Approaches to Materials Discovery in Inorganic Solid State Chemistry

无机固体构成了当今许多技术应用的基础。金属和类金属是如何结合在一起形成这些新的固体(称为“金属间化合物”)的？令人惊讶的是，要预测元素(特别是金属)如何相互反应，它们采用什么结构，以及它们拥有什么物理性质，仍然不容易。对于一些特定的固体，化学家们已经成功地应用了一些策略来“设计”新的化合物，例如操纵构建块或识别相关系列中缺失的成员。由于这些策略依赖于先例，因此从定义上讲，它们无法预见令人惊讶或违反直觉的结果。探索性工作对于扩大我们的知识库是必不可少的。实验方法包括高温反应(高达1000摄氏度)、X射线衍射、光谱学、物理性质测量和量子计算。接受过这些不同技能培训的学生，处于化学、物理和材料科学的交界处，非常适合从事先进材料研究的职业。随着成分、结构和属性之间的关系变得越来越复杂，最终模式识别技能变得太难让人类处理。为了补充这些传统方法，我们建议机器学习方法可以作为发现新材料的另一种方法，更可靠、更快。使用这两种方法，我们将合成各种金属和边界类金属(如锗、砷和硒)的新化合物。基于我们过去对开发可以将废热转化为电能的热电材料的兴趣，我们有几个令人兴奋的新应用，化合物将成为目标。首先，我们想要评估各种砷化物作为潜在的主体结构，在这种结构中，可以添加少量的磁性原子来产生铁磁半导体，这种半导体可以用于自旋电子器件，以建造更强大的计算机。其次，我们正在设计具有特殊结构的硒化物及其相关化合物，以便将它们用作非线性光学材料，以制造用于传感爆炸物或执行选择性外科手术的红外激光器。第三，我们希望发现新的有色金属间化合物，这种金属间化合物极其稀有，但因其美丽而受到高度重视。所有这些应用都有可能对许多行业产生重大影响，如能源、信息技术、环境可持续性和国防，这些都是加拿大的重大利益所在。我们更广泛的愿景是，这些方法可以推广到其他类型的材料，这样科学家就可以将人类的智慧和机器的速度结合起来，以更深入地了解自然，揭开结构和性能之间的隐藏关系，并加快材料的发现。

Inorganic solids form the basis of many of today's technological applications. How do metals and metalloids combine to form these new solids (called “intermetallic compounds”)? Surprisingly, it is still not easy to predict how elements (especially metals) react with each other, what structures they adopt, and what physical properties they possess. For some specific solids, chemists have been successful in applying strategies to “design” new compounds, such as manipulating building blocks or identifying missing members within related series. Because these strategies rely on precedent, by definition, they cannot foresee surprising or counterintuitive results. Exploratory work is essential for expanding our knowledge base. Experimental methods include high-temperature reactions (up to 1000 C), X-ray diffraction, spectroscopy, physical property measurements, and quantum calculations. Students trained in these diverse skills, which lie at the interface of chemistry, physics, and materials science, are well suited for careers in advanced materials research. As relationships between compositions, structures, and properties grow more complex, eventually the pattern recognition skills become too difficult for humans to handle. To complement these traditional approaches, we propose that machine-learning methods may serve as an alternative approach to discover new materials, more reliably and much faster. Using both these approaches, we will synthesize new compounds of various metals and borderline metalloids (such as germanium, arsenic, and selenium). Building on our past interest in developing thermoelectric materials, which can convert waste heat to electricity, we have several exciting new applications in mind for compounds to be targeted. First, we want to evaluate various arsenides as potential host structures in which small amounts of magnetic atoms can be added to give rise to ferromagnetic semiconductors, which could be used in spintronic devices to build much more powerful computers. Second, we are designing selenides and related compounds with special structures so that they can be used as nonlinear optical materials to fabricate infrared lasers for sensing explosives or performing selective surgical procedures. Third, we want to discover new coloured intermetallics, which are exceedingly rare but are highly valued because of their beauty. All of these applications have the potential to significantly impact many sectors, such as energy, information technology, environmental sustainability, and defense, which are of vital interest to Canada. Our broader vision is that these approaches can be generalized to other types of materials so that scientists can combine the wisdom of humans and the speed of machines to gain a deeper understanding of nature, to unravel the hidden relationships between structures and properties, and to accelerate materials discovery.