Physics

AI predicts quantum material properties using drastically less computational power

How the science connects

Deep learningQuantum materialsDynamical mean-fie…

AI Insight

This study presents a deep learning approach that can predict self-energies in materials using ab initio dynamical mean-field theory (DMFT) calculations while requiring only minimal training data sets. The researchers demonstrate that their neural network model can accurately reproduce complex quantum many-body effects in real materials, significantly reducing the computational cost typically associated with DMFT calculations. This method bridges machine learning and quantum many-body physics to accelerate materials property predictions.


The technique could dramatically speed up materials discovery and design by reducing the computational resources needed for accurate electronic structure calculations. This advancement may enable researchers to screen larger numbers of candidate materials for applications in electronics, energy storage, and quantum technologies without requiring expensive supercomputing resources for each calculation.


Understand the Science

Source: Deep learning-based prediction of self-energies from ab initio dynamical mean-field theory for real materials with minimal data sets