Abstract

It is now widely recognized that the long-wavelength second-order nonlinear optical properties of organic molecules which absorb in the visible or near-UV, may be reasonably well interpreted by semiempirical quantum chemical calculations. Methods such as PPP (taking into account only π electrons), CNDO or INDO (including also σ valence electrons), are generally used in conjunction with the sum-over-states expressions to obtain the tensor elements of β(ω₁ + ω₂; −ω₁, −ω₂). Such calculations may predict correct orders of magnitude, show the relative influence of donor and acceptor substituents, and enable one to identify the dominant contributions to particular tensor elements. Although these methods have a common-structure, namely a SCF part followed by some configuration interaction (CI), they also present severe limitations. They are based on a number of particular approximations, and certain intermediate quantities are calibrated on empirical data, which makes in general the results parameter-dependent^1-4). This raises the question on how the situation may be improved. With the advent of large-scale vectorized computers, the possibilities of obtaining better molecular wave-functions also by ab initio methods seem promising. How will this affect the computation of nonlinear optical susceptibilities?

© 1988 Optical Society of America