“…Since the cost of different ab initio methods scales differently with the number ( N ) of basis functions (roughly N 4 for DFTs and N 5 for MP2) and the number of basis functions tested in this study spans a large range (between 26 for LANL2DZ and 155 for aug-cc-pVTZ), the cost of tested candidate methods varies dramatically. As previous AIMD studies have demonstrated, their results are extremely sensitive with respect to the method (e.g., MP2 vs DFT). − Further, since the computationally efficient method is selected based on optimized structures, while AIMD simulations traverse through unoptimized structure, it is possible that the computationally efficient method with the smallest RMSD performs poorly in AIMD. ,, Behaviors such as convergence failures, total energy jumps, and/or unphysical charge distributions have been previously reported for AIMD with DFTs. , Therefore, one DFT functional, the rCAM-B3LYP functional, and one MP2 method, FC-MP2, are recommended. For the former, rCAM-B3LYP functional reports the overall smallest RMSD with pc-2 (5.3 kJ/mol, 105 basis functions), followed by cc-pVDZ (5.8 kJ/mol, 55 basis functions) and Sapporo-DZP-2012 (6.0 kJ/mol, 80 basis functions).…”