Chapter 3

Quantum dynamics
Time-dependent Schrödinger equation
The state of a given physical system at time t is denoted by a ket ψ (t ) ,
which is governed by

d
i ψ (t ) = H ψ (t )
dt

where H is the Hamiltonian operator of the system. The formal solution is

ψ (t ) = U (t ) ψ (0)

where U is the evolution operator which is unitary and satisfies

 dU (t )
 i = HU (t )
 dt
 U (0) = I
Evolution operator for systems with a time-independent Hamiltonian
d
i ψ (t ) = H ψ (t )
dt
ψ (t ) = U (t ) ψ (0)

2 n
 Ht  1  Ht  1  Ht 
U (t ) = e−iHt /  = 1 +  −i  +  −i  + .... +  −i  + ......
   2!    n!   

n = energy eigenvector
H n= En n → H
= ∑ En n n En = energy eigenvalue
n

= −iHt / 
U (t ) e= ∑ e −iEn t / 
n n
n

∞ ∞
=
Ψ (0) ∑ cn n →=
Ψ (t ) ∑ cn e−iE t /  n
n

=n 0=n 0
 Show that the evolution operator is unitary
+ +
U= =
U UU I

+
∞ n ∞  + n
( )
+ 1  Ht   1 H t
+
U (t ) =e −iHt / 
=∑  −i 
 =∑  +i 
 n!       
=  n 0=  n 0 n! 

∞ n
1  Ht 
= ∑  i  = eiHt / 
n =0 n !   

eiHt /  ⋅ e−iHt /  =
I But be careful if AB ≠ BA then
e A ⋅ e B ≠ e A+ B

The evolution of quantum state vectors, as governed by the

Schrödinger equation, are described by a unitary transformation.
 Consider { vn } is a complete orthonormal basis set,
and we know if the system starts at an initial state vn
the state of the system at time t is given by

φn (t ) = U (t ) vn

where U(t) is the evolution operator of the system.

Suppose φn (t ) is known for all n. Construct an explicit form of U(t).

Example: Electron spin in a constant magnetic field

• Intrinsic magnetic moment of an electron

M s = − g s µBS / 
= − g s µBσ / 2
− g s µ B (σ x xˆ + σ y yˆ + σ z zˆ ) / 2
=

=
where g s 2,= µ B e / 2m correspond to the gyromagnetic ratio and the
Bohr magneton respectively. The sigma’s are Pauli’s matrix:

0 1  0 −i  1 0 
σx =  
σy =  σz =  
 1 0  i 0   0 −1

• In a uniform magnetic field, the interaction energy is −M s ⋅ B , the Hamiltonian

is given by

H=−M s ⋅ B =γ B ⋅σ
Example: Electron spin in a constant magnetic field
d
H ψ =γ ( B ⋅ σ ) ψ =
i ψ
dt
In matrix representation:

 H ↑↑ H ↑↓  α 
H = ψ = 
 H ↓↑ H ↓↓  β 

Time evolution operator for time-independent H (constant B field) :

ψ (t ) = e−iHt /  ψ (0)

Case 1: If B is in the z-direction (i.e., B = B0 zˆ )

 e−iγ B0t /  0 
−iγ B0σ z t / 
e −iHt /  e=
=  
 eiγ B0t /  
 0

Note that the eigenvalues of the Hamiltonian H are ±γ B0

Case 2: B in the x-direction (i.e., B = B0 xˆ )

 cos Ωt
−iHt /  −iγ B0σ xt /  −i sin Ωt 
= e=  Ω =γ B0 / 
cos Ωt 
e
 −i sin Ωt

The oscillation is called Rabi oscillation.

1 −iγ B0σ xt /   1   cos Ωt 

ψ (0) =  = ψ (t ) e=  0   −i sin Ωt 
 
0    

At what initial state vectors, there are no Rabi

oscillations?
Example: Wave packet spreading of a free particle

2 ∂ 2 ∂
Schrödinger equation: − Ψ ( x , t ) = i  Ψ ( x, t )
2m ∂x 2 ∂t

Energy eigenfunctions and eigenvalues: eikx ,  2 k 2 / 2m

A general solution is a superposition of energy eigenfunctions:

k 2
1 ∞ −i t
Ψ ( x, t ) = ∫ ϕ (k )e e 2m dk
ikx
2π −∞

1 ∞ −ikx
ϕ (k )
= ∫
2π −∞
Ψ ( x ,0) e dx k-space representation

| ϕ (k ) |2 = probability density of the particle in the k-space

∞
Normalization condition: ∫−∞ | ϕ (k ) |2 dk = 1
Example: Gaussian wave packet

A general gaussian wave packet:

Ψ ( x) π −1/ 4σ 1/ 2 × exp  −σ 2 ( x − x0 ) 2 / 2 + ik0 x 

=
 

 (k − k0 ) 2  −ikx
ϕ=
(k ) (πσ ) 2 −1/ 4
exp  −
 2σ 2 

×e 0

We need 3 parameters:

k0 → Mean k or the average momentum p = k0

x0 → Mean position x = x0
1  2σ 2
σ→ Widths (∆x) 2 = (∆p ) 2
=
2σ 2 2

How do we reach the classical particle limit ?

Free evolution of a Gaussian wave packet in position space

1 ∞
∫−∞ exp  −( k − k0 )2 / 2σ 2  eik ( x − x0 ) e −ik
2
Ψ ( x, t ) t / 2m
dk
 
(πσ )2 1/ 4
2π
1/ 2
−1/ 4  σ   ik0 x − σ 2 x 2 / 2 − ik02t / 2m 
=π   × exp  
1 + iσ t / m 
2
 1 + iσ t / m
2


Probability density in space:

2
P ( x, t ) = Ψ ( x, t )

−1/ 2 σ  σ 2 ( x − k 0 t / m ) 2 
π × exp  − 
 1 + σ  t / m 
4 2 2 2
1 + σ 4  2t 2 / m 2

No need to memorize these equations! Try to see the physical picture.

A picture of the spreading of a free wave packet

Only the real part of the wave function is shown here.

w0

t=0

t=T

vgT wt
  1
=
Example: Gaussian wave packet in a harmonic oscillator H ω  a + a + 
 2
A coherent state is defined by

α is a complex number

+
where =
a a n n=
n n 0,1, 2,3.... (see the supplementary topic 5)

The coherent state is a right eigenvector of the annihilation operator with the
eigenvalue α

A coherent state in position representation 𝜓𝜓𝛼𝛼 𝑥𝑥 = 𝑥𝑥 𝛼𝛼 is a Gaussian wave

packet !

Peak momentum Peak position

(divided by ℏ)
 ∞
−|α |2 / 2 αn
Ψ (0) =
e ∑ n!
n = |𝛼𝛼〉 Gasussian wave packet
n =0

−|α |2 / 2−iωt / 2
−iωt / 2
∞
α ne−inωt
−iωt
Ψ (t ) e ∑ n!
= n e α e Gasussian wave packet
n =0

Therefore a Gaussian wave packet evolves in SHO remains in a Gasussian

wave packet, but with time-dependent position and momentum.

In terms of the evolution operator of SHO,

1
−𝑖𝑖𝑖𝑖𝑖𝑖 −𝑖𝑖𝑖𝑖𝑖𝑖(𝑎𝑎+ 𝑎𝑎+2)
𝛼𝛼𝑒𝑒 = 𝑈𝑈 𝑡𝑡 𝛼𝛼 = 𝑒𝑒 𝛼𝛼
Example: Generation of Schrödinger cat states in a harmonic oscillator
Answers:

(a)

(b)
where
Example: Generation of multiparticle entanglement (Molmer and Sorensen PRL 1998)

An interesting question is how to generate the following state in

a system of N two-level atoms:

A trick is to make use of nonlinear interaction between atoms,

where the N two-level atoms are treated as angular momentum states

By letting the system evolves under the given Hamiltonian, the state of the
system at time t is given by

In particular at
Matrix exponentials approximation

H = H1 + H 2 + .. + H K e.g. H= T + V

where [ H i , H j ] ≠ 0 and so the evolution operator cannot be expressed

as a product of evolution operators of individual hamiltonians,

e −iHt /  = e −i ( H1 + H 2 +..+ H K )t / 
≠ e −iH1t /  e −iH 2 t /  ...e −iH K t / 
Suzuki-Lie-Trotter formula:

e −iHt /  lim ( e −iH1ε / N  e −iH 2ε /  ...e −iH K ε /  ) N , ε t / N

=
N →∞
−iH1ε / N  −iH 2ε / 
(e e ...e −iH K ε /  ) N + higher terms

Useful in simulating quantum evolution when you can only realize Hj one by one.
A useful trick of engineering the evolution operator

exp ( iH 2δ t ) exp ( iH1δ t ) exp ( −iH 2δ t ) exp ( −iH1δ t )

exp ([ H1 , H 2 ]δ t ) + O(δ t 3 )

If H1 and H 2 commute with [ H1 , H 2 ] , then the above expression

is exact.
Retarded Green’s operator and resolvent

The retarded Green’s operator K is defined from the evolution operator U by:

K + (t , t ') U (t , t ')θ (t − t ')

=
Given that the system Hamiltonian is H, we have

 d 
 i  − H δ (t − t ')
 K + (t , t ') =
 dt 

Let K + (t , t ') = K + (τ = t − t ') . Taking the Fourier Transform:

−1 ∞ −iEτ / 
K + (τ ) =
2π i ∫−∞
dEe G+ ( E )

1 ∞
G+ ( E ) = ∫ dτ eiEτ /  K + (τ )
i −∞
Since K + (τ ) = e −iHτ / θ (τ )
1 ∞
G+ ( E ) = ∫ dτ eiEτ /  e −iHτ / 
i 0
1 ∞
= lim+ ∫ dτ ei ( E − H +iη )τ / 
η →0 i  0
1
= lim+
η →0 E − H + iη

The resolvent is defined by the operator :

1
G( z) = Hence=
G+ ( E ) lim+ G ( E + iη )
z−H η →0

where z is a complex number.

For a general Hamiltonian with both discrete and continuous energy spectrum,

𝑒𝑒𝑛𝑛 〈𝑒𝑒𝑛𝑛 | 𝐸𝐸 〈𝐸𝐸|

𝐺𝐺 𝑧𝑧 = � + � 𝑑𝑑𝑑𝑑
𝑧𝑧 − 𝐸𝐸𝑛𝑛 𝑧𝑧 − 𝐸𝐸
𝑛𝑛
Advanced Green’s Functions

𝐾𝐾− 𝑡𝑡, 𝑡𝑡 ′ = −𝑈𝑈 𝑡𝑡, 𝑡𝑡 ′ 𝜃𝜃(𝑡𝑡 ′ − 𝑡𝑡)

𝑑𝑑
(𝑖𝑖𝑖 − 𝐻𝐻)𝐾𝐾− 𝑡𝑡, 𝑡𝑡 ′ = 𝑖𝑖𝑖𝛿𝛿(𝑡𝑡 − 𝑡𝑡 ′ )
𝑑𝑑𝑑𝑑

1 +∞
𝐾𝐾− 𝜏𝜏 = − � 𝑑𝑑𝑑𝑑𝑒𝑒 −𝑖𝑖𝑖𝑖𝑖𝑖/ℏ 𝐺𝐺_(𝐸𝐸)
2𝜋𝜋𝜋𝜋 −∞

1
𝐺𝐺− 𝐸𝐸 = lim+ = lim+ 𝐺𝐺(𝐸𝐸 − 𝑖𝑖𝑖𝑖)
𝜂𝜂→0 𝐸𝐸 − 𝐻𝐻 − 𝑖𝑖𝑖𝑖 𝜂𝜂→0

The evolution operator can be expressed as

1 +∞
𝑈𝑈 𝜏𝜏 = 𝐾𝐾+ 𝜏𝜏 − 𝐾𝐾− 𝜏𝜏 = � 𝑑𝑑𝑑𝑑𝑒𝑒 −𝑖𝑖𝑖𝑖𝑖𝑖/ℏ [𝐺𝐺_(𝐸𝐸) − 𝐺𝐺+ (𝐸𝐸)]
2𝜋𝜋𝜋𝜋 −∞
Mathematical facts:

Therefore

which is related to the density of state of the system.

Integral equation for the Green’s operator

Consider the system Hamiltonian is

=
H H0 + V

then
dU (t , t ') U (t ', t ') = 1
i = ( H 0 + V )U (t , t ')
dt

Assuming the evolution operator is U 0 for V=0, then

1 t
=
U (t , t ') U 0 (t , t ') + ∫ dτ U 0 (t ,τ )VU (τ , t ')
i t '
1 +∞
Therefore K + (t , t ') K0+ (t , t ') + ∫ dt1 K0+ (t , t1 )VK + (t1 , t ')
=
i −∞
+∞ 𝑡𝑡
where =
K0+ (t , t ') U 0 (t , t ')θ (t − t ') � 𝑑𝑑𝑡𝑡1 𝜃𝜃 𝑡𝑡 − 𝑡𝑡1 𝜃𝜃 𝑡𝑡1 − 𝑡𝑡 ′ = � 𝑑𝑑𝑡𝑡1
−∞ 𝑡𝑡 ′
By convolution, G+ ( E ) satisfies an algebraic equation (not an integral equation)

=
G+ ( E ) G0+ ( E ) + G0+ ( E )VG+ ( E )

Series expansion of the resolvent :

=
G ( z ) G0 ( z ) + G0 ( z )VG ( z )

G ( z ) =+
G0 ( z ) G0 ( z )VG0 ( z ) + G0 ( z )VG0 ( z )VG0 ( z ) + ....

In terms of matrix elements with respect to unperturbed eigenvectors

Gkl ( z ) = φk G ( z ) φl H 0 φn = En φn

1 1 1 1 1 1
Gkl ( z ) = δ kl + Vkl +∑ Vki Vil + ....
z − Ek z − Ek z − El i
z − Ek z − Ei z − El
Example

H = E0 ψ 0 ψ 0 + ∑ ε k k k + ∑ (Vk 0 k ψ 0 + V0k ψ 0 k )
k k

1 1 1 1
G=
00 ( z ) +∑ V0k Vk 0
z − E0 k z − E0 z − εk z − E0

1 1 1 1 1
+∑ V0k ' Vk '0 V0k Vk 0 ....
kk '
z − E0 z − εk ' z − E0 z − εk z − E0

1  C C2  1
= 1 + + + ...  =
z − E0  z − E0 ( z − E0 )
2
 z − E0 − C

𝑉𝑉0𝑘𝑘 2
𝐶𝐶 𝑧𝑧 = � = Δ 𝑧𝑧 − 𝑖𝑖Γ(𝑧𝑧)
𝑧𝑧 − 𝜀𝜀𝑘𝑘
𝑘𝑘
Projecting the resolvent G(z) onto a subspace

Projection operators:

𝑃𝑃 = 𝜙𝜙1 〈𝜙𝜙1 | + 𝜙𝜙2 〈𝜙𝜙2 | + ⋯ 𝜙𝜙𝐾𝐾 〈𝜙𝜙𝐾𝐾 |

𝑄𝑄 = 𝐼𝐼 − 𝑃𝑃

where 𝜙𝜙1 , 𝜙𝜙2 , … . 𝜙𝜙𝐾𝐾 are some (but not all) eigenvectors of 𝐻𝐻0
These projection operators satisfy the following relations:

𝑄𝑄 = 𝑄𝑄 2 𝑃𝑃 = 𝑃𝑃2
𝑃𝑃𝑃𝑃 = 𝑄𝑄𝑄𝑄 = 0
𝑃𝑃, 𝐻𝐻0 = [𝑄𝑄, 𝐻𝐻0 ] = 0
𝑄𝑄𝐻𝐻0 𝑃𝑃 = 𝑃𝑃𝐻𝐻0 𝑄𝑄 = 0

Our task is to work out an equation for 𝑃𝑃𝑃𝑃 𝑧𝑧 𝑃𝑃 for the system with 𝐻𝐻 = 𝐻𝐻0 + 𝑉𝑉
 Starting from 𝑧𝑧 − 𝐻𝐻0 − 𝑉𝑉 𝐺𝐺 𝑧𝑧 = 1 , one can obtain the following equations:

𝑃𝑃 𝑧𝑧 − 𝐻𝐻 𝑃𝑃 𝑃𝑃𝑃𝑃 𝑧𝑧 𝑃𝑃 − 𝑃𝑃𝑃𝑃𝑃𝑃 𝑄𝑄𝑄𝑄 𝑧𝑧 𝑃𝑃 = 𝑃𝑃

−𝑄𝑄𝑄𝑄𝑄𝑄 𝑃𝑃𝑃𝑃 𝑧𝑧 𝑃𝑃 + 𝑄𝑄 𝑧𝑧 − 𝐻𝐻 𝑄𝑄 𝑄𝑄𝑄𝑄 𝑧𝑧 𝑃𝑃 = 0

The second equation gives

𝑄𝑄
𝑄𝑄𝑄𝑄 𝑧𝑧 𝑃𝑃 = 𝑉𝑉𝑉𝑉𝑉𝑉 𝑧𝑧 𝑃𝑃
𝑧𝑧 − 𝑄𝑄𝐻𝐻0 𝑄𝑄 − 𝑄𝑄𝑄𝑄𝑄𝑄

Substituting it into the first equation gives,

𝑄𝑄
𝑃𝑃 𝑧𝑧 − 𝐻𝐻 𝑃𝑃 𝑃𝑃𝑃𝑃 𝑧𝑧 𝑃𝑃 − 𝑃𝑃𝑃𝑃𝑃𝑃 𝑉𝑉𝑉𝑉𝑉𝑉 𝑧𝑧 𝑃𝑃 = 𝑃𝑃
𝑧𝑧 − 𝑄𝑄𝐻𝐻0 𝑄𝑄 − 𝑄𝑄𝑄𝑄𝑄𝑄

𝑄𝑄
𝑖𝑖. 𝑒𝑒., 𝑃𝑃 𝑧𝑧 − 𝐻𝐻0 − 𝑉𝑉 − 𝑉𝑉 𝑉𝑉 𝑃𝑃𝑃𝑃 𝑧𝑧 𝑃𝑃 = 𝑃𝑃
𝑧𝑧 − 𝑄𝑄𝐻𝐻0 𝑄𝑄 − 𝑄𝑄𝑄𝑄𝑄𝑄
 𝑄𝑄
𝑃𝑃 𝑧𝑧 − 𝐻𝐻0 − 𝑉𝑉 − 𝑉𝑉 𝑉𝑉 𝑃𝑃𝑃𝑃 𝑧𝑧 𝑃𝑃 = 𝑃𝑃
𝑧𝑧 − 𝑄𝑄𝐻𝐻0 𝑄𝑄 − 𝑄𝑄𝑄𝑄𝑄𝑄

1
𝑃𝑃𝑃𝑃 𝑧𝑧 𝑃𝑃 = 𝑃𝑃
𝑄𝑄
𝑃𝑃 𝑧𝑧 − 𝐻𝐻0 − 𝑉𝑉 − 𝑉𝑉 𝑉𝑉
𝑧𝑧 − 𝑄𝑄𝐻𝐻0 𝑄𝑄 − 𝑄𝑄𝑄𝑄𝑄𝑄

𝑃𝑃 1
= =
𝑃𝑃 𝑧𝑧 − 𝐻𝐻0 − 𝑅𝑅(𝑧𝑧) 𝑧𝑧 − 𝑃𝑃𝑃𝑃0 𝑃𝑃 − 𝑃𝑃𝑃𝑃 𝑧𝑧 𝑃𝑃

where 𝑅𝑅(𝑧𝑧) is called the level-shift operator defined by

𝑄𝑄
𝑅𝑅 𝑧𝑧 = 𝑉𝑉 − 𝑉𝑉 𝑉𝑉
𝑧𝑧 − 𝑄𝑄𝐻𝐻0 𝑄𝑄 − 𝑄𝑄𝑄𝑄𝑄𝑄

𝑄𝑄 𝑄𝑄 𝑄𝑄
= 𝑉𝑉 + 𝑉𝑉 𝑉𝑉 + 𝑉𝑉 𝑉𝑉 𝑉𝑉 + ⋯ . .
𝑧𝑧 − 𝐻𝐻0 𝑧𝑧 − 𝐻𝐻0 𝑧𝑧 − 𝐻𝐻0

𝑄𝑄 𝑄𝑄
Since =
𝑧𝑧 − 𝑄𝑄𝐻𝐻0 𝑄𝑄 𝑧𝑧 − 𝐻𝐻0
Differential equations

In many problems, we do not have to solve the evolution operator , but only
need the solution from a given initial state. In such cases, it would be simpler to
solve the Schrödinger equation directly.

Let ψ (t ) = ∑ cn (t ) en cn (t ) complex amplitudes

n

d
i ψ (t ) = H ψ (t )
dt

⇒ i ∑ cn (t ) en =
H ∑ cm (t ) em
n m

∑ cm (t ) en H em
⇒ icn (t ) =
m

which is a set of coupled (first order) ordinary differential equation.

Example: Two-level atom in a resonant laser field (electric dipole interaction)

A two-level atomic model requires:

• Driving field frequency is close to

to the frequency difference between
the two involved levels.

• Driving field strength is not too

strong

Where Ω is a coupling constant proportional laser field strength and

electric dipole moment
ψ (t ) cg (t ) g + ce (t ) e
=

By Schrödinger equation:
icg = Ω*eiωt ce

ice ω Ace + Ωe −iωt ce

=

But we can get rid of the time-dependent phase by working in the new variables

=cg (t ) c=
g (t ), ce (t ) ce (t )e −iωt

Then the amplitude differential equations become:

icg = Ω*ce
= (ω − ω )c + Ωc
ic e A e g
This is equivalent to a time-independent system with the Hamiltonian

 ωA − ω Ω 
H'= 
 Ω *
0 
Rabi oscillations in a two-level atom in a resonant laser field
Large detuning limit:

 ∆ Ω
Rotating frame Hamiltonian: H'= * 
Ω 0

Eigenvalues and Eigenvectors in the large detuning case:

−= g + ... ε − ≈ −Ω2 / ∆
+ = e + ... ε + ≈ ∆ + Ω2 / ∆

Going back to the original frame: No transition but pick up a phase

 Ω 2t    Ω2  
g → exp  i e → exp  −i  ω A +  t  e
 ∆ 
g
   ∆
   

Ω 2
i.e., the energies are shifted by ± which are called AC Stark shifts
∆
Trapping atoms by laser using AC Stark shift
If the laser field has a spatial profile then the atom would experience state
dependent effective potentials:
Ω 2 ( x )
Vg ( x ) = −
∆
Ω 2 ( x )
Ve ( x ) = +
∆
This is the mechanism of FORT
Quantum decay problem
Consider a state ψ0 couples to a continuum with the system Hamiltonian

H=
k k
(
 ∑ ∆ k k k +V ∑ ηk k ψ 0 + ηk* ψ 0 k )
e.g. Spontaneous emission
in two-level atom
ψ0 k
ψ 0 = e, vacuum
Set its energy =0

k = g , photon (ωk )

An initial system in ψ0 is unstable, and the system will decay to other states as time
increases

t ) a (t ) ψ 0 + ∑ ck (t ) k
Ψ (=
k
=
a (0) 1,=
ck (0) 0
Schrödinger equation:

d ia = V ∑ηk*ck
i ψ (t ) = H (t ) ψ (t ) k
dt ick =
∆ k ck + V ηk a

By formal integration,

ck (t ) = −i ∫ Vηk a (t ')e −i∆ k ( t −t ') dt '

t
0

−V 2 ∑ | ηk |2 ∫ a (t ')e −i∆ k ( t −t ') dt '

t
∴ a (t ) =
0
k

Now if ηk is independent of k, then V 2 ∑ | ηk |2 e −i∆ k ( t −t ') ≈ γδ (t − t ')

k
γ
∴ a (t ) = e −γ t / 2
− a (t ) ⇒ a (t ) =
2

ck (t ) = −i ∫ Vηk e −γ t '/ 2 e −i∆ k ( t −t ') dt '

−γ t / 2 t
=
Ψ (t ) e ψ 0 + ∑ ck (t ) k 0
k
Exponential decay
 ∞
ck (t → ∞ ) = −i ∫ Vηk e −γ t '/ 2 e −i∆ k ( t −t ') dt '
0
V ηk
= e −i∆ k t
i∆ k − γ / 2

ψ0

ψ 0 = e, vacuum ∑ ck (t ) g , photon(ωk )
k
Emitted photon has a natural linewidth γ
Why exponential decay ?

Key approximation: V 2 ∑ | ηk |2 e −i∆ k ( t −t ') ≈ γδ (t − t ')

k
This approximation assumes : (1) ηk = η a constant
(2) Summation involves −∞ < ∆ k < ∞ such that
∞
∫−∞ ei ∆ ( t − t =
')
d ∆ 2πδ (t − t ')

A more general way of solving the problem without (1) and (2) is by using the Markov
approximation:

a (t ) = −V 2 ∑ | ηk |2 ∫ a (t ')e −i∆ k ( t −t ') dt ' ≈ − a (t )V 2 ∑ | ηk |2 ∫ e −i∆ k ( t −t ') dt '

t t
0 0
k k

which is justified as long as the correlation time in the function ∑

| ηk |2 e −i∆ k ( t −t ')
is short compared with the evolution time scale of a(t). k
Next by writing
γ'
γ' −( +iδ ) t
V 2
∑ | ηk |2 t −i ∆ k ( t −t ')
0∫e dt ' ≈
2
+ iδ ⇒ a (t ) =
e 2
k
Frequency shift
Time-dependent Hamiltonian

d
Schrödinger equation: i ψ (t ) = H (t ) ψ (t )
dt
H (=
t ) H 0 + V (t )
p2
e.g. H (t ) = + V ( x ) + xE cos ωt
2m

Interaction Picture ψ (t ) I
:

ψ (t ) = e −iH 0t ψ (t ) I

d d
i ψ (t ) =H (t ) ψ (t ) ⇒ i ψ (t ) I
=VI (t ) ψ (t ) I
dt dt

where
VI (t ) = eiH 0tV (t )e −iH 0t
Formal solution in the interaction picture
Integrate the Schrödinger equation in the interaction picture:
i t
ψ
= (t ) I ψ (0) I − ∫
 0
VI (t1 ) ψ (t1 ) I dt1

i t  i t1 
ψ (t ) ψ (0) I − ∫ dt1VI (t1 )  ψ (0) I − ∫ VI (t2 ) ψ (t2 ) I dt2 
=
I  0   0 
2
i t  i t t1
= ψ (0) − ∫ VI (t1 ) ψ (0) I dt1 +  −  ∫0 dt1 ∫0 dt2VI (t1 ))VI (t2 ) ψ (t2 ) I
I  0  
2
i t  i t t1
= ψ (0) − ∫ VI (t1 ) ψ (0) I dt1 +  −  ∫0 dt1 ∫0 dt2VI (t1 ))VI (t2 ) ψ (0) I
I  0  
3
 i t t1 t2
+− 
 
∫0 dt1 ∫0 dt2 ∫0 dt3VI (t1 ))VI (t2 )VI (t3 ) ψ (0) I + ....
Dyson Series for the evolution operator in the interaction picture

2
i t  i t t1
U I (t ) = I − ∫ VI (t1 )dt1 +  −  ∫0 dt1 ∫0 dt2VI (t1 )VI (t2 )
 0  
3
 i t t1 t2
+ − 
 
∫0 dt1 ∫0 dt2 ∫0 dt3VI (t1 )VI (t2 )VI (t3 ) + ....
∞ n
 i
I + ∑− 
t t1 tn −1
= ∫0 dt1 ∫ dt2 ....∫ dtn VI (t1 )VI (t2 )......VI (tn )
n =1 
 0 0

Note that if H (t ) = H is a constant operator, then

2 n
 Ht  1  Ht  1  Ht 
U (t ) = 1 +  −i  +  −i  + .... +  −i  + ...... = e −iHt / 
   2!    n!   
Coupled equations in the interaction picture
Now consider the Hamiltonian of the system as

H (=
t ) H 0 + V (t )
We may choose the basis vectors as defined by H 0 ,

H 0 m = ωm m ψ (t ) = ∑ cn (t ) n
n

∑ cm (t ) n H (t ) m / 
∴ icn (t ) =
m

= ωn cn (t ) + ∑ cm (t ) n V (t ) m
m
−iω t
We can remove the first term in the right side by letting cn (t ) = e n cn (t )
This is precisely the interaction picture:

icn (t ) = ∑ cm (t )e ( m n ) n V (t ) m
−i ω −ω t

m
Time-dependent perturbation theory: Weak perturbation V(t)

ψ (t ) = ∑ cn (t ) n cn (t ) = e −iωn t cn (t )
n

icn (t ) = ∑ cm (t )e ( m n ) n V (t ) m
−i ω −ω t

m
Now for a weak V(t), we consider the perturbation series:

cn (t ) ≈ cn (0) (t ) + cn(1) (t ) + ...

where
cn (0) (t ) = cn (0) zero-th order (solution of V=0)

cn (1) (t ) = amplitude to first order in V

Assuming ψ (0) = k i.e., cn (0) = δ nk

(0) −i (ωm −ωn )t −i (ωk −ωn )t

∴ icn(1) (t ) ∑m

c e = n V ( t ) m e n V (t ) k
m
 e ( k n ) n V (t ) k
−i ω −ω t
∴ icn(1) (t ) =

cn(1) (t ) = −i ∫ dt 'e ( k n ) n V (t ') k

t −i ω −ω t '
0

Therefore an initial k state can make a transition to the state n due

to the time-dependent potential V(t)

k n

For the first order theory to work, how small should V(t) be?

| ∫ dt 'e ( k n ) n V (t ') k | << 1

t −i ω −ω t '
Ans:
0

V (t ) V0 eiΩt + h.c.
Specifically, if = we need

n V0 k
| | << 1
ωn − ωk ± Ω
Bound to free transition (e.g. Photo-ionization)

Free states e j
e.g. In photo-electric effect an electron
makes a transition from the bound state
to a free state by an external light field.

Bound state |g>

Assuming the model has the Hamiltonian:

ωg g
H = g +  ∑ ωk ek
k
ek +F (t )∑ ηk ek
k
( g + ηk* g ek )
where=
F (t ) F0 eiΩt + c.c. corresponds to an external harmonic driving field

The system state at time t is: ψ (t ) a (t ) g + ∑ ck (t ) ek

=
k
=
a (0) 1,=
ck (0) 0
 ia ωg a + F (t ) ∑ηk*ck
=
k
ick ωk ck + F (t )ηk a
=

−iωk t
− i ∫ dt ' F (t ')ηk a (t ')e −iωk ( t − t ')
t Formal solution
=
ck (t ) ck (0)e
0 (check it out)

−iωk t t i (ωk −ωg ) t ' −iωg t

approximation a (t ') ≈ e
≈ −iηk e
0∫dt ' F (t ')e

Then the transition probability, | ek ψ (t ) |2, to 1st order is given by:

2
t i (ωk −ωg −Ω )t ' i (ωk −ωg +Ω )t '
Pg →ek ≈ ηk F0 ∫ dt 'e Why do we drop the e part ?
0

4 | F02ηk2 | sin 2 [(ωk − ωg − Ω)t / 2]

=
(ωk − ωg − Ω)2
Mathematical remark:

sin 2 x
x2

sin 2 [ ∆t ]
lim = πδ ( ∆ )
t →∞ t ∆ 2

sin 2 [(ωk − ωg − Ω)t / 2] πt

→ δ (ωk − ωg − Ω)
(ωk − ωg − Ω) 2 2
Therefore the transition rate

Pg →ek 2π
≈ | F02ηk2 | δ (ω f − ωi − Ω)
t 

Since the total transition probability include all accessible final states,

=P ∑
= Pg →k ∫ d ωk D (ωk )Pg →k
k
Density of state
Therefore the total transition rate is

dP 2π
∴ ≈ | F02ηk2 | × D (ω=
k ωg + Ω)
dt 

Density of state at the energy

conserving frequency.
which is the Fermi’s Golden Rule.
 To summarize, based on the first order perturbation theory, when
a quantum system driven by a weak harmonic V(t), the system can
make a transition from an initial state |i> to a set of final states |f>
Allowed by energy conservation. The transition rate is given by

Transition rate ∝ |Transition matrix element between the initial and final states|2
× Density of final states whose energies satisfy the
energy conservation
Heisenberg’s picture
d
In Schrödinger's picture: i ψ (t ) = H ψ (t ) the state vector changes with time.
dt
dU (t )
Let ψ (t ) = U (t ) ψ (0) then i = HU (t )
dt

The expectation value of the operator M at time t is given by,

ψ (t ) M ψ (t ) = ψ (0) U + (t ) MU (t ) ψ (0)


Time-independent
Time-dependent
Time-dependent
Time-independent
operator
operator
M (t ) ≡ U + (t ) MU (t )

In Heisenberg’s picture, the operators change but state vector remains

unchanged.
Heisenberg’s equation of motion

Heisenberg operator: M (t ) ≡ U † (t ) MU (t )
M (0) ≡ M M is a “Schrödinger operator”, and we
assume that it is a constant operator
d d  +
M (t ) = U (t ) MU (t ) 
dt dt  

= U + (t ) MU (t ) + U + (t ) MU (t )

i + i + dU (t )
= U (t ) HMU (t ) − U (t ) MHU (t ) by i = HU (t )
  dt
i +
= U (t ) ( HM − MH )U (t )

i
= [ H (t ), M (t ) ] H (t ) ≡ U + (t ) HU (t )


Note: If [ H , M ] = 0 , then M (t ) = M (0) does not change with time

i.e., constant of motion independent of the state.
Example:

p2
Hamiltonian: =
H + V ( x)
2m
dx(t ) i p (t )
=
 =[ H (t ), x (t ) ]
 dt  m

 dp (t ) = i i ∂V
 dt [ H (t ), p (t ) ] [
= V ( x (t )), p (t ) ] = −
  ∂x x = x (t )

Same form of the Hamilton equations in classical mechanics,

except that they are operator equations !

d i
M (t ) = [ H (t ), M (t ) ]
dt 
d i
⇒ M jk (t ) = j [ H (t ), M (t ) ] k
dt 
i i i
= j H (t ) M (t ) k − j M (t ) H (t ) k = ∑  H jl (t ) M lk (t ) − M jl (t ) H lk (t ) 
   l
Example: A free harmonic oscillator

 + 1  a, a +  = 1
=H ω  a a + 
 2  

𝑑𝑑𝑑𝑑 𝑖𝑖 𝐻𝐻, 𝑎𝑎
= = 𝑖𝑖 ℏ𝜔𝜔𝑎𝑎+ 𝑎𝑎, 𝑎𝑎 /ℏ = −𝑖𝑖𝑖𝑖𝑖𝑖
𝑑𝑑𝑑𝑑 ℏ

∴ 𝑎𝑎 𝑡𝑡 = 𝑎𝑎 0 𝑒𝑒 −𝑖𝑖𝑖𝑖𝑖𝑖

1 1
𝑖𝑖𝑖𝑖𝑖𝑖(𝑎𝑎+ 𝑎𝑎+ ) −𝑖𝑖𝑖𝑖𝑖𝑖(𝑎𝑎 + 𝑎𝑎+ ) + 𝑎𝑎 + 𝑎𝑎
Since 𝑎𝑎 𝑡𝑡 = 𝑒𝑒 2 𝑎𝑎(0)𝑒𝑒 2 = 𝑒𝑒 𝑖𝑖𝑖𝑖𝑖𝑖𝑎𝑎 𝑎𝑎(0)𝑒𝑒 −𝑖𝑖𝑖𝑖𝑖𝑖𝑎𝑎

we have 𝑒𝑒 𝑖𝑖𝑖𝑖𝑖𝑖𝑎𝑎
+ 𝑎𝑎 + 𝑎𝑎
𝑎𝑎(0)𝑒𝑒 −𝑖𝑖𝑖𝑖𝑖𝑖𝑎𝑎 = 𝑎𝑎 0 𝑒𝑒 −𝑖𝑖𝑖𝑖𝑖𝑖
Example: A harmonically driven oscillator

1
𝐻𝐻 = ℏ𝜔𝜔 𝑎𝑎+ 𝑎𝑎 + + ℏ𝑔𝑔(𝑒𝑒 −𝑖𝑖Ω𝑡𝑡 𝑎𝑎+ + 𝑒𝑒 𝑖𝑖Ω𝑡𝑡 𝑎𝑎)
2

𝑑𝑑𝑑𝑑 𝑖𝑖 𝐻𝐻, 𝑎𝑎
= = −𝑖𝑖𝑖𝑖𝑖𝑖 − 𝑖𝑖𝑔𝑔𝑒𝑒 −𝑖𝑖Ω𝑡𝑡
𝑑𝑑𝑑𝑑 ℏ
𝑡𝑡
𝑎𝑎 𝑡𝑡 = 𝑎𝑎 0 𝑒𝑒 −𝑖𝑖𝑖𝑖𝑖𝑖 − 𝑖𝑖𝑖𝑖 � 𝑒𝑒 −𝑖𝑖𝜔𝜔 𝑡𝑡−𝑡𝑡 ′ −𝑖𝑖Ω𝑡𝑡 ′ 𝑑𝑑𝑑𝑑𝑑
0

−𝑖𝑖Ω𝑡𝑡 −𝑖𝑖𝜔𝜔𝑡𝑡
𝑒𝑒 𝑒𝑒
= 𝑎𝑎 0 𝑒𝑒 −𝑖𝑖𝑖𝑖𝑖𝑖 − 𝑔𝑔 −
𝜔𝜔 − Ω 𝜔𝜔 − Ω

For an initially ground state |0>, the expectation value of a is:

𝑒𝑒 −𝑖𝑖Ω𝑡𝑡 𝑒𝑒 −𝑖𝑖𝜔𝜔𝑡𝑡
0 𝑎𝑎 𝑡𝑡 0 = −𝑔𝑔 −
𝜔𝜔 − Ω 𝜔𝜔 − Ω

Show that the solution of a(t) implies that the state of the system is a coherent state if the
system is initially in a coherent state
Example: Parametric interaction

𝐻𝐻 = 𝑖𝑖𝑖𝑖𝑖 𝑎𝑎2 − 𝑎𝑎+2 𝑔𝑔 > 0

𝑑𝑑𝑑𝑑 𝑖𝑖 𝐻𝐻, 𝑎𝑎
= = 𝑔𝑔 𝑎𝑎+2 , 𝑎𝑎 = −2𝑔𝑔𝑎𝑎+
𝑑𝑑𝑑𝑑 ℏ
𝑑𝑑𝑎𝑎+ 𝑖𝑖 𝐻𝐻, 𝑎𝑎 +
= = −𝑔𝑔 𝑎𝑎2 , 𝑎𝑎+ = −2𝑔𝑔𝑎𝑎
𝑑𝑑𝑑𝑑 ℏ

Hence

𝑑𝑑 2 𝑎𝑎
2 = 4𝑔𝑔2 𝑎𝑎 ⇒ 𝑎𝑎 𝑡𝑡 = 𝑎𝑎 0 cosh 2𝑔𝑔𝑔𝑔 − 𝑎𝑎+ 0 sinh 2𝑔𝑔𝑔𝑔
𝑑𝑑𝑡𝑡

Such an interaction can lead to squeezed states (See Supplementary topic 7).

Suppose the initial state is the ground state |0>, the variance of X=
1 (a + a ) / 2
†
decreases exponentially with time as 𝑒𝑒 −4𝑔𝑔𝑔𝑔 .
Operator expansion formula

λA −λ A λ2 λ3
e Be B + λ [ A, B ] +
=  A, [ A, B ] +  A,  A, [ A, B ]  + ...
2! 3!  

The proof of this formula is based on Taylor expansion.

If [ A, B ] = c (a number ) then eλ A Be −λ A= B + cλ

e.g. eips /  xe −ips / = x + s

e.g. −α a + + α *a
A=
B=a
∴ [ A, B ] =
α e −α a + +α *a α a + −α *a
ae = a +α

α a + −α *a
D(α ) ≡ e is called the displacement operator
 1 1
Example B + [ A, B ] +
e A Be − A =  A, [ A, B ] +  A,  A, [ A, B ]  + ...
2! 3!

If=A
r 2
2
(
a − a +2 ) and B=a r = real number

[ ] 
A, B
r 2
=
2
a − a(+2
, a =−)
 r  +2 
 2  a , a

=
ra +

( )
2
r 2 + r  2 +
 A, [ A, B ] =
 2 a − a +2
, ra =
 2  a , a =
r 2
a


 A,  A, [ A, B ]  =
 (
r 2
  2 a − a +2
, r )
2 
a  =
r 3 +
a

−A  r2  + r3  +
A
e ae = a 1 + + ...  + a  r + + ...  a cosh r + a sinh r
=
 2!   3! 
   

A + −A
 r2 +
  r3 
e a e = a 1 + + ...  + a  r + = + ...  a + cosh r + a sinh r
 2!   3! 
   

( a −a )
r +2 2
) + e= e− A ⇒ (e A ) + e A =
+
Note that e A is unitary because (e A= A
e 2 = I
Campbell-Baker-Hausdorff Theorem

A, [ A, B ] =
If = B, [ A, B ] 0 then

Proof: Let C (λ ) = eλ Aeλ B

then
d
dλ
C (λ=
) ( )
A + eλ A Be −λ A C (λ )

1 1
B + [ A, B ] +
By e A Be − A =  A, [ A, B ] +  A,  A, [ A, B ]  + ...
2! 3!
∴
d
dλ
C (λ ) = ( A + B + λ [ A, B ]) C (λ ) ≡ E (λ )C (λ )
λ2
 λ   λ  λ ( A+ B ) [ A, B ]
2
λ ) exp  ∫ E (λ ')d λ=
∴ C (= ' exp  Aλ + Bλ +
  [ ]
A, B= 

e e2
 0  2 