Quantum Mechanics 1 Lecture 12:
Time Dependent Phenomena
Massimiliano Grazzini
University of Zurich
Outline
The interaction picture
Time dependent perturbative expansion
The Fermi golden rule
The driven harmonic oscillator
- time ordered product and time evolution
- the S-matrix and coherent states
- periodic perturbation
The Interaction Picture
In Lecture 3 we have seen that one can define two pictures The Schrödinger picture
Time dependence is carried by states while operators are time
independent | ψ
S(t)⟩ A
SThe Heisenberg picture
Time dependence is carried by operators and the states are time
independent A
H| ψ
H⟩
⟨A⟩
ψ(t) = ⟨ψ
S(t) | A
S| ψ
S(t)⟩ = ⟨ψ
S(t
0) | U
−1(t, t
0)A
SU(t, t
0) | ψ
S(t
0)⟩
Moving from one picture to another the matrix elements are invariant since
| ψ
S(t)⟩ = U(t, t
0) | ψ
S(t
0)⟩
with U(t, t
0) = exp{− i
ℏ H(t − t
0)}
A
H(t)
In perturbation theory it is convenient to use a third picture, called interaction picture
This picture is in some sense intermediate between the Schrödinger and Heisenberg pictures
H = H
0+ V(t)
Let us suppose that
where the perturbation V(t) may depend on time In the interaction picture we define
| ψ
I(t)⟩ ≡ exp { i
ℏ H
0t } | ψ
S(t)⟩
The operators are connected to those in the Schrödinger picture through
A
I(t) ≡ exp { i
ℏ H
0t } A
Sexp { − i
ℏ H
0t }
Therefore their time dependence is given by the free (unperturbed) Hamiltonian H
0(1)
We of course have
⟨ψ
S(t) | A
S| ψ
S(t)⟩ = ⟨ψ
S(t)exp{− i
ℏ H
0t}exp{ i
ℏ H
0t} | A
S| exp{− i
ℏ H
0t}exp{ i
ℏ H
0t}ψ
S(t)⟩ = ⟨ψ
I(t) | A
I(t) | ψ
I(t)⟩
We can now study the state evolution in the interaction picture
iℏ ∂ ∂t | ψ
I(t)⟩ = iℏ exp { i
ℏ H
0t } ( i
ℏ H
0) | ψ
S(t)⟩ + exp { i
ℏ H
0t } ( H
0+ V(t) ) | ψ
S(t)⟩
= exp { i
ℏ H
0t } V(t) exp { − i
ℏ H
0t } exp { i
ℏ H
0t } | ψ
S(t)⟩ = V
I(t) | ψ
I(t)⟩
Therefore the time evolution of states in this picture is entirely given by
which is the perturbation potential in the interaction picture V
I(t)
V
I(t)
We note that if the potential in the interaction picture depends on time even if V(t) is time independent [H
0, V(t)] ≠ 0
We now define the evolution operator in the interaction picture
| ψ
I(t)⟩ ≡ U
I(t, t
0) | ψ
I(t
0)⟩
A
I(t)
Since
| ψ
S(t)⟩ = U(t, t
0) | ψ
S(t
0)⟩
we have
| ψ
I(t)⟩ = exp { i
ℏ H
0t } | ψ
S(t)⟩ = exp { i
ℏ H
0t } U(t, t
0) | ψ
S(t
0)⟩
= exp { i
ℏ H
0t } U(t, t
0)exp { − i
ℏ H
0t
0} | ψ
I(t
0)⟩
and we can identify
and | ψ
I
(t)⟩ ≡ U
I(t, t
0) | ψ
I(t
0)⟩
U
I(t, t
0) = exp { i
ℏ H
0t } U(t, t
0)exp { − i
ℏ H
0t
0}
iℏ ∂ ∂t U
I(t, t
0) = V
I(t)U
I(t, t
0)
where is generally different than the time at which we have defined the interaction picture in Eq. t
0(1) t = 0
(2)
The evolution operator U
I(t, t
0) fulfils the equation
Note that, since in general [V
I(t), V
I(t′ )] ≠ 0 we have
U
I(t, t
0) ≠ exp { − i ℏ ∫
t
t0
V
I(t)dt′ }
The equation can be solved in iterative form (2) Indeed if V
I(t) = 0 we have
iℏ ∂ ∂t U
I(t, t
0) = V
I(t)U
I(t, t
0) = 0
and thus
U
I(t, t
0) = 1
If we now plug this zero-order solution on the right hand side of we find (2)
iℏ ∂ ∂t U
I(t, t
0) = V
I(t)U
I(t, t
0) = V
I(t, t
0)
and thus
U
I(t, t
0) = 1 − i ℏ ∫
t
t0
dt′ V
I(t′ )
This means that equation can be solved in iterative form (2)
U
I(t, t
0) = 1 − i ℏ ∫
t
t0
dt′ V
I(t′ ) + ( − i ℏ )
2
∫
t
t0
dt′ ∫
t′
t0
dt′ ′ V
I(t′ )V
I(t′ ′ ) + . . .
The solution can also be written by using the time-ordered product
T{O
1(t
1)O
2(t
2) . . . O
n(t
n)} = O
1(t
1)O
2(t
2) . . . O
n(t
n)θ(t
1− t
2)θ(t
2− t
3) . . . θ(t
n−1− t
n)
U
I(t, t
0) = 1 − i ℏ ∫
t
t0
V
I(t′ )dt′ + ( − i ℏ )
2
1 2! ∫
t
t0
T{V
I(t′ )V
I(t′ ′ )}dt′ dt′ ′ + . . . ≡ T exp { − i ℏ ∫
t
t0
V
I(t′ )dt′ }
we can write + permutations
T{O
1(t
1)O
2(t
2)} = O
1(t
1)O
2(t
2)θ(t
1− t
2) + O
2(t
2)O
1(t
1)θ(t
2− t
1)
(3)
The solution in Eq. is known as Dyson series (3)
If we now plug this first-order solution on the right hand side of we find (2)
U
I(t, t
0) = 1 − i ℏ ∫
t
t0
dt′ V
I(t′ ) + ( − i ℏ )
2
∫
t
t0
dt′ ∫
t′
t0
dt′ ′ V
I(t′ )V
I(t′ ′ )
Time dependent perturbations
We now discuss the prototype of problem that one has to face with time dependent perturbations
We have a system which is in an eigenstate of the Hamiltonian and it is subject to a perturbation | ψ
n(0)
⟩ H
0V(t)
The states and could be associated to different energies, or, in case in which there is degeneracy, could have the same energy | ψ
n(0)
⟩ | ψ
m(0)⟩
This is the typical situation of scattering problems The transition amplitude at the time is t
a
mn(t) = ⟨ψ
m(0)(t) | U(t, t
0) | ψ
n(0)(t
0)⟩
a
mn(t) = ⟨ψ
m,I(0)(t) | U
I(t, t
0) | ψ
n,I(0)(t
0)⟩
and it can be rewritten by using the interaction picture as
The perturbation is weak, so it will not change dramatically the energy levels
but it can induce a transition to another level | ψ
m(0)⟩
a
mn(t) = ⟨ψ
m(0)| U
I(t, t
0) | ψ
n(0)⟩
and use the expansion of U
I(t, t
0) in Eq. to find (3)
a
nm(t) = δ
mn+ ( − i
ℏ ) ∫
t
t0
dt′ ⟨ψ
m(0)| V
I(t′ ) | ψ
n(0)⟩
This result can be interpreted as a perturbative expansion for a
mn(t)
a
mn(t) = ∑
∞i=0
a
mn(i)(t) + ( − i
ℏ )
2
∫
t
t0
dt′ ∫
t′
t0
dt′ ′ ⟨ψ
m(0)| V
I(t′ )V
I(t′ ′ ) || ψ
n(0)⟩ + . . . .
⟨ψ
m,I(0)(t) | = exp { i
ℏ E
mt } ⟨ψ
m(0)| exp { − i
ℏ H
0t } = ⟨ψ
m(0)|
but
and analogously
| ψ
n,I(0)(t
0)⟩ = exp{ i
ℏ H
0t
0} | ψ
n,S(0)(t
0)⟩ = exp{ i
ℏ E
n(0)t
0}exp{− i
ℏ E
n(0)t
0} | ψ
n(0)⟩ = | ψ
n(0)⟩
we can write
The driven harmonic oscillator
We consider the Hamiltonian H = H
0+ V(t)
where and is a time-dependent potential
corresponding to a classical force H
0= p
2
/2m + 1/2mω
2x
2V(t) = − F(t)x F(t)
We suppose that the system in the fundamental state at and we want to find the probability to be in the -state at n t → + ∞ t → − ∞
x = ℏ
2ωm ( a + a
†) = x
02 ( a + a
†)
We first need to evaluate the potential in the interaction picture
We have V
I(t) = exp { i
ℏ H
0t } V(t)exp { − i
ℏ H
0t }
we can show that
a
I(t) = exp { i
ℏ H
0t } a exp { − i
ℏ H
0t } = ae
−iωta
I†(t) = exp { i
ℏ H
0t } a
†exp { − i
ℏ H
0t } = a
†e
iωtBy using H
0
= ℏω ( a
†a + 1 2 ) [a, a
†] = 1
This result could also be shown by using the equation of motion for an operator in the Heisenberg picture
d
dt O
H(t) = − i
ℏ [O
H, H
H] + ∂O
H(t)
∂t a
I(t) = exp { i
ℏ H
0t } a exp { − i
ℏ H
0t }
indeed from
we have
a ·
I(t) = i
ℏ H
0a
I(t) − i
ℏ a
I(t)H
0= − i
ℏ [a
I(t), H
0] = − iωa
I(t)
and thus a
I
(t) = ae
−iωtand analogously
a
I†(t) = a
†e
iωtand [a, H
0] = ℏω[a, a
†a] = ℏωa
We have therefore
V
I(t) = − F(t) x
02 ( a
I(t) + a
I†(t) ) = − F(t) x
02 ( ae
−iωt+ a
†e
iωt)
And we find for the first-order transition probability from the fundamental state to a state n
= 1 2ℏωm ∫
+∞
−∞
dt F(t) ( e
−iωt⟨n | a | 0⟩ + e
iωt⟨n | a
†| 0⟩ )
2
P
n0(t) = 1 ℏ
2∫
+∞
−∞
dt⟨n | V
I(t) | 0⟩
2
= 1 2ℏωm | F(ω) ˜ |
2δ
n1At first perturbative order we can have transitions only to the first excited state
Fourier
transform of f (t)
We now want to solve the problem exactly
S ≡ U
I(+ ∞, − ∞) = T exp { − i ℏ ∫
+∞
−∞
V
I(t′ )dt′ } = T exp { − i ℏ ∫
+∞
−∞
f (t) ( ae
−iωt+ a
†e
iωt) dt }
To do so we need to find a closed form for the time-ordered exponential
f (t)
This can be done by observing that the commutator [a, a
†] = 1
The time ordering can be shown to produce just an irrelevant phase factor
S = T exp { − i ℏ ∫
+∞
−∞
f (t) ( ae
−iωt+ a
†e
iωt) dt } = e
iΦexp { − i ℏ ∫
+∞
−∞
f (t) ( ae
−iωt+ a
†e
iωt) dt }
= e
iΦexp { β(ω)a
†− β*(ω)a } with β(ω) = i
2m ωℏ F(ω) ˜
We can now use the Baker-Campbell-Hausdorff formula e A+B = e A e B e −
12[A,B]
which is valid if [A, [A, B]] = [B, [A, B]] = 0 as in our case with
S = exp { β(ω)a
†} exp { − β*(ω)a } exp { − 1
2 | β(ω) |
2}
Neglecting the irrelevant overall phase we find Φ
A = β(ω)a
†B = − β*(ω)a
The matrix generates coherent states with S β = β(ω)
S | 0⟩ = e
−|β(ω)|2/2exp { β(ω)a
†} | 0⟩
= e
−|β(ω)|2/2∑
∞n=0
β
n(ω)
n ! | n⟩ = | β(ω)⟩
This means that
Recall coherent states (Lecture 5)
| β⟩ = e
−|β|2/2∑
∞n=0
| n⟩ β
nn !
- Eigenstates of the annihilation operator a | β⟩ = β | β⟩ with β ∈ ℂ - They have the property of having minimal uncertainty
- They are characterised by a complex number β
- The fundamental state is a particular coherent state with
(indeed a | 0⟩ = 0 ) β = 0
Their explicit expression is
- They keep minimal uncertainty under time evolution
The transition probability to the -state is given by n
| ⟨n | S | 0⟩ |
2= e
−|β(ω)|2| β |
2nn !
The result in Eq. is nothing but a Poisson distribution (4)
More generally one can show that the effect of the driving term in our harmonic oscillator is to put the oscillator into a coherent state if it was originally in a
coherent state
P
λ(n) = e
−λλ
nn !
Poisson distribution: the discrete probability distribution of a given number of independent events occurring in a fixed interval of time or space with a known average λ
For example: the number of particle emitted by a radioactive source in a given time interval
At the first perturbative order we recover the previous result
| ⟨n | S | 0⟩ |
2∼ | β(ω) |
2δ
n1(4)
When we start from the fundamental state the expected energy transfer is thus
ΔE = ℏω⟨n⟩ = ℏω | β(ω) |
2= | F(ω) ˜ |
2/2m
in agreement with what one would obtain with a classical calculation
⟨n⟩ = e
−λ∑
∞n=0
n λ
nn ! = λe
−λ∑
n=1
λ
n−1(n − 1)! = λ
The average number of “events” is
Where in our case and is related to the Fourier transform of the driving term
λ = | β(ω) |
2β(ω) = i
2m ωℏ F(ω) ˜
∞
∑
n=0P
λ(n) = e
−λ∑
∞n=0
λ
nn ! = 1
The distribution is normalised as
The Fermi golden rule
Let us now consider the case in which we have a perturbation which is active just from the time t = 0 , but is for the rest independent on time
V(t) = Vθ(t)
This allows us to discuss the case in which the system is prepared in an
initial state in a region in which the potential is negligible, let interact with the potential, and then measured in a region where the potential is still
negligible: this is the typical situation in scattering processes
V
I(t) = exp { i
ℏ H
0t } V(t)exp { − i
ℏ H
0t } a
mn(t) = δ
mn+ ( − i
ℏ ) ∫
t
0
dt′ ⟨ψ
m(0)| V
I(t′ ) | ψ
n(0)⟩
The transition amplitude at the first perturbative order is
with
a
mn(t) = δ
mn+ ( − i
ℏ ) ∫
t
0
dt′ exp{ i
ℏ (E
m− E
n)t′ }V
mnWe obtain
with V
mn= ⟨ψ
m(0)| V | ψ
n(0)⟩ The integral gives (we assume m ≠ n )
a
mn(t) = − 1
E
m− E
n( exp{i(E
m− E
n)t / ℏ} − 1 ) V
mnThe transition probability per unit time is thus
1
t P
mn= 1 t | a
mn(t) |
2= | V
mn|
2t(E
m− E
n)
2exp { i(E
m− E
n)t
ℏ } − 1
2
But we have
exp { i(E
m− E
n)t
ℏ } − 1
2
= exp{ i(E
m− E
n)t
2ℏ } ( exp{ i(E
m− E
n)t
2ℏ } − exp{− i(E
m− E
n)t 2ℏ } )
2
= 4 sin
2[ (E
m− E
n)t
2ℏ ]
(5)
The transition probability per unit time is thus
1
t P
mn= 4 | V
mn|
2t (E
m− E
n)
2sin
2[ (E
m− E
n)t
2ℏ ]
We now consider the limit of large , in which the perturbation acts for a very large
time t
The function f (x) = sin
2(xt)
tx
2For large is strongly peaked in the origin but its integral is constant* t
∫
+∞
−∞
dx sin
2(xt)
tx
2= π
It provides a representation of the distribution δ
δ(x) = 1
π lim
t→∞
sin
2(xt)
tx
2*How to compute I = ∫
+∞?
−∞
dx sin
2(xt) tx
2First of all we can change variable and show that the integral is independent
on t x → x / t
∫
+∞
−∞
dx sin
2(xt)
tx
2= ∫
+∞
−∞
dx sin
2(x)
x
2= 2 ∫
+∞
0
dx sin
2(x) x
2We define F(r) = 2 ∫
+∞such that
0
dxe
−rxsin
2(x)
x
2I = F(0)
We have F′ (r) = − 2 ∫
∞and
0
dxe
−rxsin
2x
x F′ ′ (r) = 2 ∫
∞
0
dxe
−rxsin
2x
can now be computed integrating by parts and we find
F′ ′ (r)
F′ ′ (r) = 2 ∫
∞
0
dxe
−rxsin
2x = 4
r (r
2+ 4)
Integrating twice with respect to we find r F(r) = c
2+ c
1r − 2 arctan(r /2) + r ln r
4 + r
2By using the constraint lim we can fix and and we find
r→∞
F(r) = 0 c
2= π c
1= 0
I = F(0) = π
We now go back to the probability per unit time and set x = E
m− E
n2ℏ 1
t P
mn= 4 | V
mn|
2t (E
m− E
n)
2sin
2[ (E
m− E
n)t
2ℏ ] = | V
mn|
2ℏ
2tx
2sin
2(tx)
In the large limit we find t
1
t P
mn⟶ | V
mn|
2ℏ
2πδ ( E
m− E
n2ℏ ) = 2π | V
mn|
2ℏ δ(E
m− E
n)
This formula (and its extension to the case of periodic perturbation to be discussed later) was derived by Pauli in 1928 and because of the number of different applications was called by Fermi “golden rule”
It is particularly useful in the case of the continuous spectrum, in which it
expresses the conservation of energy
Periodic perturbation
We now consider a sinusoidally varying time-dependent potential
V(t) = θ(t) ( 𝒱e
iωt+ 𝒱
†e
−iωt)
a
mn(t) = ( − i
ℏ ) ∫
t
0