Topological Data Analysis in Materials Science: Principles, Machine Learning Integration, and Application Landscapes

Abstract

The data-driven paradigms are reshaping materials science, placing unprecedented emphasis on the digital characterization and representation of the material data. Topological Data Analysis (TDA) methods offer a promising and innovative approach to uncovering data patterns in complex material systems from a topological perspective. By focusing on topological invariants across multiscale resolution, TDA offers unique robustness, interpretability, and universality for the discovery of structure–property relationships. This review presents the mathematical foundations of TDA, with a focus on key methodologies such as persistent homology, persistent GLMY homology, and Euler characteristic curves. We further introduce the emerging field of topological learning, which integrates TDA with machine learning to enhance the development of predictive models. Representative applications of TDA are discussed in diverse material systems. In addition, we briefly highlight its emerging applications in related domains such as molecular science and biochemistry. Building on these examples, we summarize the unique advantages TDA offers in materials informatics, along with the critical challenges that remain. This review aims to provide a practical guide for materials scientists seeking to apply TDA in data-driven research, and to highlight its growing potential in advancing materials discovery and design.