Andrei Tokmakoff, MIT Department of Chemistry, 3/14/2007

2.4

2-26

PERTURBATION THEORY

()

()

Given a Hamiltonian H t = H 0 + V t where we know the eigenkets for H 0 : H 0 n = En n , we

()

can calculate the evolution of the wavefunction that results from V t :

()

()

! I t = " bn t n

(2.111)

using the coupled differential equations for the amplitudes of n . For a complex time-

dependence or a system with many states to be considered, solving these equations isnt

( )

()

practical. Alternatively, we can choose to work directly with U I t,t0 , calculate bk t as:

( ) ( )

bk = k U I t,t0 ! t0

$ !i
U I t,t0 = exp+ &
%!

( )

where

t0

(2.112)

'
VI " d" )
(

()

(2.113)

Now we can truncate the expansion after a few terms. This is perturbation theory, where the

()

dynamics under H 0 are treated exactly, but the influence of V t on bn is truncated. This
works well for small changes in amplitude of the quantum states with small coupling matrix

()

()

elements relative to the energy splittings involved ( bk t ! bk 0 ; V ! Ek " En ) As well see,

the results we obtain from perturbation theory are widely used for spectroscopy, condensed
phase dynamics, and relaxation.

Transition Probability
Lets take the specific case where we have a system prepared in ! , and we want to know the

()

probability of observing the system in k at time t , due to V t .

()

()

Pk t = bk t

()

( )

bk t = k U I t,t0 !

(2.114)

2-27

$ i t
'
bk t = k exp+ & ! # d" VI " ) "
% ! t0
(

()

()

i t
d" k VI " !
" #t0

()

()

bk t = k ! !

$ !i '
+& )
% "(

So,

d" 2

t0

"2

t0

()

()

bk t = ! k! "

( ) ( )

()

!i" !k t

i t
"i$ #
d# 1 e !k 1Vk! # 1
%
t
" 0

( )

" !i %
+( $ '
# !&
m

t0

d) 2

)2

t0

d) 1 e

(2.116)

d" 1 k VI " 2 VI " 1 !

k VI t ! = k U 0 V t U 0 ! = e

using

(2.115)

()

(2.117)

Vk! t

first order

!i* mk ) 2

( ) e!i*

Vkm ) 2

"m) 1

( )

Vm" ) 1

(2.118)

(2.119)

second order
The first-order term allows only direct transitions between ! and k , as allowed by the matrix
element in V, whereas the second-order term accounts for transitions occuring through all
possible intermediate states m . For perturbation theory, the time ordered integral is truncated at
the appropriate order. Including only the first integral is first-order perturbation theory. The
order of perturbation theory that one would extend a calculation should be evaluated initially by
which allowed pathways between ! and k

you need to account for and which ones are

allowed by the matrix elements.

For first order perturbation theory, the expression in eq. (2.118) is the solution to the
differential equation that you get for direct coupling between ! and k :

!
"i " i# "k t
bk =
e
Vk" t b" 0
!t
!

() ( )

(2.120)

2-28
This indicates that the solution doesnt allow for the feedback between ! and k that accounts

()

()

for changing populations. This is the reason we say that validity dictates bk t ! bk 0 .
If ! 0 is not an eigenstate, we only need to express it as a superposition of eigenstates,

()

()

! 0 = " bn 0 n and
n

()

bk t = ! bn 0 k U I n .
n

(2.121)

Now there may be interference effects between the pathways initiating from different states:

()

()

Pk t = ck t

()

= bk t

!
n

()

k bn t n

(2.122)

Also note that if the system is initially prepared in a state ! , and a time-dependent
perturbation is turned on and then turned off over the time interval t = !" to +! , then the
complex amplitude in the target state k is just the Fourier transform of V(t) evaluated at the
energy gap ! !k .

()

bk t = !

i +$
!i# "
d" e "k Vk" "
%
!$
!

()

(2.123)

If the Fourier transform is defined as

( )

()

V! ! " F! #$V t %& =

then

1
2'

+)

()

( )

Pk! = V" ! k!

() ( )

dt V t exp i! t ,
2

(2.124)

(2.125)

2-29
Example: First-order Perturbation Theory
Vibrational excitation on compression of harmonic oscillator. Lets subject a harmonic oscillator
to a Gaussian compression pulse, which increases its force constant.

First write the Hamiltonian:

p2
1
H t = T +V t =
+ k t x2
2m
2

()

()

()

(2.126)

()

Now partition it according to H = H 0 + V t :

()

()

k t = k0 + ! k t

k0 = m!

$ t"t
0
! k t = ! k0 exp & "
2# 2
&%

()

$ t"t 2'
p2
1
1
0
)
H=
+ k0 x 2 + ! k0 x 2 exp & "
2m
2
2
2# 2 )(
&%
!##"##$ !####"####
$

"
H 0 = !! $ a a +
#

)(

(2.127)

(2.128)

V (t)

H0

H 0 n = En n

) ')

1%
2 '&

"
En = !! $ n +
#

1%
2 '&

(2.129)

2-30
If the system is in 0 at t0 = !" , what is the probability of finding it in n at t = ! ?

()

bn t =

For n ! 0 :

Using ! n0 = En " E0

!i t
d" Vn0 "
2! #t0

( ) ei$

n0"

(2.130)

! = n# :

()

bn t =

!i
" k n x2 0
2! 0

()

bn t =

So,

+$

!$

d# ein&# e!#

!i
" k0 2#$ n x 2 0
2!

2' 2

(2.131)

2 2 %2 / 2

e!n $

(2.132)

Here we used:

What about the matrix element?

x2 =

!
a + a
2m!

!
aa + a a + aa + a a
2m!

(2.133)

First-order perturbation theory wont allow transitions to n = 1 , only n = 0 and n = 2 .

Generally this wouldnt be realistic, because you would certainly expect excitation to v=1
would dominate over excitation to v=2. A real system would also be anharmonic, in which case,
the leading term in the expansion of the potential V(x), that is linear in x, would not vanish as it
does for a harmonic oscillator, and this would lead to matrix elements that raise and lower the
excitation by one quantum.
However for the present case,

So,

2 x2 0 = 2

!
2m!

!i " # k0 $
2m%

e!2$

b2 =

2 %2

(2.134)

(2.135)

2-31

and

P2 = b2 =

!" k0 2# 2
2m2$2

e%4#

2 $2

= # 2$ 2

2
! & " k0 ) %4# 2$2
e
2 (' k0 2 +*

(2.136)

From the exponential argument, significant transfer of amplitude occurs when the compression
pulse is short compared to the vibrational period.

! <<

1
"

(2.137)

Validity: First order perturbation theory doesnt allow for bn to change much from its initial
value. For P2 << 1

! 2" 2

2
# % $ k0 (
<< 1
2 '& k0 2 *)

(2.138)

Generally, the perturbation k(t) must be small compared to k0, i.e. H 0 >> V , but it should also
work well for the impulsive shock limit (<<1).

2-32
FIRST-ORDER PERTURBATION THEORY
A number of important relationships in quantum mechanics that describe rate processes come
from 1st order perturbation theory. For that, there are a couple of model problems that we want
to work through:

Constant Perturbation

(Step-Function Perturbation)

( )

! "# = ! . A constant perturbation of amplitude V is applied to t0 . What is Pk ?

() (

V t = ! t " t0 V
$0 t < 0
=%
&V t # 0

k U 0 V U 0 ! = V e

To first order, we have:

bk = ! k! "

i! k! t "t0

i t
i% # "t
d# e k! ( 0 )Vk! #
$
" t0

()

(2.139)

Here Vk! is independent of time. Now, assuming k ! ! and setting t0 = 0 we have

t
i
bk = ! Vk" # d"
0
!

=!

ei$ k""

(2.140)

Vk!
# exp i" k!t ! 1%
&
Ek ! E! $

i" t / 2

2iVk! e k!
=!
Ek ! E!

sin " k!t / 2

( )

Where I used ei! " 1 = 2i ei! 2 sin ! 2 . Now

(2.141)

(2.142)

2-33

Pk = bk =

4 Vk!

Ek ! E!

sin2 1 " k!t

(2.143)

Writing this as we did in Lecture 1:

Pk =

where ! = Ek " E!

V2
sin 2 !t / !
!2

(2.144)

2 . Compare this with the exact result we have for the two-level problem:
Pk =

V2
sin 2
2
2
V +!

!2 + V 2 t / !

(2.145)

Clearly the perturbation theory result works for V << .

We can also write the first-order result as

Pk =

()

()

V 2t 2
sinc 2 !t / 2!
2
!

(2.146)

()

where sinc x = sin x x . Since lim sinc x = 1 ,

x!0

lim Pk = V 2t 2 ! 2

(2.147)

!"0

The probability of transfer from ! to k as a function of the energy level splitting Ek ! E! :

Area scales linearly

with time.

Since the energy spread of states to which transfer is efficient scales approximately
as Ek ! E! < 2" " t , this observation is sometimes referred to as an uncertainty relation

2-34
with !E " !t # 2$ ! . However, remember that this is really just an observation of the principles
of Fourier transforms, that frequency can only be determined by the length of the time period
over which you observe oscillations. Since time is not an operator, it is not a true uncertainly
relation like !p " !x # 2$ ! .
Now turning to the time-dependence:

The quadratic growth for =0 is certainly unrealistic (at least for long times), but the expression
shouldnt hold for what is a strong coupling case =0. However, lets continue looking at this
behavior. In the long time limit, the sinc2(x) function narrows rapidly with time giving a delta
function:

lim
t!"

()

lim Pk t =

t!"

sin 2 ax 2
ax 2

2# Vk!
"

)=#$
2

( x)

(2.148)

$ Ek % E! t

(2.149)

The delta function enforces energy conservation, saying that the energies of the initial and target
state must be the same in the long time limit.
What is interesting in eq. (2.149) is that we see a probability growing linearly in time.
This suggests a transfer rate that is independent of time, as expected for simple first order
kinetics:

2-35

()

wk t =

( ) = 2" V

!Pk t
!t

k!

"

# Ek $ E!

(2.150)

This is one statement of Fermis Golden Rule the state-to-state form which describes
relaxation rates from first order perturbation theory. We will show that this rate properly
describes long time exponential relaxation rates that you would expect from the solution
to dP dt = !wP .

2-36
Slowly Applied (Adiabatic) Perturbation
Our perturbation was applied suddenly at t > t 0 (step function)

() (

) ()

V t = ! t " t0 V t

This leads to unphysical consequencesyou generally cant turn on a perturbation fast enough
to appear instantaneous. Since first-order P.T. says that the transition amplitude is related to the
Fourier Transform of the perturbation, this leads to additional Fourier components in the spectral
dependence of the perturbationeven for a monochromatic perturbation!
So, lets apply a perturbation slowly . . .

()

V t = V e!t

here is a small positive number. ! "1 is the

effective turn-on time of the perturbation.
The system is prepared in state ! at t = !" . Find Pk ( t ) .

bk = k U I ! =

bk =

!i t
i# "
d" e k! k V ! e&"
%
!$
"

!iVk! exp '(&t + i# k!t )*

"
& + i# k!

= Vk!

exp '(&t + i Ek ! E! t / " )*

Ek ! E! + i&"

Pk = bk =

Vk!
"2

exp '( 2&t )*

2
& 2 + # k!

Vk! exp '( 2&t )*

(E

! E!

) + (&" )
2

This is a Lorentzian lineshape in ! k! with width 2!! .

2-37

Gradually Applied Perturbation

Step Response Perturbation

The gradually turned on perturbation has a width dependent on the turn-on rate, and is
independent of time. (The amplitude grows exponentially in time.) Notice, there are no nodes
in Pk .
Now, lets calculate the transition rate:
2

!Pk Vk! 2" e2"t

wkl =
= 2
2
!t
" " 2 + # k!

Look at the adiabatic limit; ! "0 . Setting e2!t " 1 and using
lim
!
= $% # k!
2
2
! " 0 ! + # k!

( )

wk! =

2
2
2!
2!
Vk! " # k! =
Vk! " Ek $ E!
2
"
"

( )

We get Fermis Golden Ruleindependent of how perturbation is introduced!

2-38
Harmonic Perturbation
Interaction of a system with an oscillating perturbation
turned on at time t 0 = 0 . This describes how a light
field (monochromatic) induces transitions in a system
through dipole interactions. Again, we are looking to
calculate the transition probability between states !
and k:

()

V t = V cos ! t = " E0 cos ! t

(2.151)

()

Vk! t = Vk! cos ! t

=

(2.152)

Vk! i! t
# e + e" i! t %
&
2 $

To first order, we have:

()

bk = k ! I t

"iVk"
2!

"V
= k"
2!

"i t
i% #
d# Vk" # e k"
$
t
! 0

()

i % +% #
i % "% #
d# & e ( k" ) " e ( k" ) (
'
)

setting t 0 ! 0 (2.153)

& ei (% k" + % )t "1 ei (% k" " % )t "1 (

+
*
+
% k" " % +)
*' % k" + %

( )

Now, using ei! " 1 = 2i ei! 2 sin ! 2 as before:

# i(" k! !" )t / 2 sin # " ! " t / 2 % ei(" k! +" )t / 2 sin # " + " t / 2 % %
!iVk! ' e
$ k!
&+
$ k!
&(
bk =
(
" '
" k! ! "
" k! + "
$
&

(2.154)

Notice that these terms are only significant when ! " ! k! . As we learned before, resonance is
required to gain significant transfer of amplitude.

2-39

First Term

Second Term

max at : ! = +! k!

! = "! k!

E k > E!

E k < E!

E k = E ! + "!

Ek = E! ! ""

Absorption

Stimulated Emission

(resonant term)

(anti-resonant term)

For the case where only absorption contributes, Ek > E! , we have:

2

Pk! = bk =

or

Vk!

" ! k! " !
2

E02 k!

" ! k! " !

sin 2 #$ 12 ! k! " ! t %&

(2.155)

sin 2 #$ 12 ! k! " ! t %&

We can compare this with the exact expression:

2

Pk! = bk =

Vk!

" ! k! " !
2

+ Vk!

2
#1
sin 2 %
Vk! + ! k! " !
$ 2"

which points out that this is valid for couplings Vk!

&
t(
'

(2.156)

that are small relative to the

detuning !" = " k! # " . The maximum probability for transfer is on resonance ! k! = !

2-40

Limitations of this formula:

x3
By expanding sin x = x ! + , we see that on resonance !" = " k! # " $ 0
3!
2

V
lim
Pk t = k! 2 t 2
4"
!" # 0

()

(2.157)

This clearly will not describe long-time behavior. This is a result of 1st order perturbation theory
not treating the depletion of ! . However, it will hold for small Pk , so we require
t <<

2!
Vk"

(2.158)

At the same time, we cant observe the system on too short a time scale. We need the field to
make several oscillations for it to be a harmonic perturbation.
t>

1
1
(2.159)
"
! ! k!

These relationships imply that

Vk! << "! k! (2.160)

2-41

Adiabatic Harmonic Perturbation

What happens if we slowly turn on the harmonic interaction?

()

V t = V e!t cos" t
i"$
#i t
+ e# i"$ *
i" k" $ + !$ ' e
d
$
V
e
)
,
k"
! &#%
2
(
+
i
"
+
"
t
i
"
(
)
(
*
Vk" !t '
e k"
e k" #" )t
= e )
+
,
2!
)( # " k" + " + i! # " k" # " + i! ,+

bk =

Again, we have a resonant and anti-resonant term, which are now broadened by ! . If we only
consider absorption:
2

Pk = bk =

Vk!
4"

2
2

e 2!t

("

k!

#"

+ !2

which is the Lorentzian lineshape centered at ! k! = ! with width !" = 2# . Again, we can
calculate

the

adiabatic

limit,

setting ! " 0 .

We

will

calculate

the

rate

of

transitions ! k! = "Pk / "t . But lets restrict ourselves to long enough times that the harmonic
perturbation has cycled a few times (this allows us to neglect cross terms) ! resonances sharpen.

wk! =

2
!
%" # k! $ # + " # k! + # '
V
(
2" 2 k! &

) (