Solvent organization and QM–MM nonbonded interactions govern excitation energies of cationic Methylene Blue: A multiscale Dual-MD/ML and QM/MMPol benchmark

Methylene blue (MB) has been utilized as a therapeutic agent and has been in clinical and research use for more than a century. The cationic nature of MB + and the mutual polarization between solute and solvent constitute not just higher-order corrections but dominant contributions; accordingly, polarizable embedding approaches have been shown to substantially modify computed absorption characteristics of dyes in condensed phases, but it is still a challenge to reproduce the MB+ absorption spectrum. Here, we benchmark three multiscale simulation workflows, differing in force-field parametrization and configurational sampling strategy. In parallel, we evaluate five density-functional approximations to describe the electronic structure and to compute the associated absorption spectra of the solvated system. Consistent with experimental observations, methylene blue exhibits pronounced sensitivity to the local solvation environment, and the results indicate that the Dual-MD protocol is compatible with both conventional Born–Oppenheimer molecular dynamics (BOMD) and machine-learning (ML)-based potentials. Moreover, upon inclusion of explicit water molecules in the QM region, fully hybrid functionals yield excitation energies in excellent agreement with experiment.