UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Mechanisms of carbon mineralization from the pore to field scale : implications for carbon dioxide sequestration Harrison, Anna Lee 2014

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2015_february_harrison_anna.pdf [ 19.84MB ]
Metadata
JSON: 24-1.0166095.json
JSON-LD: 24-1.0166095-ld.json
RDF/XML (Pretty): 24-1.0166095-rdf.xml
RDF/JSON: 24-1.0166095-rdf.json
Turtle: 24-1.0166095-turtle.txt
N-Triples: 24-1.0166095-rdf-ntriples.txt
Original Record: 24-1.0166095-source.json
Full Text
24-1.0166095-fulltext.txt
Citation
24-1.0166095.ris

Full Text

MECHANISMS OF CARBON MINERALIZATION FROM THE PORE TO FIELD SCALE: IMPLICATIONS FOR CO2 SEQUESTRATIONbyAnna Lee HarrisonB.Sc., The University of Alberta, 2010A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OFDOCTOR OF PHILOSOPHYinTHE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES(Geological Sciences)THE UNIVERSITY OF BRITISH COLUMBIA(Vancouver)December 2014© Anna Lee Harrison, 2014iiAbstract Innovative technologies to stabilize atmospheric CO2 concentrations are essential in order to mitigate the harmful effects of anthropogenic greenhouse gas (GHG) emissions on the global climate system. Mineralization of carbon in solid, stable carbonate minerals through reaction of CO2 with Mg-rich mining wastes is a promising CO2 sequestration strategy that offers the potential to render certain mines GHG neutral. Here, the physical and chemical controls on rates of and capacity for CO2 sequestration in systems representative of mine tailings are examined from the mineral-fluid interface to field scale using a combination of experimental techniques.  These experimental data and existing field data are used to develop a comprehensive reactive transport model that captures the processes governing carbon mineralization in the shallow subsurface. Stirred batch reactor, microfluidic pore scale, and decimeter to meter scale column carbonation experiments using brucite [Mg(OH)2]  revealed that the primary controls on carbonation include the rate of CO2 supply, the distribution of the reactive phase, the mineral grain size/surface area, and the availability and distribution of water. The rate-limiting step during carbonation varied from CO2 supply to mineral dissolution depending on the experimental variables. Surface passivation and water-limited reaction resulted in a highly non-geometric evolution of reactive surface area. The extent of reaction was also limited at high water content because viscous fingering of the gas streams supplied to the columns resulted in narrow zones of highly carbonated material, but left a large proportion of brucite unreacted. More robust predictions of the CO2 sequestration rate and capacity that can be expected at the field scale are possible due to the incorporation of water consumption, water-limited reactivity, and surface passivation functions into the reactive transport code, MIN3P. This research imparts a better understanding of fundamental mechanisms and chemical processes relevant to CO2 sequestration in mine tailings, with implications for mineral carbonation in other settings that have greater CO2 sequestration capacity, such as shallow subsurface formations with similar mineralogy. Aspects of this research, such as water-limited iiireactivity, have broader implications for reactive transport processes in the vadose zone in general, including mineral weathering and groundwater remediation. ivPrefaceThis thesis consists of five related manuscripts that have been published in, submitted to, or prepared for submission to peer-reviewed international academic journals and a peer-reviewed paper published in international conference proceedings. A version of Chapter 2 is published in Environmental Science & Technology. This chapter is reprinted with permission from Harrison, A. L., Power, I. M. and Dipple, G. M. (2013) Accelerated carbonation of brucite in mine tailings for carbon sequestration. Environ. Sci. Technol. 47: 126–134. Copyright 2012 American Chemical Society. The Supporting Information to this manuscript is reproduced in this thesis as Appendix 1: Appendix to Chapter 2. I conducted the experiments and analytical work, interpreted results, and wrote the manuscript. I.M. Power and G.M. Dipple advised on experimental design, helped with the interpretation of results, and edited the manuscript. All authors confirm that this is an accurate reflection of the author contributions and agree to the use of the Environmental Science & Technology paper in this thesis. A version of Chapter 3 was published in the proceedings of the International Mine Water Association 2013 Mine Water Conference as: Harrison, A. L., Power, I. M. and Dipple, G. M. (2013) Strategies for enhancing carbon sequestration in Mg-rich mine tailings, in Brown, A., Figueroa, L., Wolkersdorfer, C. (Eds.), Reliable Mine Water Technology (Vol. 1). Publication Printers, Denver, Colorado, USA, pp. 593–598. It is used with permission of the International Mine Water Association. I interpreted results and wrote the manuscript. I.M. Power and G.M. Dipple advised on the interpretation of results and edited the manuscript. All authors confirm that this is an accurate reflection of the author contributions and agree to the use of this paper in this thesis. Chapter 4 has been prepared for submission to an international peer-reviewed academic journal. The appendix to this chapter (Appendix 2: Appendix to Chapter 4) will be submitted as the supporting information to the manuscript. The manuscript is authored as follows: vA.L. Harrison, G.M. Dipple, W. Song, I.M. Power, K.U. Mayer, A. Beinlich, and D. Sinton. G.M. Dipple and I conceived the study. D. Sinton, G.M. Dipple, W. Song and I designed the experiments. W. Song and I conducted the experiments. W. Song, I.M. Power, and I performed microscopy and A. Beinlich conducted Raman spectroscopy. G.M. Dipple and I interpreted the data and discussed its implications with input from K.U. Mayer on reactive transport and geochemical modeling. I wrote the manuscript with input from all authors. All authors confirm that this is an accurate reflection of the author contributions and agree to the use of this manuscript in this thesis. A version of Chapter 5 is published in Geochimica et Cosmochimica Acta. This chapter is reprinted with permission from Geochimica et Cosmochimica Acta, 148, Harrison, A. L., Dipple, G.M., Power, I.M. and Mayer, K. U., Influence of surface passivation and water content on mineral reactions in unsaturated porous media: Implications for brucite carbonation and CO2 sequestration, 477-495, Copyright (2014), with permission from Elsevier. The supporting information for this manuscript is reproduced in this thesis as Appendix 3: Appendix to Chapter 5. I designed the experiments with input from I.M. Power and G.M. Dipple. I conducted the experiments, most of the sample analysis, and all of the reactive transport modeling. I.M. Power and I performed microscopy. G.M. Dipple, K.U. Mayer and I developed a new function describing surface passivation effects in the reactive transport code, MIN3P-DUSTY. K.U. Mayer provided guidance in the implementation of this function in MIN3P-DUSTY and the modeling of experiments. I interpreted the data with input from all authors, and I wrote the manuscript. G.M. Dipple, I.M. Power, and K.U. Mayer helped with revisions and editing of the manuscript. All authors confirm that this is an accurate reflection of the author contributions and agree to the use of this manuscript in this thesis. Chapter 6 has been prepared for submission to an international peer-reviewed academic journal. The appendix to this chapter (Appendix 4: Appendix to Chapter 6) will be submitted as the supporting information to the manuscript. The manuscript is authored as follows: A.L. Harrison, G.M. Dipple, K.U. Mayer and I.M. Power. G.M. Dipple and K.U. Mayer viconceived the study. I designed the experiments with advice from all authors. I conducted the experiments and I.M. Power provided assistance in the laboratory and with sampling. I.M. Power and I performed microscopy. I interpreted the data with input from all authors. G.M. Dipple, K.U. Mayer, and I developed conceptual models to explain the observed extent of reaction in the experiments that were implemented in MIN3P-DUSTY. K.U. Mayer helped with this implementation, and the addition of a formulation in MIN3P-DUSTY to account for water loss during reaction. I wrote the manuscript, and all authors contributed to revisions and editing of the manuscript. All authors confirm that this is an accurate reflection of the author contributions and agree to the use of this manuscript in this thesis. A version of Chapter 7 is published in International Journal of Greenhouse Gas Control. This chapter is reprinted with permission from International Journal of Greenhouse Gas Control, 25, Wilson, S. A., Harrison, A. L., Dipple, G. M., Power, I. M., Barker, S. L. L., Mayer, K. U., Fallon, S. J., Raudsepp, M. and Southam, G., Offsetting of CO2 emissions by air capture in mine tailings at the Mount Keith Nickel Mine, Western Australia: Rates, controls and prospects for carbon neutral mining, 121-140, Copyright (2014), with permission from Elsevier. This manuscript was a collaborative effort with Dr. Siobhan A. Wilson, the lead author and several co-authors. As second author, I conducted reactive transport modeling and contributed to the writing and editing of the manuscript. This study consisted of three main aspects: 1) field work and analysis of field samples to quantify and verify atmospheric CO2 mineralization within an active mine tailings storage facility, 2) reactive transport modeling to explain the observed abundance and distribution of carbonate mineral precipitates and estimate the rate of CO2 sequestration, and 3) recommendations for tailings management and ore processing practices that could enhance CO2 sequestration within mine tailings storage facilities to offset greenhouse gas emissions. I was not involved in part 1; this was lead by S.A. Wilson with contributions from most of the other co-authors. I.M. Power, G.M. Dipple, and G. Southam helped with fieldwork, while S.L.L. Barker, and S.J. Fallon contributed to the analysis of field samples. These parts of the manuscript were written by S.A. Wilson. I led part 2, the reactive viitransport modeling, which involved developing the conceptual model for reactive transport at the site, tailoring the numerical model MIN3P-THM to apply to this conceptual framework, and testing hypotheses to describe the observed variation in tailings mineralogy at the mine site using MIN3P-THM. I wrote the sections of the manuscript describing the modeling methods, results, and interpretations. K.U. Mayer advised on the conceptual model and application of MIN3P-THM, and S.A. Wilson, G.M. Dipple, and I.M. Power advised on the conceptual framework and interpretation of results. Part 3 was a joint effort. Recommendations were made to accelerate CO2 sequestration based on the interpretations stemming from the field and analytical work, as well as the reactive transport modeling. The reactive transport modeling provided insight into the physical constraints (e.g., hydraulic conductivity, CO2 transport at depth) that may dictate the choice of acceleration strategy at a mine site, and helped confirm that carbon mineralization is largely restricted by the limited access of reactive minerals to CO2. The conclusions of the paper were therefore influenced by the inclusion of the reactive transport modeling. G.M. Dipple and S.A. Wilson conceived the study, with input from G. Southam and M. Raudsepp. S.A. Wilson, G.M. Dipple, A.L. Harrison, I.M. Power, K.U. Mayer, and M. Raudsepp interpreted results. S.A. Wilson, A. L. Harrison, I.M. Power, and G.M. Dipple contributed to preparation of the manuscript. Because the modeling depended upon and complemented the field results and interpretations, my supervisory committee and I agreed that the published manuscript should be included in its entirety in this thesis. All authors confirm that this is an accurate reflection of the author contributions, and agree to the use of the International Journal of Greenhouse Gas Control paper in this thesis. viiiAbstract .................................................................................................................................... iiPreface ..................................................................................................................................... ivTable of Contents .................................................................................................................viiiList of Tables .......................................................................................................................... xvList of Figures .....................................................................................................................xviiiList of Symbols and Abbreviations ................................................................................... xxivList of Supplementary Materials ...................................................................................... xxixAcknowledgements ............................................................................................................. xxxDedication .......................................................................................................................... xxxii1. Introduction ................................................................................................................... 11.1 CO2 sequestration via carbon mineralization ....................................................... 11.2 Carbon mineralization in ultramafic mine tailings ............................................... 21.3 Fundamental processes of water-rock interaction in the Earth’s shallow subsurface ............................................................................................................ 41.4 Objectives and organization ................................................................................. 52. Accelerated carbonation of brucite in mine tailings for carbon sequestration ...... 102.1 Introduction ........................................................................................................ 102.2 Methods.............................................................................................................. 122.3 Results and discussion ....................................................................................... 142.3.1 Chemical environment ............................................................................. 142.3.2 Brucite dissolution and carbonation rates ................................................ 172.3.3 Reaction mechanism ................................................................................ 212.3.4 Rate limitation ......................................................................................... 222.3.5 Implications for carbon sequestration in mine tailings ............................ 253. Strategies for enhancing carbon sequestration in Mg-rich mine tailings ............... 28Table of Contentsix3.1 Introduction ........................................................................................................ 283.2 Methods.............................................................................................................. 293.3 Results and discussion ....................................................................................... 303.4 Conclusions ........................................................................................................ 354.	 Enhanced	mineral	reactivity	driven	by	pore	fluid	mobility	in	partially	wetted	porous	media ............................................................................................................................. 364.1 Introduction ........................................................................................................ 364.2 Methods.............................................................................................................. 374.3 Results and discussion ....................................................................................... 385.	 Influence	of	surface	passivation	and	water	content	on	mineral	reactions	in	unsaturated	porous media: Implications for brucite carbonation and CO2 sequestration ........ 475.1 Introduction ........................................................................................................ 475.2 Methods.............................................................................................................. 505.2.1 Experimental design ................................................................................ 505.2.2 Assessment of reaction progress .............................................................. 545.2.3 Reactive transport modeling .................................................................... 555.3 Results ................................................................................................................ 565.3.1 Instantaneous carbonation rates and CO2 breakthrough .......................... 565.3.2 Reaction progress .................................................................................... 615.3.3 Qualitative characterization of solids ...................................................... 635.3.4 Water content ........................................................................................... 635.3.5 Reaction stoichiometry ............................................................................ 655.4 Discussion .......................................................................................................... 675.4.1 Reaction stages ........................................................................................ 675.4.2 Controls on reaction ................................................................................. 685.4.2.1 Water-limited reaction ....................................................................... 725.4.2.2 Consumption of highly reactive sites and fine grains ....................... 74x5.4.2.3 Surface passivation............................................................................ 755.5 Implications ........................................................................................................ 785.5.1 Reactive transport modeling .................................................................... 785.5.2 CO2 sequestration .................................................................................... 796.	 Physical	 and	 chemical	 influences	 of	 water	 on	 mineral	 carbonation	 in	 variably	saturated porous media: Implications for CO2 sequestration ................................. 826.1 Introduction ........................................................................................................ 826.2 Methods.............................................................................................................. 846.2.1 Experimental design ................................................................................ 846.2.2 Assessment of the rate and extent of carbonation ................................... 886.2.3 Geochemical and reactive transport modeling ........................................ 896.3 Results ................................................................................................................ 916.3.1 Rate and extent of carbonation ................................................................ 916.3.2 Characterization of precipitates ............................................................... 936.3.2.1 Composition and distribution of precipitates .................................... 936.3.2.2 Qualitative characterization of precipitates ....................................... 966.3.3 Trends in aqueous chemistry and mineral saturation states ..................... 986.3.4 Water and gas flow................................................................................. 1016.3.5 Stable carbon isotopic compositions ..................................................... 1016.4 Reaction processes ........................................................................................... 1046.4.1 Reaction rate .......................................................................................... 1046.4.2 Reaction shutdown ................................................................................ 1096.4.2.1 Surface passivation.......................................................................... 1096.4.2.2 Water-limited carbonation ............................................................... 1106.4.2.3 Conceptual models of water-limited carbonation ........................... 1146.4.2.4 Reactive transport modeling of water-limited carbonation ............. 1166.4.3 Preferential flow paths ........................................................................... 118xi6.5 Implications ...................................................................................................... 1216.5.1 Mineral dissolution-precipitation reactions in the vadose zone ............ 1216.5.2 CO2 sequestration .................................................................................. 1237. Offsetting of CO2 emissions by air capture in mine tailings at the Mount Keith Nickel Mine, Western Australia: Rates, controls and prospects for carbon neutral mining  ........................................................................................................................ 1257.1 Introduction ...................................................................................................... 1257.2 Locality and sampling strategy ........................................................................ 1287.2.1 The Mount Keith Nickel Mine .............................................................. 1287.2.2 Strategy for sampling at Mount Keith ................................................... 1307.3 Analytical and modeling methods ................................................................... 1317.4 Qualitative mineralogical results and field observations ................................. 1327.4.1 Qualitative mineralogy of Mount Keith mine tailings ........................... 1327.4.2 Occurrence of hydromagnesite mineralization ...................................... 1337.4.3 Consequences of timing and depth of tailings deposition for hydromagnesite formation ............................................................................................... 1357.5 Analytical results ............................................................................................. 1387.5.1 Rietveld refinement results .................................................................... 1387.5.2 Stable isotopic results ............................................................................ 1437.5.3 Radiocarbon results ............................................................................... 1447.6 Discussion of mineralogical and isotopic results ............................................. 1467.6.1 Mineralogical change and formation of hydromagnesite ...................... 1467.6.1.1 Occurrence of efflorescent minerals ................................................ 1467.6.1.2 Relationship of hydromagnesite to primary gangue minerals ......... 1487.6.2 Carbon reservoir fingerprinting ............................................................. 1517.6.2.1 Fingerprinting with stable isotopes ................................................. 1517.6.2.2 Fingerprinting with radiocarbon ..................................................... 153xii7.7 Reactive transport modeling of hydromagnesite formation in TSF2 ............... 1567.7.1 Reactive transport model description and initial conditions ................. 1567.7.2 Reactive transport modeling results and discussion .............................. 1597.7.2.1 Mineral dissolution and hydromagnesite precipitation rates .......... 1597.7.2.2 Occurrence of hydromagnesite at depth .......................................... 1637.8 Potential for accelerating carbon mineralization ............................................. 1647.8.1 Tailoring tailings deposition rates to enhance reaction with atmospheric CO2  ....................................................................................................... 1657.8.2 Direct injection and circulation of CO2-rich gas streams and DIC-charged waters ..................................................................................................... 1667.9 Tailings management practices for carbon neutral mining .............................. 1698. Conclusion .................................................................................................................. 1738.1 Summary of research outcomes ....................................................................... 1738.2 Accelerating carbon mineralization in mine tailings ....................................... 1768.3 Suggestions for future research ........................................................................ 178References ............................................................................................................................ 180Appendices ........................................................................................................................... 208Appendix 1: Appendix to Chapter 2 ................................................................................ 208A1.1 Detailed methods ............................................................................................. 208A1.1.1 Brucite carbonation procedure .............................................................. 208A1.1.2  Analyses ............................................................................................... 210A1.2 Detailed results ................................................................................................. 213A1.2.1 Analytical results ................................................................................... 213A1.2.2 Gas composition .................................................................................... 214A1.2.3 Rate laws used in geochemical models ................................................. 214Appendix 2: Appendix to Chapter 4 ................................................................................ 227A2.1 Supplementary methods ................................................................................... 227xiiiA2.1.1 Micromodel ........................................................................................... 227A2.1.2 Characterization of initial brucite ......................................................... 227A2.1.3 Videos  ................................................................................................... 228A2.1.4 Geochemical modeling ......................................................................... 229A2.2 Supplementary discussion ................................................................................ 230A2.2.1 Raman spectroscopy ............................................................................. 230A2.2.2 Geochemical modeling ......................................................................... 230Appendix 3: Appendix to Chapter 5 ................................................................................ 234A3.1 Detailed experimental design ........................................................................... 234A3.1.1 Initial solution composition and volume ............................................... 234A3.1.2 Gas supply ............................................................................................. 234A3.2 Detailed analytical methods ............................................................................. 235A3.2.1 Characterization of solids ...................................................................... 235A3.2.2 Initial aqueous chemistry ...................................................................... 237A3.2.3 Soil water characteristic curves and gravimetric moisture content ....... 238A3.3 Transport parameters and initial conditions employed in reactive transport models .............................................................................................................. 239A3.4 Detailed results ................................................................................................. 241A3.4.1 Particle size distribution ........................................................................ 241A3.4.2 CO2 breakthrough .................................................................................. 241A3.4.3 Extent of carbonation and water content ............................................... 250A3.4.4 Energy dispersive spectroscopy results ................................................. 259A3.5 Discussion ........................................................................................................ 260A3.5.1 Threshold model functional form.......................................................... 260A3.5.2 CO2 injection pressure calculations ...................................................... 261Appendix 4: Appendix to Chapter 6 ................................................................................ 263A4.1 Detailed methods ............................................................................................. 263xivA4.1.1 Detailed experimental methods ............................................................. 263A4.1.2 Sampling techniques ............................................................................. 263A4.1.3 Characterization of solids ...................................................................... 264A4.1.4 Aqueous chemistry ................................................................................ 266A4.1.5 Stable carbon isotopic composition ...................................................... 266A4.1.6 Soil water characteristic curves and volumetric moisture content ........ 267A4.2 Transport parameters and initial conditions employed in reactive transport models .............................................................................................................. 268A4.3 Detailed results ................................................................................................. 270A4.3.1 Particle size distribution of the brucite ore ........................................... 270A4.3.2 Soil-water characteristic curve .............................................................. 270A4.3.3 Qualitative characterization of initial material ...................................... 271A4.3.4 Water saturation and temperature .......................................................... 271A4.3.5 Reaction stoichiometry, mineralogy, and carbonation extent ............... 274A4.3.6 Energy dispersive spectroscopy results ................................................. 290A4.3.7 Stable carbon isotopic composition   .................................................... 291A4.3.8 Water film model and modeling results ................................................ 295A4.4 Pore volume change calculations ..................................................................... 297Appendix 5: Appendix to Chapter 7 ................................................................................ 299A5.1 Initial and boundary conditions and parameters used in reactive transport models .............................................................................................................. 299A5.2 Detailed CO2 flux and carbonation rate calculations ....................................... 304xvList of TablesTable 2.1 CO2 mass balance and brucite dissolution and carbonation rates ................... 18Table 5.1 Summary of experimental conditions ............................................................. 52Table 5.2 Summary of mass of CO2 sequestered, carbonation rate, reaction stoichiometry, and extent of carbonation in experiments for which solids were analyzed .... 61Table A1.1 Sources of crystal structure data for Rietveld refinement ............................. 212Table A1.2 Raw δ13C values of dissolved inorganic carbon and carbonate solids .......... 216Table A3.1 Sources of crystal structure data for Rietveld refinement ............................. 239Table A3.2 Initial conditions applied in MIN3P-DUSTY simulations ........................... 239Table A3.3 Brucite volume fractions and surface areas in experiments compared to models ........................................................................................................... 240Table A3.4 Physical and transport parameters employed for simulations using MIN3P-DUSTY ......................................................................................................... 241Table A3.5 Gas effluent composition in the very fine and fine brucite columns ............. 243Table A3.6 Gas effluent composition in the 35% and 50% saturated medium brucite columns ......................................................................................................... 246Table A3.7 Gas effluent composition in the 35% and 50% saturated medium brucite duplicate columns ......................................................................................... 248Table A3.8 Rietveld refinement and carbon content results with depth from the very fine and fine brucite columns ............................................................................... 253Table A3.9 Rietveld refinement and carbon content results with depth from the medium brucite columns with different water saturations .......................................... 254Table A3.10 Total mass of solids and water at the end of the experiment in the very fine and fine brucite columns ...................................................................................... 255xviTable A3.11 Total mass of solids and water at the end of the experiment in the medium brucite columns with different water saturations .......................................... 255Table A3.12 Mass of CO2 gained through time in the very fine and fine 35% saturated columns and the rinsed 35% saturated and 15% saturated medium brucite trial 1 columns ...................................................................................................... 256Table A3.13 Mass of CO2 gained through time in the 35% and 50% saturated medium brucite trial 1 columns ............................................................................................... 257Table A3.14 Mass of CO2 gained through time in the 15%, 35%, and 50% saturated medium brucite column duplicate experiments .......................................................... 258Table A4.1 Sources of crystal structure data for Rietveld refinement ............................. 268Table A4.2 Initial brucite volume fractions and surface areas employed in the MIN3P-DUSTY simulations and experimental brucite volume fractions and surface areas .............................................................................................................. 269Table A4.3 Physical and transport parameters employed for simulations using MIN3P-DUSTY ......................................................................................................... 269Table A4.4 Rietveld refinement and total solid phase CO2 (%CO2) data for 35% 1 ....... 277Table A4.5 Rietveld refinement and total solid phase CO2 (%CO2) data for 35% 2 ....... 278Table A4.6 Rietveld refinement and total solid phase CO2 (%CO2) data for the 60% saturated experiment ..................................................................................................... 279Table A4.7 Solid phase CO2 mass balance for 35% 1 ..................................................... 280Table A4.8 Solid phase Mg mass balance for 35% 1 ...................................................... 282Table A4.9 Solid phase H2O mass balance for 35% 1 ..................................................... 283Table A4.10 Overall reaction stoichiometry and stoichiometry of flakey carbonate phase in 35% 1 ............................................................................................................ 284xviiTable A4.11 Solid phase CO2 mass balance for the 60% saturated column ...................... 285Table A4.12 Solid phase Mg mass balance for the 60% saturated column ....................... 287Table A4.13 Solid phase H2O mass balance for the 60% saturated column...................... 288Table A4.14 Overall reaction stoichiometry and stoichiometry of flakey carbonate phase in the 60% saturated column ............................................................................. 289Table A4.15 Stable carbon isotopic composition of solid experimental products ............ 292Table A4.16 Stable carbon isotopic composition of gas phase CO2 and dissolved inorganic carbon (DIC) versus time in 35% 2 .............................................................. 293Table A4.17 Stable carbon isotopic composition of gas phase CO2 and dissolved inorganic carbon (DIC) versus time in the 60% saturated experiment ......................... 294Table A5.1 Initial mineral abundances, effective reactive surface areas, and dissolution rate laws applied in MIN3P ................................................................................. 302Table A5.2 Water chemistry of initial pore water, replenished process water, and rain water (boundary solution) ....................................................................................... 303Table A5.3 Physical and transport parameters employed for simulations using MIN3P .303xviiiList of FiguresFigure 1.1 Conceptual diagram of the scales of processes studied in this thesis ............... 6Figure 2.1 Schematic of carbon mineralization in situ in a tailings storage facility ........ 14Figure 2.2 Plots of pH in high pCO2 experiments and dissolved inorganic carbon in long-term atmospheric CO2 experiment versus time ............................................... 16Figure 2.3 Plots of dissolved inorganic carbon concentration and Mg concentration in high pCO2 experiments versus time ........................................................................ 17Figure 2.4 Experimentally determined and modeled brucite carbonation rate versus pCO2 in experiments using ligand and pH dependent brucite dissolution rate laws ......................................................................................................................... 20Figure 2.5 Deviation from calculated equilibrium δ13CDIC values in high pCO2 experiments versus time (A). Deviation from calculated equilibrium δ13C values of solids with the average δ13CDIC during carbonation in high pCO2 experiments versus time of sampling (B) ....................................................................................... 24Figure 2.6 Comparison of range of passive annual carbonation rates at various mine sites with annual mine site greenhouse gas emissions, total sequestration capacity, and potential annual carbonation rates via accelerated brucite carbonation ... 27Figure 3.1 Time to CO2 venting versus tailings deposition rate and historic tailings depth at various brucite contents with a CO2 injection rate equal to the estimated rate of point source CO2 emissions at the Mount Keith Nickel Mine . .................. 33Figure 3.2 Comparison of range of passive annual carbonation rates at various mine sites with annual greenhouse gas emissions at mine sites of various size and power sources ............................................................................................................. 34Figure 4.1 Conceptual schematic and micrographs of brucite carbonation ..................... 39Figure 4.2 Transmitted light micrographs of nesquehonite precipitates .......................... 41xixFigure 4.3 Entrainment of brucite particles by retreating water meniscus during evaporation ...................................................................................................... 42Figure 4.4 Modeled carbonation rate versus mineral-water interfacial area and volume percent brucite (A) and Damköhler number versus mineral-water interfacial area (B) ............................................................................................................ 45Figure 5.1 Schematic of experimental apparatus ............................................................. 51Figure 5.2 Mineral abundance profiles determined using Rietveld refinement of X-ray diffraction data for 35% saturated columns with very fine, fine, and medium brucite ............................................................................................................. 57Figure 5.3 Mineral abundance profiles determined using Rietveld refinement of X-ray diffraction data for columns containing medium brucite at 15%, 35%, and 50% water saturation ............................................................................................... 58Figure 5.4 Mass of CO2 sequestered versus time calculated based on the column mass gain over time (A). Instantaneous carbonation rate versus time in the column experiments (B and C) .................................................................................... 59Figure 5.5 CO2 breakthrough curves measured at column outlets versus time and MIN3P-DUSTY modeling results ................................................................................ 60Figure 5.6 Scanning electron micrographs of initial and reacted material ....................... 64Figure 5.7 Conceptual diagram of the four reaction stages .............................................. 68Figure 5.8 Absolute carbonation rate for all 35% saturated columns of different grain size versus BET measured initial brucite surface area ........................................... 69Figure 5.9 The extent of brucite carbonation and the calculated percent reduction in gaseous CO2 flux as a function of various degrees of water saturation .......... 81Figure 6.1 Schematic of experimental apparatus ............................................................. 86Figure 6.2 CO2 breakthrough curves at three positions along the flow path plotted as C/C0 versus time ...................................................................................................... 92Figure 6.3 Mass of CO2 sequestered and instantaneous carbonation rate versus time ..... 93xxFigure 6.4 Mineral and solid phase CO2 abundance profiles ........................................... 94Figure 6.5 Representative scanning electron micrographs of experimental products ...... 97Figure 6.6 Aqueous chemistry data versus time. A) Dissolved inorganic carbon concentrations, B) pH, and C) Mg concentrations versus time. D) Mineral saturation indices versus time ....................................................................... 100Figure 6.7 Stable carbon isotopic composition of gaseous CO2, dissolved inorganic carbon, and secondary carbonate solid phases ........................................................... 103Figure 6.8 Analysis of CO2 breakthrough curve features ............................................... 106Figure 6.9 CO2 sequestered versus final water saturation .............................................. 112Figure 6.10 Conceptual models of water limited carbonation. A) Dry pore model. B) Water film model ..................................................................................................... 117Figure 6.11 Comparison of modeled versus experimental CO2 breakthrough curves from 35% 1 ............................................................................................................ 119Figure 6.12 Comparison of modeled versus experimental brucite and secondary phase CO2 abundance profiles from 35% 1 and 2 ........................................................... 120Figure 6.13 Photographs of the 60% saturated column at the conclusion of the experiment.. ................................................................................................... 122Figure 7.1 The Tailings Storage Facilities at Mount Keith and location of Mount Keith Nickel Mine within Australia ........................................................................ 129Figure 7.2 Backscattered electron images of hydromagnesite in Mount Keith mine tailings ........................................................................................................... 133Figure 7.3 Mount Keith mine tailings at the surface and at depth within tailings storage facility 2 ........................................................................................................ 136Figure 7.4 Variation of hydromagnesite abundance with depth beneath the surface of tailings storage facility 2 over time ............................................................... 139Figure 7.5 Variation of the abundance of select secondary minerals with depth beneath the surface of tailings storage facility 2 over time .............................................. 140xxiFigure 7.6 Variation of the abundance of gangue carbonate minerals compared to hydromagnesite abundance with depth beneath the surface of tailings storage facility 2 over time ........................................................................................ 141Figure 7.7 Variation of the abundance of brucite and serpentine-group minerals compared to hydromagnesite abundance with depth beneath the surface of tailings storage facility 2 over time ........................................................................................ 142Figure 7.8 Stable oxygen and carbon isotope data for different modes of occurrence and mineralogy of carbonate minerals at Mount Keith ....................................... 144Figure 7.9 Stable carbon and fraction modern carbon data for secondary precipitates of hydromagnesite, bedrock carbonate minerals, and soda ash from Mount Keith.. ............................................................................................................ 145Figure 7.10 Amount of brucite remaining versus time at different rates of tailings deposition as estimated using model output ................................................................... 166Figure 7.11 Brucite carbonation rate versus saturated hydraulic conductivity ................ 168Figure 8.1 Conceptual diagram of limitations on brucite carbonation observed in the experimental studies conducted for this thesis .............................................. 177Figure A1.1 Schematic of brucite carbonation setup ........................................................ 211Figure A1.2 Mg concentration versus time in the long term atmospheric CO2 experiment and experimental controls ............................................................................. 218Figure A1.3 CO2(g) content of combined exhaust from duplicate, simultaneous, high pCO2 reactors in the 50% CO2 and 10% CO2 experiments versus time ................. 219Figure A1.4 Temperature in high pCO2 experiments versus time ..................................... 220Figure A1.5 Carbon abundance (%C) versus time in high pCO2 experiments and experimental controls and the long term atmospheric pCO2 experiment ........................... 221Figure A1.6 X-ray diffraction patterns of solids in the high pCO2 experiments versus time ............................................................................................................... 222xxiiFigure A1.7 X-ray diffraction patterns of initial solids in all experiments and final solids in the long-term atmospheric CO2 experiment .................................................. 223Figure A1.8 Scanning electron micrographs of initial solids in all experiments (A), and final solids in the high pCO2 experiments........ ..................................................... 224Figure A1.9 Comparison of brucite dissolution rate versus pH recorded in this study and by Pokrovsky and Schott (2004), and comparison of brucite dissolution rate versus log [HCO3-] in this study and recorded by Pokrovsky et al. (2005) ............. 225Figure A1.10 Nesquehonite saturation index versus time in all high pCO2 experiments ... 226Figure A2.1 Schematic of the micromodel pore network ................................................. 228Figure A2.2 Raman spectra of reaction products from five locations within the micromodel following the second (8 h) experiment ......................................................... 231Figure A2.3 Brightfield light and scanning electron micrographs of lansfordite crystals......... ................................................................................................. 232Figure A2.4 Lansfordite saturation as a function of mineral-water interfacial area and brucite dissolution ..................................................................................................... 233Figure A3.1 Particle size distribution of initial very fine, fine, and medium brucite ore .. 242Figure A3.2 Representative X-ray diffraction pattern and Rietveld refinement plot of reacted solids from the fine brucite column .............................................................. 251Figure A3.3 Final volumetric water content versus depth measured at the conclusion of the experiments ................................................................................................... 252Figure A3.4 Scanning electron micrographs and energy dispersive spectra of representative reaction products ........................................................................................... 259Figure A3.5 Calculated brucite dissolution rate and percent brucite remaining versus time in a hypothetical column containing 10 wt.% brucite ore ................................. 260Figure A4.1 Particle size distribution of the brucite ore ................................................... 270Figure A4.2 Volumetric water content versus matric suction from the Tempe pressure cell tests of duplicate initial LM50 quartz samples ............................................. 271xxiiiFigure A4.3 Representative scanning electron micrographs of initial solids .................... 272Figure A4.4 Temperature and volumetric water content versus time in all experiments .. 273Figure A4.5 Representative X-ray diffraction pattern and Rietveld refinement plot of reacted solids from the 40-50 cm depth interval in the 60% saturated column. ....... 275Figure A4.6 Color of experimental end products from 35% 2 and its correspondence to the extent of carbonation ..................................................................................... 276 Figure A4.7 Representative scanning electron micrographs and energy dispersive spectra of experimental reaction products ..................................................................... 290Figure A4.8 The functional relationship between the maximum extent of reaction possible and water saturation as employed in the MIN3P-DUSTY models ............... 295Figure A4.9 Carbonate mineral abundance versus depth in the 35% saturated experiments and predicted by the MIN3P-DUSTY models .............................................. 296Figure A4.10 Gas filled pore volume change versus the molar ratio of water to carbonate for different Mg-carbonate minerals ................................................................... 298xxivList of Symbols and AbbreviationsGHG – greenhouse gas IPCC – Intergovernmental Panel on Climate ChangeMKM – Mount Keith Nickel MinepCO2 – CO2 partial pressureDIC – dissolved inorganic carbonwt.% - weight percentXRD – X-ray diffractionLGR – Los Gatos Research[Mg] – aqueous magnesium concentrationδ13C – delta carbon-13δ13CDIC – delta carbon-13 of dissolved inorganic carbonδ13CCO2(g) – delta carbon-13 of gaseous CO2VPDB – Vienna Pee Dee BelemniteSEM – scanning electron microscopynxi – initial moles of ‘x’SA – surface areat – timemix – initial mass of ‘x’mfx – final mass of ‘x’Mx – molar mass of ‘x’Cr – reactive capacityFbrc – brucite content of tailings as fraction of tailings massmt – mass of tailingsGFWbrc – gram formula weight of bruciteb – stoichiometric coefficient for carbonation of brucitexxvrCO2 – rate of CO2 supplySbrc – surface area of brucite in tailingsrbrc – estimated carbonation rate of brucite with ‘flue gas’ (~10% CO2)USEPA – United States Environmental Protection AgencyGWI – gas-water interfacial areaGMI – gas-mineral interfacial areaMWI – mineral-water interfacial areaDa – Damköhler numbertres – residence time of the fluidk – kinetic dissolution rate constantCeq – solubility of the mineral phasevol.% – volume percentXRF – X-ray fluorescence spectroscopyBET – Brunauer–Emmett–Tellerrinstant – instantaneous carbonation rate (g CO2 h-1)mi – column mass at time ‘i’rbrc – brucite dissolution ratekeff0 – effective rate constantk0 – initial kinetic dissolution rate constant[HCO3-] – aqueous bicarbonate concentrationΩ – saturation ratioSX – reaction stage ‘X’brc – bruciteqtz – quartznsq – nesquehoniteVWC – volumetric water contentMgC –ratio of moles Mg consumed to moles CO2 sequestered during carbonationxxvinxf – final moles of ‘x’φ0 – initial brucite volume fractionφt – brucite volume fraction at time ‘t’φp – threshold brucite volume fractionkeffg – effective geometric rate constantkefft  – effective threshold rate constantLM – Lane Mountainφz0 – initial volume fractions at depth ‘z’φzp – initial volume fractions at depth ‘z’Sw – water saturation (fraction of pore volume occupied by water)CCS – carbon capture and storageδ18O – delta oxygen-18F14 C – fraction modern carbonTSF – tailings storage facilitySSAMS – Single Stage Accelerator Mass SpectrometerMAD – median absolute deviationVSMOW – Vienna Standard Mean Ocean WaterαA-B – isotopic fractionation factor between substances ‘A’ and ‘B’pO2 – O2 partial pressureKsat – saturated hydraulic conductivityAUD – Australian dollarsEU ETS – European Union Emissions Trading SchemeXRPD – X-ray powder diffractionθ – scattering angle of X-raysPDF-4+ – Powder Diffraction file, version 4+ICP-OES – inductively coupled plasma - optical emission spectrometryTOC-TN – total organic carbon - total nitrogenxxviiRsample – 13C/12C ratio in a sampleRstandard – 13C/12C ratio in a standardFE-SEM – field emission scanning electron microscopyIAP – ion activity productK – equilibrium constantkMg – brucite dissolution rate constant in the absence of ligands at pH 8.5KMg – HCO3- – adsorption constant of HCO3- on brucite surface siteskHCO3- – rate constant in the presence of HCO3-nd – no dataLLNL – Lawrence Livermore National LaboratoriesNIST – National Institute of Standards and TechnologyEDS – energy dispersive spectroscopy/spectrumα – van Genuchten parametern – van Genuchten parameterm – van Genuchten parameterSrl – residual water saturationqg – Darcy fluxkrg – relative permeability with respect to the gaseous phaseμg – viscosity of gaseous phasek – intrinsic permeabilitypg – gaseous phase pressureρg – density of the gaseous phaseg – acceleration due to gravityz – vertical distanceSea – effective saturationqg0 – initial Darcy flux at residual saturationΔVs – solid volume changexxviiivX – stoichiometric coefficient of a given reactant or productρ – densityΔVw – water volume changeΔVp – gas-filled pore volume changembgs – meters below ground surfaceA – cross-sectional areaQ – volumetric flow ratedhdl – head gradientrCO2 – CO2 sequestration rate (mol s-1)xxixList of Supplementary MaterialsVideo AM1. Brucite carbonation in a micromodel. The video consists of time lapse images taken at 5 minute intervals using a light microscope played at two frames per second. The entire duration shown is ~7.3 h with a field of view of 3.4 mm × 2.2 mm. Black frames replace frames that have been removed due to technical problems with the microscope. Red arrows indicate two occurrences of nesquehonite precipitation in the brucite-rich zone. This video can be found in the Accompanying Materials located in the Supplementary Thesis Materials and Errata Collection: http://hdl.handle.net/2429/51487Video AM2. Carbonate precipitation and brucite carbonation in a micromodel. The video consists of time lapse images taken at 5 minute intervals using a light microscope played at two frames per second. The entire duration shown is ~10 h with a field of view of 1.9 mm × 1.3 mm. Black frames replace frames that have been removed due to technical problems with the microscope. This video can be found in the Accompanying Materials located in the Supplementary Thesis Materials and Errata Collection: http://hdl.handle.net/2429/51487xxxAcknowledgementsThere are many people to whom I owe thanks for their help and support during the past four years. First, I would like to thank my supervisor, Greg Dipple who provided support and advice despite having the problems of the department resting on his shoulders. Greg, thanks for your mentorship and for encouraging me to do a PhD and inspiring me to work harder and learn more than I ever expected. I should have known from our first committee meeting when you joked that I had four years worth of work to do for my Master’s that it wasn’t really a joke…I am also very grateful to Uli Mayer for graciously helping with everything modeling and transport related. It seemed that you went out of your way to provide help and encouragement, and I really appreciated this. I would also like to thank Mark Johnson for serving on my supervisory committee and providing thoughtful commentary, and Al Lewis and Mati Raudsepp for helping with my candidacy exam. Mati, I appreciated your help with everything related to minerals and thanks for keeping your door open for questions. I’ll always keep in mind whenever my future endeavors become challenging that if it were easy “some chemist would have done it already.” I would also like to thank my external examiner, Damon Teagle, for thoughtful and insightful comments on this thesis. I appreciated the interesting and challenging questions posed to me at my defense by university examiners Roger Beckie and Dirk van Zyl, and appreciate the time they dedicated to reading this thesis. Many thanks, of course, are also due to The Carbonators (and to Trent for coining the name), a super-group of carbon sequesterers if ever there was one: Ian Power, Andreas Beinlich, Paul Kenward, and former members, Sasha Wilson and Shaun Barker. Ian, thanks for your support and advice both scientific and otherwise over the last four years, despite my constant arguments. You’ve been immensely helpful  – from lab work, to thoughtful edits, to the songs competition, and many coffee breaks and travelling adventures. I really can’t thank you enough! Thanks to Andreas for helpful discussions regarding porosity and permeability, Raman, and amusing German idioms, and for your help with formatting. Thanks, Paul, for always being available to chat about science or otherwise. Shaun provided valuable advice regarding isotopic measurements and searching for gas leaks. Thanks to Sasha for sharing your vast wisdom in the ways of Mg-carbonates and research in general, not to mention being an excellent roommate that was up for SEM parties in the hotel room. I am also forever grateful to you for your role in provoking the birdman incident at AGU. Thanks also to Gordon Southam and Jenine McCutcheon for their help with the CO2 sequestration research, particularly in the field, and for participation in epic caps games and hydromagnesite fights.  I would also like to thank my fellow graduate students and others for commiseration, xxxiincluding Leanne, Emma, Dana, Trent, Sara, Natasha, Sharon, Tim, Brendan, Craig, and Jack. Dana, thanks for putting up with my neglect of the apartment. Chelsea, Natasha, and Niamh, I appreciate your support from afar. Many people have helped with analytical and technical aspects of my research: thanks to Maureen Soon, Timothy Mah, Gethin Owen, Sally Finora, Jenny Lai, Edith Czech, Lan Kato, Elisabetta Pani, Jörn Unger, Kristal Li, and David Jones. Dave Sinton, Wen Song, and the rest of Dave’s group graciously hosted me at the U of T and provided valuable insight with respect to microfluidics.  I owe many thanks to my parents, Jed and Dawn for their support and encouragement through all my obsessions. Thanks to my dad for advice in science and outside (who would’ve thought I’d be using microfluidics?), and mom for your confidence in me and for looking out for Muse (thanks also to Cheri and Emily). Nigel, thanks for feeding me for a year so that I could babysit experiments and for providing new soundtracks that I’m sure improved my productivity. This project was funded by Carbon Management Canada (CMC) National Center of Excellence research grants to Profs. Gregory Dipple and Ulrich Mayer, and a Natural Sciences and Engineering Research Council of Canada (NSERC) research grant to Prof. Gregory Dipple. I am grateful for the funding provided from a University of British Columbia Four Year Fellowship and an NSERC Postgraduate Fellowship that allowed me to focus on research. Funding to conduct microfluidics experiments was provided by a Geological Society of America student research grant. CMC also generously provided support to attend their conferences and a summer school in Coventry, UK. xxxiiThis work is dedicated to my parents, Jed and Dawn, and to my brother, Nigel.Introduction11. Introduction1.1 CO2 sequestration via carbon mineralizationThe accumulation of anthropogenic greenhouse gases (GHGs), mainly CO2, in the atmosphere has been identified as a cause of global climate change (IPCC, 2007). Carbon sequestration is one of many strategies necessary to stabilize CO2 concentrations and prevent irreversible climate change while transitioning to alternative energy sources (Hoffert et al., 2002; Pacala and Socolow, 2004; Broecker, 2007). Carbon mineralization, also known as mineral carbonation, is a promising option for CO2 sequestration as the reaction products are environmentally benign and stable, providing little possibility of accidental release and reducing the need for post-storage monitoring (Seifritz, 1990; Lackner et al., 1995; Lackner et al., 1997; Lackner, 2003; Sipilä et al., 2008). Research in mineral carbonation has largely focused on industrial processes that use high temperatures and pressures (e.g., 185°C, 150 atm) to accelerate reaction rates (Béarat et al., 2002; Gerdemann et al., 2007; Sipilä et al., 2008; Zevenhoven et al., 2008; Fagerlund et al., 2009; Krevor and Lackner, 2009; Koukouzas et al., 2009), however the financial and energy costs of such methods has limited their implementation. Alternatives to this approach include injection of supercritical CO2 or CO2-rich gases or fluids into subsurface mafic and ultramafic formations (Kelemen and Matter, 2008; Alfredsson et al., 2008; Oelkers et al., 2008; Matter and Kelemen, 2009; Schaef et al., 2010; Gislason et al., 2010; Kelemen et al., 2011; Schaef et al., 2011; Van Pham et al., 2012; Paukert et al., 2012; Schaef et al., 2013a; Hellevang et al., 2013; Schaef et al., 2013b; Galeczka et al., 2014; Gislason and Oelkers, 2014), or alkaline industrial waste stockpiles (Bobicki et al., 2012; Power et al., 2013b and references therein), and direct air capture at Earth’s surface conditions via enhanced weathering of pulverized rock or wastes (Wilson et al., 2006; Renforth et al., 2009; Schuiling and Boer, 2010; Pronost et al. 2011; Renforth et al., 2011; Assima et al., 2012; Washbourne et al., 2012; Assima et al., 2013a; Assima et al., 2013b; Wilson et al., 2014). 2IntroductionAlthough of lower capacity than natural rock, carbonation of industrial wastes such as steel and blast furnace slag, waste from alumina production, alkaline and saline waste water, and mine tailings is a promising CO2 sequestration option, as it exploits available waste materials that tend to be inherently more reactive than natural minerals (Huijgen and Comans, 2006; Wilson et al., 2006; Dilmore et al., 2008; Eloneva et al., 2008; Ferrini et al., 2009; Khaitan et al., 2010; Power et al., 2010; Gunning et al., 2010; Wilson et al., 2010; Pronost et al., 2011; Morales-Flórez et al., 2011; Back et al., 2011; Mignardi et al., 2011; Power et al., 2011b; Bobicki et al., 2012; Power et al. 2013a; Power et al., 2013b; Power et al., 2014a; McCutcheon et al., 2014). The study of controls on reactivity of these materials provides valuable insight into reaction mechanisms that may also operate in natural mafic and ultramafic rock formations that have greater total sequestration capacity, while providing a meaningful offset of GHG emissions at the industry level (Power et al. 2013c; Power et al., 2013b). For example, Power et al. (2013b) estimate that conscription of mining wastes produced globally for carbon mineralization could offset approximately 1.5% of annual global CO2 emissions.1.2 Carbon	mineralization	in	ultramafic	mine	tailingsIt has been well documented that surface weathering of ultramafic mine tailings sequesters atmospheric CO2 under normal mining conditions via formation of hydrated Mg-carbonate phases (Wilson et al., 2006; Wilson, 2009; Wilson et al., 2009a; Wilson et al., 2011; Bea et al., 2012; Pronost et al., 2012; Beinlich and Austrheim, 2012; Oskierski et al., 2013). This passive process has been observed at historic Canadian and Australian chrysotile mines (Wilson et al., 2006; Wilson et al., 2009a; Pronost et al., 2012; Oskierski et al., 2013), in historic chromite mine shafts (Beinlich and Austrheim, 2012), and at the active Australian Mount Keith Nickel Mine (MKM) (Wilson, 2009; Bea et al., 2012) and Canadian Diavik Diamond mine (Wilson et al., 2011). Passive carbon mineralization reactions in these environments are facilitated by the high surface areas of the tailings minerals (Wilson et al., 2009a), yet are limited by the uptake 3Introductionof CO2 into solution (Wilson et al., 2010). Although the carbon sequestration potential of ultramafic tailings is significant, it remains largely under-utilized. Large mines, such as MKM, have the capacity to offset the GHG emissions of their mining operations, and potentially emissions from other sources. Complete carbonation of tailings produced annually at MKM (~11 Mt; BHP Billiton, 2005) exceed total annual mine emissions by more than a factor of ten. Yet, Wilson (2009) estimates that passive carbonation rates at MKM currently offset annual mine emissions by only ~15%. In order to exploit the full CO2 sequestration capacity at mine sites, current carbon mineralization rates must be accelerated.  A growing body of research has provided insight into the rates and mechanisms of passive carbon mineralization in mine tailings and wastes (Wilson et al., 2006; Wilson et al., 2009a; Wilson et al., 2011; Bea et al., 2012; Pronost et al., 2012; Beinlich and Austrheim, 2012; Oskierski et al., 2013), and several recent experimental studies have elucidated controls on the sequestration capacity and rate of reaction of bulk tailings (Pronost et al., 2011; Assima et al., 2012; Assima et al., 2013a; Assima et al., 2014a; Assima et al., 2014b). Recent developments in reactive transport modeling have also provided insight into the controls on passive carbonation rates (Bea et al., 2012). However, the development of strategies to accelerate carbonation rates at mine sites is still in its infancy (e.g., Power et al., 2014a), and the capability of reactive transport models to capture the governing reaction mechanisms and to predict the fate of CO2 under accelerated carbonation scenarios remains untested. Here, strategies that accelerate carbon mineralization reactions by increasing the supply of CO2 were investigated experimentally (Chapters 2-6). These experimental trials were supplemented by a collaborative study of the mechanisms of passive carbon mineralization in the active tailings storage facility at the Mount Keith Nickel Mine (Chapter 7). Experimental studies were focused on reaction of brucite [Mg(OH)2], a mineral commonly present at low abundance (~1-15 wt.%) in ultramafic mine tailings and residues (Wilson, 2009; Chrysochoou et al., 2009; Pronost et al., 2011; Assima et al., 2012; Bea et al., 2012; Assima et al., 2013a; Assima et al., 2013b; Assima et al., 2014a). Brucite is highly reactive in comparison to the more abundant silicate 4Introductionphases such as forsterite [Mg2SiO4] and serpentine-group minerals [Mg3Si2O5(OH)4] (Bales and Morgan, 1985; Pokrovsky and Schott, 2000; Pokrovsky and Schott, 2004; Pronost et al., 2011; Assima et al., 2013a; Daval et al., 2013; Thom et al., 2013; Power et al., 2013b; Assima et al., 2014a) and its carbonation could provide immediate and significant CO2 offsets at mine sites (e.g., up to ~60% offset of annual mine emissions at MKM). To better predict the fate of CO2 in tailings storage facilities and to estimate CO2 sequestration capacity, the capabilities of the existing reactive transport code, MIN3P (Mayer et al., 2002; Molins and Mayer, 2007; Bea et al., 2012), were extended to include new processes that were experimentally observed to govern carbonation. The insight provided by the experiments and improved modeling capabilities will help guide implementation of accelerated carbonation strategies at mine sites, with implications for carbonation of other industrial wastes and natural rock of similar mineralogical composition.1.3 Fundamental processes of water-rock interaction in the Earth’s shallow subsurfaceAs a naturally occurring process, carbon mineralization is also an important part of the global carbon cycle that regulates atmospheric CO2 concentrations over geologic time (Berner et al., 1983). However, identification of rate-controlling mechanisms is challenging in natural systems due to the multitude of factors that may concurrently affect reaction rates (Maher et al., 2006). The prediction of reaction rates in water unsaturated porous media is hindered by difficulties in accurately quantifying reactive surface area, capturing the effect of pore scale chemical and physical heterogeneities on reaction rates, and the response of these variables to dynamic wetting and drying processes. For instance, secondary precipitates may coat the surfaces of reacting minerals with the potential to passivate reactive surfaces, but the effect of these coatings is variable (c.f., Hodson, 2003; Park and Fan, 2004; Cubillas et al., 2005; Béarat et al., 2006; Lekakh et al., 2008; Andreani et al., 2009; Huntzinger et al., 2009; Daval 5Introductionet al., 2009a; Daval et al., 2009b; Daval et al., 2011; Stockmann et al., 2011; Stockmann et al., 2013). Pore scale heterogeneities in water distribution may also reduce the surface area available for reaction by leaving some mineral surfaces unexposed to reactive fluids (e.g., Pačes, 1983). Wetting and drying cycles may alter the reactivity over time, particularly for reactions that involve hydrous minerals whose dissolution and precipitation also modify the local water content. The experimental investigation of carbon mineralization reactions, and brucite carbonation in particular, is a fitting approach to study water-rock-gas interactions in general, due to the coupled nature of the reactions and rapid reaction rates that permit significant reaction progress on experimental timescales. The experimental studies conducted for this thesis therefore provided insight into fundamental controls on reactivity during coupled dissolution-precipitation reactions in the shallow subsurface that are addressed in Chapters 4-6. The experiments spanned reaction scales from the fluid-mineral interface to pore to porous media scale using batch and microfluidic reactors, and decimeter and meter scale column apparatus (Fig. 1.1). The broad range of reaction scales studied provided important insight into the effects of water saturation, the evolution of reactive surface area, and the previously unheralded process of evaporative-driven particle movement on mineral reactivity in conditions representative of the Earth’s shallow subsurface. The experimental results are presented in the context of their application to CO2 sequestration and the broader implications for mineral dissolution-precipitation reactions in natural and anthropogenic porous media.1.4 Objectives and organization The objectives of this thesis are to elucidate the controlling reaction mechanisms during carbonation of brucite in order to help design CO2 sequestration strategies employing carbon mineralization in near surface environments, namely ultramafic mine tailings, and to ameliorate the fundamental understanding of mineral dissolution-precipitation reactions in unsaturated porous media. A combination of experimental, analytical, and reactive transport 6Introductionmodeling techniques are used to investigate brucite and mine tailings carbonation from the mineral-solution interface to field scale (Fig. 1.1).  Five manuscripts and a conference proceedings article have been assembled that describe these investigations. The second chapter of this thesis (Chapter 2) presents an experimental study conducted using “zero-dimensional” stirred batch reactors to ascertain the effect of increased CO2 supply on brucite carbonation rates. The zero-dimensional nature of these experiments allowed the surface controlled reactions at the mineral-water interface to be examined. This chapter presents a ‘proof-of-concept’ study that demonstrates the potential to accelerate mineral carbonation rates in mine tailings by supplying gas streams with elevated 12-14 cm6 cmInterface PoreDecimeterMeter5 cm0.2 cm90 cm71.266 kg25 cmFieldMine tailings storage facility3.0 LFigure 1.1. Conceptual diagram of the scales of processes studied in this thesis. Batch reactors were used to explore reactions at the mineral-solution interface, as detailed in Chapter 2, and extrapolated from in Chapter 3. Microfluidic reactors were employed to study reactions at the pore scale in Chapter 4. Decimeter and meter scale column reactors were used to elucidate reactive transport processes at the porous medium scale (Chapters 5 and 6). Numerical reactive transport modeling helped the interpretation of reaction mechanisms and determination of rates from the porous medium to field scale (Chapters 5-7).7IntroductionCO2 concentrations. This manuscript was published in Environmental Science & Technology in 2013. Chapter 3 is a follow-up to the work presented in Chapter 2, which uses the experimental results to evaluate potential CO2 sequestration strategies that could be employed at mine sites. The experimentally determined carbonation rates are applied to estimate rates of CO2 sequestration achievable at the Mount Keith Nickel Mine in Western Australia, compared to rates of CO2 supply. The rate at which CO2 could be injected into tailings as a function of brucite content and tailing deposition rates were estimated to outline conditions under which CO2 leakage to the atmosphere could be avoided. All calculations were conducted assuming rates from the zero-dimensional reactors would be portable to field scale, an assumption fraught with uncertainty. This manuscript was published in the peer-reviewed conference proceedings for the International Mine Water Association 2013 Mine Water Conference.  The fourth chapter is a study that investigated the fundamental controls on rates and distribution of mineral dissolution-precipitation reactions at the pore scale. Microfluidic techniques were applied to elucidate the impact of evaporative processes on CO2 dissolution and fixation in the unsaturated zone at the pore scale and revealed a number of previously unheralded processes that exert considerable influence on reaction rates. It is demonstrated that the dynamic evolution of the gas-water interface, water content, and mineral to water ratio may significantly alter reaction rates as the volume of water filled pore space evolves.  Chapter 5 presents an experimental examination of the reproducibility of the reaction rates achieved in the homogeneous batch reactors to a porous medium with one-dimensional flow, and the scaling-up of reaction processes observed at the interface and pore scale. The objective of this study was to evaluate the effect of water saturation and precipitation of hydrous Mg-carbonates at the brucite surface on reaction progress in an environment physically representative of natural and anthropogenic porous media, such as mine tailings. Decimeter scale column experiments containing brucite were supplied with 10% CO2 gas streams. Numerous studies have investigated the effect of surface coatings on mineral dissolution rates 8Introductionbut have reported conflicting results (c.f., Hodson, 2003; Park and Fan, 2004; Cubillas et al., 2005; Béarat et al., 2006; Lekakh et al., 2008; Andreani et al., 2009; Huntzinger et al., 2009; Daval et al., 2009a; Daval et al., 2009b; Daval et al., 2011; Stockmann et al., 2011; Stockmann et al., 2013). Such studies typically employ zero-dimensional, fluid dominated batch reactors (Park and Fan, 2004; Daval et al., 2009a; Daval et al., 2009b; Stockmann et al., 2011; Stockmann et al., 2013), and are often complicated by concomitant precipitation of both silicate and carbonate minerals (e.g., King et al., 2010), hampering unambiguous identification of the passivating effects of carbonates alone. The experiments conducted for this thesis used brucite, a silica-free mineral. Therefore, the effect of Mg-carbonate precipitates could be assessed directly, without the complication of silica layer formation. Moreover, passivating effects may differ in the mineral dominated columns compared to batch reactors, due to the limited pore volume in which secondary phases can form, and the relative scarcity of water. Experiments of varying water saturation were conducted in order to distinguish between surface passivation-limited and water-limited reaction at conditions relevant for the shallow subsurface, with additional insight from reactive transport modeling using MIN3P-DUSTY. A version of this chapter is currently under review for publication in a peer-reviewed journal.  In Chapter 6, the results of larger (meter-scale), instrumented column experiments are presented. The objectives of this study were to ascertain the impact of water saturation on the availability of reactive surface area and reactive capacity of a porous medium, and to investigate the evolution of the effective reactive surface area of brucite during reaction. The larger size of the experimental apparatus compared to those conducted for Chapter 5 allowed for installation of instrumentation and sampling ports at multiple locations along the flow path without significant disruption of gas flow. This provided better constraints for the further development of the reactive transport model used in Chapter 5, in particular, by providing spatial and temporal evolution of the gas phase CO2 concentration and pore water chemistry, important indicators of reaction progress. Moreover, the greater range in water saturation achieved along the flow path permitted a more in depth investigation of the effects of water 9Introductionsaturation on both brucite reactivity and gas flow. These data allowed for better assessment of the effect of water saturation on reactive surface area and CO2 sequestration capacity.  Chapter 7 was a collaborative effort that studied carbon mineralization reactions at field scale in the active tailings storage facility at the Mount Keith Nickel Mine in Western Australia. The objective of this study was to identify mechanisms of carbon mineralization and quantify the rate of passive CO2 sequestration. My contribution was to conduct reactive transport modeling using MIN3P (Bea et al., 2012) to describe the observed variation in tailings mineralogy and to estimate rates of CO2 fixation. Results of the modeling were used to make recommendations for tailings management and ore processing practices that would maximize CO2 sequestration at mine sites. Siobhan A. Wilson conducted most of the fieldwork at Mount Keith and the analyses of field samples, with contributions from several co-authors. These field data were used to constrain the reactive transport modeling. The modeling depends upon and complements the field data, therefore the entire manuscript, which includes Dr. Wilson’s contributions, is included as Chapter 7 in this thesis. Chapter 7 was published in International Journal of Greenhouse Gas Control (Wilson et al., 2014). In combination, this thesis provides new information on some of the fundamental controls on mineral dissolution-precipitation reactions in conditions representative of the Earth’s shallow subsurface, using brucite carbonation as a model reaction (refer to conclusions in Chapter 8). Owing to the relevance of brucite carbonation for CO2 sequestration strategies, this thesis makes a meaningful contribution towards the development of technologies to offset greenhouse gas emissions, with particular implications for CO2 sequestration in ultramafic mine tailings. 10Accelerated carbonation of brucite2. Accelerated carbonation of brucite in mine tailings for carbon sequestration12.1 IntroductionThe accumulation of anthropogenic greenhouse gases (GHGs), predominantly CO2, in the atmosphere has been identified as a cause of climate change (IPCC, 2007). Carbon sequestration is one of many strategies necessary to stabilize CO2 concentrations while transitioning to alternative energy sources (Hoffert et al., 2002; Lackner, 2003; Pacala and Socolow, 2004; Broecker, 2007). Carbon mineralization involves dissolution of non-carbonate minerals (e.g., silicates, hydroxides, and oxides) and subsequent precipitation of carbonate minerals, which sequesters CO2 (Seifritz, 1990; Lackner et al., 1995; Lackner et al., 1997; Lackner, 2003). This is a promising option for carbon sequestration as carbonate minerals are environmentally benign and stable, providing little possibility of accidental release and reducing the need for post-storage monitoring (Lackner et al., 1995; Sipilä et al., 2008). Research in carbon mineralization has largely focused on industrial processes that use elevated temperatures and pressures to accelerate carbonation reaction rates (e.g., 185°C, 150 atm; Béarat et al., 2002; Gerdemann et al., 2007; Sipilä et al., 2008; Zevenhoven et al., 2008; Fagerlund et al., 2009; Krevor and Lackner, 2009; Koukouzas et al., 2009), however, the financial and energy costs of such methods limits their viability. To avoid such costs, research has been directed towards developing carbon mineralization strategies at low temperature and pressure conditions (Manning, 2008; Power et al., 2009; Wilson et al., 2009a; Renforth et al., 2009; Schuiling and Boer, 2010; Pronost et al., 2011; Wilson et al., 2011). Carbonation of industrial wastes such as steel and blast furnace slag, and alkaline and saline waste water is advantageous as it exploits available waste materials (Huijgen and Comans, 2006; Wilson et al., 2006; Dilmore et al., 2008; Eloneva et al., 2008a; Ferrini et al., 2009; Power et al., 2010; Gunning et al., 2010; 1A version of this chapter is published and is reprinted with permission from Harrison, A. L., Power. I. M. and Dipple, G. M.  (2013) Accelerated carbonation of brucite in mine tailings for carbon sequestration. Environ. Sci. Technol. 47: 126–134. Copyright 2012 American Chemical Society.11Accelerated carbonation of bruciteWilson et al., 2010; Morales-Flórez et al., 2011; Back et al., 2011; Mignardi et al., 2011; Power et al., 2011b; Bobicki et al., 2012).  Passive carbon mineralization in mine tailings under normal mining conditions has been well documented (Wilson et al., 2006; Wilson et al., 2009a; Wilson et al., 2010; Wilson et al., 2011; Pronost et al., 2012). This process has been recognized at historic and active Canadian chrysotile and diamond mines (Wilson et al., 2006; Wilson et al., 2009a; Wilson et al., 2011), and at the active Mount Keith Nickel Mine (MKM) in Western Australia (Wilson, 2009). Carbonation of tailings is facilitated by high surface areas (Wilson et al., 2009a) yet at MKM, reaction rates are limited by the uptake of CO2 into solution (Wilson et al., 2010; Bea et al., 2012). If passive carbonation of bulk tailings were accelerated at large mines, it could more than offset the GHG emissions of mining operations. Supplying elevated (above atmospheric) partial pressures of CO2 (pCO2) into tailings may accelerate carbonation both by enhancing mineral dissolution due to increased acidity (Vermilyea, 1969; Pokrovsky and Schott, 2000; Pokrovsky and Schott, 2004) and promoting carbonate mineral precipitation due to elevated dissolved inorganic carbon (DIC) concentration. Carbonation of brucite [Mg(OH)2], a common but minor component of ultramafic mine tailings, offers significant sequestration potential. It has been documented at between 1-15 wt.% in chrysotile tailings, nickel tailings and deposits, and in chromite ore processing residue (Wilson, 2009; Chrysochoou et al., 2009; Pronost et al., 2011). In Québec, Canada, there is an estimated 2 Gt of chrysotile mining residue (Larachi et al., 2010), some of which contains 1.8 wt.% brucite (Pronost et al., 2011). If these brucite contents are representative of the entire chrysotile stockpile, up to 27 Mt of CO2 could be sequestered via brucite carbonation alone. At active mine sites, accelerated carbonation of brucite in tailings provides the potential to significantly offset the annual GHG emissions of mining operations. At MKM, ~11 Mt of tailings are produced annually, containing 0.11-0.28 Mt brucite (BHP Billiton, 2005; Wilson, 2009). Carbonation of the brucite produced annually would offset mine emissions (370 kt yr-1 CO2 equivalent; BHP Billiton, 2005) by 22-57%. Yet, passive carbonation rates at MKM and the 12Accelerated carbonation of bruciteBlack Lake chrysotile mine in Québec are estimated at only ~56 and ~0.6 kt yr-1, respectively (Wilson, 2009; Pronost et al., 2012). Incorporation of accelerated carbonation methods in the design of prospective mines offers the potential to significantly offset emissions throughout the lifetime of mining operations. At the Dumont Nickel deposit in Québec, for instance, assuming homogeneous brucite distribution in processed nickel ore of 10-15 wt.% (Pronost et al., 2011) and a similar scale of operations as MKM, accelerated brucite carbonation could exceed that predicted for MKM. Therefore, despite its minor abundance, brucite carbonation could provide a meaningful CO2 sink on the scale of mine emissions. The goal of this study was to investigate the mechanisms of brucite carbonation in mine tailings and to identify the rate-limiting steps in order to design methods for accelerating carbon sequestration in tailings. The effect of elevated pCO2 at 1 atm total pressure on the carbonation rate of brucite was investigated experimentally in batch reactors. Investigation of brucite carbonation may also guide ongoing work to accelerate carbonation of Mg-silicate minerals in tailings, which constitute the majority of ultramafic tailings and offer greater total sequestration capacity.2.2 Methods Brucite slurries (3.0 L) were prepared in side arm flasks (A1 Fig. A1.1). The solution composition mimicked that of pore water at MKM (1.0 M NaCl, and 0.1 M MgCl2·6H2O; Stolberg, 2005). High purity pulverized brucite ore (150 g; grain size = 2-40 μm diameter; surface area = 6.6 m2 g-1) from Brucite Mine, Nevada was added to each solution just prior to supplying CO2 gas. Rietveld refinement of X-ray diffraction (XRD) data indicated the brucite ore was between 90-95 wt.% brucite with minor dolomite and trace lizardite. Mixtures of CO2 and N2 gas were bubbled into continuously stirred slurries using glass sparge tubes. Gases were blended using a two-tube gas blender. CO2 concentration of the supplied gas stream was varied between experiments for values of 10%, 50%, and 100% CO2 at a total flow rate 13Accelerated carbonation of bruciteof 540 mL min-1, which was divided evenly between duplicate experiments that were run simultaneously. Duplicate experiments are henceforth referred to as ‘X% CO2 1 and 2.’ Three experiments were run using laboratory air (pCO2 ≈ 0.04%) supplied at ~270 mL min-1 using a peristaltic pump to act as experimental controls and establish the carbonation rate at low pCO2. Slight differences in the reactivity of duplicate reactors occurred due to imperfect mixing that resulted in occasional settling of brucite particles, which were re-suspended manually when observed. Flasks were vented through the side arm to maintain atmospheric pressure. The exhaust of high pCO2 experiments was directed into a Los Gatos Research (LGR)® off-axis integrated cavity output laser spectrometer for continuous measurement of CO2 concentration (Barker et al., 2011). All gas composition data represent the combined exhaust from duplicate, simultaneous, high pCO2 reactors.  Slurry temperature and pH were measured routinely and water samples for DIC and Mg concentrations ([Mg]), and δ13CDIC (Craig, 1957) were collected regularly. Solid samples were taken at the same time for mineralogical analysis using XRD, δ13C and carbon abundance (%C) analysis, and imaging using scanning electron microscopy (SEM). System mass before and after sampling and the mass of each sample were measured. At the end of each experiment, slurries were filtered to collect and weigh all solids. The experimental durations were 198, 56, and 72 h for the 10%, 50%, and 100% CO2 experiments, respectively. Two short-term experiments using atmospheric CO2 were run for 56 and 72 h to serve as experimental controls (i.e., systems with negligible carbonation). A third atmospheric CO2 experiment was conducted for 2856 h to establish background carbonation rates. The high pCO2 experiments were terminated upon achievement of steady-state conditions. For further details regarding experimental setup and analytical techniques, refer to Appendix 1 (A1). 14Accelerated carbonation of brucite2.3 Results and discussion2.3.1 Chemical environmentThere are three main requirements to allow carbonation to occur in situ in tailings facilities: (1) CO2 is available in solution as DIC, (2) cations are available, and (3) the chemical environment promotes both tailings mineral dissolution (cation leaching), and carbonate mineral precipitation (Fig. 2.1).In all reactors, initial pH values following brucite addition were between 9.3 and 9.4 and initial DIC concentration ranged from 15 to 26 mg L-1. Initial DIC content is attributed to dissolution of laboratory air into the saline solutions prior to sparging of gas. In the 10% and 50% CO2 reactors, an abrupt initial drop in pH was followed by a period of relatively constant pH before declining to a final stable pH of ~7.6 and ~7.1 in the 10% and 50% experiments, respectively (Fig. 2.2A). In the 100% CO2 reactors pH declined rapidly to a minimum stable value of ~6.9 after 10 h. Initial drops in pH were coincident with rapid increases in both DIC and [Mg] in all high pCO2 reactors owing to CO2(g) and brucite dissolution, respectively (Figs. 2.2A, 3A and B). The pH value and duration of the first pH stabilization and final pH were dependent on the pCO2 in the reactor atmospheres (Fig. 2.2A). During the first 6-8 h in the 10% CO2 experiments, DIC and [Mg] increased, before declining to reach near constant values during the period of 12-46 h (Fig. 2.3A-B). The relatively stable DIC during this period is 2. Tailings mineraldissolution1. CO2 available in solutionmineral buering + cation leachingpH decline3. Carbonate mineralprecipitationCO2(g)CO2(aq)HCO3-   +  H+DIC + cation removalCO2 injectionChemical environmentpHpCO2Tailings carbonationUltramac tailings stableUltramac tailings dissolution onlyHCO3-Mg2+Mg2+OH-Figure 2.1. Schematic of carbon mineralization in situ in a tailings storage facility.15Accelerated carbonation of bruciteindicative of a quasi-steady-state wherein brucite carbonation was balanced by CO2 uptake into solution. The final stable pH value was lower at higher pCO2, resulting in a greater proportion of Mg and DIC remaining in solution (Figs. 2.2A, 2.3A-B, Table 2.1). Final DIC and [Mg] stabilized approximately at equilibrium with the hydrated carbonate mineral, nesquehonite [MgCO3·3H2O], as calculated with PHREEQC (Parkhurst and Appelo, 1999) using the Pitzer database (Fig. 2.3A-B). In the long-term atmospheric CO2 experiment, pH reached a minimum of 9.0. An initial increase in DIC in this experiment to 80 mg L-1 after 192 h was followed by a decline to 33 mg L-1 after 2856 h (Fig. 2.2B). This implies DIC was removed faster than it was replaced by uptake of CO2(g) into solution, with the converse true at early time. A 0.02 M increase in [Mg] in this experiment was consistent with evapoconcentration, with no significant change in [Mg] in the controls (A1 Fig. A1.2). This implies brucite dissolution was slow in the controls, and was balanced by the rate of Mg-carbonate precipitation in the long-term experiment. In all high pCO2 experiments, the decline in pH from alkaline to circumneutral conditions was driven by the uptake of CO2. Although low pH accelerates brucite dissolution (Pokrovsky and Schott, 2004), carbonate precipitation is generally favored at higher pH values (Teir et al., 2007; Eloneva et al., 2008b; Ferrini et al., 2009; Krevor and Lackner, 2011). As such, some mineral carbonation processes have been designed such that cation extraction and carbonate precipitation are achieved in separate steps (Park and Fan, 2004; Teir et al., 2007; Sipilä et al., 2008; Eloneva et al., 2008b). In situ carbon mineralization in mine tailings requires these processes to occur concurrently. The chemical environment must promote tailings mineral dissolution yet permit carbonate precipitation. The high pCO2 experiments indicated that both brucite dissolution and carbonate precipitation are promoted with elevated pCO2 at atmospheric pressure and temperature, suggesting that accelerated in situ carbonation of brucite may be achieved in ultramafic tailings, provided the brucite content is sufficient. CO2 concentration in the exhaust gas from duplicate reactors in the 10% CO2 and 50% CO2experiments was reduced by up to 45% and 44%, respectively at the onset of each 16Accelerated carbonation of brucite56789100 25 50 75 100 125 150 175 200pHTime (h)Carbonation durationA10% CO 2 50% CO 2 100% CO 2 High pCO 2PHREEQC modeled equilibrium0204060801001201400 500 1000 1500 2000 2500 3000Time (h)8.58.78.99.19.39.5pHDIC (mg L-1)DIC at equilibrium with atmospheric CO 2BpHDICCarbonation durationLow pCO 250% CO 2100% CO 210% CO 2100%50%10%Figure 2.2. Plots of pH (A) in high pCO2 experiments versus time. Dashed lines are results from PHREEQC (Parkhurst and Appelo, 1999) models assuming kinetically controlled brucite dissolution and equilibrium CO2(g) dissolution. Dissolved inorganic carbon (DIC) (diamonds) and pH (+) in the long-term atmospheric CO2 experiment, and DIC (triangles) and pH (circles) in controls versus time (B). The dashed line in (B) indicates the DIC concentration at equilibrium with atmospheric CO2. Open symbols and solid symbols in (A) represent duplicates 1 and 2, respectively. Arrows above the graph in (A) represent the duration of the carbonation reaction in the 100%, 50% and 10% CO2 experiments, respectively.experiment, implying carbonation began immediately (A1 Fig. A1.3). No change in gas composition was measurable in the 100% CO2 experiment as the entire volume of gas consisted of CO2. In all high pCO2 experiments, 43-51% of CO2 supplied during the carbonation reaction was sequestered (Table 2.1). Owing to the exothermic nature of the reaction, an increase in slurry temperature in high pCO2 reactors over that of experimental controls was noted throughout carbonation, although the experiments were not insulated (A1 Fig. A1.4). Diurnal temperature variations in the laboratory resulted in periodic fluctuations in slurry temperature in all experiments (A1 Fig. A1.4). Control temperature was exceeded by a maximum of 5.4 ± 0.5°C. The enthalpy of reaction for brucite carbonation is -75.1 kJ mol-1 (Konigsberger et al., 1999), corresponding to a total of ~129 kJ of energy released in the 100% CO2 experiments based on the average mass of nesquehonite formed, providing the potential to heat the 3.0 L of water by 10.3°C. 17Accelerated carbonation of brucite00.10.20.30.40 25 50 75 100 125 150 175 200[Mg] (M)B10% CO 2 50% CO 2100% CO 2Carbonation duration01000200030004000500060000 25 50 75 100 125 150 175Time (h)200DIC (mg L-1)A10% CO 250% CO 2 100% CO 2Carbonation durationHigh pCO 250% CO 2100% CO 210% CO 2Time (h)High pCO 2PHREEQC modeled equilibrium100%50%10%100%50%10%Figure 2.3. Plots of dissolved inorganic carbon concentration (DIC) (A), and Mg concentration (B) in high pCO2 experiments versus time. Dashed lines are results from PHREEQC (Parkhurst and Appelo, 1999) models assuming kinetically controlled brucite dissolution and equilibrium CO2(g) dissolution. Open symbols and solid symbols represent duplicates 1 and 2, respectively. Arrows above the graphs represent the duration of the carbonation reaction in the 100%, 50% and 10% CO2 experiments, respectively.This suggests heat was lost to the surroundings. Temperature may provide a useful indicator for reaction progress and be utilized to monitor and map carbonation in tailings facilities that contain sufficient brucite (Pronost et al., 2012).2.3.2 Brucite dissolution and carbonation ratesThe rates of brucite dissolution and carbonation were accelerated linearly with pCO2, resulting in a ~2400-fold increase in carbonation rate with an increase from atmospheric pCO2 (~0.04%) to 100% CO2 (Fig. 2.4). Both dissolution and carbonation rates were calculated as averages over the entire reaction, making comparison with published dissolution rates difficult, as the instantaneous rate of reaction may change with time. Brucite dissolution was calculated as follows (Eq. 2.1): Rate = nibrc / (SAibrc × tc)                                         (Eq. 2.1)18Accelerated carbonation of bruciteWhere nibrc is the initial moles of brucite, SAibrc is the total initial brucite surface area (m2), and tc is the time for complete carbonation (s) as estimated with XRD, %C, and pH data (Fig. 2.2A, A1 Figs. A1.5-A1.6). Reaction end times of 75 h for 10% CO2 1 and 2, 14 and 12 h for 50% CO2 1 and 2, respectively, and 7 h were estimated for both 100% CO2 experiments. Average dissolution rates in duplicate reactors were 9.7 × 10-9, 5.6 × 10-8 and 1.0 × 10-7 mol m-2 s-1 with 10%, 50%, and 100% CO2, respectively. The brucite dissolution rates with atmospheric CO2 were too slow to be estimated. XRD results indicated brucite was replaced by nesquehonite [MgCO3·3H2O] (Eq. 2.2) in all high pCO2 reactors (A1 Fig. A1.6):Mg(OH)2(s) + HCO3-(aq) + H+(aq) + H2O(l) → MgCO3·3H2O(s)              (Eq. 2.2)Nesquehonite was first detected within 8 h in 10% CO2 1 and within 6 h in 10% CO2 2 using carbon abundance analysis. It was detected within 2 h in the 50% and 100% CO2 experiments. Minor amounts of phases that were not identifiable by XRD were formed in all high pCO2 reactors and the atmospheric CO2 experiment (A1 Fig. A1.6-A1.7). In the atmospheric CO2 experiment, carbon abundance in the solids increased by 2.6%, identifying this phase as a carbonate. The unidentified phase was replaced by dypingite [Mg5(CO3)4(OH)2·5H2O] as the reaction progressed in 10% 1 and 2. XRD and %C data indicated carbonation was completed within 102 h in the 10% CO2 reactors, by the 22 h and 12 h sampling points in 50% CO2 1 and Table 2.1. CO2 mass balance and brucite dissolution and carbonation rates.apCO2 (% of 1 atm) Dissolution rateb (mol brucite h-1) Carbonation rateb (mol CO2 h-1) CO2(g) supplied (g) CO2 in aqueous phase (g) CO2 in solid (g) Total CO2 sequestered (g) Efficiencyd (%) 0 . 0 4 nd c 1.1  10 -4  1768 8  10 -2 14 14 0.8%  10 0.0 3  0 . 0 3 223 6 91 97 43% 50 0.22  0.15 180 9 83 92 51% 100 0. 3 6  0 .27 211 16 76 92 44% aAll values are averages for duplicate reactors.  bDissolution and carbonation rates have been converted to mol h-1 for ease of comparison. cnd = no data  d Efficiency calculated as Mass CO2 sequestered in solid ÷ Mass CO 2(g) supplied during carbonation.  19Accelerated carbonation of brucite2, respectively, and between 6 and 10 h in both 100% CO2 reactors (A1 Fig. A1.6). Reaction end points between these times were inferred by the final decline in pH indicative of brucite removal. As observed by SEM, precipitates from the 10% and 50% CO2 experiments included elongate crystals consistent with the morphology of nesquehonite (Ferrini et al., 2009), along with flaky, poorly crystalline material and pseudo-rosettes (A1 Fig. A1.8). This morphology is typical of dypingite (Power et al., 2007; Power et al., 2009), and is likely representative of both dypingite and the unidentified carbonate phase in the high pCO2 experiments. Precipitates from the 100% CO2 experiments were dominated by nesquehonite crystals (A1 Fig. A1.8).  Other experimental studies have used elevated CO2 concentrations to precipitate nesquehonite by replacement of brucite or carbonation of MgCl2 in solution (Hänchen et al., 2008; Ferrini et al., 2009; Zhao et al., 2010; Back et al., 2011), which is consistent with this study. With a Mg:C ratio of 1:1, nesquehonite optimizes the amount of carbon stored per cation, an advantage for sequestration purposes. Dypingite and the unidentified carbonate phase may be reaction intermediates (Hopkinson et al., 2011) or products of nesquehonite dehydration (Ballirano et al., 2010) between sampling and analysis, as depicted in SEM images, wherein flaky material armors nesquehonite crystals (A1 Fig. A1.8).    Initial %C values of solids ranged from 0.76-1.10% owing to the presence of primary dolomite [CaMg(CO3)2]. Carbon abundance in solid samples increased with time in each high pCO2 experiment, followed by stabilization between 7.85% and 8.29% within 100, ~12, and 10 h in the 10%, 50%, and 100% CO2 experiments, respectively (A1 Fig. A1.5). This is consistent with formation of nesquehonite, which ideally contains 8.68% carbon. Carbon abundance increased to 3.6% after 2856 h in the atmospheric CO2 experiment. The mass of CO2 sequestered in the high pCO2 experiments was calculated using %C data assuming that dolomite mass remained constant and that CO2 was stored primarily within nesquehonite. A total of 317 g of nesquehonite (~101 g CO2) and 253 g (~81 g CO2) were formed in 10% CO2 1 and 2, respectively. In 50% CO2 1 and 2, 275 g of nesquehonite (~88 g CO2) and 244 g of 20Accelerated carbonation of brucitenesquehonite (~78 g CO2) were formed, respectively. A total mass of 223 g of nesquehonite (~71 g CO2) and 254 g of nesquehonite (~81 g CO2) was precipitated in 100% CO2 1 and 2, respectively. Discrepancies between duplicate reactors are attributed to imperfect mixing causing minor amounts of brucite to remain uncarbonated. The mass of carbon mineralized in the atmospheric CO2 experiment was calculated using %C data assuming that brucite mass remained relatively unchanged, which is justified by the relative peak heights in the XRD data. A total of ~14 g CO2 was captured over 2856 h. Carbonation rates in the high pCO2 experiments were calculated as follows (Eq. 2.3):CO2 sequestration rate (g CO2 g brucite-1 h-1) = [(mfnsq / Mnsq) × MCO2] / [mibrc × th]  (Eq. 2.3)Where mfnsq is the final mass of nesquehonite (g), mibrc is the initial brucite mass (g), th is Carbonation rate (g CO2 g brucite-1 h-1)pCO2 (% of 1 atm)this study[HCO 3-] dependent brucite dissolution (P. et al. 2005)pH dependent brucite dissolution (P. and S. 2004)0 10 20 30 40 50 60 70 80 90 1000.000.040.080.120.160.20Figure 2.4. Experimentally determined and modeled (PHREEQC; Parkhurst and Appelo, 1999) brucite carbonation rate versus pCO2 in experiments using ligand (Pokrovsky et al., 2005b) and pH (Pokrovsky and Schott, 2004) dependent brucite dissolution rate laws. Open symbols and solid symbols represent duplicates 1 and 2, respectively. 21Accelerated carbonation of brucitethe time of reaction completion (h), and Mnsq and MCO2 are the molar masses of nesquehonite and CO2 (g mol-1), respectively. In the atmospheric CO2 experiment, the carbonation rate was calculated based on the rate of increase of %C in the solids. The rates of CO2 sequestration using laboratory air, and average rates between duplicate 10%, 50%, and 100% CO2 experiments were ~3.30 × 10-5, 7.78 × 10-3, 4.26 × 10-2, and 8.00 × 10-2 g CO2 g brucite-1 h-1, respectively. Dissolution and carbonation rates and CO2 mass balances are summarized in Table 2.1. 2.3.3 Reaction mechanism Brucite dissolution is a surface controlled reaction (Vermilyea, 1969; Jordan and Rammensee, 1996; Pokrovsky and Schott, 2004) that is accelerated with increased acidity and concentration of certain organic and inorganic ligands, such as HCO3- (Vermilyea, 1969; Pokrovsky and Schott, 2004; Pokrovsky et al., 2005b). Proton and ligand promoted dissolution occur in parallel; their effect is additive (Pokrovsky et al., 2005b). Ligands that promote brucite dissolution are those that form protonated ions at neutral to weakly alkaline pH (e.g., HCO3-), whereas those that are deprotonated inhibit dissolution (e.g., CO32-; Vermilyea, 1969; Pokrovsky et al., 2005b). Therefore, aqueous carbonate species may enhance or inhibit brucite dissolution depending on the solution pH. Previous studies have documented a dependence of brucite and MgO dissolution rate on HCO3- concentration (Pokrovsky et al., 2005b; Back et al., 2011). As HCO3- is the dominant aqueous carbonate species in the experimental pH range (Langmuir, 1997), dissolution and hydration of CO2 may accelerate brucite dissolution via the direct effect of increasing HCO3- concentration and the indirect effect of increasing acidity. These effects cannot be separated in the high pCO2 experiments, as neither pH nor DIC were fixed. Increased DIC concentration should also promote brucite carbonation by facilitating carbonate precipitation at lower Mg concentration, as suggested by Chen et al. (2006) to account for enhanced carbonation of olivine in NaHCO3 solutions. To discern the effect of elevated pCO2 on brucite carbonation rate, experimental conditions were modeled using PHREEQC (Parkhurst and Appelo, 1999). Experimentally 22Accelerated carbonation of brucitedetermined, steady-state, far from equilibrium rate laws dependent on pH (Pokrovsky and Schott, 2004) and ligand (HCO3-) concentration (Pokrovsky et al., 2005b) were used, and are provided in Appendix 1. Nesquehonite precipitation and CO2 uptake were simulated as equilibrium processes. Modeling revealed that experimental dissolution and carbonation rates exceeded those predicted with proton promoted dissolution and high DIC promoted carbonate precipitation alone. With 100% CO2, the experimental carbonation rate is 6-fold faster than predicted due to acidity promoted dissolution (Fig. 2.4). Comparison of experimental dissolution rates with the pH-dependent rate law of Pokrovsky and Schott (2004) highlights the discrepancy between the experimental rates and those predicted due to proton promoted dissolution (A1 Fig. A1.9). In this study, dissolution rate is calculated based on the time to dissolve brucite of a specified mass, rather than an instantaneous rate based on the flux of Mg2+; this could in part explain the divergence from the Pokrovsky and Schott rate law (Pokrovsky and Schott, 2004) as the instantaneous rate was likely not constant. Experimental carbonation rates fell between those predicted by pH and HCO3- promoted dissolution (Fig. 2.4), implying that HCO3- promoted brucite dissolution is likely an important mechanism enhancing brucite carbonation, yet the CO2 supply was insufficient to achieve the maximum predicted rates.  2.3.4 Rate limitationThe fractionation of carbon stable isotopes can be used to infer the relative rates of processes that involve CO2 and thus help identify rate limitations (Wilson et al., 2010). All isotopic data are reported in δ-notation relative to Vienna Pee Dee Belemnite (VPDB) in units of per mil (‰). The δ13CCO2(g) values of the gas supply for the 10%, 50%, and 100% CO2experiments were -35‰, -37‰, and -32‰, respectively. Differences in initial δ13CCO2(g) values are due to the use of different CO2 tanks for each experiment. Equilibrium fractionation between CO2(g) and HCO3- is +7.9‰ at 25°C (Mook et al., 1974), therefore, the equilibrium δ13CDIC values for the high pCO2 experiments should be between ~-24‰, to -29‰. Divergence from the equilibrium δ13CDIC value increased in the first 36 h in 10% CO2 1 and 2, before rising to 23Accelerated carbonation of brucitereach equilibrium composition after ~175 h (Fig. 2.5A). Wilson et al. (2010) observed a similar negative trend in δ13CDIC values during carbonation of a MgCl2 solution using atmospheric CO2, which was attributed to a kinetic isotope fractionation effect due to slow uptake of CO2(g) into solution (Wilson et al., 2010). In the 50% and 100% CO2 experiments, equilibrium δ13CDIC composition was obtained after 56 and 43 h, respectively (Fig. 2.5A). Yet, mineralogical data indicated carbonation was complete within 75, 12, and 7 h in the 10%, 50% and 100% CO2 experiments, respectively. This implies that isotopic equilibrium between CO2 in the gas and aqueous phase was not achieved during the carbonation reaction, but is only approached when the DIC sink, carbonate mineral precipitation, is ceased. Due to the presence of dolomite in the initial material, early time δ13C values of solids are not representative of precipitated carbonate (~-4.4‰). A cut off of a minimum of 4% carbon in the solids (at least 80% of carbon in nesquehonite) was chosen to be representative of nesquehonite stable isotopic composition. Nevertheless, the presence of minor dolomite results in the solids being enriched relative to the precipitated nesquehonite. The δ13C values of the solids (primarily nesquehonite) remained between -28.3‰ to -30.8‰, and -29.6‰ to -32.6‰ in 10% CO2 1 and 2, respectively. A greater range in δ13C values was recorded in the 50% CO2 experiment, with values between -29.6‰ to -37.4‰, and -31.2‰ to -36.1‰ in 50% CO2 1 and 2, respectively. In 100% CO2  1 and 2, δ13C values were between -25.0‰ to -30.1‰, and -25.6‰ to -31.0‰, respectively. Fresh carbonate could not be sampled therefore δ13C values are cumulative for carbonate formed throughout each experiment. Insufficient carbonate precipitation occurred in the atmospheric CO2 experiment to determine the 13C composition of the carbonate.   The equilibrium δ13C fractionation factor between nesquehonite and HCO3- has not been determined. However, the observed fractionation is consistent with the equilibrium carbon isotopic fractionation factor estimated by Wilson et al. (2010) for dypingite and HCO3- (3.8 ± 1.3‰ at 25°C). The δ13C values of solids generally deviate between 0-5‰ from values predicted assuming isotopic equilibrium is maintained with the average δ13CDIC value during 24Accelerated carbonation of brucite0 25 50 75 100 125 150 175 200Time (h)  15.0  10.0  5.00.0-5.0-10.0-15.0δ13 C solids (‰ VPDB)δ13C solids (Obs.-Eq.)BEquilibrium with DIC during carbonation CO 2(g) CO 2(aq) HCO 3-   +  H + HCO 3- Mg 2++Carbonation duration  15.0  10.0  5.00.0-5.0-10.0-15.0δ13 CDIC (‰ VPDB)Equilibrium 0 25 50 75 100 125 150 175 200Time (h)50% CO 2100% CO 210% CO 2δ13CDIC  (Obs.-Eq.)ACarbonation duration100%50%10%100%50%10%Figure 2.5. Deviation from calculated equilibrium δ13CDIC values in high pCO2 experiments versus time using the fractionation factor of Mook et al. (1974) (A). Deviation from calculated equilibrium δ13C values of solids with the average δ13CDIC during carbonation using the fractionation factor of Wilson et al. (2010) in high pCO2 experiments versus time of sampling (B). Shaded area represents the range of equilibrium δ13C values of solids with the observed range in δ13CDIC values during carbonation. Open and solid symbols represent duplicates 1 and 2, respectively. Arrows above the graphs represent the duration of the carbonation reaction in the 100%, 50% and 10% CO2 experiments, respectively.carbonation (Fig. 2.5B). The majority of δ13C values of solids fall within the range of equilibrium composition estimated with the observed range of δ13CDIC values during carbonation (shaded area in Fig. 2.5B). This suggests that nesquehonite precipitated approximately at isotopic equilibrium with DIC throughout carbonation. Moreover, nesquehonite saturation indices calculated using PHREEQC (Parkhurst and Appelo, 1999) indicate nesquehonite precipitated near chemical equilibrium (A1 Fig. A1.10). This implies carbonate mineral precipitation is not rate limiting for carbonation, rather the slow approach to isotopic equilibrium between CO2(g) and DIC indicates uptake of CO2 is the limiting step for brucite carbonation. During carbonation, DIC concentrations in experiments were consistently below CO2(g)↔DIC equilibrium levels as predicted by PHREEQC (Parkhurst and Appelo, 1999) modeling (Fig. 2.3A). This further indicates equilibrium between CO2 in the gas and aqueous 25Accelerated carbonation of brucitephase was not obtained during carbonation. Experimental carbonation rates were generally lower than predicted by modeling due to HCO3- promoted dissolution, with divergence increasing at lower pCO2 (Fig. 2.4). This is consistent with a kinetic limitation to uptake of CO2 into solution, as equilibrium DIC concentrations were not achieved in the experiments, and divergence from CO2(g)↔DIC equilibrium was greatest at lower pCO2. This is in agreement with the stable isotopic data, indicating that uptake of CO2 into solution is rate limiting for brucite carbonation even at elevated pCO2. The implication is that if CO2(g)↔DIC equilibrium could be attained, the brucite carbonation rate would be further accelerated. CO2 uptake into solution is divided into two steps: (1) CO2(g) dissolution and (2) CO2(aq) hydration. CO2(g) dissolution involves the phase transfer from gaseous to aqueous CO2, whereas CO2(aq) hydration is the subsequent formation of H2CO3 or HCO3- depending on solution pH, which dissociate to form CO32- (Stumm and Morgan, 1996). It is well understood that dissociation is rapid, whereas the hydration of CO2(aq) is relatively slow (Sullivan et al., 1993; Stumm and Morgan, 1996; Lasaga, 1998). An option for further enhancing CO2 uptake into solution is to accelerate CO2 hydration, which may be achieved through use of biological catalysts such as the carbonic anhydrase enzyme (Mirjafari et al., 2007; Favre et al., 2009; Sharma and Bhattacharya, 2010). 2.3.5 Implications for carbon sequestration in mine tailingsExperimental data show a linear increase in brucite carbonation rate with pCO2, resulting in a ~2400-fold acceleration with an increase from atmospheric pCO2 (~0.04% CO2) to 100% CO2. Conversely, modeling using PHREEQC (Parkhurst and Appelo, 1999) reveals a non-linear relationship between carbonation rate and pCO2 with both proton and HCO3- promoted dissolution (Fig. 2.4). This implies that use of 100% CO2 may not be necessary to achieve maximum carbonation rates if CO2 uptake into solution were enhanced (Fig. 2.4). Direct use of flue gas (e.g., ~17% CO2; Kikkinides et al., 1993) from mine site power plants could be sufficient to achieve similar carbonation rates as use of purified CO2. This is highly advantageous as it eliminates the cost of purifying CO2. 26Accelerated carbonation of bruciteDespite a kinetic limitation for uptake of CO2 into solution, brucite carbonation rates achieved in experiments exceed that required to carbonate all the brucite produced annually at MKM. MKM produces ~0.11-0.28 Mt of brucite in tailings annually, and emits 370 kt yr-1 of CO2 equivalent GHGs (BHP Billiton, 2005; Wilson, 2009). Carbonation of brucite produced annually would offset mine emissions by ~22-57% (Fig. 2.6), which is up to 4-fold faster than estimated passive carbonation rates (Wilson, 2009). A tax on carbon emissions implemented in July 2012 in Australia implies tailings carbonation may offer a significant financial benefit for Australian mines such as MKM (e.g., up to $4.8 million saved annually; Australian Government, 2011). Complete carbonation of brucite accumulated in tailings stockpiles at MKM would sequester a total of 1-3 Mt of CO2. This is a comparable sequestration rate to what is currently achieved at the largest geologic carbon sequestration demonstration projects (~1-2.8 Mt CO2 yr-1; Michael et al., 2009; Whittaker et al., 2011).  Brucite is present in ultramafic tailings from other deposit types, such as chrysotile. In Québec, Canada, there are approximately 2 Gt of chrysotile mining residues (Larachi et al., 2010). At the Black Lake mine in Québec, chrysotile milling residues contain ~1.8 wt.% brucite (Pronost et al., 2011). If these brucite contents are representative for the entire stockpile, up to 27 Mt of CO2 could be sequestered via brucite carbonation alone. Brucite carbonation represents only a small proportion of the total sequestration potential of ultramafic tailings, which consist primarily of Mg-silicate minerals. At MKM, accelerated carbonation of Mg-silicate minerals has the potential to sequester much more CO2 than is emitted annually (Fig. 2.6). Accelerated dissolution of silicate tailings minerals such as serpentine [Mg3Si2O5(OH)4] and forsterite [Mg2SiO4] is predicted with increased pCO2 owing to proton promoted dissolution (Bales and Morgan, 1985; Pokrovsky and Schott, 2000). The effect of inorganic carbon ligands on silicate mineral dissolution is less certain (c.f., forsterite, wollastonite, and anorthite; Berg and Banwart, 2000; Golubev et al., 2005), and the effect on serpentine dissolution is not well understood under conditions prevalent in mine tailings. The presence of ligands such as citrate have been found to enhance serpentine dissolution by several 27Accelerated carbonation of bruciteorders of magnitude over that predicted by pH promoted dissolution, and can be significant in pH ranges appropriate for carbonate precipitation (Krevor and Lackner, 2011). If serpentine dissolution were similarly affected as brucite by surface reactions involving the HCO3- ligand, total carbonation rates in tailings may be underestimated based on pH effects alone. passive carbonationmine emissionscarbonation capacity10-4 10-3 10-2 10-1 1Tonnes CO   Tonne Tailings -1 yr -12felsic ultramaficon-grid off-grid small mineactive brucite carbonationClinton Creek MKMDiaviklarge minebrucite contentMKMFigure 2.6. Comparison of range of passive annual carbonation rates at Diavik Diamond Mine, Northwest Territories, Canada (Wilson et al., 2011), Clinton Creek Chrysotile mine, Yukon Territory, Canada (Wilson et al., 2009), and the Mount Keith Nickel Mine (MKM), Western Australia (Wilson, 2009) with annual GHG emissions at various mine sites, total sequestration capacity, and potential annual carbonation rates at MKM via accelerated brucite carbonation.28Strategies for enhancing carbon sequestration3. Strategies for enhancing carbon sequestration in Mg-rich mine tailings23.1 IntroductionAnthropogenic greenhouse gas (GHG) emissions, particularly CO2 emissions, have been identified as a cause of global climate change (IPCC, 2007). Carbon sequestration is one of many potential strategies to stabilize CO2 concentrations and prevent irreversible climate change while we transition to non-fossil fuel based energy sources (Pacala and Socolow, 2004; Broecker, 2007). Mineral carbonation, or carbon mineralization, is a method of carbon sequestration that involves dissolution of non-carbonate minerals (e.g., silicates, hydroxides, and oxides) to release cations, and the binding of these cations with CO2 in carbonate minerals (Seifritz, 1990; Lackner et al., 1995; Lackner et al., 1997; Lackner, 2003). Many industrial wastes, including mine tailings, are rich in minerals that provide suitable feedstock for mineral carbonation such as brucite [Mg(OH)2], forsterite [Mg2SiO4], and serpentine group minerals [Mg3Si2O5(OH)4] (e.g., Wilson et al., 2009a; Renforth et al., 2011; Bobicki et al., 2012). Carbonation of industrial wastes is advantageous as it exploits available waste materials that are typically fairly reactive under ambient conditions, and it may decrease the hazardous nature of wastes such as asbestos mine tailings (Renforth et al., 2011; Bobicki et al., 2012). Carbon mineralization in mine wastes has been documented to occur passively under normal mining practices at both historic and active asbestos, diamond, chromite, and nickel mines globally (Wilson et al., 2006; Wilson et al., 2009a; Wilson et al., 2010; Wilson et al., 2011; Bea et al., 2012; Pronost et al., 2012; Beinlich and Austrheim, 2012). Carbonation reactions are facilitated by high surface areas (Wilson et al., 2009a), yet are limited by the uptake of CO2 2A version of this chapter was published in the proceedings of the International Mine Water Association 2013 Mine Water Conference as: Harrison, A. L., Power, I. M. and Dipple, G. M. (2013) Strategies for enhancing carbon sequestration in Mg-rich mine tailings, in Brown, A., Figueroa, L., Wolkersdorfer, C. (Eds.), Reliable Mine Water Technology (Vol. 1). Publication Printers, Denver, Colorado, USA, pp. 593–598. It is reproduced here with permission from all authors.  29Strategies for enhancing carbon sequestrationinto solution (Wilson et al., 2010). Although the carbon sequestration capacity of ultramafic tailings is significant, rates of passive carbonation are insufficient to take full advantage of the carbon sequestration potential. For instance, complete carbonation of tailings produced annually at the Mount Keith Nickel Mine (MKM) in Australia (~11 Mt tailings per year; BHP Billiton, 2005) would exceed annual mine emissions by more than a factor of ten. Yet passive carbonation rates currently offset annual emissions by only ~15% (Wilson, 2009). Carbon mineralization in mine tailings could be accelerated by increasing the exposure of tailings to CO2, such as by injection of CO2-rich gas streams into tailings storage facilities. Here, we generalize the experimental results from our previous work (Harrison et al., 2013a; Chapter 2) investigating the potential for accelerated carbonation of brucite, a common and highly reactive tailings mineral, to evaluate CO2 injection as a carbon sequestration strategy in mine tailings. Brucite is a Mg-hydroxide mineral that is typically present between 1 and 15 wt.% in ultramafic mine tailings and residues (Chrysochoou et al., 2009; Pronost et al., 2011; Bea et al., 2012). It is far more reactive than the more abundant silicate phases such as serpentine (e.g., Bales and Morgan, 1985; Assima et al., 2013a), and therefore provides a useful starting point for investigation of accelerated tailings carbonation strategies. 3.2 MethodsThe effects of supplying CO2-rich gas streams at ambient temperature and pressure (~21°C; 1 atm) on the carbonation rate of brucite was investigated experimentally in batch reactors, with geochemical conditions emulating those at MKM. Alkaline 3.0 L slurries containing 150 g brucite were supplied with gas streams with a range of CO2 content (~0.04%, 10%, 50%, 100% CO2 by volume) at a rate of ~270 mL min-1 (Harrison et al., 2013a; Chapter 2). Slurry samples were extracted periodically for measurement of pH, the stable carbon isotopic composition (δ13C) of dissolved inorganic carbon (DIC), and cation and DIC concentrations. Solid samples were collected for measurement of mineralogical compositions and δ13C. For 30Strategies for enhancing carbon sequestrationfurther details regarding the experimental setup, refer to Harrison et al. (2013a; Chapter 2).   3.3 Results and discussionThe experimental results indicated that brucite (brc) was carbonated to produce the hydrated Mg-carbonate mineral nesquehonite [MgCO3·3H2O] at a rate that increased linearly with CO2 partial pressure (pCO2) according to the following reaction (Harrison et al., 2013a; Chapter 2):Mg(OH)2(s) + HCO3-(aq) + H+(aq) + H2O(l) ↔ MgCO3·3H2O(s)              (Eq. 3.1)A 2400-fold increase in CO2 sequestration rate was achieved with an increase from atmospheric composition (~0.04% CO2) to pure CO2. Increasing the pCO2 serves to enhance both brucite dissolution and carbonate precipitation. However, even at elevated pCO2 the rate of CO2 uptake into solution was found to be rate limiting. This is attributed to the relatively slow transformation from gaseous CO2 to an aqueous form that can be mineralized (i.e., HCO3- and CO32-) as indicated by chemical and isotopic disequilibrium between gaseous CO2 and DIC during the carbonation reaction (Harrison et al., 2013a; Chapter 2). Nevertheless, the experimental carbonation rates would be sufficient to carbonate all the brucite produced annually at MKM (0.1-0.3 Mt brucite), offsetting total mine emissions by 20-60% (Harrison et al., 2013a; Chapter 2). Therefore accelerated carbonation of tailings minerals by supplying CO2-rich gas streams is a promising method for reducing GHG emissions at mine sites with tailings that contain brucite. Power generation at mine sites often occurs at on-site power plants, which could provide a local point source of CO2 emissions. Flue gas from power plants typically contains between 10-20% CO2 (Kikkinides et al., 1993; Uibu et al., 2011). These emissions provide a readily available source of CO2-rich gas that could be injected into tailings storage facilities to accelerate brucite carbonation (Harrison et al., 2013a; Chapter 2). An alternative 31Strategies for enhancing carbon sequestrationwould be to circulate CO2-rich water, which could help to avoid the issue of slow CO2 uptake into solution. However, the use of CO2-rich gas has the advantage that it would not increase the water consumption at mine sites, which can be an important concern particularly for mines in arid locations with high evaporative losses. A potential concern regarding CO2 injection is its effect on the mobility of hazardous metals. Metals could be mobilized via dissolution of primary tailings minerals, yet secondary precipitates can incorporate these metals and limit their mobility depending on solution pH (Power et al., 2010). Therefore further investigation is warranted to determine the potential side effects of CO2 injection on tailings geochemistry. If brucite carbonation via CO2-rich gas injection were implemented at a mine site, tailings storage facilities would need to be designed to minimize leakage of injected CO2 to the atmosphere. This requires that the rate of carbon mineralization in the tailings keep pace with the rate of CO2 supply. The rate of CO2 injection will be limited by the availability of the highly reactive phases such as brucite, as these will consume the majority of the CO2 in the short term. Greater brucite content will accommodate higher injection rates, as well as provide greater total sequestration capacity. The sequestration or ‘reactive capacity’ (Eq. 3.2) of brucite is calculated by assuming complete conversion to nesquehonite; this equates to 0.75 g CO2 stored per gram brucite. In order to prevent CO2 leakage, the rate of CO2 supply must not exceed the reactive capacity at any given time, and should therefore balance the rate of brucite deposition according to the reaction stoichiometry. The rates achieved in the brucite carbonation experiments employing 10% CO2 were considered representative of reaction in mine tailings. The predicted brucite carbonation rate using flue gas is thus ~0.3 mol CO2 m-2 brc yr-1 (after Harrison et al., 2013a; Chapter 2). Due to the relatively high reactivity of brucite, it is assumed that the rate of CO2 mineralization is dictated primarily by the brucite carbonation rate rather than carbonation of the less reactive silicates. The reactive capacity provided by brucite and the time before CO2 venting at a given CO2 injection rate can then be calculated according to Equations 3.2 and 3.3. Carbonation rates in the field will be affected by water content distribution (e.g., Assima et al., 2013a), surface 32Strategies for enhancing carbon sequestrationpassivation effects (e.g., Jeen et al., 2006), and the hydraulic properties of the porous medium. Quantification of the extent to which these effects will alter carbonation rates requires further experimentation. As such, for the purposes of this study, it is assumed that CO2 supply is the primary rate-limiting factor; therefore Equation 3.3 applies only when the maximum rate of carbonation (Eq. 3.4) is less than the rate of CO2 supply. Reactive capacity  Cr = m t F b r cGFW b r cb                                   (Eq.3.2) Time to CO 2 ventingCrr CO 2=                                        (Eq.3.3)Maximum brucite carbonation rate m t F b r c S b r c r b r c=                (Eq.3.4)Where, Cr is the reactive capacity (mol CO2), mt is the mass of tailings (g), Fbrc is the brucite content of the tailings as a fraction of tailings mass, b is a stoichiometric coefficient for the carbonation of brucite, GFWbrc is the molar mass of brucite (g mol-1), rCO2 is the rate of CO2 supply (mol CO2 yr-1), Sbrc is the surface area of brucite (m2 g-1), and rbrc is the ‘flue gas’ rate of brucite carbonation measured in batch reactors (mol CO2 m-2 brc yr-1). At MKM, approximately 0.5 m of tailings are deposited annually, assuming they are evenly distributed across the ~16.6 km2 tailings storage facility, with historic tailings reaching depths of up to ~19 m (Wilson, 2009). The total annual CO2 equivalent emissions at MKM are reported to be 0.37 Mt CO2 (BHP Billiton, 2005). GHG emissions from mining operations are typically divided between fossil fuel combustion from distributed sources like trucks and mining equipment, and emissions from electricity generation. It is estimated that ~64% of emissions are from electricity generation, and ~36% are from distributed sources (USEPA, 2008). This suggests that at MKM, approximately 0.24 Mt CO2 yr-1 is produced from point sources. If CO2 were injected at a rate equal to the rate of point source CO2 emissions (~328 mol CO2 yr-1 m-2 tailings; after BHP Billiton, 2005), current brucite production rates are nearly 33Strategies for enhancing carbon sequestrationM oun t K e ith  mi n e0 4 8 12 16 200.11101001000Time to CO2 venting(years)Historic tailings depth (m)B1.0 wt.%2.5 wt.%5.0 wt.%10 wt.%15 wt.%M ou n t Ke ith mi n e1.0 wt.%2.5 wt.%5.0 wt.%10 wt.%15 wt.%0 0.5 1.0 1.5 2.0 2.5 3.0 3.50.1110100Time to CO2 venting(years)Tailings deposition rate (m yr -1 )AFigure 3.1. Time to CO2 venting versus tailings deposition rate (A) and historic tailings depth (B) at various brucite contents with a CO2 injection rate equal to the estimated rate of point source CO2 emissions at the Mount Keith Nickel Mine (~328 mol CO2 yr-1 m-2 tailings; after BHP Billiton, 2005). Shaded areas indicate range of brucite content and tailings depth and deposition rate applicable to the Mount Keith Nickel Mine.sufficient to offset point source CO2 emissions (Fig. 3.1A). Higher injection rates could be initially applied to carbonate historic tailings. For instance, it would take up to 20 years for CO2 to vent from the deepest tailings at MKM at this CO2 injection rate (Fig. 3.1B). For highly reactive minerals like brucite, the reactive capacity rather than the mineral carbonation rate will likely limit the maximum CO2 injection rate. At mines with high tailings production rates that produce power on-site, such as MKM, it is estimated that a minimum brucite content in tailings of ~3 wt.% would be required to offset point source mine emissions, or ~64% of total emissions, if brucite is the primary sink for CO2 (Figs. 3.1A and 3.2). Smaller ‘off-grid’ mines produce greater CO2 emissions per tonne of tailings produced, and would require in excess of 6.5 wt.% brucite to offset estimated point source emissions (Fig. 3.2). Regardless of electricity source or mine size, accelerated carbonation of even minor amounts of brucite (e.g., 2.5 wt.%) could provide >4-fold acceleration over passive carbonation rates (Fig. 3.2). In the case of a carbon tax or cap-and-trade system, offsetting GHG emissions would be financially beneficial for mining companies. Economically marginal 34Strategies for enhancing carbon sequestrationdeposits that possess significant carbon capture potential, such as high brucite content, may therefore become financially viable (e.g., Bobicki et al., 2012). Although accelerated brucite carbonation would provide significant offsets of mine emissions, in order to take full advantage of the sequestration capacity offered by mine tailings, Mg-silicate carbonation must also be enhanced (Fig. 3.2). Passive carbonation of serpentine has been documented in tailings with and without brucite (Wilson et al., 2006; Wilson, 2009; Wilson et al., 2009a). This demonstrates that despite being less reactive than brucite, Mg-silicate carbonation is possible in a tailings environment. Injection of CO2-rich gas into tailings 10-4 10-3 10-2 10-1 1Tonnes CO   Tonne Tailings -1 Yr -12Clinton CkDiavik1 wt.% 5 wt.% 9 wt.% 15 wt.%small mineon-gridlarge mineMKMmine emissionsoff-gridbrucite capacityMKM Diavikpassive carbonationMg-silicate mineral reactive capacityFigure 3.2. Comparison of range of passive annual carbonation rates at Diavik Diamond Mine, Northwest Territories, Canada, Clinton Creek Chrysotile mine, Yukon, Canada, and the Mount Keith Nickel Mine (MKM), Western Australia (circles) with annual GHG emissions at mine sites of various size and power sources (stars). Orange and black stars represent estimated point source and total CO2-equivalent emissions, respectively. Total reactive capacity including Mg-silicate carbonation is indicated by the green vertical dashed-dot line. The reactive capacity per tonne tailings based on brucite carbonation alone at various brucite contents is indicated by the purple vertical dashed lines at 2 wt.% brucite intervals (modified after Harrison et al., 2013; Chapter 2).35Strategies for enhancing carbon sequestrationwill not only exploit the sequestration potential of brucite, but may also help carbonate the more abundant Mg-silicates. Due to the lower reactivity of Mg-silicates in comparison to brucite, more aggressive methods are often employed in mineral carbonation processes to accelerate dissolution, such as the use of strong acids (e.g., Alexander et al., 2007). Hence, it is unlikely that injection of CO2-rich gas alone will allow sufficient acceleration of Mg-silicate carbonation to take full advantage of the CO2 sequestration potential. As research to enhance Mg-silicate carbonation continues, deployment of accelerated brucite carbonation strategies in the interim will help guide the development of Mg-silicate carbonation techniques for mine tailings, while providing an immediate and significant GHG benefit. 3.4 ConclusionsExperimental results indicate that passive rates of carbonation could theoretically be accelerated to offset ~20-60% of total CO2 emissions at the Mount Keith Nickel Mine in Australia. This could be achieved via injection of CO2-rich gas into tailings, such as flue gas from mine site power plants, to completely carbonate brucite and partially carbonate Mg-silicates. Mines that contain >6.5 wt.% brucite in their tailings may have the potential to completely offset point source CO2 emissions via brucite carbonation alone. Carbonation of the more abundant Mg-silicates could capture >10-fold the scale of mine emissions at a large mine such as MKM. Deployment of accelerated brucite carbonation strategies would offer a first step towards the development of methods that take advantage of the total CO2 sequestration capacity of mine tailings, while providing an immediate GHG benefit.Enhanced mineral reactivity in partially wetted porous media364. Enhanced	mineral	reactivity	driven	by	pore	fluid	mobility in partially wetted porous media34.1 IntroductionMineral-fluid reactions in Earth’s shallow subsurface regulate element transport in nature, disturbed environments, and man-made materials such as industrial wastes. They are vital for nutrient availability, water quality, and carbon cycling (Blowes and Jambor, 1990; McKinley et al., 2006; Manning, 2008; Maher et al., 2009; Brantley and White, 2009). Despite the importance of mineral-fluid reactions in controlling element cycling, predictions of large-scale mass transport are hindered by the scale-dependence of mineral dissolution-precipitation rates and the use of the continuum approach in numerical models that represents porous media with continuous domains of volume-averaged properties (Molins et al., 2012). This approach can lead to significant error in the prediction of geochemical reaction rates (Li et al., 2006; Li et al., 2008; Molins et al., 2012). Global patterns in rainfall and soil moisture content will be altered in response to climate change (IPCC, 2013), therefore the impact of wetting and drying on mineral-fluid reactions is likely to be of increased significance and concern. In particular, partially wetted porous media exhibit highly heterogeneous pore microenvironments that are subject to significant change during gas and fluid mobility. Others have posited that particle entrainment by the mobile fluid meniscus during evaporation or water recharge contributes to hardening of soils, and have shown indirect textural evidence of this process (Mullins et al., 1987; Bresson and Moran, 1995), but the more important chemical impacts of particle mobility have not been previously appreciated. We anticipate that pore scale reaction rates will be highly dynamic and variable in these environments. Experiments using microfluidic reactors (i.e., micromodels) allow visualization of pore 3A version of this chapter will be submitted for publication as:Harrison, A.L., Dipple, G.M., Song, W., Power, I.M., Mayer, K.U., Beinlich, A. and Sinton, D. Enhanced mineral reactivity driven by pore fluid mobility in partially wetted porous media.Enhanced mineral reactivity in partially wetted porous media37scale processes that can inform development of pore scale reactive transport models (Yoon et al., 2012; Boyd et al., 2014). To date, most of these experiments have focused on flow dynamics; only recently have mineral precipitation reactions been studied in water filled micromodels (Zhang et al., 2010; Yoon et al., 2012; Boyd et al., 2014) and few have investigated precipitation in mixed fluid-gas conditions representative of the shallow subsurface (Kim et al., 2013). Here, we consider the reactivity of magnesium hydroxide in the form of the mineral brucite, of interest because of its fast reaction rate at ambient laboratory conditions and because it actively reacts with the atmosphere in mine wastes to immobilize carbon (Harrison et al., 2013a; Chapter 2). We investigated coupled brucite dissolution-carbonate mineral precipitation in a variably water-filled silica glass micromodel under evaporative conditions to elucidate controls on mineral-fluid reactions and highlight the importance of pore scale processes on reactivity at conditions representative of the shallow subsurface. 4.2 MethodsAs an example of mineral dissolution-precipitation reactions in the shallow subsurface, duplicate microfluidic brucite carbonation experiments were conducted using an unsaturated silica glass microfluidic chip. The microfluidic chip (micromodel) had a pore network that was 5.0 cm long by 1.5 cm wide with cylindrical glass pillars representative of quartz [SiO2] grains (Appendix 2 (A2) Fig. A2.1). The network consisted of regularly spaced pillars of two sizes: ~180 μm and ~70 μm. Pore throats were between ~215 and 260 μm, the depth was ~130 μm, and the porosity was approximately 0.87. The micromodel was fabricated using standard photolithography techniques and wet etching with hydrofluoric acid (Kim et al., 2013). Prior to experimentation and between experiments, the micromodel was thoroughly flushed with toluene, isopropyl alcohol, 12 M hydrochloric acid, deionized water, and then the experimental slurry. Experimental slurries consisted of 0.5 M MgCl2 prepared using deionized water and reagent grade MgCl2·6H2O, and 5 wt.% pulverized brucite. The brucite was sieved to <53 Enhanced mineral reactivity in partially wetted porous media38μm in diameter with a mean particle diameter of 23 μm, and surface area of 5.2 m2 g-1. The brucite was ~79 wt.% pure, with the remainder consisting primarily of dolomite [CaMg(CO3)2] and magnesite [MgCO3]. Slurries were sonicated to minimize particle aggregation prior to injection into the micromodel with a syringe. Refer to Appendix 2 for details regarding the analytical methods used to characterize the initial brucite.  After the slurry was injected, multiple pore volumes of research grade gaseous CO2 (99.99% purity; Praxair) were flushed through the chip to displace the slurry, leaving the pore network with variable water content (i.e., not all pores were filled with water). The micromodel was then immediately connected to a stream of CO2 supplied at a constant flow rate of 0.5 mL min-1 using a Teledyne ISCO model D-260 pump. A high CO2 supply rate compared to the micromodel pore volume (~85 μL) was utilized to ensure that the carbonation reaction was not limited by a lack of CO2. The first experiment was conducted for 18 h, and the second for 8 h.  Time lapse greyscale brightfield images were taken at 5 minute intervals during the experiment using an inverted Olympus CKX41 transmitted light microscope and an Orca-ER (1344 × 1024 pixels) Hamamatsu camera. These images comprised a field of view of 3.5 mm wide × 2.6 mm long. Additional images were collected in color to characterize the entire pore network following the second experiment using a Nikon eclipse E600 POL microscope and Canon EOS Rebel T2i camera at various fields of view. Raman spectra of the reaction products were collected following the second experiment using a Renishaw inVia microscope with 785 nm excitation. Mineral identifications were conducted with reference to the RRUFFTM database of Raman spectra. 4.3 Results and discussionInjection of pure CO2 gas into micromodels induced evaporation of water, dissolution of brucite, and precipitation of prismatic nesquehonite [MgCO3·3H2O] crystals and other Mg-carbonate minerals (Fig. 4.1A; A2 Figs. A2.2 and A2.3). Brucite particles dissolved Enhanced mineral reactivity in partially wetted porous media39EpoxyQtzNsqeBrcNsqBrc-richGasBrc-poorc dInert MWIGWIMWIGMIfBrcGasNsqaMineralMineralGas Waterb G M IG MIM WIM WIGWIFigure 4.1. Conceptual schematic and transmitted light and backscattered electron micrographs of brucite carbonation (see following page for caption).Enhanced mineral reactivity in partially wetted porous media40(Accompanying Materials (AM) Video AM.1) and drove pore fluid to chemical saturation with carbonate minerals. This is consistent with the replacement of brucite by carbonate minerals as has been documented in bench scale experiments and inferred from weathering of industrial wastes under similar geochemical conditions (Harrison et al., 2013a; Chapter 2). In the micromodel, and by inference in the shallow subsurface of Earth, interfaces between gas, liquid, and solid are mobile and exert a first order control on reactivity and hence element transport. We differentiate between the gas-water interfacial area (GWI), gas-mineral interfacial area (GMI), and mineral-water interfacial area (MWI) in Figure 4.1B.  Light microscopy revealed that evaporative water loss drove changes to the GWI and GMI, and that the mobility of solid particles by moving water menisci resulted in order-of-magnitude changes to the local volume-specific MWI (AM Videos AM.1-2).Many mineral reactions proceed via dissolution and precipitation mechanisms in which aqueous solutions play a critical role. This is evident in our experiments wherein brucite in Figure 4.1 (continued). Conceptual schematic and transmitted light and backscattered electron micrographs of brucite carbonation. Gas-filled areas are false colored in green. A) Prismatic nesquehonite. B) Conceptual model of key interfaces. The mineral-water interfacial area (MWI) is defined as the reactive mineral surface area per unit volume of water (red dashed line), the gas-mineral interfacial area (GMI) is the mineral surface area in effectively dry pores (green dashed line) and the gas-water interfacial area (GWI) is the surface area of the gas-water interface (blue dashed line). C) Nesquehonite precipitation in a brucite-rich, water filled zone (left-hand side) compared to lack of carbonation in a brucite-poor, water filled zone (right-hand side) and gas-filled, dry area (middle). D) Evidence of brucite entrainment by water menisci during evaporation (indicated by black arrows). The former GWI is indicated by the white dashed line and black arrows, whereas the present GWI is indicated by a red arrow and the blue dashed line. E) Backscattered electron micrograph of reaction products from a bench scale brucite carbonation experiment (Harrison et al., 2015; Chapter 5) exhibiting textures consistent with particle entrainment by the mobile fluid meniscus. The epoxy-mineral interface, indicated by the blue arrow, is an approximate representation of the location of the GWI at the end of the experiment. F) Schematic representation of the key interfaces in (E). Dashed lines are color coded according to the scheme in (B), except the black dashed line, which represents the effectively inert quartz [SiO2]-water interface. All scale bars are 200 μm.Enhanced mineral reactivity in partially wetted porous media41dry pores remained visibly unreacted, whereas hydrated carbonate precipitates were abundant in water filled pores (Fig. 4.1C; AM Video AM.1), demonstrating that the GMI is effectively unreactive. The experiments revealed that in nearly all occurrences of nesquehonite, the size, lateral extent, and morphology of crystals was limited by the location of the GWI (Fig. 4.2). Only in large, connected, water-filled zones was nesquehonite precipitation not interface-limited (Fig. 4.2D). The micromodel experiments also revealed a propensity for formation of carbonate reaction products in areas of higher brucite abundance, i.e., “mineral buffered” zones (Fig. 4.1C; AM Video AM.1). Time-lapse images clearly demonstrate the absence of carbonation in a brucite-poor water-filled zone in contrast to abundant nesquehonite precipitation in an adjacent, comparatively brucite-rich water-filled zone (Fig. 4.1C; AM Video GasGasGasNsqNsqNsqNsqGasGasNsqa bc dNsqFigure 4.2. Transmitted light micrographs of nesquehonite precipitates. Gas-filled areas are false colored in green. A-C) Nesquehonite precipitates in water-filled areas, with lateral growth limited by the gas-water interface. D) Abundant nesquehonite precipitates in a large, connected, water-filled zone. Inset showing a close-up of a nesquehonite crystal in this zone. All scale bars are 200 μm.Enhanced mineral reactivity in partially wetted porous media42AM.1). This is attributed to the higher MWI, which allows rapid dissolution due the higher mineral surface area, contributes a higher density of nucleation sites, and provides greater pH buffering capacity that promotes both increased CO2 uptake into solution and carbonate precipitation. Geochemical models predict that at low MWI, even complete dissolution of the available brucite may not be sufficient to stabilize carbonate precipitates, a likely explanation for the lack of carbonation in the brucite-poor, “fluid buffered” zone (A2 Fig. A2.4).Particle entrainment by moving water menisci during evaporation may enhance coupled mineral dissolution-precipitation reactions by increasing the MWI by a factor of 100 or more (Fig. 4.3). During evaporation, brucite particles were entrained behind the retreating fluid menisci, rather than being left behind in dry pores where they would be effectively inert (Fig. 4.1D, AM Video AM.1), similar to the coffee-ring effect in unconstrained droplet evaporation (Yunker et al., 2011). Particle movement is attributed to capillary and surface tension forces and advective water movement within water-filled pores (Maenosono et al., 1999; Xu et al., 2008). Particle entrainment by mobile fluid menisci has been posited as an explanation for changes in physical structure and strength of agricultural soils upon wetting/drying, and is Before evaporationGas-water interfaceAfter evaporationGasLiquidLiquida bFigure 4.3. Entrainment of brucite particles by the retreating water meniscus during evaporation. Initial brucite slurry (5 wt.%) (A) and brucite slurry following 10 minutes of evaporation (B) at room temperature between two glass slides, as observed using transmitted light microscopy. Circles indicate identical brucite particles before and after evaporation for spatial reference. Red circles show particles that remained stationary and the blue circle shows a particle that has been moved and rotated. Scale bars are 200 μm.Enhanced mineral reactivity in partially wetted porous media43consistent with textures in soils (Fig. 4.1E) (Kemper and Rosenau, 1984; Mullins et al., 1987; Bresson and Moran, 1995) and bench-scale experiments (Fig. 4.1F) (Harrison et al., 2015; Chapter 5). Soils in arid and semi-arid environments are particularly vulnerable, and tend to from hard surface crusts that may decrease infiltration of water, enhance erosion, and inhibit root growth (Mullins et al., 1987; Bresson and Moran, 1995). Although substantial research has been devoted to understanding the physical dynamics of drying in porous media, the more complex impacts, such as the advective transport of solutes and particle migration remain open questions (Prat, 2011). To our knowledge, the relevance of these processes to larger scale mineral weathering in the natural environment, and their potential impact on local fluid chemistry and reaction rate has not previously been recognized.Recent advances in pore scale numerical modeling reveal that heterogeneity in physical pore structure can alter mineral dissolution rates (Molins et al., 2012). Physically, both the bulk pore structure and the local GWI are changed. Porosity is augmented in areas that dry out, increasing the relative permeability for gas flow, while the collection of fine particles in the water-filled zones could decrease permeability to water upon rewetting. Water recharge may remobilize particles in some circumstances (Lazouskaya et al., 2013), although the formation of mineral cement between grains would limit their remobilization (Harrison et al., 2015; Chapter 5) (Fig. 4.1E). In addition, a significant increase in solid material at the gas-water interface, as exhibited during evaporation of a brucite slurry between two glass slides (Fig. 4.3), could alter the rate of gas uptake into solution by providing a CO2 sink near the interface.Particle entrainment also results in a dynamic evolution of the chemical environment due to changes in mineral distribution, partitioning the system into fluid buffered and mineral buffered environments. These pore scale heterogeneities can lead to the development of water chemistry microenvironments different than the bulk solution (Li et al., 2007), not unlike the effect of microbes that change localized reaction rates by orders of magnitude (Mielke et al., 2003). During evaporation of a 5 wt.% brucite slurry, the MWI was estimated to increase by approximately two orders of magnitude (Fig. 4.3). Simple batch geochemical modeling of Enhanced mineral reactivity in partially wetted porous media44[HCO3-]-dependent kinetic brucite dissolution suggests that carbonation rates may be increased by a similar magnitude (Fig. 4.4A). The increase in MWI triggered by particle movement may drive “brucite-poor,” fluid buffered zones into the more reactive “brucite-rich,” mineral buffered field (Fig. 4.4A), enhancing overall reactivity of the system as evaporation progresses. Of particular consequence is that an increased MWI in fluid buffered microenvironments could stimulate carbonation in previously unreactive pores (Fig. 4.4A). Thus particle mobility may initiate reaction in otherwise unreactive microenvironments. For example, Assima et al. (2012) observed that carbonation reaction in periodically wetted brucite-rich mine wastes was double that of experiments in which the same volume of water was introduced in a single wetting event. Manipulation of these processes could lead to enhanced control of mineral reactivity in agriculture and industrial processes such as carbon sequestration. Rates of fertilizer and mineral dissolution could be optimized by tailoring wetting/drying cycles to increase the MWI.  The dynamic evolution of the MWI in soils in a changing climate may likewise lead to changes in chemical weathering rates in the vadose zone through solute release to river catchment areas. The effect of the changing MWI on mineral reaction rate will depend on the relative rate of attachment/detachment of ions at the mineral surface during dissolution/precipitation (“reaction controlled” if rate limiting) and the timescale of reactant/product transport towards and away from the mineral surface (“transport controlled”). The dimensionless Damköhler number ( Da ), can be used to distinguish between these regimes, and is defined as the ratio of the timescale of transport to the timescale of reaction (Steefel and Maher, 2009). The Da  is related to the MWI as follows (Eq. 4.1):Da = t r e s  k MW ICeq                                               (Eq. 4.1)where t r e s  is the residence time of the fluid (s), k  is the kinetic dissolution rate constant (mol m-2 mineral s-1), Ceq  is the solubility of the mineral phase (mol m-3 fluid), and MWI is in units of m2 mineral m-3 fluid. As Da  is proportional to MWI, particle movement could shift Enhanced mineral reactivity in partially wetted porous media45the overall control of the reaction (Fig. 4.4B). Although our experiments were specifically designed to remain reaction controlled, a three order of magnitude increase in Da  indicates that a transition from reaction to transport control could be achieved, for example, via adjustment of CO2 supply rate (Fig. 4.4B).At the transport timescale typical of soil moisture in the unsaturated zone (~106 s) (Stewart and McDonnell, 1991), reaction of the common silicate minerals quartz and anorthite [CaAl2Si2O8] may change from reaction to transport controlled as a function of the MWI (Fig. 4.4B). In contrast, calcite [CaCO3] remains transport controlled over a large range of MWI. 876 4-2-3-4 -5 Log peak reaction rate (mol CO2 L-1 s-1) bulkbrucite-poor brucite-rich0.25 1.00 10.0 50.0Volume% brucite5Log MWI 0  (m 2 m-3 )3aLog Da 420 -2 -4 876 4 53Log MWI (m 2 m-3 )b6Transport controlledReaction controlled1 06 s1 06 s1 01 sC a l cit eA n ort h i t eQuar tzB r u citeFigure 4.4. Modeled carbonation rate versus initial mineral-water interfacial area (MWI0) and volume percent brucite (A) and Damköhler number ( Da ) versus MWI (B). Grey bars in (A) illustrate the approximate range of ‘brucite-poor’, bulk (initial slurry), and ‘brucite-rich’ zones observed in the micromodels. Inset photomicrographs in (A) display examples of brucite-poor, initial (bulk), and brucite-rich slurry from the evaporation experiment pictured in Figure 4.3. Lines in (B) were calculated using a transport timescale equal to the approximate residence time of soil moisture (~106 s; Stewart and McDonnell, 1991), save for brucite, which was calculated based on the CO2 velocity in the experimental systems. The residence time of CO2 in the micromodels was ~10 s. The timescale of reaction was defined as the time to reach 80% equilibrium (i.e., a saturation index of -0.1), and was determined using geochemical modeling of kinetic dissolution of each mineral (refer to Appendix 2 for details regarding modeling). Enhanced mineral reactivity in partially wetted porous media46The net solute flux derived from a catchment depends on the reaction regime (Maher, 2011): solute flux will be increased disproportionately for transport controlled regimes (Tipper et al., 2006). For example, Tipper et al. (2006) attribute seasonal variation in dissolved load of river systems to the combined influence of transport controlled carbonate dissolution, which will vary strongly between high and low runoff periods, and reaction controlled silicate dissolution, which is more strongly influenced by the fluid residence time.  Figure 4.4 implies that under the influence of extreme wetting and drying cycles, as might be induced by climate change, some silicate minerals may switch from reaction controlled to transport controlled, which in turn will affect solute fluxes and seasonal variations thereof. If the net effect of extreme wetting/drying cycles is to increase MWI, catchments may be driven towards ‘chemostatic’ behavior, as defined by Maher (2011), which maximizes solute fluxes derived from chemical weathering and thus the capture of atmospheric CO2. Alternatively, if wetting during extreme precipitation events redistributes particles and lowers the MWI, the net impact will be negligible compared to current estimates of weathering fluxes. 47Influence of surface passivation and water content5. Influence	of	surface	passivation	and	water	content on mineral reactions in unsaturated porous media: Implications for brucite carbonation and CO2 sequestration45.1 Introduction The evolution of mineral reactive surface area during dissolution-precipitation reactions is an important control on long-term reaction rates in both natural and anthropogenic environments (White and Brantley, 2003). Mineral dissolution reactions require that mineral surfaces be exposed to a reactive fluid, a criterion that is not always met in porous media. Dry conditions in unsaturated media may leave mineral surfaces insufficiently wetted for dissolution-precipitation reactions to occur, and precipitation of secondary phases at the surface of dissolving phases and within pores may occlude and passivate reactive surfaces. Understanding the effects of water scarcity and secondary precipitates on reactive surface areas of dissolving minerals is important for many areas of research in the Earth sciences, such as nutrient and element cycling, water quality, and CO2 sequestration. The reaction of silicate and hydroxide minerals, such as brucite [Mg(OH)2], with CO2 to form carbonate minerals represents a model system to study these effects due to the coupled nature of the reactions and the extensive reaction progress that can be achieved on an experimental time scale (e.g., Daval et al., 2009a; Assima et al., 2013a; Power et al., 2013b and references therein). Such carbon mineralization reactions are currently under investigation due to their potential application as CO2 sequestration technologies to help offset anthropogenic greenhouse gas emissions (Seifritz, 1990; Lackner et al., 1995; Lackner et al., 1997; Lackner, 2003). These technologies 4A version of this chapter is published and is reprinted with permission from Geochimica et Cosmochimica Acta, 148, Harrison, A. L., Dipple, G.M., Power, I.M. and Mayer, K. U., Influence of surface passivation and water content on mineral reactions in unsaturated porous media: Implications for brucite carbonation and CO2 sequestration, 477-495, Copyright (2014), with permission from Elsevier.48Influence of surface passivation and water contentinvolve injection of CO2-rich fluids or gases into subsurface mafic and ultramafic formations (McGrail et al., 2006; Kelemen and Matter, 2008; Gislason et al., 2010; Van Pham et al., 2012; Gislason and Oelkers, 2014), ex situ acceleration in industrial reactors (Gerdemann et al., 2007; Power et al., 2013b), and reaction of Mg and Ca-rich industrial wastes (Pronost et al., 2011; Bobicki et al., 2012; Pronost et al., 2012; Oskierski et al., 2013; Power et al., 2013b; Power et al., 2014a; Assima et al., 2014b; Wilson et al., 2014; Chapter 7). Similar reactions occur in natural environments during weathering and carbonation of ultramafic rock (e.g., Kelemen and Matter, 2008; Boschi et al., 2009; Power et al., 2014b). Prediction of reaction progress and the fate of CO2 in these and other environments requires accurate representation of the evolution of reactive surface area. Numerous studies have endeavored to determine the effect of surface coatings on mineral dissolution rates, with conflicting results (c.f., Hodson, 2003; Park and Fan, 2004; Cubillas et al., 2005; Béarat et al., 2006; Lekakh et al., 2008; Andreani et al., 2009; Huntzinger et al., 2009; Daval et al., 2009a; Daval et al., 2009b; Daval et al., 2011; Stockmann et al., 2011; Stockmann et al., 2013). For example, Stockmann et al. (2011; 2013) report no inhibitory effects of calcite precipitation on the dissolution of basaltic glass or diopside. Yet, several studies report inhibited dissolution of serpentine and olivine due to formation of a silica-rich layer at the mineral surface (Park and Fan, 2004; Béarat et al., 2006; Andreani et al., 2009; Daval et al., 2011; Sissmann et al., 2014; Johnson et al., 2014), and Daval et al. (2009a) documented passivation of wollastonite via calcite precipitation. Similarly, Hövelmann et al. (2012a) suggest that precipitation of magnesite during olivine carbonation clogs pores and limits reaction progress. Yet, the passivating effect of Mg-carbonates alone may be difficult to assess during silicate carbonation due to the concomitant formation of potentially passivating Si-bearing phases (King et al., 2010). Moreover, a multitude of metastable Mg-carbonate minerals are known to form during carbon mineralization reactions (Hänchen et al., 2008; Power et al., 2009; Wilson et al., 2009a; Beinlich and Austrheim, 2012; Power et al., 2013b). Differences in structure amongst these Mg-carbonates and between Mg- and Ca-carbonates 49Influence of surface passivation and water contentmeans that their passivating effects are likely highly variable, and may differ from the Ca-carbonates studied previously. The formation of hydrous Mg-carbonates is favored over precipitation of the anhydrous Mg-carbonate, magnesite [MgCO3] at low temperature (e.g., Hänchen et al., 2008); hence the precipitation of secondary carbonates is accompanied by a loss of pore water due to its incorporation in the mineral structure. The initial volume of water available may restrict the mass of CO2 that can be stored (i.e., the reaction progress) based on the stoichiometry of the hydrated phase (e.g., Schaef et al., 2011). For example, the extent of carbonation of chrysotile mining residues at ambient conditions as well as reaction of brucite and several silicates with wet supercritical CO2 is limited under low water conditions (Loring et al., 2011; Schaef et al., 2011; Schaef et al., 2013a; Assima et al., 2013a; Miller et al., 2013). Evaluation of the passivating effects of hydrated Mg-carbonates is therefore complicated by the concurrent loss of pore water. In this study, we investigate the controls on brucite [Mg(OH)2] carbonation in unsaturated column experiments supplied with 10% CO2 gas, that provide simplified representations of carbon mineralization reactions in partially water saturated porous media. The objective of this study was to evaluate the effect of water saturation and precipitation of hydrous Mg-carbonates at the brucite surface on reaction progress. Because our experiments investigated carbonation of a silica-free mineral, we can assess the effect of Mg-carbonate precipitates and their different morphologies on the extent of passivation, without the complication of silica layer formation. In addition, experiments of varying water saturation were conducted in order to distinguish between surface passivation-limited and water-limited reaction. The passivating effect of surface coatings has mainly been assessed in fluid dominated stirred reactors as opposed to mineral dominated column reactors as employed in this study. Passivating effects may differ in these environments due to the limited pore volumes in which secondary phases can form, the relative scarcity of water, and the lack of abrasion that could loosen precipitates in a stirred reactor. The reactive transport model, MIN3P-DUSTY (Mayer et al., 2002; Molins and Mayer, 2007) is applied to help elucidate the processes governing carbonation and to aid in 50Influence of surface passivation and water contentthe development of improved modeling capabilities to capture the evolution of reactive surface area during coupled dissolution-precipitation reactions, with implications for predicting the fate of CO2 in subsurface formations or alkaline waste piles.5.2 Methods5.2.1 Experimental design Two types of column experiments were used to investigate the effects of brucite grain size and water content on the carbonation reaction. The first consisted of three columns at 35% water saturation, with saturation defined as the ratio of water volume to pore volume. Each column contained a different size fraction of brucite, a ‘very fine’ fraction of <53 μm, a ‘fine’ fraction with particles between 53 and 180 μm, and a ‘medium’ fraction with particles ranging from 250 to 500 μm in diameter. The second set of experiments consisted of columns with different water saturations: 15%, 35%, and 50%, all of which contained ‘medium’ brucite. These saturation values correspond to the total bulk water saturations, while the actual saturations varied somewhat with depth. Duplicate 15%, 35% and 50% saturated medium brucite columns were set up to assess the reproducibility of carbonation rates, but no solids were sampled from these duplicate columns. In order to assess the effect of fine particulates coating larger brucite surfaces that were produced during the crushing process, a triplicate 35% saturated medium brucite column was conducted using brucite that had been repeatedly rinsed with deionized water to remove these fine particles. All experiments were conducted in 16.0 cm tall × 5.9 cm diameter polycarbonate columns manufactured by W.A. Hammond Drierite. These were filled with 10 wt.% pulverized brucite ore and 90 wt.% quartz sand to a height of 12.3-13.7 cm (Fig. 5.1). The brucite/quartz mixtures were prepared by mechanically mixing 45 g of brucite ore, and 405 g of quartz sand (total mass = 450 g). Solids were poured into columns and mechanically re-homogenized. The columns were tapped gently to allow solids to settle according to their intrinsic bulk density. The brucite ore was obtained from Premier Magnesia LLC, and was pulverized using a hammermill and sieved to separate into the appropriate size 51Influence of surface passivation and water contentfractions. The quartz sand was a product of Lane Mountain Materials that had been sieved to <600 μm (median ≈ 210 μm). The initial major oxide composition of the brucite ore and quartz sand was determined using X-ray fluorescence spectroscopy (XRF; refer to Appendix 3 (A3) for details). XRF measurements of 5 duplicate samples indicated that the oxides present in the brucite ore at ≥ 1.00% ± 1σ abundance were: MgO (60.16 ± 0.43%), SiO2 (2.71 ± 0.04%), CaO (2.04 ± 0.03%), with 34.29 ± 0.43% loss on ignition. Analysis of duplicate quartz sand samples indicated that it was 97.94 ± 1.26% SiO2 and ≤ 2.49% Al2O3. Rietveld refinement of X-ray diffraction (XRD) data from analysis of duplicate samples indicated that the quartz sand was nearly 100% pure with trace muscovite (≤ 0.9 wt.%). Analysis of triplicate samples of the medium brucite ore indicated it contained 78.8 ± 3.8 wt.% brucite, 5.5 ± 0.4 wt.% dolomite, 1.9 ± 0.3 wt.% magnesite, 7.4 ± 1.0 wt.% hydromagnesite, and <0.5 wt.% lizardite and pyroaurite. The remainder was amorphous content. Analysis of single samples of the very fine and fine size fractions indicated there was no significant difference in mineral abundance between the different size fractions, save for magnesite which was equal to 2.7 wt.% and 4.0 wt.% in the very fine and fine size fractions, respectively. The surface area of the brucite ore 12.3-13.7 cmGas exit to CO 2 sensor5.9 cm10% CO 2(g)15 mL min -1from tankperistalticpumpreservoir45 g brucite405 g quartzcolumnFigure 5.1. Schematic of experimental apparatus. 52Influence of surface passivation and water contentwas determined on duplicate samples using BET with N2 adsorption, and was equal to 5.2 ± 0.2, 3.7 ± 0.3, and 2.4 ± 0.7 m2 g-1 for the very fine, fine, and medium brucite, respectively (Table 5.1). The particle size distribution of the very fine and fine brucite was determined using a Malvern Mastersizer 2000 Laser Diffraction Particle Size Analyzer (A3 Fig. A3.1). This indicated that the mean particle radius was 12 μm and 39 μm for the very fine and fine Table 5.1. Summary of experimental conditions.  Experiment  Brucite particle diameter  (µm)  Mean  particle radius (µm)  Brucite surface area (m 2 g-1 ) Sediment  height (cm)  Porosity  Volume  water added (mL)  Saturation a (%) Grain size  trial very fine <53 12 5.2 12.3 0. 4 9 58 35 fine 53-180 39 3. 7 12.3 0. 4 9 58 35 medium 250-500 188 2.4 13.7 0. 5 4 71 35           Saturation  trial 15% 250-500 188 2.4 13.2 0.5 3 29 15 35% 250-500 188 2.4 13.7 0. 5 4 71 35 50% 250-500 188 2.4 13.4 0. 5 3 97 50 a Saturation = (volume water/volume pore space)   100%  brucite, respectively. The medium brucite size fraction had a mean particle radius of 188 μm as estimated based on sieving results (A3 Fig. A3.1). The porosity of the sediment in each column was calculated based on the density of the solids and the bulk volume of the porous media; it ranged from 0.49–0.53 (Table 5.1). Initial solutions contained 0.1 M MgCl2 added as MgCl2·6H2O from Fisher Scientific to deionized water, and <4.0 × 10-4 M dissolved inorganic carbon (DIC) from the laboratory atmosphere. The solution was slowly applied across the sediment surface in the columns and allowed to infiltrate the column under the force of gravity. The volume of solution added was dependent on the desired water saturation for each experiment (Table 5.1). Once the wetting front was observed to reach the base of the column, the gas supply was initiated.Columns were supplied with 10.0 vol.% gaseous CO2 (90.0 vol.% N2), through a hose barb at their base (Fig. 5.1). The gas streams were not humidified, and therefore induced evaporation from the columns. Gas was introduced to the base of the sediment through a 53Influence of surface passivation and water contentstainless steel grate covered with fabric mesh (300 mesh). A Cole-Parmer Masterflex® L/S precision® standard pump system fitted with an L/S Easy-Load II® pump head was used to supply the gas stream at ~15 mL min-1 (~2.7 × 10-3 g CO2 min-1) to each column at atmospheric pressure and room temperature (~21°C). The gas stream exited through a port at the top of each column to maintain close to atmospheric pressure within the column (Fig. 5.1). The CO2 concentration of the gas effluent from each column was recorded at five minute intervals using Vaisala® GMT221 CO2 concentration sensors. The accuracy of the sensor measurements varied between ± 0.5% to ± 2.0% depending on the measurement apparatus, which at times became contaminated with ambient laboratory air. Gas composition data were unavailable for the 15% saturated columns due to atmospheric contamination of the measurement apparatus. The CO2 breakthrough is expressed as ‘C/C0,’ the ratio of the CO2 concentration in the gas effluent at a given time to the measured CO2 concentration of the gas effluent after the reaction was complete. Experiments were conducted for time periods ranging from 160 to 261 h depending on the time required for the reaction to cease. Following completion of the experiments, solids were sampled by removing the entirety of the material in 2 cm intervals. The water saturation profile was estimated gravimetrically using the mass difference before and after samples were dried at room temperature. Solid samples were then analyzed for total inorganic carbon content using coulometry and mineral abundance using Rietveld refinement of XRD data (refer to A3). The solid samples were also prepared for characterization using scanning electron microscopy (SEM) both as disaggregated powders and as polished epoxy-embedded mounts to allow cross-sectional views of reacted grains. Experimental conditions including grain size, water content, and porosity are summarized in Table 5.1. For further detail regarding the experimental setup and analytical techniques, refer to Appendix 3. 54Influence of surface passivation and water content5.2.2 Assessment of reaction progressThe mass of CO2 sequestered was estimated using several lines of evidence, including (1) the total carbon content in the solid phase, (2) the mass gain of the columns during the course of the experiment, and (3) abundance of mineral phases. The measured carbon content for all experiments is expressed as %CO2 by mass. The initial brucite/quartz mixtures had an average of 0.67 ± 0.07% CO2 based on analysis of 6 samples that was contained in dolomite, hydromagnesite, and magnesite present in the initial brucite ore. This initial mass of CO2 was subtracted from CO2 content values measured for reacted samples to determine the mass of CO2 gained. This was justified by the presence of approximately the same mass of dolomite in the reaction products as in the initial material, although the data were insufficient to resolve changes in hydromagnesite and magnesite abundances. To estimate the CO2 gain gravimetrically, columns were placed on a scale with ± 0.01 g accuracy and their mass was recorded four times a day. At least one column in every experimental trial began to lose mass by the end of the experiment due to evaporation driven by the flux of dry CO2/N2 gas. Under the assumption that the evaporation rate remained relatively constant, this mass loss rate was applied to all experiments of a given trial to correct the gravimetric CO2 measurements for evaporative loss (i.e., measured mass gain + evaporative loss = total CO2 gain). Both total carbon and gravimetric measurements provided comparable estimates of the mass of CO2 sequestered. The error on total carbon measurements was ± 0.22 g and was ± 0.04 g for gravimetric measurements. The error on gravimetric measurements is small compared to the uncertainty in evaporation rate calculations. However, the accuracy of reaction progress measurement is confirmed by good agreement between these two independent analyses.The mass of CO2 contained in crystalline carbonate phases was quantified using Rietveld refinement of XRD data. In order to determine whether CO2 was also stored in non-crystalline phases, a known mass of a highly crystalline phase ([CaF2] or [Al2O3]) was added to experimental samples. This allowed for quantification of poorly- or nano-crystalline phases (e.g., Gualtieri, 2000). For further detail regarding extent of carbonation measurement methods 55Influence of surface passivation and water contentand results, refer to Appendix 3. Instantaneous carbonation rates (rinstant in g CO2 h-1) were determined based on the mass gain between each measurement as follows (Eq. 5.1):r i n s t a n t = m 2 – m 1t 2 – t 1                                                  (Eq. 5.1)where mi is the column mass (g) at a given measurement time, and ti is the time of measurement (h).5.2.3 Reactive transport modeling The experimental conditions were modeled using the multicomponent reactive transport code MIN3P-DUSTY (Molins and Mayer, 2007) to elucidate reaction mechanisms and allow for calibration of a model to capture reaction progress during carbonation of brucite. MIN3P-DUSTY comprises a suite of chemical reactions including mineral dissolution-precipitation, and can model flow and transport in both the gas and aqueous phases. Gas transport was modeled as an advective-diffusive process according to Darcy’s law and the Dusty Gas model (Mason and Malinauskas, 1983; Molins and Mayer, 2007). For a complete description of the constitutive equations refer to Mayer et al. (2002) and Molins and Mayer (2007). Brucite dissolution kinetics were modeled using a far-from-equilibrium HCO3- concentration-dependent kinetic dissolution rate law based on data from Pokrovsky et al. (2005b) and Pokrovsky and Schott (2004) (Eqs. 5.2 and 5.3):r b r c  = k e f f0 HCO 3- 0.56 (1- 2)                                           (Eq. 5.2)k e f f0  = k0SA                                                        (Eq. 5.3)where k0eff is the effective rate constant, k0 is the initial reaction rate constant equal to 10-6.13 L s-1 m-2, SA is the brucite surface area (m2), and Ω is the saturation ratio. Saturation ratio is 56Influence of surface passivation and water contentdefined as the ratio of the ion activity product to the equilibrium constant. Although the presence of multiple carbonate phases was evident in most experiments, thermodynamic data were only available for nesquehonite; therefore nesquehonite precipitation was used as a surrogate to represent precipitation of all secondary carbonate phases. Nesquehonite precipitation was effectively simulated as an equilibrium process. Porosity was set to a representative value of 0.53 for all simulations, and soil hydraulic function parameters were estimated based on results of Tempe cell tests of the solid materials (e.g., Fredlund and Rahardjo, 1993) and measured water saturation profiles. CO2 breakthrough curves measured in the experiments were used as a fitting parameter to estimate effective reactive surface areas and interpret reaction mechanisms. For a complete list of transport parameters used in the simulations, refer to Appendix 3. 5.3 Results5.3.1 Instantaneous carbonation rates and CO2 breakthrough The replacement of brucite by carbonate phases was evident in all experiments (Figs. 5.2 and 5.3). Four distinct stages in the carbonation reaction were identified, with each stage typified by a distinct trend in carbonation rate and CO2 breakthrough (Figs. 5.4 and 5.5). Although all stages were reproducible in duplicate experiments, the presence of each stage was dependent on the experimental parameters (Figs. 5.4 and 5.5). Specifically, stage 1 was absent from the very fine and fine brucite column experiments, and stage 2 was absent from the 15% saturated medium brucite columns. Stage 1 in the medium brucite columns was demarcated by rapid carbonation rates (up to ~0.16 g CO2 h-1) for the first ~1-2 h of the experiment, which corresponded to periods of negligible CO2 in the outflow (Figs. 5.4C and 5.5C and D). Because conservative flushing of one pore volume in the columns would take approximately 0.1 h in these columns, the period with negligible CO2 in the outflow is attributed to it being trapped in the column, in both the solid and aqueous phase. Stage 2 was typified by periods of relatively constant reaction rates, and consequently, 57Influence of surface passivation and water contentrelatively constant CO2 concentrations in the gas effluent (Figs. 5.4B-C and 5.5). In both the very fine and fine brucite columns, stage 2 persisted for the first ~68 h, during which little to no CO2 was measured at the column outlet (Figs. 5.4B and 5.5A-B). Carbonation rates during this period were on average 0.15 g CO2 h-1 in both columns (Table 5.2). Similarly, in the 35% and 50% saturated medium brucite columns, stage 2 was marked by relatively constant carbonation rates, and relatively constant, but elevated CO2 concentrations in the gas effluent of 4.0-5.0% and 5.0-6.0% CO2, respectively (Fig. 5.5C-D). Rates were between 0.08 and 0.11 g CO2 h-1 in the 35% saturated columns, and between 0.07 and 0.10 g CO2 h-1 in the 50% saturated columns. Carbonation rates in the finer grained brucite columns exceeded those in Depth (cm)B) fineA) very fine024681012140 4 8 12 16 0 4 8 12 16C) mediumAbundance (wt.%)experimentrinsedexperimentCO2 sequesteredmodel0 4 8 12 16nesquehoniteflakey Mg-carbonatebruciteexperimentMineralsmodelCO2 sequesteredexperimentmodelCO2 sequesteredFigure 5.2. Mineral abundance profiles determined using Rietveld refinement of X-ray diffraction data for 35% saturated columns with very fine (A), fine (B), and medium (C) brucite. The black shaded, dotted, and diagonally striped areas represent the abundance of brucite, the poorly crystalline flakey Mg-carbonate phase, and nesquehonite, respectively. Lines with symbols represent the abundance of CO2 by mass in the solid material with depth as determined using %CO2 measurements. %CO2 data from a 35% saturated column using brucite that was rinsed to remove the surface powder is plotted with a blue line and filled square data points in (C). Dashed lines represent MIN3P-DUSTY output for the abundance of CO2 by mass in the solid phase for each grain size using the threshold function. CO2 was supplied at the base.58Influence of surface passivation and water contentthe medium brucite columns on average by 1.6-1.9 times during stage 2.  In stage 3, carbonation rates declined steadily in all experiments, with a concurrent increase in the CO2 content of the gas effluent (Figs. 5.4B-C and 5.5). During stage 3, the CO2 content increased to ~7.5% and 8.0% CO2 in the very fine and fine brucite columns, respectively, and to ~9.5-10.0% in the 35% and 50% saturated medium brucite columns (Fig. 5.5). Transient sharp changes in CO2 content over short durations (i.e. ~1 h) are attributed to disturbance of the CO2 content measurement vials during column mass measurements (Fig. 5.5). Although gas composition data were unavailable, stage 3 was evident in the 15% saturated medium brucite column, with carbonation rates declining throughout the experiment and becoming negligible Abundance (wt.%)0 4 8 12 16Depth (cm)02468101214A) 15% B) 35% C) 50%experimentrinsedexperimentCO2 sequesteredmodel0 4 8 12 16 0 4 8 12 16experimentCO2 sequesterednesquehoniteflakey Mg-carbonatebruciteCO2 sequesteredexperimentMineralsFigure 5.3. Mineral abundance profiles determined using Rietveld refinement of X-ray diffraction data for columns containing medium brucite at 15% (A), 35% (B), and 50% (C) water saturation. The black shaded, dotted, and diagonally striped areas represent the abundance of brucite, the poorly crystalline flakey Mg-carbonate phase, and nesquehonite, respectively. Lines with symbols represent the abundance of CO2 by mass in the solid phase with depth as determined using %CO2 measurements. %CO2 data from a 35% saturated column using brucite that was rinsed to remove the surface powder is plotted with a blue line and filled square data points in (B). Dashed lines represent MIN3P-DUSTY output for the abundance of CO2 by mass in the solid phase for each grain size using the threshold function. CO2 was supplied at the base.59Influence of surface passivation and water contentTime (h) Time (h)Time (h)0 50 100 150 2000 50 100 150 200 0 50 100 150 200A) all experimentsB) grain size (all 35% saturated) C) saturation (all medium brucite)Mass of CO2 sequestered (g)04812160.000.040.080.120.160.200.000.040.080.120.160.20Carbonation  rate (g CO2 h-1)Carbonation  rate (g CO2 h-1)S1S2S3S4S2S1S3S4S2Legendvery finefinemediumGrain size15%35%50%SaturationGravimetric dataSolid phase %CO 2  dataGravimetric dataSolid phase %CO 2  dataS2S3S4Figure 5.4. Mass of CO2 sequestered versus time calculated based on the column mass gain over time (A). Instantaneous carbonation rate versus time in 35% saturated columns containing very fine, fine, and medium brucite (B). Instantaneous carbonation rate versus time in columns containing medium brucite at 15%, 35%, and 50% water saturation (C). Square symbols represent columns at 35% saturation, triangles are 50% saturation, and circles are 15% saturation. Grain size for columns of the same water saturation are distinguished by color, whereas all columns with medium brucite are represented by various shades of green. Open symbols represent duplicate experiments, and diamonds represent the mass of CO2 sequestered calculated using %CO2 measurements. ‘SX’ labels indicate reaction stages 1-4.60Influence of surface passivation and water contentTime (h) Time (h)0 50 100 150 20000.20.40.60.81.0CO2 breakthrough (C/C0)A) very fine, 35% saturatedB) fine, 35% saturatedS2S3S4threshold modelgeometric modelexperiment CO2 breakthrough (C/C0)0 50 100 150 200S2S3S4Time (h)0 50 100 150 200C) medium, 35% saturatedCO2 breakthrough (C/C0)S1S2S3S4CO2 breakthrough (C/C0)Time (h)0 50 100 150 200S1S2S3S4D) medium, 50% saturatedthreshold modelgeometric modelexperimentthreshold modelgeometric modelexperimentduplicate exp.experimentduplicate exp.00.20.40.60.81.000.20.40.60.81.000.20.40.60.81.0Figure 5.5. CO2 breakthrough curves measured at column outlets versus time and MIN3P-DUSTY modeling results for 35% saturated columns containing very fine (A), fine (B), and medium brucite (C), and columns containing medium brucite with 50% water saturation (D). CO2 breakthrough data were unavailable for 15% saturated medium brucite columns. CO2 breakthrough curves are expressed as ‘C/C0’, where C is the CO2 concentration of the gas effluent at a given time, and C0 is the CO2 concentration of the effluent following the carbonation reaction. Solid black, red, and green lines represent experimental data from the very fine, fine, and medium 35% and 50% saturated columns, respectively. Dashed and solid blue lines represent MIN3P-DUSTY output from simulations employing the geometric and threshold models, respectively. ‘SX’ labels indicate reaction stages 1-4. 61Influence of surface passivation and water contentafter ~97 h (Fig. 5.4C).Stage 4 was evident in all columns, and was defined by a period of negligible carbonation rate and relatively constant CO2 concentration in the gas effluent, approximately equal to the composition of the supplied gas within the estimated measurement error (Figs. 5.4B-C and 5.5). Although the final CO2 content of the gas effluent did not exactly equal the inlet composition in the very fine and fine brucite columns, the lack of mass gain after ~115 hours (Fig. 5.4A), and the near complete depletion of brucite in these experiments (Fig. 5.2) clearly indicate that stage 4 is coincident with the effective cessation of the carbonation reaction. 5.3.2 Reaction progressThe total mass of CO2 sequestered as determined using carbon content (%CO2) measurements indicated that a total of 16.8 g, 13.1 g, and 9.0 g CO2 were sequestered in the very fine, fine, and medium brucite columns at 35% saturation, respectively. A total of 5.2 g, 9.0 g, and 8.6 g CO2 were sequestered in the 15%, 35%, and 50% saturated medium brucite columns, respectively. Similarly, 9.3 g CO2 were sequestered in the 35% saturated column with rinsed medium brucite. These values are in very good agreement with those determined gravimetrically (Fig. 5.4A; Table 5.2). Table 5.2. Summary of mass of CO2 sequestered, carbonation rate, reaction stoichiometry, and extent of carbonation in experiments for which solids were analyzed. Total mass CO 2 sequestered  (g)       Experiment  Mass  CO2 in nsq (g) %C O2 data Experiment  mass gain  Mass  CO2 in flakey  phase (g) Extent of  carbonation (%) Bulk  molar Mg:CO 2 ratio Carbonation ratea  (g CO2 h-1 )  Grain size trial   very fine 4.0 16.8 16.4 12.8 94 1.5 0.15  fine 3.6 13.1 14.3 9. 4 81 1.7 0.15  medium 2.1 9.0(9. 3) b 8.3(9.2) b 6.9 59 1.8 0. 0 9               Saturation  trial          15% 0.7 5.2 4.4 4. 5 35 1.8 <0 . 0 8 35% 2.1 9.0(9. 3) b 8. 3(9.2) b 6.9 59 1.8 0. 0 9  50% 4.7 8. 6 8. 4 3. 9 58 1.8 0. 0 8  aAverage carbonation rate during stage 2.   bBracketed values are from the 35% saturated rinsed medium brucite column.   62Influence of surface passivation and water contentIn all experiments, brucite was replaced by a combination of the hydrated Mg-carbonate mineral, nesquehonite [MgCO3·3H2O], and an amorphous or nano-crystalline solid phase that could not be resolved using XRD (A3 Fig. A3.2). The abundance of this phase as determined using Rietveld refinement included any non-crystalline content in the initial materials. However, it is the relative trends in amorphous content that are important, rather than the absolute values. The total mass of CO2 sequestered in each column was significantly greater than can be attributed to the measured abundance of nesquehonite (Table 5.2). This discrepancy implies that the unidentified phase quantified using the XRD data must be a carbonate phase. The greatest extent of carbonation in terms of the mass of brucite converted to carbonate was achieved in the very fine brucite column at 94%, followed by the fine brucite column at 81%. A similar extent of carbonation was documented in the 35% and 50% saturated medium brucite columns, at 59% and 58%, respectively. The lowest extent of carbonation was attained in the 15% saturated medium brucite column at 35%. The abundance and distribution of nesquehonite, the poorly crystalline carbonate phase, and total CO2 sequestered varied significantly between experiments of different grain size (Fig. 5.2). Both the very fine and fine brucite columns exhibited similar %CO2 trends, with the greatest CO2 content near the column inlet (Fig. 5.2). Similarly, brucite abundance was lowest at the column inlet, where nesquehonite abundance tended to be highest (Fig. 5.2). Although a similar extent of carbonation was achieved in the 35% and 50% saturated columns, the distribution of brucite and carbonate precipitates differed (Fig. 5.3). In the 35% saturated column, total CO2 content and nesquehonite abundance decreased along the flow-path, whereas nesquehonite, brucite, and total CO2 were more evenly distributed throughout the 50% saturated column (Fig. 5.3). The poorly crystalline carbonate phase increased in abundance along the flow-path in all 35% saturated columns, but was more abundant with depth in the 50% saturated medium brucite column (Figs. 5.2 and 5.3). The total CO2 content in the 15% saturated medium brucite column was effectively constant along the flow-path (Fig. 5.3A). 63Influence of surface passivation and water content5.3.3 Qualitative characterization of solidsSEM micrographs of the initial material revealed that the brucite grains were coated by brucite powder, an artifact of the crushing process (Fig. 5.6A). Reaction products were comprised of elongated, narrow crystals, consistent with the morphology of nesquehonite (Ferrini et al., 2009; Harrison et al., 2013a; Chapter 2), and flakey material similar to that documented for other hydrated Mg-carbonate-hydroxide phases, such as dypingite (e.g., Power et al., 2007; Power et al., 2009; Power et al., 2013a; Fig. 5.6B-C). Energy dispersive spectroscopy confirmed that these phases consisted of Mg, C, and O (A3 Fig. A3.4). The observed abundance of flakey material was greater in the very fine and fine brucite columns than in the medium brucite columns, and generally increased along the flow path. We interpret the flakey material to represent the poorly crystalline carbonate-hydroxide phase, as the observed trend in abundance of this material along the flow path in the very fine and fine columns is consistent with the XRD data (Fig. 5.2A-B). The nanometer scale thickness of the carbonate flakes is consistent with its X-ray amorphous nature, as the crystals may be too small to exhibit long-range order that can be resolved using XRD. Cross-sectional views of reacted grains from the medium brucite columns showed that carbonate precipitates generally surrounded and coated unreacted brucite, whereas quartz surfaces typically remained uncoated (Fig. 5.6B). In some instances brucite grains had one or more sides lacking carbonate precipitates (Fig. 5.6B).  Conversely, in the very fine brucite columns, carbonate precipitates formed interstitial cement that filled pore spaces between brucite and quartz grains (Fig. 5.6D). Nesquehonite rinds appeared non-porous at a submicron scale (Fig. 5.6E-F). 5.3.4 Water content Final water content measurements are provided in terms of the ratio of water volume to bulk porous media volume (volumetric water content) rather than water saturation, owing to possible changes in pore volume due to carbonation. In the 35% saturated columns, a water loss of 28% (16.2 g), 28% (16.3 g), and 22% (15.9 g) was measured in the very 64Influence of surface passivation and water content100 μmA6 μmC100 μmBBrcQtzNsq200 μmDQtzBrcepoxyNsq50 μmEBrcNsq2 μmBrcNsqFBrcFigure 5.6. Scanning electron micrographs of initial and reacted material. Initial medium brucite (brc) showing brucite powder on the surface of large grains (A). Reaction products from the 35% saturated medium brucite column showing nesquehonite (nsq) surrounding unreacted brucite, a lesser extent of precipitates surrounding quartz (qtz), and some edges of brucite that are not coated by carbonate precipitates (B). Micrograph showing a coating of the flakey, poorly crystalline secondary Mg-carbonate phase covering a brucite grain in a sample from the fine brucite column (C). Micrograph of a sample from the very fine brucite column showing secondary carbonate infilling the space between quartz and remaining brucite grains, and cementing grains together (D). Micrographs of non-porous nesquehonite coating a ‘medium’ brucite grain from the 35% saturated column (E-F).65Influence of surface passivation and water contentfine, fine, and medium brucite columns, respectively. In the 15%, 35%, and 50% saturated medium brucite columns, a water loss of 37% (10.6 g), 22% (15.9 g), and 17% (16.5 g) was recorded, respectively. Initial water volumes are provided in Table 5.1. Water loss occurred via evaporation and incorporation into precipitates. The very fine and fine brucite columns had final volumetric water contents (VWC) ranging between 0.09-0.17, compared to an initial bulk VWC of 0.17 (35% saturation) (A3 Fig. A3.3). VWC increased with depth and ranged from 0.08-0.24 and 0.17-0.30 in the 35% and 50% saturated medium brucite columns, respectively (A3. Fig. A3.3). In the 15% saturated medium brucite column the VWC remained relatively constant with depth, at values between 0.04 and 0.06 compared to an initial bulk VWC of 0.08 (A3 Fig. A3.3). Water movement during reaction likely led to its redistribution as the reaction progressed. The mass of water lost in all experiments is much greater than expected due to evaporation alone. Evaporative losses were estimated by bulk column mass loss rates after cessation of reaction where applicable, and were between 1.8 and 3.3 g water in all experiments. Between 69% and 89% of water lost from the pore space in the columns is therefore attributed to incorporation into hydrated carbonate phases. The mass of water in nesquehonite is insufficient to account for the total water mass incorporated into the solid phase. Based on the ideal stoichiometry of nesquehonite (3 moles water per mole nesquehonite) it contained only ~8-35% of the water attributed to the solid phase. This implies that the poorly crystalline carbonate was also a hydrated phase, as is expected under the experimental conditions (Hänchen et al., 2008). 5.3.5 Reaction stoichiometryIn order to model the carbonation reaction, it is important to quantify the reaction stoichiometry to accurately represent the mass of CO2 that can be sequestered for a given mass of brucite reacted. The reaction stoichiometry for conversion of brucite to nesquehonite is 1 mole brucite per mole CO2. However, the presence of the poorly crystalline hydrated 66Influence of surface passivation and water contentcarbonate phase complicates the overall reaction stoichiometry for the brucite carbonation reaction. Mass balance calculations based on XRD and %CO2 data reveal that if the unidentified carbonate is a single phase, the Mg:CO2 ratio in this phase is ~2. In order to maintain charge balance, this implies that the unidentified phase is likely a hydroxy-carbonate with a formula of Mg2CO3(OH)2·xH2O, similar to that of the Mg-carbonate, artinite [Mg2CO3(OH)2·3H2O]. Based on the water mass balance in each experiment, it is estimated that ‘x’ is between 1 and 3 (average ≈ 2). No thermodynamic data are available to model precipitation of the poorly crystalline carbonate phase. Thus, the reactive transport models were simplified such that nesquehonite precipitation represented formation of all carbonate precipitates. This maintains the experimental CO2, OH-, and Mg balance. Because the formation of the poorly crystalline carbonate reduced the ratio of CO2 stored per mole of Mg, the initial brucite content in the models was treated as a fitting parameter to match the measured overall reaction stoichiometry. This was done rather than changing the stoichiometry of nesquehonite to avoid alteration of its solubility product and equilibrium constant. The overall reaction stoichiometry was also calculated using the experimental data. The reduction in brucite mass and the total CO2 mass gain in each column were determined with XRD and solid phase CO2 measurements, respectively. The stoichiometric ratio was then calculated according to the following equation (Eq. 5.4):MgC= n br c i  - n b r c fn CO 2 f  - n CO 2 i                                             (Eq. 5.4)where MgC is the ratio of moles Mg consumed to moles CO2 sequestered, nbrci is number of moles of brucite initially in the column, which was determined to be 0.61 ± 0.02 moles (1σ), nbrcf is the moles of brucite at the end of the experiment, and nCO2i and nCO2f are the moles of CO2 in the solid phase at the beginning and end of the experiment, respectively. The values of MgC for the 35% saturated very fine, fine, and medium brucite columns were 1.5, 1.7, and 67Influence of surface passivation and water content1.8, respectively, compared to 1.0 for conversion to pure nesquehonite. The medium brucite columns at 15%, 35%, and 50% saturation all had MgC values of 1.8, and the average MgCfrom all experiments was 1.7. Due to the inherent measurement error in the brucite and CO2 contents, and the variation of brucite content in the initial brucite ore, these values were not directly input in the models. However, the model-fitted values were in good agreement with the calculated values, and are provided in Appendix 3. The reaction stoichiometry, along with reaction rates and extent of carbonation are summarized in Table 5.2.5.4 Discussion5.4.1 Reaction stagesBoth the absolute carbonation rate and the reaction stages were strongly influenced by brucite grain size, but were relatively insensitive to water content when above 15% initial saturation (Fig. 5.4). In general, higher reaction rates were achieved with finer grain size due to the greater specific surface area of the finer particles (Tables 5.1 and 5.2). Another key difference between the grain sizes was the absence of stage 1 in the very fine and fine brucite columns. Stage 1 is distinguished from the similarly high stage 2 reaction rates in the very fine and fine brucite columns due to its comparatively short duration (Fig. 5.4B-C). This transient rapid reaction rate is attributed to the carbonation of the brucite powder that was observed at the surface of larger brucite grains (Fig. 5.6A). It is expected that these finer particles reacted rapidly and were consumed in the first ~1-2 hours of the experiment owing to their higher surface area (Fig. 5.7; Helgeson et al., 1984; Andreani et al., 2009; Assima et al., 2013a). Although stage 1 may occur in the very fine and fine brucite columns, it remains unresolved due to the similarly high reaction rate of the bulk material. Stage 2 is demarcated by a period of relatively constant carbonation rates in all experiments, and near constant CO2 concentration in the effluent gas (Figs. 5.4 and 5.5). This represents carbonation of the bulk of the brucite (Fig. 5.7). The non-zero (~4-6%) outlet CO2 68Influence of surface passivation and water contentconcentration during stage 2 is unique to the medium brucite columns (Fig. 5.5). Slight declines in CO2 concentration during this period may be due to an increase in carbonate precipitation rate after initial nucleation has occurred. In all experiments, the final decline in reaction rate and increase in outlet CO2 concentration during stage 3 (Figs. 5.4 and 5.5) is attributed to a decrease in the abundance of reactive brucite along the column length. Finally, stage 4 is reached when the reaction rate has declined sufficiently such that CO2 flow through the column is effectively conserved (Fig. 5.7).5.4.2 Controls on reactionOver this range of experimental conditions, several controls on reaction were realized, including: CO2 supply, brucite dissolution, and water content. During stage 2 in the very fine brucite column, negligible CO2 was vented (Fig. 5.5A), and the rate of CO2 supply to the column was approximately balanced by the rate of brucite carbonation (Fig. 5.8). This implies that the overall carbonation rate was limited by the rate of CO2 supply to the column (Fig. nesquehoniteStage 1 Stage 2 Stage 3 Stage 4fines bulk reaction rate decline negligible reactionbruciteflakey carbonatecarbonation reactionFigure 5.7. Conceptual diagram of the four reaction stages: stage 1) carbonation of fines, stage 2) reaction of the bulk of the brucite, stage 3) the majority of brucite is no longer available for reaction such that the reaction rate approaches zero, and stage 4) negligible reaction. Areas shaded green represent brucite grains, grey represents flakey poorly crystalline Mg-carbonate, purple represents nesquehonite, and red indicates where the carbonation reaction is predominantly occurring during each stage.69Influence of surface passivation and water content5.8); an increase in the CO2 flux would therefore accelerate carbonation. Conversely, in the medium brucite columns, only ~40-60% of the supplied CO2 is captured before reaching the column outlet, indicating that the rate of CO2 supply outpaced the rate of carbonation. Thus, the carbonation rate of medium brucite was limited either by the rate of brucite dissolution, carbonate precipitation (Fig. 5.8), or lack of available water (e.g., Schaef et al., 2011; Assima et al., 2013a). Under similar experimental conditions, it has been documented that carbonate 0 1 2 3 4 5 6Surface area (m 2 g-1 )00.050.150.200.250.300.40Carbonation rate (g CO2 h-1) 0.100.35Experimental CO 2  supply rateCO2 supply limitedB rucite dissolution limitedIncreased CO2 supplymediumfinevery finevery fine (estimated with MIN3P)Figure 5.8. Absolute carbonation rate (g CO2 h-1) for all 35% saturated columns of different grain size versus BET measured initial brucite surface area (m2 g-1) for each grain size. The medium brucite is indicated by the green square, the fine by the red square, and the very fine by the black square. The white square with the dashed line indicates the approximate carbonation rate that would be achieved for very fine brucite with a 5-fold greater CO2 supply rate, as estimated using MIN3P-DUSTY. The grey dashed line indicates the approximate trend in carbonation rate with BET surface area if CO2 supply were not limiting. The grey shaded region indicates the rate of CO2 supply to the columns as estimated using the measured gas flow rate out of the very fine column with laboratory air flow only, and at the end of the experiment once the column ceased gaining mass.70Influence of surface passivation and water contentprecipitation occurs relatively rapidly, and is therefore unlikely to be rate limiting (Wilson et al., 2010; Harrison et al., 2013a; Chapter 2). The fine brucite experiment represents the transition between these two “endmembers” of rate limitation. The low but measureable amount of CO2 vented during stage 2 (Fig. 5.5B), and the slightly lower rate of reaction compared to the very fine brucite (Fig. 5.4B) suggests that the rate of CO2 supply slightly exceeded the carbonation rate. Thus, the moderate decrease in reactive surface area between the very fine and fine brucite sufficiently lowered the rate of brucite dissolution to shift the overall reaction from being CO2 supply to mineral dissolution limited (Fig. 5.8). However, in both the fine and medium brucite columns, a significant portion of brucite was not carbonated despite the effective cessation of the reaction before the end of the experiment (Fig. 5.4C). This implies that brucite carbonation became inhibited at some point during the reaction.The extent of carbonation was strongly dependent on the initial grain size; the amount of brucite consumed in the 35% water saturated columns declined from 94% to 81% to 59% in very fine, fine, and medium brucite columns, respectively (Table 5.2). As is consistent with results of Assima et al. (2013a) for carbonation of chrysotile mining residues, reaction progress was restricted at low water content. Only 41% of the brucite was reacted in the column with the lowest water saturation (15% saturated). It is possible that the incomplete reaction of brucite within the timeframe of the experiments can be attributed to a reduction in total brucite surface area as it is consumed, leading to a decline in the bulk brucite dissolution rate. In reactive transport models, the decrease in total surface area with dissolution is commonly updated assuming particles can be represented by a uniform population of perfect spheres or cubes that shrink as they dissolve (e.g., Lichtner, 1996; Mayer et al., 2002; Appelo and Postma, 2005; Wanner et al., 2011). This is described by the following equation, henceforth designated the ‘geometric model’ (Eq. 5.5; Mayer et al., 2002):k e f fg  = k 0 SAt02 3                                           (Eq. 5.5)71Influence of surface passivation and water contentwhere φ0 is the initial mineral volume fraction, and kgeff and φt are the effective geometric reaction rate constant and mineral volume fraction at a given time greater than zero, respectively. Here, kgeff replaces k0eff in Equation 5.2. This function calculates the decline in reactive surface area that is geometrically related to the consumption of the initial material (i.e., when φt shrinks), where the ‘2/3’ exponent represents the ratio of the surface area of a sphere to its volume. This function was applied in MIN3P-DUSTY to model the experimental conditions and was compared against experimental data. Owing to the similarity in the CO2 breakthrough curves of the 50% and 35% saturated columns, and the lack of data for the 15% saturated column, the 35% saturated columns were the only medium brucite columns that were modeled. The presence of the powder coating the bulk brucite (Fig. 5.6A) was included in the model by incorporating a small mass of brucite with a significantly higher surface area than the bulk material (stage 1). The removal of this surface powder corresponds to the sharp increase in CO2 concentration in the gas effluent after ~2 h of reaction (Fig. 5.5C-D). The very similar extent of carbonation achieved in the rinsed medium brucite column confirms that reaction of the powder comprised only a small fraction of the total brucite consumed (Fig. 5.3B). Due to the lack of a mechanistic model to simulate movement of brucite particles, as documented in microfluidic experiments (Chapter 4), this process was neglected in all models. Because the reaction rate of the finer material was controlled primarily by the rate of CO2 supply rather than brucite dissolution, and the coarser grains likely bore the skeletal load of overlying material, this is not expected to significantly alter the results. Application of the geometric model to simulate the very fine brucite column reproduced the CO2 breakthrough curve and extent of reaction in the very fine brucite column very well with an adjustment of the initial reactive surface to ~20% of the BET measured surface area (Fig. 5.5A). The BET surface area is often an overestimate of the true reactive surface area, due to variability in the reactivity of surface features and the inclusion of “internal” surface area in BET measurements, which may be less reactive than the actual “external” surface area 72Influence of surface passivation and water content(Brantley and Mellot, 2000; Gautier et al., 2001; Luttge and Arvidson, 2008). The reduction of the initial brucite surface area in the model is justified to account for these effects. The same reduction in reactive surface area was thus employed to model the fine and medium brucite columns. The geometric model predicts a gradual decrease in reaction rate during stage 2 due to consumption of the bulk material and the reduction in total surface area. Yet, it cannot reproduce the precipitous shut down of reaction prior to complete consumption of the brucite in the fine and medium brucite columns, as evidenced by the poor fit between the modeled and experimental data (‘geometric model’ in Fig. 5.5). There are several hypotheses that could explain this incomplete reaction and precipitous decline in reaction rate: 1) water-limited reaction, 2) a disproportionate decline in reactive surface area over time due to preferential consumption of highly reactive surface sites and finer grains during the early stages of the reaction, and 3) a loss of effective reactive surface area due to passivation of the brucite surface via carbonate precipitation. The two latter hypotheses are mechanisms by which brucite dissolution may be limiting, whereas water-limited reaction affects both brucite dissolution and carbonate precipitation. We evaluate these hypotheses in light of the available data to discern the primary cause for the lack of reactivity in each column, and to allow reaction progress to be better modeled. 5.4.2.1 Water-limited reactionWater is required not only as a reaction medium to solubilize and transport ions, but also as a reactant to form hydrated Mg-carbonate phases that incorporate water in their crystal structures. The distribution of water may therefore be a strong control on the extent of carbonation at the grain scale. At low water contents, water is held primarily around grains and in pore throats as opposed to larger pore spaces. Thus, the wetted pore volume available for precipitation is limited, and precipitation of carbonate phases may largely be restricted to the area immediately surrounding brucite grains (Fig. 5.6B). Although pH was 73Influence of surface passivation and water contentnot directly measured, previous experimental results suggest it would drop from ~9.3 to ~7.6 during brucite carbonation (Harrison et al., 2013a; Chapter 2). However, gradients in pH and solute concentrations at the pore scale may lead to variation in saturation state within pores. In the medium brucite columns for instance, the quartz grains were generally not coated with carbonate (Fig. 5.6B), suggesting that the interfacial fluid in contact with quartz may have been undersaturated with respect to carbonate phases. Alternatively, the carbonates may favor nucleation on brucite grains compared to quartz. For example, Stockmann et al. (2014) demonstrate that carbonate precipitates preferentially nucleate on certain mineral substrates over others. In any case, without nucleation on quartz, the volume in which carbonate precipitates may form is further restricted. The incorporation of water in the crystal structure of hydrated Mg-carbonate phases means that there is a stoichiometric limit to the mass of CO2 that can be stored based on the amount of water available. For example, Schaef et al. (2011) demonstrate that the volume of water available to incorporate into nesquehonite dictates the extent of brucite carbonation in wet supercritical CO2. If all the brucite in the columns were converted to nesquehonite, two moles of water would be required per mole of brucite consumed (Eq. 5.6). Thus, 22 g (1.22 moles) of water are required to convert the initial 0.61 moles of brucite to nesquehonite. The initial volume of water added to the columns exceeded this limit for all experiments (Table 5.1). Moreover, in all but the 15% saturated medium brucite column, even the final water saturation is in excess of this value for the entire depth of the column (A3 Fig. A3.3).Mg(OH)2 + CO2 + 2H2O ↔ MgCO3·3H2O                              (Eq. 5.6) Thus, an insufficient supply of water can explain the lack of carbonation in the 15% saturated column, but cannot account for incomplete carbonation in the other columns. The distribution of water at the pore scale is likely heterogeneous and leaves some grains relatively dry, suggesting that the true reactive capacity may be lower than the theoretical stoichiometric 74Influence of surface passivation and water contentlimit. Still, the near complete consumption of brucite in the very fine brucite column with the same water content as the fine and medium brucite 35% saturated columns confirms that the lower extent of reaction cannot be attributed to insufficient water. The similar extent of reaction achieved in both the 35% and 50% saturated medium brucite columns, despite the difference in water volume is consistent with this conclusion. 5.4.2.2 Consumption of highly reactive sites and fine grainsAn alternative hypothesis to explain the incomplete reaction in these columns is that the net surface area of the brucite is progressively reduced due to the preferential consumption of the finest grains in each size fraction during the early stages of the reaction. Neglecting the powder coating, the geometric specific surface area of the finest particles within the medium brucite size fraction is within a factor of three of the coarsest grains, assuming spherical particles. This difference is too small to explain the effective shutdown of the reaction. It is also possible that the pulverization of the initial brucite induced defects at the mineral surface with higher reactivity than the underlying material (White and Brantley, 2003). As these highly reactive sites were consumed, the reaction rate would decline. For example, Petrovich (1981) shows that dissolution rates of ground quartz are higher during removal of a ~3 nm surface layer altered by grinding. However, more than 50% of the brucite is consumed prior to the stage 3 rate decline for brucite of all grain sizes; it is unlikely that reaction at these grinding induced reactive sites alone could consume such a large proportion of the brucite. Moreover, the >80% consumption of brucite in the very fine and fine columns implies that a significant proportion of coarser grained material in each size fraction is reacting along with the fines. Thus the preferential consumption of highly reactive sites and finer grains cannot explain the rapid shut off of the reaction prior to completion.75Influence of surface passivation and water content5.4.2.3 Surface passivation The surface passivation hypothesis is most consistent with the experimental data, for all but the 15% saturated, water-limited columns. Scanning electron micrographs clearly depict carbonate precipitates forming a rind (10s-100s of microns) surrounding an unreacted brucite core, and as pervasive coatings on mineral surfaces (Fig. 5.6B-F). However, the presence of a surface coating does not necessarily imply that the surface is passivated. Reports in the literature on the effect of surface coatings are highly variable. For example, experimental studies using stirred or agitated reactors found that precipitation of calcite did not inhibit dissolution of basaltic glass or diopside (Stockmann et al., 2011; Stockmann et al., 2013), and amorphous Fe-rich coatings did not affect anorthite dissolution (Hodson, 2003). On the other hand, the rate of pyrite oxidation is slowed by precipitation of oxide coatings (Nicholson et al., 1990), and the reaction of Ca-rich steel making slags and cement kiln dust with CO2 is inhibited by precipitation of calcium carbonate coatings (Lekakh et al., 2008; Huntzinger et al., 2009). Similarly, silica rich layers have been found to passivate the surfaces of olivine and serpentine, although the effect is lessened for basalt (Park and Fan, 2004; Béarat et al., 2006; Andreani et al., 2009; Daval et al., 2011; Sissmann et al., 2014; Johnson et al., 2014). Daval et al. (2009a) and Hövelmann et al. (2012a) show that precipitation of calcium carbonate and magnesite can inhibit reaction of wollastonite and olivine, respectively. Despite the disagreement in the literature as to the effect of secondary phases on dissolution, the extent of passivation is typically a function of the degree of coverage of the primary phase surface and the density and permeability of the secondary phases (e.g., Cubillas et al., 2005; Daval et al., 2009a; Daval et al., 2009b). For example, Daval et al. (2009b) show that porous silica layers are not passivating for wollastonite carbonation until the porosity of this layer becomes clogged with calcite precipitates. The degree of passivation was also dependent on the microstructure of the secondary calcite (Daval et al., 2009a). At circum-neutral pH, the precipitation of many small calcite crystals formed denser, more impermeable layers and greater passivation than at acidic pH. Similarly, Cubillas et al. (2005) found that 76Influence of surface passivation and water contentepitaxial growth of otavite [CdCO3] coating a calcite surface inhibited dissolution, but its formation as islands rather than layers on aragonite [CaCO3] surfaces did not slow dissolution. In the case of brucite, hydrated carbonates tend to favour upward growth of islands, but these islands may form a continuous layer over time (Hövelmann et al., 2012b), consistent with our observation of carbonate coatings. In our experiments, the degree to which these coatings are passivating may depend on which carbonate phase is dominant (i.e., the flakey poorly crystalline phase or crystalline nesquehonite). Velbel (1993) highlights the importance of the molar volume ratio of product to reactant for the extent of passivation, indicating that it must exceed 1 in order to completely passivate the surface. In any case, hydrated Mg-carbonates tend to have higher molar volume than brucite (e.g., brucite = 24 cm3 mol-1, nesquehonite = 75 cm3 mol-1), increasing the chances of passivation and pore clogging. SEM images of the flakey Mg-carbonate phase indicate that although it can completely coat brucite grains, it has a relatively porous structure that may not significantly inhibit transport (Fig. 5.6C). Nesquehonite, on the other hand, forms thick rinds that exceed 100 μm in places, with a low porosity, bladed microstructure (Fig. 5.6B, E-F). This low porosity coating is likely to substantially inhibit reaction.   The geometric surface area update model does not account for these effects, resulting in the poor fit with the experimental data (Fig. 5.5). We instead develop an empirical function that successfully reproduces the instantaneous carbonation rate through time (as represented by the CO2 breakthrough curves; Fig. 5.5). The proposed model relates the reactive surface area at a given time to the extent of conversion of brucite, allowing a relatively constant rate of reaction up to a threshold degree of conversion (A3 Fig. A3.5). At this threshold, the reaction rapidly shuts down with increasing brucite consumption. This function is henceforth referred to as the ‘threshold model’ (Eq. 5.7):k e f ft  = (k 0 SA) 1-0 -t0 - p5                                        (Eq. 5.7)77Influence of surface passivation and water contentthe equation is valid for φt≤ φp, where kteff is the effective threshold rate constant and φp is a threshold brucite volume fraction (i.e., conversion threshold) at which the reaction rate becomes negligible (stage 4). Here, kteff replaces k0eff in Equation 5.2. The value of φp was equal to the amount of brucite remaining in the lower half of each column, and the exponent was adjusted to provide the best fit with the experimental CO2 breakthrough curves. In order to fit the experimental data from all grain sizes, a value of 5 was required. The requirement of such a large exponent implies that the reaction shuts down rapidly when the conversion threshold is reached. The threshold model produces excellent agreement with the experimental CO2 breakthrough curves for all of the brucite grain sizes (Fig. 5.5A-C). A modest decrease in initial specific surface area equal to ~15% of the measured BET surface area for each grain size was employed. Although the threshold model is empirically derived, its form provides information as to the mechanism of reaction. Similar behavior has been documented for diffusivity in weathering rinds on basalt clasts: upon reaching a threshold porosity, the pore space becomes interconnected, and diffusivity increases considerably (Navarre-Sitchler et al., 2011). During brucite carbonation, the threshold may represent a sudden decrease in permeability of both the bulk porous media and the carbonate surface coatings due to the transformation from the poorly crystalline flakey Mg-carbonate phase to crystalline nesquehonite. The decrease in nesquehonite abundance along the flow path despite only moderate, if any, declines in total CO2 sequestered, indicates it may be forming via replacement of a metastable phase (Figs. 5.2-5.3). Such a reaction path is not uncommon in the Mg-carbonate system (e.g., Hänchen et al., 2008; Schaef et al., 2011; Felmy et al., 2012; Schaef et al., 2013a). Poorly crystalline phases are often precursors to more crystalline phases due to a lower mineral-solution interfacial energy (Steefel and Van Cappellen, 1990; Hellevang et al., 2013); the flakey Mg-carbonate may represent such a precursor. Thus, the shutdown of the reaction may relate to the conversion of the porous, poorly crystalline phase to low porosity nesquehonite. This is consistent with the lack of passivation in the very fine-grained column, as the rapid brucite dissolution rate may 78Influence of surface passivation and water contentfacilitate complete conversion to carbonate prior to the nesquehonite phase transformation. In the medium brucite columns, on the other hand, the slower brucite dissolution rate may cause simultaneous brucite dissolution and carbonate transformation, inhibiting dissolution prior to complete consumption. If this were the case, the overall reaction rate in the latter stages of carbonation would depend on the kinetics of the transition between the flakey phase and nesquehonite. It has recently been noted that the kinetics of carbonate precipitation may be an important control on the overall carbonation reaction (Pham et al., 2011; Gautier et al., 2014), but the effect of Mg-carbonate phase transitions on surface passivation have not previously been considered to our knowledge. Because the rate of transition between the carbonate phases is unknown, we cannot directly model this effect, and thus rely on the empirical function. However, the empirical formulation is able to describe the evolution of reactivity at the macroscale as a function of reaction progress and surface passivation, consistent with mineralogical observations. Detailed studies of the brucite-carbonate interface are required to further evaluate the effect of the phase transformation on the porosity of the surface coating and to develop a mechanistic model of surface passivation. Similar to the model of Navarre-Sitchler et al. (2011), an alternative explanation for the rapid shutdown in reaction, is that at a threshold brucite conversion, a sufficient volume of carbonate has formed, such that the porosity in the surface coating becomes disconnected and effectively shuts off access to the brucite surface. In either case, the effective cessation of reaction prior to complete conversion of the brucite is attributed to surface passivation, although the exact mechanism of passivation cannot be directly assessed based on the current results.5.5 Implications5.5.1 Reactive transport modeling Recent studies have highlighted the need for improved reactive transport models to predict the fate of CO2 during subsurface geologic storage. In particular, better constraints on secondary phase nucleation and precipitation kinetics are necessary (Pham et al., 2011; 79Influence of surface passivation and water contentPaukert et al., 2012; Hellevang et al., 2013; Galeczka et al., 2014; Gautier et al., 2014). Our experiments have highlighted the importance of phase transitions in the Mg-carbonate system for surface passivation and pore space clogging, a potentially important effect during subsurface injection of CO2 into Mg-bearing rock, and formation of natural, serpentinite-hosted magnesite deposits (Boschi et al., 2009). These same mechanisms may inhibit carbon sequestration in alkaline industrial wastes. While a mechanistic model for passivation remains elusive, the excellent agreement between modeled and experimental results indicates that thresholding processes represent real constraints on carbon sequestration efficiency. The mechanism behind the passivating effects requires further study in order to develop a more generally applicable mechanistic model. Conventional geometric models alone do not adequately reproduce reaction progress during brucite carbonation, hindering the ability to predict the fate of CO2 injected in the subsurface or alkaline waste storage facilities. This also has implications for modeling of reaction progress in natural systems that involve coupled dissolution-precipitation reactions, as the reactive surface area may not evolve in a straightforward manner. For example, the porosity and morphology changes associated with phase transformations may be relevant for other mineral systems. In particular, Mg-sulfate, zeolite, and smectite minerals may undergo transitions between different hydration states at low temperatures, with relevance for bioavailability of water and nutrients in dry environments (e.g., Bish et al., 2003; Wilson and Bish, 2012). In unsaturated porous media, the effective reactive surface area may be reduced due to the decreased exposure to the reactive fluid (White and Brantley, 2003 and references therein). Experimental evidence indicates that inclusion of water-limited reaction in terms of its effect on the reactive capacity, rather than just the rate, may also be required for modeling reactions in unsaturated porous media involving hydrated mineral phases.  5.5.2 CO2 sequestrationSurface passivation and water limitations are potentially important constraints for carbonation efficiency of mafic and ultramafic materials, although physical and chemical 80Influence of surface passivation and water contentheterogeneities and more complex chemistry limit direct extrapolation from the experiments to field scale. Injection of supercritical CO2 in the subsurface displaces formation water, resulting in a potentially water-limited reaction zone, and phase transitions between Mg-carbonates are common under these conditions (Loring et al., 2011; Schaef et al., 2011; Van Pham et al., 2012; Schaef et al., 2013a; Miller et al., 2013). Phase transformations in the Mg-carbonate systems may also have implications for formation of natural Mg-carbonate deposits. For example, Boschi et al. (2009) postulate that hydrous carbonate may have formed as a precursor to magnesite in serpentinite-hosted veins in Tuscany, Italy. Surface passivation may be less significant during carbonation of alkaline industrial wastes, however, which tend to be fine-grained. The grain size of mine tailings, for example, typically falls in the size range of the very fine and fine brucite employed in the experiments (e.g., Power et al., 2011b). The passivating effects of other Mg-carbonates than formed in this study (e.g., hydromagnesite, dypingite) may differ, and requires further investigation. The availability and distribution of water will play an important role controlling both the rate and extent of carbonation of mine tailings (Assima et al., 2013a), and other near surface carbon mineralization strategies that involve reaction in porous media such as engineered urban soils (Renforth et al., 2009; Washbourne et al., 2012). Low water contents will inhibit dissolution-precipitation processes, while high water content may effectively block gas transport (e.g., Collin and Rasmuson, 1988), promoting the development of preferential flow paths and reducing efficiency. If gaseous CO2 were supplied to variably saturated porous media to enhance carbonation (Harrison et al., 2013b; Chapter 3; Harrison et al., 2013a; Chapter 2; Wilson et al., 2014; Chapter 7), higher water contents would require greater pressure to maintain gas flow, increasing energy costs (Fig. 5.9). The consumption of water as reaction progresses may also provide an interesting feedback on the relative permeability of the porous media; an area that requires further study.  81Influence of surface passivation and water content0 20 40 60 80 100020406080100Water saturation (%)Extent of brucite carbonation and reduction in CO2 ux (%)ideal water contentReduction in CO2 uxthis studyAssima et al. (2013)(1 day reaction)Extent of brucite carbonationFigure 5.9. The extent of brucite carbonation measured in this study and by Assima et al. (2013a) and the calculated percent reduction in gaseous CO2 flux through a porous medium at a fixed pressure as a function of various degrees of water saturation. Circles represent data from 10-15% saturated experiments, squares represent 35% saturated experiments, triangles represent 50% saturated experiments, and diamonds represent 100% saturated experiments. The shaded region represents the approximate range of water saturation that may both optimize the extent of carbonation and minimize energy costs associated with increased CO2 injection pressure were gaseous CO2 supplied to a porous medium as part of an industrial CO2 sequestration project (e.g., Wilson et al., 2014; Chapter 7). The experiments conducted by Assima et al. (2013a) investigated carbonation of chrysotile mining residues that contained brucite using 14 vol.% CO2 with an experimental duration of 1 day. The carbonation extent of brucite was estimated assuming that 93% of the CO2 sequestration recorded was due to brucite carbonation (after Assima et al., 2013a). The CO2 flux was calculated using Darcy’s law for a 10 vol.% CO2 gas mixture (balance N2) at a constant pressure of 1 atm, assuming transport is dominated by advection (refer to A3 for details regarding the calculation).Physical and chemical influences of water on mineral carbonation826. Physical	and	chemical	influences	of	water	on	mineral	carbonation in variably saturated porous media: Implications for CO2 sequestration56.1 IntroductionMineral weathering in the unsaturated zone is an important control on nutrient availability, contaminant transport, and the carbon cycle (e.g., McKinley et al., 2006; Manning, 2008; Maher et al., 2009). The rate at which these reactions occur depends on the available surface area in contact with a reactive fluid (e.g., Helgeson et al., 1984; Brantley and Mellot, 2000). The available surface area is dependent on both the physical and chemical properties of the mineral, and environmental conditions such as water saturation (White and Brantley, 2003). In the unsaturated zone, quantification of reactive surface area is complicated by several factors, including the occurrence of dry pores and poorly connected water pathways, and secondary mineral precipitation that can dictate solution chemistry and passivate reactive surfaces (Pačes, 1983; White and Brantley, 2003; Brantley, 2008; Maher et al., 2009; Scislewski and Zuddas, 2010). Moreover, the effective reactive surface area may change over time due to consumption of the reactive phase, preferential dissolution of defects and highly reactive sites, changes in surface morphology, and formation of surface coatings (Chapter 5; Daval et al., 2009a; Helgeson et al., 1984; Maher et al., 2006; Petrovich, 1981; Scislewski and Zuddas, 2010; White and Brantley, 2003). Environmental factors are also highly variable over time in natural systems; water saturation in the shallow subsurface is altered in response to short-term weather events and longer-term climate forcings that may change the amount of surface area in contact with water, and the development of preferential flow paths may decrease exposure to reactants. Low water saturation may also reduce the reactive capacity of a porous medium, defined as the 5A version of this chapter will be submitted for publication as:Harrison, A.L., Dipple, G.M., Mayer, K.U. and Power, I.M. Physical and chemical influences of water on mineral carbonation in variably saturated porous media: Implications for CO2 sequestration.Physical and chemical influences of water on mineral carbonation83maximum extent of reaction possible (e.g., Dipple, 1995), due to a lack of water available to facilitate dissolution-precipitation reactions and incorporate into secondary hydrous minerals (Chapter 5; Assima et al., 2013a; Felmy et al., 2012; Loring et al., 2011; Miller et al., 2013; Schaef et al., 2013a, 2011; Thompson et al., 2013). Here, we aim to ascertain the impact of water saturation on the availability of reactive surface area and total reactive capacity. Experiments demonstrate that reactive surface area is not influenced by water saturation alone, therefore we use reactive transport modeling to elucidate the evolution of reactive surface area due also to changes in physical properties during carbonation of brucite [Mg(OH)2] in partially water saturated meter-scale column reactors. The reaction of brucite and other alkaline earth metal-bearing hydroxide and silicate minerals with CO2 to form carbonate minerals is known as mineral carbonation (e.g., Lackner et al., 1995). Mineral carbonation reactions are model systems with which to study reactive surface area evolution and water limited reaction, due to the coupled nature of the reaction and relatively high extent of reaction that can be achieved on experimental time scales (Chapter 5). As a natural weathering reaction, mineral carbonation is an important part of the global carbon cycle and regulates atmospheric CO2 concentrations over geologic time (Berner et al., 1983). These reactions may also be harnessed as engineered CO2 sequestration strategies, for instance by injection of CO2-rich gases or fluids in the subsurface or alkaline waste stockpiles (McGrail et al., 2006; Kelemen and Matter, 2008; Gislason et al., 2010; Bobicki et al., 2012; Power et al., 2013b), or direct air capture via enhanced weathering of pulverized rock or wastes (Wilson et al., 2006; Renforth et al., 2009; Schuiling and Boer, 2010; Pronost et al., 2011; Renforth et al., 2011; Assima et al., 2012; Washbourne et al., 2012; Assima et al., 2013a; Assima et al., 2013b; Wilson et al., 2014; Chapter 7). Brucite carbonation is of specific interest for CO2 sequestration (e.g., Zhao et al., 2010; Hövelmann et al., 2012b; Nduagu et al., 2013; Fricker and Park, 2013), particularly as a component of ultramafic mine wastes that can be exploited for its high reactivity (Chapter 5; Pronost et al., 2011; Assima et al., 2012; Bea et al., 2012; Beinlich and Austrheim, 2012; Assima et al., 2013a; Assima et al. 2013b, Harrison et al., 2013a; Chapter Physical and chemical influences of water on mineral carbonation842; Harrison et al., 2013b; Chapter 3; Wilson et al., 2014; Chapter 7) . The study of brucite carbonation therefore allows assessment of fundamental controls on reactivity during coupled dissolution-precipitation reactions in the unsaturated zone, with important implications for offsetting anthropogenic greenhouse gas emissions responsible for global climate change. 6.2 Methods6.2.1 Experimental designColumn experiments were used to investigate reactive surface area evolution during dissolution-precipitation reactions in the unsaturated zone. In order to examine the impacts of water saturation on rates, extent, and distribution of reaction, experiments with heterogeneous water content and different initial water saturation profiles were performed. A total of three experiments were conducted in 25 cm diameter × 90 cm tall acrylic columns that contained 10 wt.% pulverized brucite ore and 90 wt.% quartz sand (Fig. 6.1). Of these, two were duplicates with 35% bulk water saturation and one had 60% bulk water saturation. The duplicate experiments are henceforth referred to as “35% 1 and 2.” The brucite ore was obtained from Premier Magnesia LLC and was pulverized using a hammer mill to between 250 and 500 μm in diameter (mass weighted median ≈ 188 μm as estimated based on sieving; Appendix 4 (A4) Fig. A4.1). The quartz sand was a product of Lane Mountain Materials (“LM 50”) that had been sieved to between 53 and 425 μm. The initial major oxide composition of the brucite ore and quartz sand was determined using X-ray fluorescence spectroscopy (XRF; refer to Appendix 4 for details). XRF measurements indicated that the oxides present in the brucite ore at ≥ 1.00% ± 1σ abundance were: MgO (60.16 ± 0.43%), SiO2 (2.71 ± 0.04%), and CaO (2.04 ± 0.03%), with 34.29 ± 0.43% loss on ignition. The quartz sand was 99.10 ± 0.28% SiO2, with the remainder consisting primarily of Al2O3. Rietveld refinement of X-ray diffraction (XRD) data from analysis of the quartz sand indicated it was nearly 100% pure with trace mica (biotite or muscovite at ≤ 1.0 wt.% abundance). Analysis of triplicate samples of the brucite ore indicated it contained 78.8 ± 3.8 wt.% brucite, Physical and chemical influences of water on mineral carbonation855.5 ± 0.4 wt.% dolomite, 1.9 ± 0.3 wt.% magnesite, 7.4 ± 1.0 wt.% hydromagnesite, and <0.5 wt.% lizardite and pyroaurite. The remainder was amorphous content. The surface area of the brucite ore was determined on duplicate samples using BET with N2 adsorption, and was equal to 2.4 ± 0.7 m2 g-1. Mixtures of 10 wt.% brucite ore and 90 wt.% quartz were prepared by mechanically mixing 4.5 kg of brucite ore, and 40.5 kg of quartz sand (total mass = 45.0 kg). The mixtures were wetted with approximately half of the total desired water volume prior to being loaded into the column. The added water had 0.1 M MgCl2 and <9.0 × 10-9 M dissolved inorganic carbon (DIC). The partially wetted mixtures were loaded into the columns and compacted by manual compression under a flat surface (height of porous medium = 72 cm). After loading, additional MgCl2 solution was added to the surface to reach the desired saturation using a watering can to minimize surface disturbance. The water was allowed to infiltrate under the force of gravity. The porosity was approximately 0.52-0.54 for all columns, as estimated based on the bulk density of the solids and the volume of the porous medium.The columns were continuously flushed with N2 gas at 220 ± 25 mL min-1 overnight to remove air with atmospheric composition from the pore space prior to initiating the supply of CO2, and to allow time for the water distribution to equilibrate. The CO2 supply was begun without interrupting the flow of N2, such that the total gas flow rate was 250 ± 25 mL min-1 and the supplied gas was 10 vol.% CO2. The gas entered through a hose barb into a 4 cm tall gas filled chamber at the base of the column. The chamber was capped with a porous ceramic plate overlain by a 2 cm thick gravel layer and a sheet of fabric mesh (300 mesh) to promote homogeneous distribution of the gas into the brucite/quartz medium (Fig. 6.1). The flow rate of the gas effluent was measured at least twice a day using a Cole-Parmer® direct reading acrylic flow meter to ensure the gas supply rate was consistent. The top of the column was sealed with a gas-tight acrylic lid, and the gas effluent exited through a single 6.35 mm outer diameter tube (Fig. 6.1). Sensors and sampling ports were installed at three points along the column length at Physical and chemical influences of water on mineral carbonation861) the base (5 cm above the gravel), 2) the middle (35.5 cm above the gravel), and 3) the top (65.5 cm above the gravel). Each port consisted of a Vaisala® GMT221 CO2 concentration sensor (measurement error ± 0.5% CO2), a Decagon 5TM soil moisture and temperature probe (measurement error ± 1°C, ± 2% volumetric water content), tubing from which to extract gas samples, a SoilMoisture micro pore water sampler, and a Mininert® valve to allow pressure measurements (Fig. 6.1). A Mininert® valve was also located between each sampling port, for a total of five measurement points. The data from the CO2 and soil moisture and temperature probes was recorded every 5 or 10 minutes using National Instruments® LabVIEW™ 8.6 software (National Instruments, 2008).  Water and gas samples were removed every 2-8 hours for at least the first 24 hours, after which the sampling interval was lowered to once a day or once every two days as the reaction 0.9 mGas exit0.25 m10% CO 2(g)Pressure measurement valveGas sample portVWC/temp sensorCO 2 sensorPore water samplerPorous plate/gravel/fabric meshSafety valve71.266 kgScalerecorded with LabViewFigure 6.1. Schematic of experimental apparatus. Physical and chemical influences of water on mineral carbonation87slowed. Pressure was measured manually before each sampling time using a Dwyer Mark® II plastic manometer (measurement error ± 2.290 × 10-4 atm). Gas samples were collected for measurement of the stable carbon isotopic composition (δ13C) of the CO2 using a Los Gatos Research instrument for laser spectroscopy (Barker et al., 2011). All δ13C values are reported in δ-notation relative to Vienna Pee Dee Belemnite (VPDB) in units of per mil (‰). Water samples were collected for determination of pH, Mg and DIC concentration, and the δ13CDIC. All water samples were filtered through 0.22 μm syringe filters, and cation samples were acidified to 2% ultrapure HNO3. The solution pH was measured immediately after sampling using a portable Thermo Scientific® Orion 4-Star pH/ISE meter. DIC and δ13CDIC data were unavailable from 35% 1 due to the formation of precipitates while in storage. To avoid this issue for 35% 2 and the 60% saturated column experiments, DIC and δ13CDIC samples were diluted by 10-20 times with deionized water (<9.0 × 10-9 M DIC) immediately following sampling to prevent precipitation of carbonate minerals prior to analysis. All water samples were stored in vials with no headspace to prevent gas exchange and were kept at ~4°C prior to analysis.  The duration of the experiments varied between 430 and 504 h depending on reaction progress. Following completion of each experiment the entirety of the solid material was removed in 10 cm intervals for analysis of the mineralogical composition, the total solid phase CO2 content, and the δ13C of the carbonate minerals. The major oxide composition and volatile content of the solids was also determined for 35% 1 and the 60% saturated experiment using XRF spectroscopy. For at least three of the depth intervals in each column, triplicate discrete samples were taken to assess heterogeneity in reaction extent and mineralogy. The remaining intervals were mechanically homogenized and an aliquot of the bulk material was removed for analysis. Experimental products were examined using scanning electron microscopy (SEM) in thin section and as disaggregated particles on conventional aluminum SEM stubs. To estimate the water saturation with depth, aliquots of known volume from each depth interval were weighed immediately after sampling. The aliquots were then reweighed following drying at 70°C, and the difference in mass was equated to the mass of water present in the wet samples. Physical and chemical influences of water on mineral carbonation88Total water content at the end of the experiments was estimated for 35% 2 and the 60% saturated column experiments by collection, drying, and weighing of all reaction products. 6.2.2 Assessment of the rate and extent of carbonationThe extent of reaction was estimated using several lines of evidence, including (1) the total carbon content in the solid phase (%CO2), (2) the mass gain of the columns during the course of the experiment, (3) the abundance of mineral phases, and (4) the reduction of CO2 content in the gas phase. The %CO2 and mineral abundance data provided information as to heterogeneity in reaction extent and mineralogical distribution, whereas the gravimetric and gas phase CO2 measurements were used to estimate the total CO2 sequestered, and the latter to determine the rate of carbonation. The total carbon content for all experiments is expressed as %CO2 by mass and was measured using coulometry (refer to A4). The initial brucite/quartz mixtures had an average of 0.65% CO2 that was contained in dolomite, hydromagnesite, and magnesite present in the initial brucite ore. This initial mass of CO2 was subtracted from CO2 content values measured for reacted samples to determine the mass of CO2 gained. This was justified by the presence of approximately the same mass of dolomite in the reaction products as in the initial material as determined using XRD, although the data were insufficient to resolve changes in hydromagnesite and magnesite abundances. In order to estimate the CO2 gain gravimetrically, the columns were placed on a scale with ± 20 g accuracy and their mass was recorded three to four times a day. The abundance of secondary carbonate phases and reduction in brucite content was determined using Rietveld refinement of XRD data. A known mass of highly crystalline fluorite [CaF2] was added to solid aliquots to assess whether CO2 was also stored in X-ray amorphous carbonate phases (e.g., Gualtieri, 2000). Finally, because the flux of CO2 into the column was known, the mass of CO2 sequestered and carbonation rate were also determined based on the reduction in this flux along the column length as measured using the CO2 concentration sensors. The CO2 breakthrough curves measured with the sensors are expressed as ‘C/C0,’ the ratio of the CO2 concentration in the gas effluent at a given time to Physical and chemical influences of water on mineral carbonation89the concentration of CO2 in the supplied gas. The instantaneous rate of carbonation (rinstant in g CO2 h-1) was determined as follows (Eq. 6.1):q i n  %CO 2 i n100    PRT  q i n1%CO 2 i n1001%CO 2 o u t100      %CO 2 o u t100  PRT r i n s t a n t = n i n  n o u t MCO 2 =  (Eq. 6.1)where MCO2 is the molar mass of CO2 (g mol-1), nin and nout are the molar rate of entrance and exit of CO2 (mol s-1), respectively, %CO2in and %CO2out are the CO2 concentration of the supplied gas and gas effluent, respectively (%), qin is the total rate of supply of gas to the column (L s-1), P is the gas phase pressure (atm), R is the ideal gas constant (L atm mol-1 K-1), and T is the temperature (K). The total CO2 sequestered is the integration of the instantaneous rates throughout the experimental duration. For further detail regarding the experimental setup and analytical techniques, refer to Appendix 4. 6.2.3 Geochemical and reactive transport modelingThe variation in mineral saturation indices with reaction progress was calculated based on the measured aqueous chemistry data using PHREEQC V.3 (Parkhurst and Appelo, 2013) with the Pitzer database and mineral solubility data from the Minteq database. Reaction progress under the experimental conditions was simulated using the multicomponent reactive transport code MIN3P-DUSTY (Mayer et al., 2002; Molins and Mayer, 2007) to elucidate reaction mechanisms and evaluate conceptual models for water-limited reaction and surface area updates. MIN3P-DUSTY comprises a suite of chemical reactions including mineral dissolution-precipitation, and can model flow and transport in both the gaseous and aqueous phases. Gas transport was modeled as an advective-diffusive process according to Darcy’s law and the Dusty Gas model (Mason and Malinauskas, 1983; Molins and Mayer, 2007). For a complete description of the constitutive equations refer to Mayer et al. (2002) and Molins Physical and chemical influences of water on mineral carbonation90and Mayer (2007). Brucite dissolution kinetics were modeled using a far-from-equilibrium HCO3- concentration-dependent kinetic dissolution rate law based on data from Pokrovsky et al. (2005b) and Pokrovsky and Schott (2004) (Eqs. 6.2 and 6.3): r b r c  = k e f f0 HCO 3- 0 . 5 6 (1  2 )                                      (Eq. 6.2)k e f f  0 = k 0 SA                                                    (Eq. 6.3) where rbrc is the brucite dissolution rate (mol s-1), k0eff is the effective rate constant, k0 is the initial reaction rate constant equal to 10-6.13 L s-1 m-2, SA is the brucite surface area (m2), and Ω is the saturation ratio. Saturation ratio is defined as the ratio of the ion activity product to the equilibrium constant.The identity and rate of formation of secondary phases were selected based on experimental results. The two precipitates were an X-ray amorphous phase and nesquehonite [MgCO3·3H2O], the former having stoichiometry and solubility similar to artinite (refer to Results). Precipitation of artinite [Mg2CO3(OH)2·3H2O] was modeled effectively as an equilibrium process, but nesquehonite precipitation was slowed such that precipitation occurred at supersaturated conditions, as was consistent with experimental data. Porosity was set to a representative value of 0.52 for all simulations, and soil hydraulic function parameters were estimated based on results of Tempe cell tests of the solid materials (e.g., Fredlund and Rahardjo, 1993) and measured water saturation profiles. Model output was calibrated against both the measured CO2 breakthrough curves and mineralogy profiles. For a complete list of transport parameters used in the simulations and results of Tempe cell tests, refer to Appendix 4 (A4 Fig. A4.2; A4 Tables A4.2 and A4.3).  Physical and chemical influences of water on mineral carbonation916.3 Results6.3.1 Rate and extent of carbonationThe CO2 concentration in the gas phase (CO2 breakthrough curves) varied significantly with depth in all experiments (Fig. 6.2). Similar trends were documented in the duplicate experiments, although a slightly higher CO2 concentration was measured by the middle sensor in 35% 2 during the first ~100 hours (Fig. 6.2), which may indicate the presence of preferential flow paths. Consistent with our previous work for carbonation of brucite in smaller scale columns (Chapter 5), four distinct carbonation stages were evident in the breakthrough curves (Fig. 6.2). Stage 1 represents rapid reaction of brucite powder coating the brucite grains that was an artifact of the pulverization process (A4 Fig. A4.3). Stage 2 is the reaction of the bulk of the brucite. Stage 3 comprises a rapid decline in reaction rate due to loss of reactive brucite. Finally, stage 4 is when the reaction rate becomes negligible because any remaining brucite is effectively inaccessible for reaction. However, during stage 4 the CO2 concentration did not return to its initial value (C/C0 <1), implying that minor sequestration was still occurring. Diurnal fluctuations in pCO2 measured at the top of the columns during stage 4 were coincident with temperature variations in the laboratory (A4 Fig. A4.4A). The experimental apparatus was not insulated, thus the pore water temperature experienced diurnal fluctuations in response to variations in the laboratory air temperature (A4 Fig. A4.4A). This suggests that exothermic effects from the carbonation reaction were insignificant in comparison to changes in ambient laboratory temperature. Substantial differences were evident between the breakthrough curves of the 35% and 60% saturated columns (Fig. 6.2). In the 60% saturated column, the CO2 concentration measured at the middle sensor was virtually identical to that measured at the base. This early breakthrough is indicative of transport of CO2 along preferential flow paths. Similarly, CO2 reached the top of the column at measurable concentrations earlier in the 60% saturated column, at ~16 h compared to ~80 h in the 35% saturated columns. The CO2 breakthrough at the top of the columns was used to calculate the bulk rate of Physical and chemical influences of water on mineral carbonation92carbonation over time. Carbonation rates in Figure 6.3 are normalized to the CO2 supply rate for ease of comparison, as slight differences (10% relative) in gas flow rate were documented between experiments. These calculations revealed that the rate of carbonation was comparable between the duplicate 35% saturated columns, and these rates exceeded that in the 60% saturated column for the majority of the experimental duration (Fig. 6.3). The extent of carbonation in each experiment as calculated using the reduction of CO2 content in the gas phase and the total mass gain during the course of the experiments agreed fairly well (Fig. 6.3), except for 35% 1, for which the gravimetric estimate was ~17% lower than indicated by the gas phase measurements. As the gravimetric measurements were not corrected for evaporation due to difficulty controlling and measuring evaporation rates between experiments, these values represent the minimum mass of CO2 gained. This could account for the discrepancy between the two types of measurements, and would imply that the evaporation was highest for 35% 1. Gravimetric measurements indicated that 528 g, 566 g, and 476 g CO2 breakthrough (C/C0)0 100 200 300 400 50000.20.40.60.81.0CO2 breakthrough (C/C0)Time (h)S2 S3S2S3A) 35% saturated columns B) 60% saturated columnbasemiddletop35% 1basemiddletop35% 2basemiddletopS4S1Time (h)S4S200.20.40.60.81.00 100 200 300 400 500Figure 6.2. CO2 breakthrough curves at three positions along the flow path plotted as C/C0 (ratio of the measured CO2 concentration at the measurement port to the CO2 concentration in the supplied gas) versus time. Black, red, and green lines show the CO2 breakthrough curves measured at the base, middle, and top of the 35% saturated columns (A) and the 60% saturated column (B). “SX” labels denote reaction stages 1 through 4. Physical and chemical influences of water on mineral carbonation93were gained in 35% 1 and 2, and the 60% saturated experiments, respectively. Gas phase CO2 calculations indicated that 635 g, 587 g, and 469 g of CO2 were gained in 35% 1 and 2, and the 60% saturated experiment, respectively. Both datasets indicated that a lower amount of CO2 was sequestered in the 60% saturated column. 6.3.2 Characterization of precipitates6.3.2.1 Composition and distribution of precipitates In all experiments, brucite was replaced by nesquehonite and an X-ray amorphous Mg-carbonate phase (Fig. 6.4; A4 Fig. A4.5). A minor amount of lansfordite [MgCO3·5H2O] was also documented in 35% 1 (Fig. 6.4C). Only total amorphous content can be calculated using Rietveld refinement, therefore quantified abundances of the X-ray amorphous Mg-carbonate include any amorphous content as other phases. The XRD quantified abundance is not used to calculate the mass of this phase but allows assessment of the general trends in abundance. Nesquehonite was not resolved above ~35 cm depth, but %CO2 measurements indicate that CO2 was still sequestered (Fig. 6.4C and D). This confirms that the X-ray amorphous phase is 100 0 100 200 300 400 500 600 700 200 300 400 500 Mass CO2 (g) Time (h) 0 0.2 0.4 0.6 0.8 0 100 200 300 400 500 Carbonation rate/CO2 supply rateTime (h) 35% 135% 260%Gravimetric dataCO 2  fluxA) CO 2 sequestered B) Carbonation rate0 1.0 35% 135% 260%Figure 6.3. Mass of CO2 sequestered (A) and instantaneous carbonation rate (B) versus time. Black, grey, and blue lines represent gravimetric measurements of 35% 1, 2, and the 60% saturated columns, respectively. Diamonds in (A) show cumulative CO2 mass sequestered based on CO2 flux analysis.Physical and chemical influences of water on mineral carbonation94Depth (cm) 0 10 20 30 40 50 60 70 A) bruciteAbundance (wt.%) B) amorphous C) nesquehonite D) %CO 20 2 4 6 8lansfordite0.0 0.2 0.4 VWC60%35% 135% 2VWC35% 135% 260%bulk /unknown reaction finger35% 260%“dead zone”60%0 2 4 6 8 0 2 4 6 8 0 2 4 6 8Figure 6.4. Mineral and solid phase CO2 abundance profiles. A) Brucite abundance versus depth. B) Amorphous content versus depth. C) Nesquehonite abundance versus depth from all columns, and lansfordite abundance versus depth in 35% 1. D) Total solid phase CO2 content in the experimental products that is contained in secondary carbonate phases and final volumetric water content (VWC) versus depth. Black and grey squares represent 35% 1 and 2, respectively. Blue diamonds represent the 60% saturated column experiment. Black, grey, and blue lines represent the VWC of 35% 1, 2, and the 60% saturated experiment, respectively. Black ‘Xs’ in (C) represent lansfordite abundance in 35% 1; lansfordite was not documented in the other experiments. Orange filled symbols represent samples taken from reaction fingers, and white filled symbols represent samples taken from “dead zones” outside fingers. All others represent bulk samples or samples for which it was not possible to ascertain visually whether they were part of a reaction finger.Physical and chemical influences of water on mineral carbonation95also a carbonate, in agreement with previous experimental results (Chapter 5). The abundance of both Mg-carbonates tended to decrease with distance along the flow path in all experiments, although significant heterogeneity in the extent of carbonate precipitation was documented in triplicate samples from the same depth intervals (Fig. 6.4B and C). The greatest extent of heterogeneity for both the carbonate minerals and total %CO2 was documented in the 60% saturated column (Fig. 6.4). Due to the high degree of heterogeneity in the extent of carbonation, these data were not used to calculate the total mass of CO2 sequestered. Nowhere was brucite completely consumed, but the abundance typically increased along the flow path in parallel to the decreased %CO2 sequestered (Fig. 6.4A). An orange coloration in reaction products was found to correlate with a greater extent of carbonation (A4 Fig. A4.6).  Precipitation of hydrated Mg-carbonates is expected under the experimental conditions (e.g., Hänchen et al., 2008). The stoichiometry of the X-ray amorphous phase was estimated by mass balance with XRF, XRD, and %CO2 data for samples from 35% 1 and the 60% saturated columns (refer to Appendix 4). Only samples without measurable nesquehonite were used for the calculations, as the stoichiometry of precipitated nesquehonite may not be ideal, thereby introducing additional error into the stoichiometry calculation. The average Mg:C and H2O:Mg ratios calculated for the X-ray amorphous phase in 35% 1 were 2.2 and 1.4, respectively. The average Mg:C and H2O:Mg ratios calculated for this phase in the 60% saturated column were 1.9 and 1.6, respectively. The estimated chemical formula of the X-ray amorphous phase is thus postulated to be Mg2CO3(OH)2·~2H2O. This is similar to the composition of the known hydrated Mg-carbonate phase, artinite [Mg2CO3(OH)2·3H2O], and is consistent with the stoichiometry estimated for a phase formed under comparable conditions in smaller scale brucite carbonation column experiments (Chapter 5). We refer to this precipitate as pseudo-artinite.Physical and chemical influences of water on mineral carbonation966.3.2.2 Qualitative characterization of precipitates Brucite ore was coated with a fine brucite powder that was an artifact of the crushing process (A4 Fig. A4.3). Precipitates of two morphologies were observed in all columns; long, narrow crystals with morphology typical of nesquehonite (e.g., Chapter 5; Power et al., 2007; Schaef et al., 2011; Assima et al., 2013a), and a fine, flakey material coating surfaces (Fig. 6.5A). Energy dispersive spectroscopy confirmed that both of these phases consisted of Mg, C, and O (A4 Fig. A4.7). The morphology of the flakey material is similar to that observed for an X-ray amorphous carbonate precipitated under similar experimental conditions (Chapter 5). The flakey material is abundant in samples known to be rich in pseudo-artinite, to which we attribute this morphology. Nesquehonite needles were observed to rim pseudo-artinite (Fig. 6.5A). The distribution of these two reaction products changed significantly with distance along the flow path. Near the base of the column, where water contents were high, large crystals of nesquehonite or lansfordite were observed to form 10-100s μm thick rinds surrounding partially reacted brucite grains (Fig. 6.5E and F). Precipitates were also documented to form pervasive cements that extended across several pore bodies even where brucite grains were lacking (Fig. 6.5E). Brucite surfaces are rough, with pieces lifted off the surface in some cases and thin pieces of brucite isolated between carbonate precipitates (Fig. 6.5G and H). Cracks filled with carbonate precipitates were also observed to extend towards the center of some brucite grains (Fig. 6.5B). Further along the flow path in the relatively dry upper portion of the columns, precipitates instead formed as thin (~5 μm) coatings of pseudo-artinite surrounding both brucite and quartz grains (Fig. 6.5C and D). Consistent with pore scale observations of brucite carbonation (Chapter 4), carbonate precipitates in thin section mimicked the curvature of the water meniscus at the gas-water interface, as approximated by visible curvatures in the epoxy-mineral interface (Fig. 6.5C). The local water saturation can therefore be estimated based on the location of these interfaces. At low water saturation, carbonate precipitates are restricted to narrow regions where water films were present (Fig. 6.5C and D). The flakey carbonates were also observed to line the “gas-water interfaces,” but were typically not present Physical and chemical influences of water on mineral carbonation97ȝPApseudo-artinitensqȝPbrucitequartzepoxycarbonateȝPbrc qtznsq/lnsȝPquartzbruciteȝPbrccarbonateqtzȝPbrcnsqȝPcarbonatebrcȝPbrcqtznsq/lnsBC DcarbonateEnsq/lnsFG HFigure 6.5. Representative scanning electron micrographs of experimental products (see following page for caption). Physical and chemical influences of water on mineral carbonation98Figure 6.5 (continued). Representative scanning electron micrographs of experimental products. A) Secondary electron micrograph showing flakey pseudo-artinite with nesquehonite (nsq) needles on top from 30-40 cm depth in the 60% saturated column (~30% saturation). Images B through H are backscattered electron micrographs. B) A brucite grain with carbonate coating and channels infilled with carbonate that extend into the interior of the grain from 60-70 cm depth in the 60% saturated column (>80% saturation). C) Flakey carbonate coating (~5 μm thick; red arrows) surrounding brucite and quartz particles and lining the location of the gas-water interface (right half of micrograph) in an area with locally low water content (~10% saturation) from 20-30 cm depth in 35% 1. D) Close up image of ~5 μm thick flakey carbonate coating on brucite and quartz particles from 20-30 cm depth in 35% 1. E) Thick, dense carbonate rind (>100 μm) extending from a partially carbonated brucite particle in an area with locally high water content (~90% saturation) from 65-70 cm depth in 35% 1. F) Close up image of (E). G) The edge of a brucite grain with pieces “lifted” from the surface and encrusted in precipitated carbonate from 65-70 cm depth in 35% 1 (~90% saturation). H) Thin remnants of brucite interspersed with carbonate precipitates in an extensively carbonated area from 60-70 cm depth in the 60% saturated column (>80% saturation). Lighter grey areas are brucite, darker grey is secondary carbonate, and black is epoxy.in the middle of larger pore bodies (Fig. 6.5C), in contrast to the wetter areas of the column wherein carbonate precipitates were present throughout the pore space (Fig. 6.5E). 6.3.3 Trends in aqueous chemistry and mineral saturation states Rapid uptake of CO2 into solution caused a spike in DIC concentration in the base of 35% 2, and the base and middle of the 60% saturated column (Fig. 6.6A) early in the experiments. Within 50 h these peak concentrations declined (stage 2), and approached 0.15-0.16 M in 35% 2 (Fig. 6.6A). In the middle of 35% 2, the DIC did not exhibit a peak at early time, but reached 0.16 M within 16 h after which it remained relatively constant (stages 2-4; Fig. 6.6A). In the 60% saturated column, on the other hand, DIC concentrations were highest in the middle of the column, fluctuating between 0.19 and 0.28 M (Fig. 6.6A). At the base of the column, after the transient peak at early time, the DIC slowly increased throughout the experiment, and did not reach a stable value (stage 2; Fig. 6.6A). This indicates that the middle of the column was receiving an anomalously high CO2 flux, as is consistent with the Physical and chemical influences of water on mineral carbonation99early CO2 breakthrough measured in the gas phase. Uptake of CO2 into solution drives brucite dissolution (e.g., Pokrovsky et al., 2005b; Zhao et al., 2010; Hövelmann et al., 2012b; Harrison et al., 2013a; Chapter 2), and not surprisingly, Mg concentrations were observed to follow a similar trend as DIC concentrations (Fig. 6.6C). Rapid initial increases within 24 h were followed by a gradual reduction (stage 1-3). Mg concentrations stabilized at ~0.2 M within 120 h at both sampling locations in 35% 2, and in the middle of the 60% saturated column (stage 4; Fig. 6.6C). At the base, an early transient peak was followed by a slow increase in Mg concentration for the remainder of the experiment (Fig. 6.6C). The pH in the system is controlled by the balance between CO2 dissolution, which decreases pH, and brucite dissolution, which increases pH. As such, the rapid DIC uptake was accompanied by a decline in pH in all columns (Fig. 6.6B; stages 1-2). The pH values declined from an initial value of 9.1 in all columns to ~8.0 within 12 h at the base of the 35% saturated columns. In the middle of these columns, the pH dropped rapidly to ~8.3 within 24 h, after which the rate of decline slowed considerably (stage 3; Fig. 6.6B). Final pH values were fairly stable at 7.6-7.8 at both locations (stage 4). The 60% saturated column exhibited a unique trend; the pH at the middle of the column followed a similar development as the base of the other columns, whereas the base sustained a higher pH throughout (~8.0-8.3; Fig. 6.6B). This is consistent with CO2 largely bypassing this sampling zone due to preferential flow along a different pathway.  Mineral saturation indices were calculated from the aqueous chemistry data using PHREEQC (Parkhurst and Appelo, 2013). Prior to initiation of the CO2 supply, the solutions were moderately undersaturated with respect to brucite (-0.5 to -0.7). DIC uptake rapidly drove the solutions to be increasingly undersaturated with respect to brucite (-1.5 to -3.4), and supersaturated with respect to both nesquehonite and artinite (Fig. 6.6D). Nesquehonite remained slightly supersaturated (0.1 to 1.2) throughout the experimental duration, whereas artinite became moderately undersaturated after 100 h (-0.1 to -1.1; Fig. 6.6D). Physical and chemical influences of water on mineral carbonation100[Mg] (M) Time (h) pH Time (h) 0.00 0.10 0.20 0.30 0.40 DIC (M) Time (h) Saturation index Time (h) 0 100 200 300 400 500 A) DIC7.0 7.5 8.0 8.5 9.0 9.5 10.0 0 100 200 300 400 500 B) pH0.00 0.10 0.20 0.30 0 100 200 300 400 500 C) Mg-6 -4 -2 0 2 4 6 0 100 200 300 400 500 35% 1base middle35% 260%artinitenesquehonitebruciteS2S3S4S1D) Saturation indicesFigure 6.6. Aqueous chemistry data versus time. A) Dissolved inorganic carbon (DIC) concentrations, B) pH, and C) Mg concentrations versus time. D) Mineral saturation indices versus time as calculated using PHREEQC with the Pitzer database and mineral solubility data from the minteq database. Light red and grey lines and squares represent data from the middle and base of 35% 2, respectively. Dark blue and light blue lines and squares represent data from the middle and base of the 60% saturated column, respectively. The black and bright red lines in (B) represent the base and middle of 35% 1, respectively. In (D), the long and short dashed lines represent the saturation indices of brucite and nesquehonite, respectively, and the solid lines represent the saturation index of artinite. Physical and chemical influences of water on mineral carbonation1016.3.4 Water	and	gas	flowAlthough a gradient in gas phase pressure must have existed to drive the gas flux through the columns, differences in gas phase pressure were too low to be resolved within measurement error (± 2.290 × 10-4 atm), and remained circum-atmospheric (1.000 atm). In all columns the water content was lowest at the top of the column (~15% saturated) and highest at the base (close to 100% saturated; Fig. 6.4). Because the supplied gas was not humidified, it is expected that water was lost due both to evaporation and precipitation of hydrated Mg-carbonates. However, the volumetric water content sensors did not resolve significant changes in water saturation at any of the three measurement depths (A4 Fig. A4.4B). Conversely, estimates of the total water mass in the columns calculated by mass difference before and after the solid materials were dried at the end of the experiments indicated that up to ~30% of the initial water was lost. As some solid material and pore water was lost during deconstruction of the experimental apparatus some of the final pore water was not accounted for; thus these estimates likely represent the lower limit for the final water content. The discrepancy between these two measurements suggests either that hydraulic redistribution between measurement points resulted in little change in saturation at the sensor measurement points, or that reaction induced water loss was localized and did not significantly change the water saturation measured in the vicinity of each sensor. 6.3.5 Stable carbon isotopic compositionsThe supplied gas had δ13C values between -37.5‰ and -36.5‰; variations between experiments were due to the use of different CO2 gas tanks. The δ13CDIC plummeted rapidly in the first ~12 h of the experiments as the gaseous CO2 dissolved into solution. Gas could not be sampled from the base of the experiments owing to the high water content, but its composition can be assumed nearly equal to that at the inlet (i.e., directly from the gas tank). In addition, samples collected from the middle and top of the columns for at least the first 47 h of the experiment had insufficient CO2 content to analyze. Nevertheless, progressive depletion of 13C Physical and chemical influences of water on mineral carbonation102in the supplied gas along the flow path is evident in both datasets (Fig. 6.7A). Consequently, the δ13CDIC in the middle of the column was also depleted with respect to that at the base during stage 2 in 35% 2 (Fig. 6.7A). Within 200 h, the isotopic composition of the gas phase throughout the column returned to the composition of the supplied gas, coincident with the cessation of the carbonation reaction (stage 4, Fig. 6.7A). The δ13CDIC values also stabilized after this time at values between -28.8‰ and -25.6‰ in all experiments. These δ13CDIC values fall within the expected range for equilibrium fractionation calculated using the fractionation factor of Mook et al. (1974) for fractionation between gaseous CO2 and HCO3-, the dominant aqueous DIC species at the experimental pH. The range of equilibrium values was calculated considering the temperature fluctuations of the experimental apparatus and the total range of composition of the initial gas. Temperatures varied between 16.9°C and 24.5°C for both experiments for which δ13CDIC data were available (A4 Fig. A4.4A). Solid samples could only be collected at the end of the experiment; therefore the δ13C values of solid phases represent a cumulative signature developed over the entire experiment. The δ13C values of the secondary phases were calculated from measured bulk values by correcting for the contribution from the dolomite present in the initial brucite ore as quantified with XRD data. The δ13C value of the initial carbonates was -6.5 ± 0.3‰ as measured in triplicate unreacted samples. Although magnesite and hydromagnesite were also documented in the initial brucite ore, they are not present at sufficient abundance to be quantified in the reaction products; therefore all initial CO2 is attributed to dolomite. The fractionation factor between HCO3- and nesquehonite or pseudo-artinite is unknown. However, the equilibrium fractionation factor of Wilson et al. (2010) between dypingite [Mg5(CO3)4(OH)2·5H2O] and HCO3- was found to be relatively consistent with fractionation observed during precipitation of nesquehonite under similar experimental conditions (Harrison et al., 2013a; Chapter 2), and is therefore employed here. A wide range in 13C composition of the final solids was measured, but the majority was within the range of expected values based on equilibrium fractionation with the range of δ13CDIC measured during brucite carbonation (Fig. 6.7B). Due to the error Physical and chemical influences of water on mineral carbonation103-50 -45 -40 -35 -30 -25 -20 -15 -10 0 100 200 300 400 500 δ13 C (‰ VPDB) Time (h) 0 10 20 30 40 50 60 70 -37 -32 -27 -22 Depth (cm) δ13C (‰ VPDB)CO2 gasmiddleoutletmiddleoutletDICinletmiddleinletmiddleSolidsreaction finger“dead zone”bulk/unknownA) δ 13C of CO 2(g)  and DICB) δ 13C of secondary carbonatesgas supplyDIC eq.solids eq.35% 260%35% 135% 260%Figure 6.7. Stable carbon isotopic composition of gaseous CO2, dissolved inorganic carbon (DIC), and secondary carbonate solid phases. A) δ13C values of gaseous CO2 and DIC versus time. The blue shaded region represents the range of δ13CDIC expected for equilibrium fractionation with the supplied gas based on the fractionation factor of Mook et al. (1974) for gaseous CO2 and HCO3- and the measured experimental temperature range. The orange shaded region represents the range of initial δ13C values of the supplied gas. B) δ13C values of the secondary carbonate solid phases versus depth in the columns. Orange filled symbols represent samples taken from reaction fingers, and white filled symbols represent samples taken from “dead zones” outside fingers. All others represent bulk samples or samples for which it was not possible to ascertain visually whether they were part of a reaction finger. The grey shaded region shows the range of δ13C values of the secondary carbonate solid phases predicted for equilibrium fractionation with the DIC at any point during the experiment. The fractionation factor of Wilson et al. (2010) for HCO3- and dypingite was utilized.Physical and chemical influences of water on mineral carbonation104associated with the corrections for the initial carbonate, and the lack of a specific fractionation factor, the uncertainty in these values is considerable. Nevertheless, they are broadly consistent with equilibrium fractionation between DIC and hydrated Mg-carbonates (Fig. 6.7B). 6.4 Reaction processes The observed trends in CO2 breakthrough curves are manifestations of reaction rate and progress along the length of the columns. Importantly, the reasonably constant pCO2 measured during stage 2 reveals that reaction rates remained relatively constant at all column depths for ~85 h in the 35% saturated experiments and ~50 h in the 60% saturated experiments (Fig. 6.8A-C). The plateau occurs at similar timing and for similar durations at all three sensor locations within any single experiment (Fig. 6.8B and C). A precipitous cessation of carbonation following stage 2 is evidenced by the broader range in CO2 sequestration rate and CO2 concentration during stage 3 at each measurement depth within the columns (Fig. 6.8B and C), and the consistent approach towards the CO2 concentration of the supplied gas (C/C0 → 1; Fig. 6.8A). Finally, in the 60% saturated column, the CO2 breakthrough curves measured at the middle and outlet of the column are remarkably similar, implying that no significant reaction is occurring between these two points for most of the experimental duration. These features of the breakthrough curves are important indicators of the reaction processes governing carbonation during the different stages of reaction.  6.4.1 Reaction rateIt is typically expected that bulk mineral dissolution rates decline over time as the reactive phase is consumed, due to the geometric decline in particle size and consumption of reactive surface area (Levenspiel, 1999; Appelo and Postma, 2005). This decline in surface area is often approximated with the following relationship representative of a uniform population of shrinking spheres or cubes (Eq. 6.4; Lichtner, 1996; Mayer et al., 2002; Colón et al., 2004; Physical and chemical influences of water on mineral carbonation105Appelo and Postma, 2005; Jeen et al., 2007; Wilson et al., 2014; Chapter 7):k e f fg  = k 0 SAt023                                              (Eq. 6.4)where kgeff (L s-1) is the effective ‘geometric’ rate constant, k0 is the initial reaction rate constant equal to 10-6.13 L s-1 m-2 (after Pokrovsky et al., 2005b), SA is the surface area (m2), and φ0 and φt are the volume fraction of the reactive phase initially and a given time greater than zero. Here, kgeff replaces k0eff in Equation 6.2. Such a relationship results in a gradual decline in reaction rate associated with the shrinking geometric surface area of the particles, and is in poor agreement with the experimental observations (Fig. 6.8D). Rather, a period of constant reaction rate (stage 2; Fig. 6.8B) suggests either that the effective surface area of brucite remains relatively constant despite its consumption, or that the overall sequestration rate is not controlled by brucite dissolution kinetics. The latter implies that either the supply of CO2 or the rate of carbonate precipitation was limiting the overall reaction rate. The advancement of a sharp reaction front through the columns could also produce a constant bulk rate of reaction, but the presence of contemporaneous gas composition plateaus throughout the column disqualifies this as an explanation (Fig. 6.8A-C). A constant reaction rate would be achieved if the rate-limiting step for carbonation were the supply of CO2, as the sequestration rate would approximately balance the supply rate. However, the elevated CO2 concentration throughout most of the column depth (Fig. 6.8A and B) implies that CO2 gas is supplied to the column faster than it is locally removed. Other studies have shown that the gas-solution transfer and subsequent hydration of aqueous CO2 to a form that can be mineralized may be relatively slow compared to the flux of CO2 gas (Harrison et al., 2013a; Chapter 2; Power et al., 2013a; Wilson et al., 2010). This would result in CO2-limited brucite carbonation despite the non-zero CO2 concentration in the gas phase. CO2 uptake into solution has been demonstrated to be rate limiting under similar experimental conditions (Wilson et al., 2010; Harrison et al., 2013a; Chapter 2; Power et al., 2013a), and is Physical and chemical influences of water on mineral carbonation1060 100 200 300 400 500 05.0 ×10 -61.0 ×10 -51.5 ×10 -52.0 ×10 -5A)B)D)Time (h) CO2 sequestration rate (mol s-1) Middlegeometric modelthreshold modelSequestration rate versus timeDepth (cm) C/C 0  CO2 sequestration rate (mol L-1  bulk s -1 ) CO2 breakthrough (C/C0)00.20.40.60.81.0inletmiddleoutlet35% 1Time (h)0 100 200 300 400 500E)0 10 20 30 40 50 60 70 0 0.5 1 0 10 20 30 40 50 60 70 0 0.5 1 Depth (cm) 0 10 20 30 40 50 60 70 2×10 -61×10 -600 10 20 30 40 50 60 70 2×10 -61×10 -6025-85 h C)110-330 h 25-85 h 110-330 h0.0 0.1 0.2 0.3 0.4 0.5 0.6 0 20 40 60 80 100 Specific surface area (m2  g-1) Dissolution (%) Total surface area (m2 ) Surface area versus time1×10 -22×10 -23×10 -24×10 -2TSA calculatedSSA geometricT SA geom etricSSA calculated25-85 h110-330 hS2S3Figure 6.8. Analysis of CO2 breakthrough curve features (see following page for caption). Physical and chemical influences of water on mineral carbonation107Figure 6.8 (continued). Analysis of CO2 breakthrough curve features. A) CO2 breakthrough curves from 35% 1. Black, red, and green lines represent CO2 concentration measurements at the base, middle, and top of the column. CO2 breakthrough is expressed as C/C0, the ratio of the measured CO2 concentration at a given time to the CO2 concentration of the supplied gas. B) C/C0 versus depth in 35% 1 during reaction stage 2 (35-85 h) and reaction stage 3 (110-330 h). C) CO2 sequestration rate versus depth in 35% 1 during reaction stage 2 (35-85 h) and reaction stage 3 (110-330 h). Rates were calculated using the CO2 flux data at each sensor and are expressed as mol CO2 L bulk-1 s-1. For both (B) and (C), horizontal bars represent the range in value of C/C0 or CO2 sequestration rate at a given depth for the entire time range plotted. Data are averaged over 2 h intervals (25 measurements) to reduce measurement noise. D) Comparison of the experimental instantaneous reaction rate (solid red line) versus time in the middle of the column and the reaction rates predicted by the geometric (dashed black line) and threshold (dashed red line) models. E) Calculated evolution of total surface area and specific surface area versus extent of brucite dissolution. Solid and dashed black lines show the evolution of total and specific reactive surface area, respectively, as calculated using the geometric model. The solid red line shows the total surface area evolution required to maintain a constant reaction rate, and the dashed red line shows the evolution of specific surface area required to achieve this constant total surface area.typified by anomalously depleted 13C compositions of both DIC and precipitated carbonates due to kinetic fractionation of stable carbon isotopes during uptake of CO2 into solution (Wilson et al., 2010; Harrison et al., 2013a; Chapter 2; Wilson et al., 2014; Chapter 7). However, the δ13C values of the DIC and the secondary carbonates are consistent with equilibrium rather than kinetic fractionation in the columns (Fig. 6.7), suggesting that CO2 uptake into solution was not rate limiting. The discrepancy between these and previous experimental results (Wilson et al., 2010; Harrison et al., 2013a; Chapter 2) is attributed to the use of different experimental apparatus. Previous studies that identified CO2 uptake as the rate-limiting step were conducted in batch reactors; the increased gas-water interface provided in the column experiments likely facilitated more rapid uptake of CO2 into solution. Thus it appears unlikely that reaction rate was limited by the rate of CO2 supply.  Changes in total brucite surface area would also have less of an impact on the bulk rate of CO2 sequestration if carbonate precipitation were to limit the overall carbonation rate, a limitation which has recently been recognized for certain carbonation reactions (e.g., Pham Physical and chemical influences of water on mineral carbonation108et al., 2011; Saldi et al., 2012; Hövelmann et al., 2012a; Kampman et al., 2013; Gautier et al., 2014). A relatively slow rate of secondary phase precipitation may in some cases allow the solution to approach equilibrium with respect to the dissolving phase, retarding its dissolution (Maher et al., 2009). The moderate supersaturation with respect to nesquehonite documented in the columns (Fig. 6.6) implies that its precipitation was moderately kinetically inhibited. Nevertheless, brucite remained undersaturated (saturation index ≈ -0.5 to -3.4; Fig. 6.6), implying that reaction affinity effects due to relatively slow carbonate precipitation were unlikely to have limited the CO2 sequestration rate. Furthermore, pH values tended to plateau contemporaneously with CO2 concentrations, implying that a steady-state is achieved between brucite dissolution and carbonate precipitation. Because the reactions are sequential, the observed steady-state implies that brucite dissolution controls the overall reaction rate during stage 2 of the reaction.It is therefore concluded that the overall reaction rate during stage 2 was primarily controlled by the rate of brucite dissolution, as is consistent with the observation that an increase in brucite surface area accelerated the carbonation reaction under similar experimental conditions (Chapter 5). This implies that the total reactive surface area remained relatively constant during stage 2 in order to maintain the relatively constant rate of reaction, invalidating the assumption that reactive surface area declines geometrically (Fig. 6.8D). It is recognized that the reactive surface area of a mineral is not always equal to the geometric surface area, and therefore may not evolve geometrically. Surface roughness and mesoporosity provide additional surface area that is not accounted for by simple geometric models (e.g., Brantley and Mellot, 2000), and not all surface area is equally reactive. For example, the crystal edges of several sheet silicate minerals dissolve much more rapidly than basal sheets (Bosbach et al., 2000; Bickmore et al., 2001; Hodson, 2006), as do edge dislocations associated with exsolution lamellae (Lee et al., 1998) and defects induced by grinding (Petrovich, 1981; White and Brantley, 2003; Maher et al., 2006). Yet, walls of etch pits formed during quartz dissolution do not contribute to the reactive surface area (Gautier et al., 2001). In the case of brucite, the relatively constant Physical and chemical influences of water on mineral carbonation109reaction rate recorded during stage 2 implies that the specific surface area (i.e., surface area per unit mass) must be increasing to account for the loss of brucite mass to a greater extent than is predicted geometrically (Fig. 6.8E). This could be explained if reaction occured primarily via the stepwise removal of Mg(OH)2 sheets. Hövelmann et al. (2012b) observed this process during brucite dissolution at the nano-scale using atomic force microscopy. If the majority of reaction occurs in this manner, the total reactive surface area may not change substantially until a particle is completely dissolved, as the surface area of each newly exposed sheet would be similar throughout the reaction. In addition, the formation of secondary phases of higher volume than the primary phase can induce fracturing and generate surface area as reaction progresses (Kelemen et al., 2011 and references therein). For example, Beinlich and Austrheim (2012) documented that precipitation of lansfordite fractured serpentinite clasts during weathering. Intragranular dissolution along planes can also lead to physical weakening of the mineral grain, which induces flaking and exposure of fresh surface area (Lee et al., 1998). This is consistent with the observed brucite pieces separated from the surface and isolated between extensive carbonate precipitates (Fig. 6.5G and H). Moreover, the presence of carbonate-filled cracks extending into the interior of brucite grains reveals that the reactive fluid accessed additional surface area (Fig. 6.5B). The roughened and fractured appearance of brucite particles following carbonation may be due to a combination of reaction induced cracking and preferential dissolution along weaknesses in the particles, causing continuous exposure of fresh surface area within the first ~100 h of reaction. This would counteract the reduction in brucite mass as the reaction progressed, allowing the relatively constant rate of reaction to persist (Fig. 6.8E). 6.4.2 Reaction shutdown6.4.2.1 Surface passivation Our previous experimental results from smaller scale brucite carbonation columns revealed that the transformation from a flakey, poorly or nano-crystalline carbonate phase to Physical and chemical influences of water on mineral carbonation110low porosity, bladed nesquehonite resulted in abrupt passivation of brucite and arrested the reaction (Chapter 5). The morphology and stoichiometry of the poorly crystalline phase was similar to that of the pseudo-artinite in the present study. Due to the stoichiometric similarity between artinite and the pseudo-artinite formed in the experiments, and the good agreement between predicted artinite-buffered aqueous chemistry and measured aqueous chemistry in the experiments (Fig. 6.6D), artinite solubility data are used as a proxy for pseudo-artinite in all models. Aqueous chemistry data, which were not available from the smaller scale experiments, reveal that although nesquehonite quickly becomes supersaturated and remains so throughout the experiment, artinite becomes undersaturated within 100 h (Fig. 6.6D). The change in saturation state of artinite with time and the constant supersaturation with respect to nesquehonite are consistent with initially rapid precipitation of pseudo-artinite, coupled with relatively slow precipitation of nesquehonite replacing the earlier-formed carbonate. The observation of bladed crystals rimming pseudo-artinite is consistent with this process (Fig. 6.5A). The transformation from porous pseudo-artinite to low porosity nesquehonite as the chemical environment evolves results in abrupt passivation of the brucite surface, as is consistent with previous experiments using brucite of the same grain size and BET surface area (Chapter 5). In these experiments, surface passivation was found to limit the mass of CO2 sequestered to ~2.5-3.0 wt.% CO2 (Fig. 6.9; Chapter 5). Yet, above 35 cm depth in the 35% saturated columns, and at many locations throughout the 60% saturated column, a lower extent of reaction than this was achieved (Fig. 6.9), suggesting that there were additional processes limiting the extent of reaction. 6.4.2.2 Water-limited carbonationWater is required both as a medium to transport and solubilize ions and as a reactant that is consumed during precipitation of hydrated Mg-carbonates. Due to the formation of hydrated phases, there is a fundamental stoichiometric limit to the amount of carbonate that can form based on the mass of water available. The solid black line in Figure 6.9 shows the Physical and chemical influences of water on mineral carbonation111stoichiometric limit for conversion of brucite to nesquehonite, which consumes more water than pseudo-artinite. If the stoichiometric limit were the only restriction on the extent of reaction, the experimental data would fall on this line when water saturation was below 17%. Above 17% saturation, the reaction is no longer restricted by the stoichiometric limit, yet a large proportion of samples still exhibited a lower extent of reaction than is expected due to surface passivation or stoichiometry (Fig. 6.9). Nevertheless, the positive correlation between water saturation and the extent of carbonation (Fig. 6.9) implies that the decreased extent of reaction is likely a consequence of water content. Furthermore, SEM micrographs of the reaction products reveal that the thickness and location of carbonate surface coatings depended on the local water content (Fig. 6.5). At low water saturation, carbonate precipitates were restricted to narrow regions where water films were present (Fig. 6.5C and D), whereas at high water saturation, precipitates were observed throughout the pore space (Fig. 6.5E and F). The strikingly different textures observed between the water-poor and water-rich areas of the columns imply that the overall reaction progress fundamentally differs in low water conditions. The decreased extent of reaction above the stoichiometric limit may relate to the requirement of water as a reaction medium rather than its role as a reactant. Heterogeneous water distribution at the pore scale may leave some mineral surfaces unexposed to reactive fluids, reducing the surface area available for reaction. This is consistent with pore scale observations of brucite carbonation, which revealed a lack of reaction of particles in dry pores, whereas particles in wet pores reacted with CO2 to form nesquehonite (Chapter 4). Although water is the wetting fluid in this situation, poor connectivity between water films may leave some grains effectively unexposed to reactive fluid. In reactive transport models, effective reactive surface areas are often reduced to account for the combined effects of incomplete exposure of mineral surfaces to reactive fluid, surface aging, and surface passivation effects (Xu et al., 2007; Bea et al., 2012). Similarly, Pačes (1983) defines the mineral surface area available for reaction in natural catchments to be equal to the surface area in the saturated zone only. If mineral grains in “dry” pores were assumed to be unreactive, both the reactive Physical and chemical influences of water on mineral carbonation112Large columnSmall column (Chapter 5)FineSmall column(Chapter 5)Medium“dead zone”reactive channelPore scale observation(Chapter 4)Very fine0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 0 20 40 60 80 100 CO2 sequestered (wt%) Final water saturation (%) surface passivationd ry p ore modelwater limitedreaction fingeringSmall column(Chapter 5)stoichiometric li mitFigure 6.9. CO2 sequestered versus final water saturation. The mass of CO2 sequestered is from the total carbon measurements, as corrected for the initial carbon content. Diamonds represent data from the present study (large columns), whereas triangles represent data from our previous, smaller scale brucite carbonation columns (small columns; Chapter 5), and the square represents an observation from our previous pore scale experiments (pore scale observation; Chapter 4). Blue, red, and black data represent data from experiments that employed “medium brucite” that was 250-500 μm in diameter, the same as used in the present study, “fine” brucite that was 53-180 μm in diameter, and “very fine” brucite that was <53 μm in diameter, respectively. White and orange triangles with blue outlines show samples from obvious reaction fingers and “dead zones” that are bypassed by CO2, respectively. The solid black line represents the stoichiometric limit for reaction based on the water available for reaction of brucite to form nesquehonite. The grey-shaded area to the left of this line represents a zone with no reaction due to the lack of water available to form hydrated carbonates. The dashed blue line shows a “dry pore” calculation, wherein the volume of brucite available for reaction (i.e., the extent of reaction) is proportional to the fraction of pore space occupied by water. Physical and chemical influences of water on mineral carbonation113surface area and reactive capacity of the porous medium would decrease as a function of water saturation.On the other hand, Nishiyama and Yokoyama (2013) documented that the effective surface area of quartz in a sandstone was unaffected by water saturation because the quartz grains were coated with sufficiently connected water films that maintained a high available surface area. The efficacy of these water films to support reaction may vary over time in our experimental systems compared to those of Nishiyama and Yokoyama (2013), because water saturation evolves with reaction progress and the reaction involves precipitation of minerals that reduce the local water content. An expanding body of work has revealed that there is a water concentration threshold required for carbonation of brucite, forsterite [Mg2SiO4], and wollastonite [CaSiO3] to proceed with wet supercritical CO2 (e.g., Loring et al., 2011; Schaef et al., 2011; Felmy et al., 2012; Miller et al., 2013). Schaef et al. (2013a; 2011) documented that forsterite carbonation in wet supercritical CO2 ceases when water films are too thin to behave like bulk water, and that brucite carbonation is arrested prior to completion due to water loss to nesquehonite. Pore scale observations of brucite carbonation at ambient conditions have also revealed that carbonate precipitates tend to stop growing upon reaching the gas-water interface (Chapter 4). If all mineral grains were coated with sufficiently connected water films, the initial available surface area should not decrease at low water saturation; however, once the water volume was consumed to below a critical threshold required to support carbonation, reaction would cease. This would reduce the reactive capacity but maintain an initial effective surface area comparable to that in water saturated media. Here, we use the reactive transport model, MIN3P-DUSTY (Molins and Mayer, 2007) to evaluate whether a water limitation could explain the lack of carbonation at the top of the column, and by what mechanism it likely occurs. Three “end-member” conceptual models based on the above discussion are explored: 1) a “base case” that assumes the reaction is limited by surface passivation but is independent of water saturation, 2) a “dry pore” model that defines a portion of the brucite as unreactive because it resides in effectively dry pores, Physical and chemical influences of water on mineral carbonation114and 3) a “water film” model that assumes all brucite as initially available for reaction, but the reaction stops prematurely due to the loss of water to hydrated Mg-carbonates (Fig. 6.10). The effect of particle movement documented in microfluidic experiments (Chapter 4) are neglected, as the large size of the brucite grains meant that the majority likely bore the skeletal load of overlying material.6.4.2.3 Conceptual models of water-limited carbonationBase case. The first conceptual model represents the extent of reaction that would be achieved if water content had no effect on reactivity. The extent of carbonation depends primarily on the distance from the CO2 source and duration of exposure to the CO2-rich gas stream, and is limited by surface passivation only. The equation derived from our previous, smaller scale column experiments to model surface passivation with negligible water limitations is employed (Eq. 6.5; Chapter 5):k e f ft  =  k 0 SA 10 -t0 - p5                                       (Eq. 6.5)for φt ≤ φp, where kteff is the effective ‘threshold’ rate constant (L s-1), k0 is the initial reaction rate constant equal to 10-6.13 L s-1 m-2 (after Pokrovsky et al., 2005b), SA is the brucite surface area (m2), and φ0 and φp are the initial brucite volume fraction and a threshold brucite volume fraction at which the reaction rate becomes negligible due to surface passivation. Here, kteff replaces k0eff in Equation 6.2. This formulation is referred to as the ‘threshold’ model, and provides a relatively constant reaction rate, followed by a rapid shutdown of reaction at a threshold degree of brucite conversion (represented by the volume fraction of brucite; Fig. 6.8D), which is attributed to surface passivation (Chapter 5). Dry pore model. The dry pore model assumes that some mineral grains are not exposed to reactive fluid, resulting in a lower effective initial brucite volume fraction and surface area. Here, it is assumed that the fraction of brucite available for reaction is proportional to the Physical and chemical influences of water on mineral carbonation115fraction of water filled pore space (Fig. 6.10; Eq. 6.6). The threshold volume fraction (φp) is also proportional to the water filled pore space, such that the same extent of carbonation (i.e., percent conversion of brucite to carbonate) occurs prior to passivation (Eq. 6.7).z0  = 0   S w                                                    (Eq. 6.6)zp  = p   S w                                                    (Eq. 6.7)Where φz0 and φzp are the effective initial and threshold volume fractions at a depth ‘z’ in the column, and Sw is the water saturation (fraction of pore volume occupied by water). The variables φz0 and φzp are substituted into Equation 6.5 in place of φ0 and φp, respectively. Water	film	model. In the third conceptual model, it is assumed that all brucite grains are coated with sufficiently connected water films such that the initial effective volume fraction and surface area are the same as in the base case under water-unlimited conditions (Fig. 6.10). However, the progressive loss of water due to uptake into the solid phase decreases the volume of water in the vicinity of reactive surfaces over time until the reaction cannot proceed. Here, carbonation ceases when the local water saturation reaches a threshold value. Thus, the threshold function was again employed but the threshold volume fraction was a function of water saturation instead of surface passivation (Eq. 6.8): zp  = 0  0.102S w                                               (Eq. 6.8) This effectively reduces the reactive capacity of the porous medium, by decreasing the volume fraction of brucite available as a function of water saturation (Eq. 6.8). For simplicity, the dependence of this threshold brucite volume fraction was expressed as a linear function of water saturation (Eq. 6.8), with the slope fitted between endpoints at zero water saturation, where no reaction will occur, and an upper saturation limit above which water is Physical and chemical influences of water on mineral carbonation116no longer limiting (slope = 0.102; see A4 Fig. A4.8). At the experimental conditions, surface passivation is expected to limit the extent of brucite conversion to ~60%, when not limited by water availability (i.e., above the upper saturation limit; Chapter 5; A4 Fig. A4.8). We use an empirically derived value from the experiments of 20% saturation as the upper saturation limit for water-limited conditions. This is in fairly good agreement with both the stoichiometric limit (17%) and water thresholds reported in previous studies under similar conditions (Assima et al., 2013a; Chapter 5). Above this limit, an unmodified version of the base case threshold model accounting for passivation alone is employed (Eq. 6.5). Below this limit, Equation 6.8 is substituted for φp in Equation 6.5. The model was modified to include water as a reactant during precipitation of artinite, such that the water saturation evolved as carbonates precipitated. The net water removal from both nesquehonite and artinite precipitation was attributed to artinite, as it precipitated in greater quantities. 6.4.2.4 Reactive transport modeling of water-limited carbonationModeling results from each scenario were compared to experimental CO2 breakthrough curves (Fig. 6.11) and solid phase brucite and CO2 abundance data (Fig. 6.12). Model results for individual carbonate minerals are provided in Appendix 4 (A4 Fig. A4.9). The base case model reproduces the experimental data for the base and middle of the column well, where the water saturation exceeds ~15% (Fig. 6.11A). This is consistent with the extent of carbonation being limited primarily by surface passivation where water saturation is highest (i.e., base and middle). However, it is clear that the base case model overestimates the reactive capacity for the drier top of the column, as evidenced by the delayed CO2 breakthrough and the homogeneous extent of carbonation predicted at all depths (Figs. 6.11A and 6.12). This implies that factors other than surface passivation by nesquehonite are limiting reaction in these areas. Conversely, the dry pore model underestimates the extent of carbonation and the instantaneous rate, as indicated by the premature breakthrough and excessive pCO2 during stage 2 at the middle and top of the column (Figs. 11B and 12). The proportional relationship Physical and chemical influences of water on mineral carbonation117between water saturation and reactive capacity is clearly an oversimplification; the fraction of brucite with access to water must exceed the bulk saturation. This implies that capillary and electrostatic forces provide sufficiently connected water films to coat and connect the majority of grains (e.g., Tokunaga, 2011), providing exposure to reactive fluid. Consistent with this, the water film model produces excellent agreement between the modeled and experimental data (Figs. 6.11C and 6.12). In order to reproduce the experimental data, an initial surface area equal to that in water-unlimited conditions was required to account for the lack of CO2 breakthrough at the top of the column during the first ~100 h (i.e., water film model). This implies that the lack of reaction cannot be attributed to a lack of wetted surfaces. Instead, low water content appears to primarily limit the extent of reaction possible rather than the reactive surface area, which has important implications for quantifying reactive surface area in the unsaturated zone. Simply decreasing the effective surface area, as is often the approach, does not represent reaction progress well. Because brucite carbonation is a coupled dissolution-precipitation CO2Unreactive ReactiveA) Dry poreCO2B) Water filmReactiveFigure 6.10. Conceptual models of water limited carbonation. A) Dry pore model. B) Water film model.Physical and chemical influences of water on mineral carbonation118reaction, the restricted area within which precipitates can form at low water content, and the corresponding water loss during precipitation results in a premature cessation of the reaction compared to surface passivation alone. The precipitation of hydrous Mg-carbonates therefore provides a negative feedback on reaction progress due both to surface passivating effects and the removal of water. 6.4.3 Preferential	flow	paths The anomalously early breakthrough documented at the middle sensor of the 60% saturated column, and the heterogeneity in reaction extent in all columns is indicative of preferential flow paths. Orange coloration visibly identifies narrow (~5 cm diameter) channels of highly reacted material surrounded by white, low to moderately carbonated material (Fig. 6.13). Although some channeling was observed in the 35% saturated columns, the process was most extreme in the 60% saturated column. This is attributed to the higher percentage of water filled pores that effectively block gas transport (Collin and Rasmuson, 1988); therefore a limited volume is available for the gas to flow. Because the bases of the columns were at or close to 100% water saturation, gas flow into the columns necessitated displacement of pore water. It is well known that the differences in viscosity between two fluids, such as the supplied gas and the pore water, leads to instabilities at the their interface (Hill, 1952; Marulanda et al., 2000). This may trigger the development of viscous fingers in the direction of flow and thus heterogeneous distribution of the invading fluid in the porous medium (Hill, 1952; Marulanda et al., 2000). Because the gas flow in the experiments was upward, and it has a much lower density and viscosity than the displaced water, the invading front was inherently unstable and should lead to fingering (Hill, 1952; Kueper and Frind, 1988; Marulanda et al., 2000). Although the importance of viscosity and density differences has been demonstrated in homogeneous media (Marulanda et al., 2000 and references therein), in heterogeneous porous media, viscous fingering becomes less important, and flow paths are instead governed by the permeability field (Kueper and Frind, 1988). It is plausible that small-scale heterogeneities in Physical and chemical influences of water on mineral carbonation119water filmφ zp = φ0- 0.102S w0 100 200 300 400 500 Time (h) 0.0 0.2 0.4 0.6 0.8 1.0 C/C0 ModelExperiment (35% 1)C/C0 Time (h) base caseφ zp = φ ip  A)dry poreφ zp = φ ip   × S wC/C0 Time (h) B)C)basemiddletopbasemiddletop0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 0 100 200 300 400 500 0 100 200 300 400 500 Figure 6.11. Comparison of modeled versus experimental CO2 breakthrough curves from 35% 1. Black, red, and green lines represent CO2 breakthrough curves from the base, middle, and top of the column. Solid and dashed lines represent experimental and modeled data, respectively. A) Results from the base case model. B) Results from the dry pore model. C) Results from the water film model.Physical and chemical influences of water on mineral carbonation1200 10 20 30 40 50 60 70 Depth (cm) Abundance (wt.%) 0 10 20 30 40 50 60 70 Depth (cm) Abundance (wt.%) A) brucite B) %CO 2base casedry porewater film35% 135% 20 2 4 6 8 0 2 4 6 8Figure 6.12. Comparison of modeled versus experimental brucite (A) and secondary phase CO2 (B) abundance profiles from 35% 1 and 2. The black and grey squares illustrate measured brucite abundance in 35% 1 and 2, respectively. The grey dashed and black dash-dot lines represent the base case and dry pore models, respectively, and the solid blue line represents the water film model.the porous medium lead to preferential channelization of the flow along more permeable paths. For instance, the highly carbonated region along the column wall (Fig. 6.13C) is attributed to channelization between the porous media and the column wall. Conversely, the smaller diameter highly carbonated zones visible towards the middle of the column (Fig. 6.13C) could be the result of instabilities at the fluid interface. Although care was taken to ensure the porous media in the column experiments was homogeneous, it is likely that small-scale heterogeneities in permeability existed. Initial intrinsic and relative permeability distributions are likely to control the early Physical and chemical influences of water on mineral carbonation121infiltration of the porous medium by the gas phase, but the question as to how these initial fingers will evolve during carbonation due to reaction induced changes in porosity and permeability provides additional complexity. Mineral dissolution and precipitation can lead to positive or negative feedbacks that enhance or dampen finger development; a process known as reactive-infiltration instability (e.g., Daccord and Lenormand, 1987; Ortoleva et al., 1987; Daccord, 1987; Kelemen et al., 1995; Dipple and Gerdes, 1998; Kalia and Balakotaiah, 2007). Despite the removal of water from the pore space during hydrated Mg-carbonate precipitation, a net decrease in gas-filled pore volume during carbonation is expected due to the large solid volume increase, irrespective of the type of Mg-carbonate mineral formed (A4 Fig. A4.10). Yet, the persistence of the viscous fingers implies that permeability remained greater within the fingers than in the surrounding media, despite the porosity reduction. This is attributed to the fact that precipitation would occur in water filled pores within the fingers (e.g., Chapter 4), thus the majority of pore space clogging would not occur in the gas-filled pores through which the gas was flowing. Moreover, the evaporative removal of water from the pore space would be greatest in the fingers where the gas flux was highest, thereby increasing the gas filled pore volume and relative permeability with respect to the gas phase; a positive feedback. The implication of these effects is that a large proportion of the reactive material is inaccessible, resulting in a reduction of both reactive capacity and effective surface area. The reduced reactive capacity is evidenced by the lower mass of CO2 sequestered in the 60% saturated column despite the minimization of water-limited conditions. In practice, this is equivalent to the ‘dry pore model,’ except the number and distribution of channels cannot be estimated a priori. 6.5 Implications6.5.1 Mineral dissolution-precipitation reactions in the vadose zoneQuantification of reactive surface area, upon which mineral weathering reactions responsible for nutrient and carbon cycling and contaminant attenuation depend (Stipp, 1998; Physical and chemical influences of water on mineral carbonation122low carbonation5 cmhighly carbonatedlow carbonation5 cm10 cmA BCFigure 6.13. Photographs of the 60% saturated column at the conclusion of the experiment. A) Photograph of the column exterior showing reaction fingers: narrow zones of highly reacted material typified by an orange coloration. Off-white areas are zones of poor to moderately carbonated material. The bright orange area seen from the exterior at the column’s base is the mesh that separates the reactive material from the gravel layer. Map view of the solid materials at the conclusion of the experiment during excavation near the top of the column (B) and near the base of the column (C). Orange colored reaction fingers are evident near the base of the column, but are not visible near the top of the column to the overall lesser extent of reaction, which was insufficient to induce a visible color change.Physical and chemical influences of water on mineral carbonation123Cubillas et al., 2005; McKinley et al., 2006; Manning, 2008; Maher et al., 2009; Wilson and Bish, 2012), is complicated by low water availability in the unsaturated zone and changes in physical surface area with reaction progress. Reactive transport modeling revealed that low water saturation does not significantly reduce reactive surface area; rather it reduces the reactive capacity of the porous medium. Omission of this effect in reactive transport models may result in significant over prediction of the extent of reaction. Simply decreasing the effective surface area, as is often the approach, does not represent reaction progress well; the reactive capacity must also be considered. Moreover, the maintenance of reactive surface area despite considerable consumption of the reactive phase has implications for quantifying the evolution of reactive surface area. In the case of brucite, surface area is better represented as being constant, due in part to the roughening and fracturing of grains during reaction rather than assumed to decrease with decreasing geometric grain size. Similar mechanisms may occur during dissolution of feldspars with exsolution lamellae (Lee et al., 1998) or due to reaction induced fracturing during natural or engineered carbonation of Mg-silicates (Kelemen et al., 2011; Beinlich and Austrheim, 2012). 6.5.2 CO2 sequestrationLocal water saturation is significantly altered during injection of CO2 into subsurface formations for sequestration purposes (Loring et al., 2011; Schaef et al., 2011; Van Pham et al., 2012; Schaef et al., 2013a; Miller et al., 2013; Thompson et al., 2013; Thompson et al., 2014), and near surface CO2 sequestration strategies and natural systems will also be affected by rainfall and evaporation. The limited extent of reaction at low water contents implies that the capacity of such CO2 sequestration strategies will be strongly restricted by the available volume of water. Such effects may also be relevant for precipitation of hydrated clays, sulfates, and zeolites that have water in their crystal structures. Conversely, at high water saturation, preferential flow path development could significantly reduce the reactive capacity when CO2 is injected into porous media, although this could in part be controlled by the rate of injection Physical and chemical influences of water on mineral carbonation124(e.g., Ji et al., 1993). Preferential flow paths could likely be redirected by the addition of water, for example, by rainfall or process water recharge in mine tailings, and could be minimized via injection of gas only above the water table in engineered systems.Offsetting of CO2 emissions at the Mount Keith Nickel Mine1257. Offsetting of CO2 emissions by air capture in mine tailings at the Mount Keith Nickel Mine, Western Australia: Rates, controls and prospects for carbon neutral mining6 7.1 Introduction Storage of CO2 in carbonate minerals has been recognized as a safe and effective method for mitigating rising concentrations of atmospheric greenhouse gases (Seifritz, 1990; Lackner et al., 1995; Lackner, 2003). Unlike traditional Carbon Capture and Storage (CCS), which relies predominantly upon stratigraphic trapping of injected supercritical CO2 within rock formations, carbon mineralization technologies utilize direct production of carbonate minerals as traps for CO2. Since carbon mineralization was first proposed as a method for storing CO2 (Seifritz, 1990), most of the work on this subject has focused on the development of rapid, large-scale ex situ methods for trapping and storing CO2 at industrial point sources (reviewed in Huijgen and Comans, 2003; IPCC, 2005; Huijgen et al., 2005; Sipilä et al., 2008; Power et al., 2013b). Most processes developed to date are based on reaction of CO2 with naturally occurring non-carbonate minerals such as silicates, hydroxides or oxides. Weathering of these minerals in nature is generally a slow process and high temperatures and pressures are needed to induce carbon mineralization reactions on the short timescales (i.e., hours) required for development and deployment of industrial carbonation reactors (Sipilä et al., 2008; Zevenhoven et al., 2011). Although rapid and technologically feasible, the financial viability of this approach to carbon mineralization is currently limited by low carbon prices (Power et al., 2013a; Power et al., 2013b).6A version of this chapter is published and is reprinted with permission from International Journal of Greenhouse Gas Control, 25, Wilson, S. A., Harrison, A. L., Dipple, G. M., Power, I. M., Barker, S. L. L., Mayer, K. U., Fallon, S. J., Raudsepp, M. and Southam, G., Offsetting of CO2 emissions by air capture in mine tailings at the Mount Keith Nickel Mine, Western Australia: Rates, controls and prospects for carbon neutral mining, 121-140, Copyright (2014), with permission from Elsevier.Offsetting of CO2 emissions at the Mount Keith Nickel Mine126Recent research has focused increasingly on two alternative approaches: (1) in situ carbon mineralization and (2) ex situ carbon mineralization at industrial sites. In situ carbon mineralization aims to promote subsurface carbonation by injection of CO2, or solutions bearing dissolved CO2, into mafic and ultramafic formations (e.g., Cipolli et al., 2004; Kelemen and Matter, 2008; Gislason et al., 2010). Ex situ carbon mineralization at industrial sites focuses on low-temperature and low-pressure carbonation of alkaline industrial wastes such as smelter slag, fly ash, alkaline and saline brine, construction waste, and mine tailings (e.g., Wilson et al., 2006; Manning, 2008; Power et al., 2009; Huntzinger et al., 2009; Wilson et al., 2009a; Wilson et al., 2009b; Renforth et al., 2009; Ferrini et al., 2009; Power et al., 2010; Zhao et al., 2010; Ballirano et al., 2010; Wilson et al., 2010; Pronost et al., 2011; Renforth et al., 2011; Wilson et al., 2011; Power et al., 2011b; Bobicki et al., 2012; Pronost et al., 2012; Harrison et al., 2013a; Chapter 2; Manning and Renforth, 2013). Carbonation of the mineral waste from ultramafic mines has been documented previously (Wilson et al., 2006; Wilson et al., 2009a; Wilson et al., 2009b; Wilson et al., 2011; Pronost et al., 2012; Beinlich and Austrheim, 2012; Oskierski et al., 2013); however, previous estimates of carbonation rates have generally relied on small sample sets and have not been explained within the context of geochemical modeling. Complementary laboratory experiments and reactive transport modeling have recently been developed to investigate and quantify controls on the rate of carbon mineralization in mine tailings (e.g., Wilson et al., 2010; Pronost et al., 2011; Bea et al., 2012; Assima et al., 2013a; Assima et al., 2013b; Harrison et al., 2013a; Chapter 2). Thus, a detailed framework now exists for assessing, monitoring and modeling carbon mineralization in mine tailings. Here, for the first time, we apply isotopic and crystallographic carbon accounting methods and reactive transport modeling to quantitatively assess the rate of, and controls on, carbon mineralization on the scale of a large operating tailings facility at the Mount Keith Nickel Mine, Western Australia. This represents the first implementation of a detailed framework for carbon accounting during carbon mineralization via air capture in an active industrial setting and on a landscape scale.Offsetting of CO2 emissions at the Mount Keith Nickel Mine127At the Mount Keith Nickel Mine, the hydrated Mg-carbonate mineral, hydromagnesite [Mg5(CO3)4(OH)2·4H2O], develops as a weathering product within ultramafic mine tailings. Consequently, accounting of the amount of atmospheric CO2 that is being trapped and stored within this mineral could be used to offset the mine’s greenhouse gas emissions. The isotope system (i.e., δ13C, δ18O, and F14C) employed by Wilson et al. (2009a) has been used to assess capture and storage of atmospheric CO2 within hydromagnesite in the tailings at Mount Keith. Further to this, quantitative phase analysis using the Rietveld method and powder X-ray diffraction (XRD) data are used to estimate the amount of hydromagnesite in the tailings. Quantification of hydromagnesite at Mount Keith is non-trivial, because the mineral content of mine tailings is generally heterogeneous due to frequent changes in ore mineralogy, processing procedures, and tailings management practices. The mineralogy of tailings piles also changes with time as tailings react with the atmosphere and meteoric and process waters to produce secondary mineral phases. As a result, geostatistical methods that rely on the presence of regular and predictable geological structures cannot be used to predict the mineralogy of a mine tailings pile. Thus, in order to estimate the total amount of CO2 captured from the atmosphere and stored within secondary carbonate mineral phases at Mount Keith, it is necessary that the mineralogy of its tailings storage facilities be well constrained. We have constructed a database of quantitative mineralogical results for a large suite of tailings samples, taken from various depths below the surface of the tailings storage facility at Mount Keith.Because the ages are known for many of the tailings flows at Mount Keith, we have been able to use our quantitative mineralogical results to build a time-dependent reactive transport model that describes the geochemical evolution of its tailings storage facilities (Bea et al., 2012; this study). From this model, we illuminate the mechanisms governing carbon mineralization and obtain an empirical rate for hydromagnesite precipitation. Finally, the reactive transport model has been used to assess potential methods by which tailings management practices could be tailored to enhance carbon mineralization and maximize offsetting of greenhouse gas emissions at Mount Keith. Offsetting of CO2 emissions at the Mount Keith Nickel Mine1287.2 Locality and sampling strategy7.2.1 The Mount Keith Nickel MineThe MKD5 orebody at Mount Keith is located in the North Eastern Goldfields district of Western Australia (Fig. 7.1) and is the largest nickel producer in Australia (Grguric, 2003). The deposit at Mount Keith occurs in the NNW/SSE-trending Agnew–Wiluna greenstone belt in the Archaean Yilgarn Craton (Hill et al., 1990). The MKD5 orebody is hosted by komatiitic peridotite (primarily dunite), which attained mid–upper greenschist facies as a result of regional metamorphism (Barrett et al., 1977). Retrograde serpentinization and carbonation of the host peridotites resulted from infiltration by H2O–CO2-rich fluids (Barrett et al., 1977; Grguric et al., 2006). Resulting metamorphic assemblages (from proximal to distal) are (1) talc–magnesite, (2) antigorite–magnesite, and (3) lizardite–brucite–hydrotalcite group (Grguric et al., 2006).Conventional, staged-cutback, open pit mining methods are practiced at MKD5, yielding approximately 11 Mt of ore annually (Grguric, 2003). The mining operation at MKD5 produces approximately 370 000 t of greenhouse gases (cited as CO2 equivalent) and approximately 11 Mt of ultramafic tailings each year (BHP Billiton, 2005). Ore from the MKD5 deposit is processed using froth flotation methods to concentrate sulfide minerals (Grguric et al., 2006). Additives used in processing include citric and sulfuric acids, guar gum, Na-dithionite, Na-ethyl xanthate, and (historically) soda ash. In 2004, ore reserves contained 0.52 wt.% nickel (Grguric et al., 2006), primarily in high-Ni pentlandite [(Fe,Ni)9S8], godlevskite [(Ni,Fe)9S8], heazlewoodite [Ni3S2], and millerite [NiS]. Recovery of these minerals from the flotation circuit is typically about 70% (Grguric et al., 2006). The material rejected from the processing plant is piped to one of two tailings storage facilities (TSF2, which was the only facility in operation at the time of sampling), and is suspended in the hypersaline process water used in the flotation circuit. Tailings are deposited from spigots located on risers at nine points in TSF2 (Fig. 7.1A).The tailings storage facility at Mount Keith was constructed in two phases. The two cells of TSF1 (Fig. 7.1A) were the sole receptacles for tailings from July 1994 until the facility was decommissioned in January 1997 (Stolberg, 2005). A centralized discharge tailings storage Offsetting of CO2 emissions at the Mount Keith Nickel Mine129facility (TSF2) was commissioned to replace TSF1 by January 1997 (Stolberg, 2005) and remains in operation today. Along the circumference of TSF2, the outer 100 to 400 m of the compound are dedicated to catching and storing tailings in the event that mineral waste from the interior of the facility should overflow. This design feature has been effective and provides 2425302322203 21421614171551 5253 54 555646 4712441013 451127424350292848493256373673534339841403926381 km19 31kidney wallTSF2TSF1water returnE1E2E3E4W4W3118W2W1C1-2BMount KeithA3 45267819Figure 7.1. The Tailings Storage Facilities (TSFs) at Mount Keith (A). Location of Mount Keith Nickel Mine within Australia (B). The roughly circular facility in (A) is TSF2 and the smaller, adjoining facility is TSF1. Samples were collected at 59 locations in TSF1 and TSF2. Sampling in TSF2 was done along the perimeter of the facility and on both sides of a radial access road. Sampling in TSF1 was done proximal to the access road in Cell 2 (the right-most cell of TSF1). Squares indicate sites sampled in 2005; circles indicate sites sampled in 2006; borders on circles indicate sites for which Rietveld refinements have been done; triangles indicate locations of risers from which tailings are deposited. Shading denotes regions of the TSFs in which deposition of tailings had ceased as of 2006.Offsetting of CO2 emissions at the Mount Keith Nickel Mine130snapshots of tailings compositions from several overflow events. Tailings in the various overflow cells of TSF2 had been permitted to weather, without addition of new tailings or process water, for periods of one, three, and between seven and eight years prior to sampling in 2006. At the time of sampling, tailings in the interior of TSF2 (along the radial access road, Fig. 7.1A) had last been deposited approximately zero or one half years prior to collection. 7.2.2 Strategy for sampling at Mount KeithLimited sampling of the tailings storage facilities at Mount Keith was begun in April 2005. More extensive sampling was done in September and October of 2006. In excess of 800 samples were collected from TSF2 and the older, now-decommissioned TSF1. Bea et al. (2012) describe detailed quantitative mineralogical results and reactive transport modeling of TSF1.At the time of fieldwork, large regions of TSF2 were saturated with process water, making these regions of the tailings facility inaccessible for sampling. In order to compensate for incomplete access to the tailings at Mount Keith, sampling of the main tailings storage facility (TSF2) was carried out along a radial maintenance road and around the perimeter of the roughly circular facility (Fig. 7.1A). Samples were collected at random intervals along these two paths in TSF2. Tailings were primarily sampled by collection of single cores, coring on 5 m x 5 m and 10 m x 10 m grids (after Roselle et al., 1999) and from vertical profiles on exposed surfaces and excavated trenches. Further detail is available in the Supporting Information (SI)7.To the best of our knowledge, all of the tailings (at the surface and at depth) sampled from the overflow cells around the exterior of TSF2 are of a specific age (either one, three, or between seven and eight years). Contrastingly, only the ages of tailings sampled from near the surfaces in the central regions of TSF2 are well constrained (as zero or one half years). 7The Supporting Information to the Wilson et al. (2014) International Journal of Greenhouse Gas Control article is available with the online version of the paper: doi:10.1016/j.ijggc.2014.04.002. Supplementary data relevant to the reactive transport modeling is included as Appendix 5 (A5) in this thesis.Offsetting of CO2 emissions at the Mount Keith Nickel Mine1317.3 Analytical and modeling methodsQualitative and quantitative mineralogy were obtained from powder X-ray diffraction (XRD) data. Quantitative phase analysis with the Rietveld method (Rietveld, 1969; Hill and Howard, 1987; Bish and Howard, 1988) was done on 172 samples of mine tailings from TSF2 at the Mount Keith Nickel Mine. Samples for Rietveld refinement were selected to optimize coverage of the accessible regions of the tailings storage facility. A random number generator was used to select subsets of samples collected from 5 m x 5 m and 10 m x 10 m grids for quantitative phase analysis.Scanning electron microscopy (SEM) and energy dispersive X-ray spectrometry (EDS) were used to image and characterize mineral habits and textural relationships amongst minerals in thin section. Stable carbon and oxygen isotopic data were obtained from specimens of tailings, waste rock, ore, process additives and water. Radiocarbon data were collected from specimens of CO2 that were isolated from magnesite and hydromagnesite in bulk tailings using a selective acid extraction procedure. All radiocarbon data were collected using the Single Stage Accelerator Mass Spectrometer (SSAMS) at The Australian National University. Detailed analytical methods are provided in the Supporting Information.  The reactive transport code MIN3P (Mayer et al., 2002; Bea et al., 2012) was employed to identify the processes governing hydromagnesite precipitation in the active tailings storage facility at Mount Keith, and to estimate the rate of CO2 fixation. MIN3P is a multicomponent reactive transport code that allows direct coupling between transport and reaction processes (Mayer et al., 2002). It was recently modified by Bea et al. (2012) to better represent the influence of dynamic atmospheric conditions on reaction progress by incorporating energy balance equations and vapor transport. Mayer et al. (2002) and Bea et al. (2012) provide complete descriptions of the governing equations and verification examples for MIN3P.Offsetting of CO2 emissions at the Mount Keith Nickel Mine1327.4 Qualitative	mineralogical	results	and	field	observations7.4.1 Qualitative mineralogy of Mount Keith mine tailingsTailings at Mount Keith are composed primarily of serpentine minerals, antigorite, lizardite and minor chrysotile [Mg3Si2O5(OH)4], with hydrotalcite-group minerals including iowaite [Mg6Fe2(OH)16Cl2·4H2O] and woodallite [Mg6Cr2(OH)16Cl2·4H2O] with occasional pyroaurite [Mg6Fe2(OH)16CO3·4H2O], stichtite [Mg6Cr2(OH)16CO3·4H2O], and uncommon mountkeithite [(Mg,Ni)11(Fe,Cr)3(OH)24(SO4,CO3)3.5·11H2O]. A solid solution exists amongst iowaite, woodallite, pyroaurite and stichtite and ideal end-member compositions are generally not observed at Mount Keith (Grguric, 2001). Minor amounts of brucite, chrysotile, talc, magnetite, chromite, quartz, magnesite, dolomite, and calcite are common. Trace vermiculite is also observed in the tailings. Sulfide minerals have not been detected with XRD to a limit of approximately 0.5 wt.%. These observations are consistent with ore mineralogy (Grguric, 2003). Efflorescences of secondary minerals are abundant near surfaces within the tailings storage facilities and include carbonate minerals, halide minerals and sulfate minerals (listed in SI Tables S1 and S.2). Efflorescences of sulfate minerals commonly form at the surface of mine tailings during dry conditions or in arid climates, and are leached from tailings during rainfall events (Jambor et al., 2000). At Mount Keith, hydromagnesite, halite, hexahydrite [MgSO4∙6H2O], and blödite [Na2Mg(SO4)2∙4H2O] dominate these efflorescences, and are commonly associated with lesser amounts of epsomite [MgSO4∙7H2O], konyaite [Na2Mg(SO4)2∙5H2O], löweite [Na12Mg7(SO4)13∙15H2O], and gypsum [CaSO4∙2H2O]. Sanderite [MgSO4∙2H2O], starkeyite [MgSO4∙4H2O], pentahydrite [MgSO4∙5H2O], and carnallite [KMgCl3·6H2O] are less common and have only been observed at low abundance. Kainite [Mg(SO4)KCl·3H2O] and anhydrite [CaSO4] may or may not be present near detection in a very few of the samples analyzed. The hydration states of sulfate minerals are strongly dependent on temperature and relative humidity (e.g., Chou and Seal, 2003; Chipera and Vaniman, 2007; Chou and Seal, 2007) and double salts are known to decompose to single salts (e.g., konyaite and blödite may decompose Offsetting of CO2 emissions at the Mount Keith Nickel Mine133to produce thenardite, hexahydrite and amorphous phases; Mills et al., 2010). Thus, the relative abundances of hydrated sulfate minerals measured in the laboratory may not reflect abundance in the field.7.4.2 Occurrence of hydromagnesite mineralizationHydromagnesite was detected by XRD in the majority of samples, but is more commonly found at high abundance in samples of shallow tailings. SEM images of Mount Keith mine tailings commonly show fine crystals of hydromagnesite precipitating at the surface of grains of serpentine (Fig. 7.2). In some instances, this hydromagnesite infills both fine cracks and broad fissures in serpentine grains (Fig. 7.2A). The latter textures suggest that hydromagnesite may be forming by replacement of serpentine. Micrometer-scale crystals of hydromagnesite radiate out into the spaces between tailings grains and commonly cement them together (Fig. 7.2B). These textures are similar to those described by Wilson et al. (2009a) for dense cements of hydromagnesite and dypingite [Mg5(CO3)4(OH)2·5H2O] that form at depth within tailings SHȝPAMHHHSȝPSSSSHHMMBFigure 7.2. Backscattered electron images of hydromagnesite in Mount Keith mine tailings. Both (A) and (B) are from a particularly hydromagnesite-rich sample, 06MK9-9. (A) Hydromagnesite crystals precipitating at the surface of larger grains of serpentine and infilling cracks and fissures within them. (B) Hydromagnesite cementing both fine and larger grains of serpentine and magnetite together. H – hydromagnesite; M – magnetite; S – serpentine.Offsetting of CO2 emissions at the Mount Keith Nickel Mine134at the Clinton Creek chrysotile mine, Yukon, Canada. The textural relationship between hydromagnesite and gangue minerals such as serpentine confirms the secondary origin of hydromagnesite as a weathering product at Mount Keith.Secondary hydromagnesite is typically concentrated within and just below efflorescent crusts of sulfate minerals; however, hydromagnesite persists at depth within the tailings at Mount Keith whereas sulfate minerals generally do not. Although tailings that are heavily cemented with hydromagnesite commonly occur near the surface of TSF2, hydromagnesite-rich samples are also encountered at depth within the tailings storage facility.Excavating trenches within TSF2 to collect profile samples reveals distinct horizons of tailings material. Within these profiles, layers of clay and silt sized tailings commonly overlay layers of sand sized tailings. These well-sorted layers are likely indicative of mass settling within tailings flows during distinct depositional events [see upper 2 m of the profile at sampling site 1 (i.e., 06MKP1 within 0.5-year old tailings)], which is illustrated in Figure 7.3A and 7.3B. Tailings flows consisting of fine and coarse layers range from 1 cm to 40 cm thick. The most recently deposited tailings in 06MKP1 are covered by an efflorescent crust of blödite and halite and are enriched in hydromagnesite (samples labeled 1–3 in Fig. 7.3A). Hydromagnesite and sulfate minerals are not detectable in the layers of fine sand (4) and coarse sand (5) beneath the surficial samples. However, hydromagnesite is present in deeper samples (6–8) that are associated with filled desiccation cracks (located between samples 7 and 8). Again, hydromagnesite is absent from sample 9, which grades into medium to coarse sand, but it reappears in sample 10. Although sample 10 is of coarse sand, it is a thin horizon and is in contact with a deeper, silty layer (11) that also contains hydromagnesite. The same trend of coarsening grain size with increasing depth occurs between samples 11 and 14, all of which contain hydromagnesite. Filled desiccation cracks appear once again between sample 14 and the fine-grained, hydromagnesite-rich tailings from sample 15. Beneath this level, at a depth of approximately 160 cm, the tailings become darker in color and smell faintly of organics.Desiccation cracks are a common feature within mine tailings at the surfaces of TSF2. Offsetting of CO2 emissions at the Mount Keith Nickel Mine135Within weeks to months of tailings deposition, desiccation cracks begin to appear within the upper few centimeters of new tailings flows. Shallow, distantly spaced desiccation cracks can be seen in the newly deposited tailings at sampling site 30 (Fig. 7.3C). Stolberg (2005) notes that such cracks form within one month in column experiments conducted on water-saturated tailings from Mount Keith when these are left to drain. After approximately 6–12 months, many more cracks will have propagated within the hardened efflorescences on tailings surfaces (Fig. 7.3D). If the tailings are left exposed for a significant amount of time, the desiccation cracks can become deeper features as efflorescent sulfate minerals are leached from the surficial tailings, leaving behind hardened blocks and layers of hydromagnesite-cemented tailings (Fig. 7.3E). The prevalence of desiccation cracks at tailings surfaces, the preservation of infilled desiccation features at depth, and the higher abundance of hydromagnesite within deep samples such as 06MKP1-8 and 06MKP1-15 indicate that the horizons from which they were taken were once exposed at the surface. It is particularly notable that these former surfaces and the horizons adjacent to them are enriched in hydromagnesite compared to the hydromagnesite-poor horizons of tailings that separate them. While hydromagnesite is most abundant near tailings surfaces, including former tailings surfaces, observations of hydromagnesite in deep, sandy horizons suggests that it may continue to precipitate or become reworked at depth.7.4.3 Consequences of timing and depth of tailings deposition for hydromagnesite formationMost of the sampling in TSF2 was done to depths between zero and 130 cm using a sediment-coring device. The ~35-meter tall W1 riser (near sampling sites 1 and 2, Fig. 7.1A) was buried in tailings to a depth of 19 m at the time of sampling. This represents an average deposition of approximately 2 m of tailings each year from the time TSF2 was commissioned up until the time of sampling. However, mine tailings deposits are commonly less than 1 m deep within the exterior overflow cells and at the edges of the large central cell of TSF2. Mining at Mount Keith produces approximately 11 Mt of tailings each year. If this mass of Offsetting of CO2 emissions at the Mount Keith Nickel Mine13610 cm12345678109111213141516A B5 m20 cm20 cm3 mCDEFigure 7.3. Mount Keith mine tailings at the surface and at depth within TSF2. The upper half of the profile at sampling site 1 (06MKP1) (A). Numbers label specific samples (e.g., “1” denotes sample “06MKP1-1”) and white lines separate distinct horizons. (B) The profile excavation from (A) and a 5m x 5m grid sample (delineated by the blue tarpaulin) at sampling site 1. (C) Backhoe excavation at sampling site 30, showing near surface water and meter-scale desiccation cracks. (D) An older surface in TSF2 (~1-year old) showing well-developed efflorescences and more pervasive desiccation features. (E) The 7–8-year old surface of TSF2 at sampling site 9 (showing sample 06MK9-6), which is heavily cemented with hydromagnesite.Offsetting of CO2 emissions at the Mount Keith Nickel Mine137tailings were deposited uniformly throughout TSF2 it would be approximately 50 cm in depth. After 10 years (i.e., from January 1997 to October 2006) of uniformly thick deposition, tailings deposits would be 5.0 m deep. Clearly, the deep deposits that surround risers are not typical of all regions within TSF2, which suggests that depths on the order of meters are more common than those on the order of decameters.Stolberg (2005) gives values of 3.0 to 3.8 m for the depth of the tailings deposited near the center of Cell 2 in the now unused TSF1 (i.e., the rightmost cell in Fig. 7.1A) and values as high as 10 m for the depth of tailings near discharge points along the perimeter of Cell 2. As such, deep grid samples in TSF1 and, similarly, near the risers in TSF2 would most likely be accessing tailings deposited within one year of last deposition at the surface. Following this line of reasoning, we have attributed the same ages known for the surface deposits to the deeper tailings in our analyses.Mount Keith experiences an average pan evaporation of 2,400 mm yr-1 and limited rainfall (220 mm yr-1 on average), most of which is lost to evapotranspiration (Stolberg, 2005). The location of the watertable within most regions of TSF2 was difficult to assess by coring; however, the water table was observed at an elevation of 524.3 m above sea level at both sampling site 30 and at the water return pond (sampling site 25). At sampling site 30, samples collected with a backhoe excavator were saturated with water, indicating that they had been collected from beneath the vadose (unsaturated) zone. Recently deposited tailings from some areas of TSF2 were wetted beneath efflorescent crusts, and holes cored through these damp tailings had a tendency to swell shut within minutes of sampling, but standing water was not observed. This suggests that in parts of TSF2, only the upper few meters of tailings are unsaturated.Darkly colored, anoxic samples were collected from the base of the profile at sampling site 1, from the base of the backhoe trench at sampling site 30 (Fig. 7.3C) and from shallow depth (i.e., < 20 cm) near the water return pond. These samples may reflect a prevalent compositional change within the deeper, anoxic tailings that are located within the saturated Offsetting of CO2 emissions at the Mount Keith Nickel Mine138zone of TSF2. Hydromagnesite was present within the anoxic samples from sites 1 and 30 at abundances comparable to those in the unsaturated zone, suggesting that hydromagnesite is relatively insoluble within the saturated zone of TSF2 and would persist at depth following burial. Thus, estimates of hydromagnesite abundance, made predominantly from samples collected from within the vadose zone, should still apply to deeper, saturated tailings.7.5 Analytical results7.5.1 Rietveld	refinement	resultsRietveld refinement results for tailings from Mount Keith are shown in Figures 7.4 through 7.7 and the complete dataset is available in SI Table S5. Refined abundances and median abundances for hydromagnesite are plotted in Figure 7.4 for tailings of five different ages (i.e., 0 years, 0.5 years, 1 year, 3 years and 7 to 8 years). Shaded envelopes denote the median plus or minus the median absolute deviation of mineral abundance (e.g., Upton and Cook, 2008) for depth intervals (in cm) of [0, 25), [25,50), [50, 75), [75, 100), [100, 125), [125, 150), and [150, maximum depth]. Data are plotted similarly in Figures 7.5 through 7.7. Results for 10-year old tailings from TSF1 are discussed and modeled by Bea et al. (2012). We use the median and median absolute deviation (MAD) as measures of central tendency because Rietveld refinement results produce a sparse set of log-normally distributed data. Statistical analyses are discussed in more detail in the Supporting Information.Offsetting of CO2 emissions at the Mount Keith Nickel Mine1390 5 10 15050100150200250300350400Hydromagnesite Abundance (wt.% Rietveld)Depth Below Surface (cm)0 5 10 15 0 5 10 15 0 5 10 15 0 5 10 15Age of tailings: 0 years 0.5 years 1 year 3 years 7-8 yearshydromagnesite abundance (sampled over an interval,centred on median sample depth)median ± MAD of hydromagnesite abundance (calculated for25 cm intervals)hydromagnesite burial modelprocess water replenishment modelhydromagnesite burial model (initial abundance)Figure 7.4. Variation of hydromagnesite abundance with depth beneath the surface of TSF2 over time. Age is measured relative to 2006, the date of sampling. Open circles indicate the median depth from which a sample was taken. The associated vertical lines represent intervals of depth over which individual samples were collected. Lines drawn through the data for tailings of each age connect points that denote the mean depth of sampling (based on median depths for individual samples) and the median abundance of hydromagnesite for samples from similar depths. Shaded regions are the median value ± the median absolute deviation. Black and red lines represent model output from modeling of the hydromagnesite burial and process water replenishment hypotheses, respectively.Offsetting of CO2 emissions at the Mount Keith Nickel Mine140Depth Below Surface (cm)0 5 10 150501000 5 10 15 0 5 10 15 0 5 10 15 0 5 10 15Halite0 years 0.5 years 1 year 3 years 7-8 yearsHydromagnesite0 5 10 150501000 5 10 15 0 5 10 15 0 5 10 15 0 5 10 15Blödite0 5 10 15 20 0 5 10 15 20 0 5 10 15 20 0 5 10 15 20 0 5 10 15 20050100Mineral Abundance (wt.% Rietveld)0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5050100GypsumFigure 7.5. Variation of the abundance of select secondary minerals (hydromagnesite, halite, blödite and gypsum) with depth beneath the surface of TSF2 over time. Age is measured relative to 2006, the date of sampling.Offsetting of CO2 emissions at the Mount Keith Nickel Mine141Depth Below Surface (cm)0 years 0.5 years 1 year 3 years 7-8 yearsHydromagnesite0 5 10 150501000 5 10 15 0 5 10 15 0 5 10 15 0 5 10 15Mineral Abundance (wt.% Rietveld)Dolomite0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5050100Calcite0 0.5 1 1.5 0 0.5 1 1.5 0 0.5 1 1.5 0 0.5 1 1.5 0 0.5 1 1.5050100Magnesite0 5 10 150501000 5 10 15 0 5 10 15 0 5 10 15 0 5 10 150 5 10 15 20 0 5 10 15 20 0 5 10 15 20 0 5 10 15 20 0 5 10 15 20050100Hydrotalcite-groupFigure 7.6. Variation of the abundance of gangue carbonate minerals (hydrotalcite-group, calcite, dolomite and magnesite) compared to hydromagnesite abundance with depth beneath the surface of TSF2 over time. Age is measured relative to 2006, the date of sampling.Offsetting of CO2 emissions at the Mount Keith Nickel Mine142Depth Below Surface (cm)0 years 0.5 years 1 year 3 years 7-8 yearsHydromagnesite0 5 10 150501000 5 10 15 0 5 10 15 0 5 10 15 0 5 10 150 5 10 150501000 5 10 15 0 5 10 15 0 5 10 15 0 5 10 15BruciteSerpentine40 50 60 70 80 90 40 50 60 70 80 90 40 50 60 70 80 90 40 50 60 70 80 90 40 50 60 70 80 90050100Mineral Abundance (wt.% Rietveld)Figure 7.7. Variation of the abundance of brucite and serpentine-group minerals compared to hydromagnesite abundance with depth beneath the surface of TSF2 over time. Age is measured relative to 2006, the date of sampling.Offsetting of CO2 emissions at the Mount Keith Nickel Mine1437.5.2 Stable isotopic resultsThe δ13C values for carbonate minerals from Mount Keith mine tailings (see Fig. 7.8 and SI Table S6) vary from -8.56‰ to +1.04‰ (VPDB), with δ18O ranging from 7.81‰ to 39.89‰ (VSMOW). One specimen of bedrock magnesite from a sample of waste rock is characterized by δ13CVPDB = -6.52‰ and δ18OVSMOW = 7.91‰. A specimen of iowaite (a hydrotalcite-group mineral) from the mine pit gives values of δ13CVPDB = -2.83‰ and δ18OVSMOW = 16.11‰. Specimens of late-stage Ni-dolomite alteration give -6.58‰ ≤ δ13CVPDB ≤ -5.78‰ and 26.35‰ ≤ δ18OVSMOW ≤ 27.86‰ (δ13Cav = -6.16‰ and δ18Oav = 27.30‰). Bedrock magnesite, isolated from tailings by selective acid dissolution, gives values of -6.79‰ ≤ δ13CVPDB ≤ -1.23‰ and 10.33‰ ≤ δ18OVSMOW ≤ 18.98‰ (δ13Cav = -4.11‰ and δ18Oav = 13.27‰). Notably, magnesite extracted from bulk tailings is slightly more enriched in 13C and significantly enriched in 18O over the specimen of pure magnesite that was sampled from waste rock. Duplicate analyses of one sample of soda ash (used historically as an industrial process chemical) give average values of δ13CVPDB = -8.69‰ and δ18OVSMOW = 17.73‰. Dissolved inorganic carbon in tailings water gives values of –8.16‰ ≤ δ13CVPDB ≤ -6.20‰ (δ13Cav = -7.36‰). Water from the processing plant gives values of -6.91‰ and -4.60‰ (δ13Cav = -5.76‰).Analyses of bulk carbonate minerals give -6.70‰ ≤ δ13CVPDB ≤ -0.33‰ and 14.16‰ ≤ δ18OVSMOW ≤ 31.97‰ (δ13Cav = -3.46‰ and δ18Oav = 22.60‰). Hydromagnesite, isolated by selective acid extraction, is characterized by -8.6‰ ≤ δ13CVPDB ≤ 1.0‰ and 24.5‰ ≤ δ18OVSMOW ≤ 39.9‰ (δ13Cav = -3.0‰ and δ18Oav = 34.8‰). On average, secondary hydromagnesite is enriched over magnesite by approximately 1‰ in δ13C and approximately 22‰ in δ18O.Offsetting of CO2 emissions at the Mount Keith Nickel Mine1447.5.3 Radiocarbon resultsThe F14C values for carbonate minerals at Mount Keith range from 0.004 to 1.052 and are given in Figure 7.9 and SI Table S6. Average values of F14C and δ13CVPDB are plotted for samples on which replicate analyses have been done. Because δ13C data for four specimens of magnesite are not available, radiocarbon results are plotted for two of these specimens only (i.e., CO2 selectively extracted from magnesite in samples 06MKG-2-7-3 and 06MKG10-5-1).One specimen of bedrock Ni-dolomite gives an F14C of 0.004 and two specimens of highly pure bedrock magnesite are characterized by 0.005 and 0.009 (the latter value is -30 -25 -20 -15 -10 -5 0 5 10 150510152025303540δ13C (‰, VPDB)δ18 O (‰, VSMOW)1) atmospheric3) organicdypingiteprecipitating fromprocess/tailings waterHydromagnesite (selective) Ni-dolomite (waste rock)Magnesite (waste rock)Bulk CarbonateMagnesite (selective)Hydrotalcite-group (pit)Soda Ash (additive)2) bedrockFigure 7.8. Stable oxygen and carbon isotope data for different modes of occurrence and mineralogy of carbonate minerals at Mount Keith. Numbers label fields for Mg-carbonate minerals in equilibrium with specific reservoirs for carbon. Measurement errors (as 2σ) are typically smaller than the symbols employed.Offsetting of CO2 emissions at the Mount Keith Nickel Mine145an average of two analyses). These results for highly pure bedrock carbonate minerals are consistent with 14C-free carbon and the level of atmospheric contamination expected from gas separation in our vacuum line and preparation of graphite samples. Carbon dioxide extracted from magnesite in four bulk samples of tailings is defined by 0.024 ≤ F14C ≤ 0.365 (F14Cav = 0.147), the upper end of this range of values being significantly higher than expected from measurements on highly pure specimens. One specimen of soda ash (used historically as a process chemical in the flotation circuit at Mount Keith) has an F14C of 0.052. Analysis of CO2 F 14Cδ13 C (‰, VPDB)3) organic1) atmospheric2) bedrock0 0.2 0.4 0.6 0.8 1 1.2-30-25-20-15-10-5051015Hydromagnesite (selective)Ni-dolomite (waste rock)Soda Ash (additive)Magnesite (selective)100% moderndypingiteprecipitating fromprocess/tailings waterFigure 7.9. Stable carbon (δ13C) and fraction modern carbon (F14C) data for secondary precipitates of hydromagnesite, bedrock carbonate minerals, and soda ash (a process additive) from Mount Keith. Numbers label fields for Mg-carbonate minerals in equilibrium with specific reservoirs for carbon. Values for δ13C were determined in the Pacific Centre for Isotopic and Geochemical Research at the University of British Columbia. Values for F14C are corrected for machine fractionation using AMS results for δ13C. Measurement errors are typically smaller than the symbols employed.Offsetting of CO2 emissions at the Mount Keith Nickel Mine146selectively extracted from hydromagnesite is characterized by 0.572 ≤ F14C ≤ 1.052 (average F14C = 0.921).7.6 Discussion of mineralogical and isotopic results7.6.1 Mineralogical change and formation of hydromagnesite7.6.1.1 Occurrence of efflorescent mineralsThe most common and abundant efflorescent phases at Mount Keith are secondary hydromagnesite, sulfate minerals and halite. The distribution of these phases changes with depth below the tailings surface and evolves with time since deposition of tailings in TSF2.The abundances of both hydromagnesite and halite in tailings of all ages are greatest within 0 to 25 cm of the surface (Figs. 7.4 and 7.5). This is consistent with previous observations and model results, which indicate that secondary efflorescent minerals are more likely to form within an evaporative horizon at the contact between tailings and the atmosphere (e.g., Acero et al., 2007; Acero et al., 2009; Bea et al., 2012). The abundance of hydromagnesite becomes relatively constant at depths below 25 cm in tailings of each age. However, high abundances of hydromagnesite are sometimes measured at greater depths (Fig. 7.4), which may reflect partial preservation of crusts from former tailings surfaces or ongoing precipitation at depth. It is also possible that hydromagnesite from buried crusts is being reworked at depth as the tailings settle under their own weight. Both hydromagnesite and halite may also be remobilized to some extent by water percolating through the tailings (primarily process water in TSF2, with some influence from meteoric water). Quantitative mineralogical results also show that the mean abundance of hydromagnesite within the upper 25 cm increases with time since deposition of tailings. Thus, hydromagnesite precipitation is an ongoing process in TSF2.It is notable that hydromagnesite abundance follows a similar trend to that for halite (Fig. 7.5). The formation of halide crusts within tailings in arid climates is known to result from strong capillary transport of aqueous species toward tailings surfaces (Dold, 2006). As with hydromagnesite, the highest concentrations of halite occur at shallow depths, typically Offsetting of CO2 emissions at the Mount Keith Nickel Mine147above 25 cm (Fig. 7.5). Halite is also present at depth within the tailings and is found at low abundances in the vast majority of samples analyzed. This suggests that both halite and hydromagnesite remain saturated within tailings and are therefore relatively stable following burial under fresh tailings. Or it could imply that process water is already near saturation with respect to hydromagnesite prior to deposition of tailings, such that percolation of process waters does not significantly remobilize secondary phases.Sulfate minerals precipitate in mine tailings as a consequence of oxidative weathering of primary sulfide minerals (e.g., Jambor and Blowes, 1998; Jambor et al., 2000). Blödite, hexahydrite and other sulfate minerals are common at the surface of 0- to 1-year old tailings, but persist only at reduced abundance in a few outlier samples at the surface of 3- and 7–8-year old tailings (Fig. 7.5). Sulfate minerals are scarce below 10 cm depth and are almost entirely absent below a depth of 25 cm. Gypsum, which is relatively insoluble compared to most sulfate minerals, appears to be the only sulfate phase that persists at depth, but it is uncommon and present only at low abundance (Fig. 7.5). Bea et al. (2012) demonstrate that sulfate minerals are absent from tailings in the non-operational TSF1 at Mount Keith. Without new input of tailings process water, rainfall events likely leach sulfate minerals from the surface of TSF1 within 10 years of deposition (consistent with observations by Jambor et al., 2000). The presence of sulfate minerals in surface crusts suggests that the older tailings flows around the perimeter of TSF2 may occasionally be exposed to fresh tailings water from the interior region of the storage facility. Additionally, the presence of anoxic, darkly colored tailings at depth within TSF2 suggests that surficial sulfate dissolution may be fueling microbial sulfur reduction in deeper tailings horizons. Acero et al. (2007; 2009) demonstrate that the formation of efflorescent crusts of sulfate minerals reduces the permeability of mine tailings, which limits gas exchange with the atmosphere (i.e., evaporation, and O2 and CO2 diffusion). Thus, formation of efflorescences may be viewed as a self-limiting process that controls the depth to which alteration minerals may extend. Precipitation of carbonate and evaporite minerals, such as hydromagnesite and halite, Offsetting of CO2 emissions at the Mount Keith Nickel Mine148within tailings pores likely participates in controlling the rate and extent of mineralization.Previous studies have also shown that efflorescent crusts may trap and concentrate upward diffusing gases such as CO2 in pore spaces near tailings surfaces (Blowes et al., 1991; Tasse et al., 1997; Agnew and Taylor, 2000). Agnew and Taylor (2000) found that CO2 concentrations in pore spaces beneath a crust of hydrated Fe-sulfate minerals at the Elura Ag-Pb-Zn Mine near Booroondarra, New South Wales, Australia, were up to 10 times that of atmospheric CO2. This process only occurs where CO2 is being generated at depth within tailings as a consequence of biodegradation of underlying organics or dissolution of carbonate minerals during biologically mediated oxidation of sulfide minerals. The undetectable quantities of sulfide minerals and high abundances of carbonate minerals within both fresh and weathered tailings suggest that neither process is significant at Mount Keith. Instead, it is more likely that the ubiquitous cracks and fissures that texture efflorescences allow for limited gas exchange between the deeper tailings and the atmosphere once efflorescences have formed.7.6.1.2 Relationship of hydromagnesite to primary gangue mineralsSeveral gangue minerals including serpentine-group minerals, brucite, hydrotalcite-group minerals, magnesite and dolomite could act as sources of magnesium in hydromagnesite. Dissolution of bedrock (gangue) carbonate minerals could provide DIC into solution; essentially recycling previously fixed carbon into newly formed hydromagnesite. However, this would not constitute net sequestration of CO2 within hydromagnesite, which requires a non-carbonate mineral source for both magnesium and carbon (Wilson et al., 2009a).Variations in the abundances of mined bedrock calcite, dolomite, magnesite and carbonate-bearing hydrotalcite minerals with depth and time are plotted in Figure 7.6, and median values show no consistent trends in their distribution with depth or time. In almost all cases, the abundances of bedrock carbonate minerals stay relatively constant in comparison to results for secondary minerals such as hydromagnesite (Fig. 7.5), which typically show distinct and predictable behaviors with time and depth. Under the circumneutral pH conditions (i.e., Offsetting of CO2 emissions at the Mount Keith Nickel Mine149pH of 5 to 8) that dominate within the tailings waters at Mount Keith, dissolution of bedrock carbonate minerals will proceed most readily for calcite, then dolomite, and least readily for magnesite (Palandri and Kharaka, 2004). Hydromagnesite is less resistant to dissolution in acidic solutions than magnesite (Königsberger et al., 1999). Hydromagnesite also dissolves rapidly in 5% acetic acid (Wilson et al., 2009a), an acid that has little effect on calcite, the least resistant of the bedrock carbonate minerals. Therefore, the near-constant abundance of less resistant bedrock carbonate minerals and the increasing abundance of hydromagnesite toward the surface of the tailings is very strong evidence that bedrock carbonate minerals are not being dissolved and remineralized as hydromagnesite in TSF2. Secondary Ca-bearing minerals (i.e., gypsum and anhydrite), which could form as a consequence of calcite (and dolomite) dissolution, are relatively uncommon in the tailings at Mount Keith, particularly when compared to the high abundances of secondary Mg-minerals. Only limited dissolution of calcite may be occurring, which indicates that bedrock carbonate minerals are not a significant source of carbon and magnesium in secondary hydromagnesite.Weathering of serpentine-group minerals and brucite represents an alternative source of magnesium in TSF2. The compositional data for these minerals display a consistent trend toward decreasing abundance with proximity to tailings surfaces (Fig. 7.7). Abundances of brucite and serpentine minerals are relatively constant below a depth of approximately 25 cm in tailings aged 0 to 7–8 years, but decline rapidly within the upper 25 cm of the tailings. The decrease is coincident with increasing abundance of hydromagnesite. This effect is particularly marked for brucite, for which the median abundance and MAD decline to 0 wt.% for all surficial tailings aged 0 to 7–8 years. Contrastingly, brucite is typically present at 1.0–2.5 wt.% abundance throughout the deeper, less weathered samples from TSF2.Median values for the abundance of serpentine decline by approximately 10 wt.% over the depth interval from 25 cm to 0 cm. This is, however, a proportionally minor decrease in the abundant serpentine minerals when compared to the complete loss of brucite. Bea et al. (2012) have shown that such trends can be caused by dilution of primary gangue minerals as secondary Offsetting of CO2 emissions at the Mount Keith Nickel Mine150minerals precipitate from tailings water, solutes and atmospheric gases. Significant input of new crystalline mass to previously unaltered tailings results in a dilution effect in mineral abundances. This occurs because Rietveld refinement results provide a relative measure of crystalline mass, normalized to 100% (w/w). Consequently, it is likely that significantly less than 10 wt.% of the serpentine is lost to weathering.Weathering of brucite is known to produce hydromagnesite and other hydrated Mg-carbonate minerals under conditions of temperature and pressure that prevail at the Earth’s surface (e.g., Hostetler et al., 1966; Xiong and Lord, 2008). Recent carbonation experiments (Zhao et al., 2010; e.g., Wilson et al., 2010; Hövelmann et al., 2012b; Harrison et al., 2013a; Chapter 2) indicate that brucite carbonation can be rapid at near-ambient pressures and temperatures, even in saline and alkaline brines similar to the tailings water at Mount Keith (Wilson et al., 2010; Harrison et al., 2013a; Chapter 2). Bea et al. (2012) and Beinlich and Austrheim (2012) have also documented brucite carbonation in mine tailings from TSF1 at Mount Keith and in a sub-arctic mine shaft, respectively. Interestingly, no brucite was observed during SEM imaging of hydromagnesite-rich surface samples, which is consistent with the brucite abundance decreasing to 0 wt.% near tailings surfaces in TSF2. Grguric (2003) describes a common texture within Mount Keith ore whereby brucite forms coronas with magnetite that encompass sulfide blebs. Thus, the absence of brucite is explained by the (otherwise perplexing) overgrowths of hydromagnesite on magnetite grains (observed in Fig. 7.3B). This textural feature may play a role in limiting sulfide oxidation and carbonate mineral dissolution within Mount Keith tailings. Textural evidence from SEM imaging of particularly hydromagnesite-rich samples (Fig. 7.3A and B) also shows platy hydromagnesite crystals seeding from grains of serpentine minerals, which is consistent with serpentine carbonation. Carbonation of serpentine-group minerals in alkaline mine tailings is known to produce similar textures in brucite-free mine tailings (Wilson et al., 2009a). These results imply that the observed decreases in the abundances of both serpentine and brucite are very likely the result of two processes: (1) dissolution to produce secondary Mg-bearing minerals such as Offsetting of CO2 emissions at the Mount Keith Nickel Mine151hydromagnesite, and (2) a dilution effect caused by incorporation of water and atmospheric gases into these newly formed minerals.7.6.2 Carbon	reservoir	fingerprintingThere are four potential sources of carbon in tailings from operating mines like Mount Keith: (1) atmospheric CO2, (2) bedrock carbon from gangue carbonate minerals, (3) carbon from industrial additives, and (4) organic carbon (i.e., mine camp sewage, mined organic sediments, decay of local biota, or microbial metabolism). Net sequestration of CO2 requires that there be a non-carbonate mineral source for cations and carbon. Thus, net sequestration of CO2 can only occur in mine tailings when atmospheric carbon (either directly from the atmosphere or indirectly from modern organic matter) is fixed within the crystal structures of carbonate minerals. XRD techniques are well suited to identifying and quantifying carbonate minerals, but these techniques cannot discern which minerals are trapping and storing atmospheric CO2 nor to what extent CO2 is being sequestered. Stable and radiogenic carbon and stable oxygen isotope data can be used to: (1) identify the sources for CO2 stored in carbonate minerals, (2) elucidate the mechanisms by which minerals form, and (3) trace the processes by which carbon is cycled.7.6.2.1 Fingerprinting with stable isotopesStable isotopic data are plotted as δ18OVSMOW and δ13CVPDB in Figure 7.8. Calculated and empirically derived fields for Mg-carbonate minerals plotting in isotopic equilibrium with metamorphic, atmospheric and organic sources of carbon (and oxygen) are also given in Figure 7.8. Please refer to the Supporting Information for a detailed description of the data used to define these fields.The majority of the stable isotopic data for magnesite fall within the range of values expected for precipitation from a carbonate-bearing metamorphic fluid (Fig. 7.8), which is consistent with the known metamorphic origin for this mineral at Mount Keith. One specimen Offsetting of CO2 emissions at the Mount Keith Nickel Mine152of a hydrotalcite-group mineral (carbonate-bearing iowaite) has a similar stable isotopic signature to the specimens of magnesite. Specimens of Ni-dolomite are also similarly depleted in 13C, but are significantly enriched in 18O relative to the other bedrock carbonate minerals. This is consistent with their origin as fracture coatings and vug linings in oxidized ore (Grguric, 2003). Soda ash, which was used historically as a process chemical, is slightly depleted in 13C relative to bedrock carbonate minerals and gives δ18O values that are comparable to the upper range for specimens of magnesite. Samples of bulk carbonate in mine tailings are consistently enriched in 18O relative to bedrock magnesite and many are enriched relative to Ni-dolomite. Analyses of hydromagnesite, done using selective acid extraction, indicate further enrichment in 18O (with some δ18OVSMOW values reaching 40‰) accompanied by a small enrichment in 13C of ~1‰ on average. The significant enrichment of 18O in hydromagnesite relative to magnesite suggests precipitation in an evaporative environment. This is consistent with our interpretation that hydromagnesite forms as a secondary alteration mineral.None of the data for hydromagnesite at Mount Keith fall within the fields defined for mineral precipitation in equilibrium with atmospheric CO2 or organic carbon (i.e., mine camp sewage, which is disposed of in TSF2). However, most of the data for hydromagnesite are consistent with precipitation in equilibrium with DIC in process/tailings water at approximately 25˚C. Because equilibrium carbon isotopic fractionation factors for hydromagnesite are not known, the fields for equilibrium precipitation of hydromagnesite from atmospheric CO2 and DIC were estimated using the fractionation factor for dypingite [Mg5(CO3)4(OH)2·5H2O], a structurally and chemically related mineral, which has been estimated to be 103lnαdypingite–HCO3- = (3.8 ± 1.2)‰ between 20˚C and 25˚C (Wilson et al., 2010; refer to the Supporting Information for details). Because seasonal temperature variations at Mount Keith are greater than the temperature range over which this fractionation factor was determined, it is reasonable to expect that the observed range of δ13C values should be broader than predicted here. Based on comparison with carbon isotopic fractionation factors between other carbonate minerals Offsetting of CO2 emissions at the Mount Keith Nickel Mine153and DIC, it is likely that inclusion of temperature dependence for dypingite-DIC fractionation would increase the breadth of this field to include all the remaining data for hydromagnesite. Although results are consistent with precipitation of hydromagnesite in equilibrium with process water and DIC, the ultimate source of carbon in DIC is not apparent from stable isotope data. Process water DIC and tailings water DIC at Mount Keith (-8.16‰ ≤ δ13CVPDB ≤ -4.60‰) are consistently depleted in 13C relative to DIC in equilibrium with the atmosphere (i.e., δ13CVPDB ≈ 0‰ at 25˚C). The observed carbon isotopic signature of DIC in process and tailings water may reflect (1) kinetic depletion of 13C during diffusion of atmospheric CO2 into solution (i.e., O’Neil and Barnes, 1971; Wilson et al., 2010; Harrison et al., 2013a; Chapter 2), (2) input of organic or soil CO2 into the tailings waters, (3) the use of 13C-depleted chemicals during processing, or (4) dissolution of 13C-depleted bedrock carbonate minerals.7.6.2.2 Fingerprinting with radiocarbonF14C values for CO2 selectively extracted from hydromagnesite range from 0.572 to 1.052 and all but one specimen has F14C > 0.8 (Fig. 7.9). A value of F14C = 1.06 reflects the 14C concentration of the atmosphere in 2006, the year of sampling at Mount Keith (Levin et al., 2008). Thus, the highest values in this range are consistent with a modern source for approximately 100% of the DIC-derived carbon in hydromagnesite. The average F14C value of hydromagnesite is F14Cav = 0.921 ± 0.145, which suggests that, on average, at least 87% of carbon in these samples has a modern source (relative to the 2006 atmosphere). The only known large reservoir for modern carbon at Mount Keith is the atmosphere and, based on our radiocarbon results, it can be concluded that almost all CO2 mineralized as hydromagnesite originated from this source.It is important to note that the excursion of some of these data below a F14C of unity suggests the possibility of: (1) partial dissolution and reprecipitation of bedrock carbonate minerals as hydromagnesite, (2) mixing with carbon-bearing process additives during mineral precipitation, or (3) contamination of CO2 from modern hydromagnesite with CO2 from Offsetting of CO2 emissions at the Mount Keith Nickel Mine15414C-free magnesite during selective acid extraction.The very small amount of 14C measured in Ni-dolomite sample 06MKNi-dol (F14C = 0.004) is consistent with the amount of atmospheric contamination typically observed in our vacuum line for samples of 14C-free bedrock carbonate minerals. The sample of soda ash has F14C = 0.052, which could reflect sourcing of this process additive from a Quaternary evaporite deposit (Warren, 2006). Six specimens of magnesite give values of 0.005 ≤ F14C ≤ 0.365. Two of these analyses are of highly pure samples of magnesite found in waste rock at sampling sites 53 and 54 (i.e., 06MK53 and 06MK54), which give F14C values between 0.005 and 0.009. These values are consistent with vacuum line contamination of samples with F14C = 0, as expected of bedrock magnesite that formed by carbonate alteration of serpentinized komatiite (Barrett et al., 1977; Grguric et al., 2006). The remaining analyses were done on CO2 selectively extracted from magnesite in four samples of hydromagnesite-bearing tailings (06MKG2-6-3-mags, 06MKG2-7-3-mags, 06MKG10-5-1-mags, and 06MKP9-4-mags). Magnesite in these samples gives F14C values between 0.024 and 0.365. These values are higher than expected for bedrock magnesite at Mount Keith, which should not contain detectable 14C. Similarly to the few anomalously low F14C values for hydromagnesite, this deviation toward enrichment in 14C within magnesite could reflect (1) partial conversion of secondary hydromagnesite to magnesite or (2) contamination of CO2 from 14C-free magnesite with modern CO2 from residual, unreacted hydromagnesite during acid extraction.The decomposition of hydromagnesite to magnesite has been inferred from observation of hydromagnesite playas on the Cariboo Plateau, interior British Columbia, Canada (e.g., Renaut and Long, 1989; Renaut, 1990; Renaut and Stead, 1991). Zhang et al. (2000) estimate that the conversion of hydromagnesite to magnesite requires tens to hundreds of years. However, it may be possible that some amount of recently precipitated hydromagnesite may undergo a transformation to magnesite on the decadal scale in the mine tailings at Mount Keith. In fact, the specimen of magnesite with the highest F14C value (i.e., 06MKG10-5-1-mags with F14C=0.365) comes from the 10-year old surface of TSF1. In this case, the mixing Offsetting of CO2 emissions at the Mount Keith Nickel Mine155trend for magnesite in Figure 7.9 could result from precipitation of hydromagnesite and its decomposition to magnesite. Alternatively, the finely intergrown nature of secondary hydromagnesite (i.e., as a cement between grains of primary tailings minerals) may be preventing complete reaction with phosphoric acid during sample processing. Subsequent processing of these samples at higher temperature and for a longer period of time would then dissolve not only the desired magnesite but also the residual, unreacted hydromagnesite. This would result in higher values of F14C than anticipated for magnesite. Furthermore, the finest fractions of less acid resistant bedrock carbonate minerals such as calcite, dolomite, and hydrotalcite-group minerals may be reacting to some extent during extraction of CO2 from hydromagnesite. This would produce a dilution in 14C, resulting in artificially low values of F14C for hydromagnesite and the incipient mixing trend in Figure 7.9. Thus, we interpret this trend as an artifact of the technique used for sample preparation, which suggests that some F14C values obtained for hydromagnesite are underestimated.These observations are consistent with the results of previous isotopic studies, which conclude that dissolution of bedrock carbonate minerals does not contribute significantly to production of secondary Mg-carbonate minerals in neutral to alkaline mining environments (e.g., Wilson et al., 2009a; Wilson et al., 2011; Beinlich and Austrheim, 2012). Importantly, our radiocarbon results indicate that stable isotope data for Mount Keith carbonate minerals cannot be used to uniquely identify the source of carbon in hydromagnesite. Although radiocarbon data clearly identify a predominantly atmospheric source for carbon, δ13C data indicate only that hydromagnesite is precipitating in equilibrium with DIC, but out of equilibrium with atmospheric CO2 gas. Experiments by Wilson et al. (2010) and Harrison et al. (2013a; Chapter 2) demonstrate that kinetic depletion of 13C occurs during dissolution of atmospheric CO2 and hydroxylation of aqueous CO2 to form bicarbonate in saline and moderately alkaline (pH < 11) solutions designed to emulate tailings water at Mount Keith. Mg-carbonate minerals that form from such carbon-limited solutions precipitate in isotopic equilibrium with 13C-depleted Offsetting of CO2 emissions at the Mount Keith Nickel Mine156DIC (Figs. 7.8 and 7.9). Thus, although hydromagnesite at Mount Keith acts as a store for atmospheric CO2, this is not immediately apparent due to the impact of diffusion on the δ13C values of DIC. Most significantly, the stable isotopic data indicate that dissolution of atmospheric CO2 into mine tailings water is kinetically limited, which suggests that carbon mineralization could be accelerated by increasing t