| Molecular Dynamics (MD) simulations complement experimental methods by offering a computationally fast and accurate framework for determining the properties of materials. However, there is uncertainty associated with MD simulations in part due to the intrinsic microstructure of the material, but also due to the interatomic potential used to describe the material. To evaluate the uncertainty associated with MD predictions, this work exploits the strong dependency of MD on the interatomic potentials. Using a methodology based on surrogate models, a sensitivity analysis of the properties of graphene and aluminum when uncertainty is introduced to the parameters of their respective potentials is presented. These surrogates are utilized to identify which parameters in the potentials greatly affect the output parameters of interest. This methodology can be used to evaluate the robustness of interatomic potentials and accelerate the solution of the inverse problem to determine the parameters of the potential from observables. |
