Collaborative Research: Elements: Flexible & Open-Source Models for Materials and Devices

The project will develop first principles materials modeling software that can approach multiple length and time scales (multiscale). This software will be capable of modeling systems as complex as entire devices and materials of mesoscopic sizes. Over the course of the project the principal investigators plan to develop an open-source python-based software aimed at standardizing and generalizing multiscale simulations methods. This will enable the use of computer modeling in the design of new compounds, materials and devices. The goals are to render multiscale simulations reproducible and accessible by the broader community. In that context, the project will address the notion of 'lab 2.0', by which computer simulations replace laboratory experiments in tasks such as materials design and costly combinatorial searches for viable chemical processes. The software will be self-optimized using machine learning and exploit linear workflows approachable by nonexperts. Education and diversity will be promoted by direct participation of underrepresented minorities from high schools and colleges in hackathon workshops and summer research programs.

An approach that leverages the long-range multiscale capabilities of continuum models with accurate short-range atomistic descriptions of specific interactions, and that exploits the ideal scalability of quantum-embedding techniques, will be investigated. The main driver of the proposed implementation will be a Python codebase which will carry out the part of current software that is not computationally heavy, but instead is code heavy where many lines of code are needed in typically non-object-oriented languages. This is key to obtain the desired cluster-topology-agnostic workflows. Longstanding problems related to computational scalability and code stiffness will addressed in a three-pronged approach aimed at developing (1) modular tools implementing modules with highly object-oriented codes (e.g., quantum, classical atomistic, and continuum solvers), (2) hybrid tools implementing combinations of modular tools in a way that best exploits high-performance computing architectures, and (3) hyper tools implementing a high-level data-enabled optimization strategy that generates optimal workflows combining several hybrid tools, thereby making the software of broad applicability and accessible to nonexperts. These goals will render multiscale simulations reproducible and accessible by the broader community. The project will address the 'lab 2.0' paradigm, by which computer simulations replace laboratory experiments in tasks such as materials design and combinatorial searches for viable chemical processes. The resultant software will be self-optimized using machine learning and exploit linear workflows approachable by nonexperts. Education and diversity will include the direct participation of underrepresented minorities from high schools and colleges in hackathon workshops and summer research programs.

This award by the NSF Office of Advanced Cyberinfrastructure is jointly supported by the Division of Chemistry and the Division of Materials Research within the NSF Directorate of Mathematical and Physical Sciences.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.