Connecting elastic and inelastic scattering : Dynamic structure factor

Consider a slow enough particle ( ⃗k ) scattering off a sample (Φ)

with a periodic structure
⃗n
∣k⃗i , Φi 〉 →∣k⃗f , Φ f 〉 V ( ⃗r )=∑ v ( ⃗r − R⃗n ) V ⃗q =v ⃗q ∑e i ⃗q . R

n lattice

2π 2
Pi→ f =
ℏ ∣ ∣
〈 k⃗f , Φ f ∣V (⃗r ) ∣k⃗i , Φ i 〉 δ( E f +ϵ f −E i−ϵi )
L3 3 ⃗ L3 M
δ N i→ f = Pi→ f × 3
d k f = P i→ f × 3
k f 2 dϵf d Ω
(2 π) (2 π) ℏ
Compare this with the incident particle current
ℏ ki ℏki
⃗j = ρ ⃗v = | ψi | 2
M
=
MV ( assume : ψi =
1 i k⃗ . ⃗r
√V
e i

)
| j |d2σ = δN i → f Definition of the scattering cross-section

(MV ) k f
( )
2 2
d σ 2π
d ϵf d Ω
= 3
ki
×
ℏ
∑ 〈 Φf ∣ V q ∣Φ i 〉 * 〈 Φ f ∣ V q ∣Φ i 〉 δ( E f +ϵ f −E i−ϵi )
h f

We only measure the state of the outgoing particle. So all the possible states of
the sample must be summed over. That is the origin of the sum over f
 Connecting elastic and inelastic scattering : Dynamic structure factor

2π
∑ Pi→ f =
ℏ q ∑
|v |2
〈
Φi | ∑ e
q . R⃗
−i ⃗
|Φ f m
〉〈 Φ | ∑ e
f
q . R⃗n
i⃗
〉
| Φi δ(E f −E i +ℏ ω)
f f

1 dt i(( E − E )/ ℏ+ω)t Notice how the index f has been switched and
δ(E f −E i +ℏ ω) =
ℏ ∫2π e f i
the complex conjugation removed

H |Φf 〉 = E f |Φ f 〉 Heisenberg
representation
2π dt i ω t
. . ∑ Pi → f
.

Φf
=
ℏ
2
| v q ∑∫
|
2
2π
e Φ i | 〈
e
q . R⃗
−i ⃗
|Φfm
〉〈 Φ f |e
iHt/ ℏ i ⃗
e
⃗ −iHt / ℏ
q. R
en
| Φi 〉
fmn

2π dt i ωt
=
ℏ
2
| v q |2
× ∑ ∫ 2 π e Φi | e
mn
〈
q . R ⃗(0 ) i ⃗
−i ⃗
e
q . R ⃗(t )
m
| Φi n
〉
The time dependence has come in because the lattice
points creating the potential are not stationary – this is how R⃗m (t )= R⃗0m +u⃗m (t )
the phonons enter into the picture.

Second we need to do a THERMAL average over the

initial states, beacuse the sample sits at some finite
temperature.

Follow the standard notation for thermal average of an

expectation value.
Connecting elastic and inelastic scattering : Dynamic structure factor

dt i ωt ∑ 〈 Φi | A |Φ i 〉 e−β E
〈
q . R ⃗(0) i ⃗
q . R ⃗(t )
〉
i

∑ ∫ 2 π e Φi | e−i ⃗
e m
| Φi n

i
mn 〈 A〉 =
=∑ e
i ⃗q( R⃗m − R
0 ⃗0 )
n
dt i ω t −i ⃗q . u⃗ (0 ) i ⃗q . u⃗ (t)
∫ 2 π e 〈e m
e 〉
n ∑ e−β E i

i
mn Called the
thermal average

1 q ( R⃗ − R
⃗) dt i ω t −i ⃗q . u⃗ (0 ) i ⃗q . u⃗ (t)
0 0

∑ ∫ e 〈e 〉
i⃗
S ( ⃗q , ω) = e m n
e m n

N mn 2π
d2σ (MV )2 k f 2 π 2
= 3 2
N | v |
⃗q ×S (⃗ q , ω)
d ϵdΩ h ki ℏ

The dynamic structure factor is a characteristic of the lattice only, irrespective

of what (light, electron, neutron) is being scattered by the crystal.
Evaluating the Dynamic structure factor

For harmonic vibrations a remarkable result holds for thermal averaging

Given two matrices or operators A and B which are both linear in ⃗r and ⃗p
〈 A +B +2 A B 〉 /2
2 2

〈e A
e
B
〉 = e
in general [ A , B]≠0

− 〈(⃗
q . u⃗m (0 )) 〉/ 2 −〈 (⃗
q . u⃗n(t)) 〉/ 2
2 2

〈e i⃗
q . u⃗m (0) i ⃗
e
q . u⃗n(t )
〉 = e ×e ×e
q . u⃗m (0))(⃗
−〈 (⃗ q . u⃗n(t )) 〉
All sites m,n are
q . u⃗ (0))(⃗
q . u⃗ (t )) 〉
= e−q u / 2×e−q u / 2×e 〈
2 2 2 2
− (⃗ m n equivalent so their
r.m.s amplitude
would be same
×e 〈
2
−q 〈u 〉
2
q . u⃗ (t ))〉
− (⃗q . u⃗ (0))(⃗
= e m n

Debye-Waller
factor 2 2
q . R⃗n
0
dt i ω t 〈(⃗q . u⃗ (0 ))(⃗q . u⃗ (t))〉
S ( ⃗q , ω) = e
−q 〈u 〉
∑e i⃗
∫ 2π e e 0 n

n
The exponential can
The 1/N factor cancels after one of the now be simple expanded
indices (in this case m) is summed over. and evaluated term by
term!
Evaluating the Dynamic structure factor …..expand term by term

−q 2 〈u2 〉 q . R⃗n
0
dt i ω t
S ( ⃗q , ω) = e ∑e i⃗
∫ 2 π e [ 1+〈(⃗q . u⃗0 (0))(q⃗ . u⃗n (t ))〉 + .... ]
n

The first term is just 1 : The integral reproduces the static structure factor which gives the
correct intensity for elastic scattering including the Debye-Waller factor for temperature.

The second term can be evaluated by writing un(t) in terms of the creation annihilation
operators as applicable to a harmonic oscillator.
The operator is written in
1
√ ℏ terms of the fourier
[ ]
i (⃗ ⃗ −ω t )0 ⃗ −ω t)
+ −i (⃗k . R
0

u n (t) = ∑ 2 m ωk
ak e k .R n
+ ak e k n k
transform of the
√N ⃗
k raising/lowering operators.

( )
2 2
ℏ q 1
[ −i( ⃗ −i( ⃗
]
2 2
q ). R⃗n q). R⃗n
S 1 (⃗
q , ω) = e
−q 〈u 〉
2m
∑ ℏω e
k −⃗
n k δ(ω−ω k )+e
k +⃗
(n k +1)δ (ω+ωk )
⃗
k ,n k

Here the phonon ω( ⃗k ) is created , n k is the boson occupation probability

The wave vector of the incident particle changes by ⃗q Notice how temperature
enteres in two ways in
Notice how the exponential sum forces the requirement inelastic process
⃗
k −q
⃗ =G⃗ (a reciprocal lattice vector)
Evaluating the Dynamic structure factor …..expand term by term

−q 2 〈u2 〉 q . R⃗n
0
dt i ω t
S ( ⃗q , ω) = e ∑e i⃗
∫ 2 π e [ 1+〈(⃗q . u⃗0 (0))(q⃗ . u⃗n (t ))〉 + .... ]
n

The first term is just 1 : The integral reproduces the static structure factor which gives the
correct intensity for elastic scattering including the Debye-Waller factor for temperature.

The second term can be evaluated by writing un(t) in terms of the creation annihilation
operators as applicable to a harmonic oscillator.
The operator is written in
1
√ ℏ terms of the fourier
[ ]
i (⃗ ⃗ −ω t )0 ⃗ −ω t)
+ −i (⃗k . R
0

u n (t) = ∑ 2 m ωk
ak e k .R n
+ ak e k n k
transform of the
√N ⃗
k raising/lowering operators.

( )
2 2
ℏ q 1
[ −i( ⃗ −i( ⃗
]
2 2
q ). R⃗n q). R⃗n
S 1 (⃗
q , ω) = e
−q 〈u 〉
2m
∑ ℏω e
k −⃗
n k δ(ω−ω k )+e
k +⃗
(n k +1)δ (ω+ωk )
⃗
k ,n k

Here the phonon ω( ⃗k ) is created , n k is the boson occupation probability

The wave vector of the incident particle changes by ⃗q Notice how temperature
enteres in two ways in
Notice how the exponential sum forces the requirement inelastic process
⃗
k −q
⃗ =G⃗ (a reciprocal lattice vector)
Two generic conservation laws for scattering from a crystal

The wave-vector conservation looks

E f −E i = ±ℏ ω(⃗q ) like momentum conservation.
But it is not momentum.
k⃗f −k⃗i = q + G
⃗ ⃗
Notice the presence of an arbitrary
reciprocal lattice vector.

Incident particle → i
Outgoing particle → f : w(q) is the excitation created or destroyed

When will the additional factor of G come into play?

In a situation where the incident particle has a wave vector comparable

in size with the Brillouin zone.

In case where the G comes into play and brings back the resultant vector
into the first Brillouin zone – the process is called UMKLAPP (flip over)

It is possible with neutrons, X-rays. (estimate the wave-vector ratio)

Very improbable with visible light.

Wavecector ~ /a
Visible light ~ 10-3 of the zone vector
X-ray, thermal neutron ~ comparable in size to Brillouin zone
Some examples of phonon dispersion : Silicon

In a 3D crystal there would always be 3 accoustic modes (this can be proved)

If the unit cell has p atoms, then there should be 3p-3 optical modes.

Many degeneracies are possible. (Calculations : Lattice Dynamics + refinements)

Plotting the dispersion is done following exactly the same convention as electronic
band structure...from zone centere towards various special points.

Direct : FCC
Reciprocal : BCC

J. Phys. : Condensed Matter 20 (2008) 145213 Direct : Hexagonal

Reciprocal: Hexagonal
Some examples of phonon dispersion : FCC - Cu

Methods used to map

out phonon w(q)

Inelastic neutron
scattering

High Resolution Electron

Energy Loss scattering
(HREELS)

Inelastic Helium atom

scattering (HAS)

A. Dal Corso, J. Phys.: Condens. Matter, 25, 145401 (2013)

Model calculation results & inelastic neutron diffraction data
Some examples of phonon dispersion : GaAs

GaAs...notice that frequencies are lower than Si.

For diamond it was much higher.
Some examples of phonon dispersion : Graphene

Al Taleb et al
J Phys. Condensed
matter: 28, 103005
(2016)
Excitation spectrum of a liquid by neutron scattering : Superfluid Helium

The principles are the same.

But here momentum is real.

Unlike phonons in solid, that do
not carry real momentum.

There is NO additional reciprocal

lattice vector.

The minima at a finite Q is a

special feature of a superfluid.

C. Kittel, Quantum theory of solids.