1

CHAPTER 2

Quantum Dynamics using the time dependent Schrdinger
equation


2-1. Formal solutions

i
H
t
c+
= +
c
(2.1)
If we find the complete orthonormal basis of eigenstates of

H , {
n
}, which satisfy

n n n
H E = (2.2)
then

0
( / ) ( )
0
( ) ( ) , with ( ) ( )
n
i E t t
n n n n
n
t c t c t e c t

+ = =

(2.3)


2-2. An example: the two level system
Consider a two level system whose Hamiltonian is a sum of a simple part,
0

H , and a
perturbation

V .

0

H H V = + (2.4)
The eigenfunctions of
0

H are ,
a b
| | , with the corresponding eigenvalues E
a
, E
b
. We
will interchangeably use the notation
,

i j i j
i O j O O | | = = for any operator

O.
Without loss of generality we may assume that
, ,
0
a a b b
V V = = (otherwise we may include
the diagonal part of V in H
0
). In the basis of the functions and
a b
| |

H is then
represented by the matrix

,
,

a a b
b a b
E V
H
V E
| |
=
|
\ .
;
*
, ,
| |
a b b a a b
V V V | | = = (2.5)
The coupling elements
, i j
V are in principle complex, and we express them as

*
, ,
;
i i
a b b a
V Ve V Ve
q q
= = (2.6)
2
with V taken real and positive. It should be emphasized that the two level problem
represented by

H is not more difficult to solve than that given by

0

H , however there are

situations where it helps to discuss the problem in terms of both Hamiltonians. For
example, the system may be represented by the Hamiltonian
0

H and exists in the

stationary state
a
| , then at some time taken to be 0 t = the perturbation

V is switched on.
A typical question is what is the probability ( )
b
P t to find the system in state b following
this switch-on of the perturbation that couples between the two eigenstates of
0

H .
The simplest approach to solving this problem is to diagonalize the Hamiltonian

H , i.e. to find its eigenstates, denoted and

+
, and eigenvalues and E E
+

respectively.

=======================
Problem 2-4: Show that given ( 0) 1; ( 0) 1 ( 0) 0
a b a
P t P t P t = = = = = = , then the
probability to find the system in state b at time t is give in terms of the eigenstates and
eigenvalues of

H by the form
( ) ( )
2
/ /
( ) | | | |
i E t i E t
b b a b a
P t e e | | | |
+

+ +
= +

(2.7)
=======================

Diagonalization of

H . The functions and

+
and eigenvalues and E E
+
are
solutions of the time independent Schrdinger equation

H E = . They are found by

writing a general solution in the form

a a b b
c c | | = + (2.8)
which in the basis of
a
| and
b
| is represented by the vector
a
b
c
c
| |
|
\ .
. The Schrdinger
equation in this representation is

,
,
a a b a a
b a b b b
E V c c
E
V E c c
| || | | |
=
| | |
\ . \ . \ .
(2.9)
The requirement of existence of a non-trivial solution leads to the secular equation
( )( )
2
a b
E E E E V = (2.10)
3
which yields two solutions for E given by

2 2
( ) 4
2
a b a b
E E E E V
E

+ +
= (2.11)
The following identities

a
b a
E E V V
X
V E E E E
+
+

= =

(2.12)
where X is real and positive, are also easily verified. The coefficients c
a
and c
b
satisfy

i a
b a
E E
c e c
V
q

= (2.13)
For the absolute values we have

2 2
| |
; | | | | 1
| |
a b
a b
a
E E c
c c
c V

= + = (2.14)
Consider first E=E
+


2
2 2
2 2
1
| | ; | |
1 1
a b
X
c c
X X
= =
+ +
(2.15)
The phase factor
i
e
q
in (2.13) may be distributed at will between c
a
and c
b
and a particular
choice is expressed by

/2 /2
cos sin
i i
a b
e e
q q
u | u |

+
= + (2.16a)

/2 /2
sin cos
i i
a b
e e
q q
u | u |

= + (2.16b)
or
/ 2
/ 2
cos sin
sin cos
i
a
i
b
e
e
q
q
| u u
u u
|

| |
| |
| |
| =
|
|
|
\ .
\ .
\ .
(2.17)
where
arctan ; 0 / 2 X u u t < < (2.18)
or

( ) ( )
1/2 1/2
2 2
1
; cos
1 1
X
Sin
X X
u u = =
+ +
(2.19)
The inverse transformation is

( )
/2
cos sin
i
a
e
q
| u u
+
= (2.20a)

( )
/ 2
sin cos
i
b
e
q
| u u

+
= + (2.20b)
4
Some readers may note that that Eqs. (2.17) and (2.20) constitute a rotation in 2-
dimensional space as seen in Fig. 2a.


Fig. 2a. The components
1
1
x
y
| |
|
\ .
and
2
2
x
y
| |
|
\ .
of a vector expressed
in the two systems of coordinates 1 and 2 shown are related to
each other by the transformation
2 1
2 1
cos sin
sin cos
x x
y y
u u
u u
=
| | | | | |
| | |
\ . \ . \ .
.



Calculating the time evolution. The required time evolution is now obtained from Eq.
(2.7). Using (cf. (2.16) and (2.20))
/2
/2
| | sin( )
| | cos( )
i
b a
i
b a
e
e
q
q
| | u
| | u
+
+
= =
= =
(2.21)
we get

( )
2
2
/ /
( ) | ( ) sin cos
iE t iE t
b b
P t t e e | u u
+

= + =

(2.22)
and

2
2
/ / 2 2
( ) | ( ) cos sin 1
iE t iE t
a a b
P t t e e P | u u
+

= + = + =

(2.23)

Using Eqs. (2.11),(2.12) and ((2.18)), Eq. (2.22) can be recast in the form

( )
2
2
2 2
4
( ) sin
2
4
ab
R
b
a b ab
V
P t t
E E V
O (
=
(

+
(2.24)
where

2 2
1
( ) 4| |
R a b ab
E E V O = +

(2.25)
is known as the Rabi frequency.


u
X
1
x
y
| |
|
|
\ .
Y
1
X
2
Y
2
5

2-7.3 Time dependent perturbation theory
The interaction representation is particularly useful in cases where the time evolution
carried by
0

H can be easily evaluated, and the effect of

V is to be determined
perturbatively. The Schrodinger equation in the interaction representation is a direct route
to such a perturbation expansion. We start by integrating it to get

1 1 1
0

( ) (0) ( ) ( )
t
I I I I
i
t dt V t t + = + +
}

(2.26)
and continue by substituting the same expression (2.26) for
1
( )
I
t + on the right. This
yields
1
2
1 1 1 2 1 2 2
0 0 0

( ) (0) ( ) (0) ( ) ( ) ( )
t t t
I I I I I I I
i i
t dt V t dt dt V t V t t
| | | |
+ = + + + + +
| |
\ . \ .
} } }

(2.27)
Continuing this procedure leads to
1 1
1 2 1 2
1
0 0 0

( ) 1 ... ( ) ( )... ( ) (0)
n
t t n t
I n I I I n I
n
i
t dt dt dt V t V t V t

=
| |
| |
| + = + +
|
|
\ .
\ .

} } }

(2.28)
Note that the order of the operators ( )
I
V t inside the integrand is important: these operators
do not in general commute with each other because they are associated with different
times. It is seen from Eq. (2.28) that the order is such that operators associated with later
times appear more to the left.


2-8. Quantum dynamics of the free particles
2-8.1. Free particle eigenfunctions
The Hamiltonian of a free particle of mass m is

2
2

2
H
m
= V

(2.29)
and the corresponding time independent Schrdinger equation is

2 2
k V = ;
2
2
2mE
k =

(2.30)
6
It is convenient to use the set of eigenfunctions normalized in a box of dimensions
( )
, ,
x y z
L L L with periodic boundary conditions
( , , ) ( , , )
x x y y z z
x y z x n L y n L z n L + = + + + + ; 0, 1, 2,... n = (2.31)
A set of such functions is

( )
1
( ) ; , ,
2
; 0, 1,... ; , ,
i
x y z
j j j
j
x y z
e k k k
k n n j x y z
L
L L L

= =
O
= = =
O =
k r
k
k r k
(2.32)
with the eigenvalues

2 2
2
k
E
m
=
k

(2.33)
These functions constitute an orthonormal set

3 *
' , '
| ' ( ) ( ) d o
O
=
}
k k k k
k k r r r (2.34)
and can be used to express the time evolution of any function ( , 0) t + = r that satisfies the
periodic boundary conditions (2.31). The wavefunction at time t is

( )
2 2
/ /(2 )
( , ) ( ,0)
i k m t
t e
(

+ = +
k k
k
r r

(2.35)
where

3
1
( ,0) ( ,0)
i
d re

O
+ = +
O
}
k r
k
r r (2.36)
We have seen that the normalization condition
2
| ( )| 1 dx x
O
=
}
implies that free
particle wavefunctions vanish everywhere like
1/2
( ) ~ x

O as O. The probability
2
| ( )| x dx to find the particle at position x...x+dx vanishes like
1
O in this limit. As such,
these functions are by themselves meaningless. We will see that meaningful physics can be
obtained in two scenarios: First, if we think of the process as undergone by a distribution
of N identical independent particles where the number N is proportional to the volume O
so that the density
2
( ) | ( )| x N x = is finite. We may thus work with the single particle
wavefunctions, keeping in mind that (a) such functions are normalized by
1/ 2
O and (b)
7
that physically meaningful quantities are obtained by multiplying observables by the total
number of particles to get the single particle density factored in.
Secondly, several observables are obtained as products of matrix elements that
scale like
2
( ) x (therefore like
1
O ) and the density of states that scales like O. A well
known example is the golden rule formula (9.25) for the inverse lifetime of a local state
interacting with a continuum. Such products remain finite and physically meaningful even
when O.
Anticipating such scenarios, we use in many applications periodic boundary
conditions as a trick to represent infinite systems by taking all the periodic box dimensions
to infinity at the end of the calculation. We will see several examples below.

2-8.3. Time evolution of a one dimensional free particle wavepacket
Consider now the time evolution of a free particle moving in one dimension that starts at
t=0 in the normalized state

( )
( )
2
0 0
1/4 2
2
1
( , 0) exp
4
2
x x ip x
x t
D
D t
(

(
+ = = +
(

(2.37)
We refer to the wavefunction (2.37) in the present context as a wavepacket: It is a local
function in position space that can obviously be expanded in the complete set of free
particle waves. Below we see that this function is also localized in momentum space.
============================
Problem 2-13. Show that for the wavefunction (2.37) the expectation values of the position and
momentum are

*
0
0
( , 0) ( , 0)
t
x dx x t x x t x

= + = + = =
}
(2.38)
and

*
0
0
( , 0) ( , 0)
t
p dx x t i x t p
x

c | |
= + = + = =
|
c
\ .
}
(2.39)
Also show that the position variance associated with this wavefunction is

( )
2
2 2 2
0
0 t
x x x x D
=
= = (2.40)
and that the momentum variance is

( )
2
2
2 2
0
2
0 4 t
p p p p
D =
= =

(2.41)
8
Note that
( )
1/2
2
0
0 t
x x x
=
(
A
(

and
( )
1/2
2
0
0 t
p p p
=
(
A
(

satisfy the Heisenberg
uncertainty rule as an equality:
0 0
/ 2 x p A A = . For this reason we refer to (2.37) as a minimum
uncertainty wavepacket.

===========================

The expansion of (2.37) in the set of eigenstates
(1/ 2)
( )
ik x
k
x L e

= yields

1
( , 0)
ikx
k
k
x t c e
L
+ = =

(2.42)
where
( )
( )
( )
2
0
0
1/ 4 2
2 2
1
( ,0) exp
4
2
k k
x x
c x dx i k k x
D
L D

(
= + =
(

}
(2.43)
where
0 0
/ . k p = Evaluating the Fourier transform in (2.43) we get
( )
1/ 4
2
2
2
0 0 0
2
8
exp ( )
k
D
c D k k ix k k
L
t
| |
(
=
|
( |

\ .
(2.44)
This implies that for a particle whose quantum state is (2.37), the probability to find it with
momentum k is

2 2 2
0
8
| | exp 2 ( )
k
D
c D k k
L
t
(
=

(2.45)
The time evolution that follows from Eq. (2.37) may now be found by using Eq.
(2.35). In 1-dimension it becomes

( )
2 2
/ /(2 )
1
( , )
ikx i k m t
k
k
x t c e
L
(

+ =


(2.46)
The probability that the system described initially by ( , 0) x t + = stays in this initial state is
given by

2
( ) ( , 0) ( , ) P t x t x t = + = + (2.47)
Using Eqs. (2.42) and (2.46) as well as the 1-dimensional version of (2.34) yields

( )
2 2
2
/ /(2 )
2
( )
i k m t
k
k
P t c e
(

=


(2.48)
Inserting Eq. (2.45) and converting the sum over k to an integral, ( ) /(2 )
k
L dk t

}

finally leads to
9

2
2 2 2
0
2
( ) exp 2 ( )
2
i
P t D dk tk D k k
m t

(
=
(

}

(2.49)
Note that the dependence on L does not appear in any observable calculated above.
We can also find the time dependent wavefunction explicitly. Combining Eqs.
(2.46), (2.44), and converting again the sum over k to an integral using / 2 L t = , leads to
( ) ( )
0 0
1/4
2 2
2
2
0 0
3
( , ) exp
2
2
ik x
D i k
x t e dk ik x x D k k t
m
t

| | (
+ =
| (
|
(
\ .
}

(2.50)
This Gaussian integral can be evaluated in a straightforward way. To get some physical
insight consider the result obtained from the initial wavepacket with
0 0 0
0 x p k = = = . In
this case (2.50) yields

1/ 4 1/ 2
2
2
1
( , ) exp
2 2
4 2 /
i t x
x t D
mD
D i t m
t

(
| | | |
+ = +
(
| |
\ . \ . + (

(2.51)
that leads to

( )
1/ 2
2 2
2 2
2 2
2
2 2 2 2 2
| ( , )| 2
4
exp
2 / 4
t
x t D
m D
x
D t m D
t

(

+ = +
( `
(
)
(
(

(
(
+
(

(2.52)
This wavepacket remains at the peak position x=0, with its width increases with time
according to
( )
( )
1/2 1/2
2
2 2 2 2 2
/ 4
2
t
t
x D t m D
mD

(
A = +


(2.53)



2-9. Quantum dynamics of the harmonic oscillator
2-9.1. Elementary considerations
The classical Hamiltonian for a 1-dimensional harmonic oscillator of mass m centered
about x=0,

2
2
1
2 2
p
H kx
m
= + (2.54)
10
implies the classical equations of motion
2
/ ; x x p m p me = = with
k
m
e = (2.55)
It is convenient for future reference to define the dimensionless position and momentum
|
where
m
x
e
o o = =

(2.56)
/ p m | e = (2.57)

In terms of these variables the Hamiltonian (2.54) takes the form

2 2
( )
2
H
e
| = +

(2.58)
and the classical equations of motion become
; e| | e = =

(2.59)

In quantum mechanics the momentum corresponds to the operator / p -i x = c c , or

/ i | = c c (2.60)
the position and momentum operator satisfy the familiar commutation relationship
[ , ] x p i =

[ , ] i | = (2.61)
and the quantum Hamiltonian is

2 2
2 2
2
1

2 2
d
H m x
m
dx
e = +

(2.62)
or in dimensionless form

2
2
2
1

;
2
H e

| |
c
= = +
|
|
c
\ .
(2.63)
The solutions of the time independent Schrdinger equation

H E = are the
(orthonormal) eigenfunctions and the corresponding eigenvalues
( )
( )
2
(1/ 2)
( )
x
n n n
x N H x e
o
o

= ( ) 1/ 2
n
E n e = + (2.64)
where ( )
n
H are the Hermit polynomials that can be obtained from

1 1
( ) 2 ( ) 2 ( )
n n n
H H nH
+
= (2.65)
11

0 1
( ) 1; ( ) 2 H H = = (2.66)
and N
n
is the normalization factor

1/2
2 !
n
n
N
n
o
t
= (2.67)
chosen so that
2
( ) 1.
n
dx x

=
}
In particular the ground state wavefunction and energy
are

2 2
(1/2)
0
( )
x
x e
o
o

t

= ;
0
(1/ 2) E e = (2.68)
The Hermite polynomials also satisfy the identity

1
( ) 2 ( )
n n
d
H nH
d

= (2.69)
Wwe use use Eqs. (2.60),(2.64), (2.65) and (2.69) to derive the following identities:

( )
( )
1 1
1 1
1

1
2
1

1
2
n n n
n n n
n n
n n
i

|
+
+
= + +
= +
(2.70)


2-9.2 The raising/lowering operators formalism
Focusing again on Eqs. (2.54)-(2.63) it is convenient to define a pair of operators,
linear combinations of the position and momentum, according to

1
2
1

2

( )
2
2

( )
2
2
m i
a x p i
m
m i
a x p i
m
e
|
e
e
|
e
= + = +
= =

(2.71)
so that


( ) ; ( )
2 2
m
x a a p i a a
m
e
e
= + =

(2.72)
Using Eq. (2.61) we find that a and

a satisfy

[ , ] 1 a a = (2.73)
and can be used to rewrite the Hamiltonian in the form
12

1 1


2 2
H a a N e e
| | | |
= + = +
| |
\ . \ .
. (2.74)
Here

N a a = is called the 'number operator' for reasons given below. This operator
satisfies the commutation relations

[ , ]

[ , ]
N a a
N a a
=
=
(2.75)

To see the significance of these operators use Eqs. (2.71) and

1 ; 1 1 a n n n a n n n = = + + (2.76)
where we have used n to denote
n
. The operators

a and a are seen to have the

property that when operating on an eigenfunction of the Harmonic oscillator Hamiltonian
they yield the eigenfunction just above or below it, respectively.

a and a will therefore be

refered to as the harmonic oscillator raising (or creation) and lowering (or annihilation)
operators, respectively.
Eq. (2.71) also leads to

N n n n = (2.77)
(hence the name number operator) and to the representation of the n-th eigenstate in the
form

1
( ) 0
!
n
n a
n
= (2.78)
Furthermore it is easy to derive the following useful relations:

1 1
1
n a n n
n a n n
= + +
=
(2.79)

', 1

', 1
'| |
'| | 1
n n
n n
n a n n
n a n n
o
o

+
=
= +
(2.80)
=========================================
Problem 2-16: Use Eqs. (2.72), (2.73) and (2.79) to prove that

( ) ', 1 ', 1
| | ' 1
2
n n n n
n x n n n
m
o o
e
+
= + +

(2.81)
=========================================

13
2-9.3 The Heisenberg equations of motion
An important advantage of formulating harmonic oscillators problems in terms of raising
and lowering operators is that these operators evolve very simply in time. Using the
Heisenberg equations of motion

,
i
A H A
(
=

, the expression (2.74) and the commutation

relations for a and

a leads to


0 0
( ) ( ) ; ( ) ( ) a t i a t a t i a t e e = =

(2.82)
where now

( ) and ( ) a t a t are in the Heisenberg representation. Eq. (2.82) yields the

explicit time dependence for these operators


( ) ; ( )
i t i t
a t ae a t a e
e e
= = (2.83)
Consequently, the Heisenberg representations of the position and momentum operators are

( ) ( ) ; ( ) ( )
2 2
i t i t i t i t
m
x t a e ae p t i a e ae
m
e e e e
e
e

= + =

(2.84)
As an example for the use of this formulation let us calculate the (in-principle time
dependent) variance,
2
( ) x t A , defined by ( ) ( )
2 2
*
( ) ( , ) ( , ) x t dx x t x x x t

A + +
}
for a
Harmonic oscillator in its ground state. Using the expression for position operator in the
Heisenberg representation from Eq. (2.84) and the fact that
2 2
0 ( ) 0 0 ( ) 0 x t x t A = for
an oscillator centered at the origin this can be written in the from

( )
2
2

0 ( ) 0 0 0 0 0
2 2
02 10
2 2
i t i t
x t a e ae a a aa
m m
a a
m m
e e
e e
e e

A = + = +
= + =


(2.85)
where we have also used the commutation relation (2.73).


2-9. 4. The shifted Harmonic oscillator
Problems involving harmonic oscillators that are shifted in their equilibrium positions
relative to some preset origin are ubiquitous in simple models of quantum dynamical
processes. We consider a few examples in this section.
The position shift operator. Consider the operator
14

( )

( )
x
U e

c c
(2.86)
Since the operator / x c c is anti-hermitian (i.e. ( )

/ / x x c c = c c )

( ) U is unitary
(
1

U U

= ) for real . The identity

2
2
2
1 ( 1)
( ) 1 ... ... ( ) ( )
2 !
n n
n
x
n
e x x x
x n
x x

c
| |
c c c
= + + + =
|
|
c
c c
\ .
(2.87)
identifies this unitary operator as the position shift operator. In terms of the operators a
and

a this operator takes the form

( )

( )
a a
U e

=

(2.88)
with
2
me
=

(2.89)
Under the unitary transformation defined by

U the position and momentum

operators transform in a simple way. For example, since

U is unitary, the following

identity must holds for all functions ( ) x and ( ) x |


( )| | ( ) ( )| | ( ) ( )| | ( ) x x x U x UxU U x x UxU x | | | = = (2.90)
For this identity to be true we must have

( ) ( ) U xU x = (2.91a)
Also, since p and

U commute it follows that

( ) ( ) U pU p = (2.91b)
Using Eqs. (2.91) and (2.71) it is easy to show also that

( ) ( )

( ) ( )
U aU a
U a U a


+
=
=
(2.92)
Appendix 2A (see entry 6) presents a more direct proof of these equalities using operator
algebra relationships obtained there.
Franck-Condon factors.

( )
2
(1,2) (2) (1)*
'
, '
( ) ( ) FC dx x x
v v
v v
_ _

=
}
(2.93)
15
For simplicity we will consider the case where =0, i.e. where
(2)
( ) x
v
_ is the ground
vibrational state on the harmonic surface 2. The desired FC factor is therefore
( ) ( )
2
(1,2)
*
0
,0 ,0
( ) ( ) ( ) FC FC dx x x
v
v v
_ _

= =
}
(2.94)
where both wavefunctions are defined on the same potential surface 1 whose explicit
designation is now omitted. Note that the only relevant information concerning the
electronic states 1 and 2 is the relative shift of their corresponding potential surfaces.
Using the shift operator

* ( )
0
( ) ( ) 0
a a
I dx x x e

v
_ _ v

=
}

(2.95)
Note v and 0 are states defined on the same harmonic potential and are not shifted
with respect to each other. Using Eq. (2.108) to replace
( )
exp ( ) a a

by
2
exp( (1/2) )exp( )exp( ) a a

, and using the Taylor expansion to verify that

exp( )|0 |0 a >= > this leads to

2
(1/ 2)
0
a
I e e

v

=

(2.96)
Again making a Taylor expansion, now of the operator exp( ) a

, it is easily seen that the
only term that contributes is ( / !)( ) a
v v
v

. Using also Eq. (2.78) leads to

2
(1/ 2)
!
I e
v

= (2.97)
Using also Eq. (2.89) finally yields the result
( )
( )
2
2
2
,0
/2
( ) | | exp
2 !
m
m
FC I
v
v
e
e

v
| |
= =
|
|
\ .

(2.98)
2-9. 5. Harmonic oscillator at thermal equilibrium
Harmonic oscillators are often used as approximate models for realistic systems. A
common application is their use as convenient models for the thermal environments of
systems of interest (see Section 6-5). Such models are mathematically simple, yet able to
account for the important physical attributes of a thermal bath: temperature, coupling
16
distribution over the bath normal modes and characteristic timescales. Their prominence in
such applications is one reason why we study them in such detail in this chapter.
The treatment of quantum systems in thermal equilibrium, and of systems
interacting with their thermal environments is expanded on in Chapter 10. For now it is
enough to recall the statistical mechanics result for the average energy of a harmonic
oscillator of frequency at thermal equilibrium

1
2
T
E n e
| |
= +
|
\ .
(2.99)
where
T
n is the average excitation, i.e. the average number of quanta e in the
oscillator, given by

1
1
n
n
T n
T
n
e n a a n
n a a
e e
| e
| e | e

= = =

(2.100)
In addition we may write

0
T T
T T
a a aa a a = = = = (2.101)
because the diagonal elements of the operators involved are zero.
===================
Problem 2-19: (a) Show that

1
1
T
aa
e
| e
=

(b) Use these results to find the thermal averages
2

T
x and
2

T
p , of the squared
position and momentum operators.
============================




Appendix 2A: Some operator identities
Here we derive some operator identities involving the raising and lowering operators of the
harmonic oscillators, which are used in Section 2-9 and in many applications discussed in
this book.
1.

1 1
[ ,( ) ] [ , ] ( ) ( )
n n n
a a a a n a n a

= =

(2.102a)
17

1 1
[ , ] [ , ]
n n n
a a a a na na

= =

(2.102b)
(note that (2.102b) is the Hermitian conjugate of (2.102a)). The proof can be carried by
induction. Assume that Eq. (2.102a) holds and show that

1
[ ,( ) ] ( 1)( )
n n
a a n a
+
= +

(2.103)
follows. The left hand side of (2.103) can be manipulated as follows:
1 1 1
[ ,( ) ] ( ) ( )
n n n
a a a a a a
+ + +
=
+

1
1 1 1
1
( ) ( )
( ) ( ) [( ) ( ) ] ( ) ( ) ( ) ( 1)( )
n n
n n n n n n n n
a a
a a a
a a a a a a a n a a a aa n a a a n a
+
+ +
+
+
= = + = + = + +

which yields the right hand side of (2.102a).
2. A corollary follows after observing that (2.102a) (say) can be written as
( )

[ ,( ) ] /
n n
x a
a a d dx x
=
(
=

. Since a function f(a

) is defined by its Taylor series, we

have, for every analytic function f

[ , ( )] ( )
x a
d
a f a f x
dx +
=
(
=
(

(2.104a)
and similarly

[ , ( )] ( )
x a
d
a f a f x
dx
=
(
=
(

(2.104b)
3. The identities (2.104) are special cases of the following theorem: If the operators

A and

B commute with their commutator

[ , ] A B , i.e.

[ ,[ , ]] [ ,[ , ]] 0 A A B B A B = = , then


, ( ) [ , ] '( ) A F B A B F B
(
=

(2.105)
To prove this identity we note that since

( ) F B can be expanded in powers of

B it suffices,
as in proving (2.104), to show that
1

[ , ] [ , ]
n n
A B A B nB

= . This is shown by repeated use of
the commutation relation to get
1 1 1

[ , ] ... [ , ]
n n n n n
AB BAB A B B B A n A B B

= + = = + .
4. We can use induction as above to prove the following identity
( ) ( 1)
n n
a a a a a a = +

(2.106)
and consequently also for an analytical function f(x)
18

( ) ( 1)
( ) ( 1)
f a a a a f a a
af a a f a a a
= +
= +


(2.107)
5. The following important identity holds for any two operators

A and

B under the
condition that, as in 3 above, both commute with their commutator:

1
2

[ , ] A B
A B A B
e e e e

+
= (2.108)
In particular

A and

B can be any linear combination of x , p , a and

a .
To prove (2.108) consider the operator

( )
At Bt
F t e e = defined in terms of two
operators

A and

B , and a parameter t. Take its derivative with respect to t

( ) ( )
At Bt At Bt At At
dF
Ae e e e B A e Be F t
dt

= + = + (2.109)
Next, use the identity


[ , ] [ , ]( )
At At
B e B A t e

= that follows from (2.105). From this and
the fact that

[ , ] B A commutes with

A it follows that


[ , ]
At At At
Be e B te B A

= . Using
the last identity in (2.109) leads to

( [ , ]) ( )
dF
A B t A B F t
dt
= + + (2.110)
The two operators,

A B + and

[ , ] A B commute with each other, and can be viewed as
scalars when integrating this equation. We get

1 1
2 2
2 2

( ) [ , ] [ , ]
( )

( ) (0)
A B t A B t A B t
A B t
F t F e e e
+ +
+
= = (2.111)
It remains to substitute 1 t = in Eq. (2.111) to obtain (2.108).
6. The identities (2.92) can now be verified directly. For example
2 2
2
( ) ( )
1 ( ) (1/ 2) [ , ] (1/ 2) ( )
2 3 (1/ 2) ( ) ( ) ( )

( ) ( )

( ) ( )

a a a a
a a a a a a a a a a
a a a a a a a a
U aU e ae
e ae e e e e e ae
e e e e a e e a
a

+
+ + + +
+ + +
=
=

=

(2.112)
where, in the steps marked 1 and 3 we used (2.111) and in the step marked 2 we used
(2.104a).