Seminar: Crystal Structure Prediction for Covalent and Molecular Crys...

Seminar: Crystal Structure Prediction for Covalent and Molecular Crys...

LISTEN

Speaker: Adem Tekin, Istanbul Technical University

Title: Crystal Structure Prediction for Covalent and Molecular Crystals

Date/Time: October 2, 2019  /  13.40-14.30

Place: FENS G032

Abstract: Predicting crystal structure of a material, especialy if it now known experimentally, is one of the most important problems in computational material sciences. There are several computational methods developed for solving crystal structure prediction (CSP) problems. The fastest approach in CSP is to employ force fields, however their reliability remains a mojor issue to be addressed. There are also CSP approaches performing only local cell optimizations. Recently, Machine Learning (ML) techniques were also applied to CSP. The another and widely used approach in CSP is to use global optimization techniques such as simualted annealing (SA), evolutionary algorithms and particle swarm optimization (PSO).

Recently we developed a CSP algorithm, called as CrsytAl Structure PrEdiction via Simulated Annealing (CASPESA), to predict the crystal structures of covalent crystals such as hydrogen storage and li-ion battery materials. The main motivation of CASPESA is to globally optimize the number of peculiar interactions that might lead to lower the energy of the crystal structure. We extended the capability of CASPESA to handle molecular crystals (drug, organic electronic, dye, pigment and high energy materials) together with enhancing the strategy used in the global optimization. In particular, various modifications (such as assignment of individual step lengths for each parameter) in the SA part have been implemented, a PSO step has been placed before the SA part and the whole algorithm (PSO + SA) has been parallelized to be able to perform much deeeper global searches. Since the resulting algorithm differs from the CASPESA approach in a large extent, the new platform is called as Parallel Crystal Structure Predictor (PCSP). Moreover, PCSP is assisted with ML to analyze the resulting crystal structures.

In this talk, several applications of CASPESA and PCSP in covalent and molecular crystalline systems will be discussed.

BIO: Following BS and MSc studies in the chemistry department of Istanbul Technical University, Dr. Adem Tekin started his PhD in Theoretical Chemistry in Stuttgart (Germany, 2001) in the group of Prof. Bernd Hartke and then finished in Physical Chemistry in Kiel (Germany, 2004). After stays in Duisburg-Essen (Germany, 2005-2008) and DTU Physics (Danmark, 2008-2010) as a postdoc he joined to the informatics institute of Istanbul Technical University (Turkey, 2010) as an Assistant Prof.. His primary interests are global optimizations, intermolecular interactions, force-field development, crystal structure prediction, energy materials and computational materials design.

Contact: Emre Erdem