A General Approach to Perturbation Theoretic Calculations of Nonlinear
Susceptibility Coefficient Tensors for Ferroelectric Materials
J. F. Webb
Sirindhorn International Institute of Technology, Thammasat University,
P. 0. Box 22, Pathumthani 12121, Thailand
Abstract- Perturbation analysis in which the time variation where m and -y are mass and damping parameters respectively,
of an electric field is treated as inducing a weakly nonlinear and, following Landau-Devonshire theory [3], [8], [9], F is the
response that is expressed as a Taylor series expansion is a free energy per unit volume, which for the purposes of this
standard technique for deriving the nonlinear susceptibility . . i.
coefficients that relate the electric field to the polarization. paper is given in the general form
In this paper a generalized method is used to generate the F-1A P P 1A P P P P
coefficients and Fourier analysis is employed in order to treat -2 cla2 ct1 02 + 4 clca2ca3ca4 cti ct2 0a3 ca4
any time dependent electric field as a superposition of single + 1 Aa I Pol ... Pa (3)
frequency waves. It is then shown how this method can be ap- nF ...nF
F (
plied to the calculation of nonlinear susceptibility coefficients Here AOe,I 0e2 = a,ee2 (T - To) (summation convention not
in bulk ferroelectric materials in such a way that preserves the implied), to ensure a phase transition at a temperature related
tensorial nature of the coefficients, appropriate for crystalline to To, where each of the terms in a,,.2 are positive constants,
materials. An advantage of this general approach is that it 15
useful for the development of computer modelling packages and the rest of the A coefficients are constants with the signs of
the fourth and sixth order terms in P dependent on whether the
involving nonlinear calculations.
phase transition is first or second order (see Refs. [3], [9]) and
INTRODUCTION the higher order terms are assumed to be positive. However the
The susceptibility coefficients, which here are treated as ten- symmetry of the ferroelectric crystal will determine which of
sors so that they apply to real crystals, relate the polarization of the A coefficients are identically zero [10], [11] Also, nF > n
a dielectric material (in this case a ferroelectric) to an applied (where n is defined below as the total number of orders being
electric field E as follows considered) and is even. Also, since the multiplication of the
polarization components is commutative those A coefficients
PeO (Xi c= i Eeg + Xtla2 Ei E2 which have the same number of x, y and z components can be
+ + Xtal anEal... Ean) (1) set equal to each other. Two examples of this are setting Axy =
A and setting the 3!/2! possible A coefficients which have
in which terms up to any order n are included, the subscripts A
over xyad nttwo
x subscripts and one y subscript (one such coefficient is
range over x, y and z, and the summationrange convention for A. )
repeated indices is invoked (and is to be applied from now on A equal to each other. This is called intrinsic permutation
unless otherwise stated). In fact the most general relationship symmetry, and is a property which can also be chosen for the
between polarization and the electric field is more complicated susceptibility coefficients as pointed out below.
.. . . . .
than this involving integrals [1]. However, in this paper we ~~~~~~~~~The
equation of motion (2) describes oscillatory motion
will only deal with fields that consist of a superposition of since it has includes the parameters m and -y. But setting
monochromatic waves see (4) and the representation in (1) m is all that is required to describe relaxational motion
should be sufficient. instead.
We consider the case that the incident field E is a super-
Particular calculations for a limited number of orders have osition of monochromatic lane waves iven b
already been carried out [2]-[4], and have been extended to
any number of orders n [5], [6], but this extension was only for E 1 n tE iWjt F
the non-tensorial case. One of the major aims of the present 2Z V + , e(
paper is to give a sketch of how to extend the previous work j=1
to the tensorial case for any number of orders n, paving the where EP,, = E1>, [1].
way for a general formulation of the susceptibility coefficients SOLVING THE EQUATIONS OF MOTION
applicable to real crystals. The first step in solving (2), the equations of motion, is to
EQUATIONS OF MOTION AND ASSOCIATED EQUATIONS express the polarization as a sum of static and dynamic parts
The basic equations that will be used for finding the sus- = P,s + p1(t), and put ,9 q5(P) f,(p), so that
ceptibility coefficients are the Landau-Khalatnikov equations aF can be expanded in a Taylor series about P_ = P",s
of motion [3], [7]
02p ap &F fJ(P) =fjilpol + 4!afica2PcaiPc2
259
where p(k)(w) = (k)1 {p(k)(t . tk)}
f( fal (2 r)k j dt ... j dtkp(k)(tl, ., t)ek(1±±)
(= k <
8
)
, and, for derivatives,
''PO' k P'1l1=P'111,P1'12=P'"28'
al
k =P k 't
and P, is the spontaneous polarization found by minimizing &tk
F in (3) according to ,= However, if (1) is used to { k
substitute for the p,1 in (5) it is evident that fr, can be expressed W(l) -i(wi + + wi) P (wi + + wi) . (12)
in terms of the electric field components through the Taylor Linear Terms
series (8) can be cast into the form
f4 (P) = g4a,Ea, +92EaE
a (6)(OO+9)al) (l), (t) = al Ea,. (13)
+ *** + g/_tal**n E l * * Ea2, (6) X,uai
-t
..
But from (1), (3), (5) and (6) it can be shown that 9,,, is
where
k
proportional to x,0t1
and may be written as g, = Xal1 a,
1 ( &kf,l say; which means that g /x1 in (13) can be replaced by
YJNYit1 O{ k! k &El..**&E J EIY1
=E__ 2 =-E k,3.l . Now, with the aid of the Fourier transform equations,
(10)-(12), (13) can be expressed as
The equations of motion in (2) can now be written
coO @(1) { ( }mw2 &y + 1 VpaI ) p)
eOOx,gl Ecxl + ...
-tal n an
f=(1){/lEa1(w)}, (14)
=(g9ya,Ea, + + 9,utal Eagn) + El,, (7)
nEa, ..
where
where 0 = m Dt2 + m0*t. Equating like terms in powers of EM,1 nE
gives E,(w) 2 (i,1,,.(W
E -w + E,, 6(wo + w))
coO X4a Eal =0 (1)1(t) =(g1)Ea1+6( j=F (15)
Rearranging yields the following equation from which the
for the linear terms, where the relation E,M 6-,a.Ea. has linear susceptibility coefficients can be calculated
been used to facilitate the linear-terms calculations below, and
() P~~~~~~~~(1),(t) = /-talg Eagl (t) =~W (1) () a)}
fOr
Xteol term,Esl forEwhc > 1. Notetat) E= ((1) {sa1(w)d,,lEO1(w)} (16)
=-g_al asr Ea I.. Ea,r (9)
where
subsequent equations that involve like terms equated (in E1,1, --tal + S/1(7
including both linear and nonlinear terms) it is not necessary to The transform integrals implied in (16) can be carried out
invoke the summation convention since such equations apply systematically following Ref. [1, Chap. 2].
to all individual terms in the implied summation. Furthermore,
since the multiplication of the electric field components is Nonlinear Terms
commutative, intrinsic permutation symmetry can be applied Applying (11) to the left side of (9) and using a previous
in the same was as that described for the A components result [5], [6] also applicable here for the way in which the
in (3), which significantly reduces the number of susceptibility function s in (17) enters into the nonlinear expression, gives
coefficients that need to be calculated.
Fourier Transforms stl (wil + +wr)
We will solve (2) by using Fourier transforms defined as Next (11) is applied to the right side of (9):
follows g(1 1 { gi.h Eeil * Ex}
p(k)(t) a(k) {p(k)(wi :
=
Wk)} h(p(- 1) p(jr2)2(wr-), .P.a1iiw)): (
+ I'+ I(k) where the function h has been introduced to express the fact
'dwkPp (w1,...'
...
-y ",1 y Sl that the procedure for calculating the susceptibility coefficients
(10) for a particular order is iterative in the sense that it depends
260
on lower orders, except for the linear orders which are found is also possible to carry the calculations further resulting in
explicitly from (16). This iterative nature follows from the more explicit formulae. Such work is in progress and will be
results for the non-tensor calculations done previously [5], reported soon.
[6], and applies here when the susceptibility coefficients are ACKNOWLEDGMENT
tensors.
Hence, using (18) and (19), and moving back to the time I would like to thank Sirindhorn International Institute of
domain yields an equation for calculating any nonlinear (r > Technology for financial support.
1) susceptibility coefficient: REFERENCES
p(f (t) = tal --r(t)E0, (t) Eei, (t)
/
... [1] P N. Butcher and D. Cotter, The Elements of Nonlinear Optics.
Cambridge: Cambridge University Press, 1991.
= @(r) {p(, )(aJw, * ,wr)} [2] J. Osman, S.-C. Lim, D. R. Tilley, and Y Ishibashi, "Nonlinear optic
LA 'J coefficients in the ferroelectric phase," Journal of the Korean Physical
__ @(r).S. ( + +b ) Society, vol. 32, pp. S445-S449, 1998.
I iW1r 1 [3] J. Osman, Y Ishibashi, and D. R. Tilley, "Calculation of nonlinear
(r-1) w (1)
,. (p,o1(w)))J susceptibility tensor components in ferroelectrics," Japanese Journal of
x h(par (20) Applied Physics, vol. 37, pp. 4887-4893, 1998.
[4] R. Murgan, D. R. Tilley, Y. Ishibashi, J. F. Webb, and J. Osman, "Cal-
culation of nonlinear-susceptibility tensor components in ferroelectrics:
CONCLUSION Cubic, tetragonal, and rhombohedral symmetries," Journal of the Optical
Society of America B, vol. 19, pp. 2007-2021, 2002.
In this paper a method for calculating susceptibility coef- [5] J. F. Webb, "Derivation of nonlinear susceptibility coefficients in antifer-
roelectrics," Microelectronic Engineering, vol. 66, pp. 584-590, 2003.
ficients for ferroelectrics has been formulated in a general [6] J. F. Webb, "A general approach to perturbation theoretic analysis in non-
way, as summarized by (16) and (20). This points towards linear optics and its application to ferroelectrics and antiferroelectrics,"
automatic implementation of such calculations, for example, International Journal of Modern Physics B, vol. 17, pp. 4355-4360,
so that computer programs that model electronic circuits can 2003.
[7] Y Ishibashi and H. Orihara, "A phenomenological theory of nonlinear
be extended to incorporate nonlinear effects when modelling dielectric response," Ferroelectrics, vol. 156, pp. 185-188, 1994.
ferroelectric devices for which nonlinear effects are important, [8] A. F. Devonshire, "Theory of barium titanate (part 1)," Philosophical
as they are for things like parametric amplification. The overall [9] Magazine,
M. E. Linesvol.and
40,A.pp.M.1040-1063, 1949. and Applications of Ferro-
Glass, Principles
framework elucidated here ought to produce the correct results electrics and Related Materials. Oxford: Clarendon Press, 1977.
after appropriate calculations, and the crystal symmetry of a [10] s. V. Popov, Y P. Svirko, and N. I. Zheludev, Susceptibility Tensors for
Nonlinear Optics. Bristol: Institute of Physics Publishing, 1995.
[11] J. Nye, Physical Properties of Crystals. Their Representation by Tensors
coefficients in the free energy, (3), should appear. However it and Matrices. Oxford: Oxford University Press, 1964.
261
