Descriptors of Oxygen-Evolution Activity for Oxides: A Statistical Evaluation
Descriptors of Oxygen-Evolution Activity for Oxides: A Statistical Evaluation
复制标题
DOI：
10.1021/acs.jpcc.5b10071
复制
发表时间：
2016-01-14
影响因子：
3.7
通讯作者：
Shao-Horn, Yang
中科院分区：
文献类型：
--
作者：
Hong, Wesley T.;Welsch, Roy E.;Shao-Horn, Yang
氧电化学过程的催化剂对于燃料电池、人工光合作用和金属-空气电池等可再生能源存储和转换设备的商业可行性至关重要。过渡金属氧化物是开发可扩展的非贵金属基催化剂的优良体系，特别是用于析氧反应（OER）。新型催化剂合理设计的核心是定量构效关系的发展，它将所需的催化行为与材料的结构和/或元素描述符联系起来。最终的目标是利用这些关系来指导材料的设计。在这项研究中，51种钙钛矿的101个内在OER活性从5项文献研究和额外的测量中得到。我们使用多种统计方法，包括因子分析和线性回归模型，探索了14种金属氧键强度描述符的行为和性能。我们发现这些描述子可分为5个描述子家族，并确定电子占位和金属-氧共价是影响OER活性的主要因素。然而，为了开发强大的预测关系，仍然需要考虑多个描述符，这在很大程度上优于仅使用一个或两个描述符（在该领域中通常是这样做的）。我们证实了d电子数、电荷转移能（共价）和eg占位的最优性起着重要作用，但发现M-O-M键角和容差因子等结构因素也有相关性。通过这些工具，我们展示了如何使用统计学习来绘制新的物理见解，并结合数据挖掘在广泛的化学空间中快速筛选OER电催化剂。
Catalysts for oxygen electrochemical processes are critical for the commercial viability of renewable energy storage and conversion devices such as fuel cells, artificial photosynthesis, and metal-air batteries. Transition metal oxides are an excellent system for developing scalable, non-noble-metal-based catalysts, especially for the oxygen evolution reaction (OER). Central to the rational design of novel catalysts is the development of quantitative structureactivity relationships, which correlate the desired catalytic behavior to structural and/or elemental descriptors of materials. The ultimate goal is to use these relationships to guide materials design. In this study, 101 intrinsic OER activities of 51 perovskites were compiled from five studies in literature and additional measurements made for this work. We explored the behavior and performance of 14 descriptors of the metaloxygen bond strength using a number of statistical approaches, including factor analysis and linear regression models. We found that these descriptors can be classified into five descriptor families and identify electron occupancy and metal-oxygen covalency as the dominant influences on the OER activity. However, multiple descriptors still need to be considered in order to develop strong predictive relationships, largely outperforming the use of only one or two descriptors (as conventionally done in the field). We confirmed that the number of d electrons, charge-transfer energy (covalency), and optimality of eg occupancy play the important roles, but found that structural factors such as M-O-M bond angle and tolerance factor are relevant as well. With these tools, we demonstrate how statistical learning can be used to draw novel physical insights and combined with data mining to rapidly screen OER electrocatalysts across a wide chemical space.