Calculation of the Nonlinear Optical Properties of Molecules

Henry A. Kurtz
Department of Chemistry, Memphis State University, Memphis, Tennessee 35152
James J. P. Stewart and Kenneth M. Dieter
Frank J . Seiler Research Laboratory, U.S. A i r Force Academy, Colorado Springs, Colorado 80840

Received 19 May 1989; accepted 4 August 1989

A finite-field method for the calculation of polarizabilities and hyperpolarizabilities is developed based
on both a n energy expansion and a dipole moment expansion. This procedure is implemented in the
MOPAC semiempirical program. Values and components of the dipole moment ( p ) , polarizability ( a ) ,
first hyperpolarizability (p), and second hyperpolarizability ( 7 ) are calculated as a n extension of the
usual MOPAC run. Applications to benzene and substituted benzenes are shown as test cases utilizing
both MNDO and AM1 Hamiltonians.

INTRODUCTION and large organic systems, procedures for the

calculation of molecular hyperpolarizabilities
The polarization, P, induced in a medium have been implemented into the well estab-
by an external electric field F is given by lished semiempirical electronic structure
P = Po + x ( l ) . F + x(” * F . F program MOPAC.4 In this article, the details
+ x ( 3 ’ - F * F . F+ . . . (1) of the method of calculation will first be out-
lined and then the method evaluated by study-
where x ( ~is) the nth order susceptibility ten-
ing a series of substituted benzenes.
sor of the bulk medium.’ The bulk suscepti-
bilities can in turn be expressed in terms of
the molecular induced dipole. The dipole mo-
ment of a molecule interacting with an elec-
tric field can be written THEORY FOR CALCULATING
POLARIZABILITIES:
Pz = P ; + a$, + (1/2)PZ,kF,Fk
The energy of a system in an electric field
+ (1/6)YyklFJFkF1+ * * (2)
can be written as
where is the permanent dipole moment
and a,,, P l J k , and ylJklare tensor elements of
E(F) = E(0) - p,F, - (1/2!)a,F,F,
the linear polarizability and the first and - (~/~!)P,I,F,F,F~
second hyperpolarizabilities, respectively, of - ( 1 / 4 ! ) ~ ~ k l F , F , F k F-l . . . . (3)
the molecule.’ The second-order term gives The above equation contains implied sums
rise to sum and difference frequency mixing over repeated indices, E ( 0 ) is the energy
(including second harmonic generation) and with no field present, and F , are the compo-
optical rectification. The third-order term is nents of the applied field.
responsible for third harmonic generation and If the molecule is considered t o be in a
two-photon resonances. Typical applications uniform electric field aligned along one of
where these nonlinear effects are important the axis of the system (i.e., [F,,O,O]), the
include amplification, high-resolution spec- values of the polarizabilities along that axis
troscopy, picosecond pulse generation, infra- P,,, and yxxxx>can be obtained. For
( p x aXx,
,
red image conversion, image transmission this case the energy expression reduces to
through optical fibers, optical processing, and
information transfer in E(F,) = E(O) - PLFL - (1/2)a,,F?
As part of a long-range project t o calculate - .
- (1/6)Pz~zF?- (1/24)y,z~zF$ * * *

the nonlinear optical properties of polymer (4)

Journal of Computational Chemistry, Vol. 11, No. 1, 82-87 (1990)

01990 by John Wiley & Sons, Inc. CCC 0192-86511901011001-06$04.00
Calculation of Nonlinear Optical Properties 83

Truncating this expression after the F term rLLJJF?F,”

= -4E(0) - [E(FL7FJ)

and evaluating the energy a t four field + -FJ )

E(-FL >
strengths ( + F L -+2FL)
, leads t o four equa-
+ E(FL -FJ 9
tions in four unknowns. These can be solved
for explicitly and the results are + E(-F,,F,)I
+ 21E(FL)+ E(-F,)I
PLFL = -(2/3)lE(F,) - E(-F,)I + 2[E(F,) + E(-F,)I (7c)
+ (1/12)1E(2FL)- E(-2FL)1 (5a) Further calculations must be done to solve
for yLJJJ
and y J L LThe
L . “axial” equations are
aiiF? = (5/2)E(O)- ( 4 / 3 ) [ E ( F i+
) E(-F,)]
the same as originally derived by Bartlett
+ (1/12)[E(2Fi)+ E(-2F,)] (5b) and P u r ~ i s . ~
An alternate method for obtaining the po-
pi& = [E(F,)- E(-F,)] larizability and hyperpolarizabilities is to
- (1/2)[E(2F,)- E(-2Fz)] (5c) use the equation for the induced dipole mo-
ment (eq. ( 2 ) )instead of the energy. By using
It is also important to obtain the “nonaxial”
a similar procedure for evaluating the dipole
components of the polarizabilities. For ex-
moment at various field strengths, the fol-
ample, ai is necessary to the verify rota-
lowing equations can be derived
tional invariance of the results and to find
the principal “optical” axes. The values of PL = (2/3)[PL(FL)+ PL(-FL)I
pUjand ycij are needed for a comparison with - U / 6 )IPL(2FL)+ P/(-2Fz)1
experimental quantities. To obtain the com- (84
ponents with only two different indices ( i
andj), a linear electric field is applied in the QZLFL = (2/3)[PZ(FL)- PL(-FJl
0 plane along 45” lines between axis i and - ( 1 / W [PZ(2FL)- Pz(-2FL)1
axisj. In this case the energy expression is (8b)
= (2/3)[PL(FJ) - PL(-FJ)]
E(F,,F,) = E(0) - piFi - ~ j F -j (1/2)aicF:
- (1/2)0ljjFj2- a W. . FL FJ -
- ( w a l P L ( 2 F J )- P L ( 4 F J ) 1
(8c)
- (1/6)p,i,F: - (1/6)pjjjFj

- (1/2)p,FiFj2 - (1/2)pj,,FjF:
PZLZF? = (1/3)[PL(2W+ PL(-2FL)
- PLVL) - PZ(-FL)l (8d)
- (1/24)yiiiiF?- (1/24)yJjjF;
- (1/6)yGjjFiFj - (1/6)yj,iiF,FQ PYjFf = ( 1 / 3 )[Pz (2Fj) $- PL (-2Fj)
- P~(F,) - pL(-FJ)I (8e)
- (1/4)yiijF7 (6)
If the “axial” calculations have already YZLLLR = [PL(2FL)- PL(-2F/)1
been done, then there are only six unknowns - [PL(FL)- PL(-F/)I (8f)
in this expression. By using the results of YLIJJFLF,” = (I/’) [PU,(FL,FJ)- PL(-FL7FJ)
energy calculations at (Fi ,Fj,), (Pi,-Fj 1, + PL(FL,-q
(-Fi,Fj), (-Fi, - F j ) , it is possible to solve - PL ( - F L > -FJ 11
for four of the unknowns. The results of the
- lPL(FL) - PL(-FL)I (8g)
manipulations of the energy expressions are
Although expressed quite differently, these
ajjFiFj = (1/48) [E(2Fi,2Fj) - E(2Fi, -2Fj) expressions are equivalent to those recently
- E(-2F2,2F,) published by Williams.‘ The main differ-
+ E(-2F,, -2F,)] ences between this work and Williams’ are
- ( 1 / 3 )[E(Fi,Fj)- E(Fi, -Fj) in the number of calculations that must be
- E(-F,,F,) done and the electric fields used.
+ E(-Fe, -Fj) (7a) The numerical accuracy of the finite field
equations developed is sensitive to the preci-
sion in the energy or dipole moment calcu-
lation. For example, i t can be seen from
eq. (5d) that for a field strength of 0.001 au,
an error of about in the energies will
84 Kurtz, Stewart, and Dieter

lead to a error in y. For a similar error Table I. Benzene reults.

~ - - ~

in the dipole moment as smaller error will Polarizability Second hyperpolarizability

result in y because of the lower power of the
field. If however, in trying t o reduce this xx 89.06 xxxx 1835.7
YY 89.06 YYYY 1835.7
error by using too large a value of the field, zz 28.17 zzzz 16.7
then electron configurational changes may XXYY 612.0
take place and/or the higher terms in the fi- dau) 68.43 xxzz 532.4
nite field equations will become important. dw") 10.14 YYZZ 532.3
y(au) 1408.3
In a subsequent section a numerical study y(10-36esu) 0.71
will be given to demonstrate the importance Exp." 10.33 Exp.b 0.98
of these effects.
Both methods (the energy expansion and aDewar and Stewart, 1984.
the dipole expansion) have been implemented 'Reference 8.
in the MOPAC program. Since the results
based on SCF calculations should be identi- for y is in good agreement with the result of
cal for both methods, this provides a good Williams' and the recent experimental esti-
check on the results as to when numerical mate of the static field value.*
errors or configuration changes take place. It should be pointed out that to obtain the
If CI calculations are performed, the results type of numerical precision shown in Table I
now differ. This is because an arbitrary CI -
( Y x x x x - Y y y y y and Y x x z z = Y y y z t ) a very small
wave function does not satisfy the Hellmann- SCF convergence criterion must be used. In
Feynman Theorem. this work a value of 1.0 x lop2' was typi-
Once all the components of the hyperpo- cally used-much smaller than is normal
larizabilities have been obtained it is neces- for semiempirical energy and structure
sary t o convert them into "experimental" calculations.
q ~ a n t i t i e s . ' ,For
~ the second hyperpolariz-
ability, the quantity of interest is the mean
MONO-SUBSTITUTED BENZENES
value given by
In order to further evaluate the computa-
Y = m{Yxxxx + Yyyw + Yzzzz
tion of hyperpolarizabilities, a series of mono-
+ 2 [ Y x x y y + Y x x z z + Yy,zzI> (9)
substituted benzenes were studied. As these
For the first hyperpolarizability, P, the vec- molecules do not possess a center of symme-
tor component along the dipole moment is try, they have nonzero values for the dipole
important. The quantity of interest is moment and are good test cases for the first
P, = ( 3 / 5 ) { P * Pl/ll/-ll (10) hyperpolarizability, p. A comparison of the
values obtained and experimental values
where P * P = PxPx + P,E.,P + P Z P Z and are given in Table 11. There does not seem
PL = {PLIL + P, + P L k k l (11) t o be any systematic differences between
theory and experiment, particularly when
the large variation in experimental results
BENZENE
is considered.
Initial calculations were carried out on The results are seen to be quite good for F,
benzene, both as a base for studying the sub- CN, and OH, with essentially the same re-
stituted benzenes and as a well-studied bench- sults obtained with the MNDO and AM1
mark. Since this molecule has a center of methods. In each of these calculations the
symmetry, the dipole ( p ) and first hyperpo- optimized structure for each method was
larizability (p) are zero. MNDO based values employed. However, for aniline and nitro-
obtained for the other constants ( a and y ) benzene, the situation is different. For aniline,
are shown in Table I. It should be noted that the AM1 results agree well with experiment
in the calculation of the polarizability ( a ) , but the MNDO results are too small and of a
the atomic correction factors developed by different sign. However, if the calculations
Dewar and Stewart7 are used. The use of are carried out with the MNDO method a t
these factors gives extremely good agreement the AM1 geometry, a value of 0.66 x esu
with experimental results for a. The value is obtained; but if AM1 is used at the MNDO
Calculation of Nonlinear Optical Properties 85

Table 11.

Dipole moment ( p ,Debye)

MNDO 1.956 3.342 1.164 5.187 1.464
AM1 1.573 3.336 1.233 5.240 1.542
Exp. (Reference 9) 1.66 3.93 1.55 4.4 1.53
Hyperpolarizability Cp,, lo3' esu)
MNDO -0.65 0.43 -0.40 0.39 -0.19
AM1 -0.70 0.41 -0.51 0.02 0.83
Exp." 0.44 0.48 0.17 2.0 0.79
(Reference 10) 0.53 0.36 2.3 0.89
0.70 2.2 1.23
1.48

a Experimental values are absolute values of p,.

geometry, the result is 0.01 x lop3' esu. This The question of the numerical stability and
clearly indicates that the differences are the "best" field strengths to use in eqs. ( 5 ) ,
mainly due t o the optimized geometry from (71, and (8) is addressed by the results shown
each method. The better AM1 results agree in Figures 1 a n d 2. These figures show
with the experience that AM1 geometries are the variation of the calculated values of p
better than MNDO geometries. and y for fluorobenzene with field strength
For nitrobenzene neither of the methods (from .0001 to .1).Over a wide range of field
give correct results. For this molecule the strength, the energy and dipole expressions
calculations were both done at the AM1 ge- can be seen t o give the same results. For
ometry because of an unrealistic orientation y (Fig. 21, erratic values are found below
of the NOz group perpendicular to the ring 0.001 au and are due to the onset of numeri-
in the MNDO optimized geometry. The prob- cal errors. At higher field strengths the com-
lems with nitrobenzene do not seem to be puted value of y rises until a dramatic change
problems with the finite field methods but occurs between 0.1 and 0.05 au. The slow
with the parameterization of N in the MNDO increase in magnitude is also found for p
and AM1 approximations. This may also ac- and can be attributed to the importance of
count for some of the discrepancy found in higher order terms that were neglected in
aniline. A new version of MOPAC contain- the expansion.
ing reparameterized atoms has been devel- To further evaluate the results for the AM1
oped and will be applied to this problem." and MNDO approximations, a detailed com-

-0.61_'----'--- ?--
\0
-0.63 t \

- ',, \
X
el
-0.69
-0.71 I /
-f 0 0-0 Energy Eq.
'\,
\
1
-0.73
-0.75
4i- ___-- +-+
0-0 Dipole Eq

+ YL .

1 .OE-5 1 .OE-4 1 .OE-3 1 .OE-2

Base Field Strength (F)

Figure 1. Variation of p with field strength for fluorobenzene. Filled

circles were obtained using the energy based equations and open circles
were obtain with the dipole based equations.
86 Kurtz, Stewart, and Dieter

1.40 r-

1.30 - 0-0 Energy Eq.

/
0 l 0

0-0 Dipole Eq. 1

3
I I
1.201
(D
r)
0
7

x 1.10-
+
1 .GO-

I I
0.90&- L I 1 I ~

1 - i
I .OE-4 1 .OE- 3 1 .OE-2
Base Field Strength (F)
Figure 2. Variation of y with field strength for fluorobenzene. Filled
circles were obtained using the energy based equations and open circles
were obtain with the dipole based equations.

parison of individual components calculated axes. This comparison reveals that for ani-
for F and p is shown in Table 111. The mole- line, the agreement is quite good between
cules are all oriented such that the principal the two methods with the exception of the px
moments of inertia lie along the x, y, and z and pXn.

Table 111. Calculated Components of p and p

Dipole Moment ( p )
MNDO
X 0.770 -1.315 0.102 -2.041 -0.062
Y 0.0 0.0 0.447 0.0 0.002
2 0.0 0.0 -0.001 0.00 0.573
AM1
X 0.619 1.313 0.179 -2.062 -0.306
Y 0.0 0.0 0.451 0.0 0.0
z 0.0 0.0 0.0 0.0 0.524
Hyperpolarizability (PI
MNDO
XXX -135.19 116.59 -274.21 -165.81 -378.67
XYY 8.41 -22.19 21.74 90.37 23.69
XZZ 1.77 -12.28 -1.36 1.19 -5.73
YYY 0.0 0.0 -5.88 0.0 -0.14
YXX 0.0 0.0 - 12.23 0.0 -0.24
YZZ 0.0 0.0 -2.43 0.0 0.05
zzz 0.0 0.0 0.00 0.0 -2.39
zxx 0.0 0.0 -0.03 0.0 -66.26
ZYY 0.0 0.0 0.01 0.0 -6.46
AM1
xxx -144.12 112.57 -23 1.81 -81.40 -455.69
XYY 6.71 -21.39 20.15 78.48 28.43
xzz 1.86 - 12.89 -0.71 -0.16 -8.87
YYY 0.0 0.0 -8.00 0.0 0.03
Yxx 0.0 0.0 -10.83 0.0 0.04
YZZ 0.0 0.0 -2.44 0.0 0.00
zzz 0.0 0.0 0.00 0.0 -4.54
zxx 0.0 0.0 0.00 0.0 -58.62
ZYY 0.0 0.0 0.00 0.0 -5.36
Calculation of Nonlinear Optical Properties 87

CONCLUSIONS References
The main goal and strength of the proce- 1. C. Flytzanis, “Theory of Nonlinear Optical
dure outlined here is that everything is done Susceptibilities,” in Quantum Electronics, H.
Rabin and C.L. Tang, Eds., Academic Press,
by one program: geometry optimization and New York, 1975.
polarizabilities. All that is needed is a rough 2. D. S. Chemla and J. Zyss, Eds., Nonlinear
initial structure. Since the development is Optical Properties of Organic Molecules and
based on semiempirical methods, treatment Cr.yystals Vol. I and 2, Academic Press, Or-
lando, 1987.
of large organic molecules and polymers is 3. D. Williams, Ed., Nonlinear Optical Proper-
possible. While large basis set ab initio cal- ties of Organic and Polvmeric Materials, Am.
culations can undoubtedly do a more accu- Chem. S&., Washingtoh, D.C., 1983.
4.J. J. P. Stewart, QCPE Program - 455, 1983;
rate job on small molecules, these methods version 3.1 (1986).
are not now feasible for routine calculations 5. R. J. Bartlett and G. D. Purvis, Phvs. Rev. A ,
I ”

that are necessary for screening molecules 20, 1313 (1979).

6 . G . R. J . W i l l i a m s . J. M o l e c . S t r u c t .
for nonlinear activity. (Theochem), 151, 215 (1987).
One important extension of this work is 7. M. J. S. Dewar and J. J. P. Stewart, C h e m .
to test the usefulness of MNDO and AM1 Phvs. Lett., 111, 416 (1984).
methods t o calculate frequency dependent 8. D.P. Shelton,‘J. O p t . S O C .Am., B2, 1880
(1985).
hyperpolarizabilities. Work is currently under- 9. A. L. McClellan, T a b l e s of E x p e r i m e n t a l
way to implement this via a sum-over-states Dipole Moments, W. H. Freeman and Co., San
approach which as been shown to work quite Francisco, 1963.
10. J. F. Nicoud and R. J. Twieg, “Organic EFISH
well within a CNDO framework.12 The fre- Hyperpolarizability Dat.a,” in Nonlinear Op-
quency dependent values are necessary to tical Properties of Organic Molecules a n d
directly compare with the experimental val- Crystals, ed. D.S. Chemla and J. Zyss, Eds.,
Academic Press, Orlando, Florida, 1987.
ues and to make judgements of the potential 11. J. J. P. Stewart, J. C o m p . C h e m . , 10, 209
usefulness of new products. (1989); 10, 221 (1989).
12. V. J. Docherty, D. Pugh, and J. 0. Morley,
Research sponsored by the Air Force Office of Scien- J . C h e m . SOC.,Faraday T r a n s 2, 81, 1179
tific ResearchiAFSC, United States Air Force, under (1985).
Contract F49620-85-C-0013.