New method simulates large atomic structures effectively

A new research study has introduced a computational method that could change how scientists simulate large atomic structures. This method uses machine learning to create an effective Hamiltonian, which allows for the simulation of much larger structures than traditional quantum or classical approaches. The findings were published in the journal npj Computational Materials. An international team of physicists from the University of Arkansas, Nanjing University, and the University of Luxembourg authored the paper. The study focuses on "mesoscopic structures," which consist of millions of atoms and are often used in materials like ferroelectrics and dielectrics. Current methods struggle with these large structures, but the effective Hamiltonian method can easily handle them. This new approach is one of the fastest for atomic-scale simulations and is expected to be a valuable tool for researching these complex materials. Traditionally, obtaining the parameters needed for the effective Hamiltonian is a complicated process. The new paper proposes a machine-learning technique that simplifies this, providing a more universal and automatic way to calculate the necessary parameters for various complex systems. With this improved method, scientists can design materials with specific qualities, like ferroelectric and piezoelectric properties, through large-scale atomic simulations. Future work will aim to create a general effective Hamiltonian based on lattice structures, which could help simulate changes in structure and temperature effects. For more details, you can refer to the published article by Xingyue Ma et al in npj Computational Materials.