Uncertainties of Parameters to Predictions of Polymer Properties by Molecular Simulations

Abstract

We quantify uncertainties in the prediction of thermos-mechanical properties of polymers using atomistic simulations. We quantify uncertainties with respect to the choice of force field used to describe atomic interactions, chain molecular weight in the samples, and data analysis method in the prediction of glass transition temperature Tg, density and thermal expansion coefficients (CTE) of Poly(methyl methacrylate), PMMA. The PolymerModeler at nanoHUB was used to build amorphous samples with two different chain lengths and annealing MD simulations where performed two widely used force fields DREIDING and PCFF. In order to extract the properties of interest from the raw simulation data we use two methods: i) a bi-linear fits to the high and low temperature regions, extracting the Tg as the intersection between the fits and the CTE’s from the slopes of each line and, ii) a hyperbola to fit the entire range of data and Tg is obtained as the intersection of the two asymptotes of the resulting hyperbole, each asymptote slope gives the value of the thermal expansion coefficient. We find that our results give accurate predictions, glass transition temperature is in good agreement once the correction for experimental data is considered; a significant Tg at higher temperatures is found with the pcff force field compared with Dreiding, as well for densities at room temperature. The analysis method used was found to influence the prediction of CTE’s, but also the estimate of Tg.