[论文解读] Picturing the Gap Between the Performance and US-DOE's Hydrogen Storage Target: A Data-Driven Model for MgH2 Dehydrogenation

Chaoqun Li, Weijie Yang+6 more|arXiv (Cornell University)|Apr 16, 2024

Spacecraft and Cryogenic Technologies被引用 10

一句话总结

本文提出一种数据驱动模型，使用 Mg-H 键轨道贡献和氢原子距离来预测 MgH2 脱氢势垒，提供一种比从头算过渡态搜索更便宜的替代方案，并与 US-DOE 目标保持一致。

ABSTRACT

Developing solid-state hydrogen storage materials is as pressing as ever, which requires a comprehensive understanding of the dehydrogenation chemistry of a solid-state hydride. Transition state search and kinetics calculations are essential to understanding and designing high-performance solid-state hydrogen storage materials by filling in the knowledge gap that current experimental techniques cannot measure. However, the ab initio analysis of these processes is computationally expensive and time-consuming. Searching for descriptors to accurately predict the energy barrier is urgently needed, to accelerate the prediction of hydrogen storage material properties and identify the opportunities and challenges in this field. Herein, we develop a data-driven model to describe and predict the dehydrogenation barriers of a typical solid-state hydrogen storage material, magnesium hydride (MgH2), based on the combination of the crystal Hamilton population orbital of Mg-H bond and the distance between atomic hydrogen. By deriving the distance energy ratio, this model elucidates the key chemistry of the reaction kinetics. All the parameters in this model can be directly calculated with significantly less computational cost than conventional transition state search, so that the dehydrogenation performance of hydrogen storage materials can be predicted efficiently. Finally, we found that this model leads to excellent agreement with typical experimental measurements reported to date and provides clear design guidelines on how to propel the performance of MgH2 closer to the target set by the United States Department of Energy (US-DOE).

研究动机与目标

• 动员更快的描述符来预测氢存储材料的动力学，而超出代价高昂的从头算过渡态搜索。
• 开发一个基于数据驱动的 MgH2 脱氢势垒预测器。
• 展示该模型与实验测量的一致性，并引导设计朝 DOE 目标前进。

提出的方法

• 将 Mg-H 键的晶体哈密顿人口分析与 H 原子距离结合，形成一个距离能量比。
• 推导出与反应能垒相关的描述符。
• 确保所有参数的计算成本显著低于传统的过渡态搜索。
• 通过将预测结果与可用的实验数据进行比较来验证该模型。

实验结果

研究问题

• RQ1 基于 Mg-H 键特征和氢距离的数据驱动描述符是否能够准确预测 MgH2 的脱氢势垒？
• RQ2 所提出的描述符是否与实验测量和 US-DOE 氢存储性能目标一致？
• RQ3 从该描述符中产生哪些设计指南以改善 MgH2 的脱氢性能？
• RQ4 在计算成本和预测准确度方面，该模型与传统的从头算过渡态方法相比如何？

主要发现

• 该模型使用与 Mg-H 键轨道贡献和 H 距离相关的距离能量比来描述和预测 MgH2 的脱氢势垒。
• 预测结果与迄今为止报道的典型实验测量高度一致。
• 该方法提供设计指南，将 MgH2 的性能推近于 US-DOE 目标。
• 所有模型参数的计算成本显著低于传统过渡态搜索。