Skip to Main Content

Title ImagePublic Abstract

 
Collapse

DE-SC0019441: Bridging the time scale in exascale computing of chemical systems

Award Status: Expired
  • Institution: Brown University, Providence, RI
  • UEI: E3FDXZ6TBHW3
  • DUNS: 001785542
  • Most Recent Award Date: 08/29/2022
  • Number of Support Periods: 4
  • PM: Holder, Aaron
  • Current Budget Period: 09/15/2022 - 09/14/2023
  • Current Project Period: 09/15/2018 - 09/14/2023
  • PI: Peterson, Andrew
  • Supplement Budget Period: N/A
 

Public Abstract


Large computational facilities and improved quantum-mechanical methods are allowing the atomic-scale simulation and design of materials and chemicals at length scales previously unimaginable. However, with an increase in length scale comes an increase in complexity, requiring very long simulation times for even the most basic of simulation tasks. Interestingly, such large simulations contain substantial redundant information, since similar chemical environments arise as time or distance increases. This opens the possibility of using emerging data-science and machine-learning techniques to learn from this information and predict the properties of such chemical environments while bypassing many of the expensive quantum-mechanics calculations. This research center will design machine-learning software to be compatible with leading quantum-mechanics based software, which is anticipated to provide orders-of-magnitude improvements in the time scales accessible to such atomic-scale simulations. To demonstrate the methodologies, we will focus on one of the most challenging systems related to a number of DOE-related technologies: the electrified solid-liquid interface. Understanding the phenomena of electron transfer from a solid in solution is of crucial importance to numerous emerging technologies including fuel cells, batteries, clean coal, and solar fuel synthesis. The machine-learning framework developed in this project will integrate information from various levels of quantum-mechanical theories across large numbers of simulations to provide unprecedented insight into this important phenomenon.


Scroll to top