TY - JOUR

T1 - PSO Method for Fitting Analytic Potential Energy Functions. Application to I-(H2O)

AU - Bhandari, H. N.

AU - Ma, X.

AU - Paul, A. K.

AU - Smith, P.

AU - Hase, W. L.

N1 - Publisher Copyright:
© 2018 American Chemical Society.
Copyright:
Copyright 2018 Elsevier B.V., All rights reserved.

PY - 2018/3/13

Y1 - 2018/3/13

N2 - In this work a particle swarm optimization (PSO) algorithm was used to fit an analytic potential energy function to I-(H2O) intermolecular potential energy curves calculated with DFT/B97-1 theory. The analytic function is a sum of two-body terms, each written as a generalized sum of Buckingham and Lennard-Jones terms with only six parameters. Two models were used to describe the two-body terms between I- and H2O: a three-site model H2O and a four-site model including a ghost atom. The fits are compared with those obtained with a genetic/nonlinear least-squares algorithm. The ghost atom model significantly improves the fitting accuracy for both algorithms. The PSO fits are significantly more accurate and much less time-consuming than those obtained with the genetic/nonlinear least-squares algorithm. Eight I- - -H2O potential energy curves, fit with the PSO algorithm for the three- and four-site models, have RMSE of 1.37 and 0.22 kcal/mol and compute times of ∼20 and ∼68 min, respectively. The PSO fit for the four-site model is quite adequate for determining densities of states and partition functions for I-(H2O)n clusters at high energies and temperatures, respectively. The PSO algorithm was also applied to the eight potential energy curves, with the four-site model, for a short time ∼8 min fitting. The RMSE was small, only 0.37 kcal/mol, showing the high efficiency of the PSO algorithm with retention of a good fitting accuracy. The PSO algorithm is a good choice for fitting analytic potential energy functions, and for the work presented here was able to find an adequate fit to an I-(H2O) analytic intermolecular potential with a small number of parameters.

AB - In this work a particle swarm optimization (PSO) algorithm was used to fit an analytic potential energy function to I-(H2O) intermolecular potential energy curves calculated with DFT/B97-1 theory. The analytic function is a sum of two-body terms, each written as a generalized sum of Buckingham and Lennard-Jones terms with only six parameters. Two models were used to describe the two-body terms between I- and H2O: a three-site model H2O and a four-site model including a ghost atom. The fits are compared with those obtained with a genetic/nonlinear least-squares algorithm. The ghost atom model significantly improves the fitting accuracy for both algorithms. The PSO fits are significantly more accurate and much less time-consuming than those obtained with the genetic/nonlinear least-squares algorithm. Eight I- - -H2O potential energy curves, fit with the PSO algorithm for the three- and four-site models, have RMSE of 1.37 and 0.22 kcal/mol and compute times of ∼20 and ∼68 min, respectively. The PSO fit for the four-site model is quite adequate for determining densities of states and partition functions for I-(H2O)n clusters at high energies and temperatures, respectively. The PSO algorithm was also applied to the eight potential energy curves, with the four-site model, for a short time ∼8 min fitting. The RMSE was small, only 0.37 kcal/mol, showing the high efficiency of the PSO algorithm with retention of a good fitting accuracy. The PSO algorithm is a good choice for fitting analytic potential energy functions, and for the work presented here was able to find an adequate fit to an I-(H2O) analytic intermolecular potential with a small number of parameters.

UR - http://www.scopus.com/inward/record.url?scp=85043983979&partnerID=8YFLogxK

U2 - 10.1021/acs.jctc.7b01122

DO - 10.1021/acs.jctc.7b01122

M3 - Article

C2 - 29345938

AN - SCOPUS:85043983979

VL - 14

SP - 1321

EP - 1332

JO - Journal of Chemical Theory and Computation

JF - Journal of Chemical Theory and Computation

SN - 1549-9618

IS - 3

ER -