Automated Search for Materials for Ammonia Synthesis and Water Splitting

利用可再生电力和太阳能将水、二氧化碳和氮转化为高能量密度燃料和高价值化学品，可以改善间歇性太阳能和风能的储存和利用。这种“太阳能燃料”技术有利于将可再生能源转化为用于制造工业产品的增值化学品。该项目支持为各种氧化还原反应发现先进催化剂的基础研究。当进行新材料发现时，对于给定的有前途的材料组合物家族，只有一小部分材料将具有用于目标反应应用的期望性质。对于许多这些固态材料，化学平衡和化学反应的驱动力是未知的。已经开发了统计学习方法，其可以从大量数据中提取信息以训练高度可靠的“人工智能”模型，用于预测新材料系统的性质。在这个项目中，主要研究人员正在使用机器学习方法应用于数百种材料的实验数据，以预测数百种材料的稳定性，结构和化学反应性。然后，预测的性质可以用于识别催化技术上重要的反应的候选材料，例如将水分解为氧气和氢气，将二氧化碳转化为有用的产品，或从氮气和水合成氨的“绿色”合成。这些模型可以通过机器学习计算机代码等公共存储库和公开访问的材料数据库获得。该项目正在培训高中，本科和研究生在化学，化学工程和材料科学应用中应用最先进的机器学习方法。该研究通过PI参与科罗拉多大学领导力和多样性（BOLD）中心的扩大机会，以及在PI课程中纳入机器学习和化学的新概念，与教育和推广相结合。该项目将应用机器学习方法来大规模发现新的氧化物和氮氧化物材料，用于催化将水分解为氧气和氢气，或从氮气和水合成氨的“绿色”合成。分解水和氨合成中涉及的反应的化学驱动力关键取决于在氧化物或氮氧化物材料中产生氧空位的能量。在这个项目中，PI正在使用机器学习方法，这些方法在一组量子力学计算的氧空位形成能量上进行训练。该项目是互补的，并利用了CHE 1800592基金，该基金专注于机器学习方法和数据集的开发。预测的性质，然后可以用来识别候选氧化物和氮氧化物催化分解水或氨合成。该项目结合了电子结构，热力学，计算材料科学和机器学习方面的专业知识，研究氧化物的核心性质-氧空位形成能EV。数据驱动的方法利用了表明EV系统地取决于各种材料特性的结果，例如材料的电子带隙和生成焓。研究人员将应用机器学习方法直接对EV进行建模，使用量子力学计算的数百种材料的EV数据进行训练和描述符提取。由此产生的描述符被用来预测EV的独特的氧化物和氮化物的组合物，反过来，将使数以百万计的反应平衡的氧化和还原反应的水分解和氨合成介导的这些材料的计算。尽管先进的氧化物和氮氧化物在广泛的技术中具有巨大的技术和经济重要性，但对于引起其化学性质的详细行为仍有很多未知之处。该项目产生的新技术和热化学数据库的潜在应用包括使用氧化还原材料的热化学水裂解，化学循环合成氨，材料氧化，氧化物的碳热还原，氧分离膜，该奖项反映了NSF的法定使命，并通过使用基金会的智力价值进行评估，被认为值得支持和更广泛的影响审查标准。

The use of renewable electricity and solar energy to convert water, carbon dioxide, and nitrogen into energy-dense fuels and high-valued chemicals can improve the storage and utilization of intermittent solar and wind energy. This technology for "solar fuels" has benefits in utilization of renewable energy sources into value added chemicals used to make industrial products. This project supports fundamental research of the discovery of advanced catalysts for a wide range of redox reactions. When conducting new materials discovery, for a given family of promising material compositions, only a fraction of materials will have desirable properties for the targeted reaction applications. For many of these solid-state materials, the chemical equilibria and driving forces for chemical reactions are unknown. Statistical learning approaches have been developed which can extract information from large quantities of data to train highly reliable "artificial intelligence" models for predicting properties of a new material system. In this project, the principal investigators are using machine learning approaches applied to experimental data for hundreds of materials to predict the stabilities, structures, and chemical reactivity of hundreds of materials. The predicted properties can then be used to identify candidate materials for catalyzing technologically-important reactions, such as splitting water into oxygen and hydrogen, converting carbon dioxide into useful products, or the 'green' synthesis of ammonia from nitrogen and water. The models are being made available on public repositories such as machine learning computer codes, and through publicly-accessible materials databases. The project is training high school, undergraduate and graduate students in the application of state-of-the-art machine learning methods for chemistry, chemical engineering, and materials science applications. The research is integrated with education and outreach through the PI's participation in the Broadening Opportunity through the Leadership and Diversity (BOLD) Center at University of Colorado, and the incorporation of new concepts in machine learning and chemistry within the PI's courses. This project will apply machine learning approaches for the discovery of new oxide and oxynitride materials at scale for catalyzing splitting water into oxygen and hydrogen or the 'green' synthesis of ammonia from nitrogen and water. The chemical driving forces for the reactions involved in splitting water and ammonia synthesis depend critically on the energy to create an oxygen vacancy in the oxide or oxynitride material. In this project, The PI is using machine learning approaches trained on a set of oxygen vacancy formation energies that were calculated quantum mechanically. This project is complementary and leverages grant CHE 1800592 that focuses on the development of the machine learning methods and datasets. The predicted properties can then be used to identify candidate oxides and oxynitrides for catalyzing splitting water or ammonia synthesis. This project combines expertise in electronic structure, thermodynamics, computational materials science, and machine learning to study a central property of oxides - their oxygen vacancy formation energies, EV. The data-driven approach takes advantage of results showing that EV depends systematically on various materials properties, such as the electronic band gap and the enthalpy of formation of the material. The researchers will apply machine learning methods to model EV directly, using quantum mechanically calculated EV data for several hundred materials for training and descriptor extraction. The resulting descriptors are being used to predict EV for unique oxide and nitride compositions, and in turn, will enable the computation of millions of reaction equilibria for the oxidation and reduction reactions for water splitting and ammonia synthesis mediated by these materials. Despite the enormous technological and economic importance of advanced oxides and oxynitrides in a broad range of technologies, much is still unknown about the detailed behavior that give rise to their chemical properties. Potential applications of the new techniques and thermochemical databases produced by this project include thermochemical water splitting using redox materials, ammonia synthesis by chemical looping, oxidation of materials, the carbothermal reduction of oxides, oxygen separation membranes, and solid oxide fuel cell electrolytes.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.