Uncertainties in the predictions of thermo-physical properties of thermoplastic polymers via molecular dynamics

Abstract

We quantify the effect of various sources of uncertainties in the prediction of thermo-physical properties of polymers using molecular dynamics simulations. We quantify how the choice of polymer builder, force field, molecular weight and data analysis affect predicted values of the glass transition temperature (Tg), room temperature density and coefficient of thermal expansion of poly(methyl-methacrylate) (PMMA) and polystyrene (PS). Interestingly, we find that the data analysis introduces significant uncertainties in Tg (approximately 5%) while the other properties are insensitive to it. The force field is the only variable that significantly affects the predictions of density. Polymer-consistent force field (PCFF) resulted in a higher density for PMMA than Dreiding and the opposite trend was observed in PS; in all cases the difference in density was less than 2%. Strongly correlated with density, we find that PCFF leads to a higher Tg than Dreiding for PMMA and both force fields predict similar Tg values for PS. The trends in Tg can be explained by differences in segmental mobility of the melts predicted by the two force fields. We find that the choice of amorphous polymer builder results in uncertainties in predictions comparable to those associated with the force field due to subtle, but persistent, differences in molecular structure. The results presented here provide insight into the physics behind molecular simulations of polymers and quantitative levels of uncertainties associated with individual sources that can help practitioners of molecular simulations interested in using their results in engineering applications.

Sim2Ls: FAIR simulation workflows and data

“…Importantly, this workflow should contain all the pre- and post- processing steps required to turn inputs into outputs. While these steps are often considered unimportant and poorly described in many publications, they can significantly affect results [ 29 ].…”

Uncertainty Quantification in Atomistic Modeling of Metals and Its Effect on Mesoscale and Continuum Modeling: A Review

“…Uncertainties in classical MD primarily occur for the following reasons: a) the choice of the interatomic potential for a given MD simulation; b) the choice of inputs outlined in Section 1 for DFT calculations of reference properties and the experimental data, which was used to fit the interatomic potential; (c) simplifications to the modeled material when compared to the experimentally characterized material; d) differences in the testing procedures between experiments and simulations; and e) data analysis technique [47]. Most studies show that the choice of the force field is the main factor that affects the predictions of material properties.…”

Universality in Spatio-Temporal High-Mobility Domains Across the Glass Transition from Bulk Polymers to Single Chains

“…The nonbonded long-range electrostatic interactions were described using the particle–particle, particle–mesh method with a precision of 10 –4 kcal/(mol Å) and the nonbonded van der Waals interactions are described using the Buckingham potential with an exponential repulsion and sixth power attraction with cutoff of 12 Å. Our prior work on both thermoplastic , and thermoset , polymers has shown that this force field provides an accurate description of a range of polymers. A timestep of 4 fs was used in all of the simulations using the reversible reference system propagator algorithm (RESPA) multitime step integration with an inner layer timestep of 0.5 fs.…”

“…We calculated the glass transition temperature of bulk systems from the density versus temperature curves using bilinear fit ( T g bilinear ) and hyperbola fit ( T g hyperbola ) to the data, as described in ref , see Figure a. This results in bilinear and hyperbola T g s of 453 and 479 K for PMMA, 442 and 456 K for PS, 246 and 275 K for PP, and 195 and 208 K for PE, respectively.…”