BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Chemical transferability and accuracy of machine learning interato
 mic potentials for ionic liquid and battery solvent simulations - Zachary 
 Goodwin\, Harvard University
DTSTART:20240603T133000Z
DTEND:20240603T140000Z
UID:TALK217525@talks.cam.ac.uk
CONTACT:Eszter Varga-Umbrich
DESCRIPTION:Electrolytes\, such as those based on ionic liquids (ILs) or c
 arbonate solvents\, are typically simulated using classical force fields b
 ecause they are computational cheap enough to reach the timescales and sys
 tem sizes required. These force fields are often not quantitatively accura
 te when comparing to experiments\, however\, which motivates the developme
 nt of machine learning interatomic potentials (MLIPs) to obtain more relia
 ble results. However\, MLIPs are often only trained and deployed on a sing
 le composition\, but electrolytes are typically complex mixtures\, and we 
 want to study their physiochemical properties as a function of salt concen
 tration\, different solvents and additives\, for example. We demonstrate a
  method to train MLIPs to be chemically transferable\, using a salt-in-ion
 ic liquid as a test case. We find that to learn the energy to be transfera
 ble\, a minimum number of compositions are required\, and we generalise th
 is result to provide guidelines for investigating more complex systems [1]
 . Simulating electrolytes is also typically performed in the NPT ensemble\
 , but it has been found that training MLIPs to be stable in the NPT ensemb
 le can be difficult. We investigated this problem for various systems\, in
 cluding ILs and carbonates solvents\, and found that Allegro could be made
  to be stable for these systems\, with the chosen hyperparameters and data
 sets being key factors in determining their stability. \n[1] - Goodwin et 
 al. arXiv:2403.01980 (2024)
LOCATION:Lecture Theatre 6 at the Department of Engineering\, and Zoom lin
 k: https://zoom.us/j/92447982065?pwd=RkhaYkM5VTZPZ3pYSHptUXlRSkppQT09
END:VEVENT
END:VCALENDAR