DEVELOPMENT OF INTERATOMIC POTENTIAL BASED ON MACHINE LEARNING (CASE STUDY: HYDRAZINE MOLECULAR SYSTEM)

• Main Author: Aliffanova Ardisa, Kevin
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/86890
• Institution: Institut Teknologi Bandung

Molecular dynamics (MD) simulations play a crucial role in exploring the properties of complex materials, where the accuracy in describing interatomic interactions is vital for producing reliable results. Empirical potentials offer computational efficiency but are limited in transferability, whereas ab initio methods, such as density functional theory (DFT), provide higher accuracy at the cost of significant computational expense. Machine learning potentials (MLP), such as the Deep Potential Smooth Edition (DeepPot-SE), present a scalable and accurate alternative. This study employs DeepPot-SE to develop an MLP model for hydrazine (N2H4) in MD simulations. The multi-body representation of the atomic environment in DeepPot-SE ensures high computational efficiency while maintaining accuracy. Training data derived from DFT calculations were used to train the model, achieving a root-mean-square deviation (RMSD) of 0.040 kcal/mol for energy and 0.079 kcal/mol/Å for forces, demonstrating accuracy comparable to DFT. Using DeepPot-SE, the machine learning potential model was successfully trained with energy and force data from hydrazine simulations. The resulting model exhibits excellent performance, with low RMSD values and high consistency between reference and predicted values, indicating its ability to accurately predict energies and forces across various atomic configurations. The training process demonstrated effective learning, as evidenced by the alignment of training and validation loss curves. This model provides an efficient and accurate approach for predicting molecular properties and holds potential for application to more complex molecular systems in the future. Keywords: Machine Learning Interatomic Potentials, Molecular Dynamics, Hydrazine.