Chapter 2

Time-dependent perturbation
theory

2.1 Introduction
Postulate 2 of quantum mechanics tells us that if we wish to determine the time-
evolution of a quantum system we solve the time-dependent Schrödinger equation
(TDSE)
∂|ψi
i~ = Ĥ|ψi (2.1)
∂t
where Ĥ is the Hamiltonian operator, a Hermitian operator which specifies the
full dynamics of the system.
If the Hamiltonian is constant in time, and if we know its spectral decompo-
sition (eigenvalues Ek and eigenvectors |φk i), we can integrate the Schrödinger
equation for any initial state expressed in terms of orthonormal eigenvectors of
Ĥ. i.e.
−iEk t
X  
|ψ(t)i = exp ck |φk i (2.2)
k
~
In many systems of interest, however, we are not so lucky. Either a) we possess
no method for diagonalising the Hamiltonian (often the case, for example with
interacting many-body systems) or b) the Hamiltonian is not constant in time.
Fortunately, in a remarkably broad range of cases, we find Hamiltonians which
are close to ones we can solve exactly.
Perturbation theory is the name of a set of techniques, where the problem in
question may be seen as a small perturbation away from a problem which we do
118 Chapter 2. Time-dependent perturbation theory

know how to solve, and where the solutions are close, up to small corrections, to
the exact solutions of the solvable problem. Perturbation theory provides methods
for calculating these corrections, which often describe important phenomena in
the physics of the problem.
You are likely to have encountered time-independent perturbation theory in
your previous courses1 . This is used to study problems where Ĥ is constant in
time, but where we do not know how to exactly calculate its eigenvectors and
eigenvalues. In this chapter, you will study time dependent perturbation theory,
used to approximate the dynamics of a problem when either a) or b) or both of
the above criteria do not hold. Time-dependent perturbation theory has had many
applications in quantum theory, and plays an important role in many fields of
physics, e.g from atomic physics to laser physics, from particle physics to quan-
tum field theory.

Statement of the problem

Consider a system described by the following Hamiltonian

Ĥ(t) = Ĥ0 + λV̂ (t) (2.3)

where Ĥ0 is a time-independent Hamiltonian, whose eigenvalues Ek and eigen-

vectors |φk i are known, V̂ (t) is a (usually time-dependent) Hamiltonian, and λ
is a real number parameter which is introduced as a tool in our analysis, later we
shall set λ = 1. Perturbation theory is effective when the perturbation V̂ (t) is
weak, which, loosely speaking, means that the energies involved in this Hamilto-
nian are much smaller than the energy scales of Ĥ0 , or that the matrix elements of
V̂t are much smaller than the differences in energy Ej 0 − Ej .
Beware of the notation here, V̂ is not a simple potential energy operator, and
should not be associated with those simple spatial potential operators encountered
in previous courses. V̂ could be any Hamiltonian operator, but in practice it will
often be a Hamiltonian associated with the interaction energy of two particles or
degrees of freedom.
If no perturbation were present (V̂ (t) = 0) then we could solve the TDSE to
obtain
−iEk t
X  
|ψ(t)i = exp ck |φk i (2.4)
k
~
1
If you did not, I recommend that you read chapter 8.1 in Bransden and Joachain’s Quantum
Mechanics as background to the present chapter.

Phas0069 Advanced Quantum Theory

2.2. Dirac’s method of variation of constants 119

as above.
Note that the coefficients ck here are constant in time, and depend on the initial
state X
|ψ(0)i = ck |φk i (2.5)
k

and in this case hψ(t)|Ĥ0 |ψ(t)i = k Ek |ck |2 and the expectation value of the
P
energy is constant.
The goal of time-dependent perturbation theory is to provide approximate so-
lutions to the TDSE when perturbation V̂ (t) is small, but non-zero.

2.2 Dirac’s method of variation of constants

Since we know the solution Ĥ0 |φj i = Ek |φj i and the time-dependence in that
case, we would like to focus our attention on the effect on the dynamics of the
perturbation term λV̂ (t).
Let us represent the state in terms of the eigenstates of Ĥ0 ,

−iEj t
X  
|ψ(t)i = cj (t) exp |φj i, (2.6)
j
~

but we now allow the probability amplitudes to vary in time.

We now substitute this ansatz into the time-dependent Schrödinger equation

−iEj t
 
∂ X
i~ cj (t) exp |φj i = (Ĥ0 + λV̂ (t))|ψ(t)i (2.7)
∂t j ~

so that
−iEj t −iEj t
X   X  
cj (t)Ej exp |φj i+i~ ċj (t) exp |φj i = (Ĥ0 +λV̂ (t))|ψ(t)i
j
~ j
~
(2.8)
since Ĥ0 |φj i = Ek |φj i, we can see that the first term on the LHS and the first
term on the RHS cancel, so:

−iEj t −iEj t
X   X  
i~ ċj (t) exp |φj i = λV̂ (t) cj (t)Ej exp |φj i (2.9)
j
~ j
~
120 Chapter 2. Time-dependent perturbation theory

If we multiply from the left with hφm |, using hφm |φk i = δm,k and writing
Vmk (t) = hφm |V̂ (t)|φk i, we obtain

λ X
ċm (t) = ck (t) exp [iωmk t] Vmk (t) (2.10)
i~ k

where ωmk = (Em − Ek )/~ is called the Bohr angular frequency. Vmk (t) =
hφm |V̂ (t)|φk i is a matrix element for V̂ .
Equation (2.10) is precisely equivalent to the time-dependent Schrödinger
equation. We have not made any approximations yet. It is a series of coupled
differential equations for the coefficients cm (t), which, if one could solve them,
would provide an exact solution to the TDSE. Note that the operator Ĥ0 does not
appear in equation (2.10). Since the time-evolution due to Ĥ0 alone is already
explicitly taken care of, one can focus one’s attention solely on the perturbation
λV̂ (t), which significantly simplifies the equations used.
In many cases (2.10) cannot be solved exactly, but, provided the perturbation
is weak enough, a perturbative expansion of cm (t) can be effective with only a
small number of terms.

2.3 Perturbative expansion

We shall express cm (t) as a power series in λ:

cm (t) =c(0) (1) 2 (2)

m (t) + λcm (t) + λ cm (t) + · · ·

X (2.11)
= λj c(j)
m (t).
j

Substituting this expression into (2.10) we obtain

" #
∂ X λ X X (n)
λj c(j)
m (t) = exp [iωmk t] Vmk (t) λn ck (t) (2.12)
∂t j i~ k n

This expression should hold for any value of λ, and hence terms on the LHS
associated with λp must be equal to terms associated with λp on the RHS. Equating
powers of λ gives rise to a hierarchy of differential equations:

Phas0069 Advanced Quantum Theory

2.3. Perturbative expansion 121

To derive an expression for 0th-order terms, we identify terms to zeroth power

in λ. On the left-hand side that leaves us j = 0 on the right-hand-side no term
satisfies this, since all are multiplied by the λ outside the sum, and hence,

ċ(0)
m (t) = 0 (2.13)

An expression for 1st-order terms may be obtained by considering λ1 , i.e.

j = 1, n = 0:
1 X (0)
ċ(1)
m (t) = exp [iwmk t] Vmk (t)ck (t) (2.14)
i~ k
Similarly for 2nd-order terms, we consider λ2 , i.e. j = 2, n = 1
1 X (1)
ċ(2)
m (t) = exp [iwmk t] Vmk (t)ck (t) (2.15)
i~ k

Repeating this process we see that, in general, the n-th order equation will
have the form
1 X (n−1)
ċ(n)
m (t) = exp [iwmk t] Vmk (t)ck (t) (2.16)
i~ k

Note that from equation (2.13) the zero-order terms are constant in time, and
also that if we set λ = 0 we recover the unperturbed Hamiltonian, and equation
(2.11) reduces to the the 0th-order solution. This means that the zeroth order
coefficients are equal to the unperturbed solution.
Note also that the expressions above are independent of λ. They therefore hold
even if we set λ = 1 and write
cm (t) =c(0) (1) (2)
m (t) + cm (t) + cm (t) + · · ·

X (2.17)
= c(j)
m (t).
j

It is natural to ask whether such a series (with λ = 1) will even converge.

However, provided the perturbation is sufficiently weak, we find that this series
does provide a good approximation to the true state even after just a small num-
ber of terms. In practice, in surprisingly many cases, perturbation theory works
extremely well even to just to the first or second order correction. This means
that the results of first and second order perturbation theory are utilised widely in
physics.
122 Chapter 2. Time-dependent perturbation theory

2.4 First-order term

In the following sections we shall simplify our analysis by assuming that the per-
turbation is only “switched on” at a certain time t = 0, and that prior to that mo-
ment, the system is in an eigenstate |ψ(t)i = |φj i, in other words, the eigenstate
of Ĥ0 with energy E0 . This assumption fits many scenarios where time-dependent
perturbation theory is employed.
This allows us to immediately solve for the zeroth-order solution. At time
t = 0, when the perturbation is (instantaneously) switched on, the evolution will
still (for an instant), satisfy the unperturbed Hamiltonian, and the unperturbed
solution holds. This means that for t = 0 only the zero-th order terms are non-
zero. Since the initial state is |φj i, the zeroth order terms must then be:

c(0)
m (t) = δm,j . (2.18)

Let us now turn our attention to the first order terms. We can substitute equa-
tion (2.18) into equation (2.14) to obtain a differential equation which we can
integrate:
1
ċ(1)
m (t) = exp [iwmj t] Vmj (t). (2.19)
i~
Let us consider the m = j coefficient and m 6= j coefficients separately, and
integrate between time t = 0 and time t = τ , to give the first order term at t = τ .
First, for m = j
1 τ
Z
(1)
cj (τ ) = dt Vjj (t) (2.20)
i~ 0
and for m 6= j
Z τ
1
c(1)
m (τ ) = dt Vmj (t) exp [iwmj t] . (2.21)
i~ 0

If the matrix elements of the perturbation V̂ (t) are known, we can integrate
these expressions to calculate these coefficients.
When the perturbation is sufficiently weak, the first order solution can be a
good approximation to the exact solution, and

cm (τ ) ≈ δj,m + c(1)
m (τ ) (2.22)

where we have written the zeroth order term explicitly.

Phas0069 Advanced Quantum Theory

2.4. First-order term 123

(1)
Note that, if cm (τ ) for m 6= j is non-zero, this indicates that there is a non-
zero probability that the system, if measured at time τ will be found in state |φm i.
We call this probability the transition probability Pj→m and write
(1)
Pj→m (τ ) = |c(1)
m (τ )|
2
(2.23)

where the superscript (1) denotes that this is a first order approximation.
(1)
On the other hand |1 + cj (τ )|2 represents the probability that the system
remains in the initial state |φj i. We expect, if the perturbation is weak, that this
will stay close to 1. Comparing, cj (τ ) to first order
(1)
cj (τ ) ≈ 1 + cj
1 τ (2.24)
Z
=1+ dt Vjj (t)
i~ 0
with an exponential function expanded to first order
−i τ −i τ
Z Z
exp[ dt Vjj (t)] ≈ 1 + dt Vjj (t) (2.25)
~ 0 ~ 0
we see that to first order, the correction to cj (τ ) due to the perturbation is equiva-
lent to a phase shift
cj (τ ) ≈ eiφ cj (0) (2.26)
where
−1 τ
Z
φ= dt Vjj (t) (2.27)
~ 0
This implies that the probability that the system will remain in its initial state
is close to one, and that the first order transition probabilities will be small.
Note that we may use the above to study a more general initial condition. For
instance, if we wish to study an initial state which corresponds to a superposition
of eigenstates, we can do this via a modification of the above method since, if
Û (t) is the unitary time evolution operator describing the quantum evolution,
X X  
Û αj |φj i = αj Û |φj i (2.28)
j j

and hence one could solve the more general case, by calculating the evolution of
individual initial eigenstates |φj i and noticing that the state at time t is simply the
superposition of the evolved eigenstates (Û |φj i) .
124 Chapter 2. Time-dependent perturbation theory

2.5 Second order correction

In some cases, the perturbation will be strong enough that the second order terms
are required to get a good approximation. Once first-order terms have been de-
rived, we may substitute them into equation (2.15) to obtain explicit expressions
for the second order terms. Let us continue our analysis of the case where the
perturbation is switched on at time t = 0 and |ψ(0)i = |φj i.
Recall that
1 X (1)
ċ(2)
m (t) = exp [iwmk t] Vmk (t)ck (t) (2.29)
i~ k

Substituting in the first order terms we derived above (equation (2.21)) and
integrating, we obtain
τ t0
−1
Z Z X
0
c(2)
m (τ ) = 2 dt dt exp [iwmk t0 ]
~ 0 0 k (2.30)
0
× exp [iwkj t] Vmk (t )Vkj (t)

There are several remarks to make on the structure of these second order terms.
Notice the double time integral, first over variable t and then over t0 . In general,
due to repeated integration, an n-th order correction term will contain an n-fold
time integral. Notice also the matrix elements of V̂ involved are of the form.
Vmk (t0 )Vkj , which we can write hφm |V̂ |φk ihφk |V̂ |φj i. Notice that this is the prod-
uct of elements associated with two processes, first the transition from |φj i to |φk i
and then the transition from |φk i to |φm i. Thus, even if the first order term for the
transition Vmj is zero, and there is no direct coupling between states, |φj i to |φm i
can still take place at the level of second order terms, if a transition via a third
state is possible.

2.6 Example: A constant perturbation

Some of the power of time-dependent perturbation theory is that some very gen-
eral features of quantum processes may be seen in some simple examples. The
first such example we shall consider is a perturbation which is constant in time.
We assume that the perturbation is switched on at time t = 0, and that this “switch-
ing on” represents the only time-variation in the Hamiltonian. In other words, the
Hamiltonian Ĥ can be written

Phas0069 Advanced Quantum Theory

2.6. Example: A constant perturbation 125

Ĥ0 t<0
Ĥ = (2.31)
Ĥ0 + V̂ 0 ≤ t < τ.

We shall assume that the system is measured at time t = τ and that we are not
interested in the dynamics of the system after that time. Importantly the operator
V̂ is constant in time.
As before, we shall simplify our analysis by assuming that at time t = 0 the
system is in an eigenstate of Ĥ0 , |φj i corresponding to an
P energy Ej . We wish to
approximate the evolution of this state in time, |ψ(t)i = m cm (t) exp[−iEm t/~]|φm i,
and hence find approximations to the coefficients cm (τ ) at time t = τ when the
state is measured.
Let us first assume the perturbation is weak enough to work to first order. We
can therefore utilise directly the expression obtained in section 2.4. Substituting a
time independent perturbation V̂ into equation (2.21) we obtain for m 6= j

Z τ
1
c(1)
m (τ ) = dt Vmj exp [iwmj t] (2.32)
i~ 0

which we can integrate:

Z τ
Vkj
c(1)
m (τ ) = dt exp [iwmj t]
i~ 0
(2.33)
Vmj
= [1 − exp [iwmj τ ]] .
~ωmj

Note how this term is proportional to the ratio Vmj /(~ωmj ) = Vmj /(Em −Ej ).
Higher order corrections will be proportional to higher-order powers of such ra-
tios. This motivates the claim that perturbation theory is applicable when the
matrix elements of the perturbation Hamiltonian are much smaller than the dif-
ference in eigenvalues of Ĥ0 . Physically speaking, this tells us that the energy
scales associated with the perturbation are much smaller than the energy scales
associated with Ĥ0 , and this gives some intuition for the kinds of physical system
in which perturbation theory can be appropriately deployed.
We can calculate first order transition probabilities from equation (2.33)
126 Chapter 2. Time-dependent perturbation theory

(1)
Pj→m (τ ) = |c(1)
m (τ )|
2

|Vmj |2 (1 − exp [iwmj τ ]) (1 − exp [−iwmj τ ])

= 2
~2 ωmj (2.34)
4|Vmj |2 sin2 (ωmj τ /2)
= 2
~2 ωmj

where we have used the fact that 1−eix = eix/2 (e−ix/2 −eix/2 ) = eix/2 (−2i sin(x)).
We shall write this expression

(1) 2|Vmj |2
Pj→m (τ ) = F (τ, ωmj ) (2.35)
~2
where
2 sin2 (ωτ /2)
F (τ, ω) =
ω2
τ sin2 (ωτ /2)
2
= (2.36)
2 (ωτ /2)2
τ2
= sinc2 (ωτ /2)
2
We illustrate F (τ, ω) in figure 2.1, plotted as a function of ω. Note the sinc-
like shape of this function, dominated by one broad and high central peak, and
other smaller peaks diminishing in height as you move away from ω = 0. The
height of the central peak is τ 2 and its full-width at half maximum is 2π τ
. Thus,
as τ increases (i.e. the duration of the perturbation increases) the peak of this
function gets narrower and taller. To describe it in the limit that τ gets very large,
we require the Dirac δ-function.

Aside: Dirac delta function

You will have encountered the Dirac delta function in previous courses, so I will
just summarise some of its properties here. It has a number of definitions, one of
which, the integral definition
Z ∞
f (x) = f (x0 )δ(x − x0 )dx0 (2.37)
−∞

Phas0069 Advanced Quantum Theory

2.6. Example: A constant perturbation 127

F (τ, ω)

τ2
2

2π
τ

−4π −2π 2π 4π
ω
0
τ τ τ τ

Figure 2.1: In this figure, the function F (τ, ω) is plotted as a function of ω.

128 Chapter 2. Time-dependent perturbation theory

demonstrates the way that the delta function can be used to “remove” an inte-
gral, just as the standard delta (sometimes called Kronecker delta) δj,k allows us
to remove summations. The Dirac delta function is not a true function, since its
value on all inputs is not defined. It has a number of important properties
1. For x 6= 0 it takes the value δ(x) = 0, for x = 0 the value of δ(x) is
undefined (“infinite”).
2. It satisfies: Z ∞
δ(x)dx = 1 (2.38)
−∞

3. and
1
δ(ax) = δ(x) (2.39)
|a|
The Dirac delta can be defined as the limiting value of many functions which
have a sharp central peak, and which satisfy the property that, in some limit, the
peak becomes narrow and high, and while the integral over x from −∞ to +∞
converges to 1.

A constant perturbation: continued

Returning to our analysis of a constant perturbation, we shall now see that F (τ, ωkj )
can, in the limit τ → ∞ be written in terms of a δ-function. Namely,

lim F (τ, ω) = πτ δ(ω). (2.40)

τ →∞

To see that this is reasonable, notice that in the limit that τ gets very large,
the height of the central peak (τ 2 /2) becomes large, while its width (2π/τ ) gets
small, at the same time, in this limit, the integral over ω behaves as follows:

Z ∞ Z ∞
F (τ, ω) τ  ωτ 
= sinc2 dω
−∞ πτ 2π −∞ 2
Z ∞
τ 2
= sinc2 (x) dx
2π −∞ τ (2.41)
1 ∞
Z
= sinc2 (x) dx
π −∞
π
= =1
π

Phas0069 Advanced Quantum Theory

2.6. Example: A constant perturbation 129

R∞
where we used a standard integral −∞ sinc2 (x) dx = π. Notice that this integral
is constant, and thus remains at this value in the limit τ → ∞ as required for a
δ-function.

Perturbation Time-scales and Fermi’s Golden Rule

If we return to consider the first order transition probabilities for the constant
perturbation
(1) 2|Vmj |2
Pj→m (τ ) = F (τ, ωmj ) (2.42)
~2
we can see two different limits of behaviour, which hold independently of the
operator V̂ (i.e. independently of the physics of the problem). Recall that ~ωkj =
Ek − Ej represents the energy difference between initial and final states. For
long perturbations, we can take the limit τ → ∞ and, in this limit, we can write
equation (2.42)
(1) 2|Vmj |2
Pj→m (τ ) = πτ δ(ωjm ) (2.43)
~2
which is zero unless ~ωmj = Em − Ej = 0. In other words, in the long time limit,
regardless of the form of V̂ , first order transitions can only occur between states of
identical energy. This could be seen as a natural consequence of the conservation
of energy.
On the other, hand, if we consider the small τ limit, we see for short per-
turbations a very different behaviour. In this limit F (τ, ωmj ) has a broad central
peak, and thus transitions are allowed to states with a significant range of energies
around Ej . To analyse this more quantitatively, let ∆ω = 2π/τ be the width of
the central peak and let ∆t = τ be the length of the perturbation.
Writing ∆E = ~∆ω to represent the uncertainty in the energy of the final
state after the transition we identify the relationship
∆E∆t = ~ (2.44)
This relationship reflects the time-energy uncertainty relation. It tells us that,
over small time-scales quantum mechanics allows the energy of a system to be
uncertain. Note that the energy-time uncertainty relation is of a fundamentally
different kind to the standard Heisenberg uncertainty relation ∆x∆p ≥ ~/2 since
unlike x and p, time t is not an observable. Nevertheless, the energy time uncer-
tainty relation occurs in many physical phenomena such as the relationship be-
tween the line-width of radiation emitted from a transition and the natural decay
rate of the transition.
130 Chapter 2. Time-dependent perturbation theory

Let us return now to the long time τ → ∞ limit. Notice that equation (2.43)
is linear in time τ . We can therefore differentiate it to derive the transition rate
(1) 2π 2π
Γj→k (τ ) = 2 |Vkj |2 δ(ωjk ) = |Vkj |2 δ(Ej − Ek ) (2.45)
~ ~
This equation is called Fermi’s Golden Rule. This rule, although approximate
and derived using first-order perturbation theory, provides an accurate approxi-
mation of the transition rate of many physical processes, and is used as a useful
rule-of-thumb to estimate transition rates before a more accurate calculation is
made in many areas of physics, such as particle physics and atomic physics.

2.7 Harmonic Perturbation

In our previous example, the perturbation was assumed to be constant in time, but
time-dependent perturbation theory can also treat perturbations which are time-
varying. In this section, we shall consider the example of a harmonic perturbation,
a perturbation V̂ (t) varying sinusoidally in time. This is particularly important as
it describes, to a first approximation, how atoms interact with light.
We can construct a rudimentary model of an atom by considering an electric
dipole. An electric dipole consists of a pair of separated equal and opposite elec-
tric charges. We can think of our atom as a dipole formed by the positive nucleus
surrounded by inner electrons, and the negative outer electron.
~ has the following potential energy
An electric dipole in an electric field E
~ ·D
V = −E ~ (2.46)
where D ~ is the dipole moment.
Although this model is very naive, to a good first approximation, atoms do
indeed interact with light like an electric dipole. For a fully quantum treatment,
we would construct an atom-light interaction Hamiltonian from a quantum dipole
operator and a quantum operator for the electromagnetic field (for more details
see any quantum optics textbook). Here, however, we shall take a very simple
approach and simply study the time-dependence of such an interaction.
Consider an atom interacting with a monochromatic laser. The E-field of the
~
laser will be sinusoidal in time, ie E(t) =E ~ 0 cos(ωt), where ω is the angular fre-
quency of the radiation. The Hamiltonian representing the atom-light interaction
will therefore have the form:
V̂ = ĤI cos(ωt) (2.47)

Phas0069 Advanced Quantum Theory

2.7. Harmonic Perturbation 131

where ĤI is an operator which details the specific quantum description of the
interaction (leading to transition rules etc.). We see that this Hamiltonian varies
sinusoidally in time.
It is convenient to study a slightly more general family of harmonic Hamilto-
nians
V̂ = Âeiωt + † e−iωt (2.48)
In the special case of A being a real symmetric matrix, this reduces to equation
(2.47) with ĤI = 2Â.
As in previous examples, we shall label the unperturbed Hamiltonian Ĥ0 and
assume that the perturbation is first “switched on” at time t = 0 and that prior
to this time, the system is prepared in the state |φj i. For all times t ≥ 0, the
Hamiltonian will be
Ĥ = Ĥ0 + Âeiωt + † e−iωt . (2.49)
We can now Puse the results derived in earlier sections to calculate the coeffi-
cients of state m cm (t)|φm i to first order
cm (t) ≈ c(0) (1)
m + cm (t) (2.50)
As before, the 0-th order terms are determined by the initial condition |ψ(t = 0)i =
|φj i,
c(0)
m = δm,j (2.51)
The first order terms are given by equation (2.21)
1 τ
Z
(1)
cm (τ ) = dt Vmj (t) exp [iwmj t] . (2.52)
i~ 0
If we substitute equation (2.48) into this expression we obtain
 Z τ Z τ 
(1) 1 i(ωmj +ω)t † i(ωmj −ω)t
cm (τ ) = Amj dt e + (A)mj dt e
i~ 0 0
Amj 1 − ei(ωmj +ω)τ A∗jm 1 − ei(ωmj −ω)τ
= +
~ ωmj + ω ~ ωmj − ω (2.53)
Amj
2iτ ei(ωmj +ω)τ /2 sinc ((ωmj + ω)τ /2)


=
~
A∗jm
2iτ ei(ωmj −ω)τ /2 sinc ((ωmj − ω)τ /2)


+
~
Just as for the case of a constant perturbation, the first order terms contain sinc
functions from which in the limit τ >> 0 will become Dirac delta functions. Via
132 Chapter 2. Time-dependent perturbation theory

an analysis similar to that in the previous section, one can show that the first term
sinc ((ωmj + ω)τ /2) becomes proportional to δ(ωmj + ω) and the second term
sinc ((ωmj − ω)τ /2) becomes δ(ωmj − ω).
In this limit, the first order coefficients will only be non-zero if one of two
possible conditions are fulfilled, either a) ωmj = ω or b) ωmj = −ω. These
two cases correspond to conditions on the energies of the states |φm i to which a
transition can occur.
If case a) holds, ωmj = ω implies that (Em − Ej )/~ = ω and thus Em =
Ej + ~ω, i.e. the energy of the state |φm i is one unit of ~ω higher than Ej . If case
b) holds ωmj = −ω and therefore Em = Ej − ~ω, i.e. |φm i is one unit of ~ω
lower than Ej .
This means that first order perturbation theory has correctly predicted that light
energy can only be absorbed by an atom in units of ~ω i.e. single photons. This
is a remarkable result, since it was obtained with no detailed physical descrip-
tion of the atom-light interaction, and was achieved without even a full quantum
mechanical treatment of light! A more comprehensive approach to atom-light in-
teraction predicts similarly, absorption and stimulated emission, but also predicts
a third kind of transition, missing from the simple treatment above, spontaneous
emission of radiation, which can occur without any driving field.
In an analogous way to the constant perturbation, we can derive Fermi golden
rules which describe the rate of transition from one state to another. These take a
very similar form to the Fermi golden rule for a constant perturbation:
Absorption rate:
(1) 2π
Γj→k (τ ) = |Akj |2 δ(Ek − Ej − ~ω) (2.54)
~
Emission rate:
(1) 2π ∗ 2
Γj→k (τ ) = |A | δ(Ek − Ej + ~ω) (2.55)
~ jk
Note that by swapping the labels j and k in equation (2.55) we obtain
(1) 2π
Γk→j (τ ) = |Akj |2 δ(Ek − Ej − ~ω) (2.56)
~
where we used δ(x) = δ(−x) and |A∗kj |2 = |Akj |2 .
This means that the rate of stimulated emission from state k to state j is equal
to the rate of absorption from j to k. This is an example of “detailed balance”,
where the rate of a process is equal to the rate of an opposite process. Detailed
balance applies in many examples in physics and can be a useful calculation tool.

Phas0069 Advanced Quantum Theory