Development of neural network potential for Al-based alloys containing vacancy

Potential energy of an alloy is essential indicator for evaluating the stability structure in predicting new materials. Therefore, how to calculate potential material design has become inevitable problem. While first-principles calculations can provide chemical accuracy arbitrary atomic arrangements, they are prohibitive terms computational effort and time. To enable atomistic-level simulations both processing performance Aluminum alloys, neural network was proposed predict binding vacancy-containing aluminum alloys a highly accurate state. This method combined machine learning techniques explore intrinsic link between solid solution energies. In this study, four binary (aluminum-silicon, aluminum- zirconium, aluminum-magnesium aluminum-titanium alloys) were investigated. The mean squared errors used quantify quality models it found that trained model more stable exhibits high prediction. Monte Carlo simulation results show using successfully simulated aging process be much faster than calculations, even with accuracy.