NMR Relaxation Rates of Quadrupolar Aqueous Ions from Classical Molecular Dynamics Using Force-Field Specific Sternheimer Factors

• Published in: Journal of chemical theory and computation Vol. 17; no. 10; pp. 6006 - 6017
• Main Authors: Chubak, Iurii, Scalfi, Laura, Carof, Antoine, Rotenberg, Benjamin
• Format: Journal Article
• Language: English
• Published: Washington American Chemical Society 12.10.2021
• Subjects: Approximation; Calcium ions; Charge distribution; Chemical Sciences; Computation; Density functional theory; Dilution; Dynamics; Electric fields; Electron clouds; Errors; Estimates; Magnetic induction; Mathematical analysis; Molecular dynamics; NMR; Nuclear magnetic resonance; Parameterization; Quality assessment
• ISSN: 1549-9618; 1549-9626
• DOI: 10.1021/acs.jctc.1c00690

The nuclear magnetic resonance (NMR) relaxation of quadrupolar nuclei is governed by the electric field gradient (EFG) fluctuations at their position. In classical molecular dynamics (MD), the electron cloud contribution to the EFG can be included via the Sternheimer approximation, in which the full EFG at the nucleus that can be computed using quantum density functional theory (DFT) is considered to be proportional to that arising from the external, classical charge distribution. In this work, we systematically assess the quality of the Sternheimer approximation as well as the impact of the classical force field (FF) on the NMR relaxation rates of aqueous quadrupolar ions at infinite dilution. In particular, we compare the rates obtained using an ab initio parametrized polarizable FF, a recently developed empirical FF with scaled ionic charges and a simple empirical nonpolarizable FF with formal ionic charges. Surprisingly, all three FFs considered yield good values for the rates of smaller and less polarizable solutes (Li+, Na+, K+, Cl–), provided that a model-specific Sternheimer parametrization is employed. Yet, the polarizable and scaled charge FFs yield better estimates for divalent and more polarizable species (Mg2+, Ca2+, Cs+). We find that a linear relationship between the quantum and classical EFGs holds well in all of the cases considered; however, such an approximation often leads to quite large errors in the resulting EFG variance, which is directly proportional to the computed rate. We attempted to reduce the errors by including first order nonlinear corrections to the EFG, yet no clear improvement for the resulting variance has been found. The latter result indicates that more refined methods for determining the EFG at the ion position, in particular those that take into account the instantaneous atomic environment around an ion, might be necessary to systematically improve the NMR relaxation rate estimates in classical MD.