22 October 2024 in The Journal of Chemical Physics
We uncover the origin of unique electronic structures of single-atom alloys (SAAs) by interpretable deep learning. The approach integrates tight-binding moment theory with graph neural networks to accurately describe the local electronic structure of transition and noble metal sites upon perturbation. We emphasize the complex interplay of interatomic orbital coupling and on-site orbital resonance, which shapes the d-band characteristics of an active site, shedding light on the origin of free-atom-like d-states that are often observed in SAAs involving d10 metal hosts. This theory-infused neural network approach significantly enhances our understanding of the electronic properties of single-site catalytic materials beyond traditional theories.
27 April 2022 in Physical Review B
We study the role of electronegativity in sliding friction for five different two-dimensional (2D) monolayer systems using ab initio calculations within density functional theory with van der Waals corrections. We show that the friction depends strongly on the involved atoms’ electronegativity difference. All the studied systems exhibit almost the same magnitude of the friction force when sliding along the nonpolar path, independent of the material and the surface structures. In contrast, for sliding friction along a polar path, the friction force obeys a universal linear scaling law where the friction force is proportional to the electronegativity difference of its constituent atoms. We show that atomic dipoles in the 2D monolayers, induced by the electronegativity difference, enhance the corrugation of the charge distribution and increase the sliding barriers accordingly. Our studies reveal that electronegativity plays an important role in the friction of low-dimensional systems, and provides a strategy for designing nanoscale devices.
12 January 2026 in EdArXiv
Theoretical physical sciences-such as theoretical physics, chemistry, and materials science-feature deep abstraction hierarchies and long chains of logical dependency, making systematic learning reliant on prolonged engagement with multiple textbooks. While rigorous, such resources are largely non-personalized and inefficient for learners seeking rapid construction of a coherent physical picture. Recent advances in large language models (LLMs) offer new opportunities for educational support, but existing LLM-based educational applications-primarily centered on short-text reorganization and human-machine interaction-are most effective in domains with shallow abstraction hierarchies and are ill-suited to fields with densely interconnected conceptual structures. Grounded in theories of meaningful learning and concept mapping, we propose MyPrincipia, an LLM-based multi-agent framework for personalized first-principles knowledge explanation, particularly suited for systematic learning in theoretical physical sciences. The framework integrates an expert explainer agent with a metacognitive questioner agent that iteratively identifies unresolved conceptual ("what") and propositional ("why") gaps and generates targeted follow-up queries. Through controlled, round-based refinement with lightweight factual verification, the system reconstructs the underlying conceptual scaffold of a target learning topic and produces a personalized first-principles micro-textbook grounded in the learner's prior knowledge, accelerating the acquisition of coherent physical intuition. We welcome collaborations to further improve, extend, and apply this framework.