 10
Q) ~-------
DH AIGaAs/GaAs
E Red Autobmobile Taillight
--------------------
:::J
.......
Q) SH AIGaAs/GaAs
u Red, Yellow,
c Green GaAsP:N
co
E
(;
't:
Q) Blue SiC
a..
I
0.1
1970 1975 1980 1985 1990 1995 2000
Time
Fig. 8.7 Evolution of LED performance (1m W- I ) for the commercially important
materials. (Courtesy Hewlett-Packard Optoelectronics Division.)
levels exceeded 31m W -1. Table 8.1 summarizes the commercially import-
ant visible LED materials currently achievable.
8.3.1 Aluminium gallium arsenide (AIGaAs)

AIGaAs is an excellent compound semiconductor for high efficiency visible
red LEDs with a direct bandgap at 660 nm. In addition, since AlAs and
GaAs have approximately the same lattice constant, AIGaAs epilayers with
varying aluminium fractions are lattice matched to GaAs substrates. These
properties permit the growth of complex double-heterojunction devices for
efficient carrier injection and confinement. Using liquid phase epitaxy (LPE)
heterostructure devices have been grown on absorber substrates (ASs) [10J
and transparent substrates (TSs) [IIJ with external quantum efficiencies of
8% and 20% respectively. The brightness level of these devices ranges from
1 to 20 candela (cd) depending on the package and viewing angle. These
brightness levels are sufficiently high enough for outdoor displays and even
automotive brake lights.
The transparent substrate employed in the highest efficiency devices is
actually a thick AIGaAs layer which allows for the selective etch removal of
the GaAs substrate (Fig. 8.8). The performance advantages of AIGaAs over
GaAsP materials are shown in Fig. 8.5. Cost is the deciding factor when
selecting one material over the other. The well-established VPE growth
technology has kept GaAsP LED prices substantially lower than AIGaAs.
LPE is the only growth technology that can provide material quality
efficient enough for high brightness applications. Other growth technologies
such as metal-organic chemical vapour deposition (MOCVD) produce
Table 8.1 Summary of commercially available visible LED materials
Material Colour range Peak Luminous External Growth Device

emission efficiency quantum technique type
(nm) (imW- 1 ) efficiency (%)
AIGalnP Red-green 585 12.0 2.2 MOCVD DH

AIGaAs Red 660 9.0 20.0 LPE DH
GaP Pure green 555 1.4 0.2 LPE H
GaP:N Yellow-green 565 1.8 0.3 LPE H
GaP:NN Yellow 590 1.0 0.2 LPE H
GaP:Zn-O Red 650 4.0 5.0 LPE H
GaAs O. 15 P 0.85:N Yellow 590 1.0 0.3 VPE H
GaAs O. 35 P 0.65 Orange-red 630 0.7 0.3 VPE H
GaAs O.4 P 0.60 Red 650 0.15 0.2 VPE H
SiC Blue 470 0.02 LPE, CVD H
ZnSe Blue 470 MBE, DH
MOCVD
GaN Blue 465 0.03 MOCVD DH
DH, double heterostructure; H, homojunction device.

Absorbing Substrate Transparent Substrate
Fig. 8.8 Cross-sections of two types of red AlGaAs double-heterostructure (DR)

LEDs that are commercially available. Because of the increase in optical extraction
efficiency, the transparent substrate (TS) device is approximately twice as bright as
the absorbing substrate (AS) device. (After Steranka et al. [7].)
AIGaAs that is only half as efficient as LPE-grown material [12]. The

reduced efficiency in MOCVD-grown AIGaAs is believed to be caused by
either unintentional impurities or Ga vacancy formation. Future improve-
ments in source purity and a fundamental understanding of the relevant
material properties are necessary if MOCVD is to replace LPE for the
production of AIGaAs based LEDs.
8.3.2 Aluminium gallium indium phosphide (AIGaInP)

One of the most significant advances in LED res each in recent years was the
development of the AIGaInP materials system for high brightness applica-
tions. AIGaInP with a direct energy gap ranging from 2.3 eV (540 nm) to
1.8 eV (660 nm) is ideally suited for high brightness applications. Extensive
research on AIGaInP for optical device applications led to the commercial
availability of visible laser diodes [13]. High efficiency surface emitting
yellow LEDs were first reported in 1990 and employ double heterostructure
device concepts [14].
High quality AIGaInP can be grown by MOCVD lattice matched to
GaAs substrates. Other epitaxial synthesis techniques have not demon-
strated the quality or capability to grow these materials. For instance, the
large segregation coefficient between Al and In makes it virtually impossible
to grow AIGaInP from solution [15]. Luminous efficiencies as high as
201m A -1 at 590 nm emission have been reported [14]. This is approximate-
ly an order of magnitude better than GaAsP technology.
The external quantum efficiency of AIGaInP for various wavelengths is
shown in Fig. 8.9. For wavelengths greater than 595 nm, the external
quantum efficiency is approximately 2%. This performance is almost an
order of magnitude better than that of existing GaAsO.1SPO.8S materials.
Two improved properties of AIGaInP diodes can be attributed to the
.. :.. Ext. Quantum Eff (0/0)

- 0 Luminous Efficiency (1m/A)
3 25
a...
I ......
" ~ ............
'"
;e r
!!..... 2.5 20
c:
3
==
w
E
::J
2 .
~---;',--..--./
15
s
0
c:
fI)
m
C
0
CII
::J
1.5
I ...... 10
::2:
C')"
iii'
::J
ca
E fl ...... (')
'<
I
~
x
CD
J ....... 5
w 0.5 W
o o
560 570 580 590 600 610 620 630
Wavelength (nm)
Fig. 8.9 External quantum efficiency and luminous efficacy of the newly developed
AlGaInP high brightness LEDs as a function of emission wavelength. (Data from
Kou et al. [14].)
benefits of heterojunction technology and direct bandgap nature. When

combining the eye response with the quantum efficiency, we obtain the
luminous efficiency dependence of AIGaInP on wavelength which is also
illustrated in Fig. 8.9. The performance of AIGalnP is equivalent to GaP:N
green and 10 times brighter than GaAsP and 5 times brighter than
GaAsP:N in the orange region. With further improvements in the properties
of AIGalnP, it is conceivable that MOCVD will be used to produce LEDs
covering the green -+ orange spectrum.
The drop in luminous efficiency at shorter wavelengths is an area of
intense research. Higher efficiency green LEDs and green laser diodes have
numerous commercial applications. The drop in efficiency cannot be fully
explained by the estimated transition between the direct and indirect bands
[15]. One possible reason is that the Al composition is increasing at shorter
wavelengths and this degrades the crystal quality with unwanted impurities.
A crucial feature of the AIGaInP device structure is the incorporation of a
wide-bandgap p-type current spreading layer that is grown on top of the
p-AIGaInP cladding layer. Selection of the material for the wide-bandgap
current spreading layer is critical. Because of the low mobility in p-type
AIGalnP and doping saturation at low levels it is difficult to grow low
resistivity AIGaInP for the upper current spreading region. Efficient light
extraction is not achieved if the injection current does not spread out under
the p electrode. One approach has been to incorporate p-AIGaAs as the
wide-bandgap current spreading layer [16]. The p-AIGaAs films have low
resistivity and are transparent to emission wavelengths above 600 nm. A
typical carrier concentration of this layer is 3 x 10 18 cm - 3, which yields films
with resistivities of 0.05 n em. This is significantly lower than for the
p-AIGalnP films. The light emission area is greatly expanded in this
structure. The external quantum efficiency of the LED with the AIGaAs
current spreading layer was 40 times brighter than that of the devices
without the AIGaAs layer. The external quantum efficiency was 1.5% at
620 nm emission for the AIGaAs layer; that is 5 times higher than for
commercial GaAsP diodes at similar wavelengths.
8.3.3 Gallium phosphide (GaP)

Gallium phosphide has been the most commercially successful LED
material for green and yellow light emission. Gallium phosphide is an
indirect bandgap semiconductor and therefore interband transitions are
not typically observed. To promote radiative transitions group V elements
such as nitrogen are added. Since nitrogen has the same valence state as
the phosphorus it replaces, it does not form donor or acceptor states.
Because nitrogen has a higher electronegativity than and different size from
the P it replaces, it forms an iso-electronic trap. This combination of
hydrostatic deformation from the size difference and electronegativity dif-
ference causes the short range attraction necessary to form an efficient
iso-electronic trap [17]. When an electron is captured by the trap it
becomes spatially localized. This spatial localization Ax fixes the range of
momentum values the electron can possess. Because of the Heisenberg
uncertainty relation
Ax Ap>h/2n
where h is Planck's constant, the range of momentum values (Ap) the
electron possesses must increase. The result is that an electron bound to an
iso-electronic trap is widely distributed in momentum space. The spread in
momentum values is great enough that there is large increase in the
probability of the electrons being at the r minimum, k = O. This bending of
the momentum conservation rules allows the nitrogen-doped indirect band-
gap semiconductor to behave as if it were a direct bandgap semiconductor.
Therefore the nitrogen-doped GaP can efficiently emit radiation through
exciton recombination at the iso-electronic trap. The trap first captures a
free carrier and then the resulting Coulombic potential attracts a carrier of
the opposite charge and a bound exciton is formed. A photon is emitted on
recombination of this bound electron-hole pair. The nitrogen iso-electronic
trap has a binding energy of only 8 meV, and therefore the subsequent
emission is close to the bandgap energy. External quantum efficiencies are
fairly low (0.2%) primarily because the exciton is weakly bound and will
dissociate at room temperature.
At high nitrogen concentrations the room temperature emission shifts

from green to yellow (590 nm). This is believed to be caused by the
formation of N-N molecular complexes. Red emission is also possible when
double doping GaP with both zinc and oxygen (Fig. 8.10). In this case, the
Zn and 0 form a deep iso-electronic trap. Since the Zn-O level has a
binding energy of approximately 0.5 eV, the exciton recombination occurs at
longer wavelengths (690 nm). Efficiencies as high as 2.8 1m W- 1 can be
achieved by doping GaP with Zn-O.
Energy
GaP
Conduction Band
--- N Trapping Level
--- Zno Trapping Level
Green Red
Valence Band
Momentum ..
Fig. 8.10 Plot of momentum versus bandgap for indirect GaP materials showing
the iso-electronic trapping levels corresponding to red and green luminescence.
8.3.4 Gallium arsenide phosphide (GaAsP)

In the red display market the dominant material system is still gallium
arsenide phosphide (GaAs1-xPx). Well-developed vapour phase epitaxy
technology has played a crucial role in the economical manufacture of
GaAsP LEDs. This ternary material is composed of an alloy mixture of the
binary III-V semiconductors GaP and GaAs. GaAsP has a direct energy
gap for arsenic compositions greater than 51 %; below that the semiconduc-
tor changes to an indirect bandgap material. Below this value the X indirect
minimum becomes lower than the r direct minimum valley. At the cross-
over point band-to-band recombination is possible with an energy of 2.1 eV,
and therefore red diodes can be manufactured from GaAs O6 P O4 By
nitrogen doping the indirect bandgap material, orange and yellow emission
can be achieved. For compositions above 49% phosphorus an increase of
Growth technologies 279
more than two orders of magnitude in quantum efficiency can be achieved
with nitrogen doping. Efficiencies as high as 11m W- 1 can be achieved for
high performance devices in the orange, yellow and green spectral region
with GaAs1-xP X' Because the probability of an electron being at the direct
minimum decreases as we move from GaAs to GaP the quantum efficiency
decreases. As in the GaP LEOs, nitrogen is added to increase the probability
of radiative transitions.
8.2.5 Wide-bandgap II-VI compounds

Blue is the only colour missing from the commercially available LEO
market. A wide range of materials has been investigated for blue emission,
most notably ZnSe, ZnS, GaN, and 6H-SiC. Recent progress in crystal
growth clearly indicates that commercial blue LEOs and displays are
inevitable. Silicon carbide is commercially available but since it is an indirect
semiconductor the external quantum efficiency is approximately 0.02%.
More promising new direct bandgap materials are under intensive research.
In particular, wide-bandgap II-VI materials of ZnSe and ZnS have achieved
reasonably efficient blue luminescence. ZnSe and ZnS with energy gaps of
2.7 eV and 3.7 eV respectively are considered the ideal source for blue
emitters because of their direct bandgaps. One of the major difficulties
associated with these materials has been the inability to obtain p-type
doping effectively. A major breakthrough in the development of ZnSe was
the achievement of high p-type doping by Hass et al. [18]. This break-
through lead to the development of blue heterostructure lasers emitting at
470 nm. Efficiencies of blue ZnSe LEOs have not been measured yet.
Another promising direct bandgap material is gallium nitride (GaN).
GaN blue LEOs grown on sapphire substrates have been reported to have
external quantum efficiencies of 0.03% [19]. These layers were grown with
thin AlN buffer layers by MOCVO. Films grown with this buffer layer have
specular surfaces free from cracks. Significant improvement in the p-type
conduction of Mg-doped MOCVO-grown GaN was achieved by low energy
electron beam irradiation [20]. Further studies are necessary to understand
the mechanism of p-type conversion caused by the electron beam treatment.
8.4 GROWTH TECHNOLOGIES
8.4.1 Vapour phase epitaxy (VPE)

Vapour phase epitaxy is the most commercially important growth technol-
ogy for producing GaAs 1_ x P x LEOs. VPE relies on the gaseous transport
of the constituent column III and column V sources to the deposition zone.
Precise control of the molar flow rates of the two column V sources is
achieved by mass flow controllers. This allows independent control of the
partial pressures that govern alloy composition grading. A high degree of

flexibility in the composition and doping levels is achieved by accurate mass
flow control. The versatile nature of VPE allows synthesis of the import-
ant III-V compounds such as GaN, ZnAs, GaP, InGaP, InAsP, and
GaAsSb. However, the important ternary AlxGal-xAs compounds cannot
be grown by VPE since the AICI species reacts with the hot quartz
walls resulting in unwanted impurities in the epitaxial film [3]. A schematic
diagram of a large capacity multi vapour VPE reactor is shown in
Fig. 8.11.
rr:=:::~~~~..-
..-
PH3 + AsH3 + H2 + dopant
H2 +HCI
. . - H2
Fig. 8.11 A vertical multiwafer hydride based vapour phase epitaxy reactor for
growing GaAs,-xP x. (After Burd et al. [21].)
One of the main advantages of VPE is the ease of scale-up to a high

capacity commercial process. In this system, hydride sources AsH3 and PH 3
provided the elements As and P; therefore this technique is more specifically
called hydride VPE. Volatile GaCl is formed by introducing a mixture of
HCI and H 2 . The typical growth rate is between 12 and 30!lm h -1,
depending on gas flow.
Hydride VPE is an equilibrium process in which the crystal growth is
achieved by lowering the growth temperature of the substrate relative to the
temperature of the reaction zone. For GaAs 1 _ x P x growth the Ga reservoir
portion of the furnace is held at approximately 800 0c. GaCI is formed by
introducing a mixture of HCI and H2 to the Ga reservoir. The reaction zone
is kept at 900C to pyrolyze the hydrides PH 3 and AsH 3. During deposition
the substrate is kept at a lower temperature of approximately 750C so that

the vapour becomes supersaturated and deposition occurs.
The hydride VPE process can be described by the following set of
chemical reactions: gallium transport by chloride formation in the gallium
zone,
T-800C
2HCI(g)+2Ga(l) < ' 2GaCI(g)+H2(g)
transport and pyrolysis of the hydrides in the reaction zone,
T-900C
2PH 3(g) < ' ! P 4 (g)+3H 2(g)
T-900C
2AsH3(g) < ' !As4 (g)+3H 2(g)
deposition zone,
2GaCI(g)+ H 2(g) + HI - x)AS 4 (g)
T-750C
+!xP 4 (g) , ' 2GaAs 1 - X P X (s)+2HCI(g)
Inspection of the deposition equation indicates that the composition of the
deposited film can be controlled by adjusting the pressure of PH 3 relative to
AsH 3. The typical red GaAsP LED is grown on a GaAs(lOO) oriented
substrate by grading the alloy composition in the first layer from X = 0.0 to
X = 0.4. This buffer layer reduces the stress and defects caused by depositing
the epitaxial layer on a substrate with different lattice constant and thermal
expansion coefficient [22].
Another type of VPE system, halide VPE, uses AsCh and PCh sources in
place of the more toxic hydride sources. The governing chemical reactions
are thought to be as follows:
2PCh(g)+3H2(g)~! P 4 (g) + 6HCI(g)
2AsCI 3(g)+ 3H2(g)~! As 4 (g)+ 6HCI(g)
2HCI 3(g) + 2Ga(l) ~ 2GaCI(g) + H2(g)
Since extra HCI is being generated by the column V halide it is important to
introduce the AsCI 3, or PCh, downstream from the Ga reservoir to avoid
unwanted side-reactions. One advantage of VPE over other growth tech-
nologies, such as LPE or MOCVD, is the ability to etch the surface in situ,
thereby producing a pristine substrate. This is accomplished by eliminating
the hydrogen flow such that the deposition reaction transforms into the
following etching reaction:
2GaCI(g)+ ! As 4 (g) ~2GaAs(s) + Ch(g)
At typical growth temperatures the reaction proceeds to the left, causing a
removal of material from the substrate.
8.4.2 Metal-organic chemical vapour deposition (MOCVD)
In the past few years, MOCD has evolved into a promising technique for
production of visible light emitting opto-electronic devices. Recently, very
high brightness yellow and orange emitting LEDs have been demonstrated
in the AIGaInP materials system [14, 16]. High purity materials and
monolayer abrupt heterojunctions have been demonstrated by this tech-
nique. Because of the economy and simplicity of the technology, MOCVD is
the most viable technique for scale-up to large area epitaxial growth for
commercial applications. MOCVD is a non-equilibrium vapour phase
growth technique that is typically operated at pressures near atmospheric
pressure. In this technique a metal alkyl, such as trimethylgallium (TMGa),
is mixed in the vapour phase with a hydride, such as arsine (AsH 3), and the
reactants are simultaneously injected into the growth chamber where they
thermally decompose to form single-crystal layers.
The MOCVD technique originated from the early research of Manasevit
in 1968 [23] who demonstrated that triethylgallium (TEGa) and arsine
deposited single-crystal GaAs pyrolytically in an open tube reactor.
Manasevit and co-workers [24-26] subsequently expanded the use of this
technique for the growth of GaAs1_yPy, GaAs1_ySby, and AI-containing
compounds. Composition and growth rate are controlled by precisely
controlling the flow rate and dilution of the various components of the gas
stream. The source materials are more often than not pyrophoric and
extremely toxic. This is a special concern when it comes to gaseous sources,
such as the group V elements that are traditionally supplied from their
hydrides: arsine (AsH3) and phosphine (PH 3). Both these sources present a
serious safety problem, since they must be stored in compressed gas
cylinders, which allows for the possibility of catastrophic release. However, a
trend has emerged toward using new group V organometallic liquids, such
as t-butylphosphine (TBP) and t-butylarsine (TBA), which have, until
recently, not been commonly available in sufficient purity for quality
semiconductor growth. The group III sources are almost universally or-
ganometallic liquids, such as trimethylgallium (TMGa) and trimethyl-
aluminum (TMAI). The liquid sources are stored in bubblers through which
a hydrogen carrier gas flows. The carrier gas will saturate with vapour from
the liquid and transport the vapour to the substrate. The substrate rests on a
block of graphite called a susceptor that can be heated by an r.f. coil,
electrically, or radiantly by an infrared lamp heater.
The most common reactor chamber designs are the vertical and horizon-
tal configurations. In the vertical design, reactants are injected through
the top. Typically, the substrate is held flat on a rotating silicon carbide
coated graphite susceptor that is perpendicular to the gas flow direction
as shown in Fig. 8.12. Heating is accomplished by r.f. induction, and
temperature monitoring is accomplished by either an infrared pyrometer or
MFC's (Computer Contro lied)
Vent Flush H2
Main H2
AsH3 I nj ect ion

Manifold
Disilane Computer
Controlled
DEZn
Differential Pressure _ _H r - -
Transducer:~
Burnoff Quartz Baff] Needle

Valve
Susceptor
Fig. 8.12 Schematic of vertical reactor MOCVD system.
a thermocouple. The low gas velocities employed in a conventional

MOCVD reactor (2-4 cm s -1) promote laminar flow conditions. The addi-
tion of a quartz disc baffle at the gas inlet promotes uniform gas distribu-
tion.
The horizontal design utilizes a susceptor that is situated approximately
parallel to the gas flow direction. Minimal convection currents and highly
laminar gas flow conditions exist in the horizontal reactors [27]. Uniform
growth can be achieved after precise tilting of the susceptor to eliminate
reactant depletion along the flow direction. Heating in this design can be
accomplished by either d. induction or infrared heating from quartz-
halogen lamps. Multiple wafer design typically requires wafer rotation for
uniform deposition.
The exact chemical pathways operative in the MOCVD process are not yet
clearly understood. The basic reaction most commonly used to describe the
growth of the compound semiconductor GaAs by MOCVD is as follows:
Ga(CH3h(v) + AsH3(v)-+GaAs(s) + 3CH 4 (v)
This balanced expression ignores the facts that the specific reaction path and
reactive species are largely unknown. A more likely reaction pathway
leading to growth of the GaAs epitaxial layers involves the homogeneous
decomposition of TMGa as was reported in an earlier study [28, 29]:
Ga(CH 3h(v)+1 H2(v)-+Ga(v)+ 3CH 4 (v)
Arsine is thought to decompose heterogeneously to yield AS 4 and AS 2 at
normal growth temperatures:
Therefore the growth of GaAs that probably occurs at the solid-vapour

interface could be expressed as follows:
Ga(s-v) + t AS 4 (s-v)-+ GaAs(s)
However, the level of understanding of the growth process is inadequate at

best. The most difficult topic, and certainly the least developed, is the area of
the kinetics of the process and growth mechanisms occurring at the
solid-vapour interface during MOCVD growth. Most researchers have
optimized their MOCVD reactors by empirical studies of external par-
ameters such as growth temperature, V:III ratio, substrate tilt and mass
flow rates. However, these studies yield little information about the mechan-
istic steps that eventually lead to epitaxial growth.
These studies have identified three regions of growth: mass transport
limited, desorption and surface kinetically limited regimes. As shown in Fig.
8.13, kinetics of surface reactions dominates the growth in the low tempera-
ture regime. Conventional MOCVD is usually peformed in the mass
transport limited regime that occurs over a wide temperature range (500 0_
800C). At high temperatures, desorption of arsenic begins to limit the

growth and a decrease in the growth rate becomes evident.
In this temperature region, 55(}-850C, growth is limited by mass trans-
port of the column III reactant to the growing interface. Because the
diffusion process is slightly temperature dependent, there exists a slight
increase in the growth rate in this temperature range as shown in Fig. 8.13.
From Fick's law we can describe the flux of column III elements toward the
substrate in the mass transport limited regime as
J _ DGa(P&a-PGa)
Ga- bb RT
Temperature (0C)
850 750 650 550 450
1
Mass transport
+
~ Surface
c:
E kinetic
---
.3 It'limited
...I!!
CD
.1
...~
..r:: Ea = 19 kcal/mol
e
C)
Ea = 22 kcal/mol
.01
0.9 1.~ 1.3 1.5
1000/T(K)
Fig. 8.13 MOCVD growth rate versus reciprocal temperature for epitaxial layers
on (100) and (111) Ga oriented GaAs substrates. (After Reep et al. [30].)
where Ob is the concentration boundary layer thickness, DGa the diffusion

coefficient for Ga in the vapour phase, Pga is the input gas stream partial
pressure of TMGa, and P Ga is partial pressure at the gas-solid interface.
For temperatures above 550C, the metal alkyl pyrolysis efficiency is unity;
then we can assume Pga~PTMGa. As the partial pressure of Ga at the
interface is small in comparison with the input concentration of TMGa
(PTMGa~PGa), we can express the column III flux at the interface as
J _ DGaPTMGa
Ga- ob RT
From the above equation we can see that control of the growth rate is easily
obtained by controlling the number of moles of metal alkyls injected into the
reactor. Precise control of the growth rate is then achieved by use of
accurate electronic mass flow controllers that control the flow through the
metal alkyl bubbler. In addition, a uniform boundary layer thickness Ob
ensures uniform deposition.
In the temperature regime 300-550 e, the kinetics of surface reactions
dominates the growth. Reep and Ghandi [30] theorized that the growth
process is dominated by adsorption and decomposition of arsine at the
growing interface. This view is supported by the fact that the activation energy
for the heterogeneous decomposition of arsine on a single-crystal GaAs(l00)
surface of 18 kcal mol- 1 [28] is approximately equal to the activation energy
of 19 kcalmol- 1 observed for growth on the GaAs(l00) orientation. How-
ever, the presence ofTMGa still plays an important role since the growth rate
is strongly influenced by the partial pressure of TMGa in the reactor.
For the growth of AIGaAs the solid composition is found to be propor-

tional to the flux ratio of Al and Ga at the growing interface. Since the
diffusion coefficients of Al and Ga are approximately equal, composition
control of AlxGal-xAs is achieved by linear adjustment of the molar flows
(MF) of the column III sources. Data in Fig. 8.14 illustrate this simple
relation of composition in the solid phase to mole fraction in the gas phase.
This simple relation is described by the following equation:
MFAl
XAl=,,------:=--':"'::c:~
MFAl+MFGa
1.0
~ 0.8 ImPLl
c:
0
~
..;::;
en 0.6
0
a.
E 0.4
0
u
:2
(5 0.2
(f)
0.0
0.0 0.2 0.4 0.6 0.8 1.0
Mole fraction (AI)/[(AI) + (Ga)]
Fig. 8.14 Compositional control in MOCVD is achieved by linear adjustment of
the molar fraction of the constituents.
This equation illustrates one of the greatest assets of the MOCVD growth
process: the distribution coefficient is essentially unity for the column III
elements. This makes the growth of AI-In compounds feasible, which is in
contrast to liquid phase epitaxy where the growth of AI-In compounds is
unachievable because of the large liquid distribution coefficient [31]. For the
growth of AIGaInP, we can write a balanced MOCVD expression as:
!(x)AI(CH 3 h +!(1-x)Ga(CH 3 h +!In(CH 3 h + PH 3 --+3CH 4

+(AlxGal-x)o.5Ino.5 P
8.4.3 Liquid phase epitaxy (LPE)

In LPE growth of an epitaxial layer results from nucleation of a super-
saturated solution. The solute will freely nucleate into the coldest seeded
region, which is usually the substrate. Nucleation onto the substrate will
occur when the solution reaches a critical supersaturation. This is achieved
by different methods. The most common one is to place the substrate in
equilibrium with a saturated solution and then to begin to cool the system
to achieve a supersaturated condition (Fig. 8.15). Growth proceeds by the
transport of material from the solution into the edge of the boundary layer,
where it must diffuse across before the material is adsorbed into the surface
of the substrate. The limiting factor during LPE growth for this method is
the diffusion of the solute through the boundary layer. A different approach
requires the addition of solute to a saturated solution while its temperature
is maintained constant.
Fused - Silica
FumaceTube
Graphite Barrel
Pull
Graphite
.......... ,~Ii~~~H2
Growth Seed
Fig. 8.15 Schematic representation of a multilayer LPE apparatus employing the
sliding graphite boat technique. (After Casey and Panish [9].)
The main advantage of this technique is the ability to grow a layer of

semiconductor material at stoichiometric composition at a temperature
substantially lower than the melting points of the constituents. In addition,
the growth of the layer can be kept free of damage at the substrate interface
and the epilayer by an in situ dissolution or meltback. This is important for
the growth of high quality layers. LPE-grown epilayers contain low levels of
deep impurities because their distribution coefficients are lower than those
of shallow impurities. This makes LPE suitable for growing optical devices.
Some of the disadvantages of LPE include the inability to produce layers
of uniform thickness and the need to wipe off the solvent after the layer is
grown. In addition, the growth rates are too large to grow nanometre
precision layers. The growth of multilayer structures with extremely abrupt
interfaces is difficult. Another limitation is the difficulty in growing alloys
containing both Al and In. The major problem with LPE for manufacturing
is the difficulty of growing uniform layers over large surface areas.
There are some constraints that determine materials suitable for LPE
growth. It is required that the material to be grown be soluble and that the
resulting solution melt at a temperature lower than the melting point of the
material. The best-suited materials are those of the eutectic type that are free
from compound formation. Most III-V semiconductors behave like this and
since the resulting layer will be saturated by the solvent during growth, a
very small solid solubility of the solvent in the grown crystal is required. The
choice of solvent also depends on other factors, such as its availability in

highly pure form, its vapour pressure at growth temperature and its
reactivity with the crucible in which the LPE growth is performed.
An epitaxial layer is grown by LPE by precipitation of material from a
saturated or supersaturated solution directly onto a crystalline substrate. If
the substrate consists of a single crystal and the material to be grown is a
latticematch to the substrate, the material will precipitate onto the substra-
te surface and maintain the quality of the crystal structure. The phase
equilibria for the systems of interest typically are such that it is usually
possible to grow from a liquid solution where the major constituent of the
liquid is also a major constituent of the solid and the remaining compo-
nents of the liquid are relatively dilute.
( a) Liquid phase epitaxy - growth rates for step cooling and uniform cooling
There are two very distinct ways in which growth from a liquid solution
can be initiated. The first method involves bringing a substrate into
contact with a melt at equilibrium and then slowly cooling the melt and
substrate. This method is commonly called uniform cooling (UC). The
second method, step cooling (SC), involves the saturation of the melt at
some temperature T hot at which point the temperature is reduced to Teold,
such that the melt becomes supersaturated and the substrate is brought
into contact at the same temperature and serves as a source of nucleation
sites for growth to be initiated. In both these techniques, growth at the
substrate sets up a compositional gradient in the melt by locally depleting
the melt of the more dilute of its constituents. It is the diffusion through
this expanding depletion layer in the melt that determines the growth
rate.
For illustration, we will consider the growth of GaAs from a Ga-rich melt.
It is assumed that the concentration of As is the determining factor, since it
is the more dilute constituent of the melt. The solution will apply in general
to any growth from a melt where a single constituent is the limiting factor
for growth. The amount of solute per unit area M t leaving the solution at the
crystal surface in time t can be obtained by integration of Fick's first law of
diffusion:
M t= ft D[dC~s(Z' t)] dt
o dz z=o
where D is the diffusivity, and C~s is the concentration of As in the liquid.
z=O is then assigned to the substrate surface. The resulting layer thickness,
d, is given by
d= M t
C~s
where C~s is the concentration of As in the solid. Hsieh [32] evaluated M,
by solving the one-dimensional diffusion equation
D d2C~s(z, t) = dC~s(z, t)
dz 2 dt
under the boundary conditions for uniform cooling of C~s(z, 0) = C~s(Thot)
and C~s(O, t)= C~s(Thot)-(R/m)t, where m is given from the slope of the
liquidus curves as
The result is
d= ~ ~ (!!.)1/2 t3/2
3 C~sm 1t
The boundary conditions for pure step cooling are C~s(z, 0) = C~s(Thot) and
C~s(O, t) = C~s(Tcool). The solution is
d= 2~T (!!.)1/2 t3/2

CAsm 1t
The result for a combination of step and uniform cooling in the LPE growth
of III-V semiconductor is given by the sum of the two results above:
d = _s1_ (!!.)1/2(2 AT t1/2 + ~ Rt3/2)

CAsm 1t 3
From inspection of the above equation we can see that for short growth
times the thickness is controlled by step cooling. At longer times the uniform
cooling rate dominates.
8.4.4 Molecular beam epitaxy (MBE)

Molecular beam epitaxy is currently the only technique that can grow the
highly p-type ZnSe:N layer needed for a blue LED. The key to obtaining
p-type ZnSe was to use radiofrequency power to generate a stable nitrogen
plasma. The plasma yielded a free radical source (FRS) of nitrogen which
then acted as a p-type dopant in ZnSe [18]. MBE growth is a very
sophisticated ultra high vacuum (UHV) technique for thermal deposition of
atomic or molecular films. In conventional MBE, thin films crystallize on a
heated substrate. The material for growth typically comes from a set of
elemental furnaces that are designed for the careful control of the flux. MBE
has evolved as the premiere technique for the fabrication of high quality
semiconducting structures and has almost single handedly initiated a whole
field of research called band structure engineering. MBE owes much of its
advances to the development of reliable UHV technology that allows for

vacua better than 10- 10 Torr to be routinely achieved. The purity of
material grown by MBE is determined by the purity of the sources (which is
quite good) and the background pressure of the system. Since MBE is a high
vacuum technique, the growing surface is available for analysis. In particu-
lar, reflection high energy electron diffraction (RHEED) allows growth to be
monitored and controlled on the monolayer level. Some disadvantages of
MBE are that growth rates are slow - about one monolayer per second.
This, combined with its relatively low volume per run, leads to low overall
throughput. However, MBE has been used in an industrial setting to
fabricate laser diodes for CD players and other consumer products. Here,
the large volume and system dedication have allowed the price per unit to
drop significantly.
8.5 CONCLUSIONS
Significant progress has been made in the epitaxial growth and properties of
visible LEDs in recent years. Displays based on LEDs continue to compete
with thin film EL devices. Notable progress has been made in the develop-
ment of AIGaInP and AIGaAs efficient emitters. Efficiencies of these
materials have achieved levels high enough for large panel outdoor display
applications. An efficient blue emitter is required for full colour displays
based on LEDs to become viable. ZnSe, GaN and SiC are all being
developed as sources for blue LEDs. Currently, GaAsP and GaP are still the
dominant materials used in the marketplace. Newly developed materials
such as AIGaAs and AIGaInP will replace the existing LED materials as
production costs are reduced.
REFERENCES
1. Craford, M.G. (1977) IEEE Trans. Electron Devices, 24, 935.
2. Henry, c.H. (1976) J. Luminescence, 12-13, 47.
3. Kressel, H. (1977) Semiconductor Laser and Heterojunction LEDs, Academic
Press, New York.
4. Shockley, W. (1950) Electrons and Holes in Semiconductors, Van Nostrand, New
York.
5. Bergh, A.A. and Dean, P.I. (1976) Light Emitting Diodes, Oxford University
Press, London.
6. Varon, 1., Mahieu, M., Vandenberg, P., Boissy, M. and Lebailly, I. (1981) IEEE
Trans. Electron. Devices, 4, 416.
7. Steranka, F.N., DeFevere, D.c., Camras, M.D., Tu, C.W., McElfresh, D.K., Rudaz,
S.L., Cook, L.W. and Synder, W.L. (1988) Hewlett-Packard J., (August), 84.
8. Capasso, F. (1987) J. Appl. Phys., 24, 319.
9. Casey, H.C. and Panish, M. (1978) Heterostructure Lasers, Academic Press, New
York.
References 291
10. Tsai, M.J. (1984) Proceedings of the 11 th IntI. Symp. on Gallium Arsenide and
Related Compounds, Institute of Physics Publishing, Bristol, p. 26.
11. Cook, L.W., Camras, M.D., Rudaz, S.L., and Steranka, F.M. (1987) Proceedings
of the 14th Inti. Symp. on GaAs and Related Compounds, Institute of Physics
Conf. Series, 91, Institute of Physics Publishing, Bristol, p. 777.
12. Kellert, F. and Moon, RM. (1986) J. Electron. Mater., 15, 13.
13. Ikeda, M., Nakano, K., Mori, Y., Kaneko, K. and Watenabe, N. (1986) Appl.
Phys. Lett., 48, 89.
14. Kuo, e.P., Fletcher, RM., Ostenowski, T.D., Lardizabal, M.C., Craford, M.G.
and Robbins, V.M. (1990) Appl. Phys. Lett., 57, 2937.
15. Bour, D.P., and Shealy, J.R. (1988) IEEE J. Quant. Electron. 21, 1856.
16. Sugawara, H., Ishikawa, M. and Hatakoshi, G. (1991) Appl. Phys. Lett., 58, 1010.
17. Campbell, J.e., Holonak, N., Craford, M.G. and Keune, D.L. (1974) J. Appl.
Phys., 45, 4543.
18. Hase, M.A., Qiu, J., DePuydt, J.M. and Cheng, H. (1991) Appl. Phys. Lett., 59,
1272.
19. Akasaki, I., Amano, H., Hiramatsu, K., Koide, N. and Sawaki, N. (1987)
Proceedings of the 14th IntI. Symp. on GaAs and Related Compounds, Institute of
Physics Conf. Series, 91, Institute of Physics Publishing, Bristol, p. 633.
20. Amano, H.K., Kito, M., Hiramatsu, K., Akasaki, I. (1989) Jpn. J. Appl. Phys., 28,
L2112.
21. Burd, J.w. (1969) Trans. Metall. Soc. AIME, 245, 571.
22. Abrahams, M.S. (1969) J. Mater. Sci., 4, 223.
23. Manasevit, H.M. (1968) Appl. Phys. Lett., 12, 156.
24. Manasevit, H.M. and Simpson, W.I. (1969) J. Electrochem. Soc., 116, 1725.
25. Manasevit, H.M., (1971) J. Electrochem. Soc., 118, 647.
26. Manasevit, H.M. and Hess, K.L. (1979) J. Electrochem. Soc., 126,2031.
27. Giling, L.1. (1982) J. Phys. (Paris), Colloq. C5, 235.
28. DenBaars, S.P., Maa, B.Y., Dapkus, P.D., Danner, A.D. and Lee, H.e. (1986) J.
Cryst. Growth, 77, 188.
29. Nishizawa, J., Abe, H. and Kurabayashi, T. (1985) J. Electrochem. Soc., 132,
1197.
30. Reep, D.H. and Ghandi, S.K. (1983) J. Electrochem. Soc., 130,675.
31. Yuan, J.S., Hsu, e.e., Cohen, RM. and Stringfellow, G.B., (1985) J. Appl. Phys.,
57, 1380.
32. Hsieh, J.J. (1974) J. Cryst. Growth, 20, 49.
9
Atomic layer epitaxy of phosphor
thin films
Brian W. Sanders
9.1 INTRODUCTION
9.1.1 History of atomic layer epitaxy

The term atomic layer epitaxy (ALE) was originally coined by Suntola to
describe a new method of depositing thin films using a slightly modified
molecular beam epitaxy (MBE) system. It was originally developed to meet
the need of improved ZnS and dielectric thin films for a.c. thin film
electroluminescent displays [1-3]. By 1980, the technique was being used
by Lohja Corporation to produce large area thin film electroluminescent
(EL) displays based on ZnS:Mn. The technology improved. In 1982, even
better displays produced by this technique resulted in Lohja winning the
prize of the Society for Information Display for the greatest advance in the
display field during the preceding year. At this point, ALE began to attract
considerable attention as a method for producing thin films of high
quality. The ideas behind ALE have subsequently been applied to many
deposition methods to produce excellent quality II-VI and III-V thin
films.
One needs only to view the literature to see how important ALE has
become to thin film growth. The technique is being applied by research
groups around the world and has been reviewed by authors in many
countries (for example, Herman [4], Goodman and Pessa [5], Peng [6],
Suntola [41]). There have been international conferences concerning the
technique [8], and discussions of the method have appeared in many book
chapters (e.g. Wantanabe et al. [9]). At least one book has been written on
the subject [10].
However, what precisely is ALE? How is it implemented? What are its
advantages over conventional growth techniques? Which display materials
293
294 Atomic layer epitaxy of phosphor thin films
have been produced by this technique? What apparatus is required? These

questions will be addressed in the following sections.
9.1.2 What is ALE?

The acronym 'ALE' describes the way in which the conventional MBE
process was altered by Suntola to produce ZnS thin films. Use was made of
the great difference between the strength of a Zn-Zn or an S-S bond versus
a Zn-S bond, the bond between unlike elements being much stronger.
During the growth of the film, the effusion cells were shuttered to expose the
substrate to alternating pulses of the individual elements. A pause was
inserted between pulses to allow excess reactants to be flushed away. A
growth temperature was maintained such that only bonds between unlike
elements were maintained (bonds between like elements are weaker and
therefore more easily broken) during the pause, (i.e. the system to comes to
equilibrium, one new monolayer per cycle is formed). The growth process is
self-limiting, with only bonds between unlike elements being strong enough
to last out the pause or purging step. In this way elemental atomic layers
were deposited one by one on the substrate (more on the mechanism can be
found in section 9.2).
Because the film is grown one monolayer at a time, much better control
over the deposition is accomplished when compared with MBE. Since the
layers are built up one by one, good control over individual layer thick-
nesses results. Sharp interfaces are produced with little mixing between
layers. This offers exciting possibilities for the growth of superlattices.
One of the best controlled parameters during the film growth is film
thickness. Because the film is grown layer by layer, the film thickness can be
determined by the number of reaction sequences. This eliminates the need
for external thickness monitoring methods.
9.1.3 Conventional deposition methods altered for ALE-type growth

This deposition technique is not restricted to MBE units using elemental
source materials (as in its original application). It has since been applied to
the production of many inorganic thin films using many deposition tech-
niques. A brief survey appears in Table 9.1. The application of this growth
type to each standard deposition technique is not surprising, given its
superior control over the film growth when compared with the parent
growth method. The biggest drawback is the time required to deposit a film.
Deposition times are much longer for an ALE variation of a growth
technique than by the parent technique. This is due to the requirement that
the reactants must be fed to the substrate one at a time. Because of the large
number of groups using this technique, and the many growth methods it has
been applied to, a large variation in terminology has been used to describe
Introduction 295
Table 9.1 Some compounds grown by various atomic layer deposition techniques
Deposition Compounds Reactants Reference
technique deposited
Molecular beam ZnS Elemental Zn and S [11]

epitaxy (includes ZnCl 2 and H 2S [12]
organometallic Zinc acetate and H 2S [13]
MBE) Zn(CH3h and H 2S
ZnO Zinc acetate and water [14]
Al 20 3 AlCl 3 and water [14]
Organometallic ZnS Zn(CH3h and H 2S [15]
chemical vapour SrS, CaS, BaS M(thdh where [16]
deposition M=Ca, Sr, Ba
and thd = 2,2,6,6-
tetramethyl-3,5-
heptanedionate
MgO Mg (C2HSh and water [17]
Solution growth ZnS, CdS, ZnS04, CdS0 4, or [18]
Zn1_xCdxS ZnCI 2, CdCI 2, and
Na2S
Electrodeposition CdTe CdS0 4 and Te02 [19]
deposition methods that result in atomic layer growth of a film. This has
resulted in some confusion in accessing the literature.
9.1.4 Terminology
The terms successive ion adsorption and reaction [18], molecular layer
epitaxy [20], digital layer epitaxy [21], electrochemical atomic layer epitaxy
[19], flow-rate modulation epitaxy (FME) [22], and modulated molecular
beam epitaxy [23] have all appeared in the literature to describe thin film
deposition methods relying on the alternating exposure of reactants to a
substrate resulting in film growth of one monolayer per cycle. The prepon-
derance of these terms has become confusing especially when, often as not,
the films resulting from the growth techniques are polycrystalline and
therefore do not fit the term epitaxy.
The techniques of flow-rate modulation epitaxy and modulated molecular
beam epitaxy are a different class of technique. They do result in a net
growth of one atomic layer per cycle of a material, but they do not rely on
self-limiting reactions at the surface of the growing film. They rely on simply
limiting the amount of reactant supplied to the substrate at each step to only
that which is necessary for monolayer formation. In this way a purging step
is not required for the film growth. Although these deposition techniques do
result in good control over film growth, they will not be discussed in this
chapter. The reader is referred to the references if further information is
desired. For those techniques that rely on the historical definition of atomic
layer epitaxy, a new naming system has recently been suggested.
Most of the names mentioned above contain the term epitaxy, which is
usually used to describe the growth of single-crystal films. This has resulted
in some confusion as many films prepared by ALE are polycrystalline or
amorphous. Since the term is meant to describe the way in which the film
grows, not the crystallinity, it has been suggested by Leskela [14] that the
term atomic layer deposition (ALD) be adopted for the production of
polycrystalline films by this method of deposition. A suitable prefix can be
used to illustrate the primary deposition method. Using this terminology,
chemical vapour atomic layer deposition (CV ALD) would be used for the
deposition of a polycrystalline material through the exposure of a substrate
to alternating streams of chemical vapours. Conversely, molecular beam
atomic layer epitaxy (MBALE) refers to the production of single crystal thin
films by exposing a substrate to alternating elemental molecular beams. The
suggestion seems reasonable. If it is widely adopted it should make the
literature on this subject much more consistent and access less confusing.
This terminology will be adhered to by the author from now on when a
specific deposition technique is being explained. The more general terms
ALE and ALD will be used when the specific growth method does not
matter in the discussion.
9.2 THEORY OF ALE
9.2.1 Reaction mechanisms - elemental starting materials

ALE only occurs when conditions in a conventional deposition method
have been altered so that the substrate is exposed to only one reactant at
time. Deposition parameters (e.g. temperature or time between reactant or
purge pulses) are arranged so that chemisorption of only one monolayer of a
particular element forms at any time. Figure 9.1 gives a qualitative indica-
tion of the mechanism of ALD.
In this figure, ALE is demonstrated for the simplest case: alternating
exposure of a substrate to elemental reactants [24]. The first element
chemisorbs on the surface and subsequent atoms do not adhere to the
surface (frame 1). The nonabsorbed atoms are then swept (or pumped) away
in a purging step (frame 2). The substrate is then exposed to a pulse of the
second element (frame 3), generating a second monolayer. This is then
followed by a purge step (frame 4). Repeating this sequence builds up the
thin film one monolayer at a time. To understand the basic principles
behind ALD it is helpful to introduce a simple model based on the ideas of
chemisorption and physisorption [24].
Theory of ALE 297
IIIIIIIIII

1
IIIIIIIIII

2
IIIIIIIIII

@@@@@@@@@
Fig. 9.1 Schematic of ALD using elemental starting materials. (After Suntola and
Hyvarinen [24].)
Figure 9.2 shows an energy diagram for the absorption of an atom to the
surface of a solid. Excluding the possibility of condensation, below some
temperature Tl only physisorption occurs. Thin film growth is influenced
greatly by the substrate temperature and time between reactant pulses as
only a weak association occurs between the surface and any adsorbed
>
...CD
C)
C
CD
I
.;;
C
....oCD
0..
Chemisorption
Distance from surface

Fig. 9.2 A plot of potential energy versus distance for a chemisorbed and a
physisorbed molecule or atom. (After Suntola and Hyvarinen [24].)
atoms. The whole monolayer would desorb given enough time. Between the
temperatures Tl and T2 actual ALD occurs as the molecules or atoms will
adsorb but not desorb from the surface (chemisorption). Additional mono-
layers do not form owing to the differences in bond energies between like
and unlike elements. Above T2 the growth is again dependent on tempera-
ture and time between pulses as atoms are desorbing from the surface. ALD
only occurs when chemisorption is the sole process occurring. To attain this,
substrate temperature must be adjusted to within some optimum range.
Extending this model further to cover compound reactants results in Fig.
9.3. This figure describes processes that compete with ideal ALD when
compound starting materials are used. If the temperature is too low,
desorption and/or incomplete decomposition of the reactants occurs and an
uncontrolled growth results (region B). Condensation of the reactants may
occur, which results in a higher than ideal growth rate (region A). If the
temperature is too high, then re-evaporation of elements from the surface
may occur (region D) or, for compound reactants, undesirable decomposi-
tion or side reactions may result (region C). ALD is expected to occur only
over a limited temperature range [7].
This temperature range must be high enough to prevent condensation and
to overcome the activation energy to chemisorption, but low enough to
prevent unwanted reactions or re-evaporation. Thus, the ratio between the
vaporization temperature of source elements and the ZnS film determines
the width of the ALD 'window'. It is therefore expected that using com-
pounds with low vaporization temperatures, instead of elemental source
Theory of ALE 299
!II 2
!II
!II G)
!II C
G).,lIi:
.i .!:!
.!:! -E
Ideal ALE
-E :D 1
iij r; or ALD
~o
o C
<C 0
~
Temperature
Fig. 9.3 Schematic showing the relationship between ideal ALD growth and
competing processes with temperature when compound reactants are used. (After
Suntola [7].)
materials, will increase the temperature range over which ideal growth will
occur.
9.2.2 Growth using compound starting materials

A typical growth sequence for ideal ALD using compound reactants is
shown on the left of Fig. 9.4. A similar sequence of events occurs during
CV ALD as with elemental reactants.
The substrate is first exposed to one of the reactant compounds. This
compound then adsorbs onto the surface forming a monolayer. A subse-
quent purge step removes any excess reactant. In frame 3 the second
compound enters the system and reacts with the initial monolayer. A purge
step removes any excess of the second reactant, and then the sequence can
start again at frame 1. Again the film grows one monolayer at a time [24].
The growth of films using CV ALD can be more complicated than when
elemental sources are used. Some branching of the reaction sequence can
occur. This is illustrated in the two branches from step 2 of Fig. 9.4. Frame
2a demonstrates one possibility, the first compound to adsorb to the
substrate might 'crack' on its surface leaving elemental monolayer coverage.
The second compound might also 'crack' when exposed to the heated
substrate. The result is the same as if the cracking did not occur. More
complicated still is the possibility of partial cracking of the reactants on the
surface (frame 2b). A subsequent reaction with the second compound also
would form a complete film.
The reality of crystal growth seems to dictate that the ideal reactions
pictured above do not always hold true. Atomic layer deposition of ZnS
using compound starting materials such as zinc chloride [14] or dimethyl
~~~~I/~~
0~ 0+00 70~0 70~
0
-
0
0-
0
0
_0
(1)
0- 1
--- --- --- --- -- ..

- 2b
---r-------------~
0\~0/0\~0rJ0~0rJ0 r
2 28
~~I/~~~L/
0000Gr0

~070 15
cfJ 00 cfJ 00
~0cfJ 0
3
'~~L / ,
0000000 ,! L/
4 0_0 Compound A # Compound B
Fig. 9.4 Ideal ALD from compound reactants (after Suntola and Hyvarinen [24]).
Also shown are two possible branches that still could result in ideal monolayer per
cycle deposition.
zinc [15] and hydrogen sulphide does not always produce one monolayer per
cycle growth. The side groups on the zinc compounds take up space on the
surface of the monolayer and do not let a completed monolayer form. Thus,
less than one monolayer per cycle is observed. In the case of dimethyl zinc, only
about 80% of a monolayer forms during the film growth. When zinc chloride is
used, values as low as 33% of ideal ALD have been observed [14].
A theoretical treatment of this behaviour has been published by Pak-
kanen et al. [25] (it has also appeared in a review by Suntola [41]). The
authors use Hartree-Fock calculations for determining the thermodynamics
of absorption of dimethyl zinc on the (111) planes of zinc sulphide for
growth in the [111] direction (the direction most ZnS films deposited by
ALD grow). They discovered that, if molecules adsorb singly on the surface,
then only one-third of the surface may be covered with the zinc compound
at any time. The large chlorine atoms prevent the zinc atoms from getting
close enough to one another for a complete layer to form. Their calculations
also show that it is energetically favourable for the zinc chloride to chain
across the surface. When this occurs, 50% or even 100% of an ideal
monolayer can form from the zinc atoms. A combination of the above
reaction schemes can be used to explain almost any fraction of ideal growth
encountered. This model also might be applicable to the use of dimethyl
zinc, as a methyl group is approximately the same size as a chlorine atom.
Equipment used to deposit thin films by ALD 301
In summary, the deposition of compounds by ALD is extremely complex.

Temperature of the substrate is critical in determining whether monolayer
stepwise growth will occur. The use of compounds gives a larger 'tempera-
ture window' for ALD but decreases the coverage that can be expected for
any given monolayer owing to bulky side groups. This increases the growth
time for a thin film even further over conventional techniques.
9.3 EQUIPMENT USED TO DEPOSIT THIN FILMS BY ALD

ALE growth can be accomplished using any conventional deposition
system. All that is needed are a few alterations in the way that the reactants
are delivered to the substrate (they must be supplied in an alternating
pattern). Since the growth is slow, and many reaction sequences are
involved, all ALD systems contain some provision for computer control of a
repetitive valving, shutter, or solution exchange system. Some typical ALD
systems which have been used successfully follow.
The apparatus shown at the top of Fig. 9.5 is a typical alteration of a
CVD-type system. The reactant feed lines are valved to control the delivery
of the reactants to the substrate. This type of reactor has been used to grow
ZnS under vacuum [24] using zinc chloride and hydrogen sulphide. It was
necessary to heat the feed lines and chamber walls so a reasonable zinc
chloride vapour pressure could be achieved and condensation of the vapour
did not occur.
Substrate
InniC-' Iitehambe,
ttt
Reactants To Pump
t
or
-
Atmosphere
Substrate
To Pumping
System
- I
I
6:+:-8
I
I I I
I I
I I
I I
Shutters
WWW
Effusion
Cells
Fig. 9.5 Schematic of CVALD (top) and MBALE (bottom) systems. (After Suntola
and Hyvarinen [24].)
This type of system has also been used at ambient pressures [15J using
dimethyl zinc and hydrogen sulphide as reactants. The reactant gases in this
case have a very high vapour pressure, so high in fact that the dimethyl zinc
source had to be cooled to 273 K to lower its vapour pressure. A disadvan-
tage of this CVD application is the extremely reactive nature of the
organometallic compound. Here, the authors passed the exhaust gases
through a high temperature furnace to crack any unreacted organometallic
compound.
The bottom diagram of Fig. 9.5 is a simplified schematic of a MBE system
that could be used for ALD. A series of computer-controlled shutters allow
the necessary alternating exposure of the reactants to the substrate to occur.
A unique application of the ALE technique has been made by Gregory et
al. [19]. They have successfully grown CdTe by electrochemical ALE
(ECALE). Although this material has not found use as a phosphor, this
variant on a well-known deposition process deserves to be covered here
since many other II-VI compounds of interest to the phosphor field have
been deposited using conventional electrodeposition. A chamber similar to
the one used by these authors is pictured in Fig. 9.6. To effect ALE, these
authors change solutions repetitively during the experiment. Each solution
only contains one element necessary for film deposition. The solution
exchange is accomplished by blowing spent solution out the bottom of the
chamber using compressed nitrogen. The nitrogen flow is then shut off and
the second solution is fed into the electrode chamber through pinholes with
the use of capillary action. When this solution has been reacted it is blown
out with nitrogen and the first solution is then drawn into the electrode
chamber. ALD is possible in this system because of the phenomenon of
Electrode 0.001 in
Gap
Fig. 9.6 Schematic of a deposition chamber used for ECALE. (After Gregory et al.
[19].)
Compounds deposited by atomic layer deposition 303
underpotential deposition. The potential necessary to deposit one element

onto a second element is under that necessary to deposit the element on
itself (i.e. the process of self-deposition is of higher energy). Because of this,
the deposition can be controlled, like ALE, to be self-limiting in each cycle
through properly chosen deposition potentials.
Another ALE technique based on solution growth is pictured on the
left-hand side of Fig. 9.7. It uses a mechanical dipping system to move
multiple substrates through chemical solutions with a rinse step between
each. Use is made of a chemical reaction between elements present in the
bath and the surface of the film dipped into the bath. This apparatus has
successfully been used to deposit ZnS, CdS and Znl-xCdxS by Nicolau et
al. [18] using NazS and metal chloride solutions.
Reactants
! '--+-+--1._ -1- Substrates
Fig. 9.7 Schematic of ALE deposition systems for solution (left) (after Nicolau
[26]) and chemical vapour (right) (after Bedair et al. [27]) deposition techniques.
The last type of apparatus (pictured on the right-hand side of Fig. 9.7) to be
discussed is a variant of the CVD-type system presented earlier. Instead of
valving the source gases the substrate is rotated in and out of two or more
reactant gas streams. This system has been used in the past to produce films of
GaAs using streams of trimethyl gallium and arsine as reactant materials [28].
9.4 COMPOUNDS DEPOSITED BY ATOMIC LAYER

DEPOSITION TECHNIQUES
There have been many compounds deposited by an ALE growth variation
of a standard deposition technique. Thin films of many II-VI compounds
have been grown including the zinc and cadmium chalcogenides and the
alkaline earth sulphides. GaAs has been the most widely studied III-V
compound studied to date but some of its alloys with aluminium have also
been grown. ALD techniques have also produced many high quality oxides
useful for luminescent displays. It is also not surprising that, given the
wealth of compounds grown, complete electroluminescent devices have been
grown inside ALE systems without exposure to atmosphere.
Given the number of compounds grown, an all-encompassing summary of
the characteristics of each compound grown cannot be made. Most of the
compounds useful as phosphors that have been grown by ALD will be
covered here. However, because of the vast number of starting materials
available, and the number of research groups working on thin film growth
by ALD, only a small fraction of their results can be covered owing to space
limitations. It is hoped that the following summary will at least give a
flavour for the diversity of films produced and the ingenuity that researchers
have used to adapt conventional thin film growth techniques to produce
ALD of a material.
A word of caution should be issued at this point. Many of the compounds
mentioned in this chapter are toxic to humans in even small quantities.
Some materials are also very reactive with air or water, creating enough heat
to start a fire or releasing noxious and toxic fumes. For these reasons, the
reader is advised that he or she should obtain the necessary information on
the safe handling and use of these compounds before any experimentation
begins.
9.4.1 ll-VI compounds
(a) ZnS
Historically ZnS was the first compound deposited by ALE. It has been one
of the most widely studied thin film materials ever since (of materials grown
by ALD, that is). Since this is a book on luminescent materials, it is
fortunate that so much is known about growing this compound by ALD
since ZnS-based thin film phosphors are one of the most promising
materials available for display of information.
ZnS has been grown from elemental as well as compound reactants.
Compound metal sources have included ZnCI 2 , Zn(CH 3 COOHh, dimethyl
and diethyl zinc, and some furan adducts of zinc to name a few. Usually, the
sulphur adduct of choice has been H 2 S.
The very first implementation of ALE was the growth of ZnS from
elemental Zn and S. The material was of superior crystal quality despite
being grown at relatively low temperatures (593-643 K) [1].
Growth from elements is simpler than from compounds since only the
desired elements are used in the growth avoiding sometimes complicated
and unwanted side reactions. However, because of equipment design, or if
lower processing temperatures are desired, it is not always possible to utilize

elemental reactants and more volatile species must be employed. The next
preparation route to be discussed involves a CV ALD of ZnS using zinc
chloride and hydrogen sulphide as reactants.
The reaction
ZnCl 2+ H2 S-+ZnS + 2HCI
has been shown to occur on the surface of a growing film during ALE
growth [29]. The exchange reaction (i.e. CI for S or Zn for H) was monitored
using Auger electron spectroscopy. The study established two important
facts. The first is that surface exchange reactions during the growth of a film
using these reactants agree with the previous theoretical model of section
9.2.3. The second is that the growing film contains no traces of chlorine.
Thus, as might be predicted, a very pure film results from the use of these
reactants in CVALD. The same cannot be said of films produced with these
compounds without the use of an ALD variation of growth where carbon
and chlorine incorporation frequently occur.
Perhaps the most amount of information on film quality has been
gathered on films produced from zinc acetate and hydrogen sulphide (e.g.
ref. 5 or ref. 30). Excellent quality films have been produced from these
reactant materials at growth temperatures of 563-603 K. The films are
normally polycrystalline exhibiting columnar grains of 40-80 nm width
(depending on film thickness). Both cubic and hexagonal phases coexist in
the films. The dislocation density of 10 10 cm- 2 is an order of magnitude
lower than that of films produced by electron beam evaporation.
Polycrystalline ZnS thin films have been grown using dimethyl zinc and
hydrogen sulphide [15]. The films were grown by ALCVD over a wide
temperature range (298-773 K). Close to ideal monolayer growth per cycle
was exhibited for films deposited by this technique. As is the case in most
depositions of ZnS by ALE, the films grew with a predominantly cubic (111)
or hexagonal (0002) orientation.
Finally, we come to a solution growth variant of ALE for the production
of ZnS thin films. This variant has been dubbed SILAR for 'successive ionic
layer adsorption and reaction' process by its founder Nicolau [26]. Con-
sidering the discussion in section 9.1.4, a new acronym probably should be
used. The term SALD for 'solution atomic layer deposition' is suggested.
Films produced at room temperature using this technique compare very
favourably with films produced by other means at temperature several
hundreds of kelvins higher [18]. The mean grain diameter for films pro-
duced by this technique is 30 nm. One surprising observation was that
successful epitaxial growth of the films on (111) InP and (001) GaAs
was accomplished. Epitaxial growth of ZnS is normally quite difficult and
usually requires temperatures of at least 760 K [31]. The refractive
index varied from 2.1 to 2.25. The largest drawback of films deposited
by this technique is the entrapped voids that are revealed in TEM images
[18].
(b) Doped ZnS thin films

ZnS in itself is not very useful as a phosphor. It must be doped with some
other element (e.g. manganese or terbium) before it becomes an appreciably
bright phosphor. The chloride of Mn has been used to produce doped films
with good luminescent characteristics and efficiency (2-3 1m W- 1 ) [32, 33].
The emission is of a yellow--orange colour, is rather broad, and is centred at
580nm.
The metal halides are suitable dopants when high substrate temperatures
are used but they cannot be used at lower substrate temperatures such as
when zinc acetate is used as a source material. One chelate which has been
found to complex Mn 2 + as well as Tm3+ and Tb3+ is 2,2,6,6-tetramethyl-
3,5-heptanedionate (thd). This ligand produces compounds which volatilize
completely above 493 K but do not decompose until above 773 K [34]. The
doping of ZnS with Ln(thdh (where Ln = a lanthanide series element)
compounds has been carried out by Tammenmaa et al. [35].
(c) ZnSe
Yoshikawa et al. [36] have grown ZnSe using MBALE using dimethyl zinc
and hydrogen selenide as reactants. Ideal monolayer per cycle growth was
restricted to a substrate temperature range of 473-548 K. They also noted a
dependence of the growth rate on the dimethyl zinc flow rate with less than
monolayer per cycle deposition at low flow rates but a limit of one
monolayer per cycle at higher flow rates. This indicates that not enough
time was allowed at the low flow rates for a monolayer of Zn to form.
Yoshikawa et al. [36] also produced films by CV ALE with the same
reactants as given above. In this case it was found that ideal ALE-type
growth could only be accomplished with a hydrogen selenide flow rate at
least twice that of the MBALE growth and if a much longer pulse duration
was used. This increased the time of deposition significantly.
Despite the fact that films prepared by each technique were grown in an
ALE mode there were still marked differences in the films. The MBALE
films were much smoother in appearance than the CV ALE films. The
MBALE films also showed superior photoluminescence spectra with more
intense and better-defined peaks.
These compounds have been grown by ALD from a solution growth

technique by Nicolau et al. [18]. The films were grown at room temperature
and showed surprisingly good crystallinity considering the low growth

temperature. The reported grain sizes are 6.5 nm for pure CdS decreasing
linearly with increasing zinc content to a minimum of 4 nm when a Zn:S
ratio of 0.75 is reached. At this point, the grain size rapidly increases to
58 nm at a Zn:S ratio of 0.95. The grain size then decreases to 30 nm for pure
ZnS. The behaviour of the grain size with alloying is interesting and might
be exploited in some way in the future.
As with the ZnS case epitaxial growth was achieved at room temperature
on the following crystal surfaces: (111) InP, (001) GaAs and (110) Ge. The
ability to maintain an epitaxial relationship at room temperature coupled
with the ease of deposition and the relative safety of the reactants used
should result in a much wider research effort using this technique in the
future.
(e) ZnO
Sanders and Kitai [37J have grown zinc oxide by ALCVD. The process used
dimethyl zinc and the trace oxygen or water contained in the nitrogen purge
gas. Slightly less than one monolayer growth per cycle was exhibited for
growth temperatures ranging from 473 to 573 K. Rutherford back scattering
analysis indicated that good stoichiometric films resulted from this tech-
nique. Despite a very fine-grained material (grain size < 300 A), reasonable
resistivities (0.005 n cm) were obtained. Growth was primarily in the (0002)
direction. These films would be suitable as transparent conductors in a
variety of thin film luminescent applications.
(f) SrS. BaS. and CaS

Doped alkaline earth sulphides have shown promise as efficient phosphors
for use in flat panel displays [38]. It is therefore not surprising that these
materials have also been grown by ALE. They were first produced using
CVALD by Tammenmaa et al. [16].
Simple volatile compounds of the alkaline and rare earth elements are
difficult to find. The chelate used for the production of thin films of these
materials was 2,2,6,6-tetramethyl-3,5-heptanedionate. Chelating the metal
ions with this molecule results in compounds that volatilize completely
above 493 K. The somewhat high vaporization temperature creates a slight
problem to the mechanics of growing films from these compounds. The
whole chamber and feed lines must be heated to prevent condensation of the
compounds. The sulphur source used was hydrogen sulphide so no prob-
lems occurred with condensation of this compound.
The films were grown at substrate temperatures of 723 K. At these
temperatures the reaction between the chela ted metal and the hydrogen
sulphide was remarkably complete. No traces of carbon or oxygen were
found in the films under Auger analysis despite the high content of these
elements in the whole complex. X-ray diffraction patterns of the films
revealed that the three films grew in three different orientations. CaS grew
primarily along the (220) direction, SrS grew (111) and BaS grew along (200).
These different growth directions may be a result of the apparently
complicated growth mechanism proposed for thin film formation. Mass
spectrometric studies show that these complexes maintain their integrity in
the gas phase, and thus molecules adsorbing on the surface of a growing film
are large. They are so large, in fact, that it is impossible to generate a
complete monolayer in one growth cycle. Only 0.3-0.4 A per cycle are
achieved. This means that, after the first cycle, the incoming molecules see
several nonidentical sites in which to adsorb. Thus, different growth direc-
tions will occur according to the most energetically favourable adsorption
si te filling.
9.4.2 111-V compounds

The luminescence properties of III-V materials have been exploited for a
long time. GaAs and either AIGaAs or InAs quantum well heterostructures
have been grown by ALE. These structures show promising photolumines-
cent properties. Infrared lasers based on these materials have also been
prepared. Dapkus et al. [39J have recently reviewed the atomic layer epitaxy
of GaAs by atmospheric chemical vapour deposition techniques. Tischler
and Bedair [40J have also recently reviewed the ALE of III-V compounds
by various methods.
(a) GaAs
Growth of GaAs by CV ALD techniques has rapidly expanded since the
original paper presented by Bedair et al. in 1985 [27]. Most researchers use
trimethyl gallium and arsine with a hydrogen carrier gas when depositing
films by this technique. The dominant chamber used is similar to the one
pictured in Fig. 9.5. The substrate is rotated alternately between two or
more gas streams. Using such a system ideal ALE growth is routinely
obtained for a wide range of growth temperatures and flow rates [40].
When diethyl gallium chloride is used as the group III source a wide
deposition window is also present. Ideal monolayer growth is maintained
through a temperature range of 673-873 K for over 3000 cycles. The system
also maintained ideal ALE growth over a wide range of exposure times [40].
(b) AIGaAs
A report on the growth of this compound by CV ALE using trimethyl
gallium, trimethyl aluminium and arsine as source materials has been
presented by Bedair [28]. Films were grown at substrate temperatures of

823-973 K. Ideal monolayer growth was found to depend not only on
temperature but also on the amount of group III compound the growing
film was exposed to. The point is made that in the growth of this compound
the self-limiting picture of ALE growth should be modified. Photolumines-
cence results obtained at 77 K show a much more intense and narrower
peak for films grown by CV ALE over films grown by conventional
MOCVD and the same reactant compounds.
9.4.3 Oxide dielectrics

Thin film oxides are a very important component of any electroluminescent
device. The main roles oxides play in any device are quite varied. Some are
used as transparent conductors; others are used as insulators. For use as an
insulator the oxide layer must not only be insulating to deter electrical
breakdown but it must also be thin so that most of the applied voltage drops
across the luminescent layer. This requires that the oxide layer be applied
evenly and be of uniform thickness. Otherwise, there would be uneven
voltage drops across the device. Atomic layer growth techniques are ideally
suited to preparing thin film oxides because of the excellent control over film
thickness demonstrated during film growth. Some of the more important
oxides (from a luminescence standpoint) prepared by ALE are described
below.
(a) Al 2 0 3
Thin film aluminium oxide was one of the first oxides prepared by ALD [5].
Thin films which were pinhole free suitable for use as ion barrier or dielectric
and passivation layers in electroluminescent devices were produced. The
reactants used were AICl 3 and H 2 0. Amorphous layers were produced
despite a substrate temperature of 720 K.
(b) In20rSn03 (ITO)

Lohja Corporation has also produced thin films of transparent ITO [24].
Resistivity of 1.6 x 10- 4 n cm with transmissions greater than 80% have
been demonstrated. Films were deposited over substrate areas of 1.5 cm by
2.5 cm.
(c) MgO
High quality MgO thin films deposited on (111) silicon substrates by
CV ALD are formed by a hydrolysis surface reaction of Mg(C 2 Hs hand
H 2 0. The growth orientation ofMgO changes from (111) to (100) when the
temperature of the silicon substrate changes from 600C to 900 0c. The
growth orientation difference of MgO grown by CV ALD is rationalized in
terms of the surface reaction in CVALD [17].
9.4.4 Electroluminescent devices

One of the goals of Lohja Corp. in developing ALE was to produce a better
electroluminescent device. An all-ALE device was made by Lohja as soon as
the technology was to the point where insulators, transparent conductors
and a phosphor layer could be made. The first device made consisted of the
following structure [24]: soda lime glass (to serve as a viewing window and
to provide mechanical support for the structure), a transparent and conduc-
tive ITO electrode, a dielectric of Al2 0 3 , the phosphor layer (ZnS:Mn),
another Ab 0 3 layer, and finally a back set of ITO electrodes. The successful
demonstration of an all-ALE device showed that structures built by this
technique could be successfully generated in an industrial setting.
9.5 SUMMARY
The technique of ALE has been applied successfully to the deposition of
both single-crystal and polycrystalline thin films of II-VI and III-V
semiconductors as well as a few oxide dielectrics. The technique is easily
adapted to most conventional deposition methods and has found wide-
spread use for the production of thin films.
Because of the nature of the ALE technique, large substrate areas can be
covered homogeneously without the need for substrate rotation. Perfect
layer-by-layer growth has been realized by this technique, making it an ideal
method for producing superlattice structures.
Perhaps the most severe restriction of the technique is its low growth rate
when compared with other methods of film deposition. The high crystal
quality of films resulting from this growth method can more than compen-
sate for the longer deposition times. In fact, using this technique, structures
that cannot be grown conventionally are possible. Atomic layer deposition
techniques have become, and probably always will be, a valuable tool for
thin film phosphor growth.
REFERENCES
1. Suntoia, T. and Antson, J. (1977) US Patent 4,058,430.
2. Suntoia, T., Antson, M.J., Pakkaia, A. and Lindfors, S. (1980) SID 80 Digest,
p.108.
3. Suntoia, T., Pakkaia, A. and Lindfors, S. (1983) US Patent 4,413,022.
4. Herman, M.A. (1991) Vacuum, 42(1-2), 61-66.
5. Goodman, C.H. and Pessa, M.V. (1986) J. Appl. Phys., 60(3), R65-R81.
References 311
6. Peng, Y. (1990) Wuli, 19(12), 733-738.
7. Suntola, T. (1989) Acta Poly tech. Scand., Ser. Electr. Eng., 64, 242.
8. Niinisto, L. (ed.) (1990) Acta Poly tech. Scand., Chern. Metall. Ser., 195.
9. Wantanabe, H., Mizutani, T. and Usui, A. (1990) In Very High Speed Integrated
Circuits: Heterostructure, (ed. T. Ikama), Semiconductors and Semimetals, Vol. 30,
Academic Press, San Diego, CA, pp. 1-52.
10. Sun to la, T. and Simpson, M. (eds) (1990) Atomic Layer Epitaxy, Blackie,
Glasgow.
11. Tanninen, v.P., Oikkonen, M. and Tuomi, T. (1981) Phys. Status Solidi A, 67(2),
573-583.
12. Lahtinen, lA., Lu, A. and Tuomi, T. (1985) J. Appl. Phys., 58(5), 1851-1853.
13. Wu, Y., Toyoda, T. and Kawakami, Y. (1990) Jpn. J. Appl. Phys., Part 2, 29(5),
L727-L730.
14. Leskela, M. (1990) Acta Poly tech. Scand., Chern. Metall. Ser., 195, 67-80.
15. Hunter, A. and Kitai, A.H. (1988) J. Cryst. Growth, 91,111-118.
16. Tammenmaa, M., Antson, H., Asplund, M., Hiltunen, L., Leskela, M. and
Niinisto, L. (1987) J. Cryst. Growth, 84, 151-154.
17. Huang, R. and Kitai, A.H. (1992) App/. Phys. Lett., 61, 1450-2.
18. Nicolau, Y.F., Dupuy, M. and Brunei, M. (1990) J. Electrochem. Soc., 137(9),
2915-2924.
19. Gregory, B.W., Suggs, D.W. and Stickney, lL. (1991) J. Electrochem. Soc., 138(5),
1279-1284.
20. Nishizawa, J., Abbe, H. and Kurabayashi, T. (1985) J. Electrochem. Soc., 132(5),
1197-1200.
21. Usui, A. and Wantanabe, H. (1987) Proc. Electrochem. Soc., 87(8), 766-775.
22. Kobayashi, N., Makimot, T., Yamauchi, Y. and Horikoshi, Y. (1990) Acta
Poly tech. Scand., Chern, Metall. Ser., 195, 139-145.
23. Gerard, J.M., Marzin, lY. and Jusserand, B. (1991) J. Cryst. Growth, 111,
205-209.
24. Suntola, T. and Hyvarinen, 1 (1985) Annu. Rev. Mater. Sci., 15, 177-195.
25. Pakkanen, T.A., Nevalainen, v., Lindblad, M. and Makkonen, P. (1987) Surf
Sci., 188(3), 456-474.
26. Nicolau, Y.F. (1985) Appl. Surf Sci., 22-23, 1061.
27. Bedair, S.M., Tischler, M.A., Katsuyama, T. and EI-Masry, N.A. (1985) App/.
Phys. Lett., 47(1), 51-53.
28. Bedair, S.M. (1990) Acta Poly tech. Scand., Chern. Techno/. Metall. Ser., 195,
17-37.
29. Pes sa, M., Makela, R. and Suntola, T. (1981) Appl. Phys. Lett., 38(3), 131-132.
30. Oikkonen, M., Blomberg, M., Tuomi, T. and Tammenmaa, M. (1985) Thin Solid
Films, 124, 317-321.
31. Dryburgh, P.M. (1988) J. Cryst. Growth, 87(4), 397.
32. Tornqvist, R., Antson, l, Skarp, J. and Tanninen, V.P. (1983) IEEE Trans.
Electron Devices, 30(5), 468-471.
33. Sutela, T. (1984) Displays, 5(2), 73-78.
34. Tammenmaa, M., Koskinen, T., Hiltunen, L., Leskela, M. and Niinisto, L. (1984)
VTT Symp. Ser., 54, 18-21.
35. Tammenmaa, M., Leskela, M., Niinisto, L. and Koskinen, T. (1986) J. Less-
Common Met., 126, 209-214.
36. Yoshikawa, A., Okamoto, T., Yasuda, H., Yamaga, S. and Kasai, H. (1989) J.
Cryst. Growth, 101, 86-90. .
37. Sanders, B.W. and Kitai, A.H. (1991) Proc. SPIE-International Society of Optical
Engineering, 1398,81-87.
38. Ohnishi, H. (1989) Annu. Rev. Mater. Sci., 19, 83-101.
39. Dapkus, P.D., Maa, B.Y., Chen, Q., Jeong, W.G. and DenBaars, S.P. (1990) Acta
Poly tech. Scand., Chern. Technol. Metall. Ser., 195, 39-54.
40. Tischler, M.A. and Bedair, S.M. (1990) In Atomic Layer Epitaxy (eds T. Suntola
and M. Simpson), Blackie, Glasgow, pp. 110--154.
41. Suntola, T. (1989) Mater. Sci. Rep., 4, 261-312.
10
Lamp phosphors
T.E. Peters, R.G. Pappalardo and R.B. Hunt, Jr
10.1 PHOSPHORS FOR LOW-PRESSURE Hg PLASMAS

The operation of luminescent materials as transducers for the conversion of
electrical energy into light (in electroluminescence and in CRT displays), or
as energy converters (scatterers) of electromagnetic (EM) energy into EM
energy of different wavelength is treated in other chapters of the present
book. In what follows we shall focus on luminescent materials in powder
form (phosphors) as wavelength converters of EM radiation produced in
electrical discharge lamps.
Of paramount practical importance are phosphors utilized as converters
of ultraviolet (UV) radiation into visible light in fluorescent sources, namely
in low-pressure plasmas of rare gases (typically argon) and mercury. For a
comprehensive treatment of the properties of such plasmas the reader is
referred to Waymouth's book [1]. The striking property of these optimized
Hg-rare gas plasmas ('" 2.5 Torr Ar and", 10 mTorr Hg pressures) is that
the applied electrical energy is converted almost exclusively into a narrow
UV line, at 253.7 nm; this conversion is extremely efficient, with a ",65%
yield in the plasma positive column. In fluorescent lamps this rich harvest
of UV radiation is converted, by means of suitable phosphors, into visible
light for general illumination purposes.
The generation of radiation in Hg-containing plasmas derives mainly
from electron impact excitation of neutral mercury. The relevant lowest-
energy levels of neutral Hg are shown in Fig. 10.1. The 185 nm line is the
typical resonance line within the singlet system. The 254 nm line is actually
an intersystem transition (triplet to singlet) that is made possible by the
strong spin-orbit interaction present in heavy elements. The remaining UV
and visible lines emitted by Hg are transitions connecting excited electronic
levels.
Fluorescent lamps also find extensive use in specialized technological
areas, where the phosphor requirements differ from those for general
313
314 Lamp phosphors
Ionization
10
> Q)
c:
>-
....
C)
Q) 5
c:
Q)
Q)
>
Q) 185 nm
....J
o~--~~-------------------------
6e So)
Fig. 10.1 Lowest electronic energy levels for neutral mercury.
illumination. For instance, most of the fluorescent lamps used in xerography

have a green output. As another example, UV-to-UV down-conversion is
effected by phosphors used in lamps for the suntanning market, or simply
to convert short-UV radiation into long-wavelength UV radiation in 'black
lights'. These topics are discussed in section 10.1.9.
Fluorescent lighting technology is almost 50 years old and presents many
of the typical traits of mature technologies. The mass market of fluorescent
lighting is dominated world wide by a few manufacturers, and market shares
are acquired, or lost, on the basis of rather narrow performance differentials.
The latter pertain to efficacy (output luminosity per unit of input electrical
power, color rendition, and cost (materials and manufacturing). In most
cases these performance advantages hinge on the use of appropriate phos-
phor materials. Prior to discussing the latter, some basic ideas on luminosity
and color rendition - as relevant to phosphor selection - will be reviewed
next.
10.1.1 Basic concepts in lamp photometry and colorimetry

The output from a light source is characterized by lamp engineers by means
of its spectral power distibution (SPD), namely by the radiant power per
nanometer interval, ~iWi(A.i) AA.. Of course, a human observer will experi-
Phosphors for low-pressure Hg plasmas 315
ence no visual stimulus if the SPD contains only UV wavelengths, or only

infrared wavelengths. To extract the luminosity content from the SPD, one
has to convolute the SPD with the human eye spectral response, namely
with the luminous efficiency function V(},;) , with peak at 555 nm.
The observer is also accustomed to gauging color either under solar
illumination or under illumination from an incandescent source, such as a
hot tungsten filament. The spectral output from fluorescent lamps is
definitely unlike that of a black-body radiator, so the question arises as
to its ability to approximate the color rendition of (quasi-) black-body
radiators.
Color perception and color matching experiments can be rationalized in
terms of three different response functions associated with the human
perception of color. These response functions are the x)., y)., z). color
matching functions described [2J in the CIE Colorimetric System of 1931.
(F or convenience's sake, the y). is taken to be the same as the relative
luminous efficiency function V)..) It follows that the color information
contained in the source SPD can be extracted by convoluting the SPD with
these three color matching functions. These convolutions (scalar products in
Ai space) give three scalar quantities that can be viewed as the color
coordinates of the SPD. The color coordinates are then re-normalized, and
so one particular SPD can be described in the so-called color plane simply
by the two color coordinates x and y (the third coordinate being the
complement to unity). By way of illustration, in Fig. 10.2 the color
coordinates for red-, green-, and blue-emitting phosphors, to be discussed
later, are plotted as R, G, B in the CIE color plane, and are contained within
the color locus whose perimeter is defined by the 'pure' colors.
Of particular importance in the color plane are the color points of
black-body radiators at different temperatures that span the so-called
Planckian locus or black-body locus (BBL). The fluorescent-lamp engineer
strives to produce light sources whose color coordinates will fall on, or in
the proximity of, the Planck ian locus.
We have seen that to one SPD there corresponds in the color plane a
specific color point. The converse is not true. Taking for instance the
Planckian locus, a manifold of different SPDs can be associated with a
specific color point on it (or, equivalently, with the color temperature of a
corresponding black-body) as illustrated in Fig. 10.3, where dramatically
different SPDs are displayed for lamps having identical color points
(4100 K) on the BBL. These SPDs, although sharing a common color point,
will differ in their color rendering ability. Criteria for quantifying the latter
originated in the color industry in general, and in the fluorescent-lighting
industry in particular.
The derivation of the color rendering ability of a particular source hinges
on considering the color points of the secondary SPDs obtained by shining
the source under investigation onto eight (later expanded to fourteen)
316 Lamp phosphors
0.9
.~20
530
0.8
510 540
550
0.7
560
0.6
500 570
G
>- 580
t 0.5
590
0.4
4000 "'.
tJ.- ~ 600-
8
~,3000-2P90 T~610
~490 10000 ~ ,~, 15~0~7"" 620
r 5000 1000
0.3
~q~o ~~ 6.50
I~
~~ 8000-
Absolute
temperature 670
(K)
0.2 '0
480
0.1 B"
470
460'" 450
~400
0.1 0.2 0.3 0.4 0.5 0.6 0.7

..... x
Fig. 10.2 CIE chromaticity diagram showing the Planckian, or black body, locus
(BBL), and the color points for the red- (R), green- (G) and blue- (B) emitting
phosphors, Y203:Eu, (Ce,Tb)MgAl l1 0'9, and BaMg2Al'6027:Eu respectively.
industry standards with specific reflectance properties. The procedure is then

repeated for the black-body radiator with (approximately) the same color
temperature as the source under characterization. The corresponding pairs
of color points will show some separation, and the color rendition will more
closely approach that of an incandescent source, as these color separations
become smaller on average. (There are additional refinements to the
derivation, only of interest to the specialist, that involve performing trans-
formations on the original color plane, so as to compute the color point
separations in color spaces with more uniform metrics.)
The point to bear in mind is that, once the source SPD is available, one
can mathematically derive its luminosity content, its color coordinates and
the source color rendering, the latter expressed normally as a single
figure-of-merit quantity, the eRI, or general color rendering index. By
0.6
... 0.5
....
CI)
e0
CI)
0.4
a
...=
CI)
0.3
Q.
....
....
<II
0.2
as
~ 1\ : .,"''\....
J\
0.1
-j\ --'
~~ ~,
I
o r-..J ~
350 400 450 500 550 600 650 700 750
Wavelength (nanometers)
Fig. 10.3 SPDs for two fluorescent lamps with identical color points on the BBL,
at 4100K.
definition a black-body radiator has a CRI of 100. The CRI of fluorescent

sources ranges from '" 50 to the middle 90s, as we shall see.
The representation of the SPD in the color plane is particularly useful in
combining SPDs so as to achieve a target color point, typically a point on
the Planck ian locus. Given the nature of the transformation from SPD to
color coordinates, a target color point can be obtained from the combina-
tion of two SPDs only if the target color point is contained within the
segment joining the color points of the original SPDs. By induction, this
simple property is nowadays exploited systematically in mathematically
combining three or more SPDs so as to achieve a predetermined white field
or, equivalently, a preselected color temperature.
10.1.2. Highlights in the development of fluorescent-lamp phosphors

The topic of fluorescent-lamp phosphors was comprehensively treated in a
recent monograph by Butler [3]. Shorter review articles treat more recent
developments [4-7]. Here we briefly review what we consider the most
significant phosphor developments affecting the past history of the fluor-
escent-lighting industry.
The early fluorescent-lamp phosphors were natural fluorescing minerals
(for instance willemite, Mn-activated zinc orthosilicate) that were ground to
a powder and empirically blended together so as to obtain an approximately
white field from the fluorescent lamp. (In the history of fluorescent lighting,
empirical trial-and-error solutions to photometric and colorimetric prob-
lems have tended to precede mathematically based identification and
modeling of the relevant physical processes.)
A major turning point in the industry followed the development in the
318 Lamp phosphors
1940s of Sb-Mn co-activated halophosphate phosphors. In a single material

the blue emission from the Sb3+ activator and the orangy emission from the
Mn2+ co-activator can be adjusted, so that, in combination with the visible
emission from the Hg plasma, they produce a white field corresponding to a
wide range of color temperatures, as will be noted in section 10.1.5. Lamp
brightness was higher, with color rendering indices in the high 50s or low
60s. Phosphor blends were still employed in cases where high values of the
color rendering (high CRI) were needed, even if this meant sacrificing output
brightness.
Another major re-orientation occurred in the industry in the early 1970s.
For more than a decade laser material specialists had been investigating the
emission characteristics of rare earth ions, with a view to developing rare
earth analogs of the ruby laser. As a distant rippling effect of these research
efforts, the early 1970s saw the introduction in new fluorescent-lamp types
('tricolor' or 'triphosphor' lamps) of blends of three phosphors, each ac-
tivated with rare earth ions: a blue-emitting phosphor containing Eu2+, a
green-emitting phosphor containing the Ce3+ - Tb3+ pair, and a red-emit-
ting phosphor containing Eu 3+.
The advantage of the new 'tricolor' blend was that now one could have
the best of both worlds, namely not only higher brightness (~3400 1m at
100 h in a 40 W, 1.5 in diameter lamp), but also markedly higher color
rendering (middle 80s). The one serious objection was, and still is, the high
cost of all those rare-earth-activated materials.
One recent development - mostly affecting the European market, which
is more sensitive to color rendition than the US market - has been the
development of novel blends for high CRI lamps, with CRIs in the middle
90s, but rather low brightness. The interesting feature for the phosphor
technologist is the use in these blends of borate phosphors with high Gd
content. Gd acts as a sensitization intermediate, by funneling the UV
excitation energy, typically captured by Ce 3 +, to the emitting centers,
namely Tb3+ or Mn2+.
10.1.3 Brief survey of lamp phosphors (hosts and activators)

The phosphors used in fluorescent lamps are almost exclusively oxides, with
a strong predominance of phosphates. Sulphide and halide materials are not
at all compatible with operation in an Hg plasma. The anionic species in
many efficient phosphors are frequently mixed oxides and halides. The
empirical genesis of these halo oxides stems from the well-established
practice in oxide powder synthesis of using halides as fluxes-mineralizers.
Historically the manufacturing bias towards phosphates is due to the
relatively low temperatures required for their synthesis. However, refractory
oxides, such as yttrium oxide and complex aluminates (hexa-aluminates and
garnets), are increasing in importance as phosphor hosts. Another important
Phosphors for low-pressure H g plasmas 319
class of refractory oxides, namely the silicates, has an unfortunate chemical
affinity for Hg, conducive to poor lamp performance.
A common emitter in phosphors is divalent manganese. Depending on its
coordination, the Mn 2+ emission can be tuned from the blue-green
('" 510 nm) to the far red (650 nm). Its long decay time, in the millisecond
range, is an advantage in reducing pulsation in lamp output (flickering).
Mn 2+ is generally a poor absorber at 254 nm, and so its emission follows
excitation energy transfer from a co-activator, that absorbs efficiently at
254nm, and that plays the role of sensitizer, typically Sb3+, but also Ce3+
and Sn2+ [8].
In a few hosts where Mn 2+ is tetrahedrally coordinated (e.g. willemite,
Zn2 Si0 4 ) the charge transfer transitions of Mn 2+ are at relatively low
energy and therefore can be used to absorb 254 nm radiation; in such cases
no sensitizer is needed. Other hosts can directly absorb the 254 nm radi-
ation, and the Mn 2+ centers then intercept and trap the excitation energy as
it migrates through the host lattice. Such hosts generally contain Cd,
nowadays an unacceptable constituent.
Another versatile ion whose emission can be shifted from the UV to the
far red is Sn 2+. Special attention is required in processing Sn 2+ -activated
materials, since Sn 2+ is easily oxidized during lamp manufacturing (see
section 10.1.7).
There is generally no shortage of efficient, blue-emitting phosphors. It is
not difficult to see why. The condensed phase cations (and complex anions)
with closed-shell configurations have a first excited state at a few electron-
volts over the ground state. If the excited state relaxation is not too
pronounced, the radiative channel back to the ground state will involve
emission of either long UV or blue light. More pronounced Stokes shifts of
the emission to longer wavelengths are generally accompanied by non-
radiative de-activation in the form of potential curve crossing to the ground
state, and hence by low emission efficiency.
At present high-performance lamps employ divalent Eu-activated phos-
phors as the blue emitter, as will be seen on section 10.1.6.
10.1.4 Phosphor quantum yield and 'lamp' quantum yield

The intrinsic efficiency of the UV-to-visible conversion is simply defined by
the ratio of emitted photons to absorbed photons (quantum yield for
emission). The measurements required to arrive at this simple ratio are
difficult to carry out in practice in a fashion consistent from observer to
observer. A well-known luminescence standard is sodium salicylate, and so
the measurement of its quantum yield is central for a ranking of quantum
yields in various classes of luminescent materials. However, its quantum
yield for 254 nm excitation has been quoted in various reports [9] to vary
between 25% and 99%. For another fluorescence standard, i.e. lumogen,
320 Lamp phosphors
Bril and de Jager-Veenis [9J only report conservatively that the quantum
yield is bracketed between 40% and 60%. With this proviso in mind it can
only be stated that the quantum yield for 254 nm excitation in the traditional
lamp phosphors is in the 60-70% range. In the case of the novel rare-
earth-activated phosphors the quantum efficiency moves into the 80-90%
range, and finally for Y203: Eu the quantum efficiency may reach '" 95%.
In the context of lamp manufacturing the phosphor quantum yield is a
useful, but not a final, indicator of phosphor performance in lamps. As
discussed in later sections, both the processing of the phosphor during lamp
fabrication and exposure of the phosphor to the plasma during lamp
operation may seriously lower its conversion efficiency. From a device
standpoint the basic question in comparing the lamp performance for
phosphors emitting in different spectral regions is, having selected a lamp
type, namely a specific plasma, how many (visible) photons does one obtain
from the lamp by using a specific phosphor, or a specific blend of phos-
phors? The lamp output is normally available in the form of the SPD for
100 h of operation, for reasons explained in section 10.1.8. It is a simple
matter to convert this information in a photon-per-nanometer format, and
hence to arrive at the total photon content of the SPD. One could then
derive an absolute value of the 'lamp' quantum yield for the phosphor, if the
number of UV photons emitted from the plasma was also reliably known.
This unfortunately is not the case at present. Still, it will be a useful exercise
to rank lamp performance according to relative lamp quantum yield, as is
done in Table 10.1.
In this table Y203: Eu stands out for its superlative performance, followed
by the Ce-Tb-activated hex a-aluminate. These phosphors will be discussed
Table 10.1 Relative lamp quantum yield of commercial phosphors, as derived from
the SPDs of 40 W lamps at 100 h of operation (SPD from 380 nm to 760 nm)
Phosphor type Total power Relative lamp

(W) in SPD quantum yield
Y20 3: Eu 10.21 1.00

(Ce,Tb)MgAI 11 0 19 9.99 0.87
Cas(P04h(F,Cl):Sb,Mn (3000 K) 8.78 0.78
Strontium aluminate:Eu 2 + 9.89 0.78
BaMg2AI16027:Eu2+ 10.65 0.78
(Sr,Mgh(P0 4h:Sn 2+ 8.12 0.77
Ba2P207:Ti 8.55 0.69
Cas(PO 4h(F,Cl): Sb 9.20 0.68
Srs(P04h(F,Cl):Sb,Mn 7.50 0.64
MgGa204:Mn 7.45 0.64
CaSi0 3 :Pb,Mn 6.03 0.54
Tricolor blend for 3000 K 9.78 0.89
in detail in section 10.1.6. The halophosphate phosphor (third table entry)
is roughly 20% less efficient than Y203:Eu, but is still on a par with some
of the modern aluminates activated with divalent Eu. The Sn-activated
material is comparable in efficiency with the halophosphate phosphors.
Among the remaining phosphors red-emitting CaSi0 3:Pb;Mn is half as
efficient as Y 203:Eu.
10.1.5 Halophosphate lamp phosphors

We come now to discuss the most widely utilized phosphor type in the
industry, the so-called halophosphate phosphors. Since its introduction some
45 years ago [10, 11J, the quantity of Ca halophosphate consumed by the
fluorescent lighting industry has increased annually, reaching world-wide an
estimated production level of 10 OOOt in 1990 [12]. The halo phosphates are
structurally related to the mineral apatite, CalO(P04)6F 2 (calcium
fluoroapatite, usually abbreviated as CaF AP), which accommodates to a
wide variety of compositional modifications without changes in crystal
structure. The crystal chemistry of the apatite compounds has been extensive-
ly studied and systematized [13J and will not receive further attention here.
We will focus now on the narrower class of materials that have the widest
application in fluorescent lighting, namely the Sb-Mn co-activated calcium
fluorochloroapatites Ca5(P04h(F, CI):Sb3+ ,Mn2+, also designated by the
CaFCIAP:Sb,Mn acronym. By careful control of the chemical composition
of this phosphor type, the fluorescent lamps will have color points in close
proximity of the Planck ian locus, with color temperatures ranging from
2500 K to 7500 K.
In Fig. 10.4 we present the spectral power distribution of a 'cool white'
fluorescent lamp with a color temperature of 4300 K. The sharp narrow
0.2
......
CII
CII
0.15
S
0
ar:: A
...
V\
0.1
CII
g,
......
In
1\
0.05
~
.."
~v- J ~
o
300 400 500 600 700 800
Fig. 10.4 SPD of a Sylvania 'cool white' fluorescent lamp (40 W 1.5 in diameter).
322 Lamp phosphors
features are contributed by the Hg discharge, while the broad continuum

extending from 400 to 700 nm is the emission from the 'cool white' phosphor.
As already mentioned, the UV radiation from the Hg discharge, mainly at
254 nm, is absorbed by the Sb 3 + ion and is re-emitted as blue radiation
peaking at -477 nm. In view of the spectral overlap of this emission with
absorptions by Mn 2 +, excitation energy can be transferred to Mn 2 +,
resulting in an orangy emission peaking at 585 nm. If the Sb 3 + is omitted
from the phosphor, the Mn 2 + emission is also absent because Mn 2 + itself
does not appreciably absorb in the UV region. Thus Sb 3 + serves both as a
primary activator (emitter) and as a sensitizer for the Mn 2 + emission. The
concentration of Sb 3 + in the phosphor appears to have little effect on the
location of peak position of the blue emission band. On the contrary, the
peak position of the Mn 2 + band is influenced by both the Mn 2 + concentra-
tion and by the CI:F ratio in the phosphor. Generally speaking, increases in
either the chlorine or the manganese concentrations cause the peak position
of the Mn2+ emission band to shift to longer wavelength.
(a) Phosphor synthesis

Although the basis of the industrial-scale halophosphate synthesis was well
established decades ago, process optimization is still continuing because of
its importance for lamp output and lamp cost, and hence for market shares.
The composition of a typical reagant blend used in the synthesis of the
cool white phosphor is presented in Table 10.2. The phosphor raw materials
listed in this table were originally selected for their physical and chemical
stability, and for their availability in a sufficiently pure form at reasonable
cost. While off-the-shelf, reagent-grade chemicals can be used to prepare
CaFCIAP:Sb,Mn, the manufacture of a phosphor with reproducible effi-
ciency and chromaticity requires careful control of the purity, stoichiometry,
and particle size of the starting raw materials.
Table 10.2 Cool white phosphor formulation
Reagent Amount (mol)
CaHP0 4 3.0
CaC0 3 1.205
CaF2 0.425
NH4Ci 0.180
MnC0 3 0.089
Sb 2 0 3 0.038
CdO 0.06
Total Ca 4.630
Total Ca+Cd+Mn+Sb 4.855
Total F+Cl 1.035
Making the reasonable assumption that Cd, Sb, and Mn substitute for Ca
in the apatite lattice, and that Cl substitutes for F, it would appear that the
reagent blend described in Table 10.2 could not yield a stoichiometric
halophosphate. In particular the blend is cation deficient and has an excess
of halide.
In effect, early workers found that phosphors prepared from blends
having precisely correct cation:phosphate and halide:phosphate mole ratios
gave rise to poorly luminescent materials.
This provided the impetus for studies that employed differential thermal
analysis (DTA), thermogravimetric analysis (TGA), electron paramagnetic
resonance (EPR), X-ray diffraction, optical spectroscopy, and total chemical
analysis to follow all of the processes occurring during phosphor formation
[14-19]. It was possible to identify precisely the temperatures at which the
raw materials thermally decomposed or reacted, what transient phases were
formed, which phases acted as carriers for Sb and Mn, how the furnace
atmosphere modified the processes, at what point apatite was obtained, and
what undesirable phases might remain in the product.
Some of the more pertinent results of these studies are worthy of further
comment. The CaF AP can be prepared at temperatures as low as 700-
800C, but optimum crystallinity and brightness requires temperatures of
1000-1200C. The synthesis of CaFAP is a dynamic process, wherein the
raw materials undergo thermal decomposition or react to form new phases
that are themselves later consumed in forming the phosphor. Any change in
the raw materials, heating rate, furnace atmosphere, batch size, powder bed
depth or furnace loading will modify the characteristics of the phosphor
obtained. To maintain chromaticity and brightness within specifications in
industrial-scale syntheses, formulae have been developed that actually adjust
blend composition to compensate for raw-material changes [20].
It was also learned that a substantial part of the antimony, typically
about half of the added amount, is lost during the firing process, together
with some of the halide and cadmium. The antimony is probably volatil-
ized in the form of Sb 20 3 or SbCh. The use of cation-deficient blends
helps to minimize contamination of the phosphor by minor phases detri-
mental to phosphor brightness. The compounds Ca(Sb0 3h and ~
Ca3(P04h are particularly objectionable, while Ca2P207 appears to be
relatively harmless. Ca(Sb0 3h robs the apatite phase of needed antimony,
thereby reducing its UV absorption, and at the same time introduces a
parasitic UV absorption of its own. The ~-Ca3(P04h:Mn phase reduces
brightness because of a visible absorption (pink body color) caused by
oxidation of Mn [15]. Formation of the orthophosphate is influenced by
the actual reagent mix, and is more likely to be observed in fluoroch-
loroapatite phosphors that have been exposed to air or moisture during the
high-temperature synthesis. In effect, an inert atmosphere is needed for the
halophosphate synthesis, since oxidation of even a small fraction of the
324 Lamp phosphors
manganese has a marked deleterious effect on phosphor brightness and

lumen maintenance in lamps.
Post-synthesis processing
On completion of the CaFClAP:Sb,Mn firing, the phosphor is obtained in the
form of a sintered cake. This cake is reduced to a powder and incorporated
into a paint-like suspension capable of producing a smooth, uniform, adherent
coating on the interior of the fluorescent-lamp envelope. Butler [3] provides a
detailed description of the early attempts at lamp coating. Initially, the desired
coating characteristics were achieved by ball milling the phosphor until
particles greater than 30l1m were eliminated. However, ball milling also
increased the overall particle size distribution of the phosphor, creating many
'fines' which lowered lamp brightness, because of their reduced UV absorption
and decreased visible transmission. A phosphor with a narrow particle size
distribution was required, with neither the large particles that reduced coating
quality and adherence nor the fine particles that decreased brightness.
Milling the phosphor has other disadvantages. The particles are fractured,
exposing Mn and Sb to oxidation, reduction, and/or interaction with the Hg
plasma, all of which is harmful to lamp brightness and maintenance. Early
studies, though, showed that a sequential acid-base wash could remove
these reactive ions [21].
The major lamp manufacturers have devised their own methods for
producing halophosphates with the particle size distribution required to
provide high-quality phosphor layers of optimum brightness and lumen
maintenance. These phosphor finishing processes have been tailored to
specific phosphor suspension chemistries and lamp fabrication methods, and
do not produce equivalent results when partially transferred to a different
lamp manufacturing environment.
Although the details of these methods are proprietary, some general
informaton is available. It is now well known that the particle size and
morphology of the CaHP0 4 are reflected in the finished phosphor. Thus,
the first step in the control of particle size distribution of the phosphor is to
control the particle size of this critical component.
Various chemical washing procedures have also been developed to
eliminate surface Mn, Sb and minor non-apatitic phases from the phosphor.
The more successful of these use chelating agents such as EDTA [22] and
DTPA [23] to bind the soluble ions so they are not re-deposited onto the
phosphor. These chemical washes are conveniently sequenced with wet
milling and elutriation or decantation techniques for particle size control.
(b) Activator site preference

Still in the context of improving the efficiency and performance of the Ca
halophosphate phosphors, a considerable effort has been devoted within the
lighting industry toward the basic question of identifying the crystal sites
occupied by the activators in the apatite host. Starting from the simplest
compound, Ca fluoroapatite crystallizes in the hexagonal space group C~h
(P6 3 /m) with two molecules per unit cell and, as shown in Fig. 10.5, with two
different Ca2+ sites. The Ca(I) site has C 3 point symmetry with each calcium
ion having six oxygen neighbors at the vertices of a distorted trigonal prism.
Approximately 40% of the Ca2+ sites are of this type. The Ca(II) site has
C 1h symmetry, with the calcium ions at the corners of equilateral triangles
centered around a fluoride ion.
(a) (b)
Fig. 10.5 (a) Diagram of the fluoroapatite structure. Each large circle represents
the group of atoms shown below. These circles form planes perpendicular to the
c-axis. The Ca(I) ions are below the plane of the large circles and the dashed
rhombus outlines a unit cell with F- ions at the vertices. (b) Diagram illustrating
the orientation of the halogen chain relative to the Ca(IJ) sites in the apatite
structure [24]. Reprinted by permission of the publisher, The Electrochemical
Society, Inc.
The optical properties of CaFAP:Sb,Mn single crystals were investigated

in an attempt to identify the sites occupied by the manganese ion [25].
Similarly, in fluoroapatite single crystals Ryan et al. [26J identified, by
means of ESR techniques, Mn2+ ions in four different environments,
designated as Mn(I), Mn(II), Mn(Im) and Mn(IIm). The former two were
assigned to manganese ions occupying ordinary Ca(I) and Ca(II) sites, while
the latter were ascribed to manganese ions in Ca(I) and Ca(II) sites modified
by nearby oxygen vacancies. Correlation of the ESR and optical data from
single crystals showed that the Mn(I) centres were responsible for almost all
of the observed fluorescence. Emission from the Mn(II) centers was observ-
able only from crystals with higher Mn concentrations and the emission
differed only slightly from that of the Mn(I) centers. Emission from the
Mn(IIm) center was thermally quenched at 300 K, and that from Mn(Im)
could not be observed for other reasons. In more recent work [27J, neutron
powder diffraction techniques were employed to refine the crystal structure
of Mn-doped fluoroapatite, and to confirm that Mn goes almost exclusively
into a slightly modified Ca(I) site.
326 Lamp phosphors
The site (or sites) occupied by Sb3+ in the apatite host is still being
debated. According to the most widely accepted model [28], antimony
enters the Ca(II) site and is charge compensated by an oxygen replacing an
adjacent halogen. The Sb-O pairing explains the insensitivity of the Sb 3 +
emission to changes in the halogen (Cl or F). Davis et al. [29] confirmed the
model by demonstrating that Sb 3 + emission shifts to 400 nm when the Sb-O
pairing is precluded, and that this high-energy emission can be converted
into the normal band (Amax '" 48 nm) by re-establishing the Sb-O pairing.
Soules et al. [30] subsequently calculated the energy levels for the Sb3+
centers in fluorapatite by means of a molecular orbital model and compared
their predictions with the available spectral information. This validated the
model of Sb 3 + in the Ca(II) site, and of charge compensation via the Sb-O
pairing.
Unfortunately the direct methods used to probe the location and environ-
ment of the Mn 2 + ion in fluoroapatite are not applicable to the diamagnetic
Sb 3 + ion. However, Mossbauer spectroscopy was recently used to track
antimony during the apatite synthesis process, through the transient
Ca(Sb0 3 h phase, and the subsequent incorporation into the apatite lattice
[31]. The spectra were not sufficiently resolved to extract all of the
Mossbauer parameters, but a variation in the isomer shift values with the
CI:F ratio was observed, thereby providing evidence that Sb was incorpor-
ated into the halogen environment of the lattice. This again suggested Sb 3 +
occupancy of the Ca(II) site.
In another Mossbauer study Mishra et al. [32] reached quite a different
conclusion, reporting that they found strong support for the model [33]
wherein a (P0 4 )3- group is replaced by an (Sb0 3)3- group. This interpreta-
tion of the Mossbauer results, however, is strongly dependent on the
calculations of several models representing different Sb 3 + surroundings, and
the models selected were considered to be incomplete [34].
Perhaps the strongest direct evidence for the Sb 3 + location derives from a
recent X-ray diffraction study by DeBoer et al. [35]. By using both Reitveld
refinements and electron density difference maps in the analysis of data
obtained from undoped and strongly doped (2.2% by weight Sb) Ca
fluoroapatite, it was possible to conclude that Sb 3 + was present in a Ca(II)
site, and that a missing or displaced fluoride ion was probably associated
with the substitution. (Similar measurements on a sample containing 3% by
weight Sb, that was non-typical in several other respects, provided a
different and less acceptable result.)
It has been noted [34] that Sb3+ is not a spherical ion, but has a lone pair
of electrons approximately occupying the volume of an oxygen ligand [35].
Occupancy of the adjacent F- site by this lone pair could explain both the
missing fluorine and the insensitivity of the Sb emission-excitation to
changes in halide type, but it still leaves unexplained how charge compensa-
tion is achieved.
In further speculations on the X-ray results DeBoer et al. [35] pointed out
that the lone pair of electrons was also responsible for the one-sided
coordination of ligands in some 79 known crystal structures containing Sb 3+.
This type of one-sided coordination would also be expected in halophos-
phates and would require something like the (Sb0 3 )3 - model of Mishra et al.
[32], but with Sb 3+ residing on a Ca(II) site rather than a pS + site.
Somehow this picture must be reconciled with that obtained by Moran et
at. [36] from a study of Sb-activated CaF AP using the technique of
magic-angle spinning-nuclear magnetic resonance (MAS-NMR). In this
study the MAS-NMR of 19F and 31p were used to derive quantative
information about the number of fluorides affected by each Sb3+, as well as
information on the spatial proximity of the affected fluorides. From this it
was possible to conclude that Sb 3+ must occupy a phosphorus site.
10.1.6 Rare-earth-activated phosphors for high-performance lamps

High color rendering, de-luxe fluorescent lamps have been manufactured
and marketed for decades. In such lamps a blend of broad-band emitting
phosphors (three or four) is used to produce an SPD with a CRI of '" 90.
This is invariably accompanied by a severe drop in lamp brightness.
In the middle 1960s efficient, line-emitting, rare-earth-activated phosphors
were developed for use in color television screens where color reproduc-
tion - especially in the red - had been a long-standing problem. Rare-earth-
activated phosphors efficiently excited by 254 nm were also reported, with
green emission (Tb3+), red emission (Eu3+), and yellow emission (Dy3+).
No fluorescent lamp containing such phosphors, however, was actually
marketed until the early 1970s, when the use of a revolutionary blend
containing three rare-earth-activated phosphors (tricolor blend or triphos-
phor blend) resulted in a fluorescent lamp of excellent CRI ('" 85) and
improved luminosity (up to 10% over standard halo phosphate phosphors).
In addition to higher CRI and brightness, enhanced output maintenance
over the lamp lifetime was demonstrated [37].
In standard halophosphate lamps (namely, 40 W dissipated in a 4 ft long
plasma) good efficacy (75 1m W- 1) with adequate color rendition ('" 60) can
be had at a phosphor cost of a few cents per lamp. The corresponding cost
for the triphosphor blend was about two dollars (mostly from the expensive
yttrium, europium, and terbium ions) and this more than tripled the lamp
price.
As a compromise between performance, phosphor costs and lamp-making
costs, a combination of halophosphate and triphosphor blend is nowadays
used in many lamp types. Typically, a base layer of the halo phosphate
phosphor is topped by a second layer of triphosphor blend. In this two-layer
configuration the expensive, but durable, rare earth blend is directly exposed
to the plasma and absorbs most of the available UV radiation, shielding the
328 Lamp phosphors
more vulnerable halophosphate. Thus, the advantages of the triphosphor

blend are fully exploited in a cost-effective manner. The phosphors con-
stituting the triphosphor blend will be now reviewed.
(a) Red-emitting component (Y203:Eu3+)

Although this phosphor can be prepared by solid-state diffusion of the
refractory oxides Y203 and EU203 in sequential firing-grinding steps, a
solution--co-precipitation route leads to a more uniform distribution of Eu
in the host matrix.
After the desired molar concentrations of the constituent oxides, Y203
and EU203, are dissolved in hot nitric acid, an oxalate co-precipitate is
formed by slow addition of oxalic acid. Careful control of solution pH and
temperature ensure that the Y and Eu oxalates precipitate without segrega-
tion. The mixed Y, Eu oxalate is subsequently separated from the solution,
dried and converted to the oxide by firing in air. The decomposition of the
oxalate to oxide will occur at temperatures as low as 800C, and the particle
size of the resultant oxide is determined by that of the mixed oxalate, which
in turn is sensitive to the precipitation conditions. The Y203:Eu3 + phos-
phor is normally fired between 1200C and 1500 C in the presence of
fluxing agents that promote crystallite growth and good particle morpho-
logy. The phosphor is then washed to remove residual flux and sieved.
Nowadays most rare earth vendors supply a specific mixed rare earth
oxalate or oxide to industrial customers. The phosphor manufacturer has
then only to perform the final firing.
Excitation and emission spectra for Y203:Eu are presented in Fig. 10.6.
The Y203:Eu3+ phosphor absorbs the 253.7 nm exciting radiation via a
100
80 \ \
~Ex i+- Em
.~ 60
=
In
~ 40
20
o
200
~ 300
t-\.A J lA. L.l
600
11
700
A
Wavelength
Fig. 10.6 Excitation and emission spectra for Y203:Eu: Ex, excitation for 611 nm
Eu 3+ emission; Em, Eu3+ emission for excitation at 253.7 nm.
broad absorption band that starts at - 290 nm and peaks at - 230 nm. This
absorption is attributed to charge transfer transitions involving Eu3+ and
nearby oxygen ions.
From the charge transfer states the excitation energy relaxes to the 5Dj
excited states of the 4f 6 configuration of Eu 3 +. Line emission is normally
observed from the 5D 2 , 5D h and 5DO levels to the 7F; ground term levels.
For typical commercial phosphors, less than 1% of the emitted photons
originate from the 5D 1 excited state, and fewer still from 5D 2 The dominant
emission is from the lowest excited state 5D o, particularly in the 5Do- 7F 2
hypersensitive transition at 611 nm, with a lifetime of -1 ms. As shown in
Fig. 10.7, two inequivalent cation sites are available to Eu3+, a C 2 site and
the centrosymmetric C 3i site. At room temperature the bulk of the emission
proceeds from Eu3+ centers in the C 2 site, as a result of energy transfer
between the two sites. At low temperatures the narrow linewidth (- 1 A at
4 K) of the intra-4f6 transitions can be exploited to follow a bewildering
array of energy transfer pathways [38].
Cz Site C3i Site
o yttrium
Oxygen
Fig. 10.7 Cation sites in Y203' Oxygen-deficient fluorite structure with oxygen
vacancies (joined by dot-dashed line) along a face diagonal for C2 sites, and along a
body diagonal for C3i sites.
The quantum efficiency of this phosphor, within a few per cent of unity, is
the highest of all lighting phosphors, and remains constant over a wide
range of Eu concentrations from 1 to 20 mol. %. The brightness maintenance
in lamps also exceeds that of other phosphors. At 254 nm the phosphor has
a relatively high reflectance (25%), so in blends its ability to capture the
available UV radiation is inferior to that of other blend components. As a
result this phosphor is not only the most expensive component of the blend,
but it is also the dominant blend fraction by weight.
(b) Green-emitting component (LaMgAl ll OI9 :Ce,Tb)

This phosphor crystallizes in the magnetoplumbite (PbFe12019) structure
(Fig. 10.8) typical of many metal hexa-aluminates such as SrAI 12 0 19 , with
330 Lamp phosphors
o 02-
Mirror Plane '-~ff-+<rr7 Al3+
t
Spinel
Me2+.Me3+
Block
!
Mirror Plane .....~":O-r"1r
Fig. 10.8 The magnetoplumbite structure of SrAI 12 0 19 , as a combination of spinel
blocks and ion conduction layers. The spinel block is composed only of AI3+ and
0 2 - ions, but will accommodate Ga3+ or Mg2+ substituting for AI3+. Larger
cations (Sr2+ ,La3+ ,Ce3+ ,Tb3+) are located in the conduction layer which is a mirror
symmetry plane. Only half the unit cell is shown, the complete cell being obtained by
reflection on either mirror plane.
Sr instead of Pb and AP + standing for Fe 3+. A variety of ionic substitutions

are possible in the magnetoplumbite structure and in the structurally related
~-aluminas. The effect of ionic substitutions on the crystal chemistry of these
materials has been described by Iyi et al. [39]. The phosphor composition is
arrived at by replacing the Sr2+ -AP + combination with the La3+ _Mg2+
pair. The Ce 3+ and Tb 3+ activators fully substitute for La, leading to the
preferred formulation (CeO.67 Tb o.33 )MgAl ll 0 19
In the phosphor synthesis, luminescent-grade Ce02, Tb 4 0 7, AI(OHh, and
MgO or MgC0 3 are fired at high temperatures (> 1400C) in a reducing
atmosphere of forming gas (nitrogen-hydrogen). The latter is required to drive
Ce and Tb to their trivalent state. Since a uniform distribution of the Ce and
Tb activators is required for high emission efficiency, fine particle size reagents
are selected and wet blended in a ball mill. To enhance reactivity and to
promote particle growth, some of the magnesium is added as MgF 2. The
fluorine is released in the firing and is not retained in the phosphor.
The solution route involves dissolving La203 and Tb 4 0 7 in nitric acid,
which is then added to an aqueous solution of the nitrates AI(N0 3k9H 20,
Ce(N0 3k6H 20 and Mg(N0 3h6H 20. Mixed-hydroxide precipitation fol-
lows the addition of ammonia. This precipitate is then fired appropriately.
Terbium concentrations higher than 0.5 formula units result in the
formation of a second phase. However, in view of its high cost, there is no
incentive to approach the solubility limit of terbium [40].
The 253.7 nm radiation is absorbed in a broad 4f-5d transition of Ce3+
(see Fig. 10.9). In the absence of Tb 3+ the reverse 5d--4f transition produces
an intense UV band with peak at 350 nm and a characteristic doublet
structure (curve Em in Fig. 10.9). Remarkably, no concentration quenching
is observed even for full substitution of Ce for La. As the terbium concentra-
tion increases, the cerium band emission is monotonically quenched and
replaced by the line emissions (5D r 7 F i transitions) from terbium. The
----. A (nanometers)
Fig. 10.9 Relative excitation spectrum (Ex) of the Tb 3+ emission and diffuse
reflection spectrum (R) of (Ceo.67Tbo.33)MgAll1019' Also indicated are the excita-
tion spectrum Ex( x 20) for direct emission by Tb 3+, amplified 20 times, and the
Ce3+ emission (Em) in CeMgAI l1 0 19 [41]. Reprinted by permission of the pub-
lisher, The Electrochemical Society, Inc.
sensitization of Tb 3+ occurs through the overlap of the Ce 3+ emiSSIOn

(curve Em in Fig. 10.9) with the 50 2 levels of Tb3+ (curve Ex( x 20) in Fig.
10.9). In lighting applications it is advantageous to convert the cerium UV
emission completely into terbium visible emission. This objective is realized
in this phosphor where at 0.35Tb (formula units) the Ce emission is ~99%
quenched. After reaching the Tb centers the excitation energy cascades
through the 50 2,3,4 levels, and is finally converted into blue, green and
yellow line emission from 50 4, as shown in Fig. 10.10. The dominant
emission lines peaking at 542 nm connect 50 4 with 7F 5'
This phosphor provides a classic example of sensitized luminescence. The
quantum efficiency for visible emission from the phosphor is 90-95% that of
y 203:Eu3+. A remarkable property of the phosphor is its positive tempera-
ture response for 254 nm excitation. Up to 500 DC this phosphor shows no
emission loss and is actually brighter at 200 DC than at room temperature.
This is related, on the one hand, to the lack of concentration quenching for
cerium, and, on the other hand, to the low probability of radiationless
de-activation of Tb 3+. An extremely stable and durable material, this
phosphor can withstand heavy milling, contrary to most phosphors, and is
soluble only in concentrated acid. Even at 900 DC neither Tb nor Ce shows
any tendency to oxidize to the corresponding tetravalent states, which are
either inert or outright deleterious to UV visible conversion processes.
( c) LaPO 4: Ce, Tb (alternative green emitter)

Although (Ce,Tb)MgAI l1 0 19 is currently the standard green-emitting com-
ponent of the triphosphor blend, it is facing increasing competition from a
phosphate-based analog, LaP0 4:Ce;Tb.
332 Lamp phosphors
100
80
Em
>-
.,!:: 60
=
'"CII
...d 40
20
o
300 400
A \.A J\ 600 700
500
Fig. 10.10 Emission from the 5D4 level of Tb3+ in (Ce,Tb)MgAll1019 (254nm
excitation).
The green-emitting phosphate basically operates as the aluminate. The

cerium emission in LaP0 4 is at slightly shorter wavelength and, since the Ce
concentration is lower, the doublet splitting of the cerium emission (due to
the ground state splitting) is more pronounced, with peaks at 320 nm and
340 nm. Similarly, the splitting of the Tb 3+ emission lines is more sharply
defined. Compared with the aluminate, the peak of the dominant 5D 4 - 7 F 5
transition is shifted to longer wavelengths by '" 3 nm, that is, closer to the
peak of the luminous efficiency function. The longer-wavelength transitions
5D 4 - 7 F 4,3 are enhanced, so the overall emission is slightly yellower than the
aluminate emission. Even these small changes are commercially significant
in optimizing the triphosphor blend performance. The phosphor also shows
excellent bake resistance during lamp manufacturing and exceptional stabi-
lity in lamp operation.
The synthesis of LaP0 4 is quite unusua1. A solid solution of the rare earth
oxides La203, Ce02, and Tb4 0 7 is blended with BP04 and fired in a neutral
atmosphere. Excess oxygen from Ce02 and Tb 40 7 evolves in an auto-redox
process early in the firing. At 1000 C-1100 C a highly exothermic exchange
reaction takes place, with the formation of the phosphate phosphor and of
liquid phase B20 3. The fired product is a very hard brick of the well-reacted
phosphor embedded in vitreous boric oxide. The brick is crushed, and
washing dissolves B20 3, leaving the well-developed phosphor particles [12].
(d) Blue-emitting component (BaMg 2 AI 16 0 27 :Eu2+)

This phosphor crystallizes in the magnetoplumbite structure and its syn-
thesis is similar to that of the related hexa-aluminate green with respect to
reagent preparation, fluxing, firing temperature and atmosphere [41]. BaF 2
as well as MgF 2 when used as alkaline earth sources also provide fluxing
action. Firing in reducing atmosphere is required to maintain the europium
in the divalent oxidation state, as opposed to the Eu 3+ needed for the
red-emitting phosphor. The europium content can vary widely, but generally
amounts to a '" 10% substitution for barium.
The strong phosphor absorption - throughout the UV and almost into
the visible - is due to allowed 4f-5d transitions (see Fig. 10.11). A fairly
narrow emission band (50 nm full width at half-maximum), peaking just
short of 450 nm and with excellent quantum efficiency, originates from the
lowest 5d level.
, ....... "", ,,,,..

100~--~~----~~--------'
\ ..l ,, 'J \
,
50 Ex , I
,, I
I
,
o __---'-____.,.""'____.I,.......;"'---L----'
~
200 300 400 500 600

Fig. 10.11 Excitation and emission spectra for BaMg 2 Al 16 0 27 :Eu: Ex, excitation
for 450 nm Eu2+ emission; Em, Eu2+ emission for excitation at 253.7 nm.
Spectral overlap of the excitation and emission bands favors concentration

quenching, an effect not observed in the Ce aluminates with their larger Stokes
shifts. Europium is the most costly component in this phosphor and is kept to
a minimum, 10-15% of Ba, since higher concentrations of Eu will result in
efficiency loss - due to concentration quenching - and in higher material costs.
The phosphor normally represents only a '" 10% weight fraction of a
triphosphor blend for 4100 K, and blends for lower color temperatures
require even less of the blue-emitting component. So this phosphor is a
relatively minor blend constituent. In compact fluorescent lamps, which run
at higher temperatures and plasma densities, this phosphor is not even
required, since blue components in the lamp output are adequately contri-
buted by the visible lines of mercury (Fig. 10.1).
The Eu 2+ emission is extensively tunable by alterations of its crystalline
environment, as is the case with the 5d-4f transitions of Ce 3+. Under
253.7 nm excitation the emission band can be shifted all the way from the
UV to the red by the proper choice of the host matrix. This is in contrast to
the intra-4f shell transitions of Eu 3+ and Tb 3+, which are well known for
their insensitivity to their surroundings.
When doped with Eu2+, compounds contained in the phase diagrams for
BaO-SrO-MgO-AI203 can give UV or blue, or green emission. The
334 Lamp phosphors
stoichiometry of this phosphor is therefore very important, since it deter-

mines the spectral location of the emission band. The composition of the
phosphor as indicated above simply designates the starting formulation, and
does not represent a specific compound. The actual phosphor is generally
thought to be a solid solution of two or more ~-alumina polymorphs [71]
and usually includes other aluminates as secondary phases. The details of
the syntheses are generally considered a trade secret by the phosphor
manufacturers, and are probably strongly dependent on the specific firing
equipment employed.
This phosphor, as the other rare-earth-activated triblend components,
exhibits excellent brightness maintenance during lamp operation.
10.1.7 Lamp manufacturing

In the manufacture of a fluorescent lamp, the phosphor is coated on the
interior of the glass tube, either by filling it with a phosphor suspension or
by flushing the lamp walls with the suspension injected from above. The
excess suspension is allowed to drain out and is collected for re-use. In order
to ensure a uniform thickness of phosphor over the entire length of the lamp
the viscosity, solvent content, specific gravity of the suspension, and drying
rate must be carefully controlled. With some of the newer lamp types that
employ a halophosphate base coat and a triphosphor second coat, a
subsequent coating-and-drying cycle is required. Furthermore, the second
coat must not dissolve the base coat or physically damage it during the
drying process.
Phosphor layer uniformity is very important. Phosphor layers that are
uniformly too thick or too thin alter the chromaticity and brightness of the
lamp relative to the manufacturer's specifications, but such changes are not
readily detected by the consumer. On the contrary, when the thickness of the
phosphor layer varies along the lamp length, the attendant color variations
become visually objectionable to the user.
The components in the phosphor suspension can only be discussed in a
general way, because each lamp manufacturer has developed proprietary
formulations that are further differentiated by the lamp type coated. Gener-
ally speaking, in addition to the phosphor(s), the suspension usually includes
a film-forming binder, solvent(s) for the binder, and - if necessary -
surfactants, de-foamers, and wetting agents. Another inorganic, non-
luminescent, adherence-promoting additive is also required to bond the
phosphor to the glass after the organic constituents have been eliminated.
The liquid portion of the phosphor suspension is commonly referred to as
the coating vehicle. Two types of vehicles are currently in use. One type of
vehicle employs an organic liquid, such as butyl acetate or xylol as a solvent
for the binder, while the other type uses water. The organic-based vehicles,
in general use until the mid-1960s, have been described in some detail by
Butler [3]. These are increasingly being replaced by systems that use
water-soluble polymers, such as poly-ethylene oxide (Polyox, Union Carbide
Corp.), poly-methyl-vinyl ether-maleic anhydride [42], poly-acrylic acid
[43], and carboxymethyl cellulose. Sometimes it is necessary to combine two
or more of these materials to obtain optimum coating properties. Whenever
two-coat lamps are to be produced, an additional cross-linking ingredient is
also required to render the base coat insoluble.
Both Alon C (Degussa Corp.) and Ba(N0 3 h [44] have served as the
inorganic adherence-promoting constituent in lamp coating suspensions.
These materials are compatible with all of the aforementioned polymeric
materials, provided that established suspension formulations and com-
pounding procedures are observed.
These water-based vehicles offer the advantages of lower cost, reduced
toxicity and flammability, and eliminate the need to dry the phosphors after
the final post-synthesis surface-washing process.
The water-based vehicles are a mixed blessing. To begin with, they are
more difficult to dry, particularly when employed in the coating of U-shaped
and compact fluorescent lamps. They also require the use of antifoaming
ingredients to prevent bubble formation and inhibitors to prevent the
growth of bacteria during long-term hold-over. Phosphor blends may be
held in suspension for long periods (~12 months) and the tendency for
chemical interactions between phosphors (in the case of blends), and
between the phosphors and the other suspension components is much
greater in these water-based vehicles.
Flocculation is also much more of a problem in these suspensions. Mild
flocculation results in poor lamp appearance, while severe flocculation can
render a suspension completely unusable. While phosphor particles are too
large to be considered colloidal, many of the rules governing the stability of
colloidal dispersions have been found useful in preventing flocculation-
coagulation in lamp coating suspensions. Electrostatic forces are mainly
responsible for keeping the phosphor particles suspended in coating vehicles
based on Polyox, which is a non-ionic material. The remainder of the
aforementioned polymeric materials are anionic, and in coating vehicles
employing them the phosphor suspension is electrosterically stabilized. A
knowledge of the electric charge on the polymer molecules and of the
isoelectric points (IEPs) of all of the solid phases,as well as close control of
the suspension pH are critical in precluding floc formation. In particular, the
pH of the suspension should be controlled at a value where the zeta
potentials of all solid phases, including phosphors and adherence promoters,
have the same charge and approximately the same magnitude. Thus,
suspensions having solid phases with widely varying values of their IEPs
should be avoided.
Most of the non-phosphor components of the suspension interfere with
efficient lamp operation and longevity, and must be removed by pyrolysis in
336 Lamp phosphors
a manufacturing step known as 'lehring'. While the lehr process is an

indispensable part of the lamp maufacturing, it does have some adverse
effect on the light output of most phosphors. During the lehr, the phosphor-
coated bulbs are moved through a gas-fired furnace, heated to '" 600 C and
continuously purged with a stream of air to ensure complete combustion of
the organic binder. If the furnace temperature is too high, sodium ions from
the glass can migrate into the phosphor, resulting in a loss of lumen
maintenance. However, this type of degradation is somewhat dependent on
phosphor chemistry. On the other hand, furnace temperatures that are too
low, or furnace residence times that are too short, produce phosphor layers
containing light-absorbing carbon and/or hydrocarbon deposits. Occa-
sionally, it is also possible to relate incomplete combustion to traces of
fire-retardant impurities in the phosphor.
Lehr-induced oxidation or reduction of the activator-sensitizer ions in the
phosphor is yet another cause for serious degradation in phosphor perform-
ance. In fact, there are a number of very efficient phosphors that cannot be
used in fluorescent lamps, because they are critically damaged in the lehr.
After the lehr, flares which incorporate the exhaust tube and coated
tungsten electrodes are sealed to the end of the phosphor-coated bulb. The
bulb is then heated, exhausted, and subjected to a number of purge and
spark steps designed to rid the lamp thoroughly of residual air and water.
Finally, a drop of mercury is added to the lamp, and, following a last
exhaust cycle, it is pumped down to the correct pressure of fill gas, sealed,
and end caps are applied. In modern, high-speed fluorescent-lamp produc-
tion lines, all of the various operations described above proceed at a rate
that enables the production of between 3500 and 5000 lamps per hour,
depending on lamp type.
10.1.8 Lamp maintenance

Maintenance, or lumen maintenance, is a measure of how well a fluorescent
lamp maintains its brightness over a measured period. Usually maintenance
is expressed as a percentage of the initial brightness remaining after a given
number of hours of lamp operation. To the technical personnel involved in
phosphor research and lamp manufacturing, initial brightness, termed O-hour
brightness, is the brightness measured immediately after the lamp discharge
has stabilized, and the percent maintenance (%M) of main concern is the
0-100 h value. In the more extended time frame of interest to illumination
engineers, the initial brightness is the value recorded at 100 h, beyond the
early lumen loss, and the %M is the 100 h to 1000 h value. Unless stated
otherwise, the %M to be discussed will apply to the 0-100 h interval.
A variety of factors have been recognized as causative agents in the
temporal decline of the light output from lamps, namely glass composition
of the bulb, gas-fill pressure, choice of rare gases, electrical loading (watts per
square centimeter of lamp), coating suspension composition, color center
dynamics, phosphor surface chemistry, and photochemistry. Moreover,
these factors can, and often do, act in concert. In the early studies of lumen
maintenance [3] it was found that the lumen drop could be represented by a
rate equation for first-order chemical reactions. When several processes
operate concurrently, these reactions can be combined to give an equation
for the overall lumen loss of the form
Lo- L t =A(1-e- at )+ B(1-e- bt )+ C(1-e- ct ) (10.1)
where t is the elapsed time Lo the initial lumen output (following discharge
stabilization) L t the lumen output at time t, and A, B, and C are the
maximum lumen loss for the processes with the rate constants a, b, and c.
Rate equations of this type were found to be useful in differentiating between
various lumen depreciation mechanisms. The first term of the equation was
shown to be related to gas-phase processes [45], the second term described
solid-state reactions that affect light output up to about 2000 h [46], and the
final term was attributed to photochemical processes [47]. Some time later
it was discovered that - in addition to the longer-term depreciation
mechanisms discussed above - short-term processes of equal severity acted
to reduce the light output within the first few minutes of lamp operation
[48].
The most serious lumen loss mechanism in fluorescent lamps that is not
phosphor related is the UV-induced visible absorption, or solarization, in
the lamp envelope. For example, solarization can decrease the brightness of
a 40W cool white fluorescent lamp by as much as 60 1m, or about 2% of its
rated output. Solarization of the soda lime silica glass, from which the lamp
envelope is made, has been attributed to defect centers associated with
non-bridging oxygens acting as hole traps that absorb in the visible region
of the spectrum [49]. Antimony oxide has been found to be a beneficial glass
component that inhibits solarization and reduces the visible absorption
associated with Fe impurities in the lamp glass. While the function of Sb 2 0 3
in reducing solarization is not fully understood, it is believed that Sb 3 + acts
as an alternative hole trap that does not absorb visble light.
A similar, albeit more complex, type of solarization is also responsible for
the temporal decrease in the light output of the phosphor itself. In phos-
phors the UV-induced absorptions (from color centers) can be observed
anywhere from the deep UV to the infrared region of the spectrum. Color
centers can degrade phosphor brightness by (a) acting as a parasitic UV
absorber, that competes with the activator(s) in capturing the excitation
energy, (b) providing additional paths for the radiationless energy transfer
from the excited activator to defects and other energy sinks and (c)
absorbing phosphor emission. These adverse mechanisms can operate singly
or in concert, and the relative importance of each is dependent on the type
of phosphor being used [24].
338 Lamp phosphors
As might be expected, the role of color centers in the lumen maintenance

of calcium halophosphate has received considerable attention because of its
commercial importance. Only a brief summary of the results of this extensive
body of work can be attempted here. We refer the reader interested in a
more detailed treatment to the excellent review article by Flaherty [24].
Deviations from the stoichiometric composition cause defect centers to
form even in unactivated and unirradiated halophosphates. These defect
centers have been termed 'X' centers and 'V' centers [50J and are responsible
for UV absorptions of 250 nm and 185 nm respectively. The X center
absorption can be suppressed by stoichiometry control and by incorpora-
tion of the Mn and Sb activators. UV irradiation at 185 nm converts the Y
centers into new color centers (termed H(II) centers) that absorb - and are
bleached by - 254 nm radiation. As might be expected, under irradiation
from the 185 nm and 254 nm lines from a low-pressure Hg plasma, after the
initial, very rapid formation of H(II) centers by the 185 nm component, an
equilibrium condition results from their partial bleaching by the 254 nm
component. The degree of short-term lumen depreciation in halophosphates
is therefore dependent on the relative, rather than absolute, intensities of the
185 nm and 254 nm lines [51].
The ability of the 185 nm irradiation to create H(II) centers is reduced in
proportion to the 185 nm absorption introduced by certain intentionally
added cationic impurities [52]. Both Sb and Cd have been found to be
effective in this respect. The Sb activator absorbs both the 185 nm and the
254 nm components, the former more strongly. Consequently, as the Sb
concentration is increased, Sb can successfully compete with the Y and H(II)
absorptions, and reduces the 185 nm to 254 nm intensity ratio within the
phosphor, which leads to a lower steady-state concentration of H(II) centers
[24].
The addition of a few per cent of Cd to CaFCIAP:Sb,Mn does not
contribute to the fluorescence of the phosphor, but it introduces a strong
absorption at 185 nm that reduces the intensity of the damaging 185 nm
component [53]. This has been found to increase the initial light output of
fluorescent lamps [54]. Optical and ESR studies of color center formation in
Cd-doped crystals and powders showed that Cd occupies the Ca(II) site
[55]. For public-health reasons cadmium has now been eliminated from the
phosphors manufactured in the United States and Japan, but it is still used
in lamps in many countries.
In addition to the UV-induced absorptions and to the photochemical
processes that degrade emission by changing the valence of activator ions or
impurities located on, or near, the phosphor particle surface, there are other
even more deleterious processes taking place in the fluorescent-lamp dis-
charge. Radial cataphoresis [3, 56J draws plasma electrons to the lamp wall
and creates a sheath of negative charges near the phosphor. This attracts
Hg+ ions, which then recombine with the electrons. These recombinations
release energies capable of severing chemical bonds, as well as introducing
defects and associated color centers in the phosphor. In a maintenance study
involving an Na3(CeO.6S Tb o.3s )(P0 4h phosphor it was found that ion-
electron recombination produced a non-luminescent surface layer, whose
depth increased approximately as the square root of time, and appeared to
be associated with the oxidation of Ce3+ to Ce 4+ [57].
It is possible to minimize these lumen depreciation processes by employ-
ing phosphors with strong chemial bonds. This is evidenced by the consist-
ently high lumen maintenance (98-99%) exhibited by the rare-earth-ac-
tivated aluminates and Y203:Eu. While phosphors based on high bond
strength hosts are plentiful, few possess the required combination of excita-
tion and emission characteristics, high quantum efficiency, and economic
viability to make them suitable for fluorescent-lighting applications.
This has led several workers to explore the potential benefits of protecting
the less stable phosphors with non-emitting, high-stability, wide-band-gap
materials, such as Al 20 3 and Y203' Improvements in lumen maintenance
were reported for lamps containing Zn2Si04Mn (willemite) [58J and CaF-
CIAP:Sb,Mn [59J as a result of overcoating the phosphor layer with a layer
of colloidal A1 20 3. Similar improvements were reported for willemite and
cool white phosphors when coated by electron beam evaporation with a
surface layer of pure, or Eu-activated Y203 [60, 61, 70]. While the lumen
maintenance of the lamps was substantially improved by both the colloidal
alumina and the evaporated yttria layers, the initial brightness was dimin-
ished relative to lamps without the protective oxide over-layers.
Apparently the morphology and/or method of depositing the protective
oxide is important, because both maintenance and initial brightness were
shown to be enhanced by the application of a sputtered coating of vitreous
alumina [62]. An alternative to the sputtered alumina layers - that pro-
vides almost equivalent initial brightness and improved lumen maintenance
in combination with convenience in lamp manufacture - has been de-
scribed by Sigai [63, 64]. Each phosphor particle is encapsulated in a
conformal, non-particulate, CVD coating of Al 20 3 before the phosphor is
applied to the lamp envelope. Further enhancements in initial brightness
were provided by post-CVD coating processing that also stabilized the
alumina against hydration-solubilization in water-based lamp coating sus-
pensions [65]. Performance improvements were reported for various types
of fluorescent lamps employing coated cool white and coated willemite
phosphors.
Electrical loading has also a pronounced effect on lumen maintenance. In
modern fluorescent lamps, changes in the rare gas mixture and pressure, and
increases in loading (current density), have made it possible to bolster lamp
lumens by a factor of 1.5-2, in going from a standard of F40 (i.e. 40 W) lamp
type to high output (HO) or very high output (VHO) lamp types of the same
physical size. The effect of higher electrical loading on the lumen main ten-
340 Lamp phosphors
ance of various types of cool white lamps (1.5 in diameter) is depicted in Fig.
10.12, together with their operating current and loading characteristics in
watts per foot of lamp length. The lamps with the higher power inputs have
poorer maintenance than the standard F40 lamps because the overall
intensity of UV radiation impinging on the phosphor(s) and on the lamp
envelope has increased, and because the intensity of the 185 nm Hg emission
has increased relatively to the 254 nm component. While all the lamp types
maintain respectable levels of light output for very long times, it is evident that
an improvement in the lumen maintenance of the higher-loaded lamps is
desirable. Maintenance is of course also dependent on the type of phosphor
employed in the lamp and on the type of glass used in bulb fabrication.
, / F 40 and Slimllne (430 ma. -lOWIrt.)
=
'"
E
:s
...:I
60
~
......
S
=
8
40
""
GI
Q.,
20
0
0 4000 8000 12,000
Hours Burned
Fig. 10.12 Lamp output brightness as a function of current loading in commercial
fluorescent-lamp types.
10.1.9 Speciality phosphors

A significant market exists for fluorescent sources with emission matching
the response of photosensitive materials. Such sources will be briefly
summarized in order of decreasing photon energy of their emission.
(a) Germicidal lamps

These UV-emitting lamps do not require phosphors, but are a useful starting
point for a brief survey of the biological effects of UV radiation. In this
connection the UV spectral region has been divided into three ranges:
uve <260nm
UVB 260---320 nm
UV A 320---400 nm
Germicidal lamps are designed to emit in the uve region, and consist of
a phosphorless fluorescent tube with an envelope transparent to the
253.7 nm radiation. Absorption of this radiation disrupts the DNA mole-
cules, and so these lamps are used to kill airborne viruses, bacteria and
fungi. Such lamps can be also be used in water purification systems. uve
radiation is of course equally damaging to human tissues and must be used
with caution.
(b) Erythemal lamps

These lamps emit in the UVB region and cause reddening (erythema) of the
human skin; they are clinically used in the treatment of skin disorders, such
as psoriasis. The production of vitamin D in the human skin is also
stimulated by UVB radiation. The lamps are also employed in industrial
photochemical applications, such as polymer curing.
(c) Phosphors for skin-tanning lamps

In the last decade artificial-tanning salons have become quite popular,
particularly in Europe. Many types of tanning lamps are commercially
available and much has been learned about their effective utilization.
Generally, the bulk of the radiation from these lamps falls in the UV A
spectral region. This energy is absorbed by the human epidermal pigment,
melanin, resulting in its discoloration. Depending on the power distribution
of the lamp output in the UV A region, the color of the treated skin will vary
from yellow through bronze to tan.
The lamp must also provide some output in the UVB spectral region, for
this does redden the skin but it also stimulates the production of the melanin
pigment (melanogenesis). Without some UVB in the tanning-lamp spectrum,
all the pre-existing skin pigment is tanned and further irradiation with UV A
will have no effect. Should a person attempt deeper tanning through
prolonged exposure to UV A radiation alone, then the fine capillaries in the
subcutaneous tissues will be thermally damaged. Thus, the tanning lamp
must have a carefully balanced UV output, namely enough UVB to provide
melanogenesis without erythermal discomfort, and a well-tuned UV A
emission to provide the desired hue. To obtain a balanced UV output one
also takes advantage of the different transmission cut-offs of the lamp
envelope. For psychological reasons some visible-emitting phosphor is
generally blended with the ultraviolet emitters, so as to recreate the
sensation of sunlight. Emission and excitation spectra for several of the
phosphors used in tanning applications are given in Fig. 10.13.
The all-purpose UV -emitting phosphor is the disilicate BaSi 2 05: Pb 2 +.
Under 254 nm excitation the divalent Pb emission is a broad band peaking
at 366 nm, and with a half-width of 60 nm. The short-wavelength side of the
342 Lamp phosphors
120
:. . Hg
100
I
80
:>.
:!:l
rn
1:1 60
CD
B 40
20
0
200 250 300 350 400 450
Fig. 10.13 Excitation spectra (Ex) for peak emission and emission spectra (Em) for
254 nm excitation for phosphors used in skin-tanning lamps: spectra A, SrB 60 lO :Pb;
spectra B, Sro.3Mgo.63CeO.61All1018.3; spectra C, BaSi 2 0s:Pb.
emission band extends slightly into the UVB region for adequate
melanogenesis. The phosphor exhibits chronically poor maintenance in
lamps, especially at the high current loadings of the tanning lamps. This
requires frequent re-Iamping at the tanning salons to ensure consistent
radiation dosing. The BaSi 2 0 s :Pb 2 + phosphor is also normally present in
the 'long-wavelength UV' mineral lamps (black lights).
As the art of tanning improves, custom blends of this phosphor with other
UV emitters are being introduced in lamps to control the degree of tanning
in users of different racial stock. The UV -emitting phosphors commonly
found in these blends are cerium-activated aluminates, such as SrAI 12 0 19 :Ce.
A borate phosphor with a narrow (35 nm) emission band peaking at 300 nm,
namely SrB 6 0 1o :Pb2+, is used as an intense source of UVB radiation.
Many industrial processes requiring accelerated bleaching and fading of
materials also employ lamps containing UV -emitting phosphors and blends
thereof.
(d) Phosphors for reprographic applications

The fluorescent lamps used in photocopiers are configured as aperture
lamps. First a thick layer of phosphor (often with a highly reflective base
coat of alumina or titania) is coated onto the lamp bulb. Then the phosphor
is scraped from the bulb along a narrow longitudinal channel, 5--45 wide.
This gives the lamp a highly directional output as desirable for Xerox-type
copiers. The phosphor emission is chosen to match the action spectrum of
Phosphors for HPMV lamps 343
the photosensitive element on the copier drum. The phosphor should also
show appreciable emission in the spectral region absorbed by the document
ink.
Finally, the europium-activated strontium pyrophosphate, Sr 2 P 20 7 : Eu 2 + ,
is contained in lamps used to make architectural blueprints in the Diazo
white printing process.
10.2 PHOSPHORS FOR HIGH-PRESSURE MERCURY VAPOR

(HPMV) LAMPS
An early form of discharge lamp, that preceded both the incandescent lamp
and the standard fluorescent lamp, relied on the production of visible line
emission from a high-density Hg plasma. Figure 10.14 shows an improved
modern version of such a light source, that has 20-25% efficiency in
converting input electrical power into visible radiation, and yields much
higher brightness per unit volume than the standard fluorescent lamp. The
small bulb contains a rare gas (typically argon) and Hg.
22mmOD
I
~ ~-~~~~~~-~::::::~~ ---
~~~~~~~~~~~~
c
Fig. 10.14 Schematic of a high-pressure mercury vapor (HPMV) source: A, cath-
ode; B, exhaust tubulation; C, molybdenum foil seal. Reproduced with permission,
MIT Press.
In operation, up to 400 W of electrical power are dissipated in the small

bulb, so its temperature becomes very high, and the Hg pressure exceeds
several atmospheres. In such a dense plasma the Hg UV emissions at
185 nm (from the 61 P 1 level) and at 254 nm (from the 6 3 P 1 level) become
self-trapped, and visible emission is produced by transitions from higher
excited levels to the 6 1 P 1 and 63 P j levels. The lamp output is shown in Fig.
10.15 and is markedly deficient in red components.
Safety considerations connected with the elevated pressures and tempera-
tures during operation, as well as corrosion problems with the electrical
leads, dictate that the operating bulb be contained in an outer glass
envelope, whose inner surface temperature can reach 200 C-250 C during
344 Lamp phosphors
300 400 500 600 700
Wavelength (om)
Fig. 10.15 SPD from an HPMV lamp. Reproduced with permission, MIT Press.
lamp operation. The color rendering of such a lamp can be improved by a

phosphor capable of red emission. The requirements for such a phosphor
are
1. ability to convert efficiently into visible light (preferably red) both the
long-wavelength and short-wavelength UV output from the high-press-
ure mercury plasma, which cumulatively exceeds by '" 25% the visible
output, and
2. efficient UV -to-visible conversion at 200 C-250 C
The color-correcting phosphor for HPMV sources was originally Mn-
activated fiuorogermanate, which can be roughly viewed as the germanate
analog of the halophosphates discussed in previous sections. In this
phosphor, still used in speciality lamps for enhanced plant growth, the
red-emitting species is Mn 4 + (electronic configuration 3d 3 ), whose struc-
tured red emission is somewhat reminiscent of the line emission in ruby from
the iso-electronic Cr3+. The relatively good thermal stability of the Mn 4 +
emission can be explained by qualititative considerations: line emission is
symptomatic of a potential energy surface in the excited state that is parallel
to that of the ground state, and hence less susceptible to de-activation by
thermally activated crossing into the ground-state potential surface.
The Mn-activated fiuorogermanate phosphor has a definite yellow body
color and absorbs significantly throughout the visible range. Furthermore,
its emission is far into the red, and so, although its presence in HPMV
sources boosts the color rendering, it cannot compensate for the lumen loss
associated with the visible absorption by the phosphor.
The body color is absent in Sr-Mg orthophosphate activated with
divalent tin (see Table 10.1), so its broad red emission boosts both the CRI
and the lumen output from the lamp. Currently these beneficial effects also
derive from the use of Eu-activated yttrium vanadate (or the phosphovana-
date analog). The vanadate group absorbs over most of the UV region, and
Current issues and projected developments 345
it transfers efficiently the excitation energy to the Eu3+ ion, whose emission
is generally not susceptible to thermal quenching.
YAG:Ce is also employed in HPMV sources. Although the Ce 3 + yellow
emission is poorly excited at 254 nm, it is quite efficiently excited in the long
UV and in the blue. Absorption by the phosphor of blue plasma com-
ponents is used to advantage in providing the 'warm' visual perception
associated with sources of low color temperature.
10.3 CURRENT ISSUES AND PROJECTED DEVELOPMENTS
10.3.1 Inexpensive alternatives to the rare-earth-activated phosphors

In the phosphor area the top priority of the fluorescent-lighting industry
is the replacement of the high-performance, but very expensive, rare-
earth-activated phosphors with cheaper materials. This essentially means
replacing the rare earth ions with transition metal ions or post-transition
ions.
Since in current tricolor blends the dominant component by weight is
Y203: Eu, which is also the most expensive phosphor, there is a keen
interest in developing less expensive alternatives to Y 203: Eu. Recalling
from Table 10.1 that the lamp performance of Y2 0 3 :Eu tops that of all
known phosphors, the phosphor technologists are faced with a very arduous
task. Even if it may prove impossible to duplicate in other materials the
Y203: Eu lamp performance, the industry needs will spur efforts both to
improve the performance of known non-rare-earth, red-emitting materials,
and to seek new hosts wherein the efficiency of such red-emitting activators
is increased.
10.3.2 Vacuum UV excitation of phosphors

The present interest in phosphors excited by hard UV (vacuum UV)
radiation with A< 200 nm is driven by three potential application areas.
First, phosphors are being investigated as detectors of soft X-ray radiation.
Second, hard UV excitation of phosphors is of interest in multiphoton
emission processes. Finally, rare gas plasmas devoid of Hg are intense
sources of vacuum UV radiation.
About 15 years ago two research groups associated with the lighting
industry reported the observation of two-photon emission processes, follow-
ing absorption of radiation at ~ 220 nm, or at shorter wavelengths [66, 67].
Quantum efficiencies for visible emission well in excess of unity (up to 1.40)
were observed for Pr 3 + in fluoride hosts in a process labeled as photon
cascade emission (PCE) or photon splitting. (When the UV emission
components were included, the overall emission quantum efficiency rose to
~ 1.7.)
346 Lamp phosphors
Phosphor and discharge limitations precluded the exploitation of these high

quantum efficiencies in practical light sources. Specifically, low-pressure Hg
plasmas have relatively poor output at wavelengths shorter than 254 nm; the
Pr 3 + emission in question was mostly in the deep blue, at '" 406 nm, a
singularly inappropriate wavelength for illumination purposes; finally, the
fluoride phosphors underwent rapid discoloration during lamp operation.
Multiphoton emission processes were earlier reported by Soviet re-
searchers [68] in more standard phosphor materials, such as willemite and
y 203: Eu 3+, for excitation by the vacuum UV output from rare gas
plasmas. Recent measurements utilizing synchroton radiation [69] have
confirmed and expanded these earlier observations. This type of emission
process could therefore be exploited to convert into visible light the intense
vacuum UV output of rare gas discharges that do not contain mercury.
These multiphoton processes look quite impressive when gauged in terms of
quantum efficiency, but less so when translated in terms of energy efficiency.
10.3.3 IR-emitting materials

As we have seen in previous sections, lamp phosphors are not restricted to
the conversion of UV radiation into visible light, but are also useful as
generators of UV radiation. Conversely, the potential use of phosphors in
the near and mid infrared (IR) has not been explored yet, although
technologically the IR is becoming a very important area. For instance,
optical communications heavily rely on the infrared emission from semicon-
ductor lasers as signal sources, and on rare earth ions, such as Er 3 +, Pr 3 + ,
and Nd3+, as optical amplifiers at the 1.55 11m and 1.3 11m communication
wavelengths.
One shortcoming of rare earth ions as optical amplifiers is their narrow
tuning range, a limitation that could be overcome by the use of infrared-
emitting transition metal ions. Progress in this area is hampered by the lack
of relevant information on material properties leading to efficient, room
temperature infrared emission.
Furthermore, although semiconductor lasers are an ideal match for
individual optical fibers, the pumping of clusters of fibers and amplifiers may
be more conveniently performed with a pulsed, conventional light source.
Here IR-emitting phosphors would be of value in converting the broad
spectral output from the pulsed source into the narrower IR bands required
for efficient pumping.
REFERENCES
1. Waymouth, IF. (1978) Electrical Discharge Lamps, MIT Press, Cambridge, MA.
2. Wyszecky, G. and Stiles, W.S. (1967) Color Science. Concepts and Methods,
Quantitative Data and Formulas, Wiley, New York.
References 347
3. Butler, K.H. (1980) Fluorescent Lamp Phosphors, Pennsylvania State University
Press.
4. Blasse, G. (1987) Mater. Chem. Phys., 16, 201-236.
5. Brattain, H.G. (1987) In Chemical Analysis (ed. J.D. Winefordner) Vol. 77, Wiley,
New York.
6. Pappalardo, R.G. (1987) In Spectroscopy of Solid-state Laser-type Materials (ed.
B. Di Bartolo), Plenum, New York, pp. 397-429.
7. Welker, T. (1991) J. Lumin., 48-49, 49-56.
8. Pappalardo, R.G. and Peters, T.E. (1984) J. Lumin., 31-32, 284-286.
9. Bril, A. and de Jager-Veenis, A.W. (1976) J. Electrochem. Soc., 123, 396-398.
10. McKeag, A.H. and Ranby, P.W. (1949) US Patent 2,448,733.
11. Jenkins, H.G., McKeag, A.H. and Ranby, P.W. (1949) J. Electrochem. Soc., 96,
1-12.
12. Chenot, C.F. (1990) Personal communication.
13. Kreidler, E.R. (1967) Thesis in Ceramics 68-3549, Pennsylvania State University.
14. Wanmaker, W.L., Hoekstra, A.H. and Tak, M.G.A. (1955) Philips Res. Rep., 10,
11-38.
15. Rabatin, IG. and Gillooly, G.R. (1964) J. Electrochem. Soc., 111, 542-546.
16. Parodi, J.A. (1967) J. Electrochem. Soc., 114, 370-377.
17. Rabatin, J.G., Gillooly, G.R. and Hunter, lW. (1967) J. Electrochem. Soc., 114,
956-959.
18. Hoekstra, A.H. (1967) PhD Thesis, Technical University of Eindhoven.
19. Wachtel, A. and Ryan, F.M. (1973) J. Electrochem. Soc., 120, 693-695.
20. Chenot, C.F. (1986) US Patent 4,613,453.
21. Homer, H.H., Butler K.H. and Bergin, MJ. (1956) Illumin. Eng., 51, 214.
22. Martyny, W.C. (1962) US Patent 3,047,512.
23. Graff, E.A. (1972) J. Electrochem. Soc., 119, 118-120.
24. Flaherty, J.M. (1981) J. Electrochem. Soc., 128, 131-139.
25. Johnson, P.D. (1961) J. Electrochem. Soc., lOS, 159.
26. Ryan, F.M., Ohlmann, R.c., Murphy, 1, Mazelsky, R., Wagner, G.R. and
Warren, R.W. (1971) Phys. Rev. B, 2, 2341-2352.
27. Suitch, P.R., LaCout, J.L., Hewat, A. and Young, R.A. (1985) Acta Crystallogr. B,
41, 173-179.
28. Ouweltjes, lL. (1952) Philips Tech. Rev., 13, 346-351.
29. Davis, T.S., Kreidler, E.R., Parodi, lA. and Soules, T.F. (1971) J. Lumin., 4,
48-62.
30. Soules, T.F., Davis, T.S. and Kreidler, E.R. (1971) J. Chem. Phys., 55,1056-1064.
31. Fraknoy-Koros, V., Gelencser, P., Levay, B. and Vertes, A. (1975) J. Lumin., 49,
467-474.
32. Mishra, K.C., Patton, RJ., Dale, E.A. and Das, T.P. (1987) Phys. Rev. B, 35,
1512-1520.
33. Dale E.A. and Berkowitz, J.K. (1986) 170th Electrochemical Society Meeting,
October 1986, Abstract 708.
34. Oomen, E.W.J.L., Smit, W.M.A. and Blasse, G. (1988) Mater. Chem. Phys., 19,
357-368.
35. DeBoer, B.G., Sakthivel, A., Cagel, lR. and Young, R.A. (1991) Acta Crystallogr.
B, 47,683-692.
36. Moran, L.B., Berkowitz, lK. and Yesinowski, J.P. (1992) Phys. Rev. B, 45,
5347-60.
37. Verstegen, J.M.P.J., Radielovic, D. and Vrenken, L.E. (1975) J. IES, 4, 90-98.
38. Pappalardo, R.G. and R.B. Hunt, R.B., Jr (1985) J. Electrochem. Soc., 132,
721-730.
348 Lamp phosphors
39. Iyi, N., Takekawa, S. and Kimura, S. (1989) J. Solid State Chem., 83, 8-19.
40. Sommerdijk, J.L. and Verstegen, J.M.PJ. (1974) J. Lumin., 9, 415-419.
41. Verstegen, J.M.PJ. (1974) J. Electrochem. Soc., 121, 1623-1628.
42. Martyny, W.e. (1961) US Patent 2.987,414.
43. Repsher, R.W. (1971) US Patent 3,585,207.
44. Beaumont, H.H. and Friedman, A.I. (1969) US Patent 3,424,605.
45. Mooney, RW. and Gungle, W.e. (1956) Illumin. Eng., 51, 793.
46. Lowry, E.F. and Major, E.F. (1949) Illumin. Eng., 44, 98.
47. Lowry, E.F. (1952) Illumin. Eng., 47, 639.
48. Vrenken, L.E., de Vette, T.H. and van der Wolf, R.W. (1964) J. Illumin. Eng., 59,
59-66.
49. Onorato, P.I.K. and Alexander, M.N. (1983) Personal communication.
50. Ryan F.M., Ohlmann, R.e., Murphy, J., Mazelsky R., Wagner, G.R. and
Warren, R.W. (1971) Phys. Rev. B, 2, 2341-2352.
51. Singleton, J.H. and Such ow, L. (1963) J. Electrochem. Soc., 110, 36-41.
52. Narita, K. and Tsuda, N. (1975) Bull. Chem. Soc. Jpn., 48, 2047-2051.
53. Apple, E.F. (1963) J. Electrochem. Soc., 110, 374-380.
54. Aia, M.A. and Poss, S.M. (1960) US Patent 2,965,786.
55. Warren, R.W., Ryan, F.M., Hopkins, R.H. and Van Broekhoven, J. (1975) J.
Electrochem. Soc., 122, 752-760.
56. Ryan, F.M. (1981) J. Lumin., 24-25, 827-834.
57. Lehmann, W., Ryan, F.M., Manocha, A.S. and McAllister, W.A. (1983) J.
Electrochem. Soc., 130,171-174.
58. Hammer, E.F. and Kaduk, E.E. (1975) US Patent 3,886,396.
59. Taubner, F.R. and Roy, J.A. (1987) US Patent 4,639,637.
60. Flaherty, J.M. (1984) US Patent 4,459,507.
62. Sigai, A.G. and Flaherty, J.M. (1987) US Patent 4,670,688.
63. Sigai, A.G. (1986) US Patent 4,585,673.
64. Sigai, A.G. (1987) US Patent 4,710,674.
65. Peters, T.E. and Hunt, R.B. (1989) US Patent 4,803,400.
66. Piper, W.w., deLuca, J.A. and Ham, F.S. (1974) J. Lumin. 8, 344-348.
67. Sommerdijk, J.L., Bril, A. and de Jager, A.W. (1974) J. Lumin., 8, 341-343.
68. limas, E.R. and Savikhina, T.I. (1970) J. Lumin., 1-2, 702.
69. Berkowitz, J.K. and Olsen, J.A. (1991) J. Lumin., 50,111-115.
71. Smets, B.M.J. and Verlijsdonk, J.G. (1986) Mater. Res. Bull., 21, 1305-1310.
11
Phosphors for other applications
G. Blasse
11.1 INTRODUCTION
Other chapters on luminescent materials in this book deal with lamp
phosphors and electroluminescent thin films. There are still many other
applications of luminescent materials. Some of these will be dealt with in
this chapter. The applications in this chapter are to a large extent in the
medical field. The excitation source is for the greater part high-energy
excitation.
First we will deal with X-ray phosphors. There are two types, those used
for direct imaging, and those which store the information obtained. The
latter in particular are at this moment in the focus of interest. Subsequently
we will deal with scintillators, large crystals which are used to detect high
energy particles.
In the two following sections we deal with luminescent compositions
which can be used in laser materials and with rare earth cryptates, an
interesting and exotic class of compounds which is promising for application
in fluorescence immunoassay. A section on cathode ray phosphors and a
conclusion end this chapter.
11.2 X-RAY PHOSPHORS

After the discovery of X-rays in 1895 Rontgen realized that X-rays are not
efficiently detected by photographic film because of the weak absorption of
X-rays by film. A search for phosphors which are able to absorb X-rays
and to convert the absorbed energy efficiently into light was started.
Rather rapidly, in 1896, Pupin proposed CaW04 for this purpose (see the
other chapter [1] by this author for its luminescence properties). It served
for some 75 years in X-ray intensifying screens, an absolute record for a
luminescent material. A medical radiography system based on the use of an
intensifying screen is represented in Fig. 11.1. In order to understand the
349
350 Phosphors for other applications
----
I
I
I
X-ray source patient intensifying screen
Fig. 11.1 Schematic representation of a medical radiography system based on the
use of an intensifying screen.
success of CaW0 4 we have first to consider the main requirements which

X-ray phosphors have to satisfy [2].
Figure 11.2 shows a film combined with two intensifying screens as used
in actual radiological applications. The X-rays are absorbed by the lumines-
cent screen and converted into light which exposes the film. In order to be
useful as an X-ray phosphor the material should at least satisfy the following
requirements: high X-ray absorption, high density, and blue or green
X-ray beam S=screen
1111111]
F=film
Cassette front
Screen base
S
~~~~~~~~~~Fluorescent
Supercoat layer
Emulsion
'I ,,\
\ i\ I I,'
F ('1\' ',~ : :: - Film base
f' "
s{
upercoat
Fluorescent layer
;-~'(' ..~"'n base
-Casslette back
Fig. 11.2 Section through an X-ray cassette containing a double-coated film (F)
and a pair of intensifying screens (S).
X -ray phosphors 351
emission with high conversion efficiency (blue and green match the spectral
sensitivity of the film).
Surprisingly enough, the successful CaW0 4 does not satisfy these
requirements too well. Its X-ray absorption is relatively low, since only
one of the six constituent atoms shows strong X-ray absorption in the
range 30-80 keY which is of medical importance. Also, its density
(6.06 g em -3) is not very high. Its broadband emission (Fig. 11.3) is hard to
use completely: a blue-sensitive film does not utilize the green part of the
emission. On the other hand, it is an inexpensive and stable material. Most
detrimental, however, is the poor X-ray to light conversion efficiency of
only 5%.
350 500nm
Fig. 11.3 X-ray excited emission spectrum of CaW0 4
Still another problem with CaW0 4 is the existence of a severe afterglow.

X-ray screens with a severe afterglow (or a long persistence) of the emission
cause the appearance of a ghost image on a subsequent exposure. The
afterglow ofCaW04 can be reduced considerably by adding NaHS0 4 to the
mixture to be fired. The other drawbacks of CaW04 are intrinsic and for
this reason research for better X-ray phosphors has continued.
The first rare earth X-ray phosphor was BaFCI:Eu2+ [3]. Further
research was carried out by Fouassier et al. [4] and Stevels and Pingault
[5]. This material has a higher X-ray absorption and a better conversion
efficiency than CaW0 4 . However, its density is lower (4.56 g em - 3) and the
morphology offers problems.
The layered crystal structure of BaFCI promotes the formation of plates.
These have a tendency for light piping towards the sides. This objection has
been overcome by applying a special preparation method, i.e. spray drying.
Figure 11.4 shows the emission spectrum of BaFCI:Eu 2 +. The maximum
of the narrow emission band is near the sensitivity peak of the blue-
sensitive film. In passing we note that the emission (at 300 K) consists of two
parts, i.e. sharp-line emission within the 4f7 configuration of Eu 2 +
(6P 7W-+ 8 S), and band emission due to the interconfigurational transition
4f 65d-+4f1.
A better X-ray phosphor with related characteristics is LaOBr:Tm 3 +.
This host has the same crystal structure as BaFCI. However, the density
of LaOBr is considerably higher: 6.13 gem- 3 Rabatin has contributed
extensive work on phosphors based on this host lattice. It has been
described in several patents (see ref. 2). Its emission spectrum is given
300 500nm
Fig. 11.4 X-ray excited emission spectrum of BaFCl:Eu2+. The peak is a 4f1
intraconfigurational transition, the band a 4f 6 5d-4f1 interconfigurational transition.
360 480nm
Fig. 11.5 X-ray excited emission spectrum of LaOBr:Tm3+.
in Fig. 11.5. It consists of intraconfigurational line transitions of Tm 3+

(4f12) in the near ultraviolet and blue spectral region. This phosphor
is nowadays a commercial product because of its superb properties.
A green-emitting X-ray phosphor of high quality is Gd 2O 2S:Tb3+
described for the first time by Tecotzky [6]. This host also has a layered
crystal structure. Using suitable preparation procedures a very nice mor-
phology can be obtained (Fig. 11.6). The density is high, i.e. 7.34 g cm - 3,
and the X-ray absorption favourable. The conversion efficiency of
Gd 20 2S:Tb3+ is 18%, nearly four times that of CaW0 4. This phosphor
emits in the green, showing mainly the sD 4- 7 F J line transitions of Tb3+. In
combination with a green-sensitive film this phosphor acts as a superior
X-ray phosphor.
It seemed as if the introduction of LaOBr:Tm3+ and Gd 20 2S:Tb3+
indicated the end of the search for new X-ray phosphors. However,
this was not the case. Later other very good green-emitting X-ray phos-
phors were found, for example GdTa04:Tb3+ [7, 8], Gd 2SiOs:Tb3+ [9]
and Gd3Gas012:Tb3+ [10]. Even more surprising was the invention
of the tantalate-based phosphors which are physically similar to CaW04
[2,7].
X -ray phosphors 353
Fig. 11.6 SEM picture of the X-ray phosphor Gd 2 0 2 S:Tb3+.
The tantalate phosphors are based on M'-YTa04' This is a modification

of YTa04 which is different from M-YTa04 (fergusonite) which is a
distorted variant of scheelite (CaW0 4). In the M' modification the Ta 5 +
ions are in six coordination. The density ofM'-YTa04 is 7.55gcm- 3 To
obtain particles useful for screen applications Li 2 S0 4 has to be used as a
flux. Figure 11.7 shows the emission spectrum of M'-YTa04. It is essentially
an ultraviolet emitter. The addition of some niobium shifts the emission to
the blue. The favourable X-ray absorption of YTa04 becomes clear from
Fig. 11.8. Conversion efficiencies of up to 9% have been obtained.
300 t.OOnm
Fig. 11.7 X-ray excited emission spectrum of M'-YTa04' The band is due to
tantalate emission; the line at 312 nm is due to emission from a Gd3+ impurity.
OL---~--~--~--~--~
30 50 70keV
Fig. 11.8 X-ray absorption of layers of CaW0 4 (1), YTa04 (2) and LuTa04 (3).
Even more interesting are phosphors based on M'-LuTa04 which is the

densest white material which is not radioactive (d=9.75 gcm- 3). Also, the
X-ray absorption is higher than for the yttrium analogue (see Fig. 11.8).
However, the high price of pure LU203 may prevent commercialization. In
conclusion the best X-ray phosphors seem to be at the moment
LaOBr:Tm3+, Gd 20 2S:Tb3+ and M'-YTa04(:Nb).
11.3 X-RAY STORAGE PHOSPHORS

About ten years ago the Japanese Fuji corporation introduced a new
technique for X-ray imaging [11]. This technique is based on the use of a
photostimulable storage phosphor screen. The operation of a storage
phosphor is depicted schematically in Fig. 11.9. On irradiation electrons are
CB ., .+-
c b 6E
1~
d
a
2-<>-
VB
Fig. 11.9 Energy band model showing the electronic transitions in a storage
phosphor: a, generation of free electrons and holes; b, electron and hole trapping; c,
electron release due to stimulation; d, recombination. Solid circles are electrons, open
circles are holes. Centre 1 presents an electron trap, centre 2 a hole trap.
X-ray storage phosphors 355
promoted from the valence band to the conduction band. In a storage
phosphor a number of the created free charge carriers are trapped in
electron traps and hole traps. The traps are localized energy states in the
bandgap due to impurities or lattice defects. If the trap depth AE is large
compared with kT, the probability for thermal escape from the trap will be
negligibly small and a metastable situation is created.
The stored energy can be released by thermal or optical stimulation. In
the case of thermal stimulation the irradiated phosphor is heated to a
temperature at which the energy barrier AE can be overcome thermally. The
trapped electron (or hole) can escape from the trap and recombine with the
trapped hole (electron). In the case of radiative recombination, luminescence
is observed, which is called thermally stimulated luminescence (TSL).
Under optical stimulation the energy of an incident photon is used to
overcome AE. The luminescence due to optical stimulation is called photo-
stimulated luminescence (PSL). The phenomenon of stimulated lumines-
cence from storage phosphors is not new [12]. Storage phosphors have
found a wide range of applications, e.g. as IR detectors and in the field of
dosimetry [13].
The X-ray imaging system based on a photostimulable storage phosphor
is depicted in Fig. 11.10. The X-ray film is replaced by a storage phosphor
screen as a primary image receptor. The transmitted X-ray photons are
absorbed by the storage phosphor in the screen, where a dose-proportional
amount of energy is stored. The latent image in the screen is read out by
scanning the phosphor screen with a focused He-Ne laser. The red laser
light (A. = 633 nm) stimulates recombination, resulting in photo stimulated
luminescence. The intensity of the photostimulated luminescence is propor-
tional to the X-ray dose. For each spot of the laser beam on the screen the
intensity of the photo stimulated luminescence is measured by a photomul-
tiplier tube and stored in a computer. The X-ray photograph in the
computer can be visualized on a monitor or by making a hard copy.
He-Ne laser
screen
hard copy
I
I
I
I
X-ray patient storage I computer
source phosphor L. photomultiplier
Fig. 11.10 Schematic representation of a medical radiography system based on the
use of a photostimulable storage phosphor.
This new technique for X-ray imaging offers several important advan-
tages in comparison with conventional screen-film radiography [11]. The
response of the systems is linear over at least four decades of the X-ray
dose (10- 2-10 2 mR). This wide dynamic range prevents overexposure and
underexposure. The sensitivity of the system is higher owing to a higher
sensitivity of a photomultiplier tube to light compared with film. The
higher sensitivity enables a reduction of the exposure time. Finally, the
digitized images obtained with this system can be processed by a com-
puter which offers the possibility of digital manipulation and facilitates
archiving.
Apart from the high price, the main disadvantage of the digital
X-ray imaging system is the lower resolution. Because of scattering
of the laser beam the performance of the system is worse than the
performance of the conventional screen-film system. This prohibits applica-
tion in areas where a high resolution is required. The development of a
translucent storage phosphor screen with minimal scattering can solve this
problem.
A good storage phosphor has to meet a number ofrequirements [11].
1. The absorption coefficient for X-rays must be high. This can be achieved
by using compounds with a high density.
2. The amount of energy stored in the phosphor per unit X-ray dose has to
be large to obtain a high sensitivity.
3. A short decay time ( < 10 Ils) of the photostimulated emission is required
for a fast readout.
4. The fading of the information stored in the phosphor has to be slow (the
information must still be present several hours after exposure).
5. Stimulation has to be possible in the red or near infrared. The stimulated
emission has to be located between 300 and 500 nm where the photomul-
tiplier tube has its highest sensitivity.
The photostimulable phosphor used at present in nearly all commercial
digital X-ray imaging systems is BaFBr:Eu 2 +. A physical mechanism for
the PSL in Eu 2 + -activated barium fluorohalides has been proposed by
Takahashi et al. [14]. Optical, EPR and photoconductivity studies indicate
that on X-ray irradiation a fraction of the holes is trapped by Eu2+ ions,
giving Eu 3 +, and a fraction of the electrons in halide vacancies, giving F
centres. Illumination in the F centre absorption band stimulates recom-
bination of the trapped electron with the hole trapped on Eu 2 +, resulting
in Eu 2 + in the excited 4f 6 5d state. The Eu 2 + ion returns to the ground
state radiatively and the characteristic Eu 2 + emission around 390 nm is
observed.
Recent developments concern two topics, i.e. a refinement of the theory
and a search for new and better storage phosphors. These developments
X -ray storage phosphors 357
are reviewed here briefly. It was soon realized that the simple scheme of
Fig. 11.9 is an oversimplification. Studies on the recombination mechanism
suggest that electron-hole recombination in BaFBr:Eu 2+ does not take
place via an electron entering the conduction band, but via tunnelling. The
evidence for this originates from different sources.
1. De Leeuw et al. [15] have measured the decay of the photostimulated
luminescence under continuous stimulation in the temperature region
100-300 K and found it to be temperature independent.
2. Von Seggern et al. [16] have measured the increase of the photo-
stimulated luminescence intensity with increasing X-ray dose. This
increase appears to be linear which points to a tunnelling process.
3. Meijerink and Blasse [17] have measured the thermally stimulated
luminescence. Analysis of these results points also to tunnelling.
An important contribution to the theoretical model formation originates
from the use of vacuum ultraviolet radiation from a synchrotron source
[18]. It was shown that photostimulable centres in BaFBr:Eu2+ can be
created by irradiation into the vacuum ultraviolet region (>6.7 eV), i.e. in
the excitonic and interband region of the host lattice.
The model originating from this work invokes as a first step the relaxation
of a free exciton in the neighbourhood of a lattice distortion induced by
Eu 2+. This leads to an e + VK centre (where e stands for electron, and VK for
the VK hole centre consisting of a Bri molecule on two Br- sites).
Subsequently an off-centre self-trapped exciton is formed, i.e. a nearest-
neighbour F-H pair (where F is the well-known F centre, in which an
electron is trapped at an anion vacancy, and H is the H centre, which can be
regarded as an Xi molecule occupying an X- anion site; the anion
concerned here is Br-). This pair is assumed to be stabilized by the presence
of a substitutional Eu2+ ion which is much smaller than Ba2+. The
photostimulable centre is then a Eu 2+-F-H complex.
Photo stimulation is thought to occur as follows: on excitation the F
centre induces a certain relaxation which destabilizes the H centre and an
excited e + VK centre is formed. The excitation energy is transferred to the
nearby Eu2+ ion and emission follows. This model accounts also for the fact
that practically no Eu 3 + has been observed after X-ray irradiation of
BaFBr:Eu2+ [15-17]. Note that the simple model of Fig. 11.9 requires the
presence of Eu 3 + after irradiation.
Many other X-ray storage phosphors have been proposed. The most
promising among these are BasSi04Br6:Eu2+ and Bas Ge04Br6:Eu2+ [17,
19]. Others are Ba2Bs09Br:Eu2+ [17], Ba40Br6:Eu2+ [20], Y2SiOs:Ce3+
[21] and RbBr:TI + [22].
Since this field is at the moment still developing, it is not possible to
indicate which material will be the choice for application in the future.
11.4 SCINTILLATO RS
Scintillator materials are used to detect high-energy radiation such as y rays.
They are used in, for example, medical diagnostics and in nuclear and
high-energy physics. A spectacular example of the latter is the use of 12000
crystals of Bi4Ge3012 at CERN (Geneva) in order to detect all types of
particles in the field of high-energy physics [23]. Examples of nuclear
diagnostics in the medical field are positron emission tomography (PET),
single-photon emission computed tomography (SPECT), and computed
tomography (CT) [24].
PET is an in vivo tracer technique which makes use of the annihilation of
positive electrons emitted by a radionuclide. This results in two almost
collinear gamma quanta with an energy of 511 keY. It is possible to measure
a three-dimensional distribution of the radioactivity by detecting the y
radiation with a scintillator. Two detectors are necessary to obtain a
three-dimensional picture.
In SPECT a radioactive substance is introduced into the body and its
gamma radiation is measured. A frequently used radionuclide is 99Tc m
(E y = 140.5 keV, tl/2=6h). In PET a radioactive isotope of C, N or 0 is
used. These are the 'biological' elements in the human body, so that
metabolic processes can be followed. These isotopes are not gamma, but
positron, emitters. Positrons (positive electrons) cannot penetrate far into
the tissue (only a few millimetres). At rest the positron annihilates with an
electron yielding two photons with 511 keY energy emitted at '" 180. These
are detected by two scintillators (Fig. 11.11). From the coincidence data,
three-dimensional images can be reconstructed. PET has several advantages
over SPECT, but the detection is very expensive.
:' Sc I
I Sc
Fig. 11.11 Principle of a PET camera. There are two scintillator crystals (Sc). The
source Sp is in the body of the patient. Note that the positron decay results in the
emission of two y quanta under an angle of _180.
In CT the transmission of X-ray radiation through the body is measured.

Using a great number of transverse transmission profiles, a three-dimen-
sional density distribution can be determined. The method is fast and
accurate, but requires a relatively high radiation dose per investigation [25].
For detection an X-ray phosphor is used.
It will be clear that all these techniques require luminescent materials
which satisfy also the requirements outlined above for X-ray phosphors. In
CT the afterglow should be very fast (after 2 ms less than 0.5% of the
Scintillators 359
maximum luminescence intensity after the X-ray dose should be present).
This restricts also the radiative decay time of the luminescent centre
0.1 ms). Proposed have been NaI:TI and CsI:TI, but their afterglow is too
long and their stability too low. Better are CdW0 4, ZnW0 4 and
Bi4Ge3012, but their efficiency is not fully satisfactory (Fig. 11.12). Recently
also rare-earth activated materials have been proposed: (Y, Gdh03:Eu3+
and Gd 2 0 2 S:Pr,Ce,F. The former has too long a decay time (1.3 ms) and
too much afterglow. The Ce and F dopants in the latter decrease the
afterglow to acceptable values.
i i
400 600nm
Fig. 11.12 X-ray excited emission spectra of crystals for use in CT: spectrum a,
NaI:TI; spectrum b, CsI:TI; spectrum c, ZnW0 4; spectrum d, Bi4Ge3012 (BGO).
Scintillators to detect gamma radiation are NaI:TI, CsI:T1, CsI:Na,

Bi4Ge 30 12 and BaF 2' Some of these have already been mentioned above.
BaF 2 crystals present a special type of luminescence. The other materials
mentioned above are of the classic type, i.e. a host lattice containing a
luminescent ion (such as TI + or Bi 3+). The gamma radiation creates a large
number of electron-hole pairs in the host lattice which ultimately recombine
on the luminescent ion, resulting in emission from this ion.
The compound BaF 2, however, behaves differently. Its emission is situ-
ated in the vacuum ultraviolet with bands peaking at 220, 195 and 179 nm
with an extremely short decay time (600 ps) [26, 27]. This high-energy
emission with short decay makes the use of a multiwire proportional
chamber possible (solid scintillator proportional counter; see for example
ref. 27). The short decay time minimizes the number of measured accidental
coincidences. This is essential since PET rests on the coincidence count rate
of the two gamma quanta originating from the annihilation of the positron.
Unfortunately the emission of BaF 2 crystals shows also a long component
(T1/2 '" 600 ns) with an emission maximum at about 300 nm.
The latter emission is due to self-trapped exciton decay and has also been
observed for other fluorides. The vacuum ultraviolet emission is of a
completely different origin. It has been ascribed to a 'cross-over transition'
[28, 29]. The explanation of this phenomenon is based on excitation
experiments using synchrotron radiation. The slow component of the
emission appears for excitation above about 10 eV, the fast component
only for excitation above about 18 eV. The bandgap of BaF 2 is about
10 eV. Energy of 18 eV excites electrons from the 5p (Ba2+) core band to
the conduction band (see Fig. 11.13). The fast emission component corre-
sponds to a recombination between a hole in the lower 5p (Ba2+) core
band and an electron in the 2p (F -) valence band. The energy spacing
between these two bands is less than between the valence and the conduc-
tion bands which is a necessary condition for emission via cross-over
transitions. The structure in the emission can be described in terms of the
density of states of the valence band. The intensity of the fast component is
temperature independent between 100 and 800 K. The cross-over transi-
tions are essentially localized and not greatly influenced by the presence of
contamination. They must be considered as intrinsic transitions. Later this
type of emission has also been observed for other halides, for example
CsCI and CsBr [30].
5PBa 2+
Fig. 11.13 Energy band model of BaF 2 showing cross-over luminescence. Eg gives
the forbidden zone between the conduction band (CB) and the valence band (2PF-)'
The core band 5PBa2+ lies below the valence band. The excitation CE results in the
fast cross-over luminescence indicated by f. The slow emission s is due to a level in
the forbidden zone.
It is not so easy to find short-wavelength emission with such a short

decay time and such a high intensity. Another possibility is the Nd3+
emission of Nd 3 + -doped fluorides [31, 32]. This emission has its maximum
at 175 nm with a decay time of a few nanoseconds. An example is
LaF 3 :Nd3+. The emission corresponds to an allowed interconfigurational
transition from the 4f 2 5d to the 4f 3 configuration of the Nd3+ ion (see
Fig. 11.14). A problem with this material is that the host lattice LaF 3
always contains a small amount of Ce 3 + and Pr 3 + impurities which act
also as luminescent centres with 5d-+4f transitions at longer wavelengths
(see Fig. 11.14).
Crystals and glasses for lasers 361
160 220 280nm
Fig. 11.14 y ray excited emission spectrum of LaF 3 :Nd3+ (0.03%). In addition to
the Nd3+ emission band at 173 nm there are also emission bands due to the
impurities Pr 3+ and Ce 3+ .
This field is developing at the moment. Large numbers of materials have

been superficially characterized [33]. On the other hand, it has been shown
that a more fundamental approach can nowadays be made [34].
11.5 CRYSTALS AND GLASSES FOR LASERS

It is not the intention of this section to treat subjects such as laser physics
and laser materials. Many thorough treatments are available in the litera-
ture [35-37]. A very short introduction can be found in many general
textbooks (see for example ref. 38). Here we wish only to draw attention to
the fact that a good laser material is always an efficient luminescent
material. As an example we mention the efficient Nd3+ luminescence on
which the four-level Nd3+ laser is based. The energy levels concerned are
given in Fig. 11.15. Laser action is obtained for the 4F 3/2 --+ 4111/2 transition
(1064 nm). Population inversion is obtained by pumping in higher levels,
x----
I.
F3/2 - - . - - - -
I.
----19/2
Fig. 11.15 Part of the energy level scheme of Nd3+. The laser action of Nd3+-
containing solids is based on the emission transition 4F 3/2 ..... 4111/2 (1.06 ~m). Excita-
tion is from the ground state 419/2 to the excited levels here indicated by X. From the
4111/2 level the system relaxes rapidly to the ground state 41 9 / 2,
whereas the 4111/2 level empties easily to the 419/2 ground level. The largest
lasers in the world use Nd3+ -doped glass rods which can be rather easily
prepared in large and perfect shape. They playa role in the nuclear fusion
experiments in the Lawrence Livermore National Laboratory (USA) [39].

The Nd 3+ - YAG laser is nowadays commercially available and well known
in the field of laser spectroscopy (Y AG = Y3Ais 12). The application of
crystalline material requires the growth of large and optically perfect
crystals.
For more recent developments we turn to the transition metal ions.
Actually the first solid state laser used ruby (A12 0 3:Cr3+ ) and was develop-
ed by Maiman [40]. Figure 11.16 gives the energy level scheme of Cr3+ as a
function of the crystal field strength LI. In ruby LI is large and laser action
occurs in the 2E-+4A2 transition. This is a three-level laser because pumping
occurs in the 4A2-+4T2 and higher transitions. The 2E-+4A2 transition is
sharp, so that the ruby laser is not tunable.
---~!::.
Fig. 11.16 The position of the lower energy levels of the Cr3+ ion as a function of
the crystal field LI. In the weak-field case the emission is the broadband 4T2-+4A2
transition (b) on which a tunable near-infrared laser can be based. In the strong-field
case the emission is the line 2E-+4A2 transition (1) on which a not-tunable three-level
red laser can be based.
In other host lattices the Cr 3+ ion may have a weaker crystal field and the
emission is due to the broadband 4T2-+4A2 transition. Using this transition
a tunable infrared laser can be designed. Suitable host lattices for this
purpose are alexandrite (BeAI 20 4) [41, 42], KZnF3 [43], ScB0 3 [44J, and
GSGG (Gd3Sc2Ga3012) [45].
Also, other transition metal ions can be used to obtain a tunable laser.
Here we mention AI 20 3:Ti3+ [35,46]. Further, the luminescence of colour
centres should be mentioned in this respect. This subject has been reviewed
by Henderson and Imbusch [35]. For further examples the reader is referred
to the general literature cited.
Crystals and glasses for lasers 363
Many efficient luminescent materials cannot be used as a laser material
because of excited state absorption (ESA). Examples are Y3Ais 0 12 :Ce3+
[47], MgAI204:Mn2+ [48], and CaW0 4 [49]. It will be obvious that it is
detrimental for laser action if the emitted radiation is reabsorbed by the
other still-excited ions.
As an example we consider Y3AIs012:Ce in more detail (see also our
earlier chapter in this book [1]). In this crystal laser action is completely
quenched by excited state absorption. In the ground state absorption
spectrum of Y3AIs012:Ce3+(4f1) the lower bands are due to allowed 4f-5d
transitions on the Ce 3+ ion. A broad emission band extends from 500 to
650 nm. The excited state absorption spectrum shows a broad band overlap-
ping this region. It has an oscillator strength of 0.03.
The excited state absorption starts on the lower 5d level and ends on a
level in the conduction band. Therefore photoionization of the Ce 3+ ion
results. The minimum of the conduction band lies only 10000 cm -1 above
the lowest relaxed 5d level of Ce 3+. Therefore the higher part of the ground
state absorption spectrum (A < 300 nm) is due to direct (one-photon) photo-
ionization of the Ce 3+ ion.
In MgAh 04:Mn2+ the situation is simpler. Excited state absorption is
observed in the visible and near-infrared spectral region. The lowest excited
state from which the emission occurs is the 4T 1 level of the d S configuration
of Mn 2+. Excited state absorption empties the emitting level and promotes
the Mn 2+ ion to higher levels of its own d S configuration. This makes laser
action based on Mn2+ luminescence practically impossible. Figure 11.17
gives an energy level scheme and some of the transitions of importance.
~ iESA
M X
Fig. 11.17 Schematic configurational coordinate diagram for Mn2+ after data by
Petermann et al. [48]. The ground state level g is 6 A b the lowest excited state e from
which the emission occurs is 4T 1. Excitation is indicated by X, emission by M. The
broken line ESA indicates excited state absorption from 4T 1 to the level e' (4T 1)'
11.6 CRYPTATES FOR FLUORESCENCE IMMUNOASSAY

In this section we would like to mention a completely different group of
compounds with promising and interesting properties, i.e. rare earth cryp-
tates [50, 51]. Certain Eu3+ -cryptates can be used in fluorescence im-
munoassay, a technique from the field of medical diagnostics. The cryptate is
bonded to an antibody which couples in a specific way with a certain
biomolecule. If the labelled antibodies find the biomolecules one is looking
for, they bind to these molecules which, after suitable treatment of the
sample, show up by the red luminescence of Eu 3 + cryptate. The technique is
very sensitive and can compete with radioactive labelling. Requirements are
that the absorption strength for ultraviolet radiation is high and that the
quantum efficiency of the Eu 3 + -crypt ate luminescence is high (until now
values of 10-20% have been obtained).
A cryptand is an organic molecule with a hole in it, so that it can act as a
cage for fitting species. In certan cryptates this species can be a rare earth
ion. Some of these are given in Fig. 11.18. The use of an organic cage around
a luminescent rare earth ion presents some interesting phenomena, from a
fundamental as well as from an applied point of view.
1. It restricts expansion on excitation (compare also our previous chapter
in this book [lJ). In this way nonradiative transitions can be suppressed.
If the rare earth metal ion fits the cage, the Stokes shift and, therefore,
the nonradiative rate are small. This has been shown experimentally for
the case of [Cec2.2.1J3+ [52].
2. The cage shields the luminescent ion from its further surroundings. This
is of considerable importance in aqueous solutions. The rare earth metal
ion emission is easily quenched if high-frequency vibrations reside in the
direct neighbourhood of the ion [51]. Water molecules present very high
vibrational frequencies and quench the greater part of the rare earth
metal ion emission. This nonradiative loss can be suppressed by caging
the ion.
3. By use of cryptands with very high absorption strength, which transfer
this energy to the rare earth metal ion, it is possible to obtain a high
luminescence output from the molecular unity. This has been realized by
using cryptands containing 2,2'-bipyridine (bipy, Fig. 11.18 [51J). Excita-
tion occurs into the bipy group, which relaxes to the lowest triplet state.
From here the energy is transferred to, for example, the Eu3+ ion, which
subsequently shows red emission. This type of cryptate is promising for
application in medical diagnostics, since their absorption strength is very
high owing to the presence of the bipy groups. The highest quantum
efficiency obtained so far has been obtained for the cryptate with two
2,2' -bipyridine units and one 3,3' -biisoquinoline-2,2' -dioxide unit, i.e.
20% at room temperature.
Phosphors for cathode ray tubes 365
Fig. 11.18 The [2.2.1] and [2.2.2] cryptands. In these the central metal ion has
been removed for clarity. This is not the case for the [Lncbpy bpy bpy]3+ cryptate.
The CH 2 groups are not indicated in these formulae (cryptand is the name of the
ligand; cryptate is the name of the whole complex).
11.7 PHOSPHORS FOR CATHODE RAY TUBES

An important application of luminescent materials is their use in cathode
ray tubes for commercial television displays (black-and-white and colour).
Conventional colour television uses a cathode ray tube with three phos-
phors which generate a blue, a green and a red picture. The definite choice
for these phosphors was made years ago: ZnS:Ag+ for blue, ZnS:Cu +, AI3+
for green, and Y2 0 2 S:Eu3+ for red [53]. They are not discussed here any
further.
In recent years new cathode ray phosphors have been developed. The
introduction of high-definition television (HDTV) has increased the interest
in projection television (PTV). In this concept the light from three mono-
chromatic cathode ray tubes is projected on a screen [54]. The principle is
given in Fig. 11.19 for one cathode ray tube. Because of the large magnifica-
tion used in imaging, the phosphor screen of the cathode ray tube must be
excited with an intense and small electron spot. The maximum excitation
density may be two orders of magnitude larger than in conventional cathode
ray tubes.
"-.3''---- ---+ L
Fig. 11.19 Projection television set (schematic). The lens system L projects the
picture at the faceplate of the tube T on the viewing screen S. The mirror M provides
a folded light path to keep the cabinet depth small.
Under such high loading the efficiency of the phosphor decreases with
increasing excitation density owing to saturation effects. This effect is much
more pronounced for the semiconductor sulphide phosphors than for the
rare-earth-activated oxides. A simple qualitative explanation of this effect
runs as follows. With increasing excitation density the number of excited
activator ions increases until all activator ions are excited. This leads to
saturation of the luminescence output. Rare-earth-activated phosphors
contain usually a few per cent of luminescent ions; in the sulphides, however,
the amount of luminescent centres is far below 1%, so that saturation (or
activator depletion) appears at much lower excitation density.
Let us consider the situation for the different colours separately. For blue
emission ZnS:Ag+ is used in conventional tubes as well as in PTV. Al-
though it suffers from strong saturation, no acceptable alternative has
been found. What compensates partly for the strong saturation is the fact
that the energy conversion efficiency of ZnS:Ag+ for cathode rays is
extremely high.
For red the phosphor Y203:Eu3+ is preferred above Y20 2S:Eu3+ in
PTV tubes. The oxide shows less saturation than the oxysulphide. Also, the
Phosphors for cathode ray tubes 367
temperature dependence of the Eu3+ emission is less in Y203 than in
Y2O 2S. This is important since the temperature of the phosphor in a PTV
tube may increase up to 100 DC.
For green many candidates have been proposed. The conventional
(Zn, Cd)S:Cu + shows an unacceptable saturation in PTV tubes. First an old,
classic luminescent material was proposed, i.e. Zn2Si04:Mn2+ (willemite). It
has a stronger saturation than Tb3+ phosphors. After Gd 2O 2S:Tb 3+ (an
X-ray phosphor, see section 11.2), the phosphors Y2SiO 5 :Tb3+ and Y3(Ga,
AI)s012:Tb3+ were proposed. Unfortunately their chromaticities lie outside
the specifications given for colour television tubes. This can be remedied
using filters. A phosphor such as InB0 3:Tb3+ has excellent properties but
its decay time is too long (7.5 ms). This is due to the fact that the
crystallographic site for Tb 3+ in InB0 3 has inversion symmetry, so that all
electric dipole transitions are forbidden.
Table 11.1 shows some data on these new phosphors. Many of these have
been studied in detail and several processes to explain the saturation have
been proposed. In addition to the model mentioned above, we mention here
increased self-absorption of the emission [55] and interaction between
excited centres [56, 57].
Table 11.1 Characteristics of phosphors for PTV tubes (taken

from ref. 53)
Phosphor '1(%) a
Y2 0 3: Eu 8 0.64 0.91
Y 2 0 2 S:Eu 13 0.40 0.68
(Zn,Cd)S:Cu + 16 0.12 1.00
Zn2Si04:Mn2+ 9 0.63 0.95
Y3(AI,Ga)s 0 12 :Tb 3+ 9 0.82 0.97
Y 2SiOs:Tb3+ 9 0.88 0.94
Gd 2O 2S:Tb 3+ 11 0.51 0.80
InB0 3:Tb3+ 8 0.76 1.00
ZnS:Ag+ 20 0.36 0.89
"'1: energy conversion efficiency in per cent in conventional cathode ray
tube.
b'1;TV: relative efficiency at excitation density characteristic of PTV
tube (the lower this value, the more pronounced is the saturation).
e'1: 00: relative efficiency at 1()() 0c.
Further, it should be realized that luminescent materials under PTV

conditions show deterioration accompanied by coloration. Of course the
efficiency drops. This aspect has been discussed recently by Yamamoto and
Matsukiyo [58]. Hintzen [59] has undertaken work to improve the proper-
ties of PTV phosphors by new and carefully controlled preparation tech-
niques, e.g. spray drying. Figure 11.20 gives some of his results. Note that
~
c:
Q)
'0
i
Q)
>
~
Qj
CI: 0.5
1 : In803 : 4% Tb. standard

2: In803 : 4% Tb}
3 : In803 : 1% Tb spray-dried
4: SC803: 1% Tb
0.0 L-_ _---1_ _ _ _ _ _'---'L-._ _ _ _ _ _I....-_ _- '
102
Current density (IJA / cm 2 ) ..
Fig. 11.20 Saturation behaviour of InB0 3 :Tb and ScB0 3 :Tb prepared by the
spray dry technique as compared with InB0 3 :Tb prepared in the standard way.
(From Hintzen [59].)
the standard InB0 3 :Tb3+ (4%) shows a stronger saturation than the sample
made by the spray-drying technique. Note further that the 4% sample has
less saturation than the 1% sample as is to be expected (see above).
Research to improve the quality of luminescent materials for use in
projection television tubes is still going on. A final decision on the optimum
choice is to be expected soon.
11.8 CONCLUDING REMARKS

From many chapters in this book it will become clear that luminescent
materials are not only interesting from a fundamental point of view but also
for their application. The use of luminescent materials in several types of
cathode ray tubes, in several types of luminescent lamps, and in medical
diagnostics and X-ray photography are nowadays well known and experi-
enced by practically everybody.
It would be an incorrect conclusion that by reading this book the
reader has a complete picture of all the applications. To mention a few
examples, he will miss the thermoluminescence dating of Roman pottery
and lunar material, he will miss the luminescent artificial teeth of im-
portance for the disco visitor, and so on. Two applications not treated
in this book we wish to mention separately, i.e. the use of luminescent
Concluding remarks 369
species as local probes and the luminescence of small semiconductor
particles.
The use of luminescent centres as probes of the nature of their surround-
ings is a well-known technique and recently new examples have been
reviewed [60].
One of these is to probe the transition from crystalline to liquid crystal-
line. Here we use the luminescence of molecules which are stacked in rows
forming long polymers. In the liquid-crystalline phase the stacking is more
favourable for energy migration to quenching centres. As a consequence the
luminescence disappears at the temperature of the phase transition.
Another is the use of Gd 3 + as a probe ion. The Gd3+ emission consists of
a sharp line at about 312nm which is accompanied by weak features due to
vibronic transitions. Their position yields the frequency of vibrations around
the Gd3+ ion. In this way it can be shown that a rare earth ion on a silica
surface is on one side directly bound to a silicate group, and on the other
side to four water molecules [61].
Finally, we mention the luminescence of small particles, especially of
semiconductors. There is no doubt that this is a fascinating development in
the field of physical chemistry, although it is too early to identify ap-
plications for these particles.
The essential point is that the physical properties of small semiconductor
particles are different from both the bulk and the molecular properties. It is
generally observed that the optical absorption edge shifts to the blue if the
semiconductor particle size decreases. This is ascribed to the quantum size
effect. This is most easily understood from the electron-in-a-box model.
Because of their spatial confinement the kinetic energy of the electrons
increases. This results in a larger bandgap.
As an example we mention here colloids of ZnS. They can be prepared
with variable particle size (down to 17 A diameter, which corresponds to
particles containing about 60 molecules of ZnS). The onset of the optical
absorption shifts from 334 nm (large particles and bulk) to 288 nm (17 A
particles). The particles show luminescence. Also, the emission maximum
shifts to shorter wavelength if the particle size is decreased [62].
A recent, well-defined example of such clusters is the existence of CdS
superclusters in zeolites [63, 64]. The authors prepared very small CdS
clusters in zeolites. In zeolite Y, for example, there are sodalite cages (5 A)
and supercages (13 A). Well-defined clusters can be made by using these
cages. The zeolites were Cd2+ ion exchanged and subsequently fired in H 2 S.
The resulting zeolite is white (note that CdS is yellow). The products were
characterized by several means. It was shown that CdS is within the zeolite
pore structure. There are discrete (CdS)4 cubes in the sodalite cages. The
cubes consist of interlocking tetrahedra of Cd and S. For high enough CdS
concentrations these clusters are interconnected. As this interconnection
proceeds, the absorption spectra shift in band edge from 290 to 360 nm.
These materials show luminescence. Three different emissions have been

observed, i.e. a yellow-green (ascribed to Cd atoms), red (ascribed to sulphur
vacancies), and blue (ascribed to shallow donors). A very interesting aspect is
that the vibrational mode responsible for the nonradiative transitions in
these materials has a frequency of 500-600 cm - 1. This is higher than the
highest phonon frequency in CdS. This indicates that interface and host
(zeolite) phonons are responsible for these processes.
No doubt these semiconductor superclusters present a novel class of
materials where the three-dimensional structure can be controlled. They
present a challenge to synthetic and physical chemists.
The history of luminescent materials is full of unexpected results with a
potential for useful application. This history continues.
REFERENCES
1. BIasse, G. (1992) In Solid State Luminescence (ed. A.H. Kitai), Chapman & Hall,
London, Chapter 2.
2. Brixner, L.H. (1987) Mater. Chem. Phys., 16, 253.
3. Chenot, D.F. (1972) Canadian Patent 896453.
4. Fouassier, c., Laourette, B., Portier, 1. and Hagenmuller, P. (1976) Mater. Res.
Bull., 11, 933.
5. Stevels, A.L.N. and Pingault, F. (1975) Philips Res. Rep., 30, 277.
6. Tecotzky, M. (1968) Electrochemical Soc. Meeting, Boston, MA, May 1968.
7. Brixner, L.H. and Chen, H.Y. (1983) J. Electrochem. Soc., 130, 2435.
8. Lammers, MJ.1. and Blasse, G. (1984) Mater. Res. Bull., 19, 759.
9. Lammers, MJJ. and BIasse, G. (1987) J. Electrochem. Soc., 134, 2068.
10. Lammers, M.J.J., Severin, 1.W. and BIasse, G. (1987) J. Electrochem. Soc., 134,
2356.
11. Sonoda, M., Takano, M., Miyahara, J. and Kato, H. (1983) Radiology, 148, 833.
12. Boyle, R. (1663) Regist. R. Soc., 213.
13. McKeever, S.W.S. (1985) Thermoluminescence oj Solids, Cambridge University
Press, Cambridge.
14. Takahashi, K., Miyahara, 1. and Shibahara, Y. (1985) J. Electrochem. Soc., 132,
1492.
15. de Leeuw, D.M., Kovats, T. and Herko, S.P. (1987) J. Electrochem. Soc., 134, 491.
16. von Seggern, H., Voigt, T., Knupfer, W. and Lange, G. (1988) J. Appl. Phys., 66,
1405.
17. Meijerink, A. and BIasse, G. (1991) J. Phys. D, 24, 626.
18. Ruter, H.H., von Seggern, H., Reiniger, R. and Saile, V. (1990) Phys. Rev. Lett.,
65,2438.
19. Meijerink, A., BIasse, G. and Struye, A. (1989) Mater. Chem. Phys., 21, 261.
20. Schipper, WJ., BIasse, G., Meijer, G. and Schleid, T. (1991) Mater. Chem. Phys.,
30,43.
21. Meijerink, A., Schipper, W.J. and Blasse, G. (1991) J. Phys. D, 24, 997.
22. von Seggern, H., Meijerink, A., Voigt, T. and Winnacker, A. (1989) J. Appl. Phys.,
66,4418.
23. Rogemond, F., Pedrini, c., Moine, B. and Boulon, G. (1985) J. Lumin., 33, 455.
24. Birks, J.B. (1964) The Theory and Practice oj Scintillation Counting, Pergamon,
New York.
References 371
25. Rossner, W. and Grabmaier, B.C (1991) J. Lumin., 48-49, 29.
26. Laval, M., Moszynski, M., Allemand, R., Cormoreche, E., Guinet, P., Odru, P.
and Vacher, 1. (1983) Nucl. Instrum. Methods, 206, 169.
27. Schotanus, P., van Eijk, CW.E. and Hollander, R.W. (1988) Nucl. Instrum.
Methods A, 269, 377.
28. Aleksandrov, Yu.M., Makhov, V.N., Rodnye, P.A., Syreischikova, TI. and
Yakimenko, M.N. (1984) Sov. Phys. Solid State, 26, 1734.
29. Valbis, YaA, Rachko, Z.A. and Yansons, Ya. (1985) JETP Lett., 42,172.
30. Valbis, YaA, Rachko, Z.A. and Yansons, Ya. (1986) Opt. Spectrosc., 60, 679.
31. Schotanus, P., van Eijk, CW.E., Hollander, R.W. and Pijpelink, 1. (1985) Nucl.
Instrum. Methods A, 238, 564.
32. Dorenbos, R., van Eijk, CW.E., Hollander, R.W. and Schotanus, P. (1990) IEEE
Trans. Nucl. Sci., 37, 119.
33. Derenzo, S.E., Moses, W.W., Cahoon, 1.L., Perera, R.CC and Litton, J.E. (1990)
IEEE Trans. Nucl. Sci., 37, 203.
34. Blasse, G. (1990) IEEE Trans. Nucl. Sci., 38, 30.
35. Henderson, B. and Imbusch, G.F. (1989) Optical Spectroscopy oj Inorganic
Solids, Clarendon, Oxford, 1989.
36. Kaminskii, A.A. (1981) Laser Crystals, Springer, Berlin.
37. DiBartolo, B. (1987) Spectroscopy oj Solid-state Laser-type Materials, Plenum,
New York.
38. Atkins, P.W. (1990) Physical Chemistry, 4th edn, Oxford University Press,
Oxford.
39. Weber, MJ. (1976) In Critical Materials Problems in Energy Production, C Stein
(ed.), Academic Press, New York.
40. Maiman, T. (1960) Nature (London), 187,493.
41. Walling, J.C (1982) Laser Focus, 18,45.
42. Walling, J.C, Petersen, O.G., Jenssen, H.P., Morris, F.C and O'Dell, E.W. (1980)
IEEE J. Quantum Electron., 16, 1302.
43. Durr, U. and Brauch, U. (1986) In Tunable Solid State Lasers II (eds A.B.
Budgor, L. Esterowitz and L.G. De Shazer), Springer, Berlin, p. 151.
44. Lai, S.T., Chai, B.H.T, Long, M. and Morris, R. (1986) IEEE J. Quantum
Electron., 22, 1931.
45. Huber, G. and Petermann, K. (1985) in Tunable Solid State Lasers (eds, P.
Hammerling, A.B. Budgor and A. Pinto), Springer, Berlin, p. 11.
46. Moulton, P.F. (1982) Opt. News, (November-December 1982),9.
47. Hamilton, D.S., Gayen, S.K., Pogatshnik, G.J. and Ghen, R.D. (1989) Phys. Rev.
B,39,8807.
48. Petermann, K., Clausen, R., Heumann, E. and Ledig, M. (1989) Opt. Commun.,
70,483.
49. Koepke, Cz. and Lempicki, A. (1991) J. Lumin., 47, 189; (1992), 54, 151.
50. Sabbatini, N., Perathoner, S., Balzano, V., Alpha, B. and Lehn, 1.M. (1987) In
Supramolecular Photochemistry (ed. V. Balzani), Reidel, Dordrecht, p. 187.
51. Sabbatini, N. (1992) Coord. Chem. Rev., to be published. Mater. Chem. Phys., 31,
13.
52. Blasse, G., Dirksen, G.J., Sabbatini, N. and Perathoner, S. (1987) Inorg. Chim.
Acta, 133, 167.
53. Welker, T (1991) J. Lumin., 48-49, 49.
54. Vriens, L., Clarke, J.A. and Spruit, J.H.M. (1989) Philips Tech. Rev., 44, 201.
55. Klaassen, D.B.M. and de Leeuw, D.M. (1984) J. Lumin., 31-32, 687.
56. van der Weg, W.F., Robertson, J.M., Zwicker, W.K. and Popma, ThJ.A. (1981)
J. Lumin., 24-25, 633.
57. van der Weg, W.F., Popma, ThJ.A. and Vink, A.T. (1985) J. Appl. Phys., 57,
5450.
58. Yamamoto, H. and Matsukiyo, H. (1991) J. Lurnin., 48-49, 43.
59. Hintzen, H.TJ.M. (1990) Thesis, University of Utrecht.
60. Biasse, G. (1990) Adv. Inorg. Chern., 35, 319.
61. Hazenkamp, M.F. and Biasse, G. (1990) Chern. Mater., 2, 105.
62. Weller, H., Koch, U., Gutierrez, M. and Henglein, A. (1984) Ber. Bunsenges.
Phys. Chern., 88, 649.
63. Wang, Y. and Herron, N. (1988) J. Phys. Chern., 92, 4988.
64. Herron, N., Wang, Y., Eddy, M.M., Stucky, G.D., Cox, D.E., Moller, K. and
Beiz, T. (1989) J. Arn. Chern. Soc., 111, 530.
Index
Accelerated charges 1, 2--4 Atomic layer epitaxy 250, 253

Acceptor-related emission lines 79 acronym 294
ACPEL, see Phosphors; Powder alteration of conventional
electroluminescence deposition 294--5
Activators chemical vapour deposition 296,
lamp phophors 318-19 301, 303
see also Phosphors, activators compound starting materials
Addressing 231 299-301
line-at-a-time 240 compounds deposited 303-10
Afterglow II-VI compounds 304--8
scintillators 358-9 III-V compounds 308-9
X-ray phosphors 351 AlGaAs 308-9
Alkaline earth cha1cogenide aluminium oxide 309
degradation 259 BaS 307-8
Aluminium gallium arsenide CaS 307-8
(AlGaAs) CdS 306--7
atomic layer deposition 308-9 GaAs 308
light emitting diode material 273, indium oxide 309
275 MgO 309-10
Aluminium gallium indium oxide dielectrics 309-10
phosphide (AlGaInP), light SrS 307-8
emitting diode material ZnO 307
275-7 ZnS 304--6
Aluminium oxide, atomic layer ZnS doped thin films 306
deposition 309 ZnSe 306
Amphoteric behaviour 79 electrochemical 295, 302-3
Anharmonicity 31 electroluminescent devices 310
Anvil cell 65 elemental starting materials 296--9
Apatite 321 equipment used 301-3
APD, see Avalanche photodiode history of 293--4
(APD) molecular beam 289-90, 293, 295,
Arsine (AsH) 282 296, 302
Atomic energy levels 16--17 phosphor thin films 293-312
Atomic layer deposition, see Atomic reaction mechanisms 296--301
layer epitaxy
373
374 Index
Atomic layer epitaxy (cont.) Chelating agents 324

starting materials Chemical vapour atomic layer
compound 299-301 deposition 296, 301, 303
elemental 296-9 Chromium ion luminescence 45
terminology 295-6 eIE Colorimetric System 315
theory 296-301 CL, see Cathodoluminescence (CL)
Auger recombination 265 Colour perception and matching 315
Avalanche photodiode (APD) 72 Computed tomography (CT) 358-9
Avalanche structure in thin films 136, Concentration quenching 39, 48, 242
137, 141, 143, 154 thin film luminescence 152, 153,
Axe theory 98, 101-2 155
Continuous flow cryostats 64
Back-to-back Zener diode circuit 139
Copper in powder phosphor
Bandgap engineering 264-5, 269
Cu-S phases 183-6
Barium sulphide deposition 307-8
dopant 178-80
Bath cryostats 64
surface conduction and 181-3
Biexcitons 54
Coulomb bond 54
Bismuth ion luminescence 48-50
Coulomb impact 136, 141, 145
Blackbody radiation 12-16
Coulomb potential 10, 17
excitation sources 62-3
Cross-over transition 360
locus 315
Cross-relaxation 38
Planck's radiation energy density
Cryostats 63-5
function 15
bath cryostats 64
B1asse, G. 21-51, 349-72
continuous flow cryostats 64
Boltzmann factor 31
Cryptand 364, 365
Boltzmann statistics 12
Cryptates for fluorescence
Born-Oppenheimer approximation 24
immunoassay 364-5
Bound exciton luminescence spectra
Crystal field
(BE) 76
dipole transitions 18-19
Breathing mode 23
dopant ions and 28
Brewster angle reflection 68
splitting 17-19
Burdick, G.W. 97-131
spin-orbit 82
t-Butylarsine (TBA) 282
strong field scheme 18
t-Butylphosphine (TBP) 282
symmetry 18
CT, see Computed tomography (CT)
Cadmium sulphide deposition 306-7 Czerny-Turner mount 67
Calcium fluorochloroapatites 321
Calcium fluoroapatite 321, 325 Dating 368
Calcium sulphide deposition 307-8 DCPEL, see Phosphors; Powder
Cataphoresis, radial 338-9 electroluminescence
Cathode ray tube phosphors 365-8 Deep optical centres 56-7, 80-1
definite choice of phosphors 365 see also Optically active centres
deterioration and colouration Defects
367-8 complex defects 57
high-definition television 365 deep level centres 56-7
projection television 365-8 self-activated emission 57
saturation effects 366 DenBaars, S.P. 263-91
Cathodoluminescence (CL) 53 Detectors 69-73
Chadha, S.S. 159-227 photodetectors 69
Char~es, accelerated 1, 2-4 photomultiplier tube 70-1
Index 375
Differential thermal analysis (DTA) Electron paramagnetic resonance

184, 323 (EPR) 323
Digital layer epitaxy 295 Electron probe microanalysis
Diodes, light emitting, see Light (EPMA), formed layer 195
emitting diodes Electron transfer quenching 34-5
Dipole-dipole energy transfer 16 Electron-hole recombination
Donor-acceptor pairs 55-6 Auger recombination 265
Donor-donor energy transfer 39-43 bimolecular recombination region
Donor fluorescence, time evolution of 267
42 excess carrier recombination rate
Donor Rydberg constant 74 267
Dopants lifetime 266
crystal field at ions 28 nonradiative 266
near band-edge emission and radiative 265-8
73-80 recombination coefficient 266
optically active centres and ions 28 spontaneous recombination 266
phosphors, see Phosphors, Emissions
activators field emission 199
self-compensation 77-9 multiphoton emission 30, 346
thin film materials 148 photon cascade emission 345
Downer, M.e. 97-131 Poole-Frenkel emission 215
DTA, see Differential thermal Schottky emission 198, 199-200
analysis (DTA) self-activated emission 57, 80
Dynamic coupling 97, 103-4 spontaneous emission 9, 10, 12
stimulated emission 9, 10, 11
thermal emission 199
Edge emitters, thin film Energy
electrophotoluminescence atomic energy levels 16-17
238-9 density 2
Efficiency exciton transition energy 54
outcoupling efficiency 149 gap 34
quantum efficiency, see Quantum stationary electron energy
efficiency distribution 142
thin films 148-54 thin films 142-3
Einstein coefficients 9-10 transfer, see Energy transfer
Electrochemical atomic layer epitaxy Energy transfer
295, 302-3 cross-relaxation 38
Electrodes, transparent 161 diffusion-limited transfer 42-3
Electroluminescence dipole-dipole energy transfer 16
excitation 53 donor-donor transfer 39-43
injection luminescence 133, 134 emission transients study 89
p-n junctions 133 energy acceptor 58
photoelectroluminescence (PEL) energy donor 58
144 migration 39, 44, 58
powder, see Powder rare earth compounds 46
electroluminescence optically active centres 36-44
thin film, see Thin film single-step donor-acceptor
electroluminescence transfer 36-9
see also Luminescence spectroscopy 58, 89-90
Electron impact excitation 313 Epilayer luminescence spectra 73-80
376 Index
Epitaxy Filtering
atomic layer, see Atomic layer absorption filters 67
epitaxy birefringent filters 67, 69
digital layer 295 interference filters 66-7
flow-rate modulation 295 multicolour thin film
halide vapour phase 281 electroluminescence devices
hydride vapour phase 280-1 245, 246-9
liquid phase 273, 286-9 Fischer's model 200-1
molecular beam 79, 289-90, 293, Flat panels, see Thin film
296, 302 electro luminescence devices
modulated 295 Flickering 319
molecular layer 295 FLN, see Fluorescence line
vapour phase 279-81 narrowing (FLN)
EPMA, see Electron probe Flocculation 335
microanalysis (EPMA) Flow-rate modulation epitaxy 295
EPR, see Electron paramagnetic Fluorescence immunoassay 364--5
resonance (EPR) Fluorescence line narrowing (FLN)
Europium 40
divalent Eu-activated phosphors Forced electric dipole 98
319 Forming process, see Phosphors
luminescence of ions 45-7 Forster's law 89
one-photon transitions 118-19 Free exciton reflection spectra (FE)
Excitation 53 76
impact excitation 313 Frenkel exciton 54
optically active centres 22
see also Excitation sources;
Gadolinium, luminescence of ions
Excitons
45-7
Excitation sources
Gadolinium compounds 44
lamps 62-3
Gadolinium one-photon transitions
lasers 60
119-21
for spectroscopy 60-3
Gallium arsenide (GaAs) deposition
Excited state absorption 363
Excitons 54--5 308
Gallium arsenide phosphide (GaAsP)
biexcitons 54
diode material 273, 278-9
bound exciton luminescence
Gallium phosphide (GaP) diode
spectra (BE) 76
material 277-8
density 54
Germicidal lamps 340-1
free exciton reflection spectra (FE)
Goppert-Mayer formula 98
76
Grating spectrometer 66-8
Frenkel exciton 54
Gumlich, H.-E. 53-96
Mott density 54
transition energy 54
Wannier-Mott exciton 54, 55 Halide vapour phase epitaxy 281
Halophosphate lamp phosphors 321-7
Harmonic perturbation 10-12
Fabry-Perot elation 62 Hayne's rule 79
Faraday configuration 66 Hel bypass flow process 252
Fick's law 284 Heisenberg uncertainty principle 5,
Field quenching, thin films 143--4, 264, 277
145, 146, 148 Helmholtz pair 66
Index 377
Heterojunctions activators 318-19

double-heterostructure absorbing antimony loss during firing 323
substrate 270, 271 blackbody locus 315
single-heterostructure absorbing colour perception and matching
substrate 270 315
High-definition television 365 colour rendition 315
Hosts converters of UV to visible light
lamp phosphors 318-19 313
thin film materials 147-8 'cool white' lamps 321-2
Hot electron cold cathode (HECC) development review 317-18
239--40 divalent Eu-activated phosphors
Huang-Rhys parameter 25 319
Hunt, R.B. Jr 313--48 electron impact excitation 313
Hydride vapour phase epitaxy 280-1 energy transfer 37
Hypermaintenance phosphor 209 flickering 319
halophosphate lamp phosphors
Image intensifiers 71-2 321-7
Impact, see Coulomb impact high pressure mercury vapour
Impact excitation 313 lamps 343-5
Inconel gasket 65 colour correcting phosphor
Indium oxide deposition 309 344-5
Inhomogeneous broadening safety considerations 343--4
concentration-induced 87-8 hosts 318-19
neighbouring impurities and 85 IR emitting materials 346
zero-photon lines 81, 85 lamp quantum yield 319-21
Injection luminescence 133, 134 low pressure mercury plasmas
light emitting diodes, see Light 313--43
emitting diodes luminous efficiency function 315
Injection mechanisms, powder manufacture of lamps, see Lamps
electroluminescence, see phosphor quantum yield 319-21
Phosphors phosphor synthesis 322--4
Insulators, luminescent centres in, see Planckian locus 315, 317, 321
Optically active centres post-synthesis processing 324
Interference filters 66-7 radiation generation 313
Iso-electronic impurities 265 rare earth phosphors 327-34
Iso-electronic traps 264, 277 blue-emitting component 332--4
green-emitting component
Jahn-Teller distortion 48 329-32
Joule heating, localized 190 inexpensive alternatives 345
Judd-Ofelt parameters 100, 111, red-emitting component 328-9
121-3 sensitizers 319
Judd-Ofelt theory 97, 98-101, 110 special phosphors 340-3
closure approximation 98-9, 128-9 specific reflectance properties 316
forced electric dipole 98 spectral power distribution 314-15,
two-photon absorption 101-2 315, 316
tricolour blend 327
Kitai, A.H. 1-19 tricolour lamps 318
triphosphor blend 327
Lamp phosphors triphosphor lamps 318
activator site preference 324-7 uses, see Lamps
378 Index
Lamp phosphors (cont.) bandgap engineering 269

vacuum UV excitation 345-6 double-heterostructure absorbing
X-ray diffraction study 323, 326-7 substrate 270, 271
see also, Lamps external quantum efficiency 268-9
Lamps growth technologies
aperture lamps 342-3 liquid phase epitaxy 286-9
coating vehicles 334-5 metal-organic chemical vapour
energy migration in 44 deposition 273, 275, 279,
erythemal lamps 341 282-86
fluocculation 335 molecular beam epitaxy 289-90
germicidal lamps 340--1 vapour phase epitaxy 279-81
lehring 336 heterojunctions 269-72, 276
maintenance 336-40 homojunctions 269-72
colour centre role 338 iso-electronic impurities 265
electric loading 339-40 iso-electronic traps 264, 277
lumen loss mechanisms 336-7 materials
protective oxides 339 AIGaAs 273, 275
manufacture 334-6 AlGaInP 275-7
pyrolysis 335-6 GaAsP 273, 278-9
radial cataphoresis 338-9 GaP 277-8
solarization 337 selection 272-9
tungsten-halogen lamps 62 wide-bandgap II-VI compounds
use of lamps 313-14 279
suntanning 314, 341-2 physics of injection luminescence
xerography 314, 342-3 263-72
see also, Lamp phosphors quantum efficiency 268-9, 275
Lanthanides radiative recombination 265-8
in aqueous solution, see see also Electron-hole
One-photon transitions recombination
lanthanide oscillator strengths 101 single-heterostructure absorbing
transitions 98 substrate 270
Lasers Liquid phase epitaxy 273, 286-9
amplifying media 60--2 advantages 287
crystals and glasses for 361-3 constraints 287-8
excitation source 60--2 disadvantages 287
excited state absorption 363 limiting factor during 287
frequency doubling or tripling 62 step cooling and uniform cooling
nuclear fusion experiments 362 288-9
population inversion 361 Luminescence
pulsed 62 electron, see Electroluminescence
pumping 62 excitation of emission 53
transition metal ions 362 injection luminescence 133, 134
tunable 21, 60, 62, 362 see also Light emitting diodes
typical data 61 luminescent centres, see Optically
LEDs, see Light emitting diodes active centres
Lehring 336 photoelectroluminescence (PEL)
Ligand polarization 104 144
Light emitting diodes 263-91 principles of, see Luminescence
band-gap material, direct and principles
indirect 264-5 self-activated 80
Index 379
spectroscopy, see Spectroscopy see also Lamp phosphors; Lamps

time evolution of donor Metal~organic chemical vapour
fluorescence 42 deposition 79, 273, 275, 279,
Luminescence principles 1~ 19 282~6
accelerated charges 1, 2--4 mass transport 284~5
blackbody radiation 12~16 regions of growth 284
crystal field splitting 17~ 19 Micro-channel plate (MCP) 71~2
dipole-dipole energy transfer 16 Migration of energy, see Energy
energy levels in atoms 16-17 transfer
harmonic perturbation 10-12 MOCVD, see Metal-organic
nonradiative transitions 29~34 chemical vapour deposition
quantum description 5~6 Molecular beam epitaxy 79, 289~90,
Einstein coefficients 9~ 10 293, 302
forbidden transitions 18~ 19 modulated 295
parity selection rule 28 Molecular layer epitaxy 295
selection rules 7~9, 28 Mossbauer spectroscopy 326
spin-orbit coupling 17, 28 Mott density 54
radiation theory 1--4 Mott~Schottky barrier 201
simple harmonic radiator 4-5 Miiller, G.O. 133~57
Luminescent probes 368~9 Multiphoton emission 30, 346
Multiple interference Fabry~Perot
Mach, R. 229~62 elation 62
Magic-angle spinning~nuclear
magnetic resonance 327 Near band-edge emission 73~80
Magnesium oxide deposition 309~ 10 bound exciton luminescence
Magnetic resonance spectra (BE) 76
electron paramagnetic resonance free exciton reflection spectra (FE)
(EPR) 323 76
magic-angle spinning~n uclear self-compensation 77~9
magnetic resonance 327 Negative differential resistance (NDR)
Manganese 195
as powder phosphor activator 174~ Neodymium one-photon transitions
8 113~15
site selection spectroscopy 81~3, Nitroprussides, light-induced
85~6 metastable states in 28~9
time-resolved spectroscopy 87~9 Nonradiative transitions, optically
Zeeman splitting of emission 90-1 active centres, see Optically
Manneback recursion formulae 31 active centres, nonradiative
Materials selection, see individual transitions
materials and applications, Nuclear magnetic resonance, magic-
e.g. Zinc sulphide; Light angle spinning 327
emitting diodes
MBE, see Molecular beam epitaxy OMAs, see Optical multichannel
MCP, see Micro-channel plate analysers (OMAs)
(MCP) One-photon transitions
Medical radiography, X-ray Axe theory 98, 101~2
phosphors 349~54 closure approximation 128~9
Mercury lamps dynamic coupling 97, 103--4
high pressure 343~5 europium ions 118~19
low pressure 313--43 forced electric dipole 98
380 Index
One-photon transitions (cont.) between identical ions 38-9

g orbital terms 123-6 cross relaxation 38
gadolinium ions 119-21 diffusion-limited transfer 42-3
general theory of absorption 97- donor-donor transfer 39-43
104 migration 39, 44
ludd-Ofelt parameters 121-3 rare earth compounds 46
ludd-Ofelt theory 97, 98-101, 110, single-step donor-acceptor
128-9 transfer 36-9
lanthanide fitting 121-3 excitation 22
lanthanides in aqueous solution harmonic approximation 23
112-26 heat generation 22
see also individual elements historical study 21
ligand polarization 104 light-induced metastable states in
neodymium ions 113-15 nitroprussides 28-9
polarization dependence 127 literature 22-3
pseudoquadrupole transitions 104 no-photon transition, see zero-
samarium ions 115-18 vibrational transition
Stark levels 112-13 nonradiative return to ground
symmetry adaption 127 state 22
tensors 107-8, 112 nonradiative transitions 29-34
spherical Racah tensors 106 energy gap law 34
third-order contributions 102, 103 magnitude of rate 31
importance of 126-7 radiative rate 32
third-order terms derivation 104-7 temperature dependence 31
g orbital contributions 110-11 optical absorption 24
k = 1 terms 108-10 parity selection rule 28
rank 3 and 5 terms 112 properties of bands or lines 28
tensors 106, 107-8 quantum efficiency 22, 29
trivalent lanthanides 113-21 quenching
two-photon absorption 101-2 centre 22
Wigner-Eckart theorem 99 concentration quenching 39, 48
Optical multichannel analysers electron transfer 34-5
(OMAs) 71-2 photoionization 35-6
Optically active centres 21-51 radiative return to ground state 22
absorption transition 24-5 rare earth ions 45-7
anharmonicity 31 rubies 21, 45
bismuth ions 48-50 sensitized emission 22
Born-Oppenheimer approximation Stokes shift 22, 25, 29, 35, 48, 49
24 Struck-Fonger approach 31, 33
breathing mode 23 symmetrical stretching mode 23
case studies 44-50 temperature quenching 30, 31,
see also individual ions, e.g. 32-3
chromium ions tungstate group 47-8
chromium ions 44-5 zero-vibrational transition 25, 27
configurational coordinate diagram Orbach process 57
23-36 Outcoupling efficiency 149
deep level centres 56-7, 80-1
dopants in solids 28 p-n junctions 133
emission transition 24, 25-6 Pappalardo, R.G. 313-48
energy transfer 36-44 Parity selection rule 28
Index 381
Perturbation 98 Fischer's model 200--1

harmonic 10--12 minority carrier injection 200
time-dependent theory 10 lamp phosphors, see Lamp
time-independent theory 18 phosphors
PET, see Positron emission optoelectrical response
tomography (PET) ACPEL cells 201-3
Peters, T.E. 313--48 DCPEL devices 203
Phase sensitive detection 72 precipitation preparation methods
Phosphors 165-86 ACPEL phosphors 167-8
activators 170--86 DCPEL phosphors 168-70
copper in ACPEL phosphors preparation of II-VI compounds
172--4 166--9
copper in DCPEL phosphors quenchers 170
Cu-S phases 183-6 Schottky emission 199-200
dopant 178-80 surface analytical techniques 191
surface conduction and 181-3 thermal emission 199
doping of ACPEL phosphors X-ray phosphors 349-54
171 X-ray storage phosphors 354-7
doping of DCPEL phosphors ZnS
171-2 DCPEL history 162-5
manganese 174-8, 200 low and high temperature forms
cathode ray tubes, see Cathode ray 166
tube phosphors see also, Lamp phosphors; Powder
dopants, see activators electroluminescence
electron probe microanalysis 195 Photocells 69-70
excitation mechanisms 198-200 Photoelectroluminescence (PEL) 144
Mn centre direct excitation 200 Photoionization, quenching by
fabrication 35-6
ACPEL cells 186 Photoluminescence (PL)
DCPEL devices 186--8 excitation 53
see also forming process spectroscopy, see Spectroscopy
field emission 199 Photomultiplier tube 70--1
formed layer Photon cascade emission 345
constant voltage across 194-5 Photon emission rate 6
negative differential resistance Photon release times 9
(NDR) 195 Photon splitting 345
reciprocal capacitance 195 Photons
stuck devices 196 one-photon transitions, see One
forming process 187, 188-98 photon transitions
characteristics in DCPEL after single photon counting technique
192, 194 72-3
failure modes and 194 stimulated emission initiated by 9
localized Joule heating 190 Photostimulable X-ray storage
migration 191, 192 phosphor screen 354, 355,
models of 196--7 357
moisture in 196 Pile-up errors 73
surface environment role 196 PIN diode 72
hypermaintenance phosphor 209 PL, see Photoluminescence (PL)
injection mechanisms 200--1 Planckian locus 315, 317, 321
double injection 200--1 Planck's constant 277
382 Index
Planck's radiation energy density device failure modes 212-14
function 15 electron injection mechanisms
Pohl, U.W. 53-96 215-17
Poisson distribution 55 encapsulation technique
Polarization improvement 211, 215
dependence in one-photon further forming 212-13, 214-15
transitions 127 impedance increase 210-11
ligand polarization 104 increase in resistance 206
Polarizers 68-9 lack of stoichiometry 218
Glan-Thompson prism polarizers load line instability 214
69 softening 213, 215
Polaroid film polarizers 69 phosphors, see Phosphors
Poole-Frenkel emission 215 photon production mechanism 217
Positron emission tomography (PET) quantum efficiency 217
358 TV purposes 206
Powder electroluminescence 133, Poynting vector 4
159-227 Precipitation of phosphors
ACPEL 159 ACPEL 167-8
history 160-2 DCPEL 168-70
alphanumeric and graphical milling 169
display 205, 220 Probes 368-9
application areas Projection television 365-8
ACPEL cells 219-20 Pumping 62
alphanumeric and graphical Pyrolysis 335-6
display 220
backlighting 219
car dashboards 220, 221, 222 Quantum description of luminescence
DCPEL devices 220 5-6
dealer room applications 220, Einstein coefficients 9-10
222 forbidden transitions 9
fixed legends 219 orthogonality property 8
laptop computers 220, 222 parity selection rule 28
medical scanners 220, 222 selection rules 7-9, 28
plasma panels 219 spin-orbit coupling 8, 17, 28
backlighting 204 Quantum efficiency 29
DCPEL 159 light emitting diodes 268-9, 275,
history 162-5 277
short-lived cells 163 optically active centres 22
decay in brightness, see powder electroluminescence 217
maintenance of devices powder electrophotoluminescence
drive circuitry 217
ACPEL cells 204, 205 thin film electroluminescence 149
DCPEL devices 204-7 Quantum yield 319-21
hypermaintenance phosphor 209 Quenching
maintenance of devices centre 22
ACPEL cells 207-9 concentration quenching 39, 48,
constant current drive usage 218 242
DCPEL panels 210-19 thin film luminescence 152, 153,
decay in brightness 207-9, 210 155
degradation 210, 212 electron transfer quenching 34-5
Index 383
field quenching in thin films 143--4, Schrodinger equation 16-17

145, 146, 148 Scintillators 358-61
photoionization quenching 35-6 afterglow 358-9
spectroscopy 58 cross-over transition 360
temperature quenching 30, 31, CT 358
32-3 gamma radiation detection 359
PET 358
Racah tensors 106 short-wavelength emission 360
Radiation SPECT 358
blackbody radiation 12-16 vacuum UV emission 359-60
intensity 4 Self-activated emission 57, 80
see also Radiation theory Self-compensation 77-9
Radiation theory 1--4 Semiconductor detectors
simple harmonic radiator 4-5 avalanche photodiode 72
synchrotron radiation source 5 phase sensitive detection 72
Raman process 57 pile-up errors 73
Raman scattering, resonant 75 PIN diode 72
Raman transitions 110 single photon counting technique
Rare earth phosphors 327-34 72-3
blue-emitting component 332--4 time-to-amplitude converter (TAC)
green-emitting component 329-32 73
for lamps, inexpensive alternatives Semiconductor materials, LEDs, see
red-emitting component 328-9 Semiconductors 369
Rare earths superclusters 369-70
cryptates 364-5 Simple harmonic radiator 4-5
deep level centres 56-7 SIMS analysis 150
energy migration 46 Single-photon emission computed
luminescence of ions 45-7 tomography (SPECT) 358
one-photon transitions, see One- Site-selection spectroscopy 80-7
photon transitions deep centres 80-1
X-ray phosphors 351--4 Mn 81-3, 85-6
see also individual elements, e.g. principle demonstration 81
Europium zero-photon lines 81, 82-3
Recombination of electron-hole, see Solarization 337
Electron-hole recombination SPECT, see Single-photon emission
Refractory oxides 318-19 computed tomography
Refrigerators 64-5 (SPECT)
see also Cryostats Spectrometers 66-8
Resonant Raman scattering 75 Czerny-Turner mount 67
r.f. sputtering 254 grating spectrometer 66-8
Ruby, optically active centres 21, 45 Wood grating anomaly 68
Rydberg constant 74 Spectroscopy 53-96
acceptor-related emission lines 79
Samarium one-photon transitions bound exciton luminescence
115-18 spectra (BE) 76
Sanders, Brian W. 293-312 deep level centres 56-7
Scanning electron microscopy 191 donor-acceptor pairs 55-6
Schottky barrier 215 energy transfer 58, 89-90
Schottky emission 198, 199-200, 217 epilayers 73-80
384 Index
Spectroscopy (cant.) Spherical Racah tensors 106

equipment Spin-orbit coupling 8, 17, 28
anvil cells 65 Spin-orbit splitting 82, 83
basic set-up 59 Spin selection rule 28
cryostats 63-5 Splitting
detectors 69-73 spin-orbit splitting 82, 83
excitation sources 60-3 valence band splitting 55
external fields application 65-6 Spontaneous emission 9, 10, 12
filters 66--7 Sputtering 254
image intensifiers 71-2 Stark effect 65
Inconel gasket 65 Stark levels 112-13
interference filters 66--7 Stationary electron energy
operation principles 59-66 distribution 142
optical multichannel analysers Steradian sulphide (SrS) deposition
(OMAs) 71-2 307-8
polarizers 68-9 Stimulated absorption 11
semiconductor detectors, see Stimulated emission 9, 10, 11
Semiconductor detectors Stokes' law 59
signal processing 69-73 Stokes shift 22, 25, 29, 35, 48, 49
spectrometers 66--8 Strain probing 74, 75
temperature controlling 63-6 Struck-Fonger calculation 31, 33
types of spectra measured 59-60 Stuck devices 196
excitation sources 60-3 Suntanning lamps 314, 341-2
lamps 62-3 Symmetrical stretching mode 23
lasers 60-2 Symmetry adaption 127
tungsten-halogen lamps 62 Synchrotron radiation 346
excitons 54-5 source 5
experimental equipment, see
equipment
free exciton reflection spectra (FE) T AC, see Time-to-amplitude
76 converter (T AC)
hydrostatic pressure 65 Television displays
information from spectra 54-8 high-definition television 365
inhomogeneous broadening 57 projection television 365-8
laser use 60-2 see also Cathode ray tube
magnetic field 65 phosphors
near band edge emission, see Near Temperature control
band-edge emission cryostats 63-5
pressure measurements 90-4 refrigerators 64-5
quenching phenomena 58 spectroscopy 63-6
self-compensation 77-9 Tensors 107-8, 112
site-selection spectroscopy 80-7 rank 3 and 5 terms 112
Stark effect 65 spherical Racah tensors 106
time-resolved spectroscopy 87-9, Terbium ion luminescence 45-7
90 TFELDs, see Thin film
types of spectra measured 59-60 e1ectroluminescence devices
uniaxial stress 65, 66 TGA, see Thermogravimetric analysis
Zeeman effect 65, 66, 90-4 (TGA)
zero-photon lines 55, 56--7, 65 Thermal evaporation deposition 250
Spherical harmonics 18 Thermal quenching 30, 31, 32-3
Index 385
Thermogravimetric analysis (TGA) physical limitations 135-6

323 physical structure 135
differential 184 powder electroluminescence 133
Thermoluminescence (TL) SIMS analysis 150
dating 368 specular reflections 134
excitation 53 stationary electron energy
Thin film electroluminescence 133-57 distribution 142
applicational aspects 134-5 surface roughness 150, 152
avalanche structure 136, 137, 141, transparency advantage 134
143, 154 Thin film electrolumiriescence devices
brightness 154 229-62
concentration quenching 152, 153, addressing 231
155 all solid state 229
Coulomb impact 136, 141, 145 costs 230, 253
depth profiling 150 degradation 256-9
devices, see Thin film alkaline earth chalcogenides 259
electroluminescence devices ZnS:Mn forced aging 257-9
doping with activators 143 ZnS:Tb 259
efficiency 148-54 deposition of films 250-6
energy efficiency 152 displays 230-6
inner luminous efficiency 152 addressing 231
maximum attainable 152 disadvantages 232-3
outcoupling efficiency 149 driving 231
outer luminous efficiency 153 electronics around displays 234,
partial efficiencies 152 236
quantum efficiency 149 high capacitances 232-3
energy filter 142-3 large scale production 233-4
field quenching 143-4, 145, 146, location of electrodes 233
148 response time 231, 232
field stability 144 simplicity of structure 232
film deposition 144 using two transparent electrode
history 133-4 systems 236
impact 136, 141, 145 edge emitter 238-9
manufacture full colour, see multicolour, full
deposition of active films 250-2 colour
HCl bypass flow process 252 HCl bypass flow process 252
materials 155 hot electron cold cathode 240
dopants 148 light sources 236-7
host materials 147-8 materials 252-3, 254-5
mechanism 137-47 monochrome
back-to-back Zener diode circuit ZnS:Ln 243-4
139 ZnS:Mn 241-3
ballistic transport 141-3 multicolour, full colour 244-9
electroluminescence 137-41 colour by filtering 245, 246-9
electroluminescence 143-7 limitation 245
impact excitation 230 patterned phosphor 245, 246
loss-free acceleration 142 technology 250-6
tunnel emission 138, 142, 144 atomic layer epitaxy 250, 253
photoelectroluminescence (PEL) CVD 252, 253
144 EBE 250, 253
386 Index
Thin film electroluminescence Voigt configuration 66

devices, technology (cant.) VPE, see Vapour phase epitaxy
HCl bypass flow process 252
multisource evaporation methods Wannier-Mott exciton 54, 55
254 Wide-bandgap II-VI materials 272
rJ. sputtering 254 Wigner-Eckart theorem 99
white emission 247-8
broad band 249-50 X-ray diffraction 323, 326-7
Thin film electrophotoluminescence X-ray microanalysis 191
addressing, line-at-a-time 240 X-ray phosphors 349-54
edge emitters 238-9 afterglow 351
hot electron cold cathode 239-40 rare earths 351-4
Thin films, atomic layer epitaxy, see requirements 350-1
Atomic layer epitaxy X-ray storage phosphors 354-7
Thulium dopant 244 advantages 356
Time-dependent perturbation theory disadvantage 356
10 photostimulation 354, 355, 357
Time-independent perturbation release of stored energy 355
theory 18 requirements 356
Time-resolved spectroscopy 87-9, 90
Time-to-amplitude converter (T AC) Zeeman effect 65, 66
73 Zeeman levels 91, 92
TL, see Thermoluminescence (TL) Zeeman spectra 90-4
Tomography Zener diode model 139
computed tomography 358-9 Zener effect 199
positron emission tomography Zero-photon lines 55, 56-7, 81, 82-3
(PET) 358 inhomogeneous broadening 81, 85
single-photon emission computed spectroscopy 65
tomography (SPECT) 358 Zinc oxide (ZnO), atomic layer
Transition metals, deep level centres deposition 307
56 Zinc selenide (ZnSe)
Transitions, see individual types oj deposition 306
transition, e.g. Nonradiative near band-edge emission and
transitions dopants 73-80
Transparent conducting electrodes Zinc sulphide
161 atomic layer deposition 304-6
Triethylgallium (TEGa) 282 doped thin films 306
Trimethylaluminium (TMAl) 282 colloids of 369
Trimethylgallium (TMGa) 282 powders, see Phosphors
Tungstate group, luminescence of thin films 294
ions 47-8 Zinc sulphide:lanthanides (Zn:Ln)
Tungsten lamps 243-4
blackbody radiation from 16 Zinc sulphide:manganese (Zn:Mn)
tungsten-halogen lamps 16, 62 thin films 143-6
forced aging 257-9
Valence band splitting 55 monochrome devices 241-3
Vapour phase epitaxy Zeeman levels in crystal 90-4
halide 281 Zinc sulphide:tellurium (Zn:Te) thin
hydride VPE 280-1 films 144
light emitting diodes 279-81 Zinc sulphide:terbium (Zn:Tb) thin
Vidicon TV camera technology 71 films 259

A. H. Kitai (Auth.), A. H. Kitai (Eds.) - Solid State Luminescence - Theory, Materials and Devices-Springer Netherlands (1993)

Enviado por

Dados do documento

Título original

Direitos autorais

Formatos disponíveis

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formatos disponíveis

A. H. Kitai (Auth.), A. H. Kitai (Eds.) - Solid State Luminescence - Theory, Materials and Devices-Springer Netherlands (1993)

Enviado por

Direitos autorais:

Formatos disponíveis

Solid State Luminescence

Solid State Luminescence

I~nl SPR1NGER-SCIENCE+BUSINESS MEDIA, B.V.

Chapman & Hali, 2-6 Boundary Row, London SEI 8HN, UK

First edition 1993

@! Printed on permanent acid-free text paper, manufactured in accordance

List of contributors Vll

Chapter 7 presents the current state of thin film electroluminescence, and

Blackbody radiation Tungsten filament lamp -5%

1.2 RADIATION THEORY

Fig. 1.1 The lines of electric field E due to a point charge q.

Fig. 1.4 Lines of magnetic field B emanating from an accelerating charge. B is

where f is a unit radial vector.

since dA is a ring of area 2nr 2 sin 0 dO.

1.3 SIMPLE HARMONIC RADIATOR

1.4 QUANTUM DESCRIPTION

where a and b are time-dependent coefficients. Initially, a = 1, b = and

provided I/In is normalized, and V represents all space. Since, by definition

where n is the spatially dependent part of "'n' Hence

since the position must be a real number. This may be written as

One photon of this wavelength has energy E = hel A= 3 x 10 - 19 1. Hence, the

1.5 SELECTION RULES

Note that since we are working in three dimensions, we must consider r in

r nn ,= tXl Rn,(r)r Rn(r)dr[t

The term in brackets may be broken up into orthogonal components of unit

since tPm()=e im </>, the three integrals in may be written

13 is zero unless m'=m.

which is zero unless m' = m 1. 12 gives the same result.

is zero unless l' = 1, a property of the associated Legendre polynomials

Using the properties of associated Legendre polynomials once again, we

For a non-zero result, l' = 1 1 using the orthogonality property. The

1.6 EINSTEIN COEFFICIENTS

In order to begin making such transitions, something is needed to perturb

The proportionality constants A and B are called the Einstein A and B

1.7 HARMONIC PERTURBATION

Time-dependent perturbation theory [2J may be used to determine the

= 2iHkl {e i (o)kl- W )t / 2 sin[(wkl-w)t/2] + ei (Wkl+ W )t / 2 Sin[(W k1 +W)t/2]}

The probability of the transition (stimulated emission or absorption) is

-61t -41t -21t 21t 41t 61t

Since energy density u( v) is proportional to 1E 12, it is clear that

1.8 BLACKBODY RADIA nON

Since the populations of atoms having excited states of certain energies

where subscripts have been dropped.

Consider a cavity with metallic walls uniformly heated to temperature T.

E(x, t)=Eo Sine~x) sin(2;rrvt) where v=l

To satisfy boundary conditions, E(a, t)=O and therefore

space having axes (n x , ny, nz ). Such a space consists of a lattice of points,

n, we divide by 8 to count only one octant of the shell, and multiply by 2

N(v)= 8na 3 v2 dv= 8nV v2 dv

() I' 8nv 23kT

we can now evaluate AlB in equation (1.6) by requiring that

I. AlB AlB 8nv 2 kT

This is shown in Fig. 1.9 for three different temperatures.

1.9 DIPOLE-DIPOLE ENERGY TRANSFER

1.10 ENERGY LEVELS IN ATOMS

atom in a crystal surrounded by a space lattice of atoms. The total

involves a Coulomb potential due to the atom nucleus and appro-

1.11 CRYSTAL FIELD SPLITTING

octahedral site will have octahedral symmetry. The electric potential of an

where yr are the spherical harmonics. This is analogous to a Fourier series

Fig. 2.1 Luminescence processes in a centre A in a solid: X, excitation; M, emission;

complex ion) in a solid or a liquid. The centre is irradiated. For simplicity

2.2 THE CONFIGURATIONAL COORDINATE DIAGRAM

ug. ~:: ~=-~Ii'~~~~/.:i~