发表机构
Oak Ridge National Laboratory（橡树岭国家实验室）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本文提出开源软硬件工具包，实现RHEED控制与分析自动化，支持AI智能体引导薄膜合成，降低自主测量门槛。
AI 中文摘要
反射高能电子衍射（RHEED）在薄膜生长过程中提供了关于表面演化的丰富信息，但非自动化、依赖操作者的对准和碎片化的分析工作流程限制了其在全自主合成中的潜力。在此，我们提出一个开源硬件和软件工具包，使操作者和人工智能（AI）智能体能够使用RHEED控制和定量分析。在脉冲激光沉积系统上演示，该工具包提供可编程的电子光学控制、自动光束对准和摇摆曲线采集，以及一种无需训练的晶体学方位角对准方法。自动RHEED应用程序通过共享的图形和程序化接口提取结构和生长相关的可观测值，包括模型上下文协议（MCP）服务器。一个智能体驱动的演示展示了自然语言请求如何指导定量分析的选择、配置和执行。一个可扩展的适配器接口简化了社区开发的方法的整合，用于AI分析和RHEED模拟，并配有专为AI编码智能体设计的Markdown实现指南。这些能力通过物理测量和学习图像表示支持表面演化的互补描述，同时为纳入新的计算方法提供了实用途径。总之，这些工具降低了自动化和智能体化RHEED测量的障碍，并为未来由演化表面引导的薄膜合成的智能体控制奠定了基础。
英文摘要
Reflection high-energy electron diffraction (RHEED) provides rich information about evolving surfaces during thin-film growth, but non-automated, operator-dependent alignment and fragmented analysis workflows limit its potential in fully autonomous synthesis. Here, we present an open-source hardware and software toolkit that makes RHEED control and quantitative analysis accessible to operators and artificial intelligence (AI) agents. Demonstrated on a pulsed laser deposition system, the toolkit provides programmable electron-optics control, automated beam alignment and rocking-curve acquisition, and a training-free method for crystallographic azimuthal alignment. The auto RHEED application extracts structural and growth-related observables through shared graphical and programmatic interfaces, including a Model Context Protocol (MCP) server. An agent-driven demonstration shows how natural-language requests can guide the selection, configuration, and execution of quantitative analyses. An extensible adapter interface streamlines the incorporation of community-developed methods for AI analysis and RHEED simulation as they emerge, supported by Markdown implementation guides designed for AI coding agents. These capabilities support complementary descriptions of surface evolution through physical measurements and learned image representations while providing a practical route for incorporating new computational methods. Together, these tools reduce barriers to automated and agentic RHEED measurements and establish a foundation for future agentic control of thin-film synthesis guided by the evolving surface.
Comments30 pages, 10 figures
