210 U.V. Waghmare
capability to describe materials with strongly correlated electrons such as high T
c
superconductors. However, the DFT can be quite effectively used in simulations of
ferroelectric perovskite oxides .ABO
3
/ which are more like band-insulators with
unoccupied conduction bands constituted of the d-states of transition metal .B/.
Two flavors of DFT calculations have been extensively used in investigation of
ferroelectricity: (a) total energy calculations that involve solving electronic ground
state problem for a given potential [17], and (b) linear response (LR) calculations
that involve determination of the lowest order changes in ground state total energy
arising from small changes in the potential [18]. The former are the basic capa-
bilities of all DFT codes, and allow determination of the structure through energy
minimization and properties that correspond to the first derivative of total energy,
such as stress, forces on the atoms and electric polarization. The latter are typi-
cally available in relatively fewer codes, and allow determination of the properties
that correspond to second and third derivatives of total energy, such as phonons,
dielectric, elastic, piezoelectric and other compliances and Raman tensors. The to-
tal energy and its derivatives accessible through these DFT simulations are relevant
to the properties at T D 0 K (or at low temperatures). These can be used to de-
velop model Hamiltonians [2] of ferroelectrics that are suitable for Monte Carlo or
molecular dynamics simulations of the temperature dependence of their properties.
We comment on DFT-based determination of the properties that are fundamental
to understanding and applications of ferroelectric materials. Phonons: as phonons
are the eigenmodes of the dynamical matrix, which itself is second derivative of total
energy with respect to displacements of atoms with respect to equilibrium positions,
DFT-LR calculations provide an efficient method of determination of full phonon
dispersion of materials. For example, phonon dispersions of ABO
3
compounds in
the cubic structure exhibit a large number of unstable modes [19], which precisely
reveal which structural distortions can destabilize the cubic structure. Electric
polarization: this (first derivative of total energy with respect to electric field) is
the primary order parameter of ferroelectric transitions, whose nonzero value distin-
guishes the ferroelectric phase from the paraelectric one. While intuitively defined
as the electric dipole moment per unit volume, its estimation is rather tricky and
not possible from the knowledge of charge density of a periodic (crystal) material
alone! It was shown to be linked with an overall geometric phase of the manifold
of occupied electronic states by King-smith and Vandebilt [20, 21]. This formal-
ism has resulted in a commonly used Berry phase method to determine polarization
through a post-processing step after a total energy calculation. A formalism based
on non-abelian geometric phases [22] can be effectively used in breaking up the to-
tal polarization into different spatial regions or bonds, which can be quite insightful
into understanding of polarization of heterogeneous ferroelectrics such as super-
lattices. Born effective charge tensors: these give the strength of coupling of an
ion with electric field or IR radiation, and are the second derivatives of total energy
with respect to electric field .E/ and atomic displacement .u/: Z
i˛ˇ
D
@
2
E
tot
@E
˛
@u
iˇ
,
which can be most efficiently calculated using DFT-LR. Here, i is the atomic in-
dex and ˛; ˇ are Cartesian directions. It is relatively easier to determine the Born
charges from first-principles DFT than from experiment. Inner product of the Born