Kang, Dong-Gi; (2024) Nanocluster Structure Prediction. Doctoral thesis (Ph.D), UCL (University College London).

Preview

Text Kang_10200864_Thesis.pdf Download (15MB) | Preview

Abstract

The atomic structures of nanoclusters are predicted using global optimisation techniques that aim to locate local minima (LM) on the potential energy landscape. This research evaluates the effectiveness of replacing density functional theory (DFT) with standard interatomic potentials (IP), variable charge potentials, machine learning potentials (ML-IP: GAP), and neural network potentials (NN-IP: MACE). It also assesses the newly published “universal potentials,” which are a complete set of MACE potentials trained on data from the Materials Project Database, applicable to all elements. Monte Carlo Deterministic Quenching and an Evolutionary Algorithm are employed to find IP LM of (AlF3)n for fixed values of n, from 1 to 11, and bulk cuts for n = 27 and 64. Parameters (A = 3760.0008, ρ = 0.2220) for the Al3+-F− Born-Mayer potential were fitted to reproduce the AlF3 α-bulk phase, whilst keeping fixed F−-F− Buckingham potential parameters, which were taken from earlier work on lattice and intrinsic defect properties of bulk rare-earth fluorides. (AlF3)6 is the smallest LM nanocluster composed of octahedral corner-sharing secondary building units, a structural feature consistent with bulk AlF3 phases. Aluminium hydride nanoclusters were generated by data-mining the aluminium fluoride LM. Notably, aluminium hydride exhibits analogous configurations to aluminium fluoride and demonstrate comparable stability. To analyse the structures in terms of primary or secondary building units requires the ability to first determine their coordination numbers. A method for defining coordination numbers was developed and applied, although in many cases this proved problematic. The third system was brought to my attention by colleagues within the department: 25 gold atoms capped by 18 L-cysteine ligands, i.e., nanoclusters of Au25(Cys)18. An earlier study deduced from the XRD observations a more spherical Au25 core as compared to the new data recently obtained. Here, a model for this system was developed by first applying global optimisation to determine the atomic structure of naked Au25, to which the L-cysteine ligands were added.

Download activity - last month

Export as XMLCSVJSON

Download activity - last 12 months

Export as JSONXMLCSV

Downloads by country - last 12 months

Export as XMLCSVJSON

Archive Staff Only

View Item