北京科学智能研究院（AI for Science Institute, Beijing）近日提出了一种新的AI+大尺度电子结构模拟方法，名为DeePTB，该方法利用深度学习技术，能够高效地表示具有从头算精度的材料电子结构。这一突破性进展极大地简化了计算复杂度，并实现了百万级大尺寸结构的电子、光电响应性质的计算模拟。
DeePTB方法通过训练基于物理约束的经验公式系数，提取成键原子对的局域化学环境，并通过机器学习框架自动拟合TB参数。这种方法不仅实现了精度与效率的统一，还能够处理自旋轨道耦合相互作用，使用更小的基组，并轻松接入大规模TB算法，真正实现了器件级尺寸的量子力学模拟。
该研究以“Deep learning tight-binding approach for large-scale electronic simulations at finite temperatures with ab initio accuracy”为题，发表在《Nature Communications》上。这项成果对于推动材料科学和相关领域的发展具有重要意义，同时也为人工智能在科学领域的应用提供了新的范例。
英语如下：
News Title: “AI Breakthrough: Simulation Speed of Millions of Atoms Boosted by a Hundredfold”
Keywords: Atomic Simulation, AI Method, Efficient Tight Binding
News Content: The Beijing Institute for Artificial Intelligence for Science (AI for Science Institute, Beijing) has recently proposed a new AI+ large-scale electronic structure simulation method called DeePTB. This method utilizes deep learning techniques to efficiently represent material electronic structures with ab initio precision. This breakthrough significantly simplifies computational complexity and achieves the simulation of electronic and photonic response properties for structures of millions of atoms.
The DeePTB method trains coefficients of empirical formulas based on physical constraints to extract the local chemical environment of bonded atom pairs and automatically fits TB parameters through a machine learning framework. This method not only achieves a unification of accuracy and efficiency but also handles spin-orbit coupling interactions, utilizes smaller basis sets, and easily integrates with large-scale TB algorithms, truly realizing quantum mechanical simulations at device-level dimensions.
This research, titled “Deep learning tight-binding approach for large-scale electronic simulations at finite temperatures with ab initio accuracy,” has been published in Nature Communications. This achievement is of great significance for advancing the development of materials science and related fields, and also provides a new paradigm for the application of artificial intelligence in the field of science.
【来源】https://www.jiqizhixin.com/articles/2024-08-15-7
Views: 2