Hybrid quartic force fields with atom-specific basis sets: an efficient route to calculate anharmonic vibrational...

Successful quantum-mechanical calculations hinge on achieving an optimal balance between accuracy and computational time. The computation of quantum anharmonic PESs and vibrational spectra, particularly for large molecular systems, renders this balance especially challenging and typically demands the use of multilevel approximations. The recently developed atom-specific hybrid basis set (ASHBS) approach [M. Sharma, D. Sharma and T. K. Roy, J. Phys. Chem. A, 2025, 129, 848-859] offers one such system-bath strategy, wherein atoms in the chemically significant region are treated with a high-level basis set, while the remaining atoms of the molecule are described using a lower-level basis set for the calculation of anharmonic PESs. The modes at the chemically important site are taken as the target modes, and their accuracy defines the credibility of the approach. Such a method was validated for PESs that include single-mode and pairwise interactions between normal modes and was applied in vibrational self-consistent field (VSCF) theory and its second-order perturbative variant (VSCF-PT2) algorithms. In this work, we extend our earlier ASHBS studies by evaluating its reliability for constructing widely used quartic force fields (QFFs) and evaluating the resulting PESs within effective harmonic oscillator (EHO), VSCF, and VPT2 algorithms for the computation of vibrational transitions. A comprehensive benchmark analysis of hybrid QFFs is performed using six electronic basis sets of different sizes and families, along with DFT-based B3PW91 and B3LYP functionals, and MP2 methods, for a small prototype molecule, acrylamide. Subsequently, four different high-low basis set combinations are employed across three anharmonic vibrational algorithms for a larger molecule, isatin. It is found that, typically, for the QFFs of a chosen active site, the transitions of the corresponding modes are found to be closer to those of a high basis set while saving 60-70% of the computational time.