APPLICATION OF MACHINE LEARNING IN MOLECULAR STRUCTURE OPTIMIZATION AND NUDGED ELASTIC BAND PROCESS

• Main Author: Kristianto Wijaya, Vieri
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/54742
• Institution: Institut Teknologi Bandung

Material structure optimization and transition structure finding using Nudged Elastic Band (NEB) method are one of the works in material simulation research to study the properties of materials. Unfortunately, the computational cost of material structure optimization and NEB are very high. In this final project report, a machine learning model called gaussian process regression (GPR) was implemented in material structure optimization and NEB algorithm. The algorithm was tested on several material structures and reactions to investigate the results quality and computations duration from the algorithm. The testing shows that the implementation of GPR in material simulation algorithm produces the same quality compared to the results from the material simulation which was done without GPR. The result also shows that the implementation of GPR can accelerate the material simulation process by up to four times compared to the material simulation which was done without the implementation of GPR. Moreover, the application of GPR in NEB algorithm enables the execution of NEB simulation with more images without adding any significant computational costs. Keywords: material structure optimization, nudged elastic band, gaussian process regression.