BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Crystal Structure Search with Random Relaxations Using Graph Netwo
 rks - Gowoon Cheon\, Stanford
DTSTART:20210208T163000Z
DTEND:20210208T170000Z
UID:TALK157081@talks.cam.ac.uk
CONTACT:Bingqing Cheng
DESCRIPTION:Materials design enables technologies critical to humanity. Wh
 ile many properties of a material are determined by its atomic crystal str
 ucture\, prediction of the atomic crystal structure for a given material's
  chemical formula is a long-standing grand challenge that remains a barrie
 r in materials design. We build a novel dataset of more than 100\,000 rand
 om structure relaxations of battery anode materials using high-throughput 
 density functional theory (DFT) calculations\, which includes calculations
  with varying quantum mechanical settings for out-of-domain generalization
 . We modify graph neural network force fields to also predict stress infor
 mation\, which allows them to effectively simulate relaxations. We show th
 at models trained on data conventionally used to train interatomic potenti
 als fail to simulate relaxations from random structures\, and random struc
 ture relaxations data is crucial for crystal structure search. We find tha
 t models trained with data augmentation via random perturbations improves 
 both the accuracy and out of domain generalization\, and is able to find a
 n experimentally verified structure of a new stoichiometry.\n
LOCATION:virtual ZOOM meeting ID: 263 591 6003\, Passcode: 000042\, https:
 //us02web.zoom.us/j/2635916003?pwd=ZlBEQnRENGwxNmJGMENGMWxjak5nUT09
END:VEVENT
END:VCALENDAR