Materials Informatics

Olexandr Isayev, PhD, is Assistant Professor at UNC Eshelman School of Pharmacy, University of North Carolina at Chapel Hill. Alexander Tropsha, PhD, is K.H. Lee Distinguished Professor and Associate Dean for Pharmacoinformatics and Data Science at the UNC Eshelman School of Pharmacy, University of North Carolina at Chapel Hill. Stefano Curtarolo, PhD, is Professor in Materials Science, Electrical Engineering, and Physics and Director of the Center for Materials Genomics at Duke University, North Carolina.

• 1 Crystallography Open Database: History, Development, and Perspectives 1Saulius Gražulis, Andrius Merkys, Antanas Vaitkus, Daniel Chateigner, Luca Lutterotti, Peter Moeck, Miguel Quiros, Robert T. Downs, Werner Kaminsky, and Armel Le Bail1.1 Introduction 11.2 Open Databases for Science 31.3 Building COD 61.3.1 Scope and Contents 71.3.2 Data Sources 71.3.3 Data Maintenance 81.3.3.1 Version Control 111.3.3.2 Data Curation Policies 121.3.3.3 Quarterly Releases 131.3.4 Sister Databases (PCOD, TCOD) 141.4 Use of COD 141.4.1 Data Search and Retrieval 141.4.1.1 Data Identification 151.4.1.2 Web Search Interface 151.4.1.3 RESTful Interfaces 151.4.1.4 Output Formats 171.4.1.5 Accessing COD Records 171.4.1.6 MySQL Interface 181.4.1.7 Alternative Implementations of COD Search on the Web 201.4.1.8 Installing a Local Copy of the COD 211.4.1.9 File System-Based Queries 231.4.1.10 Programmatic Use of COD CIFs 241.4.2 Data Deposition 261.5 Applications 271.5.1 Material Identification 271.5.2 Applications for the Mining Industry 271.5.3 Extracting Chemical Information 281.5.4 Property Search 301.5.5 Geometry Statistics 301.5.6 High-Throughput Computations 311.5.7 Applications in College Education and Complementing Outreach Activities 311.6 Perspectives 321.6.1 Historic Structures 321.6.2 Theoretical Data in (T)COD 321.6.3 Conclusion 32Acknowledgments 33References 332 The Inorganic Crystal Structure Database (ICSD): A Tool for Materials Sciences 41Stephan Rühl2.1 Introduction 412.2 Content of ICSD 422.3 Interfaces 462.4 Applications of ICSD 462.4.1 Prediction of Ferroelectricity 472.4.2 Using the Concept of Structure Types 472.4.3 Two Examples of Training Machine Learning Algorithms with ICSD Data 482.4.4 High-Throughput Calculation 502.5 Outlook 51References 513 Pauling File: Toward a Holistic View 55Pierre Villars, Karin Cenzual, Roman Gladyshevskii, and Shuichi Iwata3.1 Introduction 553.1.1 Creation and Development of the PAULING FILE 573.2 PAULING FILE: Crystal Structures 573.2.1 Data Selection 583.2.2 Categories of Crystal Structure Entries 583.2.3 Database Fields 593.2.4 Structure Prototypes 623.2.5 Standardized Crystallographic Data 633.2.5.1 Checking of Symmetry 633.2.5.2 Standardization 653.2.5.3 Comparison with the Type-Defining Data Set 673.2.6 Assigned Atom Coordinates 673.2.7 Atomic Environment Types (AETs) 683.2.8 Cell Parameters from Plots 723.3 PAULING FILE: Phase Diagrams 723.4 PAULING FILE: Physical Properties 753.4.1 Data Selection 753.4.2 Database Fields 763.4.3 Physical Properties Considered in the PAULING FILE 763.5 Data Quality 803.5.1 Computer-Aided Checking 803.6 Distinct Phases 813.6.1 Chemical Formulas and Phase Names 833.6.2 Phase Classifications 843.7 Toward a Megadatabase 843.8 Applications 893.8.1 Products Containing PAULING FILE Data 893.8.2 Holistic Overviews Based on the PAULING FILE 913.8.3 Principles Defining Ordering of Chemical Elements 923.9 Lessons to Learn from Experience 993.10 Conclusion 103References 1044 From Topological Descriptors to Expert Systems: A Route to Predictable Materials 107Alexander P. Shevchenko, Eugeny V. Alexandrov, Olga A. Blatova, Denis E. Yablokov, and Vladislav A. Blatov4.1 Introduction 1074.2 Topological Tools for Developing Knowledge Databases 1084.2.1 Why Topological? 1084.2.2 Topological vs. Other Descriptors of Crystal Structures 1104.2.3 Topological vs. Crystallographic Databases 1114.2.4 Deriving Topological Knowledge from Crystallographic Data 1164.2.4.1 Algorithms for Topological Analysis 1164.2.4.2 Building Distributions of Descriptors 1184.2.4.3 Finding Correlations Between Descriptors 1234.2.5 Universal Data Storage 1264.3 Applications of Topological Tools in Crystal Chemistry and Materials Science 1314.3.1 Network Topology Prediction 1314.3.2 Prediction of Properties 1374.4 Conclusions 137References 1385 A High-Throughput Computational Study Driven by the AiiDA Materials Informatics Framework and the PAULING FILE as Reference Database 149Martin Uhrin, Giovanni Pizzi, Nicolas Mounet, NicolaMarzari, and Pierre Villars5.1 Introduction 1495.1.1 Three Key Developments Opened Up Unprecedented Opportunities 1505.1.2 Relative Few Inorganic Solids Have Been Experimentally Investigated 1515.2 Nature Defines Cornerstones Providing a Marvelously Rich but Still Very Rigid Systematic Framework of Restraint Conditions 1515.3 The First, Second, andThird Paradigms 1535.4 The Realization of the Fourth and Fifth Paradigms Requires Three Preconditions 1535.4.1 Introduction of the Prototype Classification to Link Crystallographic Databases Created by Different Groups 1535.4.2 Introduction of the Distinct Phases Concept to Link Different Kinds of Inorganic Solids Data 1545.4.3 The Existence of a Comprehensive, Critically Evaluated Inorganic Solids Database Concept (DBMS) of Experimentally Determined Single-Phase Inorganic Solids Data to Be Used as Reference 1545.5 The Core Idea of the Fifth Paradigm 1545.6 Restraint Conditions Revealed by “Inorganic Solids Overview–Governing Factor Spaces (Maps)” Discovered by Data-Mining Techniques 1565.6.1 Compound Formation Maps 1575.6.2 Atomic Environment Type Stability Maps for AB Inorganic Solids 1585.6.3 Twelve Principles in Materials Science Supporting Three Cornerstones Given by Nature 1595.7 Quantum Simulation Strategy 1615.8 Workflows Engine in AiiDA to Carry Out High-Throughput Calculation for the Creation of the Materials Cloud, Binaries Edition 1645.8.1 AiiDA 1645.8.2 SSSP (Standard Solid State Pseudopotentials) Library 1655.8.3 Workflows 1665.8.4 Workfunctions 1665.8.5 Workchains 1665.8.6 Workflows Used in This Project 1685.9 Conclusions 169Acknowledgment 169References 1696 Modeling Materials Quantum Properties with Machine Learning 171Felix A. Faber and O. Anatole von Lilienfeld6.1 Introduction 1716.2 Kernel Ridge Regression 1716.3 Model Assessment 1736.3.1 Learning Curve 1736.3.2 Speedup 1746.4 Representations 1766.5 Recent Developments 177References 1787 Automated Computation of Materials Properties 181Cormac Toher, Corey Oses, and Stefano Curtarolo7.1 Introduction 1817.2 Automated Computational Materials Design Frameworks 1827.2.1 Generating and Using Databases for Materials Discovery 1827.2.2 Standardized Protocols for Automated Data Generation 1857.3 Integrated Calculation of Materials Properties 1877.3.1 Autonomous Symmetry Analysis 1897.3.2 Elastic Constants 1917.3.3 Quasi-harmonic Debye–Grüneisen Model 1937.3.4 Harmonic Phonons 1957.3.5 Quasi-harmonic Phonons 1977.3.6 Anharmonic Phonons 1987.4 Online Data Repositories 1987.4.1 Computational Materials Data Web Portals 1987.4.2 Programmatically Accessible Online Repositories of Computed Materials Properties 2007.5 Materials Applications 2027.5.1 Disordered Materials 2027.5.1.1 High Entropy Materials 2037.5.1.2 Metallic Glasses 2037.5.1.3 Modeling Off-Stoichiometry Materials 2047.5.2 Superalloys 2057.5.3 Thermoelectrics 2057.5.4 Magnetic Materials 2087.6 Conclusion 209Acknowledgments 209References 2098 Cognitive Chemistry: The Marriage of Machine Learning and Chemistry to Accelerate Materials Discovery 223Edward O. Pyzer-Knapp8.1 Introduction 2238.2 Describing Molecules for Machine Learning Algorithms 2248.3 Building Fast and Accurate Models with Machine Learning 2348.3.1 Squared Exponential Kernel 2398.3.2 Rational Quadratic Kernel 2408.4 Searching Through Chemical Libraries 2448.5 Conclusion 248References 2499 Machine Learning Interatomic Potentials for Global Optimization and Molecular Dynamics Simulation 253Ivan A. Kruglov, Pavel E. Dolgirev, Artem R. Oganov, Arslan B. Mazitov, Sergey N. Pozdnyakov, Efim A. Mazhnik, and Alexey V. Yanilkin9.1 Introduction 2539.2 Machine Learning Potential for Global Optimization 2589.2.1 Lattice Sums Method 2589.2.2 Feature Vector 2619.2.3 Feature Vector Analysis 2629.2.4 Examples of Machine Learning Interatomic Potentials 2659.2.4.1 Aluminum 2659.2.4.2 Carbon 2679.2.4.3 Helium and Xenon 2719.2.5 Discussion 2729.3 Interatomic Potential for Molecular Dynamics 2739.3.1 General Form of the Potential 2739.3.2 Parameters Selection 2749.3.3 Thermodynamic Quantities and Phase Transitions 2779.3.4 Interatomic Potential for System of Two (or More) Atomic Types 2819.4 Statistical Approach for Constructing ML Potentials 2849.4.1 Two-Body Potential 2849.4.2 Three-Body Potential 286Acknowledgements 286References 286Index 289