EAGER: SSMCDAT2023: Deep learning Gibbs free energy functions to guide solid-state material synthesis

非技术性总结：本奖项是根据EAGER的建议而颁发的。它支持在利哈伊大学举行的SSMCDAT 2023数据会议上推进的项目进展。EAGER奖支持旨在促进我们对如何合成固体无机材料的理解的研究，这些材料是新型功能设备的基石，以满足清洁能源，环境可持续性和人类福利的社会需求。如今，许多新的无机材料已经使用计算机虚拟设计，但由于缺乏如何在实践中合成它们的知识，它们很少被转化为真实的应用。在这个项目中，一个具有材料科学和计算机科学专业知识的跨学科研究人员团队将开发一个数据驱动的框架，以预测合成这些计算设计材料的最佳方法。该团队将利用人工智能的深度学习算法，考虑温度、压力和材料化学成分的影响，有效预测待合成材料的自由能及其可能的前体。通过这些预测，研究小组将评估与合成相关的化学反应的可行性，并相应地确定最佳反应条件。为了验证和改进他们的计算预测，研究人员还将进行实验室实验，他们将使用真实的时间表征技术密切监测合成过程。为了证明他们的方法的有效性，研究人员将专注于成功合成一组被称为过渡金属氮氧化物的新型无机化合物，这些化合物在能量转换应用中很有前途，但尚未在实验上完全实现。通过该项目，该团队旨在弥合计算设计和实际合成之间的差距，为造福社会的新型功能材料开启新的可能性。该奖项还支持使科学和工程领域更加包容和多样化的活动。它将扩大妇女和少女在这些领域的参与，特别是在东非国家。它还将促进来自代表性不足背景的本科生参与研究。此外，研究人员还将开发新的教育材料，向德雷克塞尔大学的本科生教授数据驱动的材料科学。技术摘要：该奖项是根据EAGER的建议而颁发的。它支持在利哈伊大学举行的SSMCDAT 2023数据会议上推进的项目进展。EAGER奖支持旨在推进固态无机材料可合成性科学知识的研究，特别是那些已经通过计算设计但尚未通过实验实现的材料。该研究将开发一个数据驱动的框架，整合深度学习算法，计算热力学建模和验证实验，以有效预测合成途径和最佳条件。利用深度学习算法，该框架将基于无机化合物的热化学性质与其电子结构之间的关系，预测各种化合物化学计量的温度依赖性吉布斯自由能函数。通过预测要合成的化合物及其各种可能的前体的吉布斯自由能函数，该框架将在相位图计算（CALPHAD）方法的背景下对每个候选合成反应的热力学进行建模。通过评估反应自发性对实际合成关键控制参数的依赖性，合理地确定合适的前驱体和合成条件。通过原位合成方法获得的反应途径及其对反应条件的依赖性的实验理解将用于验证和改进建模预测。该团队将专注于成功合成一组计算设计的过渡金属氮氧化物，这对能量转换应用非常有趣，但尚未完全实现实验，作为该框架能力的展示。通过该项目，该团队旨在弥合计算设计和实际合成之间的差距，为造福社会的新型功能材料开启新的可能性。该奖项还支持各种教育和推广活动，这些活动将1）扩大妇女和年轻女孩在科学和工程领域的参与，特别是在东非国家; 2）促进研究参与不足的本科生; 3）开发新的学习模块，该奖项反映了NSF的法定使命，并被认为值得通过使用基金会的知识产权进行评估来支持。优点和更广泛的影响审查标准。

NONTECHNICAL SUMMARY:This award is made on an EAGER proposal. It supports progress on a project advanced at the SSMCDAT 2023 Datathon held at Lehigh University. This EAGER award supports research aimed at advancing our understanding of how to synthesize solid inorganic materials, which serve as the cornerstone for novel functional devices to meet societal needs for clean energy, environmental sustainability, and human welfare. Nowadays, many new inorganic materials have been virtually designed using computers, but they have rarely been converted into real applications because of missing knowledge of how to synthesize them in practice. In this project, a team of interdisciplinary researchers with expertise in materials science and computer science will develop a data-driven framework to predict the best ways to synthesize these computationally designed materials. The team will utilize deep learning algorithms of artificial intelligence to efficiently predict the free energies of the material to be synthesized and their possible precursors, considering the effects of temperature, pressure, and chemical compositions of the materials. With these predictions, the team will assess the feasibility of the chemical reactions associated with synthesis and correspondingly identify the optimal reaction conditions. To validate and improve their computational predictions, the researchers will also conduct laboratory experiments where they will closely monitor the synthesis process as it happens using a real time characterization technique. To demonstrate the effectiveness of their approach, the researchers will focus on successfully synthesizing a group of novel inorganic compounds known as transition metal oxynitrides, which are promising for energy conversion applications but not yet fully realized experimentally. Through this project, the team aims to bridge the gap between computational design and practical synthesis, unlocking new possibilities for novel functional materials that can benefit society.This award also supports activities to make the field of science and engineering more inclusive and diverse. It will broaden the participation of women and young girls in these fields, especially in countries in eastern Africa. It will also promote research involvement among undergraduate students from underrepresented backgrounds. Moreover, the researchers will develop new educational materials to teach data-driven materials science to undergraduate students at Drexel University. TECHNICAL SUMMARY:This award is made on an EAGER proposal. It supports progress on a project advanced at the SSMCDAT 2023 Datathon held at Lehigh University. This EAGER award supports research aimed at advancing scientific knowledge in the synthesizability of solid-state inorganic materials, especially those that have been computationally designed but not yet realized experimentally. The research will develop a data-driven framework, integrating deep learning algorithms, computational thermodynamic modeling, and validation experiments to efficiently predict synthesis pathways and optimal conditions. Utilizing deep learning algorithms, the framework will predict temperature-dependent Gibbs free energy functions for various compound stoichiometries, based on the relationship between thermochemical properties of inorganic compounds and their electronic structures. With the predictions of Gibbs free energy functions for the compound to be synthesized and its various possible precursors, the framework will then model the thermodynamics of each candidate synthesis reaction in the context of the CALculation of PHAse Diagram (CALPHAD) method. Suitable precursors and synthesis conditions will be rationally identified by assessing the dependence of reaction spontaneity on the key controlling parameters of practical synthesis. Experimental understanding of the reaction pathway and its dependence on reaction conditions obtained through an in situ synthesis approach will be used to validate and improve modeling predictions. The team will focus on successfully synthesizing a group of computationally designed transition metal oxynitrides, highly interesting for energy conversion applications but not yet fully realized experimentally, as a demonstration of the framework's capabilities. Through this project, the team aims to bridge the gap between computational design and practical synthesis, unlocking new possibilities for novel functional materials that can benefit society.This award also supports various education and outreach activities that will 1) expand the involvement of women and young girls in the fields of science and engineering, particularly in countries in eastern Africa; 2) promote research participation among underrepresented undergraduate students; and 3) develop new learning modules that incorporate data-driven materials science in the undergraduate curriculum at Drexel University.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.