Big data challenges in molecular dynamics: From force field development to trajectory post-processing

Speaker: Subramanian Sankaranarayanan (ANL)
Date: Wednesday, 2 December 2015, 17:00-18:30
Session: Molecular Dynamics II
Talk type: Extra talk (30 min)

Abstract: Nanomaterials exhibiting complex multi-property behavior and response have become important in key technological areas including solar conversion, energy storage, efficient electronics, medicine, and sensing. In functional devices employing these nanomaterials, it is critical to identify the structure-functionality correlation for an ensemble of nanoparticles. The factors determining such correlations vary over a wide range of length and time scales (intermolecular interactions at nanoscale to mesoscopic interfacial interactions and particle-particle interactions). This talk will focus on large-scale molecular dynamics simulations involving several million-atom systems to probe structure-functionality at the mesoscale as well as mesoscale phenomena that are of importance for materials synthesis and design. The rapid advances in computational power as well as the need to understand the full functionality of new materials with atomistic resolution has led to an exponential growth in the size and complexity of the data used to describe them. Using representative systems ranging from solvated thermo-sensitive polymer relevant to biomedical application to reactive complex oxides and tribological interfaces, this talk will focus on elucidating some key mesoscopic phenomena at material interfaces using multi million-atom atomistic simulations. This talk will highlight some of the computational challenges, the data-intensive nature of these calculations and the limitations of existing force-fields typically employed in molecular simulations. Finally, we will also present some of our on-going efforts on developing accurate reactive-force fields to bridge the gap between electronic structure and classical molecular dynamics simulations and describe such mesoscopic interfaces. We will discuss a stream-lined and data-driven approach to force field development using first principles density functional theory training data and machine learning algorithms. We will also discuss the validation of this approach on precious metal/oxide nanoparticles and their interfaces.