Zoek in de site...

Theme 1 colloquium: 'Looking inside ‘Dynamic Solids’ using on-the-fly Machine Learning Force Fields: from large thermodynamic ensembles to experimental observables (Lecture)

Tuesday 4 October 2022Add to my calendar
from 16:00
Dr. Menno Bokdam (MESA+ Institute for Nanotechnology, UT)

Ab-initio based Molecular Dynamics (MD) is known for its high accuracy, but also for its unfeasible demand for computational power when going beyond one thousand atoms and picosecond time scales. In recent years, several important approaches have been developed that have enabled a very successful application of machine learning in this field.  These advances are so impressive that it could mean "the end of ab initio MD". I will briefly introduce these approaches and highlight the idea of active or on-the-fly learning that we have implemented in the VASP code. Next, I will focus on the potential that the Machine Learning Force Field method has for research in physics, chemistry and materials science. I will outline a few examples where we have simulated observables of complex crystals that would have been impossible using conventional methods. The common denominator characterizing these crystals is the wide variety of anharmonic lattice dynamics they exhibit. Therefore, large thermodynamic ensembles must be simulated with MD to obtain, for example: phase transition temperatures and melting points, (crystal moment resolved) vibrational spectra, X-ray spectra and thermal conductivity. In the last part of the lecture, I will present some very recent results from a collaboration with de Wijs and Kentgens at the IMM. In a combined theoretical and experimental effort, we are studying composition engineering of high efficiency perovskite solar cells. For example, MA or Br mixed in FAPbI3 results in desirable stability effects, but detailed knowledge of local structural features, such as local (dis)order or cation interactions of formamidinium (FA) and methylammonium (MA), is limited. I will present how large isothermal-isobaric ensembles can be used to calculate the 1H-1H dipolar coupling coefficients, which can be compared with those obtained by NMR spectroscopy.
We observe a qualitative and quantitative agreement between experiment and theory on several temperature depended experimental observables. This leads us to suspect that the simulated trajectories of crystal structure provide a realistic insight into the ionic dynamics of these perovskites.

dr. Gilles de Wijs