发表机构
The Pennsylvania State University; Monash University（宾夕法尼亚州立大学; 蒙纳士大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
ReaxKit是一个模块化Python工具包，通过关注点分离架构和引擎适配器，自动化ReaxFF分子动力学模拟的输入准备、输出解析、数据分析及工作流管理，提升可重复性和可扩展性。
AI 中文摘要
经验反应力场（RFF）分子动力学能够对化学复杂系统中的键断裂、键形成、电荷重新分布和结构演化进行原子级模拟。ReaxFF方法可以说是目前可用的RFF方法中最流行且最具可转移性的。然而，ReaxFF的常规使用通常需要大量手动工作来准备输入、解释特定于引擎的输出、组织模拟工件以及开发自定义分析脚本，这限制了可重复性和可扩展性。在此，我们介绍ReaxKit，一个用于准备、解析、分析和管理ReaxFF分子动力学模拟的模块化Python工具包。ReaxKit采用关注点分离架构，区分特定于引擎的输入/输出处理、规范领域数据模型、科学分析、工作流编排、呈现、存储和图形交互。引擎适配器将来自受支持模拟环境的输出转换为类型化、引擎无关的数据结构，使分析模块能够独立于原生文件格式运行。用户请求通过一致的命令行、函数式Python和基于浏览器的图形界面执行，而专用工作区保留原始数据、规范化数据集、分析设置、结果、日志、缓存和来源信息。代表性应用展示了该工具包的广度，包括从Materials Project结构和力学性能自动生成弹性和状态方程训练数据、表征活性位点和局部结构环境，以及通过YAML定义的研究工作流执行模拟活动。这些能力表明ReaxKit支持ReaxFF工作流的所有基本阶段。
英文摘要
Empirical reactive force field (RFF) molecular dynamics enables atomistic simulation of bond breaking, bond formation, charge redistribution, and structural evolution in chemically complex systems. The ReaxFF method is arguably the most popular and transferable of the currently available RFF methods. However, routine use of ReaxFF often requires substantial manual effort to prepare inputs, interpret engine-specific outputs, organize simulation artifacts, and develop custom analysis scripts, limiting reproducibility and scalability. Here, we present ReaxKit, a modular Python toolkit for preparing, parsing, analyzing, and managing ReaxFF molecular dynamics simulations. ReaxKit uses a separation-of-concerns architecture that distinguishes engine-specific input/output handling, canonical domain data models, scientific analysis, workflow orchestration, presentation, storage, and graphical interaction. Engine adapters convert outputs from supported simulation environments into typed, engine-independent data structures, allowing analysis modules to operate independently of native file formats. User requests are executed through consistent command-line, functional Python, and browser-based graphical interfaces, while a dedicated workspace preserves raw data, normalized datasets, analysis settings, results, logs, caches, and provenance information. Representative applications demonstrate the breadth of the toolkit, including automated generation of elastic and equation-of-state training data from Materials Project structures and mechanical properties, characterization of active sites and local structural environments, and execution of simulation campaigns through YAML-defined study workflows. These capabilities show that ReaxKit supports all essential stages of the ReaxFF workflow.
Comments56 pages
