Toward a numerically efficient description of bulk-solvated anionic states

Abstract

We investigate the vertical electron attachment energy (VAE) of 1-methyl-4-nitroimidazole, a model radiosensitizer, employing quantum mechanics/molecular mechanics (QM/MM) and QM/polarized continuum (QM/PCM) solvation models. We considered the solvent-excluded surface (QM/PCM-SES) and Van der Waals (QM/PCM-VDW) cavities within the PCM framework, the electrostatic embedding QM/MM (EE-QM/MM) model, and the self-consistent sequential QM/MM polarizable electrostatic embedding (scPEE-S-QM/MM) model. Due to slow VAE convergence concerning the number of QM solvent molecules, full QM calculations prove inefficient. Ensemble averages in these calculations do not align with VAEs computed for the representative solute–solvent configuration. QM/MM and QM/PCM calculations show agreement with each other for sufficiently large QM regions, although the QM/PCM-VDW model exhibits artifacts linked to the cavity. QM/MM models demonstrate good agreement between ensemble averages and VAEs calculated with the representative configuration. Notably, the VAE computed with the scPEE-S-QM/MM model achieves faster convergence concerning the number of QM water molecules compared to the EE-QM/MM model, attributed to enhanced efficiency from MM charge polarization in the scPEE-S-QM/MM approach. This emphasizes the importance of QM/classical models with accurate solute–solvent and solvent–solvent mutual polarization for obtaining converged VAEs at a reasonable computational cost. The full-QM approach is very inefficient, while the microsolvation model is inaccurate. Computational savings in QM/MM models result from electrostatic embedding and the representative configuration, with the scPEE-S-QM/MM approach emerging as an efficient tool for describing bulk-solvated anions within the QM/MM framework. Its potential extends to improving transient anion state descriptions in biomolecules and radiosensitizers, especially given the frequent employment of microsolvation models.