Peter Cummings Automating Computational Materials Discovery through Model-Integrated Computing Professor Peter T. Cummings Department of Chemical and Biomolecular Engineering, Vanderbilt University, USA

Abstract: One of the key elements of President Obama’s manufacturing competitiveness initiative is the Materials Genome Initiative (MGI). Using high accuracy validated computational materials methods to predict and screen materials for specific properties, the MGI’s goal is to reduce by a factor of two both the time and the cost to bring new materials to the market. The development of computational materials screening tools for hard materials (metals, alloys, ceramics, etc) is relatively easy compared to the same task for soft materials (liquids, colloids, polymers, foams, gels, granular materials, and soft biomaterials), since in the former case the energy scales involved are large by comparison to thermal energy (~kBT where kB is Boltzmann’s constant and T is temperature) and the atoms/molecules are typically in crystalline lattice with no mesoscopic structuring. By contrast, for soft materials the energy scales are comparable to thermal energy, weak dispersion interactions frequently dominate, and soft matter often self-organizes into mesoscopic physical structures that are much larger than the microscopic scale yet much smaller than the macroscopic scale, and the macroscopic properties result from the mesoscopic structure. Consequently, to perform computational screening and design for soft materials involves complex molecular simulations, frequently on large petascale computing platforms and in some cases performed at multiple levels of detail. These simulations have complex workflows that experts understand and perform routinely; however, to translate these workflows into a set of tasks that can be automated, combined with other tasks and embedded within hierarchies of application-hardened engineering methodologies (such as stochastic optimization), we employ the discipline of model integrated computing (MIC). Using MIC, we develop a “meta-programming” tool that enables the establishment of a domain specific modeling language for the creation, synthesis and execution of simulation workflows. We illustrate this within two specific materials design problems: tethered nanoparticle systems and lubrication systems.