BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:𝝙-Machine Learning for Molecular Crystal Structure Prediction -
  Simon Wengert\, TU Munich
DTSTART:20210208T170000Z
DTEND:20210208T173000Z
UID:TALK157162@talks.cam.ac.uk
CONTACT:Bingqing Cheng
DESCRIPTION:The combination of modern machine learning (ML) approaches wit
 h high-quality data from quantum mechanical (QM) calculations can yield mo
 dels with an unrivalled accuracy/cost ratio. However\, such methods are ul
 timately limited by the computational effort required to produce the refer
 ence data. In particular\, reference calculations for periodic systems wit
 h many atoms can become prohibitively expensive for higher levels of theor
 y. This trade-off is critical in the context of organic crystal structure 
 prediction (CSP). Here\, a data-efficient ML approach would be highly desi
 rable\, since screening a huge space of possible polymorphs in a narrow en
 ergy range requires the assessment of a large number of trial structures w
 ith high accuracy.\nIn my talk\, I will present a workflow for the generat
 ion of tailored 𝝙-ML models that allow screening a wide range of crysta
 l candidates while adequately describing the subtle interplay between inte
 rmolecular interactions such as H-bonding and many-body dispersion effects
 . This is achieved by enhancing a physics-based description of long-range 
 interactions at the density functional tight binding (DFTB) level---for wh
 ich an efficient implementation is available---with a short-range ML model
  trained on high-quality first-principles reference data. The presented wo
 rkflow is broadly applicable to different molecular materials\, without th
 e need for a single periodic calculation at the reference level of theory.
  I will show that this even allows the use of wavefunction methods in CSP.
LOCATION:virtual ZOOM meeting ID: 263 591 6003\, Passcode: 000042\, https:
 //us02web.zoom.us/j/2635916003?pwd=ZlBEQnRENGwxNmJGMENGMWxjak5nUT09
END:VEVENT
END:VCALENDAR