AI-Guided Design and Property Prediction for Zeolites and Nanoporous Materials
A cohesive and insightful compilation of resources explaining the latest discoveries and methods in the field of nanoporous materials
In Artificial Intelligence for Zeolites and Nanoporous Materials: Design, Synthesis and Properties Prediction a team of distinguished researchers delivers a robust compilation of the latest knowledge and most recent developments in computational chemistry, synthetic chemistry, and artificial intelligence as it applies to zeolites, porous molecular materials, covalent organic frameworks and metal-organic frameworks. The book presents a common language that unifies these fields of research and advances the discovery of new nanoporous materials.
The editors have included resources that describe strategies to synthesize new nanoporous materials, construct databases of materials, structure directing agents, and synthesis conditions, and explain computational methods to generate new materials. They also offer material that discusses AI and machine learning algorithms, as well as other, similar approaches to the field.
Readers will also find a comprehensive approach to artificial intelligence applied to and written in the language of materials chemistry, guiding the reader through the fundamental questions on how far computer algorithms and numerical representations can drive our search of new nanoporous materials for specific applications.
Designed for academic researchers and industry professionals with an interest in synthetic nanoporous materials chemistry, Artificial Intelligence for Zeolites and Nanoporous Materials: Design, Synthesis and Properties Prediction will also earn a place in the libraries of professionals working in large energy, chemical, and biochemical companies with responsibilities related to the design of new nanoporous materials.
Table of Contents
List of Contributors xiii
Preface xvii
About the Cover xxiii
Acknowledgments xxv
1 The Confluence of Organo-Cations, Inorganic Species, and Molecular Modeling on the Discovery of New Zeolite Structures and Compositions 1
Christopher M. Lew, Dan Xie, Joel E. Schmidt, Saleh Elomari, Tracy M. Davis, and Stacey I. Zones
1.1 Introduction 1
1.2 Inorganic Studies 3
1.3 Organic Structure-Directing Agents (OSDAs) 9
1.3.1 Purpose and Important Properties 9
1.3.2 Classes of Ammonium-based OSDAs 10
1.3.3 Methods of Making 12
1.4 OSDA-Zeolite Energetics and Rational Synthesis 15
1.5 Role of High Throughput and Automation 22
1.6 Cataloguing, Archiving, Harvesting, and Mining Years of Historical Data 24
1.7 Concluding Remarks 25
References 25
2 De Novo Design of Organic Structure Directing Agents for the Synthesis of Zeolites 33
Frits Daeyaert and Michael Deem
2.1 Introduction 33
2.2 De Novo Design 34
2.2.1 Molecular Structure Generator 35
2.2.2 Scoring Function 36
2.2.3 Optimization Algorithm 37
2.2.4 Practical Implementation 42
2.3 Scoring Functions for OSDAs 43
2.3.1 Stabilization Energy 43
2.3.2 Other Constraints 44
2.3.3 Multiple Objectives 45
2.4 Applications 48
2.4.1 From Drug Design to the Design of OSDAs for Zeolites 48
2.4.2 Experimental Confirmation: Pure Silica STW 49
2.4.3 Experimental Confirmation: Zeolite AEI 49
2.4.4 Practical Application: SSZ-52 (SFW) 49
2.4.5 Design of Chiral OSDAs to Direct the Synthesis of Chiral STW 49
2.4.6 Design of Selective OSDAs Directed Toward BEA vs. BEB 51
2.4.7 Design of OSDAs for Chiral Zeolite BEA 52
2.4.8 Application of a Machine-Learning Scoring Function in the De Novo Design of OSDAs for Zeolite Beta 52
2.4.9 Design of OSDAs for Zeolites for Gas Adsorption and Separation 52
2.4.9.1 Carbon Capture and Storage: WEI, JBW, GIS, SIV, DAC, 8124767, 8277563 52
2.4.9.2 Carbon Dioxide/Methane Separation: GIS, ABW, 8186909, 8198030 53
2.4.9.3 Separation of Ethylene-Ethane: DFT, ACO, NAT, JRY 53
2.4.10 Design of MOFs for Methane Storage and Delivery 54
2.4.11 Multi-Objective De Novo Design of OSDAs for Zeolites Using an Ant Colony Optimization Algorithm 55
2.5 Conclusions and Outlook 55
References 56
3 Machine Learning Search for Suitable Structure Directing Agents for the Synthesis of Beta (BEA) Zeolite Using Molecular Topology and Monte Carlo Techniques 61
María Gálvez-Llompart and German Sastre
3.1 Introduction 61
3.2 Artificial Neural Networks for Modeling Zeolite-SDA van der Waals Energy Applied to BEA Zeolite 64
3.3 Virtual Screening: Identifying Novel SDA with Favorable E ZEO-SDA for the Synthesis of BEA Zeolite 69
3.4 Zeo-SDA Energy Calculation Using Atomic Models 71
3.5 Comparing Zeo-SDA Energy Calculation Using MLR, ANN, and Atomic Models 73
3.6 Conclusions 74
Acknowledgments 77
References 77
4 Generating, Managing, and Mining Big Data in Zeolite Simulations 81
Daniel Schwalbe-Koda and Rafael Gómez-Bombarelli
4.1 Introduction 81
4.1.1 Computational Materials Databases 82
4.1.2 Zeolite Databases 83
4.2 Database of OSDAs for Zeolites 85
4.2.1 Developing a Docking Algorithm 86
4.2.2 Calibrating Binding Energy Predictions 88
4.2.3 Performing and Analyzing High-Throughput Screening Calculations 91
4.2.4 Recalling Synthesis Outcomes from the Literature 94
4.2.5 Proposing OSDA Descriptors 96
4.2.6 Designing with Interactivity 99
4.3 Outlook 102
References 103
5 Co-templating in the Designed Synthesis of Small-pore Zeolite Catalysts 113
Ruxandra G. Chitac, Mervyn D. Shannon, Paul A. Cox, James Mattock, Paul A. Wright, and Alessandro Turrina
5.1 Introduction 113
5.1.1 Definitions: Templates and Structure Directing Agents; Co-templating; Dual Templating; Mixed Templating 114
5.2 SAPO Zeotypes: “Model” Systems for Co-templating 116
5.2.1 The CHA-AEI-SAV-KFI System 116
5.2.2 Development of a Retrosynthetic Co-templating Approach for ABC-6 Structure Types 118
5.3 Co-templating Aluminosilicate Zeolites 120
5.3.1 Inorganic/Organic Co-templates 121
5.3.1.1 Targeting new phases in the RHO family using divalent cations 121
5.3.1.2 Designed synthesis of the aluminosilicate SWY, STA-30 123
5.3.1.3 Co-templating and the charge density mismatch approach 124
5.3.2 Two Organic Templates in Zeolite Synthesis 125
5.3.2.1 Applications of Dual/Mixed Organic Templating 125
5.4 Intergrowth Zeolite Structures as Co-templated Materials 127
5.5 Discussion 134
5.6 Conclusions 138
Acknowledgments 138
References 138
6 Computer Generation of Hypothetical Zeolites 145
Estefania Argente, Soledad Valero, Alechania Misturini, Michael M.J. Treacy, Laurent Baumes, and German Sastre
6.1 Introduction 145
6.2 Genetic Algorithms 146
6.2.1 Codification of Genetic Algorithms 147
6.2.2 Selection Operators for Genetic Algorithms 147
6.2.3 Crossover Operators for Genetic Algorithms 149
6.2.4 Mutation Operators for Genetic Algorithms 150
6.3 Algorithms for Zeolite Structure Determination and Prediction 151
6.3.1 Zefsaii 152
6.3.2 FraGen (Framework Generator) 152
6.3.3 SCIBS (Symmetry-Constrained Intersite Bonding Search) 153
6.3.4 TTL GRINSP (Geometrically Restrained Inorganic Structure Prediction) 154
6.3.5 EZs (Exclusive Zones) 155
6.3.6 P-GHAZ (Parallel Genetic Hybrid Algorithm for Zeolites) 155
6.3.7 zeoGAsolver 156
6.4 zeoGAsolver: A Specific Example of Genetic Algorithm for ZSD 156
6.4.1 Setting Up and Coding Scheme 157
6.4.2 Initialization 157
6.4.3 Fitness Evaluation 157
6.4.4 Crossover 159
6.4.5 Population Reduction and Termination Criterion 160
6.5 Graphics Processing Units in Zeolite Structure Determination and Prediction 160
6.5.1 Quick Presentation of GPU Cards 160
6.5.2 Efficient Parallelization of Evolutionary Algorithms on GPUs 161
6.5.3 Genetic Algorithms on GPUs for Zeolite Structures Problem 162
6.5.4 GPUs in Island Model for Interrupted Zeolitic Frameworks 167
6.6 Conclusions 168
Acknowledgments 169
References 169
7 Numerical Representations of Chemical Data for Structure-Based Machine Learning 173
Gyoung S. Na
7.1 Machine Readable Data Formats 173
7.1.1 Feature Vectors 173
7.1.2 Matrices 174
7.1.3 Mathematical Graphs 175
7.2 Graph-based Molecular Representations 175
7.2.1 Chemical Representations of Molecular Structures 175
7.2.2 Molecular Graphs 176
7.2.3 XYZ File to Molecular Graph 177
7.2.4 SMILES to Molecular Graph 178
7.2.5 Multiple Molecular Graph 178
7.3 Machine Learning with Molecular Graphs 179
7.3.1 General Architecture of Graph Neural Networks 179
7.3.2 Graph Convolutional Network 181
7.3.3 Graph Attention Network 182
7.3.4 Continuous Kernel-based Convolutional Network 182
7.3.5 Crystal Graph Convolutional Neural Network 183
7.4 Graph-based Machine Learning for Molecular Interactions 183
7.4.1 Vector Concatenation Approach to Prediction Molecule-to-Molecule Interactions 184
7.4.2 Attention Map Approach for Interpretable Prediction of Molecule-to-Molecule Interactions 185
7.5 Representation Learning from Molecular Graphs 186
7.5.1 Unsupervised Representation Learning 187
7.5.2 Supervised Representation Learning 187
7.6 Python Implementations 189
7.6.1 Data Conversion: Molecular Structures to Molecular Graphs 190
7.6.2 Machine Learning: Deep Learning Frameworks for Graph Neural Networks 190
7.6.3 Pymatgen for Crystal Structures 192
7.7 Graph-based Machine Learning for Chemical Applications 193
7.7.1 Message Passing Neural Network to Predict Physical Properties of Molecules 193
7.7.2 Scale-Aware Prediction of Molecular Properties 193
7.7.3 Prediction of Optimal Properties From Chromophore-Solvent Interactions 194
7.7.4 Drug Discovery with Reinforcement Learning 195
7.7.5 Graph Neural Networks for Crystal Structures 195
7.8 Conclusion 196
References 196
8 Extracting Metal-Organic Frameworks Data from the Cambridge Structural Database 201
Aurelia Li, Rocio Bueno-Perez, and David Fairen-Jimenez
8.1 Introduction 201
8.2 Building the CSD MOF Subset 203
8.2.1 What Is a MOF? 203
8.2.2 ConQuest 204
8.3 The CSD MOF Subset 208
8.3.1 Removing Solvents With the CSD Python API 209
8.3.2 Adding Missing Hydrogens 209
8.4 Textural Properties of MOFs and Their Evolution 210
8.5 Classification of MOFs 211
8.5.1 Identification of Target MOF Families 212
8.5.2 Identification of Surface Functionalities in MOFs 217
8.5.3 Identification of Chiral MOFs 217
8.5.4 Porous Network Connectivity and Framework Dimensionality 218
8.5.5 An Insight into Crystal Quality of Different MOF Families 220
8.6 The CSD MOF Subset Among All the MOF Databases 223
8.7 Conclusions 225
Acknowledgments 226
References 226
9 Data-Driven Approach for Rational Synthesis of Zeolites and Other Nanoporous Materials 233
Watcharop Chaikittisilp
9.1 Introduction 233
9.2 Rationalization of the Synthesis-Structure Relationship in Zeolite Synthesis: Application Machine Learning and Graph Theory to Zeolite Synthesis 234
9.3 Extraction of the Structure-Property Relationship in Nanoporous Nitrogen-Doped Carbons: Dealing with the Missing Values in Literature Data 239
9.4 Acceleration of Experimental Exploration of Nanoporous Metal Alloys: An Active Learning Approach 243
9.5 Summary 247
Acknowledgments 248
References 248
10 Porous Molecular Materials: Exploring Structure and Property Space with Software and Artificial Intelligence 251
Steven Bennett and Kim E. Jelfs
10.1 Introduction 251
10.2 Computational Modeling of Porous Molecular Materials 255
10.2.1 Structure Prediction 256
10.2.2 Modeling Porosity 257
10.2.3 Amorphous and Liquid Phase Simulations 259
10.3 Data-Driven Discovery: Applying Artificial Intelligence Methods to Materials Discovery 260
10.3.1 Training Data Generation 262
10.3.1.1 Hypothetical Structure Datasets 262
10.3.1.2 Experimental Structure Datasets 263
10.3.1.3 Extraction of Data From Scientific Literature 263
10.3.1.4 Data Augmentation and Transfer Learning 263
10.3.2 Descriptor Construction and Selection 264
10.3.2.1 Local Environment Descriptors 264
10.3.2.2 Global Environment Descriptors 265
10.4 Efficient Traversal of the Chemical Space of Porous Materials 266
10.4.1 Evolutionary Algorithms 266
10.4.2 Reducing the Number of Experiments: Bayesian Optimization and Active Learning 267
10.4.3 Chemical Space Exploration with Deep Learning 268
10.5 Considering Synthetic Accessibility 269
10.6 Closing the Loop: How Can High-Throughput Experimentation Feed Back into Computation? 270
10.6.1 High-Throughput and Autonomous Experimentation 271
10.7 Conclusions 272
References 272
11 Machine Learning-Aided Discovery of Nanoporous Materials for Energy- and Environmental-Related Applications 283
Archit Datar, Qiang Lyu, and Li-Chiang Lin
11.1 Introduction 283
11.1.1 Nanoporous Materials 283
11.1.2 History and Development 283
11.1.3 Gas Separation and Storage Applications 284
11.1.4 Large-Scale Computational Screening for Gas Separation and Storage 284
11.2 Concepts and Background for Data-Driven Approaches 286
11.2.1 Dimensionality Reduction 286
11.2.2 Machine Learning Models 287
11.2.2.1 Linear Models 287
11.2.2.2 Decision Trees and Random Forests 288
11.2.2.3 Support Vector Machine 289
11.2.2.4 Neural Networks 289
11.2.2.5 Unsupervised Learning 290
11.3 Data-Driven Approaches 290
11.3.1 Nanoporous Structure Datasets 291
11.3.2 Identifying Feature Space of Materials to Screen 292
11.3.3 Methods to Search for Optimal Structures 295
11.3.4 Modeling Interatomic and Intermolecular Interactions 297
11.4 Case Studies 300
11.4.1 Post-Combustion CO2 Capture 300
11.4.2 Methane Storage 303
11.4.3 Hydrogen Storage 305
11.5 Summary and Outlook 309
References 311
12 Big Data Science in Nanoporous Materials: Datasets and Descriptors 319
Maciej Haranczyk and Giulia Lo Dico
12.1 Introduction 319
12.2 Repositories of Nanoporous Material Structures 321
12.2.1 Experimental Crystal Structures 321
12.2.2 Predicted Crystal Structures 322
12.3 Descriptors 325
12.3.1 Handcrafted Descriptors 325
12.3.2 Toward Automatically Generated and Multi-Scale Descriptors 328
12.4 Properties 329
12.5 Data Analysis 330
12.5.1 Material Similarity and Distance Measures 330
12.5.1.1 Diversity Selection 331
12.5.1.2 Cluster Analysis 332
12.6 Machine Learning Models of Structure-Property Relationships 333
12.7 Current and Future Applications 335
References 336
13 Efficient Data Utilization in Training Machine Learning Models for Nanoporous Materials Screening 343
Diego A. Gómez-Gualdrón, Cory M. Simon, and Yamil J. Colón
13.1 Descriptor Selection 344
13.1.1 Engineering of Advanced Features 344
13.1.2 Engineering of Simpler Features 347
13.2 Material Selection 349
13.3 Model Selection 351
13.3.1 Linear Regression 353
13.3.2 Supported Vector Regressors 354
13.3.3 Decision Tree-based Regressors 355
13.3.4 Artificial Neural Networks 357
13.4 Data Usage Strategies 360
13.4.1 Transfer Learning 361
13.4.2 Multipurpose Models 365
13.4.3 Material Recommendation Systems 368
13.4.4 Active Learning 370
13.4.5 Machine Learning to Speed Up Data Generation 371
13.5 Summary and Outlook 374
References 375
14 Machine Learning and Digital Manufacturing Approaches for Solid-State Materials Development 377
Lawson T. Glasby, Emily H. Whaites, and Peyman Z. Moghadam
14.1 Introduction 377
14.2 The Development of MOF Databases 379
14.3 Natural Language Processing 380
14.4 An Overview of Machine Learning Models 383
14.5 Machine Learning for Synthesis and Investigation of Solid State Materials 386
14.6 Machine Learning in Design and Discovery of MOFs 388
14.7 Current Limitations of Machine Learning for MOFs 392
14.8 Automated Synthesis and Digital Manufacturing 394
14.9 Digital Manufacturing of MOFs 401
14.10 The Future of Digital Manufacturing 403
References 404
15 Overview of AI in the Understanding and Design of Nanoporous Materials 411
Seyed Mohamad Moosavi, Frits Daeyaert, Michael W. Deem, and German Sastre
15.1 Introduction 411
15.2 Databases 411
15.2.1 Structural Databases 412
15.2.2 Databases of Material Properties 412
15.2.3 Databases of Synthesis Protocols 413
15.3 Big-Data Science for Nanoporous Materials Design and Discovery 413
15.3.1 Representations of Chemical Data 413
15.3.2 Learning Algorithms 414
15.4 Applications 415
15.5 Zeolite Synthesis and OSDAs 417
15.6 Conclusion 420
References 420
Index 425
