Master's Internship 2023

Master’s internship project under the supervision of Dr. Julien Lam. The report can be seen here.

Abstract

Observing nucleation in simulations remains challenging due to the long time needed to cross the free energy barrier required to form a nucleus. Furthermore, the structural analysis necessary to understand the process involved in nucleation can be complicated to perform in materials with multiple available structures. In this work, we present a new classification methodology for structural analysis and apply it to the study of melting and crystallization in Zinc Oxide (ZnO), using also the seeding technique to estimate the critical temperature of body-centered tetragonal (BCT) and wurtzite (WRZ) crystalline clusters in nanoparticles. In our classification methodology, we make use of the average Voronoi-weighted Steinhardt parameters as descriptors and the Gaussian Mixture Model (GMM) with the Maximum Likelihood Classifier (MLC) as a machine learning classification method. We show that our classification method makes physically meaningful structural predictions by performing multiple tests on ZnO structures. In our study of the melting of BCT and WRZ bulk crystals, the results support the hypothesis of a two-step transition between the WRZ and liquid phases. For the study of nucleation, we perform the seeding technique by inserting a crystal seed into a liquid droplet and analyzing changes in size at different temperatures. We found that the critical temperature for BCT and WRZ clusters is similar for approximate sizes of 180 and 360 atoms. For smaller clusters, around 70 atoms, we found a significantly higher critical temperature for the BCT cluster. By performing structural analysis in the crystalline clusters we observe a transition of the BCT cluster to a WRZ cluster. On the other hand, the WRZ cluster is commonly surrounded by BCT and HBN atoms. Our results support the hypothesis of the two-step transition between the WRZ and the liquid phases in the melting and crystallization processes. Furthermore, we shed light on the crystallization of ZnO nanoparticles and put into question the stability of BCT crystalline clusters.