The Effect of the Initial Guess Generator on Molecular Mechanics Calculations

Abstract

The effect of the initial guess generator on properties predicted by means of Molecular Mechanics has been investigated. We present a quantitative comparison of two generators, widely considered to be among the most efficient and advanced for the production of initial guesses used in the minimization of amorphous polymer structures. One of them (Polypack) is an excellent representative of and probably the most advanced generator available nowadays based on geometrical optimization strategies, whereas the other (End-Bridging Monte Carlo) has a strong statistical-mechanical footing and is based on one of the most powerful Monte Carlo schemes available for dense atomistic polymeric systems. A large number of initial guess structures for linear polyethylene ( C 100 ) in the united atom approximation were generated using Polypack and End-Bridging Monte Carlo. The potential energy, including torsional, non-bonded inter- and intramolecular energies, of these structures was subsequently minimized with respect to the torsional, single-chain positional (chain origin) and single-chain orientational (Euler angles) degrees of freedom. The predicted torsional angle distribution, inter- and intramolecular monomer pair distribution function, energy distribution and residual chemical potential were calculated for the minimized structures. Very significant differences in many of the computed properties were found between the structures generated using a geometrical optimization generator (Polypack) and those produced from a more physically founded generator (End-Bridging Monte Carlo). In addition, a reliable probability distribution function of total energy of minimized structures for PE has been obtained for the first time.