Vibrational and Structural Properties of Aqueous H ₂ SO ₄ and Na ₂ SO ₄ Systems from Ambient to Supercritical Conditions: A Comparative Study between GGA(-D3) and r2SCAN Functionals

Aqueous fluids containing sulfur-bearing species, such as sulfates and bisulfates, are important in several domains such as aqueous geochemistry, atmospheric chemistry, and medical sciences. For modeling their structural and thermodynamic properties, aside from empirical interatomic potentials, models developed using large data sets and machine learning have become extremely popular in recent years. Often, these data sets are produced by simulations using density-functional theory (DFT), and their quality depends on the approximations used, in particular the chosen exchange and correlation (XC) functional. Here, we determine the structural and vibrational properties of aqueous H 2 SO 4 and Na 2 SO 4 solutions from ambient to supercritical conditions from ab initio molecular dynamics using different XC functionals: BLYP, BLYP-D3, revPBE-D3 and r2SCAN. The latter three produce similar molecular and hydration shell structures of aqueous fluids from ambient to supercritical conditions. r2SCAN predicts vibrational frequencies closest to experimental data, but at significantly higher computational cost. Densities and self-diffusion coefficients at ambient conditions are largely consistent with experimental data except for the BLYP functional where larger deviations are observed. Overall, r2SCAN performs best, however, at largely increased computational cost, whereas both dispersion-corrected (D3) functionals offer a good balance between chemical accuracy and computational efficiency. Our results can serve as a guide for selecting the most suitable DFT method in large-scale data-driven projects for studying aqueous fluids from ambient to supercritical conditions.