Prediction of real tensile properties using extrapolations from atomistic simulations; An assessment on thermoplastic starch

Atomistic molecular dynamics (MD) simulations can be used to predict mechanical properties, such as stiffness and strength, for polymers. A concern is unavoidably high strain rates in simulations compared with those in physical experiments. To quantitatively capture the mechanical properties of the ‘real’ material, i.e., to predict absolute values rather than just qualitative trends, extrapolation to realistic strain rates is required. In this study, different strain-rate extrapolation strategies involving time-temperature shifting with the Williams-Landel-Ferry equation (above Tg) and the Eyring equation (below Tg) were evaluated, using thermoplastic starch as an example. MD simulations were first used to compute the stiffness and strength at three (high) strain rates over a wide range of temperatures. The mechanical MD data were then horizontally time-temperature shifted, resulting in master curves with strain-rate (x-axis) versus mechanical properties (y-axis). The precision in the prediction of experimental data was quite good in several cases, but was dependent on the extrapolation method and the specific thermoplastic starch system. A notable finding was that the simulations could be simplified using fewer simulation strain rates and temperatures. The extrapolation techniques used here are expected to be valid for other polymer systems, but this remains to be validated. •A methodology was developed to predict mechanical properties from MD simulations.•The methodology was tested on thermoplastic starch using superposition principles.•Temperature and strain-rate effects were predicted.•Extrapolation of simulated mechanical data were in good agreement with experiments.•Glass transition temperature from PVT could often be observed in mechanical trends.