Taming the Configurational Explosion Statistical Learning for Structure Search

Hörmann, L. (Speaker), Michael Scherbela (Contributor), Jeindl, A. (Contributor), Egger, A. (Contributor), Hofmann, O. (Contributor)

Activity: Talk or presentationTalk at conference or symposiumScience to science

Description

Monolayers of organic molecules on inorganic substrates often show rich polymorphism with diverse structures in differently shaped unit cells. Determining the different commensurate structures from first principles is far from trivial due to the large number of possible polymorphs. We pursue the SAMPLE approach[1][2], which is based on coarse-graining the potential energy surface and applying Gaussian Process Regression to efficiently find the energy of all polymorphs. To that end, we first determine adsorption geometries of isolated molecules as well as all possible unique substrate supercells. Then we combine each adsorption geometry with every super cell to generate all configurations. For a subset of these configurations, adsorption energies are determined using DFT. This subset serves as training data for the machine learning algorithm, that allows us to predict the adsorption energies for all polymorphs. Finally, all configurations with an adsorption energy below predefined energy threshold are reranked using DFT.
We demonstrate the capability of our approach for Naphtalene on Cu(111). We determine the adsorption energies for a large number of polymorphs and compare the results to the experimentally obtained phase diagram.
[1] Obersteiner, Hörmann, et. al., Nano Lett. 17 (7), pp 4453-4460
[2] Scherbela, Hörmann et. al., arXiv: 1709.05417
Period16 Mar 2018
Held atDPG-Frühjahrstagung 2018
Event typeConference
LocationStraße des 17. Juni 135 10623 Berlin, Germany
Degree of RecognitionInternational

Keywords

  • Advanced Materials Science