标题： 基于数据驱动的分子动力学和透射电子显微镜分析铂功能化石墨烯化学电阻传感器中的晶体生长和氢传感 显示英文标题

标题： Data-Driven Molecular Dynamics and TEM Analysis of Crystal Growth and Hydrogen Sensing in Pt-Functionalized Graphene Chemiresistive Sensors

摘要： 用催化过渡金属功能化的石墨烯通过结合石墨烯卓越的电子输运特性和金属的催化活性，提供了高性能的气体传感性能，但连接合成参数、形貌结果和传感器性能的原子级关系仍然难以捉摸。 我们开发了一种具有近密度泛函理论（DFT）精度的等变机器学习原子间势（MLIP），以进行铂（Pt）晶体在石墨烯上生长及其随后的氢（H$_2$）传感的分子动力学（MD）模拟。 通过透射电子显微镜（TEM）验证的MD模拟显示，Pt沉积始于分散的晶核聚集成多晶纳米团簇，其中主要为面心立方（FCC）内部结构；同时，MD和拉曼光谱均揭示了主要为非共价的Pt-石墨烯相互作用，这种相互作用诱导了适度的局部应变和电荷转移，但仍保持了石墨烯的结构完整性。 反应性MD模拟证实了H$_2$的解离化学吸附仅发生在Pt上，而几乎没有溢出到原始石墨烯上。然而，H在Pt上的吸附减弱了Pt-石墨烯界面结合，从而为H$_2$感知提供了一条间接的电子途径。 瞬态和稳态动力学分析表明，适中的Pt负载量可以最小化检测限；较低的负载量促进了更快的响应和恢复动力学，并增强了信号转导，而较高的负载量增加了石墨烯的掺杂水平。 DFT电荷分析表明，在未配位的Pt簇诱导下，石墨烯中产生$n$型掺杂，而在连续Pt薄膜诱导下产生$p$型掺杂，但在H吸附后这两种效应均有所减弱。 所开发的机器学习MD框架能够对石墨烯上的金属晶体生长进行量子力学准确的建模，阐明了H$_2$感知的基本机制，并将多个关键传感指标与金属负载量和形貌相关联。 显示英文摘要

摘要： Graphene functionalized with catalytic transition metals offers high-performance gas sensing by coupling graphene's exceptional electronic transport properties with the metal's catalytic activity, yet the atomistic relationships connecting synthesis parameters, morphological outcomes, and sensor performance remain elusive. We developed an equivariant machine-learning interatomic potential (MLIP) with near-DFT accuracy to perform molecular dynamics (MD) simulations of Pt crystal growth on graphene and subsequent H$_2$-sensing. MD simulations validated by TEM show that Pt deposition begins with dispersed nuclei coalescing into polycrystalline nanoclusters with predominantly FCC interiors, while both MD and Raman spectroscopy reveal a predominantly non-covalent Pt-graphene interaction that induces moderate local strain and charge transfer yet preserves graphene's structural integrity. Reactive MD simulations confirm H$_2$ dissociative chemisorption exclusively on Pt, with negligible spillover onto pristine graphene. However, H adsorption on Pt attenuates the Pt-graphene interfacial binding, providing an indirect electronic pathway for H$_2$ sensing. Transient and steady-state kinetic analyses demonstrate that an intermediate Pt loading minimizes the limit of detection; lower loadings facilitate faster response and recovery kinetics and enhance signal transduction, whereas higher loadings increase the doping level of graphene. DFT charge analysis indicates that under-coordinated Pt clusters induce $n$-type doping in graphene, whereas continuous Pt films induce $p$-type doping, with both effects attenuated upon H adsorption. The developed machine-learned MD framework enables quantum-mechanically accurate modeling of metal crystal growth on graphene, elucidates the underlying H$_2$-sensing mechanism, and correlates several key sensing figures of merit with metal loading and morphology.