Thermal Conductivity Prediction Achieves High Accuracy with Topological Descriptors in Amorphous Graphene
Summarize this article with:
Predicting how heat flows through disordered materials presents a longstanding challenge for scientists, as the lack of regular structure complicates analysis, but a new study offers a powerful solution. Kosuke Yamazaki, Takuma Shiga, and Kumpei Shiraishi, alongside their colleagues at The University of Osaka and Toyota Technological Institute, demonstrate that a technique from topology, called persistent homology, accurately predicts thermal conductivity in amorphous graphene.
The team reveals that this method not only forecasts how well heat travels through the material, but also identifies specific structural features, distorted hexagonal and triangular motifs, that hinder thermal transport, linking microscopic structure directly to macroscopic properties. This achievement establishes a framework for interpretable machine learning in materials science, offering a way to understand and predict material behaviour with unprecedented clarity. This approach not only predicts how well heat travels through the material, but also identifies specific structural features that hinder thermal transport, linking microscopic structure directly to macroscopic properties. This achievement establishes a framework for interpretable machine learning in materials science, offering a way to understand and predict material behaviour with unprecedented clarity.,.
Predicting Thermal Conductivity via Topological Data Analysis The researchers employed molecular dynamics simulations to generate a diverse set of amorphous graphene structures. Initial configurations consisted of carbon atoms arranged in a square lattice, which were then melted and quenched using a controlled cooling process. By varying the cooling rate, they created materials with different levels of structural disorder. To quantify these structural features, the team implemented persistent homology, a mathematical technique that identifies “holes” and “voids” within the atomic configurations. This involved virtually expanding spheres around each atom and tracking how these spheres connect, forming edges and loops. The birth and death radii of these topological features were recorded, generating a persistence diagram that compactly encodes structural information across multiple scales. This enabled the extraction of quantitative structural descriptors, overcoming limitations of conventional methods that struggle to capture medium-range order.,.
Cooling Rate Dictates Amorphous Graphene Conductivity The study reveals a strong dependence of thermal conductivity on the cooling rate during sample preparation. Slower cooling rates yield higher thermal conductivity, while faster rates lead to a significant decrease, eventually converging to a relatively low value. This confirms that increased structural disorder, induced by rapid cooling, suppresses heat transport, aligning with previous observations in the field. The researchers then constructed a machine learning model, specifically ridge regression, using these persistent homology descriptors to predict thermal conductivity.,.
Persistent Homology Predicts Amorphous Graphene Conductivity The researchers achieved a high degree of accuracy in predicting thermal conductivity using persistent homology-based descriptors. The model, trained and tested on independent datasets, demonstrated a strong correlation between predicted and measured values. Further analysis revealed that distorted hexagonal and triangular motifs significantly reduce thermal conductivity, a finding supported by the spatial distribution of localized vibrational modes which overlapped with these structures. This innovative methodology not only achieves accurate prediction of thermal conductivity but also provides physically interpretable insights into structure-property relationships in amorphous materials.,.
Uncovering Structural Contributions to Thermal Transport The inverse analysis technique proved particularly insightful, revealing that specific topological features correspond to structural motifs that either enhance or suppress heat conduction. Structures associated with higher thermal conductivity were found to be near-perfect hexagons and triangles, while those reducing conductivity were distorted versions of these shapes. Histograms of bond lengths within ring structures showed sharper peaks for structures enhancing conductivity and broader distributions for those suppressing it, confirming the link between order and thermal transport. These findings establish persistent homology as a promising framework for uncovering meaningful structure-property relationships within disordered materials.,. Future Directions and Implications This research demonstrates that persistent homology effectively describes and predicts thermal conductivity in amorphous graphene. The authors acknowledge that their analysis relies on classical molecular dynamics simulations, which may be influenced by the chosen interatomic potentials and simulation parameters. They also note that real-world amorphous graphene may exhibit a broader range of structural characteristics than those studied. Future work should therefore explore more realistic structure-generation methods, such as those employing machine-learned potentials, and validate these findings through comparison with experimental data. The approach offers potential for application to a wider range of amorphous and nanostructured systems, and could be further enhanced through integration with more complex machine-learning models. 👉 More information 🗞 Topological descriptor for interpretable thermal transport prediction in amorphous graphene 🧠 ArXiv: https://arxiv.org/abs/2512.13112 Tags: