Hybrid DFT-ML-MD Approach for Derivation of Lennard-Jones Interatomic Potential Parameters of Al

Atomistic simulation based on computational physics of methods is used to develop accurate interatomic potentials based on DFT (density functional theory) data. The accuracy of predicting the physical properties of a material is highly dependent on the quality of the interatomic potential used. The purpose of this study is to determine the Lennard-Jones potential parameters of Al metal (epsilon and sigma) from fitting the DFT simulation output data. The use of a “robust” fitting method to reduce the influence of outliers on the potential results is very important and therefore a machine learning method is used to help find the right potential parameters. The method used is a hybrid method using DFT to generate training data, using ML (machine learning) to fit DFT data to the Lennard-Jones (LJ) potential model, and using the MD (molecular dynamics) method to validate the LJ potential parameters. Python-based programming is applied to facilitate how the three methods can be connected. The results of this study are that Al metal has an epsilon value = 0.5000 eV and sigma Al = 3.2072 Å, with a regression coefficient R2 = 0.9441 so that it can be concluded that this study can be said to be quite good and the hybrid method can be further developed to obtain the LJ potential parameter values of various other materials, especially metals.