Atomistic simulations of H trapping at grain boundaries in ferritic alloys

实现气候中和能源生产的有效氢战略需要连接供需的基础设施。在氢的生产、储存和运输过程中，氢原子与金属微观结构的相互作用以及由此产生的降解现象，被称为氢脆，是一个相当大的挑战。大多数氢脆机制的共同起源是氢原子在金属微结构中的高迁移率。因此，氢技术优化金属材料的一个重要步骤是可靠地预测氢在微观结构中的捕获和分布。在这个项目中，我们的重点是了解氢在铁和铁素体合金中的溶解度。我们的目标是从局部几何、化学和应力/应变状态的角度，将H在gb中不同偏析位点的偏好，即偏析能，与H原子周围的局部原子环境联系起来。为此，我们通过对体心立方晶格中GB自由度空间的系统采样，构建了GB结构和能量数据库。为了从这些数据中推导出结构/属性关系，我们将开发一套分析工具，以结构单元模型的精神，使用基于结构的模型，以及更抽象的局部原子环境的数值表示。通过这种方式，我们一方面希望提供一种直观的GB表征方案，它允许对GB性质进行定性预测，例如它们的迁移率和捕获势，另一方面，通过机器学习方法提供定量预测GB能量和偏析能的描述符，即H的广义化学势，以及一般晶界。

An effective hydrogen strategy for a climate-neutral energy production requires the availability of an infrastructure for connecting supply and demand. Along the chain of H production, storage and transport, the many interactions of H atoms with metallic microstructures and the resulting degradation phenomena, summarized by the term hydrogen embrittlement, are posing a considerable challenge. Common origin of most hydrogen embrittlement mechanisms is the high mobility of H atoms in metallic microstructures. Hence, a major step towards an optimization of metallic materials for hydrogen technology is the reliable prediction of H trapping and distribution in the microstructure. In this project, we focus on understanding the solubility of H at grain boundaries (GBs) in iron and ferritic alloys. Our objective is to relate the preference of H for different segregation sites at GBs, i.e. the segregation energy, to the local atomic environment around the H atom in terms of local geometry, chemistry and stress/strain state. To this end, we construct a database of GB structures and energies via a systematic sampling of the space of the GB degrees of freedom in body centered cubic crystal lattices. To derive structure/property relationships from this data, we will develop a set of analysis tools using structure based models in the spirit of a structural unit model, as well as more abstract numerical representations of the local atomic environment. This way, we want to provide on the one hand an intuitive characterization scheme of GBs, which allows a qualitative prediction of GB properties, such as their mobility and trapping potential, and on the other hand the descriptors for a quantitative prediction of GB energies and segregation energies, i.e. the generalized chemical potential of H, also of general grain boundaries, via a machine learning approach.