Title: Part 15. Patterson Function and Diffuse Scattering
 1Part 15. Patterson Function and Diffuse 
Scattering (From Transmission Electron 
Microscopy and Diffractometry of Materials, B. 
Fultz and J. Howe, Springer-Verlag Berlin 2002. 
Chapter 9)
? Introduction ? We have briefly introduced 
Patterson function in part 9. Here we will 
pick up the subject again and give a more 
detailed look at its application. ? The 
Patterson function is a function in real space 
with variable r. One could use the 
Patterson function to explain diffraction 
phenomena involving displacement of atoms 
off periodic positions due to temperature or 
atomic size. ? diffuse scattering from 
chemical disorder, the radial distribution 
function to describe amorphous materials, 
small angle scattering phenomena. ? Phase 
factor instead of 
Definition of reciprocal lattice 
 2 ? In the Fourier transform, the prefactors of 
 are neglected when the absolute 
value of the diffraction intensity in 
unimportant.
? The Patterson function ? Atom centers at 
Points in Space Assuming there are N 
scatterers, each is a point, located at 
rj. The total diffracted waves is
 It is convenient to consider a distribution of 
scatterers, f(r), with a continuous variable r.
 f(r) is zero over most of the space, but at atom 
centers such as , is a Dirac 
delta function times a constant 
 3 Property of the Dirac delta function
 Later, the actual shape of the atomic form 
factor, , will be included.
 ? Definition of the Patterson function 
Patterson function is defined as
 It is a little bit different from convolution 
and called autoconvolution (the function is 
not inverted).
 Convolution
 Autocorrelation 
 4?
 Fourier transform of the Patterson function is 
the diffracted intensity in kinematical theorem.
 Define , and change variables 
 .
 Inverse transform 
 5 The Fourier transform of the scattering 
factor distribution, f(r), is the diffracted 
wave, ?(?k).
and
i.e.
 ? Properties of Patterson function 
  Peaks in Patterson function is 
broader than those in the scattering 
 factor distribution function (f ). 
 Periodicity for the figure on the 
right is a. 
 6  Patterson function has a higher symmetry 
than the f(r). Primary peaks occur when 
shift equal to the periodicity. For the 
above case, secondary peaks occurs at ?0.3a. 
  Friedela law diffraction experiments cannot 
 distinguish between an atom arrangement 
and the atom arrangement when it is 
inverted. (phase problem) For simple 
structure, knowing the Patterson function 
presents no difficulties in studies of simple 
crystals of pure elements. For complex 
crystals, experimental techniques has been 
developed to alter the scattering strengths 
of different atoms isomorphous substitutions 
 of atoms at known sites of the unit cell, 
isotopic substitutions in the case of 
neutron diffraction, altering the atomic 
form factor by choosing different X-ray 
wavelength (anomalous scattering) 
 7 ? Perfect Crystals It is much easier to 
handle f(r) The scattering factor 
distribution for an entire crystal, f(r) is 
the convolution of the form factor of one atom 
 with a sum of delta functions
 The Patterson function of a one dimensional 
perfect crystal, 
 8 Convolutions are commutative and 
associative ?
 Use a shape function RN(x) so that the 
summation of ?- function can be extended to 
an infinite series.
? 
 9 Now there is the function RN(x). If the 
shift of the two chains is larger than 2N, 
they do not overlap anywhere
N  9
-3a
2a
4a
0
-a
a
3a
-2a
-4a
shift 8a
-3a
2a
4a
0
-a
a
3a
-2a
-4a
a triangle of twice the total width
-a
-3a
-7a
-5a
-9a
2a
4a
0
6a
8a
a
3a
-2a
-4a
5a
7a
9a
-6a
-8a 
 10 Or one can consider the convolution of two 
identical rectanglar shape functions. ? a 
triangle shape function of twice the total 
width.
 The Fourier transform of P0(x) ? I(?k)
 Using convolution theorem two 
convolutions and one multiplication after 
Fourier transform ? two multiplications and 
one convolution 
 11 If ?ka ? 2?. The sum will be zero. The sum 
will have a nonzero value when 
 and each term is 1. 
N number of terms in the sum
1 D reciprocal lattice
 Another term is 
shape factor intensity
In 1D this function is
F.T. 
 12A familiar result in a new form.  ? -function ? 
center of Bragg peaks  These peaks are broadened 
by convolution with the shape factor intensity 
(depends on d?k, deviation parameter  Bragg 
peak of Large ?k are attenuated by the atomic 
form factor intensity.
For convenience, assume that the crystal is 
infinite in length ? shape factor offers no 
broadening 
 13? Patterson Functions for homogeneous disorder 
and atomic displacement diffuse scattering 
 ? Deviation from periodicity In many cases of 
interest, f(r) can be expressed as
Deviation function
Perfect periodic function provide sharp Bragg 
peaks
Look at the second term
Mean value for deviation is zero 
 14The same argument for the third term ? 0
1st term Patterson function from the average 
crystal, 2nd term Patterson function from the 
deviation crystal.
Sharp diffraction peaks from the average crystal
The second term is often a broad, diffuse 
intensity! Two important source of ?f(r) 
(1)atomic displacement disorder and (2) chemical 
disorder. 
 15? Uncorrelated Displacements Types of 
displacement (1) atomic size differences in an 
 alloy ? static displacement, (2) thermal 
vibrations ? dynamic displacement Consider 
a simple type of displacement disorder 
each atom has a small, random shift, ?, off 
its site of a periodic lattice
 Consider the overlap of the atom center 
distribution with itself after a shift of 
 16 No correlation in ? ? probability of overlap of 
two atom centers is the same for all 
shift except n  0
 When n  0, perfect overlap at ?  0, at ? ? 0 
no overlap
 The same number of atom- atom overlap 
 17constant deviation
FPdevs1(x) increasingly dominates over 
FPdevs2(x) at larger k.
 The diffuse scattering increases with ?k ! 
 18? Correlated Displacements Atomic size effects 
 Different atoms have different sizes. Atoms 
around the big atoms are pushed away from the 
 lattice site and atoms further away are 
gradually relaxed back to lattice site. The 
peak shapes (P(x)) are skewed to larger x. 
? Diffracted intensity is therefore shifted 
from the higher ?k sides of the peaks to the 
lower ?k sides. Overall effect causes an 
asymmetry in the shape of the Bragg peaks.
a big atoms locate 
 19? Diffuse Scattering from chemical disorder ? 
Randomness  uncorrelated chemical disorder 
Assume a statistically-random occupancy of A- 
and B-atoms on each site in the alloy. 
 f(x) comprises delta functions each 
weighted by the scattering strength of the 
individual atom. Concentration of A-atoms 
cA Concentration of B-atoms cB. 
Assume cA gt cB ? 
Deviations in the scattering factor distribution, 
?f(x). The average value of ?f is zero 
The Patterson function for the deviation crystals 
is largest for a zero shift. At shifts equal to 
lattice vectors, the positive and negative peaks 
overlap randomly. 
 20When the product is summed over x?.
More positive than negative ones, but most of the 
positive ones are small. The Patterson function 
of the deviation crystal is zero, except for 
shifts of zero.
Lets calculate Pdevs(0) cAN peaks of cBN peaks 
of 
cA
cB 
 21Just like the case of perfect crystal
Total diffracted intensity
The diffuse scattering part is also called Laue 
monotonic scattering. The intensity of the 
diffuse scattering is the difference between the 
total intensity from all atoms and the intensity 
 in the Bragg peaks 
 22 ? Short range order (SRO) parameter 
Assume a binary alloy (A-B), total number of 
atoms is N, concentration of A-atoms is cA, 
concentration of B-atoms is cB, the 
scattering factor for A atom at crystal site m 
is fA(m) the scattering factor for B atom 
at crystal site m is fB(m). Short 
range order describe correlations between 
neighboring pairs of atoms, typically separated 
by a few atomic distance. 
 23 Stronger correlations at shorter distance and 
weaker correlations at larger pair separations 
(n). When atoms are separated by very large 
distances, we should expect no correlation. 
Mathematically, two atoms at sites m and m?, 
separated by a large nth neighbor distance 
 , the lack of correlation is a 
statement of statistical independence
 The relation will be 
used for no correlation. 
 convolution is evaluated at the interatomic 
separation rn.
 Define the two conditional pair probability 
 24 In the limiting case of complete disorder and 
statistical independence between atom positions
 Define Warren-Cowley SRO parameters, ?(n)
? total 0-2
0-1
0-1
 If an alloy is random, 
 For alloy with chemical order 
for at least some value of n
at those n
 For alloy with clustering tendency 
 25 ? Patterson function for chemical SRO 
The Patterson function for the average crystal
The Patterson function for the entire crystal 
consists of two terms. (1) correlations around 
A-atoms there are NcA with scattering 
strength fA. About these A-atoms, the 
probability of finding a B-atom at a distance n 
is and that of finding an 
A-atom at a distance n is (2) similarly 
for the second term which finds the 
correlations around B-atoms. 
 26Deviation
Look at
Look at 
 27Look at
Look at
The terms having 
 28The terms having
True random
Just like previous calculation
SRO diffuse intensity 
 29For chemically centro-symmetric alloys 
sum vanished
In a random solid solution,
? Amorphous Materials ? One dimensional 
model Amorphous materials usually contains 
two or more elements  position 
correlations of atoms, local chemical 
correlations. Consider one dimensional amorphous 
 materials with single element ? provide 
qualitative features of the diffracted 
intensity arising from position disorder 
only. The average separation between atoms 
is a. 
 30The separation between each adjacent atom is a 
random Variable (statistically independent) with 
a Gaussian distribution. Atoms separated by 
larger distances have a greater uncertainty in 
their separation. The scattering factor 
distribution is
For a perfect crystal
The diffracted intensity is
First calculate the convolution 
 31The probability distribution for the separation 
of the first -nearest-neighbor (1nn) planes is
A Gaussian distribution centered at a with a 
width of ?
a
The probability distribution for the separation 
of the second-nearest-neighbor (2nn) planes the 
probability distribution of a sum of independent 
random variables is the convolution of 
the probability distributions of the variable.
2a
centered at 2a 
 32Similarly,
centered at 3a
.
Next get the Fourier transform of the above 
function 
 331
The Fourier transform of a Gaussian function is 
a Gaussian function 
 34Neglecting the common prefector
x100
? ? ? perfect lattice 
 35 ? Radial Distribution Function Consider 
the reference structure is not a periodic solid, 
 but a homogeneous distribution of atoms with 
an average density ?0. All density 
distributions are described as sums of the 
constant density plus spatial variations in 
density. The scattering factor distribution
For a perfect crystal one can use ?0  0 and 
??(r) is a set of delta functions at lattice 
sites. Here ?0 is the average bulk density, so 
that
In form of integral 
 36The first term is needed to account for the 
perfect overlap at zero shift . The 
system have N scatterers and a volume V. ? ?0V  
N.
 V  N/?0.
 0
Define a new function R(r)
A correlation function of the density 
heterogeneities in the material 
 37The diffracted intensity is
The first term is a structureless background in 
the diffraction pattern (from the sharp 
self-correlation). The second term the forward 
scattering from a large homogeneous objects 
(forward direction only).
Assume the density-density correlations in the 
material are spatially isotropic, i.e. R(r) 
depends on r only.
Look at part 8
This equation can be used to obtain R(r). 
 38First identify the contribution to the 
measured diffraction data from the term ?(?k), 
located near the transmitted beam, and we strip 
this component from the data. Define normalized 
intensity ( )
To get R(r), multiply both sides of the above 
equation by and integrate 
over
A forward Fourier transform of R(r)r into 
k-space followed by a back transform into real 
space, nonzero only when 
 39?/2 coming from the normalization factor of the 
two sine transform