Skip to Main Content

Title ImagePublic Abstract

 
Collapse

DE-SC0021009: Modeling liquids, interfaces and nucleation with Local Molecular Field theory and Artificial Intelligence sampling methods

Award Status: Active
  • Institution: University of Maryland, College Park, MD
  • UEI: NPU8ULVAAS23
  • DUNS: 790934285
  • Most Recent Award Date: 06/22/2021
  • Number of Support Periods: 2
  • PM: Fiechtner, Gregory
  • Current Budget Period: 09/01/2021 - 08/31/2022
  • Current Project Period: 09/01/2020 - 08/31/2023
  • PI: Weeks, John
  • Supplement Budget Period: N/A
 

Public Abstract

In this proposal, we will develop and apply new theoretical and computational tools for the study of thermodynamics and kinetics of nucleation and solvation in nonuniform polar liquids, focusing on processes occurring at solid-liquid and liquid-vapor interfaces. These tools will combine insights from Local Molecular Field (LMF) theory for systems with strong Coulomb interactions [Gao, Remsing, Weeks, Proc. Natl. Acad. Sci. 117 1293 (2020)] with those from advanced sampling methods such as metadynamics [Tiwary, Parrinello, Phys. Rev. Lett. 111 230602 (2013)] and more recent Artificial Intelligence (AI) based sampling methods [Wang, Ribeiro, Tiwary, Nature Commun. 10 3573 (2019)]. The combined approach should provide both new physical insights into the complex physics occurring in the liquid, solid or vapor phases, and also permit even more efficient sampling simulations. We will first study crystal nucleation of molecular systems such as urea and calcium carbonate where naive use of classical nucleation theory has been found unsatisfactory [De Yoreo, Nature Mater. 12 284 (2013)]. We also propose to develop a general AI based framework that we believe can uncover some of the physically based insights and reduced representations used in LMF and related theories, and possibly reveal new currently unappreciated insights. More generally, we hope to develop appropriately modified AI algorithms that work in conjunction with intuitive general physical ideas, each suggesting qualitatively new directions, with AI carrying out most of the detailed quantitative connections and extensions on its own.



Scroll to top