Materials science emphasizes the solid state and the applications of solid state materials. Among these applications are heterogeneous catalysis, thin films, glasses and other amorphous solids, and electronic materials. Historically, research in this area has been slow, expensive, and often Edisonian in character (try 5000 items and see if something might work).
The increased power and speed of density-functional theory (DFT) quantum mechanical software and sophisticated experimental techniques (often featuring X-rays) have become foundations for understanding solid state materials on an atomic and molecular level. However, even by themselves, they can be used only as fast as a human scientist can use them.
In slightly more than 200 pages of text, this book manages to provide a comprehensive and useful guide to the application of artificial intelligence (AI) to materials science and technology. It is an overview and cannot engage in textbook-level presentations. However, the six chapters of the book are well written despite the large amount of content and include very many references to primary sources for further study.
The first chapter is an overview that emphasizes the role of AI and machine learning in the laboratory. The emphasis is on the process of doing science rather than the science itself. Among the issues discussed are automation of the laboratory for automated and parallel experimentation, extraction and preparation of data, and the levels of machine learning that can be implemented. The next two chapters are on scientific content in materials science.
The second chapter is on catalysis. Since many processes depend on a catalyst in solid form, the composition and structure of the catalyst will determine its reactivity. This chapter discusses commonly used descriptors of chemical reactivity--hard and soft acids and bases, Fukui reactivity measure, and d-band character. DFT is the central theoretical tool. Machine learning aims to develop models for chemical reactivity prediction to be tested in the laboratory. Three examples of catalytic processes are presented: CO2 reduction, oxygen evolution, and methanol oxidation.
The topic of the third chapter is spectroscopy. The spectroscopic techniques discussed include two based on X-rays (X-ray absorption for chemical character and X-ray diffraction for crystal structure) and Raman spectroscopy (based on scattering infrared radiation). In addition to the discussion of the techniques with examples, substantial attention is paid to the problem of integrating these techniques into an automated laboratory setting where the data obtained can be part of an automated experimental program. The transition to the fourth chapter is graceful because its topic is the self-driving laboratory where rapid prototyping can be done, fast turnaround of samples, and the collection of data for machine learning can be implemented.
Chapter 5, on algorithms, is the longest chapter of the book. The algorithms discussed include those for machine learning for screening, decision tree types, optimization (gradient descent and Bayesian), linear (regression-type) models, and dimensionality reduction of a problem. A large portion of the chapter is taken up by a survey of neural networks in materials science with examples.
The last chapter is on informatics in an industrial setting. Rather than focusing on the science and engineering of materials themselves, it advocates the use of AI and machine learning into the materials laboratory and into companies within the industry with all deliberate speed.
This book would be very useful to readers in materials science--especially novices (like new graduate students). More experienced researchers who work in a traditional manner can discover ways of increasing their productivity with the tools that AI and machine learning can provide.