Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
THERMAL STATE OF CONTINENTAL AND OCEANIC LITHOSPHERE by Derrick P. Hasterok A dissertation submitted to the faculty of The University of Utah in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Geophysics Department of Geology and Geophysics The University of Utah August 2010 c Derrick P. Hasterok 2010 Copyright All Rights Reserved THE UNIVERSITY OF UTAH GRADUATE SCHOOL STATEMENT OF DISSERTATION APPROVAL The dissertation of Derrick P. Hasterok has been approved by the following supervisory committee members: , Chair 21 May 2010 John Bartley , Member 21 May 2010 Barbara Nash , Member 21 May 2010 Philip E. Wannamaker , Member 21 May 2010 Richard C. Aster , Member 21 May 2010 David S. Chapman and by the Department of D. Kip Solomon Geology and Geophysics and by Charles A. Wight, Dean of The Graduate School. , Chair of ABSTRACT The thermal state of the continental and oceanic lithosphere is reassessed on the basis of new databases for global heat flow and lithospheric heat production, recent advances in thermophysical properties measurements of minerals at high pressures and temperatures, and a better understanding of convective heat loss in young seafloor. The updated global heat flow database incorporates >60,000 records with >44,800 heat flow determinations. The update significantly increases the quantity and spatial coverage of global heat flow data since the last update in 1993. A new family of continental geotherms is proposed that is parametric in surface heat flow and takes advantage of thermophysical property data. The range of geotherms is constrained by xenolith P –T estimates; a cratonic geotherm consistent with a surface heat flow of 40 mW/m2 is particularly well constrained. Upper crustal heat production represents ∼26% of the total surface heat flow. Average heat production for the continental lower crust and mantle are 0.4 µW/m3 and 0.02 µW/m3 , respectively. Recent controversy about the interpretation of heat flow observations in young seafloor is resolved by careful filtering of data based on sediment thickness and distance from seamounts and weighting marine studies where the environment of heat flow measurements is carefully documented. Oceanic geotherms, fit to bathymetry and heat flow data, are produced for a plate model with 7 km thick crust, a plate thickness of 95 km, and mantle potential temperature of 1425◦ C. While the current estimate of global heat loss (44 TW) is reasonable, these new reference models will be instrumental in refining and estimating uncertainty in the solid Earth’s global heat loss. CONTENTS ABSTRACT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v LIST OF TABLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii ACKNOWLEDGEMENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii CHAPTERS 1. INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 2. AN UPDATED GLOBAL HEAT FLOW DATABASE . . . . . . . . . . 4 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2 Updated Global Heat Flow Database . . . . . . . . . . . . . . . . . . . . . . . . . 4 4 HEAT PRODUCTION AND GEOTHERMS FOR THE CONTINENTAL LITHOSPHERE . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 11 12 13 14 16 21 27 38 3. 4. Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Geotherm Sensitivity to Heat Production . . . . . . . . . . . . . . . . . . . . . . Previous Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Observational Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . OCEANIC HEAT FLOW: IMPLICATIONS FOR GLOBAL HEAT LOSS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 4.1 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3 Datasets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4 Thermal Model of Sea-floor Spreading . . . . . . . . . . . . . . . . . . . . . . . . . 4.5 Observed Heat Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.6 Global Filtering—Sediments and Seamounts . . . . . . . . . . . . . . . . . . . . 4.7 Environmental Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.8 Corrections to Global Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.9 Global Heat Loss . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.10 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 40 41 44 45 48 57 61 64 66 5. 6. PLATE COOLING MODELS FOR THE OCEANIC LITHOSPHERE: ARE COMPLEXITIES NECESSARY? . . . . . . . 68 5.1 5.2 5.3 5.4 5.5 5.6 5.7 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Theoretical Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Datasets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A Note on McKenzie et al. [2005] . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68 68 69 76 82 96 97 CONCLUSIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99 APPENDICES A. RADIATIVE DIFFUSIVITY AND CONDUCTIVITY OF OLIVINE REVISITED . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102 B. THERMAL CONDUCTIVITY OF AMPHIBOLES: INFLUENCE OF COMPOSITION . . . . . . . . . . . . . . . . . . . . . . . . . . 112 C. MODELS OF THERMAL CONDUCTIVITY FOR INDIVIDUAL MINERALS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124 D. EMPIRICAL CONSTANTS USED TO MODEL PHYSICAL PROPERTIES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131 E. SIMPLIFIED CONTINENTAL GEOTHERMS . . . . . . . . . . . . . . . . 135 REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138 iv LIST OF FIGURES 2.1 The global, continental and oceanic heat flow distributions. . . . . . . . . . 6 2.2 Heat flow locations in the updated global heat flow database. . . . . . . . . 7 3.1 Temperature sensitivity to 25 and 50% variations in heat production. . . 14 3.2 Heat production models of the lithosphere. . . . . . . . . . . . . . . . . . . . . . . 15 3.3 Heat production from granulite terranes and mantle xenoliths. . . . . . . . 19 3.4 Estimated P –T conditions for mantle xenoliths using the PBKN and TBKN barometer and thermometers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 3.5 Misfit of heat production models to compositionally adjusted elevation. 28 3.6 Best-fitting thermal isostatic models for each of the heat production models. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 3.7 Fits of preferred geotherm family (P = 0.74) to xenolith P –T estimates for the Kalahari craton. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 3.8 Lithospheric thickness and heat loss. . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 4.1 Oceanic datasets. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 4.2 Thermal isostasy of the oceanic lithosphere. . . . . . . . . . . . . . . . . . . . . . 45 4.3 Observed oceanic heat flow. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46 4.4 Effect of sediment cover on hydrothermal circulation and heat flow through the oceanic lithosphere. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 4.5 Median heat flow versus age in 2 m.y. bins divided into groups with 50 m increments of sediment thickness. . . . . . . . . . . . . . . . . . . . . . . . . . 49 4.6 Metrics for improvement in globally filtered datasets as a function of minimum sediment thickness and minimum distance to seamounts. . . . 52 4.7 Global filtering results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 4.8 Heat flow case study from Juan de Fuca flank. . . . . . . . . . . . . . . . . . . . 58 4.9 Filtered heat flow adjusted for sedimentation and thermal rebound. . . . 60 4.10 Estimated heat flow response to sedimentation. . . . . . . . . . . . . . . . . . . . 62 4.11 Estimated fraction of thermal rebound as a function of sediment cover and time since cessation of hydrothermal circulation. . . . . . . . . . . . . . . 63 4.12 Estimated global advective power loss from Monte Carlo analysis with 106 realizations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 5.1 Effective thermal expansivity model. . . . . . . . . . . . . . . . . . . . . . . . . . . . 73 5.2 Influence of plate thickness variations on subsidence and heat flow. . . . 83 5.3 Influence of mantle potential temperature variations on subsidence and heat flow. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84 5.4 Influence of crustal thickness variations on subsidence and heat flow. . . 86 5.5 Slices through the misfit surface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89 5.6 Preferred plate cooling model with a crustal thickness of 7 km, plate thickness of 90 km and potential temperature of 1425◦ C including heat production. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91 5.7 Model sensitivity to selected parameters. . . . . . . . . . . . . . . . . . . . . . . . . 92 A.1 Thermal diffusivity of olivine. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105 A.2 Radiative diffusivity models and data for olivine. . . . . . . . . . . . . . . . . . 107 A.3 Comparison of geotherms computed from several radiative conductivity models. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109 B.1 Thermal conductivity of amphiboles. . . . . . . . . . . . . . . . . . . . . . . . . . . . 118 B.2 Thermal conductivity of solid solution minerals . . . . . . . . . . . . . . . . . . . 120 B.3 Influence of amphibole composition on lithospheric temperatures. . . . . . 121 C.1 Effective thermal conductivity of quartz. . . . . . . . . . . . . . . . . . . . . . . . . 124 C.2 Effective thermal conductivity of muscovite. . . . . . . . . . . . . . . . . . . . . . 125 C.3 Effective thermal conductivity of orthoclase. . . . . . . . . . . . . . . . . . . . . . 125 C.4 Lattice thermal conductivity of plagioclase. . . . . . . . . . . . . . . . . . . . . . . 126 C.5 Lattice thermal conductivity of orthopyroxene. . . . . . . . . . . . . . . . . . . . 127 C.6 Lattice thermal conductivity of olivine. . . . . . . . . . . . . . . . . . . . . . . . . . 127 C.7 Lattice thermal conductivity of garnet. . . . . . . . . . . . . . . . . . . . . . . . . . 129 C.8 Lattice thermal conductivity of spinel. . . . . . . . . . . . . . . . . . . . . . . . . . . 130 vi LIST OF TABLES 2.1 Number of heat flow measurement sites . . . . . . . . . . . . . . . . . . . . . . . . . 5 2.2 Example records to the updated global heat flow database. . . . . . . . . . . 9 3.1 Compostional model used to compute geotherms. . . . . . . . . . . . . . . . . . 22 4.1 Heat flow from environmental analysis. . . . . . . . . . . . . . . . . . . . . . . . . . 58 5.1 Parameters for sediment correction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 5.2 Global bathymetry and heat flow data. . . . . . . . . . . . . . . . . . . . . . . . . . 80 5.3 Compositional model for the oceanic lithosphere. . . . . . . . . . . . . . . . . . 87 5.4 Compositional model for the oceanic lithosphere. Values as molar fraction of approximate end-member mineralogy. . . . . . . . . . . . . . . . . . . . . . 94 B.1 Amphibole compositions and conductivity. . . . . . . . . . . . . . . . . . . . . . . 115 B.2 Fitting constants for estimating conductivity coefficients . . . . . . . . . . . . 117 D.1 Physical properties and empirical constants for mineral end-members. . 132 D.2 Empirical constants for estimating conductivity. . . . . . . . . . . . . . . . . . . 133 D.3 Empirical constants for computing heat capacity of mineral end-members.134 E.1 Empirical conductivity constants. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136 E.2 Empirical expansivity constants. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137 ACKNOWLEDGEMENTS Many people helped make this work possible by contributing their own data and compilations to this database. The following researchers made substantial contributions to this end. Therefore, I would like to thank Francis Lucazeau, Jeffery Poort, Bruno Goutorbe, Dallas Abbott, Carol Stein, Valiya Hamza, Maria Richards, Jacek Majorowicz, Graeme Beardsmore, Niels Balling, Trond Slagstad, Andrea Förster, Bruno Della Vedova, Roman Kutas, Georg Delisle, Karla Bojadgieva, and Shaopeng Huang. Xenolith P –T databases provided by Roberta Rudnick and Derek Bell helped tremendously in modeling continental geotherms. Discussions with Juan Carlos Afonso, Jun Korenaga, Tom Shankland, and Christopher Grose were helpful in modeling rock properties used in the continental and oceanic studies. During the course of this research, I needed to collect hundreds of references for the global heat flow database, the xenolith database, and a crustal thickness database currently under construction for global heat flow modeling. The document delivery service and interlibrary loan at the University of Utah were indispensible and saved me weeks of hunting down articles and reports. I would like to thank Lucy Flesch at Purdue University for use of computing resources which proved invaluable at the end stages. The members of my committee, Barbara Nash, John Bartley and Rick Aster, provided excellent guidance that helped shape my scientific understanding of many subjects covered and related to this work, both through coursework and discussion. I am happy to call each a mentor and look forward to many fruitful discussions in the future. I would especially like to thank my advisors David Chapman and Phil Wannamaker for being encouraging, understanding and excellent role models for a young scientist. Most of all, I would like to thank my lab mates, Paul Gettings, Michael Davis, Melissa Mursbach, Imam Raharjo, Bryce Johnson and Christian Hardwick, and my wife Kristine Nielson for fielding numerous questions and giving excellent suggestions which helped improved the quality of this research. ix CHAPTER 1 INTRODUCTION My Ph.D. research seeks a better understanding and description of the thermal state of the continental and oceanic lithosphere. Principal components of this study of the continental lithosphere involve assembling a comprehensive, expanded database of heat flow observations, a vertical heat production profile for the continental lithosphere, a family of continental geotherms parametric in surface heat flow, and constraining geotherms using thermal isostasy and xenolith P –T estimates. For oceanic lithosphere, I focus on filtering heat flow data to reveal the background conductive heat loss of the oceanic lithosphere, estimating the spatial extent and magnitude of redistribution of heat in young seafloor by hydrothermal processes, and using plate cooling models with P –T dependent thermophysical properties to model heat flow– and subsidence–age patterns for the seafloor. Chapter 2 describes the expanded global heat flow dataset assembled for this study. Since the last update, significant advances have been made in the experimental study of mineral behavior under a large range of P –T conditions. Using this updated database and improved estimates of thermophysical properties, the distribution of heat loss within the lithosphere can be revisited in a substantial and substantive way. Chapters 3 through 5 illustrate the beginning stages of this reassessment. These reference thermal models calibrate observables, such as elevation/bathymetry, which can then be used as predictors of surface heat flow in data-poor regions. Chapter 3 explores the average contributions of radioactive heat generation to temperature profiles and heat loss out of the continental lithosphere. Because temperature governs many physical and chemical states and thermophysical properties that influence other geologic processes, it is important to establish reference geotherm 2 models; radiogenic heat production is shown to be one of the most important properties for computing lithospheric temperatures and heat flow across the asthenosphere/lithosphere boundary. Chapter 4 examines a systematic bias in oceanic heat flow data resulting from a fluid infiltration in sediments where measurements are collected. The background conductive heat loss of the oceanic lithosphere is estimated by filtering heat flow both globally and in several high resolution studies. Emphasis is placed estimating the redistribution of heat in young seafloor due to hydrothermal circulation of seawater through the oceanic lithosphere. Chapter 5 computes a number of plate-cooling models, which explore the importance of physical complexities and the addition of a crustal layer not commonly included in these models. The focus is on developing an average cooling model for the oceanic lithosphere that fits observed heat flow– and bathymetry–age relationships. The families of geotherms developed in this study incorporate several recent advances in describing pressure and temperature effects on thermal conductivity, thermal diffusivity, expansivity, density, and specific heat capacity. Because data and information are drawn from multiple sources, the thermophysical information is compiled in five appendices. Appendix A examines the radiative effect on thermal conductivity and diffusivity of olivine. This corrects an error in the model that is most commonly employed. Appendix B proposes a method for determining a relationship between composition and thermal conductivity for amphiboles. Appendix C includes figures showing the pressure and/or temperature sensitivity of thermal conductivity for a number of minerals used to calibrate the model used throughout this work. Model fits are shown with the thermal conductivity data used to calibrate them. Appendix D lists the empirical constants necessary to estimate P –T dependence of thermophysical properties used to model continental and oceanic geotherms and isostasy. 3 Appendix E provides empirical constants that can be used to more easily compute the geotherm family derived in Chapter 3. CHAPTER 2 AN UPDATED GLOBAL HEAT FLOW DATABASE 2.1 Introduction Advances in understanding the physical and chemical state of the Earth and its evolution have often followed the growth and increased spatial coverage of global geophysical data sets. For example, recognition of the geometry of plate boundaries followed the accumulation of earthquake locations in a catalog, and the details of sea-floor spreading were revealed after accumulating vast quantities of marine magnetic data. The global heat flow dataset has likewise been instrumental in identifying relationships between heat flow and tectonics on land, the importance of heat production for analysis of heat flow patterns, relationships between heat flow and age on the sea-floor, and the role of hydrothermal circulation in redistributing heat in young oceanic lithosphere. 2.2 Updated Global Heat Flow Database The first version of the global heat flow database contained a mere 63 measurements, 43 on the continents and 20 on the oceans [Birch, 1954]. The most recent update by Pollack et al. [1993] contained 20,201 heat flow determinations with roughly equal fractions on the oceans and continents. The database used in this study now contains 60,398 records (58,008 heat flow determinations), with approximately 25% oceanic and 75% continental data (Table 2.1). About 9,300 sites are located on the continental shelves which nearly triples these data. Many of the new continental data come from 14,483 BHT (bottom hole temperature) estimates in the Gulf of Mexico oil states and Cordillera foreland basin [Majorowicz et al., 1999; Blackwell and Richards, 2004]. 5 Table 2.1. Number of heat flow measurement sites Reference Birch [1954] Lee [1963] Lee and Uyeda [1965] Horai and Simmons [1969] Lee [1970] Jessop et al. [1976] Chapman and Pollack [1980] Chapman and Rybach [1985] Pollack et al. [1993] This study Continentala 43 73 131 474 597 1,699 2,808 3,601 10,337 (13,249) 33,693 (43,060) Oceanic 20 561 913 2,348 2,530 3,718 4,409 5,181 9,684 (6,952) 24,315 (14,948) Total 63 634 1,044 2,822 3,127 5,417 7,217 8,782 20,201 58,008 a Values in parentheses breakdown continental and oceanic sites when marine data on continental shelves are reclassified as continental. The most frequent occurence of heat flow is between 50 and 55 mW/m2 for the global distribution as well as the continental and oceanic subsets (Figure 2.1). The global median heat flow is 65 mW/m2 . Heat flow in the ocean basins (79 mW/m2 ) is higher than the global average with a broader distribution and a long tail containing several observations in each bin up to 1000 mW/m2 . The resulting intraquartile range is large for the oceanic observations (∼139 mW/m2 ). Continental heat flow shows a sharper peak than the oceanic data with a rapid decline in the number of observations above 150 mW/m2 . The intraquartile range for the continents is therefore much smaller (37 mW/m2 ). It should be noted that these statistics are not necessarily the true continental or oceanic heat flow because spatial bias inherent in this dataset occludes the true spatially averaged heat flow. The updated global heat flow database significantly improves the data coverage on the continents, particularly in west Africa, China, South America and in the Canadian shield (Figure 2.2). Improved spatial coverage in the oceans is most evident on the Antarctic Plate, north-central Pacific, Atlantic east of Argentina and the Arctic Ocean. However, significant gaps in data still persist, with few or no data reported in central Africa, Middle East, Brazilian rainforests, Argentina, and Himalayas. Gaps in oceanic data occur over much of the southern Oceans, Arctic Ocean, and Pacific. 6 3500 All Data N: 44861 Quartiles: [48.0, 65.1, 97.6] mW/m2 Number of Observations 3000 2500 Continental N: 29868 Quartiles: [46.0, 62.3, 83.2] mW/m2 2000 Oceanic N: 14993 Quartiles: [50.6, 79.0, 189.0] mW/m2 1500 1000 500 0 0 0 0 0 150 0 100 50 100 150 Heat Flow [mW/m2] 200 150 250 200 200 250 250 Figure 2.1. The global, continental and oceanic heat flow distributions. Median values are shown in bold and indicated by solid lines on histograms. BHT data from Majorowicz et al. [1999] and Blackwell and Richards [2004] are excluded to prevent strongly biasing the histogram. Each histogram shift is 500 observations. 7 Figure 2.2. Heat flow locations in the updated global heat flow database. Data from Pollack et al. [1993] (blue) and this update (red). Plate boundaries shown as thin black line. Oceanic regions are shown in white with light grey and dark grey regions for the subaqueous and subaerial continental regions, respectively. Data are presented using a Robinson projection. Only a handful of heat flow determinations exist on Greenland and Antarctica where rocks are not covered by ice sheets. The extensive cover by ice sheets makes determinations of heat flow logistically more challenging. I add three additional heat flow determinations to the global dataset that are made in ice-boreholes on slow moving ice sheets that may reflect the underlying lithospheric heat flow. I removed several heat flow estimates from isotopes in hot springs so that only heat flow estimates made directly from temperature measurements are included. A number of duplicates are also removed. About 50% of the database has been checked against the original references and several typographical and position errors are corrected. Beginning with the database version by Jessop et al. [1976], database records were fixed to 80 character widths, which resulted in several compromises. In previous versions, letter codes were used to indicate metadata, such as the type of instrument used to measure temperature, method of estimating thermal conductivity, and the country where the site is located. These metadata are now explicitly given to make the database more readable and some metadata has been removed because GIS programs can easily serve the same function. A single letter code was used to indicate that 8 a correction was applied to the heat flow, although multiple corrections could not be listed individually. The new update seeks to fix this issue by giving each of the corrections and value if possible. Examples of three of sites, are given in Table 2.2. Continental data are often recorded in “holes of opportunity” in boreholes drilled for mining, petroleum, or water resource exploration. The vast majority of newly added continental data are from BHTs collected in oil exploration and are generally lower quality than well-sampled and isolated boreholes. BHTs are temperatures measured at the bottom of exploration wells and may be measured at several stages during and after drilling. However, the depth of the temperature measurement, typically more than 3000 m, and the number of BHTs collected within a region add confidence to average heat flow estimates (see spatial clustering in Figure 2.2). About the time of the last global update, oceanic heat flow studies benefited from a combination of an improved apparatus used to measure temperatures in ocean sediments and high resolution positioning using GPS. These improvements led to a paradigm shift in most survey designs away from regional heat-flow mapping to high resolution transects and three-dimensional surveys that focus less on background heat flow and more on individual processes. Thus, despite the large growth in the oceanic dataset, many of the new data cluster in locations where few sites already existed. The updated global heat flow database compiled as part of this research represents considerable improvement to past updates and represents a significant contribution to the thermal geophysics and solid Earth geoscience communities. This improved and updated database, along with recent laboratory studies of pressure and temperature dependence on thermophysical properties, present an opportunity to revisit the thermal state of the continental and oceanic lithosphere. 9 Table 2.2. Example records to the updated global heat flow database. New data number Old data number Fluids Bottom water variation Refraction Topographic Climatic Sedimentation/Erosion Compaction Other Grade Site Name Latitude Longitude Elevation Bottom water temp. Bottom hole temp. Minimum depth Maximum depth No. of temp. Temperature gradient Gradient uncertainty Corr. temp. gradient Corr. gradient uncertainty No. of conductivity Conductivity Conductivity error Conductivity method 18429 XM 44 18772 CAN129 -19 8.8 -9 9.8 -4.7 B CT-29 39.3333 13.7333 -3497 Sivakasi SV-3 9.4619 77.7950 150 366 6.3 5 175 1.1 186 21 0.86 0.06 transient (needle probe) No. of heat production Heat production Heat production error Heat production method Heat flow Heat flow error Corr. heat flow 31690 151 17.15 160 18.9 0.03 A6-A7 48.5917 -123.4967 -228 2.5 6 43 29 4 2.4 0.04 6 0.72 transient (needle probe) 25 0.6 0.14 gamma-ray spectrometer 45 0.7 21 10 Table 2.2. continued. Corr. heat flow error No. of determinations Publication year Reference Verified Borehole type Environment Comments Age Lithology Sediment Thickness 1 1970/1977/1984 Erickson etal1977; DellaVedova etal1984; Erickson1970 * Ewing probe 1 2003 Ray etal2003 13 1 1976 Hyndman1976 borehole (conventional) * lake (Bullard probe) corrected heat flow was recomputed charnockite CHAPTER 3 HEAT PRODUCTION AND GEOTHERMS FOR THE CONTINENTAL LITHOSPHERE 3.1 Abstract Accurate estimates of heat production are necessary for computing lithospheric temperatures and heat flow into the base of the lithosphere. Heat production, however, is one of the most variable and difficult parameters to estimate using surface geophysical exploration methods. I propose a generalized continental lithospheric heat production model that partitions crustal heat production into upper crustal and lower crustal contributions and is constrained by thermal isostasy observations from 33 North American tectonic provinces. An average heat production for the lower crust is estimated as 0.4 µW/m3 from exposed granulite terranes and for the lithospheric mantle as 0.02 µW/m3 from chemical analyses of xenoliths. The best-fitting partition model suggests upper crustal heat production accounts for ∼26% of the observed surface heat flow. Results are relatively insensitive to mantle composition and thickness of the upper crustal heat producing layer. Continental geotherms are computed using the generalized heat production model and incorporating thermal conductivity results from a number of recent laboratory studies. Estimated P –T conditions of xenoliths provide constraints to ensure that my geotherms are reasonable. Fits to P –T conditions of 10 Precambrian regions suggest surface heat flow is ∼40 mW/m2 with a lithospheric thickness of ∼200 km. My average model for North American heat production can be used as a reference model from which observed anomalies can be identified. 12 3.2 Introduction Radiogenic heat generation, created by the decay of K, Th, and U, accounts for an estimated 30–40% of heat loss through the continents [Pollack and Chapman, 1977; Vitorello and Pollack , 1980; Artemieva and Mooney, 2001; Hasterok and Chapman, 2007a]. Accurate estimates of radiogenic heat production are important in computing lithospheric temperatures and heat flow across both the Moho and the asthenosphere/lithosphere boundary, as well as many physical parameters that depend upon temperature (e.g., density, seismic velocity, viscosity, elevation) all of which have important geodynamic implications [Jaupart and Mareschal , 1999; Flowers et al., 2004; Hyndman et al., 2005; Sandiford and McLaren, 2002]. However, reliable estimates of heat production are difficult to obtain at depths greater than a few hundred meters using common geophysical and geochemical exploration methods, making it one of the least constrained physical parameters within the lithosphere. Therefore, it is desirable to have a general heat production model that can be used as both a reference for comparison, and a starting model for thermal and geodynamic studies of the lithosphere. Several general models for heat production exist, but they are based on crustal age [Jaupart and Mareschal , 2003], which poorly correlates with heat production, or are calibrated only for shield and cratonic regions [Rudnick et al., 1998; Rudnick and Nyblade, 1999]. Neither of these model types adequately reflect the chemical differentiation of the crust. Models based on heat production–heat flow relationships derived from a collection of geologic/tectonic provinces show promise because they implicitly assume both a decrease in heat production with increasing juvenile crustal age and an increase in mantle heat flow in active tectonic regions [Chapman, 1986; Artemieva and Mooney, 2001]. These partition models are generally applied as a function of a single observable, surface heat flow. However, sampling of upper crustal heat production is generally insufficient to characterize a given province with sufficient accuracy. Continental elevation, much like ocean bathymetry, responds to lateral differences in the average thermal state of the lithosphere [Hasterok and Chapman, 2007b]. 13 Using a set of compositionally normalized elevation data I estimate the average contribution of heat producing elements to the observed heat flow. While the exact distribution of heat production is somewhat difficult to determine using this method, estimates of pressure and temperature from mantle xenoliths place limits on the vertical distribution of heat production. In this study, I develop an average heat production model for the North American continental lithosphere using compositionally normalized elevations from improved geotherm computations. The range of geotherms is constrained using estimated xenolith P –T conditions. 3.3 Geotherm Sensitivity to Heat Production The sensitivity of lithospheric temperatures and estimates of lithospheric thickness to heat production is illustrated by perturbing a standard model. Consider a three layered lithosphere, with a high heat producing upper crust (1 µW/m3 ), a depleted lower crust (0.4 µW/m3 ), and a low heat producing mantle (0.02 µW/m3 ). For this test, I define a reference geotherm with an intermediate surface heat flow of 60 mW/m2 (see Section 3.6.1 for geotherm construction). This reference geotherm reaches the 1300◦ C adiabat at ∼90 km (Figure 3.1). By varying only the heat production within each layer individually by ±25 and ±50% I investigate the sensitivity of temperatures and lithospheric thickness to the depth of the heat production anomaly. A ±50% variation in upper crustal heat production produces a difference in lithospheric thickness of ∼50 km and >400 K temperature variation at 75 km. The effect of ±25 and ±50% lower crustal heat production variance results in 10 km and 20 km lithospheric thickness estimates, respectively. The temperature difference is significantly smaller (∼175 K at 75 km for ±50% variation) than differences resulting from similar percentage variations the upper crustal heat production. For the same relative variation in heat production, the effect on temperature and lithospheric thickness is more than twice as sensitive to changes in the upper crustal layer compared to changes in the lower crustal heat production. As mantle heat production is varied, one can see that the effect is negligible for both ±25 and ±50% variation. In order to rival the 14 0 0 500 1000 1500 0 Temperature [°C] 500 1000 1500 0 500 1000 1500 1 µW/m3 25 0.4 µW/m3 Depth [km] 50 75 0.02 µW/m3 100 Reference ±25% ±50% 125 150 Figure 3.1. Temperature sensitivity to 25 and 50% variations in heat production of the (left) upper crust, (middle) lower crust, and (right) mantle. The reference geotherm is computed with 1 µW/m3 upper crust, 0.4 µW/m3 lower crust and 0.02 µW/m3 mantle. Geotherms are computed with a surface heat flow of 60 mW/m2 , upper crustal thickness of 16 km, and 39 km depth to the Moho. lithospheric thickness variation seen in the lower crustal case, mantle heat production would have to vary by greater than ±400%. It is apparent from these tests that unless upper crustal heat production is precisely known, deviations from average lower crustal and lithospheric mantle heat production can be masked by upper crustal uncertainties. Therefore, I focus my analysis only on variations in upper crustal heat production. 3.4 Previous Models Several models for crustal and/or lithospheric heat production exist, each using different constraints (Figure 3.2). Allis [1979]; Rybach and Buntebarth [1984] suggested using seismic velocities to predict heat production, but heat production for any given rock is highly nonunique. Attempts to estimate crustal heat production using xenolith P –T conditions provide reasonable averages (0.5–0.8 µW/m3 ) [Rudnick et al., 1998], but are only calibrated to Precambrian provinces. Rudnick and Nyblade [1999] used 15 0 0 1 2 1 2 3 Heat Production [µW/m3] 3 0 1 0 1 2 0 1 2 3 4 5 10 Depth [km] 20 30 Ar 40 Ph Pt (a) (b) (c) (d) 50 Figure 3.2. Heat production models of the lithosphere. (a) Heat production model from Allis [1979] is compositionally dependent; models are shown for 1, greenstone; 2, gneiss terrane; and 3, granite. (b) Rudnick et al. [1998]; Rudnick and Nyblade [1999] estimated heat production in Precambrian terranes from xenolith P –T –conditions and lithospheric thickness constraints. (c) Jaupart and Mareschal [2003] estimated the heat production within the crust of Archean (Ar), Proterozoic (Pt), and Phanerozoic (Ph) terranes by using surface heat flow constraints on the maximum mantle heat flow. (d) The heat production models of Chapman [1986] (black) and Artemieva and Mooney [2001] (dashed) are derived from reduced heat flow, surface observations, and xenolith P –T –conditions. xenolith P –T conditions to test mantle as well as crustal heat production. However, the results are less useful, as three out of five of the best-fitting grid search parameters lie on their boundaries of the allowed range, including mantle heat production. This suggests that either the ranges of allowed parameters were too restrictive or some of the physics necessary to describe system accurately is not included (e.g., P –T dependent thermal conductivity). Jaupart and Mareschal [2003] use surface heat flow constraints to make an elegant estimate of crustal heat production and find a significant variation in crustal heat production with juvenile crustal age. They suggest the present-day bulk-crustal heat production increases from ∼0.65 µW/m3 in Archean to ∼0.87 µW/m3 in Phanerozoic terranes. However, the correlation between age and heat production may be weak [Rao et al., 1982]. For example, the Wopmay Orogen and much of Precambrian Australia have anomalously high heat flow due to high upper-crustal radioactivity 16 [Lewis et al., 2003; McLaren et al., 2003]. Additionally, their analysis does not distinguish between upper and lower crustal heat production. Chapman [1986]; Artemieva and Mooney [2001] take a different approach, estimating upper crustal heat production from an empirically derived partitioning of surface heat flow between an enriched upper crustal radiogenic heat flow (qrad ) and reduced (or basal, qb ) heat flow [Pollack and Chapman, 1977]. Partition estimates range from 60:40 (qb :qrad ) [Pollack and Chapman, 1977; Vitorello and Pollack , 1980] to 67–71:33–29 [Artemieva and Mooney, 2001]. A major advantage of the partitioning model, compared with the other models discussed above, is a dependence on surface heat flow, an observable that directly responds to changes in the upper crustal radioactivity. Thus the partition model can be applied generally to any region where surface heat flow is observed or can be estimated. Because of the reliance on surface heat flow, the partition model implicitly assumes that tectonically active regions are frequently have higher than average mantle heat flow and a decrease in heat production with age. 3.5 Observational Constraints In order to derive a general heat production model for the continental lithosphere, it is important to know the range of heat production as well as the general variation with depth. Heat production observations and estimates are summarized below for each of the lithospheric layers. 3.5.1 Upper Crust Direct measurements of heat production are generally high in felsic rocks (∼2 µW/m3 ), low in mafic rocks (∼0.2 µW/m3 ) and very low in ultramafic rocks (∼0.02 µW/m3 ). Variations of heat production within individual and well-sampled terranes can vary by at least an order of magnitude over small spatial scales (<1 km) [Kukkonen and Lahtinen, 2001; Ray et al., 2003]. Individual terranes commonly have standard deviations 50–100% of the sample mean or greater, and frequently exhibit nonGaussian distributions (e.g., Ketcham [2006]; Jõeleht and Kukkonen [1998]; Brady 17 et al. [2006]; Jaupart and Mareschal [2003]). Few terranes, however, have been subjected to sufficiently detailed heat production investigations to be helpful to heat flow modeling. Airborne radiometric surveys provide good spatial averages, but the depth of penetration is ∼30 cm and often does not see through sedimentary cover [Bodorkos et al., 2004]. Even when good surface spatial averages can be obtained, heat production variation with depth is still problematic. Heat production with depth is frequently modeled using one of three functional forms: constant value layers, linearly decreasing, and exponentially decreasing [Lachenbruch, 1970]. The exponential model was developed to explain observations of heat flow–heat production patterns observed in granitic batholiths with differing levels of surface erosion [Roy et al., 1968; Birch et al., 1968; Swanberg, 1972]. I test these models for vertical heat production variation using deep boreholes and exposed crustal cross-sections, which provide insight into the nature of heat production with depth. Some constraints on vertical heat production variation come from deep boreholes and exposed crustal cross-sections. The research wells KTB in Germany and Kola SG-3 in Russia are the deepest scientific boreholes at 9 and 12.2 km, respectively. KTB is drilled into a series of layered gneiss and metabasites and SG-3 is drilled into a Precambrian volcanic and metamorphic terrane. The heat production for each well shows a reasonable correlation to rock type [Clauser et al., 1997; Popov et al., 1999] but an irregular pattern with depth. The Chinese scientific borehole (CCSD-MH) also exhibits a correlation to lithology [He et al., 2008]. However, heat production is quite variable within any given lithologic unit. Even within boreholes drilled predominantly into granite, heat production can be highly variable [Lachenbruch and Bunker , 1971; Balling et al., 1990; Vigneresse and Cuney, 1991]. No single functional form can explain the patterns of heat production with depth for these boreholes. Exposed crustal sections that can be related to pseudo-depth profiles provide greater spatial coverage, extending knowledge to greater depths than boreholes. Measurements of heat production have been collected in metamorphic core complexes (Catalina and Harquahala Mountains [Ketcham, 2006]), impact structures (Vredefort and 18 Sudbury [Nicolaysen et al., 1981; Schneider et al., 1987]), thrust sheets (Zentralgneis, Wawa-Foleyet, Pikwitonei-Sachigo, Hidaka Arc, and Arunta-Muscrave [Hawkesworth, 1974; Ashwal et al., 1987; Fountain et al., 1987; Furukawa and Shinjoe, 1997; McLaren et al., 2003]), structural folds (Egersund-Bamble, Erzgebirge [Pinet and Jaupart, 1987; Förster and Förster , 2000]), and batholiths (Idaho Batholith, and Sierra Nevada Mountains [Swanberg, 1972; Brady et al., 2006]). These exposed sections also show little correlation of heat production with estimated depth beyond a general pattern of higher values near the surface where felsic rocks are dominant and lower values in mafic rocks of the lower crust. In regions where thrust sheets duplicate parts of the crustal section, heat production often exhibits a repeating pattern [McLaren et al., 2003; Ketcham, 2006; He et al., 2008; Clauser et al., 1997]. Given the poor correlation of heat production with depth in the upper crust, I suggest a reference model with a constant upper crustal heat production. A constant heat production with depth is also the least complicated functional form, making it easy to compute anomalies. 3.5.2 Lower Crust Studies of seismic velocity and equilibrium conditions from xenoliths suggest that the lower crust is more mafic on average than the upper crust and best described by granulite metamorphic facies [Christensen and Mooney, 1995; Rudnick and Fountain, 1995]. To estimate lower crustal heat production, observations from exposed granulite terranes and xenoliths provide the best insight [Rudnick and Fountain, 1995]. Geochemical estimates of heat producing elements are converted to heat production, A, by A = 10−5 ρ [3.5CK2 O + 9.67CU + 2.63CTh] , (3.1) with concentrations of K2 O in wt.%, and of U and Th in ppm. I assume densities, ρ, of 2800, 2850, and 3000 kg/m3 for felsic, intermediate, and mafic granulites, respectively. Heat production measurements from 31 exposed granulite terranes range from 0.1 to 2.7 µW/m3 with a mean of 0.68±0.62 µW/m3 and a median 0.45 µW/m3 (Figure 3.3). Many of the higher heat production terranes are mainly felsic rather than mafic 19 10 # of Granulites 8 6 N: 41 (31 terranes) Mean: 0.68±0.62 µW/m3 Median: 0.45 µW/m3 4 2 0 0 0.5 1 1.5 2 Heat Production [µW/m3] 2.5 20 # of Xenoliths 16 12 N: 97 xenoliths Mean: 0.031±0.024 µW/m3 Median: 0.022 µW/m3 8 4 0 0 0.02 0.04 0.06 0.08 Heat Production [µW/m3] 0.1 Figure 3.3. Heat production from granulite terranes (top) and mantle xenoliths (bottom). Locations are shown by squares on inset maps. Mean values are given ± one standard deviation. Granulite data from Ashwal et al. [1987]; Fountain et al. [1987]; Pinet and Jaupart [1987]; Ray et al. [2003]; Kukkonen and Jõeleht [1996]; Förster and Förster [2000]; Brady et al. [2006]; Garrido et al. [2006]; Del Lama et al. [1998]; Attoh and Morgan [2004]; Owen et al. [68]; Martignole and Martelat [2005]; Hölttä [1997] and references therein. Xenolith data from Ackerman et al. [2007]; Bjerg et al. [2005]; Ionov et al. [1993]; Ionov [2004]; Ionov et al. [2002]; Peltonen et al. [1999]; Rudnick et al. [2004]; Wiechert et al. [1997]; Ionov et al. [2005]; Bianchini et al. [2007]; Xu et al. [1998]. 20 granulites. If only mafic granulites are considered, the mean heat production is 0.36±0.50 µW/m3 with a median of 0.15 µW/m3 . However, it is difficult to derive completely independent estimates of mafic and felsic granulites using this dataset because the reported heat production values are commonly terrane averages rather separated into mafic and felsic compositions. Xenolith-derived heat production values are lower on average than from exposed terranes with means (medians) of felsic, intermediate, and mafic granulites of 0.85 (0.57), 0.20 (0.09), and (0.13) 0.06 µW/m3 , respectively [Rudnick and Fountain, 1995]. This difference could be due to metasomatic processes that affect surface exposures or xenoliths during exhumation and/or near surface groundwater flow [Jaupart and Mareschal , 2003]. I do not include a middle crust in my modeling because the temperature sensitivity to middle and lower crustal heat production variations is small relative to the upper crust. The mid-crustal layer is not pervasive globally [Rudnick and Gao, 2003], and where it exists, tends to have heat production more similar to lower rather than to upper crust. My model includes a value of 0.4 µW/m3 for the lower crust, which is slightly higher than typical for mafic granulite and lower than most intermediate rocks like amphibolite and tonalite. 3.5.3 Lithospheric Mantle Xenoliths provide the best estimates of mantle heat production on continents. Average heat production from compilations of continental lithospheric peridotites varies from 0.006 µW/m3 in exposed off-craton massifs to 0.044 µW/m3 in cratonic xenoliths [Rudnick et al., 1998]. However, xenoliths, particularly kimberlites, are subject to significant disturbances by metasomatic processes during ascent. Excluding kimberlite xenoliths, the median heat production for cratonic peridotites is estimated to be 0.019 µW/m3 [Rudnick et al., 1998]. In general, bulk U and Th concentrations are rarely measured in xenoliths. The above heat production estimates are based on assumed abundances of U and Th relative to K concentration [Rudnick et al., 1998]. 21 Heat-producing elements U and Th are typically found in monazite, a phosphate mineral similar to apatite. Studies focusing on apatite in mantle peridotites suggest that even at very small concentrations, apatite can dominate heat production [Ionov et al., 1996; O’Reilly et al., 1997; O’Reilly and Griffin, 2000]. For example, 1% by weight apatite in a mantle rock can increase the total heat production by 0.3 µW/m3 . Since apatite is a common accessory mineral but rarely reported, it unknown whether these high heat production values are widespread or isolated locally [Ionov et al., 1996; O’Reilly et al., 1997; O’Reilly and Griffin, 2000]. However, the minor to negligible curvature in xenolith P –T estimates suggests mantle heat production is on average very small. I compiled heat production for several xenolith localities that include K2 O, U, and Th in the bulk chemical analysis (Figure 3.3). Heat production is computed using Equation 3.1 with an assumed density of 3300 kg/m3 . Total heat production for these xenoliths ranges from 0.003 to values greater than 1 µW/m3 with two exceptionally high values ≫1 µW/m3 . In general, the values >0.1 µW/m3 come from localities showing signs of metasomatism (e.g., Jericho xenoliths on the Slave craton from Russell et al. [2001]). A strong peak in the estimated mantle heat production histogram occurs below ∼0.02 µW/m3 . When restricted to values less than 0.1 µW/m3 , average heat production is 0.041±0.030 µW/m3 and a median of 0.025 µW/m3 . In my models, I assume a heat production of 0.02 µW/m3 for the mantle lithosphere. 3.6 Methods 3.6.1 Geotherms One-dimensional steady-state conductive geotherms are computed using a bootstrapping method, which requires thermal conductivity, heat production and surface heat flow as inputs (Appendix E). Because I use a complex P –T –composition-ally dependent thermal conductivity model, a Newton-Raphson iterative scheme is employed to solve for temperature. Pressure is computed using a logarithmic equation of state [Poirier and Tarantola, 1998]. 22 Table 3.1. Compostional model used to compute geotherms. Mineral quartz orthoclase albite anorthite phlogopite hornblende diopside hedenbergite enstatite ferrosillite forsterite fayalite pyrope almandine Upper 27 15 32 8 5 13 Crust Middle 15 5 35 20 20 2 3 Lower 2 10 10 18 47 1 1 1 1 9 Archon Mantlea Proton Tecton 1.96 0.14 23.22 1.78 64.04 4.96 3.19 0.81 5.47 0.53 15.47 1.53 63.65 3.35 5.58 1.42 9.97 1.03 15.37 1.63 54.20 5.80 9.57 2.43 Mantle compositions are derived from garnet xenocrysts and approximate Archean (Archon), Proterozoic (Proton), and Phanerzoic (Tecton) mantle [Griffin et al., 1999]. I assume a compositional model equivalent to a granodiorite upper crust (0–16 km), tonalite middle crust (16–23 km), and mafic granulite for the lower crust (23–39 km) (Table 3.1). Layer thicknesses for the crust correspond to estimates discussed by Rudnick and Gao [2003]. Average mantle compositions are based on garnet xenocryst estimates by Griffin et al. [1999]. Their peridotite compositions are derived from garnet xenocrysts and correspond approximately to average Archean, Proterozoic, and Phanerozoic continental lithospheric mantle. Garnet peridotites are converted to spinel-peridotites via 2 orthopyroxene + spinel ↔ garnet + olivine. The pressure of the spinel-garnet transition is estimated using the empirical relationship, Psg (T ) = 1.4209 + exp(3.9073 × 10−3 T − 6.8041), (3.2) where T is in Kelvins. This curve is calibrated by inversion of the data reported by Robinson and Wood [1998]; Walter et al. [2002]; Klemme and O’Neill [2000] and references therein. 23 3.6.1.1 Heat Production Assuming steady-state conditions, surface heat flow results from a combination of heat flow into the base of the lithosphere and the integrated heat production within the lithosphere. As discussed above, temperatures are very sensitive to variations in upper crustal heat production (Figure 3.1). Therefore, I focus only on variations in upper-crustal heat production and fix all other layers at constant values. Thus the middle crustal, lower crustal, mantle heat production, and sublithospheric heat flow can be combined into a single parameter, basal heat flow (qb ). The surface heat flow, qs , can then be written, qs = qb + AUC D (3.3) where AUC is the upper crustal heat production, D is the thickness of the upper crustal heat producing layer, and the product AUC D is the radiogenic heat flow in the upper crust. I use equation 3.3 to define three classes of heat production models. Class I – invariant upper crustal heat production where heat production is independent of the surface heat flow and all variations in surface heat production result from differences in lithospheric thickness and variation in sublithospheric heat flow. Class II – constant basal heat flow, where all variations in surface heat flow result from changes in upper crustal heat production. Hence the upper crustal heat production can be computed by AUC = (qs − qb )/D. (3.4) The invariant heat production model would be most applicable in regions such as rifts where significant variations in basal heat flow occur, but the average crustal heat production may be relatively constant [McKenzie, 1978]. Models with constant basal heat flow may best describe individual shields and cratons where the basal heat flow is believed to be relatively constant and variations in surface heat flow are still observed [Jaupart and Mareschal , 2003]. Globally, the patterns of surface heat flow are likely to result from a combination of these end-member models, hence the basal heat flow 24 represents only a fraction of the total heat flow (i.e., qb = F qs ). Thus, Class III – the partition model can be written, AUC = (1 − F )qs /D, (3.5) where F is the partition coefficient. This partition model has been used with great success in describing global heat flow/heat production patterns [Pollack and Chapman, 1977; Vitorello and Pollack , 1980; Artemieva and Mooney, 2001] and in developing a thermal isostatic relationship for North America [Hasterok and Chapman, 2007a]. I model geotherms using constant heat production within the upper crustal heat producing layer for two reasons. First, the commonly employed exponential decreasing heat production models with depth can exceed 5 µW/m3 for partition models with high heat flow. Heat production values this high are rarely observed regionally except in shields with anomalously high heat production, but can be identified by their associated low elevation anomalies such as the Wopmay orogen and North and South Australian cratons [Lewis et al., 2003; McLaren et al., 2003]. The second rationale for using constant heat production stems from highly variable heat production observations in exposed crustal sections and deep boreholes as discussed in Section 3.5.1. Heat production in the middle to lower crust is assumed to be 0.4 µW/m3 , consistent with granulite terranes. Mantle heat production is set to 0.02 µW/m3 as suggested by chemical studies of mantle xenoliths. 3.6.1.2 Thermal Conductivity Thermal conductivity is computed using a P –T –composition dependent model resulting from a combination of lattice and radiative components. Once the total conductivity of each mineral component is estimated, the effective thermal conductivity is computed using a geometric mixing model [Clauser and Huenges, 1995]. The lattice contribution is computed using a simplified form of the Equation 10 by Hofmeister [1999a], λL (P, T ) = λ ◦ 298 T n K′ 1+ TP , KT ! (3.6) 25 where temperature is in Kelvins, λ◦ is the conductivity at 0 GPa and 298 K, KT and KT′ are the isothermal bulk modulus and its first pressure derivative, and n is an empirically derived fitting constant. At lithospheric temperatures and pressures, the additional exponential factor included by Hofmeister [1999a] is very near unity and thus ignored [Beck et al., 2007]. Constants used to compute conductivity are given in Appendix D. The radiative contribution to thermal conductivity, λR , is negligible at room temperature but represents a significant fraction of the effective conductivity at high temperatures. I use an empirically derived relation for the radiative conductivity developed in Appendix A, λR (T ) = λRmax [1 + erf (ω(T − TR ))] , (3.7) where λRmax is the maximum radiative conductivity, ω is a scaling factor and TR is the temperature at 0.5λRmax . For olivine, this estimate is significantly different from previous estimates of the radiative contribution (Shankland et al. [1979] and references therein, Hofmeister [1999a, 2005]). A simplified thermal conductivity model is include in Appendix E for the specific compositional model used in this study. 3.6.2 Thermal Isostasy Much like the subsidence patterns of the oceanic sea-floor that result from the integrated cooling of the lithosphere (e.g., Parsons and Sclater [1977]), continental elevations record a thermal isostatic effect that can be estimated by examining the relationship between compositionally corrected elevation and surface heat flow [Han and Chapman, 1995; Hasterok and Chapman, 2007b,a]. Elevation responds to variations in temperature as a result of thermal expansion which changes the density of the lithosphere. In this section, I explore how continental elevation, much like oceanic bathymetry in the marine case, can be used as a constraint on continental geotherms, and consequently on heat production models for continental lithosphere. The elevation difference, ∆εT , between two regions can be computed by ∆εT = Z 0 zmax [αV (z, T )T (z) − αV′ (z, T )T ′ (z)] dz, (3.8) 26 where T (z) represents a lithospheric geotherm and αV (z, T ) is the P –T dependent volumetric expansivity computed using the method of Afonso et al. [2005]. The primed and unprimed symbols represent the reference and observed columns, respectively. The maximum depth of integration, zmax , is the depth at which the coolest geotherm reaches the mantle adiabat. Elevation data are normalized for variations in crustal thickness and density and represent 36 tectonic provinces from North America [Hasterok and Chapman, 2007a]. Three regions (Pacific Coast Ranges, Central Valley of California, and the Wopmay Orogen) are excluded from the fit due to large extraneous influences on the elevation and or heat flow (see Hasterok and Chapman [2007a] for discussion). Constants used to estimate expansivity and density are given in Appendix D. 3.6.3 Xenolith P –T Conditions While thermal elevation contributions can be used to estimate the average difference between geotherms, it is not possible to estimate absolute temperature. However, I can use xenolith P –T estimates as constraints on absolute temperature. Estimated temperatures and pressures for 1449 mantle xenoliths are shown in Figure 3.4. Temperatures range from 650 to 1500◦ C over a range of pressures equivalent to depths of 25 to 225 km. A few xenoliths from the Dabie-Sulu belt appear to be well outside the range of other localities, which reflects rapid subduction of material that thus did not have time to equilibrate to ambient mantle conditions. While there is some variation in high temperature and high pressure xenoliths >100 km, temperatures rarely exceed the 1300◦ C adiabat. In contrast to the high pressure–high temperature xenoliths, at low pressures there appear to be very few high temperatures above the 50 ppm melting curve (Figure 3.4). Either temperatures are buffered by melting, restricting maximum geotherm temperatures to the 50 ppm solidus and not the adiabat, or P –T conditions are poorly sampled. Alternatively, these xenoliths may be sampled but P –T estimates may not be possible using the Brey and Köhler [1990] thermobarometers. As melting continues to higher melt fractions, clinopyroxene is one of the first minerals to be exhausted [Katz et al., 2003]. The 27 0 sp-peridotite gt-peridotite 50 dr y 50 UHP 150 m pp Depth [km] 100 200 300 0 400 Adiabat ~2σ uncertainty wet N = 1449 250 800 1200 Temperature [°C] 1600 Figure 3.4. Estimated P –T conditions for mantle xenoliths using the PBKN and TBKN barometer and thermometers (circles). Conditions for the ultra-high pressure Dabie-Sulu belt shown in black [Zheng et al., 2006]. Estimated uncertainty (1-σ) is ±0.3 GPa (∼9 km) for pressure and 20 K for temperature [Brey and Köhler , 1990]. Melting curves by Katz et al. [2003]. The heavy black adiabat is represented by a potential temperature of 1300◦C with a gradient of 0.3◦ C/km. The spinel-garnet transition is given with a dashed line. removal of solid clinopyroxene makes T estimates from the 2-pyroxene thermometers such as TBKN used in this study impossible, suggesting that the lack of data does not preclude the higher temperatures at low pressure. This effect assumes a water concentration >50 ppm. 3.7 3.7.1 Results and Discussion Thermal Isostatic Models Misfits to compositionally normalized elevations of North America are shown in Figure 3.5. The misfit is computed by, 1/2 N N S2 X 1 X Misfit = (qi − qm )2 + (εj − εm )2 N i=1 N j=1 , (3.9) 28 2.0 Reference Heat Flow [mW/m2] 40 50 60 0 Temperature [°C] 800 1600 0 Class I 1.2 0.8 1.4 0.8 0.4 0.6 0.6 1 0.8 0.4 Invariant qref = 47.5 mW/m2 AUC = 0.7 µW/m3 300 0 Class II 0.6 25 200 0.4 Depth [km] 1 100 30 0.8 Basal Heat Flow [mW/m2] 0 35 200 Depth [km] 100 1.2 1 Invariant HP [µW/m3] 1.6 Basal qref = 39.5 mW/m2 qb = 23.5 mW/m2 20 1.0 300 0 1.4 0.4 0.8 1.2 0.6 Partition qref = 47.5 mW/m2 F = 0.74 Class III 40 50 60 2 Reference Heat Flow [mW/m ] 0 800 Temperature [°C] 200 Depth [km] 100 1 Partition Coefficient 1 0.8 0.6 0. 4 0.6 0.8 300 1600 Figure 3.5. Misfit of heat production models to compositionally adjusted elevation (left). Best-fit model represented as ‘×.’ Models where the reference heat flow does not reach the adiabat are shown in grey. Best-fitting geotherms for each associated heat production class (right). Geotherms range from 30–120 mW/m2 . 29 where N is the number of samples, qi and qm are the observed heat flow and model heat flow, and εi and εm are the adjusted province elevation and model elevation, respectively. In order to give roughly equal weights to the elevation and heat flow, the heat flow is scaled (S = 4 km/70 mW/m2 ). Results are only shown for upper-crustal heat-production thickness D = 16 km and Proton composition mantle (Table 3.1). Given the scatter in normalized elevations, it is unsurprising that the misfit surface shows that each model class has a wide range of reasonable parameters. The minimum misfit for the invariant and partition models are very similar (∼0.34) and only slightly higher for the constant basal heat flow model (0.39). Misfits using geotherms from this study show an improvement in fit over previous work using geotherm families computed using the method by Chapman [1986], which have a minimum misfit >0.7 [Hasterok and Chapman, 2007a]. The best-fitting thermal isostatic curve for the invariant model has an upper crustal heat production of 0.7 µW/m3 and reference heat flow of 47.5 mW/m2 . The reference heat flow corresponds to an adjusted elevation of zero. The elevation resulting from the invariant model is ∼2.5 km, and very similar in form to the partition model (Figure 3.6). The partition model results in the same reference heat flow as the invariant model with a partition coefficient of 0.74. This value is slightly higher than a recent global estimate of partition coefficient from Artemieva and Mooney [2001] and higher than my previous estimate of 0.6 using the Chapman [1986] style geotherms. This higher value results from my improved conductivity model and the use of P –T dependent thermal expansivity. The constant basal heat flow model results in a very different character to predicted thermal elevation resulting from the invariant and partition models (Figure 3.6), although the range of elevation for observed heat flow data is similar. Geotherms computed using the Class II model are also radically different from Class I and III (Figures 3.5). The minimum misfit has a reference heat flow of 39.5 mW/m2 , and a basal heat flow of 23.5 mW/m2 . Physically, the basal heat flow is a combination of the sublithospheric heat flow and radiogenic heat flow from the lower crust and mantle lithosphere. For a 40 mW/m2 cratonic region, my geotherm models predict 30 Adjusted Elevation [km] 3 2 I. Invariant, AUC II. Constant, qb III. Partition, P 1 0 -1 40 60 80 Heat Flow [mW/m2] 100 Figure 3.6. Best-fitting thermal isostatic models for each of the heat production models. Compositionally adjusted elevation for North American geologic provinces from Hasterok and Chapman [2007a], open squares not used in best-fitting model. Field of xenolith P –T estimates, range of adiabats and solidi curves from Figure 3.4. a lithospheric thickness of ∼200 km, and the total radiogenic contribution from the lower crust and mantle is then 12.4 mW/m2 . Using estimates of sublithospheric heat flow from the Canadian shield (11–15 mW/m2 ) [Mareschal and Jaupart, 2004], the predicted lower bound for basal heat flow is consistent with the best-fitting result. 3.7.1.1 Effect of Varying D The thickness of the upper crustal heat producing layer, D, need not be the same as the compositional upper crust as some processes such as melt migration and fluid circulation may concentrate heat producing elements toward the surface Jaupart et al. [1981]; Gosnold [1987]. Tests varying D between 6 and 18 km suggest that the partition coefficient is relatively insensitive to variations in D, ranging from 0.76 at 8 km to 0.74 at 16 km. For Class I models, iso-misfit lines closely track constant values of AUC D, implying a nearly constant upper crustal radiogenic heat flow. Best-fitting upper crustal radiative heat flow ranges from 9.9 mW/m2 at 6 km to 11.2 mW/m2 at 16 km. These observations do not hold when D and the compositional upper crustal 31 thickness are coupled as the effect of varying conductivity can trade-off for variations in D in misfit space. Independent estimates of D range from 2–16 km, with an average ∼8–10 km determined using heat flow–heat production relationships [Artemieva and Mooney [2001] and references therein]. Heat flow–heat production relationships were originally developed to explain a relationship between these two parameters for co-genetic plutons [Roy et al., 1968; Birch et al., 1968; Gosnold , 1987; Förster and Förster , 2000] and, although frequently attempted, may not work when extended to encompass all rocks in a given region [Nielson, 1987; Furlong and Chapman, 1987; Jaupart and Mareschal , 1999]. The value of D determined from heat production–heat flow relationships may also be more sensitive to the lateral scale of heat production variations as much or more than the vertical variations, representing a lower bound on the thickness of the upper-crustal heat-producing layer [Jaupart, 1983; Sandiford and McLaren, 2002]. 3.7.1.2 Effect of Mantle Composition Using Archon and Proton mantle compositions produce nearly identical results whereas the Tecton composition fits to elevation result in ∼10% higher total heat production with similar misfit. The higher heat production compensates for conductivity difference between the Archon/Proton and Tecton compositions. This result implies that younger tectonic regions are slightly enriched in upper crustal heat producing elements than older terranes. The reduction in heat producing elements with age and affinity for melts that migrate towards the surface in tectonically active regions contribute to the greater enrichment in the upper crust of young terranes. However, the conductivities of mantle phases other than olivine and garnet are not well constrained and require further study before these differences in model results can be rigorously modeled. 32 3.7.2 Xenotherms While my models fit the compositionally adjusted elevation, the scatter in modeled elevations allows for a large range of acceptable heat production models and therefore temperatures. Typical values for regional heat flow on the continents range from a little less than 40 mW/m2 to ∼100 mW/m2 and can reach 120 mW/m2 in magmatic provinces and active rift zones. However, conductive heat flow can be as low as ∼20 mW/m2 in cases of extremely low crustal heat production [Chapman and Pollack , 1974; Mareschal et al., 2000; Roy and Rao, 2000]. The best-fitting geotherm family for the constant basal heat flow model is cooler than the xenolith P –T field at low heat flow and does not reach the high temperatures recorded by xenoliths at intermediate and low pressures. Both the invariant and partition models appear to fall within the range of xenolith P –T field. The temperatures for the 40–120 mW/m2 geotherms are relatively similar. Neither geotherm family reaches the 1300◦ C adiabat by 300 km for a heat flow of 30 mW/m2 . However, at 30 mW/m2 the two geotherm models diverge considerably with a difference of ∼350 K at 300 km. In addition to examining the full P –T field from mantle xenoliths, I model the heat flow for 10 individual kimberlite provinces. Geotherms are computed using my preferred geotherm model, P = 0.74, with an Archon composition. Misfit for these xenotherms (xenolith derived geotherms) is computed using the misfit functional, N δTi2 δPi2 1 X + 2 misfit = N i=0 σT2 σP " !#−1/2 , (3.10) where δTi and δPi are the differences between the geotherm and xenolith P–T conditions. The differences are normalized by the estimated uncertainties in xenolith P–T results from the geobarometer PBKN (σP = 0.3 GPa) and TBKN (σT = 20 K) [Brey and Köhler , 1990]. Xenotherm misfits range from ∼3–5 with a surface heat flow uncertainty of ±2 mW/m2 estimated from the width of the misfit trough (insets in Figure 3.7). The xenotherm fits indicate that, at the time of xenolith equilibration, heat flow for the regions analyzed ranged from ∼37–47 mW/m2 with an average of 40 mW/m2 . 33 0 Kalahari Craton N: 323 Misfit: 3.4 HF: 40.2 4 25 Misfit Pressure [GPa] 2 6 0 30 8 0 HF 50 500 1000 Temperature [°C] 1500 Figure 3.7. Fits of preferred geotherm family (P = 0.74) to xenolith P –T estimates for the Kalahari craton. Misfit (inset) for variations in surface heat flow (HF) in mW/m2 computed using Equation 3.10. 34 0 0 500 Ouachita 1500 Slave Somerset N: 60 Misfit: 4.3 HF: 40.6 N: 111 Misfit: 5.3 HF: 37.2 N: 11 Misfit: 4.0 HF: 40 4 1000 8 0 Superior Pressure [GPa] W. Greenland N: 23 (27) Misfit: 2.3 (4.8) HF: 47 (46.2) N: 44 Misfit: 3.3 HF: 39.8 N: 44 Misfit: 3.6 HF: 39 4 Homestead 8 0 Anabar 8 0 500 Tanzania N: 16 Misfit: 4.1 HF: 37.6 N: 85 Misfit: 5.2 HF: 38.6 4 Baltic 1000 1500 N: 26 Misfit: 5.1 HF: 43.2 0 500 1000 1500 Temperature [°C] Figure 3.7 continued. Fits of preferred geotherm family (P = 0.74) to xenolith P –T estimates for selected cratonic localities. All inset misfit axes are the same as in 3.7. All xenolith P –T conditions use TBKN and PBKN thermobarometers unless noted. Open circles in Homestead are included in dashed analysis and excluded from solid. 35 Locality Kalahari Ouchita Slave Somerset Superior Homestead West Greenland Anabar Baltic Tanzania Reference Bell et al. [2003a] Grégoire et al. [2003] James et al. [2004] Saltzer et al. [2001] Simon et al. [2003] Dunn [2002] Rudnick and Nyblade [1999] and references therein Aulbach et al. [2007] MacKenzie and Canil [1999] Kopylova and Caro [2004] Kopylova et al. [1999] Russell et al. [2001] Rudnick and Nyblade [1999] and references therein Schmidberger and Francis [1999] (P estimated using method by MacGregor [1974]) Kjarsgaard and Peterson [1992] Zhao [1998] Rudnick and Nyblade [1999] and references therein Hearn [2004] Nielson et al. [2008] Sand et al. [2009] Bizzarro and Stevenson [2003] Hutchison and Frei [2008] Agashev et al. [2008] Roden et al. [1999] Roden et al. [2006] Rudnick and Nyblade [1999] and references therein Kukkonen and Peltonen [1999] Peltonen et al. [1999] Lee and Rudnick [1999] Rudnick et al. [1994] Rudnick and Nyblade [1999] and references therein Figure 3.7 continued. References to xenolith P –T data. 36 This result is consistent with heat flow collected in these and other Precambrian cratonic and shield regions [Nyblade, 1999; Mareschal and Jaupart, 2004; Roy and Rao, 2000; Roy et al., 2008; Alexandrino and Hamza, 2008]. Results using the invariant heat production are comparable. 3.7.3 Preferred Geotherm Family The constant basal heat flow model is excluded as a viable geotherm family on the basis of its inability to explain global variations in mantle xenolith P –T conditions. Elevation and xenolith P –T conditions, however, do not discriminate between the invariant and partition models. However, there are a number of factors that suggest a partitioning model is more likely. In general, heat production is higher heat production in younger terranes [Jaupart and Mareschal , 2003], making the constant heat production model unlikely. In the Canadian Shield, much of the variation in surface heat flow can be tied to regional differences in heat production [Mareschal and Jaupart, 2004]. The heat flow of most Precambrian regions is ∼40 mW/m2 as shown by the xenotherm fits above. There are, however, a number of shields and cratons with high heat flow yet little or no tectonic activity for >1 Ga. This suggests significant variations in surface heat production [Lewis et al., 2003; McLaren et al., 2005]. In light of these observations and my results, my preferred geotherm and heat production model is a partition model with ratio of 74:26 between the basal heat flow and upper crustal radiogenic heat flow, and an upper-crustal heat-producing thickness of 16 km. 3.7.4 Lithospheric Thickness and Sublithospheric Heat Flow Many geodynamic processes require an estimate of lithospheric thickness and/or the heat flow at the base of the lithosphere. Values for these parameters can be derived from my preferred geotherm models (Figure 3.8). Lithospheric thickness ranges from just under 200 km at 40 mW/m2 surface heat flow to ∼50 km at 90 mW/m2 . 37 Lithospheric Thickness [km] 250 200 150 100 50 (a) Sub-Lithospheric Heat Flow [mW/m2] 0 100 80 (b) 60 1:1 40 20 0 Sub-Lithospheric Surface Heat Flow 1.0 0.8 (c) F = 0.74 0.6 0.4 0.2 0.0 40 60 80 Surface Heat Flow [mW/m2] 100 Figure 3.8. Lithospheric thickness and heat loss. (a) Lithospheric thickness for partition model with coefficient F = 0.74, and Proton mantle composition. (b) Sublithospheric heat flow computed by subtracting lithospheric heat generation from surface heat flow. (c) Ratio of sublithospheric heat flow to surface heat flow. 38 Geotherms no longer intersect the adiabat for heat flow <34 mW/m2 . Sublithospheric heat flow is estimated by subtracting the estimated upper crustal heat production from the partition model and the contributions from the lower crust and mantle lithosphere. Minimum sublithospheric heat flow for the preferred geotherm family is 11 mW/m2 and increases to just under 65 mW/m2 for a 100 mW/m2 surface heat flow. The sublithospheric heat flow is nearly linear despite the large curvature in the lithospheric thickness because heat production within the mantle is very small in my models. Using surface heat flow models from xenotherms (Figure 3.7), one can estimate the range of sublithospheric heat flow into shields and cratons (14.5–20 mW/m2 ). The sublithospheric heat flow accounts for 39–47% of the total surface heat flow. Estimates from the Homestead xenoliths suggest a slightly higher ∼24 mW/m2 ; however, the relatively few samples at near-adiabatic temperatures make it difficult to obtain a reliable estimate. 3.8 Conclusions Radiogenic heat production is highly variable within the continental lithosphere and difficult to estimate from standard geophysical techniques. Using compositionally corrected elevations and xenolith thermobarometry, the uncertainties in upper crustal heat production can be substantially reduced. I define three types of heat production models: an invariant heat production (independent of surface heat flow); constant basal heat flow; and partitioning between upper crustal radiogenic heat production and basal heat flow. Xenolith P –T estimates suggest the constant basal heat flow models are not generally applicable. Likewise surface observations of heat production suggest upper crustal heat production can not be invariant. I propose a reference heat production model for the North American lithosphere with 26% of the surface heat flow resulting from upper-crustal heat production within a 16 km thick layer. Lower crustal and lithospheric mantle heat production are estimated at 0.4 and 0.02 µW/m3 . My model is calibrated using compositionally normalized elevation for 33 individual tectonic provinces and verified with xenolith 39 P –T suites from 10 separate cratons and shields. My preferred geotherm family estimates heat flow between 37 and 47 mW/m3 for these Precambrian xenoliths with a corresponding lithospheric thickness between 225 and 135 km, respectively. Predicted sublithospheric heat flow for these localities varies from 14.5–24 mW/m2 . CHAPTER 4 OCEANIC HEAT FLOW: IMPLICATIONS FOR GLOBAL HEAT LOSS 4.1 Abstract Heat flow measurements in seafloor younger than 55 Ma are systematically lower than model predictions for a conductively cooling lithosphere. Oceanic heat flow are typically collected in sedimented seafloor. These sedimented regions also tend to be the location of hydrothermal recharge, causing a systematic bias in measurement towards low heat flow at young ages. Hydrothermal circulation drops below detection as sedimentation and compaction sufficiently reduce permeability, causing temperatures to return to background conductive equilibrium. By filtering heat flow data to retain sites with sediment cover >325 m and located >85 km from the nearest seamount, the effect of hydrothermal circulation can be minimized. Additional adjustments for incomplete thermal rebound and sedimentation are estimated. Adjusted and filtered heat flow approaches the plate model estimate GDH1, although a deficit still persists at ages <25 Ma, possibly as a result of limitations of global filtering. An environmental analysis of heat flow co-located with seismic data suggests that heat flow at young ages is consistent with background estimates. A heat flow deficit due to hydrothermal circulation is also estimated, allowing for an estimate of the advective heat loss (7 ± 2 × 1012 W). Hydrothermal circulation in young seafloor out to ∼70 Ma accounts for ∼17% of global heat loss. 4.2 Introduction Since the first published heat flow measurements in the oceanic crust by Revelle and Maxwell [1952], over 13 000 determinations have been collected. Earliest estimates of global heat loss prior to the discovery of widespread hydrothermal circulation by 41 Lister [1972] included oceanic heat flow data on ridge flanks, arriving at a loss rate of ∼31 TW [Lee and Uyeda, 1965; Lee, 1970; Chapman and Pollack , 1975]. With the discovery of seafloor spreading, conductive cooling models were developed to explain the subsidence of oceanic lithosphere with age [Parker and Oldenburg, 1973; Davis and Lister , 1974; Crough, 1975; Parsons and Sclater , 1977; Stein and Stein, 1992]. These plate cooling models predict the background conductive heat loss through the lithosphere, highlighting a deficit in heat flow on young seafloor attributed to hydrothermal circulation [Lister , 1972]. Global heat loss models that attempt to account for the heat flow deficit by using modeled heat flow for young ages yield estimates of ∼40–44 TW [Williams and Von Herzen, 1974; Langseth and Anderson, 1979; Davies, 1980; Pollack et al., 1993]. The systematically low observed heat flow results from a spatial bias in measurements. Because the most commonly employed technique to estimate heat flow in the oceans requires sediment cover, the site of most recharge regions, the low heat flow values are preferentially captured. To estimate the true conductive heat flux, Sclater et al. [1976, 1980] attempt to remove the systematically low heat flow values by filtering sites using sediment thickness and basement morphology criteria. Their attempt focused on only a few localities and a limited dataset. I expand their analysis using similar filters to the entire oceanic heat flow dataset. In this study, I (1) illustrate that hydrothermal circulation is a major influence on upper crustal heat flow; (2) demonstrate that a systematic low in observed heat flow results from experimental design and spatial distribution of recharge and discharge zones; and (3) expand the filters developed by Sclater et al. [1976] to remove the influence of hydrothermal circulation on heat flow and reveal the background thermal regime. 4.3 Datasets Heat flow data used in this study are extracted from an updated global heat flow database (Chapter 2). Sites are assigned ages of the nearest pixel using the seafloor age model by Müller et al. [2008]. The heat flow data are prefiltered, 42 restricting heat flow values between 0 and 500 + qref (t) mW/m2 , where qref (t) is the heat flow computed by GDH1. Some previous versions of the global heat flow database use 0 mW/m2 to denote heat flow not calculated, and sites with negative heat flow are in regions where the bottom water temperatures are not stable. Therefore, I choose to exclude all of those data. Large Igneous Provinces (LIPs), which cover a small yet significant fraction of the seafloor and represent anomalous volcanism, are excluded from all oceanic data prior to any binning and filtering. The LIPs boundaries used in this study were initially defined by Coffin and Eldholm [1994]. (A 2004 digital version of LIP polygons is available from the UTIG Plates Project: http://www.ig.utexas.edu/research/projects/plates/). After pre-filtering, 13,501 heat flow determinations remain for global analysis (Figure 4.1). Some measured heat flow values are in excess of 1000 mW/m2 above estimated background. These extreme highs, most commonly on ridge flanks, are the result of very isolated and concentrated hydrothermal discharge associated with black smokers and mud volcanoes (e.g., Williams et al. [1979]; Becker and Von Herzen [1996]; Rona et al. [1996]; Eldholm et al. [1999]; Kaul et al. [2006]). Lateral variations near these discharge zones can be several hundred to 10 000+ mW/m2 in the space of a few meters. These highly anomalous regions can strongly affect the standard deviations when included. Median values (2 m.y. bins) are also affected because many measurements are typically taken near these anomalous regions. While the choice of 500 + qref mW/m2 is arbitrary, I compared the results with 1000 + qref mW/m2 and see little difference in the end result. Digital sediment thickness maps are used to estimate the sediment cover below heat flow sites (Figure 4.1). I interpolate Divins [2007] 5′ × 5′ and Laske and Masters [1997] 1◦ ×1◦ sediment thickness models to the same 2′ × 2′ grid as the age model. The lower resolution dataset is only used where the higher resolution is not available, specifically the northern Philippine Sea and Arctic Ocean. I use the seamount database by Wessel [2001] to compute the distance of heat flow sites to seamounts. The seamount database includes a height and radius fit to a truncated conical shape. Minimum height resolution is ∼1 km, but the number Heat flow sites: sediment cover ≥ 325 m all other sites Sediment Thickness [m] 20000 seamounts ≥ 85 km 8192 4096 2048 1024 784 576 400 256 144 64 16 0 43 Figure 4.1. Oceanic datasets used in this study. Basemap is 5′ × 5′ sediment thickness from Divins [2007] supplemented by 1◦ × 1◦ Laske and Masters [1997]. Seamounts (black dots) by Wessel [2001] with LIPs (grey regions) from UTIG Plates Project. Sea-floor age isochrons in 20 Ma intervals by Müller et al. [2008] with ridges identified in bold. Heat flow (red circles, included by preferred filter; white circles, excluded by preferred filter) is from an updated global database by (Chapter 2). 44 of seamounts by size is reliable above >2 km. The minimum distance from a heat flow site to a seamount is computed by determining the distance from the heat flow site to the seamount center less the radius. No attempt has been made to globally characterize basement exposures less than 1 km in height, which could potentially affect some sites. 4.4 Thermal Model of Sea-floor Spreading Before examining the effect of hydrothermal circulation on global heat loss estimates, I must first understand the background thermal regime. Continental thermal models use heat flow estimates made from temperature measurements that extend >100 m into the subsurface. The vast majority of oceanic heat flow measurements are made in the upper few meters and therefore are easily disturbed by hydrothermal circulation. Therefore, a proxy for the background thermal regime is required. Unlike heat flow, subsidence of seafloor is related to the integrated thermal state of the lithosphere and little affected by skin effects (i.e., hydrothermal circulation). Bathymetry therefore is an excellent constraint on oceanic cooling models (Figure 4.2b). Numerous cooling models have been developed [Parker and Oldenburg, 1973; Davis and Lister , 1974; Crough, 1975; Parsons and Sclater , 1977; Stein and Stein, 1992], including several recent models using temperature–pressure dependent thermophysical properties [McKenzie et al., 2005; Afonso et al., 2005, 2007]. The most common mathematical model used to describe ocean cooling is the plate. The initial thermal state of the plate is a uniform temperature with depth (Figure 4.2a). As the plate cools, the lithosphere grows in thickness and contracts, causing subsidence. In plate cooling models, the maximum plate thickness is fixed at depth by a basal temperature, which is held constant through time. The fixed basal temperature causes an asymptotic flattening of the predicted subsidence and heat flow curves (Figure 4.2b). While the cooling models mentioned above yield different plate thicknesses and mantle adiabatic temperatures based on choice of input parameters, the modeled heat flow and bathymetry are relatively similar. Subsidence of the seafloor provides a 45 0 2 0.5 5 25 50 50 t = 150 Ma 75 4 5 6 (a) 100 0 3 Bathymetry [km] Depth [km] 25 Sediment Corrected Bathymetry 500 1000 1500 Temperature [°C] (b) 0 50 100 Age [Ma] 150 Figure 4.2. Thermal isostasy of the oceanic lithosphere. (a) Oceanic geotherms in 25 Ma intervals. (b) Observed (circles) bathymetry in 2 m.y. age bins with 1-σ standard deviation and plate model bathymetry (grey line). Oceanic plate cooling model by Stein and Stein [1992], including a static 90 m reduction in bathymetry to fit the subsidence better, is shown in grey. Bathymetry data are isostatically corrected for sediment loading and exclude LIPs. reliable proxy for the integrated thermal state of the lithosphere. Since I are interested in the deficit in heat flow created by hydrothermal circulation, not the individual properties of plate cooling models, I have chosen to use the model by Stein and Stein [1992] because of its widespread use and good fit to the global bathymetry (Figure 4.2). 4.5 Observed Heat Flow Heat flow predicted from plate cooling models is high near the ridge and rapidly decreases with age (∝ t−0.5 ), asymptotically approaching a background heat flow at old ages (Figure 4.3a). Observed heat flow data for the oceans show a significant deficit at young ages relative to plate cooling estimates (Figure 4.3). The data also show considerable variability in median values at young ages. Vigorous hydrothermal circulation through the young oceanic crust is generally invoked as the explanation for this mismatch and variability [Lister , 1972; Williams and Von Herzen, 1974; Anderson et al., 1977; Sclater et al., 1980; Davis et al., 1992; Stein and Stein, 1992; Harris and Chapman, 2004]. 46 300 Heat Flow [mW/m2] Observed Heat Flow (a) 200 100 No. observations 0 400 13501 sites (b) 200 0 0 50 100 Age [Ma] 150 Figure 4.3. Observed oceanic heat flow. (a) Observed heat flow data in 2 m.y. bins with 1-σ standard deviation. shown in grey. Observed heat flow excludes data located on LIPs, with values ≤0, and > qGDH1 + 500 mW/m2 . Hatched regions are ‘reliable’ heat flow from Sclater et al. [1976, 1980]. (b) Number of heat flow sites in each bin. Hydrothermal circulation in the oceanic crust results from buoyancy-driven flow, unlike the pressure-driven flow in continental groundwater systems. Figure 4.4 illustrates how oceanic hydrothermal systems can affect heat flow. The seafloor-water interface is at constant temperature, Ts , causing isotherms beneath high bathymetry to deflect toward the surface. Cold dense seawater seeps into the oceanic lithosphere, generally through sedimented basins and exposed bathymetric lows. The water is heated as it travels through the lithosphere and then is discharged as hot buoyant seawater beneath exposed basement highs (Figure 4.4a). In this type of environment, heat flow is low in the sediments at intermediate distance surrounding exposed basement and high on exposed volcanic highs. A high may even be observed in the sediments immediately adjacent to the basement high [Stein and Stein, 1997]. 47 Ts Tb (a) Thin/No Cover qobs qref ~ 0.1-0.2 q σobs qref ~ 0.7-1 qref (b) Intermediate Extensive Cover Ts Tb qobs qref ~ 0.5 q σobs qref ~ 0.6-0.8 qref (c) Thick Extensive Cover Ts qobs qref ~ 1 q σobs qref ~ 0.4 qref Tb Figure 4.4. Effect of sediment cover on hydrothermal circulation and heat flow through the oceanic lithosphere. Top plots in each panel illustrate observed heat flow (qobs , heavy line) and reference heat flow (qref , thin line). Standard deviation of observed heat flow σobs . Bottom plots illustrate sediment cover (grey region), isotherms (thin lines), and hydrothermal circulation (solid and dashed lines). Sea-floor water interface is an isotherm, Ts , and temperatures at the sediment–basement interface, Tb , is nearly isothermal (panel (c) only). While some temperature measurements used to estimate heat flow in oceanic lithosphere are collected in crystalline basement (e.g., DSDP), >95%, are made in the upper 3–9 m of sediment owing to the types of probes typically employed. The combination of cold recharge in sediment and limitations in the measurement environment leads to a systematic bias toward anomalously low heat flow (Figure 4.3). Beneath buried volcanic basement, hydrothermal circulation is sufficiently vigorous that temperatures at the sediment–basalt interface are relatively constant, Tb . Isotherms in the sediments are deflected toward the surface above these volcanic highs, setting up the necessary density gradients for circulation (Figure 4.4b). Because both recharge and discharge zones are sediment-covered, it is possible to measure heat flow everywhere around these anomalies. Low heat flow exists in regions of thick sediment and high in regions of thin sediment. In a well sampled study, the integrated heat flow 48 can reach the background, qref . In general, however, heat flow in this environment is not well sampled, average heat flow ∼50% of the background, with high variability. As additional sediment accumulates, sediment compaction reduces the hydraulic conductivity. Eventually sediment permeability decreases fluid flow below levels detectable by heat flow (Figure 4.4c). To determine the global conductive heat flow through the lithosphere, I can exploit regions where hydrothermal circulation is reduced below detection by filtering for regions with sufficient sediment cover and far enough to fall outside the influence of basement highs and exposures. 4.6 Global Filtering—Sediments and Seamounts In order to filter out the effect of hydrothermal circulation and determine the background conductive heat flow, I need to develop criteria that can be used to find environments where hydrothermal circulation is minimal. Sclater et al. [1976, 1980] proposed three filters to oceanic heat flow data to limit the effect of hydrothermal circulation and estimate the conductive heat flow through the oceanic lithosphere. They restrict high quality heat flow determinations to regions with flat or rolling hill morphologies, thick sediment cover (>200 m) and distances >18 km from basement highs. These criteria can be applied to individual sites, requiring local bathymetric information and seismic lines that are frequently collected prior to heat flow surveys. Average heat flow values can be improved using this sort of analysis as seen by the hatched boxes in Figure 4.3a. The volume of information and format (many of the records are not digital or have been lost) make it impractical make these assessments on 13 000+ oceanic heat flow sites. Therefore, it is desirable to develop a method that can be easily automated from global sediment cover and bathymetry models in order to extract the lithospheric background heat flow. In this study, two filters are applied to the heat flow data: a minimum sediment thickness filter with a range of 0 to 1000 m; and a minimum distance to seamounts, ranging from 0 to 100 km. Figure 4.5 shows heat flow data collected in 2 m.y. age bins with sediment thickness restricted to 50 m intervals from 0 to 900 m. Because data are unevenly distributed in 49 200 0-50 m 50-100 m 100-150 m 150-200 m 200-250 m 250-300 m 300-350 m 350-400 m 400-450 m 450-500 m 500-550 m 550-600 m 600-650 m 650-700 m 700-750 m 750-800 m 800-850 m 850-900 m 150 100 50 0 150 100 50 0 150 Heat Flow [mW/m2] 100 50 0 150 100 50 0 150 100 50 0 150 100 50 0 0 50 100 150 0 50 100 Age [Ma] 150 0 50 100 150 Figure 4.5. Median heat flow versus age in 2 m.y. bins divided into groups with 50 m increments of sediment thickness. Error bars represent 1-σ. Data points without error bars represent a single measurement. Cooling model (grey) is heat flow from GDH1. 50 age and sediment thickness, many bin averages are very noisy within a given sediment interval. Sediment intervals between 50–100 and 100–150 m exhibit a very low heat flow when compared to GDH1. In fact, median values at young ages are rarely elevated much above the asymptotic limit. At 150–200 m there is increased scatter, but little trend is evident in heat flow at young ages. As the sediment interval is increased, the heat flow at younger ages generally increases and in some bins closely approaches or exceeds values of GDH1 (sediment thickness >450 m). At higher sediment intervals, >550 m, fewer and fewer heat flow medians meet or exceed GDH1. To determine the effectiveness of these filters I use two metrics, each highlighting different aspects of binned heat flow results. 4.6.1 4.6.1.1 Filter Metrics Statistical Correlation The heat flow–age pattern resulting from cooling of a half-space can be described to high accuracy as C/age−1/2 , where C is empirically determined (∼500 Ma1/2 mW/m2 ). This mathematical model also approximates the plate cooling model reasonably well. Without biasing filtered heat flow patterns towards any particular cooling model, I estimate the improvement of various filter combinations by computing the correlation coefficient of a linearized heat flow–age relationship. While the correlation between heat flow and age−1/2 is a linear space, the distance between 2 m.y. age bins is large at young ages and decreases rapidly, thereby causing any distribution to be strongly weighted by the youngest ages as older ages are progressively more closely spaced. By inverting the equation and testing the linear correlation between age and heat flow−2 , the time spacing is uniform, which gives equal weight to the entire range of ages. I apply the correlation r= PN i=1 (ti γi ) − N t̄γ̄ (N − 1)σt σγ (4.1) to filtered data, where ti and γi are the age and heat flow−2 of 2 m.y. binned data, respectively. The bar denotes mean values of all bins, σ denotes standard deviation, 51 and N is the number of bins. To avoid numerical instability, 1000 is added to each heat flow datum and zero age is set to 0.1 Ma, neither of which affect the results significantly. 4.6.1.2 Variability To complement the correlations, I seek a single value to quantify improvement in the standard deviations of age bins. The standard deviation is often higher at high heat flow. Therefore, standard deviation is normalized by the median before making a direct comparison of binned results. I define the relative variability in heat flow for a given filter as, variability = N σi 1 X , N i=1 q̄i (4.2) where q̄i and σi are the average and standard deviation of each age bin, respectively. Because hydrothermal circulation causes an increase in the spatial standard deviation in addition to a reduction in the overall average heat flow, I expect filtering to reduce the standard deviation within each bin. 4.6.2 Filtered Results Correlation coefficients between binned age and heat flow−2 range from 0.33 to 0.93 (Figure 4.6). Unfiltered data are poorly correlated with age (0.5). The highest correlation, 0.93, occurs using filters with a minimum sediment thickness of 475 m and a minimum distance to seamounts of 85 km (‘+’ in top panel of Figure 4.6). The lowest correlations occur for sediment filters below ∼100 m. The poor correlation may result from a reduction in hydraulic conductivity as sediments compact. Correlations, including sites near seamounts, remain low until the sediment thickness filter reaches ∼400 m. An increase in correlation coefficient occurs very rapidly as the minimum sediment thickness increases from 300 to 400 m when filtering distance to seamounts is small (Figure 4.5). A high plateau in correlation coefficient occurs between ∼400–700 m minimum sediment thickness suggesting all filter results within this range are relatively similar. Seamounts also appear to have very little influence on the median 1.0 Correlation 0-180 Ma 80 0.8 60 40 0.6 20 0 0 200 400 600 800 1000 Minimum Sediment Thickness [m] Correlation Coeffiient 100 0.4 0.8 100 Variablity 0-56 Ma 80 0.7 60 0.6 40 0.5 20 0 0 200 400 600 800 1000 Minimum Sediment Thickness [m] Normalized Variability Minimum Distance from Seamount [km] Minimum Distance from Seamount [km] 52 0.4 Figure 4.6. Metrics for improvement in globally filtered datasets as a function of minimum sediment thickness and minimum distance to seamounts. (top) Correlation of binned heat flow−2 with age. The ‘+’ indicates the filter combination with the maximum correlation coefficient. (bottom) Normalized variability. 53 values for sediment thicknesses >400 m. Even with coarse sediment thickness resolution, above 400 m it seems likely that local highs in the igneous crust are well sedimented, preventing fluid flow and contributing to the high correlations. Seamount distance also shows a significant influence on correlation coefficient. A slight decrease in correlation coefficient occurs as the distance from seamounts increases out to ∼30 km. By ∼40–50 km, a large increase in the correlation coefficient occurs, possibly resulting from the maximum lateral extent of most oceanic hydrothermal systems, well beyond the 18 km suggested by Sclater et al. [1976]. This 40–50 km scale is consistent with observations of the hydrothermal systems southwest of the Nicoya Peninsula on the Cocos plate [Hutnak et al., 2008]. While distance from seamounts continues to influence the correlation coefficient beyond 85 km minimum distance, the proximity to a seamount may also imply rough sediment–basement morphology and/or a higher likelihood that smaller basement penetrators below the detection limits of Wessel [2001] are near the heat flow site. Basement highs that do not penetrate the sediment cover have a smaller effect on the median heat flow. Therefore, far from seamounts, thinner sediment cover is necessary to retard flow below noise levels contributing to an increase in the correlations. Correlations begin to fall off beyond a minimum sediment thickness >700 m as a result of few data left to compute values of individual bins, making the binned results more susceptible to individual outliers. In contrast to the correlation coefficient, variability does not exhibit clearly defined peaks or troughs that can be used to pick a preferred filter. Variability is highest for the unfiltered dataset and decreases as both minimum distance to seamounts and minimum sediment thickness are increased. For 0–180 Ma ages the overall decrease in variability is from 0.54 to 0.44. A similar pattern in variability exists when the age range is restricted, but exhibits a much larger range. The variability when restricted to 0–56 Ma ranges from 0.42–0.78 (Figure 4.6 bottom panel). The higher variabilities for age restricted data result from the exclusion of bins at older ages with relatively constant and ‘low’ standard deviation. 54 4.6.3 Preferred Filters and GDH1 Misfit The preferred set of filters is chosen to minimize the effect of vigorous hydrothermal circulation on heat flow and maximize the number of data points used in the analysis. I choose filter values of 325 m for minimum sediment thickness and 85 km as the distance to the nearest seamount. The filtered dataset has a correlation coefficient of 0.89 and retains 4929 sites, or a little less than 40% of the initial pre-filtered dataset (Figure 4.7c). The preferred filter constraints improve agreement between modeled and measured heat flow as a function of seafloor age (Figures 4.7a. The large increase in heat flow from thin to thick sediment cover confirms that the effects of hydrothermal circulation can be eliminated by thick and extensive sediment cover. This assessment is further confirmed by the reduction in variability (Figure 4.7e and f ). By filtering heat flow using this simple scheme, the effect of hydrothermal circulation on average heat flow is substantially reduced. However, a heat flow deficit compared to lithospheric cooling models persists at ages <55 Ma. I explore some of the causes of this persistent deficit in Section 4.8. 4.6.4 Cautions Using Global Binned Data A site by site analysis similar to Sclater et al. [1976] is preferable to global models like the one performed here. However, a site by site analysis is impractical because of a lack of the data required to assess many, if not most, sites. A global analysis can be performed with global models of sediment thickness with some success, but several issues with global sampling and resolution must be acknowledged. Many heat flow sites collected in the previous version of the global heat flow database [Pollack et al., 1993] are isolated and individual determinations. These provide a good spatial distribution for a global analysis but a single site can be easily affected by an anomalous measurement. Most data collected since the previous database update are located in targeted surveys with tens to hundreds of measurements in clusters, lines, and grids. While providing better regional heat flow, they provide poor spatial coverage globally. The recent data are also commonly located 55 300 sed. thickness ≥ 325 m seamounts ≥ 85 km Heat Flow [mW/m2] 250 sed. thickness < 325 m seamounts - no filter (a) (b) 200 150 100 50 No. observations 0 400 unfiltered filtered, N = 4929 unfiltered excluded, N = 7370 (d) 200 0 (e) 1.2 SD/HF (c) (f ) 0.8 0.4 0 0 50 100 Age [Ma] 150 0 50 100 Age [Ma] 150 Figure 4.7. Global filtering results. (left) Heat flow locations filtered for sites with ≥325 m sediment cover and ≥85 km to the nearest seamount. (right) Sites with <325 m of sediment cover and no seamount distance constraint. (a) and (b) heat flow versus age in 2 m.y. bins (open circles contain <10 sites). One standard deviation in grey. (grey squares) Environmentally analyzed data in (a) and (g) (Table 4.1). (c) and (d) number of observations within each age bin, initial dataset in grey. (e) and (f ) normalized standard deviation (variability). (g) and (h) ratio of binned heat flow to GDH1. Black lines on (e − h) represent a 5 m.y. moving average. Grey lines from plot opposite for direct comparison. 56 1.4 Data/GDH1 (g) (h) 1.0 0.6 0.2 0 50 100 Age [Ma] 150 0 50 100 Age [Ma] 150 Figure 4.7 continued. (g) and (h) ratio of binned heat flow to GDH1. Black lines on represent a 5 m.y. moving average. Grey lines from plot opposite for direct comparison. 57 in regions which exhibit ‘interesting’ heat flow variations/anomalies, rather than conductive background values. When few data are located in a given age bin, the spatially clustered data may dominate the bin average, especially if the filtered sediment thickness is great. Uncertainties in sediment thickness are also a concern when using global datasets. At 5 arc minute resolution, a pixel covers (9.3 km)2 at the equator. For the 1◦ dataset, equatorial resolution is (111.2 km)2 . Because the sediment thickness represents an average for the region covered by the pixel, basement highs in either case may short-circuit otherwise impermeable sediment cover allowing hydrothermal circulation except where sediment cover is very thick. Numerical models of hydrothermal circulation using reasonable parameters for fluid driving forces and sediment permeability suggest that 150–200 m of sediment cover is typically required to reduce fluid flow to a level with negligible influence on heat flow [Grevemeyer and Bartetzko, 2004], although the thickness required varies based on sediment type, permeability, and the magnitude of the driving force. However, as the mean sediment thickness increases, fewer volcanic highs approach the surface and their influence on heat flow is diminished. This is consistent with improvements in filtered results with sediment thicknesses greater than 150 m. 4.7 Environmental Analysis To overcome some of the limitations in a global analysis, I choose four regions with adequate site environment data to reveal hydrothermal circulation patterns and hence to estimate the background lithospheric thermal regime. The Juan de Fuca ridge flank is by far the most extensively studied ridge flank with 1068 heat flow measurements co-located with seismic data [Davis et al., 1992, 1997]. Very few of these data are completely free of hydrothermal circulation but, properly selected, the data can still be used to estimate conductive heat flow (Table 4.1). An example of heat flow data from the Juan de Fuca flank is shown in Figure 4.8. Temperatures at the sediment-basement interface measured from nine Deep Sea Drilling Project (DSDP) boreholes with >100 m sediment cover are relatively 58 Table 4.1. Heat flow from environmental analysis. Locality Age Ma 1.4 3.5 6.0 12.9 16.7 18.2 19.8 21.4 22.8 26.6 Juan de Fuca Juan de Fuca Costa Rica rift Gulf of Aden Gulf of Aden Cocos plate Cocos plate Gulf of Aden Gulf of Aden Gulf of Aden qs ± σq a mW-m−2 361 ± 92 253 ± 61 216 ± 20 139 ± 9 118 ± 6 117 ± 20 114 ± 14 125 ± 8 110 ± 10 104 ± 8 N 98 118 33 6 3 37 24 6 17 3 qobs /qs b 0.976 0.934 0.965 0.85 0.85 0.967 0.972 0.85 0.85 0.85 a mean and standard deviation of sedimentation corrected heat flow. b sedimentation correction using a thermal diffusivity of 0.3×10−6 mm2 /s. Sediment corrections for Gulf of Aden are average of Lucazeau et al. [2008] and Lucazeau et al. [2009]. 0 1 Age [Ma] 2 3 4 5 Observed GDH1 Heat Flow [mW/m2] 600 400 200 (a) 0 80 40 (b) Sediment 4.0 4.5 (c) 1026 1027 3.5 1032 TWTT [sec] 3.0 1029 0 1023 1024 1025 1030/31 1028 Basement Temperature [oC] 120 Basement 0 40 80 120 Distance from ridge axis [km] Figure 4.8. Heat flow case study from Juan de Fuca flank. (a) heat flow, (b) basement temperature, (c) sediment and basement interfaces as determined from seismic reflection. Figure after Harris and Chapman [2004]. 59 constant despite significant variability in basement topography and sediment thickness. The heat flow pattern over the well-sedimented seafloor with variable basement topography is related to vigorous hydrothermal circulation within the basement and, while variable, the average should be close to conductive background [Davis et al., 1999]. Additional profiles from the Juan de Fuca flank above smooth basement topography exhibit a sinusoidal heat flow pattern, suggesting cellular convection within the basement beneath these sites. Because several cycles are captured, the average should represent the background heat flow [Davis et al., 1996]. A total 153 sites used from the Juan de Fuca flank allow us to estimate the conductive lithospheric heat flow in seafloor between 1 and 4 Ma. There are 48 detailed heat flow sites on the Costa Rica rift flank for which there are high resolution seismic data [Davis et al., 2004]. The sites are filtered to exclude high heat flow anomalies associated with basement highs that do not penetrate the sediment cover. The resulting heat flow is near GDH1 estimates for 6 Ma seafloor. Modeled isotherms in the region strongly up-warp as a result of variations in the basement topography and vigorous hydrothermal circulation in the upper part of the igneous crust, not hydrothermal circulation through the sediments. Thus the high anomaly is excluded from this analysis. The Gulf of Aden is a well sedimented rift and relatively young oceanic spreading center (<30 Ma). I analyzed 35 sites on the ridge flank [Cochran, 1981; Lucazeau et al., 2008, 2009]. All 35 sites have >100 m of sediment and are located above relatively smooth basement. Heat flow values are well behaved and have small standard deviations (Table 4.1). A survey of 327 sites co-located with seismic estimates of crustal thickness are used to estimate heat flow on the Cocos plate [Hutnak et al., 2008]. The authors identified regionally low basement temperatures due to fluid extraction of heat discharged through several seamounts. The affected hydrothermal zone extends several tens of kilometers beyond each exposed seamount. On the south end of the study region where there are no basement exposures, heat flow is higher and much closer to background predictions. By excluding data within the identified hydrothermally 60 affected zone and restricting to sites >100 m, median heat flow fits the GDH1 plate cooling model to within error (Figure 4.9). Data from these site-specific analyses result in heat flow values similar to the estimated conductive heat flow. Standard deviations for these data are <25% of the observed mean, significantly better than variability in global filtered results (Figure 4.7). The Juan de Fuca flank, Costa Rica Rift, Gulf of Aden, and Cocos data all show heat flow values above the globally filtered data, suggesting hydrothermally disturbed data are not completely filtered from the global analysis and/or additional corrections are necessary to account for the lower heat flow. 300 Environmental Analysis Global Filters: ≥ 10 determinations < 10 determinations Heat Flow [mW/m2] 250 200 150 100 50 (a) 0 1.4 Data/GDH1 (b) 1.0 0.6 0.2 0 50 100 Age [Ma] 150 Figure 4.9. Filtered heat flow adjusted for sedimentation and thermal rebound. (a) adjusted and filtered heat flow (black circles). An environmental analysis is performed on the grey squares (Table 4.1). (b) fraction of heat flow to GDH1. Lines represent a moving average of 5 m.y. (black) adjusted and filtered, (grey) filtered only. 61 4.8 Corrections to Global Data The systematic low at young ages between filtered heat flow and plate cooling models raises the question of whether a heat flow deficit still exists, or if plate models should be revised. The heat flow at 10 Ma is 60% of GDH1, and increases to 80% by 20 Ma but does not reach the predicted background heat flow until ∼55 Ma (Figure 4.7g). 4.8.1 Sedimentation Sediments are deposited at the surface temperature, depressing the thermal gradient and thus the heat flow. The effect of sedimentation on heat flow is near zero initially, but grows with time (Figure 4.10a). The estimated heat flow with age assuming constant sedimentation rate is shown in Figure 4.10b. Sedimentation cannot cause the observed persistent deficit at young ages unless the sedimentation rates are systematic and extremely high (>500 m/m.y.). The sediment correction can be estimated by dividing the observed heat flow by the fraction computed in Figure 4.10a. More than 80% of the seafloor has a sedimentation rate less than 50 m/m.y. and therefore at most locations there is a small or negligible sedimentation effect on heat flow. Sedimentation rates >100 m/m.y. occur only in regions near continents with high river discharge such as the northeast Pacific and Bay of Bengal. 4.8.2 Thermal Rebound In regions where sedimentation is not so rapid that hydrothermal circulation is quickly cut off, a large quantity of heat can be extracted from the lithosphere. Once hydrothermal circulation has ceased, heat flow will return to background given sufficient time. A delay in the return to background could cause a persistent low heat flow to be observed. Hutnak and Fisher [2007] explored the effect of sedimentation and thermal rebound on oceanic heat flow. Their model results show a rebound of 90% in about 1 Ma for 200 m of sediment cover or less (Figure 4.11). The rebound fraction shown in Figure 4.11 is estimated from an empirical relationship that fits the Hutnak and Fisher [2007] models for 500, 1000 and 2000 m very well. The empirical 62 1.0 10 (a) q/qref 0.8 100 0.6 1000 m/m.y. 0.4 0.2 mm2/s 0.5 mm2/s 0.2 0 -4 10 10-2 100 Time [m.y.] 102 200 Heat Flow [mW/m2] (b) 0.2 mm2/s 0.5 mm2/s 150 100 0 m/m.y. 50 10 0 100 1000 0 50 100 Age [Ma] 150 Figure 4.10. Estimated heat flow response to sedimentation. (a) ratio of heat flow with sedimentation to heat flow without. Curves are computed using Von Herzen and Uyeda [1963] with sedimentation rates of 10, 100, and 1000 m/m.y. with thermal diffusivities of 0.2 and 0.5 mm2 /s, which approximate effective diffusivity with thin and thick sediment cover. (b) heat flow as a function of seafloor age with sedimentation models in (a). 63 0.8 0.6 200 m 500 100 m 0 200 m 0m Fraction Rebound, f 1.0 0.4 0.2 0 10-4 10-2 100 Recovery Time, tr [m.y.] 102 Figure 4.11. Estimated fraction of thermal rebound as a function of sediment cover and time since cessation of hydrothermal circulation. estimate for 250 m is equivalent to the 200 m model reported Hutnak and Fisher [2007], but for my purposes will be sufficient. If I assume a constant sedimentation rate, the duration thermal rebound, tr , is given by ! hc , (4.3) tr = tp 1 − hp where tp is the age of the crust, and hp and hc are the present-day sediment thickness and thickness at cessation of hydrothermal circulation. I estimate the fully recovered heat flow, qf r from the partially recovered (observed) heat flow, qpr , and the depressed heat flow from hydrothermal circulation, qh , by qf r = qh + qpr − qh , f (4.4) where f is the fractional recovery estimated from Hutnak and Fisher [2007]. The rebound pattern I would expect on well-sedimented seafloor has a larger deficit at young ages because less time has elapsed before measurement and therefore less time to return to equilibrium. 4.8.3 Adjustments to Filtered Heat Flow I estimate the sedimentation rate by dividing the present day sediment thickness by the age of the igneous crust. While this estimate is simplistic, it is beyond the scope of this study to model the sedimentation history of the entire ocean. I then 64 compute the sedimentation effect following Von Herzen and Uyeda [1963] using a thermal diffusivity of 0.3 mm2 /s (Figure 4.10). To estimate the hydrothermal rebound, I assume that hydrothermal circulation falls below measurable levels for sites with >325 m of sediment (my preferred filter). Because it is difficult to know a priori the hydrothermally disturbed heat flow, I estimate qh from a smoothed estimate of the average deficit with age (Figure 4.7b). Since the lithosphere continues to cool following cessation of hydrothermal circulation, I choose the deficit at the present seafloor age for simplicity. Thus qr calculated in this manner is likely an underestimate of the true rebound. To avoid issues with which correction to apply first (order matters), I estimate the adjustments for incomplete rebound and sedimentation separately and add them to the filtered heat flow (Figure 4.9). The sedimentation adjustment increases the heat flow for nearly all ages. Heat flow observations within individual age bins at crust older than 100 Ma show no discernible trend in heat flow as a function of sediment thickness, implying that my sedimentation adjustment overestimates the effect of sedimentation. The adjustment for incomplete rebound is negligible for ages >70 Ma, but significantly reduces the deficit at young ages. The combination of the two adjustments reduces the underprediction of heat flow at 10 Ma from 0.6 to 0.7 of GDH1 model estimates. By 30 Ma, the deficit is reduced to near zero, indicating GDH1 heat flow is a good predictor of background conductive heat flow through the oceanic lithosphere. To estimate the remaining deficit from incomplete thermal rebound and sedimentation more accurately, one requires a thermal model specific to the history of each site. The simplistic analysis performed above suggests a persistent deficit may be small relative to the estimated background conductive heat flow model GDH1. 4.9 Global Heat Loss The current power output of hydrothermal circulation is estimated by the cumulative heat flow deficit in each age bin multiplied by the respective area over which hydrothermal circulation occurs, 65 P = N X (qi◦ − qi )Ai , (4.5) i where qi◦ and qi are the reference and observed heat flow between isochrons ti and ti−1 with associated area, Ai . A Monte Carlo analysis is performed to determine the advective heat loss using GDH1 heat flow as the reference. The data excluded by my preferred filter out to 73 Ma is used as an estimate of the hydrothermal deficit (Figure 4.7b). For 106 realizations, the estimated total power mined by hydrothermal circulation is 6.9±1.8 TW (Figure 4.12). Hydrothermal power accounts for ∼17% of the Earth’s total heat loss (44 TW [Pollack et al., 1993]). My advective power loss is similar to the estimate of power output obtained from chemical fluxes [Elderfield and Schultz , 1996], but lower than most previous estimates based on heat flow [Williams and Von Herzen, 1974; Sclater et al., 1980; Stein and Stein, 1994] with the exception of Wolery and Sleep [1976]. The previous higher estimates result from the use of total seafloor area between isochrons rather than limiting to the area over which hydrothermal circulation occurs. This restriction reduces the estimated area of hydrothermally affected seafloor to regions with sediment thickness <325 m ultimately resulting in my lower estimated hydrothermal heat transport. # Realizations 2.0 × 104 1.6 6.9 ± 1.8 TW 1.2 0.8 0.4 0 0 2 4 6 8 10 12 Advective Heat Loss [TW] 14 Figure 4.12. Estimated global advective power loss from Monte Carlo analysis with 106 realizations. 66 4.10 Conclusions My results provide an updated and improved analysis oceanic heat flow, specifically estimating the extent and magnitude of hydrothermal circulation within the crust. My global analysis highlights a systematic bias toward low heat flow by measuring temperatures predominantly in sedimented seafloor. Heat flow measurements on the seafloor are rarely used for modeling the thermal evolution of the oceanic lithosphere because they are frequently influenced by nearsurface hydrothermal circulation. A systematic low bias results from collecting an overwhelming majority of heat flow estimates in sedimented regions of the seafloor where hydrothermal recharge commonly occurs. The deficit from hydrothermal circulation is observed up to ∼65 Ma and causes an increase in the measured variability. I show that by simple filtering and adjustments for additional physical effects, binned heat flow approaches the plate model estimate from GDH1 [Stein and Stein, 1992]. Thus, heat flow can be used in addition to bathymetry to constrain cooling models of the oceanic lithosphere for nearly all seafloor ages. Filtering heat flow for localities with a minimum sediment cover of 325 m and minimum distance to seamounts of 85 km, excludes most of the data perturbed by hydrothermal circulation. Filtering also reduces the variability in heat flow from 0.6->1 to typical background variability (0.3-0.5). A deficit in heat flow at young ages persists despite filtering, but is likely due to a combination of incomplete thermal rebound following cessation of significant hydrothermal circulation and depression of the thermal gradient by high sedimentation rates. An adjustment for these effects removes the deficit for ages >25 Ma and reduces the deficit below 25 Ma. However, these adjustments are crude and may be improved by incorporating more site specific history to each observation. Adjusted heat flow approaches the estimated reference model GDH1. Thus carefully filtered and corrected heat flow data can be used as a constraint on cooling models of the oceanic lithosphere. Detailed analysis of heat flow and seismic estimates of sediment thickness from the Juan de Fuca plate, Costa Rica rift, Gulf of Aden and Cocos Plate, allow for very 67 high confidence estimates of conductive lithospheric heat flow. Heat flow in these regions are consistent with the GDH1. Total power loss from hydrothermal circulation is ∼7±2 TW, consistent with an estimate from chemical fluxes [Elderfield and Schultz , 1996], but considerably lower than previous estimates derived from heat flow studies (10-11 TW) [Williams and Von Herzen, 1974; Sclater et al., 1980; Stein and Stein, 1994]. My model is an improvement over previous heat flow estimates of advective heat loss because I obtained a better estimate of hydrothermally affected heat flow and restricted the integration to regions where hydrothermal circulation is likely. CHAPTER 5 PLATE COOLING MODELS FOR THE OCEANIC LITHOSPHERE: ARE COMPLEXITIES NECESSARY? 5.1 Abstract Plate models developed over the past five years have been growing in numerical complexity. It is difficult to compare the importance of each complexity since each study uses a different set of observational constraints. I develop a standard sediment-corrected bathymetry–age and heat flow–age dataset cleaned of hydrothermal influence. This standard dataset is used to test plate cooling models. The use of effective expansivity rather than volumetric expansivity exerts the largest influence on modeled subsidence. Crustal thickness has the greatest influence on heat flow and subsidence at young ages. Inclusion of radiogenic heat production estimated from chemical models of oceanic lithosphere produces a negligible effect on cooling models. Using a 7 km thick crust consistent with seismological observations, the minimum misfit plate cooling model has a 90 km maximum plate thickness and potential temperature of 1425◦ C. 5.2 Introduction One of the triumphs of plate tectonic theory is the ability to predict oceanic bathymetry as a function of crustal age. As hot lithosphere created at the mid-ocean ridge is transported towards the abyssal plain, it cools, contracts and subsides. This pattern can be easily modeled using simple 1-D transient cooling models with fixed surface and basal boundary conditions, also know as a plate cooling model [McKenzie, 1967]. 69 Early plate cooling models assume constant thermophysical properties [McKenzie, 1967; Sclater and Francheteau, 1970; Parker and Oldenburg, 1973; Davis and Lister , 1974; Crough, 1975; Parsons and Sclater , 1977; Sclater et al., 1980, 1981; Stein and Stein, 1992]. More recently, modeling efforts are increasingly complex and incorporate a number of additional effects including temperature- and/or pressure-dependent thermophysical properties [McKenzie et al., 2005], polymineralic composition [Afonso et al., 2005], asthenospheric melt fraction [Afonso et al., 2007], etc. Each of these models uses a different estimate for average bathymetry. For example, the Stein and Stein [1992] (GDH1) model uses bathymetry from the north Pacific and northwest Atlantic, far from seamounts and with low sediment cover, whereas the model by McKenzie et al. [2005] uses bathymetry only from the north Pacific. Heat flow data used to constrain for each of these models also vary, but are in general only used to fix the asymptotic behavior. Because many of thee models are calibrated against different bathymetry and/or heat flow datasets, it is difficult to compare models directly. Additionally, the model sensitivity to variations in parameter complexities are rarely discussed. In this study, I propose a standardized dataset for global bathymetry and heat flow that I use to investigate the importance of complexities beyond the simple constant property models. 5.3 Theoretical Formulation The equation for one-dimensional transient heat conduction with sources is given by " # ∂T ∂ ∂ k + A, [ρCP T ] = ∂t ∂z ∂z (5.1) where T is temperature, t is time, and z is depth. Thermophysical properties CP , ρ, k, and A represent heat capacity, density, thermal conductivity, and heat production, respectively. If thermophysical properties are temperature-dependent, CP and ρ are not constant in time and k is not constant in depth and Equation 5.1 expands to ∂T ∂k ∂T ∂2T ∂ [ρCP ] T + ρCP = + k 2 + A. ∂t ∂t ∂z ∂z ∂z (5.2) 70 The recent model by McKenzie et al. [2005] includes time-dependent variations of ρ and CP whereas Afonso et al. [2005, 2007] ignore them. Both studies assume negligible heat sources. I numerically solve Equation 5.2 using the Crank-Nicholson method with a central difference formulation. I initially ignore the time derivatives for CP and ρ and solve the system using tridiagonal elimination. The solution is then perturbed by a time dependent term including both CP and ρ [McKenzie et al., 2005]. Perturbations are iteratively applied until the solution converges. (The formulation given by Equation 11 of McKenzie et al. appears to be incorrect despite the correctness of their results. See section 5.7 for a discussion.) Surface heat flow is estimated from modeled temperatures from the first two nodes and the harmonic mean of conductivity at the top and bottom of the first layer, q= 2k0 k1 k0 + k1 ! T1 − T0 . z1 − z0 (5.3) Subsidence is computed from the change in density relative to a reference density column by s= N X ρobs,i − ρref,i i=0 ρref,N − ρw (5.4) where ρobs,i and ρref,i are the observed and reference density for the i-th node. 5.3.1 5.3.1.1 Thermophysical Properties Thermal Expansivity, αV and Density, ρ The temperature-dependent volumetric expansivity, αV , is estimated by αV (T ) = a0 + a1 T + a2 T −2 , (5.5) where ai are empirical constants [Fei , 1995]. The temperature-dependent density is then be easily solved by: ◦ " ρ(T ) = ρ exp − Z T Tref # αV (T )dT , (5.6) where ρ◦ is the density at reference temperature and pressure (0 GPa and 298 K), and Tref is the reference temperature. 71 The pressure- and temperature-dependent thermal expansion is computed by ρ(P ) αV (P, T ) = αV (T ) ρ0 !−δT (ρ0 /ρ(P )) , (5.7) where ρ0 and ρ(P ) are the densities at 0 GPa and at the pressure P , and δT is the Anderson-Grünison parameter [Afonso et al., 2005]. Using the logarithmic equation of state [Poirier and Tarantola, 1998], ρ(P ) ρ(P ) P = KT log ρ0 ρ0 !" KT′ − 2 ρ(P ) 1+ log 2 ρ0 !# , (5.8) where KT and KT′ are the isothermal bulk modulus and first pressure derivative, respectively, I iteratively solve for αV (P, T ) by the Newton-Raphson method. Average density and expansivity are estimated as the weighted mean of the individual phases, which are in turn computed from a weighted mean of the individual mineral endmembers. Constants used to estimate expansivity and density are summarized in Appendix D. 5.3.1.2 αV vs. αeff Laboratory studies are used to estimate the volumetric expansivity by allowing full expansion or contraction. Because the Earth’s surface is a free moving boundary, contraction can occur completely in the vertical direction. However, the strength of the lithosphere in the lateral dimensions is enough to overcome some degree of contraction. Thus, using the laboratory expansivity estimates leads to an overestimate of the total contraction and consequently subsidence as the lithosphere cools. Pollack [1980] first noted that the effective expansivity required to explain subsidence of the oceanic lithosphere is between 70–85% of the full volumetric expansivity. The viscoelastic rheology of the lithosphere depends upon temperature and allows for relaxation of stress and an increase in contraction toward the volumetric limit. Recently, Korenaga [2007a] rigorously modeled this effect, showing the significant time- as well as depth-dependence of this phenomenon with a similar effect on expansivity as estimated by Pollack [1980]. Rather than model the effect directly, which is numerically expensive, I use the results of Korenaga [2007a] to estimate an empirical approximation for the time 72 and temperature dependence of the effective expansivity. The effective thermal expansivity is reasonably approximated by, αeff = 0.5707 + 4.777 × 10−4u − 5.823 × 10−7 u2 + 5.403 × 10−10 u3 , αV (5.9) where u is given by, u = T − 552.1t−0.25 , (5.10) and is a function of both temperature, T in Kelvins, and plate age, t in m.y. Because this approximation is unstable when t is close to zero, I set u = T −552.1 for ages t < 1 m.y. The ratio of effective to volumetric expansivity is also bounded between 0.55 and 1. I multiply the ratio calculated using Equation 5.9 by the temperature-dependent expansivity before estimating density. Figure 5.1 illustrates the accuracy of the Equation 5.9 relative to the model proposed by Korenaga [2007a]. While the approximation is not perfect, it is accurate enough for my purposes given the uncertainty in estimating the effective expansivity from rheologic properties. 5.3.1.3 Heat Capacity, CP The heat capacity can be approximated as CP = c0 + c1 T −0.5 + c2 T −2 + c3 T −3 + dCP P, dP (5.11) with empirically derived constants, ci [Fei and Saxena, 1987]. The pressure contribution is computed using the thermodynamic identity [Osako et al., 2004], ∂CP ∂P ! T T ∂α =− α2 + ρ(P, T ) ∂T ! . (5.12) P The bulk heat capacity of the rock matrix is computed as the mean heat capacity of all phases weighted by their volume fractions. Constants used to estimate heat capacity are summarized in Appendix D. 73 Temperature [K] 400 800 1200 10 m.y. 50 m.y. 100 m.y. 1600 0.5 0.6 0.7 0.8 αeff/αV 0.9 1.0 Figure 5.1. Effective thermal expansivity model. The solid lines represent the modeled thermal expansivity by Korenaga [2007a] for 10, 50 and 100 m.y. The dashed lines are computed using Equation 5.9 at the same times. 5.3.1.4 Thermal Conductivity, k Thermal conductivity results from the combination of lattice (phonon-phonon) and radiative (photon-phonon) mechanisms. The total conductivity is written simply as a sum of these two components [Schatz and Simmons, 1972], k(P, T ) = kL (P, T ) + kR (P, T ). (5.13) For any given phase, the lattice thermal conductivity, kL , can be given by kL (P, T ) = k ◦ 298 T n K′ 1+ TP , KT ! (5.14) where k ◦ is the conductivity at 0 GPa and 298 K. The conductivity for a solid-solution mineral is slightly more complicated, but is easily adjusted by replacing k ◦ with a quadratic as a function of the mole fraction of an end-member. Hence, the thermal conductivity of a two-member solid-solution series is written: 2 kL (P, T ) = k0 + k1 χ + k2 χ 298 n T K′ 1+ TP , KT ! (5.15) where χ is the mole fraction of a given end-member. This equation is a simplification of Hofmeister [1999b] Equation 10, as suggested by Beck et al. [2007]. Thermal 74 conductivity of a more complex solid-solution mineral such as amphibole and garnet is more complicated still. The radiative contribution to thermal conductivity, kR , is near zero at room temperature and grows in influence with temperature. Hofmeister [1999a] estimated the radiative contribution for olivine using a third-order empirical equation. However, polynomial formulations grows quickly to unrealistic values or return negative conductivities beyond the experimentally calibrated bounds. Therefore, I prefer a formulation that is always positive and well-behaved when extrapolated beyond experimental conditions. The estimated radiative conductivity data used to calibrate my models appear to behave similar enough to that of an arctangent or error function may be used to fit the data (Appendices A and C). I estimate radiative conductivity using kR (T ) = kRmax [1 + erf (ω(T − Tm ))] , (5.16) where kRmax is the maximum radiative conductivity, ω is a scaling factor and Tm is the temperature at 0.5kRmax . The radiative contribution has been measured on very few minerals. For olivine, by far the best studied, the experimental estimates are highly variable. The most recent estimates suggest that the radiative contribution is small (Hofmeister [2005], Appendix A) with even smaller values for lherzolite [Gibert et al., 2003]. I use the estimate from lherzolite for olivine and pyroxenes. Spinel and garnet are opaque at mantle temperatures and may have little influence on radiative conductivity and therefore are ignored [Shankland et al., 2005]. No data exist for the radiative effect on plagioclase, amphibole or chlorite. I assume no radiative contribution for plagioclase, values similar to mica for chlorite, and similar to lherzolite for amphibole. A pressure component to radiative conductivity may also exist, but it is even more poorly studied than the temperature effect and therefore ignored [T. Shankland, pers. comm. 2006] For a mineral composite with N phases with mole fractions wi , the effective thermal conductivity, keff , is computed using the geometric mean [Clauser and Huenges, 1995], keff = exp " PN i=1 # wi log(kLi + kRi ) . PN i=1 wi (5.17) 75 I use this formulation for the geometric mean rather than the more common product form because it is easier to differentiate. Constants used to compute mineral conductivities are given in Appendix D. Very few measurements have been made on chlorite, clinopyroxene, or spinel from which to estimate room-temperature conductivity, and variable and P –T data are even more limited. For minerals with limited conductivity data, k ◦ is estimated to be the mean of all available determinations rather than estimating a quadratic between end-members. When temperature-dependent conductivity data do not exist, the temperature exponent, n, is estimated as 0.5 as suggested by Xu et al. [2004]; however, this assumption may not be true for all silicates. Conductivity measurements have been performed on a limited range of clinopyroxene compositions, most Di>90 [Horai , 1971; Diment and Pratt, 1988; Harrell , 2002; Hofmeister and Pertermann, 2008] and three with slightly higher iron contents (augite) [Horai , 1971; Hofmeister and Pertermann, 2008]. The conductivity of clinopyroxene measured using contact methods at room temperature ranges from 4.05–5.57 W/m-K. Room-temperature conductivity of diopside estimated by Hofmeister and Pertermann [2008] using laser flash analysis is 8.84 W/m-K, significantly higher than contact methods. Hofmeister and Pertermann [2008] suggest that laser flash analysis estimates are typically ∼20% higher than contact methods due to contact resistance, which appears to explain less than half the difference. However, a correction is often applied for contact resistance and the only studies reporting conductivities using the laser method are by Hofmeister et al., making it difficult to verify their claim. Therefore, I use the average conductivity of seven clinopyroxene samples measured using contact methods. The temperature coefficient is estimated using the results by Hofmeister and Pertermann [2008]. Because of extremely complex cation and anion substitution in amphiboles, computing the thermal conductivity from a simple quadratic of two end-members is not possible. A more complex scheme is necessary. One based on an analysis of major element chemistry and cation substitution may provide a reasonable conductivity estimate (Appendix B). I use the following approximation for the conductivity: 76 ◦ kamph = 4.0654χtr + 3.7834χfact + 1.7066χparg, (5.18) using the mole fraction of end-member compositions for tremolite (tr), ferro-actinolite (fact), and pargasite (parg). 5.3.2 Mantle Phase Changes Two major phase changes occur within the lithospheric mantle modeled in this study. The plagioclase-spinel transition, plagioclase + 2 olivine ↔ clinopyroxene + 2 orthopyroxene + spinel. (5.19) is roughly independent of temperature, with a transition pressure of ∼0.9 GPa [Herzberg, 1976]. The spinel-garnet transition, 2 orthopyroxene + spinel ↔ garnet + olivine, (5.20) varies in pressure as a function of temperature. The pressure of the spinel-garnet transition is estimated using the empirical relationship Psg (T ) = 1.4209 + exp(3.9073 × 10−3 T − 6.8041), (5.21) where T is in Kelvins. This curve is calibrated by inversion of the data reported by Robinson and Wood [1998], Walter et al. [2002], Klemme and O’Neill [2000] and references therein. 5.4 5.4.1 Datasets Bathymetry I prefilter ETOPO2 bathymetry, excluding regions defined as large igneous provinces (LIPs) [Coffin and Eldholm, 1994]. While LIPs are excluded, the flexural effect due to loading of the lithosphere can extend for several hundred kilometers. By excluding the high bathymetry associated with LIPs and accepting the associated sedimentary moats, the average bathymetry–age relationship may be biased towards greater subsidence [Crosby et al., 2006; Crosby and McKenzie, 2009]. Bathymetry 77 deviations from the average also occur as a result of variations in crustal thickness and/or mantle temperatures. For example, petrologic models predict higher melt fractions with higher adiabatic temperatures. The thickness of oceanic crust is then related to the quantity of partial melt generated within the mantle reservoir [Asimow et al., 2001]. Both the increased crustal thickness and temperatures increase bathymetry. However, depending upon how one defines ‘anomalous’ bathymetry, these natural variations may or may not be removed. Attempts to filter out ‘anomalous’ bathymetry in response to loading are made by restricting the magnitude of gravity anomalies [Crosby et al., 2006] and by using a spatial correlation and distances from seamounts [Korenaga and Korenaga, 2008]. Using Korenaga and Korenaga’s filters, average bathymetry is ∼20 m shallower with higher scatter than an unfiltered analysis. Therefore, I have chosen to use the unfiltered data. By doing so, I capture the entire flexural signals for most seamounts but may bias bathymetry to greater average depths, particularly in old ages where LIPs are more common. Sediment loading significantly influences bathymetry and is particularly apparent on passive continental margins where large sedimentary deposits result in shallowing water depth with age. Assuming negligible flexure, the equilibrated isostatic response to sediment loading is quite straightforward. It can be shown that the isostatic correction to bathymetry as a result of sediment loading is computed from the the average sediment density (ρ̄s ), density of seawater (ρw ), and density of the displaced mantle (ρm ) as ρm − ρ̄s , (5.22) ρm − ρw where h is the sediment thickness, ∆ε, is the change in water depth between the ∆ε = h adjusted and observed bathymetry [Sykes, 1996]. The effective sediment density, ρ(z), is computed as a function of the porosity, φ(z): ρs (z) = ρg + (ρw − ρg )φ(z), (5.23) where ρg is the matrix (grain) density. I use a two-layer porosity model given by φ(z) = ( φ1 e−z/h1 φ2 e−z/h2 if z ≤ H , if z > H (5.24) 78 Table 5.1. Parameters for sediment correction. Property grain density, ρg porosity, φ1 porosity, φ2 characteristic depth, h1 characteristic depth, h2 characteristic thickenss, H seawater density, ρw mantle density, ρm Carbonate Oozea 2700 0.656 0.409 1278 2591 1190 1030 3340 Units kg-m−3 m m m kg-m−3 kg-m−3 Pelagic Clayb 2500 0.812 0.504 664 2160 457 a Constants determined by inversion of chalk data reported by Mallon and Swarbrick [2002]. b Constants from Velde [1996]. where φ1 and φ2 are the extrapolated surface porosity, h1 and h2 are the decay constants (units of depth), and z is the depth. There is a fundamental change in the nature of compaction curves for most sediment types requiring a two-layered model at sediment thicknesses greater than a characteristic thickness, H [Sclater and Christie, 1980; Velde, 1996; Mallon and Swarbrick , 2002]. The average density is computed by integration from the surface to the base of the sediment column, ρ̄s = h−1 Z 0 h ρs (z)dz, (5.25) where h is the sediment thickness. The average density for thicknesses h < H is simply ρ̄s = ρg + (ρw − ρg )φ1 h1 1 − e−h/h1 , h (5.26) where hc is the compaction decay length, φ0 is the unloaded porosity, and ρs and ρw are the maximum rock and water densities, respectively. If h > H, h ρ̄s = ρg + (ρw − ρg )h−1 φ1 h1 1 − e−H/h1 + φ2 h2 e−H/h2 − e−h/h2 i , (5.27) All of the constants depend on the sediment type (Table 5.1). The two dominant oceanic sediment types are pelagic clay and carbonate ooze which spatially cover most of the oceanic crust although, near continents, terrigenous sediments are more common. Carbonates dominate shallow bathymetric regions, but are not stable below ∼4500 m water depth, allowing pelagic clay to dominate. This 79 depth is called the Carbonate Compensation Depth (CCD) and depends on water temperature, salinity, and pH [Spinelli et al., 2004]. Since the CCD depth is somewhat uncertain and some near-surface carbonate deposited before the seafloor descends below the CCD may be re-absorbed, sediment density is computed as carbonate ooze above 4000 m, pelagic clay below 5000 m and a linear combination of carbonate ooze and pelagic clay between 4000 and 5000 m. This simplification is similar to those proposed by Sykes [1996]. 5.4.2 Heat Flow I use the heat flow database from a new global compilation to estimate the heat flow as a function of seafloor age (Chapter 2). Heat flow at young ages is strongly affected by hydrothermal circulation. Binned values at young ages exhibit a significant deficit as a result of a systematic bias caused by hydrothermal circulation and by measurement constraints. Filtering heat flow data to retain only sites far from seamounts (>85 km) and with significant sediment thickness (>325 km) minimizes the effect of hydrothermal circulation on estimated heat flow (Table 5.2). The filtered heat flow values in this study are largely uncorrected for sedimentation and thermal rebound. Chapter 4 presents a detailed analysis of filtering heat flow to produce the results used in this study. Even after filtering, heat flow out of young seafloor may still exhibit a hydrothermal deficit. By examining high-density data from regional surveys with seismic control on sediment thickness and basement roughness, it is possible to identify sites with minimal hydrothermal disturbance and obtain more accurate estimates of conductive heat loss (Davis et al. [1992, 1999], and Chapter 4). I use regional estimates from several localities to supplement heat flow out of young seafloor (Chapter 4, Table 4.1). 80 Table 5.2. Global bathymetry and heat flow data. Age Ma 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 εadj ± σε m 2980 ± 552 3159 ± 548 3332 ± 559 3476 ± 580 3579 ± 589 3672 ± 542 3772 ± 564 3883 ± 570 3962 ± 594 4028 ± 561 4115 ± 522 4188 ± 532 4250 ± 576 4327 ± 623 4412 ± 568 4454 ± 566 4534 ± 568 4566 ± 597 4615 ± 618 4672 ± 689 4719 ± 643 4767 ± 657 4801 ± 667 4892 ± 665 4925 ± 688 4986 ± 718 5039 ± 713 5060 ± 671 5044 ± 676 5057 ± 725 5110 ± 757 5169 ± 809 5145 ± 719 5160 ± 745 5261 ± 681 5313 ± 686 Nq 433 622 189 165 152 1044 1082 74 119 172 115 73 52 77 63 46 66 114 45 36 39 39 14 22 38 37 30 21 18 17 28 26 27 27 36 14 q ± σq mW-m−2 205.0 ± 214.2 214.0 ± 121.3 124.0 ± 38.7 100.0 ± 37.7 101.5 ± 50.4 116.0 ± 80.1 103.2 ± 65.0 100.0 ± 40.3 95.0 ± 45.6 83.9 ± 46.2 86.6 ± 40.4 93.0 ± 37.8 80.0 ± 38.5 87.0 ± 40.8 79.0 ± 30.1 76.0 ± 29.8 69.1 ± 38.2 77.0 ± 97.9 69.9 ± 32.0 59.0 ± 29.8 67.4 ± 56.9 76.6 ± 47.4 63.0 ± 35.0 76.2 ± 31.7 62.2 ± 28.2 64.9 ± 16.7 57.0 ± 59.2 68.2 ± 9.5 69.3 ± 20.2 64.0 ± 31.2 62.8 ± 14.7 59.7 ± 30.4 65.0 ± 95.4 58.1 ± 13.1 61.5 ± 15.7 68.5 ± 24.6 Age Ma 73 75 77 79 81 83 85 87 89 91 93 95 97 99 101 103 105 107 109 111 113 115 117 119 121 123 125 127 129 131 133 135 137 139 141 143 εadj ± σε m 5321 ± 651 5297 ± 674 5292 ± 669 5268 ± 682 5283 ± 735 5268 ± 690 5279 ± 694 5291 ± 761 5368 ± 780 5388 ± 704 5385 ± 720 5391 ± 753 5402 ± 739 5419 ± 746 5455 ± 693 5443 ± 723 5424 ± 772 5458 ± 815 5477 ± 951 5498 ± 854 5416 ± 917 5412 ± 955 5399 ± 975 5347 ± 995 5397 ± 808 5439 ± 693 5398 ± 762 5477 ± 712 5506 ± 734 5501 ± 810 5524 ± 873 5565 ± 975 5523 ± 965 5486 ± 925 5567 ± 1056 5579 ± 949 Nq 22 23 15 18 95 40 133 56 35 46 46 62 32 16 9 29 19 28 38 19 11 22 15 105 38 90 77 16 11 18 16 23 34 19 13 25 q ± σq mW-m−2 57.2 ± 37.3 64.0 ± 36.3 59.3 ± 24.3 66.4 ± 31.6 72.0 ± 24.3 48.3 ± 25.7 57.0 ± 50.1 36.3 ± 42.6 45.0 ± 47.1 53.5 ± 14.4 56.9 ± 15.5 58.7 ± 15.6 43.6 ± 18.6 56.3 ± 81.0 48.0 ± 44.1 55.4 ± 31.0 49.0 ± 10.9 55.9 ± 12.5 53.2 ± 23.8 49.8 ± 25.0 38.4 ± 22.2 52.0 ± 55.5 53.6 ± 15.2 45.6 ± 8.3 49.2 ± 14.3 54.7 ± 13.6 51.1 ± 21.8 50.1 ± 14.1 50.0 ± 28.1 46.3 ± 18.9 51.1 ± 14.3 47.7 ± 16.6 49.8 ± 8.5 49.1 ± 8.4 47.7 ± 45.5 50.7 ± 4.6 81 Table 5.2. continued. Age Ma 145 149 151 153 155 157 159 161 163 εadj ± σε m 5528 ± 1019 5571 ± 1042 5520 ± 1040 5495 ± 1025 5199 ± 977 5006 ± 1138 5281 ± 1526 5660 ± 790 5578 ± 783 Nq 8 54 19 21 20 13 39 59 26 q ± σq mW-m−2 36.4 ± 10.0 51.8 ± 20.3 50.7 ± 38.6 50.7 ± 11.0 44.2 ± 9.5 40.7 ± 8.4 49.4 ± 29.4 46.9 ± 8.7 48.9 ± 19.9 Age Ma 147 165 167 169 171 173 175 177 179 εadj ± σε m 5535 ± 1051 5572 ± 788 5595 ± 772 5538 ± 737 5492 ± 698 5535 ± 815 5558 ± 897 5525 ± 997 5426 ± 1204 Italicized values excluded from misfit computation. Nq 12 16 36 12 12 15 21 44 16 q ± σq mW-m−2 50.6 ± 5.0 48.7 ± 28.0 48.4 ± 30.6 49.4 ± 18.1 53.0 ± 15.2 49.0 ± 10.0 49.0 ± 8.4 49.9 ± 10.8 48.1 ± 15.6 82 5.5 Results and Discussion 5.5.1 Plate Thickness Before fitting models to data, I explore the influence of model parameters on subsidence and heat flow. A suite of models varying the maximum plate thickness between 65 and 125 km is shown in Figure 5.2. The effect of maximum plate thickness variations on subsidence and heat flow in young seafloor is negligible. However, the models rapidly diverge as diffusion lengths reach and are limited by maximum plate thickness. As plate thickness grows, the asymptotic limits on water depth and heat flow converge as the solution approaches half-space cooling within the time interval investigated. Because of the large range of heat flow, it can be difficult to see differences between the various models on a linear scale. I therefore plot the heat flow on a nonlinear scale where q −2 ∝ t to highlight the differences between the various models at old ages. To more easily observe model differences at young ages, bathymetry and heat flow are shown relative to a preferred model (Section 5.5.5). From these plate models, I expect models with greater maximum plate thicknesses to better fit heat flow observations. 5.5.2 Mantle Adiabat Subsidence and heat flow of plate cooling models with mantle potential temperatures varying from 1200 to 1600◦ C begin to diverge at very early time (Figure 5.3). The subsidence patterns differ markedly from the half-space cooling model. Unlike the subsidence pattern resulting from plate thickness variations, subsidence due to differences in the mantle potential temperature largely occur at ages <80 m.y. At old ages, the subsidence difference for 50 K difference in mantle potential temperature is ∼200 m. At the youngest ages, the 400 K range in adiabatic temperatures tested produces >60 mW/m2 difference in predicted heat flow. However, this range is rapidly reduced to ∼20 mW/m2 by 50 m.y. At old ages, a 50 K difference in adiabatic temperature asymptotically approaches a heat flow difference of 1 mW/m2 . Most median heat flow bins are captured by the tested range of potential temperatures. Plate Thickness 0 0 HSC 4 0 50 100 Age [m.y.] 250 (d) Heat Flow [mW/m2] 200 150 100 65 75 50 0 2 75 85 95 105 115 125 3 4 150 65 0 400 200 125 100 80 70 50 100 Age [m.y.] 150 115 105 95 0 85 -400 75 -800 65 0 50 100 Age [m.y.] 150 30 (e) 65 60 75 85 50 95 105 125 HSC 40 (c) 400 50 100 150 Age [m.y.] 125 0 ∆ Subsidence [m] 3 75 85 95 105 115 125 1 0 50 100 Age [m.y.] 150 ∆ Heat Flow [mW/m2] 2 65 Subsidence [km] 1 125 800 (b) Heat Flow [mW/m2] Subsidence [km] (a) (f) 20 65 10 75 85 95 0 125 -10 -20 -30 0 50 100 Age [m.y.] 150 83 Figure 5.2. Influence of plate thickness variations on subsidence and heat flow. The dashed line represents the half-space cooling model (HSC). Bathymetry data are reduced by the approximate depth to the ridge, 2750 m. Data (dots) and reference model (solid line) same as Figure 5.6. Potential Temperature 0 0 2 1200 1300 1400 3 1500 1600 1600 1 1200 2 1300 1400 3 1500 ∆ Subsidence [m] 1 100 Age [m.y.] (d) Heat Flow [mW/m2] 200 150 100 0 4 150 250 1600 50 0 50 100 Age [m.y.] 1350 1300 -400 50 100 150 Age [m.y.] 400 200 125 100 80 70 150 1200 0 50 100 Age [m.y.] 150 30 (e) 60 1600 1500 50 1300 1200 1200 0 0 1250 ∆ Heat Flow [mW/m2] 50 1550 1500 1450 1400 -800 Heat Flow [mW/m2] 0 400 1600 HSC 4 (c) 800 (b) Subsidence [km] Subsidence [km] (a) (f) 20 1600 10 1500 0 1400 1300 1200 -10 -20 HSC 40 0 50 100 Age [m.y.] 150 -30 0 50 100 Age [m.y.] 150 84 Figure 5.3. Influence of mantle potential temperature variations on subsidence and heat flow. The dashed line represents the half-space cooling model (HSC). Data and reference model same as Figure 5.2. 85 5.5.3 Crustal Thickness At early time during the evolution of the plate models computed in this study, the crust represents a significant fraction of the total lithospheric thickness. As the system evolves, the fraction of crustal thickness to that of the entire lithosphere rapidly decreases. As a result, I expect the crustal influence on the thermal evolution to be greatest at early time. Models computed with crustal thickness varying between 1 and 15 km suggest that the impact of a crust is greatest in the first 10 m.y. with almost no influence beyond 50 m.y. However, some of this difference is due to the difference in initial temperatures conditions (Section 5.5.6.2). Variations in crustal thickness produce patterns similar to those produced by varying adiabatic temperatures except that the differences between models rapidly (20–30 m.y.) reach a nearly constant difference (Figure 5.4), whereas differences between adiabatic models increase out to ∼80 m.y. (Figure 5.3). Subsidence and heat flow are less sensitive to crustal thickness variations than to differences in the mantle potential temperature. 5.5.4 Model Misfits I run a grid search to identify the minimum misfit of plate cooling models to heat flow and bathymetric data. Misfit of the models to data are computed by, M N 1 X 1 X (εobs,i − εmod,i)2 (qobs,i − qmod,i )2 misfit = + 2 2 M i=0 σε,i N i=0 σq,i " #−1/2 , (5.28) where εi and qi are the individual water depth and heat flow associated with a total of M and N ages. The subscripts ‘obs’ and ‘mod’ are the observed and model values, respectively. The misfits are normalized by the standard deviations, σ, of the respective datasets and are therefore dimensionless. Because models compute subsidence, not bathymetry, the initial ridge depth must also be computed to determine the model misfit to the data. The initial ridge depth is determined such that the misfit between the subsidence curve and the bathymetry is minimized. The initial temperature condition is 0◦ C at the seafloor and increased linearly to the adiabatic temperature at the base of the crust. For models with no crust, temperatures are increased to the adiabat at 1 km to increase numerical stability. The adiabatic gradient is assumed to Crustal Thickness 0 0 (b) 2 15 11 3 1 5 1 2 15 11 3 1 5 HSC 50 100 Age [m.y.] 150 250 (d) Heat Flow [mW/m2] 200 150 100 1 50 0 0 400 200 125 100 80 70 50 100 Age [m.y.] 5 9 13 -400 150 (e) 60 1 50 15 HSC 40 0 50 100 Age [m.y.] 150 30 15 0 1 0 50 100 150 Age [m.y.] 0 50 100 Age [m.y.] 150 ∆ Heat Flow [mW/m2] 0 400 -800 4 Heat Flow [mW/m2] 4 (c) 800 ∆ Subsidence [m] 1 Subsidence [km] Subsidence [km] (a) (f) 20 10 1 3 5 7 9 11 15 0 -10 -20 -30 0 50 100 Age [m.y.] 150 86 Figure 5.4. Influence of crustal thickness variations on subsidence and heat flow. The dashed line represents the half-space cooling model (HSC). Preferred model in represented by the heavy line. Data used to estimate misfit are shown for reference. Data and reference model same as Figure 5.2. 87 Table 5.3. Compositional model for the oceanic lithosphere. Mineral albite anorthite clinochore daphnite tremolite ferro-actinolite pargasite diopside hedenbergite enstatite ferrosillite forsterite fayalite spinel pyrope almandine grossular A [µW/m3 ] Crust basalt gabbro 17.85 10.08 42.05 37.92 1.89 0.21 3.08 5.28 0.44 15.70 26.14 4.90 5.86 5.14 5.86 5.59 7.09 3.01 1.91 0.1 0.03 plag-lherzolite Mantle sp-lherzolite gt-lherzolite 0.71 3.82 0.41 17.49 1.94 68.43 7.20 0.003 4.49 0.41 18.77 1.93 66.55 7.15 0.70 0.003 4.49 0.41 17.32 1.78 67.00 7.20 1.19 0.43 0.18 0.003 Values as molar fraction of approximate end-member mineralogy. Crustal compositions from Hacker et al. [2003], mantle composition computed from Niu et al. [1997]. be 0.3 K/km. The model is discretized at 200 m in depth and 0.01 m.y. in time. Finer discretizations in depth and time have been tested to ensure models are not improved within my desired level of precision. A total of 2873 models are computed, with crustal thickness ranging from 0–15 km, maximum plate thickness ranging from 65–125 km and mantle potential temperatures from 1200–1600◦C. Layer compositions, including the heat production, are given in Table 5.3. The basaltic layer is assumed to be 2 km thick and all changes in crustal thickness are attributed to variations in the cumulate layer (seismic layer 3), which tends to dominate crustal structure [Mutter and Mutter , 1993]. Seismic evidence suggests average thickness of the igneous oceanic crust is 7.1±0.8 km [White et al., 1992]. Crustal thickness varies from 0 km at slow spreading centers dom- 88 inated by structural deformation [Tucholke et al., 1998, 2001] to >10 km thickness at hot fast spreading ridges, although it can be much greater in regions with anomalous volcanism such as Galapagos and Iceland [White et al., 1992; Mutter and Mutter , 1993; Asimow et al., 2001]. In this study, minimum misfit is achieved with 2 km thick crust (Figure The misfit surface shows more of a trough than a well defined global minima, suggesting that crustal thicknesses and adiabatic temperatures within this zone yield similarly fitting models. This misfit trough roughly correlates with chemically calibrated estimates of oceanic crustal thickness generated by melting MORB source as a function of potential temperature [Asimow et al., 2001]. However, the theoretical 5.5). function predicts lower adiabatic temperatures than my misfit would suggest. This may result from additional physical influences on lithospheric buoyancy, surface heat flow, or uncertainties in model parameters. Because this trough poorly estimates crustal thickness, I choose a value of 7 km consistent with seismic estimates in order to estimate other input parameters. The misfit assuming a 7 km thick crust is 0.43. Best-fitting potential temperature using a 7 km thick crust yields an estimated potential temperature of 1425◦ C. The range of potential temperatures with low misfit is ∼75 K. Adiabatic temperature estimates discussed above are only one of several proposed models derived from basaltic chemistry. Another competing model suggests potential temperatures are low (∼1300◦ C) with an additional 250 K variation resulting from heating in proximity to hotspot sources (plumes?). These hypotheses are evaluated by Herzberg et al. [2007], who find the more consistent with a lower adiabatic temperature. Below I discuss further the difference between my model and the geochemical prediction. My choice of crustal thickness results in a best-fitting plate thickness of 90 km (Figure 5.5). The misfit surface shows a well defined range of plate thickness (∼10 km) which reasonably fits the bathymetry and heat flow data. The 90 km maximum plate thickness is relatively independent of the choice of crustal thickness. Plate thickness estimates from Parsons and Sclater [1977] and Stein and Stein [1992] that use temperature and pressure independent physical properties yield 125 and 95 km, 89 Plate Thickness, 90 km Potential Temperature [°C] 1.2 (a) 1.0 0.8 1500 0.6 w mo Asi l. et a ] 01 [20 1400 0.6 1300 0.4 0.8 1.0 1200 0 Crustal Thickness, 7 km 1600 Potential Temperature [°C] 1600 1.2 5 10 Crustal Thickness [km] 15 (b) 1.2 1.0 0.8 1500 0.6 1400 0.6 1300 1200 0.8 70 80 90 100 110 120 Plate Thickness [km] Potential Temperature, 1425°C 0.8 Plate Thickness [km] 120 (c) 110 0.6 100 90 80 0.6 0.8 1.0 1.2 70 0 5 10 Crustal Thickness [km] 15 Figure 5.5. Slices through the misfit surface at (a) a plate thickness of 90 km, (b) a crustal thickness of 7 km, and (c) a mantle potential temperature of 1425◦ C. ‘×’ denote the minimum misfit for each slice. 90 respectively. McKenzie et al. [2005] showed that models incorporating temperatureand pressure-dependent properties yield ∼85% lower plate thicknesses than their constant-property counterparts (106 and 83 km). However, McKenzie et al. [2005] use volumetric rather than effective expansivity estimates. Because the volumetric expansivity plate models have overall greater subsidence than effective expansivity models for given plate thickness, the best-fitting plate thickness in models incorporating the effective expansivity is greater, consistent with my results. 5.5.5 Preferred Model My preferred model with 7 km thick igneous crust, 90 km maximum plate thickness, and a mantle potential temperature of 1425◦C is shown in Figure 5.6. Modeled subsidence is shifted by 2760 m to match bathymetric data. This model shows excellent agreement for median bathymetry. A slight under-prediction of bathymetry occurs at young ages (<50 Ma), which compensates for the increased variability in bathymetry at older ages that are over-predicted. Modeled heat flow also agrees well with the data. When a inverse square transform is applied to heat flow to make modeled results more linear, differences between the observed heat flow and model are apparent. Heat flow at young ages fit very well, a region traditionally ignored in oceanic cooling models because of hydrothermal effects on heat flow. While heat flow at older ages appears to fit less well in the transformed space, small variations that are well within uncertainty appear very large and are not a cause for concern. 5.5.6 5.5.6.1 Additional Influence on Plate Models Volumetric vs. Effective Expansivity As mentioned previously, ignoring the effects of strength within the lithosphere affects estimates of expansivity. Korenaga [2007a] showed the importance of incorporating a viscoelastic rheology into estimates of expansivity. Figure 5.7 shows the difference between model outputs computed using my empirical estimate of effective expansivity and the volumetric expansivity. The difference in subsidence quickly grows to reach half the maximum by 30 Ma. The total subsidence difference at 91 2.5 2.5 3.5 (b) Bathymetry [km] Bathymetry [km] (a) 4.5 5.5 50 100 Age [Ma] 150 400 (c) 300 0 200 100 0 50 100 Age [Ma] 150 50 Age [Ma] 400 200 125 100 80 70 Heat Flow [mW/m2] Heat Flow [mW/m2] 4.5 5.5 0 0 3.5 100 150 (d) 60 50 40 0 50 100 Age [Ma] 150 Figure 5.6. Preferred plate cooling model with a crustal thickness of 7 km, plate thickness of 90 km and potential temperature of 1425◦C including heat production. Data shown with 1-σ bars. Data used to compute misfit are in dark grey, unused in light grey. 92 1000 30 800 volumetric expansivity 600 ol-only mantle 400 no crust no crust-2 200 ∆ Heat Flow [mW/m2] ∆ Subsidence [m] (a) (b) 25 20 15 10 ol-only mantle 5 no crust 0 0 0 50 100 Age [m.y.] 150 -5 volumetric expansivity 0 50 100 Age [m.y.] 150 Figure 5.7. Model sensitivity to selected parameters. See text for description of models. Reference model (zero lines) is same as preferred (Section 5.5.5). 180 Ma is >700 m. Thus, to achieve a similar fit to bathymetry using the volumetric expansivity, the plate thickness would have to be decreased by ∼20 km. There is decrease in heat flow when using the volumetric expansivity that results from conductivity and heat capacity changes from increased lithospheric pressures. My analysis does, however, ignore the influence of thermal-cracking and crackfilling in the upper part of the lithosphere, which would reduce the difference between the effective and volumetric expansivity. While this effect is not large enough to influence the results strongly, a few percent change in effective expansivity decreases the estimated plate thickness slightly. An excellent exploration of thermal-cracking in oceanic lithosphere is made by Korenaga [2007b]. 5.5.6.2 Lithospheric Composition Thus far I have only explored one compositional model. Most plate cooling models use monomineralic composition (Fo-90) for the entire lithosphere. A few plate cooling models include a polymineralic compositions but still use Fo-90 composition to compute conductivity. However, conductivity of polymineralic peridotite can differ by 10-20% from Fo-90. Comparison between the polymineralic models lacking a crust and a model with Fo-90 olivine composition results in ∼50 m higher subsidence in the 93 Fo-90 model (Figure 5.7). The heat flow difference between the two is <1 mW/m2 . Thus, the effect of mantle composition on plate cooling models is minor and potentially a second-order effect on subsidence. While very few models include polymineralic assemblages, even fewer include a crust. The one exception is the study by Afonso et al. [2008], which again uses an Fo-90 composition to estimate conductivity. Since conductivity is required to compute heat flow, predicted heat flow should be initially higher without a crust. This difference in heat flow affects the efficacy of cooling, and as a result increases the subsidence rate. One would also expect the density difference from decreased lithostatic pressure gradients between models and differences in thermal expansion with and without crustal layers to cause a difference in subsidence, particularly at young ages where the crust represents a significant fraction of the total lithospheric thickness. As predicted, subsidence is much greater in the model without a crust. The initial depth at which temperatures reach the adiabat differs for varying crustal thicknesses in my model; therefore, a second model lacking a crust is run with the same initial temperature condition as my preferred model so they may be compared more fairly (Figure 5.7, no crust-2). The result indicates that 200 m of additional subsidence (∼2/3 total subsidence) results from my choice of initial condition as temperatures rapidly diffuse at young ages (∼10 m.y.). The inclusion of a crust reduces the overall subsidence by 100–150 m. The effect on heat flow is even more striking with an initially higher heat flow (>30 mW/m2 ) in the models lacking a crust with an asymptotically higher heat flow of ∼2 mW/m2 at ages >90 m.y. Including a crust results in a ∼20% lower minimum misfit than the minimum misfit without a crust. 5.5.6.3 Heat Production Data that can be used to estimate heat production within the oceanic lithosphere are sparse, but sufficient to make a preliminary estimate. Heat production results from the radiogenic decay of U, Th, and K which can be estimated from geochemical analyses of mid-ocean ridge basalts (MORB). Average estimates of heat production 94 Table 5.4. Compositional model for the oceanic lithosphere. Values as molar fraction of approximate end-member mineralogy. Element E-MORBa N-MORBa N-MORBb DMb DMc K2 O wt.% 0.48 ± 0.26 0.14 ± 0.07 0.09 ± 0.045 0.072 0.060 Th ppm 0.97 ± 0.68 0.20 ± 0.14 1.2 ± 0.08 0.0137 0.0079 U ppm 0.37 ± 0.31 0.083 ± 0.054 0.05 ± 0.04 0.0047 0.0032 d 3 A µW/m 0.23 ± 0.11 0.055 ± 0.020 0.033 ± 0.014 0.0036 0.0024 a Niu et al. [2002] from the East Pacific Rise b Salters and Stracke [2004] derived from modeled chemical ratios c Workman and Hart [2005] estimated from abyssal peridotites d assuming density of 3000 kg/m3 for basalts and 3340 kg/m3 for peridotites. range from 0.03 µW/m3 in depleted (normal) N-MORB to 0.2 µW/m3 in (enriched) E-MORB (Table 5.4). Most of the ocean crust is made of N-MORB, while E-MORB is more frequently found near seamounts and oceanic island basalts. However, EMORB has been found far from seamounts representing a small but significant fraction (<∼5%) of analyzed basalts [Donnelly et al., 2004]. Estimated heat production from a bulk oceanic crust consisting of 95% N-MORB and 5% E-MORB is ∼0.05 µW/m3 , representing ∼0.35 mW/m2 of the surface heat flow for a 7 km thick crust. Hydrothermal alteration of the crust tends to increase the concentration of heat producing elements substantially [Wheat and Mottl , 2004]. Weight percent K can be increased by ∼7 times, while U and possibly Th [Hart and Staudigel , 1982] can be increased even more, bringing estimated upper crustal heat production in altered N-MORB to values similar to unaltered E-MORB. However, this effect has yet to be estimated on a large scale and is estimated from three deep-sea drill holes. Radiogenic heat production within the oceanic mantle lithosphere is at least an order of magnitude smaller than within the oceanic crust. This depleted mantle (DM) contributes <0.3 mW/m2 to the surface heat flow at old ages. It is interesting to note that heat production within the DM is ∼10% of continental mantle estimates (Chapter 3). The effect of oceanic heat production on bathymetry and heat flow is very small, ∼5 m and 0.5 mW/m2 , respectively. Heat production within the oceanic lithosphere is a third-order effect and can be ignored when modeling the thermal evolution of the oceanic lithosphere. 95 5.5.6.4 Hydrothermal Alteration Aside from its influence on concentrations of radiogenic elements, hydrothermal alteration of the seafloor has been ignored. Hydrothermal alteration may affect 5–15% of the lower crust and is likely more pronounced in the upper crust. Amphiboles and phylosilicates are the most common alteration products [Carlson and Miller , 2004]. Within the mantle, serpentinization has been estimated to be as high as ∼20% [Schmidt and Poli , 1998]. The reactions producing these products are exothermic and most commonly occur in young seafloor where thermal gradients are high. Some of the heat is presumably carried out of the system as seawater fluxes through. This is a complex process and it may be difficult to estimate the total heat imparted to the lithosphere. Therefore, I choose to ignore this process. The effect of alteration on thermophysical properties influences both the efficiency of heat transport and buoyancy within the upper oceanic lithosphere. Except for density, the required physical properties of many alteration minerals are poorly understood, in particular their P –T dependence. As laboratory studies improve estimates of physical properties for alteration products oceanic cooling models may be improved. 5.5.6.5 Adiabatic Gradient Often quoted adiabatic gradients for the mantle range from 0.2 to 0.5 K/km. A comparison between my preferred model and a model with no adiabatic gradient results suggests models that do not include a gradient under-predict heat flow by ∼1 mW/m2 and under-predict bathymetry by ∼25 m. Thus the effect is well within the uncertainty and variability of the datasets and can be ignored. 5.5.6.6 Constant Melt vs. Variable Melt Fraction Melt fraction varies with time beneath a cooling plate. The predicted bathymetric profile is only affected if the melt fraction decreases to a level such that the asthenosphere can support loads for a significant period of time. Afonso et al. [2007] explore the variability in and effect of melt fraction on plate cooling models. This effect is beyond the scope of this study. 96 5.6 Conclusions Recent plate cooling models of the oceanic lithosphere have focused on incorporating selected physical effects. Because each study uses a different set of observational constraints, it is difficult to compare directly the relative importance of each complexity beyond simple analytical solutions. In this study I develop a standard bathymetry–age and heat flow–age dataset that can be used to compare plate cooling models and estimate the average thermal structure of the oceanic lithosphere. Calculated misfits from >2600 models of bathymetry and heat flow–age data suggest crustal thickness and potential temperatures form a misfit trough roughly parallel to crustal thicknesses estimated from chemical melting models of MORB source at different temperatures. By incorporating P –T dependence of polymineralic assemblages, heat production, and assuming a 7 km thick crust, the best-fitting maximum plate thickness is ∼90 km and has a potential temperature of 1425◦C. Models using polymineralic rather than monomineralic assemblages yield only minor differences in subsidence and heat flow, ∼50 m and 1 mW/m2 , respectively. However, the inclusion of a crust is a considerably larger influence on plate models, particularly at young ages where the crust represents a dominant fraction of the total lithospheric thickness. Incorporating effective expansivity rather than volumetric expansivity producing the largest influence on subsidence, reducing subsidence by several hundred meters but produces a very small change in heat flow estimates. The title of this chapter poses the question whether or not these complexities are necessary for plate cooling models. If one only wishes to model subsidence and bathymetry, the answer is no. Simple one-dimensional cooling models with constant thermophysical properties like Stein and Stein [1992] fit the data well within the bounds of heat flow uncertainties and bathymetry variations. However, if one wishes to model state variables (pressure and temperature) or physical properties (density, specific heat capacity, thermal conductivity) properly, the answer is yes to most. The adiabatic gradient and radiogenic heat production produce very small effects and can easily be ignored. 97 Future improvements to plate cooling models should focus on several additional effects not incorporated into this study that may influence subsidence and/or heat flow. These processes include metamorphic reactions driven by hydrothermal circulation, thermal-cracking and crack-filling, and partial melt in the asthenosphere. 5.7 A Note on McKenzie et al. [2005] The formulation by McKenzie et al. [2005] solves the one-dimensional diffusion equation without sources, ∂ ∂T ∂ k [ρCP T ] = ∂t ∂z ∂z ! . (5.29) If an analytic solution to G= Z k(T )dT (5.30) exists, then Equation 5.29 can be written T ∂ (ρCP ) ∂T ∂2G + ρCP = . ∂t ∂t ∂z 2 (5.31) McKenzie et al. [2005] solves Equation 5.31 iteratively using the Crank-Nicholson method and adds the time derivative of ρ and CP as a perturbation to subsequent steps. Their finite difference approximation for the initial step (their Equation 11) is written, n+1 n+1 n n Tj−1 −Hkj+1 Tj+1 + 1 + 2Hkjn Tjn+1 − Hkj−1 = Tjn + 2H Gnj−1 − 2Gnj + Gnj−1 n n n n , −H kj+1 Tj+1 − 2kjn Tjn + kj−1 Tj−1 (5.32) where H is given by H= ∆t 2ρ Tjn CP Tjn ∆z 2 , (5.33) where ∆t and ∆z are the time step and depth spacing. The indicies n and i refer to time and to space, respectively. The terms with thermal conductivity, k, appear 98 to be incorrect. The reverse-engineered PDE from Equation 5.32 with the additional time derivatives of ρ and CP is T ∂ (ρCP ) ∂T ∂ 2 G ∂(kT ) + ρCP = − . ∂t ∂t ∂z 2 ∂z (5.34) By comparision between Equations 5.34 and 5.31, I can see that their finite-difference formulation, Equation 5.32, is incorrect. CHAPTER 6 CONCLUSIONS This study addresses the thermal states of the continental and oceanic lithosphere. An enhanced global heat flow database combined with recent advances in measuring thermophysical properties of minerals at high pressures and temperatures, new compilations of lithospheric heat production, and a better understanding of convective heat loss in young sea-floor collectively permit a reexamination of current concepts regarding the thermal state of the outer 300 km of the Earth. 1. An updated global heat flow database is produced, incorporating ∼40,000 new records to the existing database, which now includes >60,000 records and >44,800 heat flow determinations. The database will be shared with the Earth science community. I anticipate considerable use of the database, in part because heat flow serves as the prime constraint in calculating temperatures through the lithosphere. This database also presents the opportunity to determine heat loss of the solid Earth to improved accuracy. 2. I develop and reassess a number of thermo-geophysical reference models of the continental lithosphere. An important part of this development is the establishment of a reference lithospheric heat production model for North America. Heat production in the lower crust estimated from exposed granulite terranes is ∼0.4 µW/m3 and geochemical estimates of heat production from xenoliths places median subcrustal lithospheric heat production at ∼0.02 µW/m3 . To estimate upper-crustal heat production I use compositionally normalized elevation and a field of allowable temperatures from xenolith P –T estimates. Geotherm models of the continental lithosphere best fit these data when, on average, 26% of the surface heat flow is derived from upper-crustal heat production. 100 Using this geotherm model, estimated surface heat flow in Precambrian regions 37-47 mW/m2 , consistent with most observations. 3. Reference models for cooling of the oceanic lithosphere typically ignore heat flow as a constraint, particularly for young sea-floor. Heat flow through young sea-floor is generally considered strongly biased to low values as a result of hydrothermal circulation. Global filtering out of heat flow sites with thin sedimentary cover (<325 m) and in proximity to seamounts (<85 km) brings heat flow–age relationships much closer to theoretical predictions for cooling lithosphere. However, a heat flow deficit still persists. Adjustments for estimated thermal rebound and sedimentation effects on heat flow further reduce deficits, bringing data in line with model estimates for regions >25 Ma. Heat flow–age estimates at younger ages are improved by examining in detail several localities with high density sampling and co-located seismic data. 4. The filtered heat flow data are used to construct plate cooling models of the oceanic lithosphere which predict a maximum plate thickness of 90 km and mantle potential temperature of 1425◦C. Model results indicate that inclusion of a crust results in an improvement compare with the models lacking a crust that are typically computed. These models incorporate several physical processes acting on the cooling lithosphere, but may be refined further by incorporating additional chemical (e.g., hydrothermal alteration) and physical processes (e.g., melting and thermal-cracking). 5. Data excluded from the heat flow–age analysis that come from areas with relatively thin sediment cover and/or sites within 85 km of an identified seamount can be used to estimate the heat loss due to hydrothermal circulation of seawater through the oceanic upper crust. Using a Monte Carlo analysis to incorporate uncertainties in heat flow, I estimate advective heat loss to be 7±2 TW, or ∼20% of the total oceanic heat loss. This result is smaller than previous heat flow estimates due to a reduction in the estimated area affected by hydrothermal 101 circulation. This estimate can be used to help constrain total fluid fluxes through the oceanic lithosphere. 6. One of the eventual goals started during the course of this dissertation is a revised estimate of global heat loss. Despite the significant improvement in spatial coverage of this global heat flow database update, large regions of the planet remain unsampled. To estimate global heat loss, heat flow in unsampled regions must first be estimated. The thermal models developed within this dissertation serve as predictors for unsampled regions. Future work will include updating additional heat flow estimators on continents such as heat flow–age and –lithology relationships. By developing a reference model of lithospheric heat production and combining it with global heat flow, it is possible to estimate the total lithospheric heat production. By incorporating estimates of radiogenic abundances within the bulk silicate Earth, and using lithospheric estimates of heat production, it may be possible to help constrain the Urey ratio. The Urey ratio is the ratio of radiogenic heat production to the total convective heat flow within the mantle, which has important implications for the initial thermal state of the Earth. This implication is one of the exciting avenues for research opened by this study as well as exploring the myriad of implications of these models. APPENDIX A RADIATIVE DIFFUSIVITY AND CONDUCTIVITY OF OLIVINE REVISITED A.1 Abstract Olivine comprises >60% of the upper mantle, making it the dominant control on heat transport through the lithosphere. Accurate estimates of the thermal diffusivity and thermal conductivity are important in modeling geotherms, the thickness of the lithosphere, depth of melt generation, etc. Diffusivity and conductivity result from lattice vibration and radiative transport of thermal energy and depend on temperature. In this study, a new model for radiative diffusivity and conductivity is based on prior measurements of effective and lattice diffusivity on olivine samples with Mg# ∼0.9. Diffusivity can be expressed as T − 1140 Dr = 0.133 1 + erf 370 , (A.1) and conductivity as kr = 0.56 1 + erf T − 1150 370 , (A.2) with misfits of 0.024 mm2 /s and 0.086 W/m-K, respectively. A geotherm estimate using this model is consistent with geophysical estimates of lithospheric thickness of the Kalahari craton, but the geotherm may be slightly cool relative to xenolith P –T conditions. Conductivity estimates using this method appear to be inconsistent with recent spectral estimates of radiative conductivity, which also depend significantly on grain size. Additional experimental data are required to resolve the disagreement between the two models. 103 A.2 Introduction Olivine is the most common mineral in the upper mantle, composing >60% by weight. Therefore, understanding the thermal transport properties of olivine is important to accurate models of lithospheric conduction and upper mantle convection. Diffusivity and conductivity affect heat production within the lithosphere, the thickness of the mantle lithosphere, sublithospheric heat flow, estimates of heat transport during convection, depth of melt generation, and thermal isostatic effects [Dubuffet et al., 1999; Michaut et al., 2007; Hasterok and Chapman, 2007a]. Thermal diffusivity and conductivity result from lattice and radiative components, the latter of which is poorly known and difficult to measure accurately [Schatz and Simmons, 1972]. Several studies have estimated radiative conductivity using spectral methods (Shankland et al. [1979] and references therein), but may suffer from resonance related to similarity between laser frequency and Si-O stretching modes in silicates [Hofmeister , 1999a]. Conductivity values from these studies range from near zero at room temperature to ∼2.8 W/m-K at 1100–1700 K. Radiative conductivity may also be estimated by subtracting the lattice conductivity from measurements of effective conductivity [Hofmeister , 1999a], hereafter referred to as a difference estimate. The difference estimate reported by Hofmeister [1999a] is considerably lower and reaches little more than 11% of the spectral estimates at 1700 K. The estimate by Hofmeister [1999a], in widespread use among modelers (e.g., Dubuffet et al. [1999], Anderson [2000], van den Berg et al. [2001], McKenzie et al. [2005], Michaut et al. [2007] and Korenaga and Korenaga [2008]), was not converted from diffusivity to conductivity as reported. In this study, I develop a radiative diffusivity and conductivity model using the difference method and propose a new functional form to fit radiative conductivity that has several advantages over standard polynomial regression. Geotherms computed using this and other radiative conductivity models are assessed by comparison to with xenolith P –T conditions and geophysical estimates of lithospheric thickness beneath the Kalahari craton. 104 A.3 Estimating Radiative Diffusivity Heat transfer through minerals occurs by two physical mechanisms, phonon-phonon interaction (lattice transfer) and black body radiation as photons (radiative transfer) [Shankland et al., 1979]. Numerous experimental studies have measured the effects of pressure and/or temperature on conductivity and diffusivity of olivine [Schatz and Simmons, 1972; Schärmeli , 1982; Katsura, 1995; Chai et al., 1996; Harrell , 2002; Xu et al., 2004]. The results of these studies show similar temperature derivatives at low temperatures with scatter in absolute values resulting from individual sample characteristics and from inclusion or exclusion of radiative conductivity. I focus on two diffusivity studies, Katsura [1995], which includes both radiative and lattice components, and Xu et al. [2004], which is controlled to measure the lattice contribution only. Both studies measure diffusivity on olivine of approximately Fo-90 composition. The diffusivity data from Katsura [1995] are shown in Figure A.1 as a function of temperature at four pressures (0.1, 3, 6 and 9 GPa). To estimate the radiative component of diffusivity, the estimated lattice-only contribution is subtracted from the effective diffusivity. The lattice-only model is estimated using the model by [Xu et al., 2004] with an additional multiplicative factor tailored to fit the low-temperature effective diffusivity. Differences between the effective conductivity data and latticeonly data at low temperatures most likely result from sample specific characteristics. This adjustment does not violate the mathematical model presented by Xu et al. [2004] to fit lattice diffusivity contributions, whereas an additive adjustment would. The constants used to reproduce the lattice contributions in Figure A.1 are 0.90, 1.11, 1.01 and 1.09, from low to high pressures respectively. Radiative diffusivity estimates derived from the difference of the Katsura [1995] data and model lattice contribution are small near room temperature and become observable above 800 K. By 1500 K the radiative contribution represents >40% of effective diffusivity at low pressures and ∼30% at high pressures. Radiative estimates for 6 and 9 GPa reach a plateau in the conductivity at high temperatures with a maximum radiative diffusivity of 0.22-0.24 mm2 /s. There appears to be a slight 105 Diffusivity [mm2/s] 1.5 0.1 GPa 3 GPa 6 GPa 9 GPa 1.0 0.5 0.0 Diffusivity [mm2/s] 1.5 1.0 0.5 0.0 500 1000 1500 Temperature [K] 500 1000 1500 Temperature [K] Figure A.1. Thermal diffusivity of olivine. Effective thermal diffusivity from Katsura [1995] (squares) measured at (a) 0.1, (b) 3, (c) 6, and (d) 9 GPa. Lattice diffusivity estimates from Xu et al. [2004] scaled to fit effective diffusivity at 500–600 K (heavy grey line). Radiative conductivity estimates at each pressure (circles) are shown with the preferred model (black line). 106 decrease in the overall radiative diffusivity with pressure but uncertainties in the experimental data a pressure effect is inconclusive. A.4 Model for Radiative Diffusivity/ Conductivity The radiative contribution is negligible at room temperature but depends strong-ly on temperature (∝ T 3 ). However, the absorption spectra show that this process need not be linear. Hofmeister [1999a] uses a third-order polynomial to fit radiative conductivity data. There are several reasons that a third-order polynomial is an undesirable fitting function. First, it can be difficult to fit data at low temperatures where the radiative contribution remains near zero. Polynomial fits may dip below zero or may be strongly influenced by noise such as the anomalously high conductivity at low temperatures. Second, extrapolation beyond the measured bounds can lead to unrealistic increases or decreases in estimated radiative component. I suggest a fitting function that can be tailored to be positive for all temperatures, near zero at low temperatures, and be constant at extrapolated temperatures. An error function formulation is used to approximate the temperature effect on radiative diffusivity by, 1 T − Tm Dr = Dmax 1 + erf 2 a , (A.3) where a is an empirically derived constant, Dmax is the maximum diffusivity and Tm is the temperature at 0.5Dmax . The resultant fit for the average radiative diffusivity is shown in Figure A.2. The best fitting diffusivity model is Dmax = 0.266 mm2 /s, a = 0.0027 K−1 and Tm = 1140 K, with an overall RMS misfit 0.024 mm2 /s. Conductivity is related to diffusivity by k = DρCP , where ρ is the density and CP is the heat capacity. The average radiative conductivity in this study ranges from near 0 below 700 K to a little greater than 1.2 W/m-K at 1600 K. The conductivity can also be fit using an empirical model identical to that used to fit diffusivity. Best fitting conductivity parameters are kmax = 1.12 W/m-K, a = 0.0027 K−1 and Tm = 1150 K with a 0.086 W/m-K RMS misfit. 107 Radiative Diffusivity [mm2/s] 0.8 0.6 0.4 1 cm 0.2 0.1 cm 0.01 cm 0 500 1000 Temperature [K] 1500 Figure A.2. Radiative diffusivity models and data for olivine. Radiative diffusivity (circles) estimated from the difference of Katsura [1995] effective diffusivity measurements and Xu et al. [2004] lattice diffusivity model. Data from Shankland et al. [1979] and references therein (squares). Difference estimates of radiative conductivity for lherzolite [Gibert et al., 2003] (diamonds). Preferred model using error function fit to circles (heavy black line). Model by Hofmeister [1999a] (grey line). Radiative conductivity estimates for 0.01, 0.1 and 1 cm grain size from Hofmeister [2005] (dashed lines). A.5 A.5.1 Discussion Comparison With Other Estimates Radiative diffusivity and conductivity estimates using the lattice models by Chai et al. [1996]; Harrell [2002] yield similar results to mine. The estimated radiative conductivity by Hofmeister [1999a] is >50% lower than the model developed in this study. Upon further inspection, I determined that the widely used result by Hofmeister [1999a] represents the radiative diffusivity, not radiative conductivity. Figure A.2 compares of my result with other radiative diffusivity data. A difference estimate of radiative diffusivity can also be made from lherzolite sample BALM4 using effective diffusivity measurements by Gibert et al. [2003]. The BALM4 lherzolite sample has ∼76 modal percent olivine with a whole rock Mg# of 0.9. Radiative diffusivity estimates for the lherzolite sample show a measurable radiative effect above ∼700 K. Above ∼900 K, the radiative diffusivity reaches a nearly constant value of 0.7–0.9 mm2 /s (Figure A.2). The estimated radiative component, while lower for 108 lherzolite, has a functional form similar to that of olivine within measured uncertainty. The differences could be due to the behavior of additional minerals in the lherzolite system, partitioning of Fe between phases, and/or differences in grain size. The earliest diffusivity values based on spectral estimates suggest significantly higher radiative diffusivity [Shankland et al. [1979] and references therein], but are subject to resonance in the Si-O bonds as a result of the similarity between the laser frequency and fundamental oscillation modes [Hofmeister , 1999a]. More recent estimates based on spectral measurements show a strong dependence on radiative diffusivity and grain size as well as temperature [Hofmeister , 2005]. Grain sizes in the Katsura [1995] samples used for effective diffusivity measurements are ∼0.005 cm. While Hofmeister [2005] did not measure grain sizes smaller than 0.01 cm, the extrapolated trend between grain the sizes of 0.01 cm to 0.1 cm suggests that the radiative conductivity is negligible within the temperature range of the Katsura [1995] experiment. The grain sizes measured in the Gibert et al. [2003] sample range from ∼0.01–0.3 cm with most of the grain diameters near 0.1 cm. Hofmeister’s spectral model for 0.1 cm is well below difference estimates for the lherzolite sample below ∼1200 K. However, the spectral model grows rapidly above 1200 K, while the radiative diffusivity for the lherzolite sample appears relatively constant above 900 K. Recent studies have also suggested an overestimation of the Hofmeister [2005] radiative conductivity model [Goncharov et al., 2008, 2009]. Clearly the results from difference estimates and spectral methods are incompatible at present. A.5.2 Implications of Radiative Conductivity Models The thermal gradient within the lithosphere is sensitive to variations in thermal conductivity. Sensitivity of geotherms to the different radiative conductivity models discussed above is tested using approximate Kalahari Craton lithospheric parameters and present day surface heat flow of 47 mW/m2 [Nyblade et al., 1990]). Figure A.3 shows geotherms computed using thermal conductivities estimated using Xu et al. [2004] for the lattice component and radiative components from this study 109 0 (a) (b) Moho Adiabat Depth [km] 50 100 1 cm 150 0.1 cm 0.01 cm 200 This Study SS72 Kalahari Craton 250 0 500 1000 Temperature [K] 1500 2 3 4 Thermal Conductivity [W/m-K] 5 Figure A.3. Comparison of geotherms (a) computed from several radiative conductivity models (b). Effective conductivity model by Schatz and Simmons [1972] (SS72) incorporated into the commonly used Chapman and Pollack [1977] geotherms (heavy grey line). Models incorporating the radiative model from this study (heavy black line). Grain size models of 0.01, 0.1 and 1 cm from Hofmeister [2005] (dashed lines). Xenolith P –T estimates (circles) for peridotite xenoliths from the Kalahari Craton [Stiefenhofer et al., 1997; Rudnick and Nyblade, 1999; Saltzer et al., 2001; Bell et al., 2003b; Grégoire et al., 2003; James et al., 2004; Katayama et al., 2008]. 110 and the 0.01, 0.1, and 1 cm grain size models by Hofmeister [2005]. The commonly used Chapman and Pollack [1977] geotherm using the effective conductivity model by Schatz and Simmons [1972] is also shown for comparison. All computed geotherms have the same temperature at the Moho and diverge below the Moho as a result of conductivity differences. Temperatures at 150 km vary by 256◦ C between the coldest (1 cm grain size) and hottest geotherms (0.01 cm grain size) with a corresponding conductivity range of ∼1.75 W/m-K. Depths of intersection to the 1300◦C adiabat range from 162 to 236 km. Lithospheric thickness estimates based on various seismic tomography techniques range from 150–300 km with several around 200 km (Begg et al. [2009] and references therein). Recent receiver function estimates for the Kaapvaal Craton are on the shallow end at 160 km. Magnetotelluric estimates of lithospheric thickness in the Kimberley region of the Kaapvaal Craton range from 190–220 km [Muller et al., 2009]. These geophysical estimates are consistent with the depth to the adiabat obtained using the radiative conductivity model in this study (197 km) and <1 cm grain size models. While seismic and electrical methods can be used to estimate lithospheric thickness, temperatures estimated using these techniques are uncertain. Xenoliths from kimberlite eruptions on the Kalahari Craton provide estimates of P –T conditions. These equilibrium conditions are believed to represent a near steady-state geotherm because of a lack of significant tectonic activity in the region for more than a billion years preceding the eruption (∼100 Ma). The geotherm for 1 cm grain size is significantly cooler than xenolith estimates, while the 0.01 cm model is on the hot end of xenolith conditions. The Schatz and Simmons [1972] estimate bounds the cool end, but is more curved than indicated by the bulk of the xenolith estimates. The geotherm using the radiative conductivity model from this study also appears cool, but is significantly less curved than the Schatz and Simmons [1972] model. The geotherm estimated for a 0.1 cm grain size appears to fit best the bulk of the xenolith data. Measured grain sizes of south African xenoliths range from 0.4–1.4 cm [Ave Lallemant et al., 1980] and appear inconsistent with the radiative models based on grain size. The fits of each of these geotherms can be improved by tailoring upper crustal 111 conductivity and heat production, but geotherms with higher grain sizes remain difficult to fit. A polymineralic model for radiative conductivity derived from the lherzolite data is likely to provide a better fit to the data because of the more modest radiative contribution. A.6 Conclusions Thermal conductivity and diffusivity are important thermophysical parameters required for calculating geothermal temperatures and geodynamic modeling. At high temperatures, thermal conductivity/diffusivity combine lattice and radiative mechanisms of heat transfer. A new radiative conductivity estimate is derived and compared with other radiative estimates. Radiative diffusivity can be estimated by, T − 1140 Dr = 0.133 1 + erf 370 , (A.4) with an RMS misfit of 0.024 mm2 /s, and a radiative conductivity estimated by, kr = 0.56 1 + erf T − 1150 370 , (A.5) with an RMS of 0.086 W/m-K. Radiative diffusivity estimates based on spectral methods show a significant grain size dependence as well as temperature, but appear to be inconsistent with difference estimates. Geotherms computed using grain sizes consistent with southern African xenoliths are too cool compared to estimated xenolith P –T conditions for the Kalahari craton. Temperatures estimated using the radiative conductivity model from this study are consistent with geophysical estimates of lithospheric thickness but may also be too cool and be too curved to fit xenolith data accurately. An estimate of radiative diffusivity within a single lherzolite sample suggests a smaller contribution to effective diffusivity than from equivalent grain size diffusivity estimates or monomineralic olivine aggregate results from this study. Additional experimental results are necessary to resolve the discrepancies between spectral and difference estimates and grain size contributions to diffusivity/conductivity. APPENDIX B THERMAL CONDUCTIVITY OF AMPHIBOLES: INFLUENCE OF COMPOSITION B.1 Abstract Amphiboles are a major rock forming mineral group and a reservoir of H2 O in the lithosphere that constitutes >15% of some igneous rocks and >50% of some metamorphic rocks, as well as small concentrations in hydrous mantle xenoliths. Hence it is important to understand its physical properties in order to properly model evolution of the lithosphere. In particular, this study examines the thermal conductivity dependence on amphibole composition. A linear-least squares approach is used to develop an empirical relationship between cation concentrations and thermal conductivity of amphiboles. Overall misfit between the empirically derived relationship and measured conductivities is 0.32 W/m-K or better with a maximum misfit of <0.8 W/m-K. Coefficients in this empirical relationship suggest that incorporation of Na and Al into the amphibole structure tends to reduce conductivity, whereas Mg and Ca tend to enhance the conductivity. While this method is promising for estimating thermal conductivity in amphiboles, it would greatly benefit from the measurements of additional compositions. This formulation should be applied with caution to non-end-member compositions. B.2 Introduction Amphiboles are a major mineral group found in crustal igneous and metamorphic rocks and are occasionally found in mantle xenoliths. Abundances of amphiboles in intermediate composition igneous rocks can be >15% and range up to nearly pure 113 hornblende (hornblendite). Amphiboles are alteration products in the oceanic crust that can reach ∼8% (locally in fracture zones >20%) of the crust [Carlson and Miller , 2004] and also are found in concentrations of >50% of some crustal metamorphic rocks. Crustal cross-sections derived from surface exposures suggest that amphibolitefacies represent approximately 10 km of the mid-crust in the Pikwitonei and Sachigo region of the Superior Craton [Fountain et al., 1987]. In magmatic arcs amphiboles may be found in concentrations greater than several tens of percent and have a large influence on the arc H2 O budget, capturing potentially 20% of water ascending in mantle-derived melts [Davidson et al., 2007]. In addition to the crust, amphibole is a major storage mineral for H2 O within the mantle [Lamb and Popp, 2009]. Because H2 O can have a disproportionately large influence on thermophysical properties, a few percent amphibole may likewise have a large influence on lithospheric dynamics. It is therefore important to understand the dependence of thermophysical properties on amphibole compositions when modeling the current state and evolution of the crust and upper mantle. Particularly important in lithospheric evolution is the thermal regime and, hence, thermal conductivity. Measurements of thermal conductivity of single amphibole crystals and monomineralic aggregates of ∼12 distinct compositions were made by Horai and Simmons [1969]; Horai [1971]; Diment and Pratt [1988]. Conductivity of amphiboles in these studies ranges from 2.2 W/m-K in glaucophane to 5.2 W/m-K in tremolite. While several compositions have been measured, given the large array of possible compositions, a relatively small fraction of compositions have been examined. Therefore, it is desirable to develop a method to predict the thermal conductivity of an amphibole by its composition. In this study, I develop a relationship between thermal conductivity to amphiboles of known composition that may be useful in predicting a wider range of compositions than has been currently studied. Unfortunately the compositions of amphiboles analyzed by Horai and Simmons [1969]; Horai [1971]; Diment and Pratt [1988] are not precise enough to develop a reliable model of thermal conductivity. Therefore, the 114 results presented in this work are preliminary and represent a proposal for a possible method of estimating the conductivity of amphiboles based on composition. B.3 Samples Amphiboles have a wide array of compositions. The general formula for amphibole is AB2 C5 T8 O22 (OH)2 [Leake and 21 others, 1997]. The A site can generally be filled with Na or K, or can be vacant. The B and C sites can have a wide variety of substitutions. In this paper, I focus on amphiboles with A, B and C sites filled with Na (A or B), Ca (B only), Mg, Fe2+ , Fe3+ (C only) and Al3+ (C or T). Conductivity measurements were performed on 20 individual amphiboles with 11 distinct compositions by Horai and Simmons [1969]; Horai [1971] using a needle probe on powdered samples. The compositions were analyzed by x-ray techniques, but were unfortunately not precisely reported so composition may deviate somewhat from reported values. Another 6 samples cut from monomineralic aggregates were measured on a divided bar [Diment and Pratt, 1988]. The compositions of these samples are somewhat less certain than those measured in the studies by Horai and Simmons [1969]; Horai [1971]. The thermal conductivity of 12 amphiboles and their general compositions used in this study given in Table B.1 . Conductivities range from 2.17 W/m-K for glaucophane to 4.66 W/m-K for tremolite with sodic and aluminous amphiboles having lower conductivities than those without. Average values for samples of similar composition were used rather than individual measurements to prevent bias from any particular composition. For tremolite, I excluded the low measurement of 2.76 W/m-K when computing the average of 4.66±0.51 W/m-K, which has a significantly lower standard deviation than when included (4.38±0.86 W/m-K). Actinolite also displays a large range in conductivity 2.15-4.52 W/m-K, which may result from a deviations from reported compositions and/or as a result of uncertainties arising from the method used to measure conductivity. Table B.1. Amphibole compositions and conductivity. Mineral 1. Anthrophyllite 2. Cummingtonite 3. Grunerite 4. Tremolite 5. Actinolite 6. Ferro-actinolite 7. Hornblende 8. Barkevikite 9. Glaucophane 10. Magnesio-riebeckite 11. Riebeckite 12. Richterite Chemical Formula Mg7 (Si8 O22 )(OH)2 Mg3.5 Fe3.5 (Si8 O22 )(OH)2 Fe7 (Si8 O22 )(OH)2 Ca2 Mg5 (Si8 O22 )(OH)2 Ca2 Mg2.5 Fe2.5 (Si8 O22 )(OH)2 Ca2 Fe5 (Si8 O22 )(OH)2 Ca2 Mg3 FeAl(Si7 AlO22 )(OH)2 3+ NaCa2 Fe2+ 2 MgFe Al0.5 (Si6.5 Al1.5 O22 )(OH)2 Na2 Mg3 Al2 (Si8 O22 )(OH)2 Na2 Mg3 Fe3+ 2 (Si8 O22 )(OH)2 2+ Na2 Fe3 Fe3+ 2 (Si8 O22 )(OH)2 Na2 CaMg5 (Si8 O22 )(OH)2 Nb 1 1 2 7 (6)c 3 1 5 1 1 1 2 1 Conductivitya [W/m-K] 3.96 3.60 3.36 4.66 3.48 3.99 2.74 2.40 2.17 3.63 3.03 3.03 a Data from Horai and Simmons [1969], Horai [1971], and Diment and Pratt [1988]. b N is the number of conductivity samples. c Only 6 samples were used to compute the average for tremolite, excluding a low value of 2.76 W/m-K. 115 116 B.4 Estimating Conductivity Amphibole conductivity, λest , is estimated using the following equation: λest = X ni ki , (B.1) where ni and ki are the number of moles and conductivity contribution from cation i, respectively. The conductivity contributions, ki , are given by Nk = λ, (B.2) where λ is a vector of measured conductivity and N is a matrix defined by ni,j , with columns representing cations from each individual amphibole composition in rows. The linear-least-squares solution is given by, k = (NT N)−1 NT λ. B.5 (B.3) Results and Discussion The results of estimated conductivity coefficients for each cation are summarized in Table B.2 and the compositionally estimated conductivities are shown in Figure B.1. Conductivity estimated using Na, Ca, Mg, Fe, Al, and Si yields a misfit of 0.32 W/m-K. Vacant sites are not treated as a separate element in this analysis as they have a negligible effect on the result. Unlike the other cations used in the analysis, Fe can be partitioned between Fe2+ and Fe3+ and hence may behave differently. By partitioning Fe into different oxidation states, the results yield a decrease of the RMS to 0.26 W/m-K. In both cases mentioned above, the misfits of all samples are within 0.8 W/m-K, with the misfits of 67% of the samples within 0.4 W/m-K (approximately 1σ of a typical thermal conductivity measurement). Some cations in many minerals can fit into different sites, amphiboles included, and as a result have differing influences on physical properties. For example, Mg in anthrophyllite sits in both the B and C sites (Table B.1). I decided to focus only on Al3+ substitution between the C and T sites because Al has a significantly larger ionic radius than Si4+ . Adding Al substitution further lowers the RMS to 0.21 W/m-K Table B.2. Fitting constants for estimating conductivity coefficients Misfit 0.32a,c 0.26a,d 0.21a,e 0.29b,c 0.22b,d 0.07b,e Na -0.1769 -0.4640 -0.3626 -0.2366 -0.5256 -0.4095 Ca 0.2014 0.1394 0.5698 0.2296 0.1675 0.7576 Mg 0.1064 0.0431 0.4245 0.0428 -0.0213 0.4714 Fe2+ Fe3+ -0.0554 0.3118 0.3144 0.9439 Fetotal 0.0496 C Al 0.2794 T Al -0.9150 -0.0101 -0.1161 0.3588 0.2555 1.0847 Altotal -0.4323 -0.3966 -0.5030 -0.4677 0.4203 -1.1908 Si 0.3918 0.4603 0.1328 0.4458 0.5150 0.0918 All table values have units of W/m-K. a Fit using all amphibole compositions. b Excluding actinolite. c No partitioning or substitution. d Fe partitioned between Fe2+ and Fe3+ . e Fe partitioned between Fe2+ and Fe3+ and Al partitioned between C and T cation sites. 117 118 Conductivity [W/m/K] 6 (a) Na 4 3 2 1 Misfit w/ actinolite Ca 5 RMS Al 0.32 W/m/K 0.22 W/m/K, partitioned 0 −1 Conductivity [W/m/K] 6 (b) 5 Na 4 3 2 1 Misfit w/o actinolite Ca RMS Al 0.29 W/m/K 0.07 W/m/K, partitioned 0 −1 1 2 3 4 5 6 7 8 9 10 11 12 Figure B.1. Thermal conductivity of amphiboles. Amphiboles containing Ca, Na and Al are illustrate by the labeled horizontal bars. Numbers correlate to samples in Table B.1. Circles represent natural samples with error bars representing 10% of the measured thermal conductivity representing typical measurement uncertainty, rather than absolute error. Misfits are computed by difference of the experimental and estimated conductivities. (a) estimated thermal conductivity using all samples: squares-no partitioning or substitution; diamonds-partitioning of Fe between Fe2+ and Fe3+ and substitution of Al in C and T sites (see text). (b) same as (a) excluding actinolite. 119 (Figure B.1a). Conductivity estimates using partitioning of Fe and substitution of Al fit nearly all amphiboles well within approximated uncertainty, with the majority of misfit in the tremolite–ferroactinolite solid-solution series. If actinolite is excluded from the analysis (see reasoning below), the RMS is slightly improved in the unpartitioned case (RMS = 0.29 W/m-K) and significantly when Fe and Al are partitioned (RMS = 0.07 W/m-K, see Figure B.1b). However, when fitting 11 independent compositions with 8 parameters, there is a fear of over-fitting the data, especially when Al substitution alone does not improve the fit over the unpartitioned formulation. More conductivity measurements on an increased number of end-member compositions are clearly needed to resolve this issue. Actinolite has a lower thermal conductivity than either tremolite or ferro-actinolite. This is unsurprising as a number of other minerals with solid solution series, such as plagioclase and olivine show a quadratic relationship between end-members with a minimum occurring at an intermediate composition (Figure B.2). The pyropealmandine and andradite-grossular series in garnets may also exhibit a quadratic behavior between end-member compositions [Giesting and Hofmeister , 2002]. Given the uncertainties in currently measured actinolite series compositions and conductivities it is difficult to reliably model intermediate compositions. Additional data are needed to model conductivities in the tremolite-ferroactinolite as well as within other amphibole solid solution series where no measurements exist. While the potential nonlinear intermediate low in conductivity reduces the usefulness of this method in predicting conductivity of intermediate compositions, endmembers may still be reliably estimated using this method. And despite this potential shortcoming, it can still yield a reasonable conductivity estimates when additional data are not available. Fitting parameters are given in Table B.2 show the effect of each cation on estimated amphibole conductivity. While the values of each constant change significantly in each case, some patterns emerge. Ca and Si are a positive influence on conductivity. However, when estimated effect of Si on conductivity is large, Ca is small and vice versa. Mg likewise has a similar relationship with the Si coefficient, but is not strictly 120 Conductivity [W/m/K] 6 5 olivine actinolite series 4 3 plagioclase 2 1 0 20 40 60 Fo/An/Tr 80 100 Figure B.2. Thermal conductivity of solid solution minerals. Olivine as a function of forsterite (Fo) content and plagioclase as a function of anorthite (An) content. The amphibole tremolite-ferroactinolite series is shown for comparison although intermediate composition is approximate. Error bars are the standard deviation from multiple samples. Olivine data from Horai [1971], Chai et al. [1996], and Harrell [2002]. Plagioclase data from Petrunin et al. [2004] and references therein. positive. Estimated conductivity is always reduced by Na. And while Al in the C site has a positive influence on conductivity, the overall effect of Al likely reduces the overall conductivity as the T site coefficient is strongly negative. Fe is more complex with no clear pattern relative to other ions. B.6 Implications To examine the influence of thermal conductivity variations I compute 1-D-steadystate temperature differences through a 10 km thick rock layer as a function of mole fraction amphibole with compositions ranging from hornblende to tremolite (Figure B.3). Heat production in the layer is assumed to be a constant at 0.6 µW/m3 , a typical value for amphibolites. Heat production has a relatively small effect on the results except when values are high, but unlikely on regional scales. Heat flow, however, strongly affects the results. At a heat flow of 20 mW/m2 , the temperature difference arising from amphibole composition are quite small at the base of the layer, even with high amphibole content. At 40 mW/m2 the temperature difference computed for tremolite to hornblende is slightly greater than 10◦ C at mole fractions similar to tonalitic composition (∼10%). 121 0.8 20 40 Fraction Amphibole 20 1 0.6 0.4 0.2 Heat Flow = 20 mW/m2 40 mW/m2 80 40 60 60 0.8 20 0.6 20 40 Fraction Amphibole 0 1 0.4 0.2 60 mW/m2 0 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 Fraction Tremolite, 1 - Fraction Hornblende 80 mW/m2 0.6 0.8 1 Figure B.3. Influence of amphibole composition on lithospheric temperatures. Temperature differences for a 10 km thick layer with varying amphibole content and composition. Radiogenic heat production is assumed constant at 0.6 µW/m3 . All temperature differences are computed relative to amphibole composition of 100% tremolite. Contour interval is 10◦ C. Each panel represents a different heat flow at the top of the layer. 122 At a mole fraction of 50% (amphibolite composition), the temperature difference is 34◦ C for pure hornblende relative to tremolite, but still small for a mole fraction of 50% tremolite (15◦ C). When values of heat flow at the top of the layer are high, temperature differences are significantly higher for mole fractions less than tremolite ∼0.4 with modest quantities of amphibole (>20%). At high heat flow, temperature differences at the base of a 10 km thick layer of 15% amphibole could reach as high as 25◦ C; high enough to possibly effect thermochronology estimates of closure depth and hence exhumation rate. If the layer thickness is greater than 10 km, the temperature differences shown in Figure B.3 will be larger. However, because heat production is included, temperature differences do not scale exactly 1:1, but are close enough that the temperature difference at 20 km is ∼5% less than double the 10 km estimate. In general, significant quantities of amphibole at modest to high heat flow are required to observe significant temperature differences due to from amphibole composition at the base of a 10 km thick layer. Thermal conductivity variations due to amphibole composition may observably change temperatures in arc terranes where heat flow is high and rocks contain a high fraction of amphibole, but are unlikely to be much of an influence in shield regions where heat flow is typically low. B.7 Conclusions Amphiboles are a major rock forming mineral within the crust and a major reservoir of H2 O in the lithosphere. Thus understanding the thermophysical properties of amphibole are important to modeling lithospheric evolution. This study examines the compositional dependence on the thermal conductivity of amphibole. Conductivity ranges from 2.2–4.7 W/m-K resulting from compositional variations and possibly from site-specific substitutions of chemical species within the amphibole structure. Si and Ca have positive influences on conductivity while incorporation of Na and Al typically reduce the conductivity. Conductivity of amphiboles can be estimated by λest = −0.237nNa + 0.230nCa + 0.043nMg (B.4) 123 −0.010nFe − 0.503nAl + 0.446nSi . to with a maximum misfit of 0.5 W/m-K and RMS of 0.3 W/m-K. By partitioning Fe between Fe2+ and Fe3+ as well as treating Al in the C and T cation sites separately misfit can be reduced to 0.1 W/m-K with a maximum misfit of 0.1 W/m-K using the resulting equation: λest = −0.410nNa + 0.758nCa + 0.471nMg (B.5) +0.359nFe2+ + 1.085nFe2+ + 0.420nC Al −1.191nT Al + 0.092nSi. Complexities in the conductivity pattern of intermediate composition amphiboles may be better predicted by a quadratic with a minimum inside the end-member compositions, thereby reducing the accuracy of this method. Temperature differences modeled using variations in amphibole composition show sensitivity to surface heat flow. Arc terranes are likely to have the greatest variation in temperature estimates due to amphibole conductivity because of their high heat flow and high amphibole content. Amphibole composition may have little effect on temperature variations in shields where heat flow is low. Additional and compositionally more accurate measurements are needed to test for nonlinear behavior at intermediate compositions, and to further refine and expand this empirical method to a greater range of compositions. This method may also hold promise for other complex minerals groups with complex compositions such as micas and clays. APPENDIX C MODELS OF THERMAL CONDUCTIVITY FOR INDIVIDUAL MINERALS This appendix includes figures of conductivity models for most of the minerals used in this dissertation. Conductivity between end-members for plagioclase, olivine, and garnet are computed using a second order polynomial by linear least squares inversion. Temperature dependence is computed using Newton’s inverse method for lattice conductivity or a joint inversion for lattice and radiative conductivity. Conductivity [W/m-K] 10 8 α-quartz β-quartz 6 keff 4 kL 2 kR 0 300 500 700 900 Temperature [K] 1100 Figure C.1. Effective thermal conductivity of quartz. Solid circles are data from Höfer and Schilling [2002] and open circles are the difference between the observed conductivity and lattice model. Radiative model is assumed the same for α- and β-quartz. Grey points are estimated from difference of observed conductivity and radiative model. Heavy, light, and dashed lines represent effective conductivity, keff , estimated lattice conductivity, kL , and estimated radiative conductivity, kR , respectively. 125 0.5 Conductivity [W/m-K] muscovite 0.4 keff 0.3 0.2 kL 0.1 kR 0 300 400 500 600 Temperature [K] 700 800 Figure C.2. Effective thermal conductivity of muscovite. Solid circles are data from Roy et al. [1981] and open circles are the difference between the observed conductivity and lattice model. Heavy, light, and dashed lines represent effective conductivity, keff , estimated lattice conductivity, kL , and estimated radiative conductivity, kR , respectively. 3 Conductivity [W/m-K] orthoclase keff kL 2 1 kR 0 300 500 700 Temperature [K] 900 1100 Figure C.3. Effective thermal conductivity of orthoclase. Solid circles are data from Höfer and Schilling [2002] and open circles are the difference between the observed conductivity and lattice model. Heavy, light, and dashed lines represent effective conductivity, keff , estimated lattice conductivity, kL , and estimated radiative conductivity, kR , respectively. 126 3.0 plagioclase Conductivity [W/m-K] (a) 2.5 2.0 1.5 1.0 0 20 40 60 % Anorthite (An) 80 100 3.0 Conductivity [W/m-K] (b) 0 2.5 20 100 2.0 55 1.5 An0 1.0 300 An20 An55 An100 400 500 Temperature [K] 600 700 Figure C.4. Lattice thermal conductivity of plagioclase. Composition data from Birch and Clark [1940], Sass [1965], and Horai [1971]. (a) conductivity as a function of anorthite at 0 GPa and 298 K. (b) conductivity as a function of temperature for four plagioclase compositions. Temperature data from Petrunin et al. [2004]. 127 4.0 Conductivity [W/m-K] orthopyroxene 3.5 3.0 2.5 2.0 200 600 1000 Temperature [K] 1400 Figure C.5. Lattice thermal conductivity of orthopyroxene. Data from Harrell [2002]. Conductivity [W/m-K] 7 (a) 6 olivine 5 4 3 2 1 0 20 40 60 % Forsterite (Fo) 80 100 Figure C.6. Lattice thermal conductivity of olivine. Composition data from Horai and Simmons [1969], Horai [1971], and Harrell [2002]. (a) conductivity at 0 GPa and 298 K as a function of the end-member forsterite. 128 Conductivity [W/m-K] 7 (b) 6 10 GPa 5 7 4 4 0 3 2 Conductivity [W/m-K] 1 7 6 (c) Fo0 Fo78 Fo91 5 4 3 2 1 200 400 600 800 1000 Temperature [K] 1200 1400 Figure C.6 continued. (b) pressure and temperature dependence of Fo90 . (c) conductivity estimates for three compositions using model established in (b) and data from Harrell [2002]. 129 Conductivity [W/m-K] 7 75 ≤ Py + Alm < 90 Py + Alm ≥ 90 Py + Alm ≥ 90 used in fit 6 (a) 5 4 3 2 1 0 20 40 60 % Pyrope (Py) 80 100 Conductivity [W/m-K] 7 (b) 6 5 4 garnet 8 4 0 3 2 1 300 500 700 900 Temperature [K] 1100 Figure C.7. Lattice thermal conductivity of garnet. (a) conductivity at 0 GPa and 298 K as a function of the end-member pyrope. Data from Giesting and Hofmeister [2002]. (b) pressure and temperature dependence of conductivity for garnet with composition Py25 Alm73 Sp1 Gr1 from Osako et al. [2004]. 130 Conductivity [W/m-K] 14 spinel 12 10 8 6 4 2 200 400 600 800 1000 Temperature [K] 1200 1400 Figure C.8. Lattice thermal conductivity of spinel. Data from Clauser and Huenges [1995]. APPENDIX D EMPIRICAL CONSTANTS USED TO MODEL PHYSICAL PROPERTIES Constants used for computing density, expansivity are given in Table D.1. Con- stants used in computing thermal conductivity are given in Table D.2. Amphibole composition is assumed to be hornblende in continental rocks (Chapter 3) and a mixture of pargasite, tremolite and ferroactinolite within the oceans (Chapter 5). Additional constants are needed for oceanic cooling models to compute heat capacity. The constants necessary to compute heat capacity are given in Table D.3. Table D.1. Physical properties and empirical constants for mineral end-members. Mineral alpha-quartz beta-quartz orthoclase albite anorthite phlogopite clinochlore daphnite tremolite ferro-actinolite pargasite hornblende diopside hedenbergite enstatite ferrosillite forsterite fayalite spinel hercynite pyrope almandine grossular w g mol−1 60.08 60.08 278.34 262.22 278.21 417.29 555.83 713.51 812.41 970.08 835.86 864.70 216.56 248.10 100.40 131.93 140.71 203.78 142.27 173.81 403.15 497.75 450.45 ρ298 kg m−3 2648 2530 2555 2620 2760 2788 2635 3343 2979 3430 3074 3248 3272 3651 3206 4003 3222 4400 3575 4264 3565 4324 3593 a0 × 105 K−1 1.4170 -0.4400 3.4000 1.9801 1.2491 5.8000 2.5 2.5 3.1310 3.1310 3.1310 2.0750 3.3300 2.9800 2.9720 2.8750 2.8540 2.3860 1.9600 0.9700 2.3110 1.7760 1.9510 a1 × 108 K−2 9.6581 a2 K -1.6973 1.0065 -0.0162 -0.9760 0.0161 1.0270 0.5711 1.0080 1.1530 1.6400 1.9392 0.5956 1.2140 0.8089 -0.3842 -0.0518 -0.4538 -0.5071 -0.4972 KT GPa 37.1 57 58.3 53.8 82.5 54 85 86.9 85 76 91.2 94 113 119 105.8 100 128 135 205 209 173 174 168 KT′ 5.99 4 4 6 3.2 7.8 3.3 4 4 4 4 4 4.8 4 8.5 8.8 4.2 4.2 4.1 4 5 6 5.5 γth 0.7 0.1 0.4 0.6 0.6 0.6 0.3 0.3 0.74 0.73 0.84 1.1 1 1.5 1.01 0.88 0.99 1.06 1.73 1.2 1.29 1.29 1.38 δT 8.42 4.11 4.44 6.57 3.47 4.55 4.3 4.3 4.74 4.73 4.84 5.1 6.04 5.21 9.39 9.05 5.19 5.26 6.5 5.19 5.3 5.52 4.57 Notes 1,2 1,2 1,2 1,2,8 1,2,8 1,2,3,4 1,12 1,12 1,2 1,2 1,2 1,2,3 1,2 1,2,3 1,5,9 1,6,10 1,2,11 1,2,11 1,7 1,2,3 1,2 1,2 1,2 Italicized values are estimated. 1 Densities and several bulk moduli from Hacker et al. [2003] and references therein. Thermal expansivity from Fei [1995] and references therein. 3 Reformulated expansivity from Holland and Powell [1998] and references therein. 4 Comodi and Zanazzi [1995]. 5 Jackson et al. [2003]. 6 Average of Yang and Ghose [1994] and Hugh-Jones [1997]. 7 Sueda et al. [2008]. 8 Average of Downs et al. [1994] and Angel [2004]. 9 Angel and Jackson [2002]. 10 Yang and Ghose [1994]. 11 Stixrude and Lithgow-Bertelloni [2005]. 12 Average of results from Pawley et al. [2002] and Welch and Crichton [2002]. 2 132 Table D.2. Empirical constants for estimating conductivity. mineral a-quartz b-quartz orthoclase plagioclase (an) biotite chlorite amphibole hornblende clinopyroxene orthopyroxene olivine (fo) spinel (sp) garnet (py) λ0 W/m-K 8.79 0.99 1.79 2.2 2.27 4.35 tr: 4.0654 2.65 4.25 3.37 3.09 11.94 4.97 λ1 W/m-K -2.18 λ2 W/m-K 1.9 fact: 3.7834 parg: 1.7066 -1.17 5.47 -7.42 3.35 8.71 8.45 n 1.48 -1.00 -0.27 -0.21 1.54 1.54 0.5 0.5 0.54 0.3 0.49 1.24 0.37 λRmax W/m-K 0.8107 0.8107 0.1949 0 0.1166 0.1166 0.1725 0.1725 0.1725 0.1725 0.1725 0 0 ω K−1 6.553e-3 6.553e-3 3.782e-3 TR K 558 558 864 2.749e-3 2.749e-3 3.900e-3 3.900e-3 3.900e-3 3.900e-3 3.900e-3 520 520 762 762 762 762 762 Notes 1,2 1,2 2 3,4,5,6 7,8 10 5,9,10,11,12 5,9,10,11,12 5,13,12,14 5,13,12 5,9,13,12 15,16 16,17,18 133 Lattice conductivity is computed as, λ◦ = λ0 + λ1 χ+ λ2 χ2 , where χ is the mole fraction of the end-member (an) anorthite, (fo) fosterite, (sp) spinel, and (py) pyrope. Amphibole conductivities are computed as a function of the end-members (tr) tremolite, (fact) ferro-actinolite, and (parg) pargasite. All other minerals do not have enough data from end-member compositions to estimated mixing behavior. Values in italics are estimated. 1 Höfer and Schilling [2002]. 2 Branlund and Hofmeister [2007]. 3 Birch and Clark [1940]. 4 Sass [1965]. 5 Horai [1971]. 6 Petrunin et al. [2004]. 7 Room temperature values [Clauser and Huenges, 1995] estimated as λeff = (2λk + λ⊥ )/3. 8 Radiative conductivity assumed same as muscovite. 9 Horai and Simmons [1969]. 10 Diment and Pratt [1988]. 11 Appendix B. 12 Radiative conductivity estimated from lherzolite Gibert et al. [2003]. See Hasterok [in prep.] for explanation. 13 Harrell [2002]. 14 Value for n estimated from augite rather than diopside sample from Hofmeister and Pertermann [2008]. 15 Spinel estimated from Clauser and Huenges [1995], hyrcenite estimated from magnetite. 16 Radiative contribution is assumed to be small from spectral measurements Shankland et al. [2005] and Shankland personal communication. 17 Giesting and Hofmeister [2002]. 18 Osako et al. [2004]. 134 Table D.3. Empirical constants for computing heat capacity of mineral end-members. c0 c1 c2 × 10−3 c3 × 10−6 Mineral kJ-K−1 kg−1 kJ-K−0.5 kg−1 kJ-K-kg−1 kJ-K2 kg−1 Notes albite 1.5012 -9.2116 -30.0990 4.0829 2 anorthite 1.5793 -4.7252 0 -1.1395 2 clinochlore 2.1726 -19.7390 -2.9739 -1.7267 1 daphnite 1.8204 -18.1750 6.0436 -1.8045 1 tremolite 1.5725 -10.7350 -11.3070 -0.4465 1 ferro-actinolite 1.4695 -12.6660 9.6112 -2.9282 1 pargasite 1.6559 -13.1260 1.5163 -2.6059 1 diopside 1.4103 -7.4109 -33.0900 4.2567 3 hedenbergite 1.3712 -10.6680 -32.7290 -0.0042 1 enstatite 1.6592 -11.9590 -22.6160 2.7805 4 ferrosillite 1.3204 -10.5590 -3.4440 -0.2858 4 forsterite 1.6572 -12.8040 0 -1.9042 4 fayalite 1.2366 -9.8820 0 -0.3052 4 spinel 1.7197 -14.0860 0 0 4 pyrope 1.4657 -7.0120 -33.0420 3.1262 4 almandine 1.2485 -6.6050 -30.2980 4.4437 4 grossular 1.1531 -0.1401 -62.1740 7.7938 4 1 Reformulated expansivity and/or reformulated heat capaciies from Holland and Powell [1998] and references therein. 2 Average of Downs et al. [1994] and Angel [2004]. 3 Berman and Brown [1985]. 4 Berman and Aranovich [1996]. APPENDIX E SIMPLIFIED CONTINENTAL GEOTHERMS Computing geotherms using my method is cumbersome, so I offer several empirical approximations to simplify computations. Thermal conductivity is approximated by the multidimensional approximation to P –T -dependent conductivity (Chapter 3, Table 3.1), λ(z, T ) = (k0 + k1 T −1 + k2 T 2 )(1 + k3 P ), (E.1) where pressure is in GPa and temperature in Kelvins. Empirical constants for the thermal conductivity of each layer are given in Table E.1. For this empirical approximation a constant density of 2850 kg/m3 and 3340 kg/m3 are used for the crust and mantle, respectively. Note that crustal rocks have two temperature ranges for each layer, which result from the α–β transtion in quartz [Höfer and Schilling, 2002; Branlund and Hofmeister , 2007]. The approximation varies by 5–10% from the mineral mixing modeled conductivity I use to compute geotherms. Geotherms is approximated using a boot-strapping solution for temperature, Ti+1 = Ti + Ai qi ∆zi − ∆z 2 , λi 2λi i (E.2) where ∆zi is the layer thickness, and Ai is the heat produced within the layer [Turcotte and Schubert, 2002]. The intra-layer conductivity, ki is computed using Equation E.1 with the average temperature of the upper and lower layer boundries as input. Heat production is estimated using my model from Chapter 3. Both Ti+1 and ki are unknowns and solved iteratively using Newton-Raphson method. Upper boundary conditions for T0 and q0 are the surface temperature and surface heat flow respectively. The resulting difference in geotherms increases from 2–21 K from 40–100 mW/m2 with an associated RMS misfit between 2 and 4 K. 136 Table E.1. Empirical conductivity constants. Upper Crust T ≤ 844 K T > 844 K Middle Crust T ≤ 844 K T > 844 K Lower Crust T ≤ 844 K T > 844 K Sp-peridotite Gt-peridotite k0 W/m-K k1 W/m k2 × 107 W/m-K3 k3 GPa−1 1.496 2.964 398.84 -495.29 4.573 0.866 0.0950 0.0692 1.733 2.717 194.59 -398.93 2.906 0.032 0.0788 0.0652 1.723 2.320 2.271 2.371 219.88 -96.98 681.12 669.40 1.705 -0.981 -1.259 -1.288 0.0520 0.0463 0.0399 0.0384 Volumetric thermal expansivity is be approximated by, αV = (a0 + a1T + a2 T −2 )(1 + a3 P ), (E.3) where temperature is in Kelvins and pressure in GPa. Coefficients for this approximation are listed in Table E.2. Using this approximation, geotherms estimated above, and Equation 3.8 the thermal isostatic relationship is easily esimated. The difference between this approximation and the numerical methods employed in this study are ∼70 m low at 40 mW/m2 and ∼70 m high at 100 mW/m2 . Additional approximations for lithospheric thickness and heat flow into the base of the lithosphere may be useful for geodynamic modeling. Lithospheric thickness using geotherms computed in Chapter 3 is approximated by, hL = 3100 , qs − 23.5 (E.4) with a misfit of 1.9 km. Using this lithospheric thickness, the heat flow into the base of the lithosphere is estimated by qL = 0.74qs − 62.0 − 8.4. qs − 23.5 (E.5) 137 Table E.2. Empirical expansivity constants. Upper Crust T ≤ 844 K T > 844 K Middle Crust T ≤ 844 K T > 844 K Lower Crust T ≤ 844 K T > 844 K Sp-peridotite Gt-peridotite a0 × 105 K−1 a1 × 108 K−2 2.355 1.741 3.208 0.500 -0.7938 -0.1193 -0.3094 -0.0778 2.020 1.663 2.149 0.602 -0.6315 -0.1059 -0.3364 -0.0745 2.198 2.134 3.036 3.026 0.921 0.711 0.925 0.906 -0.1820 -0.1177 -0.2730 -0.3116 a2 K a3 GPa−1 -0.0626 -0.0563 -0.0421 -0.0408 REFERENCES Ackerman, L., N. Mahlen, E. Jelı́nek, G. Medaris, J. Ulrych, L. Strnad, and M. Mihaljevič (2007), Geochemistry and evolution of subcontinental lithospheric mantle in central Europe: evidence from peridotite xenoliths of the Kozákov volcano, Czech Republic, J. Petrol., 48, 2235–2360. Afonso, J., G. Ranalli, and M. Fernàndez (2005), Thermal expansivity and elastic properties of the lithospheric mantle: results from mineral physics of composites, Phys. Earth Planet. Int., 149, 279–306. Afonso, J., G. Ranalli, and M. Fernàndez (2007), Density structure and buoyancy of the oceanic lithosphere revisited, Geophys. Res. Lett., 34, L10,302, doi:10.1029/2007GL029,515. Afonso, J., M. Fernàndez, G. Ranalli, W. Griffin, and J. Connolly (2008), Integrated geophysical–petrological modeling of the lithosphere and sublithospheric upper mantle: Methodology and applications, Geochem. Geophys. Geosys., 9, Q05,008, doi:10.1029/2007GC001834. Agashev, A., Y. Cherepanova, and N. Pokhilenko (2008), Whole-rock geochemistry of the deformed peridotite xenoliths from kimberlite pipe udachnaya and their bearing in kimberlite petrogenesis, in 9th Intl. Kimberlite Conf. Alexandrino, C., and V. Hamza (2008), Estimates of heat flow and heat production and a thermal model of the São francisco craton, Int. J. Earth Sci., 97, 289–306. Allis, R. (1979), A heat production model for stable continental crust, Tectonophysics, 57, 151–165. Anderson, D. (2000), The thermal state of the mantle; no role for mantle plumes, Geophys. Res. Lett., 27, 3623–3626. Anderson, R., M. Langseth, and J. Sclater (1977), The mechanism of heat transfer through the floor of the Indian Ocean, J. Geophys. Res., 82, 3391–3490. Angel, R. (2004), Equations of state of plagioclase feldspars, Contrib. Mineral. Petrol., 146, 506–512. Angel, R., and J. Jackson (2002), Elasticity and equation of state of orthoenstatite, MgSiO3 , Am. Mineralogist, 87, 558–561. Artemieva, I., and W. Mooney (2001), Thermal thickness and evolution of Precambrian lithosphere: a global study, J. Geophys. Res., 106, 16,387–16,414. 139 Ashwal, L., P. Morgan, S. Kelley, and J. Percival (1987), Heat production in an Archean crustal profile and implications for heat flow and mobilization of heatproducing elements, Earth Planet. Sci. Lett., 85, 439–450. Asimow, P., M. Hirschmann, and E. Stolper (2001), Calculation of peridotite partial melting from thermodynamic models of minerals and melts. iv. adiabatic decompression and the composition and mean properties of mid-ocean ridge basalts, J. Petrol., 42, 963–998. Attoh, K., and J. Morgan (2004), Geochemistry of high-pressure granulites from the Pan-African Dahomeyide orogen, West Africa: constraints on the origin and composition of the lower crust, J. Afr. Earth. Sci., 39, 201–208, doi: 10.1016/j.jafrearsci.2004.07.048. Aulbach, S., N. Pearson, S. O’Reilly, and B. Doyle (2007), Origins of xenolithic eclogites and pyroxenites from the central Slave Craton, Canada, J. Petrol., 48, 1843–1873. Ave Lallemant, H., J. Mercier, N. Carter, and J. Ross (1980), Rheology of the upper mantle: inferences from peridotite xenoliths, Tectonophysics, 70, 85–113. Balling, N., K. Eriksson, O. Landstroem, G. Lind, and D. Malmquist (1990), Naturgas, Vattenfall, 57. Beck, P., A. Goncharov, V. Struzhkin, B. Militzer, H. Mao, and R. Hemley (2007), Measurement of thermal diffusivity at high pressure using a transient heating technique, Appl. Phys. Lett., 91, 181,914, doi:10.1063/1.2799243. Becker, K., and R. Von Herzen (1996), Pre-drilling observations of conductive heat flow at the TAG active mound using DSV Alvin, Initial Reports ODP, 158, 23–29. Begg, G., et al. (2009), The lithospheric architecture of Africa: seismic tomography, mantle petrology, and tectonic evolution, Geosphere, 5, 23–50 (doi: 10.1130/GES00,179.1). Bell, D., G. Rossman, J. Maldener, D. Endisch, and F. Rauch (2003a), Hydroxide in olivine: A quantatative determiniation of the absolute amount and calibration of the ir spectrum, J. Geophys. Res., 108, doi:10.1029/2001JB000,679. Bell, D., M. Schmitz, and P. Janney (2003b), Mesozoic thermal evolution of the southern African mantle, Lithos, 71, 273–287. Berman, R., and L. Aranovich (1996), Optimized standard state and soultion properties of minerals: I. model calibration of olivine, orthopyroxene, cordierite, garnet, and ilmenite in the system FeO–MgO–CaO–Al2 O3 –TiO2 –SiO2 , Contrib. Mineral. Petrol., 126, 1–24. Berman, R., and T. Brown (1985), Heat capacity of minerals in the system Na2 O–K2 O–CaO–MgO–FeO–Fe2O3 –Al2 O3 –SiO2 –TiO2 –H2 O–CO2 : representation, estimation, and high temperature extrapolation, Contrib. Mineral. Petrol., 89, 168–183. 140 Bianchini, G., L. Beccaluva, C. Bondaiman, G. Nowell, G. Pearson, F. Siena, and M. Wilson (2007), Evidence of diverse depletion and metasomatic events in harzburgite—lherzolite mantle xenoliths from the Iberian plate (Olot, NE Spain): implications for lithosphere accretionary processes, Lithos, 94, 25–45, doi:10.1016/j.lithos.2006.06.008. Birch, F. (1954), The present state of geothermal investigations, Geophysics, 19, 645–659. Birch, F., and H. Clark (1940), The thermal conductivity of rocks and its dependence on temperature and composition, Am. J. Sci., 238, 529–558, 613–635. Birch, F., R. Roy, and E. Decker (1968), Heat flow and thermal history in New England and New York, in Studies of Appalachian Geology: Northern and Maritime, edited by E. an Zen et al., pp. 437–451, Interscience, New York. Bizzarro, M., and R. Stevenson (2003), Major element composition of the lithospheric mantle under the North Atlantic craton: evidence from peridotite xenoliths of the Sarfartoq area, southwestern Greenland, Contrib. Mineral. Petrol., 146, 223–240. Bjerg, E., T. Ntaflos, G.Kurat, G. Dobosi, and C. Labudı́a (2005), The upper mantle beneath Patagonia, Argentina, documented by xenoliths from alkali basalts, J. South Am. Earth Sci., 18, 125–145. Blackwell, D., and M. Richards (2004), Geothermal map of North America, Am. Assoc. Petrol. Geol., 1 Sheet, Scale 1:6,500,000. Bodorkos, S., M. Sandiford, B. Minty, and R. Blewett (2004), A high-resolution, calibrated airborne radiometric dataset applied to the estimation of crustal heat production in the Archean northern Pilbara Craton, Western Australia, Precambrian Res., 128, 57–82. Brady, R., M. Ducea, S. Kidder, and J. Saleeby (2006), The distribution of radiogenic heat production as a function of depth in the Sierra Nevada Batholith, California, Lithos, 86, 229–244. Branlund, J., and A. Hofmeister (2007), Thermal diffusivity of quartz to 1000◦C: effects of impurities and the α–β phase transition, Phys. Chem. Minerals, 34, 581– 595, doi:10.1007/s00269-007-0173-7. Brey, G., and T. Köhler (1990), Geothermobarometry in four-phase lherzolites II. new thermobarometrs, and practical assessment of existing thermobarometers, J. of Petrology, 31, 1353–1378. Carlson, R., and D. Miller (2004), Influence of pressure and mineralogy on seismic velocities in oceanic gabbros: Implications for the composition and state of the lower oceanic crust, J. Geophys. Res., 109, B09,205, doi:10.1029/2003JB002,699. Chai, M., J. Brown, and L. Slutsky (1996), Thermal diffusivity of mantle minerals, Phys. Chem. Minerals, 23, 470–475. 141 Chapman, D. (1986), Thermal gradients in the continental crust, in The Nature of the Lower Continental Crust, edited by J. Dawson, D. Carswell, J. Hall, and K. Wedepohl, 24, pp. 63–70, Geol. Soc. Spec. Pub., Denver. Chapman, D., and H. Pollack (1974), Cold spot in west Africa: Anchoring the African plate, Nature, 250, 477–478. Chapman, D., and H. Pollack (1975), Global heat flow: a new look, Earth Planet. Sci. Lett., 28, 23–32. Chapman, D., and H. Pollack (1977), Regional geotherms and lithospheric thickness, Geology, 5, 265–268. Chapman, D., and H. Pollack (1980), Global heat flow: spherical harmonic representation (abstract), Eos Trans. AGU, 61, 383. Chapman, D., and L. Rybach (1985), Heat flow anomalies and their interpretation, J. Geodyn., 4, 3–37. Christensen, N., and W. Mooney (1995), Seismic velocity structure and composition of the continental crust: a global view, J. Geophys. Res., 100, 9761–9788. Clauser, C., and E. Huenges (1995), Thermal conductivity of rocks and minerals, in Rock Physics and Phase Relations: A Handbook of Physical Constants, AGU Reference Shelf, vol. 3, edited by T. Ahrens, pp. 105–126, AGU, Washington, D.C. Clauser, C., et al. (1997), The thermal regime of the crystalline continental crust: implications from the KTB, J. Geophys. Res., 102, 18,417–18,441. Cochran, J. (1981), Simple models of diffuse extension and the pre-seafloor spreading development of the continental margin of the northeastern Gulf of Aden, Oceanologica Acta, sp., 155–165. Coffin, M., and O. Eldholm (1994), Large igneous provinces: crustal structure, dimensions, and external consequences, Rev. Geophys., 32, 1–36. Comodi, P., and P. Zanazzi (1995), High-pressure structural study of muscovite, Phys. Chem. Minerals, 22, 170–177. Crosby, A., and D. McKenzie (2009), An analysis of young ocean depth, gravity and global residual topography, Geophys. J. Int., 178, 1198–1219, doi:10.1111/j.1365246X.2009.04224.x. Crosby, A., D. McKenzie, and J. Sclater (2006), The relations between depth, age and gravity in the oceans, Geophys. J. Int., 166, 553–573, doi:10.1111/j.1365246X.2006.03015.x. Crough, S. (1975), Thermal model of oceanic lithosphere, Nature, 256, 388–389. Davidson, J., S. Turner, H. Handley, C. Macpherson, and A. Dosseto (2007), Amphibole “sponge” in arc crust?, Geology, 35, 787–790. 142 Davies, G. (1980), Review of oceanic and global heat flow estimates, Rev. Geophys., 18, 718–722. Davis, E., and C. Lister (1974), Fundamentals of ridge crest topography, Earth Planet. Sci. Lett., 21, 405–413. Davis, E., et al. (1992), FlankFlux: an experiment to study the nature of hydrothermal circulation in young crust, Can. J. Earth Sci., 29, 925–952. Davis, E., D. Chapman, and C. Forster (1996), Observations concerning the vigor of hydrothermal circulation in young oceanic crust, J. Geophys. Res., 101, 2927–2942. Davis, E., D. Chapman, H. Villinger, S. Robinson, J. Grigel, A. Rosenberger, and D. Pribnow (1997), Seafloor heat flow on the eastern flank of the Juan de Fuca ridge: data from “FLANKFLUX” studies through 1995, Proc. Ocean Drilling Prog. Init. Rep., 168, 23–33. Davis, E., et al. (1999), Regional heat flow variations across the sedimented Juan de Fuca Ridge eastern flank: constraints on lithospheric cooling and lateral hydrothermal heat transport, J. Geophys. Res., 104, 17,675–17,688. Davis, E., K. Becker, and J. He (2004), Costa Rica Rift revisited: Constraints on shallow and deep hydrothermal circulation in young oceanic crust, Earth Planet. Sci. Lett., 222, 863–879. Del Lama, E., M. de Oliveira, and A. Zanardo (1998), Geochemistry of the Guaxupé granulites, Minas Gerias, Brazil, Gondwana Res., 6, 357–365. Diment, W., and H. Pratt (1988), Thermal conductivity of some rock forming minerals: A tabulation, Open-File Rept. 88-690, USGS. Divins, D. (2007), NGDC total sediment thickness of the world’s oceans & marginal seas, Tech. rep., http://www.ngdc.noaa.gov/mgg/sedthick/sedthick.html. Donnelly, K., S. Goldstein, C. Langmuir, and M. Spiegelman (2004), Origin of enriched ocean ridge basalt and implications for mantle dynamics, Earth Planet. Sci. Lett, 226, 347–366, doi:10.1016/j.epsl.2004.07.019. Downs, R., R. Hazen, and L. Finger (1994), The high-pressure crystal chemistry of low albite and the origin of the pressure dependency of Al-Si ordering, Am. Mineralogist, 79, 1042–1052. Dubuffet, F., D. Yuen, and M. Rabinowicz (1999), Effects of realistic mantle thermal conductivity on the patterns of 3-D convection, Earth Planet. Sci. Lett, 171, 401– 409. Dunn, D. (2002), Xenolith mineralogy and geology of the Prairie Creek lamproite province, Arkansas, Ph.D. thesis, Univ. Texas-Austin. Elderfield, H., and A. Schultz (1996), Mid-ocean ridge hydrothermal flux and the chemical composition of the ocean, Ann. Rev. Earth Planet. Sci., 24, 191–224. 143 Eldholm, O., E. Sundvor, P. Vogt, B. Hjelstuen, K. Crane, A. Nilsen, and T. Gladezenko (1999), SW Barents Sea continental margin heat flow and Hakon Mosby Mud Volcano, Geo-Marine Lett., 19, 29–37. Fei, Y. (1995), Thermal expansion, in Mineral Physics and Crystallography: A handbook of physical constants, AGU Reference Shelf, vol. 2, edited by T. Ahrens, chap. 6, pp. 29–44, AGU, Washington, D.C. Fei, Y., and S. Saxena (1987), An equation for the heat capacity of solid, Geochim. Cosmochim. Acta, 51, 251–254. Flowers, R., L. Royden, and S. Bowring (2004), Isostatic constraints on the assembly, stabilization, and preservation of cratonic lithosphere, Geology, 32, 321–324. Förster, A., and H. Förster (2000), Crustal composition and mantle heat flow: implications from surface heat flow and radiogenic heat production in the Variscan Erzgebirge (Germany), J. Geophys. Res., 105, 27,917–27,938. Fountain, D., M. Salisbury, and K. Furlong (1987), Heat production and thermal conductivity of rocks from the Pikwitonei—Sachigo continental cross section, central Manitoba: implications for the thermal structure of the Archean crust, Can. J. Earth Sci., 24, 1583–1594. Furlong, K., and D. Chapman (1987), Crustal heterogeneities and the thermal structure of the continental crust, Geophys. Res. Lett., 14, 314–317. Furukawa, Y., and H. Shinjoe (1997), Distribution of radiogenic heat generation in the arc’s crust of Hokkaido Island, Japan, Geophys. Res. Lett., 24, 1279–1282. Garrido, C., J. Bodinier, J. Burg, G. Zeilinger, S. Hussain, H. Dawood, M. Chaudhry, and F. Gervilla (2006), Petrogenesis of mafic garnet granulite in the lower crust of the Kohistan paleo-arc complex (northern Pakistan): implications fro intra-crustal differentiation of island arcs and generation of continental crust, J. Petrol., 47, 1873–1914, doi:10.1093/petrology/egl030. Gibert, B., U. Seipold, A. Tommasi, and D. Mainprice (2003), Thermal diffusivity of upper mantle rocks: implications of temperature, pressure, and the deformation fabric, J. Geophys. Res., 108, doi:10.1029/2002JB002,108. Giesting, P., and A. Hofmeister (2002), Thermal conductivity of disordered garnets from infrared spectroscopy, Phys. Rev. B, 65, doi: 10.1103/PhysRevB.65.144,305. Goncharov, A., B. Haugen, V. Struzhkin, P. Beck, and S. Jacobsen (2008), Radiative conductivity in the Earth’s lower mantle, Nature, 456, 231–234 doi:10.1038/nature07,412. Goncharov, A., B. Haugen, V. Struzhkin, P. Beck, and S. Jacobsen (2009), Thermal conductivity of lower mantle minerals, Phys. Earth Planet. Int., 174, 24–32. Gosnold, W. (1987), Redistribution of U and Th in shallow plutonic environments, Geophys. Res. Lett., 14, 291–294. 144 Grégoire, M., D. Bell, and A. le Roex (2003), Garnet lherzolite from the Kaapvaal Craton (South Africa): trace element evidence for a metasomatic history, J. Petrol., 44, 629–657. Grevemeyer, I., and A. Bartetzko (2004), Hydrothermal aging of oceanic crust: inferences from seismic reflection, in Hydrology of the Oceanic Lithosphere, edited by E. Davis and H. Elderfield, pp. 128–150, Cambridge University Press, London. Griffin, W., S. O’Reilly, and C. Ryan (1999), The composition and origin of subcontinental lithospheric mantle, in Mantle Petrology: Field Observations and High Pressure Experimentation: A Tribute to Francis R. (Joe) Boyd, edited by Y. Fei, C. Bertka, and B. Mysen, Special Publication 6, pp. 13–45, Geochem. Soc., St. Louis, MO. Hacker, B., G. Abers, and S. Peacock (2003), Subduction factory 1. Theoretical mineralogy, densities, seismic wave speeds, and H2 O contents, J. Geophys. Res., 108, B1, 2029, doi:10.1029/2001JB001,127. Han, U., and D. Chapman (1995), Thermal isostasy: elevation changes of geologic provinces, J. Geol. Soc. Korea, 31, 106–115. Harrell, M. (2002), Anisotropic lattice thermal diffusivity in olivines and pyroxenes to high temperatures, Ph.D. thesis, Univ. Washington. Harris, R., and D. Chapman (2004), Deep-seated oceanic heat flux, heat deficits and hydrothermal circulation, in Hydrology of the Oceanic Lithosphere, edited by E. Davis and H. Elderfield, chap. 10, pp. 311–336, Cambridge University Press, Cambridge, U.K. Hart, S., and H. Staudigel (1982), The control of alkalies and uranium in seawater by ocean crust alteration, Earth Planet. Sci. Lett, 58, 202–212. Hasterok, D. (in prep.), Radiative conductivity and diffusivity of olivine revisited. Hasterok, D., and D. Chapman (2007a), Continental thermal isostasy II: Applications to North America, J. Geophys. Res., 112, doi:10.1029/2006JB004,664. Hasterok, D., and D. Chapman (2007b), Continental thermal isostasy I: Methods and sensitivity, J. Geophys. Res., 112, doi:10.1029/2006JB004,663. Hawkesworth, C. (1974), Vertical distribution of heat production in the basement of the eastern Alps, Nature, 249, 435–436. He, L., S. Hu, S. Huang, W. Yang, J. Wang, Y. Yuan, and S. Yang (2008), Heat flow study at the Chinese Continental Scientific Drilling site: borehole temperature, thermal conductivity, and radiogenic heat production, J. Geophys. Res., 113, B02,404, doi:10.1029/2007JB004958. Hearn, B. (2004), The Homestead kimberlite, central Montana, USA: mineralogy, xenocrysts, and upper-mantle xenoliths, Lithos, 77, 473–491. 145 Herzberg, C. (1976), The plagioclase/spinel-lherzolite facies boundary; its bearing on corona structure formation and tectonic history of the Norwegian Caledonides, in Progress in Experimental Petrology, vol. D.3, pp. 233–235, Natural Environment Research Council Publications, London. Herzberg, C., P. Asimow, N. Arndt, Y. Niu, C. Lesher, J. Fritton, M. Cheadle, and A. Saunders (2007), Temperatures in ambient mantle and plumes: constraints from basalts, picrites, and komatiites, Geochem. Geophys. Geosys., 8, doi:10.1029/2006GC001,390. Höfer, M., and F. Schilling (2002), Heat transfer in quartz, orthoclase, and sanidine at elevated temperature, Phys. Chem. Minerals, 29, 571–584, doi:10.1007/s00269002-0277-z. Hofmeister, A. (1999a), Mantle values of thermal conductivity and the geotherm from phonon lifetimes, Science, 283, 1699–1706. Hofmeister, A. (1999b), Corrections and clarifications: Mantle values of thermal conductivity and the geothermal from phonon lifetimes, Science, 284, 264. Hofmeister, A. (2005), Dependence of diffusive radiative transfer on grain-size, temperature, and Fe-content: implications for mantle processes, J. Geodynamics, 40, 51–72. Hofmeister, A., and M. Pertermann (2008), Thermal diffusivity of clinopyroxenes at elevated temperature, Eur. J. Mineral., 20, 537–549. Holland, T., and R. Powell (1998), An internally consistent thermodynamic data set for phases of petrological interest, J. Metamorphic Geol., 16, 309–343. Hölttä, P. (1997), Geochemical characteristics of granulite facies rocks in the Archean Varpaisjärvi area, central Fennoscandian Shield, Lithos, 40, 31–53. Horai, K. (1971), Thermal conductivity of rock-forming minerals, J. Geophys. Res., 76, 1278–1308. Horai, K., and G. Simmons (1969), Thermal conductivity of rock-forming minerals, Earth Planet. Sci. Lett., 6, 359–368. Hugh-Jones, D. (1997), Thermal expansion of MgSiO3 and FeSiO3 ortho- and clinopyroxenes, Am. Mineralogist, 82, 689–696. Hutchison, M., and D. Frei (2008), Diamondiferous kimberlite from Garnet Lake, West Greenland II: diamonds and the mantle sample, in 9th Intl. Kimberlite Conf. Hutnak, M., and A. Fisher (2007), The influence of sedimentation, local and regional hydrothermal circulation, and thermal rebound on measurements of seafloor heat flux, J. Geophys. Res., 112, B12,101, doi:10.1029/2007JB005,022. 146 Hutnak, M., A. Fisher, R. Harris, C. Stein, K. Wang, G. Spinelli, M. Schindler, H. Villinger, and E. Silver (2008), Large heat and fluid flux driven through midplate outcrops on ocean crust, Nature Geoscience, 1, 611–614. Hyndman, R., C. Currie, and S. Mazzotti (2005), Subduction zone backarcs, mobile belts, and orogenic heat, GSA Today, 15, 4–10. Ionov, D. (2004), Chemical variations in peridotite xenoliths from Vitim, Siberia: inferences for REE and Hf behaviour in the garnet-facies upper mantle, J. Petrol., 45, 343–367. Ionov, D., I. Ashchepkov, H.-G. Stosch, G. Witt-Eickschen, and H. Seck (1993), Garnet peridotite xenoliths from the Vitim volcanic field, Baikal region: the nature of the garnet-spinel peridotite zone in the continental mantle, J. Petrol., 34, 1141– 1175. Ionov, D., S. O’Reilly, Y. Genshaft, and M. Kopylova (1996), Carbonate-bearing mantle peridotite xenoliths from Spitsbergen: phase relationships, mineral compositions and trace-element residence, Contrib. Mineral. Petrol., 125, 375–392. Ionov, D., J.-L. Bodinier, S. Mukasa, and A. Zanetti (2002), Mechanisms and sources of mantle metasomatism: major and trace element compositions of peridotite xenoliths from Spitsbergen in the context of numerical modelling, J. Petrol., 43, 2219–2259. Ionov, D., I. Chanefo, and J. Bodinier (2005), Origin of Fe-rich lherzolite and wehrlites from Tok, SE Siberia by reactive melt percolation in refractory mantle peridotites, Contrib. Mineral. Petrol., 150, 335–353, doi:10.1007/s00410-005-0026-7. Jackson, J., J. Palko, D. Andrault, S. Sinogeikin, D. Lakshtanov, J.Wang, J. Bass, and B.-R. C.-S. Zhang (2003), Thermal expansion of natural orthoenstatite to 1473 K, Eur. J. Mineral., 15, 469–473. James, D., F. Boyd, D. Schutt, D. Bell, and R. Carlson (2004), Xenolith contraints on seismic velocities in the upper mantle beneath southern Africa, Geochem. Geophys. Geosys., 5, doi:10.1029/2003GC000,551. Jaupart, C. (1983), Horizontal heat transfer due to radioactivity contrasts: Causes and consequences of the linear heat flow relation, Geophys. J. R. astr. Soc., 75, 411–435. Jaupart, C., and J. Mareschal (1999), The thermal structure and thickness of continental roots, Lithos, 48, 93–114. Jaupart, C., and J.-C. Mareschal (2003), Constraints on crustal heat production from heat flow data, in Treatise on Geochemistry: The Crust, vol. 3, edited by R. Rudnick, chap. 2, pp. 65–84, Elsevier. Jaupart, C., J. Sclater, and G. Simmons (1981), Heat flow studies: Constraints on the distribution of uranium, thorium and potassium in the continental crust, Earth Planet. Sci. Lett., 52, 328–344. 147 Jessop, A., M. Hobart, and J. Sclater (1976), The world heat flow data collection— 1975, Geothermal Series 5, Earth Physics Branch, Energy, Mines and Resources, Ottawa, Canada. Jõeleht, A., and I. Kukkonen (1998), Thermal properties of granulite facies rocks in the Precambrian basement of Finland and Estonia, Tectonophysics, 291, 195–203. Katayama, I., Y. Suyama, J. Ando, and T. Komiya (2008), Mineral chemistry and P –T condition of granular and sheared peridotite xenoliths from Kimberley, South Africa: origin of the textural variation in the cratonic mantle, Lithos, p. doi:10.1016/j.lithos.2008.05.004. Katsura, T. (1995), Thermal diffusivity of olivine under upper mantle conditions, Geophys. J. Int., 122, 63–69. Katz, R., M. Spiegelman, and C. Langmuir (2003), A new parameterization of hydrous mantle melting, Geochem. Geophys. Geosys., 9, 1073, doi: doi:10.1029/2002GC000433. Kaul, N., J.-P. Foucher, and M. Heesemann (2006), Estimating mud expulsion rates from temperature measurements on Hakon Mosby Mud Volcano, SW Barents Sea, Marine Geology, 229, 1–14. Ketcham, R. (2006), Distribution of heat-producing elements in the upper and middle crust of southern and west central Arizona: evidence from core complexes, J. Geophysical Res., 101, 13,611–13,632. Kjarsgaard, B., and T. Peterson (1992), Kimberlite-derived ultramafic xenoliths from the diamond stability field: a new Cretaceous geotherm for Somerset Island, Northwest Territories, in Current Research, Part B, Paper 92-1B, pp. 1–6, Geol. Surv. Canada. Klemme, S., and H. O’Neill (2000), The near-solidus transition from garnet lherzolite to spinel lherzolite, Contrib. Mineral. Petrol., 138, 237–248. Kopylova, M., and G. Caro (2004), Mantle xenoliths from the southeastern Slave Craton: evidence for chemical zonation in a thick, cold lithosphere, J. Petrol., 45, 1045–1067. Kopylova, M., J. Russell, and H. Cookenboo (1999), Petrology of peridotite and pyroxenite xenoliths from the Jericho Kimberlite: implications for the thermal state of the mantle beneath the Slave craton, northern Canada, J. Petrol., 40, 79–104. Korenaga, J. (2007a), Effective thermal expansivity of Maxwellian oceanic lithosphere, Earth Planet. Sci. Lett, 257, 343–349. Korenaga, J. (2007b), Thermal cracking and the deep hydration of oceanic lithosphere: A key to the generation of plate tectonics?, J. Geophys. Res., 112, doi:10.1029/2006JB004,502. 148 Korenaga, T., and J. Korenaga (2008), Subsidence of normal oceanic lithosphere, apparent thermal expansivity, and seafloor flattening, Earth Planet. Sci. Lett, 268, 41–51. Kukkonen, I., and A. Jõeleht (1996), Geothermal modelling of the lithosphere in the central Baltic Shield and its southern slope, Tectonophysics, 255, 25–45. Kukkonen, I., and R. Lahtinen (2001), Variation of radiogenic heat production rate in 2.8–1.8 Ga old rocks in the central Fennoscandian Shield, Phys. Earth Planet. Int., 126, 279–294. Kukkonen, I., and P. Peltonen (1999), Xenolith-controlled geotherm for the central Fennoscandian Shield: implications for lithosphere–asthenosphere relations, Tectonophysics, 304, 301–315. Lachenbruch, A. (1970), Crustal temperature and heat production: implications for the linear heat flow relation, J. Geophys. Res., 75, 3291–3300. Lachenbruch, A., and C. Bunker (1971), Vertical gradients of heat production in the continental crust: 2. some estimates from borehole data, J. Geophys. Res., 76, 3852–3860. Lamb, W., and R. Popp (2009), Amphibole equilibria in mantle rocks: Determining values of mantle aH2 O and implications for mantle H2 O contents, Am. Mineral., 94, 41–52. Langseth, M., and R. Anderson (1979), Correction, J. Geophys. Res., 84, 1139–1140. Laske, G., and G. Masters (1997), A global digital map of sediment thickness, EOS Trans. AGU, 78, F483. Leake, B., and 21 others (1997), Nomenclature of amphibolites: report of the subcommittee on amphiboles of the international mineralogical association, commission on new minerals and mineral names, Can. Mineral., 35, 219–246. Lee, C.-T., and R. Rudnick (1999), Compositionally stratified cratonic lithosphere: petrology and geochemistry of peridotite xenoliths from the Labait tuff cone, Tanzania, in Proc. 7th Intl. Kimberlite Conf., edited by J. Gurney and S. Richardson, pp. 503–521. Lee, W. (1963), Heat flow data analysis, Rev. Geophys., 1, 449–479. Lee, W. (1970), On the global variations in terrestrial heat flow, Phys. Earth Planet. Int., 2, 332–341. Lee, W., and S. Uyeda (1965), Review of heat flow data, in Terrestrial heat flow, Geophys. Monogr., vol. 8, edited by W. Lee, pp. 87–100, Am. Geophys. Union, Wasington D.C. 149 Lewis, T., R. Hyndman, and P. Flück (2003), Heat flow, heat generation, and crustal temperatures in the northern Canadian Cordillera: thermal control of tectonics, J. Geophys. Res., 108, 2316, doi:10.1029/2002JB002,090. Lister, C. (1972), On the thermal balance of a Mid-Ocean ridge, Geophys. J. Roy. Astr. Soc., 26, 515–535. Lucazeau, F., et al. (2008), Persistent thermal activity at the eastern Gulf of Aden after continental break-up, Nature Geoscience, 1, doi:10.1038/ngeo359. Lucazeau, F., et al. (2009), Post-rift volcanism and high heat-flow at the oceancontinent transition of the eastern Gulf of Aden, Terra Nova, 21, 285–292. MacGregor, I. (1974), The system MgO–Al2 O3 –SiO2 : solubility of Al2 O3 in enstatite for spinel and garnet peridotite compositions, Am. Mineral., 59, 110–119. MacKenzie, J., and D. Canil (1999), Composition and thermal evolution of cratonic mantle beneath the central Archean Slave Province, NWT, Canada, Contrib. Mineral. Petrol., 134, 313–324. Majorowicz, J., G. Garven, A. Jessop, and C. Jessop (1999), Present heat flow across the western Canada sedimentary basin: The extent of hydrodynamic influence, in Geothermics in Basin Analysis, edited by A. Foerster and D. Merriam, pp. 61–80, Kluwer Academic. Mallon, A., and R. Swarbrick (2002), A compaction trend for non-reservoir North Sea chalk, Marine Petrol. Geology, 19, 527–539. Mareschal, J., and C. Jaupart (2004), Variations of surface heat flow and lithospheric heat production beneath the North American Craton, Earth Planet. Sci. Lett., 223, 65–77. Mareschal, J., A. Poirier, F. Rolandone, G. Bienfait, C. Gariépy, R. Lapointe, and C. Jaupart (2000), Low mantle heat flow at the edge of the North American continent, Voisey Bay, Labrador, Geophys. Res. Lett., 27, 823–826. Martignole, J., and J. Martelat (2005), Proterozoic dykes as monitors of HP granulite facies metamorphism in the Grenville Front Tectonic Zone (western Quebec), Precambrian Res., 138, 183–207, doi:10.1016/j.precamres.2005.05.002. McKenzie, D. (1967), Some remarks on heat flow and gravity anomalies, J. Geophys. Res., 72, 6261–6273. McKenzie, D. (1978), Some remarks on the development of sedimentary basins, Earth Planet. Sci. Lett., 40, 25–32. McKenzie, D., J. Jackson, and K. Priestley (2005), Thermal structure of oceanic and continental lithosphere, Earth Planet. Sci. Lett., 233, 337–349. 150 McLaren, S., M. Sandiford, M. Hand, N. Neumann, L. Wyborn, and I. Bastrakova (2003), The hot south continent: heat flow and heat production in Australian Proterozoic terranes, in Geological Society of Australia Special Publicaton 22, pp. 151–161. McLaren, S., M. Sandiford, and R. Powell (2005), Contrasting styles of Proterozoic crustal evolution: A hot-plate tectonic model for Australian terranes, Geology, 33, 673–676. Michaut, C., C. Jaupart, and D. Bell (2007), Transient geotherms in Archean continental lithosphere: new constraints on thickness and heat production of the subcontinental lithospheric mantle, J. Geophys. Res., 112, doi:10.1029/2006JB004,464. Muller, M., et al. (2009), Lithospheric structure, evolution and diamond prospectivity of the Rehoboth Terrane and western Kaapvaal Craton, southern Africa: constraints from broadband magnetotellurics, Lithos, 112S, 93–105. Müller, R., M. Sdrolias, C. Gaina, and W. Roest (2008), Age, spreading rates, and spreading asymmetry of the world’s ocean crust, Geochem. Geophys. Geosys., 9, Q04,006, doi:10.1029/2007GC001,743. Mutter, C., and J. Mutter (1993), Variations in thickness of layer 3 dominate oceanic crustal structure, Earth Planet. Sci. Lett., 117, 295–317. Nicolaysen, L., R. Hart, and N. Gale (1981), The Vredefort radioelement profile extended to supracrustal strata at Carletonville, with implications for continental heat flow, J. Geophys. Res., 86, 10,653–10,661. Nielson, L., M. Hutchison, and J. Malarkey (2008), Geothermal constraints from kimberlite-hosted garnet lherzolites from southern Greenland, in 9th Intl. Kimberlite Conf. Nielson, S. (1987), Steady state heat flow in a random medium and the linear heat flow–heat production relationship, Geophys. Res. Lett., 14, 318–321. Niu, Y., C. Langmuir, and R. Kinzler (1997), The origin of abyssal peridotites: a new perspective, Earth Planet. Sci. Lett., 152, 251–265. Niu, Y., M. Regelous, I. Wendt, R. Batiza, and M. O’Hara (2002), Geochemistry of near-EPR seamounts: importance of source vs. process and the origin of enriched mantle component, Earth Planet. Sci. Lett., 199, 327–345. Nyblade, A. (1999), Heat flow and the structure of Precambrian lithosphere, Lithos, 48, 81–91. Nyblade, A., H. Pollack, D. Jones, F. Podmore, and M. Mushayandebvu (1990), Terrestrial heat flow in east and southern Africa, J. Geophys. Res., 95, 17,371– 17,384. 151 O’Reilly, S., and W. Griffin (2000), Apatite in the mantle: implications for metasomatic processes and high heat production in Phanerozoic mantle, Lithos, 53, 217–232. O’Reilly, S., W. Griffin, P. Morgan, D. Ionov, and M. Horman (1997), Mantle apatite revisited: major reservoir for U and Th in the mantle and reflector of mantle-fluid sources, LPI Contrib., 921, 154. Osako, M., E. Ito, and A. Yoneda (2004), Simultaneous measurements of thermal conductivity and thermal diffusivity for garnet and olivine under high pressure, Phys. Earth Planet. Int., 143–144, 311–320. Owen, J., F. Longstaffe, and J. Greenough (68), Petrology of sapphirine granulite and associated sodic gneisses from the Indian Head Range, Newfoundland, Lithos, 2003, 91–114, doi:10.1016/S0024-4937(03)00043-4. Parker, R., and D. Oldenburg (1973), Thermal model of ocean ridges, Nature, 242, 137–139. Parsons, B., and J. Sclater (1977), An analysis of the variation of ocean floor bathymetry and heat flow with age, J. Geophys. Res., 82, 803–827. Pawley, A., S. Clark, and N. Chinnery (2002), Equation of state measurements of chlorite, pyrophyllite, and talc, Am. Mineralogist, 87, 1172–1182. Peltonen, P., H. Huhma, M. Tyni, and N. Shimizu (1999), Garnet peridotite xenoliths from kimberlites of Finland: nature of the continental mantle at an Archean Craton – Proterozoic mobile belt transition, in 7th Intl. Kimberlite Conf. Petrunin, G., V. Popov, and I. Il’in (2004), Conductive heat transfer in plagioclases (English translation), Izvestiya Phys. Solid Earth, 40, 752–759. Pinet, C., and C. Jaupart (1987), The vertical distribution of radiogenic heat production in the precambrian crust of Norway and Sweden: geothermal implications, Geophys. Res. Lett., 14, 260–263. Poirier, J.-P., and A. Tarantola (1998), A logarithmic equation of state, Physics Earth Planet. Int., 109, 1–8. Pollack, H. (1980), On the use of the volumetric thermal expansion coefficient in models of ocean floor topography, Tectonophysics, 64, T45–T47. Pollack, H., and D. Chapman (1977), Mantle heat flow, Earth Planet. Sci. Lett., 34, 174–184. Pollack, H., S. Hurter, and J. Johnson (1993), Heat flow from the Earth’s interior: analysis of the global data set, Rev. Geophys., 31, 267–280. Popov, Y., S. Pevzner, V. Pimenov, and R. Romushkevich (1999), New geothermal data from the Kola superdeep well SG-3, Tectonophysics, 306, 345–366. 152 Rao, R., G. Rao, and G. Reddy (1982), Age dependence of continental heat flow— fantasy and facts, Earth Planet. Sci. Lett., 59, 288–302. Ray, L., P. Kumar, G. Reddy, S. Roy, G. Rao, R. Srinivasan, and R. Rao (2003), High mantle heat flow in a Precambrian granulite province: evidence from southern india, J. Geophys. Res., 108, doi:10.1029/2001JB000,688. Revelle, R., and A. Maxwell (1952), Heat flow through the floor of the eastern north Pacific Ocean, Nature, 170, 199–200. Robinson, J., and B. Wood (1998), The depth of the spinel to garnet transition at the peridotite solidus, Earth Planet. Sci. Lett, 164, 277–284. Roden, M., E. Laz’ko, and E. Jagoutz (1999), The role of garnet pyroxenites in the Siberian lithosphere: evidence from the Mir kimberlite, in Proc. 7th Intl. Kimberlite Conf., pp. 714–720. Roden, M., A. Patiño-Douce, E. Jagoutz, and E. Laz’ko (2006), High pressure petrogeneisis of Mg-rich garnet pyroxenites from Mir kimberlite, Russia, Lithos, 90, 77–91. Rona, P., S. Petersen, K. Becker, R. V. Herzen, M. Hannington, P. Herzig, J. Naka, C. Lalou, and G. Thompson (1996), Heat flow and mineralogy of TAG relict hightemperature hydrothermal zones: Mid-Atlantic Ridge 26◦ N, 45◦ W, Geophys. Res. Lett., 23, 3507–3510. Roy, R., D. Blackwell, and F. Birch (1968), Heat generation of plutonic rocks and continental heat flow provinces, Earth Planet. Sci. Lett., 5, 1–12. Roy, R., A. Beck, and Y. Touloukian (1981), Thermophysical properties of rocks, in Physical Properties of Rocks and Minerals, Data Series on Material Properties, vol. II-2, edited by Y. Touloukian, W. Judd, and R. Roy, chap. 12, pp. 409–502, McGraw-Hill. Roy, S., and R. Rao (2000), Heat flow in the Indian shield, J. Geophys. Res., 105, 25,587–25,604. Roy, S., L. Ray, A. Bhattacharya, and R. Sirnivasan (2008), Heat flow and crustal thermal structure in the Late Archaean Clospet granite batholith, south India, Int. J. Earth Sci., 97, 245–256. Rudnick, R., and D. Fountain (1995), Nature and composition of the continental crust: a lower crustal perspective, Rev. Geophys., 33, 267–309. Rudnick, R., and S. Gao (2003), Composition of the continental crust, in Treatise on Geochemistry: The Crust, vol. 3, edited by R. Rudnick, chap. 1, pp. 1–64, Elsevier. Rudnick, R., and A. Nyblade (1999), The thickness and heat production of Archean lithosphere: constraints from xenolith thermobarometry and surface heat flow, in Mantle Petrology: Field Observations and High Pressure Experimentation: A 153 Tribute to Francis R. (Joe) Boyd, edited by Y. Fei, C. Bertka, and B. Mysen, Special Pub. 6, Geochem. Soc. Rudnick, R., W. McDonough, and A. Orpin (1994), Northern Tanzanian peridotite xenoliths: a comparison with Kaapvaal peridotites and inferences on metasomatic interactions, in Proc. 5th Intl. Kimberlite Conf. Rudnick, R., W. McDonough, and R. O’Connell (1998), Thermal structure, thickness and composition of continental lithosphere, Chem. Geology, 145, 395–411. Rudnick, R., S. Gao, W. Ling, Y. Liu, and W. McDonough (2004), Petrology and geochemistry of spinel peridotite xenoliths from Hannuoba and Qixia, North China craton, Lithos, 77, 609–637. Russell, J., G. Dipple, and M. Kopylova (2001), Heat production and heat flow in the mantle lithosphere, Slave craton, Canada, Phys. Earth Planet. Int., 123, 27–44. Rybach, L., and G. Buntebarth (1984), The variation of heat generation, density and seismic velocity with rock type in the continental lithosphere, Tectonophysics, 103, 335–344. Salters, V., and A. Stracke (2004), Composition of the depleted mantle, Geochem. Geophys. Geosys., 5, Q05,004, doi:10.1029/2003GC000597. Saltzer, R., N. Chatterjee, and T. Grove (2001), The spatial distribution of garnets and pyroxenes in mantle peridotites: pressure–temperature history of peridotites from the Kaapvaal Craton, J. Petrol., 42, 2215–2229. Sand, K., T. Waight, D. Pearson, T. Nielsen, E. Makovicky, and M. Hutchison (2009), The lithospheric mantle below southern West Greenland: a geothermobarometric approach to diamond potential and mantle stratigraphy, Lithos, 112S, 1155–1166, doi:10.1016/j.lithos.2009.05.012. Sandiford, M., and S. McLaren (2002), Tectonic feedback and the ordering of heat producing elements within the continental lithosphere, Earth Planet. Sci. Lett., 204, 133–150. Sass, J. (1965), The thermal conductivity of fifteen feldspar specimens, J. Geophys. Res., 70, 4064–4065. Schärmeli, G. (1982), Anisotropy of olivine thermal conductivity at 2.5 GPa and up to 1500 K measured on optically non-thick samples, in High-Pressure Researches in Geoscience, edited by W. Schreyer, pp. 349–373, Schweizerbart’sche Verlagsbuchhandlung, Stuttgart. Schatz, J., and G. Simmons (1972), Thermal conductivity of Earth materials at high temperatures, J. Geophys. Res., 77, 6966–6983. Schmidberger, S., and D. Francis (1999), Nature of the mantle roots beneath the North American craton: mantle xenolith evidence from Somerset Island kimberlites, Lithos, 48, 195–216. 154 Schmidt, M., and S. Poli (1998), Experimentally based water budgets for dehydrating slabs and consequences for arc magma generation, Earth Planet. Sci. Lett, 163, 361–379. Schneider, R., R. Roy, and A. Smith (1987), Investigations and interpretations of the vertical distribution of U, Th, and K: South Africa and Canada, Geophys. Res. Lett., 14, 264–267. Sclater, J., and P. Christie (1980), Continental stretching: an explanation of the postmid-Cretaceous subsidence of the North Sea basin, J. Geophys. Res., 85, 3711–3739. Sclater, J., and J. Francheteau (1970), The implications of terrestrial heat flow observations on current tectonic and geochemical models of the crust and upper mantle of the Earth, Geophys. J. R. astr. Soc., 20, 509–542. Sclater, J., J. Crowe, and R. Anderson (1976), On the reliability of oceanic heat flow averages, J. Geophys. Res., 81, 2997–3006. Sclater, J., C. Jaupart, and D. Galson (1980), The heat flow through the oceanic and continental crust and the heat loss of the Earth, Rev. Geophys. Space Phys., 18, 269–311. Sclater, J., B. Parsons, and C. Jaupart (1981), Oceans and contients: similarities and differences in mechanisms of heat loss, J. Geophys. Res., 86, 11,535–11,552. Shankland, T., U. Nitsan, and A. Duba (1979), Optical absorption and radiative heat transport in olivine at high temperature, J. Geophys. Res., 84, 1603–1610. Shankland, T., F. Schilling, B. Gibert, and K. Gratz (2005), Thermal diffusivity and conductivity measurements: Effect of sample length and radiation, EOS Trans. AGU, 86(52), Fall Meet. Suppl. MR23C–0075. Simon, N., G. Irvine, G. Davies, and D. P. nd R.W. Carlson (2003), The origin of garnet and clinopyroxene in “depleted” Kaapvaal peridotites, Lithos, 71, 289–322. Spinelli, G., E. Giambalvo, and A. Fisher (2004), Sediment permeability, distribution, and influence on fluxes in oceanic basement, in Hydrology of the Oceanic Lithosphere, edited by E. Davis and H. Elderfield, pp. 151–188, Cambridge University Press, London. Stein, C., and S. Stein (1992), A model for the global variation in oceanic depth and heat flow with lithospheric age, Nature, 359, 123–129. Stein, C., and S. Stein (1994), Contraints on hydrothermal flux through the oceanic lithosphere from global heat flow, J. Geophys. Res., 99, 3081–3095. Stein, C., and S. Stein (1997), Estimation of lateral hydrothermal flow distance from spatial variations in oceanic heat flow, Geophys. Res. Lett., 24, 2323–2326. 155 Stiefenhofer, J., K. Viljoen, and J. Marsh (1997), Petrology and geochemistry of peridotite xenoliths from the Letlhakane kimberlites, Botswana, Contrib. Mineral. Petrol., 127, 147–158. Stixrude, L., and C. Lithgow-Bertelloni (2005), Thermodynamics of mantle minerals – I. physical properties, Geophys. J. Int., 162, 610–632. Sueda, Y., T. Irifune, T. Sanehira, T. Yagi, N. Nishiyama, T. Kikegawa, and K. Funakoshi (2008), Thermal equation of state of CaFe2 O4 -type MgAl2 O4 , Phys. Earth Planet. Int., 174, 78–85. Swanberg, C. (1972), Vertical distribution of heat generation in the Idaho Batholith, J. Geophys. Res., 77, 2508–2513. Sykes, T. (1996), A correction for sediment load upon the ocean floor: Uniform versus varying sediment density estimations—implications for isostatic correction, Marine Geology, 133, 35–49. Tucholke, B., J. Lin, and M. Kleinrock (1998), Megamullions and mullion structure defining oceanic metamorphic core complexes on the Mid-Atlantic ridge, J. Geophys. Res., 103, 9857–9866. Tucholke, B., K. Fujioka, T. Ishihara, G. Hirth, and M. Kinoshita (2001), Submersible study of an oceanic megamullion in the central North Atlantic, J. Geophys. Res., 106, 16,145–16,161. Turcotte, D., and G. Schubert (2002), Geodynamics, 2nd ed., 456 pp., Cambridge University Press, Cambridge, U.K. van den Berg, A., D. Yuen, and V. Steinbach (2001), The effects of variable thermal conductivity on mantle heat-transfer, Geophys. Res. Lett., 28, 875–878. Velde, B. (1996), Compaction trends of clay-rich deep sea sediments, Marine Geology, 133, 193–201. Vigneresse, J., and M. Cuney (1991), Are granites representative of heat flow provinces?, in Terrestrial Heat Flow and the Lithosphere Structure, edited by V. Cermak and L. Rybach, Exploration of the Deep Contiental Crust, pp. 87–110, Springer-Verlag, Berlin, Germany. Vitorello, I., and H. Pollack (1980), On the variation of continental heat flow with age and the thermal evolution of the continents, J. Geophys. Res., 85, 983–995. Von Herzen, R., and S. Uyeda (1963), Heat flow through the eastern Pacific Ocean floor, J. Geophys. Res., 68, 4219–4250. Walter, M., T. Katsura, A. Kubo, T. Shinmei, O. Nishikawa, E. Ito, C. Lesher, and K. Funakoshi (2002), Spinel–garnet lherzolite transition in the system CaOMgO-Al2 O3 -SiO2 revisited: An in situ x-ray study, Geochim. Cosmochim. Acta, 66, 2109–2121. 156 Welch, M., and W. A. Crichton (2002), Compressibility of clinochlore to 8 GPa and 298 K and a comparison with micas, Eur. J. Mineral., 14, 561–565. Wessel, P. (2001), Global distribution of seamounts inferred from gridded Geosat/ERS-1 altimetry, J. Geophys. Res., 106, 19,431–19,441. Wheat, C., and M. Mottl (2004), Geochemical fluxes through mid-ocean ridge flanks, in Hydrology of the Oceanic Lithosphere, edited by E. Davis and H. Elderfield, chap. 19, pp. 627–658, Cambridge Univ. Press. White, R., D. McKenzie, and K. O’Nions (1992), Oceanic crustal thickness from seismic measurements and rare earth element inversions, J. Geophys. Res., 97, 19,683–19,715. Wiechert, U., D. Ionov, and K. Wedepohl (1997), Spinel peridotite xenoliths from the Atsagin-Dush volcano, Dariganga lava plateau, Mongolia: a record of partial melting and cryptic metasomatism in the upper mantle, Contrib. Mineral. Petrol., 126, 345–364. Williams, D., and R. Von Herzen (1974), Heat loss from the Earth: new estimate, Geology, 2, 327–328. Williams, D., K. Green, T. van Andel, R. Von Herzen, J. Dymond, and K. Crane (1979), The hydrothermal mounds of the Galapagos rift: observations with DSRV Alvin and detailed heat flow studies, J. Geophys. Res., 84, 7467–7484. Wolery, T., and N. Sleep (1976), Hydrothermal circulation and geochemical flux at mid-ocean ridges, The Journal of Geology, 84 (3), 249–275, doi:10.1086/628195. Workman, R., and S. Hart (2005), Major and trace element composition of the depleted MORB mantle (DMM), Earth Planet. Sci. Lett, 231, 53–72, doi: 10.1016/j.epsl.2004.12.005. Xu, X., S. O’Reilly, W. Griffin, X. Zhou, and X. Huang (1998), The nature of the Cenozoic lithosphere at Nushan, eastern China, in Mantle Dynamics and Plate Interactions in East Asia, Geodynamics, vol. 27, Am. Geophys. Union. Xu, Y., T. Shankland, S. Linhardt, D. Rubie, F. Langenhorst, and K. Klasinski (2004), Thermal diffusivity and conductivity of olivine, wadsleyite and ringwoodite to 20 GPa and 1373 K, Phys. Earth Planet. Int., 143–144, 321–336. Yang, H., and S. Ghose (1994), Thermal expansion, Debye temperature and Grüneisen parameter of synthetic (Fe,Mg)SiO3 orthopyroxenes, Phys. Chem. Minerals, 20, 575–586. Zhao, D. (1998), Diamonds and mantle xenoliths in kimberlites from the North China Craton and Canadian Northwest Territories, Ph.D. thesis, University of Michigan. Zheng, J., W. Griffin, S. O’Reilly, J. Yang, T. Li, M. Zhang, R. Zhang, and J. Liou (2006), Mineral chemistry of peridotites from Paleozoic, Mesozoic and Cenozoic lithosphere: constraints on mantle evolution beneath eastern China, J. Petrol., 47, 2233–2256.