QUANTIFYING the MINERAL CARBONATION POTENTIAL of MINE WASTE MATERIAL: a NEW PARAMETER for GEOSPATIAL ESTIMATION By

QUANTIFYING THE MINERAL CARBONATION POTENTIAL OF MINE WASTE MATERIAL: A NEW PARAMETER FOR GEOSPATIAL ESTIMATION by

ANTHONY DAVID JACOBS

M.Sci (Honours), The University of Birmingham, 2007 M.Sc., The University of Exeter (Camborne School of Mines), 2008

A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in The Faculty of Graduate and Postdoctoral Studies (Mining Engineering)

THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver) October 2014

Abstract

Direct aqueous mineral carbonation is a CO2 sequestration method that can trap anthropogenic carbon dioxide as a thermodynamically stable carbonate. Ultramafic rocks considered waste at mining operations are targeted as the substrate source for carbonation in this research. Using mining waste to permanently fix CO2 is a motivating factor for promoting industrial mineral carbonation as a viable carbon sequestration option. Experimental direct aqueous mineral carbonation was carried out on proposed Turnagain waste rock, a low-grade, high tonnage nickel sulphide deposit in northern British Columbia. 45 % magnesium silicate to magnesium carbonate conversion was achieved in two hours. The successful sequestration of

CO2 using mining waste rock and the opportunity of shared mining and mineral processing costs of a dual mining/mineral carbonation operation can aid in reducing economic and energy requirements, identified as key inhibiting factors of industrial mineral carbonation. The heterogenous mineralogy of ultramafic deposits commonly hosting mining operations makes quantifying the mineral carbonation potential (MCP) of the waste rock challenging. The MCP calculator, a novel Microsoft ExcelTM spreadsheet program was developed to estimate the modal mineral abundance of ultramafic rocks for use in MCP estimation. The calculator is intended for use by the mining industry utilising abundant lithogeochemical data as a cost-effective tool in evaluating their deposit as a supplier of substrate material for industrial mineral carbonation operation. The calculator can be tailored to estimate MCP values based on a site specific mineral assemblage. MCP values generated represent composited sample length intervals along exploration drill holes. Estimation techniques traditionally used by the mining industry for resource estimation were evaluated as methods of geospatially interpolating MCP values. Both inverse distance and ordinary kriging were successful in interpolating MCP values. A capacity of 75 million tonnes of CO2 within the proposed 28 year surface mine design at Turnagain was calculated. This capacity is significantly lower than the theoretical maximum capacity of 538 million tonnes calculated assuming all MgO within the waste rock is capable of sequestering CO2 as magnesite (MgCO3). The research highlights the importance of understanding and quantifying the mineralogy of ultramafic deposits when estimating their potential for mineral carbonation.

Preface

The dissertation is original, unpublished, independent work by the author, Anthony Jacobs.

Some external laboratory testing and analysis was undertaken and the results are reproduced herein by the author;

X-ray diffraction with Rietveld refinement analysis was undertaken by the Electron Microbeam/X-ray Diffraction Facility, Earth and Ocean Science Department, The University of British Columbia.

ICP-ES lithogeochemial and X-ray fluoresence major oxide analyses was undertaken by the Acme Analytical Laboratories, Vancouver, British Columbia.

iii

Table of contents

Abstract ...... ii

Preface ...... iii

Table of contents ...... iv

List of tables ...... ix

List of figures ...... xi

List of symbols and abbreviations ...... xiv

Acknowledgements ...... xvi

Dedication ...... xvii

1 Introduction ...... 1

1.1 Climate change and the carbon cycle ...... 1

1.2 Carbon capture and storage technologies ...... 3

1.2.1 CO2 capture ...... 4

1.2.2 CO2 transport ...... 4 1.2.3 Geological storage ...... 5 1.2.4 Mineral storage ...... 6

1.3 Carbon capture and storage in the mining industry ...... 7

1.4 Dissertation structure ...... 10

1.4.1 Dissertation focus ...... 10 1.4.2 Dissertation questions ...... 11 1.4.3 Chapter organisation ...... 12 2 Literature review: ex-situ mineral carbonation and modal mineral estimation ...... 15

2.1 Mineral carbonation background ...... 15

2.2 Mineral carbonation process pathways ...... 16

2.2.1 Direct mineral carbonation ...... 17 2.2.2 Indirect mineral carbonation ...... 19

2.3 Geology and sources of substrate material ...... 21

2.3.1 Mafic and ultramafic rocks ...... 21 2.3.2 Mining waste rock and solid industrial waste streams ...... 25

2.4 Experimental direct aqueous mineral carbonation ...... 26

2.4.1 Mineral pre-treatment ...... 28 2.4.2 Aqueous solution chemistry and experimental carbonation conditions ...... 29

2.5 Estimating the mineral carbonation capacity of ultramafic rocks ...... 32

2.5.1 Mineral carbonation controlling factors on resource estimation...... 34 2.5.2 Estimating mineralogy from lithogeochemical data ...... 41

2.6 Resource estimation in the mining industry ...... 44

2.6.1 Traditional empirical approaches to resource estimation ...... 46 2.6.2 Geostatistical approaches to resource estimation ...... 47 3 Experimental, analytical and modelling procedures ...... 50

3.1 Deposit selection and background ...... 50

3.1.1 Twin Sisters ultramafic complex ...... 50 3.2.2 Turnagain ultramafic complex ...... 52

3.2 Sample collection ...... 56

3.3 Sample preparation ...... 57

3.3.1 Density measurements ...... 57 3.3.2 Rock crushing and grinding ...... 58

3.4 Analytical methods ...... 61

3.4.1 Particle size analysis ...... 61 3.4.2 X-ray diffraction ...... 62 3.4.3 Inductively coupled plasma emission spectroscopy ...... 64 3.4.4 X-ray fluorescence...... 65 3.4.5 Scanning electron microscopy ...... 66

3.5 Experimental mineral carbonation laboratory procedures ...... 67

3.5.1 Sample selection ...... 67 3.5.2 Reactant slurry generation ...... 68

3.5.3 Autoclave testing procedure ...... 69 3.5.4 Post-carbonation product recovery ...... 71

3.6 Estimating mineralogy from geochemical analysis data ...... 71

4 Experimental direct aqueous mineral carbonation ...... 79

4.1 Twin Sisters ultramafic rocks ...... 79

4.1.1 Solid reactant particle size analysis and morphology ...... 81 4.1.2 Solid reactant geochemistry and mineralogy ...... 82 4.1.3 Bench-scale experimental results ...... 84 4.1.4 Product characterisation ...... 89

4.2 Turnagain ultramafic rocks ...... 91

4.2.1 Solid reactant particle size analysis and morphology ...... 91 4.2.2 Solid reactant geochemistry and mineralogy ...... 93 4.2.3 Bench-scale experimental results ...... 97 5 The derivation of a mineral carbonation potential parameter...... 106

5.1 Lithogeochemical data preparation and analysis ...... 107

5.1.1 Data availability and requirements ...... 107 5.1.2 Mineralogical investigation ...... 108 5.1.3 Preliminary data analysis ...... 109

5.2 The mineral carbonation potential (MCP) calculator ...... 111

5.2.1 The MCP calculator ...... 112 5.2.2 Lithogeochemical data conversion to major oxides ...... 113 5.2.3 The mineral composition calculator ...... 114 5.2.4 Operating the MCP calculator ...... 115 5.2.5 Reliability and limitations ...... 121

5.3 MCP data validation at the Turnagain ultramafic complex ...... 122

5.3.1 Major cation to oxide conversion ...... 123 5.3.2 Reactant modal mineral estimation ...... 127 5.3.3 MCP estimation ...... 133 6 Estimating a mineral carbonation resource for a mining project ...... 136

6.1 Exploratory data evaluation for MCP estimation ...... 137

6.1.1 Geological investigation ...... 137 6.1.2 Mineralogical investigation ...... 142 6.1.3 Geochemical investigation ...... 143 6.1.4 MCP data generation ...... 149

6.2 MCP estimation at the Turnagain ultramafic complex, British Columbia, Canada - a case study ...... 150

6.2.1 Preliminary MCP data analysis ...... 152 6.2.2 Spatial (structural) data analysis ...... 161 6.2.3 Three-dimensional data interpolation ...... 164

6.3 Global MCP resource estimation and validation ...... 166

7 Discussion ...... 173

7.1 Experimental mineral carbonation ...... 173

7.1.1 Mineral pre-treatment ...... 175 7.1.2 Experimental conditions ...... 176 7.1.3 Mineral carbonation extent determination ...... 179

7.2 Predicting mineral carbonation capabilities from lithogeochemical data ...... 183

7.3 Estimating the mineral carbonation potential of ultramafic deposits...... 187

8 Conclusions ...... 191

8.1 Research outcomes ...... 191

8.2 Suggestions for future research ...... 195

8.3 Personal reflection ...... 197

Bibliography ...... 199

Appendices ...... 212

Appendix A MCP calculator instructions and modifications ...... 212

Appendix A.1 Sample data preparation worksheet ...... 212 Appendix A.2 Mineral composition calculator worksheet ...... 213 Appendix A.3 MCP calculator worksheet ...... 213

vii

Appendix A.4 Multiple sample worksheet ...... 217

Appendix B Data analysis for resource estimation ...... 218

Appendix B.1 Olivine geological domain ...... 218 Appendix B.2 Wehrlite geological domain ...... 223 Appendix B.3 Olivine clinopyroxenite geological domain ...... 228

viii

List of tables

Table 2.1 Published literature on minerals tailings for carbon sequestration ...... 25 Table 2.2 Significant literature published on direct aqueous mineral carbonation ...... 27 Table 2.3 Energy consumption of mineral pre-treatment for mineral carbonation ...... 29 Table 2.4 The magnesium and calcium oxide content of common rocks considered for mineral carbonation...... 35

Table 2.5 CO2-sequestration properties of selected ultramafic complexes in North America .. 37 Table 3.1 Bulk specific gravity comparative statistics ...... 58

Table 3.2 Analytes tested and detection limits using XRF with Li2B4O7/LiBO2 fusion...... 66 Table 3.3 MINSQ and XRD comparison of reactant samples tested...... 73 Table 4.1 Experimental direct aqueous mineral carbonation conditions for Twin Sister rock samples tested...... 80 Table 4.2 Twin Sisters reactant XRF and LOI analysis results ...... 83 Table 4.3 Mineral content in XRD analyses derived from the buffered solution post- carbonation...... 85 Table 4.4 Post mineral carbonation mineral abundances using XRD with Rietveld refinement of the Twin Sisters olivine...... 87 Table 4.5 Normalised stoichiometric forsterite to magnesite conversion extents achieved over varying test durations between 1 and 6 hours...... 87 Table 4.6 Mineral abundance measured using XRD with Rietveld refinement on Turnagain reactant samples pre-carbonation...... 96 Table 4.7 The quantitative mineral abundances of product samples from the Turnagain ultramafic complex...... 99 Table 4.8 Normalised stoichiometric forsterite to magnesite conversion extents achieved using Turnagain reactant samples...... 105 Table 5.1 Lithologies selected for use in the MCP calculator at the Turnagain ultramafic deposit...... 110 Table 5.2 Estimated major oxide values from the MCP calculator compared with those measured by XRF analysis from 16 samples tested at the Turnagain ultramafic complex...... 124

Table 5.3 Measured mineral abundance by XRD with Rietveld refinement and estimated values by the MCP calculator of a Twin Sisters olivine ...... 127 Table 5.4 Major mineral abundance estimated by the MCP calculator and measured using XRD with Rietveld refinement...... 128 Table 5.5 Estimated MCP values by the MCP calculator compared with those projected from XRD with Rietveld refinement analyses...... 134 Table 6.1 Statistical properties of 11,552 geochemical samples from the Turnagain ultramafic deposit...... 144 Table 6.2 Statistical properties of 11,552 geochemical samples from the Turnagain ultramafic deposit...... 146 Table 6.3 Block model dimensions used for MCP resource estimation at the Turnagain ultramafic complex...... 151 Table 6.4 Descriptive statistics of MCP and MCP_ARC values derived using the MCP calculator program...... 153 Table 6.5 Modelled volume of each defined geological domain at Turnagain...... 154 Table 6.6 Descriptive statistics of CaO (wt.%), MgO (wt.%), MCP, and MCP_ARC values on composited drill core samples from the Turnagain ultramafic complex...... 157 Table 6.7 Descriptive statistics of CaO (wt.%), MgO (wt.%), MCP, and MCP_ARC values on composited drill core samples from the Turnagain ultramafic complex...... 160 Table 6.8 Directional variogram model parameters for each geological domain and sub-domain at the Turnagain ultramafic complex...... 163 Table 6.9 Anisotropic search ellipse dimensions for resource estimation at the Turnagain ultramafic complex...... 165 Table 6.10 CaO, MgO, MCP, MCP_ARC resource estimates at the Turnagain ultramafic complex...... 167 Table 6.11 Ordinary kriging resource estimate values for MCP at the Turnagain ultramafic complex...... 169

Table 6.12 The theoretical CO2 capacity of the ultramafic rocks at the Turnagain complex. .... 170

Table 6.13 The theoretical CO2 capacity of the ultramafic rocks within a 28 year life of mine surface mine design at the Turnagain complex...... 172 Table 7.1 Guideline cost estimate for experimental direct aqueous mineral carbonation...... 183

List of figures

Figure 1.1 Global anthropogenic GHG emissions ...... 2 Figure 1.2 World total energy consumption and consumption by fuel type, 1990-2040 ...... 3 Figure 1.3 Schematic illustrating how a potential integrated mining and mineral carbonation system could function...... 9 Figure 2.1 IUGS Classification diagram for intrusive ultramafic rocks ...... 22 Figure 2.2 A generalised distribution of dunite and serpentinite-bearing rocks in British Columbia, Canada...... 24 Figure 3.1 Simplified geological map of the Twin Sisters Mountain area, Washington State, USA ...... 51 Figure 3.2 Surface geological map of the Turnagain ultramafic complex, northern British Columbia, Canada ...... 53 Figure 3.3 Topographical map of the proposed Horsetrail pit and life of mine pit shell, British Columbia, Canada ...... 55 Figure 3.4 Drill holes selected from Turnagain for experimental mineral carbonation ...... 57 Figure 3.5 Drill core photographs of the samples tested from the Turnagain ultramafic complex ...... 60 Figure 3.6 Photograph and schematic of the autoclave setup ...... 70 Figure 3.7 Modal mineral estimation comparisons of 16 samples tested from Turnagain using XRD and MINSQ ...... 75 Figure 4.1 Twin Sisters reactant particle size analysis ...... 81 Figure 4.2 SEM image of Twin Sisters reactant olivine post-comminution ...... 82 Figure 4.3 A Rietveld refinement plot of the Twin Sisters olivine reactant material...... 83 Figure 4.4 Normalised mineral carbonation reaction extent of Twin Sisters olivine by varying reaction time...... 86 Figure 4.5 Reaction extents achieved using Twin Sisters olivine reactant in one hour at varying

PCO2...... 89 Figure 4.6 SEM images of the Twin Sisters reactant and post-mineral carbonation solid product ...... 90 Figure 4.7 Classified reactant particle size distribution from the Turnagain reactant samples tested ...... 92

Figure 4.8 SEM image of a Turnagain dunite reactant sample used in experimental mineral carbonation...... 93 Figure 4.9 Scatter plots of the mineral weight percentage of forsterite, lizardite (+ amorphous material), diopside, and spinel group minerals (magnetite and maghemite)...... 100 Figure 4.10 Relative weight percentage of the major mineral phases in reactant and pot- carbonation product samples from Turnagain...... 101 Figure 4.11 Scatter plots of the mineral weight percentage of forsterite versus magnesite and lizardite versus magnesite in Turnagain post-carbonation product samples...... 102 Figure 5.1 Generalised sequence of stages used for modal mineral estimation of ultramafic rocks in the MCP calculator program...... 117

Figure 5.2 Measured MgO, Fe2O3, and CaO oxide values from XRF analysis versus estimated values by the MCP calculator at Turnagain...... 126 Figure 5.3 Modal mineral estimation comparison of 16 samples tested from the Turnagain. . 129 Figure 5.4 Mean absolute error of serpentine versus olivine using the MCP calculator program ...... 132

Figure 5.5 Combined MgO, SiO2, and H2O modal abundances (wt.%) measured by XRF versus those estimated by the MCP calculator program ...... 132 Figure 6.1 Three-dimensional models of the geological domains used in MCP estimation at the Turnagain ultramafic complex...... 141 Figure 6.2 Histograms of MgO wt.% and CaO wt.% values from the 11,552 samples analysed at the Turnagain ultramafic complex...... 145 Figure 6.3 Histograms of MgO wt.% values from the 11,552 samples analysed at the Turnagain ultramafic complex...... 148 Figure 6.4 Plan section of the Turnagain block model at the 900 m elevation...... 155 Figure 6.5 Histogram of the MgO wt.% values within the wehrlite geological domain modelled at the Turnagain ultramafic complex...... 159 Figure 7.1 Proposed duel ore processing and mineral carbonation facility at Turnagain ...... 177 Figure 7.2 A proposed experimental direct aqueous mineral carbonation workflow for testing mine waste rock ...... 182 Figure B.1 Olivine geological domain CaO charts ...... 218 Figure B.2 Olivine geological domain MgO charts ...... 219 Figure B.3 Olivine geological domain MCP charts ...... 220

xii

Figure B.4 Olivine geological domain MCP_ARC charts ...... 221 Figure B.5 Olivine geological domain MgO vs MCP scatter plot with linear regression ...... 222 Figure B.6 Wehrlite geological domain CaO charts ...... 223 Figure B.7 Wehrlite geological domain MgO charts ...... 224 Figure B.8 Wehrlite geological domain MCP charts ...... 225 Figure B.9 Wehrlite geological domain MCP_ARC charts ...... 226 Figure B.10 Wehrlite geological domain MgO vs MCP scatter plot with linear regression ...... 227 Figure B.11 Olivine clinopyroxenite geological domain CaO charts ...... 228 Figure B.12 Olivine clinopyroxenite geological domain MgO charts ...... 229 Figure B.13 Olivine clinopyroxenite geological domain MCP charts...... 230 Figure B.14 Olivine clinopyroxenite geological domain MCP_ARC charts ...... 231 Figure B.15 Olivine clinopyroxenite geological domain MgO vs MCP scatter plot with linear regression ...... 232

xiii

List of symbols and abbreviations

ARC – Albany Research Center ASU – Arizona State University CBM – Coal Bed Methane CCS – Carbon Capture and Storage COV – Coefficient Of Variation EGR – Enhanced Gas Recovery EOR – Enhanced Oil Recovery EU ETS – European Union Emissions Trading Scheme Fo – Forsterite GHG – Greenhouse Gas GIS – Geographical Information System HTHP – High Temperature High Pressure ICP-ES – Inductively Coupled Plasma Emission Spectroscopy ID – Inverse Distance IPCC – International Panel on Climate Change KE – Kriging Efficiency KV – Kriging Variance LOI – Loss On Ignition MCP – Mineral Carbonation Potential NETL – National Energy Technology Laboratory NN – Nearest Neighbour OECD – Organisation for Economic Cooperation and Development OK – Ordinary Kriging PGE – Platinum Group Elements SEM – Scanning Electron Microscopy SG – Specific Gravity SMD – Stirred Media Detritor SMU – Selective Mining Unit TKK – Helsinki University of Technology UBC – University of British Columbia

xiv

U.S. DOE – United States Department Of Energy U.S. EIA – United States Energy Information Administration U.S. EPA – United States Environmental Protection Agency XRD – X-Ray Diffraction XRF – X-Ray Fluorescence atm – Atmosphere Btu – British thermal unit kWh – Kilowatt hour M – Mole MPa – Megapascal

PCO2 – Carbon dioxide partial pressure ppm – Parts per million t – Tonne

Acknowledgements

This dissertation is part of a broader research group focusing on CO2 sequestration by mineral carbonation in the Departments of Mining Engineering and Earth and Ocean Sciences at The University of British Columbia. This research would have not been possible without the generous funding from Hard Creek Nickel Corporation and matching funding from a Mitacs-Accelerate Internship allowing me to focus on lab work and purchase a shiny new autoclave. Hopefully this research is a success story of promoting and funding collaboration between industry and academia. I have been the recipient of an applied science graduate award and international partial tuition scholarship, of which I am extremely grateful for the financial assistance in pursuing this academic adventure. I would like to thank everyone at the NBK Department of Mining Engineering at UBC for their continued support over the last 5 years. In particular my supervisor Michael Hitch for gambling on me back in 2009 and bringing me over to this amazing city that I now call home in Vancouver. I would like to give special mentions to the professors at the Department of Earth Sciences at the University of Birmingham for giving me a solid foundation in geology and the professors at Camborne School of Mines for furthering my interest and development in the mining industry through my Master’s degree. To Dirk Van Zyl and Greg Dipple, thank you for joining my advisory committee and pushing me to expand my thought process throughout this research, your early comments along with Michael’s guidance sparked my imagination to think outside of the box when attacking my research questions. Thanks to Hard Creek Nickel, especially Eric Scheel for his assistance in sample collection, providing me with some local knowledge and helping me get to grips with their geological database for estimation. I would like to thank Dassualt Systemes GEOVIA, in particular Iain Mclean, Chawki Jreige and Gabriela Brandao for giving me the flexibility to both work and pursue this challenge simultaneously. I owe a huge thank you to Sally Finora and Pius Lo for continually answering my questions and allowing me to play with their laboratory equipment; you guys do a fantastic job. This research would not have been possible without the assistance of staff at the Electron Microbeam/X-ray Diffraction Facility, Earth and Ocean Science Department, The University of British Columbia. Special thanks to Jenny Lai, Elisabetta Pani, Edith Czech and Mati Raudsepp for all of your patience and hard work in gaining manageable XRD data from ugly serpentine filled samples. Thank you Mati for having an open door and providing me with very useful guidance and wisdom in interpreting the data provided.

xvi

Finally I owe a huge debt of gratitude to all of my friends and family in both Canada and the UK. To my amazing mum for continuously pushing and supporting me to follow my dreams. To my friends and teammates for always providing an outlet from the grind and to my partner Alice for putting up with my moaning and complete lack of free-time over these last three and half years.

Dedication

xvii

This thesis is dedicated to the memory Of Alan Jacobs 1955-2004

And, the continued support of my mum, grandparents and Alice.

xviii

1 Introduction

1.1 Climate change and the carbon cycle

The carbon cycle can be considered as two sub-cycles. The larger of these sub-cycles (approximately 99.94 % of total carbon) fixes carbon as stable carbonates (approximately 80 %) or as kerogen and graphite (approximately 20 %) dispersed within sedimentary rocks (Killops and Killops, 2005: pg. 246). This carbon within the geochemical sub-cycle has a residence time of millions of years. The smaller sub- cycle exchanges carbon through the biological recycling of dissolved and non-living particulate organic matter in water bodies (primarily the oceans), and soils, together with the biota biomass and the atmosphere. This biochemical sub-cycle has residence times of up to hundreds of years. Until the industrial era (early 18th century) the two sub-cycles were linked by a small fairly equal two-way flux. Human exploitation of fossil fuels has shifted this flux from the geochemical to the biochemical sub- cycle (Killops and Killops, 2005: pg. 246). Theories and supporting evidence concerning global warming and associated climatic change have been provided and discussed amongst the growing scientific and political communities over the last few decades. The Intergovernmental Panel on Climate Change (IPCC) has provided comprehensive evaluations on the topic. The most recent synthesis report in 2007 proposes that global warming of the earth’s climate is most directly associated with the emission of anthropogenic greenhouse gasses

(GHG’s) such as CO2, N2O and CH4. Carbon dioxide has been attributed as the primary causal greenhouse gas (approximately 77 %) toward climate change (IPCC, 2007: pg. 36). An updated IPCC synthesis report is due in 2014. Carbon dioxide is released naturally into the atmosphere through animal, plant, and microorganism respiration. A significant concentration of atmospheric CO2 is also provided by release from volcanic eruptions and hot springs (U.S. EPA, 2008). It has been suggested that the combustion of fossil fuels (approximately two-thirds) and changes in land use (approximately one-third) however are the principal contributors to the exponential global increase in atmospheric CO2 concentration since the pre- industrial era (Figure 1.1). It is estimated that the atmospheric CO2 concentration has increased from 280 ± 20 ppm in 1750 to 379 ppm in 2005 (IPCC, 2007: pg. 37).

Figure 1.1 Global anthropogenic GHG emissions

(a) Global annual emissions of anthropogenic GHGs from 1970 to 2004. (b) Share of different anthropogenic GHGs in total emissions in 2004 in terms of CO2-eq. (c) Share of different sectors in total anthropogenic GHG emissions in 2004 in 2004 in terms of CO2-eq. (Forestry includes deforestation) (Source: IPCC 2007, synthesis report: fig. 2.1, pg. 36).

Both energy demand and energy consumption over the next few decades show no sign of regression. The U.S. Energy Information Administration (U.S. EIA) in their International Energy Outlook for 2013 estimate a world marketed energy consumption growth of 56 % from 2010 to 2040 (not incorporating prospective legislation or policies that might affect energy markets) (U.S. Energy Administration Agency, 2013: pg. 9). It is expected that much of the growth in energy consumption will occur in countries outside the Organization for Economic Cooperation and Development (Figure 1.2 (a)). Energy use is projected at 90 % for non-OECD nations compared to 17 % of OECD economies (U.S. Energy Administration Agency, 2013: pg. 9). Renewable energy is the world’s fastest growing form of energy, with a projected increase of 2.5 % per year between 2010 and 2040 (U.S. Energy Administration Agency, 2013: pg. 10). It is fossil fuels however that are expected to remain the principal global energy supplier with almost 80 % use until 2040 (Figure 1.2 (b)). The modest market share increase of renewable energy sources outlines the fact that it will take time to ultimately replace the current fossil-fuel dominated global energy infrastructure.

Figure 1.2 World total energy consumption and consumption by fuel type, 1990-2040

a) World total energy consumption, 1990-2040 (quadrillion Btu) (Source: U.S. Energy Administration Agency, International Energy Outlook 2013: fig. 12, pg. 9). b) World energy consumption by fuel type (quadrillion Btu) (Source: U.S. Energy Administration Agency, International Energy Outlook 2013: fig. 16, pg. 11).

It was suggested by Pacala and Socolow (2004: pg. 968) that less than double the pre-industrial atmospheric CO2 concentration (i.e., 500 ± 50 ppm) should be adhered as the maximum level at which the most damaging and irreversible effects of anthropogenic climate change can be avoided. It is however projected that the world energy-related carbon dioxide emissions will continue to rise (43 %) from 31.2 billion tonnes in 2010 to 53.2 billion tonnes in 2040 (U.S. Energy Administration Agency, 2013: pg. 159). For global CO2 emissions to be reduced to the suggested acceptable level, it is becoming increasingly clear that stable, cost-effective and energy efficient carbon storage strategies are required. Coupled with strategies for de-carbonising energy sources and increasing the efficiency of energy production, it is possible to stabilise global anthropogenic CO2 emission levels (Lackner, 2003; Pacala and Socolow, 2004).

1.2 Carbon capture and storage technologies

Numerous methods of carbon capture and storage (CCS) have been proposed, tested and some implemented through focused research in industry, government and academic institutions. The idea of

3 zero-emission technology has dramatically increased and promoted carbon sequestration as a practical method of responding to increasing anthropogenic CO2 emission levels. To develop a widespread public acceptance of carbon sequestration technologies, any program adopted, not dissimilar to nuclear disposal must be guaranteed as safe and present no legacy issues such a CO2 leakage back in to the atmosphere. CCS requires cost effective carbon capture from the anthropogenic CO2 source. This often results in the scrubbing of CO2 from other anthropogenic gases. Both safe transport to the storage site and finally a storage method are required to fix carbon in a secure manner.

1.2.1 CO2 capture

The removal and separation of CO2 from gas streams has been undertaken by industries for a long time. Capturing CO2 at point sources such as large fossil fuel processing energy facilities and various other high anthropogenic gas emitting plants is the most effective way to ensure that the maximum concentration of CO2 possible can be captured. This simultaneously reduces CO2 emissions into the atmosphere from the greatest emitting industries. To date there are three favourable active methods of capturing CO2. These are; post-combustion, pre-combustion, and oxy-fuel. The G8 (Canada, France, Germany, Italy, Japan, Russia, UK, and the USA) have targeted a goal of 20 completed CCS demonstration projects globally including post-, pre- and oxy-combustion by 2020 (Herzog et al., 2009: pg. 27).

1.2.2 CO2 transport

A key component in CCS implementation is the transport of CO2 from the captured source to the storage site. The most practical and effective method of transporting CO2 on the large scale required for CCS involves compression to high pressures and transport via a pipeline. Gas pipelines are not a new phenomenon and much experience can be drawn from the transport of natural gas. Both natural gas and CO2 are transported at similar pressures. McCoy and Rubin (2008) developed a preliminary design model based on natural gas pipelines. Their model for CO2 transport links the design of CO2 pipelines with the operating and construction costs.

Their study produced a baseline cost to transport CO2 for six major regions of the United States. They estimated a cost of CO2 transport of US$1.16 per tonne based on 5 million tonnes of CO2 transported annually over 100 km in the Midwest region. However, this cost can vary over 30 % by region and more significantly by pipeline length and design capacity. Their model shows that costs can vary from US$ 0.15 per tonne for a 10 km pipeline to US$ 4.06 per tonne for a 200 km pipeline based on a typical 500 MW

4 power plant. The model by McCoy and Rubin (2008) outlines that the costs associated with the transport of CO2 are promising for practical implementation. An emphasis on close CO2 point sources and storage sites must however be promoted.

1.2.3 Geological storage

Geological storage includes a range of methods for storing anthropogenic CO2. These are predominantly achieved by injecting supercritical CO2 in to underground formations or cavities. CO2 is in a supercritical phase (dense phase) when temperatures reach higher than 31.1°C at pressures greater than 7.38 MPa, known as the critical point (Bachu, 2002: pg. 93). For the successful storage of CO2, a sufficient ‘trapping’ mechanism such as a cap rock, analogous to that needed for the trapping of hydrocarbons in an underground reservoir is required to prevent CO2 leakage. The predominant formations targeted for geological storage include; depleted oil and gas reservoirs, deep coal seams, and deep saline aquifers.

Two potential and demonstrated reservoirs for the trapping of anthropogenic CO2 include both active and depleted fossil fuel reservoirs in underground formations. In the case of depleted oil and gas reservoirs, CO2 can be injected when there is no further promise of hydrocarbon production. CO2 may also be injected into producing oil and gas reservoirs, known as CO2-enhanced oil recovery (EOR) and

CO2-enhanced gas recovery (EGR). These methods of CO2 storage have shown to improve the production life of the reservoir through the injection of CO2 improving oil mobility. It is proposed by Davison et al.

(2001) that improved gas recovery from such a method can more than offset the cost of CO2 capture and injection.

The greatest challenges facing this method of CO2 storage however is the possibility of leakage, which can be caused by overpressure during CO2 injection and the failure of incomplete and abandoned wells to be adequately capped. As a result post-injection monitoring of the sites is required at an added expense and the possible leakage is furthermore detrimental to the public’s perception and acceptance of the technology. Geological CO2 storage in oil and gas reservoirs remain the most advance storage option available to date with well-known demonstration projects in Norway, Canada, Algeria, Australia, and several other countries (Silpilä, 2008). A detailed CCS project database can be found at http://sequestration.mit.edu/tools/projects/index.html which follows the status of major CCS projects worldwide (Carbon Capture and Sequestration Technologies at MIT).

Anthropogenic CO2 can be injected into coal seams. The CO2 has the potential to be absorbed onto the coal and can be stored within the pore matrix of the seams for geologic time. Similar to CO2-EOR and

EGR, CO2 sequestration in applicable coal seams has the ability to enhance coal bed methane (CBM) recovery (Gunter, 2009). The sequestration of CO2 in coal seams coupled with enhanced CBM recovery is an attractive storage route with the potential of the methane retrieved to provide a product to offset the sequestration costs. However to date the most suitable physical coal characteristics remains largely unknown and the subsequent CO2 generated from burning methane has not yet been defined. Like oil and gas reservoir storage, the CO2 must be guaranteed as fixed or the monitoring of CO2 leakage must take place.

The largest capacity geological storage option for CO2 globally is in deep saline aquifers. Saline aquifers are common and widely associated with sedimentary basins worldwide (Bachu, 2002). The aquifers typically contain mineralised brines that cannot be used in agriculture or for human consumption. As such they have previously been used as a storage location for injected hazardous liquid waste (Bachu, 2002). Suitable aquifers, like hydrocarbon reservoirs must be capped by a sufficient aquitard to prevent CO2 leakage through existing fractures and incomplete wells. CO2 would be stored in applicable aquifers, by injecting supercritical CO2 into the aquifer where it will rise due to buoyancy effects before spreading into a layer beneath the cap (Gale, 2004).

The relative size and quantity of deep saline aquifers and the practiced method of CO2 injection makes the option a promising one. However, little is known about saline aquifers compared to hydrocarbon reservoirs. To maintain a cost that is acceptable for CO2 storage, investigations in to the stability of the aquifer and effectiveness of hydrodynamic and mineral trapping is required to prevent possible CO2 leakage back in to the atmosphere.

1.2.4 Mineral storage Mineral storage, also known as mineral carbonation or mineral trapping can be in-situ or ex-situ methods of anthropogenic CO2 storage. Mineral storage could also be considered a form of geological storage as they both rely on geological processes for the fixation of CO2. For clarity within this research they are considered separately. In-situ mineral storage takes advantage of the high-temperature and high-pressures associated with underground hydrothermal environments. Mineral trapping is achieved through the dissolution of olivine (forsterite, Mg2SiO4), pyroxene, and serpentine with subsequent precipitation of carbonates and silica (Giammar et al., 2005; Dufaud et al., 2009; Larachi et al., 2010). In-situ mineral carbonation can be accelerated via the injection of CO2 at elevated pressured through hydraulic fracturing. Following an initial heating step with CO2 pumped at 25-30°C, exothermic carbonation reactions result in a sustained

6 temperature and increased reaction rate. It is considered an alternative to ex-situ carbonation as it relies on in-situ hydrothermal conditions to maintain optimal reaction conditions. It does not require the pre-treatment of substrate material for carbonation. The in-situ carbonation of mantle peridotite in

Oman alone could potentially consume more than 1 billion tonnes of CO2 per year, outlining the huge potential for mineral carbonation as a permanent method of CO2 storage (Kelemen and Matter, 2008: pg. 17295). Both in-situ and ex-situ mineral carbonation technology remains in its relative infancy. With regards to in-situ mineral carbonation field tests and detailed models in to the hydraulic fracturing of peridotite, cracking due to heating, hydration, and carbonation (permeability and reactive volume fraction) are required. These processes cannot practically be simulated in laboratories. Large-scale field tests are essential for its technological development (Kelemen and Matter, 2008). The disadvantages of in-situ carbonation over ex-situ carbonation are that operating conditions are not controllable. Only site selection remains as the major degree of freedom for CCS implementation (Larachi et al., 2010). Ex-situ mineral carbonation provides an advantageous position in this regard, but to date a method has not been developed that can take full advantage of the exothermic reactions that take place during carbonation. Ex-situ mineral carbonation provides the carbon capture method chosen as the focus of this dissertation. It is discussed in detail in Chapter 2.

1.3 Carbon capture and storage in the mining industry

Mineral carbonation offers both potential and actively producing hard rock mines the most appropriate proposed CCS technology option. Surface mining operations and practice are well suited for integration with mineral carbonation technology. This is due to the promise of sharing much of the required cost both through infrastructure and processing expenditure. Cost savings coupled with the widespread availability of suitable Mg-silicate rich rocks that host many economic mineral deposits offer an attractive solution to reducing CO2 emission levels at an applicable mine. The primary commodities at such mines include nickel-sulphide, chromite, PGE and diamonds. These are often hosted in mafic and ultramafic complexes (Robb, 2005). Plans to provide market-based incentives promoting a reduction in GHG levels such as a cap-and- trade policy are gaining support. Such policy structures effectively apply a price to carbon. These have been implemented in Europe in the form of the European Union Emissions Trading Scheme (EU ETS). It remains unclear whether a policy structure if any will be emplaced in North America. Whatever GHG

7 emission regulations may be emplaced, the mining industry, would be affected both financially and in the social, environmental and legislative approval of mining operations. This would be emphasised further should the mine generate its own power using a coal-fired plant at the operation. Mine’s that have mineral carbonation capabilities could use their waste rock as a value-added product. This could effectively lower the grade of material that is economical to mine. This additional revenue stream could also lower the strip ratio of a surface mining operation by virtue of a greater proportion of the material being mined, having value. As a result there may be the potential to enhance a marginal project to a level in which that mining project could become economically feasible and environmentally attractive (Hitch et al., 2010). Due to the large quantity of substrate material required for mineral carbonation, any mineral carbonation facility will have to be situated at the mine. This would minimise the cost and CO2 generated from the transport of rock and to take advantage of shared infrastructure (Figure 1.3). The

CO2 must therefore be extracted at its point source, compressed and transported to the facility via a pipeline. Integrating a processing mill with a mineral carbonation facility would allow easy use of processing tailings rich in Mg/Ca-silicate minerals and provide a facility in which Mg-rich waste rock can be reduced to the suitable size fraction required for the mineral carbonation process. An additional benefit of such operation is the ability to dispose the stable carbonated product in the excavated mine or established tailings facility. The cost estimates of implementing mineral carbonation technology have resulted in a wide array of values. This is attributed to the vast degree of uncertainty surrounding project valuations (Hindle, 2011; Hitch and Dipple, 2012). When considering olivine as the principal substrate mineral for carbonation, cost estimates have varied between US$50-100 per tonne of CO2 sequestered (Newell et al., 2000; Lyons et al., 2003; Penner et al., 2004; O’Connor et al., 2005; Gerdemann et al., 2007). Although these initial cost estimations are too high to be economically feasible to date. Compared with geological storage, the potential for permanent storage over geologic time is still desirable. Changing CO2 policies, improved technology through better understanding of the mineral carbonation process and the development of pilot-scale operations could collectively promote the progression of ex-situ mineral carbonation.

Figure 1.3 Schematic illustrating how a potential integrated mining and mineral carbonation system could function

1.4 Dissertation structure

1.4.1 Dissertation focus A key stepping stone in gaining a better understanding of how ex-situ mineral carbonation can be implemented on a global scale is the accurate determination of potential mineral carbonation substrate resources. This dissertation approaches the difficulties faced when estimating the mineral carbonation potential of mining waste rock as a substrate source. An estimation parameter and estimation process workflow is developed. It considers the effects of mineral variability on a rock mass as the principal controlling factor on the mineral carbonation potential of mining waste rock using a direct aqueous mineral carbonation process. The research targets the further understanding of this workflow by providing a tool that can take measured cation concentrations from drill hole assay data estimating the Mg2+ and Ca2+ mineral association. The values produced by this tool can act as an initial indication of the mineral carbonation potential through reported or measured mineral carbonation extents achieved. It targets specific silicates such as olivine, serpentine and diopside. The tool can assist an initial proof of concept study. Such studies aim to promote pilot-scale experimentation at any chosen deposit. This in turn can aid in the development of the industrial mineral carbonation process. Spatial resource modeling is well established within the mining community. By adapting established resource estimation methods used by the industry and utilising available geochemical data in a cost- effective manner, it is possible to estimate mineral carbonation substrate resources within mine waste rock. This is analogous to mineral commodity resource estimation. The framework developed in Chapters 5 and 6 of this dissertation provide a guideline to how ultramafic deposits could be evaluated. The research aims to outline a workflow through a case study on the Turnagain ultramafic deposit and a starting point from which the characterisation of ultramafic deposits in respect to their mineral carbonation potential can be approached and developed. By better understanding the potential resources available at mining locations, further interest and investment in to the technology required for industrial implementation can be triggered through the promotion of the environmental and economic benefits of an integrated mineral carbonation operation. Taking advantage of previously undesirable mine waste rock can remove the need for a mine to be developed for the sole purpose of supplying material to a mineral carbonation facility.

1.4.2 Dissertation questions The three questions posed outline the broad outcomes to be determined in this dissertation. Each question targets a stepping stone that is integral to the success of promoting mineral carbonation as a method of CO2 sequestration at prospective and active mining operations. Each question is coupled with a discussion on the challenges faced at the onset of the research. The desired outcomes aim to provide a workflow and associated tools required for a mining operation to generate a preliminary capacity estimate of their waste rock to mineralogically fix CO2 by mineral carbonation.

i. Question: Can an experimental workflow be generated that will allow mining companies to successfully quantify the stoichiometric silicate to carbonate conversion achievable by mineral carbonation of waste rock at their mining operation in a cost effective manner?

Discussion: Many papers on the fundamental application of mineral carbonation have provided the basis for determining the stoichiometric silicate to carbonate conversion achievable by mineral carbonation. Many substrate material sources have been investigated and experimental results are well documented in the literature. It was hypothesised that a workflow could easily be established. The challenge proposed by this question however requires the synthesis of the information available and an analysis of the existing laboratory techniques to provide the most applicable workflow for implementation at a mining operation, at a cost that is not deemed unreasonable by a mining company. Any workflow proposed however, must promote the best practices to ensure data integrity.

ii. Question: Can a mineral carbonation potential (MCP) parameter be developed that combines minimal laboratory-scale experimentation and available exploration data at mining operations to generate an acceptable mineral carbonation potential estimate for discrete rock units at a mining operation?

Discussion: The successful conversion of magnesium and calcium silicates to their respective carbonates has been established in a large number of academic papers at a laboratory scale. The varying conversion extents have been shown to be largely controlled by substrate mineralogy. The subsequent estimation of the modal weight mineral composition of a discrete rock unit is highlighted as a key facilitating factor in the accurate determination of mineral carbonation

potential. It is postulated that modal mineral estimation through subtractive mass balance or best-fit linear programming approach coupled with the achievable stoichiometric silicate to carbonate conversion extent achievable by a given mineral can be used to better estimate the mineral carbonation potential of the rock mass. The growing research on mineral carbonation process streams makes the incorporation of mineral carbonation extents problematic as there is no one-size-fits-all solution and each deposit should be evaluated singularly.

iii. Question: Can an MCP parameter predicting the mineral carbonation potential of a discrete unit

of rock be geospatially interpolated to provide a preliminary resource estimation of the CO2 capacity of mining waste rock?

Discussion: An MCP value that quantifies the mass of CO2 that can be sequestered in a given volume of rock can be considered analogous to an economic commodity element that can be extracted from the rock. Any MCP value generated will be a product of the geochemical and mineralogical properties of the rock and controlled by those properties. Like commodity resource estimation procedures traditionally undertaken by mining companies at a pre- feasibility and feasibility stage of a mining operation, geostatistical techniques are expected to successfully spatially interpolate the mineral carbonation potential of the rock mass considered waste at a mining operation.

1.4.3 Chapter organisation Chapter 2 of the dissertation provides a review of the literature present on mineral carbonation. This includes a comparative evaluation of the methodologies used to convert magnesium silicate to magnesium carbonate minerals sequestering anthropogenic CO2 over the last two decades. A background in to the types of deposits that should be targeted for mineral carbonation is provided along with the rationale behind using mining waste rock as a principal substrate material. The benefits of developing integrated mining and mineral carbonation operations at applicable ultramafic deposits are highlighted. A direct aqueous mineral carbonation process route developed at the Albany Research Center, USA is the selected experimental process used in this research. This research does not directly offer a progression in the process development described and evaluated in this chapter, but applies the process to an ultramafic suite of rocks typical to a nickel sulphide host deposit in British Columbia, Canada. This aims to assess the applicability of the process to

12 such rocks and indirectly explores the effectiveness of the direct aqueous mineral carbonation of these rocks. The findings in turn are used to develop an estimation technique that predicts the mineral carbonation potential of such deposits. Previous attempts at modal mineral estimation from geochemical data are reviewed. Their applicability to mineral estimation at ultramafic deposits is analysed and the challenges in mineral estimation at ultramafic deposits and relevance to mineral carbonation potential estimation is put into context. In Chapter 3 the experimental methods used in the research are presented, including particle size analysis, scanning electron microscopy, X-Ray diffraction, X-Ray fluorescence, and elemental assay analysis. Descriptions and location justifications are provided for the two ultramafic complexes selected to provide the substrate material used in the experimental portion of the research. The two ultramafic complexes chosen are the Twin Sisters deposit, Washington State, USA and the Turnagain deposit, British Columbia, Canada. An experimental direct aqueous mineral carbonation process is selected to sequester CO2. The justification and details of the laboratory methods undertaken are outlined in the final section of this chapter. The experimental mineral carbonation portion of this dissertation is one of a three-step process targeted at effective mineral carbonation potential estimation. Approaches at both modal mineral estimation and three-dimensional modelling taken in this research are described analytically in the context of developing an MCP parameter. The experimental mineral carbonation results are provided in Chapter 4. The results presented include the stoichiometric magnesium silicate to carbonate conversion achievable for the Twin Sisters olivine. This uses an experimental mineral carbonation laboratory set up by the author at the University of British Columbia. The results presented are compared with published literature on experimental mineral carbonation. The established direct aqueous mineral carbonation procedure was replicated, testing 15 samples obtained from diamond drill core at the Turnagain deposit in British Columbia, Canada. These results add to the published experimental results on direct aqueous mineral carbonation. They furthermore provide an insight in to the mineral carbonation capabilities of waste rock from active or potential mining operations. Both qualitative and quantitative descriptions of the reactant and product material are provided. These include geochemical, mineralogical, textural changes that occurred during carbonation. The variable stoichiometric magnesium silicate to carbonate conversion rates is discussed. An MCP parameter is proposed and defined in Chapter 5. The chapter provides a detailed breakdown of the steps involved in generating MCP values using the MCP calculator. The MCP calculator is a Microsoft ExcelTM based series of worksheets that has been developed in this research. The

13 calculator is designed specifically to calculate the modal mineral assemblage present at Turnagain. Up to 30 sample whole rock geochemical samples can be input into the MCP calculator and an estimate of the MCP of each rock sample is produced. Results produced by the MCP calculator are analysed and compared with quantitative XRD results using the Rietveld refinement method. This is completed on a sample from the Twin Sisters ultramafic deposit and the 15 rock sample tested from the Turnagain ultramafic deposit that underwent experimental mineral carbonation discussed in Chapter 4. The degree of uncertainty associated with the results produced by the MCP calculator and the applicability of the developed MCP parameter are defined. Chapter 6 proposes a workflow to evaluate an ultramafic deposit for its potential to host an industrial ex-situ mineral carbonation operation. An example of the workflow is provided through the evaluation of the Turnagain ultramafic deposit as the mineral carbonation resource. Whole-rock geochemical data is processed by the MCP calculator to generate a suite of MCP values for the deposit. These values are subsequently interpolated using a range of geostatistical techniques to demonstrate the ability of MCP values to be geospatially estimated. An analysis of the resource estimation results and an evaluation of the proposed workflow are provided. Chapter 7 discusses the principal outcomes of this research dissertation in relation to the three questions proposed in this chapter. Chapter 8 summarises the conclusions that can be taken from the research and provides scope for future research on the topic area. Finally, a personal reflection on the findings and challenges posed from the research going forward are expressed by the author.

2 Literature review: ex-situ mineral carbonation and modal mineral estimation

The layout of the chapter mimics the layout of the dissertation as a whole. It reviews the published literature on the topic of ex-situ mineral carbonation. It synthesises the published information on the progression in mineral carbonation technology and processing regimes developed to accelerate mineral carbonation. It critically analyses the experimental direct aqueous mineral carbonation process path which has emerged as one of the most promising paths to date and how the literature has influenced the experimental mineral carbonation undertaken in this research. The results of this experimental mineral carbonation are presented in Chapter 4. Mineral variability within ultramafic rocks is exposed as an overlooked factor when developing preliminary mineral carbonation capacity estimates. A review of available modal mineral estimation techniques and their applicability to mineral carbonation estimation is given. A perspective on the challenges faced when estimating modal mineral abundance and a subsequent mineral carbonation resource capacity is provided. This is with the focus of generating a mineral carbonation potential (MCP) parameter for use by the mining sector. This developed MCP parameter is the focal point of Chapter 5. Finally a synthesis of geostatistical and empirical techniques adopted by the mining industry for traditional commodity resource estimation is given. This is in advance of analysing the application of these techniques to MCP data in Chapter 6.

2.1 Mineral carbonation background

Mineral carbonation is a naturally occurring process in which metal oxide-bearing silicate minerals such as magnesium and calcium are converted to carbonate minerals and elemental silica through the addition of carbon dioxide (Seifritz, 1990). These silicate minerals have unique advantages as raw substrate material for anthropogenic CO2 sequestration in solid form. The resultant carbonate minerals have a lower energy state than CO2. This prevents possible CO2 leakage issues over a geological timescale, a common concern with alternative CO2 carbon sequestration techniques such as oil and gas reservoir storage. Conventional carbonation pathways in nature are slow under ambient pressures and temperatures. One of the biggest challenges facing the application of mineral carbonation as a method of CO2 sequestration is identified as developing a mineral carbonation process route that is industrially, economically and environmentally viable (Goldberg et al., 2001).

Calcium and magnesium are the only common elements that form thermodynamically stable carbonate minerals. Iron forms a marginally stable carbonate and as a result has been considered as a potential substrate element for mineral carbonation (Goff et al., 2000). The presence of iron in quantities that may be required as a primary source of substrate material are however minimal. Oxides and hydroxides of calcium and magnesium would be favourable sources for carbonation due to their high reactivity, yet because of this high reactivity they are rarely found in nature (Bearat et al., 2006). Magnesium silicates are favoured over calcium silicates. This is largely due to them being more widespread, forming larger bodies and containing more reactive material per tonne of rock (Lackner et al., 1997). Two magnesium silicates; olivine (Equation 2.1) and serpentine (Equation 2.2), and one calcium silicate; wollastonite (Equation 2.3) have emerged as the most promising substrate sources.

Mg2SiO4 (olivine) + 2CO2  2MgCO3 (magnesite) + SiO2 Eq. 2.1

Mg3Si2O5(OH)4 (serpentine) + 3CO2  3MgCO3 + 2SiO2 + 2H2O Eq. 2.2

CaSiO3 (wollastonite) + CO2  CaCO3 (calcite) + SiO2 Eq. 2.3

It was shown that olivine, dissolved in hydrochloric acid (HCl), produces silica and magnesium chloride (MgCl2). Electrolysis of the MgCl2 produces magnesium and regenerates HCl. During this acid- recovery step, crystallised MgCl2 is converted to magnesium hydroxide (Mg(OH)2) that can subsequently be reacted with CO2 to form magnesite (MgCO3) (Lackner et al., 1995; 1997). Although technically feasible, Nilsen (1999) showed this process to be economically unfeasible and having four times the energy consumption than required for burning the coal to produce CO2. This initial research led by Lackner and co-worker’s (including C.H. Wendt., D.P. Butt., E.L. Joyce., D.H. Sharp., and H-J. Ziock) at the Albany Research Center (ARC) and National Energy Technology Laboratory (NETL), USA, developed a platform from which the idea of industrialised mineral carbonation as a method of CO2 sequestration could be built.

2.2 Mineral carbonation process pathways

From Lackner and co-worker’s initial work, research into developing an industrially viable mineral carbonation process pathway has diverged and progressed in varying directions. The majority of the

16 early research has focused on a direct carbonation pathway. This is where carbonation is achieved in a single process step (i.e. silicate dissolution and carbonate precipitation within a single reaction vessel). More recently, research has also started to explore indirect carbonation methodologies. Here calcium or magnesium is extracted from the reactant mineral and subsequently carbonated in multiple process steps (e.g. Park and Fan, 2004) within which, process conditions can be carefully controlled.

2.2.1 Direct mineral carbonation Direct mineral carbonation is accomplished using either a wet or a dry process. Direct gas-solid carbonation was initiated in Finland (Kohlmann et al., 2002; Zevenhoven et al., 2004). Mineral carbonation was undertaken by exposing Mg/Ca-bearing metal oxides to gaseous CO2 at elevated temperatures up to 350°C. This offers the possibility to cover the energy requirements with heat from the carbonation reaction (Zevenhoven et al., 2004; Zevenhoven et al., 2006). Very slow reaction kinetics and thermodynamic limitations practically resulted in the discontinuation of research toward this method of carbonation. Although the majority of research into direct gas-solid mineral carbonation discontinued. The early investigations however triggered research toward its use within multi-step indirect gas-solid routes (Sipilä et al., 2008). O’Connor and co-worker’s (O’Connor et al., 1999; O’Connor et al., 2000; O’Connor et al., 2001) almost simultaneously with research in Finland, proposed an alternative direct aqueous mineral carbonation pathway. Using this exothermic process route, carbon dioxide is dissolved in a slurry mixture consisting of water and a suitable mineral reactant such as olivine. The dissolved CO2 forms + - + carbonic acid (H2CO3), which dissociates into H and HCO3 (Equation 2.4) with H hydrolysing the silicate mineral, releasing the cation from the mineral reactant (i.e. Mg2+). The resulting cation is now free to

- react with HCO3 forming the solid magnesium carbonate (Equation 2.5). The combination of mineral dissolution reactions and carbonation occur within a single unit-process. This occurs under high partial pressure (PCO2) resulting in carbonic acid being continuously regenerated as it is consumed until the mineral reactant is exhausted (O’Connor et al., 1999; O’Connor et al., 2001).

+ - CO2 + H2O  H2CO3  H + HCO3 Eq. 2.4

Mg2SiO4 + 2CO2 + 2H2O  2MgCO3 + H4SiO3 Eq. 2.5

Reaction rates have shown to dramatically accelerate by varying pre-treatment methods (e.g. chemical, mechanical, thermal), decreasing the particle size, raising the reaction temperature, increasing the pressure, changing the solution chemistry and the use of a catalyst. The successful matrimony of these approaches has demonstrated consistent rates of over 80 % carbonation in less than one hour at a laboratory scale (Gerdemann et al., 2003; Gerdemann et al., 2007). A consequence of these elevated conversion rates is an additional cost, with estimates varying between US$50 and US$100/t CO2 and an elevated energy requirement of between 200 and 300 kWh/t of CO2 stored (O’Connor et al., 2005: pg. 16). These cost estimates outline the work still required to promote mineral carbonation as an industrially viable carbon sequestration option. With no ‘value’ of carbon defined in North America and the relative cost of other carbon sequestration methods such as oil and gas reservoir storage (less than US$15/t CO2 (Sipilä et al., 2008, pg. 34.)), mineral carbonation must achieve comparable costs per tonne of CO2 sequestered. Slightly increased costs could be accepted based on the permanent fixation of CO2 in carbonates and their inherent stability. Yet the economics of any storage method is likely to take precedence for industrial application. Many researchers (Chizmeshya et al., 2002; McKelvy et al., 2004; Teir et al., 2007) have attributed the rate-limiting step of direct aqueous mineral carbonation to the generation of a silica–rich passivating layer on the surface of olivine particles when olivine is used as the primary substrate mineral. The silica- rich layer was shown to thicken as magnesium ions are leached from the olivine particles as carbonation proceeds. The resulting silica-rich layer on the particle surfaces inhibits further carbonation by reducing the surface area of fresh olivine available for dissolution. Attempts to prevent or overcome this rate- limiting step by researchers in the literature are discussed in section 2.4. The still unresolved direction of carbonation process development signifies that mineral carbonation is today still an immature technology. Studies remain at laboratory scale and have not yet reached a level in which a detailed and supported assessment can be made. It is imperative that both cost and energy requirements be reduced at the mineral pre-treatment stage. This could be aided by the utilisation of the exothermic nature of the reaction and improvements to reaction kinetics. In parallel to reducing the cost and energy requirements, the development of a continuous processing system is required. A high-temperature, high-pressure (HTHP) flow reactor was designed and developed at ARC (Gerdemann et al., 2003; Penner et al., 2003; Penner et al., 2004). The initial flow reactor replicated results produced using batch reactors. However, developments ceased at the same time as much of the work on mineral carbonation at ARC and NETL. A significant concern remains on the ability to design a

18 continuous reactor capable of processing the large volume of material whilst maintain sufficient pressures to complete carbonation. The development of new continuous process systems such as that proposed by Munz et al. (2009) in which silicate dissolution and carbonate precipitation stages are separated in two pressurised reactors have the capability to advance the knowledge in the underlying carbonation mechanisms whilst advancing the technology required for an industrial pilot carbonation plant. By defining the potential substrate resources from mining operations that waste rock capable of sequestering CO2 through mineral carbonation, it may be possible to attract interest and the funding requirements to advance the laboratory-scale mineral carbonation to a pilot-plant scale. By quantifying the mineral carbonation potential of existing resources as demonstrated in this research, mining companies will be better equipped to make the investment decisions required to advance mineral carbonation from a laboratory to an industrial scale.

2.2.2 Indirect mineral carbonation Indirect mineral carbonation is a multi-stage process that involve the extraction of metal-bearing oxides or hydroxides (usually Mg and/or Ca), the reaction of these oxides and/or hydroxides with CO2 resulting in the precipitation of corresponding carbonates. Similar to direct mineral carbonation the majority of the process paths investigated has followed an aqueous regime. Research largely originating from Finland at the Helsinki University of Technology (TKK) proposed a three-stage gas-solid carbonation route. This process involves MgO extraction from Mg-silicates in a reactor under elevated temperature (>500°C) at atmospheric pressures. The MgO is then hydrated (Stage 2) before carbonation takes place by reacting the Mg(OH)2 with CO2 under elevated pressures (up to 45 bar) (Zevenhoven et al., 2008: pg. 364). Progress in improving reaction rates for dry gas-solid carbonation has been made, however the process still remains considerably too slow for industrial implementation (Sipilä et al., 2008). The most successful dry gas-solid multi-step process developed to date by Stasiulaitiene et al. (2011) shows promising results that could potentially rejuvenate interest in the process. They achieved mineral carbonation extents equivalent to those provided by direct aqueous mineral carbonation (O’Connor et al., 2005). The process produces magnesium hydroxide using ammonium sulphate as a reagent that is later recovered. Carbonation takes place in a fluidised bed at 20-40 bar and temperatures ranging between 450°C and 550°C. This method takes advantage of the carbonation reaction heat with no expensive or non-recoverable additives used. The magnesium hydroxide extracted can be carbonated to 45-50 % within 20 minutes at 20 bar and approximately 500°C (Stasiulaitiene et al., 2011: pg. 2693).

These results can be considered comparable to direct aqueous carbonation. The limited success is yet outlined by only being able to achieve comparable mineral carbonation extents shown using the direct aqueous pathway in early research without significantly reducing the pressure and temperature need. Multi-stage aqueous mineral carbonation allows both dissolution and carbonation parameters to be controlled separately. This is achieved by varying temperature, pressure and the help of various solvents and additives to leach cations in dissolution and act as CO2 carriers for carbonation. Kakizawa et al. (2001) initiated this research by proposing a two-step method using acetic acid to extract Ca2+ (Equation

2.6) from wollastonite (CaSiO3) with a subsequent precipitation step (Equation 2.7) in which calcium carbonate is produced.

2+ - CaSiO3 + 2CH3COOH (acetic acid)  Ca + 2CH3COO + H2O + SiO2 Eq.2.6

2+ - Ca + 2CH3COO + CO2 + H2O  CaCO3 + 2CH3COOH Eq. 2.7

Numerous acids and bases have been tested for their effectiveness as solutions to promote the extraction of magnesium from silicates such as serpentine. These include; HCl, H2SO4, HNO3, HCOOH,

CH3COOH, NaOH, KOH, NH3, NH4Cl, (NH4)2SO4, and NH4NO3 (Teir et al., 2007). During these extraction experiments, it was determined that the most effective acid for leaching magnesium was sulphuric acid

(H2SO4). No acid was able to target magnesium and leach it selectively. Ammonia salts is one such reagent that can effectively target magnesium for extraction, but the quantity of magnesium extracted is low. With regards to serpentine, product layer diffusion is attributed as the rate limiting step for magnesium extraction (Teir et al., 2007). Park and Fan (2004) determined that the optimum pH for aqueous carbonation is approximately 10, while the dissolution of serpentine is more likely to occur at a much lower pH to release the magnesium cations from the silicate. Based on these findings they proposed a multi-step pH swing process to achieve increased carbonation. Dissolution of serpentine was achieved by using various solvents at elevated temperature and ambient pressures. After the dissolution process, the solid product was separated by vacuum filtration. The second product, iron oxide, was precipitated from the cooled Mg- and Fe-rich solution by raising the pH of the solution with NH4OH. Finally a third product MgCO3 was precipitated by bubbling pure CO2 into the solution. Due to higher partial pressures and a further increase in pH with additional NH4OH, carbonation occurred spontaneously. The lower operating conditions of this method are promising as it does not require the energy intensive heat pre-treatment

20 step required for serpentine using the direct aqueous mineral carbonation method. Work remains however to achieve the stoichiometric conversion of heat-treated serpentine to magnesite achieved by direct aqueous mineral carbonation. The scaling-up of the process and the recycling of chemicals are essential if this process is to be considered for industrial application. The separation of the dissolution and precipitation stage required for effective mineral carbonation is a logical one as it permits the careful control on the temperature, pressure and pH required for both processes. Progress is being made on the dissolution of both serpentine (Yoo et al., 2009; Sanna et al., 2013) and olivine (Bonfils et al., 2012) to enhance carbonation efficiencies primarily through the use of solvents. The scale required for the long-term implementation of a multi-step mineral carbonation process requires significant research and testing in the near future. A process must be defined that can optimise reaction rates, recycle reagents and minimise the infrastructure and process running costs.

2.3 Geology and sources of substrate material

In nature, carbonate rocks often develop from the interaction of aqueous fluids with silicate-rich rocks enriched in calcium and magnesium (Seifritz, 1990). Silicate-rich rocks are globally abundant with silicate minerals comprising approximately 30 % of all minerals. The potential resource of silicates for mineral carbonation is thus vast. A recent evaluation of mineral reserves in the USA, estimated that there is enough material to sequester the current total CO2 emissions (approximately 7 Gt/yr) in the USA for more than 500 years (Zevenhoven et al., 2011: pg. 49). Various solid industrial waste products including mining waste and tailings, lignite fly ash, and blast furnace slag waste can also be considered a major source of substrate material for mineral carbonation. This due to the waste product’s having a small average particle size and abundant magnesium and calcium concentration.

2.3.1 Mafic and ultramafic rocks Mafic and ultramafic rocks are igneous to meta-igneous rocks. They have a chemistry derived from their mantle origin. They are typically composed of more than 90 % mafic minerals (high in magnesium and iron) and low in silica (typically less than 45 %) (Robb, 2005: pg.24). Suitable magnesium silicate-rich rocks for use in industrial mineral carbonation are often associated with ultramafic igneous intrusions (Figure 2.1). These can be divided into three major categories; Alpine, Alaskan and layered intrusive types.

Figure 2.1 IUGS Classification diagram for intrusive ultramafic rocks

IUGS Classification diagram for intrusive ultramafic rocks determined by the modal percentage of the mafic minerals (Source: modified after Le Maitre, 2002: fig. 2.9, pg. 28).

The most voluminous and widespread ultramafic rocks are the ‘Alpine’ peridotites. These are found as basal sequences of ophiolites; slabs of oceanic crust that are uplifted and eroded along past and present subduction zones and plate boundaries (Dewey, 1976; Coleman, 1977). Alpine-type ultramafic complexes are tectonically emplaced and typically comprise of pods and cumulate forsteritic (Mg-rich) dunite that can be variably serpentinised. Podiform, stratiform or thinly layered chromitite is also common. Examples of Alpine-type ultramafic complexes in British Columbia include the Nahlin Complex and Cache Creek Complex (Voormeij and Simandl, 2004). A study by Goff et al. (2000: pg. 4) demonstrated that ophiolites are found in orogenic belts throughout most of the world. They can have discontinuously exposed dimensions up to 1,000 km by 100 km. The relative size of these bodies makes them optimal sources of substrate material based on their volume of magnesium silicate minerals that can be potentially converted to magnesium carbonates. Their study also recognises that within North America, ophiolitic belts are found along the Appalachian Mountain chain. This chain stretches from the southeastern United States into Quebec and Newfoundland. They can also be found along the Cordilleran mountain chain stretching from Alaska through British Columbia to California. The quantity and size of these deposits found in conjunction with

22 the major population centres and many large CO2 industrial point sources throughout North America result in a great potential for coupled CO2 source and sequestering locations to be achieved. Many of these deposits would need mining operations to be established to extract the substrate material for use in industrial mineral carbonation. Alaskan-type complexes are podiform intrusions that typically occur in accreted island arcs. They are mafic and ultramafic intrusions that occupy a narrow belt that trends northward for approximately 600 km in southeastern Alaska. An implication of this is their remote location relative to population centres and industrial CO2 point sources. Similar ultramafic bodies can also be found in belts through the interior of British Columbia (Irvine, 1987). Examples of Alaskan-type ultramafic complexes in British Columbia include the Tulameen Complex and Turnagain Complex (Voormeij and Simandl, 2004). Voormeij and Simandl (2004) provide an overview of the ultramafic rocks present in British

Columbia as a means for delineating targets for CO2 sequestration by mineral carbonation. Their study introduces an important first stage in exploring for potential mineral carbonation substrate sources. They achieved this through determining the aerial exposure extent of Mg-rich ultramafic deposits. This research provides a next-stage exploration tool advancing from research proposed by Voormeij and Simandl (2004) by using geochemical data available from drill core analyses available at those deposits that have been drilled for metallic commodity exploration to quantify the mineral carbonation potential more accurately. British Columbia contains numerous Alpine-type and several Alaskan-type mafic and ultramafic complexes. Alaskan complexes are typically characterised by successive wehrlite, clinopyroxenite, and hornblende-rich lithologies around a dunite core. This dunite core can be massive in nature and primarily composed of forsteritic olivine exposed over large areas. They can often be found well preserved (Irvine, 1987). The mineral carbonation targets delineated by Voormeij and Simandl (2004) are illustrated in Figure 2.2.

Figure 2.2 A generalised distribution of dunite and serpentinite-bearing rocks in British Columbia, Canada

Major Alpine-type ultramafic complexes identified: 1. Cogburn Emory Zone, 2. Coquihalla Serpentine Belt, 3. Bralorne-East Liza, 4. Bridge River Complex, 5. Shulaps, 6. Chapperton Group, 7. Mount Ida Assemblage, 8. Southern Cache Creek Complex, 9. Crooked Amphibolite, 10. Antler Formation, 11. Central Cache Creek Complex, 12. Manson Lake Complex, 13. Blue Dome Fault Zone, Sylvester Allochthon, 15. Cassiar and McDame, 16. Zus Mountain, 17. Northern Cache Creek Complex (includes Atlin and Nahlin complexes). Major Alaskan-type ultramafic complexes identified: 18. Tulameen, 19. Polaris, 20. Wrede, 21. Hickman, 22. Lunar Creek, 23. Turnagain. (Source: modified after Voormeij and Simandl, 2004: fig.6 pg.162).

2.3.2 Mining waste rock and solid industrial waste streams Two predominant sources of substrate material for mineral carbonation can be considered from some active or potential mining operations, namely; mafic and ultramafic mine waste rock and associated mill processing tailings. Naturally occurring mineral carbonation has been demonstrated at a number of mine tailings sites globally. These outline the potential of an integrated mining and mineral carbonation process. The observation of magnesite (MgCO3) and hydrated Mg-carbonates including hydromagnesite (Mg5(CO4)(OH)2•4H2O), nesquehonite (MgCO3•3H2O), and dypingite

(Mg5(CO4)(OH)2•5H2O) were shown to actively trap carbon in tailings at numerous deposits. The findings are summarised in Table 2.1.

Table 2.1 Published literature on minerals tailings for carbon sequestration

Deposit Summary Source Cassiar chrysotile deposit, Negligible hydrated Mg-Carbonates Wilson et al. 2006; BC, Canada ** precipitated Wilson et al. 2009a

Clinton creek, Yukon, ~164,000 ± 16,400 tonnes of CO2 bound Wilson et al. 2006; Canada ** in tailings pile over 26 years (~2,600 Wilson et al. 2009a

tCO2/yr)

Mount Keith Nickel Mine, Hydromagnesite sequesters ~58,000 Wilson et al. 2009c

Western Australia * tCO2/yr

Diavik diamond mine, 0.5 wt.% nesquehonite traps ~1,600 Wilson et al. 2009b;

Northwest Territories, tCO2/megatonne of tailings Wilson et al. 2009c Canada *

Selected published literature on the potential of mineral tailings to passively trap anthropogenic CO2 at active (*) and closed (**) mine sites.

The findings summarised in Table 2.1 show that mineral carbonation can occur passively at mine sites within the mill processing tailings. The rates proposed are probably too slow, difficult to quantify, and store too little CO2 to be established as a single CCS option for a mine. The combination and integration of passive mineral carbonation and more aggressive, high temperature and pressure mineral carbonation of both mill processing tailings and the use of ultramafic waste rock can provide the quantity of substrate required for the industrial application of mineral carbonation technology. Due to the large amount of substrate material required for mineral carbonation, any mineral carbonation facility will have to be situated at the mine. This is to minimise the cost and CO2 generated

25 from the transport of rock and to take advantage of shared infrastructure. As a result the CO2 must be extracted at its point source, compressed and transported to the facility via a pipeline. Integrating a processing mill with a mineral carbonation facility would allow easy use of processing tailings rich in Mg/Ca-silicate minerals. It would subsequently provide a facility in which Mg-rich waste rock can be reduced to the suitable size fraction required for the mineral carbonation process. Solid industrial waste streams may also have a key role to play in introducing mineral carbonation technology. The majority of the waste streams considered for mineral carbonation use calcium silicate material as the principal substrate source. Streams considered include lignite fly ash (Back et al., 2006; Soong et al., 2006; Montez-Hernandez et al., 2009), waste incinerator bottom ash (Bertos et al., 2005; Rendek et al., 2006; Rendek et al., 2007), blast furnace slag (Eloneva et al., 2007; Teir et al., 2007; Huijgen et al., 2005; Huijgen et al., 2006), and waste cement (Katsuyama et al., 2005; Stolaroff et al., 2005; Yamasaki et al., 2006). Industrial waste material is often appropriately sized. This eliminates much of the cost associated with further size reduction of mined material. However as Lackner et al. (2008) postulate, the qualities of these resources are typically poorly defined and quantities are often small. This may restrict their potential to niche applications. Hazardous waste that can be remediated by carbonation or a company could sequester sufficient quantities of CO2 as to benefit a company that can minimise its own emissions are targeted. It is becoming clearer that such waste streams can be crucial in introducing mineral carbonation as a valid industrial technology. This is largely a result of the direct association with high CO2 emitting industries, the typical close proximity to CO2 sources and the availability of prepared mineral sources.

2.4 Experimental direct aqueous mineral carbonation

The direct aqueous carbonation process route still provides the most promising and consistent carbonation conversion rates in the literature to date (O’Connor et al., 2005; Gerdemann et al., 2007). It is the method chosen for experimental mineral carbonation undertaken in this research. This is due to it still being widely regarded as the state-of-the-art process stream. It forms the baseline from which other mineral carbonation methods are compared. Notable advances in the experimental methodology occurred between 2000 and 2007. This was predominantly from ARC and Arizona State University (ASU). Mineral carbonation experiments targeted specific minerals in ultramafic rocks from three regions of the USA, namely; southwest Oregon lizardite (serpentine), Cedar Hills antigorite (serpentine), Twin Sisters

26 olivine (dunite). These ultramafic rocks were chosen due to their relative large abundances of source rock, proximity to CO2 sources and their relatively pure nature (i.e. dominantly magnesium-silicate mineralogy) (Gerdemann et al., 2004: pg 3). Key papers studying the Twin Sisters olivine (dunite) are provided in Table 2.2.

Table 2.2 Significant literature published on direct aqueous mineral carbonation

Principal Author Year Focus of Study Reference F. Goff 1998 Sequestration potential estimation Goff and Lackner, 1998; 2000 Goff et al., 2000 W. K. O’Connor 2001 Proof of concept study; feed material O’Connor et al., 2000; analysis; solution chemistry modifications; 2001 reaction kinetics; reaction product analysis W.K. O’Connor 2002 Mineral chemistry; mineral activation; O’Connor et al. 2002 Reaction sequence S.J. Gerdemann 2003 Temperature and pressure modifications; Gerdemann et al., 2003 mineral activation S.J. Gerdemann 2004 Preliminary process feasibility; carbonation Gerdemann et al., 2004 controlling factors; flow-loop reactor studies W.K. O’Connor 2004 Energy and economic evaluation O’Connor et al., 2004 G.E. Rush 2004 In-situ sequestration potential Gerdemann et al., 2004

C. Summers 2004 Effect of SO2 on mineral carbonation Summers et al., 2004 L. Penner 2004 Energy costs, pre-treatment options and Penner et al., 2004 flow-loop reactor studies M.J. McKelvy 2006 Particle size and distribution effects through McKelvy et al., 2006 flow-loop reactor studies K.S. Lackner 2008 Sequestration potential estimation Lackner et al., 2008

Significant literature published on the research findings undertaken on the development of an accelerated direct aqueous mineral carbonation process using the Twin Sisters olivine.

The research undertaken on these ultramafic rocks played a key role in the development of the experimental mineral carbonation process. It evaluated the rate determining parameters and attempted to optimise process conditions. Experimental mineral carbonation research on ultramafic deposits to date (O’Connor et al, 2005; Gerdemann et al., 2007) has focused on surface mining operations. This is where an olivine or serpentine body is mined exclusively for potential use in mineral carbonation. It is becoming clearer that the economics of such operations are economically unfeasible. Attention toward traditional metal mining operations with ultramafic host rock geology may prove a more feasible option. Reclamation of serpentine mining waste rock (Wilson et al. 2006; Wilson et al. 2009) or industrial waste

27 streams (e.g. lignite fly ash, waste incinerator bottom ash, blast furnace slag, and waste cement) that can provide reactive substrate material for mineral carbonation should be focused on. Experimental results generated from research at ARC and ASU provide a standard reference for future experimental mineral carbonation. This is a consequence of them providing the most promising carbonation extents achieved to date. In the case of existing or potential mining operations that are commonly associated with sulphide mineralisation, the purity and homogenous nature of the ultramafic bodies for use as substrate material such as the Twin Sisters olivine is far from guaranteed and thus requires investigation. The completion of experimental mineral carbonation using the process developed at ARC and ASU on proposed mining waste rock from a potential nickel sulphide mining operation is undertaken in this research.

2.4.1 Mineral pre-treatment The stoichiometric conversion extent of magnesium silicate to magnesium carbonate minerals is principally controlled by the experimental method, the degree of mineral pre-treatment, and the reaction conditions controlling the ability to leach magnesium and precipitate magnesite. Both mechanical and thermal pre-treatment processes have been effective in the acceleration of the mineral carbonation rate and extent (Gerdemann et al., 2004). Mechanical pre-treatment effectively decreases the mean particle size, increasing the substrate surface area available for carbonation. Prior to experimental carbonation, grinding is undertaken using conventional rod- and ball-mills (-38-75 µm), with increased activation achieved (-2-4 μm) using a sealable stirred-media detritor (SMD). An increased energy penalty from 13 kW•h/t to achieve -75 µm and 233 kW•h/t to achieve -2-4 μm (Table 2.3) is incurred. Size reduction has been particularly effective in increasing the carbonation conversion rate to over 400 % when using olivine as the substrate material (O’Connor et al., 2001; Kleiv and Thornhill, 2006; Baláž et al., 2008). The energy penalty associated with the increased activation to -4 µm suggests that the industrial application of such activation is unlikely. Size reduction to -75 µm and -38 µm are size fractions often achieved by conventional mining comminution and although small are not beyond achievable for a potential mineral carbonation operation.

Table 2.3 Energy consumption of mineral pre-treatment for mineral carbonation

Feed Material Pre-treatment Energy Consumption, kW•h/t Mineral Pre-Treatment Method Crush Grinding Heat Total -75 µm -38 µm -4 µm Treated Ball Mill (-75 µm) 2 11 12 Olivine Ball Mill (-38 µm) 2 11 70 83 SMD Mill 2 11 70 150 233 Ball Mill (-75 µm) 2 11 13 Serpentine Heat Treatment (-75 µm) 2 11 293 306 Heat Treatment (-38 µm) 2 11 70 293 376

Estimated energy consumption for mineral pre-treatment by material and pre-treatment method for direct aqueous mineral carbonation (Source: modified after Gerdemann et al., 2007: table 2, pg. 2591).

Thermal pre-treatment has proven effective at removing the chemically bound water present in serpentine samples. This changes the structure of the mineral by increasing both porosity and the surface area (O’Connor et al., 1999). The most effective heat-treatment option has been shown to heat the serpentine to 630˚C in an inert atmosphere (Gerdemann et al., 2004: pg. 5). The energy cost of this process is significantly high and would almost require the energy produced by a coal-fired power plant itself (Nilsen et al., 1999). The problem of gaining a balance between decreasing the daily substrate material requirements through proven increased reactivity by mineral activation and the coincident parasitic energy loss associated with the activation process remains unresolved (O’Connor et al., 2005). These problems are most acute for the heat-treatment of serpentine and thus it is highly unlikely that such a process could be undertaken at an industrial level. High-intensity attrition grinding has proven an effective method of activating serpentine for carbonation (Gerdemann et al., 2004). The issue of process-inhibiting hydroxylation remains and is examined with respect to mineral pre-treatment in detail by Dlugogorski and Balucan (2014). Thus, when considering the scalability of mineral carbonation; a mineral pre- treatment method that maximises mineral activation for both olivine and serpentine in a single process or multiple process streams must be found that can subsequently minimise parasitic energy loss.

2.4.2 Aqueous solution chemistry and experimental carbonation conditions Initial olivine carbonation tests conducted by O’Connor et al. (1999) produced a baseline set of carbonation results. In these experiments, no mineral pre-treatment, other than size reduction and the addition of distilled water to form a slurry medium was undertaken. These experiments resulted in the

91 % stoichiometric conversion of silicate to carbonate in 24 hours under a temperature of 185˚C and a

PCO2 of 115 atm. In following work, O’Connor et al. (2000; 2001) made modifications to the carbonation solution chemistry. This included the use of a bicarbonate/salt mixture over distilled water. This resulted in a significant improvement of reaction times. The addition of NaCl provides a complexing ion (Cl-) that combines with the Mg2+ cations forming several intermediate magnesium chloride compounds. The complexing ions reduce Mg2+ cation activity in solution. This results in an increase in the solubility of the magnesium silicate. NaHCO3 is thought to + modify the solution to a slightly alkaline pH while acting as a CO2 carrier. The additional Na ions modify the surface charge of the silicate particles, enhancing ion exchange at the solid/liquid interface (O’Connor et al., 2005: pg.10). Reduced cation activity and increased speciation of magnesium through the use of a buffered solution signifies its importance in development of mineral carbonation technology. Using a buffered solution in place of distilled water results in the carbonation of naturally occurring olivine to approximately 84 % of stoichiometric conversion in 6 hours at 185°C and a PCO2 of 115 atm (O’Connor et al., 2000b: pg. 5). The experiments used a buffered solution with a bicarbonate concentration of 0.64 M. This is the maximum solubility for NaHCO3 in 1 M NaCl solution at 20°C (O’Connor et al., 2005). This solubility simulates the solution concentration that could meet the requirements for an industrial process using a solution recycling regime (O’Connor et al., 2005). It became the standard buffered solution used in the experiments at ARC (O’Connor et al., 2000). This standard buffered solution is used as the slurry solution for experimental direct aqueous mineral carbonation experiments in this research. In subsequent experiments using the Twin Sisters olivine, the maximum extent of reaction occurred at 185°C using the standard buffered solution. Two factors work against raising the temperature. The

CO2 solubility decreases and the reaction become thermodynamically less favourable. This is due to the region of magnesite stability being exceeded (Gerdemann et al., 2003). Raising the partial pressure of

CO2 extends the magnesite stability field due to the increase in CO2 activity. However the increase in pressure also pushes the reaction toward completion, due to the volume change caused by gas consumption as CO2 is converted to solid MgCO3 (Penner et al., 2004). From these experiments standard reaction conditions of 185°C at a PCO2 of 115 atm in a buffered solution of 0.64 M NaHCO3, 1 M NaCl were proposed at ARC for activated olivine substrate material. From 2004 to 2007 Chizmeshya and co-workers at ASU focused on the effects of aqueous solution chemistry and particle size on mineral carbonation (Chizmeshya et al., 2005; Chizmeshya et al., 2006;

Chizmeshya et al., 2007). The extent of carbonation was determined as a strong function of both alkali cation species and their activation. By increasing sodium and potassium bicarbonate concentration

+ [mHCO3] the extent of carbonation increased in comparison to the standard ARC solution. Na was + indicated to be more effective than K , with the extent of carbonation greater for 2.5 M NaHCO3 than

2.5 M KHCO3. Both the extent of carbonation using the ASU recommended 2.5 M NaHCO3 and the ARC recommended 0.64 M NaHCO3 was investigated during experimental mineral carbonation in this research. The results of this experimentation are discussed in Chapter 4. From experiments undertaken by Chizmeshya and co-workers (Chizmeshya et al., 2005; Chizmeshya et al., 2006; Chizmeshya et al., 2007), a 35 %, 146 %, and 158 % improvement in carbonation extent was seen for <38 µm, <75 µm, and <150 µm fractions respectively when using 2.5 M NaHCO3 in the buffered solution. The consequence of these results is being able to effectively carbonate olivine feedstock with larger particle-size fractions. This can aid in reducing the processing costs and energy need of mineral pre-treatment. However, the financial cost, environmental impact and efficiency of a potential recycle process for the more concentrated buffering reagent is yet to be evaluated. The use of organic additives remains the focus to enhance carbonation efficiency when using olivine as a substrate for carbonation (Bonfils et al., 2012). The potential of particle abrasion to improve the carbonation extent was also investigated. With the addition of 20 % quartz particles to the substrate material, a 5 % increase in the extent of carbonation was realised (Chizmeshya et al., 2006). This quartz particle addition could be applied to an industrial carbonation process due to the availability of quartz as a carbonation product. The effect of quartz particles on enhancing the extent of mineral carbonation is attributed to increased abrasion of a robust silica-rich passivating layer that forms on substrate particles. This passivating layer has been attributed to incongruent magnesium dissolution and silica precipitation (Chizmeshya et al., 2007). Alteration of the primary magnesium silicate, produces silica enriched, magnesium depleted intermediate products. While dissolution is likely surface controlled, the formation of silica-enriched zones suggests the reaction is diffusion limited. These zones appear to and inhibit further dissolution (O’Connor et al., 2001). The series of reactions involved in the direct aqueous mineral carbonation process are complex due to their multiphase nature. An un-reacted core model for particles of either shrinking or unchanging size has been suggested for the mineral carbonation process (Guthrie et al., 2001). The major distinction here is that the solid product produced provides additional resistance to the carbonation process (Ityokumbal et al., 2001). These findings triggered the research in to the separation of the dissolution and precipitation stages of mineral carbonation. This was targeted to increase the stoichiometric

31 conversion of silicate to carbonates. It is still unknown whether this will be a feasible in an industrial application of mineral carbonation. Detailed investigations by a number of research organisations continue to further the understanding of the dissolution kinetics of both olivine and serpentine. (Krevor and Lackner, 2011; Werner et al., 2011; Sanna et al., 2013). Such research is essential to the development of an industrially scalable mineral carbonation process pathway

2.5 Estimating the mineral carbonation capacity of ultramafic rocks

To date the parameter widely used to estimate the mineral carbonation potential of an ultramafic deposit is RCO2. This measure was first proposed by Lackner et al. (1995) and used by Goff and Lackner (1998) and Goff et al. (2000) to estimate the carbonation potential of the major ultramafic complexes in

North America. RCO2 is defined as the mass of ore necessary to convert a unit mass of CO2 to a carbonate. By this definition, a low RCO2 is preferable to a high RCO2. The initial measure was defined by the molar concentration of magnesium and calcium (Lackner et al. 1995; Goff and Lackner, 1998; Goff et al., 2000), but was later modified to include ferric iron in the calculation (Equation 2.8) (Penner et al., 2004). This modification was made due iron potentially being able to form stable carbonates. This permitted the calculation of the carbonation potential for alternative feedstock such as fly-ash and steel slag (Penner et al., 2004).

100 푀표푑푖푓푖푒푑 푅퐶푂2 = 2+ 2+ 2+ Eq. 2.8 (∑퐶푎 + 퐹푒 + 푀푔 )푀푊퐶푂2

Where; ΣCa2+ + Fe2+ + Mg2+ are the sum of the molar concentrations for the specified cations and

MWCO2 is the molecular weight of CO2. Theoretical minimum RCO2 values have been calculated as 2.1, 1.6, and 2.7 for serpentine, olivine and wollastonite respectively (Goff et al., 2000; Lackner et al., 2008). When considering typical minerals in ultramafic deposits, ferric iron can be included to consider the carbonation potential of fayalite (FeSi2O4). Although theoretically possible, there has been no evidence provided in the literature that has shown the successful stoichiometric conversion of fayalite to iron carbonate minerals. Considering the only marginal stability of iron carbonates, it should not be included in the carbonation potential estimation of ultramafic deposits based on the mineral carbonation methods proposed thus far.

RCO2 is a parameter capable of determining effective initial estimates of the capabilities of a resource or deposit to sequester CO2. This is achieved by applying the RCO2 ratio to the estimated volume of the

32 rock investigated (Table 2.4). It is improbable though, that all the cations will available for carbonation without the use of aggressive catalyzing solutions. This assumption is based on the stoichiometric forsterite to magnesite conversion extents achieved in this research and the literature summarised earlier in this chapter. Thus, when evaluating the potential of deposits for use in mineral carbonation, an overestimation of the carbonation extent is expected when using a more easily recoverable solution in an industrial setting (i.e. NaCl/NaHCO3) for direct aqueous mineral carbonation. Dunite zones of ultramafic complexes are preferred for use in mineral carbonation. Dunite rocks often contain the greatest magnesium by weight usually contained within olivine. Serpentinite zones are more common globally and so must also be investigated and continue to do so at numerous academic institutions. Large unaltered ultramafic deposits such as the Twin Sisters dunite in Washington State, USA typically have a homogenous mineralogy. This is comprised of forsteritic olivine (>90 %-vol) with little to no alteration throughout the deposit. The RCO2 of these deposits and the degree of stoichiometric forsterite to magnesite conversion achievable under a chosen mineral carbonation method may be accurately determined. Ultramafic complexes that host mineral deposits of economic interest such as a nickel-sulphide, kimberlitic diamond and PGE deposits have often undergone more extensive alteration. Serpentine alteration of dunite for example can be wide spreading, up to complete serpentinisation (Scheel et al.,

2004). As indicated by Goff et al. (2000) serpentine and olivine have differing RCO2 values of 2.1, and 1.6. The difference in these values plus the amount of magnesite already present in the rocks can make estimating the carbonation potential of these deposits challenging. The effect of mineral variability on the estimation of mineral carbonation potential has not been addressed in the literature. Initial studies have focused on large unaltered ultramafic deposits such as the Twin Sisters dunite. When considering a much smaller ‘niche’ operation in which mining waste rock from an active or potential mining operation could be used to offset its own CO2 emissions or additionally the emissions of a nearby large CO2 producer, a more detailed analysis of the mineral carbonation potential is required. This is due to the mineral complexity of the rocks considered. To determine the carbonation potential of mine waste rock, carbonation controlling parameters should be defined. These include the substrate mineralogy, carbonation method, and the carbonation process conditions for that particular method. An additional benefit of such operation is the ability to dispose the stable carbonated product in the excavated mine or established tailings facility. Exploratory drilling at active and prospective mines are geologically logged and geochemically assayed to spatially determine the character of the rock mass. The spatial definition of individual

33 lithologies allow a three dimensional model to be generated of each lithology present. Whole rock geochemical data created from drill core assaying allows the elemental quantities to be determined for discrete lengths of drill core. It was postulated at the onset of this research that these values can be spatially interpolated within the lithological models using estimation techniques such as inverse distance or ordinary kriging. Chapters 5 and 6 of this dissertation develop a procedure in which whole rock geochemical data can be used to estimate mineral quantities. The procedure additionally quantifies the amount of CO2 that can be sequestered by those minerals for each rock type. Identifying the principal mineral source of MgO can be used to better quantify the availability of Mg2+ for mineral carbonation. This in turn is used to spatially predict the mineral carbonation potential within each three dimensional lithological model. The stoichiometric silicate to carbonate conversion achievable by the chosen method is applied to each mineral. The mineral carbonation capability of each lithology is calculated accounting for mineral variability across a deposit.

2.5.1 Mineral carbonation controlling factors on resource estimation Numerous factors must be considered when predicting the potential of ultramafic rocks to sequester CO2 via mineral carbonation. First, the geochemical and mineralogical composition of the rocks must be understood. The chemistry or more specifically the quantity of available Mg2+ cations is the principal factor in determining the amount of magnesite that can be formed. Once the magnesium concentration of the rock is established, the mineralogical location of that magnesium must be defined. This permits the potential to extract Mg2+ from its mineral association for magnesite formation to be estimated. Coincidentally the success of the selected mineral carbonation process and associated mineral preparation should be evaluated through laboratory or pilot-scale experimentation. The relationship of these factors can be correlated to calculate the greatest achievable mineral carbonation extent for any given rock type.

Geochemical and mineralogical factors Fresh, unaltered dunite bodies such as the Twin Sisters deposit are relatively uncommon. Partially serpentinised dunite and other peridotite deposits are much more widespread. They are often greater in mass and may comprise of 20-80 % serpentinisation (Goff and Lackner, 1998). The geology of some of these serpentinised ultramafic bodies has been studied extensively. These include; the Belvidere Mountain Prospect, Vermont USA (Labotka and Albee, 1979; Goff and Lackner, 1998), and the Canyon

Mountain and Vulcan Peak deposits, Oregon, USA (Himmelburg and Loney, 1973; Thayer, 1977; Goff and Lackner, 1998). Serpentinite and serpentinised dunite/peridotite largely comprise of a heterogenous mineralogy. These minerals include; olivine, pyroxene, serpentine, enstatite, magnetite, residual chromite, brucite, carbonates and free silica (Goff and Lackner, 1998). Serpentinisation is inherently associated with a volume increase (up to 53 wt.%) and lower densities (~2.5 g/cm3) in comparison to the original peridotite (~3.3. g/cm3) (Goff and Lackner, 1998: pg. 95). This mineral variability and changing geochemistry with increased serpentinisation from unaltered peridotite/dunite has a significant impact on the RCO2 values of the differing rock types as reported by Lackner et al., (1997: pg. 4) in Table 2.4.

Table 2.4 The magnesium and calcium oxide content of common rocks considered for mineral carbonation.

Rock type MgO (wt.%) CaO (wt.%) Rc RCO2 Peridotites - Dunite 49.5 0.3 6.8 1.8 - Harzburgite 45.4 0.7 7.3 2.0 - Lherzolite 28.1 7.3 10.1 2.7 Serpentinite ~40.0 ~0.0 ~8.4 ~2.3 Gabbro ~10.0 13.0 ~17.0 ~4.7 Continental basalt 6.2 9.4 26.0 7.1

Rc and RCO2 represent the mass ratio of rock to sequester the mass of carbon or carbon dioxide respectively (Source: modified after Lackner et al., 1997: pg. 4).

Table 2.4 highlights that the preferred substrate for a mineral carbonation operation would be dunite. Harzburgite then serpentinite are also promising based singularly on their mineralogy. Goff and

Lackner, 1998, and Goff et al. 2000 determined the theoretical RCO2 for some of the largest ultramafic deposits in the USA. Table 2.5 compares the values determined by Goff and Lackner with the mineral carbonation potential of the Turnagain ultramafic complex, British Columbia, Canada using the RCO2 parameter. This assessment on the mineral carbonation capacity of ultramafic deposits is based on average magnesium content and the surface mapped geology defined extents of the deposits. The preliminary assessment of the Turnagain ultramafic deposit in Table 2.5 can be used as an analogue of a deposit that could host a mineral carbonation operation alongside a more traditional commodity based mining operation. Based on the relative size of the deposit, at an estimated 10.8 km3 volume, and using the mean magnesium content of the rocks sampled from extensive drilling at the deposit, the overall potential to sequester CO2 by mineral carbonation is vast. An RCO2 of 2.41 calculated

9 at Turnagain which could permit a theoretically possible sequestration of 13.22 x 10 tonnes of CO2. This is approximately 19.1 years of Canada’s CO2 emissions based on the 2012 emission rate (Environment Canada, 2012: pg. 4). It must be emphasised that estimate is purely based on the potential amount of

CO2 sequestered and not the quantity of CO2 that would be avoided once sequestration efficiency and energy expenditure is considered.

Table 2.5 CO2-sequestration properties of selected ultramafic complexes in North America

Twin Sisters, Vulcan Peak, Del Puerto, Belvidere Wilbur Baltimore San Turnagain, Washington Oregon California Mountain, Springs, Complex, Mateo, British Vermont California Maryland California Columbia Volume-Density Area (km2) 91 16 40 2.3 200 100 4? 24 Depth (km2) 0.6 0.5 0.3 ≤1 ≥0.2 0.3 0.25 0.45 Est.Vol. (km3) 54 8 12 2 40 30 1 10.8 Density (g/cm3) 3.3 3.2 2.8 2.9 2.65 2.7 2.6 2.95 Wt.% Mg Peridotite 29.9 27.4 27.2 29.1 _ _ _ 26.5 Serpentinite _a _ 20.9 23.1 21.8 21.2 19.9 20.7 Combined 29 27 23 26 21 21 20 23 Sequestering Properties b RCO2 1.91 2.05 2.41 2.13 2.41 2.48 2.64 2.41 Mg (109 tons) 51.68 6.91 7.73 1.51 22.26 17.01 0.52 7.33 9 CO2 (10 tons) 93.30 12.49 13.94 2.72 43.98 32.66 0.98 13.22 USA (yr)c 17.7 2.4 2.6 0.5 8.3 6.2 0.2 2.5 Canada (yr)d 134.7 18.0 20.1 3.9 63.6 47.2 1.4 19.1 World (yr)e 3.09 0.41 0.46 0.09 1.46 1.08 0.03 0.44

The physical, chemical and CO2-sequestration properties by mineral carbonation of numerous peridotite and serpentinite bodies throughout the United States (Source: modified after Goff and Lackner, 1998: pg. 98 and references therein) and the Turnagain ultramafic deposit, British Columbia, Canada. a (_) denotes that no significant occurrence has been reported b RCO2 is the calculated mass of rock processed to CO2 disposed 9 c Determined assuming a 5.28 x 10 tonnes/yr CO2 emission rate - U.S. Energy Information Administration (Source: Monthly Energy Review April 2013) 9 d Determined assuming a 0.69 x 10 tonnes/yr CO2 emission rate - Environment Canada, August 2012 (Source: Canada’s Emissions Trends 2012) 9 e Determined assuming a 30.2 x 10 tonnes/yr CO2 emission rate - U.S. Energy Information Administration (Source: International Energy Outlook 2011)

From this initial assessment, it is clear that the Turnagain ultramafic deposit is an attractive mineral carbonation target for possible exploitation. It also signifies the effectiveness of RCO2 as a parameter to quickly define deposits for a more focused feasibility study. From a geological approach, to further evaluate such deposits, both surface and sub-surface geological mapping must be undertaken. This mapping permits a more accurate volume calculation and the definition of geological domains of similar rock characteristics. Mineral carbonation capabilities assigned to these discrete rock units aids in the generation of a more reliable resource estimate. Domain definition will inherently reduce the overall mineral carbonation potential of the deposit. This is caused by accounting for rock units within the complex where rocks with limited or no mineral carbonation potential are found. Domains or lithological boundaries permit individual RCO2 values to be applied to discrete volumes of rock. The RCO2 can be calculated for each rock type based on the principal magnesium silicate within that rock type. Furthermore, the theoretical CO2 sink for each lithological unit present within a deposit by mineral carbonation can be calculated. Investigation into the specific rock units is still required to achieve an accurate MCP determination. Although these rock units can be characterised by geologists in the field, the specific mineral quantities cannot be accurately determined qualitatively. It is thus difficult to calculate the mineral variability without geochemical and mineralogical laboratory analyses. An example of this problem is posed, by which a rock unit may be mapped as a serpentinised dunite at Turnagain. What proportion of the minerals within that rock mass are olivine and what proportion of the rock are serpentine minerals?

Given the absolute RCO2 of forsterite being 1.6 and serpentine being 2.1 (Lackner et al., 2008: pg.10) significant differences in the potential CO2 capacity could be calculated. An example of the importance of mineralogical determination is proposed in the following scenario:

a) A 1,000 tonne dunite rock unit in which 90 % of the mineral mass is forsterite with an average MgO concentration of 49.5 % (Lackner et al., 1997: pg. 4) can theoretically sequester 471 tonnes

of CO2 assuming full stoichiometric conversion of forsterite to magnesite. b) A 1,000 tonne serpentinite rock unit in which 90 % of the mineral mass is lizardite with an average MgO concentration of 40 % (Lackner et al., 1997: pg. 4) will theoretically sequester 414

tonnes of CO2 assuming full stoichiometric conversion of lizardite to magnesite.

The scenario above shows the extremes between a relatively unaltered dunite to serpentinite. It supports the focused targeting of dunite over serpentine in this example based on the quantity of rock

38 that must be excavated and subsequently processed to sequester the same quantity of CO2. This has both financial and energy consequences that could significantly jeopardise the feasibility of developing a mineral carbonation resource if the carbonation efficiency of both olivine and serpentine were equal. It also signifies the importance of determining the internal mineral variability of rock units. Here in lies the challenge of estimating the quantity and ratio of magnesium silicate minerals used to fix the analysed MgO concentration within the rock mass. That is; how much of the MgO is contained within forsterite, how much of the MgO is contained within serpentine minerals. Additionally, how much of the MgO is contained within other minerals such as diopside or magnesite. A final question posed is; how does this affect the potential for that rock mass to sequester CO2 by mineral carbonation?

Mineral carbonation process methodology Numerous mineral carbonation methods have been tested at a laboratory scale. These have demonstrated variable stoichiometric magnesium silicate to carbonate conversion capabilities. The relative infancy of mineral carbonation technology makes selecting a carbonation pathway and envisioning a practical process stream difficult. This incorporates extracting substrate rock through to the deposition of magnesium carbonate product material. Determining the magnesium silicate to carbonate extent for any given mass of material (mineral or rock) is dependent on the development of mineral carbonation technology and process that is successful on a pilot plant scale. By completing laboratory scale experiments on rock samples from a deposit, the stoichiometric magnesium silicate to magnesium carbonate conversion extent achievable by a given mineral carbonation process can be established. This value can applied as a fixed factor in estimating the theoretical mineral carbonation extent and thus the mass of CO2 that can be sequestered by any given mass of rock within that deposit. As an hypothetical example; if 1,000 tonnes of waste rock from an open pit mine is comprised of 90 % forsterite with an average MgO concentration of 49.5 wt.%, and the chosen mineral carbonation pathway is capable of achieving a 75 % carbonation extent from a forsterite substrate, those 1,000 tonnes of rock are theoretically capable of sequestering 353 tonnes of CO2. Any mineral carbonation parameter created must be flexible to adaptations in the processing method used. The application of the carbonation efficiency can be applied akin to the recovery rate achieved from a processing facility at a mine site. The cation geochemistry of drill core samples can be treated the same way as a mining commodity grade. Likewise the mineralogical location of the cations allows a suitable processing method to be used to extract the cations. By estimating the qualities of the

39 feed material to a mineral carbonation processing facility, a suitable processing stream can be designed. This should be designed to maximize the mineral carbonation potential of the feed material. Consideration should be given to run simultaneous process methodologies at one mineral carbonation facility. An example of this could be where an identified rock mass that has serpentine as the dominant mineral could be sent through one process stream, whilst another identified rock mass with a forsterite dominant mineralogy could be sent through another stream. This would maximise the overall MCP. It signifies the importance of testing multiple mineral carbonation process routes at a laboratory and pilot-scale.

Carbonation efficiency and extent determination The data used for mineral carbonation extent determination can be derived from a number of laboratory sources. A geochemical/mineralogical analysis must be completed on the substrate material. Laboratory scale mineral carbonation testing must be undertaken. A geochemical/mineralogical analysis on the product material must finally be undertaken. The techniques and equipment used for all three of these phases of testing are dependent on whether a mineralogical or geochemical approach is taken. The quantity of material available for testing, the mineral carbonation process selected and the financial budget and/or experience of personnel undertaking the testing are all contributing factors. Rocks with variable silicate mineralogy can significantly impact the MCP of a deposit. By using the abundant geochemical data present from drill hole sampling at mining sites it is possible to use a geochemical mass balance approach. Predicting MCP values takes the measured magnesium concentration often by ICP-ES analysis completed on the tested samples. The effectiveness of the mineral carbonation process chosen at trapping CO2 based on the magnesium concentration can then be applied to this magnesium concentration. This application has some limitations as it assumes a consistent mineralogical host for magnesium. When undertaking experimental mineral carbonation, it is expected that large variations in the mineral carbonation extent will be seen. This is caused by the varying effectiveness of the mineral carbonation process to trap CO2 with different magnesium silicate minerals. In this research a mineralogical mass balance approach is preferred. This considers the inhomogeneity of ultramafic complexes exploited by the mining industry. A number of analytical methods can be used to determine the mineralogy of rock samples. The benefits and limitations of the quantitative XRD phase analysis using the Rietveld refinement technique are discussed in Chapter 3. Other methods are available and were analysed for use in mineral carbonation determination by Wilson

(2009). She compared successful methods of quantifying trace abundances of minerals. These include Chung’s (1974) method of normalised reference intensity ratios (RIR) and calibration curves according to the internal standard method (Alexander and Klug, 1948; Sanchez and Gunter, 2006). Wilson (2009) concluded that it is difficult to recommend the use of one method over another for the quantification of low abundances of minerals. Wilson (2009) did yet show the successful determination of minor abundances of hydrated magnesium carbonate materials in mine tailings using XRD with Rietveld refinement. That method was selected to quantify the mineralogy of Turnagain ultramafic complex rocks in this research. This is based on the experience of the laboratory staff at the Earth and Ocean Science department at the University of British Columbia Without the apparatus and necessary expertise to undertake quantitative mineral analyses, the cost of sample analyses can be expensive (i.e. between $100 and $300 per sample). It is unlikely that it will be practical or economically feasible to directly quantify the mineralogy before and after mineral carbonation experimentation. This is particularly true when using drill core samples on hundreds to thousands of drill holes that can be present at mining operations. If a mineralogical mass balance approach is chosen, a technique that can take available geochemical data and best estimate the mineralogy is preferential. Minimal direct quantitative mineralogical analysis would still be necessary to verify the effectiveness of the mineral estimation as a quality assurance. The overall cost and time for experimentation would yet be reduced. In this 16 samples were tested at Turnagain. Based on this research the XRD with Rietveld refinement analysis of both the reactant and post-carbonation samples was deemed enough to determine the mineral carbonation capabilities of dunite at Turnagain. This is based on the level of detail required to adequately generate a resource estimate at this stage of mineral carbonation as an industrial process.

2.5.2 Estimating mineralogy from lithogeochemical data As stated at the onset of this dissertation, a principal outcome of this research was to develop an MCP parameter in which abundant lithogeochemical data can be utilised at a mining operation to better estimate the mineral carbonation potential of ultramafic rocks. Estimating the modal mineral abundance of the rocks has emerged as a key component of achieving this goal. This section reviews existing methods of estimating mineralogy from lithogeochemical data. X-ray diffraction has been widely used for decades to identify the major mineral phases in rock samples. Sample preparation, expense, available resources for analysis and technical skill requirements for accurate results are an inhibiting factor when undertaking detailed quantification of mineral abundances in rocks by XRD with Rietveld

41 refinement. XRD with Rietveld refinement was used for the quantification of the pre- and post- carbonation mineral quantities in this research and the advantages and limitations of the method is discussed in detail throughout the dissertation. Point counting has been used since the 1930s to determine bulk mineralogical compositions using a regular grid. Point counting requires the preparation of thin sections through rock or drill core samples. A significant number of points are required to be counted. Increasing the number of points counted has been shown to decrease the estimation error i.e. approximately >6 % for less than 250 points to approximately 1.5 % counting 1000 points (van der Plas and Tobi, 1965). Although more affordable than XRD with Rietveld refinement, point-counting is time consuming when examining a large number of samples. Numerous techniques have been created, computer programs developed and adaptations of both published to estimate modal mineral abundances of rocks from elemental and oxide data. The techniques discussed in this research focus on those developed to determine the modal abundances in igneous rocks. The first attempt to quantify mineral abundance was made by Cross, Idding, Pirrson and Washington (CIPW). This generates an idealised mineral composition from calculated major oxide values and is known as the CIPW norm (Cross et al., 1932). The CIPW norm calculates the mineral composition by treating the parent magma as anhydrous at low-pressure. It sequentially allocates oxide values to the simplest minerals. The CIPW norm is effective in synthesising a mineral assemblage based on the initial magma crystallising to create the igneous rocks. But it is not possible to estimate the composition of more hydrous minerals and the degree of alteration such as serpentisation. As such a standalone CIPW norm calculation will not suffice in generating an MCP parameter. The reliance on a structured subtraction of element oxides to form a specific suite of minerals and the inflexibility of CIPW norms to deal with non-igneous rocks that have minerals of variable compositions such as clays triggered De Caritat et al. (1994) to devise a normative method named LPNORM. This uses linear optimisation in generating a best numerical solution through a series of linear equations. LPNORM takes quantified elemental oxide values from XRF and defined mineral compositions from XRD. It calculates the mineral quantities based on the stoichiometric conversion of all the present elemental oxides into the defined mineral assemblage through a series of equations. Concurrent to LPNORM, MODUSCALC was developed by Laube et al. (1996). This offered an alternative to the linear equation systems by generating a best-fit approximation of the mineralogy. MODUSCALC uses the Gauss-Jordan method of matrix inversion. Their system is limited to the number of minerals that can be selected to achieve the best approximation through the use of a Euclidean norm

42 on the vector system. Its use of oxides requires extra information. These include analyses of CO2 and

H2O content when recalculating a theoretical loss-on-ignition value combining geochemical and statistical data within the program. Statistical information on mineral occurrences through XRD or point counting is required for effective use of the program. MODUSCALC has shown to be a useful tool when analysing a large number of comparable samples. It also permits the study of variable mineral compositions caused by metamorphism or diagenesis (Laube et al., 1996). Notable further attempts at creating computer program solutions for estimating mineralogy include; petrological mixing computer methods through linear programming and least-square analysis (Bryan et al., 1969; Wright and Doherty, 1970; Albarede and Provost, 1977; Banks, 1979; GENMIX by Le Maitre, 1981; and LSEQIEQ by Ghiorso, 1983). Estimating modal mineralogy through linear programming and least square analysis (MONA by Metzner and Grimmeisen, 1990; SEDNORM by Cohen and Ward, 1991; MODAN by Paktunc, 1998, 2001; Gordon and Dipple, 1999; MINSQ by Herrmann and Berry, 2002; and MINLITH by Rosen et al., 2004). A common problem encountered with many of the programs is the production of erroneous results when there are minerals present that are characterised by similar variables. Such minerals include those that display a solid solution series within the mineral suite and the presence of both magnetite and hematite. All the programs are capable of converting weight percentage oxide values into an estimated modal mineralogical rock composition. This is achieved via a fixed subtractive approach first defined by the CIPW principle or generating a mineral suite from a best-fit numerical model via a linear or matrix solution. All the programs have demonstrated a tool that can estimate modal mineralogical rock composition for a particular suite of rocks or a particular deposit. Yet, the rigidity of the CIPW principle methods and the increasing information required for the best-fit numerical models have not elevated a sole leading program for modal mineral estimation. Specific challenges need to be addressed when attempting to estimate the modal mineral composition of an ultramafic suite of rocks. Major oxide data analysis of the rock mass at ultramafic complexes hosting mineral deposits of economic interest is often minimal, if existent. Whole-rock geochemical analysis of exploration drilling samples undertaken is commonly restricted to methods such as ICP-ES. This is often due to financial constraints, the need for targeted elements and the volume of samples needed to be analysed. These whole-rock analyses provide a suite of elemental weight abundances. They do not contain any silica (SiO2) quantity and LOI (H2O + CO2) values.

When attempting to predict the mineral carbonation potential of an ultramafic rock mass, at least the modal abundance of olivine and serpentine must be estimated. The modal mineralogy estimation programs commented on in the previous section require a suite of common oxide values for estimation.

These include a measured SiO2 content, typically obtained from XRF analysis. When these oxide values are available it is possible to derive the theoretical olivine content of the rock mass from many of these programs. If further analyses such as H2O and CO2 abundances are available an estimate of the hydrated-silicates such as serpentine and carbonates such as magnesite can often be estimated.

2.6 Resource estimation in the mining industry

Mineral abundance and MCP data generated over regular intervals of drill core, synonymous to metal grade values, provide data that can be estimated within a three-dimensional (3D) geological model. This estimation represents a hypothesis about the geometry, extent and character of the resource (Vann, 2005). Mine planning and scheduling is based on using a geological model for the deposit. From this blocks are typically created on an orthogonal grid superimposed on the deposit and divided in to mineable volumes. Each of these blocks is assigned a grade value (Peroni et al., 2003). MCP values applied to blocks in this manner allow the definition of a mineral carbonation potential resource model. With increasing computing power and emphasis on business risk analysis, geoscientists increasingly incorporate new factors into spatial models. As well as the grade of key economic elements, spatial estimates for deleterious elements and factors such as metallurgical recovery are generated. The intention of modelling these parameters is to permit an improved analysis of the value of the in-the- ground resource (Dunham and Vann, 2007). The modelling of MCP values can be viewed in this same light. The evaluation of previously considered waste as a substrate for mineral carbonation will aid the better understanding of a potential value-adding resource. Geological models are based on the determination of ‘domains’. These domains are geometric volumes defined by individual geological characteristics. These domains statistically assure a reasonable degree stationarity (or statistical homogeneity) (Vann, 2005). Modelling should be undertaken on a lithology-by-lithology basis or the combined modelling of a collection of lithologies considered to have a comparable geological character. One aspect of this character could be a rock type’s mineral carbonation potential.

A common concern when undertaking the estimation step involves the estimation of grades at a scale of (at least) selective mining units. Thus, a block support is estimated for samples on a much smaller support, usually several orders of magnitude smaller in volume (i.e. drill core) (Vann, 2005). Variables with smaller support are more dispersed (higher variance) than variables on larger support. This influence of support on variability is known as the volume variance or support effect (Journal and Huijbregts, 1978; De-Vitry et al., 2007). The regularisation of support through compositing is a method commonly used in the mining industry to improve the resolution of block support. This could be applied to other regionalised variables other than commodity grade such as an MCP variable. When working with sample data from intervals of drill core it is expected that these values will have more extreme grades and thus a higher variance than mining blocks. However, as Dunham and Vann (2007) posit; this may not necessarily hold true for some geochemical variables. A number of questions are considered and analysed in Chapter 6: what scale does an experimental mineral carbonation test value represent? Does it represent the volume of the original sample or even a molecule-by-molecule measurement? Can the MCP value of a 10 g sample be truly representative of a 10 m3 mining block? It is widely recognised that there is no one-size-fits-all solution to creating spatial estimates and maps of geochemical attributes. Each operation and each material type is unique and needs to be considered uniquely (Isaaks and Srivistava, 1989). Careful consideration must be given to the design sampling and implementation of any geospatial estimate. An achievable goal for mining companies using the proposed MCP estimation framework in Chapters 5 and 6 are to populate ‘mineable’ blocks with and interpolated mineral carbonation potential specific values. In this framework, traditionally interpolated metal grade values can be accompanied MCP values. These values will be composited over fixed intervals of drill core for statistical validation and providing continuous regionalised support within the selective mining units (SMU) throughout the defined geological domains. A statistical analysis can then be undertaken to estimate the MCP capabilities of waste rock into discrete volumes or SMUs within these domains. Through the careful consideration of geological and mineralogical controls on MCP values and exploratory data analysis, a study into the MCP estimation can be developed. This can use a range of estimation techniques coupled with an analysis of the degree of estimation error. The research aims to outline a workflow through a case study on the Turnagain ultramafic deposit and a starting point from which the characterisation of ultramafic deposits in respect to their mineral carbonation potential can be approached and developed.

2.6.1 Traditional empirical approaches to resource estimation Selecting an estimation method that should be used to interpolate MCP values is difficult. This is largely due to the lack of empirical data from the practical mining and reconciliation of estimates from comparable deposits. Testing geostatistical and empirical estimation techniques in this research can only be considered an introduction and an experimental approach to modelling MCP values. The validity of using one method over another requires a broader series of investigations on many ultramafic deposits than can be created from a single case study of the Turnagain ultramafic deposit. It is also important to consider the level of detail required for such analysis at this stage of MCP resource estimation. It must be stated that the research presented in this dissertation aims to couple geological and mineralogical continuity with three-dimensional MCP value interpolation to provide a baseline MCP estimate. It does not provide the assurance required for reporting legal global resource estimates or detailed mine planning. The selection of an applicable estimation method must be made based on the experience of the user and an understanding of the estimation process employed. A number of estimation methods have been used by the mining sector for decades. Empirical methods such as sectional, polygonal and triangular methods have been successfully applied to create reliable resource estimates. These methods are covered in detail by Patterson, 1959; King et al., 1982; Annels, 1991; Stone and Dunn, 1994; and Sinclair and Blackwell, 2002. The cross-sectional method typically uses a geological interpretation made on a series of cross- sections of drill holes. The principal benefit of using a cross-sectional method for resource estimation is that it uses existing geological interpretations permitting a strong geologic control on the estimation. However, it assumes a smooth transition of grades that are projected to large volumes that tends to lead to an overestimation of grade values. Polygonal estimation methods assign a raw data value as the mean value of a larger polygon or prism with no smoothing of the raw data. The use of raw values can create a conditional bias in global estimation dependent on the estimated value. This tends to overestimate the value of high-value polygons and underestimate the value of low-value polygons. The triangular method is a variation of the polygonal method in that it has a more conservative approach to assigning values to larger areas. The principle advantage of the triangular method over the polygonal method is that it smooth’s data by averaging the values at the three apices of each triangle. Limitations that restrict the application of the triangular method include an entirely empirical smoothing

46 process, a restriction to equal weighting in all directions and the development of an irregular block array (Sinclair and Blackwell, 2002). The ease of use of the methods described and the non-reliance of significant computing power make them viable initial estimation tools. These could be considered for geospatially estimating MCP values at a potential mining operation. The familiarity of inverse distance weighting (IDW) estimation and more sophisticated geostatistical estimation methods of many professional geoscientists can promote the use of these methods over the empirical methods described. IDW differs from those described in that it is used to estimate values in to a three-dimensional regularised array of blocks. This is often described as a block model at a mining operation or a deposit at the exploration stage. Each block within the array is estimated into from a series of data points. The influence of each data point on the estimated block is weighted dependent on distance. The IDW estimation process is commonly used in the mining industry due to the relative ease of use when compared to geostatistical methods. It applies uniform weighting in all directions and does not incorporate the significance of determining the degree of error associated with a single value. As a result, reconciliation of measured grades is required to determine the effectiveness of the estimation method. At this time this remains unavailable when considering reconciling MCP values at a mining operation. IDW estimates have proven to produce results comparable to those estimated by ordinary kriging (Sinclair and Blackwell, 2002). Coupled with its ease of use, speed of estimation, familiarity with many mining professionals, the estimation of MCP values using and IDW estimation method is favourable. This method is compared in Chapter 6 with MCP estimates achieved by geostatistical techniques. It can be considered as an initial attempt at the investigation of using IDW as a method that can be employed by a mining company to estimate the MCP of an ultramafic deposit.

2.6.2 Geostatistical approaches to resource estimation The theoretical base of geostatistics is well established. It has been continually advanced and endorsed by mining industry professionals for over 40 years (Armstrong and Champigny, 1989). The term geostatistics was first coined by Matheron (1963) who developed the fundamental basis for geostatistics in the mining industry. MCP values are quantities or regionalised variables that are characterised by their three dimensional distribution within the rock mass. The theory of geostatistics has some advantages over other estimation methods as it minimises the estimation variance quantitatively and considers the

47 autocorrelation of the variable through the estimation procedure. Geostatistics is developed on the basis that regionalised variables 푧 (푥) have a particular structure and a random function 푍 (푥). The two components (random and structured) can be quantified by an autocorrelation function such as a semi- variogram. The variogram or semi-variogram (half the variogram) is a fundamental tool used in geostatistics. If geostatistical estimation methods are used in estimating MCP values at ultramafic deposits, variography must be undertaken. Variography is an essential component to the kriging estimation process. Variograms can be created in multiple directions over broad sample search ranges to directionally quantify the structure of the values and incorporate them into the global estimation. Even if a geostatistical method is not used, semi-variogram modelling has been shown to improve estimates where anisotropy is present (Sinclair and Blackwell, 2002). Kriging has been referred to as a best linear unbiased estimator (BLUE) and was developed by G. Matheron and was named by Matheron and P. Cartier to honour D. Krige. The term kriging can be applied to a number of geostatistical methods including; ordinary kriging (OK), simple kriging (SK), indicator kriging (IK), multiple indicator kriging (MIK), probability kriging (PK) and universal kriging (UK) when several non-bias conditions are required, and co-kriging when a group of variables are correlated. All of the kriging methods mentioned require a high degree of confidence in a semi-variogram for block estimation. Weighting of the samples used are unbiased with regards to minimum estimation variance and are determined by a least-squares procedure (Journal and Huijbregts, 1978). Of the kriging methods, ordinary kriging (OK) is probably the most widely used for mineral resource estimation. This is largely due to its robustness and its determination of uncertainty by using kriging variance. Uncertainty can be addressed by kriging methods to calculate local block confidence levels around the mean. This assumes a normal distribution for the error (Matheron, 1963; Isaaks and Srivistava, 1989). Due to the widespread use over other kriging methods utilised in the mining industry it is selected as the geostatistical method investigated for the estimation of MCP values in this research. Furthermore, it is chosen over simple kriging due to the requirement of the mean value of the source data to be known with confidence. This cannot be defined at this stage with MCP data and the general use of SK for more local estimates than the global estimation required in this study. Likewise, OK is selected over indicator kriging methods due to the lower level of confidence in the degree of internal error of MCP data value generation. This is a requirement for accurate assignment of indicator thresholds when using IK or MIK estimation.

A limitation of kriging to measure uncertainty is that it is calculated taking into account only the geometry of samples. That is, their spatial arrangement and not their values, ignoring local variability (Matheron, 1963; Goovaerts, 1997). It is postulated that kriging may be used as an effective estimation technique for sequestration-potential in the same manner that it is typically used in grade resource estimation. This is providing there is confidence in the semi-variogram derived and sufficient experience of the professional undertaking the estimation. There is no universally accepted way to test the hypothesis that one form of estimation is better than another. That is, it has an acceptable small degree of error (Carr, 1992). The cost of trial mining and bulk sampling is extremely important in providing confidence behind an estimate in advance of mining but can be expensive. It is possible to use cross-validation to test several estimation methods by simply counting the number of locations at which one estimator performs better (i.e. with less average error) than another (Carr et al. 1989). Testing one estimator against another is time consuming, requires personnel with experience in cross validation of the methods and sufficient computing power. The factors can all contribute to added cost for a mining company. A cross validation of polygonal, IDW and OK estimation methods are provided for the estimation of MCP values at the Turnagain ultramafic complex in northern British Columbia, Canada in Chapter 6. This is provided as a first attempt to analyse the validity of using one estimation method over another. This is from the perspective of providing a reasonable estimate (i.e. an acceptable degree of average estimation error) of the mineral carbonation potential of an ultramafic deposit.

3 Experimental, analytical and modelling procedures

The laboratory procedures and techniques used in this research were based on those found in the established literature. A direct aqueous mineral carbonation laboratory procedure was selected. It has shown to achieve the greatest stoichiometric conversion of olivine to magnesite in the literature to date. When possible geochemical, mineralogical and particle size analysis was undertaken by the author at the University of British Columbia. Some additional analytical work was provided by certified analytical laboratories. A drill hole database was provided by Hard Creek Nickel Corporation. This contained bulk elemental analysis and geological information from their Turnagain property, British Columbia, Canada.

3.1 Deposit selection and background

Two ultramafic deposits were selected as the primary focus in this research, namely the Twin Sisters ultramafic complex in Washington State, USA, and the Turnagain ultramafic complex in British Columbia, Canada.

3.1.1 Twin Sisters ultramafic complex The Twin Sisters Mountains are comprised principally of alpine-type ultramafic rocks. They are located approximately 100 km north of Seattle, Washington State, 35 km east of Bellingham, Washington State and 6 km southwest of Mount Baker in Washington State, USA. The ultramafic body forms part of the mountainous Cascade Range of northwestern Washington. The area of the deposit is approximately 90 km2, roughly 16 km long and 5.5 km wide with a maximum relief of approximately 1.5 km trending N-NW. This is the largest olivine body in the United States. Unimin Corporation provides the bulk of the commercial olivine production from this ultramafic complex. This production is modest, with a throughput of approximately 100,000 tonnes of olivine per year (Harben and Smith, 2006: pg. 679).

Figure 3.1 Simplified geological map of the Twin Sisters Mountain area, Washington State, USA

Simplified geological map outlining the approximate 90 km2 Twin Sisters ultramafic deposits, Washington State, USA (Source: modified after Ferré et al., 2005: fig. 1, pg. 143).

Field observations made by Onyagocha (1978) state the Twin Sisters unaltered rock is more than 90 % medium to coarse grained dunite. The remainder is comprised of schileren of chromite and pyroxene.

A range of Fo90 to Fo93 is noted throughout most of the Twin Sisters olivine. Most olivine crystals contain between 50 % to 51 % and 52.5 % to 53.5 % MgO (Onyagocha, 1978: pg. 1464). Low alkali content and virtually no sulphides have been reported for the dunite in the Twin Sisters ultramafic complex (Onyeagocha, 1978). The unaltered nature of the Twin Sisters olivine, the size of the ultramafic complex, and its proximity to CO2 emitting sources is the reason it was selected as the substrate material for numerous mineral carbonation research projects (Table 2.2). It was used as the substrate for determining the baseline

51 experimental conditions for direct aqueous mineral carbonation at the Albany Research Center (Gerdemann et al., 2007). Twin Sisters olivine samples were used in experimental mineral carbonation in this research as a control on the experimental methodology. The primary objective of undertaking mineral carbonation experiments on the samples from Twin Sisters is to determine the degree of stoichiometric silicate to carbonate conversion that can be achieved using the autoclave instrument available. These conversion extents are compared with those published in the literature. The results provide baseline conversion extents. These are used to calibrate the effectiveness of the process and evaluate the conversion extents achieved by undertaking mineral carbonation experiments on samples from the Turnagain ultramafic complex.

3.2.2 Turnagain ultramafic complex

The Turnagain Alaskan-type ultramafic intrusive suite (Figure 3.2.) is located 65 km east of Dease Lake, north-central British Columbia, Canada. This fault bounded, 3.5 x 8 km ultramafic intrusion provides the substrate material for experimental direct mineral carbonation in this research. The ultramafic complex hosts a low grade deposit owned by Hard Creek Nickel Corporation who has completed a feasibility study. An updated preliminary economic assessment undertaken by AMC Mining Consultants (Canada) Ltd. in 2011 and was the most up-to-date assessment as of March 2014 and can be accessed at: http://www.hardcreek.com/i/pdf/HNC-Turnagain-PA-Dec-5-Final.pdf. They report an estimated 865 Mt of measured and indicated resources at 0.21 % Ni and 0.013 % Co and a further 976 Mt inferred resource grading at 0.20 % Ni and 0.013 % Co using a 0.1 % Ni cut-off (AMC Preliminary Economic Assessment, 2011:, pg. ii) at the Turnagain deposit.

Figure 3.2 Surface geological map of the Turnagain ultramafic complex, northern British Columbia, Canada

Simplified geological map of the Turnagain ultramafic comple. The location map shows the location of the deposit in relation to other ultramafic deposits of economic importance in British Columbia. The Horsetrail Zone and Northwest Zone studied in this research are highlighted (Source: Scheel, 2007: fig. 3.1, pg. 69).

The Turnagain ultramafic complex hosts one of the few magmatic nickel occurrences of economic importance in British Columbia (Hancock, 1990; Nixon, 1991). The eastern and central part of the body is comprised of dunite and is flanked by wehrlite and olivine clinopyroxenite representing cumulate sequences (Scheel et al., 2004; Scheel, 2007). Clark (1975, 1978) determined the olivine compositions of the dunite and determined them to be generally more magnesium-rich (forsterite (Fo) >88 mol-%) as opposed to pyroxenite cumulates (Fo <88 mol-%). The total range of olivine composition in the dunite varies from 87 mol-% Fo to approximately 96.5 mol-% Fo. Dunite is considered as the primary source of magnesium that could be used for mineral carbonation at Turnagain. Wehrlite, the second most abundant rock type is comprised of cumulus

53 olivine with inter-cumulus clinopyroxene. Serpentinisation is highly variable in the rock units varying from none to complete serpentinisation, but commonly represents no more than 10 vol-% of the rock in outcrop (Scheel et al., 2004: pg. 170). The updated preliminary economic assessment undertaken by AMC Mining Consultants (Canada) Ltd. propose that the Turnagain deposit will be mined using an open pit mining method. The Horsetrail pit was designed for a 28 year life of mine and includes the Horsetrail and Northwest mineralised zones (Figure 3.3). The potential in-pit resource estimated outlined the removal of 762,876 kt of mineralised material and 317,872 kt of waste material would be required from the Horsetrail pit. The optimised end of mining pit shell calculated for the Turnagain deposit would require a further push-back. This would result in the removal of a further 499,269 kt mineralised material and 641,776 kt waste material. The mine will feed the crusher at a rate of 43,400 t/day for the first five years of production and 84,600 t/day for the rest of the mine life (AMC, Updated Preliminary Assessment, 2011, pg. 1.7v-vi). Dunite-hosted sulphide mineralisation is rare, comprising of no more than 2 vol-% of the rock, and as such, the dunite would be considered largely waste rock in a mining operation at Turnagain (Scheel et al., 2004: pg. 173). Wehrlite is most commonly associated with sulphide mineralisation and would provide much of the mineral tailings that could be used for mineral carbonation at a mining operation at Turnagain.

Figure 3.3 Topographical map of the proposed Horsetrail pit and life of mine pit shell, British Columbia, Canada

Optimised Horsetrail open pit mine design with topography at the Turnagain ultramafic complex, British Columbia, Canada. The red dashed outline highlights the 28-year life of mine open pit footprint that mines the Horsetrail and Northwest Mineralised zones.

The Turnagain ultramafic deposit was chosen as the source of the substrate material for experimental mineral carbonation in this research for the following reasons:

I. The abundance of MgO-rich magnesium silicate rocks: The ultramafic complex at an estimated 3.5 x 8 km surface area could be considered as a large potential substrate

resource for mineral carbonation with regards to the amount of anthropogenic CO2 that could be sequestered. II. The mineralogical variability (mineral type and abundance of magnesium silicate) throughout the Turnagain ultramafic complex allows the challenges associated with estimating the mineral carbonation potential of ultramafic complexes to be investigated. III. The proposed nickel-sulphide mine could feed sufficient waste rock material (>40,000 t/d) to feed a pilot-scale mineral carbonation plant.

IV. The 304 drill holes at the property have been geochemically analysed supplying sufficient spatial elemental data for a resource estimation to be undertaken.

The willingness of Hard Creek Nickel Corporation to support research into the investigation of using mining waste rock for mineral carbonation is a proactive example of how mining companies can provide a basis from which research in to mineral carbonation as a method of CO2 sequestration can develop.

3.2 Sample collection

Samples for experimental mineral carbonation undertaken in this research consist of drill core collected on a visit the Turnagain property in July 2009. Approximately 75,620 metres (248,100 feet) of diamond drilling in 304 drill holes have been completed at the property. A suite of 11.532 geochemical assay results measured using inductively coupled plasma emission spectroscopy was provided by Hard Creek Nickel Corporation for all of the drill holes at the Turnagain property. Drill core samples were collected on 4 metre intervals to maintain consistency with the corresponding geochemical assay results from five pre-determined drill holes. These drill holes were chosen because of the abundance of dunite, specifically the ‘green dunite’ rock unit that makes up a large proportion of the proposed mining waste rock at Turnagain. The following drill holes; T04-24, T04- 25, T06-110, T06-111, and T06-116 (Figure 3.4.) were sampled. The core samples were split on site and sent back to the University of British Columbia. Original pulp samples from existing prepared drill core samples sent for geochemical assay by Hard Creek Nickel Corporation were obtained where available. These samples were stored in sealed sample envelopes at a dry storage location in Chilliwack, British Columbia. The pulp samples provide material that used in laboratory mineral carbonation experiments. The samples are expected to have undergone minimal, if any oxidation post sample collection. The quality assurance and quality control procedures undertaken by Hard Creek Nickel Corporation can be found life in the AMC, Updated Preliminary Assessment (2011, pg. 33-39). Some of these pulp samples were sent for geochemical assay as a quality assurance check on the original geochemical assay results provided by Hard Creek Nickel Corporation. Olivine samples from the Twin Sisters ultramafic complex were purchased from Ward’s of Canada Limited, a commercial supplier of olivine samples. Three kilograms of rock specimens were purchased and then prepared for experimental mineral carbonation at the University of British Columbia.

Figure 3.4 Drill holes selected from Turnagain for experimental mineral carbonation

Topographical map displaying the drill holes (T04-24, T04-25, T06-110, T06-111, and T06-116) selected for experimental direct aqueous mineral carbonation. The red dashed outline highlights the 28-year life of mine open pit footprint that mines the Horsetrail and Northwest Mineralised zones.

3.3 Sample preparation

Crushing and grinding of the drill core samples was undertaken to reduce the size of the material for geochemical analysis, mineralogical analysis and experimental mineral carbonation. Density measurements of the drill core samples were obtained prior to crushing and grinding.

3.3.1 Density measurements Hard Creek Nickel geologists calculated specific gravity (SG) measurements every 20 samples using drill core samples of up to 50 cm. These SG measurements were recorded and stored in the geochemical drill hole database provided by Hard Creek Nickel Corporation. All the SG values used in this research were measured using a water immersion method where the sample is weighed in both air and water. SG measurements were determined for all drill core samples collected by the author. This was done as quality assurance check on the provided measurements. The results of this quality check are provided in Table 3.1. The mean SG results for the five lithologies tested indicate that the SG results provided in the

57 drill hole database are sufficiently precise to support any resource estimation undertaken in this research. The significance of the accurate determination of SG values for the rocks studied is expressed when undertaking MCP resource estimation. To date, preliminary estimates of the mineral carbonation capacity of ultramafic rocks has used an average SG value for each ultramafic rock examined. The minimum and maximum SG values shown in Table 3.1 emphasise the inherent variability in SG values. This can lead to an inaccurate determination of the mass of ultramafic rock available as a substrate for mineral carbonation and consequently a miss-representation quantity of CO2 it is capable of sequestering. It is important that time is spent on statistically analysing the interpolation of SG values when undertaking an MCP resource estimate.

Table 3.1 Bulk specific gravity comparative statistics

Lithology Turnagain Database Recorded Quality Check Data by Author Density Values No. Min Max Mean No. Min Max Mean Samples Samples Dunite 108 2.68 3.42 2.96 37 2.73 3.36 3.01 Green dunite 212 2.62 3.33 2.96 72 2.60 3.39 2.97 Serpentinised dunite 42 2.76 3.24 3.10 6 2.92 3.26 3.07 Wehrlite 30 2.72 3.29 3.06 5 2.79 3.24 3.03 Serpentinised wehrlite 50 2.09 3.24 2.95 7 2.25 3.18 2.80

Bulk specific gravity comparative statistics from the geochemical drill hole database provided by Hard Creek Nickel Corporation and measurements undertaken by the author.

3.3.2 Rock crushing and grinding The drill core collected from the Turnagain ultramafic complex was split from the half core samples and weighed between 75 g and 200 g (Figure 3.5). The samples from the Twin Sisters ultramafic complex comprised of 3 kg of rock pieces approximately 50 g each. Each rock sample was passed through crushing circuit comprised of a primary Blake jaw crusher (set at 4 cm) and a McCully gyratory crusher (set at 1.2 cm). When 100 % of the material passed through the gyratory crusher the sample was split with a riffle splitter and approximately 25 % was retained. The

58 remaining samples were passed through a series of two cone crushers until approximately 90 % of the sample passed through a 5-mesh sieve (4 mm). Sample material was then placed in a tungsten carbide ring mill and ground in 90 second intervals for a total of six minutes. After each interval the sample was screened in a series of sieves from 5-mesh to 400-mesh (37 µm). All the material that was coarser than 200-mesh (75 µm) was then placed back in to the ring mill for re-grinding. The tungsten carbide ring mill was initially cleaned with coarse silica to minimize contamination of samples from the mill and ring. All samples were riffle split before use in experimental mineral carbonation. Varying methods of mechanical and chemical mineral activation, such as; dry and wet milling (Kleiv and Thornhill, 2006), various high-energy planetary, attritor, and nutating mills (Kalinkin et al., 2003 and, 2004; Baláž et al., 2008; Fabian et al., 2010), electrolysis and heat pre-treatment (Li et al., 2009) have been investigated in the literature. Further investigation in to mineral activation techniques are beyond the scope of this research. The mineral pre-treatment method chosen was mechanical activation by size reduction. This decision is oriented around the mineral preparation undertaken for inductively-coupled plasma emission spectroscopy and X–ray fluorescence by the mining industry and associated geochemical assay laboratories. The rationale behind this decision is based on developing a technique that mining companies can test the mineral carbonation extent of the rocks from drill core samples without having to apply extra preparation techniques than those used in traditional geochemical assay lab testing.

Figure 3.5 Drill core photographs of the samples tested from the Turnagain ultramafic complex

Drill core photographs of the samples tested from the Turnagain ultramafic complex, northern British Columbia. All of the photographs are identified by the sample name at the top. A centimetre scale is provided for each individual image. The degree of alteration is visible through all samples with the substrate matrix of each sample being dominated by olivine with the majority of the alteration veinlets comprised of serpentine. The variability and pervasiveness of serpentinisation from sample to sample is apparent.

3.4 Analytical methods

A number of laboratory based analytical techniques were used to investigate the size, morphology, geochemistry and mineralogy of both the substrate and product material used in experimental direct aqueous mineral carbonation. These include particle size analysis, X-ray diffraction, X-ray fluorescence, inductively coupled plasma emission spectroscopy, and scanning electron microscopy. The range of methods selected for analysis of the reactant and product material were based largely on the availability if the resources available to the author and the widespread use of the techniques within the mining industry. Particle size analysis was undertaken to determine the mean particle size of the substrate material used in the mineral carbonation experiments. X-ray diffraction using the Rietveld refinement method was used to determine the mineralogical composition of both the mineral carbonation reactants and products. X-ray fluorescence analysis was undertaken to determine the major oxide values of the substrate material used in the mineral carbonation experiments. Inductively coupled plasma emission spectroscopy was undertaken as a quality assurance check on the geochemical assay data provided for the Turnagain ultramafic complex. Scanning electron microscopy was used to observe the particle characteristics and textures of both the mineral carbonation reactants and products.

3.4.1 Particle size analysis The particle size of solid reactant material for use in experimental mineral carbonation was determined. The effect of particle size on mineral carbonation extent is well documented from previous experimental results in the literature. Research found that an increase in particle activation often results in a greater mineral carbonation extent (Gerdemann et al., 2007). In this research the effect of degree of mineral pre-treatment was not fully investigated. Mineral pre-treatment comprised of reactant particle size reduction alone, with no heat-treatment stage. The purpose of undertaking the particle size analysis was to ensure both consistency in the particle size of reactant material. It also ensures that previous mineral carbonation experiments such as those undertaken by Gerdemann et al. 2007 could be replicated for standardisation of the experimental mineral carbonation in this research. The particle size of all the solid reactant material investigated from both Twin Sisters and Turnagain were determined. This completed using a Malvern Mastersizer Hydro 2000(s) instrument at the University of British Columbia. The Malvern Mastersizer uses a laser diffraction measurement where particles are passed through a focused laser beam in which particles scatter light at an angle inversely

61 proportional to their size. The angular intensity of the scattered light is measured by a series of photosensitive detectors. A map of scattering intensity versus angle is used to calculate the particle size. The scattering of particles is accurately predicted by the Mie scattering model using an assumption of spherical particle morphology. This model is applied within the Mastersizer-S software to allow accurate sizing across material in the range of 0.02 μm to 2000 μm (www.malvern.co.uk/ms2000, October 2011).

3.4.2 X-ray diffraction The majority of experimental mineral carbonation research papers have used X-ray diffraction (XRD) to identify the mineral phases present (i.e. the principal magnesium silicate reactant mineral and the presence of magnesite in the product material) and used an elemental mass balance approach to determine the stoichiometric silicate to carbonate conversion extent achieved. XRD using the quantitative Rietveld refinement method was chosen to determine the stoichiometric conversion extent achieved during experiments in this research. It was chosen due to its proficiency at quantifying the crystalline phases of rocks. XRD permits an accurate mineralogical investigation to be developed rather than a purely elemental approach to the mineral carbonation extent. This is important when considering rocks that have a variable mineralogy such as those processed at mining operations. It allows the effectiveness of individual magnesium silicate minerals to sequester CO2 to be analysed. The quantitative phase (i.e. mineralogical) analysis of the reactant solid material and post- experimental mineral carbonation products was undertaken by the Electron Microbeam/X-ray Diffraction Facility in the Department of Earth, Ocean, and Atmospheric Sciences, University of British Columbia. The department services were used due to their experience in quantitative XRD analysis. The synthesis of the results produced was analysed by the author. The XRD results were not verified by an alternative laboratory. It is assumed that the expertise in analysing olivine and serpentine-rich rocks produced accurate results using the techniques chosen. The dominant mineral assemblages determined by XRD were confirmed by identification in thin section and those reported in the literature. The quantification of each mineral proportion however, is assumed to be accurate from the analysis results provided. All samples were reduced into fine powder to the optimum grain-size range for XRD analysis (<10 µm) using agate pellets in ethanol in a vibratory McCrone Micronising Mill for 7 minutes. Step-scan X-ray diffraction data were collected over a range 3-80°2 with CoKa radiation. This was undertaken on a Bruker D8 Focus Bragg-Brentano diffractometer equipped with a Fe monochromator foil, 0.6 mm (0.3°) divergence slit, incident- and diffracted-beam Soller slits and a LynxEye detector. The long fine-focus Co

X-ray tube was operated at 35 kV and 40 mA, using a take-off angle of 6°. A step size of 0.04°2θ was used 0.9 sec/step counting time. The X-ray diffractograms were analysed using the International Centre for Diffraction Database (ICDD) PDF-4 and Search-Match software by Siemens (Bruker). X-ray diffraction data of the samples were refined with Rietveld program Topas 4.2 (Bruker AXS) by the staff in the Electron Microbeam/X-ray Diffraction Facility. The Rietveld refinement method models the widths and shapes of the diffraction peaks. A model for aberrations in the geometry of the peaks and a model for the background are used to generate a diffraction pattern for each phase present (Rietveld, 1969). These individual patterns are subsequently summed and fitted to the experimental diffraction pattern. Fitting is achieved by least-squares refinement of the structural parameters of each phase and any global parameters that affect the pattern. The calculated contribution of constituent phases is used to adjust the fit between calculated and experimental patterns (Raudsepp and Pani, 2003). The Rietveld refinement method permits the characterisation of multiple phases simultaneously (Raudsepp and Pani, 2003). The refinement also shows the relative mass of each and all the phases to be derived from the diffraction pattern. This uses the relationship described by Hill and Howard (1987):

푆푟(푍푀푉)푟 푊푟 = Eq. 3.1 Σ푡푆푡(푍푀푉)푡

Where, Wr is the relative weight fraction of phase r in a mixture of t phases, S is the scale factor derived from Rietveld refinement, Z is the number of formula units per unit cell, M is the mass of the formula unit (atomic mass units), and V is the volume of the unit cell (Å3). The quantitative mineral abundances represent the relative amounts of crystalline phases normalised to 100 % providing they are in the crystalline phase. The mineral abundance cannot be quantified for any amorphous material present using Rietveld refinement. This is due to the disordered structural nature. The kaolinite-serpentine group minerals present in the samples tested are structurally disordered and of unknown poly-type. The services of the Electron Microbeam/X-ray Diffraction Facility at UBC were used rather than X-Ray diffraction analysis undertaken by the author due to experience in analysing such mineral groups. It was felt that a more accurate representation of the mineral abundance could be achieved by acquiring the services. Serpentine peaks were fitted using the space group and unit cell parameters (Pawley method) of Lizardite-2H supplemented with peak functions. This method is independent of atomic parameters and

63 no quantitative mass measurement is possible. Thus, the fitted serpentine pattern was considered as an amorphous phase. Its contribution to the pattern was back-calculated from the measured degree of crystallinity to estimate the quantity of serpentine in the samples.

3.4.3 Inductively coupled plasma emission spectroscopy Geochemical assay results obtained from diamond drill core samples are widely used within the mining industry to provide the data required to undertake spatial mineral resource estimations amongst other geochemical analyses for rock characterisation and processing requirements. The most widespread method used in determining these geochemical assay results is inductively coupled plasma emission spectroscopy (ICP-ES). Its significance as a widespread method for geospatial data generation was a driving factor in selecting this data for the geospatial generation of mineral carbonation potential values. Geochemical data provided by ICP-ES results at mining operations that could potentially provide substrate material for mineral carbonation should be used to more accurately determine the mineral carbonation resource potential. The elemental composition of the rock at known locations in a rock mass allows the magnesium and calcium quantity to eventually be spatially estimated. The ICP-ES results provided by Hard Creek Nickel Corporation for their Turnagain property was used to spatially estimate the magnesium and calcium abundance. The ICP-ES results also provide the elemental data used in the mineralogical estimation undertaken to determine the mineral carbonation potential at Turnagain. To generate the geochemical dataset provided by Hard Creek Nickel Corporation drill core samples were received by Acme from Hard Creek Nickel Corporation for all drill holes at the property between 2002 and 2008. Prior to 2004 the samples were analysed for approximately 23 major and minor elements by aqua-regia digestion followed by an ICP-ES finish. Samples collected and analysed between

2004 and 2008 were subjected to a four-acid (HNO3-HClO4-HF and HCl) digestion followed by ICP-ES analyses to determine the total values for 25 elements. The 25 elements selected provide the basic element range offered in analysis by major laboratories. This elemental suite is the most commonly selected by mining companies, predominantly as it is the cheapest option to gain the quantity of data they require to develop a resource estimate. In the four-acid method 0.25 g of the sample is heated in

HNO3-HClO4-HF to fuming and taken to dryness. The residue is then dissolved in HCl. The solutions are then analysed by ICP-ES. The sulphur content was analysed by the Leco furnace induced method. (AMC, Updated Preliminary Assessment, 2011: pg. 31). Laboratory quality controls were maintained using internal standards, sample blanks and duplicate samples by the laboratory. In addition, staff from Hard Creek Nickel Corporation inserted reference

64 sample pulps (standards) with known sample numbers and blank samples were inserted every 30 samples. Pulps from every 10th sample were sent to a check laboratory (ALS Chemex, Vancouver) (AMC, Updated Preliminary Assessment, 2011: pg. 31). As a further quality assurance check on the assay results provided by Hard Creek Nickel Corporation within the geochemical drill hole database, 18 samples were sent to Acme Analytical Laboratories Ltd by the author. for re-analysis. These samples correspond to the samples sent for XRF and were analysed using a four-acid digestion with an ICP-ES finish and Leco in. The analytes tested include; Ag, Al, As, Au, Ba, Be, Bi, Ca, Cd, Co, Cr, Cu, Fe, K, La, Mg, Mn, Mo, Na, Nb, Ni, P, Pb, S, Sb, Sc, Sn, Sr, Ti, Th, U, W, Y, Zn, and Zr using four-acid ICP-ES and total C and S using the Leco method.

3.4.4 X-ray fluorescence 18 rock powder samples were split using a riffle splitter and sent to Acme Analytical Laboratories Ltd. in Vancouver, British Columbia. These samples were analysed by X-ray fluorescence (XRF) for major oxides and loss on ignition (LOI). The analytes and detection limits for the samples tested are given in Table 3.2. 15 g of reactant rock powder was provided for each sample. The pulp samples were fused with Li2B4O7/LiBO2 at temperatures ranging between 1000°C and 1200°C in the XRF analysis. XRF uses the characteristic emission pattern of secondary (fluorescent) X-rays to determine the elemental composition. The samples were analysed using XRF at Acme Analytical Laboratories Ltd. to determine the variance of estimated oxide values calculated from the major element ICP-ES elemental analysis. Major oxide values are required to estimate the mineralogy of mineral carbonation substrate material. By quantitatively estimating the mineralogy of substrate material the mineral carbonation potential of the material can be more accurately predicted. XRF data also provides SiO2 abundance and a LOI percentage unavailable from standard ICP results. LOI can be used to estimate the volatile content including as water.

Table 3.2 Analytes tested and detection limits using XRF with Li2B4O7/LiBO2 fusion.

Analyte Tested Detection Limit Upper Limit

SiO2 0.01 % 100 % Al2O3 0.01 % 100 % Fe2O3 0.01 % 100 % CaO 0.01 % 100 % MgO 0.01 % 100 %

Na2O 0.01 % 100 % K2O 0.01 % 100 % MnO 0.01 % 100 %

TiO2 0.01 % 100 % P2O5 0.01 % 100 % Cr2O3 0.001 % 100 % Ba 0.01 % 5 % LOI 0.1 % 100 % C 0.02 % 100 % S 0.02 % 100 %

Analytes tested and detection limits using XRF with Li2B4O7/LiBO2 fusion on 18 rock powder samples from Twin Sisters and Turnagain ultramafic complexes. Testing was undertaken by Acme Analytical Laboratories Ltd., Vancouver, British Columbia.

3.4.5 Scanning electron microscopy A Philips XL-30 scanning electron microscope (SEM) was used to observe the solid reactant and product material post-experimental mineral carbonation. This was a largely completed as a visual check on the effectiveness of the mineral pre-treatment on particle size reduction and to visually identify the presence and character of the magnesite formed in the post-carbonation product material. Representative reactant and product samples were mounted and carbon coated. SEM permits the observation and characterisation of organic and inorganic material. This is done on a nanometer to micrometer scale by irradiating a confined area with a finely focused 15kV electron beam (Goldstein et al., 2003). Both secondary electron (SE) and back scattered electron (BSE) signals were used to generate images of the solid material. This was primarily to identify any particle surface and texture alterations that may have occurred during the mineral carbonation process. The SEM was equipped with a Bruker Quantax energy-dispersive X-ray (EDX) system to measure the characteristic major elements of each sample. The EDX system was used to identify unknown crystals in the SEM images taken.

3.5 Experimental mineral carbonation laboratory procedures

Experimental mineral carbonation was undertaken on both samples from Twin Sisters and Turnagain at the Norman B. Keevil Institute of Mining Engineering, University of British Columbia. The direct aqueous mineral carbonation approach pioneered at the Albany Research Center was selected. It provides the most successful stoichiometric olivine to magnesite conversion extents in the literature to date. The relative simplicity and affordable equipment requirements used in the direct aqueous mineral carbonation procedure adheres itself to the possibility of undertaking the experimental mineral carbonation experiments directly at a mining operation. The experimental mineral carbonation undertaken in this research differs from that present in the literature in the sense that it attempts to undertake direct aqueous mineral carbonation on substrate rocks that contain both olivine and serpentine minerals. Previous experiments have targeted either olivine or serpentine separately. In those examples the experimental conditions can be tailored to achieve the greatest stoichiometric silicate to carbonate conversion for that given magnesium silicate. The principle magnesium silicate mineral present in the Turnagain ultramafic complex is olivine. As such, the experimental mineral carbonation conditions chosen were targeted towards sequestering CO2 using olivine. Serpentine is present in varying quantities in each sample tested. The effectiveness of the chosen method to extract magnesium from serpentine for mineral carbonation is addressed.

3.5.1 Sample selection A total of 15 samples were tested from the Twin Sisters ultramafic complex. These samples were taken from the bulk sample created during crushing and grinding. They were divided using a riffle splitter to ensure homogenisation of the samples. These samples were tested to determine the effects of varying the testing duration, carrier solution chemistry, the use of quartz as a particle attrition additive and the effect of varying the solid to solution ratio of the slurry. These samples tested did not intend to expand on the current literature available on direct aqueous mineral carbonation, however important findings relevant to the development of mineral carbonation processing were observed and are discussed throughout the dissertation. The testing was aimed at fine tuning the process methodology and to determine the maximum stoichiometric magnesium silicate to carbonate conversion extent achievable using the experimental setup available. As a result a repeatable experimental methodology was established. This was used for the duration of the mineral carbonation testing.

A total of 16 samples were tested from the Turnagain ultramafic complex. These samples were taken from drill core samples. One sample was taken from drill hole T04-24, One sample from drill hole T04-25, five samples from drill hole T06-110, five samples from drill hole T06-111, and four samples from drill hole T06-116. These drill holes were selected due to broad spacing within the proposed open pit mine design. Samples tested from drill holes T06-110, T06-111, and T06-116 were taken at approximately 40 m intervals down each drill hole. The samples were taken at approximate 40 m spacing down hole to determine any mineral variability down the length of the drill hole. More than 16 samples would have ideally been tested for increased statistical integrity. However financial constraints due to the cost of XRD using the Rietveld method as an analysis technique to determine the mineral carbonation extent prevented this. In addition the estimation of serpentine using the Rietveld refinement method can be considered semi-quantitative. This is due to its disordered nature and unknown poly-type. Thus 16 reactant samples were considered sufficient to evaluate the mineralogy prediction technique developed in this research. In addition it permitted the determination of whether the mineral carbonation of serpentine was achieved using the direct aqueous experimental mineral carbonation procedure. The comparison of the stoichiometric silicate to carbonate conversion achieved from both Twin Sisters and Turnagain is considered adequate from the 16 samples to determine the effectiveness of the experimental setup.

3.5.2 Reactant slurry generation A buffered carrier solution was generated for the slurry used in experimental mineral carbonation. A 45 ml slurry was used for all samples tested ensuring sufficient expansion room in the autoclave reaction chamber for the safe addition the gaseous CO2. A 0.64 M NaHCO3 and 1 M NaCl solution was generated. This duplicated the solution used for experimental mineral carbonation experiments at ARC. This solution chemistry was used for 12 of the Twin Sisters samples tested. One sample (TSO-AJ14) used a distilled water carrier solution and two samples (TSO-AJ4 and TSO-AJ5) used a solution chemistry of 2.5

M NaHCO3 and 1 M NaCl. These solutions were used to investigate the effect of modifying the solution alkali cation species concentration.

The increased NaHCO3 solution was tested to see if the solution chemistry change replicated the increased mineral carbonation extent for olivine as observed by Chizmeshya et al. (2006; 2007). Due to the small substrate quantity used in the testing, the quantity of NaHCO3 and NaCl remaining in the product solids was detrimental to the stoichiometric conversion determination. As a result, the standard buffered solution proposed by O’Connor et al. (2005) was used for all the samples tested from

Turnagain. The buffered solution was selected over more aggressive, acid-based catalysing agents due to the solution proving to be affective at achieving both the Mg2+ leaching and carbonate precipitation in a single process step and the low molarity of the NaHCO3 was selected due to its ability to be effectively recycled. All samples tested from both Twin Sisters and Turnagain contained a solid feed in which at least 80 % passed through a 400-mesh (37 μm) sieve. The solid constituent of the slurry varied between 4.5 g (10 %), 6.75 g (15 %), and 9.0 g (20 %) solid rock powder when testing samples from Twin Sisters. This was carried out to investigate the effect of varying the quantity of solid rock powder in the slurry feed on the mineral carbonation extent achieved. No observable difference was seen between using 10 and 15 % solids. However, some un-reacted material was found directly below the magnetic stirrer when using a 20 % solid feed. As a result feed solid constituents when testing samples from Turnagain were fixed at 6.75 g (15 %). this was to maximize the solid product for analysis whilst ensuring thorough mixing of the slurry.

3.5.3 Autoclave testing procedure A 100 ml Hastelloy-C Parr bench-top stirred autoclave was used for the experimental mineral carbonation (Figure 3.6). This autoclave is smaller than many of those used in the literature, which are typically 2 litres in capacity for testing bulk samples of substrate material. The smaller sized reactor was chosen for financial reasons and it was considered that a 100 ml reaction chamber would be adequate to test the quantity of material often available from drill core samples. The autoclave is fitted with a CO2 gas inlet valve and a controllable magnetic stirrer. The stirrer speed, heat and pressure can all be controlled externally through an attached controlling device. The vessel was not filled over half of the maximum volume to account for safe gas and liquid expansion within the vessel, preventing overpressure. The slurry containing the rock powder and the buffered carrier solution are thoroughly mixed and added to the 100 ml chamber. The vessel is sealed by clamping and tightening the 6 bolts firmly holding the chamber in place. 99.99 % pure CO2 is supplied from a gas cylinder fitted with a flow regulator and a one-way check valve to ensure no gas can be forced back into the cylinder. The CO2 is slowly added to the chamber via a gas inlet valve until the targeted purge pressure is achieved. A purge pressure of 800 ± 25 psi (55 bar) was targeted. This was targeted as the safest possible maximum pressure range to add

CO2 into the chamber without the aid of a gas compressor and booster pump.

Figure 3.6 Photograph and schematic of the autoclave setup

On the left is a photograph of the autoclave setup at the University of British Columbia used by the author including the autoclave apparatus and external controller. On the right is a schematic of the setup used (Source: modified after Zhao et al., 2009: fig. 1, pg. 407).

Once the CO2 is added to the reaction chamber the temperature is elevated to a constant operating temperature of 185 ± 2°C. This was shown to be the most effective temperature for mineral carbonation when using olivine as the substrate mineral (O’Connor et al., 2005). This subsequently increases the internal gas pressure. The magnetic stirring rod is set to 750 rpm during the heating period until the starting operating conditions are reached. The heating period takes approximately 45 minutes. Once the operating temperature is reached the stirrer speed is increased and the CO2 pressure within the vessel is recorded. Temperature, pressure and stirring speeds are recorded every hour the test is run. The testing times for the Twin Sisters samples varied between one hour and six hours. All the samples from Turnagain were run for two hours. A two hour testing period was selected to ensure that enough mineral carbonation takes place whilst reducing the need to run long duration experiments when considering the experimental setup, autoclave warm up and cool down periods. At the completion of the test, the heater is turned off and the stirring speed is reduced to 300 rpm. The vessel was cooled to an internal temperature of 80°C taking between 45 and 55 minutes. The gas pressure was recorded before being released. The bolts and the clamp holding the vessel in place are

70 released and the contents removed into a beaker. The vessel and stirring rod is flushed with approximately 50 ml of distilled water to remove the entire remaining solid product and is added to the beaker. The beaker containing the solid product in the carrier solution is covered overnight while the product settled. The autoclave appendages and vessel are thoroughly cleaned for future use.

3.5.4 Post-carbonation product recovery Once the solid product settled overnight, approximately 75 ml of the supernatant is removed using a pipette. This solution is tested for pH using a pH electrode and reader. The remaining slurry with approximately 15-20 ml of solution is placed in an oven overnight (24 hours) at 110°C. The solid product is removed from the beaker and ground with a pestle and mortar to break up any dried clumps that may have formed. The dried solid sample is then weighed and split using a serrated blade to generate two representative samples. Approximately 2 g of each sample was sent for XRD with Rietveld analysis. The remaining solid product sample was used for product classification including SEM and particle size analysis.

3.6 Estimating mineralogy from geochemical analysis data

Typical methods of quantifying mineral volumes within a rock mass include either XRD analysis or the point- or area-counting of minerals of thin sections. With no time or financial constraints, it would be preferable to quantitatively determine the mineralogy down the length of each drill hole. Experimental mineral carbonation testing of each sample to determine the MCP at regular intervals down each drill hole would also be favourable. A mineralogical mass balance calculation quantifying the capacity of the Mg- and Ca-silicates in the rock to sequester CO2 using a variety of mineral carbonation procedures could then be analysed. This however is both time consuming and expensive. Major oxide analyses by XRF of the 16 reactant samples from the Turnagain ultramafic deposit plus a representative reactant sample from the Twin Sisters ultramafic complex were used to estimate the modal mineral abundance using the MINSQ program devised by Herrmann and Berry (2002). This program was selected from those discussed in Section 2.5 as it is one of the most recent attempts at modal mineral estimation that provides an open source to the software. The program is not aimed for specific use on a single suite of rocks making it applicable to ultramafic rocks. MINSQ uses the least squares method to quantitatively estimate the mineral proportions of a rock from whole-rock geochemical data. The spreadsheet program permits the user to select a probable or known mineral

71 assemblage. Utilising the Solver tool (a non-linear optimisation code) in Microsoft ExcelTM it converts the measured oxide data into a predicted normative assemblage. A simplified 5 mineral assemblage was estimated based on the minerals identified within the samples from XRD with Rietveld refinement analysis. This assemblage comprised of olivine, serpentine, diopside, magnetite and quartz. Each sample was calculated by pasting the major oxide values into a fixed row in the spreadsheet and then activating the Solver function. The Solver is constrained to keep the changing cells in the spreadsheet ensuring realistic estimates of the mineral proportions. By using a least squares method, the Solver minimises the sum of the squared residuals. This minimizes the difference between the actual (analyses) and estimated (calculated) percentage proportions (Herrmann and Berry, 2002). The estimated mineral proportions using the MINSQ spreadsheet are presented in Table 3.3. These values are compared with the measured mineral proportions of each sample determined by XRD with Rietveld refinement. The MINSQ program is internally consistent (Herrmann and Berry, 2002). A number of error sources in the estimation of the predicted mineral proportions of the Turnagain samples are inevitable. The olivine proportions estimated in Table 3.3 were based on a Fo91 (Mg1.82Fe0.18SiO4) composition. This is the average forsteritic value for olivine within the dunitic rocks at Turnagain. The inability to account for variations in the end-member composition is a consistent source of error when using the MINSQ spreadsheet. The average composition must be used resulting in the inaccurate proportioning of Fe2O3 and MgO on a sample by sample basis. This is an inherent problem when attempting to accurately estimate the olivine composition using any modal estimation program. The olivine composition could be adjusted on a sample by sample basis. When considering a large series of samples to be analysed at a potential mining operation, this becomes practically problematic.

Table 3.3 MINSQ and XRD comparison of reactant samples tested

Mineral Minerals (Wt.%) Sample ID estimation Olivine Serpentine Diopside Magnetite Quartz Residual method (SSQ) a TSO MINSQ 99.68 01.72 - - - 2.85 XRD 97.90 00.20 - - 0.20 Ab 01.78 01.52 - - 0.20 04-24-AJ3 MINSQ 45.37 47.25 - 5.63 - 2.35 XRD 57.12 30.53 04.63 5.27 0.33 A 11.75 16.72 04.63 0.36 0.33 04-25-AJ6 MINSQ 46.79 27.48 15.61 9.05 - 0.38 XRD 58.92 13.52 20.78 3.88 0.33 A 12.13 13.96 05.17 5.16 0.33 06-110-AJ2 MINSQ 68.80 26.69 - 3.58 - 2.31 XRD 75.43 18.00 - 4.86 0.61 A 06.63 08.69 - 1.28 0.61 06-110-AJ6 MINSQ 79.53 16.43 - 3.36 - 1.27 XRD 62.62 28.10 02.02 5.49 0.40 A 16.91 11.67 02.02 2.13 0.40 06-110-AJ10 MINSQ 79.95 17.19 - 1.49 - 5.79 XRD 77.28 15.74 01.79 3.31 0.31 A 02.66 01.45 01.79 1.83 0.31 06-110-AJ13 MINSQ 65.19 29.02 00.28 4.05 - 1.36 XRPD 73.11 16.31 03.34 5.06 0.56 A 07.92 12.79 03.06 1.01 0.56 06-110-AJ17 MINSQ 38.24 43.21 10.11 6.90 - 1.37 XRD 67.40 12.48 14.03 5.10 0.33 A 29.17 30.73 03.92 1.80 0.33 06-111-AJ1 MINSQ 57.94 34.87 00.65 5.23 - 1.53 XRD 73.84 17.21 05.14 2.69 0.22 A 15.90 17.66 04.48 2.54 0.22 06-111-AJ4 MINSQ 80.84 15.86 00.01 2.99 - 0.13 XRD 81.83 14.06 01.09 2.13 0.19 A 01.00 01.80 01.08 0.86 0.19 06-111-AJ7 MINSQ 67.59 26.88 - 4.80 - 0.64 XRD 76.29 15.51 01.40 5.32 0.37 A 08.70 11.37 01.40 0.52 0.37 06-111-AJ10 MINSQ 52.44 36.08 02.58 - - 0.95 XRD 59.92 28.27 04.47 6.25 0.20 A 07.48 07.80 01.89 1.40 0.20 06-111-AJ13 MINSQ 63.79 28.61 02.11 4.53 - 0.79 XRPD 67.18 21.90 05.87 3.94 0.27 A 03.39 06.71 03.76 0.59 0.27 06-116-AJ7 MINSQ 61.80 35.34 - 1.99 - 11.12 XRPD 64.40 31.42 - 1.45 0.21 A 02.60 03.92 - 0.54 0.21 06-116-AJ9 MINSQ 52.55 43.29 - 3.12 - 5.76 XRD 66.93 25.66 - 3.54 0.22 A 14.38 17.63 - 0.42 0.22 06-116-AJ11 MINSQ 59.78 37.36 - 2.06 - 8.94 XRD 53.41 40.42 - 2.43 0.31 A 06.37 03.06 - 0.37 0.31 06-116-AJ12 MINSQ 57.53 39.03 - 2.19 - 8.47 XRD 67.57 23.12 - 3.34 0.21 A 10.04 15.91 - 1.15 0.21

Estimated mineral proportions of 1 sample from the Twin Sisters (TSO) and 16 samples from the Turnagain ultramafic complexes comparing the results analysed from the MINSQ spreadsheet program (Herrmann and Berry, 2002) and XRD.

a Residual SSQ is the sum of the residuals squared b A is the calculated absolute difference; where, A = MINSQ (Wt.%) – XRD (Wt.%)

- The samples tested assume that LOI values are assigned as H2O in the estimation of the mineral assemblage. This can result in the over estimation of hydrated minerals when using the estimation program, serpentine in this example. The estimation of the serpentine quantities in the samples tested is assumed to be lizardite. A general formula Mg3Si2O5(OH)4 was used and does not account for any antigorite or chrysotile present in the sample. This is based on the dominant presence of lizardite from

XRD analyses. All the iron estimated is expressed as Fe2O3 due to the XRF fusion process. As a result the mineral quantity totals greater than 100%. This can cause an estimation error when proportioning the iron content in olivine estimation. The iron proportion of the olivine mineral makeup assumes a FeO contribution rather than Fe2O3. All remaining iron is assigned to magnetite (Fe2O3) even though hematite

(Fe3O4) was also determined in nearly all samples by XRD. It is yet assumed that the results are applicable for this purpose. The residual SSQ ranges from 0.38 to 11.12 for the 17 samples tested. A proportion of the residual values in this series is a product of the simplicity of the mineral assemblage tested. The results do not account for any minerals that may contain MnO, TiO2, Al2O3 and Na2O from the XRF oxide analyses. The moderately broad SSQ ranges vary on a sample by sample basis. This is due to residual MgO, SiO2 and

H2O which may be a product of inaccurate LOI measurements. This could just be an inherent estimation error when attempting to estimate the particular mineral assemblage attempted in this research. The graphical representation of the MINSQ and XRD mineral estimation methods are shown for the 16 Turnagain samples analysed in Figure 3.7. It is difficult to estimate the degree of accuracy in the mineral abundances proposed by both the MINSQ spreadsheet estimation method and the XRD with Rietveld refinement. It is assumed that the XRD with Rietveld refinement is the most accurate measurement available. This is due to the clarity of the diffractogram peaks used in the analysis. It must be stated that the interpreted serpentine mineral abundances are based on the assumption that all amorphous material in the samples tested are lizardite. As such an over estimation of the serpentine modal abundance is expected.

Figure 3.7 Modal mineral estimation comparisons of 16 samples tested from Turnagain using XRD and MINSQ

Modal mineral estimation comparisons of 16 samples tested from the Turnagain ultramafic complex, northern British Columbia using quantitative XRD with Rietveld refinement and estimated values from XRF oxide analyses converted using the MINSQ method created by Herrmann and Berry (2002).

When comparing MINSQ modal mineral abundances with those determined by XRD with Rietveld refinement, a significant proportion of the Turnagain samples show a general underestimation of olivine. A general overestimation of serpentine abundance is also seen. This is likely caused by the inaccurate measurement of LOI values during XRF analysis and the resultant proportioning of the LOI values in the form of H2O to serpentine. A notable point to highlight is that the closest approximation between XRD and MINSQ is sample 06-111-AJ4 which also had the lowest residual SSQ of 0.13 from the MINSQ analysis. This suggests that providing a successful fit with a residual SSQ of less than 1, major oxide analyses used in the MINSQ spreadsheet program could successfully estimate the modal mineral composition of ultramafic rocks, however many more samples would be required for confirmation. The use of a modal mineral estimation program such as MINSQ could be justified as an affordable mineral estimation technique of ultramafic rocks. This is particularly true when considering expensive quantitative XRD with Rietveld refinement or a time consuming point-counting technique. This is amplified when considering the large number of drill hole samples that would have to be analysed at a potential mining operation. There are however further difficulties to overcome when proposing the use of such program to estimate the mineral composition of waste rock at a mining operation. The reliance of current modal mineral estimation methods on oxide values and LOI data for the estimation of olivine and serpentine in particular means that XRF and LOI analyses must be completed. As proposed in Chapter 1 in this dissertation, the principal outcome is to develop a parameter that can estimate the MCP of an ultramafic deposit for the mining industry. This aims to maximise existing data and is both practical and affordable. Although XRF analyses are significantly cheaper than XRD with the required Rietveld refinement analysis, major oxide data is often not available from rock samples tested for resource estimation purposes at mining operations. This coupled with the time consuming nature of adjusting each sample based on rock type and manually entering sample data on an individual basis means that it will be difficult to utilise a modal mineral estimation program such as MINSQ to thousands of mining waste rock samples. A new Microsoft ExcelTM based series of spreadsheets were developed by the author in this research and is outlined and discussed in Chapter 5. It incorporates the principals of the CIPW subtractive modal mineral principal and the practical MINSQ spreadsheet format to estimate MCP values from elemental geochemical samples. Although specifically designed around the mineral assemblage observed at Turnagain, it is a template by which the modal mineral estimation of ultramafic deposits from lithogeochemical data can be approached. It utilises existing data that has already been financially

77 accounted for, incorporates mineral carbonation capabilities by a selected process and permits multiple sample entry.

4 Experimental direct aqueous mineral carbonation

This chapter provides the results obtained from the laboratory experimental mineral carbonation tests and analytical procedures undertaken on both samples from the Twin Sisters and the Turnagain ultramafic complexes. The laboratory based experimentation undertaken in this research targeted a novel series of tests on rocks with a variable mineralogy typical of those found at mining operation in mafic and ultramafic rock complexes. The first phase of experimentation on the Twin Sisters ultramafic rocks were completed to generate a benchmark suite of results that was achievable using the experimental setup designed by the author at the University of British Columbia. The second phase of experimentation on the Turnagain ultramafic rocks provide the findings on the mineral carbonation efficiency of mining waste rock from the deposit. This also provides the data utilised in developing the mineral carbonation parameter and MCP calculator.

4.1 Twin Sisters ultramafic rocks

Bench-scale experimental direct aqueous mineral carbonation experiments were undertaken on 15 olivine samples from the Twin Sisters ultramafic complex. The 15 samples were tested to determine the stoichiometric conversion of forsterite (Mg2SiO4) to magnesite (MgCO3) thus sequestering carbon dioxide using the laboratory methodology described in section 3.5. The samples were split from a 3 kg bulk rock sample. The experimental direct aqueous mineral carbonation conditions for each test are displayed in Table 4.1. The sample numbers in this research were incremented from TSO-AJ1 to TSO- AJ15 for the 15 samples.

Table 4.1 Experimental direct aqueous mineral carbonation conditions for Twin Sister rock samples tested

Sample ID Reactant Weight Carrier Solution Chemistry Test Duration Mean Pressure Mean Temperature Mean Stirring (g) (45 ml) (hrs) (bar) (°C) Speed (rpm)

TSO-AJ1 4.5 0.64 M NaHCO3/1 M NaCl 3 58 185 1427 TSO-AJ2 4.5 0.64 M NaHCO3/1 M NaCl 6 63 185 1437 TSO-AJ3 9.0 0.64 M NaHCO3/1 M NaCl 6 65 185 1431 TSO-AJ4 4.5 2.5 M NaHCO3/1 M NaCl 6 63 185 1444 TSO-AJ5 4.5 2.5 M NaHCO3/1 M NaCl 6 66 185 1439 TSO-AJ6 4.5 0.64 M NaHCO3/1 M NaCl 2 66 185 1417 TSO-AJ7 9.0 0.64 M NaHCO3/1 M NaCl 2 64 185 1432 TSO-AJ8 4.5 0.64 M NaHCO3/1 M NaCl 1 63 185 1417 TSO-AJ9 4.5 0.64 M NaHCO3/1 M NaCl 3 66 185 1441 TSO-AJ10 4.5 0.64 M NaHCO3/1 M NaCl 6 67 185 1445 TSO-AJ11 4.5 0.64 M NaHCO3/1 M NaCl 4 65 185 1440 TSO-AJ12 4.5 0.64 M NaHCO3/1 M NaCl 2 64 185 1442 TSO-AJ13 6.75 0.64 M NaHCO3/1 M NaCl 4 68 185 1440 TSO-AJ14 6.75 Distilled Water 2 67 185 1433 TSO-AJ15 6.75 0.64 M NaHCO3/1 M NaCl 2 67 185 1438

4.1.1 Solid reactant particle size analysis and morphology Particle size analysis was completed using a Malvern Mastersizer Hydro 2000S of two riffle split samples from the Twin Sisters olivine that is used as the reactant material for experimental mineral carbonation. A volume weighted mean particle size of 23.1 μm was determined. The particle size distribution was analysed to determine the percentage of material that passed through each mesh number (Figure 4.1). 88.4 % of the material passed through a 400-mesh screen (-38 μm). The size of the substrate material used in the experimental mineral carbonation is comparable to the 80 % required to classify the material under 38 μm (stage 2 activation) used in mineral carbonation experiments undertaken at the Albany Research Center (O’Connor et al., 2005).

Figure 4.1 Twin Sisters reactant particle size analysis

100.0 99.9 100.0 99.1 97.1

95.0 93.4

90.0 88.4

85.0 81.6 80.0

75.0 VolumeBelow Mesh(%)

70.0

65.0

60.0 100 170 200 230 270 325 400 Screen Mesh Number The volume of reactant material (volume percentage) from the Twin Sisters ultramafic complex classified by the following screen mesh numbers; 100 (-150 μm), 170 (-90 μm), 200 (-75 μm), 230 (-63 μm), 270 (-53 μm), 325 (-45 μm), 400 (-38 μm).

Scanning electron microscopy was undertaken on prepared Twin Sisters substrate solid samples used in experimental mineral carbonation. The mineral pre-treatment undertaken i.e. crushing and grinding resulted in size reduction, however there is little morphological change associated post pre- treatment (Figure 4.2). The shape of crystals remains fairly angular with defined forsterite crystals.

Following particle size analysis, it was determined that greater than 80 % of the material is smaller than 37 µm. The SEM images show the size of the forsterite crystals vary greatly below 75 µm. The effect of varying comminution techniques is beyond the scope of this research. It is unclear on the effect of the varying crystal size on the stoichiometric forsterite to magnesite conversion extents and whether the attrition method chosen is the most effective for use in mineral carbonation.

Figure 4.2 SEM image of Twin Sisters reactant olivine post-comminution

SEM (BSE) image of reactant olivine crystals from the Twin Sister ultramafic complex. Almost all of the crystals are smaller than the targeted 75 µm size from comminution. The sharp angular crystals vary greatly in size from 1 µm to 70 µm.

4.1.2 Solid reactant geochemistry and mineralogy Two representative samples of the Twin Sisters substrate material were sent for XRD analysis. A quantitative estimate of the mineralogy was made from the diffractograms (Figure 4.3) using Rietveld refinement. It was determined that the primary mineral phase present is forsterite (olivine), Mg2SiO4

(97.9 wt.%). Minor quartz, SiO2 (0.4 wt.%); lizardite (serpentine), Mg3Si2O5(OH)4 (0.2 wt.%); and 2+ chromite, Fe Cr2O4 (1.5 wt.%) phases were also identified. A representative sample sent for XRF analysis (Table 4.2) provided the major oxide percentages present in the substrate material. MgO (50.9 wt.%) and SiO2 (40.7 wt.%) comprise the major oxides present. Minor quantities of Fe2O3 (8.4 wt.%) and Cr2O3

(0.7 wt.%) are also present. This is comparable to the mineral constituent chemistry determined from XRD analysis.

Table 4.2 Twin Sisters reactant XRF and LOI analysis results

Analyte SiO2 Al2O3 Fe2O3 CaO MgO Na2O K2O MnO TiO2 P2O5 Cr2O3 LOI TOT/C TOT/S (%) TSOa 40.7 0.15 8.39 0.09 50.94 n/db n/d 0.12 0.02 n/d 0.677 0 n/d 0.06 Reactant

Major oxide content from X-Ray fluorescence and loss on ignition (LOI) results for the Twin Sisters olivine reactant material. a Twin Sisters olivine b n/d denotes that no oxides were measured under the detectible limit (0.01 %)

Figure 4.3 A Rietveld refinement plot of the Twin Sisters olivine reactant material

Lizardite 0.16 % 14,000 Chromite 1.54 % Forsterite 97.86 % 12,000 Quartz 0.44 %

10,000

8,000 Counts 6,000

4,000

2,000

-2,000 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50 52 54 56 58 60 62 64 66 68 70 72 74 76 2Th Degrees

The blue line is the observed intensity at each step; the red line is the calculated pattern; the solid grey line below is the difference between the observed and calculated intensities; the vertical bars are the positions of all the Bragg reflections). The coloured lines are the individual diffraction patterns of all phases. These refinement plots were provided post-analysis by the Electron Microbeam/X-ray Diffraction Facility Department of Earth and Ocean Sciences, UBC.

4.1.3 Bench-scale experimental results The principal aim of the bench-scale testing was to fine tune the experimental procedure. This established a consistent experimental methodology for undertaking direct aqueous mineral carbonation tests on rocks from Turnagain. The results are compared with those provided in the literature by researchers that have run similar experiments using the same substrate reactant material. The first two samples tested (TSO-AJ1 and TSO-AJ2) used a pressure filtration system for solid product recovery. The system was unable to recover a large proportion of the ultra-fine material (-37 µm). As a result the solid content was insufficient for XRD with Rietveld refinement analysis as the product (less than 1.5 g) had to be smear mounted. The results were considered semi-quantitative and as a result have not been produced in this research. Following these samples a solid recovery method was employed. In this method approximately 80 % of the solution was removed from the slurry and the remaining product was allowed to dry. As the post-carbonation buffered solution recovery efficiency is not considered pertinent in this research, the weight percentage of the minerals detected in the product solid content are normalised to exclude any product minerals that may have been included from the

NaHCO3/NaCl content of the buffered solution.

Samples TSO-AJ4 and TSO-AJ5 were both tested for 6 hours at 185 ± 2°C and a mean CO2 partial pressure (PCO2) of 65 ± 5 bar. These samples were tested using a 2.5M NaHCO3/1M NaCl buffered solution. The increased NaHCO3 content was tested to evaluate the effect on the carbonation extent achieved using the bench-scale experimental setup. This was based on the increased mineral carbonation extents achieved by Chizmeshya et al. (2006; 2007) using the same buffered solution conditions in which they achieve a 32 % carbonation extent at 68 bar in one hour. Both TSO-AJ4 and TSO-AJ5 showed an increased mineral carbonation extent with a normalised magnesite content of 84.4 wt.% and 74.8 wt.% respectively. This is compared to a normalised 60.1 wt.% magnesite content under the same experimental conditions using a 0.64M NaHCO3/1M NaCl buffered solution. The stoichiometric forsterite to magnesite conversion extent is consistent with those determined by Chizmeshya et al. (2006; 2007). This demonstrates that increasing the bicarbonate concentration of the solution chemistry can increase the mineral carbonation extent at lower CO2 partial pressures. However, using the increased NaHCO3 content resulted in increased weight percentages of minerals derived from the solution chemistry within the solid product. These mineral were identified and quantified using XRD with Rietveld refinement analysis and include; nahcolite (NaHCO3), thermonatrite (Na2CO3.H2O), trona

(Na3H(CO3)2(H2O)2), and halite (NaCl) (Table 4.3). It was felt that the results produced did not provide an accurate representation of the stoichiometric conversion of forsterite to magnesite. As such a buffered

84 solution chemistry of 0.64 M NaHCO3/1 M NaCl was selected for the rest of experimental mineral carbonation testing. No greater than 9.3 wt.% of the solid product in any of the samples tested comprised of minerals precipitated from the buffered solution.

Table 4.3 Mineral content in XRD analyses derived from the buffered solution post-carbonation

Phase TSO-AJ3 TSO-AJ4 TSO-AJ5 Halite 4.4 2.6 3.8 Nahcolite 0.8 15.3 25.7 Thermonatrite 1.5 n/da 1.3 Trona 2.6 11.4 6.9 Total wt.% (Rietveld) 9.3 34.3 37.7

The weight percentage mineral content in the solid product material derived from the buffered solution during experimental mineral carbonation experiments. TSO-AJ3 used a buffered solution of 0.64M NaHCO3/ 1M NaCl. TSO-AJ4 and TSO-AJ5 both used a buffered solution of 2.5M NaHCO3/1M NaCl. All three samples were tested for 6 hours at 185 ± 2°C, 65 ± 5 bar pCO2, stirrer speed 1450 ± 50 rpm in a 100 ml Hastelloy-C Parr bench-top stirred reactor. a n/d denotes that no mineral content was determined

Sample TSO-AJ14 was tested on slurry comprised of just Twin Sisters olivine reactant material in distilled water. The test was run for two hours using the standard 185°C, PCO2 65 ± 5 bar, 1450 ± 50 rpm operating conditions. A stoichiometric forsterite to magnesite conversion of only 0.9 % was determined. This supports the use of a buffered solution in the reactant slurry to improve carbonation extents. The remaining 10 samples TSO-AJ3 and TSO-AJ6 to TSO-AJ13 and TSO-AJ15 were all tested using the same operating conditions using the 0.64M NaHCO3/1M NaCl buffered solution. These samples were tested over varying time periods from 1 to 6 hours. The normalised weight percentage mineral abundance from the Rietveld refinement analysis for all the tested solid products are given in Table 4.4.

The extent of reaction (Rx) is determined for all of the samples tested (Table 4.5). The extent of reaction is a function of the stoichiometry of equation 2.1. That is, all the magnesite calculated from the Rietveld refinement is assumed to have been converted from the forsterite content of the reactant material through the addition of gaseous CO2. A steady increase in the Rx is achieved as expected by increasing the experiment duration. The mean Rx increases from 23.9 % at 1 hour to 53.8 % after a 6 hour run time. The results indicated do not account for any mineral carbonation that may have occurred during the system warm up or cool down periods. The mean reaction extents over time for the Twin

Sisters olivine samples are shown in Figure 4.4. The estimated error is based on the largest Rx variability determined between experiments tested for the same duration under equal reaction conditions.

Figure 4.4 Normalised mineral carbonation reaction extent of Twin Sisters olivine by varying reaction time

ReactionExtent (%) 20

0 0 1 2 3 4 5 6 Experiment Duration (Hours)

Normalised reaction extent (Rx) of the Twin Sisters olivine based on the stoichiometric completion of the reaction: Mg2SiO4 + 2CO2  2MgCO3 + SiO2 at varying experiment durations between 1 and 6 hours using a 0.64M NaHCO3/1M NaCl buffered solution at 185 ± 2°C, 65 ± 5 bar pCO2, stirrer speed 1450 ± 50 rpm in a 100 ml Hastelloy-C Parr bench-top stirred reactor. Standard deviation error bars are provided which are determined from the samples processed for two hours..

Table 4.4 Post mineral carbonation mineral abundances using XRD with Rietveld refinement of the Twin Sisters olivine

Phase Reactant TSO-AJ3 TSO-AJ6 TSO-AJ7 TSO-AJ8 TSO-AJ9 TSO-AJ10 TSO-AJ11 TSO-AJ12 TSO-AJ13 TSO-AJ14 TSO-AJ15 Forsterite 97.9 37.4 54.4 62.8 74.5 48.6 50.9 46.4 65.0 67.3 96.3 70.7 Magnesite n/da 60.1 44.5 35.6 23.4 49.4 47.4 51.7 32.6 30.9 00.6 27.4 Lizardite + 00.2 01.7 00.4 00.6 00.6 00.8 00.6 00.7 00.9 00.7 00.3 00.1 amorphousb Quartz 00.4 00.4 00.4 00.4 00.5 00.4 00.5 00.5 00.3 00.1 00.4 00.4 Enstatite n/d n/d n/d n/d n/d n/d n/d n/d n/d n/d 1.1 Brucite n/d n/d n/d n/d n/d 00.3 n/d n/d 00.4 n/d n/d n/d Chromite 01.5 00.4 00.3 00.6 01.0 00.5 00.6 00.7 00.8 01.0 01.3 01.4 Total % 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 (Rietveld)

The results are renormalised to exclude minerals postulated to have formed from the 0.64M NaHCO3/1M NaCl content of the buffered solution. a n/d denotes that no mineral content was determined b Lizardite content is estimated based on the degree of crystallinity and includes any amorphous material in the solid product

Table 4.5 Normalised stoichiometric forsterite to magnesite conversion extents achieved over varying test durations between 1 and 6 hours

TSO-AJ3 TSO-AJ6 TSO-AJ7 TSO-AJ8 TSO-AJ9 TSO-AJ10 TSO-AJ11 TSO-AJ12 TSO-AJ13 TSO-AJ15 Test duration (hours) 6 2 2 1 3 6 4 2 4 2 a Rx 61.6 45.0 36.2 23.9 50.4 48.2 52.7 33.4 31.5 27.9

All samples were tested using a 0.64M NaHCO3/1M NaCl buffered solution at 185 ± 2°C, 65 ± 5 bar PCO2, stirrer speed 1450 ± 50 rpm in a 100 ml Hastelloy-C Parr bench-top stirred reactor. a The renormalised extent of reaction Rx is determined based on the stoichiometric completion of the reaction: Mg2SiO4 + 2CO2  2MgCO3 + SiO2

Increased lizardite and amorphous contents in all of the reaction products shown in Table 4.4 is suggested to be a result of increased amorphous silica or nano-scale magnesite formation as identified in experiments run by McKelvy et al., 2006. These are products of the stoichiometry of equation 2.1 rather than an increase in lizardite mineral quantity. The negligible variation in quartz content within the product samples compared to the reactant material supports the theory first proposed by O’Conner et al., 2001 which was later supported by McKelvy et al., 2006 that the forsterite crystals in the solid

2+ product become increasingly silica rich as Mg is removed forming magnesite and SiO2 is not physically precipitated.

Increasing the partial pressure of CO2 has been shown in the literature to increase the stoichiometric conversion of forsterite to magnesite (O’Connor et al., 2005; Gerdemann et al., 2007). Replicating the experimental conditions used at the Albany Research Center of 185°C and a buffered solution chemistry of 0.64M NaHCO3/1M NaCl in a stirred bench reactor; the carbonation extent achieved in this research in one hour is 23.9 %. This carbonation extent was achieved using the CO2 partial pressure of 65 ± 5 bar possible without the aid of a gas booster pump. The reaction extent achieved is compared with the experimental results achieved at the Albany

Research Center, through increasing the CO2 partial pressure, compiled from results published in O’Connor et al., 2005 (Figure 4.5). The mean stoichiometric forsterite to magnesite achieved in this research is approximately 8 % lower than that extrapolated from O’Connor et al., 2005 using a comparable PCO2. The standard deviation error bars of both sets of experiments overlap. The extents produced in this research can be considered, with respect to all of the mineral carbonation extents achieved in this research to be an underestimate of what can be achieved using direct aqueous mineral carbonation based on the results published from ARC. The range of the standard from both the work done by O’Connor et al. (2005) and the samples processed in this research emphasise the variability in carbonation efficiency using the same reaction conditions. It is unclear whether this is a direct result of the direct aqueous mineral carbonation process or the experimental procedure undertaken. The successful carbonation at pressures obtained without the use of a gas booster pump, although significantly lower than the maximum Rx achieved at ARC, may still have some important implications into the future design of an applicable mineral carbonation process design at a potential mining operation. Although further testing is beyond the scope of this research, the topic is discussed further in Chapter 7 and Chapter 8.

Figure 4.5 Reaction extents achieved using Twin Sisters olivine reactant in one hour at varying PCO2

20 ReactionExtent (%)

0 0 50 100 150 200 250

PCO2 (Bar)

The diamond series represents the extents published by O’Connor et al., 2005, the error bars display the standard deviation of -75 µm samples tested for 1 hour at 185°C, 150 bar PCO2, using the 0.64M NaHCO3 1 M NaCl buffered solution. The square marker represents the carbonation extent achieved using the same experimental conditions at the maximum achievable CO2 partial pressure without the aid of a gas booster pump (65 ± 5 bar PCO2). The error bar represents the standard deviation of all samples tested under these conditions from Twin Sisters in this research.

4.1.4 Product characterisation Scanning electron microscopy was undertaken on the post-mineral carbonation solid products. In comparison to the SEM images of the Twin Sisters reactant material (Figure 4.6), the solid product is increasingly fine grained. Forsterite crystals are still abundant as expected from the XRD with Rietveld refinement analysis. The forsterite crystals however often display a cracked and honeycomb texture (Figure 4.6 (b) and (c)). The SEM image in Figure 4.6 (c) shows that almost complete rhombohedral magnesite grains are formed during the carbonation process. These predominantly form a fine grained mass smaller than 10 µm in diameter.

Figure 4.6 SEM images of the Twin Sisters reactant and post-mineral carbonation solid product

(a) Twin Sisters forsterite solid reactant. (b) Image showing the post-carbonation solid product of TSO-AJ6 with some remaining cracked forsterite crystals with a honeycomb texture and a fine-grained mass of magnesite grains. (c) TSO-AJ6 solid product under increased magnification showing the developed rhombohedra shape of the magnesite grains typically smaller than 10 µm in diameter.

The scanning electron microscopy findings are consistent with those of O’Connor et al., 2001, McKelvy et al., 2006, and Chizmeshya et al., 2007, who favour both incongruent magnesium dissolution and silica precipitation forming silica-rich passivating layer on forsterite crystals. The honeycomb texture of the forsterite in the post-carbonation solid product supports the theory of McKelvy et al., 2006 that biaxial tensile stresses can lead to cracks in the passivating layer permitting fresh forsterite surfaces to become exposed for further reaction to take place. It is postulated that the magnetic stirrer within the autoclave reaction vessel was effective in agitating the slurry to further enhance the creation of fresh reaction surfaces. However, some material found directly underneath the stirrer post-carbonation could explain the occasional reduced reaction extents. It is assumed that increased particle to particle attrition and particle abrasion against the reaction vessel walls could apply additional stresses to particles aiding in reduction or removal of the silica-rich passivating layer. Although not directly associated with the modelling of mineral carbonation potential data and the principal focus of this dissertation. Findings although not novel on the controlling factors and outcomes derived from experimental mineral carbonation in this research both validate the experimental equipment setup and process method undertaken in this research. The results are consistent with preceding studies undertaken on the Twin Sisters ultramafic rocks. This subsequently supports the potential for ultramafic rocks to sequester anthropogenic CO2 by direct-aqueous mineral carbonation.

4.2 Turnagain ultramafic rocks

Bench-scale experimental direct aqueous mineral carbonation experiments were undertaken on 16 split drill core samples. The 16 samples tested determined the stoichiometric conversion of forsterite

(Mg2SiO4) to magnesite (MgCO3) using the laboratory method described in section 3.5. The samples tested are compared with the stoichiometric forsterite to magnesite extent achieved using substrate material tested from Twin Sisters. The same experimental laboratory setup and reaction conditions: 185

± 2°C at 65 ± 5 bar PCO2, with a constant stirrer speed of 1450 ± 50 rpm was used. Bench-scale experiments used 6.75 g of solid reactant material in a 0.64 M NaHCO3/1 M NaCl buffered solution. The solution was added to a 100 ml Hastelloy-C Parr bench-top stirred reactor for a 2 hour testing period.

4.2.1 Solid reactant particle size analysis and morphology Particle size analysis was undertaken using a Malvern Mastersizer Hydro 2000S on two representative samples from each of the tested reactant samples. A volume weighted mean particle size

91 ranging between of 21.9 μm and 26.2 μm was determined. The particle size distribution quantifies the percentage of material that passed through each mesh number (Figure 4.7). Greater than 80 % of the material from each sample tested passed through a 400-mesh screen (-38 μm). The substrate material size used in the experimental mineral carbonation is comparable to the 80 % required to classify the material under 38 μm (Stage 2 activation). This material pre-treatment size was a standard in mineral carbonation experiments undertaken at the Albany Research Center (O’Connor et al., 2005: pg.3).

Figure 4.7 Classified reactant particle size distribution from the Turnagain reactant samples tested

100 04-24-AJ3 98 04-25-AJ6 06-110-AJ2 96

06-110-AJ6 94 06-110-AJ10 06-110-AJ13 92 06-110-AJ17 90 06-111-AJ1 88 06-111-AJ4 06-111-AJ7

86 06-111-AJ10 Volume Below Mesh (%) Mesh Below Volume 84 06-111-AJ13 06-116-AJ7 82 06-116-AJ9 80 06-116-AJ11 50 100 150 200 250 300 350 400 06-116-AJ12 Screen Mesh Number

The volume percentage of material passing through the following screen mesh numbers are displayed; 100 (- 150 μm), 170 (-90 μm), 200 (-75 μm), 230 (-63 μm), 270 (-53 μm), 325 (-45 μm), 400 (-38 μm).

Scanning electron microscopy was undertaken to view the particle morphology before experimental mineral carbonation. The samples, largely forsterite display angular forsterite crystals up to approximately 75 µm in diameter. The majority of the samples comprised of a matrix of angular particles less than 75 µm in diameter. The matrix is predominantly forsteritic in composition with variable quantities of serpentine or spinel group minerals. A typical SEM image of reactant material tested from the Turnagain is shown in Figure 4.8. SEM imaging was undertaken to qualitatively observe any morphological changes that occur as a consequence of mineral carbonation. This analysis helps to pinpoint any abnormalities in the carbonation extent results that may be observed.

The mineral preparation method was successful in reducing the mean particle size but did not greatly affect the particle morphology. This may be significant in the stoichiometric conversion of forsterite to magnesite calculated. Increasing the reactive surface area of forsterite by optimizing mineral pre-treatment is suggested to improve mineral carbonation extents as postulated by Kleiv and Thornhill, 2006; Baláž et al. 2008; and Fabian et al., 2010. Optimising mineral pre-treatment processes was considered beyond the scope of this research but is part of ongoing work by Jie and Hitch as part of the mineral carbonation research group at UBC and remains the focus of many studies in the literature.

Figure 4.8 SEM image of a Turnagain dunite reactant sample used in experimental mineral carbonation

Scanning electron microscopy image (reactant sample 06-111-AJ10) of the particle size distribution and morphology typically observed in substrate reactant material used in experimental direct aqueous mineral carbonation from the Turnagain ultramafic complex, northern British Columbia.

4.2.2 Solid reactant geochemistry and mineralogy All Turnagain samples tested were sent for XRD and a quantitative estimate of the mineral makeup was made using Rietveld refinement by the Electron Microbeam/X-ray Diffraction Facility in the Department of Earth, Ocean, and Atmospheric Sciences, University of British Columbia. The kaolinite- serpentine group minerals in these samples are structurally disordered and of an unknown polytype. The serpentine peaks are fitted using the Pawley space group and unit cell parameters (Pawley, 1981) of

Lizardite-2H supplemented with peak functions. The technique was developed by Wilson et al. (2006) for quantifying the serpentine content of mine tailings. As this method is independent of atomic parameters, no quantitative mass measurement is possible. The fitted serpentine pattern was considered as an amorphous phase. Its contribution to the pattern was back-calculated from the measured degree of crystallinity. The peaks of the crystalline phases were largely identifiable for the primary, secondary and minor phases (>5 wt.%) in the diffractograms. Some trace minerals with relative abundances of less than 5 % may be prone to unrepresentative quantities. This is due to the relative error being shown to increase for minor phases using the Rietveld refinement method (Raudsepp et al. 1999; Dipple et al. 2002). Although some of these minor phases may incur increased estimation errors, absolute errors are small and rarely exceed 1 % (Wilson, 2009). Some underestimation of forsterite has occurred due to incorrect compositional information. Iron is present in all the olivine samples tested. Nixon (1997) provides an olivine composition of dunite at the Turnagain complex with a range of 87 % to 96.5 % forsterite. Some of the amorphous material may not be lizardite and thus an overestimation of the lizardite abundance is possible. Within this research these phases are considered negligible to the mineral carbonation potential of the material. Little has been reported on the effectiveness of different serpentine poly-types to sequester CO2 in direct aqueous mineral carbonation. For this reason the identification of antigorite and chrysotile is not considered to influence the mineral carbonation potential of the samples. Scheel (2007) reported from extensive field study at Turnagain that lizardite is the most common serpentine mineral. He reports only minor chrysotile (near serpentine veins and shear zones) and occasional massive meshed antigorite needles may be found. The normalised quantitative mineral abundances from the XRD analysis of all the samples tested are given in Table 4.6. The samples for laboratory testing were selected based on their logged lithology by the Hard Creek Nickel Corporation geologists. The lithological naming conventions attributed to the rock types of interest at Turnagain are dunite (Du), green dunite (gDu), and wehrlite (Wh). Various prefixes such as serpentinised (Sp) are given if a section of the rock mass has undergone serpentinisation. Of the 16 samples that were used for experimental mineral carbonation 1 sample (04-25-AJ6) was logged as wehrlite, 1 sample (04-24-AJ3) as serpentinised wehrlite, 4 samples (06-111-AJ10; 06-111-AJ13; 06-116- AJ11; 06-116-AJ12) as dunite, 1 sample (06-110-AJ6) as serpentinised dunite and the remaining 9 samples (06-110-AJ2; 06-110-AJ10; 06-110-AJ13; 06-110-AJ17; 06-111-AJ1; 06-111-AJ4; 06-111-AJ7; 06- 116-AJ7; 06-116-AJ9) as green dunite.

Forsterite is the dominant mineral phase in all samples, varying between 53.5 wt.% (06-116-AJ11) and 81.8 wt.% (06-111-AJ4) respectively. The serpentine content is variable, ranging between 15.5 wt.% (06-111-AJ7) and 40.4 wt.% (06-116-AJ11). A mean serpentinisation percentage of 22 % was recorded. This is approximately 12 % greater than the average observed degree of serpentinisation on a macro level reported from fieldwork at the Turnagain property by Scheel (2007). The measured lizardite content from XRD with Rietveld was not unexpected from hand specimen investigation of the drill core. Millimeter to centimeter scale serpentine veinlets are common and found to represent 10 % to 40 % of the rock mass. Spinel group minerals are present in all samples tested with varying quantities of both magnetite and maghemite. Magnetite is present in all samples (ranging between 1.5 wt.% and 6.2 wt.%) apart from sample 06-110-AJ10. Maghemite is present in approximately half of the samples (ranging between 1.6 wt.% and 3.3 wt.%). No correlation is observed with the major mineral phase quantities and spinel quantities or the magnetite: maghemite ratio. Diopside is present as a major phase in two of the samples tested; 04-25-AJ6 (wehrlite sample) with 20.8 wt.% and 06-110-AJ17 (dunite sample) with 14 wt.%, which would suggest that this logged dunite rock is closer to wehrlite based on the XRD analysis. The low diopside content of 04-24-AJ3 at 4.6 wt.% for a wehrlite is explained by the high degree of serpentinisation. In the remaining dunite and green dunite sample, diopside content was low (ranging between 1.1 wt.% and 5.9 wt.%) with no diopside detected in four of the samples.

Table 4.6 Mineral abundance measured using XRD with Rietveld refinement on Turnagain reactant samples pre-carbonation

Phase 04-24- 04-25- 06-110- 06-110- 06-110- 06-110- 06-110- 06-111- 06-111- 06-111- 06-111- 06-111- 06-116- 06-116- 06-116- 06-116- AJ3 AJ6 AJ2 AJ6 AJ10 AJ13 AJ17 AJ1 AJ4 AJ7 AJ10 AJ13 AJ7 AJ9 AJ11 AJ12 Forsterite 57.1 58.9 75.5 62.6 77.3 73.2 67.4 73.9 81.8 76.3 59.9 67.2 64.4 66.9 53.5 67.6

Lizarditea 30.5 13.5 18.0 28.1 15.7 16.3 12.5 17.2 14.1 15.5 28.3 21.9 31.4 25.7 40.4 23.1

Quartz 0.3 0.3 0.6 0.4 0.3 0.6 0.3 0.2 0.2 0.4 0.2 0.3 0.2 0.2 0.3 0.2

Diopside 4.6 20.8 n/db 2.0 1.8 3.3 14.0 5.1 1.1 1.4 4.5 5.9 n/d n/d n/d n/d

Brucite 2.0 0.4 1.0 1.2 1.6 1.6 0.6 0.7 0.5 1.1 0.9 0.8 2.5 1.6 3.4 3.9

Magnetite 5.3 3.9 2.8 5.5 n/d 3.4 2.4 2.7 2.1 3.6 6.2 3.9 1.5 1.9 2.4 1.8

Maghemite n/d n/d 2.1 n/d 3.3 1.6 2.7 n/d n/d 1.7 n/d n/d n/d 1.7 n/d 1.6

Sjoegrenite 0.2 n/d n/d n/d n/d n/d 0.1 n/d n/d n/d n/d n/d n/d n/d n/d n/d

Troilite n/d 2.2 n/d 0.2 n/d n/d n/d 0.2 0.2 n/d n/d n/d n/d n/d n/d n/d

Clinochlore n/d n/d n/d n/d n/d n/d n/d n/d n/d n/d n/d n/d n/d 2.0 n/d 1.8

Total % 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 (Rietveld)

Mineral abundances of reactant samples from the Turnagain ultramafic complex used in direct aqueous mineral carbonation experiments determined by XRD with Rietveld refinement. a Lizardite content is estimated based on the degree of crystallinity and includes any amorphous material present in the samples b n/d denotes that no mineral content was determined

Minor phases identified by XRD include quartz (between 0.2 wt.% and 0.6 wt.%) and brucite (between 0.4 wt.% and 3.9 wt.%) that are present in all samples tested. Clinochlore, a chlorite group member is only found in 06-116-AJ9 and 06-116-AJ12. This is likely an alteration product of the abundant ferromagnesian minerals present in the rocks. Troillite (FeS) is the only sulphide mineral abundant enough to be quantified by XRD with Rietveld refinement in any of the samples. Sjoegrenite is only found in 2 samples which also contained elevated diopside quantities. Both of these samples are logged as wehrlite (04-25-AJ6 and 06-110-AJ17) with having less than 0.2 wt.% abundance. Sjoegrenite is a rare ferromagnesian hydroxide carbonate and was not identified in hand specimen. Its presence may be explained as an alteration product associated with the wehrlite rocks at Turnagain. In summary, the magnesium and calcium silicate mineralogy of these reactant samples (forsterite, lizardite, and diopside) make them applicable to direct aqueous mineral carbonation. In comparison to the dominantly forsteritic Twin Sisters dunite samples, the rocks have a much more variable magnesium silicate mineral source. The samples also contain significantly elevated spinel group mineral quantities and the presence of sulphide-rich minerals.

4.2.3 Bench-scale experimental results The bench-scale direct aqueous mineral carbonation experiments were completed to determine the stoichiometric magnesium silicate to magnesium carbonate conversion extent achievable using Turnagain ultramafic rocks. The 16 experiments used 6.75 g solid reactant material in a 0.64 M

NaHCO3/1 M NaCl buffered solution within the bench-top stirred reactor for a 2 hour testing period. The stirred reactor was heated to 185 ± 2°C at 65 ± 5 bar PCO2, with a constant stirrer speed of 1450 ± 50 rpm. The stoichiometric conversion extents achieved do not consider any mineral carbonation that may have occurred during heating up and/or cooling down periods of the reaction vessel. The mineralogy of the product samples tested was quantified using XRD with Rietveld refinement (Rietveld, 1969). Like the reactant material, serpentine was considered as an amorphous phase and Its contribution to the pattern was back-calculated from the measured degree of crystallinity (M. Raudsepp, Personal communication, October 2012). The normalised weight percentage mineral quantities for all the tested solid products are given in Table 4.7. The normalised weight percentage considers all mineral phases created from the solid components of the buffered solution in the slurry. Where present, the halite (NaCl), northupite

(Na3Mg(CO3)2Cl), and trona (Na3H(CO3)2•2H2O) mineral abundances have been removed and the remaining weight percentages renormalised. Halite was present in all sample products ranging from 1.6

97 wt.% (06-110-AJ13) to 4.1 wt.% (06-111-AJ10). Total solid product derived from the buffered solution ranged from 1.8 wt.% (04-24-AJ3) to 7.2 wt.% (06-111-AJ13), with a mean product content of 3.7 wt.%. Magnesite is the only mineral identified in the solid product material by XRD supplemental to the mineralogy of the reactant material. This does not consider the products derived from the buffered solution. Figure 4.9 and Figure 4.10 compare the reactant and product mineral abundances of the major mineral phases identified. Quartz, brucite, troilite, and clinochlore are present in trace amounts. They are not considered influential to the analysis. This is due to the high estimation error associated with minerals in such low quantities using XRD with Rietveld refinement. Sjoegrenite, which was present in two reactant samples, is not found in any of the product material. The spinel group minerals (magnetite and maghemite) are considered together in the analysis. Figure 4.10 illustrates the reduction in forsterite quantity in the product solid material compared with the reactant material in all samples. The net loss forsterite values plotted are effectively a mirror image of the net gain in weight percentage magnesite values plotted for most samples. In contrast, lizardite (+ amorphous material) quantities display a positive correlation between reactant and product solid contents. A slight net gain is seen throughout the majority of the samples.

The magnesite formation in the solid product samples are a result of gaseous CO2 sequestration by the stoichiometric conversion of forsterite to magnesite (Equation 2.1). This assumption is based on the near equal relative forsterite loss and magnesite gain in weight percentage within 14 of the 16 samples tested. Samples 06-116-AJ7 and 06-116-AJ11 do not show this trend. They display a net loss of lizardite (+ amorphous material. This suggests a proportion of the lizardite may contribute to the overall mineral carbonation potential of the samples tested. This contribution could be a result of the degree of error in the calculation of mineral abundances using XRD with Rietveld refinement. However, the trend displayed by the majority of the samples is enough to consider any magnesium cation contribution from lizardite is negligible in magnesite formation.

Table 4.7 The quantitative mineral abundances of product samples from the Turnagain ultramafic complex

Phase 04-24- 04-25- 06-110- 06-110- 06-110- 06-110- 06-110- 06-111- 06-111- 06-111- 06-111- 06-111- 06-116- 06-116- 06-116- 06-116- AJ3 AJ6 AJ2 AJ6 AJ10 AJ13 AJ17 AJ1 AJ4 AJ7 AJ10 AJ13 AJ7 AJ9 AJ11 AJ12 Forsterite 42.3 32.5 61.8 52.8 69.8 64.0 57.1 59.9 77.7 56.5 40.5 43.2 56.2 59.1 44.3 57.9

Lizarditea 34.5 17.2 19.0 28.7 16.2 19.0 13.7 18.6 13.4 18.9 27.4 22.5 25.3 25.4 36.0 22.6

Magnesite 14.5 26.9 11.6 12.6 8.0 7.4 11.0 13.2 6.1 16.3 21.0 22.7 14.4 11.5 13.6 14.2

Quartz 0.5 0.6 0.1 0.3 0.5 0.1 0.2 0.3 0.2 0.1 0.1 0.2 0.1 0.2 n/db 0.2

Diopside 3.7 17.7 2.0 n/d 2.0 3.2 15.3 14.3 n/d 1.9 4.8 5.0 n/d n/d n/d n/d

Brucite 0.7 0.3 0.7 0.4 0.6 0.8 0.5 0.3 n/d 0.6 0.4 0.8 0.8 0.4 1.7 0.5

Magnetite 3.8 3.6 1.6 5.2 1.0 2.0 n/d 3.3 2.4 2.4 5.6 4.3 1.2 1.7 2.2 1.6

Maghemite n/d n/d 3.2 n/d 1.8 3.5 2.1 n/d n/d 3.2 n/d 1.4 2.0 1.6 2.2 1.6

Troilite n/d 1.2 n/d n/d n/d n/d n/d n/d 0.1 n/d n/d n/d n/d n/d n/d n/d

Clinochlore n/d n/d n/d n/d n/d n/d n/d n/d n/d n/d n/d n/d n/d n/d n/d 1.5

Total % 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 (Rietveld)

The quantitative mineral abundances of product samples from the Turnagain ultramafic complex used in direct aqueous mineral carbonation experiments determined by XRD with Rietveld refinement. a Lizardite content is estimated based on the degree of crystallinity and includes any amorphous material present in the samples b n/d denotes that no mineral content was determined

Figure 4.9 Scatter plots of the mineral weight percentage of forsterite, lizardite (+ amorphous material), diopside, and spinel group minerals (magnetite and maghemite)

Forsterite Lizardite (+ amorphous material)

100.0 50.0

90.0 R² = 0.8995 80.0 R² = 0.729 40.0 70.0 30.0 60.0 50.0 20.0 40.0 10.0

30.0

Product (Weight %) Product (Weight %) 20.0 0.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 100.0 0.0 10.0 20.0 30.0 40.0 50.0 Reactant (Weight %) Reactant (Weight %)

Diopside Spinels (Magnetite + Maghemite)

25.0 10.0

R² = 0.9661 R² = 0.3433 20.0 8.0 15.0 6.0 10.0 4.0

5.0 2.0

Product (Weight %) Product (Weight %) 0.0 0.0 0.0 5.0 10.0 15.0 20.0 25.0 0.0 2.0 4.0 6.0 8.0 10.0 Reactant (Weight %) Reactant (Weight %)

Scatter plots of the mineral weight percentage of forsterite, lizardite (+ amorphous material), diopside, and spinel group minerals (magnetite and maghemite) in the solid reactant material versus the solid product material post-direct aqueous mineral carbonation testing.

100

Figure 4.10 Relative weight percentage of the major mineral phases in reactant and pot-carbonation product samples from Turnagain

30.0

20.0

10.0

0.0

-10.0

-20.0 Reactant/ProductDifference (Weight %)

-30.0 04-24- 04-25- 06-110- 06-110- 06-110- 06-110- 06-110- 06-111- 06-111- 06-111- 06-111- 06-111- 06-116- 06-116- 06-116- 06-116- AJ3 AJ6 AJ2 AJ6 AJ10 AJ13 AJ17 AJ1 AJ4 AJ7 AJ10 AJ13 AJ7 AJ9 AJ11 AJ12 Forsterite -14.8 -26.4 -13.6 -9.8 -7.5 -9.1 -10.3 -13.9 -4.1 -19.8 -19.4 -24.0 -8.2 -7.8 -9.1 -9.7 Lizardite 4.0 3.7 1.0 0.6 0.5 2.7 1.3 1.4 -0.6 3.4 -0.8 0.6 -6.1 -0.3 -4.4 -0.5 Magnesite 14.3 26.4 11.2 12.2 7.7 7.1 10.7 12.7 5.9 15.5 20.4 21.1 13.8 11.1 13.1 13.6 Spinels -1.4 -0.3 0.0 -0.3 -0.4 0.4 -3.0 0.6 0.3 0.3 -0.6 1.8 1.7 -0.2 2.0 -0.2 Diopside -1.0 -3.1 2.0 -2.0 0.3 -0.2 1.3 -0.8 -1.1 0.5 0.3 -0.9

Comparative analysis of the relative weight percentage (net loss or gain) of the major mineral phases in both the solid reactant and product material tested using direct aqueous mineral carbonation. The major mineral phases analysed are; forsterite, lizardite (+ amorphous material), magnesite, spinel group minerals (magnesite and maghemite), and diopside.

101

Figure 4.11 Scatter plots of the mineral weight percentage of forsterite versus magnesite and lizardite versus magnesite in Turnagain post-carbonation product samples

A 30

20 R² = 0.8121 15

10 Magnesite (Weight %) (Weight Magnesite 5

0 -30.0 -25.0 -20.0 -15.0 -10.0 -5.0 0.0 Forsterite (Weight %)

B 30 Magnesite (Weight %) 25

10 R² = 0.0237

0 -8.0 -6.0 -4.0 -2.0 0.0 2.0 4.0 6.0 Lizardite + Amorphous Material (Weight %)

Scatter plots of the mineral weight percentage of A) magnesite versus forsterite and B) magnesite versus lizardite and amorphous material in the solid product material post-direct aqueous mineral carbonation testing.

102

2+ The scatter plots in Figure 4.11 support the idea that the major Mg cation supplied for CO2 sequestration originated from forsterite within reactant material. There is a strong positive correlation (0.81) between the forsterite content lost from the reactant material and the magnesite content gain. In contrast the loss or gain of lizardite (+ amorphous material) shows no correlation (0.02) with the weight percentage gain of magnesite in the samples tested. This finding is consistent with the low direct aqueous mineral carbonation extents achieved by Gerdemann et al. (2004) attributed to not heat pre- treating the lizardite. The strong positive correlation between the reactant and product diopside mineral quantities in Figure 4.9 and the indicative low fluctuations in mineral abundance in Figure 4.10 suggest there is no contribution from diopside to the direct aqueous mineral carbonation potential of the rocks. Diopside was shown to be the primary source of calcium in the reactant samples tested. With no calcite identified from XRD analysis of the product solids, it can be assumed that magnesium should be considered as the only cation sequestering CO2 from Turnagain rocks using direct aqueous mineral carbonation. It is possible that some minor calcite may be present below the detection limit of XRD using the Rietveld refinement. It is also possible that some magnesium leached from diopside during the process could contribute to the overall magnesite product content. However this contribution is too minor to calculate. The spinel group minerals (both magnetite and maghemite) show little or no correlation between the reactant and product solid contents. The relative low abundances and associated increased estimation error using the XRD with Rietveld refinement method may explain the variation observed from sample to sample. From the mineral quantities plotted in Figure 4.10, there appears to be no link between increased spinel content and magnesite formation. Magnesite formed in all samples and appears to be a direct result of the stoichiometric conversion of forsterite to magnesite plus silica. Lizardite and diopside show little or no direct evidence of contributing to the mineral carbonation potential of the substrate material. Increased spinel group mineral abundance does not appear to have a direct negative effect on mineral carbonation. The effect of removing spinel group minerals before mineral carbonation by magnetic separation was beyond the scope of the research.

The percentage extent of reaction (Rx) was calculated for all samples (Table 4.8.). The Rx was calculated assuming magnesite formed as a result of the stoichiometric conversion of forsterite, sequestering the added CO2. The mean Rx for the samples tested at the Turnagain was 19.7 %. This is compared to the mean Rx of 37.6 % achieved under the same experimental conditions with the Twin

103

Sisters dunite samples. The lower Rx of the Turnagain ultramafic rocks is attributed to the lower amount of reactive forsterite available for mineral carbonation. The mean weight percentage of reactive forsterite in the samples tested from Turnagain is 67.7 wt.% compared with 97.9 wt.% in the Twins Sisters reactant material. As discussed with the Twin Sisters samples, passivating layers attributed to increased silica content on the rims of forsterite grains can inhibit mineral carbonation reactions. As such the lower overall reactive forsterite surface area for mineral carbonation to take place in the Turnagain samples tested may be a direct reason for the lower carbonation extents realised. Although, the lower reactant forsterite content may be a valid reason for the lower reaction extents achieved, it does not explain why the lowest stoichiometric forsterite to magnesite conversion of 5 % is observed in sample 06-111-AJ4. This sample tested had the greatest reactant forsterite weight abundance of 81.8 wt.%. From the XRD with Rietveld refinement analysis of solid reactants and products, there does not appear to be any significant mineralogical trends that indicate they influence the mineral carbonation extent. This extent may be an inherent variability in the direct aqueous mineral carbonation process. This could also be a result of the small sample size (6.75 g) used, variability in the particle size under 75 µm or insufficient mixing and agitation within the reaction vessel during experimentation. An interesting chemical aspect noticed from the Turnagain samples tested is the possible influence of the sulphide content in the reactant samples and their effect on the stoichiometric forsterite to magnesite conversion extent. Samples showing the greatest mineral carbonation potential had the highest sulphide content determined by XRF and ICP-ES testing. This may be a consequence of the sulphides lowering the pH of the solution chemistry. This allows more Mg2+ to be leached from forsterite and thus being able to bind with CO2 to form magnesite. However this lower pH is expected to limit the ability of magnesite to form as magnesite tends to favour more alkaline solution chemistry for precipitation. Unfortunately it was not possible to measure the pH as the experiments were run. There was no noticeable variation in the pre- and post-experiment pH that may aid in explaining the sulphide influence on mineral carbonation.

104

Table 4.8 Normalised stoichiometric forsterite to magnesite conversion extents achieved using Turnagain reactant samples

04-24- 04-25- 06-110- 06-110- 06-110- 06-110- 06-110- 06-111- 06-111- 06-111- 06-111- 06-111- 06-116- 06-116- 06-116- 06-116- AJ3 AJ6 AJ2 AJ6 AJ10 AJ13 AJ17 AJ1 AJ4 AJ7 AJ10 AJ13 AJ7 AJ9 AJ11 AJ12 Wt.% Forsterite 57.1 58.9 75.4 62.6 77.3 73.1 67.4 73.8 81.8 76.3 59.9 67.2 64.4 66.9 53.4 67.6 (reactant) Wt.% Forsterite 42.3 32.5 61.8 52.8 69.8 64.0 57.1 59.9 77.7 56.5 40.5 43.2 56.2 59.1 44.3 57.9 (product) Wt.% Magnesite 14.5 26.9 11.6 12.6 08.0 07.4 11.0 13.2 06.1 16.3 21.0 22.7 14.4 11.5 13.6 14.2 (product) a Rx (%) 26.0 44.9 18.1 15.7 9.7 12.5 15.3 18.8 5.0 26.0 32.3 35.7 12.7 11.6 17.1 14.4

Normalised stoichiometric forsterite to magnesite conversion extents (percentage) achieved using rocks from the Turnagain ultramafic complex.

All samples were tested using a 0.64 M NaHCO3/1M NaCl buffered solution at 185 ± 2°C, 65 ± 5 bar PCO2, stirrer speed 1450 ± 50 rpm in a 100 ml Hastelloy-C Parr bench-top stirred reactor for two hours. a The renormalised extent of reaction Rx is determined based on the stoichiometric completion of the reaction: Mg2SiO4 + 2CO2  2MgCO3 + SiO2

105

5 The derivation of a mineral carbonation potential parameter

The motivation for deriving a mineral carbonation potential (MCP) parameter is to more accurately determine the capability of any given rock mass to sequester anthropogenic CO2 by mineral carbonation. This is aided through the use of three-dimensional geospatial modelling. The idea of a parameter is essentially a variable number that can be attributed to the rock mass at a particular location. This variable defines the amount of anthropogenic CO2 that it can dispose by mineral carbonation. Applying fundamental estimation methods such as inverse distance weighting and ordinary kriging, techniques that are used widely by mining industry professionals to geospatially interpolate block models; these MCP values can be used to quantify the amount of CO2 that can be disposed for any mineable rock mass. This is achievable provided drill core is available for geochemical analysis at the deposit. The significant financial capital invested in to mining operations is based on the economic viability of mining a mineral commodity at the deposit. Using waste rock in a mineral carbonation operation requires a number of economic and practical decisions beyond the scope of this research. These factors include the development of an applicable industrial-scale mineral carbonation process, the economic viability of mining a sufficient quantity of substrate rock to feed a mineral carbonation plant and most importantly an economic incentive to develop a mineral carbonation operation in the first place. The principal aim of the MCP parameter developed in this research is to generate data that can be modelled from existing lithogeochemical data in three-dimensional space. Lithogeochemical data is commonly available at mining operations through the sampling and analysis of diamond drill core and can be used to determine the mineral carbonation potential and the inherent internal variability within a deposit. This is achievable due to data being available that is typically not available at deposits that aren’t exploited for minerals of economic importance. The parameter is thus targeted for exclusive use in the mining industry. An MCP calculator is introduced. This is a series of Microsoft ExcelTM spreadsheets program capable of generating an MCP value by estimating the modal mineral abundance from a set defined mineral assemblage and a series of lithogeochemical assay samples. The MCP value produced is the quantity of magnesium or calcium carbonate that can be formed using the selected mineral carbonation process. Due to varying deposit types, mineralogy and the range of mineral carbonation methods present at a research stage, generating an applicable parameter to suit all of these factors is challenging. This chapter aims to outline a framework for the development of an MCP parameter. An example of an

106 application of this MCP parameter is presented using lithogeochemical data from the Turnagain ultramafic complex, northern British Columbia, Canada. Although the calculator is specific to the mineral assemblage investigated at Turnagain, the estimation steps it follows can be adjusted to study comparable deposits.

5.1 Lithogeochemical data preparation and analysis

When evaluating an ultramafic deposit on its mineral carbonation potential there is no substitute for well-structured data collection program. This optimally includes a sampling program tailored towards a mineral carbonation investigation. This uses a range of geological, mineralogical and geochemical analytical techniques. This section looks at the types of data typically available at a mining exploration or operating stage. It proposes how existing data can be manipulated and highlights additional data requirements to make an informed estimation of the MCP of ultramafic deposits.

5.1.1 Data availability and requirements Geological mapping of ultramafic deposits either by a national geological survey or by a mining company can provide a first estimate on the volume of ultramafic rocks at a deposit. For example by multiplying the surface area of dunite at a deposit by the projected depth of the dunite an estimate of the volume of that rock mass can be made. By analysing some bulk rock samples taken during field surface mapping and assaying them for major cation oxides the RCO2 (Equation 2.8) of that rock mass can be derived. This can be used to provide a preliminary estimate of the capacity of that rock mass to sequester CO2. This initial approach is enough to decide whether the deposit warrants further investigation as a material source for a mineral carbonation operation. For a detailed investigation, lithogeochemical analyses of drill core are required. This allows the spatial variability of the geochemical properties of the rock mass to be determined. Extensive drilling and geochemical assay results are available at ultramafic deposits of economic interest to mining companies. It is recommended that this available data should be used to develop a more informed MCP estimation. The data preparation for the MCP calculator is minimal. Any lithogeochemical assay results from drill core available at the ultramafic complex can be synthesised for use in the calculator. Each interval of drill core assayed must be stored in a database and made identifiable to which drill hole it belongs. The depth interval at which each sample represents down the drill hole should be given unique sample

107 identification. The drill core sample intervals should have been logged by a geoscientist to geologically classify the rock. A representative sample of each interval should be assayed for the major cation quantities.

5.1.2 Mineralogical investigation A mineralogical investigation is an essential component of mineral carbonation potential estimation. Ultramafic deposits targeted for economic exploitation by mining or exploration companies often undergo a detailed mineralogical study of the major rock units present. This study typically comprises at the least, optical microscopy of thin sections cut from representative samples of each lithology. Often an XRD study to determine the principal mineral assemblage of each lithology is also conducted. Due to the variable success of using differing magnesium and calcium silicates in experimental mineral carbonation, estimating the proportions of magnesium and calcium silicates within each rock unit is important. The identification of the principal minerals in thin section can give a preliminary assessment of the abundance of each mineral. This can also provide an estimate of the degree of alteration present for each lithology. The degree of alteration can often be overlooked by examining the rocks without the aid of microscopy and lead to a qualitative overestimation of a particular magnesium or calcium silicate mineral. Furthermore it can cause an underestimation of alteration mineral abundance such as magnesium or calcium carbonate. This is significant as these minerals cannot be exploited as a substrate source for carbon sequestration. Identification of the major mineral phases for each lithology should be determined through optical microscopy and XRD can provide enough mineralogical information to make a baseline modal mineral estimation. Quantitative methods such as thin section point counting and XRD with Rietveld refinement can provide a more precise determination of the principal mineral quantities. Analysis on representative samples is recommended to determine a quantitative guideline of the mineral proportions for validation of estimates made on many rock samples using a modal mineral estimation tool such as the MCP calculator. Further mineralogical investigation can better aid any modal mineral estimation undertaken. Such investigative techniques include electron microprobe analysis of individual minerals. This can aid in determining their precise geochemical makeup. An example where this could be important with regards to MCP calculation is the Mg2+/Fe2+ ratio (magnesium number) of olivine minerals. This can vary significantly between rock types and provide a misrepresentation of magnesium cation quantity available for mineral carbonation from olivine.

108

5.1.3 Preliminary data analysis The Turnagain ultramafic deposit was identified for potential use in a mineral carbonation operation due to preliminary geological mapping estimating an approximate 10.8 km3 volume of ultramafic rocks. The deposit has been explored and drilled by Hard Creek Nickel Corporation to develop a resource estimate for a potential nickel sulphide mining operation. Approximately 75,620 metres (248,100 feet) of diamond drilling has been undertaken at the deposit. Drill core samples representative of 2 m to 4 m were collected for the length of each drill hole and sent for lithogeochemical analysis by ICP-ES. Typical with many potential or active mining operations, major oxide values from XRF or a similar procedure was unavailable for the sampled intervals of drill core. The 16 samples used in the experimental mineral carbonation from Turnagain in this research were analysed for major oxide abundances using XRF at the Acme Laboratories in Vancouver, BC. These representative samples were tested to examine the LOI values for each sample as a measure of H2O. LOI is a key indicator targeted for estimation of serpentine in the rock types. The major oxide analysis provides a measurement of the typical SiO2 content of the samples unavailable from ICP-ES geochemical testing. The XRF analyses were also undertaken as a validation tool on the accuracy of the major cation to oxide conversion achieved by the MCP calculator program. Optical mineralogy has been undertaken by geoscientists at the Turnagain ultramafic complex including Nixon et al., 1997 and Scheel, 2007. The major cumulate rock units at Turnagain comprise of dunite, wehrlite, hornblendite and minor dioritic phases. The predominant mineral assemblage of each of these major rock types is published in a Master’s thesis by Scheel (2007). This provided the basis for the mineral assemblage to be estimated using the MCP calculator in this research. Quantitative XRD analyses on the 16 Turnagain reactant samples used in mineral carbonation supported much of the optical mineralogy work done by Scheel (2007). A mineral assemblage of olivine, serpentine, diopside, clinopyroxene, magnetite, chromite, illmentite, and quartz was defined for modal mineral estimation. An Iron sulphide phase was identified in some samples and as such all sulphides in the rocks analysed were estimated as FeS.

Five specific rock types were targeted at the deposit based on their potential to sequester CO2 via direct aqueous mineral carbonation. These rock types were classified as green dunite, dunite, wehrlite, olivine clinopyroxenite and serpentinite. All sample intervals of rock are logged as any one of these rock types or a variant of them such as serpentinised dunite. The rock identification of each interval was extracted from the Turnagain exploration database. This information was added to a database to be specifically used for a mineral carbonation estimation investigation. The drill hole collar and survey

109 information as well as the logged geology codes and relevant ICP-ES geochemical assay results were added to the database. The rock types were selected based on the recorded high magnesium contents and expected high olivine contents. The green dunite lithology was identified as a separate lithology to dunite at the deposit by the geoscientists logging the drill core due to its bright green colouration. This rock type is evaluated as a separate lithology in this research due to the higher Mg/Fe2+ composition of the olivine crystals. This was identified from electron microprobe analysis undertaken by Scheel (2007). The MCP calculator requires the magnesium number to be identified (Equation 5.1) for each lithology to determine the proportions of magnesium and iron cation quantity that should be estimated in to olivine during the modal mineral estimation worksheet of the MCP calculator. The magnesium number is calculated using the following formula:

푀푔 푀푎푔푛푒푠푖푢푚 푛푢푚푏푒푟 = × 100 Eq. 5.1 (푀푔+ 퐹푒2+)

The assigned rock codes of each lithology studied and their average magnesium number can be found in Table 5.1. Despite no identifiable serpentine to magnesite conversion achieved from experimental mineral carbonation, rocks identified as serpentinite are analysed to test the effectiveness of the MCP calculator at estimating olivine and serpentine abundances. This ensures that if serpentine was able to sequester CO2 through a change of mineral carbonation process, this increase in carbonation capability could be incorporated in the calculated values. The olivine in the serpentinite is assumed to be serpentinised from dunite and for this reason is assigned the same magnesium number. It is assumed that all the rock lithologies have been accurately identified by the geoscientists at Hard Creek Nickel.

Table 5.1 Lithologies selected for use in the MCP calculator at the Turnagain ultramafic deposit

Lithology Rock Code Average Magnesium Number Green Dunite 1 93 Dunite 2 91 Wehrlite 3 88 Olivine Clinopyroxenite 4 85 Serpentinite 5 91

The assigned rock codes and average magnesium number applied for modal olivine estimation are given.

110

5.2 The mineral carbonation potential (MCP) calculator

The mineral carbonation potential calculator presented in this research is a series of four Microsoft ExcelTM worksheets. The calculator aims to provide a template from which the MCP of ultramafic rocks can be estimated from lithogeochemical assay data. The development of the MCP calculator was initiated by the need to generate a better estimate of the mineral carbonation potential of a rock mass. It uses abundant elemental geochemical assay dataset available at exploration and operating mine locations hosted in ultramafic rocks. The advantage of using the MCP calculator is that it considers the variability in silicate mineralogy throughout a rock mass as the source of the major cation targeted for use in mineral carbonation. This is helpful as differing silicates have been shown to have variable mineral carbonation capabilities dependent on the mineral carbonation process. The MCP calculator provides the volume of carbonate minerals that can potentially be formed for a discrete volume of substrate rock. This is calculated from the major cation weight percentage composition. The calculator converts these analyses first in to major oxides. It subsequently estimates their abundance based on a linear best fit stoichiometric conversion into a mineral assemblage suite defined by the user. The potential stoichiometric magnesium carbonate to magnesium silicate conversion achievable by that discrete volume of rock is estimated based on the conversion extent for a given mineral by experimental mineral carbonation. The principal advantage of using the MCP calculator is its ability to generate MCP values that can be used in resource estimation for any given rock mass at the ultramafic deposit. It provides a more in depth analysis of the rock mass than simply assessing the RCO2 of a large volume of rock with the assumption that the rock mass has a uniform mineralogy. It also provides an effective, fast and cheap alternative to expensive and time consuming XRF and quantitative XRD analyses. The MCP calculator provided in this research is not ubiquitous to all ultramafic deposits and requires adaptation on a deposit by deposit basis. This MCP calculator would need to be customised to account to the specific mineral assemblage local to the deposit. For example, if enstatite is included in the mineral estimation, the stoichiometric linear estimation would have to be modified to account for this additional magnesium silicate mineral. The MCP calculator takes advantage of the experimental mineral carbonation undertaken in this research by applying a carbonation extent (RX) for olivine to determine the MCP value. However the calculator can be used to aid MCP estimation without these values by estimating the modal mineral abundance alone. Both olivine and serpentine abundance in this example

111 may be modelled to generate an estimate of the quantity of each of these minerals that could available at the deposit for use in a mineral carbonation processing stream.

5.2.1 The MCP calculator Instructions for the data input required in each of the four worksheets are given in this chapter and within the calculator itself. The MCP calculator is provided in Appendix 1 of this dissertation. The instructions for data input into each worksheet and instructions for potential modifications that can be made to customise the program to a specific ultramafic deposit are further provided in Appendix 1. The first worksheet is titled the sample preparation worksheet. It provides a platform for converting measured major cation quantities in to major oxides for mineral estimation. The worksheet provides a data validation section in which the estimated major oxide values can be compared against measured major oxide values from the same samples. This provides an estimate of the degree of error expected from the conversion calculator. The second worksheet is a mineral composition calculator that allows the user to calculate the weight percent of each major oxide component of the minerals to be estimated in the MCP calculator. The mineral estimation and generation of the MCP values occurs in the MCP calculator worksheet. This active worksheet estimates the weight percent mineral abundance of a rock sample. These estimates are based on the data input to the sample data preparation worksheet and a pre-defined mineral assemblage from the user. The modal mineral abundance is determined by a linear stoichiometric series of calculations. The final MCP values are calculated by applying a stoichiometric silicate to carbonate conversion extent achievable for each silicate mineral by the selected experimental mineral carbonation process. This is an example of how measured reaction extents can be incorporated with modal mineral estimation to generate MCP values on a case by case basis. The MCP value estimates the amount of carbonate (e.g. MgCO3) that can be generated for the given volume of rock. This is a representation of the initial sample analysis input into the program. The final worksheet in the program is a multiple sample worksheet. This worksheet allows the user to enter up to 30 samples at a time. The sample data must be in the defined major cation series and each cation must be expressed as a weight percentage. The program as a whole is provided as a macro- enabled worksheet. By activating the multiple samples button in the worksheet, a macro designed to process up to 30 samples at a time is applied. This macro can be adapted to process more than 30 samples at a time and instructions for doing so can be found within the multiple samples macro

112 associated with the program. Once the macro is complete, the sample ID number and respective MCP value output in the worksheet.

5.2.2 Lithogeochemical data conversion to major oxides As discussed in Chapter 2, the requirement of any modal mineral estimation program is a suite of major oxide weight percentage abundances for a given volume of rock. The sample preparation worksheet designed by the author within the MCP calculator program provides a platform by which the measured major cation abundances can be input by the user on a sample by sample basis to convert the values into their respective oxides. The worksheet is comprised of four boxes, each of which requires some input from the person undertaking the analysis. The first box requires the input of each major cation measured during a whole rock geochemical analysis. In the current worksheet 37 major cations are permitted. The box allows each cation to be added as either a weight percent (column B) or as parts per million (ppm) in column C. All samples added as ppm will be converted to a weight percent by the worksheet. If the multiple samples macro is used, all major cation values must be converted to weight percentage abundance before activating the macro. Box 2 of the sample data preparation worksheet is an oxide and sulphide calculator. The upper part of the box converts ten of the major cation abundances entered into Box 1. The major oxides selected for conversion are Al2O3, CaO, CO2, Cr2O3, K2O, MgO, MnO, Na2O, P2O5 and TiO2. These oxides were selected due to them being the major components in many of the minerals found in ultramafic rocks. Each oxide is calculated by multiplying the cation quantity by the relevant conversion factor. The box contains a section for converting the iron cation abundance in to both FeO and Fe2O3. FeO is used in the

MCP calculator; however iron is typically expressed as Fe2O3 from XRF analysis and so was added to the worksheet for validation of the oxide calculator in Box 3.

The sulphides section of Box 2 requires the user type either FeS of FeS2 into cell H26. This choice controls the fate of any sulphide content that may be inserted into Box 1. This decision should be based on the typical sulphide minerals found at the ultramafic deposit. The iron and subsequent iron oxide weight percent abundances adjusted in this section for use in the MCP calculator. Iron sulphide has been selected as the principal source of sulphides in this research based on quantitative XRD identification of

FeS2 as the major sulphide source. The working cells in this section can be adjusted to account for other sulphides present at the ultramafic deposit being analysed.

113

The magnesium section of Box 2 controls the magnesium and iron ratios used in olivine estimation within the MCP calculator. The section allows up to 5 average magnesium number values for up to 5 rock codes. The rock code (a number between 1 and 5) entered into cell B2 uses the assigned magnesium number to determine the magnesium and iron ratio in the olivine modal estimation. The data validation box (Box 3) can be used to determine the accuracy of the estimated major oxide values in comparison to measured major oxide values from the same samples. The estimated component column is automatically filled from the data entered in to Box 1. Any measured major oxide values can be entered by the user in to column J of the worksheet. The difference between the estimated and measured components is calculated in column M. The validation column of the box reads as true, over or under dependent on whether the estimated component is within the accepted margin of error. This degree of error is set in the contained equation by the user. In this research the margin of error was accepted at 1 wt.%. The sum of the residuals squared is calculated in cell M44. Cell N44 reads valid or invalid for the sample as a whole dependent on whether the estimated component is within the accepted margin of error. The sum of the residual squares validity was set to 5 wt.% in this research. The upper section of Box 4 is the modal oxide profile expressed as a weight percentage. The sample number typed in to cell A2 by the user is automatically placed as an identifier of the profile. All the respective oxides converted from the sample data inserted in Box 1 are added to the profile. All remaining cations are grouped as either accessory or base metals with the exception of barium for use in barite mineral estimation if required. The modal oxide profile provides the data suite for each sample to be used in the MCP calculator. The profile comprises of all major oxides. The adjusted FeO content for any sulphides present determined in Box 2 and this iron sulphide weight percent. All remaining entered cations and a magnesium number determined from the rock code entered in cell B2 are added to the profile. Each sample can be copied and using the paste values function added to the sample library below. The sample library can hold up to 30 samples at any one time and are automatically entered into the MCP calculator from the sample library. This also permits the MCP calculator to run with 30 samples simultaneously.

5.2.3 The mineral composition calculator The mineral composition calculator worksheet (modified after Herrmann and Berry, 2002) allows users to determine the chemical composition of individual minerals. This helps to define the mineral assemblage for estimation. The working section of the worksheet, the mineral calculator is located in

114 the upper part of the worksheet. Here the major oxide values of each mineral can be calculated. The lower portion of the worksheet provides a mineral library where the results for each mineral can be pasted for storage before input into the MCP calculator. The major cations included in the mineral calculator for oxide conversion are those commonly found within minerals from ultramafic rocks. The name of the mineral being assessed must be typed in to cell A6 by the user. The number of each respective cation for the mineral formula must then be input by the user in row 6. The molecular weight of the mineral calculated is automatically updated in cell R5. This is calculated by summing the totals of the number of each cation multiplied by the molecular weight of each cation converted into their respective oxides. The weight percentage of each chemical component making up the mineral is generated in row 6 summing to 100%. The general formula used to calculate the weight percentage of each chemical component is as follows;

푁표.푐푎푡𝑖표푛푠 𝑖푛 푓표푟푚푢푙푎 × 100 × 푀표푙푒푐푢푙푎푟 푤푒𝑖푔ℎ푡 표푓 푐표푚푝표푛푒푛푡 Eq. 5.2 푀표푙푒푐푢푙푎푟 푤푒𝑖푔ℎ푡 표푓 푚𝑖푛푒푟푎푙

5.2.4 Operating the MCP calculator The MCP calculator worksheet calculates the modal mineral estimation and the estimated MCP value from the major cation assay values entered in to the sample data preparation worksheet. Like the sample data preparation worksheet, the MCP calculator worksheet comprises of four boxes. The boxes contain the data sources for use in the calculator and the final results generated by the calculator. There are a series of data grids that make up the working part of the worksheet. These require no input from the user unless the mineral assemblage being estimated differs from the test data. These working data grids comprise of a series of equations for modal mineral estimation following a general linear stoichiometric estimation procedure through numbered stages. The first box contains the imported modal profile of the entire sample data analysed. This requires no input from the user and the box can hold up to 30 samples simultaneously. This fixed data provides the data source for the iterative working data grids in the lower portions of the worksheet. Box 2 is designed to store any ideal modal mineral stoichiometric compositions that are to be estimated. The user can paste these mineral profiles into the box from the mineral composition worksheet. The cells are currently inactive; however these cells can be incorporated within the working data grids for mineral estimation. Box 3 contains the estimated mineral compositions in weight percentage summing to 100 %. The mineral assemblage can be customised to match that identified at the ultramafic deposit studied. The

115 box is populated from the working data grid labelled by a series of stages (S1, S2…..Sn). Each stage represents a calculated section of the linear stoichiometric estimation process. The calculations incorporated into each stage in this research are customised to estimate the mineral assemblage identified at the Turnagain ultramafic deposit. The stoichiometric linear mineral estimation stages are created by the author for use at Turnagain are summarised in Figure 5.1. Use of this MCP calculator at other deposits requires alteration of these calculation formulas within the worksheet. They would need to incorporate the mineral composition values assigned in Box 2. The following paragraphs focus on each stage of the estimation procedure proposed in this research based on the mineral assemblage at Turnagin. Required adjustments are highlighted that would be needed for customisation of the calculator. Stage one within the calculator is the normalisation of the modal profile to 100 %. This is achieved by assuming that the remaining weight percent of each sample tested is made up of SiO2 and H2O. Stages two, three and four assign the modal weight percentages from the sample data preparation worksheet. That is the sulphides (FeS or FeS2), base metal quantity, and all other accessory components not identified as a component of the major mineral assemblage. Stage two could be adapted to incorporate base metal sulphides with the renormalisation of the base metal and iron sulphide compositions should these phases be required in the modal mineral composition profile.

The TiO2 and Cr2O3 component of the major oxide profile are identified from quantitative XRD on the Turnagain samples as ilmentite (FeTiO3) and chromite (FeCr2O4) respectively. Both of these mineral abundances are estimated using all the respective oxides available in each sample. The iron content is renormalised as FeO for further estimation. If minerals other than ilmenite and chromite are observed as the major mineral phase of their respective oxides, the associated formulas within the data grid cells labelled as S5 and S6 can be adjusted.

116

Figure 5.1 Generalised sequence of stages used for modal mineral estimation of ultramafic rocks in the MCP calculator program

Stage • Silica ± water content estimation to renormalise the modal profile to 100 wt.% 1

Stage • Sulphide content as FeS or FeS2 2

Stage • Accessory mineral content (including; K2O, MnO, Na2O and P2O5 content of minerals) 3

Stage • Cumulative base metal content (including; Cu, Ni, Pb, Sn, W, and Zn) 4

Stage • TiO2 mineral phase content (e.g. ilmenite) 5

Stage • Cr2O3 mineral phase content (e.g. chromite) 6

Stage • Major Al2O3/K2O/Na2O mineral phase (e.g. plagioclase) 7

Stage • Major carbonate mineral phase (e.g. magnesite and/or calcite) 8

Stage • Major CaO mineral phase (e.g. pyroxene) 9

Stage • Major MgO mineral phase (e.g. olivine) 10

Stage • Secondary MgO mineral phase (e.g. serpentine) 11

Stage • Major FeO mineral phase (e.g. magnetite and/or hematite) 12

Stage • Remaining SiO2 ± H2O content (e.g. quartz and/or water) 13

Stage • Remaining MgO mineral balance (e.g. periclase) 14

Stage • Renormalisation to 100 % modal mineral balance 15

117

Major cation results from numerous samples taken from Turnagain displayed elevated aluminium contents (up to 20 wt.%) with respect to the generally low mean aluminium contents of the ultramafic rocks. To account for this elevated aluminium content, the weight percent anorthite (CaAl2Si2O8) content is estimated in stage seven. The silica portion of the estimate mineral is supplied by the surplus unassigned oxide value determined in stage one. Plagioclase is recognized in a large proportion of the samples. Anorthite can be substituted for a generic plagioclase mineral formula for estimation or substituted with other plagioclase end-member mineral formulas. Enstatite was identified as another plagioclase mineral at Turnagain. However estimating both enstatite (MgSiO3) and forsterite (Mg2SiO4) was not practical and thus due to forsterite being the dominant mineral, it was selected over enstatite for estimation.

The low concentration of K2O, MnO, Na2O, and P2O5 measured in the samples analysed mean that these major oxides are incorporated into the accessory minerals in stage three. They principally form minor components of the silica minerals already estimated. This is a deposit specific variable and the choice of plagioclase mineral estimation. This should be defined by the presence of the minerals from XRD or similar analytical technique. The magnesium-bearing minerals are highlighted as the principal controlling factor on estimated

MCP values. In stage eight, magnesite (MgCO3) content is estimated if CO2 content is accurately determined. If no accurate CO2 content is available it is assumed that the rock sample being estimated has no carbonates present. The working data grid can be adapted to include calcite (CaCO3) if required. The principal mineral source of calcium at Turnagain was identified in XRD analysis to be diopside

(CaMgSi2O6). This is a particularly as the dominant mineral present in wehrlite and olivine clinopyroxenite. Diopside was identified as the principal pyroxene mineral present at Turnagain and estimated in stage nine. Other pyroxene minerals could be substituted in the calculator at this stage. The diopside estimate uses all the CaO content of each sample. It extracts the associated stoichiometric

MgO and SiO2 contents to form the mineral estimate. The stage one value and the MgO weight abundances are renormalised for use in later stages of the estimation. Olivine, serpentine, magnetite/hematite, quartz and periclase are all estimated in stages ten to fourteen. These are estimated through a series of iterations that stoichiometrically best estimate the modal weight percentages based on user defined factors. As the most abundant mineral found in the Turnagain ultramafic rocks, olivine is first estimated. The calculator assumes all remaining MgO is unaccounted for from earlier stages of the estimation procedure is contained within the mineral. The magnesium number assigned to each sample within the modal profile determines the proportions of

118

MgO, FeO and SiO2 to be used in the olivine estimation. This uses a generic olivine mineral formula of

MgxFexSiO4. The olivine conversion matrix in the lower left data grid of the worksheet controls the active cells for use in olivine estimation. This is referred to within the series of iterations and can estimate olivine with any assigned magnesium number between 70 and 99.

The Initial olivine quantity is estimated by first calculating the required SiO2 and FeO content required to form olivine with the assigned magnesium number to fix all the MgO into olivine. The SiO2 +

H2O content from stage one and the FeO content is subsequently recalculated. If the SiO2 + H2O or FeO content is in deficit, the MgO concentration is renormalised to account for this deficit, as such to return zero values for SiO2 + H2O and/or FeO. If the MgO content is renormalised for a SiO2 deficit, the olivine content is assigned to S10. No serpentine is estimated in S11, the remaining FeO is assigned to magnetite/hematite in S12. The MgO content is assigned as periclase in S14 and the sample is renormalised in stage 15 to stoichiometrically balance the modal mineral composition to 100 %. This completes the modal mineral profile for that given sample. The serpentine content is estimated through a series of iterations in stage 11. Where necessary, the olivine abundance is adjusted in stage ten to appropriately balance the FeO, MgO, SiO2 + H2O contents within the estimated mineral assemblage. The general serpentine formula used for estimation was

Mg3Si2O5(OH)4. The active cells for internal calculations are determined from the serpentine conversion matrix to proportion the MgO and SiO2 + H2O components of the mineral. Serpentine abundance is first estimated to mineralogically fix any MgO released from a FeO deficit when estimating olivine. Enough

SiO2 + H2O must be available to fix the surplus MgO.

If there is a surplus SiO2 + H2O content above a user defined quantity (a 1 wt.% tolerance is defined at Turnagain based from XRD analyses) the required MgO to estimate the amount of serpentine is taken from the provisional olivine weight percent. This releases FeO from the estimated olivine quantity and assigns it to S12. The olivine weight percent is renormalised to account for the removed MgO, FeO and

SiO2 content. This process is repeated through three further sequences to iteratively balance the MgO,

FeO and SiO2. This normalises the olivine, serpentine, magnetite/hematite, and quartz ± water weight abundance following the iteration. The minimum olivine content is set to 10 % within the equation formulas at Turnagain assuming that each sample contains at least 10 wt.% olivine. This is to avoid generating negative olivine weight percentages. Further iterations could be added if samples have an extremely high serpentine content resulting if a large residual SiO2 + H2O estimated content remains after three iterations.

119

In stage 12 all the FeO content is assigned to magnetite/hematite. The MCP calculator has no method of determining the ratio of magnetite versus hematite in the samples being examined and as such the FeO content is grouped as both. The mineral profile is normalised to 100 % by assigning the remaining weight percent to silica ± water. The final estimated modal mineral composition profile is added to Box 3. If the reaction extents are not known for each mineral, the values produced in Box 3 could be used to model the mineral distribution within the deposit. The final MCP values for each sample estimated are determined in Box 4. This box is automatically completed dependent on values reported in Box 3 and values defined by the user in the mineral carbonation extents table within the instructions box. The mineral carbonation extents table defines either the maximum or average mineral carbonation reaction extent of the selected mineral carbonation process. The table can contain up to five minerals capable of supplying cations that can be bound to carbon to form carbonates. In the example provided from the Turnagain ultramafic deposit olivine, serpentine and diopside are included as potential silicate minerals that can provide magnesium and/or calcium to be bound with carbon dioxide. At Turnagain, olivine is targeted as the principal contributor to the MCP value, however this could easily be adapted to focus on serpentine. The MCP values are determined using the equation;

[(푆 푎 × 푅 )+(푆 푏 × 푅 )+(푆 푛 × 푅 )] × 퐶푂 푤푡.% 푀퐶푃 = 푚푖푛 푥 푚푖푛 푥 푚푖푛 푥 2 Eq. 5.3 100

Where, 푀퐶푃 is the volume weight percentage of magnesite that can be formed from a given volume of rock. 푆푚𝑖푛 is the weight percentage of substrate mineral available for carbonation, where;

(푊푡.% 푀𝑖푛푒푟푎푙 × 푊푡.% 푀푔푂 𝑖푛 푚𝑖푛푒푟푎푙 ) 푆 = Eq. 5.4 푚𝑖푛 100

푅푥 is the average or maximum extent of silicate to carbonate conversion achieved through experimental mineral carbonation for the given mineral. 퐶푂2 푤푡. % is the volume weight percent carbon dioxide that is required to sequester the available MgO as magnesite. The mass of CO2 that can be sequestered by a given rock mass is calculated as;

푀푎푠푠 표푓 푟표푐푘 × 푀퐶푃 푊푡.% 퐶푂 𝑖푛 푚푎푔푛푒푠𝑖푡푒 푀푎푠푠 퐶푂 푠푒푞 = × 2 Eq. 5.5 2 100 100

120

푊푡.% 퐶푂 𝑖푛 푚푎푔푛푒푠𝑖푡푒 Where, 2 = 0.522. 100

5.2.5 Reliability and limitations The incomplete modal profile imported into the MCP calculator from the sample data preparation worksheet i.e. having no information on the relative SiO2 and H2O contents of the rock samples to total 100 % is not preferential. As such the MCP values generated can only be considered a partial estimate in expressing the mineral proportions of the rock. Thus, the capacity of that rock sample to sequester CO2 by mineral carbonation can only be considered a partial estimate. Major cation data input in to the sample data preparation worksheet creates an internally consistent profile. This is achieved through conversion to major oxides using fixed conversion factors. At the sample data preparation stage of the calculator, the adjusted FeO content for iron incorporated as either FeS or FeS2 can result in lower FeO values than expected for use in mineral estimation when sulphide content is high for a given sample. If assigning the sulphur content as iron sulphides is problematic, the user can leave cell H26 on the worksheet blank. The sulphide value in the modal profile will read the input sulphur content and can be estimated in to a user defined sulphur mineral formula within stage two of the MCP calculator. The linear nature of the estimation within the MCP calculator has some limitations with regards to the number of minerals that can be estimated within the mineral profile. It must be noted that the MCP calculator is only a partial estimate and generic mineral formulas are recommended for accessory minerals to the principal magnesium and calcium bearing silicates, which are the main focus of the calculator. In the example provided in this research, one plagioclase feldspar end-member (anorthite) and one orthopyroxene (diopside) end-member are used in the estimation. These are identified as the principal sources of aluminium and calcium respectively. The importance of estimating these minerals were identified due to these minerals commonly occupying a large proportion of the SiO2 content of the rock. They also account for the aluminium and calcium present in the modal profile. As the SiO2 content is assumed from the surplus weight percentage not calculated in the modal profile it is important to realistically estimate the remaining SiO2 content for magnesium silicate mineral estimation. This becomes increasingly critical due to the magnesium to silica ratio being used as the controlling factor on the olivine and serpentine proportioning in the mineral profile. The iterative estimation of olivine in stage ten of the MCP calculator assumes all remaining weight percentage unaccounted by a major oxide is SiO2. Olivine is assumed as the predominant mineral as found in wehrlite, dunite and olivine clinopyroxenite at Turnagain. This estimation assumes an LOI value

121 of zero, close to what would be expected for fresh unaltered dunite. The magnesium number assigned to each rock sample being estimated should be an average for the rock sample estimated. This assumes the geoscientist logging the drill core has correctly logged the sample as the correct lithology. It also assumes that the olivine contained in the sample has the given iron to magnesium ratio defined by the magnesium number. Magnesium numbers however can vary greatly through a rock type. For example, the dunites at Turnagain have a typical measured range between 89 and 95 which can have a significant effect on the proportions of FeO, MgO and SiO2 used in olivine mineral estimation. On occasion this can result in a FeO or SiO2 deficit that has to be internally renormalised within the iteration equations. Mineralogically unaccounted oxide percentages following olivine mineral estimation are assumed to be SiO2 + H2O. MgO is removed from the estimated olivine weight percentage to satisfy the stoichiometric serpentine formula. This is based on the surplus SiO2 and H2O content. The SiO2 content removed from the olivine matrix is assumed as SiO2 and H2O to balance the olivine and serpentine proportioning. This iteration considered a best estimate based on the unknown silica and water content. It is internally consistent throughout the series of iterations completed for serpentine estimation. Using the data grids presented in the MCP calculator the mineral estimation in Box 3 will consistently provide modal mineral profile totaling 100 %. MCP value calculations utilising the user defined reaction extents are consistent for all samples. These can be adapted to account for improvements in the mineral carbonation extent for a given mineral. The MCP value calculation is reliant only on minerals with an assigned reaction extent. It does not consider external factors within the modal profile that may influence mineral carbonation on a sample by sample basis.

5.3 MCP data validation at the Turnagain ultramafic complex

Lithogeochemical data from the Turnagain ultramafic complex was used to test each individual component of the MCP calculator. 16 samples from the complex used in experimental mineral carbonation were chosen as the baseline samples for validation. This is due to a full suite of geochemical analyses being undertaken including ICP-ES for major cation abundance, XRF for major oxide abundance and quantitative XRD with Rietveld refinement for the major mineral makeup of each sample. The mean stoichiometric olivine to magnesite achieved from experimental carbonation on the samples was used as the average carbonation extent for olivine at the deposit. The XRD analysis of the mineral carbonation products were used to evaluate the effectiveness of the MCP calculator as a tool for estimating the MCP each rock unit studied at Turnagain.

122

5.3.1 Major cation to oxide conversion A representative quantity of the 16 samples from the Turnagain ultramafic deposit used in experimental mineral carbonation were sent for lithogeochemical analysis by ICP-ES to Acme Analytical Laboratories in Vancouver, British Columbia. The ICP-ES data was added to the sample data preparation worksheet of the MCP calculator program and a modal profile of the major oxides for each sample was generated. These oxide values were compared with measured oxide values determined by XRF from the same representative samples. The comparison was made using the data validation box (Box 3) of the sample data preparation worksheet. The validation results can be found in Table 5.2. The total absolute residual values determined for the estimated values all fall below 5 wt.% that is considered acceptable for the oxide conversion achieved by the MCP calculator program. The largest magnitude of error was that of the MgO estimation, not unexpectedly due to it comprising of nearly 50 % of each sample tested. The residual difference between the estimated and measured values is under 3 wt.%. The estimated and measured oxide values for MgO, Fe2O3 and CaO are plotted in Figure 5.2.

123

Table 5.2 Estimated major oxide values from the MCP calculator compared with those measured by XRF analysis from 16 samples tested at the Turnagain ultramafic complex

Sample ID Data Type Oxide Component (Wt.%)

Al2O3 CaO Fe2O3 K2O MgO MnO Na2O P2O5 TiO2 Total 04-24-AJ3 Est. Oxide 0.32 1.06 11.60 0.01 44.37 0.17 0.01 0.02 -a 57.56 Meas. Oxide 0.36 0.90 10.23 0.03 43.81 0.14 0.02 - 0.03 55.52 Residual -0.04 0.16 1.37 -0.02 0.56 0.03 -0.01 0.02 -0.03 2.21 04-25-AJ6 Est. Oxide 0.40 4.97 15.31 0.01 37.16 0.21 0.04 0.02 - 58.31 Meas. Oxide 0.41 4.35 13.91 0.06 38.28 0.18 0.03 0.00 0.06 57.28 Residual -0.01 0.62 1.40 -0.05 -1.12 0.03 0.01 0.02 -0.06 3.27 06-110-AJ2 Est. Oxide 0.11 0.29 10.01 0.01 43.71 0.15 0.01 0.01 0.02 54.32 Meas. Oxide 0.15 0.30 10.39 - 46.64 0.16 - - - 57.64 Residual -0.04 -0.01 -0.38 0.01 -2.93 -0.01 0.01 0.01 0.02 3.32 06-110-AJ6 Est. Oxide 0.26 0.36 11.85 0.01 46.01 0.14 0.01 0.01 0.02 58.68 Meas. Oxide 0.13 0.42 11.21 - 47.23 0.16 - - - 59.15 Residual 0.13 -0.06 0.64 0.01 -1.22 -0.02 0.01 0.01 0.02 2.10 06-110-AJ10 Est. Oxide 0.19 0.53 10.27 0.01 49.51 0.15 0.01 0.01 0.02 60.71 Meas. Oxide 0.20 0.63 9.31 0.01 48.51 0.13 - - 0.01 58.80 Residual -0.01 -0.10 0.96 0.00 1.00 0.02 0.01 0.01 0.01 2.12 06-110-AJ13 Est. Oxide 0.19 0.98 11.77 0.05 48.75 0.17 0.01 0.01 0.03 61.96 Meas. Oxide 0.21 0.91 10.53 0.03 45.51 0.16 - - 0.03 57.38 Residual -0.02 0.07 1.24 0.02 3.24 0.01 0.01 0.01 0.00 4.62 06-110-AJ17 Est. Oxide 0.21 3.15 11.91 0.01 39.79 0.19 0.05 0.01 0.05 55.38 Meas. Oxide 0.17 3.47 10.86 0.03 40.16 0.10 0.04 - - 54.83 Residual 0.04 -0.32 1.05 -0.02 -0.37 0.09 0.01 0.01 0.05 1.96 06-111-AJ1 Est. Oxide 0.40 1.05 11.24 0.01 43.64 0.18 0.01 0.05 0.05 56.63 Meas. Oxide 0.17 1.06 11.04 - 44.56 0.21 0.01 - - 57.05 Residual 0.23 -0.01 0.20 0.01 -0.92 -0.03 0.00 0.05 0.05 1.49 06-111-AJ4 Est. Oxide 0.13 0.29 11.74 0.01 46.19 0.18 0.01 0.01 0.02 58.59 Meas. Oxide 0.14 0.23 10.95 - 47.15 0.15 - - 0.04 58.66 Residual -0.01 0.06 0.79 0.01 -0.96 0.03 0.01 0.01 -0.02 1.90 06-111-AJ7 Est. Oxide 0.13 0.42 11.68 0.01 43.33 0.18 0.01 0.01 0.03 55.80 Meas. Oxide 0.16 0.45 11.54 - 45.63 0.18 - - - 57.96 Residual -0.03 -0.03 0.14 0.01 -2.30 0.00 0.01 0.01 0.03 2.57

124

Sample ID Data Type Oxide Component (Wt.%)

Al2O3 CaO Fe2O3 K2O MgO MnO Na2O P2O5 TiO2 Total 06-111-AJ10 Est. Oxide 0.28 1.12 13.37 0.01 41.34 0.18 0.01 0.01 0.02 56.34 Meas. Oxide 0.09 1.38 13.01 0.02 42.62 0.12 - - - 57.24 Residual 0.19 -0.26 0.36 -0.01 -1.28 0.06 0.01 0.01 0.02 2.21 06-111-AJ13 Est. Oxide 0.11 1.15 11.19 0.01 42.61 0.15 0.01 0.01 0.01 55.26 Meas. Oxide 0.14 1.18 10.89 - 44.87 0.16 0.06 - 0.01 57.31 Residual -0.03 -0.03 0.30 0.01 -2.26 -0.01 -0.05 0.01 0.00 2.69 06-116-AJ7 Est. Oxide 0.06 0.15 8.45 0.01 45.28 0.12 0.01 0.01 0.02 54.10 Meas. Oxide 0.09 0.16 8.07 - 48.01 0.12 0.01 - 0.03 56.38 Residual -0.03 -0.01 0.38 0.01 -2.73 0.00 0.00 0.01 -0.01 3.18 06-116-AJ9 Est. Oxide 0.21 0.18 9.55 0.01 45.05 0.13 0.01 0.01 0.02 55.16 Meas. Oxide 0.26 0.22 8.34 0.02 46.34 0.11 0.01 - - 55.30 Residual -0.05 -0.04 1.21 -0.01 -1.29 0.02 0.00 0.01 0.02 2.76 06-116-AJ11 Est. Oxide 0.09 0.15 8.64 0.01 45.71 0.13 0.01 0.01 0.02 54.78 Meas. Oxide 0.13 0.16 7.94 - 47.69 0.11 0.01 - 0.02 56.06 Residual -0.04 -0.01 0.70 0.01 -1.98 0.02 0.00 0.01 0.00 2.76 06-116-AJ12 Est. Oxide 0.19 0.21 8.48 0.01 47.01 0.13 0.01 0.01 0.02 56.05 Meas. Oxide 0.08 0.31 7.86 0.01 47.25 0.10 - - - 55.61 Residual 0.11 -0.10 0.62 0.00 -0.24 0.03 0.01 0.01 0.02 1.14

The residual values are the calculated difference between estimated and measured values. The total residual values are the sum of the squared differences observed between the estimated and measured components. a Value not measured or estimated

125

Figure 5.2 Measured MgO, Fe2O3, and CaO oxide values from XRF analysis versus estimated values by the MCP calculator at Turnagain

A comparison between measured MgO, Fe2O3, and CaO oxide values from XRF analysis with estimated oxide values from the MCP calculator program for 16 samples from the Turnagain ultramafic complex.

126

5.3.2 Reactant modal mineral estimation A representative sample from the Twin Sisters ultramafic complex and 16 samples from the Turnagain ultramafic complex were processed by the MCP calculator. The estimated modal mineral compositions by the MCP calculator are compared with those measured using quantitative XRD with Rietveld refinement. The Twin Sisters sample was tested in the MCP calculator program to evaluate olivine and serpentine estimation on what is known to be a relatively unaltered dunite sample with high olivine content. The results are provided in Table 5.3. The sample tested was assigned a magnesium number of 96 (Onyeagocha, 1978) in the calculator. The results show an excellent correlation between the measured (XRD) estimated (MCP calculator) values.

Table 5.3 Measured mineral abundance by XRD with Rietveld refinement and estimated values by the MCP calculator of a Twin Sisters olivine

Mineral XRD (Wt.%) MCP Calculator (Wt.%) Olivine 97.90 95.99 Chromite 1.50 0.97 Magnetite - 1.91 Serpentine 0.20 0.57 Quartz 0.40 0.19 Anorthite - 0.36 Diopside - 0.01 Total 100.00 100.00

Measured mineral abundance of reactant material from the Twin Sisters ultramafic complex by quantitative XRD with the Rietveld method used in experimental mineral carbonation and the estimated modal mineral composition from the MCP calculator program.

The modal mineral abundances estimated by the MCP calculator on samples tested from Turnagain are compared with the mineral abundances determined from the same drill core samples by XRD in Table 5.4. Some minor phases were identified in XRD and MCP calculator results; yet olivine, serpentine, diopside, magnetite and quartz are identified as common and major phases. Both magnetite and hematite were identified in XRD analyses, but the MCP calculator just totals the remaining FeO weight percentage after estimation. For graphical representation FeO is identified as magnetite here.

127

Table 5.4 Major mineral abundance estimated by the MCP calculator and measured using XRD with Rietveld refinement

Sample ID Mineral estimation Minerals (Wt.%) method Olivine Serpentine Diopside Magnetite Quartz 04-24-AJ3 MCP Calculator 79.81 13.89 2.91 0.54 0.74 XRD 57.12 30.53 4.63 5.27 0.33 Aa 22.69 16.64 1.72 4.73 0.41 04-25-AJ6 MCP Calculator 76.59 0.82 15.07 1.64 0.30 XRD 58.92 13.52 20.78 3.88 0.33 A 17.67 12.71 5.71 2.24 0.03 06-110-AJ2 MCP Calculator 55.12 39.63 0.77 2.35 0.97 XRD 75.43 18.00 - 4.86 0.61 A 20.31 21.63 0.77 2.51 0.36 06-110-AJ6 MCP Calculator 73.36 18.36 0.79 5.31 0.68 XRD 62.62 28.10 2.02 5.49 0.40 A 10.74 9.74 1.23 0.18 0.28 06-110-AJ10 MCP Calculator 94.51 - 1.42 2.52 - XRD 77.28 15.74 1.79 3.31 0.31 A 17.23 15.74 0.37 0.79 0.31 06-110-AJ13 MCP Calculator 91.07 - 2.82 4.18 - XRD 73.11 16.31 3.34 5.06 0.56 A 17.97 16.31 0.52 0.88 0.56 06-110-AJ17 MCP Calculator 53.50 27.67 9.61 6.74 1.03 XRD 67.40 12.48 14.03 5.10 0.33 A 13.90 15.19 4.42 1.64 0.70 06-111-AJ1 MCP Calculator 64.74 25.29 2.77 4.07 1.01 XRD 73.84 17.21 5.14 2.69 0.22 A 9.10 8.09 2.37 1.38 0.79 06-111-AJ4 MCP Calculator 79.90 13.98 0.75 3.25 0.99 XRD 81.83 14.06 1.09 2.13 0.19 A 1.93 0.08 0.34 1.12 0.80 06-111-AJ7 MCP Calculator 55.72 35.18 1.14 5.21 1.39 XRD 76.29 15.51 1.40 5.32 0.37 A 20.57 19.67 0.26 0.11 1.02 06-111-AJ10 MCP Calculator 52.40 34.30 3.14 6.77 1.36 XRD 59.92 28.27 4.47 6.25 0.20 A 7.52 6.03 1.33 0.52 1.16 06-111-AJ13 MCP Calculator 52.19 37.77 3.45 6.10 1.33 XRD 67.18 21.90 5.87 3.95 0.27 A 14.99 13.87 2.42 2.15 1.06 06-116-AJ7 MCP Calculator 57.56 37.62 0.40 2.30 1.19 XRD 64.40 31.42 - 1.45 0.21 A 6.84 6.20 0.40 0.85 0.98 06-116-AJ9 MCP Calculator 60.77 33.27 0.29 2.99 1.32 XRD 66.93 25.66 - 3.54 0.22 A 6.17 7.61 0.29 0.55 1.10 06-116-AJ11 MCP Calculator 58.70 35.17 0.36 3.46 1.31 XRD 53.41 40.42 - 2.43 0.31 A 5.30 5.25 0.36 1.33 1.00 06-116-AJ12 MCP Calculator 69.59 25.16 0.41 2.56 0.94 XRD 67.57 23.12 - 3.34 0.21 A 2.02 2.04 0.41 0.78 0.73 Estimated mineral proportions of 1 sample from the Twin Sisters ultramafic complex from 16 samples from the Turnagain ultramafic complex comparing the results calculated by the MCP calculator program and XRD with Rietveld refinement. a A is the calculated absolute difference; Where, A = MCP Calculator value (Wt.%) – XRD (Wt.%)

128

Figure 5.3 Modal mineral estimation comparison of 16 samples tested from the Turnagain

129

Modal mineral estimation comparisons of 16 samples tested from the Turnagain ultramafic complex, British Columbia using quantitative XRD with Rietveld refinement and estimated values using the MCP calculator program.

130

The modal mineral estimation abundances for olivine, serpentine, diopside and magnetite from the MCP calculator program are graphically compared to the mineral abundances determined by quantitative XRD in Figure 5.3. Scheel (2007) states that serpentinisation throughout the Turnagain ultramafic deposit is highly variable from zero to 100 %, averaging approximately 10 % based on field and drill core analysis. As can be seen from both the XRD and MCP calculator analysis, both the olivine and serpentine modal abundances vary between zero and 40 % for all samples tested. The absolute difference between the calculated XRD modal abundances and the MCP calculator estimated abundances for both olivine and serpentine can vary up to 20 wt.% from sample to sample. The mean absolute error for serpentine estimation using the MCP calculator in comparison to XRD analyses is 11 wt.%. The subsequent mean absolute error of olivine estimation is 12 wt.%. When comparing the absolute error of olivine versus serpentine between the calculated XRD and estimated MCP calculator results in Figure 5.4, the absolute difference between the olivine and serpentine estimated quantities display a strong positive correlation. That is the difference between estimated and measured olivine and serpentine are related in as much as an overestimation of olivine values by the MCP calculator results in an underestimation of serpentine and vice versa. This has an implication of the MCP of the sample tested if the mineral carbonation method selected can on sequester an olivine or serpentine substrate and not both as the degree of error on the MCP value estimation is a function of this absolute error in mineral estimation. It must be stated that the XRD analyses with which the MCP calculator results are renormalised to estimate the serpentine content based on the degree of crystallisation. This is discussed in Chapter 4 and as such the serpentine content can only be considered as a partial estimate. The serpentine content is only estimated by XRD using Rietveld refinement under the assumption that all amorphous material is serpentine. When analysing the components used in the olivine and serpentine estimation within the

MCP calculator in Figure 5.5. The total MgO, SiO2 and H2O contents estimated in the sample data preparation worksheet of the calculator positively correlate with the same components measured by XRF. Thus the absolute error in olivine and serpentine modal abundances between the MCP calculator program derived values and the calculated XRD values can be attributed to a combination of factors. These include the estimation error in the individual oxide components used for olivine and serpentine estimation at the sample data preparation stage of the MCP calculator, error in proportioning these components into the respective mineral formulas, and/or inaccurate serpentine and olivine quantities by renormalisation of the partially quantitative XRD analyses.

131

Figure 5.4 Mean absolute error of serpentine versus olivine using the MCP calculator program

Comparison of the mean absolute error (A) in weight percent between estimated serpentine and olivine modal abundance using the MCP calculator program on 16 samples analysed from the Turnagain ultramafic complex, northern British Columbia

Figure 5.5 Combined MgO, SiO2, and H2O modal abundances (wt.%) measured by XRF versus those estimated by the MCP calculator program

Combined MgO, SiO2, and H2O modal abundances (wt.%) measured by XRF compared with those estimated by the MCP calculator program from ICP-ES major cation analysis for mineral estimation on 16 samples from the Turnagain ultramafic complex, British Columbia.

132

Without an accurately determined serpentine modal abundance for each sample tested in the MCP calculator program it is difficult to truly quantify the estimation error expected the program. The guideline mean estimation errors of 11 wt.% and 12 wt.% when compared with semi-quantitative results of the same samples appear to be a reasonable expectancy. One key feature of the results is the ability of the MCP calculator to successfully determine samples that have high serpentine contents measured by XRD with Rietveld refinement by estimating a high serpentine content. This is supported by the analysis of samples tested from the Twin Sisters ultramafic complex. Here the calculator determined extremely high olivine content with an acceptable estimation error. Although some error is expected when estimating the olivine and serpentine contents of rock samples, the MCP calculator program, it is an effective tool at estimating the degree of serpentinisation throughout an ultramafic deposit from lithogeochemical data.

5.3.3 MCP estimation The sixteen samples analysed in the MCP calculator program from the Turnagain ultramafic complex are compared with the estimated MCP values projected from XRD analyses of the reactant samples. The measured magnesite contents from quantitative XRD analyses of the experimental mineral carbonation products are also compared in Table 5.5. Within the MCP calculation (Equation 5.3) olivine was the only mineral contributing MgO for mineral carbonation. This was selected due to it being the only mineral that shown to contribute to mineral carbonation during experimental mineral carbonation. The MgO weight percentage available for mineral carbonation from olivine in Equation 5.3 is dependent on the magnesium number of the estimated olivine content defined in Table 5.1. These magnesium numbers were applied in both the MCP calculator and the for the projected MCP values. This is based on the quantitative XRD values of the reactant samples. The projected extent of reaction i.e. the stoichiometric olivine to magnesite conversion was set as 37 %. This was determined from the mean carbonation extent achieved using the experimental mineral carbonation procedure and conditions on material from the Twin Sisters ultramafic deposit.

133

Table 5.5 Estimated MCP values by the MCP calculator compared with those projected from XRD with Rietveld refinement analyses

Sample ID MCP Calculator mineral results Projected MCP mineral results from XRD

Estimated MCP XRD Measured MCP Value Difference XRD Projected MCP XRD Measured MCP Value Difference Value (%) by MCP (%) Post-Carbonation Value (%) from (%) Post-Carbonation Calculator Reactant 04-24-AJ3 15.50 14.53 0.97 10.11 14.53 4.42 04-25-AJ6 14.84 26.93 12.09 10.43 26.93 16.50 06-110-AJ2 10.85 11.60 0.75 13.35 11.60 1.76 06-110-AJ6 15.40 12.64 2.76 11.97 12.64 0.67 06-110-AJ10 19.73 7.96 11.77 14.77 7.96 6.81 06-110-AJ13 19.02 7.45 11.57 13.97 7.45 6.53 06-110-AJ17 11.32 11.01 0.31 12.88 11.01 1.87 06-111-AJ1 13.24 13.15 0.08 13.69 13.15 0.54 06-111-AJ4 16.25 6.14 10.11 15.17 6.14 9.03 06-111-AJ7 11.45 16.32 4.87 14.15 16.32 2.18 06-111-AJ10 10.78 21.05 10.27 11.11 21.05 9.94 06-111-AJ13 11.07 22.74 11.67 12.84 22.74 9.90 06-116-AJ7 11.83 14.42 2.58 11.94 14.42 2.48 06-116-AJ9 12.46 11.50 0.96 12.41 11.50 0.91 06-116-AJ11 12.42 13.62 1.20 10.21 13.62 3.41 06-116-AJ12 14.65 14.17 0.48 12.92 14.17 1.25 Average 13.80 14.08 5.15 12.62 14.08 4.89

Estimated MCP values calculated using the MCP calculator program compared with those projected from quantitative XRD with Rietveld refinement of reactant material and calculated from quantitative XRD with Rietveld refinement of product material after experimental direct aqueous mineral carbonation of sixteen samples from the Turnagain ultramafic complex, British Columbia.

134

When comparing the average estimation difference in MCP values between those estimated by the MCP calculator program and those projected using the same reaction extent of 37 % from quantitative XRD, there is little difference at 5.15 % to 4.89 % respectively. Both the MCP calculator values and those projected from reactant XRD are consistent throughout typically ranging between 10 % and 20 %. The significant differences observed between estimated and measured values appear to be a function of the variability in the mineral carbonation extent. The use of the MCP parameter here is restricted by the industrial mineral carbonation processing regime. Until an established mineral carbonation process has been defined, the accurate use of MCP values is difficult to validate. A number of the samples tested, namely 04-24-AJ3, 06-110-AJ2, 06-110-AJ17, 06-111-AJ1, 06-116- AJ9, and 06-116-AJ12 all were estimated within 1 % of what was observed from experimental mineral carbonation. The relative consistency of estimated MCP values and those projected from XRD suggest and promising estimation results when compared with those observed from experimental mineral carbonation, the MCP calculator shows to be an effective tool at predicting MCP behavior. This is providing the variable nature of experimental mineral carbonation is considered.

135

6 Estimating a mineral carbonation resource for a mining project

To date, CO2 sequestration capabilities of mineral carbonation is restricted to estimating mineral carbonation capacities on a deposit scale. This is typically achieved from estimating the deposits volume from surface mapping and incorporating generalised major oxide values analysed from surface rock samples. The major deposits targeted by the US DOE in North America, analyse the mineral carbonation capacities of mafic and ultramafic deposits. They were targeted if they could potentially sequester a large proportion of the CO2 emissions on a regional scale.

Smaller scale deposits that may provide enough substrate material to sequester CO2 by industrial mineral carbonation for a single heavy industry proximal to the deposit have generally been overlooked. Niche operations could be a key target in initiating the development of an industrial mineral carbonation pilot-scale operation and advances in mineral carbonation technology. Mining operations hosted in ultramafic deposits are targeted in this research due to the typical availability of sufficient quantities of substrate minerals. Established infrastructure to develop a mineral carbonation pilot-scale plant is often already present at mining operations and the prospect of shared costs for mining and milling of the material are a significant benefit to the proposed collaboration. An MCP calculator was proposed in Chapter 5 that can be used to generate mineral carbonation potential values. That is, the volume of magnesite and or calcite that could be formed from a mineable volume of rock. The calculator program requires an input of major elemental assayed values. These values are commonly available at mining or potential mining operations. This chapter demonstrates geospatially estimated MCP values generated by the MCP calculator. They show that a more informed mineral carbonation resource estimate at ultramafic deposits can be achieved. By assigning MCP values from analysed intervals of diamond drill core a spatial approximation of the mineral carbonation potential of the rock can be developed. The MCP values substitute metal commodity values akin with a traditional mineral resource estimate. The estimation process is purposely similar to mineral resource estimation procedures typically undertaken by mining industry professionals for metal commodity estimation. This is to provide a workflow that can be undertaken by any mining operation to investigate the potential for mineral carbonation at their mine site. The mineral estimation workflow is demonstrated through a case study of the Turnagain ultramafic complex in northern British Columbia, Canada. The deposit is targeted for a potential nickel sulphide mine and has undergone a pre-feasibility study with geological, mineralogical and geochemical investigation as well as an extensive drilling program. This chapter presents how existing data available

136 from this pre-feasibility study coupled with data input into the MCP calculator and some preliminary experimental mineral carbonation laboratory testing can be used to generate a mineral carbonation resource estimate at Turnagain. The estimation procedure is outlined from the initial geological investigation through data analysis, three-dimensional data interpolation to the final estimate of the CO2 capture capacity via direct aqueous mineral carbonation. A resource estimate constrained within the projected open pit mine design is also provided. Any decisions made by the author contributing to the estimation results are defined. These explained in the context of how the decisions can impact a mineral carbonation resource estimate.

6.1 Exploratory data evaluation for MCP estimation

Geological, mineralogical and geochemical analyses and interpretations are critical to the successful quantitative estimation of the potential of a rock mass to be used in industrial mineral carbonation. Much has been published on exploratory analysis for traditional mineral commodity estimation. This estimation typically targets metals and industrial minerals. This section focuses on the application of established mineral estimation protocols in estimating the MCP of ultramafic rocks in three-dimensional space. Brief comments on factors that should be considered when estimating the potential an ultramafic deposit to sequester CO2 are given. Reasons for any decisions made by the author when undertaking the MCP estimation of the Turnagain ultramafic complex in this research are provided.

6.1.1 Geological investigation

A detailed geological investigation is a typical starting point for any mineral estimation. This should also be the starting point for undertaking an evaluation of the MCP of an ultramafic deposit. An evaluation of surface geological mapping and lithological identification are considered as essential for preliminary estimates to be generated. Supplemental three-dimensional modelling of the lithological units is highlighted as a key process to generating a more informed geospatial estimate. Surface geological mapping, provides a cheap and effective way of gaining an initial geologic understanding of a deposit without expensive sub-surface drilling. But, exposures only represent a small volumetric proportion of the ultramafic complex. In North America recorded databases and detailed surface geological maps are available for nearly all regions. These provide an accessible platform for locating initial targets for direct mineral carbonation. Goff et al. (2000) provided an initial evaluation of major ultramafic deposits in the United States and Puerto Rico. They reported the volume of potential

137 substrate rocks and major oxide analyses from various literature and database sources. By applying their

RCO2 parameter to the volumes of rock estimated they effectively provided a theoretical quantity of CO2 that could be sequestered by mineral carbonation at each deposit evaluated. Lackner et al. (2008) used Geographical information system (GIS) datasets to estimate the surface exposure of serpentine and ultramafic deposits in the United States. They coupled major oxide analyses evaluated the deposits for their capacity to sequester CO2 by mineral carbonation. In addition they coupled this research with the potential for each deposit to co-produce iron commonly associated with serpentine as well as the substrate material for mineral carbonation. This initial estimation process of linking GIS with available major oxide analyses is an important process that should be utilised to identify ultramafic deposits for mineral carbonation and more traditional commodity resources. This can attempt to overcome the significant expense associated with the prospect of a mining operation as a producer of mineral carbonation substrate material alone. The Turnagain ultramafic deposit was highlighted as a potential deposit for mineral carbonation in a study delineating ultramafic deposits using GIS for potential use in mineral carbonation (Figure 2.2) by Voormeij and Simandl (2004). The Turnagain complex was mapped by geoscientists at Hard Creek Nickel Corporation. This formed part of an exploration and pre-feasibility study for a potential nickel mining operation. The surface geology map created is shown in Chapter 3 of this dissertation (Figure 3.2) and shows over 60 % surface coverage of dunite, wehrlite and olivine clinopyroxenite rocks. Olivine is the principal mineral in each of the rocks. This initial surface geology mapping and the potential of a co- existent nickel sulphide mine was the driving selection factor in using the Turnagain ultramafic complex the case study for a more in-depth mineral carbonation resource estimate in this chapter. Detailed descriptive geological logging of drill core is undertaken at mineral deposits to systematically define the lithological boundaries. This can be defined at sub-surface depths and record the nature of any structures observed within the rock mass. Geological logging additionally comprises of mineral observations and any alteration to the host rock. This systematic geological mapping permits the lithology and associated characteristics to be recorded in a geological database. The formation of an electronic geological database and permits the data to be displayed in three dimensions down the length of surveyed drill holes within an applicable computer software program. Hard Creek Nickel Corporation supplied a geological database comprising of geological descriptions over logged depth intervals of drill core to the author from 273 logged drill holes at the Turnagain deposit. 196 of these drill holes were used for three-dimensional geological modelling by the author using the GEOVIA SurpacTM mining software program. The drill holes selected are drilled in the Horsetrail and

138

Northwest zones of the ultramafic complex (Figure 3.2). These two zones were chosen due to them hosting the principal mineralisation targeted by Hard Creek Nickel for exploitation. These zones comprise almost exclusively of dunite, wehrlite, and olivine clinopyroxenite. A proposed mine design was also provided to calculate and in-pit resource. Three major geological domains were defined for modelling at Turnagain for use in mineral carbonation estimation. The definition of these domains was based on largely uniform geological and mineralogical characteristics. The first domain is comprised of rocks logged as either intervals of dunite or green dunite. Both dunite and green dunite were grouped together in the same domain as the mineral assemblage was identical for both. The only variable characters are the colouration and forsteritic concentration of the olivine. The serpentinised variant of each rock was included in this domain as it was not possible to quantify the degree of serpentinisation based on the qualitative logging method. The second lithological domain modelled was wehrlite and the altered or serpentinised variant of wehrlite. The final domain comprised of intervals of drill core logged as olivine clinopyroxenite or the altered variant of the rock. Within the geological database, each domain was given an associated numeric rock code. The database was imported into the GEOVIA Surpac software program. The geometry of each of the rock types were interpreted on defined cross-sections and plan-sections of the drill holes. An interpreted two dimensional outline of each rock type were digitised by the author from the drill holes. The defined boundaries between each digitised domain were interpreted as sharp. In reality these boundaries can be gradational or irregular. For the purposes of a more simplistic geological modelling interpretation at this stage of investigation the boundaries were digitised as absolute boundaries. Any gradation or zonation in geology and mineralogy was left to interpretation from differing geochemical signatures of the rock at a later phase of the investigation. An approach of undertaking mineral carbonation estimation on a lithology by lithology basis over MCP value based modelling was chosen due to MCP values being a product of the rock type and mineralogy of that rock type. It is difficult to assign an economic cut-off to MCP values stating whether it is economical to mine the material as a substrate for a mineral carbonation operation. This is due to the relative infancy of the mineral carbonation technology and uncertainty of any costs associated with industrial mineral carbonation. The dunite, wehrlite, and olivine clinopyroxenite geological domains modelled in three-dimensions depicted in Figure 6.1. The genesis of an MCP resource is lithologically controlled. That is the resource is the principal major cation in the host rock and has not undergone any structurally controlled mobilisation or emplacement.

139

For this reason the continuity of the resource tends to be much more uniform within each rock type than traditional mineral estimation. Although structural modelling would still be an important process in developing a deposit wide model for mine design, it is not considered at this stage of MCP resource estimation. A detailed investigation of metamorphic and alteration features of the rock mass could show a contribution to the variability of MCP values on a lithology by lithology basis. Both structural modelling and investigation in to the metamorphic and alteration features of the rock mass could aid in determining the extent of structurally controlled serpentinisation and as such it would be important when serpentine is targeted as a substrate source. This can reduce the quantity of available magnesium cations with the natural alteration of olivine to serpentine in direct aqueous mineral carbonation. Such features are not modelled in this research due to the broad spacing of drill holes. More closely spaced drill holes are required for the accurate interpretation of thin veins from drill hole sections. The scale of the estimation supports this decision. Such structures are considered negligible on the deposit-wide resource estimation and the scale of the mining operation in question. Any three dimensional modelling undertaken requires a high degree of accuracy. This is emphasised when modelling the absolute limits of a deposit or lithological unit. Any modelling is subject to errors and is reliant on the geological interpretation made by the modeller. Smoothing errors are commonplace when interpolating a geological unit in three dimensions. This is a product of the unknown between recorded occurrences of the geological unit recorded from drill hole logging. This factor that can only be minimised through extra information gained through drilling or excavation. A degree of error must be accepted as part of the modelling process with limited information. Further errors that can be minimised and must be evaluated including the correct determination of drill hole locations and drill hole surveying, manual input errors during data capture. Care must be given to the input and interpretation of the rock units during drill hole logging.

140

Figure 6.1 Three-dimensional models of the geological domains used in MCP estimation at the Turnagain ultramafic complex

Three-dimensional models designed by the author of the geological domains used in MCP estimation at the Turnagain ultramafic complex, British Columbia, Canada. The topographical contours including a projected 28-year life surface mine design are provided.

141

6.1.2 Mineralogical investigation A mineralogical investigation should be undertaken for two principal purposes when considering a mineral carbonation resource estimate. The first of which is to define the mineral assemblage and mineral compositions to be estimated for MCP determination. The second is to characterise the continuity of mineralogy throughout the deposit. Standard qualitative mineral identification including hand specimen and thin section mineral identification is essential in determining the major minerals present in each rock type and determining and mineral associations. This mineral identification is important in determining the principal mineralogical source of the major cations to be used in mineral carbonation. It also allows the definition of target rock units for substrate material in mineral carbonation. The mineral assemblage identified can be used in the MCP calculator proposed in this research or within a similar modal mineral estimation program. Semi-quantitative and quantitative mineral analyses such as quantitative XRD and thin section point counting can be used to gain a better understanding of the mineralogical makeup of each rock type. Although not covered in depth in this research, quantitatively assessing any zonation of minerals can be estimated. This achieved by contouring major minerals throughout the deposit. This can identify areas of increased serpentinisation of dunite for example. This is important where there is a significant difference in the degree of mineral carbonation achievable between olivine and serpentine using the selected mineral carbonation process. Previous mineralogical investigations at the Turnagain ultramafic complex were undertaken by Clark (1975), Nixon (1997), and Scheel (2007). These predominantly focussed on sulphide mineralisation but subsequently providing information on host rock mineralogy. The three principal rock types targeted as substrate material for mineral carbonation at Turnagain namely; dunite, wehrlite and olivine clinopyroxenite all have the general mineral assemblage of olivine ± serpentine ± magnetite ± chromite ± clinopyroxene (diopside) ± orthopyroxene (Scheel, 2007). These all have varying proportions of each mineral, confirmed by quantitative XRD of representative samples in this research. The identification of this mineral assemblage provides the minerals that can be estimated within the MCP calculator for MCP value determination. The variable mineral quantities based on rock type and location within the deposit emphasises the need of mineral estimation when undertaking an MCP resource estimate.

142

6.1.3 Geochemical investigation A geochemical investigation provides the data required for the geospatial analysis of a specific quality of interest at a deposit. This can be a metallic quality grade at a known location, a specific element of interest or a specific quality such as at MCP based on an assayed geochemical signature and other quantitative and qualitative information. A geochemical investigation is required for MCP estimation in its simplest form to determine the concentration and variability of MgO and CaO. These are the major oxides with the capability to fix CO2 in a carbonate mineral form, throughout the deposit. In this research a geochemical investigation is required to provide a suite of elemental values at recorded locations throughout the deposit. These are used in the modal estimation of a defined mineral suite and the subsequent estimation of the quantity of carbonate minerals that can generated by an industrial mineral carbonation process at the deposit. Major elemental quantities were analysed by ICP-ES on samples tested at regular intervals from all the drill holes available at the Turnagain ultramafic deposit. These analyses were given to the author in 2009 and 11,552 sample analyses from 196 of the drill holes were taken from the database for use in this research. The samples extracted correspond to the logged intervals of drill core used in three- dimensional geological modelling. This coupled geological and geochemical data forms the basis for spatial MCP interpolation. All ICP-ES analyses results were given a unique identifier for the drill core length they represent. They were added to the geological database compiled for MCP resource estimation into the GEOVIA Surpac software program. Both the weight percentage magnesium and calcium values were converted into their respective oxide weights. These were statistically analysed before MCP estimation. The deposit wide statistical analysis results are given in Table 6.1. The 11,552 samples were considered a large enough sample size to make a detailed MCP estimate at the deposit. The sample lengths in Table 6.1 represent the regular intervals in which the drill holes were sampled. With the mean, median 10th and 90th percentiles of sample lengths falling at or near 4 metres in length, the geochemical data being analysed is statistically well supported down the length of each drill hole. The results highlight the principal oxide available for use in mineral carbonation at as MgO with a mean weight percentage of 39.71 %. CaO is a negligible oxide substrate source for mineral carbonation with a mean abundance of 2.04 wt.%.

143

Table 6.1 Statistical properties of 11,552 geochemical samples from the Turnagain ultramafic deposit

Statistical Method MgO (wt.%) CaO (wt.%) Sample Length (m) Number of samples 11,552 11,552 11,552 Minimum 1.12 0.01 0.50 Maximum 58.82 20.01 12.00 Mean 39.71 2.04 3.99 Median 41.14 1.02 4.00 Standard Deviation 6.66 2.70 0.52 Coefficient of Variation 0.17 1.32 0.13 Skewness -0.90 2.60 0.15 P10 30.07 0.13 4.00 P90 46.60 5.13 4.00

Statistical properties of 11,552 geochemical samples from the Turnagain ultramafic deposit, British Columbia, Canada. Weight percentage MgO, CaO values and the sample lengths represented are analysed.

The low coefficient of variation (0.17) and negative skewness (-0.90) of MgO values suggest an almost normal distribution. A histogram of MgO weight abundance and an overlain normal distribution curve in Figure 6.2 confirms this statement. It could be proposed that the creation of further geology specific domains is not required and the MgO values represent a uniform statistical population. However, the influence of MgO values alone on the practical mineral carbonation capabilities of the rock mass cannot be considered singularly. They must be analysed within the context of the lithology of the host rock and its associated mineralogy controlling the mineralogical location of all MgO. When considering the 10th (30.07 wt.%) and 90th percentile (46.60 wt.%) values of MgO in the deposit wide raw data there is still a broad range around the median value of 41.14 wt.% MgO. This suggests that there is still localised variability in MgO quantity. This could be attributed to variations in lithology and/or mineralogy. MgO wt.% and CaO wt.% values were extracted from the geochemical table within the geological database for the each of the defined rock codes, 1 to 5 in Table 5.1. The exploratory statistical analyses on each of the sets of these oxides was undertaken to analyse the value continuity from each geological domain. These statistical results are given in Table 6.2.

144

Figure 6.2 Histograms of MgO wt.% and CaO wt.% values from the 11,552 samples analysed at the Turnagain ultramafic complex

Histograms of MgO wt.% and CaO wt.% values from the 11,552 samples analysed at the Turnagain ultramafic complex, northern British Columbia, Canada. The histograms display the values with a bin spacing of 1 wt.%. The red curve in each chart is an overlain normal distribution curve.

145

Table 6.2 Statistical properties of 11,552 geochemical samples from the Turnagain ultramafic deposit

Domain Information Rock Code 1 2 3 4 5 Lithology Green Dunite Dunite Wehrlite Olivine Serpentinite Clinopyroxenite Number of samples 931 5934 3400 944 343 Statistical Method MgO (wt.%) Minimum 18.18 11.00 09.92 10.22 07.86 Maximum 57.21 58.82 57.84 51.89 53.47 Mean 44.34 41.46 38.31 30.09 37.22 Median 44.87 42.14 39.37 29.45 38.14 Standard Deviation 03.90 05.19 06.23 07.48 07.32 Coefficient of Variation 00.09 00.13 00.16 00.25 00.20 Skewness -1.68 -0.82 -0.55 00.42 -0.55 P10 40.22 34.77 29.59 20.62 27.95 P90 48.06 46.98 44.95 40.44 46.43 Statistical Method CaO (wt.%) Minimum 00.03 00.01 00.01 00.01 00.01 Maximum 13.43 14.06 17.24 20.01 13.66 Mean 00.88 01.19 02.58 06.54 02.08 Median 00.53 00.63 01.89 05.23 01.30 Standard Deviation 01.14 01.48 02.53 04.70 02.18 Coefficient of Variation 01.30 01.25 00.98 00.72 01.05 Skewness 04.57 02.57 01.68 00.77 01.70 P10 00.17 00.10 00.21 01.39 00.13 P90 01.80 03.03 05.89 13.90 04.79

Statistical properties of 11,552 geochemical samples from the Turnagain ultramafic deposit, British Columbia, Canada. Weight percentage MgO and CaO values and the number of samples represented by each rock type identified as having the potential for mineral carbonation at the deposit are given.

146

The preliminary statistical analysis on the MgO and CaO weight abundances separated by lithology, defines a lithological origin to the variability in MgO and CaO throughout the Turnagain deposit. The Green dunite displays the greatest mean MgO and lowest mean CaO weight percentages at the deposit. The low number of green dunite samples (931 samples) in comparison with dunite (5934 samples) may cause an artificial higher mean MgO value than that of dunite. The dunite rocks have a higher MgO and lower CaO mean weight than the wehrlite rocks. Wehrlite has a much broader spread of sample weight abundances as shown in the histograms of MgO wt.% in Figure 6.3. The olivine clinopyroxenite rocks have a much more variable MgO weight percentage with a mean of 30.09 wt.% this is some 11.37 wt.% lower than dunite and has a significantly greater mean CaO weight percentage of 6.54 wt.% to 1.19 wt.%. The variability in MgO and CaO shown from rock to rock at the Turnagain deposit highlights the importance of understanding the lithology and associated mineralogy. Whether applying the average

MgO and CaO quantity to an RCO2 factor proposed by Goff et al. (2000) on a rock unit by rock unit basis to determine the theoretical CO2 storage at the deposit or through determining factors such as lithological and quality value domains as part of a geostatistical investigation, a thorough investigation of the raw geochemical values analysed should be adhered.

147

Figure 6.3 Histograms of MgO wt.% values from the 11,552 samples analysed at the Turnagain ultramafic complex

148

Histograms of MgO wt.% values from the 11,552 samples analysed at the Turnagain ultramafic complex, British Columbia, Canada. The five histograms represent each of the lithologies identified as possible sources of substrate rock for mineral carbonation; green dunite (rock code 1), dunite (rock code 2), wehrlite (rock code 3), olivine clinopyroxenite (rock code 4), serpentinite (rock code 5). The histograms display the values with a bin spacing of 1 wt.%. The red curve in each chart is an overlain normal distribution curve.

6.1.4 MCP data generation The suite of 11,552 geochemical assay samples from the Turnagain ultramafic complex were used to generate MCP values using the MCP calculator proposed in Chapter 5. Magnesium numbers from Table 5.1 determined from electron microprobe analyses on olivine grains in each rock type by Scheel (2007) were assigned to each sample input into the MCP calculator based on the logged lithology. The mineral assemblage used in the modal mineral estimation calculations are the same as those defined and validated in Chapter 5. The minerals estimated for each lithology are; olivine (forsterite

149 composition determined by the assigned magnesium number), serpentine (assumed to be lizardite), clinopyroxene, magnetite, anorthite, chromite, ilmenite, quartz and iron sulphide (FeS). A modal mineral profile is normalised to equal 100 wt.% within the MCP calculator. This is based on this defined mineral assemblage and the converted oxide profile from the major cation geochemical assay values input into the program by the author. The MCP calculator worksheet of the program was manipulated to record two values for each of the 11,552 samples run through the program. Based on experimental mineral carbonation experiments undertaken in Chapter 4 of this research, it was determined that olivine was the only contributing mineral source of MgO to mineral carbonation. Negligible, if any MgO or CaO was proposed to have been leached from serpentine and diopside respectively. As such they provide no contribution to the stoichiometric silicate to carbonate conversion. Due to this it was decided that the only contributing mineral (Smin) used in the MCP calculation (Equation 5.3) would be olivine.

The first MCP value recorded by the MCP calculator used a reaction extent (Rx) factor of 0.37. This represents the mean 37 % magnesium silicate to magnesium carbonate achieved using the direct aqueous mineral carbonation extent achieved using the experimental set up at UBC by the user. The second value recorded, named MCP_ARC, used the same MCP calculator program set up but replaces the Rx factor of 0.37 with 0.81. This represents the maximum reaction extent achieved at the Albany Research Center (Gerdemann et al., 2007). This is the greatest reaction extent achieved to date using a direct aqueous mineral carbonation process. Both the MCP and MCP_ARC values determined for each of the 11,552 samples input into the MCP calculator were added to the geochemical table of the geological database. This was uploaded into the GEOVIA Surpac software program for statistical and geostatistical analysis. No duplicate data validation was deemed necessary beyond that undertaken in Chapter 5. This is based on the internal consistency of the MCP calculator and the values processed were considered applicable for use in undertaking a mineral carbonation potential resource estimate at the Turnagain ultramafic complex.

6.2 MCP estimation at the Turnagain ultramafic complex, British Columbia, Canada - a case study

The geological database was populated with the MCP values generated by the MCP calculator from the Turnagain ultramafic complex. A case study presenting MCP resource estimation is provided utilising this geological database, together with topographical, three-dimensional geological models and surface mine designs at the complex to constrain the estimation by geological and practical mining boundaries.

150

A preliminary analysis of the MCP and raw data is given including descriptive, univariate and bivariate statistics and extreme or outlier value analyses. The results of evaluating the effect of compositing data versus raw data for use in the estimation process is further demonstrated. A three-dimensional block model was generated by the author using the GEOVIA Surpac software program within which the geospatial MCP estimate at the Turnagain ultramafic complex was undertaken. The block model provides an orthogonal regularised grid in which the MCP values and the volume of substrate material for mineral carbonation can be interpolated from values contained within the geological database. This process can be undertaken on any equivalent mining software capable of resource estimation procedures. The dimensions of the block model created at Turnagain in provided in Table 6.3. The block model covers an area of 2 km by 3.25 km area to a depth of 900 m from surface of the ultramafic complex. This area was reduced from the total ultramafic outline approximately 5.5 km by 8 km mapped by Clark (1975). This is based on the drill holes used for the MCP estimation process primarily being within the Northwest Zone and Horsetrail Zones highlighted in Figure 3.2.

Table 6.3 Block model dimensions used for MCP resource estimation at the Turnagain ultramafic complex

Dimensions Northing Easting Elevation Minimum 6480500 507000 500 Maximum 6482510 510255 1400 Block size (m) 15 15 15 Sub-block size (m) 7.5 7.5 7.5

The block model is comprised of blocks 15 m3 being reduced to 7.5 m3 to improve the estimation resolution within the geological domain models. This better defines the boundaries between each domain. A 15 m block height was selected based on the bench height of the existing mine plan designed at the Turnagain property. This is considered a low enough block resolution a preliminary estimation purposes. Spatial correlation and continuity of MCP values was studied and interpreted using the variography functionality the GEOVIA Surpac software program. This analysis comprises of down drill hole and global drill hole semi-variograms being created. Anisotropic search ellipses were created to better constrain the value interpolations. Both an empirical inverse distance weighting and a geostatistical ordinary kriging estimation methods were used to interpolate MCP values within

151 individual blocks of the created block model. The parameters and justification of the parameters used in each estimation scenario are provided with the results.

6.2.1 Preliminary MCP data analysis The 11,552 MCP and MCP_ARC sample values derived from the MCP calculator program were added to the geochemical table of the geological database. This was uploaded into the GEOVIA Surpac software program. Descriptive statistical analyses on the data were undertaken on the raw MCP and MCP_ARC values generated by the calculator. These statistical results are found in Table 6.4. The preliminary statistical analysis on the MCP and MCP_ARC values separated by lithological rock code, define a lithological origin to the variability in MCP values. This mirrors that of the MgO content in Table 6.2 throughout the Turnagain deposit. The descriptive statistics provided for the MCP values supports the geological domain and boundary selection for geological modelling. The greater olivine concentration of the dunite rocks over wehrlite and both dunite and wehrlite over olivine clinopyroxenite based on their logged rock description is reflected in the greater MCP values observed in these olivine-rich rocks.

152

Table 6.4 Descriptive statistics of MCP and MCP_ARC values derived using the MCP calculator program

Domain Information Rock Code 1 2 3 4 5 Lithology Green Dunite Dunite Wehrlite Olivine Serpentinite Clinopyroxenite Number of samples 931 5934 3400 944 343 Statistical Method MCP Minimum 0.00 0.00 0.37 3.20 0.81 Maximum 25.59 26.11 22.84 20.68 22.92 Mean 14.35 12.17 11.19 8.62 9.77 Median 14.44 11.95 11.22 7.68 8.06 Standard Deviation 3.50 4.29 4.10 3.47 4.71 Coefficient of Variation 0.24 0.35 0.37 0.40 0.48 Skewness -0.34 0.36 0.19 0.98 1.00 P10 9.62 6.56 5.90 5.03 5.24 P90 18.46 17.94 16.58 13.94 18.03 Statistical Method MCP_ARC Minimum 0.00 0.00 0.37 5.73 1.49 Maximum 57.55 58.17 51.38 46.02 51.54 Mean 30.75 25.30 23.08 17.20 19.58 Median 30.94 24.69 23.10 15.27 15.37 Standard Deviation 8.85 10.79 10.22 8.51 11.67 Coefficient of Variation 0.29 0.43 0.44 0.49 0.60 Skewness -0.33 0.35 0.18 0.88 1.02 P10 18.47 11.11 9.79 8.20 8.32 P90 41.34 40.10 36.87 30.27 40.39

Descriptive statistics of MCP and MCP_ARC values derived using the MCP calculator program on 11,552 samples from the Turnagain ultramafic complex, British Columbia, Canada. The values are classified based on their assigned geological identification code.

153

Three geological domains were modelled, a coupled green dunite and dunite unit, wehrlite and olivine clinopyroxenite. The volume of each geological unit defined by physical boundaries within which MCP estimation is undertaken is provided in Table 6.5. Each block in the block model is assigned a geological code permitting the estimation parameters used to be confined by geology. Each block is assigned a unique geological code if 100 % of the block (7.5 m3) falls within the associated modelled solid. All other blocks are given an unclassified code. A typical elevation section of the block model with each block classified by its geological code is given in Figure 6.4. The redefined volume of rock based on the block model dimensions are also provided in Table 6.5. Specific gravity measurements were taken on 371 dunite drill core samples, 94 wehrlite drill core samples, and 19 olivine clinopyroxenite drill core samples. These values were added to the geological database and interpolated using a nearest neighbour estimation method. Each block was assigned a specific gravity value constrained by the defined geological domain. The interpolated specific gravity values were used to more accurately determine bulk density and mass of substrate material at the Turnagain property. The average bulk density and estimated mass of rock used in the mineral carbonation potential estimation are given in Table 6.5.

Table 6.5 Modelled volume of each defined geological domain at Turnagain

Parameter Geological Domain Dunite Wehrlite Olivine Clinopyroxenite Geological modelled 349,443,568 65,504,369 11,730,214 volume (m3) Average bulk density 3.03 3.00 3.03 (t/m3) Block model volume 348,165,422 65,486,813 11,443,781 (m3) Block model mass (t) 1,050,650,620 191,948,805 30,326,425

Modelled volume of each defined geological domain; dunite, wehrlite, and olivine clinopyroxenite at the Turnagain ultramafic complex British Columbia, Canada. The average bulk density interpolated using a nearest neighbour estimation was used to calculate the mass of each geological domain in the block model created for MCP estimation.

154

Figure 6.4 Plan section of the Turnagain block model at the 900 m elevation

Plan section of the Turnagain block model at the 900 m elevation. The blocks are coloured corresponding to their geological domain; dunite (blue), wehrlite (green), olivine clinopyroxenite (cyan) and blocks not assigned a geological domain (red).

155

The raw CaO, MgO, MCP and MCP_ARC values were extracted from the geological database and composited into 15 m samples. The compositing procedure linearly calculates the weighted average values down each drill hole. It smoothes the data and provides consistent support for resource estimation. A 15 m composite length was chosen corresponding to the block size within the block model and the selective mining unit size chosen by Hard Creek Nickel Corporation. Samples were included in the composite as long as 50 % of the composite length was made i.e. 7.5 m, subsequently the size of the sub-blocks created in the block model. All samples values were measured or calculated from the same NQ sized diamond drill core, over regularised sample intervals. The compositing procedure reduces the effect of isolated high or low values smoothing the character of the values when undertaking spatial analysis (Sinclair et al. 1983; Sinclair and Blackwell, 2002). The composited data values within each geological domain were selected for MCP estimation at the Turnagain deposit. The composited values are statistically summarised in Table 6.6 and compared to the raw values present in the geological database. The compositing process reduces the number of samples to be used in resource estimation. The data values are smoothed narrowing the range of values within each geological domain. This is expressed in Table 6.6 with a reduction in the standard deviation and a lowered coefficient of variation in MCP values from 0.34 to 0.32 for dunite, 0.37 to 0.34 for wehrlite, and 0.40 to 0.37 for olivine clinopyroxenite respectively. It was decided not to apply a bottom or top cut on any values or to create extra domains for extreme value populations. The compositing process reduced the number of extremely low or high outlying values present in the raw data. This is further indicated by the reduction in both the standard deviation and coefficient of variation values reported for both MgO and MCP in each geological domain. This is an arbitrary decision made by the author. Careful consideration should be applied to the treatment of outlying and extreme values when estimating the mineral carbonation potential of other deposits.

156

Table 6.6 Descriptive statistics of CaO (wt.%), MgO (wt.%), MCP, and MCP_ARC values on composited drill core samples from the Turnagain ultramafic complex.

Geological Domain Dunite Statistical Method Raw Values Composited Values (15 m) Value CaO MgO MCP MCP_ARC CaO MgO MCP MCP_ARC (wt.%) (wt.%) (wt.%) (wt.%) Number of samples 6865 1842 Minimum 0.01 1.13 0.00 0.00 0.01 20.95 4.88 7.65 Maximum 14.06 58.82 26.11 58.17 10.86 55.73 24.34 54.73 Mean 1.15 41.84 12.47 26.04 1.15 41.85 12.46 26.02 Median 0.62 42.55 12.48 26.11 0.76 42.34 12.26 25.43 Standard Deviation 1.45 5.15 4.26 10.71 1.16 4.63 3.96 9.94 Coefficient of Variation 1.26 0.12 0.34 0.41 1.01 0.11 0.32 0.38 Skewness 2.74 -0.97 0.24 0.23 2.24 -0.60 0.35 0.34 P10 0.10 35.31 6.68 11.38 0.18 35.64 7.20 12.68 P90 2.90 47.19 18.05 40.41 2.72 46.82 17.74 39.47 Wehrlite Value CaO MgO MCP MCP_ARC CaO MgO MCP MCP_ARC (wt.%) (wt.%) (wt.%) (wt.%) No. Samples 3400 902 Minimum 0.01 9.92 0.37 0.37 0.02 18.77 3.98 7.30 Maximum 17.24 57.84 22.84 51.38 14.33 54.21 20.84 45.85 Mean 2.58 38.31 11.19 23.08 2.55 38.34 11.17 23.04 Median 1.89 39.37 11.22 23.10 2.05 39.27 10.97 22.38 Standard Deviation 2.53 6.23 4.10 10.22 2.14 5.63 3.79 9.43 Coefficient of Variation 0.98 0.16 0.37 0.44 0.84 0.15 0.34 0.41 Skewness 1.68 -0.55 0.19 0.18 1.45 -0.43 0.25 0.25 P10 0.21 29.59 5.90 9.79 0.35 30.66 6.28 10.96 P90 7.80 46.86 17.72 39.35 5.36 44.65 16.31 36.10 Olivine Clinopyroxenite Value CaO MgO MCP MCP_ARC CaO MgO MCP MCP_ARC (wt.%) (wt.%) (wt.%) (wt.%) No. Samples 944 259 Minimum 0.01 10.22 3.20 5.73 0.02 14.26 3.56 6.71 Maximum 20.01 51.89 20.68 46.02 17.88 50.66 19.05 41.30 Mean 6.54 30.09 8.62 17.20 6.50 30.11 8.59 17.13 Median 5.23 29.45 7.68 15.27 5.55 29.21 7.65 15.07 Standard Deviation 4.70 7.48 3.47 8.51 4.23 6.80 3.16 7.71 Coefficient of Variation 0.72 0.25 0.40 0.49 0.65 0.23 0.37 0.45 Skewness 0.77 0.42 0.98 0.88 0.70 0.50 1.15 1.08 P10 1.39 20.62 5.03 8.20 1.75 22.55 5.56 9.42 P90 13.90 40.44 13.94 30.27 12.59 39.98 13.06 28.12

Descriptive statistics of CaO (wt.%), MgO (wt.%), MCP, and MCP_ARC values on composited drill core samples from the Turnagain ultramafic complex, northern British Columbia, Canada. The composites were created if 50 % of the sample length falls within the selected 15 m composite length. The values are classified based on their modelled geological domain.

157

Geostatistical methods of value interpolation used for MCP resource estimation are optimally used when the data values express a normal distribution. The descriptive statistics provided in Table 6.6 and a graphical statistical representation using histograms, probability, cumulative distribution, scatter and Q- Q plots were derived for the data values inside each define geological domain. This was completed to determine their applicability to geostatistical interpolation for MCP resource estimation. The graphical plots for CaO, MgO, MCP, and MCP_ARC values are provided in Appendix 2 and are summarised herein. Only the MCP values and not MCP_ARC values are discussed due to the values having a correlation coefficient of close to 1 as a function of the internal mineral carbonation efficiency being the only differentiation between the values determined within the MCP calculator. As such the distribution of MCP_ARC values mirrors that of MCP values. Within the dunite geological domain, the composited CaO wt.% values have a high coefficient of variation of 1.01 and a positive skewness of 2.24. When plotted (Appendix 2.1) these values are expressed as a positively skewed histogram. A steep curving cumulative frequency and probability plots, suggesting a non-normal value distribution. Following a log transformation on the CaO values, a bell shaped curve shown centralised over the mean values. This is indicative of a log normal distribution and is supported by an s-shaped cumulative frequency and a near linear probability plot on log normalised values. This has an implication on spatial semi-variogram modelling of CaO values at Turnagain and is discussed in Section 6.2.2. The close mean and median MgO wt.% vales and the low coefficient of variation of 0.11 suggested a normally distributed MgO wt.% population within the dunite geological domain. When plotted (Appendix 2.1) as a histogram, the population of values appears negatively skewed. The plotted cumulative frequency displays a strong s-shaped curve. The probability plot of the values shows a slight inflection at 26 wt.% suggesting that a possible bottom-cut of values would reduce this skewed habit. The postulated normal distribution from the descriptive statistics, histogram and cumulative frequency plots prompted the decision not to bottom-cut the MgO wt.% values. The population is thus considered to have a normal distribution for estimation purposes. A scatter plot of MgO wt.% versus MCP values within the dunite geological domain (Appendix 2.1) indicates a correlation coefficient of 0.88. This positive correlation is not surprising due to the high olivine content of the dunite rocks at Turnagain. This strong correlation is reflected in a similar histogram shape as MgO with the almost equal mean and median values peaking at the apex of a bell curve displaying normal distribution. Like the MgO wt.% values, there is a slight appearance of a sub- population that between 7 and 9. This is confirmed by a slight deflection from a near linear probability

158 plot of the histogram classes. This sub-population is not spatially distinct and comparable to MgO wt.% values it was decided not to apply a bottom-cut to the value population or create a separate value- based domain. Histograms, probability and cumulative frequency plots MgO wt.% and MCP of the 15 m composite values inside the wehrlite geological domain displayed a bimodal distribution. This is as shown in the histogram of MgO wt.% values in Figure 6.5. The three geological domains have been treated as hard boundaries. Based on the exploratory statistical analysis of the composited values inside the wehrlite domain it was decided to create two sub-domains. These are named herein as an olivine-poor wehrlite domain and an olivine–rich wehrlite domain. This domain is created on a soft-boundary around over the 25th percentile MCP value of 7.5. The naming convention was created based on the postulated origin of the bimodality being attributed to the relative proportions of olivine and clinopyroxene minerals in the rocks. Descriptive statistics of the wehrlite sub-domains are given in Table 6.7.

Figure 6.5 Histogram of the MgO wt.% values within the wehrlite geological domain modelled at the Turnagain ultramafic complex

The histogram is split in to a low olivine (red) and high olivine (green) wehrlite based on the bimodal distribution shown. A soft boundary is applied to 25 % of the values on either side of the boundary displayed as the black dividing line.

159

Table 6.7 Descriptive statistics of CaO (wt.%), MgO (wt.%), MCP, and MCP_ARC values on composited drill core samples from the Turnagain ultramafic complex

Parameter Wehrlite Geological Domain Olivine-poor Wehrlite Olivine-rich Wehrlite Values CaO MgO MCP MCP_ARC CaO MgO MCP MCP_ARC (wt.%) (wt.%) (wt.%) (wt.%) No. Samples 369 686 Minimum 0.04 18.77 3.98 7.30 0.02 26.12 7.52 13.31 Maximum 14.33 42.30 9.99 20.40 11.03 54.21 20.84 45.85 Mean 3.19 33.77 7.38 13.58 2.20 26.75 12.68 26.75 Median 2.65 33.88 7.11 12.91 1.62 26.49 12.56 26.49 Standard Deviation 2.28 4.50 1.36 3.21 1.93 4.27 3.05 7.66 Coefficient of Variation 0.72 0.13 0.18 0.24 0.88 0.11 0.24 0.29 Skewness 1.41 -0.49 0.27 0.33 1.44 -0.25 0.31 0.24 P10 0.73 27.67 5.82 9.84 0.29 34.61 8.73 16.65 P90 6.01 39.41 9.39 18.38 4.87 45.19 16.76 37.09

The values are classified based on their geological sub-domain as either olivine-poor or olivine-rich wehrlite.

The coefficient of variation is reduced for all sample values as an effect of the sub-domains created using a soft boundary within the wehrlite geological domain. This was further validated by generating histograms, probability, and cumulative frequency plots for each sub-domain (Appendix 2.2). Comparable the dunite geological domain, the CaO wt.% values are positively skewed and display a log- normal distribution when the values are log-transformed. Both the olivine-poor and olivine-rich wehrlite domains have a normal distribution for both MgO wt.% and MCP values in the three plots. The descriptive statistics, histograms, probability, and cumulative frequency plots for the olivine clinopyroxenite geological domain (Appendix 2.3) display a slightly positively skewed distribution. This attributed to the low number of samples (259). A log transformation was undertaken on all the samples within the domain. It was shown that only the MgO wt.% values have a normal distribution, whilst the CaO wt.% and MCP values had a log-normal distribution. MgO wt.% and CaO wt.% values show a negative correlation coefficient of -0.77, when plotted on a scatter plot. There is a correlation coefficient of 0.68 between MgO wt.% and MCP values. This is indicative of the increased mineral variability in the olivine clinopyroxenite geological domain. Elevated CaO values and reduced MgO values are associated with high clinopyroxene to olivine mineral ratios. The plots suggest that there could be some small sub- populations within the domain. Due to the relative small sample quantity, this could not be fully determined. The majority of the sample values have a central tendency around the mean or median and

160 as such for the purposes of this research, the domain boundaries are deemed applicable to the resource estimation purpose.

6.2.2 Spatial (structural) data analysis Semi-variogram analysis was undertaken to characterise the spatial correlation of 15 m composited MgO, CaO, MCP, and MCP_ARC values. This was completed for each geological domain created at the Turnagain ultramafic complex. The variography process provides a set of search parameters underlying the resource estimation process. Experimental semi-variograms down the length of drill holes were generated for each of the parameters within each domain. This was done to determine the nugget value

(C0) by modelling the experimental variograms. The nugget was recorded as the point at which the model intercepts the y-axis of the semi-variogram at zero lag distance. The nugget is determined down the length of drill hole variograms due to them being the most informed direction and the source of the closest sample spacing. The nugget values determined from down drill hole variogram modelling are fixed for all directional variogram modelling therein. The data was composited down each drill hole in the database at 4 m intervals (the average sample spacing) to generate the variograms. Samples were selected and composited based on an assigned code representative of the geological domain. Omni-directional variograms were generated at 10° increments generating 18 primary variograms for each of the CaO, MgO, MCP, and MCP_ARC parameters within each domain. This was completed using the 15 m composited values extracted from the geological database. A plane dip of -10° and a plunge direction and a dip direction of 100° were used to account for the slight dipping nature of the Turnagain ultramafic complex. A coned search ellipse for samples was used with a 45° spread and a 320 m search distance. The search distance was selected based on double the largest drill hole spacing, ensuring that at least two drill holes would be selected in each search. A lag distance of 20 m was used, guided by half the distance between the closest drill hole spacing at the deposit. Note that the selections of these parameters were made on a rule of thumb by the author due to their being no strict guidelines for search and variography parameters. The direction of maximum continuity was determined from the author’s interpretation of the 18 primary variograms that were generated for each parameter in each geological domain (major axis). Secondary maps were then created in which the process was repeated (semi-major axis). The experimental variograms selected were subsequently extracted along the anisotropic ellipsoid axis. This produced a third variogram in the direction of minimum continuity (minor axis) perpendicular to the dip plane. Standard variograms were generated for all normally distributed data populations. For those

161 identified with a log-normal distribution, log variograms particularly effective at viewing positively skewed data were generated. These log variograms were rescaled to equate a sill value of 1 for modelling. All variograms were validated graphically using pairwise relative variograms. This variogram type was used due to its effectiveness in interpreting small data sets. The variograms extracted along the three axis described were modelled graphically by the author.

These models were created to best fit the variograms using a fixed nugget value (C0) determined from down drill hole variography. The model is then best fit to equal a sill value of 1 (C0 + C1 ± C2) and a specified range of influence (a). This is specified in metres along the direction of maximum continuity (major axis). The range of the model is adjusted where necessary along the semi-major and minor axis to best fit the model. A spherical model was used for each variogram as it was the most appropriate model shape in each case. The variogram model parameters for each value modelled within each geological domain are provided in Table 6.8.

162

Table 6.8 Directional variogram model parameters for each geological domain and sub-domain at the Turnagain ultramafic complex

Domain Model Value Axis Variogram parameter

c0 c1 a1 c2 a2 Major 0.13 0.87 120 CaO S-Major 0.13 0.87 120 Minor 0.13 0.87 107 Major 0.08 0.92 110 MgO S-Major 0.08 0.92 110

Minor 0.08 0.92 96 Dunite Major 0.05 0.46 50 0.49 125 MCP S-Major 0.05 0.46 50 0.49 125 Minor 0.05 0.46 10 0.49 20 Major 0.08 0.29 45 0.63 105 MCP_ARC S-Major 0.08 0.29 45 0.63 105 Minor 0.08 0.29 33 0.63 80 Major 0.07 0.93 105 CaO S-Major 0.07 0.93 105 Minor 0.07 0.93 57 Major 0.08 0.42 30 0.5 135 MgO S-Major 0.08 0.42 30 0.5 135

Minor 0.08 0.42 14 0.5 64 Olivine-Poor Wehrlite Major 0.05 0.31 15 0.64 135 MCP S-Major 0.05 0.31 12 0.64 110 Minor 0.05 0.31 15 0.64 135 Major 0.05 0.305 13 0.645 130 MCP_ARC S-Major 0.05 0.305 13 0.645 130 Minor 0.05 0.305 9 0.645 95 Major 0.08 0.09 15 0.83 120 CaO S-Major 0.08 0.09 15 0.83 120 Minor 0.08 0.09 8 0.83 54 Major 0.08 0.25 40 0.67 125 MgO S-Major 0.08 0.25 40 0.67 125

Minor 0.08 0.25 19 0.67 64 Olivine-Rich Wehrlite Major 0.05 0.255 32 0.695 120 MCP S-Major 0.05 0.255 32 0.695 120 Minor 0.05 0.255 24 0.695 103 Major 0.05 0.095 14 0.855 120 MCP_ARC S-Major 0.05 0.095 14 0.855 120 Minor 0.05 0.095 7 0.855 62 Major 0.06 0.94 155 CaO S-Major 0.06 0.94 105 Minor 0.06 0.94 155 Major 0.04 0.96 100 MgO S-Major 0.04 0.96 100

Minor 0.04 0.96 52 Olivine Clinopyroxenite Major 0.05 0.95 85 MCP S-Major 0.05 0.95 85 Minor 0.05 0.95 82 Major 0.05 0.95 75 MCP_ARC S-Major 0.05 0.95 75 Minor 0.05 0.95 50

Variogram parameters modelled are c0 – nugget value, c1 – structure 1 sill, a1 – structure 1 range, c2 – structure 2 sill, a2 – structure 2 range. Variograms were created for CaO (wt.%), MgO (wt.%), MCP, and MCP_ARC within each domain.

163

6.2.3 Three-dimensional data interpolation CaO, MgO, MCP and MCP_ARC values were interpolated in to the block model created at the Turnagain ultramafic complex to generate global resource estimation for each parameter. The values were interpolated using three estimation techniques; inverse distance, ordinary kriging and nearest neighbour estimation. The 15 m composited values extracted from the geological database inside each of the four defined geological domains were used as the source of the data values from each domain. The search and estimation parameters used in each case are discussed here. The basis of which were determined from the Variogram modelling of each parameter in the previous section. Due to the low nugget values modelled from variography, a high reproducibility of data is assumed. This permits a range of powers to be used for inverse distance. As such each parameter in each geological domain was interpolated using an inverse distance interpolation technique using a power of 2 (ID2) and 3 (ID3). Three interpolations were undertaken for each parameter within the block model in three estimation passes. The search parameters used for each estimation is provided in Table 6.9. An anisotropic search ellipsoid was used to search for sample values when interpolating values into blocks with an unknown data value. The dimensions of each search ellipsoid were determined from the modelled variogram for each parameter in each domain. The search distance for samples varied between parameters for each estimation pass. Pass 1 used a search distance equaling half the range in metres from the variogram. Pass 2 used a search distance equaling the range and pass 3, twice the distance of the range. Each estimation pass did not overwrite already interpolated block values. These blocks were marked with an identifier of which pass the value was estimated. A fixed minimum number of samples were set to 3 and a maximum of 30 to be used to estimate the value of a block. These values are arbitrary and selected by the author to ensure that an adequate number of samples were used. This is to provide confidence in the block value. A maximum number was used to minimise data clustering. The volume variance effect was reduced by using a 4 x 4 x 4 discretisation of blocks for estimation in the x, y, and z orientation.

164

Table 6.9 Anisotropic search ellipse dimensions for resource estimation at the Turnagain ultramafic complex

Geological Model Ellipsoid Orientation Anisotropy Factors Maximum Search Domain Value Distance Major Semi- Minor Semi- Minor Pass 1 Pass 2 Pass 3 Axis Major Axis Major Axis (m) (m) (m) Axis Axis CaO 100 -10 40 1 1.12 60 120 240 MgO 160 -5 60 1 1.13 55 110 220 Dunite MCP 170 -5 -80 1 1.83 65 125 250 MCP_ARC 170 -5 -80 1 1.25 55 105 210 CaO 270 10 -80 1 1.46 55 105 210

MgO 100 -10 -90 1 1.53 70 135 270 Olivine-Poor Wehrlite MCP 270 10 60 1.19 1.02 70 135 270

MCP_ARC 100 -10 70 1 1.28 65 130 260

CaO 270 10 -60 1 1.45 60 120 240

MgO 250 10 -50 1 1.51 65 125 250 Olivine-Rich Wehrlite MCP 250 10 -60 1 1.14 60 120 240

MCP_ARC 260 10 -60 1 1.49 60 120 240

CaO 160 -5 -80 1.47 1 80 155 310

MgO 160 -5 -60 1 1.93 50 100 200 Olivine Clinopyroxenite MCP 170 -5 30 1.16 1 45 85 170

MCP_ARC 170 -5 -70 1 1 40 75 150

The search parameters are defined for each geological domain estimated at the deposit for CaO (wt.%), MgO (wt.%), MCP, and MCP_ARC. The search distance for all three estimation passes are given.

The search parameters used for each pass are defined in Table 6.9. Ordinary kriging requires a semi- variogram to be generated for each set of data used for the interpolation. The variography undertaken in the previous section was used and the variogram parameters in put into the modeling software are provided in Table 6.8. In addition to the value interpolation, the kriging variance and kriging efficiency was also calculated for each interpolated block. The kriging variance was calculated as a relative

165 measure of the confidence in each block estimate by measuring the data spread around an estimated block. The kriging efficiency of each block was calculated comparing the kriging variance and block variance to evaluate the effectiveness of the ordinary kriging process as an interpolation method for the parameter values studied. A final three-pass nearest neighbour interpolation was undertaken for each parameter for use in the validation of the inverse distance and ordinary kriging interpolation methods.

6.3 Global MCP resource estimation and validation

A global resource estimate of CaO, MgO, MCP and MCP_ARC values were undertaken at the Turnagain ultramafic complex. The resource estimates are constrained within wireframes designed by the author modelling three identified geological domains at the deposit. The dimensions of each geological model are confined by the diamond drilling information available from Hard Creek Nickel Corporation and the author’s interpretation. The total volume and mass of each rock mass is underestimated on the scale of the ultramafic deposit. The models however, utilise the best information available to undertake a realistic resource estimate. Each resource estimate is constrained by the confidence of the range in metres by which an orthogonal block can be estimated. This is based on exploratory data analysis and variography discussed earlier in the chapter. A resource estimate is provided using an inverse distance method to the power of 2 and 3, a nearest neighbour analysis, and an ordinary kriging in Table 6.10. The validity of each method is analysed and compared with the theoretical potential for CO2 sequestration defined by the RCO2 parameter proposed by Goff et al. (2000). This in turn is used as an evaluation of using the MCP parameter for undertaking detailed resource estimate of the mineral carbonation potential of a rock mass. It is furthermore an evaluation on the use of well-established estimation techniques within the mining industry as a tool for estimating MCP.

166

Table 6.10 CaO, MgO, MCP and MCP_ARC resource estimates at the Turnagain ultramafic complex

Geological Estimation Variable Domain Method CaO MgO Volume (m3) Mass (kt) Value (wt.%) Volume (m3) Mass (kt) Value (wt.%) ID2 337,871,672 1,020,422 0.94 330,387,609 998,078 42.64 ID3 337,871,672 1,020,422 0.94 330,387,609 998,078 42.62 Dunite OK 337,871,672 1,020,422 0.97 330,387,609 998,078 42.53 NN 337,890,234 1,020,515 1.01 332,180,578 1,003,606 42.48 ID2 62,539,172 183,486 2.47 65,168,719 191,009 37.17 ID3 62,539,172 183,486 2.47 65,168,719 191,009 37.19 Wehrlite OK 58,720,781 172,301 2.25 65,147,625 190,950 36.69 NN 63,348,328 185,781 2.35 65,034,984 190,613 37.79 ID2 11,146,781 29,780 5.52 9,410,766 26,559 30.62 3 Olivine ID 11,146,781 29,780 5.56 9,410,766 26,559 30.57 Clinopyroxenite OK 11,146,781 29,780 5.73 9,410,766 26,559 30.54 NN 11,232,000 30,080 5.73 10,270,969 28,493 30.70 Variable MCP MCP_ARC ID2 318,828,656 962,260 12.99 318,606,750 962,350 27.31 ID3 318,828,656 962,260 12.98 318,606,750 962,350 27.29 Dunite OK 318,828,656 962,260 12.88 318,606,750 962,350 27.10 NN 324,623,953 981,807 12.85 323,427,938 977,882 26.96 ID2 65,403,703 191,698 9.91 65,220,609 191,161 19.49 ID3 65,403,703 191,698 9.90 65,220,609 191,161 19.48 Wehrlite OK 65,403,703 191,698 9.43 65,220,609 191,161 18.53 NN 65,391,891 191,664 10.16 65,131,172 190,890 20.29 ID2 9,748,688 27,249 8.32 9,513,281 26,603 16.53 Olivine ID3 9,748,688 27,249 8.32 9,513,281 26,603 16.50 Clinopyroxenite OK 9,748,688 27,249 8.31 9,513,281 26,603 16.52 NN 10,510,172 29,141 8.15 10,314,844 28,619 16.20

Resource estimates for each defined geological domain is provided using inverse distance to the power of 2 and 3, ordinary kriging, and nearest neighbour interpolation techniques. The volume, mass and mean value for each parameter is given.

167

The volumes, tonnages, CaO, MgO, MCP and MCP_ARC interpolated values for all three geological domains produced consistent comparable results for all three interpolation methods analysed. The nearest neighbour populated values on the whole produce greater volume, mass and higher attribute values than those of the smoothing interpolation techniques employed. The relative low variance of the interpolated variables is expressed with the similar values produced by both the inverse distance and ordinary kriging. As postulated from the exploratory data analysis, dunite provides the greatest opportunity for mineral carbonation at the Turnagain deposit. This is attributed to its high MgO content (~42.5 wt.%) resulting in a consistent interpolated mean MCP value of approximately 13. The lower MCP values from dunite to wehrlite to olivine clinopyroxenite can be attributed to the noticeable lowering of MgO quantity and respective olivine content of the classified rocks. The estimated values were graphically checked by the author, comparing raw values displayed down the length of drill holes superimposed on the block estimates. This was undertaken as a spot check on the production of realistic block estimates based on the raw input values. Each block estimated was further flagged with an identifier from which estimation pass was used to populate the block values. All estimates consistently showed a concentric pattern around data points. From this the distance was measured at random points to ensure that the appropriate estimation ranges were being applied. Histograms created of the estimated values for each of the parameters were graphically compared to those of the raw values. The histograms generated for CaO, MgO, MCP and MCP_ARC for each domain show a strong correlation between the estimated and raw data values. This is expressed by comparable mean MCP values in Table 6.11. The significantly lower mean MCP value of the wehrlite domain is a product of the sub-domaining undertaken. A more realistic global mean for wehrlite is observed than seen from the descriptive statistics of the raw MCP values. During each ordinary kriging estimation pass, the kriging variance and kriging efficiency were calculated. This permitted a quantification of the degree of confidence to be interpreted on the estimation process (Table 6.11). The generalised high kriging efficiency normalised to 1 and the low kriging variance particularly towards the range of the variograms (pass 2), suggests that ordinary kriging is an effective and efficient method of estimating MCP values in each of the domains interpolated.

168

Table 6.11 Ordinary kriging resource estimate values for MCP at the Turnagain ultramafic complex

Geological Estimation Composited MCP Raw Estimated MCP Data (Ordinary Kriging) Domain Pass Data Mean COV Mean COV KV KE Pass 1 12.99 0.33 0.43 Dunite Pass 2 (range) 12.83 -0.33 0.87 Pass 3 12.69 -0.34 0.59 Total 12.46 0.32 12.78 0.22 -0.18 0.66 Pass 1 8.80 0.40 0.50 Wehrlite Pass 2 (range) 9.53 -0.11 0.94 Pass 3 9.51 -0.18 1.00 Total 11.17 0.34 9.43 0.12 -0.07 0.90 Pass 1 8.45 0.36 0.50 Olivine Pass 2 (range) 8.58 -0.16 0.91 Clinopyroxenite Pass 3 7.96 -0.22 0.95 Total 8.59 0.37 8.31 0.30 -0.12 0.88

The mean MCP values for each estimation pass within the defined geological domains are compared with the raw MCP composited values. The global estimation coefficient of variation (COV), kriging variance (KV) and kriging efficiency (KE) is also provided.

The amount of CO2 that could potentially be sequestered was calculated for each block for both

MCP and MCP_ARC values using Equation 5.5. The CO2 sequestered values using MCP are based on the experimental mineral carbonation extents achieved in this research. The CO2 sequestered values using MCP_ARC are based on the greatest achievable direct aqueous mineral carbonation extents achieved in the literature to date (Gerdemann et al. 2007). The estimated CO2 sequestered is provided for each geological domain and the total modelled geology based on drill hole information at Turnagain in Table 6.12. Both the inverse distance to the power of 2 and ordinary kriging estimated MCP and MCP_ARC values are used.

The RCO2, which is the theoretical mass of rock required to sequester one tonne of CO2 was calculated using Equation 2.8 from the estimated CaO and MgO block values. This is an example in which available lithogeochemical information at the deposit can be used to gain an estimate with a greater resolution of the rock’s mineral carbonation capabilities than applying a global RCO2 to the deposit. This in turn is compared with the mass required to sequester one tonne of CO2 based on the interpolated MCP and MCP_ARC values derived in this research. The results are published in Table 6.12.

169

Table 6.12 The theoretical CO2 capacity of the ultramafic rocks at the Turnagain complex

Geological Estimation Theoretical Mass of CO2 Waste rock to sequester a Domain Method Capacity sequestered (000’s t) one tonne CO2 (t)

RCO2 Mass CO2 MCP MCP_ARC MCP MCP_ARC (000’s t) Dunite ID2 2.17 443,438 65,248 137,212 14.75 7.01 OK 2.17 443,438 64,692 136,116 14.87 7.07 Wehrlite ID2 2.49 76,710 8,531 15,899 22.47 12.03 OK 2.51 76,057 7,654 14,077 25.04 13.58 Olivine ID2 3.06 8,679 1,184 2,294 23.02 11.59 Clinopyroxenite OK 3.06 8,679 1,183 2,294 23.04 11.60 Total ID2 2.24 528,827 74,963 155,405 15.76 7.59 OK 2.24 528,174 73,529 152,487 16.06 7.74

The theoretical mass of rock required to sequester one tonne of CO2 was determined using the RCO2 parameter developed by Goff et al. (2000). The theoretical mass of CO2 that can be sequestered by each modelled geological domain was calculated from the RCO2 and the domain mass. The estimated mass of CO2 that can be sequestered by each domain is calculated using the MCP and MCP_ARC interpolated values. The mass of rock

required to sequester one tonne of CO2 using the MCP and MCP_ARC parameters is also provided.

a Theoretical values assume olivine as the exclusive mineral source of MgO.

The RCO2 values calculated in Table 6.12 are derived based on the assumption that the estimated

MgO is fixed in olivine within the rock. The theoretical capacity of the area studied to sequester CO2 7 9 based on this assumption is 5.2 x 10 t CO2. This is significantly lower than the 13.2 x 10 t CO2 estimated in Table 2.5. The theoretical capacity remains an attractive potential sink for CO2, represented within an area actively explored as a nickel sulphide resource. This estimate requires no extra sampling, geochemical analyses or experimental mineral carbonation. It provides a more informed resource estimate than applying an RCO2 to the complex as a whole assuming a uniform geology and mineralogy. An estimate comparable to this is achievable at any ultramafic complex that has undergone a mineral exploration drilling program. The theoretical capacity of the rock remains the hypothetical maximum achievable sequestration at the deposit. This applies no mineral carbonation process specific information and assumes mono-mineralogy for MgO.

The mass of CO2 that can potentially be sequestered by mineral carbonation when estimating MCP and MCP_ARC values at Turnagain is significantly lower than that of the theoretical capacity. The total

6 6 modelled amount predicted is approximately 31 x 10 t CO2 when using the MCP values and 64 x 10 t

CO2 when using the MCP_ARC values. These values represent a more achievable mineral carbonation capacity using olivine as the principal substrate mineral and the current available technology. The

170 parameters used to derive these numbers also consider the variable mineralogy and MgO source for each applicable rock type and account for inefficiencies in the mineral carbonation process. This has a significant impact on the amount of material that must be mined to sequester a tonne of

CO2. This is expressed by approximately 16 tonnes of rock required using the mineral carbonation process in this research, 8 tonnes of rock using the most successful carbonation process to date. This has implications on the practicality of developing a direct aqueous mineral carbonation facility at Turnagain when compared with the theoretical 2.25 tonnes of rock that would have to be processed to sequester one tonne of CO2 utilising a 100 % efficient mineral carbonation conversion. Table 6.13 shows the in-pit potential mineral carbonation resource at Turnagain. The resource estimation is constrained within the ultimate pit limit proposed by Hard Creek Nickel Corporation in the 2011 update (AMC, Updated Preliminary Assessment, 2011, pg. 1.7vi). The resource estimate projects a

7 theoretical capacity of approximately 22.4 x 10 t CO2 could be sequestered by the ultramafic rocks contained within the pit design. An average of 2.25 tonnes of rock is required to sequester one tonne of

CO2 assuming 100 % mineral carbonation efficiency. The estimated MCP and MCP_ARC values propose a 6 6 much lower approximate sequestration potential at 31 x 10 and 64 x 10 t CO2 respectively. Although lower, the quantity of CO2 that could be sequestered by the in-pit ultramafic rocks equates to 10.5 and

21.5 years CO2 emissions of an average 500 MW coal-fired power station that emits approximately 3 million tonnes of CO2 annually (Massachusetts Institute of Technology, 2007).

171

Table 6.13 The theoretical CO2 capacity of the ultramafic rocks within a 28 year life of mine surface mine design at the Turnagain complex

Geological Estimation Theoretical Mass of CO2 Waste rock tonneto a Domain Method Capacity sequestered (000’s t) sequester one tonne CO2 (t)

RCO2 Mass CO2 MCP MCP_ARC MCP MCP_ARC (000’s t) Dunite ID2 2.17 179,964 26,059 54,403 14.99 7.17 OK 2.18 179,139 25,853 53,880 15.11 7.24 Wehrlite ID2 2.44 36,793 4,376 8,286 20.52 10.83 OK 2.50 35,910 3,580 6,626 25.08 13.55 Olivine ID2 3.00 7,846 1,031 2,002 22.83 11.50 Clinopyroxenite OK 3.00 7,846 1,030 2,002 22.85 11.51 Total ID2 2.24 224,603 31,466 64,691 16.02 7.77 OK 2.26 222,895 30,463 62,508 16.54 8.05

required to sequester one tonne of CO2 using the MCP and MCP_ARC parameters is also provided.

a Theoretical values assume olivine as the exclusive mineral source of MgO.

172

7 Discussion

7.1 Experimental mineral carbonation

At the onset of the research the following question was proposed:

Can an experimental workflow be generated that will allow mining companies to successfully quantify the stoichiometric silicate to carbonate conversion achievable by mineral carbonation of waste rock at their mining operation in a cost effective manner?

Nickel sulphide, chromite, PGE and diamond deposits hosted within mafic and ultramafic complexes have emerged as key mineral sources capable of feeding a potential mineral carbonation processing facility. Mafic and ultramafic deposits are defined by their high magnesium and iron contents and their relative low silica content. The direct aqueous mineral carbonation regime proposed and developed at ARC remains the leading process method when using olivine as the principal mineral source. The development of an alternative mineral carbonation process was beyond the scope of this research. It was decided to replicate this direct aqueous process using proposed waste rock from a nickel sulphide mine operated by Hard Creek Nickel Corporation at their Turnagain site in British Columbia, Canada. An experimental workflow was successfully demonstrated and implemented in this research to show the achievable carbonation conversion of mining waste rock in a cost-effective manner. In addition, it highlighted concerns in the use of direct aqueous mineral carbonation as a viable processing stream when using mining waste rock. The experimental mineral carbonation autoclave system arrangement used by the author at the University of British Columbia differed from that developed at ARC to produce the most successful published mineral carbonation extents to date. The autoclave at UBC has a 100 ml capacity compared to the 2 l capacity at ARC. A gas booster pump was not available to create initial internal CO2 partial pressures in excess of 70 bar at the onset of each test. To evaluate the effectiveness of the experimental laboratory-scale setup used in this research a series of tests were run using Twin Sister olivine reactant material with greater than 80 % of the material under 38 µm. This material is the same substrate used by O’Connor and co-workers at ARC. The stoichiometric olivine to magnesite conversion achieved using Twin Sisters olivine ranged from 23.9 % in one hour to 62.6 % in 6 hours. This increase in reaction extent with increased duration of

173 testing is consistent with the results produced at ARC. The experiments replicated the solution chemistry (0.64 M NaHCO3/1 M NaCl), reaction temperature (185°C), and stirring speed (1500 rpm) as used at ARC. But, a lower PCO2 of 65 bar compared with 116 bar proposed as the optimum pressure for reaction at ARC was maintained. The 23.9 % carbonation achieved in the experimental setup at UBC is approximately 8 % lower than achieved at a comparable PCO2 using the ARC experimental setup. This experimentation using Twin Sisters olivine was important as it provided a baseline mineral carbonation extent achieved using the laboratory setup for mining waste rock. It allowed the degree of variability attributable to the experimental setup to be quantified with the results published in the literature to date. Possible reasons for the lower carbonation extents observed when using the same experimental conditions could be attributed to the smaller reaction vessel size, insufficient slurry agitation highlighted by some material collecting directly under the stirrer, estimation error of mineral abundances in the reactant and product material when using XRD with Rietveld refinement and variability in the reactant crystal size between 37 µm and 75 µm. 16 samples were tested from the Turnagain ultramafic complex. Each sample was subjected to the same reaction conditions as the Twin Sisters olivine samples for 2 hour duration. The average reaction extent achieved using the Turnagain samples was 19.7 % compared to 37.6 % achieved after testing the Twin Sisters material for the same duration. The reaction extents achieved on the 16 samples were variable ranging between 5 % and 44.9 %. There was a positive correlation with a reduction in forsterite content and gain in magnesite content between XRD analyses on the reactant and product samples tested. This confirmed the principal source of Mg2+ cations for carbonation originated from forsterite rather than other potential mineral sources available including lizardite and diopside. The experimentation completed on these samples show that potential mine waste rock can successfully be used to sequester CO2 by mineral carbonation. Consideration must be made to developing a process regime that is specific to the source rock being used. This is highlighted by the lower reaction extent achieved using the same experimental setup. The following sub-sections discuss the outcomes from the experimental mineral carbonation undertaken in this research. It shows how the findings can contribute to the development of direct aqueous mineral carbonation as a viable process stream for industrial-scale CO2 sequestration. It further discusses how the experimental system used in this research can be applied to other deposits like Turnagain. This is targeted towards providing an experimental process flow that can be used by a mining company at the operation or in a laboratory. The process flow is aimed at providing results that are

174 realistic of what can be achieved using current technology available. It uses techniques and equipment that are cost effective for the mineral carbonation potential of the material to be determined.

7.1.1 Mineral pre-treatment Mineral pre-treatment prior to mineral carbonation has been extremely effective at enhancing the stoichiometric silicate to carbonate conversion. One of the principal benefits of targeting existing or principal mining operations as providing the source material for a mineral carbonation process facility is the potential for shared mining and processing costs. Using Turnagain as an example, Hard Creek Nickel Corporation propose a combined, crushing, SAG and ball mill circuit for nickel sulphide ore processing. This circuit is designed to produce feed material under 150 µm in size. This ore pre-treatment regime is typical for a mining operation that requires a feed size under 150 µm for the flotation ore liberation process. Comminution to 150 µm for rocks handled at mining operations is small butt not uncommon. For this reason, a mineral carbonation process that is optimised using substrate material at this size would be ideal. The -38 µm feed size used in this research was selected based on the baseline mineral pre- treatment size proposed by O’Connor et al. (2005). The exponential energy requirement for size reduction down to -4 µm and processing cost associated with grinding the feed material to a smaller size makes ultra-fine grinding an unlikely mineral pre-treatment option. It is suggested that this feed size is a practical and achievable size for an industrial mineral carbonation facility. The extra grinding required in reducing rock from 150 µm to 38 µm on top of the treatment of rock that would not usually be processed remains problematic. Despite this, the incurred mining cost and a large proportion of the capital processing cost remain an attractive shared component of a coupled mining and mineral carbonation operation. From this research it is proposed that the principal mineral targeted as feed material for an industrial direct aqueous mineral carbonation process should be olivine. This is exclusively due to olivine not requiring the costly and energy expensive heat pre-treatment. From reviewing previous literature, if serpentine is to be targeted as a principal substrate source a multi-step approach to mineral carbonation would be preferable. When undertaking experimental mineral carbonation to evaluate prospective mining waste rock feed material at this time, size reduction to -38 µm for olivine-rich feed material is sufficient. Research should continue to increase mineral carbonation capabilities at a larger sized feed material to maximise the opportunity of running parallel material pre-treatment stream of both ore and waste material. An

175 example of proposed parallel processing streams is given in Figure 7.1. This would also aid in the potential of using mineral process tailings as a source feed for direct aqueous mineral carbonation.

7.1.2 Experimental conditions

Varying the temperature, CO2 partial pressures and solution chemistry have all proven to enhance conversion extents using a direct aqueous mineral carbonation regime. 185°C was shown to be the optimum temperature using olivine as the principal substrate material using the standard buffered solution at ARC. As such, this temperature was used and there is no further knowledge gained on the subject from this dissertation.

Experimental mineral carbonation was undertaken at lower CO2 partial pressures than commonly published in articles from other research institutes. This is due to the author not having access to a gas booster pump. The consequence of this was CO2 had to be pumped into the sealed vessel at pressures equaling the CO2 cylinder pressure. This pressure elevated as the reaction vessel was heated to the required 185°C for testing. An average 65 ± 5 bar pressure was maintained for the duration of testing. This is approximately 50 bar lower than the optimised direct aqueous mineral carbonation conditions proposed from research at ARC. Undertaking experimental mineral carbonation at these lower pressures provide some valuable information and insight into the development of direct aqueous mineral carbonation. Average reaction extents of 19.1 % and 37.6 % for Turnagain and Twin Sisters reactant material was achieved in the batch stirred reactor for a 2 hour testing duration in this research maintaining the 65 bar CO2 partial pressure.

McCoy and Rubin (2008) suggest that CO2 be transported at pressures greater than 86 bar. This is to avoid sharp changes in compressibility that may occur as a function of temperature (McCoy and Rubin,

2008: pg. 220). For mineral carbonation to become a viable CO2 sequestration option a continuous flow reactor similar to that proposed by Penner et al. (2004) must be developed. CO2 supply for such an operation would have to be compressed at source and transported to the substrate material source i.e. the mine site. The successful storage of CO2 using waste rock by mineral carbonation at these lower pressures in this research is encouraging.

176

Figure 7.1 Proposed duel ore processing and mineral carbonation facility at Turnagain

A proposed parallel processing workflow for both ore processing and mineral carbonation of waste rock at Turnagain. The ore processing stream is modified from the design created by AMC consulting for a pre-feasibility study at the mine site (Source: modified after AMC Preliminary Economic Assessment, 2011: pg. iv). The mineral carbonation processing stream is comprised of two pebble crushers, two SAG mills and four ball mills. This would have the potential to process approximately 87,000 t/day waste rock to a continuous flow reactor for direct aqueous mineral carbonation.

177

The reaction extents achieved can be considered a conservative estimate and the reaction extents that could potentially be achieved in a continuous flow system. It is proposed here that continuous flow reactors for direct aqueous mineral carbonation should target testing mineral carbonation capabilities at these lower pressures to consider a CO2 supply that would not need further compression. A benefit of such system would be a more simplified reactor design and reduced capital expenditure and energy requirements. Two buffered solutions and one sample using just distilled water as a reactant material carrier were used in testing Twin Sisters olivine in this research. The sample tested using just distilled water as a substrate carrier showed no appreciable mineral carbonation with an Rx 0.9 % after two hours of testing. This confirms that a buffered solution in slurry generation must be used within a direct aqueous mineral carbonation regime. Increased magnesite content in the product material was measured using a 2.5 M

NaHCO3/1 M NaCl solution chemistry over a 0.64 M NaHCO3/1 M NaCl solution chemistry. This equated to a magnesite content of 84.4 wt.% and 74.8 wt.% respectively. This is compared to a normalised 60.1 wt.% magnesite content under the same experimental conditions for a 6 hour testing duration. These results are consistent with carbonation extents achieved by Chizmeshya et al. (2006; 2007). They show that increased carbonation extent can be achieved by increasing the molar concentrations within the buffered solution at lower CO2 partial pressures. A problem encountered with using the increased molar concentration of the buffered solution was the loss of cations that made up the solution into the solid product. Minerals derived from the solution chemistry were found in the XRD analyses of the solid product. This raises an issue on whether the buffered solution could be effectively recycled for use in a continuous flow mineral carbonation process. Without effective recycling of the carrying solution, the reagent costs for a mineral carbonation facility could be significant. Furthermore the large quantity of minerals added in the solid product recovered post-carbonation made the reaction extent of olivine to magnesite conversion difficult to calculate. Mineral occurrences derived from the carrier solution in samples tested using the standard 0.64 M

NaHCO3/1 M NaCl solution proposed by ARC did not contribute significantly to the XRD analyses. As such, the solution chemistry proposed remains the most practical developed to date. It is postulated that the magnetic stirrer within the autoclave reaction vessel was effective in agitating the slurry to further enhance the creation of fresh reaction surfaces for the most part, however some material occasionally collected directly underneath the stirrer and likely did not receive the required degree of agitation. It is assumed that increased particle to particle attrition and particle abrasion against the reaction vessel walls could apply additional stresses to particles aiding in reduction

178 or removal of the silica-rich passivating layer. This is widely attributed as an inhibiting factor in direct aqueous mineral carbonation. An interesting chemical aspect noticed from the Turnagain samples tested is the possible influence of the sulphide content in the reactant samples on the stoichiometric forsterite to magnesite conversion extent. Samples showing the greatest mineral carbonation potential had the highest sulphide content determined by XRF and ICP-ES testing. This may be a consequence of the sulphides lowering the pH of the solution chemistry. This allows more Mg2+ to be leached from forsterite and thus being able to bind with CO2 to form magnesite. However this lower pH is expected to limit the ability of magnesite to form as magnesite tends to favour more alkaline solution chemistry for precipitation. Unfortunately it was not possible to measure the pH as the experiments were run. There was no noticeable variation in the pre- and post-experiment pH that may aid in explaining the sulphide influence on mineral carbonation. Little, if any, research has been undertaken into the influence of sulphide concentration on direct aqueous mineral carbonation. Sulphides are a common accessory mineral in mining waste rock and the findings from experimental mineral carbonation in this research, warrants further investigation into this.

CO2 was stored in magnesite in all samples tested from Turnagain. This ensures that any CO2 sequestered using direct aqueous mineral carbonation is stored in a safe and thermodynamically stable form. This will permit the safe disposal of the processed product to be safely back filled into the mine at Turnagain as part of reclamation requirements. The magnesite formed is extremely fine grained. Detailed investigation into the use of the product material post-processing remains to be undertaken. All of the product minerals are benign and no challenging or environmentally undesirable minerals were formed by the process within the samples tested. The potential for creating a value added product could help to promote mineral carbonation technology. If as an example, the product can be confirmed as unreactive, it could be used as filler. Furthermore it may be possible to control the reaction conditions so as to create magnesite and silica to meet required characteristics for a saleable product. Although not directly associated with the modelling of mineral carbonation potential data and the principal focus of this research. Findings although not novel on the controlling factors and outcomes derived from experimental mineral carbonation in this research both validate the experimental equipment setup and process method undertaken in this research.

7.1.3 Mineral carbonation extent determination Dunite reactant samples tested from Turnagain in this research show serpentine contents of up to 30 %. Serpentine (lizardite) contributes little if any MgO for carbonation. The disordered nature of

179 serpentine means that it often is not identified in X-ray diffractograms. A consequence of this when considering dunite as the reactant rock for MCP estimation is the possibility of incorrectly assuming that all the MgO determine by lithogeochemical analyses within a reactant sample is contained within olivine. Serpentine abundance was back-calculated from the degree of crystallinity assuming it to comprise all the amorphous material in the samples. This quantification of serpentine provides data that can be used as a validation tool for mineral estimation discussed in Section 7.2. It also provides information that can be used to assess the effectiveness of the selected experimental carbonation method on the reactant samples. In particular is olivine, serpentine or both providing MgO for carbonation and if so, how much? XRD with Rietveld refinement allows the direct identification and quantification of magnesite at

Turnagain in the post-carbonation product. This analysis quantifies the amount of CO2 stored within any additional minerals that may have formed as a result of the experimentation. This has an advantage over confirming the identification of magnesite by qualitative XRD and quantifying the amount of CO2 stored through lithogeochemical analyses such as ICP-ES. XRD with Rietveld refinement avoids making the assumption that all the CO2 is thermodynamically stable and stored within magnesite. This assumption is often made using lithogeochemical analysis. This is important when evaluating waste rock on its applicability to sequester CO2 on a geologic timescale. Choosing between a mass elemental balance and a mass mineralogical balance approach to stoichiometric carbonation extent determination is subject to the level of detail needed from the analysis and mineralogical homogeneity of the rock mass being studied. In homogenous dunite deposits such as Twin Sisters, a mass elemental balance approach may be enough. This is due to the confidence on the source of the MgO concentration being forsterite at the deposit is high. This also permits bulk sampling at various locations throughout the deposit providing the substrate material for laboratory testing. Techniques such as ICP-ES for elemental analysis and XRF for oxide analysis can be used to

2+ calculate the concentration of Mg and CO2 in both the substrate and product material. This provides the necessary information to calculate quantity of CO2 that can be stored through mineral carbonation. The principal magnesium silicate reactant and magnesite in the product material can be identified qualitatively by analyses such as XRD. Deposits exploited by mining operations are typically weathered and have undergone alteration to the original mineralogy. When considering such a deposit e.g. Turnagain it is proposed that a mass mineralogical balance approach to determine mineral carbonation extent is required. This decision is largely controlled by the variable effectiveness of different magnesium silicate minerals as a substrate

180 source for direct aqueous mineral carbonation. The increased cost associated in testing both reactant and post-carbonation product samples by XRD with Rietveld over a cheaper lithogeochemical approach is not preferential. It is yet proposed based on research undertaken in this dissertation that it can provide information essential to MCP estimation. The quantification of the mineralogical source of all the MgO within each reactant sample tested aids in selecting an appropriate mineral carbonation pre- treatment and processing regime. It furthermore can be used to accurately determine the mineral carbonation extent achieved by each individual mineral from the source rock. Finally it can be used to confirm and quantify the mineralogical location of the CO2 sequestered and provide a detailed analysis of the product material to be backfilled at the mine or assessed as a potential value-added product. An experimental mineral carbonation workflow that can be adopted by a mining company wishing to test their waste rock for use in direct aqueous mineral carbonation is provided in Figure 7.2. This workflow adopts the mineralogical mass balance approach taken in this research. It is suggested that approximately 50 representative samples would be enough for a detailed analysis. This is an arbitrary number but considered by the author to provide sufficient information on the waste rock on its mineral carbonation capabilities. The number also reduces the time and cost of analyses. This is tailored around a preliminary investigation needed at this stage. The relative infancy of experimental mineral carbonation process design means there are no direct services available for mining companies to test the rocks at their deposit. Sample pre-treatment i.e. crushing and grinding can be undertaken by the laboratory used for lithogeochemical analyses. The experimental mineral carbonation methodology outlined in Chapter 3 of this research and undertaken on samples from Turnagain can be used to undertake conservative estimates of the rocks mineral carbonation capabilities using a direct aqueous regime. The Hastelloy-C Parr reactor (Figure 3.6) used can be brought directly from an autoclave supplier. It does not need any modification for testing. This allows the testing to be undertaken at the mining operation directly by the company or can be replicated at a research institution. It is recommended that XRD with Rietveld refinement analyses be undertaken by a certified laboratory with experience in testing rocks rich in olivine and serpentine.

181

Figure 7.2 A proposed experimental direct aqueous mineral carbonation workflow for testing mine waste rock

Drill core sampling Recommended 50 representative samples for the rock types targeted

Sample pre-treatment -Crushing and grinding until > 80 % is less than 38 µm

Experimental mineral carbonation XRD with Rietveld refinement analyses 2 hour testing in a stirred batch autoclave using the following Quantitatively determine the conditions: mineral proportions in the reactant material 185°C @ 65 bar PCO2 0.64 M NaHCO3/1 M NaCl solution

XRD with Rietveld refinement analyses Quantitatively determine the mineral proportions in the reactant material

Table 7.1 provides a guideline cost estimate for mining companies to calculate the quantity of CO2 using their waste rock by direct aqueous mineral carbonation. The estimate is based on the capital expense of purchasing a bench-top stirred reactor, basic experimental equipment, reagents to create the buffered solution and a pure CO2 gas cylinder. The cost is based on a 100 ml autoclave being used. It is suggested that a larger autoclave be used providing there is sufficient reactant material being available. The cost to get reactant and product material analysed is based on 50 samples tested. The cost of testing does not include any labour costs for the personnel undertaking the experimental mineral carbonation. A realistic time estimate for each mineral experiment run is 6 hours. The estimated CAD$ 45,630, including capital equipment investment is a cost effective method of determining the mineral carbonation capabilities of mining waste rock. This is supported by the increase in information supplied on the reactant material gained by XRD analysis with Rietveld refinement. The experimental workflow

182 additionally provides information on the expected mineralogical makeup of the post-carbonation product to be stored at the mine site.

Table 7.1 Guideline cost estimate for experimental direct aqueous mineral carbonation

Requirement Estimated cost (CAN$) Capital costs 25,000 Autoclave Experimental equipment 300 Solution reagents 50 Carbon dioxide 100 Sub-total $ 25,450 Reactant analyses Crushing and grinding 180 XRD with Rietveld refinement 10,000 Product analyses XRD with Rietveld refinement 10,000 Sub-total $ 20,180 Total $ 45,630

The cost guideline assumes 50 reactant and product samples are being analysed. The guideline prices estimated are based on the author’s experience.

7.2 Predicting mineral carbonation capabilities from lithogeochemical data

At the onset of the research the following question was proposed:

Can a mineral carbonation potential (MCP) parameter be developed that combines minimal laboratory- scale experimentation and available exploration data at mining operations to generate an acceptable mineral carbonation potential estimate for discrete rock units at a mining operation?

An MCP calculator was developed in this research that was shown to successfully generate MCP values from lithogeochemical data generated that is representative of regular intervals of diamond drill holes at the Turnagain ultramafic deposit in northern British Columbia, Canada. This MCP calculator can convert measured elemental quantities to oxide values and then estimate the quantities of the defined mineral assemblage at the deposit. An example of combining mineral carbonation extents to the modal mineral abundance of the samples was displayed to generate MCP values. The MCP calculator is not ubiquitous to all deposits and would require customisation on a deposit by deposit basis to generate the

183 modal mineral abundance due to its stoichiometric linear method of estimating each mineral. The generation of the MCP values is restricted to a single mineral carbonation efficiency for each mineral considered. The calculator is flexible in that olivine and/or serpentine can be targeted as the principal substrate mineral source.

The RCO2 parameter developed by Lackner et al. (1995) can be used to calculate the theoretical capacity of a mineral substrate to sequester CO2. This parameter can be successfully applied at a deposit scale in estimating the theoretical absolute capacity of an ultramafic deposit to sequester CO2 by mineral carbonation. This was achieved for numerous ultramafic deposits in North America by Goff et al.

(2000). The RCO2 parameter remains a useful tool that can be used to target deposits capable of supplying the quantity of rock required to feed a mineral carbonation operation. The parameter is particularly effective when the mineralogical source of MgO and CaO can be confidently assumed. This is true for the almost exclusive olivine source of all the measured MgO abundance at relatively unaltered dunite bodies such as Twin Sisters. The challenges of estimating the MCP of heterogeneous ultramafic deposits such as those commonly associated with mining operations was discussed in detail in Chapter 5. It is proposed by the author that like determining the carbonation extent from experimental mineral carbonation on mining waste rock a mineralogical approach should be taken. This is justified by the variable mineral carbonation extents achieved using direct aqueous mineral carbonation on magnesium silicate substrate material. The typically altered nature of ultramafic deposits associated with mining waste rock at deposits such as Turnagain is also a defining factor. Deposits that are explored and exploited for minerals of economic importance have a unique set of lithogeochemical assay results that provide the major cation (Mg2+ and Ca2+) abundances at known locations within the rock mass. This data should be used where available to generate a more informed calculation of the mineral carbonation capacity. Whole-rock geochemical analyses typically available at mining operations do not however contain any silica (SiO2) quantity and LOI (H2O + CO2) values. Existing modal mineral estimation techniques available in the literature were reviewed by the author in Chapters 2 and 3. When attempting to predict the mineral carbonation potential of an ultramafic rock mass, at least the modal abundance of olivine and serpentine must be estimated. The modal mineralogy estimation programs reviewed require a suite of oxide values for estimation. Major oxide analyses by XRF of the 16 reactant samples from the Turnagain ultramafic deposit plus a representative reactant sample from the Twin Sisters ultramafic complex were used to estimate the modal mineral abundance using the MINSQ program devised by Herrmann and Berry (2002). A

184 simplified 5 mineral assemblage was estimated based on the minerals identified within the samples from XRD with Rietveld refinement analysis. The results of the testing showed that providing a successful fit with a residual SSQ of less than 1, major oxide analyses used in the MINSQ spreadsheet program could successfully estimate the modal mineral composition of ultramafic rocks. The reliance however of current modal mineral estimation methods such as MINSQ on oxide values and LOI data for the estimation of olivine and serpentine means that XRF and LOI analyses must be completed. Although XRF analyses are significantly cheaper than XRD with the required Rietveld refinement analysis, major oxide data is often not available from rock samples tested for resource estimation purposes at mining operations. This coupled with the time consuming nature of adjusting each sample based on rock type and manually entering sample data on an individual basis means that it will be difficult to use a modal mineral estimation program such as MINSQ to thousands of mining waste rock samples. The MCP calculator developed in Chapter 5 of this dissertation provides a novel approach to estimating the modal mineral abundances at ultramafic deposits such as Turnagain. The calculator is specifically targeted toward the mining industry and coupling existing and financially accounted for lithogeochemical data with information and data collected from the experimental mineral carbonation workflow discussed in the previous section. The advantage of using the MCP calculator is that it considers the variability in silicate mineralogy throughout a rock mass as the source of the major cation targeted for use in mineral carbonation. The MCP calculator is not ubiquitous to all ultramafic deposits and requires adaptation on a deposit by deposit basis. This MCP calculator would need to be customised to account to the specific mineral assemblage local to the deposit. This is an unavoidable step in detailed MCP estimation. Like resource estimation undertaken by the mining industry, there is no one-size-fits-all approach. The program developed in this research provides an example on how MCP estimation can be approached. It uses the available data at the deposit to provide a best estimate of the carbonation capabilities. The MCP calculator provides a platform for converting measured major cation quantities in to major oxides for mineral estimation. This is no substitute for direct oxide measurements for analytical methods such as XRF. But it is proposed that the values generated are within an acceptable degree of error i.e. less than 5 % between estimated and measured oxide values on the 16 samples tested from Turnagain. This degree of error is deemed acceptable when considering that no additional expense is required from analytical testing. The mean absolute error for serpentine estimation using the MCP calculator in comparison to XRD analyses is 11 wt.% and 12 wt.% for olivine estimation.

185

A number of factors can contribute to the estimation error using the MCP calculator. The estimates of the 16 samples tested from Turnagain had a lower estimation error on the mineralogy when the oxide conversion part of the calculator less than or close to 1 % from that measured by XRF. This is an unavoidable error without directly measuring the oxide values. The acceptance of a best estimate oxide conversion from lithogeochemical data must be accepted for modal mineral estimation using only elemental data. A further source of error could be the semi-quantitative estimation of serpentine by XRD with Rietveld refinement. The serpentine content is calculated in XRD with Rietveld refinement analyses by renormalising the total mineral weight proportions. This is based on the degree of crystallisation assuming all amorphous material is serpentine. This means that the serpentine content with which the MCP calculator results are being evaluated against are only semi-quantitative. As such it is difficult predict the associated degree of error. This highlights the difficulty in estimating or even measuring serpentine content within ultramafic waste rock. An accurate serpentine estimate has emerged as a key determining factor on the mineral carbonation potential of the rocks in this research. The guideline mean estimation errors of 11 wt.% and 12 wt.% when compared with semi- quantitative results of the same samples appear to be a reasonable expectancy. One key feature of the results is the ability of the MCP calculator to successfully determine samples that have high olivine contents measured by XRD with Rietveld refinement. This is supported by the analysis of samples tested from the Twin Sisters ultramafic complex. Here the calculator determined extremely high olivine content with an acceptable estimation error. Although some error is expected when estimating the olivine and serpentine contents of rock samples, the MCP calculator program, it is an effective tool at estimating the amount of serpentinisation throughout an ultramafic deposit from lithogeochemical data than can be done qualitatively. When comparing the average estimation difference in MCP values between those estimated by the MCP calculator program and those projected using the same reaction extent of 37 % from quantitative XRD, there is little difference at 5.15 % to 4.89 % respectively. Both the MCP calculator values and those projected from reactant XRD are consistent throughout, typically ranging between 10 % and 20 %. The significant differences observed between estimated and measured values appear to be a function of the variability in the mineral carbonation extent achieved during experimentation. Further experimental testing would help to better predict the carbonation extent that could be achieved by the direct aqueous mineral carbonation process. A number of the samples tested, namely 04-24-AJ3, 06-110-AJ2, 06-110-AJ17, 06-111-AJ1, 06-116- AJ9, and 06-116-AJ12 all were estimated within 1 % of what was observed from experimental mineral

186 carbonation. The relative consistency of estimated MCP values and those projected from XRD suggest and promising estimation results when compared with those observed from experimental mineral carbonation. The MCP calculator shows to be an effective tool at predicting MCP behavior. This is providing the variable nature of experimental mineral carbonation is considered. The flexibility of the MCP calculator to apply individual carbonation extents on a mineral by mineral basis allows the user to undertake a sensitivity analysis on the MCP controlled by the mineral carbonation process. It allows a range of carbonation extents to be applied which can account for improvements in the mineral carbonation process.

7.3 Estimating the mineral carbonation potential of ultramafic deposits

At the onset of the research the following question was proposed:

Can an MCP parameter predicting the mineral carbonation potential of a discrete unit of rock be geospatially interpolated to provide a preliminary resource estimation of the CO2 capacity of mining waste rock?

The focus and motivation of predicting MCP values in three-dimensional space was to provide a next stage estimation of the potential of an ultramafic deposit and more specifically proposed mine waste rock. MCP values generated using the MCP calculator developed in this research were successfully interpolated to create a global MCP resource estimation at Turnagain. This improves on simply providing a theoretical CO2 capacity based on a rock volume assumed to have a uniform mineralogy, density and MgO content. 11,552 samples from 196 drill holes at Turnagain analysed by ICP-ES provided a lithogeochemical database that was used to determine the MCP values of regularised length intervals of the drill core. An approach of undertaking mineral carbonation estimation on a lithology by lithology basis over MCP value based modelling was chosen due to MCP values being a product of the rock type and mineralogy of that rock type. Resource estimates generated in this research are constrained within wireframes designed by the author modelling three identified geological domains at the deposit. Improved computing power and the development of three-dimensional design software provides a tool by which the dimensions of individual rock typed can be created. This is not a novel concept but can be applied as shown in this dissertation to more accurately determine the volume and tonnage of rock types capable of

187 sequestering CO2 at ultramafic deposits by mineral carbonation than can be achieved by predicting the volume of each rock type from surface mapping alone. Wireframe modelling of rock types of similar characteristics are often created by mining companies from exploration drilling for resource estimation purposes. Where available the use of these coupled with lithogeochemical assay data measured inside the confined discrete rock units should be used to improve MCP estimation. This furthermore permits rock types of similar mineralogical characteristics to be evaluated individually. CaO, MgO, MCP and MCP_ARC values were interpolated in to the block model created at the Turnagain ultramafic complex to generate global resource estimation for each parameter. The values were interpolated using three estimation techniques; inverse distance, ordinary kriging and nearest neighbour estimation. The volumes, tonnages, CaO, MgO, MCP and MCP_ARC interpolated values for all three geological domains produced consistent comparable results for all three interpolation methods analysed. Similar to more traditional commodity resource estimation the approach taken must be deposit specific. In this example ordinary kriging is preferred over inverse distance weighting due to the lack of value reconciliation available without the aid of a pilot plant to process material from the mine and due to the confidence in the variograms generated. Prior to MCP estimation both CaO and MgO values were interpolated using the available drill hole data. These data values are commonly available from lithogeochemical analyses at mining deposits and can be used to provide a more detailed evaluation of the RCO2 of the potential waste rock. The RCO2 values were derived based on the assumption that the estimated MgO is fixed in olivine within the rock.

7 The theoretical capacity of the area studied to sequester CO2 based on this assumption is 5.2 x 10 t CO2. The estimate shows that 2.25 tonnes of rock that would have to be processed to sequester one tonne of

CO2 utilising a 100 % efficient mineral carbonation conversion. This estimate requires no extra sampling, geochemical analyses or experimental mineral carbonation. It provides a more informed resource estimate than applying an RCO2 to the complex as a whole assuming a uniform geology and mineralogy. At a minimum, mining companies wishing to evaluate their waste rock for potential use in mineral carbonation should use the available data to provide a theoretical CO2 capacity and the amount or rock required to trap a tonne of CO2. The first attempt at undertaking a resource estimate of MCP values at Turnagain provide a valuable insight into the requirements for future attempts at similar deposits. Semi-variogram analysis was undertaken to characterise the spatial correlation of 15 m composited MgO, CaO, MCP, and MCP_ARC values. Variography undertaken on all rock types evaluated showed extremely good continuity

188 throughout the deposit. The value distribution was similar to that of a porphyry copper, iron ore, or industrial minerals deposit. The search ellipsoid used to locate samples for resource estimation was nearly isotropic for all parameters estimated in each rock type. If any anisotropy was found, the major axis typically followed the direction of the greatest drill hole spacing. The consequence of this finding is that for the level of detail currently required is that mining companies wishing to undertake an MCP resource estimate at this stage would likely not need to undertake variography. This is significant in that it would greatly reduce the time, skill, experience required to create an accurate estimation. Ordinary kriging requires the use of a variogram for the estimate to be made. For this reason, an inverse distance estimation method is suitable for estimating MCP values based on the Turnagain deposit. This however may no longer remain true when further deposits have been evaluated.

The mass of CO2 that can potentially be sequestered by mineral carbonation when estimating MCP and MCP_ARC values at Turnagain is significantly lower than that of the theoretical capacity. The total

6 6 modelled amount predicted is approximately 31 x 10 t CO2 when using the MCP values and 64 x 10 t

CO2 when using the MCP_ARC values. These values signify the importance of developing an industrial mineral carbonation process that can achieve the carbonation extents achieved at ARC. They also validate the importance of applying the effect of a heterogenous mineralogy based on the select direct aqueous mineral carbonation process. The estimate generate also highlights the impact on the amount of material that must be mined to sequester a tonne of CO2. This is expressed by approximately 16 tonnes of rock required using the mineral carbonation process in this research and 8 tonnes of rock using the most successful carbonation process conditions to date. This has implications on the practicality of developing a mineral carbonation facility at Turnagain when compared with the theoretical 2.25 tonnes of rock that would have to be processed to sequester one tonne of CO2 utilising a 100 % efficient mineral carbonation conversion. A feasibility study on the potential for a mineral carbonation operation at Turnagain is beyond the scope of this research. The potential mass of CO2 that could be sequestered using proven carbonation conversion extents at a laboratory scale is promising. It appears theoretically feasible that ultramafic deposit’s such as Turnagain could use their waste rock to offset the CO2 emissions of a 500 MW coal- fired plant for its planned life of mine. Consideration must be made that this estimate considers all rock within the mine design and does not discriminate between ore and waste rock. Furthermore, concerns must be raised at the quantity of rock required that must be processed to sequester one tonne of CO2 at 16 tonnes and 8 tonnes for the MCP and MCP_ARC values. This is of particular concern due to the cost

189 and energy requirements of mining, crushing, grinding and processing the quantity of rock to sequester such little CO2.

190

8 Conclusions

8.1 Research outcomes

Three pertinent questions on the evaluation and estimation of mining waste rock for use as a mineral carbonation resource were defined in the early stages of this research. The research outcomes aim to connect the three questions and define the findings in the context of taking lithogeochemical data analysed from drill core at mining operations to generating a mineral carbonation resource estimate. This considers the mineralogy of the host rock and its effectiveness as a substrate for direct aqueous mineral carbonation. Experimental direct aqueous mineral carbonation was undertaken on proposed Turnagain waste rock, a low-grade, high tonnage nickel sulphide deposit in northern British Columbia. The direct aqueous mineral carbonation regime proposed and developed at ARC remains the leading process method when using olivine as the principal mineral source. A maximum 45 % magnesium silicate to magnesium carbonate conversion was achieved using dunite waste rock from Turnagain in two hours. From this research it is proposed that the principal mineral targeted as feed material for an industrial direct aqueous mineral carbonation process should be olivine. This is exclusively due to olivine not requiring the costly and energy expensive heat pre-treatment. A direct aqueous mineral carbonation processing pathway and waste rock from a nickel sulphide mine were the focus of this research. However, the rationale and techniques employed are not ubiquitous to this selected processing pathway or the specific ultramafic deposit. The research aims to initiate further studies in the topic area by providing an example of how experimental mineral carbonation can be taken and applied directly to the assessment of potential resources (mining waste rock). Furthermore, how this integrated approach can both aid in advancing mineral carbonation technology and provide an industry specific approach to reducing carbon dioxide emissions.

Experimental mineral carbonation was undertaken at lower CO2 partial pressures than commonly published in articles from other research institutes. An average 65 ± 5 bar pressure was maintained for the duration of testing. Undertaking experimental mineral carbonation at these lower pressures provide some valuable information and insight into the development of direct aqueous mineral carbonation. An average reaction extent of 19.1 % and 37.6 % for Turnagain and Twin Sisters reactant material respectively was achieved. Successful mineral carbonation of ultramafic mining waste rock at this lower pressure than the defined standard reaction proposed by ARC is significant. As stated, a concern for the

191 development of industrial mineral carbonation is the high cost and energy requirements attributed to processing. Despite achieving relatively low carbonation extents, carbonation at these lower pressures could assist in the development of a continuous flow reactor design as the reaction pressure is similar to that in which CO2 would be transported via a pipeline to a mineral carbonation processing facility from source. This can help to overcome additional gas compression and the maintenance of these pressures for effective carbonation.

A buffered standard solution of 0.64 M NaHCO3/1 M NaCl proposed by ARC was used in the experiments on Turnagain rocks. The solution chemistry proposed was successful in enhancing mineral carbonation from slurry comprising solely of rock and water. This solution was effectively recycled, which was not the case when a higher molarity NaHCO3 solution was used, proposing that the slurry used could be recycled within an industrial mineral carbonation process. Samples showing the greatest mineral carbonation potential had the highest sulphide content determined by XRF and ICP-ES testing, this was unexpected and warrants further investigation through future research. All samples tested underwent comminution to crush and grind the rock samples to under 37 µm for experimental mineral carbonation. No knowledge was gained on the effect of the pre-treatment methods chosen on mineral carbonation efficiency. This remains the focus of the mineral carbonation research group at UBC. It remains a key research topic on improving mineral carbonation efficiency and reducing associated cost and energy requirements.

CO2 was stored in magnesite in all samples tested from Turnagain. This ensures that any CO2 sequestered using direct aqueous mineral carbonation is stored in a safe and thermodynamically stable form. The successful sequestration of CO2 using mining waste rock and the opportunity of shared mining and mineral processing costs of a dual mining and mineral carbonation operation can aid in reducing both the economic and energy requirement of industrial mineral carbonation. Choosing between a mass elemental balance and a mass mineralogical balance approach to stoichiometric carbonation extent determination is subject to the level of detail needed from the analysis and mineralogical homogeneity of the rock mass being studied. In homogenous dunite deposits such as Twin Sisters, a mass elemental balance approach may be enough. This is due to the confidence on the source of the MgO concentration being forsterite at the deposit is high. Deposits exploited by mining operations are typically weathered and have undergone alteration to the original mineralogy. When considering such a deposit e.g. Turnagain it is proposed that a mass mineralogical balance approach to determine mineral carbonation extent is required. This decision is largely controlled by the variable effectiveness of different magnesium silicate minerals as a substrate source for direct aqueous

192 mineral carbonation. A cost-effective experimental direct aqueous workflow was developed that can be used by mining companies to evaluate the potential of their waste rock to trap CO2 by the mineral carbonation process. It is apparent that each deposit considered as a substrate source for mineral carbonation must be examined in detail. A key component of this examination is experimental mineral carbonation. The mineral carbonation extent and process efficiency must be defined for the particular deposit being examined due to unique mineral assemblages within each rock and highly variable mineral quantity spatially within each rock type. This experimentation can help customise a mineral carbonation approach for that deposit and aid in the development of mineral carbonation technology as demonstrated in this research. The inherent variability in carbonation extents measured from mineral carbonation is proposed as a product of an inconsistency in the effectiveness of direct aqueous mineral carbonation process route at exposing sufficient Mg2+ from olivine for carbonation and creating the optimal environment for magnesite precipitation. Only through continued detailed experimentation on rocks from a variety of ultramafic deposits can the process be improved and optimised for use at an industrial scale. This finding exposed the lack of understanding and investigation in to mineral variability within ultramafic rocks and the challenges that this poses to estimating and quantifying the potential of the rocks to sequester CO2 by mineral carbonation. The heterogenous mineralogy of ultramafic deposits hosting mining operations makes quantifying the mineral carbonation potential (MCP) of the waste rock difficult. This is attributed to olivine achieving significantly higher carbonation capabilities than serpentine. The MCP calculator, a new and unique series of Microsoft ExcelTM spreadsheets was developed to estimate the modal mineral abundance of ultramafic rocks for use in MCP estimation. The program is targeted for use by the mining industry utilising abundant lithogeochemical data as a cost-effective tool in evaluating their deposit as a supplier of substrate material to an industrial mineral carbonation operation. The program supplementally has the ability to apply mineral carbonation extents determined from experimentation on the rock to generate MCP values i.e. the amount of magnesite that can formed by a unit of waste rock. The program can be tailored to estimate MCP values based on a site specific mineral assemblage and measured carbonation extents achieved. MCP values generated represent sampled length intervals along exploration drill holes. The MCP calculator and the generated MCP values provide an initial attempt at maximising the use of lithogeochemical data available at mining operations to create a more informed CO2 capacity

193 estimate based on the rocks available. It is flexible in that it can be customised to estimate a defined mineral assemblage identified at the deposit being investigated. It directly incorporates the mineral carbonation extents achieved on the rocks through experimentation and is flexible to advances in improvements to the mineral carbonation process. It has limitations in that the oxide values used for modal mineral estimation are converted from elemental measured values rather that direct oxide measurements and is sensitive to the inherent variability of the carbonation process. However, its use of existing data and no additional geochemical testing requirement is promising. It provides a platform by which mining companies can maximise the use of their data to provide a more informed estimate of their waste rock to trap CO2 within a reasonable degree of error at no added financial expense. Estimation techniques traditionally used by the mining industry for resource estimation were evaluated as methods of geospatially interpolating MCP values. Both inverse distance and ordinary kriging were successful in interpolating MCP values. Similar to more traditional commodity resource estimation the approach taken must be deposit specific. Variography undertaken on all rock types evaluated from Turnagain showed extremely good continuity throughout the deposit. The value distribution was similar to that of a porphyry copper, iron ore, or industrial minerals deposit. Mining companies wishing to undertake an MCP resource estimate at this stage would likely not need to undertake variography. This would greatly reduce the time, skill and experience required to create an accurate estimation. An inverse distance estimation method is suitable for estimating MCP values at this time based on this research completed at the Turnagain deposit.

The mass of CO2 that can potentially be sequestered by mineral carbonation when estimating MCP and MCP_ARC values at Turnagain is significantly lower than that of the theoretical capacity. The total

6 6 modelled amount predicted is approximately 31 x 10 t CO2 when using the MCP values and 64 x 10 t

CO2 when using the MCP_ARC values. The estimate generate also highlights the impact on the amount of material that must be mined to sequester a tonne of CO2. This is expressed by approximately 16 tonnes of rock required using the mineral carbonation process in this research and 8 tonnes of rock using the most successful carbonation process conditions to date. This has implications on the practicality of developing a mineral carbonation facility at Turnagain when compared with the theoretical 2.25 tonnes of rock that would have to be processed to sequester one tonne of CO2 utilising a 100 % efficient mineral carbonation conversion. It appears theoretically feasible that ultramafic deposit’s such as Turnagain could use their waste rock to offset the CO2 emissions of a 500 MW coal-fired plant for its planned life of mine. Knowledge was gained through attempting to answer the three questions posed at the onset of the research. Firstly

194 it is important to undertake experimental mineral carbonation and develop a framework and process path that best suits the ultramafic deposit being investigated. This in turn aids in further understanding of mineral carbonation kinetics, efficiency, limitations toward reducing cost and energy requirements. The research secondly highlighted the importance of understanding and quantifying the mineralogy of ultramafic deposits when estimating their potential for mineral carbonation. By better classifying and quantifying the mineralogy of the proposed substrate rocks, a mineral carbonation process route can be designed that maximises the effectiveness magnesium silicate dissolution and subsequent magnesite precipitation. Furthermore by better quantifying the origin of the Mg2+ cations available for mineral carbonation, an estimate of the MCP of a discrete unit of rock can be derived. Finally by calculating the theoretical capacity of a given mass of rock to sequester CO2 by the selected mineral carbonation pathway, these MCP values can be interpolated spatially in three-dimensions to generate a global resource estimate of mining waste rock.

8.2 Suggestions for future research

Mineral carbonation as a method of sequestering CO2 would become more technically and financial feasible should a process stream be developed for mineral carbonation that can achieve comparable conversion extents currently achieved using a -38 µm feed with a -150 µm feed. This would also be beneficial when exploring the possibility of carbonating MgO-rich process tailings already reduced to this size fraction. The study of particle attrition and abrasion should be investigated to improve carbonation extents without having to use more aggressive chemical approaches. This will be of particular importance in maintaining constant particle agitation within a continuous flow regime. Direct aqueous mineral carbonation using the baseline reaction conditions proposed by ARC remain the most successful using olivine as the substrate mineral to date. It is not yet clear whether a one-step or two-step process would be preferable. A two-step approach would permit the reaction conditions to be more carefully controlled and optimised. It may also permit the use of solvents to aid magnesium silicate dissolution, reducing the high energy and expensive fine grinding currently required. It would furthermore provide the opportunity for parallel process streams focusing on different magnesium silicate minerals as the substrate feed. Experimental mineral carbonation remains using batch stirred reactors. A continuous flow design would be critical to the development of a pilot-scale and industrial- scale mineral carbonation operation.

195

The classification of product material post-carbonation has not been studied within the literature in great detail. The possibility remains that a saleable product can be formed or targeted from mineral carbonation that could enhance reduce the financial restrictions by adding a value-added product. In this research samples from Turnagain showed the greatest reaction extent where sulphides were identified in the reactant material. The role and effect of sulphides commonly found in deposits of economic interest in not yet understood and definitely warrants further investigation. Financial and energy concerns remain at the forefront of the development of mineral carbonation technology. As shown in this research the trapping CO2 is possible at pressures in which it would be delivered to a mineral carbonation facility from the point source. It may be pertinent to focus research efforts toward a process that minimises capital and operating expenditure, energy consumption and accepts a lower carbonation extent. An in-depth evaluation of the energy requirements of each mineral pre-treatment and mineral carbonation process would also be beneficial. This could help to attract investment in to mineral carbonation technology. This research proposed that the potential of waste rock at Turnagain to trap the CO2 emitted from a 500 MW power plant is an example of how a niche operation such as this could be a focus for such study. This research focused on mine waste rock for a nickel-sulphide mine at Turnagain, British Columbia. Voormeij and Simandl (2004) delineated deposits capable of supplying rocks to a mineral carbonation operation of which Turnagain was one. Further identification of ultramafic deposits should be located globally to enhance the possibility of research such as this to be undertaken. Particular focus should be applied to those ultramafic deposits capable of coupling either a metal or diamond mining operation with mineral carbonation.

The coupling of CO2 source and mining operations within an acceptable distance of one another is important in reducing the transport distance of CO2 to the mine site. It is possible that the development of a mineral carbonation plant at a mining operation could attract heavy emitting operations such as a coal-fired plant as a power source towards the location of the mine. This integrated approach to mining and power generation is an interesting direction in which a feasibility study would be useful in elevating mineral carbonation as a CCS option and could lead to investment in technology developments.

A first attempt at developing an MCP parameter to better estimate the CO2 sequestration potential of mine waste rock was made in this research. Using the MCP calculator on a greater number of deposits would help to evaluate the effectiveness of it as a tool in which mining companies can use. Feedback on the ease of customising the calculator on a deposit by deposit basis would help to improve the design and advance it toward a standard method of evaluating the MCP of mining waste rock. When

196 developing the MCP calculator it was extremely difficult to derive an accurate estimation of serpentine quantities in ultramafic rock. This becomes increasingly difficult when only partial estimates of serpentine abundance can be made using XRD with Rietveld refinement by assuming that all of the non- crystalline phases are serpentine. This makes the validation of any serpentine estimation tool extremely difficult to validate. An accurate technique to estimate of serpentine abundance in ultramafic deposits is considered essential to the advancement of MCP estimation. This would ideally permit both olivine and serpentine quantities to be determined simultaneously in a cost effective and timely manner. The successful use of drill hole geochemical data in this research to develop a more accurate resource estimate is encouraging. Further studies at various ultramafic deposits will help to understand how existing resource estimation techniques can be employed on classifying waste rock at mining operations as a potential substrate source for mineral carbonation than provided here.

8.3 Personal reflection

This research set out to bridge the gap between experimental mineral carbonation and defining a resource that can be targeted to further development of mineral carbonation technology toward an industrial scale. Numerous mines such as Turnagain have waste rock that is largely ultramafic and has the potential to be a good source of substrate material for a mineral carbonation operation. Much of the previous literature has targeted ultramafic resources that have the capabilities to sequester CO2 on a regional scale. The problem with targeting deposits on the scale are that a new mine must be designed and excavated to provide material that’s sole purpose is to sequester CO2, with no commodity of interest to heavy industries and much of the resources and funding to advancing the science is now diminishing. By quantifying the MCP of mining waste rock in this research, it is an important first step at collaboration between mineral carbonation technology, largely based in academia and a niche industrial operation that can benefit from the advancement in mineral carbonation technology. It is felt that through targeting such niche operations, the required financial support to advance the technology to a pilot-scale can evolve. This can benefit both the science and the reduction of CO2 emissions from a large

CO2 emitter that already has to mine the rock to be processed. The experimental mineral carbonation completed in this research exposed the lack of a clear direction in the progression of the technology to an industrial scale. There is no one-size-fits-all solution to successfully trapping CO2 in ultramafic rocks and a more focused approach toward deposits that could realistically be exploited in the next 10 years should be taken. This would assist in increasing the

197 understanding of the mineral carbonation process and really start to develop a framework from taking a series of laboratory scale experiments to something that is practically feasible at an industrial scale. In developing the MCP calculator it became clear that when examining the MCP of ultramafic rocks you cannot simply take an elemental approach where it is assumed that all of the cations capable of forming stable carbonates will be available to actually do so. Either a process stream that first leaches Mg2+ cations from silicate minerals and precipitates carbonates in a two-step process must be successfully developed or the quantification of olivine and serpentine within a single rock sample must be significantly improved. Ultramafic deposits have invariably been altered to some degree and there is no widely accepted or used method for quantifying the serpentine content of rocks. This was a challenge when attempting to estimate the serpentine abundance from ICP-ES data with no SiO2 or H2O values to do so. Even if these values were measured, it is difficult to provide a true quantification technique for measurement or estimation. This was evident with only semi-quantitative values from XRD with Rietveld refinement that could be used to validate the estimation. To truly succeed in the short-term, the scale of mineral carbonation being targeted must be reduced. Niche operations such as mines with ultramafic rocks and the opportunity to offset their CO2 emissions appear to be a good fit for developing mineral carbonation technology. Through understanding the mineralogy of the deposit rather than taking a purely elemental approach a mineral carbonation process stream that is specific to the suite of rocks being investigated can be developed. This better understanding of the substrate material being investigated and the development of a focused processing stream can aid in deriving a realistic economic and energy assessment. This can involve the scaling-up of the process to an industrial scale and can incorporate capital expenditure and mining costs already accounted for by the mine. Through this, the advancement of mineral carbonation technology will be inevitable via focused research and research requirements to improve the feasibility of developing such an operation will become more apparent.

198

Bibliography

Albarède, F., Provost, A., 1977. Petrological and geochemical mass-balance equations: An algorithm for least-squares fitting and general error analysis. Computers and Geosciences, 3, 309-326. Alexander, L., Klug, H.P., 1948. Basic aspects of X-ray absorption in quantitative diffraction analysis of powder mixtures. Analytical Chemistry, 20, 886-889. AMC Mining Consultants (Canada) Ltd., 2011. Turnagain project, Hard Creek Nickel: preliminary economic assessment. Project 711022, 2nd December 2011. Annels, A.E., 1991. Mineral deposit evaluation, a practical approach. Chapman and Hall, London, 436 pp. Armstrong, M., Champigny, N., 1989. A study on kriging small blocks. CIM Bulletin, 82, 128-133.

Bachu, S., 2002. Sequestration of CO2 in geological media in response to climate change: road map for

site selection using the transform of the geological space into the CO2 phase space. Energy Conversion and Management, 43, 87-102.

Back, M., Vosbeck, K., Kühn, M., Stanjek, H., Clauser, C., Peiffer, S., 2006. Pretreatment of CO2 with fly ashes to generate alkalinity for subsurface sequestration. Proceedings of the 8th International Conference on Greenhouse Gas Technologies, Trondheim, Norway, June 19-22. Baláž, P., Turianicová, E., Fabién, M., Kleiv, R.A., Briančin, J., Obut, A., 2008. Structural changes in olivine

(Mg,Fe)2SiO4 mechanically activated in high-energy mills. International Journal of Mineral Processes, 88, 1-6. Banks, R., 1979. The use of linear programming in the analysis of petrological mixing problems. Contributions to Mineral Petrology, 70, 237-244. Béarat, H., McKelvy, M.J., Chizmeshya, A.V.G., Gormley, D., Nunez, R., Carpenter, R.W., Squires, K., Wolf, G.H., 2006. Carbon sequestration via aqueous olivine mineral carbonation: role of passivating layer formation. Environmental Science Technology. 40, 4802-4808. Bertos, F.M., Simons, S.J.R., Hills, C.D., Carey, P.J., 2005. Accelerated carbonation of contaminated land and waste residues as a contribution to carbon sequestration. Proceedings of the 4th Annual Conference on Carbon Capture and Storage, Alexandria, Virginia, May 2-5. Bonfils, B., Julcour-Lebigue, C., Guyot, F., Bodènan, F., Chiquet, P., Bourgeois, F., 2012. Comprehensive analysis of direct aqueous mineral carbonation using dissolution enhancing organic additives. International Journal of Greenhouse Gas Control, 9, 334-346. Bryan, W.B., Finger, L.W., Chayes, F., 1969. Estimating proportions in petrographic mixing equations by least squares approximation. Science, 163, 926-27.

199

Carbon Capture and Sequestration Technologies at MIT. Carbon capture and storage database. Accessed March 24th 2014, from http://sequestration.mit.edu/tools/projects/index.html. Carr, J.R., Roberts, K.P., 1989. Application of universal kriging for estimation of earthquake ground motion: statistical significance of results. Mathematical Geology, 21 (2), 255-266. Carr, J.R., 1992. A treatise on the use of geostatistics for the characterisation of nonrenewable resources. Nonrenewable Resources, 61-73. Chizmeshya, A.V.G., McKelvy, M.J., Sankey, O.F., Wolf, G.H., Sharma, R., Bearat, H., Diefenbacher, J.,

Carpenter, R.W., 2002. Atomic-level understanding of CO2 mineral carbonation mechanisms from advanced computational modelling. Proceedings of the 26th International Technical Conference on Coal Utilizations and Fuel Systems, Clearwater, FL, March 6-9. Chizmeshya, A.V.G., McKelvy, M.J., Squires, K., Carpenter, R.W., Béarat, H., 2005. A novel approach to mineral carbonation: Enhancing carbonation while avoiding mineral pretreatment process cost. Annual Report 2004-2005, Arizona State University. Chizmeshya, A.V.G., McKelvy, M.J., Squires, K., Carpenter, R.W., Béarat, H., 2006. A novel approach to mineral carbonation: Enhancing carbonation while avoiding mineral pretreatment process cost. Annual Report 2005-2006, Arizona State University. Chizmeshya, A.V.G., McKelvy, M.J., Squires, K., Carpenter, R.W., Béarat, H., Jarvis, K., 2006. A novel approach to mineral carbonation: Enhancing carbonation while avoiding mineral pretreatment process cost. Annual Report 2006-2007, Arizona State University. Chung, F.H., 1974. Quantitative interpretation of X-ray diffraction patterns of mixtures. II. Adiabatic principle of X-ray diffraction analysis of mixtures. Journal of Applied Crystallography, 7, 526-531. Clark, T., 1975. Geology of and ultramafic complex on the Turnagain River, northwestern British Columbia. Unpublished PhD Thesis, Queens University, 453 pp. Clark, T., 1978. Oxide minerals in the Turnagain ultramafic complex, northwestern British Columbia. Canadian Journal of Earth Sciences, 42, 1893-1903. Cohen, D.R., Ward, C.R., 1991. SEDNORM – a program to calculate a normative mineralogy for sedimentary rocks based on chemical analyses. Computers and Geosciences, 17, 1235-1253. Coleman, R.G., 1977. Ophiolites. Springer-Verlag, New York, 229 pp. Cross, C.W., Iddings, J.P., Pirsson, L.V., Washington, H.S., 1932. A quantitative chemicomineralogical classification and nomenclature of igneous rock. Journal of Geology, 10, 555-6900. Davison, J., Freund, P. and Smith, A., 2001). Putting carbon back in the ground. International Energy Agency Greenhouse Gas R&D Programme, 26 pp.

200

De Caritat, P., Bloch, J., Hutcheon, l., 1994. LPNORM: a linear programming normative analysis code Computers and Geosciences. 20, 313-347. Dewey, J.F., 1976. Ophiolite obduction. Tectonophysics, 31, 93-120. De-Vitry, C., Vann, J., Arvidson, H., 2007. A guide to selecting the optimal method of resource estimation for multivariate iron ore deposits. Proceedings of the Iron Ore Conference, Perth, WA, August 20-22. Dipple, G.M., Raudsepp, M., and Gordon, T.M., 2002. Assaying wollastonite in skarn. In Industrial Minerals in Canada, Canadian Institute of Mining, Metallurgy and Petroleum Special Volume, 53, 303-312. Dlugogorski, B.Z., Balucan, R.D., 2014. Dehydroxylation of serpentine minerals: implications for mineral carbonation. Renewable and Sustainable Energy Reviews, 31, 353-367. Dufaud, F., Martinez, I., Shilobreeva, S., 2009. Experimental study of Mg-rich silicates carbonation at 400 and 500°C and 1 kbar. Chemical Geology, 265, 79-87. Dunham, S., Vann, J., 2007. Geometallurgy, Geostatistics and project value – Does your block model tell you what you need to know?. Proceedings of the Project evaluation Conference, Melbourne, Victoria, June 19-20.

Eloneva, S., Teir, S., Savolahti, J., Fogelholm, C.-J., Zevenhoven, R., 2007. Co-utilization of CO2 and calcium silicate-rich slags for precipitated calcium carbonate production (part II). Proceedings of the ECOS’2007, Padova, Italy, June 25-28. Environment Canada, 2012. Canada’s emission trends 2012. Environment Canada report (August, 2012), 69pp. Fabian, M., Shopska, M., Paneva, D., Kadinov, G., Kostova, N., Turianicová, E., Briančin, J., Mitov, I., Kleiv, R.A., Baláž, P., 2010. The influence of attrition milling on carbon sequestration on magnesium- iron silicate. Minerals Engineering, 23, 616-620. Ferré, E.C., Tikoff, B., Jackson, M., 2005. The magnetic anisotropy of mantle peridotites: example from the Twin Sisters dunite, Washington. Tectonophysics, 398, 141-166.

Gale, J., 2004. Geological storage of CO2: What do we know, where are the gaps and what more needs to be done?. Energy, 29, 1329-1338. Gerdemann, S.J., Dahlin, D.C., O'Connor, W.K., Penner, L.R., 2003. Carbon dioxide sequestration by aqueous mineral carbonation of magnesium silicate minerals. Albany Research Center, DOE- ARC-2003-018.

201

Gerdemann, S.J., Dahlin, D.C., O'Connor, W.K., Penner, L.R., Rush, H., 2004. Ex-situ and in-situ mineral carbonation as a means to sequester carbon dioxide. Albany Research Center DOE-ARC-2004- 031. Gerdemann, S.J., Dahlin, D.C., O'Connor, W.K., Penner, L.R., 2004. Factors affecting ex-situ mineral carbonation using calcium and magnesium silicate minerals. Albany Research Center, DOE-ARC- 2004-032. Gerdemann, S.J., O’Connor, W.K., Dahlin, D.C., Penner, L.R., Rush, H., 2007. Ex-situ aqueous mineral carbonation. Environmental Science Technology, 41 (7), 2587-2593. Ghiorso, S., 1983. LSQIEQ: a FORTRAN IV subroutine package for the analysis of multiple linear regression problems with possible deficient pseudorank and linear equality and inequality constraints. Computers and Geosciences 9, 391-416. Giammar, D.E., Bruant, R.G. Jr., Peters, C.A., 2005. Forsterite dissolution and magnesite precipitation at conditions relevant for deep saline aquifer storage and sequestration of carbon dioxide. Chemical Geology, 217, 257-276. Goff, F., Lackner, K.S., 1998. Carbon dioxide sequestering using ultramafic rocks. Environmental Geosciences, 5 (3), 89-101. Goff, F., Guthrie, G., Lipin, B., Fite, M., Chipera, S., Counce, D., Kluk, E., Ziock, H., 2000. Evaluation of ultramafic deposits in the eastern United States and Puerto Rico as sources of magnesium for carbon dioxide sequestration. Albany Research Center, LA-13694-MS.

Goldberg, P., Chen, Z.-Y., O'Connor, W.K., Walters, R., Ziock, H., 2001. CO2 mineral sequestration studies in the US. Journal of Energy & Environmental Research, 1, 117-126. Goldstein, J., Newbury, D., Joy, D., Lyman, C., Echlin, P., Lifshin, E., Sawyer, L., Michael, J., 2003. Scanning electron microscopy and X-ray microanalysis. Springer, 3rd Corrected ed., 690 pp. Goovaerts, P., 1997. Geostatistics for natural resources evaluation. Oxford University Press, 483 pp. Gordon, T.M., Dipple, G.M., 1999. Measuring mineral abundance in skarn. II A new linear programming formulation and comparison with projection and Rietveld methods. Canadian Mineralogist, 37, 17-26.

Gunter, W.D., 2009. Alberta CO2 – ECBM research and field pilots summary. Alberta Research Council, 10 pp. Guthrie, G.D., Carey, J.W., Bergfeld, D., Byler, D., Chipera, S., Ziock, H.-J., Lackner, K.S., 2001. Geochemical aspects of the carbonation of magnesium silicates in an aqueous medium. Los Alamos National Laboratory, LA-UR-01-4206.

202

Hancock, K.D., 1990. Ultramafic-associated chromite and mineral occurrences in British Columbia. B.C. Ministry of Energy, Mines and Petroleum Resources, Open File 1990-27, 62 pp. Harben, P.W., Smith Jnr, C., 2006. Olivine. Contribution in Industrial minerals and rocks: commodities markets and uses. 7th Edition: edited by Kogel, J.E., Trivedi, N.C., Barker, J.M., Krukowski, S.T., 2006. SME. Herrmann, W., Berry, R.F., 2002. MINSQ; a least squares spreadsheet method for calculating mineral proportions from whole rock major element analyses. Geochemistry: Exploration, Environment, Analysis, 2 (4), 361-368.

Herzog, H., Meldon, J., Hatton, A., 2009. Advance post-combustion CO2 capture. Prepared for the Clean Air Task Force under a grant from the Doris Duke Charitable Foundation, April (2009). Hill, R.J. and Howard, C.J., 1987. Quantitative phase analysis from neutron powder diffraction data using the Rietveld method. Journal of Applied Crystallography, 20, 467-474. Himmelburg, G.R., Loney, R.A., 1973. Petrology of the V-ulcan Peak alpine-type peridotite, southwest Oregon. Geological Society of America Bulletin, 84, 1585-1600. Hindle, S., 2011. Feasibility and sensitivity analysis of integrating mining and mineral carbonation: A case study of the Turnagain nickel project. Unpublished Master of Science Thesis, University of British Columbia. Hitch, M., Ballantyne, S.M., Hindle, S.R., 2010. Revaluing mine waste rock for carbon capture and storage. International Journal of Mining, Reclamation and Environment, 24 (1), 64-79. Hitch, M., Dipple, G.M., 2012. Economic feasibility and sensitivity analysis of integrating industrial-scale mineral carbonation into mining operations. Minerals Engineering, 39, 268-275. Huijgen, W.J.J., Witkamp, G.J., Comans, R.N.J., 2005. Mineral sequestration by steel slag carbonation. Environmental Science Technology, 39, 9676-9682.

Huijgen, W.J.J., Ruijg, G.J., Comans, R.N.J., Witkamp, G.J., 2006. Energy consumption and net CO2 sequestration of aqueous mineral carbonation. Industrial Engineering Chemical Resources, 45, 9184-9194. IPCC, 2007. Climate Change 2007: Synthesis Report. Contribution of Working Groups I, II and III to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change [Core Writing Team, Pachauri, R.K and Reisinger, A. (eds.)]. IPCC, Geneva, Switzerland, 104 pp. Irvine, T.N., 1987. Layering and related structures in the Duke Island ultramafic complex, southeastern Alaska; in Ultramafic and Related Rocks, P.J. Wyllie (eds.); John Wyllie & Sons Inc., N.Y., 84-96.

203

Isaaks, E.H., Srivastava, M.R., 1989. An introduction to applied geostatistics. Oxford University Press, New York, 561 pp. Ityokumbul, M.T., Chander, S., O’Connor, W.K., Dahlin, D.C., Gerdemann, S.J., 2001. Reactor design considerations in mineral sequestration of carbon dioxide. Proceedings of the 18th Annual International Pittsburgh Coal Conference, Newcastle, NSW, Australia, December 3-7. Journel, A.G., Huijbregts, Ch.J., 1978. Mining Geostatistics. Academic Press, London, 600 pp.

Kakizawa, M., Yamasaki, A., Yanagisawa, Y., 2001. A new CO2 disposal process via artificial weathering of calcium silicate accelerated by acetic acid. Energy, 26, 341-354. Kalinkin, A.M., Boldyrev, V.V., Politov, A.A., Kalinkina, E.V., Makarov, V.N., Kalinnikov, V.T., 2003. Investigation into the mechanism of interaction of calcium and magnesium silicates with carbon dioxide in the course of mechanical activation. Glass Physics and Chemistry, 29 (4), 410-414. Kalinkin, A.M., Kalinkina, E.V., Politov, A.A., Makarov, V.N., Boldyrev, V.V., 2004. Mechanochemical interaction of Ca silicate and aluminosilicate minerals with carbon dioxide. Journal of Materials Science, 39, 5393-5398. Katsuyama, Y., Yamasaki, A., Iizuka, A., Fujii, M., Kumagai, K., Yanagisawa, Y., 2005. Development of a

process for producing high-purity calcium carbonate (CaCO3) from waste cement using

pressurized CO2. Environmental Progress, 24, 162-170.

Kelemen, P.B., Matter, J., 2008. In situ carbonation of peridotite for CO2 storage. Proceedings of the National Academy of Sciences of the United States of America, 105 (45), 17295-17300. Killops, S.D., Killops, V.J., 2005. Introduction to organic geochemistry. Blackwell Publishing, 2nd edition, 393 pp. King, H.F., McMahon, D.W., Bujtor, G.J., Scott, A.K., 1982. A guide to the understanding of ore reserve estimation. Australasian Institute of Mineralogy and Metallurgy, Supplement to Proceedings, 281, 28 pp. Kleiv, R.A., Thornhill, M., 2006. Mechanical activation of olivine. Minerals Engineering, 19, 340-347. Kohlmann, J., Zevenhoven, R., Mukherjee, A.B., Koljonen, T., 2002. Mineral carbonation for long-term

storage of CO2 from flue gases. Final Report for Finnish National Research Programme CLIMTECH (1999-2002). Krevor, S.C.M., Lackner, K.S., 2011. Enhancing serpentine dissolution kinetics for mineral carbon dioxide sequestration. International Journal of Greenhouse Gas Control, 5 (4), 1073-1080. Labotka, T.C., Albee, A.L., 1979. Serpentinisation of the Belvidere Mountain ultramafic body, Vermont: Mass balance and reaction at the metasomatic front. Canadian Mineral, 17, 831-845.

204

Lackner, K.S., Wendt, C.H., Butt, D.P., Joyce, E.L. Jnr., Sharp, D.H., 1995. Carbon dioxide disposal in carbonate minerals. Energy, 20 (11), 1153-1170. Lackner, K.S., Butt, D.P., Wendt, C.H., 1997. Magnesite disposal of carbon dioxide. Proceedings of the 22nd International Technical Conference on Coal Utilizations and Fuel Systems, March 16-19, Clearwater, FL, USA. Lackner, K.S., Butt, D.P., Wendt, C.H., 1997. Binding carbon dioxide in mineral form: a critical step towards a zero-emission coal power plant. Los Alamos National Laboratory, LA-UR-98-2237.

Lackner, K.S., 2003. A guide to CO2 sequestration. Science, 300, 1677-1678. Lackner, K.S., Duby, P.F., Yegulalp, T., Krevor, S., Graves, C., 2008. Integrating steel production with mineral carbon sequestration. U.S. Department of Energy, AISI/DOE Technology Roadmap Program.

Larachi, F., Daldoul, I., Beaudoin, G., 2010. Fixation of CO2 by chrysotile in low-pressure dry and moist carbonation: Ex-situ and in-situ characterisations. Geochimica et Cosmochimica Acta, 74, 3051- 3075. Laube, N., Hergarten, S., Neugebaur, H.J., 1996. MODUSCALC - a computer program to calculate a mode from a geochemical rock analysis. Computer Geoscience, 22, 631-637. Le Maitre, R.W., 1981. GENMIX – A generalised petrological mixing model program. Computers and Geosciences, 7, 229-247. Le Maitre, R.W., 2002. Igneous rocks: a classification and glossary of terms. Cambridge University Press, 2nd Ed, 252pp.

Li, W., Li, W., Li, B., Bai, Z., 2009. Electrolysis and heat pretreatment methods to promote CO2 sequestration by mineral carbonation. Chemical Engineering Research and Design, 87, 210-215. Lyons, J.L., Berkshire, L.H., White, C.W., 2003. Mineral carbonation feasibility study. Draft Report, Commissioned by National Energy Technology Laboratory, 56 pp. Malvern Matersizer Support. Accessed October 11th 2011 from www.malvern.co.uk/ms2000. Massachusetts Institute of Technology, 2007. The future of coal: options for a carbon-constrained world. An Interdisciplinary MIT Study, Massachusetts Institute of Technology, 2007, 192 pp. Matheron, G., 1963. Principles of geostatistics. Economic Geology, 58 (8), 1246-1266.

McCoy, S.T., Rubin, E.S., 2008. An engineering-economic model of pipeline transport of CO2 with application to carbon capture and storage. International Journal of Greenhouse Gas Control, 2008, 219-229.

205

McKelvy, M.J., Chizmeshya, A.V.G., Diefenbacher, J., Bearat, H., Wolf, G., 2004. Exploration of the role of heat activation in enhancing serpentine carbon sequestration reactions. Environmental Science Technology, 38, 6897-6903. Metzner, C., Grimmeisen, W., 1990. MONA: a user friendly computer-program for calculating the modal mineralogy of rocks from chemical analyses. European Journal Mineralogy, 2, 735-738. Montez-Hernandez, G., Pérez-López, R., Renard, F., Nieto, J.M., Charlet, L., 2009. Mineral sequestration

of CO2 by aqueous carbonation of coal combustion fly-ash. Journal of Hazardous Materials, 161, 1347-1354. Munz, I.A., Kihle, J., Brandvoll, Ø., Machenbach, I., Carey, J.W., Haug, T.A., Johansen, H., Eldrup, N., 2009.

A continuous process for manufacture of magnesite and silica from olivine, CO2 and H2O. Energy Procedia, 1, 4891-4898.

Newall, P.S., Clarke, S.J., Haywood, H.M., Scholes, H., Clarke, N.R., King, P.A., Barley, R.W., 2000. CO2 storage as carbonate minerals. An IEA Greenhouse Gas R&D Programme report. Nilsen, D.N., Hundley, G., 1999. Preliminary feasibility study of the sequestration of carbon dioxide gas with minerals: a study of the LANL aqueous process. Albany Research Center, US DOE, DOE/ARC- TR-99-003. Nixon, G.T., Hammack, J.L., 1991. Metallogeny of mafic-ultramafic rocks in British Columbia with emphasis on platinum-group elements; in Ore Deposits, Tectonics and Metallogeny in the Canadian cordillera. B.C. Ministry of Energy, Mines and Petroleum Resources, Paper 1991-47, 125-161. Nixon, G.T., Hammack, J.L., Ash, C.H., Cabri, L.J., Case, G., Connelly, J.N., Heaman, L.M., Laflamme, J.G.H., Nuttall, C., Paterson, W.P.E., Wong, R.H., 1997. Geology and platinum-group-element mineralisation of Alaskan-type ultramafic-mafic complexes in British Columbia. B.C. Ministry of Employment and Investment, Bulletin 93, 141 pp. O’Connor, W.K., Dahlin, D.C., Turner, P.C., Walters, R., 1999. Carbon dioxide sequestration by ex-situ mineral carbonation. Proceedings of the Second Annual Dixy Lee Ray Memorial Symposium, American Society of Mechanical Engineers, Washington D.C., August 29-September 2. O’Connor, W.K., Dahlin, D.C., Nilsen, D.N., Walters, R.P., Turner, P.C., 2000. Carbon sequestration by direct mineral carbonation with carbonic acid. Proceedings of the 25th International Technical Conference on Coal Utilizations and Fuel Systems, Clearwater, FL, March 6-9.

206

O’Connor, W.K., Dahlin, D.C., Nilsen, D.N., Rush, G.E., Walters, R.P., Turner, P.C., 2000. CO2 storage in solid form: a study of direct mineral carbonation. Proceedings of the 5th International Conference on Greenhouse Gas Technologies, Cairns, Australia, August 14-18. O’Connor, W.K., Dahlin, D.C., Nilsen, D.N., Rush, G.E., Walters, R.P., Turner, P.C., 2001. Carbon dioxide sequestration by direct aqueous mineral carbonation. Proceedings of the 26th International Technical Conference on Coal Utilizations and Fuel Systems, Clearwater, FL, March 5-8. O’Connor, WK., Dahlin, D.C., Nilsen, D.N., Gerdemann, S.J., Rush, G.E., Penner, L.R., Walters, R.P.,

Turner, P.C., 2002. Continuing studies on direct aqueous mineral carbonation for CO2 sequestration. 27th International Technical Conference on Coal Utilization & Fuel Systems, Clear Water, FL, March 4-7. O’Connor, W.K., Dahlin, D.C., Rush, G.E., Gerdemann, S.J., Penner, L.R., 2004. Energy and economic evaluation of ex situ aqueous mineral carbonation. National Energy Technology Laboratory, DOE/ARC-2004-028. O’Connor, W.K., Dahlin, D.C., Rush, G.E., Gerdemann, S.J., Penner, L.R., Nilsen, D.N., 2005. Aqueous mineral carbonation: mineral availability, pre-treatment, reaction parametrics, and process studies. National Energy Technology Laboratory, DOE/ARC-TR-04-002. Onyeagocha, A.C., 1978. Twin Sisters dunite: petrology and mineral chemistry. Geological Society of America Bulletin, 89, 1459-1474. Pacala, S., Socolow, R., 2004. Stabilization wedges: solving the climate problem for the next 50 years with current technologies. Science, 305, 968-972. Paktunc, A.D., 2001. MODAN – An interactive computer-program for estimating mineral quantities based on bulk composition. Computers and Geosciences, 24 (5), 425-431. Paktunc, A.D., 2001. MODAN – A computer program for estimating mineral quantities based on bulk composition: Windows version. Computers and Geosciences, 27 (7), 883-886.

Park, A-H.A., Fan, L-S., 2004. CO2 mineral sequestration: physically activated dissolution of serpentine and pH swing process. Chemical Engineering Science, 59, 5241-5247. Patterson, J.A., 1959. Estimating ore reserves follows logical steps. Engineering Minerals, 111-159. Pawley, G.S., 1981. Unit-cell refinement from powder diffraction scans. Journal of Applied Crystallography, 14, 357-361. Penner, L.R., O’Connor, W.K., Gerdemann, S.J., Dahlin, D.C., 2003. Mineralisation strategies for carbon dioxide sequestration. Proceedings of the 20th Annual International Pittsburgh Coal Conference, Pittsburgh, PA, September 15-19.

207

Penner, L.R., Gerdemann, S.J., Dahlin, D.C., O’Connor, W.K., Nilsen, D.N., 2003. Progress on continuous processing for mineral carbonation using a prototype flow-loop reactor. Proceedings of the 28th International Technical Conference on Coal Utilization & Fuel Systems, Clearwater, FL, March 9- 13. Penner, L.R., O’Connor, W.K., Dahlin, D.C., Gerdemann, S.J., Rush, G.E., 2004. Mineral carbonation: energy costs of pretreatment options and insights gained from flow loop reaction studies. Albany Research Center, USDOE, DOE/ARC-2004-042. Peroni, R., Costa, J.F., Koppe, J., 2003. Considering in situ grade variability during mining sequence. Application of Computers and Operations research in the Minerals Industries, South African Institute of Mining and Metallurgy, 15-18. Power, I.M., Dipple, G.M., Southam, G., 2010. Bioleaching of ultramafic tailings by Acidithiobacilllus spp.

for CO2 sequestration. Environmental Science Technology, 44, 456-462. Raudsepp, M., Pani, E., and Dipple, G.M., 1999. Measuring mineral abundance in skarn. I: The Rietveld method using X-ray powder-diffraction data. Canadian Mineralogist, 37, 1-15. Raudsepp, M. and Pani, E., 2003. Application of Rietveld analysis to environmental mineralogy In J.L. Jambor, J.L., Blowes, D.W., Ritchie, A.I.M., Eds., Environmental Mineralogy of Mine Wastes, 31, 165-180. Mineralogical Association of Canada, Ottawa, ON, Canada. Rendek, E., Ducom, G., Germain, P., 2006. Carbon dioxide sequestration in municipal solid waste incinerator (MSWI) bottom ash. Journal of Hazardous Materials, 128, 73-79. Rendek, E., Ducom, G., Germain, P., 2007. Influence of waste input and combustion technology on MWSI bottom ash quality. Waste Management, 27, 1403-1407. Rietveld, H.M., 1969. A profile refinement method for nuclear and magnetic structures. Journal of Applied Crystallography, 2, 65-71. Robb, L., 2005. Introduction to ore-forming processes. Blackwell Publishing, 373pp. Rosen, O.M., Abbyasov, A.A., Tipper, J.C., 2004. MINLITH — an experience-based algorithm for estimating the likely mineralogical compositions of sedimentary rocks from bulk chemical analyses. Computers and Geosciences, 30 (6), 647-661. Sanchez, M.S., Gunter, M.E., 2006. Quantification of amphibole content in expanded vermiculite products from Libby, Montana U.S.A. using powder X-ray diffraction. American Mineralogist, 91, 1448-1451.

208

Sanna, A., Wang, X., Lacinska, A., Styles, M., Paulson, T., Maroto-Valer, M.M., 2013. Enhancing Mg

extraction from lizardite-rich serpentine for CO2 mineral sequestration. Minerals Engineering, 49, 139-144. Scheel, J.E., Nixon, G.T., Scoates, J.S., 2004. New observations on the geology of the Turnagain Alaskan- type ultramafic intrusive suite and associated Ni-Cu-PGE mineralisation, British Columbia. B.C. Ministry of Energy, Mines and Petroleum Resources, Paper 2005-1, 167-176. Scheel, J.E., 2007. Age and origin of the Turnagain Alaska-type intrusion and associated Ni-sulphide mineralisation, North-Central British Columbia, Canada. Unpublished Master of Science Thesis, University of British Columbia.

Seifritz, W., 1990. CO2 disposal by means of silicates. Nature, 345, 486. Sinclair, A.J., Lonergan, E.T., McConechy, W.E., 1983. Geostatistics at the feasibility stage, Golden Sunlight deposit, Montana. Nevada Bureau of Mines and Geology Report, 36, 165-172. Sinclair, A.J., Blackwell, G.H., 2002. Applied mineral inventory estimation. Cambridge University Press, 381 pp. Sipilä, J., Teir, S., and Zevenhoven, R., 2008. Carbon dioxide sequestration by mineral carbonation: Literature review update 2005-2007. Åbo Akademi University Heat Engineering Laboratory, Rep- ort 2008-1. Soong, Y., Fauth, D.L., Howard, B.H., Jones, J.R., Harrison, D.K., Goodman, A.L., Gray, M.L., Frommell,

E.A., 2006. CO2 sequestration with brine solution and fly ashes. Energy Conversion and Management, 47, 1676-1685. Stasiulaitiene, I., Fagerlund, J., Nduagu, E., Denafas, G., Zevenhoven, R., 2011. Carbonation of serpentinite rock from Lithuania and Finland. Energy Procedia, 4, 2963-2970. Stolaroff, J.K., Lowry, G.V., Keith, D.W., 2005. Using CaO- and MgO-rich industrial waste streams for carbon sequestration. Energy Conversion and Management, 46, 687-699. Stone, J.G., Dunn, P.G., 1994. Ore reserve estimates in the real world. Society of Economic Geology, Special Publication, 3, 150 pp.

Summers, C., Dahlin, D., Ochs, T., 2004. The effect of SO2 on mineral carbonation in batch tests. Albany Research Center, DOE/ARC-2004-022. Teir, S., Eloneva, S., Fogelholm, C.-J., Zevenhoven, R., 2007. Carbonation of minerals and industrial by- products for carbon sequestration. Proceedings of the IGEC-III, Västerås. Sweden, June 18-20. Teir, S., Eloneva, S., Fogelholm, C.-J., Zevenhoven, R., 2007. Dissolution of steelmaking slags in acetic acid for precipitated calcium carbonate production. Energy, 32, 528-539.

209

Thayer, T.P., 1977. The Canyon Mountain Complex, Oregon and some problems with ophiolites. In R.G. Coleman and W.P. Irwin (Eds.) North American Ophiolites (93-105). Salem: Oregon Department of Geology and Mineralogical Industries Bulletin 95. U.S. Environmental Protection Agency, 2008. EPA's Report on the Environment (ROE) (2008 Final Report). U.S. Environmental Protection Agency, Washington, D.C., EPA/600/R-07/045F (NTIS PB2008-112484). U.S. Energy Information Administration, 2011. Annual energy outlook 2011. Independent statistics and analysis U.S. Energy Information Administration. DOE0484(2011), 301 pp. U.S. Energy Information Administration, 2013. Annual energy outlook 2013 with projections to 2040. Independent statistics and analysis U.S. Energy Information Administration. DOE0484(2013), 312 pp. Vann, J., 2005. Turning geological data in to reliable mineral resource estimates. In: Davies, T., Vann, J. The Estimation and Reporting of Resources and JORC: The role of Structural Geology. AIG Bulletin 42, The Australian Institute of Geoscientists (Perth), 9-16. van der Plas, L., A.C. Tobi., 1965. A chart for judging the reliability of point counting results. American Journal of Science, 263, 87-90. Voormeij, D.A., Simandl, G.J., 2004. Ultramafic rocks in British Columbia: delineating targets for mineral

sequestration of CO2. Ministry of Energy, Mines and Petroleum Resources. Summary of Activities, 157-167.

Werner, M., Verduyn, M., van Mossel, G., Mazzotti, M., 2011. Direct flue gas CO2 mineralization using activated serpentine: Explaining the reaction kinetics by experiments and population balance modelling. Energy Procedia, 4, 2043-2049. Wilson, S.A., Raudsepp, M., and Dipple, G.M., 2006. Verifying and quantifying carbon fixation in minerals from serpentine-rich mine tailings using the Rietveld method with X-ray powder diffraction data. American Mineralogist, 91, 1331-1341. Wilson, S.A., Dipple, G.M., Power, I.M., Thom, J.M., Anderson, R.G., Raudsepp, M., Gabites, J.E., and Southam, G., 2009. Carbon dioxide fixation within mine wastes of ultramafic-hosted ore deposits: Examples from the Clinton Creek and Cassiar chrysotile deposits, Canada. Economic Geology, 104, 95-112. Wilson, S.A., Raudsepp, M, Dipple, G.M., 2009. Quantifying carbon fixation in trace minerals from processed kimberlite: A comparative study of quantitative methods using X-ray powder

210

diffraction data with applications to the Diavik Diamond Mine, Northwest Territories, Canada. Applied Geochemistry, 24, 2312-2331. Wilson, S.A., Barker, S.L.L., Dipple, G.M., Raudsepp, M., Fallon, S.J., 2009. Carbon fixation in mineral waste from the Mount Keith Nickel Mine, Western Australia, Australia. Geological Society of America Meeting, 41 (7), Portland, Oregon, USA, October 18-21. Wright, T.L., Doherty, P.C., 1970. A linear programming and least-squares computer method for solving petrologic mixing problems. Geological Society of America, Bulletin 81, 1995-2008. Yamasaki, A., Iizuka, A., Kakizawa, M., Katsyama, Y., Nakagawa, M., Fujii, M., Kumagai, K., Yanagisawa, Y., 2006. Development of a carbon sequestration process by the carbonation reaction of waste streams containing calcium or magnesium. Proceedings of the 5th Annual Conference on Carbon Capture and Sequestration, Alexandria, VA, USA, May 8-11. Yoo, K., Kim, B-S., Kim, M-S., Lee, J-C., Jeong, J., 2009. Dissolution of magnesium from serpentine minerals in sulfuric acid solution. Materials Transactions, 50 (5), 1225-1230.

Zevenhoven, R., Kavaliauskaite, I., 2004. Mineral carbonation for long-term CO2 storage: an exergy analysis. International Journal of Thermodynamics, 7, 22-31. Zevenhoven, R., Teir, S., 2004. Long term storage as magnesium carbonate in Finland. Proceedings of the Third Annual Conference on Carbon Capture and Sequestration, May 3-6, 2004, Alexandria (VA), USA. Zevenhoven, R., Teir, S., Eloneva, S., 2008. Heat optimisation of a staged gas-solid mineral carbonation

process for long-term CO2 storage. Energy, 33, 362-370.

Zevenhoven, R., Fagerlund, J., Songok, J.K., 2011. CO2 mineral sequestration: developments towards large-scale application. Greenhouse Gas Science Technology, 1, 48-57. Zhao, L., Sang, L., Chen, J., Ji, J., Teng, H.J., 2009. Aqueous carbonation of natural brucite: relevance to

CO2 sequestration. Environmental Science Technology, 44, 406-411.

211

Appendices

Appendix A MCP calculator instructions and modifications

The MCP calculator is a Microsoft ExcelTM macro enabled worksheet program that is comprised of four active worksheets that can be modified by the user to customise it to the specific ultramafic deposit being studied. The data required setting up the program and the instructions of use are given here. Proposed modifications that could be made to customise the worksheet to a specific deposit are also proposed. Each of the four active worksheets has a protected version to show the inputs and formulas of the original worksheet cells should they be altered. A version of the macro-enabled worksheet program (MCP_Calculator_Jacobs_2014) is submitted with this dissertation.

Appendix A.1 Sample data preparation worksheet Instructions to setup the worksheet are as follows: 1. In the rock codes section of Box 2, enter the rock types studied into cells F33 to F37. 2. Enter the corresponding average magnesium number for the rock type in cells H33 to H37. Modification: Additional rock codes can be added and their corresponding information in to the rock codes section of Box 2. The formula in cell H30 will require modification to include the additional rock codes.

Instructions to use the worksheet are as follows: 1. Enter the unique sample ID in cell B1. 2. Enter the rock code for the sample in cell B2. This should correspond to an assigned rock code from the rock codes section of Box 2. 3. If the major cation data are in weight percentage values, enter the values in cells B7 to B43. If the values are in part per million, enter the values in cells D7 to D43. 4. In the sulphides section of Box 2, enter either FeS or FeS2 into cell H26. If you do not wish the sulphides to have any attributed iron concentration, leave cell H26 blank. This will convert all iron to FeO in the profile and only the S cation concentration will be attributed to sulphides in the modal profile (Box 4). Modification: The sulphides section can be modified to add any cation contribution from Box 1. The formulas in cells V7 and W7 must be adjusted to reflect the changes.

212

5. If oxide values are available for validation then enter them in Box 3, cells L7 to L43. The default error allowed for validation is 1 % for each oxide and 5 % for the entire profile. Modification: Additional oxides can be added to Box 2 for more cations. The corresponding cells in all four boxes must be adjusted to create a valid modal profile in Box 4. Modification: To adjust the validation error on each respective oxide, the formulas in cells N7 to N43 must be modified. The total error allowance for validation can be modified in cell N44. 6. If the sample is ready to be used in the MCP calculator, copy cells P7 to AG7 in Box 4 and using the right-click and the paste values function add the mineral calculated into the sample library.

Appendix A.2 Mineral composition calculator worksheet Instructions to use the worksheet are as follows: 1. Enter the name of the mineral to be calculates into cell A6. 2. Enter the number of each of the respective cations making up the mineral formula into cells B5 to P5. Modification: Additional cations and respective oxides could be added to extend the capabilities and number of minerals that can be determined by the calculator. 3. Copy rows 5 and 6 from cells A5 to R6. 4. Using the right-click and the paste values function add the mineral calculated into the mineral library. Modification: The mineral library could be expanded to store more minerals.

Note: The minerals used for estimation at the Turnagain ultramafic complex are currently stored within the mineral library.

Appendix A.3 MCP calculator worksheet There are no direct running instructions to operate the MCP calculator as the modal profile is imported directly into Box 1 from the sample data preparation worksheet. The sheet does require some minor setup as follows:

1. In the instructions box; enter the minerals capable of being used to sequester CO2 by mineral carbonation.

213

2. Enter the corresponding reaction extent (Rx) expected for each selected mineral. Modification: Additional minerals could be added to the range. Modifications must be made in Box 4 to the formulas in cells BD4 to BD34 to incorporate any changes.

The current worksheet is designed around the specific mineral assemblage identified and assessed at Turnagain in this dissertation. The following is a description of how each mineral is estimated and modifications that could be made to the worksheet in order to customise it to the generalised modal mineral composition of the rocks being studied at an alternative deposit: 1. Box 3 contains the estimated modal mineral abundances for each sample derived from the

modal profile in Box 1. Each mineral is represented by a stage number (Sn) for estimation within the labelled working cells of the worksheet.

2. S1 is the calculated SiO2 + H2O content assumed as all of the remaining components of the modal profile normalised to 100 %. 3. S2 is the value added directly from the sample data preparation worksheet. Modification: As discussed in Appendix 1.1, the sulphides are currently assessed as iron sulphides but other sulphide minerals could be directly estimated within the calculator. 4. S3 is classified as accessory. This component is the measured weight abundance of any elements not converted in the modal profile and considered as negligible toward mineral contribution and effect on the MCP values estimated. 5. S4 is the value added directly from the sample data preparation worksheet. It is currently the sum content of Co, Cu, Mo, Ni, Pb, W, and Zn. Modification: Any further base metals could be added to this list. The stage could also be combined with S2 for more accurate sulphide mineral estimation if required.

6. S5 estimates the weight percentage of ilmenite (FeTiO3). Ilmenite is assumed to be the

mineral source of all the TiO2 in the samples tested. The weight percentage fraction of iron

contributing the ilmenite formula is derived from the TiO2 content within the modal profile of each sample. The FeO contributed to ilmenite is stored in cells B39 to B69.

7. S6 estimates the weight percentage of chromite (FeCr2O4). Chromite is assumed to be the

mineral source of all of the Cr2O3 in the samples tested. The weight percentage fraction of

FeO contributing the chromite formula is derived from the Cr2O3 content within the modal profile of each sample. The FeO contributed to ilmenite is stored in cells C39 to C69. The adjusted FeO content after ilmenite and chromite estimation is stored in cells D39 to D69.

214

8. S7 estimates the weight percentage of anorthite (CaAl2Si2O8). Anorthite is assumed to be the

mineral source of all of the Al2O3 in the samples tested. The weight percentage fraction of

CaO contributing the anorthite formula is derived from the Al2O3 content within the modal profile of each sample and stored in cells E39 to E69. The re-adjusted CaO weight abundance after anorthite contribution is stored in cells F39 to F69. The weight percentage

fraction of SiO2 contributing the anorthite formula is stored in cells G39 to G69.

Modification: Alternative primary mineral sources of Al2O3 should be substituted with anorthite at this stage.

9. S8 estimates the weight percentage magnesite (MgCO3). An accurate measurement of CO2 must be generated in the modal profile to estimate magnesite. The formulas in cells H39 to

H69 determine the amount of magnesium to bind with all of the measure CO2 as magnesite. The adjusted MgO content is stored in cells I39 to I69. Modification: Another estimation stage for calcite estimation could be inserted at this point or alternatively the formulas could be adjusted to use the CaO content rather than MgO to estimate calcite abundance.

10. S9 estimates the weight percentage of diopside (CaMgSi2O6). Diopside is assumed to be the mineral source of all of the remaining CaO in the samples tested. The CaO content of the mineral formula is provided from cells F39 to F69. The MgO content required to use up all of the CaO is in diopside is given in cells J39 to J69. The adjusted MgO content is stored in cells

L39 to L69. The SiO2 required to use up all of the CaO in diopside is stored in cells K39 to K69. Modification: Alternative primary mineral sources of CaO should be substituted with diopside at this stage. 11. S10 provides the initial estimate of olivine abundance. All of the remaining MgO from the modal profile is first assumed to be contained within olivine. The magnesium number from

the modal profile defines MgO, FeO and SiO2 ratio used in olivine estimation. The ratios used are stored in the olivine conversion matrix at the bottom of the worksheet and are incorporated into the appropriate formulas within S10.

Modification: The current matrix ranges from Fo70 to Fo99. The matrix and respective formulas would require adjustment to include a greater forsterite range in the estimation.

12. In olivine estimation, the amount of FeO and SiO2 are first calculated assuming all of the

remaining MgO is contained within olivine. The amount of SiO2 required is stored in cells

215

M39 to M69 and the amount of FeO required is stored in cells N39 to N69. The remaining

SiO2 + H2O balance from S1 is recalculated following all previous mineral estimation and stored in cells O39 to O69. The remaining FeO balance following all previous mineral estimation is stored in P39 to P69. If a negative weight percentage was calculated for either

FeO or SiO2 a series of renormalization formulas are included in cells Q39 to T69 to prevent

this. This results in MgO being released from olivine and a surplus of either SiO2 or FeO is

restored. The renormalised SiO2 and H2O weight percentage is stored in cells V39 to V69. The temporary olivine weight percentage estimation is stored in cells W39 to W69. 13. S11 estimates the lizardite weight percentage if present. If there is a surplus of MgO

following olivine estimation, the amount of SiO2 and H2O required to form lizardite from the

MgO is calculated. The quantity is stored in cells X39 to X69. If there is enough SiO2 and H2O remaining to complete this, the temporary lizardite abundance is stored in cells Y39 to Y69.

14. If there is a surplus of SiO2 and H2O following S10, the quantity of MgO required to bind with

this SiO2 and H2O in lizardite is removed from the temporary olivine weight percentage to

use all of the SiO2 and H2O. This will create a surplus of MgO and FeO which is stored in cells AE39 to AE69 and AF39 to AF69 respectively. A check was emplaced in cells AG39 to AG69

so that if the temporary olivine weight abundance and the MgO, FeO and SiO2 taken from this olivine abundance is below 10 wt.% after renormalisation for lizardite estimation, no lizardite will be estimated and the temporary olivine abundance is final. This check ensures that no negative olivine abundances will be calculated. Note: This places a minimum olivine abundance of 10 wt.% in each sample tested.

15. After the adjustment to best estimate the distribution of MgO, FeO, SiO2 and H2O within olivine and lizardite, the temporary olivine abundance is stored in cells AH39 to AH69 and the temporary lizardite abundance in AI39 to AI69. Any surplus FeO from S11 is stored as temporary magnetite abundance in cells AJ39 to AJ69. 16. Iterations 1, 2 and 3 labeled S11b, S11c and S11d respectively continue to readjust the

olivine and lizardite abundances to reduce the SiO2 and H2O contents to a value below 1 wt.% (Note: This is an arbitrary degree of accuracy that can be modified). For each iteration, the temporary olivine, lizardite and magnetite weight percentages are stored. Iterations

continue until either the SiO2 and H2O balance is below 1 wt.% or the olivine content is below 10 wt.%. Modification: Further iterations could be added.

216

17. S13 sums the SiO2 and H2O values not attributed to a mineral. This is assumed to be silica

(SiO2) and remains as part of the modal mineral estimation profile. 18. S14 sums any surplus MgO values not attributed to a mineral. This is assumed to be periclase (MgO) and remains as part of the modal mineral estimation profile.

Appendix A.4 Multiple sample worksheet Instructions to use the worksheet are as follows: 1. Enter a unique sample ID for each sample in Column B 2. Enter the associated rock code for the sample being tested in Column C. Note: This rock code should correspond with the rock codes assigned in Box 2 of the sample data preparation worksheet. 3. Enter either FeS or FeS2 in Column D to define how the sulphide minerals will be estimated in the calculator. Modification: If additional sulphides are to be estimated in an earlier worksheet, this column must be adjusted accordingly. 4. Enter all of the major cation values in the order in which they are set out in the worksheet. A blank value may be inserted if there is no recorded value. Modification: The major cation values are currently set up as weight percentage and adjustments would be required to accept values in parts per million. 5. Select the yellow “Activate multiple samples worksheet” button to run the macro. 6. The MCP values are output into Column AR nest to their unique sample ID.

Note: The macro-enabled worksheet must be enabled for the macro to run. Modifications: Any adjustments to the macro can be made by following: View > Macros from the menu ribbon.

217

Appendix B Data analysis for resource estimation

Appendix B.1 Olivine geological domain Figure B.1 Olivine geological domain CaO charts a) Histogram of 15 m composited CaO (wt.%) values within the olivine geological domain with displayed normal distribution curve:

b) Cumulative frequency plot of 15 m composited CaO (wt.%) values c) Probability plot of 15 m composited CaO (wt.%) values within the within the olivine geological domain: olivine geological domain:

218

Figure B.2 Olivine geological domain MgO charts a) Histogram of 15 m composited MgO (wt.%) values within the olivine geological domain with displayed normal distribution curve:

b) Cumulative frequency plot of 15 m composited MgO (wt.%) values c) Probability plot of 15 m composited MgO (wt.%) values within the within the olivine geological domain: olivine geological domain:

219

Figure B.3 Olivine geological domain MCP charts a) Histogram of 15 m composited MCP values within the olivine geological domain with displayed normal distribution curve:

b) Cumulative frequency plot of 15 m composited MCP values within c) Probability plot of 15 m composited MCP values within the olivine the olivine geological domain: geological domain:

220

Figure B.4 Olivine geological domain MCP_ARC charts a) Histogram of 15 m composited MCP_ARC values within the olivine geological domain at the Turnagain ultramafic complex:

b) Cumulative frequency plot of 15 m composited MCP_ARC values c) Probability plot of 15 m composited MCP_ARC values within the within the olivine geological domain: olivine geological domain:

221

Figure B.5 Olivine geological domain MgO vs MCP scatter plot with linear regression

222

Appendix B.2 Wehrlite geological domain Figure B.6 Wehrlite geological domain CaO charts a) Histogram of 15 m composited CaO (wt.%) values within the wehrlite geological domain with displayed normal distribution curve:

b) Cumulative frequency plot of 15 m composited CaO (wt.%) values c) Probability plot of 15 m composited CaO (wt.%) values within the within the wehrlite geological domain: wehrlite geological domain:

223

Figure B.7 Wehrlite geological domain MgO charts a) Histogram of 15 m composited MgO (wt.%) values within the wehrlite geological domain with displayed normal distribution curve:

b) Cumulative frequency plot of 15 m composited MgO (wt.%) values c) Probability plot of 15 m composited MgO (wt.%) values within the within the wehrlite geological domain: wehrlite geological domain:

224

Figure B.8 Wehrlite geological domain MCP charts a) Histogram of 15 m composited MCP values within the wehrlite geological domain with displayed normal distribution curve:

b) Cumulative frequency plot of 15 m composited MCP values within c) Probability plot of 15 m composited MCP values within the wehrlite the wehrlite geological domain: geological domain:

225

Figure B.9 Wehrlite geological domain MCP_ARC charts a) Histogram of 15 m composited MCP_ARC values within the wehrlite geological domain with displayed normal distribution curve:

b) Cumulative frequency plot of 15 m composited MCP_ARC values c) Probability plot of 15 m composited MCP_ARC values within the within the wehrlite geological domain: wehrlite geological domain:

226

Figure B.10 Wehrlite geological domain MgO vs MCP scatter plot with linear regression

227

Appendix B.3 Olivine clinopyroxenite geological domain Figure B.11 Olivine clinopyroxenite geological domain CaO charts a) Histogram of 15 m composited CaO (wt.%) values within the olivine clinopyroxenite geological domain with normal distribution curve:

b) Cumulative frequency plot of 15 m composited CaO (wt.%) values c) Probability plot of 15 m composited CaO (wt.%) values within the within the olivine clinopyroxenite geological domain: olivine clinopyroxenite geological domain:

228

Figure B.12 Olivine clinopyroxenite geological domain MgO charts a) Histogram of 15 m composited MgO (wt.%) values within the olivine clinopyroxenite geological domain with displayed normal distribution curve:

b) Cumulative frequency plot of 15 m composited MgO (wt.%) values c) Probability plot of 15 m composited MgO (wt.%) values within the within the olivine clinopyroxenite geological domain: olivine clinopyroxenite geological domain:

229

Figure B.13 Olivine clinopyroxenite geological domain MCP charts a) Histogram of 15 m composited MCP values within the olivine clinopyroxenite geological domain with displayed normal distribution curve:

b) Cumulative frequency plot of 15 m composited MCP values within c) Probability plot of 15 m composited MCP values within the olivine the olivine clinopyroxenite geological domain: clinopyroxenite geological domain:

230

Figure B.14 Olivine clinopyroxenite geological domain MCP_ARC charts a) Histogram of 15 m composited MCP_ARC values within the olivine clinopyroxenite geological domain with normal distribution curve:

b) Cumulative frequency plot of 15 m composited MCP_ARC values c) Probability plot of 15 m composited MCP_ARC values within the within the olivine clinopyroxenite geological domain: olivine clinopyroxenite geological domain:

231

Figure B.15 Olivine clinopyroxenite geological domain MgO vs MCP scatter plot with linear regression

232