PG/KA/Chap 4 - 1

Chap 4. Time-dependent Method of Fermis Golden Rule

w
km
=
2
h
|U
km
|
2
(E
k
E
m
h)
Molecular systems : |m = |i, |k = |f,

|electrons, nuclei
U
f

,i
=
_
dR
f
(R)
_
dr
f
(r; R)U(r, R)
i
(r; R)
. .

i
(R)

f
(R)|

U
fi
(R)
. .
|
i
(R)
R

U
fi
(R)

(r; R)|U(r, R)|

(r; R)
r
Suppose : we cannot specify the nal nuclear quantum state

(e.g., have not sucient energy resolution, or not interested)

w(f i) =

w
f

,i
( Note : normally, vib. rot. energy k
B
T , electronic energy k
B
T )
PG/KA/Chap 4 - 2
And : thermal average over the initial nuclear states
w(f i) =

P
i
w(f i), P
i
= e
E
i
/Z
i
(Boltzmann distribution)
_

_
Partition function : ( normalization

P
i
= 1 )
Z
i
=

e
E
i
=

|e
H
i
| = Tr
(nuc)
[e
H
i
]
_
BO approx. H
e
(r; R)
i
(r; R) = W
i
(R)
i
(r : R)
H
i

i
(R) = E
i

i
(R) , where H
i
T
N
+ W
i
(R)
_
Density operator for nuclear states on adiabatic potential W
i
(R)

i
e
H
i
/Z
i
, Z
i
= Tr
(nuc)
[e
H
i
]
Thermal average (on W
i
(R)) of a quantity A(R)
A =

P
i
|A| =

e
E
i
Z
i
|A| =

|
e
H
i
Z
i
A|
= Tr
(nuc)
[
i
A]
_

_
PG/KA/Chap 4 - 3
Time-dependent form (Kubo-Toyozawa)
Using : () =
1
2
_

e
it
dt , (and hence ( h) =
1
h
() )
w(f i) =
2
h

|U
f

,i
|
2
(E
f
E
i
h)
=
1
h
2

U
fi
||
2
_

dte
iE
f

t/ h
e
+iE
i
t/ h
e
it
=
1
h
2

dt|

fi
e
iH
f
t/ h
|

U
fi
e
iH
i
t/ h
|e
it
=
1
h
2
_

dt|

U

fi
e
iH
f
t/ h

U
fi
e
iH
i
t/ h
. .
|e
it
A (for brevity)
w(f i) =

P
i
w(f i) =
1
h
2
_

dtTr
(nuc)
[
i
A]e
it
PG/KA/Chap 4 - 4
Application - 1 : Vibration / rotation spectra [SR 9.6, 9.7]
In the same electronic state : H
f
= H
i
= H

U
fi
=

U
ii
(dipole moment)
Vibration / rotation spectra (absorption)
()
_

dtTr
(nuc)
[

e
iHt/ h
e
+iHt/ h
. .
]e
it
(t) (Heisenberg rep.)
=
_

dt

(0)(t)e
it
Vib-rot spectra TCF of vib-rot (thermal) motion
PG/KA/Chap 4 - 5
Application - 2 : Electronic spectra, Nonadiabatic transitions
1. Gaussian Wavepacket method
2. Cumulant Expansion method
Gaussian Wavepacket [SR 9.4, 7.3.2]

p
t
,q
t
(q, t) = exp[
i
h

t
(q q
t
)
2
+
i
h
p
t
(q q
t
) +
i
h

t
]
q
t
= q, p
t
= p : average position and momenta
follow classical eq of motion : x
t
= H/p
t
, p
t
= H/x
t

t
,
t
: width and phase (time dependent)

t
= (2/m)
2
t
V
xx
/2,
t
= ih
t
/m+p
t
x
t
E
Exact on quadratic potentials
V (x) = V
0
+V
x
(x x
t
) +
1
2
V
xx
(x x
t
)
2
PG/KA/Chap 4 - 6
Using : Tr
(nuc)
[A]
1
h
n
_
dq
n
0
dp
n
0

p
0
,q
0
|A|
p
0
,q
0

w(f i) =
1
h
2
_

C(t)e
it
dt
C(t)
_
dq
n
0
dp
n
0

p
0
,q
0
|
i

U

fi
e
iH
f
t/ h

U
fi
e
iH
i
t/ h
|
p
0
,q
0

=
1
Z
_
dq
n
0
dp
n
0

i
(t ih)|
f
(t)
|
i
(t i h) =

U
fi
e
iH
i
(ti h)/ h
|
p
0
,q
0

|
f
(t) = e
iH
f
t/ h

U
fi
|
p
0
,q
0

1. Propagate wavepackets |
p
0
,q
0
on H
i
and H
f
2. Calculate the overlap
i
|
f

3. Fourier transform
PG/KA/Chap 4 - 7
Cumulant expansion method
Condon Approximation :
Neglect R dependence of

U
fi
(R)
or take 0th term of :

U
fi
(R) =

U
fi
(R
0
) +
_

U
fi
R
_
R
0
(RR
0
) +

w(f i) =
1
h
2
|

U
fi
(R
0
)|
2
_

dte
iH
f
t/ h
e
+iH
i
t/ h
. .

i
e
it
(Time ordered exponential) exp
()
[
i
h
_
t
0
dV
i
()]
V H
f
H
i
, V
i
(t) e
iH
i
t/ h
V e
+iH
i
t/ h
PG/KA/Chap 4 - 8
Time ordered exponential
( f(t) ) e
iH
f
t/ h
e
+iH
i
t/ h
= exp
()
[
i
h
_
t
0
dV
i
()]
Note : Since [H
i
, H
f
] = 0 , e
iH
f
t/ h
e
iH
i
t/ h
= e
i(H
f
H
i
)t/ h
_

_
f
t
=
i
h
e
iH
f
t/ h
(H
f
H
i
)e
+iH
i
t/ h
=
i
h
e
iH
f
t/ h
e
+iH
i
t/ h
e
iH
i
t/ h
(H
f
H
i
)e
+iH
i
t/ h
=
i
h
f(t)V
i
(t)
Integrate : f(t) = f(0) +
i
h
_
t
0
df()V
i
()
Sequential expansion (f(0) = 1) :
f(t) = 1 +
i
h
_
t
0
dV
i
() + (
i
h
)
2
_
t
0
d
_

0
d

V
i
(

)V
i
() +
exp
()
[
i
h
_
t
0
dV
i
()] ( Denition)
_

_
PG/KA/Chap 4 - 9
Note :
The original formula was
w(f i) =

P
i
w(f

i)
=
2
h

P
i
|f

U
fi
|i|
2
(E
f
E
i
h)
Condon Approx.
w(f i)
2
h
|

U
fi
(R
0
)|
2

P
i
|f

|i|
2
(E
f
E
i
h)
We will obtain w(f i) in p.7 from the Fourier transform of this.
If we correctly evaluate w(f i) in p.7, it takes into account of the
(thermal average of) the Franck-Condon overlap |f

|i|
2
Quantum eects of the nuclear motion (e.g., tunneling) are accounted
PG/KA/Chap 4 - 10
Cumulant expansion
average of exponential exponential of averages
e
ix
= e
ix
c
+
1
2
(i)
2
x
2

c
+
_

_
Expand both sides and compare (in order of ) to nd
Cumulant average : x
c
= x
x
2

c
= x
2
x
2
(= variance)
x
3

c
= x
3
3x
2
x + 2x
3
_

_
Advantages :
Average of oscillatory function e
ix

average rst, then place on exponent e

ix
c
+
Partial sum to innite order :
even if the exponent on the right-hand-side is truncated at nite order,
terms up to the innite order in is partially included
In a particular case where the variable x follows Gaussian distribution
(Gaussian process), the 3rd and higher order cumulants exactly vanish
PG/KA/Chap 4 - 11
Cumulant expansion of time-ordered exponential
exp
()
[
i
h
_
t
0
dV ()]
= exp
_
i
h
_
t
0
dV ()
c
+
_
i
h
_
2
_
t
0
d
_

0
d

V (

)V ()
c
+
_
Expand both sides and compare order by order (up to 2nd order) :
V ()
c
= V () = V (0) = V (independent of time in steady-state)
V (

)V ()
c
= V (

)V () V
2
= V (

)V ()
( V (t) V (t) V )
Line-broadening function : g(t)
1
h
2
_
t
0
d
_

0
d

V (

)V ()
In thermal equilibrium : A(

)A() = A(0)A(

)
(depends only on the time interval

)
[Verify using ... = Tr{...} and Heisenberg rep. of A(t)]
PG/KA/Chap 4 - 12
Changing integration variable (,

) (

, s

) : Jacobian = 1
g(t) =
1
h
2
_
t
0
ds
_
ts
0
d

V (0)V (s) =
1
h
2
_
t
0
ds(t s)V (0)V (s)
Thus,
w(f i) =
1
h
2
|

U
fi
|
2
_

dte
g
i
(t)
e
i(V
i
/ h)t
g
i
(t) =
1
h
2
_
t
0
ds(t s)V (0)V (s)
i
(correct up to 2nd-order cumulant)
V
i
: average over nuclear motion on potential V
i
(electronic state i)
V (0)V (t)
i
: TCF = thermal uctuation of V
.
.
PG/KA/Chap 4 - 13
Time correlation functions
C(t) = A(0)A(t) : uctuation A(t) = A(t) A
_

_
Classical : phase-space distribution function f(r, p)
C(t) =
_
dr
_
dpf(r, p)A(r, p; 0)A(r, p; t)
Statistical (ensemble) average distribution of the initial condition f(r, p)
_

_
Quantum :
C(t) = Tr[Ae
iHt/ h
Ae
iHt/ h
]
Thus, C(t) = C

(t) Re C(t) = Re C(t) ... even,

Im C(t) = Im C(t) ... odd
In classical mech. C(t) is real and even (for real quantities)
C(0) = A(0)
2
0 (variance, uctuation)
PG/KA/Chap 4 - 14
When quantity A shows stochastic random motion (eg, in solution phase),
C(t) 0 as t .
_

_
A(0) and A(t) at large t (long time interval) lose mutual correlation,
such that both A(0)A(t) > 0 and < 0 realize randomly.
In other words, at long time interval, A(0) and A(t) become statisti-
cally independent, as described by C(t) A(0)A(t) = 0 0 = 0.
_

_
Ergodic hypothesis
Statistical (ensemble) average = average over the time
C(t) = A(0)A(t) = lim
T
1
T
_
T
0
dt
0
A(t
0
)A(t
0
+t)
PG/KA/Chap 4 - 15
TCF and spectral line shape
Vib. / rot. spectra : ()
_

(t)(0)e
it
dt =
_

C(t)e
it
dt
Exponential TCF
C(t) = C(0)e
|t|
() C(0)
2

2
+
2
(Lorentzian line shape)
Gaussian TCF
C(t) = C(0)e

2
t
2
() C(0)e

2
/4
2
(Gaussian line shape)
Damped-oscillating TCF
(1) C(t) = C(0)e
|t|
e
i
0
t
()
2

2
+ (
0
)
2
(2) C(t) = C(0)e

2
t
2
e
i
0
t
() e
(
0
)
2
/4
2
(Note : oscillatory factor e
i
0
t
just introduces peak shift)
[Verify by yourself (just elementary integrations!)]
PG/KA/Chap 4 - 16
Example of Gaussian TCF
Consider : dilute solution of dipolar molecules
Short-time motion nearly free rotation (with angular velocity )
(inertial motion)
Dipole correlation : (0)(t) = ||
2
cos t
Kinetic energy of rotation = I
2
/2 (I = inertial moment)
Thermal population of : P()d e
E/k
B
T
d = e
I
2
/2k
B
T
d
(Normalization :
_

0
P()d = 1 prefactor 2(I/2k
B
T)
1/2
)
Hence,
(0)(t) =
_

0
P()||
2
cos td = ||
2
e
k
B
Tt
2
/2I
(Gaussian TCF)
PG/KA/Chap 4 - 17
Example of Exponential TCF
Langevin equation (Brownian motion model)
m v = mv + R(t)
(R(t) : random force, : friction coecient)
v(0) and statistical average
v(0) v(t) = v(0)v(t) +
1
m
v(0)R(t)
Dene C(t) v(0)v(t) , and assume v(0)R(t) = 0
(no correlation between v(0) and the random force)
d
dt
C(t) = C(t) C(t) = C(0)e
t
PG/KA/Chap 4 - 18
Example : Brownian oscillator model
Harmonic oscillator + friction + random force
m x = m
2
0
x m x + R(t)
x(0) x(t) =
2
0
x(0)x(t) x(0) x(t) +
1
m
x(0)R(t)

C(t) +

C(t) +
2
0
C(t) = 0
C(t) = C(0)(cos
1
t +

2
1
sin
1
t)e
t/2
(
2
1

2
0

2
/4 )
_

_
(i)
2
0
>
2
/4 : C(t) = damped oscillation
(ii)
2
0
=
2
/4 : C(t) = C(0)(1 +t/2)e
t/2
(iii)
2
0
<
2
/4 :
1
= imaginary C(t) = double exponential
[ (i) = under-damped, (ii, iii) = over-damped ]
PG/KA/Chap 4 - 19
Motional narrowing
(Back to page 14)
w(f i) =
1
h
2
|

U
fi
(R
0
)|
2
_

e
g
i
(t)
e
i(V
i
/ h)t
dt
g
i
(t) =
1
h
2
_
t
0
d(t )V (0)V ()
i
Assume : Exponential TCF
V (0)V () = D
2
e
|t|/
c
_
D
2
= V (0)
2
... amplitude of uctuation

c
... correlation time
_
g(t) =
1
h
2
(D
c
)
2
(e
|t|/
c
+
|t|

c
1)
PG/KA/Chap 4 - 20
1. Large
c
case (long correlation time / slow modulation)
g(t) D
2
t
2
/2 h
2
(short-time expansion in t )
w(f i) =
1
h
2
|

U|
2
_

e
D
2
t
2
/2 h
2
e
i(V
i
/ h)t
dt
=
2

U|
2
hD
exp[
h
2
( V
i
)
2
2D
2
] (Gaussian line shape)
2. Short
c
case (short correlation time / fast modulation)
e
|t|/
c
+|t|/
c
1 |t|/
c
(long time approximation in t )
w(f i) =
1
h
2
|

U|
2
_

e
D
2

c
|t|/ h
2
e
i(V
i
/ h)t
dt
=
|

U|
2
h
2
2

2
+ ( V
i
)
2
(Lorentzian line shape)
Width
c
( D
2

c
/h
2
)
motional narrowing :
shorter
c
(faster uctuation modulation) narrower spectra
PG/KA/Chap 4 - 21
in general :
_
near spectral peak large t Lorentzian shape
near spectral tail small t Gaussian shape
Overall shape parameter D
c
/h <> 0
Physical interpretation
Large
c
= slow modulation (of the nuclear conguration)
V is xed during the photo absorption/emission
distribution of V is directly reected in the spectral shape
Small
c
= fast modulation
uctuation of V is averaged out in the observation time scale
.
.