Force Field Parametrization of Metal Ions from Statistical Learning Techniques

• DOI: 10.1021/acs.jctc.7b00779
• Publication Date: 2017-11-07T19:58:22Z
• Authors: Francesco Fracchia; Gianluca Del Frate; Giordano Mancini; Walter Rocchia; Vincenzo Barone
• Keywords: Parametrization (atmospheric modeling); Force Field

A novel statistical procedure has been developed to optimize the parameters of nonbonded force fields metal ions in soft matter. The criterion for optimization is minimization deviations from ab initio forces and energies calculated model systems. method exploits combination linear ridge regression cross-validation techniques with differential evolution algorithm. Wide freedom choice functional form allowed since both nonlinear can be optimized. In order maximize information content data employed fitting procedure, composition training set entrusted a combinatorial algorithm which maximizes dissimilarity included instances. methodology validated using field parametrization five (Zn2+, Ni2+, Mg2+, Ca2+, Na+) water as test cases.