Crystallographic Calculations

Juan Rodrguez-Carvajal
Institut Laue-Langevin
Diffraction Group

The Mathematics of
Crystallography
Linear algebra and some
analysis is all you need for
doing crystallographic
calculations

Content of the talk

1: Geometry of crystals: lattices,
symmetry, metrics, crystal structure
2: Scattering theory: real crystals, powders
3: Powder diffraction: Least squares
illustrated by the Rietveld method
4: Solving structures using global optimisation
5: Magnetic structures

Summary of symmetry: Translation groups

Translation groups: symmetry groups leaving invariant an
infinite lattice of points in the R3 space: T= {t1,t2, tn }
Primitive Basis of the direct lattice: {a1, a2, a3}
A general translation can be written as: t=n1 a1 + n2 a2 + n3 a3
The corresponding operator is: O(t)= O(a1)n O(a2)n O(a3)n .
1

When acting on a vector position, the translation operator gives:

r = O(t) r = r + t
The order of the group T is infinite, but if we consider a finite
lattice having N=N1 N2 N3 primitive unit cells and periodic
boundary conditions {O(ai)Ni+1 =O(ai)} then: |T| = N
The translation group is Abelian (commutative)

Summary of symmetry: Space groups

Space groups: symmetry groups leaving invariant an
infinite crystal in the R3 space: G= {g1,g2, gn }
A general element of G can be written as: g={h|t}
Where h is the rotational part and t is the translation part
(t is, in general, composed of a lattice translation Tn plus an
intrinsic fractional translation th associated to h)
When acting on a vector position, the g operator gives:
r = g r = h r + t
The composition law is:
g1g2={h1|t1} {h2|t2} ={h1h2|h1t2+t1}
And the inverse is:
g-1 ={h|t}-1 ={h-1|-h-1t}

Summary of symmetry: Space groups

A space group is normally referred to a basis Bd={a, b, c}
that may be non-primitive
A crystal is described by the unit cell, the space group and
the list of atoms in the asymmetric unit.
Crystal:
Bd = {a , b , c} = {a1 , a 2 , a 3 }
a i a* j = ij

a *i =

Br = {a*, b*, c*} = {b1 , b 2 , b 3 } = {a1* , a *2 , a *3 }

a j ak

(i , j , k cyclic permutation )
V
Space group symbol : G = Lpqr = { g1 , g 2 ,...} = 1 + g 2 + g 3 + ... + g n
Atoms : ( Label , r j ), j = 1, 2,...na
For each site j there is a set of m(j) symmetrically equivalent atoms labelled with
the index s {s=1,2, m(j)}: rjs=gsrj1. If the atom is in a special position m(j) < n
The rotational part of symmetry operators are integers if we use the basis Bd

Summary of symmetry: Space groups

The position of an arbitrary atom in the crystal can be deduced from the
information given above. And using the following formula:
Atom positions in a crystal:
R ljs = R l + r js = R l + {h | t h }s r j = l1a+ l 2 b+ l3c + {h | t h }s ( x j a+ y j b+ z j c )
R ljs = (a

R ljs = (a

l1

c ) l 2 + (a
l
3

h11s
s
c ) h21
h31s

l1 + t1s h11s

s
c ) l 2 + t 2s + h21
s

s
l3 + t 3 h31

h12s
s
h22

h32s

h12s
s
h22

h32s

h13s x j

s
h23
y j + (a
h33s z j

t1s

c ) t 3s
t 3s

h13s x j

s
h23
y j
h33s z j

l : index of a direct lattice point (origin of an arbitrary unit cell)

j : index for a Wyckoff site (orbit)
s: index of a sublattice of the j site

For a primitive basis the indices li are integers. For a centred cell they can also be rational
numbers. In general, the fractional coordinates giving the atom positions in the asymmetric
unit are real numbers.

Direct and reciprocal metrics

Scalar products of vectors referred to crystallographic bases
Direct space:
u.v=

v j ai . a j =

i, j

v j g i j = ( u1

i, j

u2

a.a

u 3 ) b.a
c.a

a.b
b.b
c.b

a.c v1

b.c v2 = u T G v
c.c v3

Reciprocal space:
a*.a*

h . k = hi k j a . a = hi k j g = ( h1 h2 h3 ) b*.a*
i, j
i, j
c*.a*

G -1 =G*,
det(G ) = V 2 ,
det(G*) = V *2
*
i

*
j

*
ij

a*.b*
b*.b*
c*.b*

a*.c* k1

b*.c* k 2 = h T G* k
c*.c* k 3

The symmetry operators conserve the scalar products

so we have the following relations
u ' =Ru ,

v ' =Rv

u '.v ' = u . v

u '.v ' = Ru . Rv = u T R T G Rv = u T G v

R T G R=G

det( R T ) det(G ) det( R ) = det(G ) det( R ) = 1

Direct and reciprocal metrics: basis changes

Scalar product of vectors referred to both direct and reciprocal space:
r1

h3 ) r2
i, j
i, j
i
r
3
A vector/tensor is an intrinsic quantity that may be referred to whatever basis

hi rj a i . a* j =

h.r=

V=( a

hi r j i j = hi ri = ( h1

v1

c ) v2 = (e1
v
3

vc1

e 3 ) vc 2 = (a*

vc 3

e2

h2

v1*

c*) v2*
v3*

b*

Change of basis vectors:

a ' m11

A ' = b ' = m 21
c ' m31

m13 a

m23 b = M A
m33 c

m12
m 22
m32

Change of components of a vector

(a

v1

c ) v2 = a '
v
3

v1'

c v2' A T V=A 'T V ' = A T M T V ' V ' =M 1 T V
v3'

'

'

Direct and reciprocal metrics: basis changes

Relation between the components of a vector in direct and reciprocal space:

ai =

v g
i

v a

* *
j j

ik

=v

*
k

V =GV

a i .a k = v *j a *j .a k = v *j jk = vk*
j

V=G V

For the basis vector the transformation is with the inverse-transposed

A * =G * A

A=GA * (G and G * are symmetric and G * =G -1 )

Cartesian basis (e1 along a, e 2 in the a-b plane, e 3 completing the right handed frame):
a
e1


a i = mij e j A= b = M e 2 = M E
j
c
e

3
0
0
1/ a
0
a

M= b cos
b sin
0 M -1 = cos /( a sin ) 1 /(b sin )
c cos c sin cos * 1 / c *

a * cos *
b * cos *

T
G = MM

0
c *

Elements of Magnetic Groups

The group R = {1, 1} is called the time/spin reversal group
Magnetic point and space groups can be obtained as outer
direct product of point and space groups with R. The operators
of a magnetic group are called primed if they result from the
combination of the crystallographic operator with the unit
element 1 of R. Those combined with 1 are called unprimed
A general operator g = {h|th+n} acting on an atom rj in the cell at the origin
(zero-cell) having a magnetic moment mj is transformed as follows:

r j ' = g
r j = {h | t h + n}r j = hr j + t h + n = ri + a gj
m j ' = g
m j = det( h) h m j

The signature is = 1 for unprimed elements

and = 1 for primed elements

Crystallographic Magnetic (Shubnikov) Groups

Whatever crystallographic magnetic group, M, can be obtained as a
subgroup of the outer direct product of R by the crystallographic group G:
M GR.
The group G is always a magnetic group (colourless)
Paramagnetic (grey) groups are of the form: P=G+G1.
Nontrivial groups (black-white groups)
Constructive lemma: the magnetic groups derived from the
crystallographic group G can be constructed considering the index 2
subgroups H of G as constituting the unprimed elements and the rest of
operators, G H, those that are multiplied by the time reversal operator.
The magnetic group is then related to the subgroup H G (of index 2)
by the expression:
M = H + (G H) 1.
This is valid for all kind of groups: point groups, translation groups and
space groups. This is a consequence of primed x primed = unprimed

Shubnikov Groups: Magnetic Space Groups

The same procedure points groups and Bravais lattices can be applied to
space groups. One obtains a total of 1651 types of Shubnikov groups.

230 are of the form M0=G (monochrome groups),

230 of the form P=G+G1 (paramagnetic or grey groups)
1191 of the form M= H + (G H)1 (black-white,BW, groups).
Among the BW group there are 674 in which the subgroup
H G is an equi-translation group: H has the same translation
group as G (first kind, BW1).
The rest of black-white groups, 517, are equi-class group
(second kind, BW2). In this last family the translation subgroup
contains anti-translations (pure translations associated with
the spin reversal operator).

Content of the talk

1: Geometry of crystals: lattices,
symmetry, metrics, crystal structure
2: Scattering theory: real crystals, powders
3: Powder diffraction: Least squares
illustrated by the Rietveld method
4: Solving structures using global optimisation
5: Magnetic structures

Kinematics Scattering Theory

Q= kF - kI

kF=2/ uF

kI=2/ uI

In scattering experiments, the incident particle (neutron, electron, photon...)

experiences a change in its momentum and energy.
(h/2)Q =(h/2)(kF-kI)= h s

h=EF-EI

In the following we shall be concerned with elastic scattering (h=0) for

which kF=kI= 2/ and Q= Q = (4/) sin , being half the
scattering angle.
crystallographic scattering vector: s = Q/2.

Scattering expressions for arbitrary objects

In the kinematics theory (first Born approximation), the amplitude of the
wave scattered by an object is the Fourier transform (FT) of its
scattering density (SD) (r) measured in cm-2. Any object can be
considered as constituted by atoms of SD aj(r) centred at positions Rj; the
SD and the corresponding scattered amplitude and intensity can be written
Scattering density : (r ) = aj (r - R j )
as:
j

A(s ) = FT [ (r ) ]

Scattering amplitude :

A (s ) = F T [ (r ) ] =
A (s ) =

e
j

2 i sR

I (s ) = A (s ) A * (s ) =

(r ) e
aj

(u ) e

2 i sr

2 i su

d 3r =

aj

( r - R j )e 2 isr d 3 r

d u = f j (s ) e
3

2 i sR

f i ( s ) f j* ( s ) e

2 is ( R i R j )

Scattered Intensity of a Finite Crystal

Shape function: g(r) is defined as g(r)=1 for r inside the
crystal, and g(r)=0 for r outside G(s)=FT{g(r)}.
G2(s)=FT{g(r)g(-r)} is Fourier transform of the autocorrelation function:

V A S (r ) = g (u ) g (r + u ) d 3u
AS(r) is the fraction of the total volume shared in common between

the object and its "ghost" displaced by the vector r. AS(0)=1 and
decreases as r increases. G2(s)=V FT{AS(r)}

I (s )

N
F ( s ) F * (s ) G 2 ( s H )
VVc
H

Size effects in reciprocal space

VA S (r ) = g (u ) g (r + u ) d 3u

r
G 2 (s ) = FT { g (r ) g ( r )} = V FT { A S (r )}

(s H ) G H2 ( s ) = G 2 ( s )

G determines the shape of the

peak profile and it is the same
for all reciprocal nodes
Width of peak increases in 2
space
-1
as (cos
) . Except for spheres, the
width depends on the direction in
reciprocal space

Scattered Intensity of a Real Crystal

Defects: average lattice + structure factor depending on unit cell

I (s ) ( Fm Fm*+ n ) exp{2 i s R n }
n

Fm = Fm(s) is the structure factor of the cell m.

pn(Rn,s)=Fm F*m+n is independent of m, but depends on s.
The number of terms in the inner sum is given by VAS(Rn)/Vc and the
equation can be transformed to:
V
Vc

A S (R n ) F m F m* + n ex p {2 i s R n }

= N

A S (R n ) p n ( R n , s ) e x p {2 i s R n }

I (s ) =

Real Crystals: diffuse scattering

Average structure factor: F= F =1/N mFm
local fluctuation of the structure factor: n=F-Fn
pn=F2+m*m+n = F2+n,
The intensity formula can be transformed to:
I(s) =

IBragg

IDiffuse

N F2 nAS(Rn) exp {2i s Rn} + N nAS(Rn) n exp{2i s Rn} =

N F2/(VVc) HG2(s-H)

N n n exp{2i s Rn}

In the last expression, we have made the approximation AS(Rn)=1 because n

decreases with n much faster than AS(Rn).

Strained Crystals
A crystal is said to be strained if the structure factor
of the cell m can be written in the form:
1

with F = N Fn
n

D
A ( R n , s ) = exp{2 i s (u m u m + n )}

Fm = F exp{2 i s u m },

Defining the quantity:

I (s ) = NF 2 A S (R n ) A D ( R n , s ) exp{2 i s R n }
n

If s=H+s the scattered intensity around a Bragg peak is given by:

I H ( s ) = NFH2 A S (R n ) AHD ( R n ) exp{2 i s R n } FH2 f H ( s )

n

size strain

Strain effects in reciprocal space

The interpretation of the peak shape in the
case of strains is not so straightforward as
that of size because there are many possible
physical origins for strain broadening:
dislocations, interstitials, vacancies,
compositional fluctuations, etc.
The important point is that the peak shape
and width depend on both the particular
reflection and the direction in reciprocal
space.

Width of peaks increases in 2

space as tan
.

Profile for the Powder Case

For a powder (after spherical averaging), the vector Rn is projected
out along the scattering vector, s is replaced by its module s and
the total profile is a superposition of the profiles of the set of
equivalent nodes and the distribution function for size and strain
effects:

I { H } ( s ) FH2 f H ( s )
{H}

Usual notation in literature:

f hkl ( s ) { AnS AnD exp(2 i s nd hkl )}

n

Content of the talk

1: Geometry of crystals: lattices,
symmetry, metrics, crystal structure
2: Scattering theory: real crystals, powders
3: Powder diffraction: Least squares
illustrated by the Rietveld method
4: Solving structures using global optimisation
5: Magnetic structures

Experimental powder pattern

A powder diffraction pattern can be recorded in numerical form
for a discrete set of scattering angles, times of flight or energies.
We will refer to this scattering variable as : T.
The experimental powder diffraction pattern is usually given as
three arrays :

{Ti , y i , i }i =1,2,... n

The profile can be modelled using the calculated counts: yci at

the ith step by summing the contribution from neighbouring
Bragg reflections plus the background.

Powder diffraction profile:

scattering variable T: 2
, TOF, Energy
Bragg position Th

yi
zero

yi-yci
Position i: Ti

The calculated profile of powder

diffraction patterns

y c i = I h (Ti Th ) + bi
{h}

Ih = Ih ( I )
= ( xhi , P )

bi = bi ( B )

Contains structural information:

atom positions, magnetic moments, etc
Contains micro-structural information:
instr. resolution, defects, crystallite size
Background: noise, incoherent scattering
diffuse scattering, ...

The calculated profile of powder

diffraction patterns

y c i = I h (Ti Th ) + bi
{h}

The symbol {h} means that the sum is extended only to those
reflections contributing to the channel i .
This should be taken into account (resolution function of the
diffractometer and sample broadening) before doing the actual
calculation of the profile intensity.
This is the reason why some Rietveld programs are run in two
steps

Several phases ( = 1,n) contributing

to the diffraction pattern

y ci = s I ,h (Ti T ,h ) + bi

{ h }

Several phases ( = 1, n) contributing

to several (p=1, np) diffraction patterns

y cip = sp I p,h p (Ti T ,h ) + bip

{ h }

y c i = I h (Ti Th ) + bi
{h}

I h = S {L p O A C F 2 }

h
Integrated intensities are proportional to
the square of the structure factor F. The
factors are:
Scale Factor (S), Lorentz-polarization
(Lp), preferred orientation (O),
absorption (A), other corrections (C)

The Structure Factor contains

the structural parameters
(isotropic case)
F (h ) =

O
j =1

f j ( h ) T j exp 2 i h {S t }s r j
s

rj = (x j, y j, z j )

( j = 1, 2, ...n )

T j = ex p ( B j

sin 2

Structural Parameters
(simplest case)
rj = (x j, y j, z j )

Oj
Bj

Atom positions (up to 3n

parameters)
Occupation factors (up to n-1
parameters)
Isotropic displacement
(temperature) factors (up to n
parameters)

Structural Parameters
(complex cases)
As in the simplest case plus additional (or alternative)
parameters:
Anisotropic temperature (displacement) factors
Anharmonic temperature factors
Special form-factors (Symmetry adapted spherical
harmonics ), TLS for rigid molecules, etc.
Magnetic moments, coefficients of Fourier components of
magnetic moments , basis functions, etc.

The Structure Factor in

complex cases
F (h ) =

O
j =1

f j ( h ) T j g j (h s ) exp 2 i h {S t }s r j
s

h
h


h s = k = S sT k
l
l
s

g j (h s )

(s

= 1, 2, ... N G

Complex form factor of object j

Anisotropic DPs
Anharmonic DPs

The Rietveld Method

The Rietveld Method consist of refining a crystal
(and/or magnetic) structure by minimising the
weighted squared difference between the observed
and the calculated pattern against the parameter
vector:
n

= wi { yi y ci ( )}
2

i =1

wi =
i2:

i2

is the variance of the "observation" yi

Least squares: Gauss-Newton (1)

Minimum necessary condition:

2
=0

A Taylor expansion of yic ( ) around 0 allows the application of an

iterative process. The shifts to be applied to the parameters at each cycle
for improving 2 are obtained by solving a linear system of equations
(normal equations)

A 0 = b
Akl = wi
i

yic ( 0 ) yic ( 0 )
k
l

bk = wi ( yi yic )
i

yic ( 0 )
k

Least squares: Gauss-Newton (2)

The shifts of the parameters obtained by solving the normal
equations are added to the starting parameters giving rise to
a new set

1 = 0 +

The new parameters are considered as the starting ones in the next
cycle and the process is repeated until a convergence criterion is
satisfied. The variances of the adjusted parameters are calculated
by the expression:

2 ( k ) = ( A 1 ) kk 2
2

2
N - P+C

Least squares: a local

optimisation method
The least squares procedure provides (when it converges) the
value of the parameters constituting the local minimum closest
to the starting point
A set of good starting values for all parameters is needed
If the initial model is bad for some reasons the LSQ procedure
will not converge, it may diverge.

Content of the talk

1: Geometry of crystals: lattices,
symmetry, metrics, crystal structure
2: Scattering theory: real crystals, powders
3: Powder diffraction: Least squares
illustrated by the Rietveld method
4: Solving structures using global optimisation
5: Magnetic structures

Extracting integrated intensities

With single crystal the integration of individual spots allows to get
the moduli of structure factors, so the solution of the crystal
structure can be attempted using direct methods or any other
technique.
Powder diffraction suffers from peak overlap, so the extraction of
the integrated intensities is a previous step in order to solve the
crystal structure. The quality of the data is intrinsically lower, so
that traditional methods may fail.
(Situation similar in some sense to macromolecular crystallography)

The profile matching method

(Le Bail Fits)

y c i = I h (Ti Th ) + bi
{h}

When the resolution function of the diffractometer and the cell

parameters are known approximately, one can try to get the value
of the integrated intensities from the powder diffraction pattern.
Problem: If one uses least squares to refine the values of Ih
(Pawley) one has to introduce strong constraints because the
problem is not well defined, sometimes negative intensities can be
obtained due to overlap.

The profile matching method

(Le Bail Fits)

yci =

(T i Th ) + bi

{h }

Provides observed
(Ti Th )( y obs ,i Bi ) integrates intensities for
' I h ( obs ) ' = I h

( y calc ,i Bi )
i
calculating Bragg R-factor

Starting from an arbitrary set of values for Ih and using the

Rietveld formula for calculating the Bragg R-factor, it is possible
to obtain the so called Ih(obs). These intensities can be taken as
new values for Ih and repeat iteratively the procedure up to no net
improvement of the whole profile is reached.

The profile matching method

(Le Bail Fits)
(Ti Th )( y obs ,i Bi )
' I h ( obs ) ' = I h

(
y
B
)
i

calc , i
i

The values of Ih are of low quality when the overlap of reflections

is important. This is the reason one cannot use these integrated
intensities directly for solving the structure.
Many methods of solving the structure from powders (direct
methods) use a special strategy for changing the partitioning of
overlapped reflections

Solving the structure: the phase

problem
Direct space methods:
Reverse Montecarlo,
Simulated Annealing,
Genetic algorithms,

Classically, crystal structure determination is considered as a

process to determine the phases of the structure factors

(r ) = Fh exp{ 2 ihr}
{h }

(r ) = Fh exp{ 2 i ( hr + h )}
{h }

For a centrosymmetric structure

h is 0 or 1/2

The knowledge of all phases for the measured structure factors

provides a density map from which the structure is derived
(chemically recognised).

The resolution of the phase problem is the goal of

the crystal structure determination methods
Direct Methods tackle the problem looking for
phase relations (tangent formula) between structure
factors of different reflections
Direct methods need a high number of reflections and
good resolution (powders)
Direct methods are generally very efficient

But
sometimes direct methods fail in
solving particular structures
or
cannot be applied because poor
data quality (low resolution)

Structure factor calculation

F (h ) =

f j (h ) T j

j =1

exp{2 i h {S t}
s

rj}

If the chemical composition and the space group

are known we have to determine just 3n variables

Fobs (h ) r

O
j =1

f j T j exp{2 i h {S t }s r j }
s

h r = (h, k , l )r
rj = (x j, y j, z j )

r = 1, 2, ... N
j = 1, 2, ...n

Direct space methods:

Look directly for atom positions explaining the
experimental data
Minimize a reliability factor with respect to the
configuration vector or chromosome

= x1 , y1 , z1 , x2 , y2 , z2 ,...xn , yn , zn
N

R( ) = c F 2 (hr ) F 2 (hr , )
r =1

obs

calc

The Simulated Annealing Algorithm

begin
Initialise (set to zero useful quantities, do preliminary calculations )
=1
do
do
Perturb the system:
old new, =E(
new)-E(
old)
if 0 then accept, else
if exp(-/
/) > random[0,1] then accept
if accept then Update (replace old by new)
until equilibrium is approached closely enough (Ncyc)
0.9)
T+1 = f(T) (decrease temperature, usually T+1 = q T, q
=+1
until stop criterion is true (maximum , convergence, low % accepted...)
end

Content of the talk

1: Geometry of crystals: lattices,
symmetry, metrics, crystal structure
2: Scattering theory: real crystals, powders
3: Powder diffraction: Least squares
illustrated by the Rietveld method
4: Solving structures using global optimisation
5: Magnetic structures

Formalism of propagation vectors

m ljs = S kjs exp{ 2 ikR l }

{k }
l : index of a direct lattice point (origin of an arbitrary unit cell)
j : index for a Wyckoff site (orbit)
s: index of a sublattice of the j site
Necessary condition for real moments mljs

S - k js = S k js

General expression of the Fourier coefficients (complex vectors) for

an arbitrary site (drop of js indices ) when k and k are not equivalent:

1
Sk = (Rk + iIk )exp{2 ik }
2
Only six parameters are independent. The writing above is convenient
when relations between the vectors R and I are established (e.g. when
|R|=|I|, or R . I =0)

Single propagation vector k=1/2 H (SpBZ)

m lj = S kj exp {2 i k R l } = S kj ( -1)

n (l )

{k}

REAL Fourier coefficients magnetic moments

The magnetic symmetry may also be described using
crystallographic magnetic space groups

Fourier coefficients of sinusoidal structures

- k interior of the Brillouin zone (IBZ) (pair
k, -k)
- Real Sk, or imaginary component in the
same direction as the real one

mlj = Skj exp(2 ikRl ) + S-kj exp(2 ikRl )

S kj =

1
m j u j exp ( 2 ikj )
2

mlj = m j u j cos 2 (kRl + kj )

Fourier coefficients of helical structures

- k interior of the Brillouin zone

- Real component of Sk perpendicular to the
imaginary component

S kj =

1
muj u j + imvj v j exp ( 2 ikj )
2

mlj = muj u j cos 2 (kRl + kj ) + mvj v j sin 2 (kRl + kj )

Magnetic neutron scattering

a M (Q ) =

Q (m Q )
1
re f ( Q ) m
= p f (Q ) m
2
2
Q

p=0.2696 10-12 cm

f ( Q ) = m ( r ) exp(i Q r ) d 3r
Q=Q e

Only the perpendicular

component of m to Q=2h
contributes to scattering
m

Diffraction pattern of magnetic structures

Portion of reciprocal space
Magnetic reflections
Nuclear reflections
h = H+k

Magnetic reflections: indexed by a

set of propagation vectors {k}
H is a reciprocal vector of the crystallographic structure
k is one of the propagation vectors of the magnetic structure
( k is reduced to the Brillouin zone)

Magnetic Bragg scattering

Intensity (non-polarised neutrons)

I h = N h N h* + M h M * h
Magnetic interaction vector

Mh = e M(h) e = M(h) e (e M(h))

h

Scattering
vector
e
=
h = H+k
h

The magnetic structure factor

M( h) = pOj f j ( h) Tj Skjs exp 2i ( H + k ) {S t}s rj

j =1
s

j : index running for all magnetic atom sites in the magnetic

asymmetric unit (j =1,n )
s : index running for all atoms of the orbit corresponding to the
magnetic site j (s=1, nj). Total number of atoms: N = nj

{S t}

Symmetry operators of the propagation vector group

Maximum number of parameters for a general

incommensurate structure: 6N (In practice 6N-1,
because a global phase factor is irrelevant)

The magnetic representation

The vectors {js} are formed by direct sums (juxtaposition) of normal 3D
vectors ujs. Applying a symmetry operator to the vector position and the
unit spin associated to the atom js along the -axis, changes the index js to
jq and reorient the spin according to the nature of the operator g={h|th} for
axial vectors.
grsj = hrsj + t h = rqj + a gsj ; gs (q, a gsj )
( guj s ) = det(h) h n (uj s )n = det(h) h n n , = det(h)h
n

O( g )kjs = kjq , s ( g ) kjq = e

q

j
2 i k a gs

det( h)h sj, gq kjq

Matrices of the magnetic representation

Mag kjq , s ( g ) = e

j
2 i k a gs

det(h)h qj, gs

The magnetic representation as direct product

of permutation and axial representations
An inspection to the explicit expression for the magnetic representation for
the propagation vector k, the Wyckoff position j, with sublattices indexed by
(s, q), shows that it may be considered as the direct product of the
permutation representation, of dimension pj pj and explicit matrices:

Perm Pqskj ( g ) = e

j
2 i k a gs

qj, gs

Permutation representation

by the axial (or in general vector) representation, of dimension 3,

constituted by the rotational part of the Gk operators multiplied by -1 when
the operator g={h|th} corresponds to an improper rotation.

Axial V ( g ) = det( h) h
Mag kjq , s ( g ) = e

j
2 i k a gs

Axial representation

det(h)h qj, gs Magnetic representation

Basis Functions of the Irreps of Gk

The magnetic representation, hereafter called M irrespective of the
indices, can be decomposed in irreducible representations of Gk.
We can calculate a priori the number of possible basis functions of the
Irreps of Gk describing the possible magnetic structures.
This number is equal to the number of times the representation is
contained in M times the dimension of . The projection operators
provide the explicit expression of the basis vectors of the Irreps of Gk
k ( j ) =

1
*[ ] ( g ) O( g ) kjs
n(G 0k ) gG 0k

k ( j ) =

1
exp(2 i k a gsj ) det(h)h sj, gq kjq
*[ ] ( g )
n(G 0k ) gG 0k
q

( = 1,...l )

Basis Functions of the Irreps of Gk

It is convenient to use, instead of the basis vectors for the whole set of
magnetic atoms in the primitive cell, the so called atomic components of
the basis vectors, which are normal 3D constant vectors attached to
individual atoms:

k ( j ) =

Sk ( js )

, s =1,... p j

The explicit expression for the atomic components of the basis functions is:
k

( js )

*
[ ]

(g) e

j
2 i k a gs

det(h)

j
s,g[q]

gG 0k

h1

h2
h
3

Fundamental hypothesis of Symmetry Analysis

The fundamental hypothesis of the Symmetry Analysis of magnetic
structures is that the Fourier coefficients of a magnetic structure are
linear combinations of the basis functions of the irreducible
representation of the propagation vector group Gk

S kjs = C n S kn

( js )

M( h) = pOj f j ( h) Tj Cn Skn ( js ) exp{2 i hsrj }

j =1

Conventional crystallographic approach

Alternatively, a more close to traditional crystallographic approaches can
be considered in some cases. The Fourier component k of the magnetic
moment of atom j1, that transforms to the atom js when the symmetry
operator gs={S|t}s of Gk is applied (rjs=gsrj1=Ssrj1+ts), is transformed as:

Skjs = M js Skj1exp {2 ikjs }

n

M ( h ) = p O j f j ( h ) T j
j =1

S
s

kjs

exp 2 i ( H + k ) {S t}s r j kj

The matrices Mjs and phases kjs can be deduced from the relations
between the Fourier coefficients and atomic basis functions. The matrices
Mjs correspond, in the case of commensurate magnetic structures, to the
rotational parts of the magnetic Shubnikov group acting on magnetic
moments.

Conventional crystallographic approach

S kjs = M js S kj1exp {2 ikjs }

The Niggli-Indenbom theorem is related to the constructive lemma
about the subgroups of index 2 of crystallographic groups.
The theorem states that: the magnetic space groups correspond to real
1D irreducible representations of conventional crystallographic space
groups. This assures that the above formula can always be satisfied for
commensurate magnetic structures.
The Shubnikov group can be constructed just considering that the
symmetry operators of characters equal to -1 are primed elements
whereas those of character equal to 1 are unprimed elements.

Form factor models in FullProf

Scalar magnetic structure factor
n

M ( h ) = p B O jT j f j ( h ) exp {2 i hr j }
j =1

The different models existing in FullProf correspond to different

ways of describing f(h)
First model: Simple spherical form factor (free par. Wl)

f (s) =

Wl jl ( s )

l = 0,2,4,6

jl ( s ) = s 2 ( Al exp{ al s 2 } + Bl exp{ bl s 2 } + C l exp{ cl s 2 } + Dl )

for

l = 2, 4, 6

j0 ( s ) = A0 exp{ a 0 s 2 } + B0 exp{ b0 s 2 } + C 0 exp{ c0 s 2 } + D0

Form factor models in FullProf

Second model: Multipolar density approach
4

m (r ) = Rldens ( r ) lm , p ylm , p (r0 )

l =0

m, p

f (h ) = i l jl (2 h )
l=0

m, p
l

ylm , p (h 0 )

m, p

jl ( s ) = Rldens ( r ) jl ( sr ) r 2 dr
0

dens
l

( r ) = Slater( nl , l ; r ) =

ln + 3
l

( nl + 2)!

r nl exp{ l r }

Form factor models in FullProf

Third model: Multipolar orbital approach

m (r ) = l (r )

l (r ) = Rl ( r ) Alm , p y lm , p (r0 )

m, p

m, p
l

complex numbers

The form factor adopts the following form (similar to multipolar )

f (h ) =

l ' = 0,2,...2 l

i l ' jl ' (2 h )

m, p
l'

ylm' , p (h 0 )

m, p

jl ' ( s ) = r 2 Rl2 ( r ) jl ' ( sr ) dr

0

Form factor models in FullProf

Within the Third model several options exist for treating the
radial part

m (r ) = l (r )

l (r ) = Rl ( r ) Alm , p y lm , p (r0 )
m, p

1: Use the tabulated values of jl ( s ) for the ions

2: Use a hydrogen-like R(r) of quantum number nl
3: Use a Slater-type function of the form:

R ( r ) = Slater( n , ; r ) =

(2 ) 2 n + 3 n
r exp{ r }
(2 n + 2)!