Assessment of Explicit and Implicit Solvation Models for Linear and Nonlinear Optical Properties of Para-Nitroaniline...

Computational prediction of novel materials with large linear and nonlinear optical (L&NLO) responses requires models that combine sufficient accuracy with an affordable computational cost. The high sensitivity of L&NLO properties to intermolecular interactions and their dynamics in molecular materials, such as solutions, requires special attention. Using solvated para-nitroaniline as a case example, we assess the performance of a diverse set of solvation models for the evaluation of L&NLO properties, with emphasis on the comparability of the properties obtained at different levels of the underlying local-field theory. Specifically, we compare the implicit polarizable continuum model (IEF-PCM), with different classes of explicit models, such as the rigorous local field (RLF), polarizable embedding (PE), supermolecular (SMA), and semicontinuum approaches (SCA). Cyclohexane, 1,4-dioxane, and tetrahydrofuran are considered to represent environments with distinct multipolar interaction patterns. The analysis distinguishes between solute, reaction-field, and effective response properties, enabling direct comparison of results obtained at different theoretical levels. For nearly all models, a computational procedure can be chosen to achieve a fair level of mutual agreement up to first hyperpolarizabilities. Notably, the RLF approach shows particularly good performance, enabling fast screening of large molecular trajectories while providing L&NLO properties that can be directly compared with experimental measurements. The study further highlights several implementation issues and methodological limitations, including the treatment of higher multipole contributions (IEF-PCM), the role of property distributions in local-field evaluation (RLF), the implementation of effective field factors in bulk environments (PE), and practical limitations on cluster size in semicontinuum approaches (SCA), all of which can severely affect predicted L&NLO properties.