Handbook of Green Chemistry - Tools for Green Chemistry [1 ed.] 9783527326457

Table of contents :
3527326456_ftoc
01
Chapter 1: Application of Life Cycle Assessment to Green Chemistry Objectives
1.1 Introduction
1.2 Substitution of Safer Chemicals
1.2.1 Missing Inventory Data and Characterization Factors
1.2.2 Linking LCA and Chemical Risk
1.3 Design Material and Energy-Efficient Processes
1.3.1 Introduction
1.3.2 System Boundaries and Design Guidance
1.3.3 Impact Categories and Green Metrics
1.3.4 Policy Implications
1.4 Promote Renewable Materials and Energy
1.4.1 Introduction
1.4.1.1 Glycerol Case Study
1.4.2 Biochemicals Production
1.4.2.1 Life Cycle Stages of Biochemical Production
1.4.2.2 Environmental Implications of Biomass Production
1.4.2.3 Carbon Accounting and Land Use Change
1.4.2.4 Global Availability of Arable Land
1.5 Conclusion and Recommendations
References
02
Chapter 2: Shortcut Models Based on Molecular Structure for Life Cycle Impact Assessment: The Case of the FineChem Tool and Beyond
2.1 Introduction
2.2 Concept and Development of the FineChem Tool
2.3 Illustrative Applications of the FineChem Tool
2.3.1 LCA Aspects of Solvent Selection for Postcombustion CO2 Capture (PCC)
2.3.2 Bio-Based Production of Platform Chemicals
2.4 Toward A New Group Contribution-Based Version of the FineChem Tool
2.4.1 Introduction to GC models
2.4.2 Development of GC-Based LCA Models
2.4.3 Screening for Substances with Desirable Properties
2.4.4 Illustrative Example of Screening Molecules
2.5 Conclusions and Outlook
References
03
Chapter 3: Models to Estimate Fate, Exposure, and Effects of Chemicals
3.1 Introduction
3.2 Fate
3.3 Ecological Exposure
3.4 Ecosystem Effects
3.4.1 Intraspecies Variability in Populations
3.4.2 Interspecies Variability in Assemblages
3.5 Human Exposure and Effect
3.6 Environmental Impact Evaluation
3.6.1 Life Cycle Assessment
3.6.2 Risk Assessment
3.7 Recent Developments
3.7.1 New Chemicals
3.7.2 Nontoxic Stressors
3.7.3 Uncertainty and Variability
References
04
Chapter 4: Collaborative Approaches to Advance Chemical Safety
4.1 Introduction
4.2 Incentives for Collaboration and Constraints
4.3 Options for Sharing
4.3.1 Sharing Research
4.3.2 Sharing Knowledge
4.3.3 Sharing Data
4.3.4 Sharing Software Development
4.4 The Implementation of Collaborative Organizations
4.5 Collaborative Projects
4.5.1 British Industrial Biological Research Association (BIBRA)
4.5.2 The Chemical Bioactivity Information Centre (CBIC)
4.5.3 The Distributed Structure-Searchable Toxicity Database Network - DSSTox
4.5.4 ICH
4.5.5 Innovative Medicines Initiative (IMI)
4.5.5.1 CHEM21
4.5.5.2 Electronic Health Record for Clinical Research (EHR4CR)
4.5.5.3 eTOX
4.5.5.4 GETREAL
4.5.5.5 iPiE
4.5.5.6 MARCAR
4.5.5.7 MIP-DILI
4.5.6 International Life Sciences Institute (ILSI) and ILSI Health and Environmental Sciences Institute (HESI)
4.5.7 Lhasa Limited
4.5.8 OECD (Q)SAR Toolbox
4.5.9 OpenTox
4.5.10 PhUSE
4.5.11 The Pistoia Alliance
4.5.12 REACH Substance Information Exchange Forums (SIEF)
4.5.13 SEURAT-1 (Safety Evaluation Ultimately Replacing Animal Testing)
4.5.13.1 COSMOS
4.5.13.2 DETECTIVE
4.5.13.3 HeMiBio
4.5.13.4 NOTOX
4.5.13.5 SCRandTox
4.5.13.6 ToxBank
4.5.14 ToxML
4.5.15 The Traditional Chinese Medicine Database
4.5.16 United Nations - the European Agreement Concerning the International Carriage of Dangerous Goods by Road (ADR) and the Globally Harmonized System of Classification and Labeling of Chemicals (GHS)
4.5.17 US Government-Industry Collaborations
4.5.18 VEGA
4.5.19 Yale University Open Data Access (YODA)
4.6 Conclusions
References
05
Chapter 5: Introduction to Green Analytical Chemistry
5.1 Introduction
5.1.1 Defining Green Analytical Chemistry
5.1.2 Dualistic Role of Analytical Chemistry in Relation to Green Chemistry
5.1.3 Brief History of Green Analytical Chemistry
5.2 Greener Analytical Separations
5.2.1 Green Gas Chromatography
5.2.2 Greener Liquid Chromatography
5.2.3 Supercritical Fluid Chromatography
5.3 Green Sample Preparation Techniques and Direct Techniques
5.3.1 Direct Analytical Methods
5.3.2 Microextraction Sample Preparation Techniques
5.3.2.1 Solid-Phase Microextraction
5.3.2.2 Liquid-Phase Microextraction
5.3.2.3 Dispersive Liquid-Liquid Microextraction
5.3.3 Stir Bar Sorptive Extraction
5.3.4 Supercritical Fluid Analytical Extraction
5.3.5 Microwave- and Ultrasound-Assisted Extraction
5.3.6 Ionic Liquids in Extraction
5.4 Chemometrics for Signals Processing
5.5 Conclusions
References
06
Chapter 6: Cosmo-RS-Assisted Solvent Screening for Green Extraction of Natural Products
6.1 Introduction
6.2 Solvents for Green Extraction
6.2.1 Definition
6.2.2 Solute-Solvent Interaction
6.2.3 Substitution Concept
6.2.4 Panorama of Alternative Solvents for Extraction
6.2.4.1 Water: Solvent with Variable Polarity
6.2.4.2 Bio-Based Solvents
Organic Acid Esters: Ethyl Lactate, Ethyl Acetate
Fatty Acid Esters
Alcohols: Ethanol and Fusel Alcohols
Terpenes
Synthetic Biobased Solvents, Furfural Derivatives
6.2.4.3 Solvent Obtained from Chemical Synthesis
6.2.4.4 Vegetable Oils
6.2.4.5 Eutectic Solvents
6.2.4.6 Supercritical CO2
6.3 Prediction of Solvent Extraction of Natural Product
6.3.1 COSMO-RS Approach
6.3.2 Applications of COSMO-RS-Assisted Substitution of Solvent
6.3.2.1 Example 1: COSMO-RS Assisted Selection of Solvent for Extraction of Seed Oils
COSMO-RS Study
Experimental Approach
Comparison between Experimental and Simulations
6.3.2.2 Example 2: Cosmo-Rs-assisted Selection of Solvent for Extraction of Aromas
COSMO-RS Prediction
Composition of Extracts
Comparison between Experimental and COSMO-RS Study
6.4 Conclusion
References
07
Chapter 7: Supramolecular Catalysis as a Tool for Green Chemistry
7.1 Introduction
7.2 Control of Selectivity through Supramolecular Interactions
7.2.1 Catalysis with Supramolecular Directing Groups
7.2.2 Scaffolding Ligands
7.2.3 Selectivity through Confinement and Binding Effects
7.3 Reactions in Water
7.3.1 Water-Soluble Nanoreactors
7.3.2 Dehydration Reactions
7.4 Catalyst/Reagent Protection
7.4.1 Catalyst Protection
7.4.2 Protection of Water-Sensitive Reagents
7.5 Tandem Reactions
7.5.1 Synthetic Tandem Reactions
7.5.2 Chemoenzymatic Tandem Reactions
7.6 Conclusion
References
08
Chapter 8: A Tutorial of the Inverse Molecular Design Theory in Tight-Binding Frameworks and Its Applications
8.1 Introduction
8.2 Inverse Molecular Design Theory in Tight-Binding Frameworks
8.2.1 LCAP Principle in Density Functional Theory
8.2.2 LCAP Principle in Tight-Binding Frameworks
8.2.2.1 One-Orbital Tight-Binding Framework
8.2.2.2 Extended Hückel Tight-Binding Framework
8.2.3 Gradient for Optimization
8.3 How to Prepare a Molecular Framework for TB-LCAP Inverse Design?
Outline placeholder
Outline placeholder
2D Molecular Frameworks
3D Molecular Framework
8.4 How to Choose Optional Atom Types or Functional Groups?
8.5 Optimizing Molecular Properties Using the TB-LACP Methods
8.6 Conclusion
References
09
Chapter 9: Green Chemistry Molecular Recognition Processes Applied to Metal Separations in Ore Beneficiation, Element Recycling, Metal Remediation, and Elemental Analysis
9.1 Introduction
9.2 Molecular Recognition Technology as a Green Chemistry Process
9.3 Metal Separations Using Molecular Recognition Technology
9.3.1 Separation and Recovery of Individual Rare Earth Elements
9.3.2 Platinum Group Metals
9.3.2.1 General
9.3.2.2 Palladium Recovery from Native Ore
9.3.2.3 Rhodium Recovery from Spent Catalyst and Other Wastes
9.3.2.4 Platinum Recovery from Alloy Scrap
9.3.2.5 Ruthenium Recovery from Alloy Scrap
9.3.2.6 Iridium Separation from Rhodium and Base Metals
9.3.2.7 Purification of Palladium for Use in Brachytherapy
9.3.3 Gold Separation and Recovery from Process Streams
9.3.3.1 General
9.3.3.2 Gold Recovery from Plating Solutions
9.3.4 Nickel Separations and Recovery
9.3.4.1 Nickel Separations from Laterite Ores
9.3.4.2 Nickel, Aluminum, and Molybdenum Recovery from Acid Leachate of Spent Hydrodesulfurization Catalyst
9.3.4.3 Nickel Removal from Cadmium- and Zinc-Rich Sulfate Electrolyte
9.3.5 Cadmium Removal from a Cobalt Electrolyte Solution Containing a Complex Matrix
9.3.6 Bismuth and Antimony Removal from Copper Electrolyte in Production of High-Purity Copper
9.3.7 Cobalt Recovery from Zinc Streams using Iron(III) as a Pseudo-Catalyst
9.3.8 Molybdenum and Rhenium Separations
9.3.9 Indium Recovery from Etching Wastes
9.3.10 Separation of Indium and Germanium from Zinc Electrolyte Solutions
9.3.10.1 Indium Separation and Recovery
9.3.10.2 Germanium Separation and Recovery
9.3.11 Mercury Recovery from Sulfuric Acid Streams
9.3.12 Metal Recovery from Acid Mine Drainage Streams, Industrial Waste Streams, Mine Leach Streams, and Fly Ash
9.3.12.1 Metal Remediation from Berkeley Pit Acid Mine Drainage Site
9.3.12.2 Removal, Separation, and Recovery of Heavy Metals from Industrial Waste Streams using MRT
9.3.12.3 Uranium Separation and Recovery from Mine Leach Streams
9.3.12.4 Lead Separation from Fly Ash Generated by Ash Melting
9.3.13 Lithium Separation and Recovery from Brine and End-of-Life Rechargeable Batteries
9.3.14 Radionuclide Remediation
9.3.14.1 General
9.3.14.2 Cesium Separation and Recovery from Savannah River Nuclear Wastes
9.3.14.3 Cesium and Technetium Separation and Recovery from Nuclear Wastes at Hanford, Washington
Cesium Separation and Recovery from Nuclear Wastes
Technetium Separation and Recovery from Nuclear Waste
9.3.14.4 Cesium Separation and Recovery from Fly Ash
9.3.14.5 Separation and Recovery of Radioactive Cesium and Strontium from Fukishima, Dai'ichi, Japan Harbor
9.4 Analytical Applications of Molecular Recognition Technology
9.4.1 General
9.4.2 Radionuclides
9.4.2.1 Strontium Separation and Analysis using Empore™ Strontium Rad Disks
9.4.2.2 Radium Separation and Analysis Using Empore™ Radium Rad Disks
9.4.2.3 Other Radionuclide and Mixed Waste Separations
9.4.3 Precious Metals
9.4.4 Toxic Metals
9.4.4.1 Arsenic Separation and Analysis
9.4.4.2 Lead Separation and Analysis
9.4.4.3 Mercury Separation and Analysis []
9.4.5 Rare Earth Metal Separation and Analysis from Rainfall
9.4.6 Multimetal Separations and Recovery
9.5 Conclusion
References
10
Chapter 10: Shaping the Future of Additive Manufacturing: Twelve Themes from Bio-Inspired Design and Green Chemistry
10.1 Introduction
10.1.1 Disruptive Revolution of Additive Manufacturing
10.1.1.1 Basic Types
10.1.1.2 Historical Trend of the Industry
Rapid Market Growth Continues
Opening Up of the Research and Development Market
Wider Materials Choice
Shifting Products Mix: Prototyping Versus Manufacturing
Wider Spread Through More Manufacturing Sectors
10.1.1.3 Impacts and Implications
Mass Customization
Decentralization of Production
Savings in Material and Energy
Structural, Material, and Design Innovation
Cautions About Health, Safety, and Sustainability
10.1.2 Bio-inspired Design
10.1.2.1 Definition
10.1.2.2 Applications/State of the Industry
10.1.3 Green Chemistry
10.1.3.1 Definition
10.1.3.2 Applications/State of the Industry
10.1.4 Where These Three Realms Converge
10.1.5 Twelve Themes That Could Change the Way AM is Developed
10.1.5.1 Unity Within Diversity: Minimum Parts for Maximum Diversity
10.1.5.2 Systems Approach: Relationships Matter
10.1.5.3 The Optimal Activator: the Environment is the Catalyst
10.1.5.4 Taking Advantage of Gradients: Making Delta Do Work
10.1.5.5 Shape is Strength
10.1.5.6 Self Organization
10.1.5.7 Bottom-Up Construction
10.1.5.8 Hierarchy Across Linear Scales
10.1.5.9 Functionally Graded Material
10.1.5.10 Composite Construction
10.1.5.11 Controlled Sacrifice
10.1.5.12 Water is the Universal Medium
10.2 Conclusion
References
11
Chapter 11: The IFF Green Chemistry Assessment Tool
11.1 Introduction
11.2 Sustainability: An IFF Commitment
11.3 The IFF Green Chemistry Assessment Tool: Requirements
11.4 The 12 Principles of Green Chemistry
11.5 The IFF Green Chemistry Assessment Tool: Scoring and Analysis
11.6 Illustrative Example: Veridian
11.6.1 Veridian: Description of the Technology
11.6.2 Step 1: Development of a Practical Continuous Flow Technology for Grignard Addition
11.6.2.1 Original Process
11.6.2.2 Assessment (Table 11.1, Figure 11.2)
11.6.2.3 Improved Process
11.6.3 Step 2: Development of Air Oxidation Technology for Conversion of Alcohol to Ketone
11.7 Summary
References
3527326456_bindex

Citation preview

V

Contents About the Editors XIII List of Contributors XV Preface XIX

1

Application of Life Cycle Assessment to Green Chemistry Objectives 1 Thomas E. Swarr, Daniele Cespi, James Fava, and Philip Nuss

1.1 1.2 1.2.1 1.2.2 1.3 1.3.1 1.3.2 1.3.3 1.3.4 1.4 1.4.1 1.4.1.1 1.4.2 1.4.2.1 1.4.2.2 1.4.2.3 1.4.2.4 1.5

Introduction 1 Substitution of Safer Chemicals 4 Missing Inventory Data and Characterization Factors 4 Linking LCA and Chemical Risk 5 Design Material and Energy-Efficient Processes 7 Introduction 7 System Boundaries and Design Guidance 8 Impact Categories and Green Metrics 10 Policy Implications 12 Promote Renewable Materials and Energy 13 Introduction 13 Glycerol Case Study 13 Biochemicals Production 16 Life Cycle Stages of Biochemical Production 16 Environmental Implications of Biomass Production 16 Carbon Accounting and Land Use Change 18 Global Availability of Arable Land 20 Conclusion and Recommendations 20 References 21

2

Shortcut Models Based on Molecular Structure for Life Cycle Impact Assessment: The Case of the FineChem Tool and Beyond 29 Stavros Papadokonstantakis, Pantelis Baxevanidis, Effie Marcoulaki, Sara Badr, and Antonis Kokossis

2.1

Introduction 29

VI

Contents

2.2 2.3 2.3.1 2.3.2 2.4 2.4.1 2.4.2 2.4.3 2.4.4 2.5

Concept and Development of the FineChem Tool 31 Illustrative Applications of the FineChem Tool 35 LCA Aspects of Solvent Selection for Postcombustion CO2 Capture (PCC) 35 Bio-Based Production of Platform Chemicals 36 Toward A New Group Contribution-Based Version of the FineChem Tool 37 Introduction to GC models 37 Development of GC-Based LCA Models 38 Screening for Substances with Desirable Properties 40 Illustrative Example of Screening Molecules 44 Conclusions and Outlook 46 References 46

3

Models to Estimate Fate, Exposure, and Effects of Chemicals 49 Rosalie Van Zelm, Rik Oldenkamp, Mark A.J. Huijbregts, and A. Jan Hendriks

3.1 3.2 3.3 3.4 3.4.1 3.4.2 3.5 3.6 3.6.1 3.6.2 3.7 3.7.1 3.7.2 3.7.3

Introduction 49 Fate 50 Ecological Exposure 52 Ecosystem Effects 54 Intraspecies Variability in Populations 54 Interspecies Variability in Assemblages 55 Human Exposure and Effect 55 Environmental Impact Evaluation 58 Life Cycle Assessment 58 Risk Assessment 61 Recent Developments 62 New Chemicals 62 Nontoxic Stressors 63 Uncertainty and Variability 64 References 65

4

Collaborative Approaches to Advance Chemical Safety Philip Judson

4.1 4.2 4.3 4.3.1 4.3.2 4.3.3 4.3.4 4.4 4.5 4.5.1 4.5.2

Introduction 71 Incentives for Collaboration and Constraints 72 Options for Sharing 74 Sharing Research 74 Sharing Knowledge 75 Sharing Data 76 Sharing Software Development 77 The Implementation of Collaborative Organizations 78 Collaborative Projects 81 British Industrial Biological Research Association (BIBRA) 81 The Chemical Bioactivity Information Centre (CBIC) 84

71

Contents

4.5.3 4.5.4 4.5.5 4.5.5.1 4.5.5.2 4.5.5.3 4.5.5.4 4.5.5.5 4.5.5.6 4.5.5.7 4.5.6 4.5.7 4.5.8 4.5.9 4.5.10 4.5.11 4.5.12 4.5.13 4.5.13.1 4.5.13.2 4.5.13.3 4.5.13.4 4.5.13.5 4.5.13.6 4.5.14 4.5.15 4.5.16

4.5.17 4.5.18 4.5.19 4.6

The Distributed Structure-Searchable Toxicity Database Network – DSSTox 84 ICH 85 Innovative Medicines Initiative (IMI) 85 CHEM21 86 Electronic Health Record for Clinical Research (EHR4CR) 87 eTOX 87 GETREAL 87 iPiE 88 MARCAR 88 MIP-DILI 88 International Life Sciences Institute (ILSI) and ILSI Health and Environmental Sciences Institute (HESI) 89 Lhasa Limited 90 OECD (Q)SAR Toolbox 91 OpenTox 92 PhUSE 93 The Pistoia Alliance 93 REACH Substance Information Exchange Forums (SIEF) 93 SEURAT-1 (Safety Evaluation Ultimately Replacing Animal Testing) 94 COSMOS 94 DETECTIVE 94 HeMiBio 95 NOTOX 95 SCR&Tox 95 ToxBank 95 ToxML 95 The Traditional Chinese Medicine Database 96 United Nations – the European Agreement Concerning the International Carriage of Dangerous Goods by Road (ADR) and the Globally Harmonized System of Classification and Labeling of Chemicals (GHS) 96 US Government–Industry Collaborations 97 VEGA 98 Yale University Open Data Access (YODA) 98 Conclusions 99 References 99 103

5

Introduction to Green Analytical Chemistry Marek Tobiszewski

5.1 5.1.1 5.1.2

Introduction 103 Defining Green Analytical Chemistry 103 Dualistic Role of Analytical Chemistry in Relation to Green Chemistry 105 Brief History of Green Analytical Chemistry 105

5.1.3

VII

VIII

Contents

5.2 5.2.1 5.2.2 5.2.3 5.3 5.3.1 5.3.2 5.3.2.1 5.3.2.2 5.3.2.3 5.3.3 5.3.4 5.3.5 5.3.6 5.4 5.5

Greener Analytical Separations 107 Green Gas Chromatography 107 Greener Liquid Chromatography 107 Supercritical Fluid Chromatography 108 Green Sample Preparation Techniques and Direct Techniques 108 Direct Analytical Methods 108 Microextraction Sample Preparation Techniques 109 Solid-Phase Microextraction 110 Liquid-Phase Microextraction 110 Dispersive Liquid–Liquid Microextraction 111 Stir Bar Sorptive Extraction 112 Supercritical Fluid Analytical Extraction 112 Microwave- and Ultrasound-Assisted Extraction 112 Ionic Liquids in Extraction 113 Chemometrics for Signals Processing 114 Conclusions 114 References 115

6

Cosmo-RS-Assisted Solvent Screening for Green Extraction of Natural Products 117 Anne-Gaëlle Sicaire, Aurore Filly, Maryline Vian, Anne-Sylvie Fabiano-Tixier, and Farid Chemat

6.1 6.2 6.2.1 6.2.2 6.2.3 6.2.4 6.2.4.1 6.2.4.2 6.2.4.3 6.2.4.4 6.2.4.5 6.2.4.6 6.3 6.3.1 6.3.2 6.3.2.1

Introduction 117 Solvents for Green Extraction 119 Definition 119 Solute–Solvent Interaction 119 Substitution Concept 120 Panorama of Alternative Solvents for Extraction 121 Water: Solvent with Variable Polarity 121 Bio-Based Solvents 121 Solvent Obtained from Chemical Synthesis 123 Vegetable Oils 123 Eutectic Solvents 123 Supercritical CO2 124 Prediction of Solvent Extraction of Natural Product 124 COSMO-RS Approach 124 Applications of COSMO-RS-Assisted Substitution of Solvent 128 Example 1: COSMO-RS Assisted Selection of Solvent for Extraction of Seed Oils 129 Example 2: Cosmo-Rs-assisted Selection of Solvent for Extraction of Aromas 131 Conclusion 135 References 136

6.3.2.2 6.4

Contents

139

7

Supramolecular Catalysis as a Tool for Green Chemistry Courtney J. Hastings

7.1 7.2 7.2.1 7.2.2 7.2.3 7.3 7.3.1 7.3.2 7.4 7.4.1 7.4.2 7.5 7.5.1 7.5.2 7.6

Introduction 139 Control of Selectivity through Supramolecular Interactions 140 Catalysis with Supramolecular Directing Groups 141 Scaffolding Ligands 145 Selectivity through Confinement and Binding Effects 146 Reactions in Water 150 Water-Soluble Nanoreactors 150 Dehydration Reactions 156 Catalyst/Reagent Protection 158 Catalyst Protection 159 Protection of Water-Sensitive Reagents 159 Tandem Reactions 160 Synthetic Tandem Reactions 161 Chemoenzymatic Tandem Reactions 162 Conclusion 164 References 164

8

A Tutorial of the Inverse Molecular Design Theory in Tight-Binding Frameworks and Its Applications 169 Dequan Xiao and Rui Hu

8.1 8.2 8.2.1 8.2.2 8.2.2.1 8.2.2.2 8.2.3 8.3

Introduction 169 Inverse Molecular Design Theory in Tight-Binding Frameworks 170 LCAP Principle in Density Functional Theory 171 LCAP Principle in Tight-Binding Frameworks 172 One-Orbital Tight-Binding Framework 172 Extended Hückel Tight-Binding Framework 173 Gradient for Optimization 175 How to Prepare a Molecular Framework for TB-LCAP Inverse Design? 175 How to Choose Optional Atom Types or Functional Groups? 177 Optimizing Molecular Properties Using the TB-LACP Methods 182 Conclusion 186 References 187

8.4 8.5 8.6

9

Green Chemistry Molecular Recognition Processes Applied to Metal Separations in Ore Beneficiation, Element Recycling, Metal Remediation, and Elemental Analysis 189 Reed M. Izatt, Steven R. Izatt, Neil E. Izatt, Ronald L. Bruening, and Krzysztof E. Krakowiak

9.1 9.2

Introduction 189 Molecular Recognition Technology as a Green Chemistry Process 190

IX

X

Contents

9.3 9.3.1 9.3.2 9.3.2.1 9.3.2.2 9.3.2.3 9.3.2.4 9.3.2.5 9.3.2.6 9.3.2.7 9.3.3 9.3.3.1 9.3.3.2 9.3.4 9.3.4.1 9.3.4.2 9.3.4.3 9.3.5 9.3.6 9.3.7 9.3.8 9.3.9 9.3.10 9.3.10.1 9.3.10.2 9.3.11 9.3.12 9.3.12.1 9.3.12.2 9.3.12.3 9.3.12.4 9.3.13 9.3.14 9.3.14.1 9.3.14.2

Metal Separations Using Molecular Recognition Technology 194 Separation and Recovery of Individual Rare Earth Elements 194 Platinum Group Metals 196 General 196 Palladium Recovery from Native Ore 197 Rhodium Recovery from Spent Catalyst and Other Wastes 197 Platinum Recovery from Alloy Scrap 198 Ruthenium Recovery from Alloy Scrap 199 Iridium Separation from Rhodium and Base Metals 200 Purification of 103Palladium for Use in Brachytherapy 202 Gold Separation and Recovery from Process Streams 202 General 202 Gold Recovery from Plating Solutions 203 Nickel Separations and Recovery 204 Nickel Separations from Laterite Ores 204 Nickel, Aluminum, and Molybdenum Recovery from Acid Leachate of Spent Hydrodesulfurization Catalyst 205 Nickel Removal from Cadmium- and Zinc-Rich Sulfate Electrolyte 206 Cadmium Removal from a Cobalt Electrolyte Solution Containing a Complex Matrix 207 Bismuth and Antimony Removal from Copper Electrolyte in Production of High-Purity Copper 208 Cobalt Recovery from Zinc Streams using Iron(III) as a Pseudo-Catalyst 209 Molybdenum and Rhenium Separations 210 Indium Recovery from Etching Wastes 211 Separation of Indium and Germanium from Zinc Electrolyte Solutions 212 Indium Separation and Recovery 212 Germanium Separation and Recovery 213 Mercury Recovery from Sulfuric Acid Streams 213 Metal Recovery from Acid Mine Drainage Streams, Industrial Waste Streams, Mine Leach Streams, and Fly Ash 214 Metal Remediation from Berkeley Pit Acid Mine Drainage Site 214 Removal, Separation, and Recovery of Heavy Metals from Industrial Waste Streams using MRT 216 Uranium Separation and Recovery from Mine Leach Streams 217 Lead Separation from Fly Ash Generated by Ash Melting 218 Lithium Separation and Recovery from Brine and End-of-Life Rechargeable Batteries 219 Radionuclide Remediation 220 General 220 Cesium Separation and Recovery from Savannah River Nuclear Wastes 220

Contents

9.3.14.3 9.3.14.4 9.3.14.5 9.4 9.4.1 9.4.2 9.4.2.1 9.4.2.2 9.4.2.3 9.4.3 9.4.4 9.4.4.1 9.4.4.2 9.4.4.3 9.4.5 9.4.6 9.5

Cesium and Technetium Separation and Recovery from Nuclear Wastes at Hanford, Washington 221 Cesium Separation and Recovery from Fly Ash 222 Separation and Recovery of Radioactive Cesium and Strontium from Fukishima, Dai’ichi, Japan Harbor 225 Analytical Applications of Molecular Recognition Technology 227 General 227 Radionuclides 229 Strontium Separation and Analysis using EmporeTM Strontium Rad Disks 229 Radium Separation and Analysis Using EmporeTM Radium Rad Disks 229 Other Radionuclide and Mixed Waste Separations 230 Precious Metals 230 Toxic Metals 231 Arsenic Separation and Analysis 231 Lead Separation and Analysis 231 Mercury Separation and Analysis 231 Rare Earth Metal Separation and Analysis from Rainfall 232 Multimetal Separations and Recovery 233 Conclusion 233 References 234

10

Shaping the Future of Additive Manufacturing: Twelve Themes from Bio-Inspired Design and Green Chemistry 241 Thomas A. McKeag

10.1 10.1.1 10.1.1.1 10.1.1.2 10.1.1.3 10.1.2 10.1.2.1 10.1.2.2 10.1.3 10.1.3.1 10.1.3.2 10.1.4 10.1.5 10.1.5.1 10.1.5.2 10.1.5.3 10.1.5.4 10.1.5.5 10.1.5.6

Introduction 241 Disruptive Revolution of Additive Manufacturing 241 Basic Types 241 Historical Trend of the Industry 243 Impacts and Implications 245 Bio-inspired Design 249 Definition 249 Applications/State of the Industry 249 Green Chemistry 250 Definition 250 Applications/State of the Industry 250 Where These Three Realms Converge 250 Twelve Themes That Could Change the Way AM is Developed 251 Unity Within Diversity: Minimum Parts for Maximum Diversity 251 Systems Approach: Relationships Matter 252 The Optimal Activator: the Environment is the Catalyst 253 Taking Advantage of Gradients: Making Delta Do Work 254 Shape is Strength 254 Self Organization 255

XI

XII

Contents

10.1.5.7 10.1.5.8 10.1.5.9 10.1.5.10 10.1.5.11 10.1.5.12 10.2

Bottom-Up Construction 256 Hierarchy Across Linear Scales 256 Functionally Graded Material 257 Composite Construction 257 Controlled Sacrifice 258 Water is the Universal Medium 259 Conclusion 260 References 260

11

The IFF Green Chemistry Assessment Tool Geatesh Tampy

11.1 11.2 11.3 11.4 11.5

Introduction 263 Sustainability: An IFF Commitment 264 The IFF Green Chemistry Assessment Tool: Requirements 265 The 12 Principles of Green Chemistry 266 The IFF Green Chemistry Assessment Tool: Scoring and Analysis 267 Illustrative Example: Veridian 268 Veridian: Description of the Technology 269 Step 1: Development of a Practical Continuous Flow Technology for Grignard Addition 270 Original Process 270 Assessment 270 Improved Process 270 Step 2: Development of Air Oxidation Technology for Conversion of Alcohol to Ketone 274 Summary 275 References 276

11.6 11.6.1 11.6.2 11.6.2.1 11.6.2.2 11.6.2.3 11.6.3 11.7

Index 277

263

HGC113

03/27/2018

13:3:24

Page 1

1

1 Application of Life Cycle Assessment to Green Chemistry Objectives Thomas E. Swarr, Daniele Cespi, James Fava, and Philip Nuss

1.1 Introduction

Green chemistry (GC) is described by the 12 principles of green chemistry to guide the design of chemical products and processes that reduce or eliminate the generation and use of hazardous substances [1]. The guiding principles have been criticized for being qualitative and failing to provide an objective means to assess the overall “greenness” of proposed solutions or to evaluate trade-offs among conflicting principles, for example, reduced toxicity, but increased energy consumption [2]. Life cycle assessment (LCA) is the “compilation and evaluation of the inputs, outputs, and the potential environmental impacts of a product system” and provides a quantitative method to address these concerns (3], p.2). It is an international standard recognized as an effective methodology to evaluate improvement strategies and avoid shifting problems to other times and places or among various environmental media [4,5]. LCA, however, has its own set of limitations and unresolved methodology issues [6,7]. Some are particularly relevant to green chemistry, such as limited data for chemical production chains, lack of geographic specificity, and aggregation of emissions over time [8–10]. Increasingly, researchers are recognizing that the strengths and weaknesses of GC and LCA are complementary and are advocating for more effective integration of both methodologies to develop more sustainable solutions [2,11,12]. Anastas and Lankey (11], p.289) broaden the definition of green chemistry by considering chemistry to include “the structure and transformation of all matter” and hazardous impacts to address the “full range of threats to human health and the environment.” Application of LCA to GC problems promises a better understanding of the flow of toxics through the economy and provides a robust framework for organizing knowledge about inherent hazards associated with product systems [13]. LCA is comprised of four basic steps [3]. Goal and scope identifies the purpose of the study, how the results will be used, and intended audience to whom the

Handbook of Green Chemistry Volume 10: Tools for Green Chemistry, First Edition. Edited by Evan S. Beach and Soumen Kundu.  2017 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2017 by Wiley-VCH Verlag GmbH & Co. KGaA.

HGC113

03/27/2018

13:3:24

2

Page 2

1 Application of Life Cycle Assessment to Green Chemistry Objectives

results will be communicated. Clear definition of the decision context is critical to ensure the study provides objective information that enables the study commissioner and intended audience to make informed choices according to their values and priorities. Life cycle inventory (LCI) gathers data necessary to model mass and energy flows across the entire product system, from extraction or harvest of resources to the ultimate disposal of the product. Realistically, these models are always incomplete and a key decision is where to draw the boundaries on what is included in the system model. Life cycle impact assessment (LCIA) evaluates the significance of exchanges between the product system and the natural environment. Environmental interchanges are grouped, or classified into impact categories, such as acidification, climate change, human toxicity, or resource depletion. The inventory items for each impact category are then characterized for potency and mass, often in terms of a reference substance. For example, different greenhouse gases have different warming potentials, and are converted to an equivalent mass of CO2, allowing the aggregated emission to be expressed as CO2(e) (equivalents). Finally, interpretation attempts to make sense of the analytical results to provide conclusions and recommendations necessary to satisfy the intended goal and scope of the study. For additional background on LCA methods relevant to GC, the reader is referred to previous reviews. Kralisch et al. [12] provides a general overview of LCA, specific considerations to be considered in chemical design, and examples of applications to emerging research problems. Tufvesson et al. [9] reviews “green chemistry” LCA studies to identify key parameters and methodological issues. Principles of Green Chemistry 1) 2) 3) 4) 5) 6) 7) 8) 9) 10) 11) 12)

Prevention Atom economy Less hazardous chemical syntheses Designing safer chemicals Safer solvents and auxiliaries Design for energy efficiency Use of renewable feedstocks Reduce derivatives Catalysis Design for degradation Real-time analysis for pollution prevention Inherently safer chemistry for accident prevention

The present chapter is intended to provide guidance on the effective integration of LCA methods into GC initiatives. It is assumed that the primary audience is a practitioner in green chemistry, familiar with the 12 principles of green chemistry, and perhaps with only a basic understanding of LCA methods.

HGC113

03/27/2018

13:3:24

Page 3

1.1 Introduction

(Because the principles are sometimes listed differently, the American Chemistry 1) Society version is reproduced in the text box.) There is no attempt to provide a comprehensive review of current state of the art in green chemistry progress or of the latest developments in LCA. Illustrative case studies or research results are presented to emphasize key points in the application of LCA methods to GC objectives. The chapter is organized based on grouping GC objectives into the following three overarching categories:

 Substitution of hazardous chemicals with safer alternatives (Principles 3,4,5,10,11,12).

 Design of processes to be more energy and material efficient (Principles 1,2,6,8,9,11).

 Promoting a transition to renewables (materials and energy) (Principles 6,7). Further, Sjöström [14] characterized GC as a meta-discipline and described a classification model of GC research, management, and policy activities. Thus, the applicability of LCA to these various activities is also addressed for the various GC objectives. Section 1.2 on substitution addresses the challenge of quantifying the toxicity of chemicals and the products and processes that depend on those chemicals. There are fundamental differences between LCA and GC or chemical alternative assessments. While these are typically focused on use of a chemical to provide a required function in a specific application, LCA considers use of the chemical (along with other chemicals) in a product system designed to satisfy some end user need or function. Thus, the two methodologies are designed to answer different questions, and this must be considered in the application of LCA to quantify toxicity concerns. Section 1.3 on greener processes builds on this introduction to consider a broader range of environmental impacts to assess trade-offs among the various GC principles and evaluate the overall greenness of a product or process. LCA provides a robust method to quantitatively compare alternative solutions, but does not provide guidance on development of alternatives. GC complements LCA by providing specific guidance to address the issues identified in the study. Section 1.4 addresses the broad goal of promoting renewable materials and energy. This class of problems extends system considerations to include a wide range of environmental impacts and dynamic effects, such as indirect land use changes that go well beyond traditional GC assessments. Given the importance of climate change concerns, determining whether renewables provide a real benefit over synthetic alternatives is a critical area for future work. Finally, Section 1.5 concludes with recommendations to promote the effective integration of LCA and GC to develop more sustainable business practices.

1) Source: http://www.acs.org/content/acs/en/greenchemistry/what-is-green-chemistry/principles/ 12-principles-of-green-chemistry.html accessed Nov. 2015).

3

HGC113

03/27/2018

13:3:24

4

Page 4

1 Application of Life Cycle Assessment to Green Chemistry Objectives

1.2 Substitution of Safer Chemicals

A variety of chemical alternative assessment procedures have been developed to guide selection of safer substitutes [15–17]. These methods typically begin with identifying a target chemical of concern followed by an analysis of the uses of the chemical. Virtually all methods advocate “life cycle thinking,” but the focus on a specific chemical can narrow the problem definition. Alternatives are identified based on satisfying the same technical requirements for the application, or use under consideration. LCA, by contrast, has traditionally been focused on product systems and the function or service provided to the customer or end user. This broader perspective can inspire innovative approaches to satisfy the end user demand with an alternative solution that does not rely on the chemical of concern, thus eliminating the need for an alternatives assessment. If the assessment cannot identify an alternative that satisfies the technical requirements, then the analyst is advised to implement best practices to limit human exposures and environmental releases and to continue to research alternatives. The use of comprehensive LCA studies in early development stages is often dismissed as being overly complex and time and effort intensive [12]. Problems gathering inventory data for chemicals, and particularly for fine chemicals have been well documented [9,18]. Inventory data is often protected as proprietary business information. Fine chemicals tend to be produced in smaller batches, comprised of many processing steps, and produced in shared equipment in multiproduct facilities, with much data collected only at the facility level. A case study comparing alternative assessment tools for the characterization of organic solvents concluded use of LCA was limited due to data constraints that included both inventory data for the production of chemicals and characterization factors for toxicity of chemicals [19]. 1.2.1 Missing Inventory Data and Characterization Factors

Researchers have developed methods to fill inventory data gaps in chemical production systems using basic knowledge of chemical processes and proxy data based on molecular structure. One method was designed specifically to rely on information obtainable from the open literature combined with knowledge of only a few key process characteristics and a set of default estimates for all parameters [20]. A generic input–output process step was used to develop a set of equations to define a mass balance of reactants and products. Onsite production data and pilot scale data from a facility in Switzerland were used to develop default estimates for any missing parameters. Another approach developed by GlaxoSmithKline built a chemical tree of all the process materials used in production of an active pharmaceutical ingredient combined with heuristics to build gate-to-gate inventories [21]. Yet another group evaluated mass and energy flows on the petrochemical production of 338 chemicals to develop models that could

HGC113

03/27/2018

13:3:24

Page 5

1.2 Substitution of Safer Chemicals

estimate key production parameters based on molecular structure [22]. 10 descriptors, including molecular weight, number of functional groups, number of aromatic or aliphatic rings, and others were used to predict cumulative e demand (CED), the global warming potential (GWP), and the Eco-indicator 99 [23]. A tiered approach using extrapolations from existing data, substitution with generic datasets, molecular structure models, and process- based estimation methods were recommended to fill inventory gaps [10]. The life cycle inventory can then be translated to impacts using a variety of impact assessment methods, and the various methods can yield a wide range of results using differing characterization factors for a variety of toxicity endpoints. Wide variation of characterization factors for the same chemicals determined by these various impact assessment methods were recognized as a key challenge for application of LCA to the study of chemical systems. Under the Life Cycle Initiative of the United Nations Environment Programme and the Society of Environmental Toxicology and Chemistry (UNEP/SETAC LCI), a workgroup was formed to develop a consensus model for the impact assessment of chemicals [24]. Seven different models were used to determine characterization factors for a set of 45 chemicals chosen to cover a wide range of property combinations, including environmental partitioning, exposure pathways, overall persistence, long-range transport in air, and the importance of feedback between environmental media [25]. The comparison was used to identify the most important parameters and reasons for differing results. These results were then used to build a new multimedia, multipathway model that links emissions to impacts through environmental fate factors (FF), exposure factors (XF), and effects factors (EF) to calculate characterization factors for human toxicity and freshwater ecotoxicity. Human toxicity factors are expressed as number of cases per kg of chemical emitted, and ecotoxicity factors as potential affected fraction of specifies in a volume of environmental media per kg of chemical emitted. USEtox models urban air, rural air, agricultural soil, industrial soil, freshwater, and coastal marine water environmental compartments at a continental scale and all but urban air at the global scale. The general framework of the model is shown in Figure 1.1, and additional detail is available from a public web page to disseminate and continually improve the model (USEtox.org). 1.2.2 Linking LCA and Chemical Risk

While the USEtox dramatically reduced the intermodal variation of characterization factors from an initial range of up to 13 orders of magnitude down to no more than two orders of magnitude, this still represents a large band of uncertainty for impact assessment [25]. There are fundamental limitations in applying LCA to assess the chemical risk of alternative chemicals. Although the calculation procedures are similar – estimate emissions, model chemical fate and distribution in various environmental media or compartments, determine concentrations and effects, the calculations are used for different purposes [26]. LCIA aggregates best

5

HGC113

03/27/2018

13:3:24

6

Page 6

1 Application of Life Cycle Assessment to Green Chemistry Objectives

Figure 1.1 Modeling framework for USEtox.

estimates of emissions over time and space to provide a reasonable comparison of alternatives. Midpoint characterization methods link LCI emissions data to an intermediate point in the causal pathway and expresses all emissions in terms of an equivalent unit or emission, such as 1,4-dichlorobenzene equivalents. Endpoint oriented LCIA methods model the cause–effect chain up to the potential damages to human health, ecosystem health, and resources to translate LCI data to quantitative indicators of damage, such as disability adjusted life years (DALYs). Chemical risk assessment uses worst-case assumptions to develop recommendations for risk mitigation methods intended to reduce exposures to a level that results in no observable effects [26]. Thus, risk assessment and LCA answer different questions, and using them in a complementary manner may be more productive than intensive effort to improve the precision of LCA toxicity impact assessments. Even with the current level of uncertainty in toxicity characterization factors, LCIA is adequate to “. . . identify 10–30 chemicals to look at in priority and perhaps, more importantly, to disregard 400 other substances whose impacts are not significant for the considered application (25], p. 544).” Kuczenski et al. [13] have argued LCA could be made more “toxics aware” by explicit modeling of the intermediate flows between unit processes in the system model. Normally, in LCA the intermediate flows are balanced out to yield only the elemental flows to and from nature (i.e., those flows crossing the system boundary) that are used in the impact assessment methods. If toxicity information were attached to the intermediate flows, LCA practitioners could then establish a relationship between the use of a toxic chemical and the function of a product system. The real value for LCA may be in developing a better

HGC113

03/27/2018

13:3:25

Page 7

1.3 Design Material and Energy-Efficient Processes

understanding of the flow of toxics through the product system. In fact, studies have shown that a facility focus on evaluating green chemistry can give misleading results due to outsourcing of process complexity and toxicity to upstream processes, and that a life cycle perspective is critical [27]. One attempt to integrate toxicity information into LCA process models used the risk phrases (Rphrases) as defined in Annex III of European Union Directive 67/548/EEC [28]. A screening tool was developed using SimaPro LCA software. R-phrases were entered for all substances that exceeded concentration limits based on legal requirements for substance and product classification and labeling. The tool did not calculate a score, but simply compiled the data into two lists. An exposure pathway indicator provided information on the most important pathways that in turn identified the recommended risk management measures. A hazard indicator provided information about the need for hazard labeling. Another study characterized wastewater toxicity caused by detergents using data made available through REACH [29]. There are relatively few LCA studies focused strictly on toxicity factors given the limitations already mentioned. It has been used much more frequently to develop a more holistic assessment of a process and to evaluate potential tradeoffs of alternative assessments. For example, solvent use is a significant contributor to the environmental impacts for chemical processes. A variety of studies have developed guidance for solvent selection based on LCA, often evaluating cumulative energy demand, global warming potential and other impacts, as well as toxicity impacts [30,31]. Use of LCA to determine the overall greenness of a process is discussed in the next section.

1.3 Design Material and Energy-Efficient Processes 1.3.1 Introduction

Increasing environmental awareness has pressured companies to become more proactive in addressing public concerns that have expanded beyond production facilities to include all environmental effects of products during use and disposal. Corporate environmental programs have evolved from reactionary and compliance-focused efforts on end-of-pipe controls to limit environmental effects of their production facilities to forward- looking strategies to design inherently safe and green products and processes. However, the definition of what is inherently green is controversial and depends on stakeholder values and priorities and can vary across different product sectors. It is a long and complex procedure that needs a variety of analytical methods and must take into account all the life cycle considerations [32,33]. The mere application of GC principles is not sufficient to achieve a benign design. The design optimization procedure is driven by the application of the

7

HGC113

03/27/2018

13:3:25

8

Page 8

1 Application of Life Cycle Assessment to Green Chemistry Objectives

fundamental principles, but must consider the entire life cycle of a product [34]. The application of the LCA methodology is recommended as a support and screening tool to identify and quantify opportunities to reduce the environmental impact of products and processes during the design conception process. The value of LCA and life cycle thinking can be attributed to two core aspects – consideration of a broader technical system and a more comprehensive range of environmental impacts. However, both aspects create additional information and computational challenges that can conflict with goals of applying LCA in early design stages. 1.3.2 System Boundaries and Design Guidance

Consideration of the full life cycle is critical to avoid alternatives that impose unintended consequences or simply shift problems to a different place or time. Dichlorodifluoromethane or Freon-12 was originally introduced as a breakthrough safety innovation for refrigerant applications, long before its ozone damaging effects were recognized and made manifest by use as a propellant for aerosols [35]. It is much more cost effective to avoid problems by their identification early in the design than to remediate problems after a product has been put on the market. It has also been argued that aggregating life cycle inventory data across the full product system is necessary to develop a better understanding of the potential impacts of current global supply chains [36]. However, there is limited guidance available for defining appropriate system boundaries, and depending on the goal of the study, impacts can have different boundaries [37]. Further, the design process addresses both the product and the associated manufacturing processes for that product. The product and associated processes each have distinct life cycles imposing specific considerations for LCA [38]. The demands of a holistic life cycle perspective must be balanced against the constraint of limited information and time to integrate life cycle considerations in early design phases. Zheng et al. [39] developed a framework for incorporating sustainability into the conceptual design stage for chemical process development using a waste reduction algorithm. The algorithm was based on the mass and energy balance, and evaluated eight impact categories dependent on chemical properties to assess human and ecological toxicity, and atmospheric impacts for ozone depletion, global warming, acidification, and photochemical oxidation. Other researchers have attempted to use results from early laboratory experiments or pilot scale studies to project the potential impacts of full-scale production systems. Earhart et al. [40] translated data from laboratory experiments in terms of environmental impacts in order to verify the feasibility behind the use of the new starting raw material (e.g., the use of fructose to produce polyethylene furandicarboxylate-PEF). However, a review of life cycle process design concluded that advanced process development activities, such as pilot scale facilities did not yield the type and quality of data required for LCA, and recommended enhanced

HGC113

03/27/2018

13:3:25

Page 9

1.3 Design Material and Energy-Efficient Processes

collaboration between researchers and process engineers to improve data availability [41]. Another approach to simplify application of LCA in early process development has been to focus on a specific problem, and use detailed LCA studies to develop recommendations for improved design. For example, the chemical industry uses a wide range of organic solvents having properties of volatility, persistence, and toxicity that make them a priority for environmental assessments [30]. Capello et al. [31] developed a framework for assessing solvents using a simplified environmental, health and safety (EHS) screening combined with LCA. The EHS screening included a qualitative measure based on nine hazard categories: release potential, fire/explosion safety, reaction/decomposition safety, acute toxicity, irritation toxicity, chronic toxicity, persistency, air pollution, and water pollution. LCA studies were based on the combined LCIs for the petrochemical production of 45 organic solvents. The framework was then demonstrated on 26 pure organic solvents and several alcohol–water mixtures. Amelio et al. [30] expanded on their work to develop guidelines applicable in the early stages of process design that would enable the choice of solvent and the best treatment method (incineration or distillation), based on the composition of the chemical solvent. Their results demonstrated the importance of a full life-cycle perspective. The decision to select incineration or distillation was dependent on the environmental impact originating from the production of the solvents. Normalization and Weighting LCIA results can be normalized and weighted to yield a single score metric of the overall impact or greenness. These are optional elements of LCIA to calculate the magnitude of category indicator results relative to a specified reference value (normalization), and to convert and possibly aggregate indicator results across impact categories using quantitative factors based on importance (weighting) [42]. Normalization and weighting are inherently subjective, reflecting the values and priorities of the stakeholder sponsoring or conducting the study. For that reason, these are optional elements. Criteria used for normalization and weighting should be transparent, and the underlying data should be available so that other stakeholders can make independent assessments based on their values and priorities

Yet another approach to address data limitations is to develop qualitative or semi quantitative approaches to streamline the LCA. BASF developed an ecoefficiency tool that characterized chemical products for material and energy consumption, emissions, toxicity potential, and risk potential [43]. Streamlined methods were developed for the various impact categories and transparent methods to normalize and weight the data, allowing results to be displayed in simple “spider charts” (a type of two-dimensional chart that allows display of three or more quantitative variables plotted on axes starting from the same

9

HGC113

03/27/2018

13:3:25

10

Page 10

1 Application of Life Cycle Assessment to Green Chemistry Objectives

origin). Thus, the tool was effective for informing design as well as communicating results to customers and other stakeholders. The tool was further enhanced to the SEEbalance® instrument [44] that combined the eco-efficiency analysis principle with a social LCA perspective [45]. The eco-efficiency analysis was expanded to include land use impacts, and social impacts were assessed for employees, business partners, end users/consumers, the international community, society, and future generations. Results were displayed in separate charts for eco-efficiency and socio-efficiency, or combined into a three-dimensional cube for graphical displays easy to communicate to different audiences. SEEbalance® represents a successful example of how the whole life cycle perspective can be applied to the management routines in order to improve product performance and capture market advantage with effective communications. Chimex, a subsidiary of L’Oréal, launched the Eco-footprint tool in 2014 to support a corporate initiative named Made in ChimexTM that was aimed at making social responsibility central to business strategy [46]. The tool rated 10 factors grouped under ecodesign and manufacturing on a scale of 1–4. A variety of streamlined metrics and qualitative measures were defined for each impact category. The tool provided effective guidance for design and presented results in an easy to communicate format. GlaxoSmithKline (GSK) conducted a detailed assessment of the cradle-togate life cycle environmental impacts associated with the manufacturing of materials used in a typical pharmaceutical process to develop a streamlined tool for the Fast LCA of Synthetic Chemistry, FLASCTM [47]. Inventory data were generated for some 140 chemicals and were collated for eight impact categories. Statistical analysis then grouped the chemicals into 14 material classes that allowed generation of average life cycle impact profile data that could be used for materials missing LCI data. The FLASCTM evaluations were later combined with a health score to develop guides ranking commonly used reagents for 15 transformations designed to reduce the environmental impact of drug discovery and development [48]. 1.3.3 Impact Categories and Green Metrics

It has been noted that the evolution of LCA in pharmaceutical and chemical applications has been to reduce the level of detail and extend the system boundary [37]. There has also been a push to develop simple green metrics to simplify integration of life cycle insights into routine decision-making processes [49,50]. Thus, there is a continual tension between abbreviated approaches to push greater integration of life cycle approaches and more holistic assessments to avoid burden shifting and/or unintended consequences. The examples already discussed highlight that different sectors and different classes of products have different priorities and hence methods need to be tailored to the specific context. It has further been stressed that system boundaries and other details of the study should be defined with the goal and scope of the study in mind. In many cases,

HGC113

03/27/2018

13:3:25

Page 11

1.3 Design Material and Energy-Efficient Processes

simple green metrics are valid and practical, but should be supported with strategic use of LCA to define the limitations of their use. Simplified green metrics are grounded in the observation that qualitative assessments or studies involving a limited set of impact categories are sufficient in many cases to identify the key drivers of environmental impact. For example, in a study of catalytic methods to avoid phosphine oxide waste products in phosphorous-consuming pathways, cumulative energy demand was used as a proxy for total environmental burden in combination with greenhouse gas emissions [51]. One study evaluated various green metrics – reactant stoichiometry, yield, atom economy, carbon efficiency, reaction mass efficiency, mass intensity (excluding water), and mass productivity for a series of reactions [49]. The study concluded that while some of the metrics were useful as organizing concepts or for communicating with business managers, none captured the range of issues necessary to ensure sustainable solutions. The pharmaceutical industry, through the American Chemical Society Green Chemistry Institute Pharmaceutical Roundtable selected process mass intensity (PMI) as the key mass-based green metric [50]. PMI is defined as the total mass of materials used to produce a specified mass of product and is given by Eq. (1.1). PMI ˆ

total mass in a process or process step…kg† mass of product…kg†

(1.1)

Choice of the PMI metric was argued as necessary to truly integrate green chemistry and engineering into chemical processes: Considering inputs, PMI is a leading metric to facilitate changes as the processes and synthesis routes are being designed and tested. It was contrasted with E-factor, a metric that focuses on waste generated per unit of product, as shown in Eq. (1.2). E-factor ˆ

total mass of waste…kg† mass of product…kg†

(1.2)

E-factor was considered to be a legacy of end-of-pipe waste management approaches of the 1980s. The philosophical difference of these metrics reflects a broader discussion of LCIA methods and sustainability objectives. LCIA and sustainability are concerned with evaluating potential constraints imposed on human industrial systems by the natural environment. These constraints can be due to resource depletion (limits of nature to provide basic materials and fuels used by the economy) or due to damages to human and/or ecosystem health (limits of nature on absorbing wastes thrown off by the economy). It is important to take a life cycle perspective that considers the effects of both inputs and outputs. There is a tension between simple metrics or abridged LCA methods intended to promote use in business decision making processes and preserving the value of a comprehensive LCA. There are, however, inherent limitations to LCA methods yielding precise assessments of the greenness of production systems. LCA models assume static conditions, and thus will always need to be used in conjunction with more detailed process modeling to evaluate optimum operating

11

HGC113

03/27/2018

13:3:25

12

Page 12

1 Application of Life Cycle Assessment to Green Chemistry Objectives

conditions and to assess dynamic effects. In addition, even the simplest LCA could have hundreds of inventory items, making interpretation of the results extremely difficult. Classifying inventory results into a handful of impact categories greatly aids the interpretation of results and facilitates identification of trade-offs, but even interpreting and communicating LCIA results can be challenging and numerous formats have been suggested to aid the process [52]. The real value of LCA is providing a better understanding of the broader system within which the specific problem or project is encapsulated, identifying and quantifying key trade-offs, promoting improved communication and collaboration across functional units within the company and external stakeholders, and developing insights about actionable changes to improve the product or process [37,38,41]. 1.3.4 Policy Implications

The studies already discussed typically invoke an assumption of ceteris paribus, or all other things being equal. Even though the studies promoted a broader consideration of the full product system life cycle, the scope was focused on improvement actions of a specific company or organization and did not take into account broader market changes that might occur, such as substitution effects, economies of scale, and elasticity of supply and demand [53]. Thus, when considering broader sustainability initiatives or public policy reform, researchers have argued for a consequential LCA (C-LCA) that models the indirect changes induced by the proposed initiative [54,55]. The more common accounting or attributional LCA (A-LCA) uses average data to make relative assessments of environmental performance, while C-LCA uses marginal data and broader system boundaries that include indirectly impacted processes to assess “. . . how flows to and from the environment will change as a result of different potential decisions (54], p. 856).” A number of approaches have been proposed for integration of economic modeling with LCA to account for these broader market shifts [53,56]. Plevin et al. [57] argued that use of A-LCA to estimate climate change mitigation benefits can mislead policy makers. Deciding when and how to apply CLCA versus A-LCA is an ongoing debate. What is clear, however, is that decisions will be more resilient if based on assessment of a wider range of plausible scenarios [57]. But the critical question remains – how does one decide which technologies or processes might be affected by the decision under consideration? Weidema et al. [58] proposed a series of questions to help identify affected technologies and provided examples of how the questions can be applied across various sectors. 1) What time horizon does the study apply to? 2) Does the change only affect specific processes or a market? 3) What is the trend in the volume of the affected market?

HGC113

03/27/2018

13:3:25

Page 13

1.4 Promote Renewable Materials and Energy

4) Is there potential to provide an increase or reduction in production capacity? 5) Is the technology the most/least preferred? For example, the ban or lead solder to reduce toxic emissions of lead to landfills was expected to result in broad shifts, raising important questions about what happens to the lead no longer used in solder, as well as questions concerning trade-offs attributable to the higher reflow temperatures for lead-free alternatives [59]. These broader issues are particularly relevant for policies to promote a shift to renewable feedstocks, discussed in the following section. It is obvious that integrating necessary economic and social science disciplines to combine GC principles within a C-LCA perspective during early design stages is a difficult task that remains very much a work in progress.

1.4 Promote Renewable Materials and Energy 1.4.1 Introduction

The use of renewable feedstocks is one of the 12 principles of green chemistry [1], and increasing the use of biomass for the production of fuels, energy, and chemicals is seen by many as an important strategy toward sustainable development [60]. In 2012, the United States and Europe communicated their intentions to grow their bio economies [61,62]. In addition, many other countries now have bioeconomy strategies in place [63]. Globally, bioenergy is expected to contribute about one-quarter (138 EJ) of primary energy demand based on various renewable energy scenarios [64]. The share of biochemicals is foreseen to increase globally from about 3–4% in 2010 to 7–17% in 2025 [65]. Much of the strong support for the use of biomass feedstock to substitute oil derivatives is premised on the widespread assumption that they are carbon neutral, promote rural development, and provide an opportunity for countries to decrease dependence on imported oil. However, biomass production also takes up land and may compete with food production, and there is no consensus among scientists on how to evaluate biomass sustainability [63]. Even though the current ISO standards [3,42] provide a general framework for conducting LCAs, they fail to address a number of critical issues associated with bio-based products (i.e., chemicals, materials, energy) from a life cycle wide perspective. These issues include, for example, the accounting for bio-based carbon storage, impacts of land use changes, and consequential impacts of biomass diversion [66]. 1.4.1.1

Glycerol Case Study

Glycerol presents an interesting case study of the challenges of analyzing the impacts of policies to promote renewables. Biodiesel is generally produced by a transesterification reaction between triglycerides and methanol, and glycerol is a

13

HGC113

03/27/2018

13:3:25

14

Page 14

1 Application of Life Cycle Assessment to Green Chemistry Objectives

Figure 1.2 Glycerol worldwide production per year, compilation based on Ref. [68].

by-product. The booming market for biodiesel fuels driven by government programs to promote renewable fuels for transportation has created a surplus of glycerol (see Figure 1.2). Countries with large areas of available land, such as Argentina and Brazil, and countries with established palm or coconut oil plantations, such as Malaysia, Thailand, the Philippines, Indonesia, and Columbia, expanded production to sell to biodiesel producers in Europe and North America and oleochemical producers in Asia [67]. Thus, supply of glycerol is independent of the demand, and the surplus has created interest in developing new chemical uses of glycerol as a platform chemical. Morales et al. [69] studied the synthesis of lactic acid in a process building on the enzymatic production of dihydroxyacetone from crude glycerol. Their LCA study used ecoinvent datasets [70] and Aspen Plus® V8.2 process models to estimate relevant inventory data. Nonrenewable cumulative energy demand was used as a proxy for environmental impact. The ecoinvent data were based on rapeseed oil grown in Europe. Allocation of impacts assumed that glycerol was a partially utilized coproduct, and is valid as long as the supply of glycerol is not constrained by reductions in biodiesel production. Thus, policies to promote sustainable transport fuels directly impacted assessment of biochemicals produced from glycerol. Cespi et al. [71] studied the production of acrolein using glycerol as a feedstock. Acrolein is an important intermediate in industrial (acrylic acid – AcA) and agricultural (methionine) chemicals. Two synthesis routes producing glycerol as a by-product were modeled – transesterification process for biodiesel and production of fatty acids by triglyceride hydrolysis – and compared using a life cycle perspective with the traditional fossil-based pathway involving the partial oxidation of propylene. In general, the integration of the life cycle approach within the R&D stage helps to better understand production chain criticalities and to optimize the whole manufacturing process. Given the case of the AcA, the application of a simplified cradle-to-gate assessment using contribution and network analyses

HGC113

03/27/2018

13:3:25

Page 15

1.4 Promote Renewable Materials and Energy

Figure 1.3 Reduction of impacts with substitution of 20% WCO.

underlines the stages with higher environmental concerns and can help enterprises to implement their monitoring plan to find more affordable solutions. Assuming a basic scenario in which glycerol is obtained from dedicated biomass only (e.g., rapeseed), the execution of a network analysis depicted the process responsible for the greater environmental burdens expressed in terms of the main indicators, such as: climate change (GWP), cumulative energy demand, water depletion (WD), human toxicity (HT) and single score (SS). In this case, the crop production phase is the major contributor for each impact category considered (around 50–86%). Therefore, improvement efforts would be best focused on alternatives for the raw material supply. One option could be the substitution of dedicated plant feedstock with recovery of a waste product, such as waste cooking oil (WCO). As depicted in Figure 1.3, a replacement of only 20% of the virgin oil with the WCO could yield potential reduction of the impact 2) categories considered on the order of 8–19%. In addition to the environmental concerns, a broader interpretation of greenness to encompass sustainability objectives would expand the analysis to include some social evaluation such as the potential impacts on humanity. HT indicator should be always taken into consideration in product/process assessments. The SS indicator incorporates weighting of the various impact categories. Different stakeholders would have different values and priorities, and the weighting can be adjusted to accommodate differing perspectives. Substitution of WCO could create a revenue stream for restaurants and other institutional kitchens that generate the waste. The substitution would also impact the revenue of farms and plantations currently providing the plant oils. Further, the actual GHG 2) The authors acknowledge Professors Dr. Fabrizio Passarini and Dr. Fabrizio Cavani from the University of Bologna for the use of the software license and the sensible data to run the analysis.

15

HGC113

03/27/2018

13:3:26

16

Page 16

1 Application of Life Cycle Assessment to Green Chemistry Objectives

reductions achieved depend on fossil fuel market responses to substitution of biodiesel. Thus, the quantitative LCA results are necessarily supplemented with qualitative assessments of potential market responses, social impacts, and other political and technology trends to define scenarios that can test the robustness of proposed alternatives [37,38]. It is also particularly important for broad public policy initiatives to promote renewables to take a consequential perspective in LCA studies. Another area of special consideration is geographic specificity. Many land use, water pollution, and toxicity issues are localized. On the other hand, when commodity production is under investigation (e.g., butadiene) it is appropriate to focus on global concerns, for example, GWP, WD, and CED [72]. Thus, studies to evaluate the ultimate benefits of renewable feedstocks present numerous methodological challenges. 1.4.2 Biochemicals Production 1.4.2.1

Life Cycle Stages of Biochemical Production

The environmental impacts of bio-based chemicals and materials have been quantified using LCA in numerous studies (see, e.g., [73–79). Figure 1.4 shows a simplified and generic system boundary for a biorefinery system [80,81] producing chemicals, fuel, and energy from bio-based feedstocks (biomass or organic waste feedstock). A typical life cycle starts with carbon fixation from the atmosphere via photosynthesis in the biomass crop. Renewable feedstock can be obtained from various sectors, including agriculture, forestry, aquaculture, industries (process residues, construction, and demolition debris), and households (municipal wastes and wastewater). Land requirements vary with feedstock type [82]. A conceptual biorefinery is capable of supplying a wide spectrum of marketable products, including chemicals, fuels, and bioenergy [83]. In addition to biochemical conversion routes (i.e., fermentation or anaerobic digestion), thermochemical platforms apply gasification or pyrolysis as a way of transforming bio-based feedstock into fuels, energy, and chemical products [84]. Biopolymers (e.g., polyethylene or polylactic acid) may be used in a cascading manner (multiple life cycles), thereby delaying emissions of carbon stored in the polymer product to the environment [76,85]. However, most of today’s biochemicals are produced in single production chains and not yet within a biorefinery setting [86]. 1.4.2.2

Environmental Implications of Biomass Production

Several scientific studies have shown the potential of bio-based fuels, energy, chemicals, and materials to reduce both nonrenewable energy consumption and carbon dioxide emissions in comparison to their fossil-based counterparts [79]. However, biomass production and processing is also associated with adverse environmental impacts. For example, agricultural biomass production can have

HGC113

03/27/2018

13:3:26

Page 17

1.4 Promote Renewable Materials and Energy

Figure 1.4 Schematic production chain of biochemical production in a biorefinery concept (biochemical platform).

negative environmental effects such as soil erosion, eutrophication of ground and surface waters, and destruction of ecosystems resulting in diminished biodiversity [87]. Cultivation, harvesting, and subsequent processing of biomass feedstock consumes fossil energy and requires the energy intensive production and use of artificial fertilizers and hazardous chemicals. Environmental impacts from eutrophication and acidification as well as N2O emissions, which is a particularly strong greenhouse gas, are often times not properly accounted for in LCA studies on bioproducts [87]. Furthermore, the use of phosphorus as plant fertilizer is gaining increasing attention as a “critical” (and nonsubstitutable) raw material upon which crop growth depends [88]. LCI data to support impacts to eutrophication or acidification during biomass production are often scarce. Biodiversity and other ecosystem impacts are highly

17

HGC113

03/27/2018

13:3:26

18

Page 18

1 Application of Life Cycle Assessment to Green Chemistry Objectives

site specific and difficult to quantify, given current LCIA methods [89,90]. Assessment methods for water use, soil degradation, and biodiversity are immature and need further development [66]. So far, none of the LCIA methods consider aspects of criticality [91] associated with using phosphorus and minerals for fertilization. 1.4.2.3

Carbon Accounting and Land Use Change

In recent years, the carbon neutrality presumption of biomass feedstock in LCA has been challenged as indirect emissions of land use change [92,93], the dynamics of carbon flow over time [94,95], and temporary carbon storage in products [96] are receiving increased attention (see Figure 1.5). Until recently, many LCA studies presumed that biomass is inherently carbon neutral because it is part of the natural carbon cycle [97]. However, in a seminal paper, researchers challenged the greenhouse gas balance of bioethanol production in the United States, and suggested indirect links between diverting cropland for biofuels production and conversion of forest and grassland to new cropland to replace the grain diverted to biofuels [92]. Greenhouse gas emissions can occur as land is converted from one use to another (e.g., forest land to cropland) because of differences in the amount of carbon stored in the plants and potential losses of soil carbon (termed: “direct land use change”) [98]. However, “indirect land use changes” occur outside the system boundary (Figure 1.5) and are due to the displacement of services (usually food production) that were previously provided by the land now used for growing crops for the production of bioproducts [92,93,98]. LCA researchers attempt to capture such indirect impacts using consequential LCA [53] but there is no general consensus yet on how to do such assessments, and coupling LCA models with econometric

Figure 1.5 Schematic figure showing the three areas of debate in calculating the greenhouse gas emissions of bio-based products: (1) direct and indirect emissions of land use change, (2) regrowth time of biomass, and (3) temporary carbon storage.

HGC113

03/27/2018

13:3:26

Page 19

1.4 Promote Renewable Materials and Energy

models (accounting for changes in supply and demand of feedstock diversion) is challenging. Greenhouse gas emissions from indirect land use change will depend on the type of lands converted and what other product system (e.g., food production) compete with the growth of biomass feedstocks. Once land use changes have been identified and inventoried, there remains a difficult task to quantify the impacts. Land use interventions are characterized as occupational or transformational, and land quality is assessed for impacts on the intrinsic value of biodiversity, on the biotic production potential, and on ecological soil quality [99]. Although assessment methods are under development, UNEP–SETAC has issued guidance for global land use impacts on biodiversity and ecosystem services [100]. Temporary carbon storage can take place if carbon sequestered via photosynthesis from the atmosphere is stored in a bio-based product (e.g., polymer with long life time) [96]. Carbon storage can also occur at end-of-life (EoL), for example, when a bio-based product is landfilled [101,102]. Some studies have suggested that temporary carbon storage delays radiative forcing from greenhouse gases in the atmosphere and that delay provides time for technological progress and research [96] and should therefore, be accounted for. Other studies argue that biogenic carbon storage should not be considered because it is usually reversible, and therefore, inevitably adds carbon back into the atmosphere in the future. Another consideration is that delayed emissions could occur in an atmosphere with a higher CO2 concentration, producing even greater impacts [103]. The benefits of biogenic carbon storage depend on the time horizon over which the global warming potential of greenhouse gas emissions is considered [66]. A time window of 100 years (beyond which the impacts of carbon storage are not considered) is often used, although it should be noted that the choice is intrinsically subjective [96]. The majority of current LCA models ignore the time required for the harvested biomass to regrow and sequester the biogenic carbon released during the life cycle of the bio-based product [98]. While this assumption is a reasonable approximation for short rotation crops (requiring about one year for regrowth), it neglects the fact that many feedstocks (e.g., forest biomass) will need more time to regrow and sequester an equal amount of carbon as was released during the bioproduct’s life cycle. Ideally, this time component would be included by coupling LCA models with forest carbon models [94,95], but in reality this is rarely done. Multiple approaches accounting for biogenic carbon in LCAs of bio-based products exist. Simply ignoring biogenic carbon emissions has a high potential of burden shifting as many of the impacts may be located outside the general LCA system boundary (e.g., indirect land use change due to feedstock diversion may take place in other geographical regions; sequestration of biogenic carbon emitted during the production of bio-based products in new biomass feedstock may require multiple years and be outside the temporal scope of the LCA).

19

HGC113

03/27/2018

13:3:27

20

Page 20

1 Application of Life Cycle Assessment to Green Chemistry Objectives

1.4.2.4

Global Availability of Arable Land

In view of current efforts to increase commercial biofuels and biochemicals production, the availability of global arable land for nonfood purposes requires special attention. Reference [104] showed that global cropland area, which encompasses arable land and permanent crops, is in fact a scarce resource. Their study estimated that global demand for cropland area will increase mainly due to global population growth and changes in nutrition. As a result, cropland availability per capita will decrease from 2500 m2 per person in 2004 to only about 2000 m2 per person in 2030. This does not yet consider the increasing land required for providing future demands for bio-based fuels, chemical, and energy. Furthermore, climate change may lead to a higher frequency of extreme weather patterns. Considering land occupation of bio-based products together with other impact categories is important because land is a scarce resource. The pressure on global arable land can be reduced, for example, by considering the use of waste and production residues instead of virgin biomass as starting materials for green chemistry routes and by producing materials first that can be used in subsequent product life cycles (e.g., polymers produced into plastics that then serve as feedstock for subsequent chemicals or energy production) [82,87].

1.5 Conclusion and Recommendations

LCA can help to incorporate a more holistic cradle-to-grave and system- wide perspective into GC applications. It can assist in measuring the overall greenness of the 12 principles of green chemistry applied to modern product system and elucidate potential trade-offs (e.g., shifting of environmental burdens from one life cycle stage to another or from one environmental threat to another). However, LCA is not a substitute for chemical risk assessment or more detailed process system engineering studies. The potential value of LCA is in using it complementary to other analytical methods to obtain a more complete picture of the product system and to better target the detailed supporting studies. There are some inherent limitations in applying LCA to study the multitude of chemical substances used by the modern chemical industry. Missing LCI data are often mentioned as a barrier, but researchers have identified promising approaches to address this need. The use of proxy data with default values based on industry or company-specific data can provide reasonable estimates. Companies can also conduct detailed studies of existing products to develop proxies based on classes of systems with similar impacts. This suggests that there needs to be as much attention on the architecture of data sets as on the specific LCI data elements. The first LCA study will obviously be challenging, and companies should start with simple qualitative assessments. Retrospective studies of deployed products can help develop proxy data sets as already discussed. Improved collaboration across functional groups is important to minimize the administrative burden of gathering data and better coordinate data transfer between R&D, process

HGC113

03/27/2018

13:3:27

Page 21

References

engineering, product design, and environmental management. Because LCA examines inputs (resources) and outputs (emissions, wastes, and desired products) across the entire product life cycle, from resource extraction to ultimate disposal, its models are usually not site-specific that poses challenges when quantifying toxicity-related impacts and other regional environmental implications. LCA, however, can provide a useful screen for product systems with a global supply chain to identify the 10–30 chemicals that warrant more detailed risk assessment. Another challenge for LCA is that intermediate flows are balanced out to identify only the elemental flows directly to and from nature, that is, resources used and pollutants emitted. Making the LCI model “toxic aware” could provide useful information to better associate risks with specific processes. The use of risk or R- phrases to integrate toxicity information into LCI models is an interesting approach worth further development. This could provide a valuable interconnection to environmental management programs by helping to identify necessary risk management measures and hazard labeling requirements. Perhaps the most significant advantage of LCA is providing a structured framework to better integrate GC objectives into product and process design procedures. Simple green metrics or abridged LCA screening methods that are tailored to specific sectors or company objectives can be developed to promote their use in decision-making processes. These simplified approaches can be calibrated with retrospective studies of fielded products or implemented processes. Data collected to manage the operations can then be used to build proxy data sets to speed subsequent LCI efforts. This enables LCA to be used during the early design phase of new green chemistry routes in a streamlined fashion to obtain a first impression of the potential environmental implications of using different process designs and raw materials. Applied strategically, LCA can promote more effective collaboration (and information sharing) across functional groups and supply chain partners and guide organizational learning to continually improve the life cycle performance of the value chain. Recent efforts of LCA have focused on capturing indirect effects in product supply chains and across the economy through the use of economic models and techniques. The broader consideration of indirect environmental burdens is particularly important for evaluating policies to promote a transition to bio-based raw materials in chemical synthesis. This is an area in need of additional research to develop a better understanding of alternative approaches to C-LCA modeling, improved LCIA methods for land and water use, and more guidance on modeling temporary carbon storage for bio-based chemicals.

References 1 Anastas, P.T. and Warner, J.C. (1998)

Green Chemistry: Theory and Practice, Oxford University Press, New York.

2 Muñoz, I. (2012) LCA in Green

Chemistry: a new subject area and call for papers. International Journal of Life Cycle Assessment, 17, 517–519.

21

HGC113

03/27/2018

13:3:27

22

Page 22

1 Application of Life Cycle Assessment to Green Chemistry Objectives 3 ISO14040 (2006) Environmental

4

5

6

7

8

9

10

11

12

Management – Life Cycle Assessment – Principles and Framework, Geneva, Switzerland. Finnveden, G., Hauschild, M.Z., Ekvall, T., Guinée, J.B., Heijungs, R., Hellweg, S., Koehler, A., Pennington, D., and Suh, S. (2009) Recent developments in life cycle assessment. Journal of Environmental Management, 91, 1–21. Hellweg, S. and Milà i Canals, L. (2014) Emerging approaches, challenges and opportunities in life cycle assessment. Science, 344, 1109–1113. Reap, J., Roman, F., Duncan, S., and Bras, B. (2008a) A survey of unresolved problems in life cycle assessment: Part 1goal and scope and inventory analysis. International Journal of Life Cycle Assessment, 13, 290–300. Reap, J., Roman, F., Duncan, S., and Bras, B. (2008b) A survey of unresolved problems in life cycle assessment: Part 2impact assessment and interpretation. International Journal of Life Cycle Assessment, 13, 374–388. Olsen, S.I., Christensen, F.M., Hauschild, M.Z., Pedersen, F., Larsen, H.F., and Tørsløv, J. (2001) Life cycle impact assessment and risk assessment of chemicals - a methodological comparison. Environmental Impact Assessment Review, 21, 385–404. Tufvesson, L.M., Tufvesson, P., Woodley, J.M., and Börjesson, P. (2013) Life cycle assessment in green chemistry: overview of key parameters and methodological concerns. International Journal of Life Cycle Assessment, 18, 431–444. Wernet, G., Hellweg, S., and Hungerbühler, K. (2012) A tiered approach to estimating inventory data and impacts of chemical products and mixtures. International Journal of Life Cycle Assessment, 17, 720–728. Anastas, P.T. and Lankey, R.L. (2000) Life cycle assessment and green chemistry: the yin and yang of industrial ecology. Green Chemistry, 2, 289–295. Kralisch, D., Ott, D., and Gericke, D. (2015) Rules and benefits of life cycle assessment in green chemical process

13

14

15

16

17

18

19

20

21

22

and synthesis design: a tutorial review. Green Chemistry, 17, 123–145. Kuczenski, B., Geyer, R., and Boughton, B. (2011) Tracking toxicants: toward a life cycle aware risk assessment. Environmental Science and Technology, 45, 45–50. Sjöström, J. (2006) Green chemistry in perspective – models for GC activities and GC policy and knowledge areas. Green Chemistry, 8, 130–137. ICA (2013) Alternatives Assessment Guide, Interstate Chemicals Clearinghouse, pp. 166. Lavoie, E.T., Heine, L.G., Holder, H., Rossi, M.S., Lee, R.E., II, Connor, E.A., Vrabel, M.A., DiFiore, D.M., and Davies, C.L. (2010) Chemical alternatives assessment: enabling substitution to safer chemicals. Environmental Science and Technology, 44, 9244–9249. Rossi, M., Tickner, J., and Geiser, K. (2006) Alternatives Assessment Framework of the Lowell Center for Sustainable Production, pp. 21. Wernet, G., Conradt, S., Isenring, H.P., and Jiménez-González, C. (2010) Life cycle assessment of fine chemical production: a case study of pharmaceutical synthesis. International Journal of Life Cycle Assessment, 15, 294–303. Hellweg, S., Fischer, U., Scheringer, M., and Hungerbühler, K. (2004) Environmental assessment of chemicals: methods and application to a case study of organic solvents. Green Chemistry, 6, 418–427. Geisler, G., Hofstetter, T.B., and Hungerbühler, K. (2004) Production of fine and specialty chemicals: procedure for the estimation of LCIs. International Journal of Life Cycle Assessment, 9, 101–113. Jiménez-González, C., Curzons, A.D., Constable, D.J.C., and Cunningham, V.L. (2004) Cradle- to- gate life cycle inventory and assessment of pharmaceutical compounds. International Journal of Life Cycle Assessment, 9, 114–121. Wernet, G., Papadokonstantakis, S., Hellweg, S., and Hungerbühler, K. (2009)

HGC113

03/27/2018

13:3:27

Page 23

References

23

24

25

26

27

28

29

30

Bridging data gaps in environmental assessments: modleing impacts of fine and basic chemical production. Green Chemistry, 11, 1826–1831. Goedkoop, P., and Spriensma, R. (2001) The Eco-Indicator 99: A Damage Oriented Method for Life Cycle Impact Assessment, Methodology Report, PRé Consultants, Amersfoort, NL. Hauschild, M.Z., Huijbregts, M.A.J., Jolliet, O., MacLeod, M., Margni, M., van de Meent, D., Rosenbaum, R.K., and McKone, T.E. (2008) Building a model based on scientific consensus for life cycle impact assessment of chemicals: the search for harmony and parsimony. Environmental Science & Technology, 42, 7032–7037. Rosenbaum, R.K., Bachmann, T.M., Swirsky Gold, L., Huijbregts, M.A.J., Jolliet, O., Juraske, R., Koehler, A., Larsen, H.F., MacLeod, M., Margni, M. et al. (2008) USEtox - the UNEP-SETAC toxicity model: recommended characterization factors for human toxicity and freshwater ecotoxicity in life cycle impact assessment. International Journal of Life Cycle Assessment, 13, 532–546. Askham, C. (2012) REACH and LCA – methodological approaches and challenges. International Journal of Life Cycle Assessment, 17, 43–57. Roschangar, F., Sheldon, R.A., and Senanayake, C.H. (2015) Overcoming barriers to green chemistry in the pharmaceutical industry – the Green Aspiration LevelTM concept. Green Chemistry, 17, 752–768. Askham, C., Gade, A.L., and Hanssen, O.J. (2013) Linking chemical risk information with life cycle assessment in product development. Journal of Cleaner Production, 51, 196–204. Igos, E., Moeller, R., Benetto, E., Biwer, A., Guiton, M., and Dieumegard, P. (2014) Development of USEtox characterisation factors for dishwasher detergents using data made available under REACH. Chemosphere, 100, 160–166. Amelio, A., Genduso, G., Vreysen, S., Luis, P., and Van der Bruggen, B. (2014)

31

32

33

34

35 36

37

38

39

40

Guidelines based on life cycle assessment for solvent selection during the process design and evaluation of treatment alternatives. Green Chemistry, 16, 3045–3063. Capello, C., Fischer, U., and Hungerbühler, K. (2007) What is a green solvent? A comprehensive framework for the environmental assessment of solvents. Green Chemistry, 9, 927–934. Anastas, P.T. and Zimmerman, J.B. (2003) Peer reviewed: design through the 12 principles of green engineering. Environmental Science and Technology, 37, 94A–101A. Jiménez-González, C. and Constable, D.J.C. (2011) Green Chemistry and Engineering: A Practical Design Approach, Wiley-VCH Verlag GmbH, Weinheim, Germany. Gilbertson, L.M., Zimmerman, J.B., Plata, D.L., Hutchison, J.E., and Anastas, P.T. (2015) Designing nanomaterials to maximize performance and minimize undesirable implications guided by the principles of green chemistry, Chemical Society Reviews, 44, pp. 5758–5777. Tenner, E. (1996) Why Things Bite Back, Vintage Books, New York. Burgess, A.A. and Brennan, D.J. (2001) Application of life cycle assessment to chemical processes. Chemical Engineering Science, 56, 2589–2604. Jiménez-González, C. and Overcash, M.R. (2014) The evolution of life cycle assessment in pharmaceutical and chemical applications – a perspective. Green Chemistry, 16, 3392–3400. Jacquemin, L., Pontallier, P.-Y., and Sablayrolles, C. (2012) Life cycle assessment (LCA) applied to the process industry: a review. International Journal of Life Cycle Assessment, 17, 1028–1041. Zheng, K., Lou, H.H., Gangadharan, P., and Kanchi, K. (2012) Incorporating sustainability into the conceptual design of chemical process-reaction routes selection. Industrial & Engineering Chemistry Research, 51, 9300–9309. Eerhart, A.J.J.E., Faaij, A.P.C., and Patel, M.K. (2012) Replacing fossil based PET with biobased PEF; process analysis,

23

HGC113

03/27/2018

13:3:27

24

Page 24

1 Application of Life Cycle Assessment to Green Chemistry Objectives

41

42

43

44

45

46

47

48

49

energy and GHG balance. Energy & Environmental Science, 5, 6407–6422. Fazeni, K., Lindorfer, J., and Prammer, H. (2014) Methodology advancements in life cycle process design: a preliminary outlook. Resources, Conservation and Recycling, 92, 66–77. ISO 14044 (2006) Environmental Management – Life Cycle Assessment – Requirements and Guidelines, Geneva, Switzerland. Saling, P., Kicherer, A., Dittrich-Krämer, B., Wittlinger, R., Zombik, W., Schmidt, I., Schrott, W., and Schmidt, S. (2002) Eco-efficiency analysis by BASF: the method. The International Journal of Life Cycle Assessment, 7, 203–218. Schmidt, I., Meurer, M., Saling, P., Kicherer, A., Reuter, W., and Gensch, C.O. (2004) SEEbalance® : managing sustainability of products and processes with the socio- eco- efficiency analysis by BASF. Greener Management International, 45, 79–94. UNEP/SETAC (2009) Guidelines for Social Life Cycle Assessment of Products (eds C. Benoit and B. Mazjin), United Nations Environment Programme, Paris. Leseurre, L., Merea, C., Duprat de Paule, S., and Pichart, A. (2014) Eco- footprint: a new tool for “Made in Chimex” considered approach. Green Chemistry, 16, 1139–1148. Curzons, A., Jiménez-González, C., Duncan, A., Constable, D.C., and Cunningham, V. (2007) Fast life cycle assessment of synthetic chemistry (FLASCTM) tool. The International Journal of Life Cycle Assessment, 12, 272–280. Adams, J.P., Alder, C.M., Andrews, I., Bullion, A.M., Campbell-Crawford, M., Darcy, M.G., Hayler, J.D., Henderson, R.K., Oare, C.A., Pendrak, I. et al. (2013) Development of GSK’s reagent guides – embedding sustainability into reagent selection. Green Chemistry, 15, 1542–1549. Constable, D.J.C., Curzons, A.D., and Cunningham, V.L. (2002) Metrics to “green” chemistry – which are the best? Green Chemistry, 4, 521–527.

50 Jimenez-Gonzalez, C., Ponder, C.S.,

51

52

53

54

55

56

57

58

59

Broxterman, Q.B., and Manley, J.B. (2011) Using the right green yardstick: why process mass intensity is used in the pharmaceutical industry to drive more sustainable processes. Organic Process Research & Development, 15, 912–917. van Keleren, H.A., Blom, A.L., Rutjes, F.P.J.T., and Huijbregts, M.A.J. (2013) On the usefulness of life cycle assessment in early chemical methodology development: the case of organophosphorous-catalyzed Appel and Wittig reactions. Green Chemistry, 15, 1255–1263. Harding, K.G. (2013) A technique for reporting life cycle impact assessment (LCIA) results. Ecological Indicators, 34, 1–6. Earles, J.M. and Halog, A. (2011) Consequential life cycle assessment: a review. The International Journal of Life Cycle Assessment, 16, 445–453. Curran, M.A., Mann, M., and Norris, G. (2005) The international workshop on electricity data for life cycle inventories. Journal of Cleaner Production, 13, 853–862. Weidema, B. (1993) Market aspects in product life cycle inventory methodology. Journal of Cleaner Production, 1, 161–166. Marvuglia, A., Benetto, E., Rege, S., and Jury, C. (2013) Modelling approaches for consequential life-cycle assessment (CLCA) of bioenergy: critical review and proposed framework for biogas production. Renewable and Sustainable Energy Reviews, 25, 768–781. Plevin, R.J., Delucchi, M.A., and Creutzig, F. (2014) Using attributional life cycle assessment to estimate climate-change mitigation benefits misleads policy makers. Journal of Industrial Ecology, 18, 73–83. Weidema, B., Frees, N., and Nielsen, A.M. (1999) Marginal production technologies for life cycle inventories. International Journal of Life Cycle Assessment, 4, 48–56. Ekvall, T. and Andræ, A.S.G. (2006) Attributional and consequential environmental assessment of the shift to

HGC113

03/27/2018

13:3:27

Page 25

References

60

61

62 63

64

65

66

67

68

69

lead-free solders. International Journal of Life Cycle Assessment, 11, 344–353. Langeveld, H., Sanders, J., and Meeusen, M. (2010) The Biobased Economy: Biofuels, Materials and Chemicals in the Post-oil Era, Earthscan, London. European Commission (2012) Innovating for Sustainable Growth: A Bioeconomy for Europe, Brussels. The White House (2012) National Bioeconomy Blueprint, Washington. Bosch, R., van de Pol, M., and Philp, J. (2015) Policy: define biomass sustainability. Nature, 523, 526–527. Dale, B.E., Anderson, J.E., Brown, R.C., Csonka, S., Dale, V.H., Herwick, G., Jackson, R.D., Jordan, N., Kaffka, S., Kline, K.L. et al. (2014) Take a closer look: biofuels can support environmental, economic and social goals. Environmental Science and Technology, 48, 7200–7203. Kircher, M. (2012) The transition to a bio-economy: emerging from the oil age. Biofuels, Bioproducts & Biorefining, 6, 369–375. Pawelzik, P., Carus, M., Hotchkiss, J., Narayan, R., Selke, S., Wellisch, M., Weiss, M., Wicke, B., and Patel, M.K. (2013) Critical aspects on the life cycle assessment (LCA) of bio-based- materials – reviewing methodologies and deriving recommendations. Resources, Conservation and Recycling, 73, 211–228. Ciriminna, R., Pina, C.D., Rossi, M., and Pagliaro, M. (2014) Understanding the glycerol market. European Journal of Lipid Science and Technology, 116, 1432–1439. Ayoub, M. and Abdullah, A.Z. (2012) Critical review on the current scenario and significance of crude glycerol resulting from biodiesel industry towards more sustainable renewable energy industry. Renewable and Sustainable Energy Reviews, 16, 2671–2686. Morales, M., Dapsens, P.Y., Giovinazzo, I., Witte, J., Mondelli, C., Papadokonstantakis, S., Hungerbuhler, K., and Perez-Ramirez, J. (2015) Environmental and economic assessment of lactic acid production from glycerol using cascade bio- and chemocatalysis.

70

71

72

73

74

75

76

77

78

Energy and Environmental Science, 8, 558–567. ecoinvent (2010) Life Cycle inventory database v2.2 (Dübendorf, CH: Swiss Centre for Life Cycle Inventories). Cespi, D., Passarini, F., Mastragostino, G., Vassura, I., Larocca, S., Iaconi, A., Chieregato, A., Dubois, J.L., and Cavani, F. (2015a) Glycerol as feedstock in the synthesis of chemicals: a life cycle analysis for acrolein production. Green Chemistry, 17, 343–355. Cespi, D., Passarini, F., Vassura, I., and Cavani, F. (2015b) Butadiene from biomass, a life cycle perspective to address sustainability in the chemical industry. Green Chemistry, 18, 1625–1638. Chen, G.-Q. and Patel, M.K. (2012) Plastics derived from biological sources: present and future: a technical and environmental review. Chemical Reviews, 112, 2082–2099. Cherubini, F. and Jungmeier, G. (2009) LCA of a biorefinery concept producing bioethanol, bioenergy, and chemicals from switchgrass. The International Journal of Life Cycle Assessment, 15, 53–66. Moncada, J., Posada, J.A., and Ramírez, A. (2015) Early sustainability assessment for potential configurations of integrated biorefineries. Screening of bio-based derivatives from platform chemicals. Biofuels, Bioproducts & Biorefining, 9, 722–748. Nuss, P. and Gardner, K.H. (2013) Attributional life cycle assessment (ALCA) of polyitaconic acid production from northeast US softwood biomass. The International Journal of Life Cycle Assessment, 18, 603–612. Posada, J.A., Patel, A.D., Roes, A., Blok, K., Faaij, A.P.C., and Patel, M.K. (2013) Potential of bioethanol as a chemical building block for biorefineries: preliminary sustainability assessment of 12 bioethanol-based products. Bioresource Technology, 135, 490–499. Vink, E.T.H., Davies, S., and Kolstad, J. (2010) The eco-profile for current Ingeo® polylactide production. Industrial Biotechnology, 6, 212–224.

25

HGC113

03/27/2018

13:3:27

26

Page 26

1 Application of Life Cycle Assessment to Green Chemistry Objectives 79 Weiss, M., Patel, M., Heilmeier, H., and

80

81

82

83

84

85

86

87

88

Bringezu, S. (2007) Applying distance-totarget weighing methodology to evaluate the environmental performance of biobased energy, fuels, and materials. Resources, Conservation and Recycling, 50, 260–281. Kamm, B., Gruber, P.R., and Kamm, M. (2006) Biorefineries: Industrial Processes and Products – Status Quo and Future Directions, 1, Wiley-VCH Verlag GmbH, Weinheim. Kamm, B. and Kamm, M. (2004) Principles of biorefineries. Applied Microbiology and Biotechnology, 64, 137–145. Bringezu, S., Schütz, H., Pengue, W., O’Brien, M., Garcia, F., Sims, R., Howarth, R., Kauppi, L., Swilling, M., and Herrick, J. (2014) Assessing Global Land Use: Balancing Consumption with Sustainable Supply. In a report of the Working Group on Land and Soils of the International Resource Panel. UNEP International Resource Panel, Paris. IEA (2015) IEA Bioenergy Task 42 Biorefinery. In International Energy Agency (IEA) Bioenergy. Hayes, D.J. (2009) An examination of biorefining processes, catalysts and challenges. Catalysis Today, 145, 138–151. Nuss, P., Bringezu, S., and Gardner, K.H. (2012) Waste-to-materials: the longterm option, in Waste to Energy: Opportunities and Challenges for Developing and Transition Economies (ed. A. Karagiannidis), Springer, London, pp. 1–26. Cherubini, F. (2010) The biorefinery concept: using biomass instead of oil for producing energy and chemicals. Energy Conversion and Management, 51, 1412–1421. Bringezu, S., Schütz, H., Ó Brien, M., Kauppi, L., Howarth, R.W., and McNeely, J. (2009b) Towards Sustainable Production and Use of Resources: Assessing Biofuels. In UNEP Resource Panel, UNEP, Paris, pp. 120. Cordell, D., Drangert, J.-O., and White, S. (2009) The story of phosphorus: global food security and food for thought.

89

90

91

92

93

94

95

96

Global Environmental Change, 19, 292–305. Souza, D.M., Teixeira, R.F.M., and Ostermann, O.P. (2015) Assessing biodiversity loss due to land use with life cycle assessment: are we there yet? Global Change Biology, 21, 32–47. Teixeira, R.F.M., Maia de Souza, D., Curran, M.P., Antón, A., Michelsen, O., and Milà i Canals, L. (2015) Towards consensus on land use impacts on biodiversity in LCA: UNEP/SETAC life cycle Initiative preliminary recommendations based on expert contributions. Journal of Cleaner Production, 112, 4283–4287. Graedel, T.E., Harper, E.M., Nassar, N.T., Nuss, P., and Reck, B.K. (2015) Criticality of metals and metalloids. Proceedings of the National Academy of Sciences of the United States of America, 112, 4257–4262. Searchinger, T., Heimlich, R., Houghton, R.A., Dong, F., Elobeid, A., Fabiosa, J., Tokgoz, S., Hayes, D., and Yu, T.-H. (2008) Use of U.S. croplands for biofuels increases greenhouse gases through emissions from land-use change. Science, 319, 1238–1240. Searchinger, T.D., Hamburg, S.P., Melillo, J., Chameides, W., Havlik, P., Kammen, D.M., Likens, G.E., Lubowski, R.N., Obersteiner, M., Oppenheimer, M. et al. (2009) Fixing a critical climate accounting error. Science, 326, 527–528. McKechnie, J., Colombo, S., Chen, J., Mabee, W., and MacLean, H.L. (2011) Forest bioenergy or forest carbon? Assessing trade-offs in greenhouse gas mitigation with wood-based fuels. Environmental Science & Technology, 45, 789–795. Walker, T. (2010) Massachusetts Biomass Sustainability and Carbon Policy Study. Report to the Commonwealth of Massachusetts Department of Energy Resources, in Natural Capital Initiative Report, Manomet Center for Conservation Science, Brunswick, Maine. . Levasseur, A., Brandão, M., Lesage, P., Margni, M., Pennington, D., Clift, R., and Samson, R. (2012) Valuing temporary

HGC113

03/27/2018

13:3:27

Page 27

References

97

98

99

100

101

carbon storage. Nature Climate Change, 2, 6–8. Johnson, E. (2009) Goodbye to carbon neutral: getting biomass footprints right. Environmental Impact Assessment Review, 29, 165–168. Amazega, J.M., von Maltiz, G., and Boyes, S. (2010) Assessing the Sustainability of Bioenergy Projects in Developing Countries: A framework for Policy Evaluation, Newcastle University, England. Milà i Canals, L., Bauer, C., Depestele, J., Dubreuil, A., Freiermuth Knuchel, R., Gaillard, G., Michelsen, O., MüllerWenk, R., and Rydgren, B. (2007) Key elements in a framework for land use impact assessment within LCA (11 pp) The International Journal of Life Cycle Assessment, 12, 5–15. Koellner, T., de Baan, L., Beck, T., Brandão, M., Civit, B., Margni, M., i Canals, L., Saad, R., de Souza, D., and Müller-Wenk, R. (2013) UNEP-SETAC guideline on global land use impact assessment on biodiversity and ecosystem services in LCA. The International Journal of Life Cycle Assessment, 18, 1188–1202. Brandão, M., Levasseur, A., Kirschbaum, M.U.F., Weidema, B.P., Cowie, A.L., Jørgensen, S.V., Hauschild, M.Z.,

Pennington, D.W., and Chomkhamsri, K. (2013) Key issues and options in accounting for carbon sequestration and temporary storage in life cycle assessment and carbon footprinting. The International Journal of Life Cycle Assessment, 18, 230–240. 102 Christensen, T.H., Gentil, E., Boldrin, A., Larsen, A.W., Weidema, B.P., and Hauschild, M. (2009) C balance, carbon dioxide emissions and global warming potentials in LCA-modelling of waste management systems. Waste Management & Research, 27, 707–715. 103 Cherubini, F., Bird, N.D., Cowie, A., Jungmeier, G., Schlamadinger, B., and Woess-Gallasch, S. (2009) Energy- and greenhouse gas-based LCA of biofuel and bioenergy systems: key issues, ranges and recommendations. Resources, Conservation and Recycling, 53, 434–447. 104 Bringezu, S., Schütz, H., Arnold, K., Merten, F., Kabasci, S., Borelbach, P., Michels, C., Reinhardt, G.A., and Rettenmaier, N. (2009a) Global implications of biomass and biofuel use in Germany – recent trends and future scenarios for domestic and foreign agricultural land use and resulting GHG emissions. Journal of Cleaner Production, 17, S57–S68.

27

HGC113

03/27/2018

13:3:27

Page 28

29

2 Shortcut Models Based on Molecular Structure for Life Cycle Impact Assessment: The Case of the FineChem Tool and Beyond Stavros Papadokonstantakis, Pantelis Baxevanidis, Effie Marcoulaki, Sara Badr, and Antonis Kokossis

2.1 Introduction

Among various frameworks for sustainability assessment of chemical processes and products, life cycle assessment (LCA) is a well-known and widely accepted, standardized, and systems-oriented approach [1,2]. In fact, LCA provides useful insights by highlighting the relative importance of diverse environmental impacts from a “cradle-to-grave” perspective and allocating the crucial system interdependencies [3]. Thus, it is not surprising that LCA has been widely used in the assessment of the environmental footprint of chemicals, including, for example, applications in pharmaceuticals and fine chemicals production [4,5], waste management applications [6–9], and bio-based production of chemicals and fuels [10–12]. Moreover, the role of LCA in synthesis and process design [3,13] as well as its affinity to green chemistry principles [14,15] has also been highlighted in scientific literature. LCA comprises the steps of goal and scope definition (i.e., system boundaries, functional unit, and width/depth of the analysis), inventory analysis (i.e., resource requirements, resulting emissions, and assessment of data quality), impact assessment (i.e., on humans, property, and environment), and interpretation (i.e., relative importance of impacts, sensitivity analysis, and optimization potentials). The “heart” of every LCA study is the estimation of life cycle inventories for a given scope and system boundaries. During early stages of process design, model-based estimations are required to fill in inevitable data gaps. This is even more evident in the case of innovative processes and products, bio-based production of chemicals and fuels being a representative example of current academic and industrial research. Recently, Papadokonstantakis et al. [16] highlighted some challenges of model-based estimation of life cycle inventories. The authors particularly focused on issues associated with assumptions and heuristics in early stages of process design that typically do not follow a commonly accepted protocol. These issues include the degree of process integration, the

Handbook of Green Chemistry Volume 10: Tools for Green Chemistry, First Edition. Edited by Evan S. Beach and Soumen Kundu.  2017 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2017 by Wiley-VCH Verlag GmbH & Co. KGaA.

30

2 Shortcut Models Based on Molecular Structure for Life Cycle Impact Assessment

decisions made for the balancing of operating versus capital cost, which strongly depends on the process scale, as well as the allocation procedure followed, especially in multiproduct and multifunctional production systems.. Moreover, the authors identified a number of cases when the cradle-to-gate life cycle inventories of a chemical product need to be estimated:

 The product is registered in commonly used LCA databases (e.g., Ecoinvent1)), where a comprehensive list of life cycle inventories and impact assessment metrics can be found, typically with transparent background information for the considered processes, assumptions, and so on.  The product is not registered in LCA databases, but similar products can be found in terms of the technologies used for their production, and this information can be used to estimate the life cycle inventories of the investigated product depending on the modularity and level of detail of the available data.  No similar products can be found in LCA databases and/or the resolution of the information about the production processes is not high. Shortcut tools can then be used, such as the life cycle inventory estimation methods proposed by GlaxoSmithKline [17,18] or the molecular structure-based models proposed by Wernet et al. [19,20] to estimate various cradle-to-gate life cycle impact assessment (LCIA) metrics (cumulative energy demand [21], eco-indicator 99 [22], etc.).  The production process can be effectively decomposed into subprocesses (e.g., along the synthesis path and/or the value chain of the chemical product) which conform to one of the previous cases or they are amenable to either gate-to-gate short-cut approaches (e.g., proxy indicators [23], multi-input allocation for life cycle inventories of wastewater treatment [24], and waste incineration processes [8]) or detailed process modeling (e.g., if the available information allows the use of process simulation software). From the four cases already mentioned, this chapter will focus on the method and tool of Wernet et al. [19,20], also known as the FineChem tool. The tool requires the least amount of data, namely, only the molecular structure of the chemical compound whose cradle-to-gate life cycle impact assessment needs to be estimated. In the first part of this chapter, the concept and performance of the FineChem tool is briefly presented followed by some illustrative applications where this tool can be of particular use. In the second part of this chapter, we present some current work toward extending the applicability of FineChem through the more widespread functional group contribution approach. The chapter concludes by suggesting future research efforts in this direction.

1) Ecoinvent V.2. Available from http://www.ecoinvent.org/database/ecoinvent-version-2/ecoinventversion-2.html.

2.2 Concept and Development of the FineChem Tool

2.2 Concept and Development of the FineChem Tool

Filling in data gaps in LCA studies of value chains involving production of chemicals is of paramount importance for LCA practitioners because of the complexity and diversity of process synthesis and operation in chemical plants. Clearly, a detailed analysis of all mass and especially energy flows required for the production of chemicals can be a cumbersome task. This is even more evident in early phases of process design, characterized by sparse and not accurate process data. Still, it is exactly this point where considering sustainability objectives, in general, and LCA metrics, in particular, can be most advantageous because of the high degree of freedom in process related decisions. Moreover, energy flows are significantly less documented or measured compared to material flows. The case of the production of “high-value low-volume” fine chemicals and pharmaceuticals, typically produced in batch operations, is a characteristic example of not standardized documentation of energy flows [25]. From this perspective, a method to estimate life cycle impacts, especially energy related ones, in the cases of severe gaps in process-related information would be a significant step toward filling data gaps in LCA studies. Statistical “black box” models using molecular descriptors as input information to estimate a set of LCIA metrics can be an interesting approach to this aim. A simplified workflow diagram in Figure 2.1 presents the intended use of the molecular structure-based models for estimating LCIA metrics. Although the concept of molecular structure-based estimation has been widely used for the prediction of physical and thermodynamic properties (see Section 2.4.1), a similar approach for LCIA metrics poses great challenges. To start with, reliable data cannot be obtained from targeted lab experiments; they should rather be collected from industrial chemical production lines. This restricts the amount of data available for identifying the effect of a specific molecular descriptor into an observed LCIA metric and may also introduce inaccuracies in the data collection procedure as well as confidentiality issues. Another challenge is the type of molecular descriptors to be used in this type of shortcut models, since the theoretical background for interlinking, even qualitatively, certain molecular fragments with process-related information affecting the LCIA metrics is very weak, if existent. Finally, physical and thermodynamic properties are expected to be better correlated with the compound molecular structure, compared to LCIA metrics for the outcomes of many processing steps in a compound’s life cycle. Wernet et al. [19,20] introduced their methodology on how to perform molecular structure-based estimation of LCIA metrics and went on to implement it as the FineChem tool (http://www.sust-chem.ethz.ch/tools/finechem). In their first paper [20], the authors tested the feasibility of the concept by applying multilinear regression (MLR) and artificial neural networks (ANN) to a data set of 103 organic chemicals (i.e., basic chemicals, solvents, pesticides, chelating agents, etc.) coming from Ecoinvent1) and in-house generated data from industrial

31

32

2 Shortcut Models Based on Molecular Structure for Life Cycle Impact Assessment

Figure 2.1 Data availability and models for the estimation of LCIA metrics for the production of chemicals.

projects. They attempted to predict several life cycle impact related categories, such as the cumulative energy demand (CED), the global warming potential (GWP), the biological and chemical oxygen demands (BOD and COD), the total organic carbon (TOC) content of aqueous emissions, as well as the Eco-iIndicator 99 (EI99) related scores, namely, the human health (HH), the ecosystem quality (EQ), the resources (R), and the total (T) scores. They used a set of

2.2 Concept and Development of the FineChem Tool

molecular structure descriptors (e.g., molecular weight, number of hydroxyl groups, carboxyl groups, amine/amide groups, ether groups, ester groups, etc.) and they also tried some subsets of this set of descriptors to investigate the impact of the input information on the prediction accuracy. This work concluded the following:

 For all output categories, the ANN performed significantly better than the MLR models; in most of the cases the latter resulted in correlation coefficients (R2) lower than the statistically significant values for the model validation stage.  LCIA metrics, such as the CED, GWP, and the EI99 related ones, were better predicted than the BOD, COD, and TOC ones, reaching R2 of 0.6 or higher.  It was not always necessary to use the whole set of descriptors. In a second, more thorough study, Wernet et al. [19] compiled a high quality database based on industrial sources, comprising energy related data for approximately 400 basic and specialty chemicals. Note that the data often included multiple entries from different sources for the same chemical. A set of 31 candidate molecular descriptors was defined based on the results of the first study and industrial expertise. This set is presented in Table 2.1. For the development of the FineChem tool the LCIA metrics that exhibited the best fit in the first study were set as targets. Using a combination of principal component analysis, mutual information-based methods, and hard pruning techniques to simultaneously optimize a set of appropriate molecular descriptors and the ANN structure the authors concluded that the first 10 descriptors of Table 2.1 were the most promising for the development of the FineChem tool. The results demonstrated that the optimal ANN could perform within the desired accuracy to assist process developers and LCA practitioners by filling the frequently encountered data gaps. In particular, the authors estimated the expected validation performance at around 0.55–0.65 in terms of the coefficient of determination. The average relative error was estimated at 35–40%, which however, was significantly lower for molecules with higher energy requirements (e.g., dropping to around 25% for CED>100 MJ-eq/kg). The applicability range of the FineChem tool for the case of CED is defined by the ranges presented in Table 2.2. It is worth to mention that the presence of multiple entries for the same chemical allowed an estimation of the relative standard deviation in the retrieved LCIA values (e.g., for CED this was 24%). This inherent variability of the input data should set a lower bound for the best possible performance in terms of relative error of the models. To enhance the interpretability of the FineChem tool, it was important to investigate if certain molecular descriptors were associated with better prediction statistics compared to others. An analysis of trends in the relative errors did not point out such biases. For example, Figure 2.2 demonstrates that such a distinction could not be made on the basis of the molecular weight. Similar results were obtained for the distribution of the relative error with respect to the total number of functional groups. A systematic thorough analysis of this type is,

33

34

2 Shortcut Models Based on Molecular Structure for Life Cycle Impact Assessment

Table 2.1 The set of the molecular descriptors considered in the FineChem tool. 1 Molecular weight 2 Functional groups (total) 3 Oxygen atoms in keto and aldehyde groups 4 Oxygen atoms not in keto and aldehyde groups 5 Nitrogen atoms 6 Halogen atoms 7 Aromatic or aliphatic rings 8 Ternary or quarternary carbon atoms 9 Heteroatoms in rings 10 Unique substitutes on aromatic rings 11 Hydroxyl groups 12 Carboxylic acid groups 13 Amine/amide groups 14 Nitro groups 15 Chlorine atoms 16 Ether groups 17 Esters and acid anhydrides 18 Cyanide groups 19 Keto groups 20 Aldehyde groups 21 Aromatic rings 22 Fluorine atoms 23 Isocyano groups 24 Triazine ring structures 25 Carbamate groups 26 Sulfonic acid groups 27 Nonaromatic C C double bonds 28 Heteroatoms in other, not listed functional groups 29 Aromatic carbon atoms 30 Heteroatoms 31 Oxygen atoms The terms “groups,” “atoms,” “bonds,” and so on refer to the respective total number in the molecule. Source: Reproduced from Ref. [19] with permission from the Royal Society of Chemistry.

Table 2.2 Composition of the training dataset for the FineChem tool comprising the inventories of 394 chemicals. Molecular weight

Functional groups

CED [MJ-eq/kg]

Minimum

30

0

28

Maximum

1177

28

4201

Mean

149

3

167

Median

111

3

93

Source: Reproduced from Ref. [19] with permission from the Royal Society of Chemistry.

2.3 Illustrative Applications of the FineChem Tool

Figure 2.2 Relative error of the FineChem tool molecular weight trend). (Reproduced from versus molecular weight of chemicals used for Ref. [19] with permission from the Royal Socitraining (solid red line: molecular weight, solid ety of Chemistry.) black line: relative error, dashed black line:

however, still pending. Considering the “black box” nature of the underlying neural networks used to provide the predictions of the LCIA metrics, this can be considered the most significant downside of the FineChem tool. 2.3 Illustrative Applications of the FineChem Tool

In this section, we intend to provide two illustrative research areas where the FineChem tool has been successfully applied to fill in data gaps. They are not intended to designate an exhaustive set of applications where the FineChem tool (or other tools of this type) can be applied but only to illustrate the potential benefits and challenges that still exist in using this type of tool. 2.3.1 LCA Aspects of Solvent Selection for Postcombustion CO2 Capture (PCC)

Solvent makeup in amine-based PCC applications accounts for almost 10% of the operating cost in a PCC plant [26]. The main reasons of solvent make up are the solvent oxidative and thermal degradation under the PCC operating conditions as well as (but in less extent) some fugitive emissions. Degradation has an additional environmental effect, namely the release of solvent degradation products that can be associated with increased freshwater and terrestrial ecotoxicity, as well as health concerns since nitrosamines as degradation products are known to be carcinogenic.

35

36

2 Shortcut Models Based on Molecular Structure for Life Cycle Impact Assessment

The FineChem tool can thus, provide cradle-to-gate LCIA metrics associated with the solvent production phase. Other tools are required for predicting the type and effect of degradation products. Until today, there are no such tools available for estimating, in a relatively accurate way (e.g., similar to the accuracy of the FineChem tool), neither the degree of solvent degradation nor the type and amounts of degradation products on the basis of the solvent molecular structure. Although the cradle-to-gate LCIA metrics for solvent production do not provide a comprehensive estimation of the environmental impacts associated with the solvent selection stage in PCC applications, some studies attempted to use the FineChem tool estimations in the solvent screening stage [27]. The FineChem tool was used in the postassessment stage of a pool of solvents that were designed or preselected using a computer-aided molecular design (CAMD) framework. It is important to note that less than 5% of the identified solvents in this study had available life cycle inventories in databases such as Ecoinvent. Notwithstanding the importance of this application, one could argue that it would be even more beneficial if the FineChem were already integrated into the CAMD procedure to allow the evaluation of LCA criteria during the solvent selection already in the first stage. Leaving aside any arguments about the necessity of a hierarchy in decision-making (i.e., in this case one could argue that it is more beneficial to identify technologically promising solvents and only then to screen the more environmentally benign ones), there are practical problems in integrating the FineChem tool in such a procedure. Perhaps, the more significant one is that when FineChem needs to be included in a CAMD procedure, where group contribution methods are typically used (i.e., using different molecular group definitions than those identified in FineChem), additional algorithmic approaches have to be designed to ensure compatibility; these algorithmic approaches are not straightforward, because quite often the transformation from one molecular description to the other is not an injective function [16]. 2.3.2 Bio-Based Production of Platform Chemicals

Recent studies have thoroughly investigated the cradle-to-gate environmental impacts associated with the production of platform chemicals starting from biobased feedstock. For instance, Morales et al. [28] provided detailed flowsheets for various scenarios of producing bio-lactic acid from glycerol using bio– chemo–catalytic cascades. They used FineChem tool to fill in data gaps with respect to the cradle-to-gate environmental impact of chemical auxiliaries (e.g., extraction solvents) in these flowsheets that could not be retrieved from databases such as Ecoinvent. Another similar study of this kind for succinic acid downstream separation technologies was performed [29], where FineChem tool was used for the same purpose.

2.4 Toward A New Group Contribution-Based Version of the FineChem Tool

It is worth to mention that FineChem tool in its present form is not fit to predict the LCIA metrics of bio-succinic acid production. The reason is simply that no bio-based production data were used in the training set during the development of FineChem. A very recent study of this kind highlighted additional challenges in the development of a BioChem tool tailored to bio-based production [30]. The authors identified that for the not yet optimized bio-based production (e.g., either stand alone or in biorefinery systems), it may be necessary to involve some additional, yet generic, process related parameters (e.g., characterizing the degree of integration, the number of synthesis steps, etc.) in order to achieve a performance comparable to FineChem’s in the precursor feasibility study of Wernet et al. [20].

2.4 Toward A New Group Contribution-Based Version of the FineChem Tool

The following sections discuss the development of group contribution models for the prediction of three LCA indices: GWP, CED, and EI99. These are the same indices as those estimated by the FineChem tool. The incentive of a new group contribution method of this kind is to tackle the challenges identified in applications like those described in section 2.3.1, where the FineChem tool had to be used in a CAMD framework. First, we present the basic principles of group contribution (GC) models used in the prediction of properties of nontabulated compounds. Then, we describe the development of suitable GC models for the above-mentioned LCA indices. After this, a generic screening process for searching of molecular configurations that exhibit desirable and/or optimal behavior is presented. Finally, we illustrate how the developed GC tools for LCA indices can support the screening procedure. 2.4.1 Introduction to GC models

Only a very small fraction of organic molecules have been studied experimentally and their properties have been measured and made available in databases. For the case of lack of experimental data, several classes of models have been developed to enable the prediction of properties based on the molecular structure. GC methods assume that the behavior of a molecule depends mainly on a set of functional groups and their composition in the molecular structure. GC methods can provide results for a wide range of pure component properties, such as critical properties [31], phase transition enthalpies and phase change temperatures [32], heat capacity and viscosity [33], to name just a few representative examples. To use GC models, organic molecules are represented as a metaset of functional groups, for instance, the ethyl vinyl ketone (smiles: O C(C C)CC)

37

38

2 Shortcut Models Based on Molecular Structure for Life Cycle Impact Assessment

is represented as being comprised of the groups CH2 ˆCH ; CH3 ; and CH2 …CˆO† in the following composition: 1  CH2 ˆCH ; 1 

CH3 ; and 1 

CH2 …CˆO†

Properties are then formulated as functions of structurally dependent parameters, depending on the occurrence frequencies of each group in the molecule and the group’s contribution coefficient. These models provide the advantage of quick and relatively accurate estimations without requiring substantial computational resources. The drawbacks of GC methods are their inability to distinguish among isomers, due to the simplified representation of the molecular structure, and their low accuracy, due to the relatively small amount of experimental data available for the estimation of GC model coefficients. In the typical GC formulation for a specific property, X, the left-hand part is a (linear or nonlinear) function to calculate X. The right-hand part is the sum of the products between the number of occurrences of each group in the molecule and the group’s contribution coefficient. This part is linear in terms of the groups’ occurrence frequencies. More complicated formulae involve additional linear terms in the right-hand part using higher order functional groups [31]. Second- and third-order groups are multisets of first-order functional groups joined together to consider additional structural information and increase the prediction accuracy. 2.4.2 Development of GC-Based LCA Models

The development of a GC model for a property, X, consists of (a) determining the best model formulation and (b) estimating the best set of values for the group contribution coefficients. The screening between alternative formulations and the values of their coefficients is based on statistical estimators to minimize the differences between the experimental and calculated values of X in a set of molecules. This set is called the training set and it should be sufficiently large and sufficiently versatile for the procedure to be reliable. The experimental value of each molecule in the training set is an observation. The regression methodologies used in the present work for the development of GC models for the LCA indices GWP, CED, and EI99 are: multiple linear regression (MLR), principal component analysis (PCA), partial least squares (PLS), kriging, radial basis functions (RBF), and radial basis functions with PCA. MLR is a simple regression technique that is used in order to estimate coefficient values in a linear correlation. In the PCA method, the input variables (group occurrences) are transformed into a set of latent variables called principal components (PC). The system is then described by a number of PCs, typically smaller than the number of the original input variables to achieve dimensionality reduction of the input space. However, the number of PCs should be sufficiently large to represent the system’s variability (expressed as the variance of the original variables). Therefore, the user must carefully choose the number of PCs used

2.4 Toward A New Group Contribution-Based Version of the FineChem Tool

to represent the system. The correlation is then developed similar to MLR method, called principal component regression (PCR). PLS applies the same principles as PCA, with the difference that both input and output variables (LCA indices) are transformed into latent variables, called components. The correlation is now developed similar to the MLR method between input and output latent variables. Kriging is a nonlinear method that does not adopt the typical regression approach. This method projects the M training set compounds, as points {T1, T2, . . . , Tm} to an N-dimensional space where N is the number of variables. The estimation for a new random compound C will be an interpolation between T1, T2, . . . , Tm, which will be based on the calculation of the distance of C with T1, T2, . . . ,Tm separately (dis(C,T1), dis(C,T2), . . . , dis(C,TM)). These distance values will then be weighted with certain factors and their sum will provide the estimation of the index value. RBF can be formulated as a neural network methodology. The concept is similar to kriging, but it uses a set of centers instead of the {T1, T2, . . . , Tm}. These centers might represent, but not necessarily coincide with, one or more training set observations. The set of centers used to represent the system is called a grid. The estimation of a new random compound will then be interpolated between available centers, similar to the distance-based procedureof kriging. If RBF is not applied to the original variables but to a smaller number of transformed variables via PCA, then this designates the RBF–PCA approach. The following procedure describes the steps used here for the development of three models, one for each LCA index, LCAi, i∈{GWP, CED, EI99}: Step 0: Acquire a database, D, of molecular compounds and their known LCAi values. Step 1: Select a subset, DC, of D, based on a set of criteria. Step 2: Consider a formulation F to model LCAi Step 3: Consider a model development technique RTj, j∈{MLR, PCA, PLS, Kriging, RBF, RBF–PCA} Step 4: Consider a subset, DC,training, of DC, to be used for the training Step 5: Apply RTj on DC,training to generate the coefficient values for formulation F, and calculate statistical indices (errors, deviations, etc.) Step 6: Repeat steps 4–5 for all subsets of Dc and merge the outputs to a final set of F coefficients Step 7: Repeat steps 2–6 for different formulations, F, and compare the statistical indices, to find (a) the best formulation, F∗, and (b) the values of its coefficients. Step 0: As discussed, in order to develop an estimation model, a set of “experimental” values must be provided, in order to be regressed with the models input. In the LCA index case, the database (D) consists of the Ecoinvent database and a set of compounds whose LCA indices were estimated using the FineChem tool. Step 1: Not all compounds can participate in the model development process and a subset DC was finally used. This consists of compounds that can be

39

40

2 Shortcut Models Based on Molecular Structure for Life Cycle Impact Assessment

decomposed using the available molecular groups for the development of the models. Step 2: It is necessary to provide a guess for the formulation of the GC model. This formulation (output) includes the group contribution coefficients and constants of the model, and the regression tools aim to estimate the “optimal” set of their values. Optimality is in terms of how well a set of coefficient values can predict the known property values of the training set. This is by no means a trivial procedure and does not yield a single answer. Different regression tools and different tuning of these tools may well result to different “optimal” sets of coefficients. The formulation F considered for the modeling of the LCA indices is a typical, generic, linear group contribution correlation. Step 3: The model development process involves a selected model development technique of those mentioned before (i.e., MLR, PCR, etc.) and will define the coefficient and intercept values for all three LCA index models. Step 4: The initial DC will be further divided into a training set, Dc,training containing the majority of the compounds, which will be used in the regression process and a testing set, Dc,testing that will be used to test the GC model’s predictive abilities. Since, the results of a GC model depend strongly on the compounds participating in the training and testing sets, a wide range of training/testing sets combination (partitions) is regressed, in order to have an unbiased model selection. Step 5: After applying a model development technique to a partition, values for the coefficients and the intercept are estimated and a set of statistical indices, such as relative errors R2, coefficient of determination, are calculated for both Dc,training and Dc,testing, in order to assess the model’s predictive abilities. Emphasis is given on the coefficient of determination that expresses a measure of the model’s overall correlation quality and of course, the relative error is also taken under consideration. Step 6: This step is repeated for all partitions. Finally, the coefficient values of each group are merged by calculating the average between all contribution values of this group in every partition. Average values are also calculated for the statistical indices used. Step 7: This procedure is repeated for all model development methods and the average statistical values will provide an indication of the most reliable correlation. 2.4.3 Screening for Substances with Desirable Properties

Since the GC models establish a mapping between structure and property (i.e., LCA index), they can be reversed to screen for molecules of desirable behavior. Here, we will demonstrate how the models developed according to the procedure in the previous section can be used together with other group contribution

2.4 Toward A New Group Contribution-Based Version of the FineChem Tool

tools to design molecules according to a set of property constraints. The design procedure is herein perceived as an iterative process, starting from a random molecule and applying consecutive changes, until a molecule that meets the standards is obtained. This section presents the molecular representation and a set of modifications to support the screening procedure. Let a molecule be represented as a multiset, M, of all the functional groups present in the molecule, also accounting for their number of occurrences. For example, ethane is represented as the following: M1 ˆ fCH3 ; CH3 g

or f2  CH3 g

Aromatic carbon atoms, denoted as AC, are handled differently than aliphatic carbon atoms, C. In this sense, there are two types of functional groups: nonaromatic and aromatic groups. Nonaromatic groups may contain only carbon atoms that do not participate in aromatic rings (only C). Aromatic groups contain one carbon atom that is part of an aromatic ring (one AC). The default aromatic group is ACH, and larger groups result from one AC bonded to one nonaromatic substituent, for instance, ACCH3 or ACCH CH . Toluene is then represented as the following: M2 ˆ f5  ACH; AC CH3 g While non-aromatic compounds only contain non-aromatic groups, aromatic compounds may contain both aromatic and non-aromatic groups. In aromatic compounds, the number of aromatic groups must always be a multiple of 6. This rule is part of a list of (structural) constraints that a multiset of groups needs to comply to in order to represent a feasible molecular structure. Table 2.3 presents a list of common functional groups, with their valence and type. Note that, the mappings fτ(G), fσ-(G) and fσ + (G) are explained later in the text. The search for new molecular configurations that exhibit desirable properties is considered here as an iterative procedure, starting from an initial feasible multiset of groups (guess) and generating a series of feasible multisets. The iterative procedure requires a set of general and systematic modification procedures, to perturb from one feasible group multiset to another. A generic set of such procedures could involve addition of a new functional group in the multiset, to remove an existing group, to substitute an existing group with another group, to add or remove aromatic rings, and so on. Following are examples of how such modification procedures can be applied in practice:

 Addition of a new functional group: Increase the number of groups that compose a molecule. For instance, add one

CH2

group to CH3 CH2 CH(CH3)2

f3  CH3 ; CH2 ; CH CH2 >CH

3

>CH

3

Nonarom

ACCH
CH2

>C
C
CH O

CH3 NH

>CH NH

7

CH2 O

10

CH2 O

11

CH2 NH

2

Nonarom

AC

12

CH(Cl)

2

Nonarom

AC CH(Cl)
C(Cl)

13

OH

1

Nonarom

AC

—

—

COOH

AC COOH

—

—

1

Nonarom

15

14

ACCH3

0

Aromatic

CH3

16

ACCH2

1

Aromatic

17

ACCH
CH2

ACCH3

ACCH
CH

ACCH2

ACC
C < , therefore f3  CH3 ; CH2 ; CH C C CH Cl, etc. The choice to add CH CH will lead to CH3 O CH2 CH (Cl) CH CH OH. We could then replace CH3 O with CH2 C CH , replace OH with COOH, add CH2 or any other possible move. If the substitution of OH

2.4 Toward A New Group Contribution-Based Version of the FineChem Tool

45

with COOH is chosen, the resulting configuration will be CH3 O CH2 CH(Cl) CH CH COOH. This new structure can be modified by replacing CH3 O with CHC , removing CH Cl, replacing COOH with C(Cl)3, etc. The removal of >CH Cl will lead to CH3 O CH2 CH CH COOH, and so on. The screening steps described above are part of the optimization methodology developed and implemented by Marcoulaki and Kokossis [34] and adapted to the CAMD of solvents [35]. Table 2.4 reports the values of the heat of vaporization ΔHv and the boiling point temperature Tb as well as LCA indices GWP, CED, and EI99 for the configurations generated above. The thermodynamic properties are estimated according to existing models and the LCA metrics according to newly developed GC-based models. According to Table 2.4, the best option is probably CH3 O CH2 CH(Cl) CH CH COOH, due to the highest heat of vaporization and boiling point value. However, it must be noted that the same solution yields the highest LCA metrics indicating the least environment-friendly behavior. It is also possible to trade-off between performance and an efficient environmental behavior by choosing other solutions. It must also be mentioned that the GWP index appears to be proportional to the molecule size. The CED index seems to be proportional to the GWP, as the same trend with the compound’s size appears here too.

Table 2.4 Physical properties and LCIA metrics of the generated molecular configurations. Group multiset

Example of molecule

ΔHv Tb (K) GWP CED (kg-eq CO2) (MJ-eq) (kcal/mol)

{CH2 CH , CH2 O , CH2 NH , OH}

CH2 CH CH2 O CH2 NH OH

5378

454

5.47

143

{CH2 CH , CH2 O , CH2 , CH2 NH , OH}

CH2 CH CH2 O CH2 CH2 NH OH 5594

474

5.61

145

{ CH3, CH2 O , CH2 , CH2 NH , OH}

CH3 CH2 O CH2 CH2 NH OH

5262

455

4.30

114

{ CH3, CH2 O , CH2 , OH}

CH3 CH2 O CH2 OH

4144

387

2.72

74.0

{CH3 , CH2 , OH}

CH3 CH2 OH

3339

330

1.79

54.6

{CH3 O , CH2 , OH}

CH3 O CH2 OH

4325

379

2.46

66.4

{CH3 O , CH2 , CH(Cl) , OH}

CH3 O CH2 CH(Cl) OH

5519

451

3.65

82.0

{CH3 O , CH2 , CH(Cl) , CH CH , OH}

CH3 O CH2 CH(Cl) CH CH OH

5942

489

7.86

184

{CH3 O , CH2 , CH(Cl) , CH CH , COOH}

CH3 O CH2 CH(Cl) CH CH COOH

6380

533

8.63

176

{CH3 O , CH2 , CH CH , COOH}

CH3 O CH2 CH CH COOH

5533

487

7.43

160

46

2 Shortcut Models Based on Molecular Structure for Life Cycle Impact Assessment

2.5 Conclusions and Outlook

In this chapter, a state-of-the-art molecular structure-based tool (FineChem) for the cradle-to-gate estimation of life cycle impact assessment metrics is first placed within the scope of LCA as a method to fill in data gaps, which inevitably appear in complicated chemical process synthesis, especially in early stages of process design. The concept and development of the FineChem tool is briefly presented, focusing on how the required molecular descriptors were derived and highlighting the general performance of the model. We also provide information about illustrative applications where the tool has been recently applied and we highlighted some shortcomings regarding its interpretability, incorporation in automated CAMD procedures, and application in the production of biochemicals. In the second part of the chapter, we discuss current efforts to develop a new version of the FineChem tool on the basis of group contribution methods that are compatible with available CAMD tools. The approach is presented in detail, followed by an illustrative example of how molecules can be generated during a CAMD procedure, and assessed using the new version of the FineChem tool and other group contribution models. Although the new FineChem models could not be presented here since they are undergoing fine tuning, this second part aims to highlight the benefits of such an approach in terms of increased interpretability and compatibility with existing methods for screening chemicals based on diverse performance criteria. There are currently other studies under development, trying to provide shortcut models similar to the FineChem tool for bio-based production. Other attempts try to improve the performance of the model by considering a hybrid input consisting of both molecular descriptors and thermodynamic properties, carefully selected to capture LCA relevant aspects of the molecule production. Sooner or later the FineChem tool will have to be updated in order to be able to provide estimations for the ReCiPe indices [36], since the EI99 has in the meantime become an obsolete LCA metric. These are only some examples of future research in the topic of developing shortcut models for estimation of LCA relevant metrics. As the acceptability of LCA metrics is increasing among engineers and sustainability practitioners for the environmental assessment of chemical production, it is expected that the use of such shortcut models will become an indispensable tool in the hands of chemical and process engineers to introduce a sustainability dimension in early to basic stages of chemical process design.

References 1 European-Committee. ISO 14040 (2006)

Environmental management Life cycle assessment Principles and framework.

2 European-Committee. ISO 14044 (2006)

Environmental management Life cycle assessment Requirements and guidelines.

References 3 Kralisch, D., Ott, D., and Gericke, D.

4

5

6

7

8

9

10

11

12

(2015) Rules and benefits of life cycle assessment in green chemical process and synthesis design: a tutorial review. Green Chemistry, 17 (1), 123–145. Wernet, G., Conradt, S., Isenring, H.P., Jiménez-González, C., and Hungerbühler, K. (2010) Life cycle assessment of fine chemical production: a case study of pharmaceutical synthesis. The International Journal of Life Cycle Assessment, 15 (3), 294–303. Jimenez-Gonzalez, C. and Overcash, M.R. (2014) The evolution of life cycle assessment in pharmaceutical and chemical applications – a perspective. Greem Chemistry, 16 (7), 3392–3400. Barton, J.R., Dalley, D., and Patel, V.S. (1996) Life cycle assessment for waste management. Waste Management (Oxford), 16 (1–3), 35–50. Gentil, E.C., Damgaard, A., Hauschild, M., Finnveden, G., Eriksson, O., Thorneloe, S. et al. (2010) Models for waste life cycle assessment: review of technical assumptions. Waste Management (Oxford), 30 (12), 2636–2648. Seyler, C., Hofstetter, T.B., and Hungerbühler, K. (2005) Life cycle inventory for thermal treatment of waste solvent from chemical industry: a multiinput allocation model. The Journal of Cleaner Production, 13 (13–14), 1211–1224. Winkler, J. and Bilitewski, B. (2007) Comparative evaluation of life cycle assessment models for solid waste management. Waste Management (Oxford), 27 (8), 1021–1031. Khoo, H.H., Ee, W.L., and Isoni, V. (2016) Bio-chemicals from lignocellulose feedstock: sustainability, LCA and the green conundrum. Green Chemistry, 18 (7), 1912–1922. Munoz, I., Flury, K., Jungbluth, N., Rigarlsford, G., Canals, L.M.I., and King, H. (2014) Life cycle assessment of biobased ethanol produced from different agricultural feedstocks. The International Journal of Life Cycle Assessment, 19 (1), 109–119. Panichelli, L., Dauriat, A., and Gnansounou, E. (2009) Life cycle assessment of soybean-based biodiesel in

13

14

15

16

17

18

19

20

Argentina for export. The International Journal of Life Cycle Assessment, 14 (2), 144–159. Kniel, G.E., Delmarco, K., and Petrie, J.G. (1996) Life cycle assessment applied to process design: environmental and economic analysis and optimization of a nitric acid plant. Environmental Progress, 15 (4), 221–228. Anastas, P.T. and Lankey, R.L. (2000) Life cycle assessment and green chemistry: the yin and yang of industrial ecology. Green Chemistry, 2 (6), 289–295. Lankey, R.L. and Anastas, P.T. (2002) Lifecycle approaches for assessing green chemistry technologies. Industrial and Engineering Chemistry Research, 41 (18), 4498–4502. Papadokonstantakis, S., Karka, P., Kikuchi, Y., and Kokossis, A.C. (2016) Challenges for model-based life cycle inventories and impact assessment in early to basic process design stages, in Sustainability in the Design, Synthesis and Analysis of Chemical Engineering Processes (eds G. Ruiz-Mercado and H. Cabezas), Elsevier, Cambridge, USA, pp. 295–326. Curzons, A.D., Jimenez-Gonzalez, C., Duncan, A.L., Constable, D.J.C., and Cunningham, V.L. (2007) Fast life cycle assessment of synthetic chemistry (FLASC (TM)) tool. The International Journal of Life Cycle Assessment, 12 (4), 272–280. Jimenez-Gonzalez, C., Curzons, A.D., Constable, D.J.C., and Cunningham, V.L. (2004) Cradle-to-gate life cycle inventory and assessment of pharmaceutical compounds. The International Journal of Life Cycle Assessment, 9 (2), 114–121. Wernet, G., Papadokonstantakis, S., Hellweg, S., and Hungerbühler, K. (2009) Bridging data gaps in environmental assessments: modeling impacts of fine and basic chemical production. Green Chemistry, 11 (11), 1826–1831. Wernet, G., Hellweg, S., Fischer, U., Papadokonstantakis, S., and Hungerbühler, K. (2008) Molecularstructure-based models of chemical inventories using neural networks. Environmental Science and Technology, 42 (17), 6717–6722.

47

48

2 Shortcut Models Based on Molecular Structure for Life Cycle Impact Assessment 21 Huijbregts, M.A.J., Hellweg, S.,

22

23

24

25

26

27

28

Frischknecht, R., Hendriks, H.W.M., Hungerbühler, K., and Hendriks, A.J. (2010) Cumulative energy demand as predictor for the environmental burden of commodity production. Environmental Science and Technology, 44 (6), 2189–2196. Goedkoop, M. and Spriensma, R. (2000) The eco-indicator 99: a damage orientated method for life-cycle impact assessment. Methodology Annex, Pre-Consultants, The Netherlands. Bumann, A.A., Papadokonstantakis, S., Sugiyama, H., Fischer, U., and Hungerbühler, K. (2010) Evaluation and analysis of a proxy indicator for the estimation of gate-to-gate energy consumption in the early process design phases: the case of organic solvent production. Energy, 35 (6), 2407–2418. Köhler, A., Hellweg, S., Recan, E., and Hungerbühler, K. (2007) Input-dependent life-cycle inventory model of industrial wastewater-treatment processes in the chemical sector. Environmental Science and Technology, 41 (15), 5515–5522. Pereira, C., Papadokonstantakis, S., Rerat, C., and Hungerbühler, K. (2013) Industrial documentation-based approach for modeling the process steam consumption in chemical batch plants. Industrial and Engineering Chemistry Research, 52 (44), 15635–15647. Rao, A.B. and Rubin, E.S. (2002) A technical, economic, and environmental assessment of amine-based CO2 capture technology for power plant greenhouse gas control. Environmental Science and Technology, 36 (20), 4467–4475. Papadopoulos, A.I., Badr, S., Chremos, A., Forte, E., Zarogiannis, T., Seferlis, P. et al. (2014) Efficient screening and selection of post-combustion CO2 capture solvents Chemical Engineering Transactions, 39, 211–216. Morales, M., Dapsens, P.Y., Giovinazzo, I., Witte, J., Mondelli, C., Papadokonstantakis, S. et al. (2015) Environmental and economic assessment of lactic acid production from glycerol using cascade

29

30

31

32

33

34

35

36

bio- and chemocatalysis. Energy and Environmental Science, 8 (2), 558–567. Morales, M., Ataman, M., Badr, S., Linster, S., Kourlimpinis, I., Papadokonstantakis, S. et al. (2016) Sustainability assessment of succinic acid production technologies from biomass using metabolic engineering. Energy and Environmental Science, 9 (9), 2794–2805. Karka, P., Papadokonstantakis, S., Hungerbühler, K., and Kokossis, A. (eds) (2014) Environmental Impact Assessment of Biorefinery Products Using Life Cycle Analysis. 8th International Conference on Foundations of Computer-Aided Process Design – FOCAPD 2014, Elsevier, Washington, USA. Constantinou, L. and Gani, R. (1994) New group-contribution method for estimating properties of pure compounds. AlChE Journal, 40 (10), 1697–1710. Marrero, J. and Gani, R. (2001) Groupcontribution based estimation of pure component properties. Fluid Phase Equilibria, 183, 183–208. Joback, K.G. and Reid, R.C. (1987) Estimation of pure-component properties from groupcontributions. Chemical Engineering Communications, 57 (1–6), 233–243. Marcoulaki, E.C. and Kokossis, A.C. (2000) On the development of novel chemicals using a systematic synthesis approach. Part I. Optimisation framework. Chemical Engineering Science, 55 (13), 2529–2546. Marcoulaki, E.C. and Kokossis, A.C. (2000) On the development of novel chemicals using a systematic synthesis approach. Part II. Solvent design. Chemical Engineering Science, 55 (13), 2547–2561. Goedkoop, M., Heijungs, R., Huijbregts, M.A.J., De Shruyver, A., Struijs, J., and Van Zelm, R. (2009) ReCiPe 2008–A life cycle impact assessment method which comprises harmonised category indicators at the midpoint and the endpoint level, Ministerie van Volkhuisvesting, Ruimtleijke Ordening en Milieubeheer, The Netherlands.

49

3 Models to Estimate Fate, Exposure, and Effects of Chemicals Rosalie Van Zelm, Rik Oldenkamp, Mark A.J. Huijbregts, and A. Jan Hendriks

3.1 Introduction

It is common knowledge that chemical processes and substances cause environmental problems. The green chemistry principles seem clear and easy to live up to, however, how do you evaluate whether a chemical really is safer (i.e., less toxic), a renewable feedstock is more environment-friendly, or a catalytic reagent is always superior to a stoichiometric reagent? Especially since every few seconds, one new substance is added to the more than 109 000 000 already registered in the chemical abstracts service registry (www.cas.org). Performing tests on all these substances in all the species and environmental backgrounds is nearly impossible. When entering the environment, a substance can travel many paths, depending on environmental properties and physicochemical properties, to affect plant and animal species and human health issues. In the European Union alone, 100 000+ compounds are awaiting assessment, 1 500 000 chemically polluted sites potentially require cleanup, and unknown chemicals are responsible for up to 90% of the toxicity. Worldwide, 8 000 000+ species, of which 10 000+ are endangered, need protection [1]. To assist in determining and communicating the potential human and environmental health impacts of chemicals, methods, such as risk assessment and life cycle assessment, are required. These methods typically follow a driver– pressure–state–impact–response (DPSIR) framework. Environmental problems are ultimately induced by human needs such as food, shelter, mobility, prosperity and entertainment [2]. These needs can be met by intervening activities for which chemical processes are required, including manufacturing and engineering, mainly within the societal sectors industry and households. The activities and sectors cause pressures, for example, emission of contaminants, extraction of resources, and land use changes. As a result, the physical, chemical, and biological state of the environment may change, having an impact on health, prosperity, and other values. The response by society may involve a change in any

Handbook of Green Chemistry Volume 10: Tools for Green Chemistry, First Edition. Edited by Evan S. Beach and Soumen Kundu.  2017 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2017 by Wiley-VCH Verlag GmbH & Co. KGaA.

50

3 Models to Estimate Fate, Exposure, and Effects of Chemicals

link of this chain. For example, to prevent or recover from illnesses (needs), we develop medicines that require production of pharmaceuticals (activity). The use and especially excretion of pharmaceuticals cause emissions to the environment (pressure), causing changes in pharmaceutical concentrations and species composition (chemical, biological, and medical state). Such changes of the states are generally considered to affect values we care for such as health, nature, and prosperity (impact). This chapter will outline models to quantify fate, exposure, and effects of chemicals on humans and the environment. These models can be applied in risk assessment of chemicals and life cycle assessment of products. These models are able to handle multiple chemicals, based on specific properties of chemicals and species. In other words, models will be addressed that can be used in case of limited data availability (which is in fact very often the case). Subsequently, we discuss the application of the methods in risk assessment and life cycle assessment. We end by outlining the recent developments in the field to improve the existing models.

3.2 Fate

Multimedia fate models can be used for the prediction of chemical concentrations in the environment based on physical–chemical properties of the chemical of concern, environmental characteristics, and emission data [3]. In this type of model, the study area is represented by a number of homogeneous compartments, each representing a specific part of the environment (i.e., atmosphere, water, soil). The concentration of a chemical in a certain compartment can be calculated by solving the mass balance under steady state conditions with the help of linear algebra calculation rules. Steady state means that concentrations do not change over time in the compartments considered, when there is a constant emission rate. For different chemicals, different processes are important. Which processes are the most important for a certain compound depends on the physical–chemical properties of the compound. Also, the environmental conditions (temperature, rain intensity, etc.) influence the model predictions. A fate model accounts for both removal processes and intermedia transport processes of chemicals in the environment. Examples of removal processes are (bio)degradation by microorganisms, transport of the chemical to the sediment, leaching to the groundwater, and escape to the stratosphere. Intermedia transport processes account for movement of chemicals from one compartment to the other (and back). Two types of intermedia transport processes exist: advective and diffusive transport. In the case of advective transport, the chemical moves with an environmental medium from one compartment to the other (one-way transport). For example, rivers transport a chemical from freshwater to seawater and rain transports a chemical from air to the Earth’s surface. Diffusive transport between two compartments, on the other hand, is passive two-way

3.2 Fate

transport, that is, the chemical can move from one compartment to another and back. Diffusive transport from air to water is called gas absorption, while diffusive transport from water to air is called volatilization. An important characteristic of intermedia transfer rates and removal rates is that they depend strongly on the properties of a chemical. The following are a few examples:

 Chemicals that are easily transformed by microorganisms have high degrada 

 

tion rates in soil, water, and sediment, while chemicals that are not susceptible to biodegradation will be persistent in these compartments. Chemicals that tend to bind strongly to particles (hydrophobic chemicals) have a relative high removal from water to sediment via sedimentation of suspended particles. Chemicals with a high vapor pressure (and low solubility) will have a relatively high tendency for passive transport from water to air. USEtox is an example of a multimedia fate model that works with two geographical scales [4]: The continental scale with the following compartments: urban air, rural air, freshwater, sea, natural soil, and agricultural soil. The global scale with the following compartments, air, freshwater, ocean, natural soil, and agricultural soil.

The continental scale is nested in the global scale (see Figure 3.1). “Nested” means that chemicals can be transported from one scale to a higher scale and vice versa. Multimedia fate models are built with increasing spatial complexity. The advantage is that the influence of differences in landscapes on the fate of chemicals can be assessed. This can be done by expanding the number of boxes in a

Figure 3.1 Nested structure of the USEtox model [4].

51

52

3 Models to Estimate Fate, Exposure, and Effects of Chemicals

spatially explicit way [5]. An alternative is to keep the original model structure as it is, but resampling environmental properties for different landscapes [6,7].

3.3 Ecological Exposure

Models can also be used to translate environmental concentrations to concentrations in the organism itself. Physiologically based pharmacokinetic (PBPK) models to predict concentrations in various organs of humans were already developed in the early twentieth century [8]. Likewise, building of models covering different species in food webs started a few decades later [9]. While very useful for a specific chemical (e.g., DDT) or species (e.g., man), these multicompartment models require many coefficients to be parameterized. As an alternative, one-compartment models have been developed, with uptake and elimination rates linked to well-known chemical properties, such as the octanol–water partition ratio Kow [10]. To cover variability between species too, rates were also related to biological traits, for example, body size [11]. In addition, body size-independent parameters, such as assimilation efficiencies and transformation potential for substances, can, with care, be extrapolated from well-studied groups like mammals to other species. These relationships are well-embedded in theories on, for example, chemical partitioning and biological scaling. Based on these overarching principles, we can now obtain default values for uptake and elimination rates of 1000+ chemicals and 1000+ species as a function of well-known chemical characteristics and the species body size, without extensive empirical work. Such initial values may be overridden when chemical-explicit or species-specific rates become available from experiments or surveys. To allow for application to many chemicals and species, one needs to cover as much exposure routes as possible. Chemicals are usually delivered to organisms via air, water, and food, following similar principles (Figure 3.2). After delivery, compounds travel through lipid (membranes), carbohydrate (mucus, wall), and unstirred water layers. In addition, some substances are carried across membranes by protein transporters. After chemicals have entered an organism, they are transported by sap (plants) or blood (animals). The substance will partition across water, lipid, protein, or other body components. Some substances undergo biotransformation. The parent compound or the metabolite can be eliminated via the same routes, that is, via air, water, or food. In addition, there is an “apparent” route of exit called growth dilution, describing how concentrations will decrease by the addition of tissue. The ratio of the inflow and outflow rate constants can be used to estimate accumulation factors, that is, ratios of concentrations in the organism, on the one hand, and those in air, water, soil, or food, on the other. By comparing accumulation factors based on lab experiments to values obtained from field surveys, the model is validated by independent data. In general, accumulation in lab

3.3 Ecological Exposure

Figure 3.2 Exchange of chemicals by organisms with air, water, and food determined by their concentrations (Ci, C0a, C0w, Ci-1) and the sums of the resistances and delays encountered in each section. Storage indicated by regular fonts, flows in italics.

experiments is lower than in the field because of differences in exposure time and media (Figure 3.3). Following this approach, internal distribution is determined by the delay of sap or blood flow and the affinity of a chemical for body components [16]. In

Figure 3.3 Accumulation of persistent organic chemicals in food chains, especially of the Rhine delta (Adapted from Ref. [12]). Values in arrows represent typical ratios of concentrations in the upper tropic level divided by

those in the lower tropic level after correction for fat contents, based on calibration with lab data (left-hand side) versus validation with field data (right-hand side) (compilation of Refs [13–15]).

53

54

3 Models to Estimate Fate, Exposure, and Effects of Chemicals

practice, such one-compartment models are often about equally accurate as multicompartment equivalents [17]. However, in specific cases, more refinement may be needed. Exposure to, for example, volatile organics can be best modeled using a PBPK model [18]. Likewise, accumulation in very different body parts, such as above ground and below ground plant organs, also requires different compartments [19].

3.4 Ecosystem Effects 3.4.1 Intraspecies Variability in Populations

Modeling of fate and accumulation allows one to translate chemical emissions to effects by linking concentrations in air, water, soil, food, and the organism itself to levels known to be critical from toxicity experiments. In this approach, the central tool is the concentration–response relationship, traditionally describing the increasing number of individuals affected as a sigmoid function of exposure. Usually, a log-normal or log-logistic curve is fitted to data obtained from laboratory toxicity assays with a certain species (Figure 3.4). The fraction of affected individuals, reflecting intraspecific variability, is used to estimate the impact on growth, reproduction, and survival of a species. In particular, the median lethal value LC50 and the slope β of the curves can be linked to the mode of action of the chemical. For narcosis, for example, the critical body burden equals about 2–3 mmol
kg 1, independent of the chemical and

Figure 3.4 The fraction 1-pos of affected individuals in a species, for example, algae (green), fish (blue), worms (pink), and the corresponding fraction of species affected (red) versus increasing stressor levels, exemplified by chemical concentration or temperature.

3.5 Human Exposure and Effect

the species [16]. For specific modes of action similar, albeit less straightforward, simplifications are possible. While growth, reproduction, and survival are qualitatively relevant, these endpoints do not allow for a direct comparison to ecological quantities used to indicate other stressors as well. 3.4.2 Interspecies Variability in Assemblages

The relationships obtained for different species can be combined to determine the amount of species affected as a function of chemical pollution (Figure 3.4). These so-called species sensitivity distributions (SSDs), thus reflect interspecific variability in tolerance to chemicals. As for intraspecies differences, medians and slopes are related to the mode of action [20]. However, as the mode of action may differ across species, a priori parameter setting is more difficult. Fortunately, SSDs are being derived for an increasing number of chemicals. Even more, most studies indicate that species sensitivity differences between geography (polar, temperate, tropical), compartments (marine, aquatic, terrestrial), and the like are often minimal [21,22]. While SSDs usually encompass as much species as possible, one sometimes selects data on a taxon of specific (protection) interest such as native versus exotic fish [23]. The relevance of population trends estimated from indoor data can be relatively easily tested in outdoor cases. Yet, establishing the field relevance of species sensitivity distributions is more difficult. However, levels that protect 95% of the species in laboratory assays are generally protective for (semi-)field experiments, suggesting that feedbacks between species and between the environment and species weaken rather than enforce effects [24].

3.5 Human Exposure and Effect

In human effect assessment, the relationship between exposure and effect, that is, the dose–response relationship, is established. This dose–response relationship can be used to characterize the effect resulting from a specific exposure concentration. The extent of this human exposure is influenced by the concentrations in contact media, which can be either measured or estimated, and the degree of contact with them. Contact media often considered in the assessment of human exposure to environmental contaminants are drinking water, foodstuffs such as fruits and vegetables, meat products, dairy products, and fish, and environmental media such as surface water, soil, and air. Concentrations in environmental media can be estimated with fate models (Section 3.2) and used as such. Concentrations in drinking water and foodstuffs are generally estimated from those in environmental media, for example through the use of bioconcentration factors (BCFs) for fish [25], root concentration factors (RCFs) for fruit and

55

56

3 Models to Estimate Fate, Exposure, and Effects of Chemicals

vegetables [26], and biotransfer factors (BTFs) for meat and dairy products [27,28]. Concentrations in drinking water depend on the source of the water (e.g., groundwater or surface water), the purification techniques applied, and the removal efficiency of these techniques for the chemical assessed. In addition to the concentrations in contact media, the extent to which exposure to these concentrations takes place is determined by consumption patterns and behavioral characteristics of the exposed population. Food consumption databases, based on large-scale consumption surveys, are often used to characterize the daily consumption of different foodstuffs and drinking water. When these are extensive and detailed enough, specific geographical or age groups might be assessed [7]. Contact with environmental media depends on behavioral characteristics of the exposed population, such as swimming behavior or the rate of ingestion of soil particles. By combining the daily intake of different contact media with concentrations in these media, the daily intake of the chemical of interest can be calculated. This can, for example, be done with the integrated human exposure model NORMTOX [29,30] (Figure 3.5). Coherence is essential between the exposure situation assessed and the circumstances in which the dose–response relationship is derived. This might seem logical and maybe even trivial. However, the circumstances in which exposure to chemicals takes place vary widely in terms of exposure duration, exposure route, location characteristics, and recipient of the exposure (and their specific sensitivity). Clearly, controlled human experiments to derive dose– response relationships are often ethically undesirable. So, while ideally each exposure situation would be assessed via its own representative dose–response relationship, in reality this is not possible.

Figure 3.5 Schematic representation of NORMTOX [29].

3.5 Human Exposure and Effect

Figure 3.6 Schematic representation of fitting a dose–response model on experimental toxicity data.

Therefore, dose–response relationships have traditionally been derived via experimental studies with mammalian test species, such as rats, mice, or dogs, divided in a control group and a number of increasing dose groups. Distinction is made between tests assessing different routes of exposure, of which the oral and inhalatory routes are most relevant in the setting of human effect assessment. Moreover, experimental animal studies are generally classified as (sub-) acute or (sub-)chronic toxicity tests, depending on their duration relative to the life expectancy of the test animal. A typical endpoint in acute toxicity testing is mortality, while in chronic toxicity tests sublethal effects are generally assessed. Animal experiments result in an effect level per dose group, from which a dose–response relationship can be inferred through fitting of a dose–response model using, for example, the Benchmark Dose Software (BMDS) from US EPA (Figure 3.6). Selection of a specific dose–response model type can be made based on several criteria, such as its overall fit, its local fit (at the relevant dose range), its complexity, or the existence of a mechanistic understanding of the effect. Moreover, specific distinction is made between dose–response models describing continuous endpoints and those describing quantal endpoints. Continuous endpoints are gradually changing effects at the level of the individual test animal, for example, bodyweight or red blood cell count. Dose–response relationships for continuous endpoints are thus derived by fitting on the average responses per dose group. Quantal endpoints, on the other hand, reflect a dose-related change in incidence in the dose group as a whole, for example, the number of deaths or number of animals that develop tumors. Consequently, the dose– response relationship derived from quantal toxicity data (partly) depends on the variation in sensitivity within the group of test animals, and so will the magnitude of effect. Test animals with little (genetic) diversity are generally used to minimize the influence of this intraspecies variation.

57

58

3 Models to Estimate Fate, Exposure, and Effects of Chemicals

After derivation of a dose–response relationship for the test species, it should be extrapolated towards the exposure scenario of concern. This involves one or more extrapolation steps. These steps, of course, include extrapolation for differences between test species and humans, and likely also for differences between the average human and the human subpopulation of interest (e.g., children or elderly people). Depending on the coherence between the experimental exposure scenario and the exposure situation of interest, toxicity data might also be extrapolated for differences in exposure duration or differences in exposure route. For each extrapolation step, assessment factors (AFs) are traditionally used [31]. Because specific information is generally scarce or not available, these AFs are often assigned default values that are conservative values that cover a wide spectrum of chemicals and toxic mechanisms. Since the empirical basis of default AFs is weak, more specific mechanistic information should be incorporated in the extrapolation where possible. Renwick [32] proposed to subdivide the default AFs for interspecies and interindividual differences into separate subfactors for toxicokinetics and toxicodynamics, enabling the incorporation of substance-specific information. Such information can be incorporated via a probabilistic approach [33,34]. Assessment factors for individual extrapolation steps are then substituted with distributions that describe variation between individuals in their toxicokinetics or toxicodynamics. Consequently, extrapolation towards the exposure situation of interest can be done in an informed way, based on a desired level of confidence and a desired level of protection.

3.6 Environmental Impact Evaluation

Figure 3.7 provides an overview of the environmental impact pathway of chemicals that can be used in risk assessment and life cycle assessment.

Figure 3.7 Environmental impact pathway chemicals can follow.

3.6.1 Life Cycle Assessment

Life cycle assessment (LCA) addresses environmental impacts from a product or service perspective. An LCA includes the stepwise calculation of environmental

3.6 Environmental Impact Evaluation

impacts of a product or service during its full life cycle: from resource extraction to waste disposal [35]. The first step is the goal and scope definition that is followed by an inventory phase, a life cycle impact assessment (LCIA) step, and finally an interpretation phase [36]. In the goal and scope definition, the subject and the purpose of an LCA study are determined. Goals of an LCA can be to compare different products or services, or to improve, optimize, or explore the future possibilities of a product or service. During the inventory phase for each of the product systems or services considered, data are gathered for all the relevant processes involved in the life cycle. Regarding chemical pollution, the outcome of the inventory analysis is a list of emissions and their amounts. Emission amounts can be directly obtained from industry or other sources where emissions take place, but there are also examples of emission estimation models. One of them is PestLCI, to estimate pesticide emissions from field application [37], and another example is SimpleTreat [38,39], to estimate emissions from wastewater treatment plants. The removal of chemicals in wastewater treatment plants and remaining emissions to water, air, and sludge can be assessed with the same mass balance approach as used to evaluate the fate of chemicals in the environment (see Section 3.2). In the life cycle impact assessment phase, the inventory data are converted in environmental impact scores with help from so-called characterization factors (CF). Finally, the interpretation phase is to interpret the results from the previous three steps, to draw conclusions, and to provide recommendations for decision makers. To obtain the CF of a chemical, the models as described in this chapter can be applied following the impact pathway outlined in Figure 3.7. The CF accounts for the environmental persistence (fate), exposure, and effect of a chemical on humans or ecosystems CFx;i;j ˆ FFx;i;j
XFx;j
EFx;j FFx,i,j represents the compartment-specific fate factor that accounts for the transport efficiency of substance x from compartment i to, and persistence in, the environment j (FF in year). XFx,j represents the exposure factor that accounts for the exposure of chemical x in compartment i to humans or the ecosystems. EFx,j represents the effect factor of chemical x in compartment j, accounting for ecosystem or human effects. To obtain each subfactor, models are used as described in the corresponding paragraph above. The fate factor is defined as the change in the steady state concentration in an environmental compartment due to a change in emission [40]: FFx;i;j ˆ

V
dC x;j dM x;i

in which V is the volume of environmental compartment j (m3), dCx,j is the change in the steady state dissolved concentration of chemical x in environment j (kg
m 3), and dMx,i is the change in the emission of chemical x to compartment i (kg
yr 1).

59

60

3 Models to Estimate Fate, Exposure, and Effects of Chemicals

The human exposure factor is defined as the change in intake I via route r (e.g., dermal uptake, oral intake, or inhalation) over the entire population N due to a change in environmental concentration [40]: XFx;j;r ˆ

dI x;j;r
N dC x;j

The environmental exposure factor is simply the bioavailable part of the chemical in a compartment, for example, the dissolved fraction in water, leaving out the chemical concentration attached to particles. The ecosystem effect factor is defined as the change in potentially affected fraction of species (PAF) due to a change in concentration in compartment j: E x;eco ˆ

dPAF S PAF ˆ μx dC x 10

where Ex represents the effect factor of substance × (m3
kg 1); the PAF value expresses stress on ecosystems due to the presence of a single chemical or a mixture of chemicals. A PAF reflects the fraction of all species that is expectedly exposed above a certain effect-related benchmark, such as the effect concentration for 50% of the population (EC50) [41]. SPAF is the slope factor of the potentially affected fraction of species. Two main classes of methods are currently identified for the calculation of the slope factor S [42–44]: a) Methods assuming linear concentration–response relationships b) Methods accounting for the nonlinearity in concentration–response relationships. The linear methods do not take into account current background levels, but estimate the change in PAF based on 50% affected species (PAF = 0.5). In the nonlinear methods, S depends on the background situation and behavior of chemicals [44,45]. Huijbregts et al. [46] proposed a third way to estimate PAF, including background situation, but now calculating the average distance between current state and preferred state of the environment. The latter can, for example, be zero effect, but also an environmental target, such as a maximum of 5% species affected. Figure 3.8 shows the three approaches to derive SPAF in an example graph. The PAF is traditionally based on cold-blooded species, as the effects are determined for species that are in direct contact with their surroundings and have a herbivorous diet. Golsteijn et al. [47] showed the possibility of inclusion of ecotoxic impacts on warm-blooded predators in LCIA by introducing a bioaccumulation factor BF that accounts for accumulation in the food chain following a bioaccumulation model as described in Section 3.3. The chemical-specific part of the effect factor equals 1/10μ and reflects the inherent toxicity of a chemical, defined as the inverse of the average toxicity of a chemical that is the concentration of substance x where 50% of the species is exposed above an acute or chronic toxic value (kg
m 3). μx is the average sensitivity of a species community to chemical x (g/l), with sensitivity being expressed as an EC50 or another ecotoxicity test endpoint.

3.6 Environmental Impact Evaluation

Figure 3.8 Example of the derivation of effect factors for a situation with a concentration of 2.5 μg/L, following a linear approach (a), a marginal approach (b), and an average approach (c) taking 5% of potentially affected species as an environmental target.

The human effect factor is defined as the change in probability of occurrence of a disease (R) due to a change in intake of chemical x [48]: E x;human ˆ

dR SH ˆ dI x ED50x

SH is the human health slope factor of the concentration–response curve, and ED50 is the chronic dose affecting 50% of the human population of the chemical added. USES–LCA is one model specifically developed to be able to model fate, exposure, and effects in LCA [49]. The model is based on the (E)USES model family applied for risk assessment purposes in the European Union [50]. It is one of the models involved in the development of LCIA toxicity consensus model USEtox [4], but it also includes a larger number of emission compartments and in addition to freshwater ecotoxicity, it also addresses terrestrial and seawater toxicity. Furthermore, it is able to run dynamic simulations, calculating yearly concentrations, in addition to steady state outcomes. 3.6.2 Risk Assessment

Whereas life cycle assessment focuses on a product or service, risk assessment focuses on the risks of a chemical at a specific location. Both ecological and human health risks due to environmental contaminants can be assessed via a comparison of (local) exposure estimations with a level of exposure at which estimated adverse effects are considered acceptable. In ecological risk assessment, exposure is generally expressed as a predicted environmental concentration (PEC). The ratio between the PEC and the concentration at which adverse chronic effects are predicted not to occur (the predicted no

61

62

3 Models to Estimate Fate, Exposure, and Effects of Chemicals

effect concentration, or PNEC) can be interpreted as a measure of risk, through risk quotient RQ: RQ ˆ

PEC PNEC

An RQ above 1 indicates that potential adverse effects cannot be excluded. Similar to ecological risk assessment, the ratio between the predicted human daily intake (I) and the acceptable daily intake (ADI) can be used as a measure of human risk: RQ ˆ

I ADI

Since risk assessment is generally based on external levels of exposure, part of the environmental impact pathway (Figure 3.7) is generally circumvented, that is, the calculation of internal exposure from external exposure (biokinetics and accumulation). Instead, this step is inherently included in the effect calculations. Moreover, PNECs and ADIs have traditionally been derived from statistically weak no observed adverse effect concentrations (NOAEC) and no observed adverse effect levels (NOAEL), respectively. These represent the lowest chronic test concentration or dose at which no statistical difference in adverse effect is observed compared with the effect observed in the nonexposed control group. Consequently, disadvantages of these measures of effect are their dependence on the experimental setup and the fact that they can only be one of the concentrations or daily doses tested. Moreover, they make only suboptimal use of the information provided by the experimental data. Therefore, species sensitivity distributions (SSDs) (Section 3.5) are nowadays preferred to derive acceptable ecological effect concentrations, and benchmark dose models to derive acceptable human daily intakes (Section 3.7).

3.7 Recent Developments 3.7.1 New Chemicals

Most models on fate and exposure apply to neutral persistent organic chemicals because many substances belong to this class. In addition, processes determining the exchange of these substances are relatively easy to measure and model. Recently, multiple efforts have been done to address other chemical groups. Advances have been made in modeling the partitioning behavior for ionizing chemicals. Franco and Trapp [51,52] derived rules for partitioning of ionizing substances that include pKa and pH dependency. These have been applied in risk assessment and LCA modeling [7,53]. Accumulation of ionic organics has received somewhat less attention. Often uptake is assumed to be dominated by

3.7 Recent Developments

the neutral fraction, estimated as a function of pH, pKa, and pKb [54]. Likewise, substances are assumed to be mainly exchanged by passive permeation through the membrane. However, there is increasing evidence that some chemicals are taken up actively [55]. Although modeling of active uptake covering multiple substances and species is at its infancy, overarching principles appear useful there too. Rendal et al. [56] performed a literature review on toxicity tests of ionizing organic compounds performed at multiple pH levels. They found that the neutral species determines toxicity. As the number of metals is much lower than the number of organic chemicals, empirical data can cover a larger part of the parameter values needed. Despite the relatively low number of metals compared to organic chemicals, theoretical and empirical relationships to chemical properties are still needed. Metal Quantitative Structure-Activity Relations (QSARS) based on, for example, the covalent index, the ionic index, and other atomic characteristics have therefore been derived [57,58]. The environmental fate model SimpleBox was recently adapted to express engineered nanoparticle transport and concentrations in and across air, rain, surface waters, soil, and sediment, accounting for nanospecific processes such as aggregation, attachment, and dissolution [59]. Furthermore, equations describing endocytosis and exocytosis of nanoparticles and microplastics have recently been proposed. For these chemicals, particle size appears to be the most important property, next to hydrophobicity and charge [60]. Pirovano et al. [61] showed the prediction of biotransformation as function of chemical properties. However, since strong interactions between molecules become more important, variability across substances and species increases. A QSAR for biotransformation, therefore, applies to a specific group of chemicals and species only. Consequently, there are only a few models available to estimate biotransformation as a function of chemical properties and biological traits. In fate modeling, it is assumed that when a chemical is degraded it poses no harm to the environment anymore. However, with degradation, transformation products may be formed that can pollute the environment as well. This is particularly relevant when a transformation product is more toxic, more persistent, more mobile, or more bioaccumulative than its parent compound. Van Zelm et al. [62] showed that inclusion of transformation products in the characterization factor of chemicals can lead to substantially larger CFs. 3.7.2 Nontoxic Stressors

Next to chemicals, there are stressors that cause other impacts on the environment, such as global warming and eutrophication. Many chemical transport models have been available to describe fate and exposure of these stressors such as carbon dioxide, methane, and ammonia. The modeling of the effects of these stressors is less straightforward. Recently, the methods from risk assessment have been used to determine effects of other stressors. Multispecies concentration response curves have been derived for a set of nontoxic stressors, ranging

63

64

3 Models to Estimate Fate, Exposure, and Effects of Chemicals

from nitrogen and phosphorus to ground water tables [63,64]. Combining toxic and nontoxic pressures, species sensitivity distributions allow one to estimate the cumulative and relative impact of all stressors. 3.7.3 Uncertainty and Variability

Important in current developments is the assessment of uncertainty. Communication of uncertainties related to model results is vital to increase the value of risk assessment and life cycle assessment outcomes and to give practitioners a good insight in the relevancy of the results. Currently, toxicity is excluded in many LCA studies because a large uncertainty is expected. However, as outlined by Finkel [65], uncertain risk comparisons, despite their complexity, are much preferable in avoiding quantification altogether. Walker et al. [66] define any uncertainty that can be described adequately in statistical terms as statistical uncertainty. The most obvious type of statistical uncertainty is parameter uncertainty, that is, the uncertainty in the estimation of a model parameter of interest due to limited knowledge of its true value. With the exception of universal constants and physical constants that are very well determined, parameter uncertainty influences all model parameters. It is a consequence of the limited data on which parameterization is based and the variation observed within these data. This variation can be due to measurement error or intertest differences, for example, variability in results for the same chemical on the same species in the context of ecotoxicity testing. An established way to account for parameter uncertainty is via Monte Carlo simulations, in which all input parameters are assigned distributions reflecting the limited knowledge of their true values [18,29,34,45,67,68]. An extensive amount of model iterations is then performed (e.g., 10 000), each time with a new set of parameter values drawn from these distributions. The variation in the outcomes of these model runs represents the influence of parameter uncertainty. When uncertainty in an input parameter is quantified with an uncertainty distribution, the parameters of this distribution are often derived from limited data. Consequently, these distribution parameters are themselves subject to uncertainty. This type of parameter uncertainty, so-called secondary parameter uncertainty, can be quantified with sampling distributions [34,69]. Next to parameter uncertainty, other types of uncertainty that are less easily quantified might be of importance, such as context uncertainty (related to the boundaries of the system) or model uncertainty (related to the variables and relationships included in the calculations). Context and model uncertainty can be quantified via discrete choice analysis. In such an analysis, the choices that need to be made are identified. Subsequently, the various options to deal with every choice included are chosen to finally calculate results for each combination of choices [70]. Van Zelm and Huijbregts [71] show that quantifying the tradeoff between parameter and model uncertainty can help to identify optimal model complexity from an uncertainty point of view.

References

Parameters that intrinsically have one true value that is unknown due to limited knowledge can be distinguished from parameters that are intrinsically variable, for example, over time, space, between individuals or between species, but are represented in a model by a single value. Including variability in fate, exposure, and effect modeling is important to show the relevance of determining specific model estimates per region [7,72,73], over time [18,68,73,74], per individual [7,29]. Whereas uncertainty can be reduced by additional research, variability is inherent in the system and cannot be reduced. These two sources of variation, that is, uncertainty and variability, are often considered simultaneously in probabilistic risk assessment. This is mainly because they can be described statistically in a similar way (i.e., using a variance), and because variability can result in uncertainty, depending on the perspective taken [75]. However, for a meaningful probabilistic risk assessment it is important to address uncertainty and variability separately. An analysis of uncertainty can give insight into the likelihood of an adverse event, and as such provide guidance for the allocation of future research efforts to improve model predictions [76,77]. On the other hand, an analysis of the relevant variability can give insight into the magnitude of the adverse event, and as such provide guidance for the allocation of future monitoring efforts, the identification of worst-case exposure scenarios, or the selection of potential risk reduction measures [7,72]. Uncertainty and variability can be separately propagated into the model outcome via the use of nested Monte Carlo simulations [18,29,34,68], a two-step iterative process in which first all uncertain parameters are sampled and fixed, followed by a Monte Carlo simulation on the variable parameters. This process is then repeated a large number of times, each time with a new set of uncertain parameters. This way, the influence of both uncertainty and variability on the model outcome can be assessed. When uncertainty and variability are quantified, better conclusions can be drawn as to what is important, and choices can be made to what to include in each of your models.

References 1 Hendriks, A.J. (2013) How to deal with

4 Rosenbaum, R.K., Bachmann, T.M.,

100,000+substances, sites, and species: overarching principles in environmental risk assessment. Environmental Science and Technology, 47 (8), 3546–3547. 2 Smeets, E. and Weterings, R. (1999) Environmental indicators: Typology and Overview. Technical report 25. European Environment Agency, Copenhagen, Denmark. 3 Mackay, D. and Paterson, S. (1981) Calculating fugacity. Environmental Science and Technology, 15 (9), 1006–1014.

Swirsky Gold, L., Huijbregts, M.A.J., Jolliet, O., Juraske, R., Koehler, A., Larsen, H.F., Macleod, M., Margni, M., McKone, T.E., Payet, J., Schuhmacher, M., Van de Meent, D., and Hauschild, M.Z. (2008) USEtox – The UNEP-SETAC toxicity model: recommended characterisation factors for human toxicity and freshwater ecotoxicity in life cycle impact assessment. The International Journal of Life Cycle Assessment, 13 (7), 532–546. 5 Macleod, M., von Waldow, H., Tay, P., Armitage, J.M., Wöhrnschimmel, H., Riley,

65

66

3 Models to Estimate Fate, Exposure, and Effects of Chemicals

6

7

8

9

10

11

12

13

W.J., McKone, T.E., and Hungerbuhler, K. (2011) BETR global – a geographicallyexplicit global-scale multimedia contaminant fate model. Environmental Pollution, 159 (5), 1442–1445. Hauck, M., Huijbregts, M.A.J., Armitage, J.M., Cousins, I.T., Ragas, A.M.J., and van de Meent, D. (2008) Model and input uncertainty in multi-media fate modeling: benzo a pyrene concentrations in Europe. Chemosphere, 72 (6), 959–967. Oldenkamp, R., Huijbregts, M.A.J., Hollander, A., Versporten, A., Goossens, H., and Ragas, A.M.J. (2013) Spatially explicit prioritization of human antibiotics and antineoplastics in Europe. Environment International, 51, 13–26. Teorell, T. (1937) Kinetics of distribution of substances administered to the body. Archives Internationales de Pharmacodynamie et de Thérapie, 57, 205–240. Thomann, R. and Connolly, J. (1984) Model of PCB in the Lake Michigan lake trout food chain. Environmental Science and Technology, 18, 65–71. Gobas, F., Opperhuizen, A., and Hutzinger, O. (1986) Bioconcentration of hydrophobic chemicals in fish: relationship with membrane permeation. Environmental Toxicology and Chemistry, 5, 637–646. Hendriks, A.J., van der Linde, A., Cornelissen, G., and Sijm, D. (2001) The power of size. 1. Rate constants and equilibrium ratios for accumulation of organic substances related to octanolwater partition ratio and species weight. Environmental Toxicology and Chemistry, 20 (7), 1399–1420. Laane, R., van de Meent, D., de Voogt, P., Parsons, J., Hendriks, A., and Gils, V. (2011) Modeling Organic Compounds in the Estuarine and Coastal Environment. Treatise on Estuarine and Coastal Science, Geochemistry of Estuaries and Coasts, vol. 4, Elsevier, Amsterdam, pp. 161–203. Hendriks, A.J. (1995a) Modeling equilibrium concentrations of microcontaminants in organisms of the rhine delta – can average field residues in the aquatic foodchain be predicted from laboratory accumulation. Aquatic Toxicology, 31 (1), 1–25.

14 Hendriks, A.J. (1995b) Modeling

15

16

17

18

19

20

21

nonequilibrium concentrations of microcontaminants in organisms – comparative kinetics as a function of species size and octanol water partitioning. Chemosphere, 30 (2), 265–292. Hendriks, A.J., Ma, W.C., Brouns, J.J., Deruiterdijkman, E.M., and Gast, R. (1995) Modeling and monitoring organochlorine and heavy-metal accumulation in soils, earthworms, and shrews in rhine-delta floodplains. Archives of Environment Contamination and Toxicology, 29 (1), 115–127. Hendriks, A.J., Traas, T.P., and Huijbregts, M.A.J. (2005) Critical body residues linked to octanol-water partitioning, organism composition, and LC50 QSARs: metaanalysis and model. Environmental Science and Technology, 39 (9), 3226–3236. Stadnicka, J., Schirmer, K., and Ashauer, R. (2012) Predicting concentrations of organic chemicals in fish by using toxicokinetic models. Environmental Science and Technology, 46, 3273–3280. Huizer, D., Oldenkamp, R., Ragas, A.M.J., van Rooij, J.G.M., and Huijbregts, M.A.J. (2012) Separating uncertainty and physiological variability in human PBPK modelling: the example of 2-propanol and its metabolite acetone. Toxicology Letters, 214 (2), 154–165. Le, T.T.Y., Swartjes, F., Romkens, P., Groenenberg, J.E., Wang, P., Lofts, S., and Hendriks, A.J. (2015) Modelling metal accumulation using humic acid as a surrogate for plant roots. Chemosphere, 124, 61–69. Hendriks, A.J., Awkerman, J.A., de Zwart, D., and Huijbregts, M.A.J. (2013) Sensitivity of species to chemicals: doseresponse characteristics for various test types (LC50, LR50 and LD50) and modes of action. Ecotoxicology and Environment Safety, 97, 10–16. de Hoop, L., Schipper, A.M., Leuven, R.S.E.W., Huijbregts, M.A.J., Olsen, G.H., Smit, M.G.D., and Hendriks, A.J. (2011) Sensitivity of polar and temperate marine organisms to oil components. Environmental Science and Technology, 45 (20), 9017–9023.

References 22 Golsteijn, L., van Zelm, R., Hendriks, A.J.,

23

24

25

26

27

28

29

30

31

and Huijbregts, M.A.J. (2013) Statistical uncertainty in hazardous terrestrial concentrations estimated with aquatic ecotoxicity data. Chemosphere, 93 (2), 366–372. Fedorenkova, A., Vonk, J.A., Breure, A.M., Hendriks, A.J., and Leuven, R.S.E.W. (2013) Tolerance of native and non-native fish species to chemical stress: a case study for the river Rhine. Aquatic Invasions, 8 (2), 231–241. Posthuma, L., Suter, G.W., II, and Traas, T.P. (2002) Species Sensitivity Distributions in Ecotoxicology, Lewis Publishers, Boca Raton (FL), USA. Fu, W., Franco, A., and Trapp, S. (2009) Methods for estimating the bioconcentration factor of ionizable organic chemicals. Environmental Toxicology and Chemistry, 28 (7), 1372–1379. Trapp, S. (2004) Plant uptake and transport models for neutral and ionic chemicals. Environmental Science and Pollution Research, 11 (1), 33–39. Takaki, K., Wade, A.J., and Collins, C.D. (2015) Assessment and improvement of biotransfer models to cow’s milk and beef used in exposure assessment tools for organic pollutants. Chemosphere, 138, 390–397. Hendriks, A.J., Smitkova, H., and Huijbregts, M.A.J. (2007) A new twist on an old regression: transfer of chemicals to beef and milk in human and ecological risk assessment. Chemosphere, 70 (1), 46–56. Ragas, A.M.J., Brouwer, F.P.E., Buchner, F.L., Hendriks, H.W.M., and Huijbregts, M.A.J. (2009) Separation of uncertainty and interindividual variability in human exposure modeling. Journal of Exposure Science and Environmental Epidemiology, 19 (2), 201–212. Ragas, A.M.J. and Huijbregts, M.A.J. (1998) Evaluating the coherence between environmental quality objectives and the acceptable or tolerable daily intake. Regulatory Toxicology and Pharmacology, 27 (3), 251–264. Vermeire, T., Stevenson, H., Pieters, M.N., Rennen, M., Slob, W., and Hakkert, B.C. (1999) Assessment factors for human

32

33

34

35

36

37

38

39

40

health risk assessment: a discussion paper. Critical Reviews in Toxicology, 29 (5), 439–490. Renwick, A.G. (1993) Data-derived safety factors for the evaluation of food-additives and environmental contaminants. Food Additives and Contaminants, 10 (3), 275–305. WHO (2014) Guidance Document on Evaluating and Expressing Uncertainty in Hazard Characterization. vol IPCS Harmonization Project No. 11. Oldenkamp, R., Huijbregts, M.A.J., and Ragas, A.M.J. (2016) Uncertainty and variability in human exposure limits – a chemical-specific approach for ciprofloxacin and methotrexate. Critical Reviews in Toxicology, 46 (3), 261–278. Consoli, F., Allen, D., Boustead, I., Fava, J., Franklin, W., Jensen, AA., De Oude, N., Parrish, R., Perriman, R., Postlethwaite, D., Quay, B., Séquin, J., and Vigon, B. (1993) Guidelines for Life-Cycle Assessment: a Code of Practice. Society of Environmental Toxicology and Chemistry, Pensacola (FL). ISO 14040 (2006). Environmental Management – Life Cycle Assessment – Principles and Framework. Technical Standard. Geneva, Switzerland. Dijkman, T.J., Birkved, M., and Hauschild, M.Z. (2012) PestLCI 2.0: a second generation model for estimating emissions of pesticides from arable land in LCA. International Journal of Life Cycle Assessment, 17 (8), 973–986. Struijs, J., van de Meent, D., Schowanek, D., Buchholz, H., Patoux, R., Wolf, T., Austin, T., Tolls, J., van Leeuwen, K., Galay-Burgos, M., (2016). Adapting Simple-Treat for simulating behaviour of chemical substances during industrial sewage treatment. Chemosphere 159, 619–627. Lautz L, Struijs J, Nolte TM, Breure AM, Van der Grinten E, Van de Meent D, Van Zelm R. (2017). Evaluation of SimpleTreat 4.0: Simulations of pharmaceutical removal in wastewater treatment plant facilities. Chemosphere 168, 870–876. Huijbregts, M.A.J., Struijs, J., Goedkoop, M., Heijungs, R., Hendriks, A.J., and Van de Meent, D. (2005) Human population intake fractions and environmental fate

67

68

3 Models to Estimate Fate, Exposure, and Effects of Chemicals

41

42

43

44

45

46

47

48

factors of toxic pollutants in life cycle impact assessment. Chemosphere, 61 (10), 1495–1504. De Zwart, D. and Posthuma, L. (2005) Complex mixture toxicity for single and multiple species: proposed methodologies. Environmental Toxicology and Chemistry, 24 (10), 2665–2676. Larsen, H.F. and Hauschild, M. (2007) Evaluation of ecotoxicity effect indicators for use in LCIA. International Journal of Life Cycle Assessment, 12 (1), 24–33. Pennington, D.W., Payet, J., and Hauschild, M. (2004) Aquatic ecotoxicological indicators in life-cycle assessment. Environmental Toxicology and Chemistry, 23, 1796–1807. Van de Meent, D. and Huijbregts, M.A.J. (2005) Calculating life-cycle assessment effect factors from potentially affected fraction-based ecotoxicological response functions. Environmental Toxicology and Chemistry, 24 (6), 1573–1578. Van Zelm, R., Huijbregts, M.A.J., Harbers, J.V., Wintersen, A., Struijs, J., Posthuma, L., and Van de Meent, D. (2007) Uncertainty in msPAF-based ecotoxicological effect factors for freshwater ecosystems in life cycle impact assessment. Integrated Environmental Assessment and Management, 3 (2), 203–210. Huijbregts, M.A.J., Hellweg, S., and Hertwich, E.G. (2011) Do we need a paradigm shift in life cycle impact assessment? Environmental Science and Technology, 45, 3833–3834. Golsteijn, L., Van Zelm, R., Veltman, K., Musters, G., Hendriks, A.J., and Huijbregts, M.A.J. (2012) Including ecotoxic impacts on warm-blooded predators in life cycle impact assessment. Integrated Environmental Assessment and Managment, 8 (2), 372–378. Huijbregts, M.A.J., Rombouts, L.J.A., Ragas, A.M.J., and Van de Meent, D. (2005) Human-toxicological effect and damage factors for life-cycle impact assessment of carcinogenic and noncarcinogenic chemicals. Integrated Environmental Assessment and Management, 3 (1), 181–244.

49 Van Zelm, R., Huijbregts, M.A.J., and Van

50

51

52

53

54

55

56

57

de Meent, D. (2009) USES-LCA 2.0: a global nested multi-media fate, exposure and effects model. International Journal of Life Cycle Assessment, 14, 282–284. Vermeire, T., Rikken, M., Attias, L., Boccardi, P., Boeije, G., Brooke, D., De Bruijn, J., Comber, M., Dolan, B., Fischer, S., Heinemeyer, G., Koch, V., Lijzen, J., Müller, B., Murray-Smith, R., and Tadeo, J. (2005) European union system for the evaluation of substances: the second version. Chemosphere, 59, 473–485. Franco, A. and Trapp, S. (2008) Estimation of the soil-water partition coefficient normalized to organic carbon for ionizable organic chemicals. Environmental Toxicology and Chemistry, 27, 1995–2004. Franco, A. and Trapp, S. (2010) A multimedia activity model for ionizable compounds: validation study with 2,4dichlorophenoxyacetic acid, aniline, and trimethoprim. Environmental Toxicology and Chemistry, 29, 789–799. Van Zelm, R., Stam, G., Huijbregts, M.A.J., and Van de Meent, D. (2013) Making fate and exposure models for freshwater ecotoxicity in life cycle assessment suitable for organic acids and bases. Chemosphere, 90, 312–317. Birak, P., Yurk, J., Adeshina, F., Lorber, M., Pollard, K., Choudhury, H., and Kroner, S. (2001) Travis and Arms revisited: a second look at a widely used bioconcentration algorithm. Toxicology and Industrial Health, 17 (5–10), 163–175. O’Connor, I.A., Veltman, K., Huijbregts, M.A.J., Ragas, A.M.J., Russel, F.G.M., and Hendriks, A.J. (2014) Including carriermediated transport in oral uptake prediction of nutrients and pharmaceuticals in humans. Environmental Toxicology and Pharmacology, 38 (3), 938–947. Rendal, C., Kusk, K.O., and Trapp, S. (2011) Optimal choice of pH for toxicity and bioaccumulation studies of ionizing organic chemicals. Environmental Toxicology and Chemistry, 30 (11), 2395–2406. Le, T.T.Y. and Hendriks, A.J. (2013) Relationships between absorption

References

58

59

60

61

62

63

64

65

efficiency of elements in mammals and chemical properties. Critical Reviews in Toxicology, 43 (9), 800–809. Owsianiak, M., Veltman, K., Hauschild, M.Z., Hendriks, A.J., Steinmann, Z.J.N., and Huijbregts, M.A.J. (2014) Elucidating differences in metal absorption efficiencies between terrestrial soft-bodied and aquatic species. Chemosphere, 112, 487–495. Meesters, J.A.J., Koelmans, A.A., Quik, J.T.K., Hendriks, A.J., and van de Meentt, D. (2014) Multimedia modeling of engineered nanoparticles with simpleBox4nano: model definition and evaluation. Environmental Science and Technology, 48 (10), 5726–5736. Kettler, K., Veltman, K., van de Meent, D., van Wezel, A., and Hendriks, A.J. (2014) Cellular uptake of nanoparticles as determined by particle properties, experimental conditions, and cell type. Environmental Toxicology and Chemistry, 33 (3), 481–492. Pirovano, A., Brandmaier, S., Huijbregts, M.A.J., Ragas, A.M.J., Veltman, K., and Hendriks, A.J. (2015) The utilisation of structural descriptors to predict metabolic constants of xenobiotics in mammals. Environmental Toxicology and Pharmacology, 39 (1), 247–258. Van Zelm, R., Huijbregts, M.A.J., and Van de Meent, D. (2010) Transformation products in the life cycle impact assessment of pesticides. Environmental Science and Technology, 44, 1004–1009. Azevedo, LB., van Zelm, R., Leuven, R.S.E.W., Hendriks, AJ., and Huijbregts, M.A.J. (2015) Combined ecological risks of nitrogen and phosphorus in European freshwaters. Environmental Pollution, 200, 85–92. Van Zelm, R., Schipper, A.M., Rombouts, M., Snepvangers, J., and Huijbregts, M.A.J. (2011) Implementing groundwater extraction in life cycle impact assessment: characterization factors based on plant species richness for The Netherlands. Environmental Science and Technology, 45 (2), 629–635. Finkel, A. (1995) Toward less misleading comparisons of uncertain risks: the example of aflatoxin and alar.

66

67

68

69

70

71

72

73

Environmental Health Perspectives, 103 (4), 376–385. Walker, W., Harremoës, P., Rotmans, J., Van der Sluijs, J.P., Van Asselt, M., Janssen, P., and Krayer von Krauss, M. (2003) Defining uncertainty: a conceptual basis for uncertainty management in model-based decision support. Integrated Assessment, 4 (1), 5–17. Hauck, M., Hendriks, H.W.M., Huijbregts, M.A.J., Ragas, A.M.J., van de Meent, D., and Hendriks, A.J. (2011) Parameter uncertainty in modeling bioaccumulation factors of fish. Environmental Toxicology and Chemistry, 30 (2), 403–412. Huizer, D., Ragas, A.M.J., Oldenkamp, R., van Rooij, J.G.M., and Huijbregts, M.A.J. (2014) Uncertainty and variability in the exposure reconstruction of chemical incidents – the case of acrylonitrile. Toxicology Letters, 231 (3), 337–343. Oldenkamp, R., Hendriks, H.W.M., van de Meent, D., and Ragas, A.M.J. (2015) Hierarchical bayesian approach to reduce uncertainty in the aquatic effect assessment of realistic chemical mixtures. Environmental Science and Technology, 49 (17), 10457–10465. De Schryver, A.M., Van Zelm, R., Humbert, S., Pfister, S., McKone, T.E., and Huijbregts, M.A.J. (2011) Value choices in life cycle impact assessment of stressors causing human health damage. Journal of Industrial Ecology, 15 (5), 796–815. Van Zelm, R. and Huijbregts, M.A.J. (2013) Quantifying the trade-off between parameter and model structure uncertainty in life cycle impact assessment. Environmental Science and Technology, 47 (16), 9274–9280. Oldenkamp, R., Huijbregts, M.A.J., Hollander, A., and Ragas, A.M.J. (2014) Environmental impact assessment of pharmaceutical prescriptions: does location matter? Chemosphere, 115, 88–94. Hauck, M., Huijbregts, M.A.J., Hollander, A., Hendriks, A.J., and van de Meent, D. (2010) Modeled and monitored variation in space and time of PCB-153 concentrations in air, sediment, soil and aquatic biota on a

69

70

3 Models to Estimate Fate, Exposure, and Effects of Chemicals

European scale. Science of the Total Environment, 408 (18), 3831–3839. 74 De Hoop, L., Viaene, K.P.J., Schipper, A.M., Huijbregts, M.A.J., De Laender, F., and Hendriks, A.J. (2017) Time-varying effects of aromatic oil constituents on the survival of aquatic species: Deviations between model estimates and observations. Environmental Toxicology and Chemistry 36(1): 128–136. 75 Ragas, A.M.J., Huijbregts, M.A.J., Henning-de Jong, I., and Leuven, R.S.E.W. (2009) Uncertainty in environmental risk assessment: implications for risk-based

management of river basins. Integrated Environmental Assessment and Management, 5 (1), 27–37. 76 Hertwich, E.G., McKone, T.E., and Pease, W.S. (1999) Parameter uncertainty and variability in evaluative fate and exposure models. Risk Analysis, 19 (6), 1193–1204. 77 Oldenkamp, R., Huijbregts, M., and Ragas, A. (2016) The influence of uncertainty and location-specific conditions on the environmental prioritisation of human pharmaceuticals in Europe. Environment International, 91, 301–311.

71

4 Collaborative Approaches to Advance Chemical Safety Philip Judson

4.1 Introduction

Many activities that contribute to the advancement of chemical safety have collaborative components to them. Perhaps the most long-standing is the sharing of scientific findings through learned societies – in the case of chemical safety, for example, 1) through the journals and conferences of the Society of Toxicology (SOT) and the 2) Society of Environmental Toxicology and Chemistry (SETAC) . By requiring the submission of results from chemical safety studies and then using them to assess terms for registration of new products, government regulators impose a kind of asymmetrical collaboration between themselves and the submitting companies. In addition, within legal constraints, regulators may use the information they acquire to develop and disseminate generic principles or guidelines. For example, teams at the United States Food and Drug Administration (FDA) and Environmental Protection Agency (EPA) supported the development of computer programs to predict 3) 4) toxicity [1] , ; a major effort, Tox21, is in progress to develop new assays and meth5) ods in toxicity testing ; the Danish Environmental Protection Agency offers a database of predictions from Quantitative Structure–Activity Relationship (QSAR) 6) models for about 170 000 chemicals . An early example of a successful data sharing project, outside the field of 7) chemical safety evaluation, is the Cambridge Crystallographic Data Centre [2]

1) The Society of Toxicology, 1821 Michael Faraday Drive, Suite 300, Reston, VA 20190, USA www .toxicology.org/(Sept. 28, 2015). 2) The Society of Environmental Chemistry and Toxicology, Av. de la Toison d’Or, 67 b6, B-1060 Brussels, Belgium, and 229 South Baylen Street, 2nd Floor, Pensacola, FL 32502, USA www.setac. org/(Sept. 28, 2015). 3) http://www.epa.gov/oppt/exposure/pubs/episuitedl.htm. 4) http://www.epa.gov/oppt/ar/2007-2009/cross_cutting/ces.htm. 5) ntp.niehs.nih.gov/(Sept. 28, 2015). 6) qsar.food.dtu.dk/(Sept. 28, 2015). 7) www.ccdc.cam.ac.uk/(Sept. 17, 2015). Handbook of Green Chemistry Volume 10: Tools for Green Chemistry, First Edition. Edited by Evan S. Beach and Soumen Kundu.  2017 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2017 by Wiley-VCH Verlag GmbH & Co. KGaA.

72

4 Collaborative Approaches to Advance Chemical Safety

founded in 1965 and now regarded as the de facto primary repository for published crystallographic data. However, the aim of most of these operations is to release information into the public domain. This chapter is about collaborations run for the collective benefit of the participants in them. Most of them deliver public benefits and some place data or knowledge into the public domain, but publication is not their main purpose.

4.2 Incentives for Collaboration and Constraints

As individuals, humans maintain a balance between pure self-interest and collaboration for the good of the community. The two are not in conflict, or even distinct, in that what benefits a community often benefits the individuals in it and vice versa, though not always. This balance applies equally to collaborations between groups, be they teams, societies, companies, or nations. What are the factors that determine whether competition or collaboration might be preferred in the context of chemical safety research? An obvious driver for collaboration in chemical safety research is the saving of money. Chemical safety studies, especially animal studies, are expensive [3]. If several organizations need to determine the hazards associated with the same chemical – and hence how it can be used safely – there are obvious savings in doing the required studies once rather than duplicating them. A second consideration is the saving of time. If many studies are needed, individual organizations may have to schedule them in sequence because of limited resources. By agreeing a distribution of work, the time taken to completion can be reduced allowing products to reach the market sooner. This achieves faster payback on investments and increases the length of time during which a company can benefit from patent protection. Collaboration can also give participants access to a greater body of knowledge, providing opportunities to predict hazards or assess risks for a chemical by analogy with other similar ones. Even predictions that are not sufficiently conclusive to obviate the need for laboratory studies can highlight priorities, so that the most critical tests are done first and time is not wasted on others. Scientists, and the companies they work for, want to minimize the number of animal experiments they carry out for ethical reasons. There are external pressures as well. Companies care about their public image and there is widespread public concern about the use of animals in experiments. Legislators, and hence regulators, impose tight controls on animal experimentation and in some cases prohibit it altogether. In Europe, legislation requiring all marketed chemicals eventually to be tested for safety to human and environmental health (REACH – registration, evaluation, authorization, and restriction of chemicals) is in conflict with legislation prohibiting the testing of cosmetics on animals [4–6]. Quite apart from the issue with cosmetics, anecdotal estimates suggest that it could

4.2 Incentives for Collaboration and Constraints

take one hundred years or more to complete animal tests on all the chemicals in current use. United Nations and OECD international guidelines encourage a staged approach to chemical safety assessment in which animal experiments are the last resort [7,8]. There are thus powerful incentives for organizations to share information in order to use it to develop trustworthy prediction methods, based on either laboratory measurements not requiring the use of animals or computer models. The arguments in favor of collaboration are so convincing that it may seem surprising that collaboration is not automatic. However, there are important barriers – some perhaps perceived rather than absolute but arising from interpretations of international law. An important one of these is the way that patent law operates. In order to be able to register a patent, the claims to be patented must be novel and not previously published – even by the applicant. Companies worry that if they share any confidential information they might unwittingly weaken patent protection for their inventions. It is possible they worry too much but, because of the way patent laws are framed and the obligations on patent examiners in many places to challenge applications, the risks are real and the consequences potentially disastrous. Even if companies sign a secrecy agreement, there is a risk that critical information such as a chemical structure not covered under the agreement might slip through by mistake. Added to that, it is one thing to share general information about chemical toxicity with a competitor: it is quite another to tell your competitors what structures are currently hot topics in your research department. Given the position of chemical safety testing in the stages of pharmaceutical product development, revealing that research into the safety of a particular chemical is of pressing importance tells your competitors all they need to know. An added legal complication is the operation of antitrust laws in the United States and their equivalents in Europe and elsewhere. It is illegal for companies to form cartels in order to control markets. Legal advisers worry about the unclear distinction between collaborating to gain unfair competitive advantage and collaborating for more honorable reasons. Companies and their advisers apply a precautionary principle to legal issues – if in doubt, do not do it. Setting aside concerns about revealing too much, there are issues with helping competitors. If a company believes it has competitive advantage because it holds unique information or ideas (and all companies do believe it), why would it forfeit the advantage? It is not just a matter of keeping the critical information about new product lines secret – such as the chemical structure of your most promising lead and its expected commercial applications. If you have, say, a novel way of predicting chemical toxicity that avoids costly studies or lengthy delays, you will want to keep that to yourself. It is perhaps less ethically sound, but nevertheless makes business sense, that even if you have completed a costly animal experiment that rules out further development of what seemed a promising lead, it is in your interests to let your competitor spend just as much making the same findings in the same way at the same cost. Finally, if companies set up collaborations, how can they be sure that all parties will contribute, and benefit, fairly?

73

74

4 Collaborative Approaches to Advance Chemical Safety

4.3 Options for Sharing

The issues mentioned in Section 4.2 make some kinds of collaboration easier than others. Broadly, the areas in which collaboration seems to have been successful are as follows. 4.3.1 Sharing Research

The most basic form of collaboration to mutual advantage is the sponsorship of research in an academic organization by an industrial company. Typically, the sponsor supports one or more researchers to work on an agreed topic. The sponsor has rights, usually exclusive at least for a period of time, to commercial exploitation of results; the academic researchers get funding to pursue work that interests them and they are able to make ongoing academic use of their findings and to publish their work, allowing prior filing of patents where the sponsor deems it necessary. Sometimes the sponsor wants something specific done, such as creation of a particular software application, while the academic group are interested in working on the means to do it. Hence, both parties can achieve their aims with little long-term conflict of interest. From the sponsor’s point of view, one aspect of collaboration with an academic group can be an advantage or disadvantage, depending on circumstance. Academic research is, or should be, driven by the pursuit of unanticipated, significant findings during the course of work. As seen from a business planning point of view, this amounts to “project creep”: Instead of moving steadfastly to a preordained goal, researchers are likely to go off in an entirely new direction with unpredictable consequences. The negative side of this is that the sponsor does not get the deliverable that had been intended. The positive side is that, having exploitation rights, the sponsor may gain commercial control of something new, exciting, and lucrative. So, as a generalization, industrial/academic collaboration works best if either the academic interest is truly academic and the sponsor’s interest is in the creation of something less academically interesting but essential to the academic work, or if the sponsor is happy to invest speculatively in a high risk, high gain venture. This second option may not be as unproductive as it appears, even if nothing exciting comes out of it: There may be public relation benefits in sponsoring academia; sponsoring Ph. D. or postdoctoral researchers may be a way of identifying new staff and raising their interest in joining the company and working with the academic researchers and their supervisors helps to keep company staff up to date with scientific developments. It is a small conceptual step from collaboration between one commercial and one academic partner to collaborations between groups of organizations. A huge number of such groups have been set up – involving government, industrial, not-for-profit, and academic organizations as participants. Some collaborations

4.3 Options for Sharing

have included very large numbers of participants. For example, there are over 70 participating organizations in the SEURAT-1 project (see Section 4.5.13). An arrangement that is neither intended to be nor normally perceived as collaboration, but nevertheless has an element of sharing, is the use of contract research organizations (CRO) by pharmaceutical companies. The overhead costs of developing and running laboratory tests are shared between the clients of a CRO, of course, but in addition there are probably many cases where different clients submit closely related compounds for testing. CROs maintain strict secrecy and clients will not know about their shared interests but they will nevertheless benefit from the increased understanding of the field in question by staff at the CRO. A special case is collaboration between companies in REACH Substance Information Exchange Forums (SIEF), where collaborating companies may be aware of their common interest in a chemical and obliged to share information or to fund safety testing jointly (see Section 4.5.12). 4.3.2 Sharing Knowledge

There is a long tradition of competing, commercial organizations attending or organizing conferences to exchange knowledge and ideas that are not considered commercially sensitive. Industrial associations do not exist only to promote the interests of their members by representing them in discussions with governments and the media; they also facilitate appropriate sharing of ideas. Some of this collaboration is motivated by public relation considerations or the need to maintain the enthusiasm of researchers who, as scientists, want to share their discoveries, but there can be substantial benefits both in terms of cost saving and faster scientific progress. A body of specialized knowledge about chemical toxicity builds up in a research organization. One company might be interested in one group of compounds and conducting toxicity studies; another might have tested structurally related compounds. It might appear pointless, except as an act of altruism, for either company to disclose its knowledge to the other. But if, say, ten companies agree to share knowledge, for every piece of knowledge one company gives away it receives nine pieces in return. The critical question is whether such knowledge can be shared without giving away valuable secrets. In practice, generic knowledge can. It would not give away secrets about new product lines, for example, if a cosmetics company disclosed its findings on the skin sensitization potential of a functional group that had no relevance to useful cosmetic properties of compounds. It is not unusual for a company to have suspicions about the link between a particular chemical substructure and a toxicological problem, but insufficient data to confirm it or to discover what other factors determine whether toxicity is or is not manifested. Other companies may have similar theories. By sharing the theories and each testing them on their own data, rather than sharing the data, they can improve their collective understanding.

75

76

4 Collaborative Approaches to Advance Chemical Safety

For example, if a company were interested in Structure 1 and related haloindoles, they knew that enones (Structure 2) were associated with skin sensitization, and their studies suggested new knowledge about the effects of an aromatic ring on activity – perhaps even including the effects of electron donating or withdrawing ring substituents – they could share that knowledge and associate it with Structure 3 without revealing their interest in indoles. Other companies might have entirely different compounds containing Structure 3 as a substructure and be able to add to, or refute, the putative knowledge.

If data can be revealed to an appropriate “honest broker” – an organization that is not in competition with the other parties in a collaboration, and not in special relationships with any of them or third parties – the honest broker may be able to develop a theory linking chemical substructure and/or physical properties to a toxicological effect that would not have been possible with the small amount of data held by each individual party. Companies may carry out extensive literature searches in order to develop models to predict a particular kind of toxicity. Being based on published information, the models are often not commercially sensitive. Duplication can be avoided and the refinement of models speeded up, if a group of companies can agree that each will concentrate on one subset of the models to be developed. Or they may decide collectively to fund modeling by a contractor or honest broker on their behalf. Many of the examples in Section 4.5 involve knowledge sharing of the kinds mentioned here. 4.3.3 Sharing Data

The sharing of toxicological data is a particularly difficult area because of the high sensitivity of the information. Sharing knowledge derived from human interpretation of the results from studies, as described in Section 3.2, can be done without revealing details about particular products or candidate products. Sharing data involves the release of a great deal of detail. For the development of toxicity prediction models, chemical structures need to be included in the data. Various ideas have been floated for generating sets of

4.3 Options for Sharing

descriptors suitable for modeling and delivering those instead of the structures themselves – for example, sets of unconnected substructural fragments. However, there are two problems with this approach. The first is that, logically, if the underlying structures cannot be deduced from the contents of a set of descriptors, information must have been lost; modeling work is hampered from the start [9]. The second is that when experiments have been done with descriptor sets that were thought to be secure, it has been shown to be possible to reverse engineer them and discover structures that they represent. Given the importance attached by legal advisers to chemical structures, the safe option is to share information based only on structures that do not need to remain secret. That being the case, the structures themselves might as well be exchanged rather than only descriptors derived from them. There are structures that do not need to remain secret – the structures of pharmaceutical products that go onto the market are published, for example – but public relations and legal liability issues worry companies as well. Will members of the public take exception to a company’s involvement in some kinds of toxicology experiments? Will others misinterpret data and raise false alarms about the potential dangers of a product? If a product shows unexpected, dangerous side effects, might scrutiny of toxicological data provide opportunities for litigation on the grounds that a company could have anticipated the problems? However, companies do toxicological studies on structures that are not subject to most of these concerns. Many companies use the same production intermediates and reagents, and may even buy and sell them among themselves. They need to know the toxicological properties of these materials to meet regulations concerning contaminants in pharmaceutical products and to support safe working practices. Companies use many of the same materials in the formulation of candidate active ingredients for early screening and testing – solvents and wetting agents, for example – and they are no secret. It makes little economic sense for several companies to carry out expensive studies on the same intermediates and formulants, and sharing data on them (as distinct from publishing data) is now established practice (see Section 4.5.7). In response to pressure from the public and from regulatory bodies, pharmaceutical companies are starting to share data from clinical trials and allowing limited publication of some of the information (see Section 4.5.19). In Europe, as mentioned earlier, the wider chemical industry is sharing data and/or the costs of generating it to meet the requirements of REACH (see Section 4.5.12). 4.3.4 Sharing Software Development

Most of the large pharmaceutical companies set up in-house software teams, in the early days of the development of computer methods, to support research activities. They recognized the competitive advantages of being at the forefront in this new field. Few companies, if any, would think of writing their own word

77

78

4 Collaborative Approaches to Advance Chemical Safety

processing packages, for example, but activities such as molecular modeling were so new and specialized that there were no off-the-shelf products. As the technology matured, the advantages of developing complete computer applications inhouse became less clear. Independent software vendors started to offer research applications that matched or out-performed in-house applications, and as software engineering became more advanced, it was difficult for companies whose business was not software development to keep pace. It became more cost-effective to buy in solutions and for in-house teams to specialize in the customization and use of those solutions. Most, perhaps all, research-based chemical companies now license software from third parties to support chemical safety evaluation, including database management software and statistical analysis tools to handle laboratory data and toxicity prediction applications to inform teams right across research departments. However, chemical companies of all kinds still need to develop, or to support the development of, applications covering different areas of toxicology and chemistry, or making use of new technology coming from artificial intelligence research. Sponsoring academic groups to develop software is an effective way of making progress in new areas. It is less satisfactory for the creation and long-term maintenance of a new product. Academic groups exist to do research, not to provide commercial services. It is usually more effective for academic groups to create first prototypes attacking new problems or using new methods, with development into commercial products taken on by suitable software vendors (which might be existing ones, or start-up businesses). Software vendors continue to make advances for themselves but there is a shortage of data from laboratory studies to use in their research. Some collaborations allow vendors access to inhouse knowledge and data at sponsoring companies but there are concerns about releasing confidential information to an independent commercial organization and fears that the software vendor will be the main beneficiary rather than the supporting companies. An alternative is for the software company to be owned by the sponsors, and this is discussed in Section 4.4.

4.4 The Implementation of Collaborative Organizations

The costs of one-off tasks needing to be done by several companies, such as conducting laboratory tests on a particular compound to meet European REACH requirements, can be shared simply by jointly funding the work (see Section 4.5.12). These collaborations are closed, in the sense that they are between a limited group of companies, but the results of the studies, being submitted to the European authorities, become public. So for this kind of collaboration, all that is needed is an agreement on how costs will be shared – there are no issues over how results will be exploited, for example. As mentioned earlier, companies are also, in a sense, unwitting collaborators when they outsource laboratory work to contract research organizations (CRO): The CROs make

4.4 The Implementation of Collaborative Organizations

efficiency gains from overlap between projects even though the overlap remains confidential. In the case of REACH, the regulatory authority effectively imposes collaboration on groups of companies and it is not itself directly involved in the collaborations. In other cases, regulatory authorities are the instigators of and central participants in voluntary collaborations. These may be with just one or two commercial organizations that contract to create applications for the regulator, and thus qualify to be described as collaborative. Early examples were the creation of Oncologic for the prediction of chemical carcinogenesis and the Epiwin suite of programs for predicting the behavior of chemicals in the environment – both developed for the US Environmental Protection Agency [1]. A problem with these kinds of projects is that they are normally funded only for a limited period, sufficient to create the required application which goes into the public domain. Thereafter, there is no mechanism for generating income to support ongoing maintenance and development. In cases where the commercial partners have rights to exploit the results of the work, this might be expected to ensure subsequent support, but companies often decide that continuation is not profitable if the main deliverable has gone into the public domain. Alternatively, a government department may use a product from a commercial supplier (as distinct from commissioning the creation of one), through which the results of work in the department are made available to the public (see Section 4.5.17), with terms in the agreement to prevent unreasonable exploitation. The terms of agreement have to be carefully drawn up so that, on the one hand, the commercial partner does not gain unfair competitive advantage through government sponsorship while, on the other hand, the commercial partner does not end up out of pocket. There have been cases where the former has been alleged or suspected but, because inappropriate commercial sponsorship is such a serious issue for government departments and very much in the minds of staff drawing up agreements, cases where the commercial partner loses out are probably more common. By far the biggest area of collaboration involving commerce, academia, and government bodies is government sponsorship of precompetitive research. Examples in Europe are the Innovative Medicines Initiative (See Section 4.5.5) and SEURAT-1 (see Section 4.5.13). Note that the projects are precompetitive: Commercial deliverables are not expected from them and only appear if participants carry results forward into product development. In practice, this is not often the case. It probably only happens when a participant enters into a collaboration in good faith to contribute to it but with a clear and very concrete (and probably confidential) idea of what the participant intends to get out of the project for exploitation. More generally, the benefits of the collaborations are the wider advancement of science and less tangible spin-off such as the building of links between individuals, organizations, and communities that can lead to productive, new partnerships. As already mentioned, the above approaches are not well suited to provide the long-term support and development that is necessary for computer software to

79

80

4 Collaborative Approaches to Advance Chemical Safety

predict safety hazards for new compounds. Nothing is free – funding for projects has to come from somewhere. So funding needs to be secured while still making the results of collaboration accessible on fair terms. Although software could be delivered by jointly owned, profit-making companies, this is relatively unusual and perhaps outside the remit of this chapter. In practice, many of the collaborations listed in Section 4.5, some involving large numbers of participants, are based on not-for-profit membership organizations. In England, the most common form for a not-for-profit organization to take is as a company limited by guarantee. Many registered charities are companies limited by guarantee. While a profit-making company has shareholders who provide the capital for the company to operate and, in return, receive dividends, a company limited by guarantee has members who provide no capital. Instead they undertake to pay some fixed amount to meet the debts of the company in the event that it were to go into liquidation. They own and control the company but they receive no dividends. Depending on the constitution of the company, they may be entitled to a proportion of its assets if it is wound up, but more usually (and always in the case of a charity) the assets have to be transferred to another, similar, not-for-profit organization. Sponsors who contribute knowledge or data to a collaborative project need to be satisfied that the terms of membership are fair to them. Having in mind antitrust and similar laws, the terms of membership also have to be fair to potential new members and the public at large. The approach taken by CBIC (see Section 4.5.2) and Lhasa Limited (see Section 4.5.7) – used here because they are companies with which the author is familiar – illustrate how these concerns are typically met. Both companies have stipulations in their articles of association that membership is open to all organizations who support their aims and that new members shall be admitted on fair and reasonable terms. They recognize different kinds of members – commercial, government, and academic – and charge them substantially different membership and software licensing fees. CBIC further distinguishes between members based in developed and developing countries and has a category of individual (as distinct from corporate) membership. Procedures at Lhasa Limited for ensuring fair contribution of knowledge and data by collaborators are described in Section 5.7. An essential consideration in all collaborations between commercial organizations is the risk of revealing too much to a competitor: A company does not want competitors to be able to work out their forthcoming product plans from the chemical structures they are putting through safety testing, for example; being too open about as yet unpatented work jeopardizes patentability, because patent examiners may consider it to amount to publication. As mentioned in Section 4.3.2, one solution to the problem is to use an “honest broker” – a trusted independent organization given access to confidential information and working with it on behalf of the members of a collaboration. For example, the broker may be able to develop predictive models for toxicity using data assembled from a group of companies, none of which have sufficient data in isolation. As long as the data are reasonably diverse and from several companies,

4.5 Collaborative Projects

the resultant models can be shared without the member companies seeing each other’s data or learning too much about it. Companies are used to giving sensitive information to contract research organizations for use in projects that remain strictly confidential between them. The contractor may get some spin-off benefit in the form of know-how but is not entitled to supply the results to other parties. The purpose of a cooperation to create new predictive models, however, is to produce results that will be of value at least to all the parties in the cooperation. The honest broker comes to hold results financed by the sponsors that are potentially marketable and which the sponsors likely want to be maintained in the long term. This can raise concerns about giving control to a broker over which the sponsors ultimately have not authority. To take an extreme situation, one member of the consortium, or some other company, might buy the broker if it is a commercial organization, thus gaining advantage over its competing, former collaborators. One way of avoiding the difficulties is for the honest broker to be owned collectively by the collaborating companies, typically as a company limited by guarantee or similar membership organization.

4.5 Collaborative Projects

The following list of collaborative projects is not exhaustive but it illustrates the kinds of schemes operating, the topics they cover, and how some collaboration issues are handled. For want of any clear alternative way of organizing them, they are presented alphabetically. Summary information about the projects is presented in Table 4.1. 4.5.1 British Industrial Biological Research Association (BIBRA)

BIBRA was founded in 1960 as a company limited by guarantee to carry out research on behalf of its member companies in the UK. It built up a set of reports on toxicity studies, entitled “Toxicity Profiles” that were available to members and nonmembers at different prices. In 1988, for example, a Toxicity Profile cost £10 for a member and £25 for a nonmember [10]. The company changed its name to BIBRA Toxicology International in 1990, and again to BIBRA International in 1994. The organization seems to have had difficulty in continuing to raise sufficient support and it ceased operating in 2001. A new, private limited company, Toxicology Advice and Consulting Ltd, started business in 2002 and in 2013 it changed its name to BIBRA Toxicology Advice and Consulting Ltd. It offers Toxicity Profiles as well as other services. Perhaps, the case for carrying out costly laboratory research on behalf of the original BIBRA members has diminished but the information service remains useful.

81

Laboratory studies during its early years and database development

Developing shared databases and knowledge bases in environmental chemistry and traditional medicines

Toxicological data exchange

Developing consistent guidelines for chemical safety testing

BIBRA

CBIC

DSSTox

ICH

Promotion of scientific collaboration and publication

Developing shared databases, knowledge bases, and prediction models in chemistry and toxicology

Regulatory acceptance of structure–activity relationships for the prediction of chemical toxicity

Promotion of data exchange and the development of data exchange standards

Forum for users and developers of software in the pharmaceutical industry

OECD QSAR Toolbox

OpenTox

PhUSE

MIP-DILI

Lhasa Limited

Prediction of drug-induced liver damage

MARCAR

ILSI/HESI

Predicting hazards to the environment from pharmaceuticals

Biomarkers for early detection of nongenotoxic carcinogenesis by pharmaceuticals

iPiE

Prediction of difficult toxicological end points

Finding ways to make better use of data from clinical settings

GETREAL

Tools for handling and analysis of electronic health records

EHR4CR

eTOX

Alternatives to the use of finite materials such as precious metals in pharmaceuticals manufacture

CHEM21

IMIa)

Main activities

Organization/Project

Table 4.1 Collaborative projects described in this chapter.

Not for profit

Government/industry collaboration re-constituted as not for profit

Intergovernmental

Not for profit

Not for profit

Government/industry collaboration

Industrial collaboration

Government project

Not for profit

Not for profit originally. Later re-constituted as a for profit company.

Type

United Kingdom

Europe

Paris

United Kingdom

USA

Europe

Europe

USA

Cameroon

United Kingdom

Primary location

82 4 Collaborative Approaches to Advance Chemical Safety

Sharing of data, and studies to generate data, for regulatory chemical safety compliance in Europe

REACH SIEF

a) IMI and SEURAT are umbrella projects incorporating the ones listed below them.

Sharing of clinical data

Database of information about traditional Chinese medicines

Traditional Chinese Medicine Database

YODA

Promoting a standard file format for exchange of toxicological and related data

ToxML

Web access to structure–activity relationship models

Developing a data bank for chemical and toxicological data to support the other projects in SEURAT

ToxBank

VEGA

Developing in vitro assays based on cell lines from human stem cells

SCR&Tox

Guidelines and standards for chemical hazard communication

Using omics data to predict long-term toxicity

NOTOX

Promotion of precompetitive sharing and publication of data and models to predict the human and environmental hazards of chemicals

Use of human liver cells in vitro to replace rodent long-term hepatotoxicity studies

HeMiBio

US Government

Biomarkers in cellular models for repeated dose toxicity

DETECTIVE

UN ADR and GHS

Computer modeling for toxicity prediction

COSMOS

SEURATa)

Precompetitive data aggregation and sharing

Pistoia Alliance

Industry/academic collaboration

Spin-off from government/industry collaboration

Government/industry collaboration

Intergovernmental

Academic

Not for profit

Government/industry collaboration

Government-promoted industrial collaboration

Not for profit

USA

Italy

USA

International

China

United Kingdom

Europe

Europe

USA

4.5 Collaborative Projects 83

84

4 Collaborative Approaches to Advance Chemical Safety

This is an example of an unusual situation in which a not-for-profit company limited by guarantee closed and parts, at least, of its business were subsequently carried forward by a profit-making company. 4.5.2 The Chemical Bioactivity Information Centre (CBIC) 8)

CBIC is based in Buea, Cameroon, but it was registered in England, in 2012, as a company limited by guarantee, to take advantage of the stronger company legal governance there. It has recently registered a wholly owned trading subsidiary in Cameroon. Its start-up was funded by a grant from Lhasa Limited (see Section 4.5.7) and its aims are similar but in the fields of environmental toxicology and traditional medicines – mainly, but not exclusively – in Africa. Its long-term aim is to build a membership of commercial, academic, and government organizations that will both share knowledge and data and sponsor the development of computer knowledge and databases. Its articles of association also recognize a category of individual membership and at this early stage its membership mainly comprises individuals. So far, efforts at CBIC have concentrated on collecting data about the components of traditional medicines in Africa from published sources and from unpublished sources, such as internal scientific reports at universities, analyzing the data and publishing papers about the findings, and creating a prototype knowledge-based system to predict the environmental toxicity of chemicals, using technology leased from Lhasa Limited. 4.5.3 The Distributed Structure-Searchable Toxicity Database Network – DSSTox

The DSSTox project is operated by a government regulator – the US Environmental Protection Agency (EPA). The EPA has web pages carrying data relevant 9) to the development of computer prediction models for chemical toxicity [11]. The service is incorporated into the EPA ACToR online data warehouse. DSSTox files are donated by organizations recognizing the need for collaboration in this area and willing to share them. The files contain the computer-recognizable chemical structures that are essential for chemical structure-activity modeling and this was the driving force when DSSTox was set up, as there were few sources of chemical structure files together with data on biological activity. Files are in the widely used, de facto standard SD file format [12]. DSSTox was an important initiative in promoting data sharing in the field of chemical toxicity, and one of the earliest, but there do not appear to have been donations from outside the US regulators in recent times.

8) www.cbic-africa.org/index.html (Sept. 17, 2015). 9) http://www.epa.gov/comptox/dsstox/index.html (Sept. 17, 2015).

4.5 Collaborative Projects

4.5.4 ICH

The full name of this organization, the International Conference on Harmonization of Technical Requirements for Registration of Pharmaceuticals for Human Use, is rarely used, “ICH” being preferred. It was set up in 1990 at a conference hosted by the European Federation of Pharmaceutical Industry Associations (EFPIA). At the time of writing of this book, ICH appears still to be an informal association, despite its now considerable importance and influence in its area of work and international membership (covering Europe, the United States, and Japan). Formal articles of association are being finalized with a view to incorporate under Swiss law during 2016. ICH is a collaboration for the creation of consistent guidelines for chemical safety testing and related activities. The report, “ICH M7 – multidisciplinary guidelines on genotoxic impurities,” is directly relevant to data sharing and to 10) the use of toxicity prediction software such as software from Leadscope Inc. , Derek (see Section 4.5.7), and the OECD (Q)SAR Toolbox (see Section 4.5.8). ICH M7 recommends use, and acceptance by regulators, of computer prediction of genotoxicity for low-level contaminants in pharmaceutical products without the need for laboratory tests. Genotoxicity is regarded as one of the easier kinds of toxicity to predict using computer models and one of the most reliably covered by them. ICH M7 recommends the use of at least two, complementary prediction methods – ideally one that is statistically based and one that is based on human reasoning about the processes involved in interactions between chemicals and biological systems. 4.5.5 Innovative Medicines Initiative (IMI) 11)

IMI is a huge operation now costing over three billion Euros [13]. Its aim is to make progress in the most challenging areas of human health research. Sponsoring pharmaceutical companies, members of the European Federation of Pharmaceutical Industry Associations (EFPIA), make contributions in kind to projects within IMI (staff resources, use of laboratories and equipment, etc.) and the European Union provides matching cash funds. The funds are spent on supporting the participation of academic and not-for-profit research organizations. There are currently 56 projects in IMI, of which the 7 described further relate to chemical safety. The procedure for setting up projects in IMI is unusual. More widely, there have been problems with the slowness of processing bids for government-sponsored projects and, in particular, with the high costs of bidding for organizations that are not successful – often academic organizations with very limited budgets. 10) www.leadscope.com/about_us_pub.php (Sept. 17, 2015). 11) www.imi.europa.eu/(Sept. 17, 2015).

85

86

4 Collaborative Approaches to Advance Chemical Safety

In IMI, a call is put out for rather simple, outline bids from consortia for a project defined by IMI. Proposals must describe the intended scientific approach and explain how it will address the problems set out in the call, but they do not need to go into detail. A consortium must include a mix of academic and commercial organizations from across Europe (or, at least, not just from one country) and the proposal must show how each participant is needed and how, collectively, they cover all the skills and know-how that the project requires. They must include two EFPIA members who undertake to oversee the work and, because of the size of the projects, it is normally expected that a suitable organization will be included to manage the project. The winning consortium is chosen through a process of peer review and invited to submit a second stage, very detailed proposal. The invitation includes comments on potential weaknesses in the outline proposal that need to be addressed. The detailed proposal has to include sections on the legal agreements between participants on exploitation of results and similar matters. The detailed proposal is assessed and if it is accepted, the consortium is authorized to proceed. If the proposal is not accepted (following resolution of any minor issues that are identified), the call is abandoned – there are no second-choice candidates. However, the author’s understanding is that there have been only one or two failures at the second stage. Although the research is precompetitive, IMI expects the projects either themselves to roll over into ongoing self-funding projects or to produce deliverables that will be exploited. Calls from IMI to bid for contracts are open to all organizations. The mechanism for becoming involved is to identify a suitable group of organizations to work with and to prepare a joint bid to submit to IMI. There are mechanisms for new collaborators to join some of the existing projects and it would also be possible for a major sponsoring group to join the IMI parent body itself (currently there are two members – EFPIA and the European Union). IMI is open to proposals for new topics outside the scope of their current projects and these can be submitted via their web site11). It is perhaps obvious from the aims of IMI that topics need to be major scientific challenges that will require adventurous research rather than research that falls into the category of “yet more of the same.” 4.5.5.1

CHEM21

CHEM21 is seeking alternatives to the use of finite materials such as precious metals in pharmaceutical manufacturing. At first sight it may not look like a chemical safety project but researchers in the project are concentrating on finding “green chemistry” solutions – solutions that are safer for the environment and, hence for human health. The consortium includes six large pharmaceutical companies, thirteen academic organizations, and four small to medium-sized enterprises (SME). 12)

12) www.chem21.eu/(Sept. 17, 2015).

4.5 Collaborative Projects

4.5.5.2

Electronic Health Record for Clinical Research (EHR4CR) 13)

EHR4CR [14] aims to deliver tools for the handling and analysis of electronic health records. It will thus support activities in other IMI projects. Barriers to making effective research use of the information in health records include software standardization issues, ethical questions, including the need for patient confidentiality, and a variety of national and international regulations. There are 34 partners in the project plus 2 subcontractors. 4.5.5.3

eTOX 14)

eTOX was one of the early IMI projects. It is currently on 1 year extension of its original period and moving toward a follow-up project. The participants are thirteen pharmaceutical companies, twelve academic or not-for-profit organizations, and five SME. The aim of the project is to find solutions to the problem of predicting toxicity observed during repeated dose studies. Repeated dose studies are among the most expensive and they depend on the use of animals but the kinds of toxicity they reveal are difficult to predict (and currently impossible to rule out) using computer models. The project has developed an application through which researchers can access multiple prediction models and data coming from work in the project [15]. Being close to the end of its contract period, eTOX provides an example of the way IMI projects may think about ensuring long-term continuation. According to the eTOX project plan, the intention is for the project to roll over into one run by an “honest broker” – the model mentioned as a favored one for many collaborations in Section 4.4. The organization will have terms for admission of new members and will thus be an open collaboration, but one with secure methods of funding and governance – neither restricting access to its resources to a closed group nor publishing them free of charge. 4.5.5.4

GETREAL 15)

There are 29 participating organizations in the GETREAL project. The aim of the project is to show how new methods of collecting real world evidence could be developed for adoption earlier than at present in pharmaceutical research and development, and in healthcare decision making. By “real world evidence” is meant data from clinical settings in practice. The project involves industry, academia, regulatory agencies, reimbursement agencies, healthcare budget holders, and patient groups – 15 pharmaceutical companies, 11 academic organizations, 1 SME, and 1 patients’ organization. It is looking into issues such as the acceptability and usefulness of this kind of evidence; ways of analyzing the data to assess the efficacy of new medicines, including consideration of the scientific validity of the approaches used; finding practical solutions to the challenges of

13) www.ehr4cr.eu/(Sept. 17, 2015). 14) www.etoxproject.eu/(Sept. 17, 2015). 15) www.imi-getreal.eu/(Sept. 18, 2015).

87

88

4 Collaborative Approaches to Advance Chemical Safety

conducting studies in the early development of medicines; and determining best practice in modeling to assess the effectiveness of medicines. 4.5.5.5

iPiE

There is a concern about the effects of pharmaceuticals on the environment. Potential problems with endocrine disrupting chemicals such as contraceptive chemicals and their metabolites are already recognized but, more generally, it would not be surprising to find that chemicals designed to be biologically active, as pharmaceuticals are, had effects in the environment, even at low doses. The 16) iPiE project is carrying out field research as well as developing computer models to predict problems and a computer application to allow scientists to access the models via one interface. The project has been set up as a sister project to eTOX (see Section 4.5.5.3) and it is intended that they will share the same computer platform and database. Although basic aquatic lethal toxicity data for chemicals have been required by regulators for a long time, this appears to be the first time a substantial, collaborative effort has been put into studying and predicting wider, more subtle, effects on the environment and on species other than the aquatic ones usually used in existing studies (a few species of fish, daphnia, and algae). 12 pharmaceutical companies, 10 academic or not-for-profit organizations, and 3 SME are participating in the project. iPiE shares with eTOX plans for long-term continuation, most probably in the care of an honest broker. 4.5.5.6

MARCAR 17)

The MARCAR project aims to discover biomarkers for early detection of nongenotoxic carcinogenesis by pharmaceuticals during preclinical testing. There are five pharmaceutical companies, six academic organizations, and one SME in the project. There are useful methods, based on laboratory assays and computer prediction models, for detecting or predicting potential genotoxicity, which may lead to carcinogenesis, but the prediction of nongenotoxic carcinogenesis is more difficult. If the problem becomes apparent well into preclinical trials, a great deal of money has already been spent on a potential new pharmaceutical, animal tests have been run unnecessarily, and time has been lost that could have been spent on other promising compounds. The project is concentrating effort on liver carcinogenesis, as it is in this organ that problems most often manifest themselves, but other organs are not excluded from the work. 4.5.5.7

MIP-DILI

The MIP-DILI project (mechanism-based integrated systems for the prediction 18) of drug-induced liver injury) is exploring better ways to predict whether chemicals – in particular, potential pharmaceuticals – are likely to cause liver damage. 16) i-pie.org/(Sept. 21, 2015). 17) www.imi-marcar.eu/(Sept. 21, 2015). 18) www.mip-dili.eu/(Sept. 23, 2015).

4.5 Collaborative Projects

They are working on new in vitro methods, improving understanding of the links between observations in vitro and effects in vivo, and developing and improving computer prediction methods. There are 11 pharmaceutical companies, 9 academic organizations, and 6 SME in the project. 4.5.6 International Life Sciences Institute (ILSI) and ILSI Health and Environmental Sciences Institute (HESI) 19)

The International Life Sciences Institute (ILSI) has been operating since 1978. It describes itself as a public–private partnership. Although its membership of over 500 organizations appears to be all commercial, its board of trustees includes at least 50% public sector representatives such as academic scientists. It is incorporated as a not-for-profit charity. Its aim is to promote worldwide scientific collaboration to improve human and environmental health and its first work in 1978 was on the toxicology and risk/benefit of food ingredients. Currently it is active in food and water safety; toxicology and risk science; nutrition, health, and well being; and sustainable agriculture and nutrition security. Toxicology and risk science are now primarily the responsibility of a subsidiary of ILSI, the ILSI Health and Environmental Sciences Institute (HESI) [16]. ILSI places particular emphasis on putting knowledge into the public domain through conferences, workshops, and scientific publications (some of which it publishes itself). Examples are papers on chemicals that may migrate from packaging into foods [17] and on allergens that arise during food processing [18]. Note that ILSI collaborations thus differ from many of the others described in this section, where results are typically shared among the collaborators and not published (although accessible to the public through mechanisms for the admission of new members). Early this century, ILSI HESI coordinated a project entitled “International Toxicology Information Centre” (ITIC) [19,20]. Senior toxicology department staff at Procter and Gamble and Unilever wanted to promote data sharing to reduce animal toxicology studies but were aware of the potential legal and commercial difficulties. They commissioned the author to carry out a survey of scientists to find out whether they supported the idea of data sharing in principle and whether they thought their own company would join a collaboration. Broadly, the findings were that scientists were keen to share data – in order to advance science more effectively as well as to reduce animal testing – but doubted their companies’ legal departments would let them do it. ITIC was a pilot to see in practice if any data might be released for sharing and, if so, on what terms, and to demonstrate the feasibility of building a chemical structuresearchable database containing complex and varied toxicological data. It was found that appropriate data could be shared and this led to the Vitic project (see Section 4.5.7).

19) www.ilsi.org/(Sept. 23, 2015).

89

90

4 Collaborative Approaches to Advance Chemical Safety

4.5.7 Lhasa Limited 20)

Lhasa Limited [21] was founded in 1983 as a company limited by guarantee. Originally, its aim was to support the development of a computer program to assist with chemical synthesis planning, LHASA, at Harvard University [22]. Called Lhasa UK at first to distinguish it from the parent research group at Harvard, the company later changed its name because of its international nature. The team at Harvard had created one of the earliest knowledge-based systems and their interest, as an academic group, was in the design and improvement of the program itself. To be practically useful, the program needed an expanded knowledge base and the idea was for commercial members of Lhasa Limited to work on that, adding knowledge from their own chemical synthesis research. Members supported the project by making contributions in kind, since the primary aim was to draw on their knowledge and skills. The articles of Lhasa Limited make a distinction between Full Members of collaborative groups, who contribute in kind, and Associate Members who have no suitable material to offer but contribute sponsorship fees. The former have benefits such as being entitled to receive and keep the results of the collaboration, for their own use only. What constitutes equal contribution to a project is set out in Codes of Practice and each contributing member was originally required to put specified staff resources into knowledge base writing. Later, the rule was adapted for allowing the alternative of funding staff at Lhasa Limited to carry out the work, as this was more efficient. During the course of the work a spin-off application, Derek [23] was developed to predict the mammalian toxicity of chemicals and this rapidly took the focus of Lhasa Limited into toxicology. Later, a replacement system making use of more advanced reasoning was developed in collaboration with Imperial Cancer Research Fund, City University, and Logic Programming Associates [24,25]. As mentioned in Section 4.5.6, an ILSI HESI project rolled over into one at Lhasa Limited called Vitic. While companies remain wary of sharing, or unable to share, data about pharmaceutical products or potential products, there are areas where sharing is possible. Many pharmaceutical companies use the same excipients and other components in formulations of pharmaceutical compounds for research testing. While the active ingredients are secret, these other components are not. They have to be tested, to make sure they do not interfere with the studies, and companies can reduce animal use and save a good deal of expense by sharing the results of the tests [26]. Similarly, many reagents and intermediates are common to manufacturing processes used by different companies. They need to be tested both because they might be minor contaminants in the final product and to guide policy on safe working practices at manufacturing sites.

20) www.lhasalimited.org/(Sept. 23, 2015).

4.5 Collaborative Projects

In addition to the above activities, Lhasa Limited has collaborative projects on prediction of mammalian metabolism, forced chemical degradation, and assessment of potential contamination of pharmaceuticals by hazardous intermediates. The company develops its own database, knowledge base, and prediction modeling software that provide the mechanism for sharing information between its members. 4.5.8 OECD (Q)SAR Toolbox 21)

The Organisation for Economic Cooperation and Development (OECD) in its present form was founded in 1961. Its members are countries – currently 39 of them. The OECD provides guidance documents and services relating to a huge range of topics, and only two that are particularly relevant to chemical toxicology work are mentioned here. A problem for government regulators, who want companies to submit data with applications to register new chemicals, is the disparity between how toxicological data are recorded in computer files by different companies and software. The OECD has developed, and continues to develop, a set of standard file formats for toxicological data, each relating to a regulatory requirement, such as data on skin sensitization or carcinogenicity. They are called OECD Harmonised 22) Templates . Their development proceeded alongside the development of the European IUCLID submission standards, the formats have a lot in common, and it is hoped that they will eventually become the same. In this connection, the OECD has also contributed to work on ToxML (see Section 4.5.14). The OECD does not carry out or fund scientific research, but its members recognized the need for an internationally accepted computer system to support the assessment of human and environmental chemical hazards by research organizations and regulators. There is a general desire to allow the use of predictions based on statistical analysis of existing data and other computer models, as alternatives to tests that use animals, in dossiers submitted with applications to register new chemicals (including pharmaceuticals) with regulatory authorities. Two problems are hindering the acceptance of these methods. The first is that countries have different views about which prediction methods are appropriate. The second, though, is the bigger one. There are many reports that statistical and other computer predictions described in peer-reviewed scientific publications cannot be repeated because there is insufficient detail [27]. The OECD has, therefore, supported the development of the OECD (Q)SAR Toolbox, with funding from the European Union (QSAR stands for “quantitative structure–activity relationships” and the letter “Q” is in parentheses because the Toolbox also incorporates nonquantitative prediction models).

21) www.oecd.org/(Sept. 24, 2015). 22) http://www.oecd.org/ehs/templates/ (Sept. 24, 2015).

91

92

4 Collaborative Approaches to Advance Chemical Safety

The (Q)SAR Toolbox provides access to sources of existing data, including chemical structure connection tables, a set of prediction models, and tools for building new statistical models (each of which normally takes the form of an equation relating features of chemicals to a biological activity). All of the steps taken during the creation of a model are logged and preserved. So, if someone includes predictions from a model contained or developed in the (Q)SAR Toolbox to support a regulatory submission, the regulator can repeat and examine the whole modeling process. As its name implies, the (Q)SAR Toolbox is intended for experts developing computer prediction models – in particular, statistical ones – as well as applying them, and users without that level of experience and understanding may find it daunting. It is nevertheless available to 23) 24) anyone , there are training courses , and its use is increasing. 4.5.9 OpenTox 25)

OpenTox [28] was a European FP7 project to develop a web platform giving users access to toxicity predictions and data at diverse, distributed sites. In support of its work, OpenTox also developed ontologies in the fields of toxicity it covered and it contributed to the development and promotion of ToxML (see Section 4.5.14). Since the end of its period as a European FP7 project, OpenTox has run a well-attended annual series of conferences in Europe and the USA and 26) launched OpenToxipedia that is described as a community-based predictive toxicology knowledge resource. Its early focus has been on toxicology vocabulary and ontology. The OpenTox prediction application continues to be available via the OpenTox web site. A user enters a query, it is passed automatically to the remote sites, and the results are displayed to the user as they come in. The information returned is limited because the sites linked into the project are themselves mostly limited to data in the public domain. It should be noted that the objective of the project was to demonstrate that such an application could be created and would be potentially useful – it was not to create and operate a commercial service. Until now, OpenTox has suffered the disadvantage of a governmentsponsored collaboration mentioned in Section 4.5 that there was no formal organization with a long-term income stream to keep the project running beyond the grant period. However, a meeting has been announced recently to ratify proposals for formal incorporation in Switzerland and so OpenTox looks set to continue, and perhaps one of its actions will be to develop the application into a commercially useful one. Whether or not the application is developed further, OpenTox project contributions to standardization and its annual conferences in 23) www.qsartoolbox.org/(Sept. 25, 2015). 24) http://www.qsartoolbox.org/display-news/-/journal_content/title/web-training-may-2014 (Sept. 25, 2015). 25) www.opentox.org/(Sept. 25, 2015). 26) www.opentox.org/opentoxipedia (Sept. 25, 2015).

4.5 Collaborative Projects

this field of research are likely to be of significant value to the scientific community. 4.5.10 PhUSE 27)

PhUSE, Pharmaceutical Users of Software Exchange , is a not-for-profit company limited by guarantee set up as a forum for software users and developers in the pharmaceutical industry. It holds regular conferences, incorporating workshops, including annual joint meetings with the US Food and Drug Administration, and members of working groups collaborate to address particular topics – for example, at the time of writing this chapter, there are computational science working groups on semantic technology, optimizing the use of data standards, standard scripts for reporting and analysis, nonclinical activities (PhUSE has otherwise mainly concentrated on clinical studies), and emerging trends and 28) technologies . For a short period up to 2012, PhUSE produced a journal, 29) Pharmaceutical Programming . 4.5.11 The Pistoia Alliance 30)

The Pistoia Alliance is a not-for-profit organization founded in 2009 by representatives from AstraZeneca, GSK, Novartis, and Pfizer. Fourteen pharmaceutical organizations are now described as core members of the Alliance, with a further 17 participating members, most of which are software or information companies, and many additional collaborators. The Alliance does not focus on chemical safety research but three projects are of interest to people working in this area: their Controlled Substance Compliance Service, which provides access to information about what substances, in what countries, are subject to legal controls (narcotics, substances of use to terrorists, etc.) [29]; the Hierarchical Editing Lan31) 32) guage for Macromolecules (HELM) ; and the Ontologies Mapping Project . 4.5.12 REACH Substance Information Exchange Forums (SIEF) 33)

The European REACH regulations require the submission of toxicological data for all substances marketed in Europe above a specified tonnage (which started at 27) 28) 29) 30) 31) 32)

www.phuse.eu/(Sept. 25, 2015). www.phuse.eu/cs-working-groups.aspx (Sept. 25, 2015). http://www.pharmasug.org/download/PH_Programming.pdf (Sept. 25, 2015). www.pistoiaalliance.org/(Sept. 25, 2015). http://www.pistoiaalliance.org/rfp-published-helm-ambiguity-and-web-services/ (Sept. 25, 2015). http://www.pistoiaalliance.org/pistoia-alliance-announces-start-of-ontologies-mapping-project/ (Sept. 25, 2015). 33) European Commission Regulation 453/2010 and subsequent amendments.

93

94

4 Collaborative Approaches to Advance Chemical Safety

100 tonnes but is diminishing in stages so that gradually all chemicals with significant usage will be covered). As mentioned in Section 4.5, this would potentially necessitate an immense amount of animal testing and that would be in conflict with other regulations requiring animal testing to be minimized. Some testing can be avoided by appropriate use of so-called read-across – assessing the likely toxicological hazards, if any, of a compound by looking at the properties of a series of other, closely related compounds – and/or computer prediction models. The regulators have determined that where tests do need to be done they may be done only once – that is, there must be no duplication between companies. To this end, a scheme has been set up whereby companies seeking to register the same compound are required to share toxicological data on reasonable terms, if it already exists, and to share the costs of generating it, once, if not. The collaborations are termed SIEFs (Substance Information Exchange Forums) and joining a SIEF is a legal obligation for companies involved in registering chemicals. There are no rules imposed on how a SIEF is structured and run, beyond the requirement for all relevant data to be shared, and so it is likely that several approaches are being taken. 4.5.13 SEURAT-1 (Safety Evaluation Ultimately Replacing Animal Testing) 34)

SEURAT is funded under the European FP7 Research Initiative. Its aim is to find replacements for in vivo repeated dose toxicity tests, which are now prohibited for new cosmetics in Europe and are difficult to model. The participants in SEURAT are 5 industrial organizations, 18 universities, 22 research organizations, 21 SMEs, and 5 government or similar bodies. At the time of writing this chapter, SEURAT has one more year to run under 35) current funding. Annual reports provide information on progress in the projects within SEURAT that are listed further. 4.5.13.1

COSMOS 36)

COSMOS is developing molecular modeling techniques and related computational topics such as the extrapolation of observations in vitro to predictions in vivo. Particular emphasis is on making use of the concept of adverse outcome pathways – the identification and understanding of how the interaction of a molecule with a biological site triggers a sequence of events leading to an observed toxicological effect. 4.5.13.2

DETECTIVE 37)

DETECTIVE is looking for biomarkers in cellular models for repeated dose testing in vitro. The project is concentrating on human hepatic, cardiac, and renal models. 34) 35) 36) 37)

www.seurat-1.eu/(Sept. 25, 2015). http://www.seurat-1.eu/pages/library/seurat-1-annual-reports.php (Sept. 25, 2015). www.cosmostox.eu/(Sept. 25, 2015). www.detect-iv-e.eu/(Sept. 25, 2015).

4.5 Collaborative Projects

4.5.13.3

HeMiBio 38)

The aim of HeMiBio is to develop a bioreactor using human liver cells to provide a trustworthy in vitro alternative to rodent long-term hepatotoxicity studies. 4.5.13.4

NOTOX 39)

NOTOX is working toward computer prediction models that will use cellular and molecular signatures in “omics” data to predict long-term toxicity, and designing experimental systems to find predictive endpoints that can be integrated into computer models. 4.5.13.5

SCR&Tox 40)

SCR&Tox is looking at ways to develop in vitro assays using cell lines created from human stem cells. 4.5.13.6

ToxBank 41)

ToxBank is providing a service to the other projects in SEURAT by setting up a data warehouse, including a database of chemical and biological information and actual samples of compounds, cells, and cell lines. It might, thus, also become a service of potential use outside the project. 4.5.14 ToxML

The absence of an international standard for data exchange is a constraint on collaboration in toxicology work at every level. In the United States, the problem has been partially addressed by the SEND format, and in Europe by IUCLID, for submission of data with regulatory submissions. Cooperation is further strengthened by the development of the OECD Harmonised Templates mentioned in Section 4.5.8. However, all of these initiatives relate to data for routine submission to regulators. More widely, researchers need to be able to exchange data that fall outside the scope of regulatory submissions, for activities such as the development of chemical structure-based prediction models for toxicity. At the time when the OECD and collaborators started work on the Harmonised Templates (which are XML based), a project was also initiated for the development of a standard with wider coverage, entitled ToxML [30], involving many of the same individuals. ToxML was taken up by, among others, the OpenTox project (see Section 4.5.9) and some divisions within the US FDA. It was adopted jointly 42) by Leadscope Inc. and Lhasa Limited [see Section 4.5.7] initially to allow their customers/members to load data coming from projects at the US FDA.

38) 39) 40) 41) 42)

www.hemibio.eu/(Sept. 25, 2015). www.notox-sb.eu/(Sept. 25, 2015). www.scrtox.eu/(Sept. 25, 2015). toxbank.net/(Sept. 25, 2015). www.leadscope.com/(Sept. 25, 2015).

95

96

4 Collaborative Approaches to Advance Chemical Safety

Thereafter, Leadscope carried ToxML forward in support of ongoing work with the FDA, and partly in collaboration with OpenTox. 43) In 2011, the ToxML Standard Organisation (TSO) was set up as a not-forprofit organization to promote the continuing development and wider adoption of ToxML. The TSO web site incorporates a curated wiki site with an editor so that organizations wishing to exchange data with each other can extend ToxML to suit their needs and, at the same time, increase the usefulness of the standard to the wider community. Progress appears to have been slow and, even though the need for a dynamic standard is clear and ToxML is closely allied to both the OECD Harmonised Templates and the IUCLID format, the future of the project at the time of writing this chapter remains unsure. 4.5.15 The Traditional Chinese Medicine Database

There has been a lot of activity and publication in traditional Chinese medicine research in recent years. A public web-based service giving access to data on 44) traditional Chinese medicines appears to be funded at present by the YC Lab (Computational Systems Biology Lab) at China Medical University (Taiwan). Data can be downloaded free of charge and the site encourages users to upload donations of data. The site supports searching by chemical formula, SMILES codes, and some physicochemical properties. 2D and 3D chemical structures are available. 4.5.16 United Nations – the European Agreement Concerning the International Carriage of Dangerous Goods by Road (ADR) and the Globally Harmonized System of Classification and Labeling of Chemicals (GHS)

The United Nations Organization represents collaboration on a grand scale, largely outside the scope of this chapter, but ADR and GHS are of central importance to chemical hazard assessment, including chemical toxicity. Although ADR itself is described as a European agreement, it was established under the auspices of the United Nations and the guidelines it contains are consistent worldwide. They are incorporated into national law in countries across the world. Despite its reference to road transport, ADR is now also consistent with international agreements on transport by rail, sea, and air. In hard copy form, the ADR guidelines are contained in two volumes with orange covers popularly known as “the Orange Book.” The implementation of the guidelines is most visible to the public in the form of the placards on the backs of lorries showing UN numbers for goods being carried and the large, diamond-shaped

43) www.toxml.org/(Sept. 25, 2015). 44) www.chemtcm.com/(Sept. 25, 2015).

4.5 Collaborative Projects

symbols on external packaging warning of hazards in the event of an incident during transport. Consistent, albeit sometimes complicated, rules determine which symbols are to be displayed for a chemical. For example, the picture of a flame, representing flammability, is required if the flash point of a liquid is less than 60 °C. It is also required for a liquid with higher flash point but carried hot at or above its flash point temperature. Until a few years ago, it was necessary to have animal acute lethal toxicity data for a product in order to classify it for labeling for transport. However, the procedures for classifying for human and environmental health are now aligned with those used for GHS. Regional enactments of GHS became mandatory in Europe and the United States in June 2015 and had already been implemented in some other countries, such as Japan and New Zealand. GHS determines the warnings and advice to users that must be printed on packages along with warning symbols. Sets of equations allow labeling for human and environmental health hazards for “preparations” – mixtures – to be computed from the hazards associated with their ingredients, thus avoiding the need for animal tests on the preparations themselves. GHS also provides guidelines on the layout and contents of safety data sheets – documents that must accompany chemicals delivered to businesses, giving information on actions to take in an emergency and so on. The combination of allowing computation of classification for preparations and the obligation placed on suppliers to provide safety data sheets for the ingredients that go into them imposes a kind of collaboration between manufacturers and suppliers. In addition, there is strong public and government pressure on companies to collaborate where toxicity studies need to be done for individual substances, especially in Europe (see Section 4.5.12). Both ADR and GHS allow and encourage the use of alternatives to animal tests wherever possible. For example, if a product has a very high or very low pH in water, it is obvious that it is likely to be corrosive to skin and damaging to eyes. In the absence of clear information to the contrary, the product must be labeled as corrosive without any animal testing [7]. The guidelines also state that if corrosivity is predicted by (Q)SAR, the product must be labeled without animal testing. Use of (Q)SAR for other health hazards is increasingly being encouraged (see Section 4.5.4). This is a difficult area for the great majority of chemical suppliers, who have no experience or understanding of (Q)SAR modeling, which can be more subjective than is sometimes recognized and requires careful interpretation. Significant collaboration through SIEFs is imposed under European rules (see Section 4.5.12) but increasing voluntary collaboration out of necessity can be expected, even between the traditionally secretive small chemical companies. 4.5.17 US Government–Industry Collaborations

Creation of the Epiwin suite of programs giving access to data and prediction models for environmental hazards and of Oncologic, for the prediction of

97

98

4 Collaborative Approaches to Advance Chemical Safety

mammalian carcinogenicity (see Section 4.4), involved some collaboration with commercial organizations. Divisions within the US FDA have also long collaborated with academic and commercial organizations in areas of public benefit. For example, they have collaborated with Multicase Inc., Leadscope Inc., and Lhasa Limited on the development and improvement of (Q)SAR toxicity prediction models, drawing on knowledge, experience and hypotheses of staff at FDA [31–37]. Leadscope has harvested and compiled databases of FDA toxicity information under a Cooperative Research and Development Agreement with the FDA. The results are available from Leadscope for a nominal charge to cover distribution costs, in formats that can be imported into other, suitable software as well as Leadscope software [38,39]. Leadscope also has a knowledge sharing program, involving over a dozen corporate sponsors, to address specific (Q)SAR regulatory issues identified through discussions with sponsors and regulatory agencies. The US military collaborates with academic and other not-for-profit organizations in areas of wide public benefit (i.e., not necessarily for purely military purposes) providing some grant funding, and these collaborations can involve multiple organizations. 4.5.18 VEGA

Computer-Assisted Evaluation of industrial chemical Substances According to 45) Regulations (CAESAR) [40,41] was a European FP7 project to develop QSAR 46) models relevant to REACH legislation. A web-based application, VEGA , now provides access to models developed during the project. 4.5.19 Yale University Open Data Access (YODA)

Over the last couple of years, pharmaceutical companies and organizations running clinical trials, under pressure from regulators and the public, have become increasingly willing to look at ways of sharing clinical data with each other as well as with the regulators and potentially with the general public. This is a difficult area for collaboration. Pharmaceuticals in clinical trials at a research stage are commercially sensitive. They may not be fully protected by patents and there are obvious risks in letting competitors know your plans for new products. Making data from clinical trials public puts companies at risk even long after a product goes onto the market. With hindsight, a third party might claim to find a hint somewhere in a trial of a toxicity problem that does not become apparent until a pharmaceutical has been in use for a long time. This might trigger attempts at litigation that, even if groundless, can be costly and damaging to the 45) www.caesar-project.eu/(Sept. 28, 2015). 46) www.vega-qsar.eu/(Sept. 28, 2015).

References

public image of a company. But the biggest constraint to data sharing is probably the need for patient confidentiality. Companies do not want to breach that confidence, either deliberately or accidentally, for ethical reasons, quite apart from legal and public relations issues. Several major companies have nevertheless announced decisions to make the results of clinical trials more easily available. The Yale University Open Data 47) Access project, YODA [42], is promoted by a group of academically based clinical researchers to act as a trusted third party to provide access to clinical research data. Data remain the property of the originating organizations but they undertake to allow access in accordance with terms agreed with YODA. Currently, two pharmaceutical organizations are making their data accessible 48) through YODA – Medtronic Inc., an Irish medical device company and Johnson and Johnson, a company with divisions active in many areas of health care, 49) including the development of pharmaceuticals .

4.6 Conclusions

There is increasing interest among the big pharmaceutical companies, in particular, in working together wherever it can be done without threatening competitiveness and without breaching regulations designed to prevent the formation of cartels. The high costs of doing toxicological research, concerns about use of animals in experiments, the difficulties of understanding causes of toxicity without access to data for large numbers of compounds, and the desire of scientists to share their work are all powerful incentives. Companies recognize that collaboration can advance progress without damaging competitive advantage and it is evident from the examples in this chapter that it is well established in some areas. The trend is set for collaboration to increase, to the benefits of science, science-based industry, and the general public.

References 1 Posthumus, R., Traas, T.P., Peijnenburg,

W.J., and Hulzebos, E.M. (2005) External validation of EPIWIN biodegradation models. SAR and QSAR in Environmental Research, 16 (1–2), 135–148. 2 Allen, F.H., Bellard, S., Brice, M.D., Cartwright, B.A., Doubleday, A., Higgs, H., Hummelink, H.T., Hummelink-Peters, B.G., Kennard, O., Motherwell, W.D.S.,

Rodgers, J.R., and Watson, D.G. (1979) The Cambridge Crystallographic Data Centre: computer-based search, retrieval, analysis and display of information. Acta Crystallographica, B35, 2331–2339. 3 Manfred, Fleischer. (2007) Testing costs and testing capacity according to the REACH requirements – results of a survey of independent and corporate GLP

47) yoda.yale.edu/welcome-yoda-project (Sept. 17, 2015). 48) www.medtronic.com/(Sept. 28, 2015). 49) www.jnj.com/about-jnj (Sept. 28, 2015).

99

100

4 Collaborative Approaches to Advance Chemical Safety

4

5

6

7

8

9

10

11

12

laboratories in the EU and Switzerland. Journal of Business Chemistry, 4 (3), 96–114. Regulation (EC) No. 1970/2006 of the European Parliament and of the Council http://eur-lex.europa.eu/legal-content/EN/ TXT/?uri=CELEX:02006R1907-20150601 (Jan.19, 2016). Regulation (EC) No. 1223/2009 of the European Parliament and of the Council http://eur-lex.europa.eu/legal-content/EN/ TXT/?uri=CELEX:32009R1223 (Jan.19, 2016). Bergkamp, L. and Herbatschek, N. (2013) Key concepts and scope, in The European Union REACH regulations for chemicals: Law and Practice (ed. L. Bergkamp), Oxford University Press, pp. 40–81. United Nations (2015) Globalised system of classification and labelling of chemicals, New York and Geneva. Clauses 3.2.1.2 and 3.3.1.2, and Figures 3.2.1 and 3.3.1. OECD (2014) New guidance document on an integrated approach on testing and assessment (IATA) for skin corrosion and irritation Available at http://www.oecd .org/officialdocuments/publicdisplay documentpdf/?cote=ENV/JM/MONO (2014)19&doclanguage=en (Jan. 19, 2016). Filimonov, D. and Poroikov, V. (2005) Why relevant chemical information cannot be exchanged without disclosing structures. Journal of Computer-Aided Molecular Design, 19 (9–10), 705–713. The British Industrial Biological Research Association (1988) Toxicity profiles. Packaging Technology and Science, 1, 45–46. Williams, C., Wolf, M., and Richard, A.M. (2009) DSSTox chemical-index files for exposure-related experiments in ArrayExpress and gene expression omnibus: enabling toxico-chemogenomics data linkages. Bioinformatics, 25 (5), 692–694. Dalby, A., Nourse, J.G., Hounshell, W.D., Gushurst, A.K.I., Grier, D.L., Leland, B.A., and Laufer, J. (1992) Description of several chemical structure file formats used by computer programs developed at Molecular Design Limited. Journal of

13

14

15

16

17

18

19

Chemical Information and Modeling, 32 (3), 244. Goldman, M. (2012) The innovative medicines initiative: A European response to the innovation challenge. Clinical Pharmacology and Therapeutics, 91 (3), 418–425. De Moor, G., Sundgren, M., Kalra, D., Schmidt, A., Dugas, M., Claerhout, B., Karakoyun, T., Ohmann, C., Lastic, P.-Y., Ammouri, N., Kush, R., Dupont, D., Cuggia, M., Danielm, C., Thienpont, G., and Coorevits, P. (2015) Using electronic health records for clinical research: the case of the EHR4CR project. Journal of Biomedical Informatics, 53, 162–173. Sanz, F., Carrió, P., López, O., Capoferri, L., Kooi, D., Vermeulen, N.P., Geerke, D.P., Montanari, F., Ecker, G.E., Schwab, C.H., Kleinöder, T., Magdziarz, T., and Pastor, M. (2015) Integrative modeling strategies for predicting drug toxicities at the eTOX project. Molecular Informatics, 34, 477–484. Takei, A. (2015) ILSI Health and Environmental Sciences Institute: global leader in advancing translational science to create science-based solutions for a sustainable, healthier world. Genes and Environment, 37, 1. Koster, S., Bani-Estivals, M.-H., Bonuomo, M., Bradley, E., Chagnon, M.C., Garcia, M.L., Godts, F., Gude, T., Helling, R., Paseiro-Losada, P., Pieper, G., Rennen, M., Simat, T., and Spack, L. (2015) Guidance on best practices on the risk assessment of non intentionally added substances (NIAS) in food contact materials and articles. ILSI Europe Report Series, 2015, 1–70. Verhoeckx, K., Vissers, Y., Baumert, J.L., Faludi, R., Feys, M., Flanagan, S., HerouetGuicheney, C., Holzhauser, T., Shimojo, R., van der Bolt, N., Wichers, H., and Kimber, I. (2015) Food processing and allergenicity. Food and Chemical Toxicology, 80, 223–240. Judson, P.N., Cooke, P.A., Doerrer, N.G., Greene, N., Hanzlik, R.P., Hardy, C., Hartmann, A., Hinchliffe, D., Holder, J., Muller, L., Steger-Hartmann, T., Rothfuss, A., Smith, M., Thomas, K., Vessey, J.D., and Zeiger, E. (2005) Towards the creation

References

20

21

22

23

24

25

26

of an international toxicology information centre. Toxicology, 213 (1–2), 117–128. Mioussea, I.R., Currieb, R., Dattac, K., Ellinger-Ziegelbauerd, H.j., Frenche, J.E., Harrilla, A.H., Koturbasha, I., Lawtonf, M., Manng, D., Richard, R., Meehanh, R.R.j., Moggsi, J.G.j., O’Lonek, R., Rasoulpourl, R.J., Reijo Peram, R.A., and Thompson, K. (2015) Importance of investigating epigenetic alterations for industry and regulators: an appraisal of current efforts by the Health and Environmental Sciences Institute. Toxicology, 335, 11–19. Judson, P.N. (1985) Industrial collaboration in computer aided synthesis design – LHASA UK. Kemia-Kemi, 12 (2), 150–152. Corey, E.J., Wipke, W.T., Cramer, R.D., III, and Howe, W.J. (1972) Computer-assisted synthetic analysis. Facile man–machine communication of chemical structure by interactive computer graphics. Journal of the American Chemical Society, 94, 421–430. Ridings, J.E., Barratt, M.D., Cary, R., Earnshaw, C.G., Eggington, E., Ellis, M.K., Judson, P.N., Langowski, J.J., Marchant, C.A., Payne, M.P., Watson, W.P., and Yih, T.D. (1996) Computer prediction of possible toxic action from chemical structure: an update on the DEREK system. Toxicology, 106, 267–279. Langowski, J.J., Judson, P.N., Patel, M., and Tonnelier, C.A.G., StAR (1997) A knowledge-based computer system for carcinogenic risk assessment, in Animal Alternatives, Welfare and Ethics. Developments in Animal and Veterinary Sciences, vol. 27 (eds L.F.M. Van Zutphen and M. Balls), Elsevier Science, pp. 747–752. Fox, J., Hardman, D., Krause, P., Ayton, P., and Judson, P. (1995) Risk assessment and communication: a cognitive engineering approach, in Applications and Innovations in Expert Systems, vol. 10 (eds A. Macintosh and C. Cooper), Cambridge University Press, Cambridge, pp. 377–390. Stei, P., Smith, D., Pfister, R., Sparrow, S., Rabemampianina, Y., StephanGueldner, M., Stark, C., von Landenberg, F., Lammens, L., Paarup, A.M., Phillips, B., Combes, R., Briggs, K., and Judson, P.

27

28

29

30

31

32

(2007) The excipient database: an initiative of European pharmaceutical companies and Lhasa Limited to refine and avoid duplication of animal studies. Poster presented at the FELASA-ICLAS Joint Meeting, Como, Italy, June 11–14, 2007. Kristam, R., Gillet, V.J., Lewis, R.A., and Thorner, D. (2005) Comparison of conformational analysis techniques to generate pharmacophore hypotheses using Catalyst. Journal of Chemical Information and Modeling, 45, 461–476. Barry Hardy, B., Douglas, N., Helma, C., Rautenberg, M., Jeliazkova, N., Jeliazkov, V., Nikolova, I., Benigni, R., Tcheremenskaia, O., Kramer, S., Girschick, T., Buchwald, F., Wicker, J., Karwath, A., Gütlein, M., Maunz, A., Sarimveis, H., Georgia Melagraki, G., Afantitis, A., Sopasakis, P., Gallagher, D., Poroikov, V., Filimonov, D., Zakharov, A., Lagunin, A., Gloriozova, T., Novikov, S., Natalia Skvortsova, N., Druzhilovsky, D., Chawla, S., Ghosh, I., Surajit Ray, S., Hitesh Patel, H., and Escher, S. (2010) Collaborative development of predictive toxicology applications. Journal of Cheminformatics, 2, 7. Holland, R. (2015) The challenges of compliance for controlled substances in a global marketplace. Scrip Regulatory Affairs, Jan. 9, 2015. Ali, M., Patel, M., Wilkinson, D., Judson, P., Cross, K., and Bower, D. (2013) ToxML, a data exchange standard with content controlled vocabulary used to build better (Q)SAR models. SAR and QSAR in Environmental Research, 24 (6), 429–438. Cross, K.P., Kruhlak, N.L., Myatt, G.J., Stavitskaya, L., and White, A. (2014) Ensuring regulatory acceptable (Q)SAR models and expert alerts for ICH M7 reflect proprietary chemical space. Poster presented at American College of Toxicology 35th Annual Meeting, Nov 9–12th 2014, Orlando, Florida. Valerio, L.G., Balakrishnan, S., Fiszman, M.L., Kozeli, D., Li, M., Moghaddam, S., and Sadrieh, N. (2013) Development of cardiac safety translational tools for QT prolongation and torsade de pointes.

101

102

4 Collaborative Approaches to Advance Chemical Safety

33

34

35

36

37

Expert Opinion on Drug Metabolism and Toxicology, 9 (7), 801–815. Valerio, L.G. and Cross, K.P. (2012) Characterization and validation of an in silico toxicology model to predict the mutagenic potential of drug impurities. Toxicology and Applied Pharmacology, 260 (3), 209–221. Stavitskaya, L., Minnier, B.L., Benz, R.D., and Kruhlak, N.L. (2013) Development of improved QSAR models for predicting AT base pair mutations. Presented at FDA Center for Drug Evaluation and Research (CDER); Genetic Toxicology Association Meeting, University of Delaware, October 2013. Kruhlak, N.L., Cross, K.P., Minnier, B.L., Bower, D.A., and Benz, R.D. (2013) Characterization and application of an external validation set for Salmonella mutagenicity QSAR models using structural fingerprints of known toxicophores. Presented at Society of Toxicology Meeting, San Antonio, USA, March 2013. Long, A. and Valerio, L.G. (2010) The in silico prediction of human-specific metabolites from hepatotoxic drugs. Current Drug Discovery Technologies, 7 (3), 170–187. Benz, R.D., Cayley, A., Drewe, W.C., Kruhlak, N.L., and Surfraz, B. (2014)

38

39

40

41

42

Identification of adverse outcome pathways for the nephrotoxicity of nucleoside and nucleotide anti-viral drug. Society of Toxicology Annual Meeting, Phoenix, USA, March 2014. Surfraz, B., Cayley, A., Patel, M., and Benz, R.D. (2012) Androgen receptor-mediated teratogenicity: use of harvested data towards a transparent expert assay prediction system. Presented at PPTOX 10, Paris, France, May 2012. Arvidson, K.B. (2008) FDA toxicity databases and real-time data entry. Toxicology and Applied Pharmacology, 233 (1), 17–19. CAESAR (2010) CAESAR QSAR Models for REACH. Proceedings of CAESAR Workshop on QSAR Models for REACH, 10–11th March, 2009, Milan, Italy (Chemistry Central Journal, 4, Suppl. 1). Plošnik, A., Zupan, J., and Vračko, M. (2015) Evaluation of toxic endpoints for a set of cosmetic ingredients with CAESAR models. Chemosphere, 120, 492–499. Krumholz, H.M., Ross, J.S., Gross, C.P., Emanuel, E.J., Hodshon, B., Ritchie, J.D., Low, J.B., and Lehman, R. (2013) A historic moment for open science: the Yale university open data access project and Medtronic. Annals of Internal Medicine, 158, 910–911.

103

5 Introduction to Green Analytical Chemistry Marek Tobiszewski

5.1 Introduction 5.1.1 Defining Green Analytical Chemistry

One of the definitions of analytical chemistry states that it is a “scientific discipline that develops and applies methods, instruments, and strategies to obtain quality (bio)chemical information on the composition and nature of matter in space and time” [1]. Green chemistry can be defined as the design of chemical products and processes that reduce or eliminate the use and generation of hazardous substances [2]. The definition of green analytical chemistry (GAC) can be obtained from combining the two aforementioned ones. GAC is an added value to analytical chemistry – performing analytical processes bearing in mind the principles of green chemistry [3]. It stimulates research and is a new direction in the development of analytical chemistry. The principles of green analytical chemistry have been formulated recently and they can be presented with a SIGNIFICANCE mnemonic technique [3] as follows: S – Select direct analytical method I – Integrate analytical processes and operations G – Generate as little waste as possible and treat it properly N – Never waste energy I – Implement automated and miniaturized methods F – Favor reagents from renewable sources I – Increase operator’s safety C – Carry in situ measurements A – Avoid derivatization N – Note that sample number and amount should be minimized

Handbook of Green Chemistry Volume 12: Tools for Green Chemistry, First Edition. Edited by Evan S. Beach and Soumen Kundu.  2017 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2017 by Wiley-VCH Verlag GmbH & Co. KGaA.

104

5 Introduction to Green Analytical Chemistry

C – Choose multianalyte or multiparameter methods E – Eliminate or replace toxic solvents For every analytical target – group of analytes and sample matrix – there are at least several procedures developed. They do differ in terms of analytical performance but also in terms of environmental impact. Analytical procedures that are characterized by better analytical performance require more energetic and material inputs in terms of solvents and reagents, so they are environmentally problematic. Therefore, it is required to choose analytical procedures that fit given purpose [4]. Figure 5.1 summarizes the simple techniques that allow obtaining analytical results with relatively little efforts and material or energetic input. It is needless to apply analytical procedure with extremely low limits of detection if the purpose of the analyses is to perform screening. For some analytical applications it is enough to apply smartphone as stand-alone analytical device or smartphone coupled to another simplified analytical device [5]. Smartphones have been successfully applied as sensors or biosensors to detect glucose in blood, ethyl alcohol in breath, or even mercury in water. Smartphone equipped with application for color analysis was used to determine glucose and urea in human blood. The results of color images of assay plate interpretation were in excellent agreement with the reference method. What is more, the smartphone-based method covers the whole clinically important concentration

Figure 5.1 Parameters of relatively simple analytical techniques that can be successfully applied for certain analytical instead of highly laborious and technologically advances ones.

5.1 Introduction

range [6]. In a similar manner, smartphones can be used to perform determination of cortisol in saliva as the indicator of psychological stress. After simple preparation of test strip and adding four drops of saliva and waiting 10 min for the reaction, the smartphone image is interpreted for brightness that is negatively correlated with cortisol concentration. The method can be applied for cortisol determination within concentration range 0.01–10 ng mL 1 [7]. Smartphone connected to analytical device can be applied in the role of data or results displayer, such as in the case of Bluetooth connection of screen-printed electrodes [8]. 5.1.2 Dualistic Role of Analytical Chemistry in Relation to Green Chemistry

Analytical chemistry, similarly to other branches of chemical sciences, is the subject of certain activities to make it compatible with sustainable development concept. These are various actions that include material and energetic inputs reduction, waste minimization, reduction of analysts’ occupational hazards, solvent consumption reduction or elimination, and reduction of analysis time. The second role of analytical chemistry is defined by 11th principle of green chemistry – real time analysis for pollution prevention. This principle puts strong pressure on the development of analyzers that can work with little or no time delay to obtain analytical results. This requirement can be hardly fulfilled if the matrix of sample is particularly complex and/or the analyte is present in the investigated medium at very low concentration. The analytical devices that are used to perform real time analysis are mainly sensors of different types [9]. There are four modes of location of analytical device in relation to the object of analytical interest [10]. Off-line mode is based on sample collection, its transportation and analysis in laboratory. When the analytical device is taken to the investigated object for the time of measurements, it is measurement in at-line mode. Online location of analytical device is when it is installed at the measurement site and the analyses are made periodically. In-line mode involves installation of the analytical equipment in the investigated medium for the real-time analysis. From green analytical chemistry and chemical safety point of view the latter modes are the preferential ones. 5.1.3 Brief History of Green Analytical Chemistry

The term “green analytical chemistry” was used for the first time in 1999 [2,11] but the environmental impact aspects of performing analytical measurements were discussed even before [12], but not under GAC label. The idea of GAC originates from the sustainable development concept [13]. According to the principle “think globally, act locally,” analysts’ task was to introduce the values

105

106

5 Introduction to Green Analytical Chemistry

of sustainable development to their field. This could not be achieved without rapid progress in miniaturization of sample preparation techniques and analytical devices. In 1980s, the solid phase microextraction, microwave-assisted extraction, and supercritical fluid extraction were developed. In 1990s, molecularly imprinted polymers, liquid–liquid microextraction, and stir bar sorptive extraction were introduced. This was also the time of development of cloud point extraction and pressurized liquid extraction. The beginning of new millennium resulted in numerous improvements and developments of new applications of these analytical techniques. They allowed determining analytes faster, with lower detection limits, higher precision, and lowered negative environmental impact. Legislation also stimulated the development of GAC. Some of the chemicals used in analytical laboratories were banned and new, greener substitutes had to be developed. Clean Air Act (1963) put some of the volatile organic compounds on their lists leading to some limitations on extraction substances applications. Vienna Convention (1985) and Montreal Protocol (1990) banned application of ozone depleting substances, among which 1,1,1-trichloroethane and carbon tetrachloride were used in analytical laboratories. Figure 5.2 presents some ideas of bringing green analytical chemistry into life.

Figure 5.2 Basic ways to fulfill green analytical chemistry concept.

5.2 Greener Analytical Separations

5.2 Greener Analytical Separations 5.2.1 Green Gas Chromatography

Gas chromatography is greener option than liquid chromatography. Gas chromatography requires hydrogen (preferential one), helium, or nitrogen as mobile phases. Mobile phases typically used in liquid chromatography are methanol, acetonitrile, ethanol (all are solvents characterized by certain toxicity), very often as water mixtures. Therefore, gas chromatography should be the first preference over liquid chromatography, not only because of convenience of operation and good separation properties, but also because of environmental reasons. The important factor that influences the greenness of gas chromatographic separations is the energy consumption during oven heating [14]. The technology of low thermal mass allows for significant energy savings. Huge heating and cooling speed can be obtained and the system consumes ∼1% of energy compared to the traditional ones [15]. 5.2.2 Greener Liquid Chromatography

The main problem with liquid chromatography is the need to use high purity solvents as mobile phases. It was established that 34 million liters of mobile phase is used globally during single year [16]. There are three basic ways to deal with the amount of wastes generated by consumption of mobile phase. The first one is application of more benign solvents, such as ethanol and more importantly water. In this way, the generated wastes are less hazardous but they are not reduced in volume. The general trend in liquid chromatography is to apply reversed phase (mobile phase is more polar than stationary phase) rather than normal phase (mobile phase is more nonpolar than stationary phase, usually toxic solvents are applied as mobile phase) separations. In reversed-phase liquid chromatography, acetonitrile can be successfully substituted with greener ethanol. In terms of chromatographic separations, acetonitrile is characterized by considerably better performance, while ethanol performs quite well for some applications. The most important green constituent of mobile phase is water. However, it might be problematic when it comes to disposal of mobile phase. During recovery of organic constituents by distillation, azeotropes can be formed and it might be impossible to incinerate mobile phase due to high water content. The second one is recycling of the mobile phase, to reuse it as a mobile phase once more if isocratic elution is applied [17] or to apply it for different purpose. In isocratic elution, recycled, reused mobile phase is no more usable when the concentration of analyte approaches the lowest expected concentration in the sample. It has been proved that for many applications the mobile phase does not have to be extremely pure. For example, vodka was applied instead of ethanol

107

108

5 Introduction to Green Analytical Chemistry

water mixture with similar performance but worse baseline was obtained [18]. Another option is to use automatic valve that is used to separate “clean” mobile phase from mobile phase containing analytes. The third approach is reduction of analytical column dimensions that results in lower solvent consumption. The most important column parameter that contributes to reduction of mobile phase consumption is internal diameter. Reduced internal diameter also contributes to higher sensitivities, as analytes are less diluted in the mobile phase. On the other hand, column loses some of its efficiency and lower resolutions are observed. Nanoliquid chromatographic columns have been developed, with internal diameters lower than 0.1 mm. In such a system, the mobile phase flow rates are at levels of 10 4–10 2 μl min 1 [19]. Introduction of nanocolumns requires application of dedicated pumps that operate within required flow rates. Another way to reduce mobile phase consumption is to apply particles of stationary phase of finer particle size. Particle size influences the efficiency of column, columns with finer particles can be shorter and the efficiency is maintained. 5.2.3 Supercritical Fluid Chromatography

The special substitution in mobile phase with greener one is application of supercritical fluid chromatography. In such a system usually carbon dioxide in supercritical state (scCO2) is applied. This type of chromatography is characterized by excellent selectivity and low consumption of organic solvents. Polar organic solvents are sometimes added to mobile phase as polarity modifiers [20]. Supercritical fluid chromatography can be applied to separate some compounds that are problematic in liquid chromatography. Supercritical fluid chromatography is successfully applied to separate chiral substances. This can be achieved in two ways. The more common one is application of chiral stationary phase, while the other, less common one, is application of enantioselective mobile phase modifiers with application of traditional stationary phases [21].

5.3 Green Sample Preparation Techniques and Direct Techniques 5.3.1 Direct Analytical Methods

Sample preparation step is concerned as the most environmentally problematic from the whole analytical process. It requires organic, often nonpolar, though toxic, solvents. It also requires other material and energetic inputs. The aim of the sample preparation step is to change the sample matrix to the one compatible with analytical device, to enrich analytes in the sample (increase the

5.3 Green Sample Preparation Techniques and Direct Techniques

concentration to be more easily detected), and also often to remove any interfering substances. If the sample is characterized by relatively simple matrix composition and analytes are at high concentrations, direct methods can be applied. In such a case, sample is collected and then it is analyzed in unmodified state. There are relatively many direct spectroscopic methods [22]. These include X-ray fluorescence spectrometers, atomic absorption spectrometers, and atomic emission spectrometers. The details of spectroscopic methods and technical solutions strongly influence the possibility to apply spectroscopy directly. For example, flame atomic absorption spectroscopy allows for direct analysis of liquid samples, while solid ones require mineralization as a sample treatment step. Electrothermal atomic absorption spectroscopy allows applying direct analysis approach to liquid and solid samples. In fact, thermal mineralization of small mass sample, in such a case is performed in graphite furnace after sample injection. It is much harder to apply chromatographic direct methods but there are also some developed methods [23]. Direct gas chromatographic methods were developed for determination of volatile organic compounds in water samples [24], methanol in wine and whiskey [25], and MTBE in gasoline [26]. Liquid chromatographic methods without sample preparation were developed for the determination of herbicides in water samples [27] or phenols in olive oil [28]. Apart from the lack of sample preparation step and related consumption of solvents and reagents direct techniques allow obtaining better precision. On the other hand, no analytes enrichment during sample preparation results in relatively high procedural detection limits. Most of the applications of chromatographic techniques and many of spectroscopic ones unfortunately require some kind of sample preparation. 5.3.2 Microextraction Sample Preparation Techniques

Great numbers of the sample matrices require analytes to be extracted before the analysis. The first and the most easily applied technique is solvent extraction. Historically, frequently applied solvents were dichloromethane, chloroform, carbon tetrachloride, tetrachloroethene, benzene, and other nonpolar hydrocarbons or halocarbons. These solvents are characterized by good extraction efficiency but they are toxic to humans via digestion and inhalation tracks, toxic to aquatic ecosystems, some of them are carcinogenic. Extraction with liquids is rather hard to be automated, which results in high occupational exposure of analytical chemists. Therefore, there was a strong motivation for searching new, environment-friendly techniques. Some efforts are made to incorporate less toxic ones and solvents originating from renewable feedstocks [29], or natural deep eutectic solvents [30]. There are also many techniques developed that can substitute liquid extractions. Only the most commonly used will be discussed further, some techniques involving minor improvements will be omitted.

109

110

5 Introduction to Green Analytical Chemistry

Figure 5.3 The scheme of SPME sample preparation setup: (a) Direct injection mode. (b) Headspace mode.

5.3.2.1

Solid-Phase Microextraction

Solid-phase microextraction (SPME) is the technique that utilizes fiber covered with stationary phase that allows for sorption of analytes. The two basic modes to SPME include adsorption and absorption modes. SPME has been developed by Pawliszyn and co-workers in 1992 [31]. The principle of operation involves the extraction of the fraction of analyte present in the sample by the stationary phase of the fiber. After exposition of the fiber to the sample (see Figure 5.3), the fiber is inserted to the injector of the chromatograph and thermal desorption of the analytes takes place. The technique does not involve utilization of any solvents, so it is concerned as green. Many technical solutions have been developed in terms of SPME fibers coating. This allowed for great variety of analytes to be determined by this technique. The fibers can be exposed to gaseous and liquid samples, and also they can be introduced to biological tissues that greatly increase the area of applications. Recently, there is huge development seen in terms of stationary phases, including applications of ionic liquids [32] and polar phases soluble in water that are protected with nonpolar membranes [33]. There is also rapid development in terms of analytes determined and sample matrices that are involved in SPME applications. 5.3.2.2

Liquid-Phase Microextraction

The efforts in miniaturization of liquid phase extraction have also been made. There are two basic modes of liquid phase extraction – single drop microextraction (SDME) and hollow fiber liquid phase microextraction (HF-LPME). The first of the methods involves exposition of one drop of organic solvent placed at the tip of the chromatographic needle to the liquid sample. After the extraction, the drop of solvent is sucked into syringe and can be injected into the chromatograph. As only few microliters of organic solvent are consumed during the analysis, the

5.3 Green Sample Preparation Techniques and Direct Techniques

technique is considered as green [34]. One of the main problems related to this technique is poor stability of the organic solvent drop. The droplet can fall of the tip of syringe needle that results in the need to repeat the analysis. On the other hand, organic solvent may undergo dissolution in sample that results in erroneous results. Additional internal standard compound is added to solvent droplet to establish solvent volume at the moment of injection to analytical device. The approach involving HF-LPME overcomes the problem of organic solvent stability. In this case, the solvent is trapped inside polymeric porous fiber. Both problems of dissolution and lack of stability are overcome. Similar to SDME, this technique is considered as green as it consumes microliter amounts of organic solvents. After the extraction solvent can be easily recovered from the hollow fiber and injected to analytical device. 5.3.2.3

Dispersive Liquid–Liquid Microextraction

Different approach to liquid phase extraction is presented by dispersive liquid– liquid microextraction (DLLME). The main problem of SDME and HF-LPME is that the extraction takes too much time. It is due to limited transfer of analytes, because of small boundary between organic and aqueous phases. The system of two solvents is applied in DLLME to obtain extraction in one or few minutes. The first solvent applied is extraction one that has to be immiscible with water. The second one is dispersive solvent that has to be miscible with both extraction solvent and water-based sample. After the injection of extraction and dispersive solvents mixture to the sample, rapid dissolution of dispersive solvent takes place and extraction one is highly dispersed in the liquid sample. The extraction solvent can be recovered after centrifugation (see Figure 5.4). The extraction solvents frequently used in DLLME are chlorobenzene, toluene, aliphatic hydrocarbons, and also long chain alcohols. The typical required volume of this solvent is from few to 80 μl. Three most commonly used dispersive solvents are methanol,

Figure 5.4 The scheme of dispersive liquid–liquid microextraction.

111

112

5 Introduction to Green Analytical Chemistry

acetonitrile, and acetone. The typically reported volumes of disperser solvent are within range 0.4–1 ml [35]. It is worth to mention one interesting solution involving the long chain alcohols with melting points close to room temperature. In this case, the separation of extraction solvent is performed by lowering the temperature and removal of the solid droplet with tweezers. 5.3.3 Stir Bar Sorptive Extraction

Another sample preparation technique is stir bar sorptive extraction (SBSE) developed by Sandra. The procedure is based on application of magnetic stirrer covered with stationary phase capable of absorption of analytes. The sorptive phase can be selected according to the polarity to maximize sorption efficiency of target analytes. After the extraction, magnetic stirrer is removed from the sample solution and the analytes are desorbed in one of two methods. The first one, less green, is desorption with organic solvent. The second one, definitely greener, is thermal desorption of analytes in the injection port of chromatograph. The main advantages of SBSE are excellent enrichment factors that are obtainable and therefore, good performance at ultratrace analytes levels [36]. There are numerous applications of SBSE technique but mainly used in determination of nonpolar compounds as polar ones are characterized by worse performance. 5.3.4 Supercritical Fluid Analytical Extraction

Supercritical fluids, apart from being used to extract valuable compounds from plant material in the industrial scale, are used in sample preparation in analytical chemistry. This type of extraction has been used to extract analytes from biological samples, soil, sediments, and food samples. The most commonly used solvent in supercritical fluid extraction is CO2. The other solvents that can achieve their supercritical state and can potentially be applied are N2O, pentane, ethane, propane, and sulfur hexafluoride of fluoroform. The compound of the first choice is CO2 as it is nontoxic, nonflammable, cheap, and easy to be obtained in high purity. Its supercritical state conditions are relatively mild and therefore, easy to be obtained [37]. It is possible to develop targeted extractions with supercritical fluids, as the solubility of target compounds depend on the temperature and pressure of the fluid [38]. Since scCO2 is concerned as nonpolar solvent, it is common to apply polar organic cosolvents (i.e., methanol) as polarity moderators if analytes are of more polar nature. 5.3.5 Microwave- and Ultrasound-Assisted Extraction

Microwave- and ultrasound-assisted extractions are also considered as greener techniques. It should be noted that these sample preparation techniques do not

5.3 Green Sample Preparation Techniques and Direct Techniques

allow eliminating the application of solvents. Both microwaves and ultrasounds stimulate the extraction efficiency and allow obtaining desired extraction efficiency in shorter time. As a consequence, less of organic solvents can be used, applied solvents do not have to be characterized by very strong extraction potential, so less toxic solvents can be applied. Another aspect is that reduced extraction time allows for energy savings. Ultrasound-assisted extraction is attractive alternative in relation to traditional extraction, not only because of good efficiencies and reproducibilities, but also due to reasonable costs of the equipment and its convenient use [39]. Ultrasound-assisted extraction apart from organic pollutants is applied for nanoparticles extraction from environmental samples [40]. It is also used to enhance digestion processes before elemental analysis. It is compatible with enzymatic digestion, which is its advantage [41]. Microwave-assisted extraction can be operated in solvent-free mode and it is applied to extract compounds from solid biological samples. The requirement is that the sample should be characterized by some water content. The increase of temperature due to microwave radiation results in expansion of plant cells and destruction of cell walls that allows for extraction of the compounds [42]. Microwave-assisted extraction was successfully applied to extract variety of organic compounds from different sample matrices [43]. 5.3.6 Ionic Liquids in Extraction

Ionic liquids (ILs) have gained much attention as green solvents. This attention is also clearly visible in analytical chemistry [44]. As extraction solvents they have the advantage to be selected from their great variety to maximize the extraction efficiency of target groups of compounds. They are applied in all aforementioned microextraction techniques. They are applied in HF-LPME and SDME, where they are advantageous by their low volatility (when extraction is performed in headspace) and low solubility in water (when extraction is in direct immersion mode). They are applied in DLLME as the extraction solvents, where they are injected to sample in such a cation–anion combination that makes IL water-soluble. Then a solution, containing anion that in combination with cation makes insoluble IL, is injected. Newly formed insoluble IL is highly dispersed in liquid sample and rapid analytes extraction takes place. It should be noted that ILs used in liquid extraction techniques, due to their negligible volatility, are incompatible with gas chromatographs without modification of the injectors. This problem does not occur if ILs are applied as fiber coatings in SPME. They need to be polymerized and bonded to the surface of fiber to withstand the high temperatures in the injectors and to be stable during extensive mixing of liquid sample during extraction. Their unique properties allow performing task-specific extractions, if ILs are properly selected toward given groups of analytes [45]. There are some applications of ILs as modifiers of mobile phases in liquid chromatography. They allow modifying retention times of eluted compounds

113

114

5 Introduction to Green Analytical Chemistry

and allow obtaining better peak performance [46]. However, the application of ILs as mobile phase modifiers is not common so far. More common application is their incorporation as matrix in MALDI (matrix-assisted laser desorption ionization) mass spectrometry. They easily fulfill two requirements toward matrix – low volatility in the vacuum conditions and containing chromophore to absorb part of the laser radiation. Recently, the greenness of ILs has been questioned. Some of ILs are toxic to humans and aquatic life, some are flammable or even explosive. ILs lost the undoubted status of being green solvents, however, they are still attractive to analytical chemists because of nonvolatility and tunable properties.

5.4 Chemometrics for Signals Processing

Quite another approach to analysis of sample constituents is not separation of compounds of interest from the sample matrix but separation of signals corresponding to these compounds from relatively complex datasets using multivariate statistics [47]. What is important is, nondestructive sample analysis can be done by means of infrared, X-ray fluorescence, or Raman spectroscopy. With chemometrics it is relatively easy to obtain qualitative information about sample composition, more complex algorithms are needed to obtain quantitative information. Moros et al. [48] list the most important features that allow considering spectroscopic methods supported by chemometrics as green ones [48]. These are significantly reduced risk related to sample preparation, replacement of many analytical instruments into a single one, lowered costs of analysis, and obtaining similar analytical performance without environmental negative effects. The procedure based on mid-infrared spectroscopy and chemometric algorithm allows determining benzo[a]pyrene in cigarette smoke [49]. Traditional procedure would be based on isolation of benzo[a]pyrene by sorbent, its release by solvent extraction, and GC or LC analysis. Near-infrared spectroscopic procedure was developed to determine adulterants in starch. It was successfully used to detect kudzu starch falsifications with cheaper starches or talcum powder [50]. Chemometrics used to resolve near-infrared spectra was applied to determine fats, proteins, and sodium chloride concentrations in cheese samples. The results were in acceptable agreement with reference method measured values. The proposed methodology substitutes Kjeldahl method for protein determination, gravimetric method for fat content determination, and titration for chlorides determination [51].

5.5 Conclusions

Green analytical chemistry is rapidly developing branch of sustainability sciences. There are numerous ways to make analytical procedures greener and this

References

introduction to green analytical chemistry overviews only the some major ones. All of the steps of analytical procedures are subject to reduction of their environmental impact. The most important strategies, such as miniaturization, application of greener solvents, and reagents, allow for significant reduction of hazards related to the analysts’ activities.

References 1 Kellner, R., Mermet, J.M., Otto, M.,

2 3

4 5

6

7

8

9

10

11 12

13

14

Valcárcel, M., and Widmer, H.M. (eds) (2004) Analytical Chemistry, 2nd edn, Wiley-VCH Verlag GmbH, Weinheim, Germany, Chapter 1. Anastas, P.T. (1999) Critical Reviews in Analytical Chemistry, 29, 167. Gałuszka, A., Migaszewski, Z., and Namieśnik, J. (2013) Trends in Analytical Chemistry, 50, 78. Koel, M. (2016) Greem Chemistry, 18, 923–931. Roda, A., Michelini, E., Zangheri, M., Di Fusco, M., Calabria, D., and Simoni, P. (2016) Trends in Analytical Chemistry, 79, 317–325. Wu, Y., Boonloed, A., Sleszynski, N., Koesdjojo, M., Armstrong, C., Bracha, S., and Remcho, V.T. (2015) Clinica Chimica Acta, 448, 133. Choi, S., Kim, S., Yang, J.-S., Lee, J.-H., Joo, C., and Jung, H.-I. (2014) Sensing and BioSensing Research, 2, 8. Zhang, D., Jiang, J., Chen, J., Zhang, Q., Lu, Y., Yao, Y., Li, S., Liu, G.L., and Liu, Q. (2015) Biosensors and Bioelectronics, 70, 81–88. Barsan, M.M., Ghica, M.E., and Brett, C.M.A. (2015) Analytica Chimica Acta, 881, 1. Tobiszewski, M., Mechlińska, A., and Namieśnik, J. (2010) Chemical Society Reviews, 39, 2869. Namieśnik, J. (1999) Environmental Science and Pollution Research, 6, 243. de la Guardia, M., Khalaf, K.D., Carbonell, V., and Morales-Rubio, A. (1995) Analytica Chimica Acta, 308, 462. Holden, E., Linnerud, K., and Banister, D. (2014) Global Environmental Change, 26, 130. Płotka, J., Tobiszewski, M., Sulej, A.M., Kupska, M., Górecki, T., and Namieśnik, J.

15

16

17

18

19

20

21

22

23 24

25

26

27

(2013) Journal of Chromatography A, 1307, 1. Luong, J., Gras, R., Mustacich, R., and Cortes, H. (2006) Journal of Chromatographic Science, 44, 253. Gaber, Y., Tornvall, U., Kumar, M.A., Amin, M.A., and Hatti-Kaul, R. (2011) Greem Chemistry, 13, 2021. Garrigues, S., Armenta, S., and de la Guardia, M. (2010) Trends in Analytical Chemistry, 29, 592. Welch, C.J., Wu, N., Biba, M., Hartman, R., Brkovic, T., Gong, X., Helmy, R., Schafer, W., Cuff, J., Pirzada, Z., and Zhou, L. (2010) Trends in Analytical Chemistry, 29, 667. Sandra, P., Vanhoenacker, G., David, F., Sandra, K., and Pereira, A. (2010) LC GC Europe, 23, 1. Bernal, J.L., Martín, M.T., and Toribio, L. (2013) Journal of Chromatography A, 1313, 24. Kalíková, K., Slechtová, T., Vozka, J., and Tesarová, E. (2014) Analytica Chimica Acta, 821, 1. Bendicho, C., Lavilla, I., and Pena-Pereira, F., and Romero, V. (2012) Journal of Analytical Atomic Spectrometry, 27, 1831. Tobiszewski, M. and Namieśnik, J. (2012) Trends in Analytical Chemistry, 35, 67–73. Tobiszewski, M. and Namieśnik, J. (2011) Analytical and Bioanalytical Chemistry, 399, 3565. Wang, M.-L., Wang, J.-T., and Choong, Y.-M. (2004) Journal of Food Composition and Analysis, 17, 187. Perez Pavon, J.L., del Nogal Sanchez, M., Garcıa Pinto, C., Fernandez Laespada, M.E., and Moreno Cordero, B. (2004) Journal of Chromatography A, 1048, 133. Huang, S.-B., Mayer, T.J., and Yokley, R.A. (2008) Journal of Agricultural and Food Chemistry, 56, 2595.

115

116

5 Introduction to Green Analytical Chemistry 28 Jimenez, M.S., Velarte, R., and Castillo, J.R. 29

30

31

32

33

34

35

36 37

38

39 40

(2007) Food Chemistry, 100, 8. Pena-Pereira, F., Kloskowski, A., and Namieśnik, J. (2015) Greem Chemistry, 17, 3687. Espino, M., de los Ángeles Fernández, M., Gomez, F.J.V., and Silva, M.F. (2016) Trends in Analytical Chemistry 76, 126–136. Arthur, C.L., Killam, L.M., Buchholz, K.D., Pawliszyn, J., and Berg, J.R. (1992) Analytical Chemistry, 64 (17), 1960. Ho, T.D., Canestraro, A.J., and Anderson, J.L. (2011) Analytica Chimica Acta, 695, 18. Kloskowski, A., Pilarczyk, M., and Namieśnik, J. (2009) Analytical Chemistry, 81, 7363. Spietelun, A., Marcinkowski, Ł., de la Guardia, M., and Namieśnik, J. (2014) Talanta, 119, 34. Zgoła-Grześkowiak, A. and Grześkowiak, T. (2011) Trends in Analytical Chemistry, 30, 1382. Nogueira, J.M.F. (2012) Analytica Chimica Acta, 757, 1. Zougagh, M., Valcarcel, M., and Rıos, A. (2004) Trends in Analytical Chemistry, 23, 399. Herrero, M., Castro-Puyana, M., Mendiola, J.A., and Ibanez, E. (2013) Trends in Analytical Chemistry, 43, 68. Pico, Y. (2013) Trends in Analytical Chemistry, 43, 84. Albero, B., Sanchez-Brunete, C., GarciaValcarcel, A.I., Perez, R.A., and Tadeo, J.L.

41

42

43

44 45 46

47

48

49

50

51

(2015) Trends in Analytical Chemistry, 71, 110. Bendicho, C., De La Calle, I., Pena, F., Costas, M., Cabaleiro, N., and Lavilla, I. (2012) Trends in Analytical Chemistry, 31, 50. Li, Y., Fabiano-Tixier, A.S., Abert Vian, M., and Chemat, F. (2013) Trends in Analytical Chemistry, 47, 1. Sanchez-Prado, L., Garcia-Jares, C., Dagnac, T., and Llompart, M. (2015) Trends in Analytical Chemistry, 71, 119. Sun, P. and Armstrong, D.W. (2010) Analytica Chimica Acta, 661, 1. Tan, Z.-q., Liu, J.-f., and Pang, L. (2012) Trends in Analytical Chemistry, 39, 218. García-Alvarez-Coque, M.C., Ruiz-Angel, M.J., Berthod, A., and Carda-Broch, S. (2015) Analytica Chimica Acta, 883, 1. Gredilla, A., Fdez-Ortiz de Vallejuelo, S., Elejoste, N., de Diego, A., and Madariaga, J.M. (2016) Trends in Analytical Chemistry, 76, 30–39. Moros, J., Garrigues, S., and de la Guardia, M. (2010) Trends in Analytical Chemistry, 29, 578. Zhang, Y., Zou, H.-Y., Shi, P., Yang, Q., Tang, L.-J., Jiang, J.-H., Wu, H.-L., and Yu, R.-Q. (2016) Analytica Chimica Acta, 902, 43. Xu, L., Shi, W., Cai, C.-B., Zhong, W., and Tu, K. (2015) LWT - Food Science and Technology, 61, 590. Pi, F., Shinzawa, H., Ozaki, Y., and Han, D. (2009) International Dairy Journal, 19, 624.

117

6 Cosmo-RS-Assisted Solvent Screening for Green Extraction of Natural Products Anne-Gaëlle Sicaire, Aurore Filly, Maryline Vian, Anne-Sylvie Fabiano-Tixier, and Farid Chemat

6.1 Introduction

Solvents play an important role in great number of unit operations in chemistry and chemical engineering. In fact, nowadays there is no production process in perfume, cosmetic, pharmaceutical, food ingredients, nutraceuticals, biofuel, or fine chemicals industries that do not use a solvent step [1]. Solvents can be used as diluents or additives in paints and inks, as cleaning agents, or solvents for synthesis and extraction. Figure 6.1 [2] shows the distribution of the solvent market in Europe according to the business sector. The problem of commonly used solvents is their negative impact on HSE (Health, Safety, and Environment) as most solvents currently available on the world market and come from the petrochemical industry are volatile organic compounds (VOCs). In 2009, the global market of chemicals compounds represented around $100 billion and it is estimated to reach $3000 billion in 2025 [3]. Only 3% of these chemicals are obtained from renewable resources, after chemical processing, fermentation, or enzymatic conversion. This share is predicted to attain around 15% by 2025 [3]. With the geopolitical environment (oil prices increase and decrease of reserves), societal demand for more sustainable products, and the arrival of the new regulations (REACH, European guidelines, etc.), much interest has been given to the development of new eco-friendly ways of replacing conventional solvents. In other words, alternatives to petrochemical solvents have to fulfill the principles of green chemistry (Figure 6.2). Finally, new technologies such as solvent-free method, aqueous formulations, microwaves, ultrasound, or alternative solvents appear to be good candidates. Among these solutions, the use of greener solvents, such as bio-based solvents, constitutes one of the most important alternative routes for the substitution of petrochemical solvents. Pharmaceutical industries such as GSK [5], Sanofi [6], or Pfizer [7] developed their own solvent selection guides that provide technical data and clear instructions for the development of more sustainable processes. This allows reporting the involvement and commitment of industries in the

Handbook of Green Chemistry Volume 12: Tools for Green Chemistry, First Edition. Edited by Evan S. Beach and Soumen Kundu.  2017 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2017 by Wiley-VCH Verlag GmbH & Co. KGaA.

118

6 Cosmo-RS-Assisted Solvent Screening for Green Extraction of Natural Products

Figure 6.1 Market allocation of solvents by industries in Europe.

investigation on greener alternative solvents and proves their concern for the integration of the sustainable development approach. In the field of extraction of natural products, extraction using petroleumsourced solvents (mostly hexane) remains the most commonly practiced method. With the global context around fossil resources and COV emissions, the investigation on the substitution of these petroleum solvents has become an important issue for the research in the extraction field; several examples of potential alternatives to petroleum solvents for the extraction of various natural products can be found in the literature [1,8–11]. Moreover, theoretical approach such as COSMO-RS is now a useful tool in the investigation of potential candidates for the replacement of petroleum solvents.

Figure 6.2 The 12 principles of green chemistry [4].

6.2 Solvents for Green Extraction

In this chapter, some definitions and general information about the solvents will be given as well as some essential concepts concerning solute–solvent mixtures will provided in order to understand and undertake the substitution process of solvents used in extraction. An overview of potential alternative solvents will then be given, and finally the COSMO-RS approach for the prediction of physicochemical properties will be introduced with examples of application for the extraction of oils and aromas.

6.2 Solvents for Green Extraction 6.2.1 Definition

A solvent is defined as “a liquid that has the property to dissolve, dilute or extract other materials without causing chemical modification of these substances or itself. The solvents are able to implement, apply, clean or separate products” [12]. There is a considerable variety of solvents, including water on one hand and organic solvents on the other. These solvents are grouped according to their chemical nature and thus form different families such as aromatic hydrocarbons (benzene, toluene, xylene, etc.), nonaromatic hydrocarbons (alkanes, alkenes, etc.), alcohols (methanol, ethanol, propanol, etc.), ketones (acetone, methylethylketone, etc.), esters (acetates etc.), ethers (ethyl ether, tetrahydrofuran, dioxane, etc.), glycol ethers, halogenated hydrocarbons (chlorinated, brominated, or fluorinated derivatives), amines, and amides. Among these compounds, we can also define a class of oxygenated solvents consisting of alcohols, ketones, and so on. From a macroscopic viewpoint, a solvent is a continuum characterized by macroscopic physical constants [13] – boiling point, melting point, vapor pressure, relative permittivity, thermal conductivity, surface tension, density, viscosity, refractive index, etc., whereas from a microscopic viewpoint, it is a discontinuum that consists of individual solvent-interacting molecules characterized by molecular properties (dipole moment, electronic polarizability, hydrogen bond donor or acceptor character, electron donor or acceptor character, etc.). These different properties are at the origin of solute–solvent interactions during the dissolution process. 6.2.2 Solute–Solvent Interaction

During the dissolution of a solute in a solvent, cohesive forces holding solvent and solute molecules together are disturbed and new solute–solvent interactions appear to form a homogeneous and stable solution. From a thermodynamic viewpoint, this phenomenon involves two types of energy: first, an enthalpy of

119

120

6 Cosmo-RS-Assisted Solvent Screening for Green Extraction of Natural Products

dissolution, corresponding to the balance between bonds rupture (solute–solute and solvent–solvent) and bonds formation (solute-solvent); second, an entropy of dissolution, resulting from the disorder caused by the dispersion of the solute in the solvent. However, to make the process of dissolution possible, it is necessary that the solvent and solute molecules have enough affinity so that they can be distributed uniformly in the continuum formed by the solvent. Cohesion forces of both species have to be similar so that the solution spontaneously takes place (or with very little energy). Knowing the intensity of the cohesion forces is essential to be able to effectively select a solvent for a specific application. The polarity of the solvent is one of the key properties resulting from these forces. According to IUPAC comity [14], the solvent polarity is described as the global ability of solvation of a compound, depending on the action of all specific (hydrogen donnor/acceptor, electron donnor/acceptor) and nonspecific (London, Debye, Keesom, and Coulomb forces) interactions between a solute and a solvent, except interactions leading to definitive alteration of the solute. 6.2.3 Substitution Concept

To replace potentially harmful petroleum solvent and limit VOC emissions, a preventive approach can be considered. This approach consists in using innovative technologies such as ultrasound and microwaves, solvent-free methods, aqueous formulations, and supercritical fluids or in finding less toxic alternative solvents. Replacing a solvent by another does not necessarily mean eliminating all the hazards and issues related to the implementation of a process. Indeed, the modification of a process is generally associated with new risks. Precautions should therefore be taken into account in the selection of an alternative solvent, for example, technical, environmental, or sanitary criteria, the ecocompatibility of the process, and the price of the solvent besides the technical criteria related to the properties of the solvent such as the solvation power [15]. The solvation power is a key criterion that can be evaluated using various methods such as the Kauri-butanol index, Kamlet–Taft scale, or Hildebrand and Hansen solubility parameters; however, thanks to a much more powerful tool COSMO-RS that can also be used as a real decision tool for the choice of alternative solvents. To sum up, an ideal alternative solvent must meet the following criteria:

      

Should not emit VOC Be of low toxicity for humans Have a limited impact on environment (be eco-friendly) Can be obtained from renewable resources Have a high solvation power Be easy to recover Not significantly change a process setup.

6.2 Solvents for Green Extraction

6.2.4 Panorama of Alternative Solvents for Extraction 6.2.4.1

Water: Solvent with Variable Polarity

Water is recognized as the safest and cheapest solvent for the extraction of natural products. It is an abundant resource, nontoxic, and nonflammable. It allows the solubilization of a large range of polar molecules but also apolar compounds under certain conditions. As the distribution of molecules in a fluid is governed by the interaction energy between them, the nature of these interactions is dependent on the molecular geometry and charge distribution. In the case of water, the hydrogen bond mainly contributes to the interaction energy with an electrostatic attraction force of 8–40 kJ mol 1 between the hydrogen and oxygen. The cohesion of the resulting network is responsible among other things at high surface tension and high boiling point [16]. At room temperature, the dielectric constant of water (εr = 78.3) is much higher than that of ethanol (εr = 24.55) or n-hexane (εr = 1.88), which are two common solvents used in extraction. Therefore, water is a highly polar solvent at ambient conditions. However, when the temperature and pressure increase, the dielectric constant rapidly decreases due to the rupture of the hydrogen bonding network. The physicochemical properties of the liquid water allow hydration of the polar molecules, nonpolar and ionic as in the case of subcritical water. In addition, cosolvents, hydrotropes, or surfactants [17] significantly alter the properties of water. Thus, aggregates or structures such as micelles and liquid crystals form, which has an impact on the solubility and the extractive capacity of aqueous solutions. 6.2.4.2

Bio-Based Solvents

Defined as solvents coming from at least some renewable raw materials, biobased solvents have the advantage of offering a positive impact on environmental and health (no emission of volatile organic compounds, biodegradable, and nontoxic). Most of the currently available bio-based solvents are principally used in organic synthesis or in the formulation field, but can be promising candidates as solvents for extraction application. Organic Acid Esters: Ethyl Lactate, Ethyl Acetate

Organic acid esters are derived, as the name indicates, from the esterification of organic acids such as acetic, citric, gluconic, or lactic acid. Ethyl lactate and ethyl acetate are examples of these organic acid esters. Ethyl lactate can directly be obtained by corn fermentation or by the esterification of lactic acid obtained from corn starch [18]. It is 100% biodegradable, easy to recycle, noncorrosive, noncarcinogenic, and nontoxic, and it is approved by FDA as food aroma additive. It has a high boiling point, low vapor pressure, low surface tension, and a high solvation power. Because of these various advantages, ethyl lactate is being used in different fields such as pharmacy, perfumes, inks, coatings, or food industry [19–21].

121

122

6 Cosmo-RS-Assisted Solvent Screening for Green Extraction of Natural Products

Traditionally, ethyl acetate resulted from the esterification of ethanol and acetic acid. Nowadays, it can be obtained through dehydrogenation of bioethanol [22,23]. The advantage of this new process is the use of only one raw material, bioethanol, principally produced by fermentation. It is a biodegradable solvent largely used in various industrial applications such as decaffeination, adhesives, cleaning products, or paint and coatings, thanks to its high dilution rate of aliphatic and aromatic compounds. It is nevertheless classified as highly flammable. Fatty Acid Esters

Fatty acid esters are vegetable oil derivatives (from rapeseed or sunflower, for example) transesterified from an alcohol such as methanol, ethanol, or 2-ethoxyhexanol. The most prevalent fatty acid ester in terms of volume is Diester® (methyl ester of rapeseed oil) and it is used as biofuel. These kinds of compounds are also used in formulations of ink, pesticides, or cleaning products. Alcohols: Ethanol and Fusel Alcohols

Bioethanol, or ethanol from natural origin, is sourced from the fermentation of sugars (from sugar cane or sugar beet), from enzymatic hydrolysis of starch (from corn or wheat) or from lignocellulosic raw material (wood, herbs, agricultural waste) [24]. Its production significantly reduces the emission of greenhouse gases that characterizes the combustion of conventional fuels. In addition, it is also biodegradable and less toxic than many petroleum-based solvents. It can be used as biofuel and it appears in the composition of nail polish, for example. Moreover ethanol is a well-known solvent for plant extraction [25], thereby bioethanol can be immediately used in these processes. Fusel alcohols are a mixture of alcohols obtained after the distillation of bioethanol. Fusel alcohols are currently mostly used as energy source but they can be precursor of acetates, like isoamyle acetate, that appeared to be a good solvent as nail polish remover [15,26]. Terpenes

Terpenes represent a broad class of organic compounds of natural origin that are found in essential oils, oleoresins, fruit, and herbs. They can be isolated from vegetal raw materials by physical methods such as hydrodistillation or cold pressing in the case of citrus peel oils. Terpenes are found in many products such as perfumes, pharmaceuticals, dyes, food, and so on where they are mainly used for their fragrant power. Nevertheless terpenes, because of their abundance, can also be used as solvents. The most commonly used terpene as a solvent is probably limonene, an important byproduct of the citrus industry [27,28]. Terpenes have already been used as solvents in the extraction of several products such as oil from microalgae [29], aromas [10], vegetable oils [30], and so on.

6.2 Solvents for Green Extraction

Synthetic Biobased Solvents, Furfural Derivatives

In the context of biorefinery and recovery of the whole plant, lignocellulosic residues from cereal production can be exploited to produce furfural. Furfural is the precursor of many molecules such as 2-methyltetrahydrofuran (MeTHF). MeTHF is synthetized by hydrogenation of products obtained from carbohydrate fractions of hemicellulose from various feedstocks such as corn cobs or sugar cane bagasse [31]. MeTHF is a commonly used solvent in organic synthesis [32–34], but thanks to its biodegradability, promising environment footprint, easy of recycling, and good solvation power it appears as good potential alternative for the replacement of petroleum solvent in the extraction field like in the replacement of hexane in the extraction of vegetable oils [11], aromas [10], or carotenoids [35]. 6.2.4.3

Solvent Obtained from Chemical Synthesis

More healthy and environment-friendly solvents are not always from plant origin, they can be derived from by-products of the petrochemical industry or be obtained by “green” chemical synthesis as for example dimethyl carbonate (DMC) or cyclopentyl methl ether (CPME). DMC, traditionally resulting in the reaction of phosgene, methanol, and methyl chloroformiate, is nowadays obtained by transesterification of propylene carbonate or by reaction of carbon monoxide, methanol, and oxygen. Thanks to nontoxicity and biodegradability properties, DMC is recognized as a “green” solvent. Examples of studies using DMC for biodiesel production can be found in literature [36,37]. CPME is also a solvent resulting from chemical synthesis. It is an alternative to traditional ether solvents like tetrahydrofuran (THF) or dioxane. It is not produced from renewable feedstock but has numerous advantageous properties that make CPME a good candidate for direct replacement of ethers, thanks for example to its boiling point (106 °C) allowing the reduction of peroxide formation [38,39]. CPME outperforms other ether solvents in terms of environment, health, and safety, which awakens interest in the field of plant extraction. 6.2.4.4

Vegetable Oils

Vegetable oils are nonpolar lipophilic systems whose composition considerably vary according to their origins, the quality, and the methods with which they were obtained. Commonly used in cosmetics or in food industry, they can also be applied to extraction field as, for example, to obtain bioactive phytochemicals from natural resources. Vegetable oils have successfully been used as solvent for the extraction of carotenoids from by-products of crustaceans or from fresh carrots [9,40,41] and also for the extraction of aromas from basil [42]. 6.2.4.5

Eutectic Solvents

A eutectic solvent is a type of ionic liquid consisting of a mixture of compounds that forms a eutectic having a much lower melting point than its components taking individually. At first, eutectic solvents were mixtures of quaternary

123

124

6 Cosmo-RS-Assisted Solvent Screening for Green Extraction of Natural Products

ammonium salts in combination with hydrogen donors such as amines or carboxylic acids. Today, new solvent called natural deep eutectic solvents (NaDES) are considered as a new class of ionic liquids [43]. NaDES are inspired by the way metabolites are dissolved in the plant cell, by eutectic combinations of sugars and nitrogen compounds. Bi et al. [44] investigated the potential of Nades for the extraction of polyphenols from Japan cypress leaf but to date, the extent of research dealing with the effectiveness of NADES as solvents in processes of extraction or separation of molecules is relatively low. 6.2.4.6

Supercritical CO2

Extraction using supercritical CO2 uses compressed carbon dioxide at moderate temperature (31 °C and 74 bar) that acts as a solvent. This technique allows working at a moderate temperature that does not denature the thermolabile compounds of the plants. CO2 is not toxic and is completely eliminated as gas by returning to atmospheric pressure at the end of the extraction. This technique allows obtaining pure extracts free of all traces of solvent [45]. The extraction with supercritical CO2 is applied to molecules with low molecular weight and having low polarity such as carotenoids, triglycerides, fatty alcohols, light aromas, and so on. A wide range of applications using supercritical CO2 extraction are available today and the most common one is the extraction from natural plant material. This technique is nowadays used as industrial scale for the extraction of spices (India, Germany), the coffee and tea decaffeination (Germany, USA, Italy), or the extraction of hops for the brewery (Europe).

6.3 Prediction of Solvent Extraction of Natural Product

The choice of an alternative solvent remains a crucial step in a substitution process. This can be done, thanks to computational methods and especially continuum solvation models allowing the description of electrostatic interaction between a solute and a solvent and also the calculation of thermodynamic properties for solvation without any experimental data. COSMO-RS is one of these models that can be seen as an evaluation tool of solvation power, important criteria in the field of the extraction of natural products. 6.3.1 COSMO-RS Approach

First published by Klamt and Schüürmann [46], the conductor-like screening model, usually noted COSMO, has become very popular in computational chemistry, as it is an efficient variant of the apparent surface charge dielectric continuum solvation models [47]. The solute is treated as if embedded in a dielectric medium via a molecular surface constructed around the molecule. This is a macroscopic approach using dielectric constant of the solvent.

6.3 Prediction of Solvent Extraction of Natural Product

COSMO-RS, with the extension RS for “real solvent,” is a combination of an electrostatic theory, COSMO, with the statistical thermodynamics treatment of interacting surfaces [47]. It describes the liquid-phase interactions such as interaction contacts of molecular surfaces. This tool can generate a charge density surface that reflects the most stable state of the molecule in a perfect conductor. Unlike models by contribution groups (UNIFAC, Universal Functional Activity Coefficient) for which all functional groups and all interactions are not taken into account, the COSMO and COSMO-RS calculations can be applied to all types of solute–solvent organic systems. The molecule of solute is immersed in a continuous medium of dielectric constant ε. This can be imaged by a perfect virtual conductor [48]. In this environment, a surface is built around the molecule that generates a large number of electrostatic charges. Each segment i of this surface is characterized by its area ai and the screening charge density σ i on this segment that takes into account the electrostatic screening of the solute molecule by its surrounding and the back-polarization of the solute molecule [48]. Each piece of the molecular surface is in close contact with another one. The structure and the distribution of charges are optimized in order to find the minimum energy of the system (molecule in its most stable state) through calculations based on algorithms, called DFT (density functional theory). This charge density is called σ-surface. The color code of this σ-surface gives information on the charge density at each point of the molecule. The green zone corresponds to zero charge density, the blue characterizes a density of positive charge δ+ and red color means densities of negative charge δ . In other words, the green areas are usually linked to the carbon skeleton, the red area to oxygenated or nitrogenous groups, and blue areas to hydrogen atoms. This optimum charge distribution is segmented and reduced to a histogram called σ-profile as can be seen in Figure 6.3. The σ-profile noted p(σ) translates into 2D the information contained in the 3D surface. Thus, the σ-surface and the σ-profile permit to characterize a reference molecule (minimum energy, stable molecule) as a solute in a perfect conductor. At this point, the molecule is isolated and does not detect the molecules in its vicinity. In order to consider a molecule as a solvent or a molecule in a solvent and to quantify the associated interaction energies, an additional step of statistical thermodynamic calculations is required. Solvent interactions are reduced to local interactions by pairs of surface portions represented by charge densities σ and σ ´ (Figure 6.4). All interactions of surfaces are assumed to be in close contact. These contacts can be ideals, noncomplementary or can highlight hydrogen bonding that generate energies of interactions. The sum of these energies for a defined area is the interaction energy functional. If a contact on a surface area aeff (effective contact area) is considered in the particular situation of complementary molecule (where σ = σ ´ ), the interaction energy, also called misfit energy Emisfit, is equal to zero and the contact is “ideal.”

125

126

6 Cosmo-RS-Assisted Solvent Screening for Green Extraction of Natural Products

Figure 6.3 Distribution of charge.

In the general case, there is a mismatch between the partners, σ ˆ 6 σ ´ , therefore Emisfit is not zero. In the case where two opposite polarity surfaces are in contact, additional energy appears, hydrogen bonding (HB) energy Ehb. HB donors have a strongly negative screening charge density whereas HB acceptors have strongly

Figure 6.4 Charge interactions (as described in Ref. [49]).

6.3 Prediction of Solvent Extraction of Natural Product

positive screening charge densities. HB interaction can generally be expected if two sufficiently polar pieces of surface of opposite polarity are in contact. This energy is added to the electrostatic Coulomb energy from a σ hb threshold. In addition to electrostatic misfit and HB interaction, Van der Waals interactions between surface segments are taken into account. The contribution of Van der Waals interactions is not dependent of its vicinity, this is not an energy of interaction but rather a contribution to the energy of the reference state. The interactions of molecular surfaces are given by an interaction energy functional, noted Eint E int ˆ E misfit ‡ E hb ‡ E vdw It depends on the polarity of charge densities (hydrogen bonding) interactions surfaces (Emisfit, “noncomplementary” contacts), plus a contribution of Van der Waals interactions (energy of the reference state). The link between the microscopic surface interaction energies and macroscopic thermodynamic properties of a liquid is statistical thermodynamic calculations in order to obtain a consistent model of molecules in solution. These statistical calculations can be done for the ensemble of interacting surface pieces. To describe the composition of the surface segment ensemble with respect to the interactions, only the probability of σ has to be known for all compounds i. In this case the σ-profile pS(σ) is simply the sum of σ-profiles of components pi, weighted by their molar fraction in the mixture (xi). pS …σ† ˆ Σxi :pi …σ† The chemical potential of a surface segment with screening charge densities in an ensemble described by normalized distribution function ps(σ) is noted μs(σ). This approach can first allow estimating the affinity of a molecule depending on the contacting charge density σ. The molecule is considered as a solvent and this affinity is represented as a σ-potential curve μs(σ). Second, the estimation of the affinity of a molecule i in a solvent S is translated by computing its chemical potential μiS (kcal (mol 1 A 2). This chemical potential is obtained by integration of the σ-potential over the surface of i. The σ-potential is calculated from the statistical thermodynamics of molecular interaction based on the obtained σ-profile. This chemical potential μiS, given in the equation below, is the standard chemical potential minus RTln(xi), which allows the prediction of almost all thermodynamic properties of compounds or mixtures including solubility. Z μSi ˆ μiC;S ‡ pi …σ †μS …σ †dσ With μiC;S , the combinatorial contribution to the chemical potential resulting from the derivation of the combinatorial free energy expression [49]. All calculation steps are summarized in Figure 6.5. In a general way, a compromise between computational demands of quantum chemistry and quality of the predictions has to be made.

127

128

6 Cosmo-RS-Assisted Solvent Screening for Green Extraction of Natural Products

Figure 6.5 Different calculation steps of COSMO-RS. The quality, accuracy, and systematic errors of the electrostatics resulting from the underlying quantum chemical COSMO calculations depend on the quantum chemical method as well as on the basis set. In fact the software needs a special parameterization for each method/basis set combinations. The choice of the appropriate quantum chemistry method and basis set level generally depends upon the required quality and the later application of the predictions. The following three main areas of application require different proceeding regarding quantum chemistry [50]:



 

Applications in chemical and engineering thermodynamics; they typically require high quality prediction of thermophysical properties (such as prediction of binary VLE or LLE data, activity coefficients in solution

or vapor pressure) of mixtures of small/ medium molecules (up to 25 nonhydrogen atoms). Application for the purpose of screening a large number of compounds typically requires a predictive quality that is rather lower than for applications in chemical engineering (such as rediction of solubility of compound in various solvents or prediction of solvent partition coefficients like log P for a large number of solutes. However, the molecules involved are often larger (>100 atoms) and an overall large number of compounds have to be computed by quantum chemistry. Treatment of ionic species does not require any special parametrization of calculation methods, nevertheless high quality quantum chemistry method in combination with a large basis set is required to capture the strong polarity of the ionic species.

6.3.2 Applications of COSMO-RS-Assisted Substitution of Solvent

Today, various examples of the use of COSMO-RS are available in the literature. Moity et al. [51] applied the predictive approach COSMO-RS on conventional organic solvents and “green” solvents. The study presents an overview of their

6.3 Prediction of Solvent Extraction of Natural Product

physicochemical properties that facilitates the comparison between the conventional solvents and these new solvents. This approach via COSMO-RS has already been used in various application fields such as in pharmaceutical formulation for the selection of an excipient [52], in the improvement of the purification liquid–liquid extraction of monoethylene glycol [53], for separating long chain fatty acid [54] or in cosmetics for solubilizing ingredients [55]. Moreover, COSMO-RS has successfully been applied to extraction field with the screening of alternative solvents for the replacement of hexane in the extraction of vegetable oil or aromas. 6.3.2.1 Example 1: COSMO-RS Assisted Selection of Solvent for Extraction of Seed Oils

As described by Sicaire et al. [11,56], COSMO-RS has been used as a tool to get a first idea on the potential of biobased solvents for the extraction of rapeseed oil for the replacement of hexane used in the industrial process. Currently, hexane is the most commonly used solvent for extraction of vegetable oils, thanks to his various advantages such as ease of removal or low boiling point, but it also provides ideal functionalities in terms of lipid solubility. However, it is produced from fossil energies and n-hexane, which is one of the main constituents of industrial hexane is classified as category 2 reprotoxic and as category 2 aquatic chronic toxic [57], which makes its use at industrial scale questionable. COSMO-RS Study

A COSMO-RS simulation was conducted on four triglycerides (TAGs), TAG1 (R1:C18:3n-3, R2: C18:2n-6, R3:C18:2n-6), TAG 2 (R1:C18:3n-3, R2:C18:2n-6, R3:C18:2n-6), TAG 3 (R1:C18:1n-9, R2:C18:1n-9, R3:C18:1n-9), TAG 4 (R1: C18:1n-9, R2:C18:2n-6, R3:C18:2n-6). The calculation has been performed for each molecule of interest (solvent and TAGs), using the Cosmotherm program, (Version C30 Release 13.01). The DFT/Cosmo calculation has been with using the quantum chemical program Turbomole. The aim is to determine their relative solubility, given by log(xj ), in various solvents (n-hexane, 2-methyltetrahydrofuran, cyclopentyl methyl ether (CPME), dimethyl carbonate (DMC), isopropanol (IPA), ethanol (EtOH), ethyl acetate (EtOAc)). n-Hexane was taken as reference as it is the solvent of choice for the extraction of lipids. The temperature was fixed at 55 °C that corresponds to the temperature of extraction in industrial conditions. Results of the COSMO-RS simulation are presented in Figure 6.6. The logarithm of the best solubility is set to 0 and all other solvents are given relatively to the best solvent. It can be noticed that log(xj ) for TAG2, TAG3, and TAG4 with n-hexane is equal to 0. It means that it has the best solubility compared to other tested solvents. Considering these TAGs, log(xj ) with MeTHF, CPME, and ethylacetate are equal to 0, which means that in terms of relative solubility these five solvents are equivalent to n-hexane (and even better for TAG1). The other solvents, DMC, IPA, and ethanol, are theoretically not good substitute to n-hexane for the extraction of TAGs as log(xj ) for these

129

130

6 Cosmo-RS-Assisted Solvent Screening for Green Extraction of Natural Products

Figure 6.6 Relative solubility of four triglycer- 6, R3:C18:2n-6), TAG 3 (R1:C18:1n-9, R2:C18:1nides in alternative solvents given by COSMO- 9, R3:C18:1n-9), TAG 4 (R1:C18:1n-9, R2:C18:2nRS simulation (TAG 1 (R1:C18:3n-3, R2:C18:2n- 6, R3:C18:2n-6)). 6, R3:C18:2n-6), TAG 2 (R1:C18:3n-3, R2:C18:2n-

constituents are lower than with n-hexane. Regarding the global results, MeTHF, CPME and, EtOAc appear to theoretically be the best candidates to hexane among all other tested solvents considering the relative solubility of the four major TAGs of rapeseed oil. Experimental Approach

These solvents were then experimentally tested for the actual extraction of rapeseed oil in order to correlate the results of the actual extraction to those computed using COSMO-RS. The lipid yields given in Figure 6.7 show that n-hexane enables the extraction of around 47 g lipids/100 g dry matter as well as MeTHF and ethanol. IPA gives a yield of at least 45 g lipids/100 g dry matter, thanks to the extraction of phospholipids. Ethylacetate and DMC are also good extraction solvents considering the standard deviation of the yields. The last solvent, CPME, gives slightly lower yields but allow at least an extraction of 37 g lipids/g dry matter that represent around 80% of the amount extracted with n-hexane. Concerning the lipid profile of extracts, no significant selectivity has been noticed as the composition in fatty acids remains the same. The main fatty acids in extracted oils are oleic (C18:1), linoleic (C18:2), linolenic (C18:3), and palmitic (C16:0) that represent more than 90% of the total fatty acids in extracted oil. Moreover, a HPTLC analysis allows to confirm that more than 80% of the constituents extracted with these solvents are triglycerides (TAG) as shown in Figure 6.7. Other constituents found in oils extracted with MeTHF, CPME, IPA, and ethanol, are phospholipids due to the higher polarity of these solvents compared to n-hexane.

6.3 Prediction of Solvent Extraction of Natural Product

Figure 6.7 Lipid yield and lipid classes in total extracts (TAG: triglyceride, PE: phosphatidylethanolamine, PI: phosphatidylinositol, PC: phosphatidylcholine, LPC: Lysophosphatidylcholine).

Comparison between Experimental and Simulations

The COSMO-RS calculations indicate that MeTHF and CPME were theoretically the best alternatives to n-hexane. Nevertheless, the experiments showed that among the both solvents only MeTHF was as good as n-hexane qualitatively and quantitatively. CPME gives a lower yield than nearly all the other tested solvents. Ethanol and IPA experimentally give very good lipid yields, thanks to the extraction of phospholipids. Ethyl acetate that was a good candidate from theoretical point of view show good results given the standard deviation as well as DMC even if it was not expected after the COSMO-RS simulations. The differences between theory and experiments can be explained by the fact that COSMO-RS only generates solubility simulations and does not take into account the phenomenon of extraction itself. The extraction does not just depend on the solubility of the target solute; however, it is also affected by the penetration of the solvent inside the matrix depending on the viscosity, for example, the disruption of biological structures that contain the solute as well as the diffusion of the solute to the solvent. Some tested solvent can appear as very good solvent for the free solute solvation but the performances are likely to be affected by the other components of the matrix. 6.3.2.2

Example 2: Cosmo-Rs-assisted Selection of Solvent for Extraction of Aromas

The predictive method COSMO-RS [47], has been performed to evaluate the performance of five alternative solvents (2-methyltetrahydrofuran (MeTHF), ethyl acetate (EtOAc), ethyl lactate, ethanol (EtOH), and dimethyl carbonate

131

132

6 Cosmo-RS-Assisted Solvent Screening for Green Extraction of Natural Products

(DMC)) compared to n-hexane for extraction of food aromas from caraway (Carum carvi L) seeds. COSMO-RS Prediction

Figure 6.8 shows the step calculation with COSMO-RS. The calculation has been performed for each molecule of interest (solvent and aromas of caraway), using the Cosmotherm program, (Version C30 Release 13.01). The DFT/Cosmo calculation has been with using the quantum chemical program Turbomole. The 3D distribution of the polarization charges σ on the surface of each molecule is converted into a surface composition function (σ-profiles). Such σ-profiles provided detailed information about the molecular polarity distribution. The chemical potential of the surface segment (σ-potential) is calculated from thermodynamics of the molecular interactions based on the obtained σ-profile. Based on the σ-profile of the solvent, its chemical potential, σ-potential (Figure 6.8d), which can be interpreted as the affinity of the solvent S for the surface of polarity σ. The chemical profile and potential can be used to understand the interaction between aroma and solvent in the mixture state. It is for this reason that

Figure 6.8 Step calculation with COSMO-RS: (a) molecule emmerged; (b) molecular surface; (c) energies of local surface interactions between σ-profiles of limonene and three solvents (Hexane, EtOAc, MeTHF); (d) σ-potentials of limonene and three solvents.

6.3 Prediction of Solvent Extraction of Natural Product

Cosmotherm is a useful tool for the substitution of n-hexane. Figure 6.8d shows the potentials of three solvents, namely, hexane, MeTHF, EtOAc, and the aroma compounds of caraway: Limonene. The σ-potential of a selected compound describes the likeliness of this compound to interact with solvents with polarity and hydrogen bonds. Analysis of the profile and potential of the components of the mixture (aromas and alternatives solvents) gives some important information about the molecules and can be used to predict possible interactions in the fluid phase. σ-profile of hexane (Figure 6.8c) shows two peaks, resulting from the hydrogens on the negative side and from the carbons on the positive side. But the two peaks are close enough to be of no relevance for the σ-potential of hexane. As depicted in Figure 6.8d, σ-potential of hexane is close to a parabola centered at σ = 0. It is characteristic of apolar alkane. Thus, the same remarks can be made for the σ-potential of Limonene. Otherwise, Figure 6.8c shows a σ-profile for the EtOAc with three peaks. The positive contribution resulting from the carbonyl oxygen. The carbons appear to be responsible at σ = 0.1 eA 2. The last peak resulting from the polarized hydrogens. As a consequence, the σ-potential (Figure 6.8d) is asymmetric. For example, EtOAc σ-potential is attractive, thanks to electron pairs in carbonyl (on the negative side) and it is almost an unattractive (on the positive side) as the nonpolar hexane. The same remarks can be made for the σ-potential of MeTHF. Another suitable criterion is the similarity of the σ-potential, which can be quantified by the sigma potential match similarity (SPS). In addition COSMOtherm allows the calculation of a molecular σ-potential similarity SPi,j of two compounds i and j. This SPi,j values depends on the temperature. The calculation has been performed at boiling point of each solvent. Thus, SPi,j will be small if the overlap between the compounds σ-potentials is small. The SPi,j values are listed in Table 6.1. For example, limonene has a SPi, P j = 51 with hexane and a S i,j = 9 with EtOAc. The σ-potentials of limonene and hexane are very similar unlike limonene and EtOAc or 2-MeTHF. According to the rule “like dissolve like,” EtOAc or 2-MeTHF seems thus more promising to extract oxygenated monoterpenes as hexane. It is interesting to note that most flavors of caraway are oxygenated monoterpenes. Which explains that, EtOAc extract shows an aromatic profile interesting. (Table 6.1) Composition of Extracts

This simulation results have been compared by experimental solvent extraction of caraway seeds using various alternative solvents (biobased and petrochemical). Extraction of caraway at the boiling point of each solvent has been done with conventionally heated reflux. The resulted food aroma’s extracts were analyzed by GC-FID or GC-MS to compare the solvent’s performance in terms of major flavors compounds and in terms of selectivity between the two major compounds: limonene and carvone. The chemical class characterization of caraway seed essential oil showed the prevalence of ketones, represented by carvone (73.20%, of the total GC area), trans-dihydrocarvone (0.37%), and cis-dihydrocarvone (0.36%). The monoterpene hydrocarbons formed the second main class

133

134

6 Cosmo-RS-Assisted Solvent Screening for Green Extraction of Natural Products

Table 6.1 List of SPi;j values.

Boiling point (°C)

A

C

D

E

H

I

Hexane

MeTHF

Ethyl acetate

Ethyl lactate

Ethanol

DMC

72

80

77C

119

73

90

Monoterpenes 1

Limonene

51

9

9

9

3

7

2

α-Terpinene

56

8

8

9

3

7

27

Monoterpenes oxygenated 3

Linalool

7

15

26

34

21

4

Limonene oxide

9

40

48

30

15

32

5

β-terpineol

4

33

55

51

33

52

6

Dihydrocarvone

7

42

59

36

18

38

7

α-Terpineol

4

78

25

15

18

18

8

Carveol

6

43

69

38

23

45

9

Carvone

5

53

82

48

26

54

10

(E) Anethole

22

13

20

18

6

15

65

7

7

8

2

6

10

41

46

27

15

30

Low

Medium

High

Sesquiterpenes 11

β-caryophyllene Sesquiterpenes oxygenated

12

Caryophyllene oxide

and contain limonene (23.57%) as the principal constituent. The remaining fractions, such as aldehydes, oxygenated monoterpenes, sesquiterpenes, and oxygenated sesquiterpenes, formed the minor essential oil chemical classes of caraway seeds. In agreement with the simulation results, ethyl acetate extract shows an aromatic profil similar to essential oil of caraway. Comparison between Experimental and COSMO-RS Study

Carvone causes typical freshness character of caraway aromas. With limonene, they are the main compounds in caraway extract. To compare COSMO-RS study and experimental data, we studied these two compounds more in detail. The σ-potential of limonene is similar to n-hexane and α-pinene and the σ-potential of carvone is similar to EtOAc, DMC, MeTHF, or ethyl lactate with different indice (Table 6.1). Thus, α-pinene appears as alternative solvent and most promising for limonene extraction and ethyl acetate especially for carvone. A quantitative analysis has been performed for each solvent extract, in view to determine the amount of carvone and limonene. The results of the GC-FID analysis are shown in Figure 6.9. Clear differences are observed in the relative

6.4 Conclusion

Figure 6.9 Quantitative analysis of the two main flavors of caraway.

composition of the extracts. The amount of limonene, isolated from ethyl acetate or MeTHF was highest. The difference of α-pinene extract, between simulation and experimental result, can be explained by the difficulty to evaporate the solvent. The resulting azeotrope during the step of evaporating the solvent contain as well α-pinene as monoterpenes. MeTHF and EtOAc extract contain limonene and carvone, while than IPA, butanol and ethyl lactate extract are a content in limonene and carvone considerably lower. The DMC extract showed a selectivity for carvone. According to simulation COSMO-RS the best solvent for carvone extraction is EtOAc (Si,j = 82), followed by DMC (Si,j = 53) and MeTHF (Si,j = 54). Nevertheless, it is interesting to note that DMC extract showed a high concentration in carvone (c = 8.9 g l 1), followed by EtOAc (4.5 g l 1) then MeTHF (4.4 g l 1), experimentally. This purely predictive approach remains valid for the screening of solvents and allows including emerging biobased molecules. Finally, EtOAc extract has a rich flavor profile and a high concentration in carvone and limonene. DMC extract shows selectivity against carvone. EtOAc [58] and DMC are not CMR. Consequently, coupling the COSMO-RS study to experimental solubility profile, EtOAc and DMC appeared as the alternatives solvents most likely to replace n-hexane for extraction of aromas from caraway seeds.

6.4 Conclusion

In the substitution of toxic solvents from petrochemical origin, the use of greener products such as biobased solvents sourced from industrial by-products appears as an option. These solvents are synthesized from at least one renewable raw material generally coming from the sugar or oilseed industry. Many of the alternative solvents listed in this chapter can be used for extraction. The choice

135

136

6 Cosmo-RS-Assisted Solvent Screening for Green Extraction of Natural Products

of a substitution solvent is an important measure in the implementation of more sustainable processes. Decision tool as COSMO-RS can be very useful and is now commonly used in various fields of applications. As described in this chapter, COSMO-RS has successfully been used to have a first idea on the potential of solvents for the extraction of various compounds from matrices as for example oil from oilseeds or aromas from caraway. Most of the time the results simulation are correlated by the experimental investigations even if sometimes there are some differences. COSMO-RS is very useful tool to help in the substitution but results have to be correlated with experiments as it can’t simulate all the phenomena that occur during the extraction.

References 1 Chemat, F. and Vian, M.A. (eds) (2014)

2

3

4

5

6

7

8

9

Alternative Solvents for Natural Products Extraction, Springer, Berlin. ALCIMED (2002) Analyse des marchés potent iels des agro-solvants et recommandations pour la conduite du programme AGRICE. Vijayendran, B. (2010) Bio products from bio refineries: trends, challenges and opportunities. Journal of Business Chemistry, 7, 109–115. Anastas, P.T. and Warner, J.C. (1998) Green Chemistry: Theory and Practice, Oxford University Press, Oxford. Henderson, R.K., Jiménez-González, C., Constable, D.J.C. et al. (2011) Expanding GSK’s solvent selection guide: embedding sustainability into solvent selection starting at medicinal chemistry. Greem Chemistry, 13, 854. Prat, D., Pardigon, O., Flemming, H.-W. et al. (2013) Sanofi’s solvent selection guide: a step toward more sustainable processes. Organic Process Research and Development, 17, 1517–1525. Alfonsi, K., Colberg, J., Dunn, P.J. et al. (2008) Green chemistry tools to influence a medicinal chemistry and research chemistry based organisation. Greem Chemistry, 10, 31–36. Johnson, L. and Lusas, E.W. (1983) Comparison of alternative solvents for oils extraction. Journal of the American Oil Chemists Society, 60, 229–242. Li, Y., Fabiano-Tixier, A.S., Tomao, V. et al. (2013) Green ultrasound-assisted

10

11

12

13

14

15 16

extraction of carotenoids based on the bio-refinery concept using sunflower oil as an alternative solvent. Ultrasonics Sonochemistry, 20, 12–18. Filly, A., Fabiano-Tixier, A.-S., Lemasson, Y. et al. (2014) Extraction of aroma compounds in blackcurrant buds by alternative solvents: theoretical and experimental solubility study. Comptes Rendus Chimie, 17, 1268–1275. Sicaire, A.-G., Vian, M., Fine, F. et al. (2015b) Alternative bio-based solvents for extraction of fat and oils: solubility prediction, global yield, extraction kinetics, chemical composition and cost of manufacturing. International Journal of Molecular Sciences, 16, 8430–8453. Fréville, V. (2011) Caractérisation et évaluation des performances de nouveaux solvants synthétisés à partir d’agroressources. Université de Technologie de Compiègne. Reichardt, C. (2007) Solvents and solvent effects: an introduction. Organic Process Research and Development, 11, 105–113. Nič, M., Jirát, J., and Košata, B. et al. (eds) (2009) IUPAC Compendium of Chemical Terminology: Gold Book, 2.1.0 edn, IUPAC, Research Triagle Park, NC. De Caro, P. and Thiebaud Roux, S. (2008) Biosolvents. Petigny, L., Périno, S., Minuti, M. et al. (2014) Simultaneous microwave extraction and separation of volatile and non-volatile organic compounds of boldo leaves: from

References

17

18

19

20

21

22

23 24

25

26

27

28

lab to industrial scale. International Journal of Molecular Sciences, 15, 7183–7198. Fairchild, E.H., Komor, J.A., Petro, A.J., and Baird, G.T. (1999) Hydrotropes and their applications, in Industrial Applications of Surfactants IV (ed. D.R. Karsa), Woodhead Publishing, pp. 50–62. Burtin, E., Teissier, R., and Tretjak, S. (2006) Continuous ethyl lactate preparation method. U.S. Patent US20060041165 A1. Nikles, S.M., Piao, M., Lane, A.M., and Nikles, D.E. (2001) Ethyl lactate: a green solvent for magnetic tape coating. Greem Chemistry, 3, 109–113. Muse, J. and Colvin, H. (2005) Use of ethyl lactate as an excipient for pharmaceutical compositions. U.S. Patent US8048919 B2. McConville, J.T., Carvalho, T.C., Kucera, S.A., and Garza, E. (2009) Ethyl lactate as a pharmaceutical-grade excipient and development of a sensitive peroxide assay. Colley, S.W., Fawcett, C.R., Rathmell, C., and Tuck, M.W.M. (2004) Process for the preparation of ethyl acetate. U.S. Patent US6809217 B1. Gadewar, S.B. (2013) Ethyl acetate production. U.S. Patent US8562921 B2. Gupta, A. and Verma, J.P. (2015) Sustainable bio-ethanol production from agro-residues: a review. Renewable and Sustainable Energy Reviews, 41, 550–567. Azmir, J., Zaidul, I.S.M., Rahman, M.M. et al. (2013) Techniques for extraction of bioactive compounds from plant materials: a review. Journal of Food Engineering, 117, 426–436. Deswartvaegher, A., Forestier, B., Miard, S. et al. (2006) Varnish composition based on a solvent exclusively of vegetable origin. U.S. Patent EP1819400 B1. Mira, B., Blasco, M., Berna, A., and Subirats, S. (1999) Supercritical CO2 extraction of essential oil from orange peel: effect of operation conditions on the extract composition. Journal of Supercritical Fluids, 14, 95–104. Njoroge, S.M., Koaze, H., Karanja, P.N., and Sawamura, M. (2005) Essential oil constituents of three varieties of Kenyan

29

30

31

32

33

34

35

36

37

38

sweet oranges (Citrus sinensis). Flavour and Fragrance Journal, 20, 80–85. Dejoye Tanzi, C., Abert Vian, M., Ginies, C. et al. (2012) Terpenes as green solvents for extraction of oil from microalgae. Molecules, 17, 8196–8205. Fine, F., Vian, M.A., Tixier, A.-S.F. et al. (2013) Les agro-solvants pour l’extraction des huiles végétales issues de graines oléagineuses. OCL, 20, A502. Zheng, H.-Y., Zhu, Y.-L., Teng, B.-T. et al. (2006) Towards understanding the reaction pathway in vapour phase hydrogenation of furfural to 2-methylfuran. Journal of Molecular Catalysis Chemical, 246, 18–23. Aycock, D.F. (2007) Solvent applications of 2-methyltetrahydrofuran in organometallic and biphasic reactions. Organic Process Research and Development, 11, 156–159. Pace, V. (2012) 2-Methyltetrahydrofuran: a versatile eco-friendly alternative to THF in organometallic chemistry. Australian Journal of Chemistry, 65, 301–302. Pace, V., Hoyos, P., Castoldi, L. et al. (2012) 2-Methyltetrahydrofuran (2-MeTHF): a biomass-derived solvent with broad application in organic chemistry. ChemSusChem, 5, 1369–1379. Sicaire, A.-G., Vian, M.A., Filly, A. et al. (2014) 2-Methyltetrahydrofuran: main properties, production processes, and application in extraction of natural products, in Alternative Solvents for Natural Products Extraction (eds F. Chemat and M.A. Vian), Springer, Berlin, pp. 253–268. Kai, T., Mak, G.L., Wada, S. et al. (2014) Production of biodiesel fuel from canola oil with dimethyl carbonate using an active sodium methoxide catalyst prepared by crystallization. Bioresource Technology, 163, 360–363. Dawodu, F.A., Ayodele, O.O., Xin, J., and Zhang, S. (2014) Dimethyl carbonate mediated production of biodiesel at different reaction temperatures. Renewable Energy, 68, 581–587. Watanabe, K., Yamagiwa, N., and Torisawa, Y. (2007) Cyclopentyl methyl ether as a new and alternative process solvent. Organic Process Research and Development, 11, 251–258.

137

138

6 Cosmo-RS-Assisted Solvent Screening for Green Extraction of Natural Products 39 Kobayashi, S., Kuroda, H., Ohtsuka, Y.

40

41

42

43

44

45

46

47

48

et al. (2013) Evaluation of cyclopentyl methyl ether (CPME) as a solvent for radical reactions. Tetrahedron, 69, 2251–2259. Handayani, A.D., Sutrisno, I.N., and Ismadji, S. (2008) Extraction of astaxanthin from giant tiger (Panaeus monodon) shrimp waste using palm oil: studies of extraction kinetics and thermodynamic. Bioresource Technology, 99, 4414–4419. Pu, J., Bechtel, P.J., and Sathivel, S. (2010) Extraction of shrimp astaxanthin with flaxseed oil: effects on lipid oxidation and astaxanthin degradation rates. Biosystems Engineering, 107, 364–371. Li, Y., Fabiano-Tixier, A.S., Ginies, C., and Chemat, F. (2014) Direct green extraction of volatile aroma compounds using vegetable oils as solvents: theoretical and experimental solubility study. LWT - Food Science and Technology, 59, 724–731. Dai, Y., van Spronsen, J., Witkamp, G.-J. et al. (2013) Natural deep eutectic solvents as new potential media for green technology. Analytica Chimica Acta, 766, 61–68. Bi, W., Tian, M., and Row, K.H. (2013) Evaluation of alcohol-based deep eutectic solvent in extraction and determination of flavonoids with response surface methodology optimization. Journal of Chromatography A, 1285, 22–30. Jesus, S.P. and Meireles, M.A.A. (2014) Supercritical fluid extraction: a global perspective of the fundamental concepts of this eco-friendly extraction technique, in Alternative Solvents for Natural Products Extraction (eds F. Chemat and M.A. Vian), Springer, Berlin, pp. 39–72. Klamt, A. and Schüürmann, G. (1993) COSMO: a new approach to dielectric screening in solvents with explicit expressions for the screening energy and its gradient. Journal of the Chemical Society, Perkin Transactions, 2, 799–805. Klamt, A. (2011) The COSMO and COSMO-RS solvation models. Wiley Interdisciplinary Reviews. Computational Molecular Science, 1, 699–709. Klamt, A. and Eckert, F. (2000) COSMO-RS: a novel and efficient method for the a priori

49 50 51

52

53

54

55

56

57

58

prediction of thermophysical data of liquids. Fluid Phase Equilibria, 172, 43–72. COSMOlogic GmbH & Co. KG (2013) COSMOthermX tutorial 2013. COSMOlogic GmbH & Co. KG (1999) COSMOtherm User Manual. Moity, L., Durand, M., Benazzouz, A. et al. (2012) Panorama of sustainable solvents using the COSMO-RS approach. Green Chemistry, 14, 1132. Pozarska, A., da Costa Mathews, C., Wong, M., and Pencheva, K. (2013) Application of COSMO-RS as an excipient ranking tool in early formulation development. European Journal of Pharmaceutical Sciences, 49, 505–511. Garcia-Chavez, L.Y., Hermans, A.J., Schuur, B., and de Haan, A.B. (2012) COSMO-RS assisted solvent screening for liquid–liquid extraction of mono ethylene glycol from aqueous streams. Separation and Purification Technology, 97, 2–10. Xing, H., Zhang, X., Yang, Q. et al. (2014) Separation of long chain fatty acids with different number of unsaturated bonds by fractional extraction: experimental and COSMO-RS study. Food Chemistry, 143, 411–417. Benazzouz, A., Moity, L., Pierlot, C. et al. (2014) Hansen approach versus COSMO-RS for predicting the solubility of an organic UV filter in cosmetic solvents. Colloids and Surfaces A: Physicochemical and Engineering Aspects, 458, 101–109. Sicaire, A.-G., Abert Vian, M., Fine, F. et al. (2015a) Experimental approach versus COSMO-RS assisted solvent screening for predicting the solubility of rapeseed oil. OCL, 22, D404. ECHA (2008) Classifications – CL Inventory. Available at http://echa.europa.eu/en/ information-on-chemicals/cl-inventorydatabase/-/cl-inventory/view-notificationsummary/115449;jsessionid= 1E44CEF20283ABAD4FB3EAFC8290CE3A .live2 (accessed February 6, 2015). INRS (2011) Acétate d’éthyle – Fiche toxicologique. Available at http://www. inrs.fr/publications/bdd/doc/fichetox. html?refINRS=FT%2018 (accessed November 12, 2015).

139

7 Supramolecular Catalysis as a Tool for Green Chemistry Courtney J. Hastings

7.1 Introduction

Catalysis is central to advancing green chemistry in the area of synthetic chemistry [1,2]. Beyond replacing stoichiometric reagents, catalysts have the potential to streamline multistep synthesis by enabling new bond-forming processes to shorten synthetic sequences and achieve better step economy [3,4]. Supramolecular catalysis and the application of supramolecular concepts to catalytic reactions is emerging as a valuable tool for improving catalytic reactions for synthetic chemistry. Supramolecular catalysis can enable aqueous reaction conditions, improve reactions selectivity, improve catalyst lifetime, and enable tandem reactions, all of which can have positive impacts on the cost, waste, and energy associated with a reaction. The field of supramolecular chemistry concerns the design of molecular entities that are defined by reversible, noncovalent interactions. While each supramolecular interaction is quite weak individually, the effect of many such interactions working in concert can produce strongly associated and structurally well-defined molecular species [5–7]. Such additive effects are responsible for the spectacular structural complexity found in biomacromolecules such as proteins. Efforts to characterize these interactions have provided chemists with a “toolbox” of reliable methods to program the association between two or more molecules to form a single complexed species. Thus, supramolecular chemistry represents a complementary approach toward molecular construction, and one that offers certain advantages over covalent chemistry [5–8]. Like supramolecular interactions, host–guest binding relies on manifold noncovalent interactions, with the added requirement that the host possess an interior cavity that is complementary in size and shape to the guest molecule [9–11]. Quite frequently, the “inner phase” of a synthetic host presents a dramatically different chemical environment to a bound guest than what it would experience in the surrounding bulk solvent. In fact, the environment within a synthetic host is frequently unlike anything that a molecule would experience in any solvent,

Handbook of Green Chemistry Volume 12: Tools for Green Chemistry, First Edition. Edited by Evan S. Beach and Soumen Kundu.  2017 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2017 by Wiley-VCH Verlag GmbH & Co. KGaA.

140

7 Supramolecular Catalysis as a Tool for Green Chemistry

particularly with respect to confinement effects. Many hosts themselves are constructed through supramolecular interactions, self-assembling from relatively simple subunits into highly complex and symmetric structures [12–16]. The design of synthetic self-assembled host molecules requires control over the geometry of the individual components and how the components interact with each other. This control can be achieved by choosing the subunits to interact with each other through known and predictable noncovalent interactions. Supramolecular catalysis relies upon noncovalent interactions to provide the primary associative interaction between catalyst and substrate, a factor that is responsible for the spectacular selectivity and reactivity of enzymes. Supramolecular interactions can be involved in catalysis in a number of ways. Supramolecular encapsulation of one or more substrate molecules within a host (which itself is often self-assembled through supramolecular chemistry) can promote or modulate reactivity. Supramolecular binding can enforce substrate–catalyst interactions through molecular recognition processes that function independent of the reactive functional groups. Finally, it is possible to install catalytic moieties within the cavity of a molecular host, which can then bind substrate molecules. Since the field of supramolecular catalysis and related research areas have been the subject of many excellent reviews, the aim of this chapter is not to provide a comprehensive review of supramolecular catalysis [17–39]. Rather, the goal is to summarize the types of reaction improvements that can be made, and to provide representative examples where supramolecular catalysis was used a tool for obtaining a favorable reaction outcome. Special emphasis is placed on examples that involve widely used and synthetically useful transformations, such as cross-coupling, hydroformylation, and C H functionalization reactions. Finally, conceptually related work on encapsulation-mediated reaction control using metal–organic frameworks [40–44], the inner phase of polymers [45–48], and dendrimers [49–51], and other such species are beyond the scope of this chapter, and will be omitted.

7.2 Control of Selectivity through Supramolecular Interactions

Supramolecular binding and encapsulation can exert large effects on reaction selectivity, influencing which products are formed (regioselectivity, stereoselectivity) and which substrates are allowed to react (substrate gating). This aspect of supramolecular catalysis parallels the high levels of selectivity achieved by enzymes, which are also due in large part due to the cumulative influence of many noncovalent interactions between enzyme and substrate. Imposition of selectivity on synthetic reactions is an important goal, since separation of products typically requires energy- or solvent-intensive purification steps. Supramolecular control of selectivity is particularly attractive in reactions where many sites in a substrate molecule are equally reactive (e.g., C–H functionalization) or

7.2 Control of Selectivity through Supramolecular Interactions

where selectivity is difficult to achieve using traditional catalyst engineering approaches (e.g., photochemistry and hydroformylation). Thus, representative reactions in which supramolecular interactions improve selectivity to synthetically useful levels will be the focus of this section. 7.2.1 Catalysis with Supramolecular Directing Groups

Reactions in which attractive substrate–reagent (or substrate–catalyst) interactions exist often proceed with greater selectivity or altered selectivity compared to cases where a directing group is absent, and as such substrate-directed reactions are valuable in synthesis (Scheme 7.1) [52,53]. Typical directing groups influence selectivity by binding directly to the group that is reacting with the substrate. In the case of transition metal catalysis, this means that the metal center is both the reactive center and the site of molecular recognition. This strategy limits the possible substrate directing groups to those that will bind to, but not inhibit the catalytic metal. A more flexible strategy is for the molecular recognition element to be separate from the reactive center (Scheme 7.1). In addition to expanding the toolbox of noncovalent interactions that may be used for molecular recognition, this approach also enables remote functionalization [54–57], while traditional directing groups tend to favor activation of proximal positions.

Scheme 7.1

This approach was pioneered by Breslow and coworkers, who developed a cyclodextrin-modified Mn–porphyrin catalyst for aliphatic C H hydroxylation. This catalyst selectively hydroxylates an unactivated position of a steroid derivative (Scheme 7.2) [58]. The steroid substrate androstanediol was derivatized with

141

142

7 Supramolecular Catalysis as a Tool for Green Chemistry

Scheme 7.2

two ester groups bearing both water-solublizing moieties and a tert-butylphenyl for binding to the cyclodextrin. When this substrate is subjected to the catalyst in the presence of iodosobenzene as the terminal oxidant in water, the steroid is regio- and stereospecifically hydroxylated. It is noteworthy that the methylene position where hydroxylation occurs is not the most intrinsically reactive site on the substrate, and that supramolecular binding is responsible for the observed selectivity. Crabtree and coworkers reported a dimanganese terpyridine catalyst bearing two molecular recognition sites for binding carboxylic acid substrates. The terpyridine ligands are functionalized with a phenylene group and then the Kemp triacid, which provides a U-turn geometrical element. This orients a carboxylic acid directing group that is capable of binding carboxylic acid-containing substrates such as ibuprofen (Scheme 7.3). The binding mode positions a single C H bond near the active metal center, and the substrate is regio- and stereoselectively hydroxylated. Lower selectivity is observed when the reaction is performed using a dimanganese terpyridine catalyst lacking the molecular recognition element [59]. Bach and coworkers recently reported the design of a ruthenium–porphyrin catalyst bearing a chiral molecular recognition element. A chiral lactam moiety is linked to the porphyrin through a rigid alkyne linker and is responsible for binding lactam substrates (Scheme 7.4). The catalyst is capable of

7.2 Control of Selectivity through Supramolecular Interactions

Scheme 7.3

Scheme 7.4

enantioselective C H oxidation of prochiral spirocyclic lactam substrates with high enantioselectivity, but modest yields [60]. Recently an iridium–bypyridine complex with an attached urea group was disclosed by Kanai and coworkers for the catalytic C–H borylation of arenes. The pendant urea moiety complexes the carbonyl group of substrate benzamides, and the rigid ligand framework positions the iridium center closest to the C H bond meta to the amide substituent (Scheme 7.5). As a result, the borylation reaction is meta-selective, while analogous complexes lacking the substrate-binding group give a mixture of isomers [61].

143

144

7 Supramolecular Catalysis as a Tool for Green Chemistry

Scheme 7.5

Photochemical reactions are extremely valuable synthetic transformations, but controlling the selectivity of such reactions can be challenging. This is due to the extremely short lifetimes and high intrinsic reactivity of excited-state reaction intermediates, which provide little opportunity for directing the reaction outcome. Chiral, supramolecular triplet sensitizers have been developed by Bach and coworkers to perform enantioselective photochemical reactions. The catalyst design links a triplet sensitizer) to a chiral lactam-binding group derived from the Kemp triacid, and it both ensures close contact between the sensitizing group and the substrate while controlling the stereochemical outcome of the reaction (Scheme 7.6). This family of catalysts enables the synthesis of enantioenriched products via photochemical cyclization and [2 + 2] cycloaddition [62–65].

Scheme 7.6

7.2 Control of Selectivity through Supramolecular Interactions

7.2.2 Scaffolding Ligands

As an alternative to catalyst directing groups that operate through noncovalent bonds, it is also possible to use reversible covalent bonding to colocate substrate and a metal catalyst. Scaffolding ligands, which contain both a catalyst binding unit and a site for reversible covalent substrate binding, are used for this purpose [66,67]. The reversibility of the substrate binding allows the scaffolding ligand to be used in catalytic quantities. The initial application of this approach to the rhodium-catalyzed hydroformylation reaction was independently reported by the groups of Breit and Tan [68,69]. The scaffolding ligands employed in this system contain a phosphine group for metal binding and a site for reversible, covalent bonding of substrate (Scheme 7.7).

Scheme 7.7

The reversibly bound directing group is able to effectively impose regiocontrol over the hydroformylation of homoallyllic alcohols, which is followed by oxidation to provide lactone products. In the absence of the scaffolding ligand, a mixture of products favoring the linear aldehyde, which cyclizes to form a sixmembered lactone after oxidation (Scheme 7.8). The scaffolding ligand is able to override the intrinsic selectivity of the reaction, selectively producing the branched product (which forms a five-membered lactone after oxidation). This approach could also be applied to hydroformylation of alkene substrates bearing sulfonamide and aniline directing groups [70,71].

145

146

7 Supramolecular Catalysis as a Tool for Green Chemistry

Scheme 7.8

7.2.3 Selectivity through Confinement and Binding Effects

The chemical environment and confined space inside of self-assembled hosts can impart selectivity to reactions mediated by supramolecular catalysts. When catalysis occurs within a confined space, it is possible to impart product selectivity that is difficult to achieve with conventional catalysts. A second type of selectivity is the control over which substrates are allowed to react by limiting the size and shape of molecules that penetrate the host interior. Both of these types of selectivity are also hallmarks of enzymatic catalysis. Seminal work published by the van Leeuwen and Reek groups has explored the effect of supramolecular encapsulation on the selectivity of rhodium hydroformylation catalysts, an important reaction in which selectivity is difficult to control [72]. A monodentate tripyridylphosphine ligand is capable of binding a zinc porphyrin panel through each pyridine, creating a well-defined ligand-templated assembly that encapsulates a phosphine-bound rhodium center (Scheme 7.9). Compared to the rhodium complex without the associated porphrins, the ligand-templated catalyst is more active and more selective for the branched isomer. The increased selectivity produced by the encapsulated catalyst is due to the steric restrictions imposed by the assembly interior. A related ligand-templated assembly was created from the tris(zinc (II) porphyrin)phosphite ligand, which self-assembles in the presence of three bridging diamines to form a sandwich structure with a rhodium center in the interior cavity. This supramolecular catalyst is an active hydroformylation catalyst and is highly selective for the linear hydroformylation product [73]. Rebek and coworkers have designed a family of open-ended resorcinarenederived hosts in which a diversity of functional groups are positioned over the host rim, protruding into the cavity. The host is functionalized with a carboxylic acid group, which is attached to the host rim and dangles into the binding pocket (Scheme 7.10). The intramolecular epoxide ring opening of a 1,5-epoxyalcohol is catalyzed by the host to form a hydroxymethyltetrahydrofuran product [74]. The host-catalyzed reaction is substantially accelerated when compared to the reaction catalyzed by a carboxylic acid that is electronically similar but lacks any substrate-recognizing cavity. This difference underscores the enhanced reactivity that results from enforcing the close proximity of substrate and a

7.2 Control of Selectivity through Supramolecular Interactions

Scheme 7.9

Scheme 7.10

catalytic functional group. Additionally, the 26-catalyzed reaction produces a mixture of regioisomers, the result of intramolecular nucleophillic attack at both epoxide positions, while the host-catalyzed reaction yields a single regioisomer. An important self-assembling catalyst system for epoxide formation was published by the Hupp and Nguyen groups. The supramolecular box self-assembles from rigid porphyrin-based components and forms a large, cavity-containing structure bearing interior manganese porphyrins [75,76]. While alkenes such as stillbene can be oxidized to the corresponding epoxide by the encapsulated catalyst, larger derivatives do not interact with the catalyst as easily and are undergo epoxidation less efficiently (Scheme 7.11). The ability to discriminate between

147

7 Supramolecular Catalysis as a Tool for Green Chemistry

Scheme 7.11

148

7.2 Control of Selectivity through Supramolecular Interactions

substrates on the basis of size alone is due to the steric constraints imposed by encapsulation. A recent from deBruin and coworkers report detailed the cyclopropanation behavior of a cobalt–porphyrin catalyst encapsulated within a M8L6 cubic assembly [77]. The constricted cage pores of the host modulate how easily substrates reach the encapsulated catalyst, with smaller substrates having easier access. In a competition experiment, 8:2 selectivity for the smaller substrate was exhibited (Scheme 7.12). In contrast, no selectivity is observed when the same experiment is conducted with the unencapsulated catalyst.

Scheme 7.12

A micellar system disclosed by Scarso and coworkers exhibits high levels of substrate selectivity in the palladium-catalyzed hydrogenation of α,β-unsaturated aldehydes [78]. In this example, the catalytic species is a surfactant-encapsulated Pd nanoparticle. Lipophilic substrates bearing long alkyl chains react faster than C4 and C5 substrates by a factor of 300 (Scheme 7.13). The opposite trend in reaction rates is observed when the reaction is conducted in organic solvent. This is due to their increased ability to associate with the micellar phase due to the hydrophobic effect, allowing easier access to the catalytic nanoparticle surface. Similar effects are seen in the Diels–Alder and Heck reactions, catalyzed by micelle-encapsulated Cr(III)–salen and Pd(II) catalysts, respectively [79,80]. Interestingly, larger substrates react faster in these systems, in contrast to the selectivity typically seen in other supramolecular systems.

149

150

7 Supramolecular Catalysis as a Tool for Green Chemistry

Scheme 7.13

7.3 Reactions in Water

Solvents account for a large fraction of waste generated in chemicals reactions, and switching to environmentally innocuous solvents is one of the twelve principles of green chemistry [1,2,81,82]. Water is perhaps the most obvious green solvent because it is nontoxic, nonflammable, inexpensive, and requires no synthesis. Despite these advantages, water is seldom used as a solvent for organic reactions because many substrates and reagents are either insoluble in or incompatible with water [83]. A related issue is that water can react with some reaction intermediates, producing undesired side products. Many supramolecular hosts are water soluble while possessing a hydrophobic interior cavity, and the host interior presents a chemical environment to encapsulated guests that is dissimilar to water. Thus, water can be used as the bulk solvent while the reaction itself takes place within the inner phase of the host, where reaction conditions are more favorable. Enzyme mimicry under biologically relevant reaction conditions (e.g., water as the solvent, physiological temperature, and pH) has been a long-standing goal of supramolecular chemistry, and many reviews have been published that summarize these efforts [23,25,27,29–33,35,38,54,84–87]. Likewise, conducting organic reactions in water using self-assembled micellar nanoreactors is a research area that has received considerable interest, and several reviews have been published [20,37,88,89]. Because these excellent reviews are quite comprehensive, this section will discuss selected examples that illustrate how supramolecular catalysis can improve reactions in water. 7.3.1 Water-Soluble Nanoreactors

A substantial fraction of self-asssembled molecular hosts are soluble in water and possess hydrophobic interiors. When the reactants and/or catalyst of an organic reaction are encapsulated within a hydrophobic cavity, the molecular host can act as a nanometer-sized reaction flask, bringing together reactants that

7.3 Reactions in Water

would otherwise be insoluble [31]. While many examples of supramolecular catalysis in water now exist, the most practical and synthetically useful strategy that has emerged is the use of micellar hosts that spontaneously self-assemble in water. Advantages of these systems over other water-soluble hosts include their reliable self-assembly under a wide range of conditions, the commercial availability and low cost of many micelle-forming surfactants, and wide range of hydrophobic molecules that are encapsulated [20,37,88,89]. Micelles lack defined structure compared to other supramolecular structures, which is responsible for their broad encapsulation behavior. A corresponding shortcoming of these systems is that they do not produce confinement effects found in other hostcatalyzed systems, such as shape-based substrate gating or the enforcement of specific substrate orientations. Kobayashi and coworkers have developed a number of useful reactions in water using Lewis acid–surfactant-combined catalysts (LASCs). Crucial to these reactions was the counterintuitive discovery that rare earth metal triflates are water-compatible Lewis acid catalysts for the Mukaiyama aldol reaction, and that water is in fact required for catalyst activitiy [90–92]. This led to the discover of the prototypical LASC, scandium tris(dodecyl sulfate) (Sc(DS)3), in which the Lewis acidic scandium atom possesses ligands with surfactant properties. While the LASC is soluble in water, it creates a hydrophobic environment for reactants that slows the rate of silyl enolate hydrolysis, a major decomposition pathway in water. The Mukaiyama aldol reaction between silyl enolates and aldehydes proceeds rapidly and in high yield using Sc(DS)3 as a catalyst in pure water as the solvent (Scheme 7.14) [93,94]. In addition to the aforementioned advantages of using water as the reaction solvent, this system allows the use of aqueous formaldehyde instead of gaseous or polymeric forms of the valuable C1 electrophile, which is not possible under anhydrous conditions [91].

Scheme 7.14

Kobayashi and coworkers have disclosed several additional reactions that are amenable to LASC catalysis in water (Scheme 7.15). The three-component Mannich reaction of amines, aldehydes, and silyl ketene acetals is catalyzed by Sc(DS)3 and in higher yield by a related Cu(II) LASC, copper bis(dodecyl sulfate) (Cu(DS)2) [94,95]. A similar three-component Abramov-type reaction of amines,

151

152

7 Supramolecular Catalysis as a Tool for Green Chemistry

Scheme 7.15

aldehydes, and phosphite ester nucleophiles is catalyzed by Sc(DS)3 [96]. Conjugate additions to electron-deficient olefins with beta-ketoester and indole nucleophiles in water using Sc(DS)3 as a catalyst were reported [97,98]. It is noteworthy that these reactions are all operationally simple and conducted at ambient temperature. Asymmetric reactions are also possible using LASC catalysis (Scheme 7.16). The ring opening of meso-epoxides with aromatic amines, catalyzed by Sc(DS)3

Scheme 7.16

7.3 Reactions in Water

and a chiral a bipyridine ligand, proceeded with high enantioselectivity [99]. Asymmetric catalysis of the Mukaiyama aldol reaction was also achieved using Cu(DS)2, a chiral bisoxazoline ligand, and a Brønsted acid additive [93,100]. In this case, however, the enantioselectivity was modest. Lipshutz and coworkers have made important contributions by advancing a series of designer surfactants that serve as nanoreactors for green, practical, and synthetically useful reactions in water (Scheme 7.17) [88,89]. Central to the success of this research effort has been the design of surfactants that self-assemble to form nanoreactors with optimal properties for mediating organic reactions in water. The size and morphology of particles formed by these surfactants were found to be particularly important, with 50–100 nm diameter particles being most effective. The structures of surfactants TPGS-750-M and Nok were both optimized with this property in mind, and accordingly are the most effective for performing reactions in water. It should also be noted that these surfactants are environmentally innocuous, being derivatives of nontoxic compounds vitamin E and β-sitosterol [101,102].

Scheme 7.17

153

154

7 Supramolecular Catalysis as a Tool for Green Chemistry

These designer surfactants, particularly TPGS-750-M and Nok, have been explored extensively for performing organic reactions in water as the bulk reaction solvent. All of the most common transition-metal-catalyzed reactions can be performed efficiently under micellar conditions, including olefin metathesis, Sonogashira coupling, Suzuki–Miyaura coupling, Heck coupling, Stille coupling, Miyaura borylation, and Buchwald–Hartwig amination (Scheme 7.18) [101–113].

Scheme 7.18

7.3 Reactions in Water

Other indispensable organic transformations, such as amide formation, nucleophilic aromatic substitution, and nitroarene reduction, can also be performed in water under micellar conditions (Scheme 7.19) [114–116]. Excellent yields are obtained for each of these micellar reactions at room temperature, while many of the corresponding reaction run in organic solvent under conventional conditions require elevated temperatures. This is due to the high local concentration found within the micellar nanoreactors, which accelerates reaction rates.

Scheme 7.19

Beyond the advantages of switching the reaction solvent to water, Lipshutz and coworkers have demonstrated that using micellar nanoreactors leads to dramatic reductions in waste when compared to traditional methods. The E factors (the ratio of waste to product produced by a chemical product) [82] of some representative reactions run under micellar and conventional conditions were compared, showing that micellar reaction E factors were typically reduced by an order of magnitude relative to conventional reactions. Finally, the aqueous reaction mixture left over after product extraction contains surfactant and catalyst, and can be recycled several times without detrimental effect on yield, further reducing the amount of generated waste [117]. Very recently, scientists at Novartis published an analysis of the environmental and economic benefits of using TPGS-750-M in water instead of conventional organic solvents for the kilogram-scale production of an Active Pharmaceutical Ingredient (API) [118]. Although the exact nature of the API and intermediates had to be obscured due to the commercial sensitivity of the project, the authors reported a 50% reduction in the quantity of organic solvents used and a 50% reduction in the quantity of substrates and reagents.

155

156

7 Supramolecular Catalysis as a Tool for Green Chemistry

7.3.2 Dehydration Reactions

Performing dehydration reactions in aqueous solvent is typically a challenging task, due to either thermodynamic (such as forming an amide or ester through condensation) or kinetic factors (such as when a carbocation can undergo elimination or nucleophilic attack by water). Counter to intuition, however, it is possible to bias reactions toward dehydration products in pure water if the reaction takes place within the hydrophobic interior of a supramolecular nanoreactor. In 2002, Kobayashi and coworkers reported that a Brønsted acid-functionalized surfactant acted as a catalyst for esterification of carboxylic acids and alcohols in water as the sole solvent [119]. Run under conventional conditions, removal of water formed as a coproduct of this reaction is typically necessary to bias the equilibrium toward the desired ester and achieve high yields. Esterification is successful in water using a catalytic quantity of dodecylbenzenesulfonic acid (DBSA), a micelle-forming Brønsted acid (Scheme 7.20). This reaction is limited to reaction partners that are quite hydrophobic, which is necessary for them to partition into the micelle interior. The inability for water to penetrate into the micelle core alters the thermodynamics of the system, producing high yields of ester. A similar approach was successful with other dehydration reactions, such as ether formation, thioether formation, and thioacetal formation. Since this report, several micelle-mediated dehydration reactions have been reported.

Scheme 7.20

A tetrahedral, self-assembled metal–organic cage (Ga4L6, where L is a bisbidentate organic ligand) developed catalyzes the monoterpene-like Prins cyclization of citronellal, as reported by the Raymond and Bergman groups in 2012 [120,121]. This cyclization proceeds through the intermediacy of a carbocation, which can be deprotonated to form an alkene product, or trapped with water to form the corresponding diol (Scheme 7.21). When the reaction is

7.3 Reactions in Water

Scheme 7.21

conducted in buffered acidic water, the diol is the major product, while the alkene product predominates when the reaction is catalyzed by encapsulation within the Ga4L6 assembly. This effect is also seen in the gold-catalyzed eneyne cycloisomerization, which similarly proceeds through a cationic intermediate (Scheme 7.22). In this case, the gold catalyst, PMe3AuBr, produces a product resulting from water incorporation. When a gold catalyst encapsulated within the Ga4L6 assembly (PMe3Au+  Ga4L6, where  denotes encapsulation) is used instead, a formally dehydrated product is also produced. In both of these cases, rate of water addition is substantially decreased within the hydrophobic interior of the supramolecular assembly, an effect that is responsible for the stabilization of various water-sensitive species within the same host [122–126].

Scheme 7.22

Fujita and coworkers have reported the catalysis of the Knoevenagel condensation by inclusion within a self-assembled metal–ligand cage (Pd6L4) bearing a 12+ charge. Low catalyst loading (1 mol%) of the cage is sufficient to catalyze the condensation of 2-naphthaldehyde with Meldrum’s acid in high yield in

157

158

7 Supramolecular Catalysis as a Tool for Green Chemistry

neutral water as the solvent (Scheme 7.23). The Pd2+ centers located near the cage openings stabilize the anionic reaction intermediates, increasing the reaction rate. Finally, the reaction product is too large to fit within the host cavity, facilitating catalytic turnover [127].

Scheme 7.23

Lipshutz and coworkers recently reported the gold(III)-catalyzed dehydrative cyclization of propargyl diols and propargyl amino acids in water under micellar conditions. When the equivalent reaction is conducted in organic solvent, activated molecular sieves are added to remove water, which is generated as a stoichiometric by-product [128,129]. Using TPGS-750-M as a micellar host for this reaction, the cyclization reaction proceeds smoothly, producing furan and pyrrole products in high yields (Scheme 7.24). No dehydrating agents are required to drive the reaction forward, despite the presence of a vast excess of water [130].

Scheme 7.24

7.4 Catalyst/Reagent Protection

Catalyst and reagent stability is an important issue in many synthetic reactions, and the prevention of off-pathway decomposition is critical for achieving low catalyst loadings and good atom economy. Supramolecular encapsulation of a reactive catalyst or reagent within a host cavity can protect it from detrimental interactions, providing longer catalyst lifetimes. This can produce higher yields

7.4 Catalyst/Reagent Protection

and allow lower catalyst loading, which is particularly important when considering the low earth abundance of many precious metals used in catalysis. 7.4.1 Catalyst Protection

Manganese porphyrins are useful catalysts for the oxidation of unactivated C H bonds, but they rapidly decompose, limiting their synthetic utility. The decomposition occurs through a bimolecular mechanism, forming an oxo-bridged Mn dimer (Mn-O-Mn). The supramolecular metal-ligand square published by Hupp and coworkers binds a single Mn porphyrin molecule through pyridine-Zn association (Scheme 7.11). The lifetime of the encapsulated Mn porphyrin is increased by 18-fold, and the turnover numbers are increased 10–100-fold as well. The stabilization is due to the suppression of the bimolecular decomposition by supramolecular protection [75]. Similar stabilization of Co–porphyrin catalysts are provided by encapsulation in work reported by de Bruin and coworkers (Scheme 7.12). The Bergman and Raymond groups reported the isomerization of allyl alcohols (Scheme 7.25) is catalyzed by a ruthenium(II) catalyst encapsulated within a self-assembled metal–ligand cage Ga4L6 (Scheme 7.21) [131]. This reaction exhibits host-mediated size selectivity, substrate inhibition, and the reaction proceeds in water. The catalyst lifetime is prolonged by encapsulation, and compared to the performance of the unencapsulated catalyst in organic solvent, encapsulation leads to higher turnover numbers.

Scheme 7.25

7.4.2 Protection of Water-Sensitive Reagents

An intriguing and surprising finding from the Lipshutz group is that their micellar systems allow for the generation and reaction of several moisture-sensitive organometallic reagents, even when water is the reaction solvent [102]. Negishilike couplings between alkyl and aryl halides can be accomplished using zinc

159

160

7 Supramolecular Catalysis as a Tool for Green Chemistry

dust and a palladium catalyst under micellar conditions in water (Scheme 7.26). The water-sensitive organozinc reagent is formed in situ at the metal surface and then partitions into the anhydrous micelle interior more rapidly that it can react with water. At this point, transmetallation to a less sensitive organopalladium species occurs.

Scheme 7.26

This concept was extended to perform cuprate conjugate addition reactions. The reaction of alkyl halides with zinc dust forms organozinc species, which undergoes transmetalation to form an organocopper species [132]. The organcopper reagent undergoes conjugate addition to an enone, catalyzed by AuCl3 as a Lewis acid (Scheme 7.27). Remarkably, this reaction proceeds smoothly despite the fact that the reaction must proceed through two water-sensitive intermediates in water as the bulk solvent. An additional feature is that this chemistry proceeds at room temperature instead of the cryogenic temperatures often necessary for organocopper chemistry.

Scheme 7.27

7.5 Tandem Reactions

Tandem reactions, in which multiple reaction events occur sequentially in a single reaction vessel, offer an appealing alternative to iterative chemical synthesis [133]. The isolation and purification of intermediate products is energy consuming, produces large quantities of chemical waste, and costly, so tandem reactions are particularly desirable from a green chemistry standpoint. The execution of tandem reactions is difficult because the conditions required for multiple reactions are often incompatible. It is possible to use supramolecular encapsulation as a tool to circumvent this problem by partitioning an incompatible reaction event into a host interior, where it will no longer interfere

7.5 Tandem Reactions

with other reactions. This approach is inspired by enzymatic catalysis, in which incompatible reactions occur in active sites that are isolated from other reaction processes in solution. Even in vitro, extremely impressive tandem processes are possible using enzymatic catalysis. For instance, the one-pot synthesis of the polyketide natural product enterocin from simple precursors benzoic acid and malonyl-CoA was accomplished. The tandem process forms 10 C C bonds, 5 C O bonds, and 7 stereocenters, is catalyzed by 12 purified enzymes, and proceeds in 25% overall yield [134]. 7.5.1 Synthetic Tandem Reactions

An impressive three-reaction tandem process enabled by supramolecular encapsulation was recently disclosed by the Nitschke group [135]. In this process, furan first reacts with singlet oxygen (photogenerated by methylene blue) to form an endoperoxide, which is converted to fumaraldehydic acid in a step catalyzed by encapsulation within a self-assembled metal–organic host. Finally, the proline-catalyzed aldol reaction between nitromethane and fumaraldehydic acid yields the final lactone product in 30% yield (Scheme 7.28). All reactants and catalysts are present at the beginning of the reaction, which proceeds without any of the three reaction cycles interfering with each other. Not only does encapsulation within the cage catalyze the endoperoxide rearrangement, but it also suppresses nonproductive reaction pathways that occur when the cage is absent. It is also noteworthy that the cage itself self-assembles in the reaction mixture, and that the process occurs in water.

Scheme 7.28

As described in Section 3.1, a large number of transition-metal-catalyzed reactions can be conducted in water within the hydrophobic core of a selfassembled micelle. Except for the catalysts and reagents, the conditions required are nearly identical for each reaction. The generality of these reaction conditions has enabled the Lipshutz group to design several multireaction tandem processes

161

162

7 Supramolecular Catalysis as a Tool for Green Chemistry

in water, which proceed in a single reaction flask and require no purification of intermediates [111,113,136]. For instance, the diamination of 1-iodo-4-bromobenzene was accomplished by an initial installation of a carbamate group, followed by a subsequent coupling with a second carbamate (Scheme 7.29). Both amination events are facilitated by the same catalyst, and the second reaction event is conducted by simply adding the second carbamate and increasing the reaction temperature.

Scheme 7.29

Micellar conditions allow for the formation of C C bonds using the Suzuki or Sonogashira coupling of 1-iodo-4-bromobenzene can be followed by a Pdcatalyzed amination step, furnishing the difunctionalized product (Scheme 7.30). In this case, it is worth noting that the reactions proceed with good yields, despite requiring two different Pd catalysts for the two coupling steps.

Scheme 7.30

7.5.2 Chemoenzymatic Tandem Reactions

The importance of enzymes in organic synthesis is growing [137–140], and given that enzymes are particularly suited for tandem processes, the use of supramolecular encapsulation to enable chemoenzymatic tandem reactions under biological conditions is desirable. Bergman and Raymond have reported two tandem chemoenzymatic systems involving a self-assembled metal–ligand host [141]. In the first example, the initial allenic ester or amide is hydrolyzed by an esterase or lipase, followed by allene hydroalkoxylation catalyzed by an encapsulated gold

7.5 Tandem Reactions

complex (Scheme 7.31). This one-pot process proceeds in water affords product tetrahydrofurans in high yield. Supramolecular encapsulation prevents unwanted interactions between the gold catalyst and enzyme; in the absence of the host, both catalytic reactions are negatively impacted.

Scheme 7.31

A second chemoenzymatic tandem process involves an allyl alcohol isomerization catalyzed by an encapsulated Ru(II) catalyst followed by NADPH-dependent enzymatic reduction of the resulting aldehyde (Scheme 7.32). Overall, the process represents a formal alkene reduction. Cofactor regeneration was accomplished by a second enzyme, allowing sodium formate to be the terminal reductant instead of the expensive NADPH. In tandem processes involving synthetic reactions, incorporating additional reaction cycles is extremely challenging, if not impossible. In this example, however, the second cofactor recycling enzyme was added without any change in reaction conditions. The design of more complex chemoenzymatic processes will no doubt be aided by supramolecular substrate gating and other forms of reaction control [23,86] to minimize crosstalk between reaction cycles.

Scheme 7.32

163

164

7 Supramolecular Catalysis as a Tool for Green Chemistry

7.6 Conclusion

Supramolecular catalysis is still in an early stage of development, and much work has been dedicated to establishing proof-of-concept rather than developing practical processes. However, the examples in this chapter demonstrate how supramolecular binding can be applied as a tool to achieve reaction outcomes that are desirable from a green chemistry standpoint. Indeed, the past decade has seen the productive application of supramolecular binding and encapsulation toward synthetically useful and green reactions, particularly in the area of micellar catalysis, directed C–H functionalization, and hydroformylation. As supramolecular chemistry concepts continue to be adopted by the wider synthetic community, further practical and green applications can be expected. Finally, although using supramolecular protection to enable tandem reactions is still a research area in the earliest stages of development, it holds great potential for reducing the time, waste, and energy involved in chemical synthesis. Particularly if one considers the integration of synthetic and enzymatic chemistry into chemoenzymatic tandem processes, it is now possible to imagine a truly ideal reaction process in which a desired product is made from simple reactants in a single operation at ambient temperature using water as the solvent. Again, it will be the adoption of these concepts by the wider synthetic community that will lead to practical and impactful new processes.

References 1 Anastas, P.T. and Warner, J.C. (1998)

2

3 4

5

6 7 8

Green Chemistry: Theory and Practice, Oxford University Press, New York. Lancaster, M. (2002) Green Chemistry: An Introductory Text, Royal Society of Chemistry, Cambridge. Wender, P.A. and Miller, B.L. (2009) Nature, 460, 197. Wender, P.A., Verma, V.A., Paxton, T.J., and Pillow, T.H. (2007) Accounts of Chemical Research, 41, 40–49. Lehn, J.M. (1990) Angewandte Chemie International Edition in English, 29, 1304. Cram, D.J. (1988) Angewandte Chemie International Edition in English, 27, 1009. Lehn, J.M. (1985) Science, 227, 849. Nguyen, S., Gin, D., Hupp, J., and Zhang, X. (2001) Proceedings of the National Academy of Sciences of the United States of America, 98, 11849.

9 Hof, F., Craig, S., Nuckolls, C., and Rebek,

10 11 12

13

14

15 16

J. (2002) Angewandte Chemie International Edition in English, 41, 1488. Pedersen, C.J. (1998) Angewandte Chemie International Edition in English, 27, 1021. Pedersen, C.J. (1967) Journal of the American Chemical Society, 89, 2495. Chakrabarty, R., Mukherjee, P.S., and Stang, P.J. (2011) Chemical Reviews, 111, 6810–6918. Northrop, B.H., Zheng, Y.-R., Chi, K.-W., and Stang, P.J. (2009) Accounts of Chemical Research, 42, 1554. Caulder, D. and Raymond, K. (1999) Journal of the Chemical Society, Dalton Transactions, 1999, 1185. Rebek, C.J. (1997) Chemical Reviews, 97, 1647. Saalfrank, R.W., Maid, H., and Scheurer, A. (2008) Angewandte Chemie International Edition in English, 47, 8794.

References 17 Singh, N., Tena-Solsona, M., Miravet, J.F.,

18 19

20 21

22 23

24 25

26

27 28

29 30

31

32

33

34

35

36

and Escuder, B. (2015) Israel Journal of Chemistry, 55, 711. Ohmatsu, K. and Ooi, T. (2015) Tetrahedron Letters, 56, 2043–2048. Leenders, S.H.A.M., Gramage-Doria, R., de Bruin, B., and Reek, J.N.H. (2015) Chemical Society Reviews, 44, 433. La Sorella, G., Strukul, G., and Scarso, A. (2015) Greem Chemistry, 17, 644. Brown, C.J., Toste, F.D., Bergman, R.G., and Raymond, K.N. (2015) Chemical Reviews, 115, 3012. Assaf, K.I. and Nau, W.M. (2015) Chemical Society Reviews, 44, 394. Zarra, S., Wood, D.M., Roberts, D.A., and Nitschke, J.R. (2014) Chemical Society Reviews, 44, 419. Yang, C. and Inoue, Y. (2014) Chemical Society Reviews, 43, 4123. Raynal, M., Ballester, P., Vidal-Ferran, A., and van Leeuwen, P.W.N.M. (2014) Chemical Society Reviews, 43, 1734. Raynal, M., Ballester, P., Vidal-Ferran, A., and van Leeuwen, P.W.N.M. (2014) Chemical Society Reviews, 43, 1660. Kataev, E.A. and Müller, C. (2014) Tetrahedron, 70, 137. Hapiot, F., Bricout, H., Menuel, S., Tilloy, S., and Monflier, E. (2014) Catalysis Science and Technology, 4, 1899. Marchetti, L. and Levine, M. (2011) ACS Catalysis, 1, 1090. Dong, Z., Wang, Y., Yin, Y., and Liu, J. (2011) Current Opinion in Colloid and Interface Science, 16, 451. Yoshizawa, M., Klosterman, J.K., and Fujita, M. (2009) Angewandte Chemie International Edition in English, 48, 3418. Koblenz, T.S., Wassenaar, J., and Reek, J.N.H. (2008) Chemical Society Reviews, 37, 247. Oshovsky, G.V., Reinhoudt, D.N., and Verboom, W. (2007) Angewandte Chemie International Edition in English, 46, 2366. Wilkinson, M., van Leeuwen, P., and Reek, J. (2005) Organic and Biomolecular Chemistry, 3, 2371. Vriezema, D., Aragones, M., Elemans, J., Cornelissen, J., Rowan, A., and Nolte, R. (2005) Chemical Reviews, 105, 1445. Lutzen, A. (2005) Angewandte Chemie International Edition in English, 44, 1000.

37 Dwars, T., Paetzold, E., and Oehme, G.

38 39 40

41 42

43

44

45

46

47

48

49

50

51

52 53

54 55

(2005) Angewandte Chemie International Edition in English, 44, 7174. Breslow, R. (1995) Accounts of Chemical Research, 28, 146. Lehn, J.M. (1994) Applied Catalysis A, 113, 105. Uemura, T., Yanai, N., and Kitagawa, S. (2009) Chemical Society Reviews, 38, 1228. Ma, L., Abney, C., and Lin, W. (2009) Chemical Society Reviews, 38, 1248. Lee, J., Farha, O.K., Roberts, J., Scheidt, K.A., Nguyen, S.T., and Hupp, J.T. (2009) Chemical Society Reviews, 38, 1450. Suslick, K.S., Bhyrappa, P., Chou, J.-H., Kosal, M.E., Nakagaki, S., Smithenry, D.W., and Wilson, S.R. (2005) Accounts of Chemical Research, 38, 283. Kesanli, B. and Lin, W. (2003) Coordination Chemistry Reviews, 246, 305. Lensen, D., Vriezema, D.M., and van Hest, J.C.M. (2008) Macromolecular Bioscience, 8, 991. Helms, B., Liang, C.O., Hawker, C.J., and Frechet, J.M.J. (2005) Macromolecules, 38, 5411. Liu, L., Rozenman, M., and Breslow, R. (2002) Journal of the American Chemical Society, 124, 12660. Liu, L. and Breslow, R. (2002) Journal of the American Chemical Society, 124, 4978. Chi, Y., Scroggins, S.T., and Fréchet, J.M.J. (2008) Journal of the American Chemical Society, 130, 6322. Liu, L. and Breslow, R. (2003) Journal of the American Chemical Society, 125, 12110. Diederich, F. and Felber, B. (2002) Proceedings of the National Academy of Sciences of the United States of America, 99, 4778. Hoveyda, A.H., Evans, D.A., and Fu, G.C. (1993) Chemical Reviews, 93, 1307–1370. Rousseau, G. and Breit, B. (2011) Angewandte Chemie International Edition in English, 50, 2450. Breslow, R. (1980) Accounts of Chemical Research, 13, 170. Breslow, R. and Maresca, L.M. (1978) Tetrahedron Letters, 19, 887–890.

165

166

7 Supramolecular Catalysis as a Tool for Green Chemistry 56 Breslow, R. and Maresca, L.M. (1977) 57

58

59

60

61

62

63

64

65 66 67

68

69

70 71

72

73

74

Tetrahedron Letters, 18, 632–626. Breslow, R. and Winnik, M.A. (1969) Journal of the American Chemical Society, 91, 3083. Breslow, R., Zhang, X., and Huang, Y. (1997) Journal of the American Chemical Society, 119, 4535. Das, S., Incarvito, C.D., Crabtree, R.H., and Brudvig, G.W. (2006) Science, 312, 1941. Frost, J.R., Huber, S.M., Breitenlechner, S., Bannwarth, C., and Bach, T. (2014) Angewandte Chemie International Edition in English, 54, 691–695. Kuninobu, Y., Ida, H., Nishi, M., and Kanai, M. (2015) Nature Chemistry, 7, 712. Maturi, M.M. and Bach, T. (2014) Angewandte Chemie International Edition in English, 53, 7661. Alonso, R. and Bach, T. (2014) Angewandte Chemie International Edition in English, 53, 4368. Müller, C., Bauer, A., and Bach, T. (2009) Angewandte Chemie International Edition in English, 48, 6640. Bauer, A., Westkämper, F., Grimme, S., and Bach, T. (2005) Nature, 436, 1139. Tan, K.L. (2011) ACS Catalysis, 1, 877–886. Yeung, C.S. and Dong, V.M. (2011) Angewandte Chemie International Edition in English, 50, 809. Lightburn, T.E., Dombrowski, M.T., and Tan, K.L. (2008) Journal of the American Chemical Society, 130, 9210. Grünanger, C.U. and Breit, B. (2008) Angewandte Chemie International Edition in English, 47, 7346. Joe, C.L. and Tan, K.L. (2011) Journal of Organic Chemistry, 76, 7590. Worthy, A.D., Gagnon, M.M., Dombrowski, M.T., and Tan, K.L. (2009) Organic Letters, 11, 2764. Slagt, V.F., Reek, J., and Kamer, P. (2001) Angewandte Chemie International Edition in English, 40, 4271–4274. Slagt, V.F., van Leeuwen, P.W.N.M., and Reek, J.N.H. (2003) Angewandte Chemie International Edition in English, 42, 5619. Shenoy, S.R., Pinacho Crisóstomo, F.R., Iwasawa, T., and Rebek, J. (2008) Journal

75

76

77

78

79 80

81 82 83 84 85 86

87 88 89 90 91

92 93

94

of the American Chemical Society, 130, 5658. Merlau, M., Mejia, M., Nguyen, S., and Hupp, J. (2001) Angewandte Chemie International Edition in English, 40, 4239. Lee, S.J., Cho, S.-H., Mulfort, K.L., Tiede, D.M., Hupp, J.T., and Nguyen, S.T. (2008) Journal of the American Chemical Society, 130, 16828. Otte, M., Kuijpers, P.F., Troeppner, O., Ivanovic ́-Burmazovic ́, I., Reek, J.N.H., and de Bruin, B. (2014) Chemistry – A European Journal, 20, 4880. La Sorella, G., Canton, P., Strukul, G., and Scarso, A. (2014) ChemCatChem, 6, 1575. Trentin, F., Scarso, A., and Strukul, G. (2011) Tetrahedron Letters, 52, 6978. La Sorella, G., Bazan, M., Scarso, A., and Strukul, G. (2013) Journal of Molecular Catalysis, 379, 192–196. Breslow, R. (2013) Green Solvents: Reactions in Water, vol. 5, Wiley-VCH. Sheldon, R.A. (2007) Greem Chemistry, 9, 1273. Butler, R.N. and Coyne, A.G. (2010) Chemical Reviews, 110, 6302. Breslow, R. (1982) Science, 218, 532. Kirby, A. (1996) Angewandte Chemie International Edition in English, 35, 708. Wiester, M.J., Ulmann, P.A., and Mirkin, C.A. (2010) Angewandte Chemie International Edition in English, 50, 114. Bistri, O. and Reinaud, O. (2015) Organic and Biomolecular Chemistry, 13, 2849. Lipshutz, B.H. and Ghorai, S. (2012) Aldrichimica Acta, 45, 3. Lipshutz, B.H. and Ghorai, S. (2008) Aldrichimica Acta, 48, 59. Kobayashi, S. and Hachiya, I. (1992) Tetrahedron Letters, 33, 1625. Kobayashi, S. and Hachiya, I. (1994) Journal of Organic Chemistry, 59, 3590. Kobayashi, S. and Nagayama, S. (1998) Journal of the American, 120, 8287–8288. Kobayashi, S., Mori, Y., Nagayama, S., and Manabe, K. (1999) Greem Chemistry, 1, 175. Manabe, K., Mori, Y., Wakabayashi, T., Nagayama, S., and Kobayashi, S. (2000) Journal of the American Chemical Society, 122, 7202.

References 95 Manabe, K.M. and Kobayashi, S. (1999) 96

97

98

99 100

101 102

103 104 105

106

107 108

109

110

111 112 113

Organic Letters, 1, 1965–1967. Manabe, K. and Kobayashi, S. (2000) Chemical Communications (Cambridge, UK), 669. Mori, Y., Kakumoto, K., Manabe, K., and Kobayashi, S. (2000) Tetrahedron Letters, 41, 3107–3111. Manabe, K., Aoyama, N., and Kobayashi, S. (2001) Advanced Synthesis and Catalysis, 343, 174–176. Azoulay, S., Manabe, K., and Kobayashi, S. (2005) Organic Letters, 7, 4593. Kobayashi, S., Nagayama, S., and Busujima, T. (1999) Tetrahedron, 55, 8739–8746. Klumphu, P. and Lipshutz, B.H. (2014) Journal of Organic Chemistry, 79, 888. Lipshutz, B.H., Ghorai, S., Abela, A.R., Moser, R., Nishikata, T., Duplais, C., Krasovskiy, A., Gaston, R.D., and Gadwood, R.C. (2011) Journal of Organic Chemistry, 76, 4379. Lipshutz, B.H. and Abela, A.R. (2008) Organic Letters, 10, 5329. Lipshutz, B.H. and Taft, B.R. (2008) Organic Letters, 10, 1329. Nishikata, T. and Lipshutz, B.H. (2009) Chemical Communications (Cambridge, U.K.), 6472. Nishikata, T., Abela, A.R., and Lipshutz, B.H. (2010) Angewandte Chemie International Edition in English, 49, 781. Nishikata, T. and Lipshutz, B.H. (2010) Organic Letters, 12, 1972. Lipshutz, B.H., Ghorai, S., Leong, W.W.Y., Taft, B.R., and Krogstad, D.V. (2011) Journal of Organic Chemistry, 76, 5061. Lipshutz, B.H., Isley, N.A., Moser, R., Ghorai, S., Leuser, H., and Taft, B.R. (2012) Advanced Synthesis and Catalysis, 354, 3175. Isley, N.A., Gallou, F., and Lipshutz, B.H. (2013) Journal of the American Chemical Society, 135, 17707. Isley, N.A., Dobarco, S., and Lipshutz, B.H. (2014) Greem Chemistry, 16, 1480. Handa, S., Slack, E.D., and Lipshutz, B.H. (2015) Angewandte Chemie, 127, 12162. Handa, S., Wang, Y., Gallou, F., and Lipshutz, B.H. (2015) Science, 349, 1087–1091.

114 Kelly, S.M. and Lipshutz, B.H. (2014)

Organic Letters, 16, 98. 115 Gabriel, C.M., Keener, M., Gallou, F., and

116

117 118

119

120

121

122

123

124

125

126

127

128 129

130

Lipshutz, B.H. (2015) Organic Letters, 17, 3968. Isley, N.A., Linstadt, R.T.H., Kelly, S.M., Gallou, F., and Lipshutz, B.H. (2015) Organic Letters, 17, 4734. Bruce, H. and Lipshutz, S.G. (2014) Greem Chemistry, 16, 3660. Gallou, F., Isley, N.A., Ganic, A., Onken, U., and Parmentier, M. (2016) Greem Chemistry, 18, 14. Manabe, K., Iimura, S., Sun, X.-M., and Kobayashi, S. (2002) Journal of the American Chemical Society, 124, 11971. Hart-Cooper, W.M., Zhao, C., Triano, R.M., Yaghoubi, P., Ozores, H.L., Burford, K.N., Toste, F.D., Bergman, R.G., and Raymond, K.N. (2015) Chemical Science, 6, 1383. Hart-Cooper, W.M., Clary, K.N., Toste, F.D., Bergman, R.G., and Raymond, K.N. (2012) Journal of the American Chemical Society, 134, 17873. Fiedler, D., Bergman, R.G., and Raymond, K.N. (2006) Angewandte Chemie International Edition in English, 45, 745. Dong, V.M., Fiedler, D., Carl, B., Bergman, R.G., and Raymond, K.N. (2006) Journal of the American Chemical Society, 128, 14464. Brumaghim, J., Michels, M., and Raymond, K. (2004) European Journal of Organic Chemistry, 2004, 4552. Brumaghim, J., Michels, M., Pagliero, D., and Raymond, K. (2004) European Journal of Organic Chemistry, 5115. Ziegler, M., Brumaghim, J., and Raymond, K. (2000) Angewandte Chemie International Edition in English, 39, 4119. Murase, T., Nishijima, Y., and Fujita, M. (2012) Journal of the American Chemical Society, 134, 162. Egi, M., Azechi, K., and Akai, S. (2009) Organic Letters, 11, 5002. Aponick, A., Li, C.-Y., Malinge, J., and Marques, E.F. (2009) Organic Letters, 11, 4624. Minkler, S.R.K., Isley, N.A., Lippincott, D.J., Krause, N., and Lipshutz, B.H. (2014) Organic Letters, 16, 724.

167

168

7 Supramolecular Catalysis as a Tool for Green Chemistry 131 Brown, C.J., Miller, G.M., Johnson, M.W.,

132

133

134

135

Bergman, R.G., and Raymond, K.N. (2011) Journal of the American Chemical Society, 133, 11964. Lipshutz, B.H., Huang, S., Leong, W.W.Y., Zhong, G., and Isley, N.A. (2012) Journal of the American Chemical Society, 134, 19985. Fogg, D.E. and dos Santos, E.N. (2004) Coordination Chemistry Reviews, 248, 2365. Cheng, Q., Xiang, L., Izumikawa, M., Meluzzi, D., and Moore, B.S. (2007) Nature Chemical Biology, 3, 557. Salles, A.G., Zarra, S., Turner, R.M., and Nitschke, J.R. (2013) Journal of the American Chemical Society, 135, 19143.

136 Bhattacharjya, A., Klumphu, P., and

137 138

139

140

141

Lipshutz, B.H. (2015) Organic Letters, 17, 1122. Pàmies, O. and Bäckvall, J.-E. (2003) Chemical Reviews, 103, 3247. Leresche, J. and Meyer, H. (2006) Organic Process Research and Development, 10, 572. Meyer, H. (2011) Organic Process Research and Development, 15, 180–188. Clouthier, C.M. and Pelletier, J.N. (2012) Chemical Society Reviews, 41, 1585. Wang, Z.J., Clary, K.N., Bergman, R.G., Raymond, K.N., and Toste, F.D. (2013) Nature Chemistry, 5, 100.

169

8 A Tutorial of the Inverse Molecular Design Theory in Tight-Binding Frameworks and Its Applications Dequan Xiao and Rui Hu

8.1 Introduction

The principle #4 of Green Chemistry states, “chemical products should be designed to preserve of function while reducing toxicity” [1]. Thus, developing efficient computational methods for molecular design is a critically important step toward the successful development of green or sustainable chemicals. First, proper molecular design can avoid the use of toxic chemicals. For example, toxic metals may be avoided for use in catalysis, and be replaced by earthabundant metals with low toxicity. Second, molecular design can lead us to an optimal material in an efficient way. Materials discovery process is time-consuming and usually takes many trial-and-errors. Developing effective molecular design methods may take much fewer trials to reach the desired materials, and thus avoid the waste of chemicals and materials. In this chapter, we provide a tutorial on the inverse molecular design theory and its applications in materials design and discovery. In the followings, we will introduce the inverse molecular design in tight-binding frameworks, as shown in Section 8.2. We will provide detailed description on how to define molecular frameworks (Section 8.3) and select optional functional groups (Section 8.4). Finally, we will show four examples on using the inverse molecular design programs to search for optimal linear and nonlinear optical materials and photosensitizers for solar cell applications. Due to the low computational cost of tight-binding electronic structure calculations, we anticipate that the inverse molecular design theory in tight-binding frameworks will be a promising tool for attacking challenges in materials discovery such as catalysts design for renewable energy applications.

Handbook of Green Chemistry Volume 10: Tools for Green Chemistry, First Edition. Edited by Evan S. Beach and Soumen Kundu.  2017 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2017 by Wiley-VCH Verlag GmbH & Co. KGaA.

170

8 A Tutorial of the Inverse Molecular Design Theory in Tight-Binding Frameworks and Its Applications

8.2 Inverse Molecular Design Theory in Tight-Binding Frameworks

Designing materials with optimum properties is a long-term dream for both experimental and theoretical researchers. Historically, researchers rely on the approach of “trial and error” to find new materials with desired properties. Owing to the development in modern theoretical and computational chemistry (e.g., density functional theory [2]), predicting molecular properties using accurate and efficient quantum chemistry methods becomes more and more practical. As a result, inverse molecular design theory is emerged as an attractive tool to tackle the challenges in materials design. Inverse molecular design is a general term for the strategies in molecular design, relative to the strategy of direct design. In direct design, a new molecule is proposed first, and then the molecular property is computed or tested to check its feasibility. In contrast, inverse molecular design theory aims to search for the optimum points in the hypersurface of the property–structure relationship, and then map out the molecular structures at the optimum points [3]. Hence, using the idea of inverse molecular design could significantly enhance the success rate of molecular design, thus leading to successful materials discovery. Inverse molecular design was formulated as an optimization problem in theory, that is, searching for the optimum chemical structures with the aid of optimization algorithms. f inv  min jO‰H ‰λ1 ; λ2 ; . . .ŠŠ λ1 ;λ2 ;...

O∗ j

(8.1)

Here, finv is a notation for the operation of inverse molecular design. O denotes a molecular property (an observable), which is a functional of the Hamiltonian H. λ1, λ2, and so on are a set of variables of the Hamiltonian through which different molecular structures are represented. O∗ denotes the target molecular property, for example, a maximum point of the molecular property. Notation “min” means adapting different optimization algorithms for minimizing the absolute value of O‰H ‰λ1 ; λ2 ; . . .ŠŠ O∗ . Thus, finv aims to find a particular set of variables λ1, λ2, . . . (and thus a molecular structure) that shows the closest matching to the target molecular property. In the molecular structure space, the Hamiltonian variables λ1, λ2, and so on are associated with the atoms types and their connectivity. With the choice of the Hamiltonian variables, that is, the way of varying molecular structures in a chemical structure space, different optimization algorithms including stochastic and deterministic search methods have been adapted for inverse molecular design [3]. Adapting effective and efficient optimization algorithms is the key to successful searches of optimal molecules out of all the possible molecular structures (i.e., molecular space), as the general molecular space is so large. For examples, the number of accessible drug-like molecules is estimated to be larger than 1060;

8.2 Inverse Molecular Design Theory in Tight-Binding Frameworks

only approximately 108 of these structures have been synthesized [4,5]. The estimated cost of high-throughput screening to assess the viability for drug activity of known structures is in the tens of billions of dollars [5]. Collecting this significant body of data would only begin to mine the vastness of molecular space, one goal of molecular design. Strategies of combinatorial chemistry provide additional means to explore the richness of molecular structure in real and virtual libraries [6–11]. A very large number of solid-state materials can be imagined as targets of optimization as well [12]. An open theoretical challenge is to establish strategies to explore molecular space effectively, in order to discover molecular structures and materials with the most useful properties. Current optimization strategies to discover optimal molecular structures are of two basic kinds: those that examine molecules “one by one” (i.e., stochastic methods such as like genetic algorithm [13] and Monte Carlo methods) and those that employ continuous (i.e., deterministic) optimization strategies [14,15]. In order to deterministically search for molecules, Beratan and Yang introduced a continuous optimization strategy based on the approach of linear combination of atomic potential (LCAP) for inverse molecular design in 2006 [14]. The LCAP approach provides a continuous interpolation between discrete molecular structures. This characteristic distinguishes the approach from an earlier strategy [16]. As a result, a continuous molecular property hypersurface with regarded to the chemical structure variations is constructed. Such a continuous property surface allows property gradients to be defined and evaluated, leading to efficient structure optimization. In 2008, Xiao et al. first adapted the LCAP principle into the Hückel tightbinding framework [17], denoted TB-LCAP. The tight-binding strategy defines an effective Hamiltonian (matrix) to represent the electronic structure. The TBLCAP method was proved to be efficient when searching for optimum structures in large molecular spaces including 106–1019 possible molecules. In 2011, Xiao et al. extended the TB-LCAP approach into the extended Hückel tight-binding framework, and used it to optimize photosensitizers for dye-sensitized solar cell applications, and the optimized lead chromophore was then proved in experiments [18]. In the following, we first describe the LCAP principle in the framework of density functional theory, and then we will describe the LCAP principle in the tight-binding frameworks, which provides a fast way for molecule search. 8.2.1 LCAP Principle in Density Functional Theory

In density functional theory [2], the Schrodinger equation is ‰T ‡ vext …r† ‡ vH …r† ‡ vxc …r†Šϕn …r† ˆ εn ϕn …r†

(8.2)

r where T ˆ 2m is the kinetic energy operator, vext …r† is the external potential operator, vH …r† is the Hartree energy operator, and vxc …r† is the exchangecorrelation energy operator, ϕn …r† is the eigen-function, and εn is the eigen-energy. 2

171

172

8 A Tutorial of the Inverse Molecular Design Theory in Tight-Binding Frameworks and Its Applications

Given the option of a few atom type (or functional groups) in one or more atomic sites, the external potential is a linear combination of atomic potential [14], vext …r† ˆ

X

bRA vRA …r†

(8.3)

R;A

Here, vRA …r† is the external potential of atom or group type A at position R. Specifically, if A represents an atom type, vRA …r† ˆ Z RA =jr Rj where Z RA is the atomic number of atom A. The optimization coefficients bRA define the weighting of a particular atom or group type, and are varied during the optimization. For real molecules, one bRA value must be equal to one and all others must be zero for the site R. That is, no more than one chemical unit may exist at any position P R. The constraints on bRA during the optimization are A bRA ˆ 1 and 0  bRA  1. The potential v…r† and the number of electrons completely specify the input for the ground-state electronic structure and its properties. The external potential establishes a space on which optimization/search is conducted to design molecules. When the optimum is reached, in a continuous optimization, the bRA values are rounded to zero or one. Similar in spirit, a method of variational particle numbers was developed for inverse molecular design. In this method, the Hamiltonian variables are directly defined by the number of electrons N, atom-types Z, and spatial configurations R [15]. 8.2.2 LCAP Principle in Tight-Binding Frameworks 8.2.2.1

One-Orbital Tight-Binding Framework

The LCAP principle has been implemented in a Hückel tight-binding framework by Xiao et al. [17], denoted TB-LCAP. The tight-binding strategy defines an effective Hamiltonian (matrix) to represent the electronic structure. The bRA coefficients are parameterized in the tight-binding Hamiltonian matrix. Since molecular electronic properties are determined by the Hamiltonian matrix, optimizing the property is achieved by optimizing the bRA coefficients. In the tight-binding framework, the LCAP principle is implemented using a linear combination of Hamiltonian matrix elements. In the Hückel approximation, only π-electrons are included, basis states are assumed orthonormal, and only nearest neighbor sites interact. The Hückel Hamiltonian matrix is [19] 2

H 11

6 6 6 H ˆ 6 ... 6 4

O

3

∙∙∙

H ij H ii

H ji ∙∙∙

O

7 7 .. 7 . 7 7 5 H nn

(8.4)

8.2 Inverse Molecular Design Theory in Tight-Binding Frameworks ´

Table 8.1 Hückel site energies (h(A)) and nearest neighbor interaction energies (h(A,A )) [20]. ´

h(A,A ) (eV)

h(A) (eV) C C

0.00

N

P

1.00

N

0.51

1.02

1.09

P

0.19

0.77

0.78

0.63

where i and j are indices of atomic sites, and n equals the total number of atomic sites in the molecule. For these n sites, we search for the optimal atom (or group) type for some (or all) of these sites for the target property. For each variable site, there are N type atom or group choices. Hückel theory is a π-electron theory, so only one pz basis function appears per site. The H matrix element indices (i or j) are associated with atoms and the atomic position R. The coefficients bRA in Eq. (8.3) may be expanded to include the biA coefficients for each site. The site energy Hii is determined by the available atoms for a site, so the LCAP value of this matrix element is …var† H ii

N itype

ˆ

X

…A†

biA hii

(8.5)

Aˆ1 …A†

where hii is the energy of atom type A at site i, biA is the LCAP weighting coefficient for atom type A on site i, and N itype is the total number of atom types possible at site i. For chemically fixed sites, the site energy equals the energy of j ˆ h…A† the atom type at that site: H …fixed† jj jj and N type ˆ 1. The interaction between sites i and j is j

N itype N type

H ij ˆ

XX Aˆ1 A0 ˆ1

´

biA bA´ hij…A;A † j

(8.6)

´

where hij…A;A † is the interaction strength between atom type A (at site i) and A´ (at site j). For an example, the Hückel parameters [20] for a few elements are shown in Table 8.1. The elements are C, N, and P. 8.2.2.2

Extended Hückel Tight-Binding Framework

The Extended Hückel approach has been extensively applied for calculations of electronic structures and photoabsorption spectra for a wide range of molecular and solid-state structures [21,22], as well as for studies of sensitized TiO2 surfaces including the analysis of interfacial electron transfer [23–26]. The extension of TB-LCAP into the extended Hückel tight binding frameworks allows us to search for optimum chemical structures beyond the π-conjugated molecules, for many different chemical properties.

173

174

8 A Tutorial of the Inverse Molecular Design Theory in Tight-Binding Frameworks and Its Applications

The time-independent Schrödinger equation is represented in matrix form, HQ ˆ ESQ

(8.7)

where H is the extended Hückel Hamiltonian in the basis of Slater-type atomic orbitals (STO’s), Q is the matrix of eigenvectors, E is the diagonal matrix of eigenstate energies, and S is the overlap matrix of the STOs. The (EH) LCAP Hamiltonian is defined in terms of the participation coefficients {b…A† i } for the complete set of atom-types A and sites i, giving the probability weight of such an atom type A at site i while undergoing a continuous transformation of each site through N itype PN i …A† possible atom types. The constraints are A type b…A† i ˆ 1 and 0  bi  1. Using the participation coefficients b…A† i , the diagonal matrix elements of the EH/TB-LCAP Hamiltonian are defined as follows: N itype

H iα;iα ˆ

X

…A† b…A† i hiα;iα

(8.8)

Aˆ1

where α represents a STO of atom type A and h…A† iα;iα is the EH diagonal Hamiltonian matrix element for atom type A at site i. Note that for the specific case of ˆ 1, H iα;iα ˆ h…A† b…A† i iα;iα since all other participation coefficients at the i-site are equal to zero. More generally, H iα;iα is an arithmetic average of EH matrix elements associated with all possible atom types at site i weighted by their corresponding participation coefficients. Analogously, the off-diagonal matrix elements are defined as follows: j

N itype N type

H iα;j⠈

XX Aˆ1 A´ ˆ1

´

´

…A † …A;A † b…A† i bj hiα;jβ

(8.9) ´

…A;A † the with β representing the atomic orbital β of atom type A´ at site j, and hiα;jβ original off-diagonal Hamiltonian element for the case of atom type A at site i, and atom type A´ at site j:   ´ …A;A´ † † …A´ † (8.10) hiα;j⠈ K ´ S …A;A h…A† iα;iα ‡ hjβ;jβ =2 iα;jβ

where K = 1.75 and K ´ ˆ K ‡ Δ2 ‡ Δ4 …1

K † is defined according to the original     …A´ † …A´ † Wolfsberg–Helmholz formula [27], with Δ ˆ H …A† H jβ;jβ = H …A† iα;iα iα;iα ‡ H jβ;jβ . ´

…A;A † for atom types A at site i and The off-diagonal overlap matrix elements S iα;jβ atom types A´ at site j are defined as follows: j

N itype N type

S iα;j⠈

XX Aˆ1 A´ ˆ1

´

´

…A † …A;A † b…A† i bj S iα;jβ

(8.11)

The participation coefficients are initialized randomly. These coefficients are subsequently optimized with respect to the property of interest (e.g., solar light absorption) computed from the solar spectrum and the´ oscillator strengths of …A † …A;A´ † , h , and h as defined allowed transitions obtained by using elements h…A† iα;iα jβ;jβ iα;jβ

8.3 How to Prepare a Molecular Framework for TB-LCAP Inverse Design?

by the EH Hamiltonian, with STO parameters reported by Hoffmann et al. [27] without any further corrections. 8.2.3 Gradient for Optimization

We optimize molecular properties, such as hαi…π† , β…==π† , μ2HL , and E gap using a continuous searching LCAP approach with the quasi-Newton method [28]. The gradients are calculated numerically using the finite-difference   f …‡δbiA † f … δbiA † @f ˆ (8.12) @biA 2δbiA Here, the function f represents any of the four property functions and δbiA represents variation of biA . At the end of the optimization, biA values are rounded to 0 or 1.

8.3 How to Prepare a Molecular Framework for TB-LCAP Inverse Design? 2D Molecular Frameworks

For the Hückel TB-LCAP approach, 2D molecular framework (π-conjugated structures) can be used for the search. Figure 8.1 is a typical 2D framework based on the anthracene molecule. As the first step, we label all atomic sites (where C atoms are located) by sequential integer numbers. We then define the connectivity in a 2D framework as shown in Figure 8.2. The first line represents the total number of the atoms. The second line represents the total number of sites that are allowed to change. The third line represents the properties of the first site. For input “1 2 1,” the first number means the first site, the second number means that the first site is connected to two other sites; the third number means that the first site is allowed to vary. If it is not variable, then it is a “0.” The forth line represents the two sites that the first is connected to. In the following, we will just repeat the format similar to lines 3–4 for the rest of the sites. Third, we will prepare a geometry file (see Figure 8.3) for the 2D molecular framework. The first line represents to the number of dimensions for the coor-

Figure 8.1 Illustration of the numbering of atomic sites in 2D molecular framework based on anthracene.

175

176

8 A Tutorial of the Inverse Molecular Design Theory in Tight-Binding Frameworks and Its Applications

Figure 8.2 Illustration of the connectivity file for the molecular framework in Figure 8.1.

Figure 8.3 Illustration of the geometry file for the molecular framework in Figure 8.1.

8.4 How to Choose Optional Atom Types or Functional Groups?

Figure 8.4 Illustration of the numbering of atomic sites of in a 3D framework based on phenylacac.

dinates. For a structure having only x and y coordinates, the number is 2. The following lines are the x and y coordinates for each sites. 3D Molecular Framework

For the extended Hückel TB-LCAP approach, 3D molecular framework can be used. Figure 8.4 shows an example of a 3D molecular framework, where all the atoms are numbered in a sequential order. Since this molecular framework is not limited to π-conjugated structures, we need to number all the atoms including hydrogen atoms. A geometry file is then prepared (see Figure 8.5) for the 3D molecular framework. The first line indicates the total number of atoms in the molecular framework. The second line represents the net charge of the molecular framework. After that, all the atoms are listed following the order of the numbering in Figure 8.4, where x, y, and z coordinates are shown following the atomic symbols.

8.4 How to Choose Optional Atom Types or Functional Groups?

Once the molecular frameworks are prepared, optional functional groups can be chosen based on the particular molecular framework. As an example, a set of chosen functional groups in a 2D molecular framework are shown in Figure 8.6. Here, three optional atom types C, N, and P are chosen for atomic sites 4 and 11. And eight functional groups (CC, NN, CN, NC, CP, PC, NP, and PN) are chosen for diatomic sites 1–2, 6–7, 8–9, and 13–14. The input file for choosing these optional functional groups or atom types is shown in Figure 8.7. The text inside are for commenting purpose only here, which should not be included in the actual input file for the TB-LCAP program.

177

178

8 A Tutorial of the Inverse Molecular Design Theory in Tight-Binding Frameworks and Its Applications

Figure 8.5 Illustration of the geometry file for the 3D molecular framework in Figure 8.4.

Figure 8.6 Optional atom types (C, N, and P) in the circle sites, and optimal diatomic functional groups (CC, NN, CN, NC, CP, PC, NP, and PN) in the olive sites for the 2D framework.

For the 3D molecular framework in Figure 8.4, the choice of optional functional groups is shown in Figure 8.8. Four diatomic functional groups: C H, N X, O X, and S X are chosen for the atomic sites 5 and 8, and atomic sites 2 and 7, respectively, and three diatomic functional groups: C H, C O, and C S are chosen for the diatomic sites 6 and 9, and diatomic sites 3 and 23, respectively. Here, “X” denotes a dummy atom, which is an empty site in Figure 8.8. In Figure 8.8, you will learn how to write the selection into an input file for the TB-LCAP program to read upon implementation. The text inside are for

8.4 How to Choose Optional Atom Types or Functional Groups?

Figure 8.7 Illustration of the input file for choosing the optional atom types and functional groups shown in Figure 8.6.

179

180

8 A Tutorial of the Inverse Molecular Design Theory in Tight-Binding Frameworks and Its Applications

Figure 8.7 (Continued)

Figure 8.8 Optional diatomic function group types for the selected diatomic sites.

commenting purpose only here, which should not included in the actual input file for the TB-LCAP program (Figure 8.9). The input files here can also be prepared using the Web-based Interface shown in Figure 8.10. This web interface is for preparing the input file for 3D molecular framework. Optional atom types can be assigned for any particular

8.4 How to Choose Optional Atom Types or Functional Groups?

Figure 8.9 Illustration of the input file for choosing the optional atom types and functional groups shown in Figure 8.8.

181

182

8 A Tutorial of the Inverse Molecular Design Theory in Tight-Binding Frameworks and Its Applications

Figure 8.10 A web-based JAVA interface for setting up the calculations based on the TB-LCAP methods for materials design.

atom sites in your prepared molecular framework. The selection of diatomic functional groups is not allowed in the current version, but will be included in the future version. Detailed description for using the web-based JAVA interface can be found in the Web site of chemistry-hpc.newhaven.edu/tbimd.php. The available molecular properties for TB-LCAP searches are oscillator strength (1), HOMO-LUMO energy gap (2), and ground-state electronic energy (3).

8.5 Optimizing Molecular Properties Using the TB-LACP Methods

Once the TB-LCAP Hamiltonian is formulated, the molecular property functions can be computed from the TB-LCAP Hamiltonian. Following the gradient of the molecular properties with respect to the TB-LCAP coefficients, we can reach the optimum points of molecular property space. The chemical structures

8.5 Optimizing Molecular Properties Using the TB-LACP Methods

that are closest to the optimum properties are the sought structures for the inverse molecular design. In the followings, we will show how the TB-LCAP approaches were used to optimize the electronic polarizability, electronic hyperpolarizability, HOMO-LUMO energy gap, and photoabsorption due to the HOMO-LUMO transition. More details of these TB-LCAP searches can be found in our previous publications [17,18]. The optimizations of these molecular properties have important implications in real applications. For example, optimizing electronic hyperpolarizability can lead to the discovery of optimal nonlinear optical (NLO) materials. When photons interact with the NLO materials, the frequencies, phases, and other properties get altered. Because of their abilities to manipulate photonic signals, NLO materials are highly demanded in technologies such as optical communication, optical computing, and dynamic image processing [29,30]. For molecular design, materials with large NLO responses, such as the first hyperpolarizability [30], are desired to improve the efficiency of manipulating photonic signals, such as changing photon frequencies. Example 8.1: Optimizing Molecular Electronic Polarizability in a 2D Framework The molecular polarizability α components are defined as second derivatives of E with respect to electric filed F in directions i and j, αij ˆ @ 2 E=@F i @F j . We optimize the following molecular polarizability function [31]:  1 hαi…π† ˆ αxx ‡ αyy ‡ αzz (8.13) 3 where, the polarizability components (αxx , αyy , and αzz ) are computed using the finite-field methods: αii ˆ

2E…0†

‰E…F i † ‡ E… F i †Š F 2i

(8.14)

A typical search path using the TB-LCAP program is shown in Figure 8.11, for maximizing the electronic polarizability based on the molecular framework in Figure 8.6.

Figure 8.11 Illustration of a typical search path for maximizing electronic polarizability (α) based on the 2D framework in Figure 8.6.

183

184

8 A Tutorial of the Inverse Molecular Design Theory in Tight-Binding Frameworks and Its Applications

Example 8.2: Optimizing Molecular Electronic Hyperpolarizability in a 2D Framework The first hyperpolarizability β components are defined as the third derivatives of E with respect to F in the directions of i, j, and k, @3E @F i @F j @F k

βijk ˆ

(8.15)

We optimize the following objective function of molecular hyperpolarizability [31]: 1=2 1 2 βx ‡ β2y ‡ β2z (8.16) 5  P where βa ˆ j βajj ‡ βjaj ‡ βjja and the subscript a represents x, y, or z. The hyperpolarizability components are computed using the finite-field methods β…π† == ˆ

βiii ˆ

‰E…F i †

E… F i †Š ‡ 0:5‰E…2F i † F 3i

E… 2F i †Š

(8.17)

and  βijj ˆ

0:5E… F i ; F j †

E…F i ; F j † ‡ E… F i ; F j † F i F 2j

 E…F i ; F j † ‡ ‰E…F i †

E… F i †Š

(8.18)

The electronic energies E…0† and E…F i † (with field) are calculated from the trace of the Hamiltonian and the density matrices. The Hamiltonian matrix in the presence of the field is Hkl …F i † ˆ Hkl ‡ < kjr i jl > F i ;

(8.19)

Here, k and l are indices for the p-orbital basis functions. In the Hückel approach,  the transition dipole integral is approximated kjr i jl  Rki δkl , where Rki is the ithdirection Cartesian coordinate of the kth atom. In the finite-field method, the numerical value of the field F is taken to be 0.01 V/Å. A typical search path using the TB-LCAP program is shown in Figure 8.12, for maximizing the electronic first hyperpolarizability based on the molecular framework in Figure 8.6.

Figure 8.12 Illustration of a typical search path for maximizing electronic hyperpolarizability (β) based on the 2D framework in Figure 8.6.

8.5 Optimizing Molecular Properties Using the TB-LACP Methods

Example 8.3: Optimization of HOMO-LUMO Energy Gap in a 2D Framework The objective function of HOMO-LUMO energy gaps is computed as the difference between the LUMO energy and the HOMO energy. A typical search path using the TB-LCAP program is shown in Figure 8.13, for minimizing the HOMOLUMO energy gaps based on the molecular framework in Figure 8.6.

Figure 8.13 Illustration of a typical search path for minimizing HOMO-LUMO energy gap (Egap) based on the 2D framework in Figure 8.6.

Example 8.4: Optimizing Photoabsorption due to HOMO-LUMO Transition in a 3D Framework We optimized the oscillator strength fHL corresponding to the HOMO-LUMO electronic transitions. The oscillator

Estrength

Ef pq of correspond to the transition between electronic eigenstates ψp and ψq [32], f pq ˆ

8π 2 υpq cme



2 μpq 3he2

(8.20)

where υpq is the transition frequency in cm 1, c is the speed of light in vacuum, me is the D electron E mass, h is the Planck constant, e is the electron charge, and μpq ˆ ψq jr jψp is the transition dipole moment. A typical search path using the TB-LCAP program is shown in Figure 8.14, for maximizing the oscillator strength due to the HOMO-LUMO electronic transition based on the 3D molecular framework in Figure 8.8.

185

186

8 A Tutorial of the Inverse Molecular Design Theory in Tight-Binding Frameworks and Its Applications

Figure 8.14 Illustration of a typical search path for maximizing HOMO-LUMO transition oscillator strength based on the 3D framework in Figure 8.8. Reproduced with permission from Ref. [18]. Copyright 2016, American Chemical Society.

8.6 Conclusion

Due to the fast development of modern quantum chemistry calculations, inverse molecular design theory has emerged as an attractive tool to take on the challenges in effective molecular design of novel materials, which is consistent with one of the goals (Principle #4) of green chemistry [1]. Inverse molecular design theory aims at searching for optimum points on the hypersurfaces defining property–structure relationships, and mapping out the molecular structures at these points [3]. Historically, discrete optimization algorithms (such as genetic algorithm [13] and Monte Carlo methods) are used for inverse molecular design. Developed by Beratan and Yang in 2006 [33], a new inverse molecular design strategy was proposed based on the principle of linear combination of atomic potential. The LCAP approach provides continuous interpolation between discrete molecular structures, enabling us to search optimum molecules efficiently using deterministic or continuous optimization algorithms. We present the basic principle of LCAP in the framework of density functional theory, as originally proposed. Similar in spirit, another deterministic search methods called variational Hamiltonian was also developed in 2005 [15]. Developed by Xiao et al. [17], the LCAP strategy adapted in Hückel tightbinding frameworks allows us to effectively and efficiently search for optimum molecular guided by the gradients of molecular properties versus the LCAP coefficients. In 2011 [18], the TB-LCAP strategy was extended by Xiao et al. to the extended Hückel tight-binding frameworks, allowing us to search for almost any possible molecules beyond π-conjugated systems.

References

The implementation of the TB-LCAP methods include two key procedures: defining the molecular framework, and choosing optional functional groups. A web-based service program TB-IMD is created by the authors for running the TB-LCAP searches. Due to the low computational cost of tight-binding electronic structure calculations, we envision that the TB-LCAP approaches will play a key role on tackling the challenging problems of materials design (thus, leading to materials discovery) such as catalysts design in renewable energy applications.

References 1 Anastas, P.T. and Warner, J.C. (1998)

2 3

4

5

6 7

8

9 10 11

12

13 14

Green Chemistry: Theory and Practice, Oxford University Press, New York. Hohenberg, P. and Kohn, W. (1964) Physical Review, 136, B864. Xiao, D., Warnke, I., Bedford, J., and Bastista, V.S. (2014) RSC Specialist Periodical Report – Chemical Modelling 10, 1. Ertl, P. (2003) Journal of Chemical Information and Computer Science, 43, 374. Hann, M.M. and Opera, T.I. (2004) Current Opinion in Chemical Biology, 8, 255. An, H. and Cook, P.D. (2000) Chemical Reviews, 101, 3311. Ellman, J., Stoddard, B., and Wells, J. (1997) Proceedings of the National Academy of Sciences of the United States of America, 94, 2779. Pirrung, M.C. (2004) Molecular Diversity and Combinatorial Chemistry: Principles and Applications, Elsevier, Oxford. Shoichet, B.K. (2004) Nature, 432, 862. Stockwell, B.R. (2004) Nature, 432, 846. Terrett, N.K. (1998) Combinatorial Chemistry, Oxford University Press, New York. Xiang, X.-D., Sun, X., Briceno, G., Lou, Y., Wang, K.-A., Chang, H., WallaceFreedman, W.G., Chen, S.-W., and Schultz, P.G. (1995) Science, 268, 1738. Franceschetti, A. and Zunger, A. (1999) Nature, 402, 60. Wang, M., Hu, X., Beratan, D.N., and Yang, W. (2006) Journal of the American Chemical Society, 128, 3228.

15 von Lilienfeld, O.A., Lins, R.D., and

16 17

18

19 20 21 22

23

24

25

26

Rothlisberger, U. (2005) Physical Review Letters, 95, 153002. Kuhn, C. and Beratan, D.N. (1996) Journal of Physical Chemistry, 100, 10595. Xiao, D., Yang, W., and Beratan, D.N. (2008) Journal of Chemical Physics, 129, 044106. Xiao, D., Martini, L.A., Snoeberger, R.C., Crabtree, R.H., and Batista, V.S. (2011) Journal of the American Chemical Society, 133, 9014. Hückel, E. (1931) Zeitschrift für Physik, 70, 204. Van-Catledge, F.A. (1980) Journal of Organic Chemistry, 45, 4801. Hoffmann, R. (1988) Reviews of Modern Physics, 60, 601. Burdett, J.K. (1995) Chemical Bonding in Solids, Oxford University Press, Oxford. Abuabara, S.G., Cady, C.W., Baxter, J.B., Schmuttenmaer, C.A., Crabtree, R.H., Brudvig, G.W., and Batista, V.S. (2007) The Journal of Physical Chemistry C., 111, 11982. McNamara, W.R., Snoeberger, R.C., Li, G.;., Schleicher, J.M., Cady, C.W., Poyatos, M., Schmuttenmaer, C.A., Crabtree, R.H., Brudvig, G.W., and Batista, V.S. (2008) Journal of the American Chemical Society, 130 14329. Rego, L.G.C. and Batista, V.S. (2003) Journal of the American Chemical Society, 125, 7989. Abuabara, S.G., Rego, L.G.C., and Batista, V.S. (2005) Journal of the American Chemical Society, 127, 18234.

187

188

8 A Tutorial of the Inverse Molecular Design Theory in Tight-Binding Frameworks and Its Applications 27 Ammeter, J.H., Buergi, H.B., Thibeault,

31 Kurtz, H.A. and Dudis, D.S. (eds) (1998)

J.C., and Hoffmann, R. (1978) Journal of the American Chemical Society, 100, 3686. 28 Rao, S.S. (1978) Optimization Theory and Application, 2nd ed., Halsted, New York. 29 Boyd, R.W. (1992) Nonlinear Optics, Academic Press, New York. 30 Marder, S.R., Beratan, D.N., and Cheng, L.-T. (1991) Science, 252, 103.

Reviews in Computational Chemistry, vol. 12, Wiley-VCH Verlag GmbH, New York. 32 Calzaferri, G. and Rytz, R. (1995) Journal of Physical Chemistry, 99, 12141. 33 Wang, M., Hu, X., Beratan, D.N., and Yang, W. (2006) Journal of the American Chemical Society, 128, 3228.

189

9 Green Chemistry Molecular Recognition Processes Applied to Metal Separations in Ore Beneficiation, Element Recycling, Metal Remediation, and Elemental Analysis Reed M. Izatt, Steven R. Izatt, Neil E. Izatt, Ronald L. Bruening, and Krzysztof E. Krakowiak

9.1 Introduction

Introduction of green chemistry principles into the chemical and pharmaceutical industries has attracted much interest [1]. Application of these principles in these industries has resulted in processes that are cleaner, more eco-friendly, simpler, and more cost-effective than those they replaced. Additional benefits of the green chemistry processes include more positive publicity, less generation of waste, fewer chemical accidents, and less product contamination [1]. The majority of these applications involve the introduction of cleaner, safer, and simpler pathways in the synthesis of organic compounds, improvement of processes in the manufacture of chemical and pharmaceutical products, and substitution of alternatives for solvents. Metal separations in mining and related industries are a fruitful field for the application of similar green chemistry principles [2]. Global awareness of the need to protect the environment, domestic animals, and human beings from the effects of metal pollutants has increased during the past few decades. This awareness is evident in press releases, media coverage, and published literature [2,3] and is expected to be of increasing importance in driving future legislative action. It has been recognized that methods for metal separations are needed that come closer to meeting the standards set by green chemistry principles [1,2]. Molecular Recognition Technology (MRT) is one of the most successful of these methods from a commercial standpoint. Chemical separations, as presently used, often result in extensive waste generation, due, in large part, to low metal selectivity by reagents used in the procedures [1,2]. Metal separations are required in a wide range of fields, including mining, ore beneficiation, radionuclide remediation, element recycling, environmental remediation, and sample preparation for metal analysis. Metals are central components in each of these fields, but application of green chemistry principles to separations of these metals is rare. Failure to control the movement of metals in processes results in excess waste generation; lost income; loss of a valuable and, often, irreplaceable resource; potential severe environmental

Handbook of Green Chemistry Volume 10: Tools for Green Chemistry, First Edition. Edited by Evan S. Beach and Soumen Kundu.  2017 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2017 by Wiley-VCH Verlag GmbH & Co. KGaA.

190

9 Green Chemistry Molecular Recognition Processes Applied to Metal Separations in Ore Beneficiation

damage; potential human health problems; and possible geopolitical concerns as nations compete for available resources [4,5]. Entrenched separation methods, such as solvent extraction (SX), ion exchange (IX), and precipitation, used extensively for metal separations in the industries mentioned [6,7], generally, run counter to green chemistry principles [1,2,8]. Successful functioning of traditional separation procedures is especially difficult in highly acidic or basic solutions; in the presence of complex matrices containing large amounts of other solutes; and in systems where the target metal is present in mg l 1 or lower concentrations, especially in the presence of complex matrices. A discussion of advantages and disadvantages of conventional separation systems, including MRT, is available [2,9–11]. MRT is a green chemistry method for metal separations at the molecular level utilizing nanochemistry principles [2,11–13]. MRT has been shown in commercial metal separation processes to be simpler, more cost-effective, more eco-friendly, and more efficient than conventional procedures [2,8–10,13]. It is desirable in the long term to replace processes that generate large amounts of waste with cleaner technologies, such as MRT. However, there are significant barriers to the replacement of existing technologies, as indicated by Matus et al. [14]. The purpose of the present paper is to provide a review of the extensive use of MRT green chemistry processes in metal separations. These separations cover a wide variety of areas, including processing and beneficiation of mined ore, metal recovery from acid mine drainage streams, metal recycling, toxic metal recovery from environmental streams, recovery of radionuclides from wastes, and preparation of solutions for metal analyses. Applications are presented of commercial operations using MRT in each area. A major aim is to demonstrate that green chemistry processes, such as MRT, are capable of clean chemistry separations in a competitive market and to advocate development and greater use of such processes by members of the separations community.

9.2 Molecular Recognition Technology as a Green Chemistry Process

The green chemistry approach of MRT to chemical separations and recovery has been described [2,9,11]. Green chemistry principles (Table 9.1) emphasize the importance of preventing generation of waste instead of cleaning up waste after it is formed [15]. Waste generation with little or no attempt to clean it up is an excellent example of negative externality. In such circumstances, the burden of cleaning up waste and any resulting environmental and/or human or animal health effects are borne by society, not by the generator of the waste. Large amounts of critical and/or valuable metal resources are, subsequently, lost. Furthermore, in the case of metals, the problem remains with society for long term, since metals are indestructible [15], can change their form and can move through ecological systems in uncontrolled and, often, unpredictable ways when

9.2 Molecular Recognition Technology as a Green Chemistry Process

Table 9.1 Green chemistry principles applied to metal separations. Design for Energy Efficiency: Energy requirements of chemical processes should be recognized for their environmental and economic impacts and should be minimized Use of Renewable Feedstocks: A raw material or feedstock should be renewable rather than depleting whenever technically and economically practicable Prevention: It is better to prevent waste than to treat or clean up waste after it is formed Safer Solvents & Auxiliaries: Use of auxiliary substances (e.g., solvents, separation agents, etc.) should be made unnecessary, whenever possible and, when used, innocuous

exposed to environmental conditions [4,15,16], and can have dire human and animal health effects, if ingested [3,16]. Negative environmental, energy, and health impacts from metal pollution can be severe [3,15,16], unless care is taken to minimize them. Prevention of waste generation is expensive, since it requires recovery of the metal or disposal of the waste in an environmentally safe manner. Adequate disposal of such waste is, likewise, expensive and, usually, is not factored into the cost of the ultimate product. Generally, there is much talk about waste control, but little action for many reasons, largely economic [15]. The majority of waste worldwide is incinerated, sent to landfill, or stored in tailings [17]. The magnitude of the problem of coping with such waste is enormous and is growing [18]. Matus et al. [14] have identified and discussed barriers to the implementation of green chemistry in the chemical and pharmaceutical industries. These barriers comprise a complex set of intertwined issues that act to impede the effective implementation of green chemistry within the chemical enterprise. Barriers broadly fall into categories of economic and financial, inertia, regulatory, technical, organizational, cultural, and even definition and measurement. These categories themselves have significant interactions and overlaps. It is to be expected that attempts to introduce green chemistry operations into separation processes in the metals and related industries would experience similar barriers. In the metal separations area, SX, IX, and precipitation are used extensively. These low-selectivity separation methods use solvents and, often, corrosive chemicals resulting in the generation of large amounts of waste. They are the main competitors of highly selective separation processes. Technological and economic advantages of green chemistry MRT over these traditional technologies have been presented and discussed [2,8,9,13]. A major impediment to the introduction and use of new technologies is the investment already made in existing technologies. Major expenses are usually required to install new technologies. Legislative action is often the spur that leads to cleaner processes, as has been pointed out by Taylor et al. [19]. One reason for the movement of metals industries away from Organization for Economic Co-operation and Development (OECD) nations to non-OECD nations in the past few decades is the passage of legislation requiring cleaner operations in OECD countries. Legislative restraints are usually much less severe in non-OECD nations. As a result, there is extensive metal pollution in these countries resulting in major costs to society [15,20,21].

191

192

9 Green Chemistry Molecular Recognition Processes Applied to Metal Separations in Ore Beneficiation

Figure 9.1 Representation of a SuperLig® resin consisting of a ligand (18-crown-6) bound by a tether to a solid support (silica gel).

Highly selective metal separations are achieved using MRT systems. Predesigned metal-selective ligands are attached by chemical binding through a tether to solid supports, such as silica gel or polymer substrates, as illustrated in Figure 9.1 [2,11]. Resulting products, termed SuperLig®, operate at nanometer scale [12]. If used in analytical applications, MRT products are termed AnaLig®. Organic solvents are not used in these MRT systems. The controlled, highly selective separations that occur at the molecular level are an example of nanochemistry being used in a novel way to demonstrate supramolecular host–guest assembly [12,22]. The availability of SuperLig® systems capable of highly selective separations at the molecular level covering a wide range of elements in the periodic table provides a valuable tool for use in green chemistry metal recovery. MRT has been used in commercial systems for selective metal separations and recovery for more than two decades. Such systems, normally, operate in a column mode although other modes may be used, as presented in Section 9.4. The column is packed with the appropriate SuperLig® resin. A feed solution containing the target metal and any gangue metals is flowed through the column, where the target metal is extracted by the SuperLig® resin, and the remaining solution goes to raffinate. The target metal is eluted with a small volume of eluent to form an eluate solution containing the pure target metal, which is highly concentrated over its concentration in the feed solution. Green chemistry characteristics of MRT systems that differentiate them from conventional separation technologies have been presented and discussed [2,13]. These characteristics arise, primarily, from large SuperLig®-target metal binding energies (log K), rapid reaction kinetics, high degree of selectivity by the bound ligand for target metal (s) even in complex matrices such as highly acidic or basic solution and/or in solutions containing high concentrations of competing ions, regenerability of the resin for multiple uses, use of relatively benign solutions throughout, small floor space requirements, marked reduction of metal inventory time, and the ability to carry out selective metal separations and recover pure metals over wide metal concentration ranges from g l 1 to mg l 1 or lower. The metal separated can be recovered either for value or in a form allowing it to be disposed of in an environmentally safe manner. Movement of the metal is controlled throughout the process. The indicated steps are summarized in Table 9.2.

9.2 Molecular Recognition Technology as a Green Chemistry Process

Table 9.2 Steps in operation of MRT processes in column mode. Loading phase – target metal ion loaded from feed onto SuperLig® product charged into column(s) Pre-elution wash phase – remaining feed solution washed from column Elution phase – target metal ion eluted with small amount of eluent to form concentrated solution containing the metal product Postelution wash phase – remaining eluent washed from column; cycle begins again with first step

The steps in Table 9.2 are illustrated by the example of commercial Pd separation in Figure 9.2 [23]. A feed solution containing Pd, other platinum group metals (PGM), and base metals has been added to a column packed with Pd-selective SuperLig® 2. The Pd is retained on the column by the SuperLig® 2 and can be seen as the red material near the top of the column. The remaining feed solution goes on to raffinate. Following washing of the column to remove residual feed solution, the bound Pd is eluted in concentrated form by a small amount of ammonium bisulfite eluent [23]. Selective separation of Pd in a single stage and its subsequent concentration by elution in the eluate allows retrieval of this metal in pure form. Removal of Pd from the feed solution ensures that the raffinate is free of Pd eliminating any need to further purify the raffinate downstream. Concentration of separated Pd by elution results in an eluate solution that contains only Pd.

Figure 9.2 Columns used for palladium recovery at Impala Platinum, Ltd. installation, South Africa. Reproduced with permission from Ref. [23].

193

194

9 Green Chemistry Molecular Recognition Processes Applied to Metal Separations in Ore Beneficiation

Metal separation technologies based on green chemistry principles need to be applicable to a wide range of systems, since the number of metals used in new products has increased markedly in recent decades and now includes most of the metals in the periodic table compared to, perhaps, a dozen a half century ago [24]. Design of metal selectivity into MRT systems requires knowledge of metal, ligand, and system parameters involved in separation processes. Insight into this design procedure is seen in the examples that follow. In each case, the separation follows the procedure illustrated in Table 9.2 and Figure 9.2. Several examples show the variety of forms in which the metal can be attached to the SuperLig® resins. These forms are important in the design process and include cations, constituents of anions, and ion pairs. Potential applications are openended and, in principle, exist wherever metals need to be separated and recovered. MRT applications are found in a wide variety of fields where control of metal pathways is essential. The challenge is to develop SuperLig® resins capable of high selectivity for the metal to be targeted combined with an elution protocol that concentrates the extracted metal in an eluate that is compatible with downstream processes. Application of host–guest molecular recognition principles to the design process has resulted in a large number of highly selective SuperLig® and AnaLig® resins that operate in a wide variety of fields as will be seen in the examples that follow. These examples will show green chemistry separations, using these resins, of a large fraction of the metals in the periodic table.

9.3 Metal Separations Using Molecular Recognition Technology 9.3.1 Separation and Recovery of Individual Rare Earth Elements

China dominates the global rare earth element (REE, singular or plural) market [25,26]. Essentially all REE used worldwide are mined in China and nearly 100% of those mined are separated there into individual metals [6]. China currently has an annual REE output of around 150 000 tonnes [21]. About 40% of this output derives from illegal mining operations, particularly of ionic clay REE deposits located in South China. There are very negative externality effects of illegal mining. In the Quanzhou region, it has been estimated that negative externality reclamation costs (38 billion RMB) of illegal mining outweigh local revenues (32.9 billion RMB) from sale of the mined REE by about 5 billion RMB. Not included in these costs are the large negative externality health and ecological costs to the local inhabitants [21,27]. Separations of individual REE from the mined ore use SX and IX processes, which are the source of extensive land, water, and air pollution from the wastes generated [15,25,26]. The substantial quantity of waste generated is, in large part, a result of use of low-selectivity processes for the separations. REE are among the most difficult elements in the periodic table to separate from each other

9.3 Metal Separations Using Molecular Recognition Technology

because of the close similarity of their chemical and physical properties. Separations were carried out a century ago by the laborious method of fractional crystallization that required thousands of stages [28]. Development of atomic weaponry in World War II resulted in characterization of REE, which were products of nuclear reactions. Ion exchange proved to be an excellent method, at that time, to obtain very pure REE, but only in small amounts [7]. The need for commercial quantities of these metals arose in the 1960s with the advent of color TV, which required Eu for use as a red phosphor. Solvent extraction proved to be an effective method for the production of the large quantity of REE needed for phosphors, high strength magnets, and other products during the latter part of the twentieth century [7]. Solvent extraction is effective in making required separations, but hundreds of stages are needed to obtain individual REE of high purity [6] because of the low selectivity of the process. An MRT process, based on green chemistry principles, has been developed and commercialized for the separation and recovery of REE from virgin ore [13]. In this proprietary process, the entire suite of 16 REE, Pm excluded, has been separated at the bench level at high purity of >99% [29,30]. A pregnant leach solution (PLS) derived from Bokan-Dotson Ridge, Alaska, feedstock was the feed material for the separations. The PLS was prepared by Hazen Research Inc., of Golden, Colorado from beneficiated ore using a metallurgical process that has been described [29]. The Bokan deposit contains high levels of certain of the “heavy” REE, such as Dy, Tb, and Y, which are critical to the US economy [31]. The entire set of 16 REE was initially separated from the non-REE gangue metals present in the PLS [13]. This separation of REE, accomplished at the >99% level, is an important first step in the process. Having the entire suite of REE separated from accompanying gangue material simplifies subsequent separation of individual REE, since impurity metals are not present downstream. Furthermore, the >99% pure concentrated REE solution contains essentially all of the REE present in the original PLS. This exceptionally high REE recovery rate is a key economic differentiator between MRT and other, less selective, processes, which leave appreciable quantities of REE in tailings as waste. The MRT process maximizes the value obtained from the PLS and ensures that minimal amounts of REE are discarded to tailings. Separation of the 16 REE into individual elements at >99% purities was achieved in a three-step process [13]. In Step 1, Sc and Ce were removed. In Step 2, light REE (Y, La, Pr, Nd) were separated from heavy REE (Sm through Lu). In Step 3, individual members in the light and heavy groups were separated from each other at the >99% level. Individual REE were collected as carbonates and analyzed for purity by an ICP procedure at IBC. Analyses by an independent laboratory confirmed the purity levels at >99%. Comparison of opex and capex costs for MRT versus SX in REE separations and recovery indicated the advantages of MRT leading to the lower costs [13]. A pilot plant capable of making group and individual separations of REE has been completed [32a]. Feedstock for the pilot plant is derived from the Bokan– Dotson Ridge REE deposit. Mining of this ore and preparation of the PLS from it

195

196

9 Green Chemistry Molecular Recognition Processes Applied to Metal Separations in Ore Beneficiation

are also carried out using green chemistry principles [33]. Separation of the group of REE from gangue metals at the >99% level and REE recovery at the 99% level was achieved followed by separations of REE into light and heavy groups and separation of Dy from the heavy group at the 99.99% level [32b]. The large REE recovery is significant considering that the Mt. Pass, California REE operation, when it was functioning, had overall recovery of REE values of 65–70% and tailings that assayed ∼1–2% REE [32c]. Achievement of individual REE separations illustrates how barriers to implementation of green chemistry principles can be met and overcome. Matus et al. [14] indicate that a nearly universal barrier is the perception by those who are established and have built careers in a particular field that implementation represents a threat to their established position. As a result, introduction of green chemistry processes often works best under enlightened, supportive, and progressive management and a technical staff who possess a different set of ideas and abilities than those found in previous technologies. These factors came together in the present case and the result is the development of a green chemistry separation scheme for REE separation and recovery at the molecular level based on principles of molecular recognition [34]. 9.3.2 Platinum Group Metals 9.3.2.1

General

Sequential separations of individual PGM from a mixture of these metals in the presence of gangue metals have been described [2,8,13]. A key aspect of the separations is the use of SuperLig® products designed to bind selectively with specific PGM in the presence of competing metal species, including other PGM. Achievement of individual metal selectivity requires that SuperLig® resin design take into account the speciation of the PGM as a function of the anion present and the pH of the solution as well as the relative binding strengths of the several possible host–guest combinations. For example, SuperLig® 2 is selective for the cation, Pd2+, allowing selective separation of this cation from remaining PGM, which do not form cations in the solutions involved. Rhodium is separated as an 1) anion, RhCl52 and/or RhCl63 , using SuperLig® 190, which is designed to selectively separate this particular species. Selectivity of SuperLig® 133 for PtCl62 over other PGM chloro anions and over other anions and cations is very high. Ruthenium binds to SuperLig® 187 as a chloro anion, with high selectivity over Pt and Pd and base metal species. Iridium is selectively separated from Rh at high purity using SuperLig® 182. Selectivity is of critical importance in these separations. Single pass separations for the target PGM of 99%+, and individual PGM purities of 99.95–99.99% are achieved. High selectivity also eliminates the need for multiple separation stages 1) Charges on metals are, generally, omitted in the text, except where needed for clarity. Charges on ionic complex species formed by metals are given.

9.3 Metal Separations Using Molecular Recognition Technology

downstream that would increase use of hazardous and/or contaminating chemicals and increase operating costs. Selective separations of PGM are, in all cases, single-pass resulting in efficient operations with minimal waste. The time required to make separations is short resulting in reduced inventory time for PGM over that required in conventional separations. Reagents used for wash and elution are chosen to be as mild as possible, and include water, mineral acids, and simple salts. Specific wash and elution reagents used are indicated in the descriptions that follow of PGM separations and elutions. These PGM separations provide examples of green chemistry processes occurring at the molecular level resulting in economic benefits to an industrial company by eliminating steps; shortening the time required to make separations; eliminating use of many, often hazardous, chemicals; reducing labor costs; reducing environmental impacts; and shortening inventory time for the valuable PGM involved. Target metals are recovered in pure concentrated form from eluate solutions following elution from the column with a small amount of eluent, as is described in Table 9.2. 9.3.2.2

Palladium Recovery from Native Ore

An MRT system is used at Impala Platinum, Ltd. in South Africa for commercial recovery of Pd from native ore [8,23]. The operation of the system is illustrated in Figure 9.2. MRT was chosen by Impala after extensive internal laboratory and pilot plant tests comparing the MRT product, SuperLig® 2, with IX and SX products. High selectivity and high affinity for Pd in the presence of matrix species, including other PGM, led to the selection of MRT. There are a number of benefits of this separation. First, quantitative and selective removal of Pd dramatically increases Pd yield with reduced amounts of Pd going to recycle. An important advantage of high selectivity is that Pd is the only metal removed from the feed by SuperLig® 2, thus ensuring its subsequent elution from the resin in pure form. Second, with no Pd going downstream, recovery of other Pt metals from the raffinate is greatly simplified with significant time and cost savings. Third, several process steps, in the old Pd recovery process, needed to clean up Rh/Ir streams containing impurity Pd downstream are no longer required. Thus, high selectivity results in more rapid and economical Pd recovery by elimination of several process steps together with the additional reagents and labor needed for those steps. Control of Pd throughout the process minimizes waste generation, consistent with the principles of green chemistry (Table 9.1). 9.3.2.3

Rhodium Recovery from Spent Catalyst and Other Wastes

Tanaka Kikinzoku Kogyo K.K. (TKK) in Japan uses MRT to recover Rh from spent automotive catalysts and other feeds [8,35,36]. The MRT process is highly selective for Rh in this process. Rhodium is recognized as a chloro species, such as RhCl52 and/or RhCl63 , by SuperLig® 190. If Pt is present in excess of Rh, some Pt impurity, present as an anion, is found in the concentrated Rh solution. In such cases, Pt is removed using another SuperLig® product, resulting in commercially pure Rh. Benefits of the MRT process include high first pass recovery

197

198

9 Green Chemistry Molecular Recognition Processes Applied to Metal Separations in Ore Beneficiation

rates, significant reduction in process time required to refine Rh, and reduction in required floor space. This example demonstrates the economic benefits of SuperLig® resins capable of selective interaction with chloro anions of the platinum metals. 9.3.2.4

Platinum Recovery from Alloy Scrap

A commercial MRT system has been used for extraction, recovery, and purification of Pt from alloy scrap containing Co, Cr, Pt, and, in some cases, Cu, resulting from the sputtering process [37]. The feed solution consists of dissolved Co, Cr, and Pt derived from leaching alloy scrap with either HCl/H2O2 or HCl/ bleach. Platinum dissolves in Cl matrices as PtCl62 . The selectivity of SuperLig® 133 for PtCl62 over some of the other PGM chloro anions and other anions and cations is very high. The eluant is H2O at ambient temperature. Washes are 5 M NaCl/0.1 M HCl and/or 1 M NaCl/0.1 M HCl. Dissolution of Pt in the scrap is 100% (within the limits of analytical error). The single pass recovery of Pt with SuperLig® 133 is 99.9% (not including the Pt on the trail columns). Overall recovery is in the range of 99.99%. Ammonium chloride is charged into the concentrated eluate solution precipitating (NH4)2PtCl6 (Pt yellow salt). Postprecipitation barrens are extremely low in Pt, indicating minimal loss of this valuable metal. The purity of the Pt yellow salt produced is normally four nines versus metal. The Pt yellow salt is converted to sponge using conventional reduction technology. Melting/forming processes are then used to get a thin film sputtering target. A flow sheet of the process is given in Figure 9.3.

Figure 9.3 General flow sheet of the platinum refining process incorporating the MRT system. Reproduced with permission from Ref. [37].

9.3 Metal Separations Using Molecular Recognition Technology

Figure 9.4 Photograph of a commercial system for platinum refining at an Asian refinery. Reproduced with permission from Ref. [37].

A photo of the commercial size MRT refining system for Pt extraction and refining is shown in Figure 9.4 [37]. The system consists of three columns in series in polishing mode. The MRT system operates as described in Table 9.1 with the following conditions: (1) Pt is loaded onto the column as the anion, PtCl62 , (2) elution of PtCl62 from the column is accomplished with H2O at ambient temperature, and (3) column wash solutions consist of mild reagents, including 5 M NaCl/0.1 M HCl or 1 M NaCl/0.1 M HCl. 9.3.2.5

Ruthenium Recovery from Alloy Scrap

MRT has been used for extraction, recovery, and purification of Ru from Ru alloy scrap [37]. The MRT system extracted Ru from a matrix of Al, Fe, Na, and other base metals present in 6 M HCl. Ruthenium binds to SuperLig® 187 as either RuCl5(H2O)2 or RuCl63 . A low oxidation–reduction potential (ORP) of 300–400 mV versus Ag/AgCl electrode and 6 M excess Cl are required to ensure that all Ru is present as Ru(III) and the presence of any Ru(II)-chloro species is minimized. Elution is performed using 5 M NH4Cl at ambient temperature. Washes are made using HCl. The oxidized and precipitated Ru product from the eluate concentrate is (NH4)2RuCl6. The purity of the Ru salt is usually 99.98–99.99% as metal equivalent. Pyrometallurgical techniques are used to reduce the Ru salt to Ru sponge material. A process block diagram is shown in Figure 9.5. The MRT system operates as described in Table 9.2 with the following conditions: (1) Ru is loaded onto the column in anionic form, as RuCl5(H2O)2 or RuCl63 , (2) elution of these anions from the column is

199

200

9 Green Chemistry Molecular Recognition Processes Applied to Metal Separations in Ore Beneficiation

Figure 9.5 Process block diagram for refining ruthenium using MRT. Reproduced with permission from Ref. [37].

accomplished with NH4Cl at ambient temperature to form a high purity (NH4)2RuCl6 solution, and (3) column wash solutions consist of mild reagents, including dilute HCl solutions. This application of MRT is of interest because of its contrast with the conventional Ru distillation process, which it displaces. In the latter process, Ru is carefully distilled as RuO4, a powerful oxidizing agent, which can decompose explosively to RuO2 above 100 °C and may do the same at ambient temperatures if brought into contact with oxidizable organic solvents or other organic materials [38]. These properties, together with the poisonous character of RuO4, necessitate careful containment of the Ru system during processing, resulting in high capex and opex costs. Both conventional and MRT processes begin with the same fusion/dissolution step, and both use conventional chloride reduction technology to produce metal sponge. The difference between the two processes lies in the intermediate step, where distillation is used in the conventional process. A comparison of distillation and MRT parameters in Ru production is given in Table 9.3. The capex and opex costs of the distillation procedure exceed those of the green chemistry MRT procedure by a large amount. In addition, the potential for additional costs if an accident occurs in the distillation process is large. This example illustrates the significant reduction of capex and opex costs that result by replacement of a system that violates green chemistry principles, with a less dangerous and more eco-friendly one. 9.3.2.6

Iridium Separation from Rhodium and Base Metals

A flow sheet is shown in Figure 9.6 for the selective separation of Ir from Rh and base metals [10,39].

9.3 Metal Separations Using Molecular Recognition Technology

Table 9.3 Comparison of distillation and MRT parameters for ruthenium production [37]. Parameter

Distillation

MRT

Major capital equipment

Absorbers, scrubbers, glass-lined vessels, explosive resistant equipment, tanks, instrumentation, program logic controller, still, protective casings, extensive ducting

Conventional columns, tanks, valves, instrumentation, PLC

Health and safety issues

RuO4 explosive and poisonous, NaBrO3 and CCl4 carcinogenic, generation of Cl2 and Br2

None

Control issues

Oxidation potential important requiring removal of all base metals prior to beginning of operation

None

Level of Ru purity obtained

High

High

Level of Ru recovered

>95%

>99%

Number of process steps

One or two, labor intensive

One

Pipeline

Weeks

Days

Relative capital cost index (1 lowest)

5–10

1

Relative operating cost index (1 lowest)

3–5

1

SuperLig® 182 has high selectivity for Ir over Rh and the base metals present. Thus, Ir is separated from the feed stream at high purity in one pass through the column. Use of an appropriate SuperLig® product allows further separation of Rh from the base metals in the raffinate. Possible eluants are ambient water, hot (70 °C) water, and 0.25 M Na2SO3/2.5 M NaCl solution. Which eluant is chosen depends on the process conditions, such as the Ir speciation present. Similar

Figure 9.6 Flow sheet for selective separation of iridium from rhodium and base metals. ORP, oxidation–reduction potential. Reproduced with permission from Ref. [39].

201

202

9 Green Chemistry Molecular Recognition Processes Applied to Metal Separations in Ore Beneficiation

options are available for wash solutions. This example illustrates the simplicity of the MRT procedure and the use of mild reagents and conditions to accomplish the desired separation. 9.3.2.7

Purification of 103Palladium for Use in Brachytherapy

Purification of isotopes for medical and other uses can be accomplished using MRT. An IBC product, AnaLig® Pd-03, is used by manufacturers of brachytherapy seeds in the United States and Europe for purification of 103 Pd for use in the treatment of prostate cancer [40]. This important and widely used therapy requires insertion of radioactive 103 Pd directly into the prostate gland. The therapy is especially useful for patients with aggressive prostate cancer. The MRT product is used to separate 103 Pd from a matrix in which it is present and purify it prior to preparing the pellets involved in the treatment. High purity 103 Pd is essential in this application. The 103 Pd composition offers distinct advantages over a competitor, iodine-based seeds, by eliminating steps and reagents. In addition to brachytherapy, 103 Pd is used for treatment of other conditions, such as vascular disease and macular degeneration. This example suggests the possibility of using MRT for other medical metal isotope purifications. 9.3.3 Gold Separation and Recovery from Process Streams 9.3.3.1

General

Gold usually occurs at low concentrations in ore bodies, typically