Emerging Frontiers in Computational Physics: From DFT to AIAssisted Modeling

Abstract

Computational physics has traversed a remarkable journey over the past five decades, evolving from rudimentary numerical methods to highly sophisticated quantum-mechanical simulations. The discipline today occupies a central position in both fundamental research and applied materials science. Density functional theory (DFT) provided the first practical framework for solving the many-body electron problem with acceptable computational cost, and its widespread adoption transformed how researchers study electronic structure, bonding, and material properties. In recent years, machine learning and artificial intelligence have begun reshaping this landscape once more, enabling models that combine the accuracy of quantum mechanics with computational speeds that were previously unimaginable. This review surveys the historical development of computational physics methods, examines landmark contributions that defined each era, and assesses the current state of AI-assisted modeling. Particular attention is paid to the integration of neural network interatomic potentials, generative models for materials discovery, and physics-informed neural networks. Challenges associated with transferability, interpretability, and data quality are discussed,along with an assessment of the most promising directions for future research.