一个机器学习原子间潜能数据集和模型用于催化,具有局部微调到化学精度的微调
Zhihong Wu1, Lei Zhou1, Pengfei Hou1
1Center for Rare Earth and Inorganic Functional Materials, School of Materials Science and Engineering & National Institute for Advanced Materials, Nankai University, Tianjin 300350, China.
JACS Au
|December 26, 2025
概括
我们开发了催化大原子模型 (CLAM),这是一种用于复杂催化的机器学习工具. CLAM准确地预测了催化反应,并加速了模拟,超过了传统方法.
科学领域:
- 计算化学计算化学
- 材料科学 材料科学 材料科学
- 化学工程是化学工程的重要组成部分.
背景情况:
- 异质催化包括具有动态催化剂变化的复杂反应.
- 传统的密度函数理论 (DFT) 方法与这些复杂性作斗争.
研究的目的:
- 介绍催化大原子模型 (CLAM) 用于机器学习异质催化中的原子间潜力.
- 提高催化模拟的准确性和效率.
主要方法:
- 在多样化的数据集上训练CLAM,包括金属/合金板,氧化物,集群,2D材料和小分子.
- 开发了一个"局部微调"算法,以改善ML的原子间潜力,用于结构优化和过渡状态搜索.
- 利用分子动力学模拟来评估CLAM复制动态催化现象的能力.
主要成果:
- 在化学准确度值内,在过渡金属表面上吸附能量的预测准确度达到了94%.
- 与DFT相比,演示了3.4倍的计算加速.
- 在过渡状态搜索中显示了81%的准确性,与基于DFT的CI-NEB相比,速度提高了10.1倍.
- 在没有额外的微调的情况下,使用分子动力学模拟成功复制了动态催化现象.
结论:
- 对于异质催化剂,CLAM提供了一个高度准确和高效的机器学习方法.
- "局部微调"算法显著提高了ML原子间潜力的预测能力.
- 在加速催化剂发现和理解动态催化过程方面,CLAM显示出有前途的潜力.
相关概念视频
Molecular Models
43.3K
Physical models representing molecular architectures of chemical compounds play essential roles in understanding chemistry. The use of molecular models makes it easier to visualize the structures and shapes of atoms and molecules.
43.3K
Ligand Binding Sites
14.8K
Proteins are dynamic macromolecules that carry out a wide variety of essential processes; however, the activities of most proteins depend on their interactions with other molecules or ions, known as ligands.
Protein-ligand interactions are quite specific; even though numerous potential ligands surround a cellular protein at any given time, only a particular ligand can bind to that protein. Moreover, a ligand binds only to a dedicated area on the surface of the protein, known as the...
Protein-ligand interactions are quite specific; even though numerous potential ligands surround a cellular protein at any given time, only a particular ligand can bind to that protein. Moreover, a ligand binds only to a dedicated area on the surface of the protein, known as the...
14.8K
Introduction to Mechanisms of Enzyme Catalysis
10.4K
For many years, scientists thought that enzyme-substrate binding took place in a simple "lock-and-key" fashion. This model stated that the enzyme and substrate fit together perfectly in one instantaneous step. However, current research supports a more refined view scientists call induced fit. The induced-fit model expands upon the lock-and-key model by describing a more dynamic interaction between enzyme and substrate. As the enzyme and substrate come together, their interaction causes...
10.4K


