Markos Katsoulakis and Matthew Dobson, along with students, postdocs, and a team of researchers from Brown University, University of Delaware, University of North Carolina-Chapel Hill, and University of California San Diego recently received a $3.1 million award from U.S. Defense Advanced Research Projects Agency (DARPA). This grant focuses on developing a predictive modeling framework for the reliable computational design of novel, superior, and/or lower cost materials for applications to catalysis, energy production, and energy storage.

To date there is strong experimental evidence that novel material architectures can provide unprecedented performance. Therefore a reliable computational framework for the prediction of such materials can be a critical ingredient for decreasing development cost and time-to-market. In view of our ever-increasing computational capabilities, such a computational framework is already emerging. However, despite their sophistication, computational models are not always predictive. This can be attributed to both model uncertainties as well as the vast spatiotemporal discrepancies between the scales at which we need to make predictions and the scales at which reliable simulations are feasible. In particular, although Uncertainty Quantification (UQ) for this broad class of problems is still at an embryonic stage, nevertheless, because it is intimately linked to reliable computational predictions, UQ is a critical component for guiding the synthesis of materials and experimental assessment of model predictions. The goal of the new grant is to bridge this scientific and ultimately technological gap via the interdisciplinary expertise and collaborations of the participating teams and their researchers. From a mathematical perspective, this research represents a new direction for applied and computational mathematics, bringing together applied probability, uncertainty quantification, high-performance computing, and data science.

The team at UMass Amherst leads the research thrust on developing scalable algorithms and the enabling mathematical methods for uncertainty quantification and predictive computing for multi-scale, multi-physics systems with from hundreds to millions of uncertain parameters, often correlated due to underlying physicochemical principles and constraints. Besides applications to energy and materials, such challenges are typically encountered in many chemical and biological processes, geosciences, and epidemiology as well as in various types of complex networks. Furthermore, the UMass team will develop fast uncertainty-quantification and sensitivity-screening methods for extreme and rare events and will study how the overall behavior of complex, stochastic systems such as network dynamics is ultimately determined by the occurrence of rare events. Additional background on this research project is given at http://www.supercomputingonline.com/latest/58851-katsoulakis-develops-ne....