IONIC CRYSTALS

Definition of an ionic crystal composed of atoms A and B

Cohesion results from transfer of an electron from element A to element B, producing
+
!
closed-shell ions A and B which attract each other by Coulomb forces and repel
each other at short-range due to Pauli repulsion. A schematic picture of an ionic solid
is shown in Fig. 1.

Fig. 1. Schematic picture of an ionic solid with the NaCl type structure

Metal halides
LiF, LiCl, NaF, NaCl, NaBr, KF, KBr, AgCl, AgBr, AgI etc.
crystallize in the sodium chloride or cesium chloride structure.
Cesium chloride:

Simple cubic lattice of lattice of Cs ions with Cl ions on simple

cubic lattice displaced relative to the Cs lattice by 12 111
vector. Alternatively, a bcc lattice with Cs in cube corners and
Cl at cube centers (Fig. 2a).

Sodium chloride

A simple cubic lattice of alternate positive and negative ions.

Each species is forming a face-centered-cubic lattice and these
are displaced with respect to each other by 12 110 (Fig. 2b).

Fig. 2a. CsCl structure

Fig. 2b. NaCl structure

2

Metal oxides
Sodium chloride

MgO, CaO, SrO, BaO, MnO, FeO, CoO, NiO, TiO, VO, CrO,
ZnO, CdO

Spinel

MgAl2O4, FeAl2O4, ZnCr2O4, ZnFe2O4

Perovskite

BaTiO3 , SrTiO3 (see Fig. 3).

Corundum

Al2O3 i. e sapphire (hexagonally close-packed oxygen planes

with aluminum filling two-thirds of the available octahedral
sites).

Fig. 3 Structure (cubic symmetry) of SrTiO3

Electrostatic neutrality of ionic crystals
When studying defects in ionic crystal the block has to remain electrostatically
neutral i. e. no surplus of a positive or negative charge must occur.
Examples: Vacancies occur in pairs of vacancies at positive and negative ions.
Frenkel defect - vacancy + interstitial.

INTERACTIONS BETWEEN IONS

Interaction between ions in an ionic crystal can be divided into two principal parts:

Long-range Coulomb interaction

The energy associated with this interaction between ions can be written as

EC =

1 Z iZ j
"
2 i,j rij

(I1)

i! j

where Z i is the charge at ion i and rij is the separation between ions i and j. The
summation in (I1) converges because there are positive and negative ions but the
convergence is very slow. Summation methods, such as the Ewald method, described
below, have to be employed.
The charges associated with ions may be either full charges corresponding to isolated
ions or partial charges. In the former case the corresponding Z i is an integer multiple
of the electron charge and in the latter case it is a non-integer multiple of the electron
charge corresponding, for example, to the charge found in ab initio calculations for a
given ionic crystal. In both cases there is zero total charge within the repeat cell.

Short range interaction

These are most commonly represented by pair potentials of the Buckingham form

$ r ' D ij
! ij = Aij exp& " ) " 6
% #ij ( r

(I2)

The first term is the (Pauli) repulsion of ions at short separations described by the
Born-Mayer potential. The second term is the (weak) Van der Waals attraction and it
is often omitted in studies of ionic solids. The short range interactions can be
regarded as significant only for separations smaller or equal to that of the first nearest
neighbors. The parameters Aij , D ij and !ij can be determined, similarly as in the case
of any empirical pair potentials, by fitting equilibrium quantities, e.g. lattice
parameter, elastic moduli, cohesive energy, of the material that are determined
experimentally and/or by ab initio calculations.

Since both long-range Coulomb interactions and short-range repulsions are described
by pair potentials model the Cauchy relations between elastic moduli ( C 12 = C 44 for
cubic symmetry) will be satisfied in this scheme.

Rigid ion model

The ions are considered as possessing either full or partial charges and the total
energy of the system is composed of the Coulombic term (I1) and the term given by
the short-range pair potentials (I2). This determines the total energy of the solid as a
function of atomic positions and atomistic calculations (molecular statics, molecular
dynamics or Monte Carlo) can be carried out as in the case of other pair-potential
models.
However, the Coulombic term cannot be evaluated by a direct summation over ions
since the convergence would be extremely slow and Ewald type methods, discussed
below, must be employed.
Example: Potentials for Al2O3 discussed in the study of P. R. Kenway: J. American
Ceram. Soc. 77, 349, 1994.

Shell model
An important difference between ionic crystals and metals is that ionic crystals are
insulators and internal electric fields can be present in the vicinity of defects.
Therefore, defect energies may depend sensitively on the electronic polarization of
the lattice. This is particularly significant in the case of charged defects. Hence, the
empirical model needs to be compatible with dielectric properties of the materials
studied if these effects are to be taken into account. This can only be achieved if one
allows for polarization of ions. In this case we can evaluate, and/or fit, the low- and
high-frequency dielectric constants ! o and ! " , respectively, in the framework of the
model. (For a brief explanation of dielectric properties see Appendix).
The most successful empirical model through which the polarizability of ions has
been introduced into the atomic level analysis of ionic crystals, is the shell model,
first advanced by B. G. Dick and A. W. Overhauser (Phys. Rev. 112, 90, 1958). It is
shown schematically in Fig. 4.

Spring
Core

Shell

Fig. 4. The schematic picture of the shell model of an ion.

In the framework of this model ions can be polarized and each ion i, possessing a
charge Zi, consists of two components:
(i)

A core having the same mass as the ion i and a charge Zi

s

(ii) A 'shell' having no mass and a charge Zi .

c

The total charge of the ion Zi = Z i + Z i .

The core and the shell are coupled by a harmonic spring with the force constant ! i so
that the force acting between the core and the shell depends only on their relative
displacement, s, and no directional forces are involved. (In this sense the shell is
spherical).
Interactions involved in the shell model
(i) The Coulomb interactions: include shell-shell, shell-core and core-core
interactions between different ions.
(ii) Short range shell - shell repulsive interactions between different ions described by
potentials given by equation (I2). No short-range core - core interactions are usually
included.
(iii) Core - shell interactions inside a particular ion: Described by the harmonic spring
linking the core and the shell, characterized by a spring constant ! i .

Fitting of shell parameters Zi and ! i

Empirical parameters:

Low- and high-frequency dielectric constants ! o and ! "

! o the transverse optical frequency

The dipole moment induced at an ion, i, due to the displacement, s, of the shell is
s
equal to pi = Z i s . Its potential energy in an external electric field, E a , is
s
U pot = !p i E a = !Zi s Ea . At the same time the self-energy of the ion, arising due to

the harmonic spring linking the core and the shell, is Uself = 12 ! i s . Minimization of
6

the total energy, U pot + U self , with respect to s yields s =

Z si E a
!i

so that the

polarizability of the ion i, defined as ! i = p i / E a is

!i =

(Z si )2
"i

(I3)

The polarizabilities of the ions, ! j , are directly related to the two dielectric constants
and the transverse optical frequency. For example, in the case of metal helides of the
type NaCl with polarizabilities of the corresponding two elements ! 1 and ! 2
(Sangster, M. J. L., Schrder, U. and Arwood, A. M., J. Phys. C 11, 1523, 1978)

!1 + ! 2 =

$% + 2
3" $% &1
2
+
(Z & Z ( )
2
4# $% + 2 'o ($ 0 + 2)

(I4.1)

and for metal helides such as CaF2, SrCl2

!1 + 2!2 =

$% + 2
3" $ % & 1
2
+
(Z & Z ( )
2
4# $ % + 2 2' o ($ 0 + 2)

(I4.2)

! is the volume of the repeat cell, the reduced mass of the two particles1,
Z = Z1 = Z2 the ionic charge and Z' so called Szigetti charge (B. Szigetti, Proc. Roy.
Soc. London A 204, 51, 1950) given in this case by the relation
2

(Z ! ) =

9" 2o# % o & % '

4$

(%' + 2)2

(I4.3)

Analogous relations can also be used for oxides of the type MgO, CaO etc. (see, for
example, M. J. L. Sangster and A. M. Stoneham, Philos. Magazine 43, 597, 1981).
We now make the assumption that the free ion polarizabilities ! i are properties
that depend only on the ions and not on the crystal in which they are placed. By
analyzing a large number of ionic compounds that combine the same elements, the
values of ! i are determined by the least square fit. For example, for alkali helides we

1 $
! 1
+
& , where m1 and m 2 are the masses of the two ions.
" m1 m 2 %

=#

consider all combinations of Li, Na, K, Rb, Cs with F, Cl, Br, I, i. e. compounds: LiF,
LiCl, LiBr, LiI; NaF, NaCl, NaBr, NaI; KF, KCl, KBr, KI; RbF, RbCl, RbBr, RbI;
CsF, CsCl, CsBr, CsI. Thus we use experimental data, ! o , ! " and ! o , for twenty
compounds to determine polarizabilities ! i for eight elements. Using these
s

polarizabilities and equation (I3), the shell charges Zi and corresponding spring
constants ! i are then determined such as to fit with the shell model dielectric
constants ! o and ! " and the transverse optical frequency ! o ; for details see Sangster,
M. J. L., Schrder, U. and Arwood, A. M., J. Phys. C 11, 1523, 1978.
Frequently, but not always, only the negative ions (owing to the surplus of electrons)
are treated as polarizable while the positive ions are regarded as point charges; the
shell model is then applied only to negative ions.
Fitting of parameters of short-range potentials given by equation (I2).
Parameters Aij , D ij and !ij can be ascertained, similarly as in the case of empirical
pair potentials for metals, by fitting the experimentally determined equilibrium
properties of the material and/or results of DFT based calculations. For example,
potentials for oxides have been constructed by M. J. L. Sangster and A. M. Stoneham
(Phil. Magazine 43, 597, 1981). During this fitting the parameters of the shell,
determined from fitting the dielectric properties, are, of course, used.
Since this is still a pair potential model the Cauchy relations between elastic moduli
( C 12 = C 44 for cubic symmetry) are satisfied. A modification, the so called breathing
shell model, has been developed in which non-spherical deformation of the shell is
included (M. J. L. Sangster: J. Phys. Chem. Solids 34, 355, 1973; also C. R. A.
Catlow, I. D. Faux and M. J. Norgett, J. Phys. C 9, 419, 1976). In this model the
Cauchy relations need not be satisfied.

Atomistic calculations employing the shell model

In the framework of the shell model each ion possesses six degrees of freedom.
Three are associated with the position of the core and three with the position of the
shell. The energy of the system and corresponding interatomic forces are composed
from the following contributions:
Coulomb interaction between different ions
(a)
(b)
(c)

Core Core
Shell Shell
Core Shell
8

In general, if there are N ions in the system and each is composed of a core and a
shell, interactions between 2N different charges must be considered. Since Coulomb
interactions are very long range, evaluation of the corresponding interaction energies
requires a special treatment employing Ewald summations. This approach is
described in more details below.
Short-range interaction between shells
Interactions between shells associated with individual ions are described by potentials
of the type (I2) and if there are M different types of ions then there are, in general,
1 2 M(M + 1) such potentials. For example in a binary compound composed of ions
A and B there are three potentials ! AA ,! BB and ! AB .
No short-range interactions between the cores are usually considered since the cores
are screened by shells.
Core - shell interactions inside an ion
Described by harmonic spring with a spring constant ! i that links the core with its
shell.
All these interactions have to be included when evaluating the energy of the system
and forces on cores and shells. In molecular statics calculations the energy of the
system is then minimized with respect to the positions of both cores and shells of the
ions. In molecular dynamics calculations both cores and shells may move or only the
unit core plus shell move.

Treatment of Long-range Coulomb interactions

(See C. Kittel: Introduction to Solid State Physics)
We consider a periodic structure with s point charges, Z j , in the repeat cell and the
s

sum of the charges in every repeat cell is zero, i. e.

! Z j = 0.

Periodic boundary

j=1

conditions are naturally assumed. To compute the energy, E C , associated with the
Coulomb interaction of these charges we need to evaluate the potential, ! i , at each
charge site i in the repeat cell induced by all the other charges in the structure that
includes all the periodically repeating cells. The interaction energy of this system of
charges is then
1
(I5)
E C = " Z i! i
2 i
9

10

Ewald method for evaluation of ! i - Three-dimensionally periodic

structures
The electrostatic potential experienced by charge i in the presence of all the other
charges can, in principal, be calculated by direct summation of the Coulomb
potentials. However, convergence of such summation is very slow and Ewald
proposed the following trick in which the potential is decomposed into two parts:
(a) The potential, ! a , arising from Gaussian densities of charges centered at the sites
of point charges, each giving the same total charge as the corresponding point charge.
This is shown schematically in Fig. 5. For the charge at position n this density is2
% r$r 2(
1
n *
n
!a (r) = Z n 3 3 /2 exp ' $
(I6)
2
'
"#
" *
&
)
The width of the Gaussian, determined by ! , is chosen such as to assure fastest
possible convergence in the summation.
(b) The potential, ! b , arising from point charges and additional Gaussian
distributions of charges of opposite signs centered at the positions of the charges.
However, no charge is placed at the position where the potential is being evaluated, i.
e. position marked 0 in Fig. 5. This is shown schematically in Fig. 6. The Gaussian
densities are given by expressions analogous to (I6).

Gaussian densities
of charge

-1

-2
Positions of
point charges

Fig. 5. Gaussian densities of charge centered at positions of individual point charges

giving the same total charge as point charges. The position at which the potential is
calculated is marked 0.
#

The Gaussian density is normalized such that 4! $ "na (r)r 2 dr = Z n .

0

10

11

Gaussian densities
of charge

-2

-1

Positions of
point charges

Point charges

Fig. 6. Gaussian densities of charges centered at positions of individual point charges

giving the opposite total charge as do the point charges, together with point charges.
The position at which the potential is calculated is marked 0 and no charge is placed
at this position.
Potential ! a
We expand this potential into a Fourier series3

!a (r) =

#C

exp(iK " r)

(I7.1)

where K are the reciprocal lattice vectors of the periodic structure considered4.
s

Similarly, the charge density invoking this potential, !a = " " !a , where the
k

k,j

j=1

summation over j extends over one repeat cell and summation over k extends over
various repeat cells, can be expanded as
3

The potential possesses the periodicity of the structure so that

quantities calculated.

! a (r) = ! a (r + rn ) . The same applies to all other

3

4A

reciprocal vector K is generally given as K =

!j b
i

where j i are integers and

i=1

b 1 = 2!

a2 " a 3
a 3 " a1
a1 " a 2
, b 2 = 2!
, b 3 = 2!
a 1 # (a 2 " a 3 )
a 1 # (a 2 " a 3 )
a 1 # (a 2 " a 3 )

vectors of the reciprocal lattice.

11

are

the

basis

12

!a (r) =

#!

exp(iK " r)

(I7.2)

! a and !a are linked by the Poisson equation ! 2 " a = #4$% a . Inserting (I7.1) and
(I7.2) into the Poisson equation yields

"K C
2

exp(iK ! r) = 4#

"$

exp(iK ! r)

(I8)

and, therefore,

CK =

4!"K
K2

(I9)

By definition, the Fourier coefficient of the charge density !a is

!"K =

% " (r) exp(#iK $ r)dr

a

(I10)

repeat
cell

where ! is the volume of the repeat cell. Here the density !a (r) originates from the
charges within the repeat cell as well as from the charges in all other cells. Hence,
this integral is the same as when integrating the density originating in the repeat cell,
multiplied by exp(! iK " r) , over the whole space, i. e.
s

& $ " (r # r ) exp(#iK % r)dr

j
a

!"K =

(I11)

All space j=1

Inserting (I6) into (I11) yields

!"K =

1
# $3/ 2
3

1
#3$ 3/2

2
'
*
r % rj
)
Zj
exp %
% iK & r ,dr
2
)
,
#
j=1
All space
)(
,+
s
' /2
*
)
Z j exp(%iK & rj )
exp % 2 % iK & / ,d/
) #
,
j=1
All space
(
+
s

. .

(I12)

where the substitution r ! rj = " was made. The last integral can be evaluated with
the help of complex variables and we obtain
12

13

% $2K 2 (
!" K = S(K) exp ' #
4 *)
&

(I13.1)

where
s

S(K) =

# Z exp(!iK " r )
j

(I13.2)

j=1

is the structure factor in which charges Z j are the form factors.

After inserting (I13.1) into (I9), equation (I7.1) yields for the potential at the position
I

4"
!a (ri ) =
#

,
K

Si (K )
K2

& %2 K 2 )
exp ( $
4 +*
'

(I14)

where the structure factor Si (K) is evaluated such that the origin is taken at the
position i. Owing to the exponential dependence only a few shortest reciprocal lattice
vectors need to be included in this summation.

Potential ! b
The potential at the position i arising from a charge at a position rj , can be written as

%
1 1
Zj ' !
'r r
j
'& j

rj

$ " (r)dr ! $
b

rj

"b (r )
r

(
dr *
*
*)

(I15)

when taking the position i as the origin. The first term arises from the point charges,
the second term from the part of the Gaussian distribution lying inside the sphere of
radius rj and the third part from the part of the Gaussian distribution lying outside
this sphere. Substituting for !b (r) in (I15) the Gaussian function analogous to (I6)
and summing over all the charge positions, both in the repeat cell and in all other
cells, yields
& ri " rn )
Zn
(I16.1)
! b (ri ) = $
F(
+
n #i ri " rn
' % *
where
13

14

F(x) =

exp(" y )dy
$
!
2

(I16.2)

and the summation extends over all the charge positions both in the repeat cell at the
origin and all other cells. Obviously, F converges very rapidly to zero as x increases
and thus only a small number of cells neighboring the repeat cell at the origin need to
be included.
The potential at the position i is then ! i = ! a (ri ) + ! b (ri ) " ! 0 (ri ) where
%

!0 (ri ) =

&

4" r # ri $a (r # ri )

r # ri

d(r # ri ) =

2Zi
' "

is the contribution of the Gaussian density at the position i that was incorporated into
the case (a) and must be subtracted since only contributions of charges other than the
charge at i are to be included.
Hence, the potential at the site i is

4"
!i =
#

Si (K )
K2

& %2 K 2 ) 2Z i
exp ( $
$
+
4 +* % "
'

,
n -i

& ri $ rn )
F(
+
ri $ rn ' % *
Zn

(I17)

The parameter ! is arbitrary and ! i does not depend on its choice. However, the convergence of
the sums in (I17) does and the trick is to choose ! such that these sums converge rapidly5.

Ewald method for two dimensional periodicity

An analogous Ewald type summation of the Coulomb energies has been developed
(D. E. Parry, Surface Science 49, 433, 1975) when only two-dimensional periodicity
is present as, for example, in the case of interfaces. The summations are then done
plane by plane in the crystallographic planes parallel to the periodic plane, e. g. the

In an ideal ionic crystal the Coulomb energy can be written as E C = !

parameter and ! =

"
a

, where a is the lattice

#Z "
2
i

is the Madelung constant; the summation over i extends over all the

i =1

ions in the repeat cell.

14

15

interface (see also R. E. Watson, J. W. Davenport, M. L. Perlman and T. K. Sham,

Phys. Rev. B 24, 1791, 1981).
In the case of an interface or free surface the bloc is divided into two regions as
follows:
Region I: In this region full atomic relaxation is taking place and the repeat cell of
the planes parallel to the boundary is defined by the vectors b I1 and b I2 so that in the
region I any vector in a plane parallel to the boundary can be written as their linear
combination.
Region II: In this region atoms are not relaxed and they occupy ideal lattice
positions. This region may, however, be shifted as a rigid block. The repeat cell of
the planes parallel to the boundary is defined by the vectors b II1 and b II2 so that in the
region II any vector in a plane parallel to the boundary can be written as their linear
combination.
In general vectors b I1 and b I2 are not the same as vectors b II1 and b II2 but, for example,
in the case of a grain boundary with the smallest available repeat cell, the vectors
determining this cell will be the same in both the region I and the region II.
The Coulomb energy of a charge at the position j in the region I may be decomposed
into contributions arising from its interaction with charges in regions I and II,
respectively, i. e.
Zj Zk
Z j Zk
Zj Zk
(I18)
=
+
r
r
r
jk
jk
jk
k
k from I
k from II

According to Parry the contribution from the region I is

k from I

Z jZ k
rjk

4"Z j
AI

+ Zj

!Z!

exp % # K I
&'

k from I

!!

!,m k from I

KI

Zk

KI

2
4$ (
)*

1 # erf rjk # R

I
!m

cos % K I rjk ( #
&
)

2Z 2j $
"

(I19)

rjk # R Ilm

where K I is the reciprocal lattice vector related to planar lattice based on b I1 and b I2 ,
I
I
I
R !m = !b 1 + mb 2 ( ! and m are integers), AI the area of the repeat cell in region I and
the convergence parameter ! have the same meaning as in the previous case.

15

16

The contribution from the region II is

k from II

Z j Zk
rjk

2"Z j
A II

!Z!
k

k from II

K II (0

exp $ # K II (z j + z k )&
%
'
cos $%K II rjk &'
$
&
K II %1 # exp(# K II d)'

(I20)

where (rn ,z n ) are cylindrical coordinates of the position of charge n with the z axis
perpendicular do the interface, K II is the reciprocal lattice vector related to the planar
lattice based on b II1 and b II2 , AII the area of the repeat cell in region II and d is the
magnitude of the repeat vector in the z direction, perpendicular to the interface.

16

17

APPENDIX POLARIZATION IN IONIC CRYSTALS

Macroscopically the applied electric field, E a , induces polarization of the medium

which is described by the polarization vector P = !E a , where ! is the electric
susceptibility. The polarization vector is defined as the dipole moment per unit
volume. The dipole moment associated with a charge q j is p j = q j rj , where rj is the
position vector of the charge q j and the polarization vector is P = 1 ! " p j , where
j

the summation extends over all the dipole moments within the repeat cell and ! is
the volume of this cell.
The dielectric displacement vector D = E a + 4!P satisfies the same Poisson equation
in the dielectric medium as does E a in the vacuum and we define the dielectric
constant, , via the relation D = !E a ( ! = 1+ 4"# ).
If p j is the dipole moment associated with the ion j, then the polarizability of this
ion, ! j , is defined by the relation

p j = ! j E loc ( j)
where E loc ( j) is the local electric field at the position of ion j. This field is not
necessarily just the applied field E a but because of the effects of polarization of the
4!P " + 2
=
Ea.
medium it is given by the so-called Lorentz relation E loc = E a +
3
3
(See e.g. C. Kittel: Introduction to Solid State Physics).

17