The Department of Energy Office of Science announced that five research projects involving Berkeley Lab researchers have been selected for awards through the Funding for Accelerated, Inclusive Research (FAIR) initiative, which aims to build research capacity at institutions historically underrepresented in the office’s portfolio. The awards will support basic research projects across the Office of Science portfolio and are designed to help faculty at Minority Serving Institutions (MSIs) and Emerging Research Institutions (ERIs) foster mutually beneficial relationships with partner institutions, including DOE national labs.
The projects involving Berkeley Lab researchers range from developing machine learning and computational tools for scientific discovery to energy storage-related research. Berkeley Lab co-investigators involved in these projects are: Paolo Calafiura, Khaled Ibrahim, Andrew Nonaka, David Prendergast, and Wanli Yang. The projects, listed below, are expected to begin this fall and run for three years.
Investigating Large-Scale Models for High-Energy Physics Pattern Recognition
Investigators: Alina Lazar, Professor, Youngstown State University (PI) and Paolo Calafiura, Senior Scientist, Computing Sciences Area, Berkeley Lab (Co-investigator)
Pattern recognition from datasets of high-density particle 3D points is essential for the next generation of high energy physics experiments. However, the computational complexity of traditional tracking algorithms is challenging for large projects. Graph representations of data can be useful to capture complex relationships between data points. The proposed work is to develop algorithms that scale the training and inference of Graph Neural Networks (GNNs) for particle tracking in high energy physics applications, and potentially other applications such as scene reconstruction in self-driving car applications.
An Efficient Storage-Driven Machine Learning Model for Use with Multimodal Scientific Data
Investigators: Vijayalakshmi Saravanan, Assistant Professor, University of South Dakota (PI) and Khaled Ibrahim, Scientist, Computing Sciences Area, Berkeley Lab (Co-investigator)
The project aims at enhancing scientific workflows, which increasingly rely on machine learning (ML), simulation, and hybrid techniques to predict, understand, and optimize the behavior of complex systems in scientific inquiries. It will do so in high-performance computing environments, focusing on workflows reliant on diverse data modalities.
Machine-Learning-Based Surrogate Modeling for Stochastic Multiscale Simulation Methodology
Investigators: Changho Kim, Assistant Professor, UC Merced (PI) and Andrew Nonaka, Staff Scientist, Computing Sciences Area, Berkeley Lab (Co-investigator)
Complex interactions among different processes in multi-physics systems often lead to scientific advances and technological innovations, however, the disparate natures of these processes, particularly significantly different time and length scales, make computational modeling of such systems challenging. The proposed research aims to develop a machine-learning-based, stochastic multiscale simulation methodology for systems at mesoscopic scales (in-between microns and nanometers). Science applications include chemically-reacting microfluids and fluid-solid surface interactions (for example, for a gas-solid interfacial system where heterogeneous catalytic reactions for clean energy occur).
Super-resolution 3D Atomic Force Microscopy of Electrochemical Interfaces
Investigators: Tao Ye, Professor, UC Merced (PI); Ashlie Martini, Professor, UC Merced (co-PI), and David Prendergast, Senior Scientist, Energy Sciences Area, Berkeley Lab (Co-investigator)
The electrochemical processes of batteries, supercapacitors, and fuel cells predominantly take place at the interface between the electrode (electron conductor) and the electrolyte (ion conductor). This project aims to develop super-resolution 3D atomic force microscopy (AFM) capable of mapping the 3D arrangement of solvent molecules and ions under electrochemical control. The project team aims to combine AFM images of charged solid-liquid interfaces with molecular dynamics simulations to train a machine learning approach to efficiently associate measured features with specific molecular structure. Such advances are expected to aid the development of materials for electrochemical energy storage and conversion.
Emerging Properties through Controlled Phase Transformations for High Energy Sodium Ion Batteries
Investigators: Hui (Claire) Xiong, Professor, Boise State University (PI), and Wanli Yang, Senior Scientist, Energy Sciences Area, Berkeley Lab (Co-investigator)
This project targets both the fundamental understanding and practical material optimizations for achieving high energy-density and low-cost sodium ion batteries. The characterizations of the oxide based sodium ion battery materials focus on soft X-ray spectroscopy at the Advanced Light Source. The goal is to obtain a deeper mechanistic understanding of the chemical states at the high voltage range of the sodium ion battery electrode, which will support the development of material optimization guidelines for developing better cathodes for high-energy sodium ion batteries.
The FAIR initiative complements the office’s Reaching a New Energy Sciences Workforce (RENEW) initiative, which focuses on supporting training opportunities for undergraduate and graduate students at institutions underrepresented in its portfolio.
