Deep materials informatics: Applications of deep learning in materials science

  • PDF / 1,616,818 Bytes
  • 14 Pages / 612 x 792 pts (letter) Page_size
  • 84 Downloads / 190 Views

DOWNLOAD

REPORT


Artificial Intelligence Prospective

Deep materials informatics: Applications of deep learning in materials science Ankit Agrawal and Alok Choudhary, Department of Electrical Engineering and Computer Science, Northwestern University, Evanston, IL 60201, USA Address all correspondence to Ankit Agrawal at [email protected] (Received 27 January 2019; accepted 24 May 2019)

Abstract The growing application of data-driven analytics in materials science has led to the rise of materials informatics. Within the arena of data analytics, deep learning has emerged as a game-changing technique in the last few years, enabling numerous real-world applications, such as self-driving cars. In this paper, the authors present an overview of deep learning, its advantages, challenges, and recent applications on different types of materials data. The increasingly availability of materials databases and big data in general, along with groundbreaking advances in deep learning offers a lot of promise to accelerate the discovery, design, and deployment of next-generation materials.

Introduction In this era of big data, we are being bombarded with huge volumes of data from a variety of different sources (experiments and simulations) at a staggering velocity in practically all fields of science and engineering, and materials science is no exception. This has led to the emergence of the fourth paradigm of science, which is data-driven science, and builds upon the big data created by the first three paradigms of science (experiment, theory, and simulation). Advanced techniques for datadriven analytics are needed to analyze these data in ways that can help extract meaningful information and knowledge from them, and thus contribute to accelerating materials discovery and realize the vision of Materials Genome Initiative (MGI).[1] The fourth paradigm of science utilizes scalable machine learning (ML) and data mining techniques to extract actionable insights from such big data and inform materials design efforts at various levels. Figure 1 depicts the four paradigms of science.[2] Materials science and engineering researchers rely on experiments and simulations to try to understand the processing–structure–property–performance (PSPP) relationships,[2,3] which are far from being well-understood. In fact, almost everything in materials science depends on these PSPP relationships, where the cause–effect relationships of science go from left to right, and the goals–means relationships of engineering go from right to left. In order to discover and design new improved materials with desired properties, we need to better understand this complex system of PSPP relationships. Figure 2 depicts these PSPP relationships of materials science and engineering.[2] The scalable data-driven techniques[4–8] of the fourth paradigm of science have found numerous applications in a lot

of diverse fields such as marketing and commerce,[9,10] healthcare,[11,12] climate science,[13,14] bioinformatics,[15,16] social media,[17,18] materials science,[19,20] and cosmol