BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Europe/Stockholm
X-LIC-LOCATION:Europe/Stockholm
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20210916T132448Z
LOCATION:Michel Mayor
DTSTART;TZID=Europe/Stockholm:20210705T160000
DTEND;TZID=Europe/Stockholm:20210705T163000
UID:submissions.pasc-conference.org_PASC21_sess119_msa277@linklings.com
SUMMARY:Energy Functionals of Atom-Based Electron Populations: An Approach
  to Robust ML PESs
DESCRIPTION:Minisymposium\n\nEnergy Functionals of Atom-Based Electron Pop
 ulations: An Approach to Robust ML PESs\n\nXie\n\nML PESs have shown great
  promise in reducing the computational cost of DFT calculations, however, 
 most methods are not suitable for describing systems with variable electro
 nic structure or where long-range interactions are essential, e.g. molecul
 es with different charge states. In order to solve this problem, we are in
 terested in developing ML energy functionals of atom-based electron popula
 tions. The first series of this is the BpopNN method, where we used neural
  networks to map atomic species, coordinates and populations to the DFT en
 ergies. Optimizing the model energy with respect to the electron populatio
 ns self-consistently naturally allows for the incorporation of long-range 
 environmental effects, similar to the SCF calculations in DFT.  The e
 ffectiveness of this approach has been tested on a series of LinHn cl
 usters of total charges -1, 0, 1. We will also discuss improvements to the
  current model, e.g. how to incorporate long-range electrostatics, and ext
 ending the application to Li-organic systems.\n\nDomain: Chemistry and Mat
 erials
END:VEVENT
END:VCALENDAR