The Center for Accelerating Materials Modeling

Team members: Jose Borreguero, Olivier Delaire, Monojoy Goswami, Mark Hagen (PI), Vickie Lynch, Andrei Savici,
Galen Shipman, and Bobby Sumpter

Great progress has been made in the last decade towards the predictive materials modeling paradigm that lies at the heart of our ability to solve grand scientific challenges. Theoretical and computational advances enable complex problems to be tackled, including the behavior of correlated systems (electron-electron, electron-phonon), nanoscale materials and assemblies, and functional systems without the simplifying characteristic of periodic long-range order. Similarly, experimental capabilities for the characterization of new materials have expanded in speed and resolving power, with a corresponding growth in the rate and volume of the data acquired. Free-electron lasers and next-generation neutron and synchrotron facilities are pushing the limits of the abilities of research teams to interpret experimental results for scientific impact. Research aimed specifically at meshing new capabilities in theory, data management and modeling, and information-intensive characterization is needed in order to harness effectively the power of individual advances to heighten the impact of materials-by-design.
The goal of this research is to break down remaining barriers to the synthesis and characterization of next-generation materials based on predictive theory, and the corresponding use of these results to enhance the accuracy and accessibility of model predictions.

Specifically, our goal is to improve the predictive capability of materials models and significantly accelerate the rate of scientific discovery from experimental data taken at the Spallation Neutron Source (SNS) by integrating modeling into all aspects of the experimental chain. The leading edge of predictive theory is often remote from research groups with expertise in the synthesis and characterization of novel materials, extending the time required to translate prediction to demonstration. Conversely, this separation also hinders the use of important new experimental results in the development and validation of predictive models. Research aimed toward truly closing the gap between prediction and realization is needed in order to gain the optimum advantages presented by developments in both theory and experiments.

To realize the promise of predictive modeling in advancing experimental research it is necessary to use models to predict directly the results of individual experiments, to compare model and experimental results in near-real time, and to establish clearly the relationships between changes in model assumptions or experimental basis and the resulting impact on predicted quantities. This project will establish the Center for Accelerating Materials Modeling (CAMM). CAMM researchers will develop the understanding of experimental methods, theoretical tools and model sensitivity needed to test and validate predictive materials models against experimental results. Platforms developed in this research for computational, data management, and communications capabilities will be made available to the broader scientific community through interactions with user-driven research at DOE-supported user facilities.

The direct use of predictive modeling to guide experimental studies will be pioneered by linking leadership-class computational capabilities with streaming experimental results from ongoing neutron scattering experiments at the Spallation Neutron Source, providing near-real-time feedback from theory and simulation to optimize the information obtained from individual experiments. The initial focus of this project has been the construction and demonstration of one of the core components of the software toolkit. Specifically, this component is the software to refine potential (or force field) models in molecular dynamics (MD) simulations against data from inelastic/quasi-elastic neutron scattering experiments. Ultimately, we will demonstrate this software for two example cases from "soft" and "hard" condensed matter domains, polyethylene oxide (PEO)-acrylic acid (AA) and KTa1-xNbxO3, respectively.

CAMM began work at the start of the first quarter FY2013. Since this time we have completed the design and initial development for version 1.0 of the analysis workflow involving the interaction of the refinement software package (Dakota[1]) with the codes for the molecular dynamics (MD) simulations and the calculation of the dynamic structure factor for the neutron scattering. Workflow management is based on the Kepler[2] scientific workflow management package (developed in research sponsored by the Office of Advanced Scientific Computing Research) to manage the workflow of submitting and monitoring the computing jobs for the various MD simulations and neutron scattering calculations. Initial tests have been successfully performed using Dakaota and Kepler together. For the science demonstration on polyethylene oxide-acrylic acid initial molecular dynamics simulations have been performed to determine the initial equilibration conditions of the model that will need to be performed as a precursor to the refinement against experimental data.