Professional Documents
Culture Documents
Chemical Physics Lettres. 1990, 171, 209-216
Chemical Physics Lettres. 1990, 171, 209-216
The static dipole polarizability, first and second hyperpolarizability tensors of molecules with nonlinear optical characteristics
are calculated via ab initio coupled-perturbed Hartree-Fock theory. A study of the effect of basis set augmentation on the calcu-
lated properties for the nitrobenzene molecule emphasizes the importance of diffuse functions. The method is then applied to a
series of methyl substituted nitroanilines, and comparison with existing experimental data and other computational data is pre-
sented. The results indicate that the present approach works well to reproduce experimentally observed trends.
210
Volume 17I, number 3 CHEMICAL PHYSICS LETTERS 3 August 1990
be required on a systematic set of organic molecules truncation of the sum over states. The results are
to calibrate the accuracy of quantum mechanical cal- critically dependent on the level of truncation and
culations. Finally it is important to keep in mind that thus the computational procedure does not appear to
the experimental quantities are frequency dependent be well defined at the present time.
and that the formalism used in this study does not Relatively few ab initi calculations of (hy-
take into account frequency dependence. per)polarizabilities have been published, and most
Several theoretical approaches to calculating mo- have dealt with small molecular systems [ 14,15 1. In
lecular nonlinear optical properties have appeared in this paper we report on the calculation of these prop-
the literature. They differ first in the form of the mo- erties for larger systems which have been the subject
lecular wavefunction which defines the ingredients of numerous experimental investigations, in panic-
from which the nonlinear optical properties .are cal- ular we present the results of calculations of these
culated, i.e. the electron density and its changes in properties for methyl substituted nitroanilines per-
the presence of an electric field. Simple and attrac- formed in the framework of the Hartree-Fock SCF
tive approaches are the semi-empirical model Ham- theory using extended basis sets. The properties of
iltonian methods such as the all-electron CNDO interest were evaluated from analytic expressions for
model [ 7 ] or the n-electron-only PPP model [ 8 1. the derivatives of the SCF energy with respect to an
Owing to their computationally inexpensive char- electric field, obtained through CPHF theory. Sec-
acteristics these methods can be relatively easily ap- tion 2 of this paper gives the most important tech-
plied to larger molecular systems [ 5,9-l 31. First nical details related to the calculations. The basis set
principles methods deal traditionally with smaller dependence of the calculated properties for nitro-
molecules because of their greater computing re- benzene is described in section 3. This study resulted
quirements. They have however the potential to pro- in the selection of a basis set used for the calculations
vide results of a more reliable nature. The second on nitroaniline derivatives which are discussed in
difference among theoretical approaches lies in the section 4. Conclusions are given in section 5.
manner in which the nonlinear response is calcu-
lated in the framework of perturbation theory, either
by means of the CPHF approach [ 2,3,13 ] or the sum-
over-states (SOS) approach [9-l 21. Especially rel- 2. Computationaldetails
evant to the research presented here are the recent
calculations of /3and y reported by Morley et al. [ 12 ] The molecules considered in this work are shown
for the same systems as the ones investigated here. in fig. 1 along with the system of Cartesian coordi-
We note that because of the size of molecules con- nates adopted. All geometrical parameters of aniline,
sidered here these authors must resort to a drastic nitrobenzene (NB), and p-nitroaniline (PNA) were
/N\ 0
0 a
/N\ .YN\ 0
0
/N\ 0 0 0
Fig. 1.
211
Volume I7 1, number 3 CHEMICALPHYSICSLETTERS 3 August 1990
optimized at the SCF level of theory using a (9s,Sp) / tance in ab initio calculations of NLO properties. It
[ 3s,2p] double-zeta quality basis set for the second is now generally accepted that energy-optimized ba-
row atoms and a (4s) / [ 2s ] double-zeta basis set for sis sets which yield satisfactory geometrical param-
H. This basis set is denoted DZV thereafter [ 161. eters in molecular structure determination need be
The geometry of the nitroaniline derivatives o- augmented with diffuse functions for a more flexible
methyl-p-nitroaniline (MNA), p-N-methylamino- description of the electron density away from the nu-
nitrobenzene (MNB), and pN,N-dimethylaminon- clei. For small molecules the basis set dependence of
itrobenzene [ 61 borrowed parameters from the op- the calculated properties is already well documented
timized geometry of pnitroaniline, with “idealized” [ 2,141. For larger molecules it is currently believed
methyl groups substituted for the H atoms where ap- that functions on other centers may lead to an ad-
propriate. For nitrobenzene and p-nitroaniline the equate description of the polarization effects asso-
calculated structures are in overall excellent agree- ciated with a given center. Until recently no com-
ment with experimental structures. For this latter putational evidence of this hypothesis was available.
molecule the (hyper)polarizability tensors calcu- The effect of basis set augmentation on the calcu-
lated at the theoretical geometry and at the experi- lated properties for large systems was first explored
mental geometry showed no significant differences. for polyenes C4H6 to CzzH2+ For short polyenes ex-
Therefore we are confident that the conclusions based tra diffuse functions and diffuse polarization func-
on the present study would hold, had we used slightly tions are crucial for describing the second hyper-
“better” geometries. polarizability. However for the longer chain the
Detail of the equations and their computer imple- importance of diffuse functions are notably reduced.
mentation in the program HONDO [ 171 can be We are concerned here with the effect of the basis
found elsewhere. The calculations were performed set augmentation on the calculated polarizability, first
on the LCAP-2 and LCAP-3090 computer systems and second hyperpolarizabilities for the nitroben-
[1X1* zene molecule. Starting from the double-zeta quality
Before describing the computational results, we basis set used for the geometry determination, we
must underscore the fact that the calculated tensors added up to three diffuse functions on the second-
/I and y reported throughout this paper are in accord row atoms of p and/or d type. No diffuse functions
with eq. ( 1). They are divided by a factor of 2 and were appended to the double-zeta basis set for the
6 respectively from the values printed by the com- hydrogen atoms. This choice was made since we are
puter program HONDO. This is due to a difference concerned with the effects associated with the out-
in definition used by experimentalists which follow of-plane Kelectrons mainly. In doing so we sacrificed
the Bloembergen model [ 191 and by theoreticians somewhat the accuracy of some of the components
who follow the SOS approach, both of them based on of the polarizability tensors. The fact that the pzII
a perturbation expansion of the energy and of the in- component dominates all other quantities entering
duced polarization (eq. ( 1)) on the one hand (the the observable Bv,, and that the diagonal component
numerical factor in front of the term at order n is l/ yzzzzis very much larger than any other component
n, and theoreticians which adopt the CPHF ap- of the y tensor provides a justification of this choice.
proach based on a Taylor series expansion of the en- Our previously published results on C4H6 suggested
ergy on the other hand (the numerical factor in front that (Yand /3 should not be affected too much by the
of the term at order n is 1 /n! ). For the sake of com- lack of diffuse functions on the hydrogen atoms. It
parison with already published experimental data, is worth remarking that we have not explicitely in-
we report tensor values properly scaled to the Bloem- cluded diffuse s functions. This is because we make
bergen model. use of Cartesian Gaussian functions. In doing so a
single exponent cd is associated with six functions,
live of the 3d type and one of the 3s type. The radial
3. Basis set effects for nitrobenzene part of the latter one has a maximum in the region
where a 1s function with exponent csx2.Old would
The choice of basis set is of paramount impor- have its maximum. Therefore some of the basis sets
212
Volume 171, number 3 CHEMICAL PHYSICS LETTERS 3 August 1990
described below include diffuse s functions implic- /I tensor. The components jxxz and jl~~~are relatively
itely. The Gaussian exponents of the diffuse func- insensitive to the basis set. The quantity AM is how-
tions for the second row atoms were varied in a dis- ever sensitive to the basis set owing to the negative
cret manner to check the importance of diffuseness sign of /I211which enters its definition.
of the functions. Ultimately we wanted to select a In contrast at least two diffise d-type functions ap-
basis set which would offer a satisfactory compro- pear to be necessary to obtain satisfactory values of
mise between size and quality. y. Basis sets A through D have no or at most one dif-
The individual independent components along the fuse d set. Basis sets E through I have two diffuse d
Cartesian axes of the (hyper)polarizability tensors functions. The components Y,-_, rxxyv,y===,yW and
calculated with different basis sets are given in table ryvzzhave dramatically different values for bases A
1. The notation adopted indicates which functions to D than for bases E to I, some differ by more than
are appended to the DZV basis set to form bases A one order of magnitude. This is not the case for JJ====,
through I. The exponents of the additional functions which happens to be the largest component of them
are given next to their types. It is striking to note that all. Again we attribute this behavior to the cooper-
for a! and /I the various basis sets yield very similar ative effects of neighboring basis functions. The end
results. The diffuse functions affect most aY,,and fl,,,,=, result however is a mean value 7 which reaches a sec-
which happen to be the smaller components in mag- ond plateau with bases E through I. Somewhat sur-
nitude. The CX, and CL components, and the mean prisingly though, the inclusion of a tight d function
value & change by at most 15%. Even the DZV re- appears to result in a slightly smaller mean value of
sults are satisfactory. This may be the most vivid il- jIYecand y. The diffuse p function added to set E to
lustration yet of how basis functions on neighboring create set I has also a marginal effect. Clearly we do
atoms help in describing the polarization of the va- not observe basis set convergence for the properties
lence electrons. The same conclusion applies to the of interest. However in accord with our previous
Table 1
Nitrobenzene (hyper)polarizabilities: basis set dependence
A B C D E F G H I
DZV Id Id 2d 2d 2d 2d 3d 3d lp,2d
d-8 d.2 d.8 d.8 d.2 d.2 d.8 d.8 p.05
d.2 d.15 d.05 d.015 d.2 d.15 d.2
d.015 d.085 d.015
P .) 6.0 5.5 5.6 5.3 5.4 5.6 5.6 5.4 5.4 5.6
%z b) 13.0 13.0 13.5 13.6 13.7 14.0 13.7 13.7 14.0 14.0
%Y 4.3 4.8 6.4 6.5 6.9 7.1 6.9 7.0 7.2 7.1
t&z 13.7 14.2 15.3 15.4 15.6 15.8 15.5 15.7 15.9 15.7
a! 10.3 10.7 11.7 11.8 12.1 12.3 12.0 12.1 12.3 12.3
/%xzC~ 48.3 41.7 38.5 37.0 38.1 45.4 42.3 40.1 43.4 43.2
Bw* 5.0 4.1 5.8 6.2 8.3 16.9 16.7 16.8 15.5 20.3
- 143.2 - 127.3 - 134.7 - 139.8 - 141.3 - 127.8 -131.4 - 142.7
k - 138.5
90.2 -117.3
71.5 104.6 84.1 88.3 94.4 99.0 70.9 72.5 99.5
Y- d) 33.9 55.4 145.6 159.4 222.9 631.6 448.6 446.3 524.2 564.6
YWY 45.6 41.0 60.6 66.6 92.7 300.4 294.6 289.3 234.2 327.8
Yzw -75.8 -60.4 -18.6 - 10.4 12.3 123.1 100.3 96.1 97.6 109.4
YYJW 46.7 38.9 72.8 77.3 123.4 632.9 795.9 751.0 414.8 725.5
VW,, 23.3 22.8 44. I 49.5 72.5 244.7 267.7 259.6 189.5 250.5
Yz. 2051.5 1709.9 1920.6 1748.3 1923.7 2591.8 2460.7 2250.2 2354.5 2565.3
P 423.7 362.2 462.2 439.3 525.0 1038.6 1006.1 947.5 867.2 1046.2
213
Volume 17I, number 3 CHEMICAL PHYSICS LETTERS 3 August 1990
work [2], the very distinct plateaux of calculated We first note that the SCF calculated dipoles are
values are sufficient for our purpose of defining a ba- in excellent agreement with experiment, except for
sis set of semiquantitative accuracy which should the PNA value which is somewhat too large. The
make it possible to differentiate substituent effects in agreement for & is also excellent, and these results
chemical structures. contribute to the general belief that the ab initio SCF
The present results combined with our earlier methodology is able to predict dipole polarizabilities
findings for the polyenes suggest that basis set E with with a high degree of accuracy. The calculated &
a diffuse d function and another very diffuse one values extracted from eqs. (5) and (6 ) show some
constitutes a satisfactory compromise between size interesting trends as well: for aniline and nitroben-
and quality. A slightly smaller basis set would have zene, IL, is small; for PNA and its methyl substi-
been sufficient, had we limited our research to the tuted derivatives aec is much larger although they
calculation of (Y and /3. The calculations reported appear to be a factor of 2 to 3 smaller than the ex-
thereafter were carried out with basis set E. perimental values measured at a wavelength of 1.9 1
pm. The jj values are found to be in qualitative ac-
cord with experiment [20] although their absolute
4. Application to nitroaniline derivatives values are in error by a factor z 5. We note that these
results are in line with calculations recently reported
To establish the level of reliability of ab initio SCF for benzene for (Yand y [ 2 I], and for the haloform
derived (hyper )polarizabilities, we calculated these series CHX3 for (Y,/?, and y also reported recently
properties for aniline, PNA, MNA, MNB, and [ 23 1. For the haloform series, the calculated values
DMNB using the basis set denoted E. The results are of p were found to be z l/6 the experimental value.
gathered in table 2 along with some available ex- We note that 7 shows a behavior which parallels a,,.
perjmental data from refs. [ 5,6,20]. The mean value of y for PNA and its derivatives are
Table 2
(Hyper)polarizabilities ‘) of benzene derivatives
5.7f’
1.7
214
Volume 171, number 3 CHEMICAL PHYSICS LETTERS 3 August 1990
at least twice as large as for aniline and NB. The in- properties. Using the approach described in ref. [ 15]
cremental factors for 7 are not as large as forjIvecwhen it will be possible to remove this part of uncertainty.
going from PNA to DMNB. Computed frequency dependent (hyper)polariza-
The key ingredients for large fl are well under- bilities for some of the same molecules will be re-
stood: they are linked to the presence of a “donor” ported in a future publication.
group and of an “acceptor” group in the molecule Finally we note a potential unvaluable contribu-
which result in a highly polarizable system. Aniline tion of these and similar studies to the interpretation
and nitrobenzene are lacking one of the two, and of experimental data. Traditionally, the second hy-
therefore their j? values are small. This is observed perpolarizability contribution to EFISH measure-
experimentally for p and the present calculations are ments is ignored when extracting b, This is because
in accord with this notion. Nitroanilines contain a there is often no knowledge of the order of magni-
donor and an acceptor group, they have large /3’soh- tude of y which is then ignored. Calculations of the
served experimentally and again the present calcu- type described here, by yielding values of y may help
lations reflect this fact. The present study suggests improve the interpretation of experimental data.
that the same ingredients contribute to large y values.
In table 2 we collected also other calculated values
previously published and obtained with the PPP and References
CNDO/S methods. We note that great care and cau-
tion must be used when comparing data, as the pub- [ 1] D.S. Chemla and .I. Zyss, eds., Nonlinear optical properties
lished data do not appear to be always consistent. of organic molecules and crystals, Vols. 1,2 (AcademicPress,
New York, 1987).
See for example refs. [ 9, lo] _The values listed as PPP [ 21 J.B. Hurst, M. Dupuis and E. Clementi, J. Chem. Phys. 89
are taken from ref. [ lo]. However we note the sen- (1988) 385.
sitivity of the PPP results on the choice of the PPP [ 31 C.P. DeMelo and R. Silbey, J. Chem. Phys. 88 (1988) 2558.
parameters, as the jl values may vary by as much as [4] B. K&man, Chem. Phys. Letters 143 (1986) 81.
a factor of two. The SCF method used in this work [ 51 D. Pugh and J.O. Morley, in: Nonlinear optical properties
and the PPP method yield similar results for aniline of organic molecules and crystals, Vol. 1, eds. D.S. Chemla
and J. Zyss (Academic Press, New York, 1987) p. 193.
and nitrobenzene, while the j$, for nitroaniline is
[6] J.F. Nicoud and R.J. Tweigh, in: Nonlinear optical
significantly larger than for PNA using the PPP properties of organic molecules and crystals, Vol. 2, eds. D.S.
method. The CNDO/S values listed in table 2 are Chemla and J. Zyss (Academic Press, New York, 1987) p.
taken from refs. [ 11,221. 255.
[7] J.A. Pople, D.P. Santry and G.A. Segal, J. Chem. Phys. 43
(1965) 9219.
[ 81 J.A. Pople, Trans. Faraday Sot. 49 (1953) 1375;
5. Conclusions
R. Pariser and R.G. Parr, J. Chem. Phys. 21 ( 1953) 466;
R. Pariser, J. Chem. Phys. 21 (1953) 568.
To the best of our knowledge this study represents [ 9) D. Li, T.J. Marks and M.A. Ratner, Chem. Phys. Letters
the first attempt to calculate the (hyper)polarizability 131 (1986) 370.
tensors of experimentally well studied organic mol- [IO] D. Li, M.A. Ratner and T.J. Marks, J. Am. Chem. Sec. 110
ecules by means of a method based on ab inito SCF- (1988) 1707.
LCAO formalism. The trends extracted from cal- [ 11I C.G. Teng and A.F. Garito, Phys. Rev. Letters 50 ( 1983)
350;
culatedb, (Y,8, and y are in accord with experimental S.L. Lalama and A.F. Garito, Phys. Rev. A 20 (1979) 1179.
observations. We note however that calculated j?“,‘s [ 121J.O. Morley, P. Pavlides D. Pugh, J. Chem. Sot. Faraday
are not quite within a factor of 2 of the experimental Trans. II 85 (1989) 1789.
data, and that calculated y’s are x 5 times too small [ 131J. Waite and M.G. Papadopoulos, J. Phys. Chem. 93 (1989)
consistently. While this finding is somewhat disap- 43; J. Chem. Phys. 85 (1986) 2831;
pointing, it is in line with results obtained for other M.G. Papadopoulos and J. Waite, Chem. Phys. Letters 135
(1986) 361.
molecules using the same approach. Of course com-
[ 141 C.E. Dykstra, J. Chem. Phys. 82 (1985) 4120;
parison was made between experimentally measured S. Liu and C.E. Dykatra, Chem. Phys. Letters 119 ( 1985)
frequency dependent properties and calculated static 407;
215
Volume I 7 1, number 3 CHEMICAL PHYSICS LETTERS 3 August 1990
S.Y. Liu, C.E. Dykstra, K Kolenbrander and J.M. Lisy, J. [ 181 E. Clementi, J. Chem. Phys. 89 ( 1985) 4426.
Chem. Phys. 85 (1986) 2077; [ 191N. Bloembergen, Nonlinear optics (Benjamin, New York,
S.Y.LiuandC.E.Dykstra, I. Phys. Chem. 91 (1987) 1749. 1965).
[ 151 H. Sekino and R.J. Barlett, J. Chem. Phys. 85 (1986) 976. [ 201 G.R. Meredith, B. Buchalter and C. Hanzlik, J. Chem. Phys.
[ 16 ] T.H. Dunning and P.J. Hay, Modern theoretical chemistry, 78 (1983) 1543.
Vol. 2, ed. H.F. Schaefer III (Plenum Press, New York, (211 E. Pert-in, P.N. Prasad, P. Mougenot and M. Dupuis, J.
1976). Chcm. Phys. 91 (1989) 4728.
[ 17 ] M. Dupuis, J. Rys and H.F. King, J. Chem. Phys. 65 ( 1976) [22] V.J. Docherty, 0. Pugh and J.O. Morley, J. Chem. Sot.
111; FaradayTransII81 (1985) 1179.
M. Dupuis, P. Mougenot, J.D. Watts, G.J.B. Hurst and H.O. [23] S. Kama, M. Dupuis, E. Pen-in and P.N. Prasad, J. Chem.
Villar, HONDG (8.0). in: Modem technique in Phys., submitted for publication.
computational chemistry, ed. E. Clementi (Escom, 1989).
216