40 1 Atoms: general aspects

Appendix I.2 Perturbation effects in two-levels system

We shall refer to a model system with two eigenstates, labeled |1 > and
|2 >, and correspondent eigenfunctions φ01 and φ02 forming a complete or-
thonormal basis. The model Hamiltonian is H0 and H0 φ0m = Em φ0m , with m =
1, 2. In real systems the Hamiltonian H can differ from H0 owing to a small
perturbation HP . Following a rapid transient (after turning on the perturba-
tion) the stationary states are described by eigenfunctions that differ from the
ones of the model system by a small amount, that can be written in terms of
the unperturbed basis. This is equivalent to state that the eigenfunctions of
the equation
Hφ = Eφ (A.I.2.1)
are
φ = c1 φ01 + c2 φ02 (A.I.2.2)
with c1 and c2 constants. By inserting φ in A.I.2.1 and multiplying by < φ01 |
and by < φ02 | in turn, in the light of the orthonormality of the states, one
derives for c1,2

c1 (H11 − E) + c2 H12 = 0
c1 H21 + c2 (H22 − E) = 0

with Hmn =< m|H|n >. Non-trivial solutions imply

! "
H11 − E H12
det =0
H21 H22 − E

and the eigenvalues turn out

1 1#
E∓ = (H11 + H22 ) ± (H11 − H22 )2 + 4H12 H21 (A.I.2.3)
2 2
When the diagonal elements of HP are zero, A.I.2.3 reduces to
1 1#
E∓ = (E1 + E2 ) ± (E1 − E2 )2 + 4ε2 (A.I.2.4)
2 2
where ε2 = | < 2|HP |1 > |2 , HP being Hermitian.
The perturbation effects strongly depend on the energy separation ∆E =
E2 − E1 . For degenerate energy levels (∆E = 0) the largest shift of the levels
occurs, given by 2ε. For perturbation much weaker than ∆E, Eq.A.I.2.4 can
be expanded, to yield the second-order corrections

ε2
E∓ = E2,1 ± (A.I.2.5)
∆E
 Appendix I.2 Perturbation effects in two-levels system 41

The corrections to the unperturbed eigenvalues as a function of ∆E are

illustrated below

≈ε /∆E
Energy

E E
ε ∆E=E -E
ε

E E

≈-ε /∆E

The eigenfunctions in the presence of HP can be obtained by deriving the

coefficients c1,2 in A.I.2.2 in correspondence to E = E+ and E = E− . For
widely separated unperturbed states one obtains
H12 0 H12 0
φ+ " φ01 − φ , φ− " φ02 + φ
∆E 2 ∆E 1
while for degenerate eigenstates
1 H12 0 1 H12 0
φ+ = √ (φ01 + φ ), φ− = √ (φ01 − φ ) (A.I.2.6)
2 |H12 | 2 2 |H12 | 2

Now we turn to the time evolution of the system, by considering two

cases. The first is the evolution of the system after a static, time-independent
perturbation has been turned on, the second (to be discussed as Appendix
I.3) when a periodic time-dependent perturbation is applied.
To deal with the time dependence one has to refer to the complete unper-
turbed eigenfunctions and to the time-dependent Schrodinger equation:
∂ψ
[H0 + HP (t)]ψ = ih̄ (A.I.2.7)
∂t
The eigenfunction A.I.2.2 is now written with time dependent coefficients
ψ = c1 (t)ψ10 + c2 (t)ψ20 (A.I.2.8)
with |c1 |2 + |c2 |2 = 1. Let us assume the initial condition c1 (t = 0) = 1
and c2 (t = 0) = 0. The probability that at the time t after turning on the
perturbation the system is found in the state |2 > is given by
42 1 Atoms: general aspects

P2 (t) = |c2 (t)|2 (A.I.2.9)

The equation for c2 (t) is obtained by inserting A.I.2.8 into A.I.2.7. Recall-
ing that
0
∂ψ1,2
0
H0 ψ1,2 = ih̄
∂t
one has
dc1 dc2
HP (c1 ψ10 + c2 ψ20 ) = ih̄(ψ10+ ψ20 ) (A.I.2.10)
dt dt
By multiplying this equation by (ψ10 )∗ , integrating over the spatial coordinates
and by taking into account that ψ1,2
0
(t) = φ01,2 exp(−iE1,2
o
t/h̄) one finds
dc1
c1 < 1|HP |1 > +c2 < 1|HP |2 > e−iω21 t = ih̄ (A.I.2.11)
dt
$
where ω21 = (E20 − E10$)/h̄; < 1|HP |1 >= H11 ≡ (φ01 )∗ HP φ01 dτ and
< 1|HP |2 >= H12 ≡ (φ01 )∗ HP φ02 dτ are the matrix elements of the per-
turbation between the stationary states of the unperturbed system 5 .
In analogous way, from A.I.2.10, multiplying by (ψ20 )∗ one derives
dc2
c1 H21 e+iω21 t + c2 H22 = ih̄ (A.I.2.12)
dt
In order to illustrate these equations for c1,2 , let us refer to a perturba-
tion which is constant in time, with no diagonal elements. Then (HP )11 =
(HP )22 = 0 and (HP )12 = h̄Γ, (HP )21 = h̄Γ ∗ . Eqs. A.I.2.11 and A.I.2.12
become
dc1 dc2
= −iΓ e−iω21 t c2 = −iΓ ∗ e+iω21 t c1
dt dt
By taking the derivative of the second and by using the first one, one has

d2 c2 dc2
2
= iω21 − c2 Γ 2
dt dt
of general solution
iω21 t
c2 (t) = (AeiΩt + Be−iΩt )e 2

#
with Ω = (1/2) ω21 2 + 4Γ 2 . The constants A and B are obtained from the

initial conditions already considered, yielding

iΓ iω21 t
c2 (t) = − sinΩte 2
Ω
5
In the Feynman formulation the coefficients ci =< i|ψ(t) > are the ampli-
tudes
% that the system is in state |i > at the time t and one has ih̄(dci /dt) =
j
H ij (t)cj (t), Hij being the elements of the matrix Hamiltonian.
 Appendix I.3 Transition probabilities and selection rules 43

and therefore
#
4Γ 2 2 ( ω21 + 4Γ )t
2 2
P2 (t) = |c2 (t)| = 2
2
sin , (A.I.2.13)
ω21 + 4Γ 2 2

known as Rabi equation. P1 (t) = 1 − P2 (t).

It is worthy to illustrate the Rabi equation in the case of equivalent states,
so that E10 = E20 . We shall refer to such a situation in discussing the molecular
Hydrogen ion H+ 2 where an electron is shared between two protons (§8.1).
Then ω21 = 0 and Eq. A.I.2.13 becomes

P2 (t) = sin2 Γ t (A.I.2.14)

namely the system oscillates between the two states. After the time t = π/2Γ
the system is found in state |2 >, even though the perturbation is weak. For
H+2 one can say that the electron is being exchanged between the two protons.
For widely separated states so that ω212
% 4Γ 2 Eq. A.I.2.13 yields

2Γ 2 2 ω21 t
P2 (t) = ( ) sin (A.I.2.15)
ω21 2
predicting fast oscillations but very small probability to find the system in
state |2 >.
Pulse resonance techniques (see Chapter 6) can be thought as an applica-
tion of the Rabi formula once that the two spin states (spin up and spin down
in a magnetic field) are “forced to become degenerate” by the on-resonance
irradiation at the separation frequency (E20 − E10 )/h.
In the presence of a relaxation mechanism driving the system to the low-
energy state, a term −ih̄γ (with γ the relaxation rate) should be included in
the matrix element H22 . In this case, from the solution of the equations for
the coefficients c1,2 (t) the probability P2 (t) corrects Eq. A.I.2.14 for the Rabi
oscillations with a damping effect. For strong damping the oscillator crosses
to the overdamped regime: after an initial raise P2 (t) decays to zero without
any oscillation (see the book by Budker, Kimball and De Mille quoted
in the preface). Some more detail on the relaxation mechanism for spins in a
magnetic field will be given at Chapter 6.

Appendix I.3 Transition probabilities and selection

rules
The phenomenological transition probabilities induced by electromagnetic
radiation are defined in Problem F.I.1, where the Einstein relations are also
derived. To illustrate the mechanism underlying the effect of the radiation one
has to express the absorption probability W12 between two levels |1 > and
|2 > in terms of the Hamiltonian describing the interaction of the radiation
44 1 Atoms: general aspects

with the system. Here this description is carried out by resorting to the time-
dependent perturbation theory.
The perturbation Hamiltonian HP (t), already introduced in Appendix I.2,
is then specified in the form

HP (t) = H1 eiωt , (A.I.3.1)

appropriate to the electromagnetic (e.m.) radiation. In fact, from the one-

electron Hamiltonian in fields (see Eq. 1.26)

c )
(p + eA 2
H= − eϕ (A.I.3.2)
2m
(A and ϕ vector and scalar potentials), recalling that [p, A] = −ih̄(∇.A) ∝
divA = 0 in the Lorentz gauge and that for electromagnetic radiation
A(r, t) = A0 exp[i(k.r − ωt)], the first order perturbation Hamiltonian turns
out
ih̄e
Hrad = − A.∇ (A.I.3.3)
mc
By expanding A(r, t)

A(r, t) = A0 e−iωt [1 + i(k.r) + ...] (A.I.3.4)

and limiting the attention to the site-independent term (electric dipole ap-
proximation or long-wave length approximation) one can show that6
c
Hrad ∝ A0 .∇ ∝ A0 .r ∝ E0 .r
ω21
Therefore H1 in A.I.3.1 takes the form H1 = −er.E0 , with E0 amplitude of
the e.m. field (electric dipole mechanism of transition).
Now we use the results obtained in Appendix A.I.2, again considering
that (Hrad )11 = (Hrad )22 = 0 and (Hrad )12 = (Hrad )∗21 . The equations for
the coefficients c1,2 become

dc1
ih̄ = c2 e−iω21 t cosωt < 1|x|2 > eE0
dt
dc2
ih̄ = c1 e+iω21 t cosωt < 2|x|1 > eE0 (A.I.3.5)
dt
for a given x-component of the operator r. For Eqs. A.I.3.5 only approximate
solutions are possible, essentially based on the perturbation condition
6
It is recalled that E = −(1/c)∂A/∂t and that the matrix element of the ∇
operator can be expressed in terms of the one for r:

< 2|∇|1 >= −mω21 < 2|r|1 > /h̄

 Appendix I.3 Transition probabilities and selection rules 45

HP ' H0 (while they are solved exactly for ω = 0, as seen in App.I.2).

For ω around ω21 one finds
! "
2 (ω−ω21 )t
sin
t2 2
|c2 (t)| = 2
2
| < 2|H1 |1 > |2 . (A.I.3.6)
h̄ (ω − ω 21 )2 t2

|c2 (t)|2 has the time dependence depicted below, with a maximum at ω = ω21
proportional to t2 .

2
|c2(t)|

-4π/t -2π/t 0 2π/t 4π/t

ω-ω21

On increasing t the zeroes of the function tend to the origin while the
maximum increases. Thus for t → ∞ one has |c2 (t)|2 ∝ δ(ω − ω21 ), δ being
the Dirac delta function. By taking into account the spread of the excited
state due to the finite width (see Prob. F.I.1) or by resorting to the non-
monocromatic character of the radiation, one writes
&
t2 sin2 (ω−ω 21 )t

|c2 (t)| ∝ 2
2
ρ(ω) 2
2 dω (A.I.3.7)
h̄ (ω − ω21 )2 t4
where the frequency distribution ρ(ω) of the radiation is a slowly varying
function around ω21 . Then one can set ρ(ω) " ρ(ω21 ). The integration over ω
yields 2π/t, and thus the transition probability per unit time becomes
2π
W12 = |c2 (t)|2 /t =2 | < 2|H1 |1 > | δ(ω − ω21 )
2
h̄
For the electric dipole mechanism and linear polarization of the radiation
along ε̂ this Equation reads
2π
W12 = 2 | < 2| − er.ε̂|1 > | E0 δ(ω − ω21 )
2 2
(A.I.3.8)
h̄
For random orientation of r with respect to the e.m. wave one has to average
cos2 θ over θ, to obtain 1/3. By introducing the energy density ρ(ω21 ) or ρ(ν21 )
(ρ =< E 2 > /4π) one finally obtains
46 1 Atoms: general aspects

2π
W12 = 2 ρ(ν21 )|R21 |
2
(A.I.3.9)
3h̄
where |R21 |2 = | < 2| − ex|1 > |2 + | < 2| − ey|1 > |2 + | < 2| − ez|1 > |2 .
R21 represents an effective quantum electric dipole associated with
a pair of states. The selection rules arise from the condition

R21 ≡< 2| − er|1 >*= 0.

In the central field approximation the selection rules are

i) each electron makes a transition independently from the others ;
ii) neglecting the spin, the electric dipole transitions are possible when
∆l = ±1 and ∆m = 0, ±1 (according to parity arguments involving the
spherical harmonics).
When the spin-orbit interaction is taken into account the selection rules
are
∆j = 0, ±1 and j = 0 ↔ j = 0 transition not allowed ;
∆m = 0, ±1 and no transition from m = 0 ↔ m = 0, when ∆j = 0 (in
view of the conservation of the angular momentum).

The magnetic dipole transitions (mechanism associated with the term

(ik.r) in A.I.3.4) are controlled by the selection rules
∆l = 0 and ∆m = 0, ±1
while for the transition driven by the electric quadrupole mechanism
∆l = 0, ±2 and ∆m = 0, ±1, ±2 (l = 0 ↔ l$ = 0 forbidden)
Further details on the selection rules will be given at §3.5. Here we remark
that the transition probabilities associated with the magnetic dipole or with
the electric quadrupole mechanisms are smaller than W12 in A.I.3.9 by a factor
of the order of the square of the fine structure constant α = e2 /h̄c " 1/137
(for e.m. radiation in the visible spectral range).