@prefix vivo: . @prefix edm: . @prefix ns0: . @prefix dcterms: . @prefix dc: . @prefix skos: . vivo:departmentOrSchool "Science, Faculty of"@en, "Earth, Ocean and Atmospheric Sciences, Department of"@en ; edm:dataProvider "DSpace"@en ; ns0:degreeCampus "UBCV"@en ; dcterms:creator "Alt-Epping, Peter"@en ; dcterms:issued "2009-09-17T15:42:21Z"@en, "2000"@en ; vivo:relatedDegree "Doctor of Philosophy - PhD"@en ; ns0:degreeGrantor "University of British Columbia"@en ; dcterms:description "The permeability of the oceanic crust is the primary hydrologic parameter that controls the geometry, the vigor, and the duration of hydrothermal fluid flow at mid-ocean ridges. The coupling of fluid flow, temperature, and chemistry and the effect of the permeability on these coupled processes are assessed to determine whether geochemical data can be used to constrain estimates of basement permeabilities and the vigor of convection. The coupling of flow, temperature, and chemistry is investigated for an open, sedimented ridge setting from the onset of fluid motion after an initial state of a conductive temperature distribution and fluid stagnation, to a near steady-state convective system. Fluid residence times, physical water/rock ratios, temperature conditions, recharge and discharge rates, flow geometries, and the degree of fluid mixing are calculated for the evolving hydrothermal system and the influence of these parameters on mineral alteration and aqueous phase concentrations are discussed. The chemical evolution of the system suggests that despite differences in alteration patterns and the intensity of alteration for different basement permeabilities, the nature of the alteration reactions is unlikely to be a useful parameter for constraining the vigor of convection. Estimates of fluid velocity can be obtained using a formulation that relates the mass transfer between the solid and the fluid phase along the fluid's flowpath to the average velocity along the flowpath. Different regions in the hydrothermal system and a range of chemical species are examined to assess their usefulness in constraining estimates of average flow velocities. The results of these calculations suggest that the mass transfer of aqueous silica may be useful for estimating fluid flow velocities in hydrothermal systems, in particular in those regions of the system at or near quartz equilibrium so that aqueous silica concentration is buffered by quartz and correlated with the temperature distribution."@en ; edm:aggregatedCHO "https://circle.library.ubc.ca/rest/handle/2429/12860?expand=metadata"@en ; dcterms:extent "14588079 bytes"@en ; dc:format "application/pdf"@en ; skos:note "INSIGHTS INTO THE EVOLUTION OF A N OCEANIC H Y D R O T H E R M A L S Y S T E M A N D A METHOD FOR CONSTRAINING ESTIMATES OF THE VIGOR OF H Y D R O T H E R M A L CONVECTION by PETER ALT-EPPING B.Sc., Brock University, 1991 M . S c , Freiburg University, 1994 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in Department of Earth and Ocean Sciences We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH C O L U M B I A November 2000 © Peter Alt-Epping, 2000 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of .CftOZT/tt OC£At~> Sg£*JCj>S The University of British Columbia Vancouver, Canada Date l^Oi/g/Xgfc^- 0Xl ZB® DE-6 (2/88) Abstract The permeability of the oceanic crust is the primary hydrologic parameter that controls the geometry, the vigor, and the duration of hydrothermal fluid flow at mid-ocean ridges. The coupling of fluid flow, temperature, and chemistry and the effect of the permeability on these coupled processes are assessed to determine whether geochemical data can be used to constrain estimates of basement permeabilities and the vigor of convection. The coupling of flow, temperature, and chemistry is investigated for an open, sedimented ridge setting from the onset of fluid motion after an initial state of a conductive temperature distribution and fluid stagnation, to a near steady-state convective system. Fluid residence times, physical water/rock ratios, temperature conditions, recharge and discharge rates, flow geometries, and the degree of fluid mixing are calculated for the . evolving hydrothermal system and the influence of these parameters on mineral alteration and aqueous phase concentrations are discussed. The chemical evolution of the system suggests that despite differences in alteration patterns and the intensity of alteration for different basement permeabilities, the nature of the alteration reactions is unlikely to be a useful parameter for constraining the vigor of convection. Estimates of fluid velocity can be obtained using a formulation that relates the mass transfer between the solid and the fluid phase along the fluid's flowpath to the average velocity along the flowpath. Different regions in the hydrothermal system and a range of chemical species are examined to assess their usefulness in constraining estimates of average flow velocities. The results of these calculations suggest that the mass transfer of aqueous silica may be ii useful for estimating fluid flow velocities in hydrothermal systems, in particular in those regions of the system at or near quartz equilibrium so that aqueous silica concentration is buffered by quartz and correlated with the temperature distribution. iii Table of Contents Abstract ii Table of Contents iv List of Tables ix List of Figures x Acknowledgements xvi 1 Introduction 1 1.1 Objective of Study 1 1.2 Background and Review of Study Area - Middle Valley, Juan de Fuca Ridge 2 1.2.1 Geographical Setting, Site Description and Drilling Objectives 3 1.2.2 Geology 4 1.2.2.1 Tectonic Setting 4 1.2.2.2 Sedimentology 6 1.2.2.3 Petrology 8 1.2.3 Geophysics 9 1.2.3.1 Seismic Profiles and Logs 9 iv 1.2.3.2 Heat Flow Distribution and Temperature Measurements 11 1.2.4 Hydrology 14 1.2.4.1 Porosity and Permeabilities 14 1.2.4.2 Pressure Measurements 16 1.2.5 Geochemistry 18 1.2.5.1 Rock Geochemistry 18 1.2.5.1.1 The Sediments - Mineralogy and Geochemistry 18 1.2.5.1.2 The Basalts - Mineralogy and Geochemistry 25 1.2.5.2 Fluid Geochemistry 28 1.2.5.2.1 Pore Fluids 28 1.2.5.2.2 Vent Fluids 32 1.3 Conclusion 33 The Evolution of a Transient Convective System 43 2.1 Introduction 43 2.1.1 Review of Permeability Estimates and Concepts of Crustal Hydrothermal Convection 44 2.1.2 Mathematical Model 57 2.1.2.1 Governing Equations 57 2.1.2.2 Numerical Scheme 60 2.1.3 Model Design : 61 2.1.3.1 Flow Domain : 61 2.1.3.2 Initial and Boundary Conditions 62 2.1.3.3 Choice of Parameters 63 2.1.3.4 Discussion of Model Uncertainties 65 2.2 The Evolution of a Hydrothermal Convective System 68 2.2.1 The Evolution of Temperature and Velocity Fields 68 2.2.2 Average Basement Temperatures and Temperatures at Recharge and Discharge Zones 73 2.2.3 Fluxes through Recharge and Discharge Zones and the Sediment/Basement Interface 77 2.2.4 Flowlines and Travel Times -. 80 2.2.5 Water/rock Ratios (time-integrated fluxes) 92 2.2.6 Fluid Mixing 99 2.3 Conclusion 107 Chemical Evolution of a Hydrothermal Convective System 141 3.1 Introduction 141 3.1.1 Chemical Processes and Permeability Changes -Methods and Previous Results 142 3.1.2 Mathematical Model 147 3.1.2.1 Governing Equations 147 3.1.2.1.1 Reaction Rate Law 154 3.1.2.1.2 Activity Coefficient Model 162 3.1.2.1.3 Porosity and Reactive Surface Area 163 3.1.3 Numerical Model 167 vi 3.2 Reactive Transport Modeling 170 3.2.1 Hydrothermal Alteration of Oceanic Crust - Overview 170 3.2.2 Model Design 177 3.2.3 The Influence of Hydrothermal Circulation on Mineral Alteration, Fluid Composition, and Porosity 180 3.2.3.1 Recharge Zone 181 3.2.3.2 Center of Domain (Reaction Zone) 184 3.2.3.3 Discharge Zone 189 3.2.3.4 Lateral Flow into the Sediments 193 3.2.3.5 Alteration Patterns 195 3.2.3.6 Comparison of Alteration Patterns for Basement Permeabilities lx lO\" 1 4 m 2 and5xl0\" 1 4 m 2 200 3.2.3.7 Porosity Changes - The Impact of Alteration Patterns on Fluid Flow 203 3.3 Conclusion -207 4 A Method to use Geochemical Data for Constraining Estimates of the Vigor of Convection in a Hydrothermal System 234 4.1 Introduction 234 4.1.1 Chemical W/R Ratios in Dynamic Systems 236 4.1.2 Calculation of Chemical Water/Rock Ratios 237 4.2 Application of the Method 243 4.2.1 One-Dimensional Simulations 243 4.2.1.1 Results of Reactive Transport Calculations 246 vii 4.2.1.1.1 Effect of Dispersion 255 4.2.1.1.2 Effect of Flow Velocity 258 4.2.1.2 Physical and Chemical W/R Ratios 260 4.2.1.2.1 Intervals of Uniform Temperature Gradient (T1-T3) 261 4.2.1.2.2 Intervals of Uniform Alteration Assemblage (M1,M2) 263 4.2.1.3 Two-Dimensional Simulations 265 4.2.1.3.1 Recharge Zone 268 4.2.1.3.2 Discharge Zone 272 4.2.1.3.3 Lateral Flow 275 4.2.1.3.4 Vertical Flow through Sediments 279 4.3 Conclusion 281 5 Conclusion of Thesis 303 References 308 viii List of Tables Table 2.1: Summary of physical and hydrological properties of the sediments, the basalt, and the recharge/discharge zones 112 Table 3.1: Summary of the composition and the kinetic and thermodynamic parameters of the solid phase 211 Table 3.2: Major ion composition of seawater and secondary species 212 ix List of Figures Figure 1-1. Location map of Sites 855-858. The insert shows a map of the Juan de Fuca Ridge and the location of Middle Valley. (From: Davis et al. 1994) 36 Figure 1-2. Conceptual cartoon of hydrothermal circulation at Middle Valley, based on results of Leg 139. (From: Davis and Fisher, 1994) 37 Figure 1-3. Seafloor heat-flow measurements (W/m ) (From: Davis et al. 1994) 38 Figure 1-4. Permeability and temperature versus depth. The depth of the top of the first sill is shown as the horizontal dotted line. (From: Davis and Fisher, 1994) 39 Figure 1-5. Profile across Site 858 showing the distribution of alteration zones. A description of the zones is given in the text. (From: Leybourne and Goodfellow, 1994) 40 Figure 1-6. Mineralogical profile at Site 857. (From: Leybourne and Goodfellow, 1994)41 Figure 1-7. Profile of fluid composition at Site 857 42 Figure 2-1. Model flow domain, including initial and boundary conditions and solid medium properties 113 Figure 2-2. Evolution of temperature and velocity fields - basement permeability lx lO\" 1 4 m 2 114 Figure 2-3. Evolution of temperature and velocity fields - basement permmeability 1x10\" 1 5 m 2 115 • • 13 Figure 2-4. Evolution of temperature and velocity fields - basement permeability 1x10\" m 2 116 Figure 2-5. Temperature evolution at the recharge and discharge zones and evolution of average basement temperatures 117 Figure 2-6. Fluid mass fluxes (kg/year) through the recharge and discharge zones and across the sediment/basement interface 118 13 2 Figure 2-7. Particle flowpaths and traveltimes for basement permeabilities 1x10\" m -l x l 0 \" 1 5 m 2 119 Figure 2-8. Particle flowpaths and rate of temperature change for basement permeabilities l x l 0 \" 1 3 m 2 - l x l 0 \" 1 5 m 2 120 Figure 2-9. Particle flowpaths for steady state flow field for basement permeabilities lx lO\" 1 3 m 2 , lx lO\" 1 4 m 2 , and lx lO\" 1 5 m 2 121 Figure 2-10. Evolution of water/rock ratios (fluid mass (kg)/volume of rock (m )) -basement permeability lx lO\" 1 4 m 2 122 Figure 2-11. Evolution of water/rock ratios (kg/m3) - basement permeability lx lO\" 1 5 m2.123 3 13 2 Figure 2-12. Evolution of water/rock ratios (kg/m ) - basement permeability 1x10\" m .124 Figure 2-13. Evolution of average basement water/rock ratios (kg/m ) for full transient (black curves) and steady state flow simulation (grey curves) - basement permeabilities lx lO\" 1 3 m 2 - lx lO\" 1 5 m 2 125 Figure 2-14. Average basement water/rock ratios (kg/m ) versus permeability at various times for full transient and steady state flow simulation (20000 years) 126 Figure 2-15. Locations and labels of tracer release and sampling points 127 Figure 2-16. Tracer breakthrough curves for sampling locations X I , X2 - basement permeability lx lO\" 1 4 m 2 128 Figure 2-17. Tracer breakthrough curves for sampling locations X3 , X4 - basement permeability lx lO\" 1 4 m 2 129 Figure 2-18. Tracer breakthrough curves for sampling locations Y l , Y2 - basement permeability lx lO\" 1 4 m 2 130 Figure 2-19. Tracer breakthrough curves for sampling locations Y3, Y4 - basement permeability 1 x 10\"14 m 2 131 Figure 2-20. Tracer breakthrough curves for sampling locations X I , X2 - basement permeability 1x10\"15 m 2 132 Figure 2-21. Tracer breakthrough curves for sampling locations X3 , X4 - basement permeability lx lO\" 1 5 m 2 133 Figure 2-22. Tracer breakthrough curves for sampling locations Y l , Y2 - basement permeability lx lO\" 1 5 m 2 134 Figure 2-23. Tracer breakthrough curves for sampling locations Y3, Y4 - basement permeability lx lO\" 1 5 m 2 135 Figure 2-24. Tracer breakthrough curves for sampling locations X I , X2 - basement permeability 1x10\"13 m 2 136 Figure 2-25. Tracer breakthrough curves for sampling locations X3 , X4 - basement permeability 1x10\" m 137 Figure 2-26. Tracer breakthrough curves for sampling locations Y l , Y2 - basement permeability lx lO\" 1 3 m 2 138 Figure 2-27. Tracer breakthrough curves for sampling locations Y3, Y4 - basement permeability lx lO\" 1 3 m 2 139 xii Figure 2-28. Tracer breakthrough curves for the discharge zone (Zl) - basement permeabilities lx lO\" 1 5 m 2 , lx lO\" 1 4 m 2 , and lx lO\" 1 3 m 2 140 Figure 3-1. Compilation of chemical water/rock ratio versus depth measurements (after Fisher, 1998) and comparison with a collection of permeability versus depth data from various ridges (From: Davis and Fisher, 1994) 213 Figure 3-2. Chemical model domain - solid phase and fluid composition and initial and boundary conditions 214 Figure 3-3. Profile locations 215 Figure 3-4. Mineralogical profiles through the recharge zone at various times 216 Figure 3-5. Fluid composition profiles through the recharge zone at various times 217 Figure 3-6. Mineralogical profiles through the center of the domain at various times....218 Figure 3-7. Fluid composition profiles through the center of the domain at various times219 Figure 3-8. Temperature and aqueous silica distribution (in moles/kg) at 15000 years. .220 Figure 3-9. Mineralogical profile through the discharge zone at various times 221 Figure 3-10. Fluid composition profiles through the discharge zone at various times....222 Figure 3-11. Evolution of the vent fluid composition (a) and comparison with seawater and vent fluids from Middle Valley (Site 858) (b) 223 Figure 3-12. Distribution of the pH at 15000 years 224 Figure 3-13. Fluid composition profile from 30 m left of the discharge zone. The fluid composition profile through the sediments at the center of the domain is plotted below for comparison 225 XIII Figure 3-14. Mineralogical profile from 30 m left of the discharge zone. The mineralogical composition profile through the sediments at the center of the domain is plotted below for comparison 226 Figure 3-15. Transient precipitation patterns of quartz (in percent volume change) 227 Figure 3-16. Transient precipitation patterns of albite (in percent volume change) 228 Figure 3-17. Transient precipitation patterns of clinozoisite (in percent volume change).229 Figure 3-18. Transient precipitation patterns of chlorite (in percent volume change)...:.230 Figure 3-19. Mineralogical profiles for basement permeability 5xl0\" 1 4 m 2 231 Figure 3-20. Porosity changes through time at the recharge zone and discharge zone (in percent volume change) 232 Figure 3-21. Porosity changes through time for model domain (in percent volume change) 233 Figure 4-1. Column design, temperature distribution, solid phase composition, and important hydrologic parameters 286 Figure 4-2. Fluid composition profiles through the column at 500 and 15000 years 287 Figure 4-3. Mineralogical profile through the column at 15000 years. The profile includes the mineral assemblages M l , M2, and M3 288 Figure 4-4. The effects of dispersion on the concentration profile. Dispersivities 0 m and 100 m 289 Figure 4-5. The effect of the flow velocity on the concentration profile. Flow velocities 5 m/year and 50 m/year 290 Figure 4-6. Water/rock ratios for interval T l 291 Figure 4-7. Water/rock ratios for interval T2 292 Figure 4-8. Water/rock ratios for interval T3 293 Figure 4-9. Water/rock ratios for mineral assemblage M l 294 Figure 4-10. Water/rock ratio calculations for mineral assemblage M2 295 Figure 4-11. Flow and temperature fields at 10000 years, including the 'columns' for which average flow velocities are calculated (top) and temperature field, flow field, and flowlines at 15000 years (bottom) 296 Figure 4-12. Average fluid flow velocities for recharge zone for sediment (top) and basalt (bottom) sections 297 Figure 4-13. Average fluid flow velocities for discharge zone for sediments (top) and basalt (bottom) sections 298 Figure 4-14. Fluid composition and mineralogical profile through horizontal 'column' below the sediment/basement interface 299 Figure 4-15. Average fluid flow velocities for horizontal 'column' below the sediment/basement interface 300 Figure 4-16. Velocity estimates for the reaction zone, based on short flowpaths (20 m) and the assumption of strictly horizontal flow. Velocity in m/year 301 Figure 4-17. Velocity estimates for the sediment section at the center of the domain. ...302 XV Acknowledgements I would like to thank my supervisor Leslie Smith for his helpful suggestions and comments, his time and patience. I am also thankful to the members of my committee Roger Beckie and Greg Dipple whose doors were always open for questions and suggestions. I appreciate the help of Motomu Ibaraki who provided me with his flow code and introduced me to the mysteries of the UNIX operating system and Carl Steefel and Steve Yabusaki who provided me with the reactive transport code OS3D. This work has been supported by a grant from the Natural Sciences and Engineering Research Council of Canada. xvi 1 Introduction 1.1 Objective of Study The permeability of the oceanic crust is the primary hydrologic parameter that controls the geometry, the vigor, and the duration of hydrothermal fluid flow at mid-ocean ridges. Reliable data on the magnitude and distribution of crustal permeabilities are scarce. The range of measured or estimated permeabilities for the upper oceanic crust spans more than four orders of magnitude. One way of reducing our uncertainty in the magnitude of permeability is to use geochemical data. The intent of this study is to find geochemical constraints on estimates of the vigor of convection in the oceanic crust at a sedimented mid-ocean ridge setting. This study provides ideas about geochemical processes that are a consequence of the temperature and fluid flow regimes in this environment, how these processes may affect the evolution of the convective system as a result of porosity changes, and how geochemical processes can be used to estimate fluid flow velocities. On one hand the results help in the interpretation of data collected at mid-ocean ridges and suggest which data and where it should be collected to best constrain estimates of fluid flow. On the other hand the results illustrate the nature of the uncertainties in the estimates that originate from the complex interaction and coupling of fluid flow, temperature changes, and chemical reaction. The permeability of a medium is a key parameter for defining and understanding hydrological processes but it is also difficult to measure in-situ due to its scale dependence. While geochemical data can help 1 constrain the estimates of permeability, geochemical reactions, such as mineral dissolution and precipitation, and the concomitant porosity change make permeability a time dependent parameter and more difficult to assess. This first part of this study gives insight in the evolution of flow and temperature regimes in an oceanic hydrothermal system at a sedimented ridge. The evolution will be discussed with regards to flow geometries, particle flowpaths and travel times, water/rock ratios, advective fluid mixing, recharge and discharge rates, and temperature conditions. These parameters are assessed for a range of basalt permeabilities to search for trends or correlations that may be useful to constrain permeabilities. In addition, emphasis is placed on the geochemical significance of these parameters. The second part of this study describes the effect of the transient state of the hydrothermal system on chemical alteration patterns, its consequences in terms of porosity/permeability changes, and possible implications on the nature of fluid flow. In the third part of this study, a method to estimate fluid flow velocities based on the coupling of chemical mass transfer and fluid flow velocities is derived. The potential of the method and its limitations in constraining estimates of the vigor of fluid flow in an oceanic hydrothermal system is demonstrated. The method can be generalized to provide geochemically-based estimates of fluid velocity for any non-isothermal flow system. 1.2 Background and Review of Study Area - Middle Valley, Juan de Fuca Ridge 2 1.2.1 Geographical Setting, Site Description and Drilling Objectives The design of the model and the parameters used in this study are based on results from Legs 139 and Leg 169 of the Ocean Drilling Program to Middle Valley, a northern segment of the Juan de Fuca Ridge located about 300 km west of Vancouver Island in British Columbia, Canada (Figure 1.1). Both legs provided data on and insights into hydrothermal processes at a sediment-covered, mid-ocean ridge. Sedimented spreading centers are relatively rare but they are a very attractive target for drilling. A regionally continuous blanket of sediments over young oceanic crust acts as a hydrological seal and a thermal insulator of the underlying crust. This feature leads to a regime of high temperatures at relatively shallow depths in the crust. Recharge of seawater into the basement is reduced and discharge of hydrothermal fluids is restricted to distinct zones. At these zones of focused discharge of hydrothermal fluids, large massive sulfide deposits can be produced. The sediment cover makes it possible to measure heat flow variations on the seafloor, which may help determine the geometry of hydrothermal convection in the upper oceanic crust, the 'hydrothermal reservoir'. The sediments can preserve magmatic, tectonic and thermal events helping constrain the spatial and temporal relationship of these events (Davis et al. 1992b). The four drilling sites of Legs 139 and 169 (Sites 855-858) are outside the deepest part of the rift, in the southeastern area of Middle Valley (Figure 1.1) and are situated on an eastward tilted fault block (Davis and Fisher, 1994). Site 855 is located at the normal fault, which forms the eastern boundary of the valley (Figure 1.1). This fault is thought to be a zone where fluid recharge into the basement may occur. The objectives of 3 drilling at this site were to determine the geometry of the fault and its hydrologic significance by characterizing the nature and rate of fluid flow along this fault into the basement. Site 856 is located on and around \"Bent Hi l l \" , some 3 km southwest of Site 855, one of several small hills in the eastern part of the valley (Figure 1.1). It is a site of former active venting and the only site at which massive sulfides were found. Studying the genesis and nature of the massive sulfide body was the objective of drilling at this site. Site 857 is located on a fault-bounded block of oceanic crust, covered with a 470 m thick sequence of turbidites (Figure 1.1). The bounding faults are the normal fault at Site 855, about 5.2 km to the east and a major normal fault 1.5 km to the west. Site 857 was chosen to study hydrothermal convection in the relatively permeable basement which is sealed by a less-permeable sediment cover and to gain insight into chemical reactions that are controlling the composition of hydrothermal vent fluids. Site 858 is an active vent site where hot hydrothermal fluids discharge through the seafloor. It is part of the larger \"Dead Dog\" hydrothermal vent field, roughly 2 km north of Site 857 (Figure 1.1). Besides sampling the vent fluids and rocks at and around the upflow zone, determining the nature of fluid flow were the objectives of drilling at this site. 1.2.2 Geology 1.2.2.1 Tectonic Setting 4 The Juan de Fuca Ridge stretches 550 km between the Blanco and Savanco transform faults in the north-eastern Pacific (Figure 1.1). The ridge crest morphology is typical of a ridge with a medium to fast spreading-rate (Davis and Villinger, 1992). The full spreading rate in the northern part of the Juan de Fuca Ridge is about 54 mm/year (Currie and Davis, 1994). The Juan de Fuca Ridge is divided into generally continuous 50- to 100 km long ridge-segments and spreading takes place in small axial rift zones at the crest of these segments (Figure 1.1). The northern end of the Juan de Fuca Ridge however, has a more complex crest morphology. Here, probably because of the intersection with the cooler Sovanco fracture zone and a resulting diminished magma supply, spreading currently takes place in three deep axial valleys; West Valley, Endeavour Valley, and Middle Valley (Davis et.al., 1992b) (Figure 1.1). Middle Valley, which is roughly 15-20 km wide and 75 km long, is a sediment-filled rift that until recently was the primary axis of seafloor spreading. Less than 200,000 years ago and probably within the last 10,000 to 15,000 years, spreading jumped from Middle Valley to West Valley, located 15 km to the west (Davis and Villinger, 1992). The geologic structure at Middle Valley is characterized by sets of normal faults, in general striking parallel to the valley axis, which cause the valley to be divided into fault bounded blocks (Figure 1.2). Offset rates may have been different at individual faults, causing the blocks to tilt and leading to varying sediment thicknesses and even to the exposure of basement in some areas. The block of oceanic crust, which was the focus of investigation of Leg 139 is bounded by two west-facing normal faults. Site 855 is on the eastern side of this block and Site 858 is near the western fault. Sites 857 and 856 are 5 located in the interior of the block (Figure 1.1). Both faults may have been active in the late Pleistocene and Holocene and subsidence and tilting of the block have led to the deposition of a 5 km wide wedge-shaped sediment cover which thickens to the east. The age of the fault-bounded block of oceanic crust is estimated to be between 200,000 -400,000 years (Davis and Villinger, 1992). 1.2.2.2 Sedimentology The sediment cover at Middle Valley has important implications in that it thermally and hydrologically seals the oceanic crust. Sediments are mostly terrigeneous, derived from the western margin of the North American continent. The sediment sequence is locally more than 2000 m thick; sediment thickness generally increases toward the center of the valley and toward the north (Buatier et al. 1994). Sedimentation took place mainly during the Pleistocene in the form of turbidites, which were supplied from the continental margin, primarily from Queen Charlotte Sound to the north (Davis and Villinger, 1992) (Figure 1.1). No barriers were blocking the sediment supply and sedimentation rates were sufficiently high to keep the valley full. Subsidence and the production of the current relief in the valley must have occurred in the most recent interglacial period (that is, during the last 10000 years), (Davis and Villinger, 1992). In the late Pleistocene and Holocene, continental sediment supply decreased and recent sedimentation rates are quite low (Buatier et al. 1994). The sedimentary sequence is characterized by units of turbidite layers of variable thickness and biogenic pelagic 6 horizons of Pleistocene age, overlain by hemipelagic sediments with few turbiditic horizons but abundant biogenic components of Holocene age. Sediments recovered during Leg 139 are divided into units and subunits according to the amount of turbidites and their degree of alteration (Buatier et al. 1994). Six main lithologic units can be distinguished (note that these units are based on lithologic character and do not necessarily represent a stratigraphic sequence): Unit I makes up the shallowest part of the sedimentary sequence and is composed of hemipelagic sediments of Holocene age with a minor content of turbidites. Biogenic components are common, but their amount decreases toward the hydrothermal discharge at Site 858. Unit II is a sequence of interbedded hemipelagic and turbiditic sediments which can be subdivided into four subunits according to their degree of alteration: II A: Weakly altered, abundant carbonate and/or dolomite nodules and concretions IIB: Partially indurated and brecciated; carbonate nodules, concretions, fracture fillings are common II C: More altered, fractured, brecciated IID: Strongly indurated and brecciated; silicified (near Site 858) 7 Unit III (only in Hole 858B) is a metalliferous, oxidized, smectite-rich mud interbedded with Unit I. Unit IV is the massive sulfide deposit recovered at sites 856 and 858. Unit V is the extrusive igneous basement drilled at Site 858 and Unit VI consists of mafic intrusions. 1.2.2.3 Petrology Whereas at unsedimented ridges the oceanic crust is made of volcanic units, the upper part of sedimented ridges is characterized by a sill-sediment complex which formed as volcanic sills intruded the sedimentary units. Both N-MORBS (normal mid-ocean ridge basalt) and T-MORBS (transitional mid-ocean ridge basalt) were found at the Leg 139/169 drilling sites. At Site 855, N-MORBS were identified which represent the oldest magmas and have the chemical and petrological characteristics of ridges with slow to medium spreading rates (Stakes and Franklin, 1994). T-MORBS were found at Site 858 indicating a different magma chamber or a different degree of melting. Intrusion of magmas of both MORB-types and mixing of both mantle sources or alternatively variable melting of a single source occurred at Site 857. The sills at Site 856 are thought to be the youngest volcanic rocks, which may have originated at the time when spreading at Middle Valley ceased. Therefore they postdate a magma chamber which lead to a 8 hydrothermal system and the formation of massive sulfides. The sills at this site consist of highly primitive lavas. 1.2.3 Geophysics 1.2.3.1 Seismic Profiles and Logs The sediment-sill structure of the upper oceanic crust, overlain by a cover of sediments, is recognizable in seismic profiles. Seismic profiles from Site 857, for example, show the occurrence of the first sill as a strong reflector. The thickness of the sediments is about 500m (Figure 1.2). This is confirmed by results from drilling which showed that roughly 465 m of sediment were penetrated before the first sill was encountered (Langseth and Becker, 1994). At the deepest point of penetration, at a depth of 936 m, the sediment-free igneous basement still was not reached. By means of gamma logs, at least 26 individual basaltic sills with thicknesses varying from 1 m to 25 m could be identified. (Langseth and Becker, 1994). The sills combined make up 41% of the total thickness of the lower part of hole 857D (440 m - 920 m). Seismic profiles and resistivity logs indicate that sediments in the sediment-sill sequence are compacted and indurated due to heating and loading from the intrusives. Extensive fracturing of the sediment-sill complex is seen on Formation Microscanner Scanner (FMS) images of borehole walls. The fractures are wide (up to 1 cm) and although mineralized contain visible void space (Davis and Fisher, 1994), which may suggest a relatively high permeability for this 9 sequence (Langseth and Becker, 1994). Pervasive fracturing of sedimentary and igneous rocks was also imaged at the top of the igneous units in hole 858F. Many fractures have a subvertical orientation and are aligned with the trend of the spreading center and bounding faults. Some fractures, however, may be drilling-induced thermal cracks, which formed due to thermal stresses created by circulating cool water into formations where temperatures are greater than 250°C. At the other sites, geophysical measurements reveal the basement to be relatively shallow. At Site 858, this is due to a small dome-like structure located below the vent-field (Langseth and Becker, 1994) and adjacent to a buried basement fault (Figure 1.2). This structure is, unlike the sediment-sill sequence at Site 857, true igneous basement entirely composed of extrusive rocks from lava flows. Langseth et.al.(1994) interpret this structure as a small seamount that stood as a topographic high above the seafloor before it was buried by turbidites. This edifice serves as conduit for the upwelling plume of hydrothermal fluids at this site. Site 856 is located on and around Bent Hi l l , a roughly circular hill which is about 60m high and about 500m in diameter (Stakes and Franklin, 1994) and is composed of sediments and sulfide talus (Tivey, 1994). Uplift of Bent Hi l l is probably associated with a laccolithic intrusion which was found at a depth of 112-120 mbsf (Davis and Fisher, 1994, Rohr and Schmidt, 1994). At Site 855 the depth to the uppermost sill ranges between 50 -100 m. Between 200 - 300 m a group of sills was imaged, which is believed to be the top of the actual sediment-sill sequence at this location (Rohr and Schmidt, 1994). The character of sediments and basement is different on each side of the fault at Site 855, indicating that this fault has been active for some time. The throw of the fault was determined to be 115 m, which exceeds the local 10 sediment thickness (about 100 m) and therefore suggests that igneous basement may be exposed (Davis and Fisher, 1994). Sediments adjacent to the fault are disturbed possibly due to the faulting process, hydrothermal alteration, or debris flows from the fault face (Rohr and Schmidt, 1994). 1.2.3.2 Heat Flow Distribution and Temperature Measurements Extensive geothermal exploration of Middle Valley has produced a comprehensive map of seafloor heat flow. Measured values range from as low as 0.15 W/m 2 to as high as nearly 25 W/m 2 . Values show both large regional and local variations (Figure 1.3). In general, heat flow is inversely correlated with the sediment thickness, which is likely due to their thermal and hydrological sealing of a near isothermal, permeable basement (Davis and Villinger, 1992). In the eastern section of the valley a number of high-amplitude, localized heatflow anomalies occur. They exceed heat flow values of 1 W/m 2 and typically extend several hundred meters to kilometers (Davis and Villinger, 1992). These heat flow highs are associated with vent sites at which high temperature fluid (= 270°C) discharges into the seawater. Heat flow anomalies were commonly found near the normal faults that bound the valley, which suggests that the faults or basement exposed at these faults act as conduits for both discharge and recharge of hydrothermal fluids (Figure 1.3). Due to technical difficulties on Leg 139, reliable temperature measurements were restricted to the upper 80 m of the sediment sequence (Davis and Wang, 1994) 11 (Figure 1.4). Formation temperatures at depth had to be estimated from shallow temperature measurements, heat flow- measurements, and the physical properties of the sediments. If the thermal conductivity of the sediments is known or can be estimated, heat flow measurements on the seafloor allow the extrapolation of temperatures at the interface of the sediment cover and the upper basement. This approximation however holds only under the assumption that vertical heat flow through the sediments is purely conductive. This model may be too simple for conditions at Middle Valley as slow advective heat flow does occur locally which is evident by anomalous pore-fluid composition measured on fluid samples from gravity cores and by heat flow variations with depth (Villinger et al. 1994). Nevertheless, as described in more detail below, models were developed which are able to estimate thermal conductivity profiles of the sediments. At Site 857, heat flow is 0.7 - 0.8 W/m 2 (Figure 1.3) and temperatures calculated for the top of the hydrothermal basement, which was thought to be a permeable basement 'reservoir' where hydrothermal circulation was inferred to occur beneath a sediment seal (below 470 mbsf) were an estimated 280°C (Figure 1.4) (Davis and Fisher, 1994). This temperature value was also found to be the maximum temperature of the vent fluids at Site 858 roughly 1.6 km to the north. (Davis and Wang, 1994). This coincidence is in agreement with the hypothesis that the sediment-sill complex is a permeable hydrothermal reservoir hydrologically sealed by a thick sequence of sediments, through which heat transport is mainly conductive. Average heat flow at the vent field at Site 858 is about 12 W/m 2 (Figure 1.3), temperatures of vent fluids range from 255 to 276 °C (Davis and Fisher, 1994). Diffuse or channelized upflow of hydrothermal fluids through permeable sections in the sediments underneath the vent field must be fast enough to 12 maintain these high venting temperatures and the high temperatures measured at shallow depths in the sediments (= 200°C at 20 mbsf; Langseth and Becker, 1994). Focussed upflow of fluid is indicated by high lateral gradients in heat flow around the vent field which decrease systematically with distance from the field, reaching background levels at a distance of a few hundred meters (Buatier et al. 1994). This observation is consistent with a model of conductive heat transport surrounding a hot, isothermal upflow zone. Anomalously high heat flow was also found at Bent Hi l l (>1 W/m ) (Figure 1.3). The maximum heat flow, however, is not associated with the hill but is located over an active vent field with fluids venting at temperatures of up to 265°C, some 300 m south of Bent Hi l l (Davis and Fisher, 1994). The massive sulfide deposit is not associated with the current hydrothermal activity and the uplift of Bent Hi l l . Fluid temperatures of >350°C are required to form such a deposit and it has been suggested by Mottl et al. (1994) that the formation of the deposit occurred during high temperature venting in the Pleistocene. At Site 855 where fluid is thought to recharge the hydrothermal basement, heat flow near the fault is on the order of 350mW/m2 (Figure 1.3) and temperatures at the bottom of the sediments were determined to be about 33°C (Davis and Fisher, 1994). At the distances from the fault of a few hundred meters, heat flow is about 580 mW/m and is remarkable constant along most of the fault where measurements were made. The extent of the heat flow anomaly associated with fluid recharge of less than a few hundred meters can be used as a constraint on the inflow rate along the fault zone (Davis and Fisher, 1994). The authors conclude that the Darcian velocity for water flowing into the 13 basement may be as high as 0.5 m/yr. With the aquifer dimensions used the volumetric flux would total 25 m3/yr per meter length of fault. These inflow rates are too low (roughly by a factor of 30) to account for the observed extent of alteration in the hydrothermal basement and the fluxes observed at hydrothermal vents. Therefore recharge of fluids into the basement has to occur elsewhere, possibly via diffuse flow through the sediment sequence (Davis and Fisher, 1994). A few high heat flow anomalies were encountered along the fault at Site 855 (Figure 1.3), giving evidence of discrete areas along the fault where fluid discharge may occur. 1.2.4 Hydrology 1.2.4.1 Porosity and Permeabilities Permeability is probably the most important parameter in the assessment of fluid flow patterns and velocities. However, permeability is difficult to quantify ' in situ' because of its scale dependence. In particular in fractured or faulted rocks the bulk permeability may be controlled by a few, widely spaced fracture zones which require the scale of measurement to be large enough to assign one average, representative value of bulk permeability to the formation. During Leg 139, permeability measurements were performed at Sites 857 and 858 with the help of drill string packers. Drill string packers are used to hydraulically isolate sections of the drilling hole to perform slug and injections tests for measurements of permeabilities. 14 Despite the different nature of hydrothermal basement, either true igneous basement at Site 858 or the sediment/sill sequence at Site 857, permeabilities measured at both locations are very similar and average about lx lO\" 1 4 m 2 (Figure 1.4). This similarity may indicate that as a result of hydrothermal processes the sediment-sill sequence is indurated enough such that its hydrologic properties are similar to those of a fractured igneous rock. At Site 857 a relatively thin, highly transmissive zone was encountered at a depth of about 600 mbsf, interpreted to be a fault zone (Becker et al. 1994). In this zone permeabilies are as high as lx lO\" 1 0 m 2 . Depending on the spatial extent of the fault zone it may have an important effect on the flow pattern by strongly focusing hydrothermal fluxes. This zone possibly reflects the general flow regime in the upper oceanic crust which is that hydrothermal convection is controlled by irregular zones of high transmissivity. Results of laboratory measurements by Fisher et al. (1994) on sediment samples yielded average permeability values on the order of lx lO\" 1 6 m 2 , which is two orders of magnitude less than in-situ permeabilities estimated for the hydrothermal basement. Permeabilities in the sediments typically decrease with depth due to increased compaction, consolidation and thermal alteration. In particular the clay fraction of the turbidites may contribute to consolidation and permeability reduction at depth (Fisher et al. 1994). Calculated and measured permeabilites in the sediments at Site 857, for example, decrease from lx lO\" 1 5 m 2 at the top of the sequence to lx lO\" 1 6 m 2 at 150 mbsf. At sites 856 and 858 the reduction in permeability is more pronounced, reaching values on the order of lx lO\" 1 7 m 2 at 80 mbsf (e.g. hole 856B), suggesting that permeability 15 reduction is also associated with the proximity to hydrothermal vents and alteration of the sediments (Fisher et al. 1994). Leybourne and Goodfellow (1994) propose a model in which alteration processes within the core of hydrothermal upflow of vent fluids, such as recrystallization of primary minerals, in particular quartz and Mg-chlorite, and crystallization of secondary minerals in voids cause cementation and induration of the sediments, creating a zone of increased, fracture-controlled permeability. The front of this high permeability zone propagates upward and expands laterally into the sediments. Lateral flow into and within the sediments is supported by the anisotropy caused by grading of grain sizes within individual, continuous turbidite layers. The clay-size fraction inhibits vertical flow and at the same time coarser grained, lower parts of the turbidite layer improve lateral hydraulic communication (Fisher et al. 1994). Evidence for larger scale lateral fluid flow through the sediments was found at Site 857, but its significance in terms of fluid flow geometries is difficult to assess due to the large horizontal scale of the valley (Fisher et al. 1994). 1.2.4.2 Pressure Measurements One goal of Leg 139 was the installation of CORKs (circulation obviation retrofit kit) at sites 857 and 858. CORKs are instruments which are able to seal boreholes to prevent the exchange of ocean and formation water and therefore minimize any drilling-induced thermal and chemical disturbances in order to monitor long term (up to 2 - 3 years) in-situ temperature and pressure conditions (Davis and Becker, 1994). It also 16 allows the sampling of fluids long after holes are drilled under close to undisturbed formation conditions. Preliminary results of 1 year of C O R K operation were reported by Davis and Becker, (1994) and are briefly summarized below. Formation temperatures measured with CORKs were compromised due to technical difficulties. Data are noisy at times and absolute temperatures are probably too high. Pressure measurements however are thought to be reliable at Site 857. Results show that during the first 3 weeks of data recording following the C O R K installation in hole 857D, pressure and temperatures recovered toward undisturbed conditions. Pressure recordings from Site 857 indicate that undisturbed formation pressures are very close to hydrostatic which is most likely due to its location away from hydrothermal upflow or downflow. It could also indicate that average permeabilities in the hydrothermal basement are high (> 5x10\"13 m2) to cause convection to be vigorous enough to reduce pressure differences along the sediment/basement interface to a nearly uniform value. This hypothesis however remains untested as long as the geometry of hydrothermal convection at Site 857 is unknown (Davis and Becker, 1994). Basement fluids at hole 858G are overpressured. An accurate estimate of overpressure is difficult because of the potentially significant thermal and pressure perturbations due to leakage from neighboring hole 858F. A lower limit of formation pressure was estimated to be 180kPa. (Davis and Becker, 1994) but the value may be as high as 450 kPa with respect to the local hydrostat. Evidence for overpressured basement fluids is given by hydrothermal venting at the 'Dead Dog' vent field and indications of lateral flow in the sediments at Site 857 at a level that coincides with the depth of the top 17 of the basement edifice at Site 858. At this depth anomalous porewater compositions were found in the sediments at Site 857, in composition similar to the vent fluids at Site 858, suggesting a lateral flow component from the discharge zone at Site 858 into the sediment sequence at Site 857, 1.6 km away (Davis and Becker, 1994). Davis and Becker (1994) speculate that i f one assumes a hydrologic link between Site 858 and Site 857 via a highly permeable basement, at a depth below 470m, and then shifts the local hydrostat at Site 858 such that it intersects the 857 hydrostat at 470m, hydrostatic fluid pressure exceeds the lithostatic pressure at about 70mbsf at Site 858. This means that hydrofracturing would be possible anywhere above this depth. Evidence of brecciation in core samples from depths above 50mbsf was indeed commonly found. (Lithologic Subunit IIB; Shipboard Scientific Party, 1992b). 1.2.5 Geochemistry 1.2.5.1 Rock Geochemistry 1.2.5.1.1 The Sediments - Mineralogy and Geochemistry The minerals in the sediments can be divided according to their primary, detrital or secondary, and authigenic origin. The occurrence of authigenic minerals or mineral assemblages allows the definition of alteration zones which reflect the degree and extent of hydrothermal activity. 18 Mineral composition of unaltered sediments reflects their terrigeneous origin. They are predominantly composed of quartz (up to 40%), plagioclase and minor amounts of amphibole and mica. The clay-size fraction consists mainly of smectite, chlorite, illite, and irregular, mixed-layer clay (Leybourne and Goodfellow, 1994). The most important secondary hydrothermal minerals include quartz, Mg-chlorite, a corrensite-like mineral, smectite, epidote, wairakite,analcime, anhydrite and gypsum, pyrite, and calcite. These minerals typically occur in interstitial spaces (cracks, pores etc.) or replace primary minerals (Kurnosov et al. 1994). Leybourne and Goodfellow (1994) distinguish several alteration zones for each of the Sites 856, 857,and 858, and number them according to the degree of alteration with the highest zone number being the least altered and Zone I showing the strongest alteration. At Sites 856 and 858, which both represent hydrothermal upflow zones, these alteration zones occur at different depth intervals or are distributed with different lateral extent about a hydrothermal upflow zone and reflect the nature and conditions of reactions between porewater, minerals and hydrothermal fluids at each of the sites. Due to the impact of the high temperature conditions during the formation of the now inactive massive sulfide deposit on the sediment alteration at Site 856, the nature and pattern of alteration at this site does not correspond to the current conditions and is therefore not immediately relevant to the purpose of this study. For this reason we omit the description of Site 856 and focus on Sites 858 and 857 only. The following descriptions are taken from Leybourne and Goodfellow (1994). 19 At Site 858, six different alteration zones can be identified. They are distributed laterally and vertically about the center of recent hydrothermal fluid discharge (Figure 1.5). The zonation pattern is consistent with the concept of upflow of hydrothermal fluids and outward migration of these fluids and their mixing with entrained seawater. Fluid temperatures near the core of active hydrothermal upflow are moderate and range between 260 and 300°C, which is too low for the formation of massive sulfides. In the outermost and/or uppermost alteration zone temperatures are generally <100°C. Zone VI (Saponite-Pyrite) This zone covers the depth interval from 24 to 38 mbsf in Hole 858B and is characterized by the alteration of hemipelagic clay to saponite (Mg-smectite) and pyrite. The substantial amount of saponite in the sediments results in large MgO contents of bulk sediments (17.9-34.2 wt%). Formation temperatures of saponite measured from oxygen isotopes are between 119 and 134°C. Secondary quartz and anhydrite occur as vein fillings. Below 24 mbsf the sediments are brecciated. Zone V (Calcite(+/-Dolomite)-Illite-Pyrite) This zone occurs between 15 and 80 mbsf in Hole 858A, and above about 20 mbsf in Hole 858 C and D, which are both closer to the site of active venting. Sediments are weakly indurated and locally brecciated. Calcite occurs as.cement, concretions and veins, and even as massive breccia in Hole 858C (at 15 mbsf). Quartz and clay minerals are variably recrystallized. Detrital plagioclase is altered to albite at depths >25 mbsf, k-feldspar shows an increasing degree of chloritization and sericitization with depth. Pyrite 20 occurs disseminated throughout the clay layers, interstitially within the silt -and sandlayers, and as a replacement of worm-borrows. Zone IV (Anhydrite-Illite-Pyrite) This zone extends from 70 to 175 mbsf in Hole 858A, between 11.5 to 24 mbsf in Hole 858B and between 17 and 68.5 mbsf in Hole 858C. Sediments are moderately indurated. Anhydrite is present throughout this zone and occurs as disseminated crystals, crystal aggregates, veins, vugs or concretions and can make up to 1 -2 vol% (Leybourne and Goodfellow, 1994). Authigenic illite and pyrite are omnipresent. Estimated alteration temperatures for zone IV and zone V range between 112 to 247°C. (Peter et al. 1994) Zone III (Albite-Chlorite-Pyrite) This zone occurs between 175 and 320 mbsf in Hole 858A and 33 and 92 mbsf in Hole 858C. It contains indurated and brecciated sediments with an alteration mineralogy that is similar to the previous two zones (overgrowth and recrystallization of quartz, recrystallization of chlorite, albitization of plagioclase, chloritization and sericitization of k-feldspar). In contrast to the previous two zones is the general absence of anhydrite and carbonate. Alteration temperatures measured on fluid inclusions yielded a range from 230 to 280°C for this zone. Zone II (Quartz-Chlorite-Epidote-Pyrite) This zone is only present in Hole 858F between 120 and 143 mbsf and 173 and 249 mbsf. The sediment is lithified, brecciated, fractured, and veined. Epidote is abundant as granular masses or disseminated and makes up to 20 vol% (Leybourne and 21 Goodfellow, 1994). The chlorite in this zone is relatively Mg-rich, indicating the mixing of seawater and hydrothermal fluids. The increase of Mg in chlorite toward the conduit of fluid upflow points out a higher fluid/rock ratio due to enhanced, fracture-controlled permeability. Zone I (Quartz-Wairakite-Epidote-Pyrite) Zone I represents the center of the upflow zone. The sediments are fractured, veined, and indurated. Quartz and wairakite occur in veins and molds of what possibly used to be anhydrite. Anhydrite is found in veins and as concretions which suggests that seawater entered the hydrothermal upflow zone, possibly due to fluctuating hydrothermal activity. Wairakite locally replaces the matrix and can form up to 35% of the sediment. Measurement of fluid inclusions in wairiakite yielded temperatures of formation between 250°C and 300°C. Plagioclase is present in highly variable compositions, generally its Ca content increases (and Na decreases) with temperature (Seyfried et al.1991, Leybourne and Goodfellow, 1994). Similarly, the Na/Ca ratio in wairakite is lower than in the wairakite-bearing zone at Site 857. The large amount of Ca-bearing alteration minerals and the loss of k-bearing phases (k-feldpsar, micas, illite) lead to low K2O and high CaO contents of the sediments. The MgO content of the sediments, in particular in chlorite, is also relatively low. The mineralogy and bulk sediment composition is consistent with the low (zero) Mg and high Ca concentrations in the endmember hydrothermal fluids passing through this zone. At Site 857, four alteration zones could be distinguished (Figure 1.6). Generally there is an increase in hydrothermal alteration downhole and there is a similarity between 22 mineralogy, bulk chemical composition and fluid inclusions between the alteration assemblages at this site and Site 858, lending support to the idea of the generation of vent fluids discharging at Site 858 in a reaction zone at Site 857 and the hydrological connection of these sites via a high-permeability hydrothermal basement. Zone IV: (Calcite-Clay-Pyrite) This zone comprises the upper section of the sediment sequence (70 to 300 mbsf). Clay minerals are increasingly recrystallized with depth, calcite is common as cement or concretions. Above 70 mbsf calcite is present only as a minor or trace alteration product. Zone III: (Albite-Chlorite-Pyrite) Zone III occurs between 300 and 450 mbsf. The most important alteration processes are albitization of plagioclase, chloritization and sericitization of k-feldspar, and an increasing degree of quartz recrystallization below 330 mbsf. Zone II: (Quartz-Chlorite-Epidote-Pyrite+/-Sphene) In this zone detrital plagioclase is altered to albite and detrital k-feldspar to chlorite which is the dominant clay mineral in this zones. Quartz shows strong evidence of recrystallization (ragged, diffuse grain boundaries). Authigenic epidote is locally abundant (up to 30 vol%). Micas are almost completely destroyed, leading to a low K2O content of the sediments. Sulfides occur in both veins or as disseminated minerals. Zone I: (Quartz-Wairakite-Epidote) This zone is situated within the sediment-sill complex at Site 857 at a depth of 628-736 mbsf and is similar to zone I at Site 858 in terms of mineralogy, fluid inclusion 23 temperatures, Sr-isotope ratios and composition of fluid inclusions, indicating equilibration with similar hydrothermal fluids (Goodfellow and Peter, 1994). Wairakite is common in veins, interstitially or in vugs within the sand and silt fraction of the sediments. In veins it is commonly associated with quartz and epidote. Temperatures measured on fluid inclusions in wairakite yielded a range from 242 to 350°C between 450 and 695mbsf in Hole 857D, without showing any systematic downhole variation (Peter et al. 1994). This, the stratabound nature of this zone, and its pervasive alteration indicates that this zone was essentially isothermal during much of its history. This is in agreement with the concept of an isothermal reaction zone, which supplies fluids to the vent areas such as Site 858. In this zone the only primary mineral present is quartz (Goodfellow and Peter, 1994). Downhole profiles of major element bulk compositions for the sediments reflect the main alteration processes and in combination with porewater profiles help quantify the mass transfer between the solid and fluid phases in each zone. The character of alteration and thus the trends in the compositional profiles from Site 857 and 858 are similar, but profiles from Site 857 show the compositional changes most clearly due to a relatively low, uniform geothermal gradient with a maximum temperature of about 300°C at 450 mbsf. (Baker et al. 1994) and will be briefly discussed here and in greater detail in the chapter on porewater profiles below. In Zone I and II, Si02, CaO and AI2O3 have been added in substantial amounts, Na20, K2O and CO2T on the other hand are depleted. Si02 typically ranges between 50-70 wt% but may reach values greater than 70 wt% in zones I and II due to the addition of 24 hydrothermal quartz. The source of Ca may be dissolved biogenic carbonate, destruction of plagioclase, downwelling seawater, or dissolution of anhydrite, precipitated during an earlier stage (Goodfellow and Peter, 1994). Goodfellow and Peter, (1994) come to the conclusion that most of the elements depleted from this zone are enriched in the hydrothermal precipitates at Site 858, which again suggests a hydrologic link between the reaction zone at Site 857 and the site of active venting 858. In Zone III the replacement of k-feldspar, plagioclase, mica and illite with an albite-chlorite-pyrite assemblage leads to the addition of Na, Mg, and Fe and the depletion of K , Rb, Ca, Sr, and Ba. The addition of these elements and an increase in 8 7 Sr/ 8 6 Sr ratios from zone I to zone III suggest upward advection of fluids from the reaction zone (Goodfellow and Peter, 1994). CaO decreases with depth in zone IV due to the dissolution of biogenic carbonate. In zones IV and III (60 and 380 mbsf) it increases in an irregular fashion due to the formation of carbonate concretions and in zones I and II due to the occurrence of wairakite and epidote (Figure 1.6). 1.2.5.1.2 The Basalts - Mineralogy and Geochemistry Igneous rocks were recovered from all of the drilling sites in Middle Valley. From Site 855 a MORB-type basalt, which probably represents basement rock that extruded during the early rifting stages of Middle Valley (Stakes and Franklin, 1994), from Site 856 late mafic sills made of primitive picritic basalts, from Site 858 at 250 mbsf 25 moderately evolved basalt, and from Site 857 a series of highly altered sills of variable chemistry. Plagioclase, clinopyroxene, and olivine are the most abundant minerals in the unaltered basalt, differences between basalts from each site exist in the amount and composition of each of these main minerals, the type and amount of accessory minerals, and the degree of alteration. At Site 855 the average composition of olivine is Fog3 (F077. 86), the composition of plagioclase phenocrysts varies from a calcic core Angs to more sodic rims A1168-74, plagioclase in the groundmass averages around An7o-75. Alteration is only minor in these basalts (Stakes and Franklin, 1994). In unaltered basalts at Site 856 olivine is more magnesium-rich (F088-90) and makes up 2% - 8% of the rock. Plagioclase occurs as both phenocrysts and matrix minerals, compositions range from A n 7 4 . 7 9 to An64-7o, with the latter being found in the groundmass and in the rims of phenocrysts. Clinopyroxene exists only as groundmass phase. The most important accessory minerals are sphene, titanomagnetite and sulfides. The sills at Site 857 show a characteristic variability in texture, crystal size, and mineral composition due to different cooling rates from the sill margins to the interiors. Olivine as a primary mineral was not found in any of the sills. Pseudomorphs of either olivine or clinopyroxene were observed, however, composed of smetite/chlorite or chlorite. Plagioclase phenocrysts have a core composition of An6o-73 and become more sodic toward the rims (An5o-6o)- Calcic megacrysts (Anso-89) were also found. Clinopyroxene occurs as both phenocrysts and as groundmass phase. Accessory minerals are sphene, ilmenite (up to 3%), and sulfides. Mineralogy and geochemistry of the sills 26 have clearly been modified by alteration. In particular the fine-grained, chilled margins of the sills are metasomatized or exhibit hydrothermal recrystallization. Alteration to Mg-rich chlorite is pervasive in the sill margins, leading to anomalously high Mg, A l and Ti contents compared to the unaltered rock. Microlites and phenocrysts in the margins are replaced by chlorite, epidote and actinolite. Vesicles are abundant in the sill margins and are typically filled with quartz, epidote, prehnite, smectide, or sulfides. Veins, cutting through the altered margins into the fresher rock, are filled with secondary minerals such as Fe-chlorite/smectite, sulfides, quartz, zeolites (wairakite), and epidote. Toward the coarser grained interior of the sills the extent of chloritization is variable and may include over 50% of the rock. Plagioclase is commonly replaced by epidote, chlorite, and prehnite. Albitic rims on plagioclase (An<2o) are present but relatively rare. Chlorite, epidote, as well as titanite are the dominant vein minerals. The replacement of primary minerals with Mg-rich alteration products lead to high MgO contents of bulk rock compositions. The correlation of MgO with H2O in bulk rock compositions points to seawater as the source of Mg. The abundance of Ca-rich vein minerals suggests that uptake of Mg was accompanied by a release of Ca . At Site 858 all the recovered igneous basement rocks were interpreted as basalt flows. These rocks are rather uniform in composition and have numerous quenched contacts. Plagioclase, if not replaced by epidote and albitic plagioclase, is zoned with a core composition of An70-75 and a rim composition of An5g_69. Clinopyroxene and olivine are usually replaced by chlorite and smectite. Common accessory minerals are spinel, ilmentite, magnetite and sulfides (1-5% - less than at Site 857). Alteration appears to be less pervasive than at Site 857 and the formation of secondary phases is typically 27 restricted to vein and vesicle fillings rather than bulk replacement. Alteration assemblages vary from low-temperature phases (e.g. calcite, smectite, and celadonite) at shallower depths to higher temperature phases such as epidote, chlorite/smectite, talc in the deeper parts of the borehole. 1.2.5.2 Fluid Geochemistry 1.2.5.2.1 Pore Fluids Pore-water compositions differ from bottom seawater values as a result of bacterial degradation of organic matter, reaction with detrital silicates and basaltic sills, recrystallization of biogenic carbonates and silicates, hydration of silicates, transport processes, and increasing temperature (Davis et al. 1992b). At Site 855 basement fluid compositions are similar to seawater composition, in particular near the fault which constitutes the eastern boundary of Middle Valley (Figure 1.1) (Davis et al. 1992b). This suggests that bottom seawater may travel along the fault plane and outward into the basement rocks. The major ion concentrations (e.g. Ca, Mg) change only slightly, most likely due to low temperatures and the low reactivity of the system. At Site 858, pore fluid compositions vary with distance from the vent area and steepness of the temperature gradient. Anomalies in the CI\" concentration profiles from Site 858 suggest lateral flow from the hydrothermal upflow zone outward into the sediments Davis et al. (1992b). Fluctuations in the Cl-profiles may suggest that fluids of 28 different origin (i.e. seawater or hydrothermal fluids) are flowing at different depth intervals and local mixing of these different fluids may occur. This in turn may lead to local chemical reactions such as anhydrite precipitation. At greater distance from the vent field, the chlorinity profile shows a linear increase with depth (to 120 mbsf) suggesting mainly diffusive transport. Below that depth chlorinity decreases to seawater levels and remains at this level to a depth of 240 m. Below 240 m chlorinity increases again. The profile again strongly suggests that lateral flow through the sediments is occurring. These transport processes in combination with mineral alteration reactions are the reason for fluctuation in all major ion concentration profiles. Calcium generally increases with depth, sulfate decreases below 190 mbsf due to anhydrite precipitation. Anhydrite was found between 69 and 333 mbsf. The reason for the increase in calcium in the porewater despite the precipitation of anhydrite is most likely a large amount of calcium added as a result of silicate alteration. Magnesium decreases with depth and then increases slightly. Dissolved silica shows variations above 100 mbsf, which may be controlled by the equilibrium with and availability of biogenic silica. Below that depth the silica increases and correlates with the increasing temperature. Quartz is common as cement and vein filling at greater depths (near bottom of Hole 858A). The potassium profile shows K-leaching at greater depths and higher temperatures and K-uptake into clay minerals in the cooler, upper parts of the sediment sequence. Sodium shows the opposite trend, probably due to albitization and analcime formation at depth. 29 The most complete porewater composition profile was obtained from Site 857. A compilation of fluid compositions versus depth is shown in Figure 1.7. Calcium concentrations remain at bottom seawater levels throughout the upper 50 m, then increases to a depth of 285 mbsf (Figure 1.7). The relatively uniform calcium concentration above 50 mbsf reflects the balance between alteration of Ca-bearing silicates and formation of secondary carbonates. The increase between 50 and 285 mbsf is due to the continued release of Ca by alteration of detrital silicates, exceeding the amounts of Ca removed by precipitation of calcium carbonate and anhydrite. Between 285 and 400 mbsf the calcium concentration is about the same as the endmember concentration at Site 858 (Davis et al. 1992b). Sulfate is consumed at shallow depths due to bacterial degradation of organic matter. Below this depth, between 70 -100 mbsf, sulfate concentrations show a small maximum possibly as a result of redissolution of anhydrite or oxidation of sulfide minerals. At greater depths downhole concentrations uniformly decrease which may be the result of anhydrite precipitation, biogenic degradation of organic matter, or both. Magnesium concentrations are roughly equal to those of bottom seawater in the upper 50 mbsf and then decrease to a minimum at a depth of 193 mbsf. The decrease indicates the alteration of detrital minerals and the formation of Mg-rich alteration phases. Below 193 mbsf Mg slightly but uniformly increases, which may reflect the equilibrium with the different mineral assemblage as temperatures increase. The potassium profile shows a minimum at 100 mbsf and a maximum at 400 mbsf. The decrease in the upper levels may be attributed to the formation of secondary phases such as chlorite at low temperatures (80-120°C), which also consumes much of the Mg, and the concomitant release of Ca. The increase below the minimum is a 30 consequence of water/rock interaction at higher temperature (+/- 150°C). Sodium concentrations are roughly equal to bottom seawater values in the upper 175 mbsf, then decrease to a minimum at 300 mbsf, which is a result of albitization of detrital plagioclase and the formation of zeolites. Mass balance calculations performed by Davis et al. (1992b) revealed that in the interval from 300-400 mbsf the anomalously large amount of Na in the sediments requires a long-lived supply of Na from fluids and albitization, making it necessary that lateral flow must have occurred or is still occurring. Consistent with lateral flow in certain depth intervals are fluctuation in the CI\" concentration profile at this site. The shallow peak in the dissolved silica concentration at 40mbsf is probably a result of dissolution of biogenic silica. Concentrations decrease below this maximum which may indicate the equilibrium with a different silica phase. At depths below 80mbsf concentrations are roughly uniform, which is not consistent with what to expect with increasing temperature (Figure 1.7). In summary, there is strong evidence of lateral flow in the sediment section at Site 857. Pore water in the depth interval from 300 to 400 mbsf is similar in its chlorinity, sodium, potassium and calcium concentrations to the vent fluids at Site 858, 1.6 km away. These fluids are likely to be the source for the vents and it is interesting to note that this depth interval is above the first sill, which may suggest that the depth to the permeable hydrothermal basement does not coincide with a change in lithology or a strong seismic reflection, i.e. the transition from sediments to the sediment/sill sequence. Evidence for the flow direction, i.e., from Site 857 to Site 858, can be drawn from the presence of magnesium and sulfate in the pore-water and their absence in the vent fluids. Sources of these ions at 200 to 300°C are highly unlikely (Davis et al. 1992b). Another 31 evidence is the degree of albitization in the sediments, which could not have been achieved with endmember hydrothermal fluids. However, it should be noted that as pointed out earlier, the overpressured basement at Site 858 may lead to lateral flow of high temperature fluids into the sediments, feeding and interacting with sediment pore fluids. Evidence for this hypothesis are porewater anomalies in the sediment sequence at Site 857 in an interval that coincides with the top of the basement edifice at Site 858 (150-200 mbsf). Therefore directions of lateral flow within the sediment section possibly change at different depth intervals. 1.2.5.2.2 Vent Fluids Vent fluid chemistry is in general agreement with the concept of a hydrothermal reaction zone below the sediment-basalt interface in which hot fluids react with basalts (as indicated by the basaltic character of the 8 7 / 8 6 S r in barite (0.7054-0.7060), gypsum (0.70655) and calcite (0.7062-0.7096) (Goodfellow, in press)) and to a lesser degree the vent fluids react with the sediments in the upflow zone. The important characteristics of the Middle Valley vent fluid are first of all the enrichment of calcium and depletion of magnesium and sodium. Based on isotopic data (Butterfield et al. 1994) inferred a basaltic source for the calcium. The destruction of Ca-bearing phases in the basalt (such as anorthite) is accompanied by the formation of Na or Mg-rich alteration phases. The main Mg-bearing phases are chlorite and smectite. Despite the Ca enrichment in the vent fluids, these fluids are undersaturated with respect to calcite. Silica is also strongly 32 enriched in the vent fluids. Quartz solubility calculations allowed Butterfield et al. (1994) to conclude that the vent fluids are in equilibrium with quartz, possibly within ~500 m of the Middle Valley seafloor. Sulfate is nearly completely removed in the vent fluids, probably as a consequence of anhydrite precipitation . Potassium is enriched in the vent fluids due to the leaching of K-bearing phases (micas, k-feldspar). 1.3 Conclusion The data collected during Leg 139 at Middle Valley constrain a range of the key hydrological parameters and help understand the geochemical processes which occur in this sedimented mid-ocean ridge setting. The data allow one to make assumptions about the vigor and geometry of hydrothermal convection in the upper oceanic crust and the nature of the important alteration processes in the basalts and the sediments. The sediments, which are locally up to 2000 m thick, consist of a sequence of Pleistocene turbidites and intercalated pelagic, biogenic horizons, overlain by Holocene, hemipelagic clays. The sediment cover acts as a hydrological seal and a thermal insulator to the underlying crust. Consequently, temperatures within the upper crust are generally much higher than those typical in non-sedimented ridge environments. At sedimented areas, the combination of heat flow and seismic reflection observations indicate an inverse correlation between heat flow and the sediment thickness (e.g. Davis and Villinger, 1992). This has led to the concept that hydrothermal convection in the permeable crust is vigorous enough to thermally homogenize the basement. 33 Estimated temperatures of basement fluids are of the order of 280°C, which is also the maximum temperature of the vent fluids discharging at Site 858. Field tests and laboratory experiments indicate that the permeability contrast between the low-permeability sediments and the permeable basement amounts to roughly two orders of magnitude ( lx lO\" 1 6 m 2 and lx lO\" 1 4 m 2 , respectively). The permeability of the basement is thought to be fracture controlled but very little is known about the frequency, extent, geometry, and orientation of these fractures. This means that the basement permeability which was measured over a relatively small volume of oceanic crust, is not necessarily representative for the basement on a formation scale. Indeed, the encounter of a fracture zone with a permeability of about 1x10\"10 m 2 within the igneous basement suggests that the bulk permeability of the basement could be orders of magnitude higher. As a consequence of the sediments overlying the permeable basement, the exchange of basement fluids and seawater is reduced and commonly restricted to permeable conduits through the sediments. Fluid recharge to the basement occurs along fault zones, basement outcrops (e.g. Site 855), or as slow advective flow through the sediments. Discharge of hydrothermal fluids through the sediments tends to be highly focussed in form of hydrothermal vents and is commonly related to fault zones or topographic basement highs. Because of the low permeability, the sediments are characterized by mainly conductive heat transport and diffusive solute transport. Geochemical reactions in the sediments are controlled by fluid and sediment compositions and the local geotherm. Thermal and geochemical anomalies exist in and around areas of fluid recharge and discharge. Lateral flow at different depth intervals, promoted by graded grainsizes in the 34 sedimentary units, causes local pore-water anomalies. The basalts of the upper oceanic crust form the permeable layer in which convective fluid flow and heat transfer occurs. The upper oceanic basement occurs as either 'true' extrusive igneous units (at Site 858), or as a series of mafic sills interlayered with sediments (at Site 857). The basalts show clear evidence of hydrothermal alteration and exhibit an alteration mineralogy typical of a greenschist metamorphosis. Compositions and isotope ratios of vent fluids indicate that the characteristics of these fluids are mainly derived by the interaction with the basalts and to a lesser degree with the sediments. Little is still known about the effect of vigorous fluid circulation on the geochemistry and the feedback of geochemical reactions on hydrological properties and flow conditions. The concept of a 'hydrothermal basement' or a 'reaction zone' in the sediment-sill sequence where hydrothermal fluids circulate and react with the basalt and the sediments, and fluid recharge into and discharge from the upper crust through the sediment cover via high-permeability pathways (e.g., fault-zones, basement highs) (Figure 1.2) seems to be in agreement with the available data. 35 Figure 1-1. Location map of Sites 855-858. The insert shows a map of the Juan de Fuca Ridge and the location of Middle Valley. (From: Davis et al. 1994) 36 Figure 1-2. Conceptual cartoon of hydrothermal circulation at Middle Valley, based on results of Leg 139. (From: Davis and Fisher, 1994) 37 38 Log p e r m e a b i l i t y (m*) Temperature (°C) :.: ._•„;;. „1t} 0 SO J.Q0 150 200 250 3 0 0 o. CJ O 1 0 0 2 0 0 3 0 0 4 0 0 5 0 0 6 0 0 7 0 0 8 0 0 9 0 0 1 0 0 0 ~- J-t\": 3 | J L _ _ J J_—J-Permeability data from: 1. Derived from porosity/depth curve in Fisher et al. (1994) 2. Model result from Bessler et al. (1994) 3. + 4. Downhole packer measurements (Becker et al. 1994) Temperature data from temperature probe data and conductivity profiles, (Villinger at a l , 1994; Davis and Wang, 1994). Note the isothermal basement below 470 mbsf. Figure 1-4. Permeability and temperature versus depth. The depth of the top of the first sill is shown as the horizontal dotted line. (From: Davis and Fisher, 1994) 39 w HOLE MOLE 858C I' — HOLE HOLE \\ / \\ l \\LJl-: ZONE 111 b • v 7 0 N F S n f ZONE 111 b A N D n b <*— v ' . • uttlOe mm— AntlYi Figure 1-5. Profile across Site 858 showing the distribution of alteration zones. A description of the zones is given in the text. (From: Leybourne and Goodfellow, 1994) 40 100 200 -300 400 I 5 0 0 600 1 700 -< 800 900 -M 1 0 0 0 o o o rXIUXUUUUUJUUlJ c g 5 2 < U A III c Ic He NS Calcite-Clay-Pyrite Albite-Chlorite-Pyrite Quartz-Chlorite-Epidote-Pyrite+/-Sphene Quartz- Wairakite-Epidote jure 1-6. Mineralogical profile at Site 857. (From: Leybourne and Goodfellow, 1994) 41 gure 1-7. Profile of fluid composition at Site 857. 42 2 The Evolution of a Transient Convective System 2.1 Introduction Legs 139 and 169 to Middle Valley, at the northern end of the Juan de Fuca Ridge, provided important insights into the flow conditions in the oceanic basement at a sedimented ridge crest. Despite the success of both legs, little detail is known about the geometry, intensity and life-span of the convective system. Other information such as from chemical data may be helpful in providing additional constraints on these issues. The composition of the fluid and the rock represent the integrated effects of the fluid's reactive history along its flowpath and the reactive history at a particular location, respectively. Therefore, to make effective use of chemical data, an understanding of the fluid's flowpath and the changing physical and chemical conditions at a location within the system is required. The intent of this study is to examine transient physical conditions between the onset of fluid motion and the steady state convective fluid flow and how these conditions may affect the composition of the fluid and the rock. The duration of the transient state, the thermal conditions, the vigor of fluid flow, and the rate at which the conditions change are primarily dependent on the magnitude and distribution of the permeability. To assess this dependency, the transient state of the system is calculated as a function of basement permeabilities while boundary conditions, sediment permeability, and all other parameters remain unchanged. The focus of this study is on recharge into and discharge out of the convective layer, physical water/rock ratios, the geometry of flowpaths, particle travel times, rates of temperature change, and fluid mixing during the 43 transient state. The possible implications of these aspects of the system on the chemical conditions in the hydrothermal system are emphasized. 2.1.1 Review of Permeability Estimates and Concepts of Crustal Hydrothermal Convection Numerous studies have addressed the magnitude and distribution of permeability in the upper igneous oceanic crust. An excellent summary of recent advances and insights can be found in Fisher (1998). In the following section a brief overview of constraints on the basement permeability and concepts of crustal hydrothermal convection is presented. The intent of this overview is to document which elements of the model developed here have been adopted from previous studies, which elements are new to this work, and to indicate the limitations of the model developed for this study. Crustal permeabilities have been measured directly in the laboratory on basalt samples and in-situ as formation permeability from packer measurements, or were obtained indirectly from thermal measurements, porosity measurements, or borehole imaging and core fracture analysis. Results from packer experiments yield a range of formation permeabilities from less than lx lO\" 1 7 m 2 (e.g. Becker, 1989, Costa Rica Rift) to about lx lO\" 1 3 m 2 (Larson et al. 1993, Pigafetta Basin). A thin interval in the sediment/sill 10 2 sequence at Middle Valley was measured to have a permeability as high as 1x10\" m . In general, the permeability of the oceanic crust appears to be correlated with the lithological and structural properties of the crust (e.g. pillows, faults). Relatively high permeabilities 44 (> l x l 0 \" 1 4 m 2 ) occur primarily in the upper few hundred meters, the extrusive part of the basaltic crust (Fisher, 1998). Permeability values determined on basalt samples range from lx lO\" 2 0 m 2 to lx lO\" 1 6 m 2 (Johnson, 1980a; Hamano, 1980; Karato, 1983a; Christensen and Ramananantoandro, 1988). Lower values are attributed to the clogging of microcracks by clay particles (Johnson, 1980a). Indirect estimates of permeabilities for the upper crust based on borehole thermal measurements, porosity measurements, borehole imaging and core fracture analysis, or seafloor heat flow measurements, range from lx lO\" 1 7 m 2below 200 m (Fisher et al. (1990), Costa Rica Rift, south flank) to lxlO\" 9 m 2 (e.g. Davis et al. (1997b), Juan de Fuca Ridge, east flank, 60 mbsf). A number of numerical models have been used to simulate hydrothermal convection and to estimate the magnitude of permeability and its distribution in the upper oceanic crust by using observations from various ridges as calibration targets. Most model studies were performed to simulate processes on the ridge flank rather than at the ridge crest. The oceanic crust at the ridge flank is older (several million years (Parsons and Sclater, 1977)) so that thermal conditions are considerably cooler and fluid convection is much less vigorous than at the ridge crest. As well, changes in the physical properties of sediments and basalts, resulting from geochemical alteration reactions, are likely to have slowed. Thus, difficulties such as phase separation, turbulent flow, or rapidly changing and spatially variable hydrological conditions and properties need not be considered under ridge flank conditions. In addition, modeling ridge crest conditions is 45 difficult as permeabilities at the ridge crest are less constrained because there are no direct measurements of crustal permeabilities from unsedimented ridge crests (Fisher, 1998) The modeling of convective fluid flow can be classified into two methodologies (Lowell et al. 1995): a porous/fractured medium and a fracture loop (or pipe model) approach. The porous medium approach is based on the continuum concept which states that there is a representative elementary volume (REV) (Bear, 1972), throughout which the properties of fluid and solid medium can be represented by a single value in the form of a meaningful statistical average. Fractures are not accounted for explicitly but rather are represented as localized increases in porosity and permeability. A fracture loop model consists of an open channel through which seawater enters at the seafloor. While passing through the channel the fluid constantly interacts with the surrounding rock by chemical reactions and heat exchange until it discharges back into the ocean. A fracture loop model is generally a single pass and has been modeled as flow along a single channel with (e.g. Bodvarsson and Lowell, 1972) or without (e.g. Strens and Cann, 1982) upflow and downflow zones. Fracture loop or pipe models have been used for examining the general behavior of a hydrothermal system but without considering details of the temperature and velocity distribution. Fracture loop models provide the advantage of mathematical simplicity and are justified by the evidence for hydrothermal vent fluids passing through the crust only once (Lowell and Germanovich 1997). This evidence is given by comparison of vent chemistry with seawater/basalt interaction experiments (Seyfried and Mottl, 1982), the determination of hydrothermal 46 fluid residence times (Kadko et al. 1986), and structural evidence from ophiolites (Richardson et al. 1987; Nehlig,1994). Early models of hydrothermal fluid flow focussed on concepts, analytical methods, and theoretical considerations to constrain estimates of the magnitude of crustal permeabilities and its distribution. A one-dimensional fluid flow and heat flow model by Lister (1972, 1974)) included a cracking front, which propagates into the crust as a result of the cooling and shrinking of the rock which comes in contact with seawater. The cracking front separates the convective regime in the cracked, permeable region from a conductive layer below. The percolation of water causes the formation of a system of columns, leading to overall high permeabilities in the upper crust and an anisotropic permeability distribution with vertical permeabilities twice as great as horizontal permeabilities. Fracture fatigue may lead tp a sudden collapse of the system of columns resulting in a zone of reduced permeability. The permeability decrease in those zones may be as large as four orders of magnitude (Lister, 1974). Lister assumed overall crustal permeabilties'of lx lO\" 1 6 m 2 (Lister, 1972) and higher permeabilities of up to lx lO\" 1 1 m 2 for the upper oceanic crust (Lister, 1972, 1974). Early fracture models revealed that flow rates are highly sensitive to fracture widths (e.g. Bodvarsson and Lowell, 1972). Furthermore, the duration of fluid venting and the fluid's temperature are controlled by the fracture widths and depths. Variations in fracture distribution and geometry were suggested to be the cause of heat flow variations at the seafloor (Bodvarsson and Lowell, 1972). 47 The thermal conditions leading to the formation of massive sulfides in the upper oceanic crust were used by Lowell and Rona (1985) to calibrate a series of fracture loop models. The necessary thermal conditions could be reproduced with a 1 km thick reaction zone having a permeability of 1 x 10\"15 m 2 and a permeability of the upflow zone of 1 x 10\" 1 3 m 2 . A thinner reaction zone would require a higher permeability. Mass fluxes observed at black smoker hydrothermal vents such as on the East Pacific Rise provided a constraint for discharge rates through a system of fractures in a model by Strens and Cann (1982). The fracture geometry capable of producing sufficient 8 2 mass flux was converted to an equivalent porous medium permeability of 1x10\" m . Lowell and Burnell (1991) used temperatures (~350°C) and flow rates (1-5 m/s) of \"black smoker\" vents to estimate the permeability of a hydrothermal system. A permeability of l x l 0 \" 1 3 m 2 yielded mass flux and heat output values that are close to those observed for black smoker systems but only during the first 100 years. The authors suggested that other mechanisms to stabilize the temperature and heat output of black smoker vents, such as downward migration of fluids into the underlying gabbros, were necessary. A number of studies investigated the causes leading to mega-plume events and the resulting heat output to infer hydrological conditions in the oceanic crust. Cann and Strens (1989) simulated mega-plume events by means of a low permeability cap within the discharge zone. Hydrofracturing or crustal extension then lead to an abrupt increase in the permeability of the discharge zone causing the formation of a transient plume. The 13 1 1 2 model assumed permeabilities in the recharge area of lx lO\" 1 J to 1x10\"\" nf . Similarly, Wilcock (1997) suggested a sudden increase in the permeability of the discharge zone to 48 lx lO\" 9 to lx lO\" 1 0 m 2 in addition to a high permeability reaction zone as a cause for mega-plume events. The permeability increase in the discharge zone leads to fluid decompression at depth and the formation of a plume event. In contrast, Cathles (1993) assumed the upflow zone to consist of a narrow zone at the edge of the magmatic intrusion without including a low-permeability cap. The required permeability for the discharge zone is 3xl0\" 1 4 m 2 if the permeability in the recharge zone is at least lx lO\" 1 6 m 2 . No impermeable cap was required in a fracture loop model by Lowell and Germanovich (1995) to simulate mega-plume discharge. Instead, high initial permeabilities of > lx lO\" 1 2 m 2 for the recharge zone and even higher permeabilities for the discharge zone (lxlO\" 9 m2) were suggested. As observations and measurements from various ridges provided new and more accurate calibration targets for modeling studies, more emphasize was placed on understanding the processes that occur at specific sites. For instance, mass balance considerations from the Endeavor segment of the Juan de Fuca Ridge enabled Wilcock and McNabb (1996) to estimate the effective permeability of the upper oceanic crust to be 6xl0\" 1 3 m 2 to 6x l0\" 1 2 m 2 . Probably the most common basis for the calibration of fluid flow models are seafloor heat flow measurements and heat flow patterns at various ridges. Models by Fehn and Cathles (1979,1886), Fehn et al. (1983) and Fehn (1986) incorporated heat flow components from the ridge crest as well as from below the crust. A uniform permeability model with permeability values ranging from 2.5xl0\" 1 6 m 2 to 5xl0\" 1 6 m 2 as well as an 15 2 exponentially decreasing permeability with depth from a permeability of 2.5x 10\" m 49 near the surface were assumed. Heat flow variations close to the Galapagos Spreading Center were thought by Ribando et al. (1976) to be consistent with the concept of cellular convection. A two-dimensional, porous medium model and varying boundary conditions and permeability distributions was used to match the wavelength of the heat flow highs and lows. An isotropic permeability of 4.5xl0\" 1 6 m 2 was estimated. A heat flow distribution consistent with heat flow data from the Galapagos Ridge was obtained in a three-dimensional model by Travis et al. (1991). A three layer system was used, which 15 2 • consists from top to bottom of a 1.5 km thick permeable layer (1x10\" m ), representing the basaltic extrusive rocks (pillows and flows) and dikes, a layer of intermediate permeability (2.5xl0\" 1 6m 2 , gabbros), and a low-permeability conductive layer ( lx lO\" 1 7 m 2 , upper mantle). Discrete magmatic intrusions were added to a background convective flow-pattern, which was obtained by placing a constant high temperature condition on the base. The background convective flow alone introduced significant heterogeneity in temperature and heat flux fields, similar to observations from mid-ocean ridges. The effects of a localized intrusion are limited to vigorous convection above the intrusion and have little impact on the flow on a regional scale. Heat flow measurements on the seafloor in Middle Valley indicated a near isothermal sediment/basement interface (temperature variations of less than 20°C) suggesting that the effective permeability of the basement is high enough to permit vigorous convection and thermal homogenization. A two-dimensional model study by Bessler et al. (1994) showed that stable convection can occur at a permeability higher than 8xl0\" 1 6 m 2 . The overall cooling of the system increases with a more vigorous 50 convection. Temperature variations along the interface decrease with increasing permeability of the hydrologic basement as a more permeable medium leads to lower temperatures above the upwelling and higher temperatures above the downwelling limb of the convection cell. The authors come to the conclusion that in order to obtain temperature differences of less than 20°C along the interface, vigorous flow and a basement permeability on the order of 10\"13 m 2 or higher are required, regardless of the geometry of the convection cells. This permeability value is considerably higher than the values measured for the sediment/sill sequence (Site 857) and the igneous basement (Site 858) at Middle Valley. It could imply that the discrete zones of higher permeability encountered in Hole 857D are sufficiently connected and extensive to increase the bulk permeability of the sediment/sill sequence several orders of magnitude. Heat flow measurements from Middle Valley were also used to calibrate a 3D model by Rabinowicz et al. (1998). Results of this study include an assessment on the increased accuracy of 3D models over 2D models. The authors assume a homogeneous permeability distribution in the permeable convective layer and the overlying layer of low-permeability sediments as well as an exponential permeability increase with depth based on permeability measurements taken on samples from sediments at Middle Valley (Fisher at al., 1994). For the simulations with uniform permeabilities, values of 5xl0\" 1 5 m 2 and 2x 10\"16 m 2 were assigned to the convective basement and the sediments, respectively. For the depth-dependent permeability case, these permeabilities represent the values assigned to the bottom and top boundary of the domain, respectively. With the three-dimensional simulations the authors were able to obtain heat flow values close to 51 the observed values and successfully modeled the characteristic circular shape of the regions of high heat flow at the seafloor of Middle Valley. The calculated pressure distribution in the sediments is such that lithofraction and associated with it an increase in permeability is possible near the top of the sediment cover. A number of modeling studies have been carried out for ridge flanks. Despite the older crust at the flank and the resulting lower heat-flow and less vigorous convection, certain constraints on fluid flow are similar at both the ridge flank and the ridge crest. However, the oceanic crust at the ridge flank is likely to have undergone chemical alteration, which may result in an overall permeability reduction. Thus permeability estimates from the ridge flank may be seen as a lower limit of permeabilities at the ridge crest. In addition, modeling studies of the ridge flank have revealed a great deal of general insight into the permeability distribution as well as the effects of certain conditions, such as basement relief or thickness of the sediment layer, on fluid convection. Therefore it is useful to summarize a number of key results from flank models. In a study to investigate hydrothermal circulation at the eastern flank of the Juan de Fuca Ridge, Davis et al. (1992) found a low-amplitude but coherent heat flow signal in an area of the crust that exhibits only minor basement relief and is covered by a uniform sheet of sediments. The authors suggest that this signal may be caused by cellular convection in the permeable oceanic crust. The eastern flank of the Juan de Fuca Ridge was also the subject of a modeling study performed by Davis et al. (1996). Similar to the analysis at Middle Valley, the 52 model results were constrained by thermal homogenization at the sediment/basement interface (T-variations <10°C). The results indicate that two basement permeabilities are possible to provide the degree of isothermal conditions along the interface, one at approximately 3x10\"14 m 2 , slightly above the critical value for the onset of convection (2xl0\" 1 4m 2), the other at about 2xl0\" 1 2 m 2 . The former value is close to results obtained from packer experiments done in boreholes drilled into the oceanic crust. Because of the proximity of the lower permeability value to the critical permeability, the authors favor the higher permeability value because hydrothermal circulation would be much more robust and persistent. The large discrepancy between the calculated and measured permeabilities may be explained by fractures that are too widely spaced to be sampled representatively by drilling. Many convection models are based on the assumption of a flat layer geometry where a permeable basement layer is over - and underlain by less permeable sediments and the lower crust, respectively. However, basement relief was reported to have a strong effect on the geometry and vigor of convective circulation (e.g. Lister, 1972). The depth to which the effect of basement topography extends is limited by the amplitude of relief (Lowell (1980) and Hartline and Lister (1981)). Fisher et al. (1990) showed that geothermal, geochemical, and hydrological observations from Site 504 (Costa Rica Rift, south flank) could be explained by topographically driven flow and the formation of large aspect ratio convection cells. The model combined basement relief with an amplitude of several hundred meters and a wavelength of several kilometers, varying thickness of low-permeability sediments 53 overlying the basement, and thermal and geochemical homogenization of fluids at the sediment/basement interface. Downhole logs and packer measurements at Site 504 indicate that the upper 100-200 m of basement are significantly more permeable than the remaining upper crustal section. Consequently, they adopted a three-layer model: a 100 m thick upper layer with a bulk permeability of 1x10\"13 m 2 , an intermediate layer of 100 m with a permeability of 5x10\"'5 m 2 and a lower layer with a permeability less than 1x10\"17 m 2 . In a later study Fisher et al. (1994) found that thin, highly permeable zones (permeabilities as high as 1x10\" m ) within the uppermost basement are necessary to explain the heat flow around Site 504. Despite the presence of these high permeability conduits, the average permeability of the basement was assumed to be consistent with in situ measurements of bulk permeabilities. Davis et al. (1997b) used data obtained from the eastern flank of the Juan de Fuca Ridge and included basement topography in order to assess its effect on the efficiency of heat transfer. In addition, Davis et al. (1997b) argued that by introducing basement topography the ambiguity of the earlier study (Davis et al. 1996), in which near isothermal interface conditions were obtained for both near-critical and vigorous convective circulation, will be eliminated. The modeling domain consisted of a tilted convecting layer, which leads to an increase in sediment thickness from one side of the domain to the other. Permeabilities required to obtain near isothermal conditions along the interface are of the order of 10\"11 m 2 for a 600 m thick convecting layer. By reducing the thickness of the permeable layer by one order of magnitude, the permeability has to be 54 increased by two orders of magnitude to obtain the same temperature distribution at the interface. Yang et al. (1996) were able to match observed seafloor heat flow magnitudes and wavelengths without invoking basement relief or differential sediment or basement thicknesses by introducing large-scale fracturing within the oceanic crust. Their modeling domain consists of two basement layers and rock-matrix permeabilities of 5xl0\" 1 4 m 2 and lx lO\" 1 6 m 2 , respectively. Fractures were randomly distributed, vertically or horizontally 9 2 * oriented, 0.22 wide discrete zones of high permeability (4x10\" m ) with a spacing of at least 10 m. A general study designed to demonstrate the significance of the permeability distribution on a ridge flank was carried out by Rosenberg et al. (1993). The authors discussed the relation between the types of permeability fields likely to exist at mid-ocean ridges and the associated patterns of fluid flow and heat flux. The model domain consisted of a permeable ( lxlO\" 1 3 m2) layer sandwiched between two low-permeability layers. Under the imposed boundary conditions, a uniform, isotropic permeability of the permeable layer led to temperatures along the sediment/permeable layer interface of 37 +/- 3°C. A n enhanced horizontal permeability, which was intended to represent horizontal fracturing in the shallowest basement, results in greater temperature differences along the interface. For a scenario where the vertical permeability is greater than the horizontal permeability, which is likely to occur in the vertically jointed dyke section of the deeper crust, the temperatures along the interface become more uniform. The bulk permeability controls the depth to which significant fluid flow extends, the introduction of anisotropy 55 tends to affect the number of convection cells. Increasing the thickness of the permeable zone results in higher maximum flow velocity, fluid flux, differences in temperatures along the sediment/permeable layer interface, variations in heat flux at the seafloor, and aspect ratios of the convection cell. The question of how sediment permeability and thickness control the transition from an open convective system with exchange between the ocean and the basement, and a closed system without fluid exchange, was addressed in a model study by Snelgrove and 13 2 Forster (1996). The permeability of the convective basement is fixed at 1x0\" m . The bulk permeability of the sediments is varied within a range typical for silt-rich and clay-rich sediments. They also vary permeability with depth according to empirical depth/permeability relationships. Using this relationship for silt-rich sediments, permeabilities range from about lx lO\" 1 3 m 2 near the seafloor to about 3 x l 0 \" 1 5 m 2 at the bottom of the sediment layer. Clay-rich sediment permeabilities range from about 3 x 1 0 \" ' 5 m 2 to lx lO\" 1 7 m 2 over the same depth interval. The authors conclude that for a given basal heat flux of 0.27 W/m 2 , the transition from a closed to an open system takes place over a narrow range of permeabilities, between lx lO\" 1 5 and 5 x l 0 \" 1 5 m 2 . Testing the effects of different sediment thicknesses demonstrates that the thickness plays a relatively minor role in influencing the circulation pattern in the convective basement, the effective permeability of the sediments exerts a much stronger control. In conclusion, despite the increasing database and sophistication of models there is still a large degree of uncertainty with regard to the magnitude and distribution of permeability in the upper oceanic crust. The range of measured and estimated 56 permeabilities is wide and too little is known about the permeability distribution to constrain either one with confidence. It appears that observations from various ridges, such as an approximately isothermal upper basement surface, can be created by any form of high Nusselt number hydrothermal regardless of the details of the permeability distribution or anisotropy. The Nusselt number expresses the ratio of the total heat transfer to the heat that would be transferred by conduction alone in the absence of convection. For this reason, additional information, such as that which can be derived from geochemical data, appears necessary to advance our understanding of the actual flow patterns and permeability distributions. To make efficient use of geochemical data, information of the chemical conditions in the past and at present are required. The model study described in this chapter is aimed at gaining insight into possible implications of transient fluid flow and temperature distributions on the chemistry. 2.1.2 Mathematical Model 2.1.2.1 Governing Equations The governing equations for hydrothermal fluid flow in porous media are derived by considering the conservation of fluid mass and heat. The specific discharge is given: q = dL (VP - pwgVZ) (2.1) where q is the specific discharge, k is the permeability of the rock matrix, P is the pressure, g the gravitational acceleration, and Z is the elevation relative to datum. The variables p w and p w are the dynamic fluid viscosity and fluid density, respectively, both 57 are temperature and to a lesser degree pressure dependent. For the case where the fluid is compressible and the rock mass is treated as non-deformable, and there exists a thermal equilibrium between fluid and solid matrix, the equation for conservation of fluid mass can be written as v • \\ P w — ( V P + Pwgvz)\\ = frw[pw ^r+rw (2.2) [ M w J v dt ' w dt j where T is the temperature, § is the porosity, and (3W and y w are the isothermal compressibility and the isobaric thermal expansivity of the fluid, respectively. The governing equation of transient heat transport is v • (^,vr)- v • {pwhwq}= ^y>p„K + (i - *)pmK ] (2.3) where h w is the enthalpy of the fluid, p m and h m are the density and enthalpy of the rock, respectively, and X,y is the effective thermal conductivity of the rock-fluid mixture. The enthalpy of the rock matrix (hm) is assumed to change linearly with temperature according the relation K=cm*{T-T0) (2.4) where c m is the specific heat of the rock matrix and To is the standard temperature (273.14K). The effective thermal conductivity h,j of the rock-fluid mixture is represented as: (a, - a\\q.q{ . . *•» = 11 + AM\\^ + W»>S>J (2-5> 58 where ai and a t are the longitudinal and the transverse dispersivities, respectively, 5y is the Kronecker delta, and Xw and A,m are the thermal conductivity of fluid and rock, respectively. The thermal conductivity of solids and fluids of uniform composition is primarily dependent only upon temperature. The temperature dependence of the thermal conductivity is not accounted for in this study. Also, the transient fluid flow and heat transport simulation does not include the feedback of chemical reactions, tectonic processes, sediment deposition and compaction on the permeability and other properties of the solid medium. The fluid properties, such as thermal conductivity, density, and viscosity are calculated using the equations from the A S M E Steam Table (Meyer et al. 1983). The effect of salinity on the fluid properties is not taken into account. The velocity fields obtained from the coupled flow and heat transport simulation are used to carry out a non-reactive transport simulation in section 2.2.6. The conservation equation for transport of solute species in a multicomponent aqueous fluid can be written as where C\\ is the concentration of the M i species, Dy is the dispersion coefficient which combines the effects of mechanical dispersion and diffusion, andv is the average linear fluid velocity defined as V - ( D f f V C t ) - V . ( v C t ) dCk (2.6) (2.7) 59 A detailed outline of the mathematical formulation of the transport routine is presented in chapter 3.1.2. 2.1.2.2 Numerical Scheme The coupled flow and heat transport code used in this study was developed at the U.B.C. by Motomu Ibaraki. In this code the governing equations 2.2 and 2.3 are discretized using a finite-volume approach (Patankar,1980). There are three variables in equations 2.2 - 2.4, fluid pressure (P), the enthalpy of the fluid (hw) and the temperature (T). Pressure and either enthalpy or temperature are used as primary variables. If the temperature of a node is specified explicitly, temperature is a primary variable, otherwise it is the fluid enthalpy. The two governing equations for the two unknowns, i.e. pressure and either temperature or enthalpy, are solved simultaneously at each grid node at each time step using the Newton-Raphson method. A matrix solver which takes advantage of the block structure in the matrix is employed (Forsyth and Sammon, 1986; Ibaraki et al. 1996). The numerical scheme fully accounts for density changes of the fluid, making the Boussinesq approximation, which is commonly used in studies on thermally-driven free convection, unnecessary. The numerical formulation of the transport routine used in chapter 2.2.6 is presented as part of the description of the reactive transport code (chapter 3.1.2). 60 2.1.3 Model Design 2.1.3.1 Flow Domain The model domain is a two-dimensional cross-section of a sedimented oceanic crust (Figure 2.1). The domain is 2000 m wide and 1020 m deep. The upper 500 m of the domain represent low-permeability sediments (layer 1), overlying 500 m of higher-permeability basalt (layer 2). Below the permeable basalt there is a thin (20 m) conductive layer. This conductive bottom layer was added to account for lateral heat flow variations below the permeable layer that originate as a consequence of the convective circulation. The choice of 20 m for the thickness of this layer was made based on numerical convergence problems for thicker bottom layers. On both sides of the domain and penetrating the sediments are 5 m wide 'windows' of higher permeability that act as a recharge zone for seawater entering the basement and a discharge zone of hydrothermal fluids into the ocean. The design of the domain is loosely constrained by observations from Middle Valley. The thickness of the sediment layer approximately corresponds to the thickness of the sediment cover at Site 857. Boreholes at this site were drilled into the reaction zone, the section of oceanic basement in which the circulating fluids obtain the geochemical fingerprint of a hydrothermal fluid. Recharge into the reaction zone occurs through higher permeability pathways, such as fault zones encountered at Site 855. The hydrothermal fluids exit the basement either as slow, diffuse discharge through the seafloor or as focussed discharge through a higher permeability upflow zone, such as at Site 858. There 61 is evidence that Sites 857 and 858 are hydrologically connected. The sites are approximately two kilometers apart. The width of the recharge and discharge zones is similar to the estimated width of a fracture zone in the basement at Site 857 (Becker et al. 1994). According to observations, at Site 857 basement relief is probably not significant enough to affect fluid flow in the basement. Thus the assumption of equal sediment thickness and a flat sediment/basement interface is reasonable. 2.1.3.2 Initial and Boundary Conditions A constant fluid pressure of 25 M P a is assigned to the upper boundary of the domain, representing the seafloor hydrostat (Figure 2.1). The isobaric seafloor permits limited fluid discharge and recharge through the sediment cover. We assume a constant 2 2 heat flux of 850 mW/m along the base of the domain. The value of 850 mW/m is in the range of measured heat fluxes at Middle Valley (Figure 1.3). The temperature at the seafloor is specified as 1°C, which corresponds to the bottom ocean temperature. Note that a fixed temperature at the seafloor will lead to an overestimate of the temperature gradient in the upper part of the discharge zone. Where appropriate, the influence of this boundary condition on the model results will be highlighted. Both side boundaries of the domain are impermeable and adiabatic. The initial condition used for all flow and heat transport simulations is a conductive temperature field and no flow within the domain. 62 2.1.3.3 Ch oice of Parameters The underlying assumption of the use of Darcy's Law is to replace the actual medium with a representative continuum in order to use a macroscopic law to describe flow at a microscopic scale (Bear, 1972). This implies that the complex flow configuration at the pore scale or, in a fractured medium, at the fracture scale, is ignored. We assume that our modeling scale is large enough to eliminate the dominance of single fractures within the system and that it is reasonable to approximate the hydrologic basement with a homogeneous porous medium model. Under these assumptions we can assign an average uniform effective permeability and porosity to the medium. At Middle Valley, the results from packer and flowmeter experiments in the basement indicate that the lowermost 180 m of the sediment/sill sequence in Hole 857D and the upper igneous basement in Hole 858G both have an average permeability on the order of 1 x 10\"14 m 2 (Becker et al. 1994) (Figure 1.4). Therefore we choose a basement permeability of lx lO\" 1 4 m 2 to describe the evolution of a convective system in detail (Figure 2.1). These results are then compared to those obtained using higher and lower basement permeability values. The range of permeabilities used in this study spans two orders of magnitude, from l x l 0 \" 1 5 m 2 t o lx lO\" 1 3 m 2 (Table 2.1, Figure 2.1). The sediments are mainly of a thick sequence of turbidites. The sequence of low-permeability clays and the more permeable silts and sands in the sediment layer causes the sediments to be anisotropic. Bulk permeability measurements taken on samples from Middle Valley yielded values between 9.9xl0\" 1 7 m 2 and 1.8xl0\"15 m 2 (Fisher et al. 1994). 15 2 In-situ estimates resulted in a permeability range of about 5x10\" m near the seafloor to 63 about l x l 0 \" 1 7 m 2 at 100 m depth (Figure 1.4). According to Davis et al. (1992b) and Rigsby et al. (1994) the sand/mud ratio in the upper part of hole 85 8A is about 1:10. Average permeability values in the vertical and horizontal direction used in this study are obtained by using bulk permeability values of 5x 10\" m for the sand and a value of lx lO\" 1 7 m 2 for the mud/clay fraction and taking the weighted harmonic and arithmetic average, respectively. The calculated average vertical permeability remains close to the permeability value of the clays, around lx lO\" 1 7 m 2 . This indicates the strong control of the low-permeability layers on vertical flow rates. For the horizontal permeability an average of 5xl0\" 1 6 m 2 was obtained. According to Snelgrove and Forster (1996) these permeability values are low enough to effectively seal the convective basement from the seafloor. The permeability of the recharge and discharge zones is specified to be 7.5x10\"14 m 2 . There is no observational constraint for this value. Based on simulations presented later, this permeability value is chosen to obtain temperatures of basement fluids in agreement with observations from Middle Valley. The porosity and density of the medium are uniform throughout the domain and are fixed at 10% and 2000 kg/m 3, respectively. The dispersivities of the medium in both layers are 10 m in the longitudinal direction and 1 m in transverse direction. Thermal properties of the solid phase are also constant and uniform throughout the domain. The thermal conductivity is 1.728 W/m°C and the heat capacity is fixed at 1000 J/kg°C (Table 2.1). 64 2.1.3.4 Discussion of Model Uncertainties A mathematical model is only a simplified image of reality. Our goal is to model processes within a framework of realistic assumptions and to obtain results that allow general conclusions, subject to constraints imposed by computational requirements. In order to be able to assess the validity of the results it is important to be aware of the limitations and inaccuracies introduced by the approximations and limitations that have been made. The review of permeability estimates and measurements demonstrated that there is great variability and significant uncertainty in the magnitude and the distribution of permeability in the upper oceanic crust. Rather than a conceptual model based on cellular convection in a homogeneous basement, it has been suggested that fluid flow may be focussed in high permeability layers. Similar to the permeability the porosity of the medium directly affects the magnitude of the velocity. Therefore the choice of porosity exerts an important control on the evolution of a hydrothermal system. The volume of cold seawater recharging the basement is likely to influence the. rate of cooling and thus the evolution of the hydrothermal system. Thus the transmissivity of the recharge zone and the discharge zone are important parameters but the values are not constrained by observations from Middle Valley (chapter 1). Ignoring the feedback on the permeability and other properties of the solid medium of chemical reactions, tectonic processes, sediment deposition and compaction in this analysis will also affect the evolution of the hydrothermal system. Furthermore, some 65 of the dependencies between model parameters (e.g. the temperature dependence of the thermal conductivity or the heat capacity of the solid phase) as well as other conditions acting on the magnitude of model parameters (e.g. the effect of salinity on fluid properties) are not accounted for. Another limitation of the model arises from its two-dimensional geometry. Laboratory experiments (e.g. Hartline and Lister, 1977) and modeling results (e.g. Rabinowicz et al. 1998) demonstrate that the flow geometry is three dimensional with asymmetrical cells, consisting of plume-like ascending flows and sheet-like descending flow. A modeling study by Rabinowicz et al. (1998) demonstrates that a 2D model generally underestimates the subsurface temperatures and the Nusselt number. The choice of a conductive temperature field and no flow as the initial condition is unavoidably arbitrary but the non-uniqueness of the initial conditions in nature makes it difficult to make a well-constrained choice. Rosenberg et al. (1993) investigated the effect of different initial conditions on the evolution of a convective flow system. For an initial temperature field with a temperature equal to the top boundary temperature everywhere, for instance, the authors found that it takes considerably more time to reach steady state than for an initial conductive temperature field. Thus the time scales presented in this study are subject to our choice of the initial conditions. Nevertheless, the chosen initial condition should not greatly affect the trends during the evolution of a hydrothermal system, which is the emphasis of this study. It has also been reported by these authors that the choice of the initial condition influences the steady state geometry (aspect ratio and number of convection cells) of a closed convective system. This should 66 not be an important aspect in an open system because the flow geometry in the basement is strongly controlled by the recharge and discharge zones. Inaccuracies may also originate from the assigned boundary conditions. A fixed temperature along the seafloor results in an overestimate of temperature gradients in the upper part of the discharge zone. To avoid problems imposed by the upper boundary, all calculations that include processes within the discharge zone will be performed for the lower part of the conduit. The assumption of a constant heat-flux at the bottom boundary over thousands of years is another simplification. Keeping the heat flux constant throughout the simulation implies that the heat source at depth is replenished. However, the heat source may often be a relatively small magmatic intrusion and the duration of heat production may not be sufficient for the hydrothermal system to reach a steady state. In other words, the boundary conditions of the system may vary at a faster rate than the hydrothermal system is able to respond. Another factor to consider is the fixed horizontal extent and the impermeable side boundaries of the domain, which impose vertical upflow or downflow parallel to these boundaries. The number and the dimensions of the cells produced by the model will not have the naturally preferred values and the magnitude of the calculated heat transfer will be affected. However, the high permeability windows in the sediments would probably impose a similar effect on the convection geometry, i.e., creating zones of vertical fluid flow in the basement. Thus the impact of the no-flow boundaries may be less significant. In summary, despite several key simplifications, the results presented in this chapter are based on reasonable constraints and assumptions. The results may not be 67 representative in every detail but the general conclusions are valid and of general significance. 2.2 The Evolution of a Hydrothermal Convective System In this section, the evolution of the convective system is described in detail for a basement permeability of lx lO\" 1 4 m 2 . Differences and similarities to conditions at lower and higher basement permeabilities are pointed out where appropriate or necessary. 2.2.1 The Evolution of Temperature and Velocity Fields Figure 2.2 illustrates the evolution of fluid flow and temperature fields using snapshots at various times for a basement permeability-of l x lO\" 1 4 m 2 . Fluid flow is initiated by a random disturbance of the temperature field except in regions where temperatures are fixed. For time periods less than 500 years after the disturbance of the initial conductive, no-flow state, fluid flow is highly vigorous. The flow pattern is irregular and characterized by small unstable convection cells with low aspect ratios. The flow field evolves by the merging of these small convection cells into larger cells. The range of fluid velocities in the basement is relatively large. The highest fluid velocities occur in the upflow and downflow zones of the convection cells, the lowest towards the center of the cells. The flow pattern leads to a wave-like arrangement of the isotherms, 68 they are bent upwards where hot fluids ascend from the bottom boundary towards the sediment/basement interface and bent downwards where fluids move from the interface towards the bottom boundary. A similar wave-like pattern but with a much lower amplitude can be observed in the bottom section of the sediments. Here the isotherms are bent upwards above the hot rising plumes in the basement. The wavelength of the lateral temperature variations in the sediments corresponds to roughly twice the width of a single convection cell in the basement. Due to the vigorous convection the range of basement temperatures is relatively low. The continuous inflow of cold seawater leads to the formation of a thermal plume which enters and spreads through the basement. The front of the plume moves at a rate that is slower than the front of fluids of seawater composition by a factor that is equivalent to the porosity (10%). As soon as the thermal plume enters the basement (at around 500 years) the convection pattern changes dramatically. On the left-hand side of the domain, the steady progression of the thermal front leads to the formation and expansion of a single large convection cell. The front of the plume exhibits high temperature gradients and it separates the basement into a region of relatively low fluid temperatures within the plume and a region of higher fluid temperatures outside the plume. The result is a considerable temperature change over a short period of time in regions affected by the plume. Below the discharge zone, fluid temperatures are still relatively unaffected by the thermal evolution on the opposite side of the domain. Hot fluids ascend from the bottom boundary and either exit the system through the discharge zone or become 'recycled' in the basement. 69 The number of convection cells in the basement decreases as the convection cell adjacent to the recharge zone continues to grow. At 1000 years it fills nearly the entire left half of the basement. Eventually, the hydrothermal system consists of only one large convection cell and the basement fluids circulate around a single stagnant area (Figure 2.2). This state has evolved at about 2500 years. The height of this cell is confined by the top and bottom boundaries of the basement layer and its horizontal dimensions extend almost from one side of the domain to the other. This leaves only a narrow zone for fluids that recharge, pass through, and exit the basement. After the single convection cell has formed, changes in flow geometry become relatively insignificant, i.e. the hydrothermal system has obtained a stable flow pattern. Despite the stable flow pattern, however, the system is not at a steady state as the overall vigor of convection continues to decrease. In this stable, single cell flow field the highest fluid velocities occur in the regions of horizontal flow and in the region of upward flow. The lowest basement fluid velocities are present near the axis of rotation of the cell. The topology of the isotherms is consistent with the inflow of cold seawater, which becomes heated while flowing along the bottom boundary. The counterflow along the sediment/basement interface causes a moderate cooling of the basement fluids. The highest fluid temperatures in the domain occur at the bottom boundary below the discharge zone. The high flow velocities in the regions of upward flow result in low temperature gradients so that fluids exiting the domain undergo only moderate cooling during their ascend towards the seafloor. The evolution of the convective system shows similar characteristics at different basement permeabilities. The main differences occur in terms of time-scales and magnitudes of temperature and velocity values. The duration of unstable convection, i.e. 70 the state before the onset of a single convection cell, decreases with increasing basement permeability. From the initial conductive, no-flow state to the onset of a single convection cell, the times required for basement permeabilities of 1 x 10\"15 m 2 , l x l 0\"14 m 2 and 1x10\" 1 3 m 2 are about 15000, 2500, and 1500 years, respectively. The evolution of the convective system for a basement permeability of 1 x 10\" m is illustrated in Figure 2.3. Over the first 3000 years hardly any fluid motion is present and the system essentially remains in its intial state. The pattern of convective circulation which begins to form after this long period of near-stagnation is characterized by fewer cells with higher aspect ratios than for a basement permeability of lx lO\" 1 4 m 2 (Figure 2.3). Also, flow velocities are low and do not show the same degree of variations as at higher permeabilities. In contrast, the range of basement temperatures is wide and the overall temperatures are high. The sluggish flow and the slowly changing convection patterns lead to more significant thermal variations at the sediment/basement interface and stronger lateral temperature variations within the sediments. These lateral temperature variations in the sediments have a higher amplitude and a lower wavelength than at higher basement permeabilities and extend almost over the entire bottom half of the sediments. At 5000 years the steady inflow of cooler seawater leads to the formation and growth of a thermal plume within the basement. The evolution of the plume triggers the onset and growth of a single convection cell that eventually constitutes the stable flow pattern (at about 15000 years) (Figure 2.3). While at higher basement permeabilities this convection cell essentially fills the entire basement, the cell is much smaller at a permeability of lx lO\" 1 5 m 2 and it is confined to the right half of the domain. Thus fluids recharging the basement can spread throughout the left half of the domain, affecting a much larger portion of the 71 basement. However, considering the relatively low fluid volumes entering the basement the potential for significant chemical alteration as a result of a thermal and chemical disequilibrium between fluid and rock is probably small. An interesting observation can be made from the topology of the isotherms once the stable flow pattern is obtained. The lower recharge rates at low basement permeabilities in combination with the greater spread of recharging fluids leads to only a mild depression of the temperatures near the recharge zone. Over a large portion of the basement, the pattern of the isotherms does not deviate greatly from that of a conductive temperature field. As a consequence, lateral temperature variations at the sediment/basement interface and within the sediments become small as the system approaches steady state, thus mimicking the same effect as thermal homogenization due to vigorous convection in terms of heat flow distribution at the seafloor. 13 2 An increase of the basement permeability to 1x10\" m causes the early convection to be vigorous enough to thermally homogenize the basement (Figure 2.4). The temperatures between the bottom boundary and the sediment/basement interface do not vary by more than 20°C in the first few hundred years. While at lower permeabilities cellular convection is clearly recognizable and individual convection cells can be identified even at the early stages of the evolution of the flow field, this is not the case at a permeability of lx lO\" 1 3 m 2 . At a permeability this high the early flow field is extremely irregular. Only when the thermal plume enters the basement a pattern of cellular convection becomes recognizable (Figure 2.4). This process of the stabilization of the flow field is similar to that at lower permeabilities in that a convection cell begins to form 72 adjacent to the recharge zone and continues to grow until it fills almost the entire basement (Figure 2.4). The stage of a single large convection cell occurs much earlier, however, at roughly 1500 years. As the system evolves, flow velocities remain high. At steady state, lateral temperature variations in the basement amount to about 50°C . In summary, for the range of basement permeabilities the evolution of hydrothermal convection begins with a phase of highly vigorous, irregular fluid flow and high temperature conditions. The steady inflow of cooler seawater leads to the formation of a thermal plume. As soon as the front of the plume enters the basement, the flow field begins to stabilize towards a single convection cell. As the plume enters and spreads through the basement, the rate of cooling accelerates and the overall vigor of convection decreases correspondingly. When a single convection cell has formed, the flow geometry is stable but the overall cooling and decreasing vigor of convection continues until a steady state is reached. The evolution differs for the range of basement permeabilities in terms of timing of these phases, the vigor of convection and convection patterns, and the thermal conditions of the system. At high basement permeabilities, the evolution proceeds faster, fluid flow is initially more erratic and generally more vigorous, and the basement is more thermally homogeneous. 2.2.2 Average Basement Temperatures and Temperatures at Recharge and Discharge Zones 73 The evolution of average basement temperatures (i.e. average temperatures for the entire region below the sediment/basement interface) and temperatures at the bottom of the recharge and discharge zones (50 m above the sediment/basement interface) for a range of basement permeabilities are illustrated in Figure 2.5. Average basement temperatures are quite similar in terms of magnitude and trends for the permeability range from 5xl0\" 1 5 m 2 to lx lO\" 1 3 m 2 . The temperature evolution is characterized by a smooth, rather steady decline towards a steady state value. The difference at 20000 years for the above-mentioned permeability range is only about 30°C. It should be emphasized that the 15 2* evolution of the average basement temperature for a low permeability of 1 x 10\" m i s quite different from higher permeabilities. Comparing the trends for permeabilities of lx lO\" 1 5 m 2 and 5xl0\" 1 5 m 2 , for instance, it appears that the evolution towards a stable convection pattern becomes much more sluggish for a permeability difference of only half an order of magnitude. For the lower permeability, it takes about 2500 years before any noticeable changes in basement temperatures from the initial state occur. It is only after the cold plume enters the basement at about 2500 years that the resulting temperature gradients are sufficient to trigger convective circulation and the temperature evolution approaches that of higher basement permeabilities. In contrast to the smooth decline of average basement temperatures, the temperatures at the bottom of the recharge zone decrease rapidly when the thermal plume reaches the bottom of the recharge zone and then decrease only slightly before reaching a steady state value. The same trend occurs for all permeability values. At 20000 years the temperature at the base of the recharge zone is higher for a higher basement permeability. 74 15 2 The difference amounts to almost 100°C between a permeability of 5x10\" m and 1x10\" 13 2 15 2 m . In contrast, the temperature difference between permeabilities of 5x10\" m and lx lO\" 1 4 m 2 is only minimal. The temperature evolution within the discharge zone for a permeability of lx lO\" 1 4 m 2 shows a steep increase shortly after the initial disturbance of the conductive temperature distribution. Temperatures of discharging fluids rise from about 230°C to 345°C within the first 500 years after the onset of fluid movement. From then on discharge temperatures undergo low-amplitude (tens of degrees) temperature increases and decreases superimposed on a weak overall decline. These temperature fluctuations have a wavelength of roughly 1000 years and last until about 2800 years. After 2800 years the temperature decline is steady, rather steep at first and then leveling off to a steady state value which amounts to about 275°C. The transition from the temperature fluctuations to the steady decline coincides with the onset of the stable convection pattern of a single convection cell. A similar pattern of discharge temperatures can be observed for higher and lower basement permeabilities. Differences occur in terms of maximum discharge temperatures as well as the amplitude and duration of the initial temperature fluctuations. Maximum discharge temperatures are generally higher and occur earlier at higher permeabilities. As the evolution towards a stable convection pattern takes more time at lower basement permeabilities, the period of temperature fluctuations is longer for lower basement permeabilities. Whereas these fluctuations are short-lived and hardly detectable for a high permeability of lx lO\" 1 3 m 2 , at lower permeabilities of lx lO\" 1 5 m 2 these fluctuations are 75 more significant in terms of amplitude (up to 25°C) and wavelength (on the order of 1000's of years). The general trend is therefore that at higher basement permeabilities the initial discharge temperatures are higher but the duration of the high temperature state is shorter. This trend is consistent with the earlier spread of the thermal plume through the basement. In addition, the early state temperature fluctuations have a considerably greater wavelength and a larger amplitude. At 20000 years, close to steady state, discharge temperatures tend to be lower at higher basement permeabilites but fall into a relatively small range. The temperature 13 2 differences at this time are only minimal between permeabilities of 1 x 10\" m and 5x10\" 1 4 m 2 and amount to about 40°C between permeabilities of lx lO\" 1 3 m 2 and 5xl0\" 1 5 m 2 . The small range of discharge temperatures suggests that the temperature of vent fluids, which controls the formation of massive sulfides among other things, does not appear to be primarily dependent on the basement permeability or vigor of convection in the basement. Large temperature differences at the recharge zone and a narrow range temperatures at the discharge zone for the permeability range from 5xl0\" 1 5 m 2 to lx lO\" 1 3 m 2 imply that a significantly larger temperature gradient across the domain exists at lower basement permeabilities. For a permeability of 5x10\" m , for instance, this difference amounts to 265°C, whereas for a higher permeability of lx lO\" 1 3 m 2 this difference is only about 120°C. This observation is consistent with a higher degree of thermal homogenization at higher basement permeabilities due to a more vigorous convection. Thus at low basement permebilities a fluid parcel undergoes a much greater temperature 76 increase as it travels from the recharge zone to the discharge zone. This large temperature increase is likely to lead to chemical alteration patterns that are more heterogeneous at lower permeabilities than at higher permeabilities. 2.2.3 Fluxes through Recharge and Discharge Zones and the Sediment/Basement Interface To assess the evolution of recharge and discharge rates for the hydrothermal basement, total fluid volume and mass fluxes across a horizontal line at 500 mbsf are calculated. This line is divided into three intervals consisting of the recharge zone, the discharge zone, and the sediment/basement interface, i.e. the line separating the low-permeability sediments from the basement. The fluxes are calculated for each of the intervals separately; downward flow rates through the recharge zone, net fluxes across the sediment/basement interface and upward flow rates through the discharge zone. The evolution of mass fluxes through each of the three intervals is shown in Figure 2.6. Fluid recharge into the basement occurs through the recharge zone and as recharge through the sediments. Fluxes through the recharge zone are generally more than twice as large as the net recharge rates across the sediments/basement interface. The total recharge into the basement is balanced by the discharge of fluids through the discharge zone. Fluid flow through the conduit of the discharge zone is therefore much more vigorous than flow through the recharge zone into the basement. 77 For each permeability value the evolution of mass fluxes through the recharge zone, sediment/basement interface, and discharge zone shows parallel trends as required by mass balance constraints. As well, the evolution of fluxes for each permeability value correlates with the evolution of discharge temperatures for the same permeability (Figure 2.5). At a basement permeability of lx lO\" 1 4 m 2 for instance, the flux curves show distinct peaks and lows for about 2800 years before starting a decline towards a steady state value. Each peak and low coincides with the variations in the discharge temperature. As with the temperature curve, the onset of the smooth decline approximately coincides with the onset of a large single convection cell so that the initial variations of the flux curves are attributed to the changing convection pattern in the basement. At lower basement permeabilities the initial fluctuations are longer lasting and the highs and lows are more pronounced. It should be noted that fluctuations of discharge rates at hydrothermal vents on the ocean floor have indeed been observed. Once the stable convection pattern is established, recharge rates through the recharge zone are remarkably similar for all basement permeabilities. In contrast, recharge across the sediment/basement interface is lower at higher basement permeabilities. Differences in total recharge rates for all basement permeabilities are thus predominantly a result of different flow rates through the sediment/basement interface. These differences are balanced by the total discharge rates, which decrease with increasing basement permeability. The correlation of fluid fluxes and discharge temperatures during the state of unstable convection may raise the question why there is not a drop in discharge 78 temperatures as the recharge of cold seawater increases. The reason why this is not the case is that the total recharge consists of a component that enters the basement through the recharge zone and one that enters the basement through the sediments. Even though the latter component is generally less than half of the former, the fluctuations of the recharge through the sediments are much more pronounced. In addition, considering the higher temperature of the fluids at the bottom of the sediments compared to the temperature of the fluids at the bottom of the recharge zone, the correlation between the discharge temperatures and the fluxes may arise as a result of the variations in the fluxes of high temperature fluids through the sediments. A larger flux of higher temperature fluids across the sediment/basement interface leads to an increase in the temperature of the vent fluids. In conclusion, recharge into the basement occurs as focussed flow through the recharge zone and as slow flow across the sediment/basement interface. The amount of sediment pore water entering the basement becomes more significant at lower basement permeabilities and we would expect to see a more pronounced geochemical fingerprint of the sediments in the composition of the vent fluids. Moreover, temporal variations in flow rates through the sediments are significant enough to affect discharge rates and the thermal state of the discharging fluids. Lower basement permeabilities lead to a higher total mass discharge as a consequence of higher flow rates through the sediment/basement interface at lower basement permeabilities on one hand but similar recharge rates for the range of basement permeabilities on the other hand. 79 2.2.4 Flowlines and Travel Times A helpful method of visualizing and understanding the geometry and the velocity distribution of the flow field is the use of flowlines. Flowlines can be regarded as centerlines of streamtubes or as advective travel paths of particles released at some point in the domain. With the calculation of flowlines we can track advected particles and determine their position as a function of time. It has to be pointed out, however, that it is impossible to predict the path of an individual solute particle accurately without accounting for dispersion. The error introduced through neglect of dispersive processes becomes larger with the distance traveled by the particle and higher flow velocities (assuming the same dispersivity of the medium). Rather than travel times for an individual particle they should be regarded as average travel times for a large number of particles released from the same point and at the same time. From a chemical point of view the flowpath of a fluid packet exerts an important control on the fluid's reactive history. As the fluid packet travels through the flow system, it experiences continuous compositional changes as physical conditions (temperature and pressure) and the composition of the rock vary along the flowpath. In addition to tracking solute particles, we calculate the travel times of these particles and a parameter which expresses the rate of temperature change that a fluid packet experiences as it moves through the flow system. Travel times quantify the period of time a fluid packet experiences certain physical and chemical conditions which may explain some of the chemical characteristics of the fluid or the rock. For instance, the understanding of the 80 flow geometry and residence times may enable us to constrain the sedimentary and basaltic fingerprints on the composition of endmember hydrothermal fluids. The parameter that expresses the rate of temperature change that a fluid packet undergoes while traveling along the flowpath, combines the temperature gradient with the fluid velocity. Denoting this parameter A, its mathematical expression is A = ATv, (2.8) where AT is the temperature gradient (per meter flowpath) and v is the average linear pore velocity. Its physical meaning is the temporal rate of temperature change that a fluid packet undergoes while traveling one meter along the flowpath. Thus a high fluid velocity and a low temperature gradient would be equivalent to a high temperature gradient and a low velocity with respect to the parameter A . As chemical reaction rates are strongly controlled by the temperature, the rate of temperature change is correlated with the rate and the intensity of mass transfer between the solid and the fluid phase. Thus we may be able to identify intervals along the fluids flowpath where a fluid packet and the solid phase are likely to experience major compositional changes. In the first set of calculations, flowpaths, travel times, and the rate of temperature change for a full transient simulation of the convective system are determined. Particles enter the system at three different release points at the seafloor; one at the recharge zone, the second at the center of the domain, the third near the discharge zone. According to these release points, travel times and temperature changes are plotted with the suffices a, b, and c, respectively. Particle travel times are calculated and plotted at the same locations as the temporal rates of temperature change in Figures 2.7 and 2.8, respectively. 81 At a basement permeability of lx lO\" 1 4 m 2 (middle panel, Figure 2.7) the particle that is released in the recharge zone (particle a) reaches the basement at about 310 years and exits the system after 1410 years. As this particle travels ahead of the advancing thermal plume, it enters the basement before the flow system has stabilized. Thus it passes through the basement when the flow field is highly transient and characterized by low aspect-ratio convection cells. The particle moves around these convection cells in a wave-like fashion. Between about 977 and 1154 years it travels on a near-circular, closed flowpath within a convection cell for a single loop before exiting the basement through the discharge zone. Particles traveling through the low-permeability sediment section (particles b and c) reach the basement only after more than 7000 and 10000 years, respectively. By that time, the flow system has stabilized toward a single convection cell flow geometry but has not reached a steady state. When these particles enter the basement, they travel in a narrow zone along the outer margin of the convection cell. Particle b exits the system through the discharge zone after a single pass while particle c becomes trapped in the outermost region of the convection cell and recirculates through the basement. Note that the trapping of particle c indicates that the system has not yet acquired a steady state. Particle c passes through the lower 50 m of the sediments between x = 1800 m - 1950 m, that is, above the region of fluid upflow in the basement. This implies that this area in the sediments is part of the large convection cell. The travel time through the basement for particle b is of the order of 1000 years and is thus somewhat shorter than the travel time through the basement of particle a. However, the distance traveled by particle a is significantly longer and its velocity therefore has to be higher. On its way from the 82 recharge zone to the discharge zone, particle a passes more than 10 times vertically through the basement. Thus the total vertical distance traveled is more than twice as long as the total horizontal distance traveled. The rate of temperature change per meter flowpath for a basement permeability of l x l O \" 1 4 m 2 (middle panel, Figure 2.8) shows that a fluid packet entering the domain through the recharge zone experiences very large temperature changes as it moves from the cool seafloor, ahead of the thermal front, into the basement. O n reaching the basement, the temperature change decreases by about two orders of magnitude. Within the basement the values vary within one order of magnitude. The magnitude of the values appears to be primarily dependent on the location of the particle with respect to the convection cell. The lowest values occur in the region where the particle undergoes a change in the flow direction from dominantly vertical to horizontal flow or vice versa. The rather uniform values in the basement are a reflection of the initial vigorous convection and the relatively high degree of thermal homogenization. The rate of temperature change increases dramatically as the fluid ascends through the discharge zone. Near the seafloor this increase is in part a reflection of the fixed-temperature boundary condition at the seafloor. Due to the low flow rates through the sediments, the rate of temperature change is low. The rates increase by roughly one order of magnitude as soon as the fluid packet enters the basement. It remains rather uniform in the regions of horizontal flow that is along the sediment/basement interface and the bottom boundary. O n flowpath c the rate of temperature change reaches a minimum where the fluid turns from horizontal into 83 downward flow. In contrast, flowpath b, which is located within the outer regions of the convection cell where flow velocities are higher, and where it approaches the recharge zone where relatively cool conditions prevail, retains a larger rate of temperature change. Overall, comparing the rates of temperature change within the basement for the three flowpaths a, b, and c, the values are quite similar. Considering the fact that particle a moves through the basement during the high temperature state of the system, the more homogeneous initial temperature distribution and the resulting low temperature gradients are probably the reason for values similar to particles b and c. At a basement permeability of 1 x 10\" m the particle released at the recharge zone reaches the basement after about 70 years (upper panel, Figure 2.7). When it enters the basement it becomes trapped in highly unstable convection cells near the recharge zone. The flowpath is erratic and the travel times indicate very high flow velocities, suggesting thorough fluid mixing during this early state of flow. As time progresses the flow field tends to stabilize by forming a single convection cell. The particle travels along the outer regions of this growing cell and does not exit the system through the discharge zone. Thus the flow field stabilizes so rapidly, that the fluids recharging the basement during the early phase of fluid flow become trapped in the basement and do not discharge back into the ocean. Note that the flowpath of particle a was cut off arbitrarily at about 1221 years. Particles that migrate through the sediments reach the basement after more than 10000 years. By then the flow field in the basement has long stabilized toward a single convection cell. Similar to the case where the basement permeability is l x l O \" 1 4 m 2 , 84 particle b moves through the basement in a single loop within the outer regions of the convection cell. Particle c moves laterally out of the discharge zone into the sediments which signifies that the system is not in a steady state because at steady state particles recharging the convective system either through the recharge zone or through the sediment sections migrate through the system in a single pass. This issue will be explored in greater detail below. Particle c traverses the sediments and reenters the basement and the same process repeats itself. The total travel time of particle b and c through the basement is about 200 years and is thus considerably shorter than for a basement permeability of lx lO\" 1 4 m 2 . The rates of temperature change for particle a are lower than for a basement permeability of lx lO\" 1 4 m 2 (upper and middle panel, Figure 2.8). This suggests that the initial vigorous flow does not compensate for the thermal homogenization in the basement to acquire the same rate of temperature change. The rate values are also more uniform than at lower permeabilities, which is consistent with the highly irregular and vigorous flow and the thermal homogenization that is associated with it. Within the recharge zone the temperature changes are initially very low, which may be due to an initial sluggish onset of fluid recharge. Within the sediments the rates of temperature change are less than one order of magnitude lower than those for particle a, despite velocity differences of several orders of magnitude. After crossing the sediment/basement interface into the basement, particles b and c undergo temperature changes that are higher than for the same particles at lower basement permeabilities and rather uniform along the sediment/basement interface and 85 the bottom boundary. Significant temperature changes occur as the particles approach the downflow limb of the convection cell (between x = 0 - 200 m). Once these particles get near the recharge zone, the rate of temperature change increases steeply. The rapid temperature change is a consequence of the proximity of the downflow limb of the convection cell to the recharge zone and the resulting high temperature gradients. At a low basement permeability of lx lO\" 1 5 m 2 , the particle released at the recharge zone reaches the basement at about 2400 years and the discharge zone after more than 5500 years (lower panel, Figure 2.7). These long travel times are in part due to the initial long period of near fluid stagnation (Figure 2.3). Particle a moves through the basement in a wave-like fashion and exits the system through the discharge zone. Compared to the flowpath of particle a at a basement permeability of 1x10\"14 m 2 , the waves have a higher wavelength and a lower amplitude, which is consistent with a higher aspect ratio of the initial convection cells (Figure 2.3). Particles moving through the sediments reach the basement well after 10000 years. Particle b exits the basement after a single loop. Particle c becomes trapped in the single convection cell, which will eventually represent the steady state flow geometry. Compared to higher basement permeabilities, the convection cell is considerably smaller so that the flowpath of particles through the basement, which enter the basement through the sediment/basement interface, is shorter. Despite the shorter flowpath, however, the travel time through the basement of particle b amounts to about 2500 years, which is significantly longer than basement travel times of the same particle at higher basement permeabilities. 86 The rate of temperature change is highest within the discharge zone. Similar to a permeability of l x l O \" 1 3 m 2 , the low rate values in the recharge zone is probably due to sluggish onset of fluid recharge. For particle a which moves through the basement ahead of the thermal front, the temperature changes along the flowpath are slightly larger and tend to exhibit somewhat greater variations than for a basement permeability of l x l O \" 1 4 m 2 . This indicates that the initial temperature gradients are steeper and the effect of thermal homogenization is not as strong as at higher basement permeabilities. The particles moving through the sediments undergo temperature changes at a rate similar to those at higher basement permeabilities. On entering the basement, these rates hardly change, in particular for particle b, which reflects the low permeability contrast between the basement and the sediments. Thus the sediment/basement does not constitute the same sharp thermal and hydrologic boundary as for higher basement permeabilities. Particle c undergoes small variations in the rates of temperature change after becoming trapped in the convection cell. These variations are a function of the particle position within the cell. In summary, particles moving through the recharge zone at t = 0 travel ahead of the thermal front. These particles either move through the basement in a wave-like pattern and exit the system through the discharge zone or become trapped within the evolving flow field and never exit the domain. The latter case was observed only at high basement permeabilities of l x l O \" 1 3 m 2 . The flowpath of particle a reflects the early, highly irregular state of the flow field. At high permeabilities, the flowpath is erratic, suggesting thorough fluid mixing during this early state of flow. At lower permeabilities, the more organized 87 wave-like flowpath indicates discrete zones in which recharging fluids pass through the basement around small, transient convection cells. Individual convection cells appear to be isolated hydrological systems. Recharge through the sediments proceeds at a very slow rate such that the fluids reach the basement after the flow field has stabilized. Particles that cross the sediment/basement interface move within the outer regions of the single stable convection cell through the basement. A l l particles released at the center of the domain exit the system after a single loop. Those released near the discharge zone become trapped in the outermost regions of the cell and are subject to repeated flow through the basement and the sediments. The parameter expressing the rate of temperature change along the flowpath indicates that the rates are highest in the discharge zones. Values for the recharge zone were calculated only for particle a, immediately after the onset of fluid motion. Therefore these values are subject to the sluggish initiation of fluid recharge and the values should not be considered representative of conditions in the recharge zone. The lowest rates of temperature change occur in the sediments. The magnitude of these values is similar for all basement permeabilities. During the early phase of fluid flow in the basement, which is reflected by particle a, the rates of temperature change are lowest and show the least variation for high basement permeabilities. This indicates the initial intense fluid and thermal mixing. After the stabilization of the flow field, particles that move through the basement (particles b and c), show the highest rates of temperature change at high 88 basement permeabilities, in particular near the recharge zone and along the bottom boundary. A n important implication of these results is that despite significant initial flow rates at high basement permeabilities, the rate at which the fluid temperature changes is relatively low which may limit the potential for chemical mass transfer and favor a fluid composition that is buffered by the rock composition. Consequently, the overall intensity of alteration may not be correlative to the vigor of fluid flow. This scenario changes after the stabilization of the flow field. Then the high flow velocities in the outer regions of the convection cell lead to faster temperature changes, in particular near the recharge zone, and to a greater potential for chemical mass transfer. In a second set of flowpath calculations, our aim is to investigate the steady state flow geometry for different basement permeabilities. It was discussed earlier that fluid flow in the basement stabilizes to form a single convection cell. After the convection cell has formed, the system continues to cool and the vigor of convection continues to decrease towards a steady state but the geometry of the flow field undergoes only minor changes. The question arises, how do recharging fluids move through the basement once the single convection cell has formed? How extensive is fluid exchange? To resolve these issues we track recharging fluids by placing particles at the seafloor with equal spacing, as well as along the length of the recharge and the discharge zones. The particles released along the recharge and the discharge zone also indicate the extent of lateral flow into the sediments. Particles are tracked after the flow field has 89 reached a steady state, which occurs at different times for the different permeability regimes. The geometry of the flowpaths confirms that at each basement permeability the steady state convective system consists of a single convection cell that is hydrologically isolated (Figure 2.9). At steady state, fluids enter the basement either through the recharge zone or the sediment section, move through the basement along the outer margin of this cell, and exit the system through the discharge zone after a single pass. No exchange of fluids within the interior of the cell occurs. The flow geometry is therefore similar to a single pass fracture-loop model (e.g. Bodvarsson and Lowell, 1972), in which seawater passes through the system only once. The size of the hydrologically-isolated zone is strongly dependent on the basement 15 2 permeability. Whereas at a basement permeability of 1x10\" m the isolated core has a width of not more than 300 m, at a permeability of lx lO\" 1 3 m 2 it extends almost across the entire width of the domain (Figure 2.9). Corresponding to the different size of the isolated core, the size of the region undergoing fluid exchange with recharging fluids is different. At high permeabilities only a narrow zone of fluid recharge into the basement exists, bounded by the core of the convection cell and the domain boundaries. At a 15 2 permeability of 1 x 10\" m roughly two thirds of the basement undergo exchange with fluids recharging through the recharge zone or the sediments. Considering that the total amount of fluid recharge into the basement is relatively insensitive to the basement permeability (Figure 2.6) we can infer that at high permeabilities recharging fluids move through the basement in a confined area but at high velocities. At low permeabilities a 90 larger part of the basement comes in contact with recharging fluids but those fluid volumes are rather low. Figure 2.9 indicates that as fluids ascend through the discharge zone, lateral flow into the sediments occurs. After entering the sediment sequence fluid movement is downward towards the basement in an oblique fashion, resulting in a triangular shape of the region affected by lateral flow. The size of the triangular region is similar for all permeabilities, it has a width of 600 m-800 m at the sediment/basement interface. At basement permeabilities lx lO\" 1 5 m 2 and lx lO\" 1 4 m 2 a small isolated convection cell is present within this triangle. If all recharging fluids pass through the steady state convective system only once then the triangular region of lateral flow surrounding the discharge zone has to be part of the hydrologically isolated convective system, i.e. it does not experience exchange with recharged fluids. Lateral flow is less significant around the recharge zone. The extent and direction of lateral flow is dependent on the basement permeability. Whereas at a low permeability of 1x10\"15 m 2 fluid flow from the recharge zone into the sediments affects a triangular region measuring about 400 m in width at the sediment/basement interface, at a permeability of 1x10\"13 m 2 fluid flow is reversed, that is, from the sediments towards the recharge zone. In general, fluids in the sediments tend to move towards the downflow limb of the convection cell in the basement. As the basement region affected by fluid recharge is narrow at higher permeabilities, the downflow limb is located closer to the recharge zone. Thus there is a stronger tendency of lateral fluid motion towards the recharge zone in the sediments at higher basement permeabilities. 91 The hydrological isolation of a large part of the convective system implies also its chemical isolation, i.e. it can be regarded as a closed chemical system without fluid exchange or fluid mixing. The fluids trapped in the cell experience minor temperature changes and to some extent compositional changes of the rock as small fluid volumes pass through both the basalt and the sediments. The important geochemical implication of lateral flow of hydrothermal fluids from the discharge zone into the sediments is that these fluids can be expected to be distinct from the sediment pore waters in terms of composition and temperature and in strong disequilibrium with the mineralogy of the sediments. Consequently, the flow of hydrothermal fluids into the sediments will cause pore water anomalies and distinct mineral alteration patterns. Alteration reactions should be most significant near the seafloor and along the outer margin of the triangular zone due to high temperature gradients and the possible interaction with downflowing seawater/sediment pore water. Note that the triangular shape of the zone affected by lateral flow is consistent with alteration patterns observed at Middle Valley (Site 858, Figure 1.6). 2.2.5 Water/rock Ratios (time-integrated fluxes) Water/rock ratios (w/r ratios) are an expression for time-integrated fluid mass fluxes through a unit mass of rock. Estimates of w/r ratios in natural systems are typically based on chemical characteristics such as certain \"soluble elements\" in the rock that are quantitatively leached into the fluid (von Damm, 1995). Thus w/r ratios are useful for 92 providing a link between hydrological and chemical properties of the system. This link will be further explored in the next two chapters. In this section, the focus is on the pattern and distribution of time-integrated fluid fluxes, which are expressed as the total fluid mass flux per unit volume of rock. The advantage of a mass/volume ratio is its independence on the chemical composition of the rock. To obtain a mass/mass ratio, which would be the water/rock ratio in its proper sense, one would have to be divide the mass/volume ratios by the rock's density. Water/rock ratios (mass/volume) are calculated as snapshots at various times and are superimposed on the velocity fields at these same times. Contrary to the instantaneous view of the velocity field, w/r ratios represent the integrated effects of flow conditions. A comparison of both the velocity and the w/r ratio field demonstrates to what degree the current flow (and temperature) field is out of equilibrium with the past flow (and temperature) history. If the disequilibrium is significant, we can expect that there may also be a disequilibrium between the rock composition, which may have preserved the history of physical and chemical conditions, and the current conditions. Figure 2.10 illustrates the evolution of w/r ratios for a basement permeability of l x l O \" 1 4 m 2 . It shows that during the early, highly-irregular state of flow, the distribution of w/r ratios reflects the overall flow geometry in that highest w/r ratios occur in regions of vertical flow, i.e. within the limbs of the convection cell and low w/r ratios occur at the points of transition from vertical to horizontal flow (500 years - Figure 2.10). The flow pattern leads to vertically oriented, elongated patch-like zones of higher w/r ratios. The unstable flow field and the vigorous convection, however, do not lead to significant gradients in w/r ratios. The scenario changes when cooler seawater enters the basement and a stable 93 convection cell begins to form adjacent to the recharge zone (2000 years - Figure 2.10). The initial, patchy distribution of w/r ratios becomes overprinted by this growing convection cell. This process occurs particularly fast in the outer regions of the convection cell where the highest fluid velocities are present. A s fluid velocities decrease toward the interior of the cell, the initial pattern of w/r ratios tends to be preserved. This means that the distribution of w/r ratios is out of equilibrium with the current velocity distribution and fingerprints of the early-state, highly vigorous convection are still detectable at much later times, especially in regions that experience only low fluid fluxes during stable fluid flow. At 20000 years the stable convection pattern has overprinted most of the traces of the initial flow geometry, except for some small patches in the center of the cell (Figure 2.10). The geometry of the w/r ratio contours is now more in accordance with the current velocity field. The velocity differences between the inner and outer regions of the cell leads to increasingly steeper w/r ratios gradients as time progresses. The distinctly higher w/r ratios below the discharge and recharge zones should be detectable in the rock's composition, in particular below the recharge zone where cool fluids with a strong chemical seawater component enter the basement in a strong disequilibrium with the rock. The result of such focussed flow will be localized zones of strong chemical alteration. In the sediment layer water/rock ratios have a range which is roughly two orders of magnitude lower than in the basalt (Figure 2.10). Small volumes of lateral flow from the discharge zone into the sediments leads to lower values near the discharge zone than in the central parts of the domain. This suggests that lateral fluxes into the sediments under these conditions are small. A small, isolated convection cell in the bottom right 94 corner of the sediments causes a circular arrangement of the contours and relatively steep gradients. 15 2 At a lower basement permeabilities of 1 x 10\" m and at higher basement permeabilities of lx lO\" 1 3 m 2 , the same phases during the evolution of the hydrothermal system can be recognized. However, as with the evolution of the flow field, the duration and timing of these phases are different. In addition, as the geometry of flow changes with the basement permeability, so does the arrangement of the w/r ratio contours. At a low permeability of 1x10\"15 m 2 , the early flow field that consists of convection cells of near unit aspect ratio (Figures 2.3 and 2.7), leads to almost circular zones of high and low w/r ratios at the limbs and the center of the convection cells, respectively (Figure 2.11). Due to the sluggish evolution of the flow field and the extended transient state, the velocity and the w/r ratio distribution are still in disequilibrium after 20000 years (Figure 2.11). For instance, even though the flow field has stabilized by then, the center of a convection cell shows high w/r ratios, a result of an upflow zone that existed earlier in this region (at around 10000 years, x = 1500 m). Overall, w/r ratios in the basement are relatively uniform and only one order of magnitude higher than those in the sediments. The much wider basement region affected by fluid recharge in addition to the initial sluggish flow causes only moderate w/r ratio gradients below the recharge and the discharge zones. The stable fluid circulation leads to a zone of near fluid stagnation at the sediment/basement interface, adjacent to the recharge zone. At a permeability of 1x10\"13 m 2 the patchy arrangement of the contours during the early state is small-scale and highly disorganized (Figure 2.12). Nevertheless, even at 95 high permeabilities an overall vertical orientation of these patches is present. Despite the rapid stabilization of the flow field, fingerprints of the early, highly vigorous convection are recognizable in the interior of the convection cell in form of small patch-like w/r ratio contours at 20000 years (Figure 2.12). For basement permeabilities of l x l O \" 1 4 m 2 and l x l O \" 1 3 m 2 the order of magnitude of the w/r ratios in the interior of the convection cell is similar at 20000 years. The values are of roughly the same order of magnitude as those found throughout the basement at a permeability of l x l O \" 1 5 m 2 at the same point in time. Major differences between w/r ratios for different permeabilities occur and become more significant towards the outer regions of the convection cell. For instance, at 20000 years, the basement with a permeability of l x l O \" 1 3 m 2 exhibits w/r ratios in the outer regions of the convection cell which are more than twice as high as for a permeability of l x l O \" 1 4 m 2 . Thus, at stable convective circulation we find higher w/r ratio gradients at higher permeabilities. In the sediments, the order of magnitude of w/r ratios is about the same for the range of basement permeabilities from 1x10\" m to 1x10\" m . From the w/r ratio values it appears that lateral flow from the discharge zone into the sediment section is more significant at lower basement permeabilities. In conclusion, the evolution of physical w/r ratios undergoes phases in accordance with the evolution of the velocity field. A n initial irregular, patch-like distribution of w/r ratios becomes overprinted by a growing, stable convection cell with high w/r ratios at its margin and low ratios in its interior. The low fluid velocities in the interior of the cell lead to only small changes in w/r ratios and the rocks in this region have been involved in 96 fluid/rock interactions mostly during the early, high temperature state. The central parts of the convection cell are therefore characterized by cooling but relatively little fluid exchange. Thus the possibility exists that mineral alteration products of this early state are still preserved in the rock. Toward the outer regions of the cell, flow velocities increase, and fingerprints of the initial convective pattern are quickly removed. Rocks in this region can be expected to show evidence of much greater mass transfer with the fluid phase. The results indicate that the distribution and the range of w/r ratios at a certain point in time during the evolution of the convective system vary with the permeability of the basement. The question arises if there is a single parameter that can express the dependency of w/r ratios on the basement permeability. One possibility is to calculate average w/r ratios for the entire basement section. Average w/r ratios versus time for the range of basement permeabilities are shown in Figure 2.13. The figure contains two sets of curves, one for a transient simulation and the other for a steady state flow simulation. A l l transient curves show the same trend, an initial steep increase in w/r ratios that later approaches the slope of the linear steady state curve for the same basement permeability. The steep initial slope is consistent with the initial, highly vigorous convection. The higher the basement permeability the steeper the initial increase. A comparison of transient and steady state curves for the same permeability shows that due to the initial vigorous convection that average w/r ratios are higher if the transient state of the flow field is accounted for. The difference between the transient and steady state curves increases with the basement permeability. Taking into account the shorter transient state at higher permeabilities, the fluid volumes passing through the system immediately after 97 the onset of fluid movement are indeed significant. Thus the fingerprints of the early-state irregular fluid flow or the deviations of w/r ratios from the current velocity field should become particularly obvious near the center of a convection cell at high basement permeabilities, that is in regions that undergo a large decrease in fluid flow rates as the system evolves. Indications of initially vigorous flow can for instance consist of chemical alteration reactions, which are sensitive to w/r ratios. Or taking into account the high temperature conditions in addition to the vigorous flow during the early state, one would expect the alteration to be generally much more intense and/or farther from equilibrium with the current physical and chemical conditions in the pore water To test whether there exists a correlation between average w/r ratios and the basement permeability, we plot average w/r ratios versus permeability at four different times of the transient simulation (1000, 5000, 10000, and 20000 years) as well as 20000 years of the steady state flow field on a log-log plot (Figure 2.14). Both the steady state values of the transient flow simulation (here represented as 'near' steady state after 20000 years) and the steady state-only simulation plot as a straight line on a log-log plot. The curves showing snapshots of average w/r ratios during the transient state indicate that as time progresses the curves become increasingly log-log linear, approaching the slope of the steady-state curve. Consistent with the sluggish evolution of the flow field, at low basement permeabilities average w/r ratios need a longer time to fall on the log-log linear w/r ratio/permeability curve. At 20000 years the slopes of the transient and the steady state simulation are quite similar. The transient curve, however, is offset towards higher w/r ratio values which is a consequence of the initial high flow rates. 98 The log-log linear correlation between average w/r ratios and the basement permeability at steady state is a consequence of the linear correlation between the fluid velocity and the permeability of the medium on the one hand, and the linear increase of w/r ratios with time on the other hand. If the fluid velocity was proportional to the permeability, the slope of the curve would be equal to one. The slope of around 0.4 indicates that when the basement permeability is increased, the average velocity, and thus the average w/r ratios, increases at a slower rate. The exact nature of this correlation remains to be investigated. Nevertheless, average w/r ratios may be a useful way of constraining bulk permeabilities. 2.2.6 Fluid Mixing During the transient state of a flow system advective fluid mixing occurs. In other words, particle flowpaths may crosscut each other when looking at the flow field from a time-integrated perspective. Where fluids of different temperature or composition mix a chemical disequilibrium is induced and a mass transfer between the solid and fluid phase may take place. The transient particle tracking calculations suggest that fluid mixing is likely to be most intense at high permeabilities when fluid flow is irregular and vigorous. It was shown that the transient state has little impact on the particle flowpaths through the low-permeability sediments, so that fluid mixing is not an important process within the sediments. As the flow system approaches steady state and fluids tend to travel within stationary streamtubes (Figures 2.9) advective mixing becomes less intense. Nevertheless, 99 mass transfer between streamtubes still occurs in the form of lateral dispersion. In comparison to advective mixing, however, the effect of lateral dispersion is undoubtably much smaller. In this chapter we assess the significance and nature of mixing (advective, dispersive, and diffusive) and evaluate the transport-controlled distribution of fluid compositions for different basement permeabilities. For this purpose the output from the transient flow simulation is coupled with the transport code OS3D (Steefel and Yabusaki, 1996). For details of the mathematical and numerical formulation of the transport calculations the reader is referred to the next chapter. In the simulations described in this chapter, no chemical fluid/rock interaction is taken into account. The calculations are essentially a series of tracer tests in which we specify a constant source concentration of a non-reactive species at certain locations within the domain and measure the fluid composition with respect to these tracers at various locations through time. The locations of the tracer release points and the 'sampling points' are shown in Figure 2.15. The release points are chosen such that a range of fluids of different origins can be traced. These fluids include sediment pore water from just above the sediment/basement interface, fluids recharging the basement through the recharge zone, fluids released near the seafloor, as well as fluids from different locations within the basement. Each source releases a tracer at a fixed concentration of 100 moles/kg. The tracers released at the various sources are labeled according to Figure 2.15. The initial concentration of all tracers throughout the domain is zero. It has to be emphasized that because the concentration is fixed at the release point, the flow rate at this point determines the mass of the tracer that is released into the system. The sampling locations are located at various 100 points in the basement, as well as within the discharge zone (Figure 2.15). The results are presented as tracer breakthrough curves at these locations. Figures 2.16 to 2.19 show tracer breakthrough curves for a basement permeability of lx lO\" 1 4 m 2 at sampling points X l - 2 , X3-4, Y l - 2 and Y3-4, respectively. At first glance the 'spiky' appearance of the curves is apparent which indicates that despite the stabilization of the flow field with the formation of a single convection cell at about 2500 years, there is significant variation in the fluid composition at these various sampling locations. The tracers tend to pass through these locations in pulses that may or may not reappear. These pulses are initially sharp and pronounced but as time progresses the pulses spread and concentration changes become less significant (e.g. tracer 1A at X4, 4A at X2). This spreading of the tracer pulses reflects the transition to a stable flow pattern and may also be attributed to the effects of dispersion. The latter is indicated by the dilution of tracers at various sampling locations (e.g. tracer 4A at location Y2 or tracer 4E at location X3). As a result, when the breakthrough pattern changes from erratic to more uniform after the flow field has stabilized, the fluid composition tends to be dominated by one or two tracers, typically those that are released on or near the same flowpath as the sampling point is located on. Other tracers appear in low background concentrations. The sequence of tracers passing through a location is not always controlled by the distance to the release point. In some cases tracers that are released close to the sampling point are never sampled. At X I , for instance, the number and concentration levels of basement tracers are very low despite the fact that release points are not far away. Other 101 tracers breakthrough the sampling location only after the flow field has stabilized (e.g. tracer 4A at Y l , first appears after 5000 years). Tracer 1 A, the tracer released at the recharge zone, is sampled at each location and occurs at each location in significant amounts. This means that compared to other tracers, it is more evenly distributed throughout the basement. It may also mean that more mass of tracer 1A is brought into the system due to high flow velocities in the recharge zone. Unlike tracers that are released in the basement, the time of the first breakthrough of tracer 1A at the various sampling locations correlates with the distance of the location to the recharge zone. Both observations are consistent with the early flow pattern as it is illustrated in Figures 2.2 and 2.7. A tracer released at the recharge zone moves through the basement in a wave-like fashion and spreads throughout most of the basement. Large portions of the flow field in the basement consist of isolated convection cells and a tracer released in one of those cells would be trapped and could not undergo the same degree of mixing as the tracer released at the recharge zone. Tracers that are released in the sediments are insignificant in basement fluids throughout most of the basement. Only at sampling location X I do sediment tracers from the bottom of the sediments occur in measurable amounts. The first breakthrough of the sediment tracers at X I coincides with the stabilization of the flow field. The sequence at which the tracers first appear at the sampling point in measurable amounts as well as the concentrations are proportional to the distance between release point and sampling point. After the first breakthrough, the sediment tracers maintain a nearly uniform concentration. An interesting observation is that after 4000 years, the sediment tracers are 102 the dominant tracers in solution at X I , as all basement tracers have decreased to even lower concentrations. The implication of this is that when the flow geometry is stable, regions in the basement exist that are distinct from the rest of the basement in terms of pore fluid composition, that is in this case the sedimentary component in the pore waters. Breakthrough curves for the same set of sampling locations were calculated for a basement permeability of lx lO\" 1 5 m 2 and lx lO\" 1 3 m 2 and are shown in Figures 2.20-23 and Figures 2.24-2.27, respectively. At the lower basement permeability, only sampling locations Y l - 4 are located in the region in which the single stable convection cell evolves. Sampling locations XI-4 are located in the region that is affected by fluid recharge through the recharge zone and the sediments (Figure 2.9). As a consequence, the breakthrough curves on each side of the domain show a different pattern. At sites XI-4, the breakthrough of tracers 1A and 3 A, which are released in and adjacent to the recharge zone, respectively, occurs as a front that is a more or less steep increase and a uniform concentration after that. Unlike the pulse-like breakthrough of the basement tracers at a permeability of lx lO\" 1 4 m 2 , the breakthrough of these tracers resembles that of a continuous plume. The first appearance of the tracers 1A and 3 A occurs earlier at the bottom (X4) and later in the upper region (XI) of the basement. The sediment tracer 3 A makes up a significant component in the fluids sampled at X2 and X3 , reaching about 20% of the total tracer concentration in solution. Preceding the breakthrough of the tracers 1A and 3 A are periods of pulse-like breakthroughs of various tracers, but not with the variability and the diversity of tracers as at a permeability of 1 x 10\"14 m 2 . These periods of tracer appearances may be interrupted by 103 periods during which no tracer is sampled at all (e.g. between 6500 and 10000 years, at XI ) . For all locations in the basement, no tracer is sampled during the first 2000 years of simulation time, which is consistent with a long initial period of fluid stagnation. At sampling locations Y l - 4 , the pattern of tracer breakthrough curves has a greater resemblance to that of a permeability of lx lO\" 1 4 m 2 . The tracers appear in pulses and exhibit considerable concentration changes throughout most of the simulation time. The amplitude of these pulses is significantly larger than those at a basement permeability of lx lO\" 1 4 m 2 . Only between 12000 and 15000 years is a transition towards smoother breakthrough curves and decreasing tracer concentrations recognizable. As discussed earlier, this transition is a result of the formation of a single convection cell probably in combination with the spreading and dilution effects of dispersion. The number of tracers that are sampled at these locations Y l - 4 is initially low but it increases after the transition towards more uniform breakthrough curves. This diversity of tracers is in part due to sediment tracers that occur at these locations in measurable amounts and indicate a small but recognizable sedimentary signature in the basement fluids. At a high basement permeability of lx lO\" 1 3 m 2 (Figure 2.24-27) the transition from a pulse-like tracer breakthrough to a smooth, steady breakthrough curve with decreasing or in some cases increasing tracer concentrations occurs much earlier. The initial period of erratic tracer breakthrough comprises around 1000 years. From then on the composition of the fluid tends to be dominated by one or two tracers, typically those that are released on the same flowpath as the sampling point is located on, and a large number of tracers that constitute a low background concentration. A n interesting 104 observation is the similarity of the breakthrough curves on both sides of the domain, i.e. between locations XI-4 and Y l - 4 as well as between X2 and X4 and Y2 and Y4. This observation is consistent with the stable flow geometry of a single convection cell, that, at permeabilities this high, essentially fills the basement and is thus symmetrical. Of particular interest is the composition of vent fluid with respect to its content of basement fluids, sediment pore water, and seawater recharging through the recharge zone. The pattern of steady state flowlines in Figure 2.9 suggests that the fluids that recharge the hydrothermal system pass through the system in a single pass. Therefore one would expect to find a relatively strong signature of recharged seawater (tracer 1 A) in the vent fluids. To test this hypothesis we place one sampling location within the discharge zone (sampling point Z l ) . The breakthrough curves at this location for the range of basement permeabilities are shown in Figure 2.28. The first breakthrough of tracer 1A occurs at 15 2 different times, at 5500 years, 3000 years, and 500 years for a permeability of 1x10\" m , lx lO\" 1 4 m 2 , and lx lO\" 1 3 m 2 , respectively. Whereas at a permeability of lx lO\" 1 5 m 2 the breakthrough curve shows some fluctuations after the first appearance, at higher permeabilities the concentration remains roughly uniform. The first breakthrough at higher permeabilities approximately coincides with the formation of a stable convective pattern in the basement. The fluctuations at a permeability of lx lO\" 1 5 m 2 indicate that the flow field has not yet fully stabilized. The concentrations of tracer 1A are relatively similar for permeabilities lx lO\" 1 4 m 2 and lx lO\" 1 5 m 2 and somewhat lower for a permeability of lx lO\" 1 3 m 2 . Overall, for the range of basement permeabilities, the concentration of tracer 1A amounts to about one third or less of its release concentration. 105 Most of this dilution is attributable to dispersion during the passage of the tracer through the system. Another major component in the vent fluids is tracer 4F, which is released about midway into the basalts, below the discharge conduit through the sediments. Its concentration in the vent fluids increases with increasing basement permeability. This correlation is a result of the different flow rates at release point 4F, which are higher for higher basement permeabilities, in combination with a more focussed upflow at higher basement permeabilities (Figure 2.9). A n interesting result is that the number of tracers passing through the discharge zone and thus the degree of mixing is larger for a low basement permeability. Even tracers released in the sediments, which are present only in traces in the vent fluids at 14 2 13 2 basement permeabilities 1x10\" m and 1x10\" m , occur in significant concentrations in 15 2 the vent fluids at a basement permeability of 1 x 10\" m . Thus it appears that due to the extensive transient state at low basement permeabilities on the one hand, and the early hydrologic isolation of a large part of the basement in combination with a more pronounced effect of dispersion at higher permeabilities on the other hand, vent fluids at low basement permeabilities have a more heterogeneous composition. In summary, the pattern of the breakthrough curves suggests that during the early, unstable flow regime the tracers released in the basement become distributed in an irregular, disordered pattern. The erratic distribution throughout the basement causes pulse-like breakthrough curves of individual basement tracers. As time progresses and the flow field evolves to a single convection cell, these tracers become more evenly distributed and diluted. The dilution reflects the effect of dispersion. Thus, the general 106 trend is that when the stable flow field and the hydrologic isolation of a large part of the basement is established the fluid composition at a point within the domain becomes less affected by the initial state of large-scale fluid mixing and more by the conditions along the flowpath on which the sampling point is located. Overall, consistent with the evolution of the flow field, at high basement permeabilities, mixing of basement fluids occurs over a short period of time and the condition of a 'well-mixed' fluid occurs earlier. The early formation of a single convection cell and the resulting hydrologic isolation of a large part of the basement leads to a rapid dilution of background tracer concentrations and a pore fluid composition that is dominated by tracers that are released near or on the same flowpath on which the sampling location is located. At lower basement permeabilities, the initial mixing extends over a longer period of time. As a consequence, the pore fluids contain a heterogeneous composition, that is, significant amounts of various background tracers over a longer period of time. At a permeability of lx lO\" 1 5 m 2 , this includes a considerable amount of the tracers released in the sediments. 2.3 Conclusion In this chapter we investigated aspects of the evolution of a hydrothermal system in terms of processes that may effect the chemical characteristics of the fluid and the rock. It is important to see the composition of the rock as the integrated effects of changing physical (temperature and pressure) and chemical (fluid composition) conditions, and the 107 composition of the fluid as the result of continuous water/rock interaction along the fluids flowpath. For this reason, in order to interpret and make use of chemical data, information on the history of fluid flow and temperature conditions in the hydrothermal system must be present. In our simulations, the history of fluid flow and temperature begins with the disturbance of the initial conductive, no-flow state and ends with a stable convection pattern. In general, the duration of the transient state increases with decreasing basement permeability. For all basement permeabilities the transient state can be divided into two phases; one during which the flow pattern is variable, irregular, and dominated by small convection cells, and the other phase during which the flow field has stabilized in form of a single large convection cell. Once the stable flow pattern is obtained, temperatures and the overall vigor of convection decrease toward a steady state value. The transition of these phases occurs gradually as the continuous recharge of cold seawater leads to the evolution and the slow expansion of a thermal plume in the basement. At the same rate as the plume expands, a single convection cell evolves which eventually extends from the recharge zone to the discharge zone. Both phases are shorter and the initial temperatures and the vigor of convection are higher for higher basement permeabilities. At high permeabilites of lx lO\" 1 3 m 2 the initial vigorous flow leads to nearly thermally homogeneous conditions in the basement. In contrast, at lower permeabilities of lx lO\" 1 5 m 2 the first phase is characterized by a long initial period of fluid stagnation. Despite vigorous fluid flow at high basement permeabilities, the thermal homogenization leads to low temperature changes of the fluid and thus a greater likelihood that the fluid composition is buffered by the rock composition. 108 With the onset of a stable flow geometry in the second phase, the convective system has become similar to a single pass fracture loop, in which fluids that recharge the convective basement exit it after only a single pass. Thus only a relatively small region in the basement is affected by interaction between recharged seawater and the rock. The bulk volume of the basement, as well as the sediments surrounding the discharge zone are part of a hydrologically isolated system, in which fluids rotate around the axis of the convection cell. This hydrologically isolated system is also chemically closed, i.e. no mass enters or leaves the system and fluids trapped in the convection cell only undergo small temperature changes. Fluid packets travelling in the outer regions of the cell experience compositional changes of the rock as they may pass through the sediments as well as the basalt. Lateral flow of hydrothermal fluids into the sediments along the discharge conduit is likely to result in distinct alteration reactions. However, as the later fluxes into the sediments are relatively low, the intensity of alteration is probably strongly dependent on the age of the system. The irregular flow geometry during the first phase is reflected in fluctuations of recharge and discharge rates into and out of the basement as well as periodic variations in discharge temperatures. Only with the beginning of the second phase does an overall decline of both recharge and discharge rates and discharge temperatures set in. The decline is steep at first and then the values slowly approach a steady state condition. Fluxes through the sediment/basement interface may have an effect on the thermal evolution of the discharging fluids. ' 109 During the initial vigorous convection, water/rock ratios increase rapidly. The distribution of w/r ratios reflects the pattern of the initial flow geometry, for example vertical zones of high water/rock ratios during low-aspect ratio convection at a basement permeability of lx lO\" 1 4 m 2 . The evolution of the distribution of w/r ratios occurs at a much slower pace than the evolution of the flow field. As the single convection cell evolves, the early pattern of w/r ratios becomes slowly overprinted. Highest flow velocities now occur in the outer regions of the cell and as a consequence w/r ratios at the sediment/basement interface, the bottom boundary, and the limbs of the cell continue to increase rapidly. In the cell's interior, in contrast, near stagnant conditions lead to w/r ratio values almost unchanged from the initial values. Thus signatures of the early state of fluid flow may still be preserved in the rock in the form of alteration products. At steady state, average basement w/r ratios show a log-log linear relationship with the basement permeability. The same log-log linear relationship between average basement w/r ratios and the basement permeability is approached i f the transient state is accounted for. The position of the transient curve, however, is offset towards higher average w/r ratios relative to the steady state curve as a result of the initial vigorous state. The significance of the transient state in terms of total flow rates increases with the basement permeability despite a shorter duration of this state. A tracer test demonstrated that long after the flow field has stabilized, different tracer contents can be sampled at different locations. It shows that during the initial irregular state of flow the basement tracers are distributed in a heterogeneous fashion throughout the basement. When the flow field has stabilized, these tracer pulses spread 110 and become diluted due to dispersion until these tracers appear only as uniform or slowly decreasing background concentration in the pore waters. The composition of the sampled waters is then dominated by tracers that are released on or near the flowpath on which the sampling location is situated. Consistent with the duration of the transient state, the dilution of the initial heterogeneous composition of pore waters occurs faster at higher basement permeabilities. At low permeabilities the effect of dilution is low enough such that sediment pore waters become a significant component in basement pore fluids as well as vent fluids. In summary, it was demonstrated that the evolution of a hydrothermal system shows similar trends or phases for the range of basement permeabilities used in these simulations. A promising approach to constrain the vigor of convection or the permeability of the convective layer may be the correlation between the bulk permeability and the average w/r ratios and the possibility to use geochemical information to estimate w/r ratios. Implications of the evolution of the flow and temperature fields on a simple geochemical system will be assessed in the next chapter. The results are then used to develop and test a method to derive w/r ratios and use them to constrain flow velocities in chapter 4. i l l Permeability (m 2) Porosity Density (kg/m 3) Dispersivity Thermal Conductivity (W/m°C) Heat Capacity (J/kg m°C) Sediments k x = 5 x l 0 \" 1 6 k y = l x l 0 \" 1 7 10% 2000 OCL = 10 m ax = 1 m 1.728 1000 Basalt k = l x l O \" 1 5 l x l O \" 1 3 10% 2000 a L = 10 m a j = 1 m 1.728 1000 Windows k=7 .5x l0\" 1 4 10% 2000 a L — 10 m OLJ = 1 m 1.728 1000 Table 2.1.: Summary of physical and hydrological properties of the sediments, the basalt, and the recharge/discharge zones. 112 Temperature: 1°C Pressure: 250 bars recharge zone discharge zone k = 7.5e-14m S E D I M E N T S k = 5e-16 m k = l e - 1 7 m k-=7.5e-14m B A S A L T / P E R M E A B L E BAS1 .MI N T k - le-15 in - le-13 in 20 m Conduct ive Layer 2000 m — O Heat Flow: 850 mW/m Initial Conditions: Conductive Temperature Field No Flow Figure 2-1. Model flow domain, including initial and boundary conditions, and permeabilities. 113 114 115 0 1 0 0 0 2 0 0 0 X(m) Figure 2-4. Evolution of temperature and velocity fields - basement permeability lxlO\" 1 3 m2. 116 0 0 5000 . 10000 15000 20000 Time (years) le-15 i»2 5e-lS m 2 • recharge zone recharge zone discharge zone discharge zone basement - basement le-14 m 5e-14 m 2 recharge zone recharge zone discharge zone discharge zone basement basement le-13 m2 recharge zone discharge zone basement Figure 2-5. Temperature evolution at the recharge and discharge zones and evolution of average basement temperatures. 117 Time (years) l e - 1 5 m recharge zone interface discharge zone l e - 1 4 m* recharge zone — - • - interface discharge zone 2 l e - 1 3 m recharge zone interface discharge zone 5 e - 1 5 m recharge zone interface — „ discharge zone 5 e - 1 4 m 2 recharge zone interface discharge zone Figure 2-6. Fluid mass fluxes (kg/year) through the recharge and discharge zones and across the sediment/basement interface. 118 52-a le-13m2 4194-c lSpO-b ,32J4,Q-c 10S20-b / 62550-c Q ... .,*1 4 « « k 'V--...0*>?a '*'*•.. / * ' •** \\ \\ \"ik v**»..C3?# | f 1 N , i \\ 1 f - a f\\ \"' \\ \\ .> -:J|sk?2-a,f*;* L/ 1 \\ ..1020-a C: V v ^ ^ J l ^ f e ^ ^ ^ S ^ M ^ t e ^ 9 ^ ^ - - m i i / _ [ QQQ- ^^^»?»..^.„..».2fli(.MUc—f.f.j2lf-f —L522U-b.(.4i}f!{lr - .2ttoijli-e-i52&tUi.. ^ ^-Wt- 64JJIkc 0 1000 2000 X(m) 1 2M90-1 15%90-b 2?8-a \"S Q 3. t= -500 o.le-14m2 6f95-b 14J10-a -1000-....... 1 , 0 5 - a °\"aI'llk-c ( \\ \\ \\ 1 / N M * 4I9-a 2350Q-c\\ \"~92?7-b 1000 X(m) 2000 1208-a % -500 Q -1000 le-15m2 9569-b 9454-c 18f80-b 16340-b -.USfiO-c ^ 4446-a 140Qft>-^—~-S4Il4 / X>., I jb*<570-| v W 8 P T « « f e ^ *7Q£hjL^*^ • *** 1000 X(m) 2000 Figure 2-7. Particle flowpaths and traveltimes for basement permeabilities lxlO\"' 3 m2 - lxlO' 1 5 m2. 119 0.02-a •3 -500 a. D 0! 31 le-13m 2 0.03-c 23\\03-b yo.oi,-c 0.01-b / O'.Ol-c Iftp*',:;;-.. ( ( . ( H i a ; ^ : w ^ ^ f e - \" ' - - t t ^ .. 0.0fea. l i i ^ - ^ : : : ^ , . / • • ^.r-^rn^:^^^ „. ni1 t a S \\ ° - . 8 5 - a r \\ Qisk. „_'H.f 6 « % 0 7 - a S . / | \\ ...O-U-a , p V V ^ I « ^ i -1000-1000 X(m) r v 2000 69 53-a -500-6.E5-a le-14m 2 79(75-i 0.0'4-c 0.03-b -1000-kili'c / 0.01-a / l \\ \\ | / \\ / / 0.}9-a 1 ° -« f°A / / I i : 5 0.36-a / f | \\ \\ / / X b.37rV 0.28-a \\ 0^0-a\\ \\ x / \\ ¥).S5-a»* ' ' ' ' ! \" ^ -^ - f \\ 0.38-c\\0.34-c ) . . , , V | _ 0,.37-c '•—•»\"'\" ' N.n.31r,a, . ,— 0.41-C0.6I7-A *6!f£b 0.19-t •5 -500i -1000-1000 X(m) I 2000 le-15m 2 0.02-b 0.62-c 1 V 5 - b 13-a 0-a O.Oj'-o'' s «Ul4-a| / \"•AQSJ) 1000 X(m) 2000 Figure 2-8. Particle flowpaths and rate of temperature change for basement permeabilities 1x10\" l x l 0 \" 1 5 m 2 . 120 121 0 1000 2000 X(m) Figure 2-10. Evolution of water/rock ratios (fluid mass (kgVvolume of rock (m3)) - basement permeability lxl0- ' 4 m 2 . 122 X(m) Figure 2-11. Evolution of water/rock ratios (kg/m3) - basement permeability lxlO\"15 m2. 123 124 1 o > 1.4e+6 1.2e+6 1.0e+6 H 8.0e+5 6.0e+5 4.0e+5 2.0e+5 H 5000 10000 Time (years) 15000 20000 Figure 2-13. Evolution of average basement water/rock ratios (kg/m3) for full transient (black curves) and steady state flow simulation (grey curves) - basement permeabilities lxlO\"13 m 2 lxlO-1 5 m2. 125 le+7 B le+4 -• 2> le+3 : 1000 years - 5000 years 10000 years 20000 years • steady state (20000 years) le+l -| 1 . . . , . r—,—I , , , , , , r - ^ H le-15 le-14 le-Basement Permeability Figure 2-14. Average basement water/rock ratios (kg/m3) versus permeability at various times for full transient and steady state flow simulation (20000 years). 126 1A IB 1C ID 3A \"xiO\" 3B 4A.. ^ 413 X3 P P 2A 3C 4C 3D 41) Y l Q V2o v? Y4\" o z i '3EV.3F.(D £ 4f. 4F o o o 2000 m Release Point O Sampling Point Figure 2-15. Locations and labels of tracer release and sampling points. 127 0 2000 4000 10000 12000 14000 0 2000 Release Points 6000 8000 Time (years) 12000 14000 1 A 1 B 1 C 1 D 2 A 3 A 3 B 3 C 3 D 3 E 3 F — 4 A — - 4 B — - - 4 C — 4 D •— 4 E — - 4F Figure 2-16. Tracer breakthrough curves for sampling locations X I , X2 - basement permeability 1x10\" m 128 129 130 132 2000 4000 10000 12000 70 - j 60 -50 -60 40 -M 30 -20 -10 : 0 -Release Points 6000 8000 T ime (years) 10000 12000 14000 1 A 1 B 1 C 1 D 2 A 3 A — 3 B 3 c • 3 D — 3 E — 3 F 4 A 4 B 4 C 4 D 4 E 4 F Figure 2-21. Tracer breakthrough curves for sampling locations X 3 , X4 - basement permeability 1x10' m . 133 80 Y L — 3 E 4 E 3 F 4 F FIGURE 2-22. TRACER BREAKTHROUGH CURVES FOR SAMPLING LOCATIONS Y L , Y 2 - BASEMENT PERMEABILITY 1X10' m . 135 136 137 le-15m2 oo •g \"3 S 12000 14000 le-14m 2 40 le-13m 70 60 4 50 H 2000 4000 8000 8000 10000 12000 14000 2000 4000 6000 8000 10000 12000 14000 Time (years) Release Points Figure 2-28. Tracer breakthrough curves for the discharge zone (Zl) - basement permeabilities lxlO\"1 m2, lxlO\"14 m2, and lxlO - 1 3 m2. 140 3 Chemical Evolution of a Hydrothermal Convective System 3.1 Introduction Chemical alteration patterns in the oceanic crust reflect the time-integrated effects of fluid flow, temperatures, pressures, fluid residence times, and fluid compositions (Spooner et al. 1977). The composition of the fluid represents the integrated effects of the fluid's reactive history along its flowpath. Reactive conditions change constantly due to changing boundary conditions and/or during the slow evolution of the hydrothermal system toward a steady state. This leads to highly complex alteration patterns that represent the evolutionary history of the system and which may be in disequilibrium with the current temperature conditions and the local fluid composition. Thus a reliable interpretation of chemical data is often not possible without coupling information on fluid flow, temperature, chemical reaction, and time. This chapter is aimed at investigating the chemical processes that affect the basalt and the sediment cover and the fluid's major ion chemistry. The emphasis is on alteration patterns during the evolution of the hydrothermal system and how these patterns may feed back on the evolution of the system as a consequence of changes in porosity. Then the evolving chemical processes and patterns for two different basement permeabilities are compared to assess whether alteration patterns or the nature of alteration can be used for constraining estimates of basement permeabilities. 141 The simulations reported here were carried out using the reactive transport code OS3D (Steefel and Yabusaki, 1996), the structure i f which is outlined in a later section of this chapter. 3.1.1 Chemical Processes and Permeability Changes -Methods and Previous Results In the absence of direct means of observing hydrothermal processes in the oceanic crust, constraints on the nature of fluid circulation must be inferred from indirect evidence of these processes, such as from the examination of the alteration of rock samples, locations and size of sulfide deposits, and pore or vent fluid compositions. The composition and the relative abundance of alteration minerals may provide information on conditions of fluid/rock interactions, such as temperature, pressure, water/rock ratios, rock and fluid composition, or the oxidation state of the system. Water/rock ratios provide a link between the hydrology and the chemistry of a hydrothermal system. The correlation between the average w/r ratios and the permeability of the medium was discussed in the previous chapter. A comparison of chemical water/rock ratio (that is, w/r ratios based on chemical data) and permeability versus depth estimates obtained from various ridges shows the same vertical layering and thus confirms the correlation between the two parameters (Figure 3.1). Crustal alteration therefore appears to be influenced by the permeability distribution. However, as the chemical characteristics of the rocks reflect the time integrated reactive history, one has to 142 take into account that the alteration of rocks is not only a function of the fluid volume passing through it. In addition, the fluid residence time may be too short to establish local equilibrium and/or fluids may have passed through previously altered rock. Thus owing to these uncertainties, water/rock ratio estimates based on chemical information have been regarded as a semi-quantitative analysis which provides values that are equal to or less than actual or physical water/rock ratios (Fisher, 1998). Other chemical characteristics that have been used to constrain the distribution and evolution of permeability in the upper basement include alteration patterns in ophiolites (e.g. Gillis and Robertson, 1988, 1990; Haymon et al.1989; Valsami-Jones and Cann, 1994). Typically ophiolites are characterized by extensive geochemical variability (over a centimeter- and meter-scale distance (Alt, 1995)) and irregular alteration patterns (Gillis and Robinson, 1988,1990; Stern and Elthon, 1979; Harper et al. 1988). Gallahan and Duncan (1994) proposed that the low-temperature alteration pattern in the uppermost basaltic crust of the Troodos ophiolite is the result of irregular fluid flow due to the discontinuous sealing of cracks rather than of a well-defined flow geometry. There appears to be a link between the primary composition of a rock or its location within the oceanic crustal stratigraphy and the form of alteration, such as low-temperature oxidation and reactions binding alkalis in the uppermost pillow layer. Reactive transport models have been used to gain insight to the complex coupling of temperature, fluid flow and chemistry in a hydrothermal system. These models are usually based on a number of simplifying assumptions and the modeling results are therefore rather approximations of general processes or conditions in the hydrothermal 143 system. Most of these coupled models were used to simulate processes in continental systems, in particular in sedimentary basins, and addressed a variety of issues such as geochemical self-organizing phenomena (Ortoleva et al., 1987), the effects of pressure solution on diagenesis and the compartmentalization of sedimentary basins (Dewers and Ortoleva, 1990, 1994), the patterns of porosity alteration under non-isothermal conditions (Hewett, 1986), or the formation of unconformity-type uranium deposits (Raffensperger and Garven, 1995 a, b). An excellent review of basin-scale hydrogeologic modeling is given in Person et al. (1996). Most of these modeling studies are restricted to a local equilibrium assumption. Knapp (1989) concluded that local equilibrium is a good approximation i f an equilibrium state can be obtained over a distance and time period that is less than the problem time and length scales. This situation is likely to occur in sedimentary basins. In contrast, in oceanic hydrothermal systems where convection is vigorous and temperature gradients are steep a kinetic approach to heterogeneous reactions is more appropriate. However, because of the scarcety of kinetic data, models that do account for kinetic processes are rare and often limited to geochemically simple systems. For example, the quartz system has been used to address the kinetic controls on physical phenomena rather than a full chemical system with complex mineral/water interaction (Bolton et al. (1999); Wells and Ghiorso, 1991). The advantage of the quartz system is that it is simple and that its kinetic behavior is reasonably well understood and constrained by data. In a study addressing the effects of a heterogeneous permeability distribution (and thus a heterogeneous distribution of reactive surface area) on kinetically-controlled quartz dissolution and precipitation, Bolton et al. (1999) found that a fluid initially in 144 equilibrium with quartz becomes oversaturated when travelling in a direction of high to lower temperature and undersaturated when moving from lower to high temperature. The transition from oversaturated to undersaturated conditions roughly coincides with the region where flow changes from up-temperature to down-temperature and the distribution of the solute concentrations closely follows that of the temperature. However, fluids that are far out of equilibrium can still be undersaturated while cooling and oversaturated while heating. Fluid disequilibrium is most pronounced in zones of low surface to fluid volume ratio, i.e. coarse grained, high permeability zones. Bolton et al. (1999) concluded that due to time-dependent phenomena, the system or large portions of it cannot achieve geochemical steady state, even over long geologic times. One of the first multicomponent and multiphase reactive transport simulations using a kinetic approach to heterogeneous reactions was carried out by Steefel and Lasaga (1994). The authors modeled a hydrothermal system in a rock of granitic composition. In addition to successfully simulating processes in regions of fluid flow under increasing or decreasing temperature conditions, the significance of chemical processes on porosity and permeability changes were highlighted. The transient simulation coupled with porosity and permeability change indicated a reduction in the permeability and the channelization of flow in the upwelling plume and an increase in permeability and a divergence of flow in the downflow regions of the system. As the flow system evolves, upwelling fluid continuously advects heat upward and downwelling fluids bring relatively cool water to deeper parts of the domain. This process leads to larger horizontal gradients in the density field, which in turn enhances the buoyancy force and increases the Darcy flux. In addition, the upwelling, hot fluids are of significantly lower viscosity than the 145 surrounding fluids which enhances the channelization of flow in the high permeability zones. Channelization is inhibited for downwelling cold fluids as their viscosity increases and pore space may become reduced due to mineral precipitation. Thus, the spatial and temporal variations of the permeability field lead to a modification of the flow geometry and the topology of the isotherms in the system with time. These processes are self-enhancing, zones of increased permeability will lead to larger fluid fluxes which will cause the fluids to be more undersaturated with respect to quartz, which wil l then result in more quartz dissolution. These self-enhancing, coupled effects may prevent the system from reaching a steady state. One of the few reactive transport studies that addressed processes in the oceanic crust was carried out by McCollom and Shock (1998). The model included the lower crust in which temperatures range from 300-900°C and pressures range up to 200 MPa. The authors report that mineral assemblages are essentially identical for fluids moving in the direction of decreasing temperature as they are for fluids moving in the direction of increasing temperature. This result implies that fluids circulating in the oceanic crust, regardless if they are moving upward or downward, may produce the same mineral alteration assemblage, making the alteration mineralogy independent of the prior history of the fluid. In summary, reactive transport simulations are in the process of becoming a major tool in understanding coupled processes in a hydrothermal system. The focus of most of the modeling studies has been on alteration patterns and how these alteration patterns may influence the permeability of the medium under steady state conditions. In this chapter we 146 carry out a transient reactive transport simulation and evaluate how the porosity of the rock changes as the hydrothermal system evolves and discuss how these changes may feed back on the permeability of the rock and the evolution of the hydrothermal system with regard to temperature and fluid flow. Note that no feedback loop is incorporated in the model to relate changing porosity values to changes in the permeability of the system. 3.1.2 Mathematical Model 3.1.2.1 Governing Equations The governing equation for the conservation of solute mass can be written as + V • ( j ^ + J a , v + id,ff )= R,, (i=l,2,...,N t o t) (3.1) (Steefel and Lasaga, 1994), where Q is the molar concentration (moles/unit volume solution) of species in solution, Jdisp, Jdiff, Jadv are the dispersive, diffusive and advective fluxes respectively (all in units of moles per unit area rock per unit time), -> N x)> ( 3 - 1 2 ) where Kj is the equilibrium constant associated with each of these reactions for the appropriate temperature (Steefel and Lasaga, 1994). The rate term for the reversible reactions (Rjaq) from the rate term Rj in the governing equation (equation 3.1) can be eliminated. Following the formulation of Steefel and Lasaga (1994) the governing partial differential equations for primary and secondary species, respectively, can be written a + V-(qCj-BVCj)=R? (j=l>...,Nc) (3.13) + V • (qX, - DVX,) = R? (i=l,....,NX) (3.14) a where D is assumed to be the same for every species. Equation 3.11 states that there are vy moles of primary species Aj in one mole of secondary species Aj. Then vur,=-R? (3-15) where rj is the rate of production of secondary species i . By definition: Nr Combining equations 3.15 and 3.16 results in 152 (3.17) (=1 Taking the equation for mass conservation of the secondary species (equation 3.14), multiplying it with vy, and summing over all the N x secondary species yields: d_ a + v V '=i •DV \\i=\\ -Rj9 (3.18) Adding this equation and the equation for mass conservation of the primary species (equation 3.13) results in an equation with N c (the number of primary species) unknowns (Steefel and Lasage, 1994): d_ a +v- J \\ '=1 = Rmn . (3.19) forG = l , . . ,N c ) . Because one underlying assumption is that only the heterogeneous reactions are irreversible, R j m i n is the only term that remains on the right hand side of the equation. Theoretically any irreversible reaction can be included on the right hand side of the equation in the same way (Kirkner and Reeves, 1988). A total concentration U of the mobile aqueous species j can be defined as (3.20) which expresses the total elemental concentrations in solution (for example, total calcium) (Reed, 1982; Lichtner, 1985). Rewriting the governing differential equations in terms of the total concentrations (Kirkner and Reeves, 1988), gives: 153 ^ + V • (qUj - DVUj) = RJm Q=l v . . ,N c ) (3.21) dt For H + and redox species, this simple concept does not hold and the total concentrations may take on negative values. Redox reactions may be treated in the same way as complexation or heterogeneous reactions but they have to be charge balanced (Reed, 1982; Lichtner, 1985). 3.1.2.1.1 Reaction Rate Law Most of the computer programs which are used for modeling water/rock interactions assume the existence of a local equilibrium state between the fluid and the rock (Wells and Ghiorso, 1991 ). The local equilibrium assumption requires that all phases in contact with one another react reversibly to changes in temperature, pressure and fluid composition (Thompson, 1959). A common practice, when an equilibrium model is used to simulate a geochemical system which is out of equilibrium as a whole, is to divide the system into a number of subsystems, each of which may be considered to be in equilibrium and analyzed independently. This means the system is at local equilibrium (Wells and Ghiorso, 1991). Despite the computational advantages of a local equilibrium assumption, the choice of a kinetic modeling approach rather than a local equilibrium model arises from the problem at hand, namely the effect of the vigor of convection on the geochemistry in an evolving hydrothermal system. The transient nature of the flow and temperature and locally high flow velocities and temperature gradients make a full chemical equilibrium (the complete alteration of the rock) unlikely. A partial equilibrium 154 and/or chemical steady state is probably a more realistic assumption than a full local equilibrium. At partial equilibrium, only some of the minerals in the assemblage are at equilibrium (incomplete alteration). At chemical steady state a net mass transfer between the fluid and the rock occurs and mineral abundances, mineral surface areas, porosity, permeability, and the boundaries between mineral assemblages all change (Wells and Ghiorso, 1991). The kinetics of mineral precipitation/dissolution reactions depend on the rate of transport of various aqueous species to the mineral surface as well as on the rate of attachment or detachment of ions from the active sites on the mineral surface (Lasaga, 1990). A quantitative rate law for a mineral precipitation/dissolution reaction should therefore include a formulation which describes both transport and surface processes. Because of the complexity in multicomponent systems one commonly tries to identify the rate-limiting step and defines the mineral precipitation or dissolution as being either transport or surface controlled. Transport-controlled reaction rates occur i f the transport rate of aqueous species to and from the mineral surface is slower than the rate of attachment/detachment of ions on the surface. Reaction rates are surface-controlled i f the attachment/detachment rate of ions is slower than the transport rate of species to the mineral surface. Quite often, however, both processes are equally important and the definition of the rate limiting process is somewhat arbitrary. The kinetic law used in the code OS3D is based on the assumption that attachment and detachment of ions from the mineral surface is rate-limiting (surface-controlled reaction rate). The surface controlled reaction rate is dependent on the potential energy barrier which needs to be overcome to form an activated complex on the mineral surface (Lasaga, 1981; Wells and Ghiorso, 155 1991). The advantage of this assumption is that concentration gradients near the mineral surface and diffusion to and from the mineral surface don't have to be accounted for. Therefore there is no need to solve for aqueous concentrations immediately adjacent to the mineral because these species concentrations are just the concentrations of the same species in the bulk solution. Consequently, surface-controlled growth or dissolution is a much simpler problem from a computational point of view (Steefel and Lasaga, 1994). According to Steefel and MacQuarrie (1996), for an elementary reaction written as A + B » C (3.22) a forward rate can be written as Rjw=kJW[A][B) (3.23) and the backward or reverse rate as Rhw=khw[c) (3.24) where [A], [B], and [C] indicate the concentrations of the species A , B, and C, respectively, and kfw and kbw are the forward and reverse rate constants respectively. The complete rate expression for species A is therefore (Steefel and MacQuarrie, 1996): At equilibrium the forward rate equals the reverse rate which makes the equilibrium state a dynamic state at which forward and reverse rates balance. Thus, at equilibrium: *>M*]=**[c] (3-26) rearranging and applying the law of mass action gives 156 where K m is the equilibrium constant for the reaction (Steefel and MacQuarrie, 1996). In reality, this principle holds only where elementary reaction mechanisms are involved (Lasaga, 1984; Steefel and MacQuarrie, 1996), or in certain cases where a single elementary reaction is the rate-limiting step within the overall reaction (Aagaard and Helgeson, 1982). For the case of local equilibrium the actual values of the forward and reverse rate constant are not important, as long as the ratio kfw / kbw is equal to K m and the rates are fast enough such that local equilibrium is attained during the time scale of interest. After rearranging, a rate expression in terms of a single forward rate and an equilibrium constant can be obtained: ^-[AlB^-rMA-UlBh^ Q' UIBK ) V K m J (3.28) Q m is the ion activity product QM = ft A7 (3-29> where the aj are the activities of the components making up the mineral and vmj are the stoichiometric coefficients (Steefel and MacQuarrie, 1996). This formulation has been widely used as the basis for describing mineral dissolution and precipitation reactions (Aagaard and Helgeson, 1982; Lasaga, 1984). However, realizing that it is the reactive surface area of the mineral and not the actual mineral concentration which determines the rate an expression for the interfacial surface area A m replaces the mineral concentration 157 term. Also, mineral dissolution reactions are commonly dependent on species present in the solution which makes it necessary to include a term that describes the dependence of the dissolution rate on ion activities. For instance, many mineral dissolution reactions are pH dependent, such as the dissolution of calcite: C a C 0 3 + H + = C a + 2 + H C 0 3 (3.30) This leads to the following form for the rate law of crystal growth and dissolution of a mineral, expressed as the rate of change rk in the mole number n m of mineral A m (Lasaga, 1981, 1984; Aagaard and Helgeson, 1982; Steefel and Van Cappellen, 1990), rm = sgn log On AJk. (Nc+Nx \\ nx / \\ M Q (3.31) where A m is the mineral's reactive surface area (cm ) and k m is its intrinsic rate constant 2 (either the growth or dissolution rate constant (mole/cm sec), ai is the activity of a basis species raised to an empirically-determined power p which affects the far from equilibrium dissolution rate, and M and n are positive, experimentally-determined numbers. In this form of rate law the three components briefly discussed above can be recognized. ( A m k m ) states that the reaction rate is proportional to the reactive surface area and the dissolution rate constant. This implies that i f the surface area is zero, i.e. the mineral does not exist, it cannot precipitate until a crystal nuclei is formed. The FI-term represents the catalyzing or inhibiting effect of species in solution on the reaction. Depending on whether a species promotes or inhibits the formation of the activated complex, the exponent p is either positive or negative. The third term represents the link 158 between thermodynamics and the kinetic state of the system. When mineral A m is oversaturated, Qm>Km, and the mineral precipitates. When Q m < K m the mineral is undersaturated and dissolves. K m is the equilibrium constant for the mineral written as the destruction of one mole of mineral m. The rate constant is temperature dependent and this dependence is commonly expressed by an Arrhenius equation (Lasaga, 1984) where A is a pre-exponential factor which may be a function of temperature, but is commonly considered to be a constant (Lasaga, 1981; Wells and Ghiorso, 1991 ), E a is the activation energy (J/mole), T is the absolute temperature, and R is the gas constant (8.3143 J/K mole). Since most rate constants are reported at 25°C, the code OS3D calculates the temperature dependence of k+ in the form of an interpolation function as where k 2 5 is the dissolution rate at 25°C (both in mole/cm sec). Values of k+ for any mineral can be calculated if the appropriate activation energy and k25 of the mineral are known. The value of E a is normally determined experimentally. The equations indicate that, in general, reaction rates increase exponentially with respect to temperature which means that high temperature systems are more likely to approach equilibrium than low temperature systems over a given period of time. The advantage of a kinetic treatment of heterogeneous reactions is that it does not require the constraints and assumptions that are necessary for an equilibrium model (e.g. k+ - Ae R'k (mol/cm2 sec) (3.32) (3.33) 159 partial or local equilibrium). Moreover, equilibrium between fluid and rock is often not achieved in nature (Bethke, 1996). The kinetic formulation is a more general approach because even if the system is at equilibrium it is possible to use a fully kinetic formulation. The most significant shortcoming of the kinetic approach is the scarcity of well constrained kinetic parameters. Very few details are known about mineral/water reaction rates (Steefel and Lasaga, 1994). Rate constants and the dependence of the reaction rates on parameters such as ionic strength, temperature, pH, or the reactive mineral surface area all have to be known. Kinetic data, i f available, are often restricted to limited ranges of temperatures and pressures for the species and phases present. Kinetic formulations are typically based on results and measurements from lab experiments. But there is a significant discrepancy between the results from lab experiments and field observations. For albite, for example, this discrepancy can be as large as four orders of magnitude (Bethke, 1996). Besides measurement errors, one reason for the discrepancy between lab experiments and field observations could be the difficulty in defining the reactive surface areas of the mineral grains. Not only is it difficult to determine in the field but it is also not constant with time (Steefel and Van Capellen, 1990). The distribution of phases, grain sizes, and pore spaces or, in a fractured medium, the fracture distribution, geometry and interconnectivity all have an effect on the reactive surface area of the medium but none of these parameters can be known with certitude because they are difficult to measure or estimate in the field (Steefel and Lasaga, 1994). Lab experiments are commonly performed under ideal conditions where mineral surfaces are fresh and not occluded. In the field, however, minerals may be coated with oxides, hydroxides, or organic material. 160 They may also be occluded by contact with other materials such as other grains, reaction products, or organic matter (Bethke, 1996). Also, the roughness of the mineral surface is an important factor. Smooth surfaces, which are more common in the field are less reactive than rough, edgy surfaces as typical of the fresh material used in lab experiments. Furthermore, in the saturated zone in the field there exist small scale variations in pore velocities and even stagnant, hydrologically isolated fluid in the pores can occur, resulting in fluids which are in contact with mineral surfaces but isolated from the bulk chemistry of the pore water. In addition, in fractured rocks typically only the mineral surfaces lining the fractures are reactive as fluids bypass many of the mineral grains in the rock. Moreover lab experiments may use deionized water as aqueous solution which may be far from equilibrium with the minerals, unlike in nature where a range of saturation states is likely to exist (Bethke, 1996). Mineral reactions may be catalyzed or inhibited in strong electrolyte solutions such as seawater. Dove and Crerar (1990), for example, demonstrated that quartz dissolves into electrolyte solutions up to 30 times more quickly than it does in pure water. Trace minerals in the solid medium may exert a significant control on fluid compositions which cannot be accurately modeled by reactions between the major minerals and the fluid. Similarly, the effect of organic material and the action of bacteria are often impossible to include in model calculations or lab experiments (Bethke, 1996). Despite the uncertainties and shortcomings of the kinetic modeling approach, it is believed that the conclusions and ideas drawn from this reactive transport study are general enough to be representative and realistic. Furthermore, it may be argued whether an alternative approach, e.g. a local equilibrium model, would be more representative. 161 Considering the heterogeneity and the transient nature of the flow and temperature fields, which were discussed in the previous chapter, the restriction to a local equilibrium seems problematic. The reader will see that despite the uncertainties regarding the kinetic formulation, the model produces results that are largely in accordance with observations from Middle Valley and other oceanic ridges. 3.1.2.1.2 Activity Coefficient Model The code O S 3 D can either be run with a unity activity coefficient or with an extended Debye-Hueckel formulation to calculate activity coefficients for all charged species. The extended Debye-Hueckel Formulation is given by where y{ and zs are the activity coefficient and the charge of each species i , respectively, I is the ionic strength of the solution, A and B are constants that depend on the dielectric constant and the temperature, a( is the ion size parameter, and bs is an ion specific parameter that accounts for the decrease in solvent concentration in concentrated solutions. The ionic strength (I) is calculated from log yt = -Az){iy + hi (3.34) 1 + Bat{I)2 1 m (3.35) where: Q = concentration of ion species I m = number of ions present in solution 162 Zj = charge of ion i The coefficients a; and b; describing the temperature dependence of the activity coefficient are taken from the E Q 3 database (Wolery et al. 1983). It should also be noted that the Debye-Hueckel formulation becomes increasingly inaccurate in more concentrated solutions (ionic strengths greater than 0.5). For marine conditions (I = 0.7) the extended Debye-Hueckel equation may still provide reasonable results. 3.1.2.1.3 Porosity and Reactive Surface Area A s mentioned above, the reactive surface area of a mineral is probably the least constrained parameter in the rate expression. The bulk porosity and thus the permeability of the aquifer material changes as the total volume of the mineral phases changes. The change of the volume fraction of an individual mineral (j)m as a result of precipitation/dissolution reactions can be written as: - Vmrm, (3.36) dt where V m is the molar volume of the mineral and r m the reaction rate (Steefel and Lasaga, 1994). Converting this equation into the finite difference form, the individual mineral volume fractions m (t + At) = m (t) + Vmrm (t + At)At (3.37) The porosity of the medium is then related to the sum of the volume fractions of each mineral phase in the system as (Steefel and Yabusaki, 1995) 163 m (3.38) A formulation for the dependence of the total reactive surface area on the permeability structure of the rock is commonly inferred from simple geometrical assumptions about pore - or fracture geometry and distribution. In general, fluids passing through a relatively wide fracture will encounter a smaller amount of surface area than fluids passing through a narrower fracture. And typically, fluids passing through a porous medium will be in contact with more reactive mineral surface area than fluids travelling through a fractured medium. For constraining the reactive surface area of the medium Steefel and Lasaga (1994) use a formulation for an idealized, single set of smooth, parallel fractures with constant aperture, 5, and a constant spacing d. The assumption is that mineral/water reactions are confined to the fracture walls, the third dimension into the rock does not contribute to these reactions. The fracture porosity for a single fracture is then (e.g. Snow, 1970; Marsily, 1986; Phillips, 1991) (3.39) and for a set of N F parallel fractures A 8, (3.40) where the subscript i refers to an individual fracture set. The total surface area A t o t in contact with fluid per unit volume rock becomes 164 4* = ^ j (3-41) and for a single set of parallel fractures, Ao, = (3-42) which states that the total surface area of a fractured rock in contact with fluid is directly dependent on the fracture spacing d. The assumption of smooth fracture walls underestimates the actual surface area in contact with the pore water. Actual BET measurements of natural mineral surfaces suggest that surface roughness can account for one to two orders larger mineral surface areas than for idealized geometries (White and Peterson, 1990; Wells and Ghiorso, 1991). Also, wider fractures with greater fluid flow and higher water/rock ratios will exhibit a different mineral alteration assemblage and lead to different proportions and compositions of the minerals in comparison to narrow fractures with low water/rock ratios. For hydrothermal alteration at 300°C, for instance, epidote would be a typical alteration phase at low water/rock ratio but it would be replaced by chlorite at higher water/rock ratios (McCollom and Shock, 1998). Thus, we can expect to find in a rock that is pervasively altered different mineral assemblages and different compositions of individual minerals depending on whether they occur in the bulk of the rock, in relatively narrow fractures, or in relatively wide fractures. In addition, minerals are often not evenly distributed within the rock or on the fracture wall. This may occur where secondary quartz or calcite line the fracture walls. Thus the bulk mineralogy of a rock does not necessarily represent the mineral composition of the fracture wall. Further complicating the quantification of the reactive surface area of minerals is that only a portion of the total mineral surface area in a system (the reactive surface area) 165 actually participates in chemical reactions (Helgeson et al. 1984). The theory is that the reactive surface area is related to the density of crystallographic defects of mineral grains because such defects are sites of enhanced dissolution (Holdren and Speyer, 1985; Brantley et al. 1986; Wells and Ghiorso, 1991). As the reactive surface area is poorly constrained by observations or experiments the modeler has to assume a representative number for the model. The code OS3D checks i f the time step is small enough such that the mineral volume fraction does not go below zero. A negative mineral volume fraction may occur where mineral solubilities are high and/or the amount of a mineral present is very small. Presently, reactive surface areas of minerals are updated differently, depending on whether the mineral is primary or secondary (i.e., it is not initially present). For primary minerals, the surface areas as a function of time are given by where mj0 is the initial volume fraction of the mineral m and 0)2/3 implies that when the porosity of the medium decreases to zero, the reactive surface area also decreases to zero. For secondary minerals, the reactive surface areas as a function of time are given by: (3.43) precipitation dissolution (3.44) £ f*0 precipitation 166 3.1.3 Numerical Model The software used for simulating multicomponent reactive transport is based on the OS3D-software package by Steefel and Yabusaki, (1996). The code was modified so that it could be coupled with the flow code that was used to produce flow, temperature, and density fields at specified times. Minor modifications had to be made to the code OS3D to allow for a full transient reactive transport simulation. The code simulates reactive transport by splitting the reaction and transport steps in time and is based on the continuum representation of transport and reaction in porous media. For the solution of the non-linear set of partial differential equations which describe coupled reaction and transport, the integrated finite volume technique is applied (Patankar, 1980). The time-splitting of transport and reaction terms is useful for simulations of advection-dominated flow systems, as it creates less numerical dispersion compared to other procedures. The idea behind the operator splitting approach is to solve the transport terms and reaction terms separately. The technique carried out in the code is the Strang scheme, in which a half transport step is followed by a full reaction step which is in turn followed by another half transport step (Strang, 1969; Zysset et al. 1994). This three-step scheme requires no iteration and can be written as: ^ pre-reaction CJ\" ^ At / 2 followed by the reaction step 167 = L(Cj, (k=l,.. . ,N c+N x), (3.45) followed by another half transport step, (c ; + 1 - c ; reacted ) Z ( C , r c , (k=l,.. . ,N c+N x) (3.47) At I 2 where Ck pre-reaction are the concentrations of the individual species at the end of the first 1/2 transport step, U j r e a c e are the total concentrations at the end of the full reaction step, and Ck\" + 1 are the individual species concentrations at the end of the second 1/2 transport step. L refers to the spatial differential operator defined by L=[V-(u - Dgrad)]. Zysset et al. (1994) state that this scheme is second order accurate in time. Advection-dominant transport problems require accurate resolution of sharp concentration fronts. The solution of these kinds of problems is typically achieved by the use of explicit numerical schemes, i.e. forward schemes in which the known concentrations at time t are used to calculate the unknown concentrations at time t + At. The transport algorithm used in OS3D uses an explicit, third order accurate total variation diminishing or T V D method, proposed by Gupta et al. (1991). This scheme is based on the calculation of the solute concentrations at the face of a grid-cell in a way that includes a dependency of the concentrations on the flow direction. For flow in the positive x-direction, for example, the estimated concentration would be: V J (3.48) where 168 f /?(,.) = Max 0,Min 2,2r, 2 + -V 3y (3.49) and r = C\" - C\" Sx, 8x, C\" ^ 1 + 1 C\" Sx, Sx /+1 (3.50) and similarly for negative velocities. The method is generalized for multiple dimensions. After the calculation of advective transport, the solutions for diffusion and dispersion are obtained. Newton's method is used to solve the set of non-linear equations. Explicit schemes such as the TVD scheme used in OS3D are dependent on stability criteria, such as the Courant criterion. The Courant number Co controls the time step At such that the ratio of the distance traveled by fluid advection in a single timestep to the grid spacing (Pinder and Gray, 1977) Co. A r v . Ax < 1 (3.51) for advective transport in the x-direction. For explicit schemes, the transport equation is stable for values of the Courant number of Co x < 1, which implies that the solute concentration cannot be advected more than one grid cell in a single time-step. For multidimensional domains the Courant number has to fulfill the criterion in each direction. The Peclet number is a non-dimensional expression which controls the spatial discretization in each of the coordinate directions by comparing the characteristic time for dispersion and diffusion for a given length scale with the characteristic time for advection. It is represented as 169 vrAx Pe=— (3.52) D XX where v x = q/(|)w is the pore water velocity in the x-direction Ax is the cell size in the x-direction and D x x is the dispersion coefficient in the x-direction. A t grid Peclet numbers > 2, numerically calculated concentration fronts using unmodified higher order schemes tend to exhibit spurious oscillations in the vicinity of the front. The problem becomes more significant as the Peclet number is increased. A common way to reduce the grid Peclet number to a value of < 2 is to refine the grid, i.e. to decrease Ax. This, however, has the disadvantage of leading to more grid points for which a solution has to be obtained and thus requires more computer time and resources. Peclet numbers in this model study do not exceed a value of 2. 3.2 Reactive Transport Modeling 3.2.1 Hydro thermal Alterat ion of Oceanic Crus t - Overview Early studies of oceanic basement rocks found that the metamorphic grade in the upper oceanic crust gradually increases downward from zeolite to greenschist facies (e.g. Cann, 1969; Shido et al. 1974). The discovery of hot springs and black smoker vents at 170 the seafloor demonstrated the importance of the alteration of the crust on the composition of hydrothermal fluids. Since then numerous studies elucidated how alteration may vary in different regions of submarine hydrothermal systems and how fluid compositions evolve. These regions include a recharge zone, a discharge zone, and a reaction zone at depth. Seawater, which enters the recharge zone is heated and reacts as it moves generally downward to the reaction zone. The reaction zone is situated near the magmatic heat source. In this zone temperatures are high (>350°C) and fluids are thought to acquire their final chemical fingerprint. Due to buoyancy forces at these high temperatures, fluids move upwards toward the seafloor through the discharge zone. Discharge zones can be classified as \"focussed\", where high-temperature hydrothermal fluids are sufficiently channeled in high permeability conduits to vent directly onto the seafloor, or \"diffuse\", where hydrothermal fluids do not reach the seafloor but rather mix with seawater in the subsurface (Alt, 1995). We wi l l concentrate on focussed upflow zones only. The main chemical processes that occur in the recharge zone include low-temperature oxidation, low-temperature alkali fixation, Mg-fixation, anhydrite formation, and alkali loss at elevated temperatures (Alt, 1995). The major alkali sink are micas, chlorite and smectite (caledonite - nontronite/saponite). The fixation of magnesium occurs in smectites (e.g. saponite) at low temperatures (< 200°C) and chlorite at higher temperatures (>200°C). Coincident with the uptake of magnesium by the rock at temperatures greater than 150°C is the loss of alkalis from the rocks. Another main reaction occurring during seawater recharge is the formation of anhydrite. Heating seawater to temperatures of 150 - 200°C wi l l precipitate most of calcium and about two-171 thirds of the sulfate content as anhydrite (Bischoff and Seyfried, 1978). The addition of calcium from fluid/basalt interaction can increase the amount of sulfate depletion (Bischoff and Dickson, 1975). Anhydrite precipitation inhibits the formation of Ca-bearing silicates (such as epidote) and limits the amount of seawater sulfate that enters the high temperature portion of the hydrothermal system (>250°C) where sulfate reduction can take place (Shanks et al. 1981; Janecky and Shanks, 1988; Bowers, 1989, Alt et al. 1989a). The hydrothermal fluid obtains its chemical signature in the reaction zone where fluid temperatures can be higher than 350°C. The mineral paragenesis found in the reaction zone is consistent with the greenschist facies metamorphic grade. Common alteration reactions include the formation of chlorite (chloritization) which may replace primary plagioclase and mafic minerals; the alteration of plagioclase to albite (albitization) according to C a A l 2 S i 2 0 8 + 2 N a + + 4 S i 0 2 -> 2 N a A l S i 3 0 8 + Ca +2 (3.53) anorthite quartz albite and to epidote/clinozoisite according to 3CaAl 2 Si 2 0 8 + C a + 2 + 2 H 2 0 -> 2 Ca 2 Al 3 Si 3 0 , 2 (OH) + 2 H + (3.54) anorthite clinozoisite 111 Adding both equations from above results in an equation which describes the equilibrium between anorthite, albite, and clinozoisite: 2 C a A l 2 S i 2 0 8 + 2 S i 0 2 + N a + + H 2 0 -> N a A l S i 3 0 8 + C a 2 A l 3 S i 3 O i 2 ( O H ) + H + (3.55) anorthite quartz albite clinozoisite This equation implies that during hydrothermal alteration of plagioclase, the fluid may approach equilibrium with respect to anorthite, quartz, albite, and clinozoisite and that the alteration reactions involving these minerals may affect the ratios of Na , Ca, S i 0 2 , and H + ratios in the fluids. Moreover, the equation suggests that there is an inverse relationship between the p H and Si0 2 , aq , a decrease in p H would cause an increase in the aqueous silica concentration. Thus the S i0 2 , aq concentration and the p H play an important role in controlling the concentrations and ratios of other aqueous species (Berndt et al. 1989). Depending on the extent and the direction of silicate alteration reactions, hydrothermal fluids may or may not become saturated with respect to quartz. Quartz saturation at depth can only be rigorously assumed when quartz occurs as an alteration phase (Berndt et al. 1989). A sink for aqueous silica is the albitization reaction, equation 3.53. Another major sink is the hydration of Si0 2-depleted, mafic minerals such as the alteration of olivine or pyroxene to talc (here used as an analogue for smectite-chlorite alteration (Berndt et al. 1989), according to: 3 M g S i 0 3 + H 2 0 + S i 0 2 - » M g 3 S i 4 O i 0 ( O H ) 2 (3.56) enstatite talc 173 This reaction proceeds from left to right and it can be concluded that the Si02,,aq production from the alteration of silica saturated phases has to exceed the silica loss from the replacement of silica-depleted phases in order to obtain quartz saturation. Si02,aq production may be favored at high initial Ca-concentrations in the fluids, a high albite component in the plagioclase, and high water/rock ratios. If mafic phases are abundant, the uptake of aqueous silica w i l l result in low Ca concentrations, high p H , and high N a concentrations, consistent with the feldspar alteration reactions from above. In contrast, at quartz saturation C a concentrations wi l l be high, the p H and N a concentrations low. Aqueous silica concentrations wi l l therefore exert an important control on the p H and the Ca/Na ratios in the fluid (Berndt et al. 1989). Discharge zones occur as focussed or diffuse and can extend as deep as to the sheeted dikes in the lower oceanic crust. Here, we focus on the shallow portions of the focussed upflow zones as they occur and have been sampled at Middle Valley. Typical alteration assemblages in these zones include quartz/epidote (and possibly some albite), which commonly indicate regions of high flow rates and minor compositional changes within the fluid (Seyfried et al. 1988); chlorite or illite assemblages (e.g. Zierenberg et al. 1988; Richards et al. 1989), which may indicate mixing of hydrothermal fluids with seawater; silicification, by cooling and mixing of hydrothermal fluids with seawater; and the formation of massive sulfides. When a fluid moves through the different regions of a hydrothermal system, the recharge zone, reaction zone, and discharge zone, its composition is affected by: (1) seafloor and subsurface temperature, (2) pressure, (3) rock composition and state of 174 alteration, which may include layers of sediments, (4) chemical kinetics, (5) phase separation, (6) subsurface mixing, and (7) subsurface heat conduction (von Damm, 1995). A review of fluid and rock compositional data from Middle Valley and various other ridges as well as from experimental studies reveal certain consistent patterns in major ion chemistry. A principal characteristic of seawater/basalt interaction is the almost complete removal of magnesium from the solution at temperatures > 150°C due to the precipitation of M g - O H - S i minerals with a coupled release of protons to solution (e.g. Bischoff and Dickson, 1975; Seyfried and Bischoff, 1979). Seawater sulfate is almost completely removed and the discharging hydrothermal fluids contain very little, i f any, sulfate. The sulfate depletion is a result of both anhydrite precipitation when the downflowing seawater is heated above about 150°C, and of reduction processes in the hottest parts of the hydrothermal system. Other features of vent fluids are: (1) alkalis other than N a are enriched with respect to seawater even though the alkali metal concentration in basalt is low; N a may be enriched or depleted, a major sink of N a is the albitization of basalt; alkalis tend to be incorporated into low temperature (<100 °C) alteration phases; (2) alkaline earths Be and Ca are enriched, the latter due to the albitization process or other cation exchange reactions, Ca may also be depleted by its removal into anhydrite or secondary Ca-silicates, such as epidote; Sr is either enriched or depleted; (3) trace metals, such as Fe and M n are greatly enriched in the endmember hydrothermal fluids; and (4) silica contents are much higher than in seawater and are commonly constrained by quartz equilibrium at high temperatures (von Damm, 1995). While these are general characteristics, the exact chemistry of vent fluids is known to vary over short distances within a vent field (Butterfield et al. 1990) and over 175 short timescales, in particular in the early stages of venting activity (von Damm et al. 1992). The average lifespan of a hydrothermal system is unknown, as is the fluid's residence time in the oceanic basement. A t Middle Valley, most of the general characteristics of an oceanic hydrothermal system have been observed (see Chapter 1). However, Middle Valley differs from the above mentioned characteristics in that it is covered by a thick sequence of sediments. The sediment cover adds some complexity to the general features of an oceanic hydrothermal system as a result of the different primary chemical composition and its hydrologic properties. The primary sediments are composed of quartz, feldpars (including k-feldspar), mica, and carbonates of biogenic origin (Kurnosov et al. 1994). Hydrologically, the sediments are of low permeability such that flow rates through the sediments are low. Alteration in the sediments is therefore dominated by the temperature distribution. The layered and anisotropic nature of the sediments promotes lateral fluid flow from the discharge zone into the sediment section. This factor and the elevated temperatures at the discharge zone lead to concentric zones of different mineral assemblages surrounding the discharge conduit as described earlier (Figure 1.5). In the vicinity of discharge zones, greenschist facies metamorphism may form only a few tens of meters below the seafloor (Kurnosov et al. 1994). Away from the discharge zone, the sediments are generally subject to low- to medium-temperature alteration. Typical alteration reactions such as chloritization and the formation of smectite occur in the sediments as well . Alteration reactions that only occur in the sediments include the alteration of primary k-feldspar and micas to a sericite/chlorite phase. Calcite may be present in the sediments either as primary, biogenic calcite or as secondary calcite. 176 Secondary calcite occurs at Middle Valley as cement and concretions in the upper parts (0 - 300m, Site 857) of the sediments (Leybourne and Goodfellow, 1994). The chemistry of the vent fluids at Middle Valley (Site 858) appears to be largely controlled by basalt-seawater interaction. The magnitude of the sediment fingerprint (in terms o f alkali systematics, strontium isotope ratios, or the content of organic material) is low, even compared to other sedimented ridges (Butterfield et al. 1994). Thus, the overall effect of the sediments on the chemistry of the hydrothermal system appears to be restricted to alteration patterns within the sediments, in particular in the regions surrounding discharge zones. 3.2.2 M o d e l Design The design of the model domain corresponds to that used for the flow simulations in the previous chapter. The chemical conditions specified for the model are summarized in Table 3.1 and Figure 3.2. The initial composition of the sediment and the basalt layers is based on the modal analysis of samples from Middle Valley (Davis et al. 1992b) but it should also be representative for other sedimented ridge areas. The basalt is composed of clinopyroxene (with Mg.Fe = 6:1) and plagioclase (an 5 5 - an 7 5). The sediments consist of a mixture of quartz, plagioclase (an 25 and an 5 5), k-feldspar, and calcite. Solid solutions of feldspars and pyroxenes were modeled as ideally mixed solutions. The range of secondary minerals which are allowed to precipitate but do not appear in the initial composition of the rock, include chlorite (the endmembers daphnite and clinochlore), illite, albite, anhydrite, clinozoisite, and muscovite. The choice of 177 secondary minerals is based on the one hand on observations from Middle Valley, and on the other hand on our goal to model a set of mineral alteration reactions in which species are distributed between a dissolving primary mineral phase and a precipitating secondary mineral phase (e.g. the alteration of feldspar, equation 3.55). In addition, emphasis is placed on general occurrence rather than site specific occurrence. For this reason, wairakite, which is a common mineral at Middle Valley but not a mineral that has been reported to be a major alteration phase at other ridges, is not included in our simulations. Also , we do not account for redox reactions which implies that minerals involved in redox processes (e.g. pyrite or epidote) are not considered. A s an approximation for epidote, clinozoisite is part of the set of secondary minerals. Muscovite was included to represent sericite. The major ion composition of seawater is based on samples from Middle Valley and is summarized in Table 3.2. In addition, a range of secondary species which are summarized in Table 3.2 is allowed to form. Fluid of seawater composition occurs along the top boundary throughout the simulation time. A s initial (t = 0) porewater of the sediments and the basalts, seawater is equilibrated with the rock at temperatures consistent with the temperature distribution 300 years after the onset of fluid motion. The reason for the 300 year time gap between the flow and temperature simulation and the chemical simulation relates to the initial temperature (and pressure) conditions (> 400°C) close to the two phase curve (the curve that separates the pressure-temperature space in which only a liquid exists from that in which both liquid and vapor exist). The reactive transport code is not designed to handle phase separation. In addition, the code showed 178 convergence problems resulting from the calculation of erroneous activity coefficients at temperatures of around 400°C. The geochemical database was produced with the program S U P C R T 9 2 (Johnson et al. 1992). Equilibrium constants of minerals and aqueous complexes were calculated for a temperature range from 1 to 380°C. A n interpolation function is used to calculate the equilibrium constant for any desired temperature within this range. The pressure is fixed at 25 Mpa , which means that the effect of pressure on the equilibrium constant is not accounted for. The reaction rate constant at 25°C for quartz is specified at 4 .3xlO\" 1 4 m\" 2 s\"1 (Rimstidt and Barnes, 1980) and the rate constant of all other minerals in the system is specified at l .OxlO\" 1 2 m\" 2 s\"1 (Table 3.1). This order of magnitude estimate of l .OxlO\" 1 2 m\" 2 s\"1 is loosely constrained by estimates of reaction rates for feldpars by Helgeson et al., (1984). Similarly, the activation energy for quartz is specified at 18 kcal mole\"1 (Rimstidt and Barnes, 1980) and the activation energy of all other minerals at 5 kcal mole\" 1, which is loosely constrained by estimates of activation energies for feldpars by Helgeson et al., (1984) (Table 3.1). The lack of accurate kinetic data and the resulting weak constraints on the kinetic parameters for most minerals in the system are obviously a limitation to the representativeness of this study. A principal objective of the simulations is to reproduce the dominant alteration reactions of the system, rather than reproducing details of the chemical conditions. The model results, which w i l l be presented in the following sections, show that with the assumptions underlying this model we obtained results that are for the most part consistent with observations from oceanic ridges. These include the dominant alteration reactions and many characteristics regarding the evolution of the fluid 179 composition from seawater to a hydrothermal fluid (e.g. the pH). Observations that could not be reproduced wi l l be pointed out where appropriate. The initial porosity of the medium is 10%. The porosity is updated during the simulation based on the amounts of minerals precipitated or dissolved. The initial surface area of each mineral is fixed at 10 m 2( minerai)/ni 3(rock) (Table 3.1). Wi th the number of primary minerals, the total initial reactive surface area is then 60 m 2 /m 3 and 40 m 2 / m 3 for the sediments and the basalt, respectively. Values for the initial total reactive surface area are within a range of values used by Steefel and Lasaga (1994). According to equations 3.42 and 3.39 this corresponds to a fracture spacing of 0.03 m and 0.05 m for a single set of parallel fractures and (with a fracture porosity of 10%) an aperture of a single fracture of 0.003 m and 0.005 m for the sediments and basalt, respectively. Note that the sediments exhibit a fracture porosity only in regions where they have been indurated due to hydrothermal alteration. A sensitivity analysis on the magnitude of the initial surface area suggests that it affects the amounts of mineral precipitation/dissolution slightly but does not affect the nature of alteration reactions. 3.2.3 The Influence of Hydrothermal Circulation on Mineral Alteration, Fluid Composition, and Porosity The results of the reactive transport modeling w i l l be presented according to the regions of an oceanic hydrothermal system; the recharge zone, discharge zone, and reaction zone. The locations of the mineralogical and fluid composition profiles are 180 shown in Figure 3.3. The amounts of minerals precipitated or dissolved are given as changes in the volume percentage per unit volume. A negative volume change implies a net dissolution of the mineral, a positive change a net precipitation. Where not otherwise specified, the volume change of a mineral is calculated with respect to its inital volume fraction in the system. The porosity change (in percent per unit volume) of the medium is calculated from the net volume change of the solid phase. A negative porosity change corresponds to a porosity decrease, a positive change to an increase. A l l aqueous species are presented in moles/kg. 3.2.3.1 Recharge Zone The dominant reactions in the recharge zone (which comprises the high-permeability window in the sediments as well as the region in the basement below the window) are the precipitation of anhydrite, chlorite, illite, muscovite, and quartz (Figure 3.4). Anhydrite formation begins just above the sediment/basement interface and shows a distinct peak just below the interface at 500 years (Figure 3.4). A s time progresses, the region of anhydrite formation extends further downward toward the bottom boundary of the domain. The large quantities of anhydrite that precipitated initially below the sediment/basement interface start to redissolve. A t 5000 years the maximum amount of anhydrite precipitation is 1.2%; at 15000 years the highest value has decreased to about 1.1%. A t 15000 years, a second anhydrite peak is present near the bottom boundary. Thus the evolution of the anhydrite precipitation pattern suggests a general increase in the 181 region affected by anhydrite precipitation and a shift of maximum anhydrite precipitation downward, that is, in the direction of flow. The second most important mineral in terms of volume percentage precipitated is chlorite (clinochlore, daphnite). Chlorite formation is initiated about halfway through the sediments. A t 500 years a slight maximum is present just below the sediment/basement interface, at 5000 years the amount of chlorite in the basement section is nearly uniform, at 15000 years chlorite shows a steady increase downwards towards the bottom boundary. Illite is not present in the basement at early times, presumably as a result of the high temperature conditions. It is present within the basement section of the profile at 5000 years with up to 0.2 volume % precipitated. Illite shows a steady increase from the sediment basement interface to the bottom boundary where it makes up more than 1 % of the solid phase at 15000 years. Muscovite occurs in the basement section of the recharge zone in small and nearly constant amounts throughout the simulation time. Quartz precipitation begins just above the sediment/basement interface but significant quantities of quartz formation are restricted to the lower half of the basement section. The change of the fluid composition with depth (Figure 3.5) is consistent with the alteration of the solid phase. Throughout the simulation time no or little compositional change occurs within the low- temperature, upper part of the sediments. The most noticeable change within the sediments is an increase in silica in the lower half of the sediment section, in particular during the early convective phase of the system (here represented as a snapshot at 500 years). During this phase M g , Ca, and SO4 all decrease significantly below the sediment basement interface. Calcium recovers in the lower 200 182 m of the basement to about seawater levels. Potassium shows a steady decrease, which steepens near the bottom boundary. Sodium exhibits a slight drop in its concentration as the fluid crosses the interface. Iron and silica are the only species that show an increase in the aqueous phase in the basement section of the profile. A s time progresses the large compositional changes across the sediment/basement interface that are present at early times (500 years) become less intense. The location of the main compositional changes has shifted towards the bottom boundary. The slopes of the calcium and sulfate profiles show a reversal about midway through the basalt; while both species concentrations drop considerably below the interface during early time, sulfate and in particular calcium show an increase at later times. This slight increase is consistent with the redissolution of anhydrite near the interface, and the shift of anhydrite precipitation further into the reaction zone. While the slopes of the profiles are less steep than during the early phase, the general trends of all other species are essentially unchanged. In conclusion, within the downflow zone, large amounts of minerals precipitate just below the sediment/basement interface; in particular anhydrite, chlorite, illite, and muscovite. The precipitation is initiated by the increasing temperature as seawater moves down through the recharge zone. The high precipitation rates lead to a decrease of all major species in solution with depth, except iron and silica. Calcium and sulfate are taken up by anhydrite, magnesium is removed through chlorite and illite precipitation, and potassium is incorporated in the phases illite and muscovite. The mineral precipitation pattern and the aqueous concentration profiles are variable through time. Anhydrite, for 183 instance, shows a distinct peak just below the sediment/basement interface at early times. A t 15000 years anhydrite takes on a more homogeneous distribution as it redissolves in the upper half of the basement section and reprecipitates in the lower half, leading to the formation of a second maximum near the bottom boundary. The redissolution of anhydrite causes a change in the slope of the calcium and sulfate profiles. The changes in the chemical profiles can be explained by the evolution of the flow and temperature fields. In the early phase of the evolution before the thermal plume has passed through the system, temperatures in the basement are near their initial values (Figure 2.5). The irregular flow leads to a relatively large degree of thermal homogenization. Consequently, recharging fluids experience a sudden increase in temperature as they cross the sediment/basement interface, leading to steep chemical gradients just below the interface. A s the plume works its way through the system, the cooling rate increases in comparison to the cooling due to the onset of convection. The magnitude and the gradients of temperature decrease within the recharge zone, causing a shift in alteration patterns and fluid concentration profiles. This shift has two effects: an expansion of the zone affected by alteration reactions typical of low to medium temperature (a result of the decreasing temperature gradients) and a shift in the location of maximum mineral precipitation/dissolution rates in the direction of flow (as favorable temperature conditions have also moved in the direction of flow). 3.2.3.2 Center of Domain (Reaction Zone) 184 The mineralogical profile through the center of the domain is nearly symmetrical with respect to the percentage of minerals precipitated and dissolved (Figure 3.6), indicating that fluids are buffered by the rock composition. The dominant reactions in the basement section (the reaction zone) are the alteration of plagioclase and clinopyroxene to a secondary assemblage consisting of clinozoisite, albite and small amounts of quartz and chlorite. In the sediments, the intensity of primary feldspar alteration increases with depth. Secondary phases include illite, albite, muscovite, clinozoisite, and quartz, with minor amounts of chlorite, anhydrite and calcite. The occurrence of chlorite, illite and anhydrite reflects the low rates of seawater percolation through the sediments. During the early state of the flow system (500 years, Figure 3.6), calcium-rich plagioclase precipitates in the lower half of the reaction zone and quartz and clinozoisite are present in the upper part (above about 850 m). A t this time, mineral phases (either precipitating or dissolving) show a peak just below the sediment basement interface. This peak may originate due to a combination of the compositional differences between the reaction zone and the sediments, the temperature differences, and possibly some mass transfer between the sediments and the basement. A t 5000 and 15000 years the mineralogical profiles through the center of the domain have roughly the same slope and the differences are mostly of a quantitative nature (the amounts roughly double). The profiles at these times show that the quartz/clinozoisite assemblage has shifted to the lower section of the reaction zone. In the sediments, the same alteration phases are present throughout the simulation time. Changes occur in terms of the relative proportions of these phases and the depth and extent of their occurrence. 185 The fluid concentration profile through the center of the domain (Figure 3.7), shows significant changes, both spatially and through time, in particular within the sediments. The profiles through the sediments can be divided into an upper and lower interval. In the upper interval, spanning the upper 300 m of the sediments, fluids are at or near seawater concentrations. Only silica shows a major concentration increase with depth. The second interval comprises the lower 200 m of the sediments. The transition between the intervals is marked by sudden drop in magnesium, sulfate, and bicarbonate. A t 500 years silica, potassium, sodium and calcium are not or only little affected by this transition. Sil ica continues to increase with depth. Both the calcium and the potassium profiles shows some minor highs and lows superimposed on a slight general increase towards the sediment/basement interface. Sodium concentrations are essentially uniform throughout the sediments. A t 5000 and 15000 years, the decline in M g , SO4, and HCO3\" coincides with an increase in potassium in solution. Calcium concentrations are nearly uniform throughout the lower interval within the sediments. The silica curve does not change significantly throughout the simulated period of time. The silica concentration at the bottom of the sediments is somewhat lower at 15000 years than at 500 years. The transition from the sediments into the basement reaction zone is marked by a sudden drop in potassium, sulfate, and bicarbonate in solution. Magnesium and iron increase; changes in silica, calcium, and sodium are less significant. Within the reaction zone, the compositional changes of the fluid are less dramatic than in the sediments and they generally reflect the trends of the mineralogical profiles. This suggests that the fluid in the reaction zone has lost all or most of its seawater components and is buffered by the rock composition. The snapshots of the concentration profiles at different times suggest 186 that the evolution of the convective system in the basement is only weakly reflected by changes in the shape of the profiles. The silica profile, for instance, shows a transition from a straight profile at 500 years to a slightly curved profile at 5000 and 15000 years, indicating the formation of a large convection cell with somewhat higher silica concentrations in the upper half (down-temperature flow) and lower concentrations in the lower half (up-temperature flow). The same trend is present in the potassium profile at 15000 years. Bicarbonate, which essentially acts as a tracer within the reaction zone, shows a concentration peak in the center of the reaction zone, within the isolated core of the convection cell. Bicarbonate accumulated in the reaction zone during the early state of convection, at relatively high rates of mass transfer between the sediments and the basalt, and is then maintained at the initial high concentrations in the hydrologically isolated core of the stable convection cell. The migration of seawater along the bottom boundary through the basement is indicated by a major compositional change of the fluids in the lowermost 50 m of the reaction zone. This interval is marked by elevated magnesium, sulfate, and iron, and decreased amounts of potassium, silica, and calcium. The steady increase of aqueous silica with depth eventually leads to quartz precipitation in the reaction zone. Furthermore, the increase with depth and the slightly curved concentration profile in the reaction zone suggest that the release of aqueous silica and/or the uptake of silica by quartz may be dependent on the temperature. Figure 3.8 confirms that aqueous silica in the reaction zone is indeed correlated with temperature. A s shown later (Section 3.2.3.5), the oversaturation with respect to quartz throughout most of the reaction zone suggests that this correlation is an effect of quartz buffering. 187 In summary, the geochemical profile through the center of the domain suggests that the alteration reactions in the sediments are controlled by the increasing temperature with depth in combination with low rates of seawater recharge. The transition from a low to higher degree of alteration occurs across a small interval with the removal of magnesium from the porewater through the precipitation of illite and chlorite at about 300 mbsf. The removal of magnesium from solution leads to a release of calcium and potassium into solution. The release of calcium is balanced by its removal through the precipitation of anhydrite, clinozoisite, and calcite. The latter causes a sharp drop in the bicarbonate concentration at 300 mbsf. A net increase in potassium concentrations occurs in the lower part of the sediments, despite its fixation in illite at lower temperatures and muscovite at higher temperatures. Because the fluids have obtained a higher temperature signature at the bottom of the sediments, some of the trends within the lower interval of the sediments continue into or are also present within the reaction zone. These trends include low concentrations of sulfate, bicarbonate, and magnesium. The abundance of clinopyroxene in the reaction zone leads to a recovery of magnesium concentrations at 15000 years. However, these concentrations remain considerably below the magnesium content of seawater. The lack of potassium-rich phases in the reaction zone leads to a sharp drop in potassium in solution. The controlling reactions in the reaction zone are the alteration of plagioclase to albite and clinozoisite, and of clinopyroxene to chlorite. A t 500 years, the distribution of alteration phases suggests a division of the reaction zone in an upper half with quartz/clinozoisite alteration and a lower half with the formation of a secondary calcium-rich plagioclase. This division is probably a result of the dominantly vertical flow during the phase of unstable fluid flow (Figure 2.7). The cooling of the 188 system leads to a shift in the occurrence of alteration assemblages in both the sediments and the reaction zone. For instance, the quartz/clinozoisite assemblage which occurs in the upper half of the reaction zone at 500 years has shifted downward at 5000 and 15000 years, replacing the zone of secondary plagioclase formation. The precipitation of quartz in the reaction zone leads to a temperature correlation of aqueous silica in the reaction zone. However, the low temperature gradients within the reaction zone do not affect the nature of reactions, the distribution of the alteration phases, or the fluid composition significantly. 3.2.3.3 Discharge Zone In the discharge zone (the zone of upward fluid flow, including the higher permeability window in the sediments and the basalt section beneath the window) the percentages of minerals precipitated and dissolved are roughly uniform throughout the depth of the basalt section (Figure 3.9). The dominant secondary mineral phases at 5000 and 15000 years are clinozoisite, albite, quartz, and chlorite. The amount of these minerals generally increases with time. At 500 years, these minerals are not yet present. Instead, a small percentage of secondary Ca-rich plagioclase precipitates. In the sediments the mineral pattern is more complex. The amounts of secondary quartz increase steeply from above the sediment/basement interface toward the seafloor. Near the seafloor, quartz fills almost the entire initial pore space after 15000 years. (Here it has to be reminded that due to the fixed temperature conditions at the seafloor, the 189 temperature gradient and thus the amounts of quartz precipitated are overestimated.) Other secondary minerals that occur in significant amounts within the sediments are albite, clinozoisite, and chlorite. A t 500 years anhydrite occurs below the seafloor, indicating initial downward flow of seawater into the discharge zone. A t the discharge zone, the fluid has obtained the characteristics of an endmember hydrothermal fluid, that is, it is depleted of all major seawater components, in particular M g and SO4, and has acquired the chemical signature of high temperature water/rock interaction (Figure 3.10). Throughout the discharge zone the main compositional changes of the fluids occur in the sediments, especially near the seafloor. These changes are due to a combination of both a compositional change of the rock as the fluid passes from the basalt into the sediments (e.g. reflected by the increase in potassium and bicarbonate) and the steep temperature gradients (the decrease in silica). In the basalt section, the compositional gradients in the fluids are lower. This behavior is a consequence of the high upflow velocities, low temperature gradients, and a fluid at or near equilibrium with the rock. The compositional changes of the fluid as it moves through the discharge zone towards the seafloor exhibit the same trends at all three times; 500, 5000, and 15000 years. This indicates that the fluid composition is not sensitive to the overall evolution of the convective system (except for the initial downward flow of small volumes of seawater, which causes anhydrite to precipitate below the seafloor). The fluid that exits the hydrothermal system (the vent fluid) is of particular interest as it reflects the integrated physical and chemical conditions along its flowpath and it is much easier to sample than porewater from within the reaction zone, for 190 instance. Figure 3.11a shows the compositional evolution of the vent fluid (taken from within the discharge zone 50 m above the sediment/basement interface to avoid boundary effects), as well as a comparison between these fluids and seawater. A s indicated by a tracer which was released into the seawater that enters the system through the recharge zone, the first seawater that has passed through the reaction zone arrives at the discharge zone at about 500 years. The first occurrence of the recharged seawater is the dispersed front of the 'seawater plume', however. The core of the plume arrives at the discharge zone at about 1400 years. The arrival of the core of the plume initiates the main compositional changes in the vent fluid. The fluid composition changes from that of the initial porewater in the reaction zone to a fluid that has undergone a significant loss in calcium and sodium, and a gain in bicarbonate, potassium, and sulfate. The loss in calcium and sodium can be explained by the uptake in anhydrite and albite, respectively. The increase in bicarbonate, as well as a small peak in the potassium concentrations, are most likely a result of the dissolution of calcite and K-bearing phases within the sediments of the recharge zone and the discharge zone. A s bicarbonate is essentially conservative in the reaction zone, its amounts may be regarded as a reflection of the 'sedimentary fingerprint' in the vent fluids. The increase in sulfate indicates that not all of it has precipitated as anhydrite along the flowpath. After the breakthrough of recharged seawater, the vent fluid composition shows a tendency to approach the initial porewater concentrations. Species that do not follow this trend are bicarbonate, sulfate and magnesium. These species increase to concentrations significantly higher than in the initial porewater. Bicarbonate may increase as a result of more calcite dissolution or less calcite precipitation in the sediments of the recharge zone 191 and the discharge zone, respectively The increase in magnesium may be due to magnesium buffering of clinopyroxene and/or a lower rate of chlorite or illite precipitation. The increase in sulfate may be attributed to the cooling of the system and the lower rate of anhydrite precipitation in combination with local anhydrite redissolution. Overall, after the breakthrough of recharged seawater, the compositional evolution of the vent fluid is smooth and appears to approach a steady state composition. The compositional changes reflect the gradual cooling and decreasing vigor of convection of the hydrothermal system. A comparison between seawater, modeled vent fluids, and vent fluids from Middle Valley (Site 858) indicates that many of the characteristic differences between seawater and vent fluids could be modeled successfully (Figure 3.116). These include elevated Ca contents, the depletion of M g and SO4, and the increase in silica. Not consistent with typical vent fluids are the low contents of potassium and iron. These underestimates may be a result of too low initial iron and potassium contents in the basalts in combination with an overestimate of the precipitation rate of K-bearing (illite, muscovite) and Fe-bearing phases. Figure 3.12 shows the distribution of the p H at 15000 years. Within the discharge zone a p H of approximately 5.5 is obtained, which matches the p H of the vent fluids at Middle Valley well (Butterfield et al. 1994). A s most dissolution/precipitation reactions are p H dependent, the close match between observed and modeled p H lends support to the representativeness of the model results. In summary, the processes and chemical conditions at the discharge zone are controlled by high flow rates, a low degree of fluid/rock mass transfer in the basalt 192 section, and a considerably higher degree of fluid/rock mass transfer in the sediments due to a compositional change of the rock and the steep temperature gradients near the seafloor. The evolution of the mineralogical pattern and fluid composition suggest that chemical conditions within the discharge zone change as soon as the initial seawater recharge passes through the discharge zone. During the earliest state, (probably immediately after the disturbance of the conductive, no-flow initial state) a short period of buoyancy driven sinking of cold seawater into the discharge zone is evidenced by small amounts of anhydrite near the seafloor. After fluid upflow is established within the discharge zone, the composition of the fluid is that of the initial porewater in the reaction zone. The only mineral precipitating in the basalt section during this time is Ca-rich plagioclase. The breakthrough of the core of the 'seawater plume' initiates the main compositional changes in both the solid and the fluid phase. The vent fluid composition slowly recovers from these changes and tends to approach the composition of the initial porewater. Species that do not follow this trend are bicarbonate, sulfate, and magnesium. 3.2.3.4 Lateral Flow into the Sediments The region surrounding the discharge zone is of particular interest, due to the chemical complexities induced by elevated temperatures, lateral flow of hydrothermal fluids into the sediments, and the downward percolation of seawater (Figure 2.9). Results from Middle Valley indicate the existence of reaction halos surrounding the discharge zone (Figure 1.5). A s described earlier, these halos are characterized by mineral 193 assemblages, which originate as a result of local temperature conditions and various degrees of mixing of seawater, hydrothermal fluids, and porewater. To assess the processes in the vicinity of the discharge zone, mineralogical and fluid composition profiles through the sediments are taken 30 m to the left of the discharge zone at 15000 years. Both profiles are plotted together with profiles through the sediments at the center of the domain in Figures 3.13 and 3.14. As pointed out earlier, the profiles through the sediments at the center of the domain can be divided in two intervals; one upper interval with low temperatures and low reactivity in which fluid compositions deviate only slightly from the seawater composition, and a lower interval of elevated temperatures where more dramatic compositional changes in the fluid phase occur. The alteration of the solid phase increases gradually from the seafloor to the bottom of the sediment section. These two intervals of distinct fluid compositions are absent in the region surrounding the discharge zone. Instead, fluid composition and mineralogical profiles are roughly uniform, except in the upper 100 m, where the profiles approach the fixed seafloor conditions. In addition to the different slope of the profiles, the degree of alteration is significantly higher near the discharge zone as reflected by the larger amounts of minerals dissolved and precipitated. The nature of the alteration reactions at the discharge zone is similar to the lower interval at the center of the domain, which is consistent with the high temperatures around the discharge zone. A n important difference is the lack of illite throughout the lower 350 m of the profile near the discharge zone. Illite precipitates and effectively removes magnesium near the seafloor (0-150 mbsf). The 194 cessation of illite precipitation leads to higher potassium contents below 150 m and the precipitation of muscovite. Sulfate is strongly depleted just below the seafloor. Si l ica contents and the amount of precipitated quartz are higher throughout the lower 350 m of the sediments near the discharge zone than at the center of the domain. Indirect evidence in the fluid composition profile for lateral flow of hydrothermal fluids into the sediments can be inferred from the lack of seawater components in the fluid composition profile and the almost complete lack of anhydrite in the mineralogical profile. The lateral flow of hydrothermal fluids is probably also the reason for the near uniform slope of the fluid composition and mineralogical profiles. In conclusion, major differences occur between the profiles through the sediments near the discharge zone and in the rest of the domain. These differences are primarily a result of a different temperature profile at these locations and lateral flow from the discharge zone into the sediments. 3.2.3.5 Alteration Patterns A s the system evolves from its initial state of high temperatures and highly irregular, vigorous fluid flow towards a stable convection cell and cooler conditions, the overall reactivity of the system, the magnitude of mass transfer between the solid and the fluid, and the redistribution of the mass throughout the domain decreases. It was demonstrated that the initial unstable convection can last for a significant period of time. But it can be expected that the flow and temperature fields obtain a stable pattern earlier 195 than geochemical alteration patterns in the solid phase. A s a consequence, alteration phases may be in disequilibrium with the flow field, the temperature conditions, and the fluid composition. This assumption wi l l be tested for a range of minerals. It is likely that minerals that precipitate in the recharge zone or the discharge zone are not as strongly affected by the transient flow field as the minerals that precipitate in the reaction zone. This is because the inflow zone and the outflow zone provide preferred pathways of fluid flow due to the permeability contrasts with the surrounding rocks, and are consequently only marginally affected the initial, irregular flow pattern. These minerals, which include anhydrite and chlorite, w i l l be primarily affected by the overall cooling and the resulting shift of the isotherms. The consequences of the shifting temperatures, the resulting redissolution of anhydrite in certain intervals, and the migration of the region of highest anhydrite precipitation were pointed out earlier. The reaction zone experiences alteration reactions that are dominated by a fluid that is close to or at local equilibrium with the rock. Therefore, compared to the recharge and discharge zones, the amounts of individual minerals precipitated or dissolved are relatively small. Nevertheless, the net total amounts of minerals precipitated or dissolved and the reduction or increase in pore space, may be significant enough to have an effect on the permeability distribution in the basement and the fluid flow pattern. In addition, the reactions in the reaction zone are spatially much more significant and w i l l therefore be important in the overall chemical budget. Secondary minerals that precipitate in the reaction zone in significant amounts are quartz, clinozoisite, albite, and chlorite. The 196 precipitation patterns (as volume change in percent) of these minerals are shown in Figures 3.15-3.18. Each figure depicts the mineral patterns for the model domain but the emphasis in the following discussion w i l l be on the reaction zone only. After 1000 years of simulation time, the quartz pattern in the reaction zone is very heterogeneous (Figure 3.15). The contours reflect the vigorous flow in high aspect ratio convection cells; quartz maxima and minima tend to occur in concentric, patch-like zones. The highest amounts of quartz occur near the sediment basement interface, in particular within regions of downward fluid flow. Large amounts of quartz precipitate also at the front of the thermal plume that begins to expand near the recharge zone. The lowest amounts of quartz occur in regions close to the bottom boundary. A t 5000 years, the contours are still undulatory and patch-like, the latter occurs near the sediment/basement interface. Three trends are beginning to become recognizable at this time; relatively large amounts of quartz precipitation at about midway through the basalt (at depth 700 m) in the left half of the domain, a decrease in quartz from right to left along the sediment/basement interface, and an increase in quartz from left to right along the bottom boundary. These trends are characteristic for the stable flow geometry, as indicated by the quartz pattern for the time interval between 10000 and 15000 years (lower panel, Figure 3.15). Thus, at 5000 years the quartz pattern still shows fingerprints of the initial, irregular flow conditions but these are increasingly overprinted by the quartz pattern which corresponds to the stable flow geometry. A comparison of the amounts of quartz precipitated within the two time intervals, up to 5000 years and between 10000 -15000 years, shows that in the left half of the reaction zone the amounts are roughly the same and in the right half, the amounts are considerably larger for the later time interval. 197 The geometry of the contours showing the distribution of albite precipitation at 5000 years is quite similar to that of the quartz-contours (Figure 3.16). However, unlike the quartz pattern, the largest amounts of albite occur along the sediment/basement interface and in the lower left quarter of the reaction zone. The percentage of albite precipitation passes through a minimum at about mid-depth (700-800 m). Thus, compared to the quartz alteration pattern at 5000 years, the gradients are reversed; where large amounts of albite precipitate, quartz occurs in small amounts and vice versa. The inverse correlation between quartz and albite is consistent with equation 3.53. This inverse correlation between quartz and albite is lacking in the time interval between 10000 and 15000 years, that is, after the onset of a single convection cell (Figure 3.16). The highest amounts of precipitated albite occur within a large region in the right half of the domain. Similar to the pattern at 5000 years, the amounts of albite decrease somewhat towards mid-depth (800-850 mbsf). From this region to the left and to the right, the percentage of albite decreases. To the left, the contours tend to approach the shape of the isotherms near the inflow zone, suggesting a weak temperature control on albite precipitation in this region. To the right, toward the outflow zone, the amounts of albite decrease over a much shorter interval and reach a minimum within the discharge zone. A comparison of the patterns of both time intervals, 0 - 5000 years and 10000 -15000 years, indicates that the region of maximum albite formation has shifted from left to right. The amounts of albite precipitated during these two intervals are remarkably similar despite the fact that the interval between 0 - 5000 years includes the early state of vigorous flow and high temperatures. This result suggests that the reaction kinetics of albite are not affected by the overall cooling of the system as it evolves from the initial to 198 the steady state. A s a consequence, the amount of albite in the system should be roughly proportional to its age. Interestingly, the albite pattern at 15000 years, which represents the time-integration of the shifting albite pattern, is almost symmetrical. The albite maximum occurs at x = 1000 m in the lower half of the basement section, and somewhat offset at x = 1500 m at the sediment/basement interface (Figure 3.16). In contrast to quartz and albite, the clinozoisite precipitation pattern is nearly identical for both time intervals (0 - 5000 years and 10000 - 15000 years) (Figure 3.17). The shape of the contours indicates a correlation with temperature. The amounts of precipitated clinozoisite are significantly greater for the earlier time interval, reflecting favorable conditions for clinozoisite during the early, highly reactive state of the system and a general dependence of the occurrence of clinozoisite on the temperature. Chlorite precipitation patterns are shown in Figure 3.18. Chlorite is the main sink for magnesium. It occurs in large quantities within and beyond the recharge zone where it precipitates by removing seawater magnesium. In the reaction zone it is predominantly an alteration product of clinopyroxene. The chlorite pattern in the reaction zone at 5000 years weakly indicates the pattern of the early state of flow by occurring in patch-like zones along the sediment/basement interface. These zones may be remnants of early-state upflow or downflow zones. To the left, that is, where the stable convection cell starts to spread through the basement, the transition to the stable flow field is recognizable. Note that at 5000 years the stable convection pattern has been active for more than 2000 years. Thus it appears that the transition from a geometry of alteration patterns that evolved at early times to patterns in accordance with the stable convection flow field, occurs first in 199 regions that are initially affected by the growing cell but with a time lag of several thousands of years. A t 15000 years the pattern is essentially the same as that between 10000 and 15000 years, which indicates that the remnants of the early state of the convective system have been overprinted. The pattern is characterized by vertical, slightly tilted contours that suggest a weak temperature dependency of the amounts of chlorite. Examination of the amounts of chlorite precipitated shows no significant change in the precipitation rates during the first 5000 years and during 5000 years of stable convection. The patch-like regions of chlorite that are present near the sediment/basement interface at 5000 years undergo redissolution after the stabilization of the flow pattern. In conclusion, the evolution of alteration patterns of individual minerals suggests that patterns that originate during the phase of vigorous flow and high temperatures may be preserved within the rock for a considerable amount of time. Thus there may be a significant time lag between the evolution of the flow and temperature fields and the alteration patterns. Consequently, a rock that is sampled from the oceanic crust may not represent the current chemical conditions at the sampling location but the reactive history at this location. Furthermore, the sluggish evolution of the mineral alteration patterns make a complete equilibrium between the mineralogy of the rock and the fluid unlikely. 3.2.3.6 Comparison of Alteration Patterns for Basement Permeabilities 1 xlff14 m2 and 5xl(T14 m2 It was discussed earlier that the transient state of the hydrothermal system consists of an early phase, during which fluid flow is highly vigorous and temperatures are high, 200 and a later phase, during which the flow field has stabilized to a single convection cell and the vigor of convection and the temperatures decrease toward a steady state value. It was shown in the mineralogical profiles that each phase can produce a separate mineral assemblage. Different assemblages therefore occur preferentially in regions of high initial temperatures, e.g. within the basalt section of the discharge zone or in the lower half of the reaction zone. In regions that have a lower initial temperature, the assemblage remains unchanged and only the proportions of individual minerals and the location of their occurrence may change. A second reactive transport simulation was performed for a basement permeability of 5 x l 0 \" 1 4 m 2 and a simulation-time of 5000 years with the intention to assess whether the higher permeability and the consequent different thermal evolution of the system can lead to recognizable differences in the mineral assemblages within the recharge zone, the reaction zone, or the discharge zone. The results are presented in Figure 3.19. A s details of the dominant alteration reactions have been discussed earlier, the focus here w i l l be on a comparison of mineral patterns. Within the recharge zone, the nature of the dominant reactions is the same for both permeabilities, that is the precipitation of anhydrite and chlorite (Figures 3.19 and 3.4). The percentage of these minerals is significantly larger for the higher basement permeability, particularly within an interval extending from the sediment/basement interface to about 750 mbsf. The amount of anhydrite and chlorite that precipitate suggest the recharge zone could be sealed at 550 mbsf. The large amount of these minerals may indicate a trend which was discussed earlier and is illustrated in Figures 2.2-2.4, 2.8 and 2.9. The size of the stable convection cell is dependent on the basement permeability. A 201 higher permeability leads to a larger cell, a narrower zone affected by the recharge of cool seawater in the basement, and consequently to higher temperature gradients around the recharge zone. Another difference between the mineralogical profiles in the recharge zone is the lack of illite within the basalt section at higher basement permeabilities. The lack of illite may be a result of the more favorable temperature conditions for chlorite precipitation and/or a lack of sufficient potassium in solution. A t the center of the domain (Figures 3.19 and 3.6), the mineralogical profiles are virtually identical for both basement permeabilities. Thus the mineral alteration pattern would not allow one to discriminate between the two basement permeabilities. The same can be said for the mineralogical profile through the discharge zone (Figure 3.19 and 3.9). In conclusion, only at the recharge zone are differences in the alteration pattern for both basement permeabilities clearly exhibited. Within the reaction zone and the discharge zone, the temperature differences are too small and the fluid too closely buffered by the rock composition to be able to recognize any distinct differences between the mineralogical profiles. In practice, however, it is not likely that the observed changes in the alteration pattern within the recharge zone can be used to draw general conclusions or apply constraints on the basement permeability or the vigor of convection in the basement as the alteration-patterns are dependent on a specific flow geometry and temperature distribution. It should be kept in mind, however, that circumstances may arise under which the amount of secondary minerals can completely seal off the recharge (and/or discharge) conduit. This would certainly have a significant impact the evolution of the hydrothermal system. 202 This conclusion probably holds for a greater contrast in permeability than half an order of magnitude. Considering the small differences in average basement temperatures for a range of basement permeabilities of two orders of magnitude (Figure 2.5), the metamorphic grade of the rock in the reaction zone is not likely to change over this permeability range. 3.2.3.7 Porosity Changes - The Impact of Alteration Patterns on Fluid Flow In the previous section the emphasis was on understanding transient patterns in individual minerals. In order to obtain detailed information on the impact of alteration patterns on fluid flow, the permeability change has to be determined and the flow field continually updated. The coupling of porosity change and permeability change and the feedback on the flow field was not performed in this study. Only the porosity change was calculated as the net volume change of all minerals dissolved or precipitated. Thus the porosity change is a rather crude assessment of the impact of alteration reactions on fluid flow, but some general insights can be gained. Porosity changes are calculated for the recharge zone, the discharge zone, and the reaction zone. Within the recharge zone, large quantities of anhydrite and chlorite precipitate just below the sediment/basement interface, as soon as seawater enters the hot basement. The porosity profile through the recharge zone in Figure 3.20 reflects the distribution of anhydrite in Figure 3.4, suggesting that anhydrite is the crucial mineral in terms of porosity/permeability changes in the recharge zone. The magnitude of the porosity change is rather small (about 1.6 % max. reduction). The transient nature of the anhydrite pattern 203 was discussed earlier. The effect of the shifting anhydrite pattern is a reversal of the trend in the porosity change from a porosity reduction to an increasing porosity in the upper half of the basement section at some point in time between 5000 and 10000 years. In the lower half of the basement section the porosity continues to decrease. Thus the alteration pattern in the recharge zone potentially induces rather long-term changes or fluctuations in the recharge rates into the basement. The profile of porosity changes through the discharge zone shows a slight increase in the porosity within the basalt section. In the sediment section, a porosity increase occurs in the lowermost sediments. The increasing porosity turns into a reduction of pore space above about 400 mbsf at 15000 years. The location of the reversal from a porosity increase to a decrease shifts downward through time. A t 500 years this reversal occurs at about 200 mbsf. Thus, through time the discharge zone becomes progressively sealed from top to bottom. Even though the large porosity reduction at the seafloor is in part a result of the overestimate of the temperature gradient due to the fixed temperature at the seafloor, the error in the magnitude of porosity reduction should not affect this trend. The evolution of the porosity in the reaction zone is best described considering the evolution in terms of the same two phases used earlier. During the first phase (500 years), the period of unstable convection, the contours of porosity change follow the evolution of the isotherms with a small time lag (Figure 3.21). A t 5000 years, the arrangement of the porosity contours is still irregular and patch-like in places, reflecting the effects of the initial state of the system. But the trend that w i l l be characteristic for the stable convective system is beginning to overprint the first phase; the up-temperature flow along 204 the bottom boundary leads to a porosity increase, the down-temperature flow along the sediment/basement interface leads to a reduced porosity. This trend is clearly exhibited in the porosity changes for the interval between 10000 and 15000 years. This trend of increasing porosity in the lower half of the reaction zone and decreasing porosity in the upper half of the reaction zone w i l l be self-enhancing. Theoretically, one would expect a more channeled and vigorous flow along the bottom boundary and a wider region of flow with lower flow velocities in the upper half of the reaction zone. However, the magnitude of the porosity change is quite low and the question arises whether these trends w i l l ever have a significant feedback on the flow field. Interestingly, the results show an overall net increase in porosity (the change of porosity in volume percent is positive) which is counter to the common belief that the porosity and the permeability of a hydrothermal system decreases with time as a result of alteration reactions. Moreover, the porosity increase is more significant during the later interval (10000 - 15000 years) than it is during the first 5000 years, in particular towards the discharge zone. Factors that may contribute to an overall porosity increase may be: (1) we did not account for all secondary minerals that may precipitate in the system (e.g. wairakite or other hydroxides), (2) the cooling rate of the system has been underestimated, for example, due to a decrease in heat supply, leading to the early precipitation of lower temperature alteration phases and potentially a decrease in the porosity, and (3) the pattern of hydrologically and chemically isolated reaction zone may only be short-lived or not established so that fluid exchange with seawater occurs in the reaction zone. 205 Even though there may be some uncertainty regarding the magnitude of the porosity change in the hydrologically isolated convection cell, it was demonstrated in the previous chapter that i f localized sites of seawater recharge and discharge are present, a single convection cell is the preferred pattern of convection. A fluid circulating in this convection cell is chemically at or close to equilibrium with the rock and only moderate amounts of mass transfer between the fluid and the rock can occur. Under these circumstances the hydrologic properties of the rock throughout much of the convection cell are not likely to change significantly. Our model suggests that over a period of thousands of years a significant porosity/permeability reduction due to chemical reactions occurs only in regions affected by the recharge of seawater and within the discharge zones, but regions in the basement that are isolated from the exchange with seawater or do not exhibit significant temperature gradients undergo only minor changes in the hydrological properties. Because the longevity of the patterns of convection remains an open question, many potentially transient effects (which could include, for instance, the formation of new pathways for seawater into or out of the reaction zone) are not accounted for in this model study. Thus, an overall permeability reduction in the oceanic crust consistent with the common belief could be obtained i f the effects of chemical alteration of the rock are integrated over a sufficiently long period of time. In summary, the effects of mineral alteration on the porosity are most significant in the recharge and the discharge zone. The trend within the discharge zone is a progressively sealing of the conduit from top to bottom. Within the recharge zone, the effect of the porosity reduction depends on the distribution pattern of anhydrite. The effect may range from a partial clogging of the recharge zone with some temporal 206 variations in the porosity due to the partial redissolution of anhydrite, to a complete clogging (as was observed for a permeability of 5x10~ 1 4 m 2 ) . Within the reaction zone the pattern of the porosity change reflects the heterogeneity induced by the initial unstable state of the system. But the magnitude of porosity change is low and the overall effect on the flow field, even though self-enhancing, may not be significant over the simulated period of time. 3.3 Conclusion In this chapter, the complex coupling and interaction of fluid flow, temperature, and chemical reactions in a transient, oceanic hydrothermal system was modeled. Despite the simplifications underlying our modeling approach, we were able to simulate many of the trends, processes, and observations from Middle Valley and other oceanic ridges. In the recharge zone, large amounts of anhydrite and chlorite precipitate, causing a depletion of magnesium and sulfate and an increase of calcium within the solution. In the reaction zone, the alteration reactions include the formation of albite, clinozoisite and chlorite. The fluids in the reaction zone become buffered by the rock composition and obtain the signature of an endmember hydrothermal fluid, which has lost the characteristics of seawater. Within the discharge zone, the rapid cooling of fluids near the seafloor leads to extremely high chemical gradients in the upper sediment section. Large amounts of quartz precipitate in the upper 300 m of the discharge conduit. Alteration 207 reactions in the sediments overlying the basalt are dominantly temperature controlled. The low rates of seawater percolation lead to only small contents of seawater-derived alteration products, such as anhydrite. Elevated temperatures and lateral flow cause anomalies in the porefluid and mineralogy profiles through the sediments surrounding the discharge zone. The sedimentary component in the hydrothermal fluids is insignificant. We successfully modeled characteristic differences between seawater and vent fluids, including elevated silica and calcium, depletion of magnesium and sulfate, and a p H of 5.5 in the vent fluids. Inconsistencies between observations and modeling results include too low concentrations of potassium and iron in the modeled vent fluids. Metamorphism in a hydrothermal system within the oceanic crust has an important impact on the evolution of this hydrothermal system in terms of thermal history, vigor of convection, and the pattern of fluid flow. It was shown that the evolution of alteration patterns occurs at a much slower pace than the evolution of the temperature and flow fields. This behavior is a mainly a consequence of the cooling of the system and the decreasing reactivity within the system. A s a result, mineral alteration patterns may contain products of the initial high temperature, highly reactive phase of convection, causing a disequilibrium of the mineral phases with the current temperature conditions and the composition of the pore fluids. The impact of chemical alteration reactions on the flow system are most significant in regions where highly concentrated fluids are far from equilibrium with the rock and/or in regions of steep temperature gradients. These regions occur at the recharge zone and the discharge zone of the hydrothermal system. A t the recharge zone, seawater 208 enters the high temperature basement and large amounts of anhydrite and chlorite precipitate, reducing the pore space and the recharge rate. A t the discharge zone, quartz precipitation leads to a rapid decrease in the pore volume near the seafloor and an almost complete clogging of the discharge conduit. Despite partial redissolution and shifting of the alteration assemblages, the general trend within the recharge and the discharge zone is to a reduction of recharge and discharge rates which may ultimately cause the transition from an open to a closed hydrothermal system. This in turn would lead to the onset of smaller aspect ratio convection and the perturbation of the hydrologically isolated region. The rate and magnitude of porosity/permeability changes in the recharge and discharge zones are considerably larger than the porosity/permeability changes in the reaction zone. Here, the fluid is at or near local equilibrium with the rock and the mass transfer between the fluid and the rock and the potential for porosity changes is rather low. A s a large region of the reaction zone is hydrologically and geochemically isolated (Figure 2.9) no mass enters or leaves this region and only redistribution of mass occurs. The fluid circulation within a single convection cell leads to an increase in pore space in the flow direction in the lower half of the reaction zone, and a decrease in the pore space in the flow direction in the upper half. The calculations indicate an overall small net porosity increase within the reaction zone at 15000 years. It appears that a more significant porosity change in the reaction zone can only be achieved i f the isolated convection pattern is perturbed and fluids that are farther from equilibrium with the local chemical conditions interact with the rock. For instance, the processes in the recharge zone suggest that i f fluids that contain a large fraction of seawater components enter and 209 are being distributed throughout the reaction zone, a significant and large-scale porosity reduction could occur. 210 Initial Amounts Reactive Surface Area (m 2 /m 3 ) Dissolution rate constant at 25 °C (m-V 1 ) Activation Energy (kcal mol\" 1) Equilibrium constant (log K ) at 25 °C Sediments Calcite 15% 10 l .OxlO\" 1 2 5 1.979 Sediments Quartz 30% 10 4 .3x l0\" 1 4 18 -3.971 Sediments An25Ab73.5Ori.5 20% 10 l .OxlO\" 1 2 5 8.186 Sediments An 5 5 Ab 4 3.50ri .5 10% 10 l .OxlO\" 1 2 5 15.179 Sediments K-feldspar 15% 10 l .OxlO\" 1 2 5 -0.876 Basalt An 5 5Ab 43.50ri.5 5% 10 l .OxlO\" 1 2 5 15.179 Basalt An 6 5 Ab 3 40r i 20% 10 l .OxlO\"1 2 5 17.481 Basalt An 7 5Ab240ri 25% 10 l .OxlO\" 1 2 5 19.757 Basalt Clinopyroxene 40% 10 l .OxlO\" 1 2 5 21.117 Secondary Minerals Illite l .OxlO\" 1 2 5 7.0602 Secondary Minerals Muscovite l .OxlO\" 1 2 5 11.481 Secondary Minerals Anhydrite l .OxlO\" 1 2 5 -4.089 Secondary Minerals Fe-chlorite l .OxlO\" 1 2 5 50.61 Secondary Minerals Mg-chlorite l .OxlO\" 1 2 5 66.117 Secondary Minerals Clinozoisite l .OxlO\" 1 2 5 41.248 Table 3.1: Summary of the composition and the kinetic and thermodynamic parameters of the solid phase 211 Species Initial Concentration (mmol/kg) cr 540 S0 4 \" 2 28 Si02,aq 0.18 N a + 460 K + 10 M g 2 + 53 C a 2 + 10.28 Fe 2 + 5xl0\" 6 Ali+ lx lO\" 1 2 HCO3\" 1.5 pH 7.5 Secondary Species: OH\", NaCl(aq), M g C l + , CaS04(aq), KS0 4 \" , C a H C 0 3 + , CaC0 3(aq), MgC0 3(aq), M g H C 0 3 + , CaCl + , CaCl2(aq), C0 3 \" 2 , FeCl + , FeCl2(aq), KCl(aq), C0 2(aq), A10H + 2 , HSO4\" Table 3.2: Major ion composition of seawater (after Butterfield et al. 1994 and Davis et al. 1992b) and secondary species. 212 log permeability CD - 1 ? - 1 6 - I S - U - 1 3 -12 chemical water/rock ratio 10\"1 1(P 101 102 103 1Q4 1 0 5 Permeability data from various O D P Boreholes, including Hole 858G from Middle Valley. Chemical water/rock ratios based on samples from: D S D P Hole 396B (Bohlke et al. 1981) Hole 504B ( A l t e t a l . 1986a) D S D P Sites 261 and 462 (Hart and Staudiegel, 1986) Holes C Y - 1 and C Y - 1 A in the Troodos ophiolite (Gill is and Robinson, 1990) Figure 3-1. Compilation of chemical water/rock ratio versus depth measurements (after Fisher, 1998) and comparison with a collection of permeability versus depth data from various ridges (From: Davis and Fisher, 1994). 213 Fluid of Seawater 1 Sediment 30% Quartz 20% A n 2 5 A b 7 3 50r,; 5 - 10% An 5 5 Ab 4 3. 5 Or 1 . 5 15% Calcite • 1 - j 15%K-feldspaf Basal l 5% 'An55Ab4.-5.5Or1 40% Clinop\\ roxene 2000 m Secondary minerals: Anhydrite, Mg-Chlorite, Fe-Chlorite, Albite, Illite, Muscovite, Clinozoisite Initial conditions: Porewater in equilibrium with the rock Porosity 10% (updated during simulation) Reactive surface area: 10m 2( mi n erai)/m 3(rock) (updated during simulation) Flow/temperature fields: 300 years into transient flow simulation Figure 3-2. Chemical model domain - solid phase and fluid composition and initial and boundary conditions. 214 Recharge zone Center of Domain Discharge zone S E D I M E N T S 1 ' \" iii!\" t< S P B A S A L T R l ' \\ ( T K ) \\ 2000 m Figure 3-3. Profile locations. 215 Volume% 0.2 0.4 0.6 O.f .200 .400 f a •600 .800 .1000 . 1 0 0 0 Volume% — Quartz — Anhydrite Calcite — Clinochlore — Daphnite — • Albite — - An25 - - An55 — An65 •• An75 — K-feldspar Muscovite •— Illite • • • Augite — Clinozoisite -1000 Figure 3-4. Mineralogical profiles through the recharge zone at various times (units in %volume change) 216 500 years moles/kg 5000 years moles/kg le-5 le-4 le-3 le-2 le-1 le-0 -200 -400 -600 -800 -1000 0 -200 I i III -400 \\J III A// v (4.6) Rearranging equation 4.1 and substituting equation 4.4 and 4.6 gives v = 'S±S^n—ZLL (4.7) which relates the fluid velocity to the mass transfer between the solid and fluid phase and a proportionality factor which includes physical properties of the fluid and the rock as well as the length of the flowpath and the time interval over which ACj(r) is being estimated. 239 Reformulating equation 4.7 by separating out the information on the geochemistry and the length of the flowpath we can write AC, AC, L = (4.8) where the expression on the left-hand side is the chemical w/r ratio weighted by the length of the flowpath. Note, in the following calculations the expression on the left-hand side will be referred to as the chemical w/r ratio. The expression on the right is the physical w/r ratio for an open, flow-through system. Note that the expression L/ACj(W) is the inverse of the concentration gradient over the length of the streamtube. Equations 4.5 to 4.8 show that there is a linear relationship between w/r ratios, the fluid velocity, and the permeability of the medium. Underlying equations 4.7 and 4.8 is the assumption that the flow system is isothermal and that A is uniform along the streamtube. In a non-isothermal system p(W) and q will vary along the flowpath. Thus, in a manner similar to the mass loss or gain in the solid phase and the properties of the solid phase, q and p(W) are regarded as averages over the flowpath. The magnitude of the error introduced by this simplification depends on the temperature difference between both ends of the streamtube; the larger the temperature difference the larger the error. In this study, various temperature conditions along the flowpath are tested and it will be shown that despite the simplifications, equations 4.7 and 4.8 provide reasonable results. 240 A n prerequisite for applying this method is a basic understanding of the large-scale flow geometry in order to identify appropriate sampling or measurement locations. It is not essential to know the direction of flow but rather to be able to recognize regions with a consistent flow geometry. Estimates of the fluid velocity obtained from equation 4.7 are averages in time and space. This equation is based on the assumption that the system is at a hydrological and geochemical steady state. The fluid velocity is constant in time throughout the region of flow and only advective mass transport is considered. Geochemical steady state implies that the net mass transfer of an element between the liquid and solid phase at any location along the streamtube is constant. This means that the average mass loss or mass gain of a species in the solid phase varies in a linear manner and the concentration gradient of this species in the aqueous phase over the length of the flowpath remains constant. Hydrodynamic dispersion, which occurs as a result of mechanical mixing and molecular diffusion, is also not accounted for in equations 4.7 and 4.8. The magnitude of the error introduced by neglecting the dispersion of solutes in the mathematical model is a function of the dispersivity of the medium, the flow velocity, and the magnitude of molecular diffusion of the species under consideration. The significance of hydrodynamic dispersion on the accuracy of w/r ratio calculations will be assessed below. In order to reduce the effect of measurement inaccuracies, it is advantageous to have a large net mass transfer of a species between the fluid and solid phase. In addition, favourable reactions are those which produce a monotonic slope of the concentration gradient along the streamtube. The magnitude of the mass transfer is a function of the 241 reaction kinetics and the fluid residence time R. A n expression for the fluid residence time can be formulated by rewriting equation 4.8 as: R = ACi(w)p(w)At (4.9) A C , A t uniform and constant reaction rates, the fluid residence time must be sufficient to allow for a measurable mass transfer between the solid and the fluid phase but short enough to avoid thermodynamic equilibrium between the two phases. A t thermodynamic equilibrium no mass transfer between the solid and fluid phase would occur such that and the calculation of w/r ratios would be invalid. It has been assumed that many high temperature, highly reactive chemical systems such as the one described in this study are likely to be at thermodynamic equilibrium (e.g. von Damm, 1995). In non-isothermal flow-through systems, however, the issue of residence time versus reaction kinetics becomes less important for w/r ratio calculations as the length of the flowpath can be chosen such that the fluid moves across a temperature gradient that provides for a sufficiently large mass transfer in the direction of flow. A C / ( R ) = A C , = 0 (4.10) 242 4.2 Application of the Method 4.2.1 One-Dimensional Simulations The first set of simulations is aimed at testing the proposed approach and understanding the geochemical processes that occur as a fluid of initial seawater composition passes through a solid medium of basaltic composition under non-isothermal conditions. A summary of the hydrologic and chemical parameters used in the one-dimensional simulations is given in Figure 4.1. The calculations are performed for a column with an arbitrary cross-sectional area of 20 m 2 and a length of 3120 m, for a period of time of 15000 years. A steady state flow field with a uniform fluid velocity of 10 m/year is specified. The magnitude of the velocity is in the upper range of flow velocities in the basement rocks and in the lower range of recharge or discharge velocities calculated for the two-dimensional convective flow field for a basement permeability of lx lO\" 1 4 m 2 (Chapter 3). With impermeable boundaries along the long sides of the column the flow field is one-dimensional. A longitudinal dispersivity of 10m is specified to account for mechanical dispersion along the flowpath. The column is filled with a solid medium, which is hydrologically and geochemically homogeneous. Its initial composition is that of an idealized basalt consisting of plagioclase and clinopyroxene. The plagioclase and clinopyroxene are represented as ideally mixed solid solutions. The compositional varieties of plagioclase are an 5 5, an65, and an 7 4; clinopyroxene is specified with an Fe:Mg ratio of 1:6. The 243 secondary minerals, quartz, chlorite (clinochlore, daphnite), clinozoisite, anhydrite, albite, and illite are allowed to precipitate. The basalt is modeled as a uniform porous medium with a porosity of 10%. For simplicity, redox processes are ignored in the reactive transport calculations. Clinozoisite is used to approximate epidote as an alteration phase. A t t = 0, the porewater in the column is in local equilibrium with the solid phase composition. The fluid entering the column is of seawater composition. In all mass transfer calculations the focus w i l l be only on the primary species (i.e. species that constitute the basic building blocks of the system) M g , Ca, Fe, Na , K, Si02,aq, and SO4 (Table 3.2). CI, A l , and HCO3\" are defined as a primary species but are not included in the mass transfer calculations. A range of secondary aqueous species which are summarized in Table 3.2 is allowed to form as well . Aqueous concentrations of the primary species in the following discussion are presented as total concentrations. Note that the neglect of redox processes w i l l certainly have an affect on the calculated fluid compositions, in particular for Fe and SO4 concentrations in solution. However, i f S 0 4 is depleted by anhydrite precipitation before the fluid reaches the high temperature regime (>250°C) where sulfate reduction can take place, the impact of neglecting redox processes involving sulfate may be small. Because of the low concentration of iron in solution, the overall effect on the system due to the neglect of redox processes involving iron is probably insignificant. Simulations are performed for non-isothermal conditions which are designed to simulate the thermal history of a fluid passing through the oceanic crust, that is through the recharge zone, reaction zone, and discharge zone. Note that in this simple model the 244 fluid flow regime has no influence on the imposed temperature distribution. The temperature conditions divide the column into three intervals, each representing a different thermal regime in a typical mid-oceanic ridge setting. The three intervals are characterized by different temperature values and gradients and are designated as T l , T2, and T3 (Figure 4.1). The temperature gradient is uniform throughout each interval. The temperature range and the length of each interval are loosely constrained by conditions at Middle Valley. The choice of a constant temperature gradient is somewhat arbitrary but should not affect the validity of the results. The first interval (Tl) extends from 0 to 580 m and represents the recharge zone in which the fluid is heated to a basement temperature of 290°C at a rate of 0.5°C/m. Once being heated to 290°C the fluid enters a 2000m long interval (T2) in which it is slightly cooled from 290°C to 270°C at a constant rate of 0.01°C/m. This interval represents the near isothermal, high temperature conditions encountered in the reaction zone at Middle Valley. After passing through the high temperature interval the fluid enters the third interval (1580-3120 m) (T3) which simulates the processes in the discharge zone by cooling the fluid to ambient seawater temperature (1°C ) at a rate of 0.5°C/m. The temperature affects the density of the fluid but does not feed back onto the velocity field. Also porosity and permeability changes by mineral precipitation or dissolution are assumed to not affect the velocity field. Note that in order to facilitate convergence of the solution of the reactive transport calculations, we extended the column to 3200 m and specified the fluid concentration to be at equilibrium with the rock at the end of the column. The additional interval of 100 m has a uniform temperature of 1°C. It was truncated before any analysis and mass transfer calculations were carried out. 245 4.2.1.1 Results of Reactive Transport Calculations In order to assess the suitability of a species for w/r ratio calculations the controlling reactions that give rise to its concentration distribution in the aqueous phase and its loss or gain in the solid phase need to be understood. For example, it was mentioned earlier that concentration profiles with a monotonic slope are one prerequisite for accurate w/r ratio calculations. The understanding of chemical conditions along the flowpath can help identify those sections of the flowpath in which these conditions are likely to occur. In addition, the fluid that enters the column is in disequilibrium with the rock and, while passing through the column, undergoes continuous compositional changes that are a function of both physical conditions, such as fluid velocity and temperature, and chemical conditions such as the fluid's starting composition and the composition of the rock. This implies that the nature of the alteration reactions and thus the shape of the concentration profile does not necessarily correlate with the temperature conditions in the column but could also be dependent on compositional changes in the fluid or the rock. Therefore, in addition to dividing the column into intervals characterized by certain temperature conditions (i.e. in this case the magnitude of the temperature gradient), it is useful to identify regions of consistent alteration reactions which allow predictions on the behaviour of the species in solution along the flowpath. Aqueous concentration and mineralogical profiles along the column are illustrated in Figures 4.2 and 4.3, respectively. We consider first the species total aqueous 246 concentrations at 15000 years and the percent volume changes of the minerals over a period of 15000 years. A time period of 15000 years corresponds to nearly 50 pore volumes of seawater migrating through this column. Concentrations at 500 years shown in Figure 4.2 are discussed later. The concentration profile shows that over the first 300 m (up to a temperature of about 150°C) no significant changes occur and the aqueous concentrations remain similar to initial seawater levels. Beyond 300 m, sulfate, calcium, potassium and magnesium contents decrease. Sulfate in solution decreases to near-complete removal at about 900 m. Calcium concentrations decrease in parallel to the depletion of sulfate, pass through a minimum at about 500 m, and then increase to almost four times the initial seawater concentration in the middle third of the column. Potassium and magnesium concentrations decrease strongly between 300 m and 1100 m and 1700 m, respectively. Potassium shows a slight recovery towards the end of the column with a weak maximum at 2750 m, but generally its concentration remains much lower than that in seawater. Magnesium levels remain low and do not show significant changes past 1700 m. The sodium concentration profile shows a concentration maximum at 1100 m and a minimum at 1750 m. Beyond a section of increasing concentration that occurs between 1750 m and 2580 m the sodium concentration levels off. Final amounts of sodium in solution are slightly higher than in seawater. Iron concentrations increase to a maximum at about 800 m and gradually decrease toward the end of the column. The aqueous silica concentration profile is nearly symmetrical with an exponentially-shaped increase in interval T l and a similar decrease in interval T3. In interval T2 the silica concentration profile shows a slight depression between 1100 m and 1450 m with a weak local minimum at 1200 m. With the exception of this depression the silica profile shows a 247 slight, almost linear decrease in T2. The linearity of the silica profile at high temperatures suggests a temperature control on the amount of silica in solution. The mineralogical profile in Figure 4.3 represents the net amount of minerals precipitated or dissolved expressed as a volume percent over the simulation time of 15000 years. Negative values are mineral volumes dissolved, positive values are mineral volumes precipitated. A l l primary minerals (i.e. clinopyroxene and plagioclase) are plotted in black in Figure 4.3. The volumes of dissolved primary minerals increase with increasing temperature in T l and decrease with decreasing temperature in T3. Across interval T2 the dissolved amounts of clinoyroxene and plagioclase show for the most part a steady, near-linear decrease that can be correlated with the slowly decreasing temperature. A small, step-like drop in those amounts occurs at 1100 m and 1650 m for plagioclase and clinopyroxene, respectively. Overall, the volumes of primary minerals that are dissolved appear to be dominantly temperature controlled. In contrast, the distribution of precipitated secondary minerals is much more irregular. The profile suggests a division of the column in (at least) three segments, each characterized by a distinct alteration mineral assemblage. The interval containing these assemblages will be labeled according to Figure 4.3 as M l , M2, and M3. The alteration assemblage in M l , which extends from 0 to about 1100 m, consists of chlorite (clinochlore/daphnite), anhydrite, quartz, and illite. The second interval (M2), extending from 1700 m to 2580 m, contains a group of minerals comprising mostly albite, clinozoisite, chlorite and quartz. In interval M3, which coincides with interval T3, calcite, illite, quartz, and albite precipitate. The transitions of these assemblages are either gradual 248 and marked by the steady removal of species from solution and the concomitant release of different species into solution, such as between M l and M2, or the transitions can be rather abrupt and correlate with sudden changes in the temperature conditions such as between M2 and M3. The change in alteration mineralogy and the end of interval M l at 1100 m is triggered by the removal of seawater sulfate by anhydrite precipitation and the depletion of potassium by its fixation in illite. It indicates the transition of the fluid from that of modified seawater toward that of an evolved hydrothermal fluid that is in the process of becoming buffered by the rock composition. Anhydrite formation occurs as soon as the fluid of seawater composition reaches a temperature above about 120°C (at 230 m) and it continues until virtually all sulfate is removed from solution (at about 1100 m). This is an important process because it limits the Ca concentrations in the fluid, inhibiting the formation of Ca silicates in the mineral assemblage, and limiting the amount of seawater sulfate reaching the high-temperature portion of the system, where sulfate reduction can take place. Consequently, the cessation of anhydrite precipitation at 1100 m is followed by the first appearance of clinozoisite, a calcium silicate. Seawater potassium is largely removed by illite precipitation at 1100 m, but potassium levels increase again in the high temperature interval as plagioclase dissolves and releases small amounts of potassium. The fixation of potassium in illite coincides with an increase in sodium in solution and a rather sudden appearance of albite as an alteration phase as soon as illite ceases to precipitate. The formation of albite consumes aqueous silica according to equation 3.56, causing the solution to drop below quartz 249 saturation and quartz to be removed from the alteration assemblage at 1100 m. The occurrence of quartz in M l is limited to only a short segment between about 700 m and 1100 m. At 1350 m quartz reappears and becomes an important alteration phase due to a decline in the precipitation rate of albite and an increase in the amount of precipitated clinozoisite. In M2 the amount of precipitated quartz gradually decreases but shows a distinct peak at the transition to the interval M3/T3 in which the fluid undergoes rapid cooling. Up to about 1700 m chlorite constitutes an important alteration phase but it continues to precipitate in lesser amounts up to 2800 m. Chlorite formation is the main sink of magnesium in solution. The similar but opposite slopes of the magnesium and calcium concentrations profiles between 600 m and 1700 m suggest a balance between Mg uptake and Ca release. At about 1700 m all magnesium of seawater origin is removed from solution and Ca concentrations decrease slightly before leveling off, indicating the termination of the exchange of these elements. The small amount of magnesium that still occurs in solution is balanced by the rates of clinopyroxene dissolution and chlorite precipitation as indicated by the inverse correlation between the clinopyroxene and chlorite profiles. The point at which seawater magnesium is removed (1700 m) marks the transition to what can be called an endmember hydrothermal fluid which has lost all the characteristics of seawater. The endmember hydrothermal fluid is buffered by the mineral assemblage M2 which is composed of quartz, clinozoisite, albite and some chlorite. Mineral buffering is indicated by linear, monotonic slopes of the mineral and the aqueous concentration profiles in this interval. The linearity of the profiles suggests that mass transfer between the solid phase and the endmember fluid is only controlled by the temperature and the fluid residence time. While seawater magnesium is removed, cation 250 exchange between the solid and the fluid phase continues. For instance, the removal of calcium from solution appears to be coupled with the release of sodium into solution, as indicated by inversely correlated slopes. The magnitude of exchange is lower than that between seawater magnesium and calcium however, as both the calcium and the sodium profiles do not exhibit significant changes. The beginning of interval T3, at 2580 m, is marked by the steep decline of the temperature to seawater levels and the mineral assemblage M3, characterized by the transition of the high temperature mineral assemblage to lower temperature mineral phases such as calcite and illite. Potassium concentrations, which have increased along most of the high temperature interval, have reached a level sufficiently high to allow small amounts of illite to precipitate. The modeling results are similar to the results obtained from the two-dimensional simulations discussed in the previous chapter. The evolution of seawater towards an endmember hydrothermal fluid that is buffered by the mineral assemblage of the rock is mainly controlled by the exchange of species, i.e. the removal of seawater sulfate and magnesium, and the release of elements such as calcium from the rock. This process is consistent with predictions from experimental reaction of seawater with basalts (e.g. Mottl, 1983; Seyfried, 1987; Seyfried et al. 1988). The alteration reactions that include silicates lead to the release or the fixation of aqueous silica. Depending on the net mass transfer of silica the solution can be either over or undersaturated with respect to quartz. Significant differences between the model calculations and field observations from ridges or experiments occur in the behavior of iron and potassium. Both elements are greatly 251 enriched in endmember hydrothermal fluids. Our calculations indicate that iron increases to significant concentrations only from 0 to 900 m, beyond which it decreases again to about seawater levels. Iron concentrations in the system are governed by a balance between clinopyroxene dissolution and chlorite precipitation. The low levels of iron in solution beyond 900 m indicate that all of the released iron is taken up by chlorite. Similarly, potassium concentrations decrease to very low levels in M l due to its fixation in illite and recover to somewhat higher levels throughout T2, but generally potassium remains far below seawater levels. The increase at high temperatures is consistent with experimental results which indicate the release of potassium from basalt at temperatures > 150°C (Seyfried and Bischoff, 1979; Seyfried, 1987). One may speculate whether the discrepancy between the magnitude of observed and our calculated potassium concentrations in hydrothermal fluids is an effect of the length of the column, the temperature or release rate, the potassium contents in the feldspars, or a combination of those effects. Nevertheless, despite the inconsistency in terms of concentration values, the trend of the potassium concentration profile appears to be consistent. The strong disequilibrium between the fluid of seawater composition and the rock of basaltic composition in interval M l leads to a substantial precipitation of secondary minerals, in particular anhydrite, chlorite, and illite. A s the fluid evolves towards its endmember composition and becomes more and more buffered by the rock composition, the amounts of minerals dissolved or precipitated decrease. A s soon as seawater sulfate and magnesium are removed, the amounts of secondary minerals formed are rather moderate. When the fluid enters interval T3 the precipitated volumes of most secondary 252 minerals show a small initial peak and tend to decrease rapidly as the temperature approaches that of seawater. A question of interest for the calculation of w/r ratios is whether the hydrothermal fluid reaches an equilibrium or a steady state composition with respect to the minerals in the system. As our system is non-isothermal a thermodynamic equilibrium is unlikely. Provided that the composition of the solid phase is homogeneous, the source composition of the fluid is constant, and flow and temperature fields are at steady state, the answer to the question of whether local equilibrium or steady state is present depends on the velocity at which fluids pass through the system, that is the fluid residence time, and the temperature-dependent reaction rate of the mineral. Local equilibrium is more likely at high temperatures, low temperature variations, and low fluid velocities. In contrast, geochemical steady state can occur wherever or whenever fluid flow and temperatures are at steady state. Steady state may be more likely than local equilibrium at high fluid velocities and steep temperature gradients. To determine whether the species concentrations are at local equilibrium or steady state, the concentration profiles through the column shortly after one pore volume has moved through the column (500 years) are added to Figure 4.3 and shown in grey. If a concentration profile shows a significant change in slope through time, the species concentration would not be at steady state. The concentration profiles are only weakly shifted from left to right over time and this shift is largely restricted to the high temperature interval (T2) of the column. The shift is small enough to conclude that the chemical system within the column is essentially at steady state. To assess whether the 253 solution is at local equilibrium with a mineral, the mineral saturation states were determined. The saturation indices of the mineral indicated that in intervals T l and T3 of the column, that is the high temperature gradient intervals, the solution is far from equilibrium with any of the minerals in the system. This result indicates that the fluid residence times are too short and temperature conditions change too rapidly to allow the solution to equilibrate, so that only steady state concentrations can be obtained. In interval T2, at high temperatures and a low temperature gradient, the solution reaches equilibrium with anhydrite and quartz. The anhydrite equilibrium is maintained by the increase in Ca in solution in exchange for the uptake of seawater magnesium. It constrains the sulfate concentration to be at low levels. Anhydrite becomes undersaturated as soon as calcium concentrations level off beyond 1700 m. The solution reaches equilibrium with quartz at about 750 m which is well within the high temperature interval. Between 1100 and 1350 m the silica in solution drops below quartz saturation due to its uptake by albite but reaches saturation levels again between 1350 m and 2580 m. Quartz saturation implies that the aqueous silica is buffered by local equilibrium with quartz. This explains the temperature correlation of aqueous silica concentrations where quartz saturation is present. Beyond 2580 m, the fluid cools rapidly to seawater temperature and the large amount of silica in solution leads to quartz oversaturation. Similar to interval T l , the residence times are too short to obtain local quartz equilibrium and silica levels are at steady state. In summary, the results confirm that if a system is at a hydrologic steady state and if the source composition of the fluid and the primary composition of the rock are constant, the species' concentration profiles through the column are at or close to steady 254 state. Furthermore, the silica concentration profile tends to correlate with the temperature profile. In general, how well the temperature and silica profiles correlate depends on the saturation state of the fluid with respect to quartz. The best correlation occurs at quartz saturation, which is likely to occur at high temperatures and low temperature variations and/or long fluid residence times such as in interval T2. A t over - or undersaturation, the strength of the correlation depends on the source concentration, that is, how far from equilibrium the silica level is at the source, and how quickly the fluid undergoes temperature changes while traveling along the flowpath. The latter is a function of the fluid velocity and the steepness of the temperature gradient. 4.2.1.1.1 Effect of Dispersion A s solutes are transported through a porous or fractured medium both mixing during fluid advection and molecular diffusion cause the solutes to spread along the flowpath leading to dilution. This spreading phenomenon is called hydrodynamic dispersion and it w i l l influence the concentration profiles of the aqueous species along the flowpath. Dispersive mixing is commonly related to the dispersivity of the porous medium and the fluid flow velocity according to equations 3.6-3.8. Diffusive transport is proportional to the concentration gradient (equation 3.10) and its effects are of importance only at low velocities. In our simulations, diffusion can be regarded as negligible so that dispersion is mainly a result of mixing and thus a 255 function of the dispersivity and the flow velocity. The consequences of dispersion are largely restricted to the aqueous phase and should not affect the composition profile of the solid phase. The impact of dispersion on the accuracy of w/r ratios is therefore a question of how dispersion influences the concentration gradients. The effect of dispersion on the results is assessed by running two additional simulations with dispersivity values of 0 m and 100 m. Results are plotted in Figure 4.4. The spreading and dilution effect of hydrodynamic dispersion is obvious for most species. In general, where the slopes in the concentration profiles are monotonic, dispersion causes species concentrations to be lower than species that have been transported by advection only. Magnesium and sulfate for instance become depleted as the temperature of the fluid of seawater composition rises and the profiles for the range of dispersivities have roughly the same slope. Thus concentration gradients measured in this section of the column should be similar and thus more or less independent of the dispersivity of the medium. However, at and beyond concentration minima, higher concentrations occur with higher dispersivities as dispersion smoothes out the concentration profile. Calcium for instance decreases in the first 500 m of the column, then increases before it levels off beyond 1700 m. A larger dispersivity actually leads to higher calcium concentrations before and beyond the minimum. The behaviour of calcium shows the difficulty of predicting net concentration changes in the aqueous phase in regions along the flowpath which exhibit concentration increases and decreases, even more so i f dispersion is an important process. 256 Aqueous silica seems to be unaffected by dispersion in regions along the flowpath where it is in local equilibrium with quartz, for example in interval M 2 . A l so the transition from intervals M 2 to M 3 does not show any visible effects of dispersion in the silica profile. This indicates the strong tendency of quartz buffering i f silica in the solution is close to or at quartz equilibrium. In conclusion, dispersion is a process which, even though difficult to quantify, should be accounted for i f average mass transfer between the fluid and rock are calculated. The inaccuracy introduced by dispersive processes on the calculation of concentration gradients can be minimized i f the location of measurement is chosen carefully. Provided the net mass transfer of a species along the flowpath occurs either as mass gain or mass loss in either phase and dispersivities are constant, the concentration gradients based on calculations assuming only advective transport should be similar to the values based on advective-dispersive transport. This allows for an accurate estimate of mass transfer even i f dispersion processes are significant. Even in two or three dimensions, when dispersion between streamtubes occurs (transverse dispersion) the general trend of the concentration profiles should not change. Where aqueous silica is close to equilibrium with quartz, dispersion has no detectable effect on the silica concentration profile. Therefore dispersion does not have to be accounted for explicitly in equation 4.8, but one should be aware of its effects i f the direction of mass transfer between the solid and fluid phase changes along the flowpath. 257 4.2.1.1.2 Effect of Flow Velocity A set of simulations is presented to show the consequences of different fluid flow velocities on the concentration profile. These simulations are aimed at spanning a range of flow velocities likely to occur in the upper oceanic crust to test the sensitivity of mass transfer to fluid residence times. Fluid concentration profiles are calculated for fluid velocities of 5 m/year and 50 m/year and plotted at 15000 years in Figure 4.5. The results show that the nature of alteration reactions does not change. The main difference is that at lower fluid fluxes less solute mass enters the column and thus secondary minerals that incorporate seawater components, such as anhydrite, chlorite and illite, occur in lower amounts, that is, maximum precipitated volumes are smaller and the extent of these reactions do not reach as far into the column. This behaviour implies that at low fluid velocities the transition from seawater to an endmember hydrothermal fluid occurs earlier, which means that the extent of mineral assemblage M l is shorter, while that of assemblage M2 is longer. At a fluid velocity of 5 m/year the fluid is depleted of seawater magnesium at about 1100m into the column, instead of 1700 m for a fluid velocity of 10 m/year (Figure 4.5). Accordingly, the concentration profiles of cations that are involved in the exchange with magnesium (e.g. calcium) are shifted to the left and their concentrations at the end of the column are lower. At the same time, longer residence times cause a larger mass transfer from solid to the fluid phase. Species that are leached from the rock, such as potassium and sodium, tend to obtain higher concentrations at low velocities. The potassium profile for a velocity of 50 m/year illustrates that i f the fixation rate of potassium into secondary K-bearing phases is considerably lower than the flow 258 rate, significant amounts of initial seawater components can be transported far into the system. At the end of the column, potassium concentrations are still about half of the seawater concentrations at a velocity of 50 m/year. The same can be said for magnesium, which at high velocities loses only a small fraction of the seawater content over the length of the column. Sulfate, in contrast, seems to be readily removed by anhydrite at the beginning of interval T2 for the range of velocities. The silica profiles illustrate the dependence of the fluid's saturation state on the source concentration and the residence time. At a velocity of 5 m/year the silica mass transfer from the solid phase to the fluid phase is large enough such that the fluid is at quartz saturation when it enters the high temperature interval T2. Throughout the interval T2 the silica profile is essentially linear, i.e. at quartz saturation. At a velocity of 50 m/year the profile is shifted to the right and quartz saturation is obtained only at about 2150 m, well within the high temperature interval. Once silica levels are at quartz equilibrium, the profiles for all fluid velocities overlap indicating that the silica concentration is now solely controlled by the temperature. In summary, it appears that the temperature and the fluid composition are the controlling factors on the nature of the alteration reactions. A change in flow velocities affects the distribution and amounts of minerals precipitated where the system is not at equilibrium. At low velocities the direction of mass transfer between the solid and the fluid phase and thus the shape of the concentration profiles can be predicted if the nature of the dominant alteration reactions is known. At high flow velocities, however, a significant disequilibrium between the fluid and the rock composition may occur. 259 Seawater-derived potassium and magnesium can be transported far into the high temperature portions of the hydrothermal system. This is important in its general implications for the reduction of the porosity and permeability within the reaction zone of a hydrothermal system, an issue which was discussed in the previous chapter. The sulfate and calcium removal through anhydrite precipitation is not significantly affected by velocity changes. The same can be said for the silica/quartz system. In general, in order to predict the direction of mass transfer for the calculation of w/r ratios at high fluid velocities, the rates of a reaction have to be fast (e.g. the precipitation of anhydrite) or a sufficiently long flowpath has to be chosen (e.g. for the removal of magnesium). 4.2.1.2 Physical and Chemical W/R Ratios As mentioned earlier, the column is divided into three intervals according to the temperature conditions, an interval of increasing temperatures and a high temperature gradient (Tl), an interval high, slighly decreasing temperatures (T2), and an interval of decreasing temperatures and a high temperature gradient (T3). It was demonstrated that the chemical conditions in the column do not always correlate directly with these temperature conditions despite the strong effect that temperature has on individual chemical reactions. Thus the boundaries between the intervals defined by a uniform temperature gradient do not necessarily coincide with boundaries between intervals that are defined by a consistent mineral assemblage. Instead the transition between alteration assemblages is strongly controlled by the fluid composition as long as the fluid is not 260 buffered by the mineral assemblage, that is where the fluid still contains a large amount of seawater components. We identified additional intervals (M1-M3) along the flowpath which are characterized by a uniform mineral assemblage. Physical w/r ratios and chemical w/r ratios calculated by using equation 4.8 across the length (L) of each temperature interval and each mineral assemblage are plotted in Figures 4.6-4.8 and 4.9-4.10, respectively. The total simulated time span for all runs is 15000 years. To ensure that reactions between the initial porewater and the minerals are not included in the w/r ratio calculations, the beginning composition of the solid phase for determining ACj(r) is evaluated shortly after one pore volume has passed through the column at 500 years. The time axis of the figures corresponds to the simulation time, At in equation 4.8 is the simulation time minus the time for the first pore volume to pass through the column. Because flow velocities are constant throughout the column, the curves of physical w/r ratios versus time increase linearly. As density changes of the fluid are accounted for physical w/r ratios are calculated as average mass fluxes through the T and M intervals. This explains the slight variation of physical w/r ratios between each interval. Physical w/r ratios can be regarded as the 'truth' which we aim to match with chemical w/r ratios. 4.2.1.2.1 Intervals of Uniform Temperature Gradient (T1-T3) Figure 4.6 illustrates chemical w/r ratios calculated from the mass transfer of major ions in interval T l . It indicates that all chemical w/r ratios show linear trends and 261 the curves fall close to each other but below the physical w/r ratio curve. W/r ratios calculated for interval T2 are shown in Figure 4.7. A l l major species show a linear increase and provide accurate w/r ratio values, except SO4, which is strongly depleted. A s indicated by the concentration profile in Figure 4.2, magnesium is still present in significant concentrations when the fluid enters this interval. However, similar to sulfate, the concentration level of magnesium is specific to the given temperature and flow conditions and magnesium yields accurate results because we are able to integrate the mass transfer over the length of this interval. A n increase in the fluid residence time in the first interval could lead to a complete removal of magnesium from solution before it reaches the second interval and, similar to sulfate, would not provide useful w/r ratio estimates. Also , calcium concentrations reach a minimum at about 500 m which coincides with the depletion of sulfate and the cessation of anhydrite precipitation. The concentration minimum does not affect the accuracy of the w/r ratio estimates based on calcium as its net mass transfer across the interval is large enough. However, because the location of the minimum is not directly related to the temperature distribution, calcium has to be used with caution for similar reasons as for magnesium. Figure 4.8 shows the estimates w/r ratios for interval T3. A l l chemical w/r ratios increase linearly but deviations from physical w/r ratios are more significant than in intervals T l and T2. The scatter of the curves in this interval is partly due to the depletion of ions such as magnesium and sulfate and in part a result of near-uniform concentration profiles through this interval as is the case for sodium. The best match is obtained by calcium and silica. Calcium shows a weak net increase throughout interval T3. Sil ica 262 decreases steeply as the temperature drops to seawater values. The large silica mass transfer ensures reliable w/r ratio estimates. 4.2.1.2.2 Intervals of Uniform Alteration Assemblage (M1,M2) W/r ratios for the first mineral assemblage are illustrated in Figure 4.9. Similar to the results from the interval T l , the mass transfer of all species produces chemical w/r ratios that are lower than the actual physical w/r ratios. However, the chemical w/r ratios fall closer to the physical w/r ratio curve. Potassium, for instance, which showed the greatest underestimate in T l matches the physical w/r ratios very well in M l . The generally better fit of the chemical w/r ratios may be explained by the larger mass transfer that takes place over the longer interval. Sodium and silica show a weak non-linear slope as a result of a small transient shift in their concentration profiles in flow direction between 500 and 15000 years (Figure 4.2). A l l chemical w/r ratios for interval M2, which contains the high temperature, endmember hydrothermal fluid provide reasonable values with the exception of sulfate (Figure 4.10). Sulfate deviates considerably from the physical w/r ratio curve as it is virtually removed in this interval. The cations calcium, magnesium, iron and sodium show a weak non-linearity which is likely to be a transient effect. Quartz buffering of aqueous silica in M2 ensures a silica mass transfer that is temperature dependent. Despite the low temperature gradient and the resulting small silica mass transfer in this interval, the w/r ratios fall relatively close to the physical values. This suggests that in general 263 when a solution is at quartz saturation at two different locations along the flowpath and a temperature gradient is present between these locations, the mass transfer of silica is likely to provide accurate values of the w/r ratio. Interval M 3 is identical to T 3 , for which w/r ratio calculations have been performed in the previous paragraph. In conclusion, where temperature is the primary factor controlling the nature of reactions, such as during rapid heating and cooling of seawater and hydrothermal fluids, respectivly, the correlation of temperature and chemical mass transfer leads to reasonable w/r ratio estimates for intervals of the flowpath that are either defined by consistent temperature conditions or alteration assemblages. It is likely, however, that over intervals of the flowpath where temperature changes are less dramatic, it is more likely that the change in the fluid composition triggers the beginning of a new mineral assemblage and resulting changes in the rate or direction of species mass transfer. For this reason, identifying and estimating w/r ratios over intervals of consistent mineralogy can increase the accuracy of these estimates. The nature of w/r ratios and testing various ways of deriving w/r ratio estimates based on the chemistry in an open, flow-through system was the focus of the one-dimensional simulations. It was demonstrated earlier that it is possible to relate w/r ratios with fluid velocities and thus to establish a link between the chemistry of a system and its hydrologic conditions (equation 4.7). In the next section, calculations are performed for the more complex, two-dimensional reactive transport simulations that were the focus of the previous chapter. 264 4.2.1.3 Two-Dimensional Simulations In the previous two chapters the evolution of the fluid flow and thermal fields, and the alteration mineralogy were presented for a system with a basement permeability of 1x10\"14 m 2 (that is the permeability value which was obtained by results from packer and flowmeter experiments in the basement at Middle Valley (Becker et al. 1994)). In this chapter we build on these results and present mass transfer calculations and estimates of fluid velocities derived for different regions in the modeling domain. The flow and temperature fields and the subsequent coupling with the reactive transport simulation provide us with a representation of the geometry of the flow field and the geochemistry at any point of the domain. In a field setting, however, information on the flow geometry is limited and has to be inferred. For instance, fluid flow is likely to be predominantly vertical in areas such as high permeability recharge and discharge zones, which can be identified by means of heatflow measurements. Between recharge and discharge zones of the same hydrothermal system lateral flow may occur. For this reason we initially limit our calculations to flowpaths or streamtubes within the recharge and discharge zones on both sides of the domain as well as within a region of dominantly horizontal flow near the sediment/basement interface. The method is also applied to a vertical section through the sediments at the center of the domain (upper panel, Figure 4.11). Here we assume the flow direction to be vertically downward, even though a small lateral component toward the downflow limb of the convection cell exists (Figure 2.9). 265 Not only do these calculations serve to test the method under different chemical and hydrologic conditions (rock composition and flow velocity, respectively), but the sediments are generally easier to sample and, most importantly, constraints on At, that is the life-span of the flow system, are easier to derive. The hydrologic and chemical conditions and parameters used in the two-dimensional simulation were described earlier and are summarized in Figures 2.1 and 3.2, respectively. Mass transfer between the solid and fluid phase is integrated over a time interval of 5000 years, between 10000 and 15000 years. For the sediment section at the center of the domain, mass transfer is calculated between 12500 and 15000 years to ensure that at least one pore volume has passed through the sediments. A large time interval ensures that sufficient mass transfer along the flowpath occurs. In general, however, any time interval could be used provided that the change in the rock composition over this interval can be determined. In an oceanic system where the accuracy of measurement is low At should be large. If the onset of hydrothermal circulation can be constrained, At could be assumed to represent the age of the hydrothermal system and thus be sufficiently large. This assumption would then require that fluid flow is stable (i.e., it does not change its geometry significantly) throughout regions of the hydrothermal system from the onset of hydrothermal convection until the time of measurement. F low and temperature fields at 10000 and 15000 years and the stretches of flowpath over which the velocity estimates are performed are illustrated in Figure 4.11. A t 10000 years the flow field is essentially stable, that is, it does not change its geometry 266 significantly. However, the system has not reached a steady state, which implies that temperatures and the vigor of convection decrease with time. Transient effects during this interval, in particular with respect to changes in temperature, are not accounted for and constitute a possible source of inaccuracy in the calculation of w/r ratios. The pattern of flowlines superimposed on the temperature and velocity fields at 15000 years (lower panel, Figure 4.11) shows that the stable flow field consists of a single large convection cell that extends across the entire domain. The figure shows what was discussed in the previous chapter, that is the convection cell is essentially an isolated hydrological system. Fluids recharging the basement from the recharge zone travel through a zone along the boundaries of the domain and exit the system after passing through the basement only once. Consequently, seawater entering the basement obtains the chemical fingerprint of a hydrothermal fluid during a single pass through the convective system. In contrast to the fluids moving along the boundaries, the fluids trapped within the isolated convection cell undergo low-gradient temperature changes as the fluid circulates around the axis of the cell. In the upper half of the cell the fluid undergoes cooling, in the lower half heating (Figure 4.11). Equation 4.7 is used to relate chemical w/r ratios to estimates of the flow velocity. The estimated velocities are then compared with the velocities (the magnitude of the velocity vector) obtained from the flow simulation. A l l estimates of the flow velocity based on chemical mass transfer assume one-dimesional flow, that is in a vertical direction for the up- and downflow zones and in a horizontal direction below the interface. 267 The mass transfer calculations are performed as a function of the length of the flowpath. This method involves a step-by-step increase in the length of the flowpath over which the calculations are performed. The length of each step is 10 m. The flowpath is extended in the direction of flow and the results are plotted as the inferred average velocity versus the length of the flowpath. The velocities derived from the flow code are averaged and plotted in the same manner. This averaging method allows us to test the consistency of spatial trends in the mass transfer, provide constraints on sampling densities and locations, and determine over what distance the assumption of one-dimensional flow in a two-dimensional domain is reasonable. 4.2.1.3.1 Recharge Zone The recharge zone designates the region of vertical downward flow and is located at the left-hand boundary of the domain. The upper half of the recharge zone constitutes a 5 m wide, higher-permeability conduit through the sediment section that establishes the hydrological connection between the ocean and the hydrothermal basement. The high permeability contrast between the recharge conduit and the sediments leads to focussed flow of seawater into the basement through this part of the recharge zone. F low velocities within the conduit are about one order of magnitude higher than flow velocities in the basement below. Therefore, on reaching the basement the fluid experiences a rapid decrease in velocity and at the same time enters a much higher temperature regime. Because of the drop in flow velocities below the sediment/basement interface two 268 separate average velocity calculations are performed, one for the recharge conduit through the sediments layer and the other for a vertical 'column' in the basalt directly beneath the recharge conduit. This division also has the advantage that separate calculations are performed for the geochemically different sediments and basalts. Profiles of the mineral volume percentage that is precipitated or dissolved and total species concentrations for the total length of the recharge zone (i.e. from the seafloor to the bottom of the domain) after 15000 years were plotted in Figures 3.4 and 3.5, respectively. Only the important characteristics are summarized here. High flow velocities, low temperature gradients and overall temperatures below 50°C within the recharge conduit cause relatively small amounts of mass exchange between 0 - 250 m as indicated by very small concentration changes and a limited alteration of the primary mineralogy. Below 250 m chlorite starts to precipitate. Just above the sediment/basement interface anhydrite begins to form. These reactions initiate the first noticable changes in the slope of the concentration profiles. On entering the basalt the fluid temperature increases rapidly so that the amount of precipitated chlorite and anhydrite increase towards a maximum just below the interface. Other alteration phases that occur in the basalt are illite and small amounts of muscovite and quartz. Most of the alteration and the bulk of the mass exchange occurs within the basalt section of the recharge zone while the primary mineralogy and the concentration profiles are essentially unchanged throughout most of the sediment section. Within the basalt the concentration of magnesium drops by about 50% from the sediment/basement interface to the bottom boundary. Potassium and sodium decrease slightly with depth. Calcium and sulfate concentrations increase from the sediment/basalt interface downward to about halfway into the basalt section before 269 decreasing toward the bottom boundary. The increase, which appears to be in disagreement with the large quantities of precipitated anhydrite in the solid phase, is due to the redissolution of previously formed anhydrite. This behaviour indicates that the system at 15000 years has not yet reached a steady state. Silica increases throughout the recharge zone, including the recharge conduit, but its concentration remains low in comparison to that of other species in solution. Velocity estimates and the true average velocity for both the conduit through the sediments and for the basalt section of the recharge zone are shown in Figure 4.12. Velocities in the sediment section are of the order of 37 m/year and are nearly constant throughout its length except for the lower 20 m where velocities begin to decrease as the fluid approaches the basement. In the basement beneath the high permeability recharge conduit, fluid velocities continue to decrease between 500 m and 600 m and exhibit only minor changes in an interval extending from 600 to 750 m. Below this interval, velocities decrease again by more than one order of magnitude near the bottom boundary of the domain. Note that due to the averaging of the flow velocities for the basalt section of the recharge zone details of the flow field are lost as the flowpath is extended. The average velocity curve does account for the decreasing velocities near the sediment/basement interface, but the decrease at the bottom boundary is less apparent. The results indicate that only the mass transfer of silica captures the magnitude of the high velocities in the recharge conduit reasonably well. The very limited mass transfer of sulfate and cations within this zone causes velocity estimates to deviate quite significantly from the actual velocities. 270 Higher temperatures, larger temperature gradients, and lower velocities within the basalt section of the recharge zone result in a much larger mass transfer. Consequently, most species lead to a good match with actual fluid velocities over large segments of the total flowpath. A n exception are velocity estimates derived from the sodium profile which underestimate the true velocity values and do not capture the trends of the velocity variation accurately. Due to the concentration maximum of calcium and sulfate in the basalt, erroneous velocity estimates are calculated in the lower half of the basalt using these species. The magnesium and potassium concentration profiles show similar trends and consequently provide nearly identical velocity estimates. In summary, the trends that were observed in interval T l of the column simulation are recognizable in the downflow zone of the two-dimensional flow and temperature field, in particular in the basalt. Silica is the most robust species, which provides good estimates for the entire temperature and velocity range within the sediment and the basalt section of the recharge zone. Magnesium and potassium are reliable species in the basalt as they capture the velocity trends and magnitude very well. Sulfate and calcium provide good results only in regions of the basalt where their concentration profiles show a monotonic increase or decrease. The slightly unsteady nature of the temperature field leads to the redissolution of anhydrite below a depth of 700 m, causing a concentration maximum for both species and thus erroneous velocity estimates in the lower half of the basalt 271 4.2.1.3.2 Discharge Zone The discharge zone is a region of dominantly vertical upward flow positioned at the right hand boundary of the domain. The upper half of the upflow zone consists of the discharge conduit which penetrates the sediment section and is a vertical, 5 m wide zone of higher permeability forming the hydrologic connection between the basement and the ocean. Within the discharge conduit fluid velocities are as high as 70 m/year. The velocities are nearly constant throughout the conduit and decrease only close to the seafloor. The rapid upflow of hot hydrothermal fluids towards the relatively cool ocean floor leads to very steep temperature gradients in the upper third of the conduit. It should be noted that due to the imposed constant temperature along the seafloor, the calculated temperature gradient near the seafloor is oversteepened. This artifact results in an overestimate of the mass transfer between the solid and the fluid phase but this should not have a significant effect on the nature of alteration reactions. In the basement section of the discharge zone, fluid velocities are about one order of magnitude lower. The lowest velocities occur near the bottom boundary of the domain. The profiles of the mineral volume percentage precipitated or dissolved and total species concentrations after 15000 years of simulation time were shown in Figures 3.9 and 3.10, respectively. The profiles indicate that the main alteration phases in the basalt are quartz and clinozoisite, in addition to small amounts of chlorite and albite. This assemblage is similar to the mineral assemblage M2 that buffers the fluid of endmember composition in the one-dimensional simulation. This assemblage is consistent with the fluid composition which shows the characteristics of the endmember hydrothermal fluid 272 described for the interval M2, that is a fluid that has lost its seawater-derived components. Calcium shows a weak concentration maximum in the lower third of the basalt, silica decreases and sodium and potassium slightly increase toward the sediment/basement interface. The small amounts of iron and magnesium within the discharge zone are controlled by the balance of clinopyroxene dissolution and their fixation into chlorites (daphnite and clinochlore, respectively). In addition, magnesium is taken up by illite at lower temperatures (above 200 m). In the discharge conduit through the sediment section, the steep temperature gradient causes a far more dramatic mass transfer and more significant changes in the fluid and solid phase composition than in the basement, especially near the seafloor. The most important reaction in this zone is the precipitation of quartz. The very large amounts of quartz that have precipitated near the seafloor after 15000 years of simulation time are likely to be overestimated due to the constant temperature assigned to the seafloor. Nevertheless, compared to other secondary minerals, for example, albite and clinozoisite, which precipitate throughout most of the discharge zone, quartz is by far the most abundant. The massive precipitation of quartz is accompanied by a drop in aqueous silica concentrations. In contrast, the concentrations of sodium, calcium and potassium increase as the fluid reaches the seafloor. The increase in potassium in the upper 200 m of the discharge zone causes the precipitation of illite and muscovite. The flow velocity estimates are plotted for both the discharge conduit through the sediments and the basalt section of the discharge zone in Figure 4.13. The plot shows that silica captures the magnitude and the constancy of the velocity in the discharge conduit almost perfectly. Estimates start to deviate from the velocity curve only in the upper 50 m 273 of the conduit. Magnesium and iron also provide reliable velocity estimates despite being present in only minute amounts. Potassium underestimates the true velocities by about 50%. A l l other major species provide erroneous velocity estimates for the discharge conduit. Deviations from the velocity curve are more severe in the basalt section of the discharge zone, for all major species in solution. Velocity estimates based on silica may still be considered meaningful as the estimates are reasonably close to the true velocities in the lower 300 m of the basement section. However, these estimates become unreliable near the sediment/basement interface. Of limited use are also iron and sodium, which both underestimate velocities by more than 50% but show the correct trend in the velocity distribution. The likely reasons for the large deviations of the velocity estimates from the true fluid velocities are the small extent of mass transfer in the central and upper section of the basalt, in addition to steep, non-linear concentration gradients near the bottom boundary. In summary, the fluid, which ascends from the bottom boundary to the ocean floor resembles that of the endmember fluid in the one-dimensional simulations, that is, it is depleted in sulfate and magnesium. Despite low concentrations, magnesium and iron provide useful velocity estimates for the discharge conduit through the sediments. Silica provides reasonably accurate velocity estimates over most of the sediment and the basalt section of the discharge zone. However, significant deviations from the true velocity profile occur near the sediment/basement interface and at the seafloor. Unlike the results 274 from the one-dimensional simulation, all other cations provide estimates that deviate substantially from the true velocity curve. 4.2.1.3.3 Lateral Flow Predominantly lateral flow may be present between the limbs of a convection cell in regions of the oceanic basement where cellular convection occurs, or i f the permeability in the upper basement is anisotropic such that horizontal flow is favored. The latter case has been suggested by Fisher et al. (1998) who argue that the basement anisotropy is largely due to the fact that the basement consists of a sequence of interbedded sediments and sills. Where the convective layer is overlain by a lower permeability cap of sediments, fluid flow will be forced in a direction more or less parallel to the interface of the two layers. We use this assumption and perform calculations of velocity estimates for a horizontal 'column' which is located 100 m below the sediment basement interface, 10 m wide, and which extends from 250 m to 1750 m in the horizontal direction of the modeling domain. We eliminate 250 m on each side of the domain to exclude the recharge and discharge zones and most of the limbs of the convection cell. Flow through this column is from right to left and fluids experience a slight decrease in velocity and cooling of about 70°C along their flowpath (Figure 4.11). The column is located well within the convection cell which is, as stated above, an isolated chemical and hydrological system not effected by recharging 'fresh' seawater. The 275 fluids are therefore endmember hydrothermal solutions depleted of magnesium and sulfate. The profiles of total species concentrations and mineral volume percentage precipitated and dissolved show that the dominant reactions are epidotization and albitization, that is the conversion of plagioclase to clinozoisite and albite according to equations 3.53-3.55, respectively (Figure 4.14). Less important reactions include the precipitation of quartz, chlorite, and the potassium-bearing phases muscovite and illite. The depletion of seawater components in the fluids leads to an exchange of components between primary and secondary minerals, resulting in a 'replacement' of primary mineral phases by secondary phases and a near-symmetrical mineralogical profile in Figure 4.14. The amounts of dissolved primary minerals decrease from right to left with decreasing temperature. A similar temperature correlation is shown by the precipitated volumes of the secondary minerals quartz, chlorite and clinozoisite. The amounts of these minerals decrease from right to left. In contrast, the process of albitization correlates with temperature only in the left half of the profile. Muscovite is favored over illite at higher temperatures as it appears only in the right half of the profile, illite is only present in the left half of the profile. In the concentration profiles, a temperature correlation as it occurs in the mineralogy is not quite as obvious except in the aqueous silica profile. Aqueous silica in solution decreases from right to left. The similar slopes of the aqueous silica curve and the quartz curve is consistent with quartz buffering of aqueous silica, a process that was described for the high temperature interval (T2 and M2) of the one-dimensional 276 simulation. The temperature correlation of aqueous silica and the resulting monotonic slope of the silica concetration profile provide accurate velocity estimates (Figure 4.15). In addition to silica, only calcium, sodium, and potassium are present in significant concentrations but their profiles do not show the same monotonic trend as silica. Calcium and potassium exhibit a weak concentration maxium at about 900m, sodium concentrations peak at 400 m (Figure 4.14). Consequently, velocity estimates derived from the mass exchange of these cations are highly erratic and essentially of little use (Figure 4.15). Magnesium, which is present in minute amounts captures the magnitude of the velocity in the left half of the profile, that is at higher temperatures, but deviates from the velocity curve in the right half where temperatures are lower. Iron is also present in low concentrations but it captures the slope of the velocity curve well but underestimates the average velocity througout the profile. Sulfate is virtually removed and is not suited for velocity estimates in this region of the reaction zone. In summary, the processes that cause mass exchange in the region of lateral flow are similar to those described for the one-dimensional column experiments, in particular for the interval M2 which exhibits mineral buffering of the endmember hydrothermal fluid. However, the near-linear concentration profiles of the cations that were observed in M2 and provided accurate chemical w/r ratios are not present in the region of lateral flow within the two-dimensional domain. The concentration profiles of the cations are rather non-linear or erratic and velocity estimates derived from the profiles are inaccurate. Nevertheless, the process of quartz buffering of aqueous silica that was described for interval M2 occurs in the region of lateral flow as well. This process, which controls silica mass transfer and causes a temperature dependence of the amount of silica in 277 solution, occurs when silica levels are near or at local quartz equilibrium. Under these conditions, the temperature correlation provides a prerequisite for reliable velocity estimates. This is confirmed by the very good match of the velocity estimates based on silica mass transfer with the true average velocities in Figure 4.15. In addition, the good match suggests that the assumption of dominantly horizontal flow near the sediment/basement interface is reasonable. We can take the method one step further and use silica mass transfer to estimate fluid velocities for a larger portion of the basement. The flowpaths over which the mass transfer is calculated and average velocities are estimated have a length of 20 m. The calculations are performed for the entire basement section, excluding the limbs of the convection cell and the upflow and downflow zones, such that strictly horizontal flow can be assumed. The results of these calucations are shown and compared with the true average velocities (the magnitude of the velocity vector) in Figure 4.16. Erroneous estimates occur near the bottom boundary on the left-hand side of the domain, where silica has not reached quartz saturation levels. A deviation of the velocity pattern is present near the sediment/basement interface between 250 m and 1000 m where recharge from the sediments into the basement leads to a local disturbance of the otherwise quartz saturation and thus temperature-controlled silica levels in solution. In general, however, the estimated velocity values as well as their variation through the domain match the true velocity field very well. 278 4.2.1.3.4 Vertical Flow through Sediments One of the most difficult parameters to constrain in equation 4.7 is At. The difficulty arises because At has to be large enough to allow sufficient mass transfer between the solid and the fluid phase and at the same time the flow conditions should not change significantly throughout At. Regarding the often transient nature of flow in the basement, the most suitable regions are those that exhibit fast rates of mass transfer, such as in the recharge zone or the discharge zone. In contrast, flow through the sediments, even though extremely slow, is less transient. Figure 2.9 demonstrates that fluid flow through the sediments tends to be downward with a lateral component towards the downflow limb of the convection cell in the basement. The lateral component is stronger toward the discharge zone. At the center of the domain, where the mass transfer calculations are carried out, the lateral component is more important in the lower half of the sediments but it is overall weaker than the vertical component. Thus the assumption of purely vertical downward flow seems reasonable. Fluid flow through the sediments probably begins as soon as they have been deposited. Stratigraphical or isotope dating methods enable estimates of the time of deposition and thus constrain the time of fluid percolation through the sediments, that is At. In addition to the advantage of providing constraints on the duration of fluid flow, sediments are a porous medium in its proper sense (that is, i f not indurated by alteration as may be the case in the lower sections of the sediment layer), as opposed to the fracture-permeability of a basalt which on a macroscopic scale is only an approximation of a porous medium. Fluid flow patterns through the sediments are therefore easier to predict 279 and to model. Finally, it may be easier from a technical perspective to obtain porefluid and rock samples from the sediments at or near the seafloor than from the much deeper basalts. The mineralogical and fluid composition profiles through the sediments (Figure 3.6 and 3.7) indicate that the nature of the chemical reactions is different from those at the recharge zone. Whereas chemical reactions at the recharge zone are controlled by the heating of seawater and the associated exchange of cations between the solid and the fluid phase, the lower flow rates through the sediments cause reactions that are rather controlled by the mass transfer between primary and secondary mineral phases. Similar to the region of horizontal flow below the sediment/basement interface, this leads to far lesser amounts of mineral phases that incorporate seawater components, such as anhydrite and clinochlore, and to a roughly symmetrical mineralogical profile (Figure 3.6). The dominant reactions are the alteration of primary feldspars and the formation of illite, albite, clinozoisite, and small amounts of quartz and muscovite. In the porefluids (Figure 3.7), magnesium, potassium, sodium and sulfate all decrease from the seafloor to the sediment basement interface due to their fixation into alteration phases. Potassium recovers to almost its seawater concentration from about 250 m to the interface. Calcium concentrations increase between the seafloor and 250 m and tend to level off with minor fluctuations beyond 250m. Silica concentrations show a slight net increase from the seafloor to the interface. Mass transfer calculations were performed for the period of time between 12500 and 15000 years, after the first pore volume of seawater has passed through the 280 sediments. Figure 4.17 shows that silica, calcium, magnesium, and in the lower half of the sediments sulfate provide reasonable velocity estimates. The good agreement of both the true velocity and the velocity estimates indicates that the assumption of purely vertical flow through the sediments is appropriate (under the conditions specified) and the mass transfer of species can be used to estimate flow velocities even at very slow flow rates. 4.3 Conclusion The goal of this chapter was to develop a formulation which helps to constrain estimates of the vigor of convection in the oceanic crust by using geochemical information on fluid and rock compositions. The formulation is based on w/r ratios for a dynamic flow system. It relates chemical mass transfer to the flow velocity, the densities of the fluid and the rock, the rock's porosity, the age of the hydrothermal system, and the length of the flowpath. The mass transfer between the solid and fluid phase is integrated over a certain length of the fluid's flowpath, thus estimates of flow velocities obtained from this mass transfer are spatial and temporal averages over the same length. The integration of the mass transfer over the flowpath ensures that under non-isothermal conditions sufficient mass transfer between the solid and the fluid phase occurs and the possibility of thermodynamic equilibrium (which would result in invalid results) is avoided. The nature of the alteration reactions determines whether a species can provide useful estimates of flow velocities. Favourable for accurate estimates is i f the 281 concentration profile of a species along the fluids flowpath either monotonically increases or decreases. The assessment indicates that a difference in the average fluid velocity changes the extent to which certain alteration reaction occur and the magnitude of mass transfer. At low flow velocities, the fluid obtains its endmember composition earlier as less seawater enters the system and less species exchange takes place. Species which are leached from the solid phase and not incorporated into secondary phases tend to acquire higher concentration levels at lower velocities. With decreasing velocities the concentration profiles approach those of local equilibrium conditions. In contrast, at high fluid flow velocities the concentration profiles become more unpredictable. Depending on the saturation state of the solution and the magnitude of mass transfer over a certain length of the flowpath, a significant disequilibrium between the solid and the fluid phase may be present. The silica concentration profile is conservative with respect to velocity changes along sections of the flowpath where quartz saturation is present. Dispersion influences solute concentrations through spreading and dilution. Simple dilution of aqueous solutes prevails as long as concentration profiles have a consistent slope, for example, the steady removal of magnesium. Then the impact of dispersion on the magnitude of the concentration gradient is relatively insignificant. The impact of dispersion becomes more important where the concentration profile exhibits an extrema. At and beyond a minimum, for example, dilution does not occur and instead higher concentrations are obtained for larger dispersivities. The silica profile is conservative with respect to dispersivity changes where quartz saturation is present which 282 implies that at quartz saturation the impact of dispersion at silica concentration extrema would not be a concern. In order to apply this method to multi-dimensional flow regimes, an approximate understanding of the flow geometry is necessary because the mass transfer has to be integrated over certain length of the flowpath. The flowpath of a fluid parcel moving through an oceanic hydrothermal system can be divided in a recharge zone, reaction zone, and discharge zone. In each of these regions inferences on the flow geometry can be made if stable cellular convection occurs: downward flow in the recharge zone, upward flow in the discharge zone, and predominantly horizontal flow in the reaction zone. Each of these regions has distinct flow and temperature conditions and may also contain a distinct rock composition. As a consequence, within each region characteristic chemical alteration reactions occur that may provide the basis for mass transfer calculations and the application of the proposed method. In the recharge zone the fixation of alkalis and magnesium, the precipitation of anhydrite and the removal of calcium and sulfate from solution, and the increase in dissolved silica provide a wide range of species that are involved in an exchange between the solid and the fluid phase and are useful for estimating flow velocities. Silica yields accurate velocity estimates over the widest range of temperatures and flow velocites. In the reaction zone the fluid is depleted of seawater components such as magnesium and sulfate. In addition, the temperature gradients within the reaction zone are low such that the rates of overall mass transfer are slow. Thus the nature of alteration reactions and the slope of concentration profiles along the flowpath may be difficult to 283 predict. If the system obtains quartz saturation, however, the aqeous silica concentrations are correlated with the temperature and predictions about the direction and the intensity of silica mass transfer along the flowpath can be made if the temperature distribution is known. Then, in regions of the reaction zone where quartz occurs as an alteration product, w/r ratios can be reliably estimated based on the quartz/aqueous silica mass transfer. In the discharge zone the rapid cooling of ascending hydrothermal fluids leads to high temperature gradients, in particular near the seafloor. The substantial precipitation of quartz and the drop in aqueous silica can be used to infer fluid flow velocities for the discharge zone. A parameter which is difficult to constrain in this method is At, that is the period of time over which the flow system is stable and measurable mass transfer occurs. Unless clearly defined pathways are present (e.g. high permeability fracture zones) in which fluid flow is steady, At may be difficult to determine. These concerns do not arise for the sediments, as fluid flow is likely to have initiated at or shortly after sediment deposition and the time of deposition is relatively straight forward to constrain. Our calculations show that the chemical mass transfer provides the basis for reasonable flow velocity estimates through the sediments despite the low flow rates. It was demonstrated that the quartz/silica system provides reliable estimates for flow velocities under a wide range of temperature and flow conditions. The quartz/silica system may provide a useful option for estimating w/r ratios and flow velocities as an alternative or in addition to the use of isotopes or conservative species involved in mass exchange. What makes the quartz/silica system useful is that silica is a major component 284 in hydrothermal fluids and the correlation between the aqueous silica concentration and temperature at quartz saturation makes its behaviour predictable once quartz saturation is obtained. 285 \"C Fluid of Seawater Composition^ T l T2 T3 1500 Distance (m) 3000 280 260 240 220 200 180 160 140 120 100 81) 60 40 20 0 Primary Minerals: An55, An65, An75, Clinopyroxene Secondary Minerals: Quartz, Anhydrite, Calcite, Chlorite (Clinochlore, Daphnite), Albite, Illite, Clinozoisite Figure 4-1. Column design, temperature distribution, solid phase composition, and important hydrologic parameters. 286 0.48 Distance (m) 500 years Ca+2 Fe+2 K+ Mg+2 Na+ Si02,aq S04-2 15000 years Ca+2 Fe+2 K+ Mg+2 Na+ Si02,aq S04-2 Figure 4-2. Fluid composition profiles through the column at 500 and 15000 years. 287 M l M2 M3 - i — i — i — i — r -500 1 1 1 1 1 1 1 1 c 1 1 r 1000 1500 2000 - i 1 c r -2500 - Quartz - Anhydrite - Calcite • Clinochlore | Daphnite - Albite An55 An65 • An75 Illite • • • • Augite Clinozoisite! 3000 Distance (m) Figure 4-3. Mineralogical profile through the column at 15000 years. The profile includes the mineral assemblages M l , M2, and M3. 288 0.48 A Distance (m) 0 m Ca+2 Fe+2 K+ Mg+2 Na+ Si02,aq SC4-2 100 m - • - Ca+2 • • Fe+2 K+ - - Mg+2 Na+ Si02,aq •— S04-2 Figure 4-4. The effects of dispersion on the concentration profile. Dispersivities 0 m and 100 m. 289 A / 0 0 8 0.04 I 0.02 0.00 \\ \\ X A 1 ' \\ 500 1000 1500 Distance (m) 2000 2500 3000 5 m/yr Ca+2 Fe+2 K+ Mg+2 Na+ Si02,aq S04-2 50 m/yr Ca+2 Fe+2 K+ Mg+2 Na+ Si02,aq S04-2 Figure 4-5. The effect of the flow velocity on the concentration profile. Flow velocities 5 m/year and 50 m/year. 290 ]\\ Ca+2 2000 4000 6000 8000 10000 12000 14000 Time (years) Figure 4-6. Water/rock ratios for interval T l . 291 Figure 4-7. Water/rock ratios for interval T2. 292 2000 4000 6000 8000 10000 12000 14000 Time (years) Figure 4-8. Water/rock ratios for interval T3. 293 5 2000 4000 6000 8000 10000 12000 14000 Time (years) Figure 4-9. Water/rock ratios for mineral assemblage M l . 294 5 2000 4000 6000 8000 10000 12000 14000 Time (years) Figure 4-10. Water/rock ratio calculations for mineral assemblage M2. 295 recharge zone discharge zone 0 1000 2000 X(m) 15000 years 0 1000 2000 X(m) Figure 4-11. Flow and temperature fields at 10000 years, including the 'columns' for which average flow velocities are calculated (top) and temperature field, flow field, and flowlines at 15000 years (bottom). 296 0 Seafloor 200 300 Length of Flowpath (m) 500 Sediment/Basement Interface Flow Direction Sediment/Basement Interface 200 300 Length of Flowpath (m) Flow Direction 500 Bottom Boundary Figure 4-12. Average fluid flow velocities for recharge zone for sediment (top) and basalt (bottom) sections. 297 80 60 >^ £, >> > 00 u > 40 H 20 / / • \\^ V 0 500 Seafloor i 1 1 r-400 300 200 Length of Flowpath (m) 100 Sedimenl/Basement Interface Flow Direction c3 O 15 > < 500 Sediment/Basement Interface 400 300 200 Length of Flowpath (m) Bottom Boundary Flow Direction Figure 4-13. Average fluid flow velocities for discharge zone for sediments (top) and basalt (bottom) sections. 298 Quartz Anhydrite Calcite Clinochlore Daphnite • Albite - An25 - - An55 - An65 •••• An75 ••- K-feldspar Muscovite - Illite • • Augite - - Clinozoisite Distance (m) Figure 4-14. Fluid composition and mineralogical profile through horizontal 'column' below the sediment/basement interface. 299 Length o f Flowpath (m) Flow Direction Figure 4-15. Average fluid flow velocities for horizontal 'column' below the sediment/basement interface. 300 True Velocities 400 600 800 1000 1200 1400 1600 X(m) Si02 X(m) Figure 4-16. Velocity estimates for the reaction zone, based on short flowpaths (20 m) and the assumption of strictly horizontal flow. Velocity in m/year. 301 0.06 •n 0.04 0.02 0.00 Si02: S04-2 Average Velocity —i— 100 i 1 r-200 300 Length of Flowpath (m) Seafloor 400 500 Sediment/Basement Interface Flow Direction Figure 4-17. Velocity estimates for the sediment section at the center of the domain. 302 5 Conclusion of Thesis The objective of this study was to gain insights into the vigor and geometry of fluid flow in a sedimented oceanic ridge setting. Both the vigor and geometry of flow are controlled by the permeability and the permeability distribution in the hydrothermal basement. Furthermore, fluid flow is directly related to the thermal and geochemical evolution of the hydrothermal system. The coupling of flow, temperature, and geochemistry and the effect of the permeability on these coupled processes was the emphasis of this study. These coupled processes were investigated for an open system from the onset of fluid motion after an initial state of a conductive temperature distribution and fluid stagnation, to a near steady-state convective system for permeabilities ranging from 1x10\" m to 1x10\" m . The system included higher permeability recharge and discharge zones through the sediments, which led to focussed recharge of cool seawater into the convective system and discharge of hot basement fluids out of the convective system, respectively. The results show that the basement permeability strongly affects the evolution of the hydrothermal system. The higher the basement permeability, the higher the initial flow rates, the higher the degree of fluid mixing and thermal homogenization, and the faster the rate of cooling and the stabilization of the flow field. For the range permeabilities, the flow field evolves from an inital state of unstable, irregular flow to a stable convection geometry of a single large convection cell. The inital state is characterized by extremely vigorous and erratic flow at high permeabilities and a more 303 organized flow in small transient convection cells at lower permeabilities. The aspect ratio of these transient convection cells is near unity at lowest permeabilities and decreases if the permeability is increased. The stabilization of the flow field coincides with the migration of the cooling front induced by the steady inflow of cold seawater through the basement. The stable convection cell leads to the hydrological isolation of a portion of the basement and restricts the recharge of seawater to the confined region outside the isolated portion. Recharge into the basement occurs either through the recharge zone or through the sediments. Fluid flow through the sediments tends to become focussed towards the sediment/basement interface and only a small zone in the basement rocks outside the isolated convection cell transmits the recharge through the sediment cover. At steady state recharging fluids exit the convective system after a single pass through the basement. The size of the stable convection cell and thus the size of the hydrologically isolated basement region is larger for higher basement permeabilities. Accordingly, the region affected by recharging fluids, that is, fluids that enter the basement either through the recharge zone or the sediments, is smaller. Time-integrated fluid fluxes or water/rock ratios can be correlated with the basement permeability. For a steady state flow field, this correlation is log-log-linear. When the transient state is accounted for, the slope of average w/r ratios versus permeability approaches the same log-log-linear correlation. The longer the system has been in operation, the better the correlation. The reactive transport simulation suggests that the nature of alteration reactions is not likely to change over the range of realistic permeability values as the thermal 304 conditions in the basement would not change sufficiently to change the metamorphic grade. The permeability does have an effect on the intensity of alteration and the distribution of alteration products, however. As the hydrothermal system cools, the reactivity in the system decreases. Thus the transient state of the hydrothermal system, which is characterized by high temperatures and high flow velocities, leads to alteration patterns that can be recognized after significant periods of time and may be in a disequilibrium with the current chemical conditions. Therefore, alteration patterns in the rock have to be interpreted as the time integrated reactive history of the system. The feedback of alteration reactions on the porosity/permeability of the medium and back on the flow conditions is most significant in regions where the fluid is far from equilibrium with the rock. Such conditions exist where temperature gradients are steep, such as in the recharge zone and the discharge zone. A large reduction in porosity occurs in the higher-temperature part of the recharge zone and in the upper section of the discharge zone. The reduction may be significant enough to cause a complete sealing of the recharge or the discharge zone. The reaction zone, in contrast, exhibits only minor porosity changes, as a result of low temperature gradients and a fluid that is buffered by the rock composition. A contributing factor to the rock buffering of the fluid composition is the hydrological isolation of a large part of the reaction zone and the lack of seawater components. In fact, within this isolated region the porosity shows a net increase, which is contrary to the common believe that the porosity/permeability of the hydrothermal system decreases as it cools. Thus it appears that in order to obtain a porosity decrease of significant magnitude, fluids with a strong seawater signature have to enter and be distributed throughout the reaction zone. 305 As the nature of alteration reactions, the degree of alteration, and alteration patterns do not provide a direct constraint of the vigor of fluid flow and the basement permeability, the correlation between the average basement water/rock ratios and the permeability and the link between chemical water/rock ratios and the fluid flow velocity was examined in more detail. Measurements of water/rock ratios are commonly based on the transfer of isotopes or conservative species between the fluid and the rock. None of the species in our chemical model behaves in a truly conservative manner. Instead, it was found that the silica/quartz system provides a reliable basis for mass transfer calculations and flow velocity estimates along the fluid flowpath. Aqueous silica is at (local) quartz equilibrium in large portions of the hydrothermal system and, as a result, correlates with the temperature. Thus, i f the temperature distribution along the fluid's flowpath is known, predictions can be made on the behavior of aqueous silica. Mass transfer has to be calculated along the fluids flowpath, which obviously requires not only a basic understanding of the flow geometry but also a stable flow field over the period of time, for which the mass transfer is integrated. Constraints on the geometry of flow can be made where preferential pathways exist, such as the higher permeability recharge and discharge zones, or, provided the flow field has stabilized, along the sediment basement interface. The sediments, covering the oceanic basement rocks, also offer suitable conditions for the application of this method. Not only can the flow direction be constrained, but also the duration of stable fluid flow if fluid flow was initiated at or shortly after the deposition of the sediments. In summary, the principal contributions of this thesis are: 306 1. A detailed investigation of transient characteristics of fluid flow and temperature in an open hydrothermal system and the possible implications on the chemistry of the system as a function of the basement permeability is presented. Average water/rock ratios or average time integrated fluid fluxes can be correlated with the basement permeability. 2. The coupling of the evolving flow and temperature conditions with chemical reactions provide a reasonable representation of chemical processes observed at oceanic hydrothermal systems such as that at Middle Valley. The model calculations suggest that the metamorphic grade of the alteration reactions is not sensitive to changes in the basement permeability. 3. A formulation of water/rock ratios for an open flow-through system and the relation of water/rock ratios to the chemical mass transfer along the fluid's flowpath are developed. Based on this formulation a method is presented which allows the estimate of average velocities along the flowpath. 4. The method suggests that silica mass transfer provides reliable velocity estimates under a wide range of flow and temperature conditions, in particular in those region of the system at or near quartz equilibrium so that the aqueous silica concentration is buffered by quartz and correlated with the temperature distribution. 307 References Aagaard, P. and H . C . Helgeson, Thermodynamic and kinetic constraints on reaction rates among minerals and aqueous solutions, I. Theoretical considerations. Am. J. Sci., 282, 237-285, 1982 Albarede, F. , A Michard, J.F. Minster, and G . Michard, 8 7 S r / 8 6 S r in hydrothermal waters and deposits from the East Pacific Rise at 21 °N, Earth Planet. Sci. Lett., 55, 229-236, 1981 Al t , J.C., J. Honnorez, C. Lavarne, and R. Emmermann, Hydrothermal alteration of a 1-km section through the upper oceanic crust, D S D P Hole 504B: Mineralogy, chemistry, and evolution of seawater/basalt interactions. J. Geophys. Res., 91, 10, 309-10,225, 1986a Al t , J.C., T.F. Anderson, and L . Bonnell, The geochemistry of sulfur in a 1.3 km section of hydrothermally altered oceanic crust, D S D P Hole 504B, Geochim. Cosmochim. Acta, 53,1011-1023, 1989a Al t , J .C. Subseafloor processes in mid-ocean ridge hydrothermal systems, in Seafloor Hydrothermal Systems: Physical, Chemical, Biological, and Geological Interactions, Geophys. Monogr. Ser., V o l . 91, edited by S.E. Humphris et al. 85-114, A G U , Wahington, D .C . , 1995 Baker, P .A . , S.L. Cross, and S.J. Burns, Geochemistry of carbonate nodules and cements and implications for hydrothermal circulation, Middle Valley, Juan de Fuca Ridge. Proceedings of the Ocean Drilling Program, Scientific Results, V o l . 139, Ocean Dri l l ing Program, College Station, Texas, 313-329, 1994 Bear, J., Dynamics of fluids in porous media, Elsevier, New York, 1972 Bear, J., Hydraulics of Groundwater, M c G r a w - H i l l , N e w York. 1979 Becker, K . et al. Deep drilling into young oceanic crust, Hole 504B, Costa R ica Rift, Reviews ofGeophysic, 27, 79-102, 1989 Becker, K . , R . H . Mor in , and E .E . Davis, Permeabilities in the Middle Val ley hydrothermal system measured with packer and flowmeter experiments. Proceedings of the Ocean Drilling Program, Scientific Results, V o l . 139, Ocean Dri l l ing Program, College Station, Texas, 613-627, 1994 Berndt, M . E . , W . E . Seyfried, and D.R. Janecky, Plagioclase and epidote buffering of cation ratios in mid-ocean ridge hydrothermal fluids: Experimental results in and near the supercritical region. Geochim. Cosmochim. Acta, 53, 2283-2300, 1989 308 Bessler, J .U. , L . Smith, E.E. Davis, Hydrologic and thermal conditions at a sediment/basalt interface: implications for interpretation of field measurements at Middle Valley. Proceedings of the Ocean Drilling Program, Scientific Results, V o l . 139, Ocean Dri l l ing Program, College Station, Texas, 667-679, 1994 Bethke, C M . , Geochemical Reaction Modeling. Oxford University Press, N e w York-Oxford, 1996 Bischoff, J .L. and F .W. Dickson, Seawater-basalt interaction at 200°C and 500 bars: inplications for orogin of seawater heavy-metal deposits and regulation of seawater chemistry, Earth Planet. Sci. Lett., 25, 385-397, 1975 Bischoff, J .L. and W.E. Seyfried, Hydrothermal chemistry of seawater from 25 to 350°C, Am. J. Sci., 278, 838-860, 1978 Bodvarrson, G . and R.P. Lowel l , Ocean-floor heat flow and the circulation of interstitial waters,/. Geophys. Res., 77, 4472-4475, 1972 Bohlke, J .K. , J. Al t , and K . Muehlenbachs, Oxygen isotope-water relations in altered deep sea basalts, Can. J. Earth Set, 21, 67-77, 1984 Bolton, E.W., A . C . Lasaga, D . M . Rye, Long-term flow/chemistry feedback in a porous medium with heterogeneous permeability; kinetic conraql of dissolution and precipitation. Am. J. Sci., 299, 1-68, 1999 Bowers, T.S., Stable isotope signatures of water-rock interaction in mid-ocean ridge hydrothermal systems: sulfur, oxygen, and hydrogen, J. Geophys. Res., 94, 5775-5786, 1989 Brantley, S.L., S.R. Crane, A . D . Crerar, R. Hellmann, and R. Stallard, Dissolution at dislocation etch pits in quartz. Geochim. Cosmochim. Acta, 50, 2349-2361, 1986 Buatier, M . D . , A . - M . Karpoff, M . Boni , G . L . Friih-Green, and J .A. McKenz ie , Mineralogic and petrographic record of sediment-fluid interaction in the sedimentary sequence at Middle Valley, Juan de Fuca Ridge, Leg 139. Proceedings of the Ocean Drilling Program, Scientific Results, V o l . 139, Ocean Dri l l ing Program, College Station, Texas, 133-155, 1994 Butterfield, D . A . , G.J . Massoth, R.E McDuff, J.E. Lupton, and M . D . Li l ley , Geochemistry of hydrothermal fluids from axial seamount hydrothermal emission study vent field, Juan de Fuca Ridge: subseafloor boiling and subsequent fluid-rock interaction. J. Geophys. Res., 95, 12895-12921, 1990 309 Butterfield, D . A . , R . E . McDuff, J. Franklin, and C . G . Wheat,Geochemistry of hydrothermal vent fluids from Middle Valley, Juan de Fuca Ridge. Proceedings of the Ocean Drilling Program, Scientific Results, V o l . 139, Ocean Dri l l ing Program, College Station, Texas, 395-411, 1994 Cann, J.R., Spilites from the Carlsberg Ridge, Indian Ocean. J. Petrol, 10, 1-19, 1969 Cann, J.R. and M . R . Strens, Modeling periodic megaplume emissions by black smoker systems, J. Geophys. Res., 94, 12227-12237, 1989 Cathles, L . , A capless 350°C flow zone model to explain megaplumes, salinity variations, and high temperature veins in ridge axis hydrothermal systems, Econ. Geol, 88, 1977-1988, 1993 Christensen, N.I . and R. Ramananantoandro, Permeability of the oceanic crust based on experimental studies of basalt permeability at elevated pressures, Tectonophysics, 149, 181-186, 1988 Currie, R . G . and E . E . Davis, L o w crustal magnetization of the Middle Valley sedimented rift inferred from sea-surface magnetic anomalies. Proceedings of the Ocean Drilling Program, Scientific Results, V o l . 139, Ocean Dri l l ing Program, College Station, Texas, 19-29, 1994 Davis, E .E . , J. Mott l , and A . T . Fisher (Eds.), Proceedings of the Ocean Drilling Program, Initial Results, V o l . 139, Ocean Dri l l ing Program, College Station, Texas, 1992b Davis, E . E . and H . Vill inger, Tectonic and thermal structure of the Middle Valley sedimented rift, northern Juan de Fuca Ridge. Proceedings of the Ocean Drilling Program, Initial Results, V o l . 139, Ocean Dri l l ing Program, College Station, Texas, 9-41, 1992 Davis, E .E . , D.S. Chapman, M . J . Mott l , W.J . Bentkowski, K . Dadey, C. Forster, R. Harris, S. Nagihara, K . Rohr, G . Wheat, and M . Whiticar, FlankFlux: A n .experiment to study the nature of hydrothermal circulation in young oceanic crust. Can. J. Earth Sci, 29(5), 925-952, 1992 Davis, E . E . and K . Becker, Formation temperatures and pressures in a sedimented rift hydrothermal system: 10 months of C O R K observations, Holes 857D and 858D. Proceedings of the Ocean Drilling Program, Scientific Results, V o l . 139, Ocean Dri l l ing Program, College Station, Texas, 649-667, 1994 Davis, E .E . , J. Mott l , and A . T . Fisher (Eds.), Proceedings of the Ocean Drilling Program, Scientific Results, V o l . 139, Ocean Dri l l ing Program, College Station, Texas, 1994 310 Davis, E.E. and A . T Fisher, On the nature and consequences of hydrothermal circulation in the Middle Valley sedimented rift: Inferences from geophysical and geochemical observations, Leg 139, Proceedings of the Ocean Drilling Program, Scientific Results, V o l . 139, Ocean Dri l l ing Program, College Station, Texas, 695-721, 1994 Davis, E.E. and K . Wang, Present and past temperatures of sediments at Site 857, Middle Valley, northern Juan de Fuca Ridge. Proceedings of the Ocean Drilling Program, Scientific Results, V o l . 139, Ocean Dri l l ing Program, College Station, Texas, 565-573, 1994 Davis, E.E. and D.S. Chapman, and C . B . Forster, Observations concerning the vigor of hydrothermal circulation in young volcanic crust. J. Geophys. Res., 101, 2927-1942, 1996 Davis, E.E., K . Wang, J. He, D.S. Chapman, H . Vill inger, and A . Rosenberger, A n equivocal case for high Nusselt-number hydrothermal convection in sediment-buried igneous oceanic crust, Earth Planet. Sci. Lett., 146, 137-150, 1997b Dewers, T. and P. Ortoleva, A coupled reaction/transport/mechanical model for intergranular pressure soltuion, stylolites, and different compaction and cementation in clean sandstones, Geochim. Cosmochim. Acta, 54, 1609-1625, 1990 Dewers, T. and P. Ortoleva, Nonlinear dynamical aspects of deep basin hydrology: Fluid compartment formation and episodic fluid release, Am. J. Sci., 294, 713-755, 1994 Dove, P . M . and D . A . Crerar, Kinetics of quartz dissolution in electrolyte solutions using a hydrothermal mixed flow reactor. Geochim. Cosmochim. Acta, 54, 955-969, 1990 Dullien, F . A . L . , Porous Media, Academic Press, San Diego, 396p. 1979 Fehn, U . and L . Cathles, Hydrothermal convection at slow-spreading mid-ocean ridges, Tectonophysics, 55, 239-260, 1979 Fehn, U . , K . Green, R.P . V o n Herzen, and L . Cathles, Numerical models for the hydrothermal field at the Galapagos Spreading Center, J. Geophys. Res., 88, 1033-1048, 1983 Fehn, U . , The evolution of low-temperature convection cells near spreading centers: A mechanism for the formation of the Galapagos Mounds and similar manganese deposits, Econ. Geol, 81, 1396-1407, 1986 Fehn, U . and L . Cathles, The influence of plate movement on the evolution of hydrothermal convection cells in the oceanic crust, Tectonophysics, 125, 289-312, 1986 311 Fisher, A . T . , K . Becker, T . N . Narasimhan, M . G . Langseth, and M . J . Mott l , Passive, off-axis convection in the southern flank of the Costa Rica Rift, J. Geophys. Res., 95, 9343-9370, 1990 Fisher, A . T . , K . Becker, and T . N . Narasimhan, Off-axis hydrothermal circulation: Parametric tests of a refined model of processes at Deep Sea Dri l l ing Project/Ocean Dri l l ing Program Site 504. J. Geophys. Res., 99, 3097-3121, 1994 Fisher, A . T . , K . Fischer, D.Lavoie, M . Langseth, and J. X u , Geotechnical and hydrogeological properties of sediments from Middle Valley, northern Juan de Fuca Ridge. Proceedings of the Ocean Drilling Program, Scientific Results, V o l . 139, Ocean Dri l l ing Program, College Station, Texas, 1994 Fisher, A . T . , Permeability within the basaltic oceanic crust. Reviews of Geophysics, 36, 2, 143-182, 1998 Forsyth, P .A . , and R . B . Simpson, A two-component model for natural convection on a propis medium. International Journal for Numerical Methods in Fluids, 12, 655-682, 1991 Gallahan, W . E . and R . A . Duncan, Spatial and temporal variability in crystallization of celadonites within the Troodos ophiolite, Cyprus: Implications for low-temperature alteration of the oceanic crust, J. Geophys. Res., 99, 3147-3161, 1994 Gi l l i s , K . and P.T. Robinson, Distribution of alteration zones in the upper oceanic crust, Geology, 16, 262-266, 1988 Gi l l i s , K . and P.T. Robinson, Patterns and processes of alteration in the lavas and dikes of the Troodos ophiolite, Cyprus, . / Geophys. Res., 95, 21523-21548, 1990 Goodfellow, W . D . and J . M . Peter, Geochemistry of hydrothermally altered sediment, Middle Valley, northern Juan de Fuca Ridge. Proceedings of the Ocean Drilling Program, Scientific Results, V o l . 139, Ocean Dri l l ing Program, College Station, Texas, 207-291, 1994 Gupta, A . D . , L . W . Lake, G . A . Pope, K . Sephernoori, and M . J . K i n g , High-resolution monotonic schemes fro reservoir fluid flow simulation. In Situ, 15, 289-317, 1991 Hamano, Y . , Physical properties from basalts from Hole 417D and 418A, Initial Rep. Deep Sea Drilling Prof, 51-53, 1457-1466, 1980 Harper, G .D. , J.R. Bowman, and R. Kuhns, A field, chemical, and stable isotope study of subseafloor metamorphism of the Josephine ophiolite, California-Oregon, J. Geophys. 93, 4625-4656, 1988 312 Hart, S.R. and H . Staudigel, Ocean crust vein mineral deposition: Rb/Sr ages, U-Th-Pb geochemistry, and duration of circulation at D S D P Sites 261, 462, and 516. Geochim. Cosmochim. Acta, 50, 2751-2761, 1986 Hartline, B . K . , and C . R . B . Lister, Topographic forcing of supercritical convection in a porous medium such as the oceanic crust, Earth. Planet. Sci. Lett., 55, 75-86, 1981 Haymon, R . M . , R . A . Kosk i , and M . J . Abrams, Hydrothermal discharge zones beneath massive sulfide deposits mapped in the Oman ophiolite, Geology, 17, 531-535, 1989 Helgeson, H . C . , W . M . Murphy, and P. Aagaard, Thermodynamic and kinetic constraints on reactions rates among minerals and aqueous solutions. II. Rate constants, effective surface area, and the hydrolosis of feldspar. Geochim. Cosmochim. Acta, 48, 2405-2432, 1984 Hewett, T .A . , Porosity and mineral alteration by fluid flow through a temperature field, in Reservoir Charcterizsation, edited by Lake, L . W. and H . B . Carroll, 83-140, Academic, San Diego, Calif. 1986 Ibaraki, M . D . , D . Haley, and E . A . Sudicky, XMD version 1.3 User's Guide - An efficient sparse matrix solver library. Ohio State University, Columbus, Ohio: Department of Geological Sciences, 1996 Janecky, D . A . and W . C . Shanks, Compositional modeling of chemical and sulfur isotope reaction processes in seafloor hydrothermal systems: Chimneys, massive sulfides, and subjacent alteration zones. Can. Mineral, 26, 805-825, 1988 Johnson, D . M . , Crack distribution in the upper oceanic crust and its effects upon seismic velocity, seismic structure, formation permeability, and fluid circulation, Initial Rep. Deep Sea Drilling Prof, 51-53, 1479-1490, 1980a Johnson, J.W., E . H . Oelkers, and H . C . Helgeson, SUPCRT92: A software package or calculating the standard molal thermodynamic properties of minerals, gases, aqueous species, and reactions from 1 to 5000 bars and 0°C to 1000°C. Computers in Geosciences, 1992 210 210 Kadko, D . , M . P . Bacon, and A . Hudson, Enhanced scavenging of Pb and Po by processes associated with the East Pacific Rise near 8°45' N . Earth Planet. Sci. Lett., 81, 349-357, 1986 Karato, S., Physical properties of basalts from the Galapagos, Leg 70, Initial Rep. Deep Sea Drilling Prof, 69, 687-695, 1983a 313 Kirkner, D.J. and H. Reeves, Multicomponent mass transport with homogeneous and heterogeneous chemical reactions: Effect of chemistry on the choice of numerical algorithm. I. Theory. Water Resources Research, 24, 1719-1729, 1988 Knapp, R.B., Spatial and temporal scales of local equilibrium in dynamic fluid-rock systems, Geochim. Cosmochim. Acta, 53, 1955-1964, 1989 Kurnosov, V. , I . Murdmaa, T. Rosanova, G. Kashintzev, V . Eroshchev-Shak, and S. Krasnov, Mineralogy of hydrothermally altered sediments and igneous rocks at Sites 856 - 858, Middle Valley, Juan de Fuca Ridge, Legl39. Proceedings of the Ocean Drilling Program, Scientific Results, Vol . 139, Ocean Drilling Program, College Station, Texas, 113-133, 1994 Langseth, M . G . and K. Becker, Structure of igneous basement at Sites 857 and 858 based on Leg 139 downhole logging. Proceedings of the Ocean Drilling Program, Scientific Results, Vol . 139, Ocean Drilling Program, College Station, Texas, 573-585,1994 Larson, R.L., A.T. Fisher, and R. Jarrard, Highly layered and permeable Jurassic oceanic crust in the western Pacific, Earth Planet. Sci. Lett., 119, 71-83, 1993 Lasaga, A.C. , Rate laws in chemical reactions, in Kinetics of Geochemical Processes, Reviews in Mineralogy, Mineral. Soc.Am., Lasaga, A .C . and R.J. Kirkpatrick (eds.), Vol 8, 135-169, 1981 Lasaga, A . C , Chemical kinetics of water-rock interactions. J. Geophys. Res., 89, 4009-4025,1984 Lasaga, A . C , Atomic treatment of mineral-water surface reactions, in Mineral-Water Interface Geochemistry: Reviews in Mineralogy, Hochella, M.F.Jr . and A.F . White (eds.), Vol 23, 17-85, 1990 Leybourne, M.I. and W.D. Goodfellow, Mineralogy and mineral chemistry of hydrothermally altered sediment, Middle Valley, Juan de Fuca Ridge. Proceedings of the Ocean Drilling Program, Scientific Results, Vol . 139, Ocean Drilling Program, College Station, Texas, 155-207, 1994 Lichtner, P . C , Continuum model for simultaneous chemical reactions and mass transport in hydrothermal systems. Geochim. Cosmochim. Acta, 49, 779-800, 1985 Lister, C.R.B., On the thermal balance of a mid-ocean ridge, Geophys. J. R. Astron. Soc, 26,515-535, 1972 Lister, C.R.B., On the penetration of water into hot rock, Geophys. J. R. Astron. Soc, 39, 465-509, 1974 314 Lowel l , R.P. , Topographically driven subcritical hydrothermal convection in the oceanic crust. Earth. Planet. Sci. Lett., 49, 21-28, 1980 Lowel l , R.P. , and P . A . Rona, Hydrothermal models for the generation of massive sulfide deposits, J. Geophys. Res., 90, 8769-8783, 1985 Lowel l , R .P . and D . K . Burnell, Mathematical modeling of conductive heat transfer from a freezing, convecting magma chamber to a single-pass hydrothermal system; implications for seafloor black smokers. Earth Planet. Sci. Lett., 104, 1, 59-69, 1991 Lowel l , R.P. , P . A . Rona, and R.P. von Herzen, Seafloor Hydrothermal Systems, J. Geophys. Res., 100, 327-352, 1993 Lowel l , R .P . and L . N . Germanovich, On the temporal evolution of high-temperature hydrothermal systems at ocean ridge crests, J. Geophys. Res., 99, 565-575, 1994 Lowel l , R.P. and L . N . Germanovich, Dike injection and the formation of megaplumes at ocean ridges, Science, 267, 1804-1807, 1995 Marsily, G . de, Quantitative Hydrogeology. Academic Press, N e w York, 1986 M c C o l l o m T. and E . L . Shock, Fluid-rock interaction in the lower oceanic crust; thermodynamic models of hydrothermal alteration. J. Geophys. Res., 103, 547-575, 1998 Meyer, C . A . , R . B . McClintock, G.J. Silverstri, and J.R.C. Spencer, ASME steam tables: Thermodynamic and transport properties of steam (6 t h ed.). N e w York, N . Y . : American Society of Mechanical Engineers, 1983 Mott l , M. J . , Metabasalts, axial hot springs, and the structure of hydrothermal systemsat mid-ocean ridges, Geol. Soc. Am. Bull.,94, 161-180, 1983 Mott l , M. J . , C . G . Wheat, and J. Boulegue, Timing of ore deposition and sil l intrusion at Site 856: evidence from stratigraphy, alteration, and sediment pore-water composition. Proceedings of the Ocean Drilling Program, Scientific Results, V o l . 139, Ocean Dri l l ing Program, College Station, Texas, 679-695, 1994 Nehlig, P. Fracture and permeability analysis in magma-hydrofhermal transition zones in the Samail ophiolite (Oman), J. Geophys. Res., 99. 589-601, 1994 Ortoleva, P. E . Merino, C. Moore, and J. Chadam, Geochemical self-organization, I, Reaction-transport feedbacks and modeling approach, Am. J. Sci., 287, 979-1007, 1987 Parsons, R . L . and J.G. Sclater, A n analysis of the variation of ocean floor bathymetry and heat flow with age. J. Geophys. Res., 82, 803-829, 1977 315 Patankar, S.V., Numerical Heat Transfer and Fluid Flow. Hemisphere Publishing Corporation, 1980 Person, M . , J.P. Raffensperger, S. Ge, and G . Garven, Basin-scale hydrogeologic modeling, Review of Geophysics, 34, 1, 61-87, 1996 Peter, J . M . , W . D . Goodfellow, and M.I . Leybourne, Fluid inclusion petrography and microthermometry of the Middle Valley hydrothermal system, northern Juan de Fuca Ridge. Proceedings of the Ocean Drilling Program, Scientific Results, V o l . 139, Ocean Dri l l ing Program, College Station, Texas, 411 -429, 1994 Phillips, O . M . , F low and reactions in permeable rocks. Cambridge University Press, Cambridge, 277 p. Pinder, G.F. and W . G . Gray, Finite element simulation in surface and subsurface hydrology. Acad. Press, N e w York, N . Y . , U . S . A . , 295 pages, 1977 Rabinowicz, M , J. Boulegue, and P. Genthon, Two -and three-dimensional modeling of hydrothermal convection in the sedimented Middle Valley segment, Juan de Fuca Ridge. J. Geophys. Res, 103, 24,045-24,065, 1998 Raffensperger, J.P. and G . Garven, The formation of unconformity type uranium ore deposits, 1, Coupled groundwater flow and heat transport modeling, Am. J. Sci., 295, 581-636, 1995a Raffensperger, J.P. and G . Garven, The formation of unconformity type uranium ore deposits, 2, Coupled hydrochemical modeling, Am. J. Sci., 295, 639-696, 1995b Reed, M . H . Calculation of multicomponent chemical equilibria and reaction processes in systems involving minerals, gases, and an aqueous phase. Geochim. Cosmochim. Acta, 46, 513-528, 1982 Ribando, R., K . Torrence, and D . Turcotte, Numerical models for hydrothermal circulation in the oceanic crust. J. Geophys. Res., 81, 3007-3012, 1976 Richards, H . G . , J.R. Cann, and J. Jensenius, Mineralogical zonation and metasomatism of the alteration pipes of Cyprus sulfide deposits, Econ. Geol, 84, 91-115, 1989 Richardson, C.J , J.R. Cann, H . G . Richards, and J .G. Cowan, Metal-depleted root zones of the Troodos ore-forming hydrothermal systems, Cyprus. Earth Planet. Sci. Lett., 84, 243-253,1987 Rigsby, C . A . , R . A . Zierenberg, and P . A . Baker, Sedimentary and diagenetic structures and textures in turbiditic and hemiturbiditic strata as revealed by whole-core X -radiography, Middle Valley, northern Juan de Fuca Ridge. Proceedings of the Ocean 316 Drilling Program, Scientific Results, V o l . 139, Ocean Dri l l ing Program, College Station, Texas, 105-113, 1994 Rimstidt, J .D. and H . L . Barnes, The kinetics of silica-water reactions, Geochim. Cosmochim. Acta, 44, 1683-1700, 1980 Rohr, K . M . M . and U . Schmidt, Seismic structure of Middle Valley near Sites 855-858, Leg 139, Juan de Fuca Ridge. Proceedings of the Ocean Drilling Program, Scientific Results, V o l . 139', Ocean Dri l l ing Program, College Station, Texas, 3-19, 1994 Rosenberg, N . D . , F.J . Spera, and R . M . Haymon, The relationship between flow and permeability field in seafloor hydrothermal systems. Earth. Planet. Sci. Lett., 116, 135-153,1993 Seyfried, W . E . and J .L. Bischoff, L o w temperature basalt interaction with seawater: an experimental study at 70°C and 150°C. Geochim. Cosmochim. Acta,A3, 1937-1947, 1979 Seyfried, W . E . and M . J . Mott l , Hydrothermal alteration of basalt by seawater under seawater-dominated conditions. Geochim. Cosmochim. Acta, 46, 985-1002, 1982 Seyfried, W . E . , Experimental and theoretical constraints on hydrothermal alteration processes at mid-ocean ridges, Ann. Rev. Earth Planet. Sci., 15, 317-335, 1987 Seyfried, W . E . , M . E . Berndt, and J.S. Seewald, Hydrothermal alteration processes at mid-ocean ridges: constraints from diabase alteration experiments, hot-spring fluids, and composition of the oceanic crust, Can. Mineral., 26, 787-804, 1988 Seyfried, W.E.,Jr. , Ding, K . , and Berndt, M . E . , Phase equilibria constraints on the chemistry of hot spring fluids at mid-ocean ridges. Geochimica, Cosmochimica Acta, 55, 3559-3580, 1991 Shanks, W . C . I l l , J .L. Bischoff, and R.J . Rosenbauer, Seawater sulfate reduction and sulfur isotope fractionation in basaltic systems: Interaction of seawater with fayalite and magnetite at 200-350°C. Geochim. Cosmochim. Acta, 45, 1977-1995, 1981 Shanks, W . C , III, J .K. Bohlke, and R.R. Seal II, Stable isotopes in mid-ocean ridge hydrothermal systems: interaction between fluids, minerals, and orgeanisms, in Seafloor Hydrothermal Systems: Physical, Chemical, Biological, and Geological Interactions, Geophys. Monogr. Ser., 91, edited by Humphris et al. 222-247, A G U , Washington, D C , 1995 Shido, F., A . Miyashiro, and M . Ewing, Compositional variation in pi l low lavas from the Mid-Atlantic Ridge, Mar . Geol, 16, 177-190, 1974 317 Shipboard Scientific Party, Site 857, Proceedings of the Ocean Drilling Program, Initial Reports, V o l . 139, Ocean Dri l l ing Program, College Station, Texas, 283-429, 1994 Snelgrove, S.H., and C . B . Forster, Impact of seafloor sediment permeability and thickness on off-axis hydrothermal circulation: Juan de Fuca Ridge eastern flank, J. Geophys. Res., 101,2915-2925,1996 Snow, D.T. , The frequency and apertures of fractures in rock. International Journal of Rock Mechanics Mineral Science, V o l . 7, 23-40 Spooner, E .T .C . , H.J . Chapman, and J.D. Smewing, Strontium isotopic contamination and oxidation during ocean floor hydrothermal metamorphism of the ophiolitic rocks of the Troodos massif, Cyprus, Geochim. Cosmochim. Acta, 41, 873-890, 1977 Stakes, D.S. and J . M . Franklin, Petrology of igneous rocks at Middle Valley, Juan de Fuca Ridge. Proceedings of the Ocean Drilling Program, Scientific Results, V o l . 139, Ocean Dri l l ing Program, College Station, Texas, 79-105, 1994 Steefel, C.I. and P. van Cappellen, A new kinetic approach to modeling water-rock interaction,: The role of nucleation, precursors, and Ostwald ripening, Geochim. Cosmochim. Acta, 54, 2657-2677, 1990 Steefel, C.I. and A . C . Lasaga, A coupled model for transport of multiple chemical species and kinetic precipitation/dissolution reactions with application to rective flow in single phase hydrothermal systems, Am. J. Sci., 294, 529-592, 1994 Steefel, C.I. and S.B. Yabusaki, O S 3 D / G I M R T . Software for Multicomponent-Multidimensional Reactive Transport, User Manual and Programmer's Guide. P N L -11166, Pacific Northwest National Laboratory, Richland, Washington, 1996 Steefel, C.I. and K . T . B . MacQuarrie, Approaches to modeling of reactive transport in porous media, in Reactive Transport in Porous Media, Reviews in Mineralogy, Mineral. Soc.Am., Lichtner, P .C. , C.I. Steefel, E . H . Oelkers (eds.), V o l . 34, 83-125, 1996 Stern, C. and D . Elthon, Vertical variations in the effects of hydrothermal metamorphism in Chilenean ophiolites: Their implications for ocean floor metamorphism, Tectonophysics, 55, 179-213, 1979 Strang, G . , On the construction and comparison of difference schemes. SIAMJ. Numer. Anal, 5, 506-517, 1969 Strens, M . R . and J.R. Cann, A model of hydrothermal circulation in fault zones at mid-ocean ridge crests, Geophys. J. R. Astron. Soc, 71, 225-240, 1982 318 Thompson, J.B., Local equilibrium in metasomatic processes, in Researches in Geochemistry, P.H. Abelson (ed.), John Wiley& Sons, New York, 427-457, 1959 Tivey, M.A. , High-resolution magnetic surveys over the Middle Valley mounds, northern Juan de Fuca Ridge. Proceedings of the Ocean Drilling Program, Scientific Results, Vol . 139, Ocean Drilling Program, College Station, Texas, 29-39, 1994 Travis, B.J., D.R. Janecky, and N.D. Rosenberg, Three-dimensional simulation of hydrothermal circulation at mid-ocean ridges, Geophys. Res. Lett., 18, 1441-1444, 1991 Valsami-Jones, E. and J.R. Cann, Controls on the Sr and Nd isotopic compositions of hydrothermally altered rocks from the Pindos ophiolite, Greece, Eart. Planet. Sci. Lett., 125,39-54, 1994 Villinger, H.W. M.G. Langseth, H . M . Groschel-Becker, and A.T. Fisher, Estimating in-situ thermal conductivity from log data. Proceedings of the Ocean Drilling Program, Scientific Results, Vol . 139, Ocean Drilling Program, College Station, Texas, 545-553, 1994 Von Damm, K.L . , Seafloor hydrothermal activity; black smoker chemistry and chimneys, Annual Review of Earth and Planetary Sciences, 18, 173-204, 1990 Von Damm, K . L . , D.C. Colodner, and H.N. Edmonds, Hydrothermal fluid chemistry at 9-10°N EPR '92: Big changes and still changing, EOS Trans. AGU, 75, Fall Mtg. Suppl., 524,1992 Von Damm, K . L . , Controls on the chemistry and temporal variability of seafloor hydrothermal fluids, in Seafloor Hydrothermal Systems: Physical, Chemical, Biological, and Geological Interactions, Geophys. Monogr. Ser., 91, edited by Humphris et al. 222-247, A G U , Washington, D . C , 1995 White, A.F. and M . L . Peterson, Role of reactive-surface-area characterization in geochemical kinetic models, in Chemical modeling of aqueous systems II, edited by D.C. Melchior and R.L. Bassett, ACS Symposium Series, 416, 461-475, 1990 Wilcock, W.S.D., and A. McNabb, Estimates of crustal permeability on the Endeavour segment of the Juan de Fuca mid-ocean ridge, Earth. Planet. Sci. Lett., 138, 83-91, 1996 Wilcock, W.S.D., A model for the formation of the transient event plumes above mid-ocean ridge hydrothermal systems, J. Geophys. Res., 102, 12109-12121, 1997 Wells, J.T. and M.S. Ghiorso, 1991 Coupled fluid flow and reaction in mid-ocean ridge hydrothermal systems; the behavior of silica. Geochim. Cosmochim. Acta, 55, 9, 2467-2481, 1991 319 Wolery, T.J., EQ3NR, a computer program for geochemical aqueous speciation-solubility calculations: User's guide and documentation: Lawrence Livermore Laboratory U R C L -53414, 1983 Yang, J., R.N. Edwards, J.W. Molson, and E.A. Sudicky, Three-dimensional numerical simulation of the hydrothermal systemwithin TAG-like sulfide mounds. Geophys. Res. Lett., 23, 3475-3478, 1996 Zierenberg, R.A., W.C. Shanks, W.E. Seyfried, R.A. Koski, and M.D. Strickler, Mineralization, alteration and hydrothermal metamorphism of the ophiolite-hosted Turner-Albright sulfide deposits, southwest Oregon. J. Geophys. Res., 93, 4657-4674, 1988 Zysset, A. , F. Stauffer, and T. Dracos, Modeling of chemically reactive groundwater transport. Water Resources Research, 30, 2217-2228, 1994 320 "@en ; edm:hasType "Thesis/Dissertation"@en ; vivo:dateIssued "2000-11"@en ; edm:isShownAt "10.14288/1.0053172"@en ; dcterms:language "eng"@en ; ns0:degreeDiscipline "Oceanography"@en ; edm:provider "Vancouver : University of British Columbia Library"@en ; dcterms:publisher "University of British Columbia"@en ; dcterms:rights "For non-commercial purposes only, such as research, private study and education. Additional conditions apply, see Terms of Use https://open.library.ubc.ca/terms_of_use."@en ; ns0:scholarLevel "Graduate"@en ; dcterms:title "Insights into the evolution of an oceanic hydrothermal system and a method for constraining estimates of the vigor of hydrothermal convection"@en ; dcterms:type "Text"@en ; ns0:identifierURI "http://hdl.handle.net/2429/12860"@en .