Crystal Structure and Dynamics

Paolo G. Radaelli, Michaelmas Term 2015

Part 2: Scattering theory and experiments

Lectures 6-8
Web Site:
http://www2.physics.ox.ac.uk/students/course-materials/c3-condensed-matter-major-option

Bibliography
◦ B.E. Warren, X-ray diffraction Dover Publications, Inc., New York, 2nd Ed. 1990. is a rather
old book with a very dated notation, but somehow I always find myself going back to it.
Some of the explanations are very clear.

◦ G.L. Squires Introduction to thermal neutron scattering Dover Publications, Cambridge, New
York, 1978is a classic introductory book, and has all the basic derivations of nuclear and
magnetic cross sections.

◦ John D. Jackson ‘Classical Electrodynamics John Wiley & sons, New York, Chichister, Bris-
bane, Toronto, Singapore, (1975). This book provides detailed derivation of the classical
X-ray scattering cross section. It is a very complete compendium of electrodynamics,
although not always easy to digest...

◦ Stephen W. Lovesey,Theory of neutron scattering from condensed matter , Oxford Science

Publications, Clarendon Press, Oxford (1984) — in 2 volumes. This is “the” texbook on
the theory of neutron scattering and related matters.

◦ E. H. Kisi Applications of neutron powder diffractionOxford University Press, Oxford, New

York, 2008 is a more recent book on neutron diffraction, focussing on the powder method.
◦ André Authier Dynamical Theory of X-ray Diffraction, International Union of Crystallography,
Oxford University Press (Oxford, New York), 2001. This is a very complete book on the
dynamical theory of X-ray scattering. It has a good historical introduction and some simple
explanations of the phenomena.

◦ P.J. Grundy & G.A. Jones Electron Microscopy in the Study of Materials, (Edward Arnold
Ltd. London: UK, 1976). A handy booklet on electron microscopy, a bit old but still useful
to understand how dislocations can be imaged.

◦ S.J.L. Billinge & M.F. Thorpe Local structure from diffraction, Kluwer Academic Publishers
New York, Boston, Dordrecht, London, Moscow, 2002. A good collection of articles on
diffuse scattering and scattering from disordered materials.

1
Contents
1 Lecture 6 — Scattering from individual atoms and spins 3
1.1 Thomson scattering of X-rays from a single electron . . . . . . . . . . . . . . . . 3
1.2 Thomson scattering from many quasi-free electrons . . . . . . . . . . . . . . . . 5
1.3 X-ray scattering from bound electrons — anomalous scattering . . . . . . . . . . 7
1.4 Thermal neutron scattering from atoms and spins . . . . . . . . . . . . . . . . . . 8

2 Lecture 7 - Scattering from crystals 12

2.1 Cross section for a “small” perfect single crystal . . . . . . . . . . . . . . . . . . . 12
2.2 Laue and Bragg equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.3 The effect of atomic vibrations — the Debye-Waller factor . . . . . . . . . . . . . 15
2.4 Finite size effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

3 Lecture 8 - Diffraction experiments and data analysis 19

3.1 Geometries for diffraction experiments - single crystal diffraction . . . . . . . . . 19
3.1.1 Scattering triangles for elastic scattering and the Ewald construction . . . 19
3.2 Scattering triangles for inelastic scattering . . . . . . . . . . . . . . . . . . . . . . 20
3.3 Powder diffraction and the Debye-Scherrer cones . . . . . . . . . . . . . . . . . . 22
3.4 Integrated Intensities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.5 Structural solution from diffraction data . . . . . . . . . . . . . . . . . . . . . . . . 24
3.5.1 The phase problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.5.2 The Patterson method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.5.3 Structural optimisation: least-square refinements . . . . . . . . . . . . . . 26

4 Appendix I: more on the theory of scattering 27

4.1 Scattering from a free electron (extended derivation) . . . . . . . . . . . . . . . . 27
4.2 Thomson scattering from many quasi-free electrons (extended derivation) . . . . 32

5 Appendix II: X-ray absorption and resonant scattering 33

5.1 X-ray absorption: the photo-electric effect and X-ray fluorescence . . . . . . . . . 33
5.2 X-ray scattering beyond the free-electron approximation . . . . . . . . . . . . . . 35
5.3 Resonant scattering and anomalous corrections to the form factor . . . . . . . . 36
5.4 Compton scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

6 Appendix III: Scattering of particle beams 37

6.1 Wavefunction equation for the static scattering problem . . . . . . . . . . . . . . 38
6.2 Elastic scattering of electrons in the 1st Born approximation . . . . . . . . . . . . 40
6.3 Nuclear scattering of “slow” neutrons . . . . . . . . . . . . . . . . . . . . . . . . . 42
6.4 The Fermi golden rule and its connection with the first Born approximation . . . . 43
6.5 Magnetic scattering of neutrons . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

7 Appendix IV: Useful constants 46

2
1 Lecture 6 — Scattering from individual atoms and
spins

In order to extract information about the atomic structure of a crystal, liquid

or glass by diffraction, the wavelength of the probe must be compara-
ble or smaller than the interatomic distances, which are typically a few
Ångstroms (10−10 m, or 10−1 nm). Tab. 1 illustrates the typical wavelengths
and energies employed for X-ray, neutron and electron diffraction.

Table 1: Typical wavelenghts and energies employed for X-ray, neutron and
electron diffraction. For electromagnetic radiation, E = hc/λ, with hc = 12.4
2 ~2 2 2
KeV · Å; for a non-relativistic particle beam, E = 2π
mλ2
=, where 2πm~ = 82
meV · Å2 for neutrons and 150 eV · Å2 for electrons. A typical Transmission
Electron Microscope (TEM) can operate at 200 KV raising the electron ve-
locity to 70 % of the speed of light, and some state-of-the-art microscopes
can reach the MV range; therefore, relativistic effects need to be taken into
account in converting between energy and wavelength.
λ E
X-rays 0.1–6 Å 2–150 KeV
neutrons 0.3–10 Å 1–1000 meV
electrons 0.02–3 Å 20 eV–200 KeV

Powerful X-ray, neutron and electron sources and diffraction instruments are
available to the experimentalists.

1.1 Thomson scattering of X-rays from a single electron

• Bragg diffraction of X-rays is primarily due to the scattering from elec-

trons bound to the atoms of the crystal structure. It is generally a very
good approximation to employ the so-called Thomson formula (from J.J.
Thomson, Nobel Prize 1906) to calculate the relevant scattering ampli-
tudes and cross sections. This is a bit of a paradox, since the Thomson
formula assumes free electrons1 , but the agreement with experiments
in nonetheless very good. The Thomson formula is valid in the non-
relativistic limit (velocity of the particle << c).

• By using classical electrodynamics, we can calculate the time-dependent

amplitude of the electric field generated at a distance R from an electron
that has been accelerated by an electromagnetic plane wave.

• If the time-dependent electric field at the position of the electron is

1
A description of the scattering process beyond the free electron approximation is con-
tained in Appendix II.

3
 E = E0 e−iωt (1)

and the incident polarisation (see fig. 11) is

 = cos ξ σ + sin ξ π (2)

then2

ei(kR−ωt) 
cos ξ 0σ + sin ξ cos γ 0π

E(R, t) = −r0 E0 (3)
R

where

e2
r0 = = 2.82 × 10−15 m (4)
4π0 mc2

is the classical radius of the electron.

• The scattering angle γ is the angle between the incident and scattered
wavevector (this angle is also known, by longstanding diffraction con-
vention, as 2θ). Based on eq. 55, we can make the following observa-
tions:

Figure 1: Diagram illustrating the conventional σ and π reference directions

for the incident and scattered polarisation. Note that σ · 0σ = 1 always.
Conversely, π ·0π = cos γ depends on the scattering angle γ, and vanishes
for γ = π/2.

• A plane wave impinging on a quasi-free charge produces a scattered spher-

ei(kR−ωt)
ical wave , with an amplitude that in general depends on the
R
scattering angle γ.
2
See Appendix I for a complete derivation.

4
• If the incident wave is σ-polarised, the scattered wave is σ 0 -polarised, and
r0
has amplitude R E0 .

• If the incident wave is π-polarised, the scattered wave is π 0 -polarised, and

r0
has amplitude R E0 cos γ.

• The intensity of the scattered wave is zero for scattering of π polarisation

at 90◦ .

• The scattered wave has a phase shift of π upon scattering (minus sign).

• For generic incident polarisation, the scattering amplitude (amplitude of the

spherical wave) in each scattered polarisation can be written as:

A = r0 [ · 0 ] (5)

• The cross section is defined as the average power radiated per unit solid
angle divided by the average incident power per unit area (power flux,
Φ)

• For an unpolarised X-ray beam, all the angles ξ are equally represented
(i.e., there will be photons with all polarisations). The cross section in
this case is

1 + cos2 γ
 
dσ
= r02 (6)
dΩ 2

1.2 Thomson scattering from many quasi-free electrons

• When many electrons are inclose proximity to each other, the spherical
waves emitted by them will interfere.

• The combined effect of these waves can be calculated in the following ap-
proximations:

 The amplitude of the motion of the electrons is much smaller than the
wavelength.
 The distance at which the process is observed is also much larger
than the distance between electrons (Fraunhofer diffraction or far-
field limit).
 The distribution of electrons can be considered continuous, with num-
ber density ρ(x) [electrons/m3 ].

5
 With these approximations, which are always obeyed for the electron
cloud around an atom and reasonable values of the electric field)

ei(kR−ωt) 
Z
cos ξ 0σ + sin ξ cos γ 0π ρ(x)e−iq·x dx

E(R, t) = −r0 E0
R
(7)

• The vector q = kf − ki is the scattering vector and ki and kf are the

incident and scattered wavevectors.

• Note the important formula, valid for elastic scattering (we recall that γ =
2θ):

4π sin θ
q = |q| = (8)
λ

Eq. 8 is illustrated graphically in fig. 2

Figure 2: Scattering triangle for elastic scattering.

• The integral

Z
f (q) = ρ(x)e−iq·x dx (9)

is known as the atomic scattering factor or form factor.

We have arrived here at an important result: the scattering amplitude for many quasi-free
electrons is proportional to the Fourier transform of the charge density. Note that the
integral for q = 0 is the total charge, which for an atom is the atomic number Z (fig. 3).

A key fact to remember: the more spread out the charge is around the atom, the faster
f (q) will decay at high q.

High q ≡ high scattering angles, short wavelengths.

• The cross sections are obtained in the same way as for a single charge —
for instance, the unpolarised cross section for an atom is:

6
 1 + cos2 γ
   
dσ
= r02 |f (q)|2 (10)
dΩ atom 2

which, in forward scattering, becomes:

 
dσ
= r02 Z 2 (11)
dΩ atom

Figure 3: Atomic scattering factors (form factors) for selected neutral atoms
and ions. Note that sin θ/λ = q/4π.

• One can find tabulated values of the form factors for neutral and ionised
atoms in the International Tables for Crystallography, volume C, p 555
and p 566, respectively.

1.3 X-ray scattering from bound electrons — anomalous scatter-

ing

• Truly “free’ electrons (e.g., conduction electrons in a metal) hardly con-

tribute to the scattering of X-rays, because their probability distribution

7
 extends a long way throughout the crystal, and, from eq. 9, the form
factor decays very rapidly away from forward scattering (fig. 3).

• The largest contribution to X-ray scattering from atoms is given by “core”

electrons, which are close to the nucleus and have slowly decaying from
factors — but these electrons are certainly not free!

• There are large departures from the Thomson scattering formula near atomic
resonances, where the energy of the photon is just sufficient to eject an
electron from a core state into the continuum. Away from resonances,
the Thomson formula can be corrected to a very good approximation by
replacing the form factor by the complex quantity

f (q) = fThom (q) + f 0 (~ω) + if 00 (~ω) (12)

where the so-called anomalous terms, f 0 and f 00 , away from atomic

resonances do not depend on q and are weak functions of the photon
energy 3 .

It can be shown, as a consequence of the so-called optical theorem, that the imaginary
part of the scattering factor is proportional to the linear absorption coefficient due to the
photoelectric effect.
ω
f 00 (~ω) = µ (13)
4πr0 cNa
where Na is the number of atoms per unit volume, and the other symbols have the usual
meaning. The quantity µ is the linear absorption coefficient, defined as:

I = I0 e−µL (14)

1.4 Thermal neutron scattering from atoms and spins

• As in the case of X-rays, the neutron scattering process generates a spher-

ical wave, the squared amplitude of which is proportional to the cross
section.

• Thermal neutrons, with energies in the meV range, are commonly used
to probe condensed matter. Their properties can be summarised as
follows:

 Free neutrons are unstable, with half-life τ = 10.6 min. (β-decay)

3
See Appendix II for more details about resonant X-ray scattering.

8
  Neutrons bound in nuclei are (generally) stable.
 Mass: 1.67492729(28) × 10−27 kg
 Electric dipole moment D < 10−25 (e cm)
1
 Spin: s = 2 — neutrons are fermions.
e~
 Magnetic dipole moment: µ = −1.9130418 µN , where µN = 2mp =
5.05078324(13) × 10−27 JT−1 is the nuclear magneton.

Table 2: Neutron wavelenghts and kinetic energies in different “slow” ranges.

The thermal energy per particle at room temperature is 25 meV.
λ (Å) E (meV)
Cold 3–30 0.1-10
Thermal 1–3 10–100
Hot 0.4–1 100–500
Epithermal < 0.4 > 500

• Neutrons are elastically or inelastically scattered by the condensed mat-

ter through two completely different mechanisms, but, as it turns out,
yielding comparable scattering lengths.

1. By nuclear interaction with the atomic nuclei.

2. By dipole interaction with the unpaired spin and orbital mag-
netic moments of the atoms. This is only present if the atom or
ion has a magnetic moment.

• A summary of the key results is provided here below4 .

4
See Appendix III for a complete derivation of the nuclear and magnetic neutron cross
sections

9
 Neutron-nuclear interaction
• The neutron-nuclear interaction is isotope and elements specific, and depends on the
mutual orientation of the neutron and the nuclear spin.
• As far as neutron crystallography is concerned, the key parameter is the scattering ampli-
tude averaged over the nuclear spin states, known as the coherent scattering ampli-
tude.
• The neutron nuclear coherent scattering amplitude is independent on q — it carries no
form factor, and is therefore expressed by a single number, known as the Fermi length.
• Fermi lengths can be positive or negative, depending on whether the neutron-nuclear
interaction is attractive or repulsive. For typical nuclei, they are of the order of a few fm
(10−15 m) (see fig 4), which means that they are comparable to the classical electron
radius. However, atoms have a single nucleus and many electrons, so X-ray scattering
cross sections in the forward direction are typically much larger than neutron cross sec-
tions (X-ray cross sections decay at high q due to the form factor).
• Fermi lengths do not vary in a systematic way across the periodic table (fig 4), which means
that with respect to X-rays, neutrons are uniquely sensitive to some light elements
— notably oxygen. The different scattering lengths of different isotopes is also widely
exploited in the so-called contrast variation techniques.

Figure 4: Variation of the Fermi length as a function of atomic weight.

10
Neutron-magnetic interaction
• When the scatterer carries a magnetic moment, in addition to the normal nuclear interaction,
neutron are also scattered by dipole-dipole interaction from the magnetic moment of the
atom.
• Magnetic scattering of neutrons is governed by the following vector scattering amplitude.

An = γN r0 fm (q)M⊥ (15)
where γN is the neutron gyromagnetic ratio (−1.9130418), r0 is the familiar classical elec-
tron radius and M⊥ is the projection of the atomic magnetic moment perpendicular
to the wavevector transfer q, and is expressed in Bohr magnetons.
• The quantity fm (q) is known as the neutron magnetic form factor, and is normalised so that
fm (0) = 1. It is similar to the X-ray form factor, except for the fact that it only include the
more extended density of unpaired electrons. Therefore magnetic neutron scattering
decays very rapidly at high q.
• From eq. 15 one can obtain a number of cross sections, accounting for the different ori-
entations of the neutron spin with respect to the atomic magnetic moment (neutron po-
larisation). The most important cross section is the unpolarised neutron cross section
(averaged over all the possible neutron polarisations), which, for a single atom, is:

dσ 2 2 2
= γN r0 fm (q)|M⊥ |2 = γN
2 2 2
r0 fm (q)M 2 sin2 α (16)
dΩ
where α is the angle between M and q. Note that the cross section is zero if q is
parallel to M.
• Typical magnetic moments for atoms and ions are a few Bohr magnetons. Therefore, from
eq. 16, one finds that neutron nuclear and magnetic scattering cross sections are
typically comparable in magnitude for magnetic atoms.

11
2 Lecture 7 - Scattering from crystals

2.1 Cross section for a “small” perfect single crystal

• We want to calculate the scattering cross section from a “small” single crys-
tal. Here, “small” means that we can ignore multiple scattering events.
We still consider the crystal as perfectly periodic.

• We will employ the same approximations that we have used for the scat-
tering from many quasi-free electrons (the far-field approximation is not
obeyed in some relevant cases, but we will not concern ourselves with
them).

• We will consider the case of X-rays, but the calculation is analogous for
neutrons. The scattering amplitude in each final polarisation direction
is obtained by integrating over the whole crystal (instead that around an
atom, as for eq. 7):

−iq·r [ · 0 ]
R
A(q) = r0 Crystal drρ(r)e (17)

• We can exploit the fact that the charge density is periodic, so that, if ri is a
lattice translation and x is restricted to the unit cell containing the origin:

ρ(r) = ρ(ri + x) = ρ(x) (18)

whence the scattering amplitude becomes

XZ
dxρ(x)e−iq·(ri +x)  · 0
 
A(q) = r0
i U nit Cell
X Z
−iq·ri
dxρ(x)e−iq·x  · 0
 
= r0 e (19)
i U nit Cell

where the summation runs over all the unit cells in the crystal. The
expression

Z
F (q) = r0 dxρ(x)e−iq·x (20)
U nit Cell

is known as the structure factor. Note the close analogy with eq. 32,
Part I.
The structure factor is proportional to the Fourier transform of the charge density (or,
more in general, scattering density) integrated over the unit cell.

12
• If the electron density ρ(r) is a superposition of atomic-like electron densi-
ties (i.e., a series of δ functions), F (q) can be written as

X
F (q) = r0 fn (q)e−iq·xn (21)
n

where the sum runs over all the atoms in the unit cell and fn (q) are the
form factors of each species and xn are their positions within the unit
cell.

• We can now calculate the cross section:

 
dσ X X 2
= A(q)A∗ (q) =  e−iq·(ri −rj )  |F(q)|2  · 0

(22)
dΩ
j i

• We now introduce the fact that the double summation in parentheses can
be consider as running over an infinite lattice. Consequently, all the
summations over i labelled by rj are the same (they only differ by a shift
in origin), and the summation over j can be replaced by multiplication
by Nc — the number of unit cells in the crystal (→ ∞).

• The remaining single summation is only non-zero when q is a RL vector. If

q is restricted to the first Brillouin zone, we can write:

Z
1 v0 X −iq·ri
δ(q) = dxe−iq·x ' e (23)
(2π)3 (2π)3
i

where v0 is the unit cell volume. For an unrestricted q, the same ex-
pression holds with the left-hand member replaced by a sum of delta
functions centred at all reciprocal lattice nodes, indicated with τ in the
remainder. With this, we can write the final expression for the cross
section:

dσ (2π)3 X 2
δ(q − τ )|F (τ )|2  · 0

= Nc (24)
dΩ v0 τ

• The term [ · 0 ]2 needs to be averaged over all this incident and scattered
polarisations, yielding a polarisation factor P(γ), which depends on
the experimental setting. For example, for an unpolarised incident beam
and no polarisation analysis:

13
 1 + cos2 γ
 
P(γ) = unpolarised beam (25)
2

• The final general expression for the average cross section is:

dσ (2π)3 X
= Nc δ(q − τ )|F (τ )|2 P(γ) (26)
dΩ v0 τ

Let’s recap the key points to remember:

 The cross section is proportional to the number of unit cells in the crystal. The bigger
the crystal, the more photons or particles will be scattered. We can clearly see that
this result must involve an approximation: the scattered intensity must reach a limit
when all the particles in the beam are scattered.
 The cross section is proportional to the squared modulus of the structure factor (no
surprises here — you should have learned this last year).
 Scattering only occurs at the nodes of the RL. For a perfect, infinite crystal, this is in
the form of delta functions.
 The cross section contains the unit-cell volume in the denominator. This is necessary
for dimensional reasons, but it could perhaps cause surprise. After all, we could
arbitrarily decide to double the size of the unit cell by introducing a “basis”. The
answer is, naturally, that the |F (τ )|2 term exactly compensates for this.

• The limitations of the small crystal approximation can be overcome by com-

plete description of the scattering process, including multiple scattering,
in what is known as dynamical theory of diffraction (see extended
version of the notes for a very short introduction).

2.2 Laue and Bragg equations

• The δ function in eq. 24 expresses the fact that the cross section is zero
unless q is equal to one of the RLV .

• This is more traditionally expressed by the Laue equations:

q = ha∗ + kb∗ + lc∗

(27)
q · a1 = (kf − ki ) · a1 = 2πh
q · a2 = (kf − ki ) · a2 = 2πk
q · a3 = (kf − ki ) · a3 = 2πl (28)

14
• h, k and l are the Miller indices that we have already encountered.

• The modulus of the scattering vector (fig. 2 and eq. 8) is:

4π sin θ
q = |q| = (29)
λ

• Given a RLV q with Miller indices hkl, it can be shown by simple geometry
that there exist a family of parallel real-lattice planes perpendicular to it,
and that the distance between two adjacent planes is d = 2πn/q, where
n is the greatest common divisor of h, k and l.

• From this and eq. 29, one can deduce that the scattering process can be
thought as a mirror reflection from this family of planes, with the addition
of Bragg law:

2d sin θ = nλ (30)

2.3 The effect of atomic vibrations — the Debye-Waller factor

• Atoms are always displaced away from their “ideal” positions, primarily due
to thermal vibrations, but also due to crystal defects. This has an effect
on the scattering cross section.

• We can re-write the expression of the scattering amplitude (eq. 19), taking
into account the effect of these displacements (we omit the polarisation
factor [ · 0 ] for simplicity):
X X
A(q) = r0 e−iq·ri fn (q)e−iq·(xn +un,i ) (31)
i n

where un,i is the displacement vector characterising the position of the

atom with label n in the ith unit cell.

• Bragg scattering results from time averaging of the scattering amplitude

(not the cross section). The effect of this is:

15
 Atomic vibrations “smear out” the scattering density, acting, in a sense as an additional
“form factor”.
 The higher the temperature, the more the atoms will vibrate, the more the intensity will
decay at high q. This is easily understood by analogy with the form factor f (q): the
more the atoms vibrate, the more “spread” out the scattering density will be, the
faster the scattering will decay at high q.
 The softer the spring constants, the more the atoms will vibrate, the more the intensity
will decay at high q.
 The lighter the atoms, the more the atoms will vibrate, the more the intensity will decay
at high q.

• The time averaging can be expressed as:

X X
e−iq·ri fn (q)e−iq·xn e−iq·un,i


A(q) = r0 (32)
i n

• The term in hi represents the time averaging and does not depend on the
specific atomic site i. One can show (see extended version of the notes
on web site) that

−iq·u 
e n,i
= e−W (q,n) . (33)

where, in the simple isotropic case

W (q, n) = Un q 2 (34)

• With this, we obtain the general expression for the X-ray structure fac-
tor in the isotropic case

−iq·xn e−Un q 2
P
F (q) = r0 n fn (q)e (35)

• A very similar expression is found for the coherent neutron structure fac-
tors for nuclear scattering.

−iq·xn e−Un q 2
P
F (q) = n bn e (36)

• The corresponding formula for magnetic scattering of neutrons from a

collinear ferromagnet or antiferromagnet is

16
 −iq·xn e−Un q 2
P
F (q) = γN r0 n fn (q)Mn sin α e (37)

where Mn (expressed in Bohr magnetons) reflects both the magnitude

and the sign of the magnetic moment of atom n, and f mn (q) is the
corresponding magnetic form factor and α is the angle between M and
q

• Real crystals are not perfectly periodic, due to the presence of defects,
lattice vibrations and, quite simply, the fact that they are of finite size.

• In scattering experiments, deviation from perfect periodicity results in scat-

tering outside the RL nodes.

• Static defects produce elastic elastic scattering, known as diffuse scatter-

ing because of the fact that it not strongly peaked as Bragg scattering.

• Dynamic effects (such as lattice vibrations) produce inelastic scattering.

• In a diffraction experiment, one does not analyse the energy of the scat-
tered particle, and both effects contribute to the diffuse scattering. Scat-
tering from phonons is known as thermal diffuse scattering.

• Extended defects (involving planes or lines of defects) are described in the

extended version of the notes.

2.4 Finite size effects

• In the case of an infinite perfect crystal, the cross section is a series of δ

functions centred at the RL nodes, This is a result of the infinite sum-
mation over all the real lattice nodes.

• If we carry out a finite summation instead, for example over N1 ,N2 and N3
unit cell in the a1 , a2 and a3 directions, and remembering that

N
sin (2N + 1) x2


X
−inx
e = (38)
sin x2
n=−N

we obtain (in each direction, x in eq. 38 is = q · ai )

"
#
dσ Y sin2 Ni 1 q · ai 2
2
|F (q)|2 ε · ε0

= 2 1 (39)
dΩ sin ( 2 q · ai )
i

17
• This oscillatory function will be in general smeared out by coherence effects
(see long versions of the notes), and can be approximated as

" #
dσ Y 1 2 2
= Nc2 e−(Ni 2 q· ai ) /π |F (q)|2 ε · ε0

(40)
dΩ
i

where the Gaussian functions have variance and FWHM

2π
σi2 =
Ni2 a2i
√
4 π ln 2
F W HM = (41)
Ni ai

• We can therefore conclude that:

 The cross section at a given q is proportional to Nc2 .

 The width in q is inversely proportional to the number of unit cells
along that direction.
 The integrated cross section in three dimensions (remember the
√
Gaussian integral 2πσ 2 ) is therefore proportional to Nc , which
reproduces the result we obtained for the infinite crystal (eq. 8,
lecture 6).

18
3 Lecture 8 - Diffraction experiments and data analy-
sis

3.1 Geometries for diffraction experiments - single crystal diffrac-

tion

In general, the experimental apparatus to perform a diffraction experiment on

a single crystal or a collection of small crystals (powder diffraction) will consist
of (fig. 5 ):

 An incident beam, which can be monochromatic or polychromatic.

 A sample stage, which enables the sample to be oriented and also incor-
porates the sample environment to control a variety of physical (P , T ,
H...) and/or chemical parameters.

 A detector, which includes a detector of photons or particles. This is nor-

mally mounted on a separate arm, enabling the 2θ angular range to be
varied.

Figure 5: The geometry of a “four circle” single-crystal diffractometer. The

“four circles” (actually four axes) are marked “φ”, “χ”, “Ω” and “2θ”. The 2θ
and Ω angles are also known as γ and η.

3.1.1 Scattering triangles for elastic scattering and the Ewald construc-
tion

• As we have seen, the scattering cross section for a single crystal is a se-
ries of delta functions in reciprocal space, centred at the nodes of the
reciprocal lattice.

19
• When a single crystal is illuminated with monochromatic radiation, the scat-
tering conditions are satisfied only for particular orientations of the crys-
tal itself — in essence, the specular (mirror-like) reflection from a family
of lattice planes must satisfy Bragg law at the given wavelength.

With monochromatic radiation, for a generic crystal orientation, no

Bragg scattering will be observed at all.

• Fig. 6 show the geometrical construction used to establish when the scat-
tering conditions are satisfied. Note that here we employ the diffraction
convention: q = kf − ki (see below for the inelastic conventions).

• A typical problem will state the wavelength λ of the incident and scattered
radiation (which are the same, since the scattering is elastic), the sym-
metry and the lattice parameters of the material and the Bragg reflection
to be measured (given in terms of the Miller indices hkl). These data
are sufficient to determine ki = kf = 2π/λ and q (for a right-angle lat-
p
tice q = 2π h2 /a2 + k 2 /b2 + l2 /c2 ; see previous lectures for formulas
to calculate q in the general case).

• Since all the sides of the scattering triangle are known, it is possible to
determine all the angles — in particular the scattering angle γ = 2θ and
the orientation of the incident beam with respect to the lattice required
to be in scattering condition.

• The circle shown in fig. 6 is actually a sphere in 3D, and defines the locus
of all the possible scattering vectors for a given ki . This is known as
the Ewald sphere, from the German physicist Paul Peter Ewald (1888,
1985).

• The maximum value of q is achieved for γ = 2π (backscattering), and is

q = −2ki

• The nodes ”accessible” by scattering are contained within a sphere of ra-

dius 2ki = 4π/λ, centered on the origin of reciprocal space, so that

0 ≤ q ≤ 2ki (42)

3.2 Scattering triangles for inelastic scattering

• For inelastic scattering, the inelastic convention: q = ki − kf is generally

employed, so that qdif = −qine .

20
Figure 6: The procedure to construct the scattering triangle for elastic scat-
tering.

• In an inelastic scattering experiment, the scattered particle loses (energy

loss scattering) or gains (energy gain scattering) part of its energy,
and a corresponding amount of energy is transferred to or from an exci-
tation in the crystal such as a phonon or a magnon. In constructing the
scattering triangle, we should therefore allow for the fact that kf will be
either larger or smaller than ki .

Figure 7: The procedure to construct the scattering triangle for inelastic scat-
tering. (a) energy gain; (b) energy loss.

• The corresponding constructions are shown in 7. One can see that eq. 42
should be replaced by

|ki − kf | ≤ q ≤ ki + kf (43)

• The region of reciprocal space ”accessible” by scattering is bounded by

21
 two spheres of radius ki + kf and |ki − kf |, centered on the origin of
reciprocal space.

• Maximum and minimum q are achieved in backscattering and forward scat-

tering, respectively.

3.3 Powder diffraction and the Debye-Scherrer cones

• A “powder” sample is a more or less “random” collection of small single

crystals, known as “crystallites”.

• The cross section for the whole powder sample depends on the modulus
of the scattering vector q but not on its direction. Therefore, when a
powder sample is illuminated, scattering is always observed (unlike the
case of a single crystal).

• For a monochromatic incident beam, the 2θ angle between the incident

and scattered beam is fixed for a given Bragg reflection, but the angle
around the incident beam is arbitrary. The locus of all the possible
scattered beams is a cone around the direction of the incident beam
(fig. 8).

• All the symmetry-equivalent RL nodes, having the same q, contribute to

the same D-S cone (fig. 9).

• Accidentally degenerate reflections, having the same q but unrelated hkl’s,

also contribute to the same D-S cone. This is the case for example,
for reflections [333] and [115] in the cubic system, since 32 + 32 + 32 =
12 + 12 + 52 .

Key points to retain about powder diffraction

 In powder diffraction methods, the intensity around the D-S cones is

always integrated, yielding a 1-dimensional pattern.

 Powder diffraction peaks are usually well-separated at low q, but be-

come increasingly crowded at high q often becoming completely
overlapped. This substantially reduce the amount of information
available to solve or refine the structure precisely (see below).

3.4 Integrated Intensities

• Exam problems will not be concerned with peak fitting, and you will be given
integrated intensities of some form. These intensities will be usually

22
Figure 8: Debye-Scherrer cones and the orientations of the sets of Bragg
planes generating them.

Figure 9: Ewald construction for powder diffraction, to represent the crystal

being rotated randomly around the direction of the incident beam (the figure
actually shows the opposite, for clarity). Note that many RL nodes are simul-
taneously in scattering — those that have the same |q|. Symmetry-equivalent
reflections have the same |q| and also the same structure factor. The powder
intensity is therefore multiplied by the number of symmetry-equivalent reflec-
tions, known as the multiplicity.

corrected for the Lorentz - polarisation (i.e., P(γ) ), angle-dependent

attenuation (typically due to X-ray or neutron absorption) and incident
flux terms, but the role these terms may be requested as part of the
discussion. The integrated intensity can always be reduced to a dimen-
sionless quantity (counts).

• The general expression for the integrated intensity (number of particles) is

23
 d3
 
Pτ = N c mτ |F (τ )|2 P(γ)L(γ)Aτ (λ, γ)Finc (44)
v0

- Nc is the number of unit cells in the sample.

- d is the d-spacing of the reflection.
- v0 is the unit-cell volume.
- mτ (powder diffraction only) is the number of symmetry-equivalent
reflections. This accounts for the fact that in powder diffraction
these reflections are not separable, and will always contribute to
the same Bragg powder peak (see previous discussion and fig. 9.
- P(γ) is the polarisation factor (dimensionless), which we have already
introduced.
- L(γ) is the so-called Lorentz factor (dimensionless), and contains
all the experiment-specific geometrical factors arising from the δ-
function integration.
- Aτ (λ, γ) (dimensionless) is the attenuation and extinction coefficient,
which account for the beam absorption and for dynamical effects.
- Finc is the incident time-integrated flux term (counts per square me-
tre), which accounts for the strength of the incident beam and for
the counting time.

• The product (d3 /v0 )P(γ)L(γ)Aτ (λ, γ), is sometimes called the LPGA fac-
tor (Lorentz-Polarisation-Geometrical-Attenuation), and is used to cor-
rect the raw data. When absolute incident flux measurements are not
available, one obtains a pattern where the intensities are proportional
to the square of the structure factor, the proportionality constant being
a scale factor.

3.5 Structural solution from diffraction data

3.5.1 The phase problem

• From eq. 20, we can see that the structure factor is proportional to
the Fourier transform of the charge density (or, more in general,
scattering density) integrated over the unit cell.

• By the elementary theory of the Fourier transform over a finite interval (ex-
tended to 3 dimensions) we can calculate the charge density given all
the structure factors:

24
 1 X
ρ(x) = F(τ )eiτ ·x (45)
r 0 v0 τ

• From eq. 45 follows that if we were able to measure all the structure factors,
we could reconstruct the charge density exactly. Clearly, it is impossible
to measure all the infinite nodes of the reciprocal space, but it can be
shown that it would be sufficient to measure up to a value of qmax to
obtain a Fourier map with resolution 2π/qmax in real space.

• However, direct reconstruction of the charge density from diffraction data

is impossible, because only the amplitudes of the structure factors are
known (through the term |F |2 in the cross section), while the phases are
unknown. Solving a crystal structure is therefore equivalent to phasing
the reflections.

• A set of mathematical methods, known as direct methods, have been

developed to phase reflections without any a priori knowledge of the
crystal structure. They exploit the fact that the Fourier maps are not
completely arbitrary, but are positive (for X-rays) and atomic-like.

3.5.2 The Patterson method

• It is nonetheless possible to obtain some degree of information about scat-

tering densities without any knowledge of the phases. Again, from eq.
20, we obtain:

ZZ
0
|F (q)|2 = r20 dxdx0 ρ(x)ρ(x0 )e−iq·(x−x ) (46)
unit cell

With some manipulations we obtain:

Z
1 X 2 iτ ·x
|F (τ )| e = dx0 ρ(x0 )ρ(x + x0 ) = P (x) (47)
r02 v0 τ unit cell

• The function defined in eq. 47 is known as the Patterson function (from

Lindo Patterson, 1934). One can perhaps recognise in eq. 47 that the
Patterson is the autocorrelation function of the scattering density.

• Patterson functions are 3-dimensional functions defined within one unit cell,
and are usually presented in the form of 2-dimensional “slices”.

25
• Atomic-like scattering densities are mostly zero, except at the atomic po-
sitions. Therefore the Patterson function will be mostly zero as well,
except at the origin (x = 0) and for values of x corresponding to vectors
joining two atoms. At these vectors, the Patterson function will have
peaks.
2
P
• The integral of the x = 0 peak is easily calculated to be i Zi , where Zi
is the atomic number of atom i and the sum is over all atoms in the unit
cell.

• Likewise, the integral of a peak corresponding to the interatomic distance

rij is Zi Zj .

3.5.3 Structural optimisation: least-square refinements

• Most problems in physical crystallography involve determining subtle struc-

tural variations from well-known and rather simple structural motifs. There-
fore, structural optimisation is usually the method of choice for the struc-
tural condensed-matter physicist.

• If one is reasonably close of the solution, with only a few free parameters
left to determine, it is possible to minimise the agreement between ob-
served and calculated squared structure factors |F |2 as a function of
the free parameters. This is clearly a non-linear optimisation problem,
and a number of strategies have been developed to solve it in a variety
of cases.

• The best known structural optimisation method is known as the Rietveld

method, and is applied to powder data.

• In the Rietveld method, one performs a nonlinear least-square fit of the

measured profile, rather than of the |F |2 as in the single-crystal meth-
ods. This could appear more complicated, since one has to fit the mi-
crostructural and instrumental parameters controlling peak broadening
and the background at the same time, but has the great advantage of
accounting automatically for peak overlap.

26
4 Appendix I: more on the theory of scattering

4.1 Scattering from a free electron (extended derivation)

Thomson scattering is a classical phenomenon, and can be explained as fol-

lows: as electromagnetic radiation impinges on a free electron, the oscillating
electric field subjects it to an acceleration. The scattered radiation is the ra-
diation emitted by the accelerated charge. If a charge is accelerated and is
observed in a reference frame where its velocity is small compared to that of
light, the electric field can be written as

 
(−e) n × (n × a)
E(R, t) = (48)
4π0 c2 R ret

where a is the acceleration and n is a unit vector along the segment con-
necting the particle to the observer and pointing towards the latter (fig. 10).
The subscript “ret” means “retarded” (see below).

Note that if we keep rotating the scattered beam in the direction shown in fig. 10, i.e., in
the plane containing the incident polarisation, the projection a⊥ will eventually become
zero at 90◦ . On the other hand, if we rotated perpendicular to the incident polarisation,
the projection would obviously stay the same. In a nutshell, this is the essence of the
polarisation factor.

Before we discuss eq. 48, it is worth reminding the important vector relation
(used also later on in the course). For any three vectors:

a × (b × c) = (a · c)b − (a · b)c (49)

n × (n × a) = − [a − (a · n) n] (50)

The expression in square brackets on the right side of eq. 50 is the projection of a on
the plane perpendicular to n

In eq. 48, the square bracket with subscript “ret” means that the quantity in
the bracket is to be evaluated at the “retarded” (i.e., earlier) time t − R/c. R is
the distance between the “retarded” position and the observer (fig. 10). Since
we are interested in the radiation at a great distance from the particle and at
oscillatory, small-amplitude motions of the latter, we can replace R and n with
constant quantities referring to the average position of the particle.

27
Figure 10: Diagram illustrating the Thomson scattering of X-rays from a free
electron. The acceleration due to the incoming wave is a, whereas its projec-
tion perpendicular to the line of sight is a⊥ .

The electric field due to an accelerated particle is proportional to the projection of the
acceleration perpendicular to the line of sight.

Let us consider a beam of polarised X-rays (i.e., with a well-defined direction

of E, identified by the unit vector ) impinging on a free electron (again, refer
to the scheme in fig. 10). The particle with be accelerated by the electric field
E0 e−iωt of the incoming wave so that

(−e)
a(t) = E0 e−iωt (51)
m

Substituting eq. 51 into eq. 48 we obtain:

28
 e2 e−iω(t−R/c) e2 ei(kR−ωt)
E(R, t) = E0 [n × (n × )] = E0 [n × (n × )]
4π0 mc2 R 4π0 mc2 R
(52)

ω e2
where k = is the wavenumber. The quantity r0 = = 2.82 × 10−15
c 4π0 mc2
m is known as the classical electron radius. By applying eq. 49 once again
we find that the expression in square bracket on the right side of eq. 51 is

[n × (n × )] = − [ − ( · n) n] (53)

i.e., it is minus the component of  perpendicular to n.

The scattered polarisation 0 is the projection of the incident polarisation perpendicular

to the line of sight.

Figure 11: Diagram illustrating the convenstional σ and π reference directions

for the incident and scattered polarisation. Note that σ · 0σ = 1 always.
Conversely, π ·0π = cos γ depends on the scattering angle γ, and vanishes
for γ = π/2.

It is useful at this point to introduce two orthogonal reference components of 

and of the scattered polarisation 0 , according to the scheme shown in fig. 11:
π and 0π are in the “scattering plane”, defined by the incident and outgoing
directions, whereas σ = 0σ are perpendicular to the scattering plane and
are equal. We can decompose the incident polarisation as

29
  = cos ξ σ + sin ξ π (54)

so that the incident wave has σ-polarisation for ξ = 0 and π-polarisation for
ξ = π2 . The incident polarisation makes an angle ξ with the vector σ . With a
little geometry we conclude

ei(kR−ωt) 
cos ξ 0σ + sin ξ cos γ 0π

E(R, t) = −r0 E0 (55)
R

where γ is the angle between the incident and scattered wavevector (this
angle is also known, by longstanding diffraction convention, as 2θ). Based on
eq. 55, we can make the following observations:

• A plane wave impinging on a quasi-free charge produces a scattered spherical wave

ei(kR−ωt)
, with an amplitude that in general depends on the scattering angle γ.
R
• If the incident wave is σ-polarised, the scattered wave is σ 0 -polarised, and has amplitude
r0
R E0 .

• If the incident wave is π-polarised, the scattered wave is π 0 -polarised, and has amplitude
r0
R E0 cos γ.

• The intensity of the scattered wave is zero for scattering of π polarisation at 90◦ .

• The scattered wave has a phase shift of π upon scattering (minus sign).

The instantaneous energy flux of the scattered wave is given by the Poynting
vector:

S = E × H = 0 c|E|2 n (56)

The average power radiated per unit solid angle in both polarisations is there-
fore

dP 0 c 2 2 0 c r02 2  2
= R2 h|S|i = E0 cos ξ + sin2 ξ cos2 γ

R |E| = (57)
dΩ 2 2

It can also be shown that the power radiated for an arbitrary final polarisation
0 is

0 c r02 2 
 
dP 2
= E0  · 0 (58)
dΩ 0 2

30
As appropriate for a scattering process, it is convenient at this point to intro-
duce the cross section, defined as the average power radiated per unit solid
angle divided by the average incident power per unit area (power flux, Φ),
which is

0 c 2
Φ= E (59)
2 0

The cross section into both final polarisation channels (i.e., if the scattered
beam is measured without an analyser) is therefore

dσ
= r02 cos2 ξ + sin2 ξ cos2 γ
 
(60)
dΩ

whereas for an arbitrary final polarisation 0 is

 
dσ
= r02 [ · 0 ]2 (61)
dΩ 0

For an unpolarised X-ray beam, for which all the angles ξ are equally repre-
sented, the cross section becomes

1 + cos2 γ
 
dσ
= r02 (62)
dΩ 2

NOTE: our discussion on the incident and final polarisations is relevant be-
cause different X-ray sources have different characteristics. The beam from a
typical lab X-ray machine is unpolarised, and so eq. 62 applies. Synchrotron
radiation is naturally polarised in the plane of the electron orbit,so that ξ = 0 in
eq. 60. Unlike the case of lab machines, synchrotron diffractometers have a
vertical scattering plane, so that the 90◦ -cross section does not vanish. Syn-
chrotron beamlines specifically designed for resonant or magnetic scattering
often have the ability to change the incident polarisation to the π channel
or even to produce circularly-polarised X-rays. in addition, it is possible to
analyse the scattered polarisation as well.

31
4.2 Thomson scattering from many quasi-free electrons (extended
derivation)

The Thomson formula can be easily extended to the case of multiple scat-
tering centres, provided that the amplitude of the motion of each electron
is small with respect to the wavelength. What we aim to achieve is to find
an expression for the X-ray scattering amplitude and cross section of a multi-
electron atom. In this case, the radiation emitted by each electron at position
xi will be characterised by an approximately time-independent phase factor
eik·xi , accounting for the fact that the charge is not at the origin, k being the
wavevector of the incident radiation. Also, the radius R in eq. 55 needs to be
replaced with individual radii Ri . Here, we can employ the very useful trick of
approximating

Ri ≈ R − n · xi (63)

This is equivalent to considering diffraction in the “far field” limit (Fraunhofer

diffraction). Eq. 63 can be obtained by writing the vector relation;

Ri = |Ri | = |R − xi | = |Rn − xi | (64)

and expanding it to the quadratic term in xi /R; is certainly valid in the case
we are interested in, where the distances between scattering centres are
comparable to atomic sizes whereas R (the experimental scattering path) is
macroscopic. By summing the amplitudes of individual scattering centres we
obtain

i(kR−ωt)
E(R, t) = −r0 E0 e [ · 0 ] i(ki −kf )·xi
P
R ie
i(kR−ωt)
= −r0 E0 e [ · 0 ] −iq·xi
P
R ie (65)

where ki and kf are the incident and scattered wavevectors and q = kf − ki .5

5
Throughout this part of the course, we will employ the convention that q is the change of
wavevector of the particle or photon, so q = kf − ki . the convention q = ki − kf identifies
q with the wavevector transferred to the crystal, and is widely employed particularly in the
context of inelastic scattering

32
5 Appendix II: X-ray absorption and resonant scatter-
ing

5.1 X-ray absorption: the photo-electric effect and X-ray fluores-

cence

When the beam of X-rays impinges at normal incidence on a slab of mate-

rial of thickness L, it suffers both scattering and absorption, and is therefore
attenuated according to the familiar equation:

I = I0 e−µL (66)

where µ is the linear attenuation coefficient, which is related to the total

cross section σtot (scattering plus absorption) by the equation:

µ = σtot Na (67)

where Na is the number of scattering/absorption centres (here atoms) per

unit volume. Fig. 12 shows a comparison of the cross sections of different
processes leading to X-ray attenuation in materials.

At X-ray energies used in typical experiments (5–50 KeV), the photo-electric absorption
cross section is by far the largest contributor to X-ray attenuation.

Key facts about the photoelectric absorption of X-rays

• In the photo-electric absorption process, a photon is completely absorbed and the energy is
transferred to a core electron (i.e., an electron in the inner atomic shell), which is excited
into unoccupied bound states above the Fermi energy or in the continuum.

• The key parameter controlling this effect is, naturally, the electron binding energy, which
is typical of each shell and atom or ion. No photoelectric absorption can take place
if the photon energy is lower than the binding energy of the electron. This is in
complete analogy with the more familiar photo-electric effect in semiconductors. As the
photon energy is increased through a binding energy “threshold”, additional absorption
can take place, so absorption increases abruptly. This is known as an absorption edge
(see fig. 13).

• Binding energy are classified based on the quantum numbers of the core electron. A
letter indicates the principal quantum number of the core electron, so “K” for n=1,
“L” for n=2, “M ” for n=3 “N ” for n=4 and so on. This is followed by a roman subscript
indicating energy sub-levels. Therefore, the K edge indicates a transition from the 1s
core state. LI , LII and LIII indicate transitions from the 2s, 2p1/2 and 2p3/2 , respectively
(2p1/2 and 2p3/2 having distinct values of the total angular momentum J).

33
Figure 12: Contributions to the X-ray scattering and absorption cross section
for the element carbon (C), from the International Tables for Crystallography,
vol. C, p 213. In this figure, σel is the Thomson cross section, σinel is the
Compton inelastic cross section and σpp is the pair-production cross section,
whereby a high-energy photon produces an electron-positron pair. Note the
K absorption edge in the photo-electric cross section at 284.2 eV

• An X-ray photo-electric absorption event is followed by a chain of X-ray emission (or fluo-
rescence) events whereby the excited atom gradually relaxes to the ground state. The
processes of absorption and subsequent re-emission are shown schematically in fig. 14.
Strong emission lines are those that follow the dipole selection rules, i.e., ∆l = ±1 and
∆j = 0, ±1. For example, for transition metals, there are 3 strong emission lines: Kα1 and
Kα2 for transitions from 2p1/2 and 2p3/2 to 1s and Kβ for transitions from 3p to 1s. X-ray
emission is extensively employed to produce monochromatic X-ray radiation (see below).

• Far from absorption edges, photo-electric absorption decreases as a function of pho-

ton energy, following the very approximate law:

Zn
σph ∝ (68)
(~ω)3

where Z is the atomic number and the exponent is a number between 4 and 5.

• A list of characteristic absorption and emission X-ray energies can be found in the Interna-
tional Tables for Crystallography, vol. C, starting from p 206.

34
Figure 13: Attenuation length (1/µ) in microns for elemental
iron (Fe), in the energy interval between 30 eV and 25 KeV.
The figure was generated using the attenuation calculator in
http://henke.lbl.gov/optical constants/atten2.html. Note the three absorption
edges: K at 7.112 KeV, L (actually three edges at 845 eV, 720 eV and 707
eV) and M (edges at 90 eV and 50 eV).

Figure 14: Schematic representation of the absorption and subsequent emis-

sion for a K edge event. Left: a photon with energy above the K edge is
absorbed and an 1s electron is ejected above the Fermi energy or into the
continuum. Right: an electron in the 2p1/2 shell makes a transition to fill
the 1s core hole, resulting in the emission of a photon with the characteristic
energy Kα2 .

5.2 X-ray scattering beyond the free-electron approximation

Up to now, we have left the issue of “free” electrons somewhat ambiguous.

Truly “free’ electrons (e.g., conduction electrons in a metal) hardly contribute
to the scattering of X-rays, because their probability distribution extends a
long way throughout the crystal, and the form factor decays very rapidly away
from forward scattering. Conversely, the largest contribution to X-ray scatter-
ing from atoms is given by “core” electrons, which are close to the nucleus
and have slowly decaying from factors — but these electrons are certainly

35
not free! Indeed, there are large departures from the Thomson scattering
formula near atomic resonances, where the energy of the photon is just suf-
ficient to eject an electron from a core state into the continuum. Away from
resonances, the Thomson formula can be corrected to a very good approxi-
mation by replacing the form factor by the complex quantity

f (q) = fThom (q) + f 0 (~ω) + if 00 (~ω) (69)

where the so-called anomalous terms, f 0 and f 00 , away from atomic reso-
nances do not depend on q and are weak functions of the photon energy.

It can be shown, as a consequence of the so-called optical theorem, that the imaginary
part of the scattering factor is proportional to the linear absorption coefficient due to the
photoelectric effect.
ω
f 00 (~ω) = µ (70)
4πr0 cNa
where Na is the number of atoms per unit volume, and the other symbols have the usual
meaning. The quantity µ is the linear absorption coefficient, defined in eq. 66.

5.3 Resonant scattering and anomalous corrections to the form

factor

Here, we repeat essentially the same derivation that led to the calculation of
the Thomson cross section, but replacing the “free-electron” acceleration in
eq. 51 with an expression appropriate to a damped oscillator.

m ẍ + γ ẋ + ωi2 x = (−e)E(t)
 
(71)

(−e) ω2
a(t) = E0 e−iωt 2 (72)
m ω − ωi2 + iωγi

where ωi is the resonance frequency of each electron. If we use 72 instead

of 51 in the previous derivations, we obtain in general a complex scattering
amplitude. It is easy to evaluate this in the case ω  ωi , i.e., for example, for
scattering of high-energy X-rays (> 20KeV ) from first-period transition-metal
ions. In this case

36
 ω2
 
X γi
f (q) = [f (q)T hom ]i 1 − i2 + i (73)
ω ω
i

5.4 Compton scattering

One important issue related to the bound nature of electrons is the fact that no
elastic scattering is possible for a truly free electron, but the Compton formula
applies instead:

k0 1
= (74)
k 1+ ~ω
mc2
(1 − cos γ)

At low photon energy, the photon momentum is transferred to the atom as a

whole (much heavier) or to the entire crystal, and the scattering is elastic to
a very good approximation. As the energy increases nearing the rest mass
energy of the electron (mc2 = 511 KeV), the Compton scattering cross section
increases.

6 Appendix III: Scattering of particle beams

A plane wave of X-rays impinging on a quasi-free distribution of charges pro-

duces a spherical wave, the amplitude of which is proportional to the incident
amplitude E0 . The coefficient multiplying the spherical wave is

r0 f (q) cos ξ 0σ + sin ξ cos γ 0π

 
(75)

can be complex. In the general case, for a given wavelength, this coefficient
depends on both angular variables of the scattered beam, but for spherically-
symmetric atoms it depends only on the scattering angle γ = 2θ. Impor-
tantly, the cross section is equal to the square modulus of the spherical
wave coefficient. As we shall see shortly, the same principles applies to
the scattering of particle beams, provided that the particle beam is described
quantum-mechanically, so that the wave-like nature of the particles is appar-
ent. Indeed, quantum mechanics is essential to obtain exact results, although
the essential features are often classical or semi-classical in origin. In addi-
tion, it is often convenient to discuss the stationary problem rather than the

37
time-dependent problem of a single particle starting off far away from the scat-
terer. The stationary problem is equivalent to considering a steady streams
of particle coming from infinity, which is partly converted into a current of
scattered particles in the form of a spherical wave. Finally, provided that we
operate in the non-relativistic limit, we can consider the 2-particle scattering
problem to be equivalent to that of scattering from a static potential V (r),
provided that the mass of the particle is replaced by the effective mass

m1 m2
mef f = (76)
m1 + m2

Hereafter, we will simply refer at the effective mass as m.

6.1 Wavefunction equation for the static scattering problem

We are looking for the solutions ψ(r) for the following eigenfunction equation

~2 2
 
− ∇ + V (r) ψ(r) = Eψ(r) (77)
2m

For the scattering problem, we are not interested in the bound states, so we
will focus on the continuous spectrum with positive eigenvalues. We also
expect that, at long distances from the origin of the potential, the solution will
~2 k2
approximate a plane wave (at least on one side), with energy E = 2m . By
2m
defining the new potential U (r) = ~2
V (r) we arrive at the equation:

∇2 + k 2 ψ(r) = U (r)ψ(r)


(78)

The solution of eq: 78 with the right side set to zero is clearly a plane wave

ϕ(r) = eik·r (79)

Eq. 79 represent the free-particle limit of the wavefunction, i.e., the incident
wave. Therefore, we will later employ 79 with k = ki .

An important step towards the solution of the general eq. 78 is to solve the
point-source equation:

∇2 + k 2 ψ(r) = δ(r)


(80)

38
The solutions of equations of the type 80 are known as Green’s functions.
It can be shown rather straightforwardly that the following two functions are
solutions of eq: 80

1 eik|r|
G+ (r) = −
4π |r|
1 e−ik|r|
G− (r) = − (81)
4π |r|

To verify that 81 are indeed solutions of eq: 80 it is sufficient to use the relation

∇2 (ab) = a∇2 b + b∇2 a + 2∇a · ∇b (82)

The two solutions in eq. 81 are expanding and contracting spherical waves,
respectively. By selecting the former, we can rewrite eq. 78 in an integral
form (we only write the solution with the expanding Green function):

Z
ψ(r) = ϕ(r) + dr0 G+ (r − r0 )U (r0 )ψ(r0 ) (83)

Since we are primarily interested in solutions far away from the region where
the potential is non-zero, we can employ the so-called far-field approximation.
The key observation here is that r0 is small, because the potential is non-
zero only near the scatterer. Also, importantly, |ki | = |kf | = k, since we are
dealing with elastic scattering. The correspondence between eq. 63 and
eq. 84 is clear if we write kf = kn

k|r − r0 | ≈ kr − kf · r0 (84)

after some simple algebra we obtain (again, at long distances from the scat-
tering centre):

 Z  ikr
ik·r 1 0 −ikf ·r0 0 0 e
ψ(r) = e + − dr e U (r )ψ(r ) (85)
4π r

Eq. 85 has a very pleasing form: it is very similar to eq. 55, and we could
surmise that the squared modulus of the expression in square brackets (the
“scattering amplitude”) is the cross section (this is proven in most standard
quantum mechanics textbooks by introducing the probability current density).

39
Here, we will take it for granted without further proof. Unfortunately, the scat-
tering amplitude in eq. 85 depends on the wavefunction itself. We can make
progress by assuming that potential is in some sense “small”, so that the
spherical wave is a small component of the overall wavefunction. We can
therefore attempt to expand the integral term in series:

Z
1 0
− dr0 e−ikf ·r U (r0 )ψ(r0 ) ≈
4π
2 Z 0
eik|r−r |
Z 
1 0 −ikf ·r0 0 0 1 0 0
− dr e U (r )ϕ(r ) + drdr ϕ(r ) U (r)U (r0 ) +(86)
...
4π 4π |r − r0 |

The series in eq. 86 is known as the Born series; taking the first term alone
is know as the first Born approximation. It is easy to obtain the expression
for the scattering cross section in the first Born approximation; remembering
that ϕ(r0 ) is the plane wave exp(iki · r0 ), we obtain:

   Z 2
dσ  1 −iq·r

=
 dre U (r) (87)
dΩ Born 4π

where q = kf − ki . In other words, in the first Born approximation, the

cross section is proportional to the Fourier transform of the potential
energy.

6.2 Elastic scattering of electrons in the 1st Born approximation

As a first example of scattering of a particle beam, we will consider the elas-

tic scattering of electrons from the Coulomb potential produced by the nu-
cleus and by the electrons bound in an atom. For simplicity, we will consider
electrons as spinless particles. Based on eq. 87, we need to calculate the
potential, and for this is useful to recall that

∇2 e−iq·r = −q 2 e−iq·r (88)

2m
Recalling that U (r) = ~2
V (r) , we can rewrite the integral in eq. 87 as

Z Z Z
1 m m
dre−iq·r U (r) = − drV (r)∇2 e−iq·r = − dre−iq·r ∇2 V (r)
4π 2πq 2 ~2 2πq 2 ~2
(89)

40
where the rightmost expression is obtained by integrating twice by part and
assuming that both the potential and its first derivative are zero at infinite dis-
tance from the origin. We can now use Poisson’s equation and the expression
for the charge density of an atom of atomic number Z:

ρ
∇2 Φ = (90)
0
V = (−e)Φ (91)
ρ(r) = Zeδ(r) + (−e)ρel (r) (92)

where Φ is here the electrostatic potential and ρel (r) is the same electron
density we have employed for X-rays. To find the expression for the elastic
scattering cross section:

2  2
e2 m
 Z
dσ Z − dre−iq·r ρel (r)
 
= (93)
dΩ 0 2πq 2 ~2  

We can see the immediate analogy with the scattering cross section for X-
rays. The integral term is exactly the same as the X-ray form factor, and is
referred to as fX in the text. The term in brackets takes the place of the
classical electron radius, and is numerically

e2 m
= 3.38 × 10−10 m Å−2 (94)
0 2π~2

e2
whereas, for comparison, the classical electron radius is r0 = =
4π0 mc2
2.82 × 10−15 m. We can see therefore that for all typical values of q (1–10 Å−1 )
employed for electron diffraction experiments, the scattering amplitudes for
electrons is much larger than for X-rays.

The elastic scattering amplitude for electrons f B is often expressed in wave-

length/angle units as in eq. 95. One can find tabulated values for neutral and
ionised atoms in the International Tables for Crystallography, volume C, p 263
(λ is the electron wavelength and both λ and f B in eq. 95 are expressed in
Å):.

f B (sin θ/λ) = 0.023934λ2 [Z − fX (sin θ/λ)]/ sin2 θ (95)

41
6.3 Nuclear scattering of “slow” neutrons

The kinetic energies of neutrons employed in typical scattering experiments

is much lower than the energies of nuclear reactions, which are typically in
the MeV range. In fact, it is established from nuclear physics that the neutron-
nuclear interaction can be approximated with a square potential well of depth
1
V0 ≈ 50 MeV and range d = 1.3A 3 × 10−15 m, where A is the mass number
of the nucleus. It is therefore implausible that the first Born approximation
could be applied as such to the scattering of slow neutrons. In fact, a more
detailed calculation shows that the first Born approximation is applicable only
to neutrons with E > 25 MeV.

Nevertheless, Fermi proposed in 1936 that slow neutron scattering could still
be satisfactorily treated in the first Born approximation. The general idea
here is that the wavelength of the neutrons is so large that one can replace
the real potential with a “pseudopotential” (known as the Fermi pesudopoten-
tial), which yields the same first Born scattering amplitude and for which the
approximation definitely applies. For this, one observes that, over the range
of the true potential, the phase of the neutron is essentially constant in eq.
86, and the scattering amplitude can be written as:

Z
m m
f =− drV0 (r) = − V0 d3 (96)
2π~2 2π~2

We could therefore think of decreasing the depth of the potential well into the
thermal ranges by increasing the range of the potential by a factor of, ∼ 1000,
while maintaining the same value of the scattering amplitude and, crucially,
still keeping the range much smaller than the neutron wavelength — a set
of conditions for which the first Born approximation definitely applies. This
conjecture led to the development of the Fermi pseudopotential, which has
the form:

2π~2
VF (r) = bF δ(r) (97)
m

where bF is the scattering length, known as the Fermi length. In practice, the
Fermi pseudopotential is completely satisfactory to describe nuclear neutron
scattering for diffraction experiments, and only need corrections (analogous
to the X-ray anomalous corrections) only for energies near neutron-nuclear
resonances.

Here are some important facts about neutron scattering lengths and cross

42
sections:

• Neutron scattering amplitudes do not depend on q, i.e., they carry no form

factor. For diffraction experiment this is crucially important, because it
means that the intensity of the diffraction features does not decay at
high angles as fast as in the case of X-rays (we will see in the next
lecture that thermal motion causes high-q Bragg intensity decay even in
the case of neutrons). They are also largely independent on the neutron
energy, at least in the regime of interest for neutron diffraction.

• Fermi lengths for typical nuclei are of the order of a few fm (10−15 m), which
means that they are comparable to the classical electron radius. How-
ever, atoms have a single nucleus and many electrons, so X-ray scat-
tering cross sections in the forward direction are typically much larger
than neutron cross sections. Neutron cross sections are traditionally
measured in barns (10−28 m2 ).

• Neutron absorption is also much lower than for X-rays, which, together
with the previous observation, means that attenuation (absorption +
scattering) lengths are of the order of cm for neutrons and of µm for
X-rays. This has a profound effect on the design of diffraction experi-
ments exploiting the two types of radiation.

• Fermi lengths vary across the periodic table without any particular regu-
larity. They can be positive or negative depending on the sign of the
nuclear potential. They depend on the isotope, often very strongly, and
also on the direction of the nuclear spin with respect to the neutron
spin. This has two main consequences, one “positive” and one “nega-
tive”: the “negative” consequence is the presence of “incoherent” scat-
tering, due to the random mixture of different isotope and spin orienta-
tions, which contributes to the experimental background. The “positive”
consequence is the possibility to exploit different isotopes of the same
element to gain additional contrast.

6.4 The Fermi golden rule and its connection with the first Born
approximation

In the previous sections, we have considered the Born series for a spinless
particle beam, so that there was no internal degree of freedom to consider.
In addition, the internal state of the scatterer (in our case a time-independent
potential) was also unchanged during the collision. Consequently, the energy

43
of the scattered particle remains the same after the collision. It is useful to
extend our description of the scattering to cases in which there are internal
degrees of freedom which may be altered in the collision. The Fermi golden
rule (the development of which is mainly due to Dirac) is widely employed
to calculate the transition probability of an eigenstate (in our case, a plane
wave), into a continuum of “final” states due to a perturbation. As in the case
of the first Born approximation, it is the first term of a perturbative expansion,
and is applicable within very similar conditions to the first Born approximation,
namely that the final states do not significantly deplete the original eigenstate.
Here, we only quote the general result for the scattering cross section be-
tween states with wavevectors ki and kf and an internal transition between
states λi and λf with energies Ei and Ef , respectively.

dσ kf    2
= (m/2π~2 )hkf , λf V̂  ki , λi i δ(~ω + Ei − Ef ) (98)
  
dΩdEf ki

where we have employed the familiar hbra| – |keti notation, V (r) is the “per-
turbing” potential and ~ω = ~2 (kf2 − ki2 )/2m, ensuring energy conservation.

It is a simple exercise to show that, for elastic scattering in the absence of in-
ternal degrees of freedom, the “golden rule” cross section in eq. 98 is exactly
the same as the first Born approximation result in eq. 87.

6.5 Magnetic scattering of neutrons

We will exemplify the application of the Fermi golden rule by outlining the
calculation of the scattering of slow neutron onto a magnetic atom. We will
assume that no energy is exchanged in the process. Here, the interaction
potential is naturally the dipole interaction between the neutron spin S and the
magnetic field B(r) generated by the electrons(we will assume zero external
magnetic field for simplicity). When dealing with elastic magnetic scattering,
the initial and final states of the atoms are assumed to be the same, implying
that conservation of linear and angular momenta is ensured by the crystal as
a whole. Therefore, the operator quantities (here indicated explicitly with a “
b ”) are only those acting on the neutron coordinates.

V (r) = −γN µN Ŝ · B(r) [Joules] (99)

where Ŝ is the neutron spin operator, γN is the neutron gyromagnetic ratio

e~
(−1.9130418) and µN is the nuclear magneton ( 2m p
= 5.05078324(13) × 10−27

44
JT−1 ). The magnetic field of a single electron moving with velocity v is

 
µ0 µe × r̂ −eµ0 v̂ × r̂
B(r) = ∇ × 3
+ (100)
4π r 4π r3

Where µe is the magnetic moment of the electron, given by (µB is the Bohr
magneton, s is the spin of the electron)

µe = −2µB s (101)

The two terms in eq. 100 represent the spin and orbital part of magnetic
moment. The “hatted” vector quantities in eq. 100 are meant as operators
(so, for instance,
v̂ = −i~∇ (102)

The derivation of the cross section in in eq. 98, even in the general case of
inelastic scattering and is quite straightforward, and is reported, for example,
in Stephen W. Lovesey, “Theory of neutron scattering from condensed mat-
ter”, Oxford Science Publications, Clarendon Press, Oxford (1984) — Volume
2. Here, for simplicity, we only report the final result for elastic scattering of
unpolarised neutrons, i.e., we are averaging on the initial and final neutron
spins:

dσ 2 2 †
= γN r0 Q⊥ · Q⊥ (103)
dΩ

where

• r0 is the classical electron radius. This means that the scattering ampli-
tude for a neutron by the magnetic field of a single electron is compara-
ble to the Thomson scattering amplitude of X-rays.

• Q is given by the formula (spin only scattering)

X
Q= eiq·ri si (104)
i

and, in the general case, is the Fourier transform of the magnetisa-

tion density (spin + orbital) for unpaired electrons. In simple cases,

Q = fm (q)µ (105)

45
 µ is the magnetic moment of the atom and and fm (q) — the magnetic
form factor — takes the place of the X-ray form factor in the analogous
expression for Thomson scattering.

• Q⊥ is the projection of Q perpendicular to the scattering vector q

1
Q⊥ = q×Q×q (106)
q2

in practical terms, this means that neutrons are only sensitive to the
components of the magnetic moments perpendicular to the scat-
tering vector.

• If we take into account the direction of the neutron spins, the cross section
will contain terms of the kind Ŝ · Q⊥ and depends on the initial and
final states of the neutron. For pure magnetic elastic scattering, the
cross section is non-zero only for opposite incident and scattered
neutron spins — we say that it is “spin flip” only.

7 Appendix IV: Useful constants

0 = 8.854187 × 10−12 F m−1 — vacuum permittivity

µ0 = 4π × 10−7 N A−2 — vacuum permeability (exact)
c = 2.99792458 × 108 m s−1 — speed of light in vacuum
e = 1.602176487(40) × 10−19 C — unit charge

me = 9.10938215(45) × 10−31 Kg — electron rest mass

mp = 1.672621637(83) × 10−27 Kg — proton rest mass
mn = 1.674927729(28) × 10−27 Kg — neutron rest mass

e2
r0 = = 2.82 × 10−15 m — classical electron radius
4π0 mc2
e~
µN = = 5.05078324(13) × 10−27 JT−1 — nuclear magneton
2mp
γN = −1.9130418 — neutron gyromagnetic ratio
e~
µB = = 9.27400915(23) × 10−24 JT−1 — Bohr magneton
2me

46