Finite Field Method for Nonlinear Optical Property Prediction Using Rational Function Approximants.

The finite field (FF) method is a quick, easy-to-implement tool for the prediction of nonlinear optical properties. Here, we present and explore a novel variant of the FF method, which uses a rational function to fit a molecule's energy with respect to an electric field. Similarly to previous FF methods, factors crucial for the method's accuracy were tuned. These factors include the number of terms in the function, the distribution of fields used to construct the approximation, and the initial field in the approximation. It was found that the approximant form that best fits the energy has four numerator terms and three denominator terms. To determine a reasonable field distribution, the common ratio of a geometric progression was optimized to √2. Finally, an algorithm for determining a good initial field guess was devised. The optimized FF method was used to compute the polarizability and second hyperpolarizability for a set of 121 molecules and the first hyperpolarizability for a set of 91 molecules. The results from this were compared to a previous polynomial-based FF method. It was found that using a rational function gives higher errors compared to the polynomial model. However, unlike the polynomial model, no subsequent refinement steps were needed to obtain usable results. An overall comparison of the behavior of the two methods also shows that the rational function is less sensitive to the chosen initial field, making it a good choice for new quantum chemistry codes.