Dissolved CO2 effect on the reactivity of the Hontomín reservoir rocks (limestone and sandstone)

DISSOLVED CO2 EFFECT ON THE REACTIVITY OF THE HONTOMÍN RESERVOIR ROCKS (LIMESTONE AND SANDSTONE) María García Ríos PhD Thesis Department of Geotechnical Engineering and Geo-Sciences (ETCG) Technical University of Catalonia (UPC) Supervisors: Dr. Jordi Cama Dra. Linda Luquot Dr. Josep M. Soler Institute of Environmental Assessment and Water Research (IDAEA-CSIC) Acta de calificación de tesis doctoral Curso académico: Nombre y apellidos: MARIA OLIMPIA GARCIA RIOS Programa de doctorado: Unidad estructural responsable del programa Resolución del Tribunal Reunido el Tribunal designado a tal efecto, el doctorando / la doctoranda expone el tema de la su tesis doctoral titulada ____________________________________________________________________________________ __________________________________________________________________________________________. Acabada la lectura y después de dar respuesta a las cuestiones formuladas por los miembros titulares del tribunal, éste otorga la calificación: NO APTO APROBADO NOTABLE SOBRESALIENTE (Nombre, apellidos y firma) Presidente/a (Nombre, apellidos y firma) Secretario/a (Nombre, apellidos y firma) Vocal (Nombre, apellidos y firma) Vocal (Nombre, apellidos y firma) Vocal ______________________, _______ de __________________ de _______________ El resultado del escrutinio de los votos emitidos por los miembros titulares del tribunal, efectuado por la Escuela de Doctorado, a instancia de la Comisión de Doctorado de la UPC, otorga la MENCIÓN CUM LAUDE: SÍ NO (Nombre, apellidos y firma) Presidente de la Comisión Permanente de la Escuela de Doctorado (Nombre, apellidos y firma) Secretario de la Comisión Permanente de la Escuela de Doctorado Barcelona a _______ de ____________________ de __________ TECHNICAL UNIVERSITY OF CATALONIA (UPC) DEPARTMENT OF GEOTECHNICAL ENGINEERING AND GEO-SCIENCES (ETCG) Dissolved CO2 effect on the reactivity of the Hontomín reservoir rocks (limestone and sandstone) Thesis presented by María García Ríos Work conducted in the Institute of Environmental Assessment and Water Research (IDAEA-CSIC) under the supervision of Dr. Jordi Cama i Robert Institute of Environmental Assessment and Water Research (IDAEA), CSIC Dra. Linda Luquot Institute of Environmental Assessment and Water Research (IDAEA), CSIC Dr. Josep M. Soler Institute of Environmental Assessment and Water Research (IDAEA), CSIC Barcelona, February 2015 This thesis has been funded by CIUDEN (project ALM11/009), Spanish Government Project CGL2010-20984-CO2-01 and the PANACEA project (European Community’s Seventh Framework Programme FP7/2007-2013 under grant agreement number 282900). to The Rivers Abstract A test site for CO2 geological storage is situated in Hontomín (Burgos, northern Spain) with a reservoir rock that is mainly composed of limestone (80-85%) and sandstone (15-20%). The reservoir rock is a deep saline aquifer that is covered by a very low permeability formation which acts as a cap rock. During and after CO2 injection, since the resident groundwater contains sulfate, the resulting CO2-rich acid solution gives rise to the dissolution of carbonate minerals (calcite and dolomite) and secondary sulfate-rich mineral precipitation (gypsum or anhydrite) may occur. These reactions that may imply changes in the porosity, permeability and pore structure of the repository could vary the CO2 storage capacity and injectivity of the reservoir rock. Therefore, better knowledge about the overall process of gypsum precipitation at the expense of carbonate mineral dissolution in CO2-rich solutions and its implications for the hydrodynamic properties of the reservoir rocks is necessary. A first aim of this thesis is to better understand these coupled reactions by assessing the effect that P, pCO2, T, mineralogy, acidity and solution saturation state exert on these reactions. To this end, experiments using columns filled with crushed limestone or dolostone are conducted under different P–pCO2 conditions (atmospheric: 1–10-3.5 bar; subcritical: 10–10 bar; and supercritical: 150–34 bar), T (25, 40 and 60 °C) and input solution compositions (gypsumundersaturated and gypsum-equilibrated solutions). The CrunchFlow and PhreeqC (v.3) numerical codes are used to perform 1D reactive transport simulations of the experiments to evaluate mineral reaction rates in the system and quantify the porosity variation along the column. Within the range of P–pCO2 and T of this study only gypsum precipitation takes place and this only occurs when the injected solution is equilibrated with gypsum. Under the P–pCO2–T conditions, the volume of precipitated gypsum is smaller than the volume of dissolved carbonate minerals, yielding always an increase in porosity (Δ up to ≈ 4%). A decrease in T favors limestone dissolution regardless of pCO2 owing to increasing undersaturation with decreasing temperature. However, gypsum precipitation is favored at high T and under atmospheric pCO2 conditions but not at high T and under 10 bar of pCO2 Resumen Una planta piloto para el almacenamiento geológico de CO2 se ubica en Hontomín (Burgos). El reservorio es un acuífero salino profundo que se compone principalmente de roca caliza (80-85%) y arenisca (15-20%). Éste está situado entre dos capas de muy baja permeabilidad que actúan como rocas sello. La disolución de CO2 en el agua presente en el reservorio provocará una disminución del pH y, en consecuencia, la disolución de los carbonatos presentes en el reservorio. Además, como la solución residente es rica en sulfato, es posible la precipitación de minerales secundarios (yeso o anhidrita). Estas reacciones pueden provocar cambios en la porosidad, permeabilidad y estructura de poro del reservorio que, a su vez, pueden hacer variar su inyectabilidad y capacidad de almacenamiento. Por todo ello, es necesario profundizar en el conocimiento sobre los procesos acoplados de precipitación de yeso y disolución de carbonatos (calcita y dolomita) en soluciones ricas en CO2 disuelto y sus implicaciones en las propiedades hidrodinámicas de la roca reservorio. Un primer objetivo de esta tesis es comprender mejor estas reacciones acopladas mediante la evaluación del efecto que ejercen la presión P, la presión parcial de CO2 pCO2, la temperatura T, la mineralogía, la acidez y el estado de saturación de la solución sobre estas reacciones. Con este objetivo, se han realizado una serie de experimentos utilizando columnas llenas de roca caliza o dolomía triturada bajo diferentes condiciones de P-pCO2 (atmosférica: 1-10-3.5 bar; sub-crítica: 10-10 bar, y supercrítica: 150-34 bar), T (25, 40 y 60 ° C) y composición de la solución de entrada (soluciones sub-saturadas o equilibradas con respecto al yeso). Los códigos numéricos CrunchFlow y PhreeqC (v.3) se han utilizado para realizar simulaciones de transporte reactivo de los experimentos en columna con el objetivo de evaluar las velocidades de reacción en el sistema y cuantificar la variación de la porosidad a lo largo de la columna. En las condiciones de P-pCO2-T estudiadas, la precipitación de yeso únicamente tiene lugar cuando la solución inyectada está en equilibrio con yeso. Además, el volumen de yeso precipitado es menor que el volumen de carbonato disuelto, originando siempre un aumento de porosidad (Δ hasta ≈ 4%). Una disminución en la T favorece la disolución de la caliza independientemente de la pCO2 debido al aumento de la sub-saturación. Sin embargo, hay un aumento en la precipitación de yeso a alta T para condiciones atmosféricas, viéndose el efecto contrario para condiciones sub-críticas. El aumento de la pCO2 conlleva un aumento en la disolución de caliza, hecho que es directamente atribuido al efecto del pH, que es más ácido a mayor pCO2. La disolución de caliza conlleva un retraso en la precipitación de yeso (largo tiempo de inducción), lo contrario que ocurre con la disolución de dolomía que promueve una rápida precipitación de yeso. Además, debido a la lenta cinética de disolución de la dolomita con respecto a la de la calcita, el volumen de mineral disuelto y el aumento de porosidad son mayores en los experimentos con caliza bajo todas las condiciones de pCO2 estudiadas. La disolución del carbonato se produce a lo largo de toda la columna cuando la pCO2 es alta (10 and 34 bar) y, en cambio, se localiza en la entrada de la columna bajo condiciones atmosféricas. Esta diferencia es debida a la capacidad tampón del ácido carbónico, ya que mantiene el pH alrededor de 5 y mantiene la solución sub-saturada con respecto a la calcita y a la dolomita a lo largo de la columna. Las simulaciones de transporte reactivo (1D) reproducen los datos experimentales (disolución de carbonato y precipitación de yeso para las diferentes condiciones de P-pCO2-T). En base a las leyes de velocidad de reacción que se encuentran en literatura, se han usado los valores de las áreas reactivas para realizar el ajuste del modelo a los datos experimentales. Estos valores son bastante inferiores a los inicialmente calculados a partir de las áreas geométricas, debido a que las reacciones de disolución estás controladas por el transporte. La roca reservorio en Hontomín está significativamente fracturada. Por lo tanto, entender los cambios en las propiedades hidrodinámicas de las fracturas, inducidos por reacciones de disolución/precipitación, es esencial para predecir los posibles flujos subterráneos como fugas, inyectabilidad o producción de fluidos. Teniendo en cuenta esto, el segundo objetivo de esta tesis es caracterizar la evolución de fracturas que estén en contacto con soluciones ricas en CO2 a diferentes caudales. Además, se compara la respuesta geoquímica a la inyección de CO2 de las dos rocas principales del reservorio (caliza y arenisca). Para ello, se realiza un conjunto de experimentos de percolación que consisten en inyectar soluciones ricas en CO2 (sin sulfato y ricas en sulfato) a través de rocas de caliza y arenisca fracturadas. Todos ellos bajo P = 150 bar y T = 60 ºC y a diferentes caudales comprendidos entre 0,2 y 60 mL/h. La variación del volumen de fractura, inducida por la disolución de calcita y la precipitación de yeso, se mide mediante micro-tomografía de rayos X (XCMT) y la química de la solución. Se evalúa también la influencia del caudal en la evolución de la fractura, obteniéndose que aumentando el caudal el volumen de calcita disuelta por unidad de tiempo aumenta, confirmando así que la disolución en la fractura está controlada por el transporte. También se observa que la formación de geometrías más uniformes a caudales más rápidos puede favorecer la disolución de calcita. Los patrones de disolución varían de ‘face dissolution’ a ‘wormhole’ y a ‘uniform dissolution’ a medida que aumenta el caudal (es decir, números de Péclet Pe de 1 a 346). Se mide también la variación de permeabilidad de la fractura encontrando que su evolución depende del caudal y del patrón de disolución desarrollado. La permeabilidad de la fractura siempre aumenta independientemente del contenido de sulfato de la solución de entrada. En base a los resultados experimentales obtenidos con las rocas de Hontomín, se evalúa cuál sería el contexto geológico más favorable en el reservorio para la inyección y almacenamiento del CO2. Además, se realizan modelos de transporte reactivo (2D) de los experimentos de percolación con fracturas para estimar los parámetros cinéticos y de flujo. Agraïments Quin moment!...això és el que diria un gran amic meu! Ja ha arribat! i he de reconèixer que el camí no ha estat fàcil. Un camí on hi ha tants forats que agraeixes sincerament que algú et doni un cop de mà. I per això estic aquí, per agrair a tots aquells que en el algun moment s'han parat al costat meu, han vist que necessitava ajuda i me l’han ofert. Ajuda de tota mena, és clar! ....durant les hores llargues del dia, quan estem treballant: gràcies al Víctor, al Francesco, a la Cris, a la Gaby, a l'Anna Russian i a la Yoar per fer que aquestes hores passin de manera més lleugera, per distreure’m, fer-me riure i per aguantar-me en aquells moments on sembla que tot és negre. A l'Ester per la seva disponibilitat a l'hora d'ajudar-te en qualsevol problema o de passar una estona agradable xerrant. Al Carles Ayora i al Josep Soler per la seva ajuda incondicional. A la Linda per la seva ajuda amb condicions (li dec molta pasta!) però sempre amable i profitosa. ...quan arribes a casa tot cansat: gràcies al Josevi per ser tant pacient i agradable amb mi sempre, per treure'm un somriure i per calmar el monstre que tinc a dintre que, a vegades, vol sortir. Abans del Josevi altres m’han patit i/o gaudit. Agrair també a tots aquells que han compartir pis amb mi durant tots aquests anys menjant-se amb patates els dies bons i també els dolents: gràcies a la gent de Sant Eusebi (el Davik, el Kike, l’Elisa i la Dèlia) i el meu estimat amic Lluis. ...quan és divendres i vols trobar-te amb la teva gent, la de tota la vida o no, on tot val i tot s'entén perquè et fan sentir sempre com a casa (la família!): gràcies al Rubén, a la Laura, a l'Anna, a la Rosa, a la Mònica, al Servando i a la Marta. Gràcies també a les meves nenes amb les que no només he compartit això sinó quasi tota una vida: gràcies a la Patri, a la Susana i a la Lorena. Gràcies també a l’Ana pels moments tant especials juntes. ...en tots els moments: gràcies a la meva mare, el meu pare i la meva cosina Txell per formar una família acollidora i agradable i per sempre donar-me suport. I sobretot gràcies al meu director de tesi, el Jordi Cama, perquè sense ell tot això no hagués estat possible. Gràcies per totes les vegades que m’has vist fluixa i m’has aixecat i també per aquelles que m’has vist forta i has compartit amb mi el moment. Ha estat un plaer! Table of contents PART I: INTRODUCTION and MATERIALS AND METHODS .................................... 1 Chapter 1 Introduction ....................................................................................................................... 3 1.1 Background and objectives ..................................................................................................... 3 1.2 Thesis outline ........................................................................................................................ 11 Chapter 2 Materials and methods .................................................................................................... 13 2.1 Experimental methodology ......................................................................................................... 13 2.1.1 Sample characterization and analytical techniques .............................................................. 13 2.1.2 Injected solutions ................................................................................................................. 17 2.1.3 Experimental setups and conditions ..................................................................................... 20 2.1.3.1. Atmospheric pressure setup (P = 1 bar) ....................................................................... 21 2.1.3.2. Subcritical pressure setup (P = 10 bar) ........................................................................ 21 2.1.3.3. Supercritical pressure setup (P = 150 bar) ................................................................... 23 2.1.4 Mass transfer calculations .................................................................................................... 24 2.1.5 Permeability changes ........................................................................................................... 26 2.2 Reactive transport modeling ....................................................................................................... 27 2.2.1. Description of the CrunchFlow reactive transport code...................................................... 27 2.2.2 One-dimensional model (Part II: crushed rock) .................................................................. 29 2.2.2.1. Numerical discretization .............................................................................................. 29 2.2.2.2 Rock and solution composition ..................................................................................... 30 2.2.2.3 Flow and transport properties ....................................................................................... 30 2.2.2.4 Thermodynamic and kinetic data .................................................................................. 31 2.2.2.5 Reaction rates ................................................................................................................ 32 2.2.3 Two-dimensional model (Part III: fractured cores) ............................................................ 32 2.2.3.1 Numerical discretization ............................................................................................... 33 2.2.3.2 Rock and solution composition ..................................................................................... 35 2.2.3.3 Flow and transport parameters ...................................................................................... 35 2.2.3.4 Thermodynamic and kinetic data .................................................................................. 37 2.2.3.5 Reaction rates ................................................................................................................ 37 PART II: CRUSHED ROCK. .......................................................................................... 39 Chapter 3 Interaction between CO2-rich sulfate solutions and carbonate rocks: column experiments and 1D modeling ..................................................................................................................... 41 3.1 Introduction ................................................................................................................................. 41 3.2 Results ......................................................................................................................................... 42 3.2.1 Experiments under atmospheric conditions (P = 1 bar; pCO2 = 10-3.5 bar) .......................... 44 3.2.1.1 H2SO4 solution (s) ......................................................................................................... 44 3.2.1.2 Acidic gypsum-equilibrated solution (a2.1 and a3.5) ................................................... 46 3.2.2 Experiments under subcritical conditions (P = pCO2 = 10 bar) ........................................... 47 3.2.2.1 Gypsum-undersaturated solution (gp-u) ....................................................................... 47 3.2.2.2 Gypsum-equilibrated solution (gp-e) ............................................................................ 47 3.2.3 Experiment under supercritical conditions (P = 150 bar; pCO2 = 34 bar) ........................... 49 3.3 Discussion ................................................................................................................................... 50 3.4 Summary and conclusions .......................................................................................................... 56 PART III: FRACTURED CORES ................................................................................... 59 Chapter 4 Influence of the flow rate on dissolution and precipitation features during percolation experiment with fractured limestone and sandstone cores ............................................... 61 4.1 Introduction ................................................................................................................................. 61 4.2 Results ......................................................................................................................................... 62 4.2.1 Initial fracture characterization ............................................................................................ 63 4.2.2 Aqueous chemistry ............................................................................................................... 65 4.2.3 Permeability ......................................................................................................................... 66 4.2.4 Identification of dissolution and precipitation processes ..................................................... 69 4.2.4.1. Limestone samples ....................................................................................................... 70 4.2.4.2. Sandstone samples ....................................................................................................... 75 4.3 Discussion ................................................................................................................................... 79 4.3.1 Fracture volume calculated from mass balance and XCMT ................................................ 79 4.3.2 Influence of flow rate on reaction ........................................................................................ 82 4.3.3 Dissolution patterns.............................................................................................................. 86 4.3.4 Permeability changes during fracture dissolution ................................................................ 90 4.4 Summary and conclusions .......................................................................................................... 92 Chapter 5 Dissolved CO2 effect on two fractured reservoir rocks: comparison and 2D modeling ..... 95 5.1 Introduction ................................................................................................................................. 95 5.2 The role of silicate minerals on the CO2 storage capacity and injectivity .................................. 96 5.3 (2D) Reactive transport modeling ............................................................................................... 99 5.3.1 Face dissolution ................................................................................................................... 99 5.3.2 Wormhole .......................................................................................................................... 101 5.3.3 Uniform dissolution ........................................................................................................... 104 5.3.4 Flow and reaction kinetics parameters ............................................................................... 105 5.4 Summary and conclusions ........................................................................................................ 106 PART IV: CONCLUSIONS............................................................................................ 107 Chapter 6 Conclusions ..................................................................................................................... 109 References ...................................................................................................................... 115 (dashed lines): Variation in the experimental and simulated Ca (a) and S (b) concentration versus time. r indicates the initial radius of the cylinder (see text)........................................ 102 Fig. 5.6 Experiment L1-gp-e (wormhole); simulation with rectangular + cylindrical coordinates and fixed flow: (a,b) Variation in the experimental and simulated Ca and S concentration versus time and variation in simulated porosity (c) and mineral content (d) with distance normal to fracture..................................................................................................... 103 Fig. 5.7 Experiment S60-no-s (uniform dissolution); simulation with rectangular coordinates and flow update: (a) Variation in the experimental and simulated Ca concentration versus time and (b) simulated porosity variation with the distance normal to fracture. Grey area in (b) indicates the zone with high porosity (96%) measured by SEM along most the fracture length...................................................................................................................................... 104 Fig. B.1 Variation of experimental (Exp) and simulated (Sim) output pH (a) and output concentration of Ca (b), Mg (c) and S (d) with time in limestone (L; in green) and dolostone (D; in blue) column experiments (L25-atm-s and D25-atm-s, respectively). Solid lines indicate input solution except for Mg which is smaller than 3 10-4 mol/kgw (Table 2.2). Dashed and dotted lines indicate simulated values of limestone and dolostone column experiments, respectively. ...................................................................................................... 140 Fig. B.2 Variation of the experimental (Exp) and simulated (Sim) increase in Ca concentration (a) and output pH (b) with time in limestone column experiments at 25 ºC (in green; L25-atm-a3.5) and 60 ºC (in red; L60-atm-a3.5). Solid line in (b) shows input pH. Dashed and dotted lines show simulated values at 25 and 60 ºC, respectively. .................... 141 Fig. B.3 Variation of the experimental (Exp) and simulated (Sim) increase in Ca (a) and output pH (b) with time in limestone column experiments at 25 ºC (in green; L25-10-gp-u) and 40 ºC (in orange; L40-10-gp-u). Colored solid lines in (a) represent the Ca equilibrium with calcite and black-dashed line in (b) indicates input pH. Dashed and dotted lines show simulated values at 25 and 40 ºC, respectively. ..................................................................... 141 Fig. B.4 Experimental variation of volume of dissolved limestone VL-diss (a) and porosity (b) versus time in experiments performed at 25 ºC (in green; L25-10-gp-u and L25-10-gp-e), 40 ºC (in orange; L40-10-gp-u and L40-10-gp-e) and 60 ºC (in red; L60-10-gp-e). Solid and dashed lines indicate experiments with gypsum-equilibrated and gypsum-undersaturated solutions, respectively. ........................................................................................................... 142 List of tables Table 1.1 Average composition of the Hontomín groundwater (± 10 %) in terms of total concentration (mol/kgw) and pH. It was provided by CIUDEN after extraction from the H-2 well. ............................................................................................................................................ 8 Table 2.1 Rock samples: origin and mineralogical composition (wt.%). See also Fig. 1.2. ... 14 Table 2.2 Injected solutions: amount of reagents, experimental conditions, average concentration (from ICP-AES), experimental pH, and calculated saturation indexes (SI), pH and ionic strength (I). ............................................................................................................... 19 Table 2.3 Reactive surface area (Am) and input boundary conditions (SI, I and pH) used in simulations under atmospheric conditions (CrunchFlow code)............................................... 31 Table 2.4 Reactive surface area (Am) and input boundary conditions (SI, I and pH) used in simulations under subcritical and supercritical conditions (CrunchFlow and PhreeqC (v.3) codes). ...................................................................................................................................... 32 Table 2.5 Initial mineralogical composition of both the rock matrix and the high-permeability zone (fracture) and input solution used in the 2D simulations. ............................................... 36 Table 2.6 Initial transport properties assumed in the 2D calculations. ................................... 37 Table 3.1 Experimental conditions and results (pH, volume of dissolved rock and precipitated mineral, porosity variation, and measured and calculated loss of mass) of the column experiments. ................................................................................................................ 43 Table 4.1 List of the percolation experiments. ........................................................................ 62 Table 4.2 Measured (weighted) mass of fractured core (Mmeas), fracture permeability (k), and fracture geometry (a and V) obtained by hydraulic measurement (ah and Vh), XCMT (aXr and VXr) and SEM (as and Vs) at initial time (t = t0). ...................................................................... 63 Table 4.3 Measured mass, measured and calculated loss of mass and variation in fracture volume determined from aqueous chemistry and XCMT at the end of the experimental runs (t = tf). .......................................................................................................................................... 81 Table 4.4 Péclet (Pe) and Damköhler (Da) numbers and net reaction rates expressed as volume of dissolved calcite, precipitated gypsum and variation in fracture volume per time and injected volume. ................................................................................................................ 85 Table A.1 Experimental and input boundary conditions, transport parameters, numerical discretization and rock composition used in simulations of column experiments under atmospheric CO2 conditions. ................................................................................................. 130 Table A.2 Experimental and input boundary conditions, transport parameters, numerical discretization and rock composition used in simulations of column experiments under subcritical and supercritical CO2 conditions. ......................................................................... 131 Table A.3 Equilibrium constants (log K) and stoichiometric coefficients for equilibria in solution (column experiments and fractured core experiments L0.2-gp-e and L1-gp-e). Reactions are written as the destruction of 1 mol of the species in the first column. * indicates species used in the atmospheric CO2 experiments with H2SO4 input solution (s)................. 132 Table A.4 Equilibrium constants (log K) and stoichiometric coefficients for mineral reactions (column experiments). Reactions are written as the dissolution of 1 mol of mineral. .......... 133 Table A.5 Parameters for the mineral reaction rate laws (column experiments). All parameters are from Palandri and Kharaka (2004), except for the coefficients m1 and m2 for calcite, which are based on the data reported by Xu et al. (2012). ........................................ 133 Table A.6 Experimental and input boundary conditions, fracture dimensions, numerical discretization, transport parameters and rock and fracture composition used in simulations of fractured core experiments. .................................................................................................... 135 Table A.7 Equilibrium constants (log K) and stoichiometric coefficients for equilibria in solution in fractured core experiment S60-no-s. Reactions are written as the destruction of 1 mol of the species in the first column .................................................................................... 136 Table A.8 Equilibrium constants (log K) and stoichiometric coefficients for mineral reactions (fractured core experiments). Reactions are written as the dissolution of 1 mol of mineral. 137 Table A.9 Parameters for the mineral reaction rate laws (fractured core experiments). Parameters for calcite, gypsum and quartz are from Palandri and Kharaka (2004), except for the coefficients m1 and m2 for calcite, which are based on the data reported by Xu et al. (2012). Parameters for microcline are from Bandstra et al. (2008). ...................................... 137 Part I: Introduction and Materials and Methods Chapter 1 Introduction 1.1 Background and objectives Energy is a key input into almost all activities and is fundamental to society wellbeing. However, as recently reported by the International Energy Agency (IEA), its use represents by far the largest source of greenhouse gas (GHG) emissions (83% in 2011; IEA 2013). Smaller shares correspond to agriculture, producing mainly CH4 and N2O from domestic livestock and rice cultivation, and to industrial processes not related to energy, producing mainly fluorinated gases and N2O. Fossil fuels currently supply 81% of the energy consumed globally and CO2 resulting from the oxidation of carbon in fuels during combustion dominates the total GHG emissions (65% in 2010; Fig. 1.1). Total anthropogenic GHG emissions have risen more rapidly from 2000 to 2010 than in the previous three decades and have reached human history record in 2010 (49 ± 4.5 GtCO2eq/yr; Fig. 1.1). The global economic crisis 2007-2008 has temporarily reduced global emissions but not changed the longer-term trend. Whereas more recent data are not available for all gases, initial evidence suggests that growth in global CO2 emissions from fossil fuel combustion has continued with emissions increasing by about 3% between 2010 and 2011 and by about 1–2% between 2011 and 2012 (IPCC, 2014). These high levels of GHG emissions are the direct source of the global climate change. The United Nations Framework Convention on Climate Change (UNFCCC) provides a structure for intergovernmental efforts to tackle the challenge posed by climate change. The 4 Chapter 1: Introduction Convention’s ultimate objective is to stabilize GHG concentrations in the atmosphere at a level that would prevent dangerous anthropogenic interference with the climate system. Fig. 1.1 Total annual anthropogenic GHG emissions (GtCO2eq/yr) by groups of gases 1970-2010: CO2 from fossil fuel combustion and industrial processes; CO2 from Forestry and Other Land Use (FOLU); methane (CH4); nitrous oxide (N2O); fluorinated gases covered under the Kyoto Protocol (HFC-PFC-SF6). Average annual growth rate for each decade is highlighted with the brackets (IPCC, 2014). The Conference of Parties (COP) in 2010 further recognized that deep cuts in global GHG emissions are required, with a view to hold the increase in global average temperature below 2 ºC above preindustrial levels (IEA, 2013). Consequently, given that continued global economic growth will further increase energy consumption needs, meeting climate challenge will require changes in energy consumption and in the technologies used to produce energy (Global CCS Institute, 2013). However, the deployment of existing and new low-carbon technologies is not an immediate process and may take several decades. Therefore, bridge technologies are needed. Carbon Capture and Sequestration (CCS) may indeed be one of such bridge technologies that will permit the reduction of CO2 emissions over the coming decades while a change in the energy market occurs (IEA, 2010). 5 CCS technology consists of the separation of CO2 from industry and energy-related sources, transport to a storage location and long-term isolation from the atmosphere. Geological storage options for CO2 include depleted oil and gas reservoirs, use of CO2 in enhanced oil recovery (CO2-EOR), use of CO2 in enhanced coal bed methane recovery, deepunminable coal seams and deep-saline aquifers. The latter have received particular attention due to their high CO2 storage capacity and wide availability throughout the world (Bachu and Adams, 2003). However, CO2-EOR projects currently dominate geological storage. Nowadays, all twelve projects in operation are in industries that separate CO2 as part of their normal procedures – natural gas processing, fertilizer production, hydrogen production, and synthetic natural gas – and nine of these projects use the captured CO2 for enhanced oil recovery (EOR). The remainder is dedicated to storage in geological formations containing brine or non-potable water (Global CCS Institute, 2013). The storage capacity of the twelve projects in operation totals 25 Mt/yr and the remaining projects in planning (45 projects) have the potential to store 84 Mt/yr. The European Energy Programme for Recovery (EEPR) was established in 2009 to address both Europe’s economic crisis and European energy policy objectives. Almost €4 billion were assigned to co-finance EU energy projects that would boost the economic recovery, increase the security of energy supply and contribute to the reduction of greenhouse gas emissions. The three sectors meeting these conditions were gas and electricity infrastructure, offshore wind energy and carbon capture and storage (CSS) projects. One of these CCS projects, namely The Compostilla project, is leaded by a three partner consortium (Endesa, CIUDEN (CIUDad de la ENergía foundation) and Foster Wheeler Energia Oy (FWEOy)), and is located in Ponferrada, northern Spain. The project is in charge of three Technology Development Plants (TDPs) at pilot scale: the CO2 capture and transport TDPs in Cubillos del Sil (León, Spain) and the CO2 geological storage TDP in Hontomín (Burgos, Spain). This thesis falls within the context of CO2 geological storage in Hontomín. The Hontomín reservoir formation for CO2 storage is a deep saline aquifer in Mesozoic sedimentary sequences that is covered by a very low permeability formation which acts as a cap rock. At the storage TDP, it is programmed to inject less than 100000 tonnes of CO2 into a 1500-meter-deep dome-like saline aquifer over a period of five years. Five existing wells (H1-5) were used to characterize the structure of the geological formation and two new wells were drilled, one for CO2 injection (H-I) and another one for monitoring (H-A). 12 Chapter 1: Introduction Two appendixes are supplied:  Appendix A provides the model parameters used to perform the 1D simulations of the column experiments and the 2D simulations of the percolation experiments (fractured cores).  Appendix B includes additional experimental and modeling results from the column experiments. Chapter 2 Materials and methods 2.1 Experimental methodology This section describes the experimental procedure followed to perform the column experiments presented in Part II: crushed rock and the percolation experiments with fractured cores shown in Part III: fractured cores. Analytical techniques used to characterize rock samples and injected solutions are detailed. In addition, the different experimental equipments used under different P conditions and the mass transfer and fracture permeability calculations are described. 2.1.1 Sample characterization and analytical techniques Four different sedimentary rock samples were used in this study: vuggy limestone, dolostone, oolitic limestone and sandstone. The vuggy limestone, the oolitic limestone and the sandstone were provided by CIUDEN and belong to the Bercedo series and different formations in the Hontomín reservoir rock (Fig 1.2; Table 2.1; Pujalte et al., 1998). The dolostone was provided by the Department of Mineralogy (Faculty of Geology, Barcelona University) and comes from Peñarroya, Teruel (Spain). The mineralogical composition of the samples was obtained by X-ray diffraction (XRD), performed using a Bruker diffractometer model D-5005 with Cu K- α1 radiation, and Rietveld analysis (Young, 1995) (Table 2.1). 14 Chapter 2: Materials and methods The vuggy limestone and the dolostone rock samples were used to carry out the column experiments that are presented in Part II: crushed rock and the oolitic limestone and the sandstone rock samples were used to perform the percolation experiments with fractured cores that are shown in Part III: fractured cores (Table 2.1). Table 2.1 Rock samples: origin and mineralogical composition (wt.%). See also Fig. 1.2. Rock samples used in the column experiments (Part II: crushed rock) were crushed to a grain size between 1 and 2 mm for the atmospheric and 10 bar pCO2 experiments. For the 34 bar pCO2 experiment, the limestone was ground to a grain size between 250 and 500 μm owing to the smaller diameter of the column (Fig. 2.1). Thereafter, in order to remove microparticles due to grinding, the crushed and ground samples were washed three times with pH 1 solution (HCl) and three times with deionized water. Finally, the washed samples were dried in the oven at 40 °C and were put into cylindrical columns. X-ray fluorescence analysis of the vuggy limestone and the dolostone was performed using a Bruker spectrometer model AXS-S2 Ranger to identify minor-element composition. In the limestone sample, Ca and Mg were the major elements, Si (0.37 wt.%) and Fe (0.20 wt.%) appeared as minor components and S and Sr as trace elements (< 0.1 wt.%). In the dolostone sample, Al, Si, S, Mn and Cl appeared as minor components (from 0.1 wt.% to 1 wt.%) and Cu, Pb and K as trace elements. Fe (1.48 wt.%) was the only element, besides Ca and Mg, with a concentration higher than 1 wt.%. Sample Experiments Series Formation Calcite Dolomite Quartz Microcline Bercedo Puerto de la Palombera Bercedo Sopeña Bercedo Areniscas del Río Polla *coarse-grained sandy limestone according to Mount (1985). - 100 - - Vuggy Limestone Dolostone 90.7 9.3 - Mineralogical composition (wt.%) (XRD and Rietveld) Oolitic limestone Sandstone* Part III: fractured cores 65.7 - 27.8 6.5 100 - - Part II: crushed rock - - - 15 Fig. 2.1 Rock samples used in the column experiments described in Part II: crushed rock. Top: vuggy limestone (core, crushed grains (1-2 mm)/ground grains (250-500 μm) and SEM image); bottom: dolostone (fragments, crushed grains (1-2 mm) and SEM image). Flow-through experiments with the dolostone rock sample were conducted to obtain dolomite dissolution rates at different pH values from 1.5 to 3.5. Assuming stoichiometric dolomite dissolution, the Ca/Mg ratio measured from the steady-state output Ca and Mg concentrations was used to obtain the structural formula of dolomite (Ca1.048Mg0.952(CO3)2). Scanning electron microscopy (SEM) was performed on C-coated samples before and after the column experiments using a Hitachi H-4100 instrument under a 15-20 kV potential. The surface of the unreacted limestone grains was rough, whereas the surface of unreacted dolostone grains was flat and terraced. Microparticles attached onto the surfaces were not observed (Fig. 2.1). Cylindrical rock cores used in Part III: fractured cores were cored side-by-side from the provided samples. Fifteen cores of 9 mm in diameter (d) and 18 mm in length (L) were obtained; six limestone cores and nine sandstone cores. Limestone and sandstone porosities were 5% and 6%, respectively, as reported by CIUDEN (ALM-09-008, 2010). The permeability of the rock cores (k < 10-18 m2) was measured by performing a permeability test using the Icare Lab CSS II apparatus (Luquot et al., 2012). Thereafter, a fracture was artificially created by sawing each core with a circular saw, during which formation of microcracks could happen. Nonetheless, as discussed in Section 4.2.4.1, their existence did not 5 mm 5 mm 25 mm 500 μm 5 mm 1 mm 25 mm VuggyLimestone Dolostone 16 Chapter 2: Materials and methods intervene in the overall fracture dissolution. To guarantee flow exclusively through the fracture, all fractures were laterally sealed using a fiber glass thread and Duralco 4525 epoxy resin (stable mechanical and chemical properties up to 690 bar, 260 ºC and low pH) (Fig. 2.2a). Fig. 2.2 Fracture core dimensions (a) and SEM images (b) of the rock samples used in the percolation experiments with fractured cores shown in Part III: fractured cores. Cal = calcite; Qz = quartz and Mc = microcline. Some fractured cores were characterized by XCMT before and after the experiments. Data was acquired at the National Institute for Lasers, Plasma and Radiation Physics (NILPRP) (Bucharest-Magurele, Romania) using the Cone beam CT rapid scan (180° + ½ fan angle), Oblique View Cone Beam. X-ray energy was 225 kVp and maximum power was 10/15 W, using a tungsten filament source. The cores were mounted on a rotary stage, and images were collected every 0.5º. The linear detector, using 1,024 scintillator-photo diode assemblies, yielded 16-bit output digital files. The resulting pixel size was 14 μm. The processing of these X-ray microtomography data was carried out by Voxaya (Montpellier, France), providing characterization of the fracture geometry. Segmentation of the images was performed using the method “edge based snakes” (Yushkevich et al., 2006) and error calculation was carried out by changing the iteration number which corresponds to the propagated time of the segmented volume. Calculated errors of fracture volumes ranged from 0.4% to 5.6%. Oolitic limestone Sandstone (a) (b) Cal Qz Mc 400 mm fracture Cal fracture 200 mm 17 After the reacted samples had been scanned, the cores were flooded with epoxy to allow sectioning and further analysis with SEM. The reacted cores were sectioned along different planes perpendicular to the fracture depending on fracture evolution (sections 1 and 2 in Fig. 2.2a). SEM analyses were performed using a Hitachi H-4100 instrument under a 15-20 kV potential to obtain the dimensions of the fracture and observe features of mineral dissolution and precipitation (Fig. 2.2b) MicroRaman spectra, using a Jobin-Yvon LabRam HR 800 apparatus equipped with an Olympus BXFM microscope and using a wavelength of 532 nm, were obtained to identify the secondary phases in the reacted fractured cores. 2.1.2 Injected solutions Two types of solutions were prepared. The first type was a synthetic version of the Hontomín groundwater, which is nearly in equilibrium with calcite, dolomite and gypsum/anhydrite, with a 0.6 M ionic strength and neutral pH (Table 1.1). Four modifications of this solution (named gypsum-equilibrated solution, gp-e) were performed varying gypsum saturation state and acidity, and yielding one solution undersaturated with respect to gypsum (gypsum-undersaturated solution, gp-u), one solution sulfate-free (no-s) and two acidic gypsum-equilibrated solutions (pH 2.1 acid solution, a2.1; pH 3.5 acid solution, a3.5) (Table 2.2). TDS of these solutions was around 30 g/L. Solutions were prepared by adding appropriate amounts of reagents CaCl2·2H2O, MgCl2·6H2O, NaCl, KCl, Na2SO4 and NaBr to Millipore MQ water (18 MΩ·cm) at room T (25 ± 2 ºC) and under atmospheric pressure. The amounts of reagents were based on equilibrium calculations using the CrunchFlow code (Steefel, 2009) and the EQ3/6 database (Wolery et al., 1990). The solubility product of gypsum was that included in the database (logKGp = -4.4729 at T = 25 ºC). However, in the gypsum-equilibrated solutions unexpected precipitation of gypsum (≈ 1.3 g in 2 L solution) occurred while stirring the solutions for 12 h. Thus, final mixtures were filtered using a 0.22 μm filter to eliminate the precipitated gypsum particles and analyzed to measure the total concentrations by inductively coupled plasma-atomic emission spectrometry (ICP-AES) using a Thermo-Jarrel Ash spectrometer equipped with a CID detector (Table 2.2). The detection limits for Ca, S, Mg, K, Na and Fe were 1.25 10-6 M, 1.34 10-6 M, 2.06 10-6 M, 1.28 10-6 M, 1.30 10-4 M and 3.58 10-7 M, respectively. Using the measured equilibrium concentrations, a new logKGp value was calculated to be logKGp-25 = -4.5978 at T 18 Chapter 2: Materials and methods = 25 ºC. Hence, applying the same correction factor at different temperatures, the newly obtained logKGp-40 and logKGp-60 were -4.6368 (40 °C) and -4.7383 (60 °C). These values turned out to be similar to those recently reported by Nordstrom (2013). The new gypsum logK values at 25, 40 and 60 °C were used in the calculations of this study. For the atmospheric pCO2 experiments the solutions were acidified to pH 2.1 and pH 3.5 (a2.1 and a3.5 input solutions) by adding appropriate amounts of 1M HCl solution. For the rest of experiments run under pCO2 higher than the atmospheric one, the acidity of the solution was obtained from dissolution of CO2. Relationship between CO2 partial pressure and aqueous CO2 concentration, as well as the saturation state of the injected solutions under the experimental conditions were calculated using the PhreeqC (v.3) code (Parkhurst and Appelo, 2013) and the PhreeqC database (Table 2.2). The aqueous solubility of CO2 is temperature, pressure and ionic strength dependent, generally lower at elevated temperature and salinity and higher at elevated pressure (Duan and Sun, 2003; Takenouchi and Kennedy, 1964). pH and saturation indexes (SI) of input solutions injected in experiments performed under P = 150 bar were calculated using PhreeqC-v.3 (P effect included) and CrunchFlow (P effect not included) to quantify the non-linear effect of P (P > 20 bar) on CO2 solubility and mineral equilibria (through the molar volume of solutes) reported by Appelo et al. (2014) (Table 2.2). The second type of prepared solution was a sulfuric acid solution (H2SO4, pH = 2.5; TDS = 1.2 g/L) in equilibrium with gypsum at T = 25 °C (Table 2.2; H2SO4 solution, s). 3.162 mL of 1 M H2SO4 solution were poured into 2 L of Millipore MQ water. Then, approximately 20 g of fragmented gypsum were added and the mixture was stirred for 12 h. Finally, gypsum-equilibrated solutions were filtered using a 0.22 mm filter. Five input solutions were used to perform the column experiments presented in Part II: crushed rock and three input solutions were used to run percolation experiment through fractured cores shown in Part III: fractured cores (Table 2.2). Table 2.2 Injected solutions: amount of reagents, experimental conditions, average concentration (from ICP-AES), experimental pH, and calculated saturation indexes (SI), pH and ionic strength (I). Input label s no-s CaCl2·2H2O- 12.67 MgCl2·6H2O- 13.33 NaCl - 45.05 KCl - 1.69 Na2SO4- - NaBr - 2.34 Exp. Label L25-atm-s L25-atm-a3.5 L60-atm-a3.5 L25-atm-a2.1 L60-atm-a2.1 L25-10-gp-u L40-10-gp-u L25-10-gp-e L40-10-gp-e L60-10-gp-e L60-34-gp-e Part II: Crushed rock D25-atm-s D40-10-gp-e Exp. Label L0.2-gp-e, L1-gp-e, L5-gp-e, L60-gp-e, L1-no-s, L60-no-s, Part III: Fractured cores S5-gp-u, S60-gp-u S0.2-gp-e, S1-gp-e, S5-gp-e, S60-gp-e S1-no-s, S5-no-s, S60-no-s T (°C) 25 25 60 25 60 25 40 60 25 40 60 60 P (bar) 150 150 150 150 pCO2 (bar) 62 34 62 62 Ca2+ 1.55E-02 4.15E-02 SO42- 1.85E-02 - Mg2+ 1.73E-04 3.34E-02 K+- 1.16E-02 Na+- 3.99E-01 Cl-- 5.48E-01 Br-- 1.14E-02 CO21.29E-05 1.30E-05 6.25E-06 1.29E-05 6.23E-06 2.79E-01 2.02E-01 6,15E-01 2.79E-01 2.02E-01 1.43E-01 3.85E-01 6.15E-01 6.15E-01 pH 2.50 3.56 3.65 - 3.51 3.62 3.73 - - - Cal -10.34 -8.01 -7.77 -10.84 -10.60 -3.45 -3.22 -3.03 (-3.35) -3.48 -3.21 -2.88 -2.96 (-3.25) -3.01 (-3.32) -3.10 (-3.46) Dol -23.19 -16.55 -15.74 -22.18 -21.36 -7.34 -6.71 -6.16 (-6.48) -7.47 -6.77 -5.96 -6.07 (-6.35) - - Gp-25 0.09 0.05 - -0.03 - -0.19 - - 0.003 - - - - - Gp-40 - - - - - - -0.20 - - 0.006 - - - - Gp-60 - - 0.04 - -0.09 - - -0.21 (-0.24) - - -0.03 -0.04 (-0.07) -0.01 (-0.05) - pH 2.50 3.65 3.70 3.40 (3.26) 3.61 3.68 3.78 3.53 (3.40) 3.40 (3.26) 3.37 (3.21) I0.05 0.63 0.62 0.61 0.59 0.60 0.61 0.60 0.61 0.62 0.58 0.6 0.62 0.64 (1) refers to the average concentrations of several prepared solutions of the same type. The largest %RSD (relative standard deviation) of the average concentration in s, a2.1-3.5 , gp-u , gp-e and no-s solutions is 2 (for SO42-), 0.8 (for Ca2+), 3.5 (for Mg2+), 4.9 (for SO42-) and 1.2 (for Mg2+), respectively. (2) SI, I and pH values are calculated using CrunchFlow and SI and pH values in brackets are calculated using PhreeqC-v.3. Relationship between pCO2 and aqueous CO2 concentration was calculated using PhreeqC-v.3. TDS (Total Dissolved Solids) of all solutions is around 30 g/L except for solution s which is 1.2 g/L. Cal = calcite; Dol = dolomite; Gp = gypsum; s = H2SO4 solution; a3.5 = acid solution pH 3.5; a2.1 = acid solution pH 2.1; gp-u = gypsum-undersaturated solution; gp-e = gypsum-equilibrated solution; no-s = sulfate-free solution. a3.5 and a2.1 input solutions were acidified to pH 3.5 and 2.1 by adding appropriate amounts of 1M HCl solution. a3.5 a2.1 gp-u gp-e reactant mass (g) in 2L MQ water 16.12 12.67 16.12 13.34 13.34 13.34 37.84 40.57 37.84 1.69 1.69 1.69 8.77 5.45 8.77 2.34 2.34 2.34 10-3.5 10 10 (1)average concentration (mol/kgw) and experimental pH 5.30E-02 4.83E-02 4.40E-02 experimental conditions 60 1 10 10 4.94E-02 3.28E-02 3.22E-02 3.33E-02 3.26E-02 2.90E-02 2.74E-02 1.92E-02 2.68E-02 1.19E-02 1.12E-02 1.15E-02 1.14E-02 3.77E-01 3.99E-01 3.92E-01 4.02E-01 5.18E-01 5.14E-01 5.15E-01 5.02E-01 (2)SI, pH and I (CrunhFlow and PhreeqC-v.3) 1.14E-02 3.50 2.10 1.13E-02 1.13E-02 1.14E-02 3.50 2.10 20 Chapter 2: Materials and methods 2.1.3 Experimental setups and conditions Three experimental setups were used to work under three different P conditions (P = 1, 10 and 150 bar) in the column experiments presented in Part II: crushed rock and two experimental setups were used to work under P = 150 bar in the percolation experiments with fractured cores shown in Part III: fractured cores (Fig. 2.3). Experiments run at P = 1 bar and P = 10 bar were carried out at the IDAEA-CSIC laboratory (Barcelona, Spain) and the experiments conducted at P = 150 bar were performed at the Geosciences Department of Montpellier University-CNRS (Montpellier, France). Fig. 2.3 Experimental setups used to work under atmospheric (a), subcritical (b) and supercritical (c) CO2 conditions. 21 2.1.3.1 Atmospheric pressure setup (P = 1 bar) Transparent methacrylate cylindrical columns of 2.6 cm in inner diameter (d) and 4 cm in length (L) were filled with approximately 20 g of crushed rock fragments with a grain size of 1-2 mm. A bed of glass beads of 0.7 mm in diameter was placed at the top and bottom of the cylinder to homogenize the inlet and outlet solutions. The thickness of the beds was about 0.7 cm, yielding an effective column volume (Vbulk) of 13.80 cm3 (Fig. 2.3a). A 0.22 mm filter was placed at the top of the column to prevent any particle release into the output solution. Column porosity was calculated using the mass of the rock sample, the density of the rock and the effective column volume. The density of the vuggy limestone (2.72 g/cm3) was obtained from the weight fractions of calcite (90.7 wt.%) and dolomite (9.3 wt.%) and the respective densities (2.71 g/cm3 and 2.86 g/cm3; Downs, 2006). Initial porosities ranged between 45% and 48%. Given the mineral mass, the rock density and the effective column volume, and assuming that the rock fragments were spheres of 1.5 mm in diameter, the geometric surface areas of vuggy limestone and dolostone fragments were approximately 2000 m2m/m3bulk (m: mineral; bulk: column). Input solutions were injected from bottom to top of the column using a peristaltic pump under a constant flow rate of 0.021 ± 0.002 mL/min. The outlet solution was periodically collected. Darcy velocity ranged from 6.28 10-7 to 7.06 10-7 m3/m2/s, yielding residence times (τ) between 4.6 and 5.6 h (Table 3.1). Collected samples were immediately acidified with 1% HNO3 to avoid changes in chemical composition. Most of the experiments were undertaken at room temperature (22-25 °C). For the experiments performed at 60 °C the columns were fully immersed in a thermostatic bath. Input and output pH at the desired temperature was measured with a conventional glass pH electrode (accuracy ± 0.02) that was calibrated using Crison buffer solutions of pH 2.02 and 7.00. Total input and output concentrations were measured by ICP-AES (see Section 2.1.2). 2.1.3.2 Subcritical pressure setup (P = 10 bar) The experimental setup was completely designed and assembled at the Geosciences Department of the IDAEA-CSIC in collaboration with the GASLI company, a manufacturer of pressure gauges and measurement instruments located in Barcelona (Spain). A 316 stainless steel column (3.2 cm in diameter (d) and 6 cm in length (L)) was filled with approximately 60 g of crushed rock fragments with a grain size of 1-2 mm. The mass of 28 Chapter 2: Materials and methods supplied by the user as input. In this set of simulations, the reaction rate laws used in the calculations are expressed as (2.14) where is the reaction rate for a given mineral in units of mol/m3bulk/s, is the mineral surface area (m2m/m3bulk), is the reaction rate constant (mol/m2/s) at the temperature of interest, is the term describing the effect of pH on the rate, is a term describing a catalytic/inhibitory effect by another species on the rate, and is the function describing the dependence of the rate on the solution saturation state. The summation term indicates that several parallel rate laws may be used to describe the dependence of the rate on pH or on other species. The rate constant at temperature T (K) is calculated from (2.15) where is the rate constant of a mineral at 25 °C, is the apparent activation energy of the overall reaction (J/mol), is temperature in Kelvin and is the gas constant (J/mol/K). The function is defined as Δ (2.16) where is the Gibbs energy of the reaction (J/mol), is the ionic activity product of the solution with respect to the mineral, is the equilibrium constant for that mineral reaction (ionic activity product at equilibrium) and and are empirical coefficients. Changes in mineral surface area (m2m/m3bulk) due to reaction are calculated according to dissolution (2.17) precipitation (2.18) 29 where is the initial volume fraction of the mineral and is the initial porosity of the medium. This formulation ensures that as the volume fraction of a mineral goes to 0, its surface area does too. Moreover, for both dissolving and precipitating minerals, the term requires that the surface area of a mineral in contact with fluid goes to 0 when the porosity of the medium goes to 0. This formulation is used primarily for primary minerals (i.e., minerals with initial volume fractions > 0). For secondary minerals which precipitate, the value of the initial bulk surface area specified is used as long as precipitation occurs. If this phase later dissolves, the above formulation is used with an arbitrary initial volume fraction of 0.01. Regardless the changes in permeability in the 2D calculations where flow was updated, the code solved Darcy’s law (neglecting the buoyancy term; Eq. (2.19)), and permeability was updated at each time step according to Eq. (2.20). (2.19) (2.20) 2.2.2 One-dimensional model (Part II: crushed rock) Results and discussion of the 1D reactive transport simulations that reproduce the experimental data of the column experiments are shown in Chapter 3. Model parameters used in 1D simulations are given in the following sections. 2.2.2.1 Numerical discretization The one-dimensional numerical domain is composed of 20 elements in most simulations. Two simulations under atmospheric conditions with a2.1 input solution (Table 3.1; experiments L25-atm-a2.1 and L60-atm-a2.1) were run using a two-zone domain composed of 40 shorter elements at the start of the column and 16 longer elements along the rest for better resolution of changes in solution composition. See Tables A.1 and A.2 (Appendix A) for details of the spatial discretization. 30 Chapter 2: Materials and methods 2.2.2.2 Rock and solution composition Two rocks were considered in the calculations: vuggy limestone and dolostone (Table 2.1). Mineral volume fractions are shown in Tables A.1 and A.2 (Appendix A). For each experiment, only the precipitated sulfate phase was taken into account (gypsum at all temperatures). Rate laws for the reacting minerals were taken from the literature (Palandri and Kharaka, 2004). The fit of the model to the experimental data (aqueous Ca, S, Mg concentrations and pH) was performed by adjusting the values of the reactive surface areas (Am in Eq. 2.14). A sensitivity study of the reactive surface areas was performed. As a result, ranges of values that could also fit the experimental concentrations within the ± 2% analytical error and experimental pH are provided. In the experiments where gypsum precipitated, the initial surface area of gypsum (area used to fit the early evolution of the system) was increased in separate calculations to fit S and Ca concentrations along the experiment (see initial and final values of the reactive area of gypsum in Tables 2.3 and 2.4). Calculated (CrunchFlow) saturation index, ionic strength and pH of the injected solutions (input boundary condition) are given in Tables 2.3 and 2.4. In the case of the 34 bar pCO2 experiment, the PhreeqC (v.3) calculated undersaturations with respect to calcite and gypsum are slightly larger (Table 2.4; SICal = -3.25 and SIGp = -0.07) than the CrunchFlow values owing to the P effect on equilibrium constants. In all simulations, the initial solution composition inside the column was considered to be in equilibrium with calcite, dolomite and gypsum. 2.2.2.3 Flow and transport properties The Darcy velocity q (m3/m2/s) in the column experiments was obtained dividing the flow rate Q (m3/s) by the cross section S (m2) calculated as S = π·r2, being r the radius of the column. Darcy velocity q, effective diffusion coefficient De and longitudinal dispersivity αL used in the simulations are shown in Tables A.1 and A.2 (Appendix A). The effective diffusion coefficient was derived from (2.21) The diffusion coefficient in water at 25 ºC D0 and the cementation exponent n used in the simulations were 10-9 m2/s and 2-2.5, respectively. They are based on common values 31 reported in the literature (e.g., Ullman and Aller, 1982; De Marsily, 1986; Domenico and Schwarthz, 1990). 2.2.2.4 Thermodynamic and kinetic data Thirty-seven aqueous species were considered in most simulations (experiments with a2.1, a3.5, gp-u and gp-e as input solution). Two experiments performed under atmospheric conditions used H2SO4 input solution in equilibrium with gypsum (s). In these cases, seventeen aqueous species were taken into account (Table A.3, Appendix A, species with *). All the equilibrium constants (log K at 25, 40 and 60 °C) and stoichiometric coefficients, which were taken from the EQ3/6 database (Wolery et al., 1990; included in the CrunchFlow code), are shown in Table A.3 (Appendix A). Activity coefficients were calculated using the extended Debye-Hückel formulation (b-dot model) with parameters obtained from the CrunchFlow database (EQ3/6). Three solid phases were considered in the calculations (calcite, disordered dolomite and gypsum). The equilibrium constants for the mineral reactions were also taken from the EQ3/6 database (Table A.4, Appendix A). At 25, 40 and 60 °C, the values of the gypsum equilibrium constants were decreased by approximately 25% to fit the experimentally observed equilibrium condition, as explained in Section 2.1.2. Table 2.3 Reactive surface area (Am) and input boundary conditions (SI, I and pH) used in simulations under atmospheric conditions (CrunchFlow code). initial final initial final initial final 0.001(0-0.01) 0.015(0.01-0.03) 0.01(0-0.1) 0.1(0.02-0.2) - - - - - - - - - - - - 0.01(0.015-0.03) 0.04(0.02-0.05) Am = Reaction surface area. Am in brackets indicates the range of values that fits the experimental concentration data (within ± 2% uncertainty) in the sensibility study. Cal = calcite; Dol-dis = disordered dolomite; Gp = gypsum. L60-atm-a2.1 ROCK COMPOSITION Am (m2m/m3bulk ) Cal - 120(80 - 180) Exp. label D25-atm-s L25-atm-s L25-atm-a3.5 L60-atm-a3.5 L25-atm-a2.1 Dol -dis 17(16 - 20) 5(4 - 7) 5(0 - 20) Gp-25 0.03(0.01-0.06) 0(0-0.03) - SI, I and pH (CrunchFlow) Gp-40 - - - Gp-60 - - 0(0-0.005) INPUT BOUNDARY CONDITIONS Input label s a3.5 a2.1 -10.60 Dol-dis -22.85 -23.53 -16.55 -15.74 -22.18 -21.36 Cal -10.33 -10.34 -8.01 -7.77 -10.84 - - Gp-40 - - - - - - Gp-25 0.09 0.08 0.05 - -0.03 pH 2.50 3.50 2.10 -0.09 I 0.05 0.63 0.62 0.61 0.59 Gp-60 - - - 0.04 32 Chapter 2: Materials and methods 2.2.2.5 Reaction rates Kinetic rate laws used for primary minerals (calcite and dolomite) and secondary minerals (gypsum) are from Palandri and Kharaka (2004) and Xu et al. (2012). Xu et al. (2012) proposed a calcite dissolution rate that improves the rate-G dependence under close- to-equilibrium conditions (-12 ≤ G ≤ 1.7 kJ/mol) with respect to the simplest TST-based rate law (m1 = m2 = 1 in Eq. (2.16)). Hence, all simulations were run based on this rate law (i.e., m1 = 3 and m2 = 1 in Eq. (2.16)). Rate parameters and apparent activation energies are listed in Table A.5 (Appendix A). The two parallel rate laws for each primary mineral describe the explicit dependence of the rates on pH. Rate constants at temperatures different from 25 ºC were calculated according to Eq. (2.15). Table 2.4 Reactive surface area (Am) and input boundary conditions (SI, I and pH) used in simulations under subcritical and supercritical conditions (CrunchFlow and PhreeqC (v.3) codes). 2.2.3 Two-dimensional model (Part III: fractured cores) Two-dimensional reactive transport simulations were performed to reproduce the experimental data of three percolation experiments with fractured cores which developed initial final initial final initial final 0.01(0-0.1) 10(1-*) - - - - - - 0.01(0-0.04) 0.3(0.1-*) - - - - - - 0.01(0-0.1) 0.1(0.01-*) Am = Reaction surface area. Am in brackets indicates the range of values that fits the experimental concentration data (within ± 2% uncertainty) in the sensibility study. * Maximum value is not constrained within the experimental error; solution reached equilibrium in the column. ** CrunchFlow (charge balance) calculated pH, similar to the measured averaged pH (see Table 3.1). Cal = calcite; Dol-dis = disordered dolomite; Gp = gypsum. SI and pH in brackets is calculated using PhreeqC (v.3). D40-10-gp-e L60-34-gp-e ROCK COMPOSITION Am (m2m/m3bulk ) Cal 120(80 - 180) - 120(80 - 180) Exp. label L25-10-gp-u L40-10-gp-u L25-10-gp-e L40-10-gp-e L60-10-gp-e Dol -dis 5(0 - 10) 5(3 - 20) 10(5 - 30) Gp-25 - - - - Gp-60 - - - 0.5(0.2 -*) Gp-40 - - 0.3(0.08-*) - INPUT BOUNDARY CONDITIONS Input label gp-u gp-e SI, I and pH (CrunchFlow) -2.88 -3.19 -2.96 (-3.25) Dol-dis -7.34 -6.71 -7.47 -6.77 -5.96 -6.74 Cal -3.45 -3.22 -3.48 -3.21 -6.07 Gp-25 -0.19 - 0.00 - - - - 0.01 - -0.03 - -0.04 (-0.07) Gp-40 - -0.20 - -0.01 - Gp-60 - - - - 0.63 0.60 pH 3.65** 3.7** 3.61** 3.68** 3.78** 3.68** 3.53 (3.40) I 0.60 0.61 0.61 0.62 0.58 33 three different dissolution patterns: face dissolution (L0.2-gp-e), wormhole (L1-gp-e) and uniform dissolution (S60-no-s). Results and discussion of the simulations are shown in Chapter 5. Model parameters used in 2D simulations are given in the following sections. 2.2.3.1 Numerical discretization Rectangular coordinates were used to model experiments which developed face and uniform dissolution patterns (L0.2-gp-e and S60-no-s), whereas rectangular and cylindrical coordinates were used in the experiment where a wormhole developed (L1-gp-e). Reasons to apply this modeling approach are given in Chapter 5. To simplify the models with rectangular coordinates only half of the core was considered given the symmetry of the fractured cores (Fig. 2.4a).  Rectangular coordinates: The dimensions of the domain were: Ry: 20 mm (core length). Rx: 3.5 mm. where Rx was computed by considering that half the core section S (S = π(d/2)2; d = 9 mm) was equivalent to the area of a rectangle whose sides were d and Rx (Fig. 2.4a). The domain was composed of two parts: (1) high-permeability zone (fracture, large porosity; in red in Fig. 2.4a) and (2) rock matrix (small porosity; in green in Fig. 2.4a). The fracture zone was at the left side of the rectangular domain, parallel to the flow direction (y axis) and had a thickness equal to half the experimental fracture aperture (i.e., the first node along the x direction). The domain in experiment L0.2-gp-e was composed of 27 elements in the x direction and 36 elements in the y direction. In experiment S60-no-s the domain was composed of 28 elements in the x direction and 20 elements in the y direction. This coarser discretization was applied to reduce the computational time constraint by the Courant number. The detailed spatial discretization is given in Table A.6 (Appendix A).  Cylindrical coordinates: The dimensions of the domain were: Cy: 20 mm (core length) 34 Chapter 2: Materials and methods Cx: 4.5 mm (radius of the core) In this case, the 2D grid had symmetry around the y axis (the core sample axis). The length of the interface (z) between two cells along x was proportional to the radial distance to the core sample axis. In rectangular coordinates, this length was independent of radius. Fig. 2.4 Scheme showing the geometry and boundary conditions of the flow domain used in the models: (a) rectangular and (b) cylindrical coordinates. Left and right boundaries are no-flow boundaries. Plots on the left show the conceptual model, and plots on the right show the implemented grid. The domain was composed of two zones: (1) cylindrical fracture (high porosity) and (2) rock matrix (small porosity). The fracture zone was at the left side of the rectangular domain, 35 parallel to the flow direction (y axis) and 402 mm thick (i.e., 52 nodes along the x direction). The thickness of the fracture was varied from 6.7 mm (initial fracture aperture) to 402 mm in a sensitivity study (see Chapter 5, Section 5.3.2). When cylindrical coordinates were used, it was assumed that the initial fracture was already cylindrical (i.e., a very small wormhole). The domain in experiment L1-gp-e was composed of 133 elements in the x direction and 36 elements in the y direction. Elements along the x axis were rather short to avoid errors in the flux calculations. The detailed spatial discretization is given in Table A.6 (Appendix A). 2.2.3.2 Rock and solution composition The initial mineralogical composition and porosity of the two described zones are given in Table 2.5. Two rocks were considered in the calculations: oolitic limestone and sandstone (Table 2.1). Fracture porosity was defined to be 100% whereas limestone and sandstone porosities were those reported by CIUDEN (ALM-09-008, 2010). The reactive surface areas (Am in Eq. 2.14) of calcite, quartz and microcline were initially calculated considering the geometric area (see Section 2.1.3.3). The calcite reactive surface area was fitted to reproduce the variation in Ca concentration. Given that quartz and microcline did not intervene in the overall process, the values of their reactive surface areas were not modified. Calculated values of saturation index, ionic strength and pH of the injected solutions are given in Table 2.2. In all simulations, the initial solution composition in the rock matrix (pore solution) was considered to be in equilibrium with calcite and gypsum. 2.2.3.3 Flow and transport parameters Flow field was updated according to porosity and permeability changes in the simulations of the experiments that developed face and uniform dissolution (rectangular coordinates). Flow update is not implemented in CrunchFlow when cylindrical coordinates area used, and therefore fixed flow was assumed in this case. Notice that De was a fitted parameter in the calculations. The molecular diffusion coefficient D0, cementation exponent n (e.g., Ullman and Aller, 1982; De Marsily, 1986; Domenico and Schwarthz, 1990), longitudinal dispersivity αL and effective diffusion coefficient De derived from Eq. (2.21), are listed in Table 2.6. 36 Chapter 2: Materials and methods Table 2.5 Initial mineralogical composition of both the rock matrix and the high-permeability zone (fracture) and input solution used in the 2D simulations. In the simulation of the wormhole experiment (i.e., rectangular + cylindrical coordinates and fixed flow), q was fixed in the fracture zone of the numerical domain. In simulations of the experiments with face and uniform dissolution patterns (i.e., rectangular coordinates and flow update), Q (m3/s) was imposed at the first node of the high-permeability zone (node 1,1) to obtain the corresponding q (m3/m2/s). In the experiments where flow field was updated, rock and fracture permeability varied according to Eq. (2.20). Initial fracture permeability in experiment S60-no-s was obtained by hydraulic measurement using Eqs. (2.10 and 2.11). However, in experiment L0.2-gp-e, hydraulic measurement could not be performed and hence initial fracture permeability was calculated using Eq. (2.11) and the fracture aperture measured from SEM image (as). In both cases, initial limestone and sandstone permeabilities (rock matrix) were assumed to be 10-20 m2 based on the measured rock permeabilities (see Section 2.1.1). In these two simulations, left and right boundaries were no-flow boundaries. Initial pressure was set to 150 bar in the entire domain. Exp. label Rock matrix Rock Initial porosity φ0% Vol. F. AmVol. F. AmVol. F. Am Cal 0.95 2000 0.95 9500 0.612 300000 Qz - - - - 0.264 3605 Mic - - - - 0.060 2180 Gp-60 0.00 100 0.00 6 - - High-permeability (fracture) zone Initial porosity φ0% Vol. F. AmVol. F. AmVol. F. Am Gp-60 0.00 100 0.00 6 - - Input solution Input label Am = Reactive surface area in m2m/m3bulk. Vol. F. = Volume fraction Chemical composition, calculated pH, SI and I of the input solutions are in Table 2.2. gp-e gp-e no-s 6 5 100 100 100 Sandstone L0.2-gp-e L1-gp-e S60-no-s Oolitic limestone 37 Table 2.6 Initial transport properties assumed in the 2D calculations. 2.2.3.4 Thermodynamic and kinetic data Thirty-seven aqueous species were considered in limestone simulations (experiments L0.2-gp-e and L1-gp-e; Table A.3, Appendix A) and forty-four aqueous species in the sandstone simulation (experiment S60-no-s; Table A.7, Appendix A). Equilibrium constants (log K at 60 °C) and stoichiometric coefficients, which were taken from the EQ3/6 database (Wolery et al., 1990; included in the CrunchFlow code), are shown in Tables A.3 and A.7 (Appendix A). Activity coefficients were calculated using the extended Debye-Hückel formulation (b-dot model) with parameters obtained from the CrunchFlow database (EQ3/6). Two solid phases were considered in the limestone calculations (calcite and gypsum) and three solid phases in the sandstone calculations (calcite, microcline and quartz). The equilibrium constants for the mineral reactions were also taken from the EQ3/6 database (Table A.8, Appendix A). The values of the gypsum equilibrium constants at 60 °C were decreased by approximately 25% to fit the experimentally observed equilibrium condition, as explained in Section 2.1.2. 2.2.3.5 Reaction rates Kinetic rate laws used for calcite, quartz and gypsum are from Palandri and Kharaka (2004) and Xu et al. (2012) and those for microcline are from Bandstra et al. (2008). Rate parameters and apparent activation energies are listed in Table A.9 (Appendix A). The two parallel rate laws for calcite and microcline describe the explicit dependence of the rates on pH. Exp. label L0.2-gp-e L1-gp-e S60-no-s Coordinates rectangular rectangular + cylindrical rectangular Fracture permeability km26.75E-12(1) - 4.32E-12(2) Rock permeability km21.00E-20 - 1.00E-20 Initial Darcy velocity (N) qm3/m2/s 6.86E-04 4.61E-3 (1) + 5.47E-4 (52) 2.56E-01 Diffusion coeff. D0m2/s 1.1E-10 5.36E-09 1.1E-08 Cementation exponent n2.5 2.5 2.5 Eff. diffusion coeff. Dem2/s 6.0E-14 3.0E-12 9.5E-12 Long. dispersivity αLm 0.01 0.01 0.01 Trans. dispersivity αTm - - - N = number of nodes along the x direction with fixed flow (1)Fracture permeability calculated using Eq. (2.11) and as. (2)Fracture permeability calculated using Eq. (2.11) and ah. 44 Chapter 3: Column experiments and 1D modeling Fig. 3.1 SEM images of the reacted limestone (a) and dolostone (b). Gypsum needles precipitated on the carbonate surfaces. As stated above, simulations successfully reproduced the experimental variation of Ca, Mg and S concentration over time. However, the simulated output pH matched the measured output pH of the 10 bar pCO2 experiments (Fig. B.3b, Appendix B) but was systematically smaller than the measured one in experiments under atmospheric pCO2 (Fig. B.1a, Appendix B). This mismatch was probably caused by CO2 degassing when measuring the output pH under atmospheric conditions. pH was measured in line under 10 bar pCO2 (no degassing). 3.2.1 Experiments under atmospheric conditions (P = 1 bar; pCO2 = 10-3.5 bar) 3.2.1.1 H2SO4 solution (s) Fig. 3.2a, b shows the output Ca and S concentrations measured during the experiments with s input solution (Table 3.1; experiments L25-atm-s and D25-atm-s). The measured output pH rapidly increased from 2.50 to 7.55 and reached steady state from the start of the experiment in both limestone and dolostone column experiments (Fig. B.1a, Appendix B). The output Ca concentration was larger than the input one in both experiments (Fig. 3.2a). The high output pH and Ca and Mg concentrations indicate that calcite and dolomite dissolved. 45 Fig. 3.2 Top row (experiments L25-atm-s and D25-atm-s): Variation of the experimental (Exp) and simulated (Sim) output concentration of Ca (a) and S (b) with time in limestone (L; in green) and dolostone (D; in blue) column experiments. Black-solid lines indicate input solution. Dashed and dotted lines depict simulated values of limestone and dolostone column experiments, respectively. Bottom row (experiments L25-atm-a2.1 and D25-atm-a2.1): Variation of the experimental (Exp) and simulated (Sim) increase in Ca (c) and S (d) concentration with time in limestone column experiments at 25 ºC (in green) and 60 ºC (in red). Solid lines indicate the Ca concentration increase at equilibrium with calcite. During limestone dissolution, output pH and output Ca concentration reached equilibrium with calcite (equilibrium Ca concentration is 1.98 10-2 M). However, during 1.70e-2 1.75e-2 1.80e-2 1.85e-2 1.90e-2 1.95e-2 0 500 1000 1500 2000 2500 3000 S (mol/kgw) time (h) 1.5e-2 1.6e-2 1.7e-2 1.8e-2 1.9e-2 2.0e-2 2.1e-2 0 500 1000 1500 2000 2500 3000 Exp L (out) Sim L (out) Exp D (out) Sim D (out) L and D (inp) Ca (mol/kgw) time (h) First set L25-atm-s and D25-atm-s 1.5e-2 1.6e-2 1.7e-2 1.8e-2 1.9e-2 2.0e-2 2.1e-2 0 500 1000 1500 2000 2500 3000 Exp L (out) Sim L (out) Exp D (out) Sim D (out) L and D (inp) Ca (mol/kgw) time (h) 8.0e-3 1.2e-2 1.6e-2 2.0e-2 2.4e-2 2.8e-2 0100 200 300 400 500 600 700 800 Exp 25 (out) Sim 25 (out) Exp 60 (out) Sim 60 (out) 25 (Cal EQ) 60 (Cal EQ) Ca (mol/kgw) time (h) -5e-3 -4e-3 -3e-3 -2e-3 -1e-3 0 0100 200 300 400 500 600 700 800 S (mol/kgw) time (h) First set L25-atm-a2.1 and L60-atm-a2.1 8.0e-3 1.2e-2 1.6e-2 2.0e-2 2.4e-2 2.8e-2 0100 200 300 400 500 600 700 800 Exp 25 (out) Sim 25 (out) Exp 60 (out) Sim 60 (out) 25 (Cal EQ) 60 (Cal EQ) 25-60 (inp) Ca (mol/kgw) time (h) 8.0e-3 1.2e-2 1.6e-2 2.0e-2 2.4e-2 2.8e-2 0100 200 300 400 500 600 700 800 Exp 25 (out) Sim 25 (out) Exp 60 (out) Sim 60 (out) 25 (Cal EQ) 60 (Cal EQ) 25-60 (inp) Ca (mol/kgw) time (h) (a) (b) (c) (d) 46 Chapter 3: Column experiments and 1D modeling dolostone dissolution output Ca and Mg concentration did not achieve equilibrium with dolomite (equilibrium Ca and Mg concentrations are 1.82 10-2 M and 2.88 10-3 M, respectively). Equilibrium concentrations were calculated with PhreeqC (v.3; Parkhurst and Appelo, 2013) and the PhreeqC database. In the limestone column experiment, the output S concentration was only slightly smaller than the input one up to around 1500 h, after which it decreased and reached steady state. By contrast, in the dolomite column experiment, the output S concentration was smaller than the input concentration and steady state was reached from the start of the experiment (Fig. 3.2b). This decrease indicates that a sulfate-rich mineral precipitated after around 1500 h in the limestone column experiment and almost from the start in the dolomite column experiment. The precipitated sulfate-rich mineral was identified by XRD to be gypsum in both cases. In the column experiments performed by Singurindy and Berkowitz (2003), clogging of the column due to gypsum precipitation was favored under flow rates between 1 and 2.2 mL/min and H+/SO42- ratios between 2 and 2.5. The range of H+/SO42- ratios and flow rates used in their study were significantly higher than those used in our experiments. 3.2.1.2 Acidic gypsum-equilibrated solution (a2.1 and a3.5) Two limestone column experiments with a2.1 input solution were performed at 25 and 60 ºC (Table 3.1; experiments L25-atm-a2.1 and L60-atm-a2.1). Fig. 3.2c and d shows the experimental and simulated Ca and S concentration increase versus time. In these experiments, the measured output Ca concentration and pH were higher than the respective input values. The output Ca concentration was slightly larger at 25 than at 60 ºC. The solution saturation state was close to calcite equilibrium at both temperatures, but closer at 60 ºC (Fig. 3.2c; PhreeqC (v.3)-calculated equilibrium Ca concentrations at 25 and 60 ºC were 7.28 10- 2 M and 6.45 10-2 M, respectively). At both temperatures, S concentration was immediately smaller than the input one, indicating immediate precipitation of sulfate-rich mineral. According to the S deficit, mineral precipitation at 60 ºC was higher than at 25 ºC (Fig. 3.2d). The precipitated sulfate-rich mineral at 25 and 60 ºC was identified by XRD to be gypsum even if thermodynamically the most stable phase at 60 ºC was considered to be anhydrite. The fact that gypsum was the precipitated phase at 60 ºC was not rare owing to the existence of a marked inconsistency between the thermodynamics of calcium sulfate and its crystallization behavior (Van Driessche et al., 2012). Ossorio et al. (2013) reported that 47 salinity and temperature strongly influence the type and stability of the precipitated phase (gypsum, bassanite and anhydrite), yielding gypsum stability up to 10 months at 80 ºC under 0.8 M NaCl conditions. Increasing salinity (4.3 M NaCl), bassanite precipitation prevails and gypsum stability decreases. In the limestone column experiments run with a3.5 input solution (Table 3.1; experiments L25-atm-a3.5 and L60-atm-a3.5), the output Ca concentration was only slightly larger than the input one (Fig. B.2a, Appendix B), and the input and output Mg and S concentrations were the same within error. Therefore, at this input pH, limestone dissolution was small and there was no precipitation. The output pH increased up to 6.75 (25 ºC) and 6.82 (60 ºC) (Fig. B.2b, Appendix B). 3.2.2 Experiments under subcritical conditions (P = pCO2 = 10 bar) 3.2.2.1 Gypsum-undersaturated solution (gp-u) In the limestone column experiments with gp-u input solution (Table 3.1; experiments L25-10-gp-u and L40-10-gp-u), the output Ca concentration was larger than the input concentration at both temperatures (25 and 40 ºC), reached steady state at the start of the experiment, and was greater at 25 ºC than at 40 ºC (Fig. B.3a, Appendix B). The output pH increased up to 5.22 (25 ºC) and 5.17 (40 ºC) (Fig. B.3b, Appendix B). The input and output Mg and S concentrations were the same within error. Therefore, calcite dissolution was the only occurring reaction. In both experiments, the solution saturation state did not reach equilibrium with calcite (Fig. B.3a, Appendix B; PhreeqC (v.3)-calculated equilibrium Ca concentrations at 25 and 40 ºC were 7.04 10-2 M and 6.31 10-2 M, respectively). In the experiment performed at 25 ºC, the measured loss of mass calculated by subtracting the final weight of the sample from the initial weight was 8% higher than the calculated loss of mass based on the aqueous chemistry (comparison of ΔMmeas and ΔMcalc-Ca in Table 3.1). However, these values were within the calculated propagated error (≈ 10%). 3.2.2.2 Gypsum-equilibrated solution (gp-e) Three limestone column experiments with gp-e input solution were performed at 25, 40 and 60 ºC (Table 3.1; experiments L25-10-gp-e, L40-10-gp-e and L60-10-gp-e). 48 Chapter 3: Column experiments and 1D modeling Fig. 3.3 Top row (experiments L25-10-gp-e, L40-10-gp-e and L60-10-gp-e): Variation of the experimental (Exp) and simulated (Sim) increase in Ca (a) and S (b) concentration with time in limestone column experiments at 25 ºC (in green), 40 ºC (in orange) and 60 ºC (in red). Dashed, dotted and red-solid lines show simulated values at 25, 40 and 60 ºC, respectively. Bottom row (experiments L40-10-gp-e and D40-10-gp-e): Variation of the experimental (Exp) and simulated (Sim) increase in Ca (c) and S (d) concentration versus time in limestone (L; in orange) and dolostone (D; in blue) column experiments. Dotted and dashed lines show simulated values of limestone and dolostone experiments, respectively. Solid lines in (c) represent the Ca concentration increase in equilibrium with calcite (in orange) and dolomite (in blue). The results of this set of experiments showed that the output pH immediately increased up to 5.21 (25 ºC), 5.16 (40 ºC) and 5.13 (60 ºC) and reached steady state. The output Ca -4.7e-3 -3.1e-3 -1.6e-3 0.0 1.6e-3 0 152 304 456 608 760 S (mol/kgw) time (h) (d) (c) Second set L40-10-gp-e and D40-10-gp-e Second set L25-10-gp-e, L40-10-gp-e and L60-10-gp-e -5.0e-3 0.0 5.0e-3 1.0e-2 1.5e-2 2.0e-2 2.5e-2 0 180 360 540 720 900 Exp L (out) Exp D (out) Sim L (out) Sim D (out) L (Cal EQ) D (Dol EQ) Ca (mol/kgw) time (h) -6.0e-3 -5.0e-3 -4.0e-3 -3.0e-3 -2.0e-3 -1.0e-3 0.0 1.0e-3 0 180 360 540 720 900 S (mol/kgw) time (h) -5.0e-3 0.0 5.0e-3 1.0e-2 1.5e-2 2.0e-2 2.5e-2 0 180 360 540 720 900 Exp L (out) Exp D (out) Sim L (out) Sim D (out) L (Cal EQ) D (Dol EQ) Ca (mol/kgw) time (h) (a) (b) 0.0 6.2e-3 1.3e-2 1.9e-2 2.5e-2 0 152 304 456 608 760 Ca (mol/kgw) time (h) 4e-2 5e-2 6e-2 7e-2 8e-2 0 110 220 330 440 550 Exp 25 (out) Exp 40 (out) Sim 25 (out) Sim 40 (out) Exp 60 (out) Sim 60 (out) Ca (mol/kgw) time (h) 49 concentration exceeded that of the input in all experiments and decreased with time at 25 and 40 ºC, reaching steady state at 60 ºC. Ca release decreased with temperature (Fig. 3.3a;Ca25ºC >Ca40ºCCa60ºC). In all experiments, the solution saturation state did not reach equilibrium with calcite, being slightly further from equilibrium at 25 ºC (PhreeqC (v.3)-calculated equilibrium Ca concentrations at 25, 40 and 60 ºC were 7.55 10-2 M, 6.93 10-2 M and 6.08 10-2 M, respectively). The input and output Mg concentrations were the same within error. The output S concentration was smaller than the input concentration in all experiments. This S deficit gradually increased with time at 25 and 40 ºC and showed little variation with time at 60 ºC. Output S concentration increased with temperature, which implies that │S60ºC│<│S40ºC│<│S25ºC│ (Fig. 3.3b). Thus, according to the resulting trend of Ca and S, the amounts of dissolved calcite and precipitated gypsum (identified by XRD) were larger at lower temperatures. Using gp-e as input solution, the dolostone column experiment at 40 ºC (Table 3.1; experiment D40-10-gp-e) showed that the output pH increased up to 4.62. ΔCa and ΔS variations with time are shown in Fig. 3.3c and d. The output Ca and Mg concentrations only increased very slightly (zero within error) and immediately reached steady state. The solution saturation state did not reach equilibrium with dolomite. In contrast to the limestone experiment, the output S concentration decreased immediately and reached steady state (Fig. 3.3d). However, the drop in concentration was very small (zero within error). In this set of experiments, the measured loss of mass was different from the calculated loss of mass by a factor of 5-28% (Table 3.1). 3.2.3 Experiment under supercritical conditions (P = 150 bar; pCO2 = 34 bar) The output pH in the limestone column experiment with gp-e input solution at 34 bar of pCO2, P of 150 bar and T of 60 ºC (Table 3.1; experiment L60-34-gp-e) could not be measured. The simulated output pH rapidly increased up to 4.72 (CrunchFlow) and 4.47 (PhreeqC (v.3)) and reached steady state. The output Ca and Mg concentrations were larger than the input concentrations (Fig. 3.4a), and the output S concentration was slightly smaller (Fig. 3.4b). Hence, limestone dissolved and gypsum (identified by XRD) precipitated. The calculated loss of mass (0.07 g) was 40% higher than the measured loss of mass (0.05 g) (Table 3.1), the difference being due to the small mass value. 50 Chapter 3: Column experiments and 1D modeling Fig. 3.4 Variation of the experimental (Exp) and simulated (Sim) output Ca and Mg (a) and S (b) concentration with time in the limestone column experiment under supercritical conditions (L60-34- gp-e). Red and green dashed lines indicate simulated values of output concentrations using CrunchFlow and PhreeqC (v.3), respectively. In the PhreeqC (v.3) calculation, dolomite was not considered and the calculated output S concentration coincides with the input value. 3.3 Discussion The influence of the variation in T, mineralogy and pCO2 on the dissolution and precipitation processes was evaluated using the experimental and modeling results. Overall, it was observed that (1) by lowering temperature, the amount of dissolved limestone increased under any pCO2 condition and gypsum precipitation was only favored at high pCO2; (2) as expected, the amount of mineral dissolved and porosity increase were noticeably larger in the limestone experiments than in the dolostone ones, regardless of solution composition, temperature and dissolved CO2; gypsum precipitated immediately as dolostone dissolved and only after some time as limestone dissolved; and (3) by increasing pCO2 the amounts of dissolved limestone and precipitated gypsum increased, enhancing the porosity over a longer column length. Moreover, under all P-pCO2-T conditions, the volume of precipitated gypsum was always smaller than the volume of dissolved rock (either limestone or dolostone), yielding in all cases a porosity increase. Detailed explanation of the mechanisms that control the occurring processes is given as follows and is illustrated by the plots in Fig. 3.5 and 3.6. (b) (a) Third set L60-34-gp-e 2.5e-2 3.0e-2 3.5e-2 4.0e-2 4.5e-2 5.0e-2 5.5e-2 6.0e-2 6.5e-2 020 40 60 80 100 120 Exp (inp) Exp (out) Sim CrunchFlow Sim PhreeqC Concentration (mol/kgw) time (h) Exp Ca (inp) Mg Exp Mg (inp) Ca 2.2e-2 2.3e-2 2.4e-2 2.5e-2 2.6e-2 2.7e-2 2.8e-2 020 40 60 80 100 120 Concentration (mol/kgw) time (h) S Exp S (inp) 2.5e-2 3.0e-2 3.5e-2 4.0e-2 4.5e-2 5.0e-2 5.5e-2 6.0e-2 6.5e-2 020 40 60 80 100 120 Exp (inp) Exp (out) Sim CrunchFlow Sim PhreeqC Concentration (mol/kgw) time (h) Exp Ca (inp) Mg Exp Mg (inp) Ca 51 Fig. 3.5 Experimental variation of volume of dissolved rock Vrock-diss (a and d), percentage of volume of dissolved limestone (g), volume of precipitated gypsum VGp-ppt (b and e), percentage of volume of precipitated gypsum (h) and porosity (c, f and i) with number of pore volumes Vp in experiments performed at 25 ºC (in green), 40 ºC (in orange) and 60 ºC (in red). Solid, dashed and dotted lines (plots of T and pCO2) represent atmospheric, 10 bar and 34 bar pCO2 conditions, respectively. Solid lines and solid lines with empty squares (plots of mineralogy) indicate experiments with limestone (L) and dolostone (D), respectively. VL-diss (%) and VGp-ppt (%) are percentages of dissolved and precipitated volumes with respect to each initial sample volume. (1) T effect: Comparison between two groups of experiments with gypsumequilibrated solution and temperature ranging from 25 to 60 ºC was used to assess the T effect on porosity changes (Table 3.1; the first group was conducted under atmospheric pCO2 (L25-atm-a2.1 and L60-atm-a2.1) and the second group under 10 bar of pCO2 (L25-10-gp-e, L40-10-gp-e and L60-10-gp-e)). Fig. 3.5a shows that the amount of limestone dissolved 0 600 1200 1800 030 60 90 120 150 Exp L (25 ºC, no EQ brine) Exp L (40 ºC, no EQ brine) Exp L (25 ºC, brine) Exp L (40 ºC, brine) Exp L (60 ºC, brine) VL-diss (mm3) Vp 0 100 200 300 400 500 030 60 90 120 150 Vppt (25)EQ Vppt(40)EQ Vppt(60) EQ Vppt(25)atm Vppt(60)atm VGp-ppt (mm3) Vp 0 0.5 1 1.5 2 2.5 3 3.5 4 030 60 90 120 150 Exp (10bar, 25ºC, NoEQ Brine) Exp (10bar, 40ºC, NoEQ Brine) Exp (10 bar, 25ºC, EQ Brine) Exp (10 bar, 40ºC, EQ Brine) Exp (10bar, 60ºC, EQ Brine)  Vp 0 2 4 6 8 10 040 80 120 160 200 Exp L (atm, 60ºC) Exp L (10bar, 60ºC) Exp L (56bar, 60ºC) VL-diss (%) Vp 0.0 0.9 1.8 2.7 3.6 4.5 040 80 120 160 200 Exp (atm, 60ºC) Exp (10bar, 60ºC) Exp (56bar, 60ºC)  Vp dissolution precipitation porosity mineralogy temperature pCO2 0 600 1200 1800 0100 200 300 400 500 600 VRock-diss (mm3) Vp 0 100 200 300 400 500 0100 200 300 400 500 600 VGp-ppt (mm3) Vp 0 0.5 1 1.5 2 2.5 3 3.5 4 0100 200 300 400 500 600 L25-atm-s D25-atm-s L40-10-b D40-10-b  Vp (a) (b) (c) (d) (e) (f) (g) (i) (h) 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 040 80 120 160 200 Exp (atm, 60ºC) Exp (10bar, 60ºC) Exp (56bar, 60ºC) VGp-ppt (%) Vp 0 600 1200 1800 030 60 90 120 150 L25-atm-a2.1 L60-atm-a2.1 L25-10-gp-e L40-10-gp-e L60-10-gp-e VL-diss (mm3) Vp 0 0.5 1 1.5 2 2.5 3 3.5 4 0100 200 300 400 500 600 L25-atm-s D25-atm-s L40-10-gp.e D40-10-gp-e  Vp 0.0 0.9 1.8 2.7 3.6 4.5 040 80 120 160 200 L60-atm-a3.5 L60-10-gp-e L60-34-gp-e  Vp 52 Chapter 3: Column experiments and 1D modeling increased by decreasing the temperature, and that the difference between VL-diss at 25ºC and VL-diss at 60ºC considerably increased with pCO2. Indeed, after 100 pore volumes, VL-diss (25ºC)/VL-diss (60ºC) was 1.08 under atmospheric conditions and 1.99 under 10 bar of pCO2. Fig. 3.6 Modeled porosity variation along the column during experiments at 25 ºC (in green), 40 ºC (in orange) and 60 ºC (in red). Solid, dashed and dotted lines represent atmospheric, 10 bar and 34 bar pCO2 conditions, respectively, and plain and empty-square lines indicate experiments with limestone and dolostone, respectively. Colored areas indicate simulated values using the initial (AGp-0) and final (AGp-f) gypsum reactive area. T effect under atmospheric and 10 bar of pCO2, respectively (a and b), mineralogy effect (c) and pCO2 effect on porosity changes (d). Variation of calcite saturation index (SICal) along the column is shown in (a, b and d). It should be noted that the limestone dissolution rate decreased with temperature under both pCO2 conditions in contrast to the reported increase in the calcite dissolution rate constants with temperature from 25 to 100 ºC and under pCO2 from atmospheric to 55 bar (Pokrovsky (a) (b) (c) (d) 53 et al. 2005, 2009). An explanation for the observed T effect on the calcite dissolution rate was provided by the different solution saturation state along the column at different temperatures. From the CrunchFlow simulations run with the same reactive surface area of calcite (ACal = 120 m2m/m3bulk), it was deduced that dissolution at 25 ºC took place under more undersaturated conditions than at 60 ºC under all pCO2 conditions (i.e., SICal is further from equilibrium at lower T; see SICal variation of atmospheric and 10 bar pCO2 experiments in Fig. 3.6a, b, respectively). Hence, the trend (faster dissolution rate at lower T) is driven by the solution saturation state rather than by the temperature dependence of the dissolution rate constant. The fact that CO2 solubility is higher with decreasing temperature (Duan and Sun, 2003) also contributes to the observed trend of faster dissolution rate with decreasing temperature. The higher solubility results in buffering of the solution pH at lower values (see smaller pH values at lower T in Tables 2.4 and 3.1). As regards the T effect on gypsum precipitation, results show that by lowering temperature, gypsum precipitation rate was favored under 10 bar of pCO2 and not favored under atmospheric pressure (Fig. 3.5b). Under all pCO2 conditions, two stages were distinguished in limestone column experiments where gypsum precipitated. In the first stage, to match the small initial S deficit, the same value of gypsum reactive area for all experiments was used (AGp ≈ 0.01 m2m/m3bulk). In the second stage, this initial value was increased to match the growing S deficit as gypsum kept on precipitating. In the former stage, although the solution was slightly more supersaturated with respect to gypsum at the lowest T, the gypsum precipitation rate was faster at the highest T. This demonstrated that in this first stage the increased rate constants prevailed over the solution saturation state. In the latter stage, the dependence of the gypsum precipitation rate on T shows differences between experiments run under atmospheric and 10 bar of pCO2 conditions. Under atmospheric pCO2 conditions, since the difference in calcite dissolution rates at 25 and 60 ºC was slight, the initial trend remained and the gypsum precipitation rate was faster at the highest T. By contrast, under 10 bar of pCO2 conditions the trend changed and the gypsum precipitation rate was faster at the lowest T owing to the larger Ca concentration at this T (Fig. 3.3a, b). Overall, the coupled process of limestone dissolution and gypsum precipitation always increased porosity (under any pCO2), the increase being higher at the lowest temperature (Fig. 3.5c). Simulations showed that an increase in temperature does not affect the trend of porosity variation along the column but reduces porosity creation under all pCO2 conditions (Fig. 3.6a, b). Note: This chapter is based on the submitted article to Int. J. Greenh. Gas Control: Garcia-Rios, M., Luquot, L., Soler, J.M., Cama, J. 2015. Influence of the flow rate on dissolution and precipitation features during percolation of CO2-rich sulfate solutions through fractured limestone samples. Chapter 4 Influence of the flow rate on dissolution and precipitation features during percolation experiments with fractured limestone and sandstone cores 4.1 Introduction This chapter presents the experimental results of a set of percolation experiments which consist of injecting CO2-rich solutions through fractured core samples under Hontomín reservoir conditions. Cores were made of limestone and sandstone rocks from the Hontomín reservoir. Experiments were run under different flow rates and sulfate content of the injected solution. X-ray computed microtomography (XCMT) was used to characterize changes in fracture volume induced by dissolution and precipitation processes. In addition, measurement of the pressure difference between the inlet and the outlet of the sample and of the aqueous chemistry enabled the determination of permeability changes and net reaction rates. A discussion including the influence of flow rate on the reactions, evolution of dissolution patterns and permeability changes during fracture dissolution is likewise presented. 62 Chapter 4: Percolation experiments through fractured cores 4.2 Results Table 4.1 shows the list of the percolation experiments. Core and fracture dimensions and mineralogical composition of limestone and sandstone samples were described in Section 2.1.1 (Table 2.1, Fig. 2.2). Three different input solutions were injected through the fractured limestone and sandstone cores (sulfate-free solution no-s, gypsum-undersaturated solution gp-u and gypsum-equilibrated solution gp-e; Section 2.1.2; Table 2.2). All experiments were run under P = 150 bar, pCO2 = 62 bar and T = 60 ºC, and flow rates varied from 0.2 to 60 mL/h. Two flow-through apparatus (Icare Lab CSS I, Luquot and Gouze, 2009, and Icare Lab CSS II, Luquot et al., 2012) were used to reproduce the in situ reservoir conditions for CO2 sequestration (Section 2.1.3; Fig. 2.3c). Table 4.1 List of the percolation experiments. Experiment label QInput solution label (mL/h) L1-no-s 1no-s L60-no-s 60 no-s L0.2-gp-e 0.2 gp-e L1-gp-e 1gp-e L5-gp-e 5gp-e L60-gp-e 60 gp-e S1-no-s 1no-s S5-no-s 5no-s S60-no-s 60 no-s S5-gp-u 5gp-u S60-gp-u 60 gp-u S0.2-gp-e 0.2 gp-e S1-gp-e 1gp-e S5-gp-e 5gp-e S60-gp-e 60 gp-e Labels of the experiments are coded by rock type, flow rate Q and input solution label: rockQ-input label. All experiments were run under P = 150 bar, pCO2 = 62 bar and T = 60 °C. no-s = sulfate-free solution. gp-u = gypsum-undersaturated solution. gp-e = gypsum-equilibrated solution. LIMESTONE (L) SANDSTONE (S) 63 4.2.1 Initial fracture characterization The initial fracture aperture was obtained by means of three independent experimental measurements: (1) hydraulic aperture determined from hydraulic test using Eq. (2.10), (2) geometric aperture obtained from X-ray computed microtomography (XCMT) data, and (3) geometric aperture obtained from scanning electron microscopy (SEM) examination. The relationship between the aperture and the volume of the fracture at initial time ( ) was considered to be , assuming that the initial fracture volume was defined as a parallelepiped. Table 4.2 shows the obtained fracture aperture and fracture volume values at (comparison of , and ). Table 4.2 Measured (weighted) mass of fractured core ( ), fracture permeability ( ), and fracture geometry (a and V) obtained by hydraulic measurement ( and ), XCMT ( and ) and SEM ( and ) at initial time ( ). Experiment Mmeas kahVhaXr VXr asVsah/asaXr/as label g m2μm mm3μm mm3μm mm3 L1-no-s 2.995 9.25E-13 3.33 0.56 - 7.47 6.4 1.07 0.5 - L60-no-s 3.015 - - - - 4.84 4.1 0.70 - - L0.2-gp-e - - - - - - 9.0 1.48 - - L1-gp-e 2.739 - - - 30.91 8.06 6.7 1.06 - 4.6 L5-gp-e 2.850 4.64E-13 2.36 0.37 - 5.71 5.3 0.82 0.4 - L60-gp-e 2.690 8.49E-12 10.10 1.57 38.07 7.94 8.1 1.26 1.2 4.7 S1-no-s 2.974 6.72E-13 2.84 0.49 - - 7.1 1.22 0.4 - S5-no-s 2.896 3.44E-13 2.03 0.33 - 2.85 2.5 0.41 0.8 - S60-no-s 2.831 4.34E-12 7.22 1.19 - 10.52 9.2 1.51 0.8 - S5-gp-u 2.852 - - - - 4.33 3.8 0.62 - - S60-gp-u 2.685 - - - - 7.67 7.2 1.10 - - S0.2-gp-e 3.101 - - - - - 9.2 1.56 - - S1-gp-e 2.865 - - - 35.78 7.44 6.4 1.06 - 5.6 S5-gp-e 2.850 1.76E-12 4.60 0.74 - - 5.2 0.84 0.9 - S60-gp-e 3.092 1.24E-11 12.19 2.10 - 11.38 9.50 1.63 1.3 - ah and Vh = hydraulic aperture and volume. aXr and VXr = geometric aperture and volume from XCMT; VXr values with grey background correspond to experiments where XCMT was not performed, and were calculated using the relationship between SEM and XCMT results of experiments L1-gp-e, L60-gp-e and S1-gp-e (see Section 4.3.1 ). as and Vs = geometric aperture and volume from SEM. comparison LIMESTONE SANDSTONE Hydraulic measurement XCMT SEM t = t0 64 Chapter 4: Percolation experiments through fractured cores Initial hydraulic aperture and fracture permeability ) could only be calculated according to Eqs. (2.10) and (2.11) in the eight percolation experiments where was measurable (Table 4.2; , and ). In the other seven experiments was smaller than the minimum measurable value from the beginning of the experiment. Three unreacted fractured cores were characterized by XCMT (Table 4.2; and ; exp. L1-gp-e, L60-gp-e and S1-gp-e). Data was acquired at the National Institute for Lasers, Plasma and Radiation Physics (NILPRP; Bucharest-Magurele, Romania), and the processing of the X-ray microtomography data was carried out by Voxaya (Montpellier, France), providing characterization of the fracture geometry. Once the percolation experiments were finished, SEM analyses of all fractured samples were performed to obtain the dimensions of the fracture and observe features of mineral dissolution and precipitation. From SEM images, the initial fracture aperture could be measured in unaltered fracture regions, near the outlet and far away from the dissolution front (Table 4.2; and ). Detailed information about XCMT data acquisition and SEM analyses is presented in Section 2.1.1. Initial fracture characterization was performed by all three methods mentioned only for experiment L60-gp-e. Good agreement was observed between the measured parameters using hydraulic and SEM methods (Table 4.2; = 1.2). In contrast, results from XCMT were noticeably higher (Table 4.2; = 4.7). Good agreement between hydraulic measurement and SEM was also obtained in the four other experiments where varied from 0.8 to 1.3, and was poorer in the three other experiments where ranged from 0.4 to 0.5 (Table 4.2). Discrepancy between these two methodologies could be explained by the fact that the initial fracture apertures obtained from hydraulic measurement correspond to the minimum aperture of the fracture (controlling permeability), whereas those from SEM were an average of four measured values where reaction was not supposed to occur. However, significantly larger discrepancies existed when XCMT results were compared with those obtained with hydraulic measurement and SEM, which could be attributed to the limited resolution of the technique (14 μm of pixel size) and the high background noise from the data. For the three experiments where the sample was characterized by XCMT, was around 5. 65 4.2.2 Aqueous chemistry Fig. 4.1 illustrates the variation in Ca and S concentration over time during limestone experiments, and Fig. 4.2 shows the variation in Ca, S and Si concentration over time during sandstone experiments. Note that variation in chemical composition of experiment S5-no-s is not shown in Fig. 4.2 because technical problems prevented the measurement. Fig. 4.1 Variation in the increase of Ca (a) and S (b) concentrations over time in the percolation experiments with fractured limestone cores, using no-s input solution (open symbols) and gp-e input solution (solid symbols) at = 0.2 mL/h (in violet), = 1 mL/h (in green), = 5 mL/h (in red) and = 60 mL/h (in black). Time for experiments at = 0.2, 1 and 5 mL/h is plotted in the lower x-axis and time for experiments at = 60 mL/h is plotted in the upper x-axis. The output Ca concentration in both limestone and sandstone experiments was always higher than the input concentration throughout the entire experimental run (ΔCa > 0), indicating calcite dissolution (dotted lines in Fig. 4.1a and Fig. 4.2a). Overall, the Ca concentration increase was larger at slow flow rates ( from 0.2 to 5 mL/h) than at the highest flow rate ( = 60 mL/h). At the slowest flow rate ( = 0.2 mL/h), the output Ca concentration continuously increased throughout the experiment. In some experiments with from 1 to 5 mL/h, Ca was released in two stages defined by an initial peak of ΔCa followed by a sharp decrease to approach an almost unvarying concentration (e.g., exp. L1-gp-e and 66 Chapter 4: Percolation experiments through fractured cores L5-gp-e). At = 60 mL/h, ΔCa was fairly constant during the entire experimental run. Moreover, in the experiments run at the same flow rate ( = 1, 5 or 60 mL/h) but with different input solution (no-s, gp-u or gp-e), ΔCa was larger in those with lower input S concentration (open and semi-solid symbols in Fig. 4.1a and Fig. 4.2a). In both the limestone and sandstone experiments with S concentration in the injected solution (gp-u and gp-e), the output S concentration was always lower than the input one, leading to a permanent S deficit (ΔS < 0; dashed lines in Fig. 4.1b and Fig. 4.2b). In experiments run under slow flow rates ( = 0.2, 1 and 5 mL/h) ΔS was very small (zero within the analytical error) during the early stage, whereas in the experiments with the fastest flow rate ( = 60 mL/h) this small value was observed during the whole experiment. Sulfur deficit indicated precipitation of a sulfur-rich mineral. In sandstone experiments the measured output Si concentration was slightly higher than the input one (solid lines in Fig. 4.2b). Given that the calculated pH ranged from 3.3 to 4.4 during the experiments (Chapter 5), Si was only released from dissolving microcline since quartz dissolution rate is negligible at acid pH (Bandstra et al., 2008). In addition, Si concentration increased by decreasing flow rate. Owing to the obtained low output Si concentration, microcline dissolution was not taken into account to calculate the changes in fracture volume shown in Section 4.3.1. 4.2.3 Permeability Fracture permeability was calculated according to Eq. (2.11) in the eight percolation experiments where could be measured (Table 4.2). In the other seven experiments, was initially smaller than the minimum measurable value ( ) preventing calculation. Fig. 4.3 shows that fracture permeability increased over time in the eight experiments regardless of the sulfur content of the injected solution. Note that precipitation of a S-rich phase did not prevent the permeability increase. 67 Fig. 4.2 Variation in the increase of Ca (left column-a) and S and Si (right column-b) concentrations over time in the percolation experiments with fractured sandstone cores, using no-s input solution (open symbols), gp-u solution (semi–solid symbols) and gp-e input solution (solid symbols) at = 0.2 mL/h (in violet), = 1 mL/h (in green), = 5 mL/h (in red) and = 60 mL/h (in black). Dotted lines in (a) indicate Ca concentrations and solid and dashed lines in (b) indicate Si and S concentrations, respectively. 68 Chapter 4: Percolation experiments through fractured cores In all limestone experiments, two stages were observed (Fig. 4.3a). In the initial stage the increase in permeability was much slower than in the second one. In particular, in the S- rich experiments (L5-gp-e and L60-gp-e), the initial stage was longer than in the S-free experiment (L1-no-s) and shorter in the experiment with the fastest flow rate (L60-gp-e). All sandstone experiments showed a stepped increase in permeability with steps becoming shorter by increasing flow rate (Fig. 4.3b). Exceptionally, in experiment S5-gp-e, the permeability increased similarly to that in limestone experiments (i.e., an initial slow stage followed by a fast one). The different behavior of permeability increase between limestone and sandstone experiments is related to different developed dissolution patterns, as explained in Section 4.3.4. Note also that in experiment S60-gp-e the permeability increase was suddenly interrupted by a sharp fall, likely caused by detachment of quartz grains. Nonetheless, an immediate recover of permeability indicated that this phenomenon itself was unable to prevent the permeability increase in contrast to reported permeability reduction by transport of particles in limestone percolation experiments (Luquot et al., 2014). Fig. 4.3 Variation in fracture permeability during limestone (dashed lines-a) and sandstone (solid lines-b) experiments. = 1 mL/h (green line), = 5 mL/h (red and pale red lines) and = 60 mL/h (black and grey lines). In the plots with grey background, upper-x axis indicates time for exp. L1-no-s and S5-gp-e and lower-x axis shows time for exp. L5-gp-e, S1-no-s and S5-no-s. 69 4.2.4 Identification of dissolution and precipitation processes The most prominent mineralogical change during any CO2-flooding experiment is carbonate dissolution (Weibel et al., 2011). Indeed, the changes in solution chemistry observed in the current limestone and sandstone experiments were mostly bound to calcite dissolution (ΔCa > 0). In the sandstone experiments, the low Si release (ΔSi > 0) was associated to microcline dissolution. Feldspar dissolution is commonly reported in CO2 experiments (e.g., Fisher et al., 2010; Wandrey et al., 2011). However, given the slow feldspar dissolution rate relative to that of calcite (about five orders of magnitude) at the pH of this study (3.3-4.4), and the short duration of the experiments (up to ≈ 100 h), feldspar dissolution is of little significance in these experiments. Changes in solution chemistry in the experiments with S-rich injected solution indicated precipitation of a S-rich phase (ΔS < 0). MicroRaman analysis showed that this S- rich phase was always gypsum (Fig. 4.4). Fig. 4.4 SEM images and MicroRaman spectrum of a thin section (section 2 in Fig. 2.2a) from the reacted fracture in experiment L5-gp-e: (a) dissolved calcite in the fracture surfaces and precipitated gypsum crystals. (b) Detailed view of the gypsum (Gp) crystals that grow at the expense of calcite (Cal) dissolution. Note the strong alteration of the fracture surfaces leading to formation of high microporosity. The y values indicate the distance from the inlet (y = 0) of the fracture along the flow direction (y). (c) MicroRaman spectra. The presence of the two characteristic peaks of water at ≈ 3500 cm-1 confirms that gypsum is the sulfate precipitated phase. The standard spectra of gypsum and anhydrite are from Downs (2006). XCMT and SEM were used to identify and localize these reactions along the fractures. In the limestone samples, two different types of thin sections were prepared from the reacted cores according to the observed fracture evolution. Sections parallel to the flow 200 μm50 μm a) b) c) Gp Cal y = 1.5 mm y = 1.5 mm L5-gp-e 2H2O 76 Chapter 4: Percolation experiments through fractured cores SEM images of the experiments with no-s input solution showed only calcite dissolution, leaving non-dissolved grains of quartz and microcline along the fracture. This phenomenon led to non-uniform aperture increases, independently of the flow rate, but the faster the flow rate the larger the non-uniformity (Fig. 4.9). XCMT images showed that at Q = 1 mL/h calcite dissolution was little and uniform along the fracture (Fig. 4.11a). XCMT was not performed in this experiment. At Q = 5 ml/h, calcite dissolution yielded uniform dissolution from the inlet to the middle of the fracture and a wormhole from the middle to the outlet (Fig. 4.11b). At Q = 60 mL/h, uniform dissolution took place along the fracture but slightly localized at the inlet (Fig. 4.11c). In the S-rich experiments with gp-u solution XCMT images showed only calcite dissolution. Although gypsum precipitation was inferred from aqueous chemistry (Table 4.3), gypsum crystals were not observed by XCMT. Increasing the flow rate dissolution patterns changed from wormhole (Q = 5 mL/h; Fig. 4.11d) to nearly uniform dissolution (Q = 60 mL/h; Fig. 4.11e). SEM images of the S-rich experiments using gp-e solution showed negligible dissolution at the slowest flow rate experiment. Noticeable dissolution, but differently distributed along the fracture, was observed in the rest of experiments with higher flow rate (Fig. 4.10). In the 1 and 60 mL/h experiments, dissolution was mostly homogeneous along the fracture except at the outlet, whereas in the 5 mL/h experiment it was localized around the middle of the fracture (Fig. 4.10). Other than calcite dissolution, precipitation of gypsum and an unidentified phase was observed. While gypsum crystals grew at expense of calcite dissolution, the unknown phase formed at expense of microcline dissolution (close-up images in Fig. 4.10b). SEM-EDX analysis indicated that this phase was formed of Si, Al, and K but it could not be identified by microRaman since fluorescence emission masked the Raman signal. XCMT images revealed that the dissolution was initially controlled by wormhole formation to finish up as a uniform in the 1 mL/h experiment. In contrast, at 60 mL/h the process occurred inversely (Fig. 4.11f, h). X-ray microtomography was not perfomed in the 5 mL/h experiment. 77 Fig. 4.10 SEM images of the reacted sandstone fractures in experiments with gp-e input solution at (a) Q = 0.2 mL/h (S0.2-gp-e), (b) Q = 1 mL/h (S1-gp-e), (c) Q = 5 mL/h (S5-gp-e) and (d) Q = 60 mL/h (S60-gp-e). Close-up images in (b) show precipitated gypsum (left) and precipitated unidentified aluminosilicate (right). The y values indicate the distances from the inlet (y = 0) of the fracture along the flow direction (y). Yellow circle in (d) indicates a possible site of a grain detachment. Fig. 4.12 depicts the fracture-length profiles from XCMT data that show distinct evolution of the dissolution processes along the sandstone fractures (detailed explanation of the figure legend is given in the previous section (Section 4.2.4.1)). Due to a more uniform a) S0.2-gp-e b) S1-gp-e c) S5-gp-e d) S60-gp-e 1 mm 50 mm 50 mm y = 0 mm y = 18 mm 1 mm y = 0 mm y = 18 mm 1 mm y = 18 mm y = 0 mm y = 0 mm y = 18 mm 1 mm 78 Chapter 4: Percolation experiments through fractured cores dissolution in the sandstone samples than that observed in the limestone ones, black and orange profiles coincided in most of the experiments (Fig. 4.12b, c, e, h). In fact, lack of coincidence was observed where a localized dissolution (e.g., wormhole) developed (Fig. 4.12d, f). Fig. 4.11 XCMT results. Total volume of reacted (A) and unreacted (A’) fractures in sandstone experiments with no-s solution at (b) Q = 5 mL/h and (c) Q = 60 mL/h, with gp-u solution at (d) Q = 5 mL/h and (e) Q = 60 mL/h and with gp-e solution at (f) Q = 1 mL/h and (h) Q = 60 mL/h. XCMT was not performed in experiment S1-no-s and XMT analysis was not carried out in experiment S5-gp- e. Color scale bars are in pixels (1 pixel = 14 μm). Black and white sections (perpendicular to flow) show the fracture morphology with associated dissolution patterns from the inlet (left) to the outlet (right) of the cores. (d) S5-gp-u (a) S1-no-s (c) S60-no-s no-sinput gp-u input Q= 1 mL/h Q= 5 mL/h Q= 60 mL/h (e) S60-gp-u ---------no experiment------- --no XCMT-- (b) S5-no-s flow gp-einput (f) S1-gp-e (g) S5-gp-e (h) S60-gp-e -----no XMT analysis----- inlet outlet inlet outlet inlet outlet inlet outlet inlet outlet inlet outlet inlet outlet flow totalfracture volume (unreacted) totalfracture volume (reacted) (initial and from dissolution; VXr + VXr-ppt) A’ A A’ AA A A AA 79 Fig. 4.12 Fracture-length profiles that show the volume of unreacted (blue lines) and reacted (black lines) fractures and the largest connected volume from dissolution (orange lines) in sandstone experiments with no-s solution at (b) Q = 5 mL/h and (c) Q = 60 mL/h, experiments with gp-u solution at (d) Q = 5 mL/h, (e) Q = 60 mL/h and experiments with gp-e solution at (f) Q = 1 mL/h and (h) Q = 60 mL/h. 4.3 Discussion 4.3.1 Fracture volume calculated from mass balance and XCMT Once the percolation experiments were finished, five limestone and six sandstone fractured samples were characterized by XCMT to determine the geometry of the reacted fracture (Table 4.3; ). The changes in fracture volume obtained from aqueous chemistry original 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 0 2 4 6 8 10 12 14 16 18 Core length (mm) Slice area (mm2) (d) S5-gp-u (a) S1-no-s (c) S60-no-s no-sinput gp-u input Q= 1 mL/h Q= 5 mL/h Q= 60 mL/h (e) S60-gp-u totalfracture volume (unreacted) totalfracture volume (reacted) (initial andfrom dissolution; VXr ) connected volume from dissolution ---------no experiment------- -----no XCMT----- (b) S5-no-s flow gp-einput 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 0 2 4 6 8 10 12 14 16 18 Core length (mm) Slice area (mm2) 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 0 2 4 6 8 10 12 14 16 18 Core length (mm) Slice area (mm2) (f) S1-gp-e (g) S5-gp-e (h) S60-gp-e -----no XMT analysis----- 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 0 2 4 6 8 10 12 14 16 18 Core length (mm) Slice area (mm2) 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 0 2 4 6 8 10 12 14 16 18 Core length (mm) Slice area (mm2) 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 0 2 4 6 8 10 12 14 16 18 Core length (mm) Slice area (mm2) 80 Chapter 4: Percolation experiments through fractured cores ( ) and from XCMT data ( ) were compared at the end of the experiment (Table 4.3; ; comparison of and ). From aqueous chemistry, the volume of dissolved calcite and precipitated gypsum and the variation of fracture volume at the end of the experiment ( ) were calculated according to Eqs. (2.2) and (2.7) and the mass balance equations (Eqs. (2.3) and (2.5); Table 4.3). The loss of mass calculated from the aqueous chemistry ( ) was in very good agreement with the measured loss of mass ( ) giving high reliability of the chemical analyses (Table 4.3). Information about the total volume of fracture after reaction (void space including the initial fracture volume and the dissolution-induced volume) and the volume of gypsum precipitated in the fracture was obtained by processing the X-ray microtomography data sets. was always an amount of gypsum precipitated inside the fracture or in a wormhole. The change in fracture volume shown in Table 4.3 was determined as (4.1) where the total volume of the unreacted fracture was obtained using XCMT analysis only for three experiments (Table 4.2; ). The unreacted fracture volume obtained from XCMT was ≈ 7 times larger than that from SEM. Hence, this factor of ≈ 7 was used to calculate for the rest of experiments, from which XCMT analysis of the unreacted fracture was not performed (Table 4.2; values in grey background). Comparison between the variation in fracture volume obtained from aqueous chemistry ( ) and that obtained from XCMT analysis ( ) showed very good agreement in six experiments in which ranged from 0.92 to 1.05 (Table 4.3). Agreement was poorer in the other four experiments (Table 4.3; 0.72 ≤ ≤ 0.86). Problems in the XCMT segmentation process could be the cause of the observed discrepancies in the latter experiments. Void space was probably interpreted to be epoxy resin due to the large pixel size (14 μm) and the high background noise from the data. As a result, in the experiments where dissolution patterns developed in contact with the epoxy resin area, discrepancy increased. In most of the experiments where a wormhole developed in a central position (e.g., Figs. 4.7a, d and 4.11f), a very small discrepancy was observed. 81 Table 4.3 Measured mass, measured and calculated loss of mass and variation in fracture volume determined from aqueous chemistry and XCMT at the end of the experimental runs (t=tf). Regarding the volume of precipitated gypsum in the limestone experiments, volume calculated from aqueous chemistry was greater than that obtained from XCMT analysis in all experiments (Table 4.3; / = 2.9 in experiments L1-gp-e and L5-gp-e). The largest difference between measurements was obtained in the experiment with the fastest flow rate (L60-gp-e), where precipitation of gypsum was not observed by XCMT (Table 4.3; not observed) in contrast to the volume of precipitated gypsum calculated from aqueous chemistry (Table 4.3; = 6.73 mm3). In this experiment, the output Ca concentration was smaller than that obtained in experiments under slower flow rate ( from 0.2 to 5 mL/h). This fact and the shorter residence time probably induced the precipitation of smaller gypsum crystals, which would be more difficult to identify by XCMT. Moreover, in experiment L60-gp- Comparison Experiment tf Mmeas ΔMmeas VCal-diss VGp-ppt ΔVch ΔMcalc VXr VXr-ppt ΔVXr ΔVXr/ΔVch label (h) (g) L1-no-s 30.75 2.957 -0.039 14.92 - 14.92 -0.040 22.60 - 15.13 1.01 L60-no-s 4.70 2.945 -0.069 23.50 - 23.50 -0.064 21.73 - 16.89 0.72 L0.2-gp-e 45.72 2.809 - 2.37 0.53 1.84 -0.005 - - - - L1-gp-e 71.91 2.690 -0.049 21.53 5.47 16.06 -0.046 21.84 1.91 13.78 0.86 L5-gp-e 18.50 - - 23.76 8.34 15.42 -0.045 21.77 2.87 16.06 1.04 L60-gp-e 4.51 2.648 -0.042 22.11 6.73 15.38 -0.044 19.44 not detected 11.50 0.75 S1-no-s 6.31 2.970 -0.004 1.76 - 1.76 -0.005 - - - - S5-no-s 1.38 2.892 -0.004 - - - - 4.71 - 1.86 - S60-no-s 1.32 2.797 -0.034 12.45 - 12.45 -0.034 23.59 - 13.07 1.05 S5-gp-u 5.58 2.838 -0.014 6.72 0.50 6.23 -0.017 10.03 not detected 5.70 0.92 S60-gp-u 5.59 2.564 -0.121 45.76 2.97 42.79 -0.117 48.2 not detected 40.57 0.95 S0.2-gp-e 102.20 - - 7.34 0.65 6.69 -0.018 - - - - S1-gp-e 37.34 2.825 -0.040 14.05 0.16 13.89 -0.038 21.88 not detected 14.44 1.04 S5-gp-e 17.66 - - 14.24 4.66 9.58 -0.028 - - - - S60-gp-e 4.03 - - 27.42 1.28 26.14 -0.071 32.01 not detected 20.63 0.79 tf = experimental time. Mmeas = measured mass of fractured core (weighted). ΔMmeas and ΔM calc = measured and calculated (from aqueous chemistry) loss of mass. VCal-diss and VGp-ppt = volume of dissolved calcite and precipitated gypsum (aqueous chemistry). VXr and VXr-ppt = volume of fracture (void space) and volume of precipitated gypsum (XCMT). ΔVch and ΔVXr = variation of fracture volume from aqueous chemistry (ch) and XCMT (Xr). Initial values used to calculate ΔMmeas and ΔVXr are given in Table 4.2. The propagated error of ΔVch is 10-12 %. LIMESTONE SANDSTONE t = tf Aqueous chemistry XCMT (g) (mm3) (mm3) 82 Chapter 4: Percolation experiments through fractured cores e was probably overestimated owing to the high uncertainty in the aqueous chemistry (Fig. 4.1b; ΔS ≈ 0 within the analytical error). In the experiment with the slowest flow rate ( = 0.2 mL/h) this comparison could not be performed because the XCMT analysis was not carried out. Gypsum was not observed either in the SEM images probably due to the minor calcite dissolution. In the sandstone experiments, precipitated gypsum was not detected by XCMT. In experiments S60-gp-u and S60-gp-e with remarkable , precipitation of small gypsum crystals owing to the fast flow rate ( = 60 mL/h) was likely the cause to impede detection of gypsum by XCMT. In the experiment with the largest (S5-gp-e) this comparison could not be performed because the XCMT analysis was not carried out. Regardless of the methodology used to calculate changes in fracture volume, it was observed that the final fracture volume was always larger than the initial one even when gypsum precipitated in experiments with sulfate-rich solutions (gp-u and gp-e). 4.3.2 Influence of flow rate on reaction The effect of flow rate on the resulting dissolution and precipitation processes was investigated by varying the flow rate from 1 to 60 mL/h in the experiments with no-s and gp- u input solutions and from 0.2 to 60 mL/h in the experiments with gp-e input solution (Table 4.1). Mineral dissolution is limited by reaction rates in the fracture surfaces and solute transport (advection vs. diffusion) within the fracture. The Damköhler ( ) and Péclet ( ) numbers parameterize the relative magnitude of these processes. In the formulation that we use, compares the magnitudes of the advective solute flux along the fracture with diffusion from the fracture surface, and compares mineral reaction rates with advective flux along the fracture. The number is given by , where is the mean fluid velocity in m/s defined as , is the mean aperture of the fracture in m, is the molecular diffusion coefficient of the reactants in m2/s, is the volumetric flow rate in m3/s and is the width of the fracture in m. The molecular diffusion coefficient at 60 ºC ( = 5.36. 10-9 m2/s) was calculated using Arrenhius equation and the molecular diffusion coefficient in water assumed for all species at 25 ºC ( ≈ 10-9 m2/s). 83 To assess the flow rate effect on calcite dissolution, experiments with similar calcite dissolution rate ( ; Xu et al., 2012) must be compared. The initial calcite dissolution rate at the inlet of all fractures could be considered identical because the reactive surface area (same geometry), the dissolution rate constant (same T, P), the proton activity (same pH) and the saturation state (similar input solutions) were almost the same at this initial point (y = 0 mm and ). Once the injected solution reacts with the fracture surface (calcite) under different flow rates, the dissolution rate parameters ( , , and ) will be differently affected with time and distance. Along the fractures, P and T were constant and changes in pH were very similar in all experiments (average input pH = 3.23 ± 0.02; average output pH = 3.91 ± 0.20), yielding no variation in and very similar variation in . Pokrovsky et al. (2009) demonstrated that was basically not affected by pCO2 directly (only through the term). Additionally, saturation state should not noticeably change between experiments because all output solutions were undersaturated with respect to calcite (SICal ≈ -2.4 and = 4 10-3 for experiments at = 60 mL/h and SICal ≈ -1.6 and = 2.5 10-2 for experiments at = 0.2, 1 and 5 mL/h) and considered to be far from equilibrium (SICal < -1.5; Cubillas et al., 2005; Xu et al., 2012). Hence, differences in between experiments at different flow rates were probably not caused by the small differences in . As for the reactive surface area term, it should be similar in all experiments. However, as a consequence of the coupled chemical reactions and transport of elements through the fracture, different dissolution patterns (homogeneous vs. heterogeneous/localized dissolution) may occur, yielding a noticeable variation in the accessible area of mineral that can react with the solution along the fracture. Consequently, differences in the amount of Ca obtained from dissolution at the outlet of the fracture (net reaction rate) between experiments were probably related to changes in the reacted area. At this large scale, Luhmann et al. (2014) and Luquot and Gouze (2009) measured a decrease in during wormhole formation. In addition, at the micrometer scale the flow rate also affects , favoring an increase for slow flow rates (microporosity development; e.g., Fig. 4.6b and c) and an decrease for fast flow rates (surface smoothing; e.g., Fig. 4.6d). Similar observations were reported by Deng et al., (2013), Luquot et al. (2014) and Noiriel et al. (2009). These authors characterized microporosity formation during limestone dissolution for several flow-through experiments. Noiriel et al. (2009) proposed the so-called sugar-lump model that reproduces the 84 Chapter 4: Percolation experiments through fractured cores experimental results (formation of microporosity) by a dissolution mechanism that induces the formation of small calcite particles with larger reactive surface area. To calculate the Damköhler number ( ), a velocity for calcite dissolution far from equilibrium in m/s ( ) is used. This type of rate law (linear dependence on ) applies to calcite dissolution under acidic conditions (Atanassova et al., 2013; Palandri and Kharaka, 2004; Pokrovsky et al., 2009). Experimental studies have reported that below pH 5.5, the calcite dissolution rate is pH-dependent and limited by mass transfer processes (e.g., Plummer et al., 1978; Sjöberg and Rickard, 1984). Therefore, calcite dissolution rate should increase with fluid velocity because the mass transfer of reactants and products is enhanced by the local fluid velocity near the mineral surface. This trend has been observed by several authors based on discontinuities in the evolution of the output solution concentration due to changes in the flow rate regime (Elkhoury et al., 2013; Noiriel et al., 2007). Noiriel et al. (2007) showed that the chemical flux of Ca2+ in mol/s increased with increasing flow rate. In this study, the calculated volumes of dissolved calcite per hour (mm3/h) also increased with increasing flow rate (Table 4.4; e.g., from 0.05 to 4.90 mm3/h in limestone experiments and from 0.07 to 6.80 mm3/h in sandstone experiments by increasing from 0.2 to 60 mL/h with gp-e solution). Additionally, the volume of dissolved calcite per injected volume (mm3/mL) was calculated and showed the opposite tendency: e.g., decreased by a factor of ≈ 3.2 with increasing the flow rate from 0.2 ( = 1) to 60 mL/h ( = 346) in both limestone and sandstone experiments with gp-e (Table 4.4). It was probably caused by the shorter residence time in experiments under faster flow rate. Note that at = 0.2 mL/h, was lower than expected in both limestone and sandstone experiments. The explanation for this was unclear as the output solution was far from equilibrium and there was no evidence of the existence of more diffusive zones that limited the overall dissolution process and retarded the exit of the reaction products. To compare experiments performed with the same rock and flow rate, the dissolution rates were calculated at the same experimental time (experiments with (*) in Table 4.4), in particular at the shortest experimental time . It was expected to obtain similar values of and in experiments under the same flow rate and similar evolution of fracture geometry (equivalent reacted area) (Table 4.4). This was evidenced, for 85 instance, in limestone experiments at = 60 mL/h (wormhole formation: L60-no-s and L60- gp-e) and in sandstone experiments at = 5 mL/h (wormhole formation: S5-gp-u and S5-gp- e*). Table 4.4 Péclet (Pe) and Damköhler (Da) numbers and net reaction rates expressed as volume of dissolved calcite, precipitated gypsum and variation in fracture volume per time and injected volume. However, in the two limestone experiments run at = 1 mL/h, and were larger in the experiment using no-s input solution (L1-no-s and L1-gp-e*). The reacted area for the no-s experiment was larger because of the formation of channel branching near the outlet (Fig. 4.7a, c). Likewise, and were smaller in the sandstone experiment run at = 60 mL/h with gp-e owing to a more localized dissolution near the outlet than those for the other two sandstone experiments with nil or lower sulfate content in the input solution (Fig. 4.10 c, e, g). This fact confirmed the effect of the reacted area on the net calcite dissolution rate (including both reaction and transport). QPe Da tfVCal-diss/tfVGp-ppt/tfΔVch/tfVinj (tf)VCal-diss/Vinj (tf) VGp-ppt/Vinj (tf)ΔVch/Vinj (tf) (mL/h) (h) (mL) L0.2-gp- e 0.2 1 1.9E-05 46 0.05 0.01 0.04 9 0.26 0.06 0.20 L1-no-s 31 0.49 - 0.49 31 0.49 - 0.49 L1-gp-e* 72 0.30 0.08 0.22 72 0.30 0.08 0.22 L5-gp-e 529 7.6E-07 19 1.28 0.45 0.83 93 0.26 0.09 0.17 L60-no-s 5 5.00 - 5.00 282 0.08 - 0.08 L60-gp-e 5 4.90 1.49 3.41 271 0.08 0.02 0.06 S0.2-gp- e 0.2 1 1.9E-05 102 0.07 0.01 0.07 20 0.36 0.03 0.33 S1-no-s 6 0.28 - 0.28 6 0.28 - 0.28 S1-gp-e* 37 0.38 0.004 0.37 37 0.38 0.004 0.37 S5-no-s 1 - - - 7 - - - S5-gp-u 6 1.20 0.09 1.12 28 0.24 0.02 0.22 S5-gp-e* 18 0.81 0.26 0.54 88 0.16 0.05 0.11 S60-no-s 1 9.43 - 9.43 79 0.16 - 0.16 S60-gp- u6 8.19 0.53 7.66 335 0.14 0.01 0.13 S60-gp-e 4 6.80 0.32 6.49 242 0.11 0.01 0.11 tcVCal-diss/tcVGp-ppt/tcΔVch/tcVinj (tc)VCal-diss/Vinj (tc) VGp-ppt/Vinj (tc)ΔVch/Vinj (tc) (h) (mL) L1-gp-e* 31 0.30 0.08 0.22 0 0.30 0.08 0.22 S1-gp-e* 6 0.27 0.003 0.26 0 0.27 0.003 0.26 S5-gp-e* 6 1.26 0.10 1.16 0 0.25 0.02 0.23 Pe and Da = Péclet and Damkhöler numbers. tf = total experimental time; tc = time of the shortest experiment in those run at the same rock and flow rate (see text in Section 4.3.2). * indicates experiments where tc was calculated. Vinj = injected volume. Exp. label (mm3/h) (mm3/h) (mm3/mL) 7.6E-07 346 6.3E-08 Rate values calculated at tc 60 60 1 1 5 6 3.8E-06 29 Rate values calculated at tf 6 346 3.8E-06 6.3E-08 (mm3/mL) 92 Chapter 4: Percolation experiments through fractured cores Fig. 4.14 shows variation in fracture permeability versus the number of equivalent fracture volumes for the experiments where was measurable. Variation in could not be measured in the two experiments with the slowest flow rate (Table 4.2; = 1; L0.2-gp-e and S0.2-gp-e) from which, the slowest increase would be expected, leading to the maximum number of fracture volumes until breakthrough. All experiments under faster flow rate, where was measurable, showed a remarkable increase. The increase was very sharp for experiments where a single dominant wormhole developed (e.g., L1-no-s, S5-gp-e) and more gradual for the experiments with more uniform dissolution patterns and, in general, under faster flow rates (e.g., L5-gp-e, S60-gp-e). In addition, in experiments where a single wormhole developed, the significant increase in occurred after a relatively few fracture volumes. For the 5 and 60 mL/h experiments the number of fracture volumes until breakthrough increased due to formation of more uniform dissolution structures. In this case, dissolution took place over larger fracture surface areas. The trend in the permeability responses showing the typical transition from a single dominant wormhole channel to a more uniform structure was likewise observed in percolation experiments performed in porous media, but more ramified wormhole structures formed (Fredd and Fogler, 1998, Luhmann et al., 2014). 4.4 Summary and conclusions Injection of CO2-rich sulfate solutions through fractured limestone and sandstone cores always produced an increase in permeability and fracture volume even when gypsum precipitated in the fractures (experiments using gypsum-equilibrated solutions). In general, the two methods (aqueous chemistry and XCMT analysis) used to calculate the variation in fracture volume induced by calcite dissolution and gypsum precipitation showed good agreement, yielding differences around 5%. Major discrepancies between them arose when wormholes developed along the lateral sealing and small gypsum crystals precipitated in experiments under fast flow rate. In both cases, the XCMT resolution and image quality limited an accurate quantification of the resulting volumes of fracture and precipitated gypsum. By increasing the flow rate, under the same pH and far from equilibrium conditions, (1) the volume of dissolved calcite per time increased, confirming that calcite dissolution in 93 the fracture was transport controlled, and (2) the volume of dissolved calcite per injected volume decreased, likely caused by the smaller residence time in experiments run under faster flow rate. In addition, the formation of more uniform geometries under fast flow rates (i.e., increase of the reacted area) led to an increase in the amount of calcite dissolved. At the micrometer scale, slow flow rates (e.g., Q = 1 and 5 mL/h) led to an increase in reactive surface area caused by an enhancement of microporosity, whereas fast flow rates (e.g., Q = 60 mL/h) tended to form smooth fracture surfaces which did not favor an increase in reactive surface area. At the same flow rate, the amount of calcite dissolved was larger in some experiments with sulfate-free solution owing to the larger reacted area related to the developed dissolution feature. Another factor contributing to the larger calcite dissolution in sulfate-free experiments than in sulfate-rich ones could be the sulfate inhibitory effect on calcite dissolution, together with gypsum coatings leading to calcite passivation. In general, dissolution patterns in limestone and sandstone experiments varied from face dissolution to wormhole formation and uniform dissolution by increasing the flow rate (larger and smaller ), confirming the tendency observed in previous studies (Detwiler et al., 2003; Szymczak and Ladd, 2009). Nonetheless, to predict them, initial surface heterogeneities must be taken into account since they were decisive in the evolution of the dissolving fractures. Variation in fracture permeability was found to be highly dependent on flow rate and developed dissolution pattern. Thus, permeability increase was very sharp when a single dominant wormhole developed, whereas it was more gradual when uniform dissolution occurred under fast flow rate. The number of fracture volumes to breakthrough increased with flow rate. This phenomenon was attributed to the formation of more uniform dissolution structures, which induced dissolution over larger fracture surface areas. 94 Chapter 4: Percolation experiments through fractured cores Chapter 5 Dissolved CO2 effect on two fractured reservoir rocks: comparison and 2D modeling 5.1 Introduction This chapter is divided in two parts. In the first one, the geochemical response of two fractured reservoir rocks (limestone and sandstone) to the injection of a CO2-rich solution under the Hontomín reservoir conditions was compared. The main difference between the limestone and the sandstone is the presence of quartz and microcline grains (≈ 35%) in the latter rock, which affects the evolving geochemical processes and, consequently, could influence the CO2 storage capacity and injectivity of the reservoir. In the second part, 2D reactive transport simulations that reproduce the variation in aqueous chemistry and fracture geometry of the experiments were performed to estimate flow and reaction kinetics parameters. Under the experimental conditions, pH could not be measured. The calculated pH (CrunchFlow) increased from 3.3 (input solution) to 4.4 (steady-state output solution). 96 Chapter 5: Comparison and 2D modeling 5.2 The role of silicate minerals on the CO2 storage capacity and injectivity The reactivity of two fractured reservoir rocks (limestone composed of 100% calcite and sandstone composed of 65.7% calcite, 27.8% quartz and 6.5% microcline) in contact with a CO2-rich sulfate solution was compared. As shown in Chapter 4, calcite dissolution was the dominant reaction in both limestone and sandstone experiments, becoming the only process considered to calculate the volumes of dissolved rock. Negligible dissolution of K-feldspar (microcline) and nil dissolution of quartz led to the existence of inert regions in reactive zones of the sandstone experiments. The influence of the inert regions on rock dissolution, mineral precipitation and fracture volume variation was evaluated. Two sets of experiments with gp-e input solution run under flow rates of 0.2, 1, 5 and 60 mL/h were compared (Table 4.1; four limestone experiments: L0.2-gp-e, L1-gp-e, L5-gp-e and L60-gp-e; four sandstone experiments: S0.2-gp-e, S1-gp-e, S5-gp-e and S60-gp-e). Overall, the volume of dissolved rock was larger in the sandstone experiments than in the limestone ones (Fig. 5.1a), likely caused by the evolved dissolution pattern. The presence of inert silicate grains in the sandstone experiments favored more extended dissolution structures than the localized ones in the limestone experiments. As a result, a larger area of calcite could interact with the CO2-rich solution resulting in a larger amount of calcite dissolution. The XCMT images show the difference in dissolution patterns between limestone and sandstone experiments (e.g., comparison of Fig. 4.7c and Fig. 4.11f). Only when was 5 mL/h was larger in the limestone experiment. This was related with the zone in which dissolution started: in the fracture (limestone experiment L5- gp-e) or in the rock matrix through initial heterogeneities (sandstone experiment S5-gp-e). Net rock dissolution was likely favored when started in the fracture. The volume of gypsum precipitated was always larger in the limestone experiments, even if calcite dissolution was smaller (Fig. 5.1b). In the sandstone experiments, dispersion of nuclei for precipitation on the inert surfaces (quartz grains) was likely the cause of less precipitation. This phenomenon was already observed when inert wood chips (Rötting et al., 2008) and glass beads (Offeddu et al., 2015) were added in columns packed with calcite grains to prevent calcite passivation. 97 Fig. 5.1 Experimental variation of volume of dissolved rock (a), volume of precipitated gypsum (b) and fracture volume (c) with time in limestone (dashed lines) and sandstone (solid lines) experiments with gp-e input solution, under = 0.2 mL/h (in violet), = 1 mL/h (in green), = 5 mL/h (in red) and = 60 mL/h (in black). The resulting and always yielded a larger increase in fracture volume in sandstone experiments (Fig. 5.1c). Hence, it appears that the CO2 storage capacity would be more favored in a sandstone reservoir than in a limestone one because the increase in porosity is higher and larger extended distribution of created volume occurs (uniform dissolution). The different distribution of created volume between limestone and sandstone experiments was responsible for the different variation in fracture permeability (Fig. 5.2). Limestone dissolution tended to be localized (wormhole), whereas sandstone dissolution tended to be extended (uniform). To illustrate it fracture permeability variation was compared between a limestone experiment where a wormhole developed (L60-gp-e; Fig. 5.2a) and a sandstone experiment where uniform dissolution occurred (S60-gp-e; Fig. 5.2b). In the limestone experiment, the increase in fracture permeability started with the formation of a localized preferential path (t1), which continuously enlarged (t2 and t3), resulting in a gradual increase in permeability (from t1 to t3). In the sandstone experiment, the fracture permeability increase was more complex. In this case, the enlargement of a first preferential path (first permeability increase; t1) was constrained by the presence of inert grains. The initial path 98 Chapter 5: Comparison and 2D modeling enhancement stopped as calcite totally dissolved and silicate grains contacted the solution. Thereafter, additional paths developed (t2), keeping permeability constant until a larger channel originated (t3 with a3 > a1, a2), allowing permeability to increase. This stepped increase in fracture permeability contrasts with the sharper increase observed in limestone experiments. Given the resulting differences in fracture permeability variation, an advantage of a progressive stepped increase in permeability in the sandstone experiments is that risks associated with changes in the mechanical properties of a reservoir, induced by a sharp permeability increase during CO2 injection, could be minimized. Moreover, in sandstone experiments, the lower number of fracture volumes necessary to start the permeability increase could facilitate the CO2 injection and, hence, reduce the energetic storage costs. Another benefit that stems from the more extended distribution of created volume in sandstone experiments is the enhancement of porous connectivity, which favors capillary trapping. Fig. 5.2 Variation in fracture permeability as a function of number of fracture volumes (t/ τ) and the associated distribution of created volume in the experiments L60-gp-e (a) and S60-gp-e (b). k(t0) indicates initial fracture permeability. a0 •t0, a0 •t1, a1 (a1> a0) •t2, a2 (a2≤ a1) •t3, a3 (a3> a1, a2) a1 a2a2 a3 a0 a1 a2 a3 •t0, a0 •t1, a1 (a1> a0) •t2, a2 (a2> a1) •t3, a3 (a3> a2) (a) Limestone L60-gp-e (b) Sandstone S60-gp-e 99 5.3 (2D) Reactive transport modeling Simulations of experiment L0.2-gp-e (face dissolution), L1-gp-e (wormhole) and S60- no-s (uniform dissolution) are shown in this section. In addition, discussion on the fitting parameters used to adjust the model to the experimental data (measured Ca and S concentrations and porosity variation) is also provided. Rectangular coordinates were used to model experiments which developed face and uniform dissolution patterns (L0.2-gp-e and S60-no-s), whereas rectangular and cylindrical coordinates were used in the experiment where a wormhole developed (L1-gp-e). When cylindrical coordinates were used, it was assumed that the initial fracture was already cylindrical (i.e., a very small wormhole). 5.3.1 Face dissolution A model with rectangular coordinates and flow update was used to simulate the experimental data of experiment L0.2-gp-e (face dissolution). A good match between the experimental and simulated Ca and S concentrations versus time was achieved by considering an initial De value of 3.0 10-12 m2/s in the rock matrix and adjusting the initial calcite and gypsum reactive surface areas (ACal = 250 m2m/m3bulk; AGp = 10 m2m/m3bulk) (SIM_A in Fig. 5.3a, b). Nonetheless, the simulated variation in porosity (Fig. 5.3c) did not match the actual variation measured by SEM (Fig. 4.6a). Considering half of the fracture core (due to symmetry), the SEM image shows 100% porosity up to 12 m in the x direction (normal distance to fracture) at the inlet of the fracture (colored area in Fig. 5.3c, d), and no porosity increase at the outlet. The initial De value was calculated using Eq. (2.21) with D0 and n values equal to 10-9 m2/s and 2.5, respectively. These values are common values reported in literature (see Section 2.2.2.3). A reasonable match of the variation of both the Ca and S concentrations with time (SIM_B in Fig. 5.3a, b) and porosity with distance normal to fracture (Fig. 5.3d) was only obtained by reducing the initially estimated De value to 6.0 10-14 m2/s and adjusting the values of ACal and AGp (2000 and 100 m2m/m3bulk, respectively). Model parameters used in this simulation are given in Table A.6, Appendix A. Experimental porosity variation was plausibly reproduced by the model: at the inlet of the core the calculated porosity was higher than 85% over the first 11.3 m in the x direction (the measured porosity was 100% up to 12 m; Fig 100 Chapter 5: Comparison and 2D modeling 4.6a, y = 0 mm). At the outlet, the calculated porosity was slightly higher than the measured one (20% instead of 5%; Fig 4.6a, y = 20 mm). The changes in mineral content caused a noticeable increase in porosity at the inlet of the fracture and almost no increase afterwards (Fig. 5.3d). The different increase in porosity between inlet and outlet of the fracture (markedly larger at the inlet) compared well with the observed dissolution pattern (face dissolution). Fig. 5.3 Experiment L0.2-gp-e (face dissolution); simulations with rectangular coordinates and flow update: (a,b) Variation in the experimental and simulated Ca and S concentration versus time and (c,d) simulated porosity variation with distance normal to fracture. Colored areas in (c,d) indicate the zone with 100% porosity measured by SEM at the inlet of the fractured core. These changes in porosity and the associated changes in permeability have translated into a slight change in Darcy velocity in the fracture. At the end of the experiment, solution 3.5e-2 4.0e-2 4.5e-2 5.0e-2 5.5e-2 6.0e-2 6.5e-2 7.0e-2 010 20 30 40 50 Exp Ca(out) SIM_B Exp Ca(in) SIM_A Ca (mol/kgw) time (h) 2.2e-2 2.3e-2 2.4e-2 2.5e-2 2.6e-2 2.7e-2 2.8e-2 010 20 30 40 50 Exp S(out) SIM_B Exp S(in) SIM_A S (mol/kgw) time (h) (a) (b) (c) (d) SIM_A SIM_B 0 20 40 60 80 100 0 2e-5 4e-5 6e-5 8e-5 inlet (y = 0 mm) middle (y = 10 mm) outlet (y = 20 mm)  distance normal to fracture (m) 0 20 40 60 80 100 0 2e-5 4e-5 6e-5 8e-5 inlet (y = 0 mm) middle (y = 10 mm) outlet (y = 20 mm)  distance normal to fracture (m) 101 still flowed preferentially along the fracture (Vy; Fig. 5.4b), with a negligible deviation towards the rock matrix (Vx; Fig. 5.4a). Fig. 5.4 Velocity field for experiment L0.2-gp-e (face dissolution) at t = 46 h; Velocity (m3/m2/yr) in the x direction (Vx; left plot) and in the y direction (Vy; right plot). 5.3.2 Wormhole Simulation of the wormhole experiment would require the use of cylindrical coordinates and flow update. However, flow update is not implemented in CrunchFlow when cylindrical coordinates are used. Given this limitation, a model with rectangular coordinates and flow update was used to simulate just the very initial stage of the experiment by assuming that wormhole formation was not initiated yet. This approach allowed a successful match of the initial variation in Ca and S concentration with time (black-solid lines in Fig. 5.5a, b) but, as expected, did not reproduce the porosity variation with distance normal to fracture. The match of this initial stage was achieved by considering a De value of 3.0 10-12 m2/s and adjusting the calcite and gypsum reactive surface areas (ACal = 2000 m2m/m3bulk; AGp = 100 m2m/m3bulk). Note that these areas were one order of magnitude higher than those used to fit the face-dissolution experiment using the same De value. The measured (SEM) porosity variation was defined by a wormhole with a radius of 600 m at the inlet and a radius of 430 m at the outlet (Fig. 4.6b) whereas the calculated porosity was only higher than 80% in the Chapter 6 General conclusions The main conclusions of this thesis are: 1) Regarding the effect of P, pCO2, T, mineralogy, acidity and solution saturation state on the coupled reactions of calcite/dolomite dissolution and gypsum precipitation (crushed rock): - Under all pCO2 conditions, low temperature favored limestone dissolution rate although the calcite dissolution rate constants increase with temperature. This inverse tendency was explained by the fact that limestone undersaturation increased by decreasing the temperature, which suggested that the process was thermodynamically controlled. - In experiments using gypsum-undersaturated solutions, gypsum did not precipitate and the amount of dissolved limestone was found to be slightly higher than that obtained in experiments using gypsum-equilibrated solutions. The decrease in calcite dissolution rate could be associated to the sulfate inhibitory effect and/or passivation of the calcite grain surfaces. - As expected under the conditions of this study, the volume of dissolved limestone was larger than that of dolostone owing to the well-known faster calcite dissolution kinetics. Likewise, a pCO2 increase implies a pH decrease that enhances substantially calcite dissolution rate with respect to that of dolomite. In addition, gypsum induction 110 Chapter 6: Conclusions time was longer when limestone dissolved and precipitation increased gradually. When dolostone dissolved, gypsum precipitated quickly and precipitation remained steady. - When raising pCO2, the limestone dissolution rate increased along the column because of the direct pH effect on the calcite dissolution rate. Dissolution of the carbonate minerals in acidic pH was controlled, under atmospheric pressure, by the protons provided by the strong acid (HCl or H2SO4), whereas under high pCO2, H2CO3 partial dissociation controlled the dissolution. Model results showed that if brine acidity was controlled by a strong acid, dissolution occurred exclusively at the first rock-brine contact, raising the pH at ≈ 7 and limiting the limestone dissolution further away. In contrast, simulations under high pCO2 conditions showed that pH remains acidic (≈ 5) and the brine was permanently undersaturated with respect to calcite and dolomite (due to the carbonic acid buffer capacity), yielding a higher increase in porosity all over the rock-brine contact. This suggested that calcite/dolomite dissolution induced by CO2-rich solutions tends to extend the dissolution fronts (1) favoring the CO2 storage capacity of the reservoir and (2) preventing localized dissolution which would lead to sharp changes in hydrodynamic rock properties, harmful for CO2 injection. - A good match between the CrunchFlow and PhreeqC (v.3) reactive transport calculations and the experimental data was obtained. Rate laws including the values of the rate constants were taken from literature. The fit of the model to the experimental data was performed by adjusting the values of the reactive surface areas. The calcite and dolomite reactive surface area values had to be diminished by two orders of magnitude from the initially calculated geometric surface areas. A possible explanation for the small areas could be given by the transport (diffusion) control of the dissolution reactions at pH < 5. It should be noted that a single value of the reactive area for calcite provided a good fit of the model to all experimental results, supporting the applicability of this modeling approach. The values for dolomite were more variable but continued to be within the model uncertainty. - Overall, the coupled process of limestone/dolostone dissolution and gypsum precipitation always increased porosity (under any pCO2). This suggested that 111 gypsum precipitation cannot decrease the reservoir rock porosity nor impede CO2 injection. 2) Regarding the influence of the flow rate on dissolution and precipitation features in the percolation experiments with fractured limestone and sandstone cores (fractured rock): - Injection of CO2-rich solutions through fractured limestone and sandstone cores always produced an increase in permeability and in fracture volume even when gypsum precipitated. - In general, the two methods (aqueous chemistry and XCMT analysis) used to calculate the variation in fracture volume induced by calcite dissolution and gypsum precipitation showed good agreement, yielding differences < 4%. Major discrepancies between them arose when wormholes developed along the lateral sealing and small gypsum crystals precipitated in experiments under fast flow rate. The XCMT resolution and image quality limited an accurate quantification of the resulting volumes of fracture and precipitated gypsum. - By increasing the flow rate, under the same pH and far from equilibrium conditions, (1) the volume of dissolved calcite per time increased, confirming that calcite dissolution in the fracture was transport controlled, and (2) the volume of dissolved calcite per injected volume decreased, likely caused by the smaller residence time in experiments under faster flow rate. In addition, the formation of more uniform geometries under fast flow rates (i.e., increase of the reacted are) led to an increase in the amount of calcite dissolved. At the micrometer scale, slow flow rates (Q = 1 and 5 mL/h) led to an increase in reactive surface area caused by an enhancement of microporosity, whereas fast flow rates (Q = 60 mL/h) tended to form smooth fracture surface which did not favor an increase in reactive surface area. - At the same flow rate, the amount of calcite dissolved was larger in some experiments with sulfate-free solution owing to the larger reacted area related to the developed dissolution feature. Another factor contributing to the larger calcite dissolution in sulfate-free experiments than in sulfate-rich ones could be the sulfate inhibitory effect on calcite dissolution, together with gypsum coatings leading to calcite passivation. 112 Chapter 6: Conclusions - In general, dissolution patterns in limestone and sandstone experiments varied from face dissolution to wormhole formation and uniform dissolution by increasing the flow rate (larger and smaller ), confirming the tendency observed in previous studies (Detwiler et al., 2003; Szymczak and Ladd, 2009). Nonetheless, to predict them, initial surface heterogeneities must be taken into account since they were decisive in the evolution of the dissolving fractures. - Variation in fracture permeability was found to be highly dependent on flow rate and developed dissolution pattern. Thus, permeability increase was very sharp when a single dominant wormhole developed, whereas it was more gradual when uniform dissolution occurred under fast flow rate. The number of fracture volumes to breakthrough increased with the flow rate. This phenomenon was attributed to the formation of more uniform dissolution structures, which induced dissolution over wider fracture surface areas. 3) Regarding the geochemical response of the two main Hontomín reservoir rocks (limestone and sandstone) to injection of a CO2-rich sulfate solution, and the 2D simulations of the percolation experiments: - On the one hand, during the interaction between CO2-rich sulfate solutions and fractured reservoir rocks, the presence of inert silicate grains in sandstone favored the occurrence of largely distributed dissolution structures in contrast to localized dissolution in limestone. Consequently, a larger area of calcite could interact with the CO2-rich solution resulting in a larger amount of calcite dissolution. On the other hand, dispersion of precipitation nuclei on the surface of the inert silicates yielded smaller volume of gypsum precipitated in sandstone than in limestone. As a result, in sandstone reservoirs, the larger increase in fracture volume, along with the more extended distribution of the created volume, would favor the increase of the CO2 storage capacity. - The different distribution of created volume between limestone and sandstone fractured reservoir rocks promoted a different variation in fracture permeability. A progressive stepped increase in permeability for sandstone is preferred to a sharp increase in permeability for limestone to minimize risks related with CO2 injection, favor capillary trapping and reduce the energetic cost of storage. 113 - The 2D reactive transport models reproduced the variation in aqueous chemistry and in porosity of the experiments by adjusting the calcite reactive surface area value (ACal) and, in some cases, the effective diffusion coefficient value (De) derived from literature (3.0 10-12 m2/s). As in the experiments with crushed rock, ACal values had to be diminished from the initially calculated geometric surface area to account for the transport control of the calcite dissolution reaction at pH < 5, which increased by decreasing the flow rate. De values for sandstone (9.5 10-12 m2/s) were higher than those for limestone (3.0 10-12 m2/s and 6.0 10-14 m2/s), as it is found in literature for this type of sedimentary rocks and similar porosity. 114 Chapter 6: Conclusions References Alcalde, J., Marzán, I., Saura, E., Martí, D., Ayarza, P., Juhlin, C., Pérez-Estaún, A., Carbonell, R., 2014. 3D geological characterization of the Hontomín CO2 storage site, Spain: Multidisciplinary approach from seismic, well-log and regional data. Tectonophysics 627, 6–25. Alkattan, M., Oelkers, E.H., Dandurand, J., Schott, J., 2002. An experimental study of calcite dissolution rates at acidic conditions and 25 ºC in the presence of NaPO3 and MgCl2. Chem. Geol. 190, 291–302. ALM-09-008, 2010. Caracterización petrofísica de las facies jurásicas-cretácicas aflorantes en las inmediaciones de Aguilar de Campoo, equivalentes a los niveles almacén y sello presentes en la estructura geológica a utilizar como Planta de Desarrollo Tecnológico de almacenamiento geológico de CO2. Hito 1: Informe de muestreo. CIUDEN. Appelo, C.A.J., Parkhurst, D.L., Post, V.E.A., 2014. Equations for calculating hydrogeochemical reactions of minerals and gases such as CO2 at high pressures and temperatures. Geochim. Cosmochim. Acta 125, 49–67. Arvidson, R.S., Collier, M., Davis, K.J., Vinson, M.D., Amonette, J.E., Luttge, A., 2006. Magnesium inhibition of calcite dissolution kinetics. Geochim. Cosmochim. Acta 70, 583–594. Atanassova, R., Cama, J., Soler, J.M., Offeddu, F.G., Queralt, I., Casanova, I., 2013. Calcite interaction with acidic sulphate solutions: a vertical scanning interferometry and energydispersive XRF study. Eur. J. Mineral. 25, 331–351. 116 References Bachu, S., Adams, J.J., 2003. Sequestration of CO2 in geological media in response to climate change: capacity of deep saline aquifers to sequester CO2 in solution. Energy Convers. Manag. 44, 3151–3175. Bachu, S., Gunter, W.D., Perkins, E.H., 1994. Aquifer disposal of CO2: Hydrodynamic and mineral trapping. Energy Convers. Manag. 35, 269–279. Bandstra, J. Z., Buss, H. L., Campen, R. K., Liermann, L. J., Moore, J., Hausrath, E. M., Navarre-Sitchler, A.K., Jang, J., Brantley, S. L., 2008. Appendix: compilation of mineral dissolution rates. Kinetics of Water–Rock Interaction. Springer, New York, 737-823. Barrante, J.R., 1974. Applied Mathematics for Physical Chemistry. Prentice-Hall. Benson, S.M., Cole, D.R., 2008. CO2 sequestration in Deep Sedimentary Formations. Elements 4, 325-331. Bertier, P., Swennen, R., Laenen, D., Dreesen, R., 2006. Experimental identification of CO2- water-rock interactions caused by sequestration of CO2 in Westphalia and Buntsandstein sandstones of the Campine Basin (NE-Belgium). J. of Geochem. Exploration 89, 10-14. Boving, T.B., Grathwohl, P., 2001. Tracer diffusion coefficients in sedimentary rocks: correlation to porosity and hydraulic conductivity. J. Cont. Hydro. 53, 85–100. Carroll S., Smith M., Hao Y., 2012. Evolution of carbonate dissolution features produced under variable pCO2 conditions relevant to CO2-EOR and geologic CO2 storage. AiChE Annual meeting, conference proceedings 415b. Chopping, C., Kaszuba, J.P., 2012. Supercritical carbon dioxide–brine–rock reactions in the Madison Limestone of Southwest Wyoming: An experimental investigation of a sulfurrich natural carbon dioxide reservoir. Chem. Geol. 322-323, 223–236. Cubillas, P., Köhler, S., Prieto, M., Chaïrat, C., Oelkers, E.H., 2005. Experimental determination of the dissolution rates of calcite, aragonite, and bivalves. Chem. Geol. 216, 59–77. De Giudici, G., 2002. Surface control vs. diffusion control during calcite dissolution: Dependence of step-edge velocity upon solution pH. Am. Mineral., 87, 1279-1285. 117 De Marsily, G., 1986. Quantitative Hydrogeology. Academic Press, San Diego. Deng, H., Ellis, B.R., Peters, C.A., Fitts, J.P., Crandall, D., Bromhal, G.S., 2013. Modifications of carbonate fracture hydrodynamic properties by CO2-acidified brine flow. Energy & Fuels 27, 4221-4231. Detwiler, R.L., 2008. Experimental observations of deformation caused by mineral dissolution in variable-aperture fractures. J. Geophys. Res. 113, B08202. Detwiler, R.L., Glass, R.J., Bourcier, W.L., 2003. Experimental observations of fracture dissolution: The role of Peclet number on evolving aperture variability. Geophys. Res. Lett. 30, 1648. Detwiler, R.L., Rajaram, H., 2007. Predicting dissolution patterns in variable aperture fractures: Evaluation of an enhanced depth-averaged computational model. Water Resour. Res. 43, W04403. Dijk, P.E., Berkowitz, B., Yechieli, Y., 2002. Measurement and analysis of dissolution patterns in rock fractures. Water Resour. Res. 38, 1013. Domenico, P.A., Schwartz, F.W., 1990. Physical and Chemical Hydrogeology. John Wiley and Sons, New York. Downs, R.T., 2006. The RRUFF Project: an integrated study of the chemistry, crystallography, Raman and infrared spectroscopy of minerals. Program and Abstracts of the 19th General Meeting of the International Mineralogical Association in Kobe, Japan. O03-13 Duan, Z., Sun, R., 2003. An improved model calculating CO2 solubility in pure water and aqueous NaCl solutions from 273 to 533 K and from 0 to 2000 bar. Chem. Geol. 193, 257–271. Durham, W.B., Bourcier, W.L., Burton, E.A., 2001. Direct observation of reactive flow in a single fracture. Water Resour. Res. 37, 1–12.