Fluid Inclusions in Fibrous and Octahedrally-Grown Diamonds

FLUID INCLUSIONS IN FIBROUS AND OCTAHEDRALLY-GROWN DIAMONDS

by Evan Mathew Smith

A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF

DOCTOR OF PHILOSOPHY

The Faculty of Graduate and Postdoctoral Studies

(Geological Sciences)

THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)

March 2014

Abstract My thesis puts forth new models for diamond formation that explain the difference between octahedral and fibrous diamond growth, as well as the difference between octahedral diamond growth in the lithospheric and the sublithospheric mantle. Diamond growth in the mantle involves reactions between carbon-bearing fluid and the host rocks it infiltrates. This fluid is sometimes included in diamond. Fluids in dendritically-grown, fibrous diamonds from Wawa, Superior craton, were analysed in a novel way, using transmission X-ray diffraction. The technique allows bulk analysis of daughter minerals within fluid inclusions. The mineralogy, major and trace elements, Sr isotopes, volatiles, and nitrogen characteristics of the hydrous saline–high-Mg carbonatitic fluid in these Archean diamonds strongly resemble those of Phanerozoic fibrous diamonds. This implies that some mantle processes, including the formation of fibrous diamonds, can be extended unvaryingly back to 2.7 Ga. Fluid equilibrated with octahedrally-grown diamonds from the Siberian, Kaapvaal, and

Congo cratons is trapped in healed fractures in the diamonds. They contain anhydrous CO2–N2

fluid inclusions with 40±4 mol% N2 and inclusions of former silicate melt that had an original N2 content of ~0.1 wt%, as shown by Raman, electron microprobe, and microthermometry analyses.

The liberation of N2 from the convecting mantle is proposed to be controlled by increasing oxygen fugacity that destabilizes host phases. The observed distinct fluid compositions between hydrous fluids in fibrous and anhydrous fluids in octahedrally-grown diamond entail distinct processes of diamond formation that, ultimately, govern the growth habit. Water may trigger fibrous growth by inhibiting the expansion of {111} layers and lowering the interfacial energy between the diamond and fluid.

Certain features in diamond fluids, such as Eu anomalies and potential carbonate–CO2 isotopic fractionation, show that several mantle processes can produce geochemical signatures that may be mistaken as input from subducted materials.

The finding of N2 in diamond-forming fluids leads to an explanation for the characteristically low N content of sublithospheric diamonds. I propose this compositional trait is due to growth in a metal-saturated environment. Metallic Fe in the mantle below ~250 km should trap N and may be the largest mantle N reservoir.

Preface The research program described herein began with the initial plan to analyse fibrous diamonds with X-ray diffraction, based on the preliminary work of my research supervisor, Dr. Kopylova. The subsequent directions taken in research were identified and designed principally by myself, with the help of Dr. Kopylova. I performed all parts of the primary research and led the analysis of data. Research results were written into 4 manuscripts, 2 of which are published, and the other 2 have been submitted. These 4 manuscripts form the chapters of this thesis. The published chapters are reproduced with permission of the copyright holders. Co-authorship details are summarized below along with related conference abstracts where this content was also presented. Chapter 2 is published as: Smith, E.M., Kopylova, M.G., Dubrovinsky, L., Navon, O., Ryder, J., and Tomlinson, E.L., 2011. Transmission X-ray diffraction as a new tool for diamond fluid inclusion studies. Mineralogical Magazine, 75(5): 2657-2675. Overall, I was responsible for 90% of the research and 80% of the writing. Kopylova provided the samples, the initial plan for this investigation and helped to identify implications of the work and also helped with the organization and editing of the written manuscript. Dubrovinsky helped oversee the majority of the analyses and verified the data interpretation. Navon and Tomlinson provided diamond samples and comments on the written manuscript. Ryder provided diamond samples. This work was also presented at 3 conferences: • Smith, E., Kopylova, M., Dubrovinsky, L., and Tomlinson, E., 2010. X-ray diffraction study of the mineral and fluid inclusions in fibrous diamond. Yellowknife Geoscience Forum 2010. (poster presentation) • Smith, E., Kopylova, M., and Dubrovinsky, L., 2010. X-ray diffraction study of the mineralogy of microinclusions in fibrous diamond. GAC MAC GeoCanada 2010. (oral presentation) • Smith, E., Kopylova, M., and Dubrovinsky, L., 2010. X-ray diffraction study of the mineralogy of microinclusions in fibrous diamond. Geophysical Research Abstracts. 12: EGU 2010-4741-1 (European Geosciences Union – oral presentation)

iii

Chapter 3 is published as: Smith, E.M., Kopylova, M.G., Nowell, G.M., Pearson, D.G. and Ryder, J., 2012. Archean mantle fluids preserved in fibrous diamonds from Wawa, Superior craton. Geology, 40(12): 1071-1074. Overall, I was responsible for 60% of the research and 80% of the writing. Kopylova helped with sample acquisition, data interpretation, as well as manuscript editing and restructuring. Nowell oversaw the extensive procedures for data collection. Pearson provided help with data quality and interpretation. Ryder provided diamond samples. This work was also presented at 3 conferences: • Smith, E.M. and Kopylova, M.G., 2013. A fresh look at Eu anomalies: The effect of Cl- rich fluids. GAC MAC Winnipeg 2013. Abstract No. 142 (oral presentation) • Smith, E.M. et al., 2012. The contrast in trace element chemistry and volatile composition between fluid inclusions in fibrous and octahedral diamonds. 10th International Kimberlite Conference, Extended Abstract No. 10IKC-102. (poster presentation) • Smith, E.M., Kopylova, M.G. and Ryder, J., 2011. Fluid inclusions in Archean diamonds from Wawa, Ontario. GAC MAC Ottawa 2011, Abstract No. 95. (oral presentation)

Chapter 4 is published as: Smith, E.M., Kopylova, M.G., Frezzotti, M.L. and Afanasiev, V.P., 2014. N-rich fluid inclusions in octahedrally-grown diamond. Earth and Planetary Science Letters, 393(0): 39-48. Overall, I was responsible for 70% of the research and 80% of the writing. Kopylova helped with sample acquisition and the organization and editing of the manuscript. Frezzotti oversaw the data collection and interpretation. Afanasiev provided diamond samples. This work was also presented at 3 conferences: • Smith, E.M., Kopylova, M.G., Frezzotti, M.L. and Afanasiev, V.P., 2013. Diamond inclusions reveal fugitive mantle nitrogen. Goldschmidt 2013, Abstract ID 1601. (oral presentation - invited speaker) • Smith, E.M., Kopylova, M.G., Frezzotti, M.L. and Afanasiev, V.P., 2013. “Vapour” vs. melt inclusions in Siberian placer diamonds. GEM (Geoscience for Energy and Minerals) Geological Survey of Canada, Diamond Project Workshop, Vancouver. (oral presentation)

• Smith, E.M., Kopylova, M.G., Frezzotti, M.L. and Afanasiev, V.P., 2013. Nitrogen bubbles in the mantle: Evidence from diamond inclusions. GAC MAC Winnipeg 2013. Abstract No. 141 (oral presentation)

Chapter 5 has been accepted (March 2014) for publication in the Canadian Journal of Earth Sciences as: Smith, E.M., and Kopylova, M.G., Implications of metallic iron for diamonds and nitrogen in the sublithospheric mantle. Overall, I was responsible for 90% of the research and 70% of the writing. Kopylova provided significant help with the structure and presentation of the text and figures, as well as help with editing.

Table of contents Abstract ...... ii Preface...... iii Table of contents ...... vi List of tables ...... ix List of figures ...... x Acknowledgments...... xii 1 Introduction ...... 1 1.1 The importance of diamond research ...... 1 1.2 What is known about diamond and related fluids ...... 2 1.2.1 Metasomatic fluid as an agent of diamond growth ...... 2 1.2.2 Diamond growth habits ...... 3 1.2.3 Mantle host rocks for diamond ...... 7 1.2.4 Fluid inclusions in fibrous diamond ...... 9 1.2.5 Fluid inclusions in octahedrally-grown diamond ...... 11 1.2.6 Temporal evolution of diamond-forming fluid ...... 12 1.2.7 Implications for mantle petrology and diamond growth...... 13 1.2.8 Nitrogen and diamond...... 14 1.2.9 Fluids and experimental diamond synthesis ...... 15 1.3 Outstanding gaps in our knowledge ...... 18 1.4 Organization of thesis...... 20 1.4.1 Manuscript chapters ...... 20 1.4.2 Appendix overview ...... 21 2 Transmission X-ray diffraction as a new tool for diamond fluid inclusion studies ...... 23 2.1 Introduction ...... 23 2.2 Samples ...... 24 2.3 Methods ...... 26 2.3.1 Sample cleaning ...... 26 2.3.2 X-ray diffraction with a high-brilliance lab diffractometer ...... 26 2.3.3 Synchrotron X-ray diffraction...... 27 2.3.4 Detection limits ...... 27

2.3.5 Phase identification ...... 29 2.4 Results ...... 31 2.5 Discussion ...... 40 2.5.1 Missing daughter mineral patterns ...... 40 2.5.2 Daughter mineral species ...... 47 2.6 Summary and conclusions ...... 48 3 Archean mantle fluids preserved in fibrous diamonds from Wawa, Superior Craton ...... 50 3.1 Introduction ...... 50 3.2 Wawa metaconglomerate diamonds...... 51 3.3 Discussion ...... 55 3.3.1 Parallels between diamond formation in the Neoarchean and Phanerozoic ...... 55 3.3.2 Fluid origins from non-enriched mantle ...... 56 3.3.3 Mantle Eu anomalies and saline fluid ...... 57 3.4 Summary and conclusions ...... 58 4 N-rich fluid inclusions in octahedrally-grown diamond ...... 60 4.1 Introduction ...... 60 4.2 Diamond samples ...... 62 4.3 Analytical methods ...... 64 4.4 Results ...... 66

4.4.1 CO2–N2 fluid inclusions ...... 66

4.4.2 N2-rich silicate melt inclusions ...... 67 4.5 Discussion ...... 71 4.5.1 Nitrogen as a major mantle volatile phase ...... 71

4.5.2 Attaining high N2 contents ...... 74 4.5.3 Implications for the mantle nitrogen cycle ...... 76 4.6 Summary and conclusions ...... 81 5 Implications of metallic iron for diamonds and nitrogen in the sublithospheric mantle ...... 82 5.1 Introduction ...... 82 5.2 Siderophile character of N...... 84 5.3 Consequences of metallic Fe in the sublithospheric mantle ...... 86 5.3.1 Low N content of sublithospheric diamonds ...... 86

vii

5.3.2 Model of diamond growth in the sublithospheric mantle ...... 90 5.3.3 Metallic Fe as a host of mantle nitrogen ...... 92 5.4 Summary and conclusions ...... 94 6 Concluding discussion ...... 96 6.1 Volatiles involved in fibrous and octahedral diamond formation ...... 96 6.2 Water as a trigger for fibrous diamond habit ...... 98 6.3 Temporal evolution of water content in fluid...... 100

6.4 Behaviour of carbonate-bearing media in eclogite: Separating CO2 ...... 104 6.5 Implications for subduction signatures in the mantle ...... 107 6.6 On the incompatibility of nitrogen ...... 108 6.7 Future studies ...... 114 References ...... 119 Appendices ...... 135 Appendix A: Diamond sample catalogue...... 135 Appendix B: X-ray diffraction patterns ...... 140 Appendix C: Detailed methods for Chapter 3 ...... 156 Appendix D: Wawa electron microprobe data ...... 160 Appendix E: Trace element and Sr isotope analyses ...... 167 Appendix F: Infrared spectra...... 171 Appendix G: Microthermometry ...... 177 Appendix H: Eclogitic inclusions in Siberian diamonds ...... 179 Appendix I: Raman quantification of nitrogen ...... 182 Appendix J: Nitrogen in sublithospheric diamonds ...... 183 Appendix K: Diamond polishing equipment ...... 193 Appendix L: Dissemination of PhD research ...... 194

viii

List of tables Table 1-1. Comparison of diamond characteristics...... 7 Table 2-1. Summary of XRD results ...... 31 Table 2-2. Reference intensity ratios of some daughter mineral phases ...... 41 Table 2-3. Concentrations and mass of each mineral intersected by the X-ray beam ...... 43

Table 4-1. Raman quantification of CO2 and N2 ...... 66 Table 4-2. Electron microprobe data for melt inclusions in the Congo diamond ...... 69

List of figures Figure 1-1. Examples of common diamond morphology ...... 4 Figure 1-2. Fluid inclusions in fibrous diamond...... 5 Figure 1-3. Diamond growth mechanisms operating at a {111} interface ...... 6 Figure 1-4. Mineralogical phase changes with depth for pyrolite and MORB ...... 8 Figure 1-5. Composition of fluid in fibrous diamonds worldwide ...... 10 Figure 2-1. Representative sample images for all suites studied ...... 25 Figure 2-2. Diffraction patterns for 10 µm3 and 100 µm3 of corundum powder ...... 28 Figure 2-3. Examples of diffraction pattern textures ...... 30 Figure 2-4. Two-dimensional diffraction patterns ...... 36 Figure 2-5. Diffraction pattern and integrated profile ...... 37 Figure 2-6. Diffraction patterns showing the unidentified phase ...... 39 Figure 3-1. Molar composition of fluid inclusions in Wawa (Canada) fibrous diamonds ...... 51 Figure 3-2. Infrared spectra of Wawa (Canada) diamonds, shifted vertically for clarity ...... 53 Figure 3-3. Trace element patterns of Wawa (Canada) fibrous diamonds ...... 54 87 86 Figure 3-4. Calculated Sr/ Sri for Wawa (Canada) diamonds at 2.7 Ga ...... 54 Figure 3-5. Fibrous diamond Eu anomalies, grouped by fluid composition ...... 58 Figure 4-1. Octahedrally-grown diamonds containing the studied fluid and melt inclusions ...... 62 Figure 4-2. Studied inclusions in diamonds ...... 64 Figure 4-3. Representative Raman spectra for fluid inclusions ...... 67 Figure 4-4. Compositional range measured in 3 melt inclusions...... 70 Figure 4-5. Comparison of C/N ratios and N concentrations in diamonds ...... 72 Figure 4-6. Redox-controlled nitrogen escape ...... 79 Figure 5-1. Schematic mantle section ...... 83 Figure 5-2. Electron backscatter image of metallic Fe-Ni blebs ...... 84 Figure 5-3. Nitrogen concentration distributions for inclusion-bearing diamonds ...... 87 Figure 6-1. Synthetically grown dendritic diamond ...... 99 Figure 6-2. Isobaric phase diagrams, for a generalized rock system ...... 102 Figure 6-3. Carbon phase changes with oxygen fugacity and temperature ...... 105 Figure 6-4. Carbon isotopic characteristics of eclogitic diamonds ...... 106 Figure 6-5. Variations in diamond N concentration versus δ13C...... 112

Figure 6-6. Nitrogen concentration and aggregation diagram ...... 117

Acknowledgments The past several years of research have been motivated and supported by many people, to all of whom I owe my gratitude. First and foremost I thank Dr. Maya Kopylova, my supervisor, for her guidance and wisdom. She has invested a great deal into assuring my success. She has been a valuable asset to my research by being able to navigate and incorporate Russian diamond literature, as well as draw upon her broad professional network. One such connection, Dr. Leonid Dubrovinsky, was instrumental in the initial stages of research, as a secondary supervisor at the University of Bayreuth. He generously provided his XRD laboratory for my use over several months. I also thank the other members of my supervisory committee, Dr. Kelly Russell and Dr. Mati Raudsepp, for their help in refining my research plan and their counsel regarding obstacles I encountered. Tackling a PhD project depends heavily on having adequate funding. I am thankful to have been provided with this support and I recognize that it is a privilege to be able to pursue one's own research questions in detail. Financial support has largely been provided by the Natural Science and Engineering Research Council of Canada (NSERC) through a Canadian Graduate Scholarship. NSERC has also provided support through the form of research grants to Dr. Kopylova. Further research grants were provided to me by the Society of Economic Geologists and to L. Dubrovinsky by the German Research Council, which allowed me to conduct research abroad. I also thank the U.S. Department of Energy for granting time at the Advanced Photon Source, and the GSECARS support staff for ensuring a productive visit. Research support for Raman spectroscopy was provided by the Italian Research Group for Antarctica. Diamond samples were freely provided by several sources, including John Ryder (Dianor Resources Ltd.), Emma Tomlinson, Valentin Afanasiev, Elena Deljanin, Maureen Morrison, and Diavik Diamond Mines Ltd. A suite of diamonds was also borrowed from Adrian Van Rythoven and Dan Schulze, although these did not end up being studied. Analyses of diamond samples were made possible with the expert help of: Leonid Dubrovinsky and Matthew Newville, for X- ray diffraction; Geoff Nowell, for trace element and Sr isotope determinations; Dan Marshall and Andrew Beeby, for valuable preliminary Raman spectroscopy; Mati Raudsepp and Edith Czech, for electron microprobe analyses; Thomas Stachel, Zhihai Zhang, and Amir Mehdi Dehkhoda, for infrared spectroscopy; and Maria Luce Frezzotti, for detailed Raman spectroscopy.

xii

Through my years in the Diamond Exploration Lab at UBC it has been a pleasure to work alongside Wren Bruce, Christine Miller, Yvette Beausoleil, Bram van Straaten, Andrea De Stefano, Dave Newton, Chuck Kosman, and Matt Gaudet, as well as all the members of our sister lab, the VPL. They have fostered a positive and energetic atmosphere. Many thanks to the entire EOAS department for the sense of community that I have found here. I offer my utmost gratitude to my friends and family who have encouraged me. The support and enthusiasm of my family has given me much needed motivation. Ultimately, I am grateful for the mysteries that make us dream and the people who make us dreamers.

xiii

1 Introduction

1.1 The importance of diamond research Diamond is best known as an economic commodity, most highly valued as a gemstone, but diamond also carries valuable information about its genetic history and environment of formation in the Earth's mantle, both in the form of inclusions and the diamond crystal itself. Its strength, high melting point, and chemical inertness allow diamond to remain preserved for billions of years and survive volcanic transport to surface. Studying natural diamond, in my opinion, is one of the most promising ways to understand the mantle environment and its evolution through time. Like other minerals, diamonds can entrain samples of its genetic medium during growth, in the form of fluid inclusions. These direct samples of the natural diamond-forming fluid are critical for understanding diamond genesis. In an applied sense, understanding this process would aid diamond exploration efforts. Genetic models form a theoretical basis to guide mineral exploration. Currently, indicator minerals associated with diamonds provide a means of targeting prospective diamond deposits. Recognizing an association between diamond and certain minerals like G10 garnets was a large step forward for diamond exploration (Gurney et al., 1993). However, the genetic reason for those associations, like the G10–diamond relationship, remains unclear (Malkovets et al., 2007; Klein-BenDavid and Pearson, 2009). As our understanding of diamond genesis improves, so too should our ability to recognize prospective mantle regions where diamonds are likely to have formed. Beyond economic incentives, diamond genesis sheds light on the deep carbon cycle. Diamond formation in the mantle immobilizes carbon, forming a sink in the carbon cycle. Mantle carbon can be mobilized or immobilized according to its redox state, sometimes called "redox melting/freezing" (Rohrbach and Schmidt, 2011). More generally, diamond formation can be considered a product of mantle metasomatism. In this sense, the fluid inclusions they contain are not merely diamond-forming agents, but are also agents of upper mantle refertilization (Weiss et al., 2011; Miller et al., 2013), and crucial pathways in the cycles of mantle volatiles like nitrogen. The fluid composition, temperature, and flux may lend insight into how mantle 1

fluids migrate (Iwamori, 1998; Kelemen et al., 2000). For example, fluid temperature and composition affect the dihedral angle, thereby controlling the connectivity of fluid networks (Luth, 2005) and therefore the degree of melting of the protolith and the composition of incipient melt. As diamonds have formed episodically in the cratonic lithospheric mantle since at least 3.5 Ga (Gurney et al., 2010), diamonds may provide unique long-term records of mantle fluids. Below I summarize the current knowledge of diamond formation, the solid and fluid media of this formation, and methods used by diamond geologists.

1.2 What is known about diamond and related fluids

1.2.1 Metasomatic fluid as an agent of diamond growth Diamond growth in the mantle is regarded as a metasomatic process, whereby mobile carbon, as hydrocarbons or carbonate, percolates into a host rock and consequent redox reactions oxidize or reduce, respectively, the carbon to diamond (Deines, 1980; Luth, 1993; Gurney et al., 2005). The details of this process depend on both the nature of the fluid or melt and the host rock. For the majority (~99%) of mined diamonds, their environment of formation is in thick, ancient portions of continental lithospheric mantle, sometimes called cratonic roots, where the lithosphere intersects the diamond stability field. Evidence for a metasomatic origin of diamond is abundant and includes trace element compositions of diamond inclusions (Stachel and Harris, 1997a; Stachel et al., 2004; Tomlinson et al., 2009), isotope geochemistry (Klein-BenDavid and Pearson, 2009; Klein-BenDavid et al., 2010), textural relationships within diamondiferous xenoliths (Taylor et al., 2000; Spetsius and Taylor, 2002), as well as direct observation of diamond-forming agents preserved as fluid inclusions in diamonds (e.g. Navon et al., 1988). The diamond-producing metasomatic agent is often referred to as a fluid. For diamond- bearing mantle (>150 km), the high pressure is near the second critical end-point of the solidus in

a lherzolite-H2O-CO2 system, where the phase boundary between melt and hydrous vapour ceases to exist (Wyllie and Ryabchikov, 2000). There is likely to be a continuum between melt and vapour during some diamond-forming events (Stachel and Harris, 2008). Therefore distinguishing between melt, vapour, or miscible mixtures of the two at such conditions can be ambiguous, and mantle petrologists use the term “fluid” (Wyllie and Ryabchikov, 2000). This is 2- especially useful for carbonatitic fluids, which are rich in CO3 complexes, dissolved or molten, 2

below or above the melting temperature. Based on final crystallization products of inclusions formed from such fluid, i.e. carbonates, it is “almost impossible to distinguish between carbonatitic melt and hydrothermal fluid” (Veksler and Lentz, 2006). In keeping up with the traditional use of “fluid” in mantle petrology, the term fluid is used in this thesis to denote mobile, carbon-bearing phases percolating into solid host rocks.

1.2.2 Diamond growth habits Fibrous diamond is a variety of diamond produced by dendritic growth (Figure 1-1 e-h), which tends to incorporate many sub-micron sized inclusions of the growth medium (Figure 1-2). In fact it is the only diamond variety that readily contains fluid inclusions. Fibrous diamond is thought to grow rapidly, under conditions of high chemical driving force (Figure 1-3) (Sunagawa, 1984; Gurney, 1989; Boyd et al., 1994). The exact mechanism(s) that governs growth habit is not known. One clue comes from synthetic diamond growth in metal catalysts, where increasing water content triggers dendritic growth, although this cannot be directly applied to nature (Kanda, 1985; Palyanov et al., 2012). Octahedrally-grown diamond is the more common counterpart to fibrous diamond (Figure 1-1), which is believed to grow more slowly, at conditions of lower chemical driving force where crystal growth dominates over nucleation (Tolansky and Sunagawa, 1959; Sunagawa, 1984). The term octahedrally-grown means the diamond crystallized by smooth growth in {111} layers (Figure 1-3). The eventual morphology of octahedrally-grown diamond is not necessarily an octahedron. They are commonly subject to modification by twinning and/or resorption (Figure 1-1). Resorption is a dissolution process that can occur at depth in the mantle or during volcanic transport to surface (Orlov, 1977). Polycrystalline aggregates are also not uncommon.

a b c e f

g h

octahedron dodec.

mm scale macle Figure 1-1. Examples of common diamond morphology. The most basic form is an octahedron (a), bound by flat {111} growth faces. Diamonds are commonly resorbed by fluids in the mantle and/or by interaction with the kimberlite during transport to surface (Fedortchouk et al., 2007). Resorption dissolves the edges of the octahedral faces, producing curviplanar surfaces with nearly {110} orientation (see cartoon). With progressive resorption, the diamond morphology changes from (a) octahedron, to (b) resorbed octahedron, to (c) dodecahedroid (also called tetrahexahedroid, as the rhomboid "faces" are often bisected by an irregular edge). Macles are a twinned variation of the octahedral morphology. Macles can also be resorbed (d) into a triangular lensoid shape. e, f, g, h) show the appearance of fibrous diamond. Octahedra can be overgrown by a fibrous diamond coat (e, g). In the absence of a pre-existing octahedral core, fibrous diamonds grow as cuboids (f, h). Note the turbid, greyish appearance of fibrous diamond. All examples of non-fibrous diamond above (a-d, and the cores within e and g) are examples of octahedrally-grown diamond. Diamonds shown are from the Diavik diamond mine, except for (d), which is of unspecified African origin.

a b

[010] [100] [010] [100] 1 mm 100 µm

c d Cl

Na halite sylvite

Figure 1-2. Fluid inclusions in fibrous diamond. a, b) Polished sections of fibrous diamond in transmitted light to show the distribution of the sub-micron-sized inclusions. Both sections are oriented with the cube plane (001) within the page. The inclusions are aligned in trails parallel to the <111> growth fibres. In the cuboid diamond (a), fibres appear to run at 45° angle to the exterior cuboid faces. In (b), showing part of a fibrous coat, the fibres run left-right. They grew on an octahedral core and have been deposited in multiple (111) layers, coinciding with the dark vertical bands. c) Electron backscatter image of a polished surface of fibrous diamond. Each faint bright spot is an inclusion (<1 µm). d) A larger fluid inclusion exposed at the diamond surface, with Cl, Na, K chemical maps (left), showing these are chloride daughter minerals. The dark region surrounding the cubic crystals would have been occupied by residual, hydrous fluid.

Figure 1-3. Diamond growth mechanisms operating at a {111} interface. The dominant mechanism is determined by the chemical driving force. Higher driving forces lead to the dominance of mechanisms with higher growth rates. At low driving forces, growth proceeds on smooth {111} faces and produces octahedrally-grown diamond. At high driving force in this diagram, growth becomes dendritic. Dendritic growth fibres advance and branch along <111> vectors, perpendicular to {111} surfaces. Dendritic or fibrous growth leads to cuboid morphologies with rough faces. Fluid and mineral inclusions are trapped between fibres as they grow and branch. Diagram is based on Sunagawa (1984). Crystal sketches are from Goldschmidt (Atlas der Krystallformen, 1913-1923).

There are some important differences between fibrous and octahedrally-grown diamond, summarized in Table 1-1. In addition to their obvious differences in appearance (Figure 1-1), and nitrogen (N) content, the two habits tend to form at different times. Fibrous diamond commonly occurs as a coat on an octahedrally-grown core, but the reverse situation is rare. Fibrous diamonds are generally much younger than octahedrally grown diamond. This age discrepancy is based on nitrogen characteristics. When diamond grows, it often incorporates N in the crystal lattice, where it substitutes for carbon. Nitrogen is the most common impurity in diamond. The N atoms are initially isolated, as singular N defects called C centres. N atoms quickly aggregate, as a function of temperature and time, to form N-N pairs called A centres (Evans, 1992). Normally, C centres are rare in natural diamond. When they are found, they indicate that the diamond spent only a brief time in the mantle (millions of years or less) or were stored at exceptionally cool temperatures (Taylor et al., 1996). With additional time and heat, A centres (N-N pairs) further aggregate into groups of 4 N atoms plus a vacancy, called B centres (Evans, 1992).

Nitrogen is generally poorly aggregated in fibrous diamond, being predominantly A centres and some C centres. This immature aggregation state is taken as an indication that fibrous diamonds cannot have spent long at the high temperatures of the mantle (Boyd et al., 1987; Boyd et al., 1994; Gurney et al., 2010). A caveat here is that models of aggregation kinetics (e.g. Taylor et al., 1990; Taylor et al., 1996) are based solely on octahedrally-grown diamond. Aggregation kinetics in fibrous diamond might actually be slightly different (Zedgenizov et al., 2006; Howell et al., 2012). Nevertheless, the poorly aggregated N in fibrous diamond is consistent with relatively short mantle residence times. For this reason, fibrous diamonds are thought to have grown shortly before volcanic sampling and transport to the surface.

Table 1-1. Comparison of diamond characteristics. Fibrous diamond Octahedrally-grown diamond Fluid inclusion content abundant sub-micron fluid (and rarely contains fluid (or melt) mineral) inclusions, giving a turbid inclusions appearance to the diamond Nitrogen content typically 600–1600 ppm (Cartigny, typically <600 ppm (Cartigny, 2005) 2005; Stachel and Harris, 2009) Mantle residence time <5 Ma, based on N aggregation often up to several Ga (Gurney et state (Boyd et al., 1994; Gurney et al., 2010) al., 2010) Age typically Phanerozoic, close to age typically 3.5–1.0 Ga (Gurney et al., of hosting kimberlite 2010) Age relationship between • fibrous growth is almost always younger fibrous and octahedral • fibrous coat over pre-existing octahedral diamond is common growth • the switch from octahedral to fibrous growth is accompanied by a change in N aggregation state, that indicates the coat is much younger (Boyd et al., 1987) • very rarely, fibrous diamond may be overgrown by octahedral diamond (Zedgenizov et al., 2006; Rondeau et al., 2007)

1.2.3 Mantle host rocks for diamond Studies of mineral inclusions in diamonds provided some of the initial evidence that they did not grow as phenocrysts in kimberlite magmas. Instead, the inclusions indicate that diamonds have grown predominantly within peridotite and eclogite host rocks. The common mineral inclusions of the peridotitic suite are olivine, Cr-pyrope, orthopyroxene, Cr-diopside, Mg- chromite, and Ni-Fe sulfides, whereas common inclusions of the eclogitic suite are grossular- almandine-pyrope, omphacitic clinopyroxene, rutile, and coesite (Stachel and Harris, 2008). The

inferred host rocks from mineral inclusion studies are verified by the occurrence of diamondiferous xenoliths. In addition to eclogite and peridotite, a small proportion (~2%) of inclusions-bearing diamonds have characteristics consistent with a websteritic host, but this suite is not as well defined (Stachel, 2007). In the convecting mantle below the lithosphere, the mineralogy changes with increasing pressure and temperature. Rare diamonds derived from the sublithospheric mantle contain mineral inclusions that, like lithospheric diamonds, also reflect host rocks with bulk mafic and ultramafic compositions, although the mineralogy differs according to phase changes with depth (Figure 1-4) (Stachel et al., 2005; Kaminsky, 2012). Pyrolite MORB

Opx Depth Cratonic (km) lithosphere HP Garnet Cpx Clinopyroxene 200 Garnet Olivine Cpx Coesite

“Asthenosphere” 400

Wadsleyite

Majorite Majorite

Ringwoodite zone ransition 600 T Convecting mantle MgSi-perovskite

MgSi- 800 Stishovite Fe-periclase Cf

perovskite L Lower mantle Lower NA CaSi-perovskite CaSi-perovskite

0 20 40 60 80 100 0 20 40 60 80 100 Mineral proportions (wt.%) Mineral proportions (wt.%) Figure 1-4. Mineralogical phase changes with depth for pyrolite and MORB (mid-ocean ridge basalt) compositions. Pyrolite is ultramafic and MORB is mafic. The base of cratonic (ancient, continental) lithosphere varies, but is shown at 200 km for simplicity. Opx: orthopyroxene; Cpx: clinopyroxene; HP Cpx: high-pressure clinopyroxene phase; Cf: calcium ferrite type structure; NAL: new aluminous phase. The pyrolite (left) panel is after Irifune and Tsuchiya (2007). The MORB (centre) panel is after Perrillat et al. (2006). The depth divisions into lithosphere, "asthenosphere," transition zone, and lower mantle are after Stachel et al. (2005).

1.2.4 Fluid inclusions in fibrous diamond: Analytical techniques and fluid compositions To date, most studies of fluid inclusions in diamond have dealt with fibrous diamond. Several methods have been used to study these fluid inclusions. After cooling from mantle conditions, the inclusions contain residual fluid along with daughter minerals that have crystallized from the fluid, including micas, carbonates, chlorides, apatite, and quartz (Navon et al., 1988; Guthrie et al., 1991; Walmsley and Lang, 1992; Klein-BenDavid et al., 2006; Kopylova et al., 2010). The most popular technique for analysing fibrous diamond microinclusions has been electron microprobe analysis (EPMA). Each analysis is of a single inclusion. A fibrous diamond may contain millions of inclusions, so many analyses must be performed to build up a statistically representative sample of the bulk fluid composition. Analyses are performed on polished diamond surfaces, to expose the interior of the diamond. Exposed inclusions may have lost some contents or been contaminated during polishing. For these reasons, inclusions lying just below the diamond surface are analysed. They are pristine and protected by diamond. The oxide total from each analysis is much less than 100%, often less than 5%. Several factors are responsible. Firstly, the inclusions are small (<1 µm), so they only make up part of the volume excited by the electron beam. Secondly, the inclusions are subsurface, so they are not efficiently excited by the electron beam and emitted X-rays are partly attenuated. Thirdly, the inclusions are rich in carbonate and water, neither of which will contribute to the oxide totals. The variable shape and heterogeneous composition of the inclusions also complicates measurement. Despite these complications, the data collected are satisfactory and the accuracy is better than 15% for major elements (Izraeli et al., 2004; Weiss et al., 2008). The fluid inclusions define two compositional ranges (Figure 1-5), from silicic (Si, Al, and water rich) to low-Mg carbonatitic (Ca and carbonate rich) and from saline (K, Cl, and water rich) to high-Mg carbonatitic (Ca, Mg, and carbonate rich) (Izraeli et al., 2001; Klein-BenDavid et al., 2009; Weiss et al., 2009b). The Mg content differs toward carbonate-rich end of each trend, despite overlapping in the commonly used ternary plot (Weiss et al., 2009b).

a Si+Al b

silicic 40 saline

–low-Mg carbonatitic –high-Mg carbonatitic 30

20 Cl (wt%) 10

0 0 10 20 30 MgO (wt%) K+Na saline–high-Mg carbonatitic

Ca+Mg+Fe

Figure 1-5. Composition of fluid in fibrous diamonds worldwide. a) The most common ternary diagram used to separate fibrous diamond fluid molar compositions into silicic (Si+Al), carbonatitic (Ca+Mg+Fe), and saline (K+Na) end members. Data in (a) compiled from (Zedgenizov et al., 2007a; Klein-BenDavid et al., 2009 and references therein; Kopylova et al., 2010; Smith et al., 2012a). b) Cl versus MgO for saline–high-Mg carbonatitic (blue field) and silicic–low-Mg carbonatitic trends (green field). The two compositional trends are distinct, despite overlapping in the commonly used ternary diagram. Fields shown in (b) are reproduced from (Weiss et al., 2009b).

In addition to EPMA, transmission electron microscopy (TEM), and infrared spectroscopy (FTIR) have proven useful for studying fluid inclusions in fibrous diamond. TEM can measure chemical compositions of individual daughter minerals in the fluid inclusions, and TEM electron diffraction can identify crystal structures, but only one microinclusion can be analysed at a time (Klein-BenDavid et al., 2006; Logvinova et al., 2008). Any variation in composition complicates the process. FTIR, however, provides a bulk analysis that incorporates many inclusions into one spectrum. A spectrum reveals molecular vibrational groups, which correspond to minerals crystallized within the fluid inclusions (Weiss et al., 2010). However, mineral identification is difficult, due to overlap between mineral peaks and the large nitrogen absorption bands of the diamond lattice (Kopylova et al., 2010), and also due to the limited reference spectra for complex carbonates and silicates at high pressure (Navon, 1991). In addition, the chloride minerals which prevail in saline inclusions are transparent to infrared and therefore undetectable.

Trace elements have been measured in diamond fluid inclusions using laser ablation inductively-coupled plasma mass spectrometry (LA-ICPMS), both by on-line (Rege et al., 2010) and off-line (Klein-BenDavid et al., 2010) ablation techniques. The off-line technique has the advantage of lowering minimum detection limits and allowing analysis of radiogenic isotopes from the same sample (Klein-BenDavid et al., 2010). The largest study of fibrous diamonds from various localities worldwide showed two general groups of trace element patterns: a group high in light rare earth elements (LREE), Ba, and K, and a group with depletion in LREE, Ba, and K, and negative Sr and Y anomalies (Rege et al., 2010). The pattern type is not obviously correlated with fluid major element composition, although saline–carbonatitic fluid inclusions more often have high-LREE patterns and silicic fluids more often have lower LREE patterns (Rege et al., 2010). Other trace element studies of fibrous diamonds have shown similar enrichment and depletion features, although there is much variation and interpretations are not straightforward (Schrauder et al., 1996; Zedgenizov et al., 2007b; Tomlinson et al., 2009; Weiss et al., 2009b; Klein-BenDavid et al., 2010). To complement trace elements, radiogenic isotopes incorporate time-integrated fluid evolution into the picture. Using off-line laser ablation to collect a large sample, an aliquot of the sample can be used for radiogenic isotope analyses, such as using thermal ionization mass spectrometry (TIMS) to measure 87Sr/86Sr isotopic ratios (McNeill et al., 2009; Klein-BenDavid et al., 2010).

1.2.5 Fluid inclusions in octahedrally-grown diamond For fluid inclusions in octahedrally-grown diamond, there has been only one previous, unambiguous study of genuine fluid inclusions. Tomilenko et al. (1997) reported relatively large (1–35 µm across) fluid inclusions, compared to the sub-micron-sized inclusions in fibrous diamond. The inclusions were studied using microthermometry and laser Raman spectroscopy

and concluded to contain CO2, N2, and hydrocarbons. However, the Raman spectroscopy was

only able to identify N2, as the spectral resolution was too low to separate potential CO2 peaks from the diamond peak. Raman also failed to detect hydrocarbon species. What Tomilenko et al. described as a frozen hydrocarbon phase in microthermometry may have actually been a vapour

bubble in a high-density CO2–N2 mixture (Van den Kerkhof and Thiéry, 2001). In such a

mixture, the phase boundaries can be so subtle that they cannot be reliably recognized in some samples. Microthermometry is difficult to apply to inclusions of unknown composition, especially at the high fluid densities that should be expected for mantle-derived fluid inclusions. Raman spectroscopy should be the most useful technique for studying these inclusions, provided the instrument setup is optimized for the application. The comparison of fluids equilibrated with fibrous and octahedrally-grown diamond is undoubtedly complicated by the scarcity of fluids found in octahedral diamonds. It is possible to collect bulk trace element analyses of non-fibrous, gem diamond without any visible inclusions. Initially, these patterns were thought to reflect incorporation of miniscule amounts of diamond- forming fluids (McNeill et al., 2009; Rege et al., 2010). However, syngenetic garnet inclusions indicate that the true growth medium must have had higher LREE enrichment and higher Ce/Eu and Ce/Ti ratios than what is shown by the trace element patterns of the octahedrally-grown diamond host (Melton et al., 2012). Thus, the trace element patterns of octahedrally-grown diamond do not directly reflect the diamond growth medium. This leaves researchers with only indirect ways to gain insight into the composition of the parental fluids of octahedrally-grown diamond. Trace element characteristics of this fluid have been reconstructed based on experimentally-determined partition coefficients and trace element patterns of mineral inclusions in octahedrally-grown diamonds. The reconstructions suggest the metasomatic fluid responsible for octahedrally-grown diamond may be similar to the fluids found in fibrous diamond (Tomlinson et al., 2009; Weiss et al., 2009b).

1.2.6 Temporal evolution of diamond-forming fluid Changes in the composition of the diamond-forming fluid have been documented within progressive layers of fibrous diamond growth. Progressions toward carbonatitic compositions within the saline–high-Mg carbonatitic trend (Klein-BenDavid et al., 2004) and progressions toward or away from carbonatitic compositions within the silicic–low-Mg carbonatitic trend (Weiss et al., 2009b; Kopylova et al., 2010) have been measured using EPMA. Abrupt changes in trace element patterns from core-to-rim in fibrous diamonds have also been measured and were interpreted as a change from high-Mg to low-Mg carbonatitic fluid composition, although the accompanying major elements were not analysed (Rege et al., 2010). These features in

fibrous diamond demonstrate the ability of diamond-forming fluids to change character over the course of diamond growth, either due to evolution of a single fluid or influx of new, distinct fluids. Temporal changes may also be reflected in the contrasting nature of primary and secondary fluid inclusions within a population of non-fibrous, Siberian alluvial diamonds. Within these diamonds, hydrous-carbonatitic fluid is trapped along grain boundary trails, similar in character to fibrous diamond fluid inclusions (Logvinova et al., 2011). A later generation of fluid

inclusions, in healed fractures, described as CO2, N2 and possibly hydrocarbon mixtures (Tomilenko et al., 1997), show that fluids in this system have evolved in time. On a larger time scale, Stachel and Harris (2009) have suggested that there has been a temporal shift in the carbon-bearing fluid species that form diamonds, from methane-based in the Archean, to carbonate-based in the Proterozoic onward. This possibility is worth considering, given the episodic rather than continuous nature of diamond growth through time and the change in mantle geodynamics that might accompany the start of modern subduction around 3 Ga (Shirey and Richardson, 2011). In fact many octahedrally-grown diamonds formed in a relatively narrow interval of 3.2–3.5 Ga (Gurney et al., 2010). These diamonds are harzburgitic, having been grown in a clinopyroxene-poor peridotite, and make up about 50–60% of lithospheric diamonds (Stachel and Harris, 2008).

1.2.7 Implications for mantle petrology and diamond growth A comparison of major element compositions of fibrous diamond fluid inclusions and coexisting mineral inclusions suggests that these saline-carbonatitic-silicic fluids may be responsible for some metasomatic refertilization of the lithospheric mantle. Mineral inclusions in fibrous diamonds have been metasomatised by the fluid, causing Ca and Fe enrichment in peridotitic garnet and pyroxene and a decrease in olivine Mg number (Tomlinson et al., 2006; Miller et al., 2013). As agents of mantle metasomatism, trace elements in the fluids in fibrous diamonds further show that they are efficient carriers of incompatible elements. Rare earth element (REE) contents of the fluid inclusions exhibit similar LREE enrichment over the heavy REE's typical of kimberlites and carbonatites, along with negative anomalies in Zr, Hf, and Ti (Tomlinson et al.,

2009; Weiss et al., 2009b; Klein-BenDavid et al., 2010). In particular, the high-Mg carbonatitic variety of fibrous diamond fluids most closely resembles Group I kimberlites and average compositions of post-Archean xenoliths (Weiss et al., 2011). Weiss et al. (2011) proposed that the three are genetically linked, with one possibility being that the addition of small amounts of diamond-forming fluids or kimberlite magma could produce the patterns of post-Archean, metasomatised lithosphere, represented by xenoliths. However, despite their similar trace element features, fibrous diamond fluids ultimately contain much higher concentrations of K and volatiles than kimberlites or carbonatites (Klein-BenDavid et al., 2009). On the basis of trace element patterns and associations with major element composition of fluid inclusions, Weiss et al. (2009b) proposed a model whereby saline fluid interaction with peridotite leads to high-Mg carbonatitic fluids, and fluid interaction of a hydrous, K-rich fluid with eclogite leads to the silicic to low-Mg carbonatitic compositional array. However, if these fluid compositions depend on interaction with peridotite versus eclogite, the dependence is limited to fluid genesis. When the fluid crystallizes diamond, it has been shown that its major element (Tomlinson et al., 2006) and trace element (Rege et al., 2010) composition does not correlate with host rock type. Within silicic to low-Mg carbonatitic fluids, there is evidence for mixing of at least two distinct sources with unique 87Sr/86Sr ratios isotopes (Klein-BenDavid et al., 2010). Klein- BenDavid et al. (2010) interpreted the range of 87Sr/86Sr ratios as a mixture between an ancient, radiogenic component derived from the breakdown of hydrous, K-rich, mica-rich metasomes in the lithosphere and a carbonatitic type of fluid from the convecting mantle. In all cases, the need for a prominent convecting mantle input is made clear by the fact that C and N isotopic signatures of fibrous diamond are restricted and lie close to convecting mantle values (Cartigny, 2005).

1.2.8 Nitrogen and diamond There has long been debate about the concentration of N in diamond-forming media, and its compatibility within diamond (Cartigny et al., 2001; Thomassot et al., 2007; Stachel et al., 2009). Cartigny et al. (2001) argue that N is incompatible in diamond and that its uptake is controlled by kinetics, explaining why N concentration frequency distributions for octahedrally-

grown diamond tend toward values <50 ppm and rapidly-grown, fibrous diamond has much higher N concentrations. However, Stachel et al. (2009) argue that N is compatible, based on modelling N concentration as a function of δ13C during C isotopic fraction between the diamond and growth medium. It is not yet clear whether differences in N content of diamond, such as the drastic difference between fibrous and octahedral diamond (Table 1-1) reflect growth kinetics or the concentration of N in the diamond-forming medium. If diamond growth media can be relatively N-rich (e.g. >1 mol%), like the fluids described by Tomilenko et al. (1997), then growth kinetics or host rock effects must impose controls that limit N uptake into the diamond, which seldom have concentrations over 2000 ppm, or 0.2 mol% (Stachel et al., 2009). Another difference in N content is observed between diamonds formed in the lithosphere and those from the sublithospheric, convecting upper/lower mantle. Sublithospheric diamonds contain significantly less N (e.g. Davies et al., 1999a), which remains an unexplained curiosity. This difference suggests that host rock specific behaviour could also be an important consideration in deciphering processes of diamond formation. The isotopic composition of N in both fibrous and octahedrally-grown diamond mimics the negative δ15N values of mid-ocean ridge basalts, suggesting the N is derived from the mantle and does not support a contribution from heavier subducted, crustal N with δ15N>0 (Cartigny, 2005). Cartigny (2005) notes that negative δ15N values in eclogitic diamonds therefore do not support the view that their low δ13C values are due to subducted organic C (e.g. Sobolev and Sobolev, 1980; McCandless and Gurney, 1997). The debate over low δ13C remains open, however, with new data from oxygen isotopes appearing to favour a subducted origin (Ickert et al., 2013).

1.2.9 Fluids and experimental diamond synthesis Experiments have helped to clarify diamond growth and dissolution interactions with different types of fluids. Variations in the composition and temperature-pressure conditions of the growth medium can lead to a range of diamond morphologies (Sunagawa, 1984; Palyanov et al., 2012), which suggests the morphology of a natural diamond communicates something of its

growth history. Experiments have also shown how other expected natural fluids like H2O and

CO2 can dissolve diamond and reproduce the kinds of resorption features exhibited by natural samples (Fedortchouk et al., 2007; Zhang and Fedortchouk, 2012). Significant versatility has been demonstrated in the ability of diamond to crystallize from a broad range of media, including C-O-H fluids, sulfide melts, transition metal melts (e.g. Fe, Ni, Co) and carbonatitic–silicic–hydrous–chloride melts (Bundy et al., 1955; Strong and Hanneman, 1967; Arima et al., 1993; Kumar et al., 2000; Borzdov et al., 2002; Pal'yanov et al., 2002a; Sokol and Pal'yanov, 2008; Spivak et al., 2008; Pal'yanov and Sokol, 2009; Palyanov et al., 2010; Bureau et al., 2012). Many of these experiments use graphite as the C source for diamond, through dissolution and precipitation at high temperature and pressure, within the diamond stability field. Although graphite is not necessarily thought to be the C source for natural diamond, such experiments are instructive for showing how diamond nucleation and growth are affected by the composition of the growth medium. Within C-O-H fluids, diamond crystallization is most intense in H2O-rich compositions, whereas no diamond formation is

observed in H2-rich fluids (Sokol et al., 2009). Within sulfur-based media, diamond has been grown both from molten sulfur (Pal'yanov et al., 2001) and from Fe-sulfide melt (Shushkanova and Litvin, 2006; Shushkanova and Litvin, 2008; Spivak et al., 2008). The relatively high nucleation density, reaching 105 mm-3, observed for diamond growth in sulfide melt suggests it may play a role in the growth of natural polycrystalline diamond (Shushkanova and Litvin, 2008). In comparison, much more work has been done on diamond synthesis in metal catalysts, especially (Fe,Ni) alloys, for industrial and technical applications (Field, 1992b). The participation of (Fe,Ni) metal in some natural diamond formation is suggested by metallic (Fe,Ni) inclusions in natural diamond (Hayman et al., 2005; Bulanova et al., 2010; Gurney et al., 2010; Kaminsky and Wirth, 2011) and further supported for sublithospheric mantle settings, which may be saturated in (Fe,Ni) metal (Rohrbach et al., 2007; Rohrbach et al., 2011). In efforts to more closely simulate the composition of fluid inclusions observed in fibrous diamonds, many recent experiments have focused on carbonatitic–silicic–hydrous–chloride diamond-forming media. The effect of growth medium composition on diamond crystallization behaviour is complex. In the simplest case, diamond growth can occur readily in Ca- and Mg- carbonate melts and is enhanced by the presence of CO2 and H2O (Sokol et al., 2000). The

addition of CO2 and H2O also promote diamond growth in alkaline carbonate melts (Pal'yanov et 16

al., 2002a). Within alkaline carbonate systems, Na and K have been noted to have differing effects on the morphology and growth rate of diamond. Faster, cubo-octahedral and slower, octahedral diamond growth are facilitated by Na-carbonate and K-carbonate, respectively (Pal'yanov et al., 2002a). The addition of KCl to K-carbonate melt enhances diamond crystallization by shortening the induction period, possibly due to the effect of Cl destabilizing the carbonate groups (Tomlinson et al., 2004). Similarly, KCl enhances diamond growth in hydrous silicate melts, although the addition of NaCl appears to inhibit diamond growth (Fagan and Luth, 2011). The different diamond growth behavior for KCl and NaCl may be related to the contrasting diamond morphologies noted by Pal’yanov et al. (2002a) for growth in K- and Na- carbonate. As in carbonate, H2O catalyzes diamond growth in silicate melts. Diamond

crystallization is markedly inhibited in H2O-poor silicate melts, especially below H2O/(H2O +

SiO2) mass ratios of 0.16 (Sokol and Pal'yanov, 2008). An important finding from experiments in hydrous carbonatitic–silicic media is the incorporation of multi-phase inclusions in the synthesized diamonds that resemble those in natural fibrous diamonds (Bureau et al., 2012). The inclusions confirm that silicic melts and aqueous fluids can coexist in regions of active diamond growth and may be miscible as a supercritical fluid (Bureau et al., 2012). In furthering these efforts to understand natural diamond growth, carbonate reduction to diamond must also be considered, as opposed to conversion of graphite to diamond. Thereby, the

carbonate or CO2 in the growth medium itself is the source of C for diamond growth, rather than added graphite. In this case, C reduction is a prerequisite for diamond crystallization. Factors that influence the redox reaction may, in turn, influence the nature of diamond growth. Synthetic diamond growth has been achieved by the reduction of carbonate to neutral carbon (C0) using a

variety of reducing agents, including H2 (Pal'yanov et al., 2002b; Palyanov et al., 2005), metallic Si or SiC (Arima et al., 2002), and metallic Fe (Palyanov et al., 2013). A redox gradient was established in the latter experiment by surrounding a pellet of metallic Fe or Fe-carbide with carbonate. In addition to diamond crystallization, investigation into the incorporation of N into synthetic diamonds has been useful in showing how N might partition between different phases in nature. N incorporation into diamond can vary not only according to the N concentration of the growth medium (Yu et al., 2008), but also according to growth rate (Babich et al., 2012) and 17

partitioning effects with the growth medium (Palyanov et al., 2013). Aggregation of substitutional N in the diamond crystal lattice is another feature of natural diamond that can be reproduced experimentally. Aggregation kinetics provide constraints on the temperature and duration of mantle storage for diamonds (Taylor et al., 1990; Taylor et al., 1996). Most of the experimentation that has considered N incorporation in diamond has been primarily for industrial applications, and therefore has been done in metal catalysts because this is the primary method of industrial diamond synthesis (Field, 1992b). Monitoring and controlling uptake of N during diamond synthesis (Kanda and Sekine, 2001) has built a framework for understanding N partitioning between metal and diamond. This framework may apply to diamonds from mantle settings saturated with metallic iron (Frost et al., 2004; Rohrbach et al., 2007; Rohrbach et al., 2011).

1.3 Outstanding gaps in our knowledge Fluid inclusions in diamonds are invaluable samples, but they are inherently limited by two factors. Firstly, studies of diamond fluid inclusions are almost entirely restricted to fibrous diamond, as it is the only diamond variety in which fluid inclusions have been readily found. Secondly, these fibrous diamonds are generally young. Their low N aggregation states place them at a similar age to their volcanic hosts. As cratonic diamonds in general are hosted primarily by Phanerozoic (<541 Ma) volcanics (Gurney et al., 2010), most fibrous diamonds are Phanerozoic by association. Such is the case for fibrous diamonds reported in previous studies. These two limiting factors mean that the previous studies into diamond-forming fluids are restricted to Phanerozoic fluids and restricted to fibrous diamond. As shown in Table 1-1, the more common and commercially viable variety of diamond, octahedrally-grown diamond, is typically Archean or Proterozoic in age, and may form in a distinct manner from fibrous diamond. All fibrous diamonds analysed to date have fluid inclusions of generally carbonatitic character, indicating that diamond growth may have occurred by the reduction of carbonate. The temporal restriction of analysed fluid to the Phanerozoic is problematic because there is convincing evidence of Archean diamond growth by the oxidation of methane, based on C and N isotopic variations in coexisting octahedrally-grown diamonds in one xenolith (Thomassot et al., 2007). Stachel and Harris (2009) have suggested that there has been a temporal shift in the

fluid/melt species, and therefore redox reactions, that form diamonds, from methane oxidation in the Archean, to carbonate reduction in the Proterozoic onward. This dichotomy underscores the need to analyse Archean diamond melt/fluid inclusions. A third deficiency of previous fibrous diamond studies is that the saline–carbonatitic variety of fluid inclusions have not been analysed for radiogenic isotopes. Klein-BenDavid et al. (2010) found evidence for fluid mixing within Sr isotopes in silicic–carbonatitic. These authors proposed that the silicic end member required input from an ancient, lithospheric mica-rich metasome source that was mixed with carbonatitic melt from the convecting mantle. This raises the question of whether or not the saline character in some fibrous diamond fluids might similarly have its own distinct radiogenic signature. The results of EPMA and TEM studies of fibrous diamond, while valuable, are limited by the restriction of only analysing one inclusion at a time. FTIR analysis remedies this problem, but only to a certain extent, as it is limited by its inability to detect the chlorides that dominate the daughter mineral population in the saline variety of fluid compositions. There is need for a better methodology for bulk characterization of fibrous diamonds. Preliminary work using X-ray diffraction has shown it may be a valuable complementary bulk analysis technique. It would be able to detect chloride minerals. Furthermore, the task of mineral identification would benefit from the extensive ICDD diffraction database. The need to analyse Archean diamond fluid inclusions and to further characterize the saline–carbonatitic fluids, however, still falls within the realm of fibrous diamond. Our view of diamond-forming media remains grossly incomplete in the absence of fluid or melt inclusions from octahedral diamond. Expanding upon our knowledge from fibrous to octahedrally-grown diamond requires recognizing more fluid and melt inclusions in the latter. In particular, it remains to be demonstrated whether the fluids or melts equilibrated with octahedrally-grown diamond contain oxidized or reduced carbon species, whether the fluids are similar to fibrous diamond fluids, and what other volatiles are involved. The fluids in Siberian octahedral diamonds described by Tomilenko et al. (1997) provide a good starting point, but these have been inadequately characterized. More detailed Raman spectroscopy is required to clarify the peculiar N-rich composition suggested by Tomilenko et al. (1997), and to check for methane or other hydrocarbons in the fluid. 19

1.4 Organization of thesis

1.4.1 Manuscript chapters This PhD project tackles several broad problems surrounding fluid inclusion studies in diamonds. These include shortcomings in techniques for studying fluid/melt inclusions in diamond, limitations in previously studied fluid/melt inclusions to Phanerozoic, fibrous diamond varieties, uncertainties in the composition of volatiles trapped by octahedrally-grown diamond, and uncertainties in the behaviour of volatiles, especially N and water, during diamond formation. These are addressed in the following chapters: • Chapter 2 addresses the issue of analytical techniques. Specifically, it explores the use of transmission X-ray diffraction for the non-destructive determination of the bulk mineralogy of crystallized fluid inclusions in fibrous diamond. It was applied to several fibrous diamond suites.

o This chapter is published as: Smith, E.M., Kopylova, M.G., Dubrovinsky, L., Navon, O., Ryder, J., and Tomlinson, E.L., 2011. Transmission X-ray diffraction as a new tool for diamond fluid inclusion studies. Mineralogical Magazine, 75(5): 2657-2675.

o Results described in this chapter were also presented in: (Smith et al., 2010a; Smith et al., 2010c; Smith et al., 2010b; Smith et al., 2010d). • Chapter 3 describes major and trace element compositions, as well as Sr isotopes, of Archean fibrous diamond fluids to compare with those in the literature, which are restricted to the Phanerozoic, in order to examine potential temporal variations in diamond-forming media through Earth history. Analysis of Sr isotopes in these samples is also an important benchmark for comparison. The analysed fluid inclusions are saline–carbonatitic, a composition whose Sr isotopes have never been measured.

o This chapter is published as: Smith, E.M., Kopylova, M.G., Nowell, G.M., Pearson, D.G. and Ryder, J., 2012. Archean mantle fluids preserved in fibrous diamonds from Wawa, Superior craton. Geology, 40(12): 1071-1074.

o Results described in this chapter were also presented in: (Smith et al., 2011b; Smith et al., 2012b; Smith and Kopylova, 2013).

• Chapter 4 reports fluid and volatile-rich melt inclusions in octahedrally-grown diamonds, focussing on the strikingly nitrogen-rich composition of the volatiles.

o This chapter has been published as: Smith, E.M., Kopylova, M.G., Frezzotti, M.L. and Afanasiev, V.P., 2014. N-rich fluid inclusions in octahedrally-grown diamond. Earth and Planetary Science Letters, 393(0): 39-48.

o Results described in this chapter were also presented in: (Smith et al., 2013a; Smith et al., 2013c; Smith et al., 2013b). • Chapter 5 builds on the behaviour of mantle nitrogen and diamond genesis established in Chapter 4, but explores the importance of host rocks by considering recent proposals that the deeper mantle is saturated in metallic iron. The presence of metallic iron is argued to have a profound effect on the nitrogen content of diamond.

o This chapter has been accepted (March 2014) for publication in the Canadian Journal of Earth Sciences as: Smith, E.M., and Kopylova, M.G., Implications of metallic iron for diamonds and nitrogen in the sublithospheric mantle. • Chapter 6 summarizes all new results and presents a model of that explains the cause for the growth habit of fibrous diamond. Attention is drawn to the volatiles in diamond fluid

inclusions, particularly water, nitrogen, and CO2, and their behaviour during diamond formation are discussed. Suggestions are also given for future studies in this field.

1.4.2 Appendix overview • Appendix A: Diamond sample catalogue A summary of the diamonds examined in this thesis and the analytical techniques applied to them. • Appendix B: X-ray diffraction patterns Images of the two-dimensional X-ray diffraction patterns in fibrous diamonds, to supplement Chapter 2. These were collected using the high-brilliance lab diffractometer at Bayerisches Geoinstitut (BGI). • Appendix C: Detailed methods for Chapter 3 The electron microprobe, infrared spectroscopy, and mass spectrometry techniques used for Chapter 3 are described. 21

• Appendix D: Wawa electron microprobe data Major element compositions of all analysed fluid microinclusions in the Wawa fibrous diamonds are given, to supplement Chapter 3. • Appendix E: Trace element and Sr isotope analyses The trace element and Sr isotope data tables, to supplement Chapter 3 are given. These measurements were obtained by mass spectrometry at Durham University. • Appendix F: Infrared spectra Diamond infrared spectra and nitrogen characteristics are shown for Wawa fibrous diamonds (Chapter 3) and two octahedrally grown-diamonds (Chapter 4). • Appendix G: Microthermometry A description of the microthermometry for Siberian alluvial diamond samples, to supplement Chapter 4. • Appendix H: Eclogitic inclusions in Siberian diamonds Images of mineral inclusions in some Siberian alluvial diamonds, along with their energy- dispersive X-ray or Raman spectra, to supplement Chapter 4. • Appendix I: Raman quantification of nitrogen

Table of CO2 and N2 peak areas from Raman spectroscopy and their calculated molar proportions. • Appendix J: Nitrogen in sublithospheric diamond Literature data compilation for nitrogen characteristics of sublithospheric diamonds, used in Chapter 5. • Appendix K: Diamond polishing equipment A description of the equipment used to prepare diamond samples for analyses in Chapters 3 and 4. • Appendix L: Dissemination of PhD research A summary of the technical output during the course of this PhD degree program.

2 Transmission X-ray diffraction as a new tool for diamond fluid inclusion studies1

2.1 Introduction The millions of sub-micrometre-size mineral and fluid inclusions that may be trapped in fibrous diamond represent direct samples of the natural diamond-forming environment and are crucial for understanding diamond genesis. The compositions of fluid inclusions in fibrous diamond worldwide define two trends: (1) a range from silicic (Si-, Al- and water-rich) to low- Mg carbonatitic (Ca- and carbonate-rich) and (2) a range from saline (K-, Cl- and water-rich) to high-Mg carbonatitic (Ca-, Mg- and carbonate-rich) (Izraeli et al., 2001; Klein-BenDavid et al., 2009; Weiss et al., 2009b). The fluid inclusions contain daughter minerals including mica, carbonates, chlorides, apatite, and quartz (Navon et al., 1988; Guthrie et al., 1991; Walmsley and Lang, 1992; Klein-BenDavid et al., 2006; Kopylova et al., 2010). Electron microprobe (EPMA), transmission electron microscopy (TEM), and infrared spectroscopy (FTIR) methods have been used effectively for the study of fluid inclusions. Analyses by EPMA and TEM reveal inclusion chemistry, and TEM electron diffraction can identify crystal structures, but these methods only analyse one inclusion at a time out of millions. Infrared spectroscopy gives a bulk analysis of vibrational groups, but unique mineral identification is difficult, partly due to the limited data for complex carbonates and silicates at high pressure (Navon, 1991). A novel application of transmission-geometry X-ray diffraction (XRD) was used to identify daughter minerals in fibrous diamond fluid inclusions. The abundance and variable orientation of daughter minerals in fibrous diamond make them analogous to a powder spread out in a volume. In this way, the approach is similar to powder XRD in transmission-geometry

1 This chapter has been published. Smith, E.M., Kopylova, M.G., Dubrovinsky, L., Navon, O., Ryder, J., and Tomlinson, E.L., 2011. Transmission X-ray diffraction as a new tool for diamond fluid inclusion studies. Mineralogical Magazine, 75(5): 2657-2675. Additional, unpublished data has been provided in appendices. 23

(e.g. Cullity and Stock, 2001). It is a quick and non-destructive tool that gives an in-situ bulk analysis of inclusion mineralogy. The aim of this study was to develop the methodology of XRD for fibrous diamond fluid inclusions. It is shown that the detection limit for this technique is reasonably low, but the patterns from some minerals are missing, such that XRD does not give a reliable representation of the daughter mineral assemblage. The need for high-brilliance X-rays also limits the application of this technique. Nevertheless, XRD can accurately detect some daughter minerals, making it a potential complement to EPMA or FTIR methods.

2.2 Samples Diamond samples from 4 localities were studied, comprising 19 diamonds from Mbuji- Mayi, Democratic Republic of Congo; 10 diamonds from Wawa, Ontario, Canada; 8 diamonds from the Panda kimberlite, Ekati Mine, Northwest Territories, Canada; and 1 diamond from the Jericho kimberlite, Nunavut, Canada (Figure 2-1). The Democratic Republic of Congo diamonds (prefix MMZ) include 3–8 mm fibrous cuboids and coated octahedra with grey, greenish-grey, brown-grey, and yellow-grey colours. In some cases the fibrous regions are concentrically zoned with multiple growth layers punctuated by turbidity variations. All of the samples had been laser cut to extract ~0.5 mm plate sections from the centre. Cut surfaces were polished. Previous EPMA, FTIR, and Raman spectroscopy on these samples showed they have silicic–low-Mg carbonatitic fluid inclusions containing sheet silicates, carbonates, and apatite (Kopylova et al., 2010). The Wawa samples (prefix W) include 1–2 mm grey and black fibrous cuboids, fibrous dodecahedra, fibrous diamond coat and one non-fibrous diamond with a high density of inclusions. The non-fibrous sample (W8) has a granular texture and irregular, rough surfaces. The Wawa diamonds were recovered from a polymictic metaconglomerate unit of the Michipicoten Greenstone Belt, which is located in the southwest part of the Superior craton. The diamonds from the Panda kimberlite (prefix Pan) at the Ekati mine include coated octahedra with a grey coat colour, 1–4 mm on edge. Sample fragments had been double-polished to produce ~0.5 mm thick plates, and they had been investigated by EPMA and FTIR (Tomlinson et al., 2006) and by secondary ion mass spectrometry (SIMS) and laser ablation

inductively coupled plasma mass spectrometry LA-ICP-MS (Tomlinson et al., 2009) already. The fluid inclusions are saline-rich and coexist with peridotitic (7 samples) and eclogitic (1 sample) mineral inclusions (Tomlinson et al., 2006). The single sample from the Jericho kimberlite (J300339G) is a grey fibrous cuboid that is 1 mm on edge.

MMZ-14 MMZ-85

1mm 1mm

W1 W2

1mm 1mm

Pan5 J300339G

1mm 1mm

Figure 2-1. Representative sample images for all suites studied. MMZ-14, W2, and J300339G are fibrous cuboids. MMZ-85 is polished slab of a thickly-coated octahedron, shown in transmitted light. W1 is a fragment of coated octahedron, broken roughly along (110). Pan5 is a polished fragment of a coated octahedron. The fibrous coat (left) has circular ablation pits from previous work.

2.3 Methods Transmission-geometry XRD was used to investigate the inclusions. An X-ray beam passes through the sample and the diffraction pattern is recorded by an area detector. An extensive XRD investigation of the samples was carried at the Bayerisches Geoinstitut (BGI), Germany, followed by synchrotron XRD at the Advanced Photon Source (APS) at Argonne National Laboratory in the USA. Following XRD investigation, some of the Wawa samples were examined with a scanning electron microscope (SEM) coupled with energy-dispersive X-ray spectrometry (EDS) to aid with the interpretation of the XRD data.

2.3.1 Sample cleaning The XRD technique used here is a bulk analysis tool that is very sensitive to any surface contaminants. To remove surface contaminants, the samples were soaked in concentrated HF and

HNO3 (in a 3:1 mixture) at 50 °C for 24 h prior to analysis at the BGI. However, diffraction patterns obtained at the BGI showed that some samples retained traces of minerals and other material in cracks and on rough surfaces. Therefore, all samples were cleaned again prior to the synchrotron XRD studies at the APS. In this cleaning process, the samples were sealed in Teflon vials containing concentrated HF and HNO3 (in a 3:1 mixture) and heated to 140 ºC for 5 h. Subsequent XRD results were free from most suspected contaminant phases.

2.3.2 X-ray diffraction with a high-brilliance lab diffractometer The XRD analysis carried out at the BGI, used a transmission-geometry diffractometer, with a Rigaku FR-D high-brilliance rotating anode X-ray source running at 56 kV and 60 mA, with Osmic Confocal Max-Flux optics and a Smart Apex 4K CCD area detector (Dubrovinsky et al., 2006). The beam diameter was 40 µm and Mo-Kα radiation was used. The X-ray flux is ~100 times greater than a conventional sealed X-ray tube, which is important for detecting small amounts of material. Samples were fixed to a glass fibre using lacquer and mounted on a motorized goniometer stage. A video camera that was aligned relative to the X-ray beam allowed visual selection of the regions to be analysed.

The XRD patterns were collected both in a non-scanning, stationary mode and in a scanning mode where the sample rotated 360° about an axis (ϕ), which intersects the beam path at a high angle. Although the rotation helped to analyse a greater sample volume, the resulting diffraction patterns were dominated by diamond reflections. The intensity of the signal from non- diamond phases was sensitive to changes in sample orientation in many cases. Therefore, a considerable amount of time was spent carefully varying the orientation of each sample. The best patterns were obtained by first using rapid 10 or 20 s collections at varying orientations about 2 rotational axes (ϕ and ω) to find an orientation that produced strong non-diamond reflections. A pattern was then collected for 900 or 1800 s in this orientation. The resulting data is a two- dimensional image of X-ray intensity with 2θ angles increasing radially from the beam centre, or the shadow cast by the beamstop. Interpretation of the pattern was made by comparison to the PDF-2 database of the International Centre for Diffraction Data (ICDD).

2.3.3 Synchrotron X-ray diffraction The synchrotron XRD patterns were collected using an X-ray wavelength of 0.588 Å (21.1 keV) and a 20 µm beam diameter, with a mar345 image plate detector. The beam flux was several thousand times greater than at the BGI. Samples were mounted with double-sided tape on a motorized xyz stage. A video camera was used to position the samples in the beam path. Each analysis was a stationary non-scanning measurement with a 600 s collection time. Compared to the analyses carried out at the BGI, much less time was available at the synchrotron to investigate each sample thoroughly and only two or three points were analysed. As a result, some sporadically distributed mineral inclusions may have been missed, but the analysis of fluid inclusions should not be affected as they have a much more uniform distribution within the turbid regions that were analysed.

2.3.4 Detection limits It is difficult to define a detection limit in XRD because many variable factors affect the detectability of any phase. Nevertheless, the concept of a detection limit is critical in this XRD study because the inclusions in fibrous diamond account for <1% of the sample volume. The high-brilliance lab diffractometer at the BGI is capable of recording a clear XRD pattern from as

little as 10 µm3 of material in a sample. This was demonstrated using an experimental setup designed to imitate microinclusions in fibrous diamond which also accounted for particle size and attenuation. Figure 2-2 shows diffraction patterns collected from 10 µm3 and 100 µm3 of corundum powder with a particle size of 7–10 nm in a non-scanning, stationary analysis with 3600 s collection time. This particle size is at the lower end of the size range for daughter minerals in fibrous diamond, which are normally 10–200 nm (Guthrie et al., 1991; Klein- BenDavid et al., 2006; Logvinova et al., 2008). Larger crystal sizes tend to produce sharper diffraction patterns. The corundum sample was placed between two diamond anvils, each 2 mm thick, in a diamond anvil cell to simulate the attenuation which affects inclusions in diamond. The attenuation by neighbouring inclusions is negligible as these inclusions are extremely small and thin. The incident X-rays are attenuated by 36% before reaching the corundum for the Mo-Kα radiation used. The diffracted X-rays are attenuated by a further 36% as they pass through the second diamond anvil on their way to the detector. Nearly all the diamond samples analysed are thinner than 4 mm, so 10 µm3 corundum in the analysed volume is considered to be a conservative benchmark for the detection limit of a high-brilliance lab diffractometer, like the one at BGI.

α-Al2O3 (corundum) 10 nm particle size

10 μm3

100 μm3

Figure 2-2. Diffraction patterns for 10 µm3 and 100 µm3 of corundum powder using the high-brilliance lab XRD at the BGI to demonstrate its low detection limit. The sample is within a diamond anvil cell, to simulate attenuation affecting daughter mineral crystals within fibrous diamond.

The APS synchrotron setup is expected to be able to collect patterns from even smaller amounts of material because the X-ray brilliance is several orders of magnitude greater and the wavelength is slightly shorter, resulting in less attenuation. For comparison, the size of a typical microsample used for transmission XRD work is >106 µm3 (Denaix et al., 1999). In normal powder XRD, 10 µm3 of corundum would be utterly lost even in a small powder sample of 1 mm3, falling well below a typical detection limit of 0.5% (Chung, 1974).

2.3.5 Phase identification The image recorded by the area detector during each measurement constitutes a two- dimensional (2D) XRD pattern. Patterns were integrated using Fit2D v. 12.077 to give 2θ profiles (Hammersley et al., 1996). Areas of the pattern where the diffracted intensity might include undesirable reflections, such as those produced by diamond, can be selected and excluded during integration. Background removal was performed manually using Fityk version 0.9.0 before importing the profiles into Match! version 1.9 for phase identification using the ICDD PDF-2 database. Weaker peaks that are discernible in the 2D patterns can be lost in the background noise during integration. Careful comparison of peak locations and intensities between the 2D patterns and the integrated profiles helped to discriminate between noise and real peaks. The assignment of each of the phases listed in Table 2-1 as either a contaminant, an inclusion, or a secondary phase is based on the 2D pattern texture, reaction to acid cleaning and agreement with expected mineralogy. The identified phases can occur as solid mineral inclusions, daughter minerals in fluid inclusions, or secondary phases in cracks or on the diamond surface. Continuous diffraction rings are produced by large numbers of crystals in different random orientations. A ring with no distinguishable diffraction spots indicates very small, randomly oriented crystallites. A small number of discrete diffraction spots indicate the presence of a small number of larger crystals. Spotted rings are produced by phases with a few large and many small crystals. Preferred orientation produces more intense diffraction in certain regions around the circumference of the diffraction rings. Examples are shown in Figure 2-3.

preferential spots rings spoted rings orientation Figure 2-3. Examples of diffraction pattern textures that can be useful for interpreting the orientations and size distribution of crystallites.

Phases that were apparently removed or that reacted to produce fluorides during acid cleaning were generally interpreted as contaminants on the external surfaces of the crystals. Clinochlore was observed by SEM along cracks that reached the crystal surfaces and filling cavities inside some Wawa diamonds. Clinochlore reacted with HF and is labelled as a secondary phase because it was probably not incorporated during the growth of the diamond. Goethite crystals 5–10 µm long were also observed by SEM in fractures in some diamonds of the MMZ suite. Any samples with extraneous visible material that resisted acid cleaning were treated with extra caution. In several cases, aiming the X-ray beam at foreign materials on rough diamond surfaces revealed clay minerals, quartz and lizardite. The interpretation of the integrated XRD profiles was assisted greatly by noting the similarities and differences in the texture of the reflections in the 2D patterns. For example, a 2D pattern containing sharp diffraction spots along with diffraction rings at different 2θ angles must be produced by at least 2 phases. Existing chemical and mineralogical data for the MMZ (Kopylova et al., 2010) and Panda (Tomlinson et al., 2006) suites was used as a guide when interpreting the diffraction patterns. The quality of fit between measured patterns and phases in the ICDD database is expressed qualitatively. The Match! software, which was used for phase identification, produces ‘figure of merit’ values for the fits, but these vary greatly with user settings. Placing too much reliance on the ‘figure of merit’ values can lead to false identifications. The fit quality is indicated in Table 2-1 as being excellent, good, or fair , denoting matches ranging from nearly perfect to reasonable, with corresponding ‘figure of merit’ values of >0.8, 0.8–0.7, and 0.7–0.5 produced 30

by the Match! software. If no satisfactory match could be made the phases are noted as unidentified. Phases corresponding to typical daughter minerals such as micas, carbonates, halides and quartz, or typical non-fibrous diamond inclusions such as olivine and garnet, were considered likely to be genuine inclusions rather than secondary materials on the outer surfaces of the diamond or in internal cracks. This rationale is discussed further in the results section. Phases that were identified by XRD and also found as monomineralic inclusions using SEM are noted in the results (Table 2-1).

2.4 Results Of the 38 diamonds examined by XRD, 29 diamonds produced some non-diamond reflections whereas nine produced none (Appendix B: X-ray diffraction patterns). Identification of the non-diamond phases showed that only twenty of the diamonds produced patterns that are considered to be from inclusions. Of those twenty diamonds only ten produced patterns from phases that are interpreted as daughter minerals in fluid inclusions (Table 2-1). The daughter minerals are expected to be consistent with the fluid compositions reported for fibrous diamonds and their 2D XRD patterns are expected to indicate that they are made up of many small crystals, with or without preferred orientation. The XRD patterns from several of the diamonds are produced by contaminant phases or secondary minerals on the surfaces or in cracks.

Table 2-1. Summary of XRD results from high-brilliance lab (L) and synchrotron (S) diffractometers. ICDD PDF-2 XRD Pattern fit card Sample used Identified phase Formula Texture (Figure of Merit) Occurrence number

a MMZ-8 L + S quartz SiO2 spotted rings excellent (>0.8) contaminant 33-1161 MMZ-9 L none

MMZ-10 L none

clay (possibly (29- a MMZ-11 L + S sepiolite) (Mg4Si6O15(OH)2*6H2O) rings fair (0.5-0.7) contaminant 1492)

spotted rings, celadonite K(Mg,Fe,Al)2(Si,Al)4O10(OH)2 preferential 17-0521 b MMZ-14 L + S or phlogopite KMg3(Si3Al)O10(OH)2 orientation good (0.7-0.8) inclusions 10-0495

unidentified phase(s) MMZ-15 L + S (possibly mica) spots inclusions goethite FeO(OH) rings good (0.7-0.8) secondary 29 -0713

iron fluoride hydrate FeF3*3H2O rings good (0.7-0.8) contaminant 32-0464

a MMZ-16 L + S corundum Al2O3 spots fair (0.5-0.7) contaminant 43-1484 31

ICDD PDF-2 XRD Pattern fit card Sample used Identified phase Formula Texture (Figure of Merit) Occurrence number MMZ-19 L none

MMZ-22 L none MMZ-25 L + S goethite FeO(OH) rings excellent (>0.8) secondary 29 -0713 MMZ-27 L + S goethite FeO(OH) rings excellent (>0.8) secondary 29-0713

iron fluoride hydrate FeF3*3H2O rings good (0.7-0.8) contaminant 32-0464

MMZ-28 L + S none

unidentified phase(s) MMZ-29 L + S (possibly mica) spots inclusions

goethite FeO(OH) rings fair (0.5-0.7) secondary 29-0713

a corundum Al2O3 spotted rings excellent (>0.8) contaminant 43-1484

MMZ-31 L none

MMZ-75 L + S none MMZ-76 L none

MMZ-79 L + S none

MMZ-81 L none

a MMZ-85 L + S lizardite Mg3Si2O5(OH)4 rings fair (0.5-0.7) contaminant 18-0779 12-0204, 26-1226, clays, 12.5 Å and 12-0219, 18.5 Å component rings fair (0.5-0.7) contaminanta 6-0002

W1 L + S sylvite KCl spotted rings excellent (>0.8) inclusionsb 41-1476 halite NaCl spotted rings good (0.7-0.8) inclusionsb 5-0628

b dolomite CaMg(CO3)2 spotted rings good (0.7-0.8) inclusions 36-0426

b benstonite Ca7Ba6(CO3)13 spotted rings fair (0.5-0.7) inclusions 14-0637

b ikaite CaCO3*6H2O spotted rings fair (0.5-0.7) inclusions 37-0416

forsteritic olivine (Mg,Fe)2SiO4 spots good (0.7-0.8) inclusions 34-0189

rutile TiO2 rings excellent (>0.8) uncertain 21-1276

a clinochlore (Mg,Fe,Al)6(Si,Al)4O10(OH)8 rings fair (0.5-0.7) secondary 29-0701 spotted arc, phase with 2.63 Å preferential component orientation inclusions

c W2 L + S forsteritic olivine (Mg,Fe)2SiO4 spots good (0.7-0.8) inclusions 31-0795

pentlandite (Fe,Ni)9S8 rings fair (0.5-0.7) inclusions 8-0090 29-0701, 16-0351, clinochlore (Mg,Fe,Al)6(Si,Al)4O10(OH)8 spotted rings good (0.7-0.8) secondary 16-0362

c W3 L + S forsteritic olivine (Mg,Fe)2SiO4 spotted rings excellent (>0.8) inclusions 31-0795 29- 0701,24- W4 L + S clinochlore (Mg,Fe,Al)6(Si,Al)4O10(OH)8 spotted rings excellent (>0.8) secondary 0506

MgAlF5*1.5H2O MgAlF5*1.5H2O rings excellent (>0.8) contaminant 39-0665

ICDD PDF-2 XRD Pattern fit card Sample used Identified phase Formula Texture (Figure of Merit) Occurrence number

9-0353, Mg-chromite (Mg,Fe)(Cr,Al)2O4 11-0009 or chromite FeCr2O4 34-0140 W5 L + S or magnetite Fe3O4 spotted rings excellent (>0.8) inclusions 19-0629

c forsteritic olivine (Mg,Fe)2SiO4 spotted rings excellent (>0.8) inclusions 34-0189

c W6 L + S pyrope (Mg,Fe)3Al2(SiO4)3 spotted rings excellent (>0.8) inclusions 2-1008

c forsteritic olivine (Mg,Fe)2SiO4 spots fair (0.5-0.7) inclusions 31-0795

W7 L + S sylvite KCl spotted rings excellent (>0.8) inclusionsb 41-1476 halite NaCl spotted rings excellent (>0.8) inclusionsb 5-0628

b dolomite CaMg(CO3)2 spotted rings excellent (>0.8) inclusions 36-0426

b norsethite BaMg(CO3)2 spotted rings good (0.7-0.8) inclusions 12-0530 spotted arc, phase with 2.63 Å preferential component orientation inclusionsb

forsteritic olivine (Mg,Fe)2SiO4 spots good (0.7-0.8) inclusions 34-0189

W8 L + S clinochlore (Mg,Fe,Al)6(Si,Al)4O10(OH)8 spotted rings excellent (>0.8) secondary 24-0506

c titanite CaTiSiO5 spots fair (0.5-0.7) inclusions 25-0177

W9 L + S sylvite KCl spotted rings good (0.7-0.8) inclusionsb 41-1476 halite NaCl spotted rings good (0.7-0.8) inclusionsb 5-0628

b dolomite CaMg(CO3)2 spotted rings good (0.7-0.8) inclusions 36-0426 spotted arc, phase with 2.63 Å preferential component orientation inclusionsb

b ikaite CaCO3*6H2O spotted rings fair (0.5-0.7) inclusions 37-0416

b pyrope (Mg,Fe)3Al2(SiO4)3 spotted rings good (0.7-0.8) inclusions 2-1008

forsteritic olivine (Mg,Fe)2SiO4 spots fair (0.5-0.7) inclusions 34-0189 29-0701, W36 S clinochlore (Mg,Fe,Al)6(Si,Al)4O10(OH)8 spotted rings excellent (>0.8) secondary 24-0506

b PAN1 L + S norsethite BaMg(CO3)2 spots good (0.7-0.8) inclusions 12-0530

b dolomite CaMg(CO3)2 spots fair (0.5-0.7) inclusions 36-0426

sylvite KCl spots fair (0.5-0.7) inclusionsb 41-1476

spotted arc, phase with 2.63 Å preferential component orientation inclusionsb

forsteritic olivine (Mg,Fe)2SiO4 spots fair (0.5-0.7) inclusions 31-0795

sellaite MgF2 41-1443 a or LiMgFeF6 LiMgFeF6 rings good (0.7-0.8) contaminant 23-1193

a hectorite-15A Na0.2(Mg,Li)3Si4O10(OH)2*4H2O rings fair (0.5-0.7) contaminant 25-1385

PAN2 L + S sylvite KCl spotted rings excellent (>0.8) inclusionsb 41-1476

ICDD PDF-2 XRD Pattern fit card Sample used Identified phase Formula Texture (Figure of Merit) Occurrence number

spotted arc, phase with 2.63 Å preferential component orientation inclusionsb

b eitelite Na2Mg(CO3)2 spots fair (0.5-0.7) inclusions 24-1227

barium chloride b hydrate BaCl2*H2O spots fair (0.5-0.7) inclusions 39-1305

forsteritic olivine (Mg,Fe)2SiO4 spots fair (0.5-0.7) inclusions 34-0189

PAN3 L + S forsteritic olivine (Mg,Fe)2SiO4 spotted rings excellent (>0.8) inclusions 34-0189 42-0568, PAN4 L + S omphacite (Na,Ca)(Al,Mg)Si2O6 spotted rings excellent (>0.8) inclusions 17-0522 41-1476, PAN5 L + S sylvite KCl spotted rings good (0.7-0.8) inclusionsb 26-0921

pyrope (Mg,Fe)3Al2(SiO4)3 spotted rings excellent (>0.8) inclusions 2-1008

unidentified phase(s) (possibly olivine, PAN6 L + S garnet, carbonate) spots inclusions

PAN7 L + S sylvite KCl spotted rings fair (0.5-0.7) inclusionsb 41-1476

b dolomite CaMg(CO3)2 spotted rings fair (0.5-0.7) inclusions 36-0426

halite NaCl spotted rings fair (0.5-0.7) inclusionsb 5-0628 spotted arc, phase with 2.63 Å preferential component orientation inclusionsb

forsteritic olivine (Mg,Fe)2SiO4 spots fair (0.5-0.7) inclusions 31-0795

PAN8 L + S sylvite KCl spotted rings good (0.7-0.8) inclusionsb 41-1476 halite NaCl spotted rings good (0.7-0.8) inclusionsb 5-0628 spotted arc, phase with 2.63 Å preferential component orientation inclusionsb

forsteritic olivine (Mg,Fe)2SiO4 spots fair (0.5-0.7) inclusions 31-0795

pyrope (Mg,Fe)3Al2(SiO4)3 spots fair (0.5-0.7) inclusions 2-1008

b J300339G L + S dolomite CaMg(CO3)2 spotted rings excellent (>0.8) inclusions 36-0426

phlogopite KMg3(Si3Al)O10(OH)2 spots good (0.7-0.8) inclusions 10-0493

clinochrysotile Mg3Si2O5(OH)4 spots fair (0.5-0.7) contaminant 43-0662

a fluorite CaF2 rings excellent (>0.8) contaminant 35-0816 anot detected following vigorous acid re-cleaning binterpreted as daughter minerals within fluid inclusions cobserved as monomineralic inclusions in SEM, mostly 0.1-5 µm

Table 2-1 summarizes the phases identified by XRD. Remnants of gold sputter coating from previous work produced gold diffraction rings in some patterns from MMZ-75, MMZ-85, and PAN7. The Democratic Republic of Congo suite has one diamond (MMZ-14) with 34

celadonite (high-Si mica) or phlogopite mica (Figure 2-4). This phase is expected to be a daughter mineral based on the sample’s bulk fluid composition and infrared spectrum (Kopylova et al., 2010). The diffraction pattern texture indicates many small crystals with some preferential orientation. Oriented growth of phlogopite with respect to diamond has been observed in fluid inclusions using TEM (Logvinova et al., 2008). Two other samples (MMZ-15 and MMZ-29) produced a few faint diffraction spots that could not be uniquely identified, but which may originate from one or a small number of phlogopite-like mica inclusions. The corundum found with MMZ-16 and MMZ-29 is most likely to be a contaminant from the polishing paste that was used to remove the sputtered coating used for EPMA. The corundum pattern is similar in both samples and disappeared completely following more thorough cleaning. Corundum is rare and occurs only in eclogitic diamonds; it would be unusual to find it in such abundance in two separate samples. Visual X-ray targeting allowed the dark brown material on the surfaces of several MMZ diamonds to be identified as goethite. The goethite also extends into cracks, producing a red colour. Iron fluoride hydrate was found in association with goethite. This fluoride-bearing phase is believed to have formed during acid cleaning and attests to the presence of goethite as a surficial, non-inclusion phase. All other phases found in the MMZ suite are thought to be secondary minerals because they were detected only in certain areas of the rough, external surfaces of the diamonds and they could not be found following the more rigorous acid cleaning. The Wawa diamond suite includes three samples that gave diffraction patterns with sylvite, halite and dolomite, as well as benstonite (Ca7Ba6(CO3)13), norsethite (BaMg(CO3)2), and ikaite

(CaCO3·6H2O), a hydrous carbonate that is unstable at low pressures (Shahar et al., 2005). These phases are interpreted as daughter minerals in fluid inclusions, indicating that at least some Wawa fibrous diamonds contain saline–high-Mg carbonatitic-type fluid inclusions. Interestingly, the three diamonds with detected daughter minerals (W1, W7, and W9) are all fibrous diamond coat material. The Wawa fibrous cuboids and the one non-fibrous polycrystalline diamond did not reveal daughter mineral patterns.

a Synchrotron XRD, sample MMZ-14 b High-brilliance lab XRD, sample MMZ-14

60 celadonite mica, K(Mg,Fe,Al)2(Si,Al)4O10(OH)2 1000 celadonite mica, K(Mg,Fe,Al)2(Si,Al)4O10(OH)2 50 (0.588 Å wavelength) 800 (0.709 Å Mo wavelength) 40 600 30 400 20 10 200

0 0 0 5 10 15 20 25 0 5 10 15 20 25 30 35 2θ (degrees) 2θ (degrees) Figure 2-4. Two-dimensional diffraction patterns collected from sample MMZ-14 at (a) the synchrotron and (b) using the high-brilliance lab diffractometer. The white ‘hole’ in each image is the beamstop shadow. Corresponding integrated profiles (bottom) are given with the matched reference pattern for celadonite (ICDD 17-521). Arrows mark the position of the 10 Å (001) reflection for reference.

Eight of the ten Wawa diamonds, including fibrous coats and cuboids, contain mineral inclusions typical of non-fibrous diamond such as forsteritic olivine, pyrope garnet, spinel, and pentlandite. Figure 2-5 shows the diffraction pattern and integrated profile for forsteritic olivine in sample W3. The pattern shows that the olivine is present as numerous small crystals. Reconnaissance studies using SEM showed the olivine inclusions to be almost entirely sub- micrometre size. Sample W1 produced well defined diffraction rings which were identified as rutile, but only from near the diamond’s surface layer. The nature of the rutile is uncertain, as it was not observed at the APS synchrotron, but it is unlikely to be a contaminant. The non-fibrous diamond aggregate (sample W8) produced diffraction spots corresponding to titanite. Five of the ten Wawa diamonds produced clinochlore diffraction patterns. Clinochlore is likely to be secondary because it was found to be associated with a fluoride phase (MgAlF5·1.5H2O) 36

following acid cleaning. Chlorite has been observed along surface-reaching cracks in diamonds from another Wawa diamond deposit where it was attributed to greenschist facies metamorphism (De Stefano et al., 2006).

sample W3, olivine, (Mg,Fe) SiO 90 2 4 0.588 Å wavelength 80 70 60 50 40 30 20 10 0 5 10 15 20 25 2θ (degrees) Figure 2-5. Diffraction pattern and integrated profile showing the presence of forsteritic olivine in sample W3. Follow-up SEM investigation showed the olivine inclusions to be monomineralic, with a dominant inclusion size <1 µm, but ranging up to 4 µm. The pattern was collected at the APS synchrotron.

In the Panda kimberlite diamond suite, five of the eight samples produced diffraction patterns which indicated the presence of sylvite and four of these revealed carbonates as well.

The carbonates include dolomite, norsethite (BaMg(CO3)2), and eitelite (Na2Mg(CO3)2). Sylvite, halite and carbonates in these samples are interpreted as daughter minerals in fluid inclusions, which is consistent with the fluid composition found by EPMA in these samples (Tomlinson et al., 2006). In common with the Wawa diamonds, the Panda kimberlite samples produced diffraction spots for minerals that are found in non-fibrous diamonds, including forsteritic olivine, pyrope garnet and omphacite. The diffraction spots from sample PAN6 cannot be identified uniquely, but they may be produced by olivine, garnet, or a carbonate. These mineral inclusions were reported for these Panda samples using FTIR and EPMA (Tomlinson et al., 2006). However, several of the mineral inclusions found by Tomlinson et al. (2006) were not identified in the XRD patterns. For example, neither clinopyroxene nor orthopyroxene were found in the XRD patterns from PAN1, PAN3, or PAN5. An unidentified phase with a prominent reflection 2.63 Å was found in several Wawa and Panda samples (Figure 2-6); W1, W7, W9, PAN1, PAN2, PAN7, and PAN8 all contain this phase. The phase has a preferential orientation with respect to the host diamond. All the samples 37

containing this 2.63 Å phase also host chloride and carbonate daughter minerals in fluid inclusions, which suggests that the unidentified phase is also a daughter mineral from the saline– high-Mg carbonatitic fluids. Although a reasonable match could not be found, the unidentified phase is suspected to be a carbonate, chloride, or phosphate mineral. The single Jericho sample (J300339G) produced diffraction patterns for dolomite and phlogopite when examined by XRD. Dolomite is likely a daughter mineral in fluid inclusions. Phlogopite may also be present as a daughter mineral, but its pattern at the APS synchrotron only showed several non-continuous diffraction spots, rather than an arc or ring as would be expected for abundant daughter crystals in fluid inclusions. The Jericho sample also showed faint diffraction rings for clinochrysotile, which is suspected to be a contaminant, or possibly a secondary mineral along cracks in the diamond. Aside from the inclusions and other phases, the diffraction patterns from the diamonds themselves deserve mention. Most of the XRD pattern that were collected without rotation, in random sample orientations contained either one (111) diamond reflection or no diamond reflections at all. Patterns collected with 360° scanning sample rotation revealed a symmetrical pattern of (111) and (220) diffraction spots. These observations serve as a reminder that fibrous diamond cuboids and fibrous diamond coatings are not polycrystalline. Although fibrous diamond has less crystal lattice perfection than gem-quality diamond, the diamond “fibres” share a common crystallographic orientation and a common nucleus, like the branches of a snowflake. Ballas is the truly polycrystalline variety of fibrous diamond. As a matter of terminology, “monocrystalline diamond” should not be used to refer to non-fibrous, gem-quality diamond when trying to distinguish it from fibrous diamond.

W1 W9

Pan7 Pan8

Figure 2-6. Diffraction patterns showing the unidentified phase in samples W1, W9, Pan7 and Pan8. Arrows mark the common 2.63 Å reflection. Note the similar appearance of this reflection in all four patterns as a central smudged reflection with two smaller adjacent satellite reflections. Other phases present in these patterns are sylvite, halite, carbonates and forsteritic olivine. The patterns shown here were recorded with the high-brilliance lab diffractometer.

2.5 Discussion

2.5.1 Missing daughter mineral patterns The XRD analysis was expected to reveal the daughter mineralogy in all the fibrous diamond samples. However, less than one third of the fibrous diamonds that were examined produced diffraction patterns for daughter minerals in fluid inclusions, despite the fact that all the fibrous diamond samples are turbid and clearly inclusion rich. Daughter mineral phases that were expected on the basis of EPMA and FTIR studies include phlogopite or celadonite-like high-Si mica, Ca-Mg-Fe-Ba carbonates, apatite, quartz, and chlorides (Navon, 1991; Klein-BenDavid et al., 2009; Kopylova et al., 2010). A crucial observation from TEM studies is that fluid inclusions in fibrous diamond contain daughter crystals that are capable of producing electron diffraction patterns (Lang and Walmsley, 1983; Guthrie et al., 1991; Klein-BenDavid et al., 2006). On the basis of EPMA, FTIR, and TEM studies, the mineralogy of fluid inclusions in fibrous diamond is dominated by phlogopite or high-Si mica, Ca-Mg-Fe carbonates, apatite, K- Na chlorides, and quartz (Guthrie et al., 1991; Navon, 1991; Klein-BenDavid et al., 2006; Weiss et al., 2010). The composition is fairly consistent from one fluid inclusion to the next within single diamonds, although some zonation has been observed in EPMA (Klein-BenDavid et al., 2004; Weiss et al., 2009b; Kopylova et al., 2010). Chemical variability between fluid inclusions is also observed in TEM, but it is obscured by the loss of material from ruptured inclusions (Klein-BenDavid et al., 2006). Three hypotheses to explain the poor XRD response from daughter minerals are presented: (1) the abundance of daughter minerals is low and often falls below the detection limit in XRD; (2) a portion of the daughter mineral population is actually amorphous or dissolved in residual fluid; (3) the daughter mineral population comprises many minerals with similar chemistry, such that any one mineral is often not detectable in XRD. These three explanations are discussed below.

2.5.1.1 Model 1: Non-detectable daughter minerals In section 2.3.4 it was demonstrated that the high-brilliance lab diffractometer can detect the diffraction pattern from 10 µm3 of corundum, with 7–10 nm particle size and random orientation, within the analysed volume. Aside from total volume and particle size, the 40

detectability of any phase is dependent on crystal symmetry and average electron density. Minerals that have highly symmetrical crystal structures produce strong signals in XRD. Dense minerals with closely spaced atoms or high average atomic numbers have high overall electron density and also tend to produce strong signals in XRD. Crystal symmetry and electron density are mineral-specific factors. Accordingly, the intensity of the XRD pattern produced by 10 µm3 of corundum will be different from the intensity produced by 10 µm3 of another mineral. Empirical comparisons of XRD pattern intensity can be used to account for this effect. Many

ICDD powder diffraction files contain published reference intensity ratios. The value, I/Ic, is a

ratio of the integrated profile intensity of any mineral (I) compared to that of corundum (Ic) in a 50:50 mixture by mass. Despite the problems of different minerals having different numbers of peaks and peak positions, the intensity ratios provide a straightforward approach to the variation

in XRD pattern intensity between minerals. For our purposes, the I/Ic values provide a good approximation for judging the detection limits of other minerals based on the detectability of 10 3 µm of corundum. Table 2-2 lists I/Ic values for some minerals of relevance to the fluid inclusions. The intensity of the pattern of 10 µm3 or about 40 pg of corundum collected with the

high-brilliance lab diffractometer (Figure 2-2) should be comparable to 40/( I/Ic) pg of another mineral. For randomly oriented particles, this mass accounts for the fact that only a fraction of particles may be favourably oriented to diffract in a stationary analysis. Diffraction from particles with preferred orientation will vary somewhat with sample orientation, although sample rotation during XRD measurement did not improve the results.

Table 2-2. Reference intensity ratios of some daughter mineral phases with respect to corundum. Detectable mass, based ICDD PDF-2 card Mineral I/Ic 3 on 10 µm corundum (pg) number for I/Ic corundum 40

calcite 2 20 5-0586 ankeritea 2.8 14 41-0586 phlogopite 0.97 41 34-0159 apatite 1.5 27 15-0876 quartz 3.6 11 33-1161 sylvite 3.9 10 4-0587 a Ankerite I/Ic value should be similar to dolomite. A dolomite value is not in the ICDD database.

The calculated detectable masses for each mineral can now be compared to the mineral concentrations measured by FTIR for some of the diamond samples. Table 2-3 shows the concentrations of phlogopite and calcite using the conversion factors developed by Weiss et al. (2010) specifically for fibrous diamond. To estimate the amount of each mineral intercepted by the X-ray beam during XRD analysis, concentrations were multiplied by the mass of diamond analysed by the X-ray beam for the high-brilliance lab diffractometer. A conservative estimate of the volume analysed was calculated using the beam diameter of 40 µm and the sample thickness in its thinnest dimension. The ‘intersected mass’, in picograms, is an estimate of the amount of mineral in the volume analysed by the X-ray beam with the high-brilliance lab diffractometer, assuming a homogeneous inclusion distribution. These amounts can be compared to the values in Table 2-2. All the samples measured contain enough phlogopite to exceed the 41 pg ‘detectable mass’, which is expected to produce an XRD pattern with an intensity comparable to that of 10 µm3 of corundum. Moreover, 11 of the 17 samples contain more than ten times the detectable phlogopite mass. The XRD results revealed celadonite or phlogopite mica in MMZ-14, which has the third highest phlogopite concentration according to FTIR and the third highest calculated intersected mass of 1550 pg. The phlogopite concentration is also reasonably high in MMZ-29, which showed faint diffraction spots from an unidentified phase that may be phlogopite mica. Calcite concentrations from FTIR are also shown in Table 2-3. Some FTIR spectra did not have clear carbonate peaks and the calcite entry for those samples is blank. Eight of the 17 samples have enough calcite to meet or exceed the 20 pg mass that is expected to be detectable with the high-brilliance lab diffractometer. Two samples contain more than ten times the detectable mass of calcite. Compared to other fibrous diamonds, the concentrations of phlogopite and calcite in the 17 MMZ samples are generally low. Weiss et al. (2010) reported concentrations in the range of 186–1188 ppm for phlogopite/mica and 41–480 ppm for calcite from 13 fibrous diamonds from six different localities. This suggests that fibrous diamonds, in general, should have daughter mineral concentrations above the known detectable levels for the high-brilliance lab XRD, at least for dominant daughter minerals.

Table 2-3. Calculated concentrations and mass of each mineral intersected by the X-ray beam. Phlogopite Calcite

XRD minimum Concentration intersected Concentration intersected Sample thickness (mm) (ppm) mass (pg) (ppm) mass (pg) MMZ-8 2.1 60 557 3 28 MMZ-9 0.3 43 57

MMZ-10 1.1 92 447

MMZ-14 1.4 250 1550 27 168 MMZ-16 2.0 133 1180 64 566 MMZ-19 1.9 100 841

MMZ-22 2.1 85 791 7 61 MMZ-25 1.1 319 1554

MMZ-27 1.3 789 4536

MMZ-28 0.8 67 236 42 150 MMZ-29 1.8 152 1210 17 135 MMZ-31 1.7 119 898

MMZ-75 4.0 65 1157 15 268 MMZ-76 0.5 87 193 8 17 MMZ-79 0.4 23 42 5 9 MMZ-81 0.6 97 256 12 33 MMZ-85 0.6 89 236

The estimated phlogopite and calcite contents from FTIR suggest all 17 MMZ samples shown in Table 2-3 should yield diffraction patterns for phlogopite using a high-brilliance lab diffractometer. Several of these samples should also yield diffraction patterns for calcite. However, the actual XRD results only had one sample (MMZ-14) with a clear pattern for celadonite or phlogopite mica. The results from the APS synchrotron were similar to those from the high-brilliance lab diffractometer, despite the significantly lower expected detection limit. The intensity of an XRD pattern also depends on crystal perfection, orientation and attenuation from surrounding phases. Crystal perfection and orientation effects are less significant than symmetry and electron density (Cullity and Stock, 2001). Attenuation from diamond has been more than accounted for by placing the test 10 µm3 corundum sample in a diamond anvil cell. Attenuation from fluid and other inclusions surrounding each daughter mineral is insignificant because the beam passes through very little of these materials.

Peak broadening due to small crystallite size introduces another potential limitation to detecting daughter minerals. Peak broadening varies inversely with crystallite size, increasing rapidly below ~10 nm as the crystallite size approaches the X-ray wavelength. Daughter minerals imaged using TEM are normally 10–200 nm in size (Guthrie et al., 1991; Klein-BenDavid et al., 2006; Logvinova et al., 2008) and they should produce less broadening than the 7–10 nm particles in the corundum reference sample. The XRD results could be taken as an indication that a significant number of daughter crystals fall at the lower end, or perhaps below, the crystal size range observed in TEM. Broadening is exaggerated by the thickness of each sample when the thickness exceeds ~1% of the sample-to-detector distance because the variable distance between each crystallite and the detector begins to have a significant effect. For the high-brilliance diffractometer used, the distance is only 50 mm, meaning the broadening exhibited in Figure 2-2 will roughly double for a 2 µm thick sample. Thickness broadening is insignificant at the 570 mm sample-to-detector distance for the synchrotron XRD measurements. The expression below includes the effects of broadening due to crystallite size in the first term, which is the Scherrer equation (Scherrer, 1918) and sample thickness in the second term:

Equation 2-1: = + 2 tan 퐾휆 −1 푑 tan 2휃 where β is angular broadening,훽 �휀 cos K휃 �is a �shape휃 − factor,� λ is푑+ the푡 � X� -ray wavelength, ε is particle size, θ is the Bragg angle, d is the sample-to-detector distance, and t is the sample thickness intersected by the beam. The crystallite shape factor (K) is usually near 1 (Langford and Wilson, 1978) and has a smaller effect than crystallite size (ε). Although thickness broadening is not angular, it is expressed here as an angular broadening so it may be added directly to the Scherrer equation. Another consideration which was taken into account is the inclusion spacing compared to the X-ray beam width. It is very unlikely that the X-ray beam would pass through the sample without intersecting fluid inclusions. Images of fibrous diamond produced by TEM show a fluid inclusion spacing on the order of ~5 µm (Klein-BenDavid et al., 2006). The 40 µm diameter X- ray beam produced by the high-brilliance lab diffractometer would intersect ~104 inclusions in a 1 mm path in such a diamond. The 20 µm X-ray beam produced by the synchrotron would intersect about 103 inclusions. Given the size of the X-ray beams, the spacing of the fluid

inclusions and the sample dimensions, it is likely that many inclusions were intersected by the X- ray beam during each XRD analysis. Thus the comparatively small volume analysed by XRD should truly reflect the mineral concentrations derived from the FTIR data. The XRD detection limits are a serious limitation and it is possible that some minor daughter mineral phases such as apatite fall below the detection limits. However, the detection limits do not provide a clear-cut explanation for the absence of XRD patterns from daughter minerals for the samples examined. Even though it is difficult to accurately define a detection limit for the technique used, conservative estimates based on XRD patterns from corundum and FTIR spectra from 17 samples of the MMZ suite suggest that the amount of phlogopite in all 17 samples should be sufficient to produce diffraction patterns.

2.5.1.2 Model 2: Amorphous or dissolved material A second explanation for the unexpectedly poor detection of daughter minerals using XRD is that some of the daughter phases are amorphous solids or partly dissolved in residual fluid (Kopylova et al., 2010). There are a few limited accounts of amorphous solids in fibrous diamond. Quartz was found coexisting with amorphous silica using TEM analysis in a fluid inclusion in a fibrous diamond coat (Guthrie et al., 1991). Analyses of some turbid cuboid regions at the centre of octahedral diamonds by TEM have identified amorphous carbonate coexisting with crystalline daughter minerals in fluid inclusions (Logvinova et al., 2008). Glassy silicates have been reported on the basis of FTIR and Raman studies of fibrous cuboid diamonds from Udachnaya (Zedgenizov et al., 2004). Dissolution cavities have also been found to contain varying amounts of amorphous alumina, silica and carbonate (Klein-BenDavid et al., 2007b). Although these cavities are interpreted to be distinct from fluid microinclusions, they provide another example of amorphous solids in fibrous diamond. Studies of fibrous diamonds by TEM more commonly reveal crystalline daughter minerals inside fluid inclusions (Guthrie et al., 1991; Walmsley and Lang, 1992; Klein-BenDavid et al., 2006; Logvinova et al., 2008). These results do not support the hypothesis that amorphous solids are a major constituent of fibrous diamond fluid inclusions. The possibility that the observed crystals were produced inadvertently during sample preparation via ion beam milling is considered unlikely. Ion beam milling of the sample foils produces a temperature increase of less

than 10 ºC at the ion beam site (Ishitani and Yaguchi, 1996). Furthermore the pressure drop caused by rupturing the fluid inclusions cannot be responsible for triggering crystal growth because intact inclusions contain crystalline daughter minerals (Guthrie et al., 1991). If anything, the beam energy would be more likely to cause amorphization of existing crystals. Aside from amorphous solids, the dissolved mineral content in the residual fluid phase should be considered. The fluid inclusions typically contain 10–25% water, by mass (Weiss et al., 2010). Mineral solubility in the water will be elevated due to the high inclusion pressure and salinity, but some simple calculations show the dissolved load cannot be substantial. For example, the molality of calcite in a H2O-NaCl system at 300 K and 1 GPa is well below 0.05 mol/kg (Newton and Manning, 2002; Duan and Li, 2008). Translating this solubility to a fluid inclusion with 25% water means that the inclusion will contain <<1% dissolved calcite. Therefore, it is likely that the only a minor component of most carbonates, silicates, and other daughter minerals could be dissolved in the fluid phase at room temperature. Overall, it appears unlikely that amorphous solids and/or dissolved minerals could dominate the daughter mineral population. Some mineral peaks in infrared spectroscopy may be enhanced by the combined contribution of crystals and dissolved or amorphous material, giving the impression that mineral concentrations are higher than they really are. The poor XRD response of daughter minerals may be partly, but not entirely, explained by amorphous or dissolved material.

2.5.1.3 Model 3: Daughter mineral diversity Another explanation for the poor general detectability of daughter minerals by XRD may be that the mineral population is made up of a wide variety of minerals, rather than a few dominant ones. As a result of this variety, each mineral would be present in lower concentrations and might not be detectable by XRD. Some mineral concentrations calculated from FTIR (Table 2-3) may not represent single minerals, but may be due to overlaps in the spectral features produced by related minerals. For example, a diamond could contain many related carbonates rather than calcite or dolomite alone. Those daughter minerals that were detected by XRD may come from growth regions with less mineralogical variability. Several examples of adjacent fluid microinclusions with different

daughter minerals have been shown in TEM analyses (Klein-BenDavid et al., 2006; Logvinova et al., 2008). Mineral variability may also be exemplified in the Raman spectra for the MMZ

suite, which revealed peaks for possible apatite and anapaite (Ca2Fe(PO4)2·4H2O) among the phosphates; brucite, talc, serpentine and clinochlore among the hydrous magnesium-rich sheet

silicates; dolomite, calcite and kalicinite (KHCO3) among the carbonates; biotite and phlogopite among the micas; Mg-chromite, ilmenite, magnetite and hematite among oxides; as well as graphite, pyrope, forsterite, monticellite, orthopyroxene, quartz, halite, and bultfonteinite

(Ca2(HSiO4)F·H2O) (Kopylova et al., 2010). Another example of variability comes from two Wawa samples (W1 and W9) that appear to contain ikaite, a hydrous carbonate which is stable only at high pressure (Shahar et al., 2005) along with two anhydrous carbonates. Aside from indicating mineralogical variability, these phases may signify pressure variability amongst fluid inclusions. The range in the size and shape of fluid inclusions could elicit different elastic strains in the diamond lattice, leading to a range in inclusion pressures upon cooling from mantle conditions. It should be noted that many of the fluid inclusions have dislocations around them (Klein-BenDavid et al., 2006; Logvinova et al., 2008) that will affect the accommodation of stress from the inclusion. These observations support the idea that there could be sufficient compositional and pressure variability from inclusion to inclusion to produce a range of different daughter minerals. Daughter mineral diversity may be able to explain why many samples produced no response on a high-brilliance lab diffractometer. However, the results from the APS synchrotron are harder to explain in terms of mineral diversity. It would have to be extreme, perhaps >100 different minerals, given that the detection limit for the synchrotron diffractometer is at least an order of magnitude lower than the lab instrument.

2.5.2 Daughter mineral species Sylvite and halite are confirmed for the first time as prominent daughter minerals in fluid inclusions in fibrous diamond. Chlorides are transparent in the mid-IR range and they decompose rapidly in the TEM (Klein-BenDavid et al., 2006) which makes them difficult to identify. Previous identifications of these halide minerals were inconclusive, as they were based on fluid

inclusion chemistry (Klein-BenDavid et al., 2006; Rondeau et al., 2007; Logvinova et al., 2008; Kopylova et al., 2010) or non -unique Raman peaks (Kopylova et al., 2010). Sylvite and halite were found in the Wawa and Panda diamond suites. Fluid inclusions in the Panda suite have been shown to have saline–high-Mg carbonatitic-type compositions (Tomlinson et al., 2006). The Wawa samples with sylvite and halite were also found to have chloride-rich microinclusions using SEM. Another mineral of interest was also identified in the Wawa sample suite. Two Wawa samples produced XRD patterns that were a fair match to the high-pressure mineral ikaite

(CaCO3·6H2O). This match is strengthened by the fact that the fluid inclusions are expected to be carbonate- and water-rich and have high residual pressures (Navon, 1991). Ikaite would be destroyed by the loss of inclusion pressure during ion beam milling for TEM analysis, leaving calcite or aragonite in its place. Calcite formation is inhibited by Mg2+ and phosphate, but these ions have much less effect on ikaite formation (Lippmann, 1959; Dickens and Brown, 1970). The stability of ikaite at room temperature requires an inclusion pressure of at least 0.45 GPa (Shahar et al., 2005). The actual pressure is probably higher. Quartz peak shifts in infrared spectra indicate pressures of 1.5–2.1 GPa in carbonatitic–silicic fluid inclusions (Navon, 1991).

Navon (1991) estimated entrapment pressures of 4–7 GPa by extrapolating along H2O–CO2 isochores to upper mantle temperatures. Similar extrapolation, starting with 0.45 GPa, means the fluid inclusions in the Wawa diamonds were trapped at pressures greater than 2.5 GPa.

2.6 Summary and conclusions Transmission X-ray diffraction is demonstrated as a new tool for examining bulk daughter mineralogy within fluid inclusions in fibrous diamond as well as mineral microinclusions in diamond. In transmission geometry, the X-ray beam passes through the sample, interacting with a volume of material. Fibrous diamonds from the Democratic Republic of Congo, Wawa, the Ekati mine, and the Jericho kimberlite were analysed. The low daughter mineral concentrations within fibrous diamond require a high-brilliance X-ray source. The detection limit for such a diffractometer is <10 µm3 for corundum powder within the analysed volume. Identified daughter mineral phases include celadonite or phlogopite-like mica, dolomite,

sylvite, and halite as well as probable benstonite (Ca7Ba6(CO3)13), norsethite (BaMg(CO3)2), 48

eitelite (Na2Mg(CO3)2), ikaite (CaCO3·6H2O) and an unidentified phase with a prominent reflection at 2.63 Å. In addition to fluid inclusions, mineral inclusions of forsteritic olivine, pyrope garnet, pentlandite and other phases typical of non-fibrous diamonds were identified. Unexpectedly, only 10 out of 38 diamonds examined produced diffraction patterns with daughter minerals. Low detection limits cannot readily account for the poor response from major daughter mineral phases. The presence of significant amounts of amorphous or dissolved material appears unlikely, but cannot be ruled out. Presently, the preferred explanation is that there is a wide variety of daughter minerals, thereby lowering the concentration of any one mineral. It is plausible that some samples contain sufficient daughter mineral diversity such that no single phase exceeds the detection limits for the XRD techniques used. Overall, transmission X-ray diffraction is capable of identifying common daughter minerals, but tends to give an incomplete account of fluid inclusion mineralogy. It may therefore be more valuable when accompanied by other techniques like EPMA, TEM, or FTIR.

3 Archean mantle fluids preserved in fibrous diamonds from Wawa, Superior Craton2

3.1 Introduction Fluids inclusions in diamond are the best preserved samples of any mantle fluid due to the robustness of their host. They provide information on processes of diamond growth and mantle metasomatism in general. Studies of fluids in diamond revolve around “fibrous” diamond, a habit formed by symmetric <111> dendritic growth. The habit leads to a cuboid shape. Fibrous growth on an older, preexisting octahedral diamond produces a fibrous coat. Fibrous diamond is usually turbid due to a high density of submicron-sized fluid and mineral inclusions. Low nitrogen aggregation states indicate short mantle residence times for most fibrous diamonds, probably less than 5 Ma prior to volcanic transport to the surface (Gurney et al., 2010). This observation, in addition to Sr isotopes (Akagi and Masuda, 1988) and trace element characteristics, has led several authors to propose a genetic link between fibrous diamond and kimberlitic fluids (e.g. Navon et al., 1988; Tomlinson et al., 2005; Weiss et al., 2011). Given that the majority of known kimberlites are Phanerozoic, most fibrous diamonds are also relatively young compared to octahedral diamonds, which are mostly Archean and Proterozoic in age (Gurney et al., 2010). This means that diamond-forming fluids studied to date are essentially limited to the Phanerozoic record. The present study focuses on Archean fibrous diamonds to explore more ancient diamond- forming fluids. These diamonds were recovered from a 2.697–2.701 Ga diamondiferous metaconglomerate in the Wawa subprovince of the Superior craton, Canada. Diamonds in the metaconglomerate were originally brought to surface by a kimberlite, emplaced shortly before 2.701 Ga (Kopylova et al., 2011). Consequently, fluid inclusions in Wawa fibrous diamonds may be the oldest mantle fluids available for study.

2 This chapter has been published. Smith, E.M., Kopylova, M.G., Nowell, G.M., Pearson, D.G. and Ryder, J., 2012. Archean mantle fluids preserved in fibrous diamonds from Wawa, Superior craton. Geology, 40(12): 1071-1074. Additional, unpublished data has been provided in appendices. 50

3.2 Wawa metaconglomerate diamonds The studied diamonds include three fibrous coats over octahedral cores (W1, W7, W9) and three fibrous cuboids (W14, W29, W50) (Figure 3-1). These were analysed using infrared spectroscopy and electron probe microanalysis (EPMA), in addition to mass spectroscopy for trace element and Sr isotope characteristics (Appendix C: Detailed methods for Chapter 3). Consideration was given to select unfractured diamonds, impervious to fluids from crustal metamorphism. Greenschist metamorphism of the conglomerate host rock has introduced clinochlore along surface-reaching cracks in some diamonds (Smith et al., 2011a).

W1 Fibrous coat Si + Al Fluid composition in fibrous coats W1 S ili c W7 ic – lo W9 1 mm w -Mg

carbona

W14 t iti Cuboid c Saline–high-Mg caarbonatitic

K + Na

Ca + Mg + Fe Figure 3-1. Molar composition of fluid inclusions in Wawa (Canada) fibrous diamonds. Inset shows a fibrous coated (W1) and a fibrous cuboid (W14) diamond. Fields are given for fibrous diamonds worldwide (Zedgenizov et al., 2007a; Klein-BenDavid et al., 2009, and references therein; Kopylova et al., 2010).

Bulk fluid compositions, determined by EPMA of unexposed inclusions lying just below a polished and cleaned surface, were consistent between the three fibrous coat samples (Appendix D: Wawa electron microprobe data; Appendix K: Diamond polishing equipment). No systematic intra-diamond variation was found. Fluid inclusions in the cuboids were prohibitively sparse compared to mineral inclusions and were not analysed by EPMA. Measured fluid inclusions fall within the established compositional range of Phanerozoic fibrous diamonds, plotting toward the saline end member of the saline–high-Mg carbonatitic trend (Figure 3-1). The average composition of the 174 analyses is: 38% Cl, 26% K, 18% Na, 3.9% Ca, 4.2% Mg, 3.4% Fe, 4.6% 51

Ba, 1.6% Si, and 0.1% Al. Previous X-ray diffraction analyses revealed sylvite, halite, and Ca- Mg-Fe-Ba carbonate daughter minerals that crystallized within the trapped fluid (Smith et al.,

2011a). The fluid inclusions coexist with discrete olivine (Fo92) and Cr-pyrope mineral microinclusions, indicating a harzburgitic paragenesis. This paragenesis also characterizes octahedral, non-fibrous diamonds in the Wawa metaconglomerate (Miller et al., 2012), including the cores of coated diamonds, which are significantly older than the fibrous growth. Infrared spectroscopy reveals nitrogen contents of 1000–2000 ppm in the fibrous diamond coats and 230–330 ppm in the fibrous cuboids, present as paired nitrogen atoms (Appendix F: Infrared spectra). Aggregated groups of four nitrogens were not observed, and the diamonds can be classified as Type IaA. The same poorly aggregated nitrogen characteristics prevail in Phanerozoic fibrous diamonds (Boyd et al., 1992) and are thought to signify a short mantle residence time. For example, the most nitrogen-rich diamond (1994 ppm) spent <1 Ma in the mantle, based on aggregation kinetics at 1150 °C (Taylor et al., 1990). Infrared spectra also show the presence of water (Figure 3-2) along with carbonate peaks from daughter minerals. An average water concentration (bulk diamond) of 80 ppm was calculated from the peak height at 3400 cm-1 (Weiss et al., 2010) for the fibrous coats. These results are consistent with Phanerozoic fibrous diamonds, which typically contain 50–500 ppm water (Weiss et al., 2010)

and only minor amounts of CO2, as most of the carbon is present as carbonate (Navon et al., 1988). The lower overall inclusion content of the cuboids gives an average of ~10 ppm water and correspondingly diminished carbonate peaks. The cuboids crystallized from a different fluid, as evidenced by contrasting nitrogen concentrations. This conclusion is supported by trace element characteristics (Figure 3-3) (Appendix E: Trace element and Sr isotope analyses). Here, pattern shapes are more meaningful than absolute concentrations of analysed elements, as the latter are mostly controlled by variable inclusion contents within diamond. Wawa fibrous diamond coats show high Th/U and Rb/Sr ratios, higher levels of Ba and Th, pronounced positive Eu anomalies, and stronger light rare earth element

(LREE) enrichment (Lan/Prn = 8–16; subscript n represents normalization by C1-chondrite)

compared to cuboid samples (Lan/Prn = 2). These differences between coats and cuboids relate to distinct compositions of fluid batches, with cuboid diamond fluid being less N-rich and possibly less hydrous. It precipitated diamond within a diamond-barren host that had not been exposed to 52

the previous metasomatic activity of octahedral diamond growth. However, the cuboids trapped fewer fluid inclusions and more mineral inclusions than the coats, making direct geochemical comparison difficult. Five of the six studied coated and cuboid diamonds have clear Nb troughs, which may reflect fluid interaction with, and Nb partitioning into, Ti oxides, such as rutile or perovskite. Wawa fibrous diamonds have 87Sr/86Sr values of 0.7042–0.7078. Estimates of the initial Sr isotope ratios of the fluid source have relatively large errors because of uncertainty in the Rb content of the fluids and their old age (Figure 3-4). The data have a general positive correlation between Rb/Sr and 87Sr/86Sr ratios but they do not form a geologically reasonable isochron greater than or equal to the age of the metaconglomerate host. This, and the variability in their initial Sr isotope compositions indicate that the contributing fluids were isotopically heterogeneous. The variability is in line with Sr isotope heterogeneity observed in fibrous diamonds from the Democratic Republic of Congo (Akagi and Masuda, 1988) and Botswana (Klein-BenDavid et al., 2010).

) d e Nitrogen z li a Carbonate m ater r ater Diamond Hydrogen W Carbonate W Olivine, garnet Olivine, n o 2 ss CO n e

k W1 c

h i W7 t ( n o i W9 p t Fibrous coat Fibrous r o

b s W29 A W14

W50 Cuboid 4000 3000 2000 1000 -1 Wavenumber (cm ) Figure 3-2. Infrared spectra of Wawa (Canada) diamonds, shifted vertically for clarity.

“Fibrous high” “Fibrous low” Rege et al., 2010

1 ×10-2 1 ×10-3 Weiss et al., 2009 Saline “table” in Rege et al., 2010 1 ×103 Saline “bench” 1 ×102 d 2

e 1 ×10 z 1 ×101 Ekati (Tomlinson ali 1 ×101 et al., 2009) m r 0

no 1 ×10 1 ×102 Wawa samples e t

i -1

r 1 ×10 1 ×101 W1(A1) -2 1 ×10 W1(A2) hond c -3 W7 1 1 ×10 C

W9 coat Fibrous -4 1 ×10 Wawa Ti data W14 1 ×10-5 are below LOD W29 Rb Ba Th U Nb Ta La Ce Pr Sr Nd Pb Zr Hf SmEu Ti Gd Tb Dy Y Ho Er Yb W50 Cuboid

Figure 3-3. Trace element patterns of Wawa (Canada) fibrous diamonds, normalized to C1 chondrite (McDonough and Sun, 1995), with patterns from the literature for comparison. LOD—limit of detection.

0.700 0.704 0.708 0.712 0.716 0.720

0 Ga

DRC (71 Ma) 2.7 Ga (Akagi and Masuda, 1988) Botswana (93 Ma) 0.700 0.704 (Klein-BenDavid et al., 2010) 87 86 Wawa (2.7 Ga) (2σ error bars) Sr/ Sri (cuboid, coat) Bulk silicate Earth Depleted MORB-source mantle 87 86 Figure 3-4. Calculated Sr/ Sri for Wawa (Canada) diamonds at 2.7 Ga, with literature data for fibrous diamonds, which are ca. 0 Ga in age. The Wawa fibrous diamond coats have smaller errors than cuboids, due to higher Rb levels. Bulk silicate Earth at 0 Ga and 2.7 Ga are plotted to show relative isotopic enrichment/depletion. DRC—Democratic Republic of Congo; MORB—mid-oceanic ridge basalt.

3.3 Discussion

3.3.1 Parallels between diamond formation in the Neoarchean and Phanerozoic The following evidence suggests conformity between the ca. 2.70 Ga Wawa fibrous diamond fluids and much younger, Phanerozoic, examples. Firstly, nitrogen contents of the Wawa diamonds fall into the range of Phanerozoic fibrous diamonds, 150–2400 ppm with a 1000–1200 ppm mode (Cartigny, 2005; Klein-BenDavid et al., 2010), attesting to similar nitrogen contents of the diamond-forming fluids. Secondly, the saline fluids in Wawa fibrous diamond mimic the compositions of saline varieties of Phanerozoic fibrous diamonds worldwide (Izraeli et al., 2001; Klein-BenDavid et al., 2007a; Zedgenizov et al., 2007a) (Figure 3-1). Carbonatitic and silicic fluid compositions are not observed, either due to nonexistence of such fluids in the Archean or the limited number of samples. A similar paucity of carbonatitic and silicic fluids in some Phanerozoic diamond populations, such as at Ekati (Canada) or Koffiefontein (South Africa), favors the latter option. A third similarity between Neoarchean and Phanerozoic fibrous diamonds is demonstrated by trace element characteristics. Remarkably, Wawa fibrous diamonds exhibit nearly identical major and trace element ratios to some Ekati diamonds from the 53 Ma Panda kimberlite (Tomlinson et al., 2009) (shaded band in Figure 3-3). The similarity between Ekati and Wawa fibrous coats includes overlapping La/Nb and Th/Nb ratios, which have been suggested to characterize different types of trace element pattern shapes in fibrous diamonds (Weiss et al.,

2009b; Navon et al., 2011). They also have Srn/Ndn > 1, a feature that has only been observed in saline diamond fluid inclusions (McNeill, 2011). Wawa diamond trace element systematics (Figure 3-3) generally match Phanerozoic fibrous diamonds, in that they have high degrees of inter-element fractionation, particularly between Rb, Ba, Th, U, Nb, and REEs. The patterns are similar to the “table” type patterns of Weiss et al. (2009a), characterized by high Ba, Th, U, and La compared to Nb. However, the Wawa fluids have higher Sr/Nd, Zr/Sm, and Hf/Sm. From this evidence, it appears that saline diamond-forming fluids are not unique to the Phanerozoic, but were also generated in the Neoarchean. Diamond-generating processes that operated in the Slave mantle at ca. 60 Ma, shortly before Ekati and Diavik kimberlite volcanism (Creaser et al., 2004), may resemble the Neoarchean scenario that led to the growth of Wawa

fibrous diamonds. Possible fluid sources are long-term, episodic degassing from the lower mantle (Allègre et al., 1987) or a repeatable style of volatile escape from episodically subducted or delaminated lithosphere (Helmstaedt et al., 2010), and might be linked to kimberlite source regions.

3.3.2 Fluid origins from non-enriched mantle While the major and trace element characteristics of a fluid elucidate chemical processes at the time of the diamond formation, radiogenic isotopes offer a complementary glimpse into the 87 86 longer, time-integrated history of the fluid source. Calculated initial Sr isotope ratios ( Sr/ Sri) for the diamonds at 2.7 Ga yield values that are within error of estimates of depleted mantle (Machado et al., 1986) or bulk silicate Earth (BSE) at 2.7 Ga (Figure 3-4). While the 87 86 uncertainties in Sr/ Sri are necessarily large, these unradiogenic fluids strongly favor a source that had experienced no enrichment of Rb/Sr relative to BSE at 2.7 Ga. It is possible the source is an Archean equivalent of modern-day group I kimberlite sources, with 87Sr/86Sr near BSE (Nowell et al., 2004). Fluid derivation from a kimberlite-like source is an attractive option, considering the brief timing between Wawa fibrous diamond growth and volcanic sampling via 87 86 kimberlite. Furthermore, kimberlite-like fluids have been tied to Sr/ Sri values in ca. 93 Ma fibrous diamond from the Democratic Republic of Congo (Figure 3-4) (Akagi and Masuda, 1988). A non-lithospheric, convecting mantle, kimberlitic or carbonatitic source was also proposed for Botswanan fibrous diamond, but in this case, the infiltrating non-enriched fluid was variably mixed with an enriched component with high 87Sr/86Sr (Klein-BenDavid et al., 2010). The high-87Sr/86Sr component was proposed to represent ancient mica-rich veins in the lithosphere that formed or matured over billion-year time scales. As the Botswanan, and probably the Democratic Republic of Congo, fibrous diamonds fluids fall within the silicic—low-Mg carbonatitic compositional trend, it should be noted that the Wawa diamonds provide the first Sr isotope data for saline diamond fluids. Thus, their non- 87 86 enriched Sr/ Sri values suggest that the more enriched component observed in Botswanan diamonds may be restricted to the silicic—low-Mg carbonatitic trend, perhaps associated with the silicic end member as input from mica-rich veins. The lack of involvement of an enriched mantle component in the genesis of the Wawa diamonds, despite the evidence for the existence

of such components in Archean mantle (Richardson et al., 1984; Pearson et al., 1995), indicates that fluids from enriched lithospheric mantle are not a prerequisite for fibrous diamond growth. 87 86 Future isotopic studies will show if unradiogenic Sr/ Sri values are also typical for other suites of saline fibrous diamonds. 87 86 The low Sr/ Sri in Wawa fluid inclusions was inherited from a source with a low time- integrated Rb/Sr ratio. An additional input of Rb at the time of diamond growth is indicated by the presently “enriched” Rb/Sr ratios that are 2–8 times higher than BSE. The high Rb/Sr, extreme K content, and “enriched” trace element signature of the fluid are probably linked, reflecting a transient event, closely preceding diamond and kimberlite formation. This event might involve low-degree partial melting of the source, exsolution of the fluid, or mixing between a low-87Sr/86Sr component and a high-Rb/Sr component with relatively little Sr, so as to not destroy the low 87Sr/86Sr signature.

3.3.3 Mantle Eu anomalies and saline fluid

Some samples of Wawa fluids show positive Eu anomalies (Eun / Eun* = 7–14 in fibrous coat samples) (Figure 3-3). Eu anomalies in mantle rocks are commonly ascribed to plagioclase accumulation or removal in the crust or shallow mantle, and a subduction origin is inferred (e.g. Jacob, 2004). The findings presented here suggest that this interpretation may not be correct if samples interacted with saline, Cl-rich fluid. A combined analysis of REE concentrations in fluids of different compositions in fibrous diamonds worldwide reveals the apparent restriction of Eu anomalies to saline fluids, including those in Wawa and Ekati fibrous diamonds (Figure 3-5). A subduction origin for this anomaly, however, would disagree with the consistent, convecting mantle-like carbon and nitrogen stable isotopes in fibrous diamonds (Boyd et al., 1992). A comparable association between Eu and saline fluid is regularly observed at seafloor vents, where emerging hydrothermal fluids have pronounced positive Eu anomalies (Michard et al., 1983) that correlate with fluid ligand concentrations rather than with rock composition (Craddock et al., 2010). As dissolved Cl− is an efficient transporter of REEs, but forms stronger complexes with Eu relative to other REEs (Flynn and Burnham, 1978), the fluid itself is capable of fractionating Eu. Such a relationship has been experimentally demonstrated for aqueous fluid and peridotite at 400

°C and 50 MPa, producing anomalies up to Eun / Eun* ≈100 (Allen and Seyfried, 2005), as well as for aqueous fluid and granitic melt at 800 °C and 400 MPa (Flynn and Burnham, 1978). These experiments show that the affinity of Eu for Cl-rich fluid increases with temperature, pressure, and Cl concentration. Furthermore, the reducing conditions that help convert Eu3+ to Eu2+ at surface broaden with increasing temperature (Bau, 1991), permitting enhanced Eu mobility over a wider range of redox conditions at mantle temperatures. Cl-rich diamond-forming fluid should preferentially scavenge Eu as it percolates through the mantle, giving the fluid a positive Eu anomaly. Moreover, the occurrence of Eu anomalies in both Wawa and Ekati samples demonstrates that this metasomatic mechanism for the creation of Eu anomalies has operated in both the Neoarchean and Phanerozoic. Saline fluids in general may be an important agent for producing Eu anomalies in the upper mantle, such as the intra-xenolith Eu variations for clinopyroxene reported by Taylor et al. (2000). It is therefore suggested that Eu anomalies in mantle rocks cannot advocate a shallow origin unless possible interaction with Cl-rich fluid is considered and discounted.

Saline (n = 16)

High-Mg carb. (n = 13) Wawa (n = 10) Low-Mg carb. (cuboid, coat) Silicic (n = 20) Other fibrous diamonds Uncertain (n = 8) 0.1 1 10

Eu /Eu * n n Figure 3-5. Fibrous diamond Eu anomalies, grouped by fluid composition within the saline–high-Mg carbonatitic and silicic–low-Mg carbonatitic trends. (Data: Tomlinson et al., 2009; Klein-BenDavid et al., 2010; Rege et al., 2010, and references therein).

3.4 Summary and conclusions Reported here are the first analyses of samples of Neoarchean mantle fluid captured as inclusions in fibrous diamonds from the 2.701–2.697 Ga Wawa metaconglomerate (Canada). The K-, Na-, and Cl-rich carbonate-bearing saline fluid inclusions bear a strong resemblance to those of Phanerozoic fibrous diamonds. This similarity in major and trace elements, volatile 58

components, and nitrogen characteristics implies the uniformitarian extension of some mantle processes, including the formation of fluid-rich fibrous diamonds, back to the Neoarchean. Low 87 86 initial Sr isotope ratios ( Sr/ Sri = 0.700–0.702) for the diamonds at 2.7 Ga are within error of estimates of Archean depleted mantle or bulk silicate Earth at 2.7 Ga, consistent with a non- enriched, convecting mantle/group I kimberlite-like source of the fluid. Eu anomalies occur in both Neoarchean and Phanerozoic saline fibrous diamond fluid, but not in silicic or carbonatitic fluid varieties. This restriction suggests that interaction with Cl-rich fluid may lead to Eu anomalies in mantle rocks, which therefore should not be automatically assigned to a subduction origin when a petrogenetic history involving saline fluids is suspected.

4 N-rich fluid inclusions in octahedrally-grown diamond3

4.1 Introduction Inclusions in diamond offer valuable samples of the mantle. This is especially true for mantle fluids, which diamond can preserve due to its inertness and strength. However, most samples of mantle fluids studied to date come from a distinct, late-forming variety of "fibrous" diamond (e.g. Navon et al., 1988; Izraeli et al., 2001; Klein-BenDavid et al., 2006; Logvinova et al., 2008; Klein-BenDavid et al., 2009). The more familiar, gemmy, octahedrally-grown variety of diamond has rarely been reported to contain fluid inclusions (Gurney et al., 2010) and it is unclear if it precipitates from the same types of fluids as fibrous diamond (McNeill et al., 2009). Direct analysis of the fluid equilibrated with this type of diamond is invaluable for testing models of diamond formation, while scientists argue whether diamond precipitates from the

CH4- or CO2-bearing fluids (Stachel and Harris, 2009). Thus it is uncertain whether reduction or oxidation fixes deep carbon, and the mobility and fluxes of carbon in the deep ancient mantle are poorly understood. Here we describe fluid inclusions in octahedrally-grown diamonds from 3 different localities. At the conditions of diamond formation there can be a supercritical continuum between melt and fluid, making the distinction between the two ambiguous (Stachel and Harris, 2008). At room temperature, however, the inclusions described here are either entirely liquid/gas or are dominated by solids, and will be referred to as fluid and melt, respectively. While recognizing melts and fluids in octahedrally-grown diamond is in itself a significant finding, the unexpectedly nitrogen-rich volatile content of the inclusions is perhaps more surprising. N2 is scarcely detected in fluid inclusions in mantle-derived xenoliths (Andersen and Neumann, 2001). The inventory and geodynamic behaviour of the ~1 ppm nitrogen content of the Earth's mantle is poorly understood (Marty, 1995; Marty and Dauphas, 2003; Cartigny and Marty, + 2013). Mantle nitrogen is believed to be mostly recycled NH4 from surface, based on the high

3 This chapter has been published. Smith, E.M., Kopylova, M.G., Frezzotti, M.L. and Afanasiev, V.P., 2014. N-rich fluid inclusions in octahedrally-grown diamond. Earth and Planetary Science Letters, 393(0): 39-48. Additional, unpublished data has been provided in appendices. 60

36 N2/ Ar ratio of the mantle relative to the atmosphere (Zhang and Zindler, 1993; Marty and + Dauphas, 2003). Surficial nitrogen enters the mantle primarily as NH4 substituting for K in micas (Juster et al., 1987; Williams et al., 1992; Goldblatt et al., 2009; Watenphul et al., 2009).

In contrast, nitrogen is released from the convecting mantle as molecular N2, appearing as a minor component in volcanic gases in mid-ocean ridge basalts and mantle-derived fluid + inclusions (Marty, 1995; Fischer and Marty, 2005). The relationship between incoming NH4 + and outgoing N2 is unclear. If the outgoing N2 is derived from recycled NH4 , then there should + be a geological process in the mantle oxidizing NH4 . The mechanism for liberating nitrogen from the convecting mantle remains uncertain. A further complication in the deep nitrogen cycle arises upon considering nitrogen isotopes. Nitrogen influx to the mantle is dominated by isotopically heavy, sedimentary nitrogen, with positive δ15N values around 5‰ (parts per thousand (‰) deviation in 15N/ 14N relative to

atmospheric N2), whereas effluxed nitrogen in mid-ocean ridge basalts, diamonds, and other upper mantle samples falls around -5‰, being isotopically light (Cartigny et al., 1998b; Marty and Zimmermann, 1999; Marty and Dauphas, 2003; Cartigny and Marty, 2013). This puzzling isotopic mismatch has been interpreted as the isolation of the upper mantle, light nitrogen reservoir from a heavier, subducted, lower mantle reservoir (Marty and Dauphas, 2003). This interpretation, however, is difficult to reconcile with current ideas of mantle convection involving material exchange between the upper and lower mantle (Bercovici and Karato, 2003; Tackley, 2012) and isotopically light N compositions measured in lower mantle diamonds (Palot et al., 2012). Developing a better understanding of N exchanges between surface and mantle reservoirs is a key factor for understanding the evolution of the atmosphere (Marty, 2012). Fluid and melt inclusions in diamonds reported in this study hold the clues to the behaviour of mantle N. Here we show that diamond-forming processes can lead to unexpectedly high

residual N2 concentrations that have no clear mineral repository. The inclusions further suggest + that the oxidation of mineralogically bound NH4 to mobile N2 liberates stored nitrogen from the + mantle. Isotopic fractionation between NH4 and N2 during this redox process could help explain the isotopic mismatch between incoming and outgoing mantle nitrogen.

4.2 Diamond samples The studied diamonds include samples from the Siberian, Congo, and Kaapvaal cratons (Figure 4-1, Table 4-1). The 4 studied Siberian samples were selected from 12 diamonds sourced from alluvial deposits of the Ebelyakh River of the northeastern Siberian Platform. They belong to a common variety of alluvial diamonds in the region that have octahedral growth forms, with overall rounded morphologies due to varying amounts of dissolution and mechanical abrasion. These diamonds contain cracks and etched channels in their surfaces, and abundant flat graphitic inclusions that discontinuously trace healed internal fracture networks and give the diamonds an overall greyish colour. Some diamonds appear yellow/orange due to iron oxides precipitated within surface-reaching cracks. This distinctive population of Siberian diamonds has been noted to contain fluid inclusions (Tomilenko et al., 1997; Logvinova et al., 2011) and mineral inclusions of coesite (Ragozin et al., 2002), rutile (Afanasyev et al., 2009), and kyanite in unfractured parts of the samples, indicating these diamonds grew in eclogitic host rocks (Appendix H: Eclogitic inclusions in Siberian diamonds).

Congo Kaapvaal

a b

Siberian

c d e f

Figure 4-1. Octahedrally-grown diamonds containing the studied fluid and melt inclusions, from the Congo (a), Kaapvaal (b), and Siberian (c-f) cratons, with a 1 mm scale bar for each.

The diamonds from the Congo craton, from the Democratic Republic of the Congo, and Kaapvaal craton, from the Roberts Victor kimberlite in South Africa, are single octahedral crystals with minor dissolution of their edges. The Congo and Kaapvaal craton diamond samples (1 each) were selected from a few tens of diamonds after polishing windows in samples and recognizing potential melt/fluid inclusions visually (Appendix K: Diamond polishing equipment). Nitrogen incorporated into the diamond lattice during initial growth was characterized with infrared spectroscopy. Nitrogen contents in one of the Siberian samples, Sib4, were determined to be 1160 ppm, with 65% as A centres (N pairs) and 35% aggregated to B centres (N clusters of 4). Other diamonds of this distinctive Siberian population possess similar nitrogen characteristics (Logvinova et al., 2011). The diamond lattice in the Kaapvaal sample contains 520 ppm nitrogen, with 60% aggregated to B centres. These lattice-bound nitrogen contents and aggregation states compare well with other cratonic diamonds (Cartigny, 2005). All the selected diamond samples have inclusions hosted along healed fractures. The healed fractures tend to be optically invisible, but can be seen in cathodoluminescence images of polished sections through a diamond's interior. The fractures are secondary to the initial diamond growth, but predate later diamond formation that healed the fractures and, in the Congo diamond, continued to form a new layer of octahedral diamond (Figure 4-2). Although diamonds tend to deform plastically above about 900 °C, high strain rates inevitably lead to fractures due to brittle failure (Brookes et al., 2000). Brittle deformation and subsequent healing has been recognized previously in cathodoluminescence studies of natural diamonds (e.g. Anand et al., 2004; Taylor and Anand, 2004). Fractures like these make ideal traps for mantle volatiles, capturing samples of diamond-forming melts or fluids. The relative ease with which these inclusions were found for the present study suggests they are not uncommon. In spite of this, melt or fluid inclusions are rarely documented in octahedrally-grown diamond, the most common variety of natural diamond (Gurney et al., 2010). They might be underreported in previous work, being dismissed when fractures are recognized or overlooked in favour of larger, more pristine-looking mineral inclusions. Indeed, diamonds with conspicuous fractures are generally avoided in inclusion studies, for fear that any entrained inclusions may be contaminated or otherwise modified by secondary processes. 63

a Congo (cathodoluminescence) 1 2 3 b Congo

c Siberian d Siberian e Kaapvaal

Figure 4-2. Studied inclusions in diamonds. a) Cathodoluminescence image of a {100} polished surface through the Congo diamond with a healed fracture, hosting melt inclusions. The diamond growth healing the fracture and forming the outermost layer has darker luminescence. Scale bar is 100 µm. The cartoon inset shows the initial octahedral diamond (snapshot 1) was fractured (snapshot 2) and subsequently healed upon the arrival of new fluid/melt, some of which became trapped in the fractures, that grew a new outermost layer of diamond (snapshot 3). Microphotographs show silicate melt inclusions with exsolved N2 bubbles in diamonds from the Congo (b) and Kaapvaal cratons (e) (arrow indicates bubble). c,d) Mixed CO2–N2 Fluid inclusions with polygonal boundaries crystallographically imposed by diamond in Siberian diamonds. Scale bars in b, c, d, and e are each 10 µm.

4.3 Analytical methods Raman spectra were collected with a Jobin-Yvon LabRAM spectrometer at the University of Siena, Italy. The spectrometer has a Peltier-cooled CCD detector and an Argon-ion (514.5 nm) laser set to an emission power of 450 mW. The laser power reaching the sample surface was 80%. Confocal optics with an Olympus 100× lens focussed the laser to analyze a volume of about 1×1×5 µm3. The confocality eliminates out-of-focus light, thereby rejecting Raman scattered light that might come from gases such as N2 or CO2 in the air space above the sample.

When N2 or CO2 are detected in an inclusion, atmospheric contributions to that signal can be

checked and ruled out by analyzing the mineral volume just adjacent to the inclusion. This should give a signal without any contribution from the inclusion, but that still contains any potential atmospheric signal. Scattered light was analyzed with a 100 µm slit and a spectral -1 resolution of 1.5 cm , using a 1800 grooves/mm grating, allowing discrimination between CO2 and diamond peaks. Collection times varied between 5 and 900 s, with 1 to 5 accumulations depending on the sample fluorescence and signal intensity. Quantitative Raman determinations

of CO2 and N2 molar proportions were made using the following equation (Wopenka and Pasteris, 1986; Frezzotti et al., 2012b):

Equation 4-1: Xa = [Aa/(σa ξa)]/Σ[Ai/(σi ξi)] where Xa, Aa, σa , and ξa are the molar fraction, peak area, Raman cross-section, and

instrumental efficiency for species a, and Ai, σi , and ξi are respectively corresponding values for all species, while Σ denotes the sum. Peak areas for all species should be taken from spectra of the same collection time, at the same point of analysis. Movement, repositioning of the sample, or a change in microscope focus between measurements would result in inaccurate quantification. Fluid inclusion microthermometry for the Siberian diamond fluid inclusions was also carried out. The fluid behaviour upon cooling agrees with the Raman measurements, but is otherwise not instructive (Appendix G: Microthermometry). Cathodoluminescence imaging was done with a CLmk3A cold-cathode optical cathodoluminescence stage, operating at 15–17 kV and 340–400 µA, at the University of British Columbia. Electron probe microanalysis (EPMA) was conducted with a Cameca SX-50 microprobe at the University of British Columbia, using a 15 kV accelerating voltage and 20 nA beam current. Na, Mg, Al, Si, Ti, Cr, Fe, and Ni peaks were measured with a 20 second count time. Ca was measured with a 60 second count time in an effort to lower the detection limit. Infrared spectroscopy was done at the University of British Columbia with a benchtop Varian 3100 FTIR connected to a Varian 610-IT microscope, with a liquid nitrogen cooled MCT detector. Spectra were collected in transmission mode in the range 650–4000 cm-1 and resolution of 8 cm-1. Each spectrum was given an appropriate baseline and normalized for sample thickness using the intrinsic diamond absorption at 2000 cm-1. Spectra were then processed using a deconvolution macro in an Excel spreadsheet developed by D. Fisher (Diamond Trading

Company). The macro fits the nitrogen absorption region with reference nitrogen peaks to determine contributions of A and B centres to the spectrum. The detection limit for nitrogen in the diamond lattice is approximately 10 ppm. Nitrogen contents were determined in Kaapvaal sample and the thinnest Siberian sample (Sib4) (Appendix F: Infrared spectra). Thicker Siberian samples are expected to be similar, but were not quantified as due to complete absorption of the infrared signal in the nitrogen region. For the Congo diamond, its large size and internal fracturing prohibited transmission of the infrared beam, so nitrogen content of the diamond lattice could not be measured.

4.4 Results

4.4.1 CO2–N2 fluid inclusions The Siberian diamonds trapped platelike (~1 µm thick) volatile inclusions with polygonal boundaries imposed by the host diamond, while the Congo and Kaapvaal diamonds trapped melt

that subsequently exsolved its volatiles as a small bubble (Figure 4-2). N2 and CO2 were identified by Raman spectroscopy within fluid inclusions and volatile bubbles (Figure 4-3; Table

4-1). No other volatiles were detected. If hydrocarbons were present, notably CH4, they would be readily detected due to higher scattering efficiencies compared to N2 and CO2 (Burke, 2001).

Table 4-1. Raman quantification of CO2 and N2 in analyzed diamond inclusions. Average Molar Standard Paragenesis No. of No. of Volatile Craton Locality % deviation (host rock) diamonds inclusions occurrence CO2 N2 (molar %) Northeastern Siberian Siberian Platform, eclogitic 4 17 fluid inclusion 60 40 4 Russia bubble in melt 1 3 7 93 2 inclusion Democratic Republic Congo peridotitic of Congo bubble in melt not 9 100 inclusion detected

Roberts Victor Mine, bubble in melt not Kaapvaal peridotitic 1 1 100 South Africa inclusion detected

The mixed CO2–N2 fluids in 17 inclusions from 4 Siberian diamonds consistently contain

40±4 mol% N2 (Appendix I: Raman quantification of nitrogen). Spacing between the two main

CO2 peaks (Fermi diad), which increases with fluid density (Wright and Wang, 1973), reaches

extreme values, up to 108.2 cm-1. This value implies even greater pressures than the 106 cm-1

value that qualifies pure CO2 fluid inclusions as "superdense" and likely mantle-derived

(Frezzotti et al., 2012b). However, barometric calibrations for CO2–N2 mixtures are unavailable. Trapping conditions would be difficult to constrain even with the calibration, as the fluid density variation with depth, for example from 95 to 190 km along a typical cratonic geotherm, would be

small, only 2% (Soave, 1972). Nevertheless, the extreme CO2 peak splitting is consistent with a mantle origin. More importantly, the evidence for subsequent growth of diamond to heal the fractures and trap fluid (Figure 4-2) signifies fluid trapping in the diamond stability field. Trapping at 4.5–7 GPa and 1000–1300 °C would give a residual pressure on the order of 1 GPa at room temperature (Soave, 1972). Contamination of the fluid inclusions by atmospheric nitrogen can be ruled out given the high remnant inclusion pressures and negligible ability for nitrogen diffusion through the diamond lattice (Koga et al., 2003). 2,330 5

1,283 Congo 1,387 4

Siberian 2,332 1,388 2,332

Relative intensity Relative 2 1,281 Kaapvaal 1 2 2 2 CO CO diamond N 0 1,200 1,300 1,400 2,300 2,400 -1 Raman shift (cm ) Figure 4-3. Representative Raman spectra for fluid inclusions in a Siberian diamond, and volatile bubbles in melt inclusions in diamonds from the Congo and the Kaapvaal cratons.

4.4.2 N2-rich silicate melt inclusions Volatile bubbles were analyzed within 12 irregularly shaped melt inclusions of the Congo diamond and 1 melt inclusion in the Kaapvaal diamond (Figure 4-2). The bubbles are dominated

by N2, and only 3 bubbles had detectable amounts of another species, CO2, which was <9 mol.% 67

in all cases (Table 4-1). This is only the second time mantle volatile compositions reaching 100

mol.% N2 have been reported for any sample (Andersen and Neumann, 2001). Consistent

occurrence of bubbles in the larger inclusions attests to the N2 having been dissolved in the melt at the time of trapping. Smaller inclusions lack bubbles, which is a feature commonly observed in melt inclusion studies due to the difficulty of bubble nucleation (Lowenstern, 2003).

The surrounding solid (former melt) (Figure 4-2) in the Congo diamond gives sharp Raman

peaks for olivine ((Mg,Fe)2SiO4) (Figure 4-4) and, in some inclusions, a broad band centred on 880 cm-1 from silicate glass with low polymerization of silica tetrahedra (McMillan, 1984). Raman spectra confirm the presence of olivine crystal lattices, although there are no visible grain boundaries delineating discrete olivine crystals. Electron microprobe analysis of inclusions exposed at the polished surface (Table 4-2) shows micron-scale compositional variability, with compositions near olivine, but having excess Si (up to 1.03), coupled with a deficit in Mg+Fe (down to 1.94), compared to olivine stoichiometry. The presence of volatile bubbles, the round, droplet shapes of these inclusions (Figure 4-2), as well as non-stoichiometric composition of the material all attest to the interpretation that these inclusions were trapped as melt. Relative olivine peak intensities remain constant across an inclusion, indicating a single crystallographic orientation for the olivine. This uniform orientation and lack of visible crystal boundaries could be the product of rapid dendritic olivine growth within the melt after trapping. Measured compositions of the solid portions of the inclusions fall on a tie line from (Mg,Fe)2SiO4 to

(Mg,Fe)2Si2O6, between olivine and orthopyroxene compositions (Figure 4-4), with average modal proportions of 96% (Mg,Fe)2SiO4 and 4% (Mg,Fe)2Si2O6. A peridotitic host rock paragenesis for the Congo sample is inferred from this composition. A peridotitic paragenesis was also inferred for the Kaapvaal diamond based on the Raman shift (880 cm-1) of a silicate glass inclusion, characteristic of a low degree of polymerization of silica tetrahedra (McMillan, 1984). The near-olivine melt composition measured in the Congo sample is unusual for upper mantle melts. It would require high temperatures to remain stable as liquid. For example, the liquidus for the average composition lies above 1900 °C at 2 GPa, even with the addition of a several weight percent H2O, calculated using MELTS (Java applet 1.2.1). The melt could be 68

Table 4-2. Electron microprobe data for traverses across 3 melt inclusions in the Congo diamond sample. An olivine standard is given for comparison.

Inclusion 1 Inclusion 2 Inclusion 3 Average Olivine Points 1 2 3 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 (n = 19) standard Wt% SiO2 42.56 41.88 41.97 41.80 41.63 41.42 41.65 41.93 41.95 41.02 40.64 41.11 42.03 42.02 41.89 41.98 41.49 41.56 41.08 41.66 41.31 TiO2

Total 100.15 99.65 99.51 99.57 99.69 98.51 99.87 98.26 99.65 98.77 99.23 100.79 99.25 99.57 98.96 99.77 99.02 98.80 98.13 99.32 100.76

Cations per 4 oxygens (olivine stoichiometry) Si 1.03 1.02 1.02 1.01 1.01 1.02 1.01 1.03 1.02 1.00 0.99 0.99 1.02 1.02 1.02 1.02 1.01 1.02 1.01 1.01 1.00 Al 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Mg 1.79 1.81 1.80 1.82 1.82 1.81 1.82 1.78 1.80 1.84 1.86 1.86 1.80 1.81 1.80 1.81 1.81 1.81 1.82 1.82 1.83 Fe 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.17 Ca 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Ni 0.00 0.00 0.01 0.00 0.00 0.01 0.01 0.00 0.01 0.00 0.00 0.01 0.00 0.01 0.01 0.01 0.01 0.01 0.00 0.01 0.00 Cr 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

Total 2.97 2.98 2.98 2.98 2.99 2.98 2.99 2.97 2.98 3.00 3.01 3.01 2.98 2.98 2.98 2.98 2.99 2.98 2.99 2.99 3.00

Mg/(Mg+Fe) 0.924 0.922 0.922 0.925 0.923 0.924 0.924 0.921 0.922 0.926 0.925 0.925 0.924 0.925 0.925 0.923 0.923 0.924 0.925 0.924 0.916

produced by localized high-degree melting in a dunite. Alternatively, efficient reduction of an original carbonate melt component to diamond (i.e. melt decarbonation) may have left residual (Mg,Fe)O in the melt and increased its normative olivine content. If the temperature during lithospheric diamond formation is near 1200 °C (Stachel and Harris, 2008), then the degree of undercooling for the trapped melt would be severe and would strongly encourage dendritic olivine growth (Faure et al., 2003). Generating the observed ultramafic melt composition via melt decarbonation eliminates the need to invoke extreme temperatures, so long as decarbonation outpaces solidification. More work is necessary to determine whether decarbonation could account for the observed melt composition. Pressure-induced Raman peak shifts of olivine crystals precipitated from the melt indicate residual inclusion pressures on the order of ~1 GPa (Wang et al., 1993; Izraeli et al., 1999), with -1 -1 the main olivine (Fo92) 856 cm peak reaching up to 861 cm (Figure 4-4). Olivine barometry (e.g. Howell et al., 2010) cannot constrain the temperature and pressure conditions of trapping more accurately due to presence of glass and a volatile bubble in the inclusion.

68 t 861 i

s 886 en 66 t 830 n i

1.0 e v 64 i t a l O e g R 62 856

M 0.8

+ 60 750 800 850 900 950 -1

O Raman shift (cm ) e 58 0.6

F Olivine composition, (Mg,Fe)2SiO4

% 56 l

o 0.4

m 54 52 Orthopyroxene composition, 0.2 50 (Mg,Fe)SiO 3 48 32 34 36 38 40 42 44 46 48 50 mol% SiO2 Figure 4-4. Compositional range measured in 3 melt inclusions, defining a linear trend (R2 = 0.99) between olivine and orthopyroxene compositional end members. Gradations are shown for molar fraction of olivine from 1.0 to 0.0. The inset shows olivine peaks in the Raman spectrum from a melt inclusion. The peaks are shifted due to residual pressure in the inclusion, with the main peak shifted from 856 cm-1 to 861 cm-1. 70

The original dissolved N2 content of the melt, prior to bubble exsolvation, can be roughly

estimated by comparing bubble to inclusion size (Figure 4-2) and assuming that the density of N2 reflects mantle conditions, consistent with the remnant pressure in the olivine. A relatively large

inclusion with a defined spherical bubble was selected to estimate the bulk inclusion N2 content. The large size of the inclusion makes it liable to be representative of the bulk melt. The bubble is spherical with a 3.5 µm diameter, set in a tabular inclusion with an area of 530 µm2 and thickness of 10 µm. This gives a bubble/inclusion volume ratio of 0.004, which can be combined 3 with the density of the bubble and solid to give N2 content by mass. A density of ~0.93 g/cm is

expected for an N2 bubble in melt solidified within 4–6 GPa and 1000–1200 °C, in the diamond stability of the Congo cratonic lithospheric mantle (Soave, 1972). Melt solidification at less

extreme mantle conditions would still yield relatively high N2 densities. For example, a density 3 3 of >0.8 g/cm , or molar volume <35 cm /mol, is expected for N2 in the upper mantle, at conditions exceeding 1.2 GPa and 900 °C (Soave, 1972). Using a bubble density of 0.9 g/cm3 and surrounding solid density of 3.3 g/cm3, after forsteritic olivine, a density ratio of 0.27 is obtained. Multiplying the density ratio, 0.27, by the volume ratio, 0.004, gives an original

dissolved N2 content of ~0.1 wt%. The possibility that appreciable amounts of CO2 remained in the solid portion, rather than exsolving into the bubble, is deemed unlikely because the carbon solubility in olivine is low, around 0.1–1 ppm (Keppler et al., 2003), and the Raman spectra did not exhibit carbonate mineral peaks, which have a strong scattering ability compared to silicates (Frezzotti et al., 2012b).

4.5 Discussion

4.5.1 Nitrogen as a major mantle volatile phase Nitrogen is typically regarded as a minor mantle volatile. It is scarcely detected in mantle- derived xenolith fluid inclusions, which are overwhelmingly dominated by CO2 (Andersen and Neumann, 2001). The discovery of high nitrogen contents in melts and fluids from the Siberian, Congo, and Kaapvaal cratons (Table 4-1) establish that nitrogen can constitute major mantle volatile, at least in the context of diamond-related melts/fluids. Some previous reports have also shown nitrogen-rich volatiles in diamonds. Nitrogen was recognized earlier in fluid inclusions in Siberian alluvial diamonds, similar to the ones reported 71

on here, using Raman spectroscopy and microthermometry, but the nitrogen content could not be quantified (Tomilenko et al., 1997). Minute inclusions of pure nitrogen, several tens of nanometres across, have been found in recent transmission electron microscopy studies of microinclusions in diamonds from Juina, Brazil (Kaminsky and Wirth, 2013; Wirth and Yang, 2013). The relative nitrogen enrichment in the inclusions is also evident when expressed in the form of molar C/N ratios. These are useful for gauging the relative N enrichment in the volatile budget, which is helpful for comparing the Siberian fluid inclusions to the Congo and Kaapvaal melt inclusions that have markedly different absolute N concentrations (Figure 4-5). They are also useful when absolute concentrations for C and N cannot be deduced, as would be the case if the volatile bubbles in the melt inclusions were extracted and analyzed by gas chromatography.

1,000,000 Cratonic diamonds

100,000

10,000 Diamonds in

) this study r

a 1,000 l Mantle C/N o m

( 100 N /

C 10 C/N in gas from crushed diamonds 1 Siberian Congo 0.1 CO -N fluid melt 2 2 0.01 1 100 10,000 1,000,000 Nitrogen concentration (mass ppm) Figure 4-5. Comparison of C/N ratios and N concentrations in diamonds and the studied melt and fluid inclusions. The Congo melt inclusion C/N ratio shown is the maximum value, from the bubble with 9 mol% CO2 and 91 mol% N2. The two studied diamonds shown are from infrared spectroscopy for the one Kaapvaal diamond and one of the Siberian diamonds. The inclusions exhibit significantly lower C/N ratios, as well as higher N concentrations, than the majority of cratonic diamonds. Range of cratonic diamonds shown is based on 917 diamonds, and terminates at the 99th percentile, at 1400 ppm N (Stachel and Harris, 2009). The median lies at 116 ppm N (square symbol). The mantle C/N value shown is 535±224 (Marty and Zimmermann, 1999). Average C/N in gases for crushed diamonds are given for several localities: Africa, 2.1 (n = 42, non-fibrous), 8.6 (n = 3, fibrous cuboid) (Melton and Giardini, 1974); Siberia, 1.5 (n = 14) (Logvinova et al., 2011); India, Panna mine, 14 (n = 3) (Melton and Giardini, 1981); Brazil, fibrous cuboid, 2.9 (n = 1) (Melton and Giardini, 1974); Arkansas, 0.8 (n = 7) (Melton and Giardini, 1975). The overall average C/N of the gases from crushing is 2.7 (n = 70). 72

Mantle volatiles sampled at mid-ocean ridges have C/N values of 535±224, being

dominated by CO2 (Marty and Zimmermann, 1999). However, the C/N molar ratios in the studied inclusions are all below 0.8, significantly lower than the typical mantle range (Table 4-1, Figure 4-5. Similarly, earlier reports of gases analyzed from a total of 70 diamonds crushed in vacuo also show relatively low C/N, within 0.8–14 (Figure 4-5) (Melton and Giardini, 1974; Melton and Giardini, 1975; Melton and Giardini, 1981), supporting the notion that low C/N ratios are common for inclusions in diamonds. Of those 70 diamond, 66 were octahedrally- grown, non-fibrous diamonds, and only 4 have a description consistent with being fibrous

diamond, which is the variety that is normally rich in fluid micro-inclusions. The most N2-rich

gas released by crushing was 87 vol% N2 from a diamond from Arkansas (Melton and Giardini, 1975). Examination of the crushed fragments of this sample using secondary electron microscopy revealed 1–30 µm cavities inside the diamond, devoid of solids (Giardini and

Melton, 1975). Gas chromatography further shows that N2 is a significant and sometimes

dominant volatile in diamond-related melts/fluids. It can surpass CO2 and H2O in mantle melts and fluids. Inclusions like these may be more common than currently recognized because, if trapped in other minerals besides diamond, they would probably not survive decompression. Fluid inclusions trapped at extreme depth (>100 km) in most minerals would likely rupture upon transport to surface (Touret and Huizenga, 2011; Frezzotti et al., 2012a). Owing to its physical and chemical robustness, diamond may be the only mineral that can readily retain these highly pressurized volatiles near surface, to give direct samples of the nature of volatiles from extreme depths. In addition to diamond, mantle xenoliths from the Canary Islands have been found with

nearly pure N2 fluid inclusions (Andersen et al., 1995). These inclusions demonstrate that N2 can be a major volatile in other mantle settings, unrelated to diamond. Thus, although N tends to be a trace component at ~1 ppm in the silicate Earth, with a usual C/N ratio of 535±224 (Marty, 1995;

Marty and Zimmermann, 1999), there are processes that can render N2 a major mantle volatile species and drastically lower the C/N ratio.

4.5.2 Attaining high N2 contents

The high N2 content of fluids observed in our samples could be interpreted either as an inherent characteristic of the diamond-forming media or as residual after diamond precipitation. The former option would require a uniquely N-rich mantle source, with low C/N ratios. The latter option is preferred because it is well established that N isotopes associated with cratonic diamonds are consistent with a non-unique, ambient convecting mantle source (Cartigny, 2005). A non-unique mantle N source fits well with the observation that the inclusions come from 3 different cratons. The N-rich fluid and melt inclusions reported here indicate that the growing diamond left a substantial amount of N in the inclusions. The C/N ratios in all inclusions measured (Figure 4-5) are significantly lower than typical mantle C/N values of 535±224 (Marty and Zimmermann, 1999), which characterize not only mid-ocean ridge basalts, but also carbonatites, such as at the Oldoinyo Lengai volcano (Fischer et al., 2009). A straightforward explanation for the contrast is that the low C/N values in the inclusions have evolved from higher, initially mantle-like values. An initial C/N ratio nearer to 535±224 is supported by calculations of original C/N ratios of diamond forming media based on δ13C and N concentrations in diamonds (Cartigny et al., 2001). These calculated values, combined with the fact that the lowest C/N ratios in cratonic diamonds approach ~500, indicate that initial C/N ratios in diamond growth media are likely to be near mantle C/N values (Figure 4-5). Diamond growth or some independent mechanism must have lowered the C/N ratios. For

the case of the CO2– N2 fluids in the Siberian diamonds, a potential contributing factor to the low C/N ratios may be the exsolution and escape of these fluids from a precursory carbonatitic melt

in an eclogite host rock setting. As modelled by Cartigny et al. (2001), N2 will preferentially be

incorporated into the escaping CO2 phase, resulting in a CO2– N2 fluid with lower C/N than the melt. However, the effect will only decrease the C/N of the escaping phase by a factor of 2, meaning that this mechanism cannot produce the observed low C/N ratios in the Siberian diamond fluid inclusions. Reducing the C/N ratio of the diamond growth medium from ~535 down to 0.8 requires removal of at least 99.8% of the C, or more if N is also removed. While the residual fluid and melt inclusions tend towards low C/N, diamonds themselves have higher C/N values, above 74

mantle values (Figure 4-5). Based on measurements from 917 cratonic diamonds, 99% of cratonic diamonds have C/N ratios >800 (Stachel and Harris, 2009). Reaching the median C/N value of 11,000 (91 at. ppm N) (Stachel and Harris, 2009) from the starting C/N of ~535 would require that no more than 5% of the N is sequestered into the diamond from the parent medium (535×(1/0.05) = 11,000). C/N ratios in the inclusions and diamonds agree that the majority of C is drawn out of the medium into the diamond, while the majority of N remains in the medium. The low C/N ratios observed in the melt and fluid inclusions can therefore be attained by diamond growth. This finding contradicts aspects of recent models of N partitioning behaviour during diamond crystallization (e.g. Thomassot et al., 2007; Stachel et al., 2009; Smart et al., 2011; Wiggers de Vries et al., 2013). These models are unable to account for the low C/N observed in the inclusions reported here. These models, which are intended to describe C isotopic fractionation, may need to be re-evaluated from the perspective of N incorporation into diamond as a function of growth kinetics rather than equilibrium partitioning (Cartigny et al., 2001). Absolute N concentrations in the inclusions are less straightforward to explain, although it is clear from the observed C/N ratios that most N will remain in the residual melt or fluid, regardless of whether N is considered compatible or incompatible. As long as no new

components are added to the melt/fluid, its absolute N concentration ([N]melt) relies on the ratio of N to non-nitrogen components removed, due to the crystallization of diamond and/or other minerals. If we consider a precursory, carbonatitic mantle melt with a C content of 10 wt%

(Dalton and Presnall, 1998) and C/N of ~535 (Marty and Zimmermann, 1999) the [N]melt will be

0.02 wt%. Using 0.02 wt% as a benchmark for the highest expected initial [N]melt, reaching a

[N]melt of 0.1 wt%, like the melt inclusions in the Congo diamond, requires the removal of at

least 78% of the non-nitrogen melt component. Removal of C or CO2 from the melt, by diamond growth, cannot account for this much mass. Removal of additional melt components through

crystallization would further allow [N]melt to increase. Considering the ultramafic melt composition in the Congo diamond, for which olivine crystallization is to be expected, crystallization of minerals other than diamond will likely affect [N]melt during diamond growth. Taking the melt inclusions as a proxy for the fate of non-included melt would indicate that crystallization can drive [N]melt to the point of saturation and exsolution of N2. Diamonds grown 75

from this melt should tend towards extreme N contents if N is compatible. The absence of such zoned N-rich diamonds suggests either that N is incompatible in diamond (Cartigny et al., 2001)

or diamond growth tends to outpace the increase in [N]melt due to melt crystallization.

4.5.3 Implications for the mantle nitrogen cycle

4.5.3.1 Nitrogen degassing and efflux When considering N effluxes from the mantle, the focus is typically on volcanic activity such as at mid-ocean ridges (Busigny et al., 2011). N is considered as a dissolved component carried in mantle melts, with concentrations at parts per million levels. However, our observations show that surplus N sourced from the convecting mantle can become concentrated at depth in some mantle settings, and potentially degas in a mobile fluid phase. Once concentrated to high degrees, the N has few potential mineral depositories, either within mineral lattices or as fluid inclusions. Even diamond, which routinely contains two orders of magnitude more nitrogen in its crystal lattice than other upper mantle minerals (Cartigny et al., 2001), was unable to accommodate the nitrogen budget of the diamond-forming media trapped in our

samples, leaving excess N2 in the melt and fluid inclusions. In instances where mantle melts react with and solidify in the lithospheric mantle, it is

probable that N2 becomes concentrated in the residuum. In the absence of upper mantle mineral

depositories, the N2 may continue to percolate upward, and escape non-volcanically at surface. In support of this possibility, isotopic signatures of N2, CO2, and noble gases sourced from the sublithospheric mantle have been recognized in soil gases, dry gas vents, and thermal and cold springs associated with faults and deep fractures (Bräuer et al., 2004; Frezzotti et al., 2009). Mantle volatiles, released into the base of the crust from crystallizing mantle melts, have also been recognized in fluid inclusions in lower continental crustal granulites (Touret, 1992).

The possibility of additional, previously unrecognized N2 degassing from the sublithospheric mantle warrants consideration from the perspective of N fluxes into and out of the mantle. Considerable net influx of surficial nitrogen to the mantle is suggested by comparing recent estimates of nitrogen input by subduction (13.2×1011 g/yr) to the total output (3.6×1011 g/yr), at mid-ocean ridges, back-arc basins, and arc/intraplate volcanics (Busigny et al., 2011). These fluxes mean that about two thirds of subducted nitrogen is not balanced by a return flux to 76

surface. Flux imbalance is also invoked in the proposal that long term sequestration of

atmospheric N2 in the mantle has diminished atmospheric pressure, and thereby, the greenhouse effect since the Archean, as a resolution to the faint young sun paradox (Goldblatt et al., 2009). However, extrapolating the 9.6×1011 g/yr net flux (Busigny et al., 2011) back over 3 billion years of modern plate tectonics (Shirey and Richardson, 2011), would add 2.6×1021 g of recycled N to the mantle. Adding 2.6×1021 g of N to the mantle is unfeasible when compared to a recent estimate of the total mantle N content of 6.4×1020 g, for a non-chondritic Earth model (Halliday, 2013), or feasible but substantial compared to another estimate of 4.3×1021 g (Marty, 2012).

Furthermore, such a dramatic change in atmospheric N2 pressure is not supported by Archean

quartz fluid inclusions, which suggest no significant change in atmospheric N2 content since the Archean time (Marty et al., 2013). Conflicting lines of evidence like these leave room for additional, unaccounted for N efflux from the mantle. Our calculations show that mantle degassing from low-degree melts is unlikely to produce the enormous N efflux needed for steady state conditions, but it may constitute a sizable N flux. The N efflux from the asthenosphere due to carbonate-rich melts has been estimated in two ways. The first is based on the supply rate of kimberlites in the Lac de Gras region of the Slave craton. Kimberlite volcanism has emplaced about 270 pipes here over a 29 million year span, which works out to an estimated average mass supply of 107– 109 g/yr (Patterson and Francis, 2013). No matter what its N concentration, this amount of melt is far too small to contribute significantly to N efflux from the mantle. Using the N concentration from the melt inclusions in the Congo diamond, 0.1 wt%, a melt supply of 9.6×1014 g/yr would be needed to expel enough N from the mantle to balance out the apparent net influx of 9.6×1011 g/yr (Busigny et al., 2011). This amount of melt is at the upper end of output rates for the range of worldwide volcanoes (Crisp, 1984; White et al., 2006), but exceeds the output from the carbonatite volcano Oldoinyo Lengai during active eruption by two orders of magnitude (Dawson et al., 1994). Carbonatitic volcanoes may be the most appropriate for comparing with diamond-forming melts, as they are both relatively low-degree melt products. The combined sense from examining kimberlite and carbonatite volcanic output rates is that low-degree mantle melts, as might be expected for diamond formation, cannot carry a N flux of 9.6×1011 g/yr that may escape and percolate to the surface. 77

The question of degassing of N from mantle melt can also be approached by considering the amount of mantle material required to yield 9.6×1011 g/yr of N. Assuming the carbonatitic melt has 10 wt% C (Dalton and Presnall, 1998) and inherits a mantle C/N ratio of ~535 (Marty and Zimmermann, 1999), it will contain 0.02 wt% N. To reach 9.6×1011 g/yr of N requires a melt supply of 4.8×1015 g/yr. If the melt has a density of 2.7 g/cm3 (at 6 GPa) (Dobson et al., 1996), it could be produced by 0.1 vol% partial melting of ~1800 km3 of mantle. This is an unreasonably large mantle volume and such a process is therefore considered unlikely. To put it in perspective, the annual production of 21 km3 of mid-ocean ridge basalt (Crisp, 1984) is generated by ~20 vol% partial melting of ~100 km3 of mantle. It is unlikely that sufficient carbonatitic mantle melt 11 could be generated to return 9.6×10 g/yr of N to surface via degassing of its N2. The magnitude of the N efflux from carbonatitic magmatism is not negligible. Using the same assumptions as above, a N efflux equivalent to the sum from intraplate volcanism, 5.7×107 g/yr (Sano et al., 2001) could be generated by 0.1 vol% melting of 0.1 km3 of mantle. Constraints on the actual N efflux from this activity may come from obtaining a better understanding of the production rate of low-degree mantle melts that have contributed to diamond-formation and other lithospheric metasomatism through time.

4.5.3.2 Nitrogen speciation and redox control The speciation of N in diamond-forming media is important for understanding how it is incorporated into the diamond lattice (Boyd et al., 1987; Iakoubovskii and Adriaenssens, 2002) and how N is liberated from the convecting mantle. In the inclusions, N is present as molecular

N2 and is associated with oxidized carbon, CO2, which has also been released from the convecting mantle. This association suggests redox conditions bear control over nitrogen + speciation and mobility. Nitrogen oxidation and liberation from NH4 -bearing mantle 2- (Watenphul et al., 2010) could mimic the C to CO3 "redox melting" mechanism proposed to

control the release of carbonate as a function of oxygen fugacity (fO2) (Rohrbach and Schmidt,

2011) (Figure 4-6). Carbonate melt production should occur near fO2 about 2 to 3 log units above the iron-wüstite buffer, as upwelling mantle approaches a depth of ~250 km (Rohrbach + and Schmidt, 2011). Similar fO2 conditions could trigger the oxidation of NH4 groups to N2 and

water (Watenphul et al., 2010), given that the transition from NH3 to N2 in H-N-O fluids also lies 2 to 3 log units above the iron-wüstite buffer (Watenphul et al., 2009).

Redox conditions would also govern N2 liberation if metal nitrides store mantle nitrogen

(Dobrzhinetskaya et al., 2009; Kaminsky and Wirth, 2011). Increasing fO2 with decreasing depth in the mantle, especially as the mantle crosses from buffering by Fe0↔Fe2+ to buffering by Fe2+↔Fe3+ (Figure 4-6), would destabilize any small amounts of metal (Rohrbach et al., 2007;

Rohrbach and Schmidt, 2011) that might contain nitrogen, and liberate it as N2. During diamond

growth in the lithosphere, carbonate or CO2 reduction to diamond (Stachel and Harris, 2009) is

not necessarily accompanied by N2 reduction to NH3, as the latter would require lower fO2 and depends on hydrogen fugacity (Andersen et al., 1995). We conclude that redox control over mantle nitrogen may be fundamental for its overall nitrogen budget.

-2 reduction to diamond mid-ocean ridge CO3 lithosphere leaves low C/N, excess N2 asthenosphere N2 15 δ15N ≈ -5‰ δ N ≈ -5‰ + stite N 4 ü 2 carbonate-bearing melt w NH – + dissolved N 2 isotopic CO -2 N fractionation? 3 2 15 15 δ NNH4+ > δ NN2

N ≈ 5‰ subduction15 of δ relative to iron to relative fO control

2 2 0 oxidation + O

f C NH 3+ 2+ 4 Fe ↔Fe 0 +1 +2 +3 ...

log ~250 km Δ + Fe0 ↔Fe2+ NH4 in silicates

Figure 4-6. Redox-controlled nitrogen escape, with carbonate production, from the convecting mantle. This + upper mantle section depicts the release of N2 from NH4 -bearing silicates, as a function of oxygen fugacity, in parallel with release of carbonate melt (Rohrbach and Schmidt, 2011), yielding a N2-bearing diamond- forming melt. When the carbonate is reduced to diamond in the cratonic lithosphere, the N2 becomes concentrated in the residuum.

4.5.3.3 Nitrogen speciation and isotopic fractionation The speciation of N in diamond-forming fluids and melts also controls how N isotopes might fractionate (Thomassot et al., 2007). Nitrogen isotopes in diamonds are thought to

replicate the isotopic signature of the parental source for diamond-forming media (Cartigny, 2005). Their generally negative δ15N values centering around -5‰ are taken to reflect a mantle source of the same value (Cartigny, 2005). The same principle is used in the interpretation of mid-ocean ridge basalts, and the value of -5‰ is taken to represent the convecting upper mantle (Marty and Zimmermann, 1999; Marty and Dauphas, 2003; Cartigny and Marty, 2013). This assumes there is no isotopic fractionation of N during the production of diamond-forming fluid and mid-ocean ridge basalts, based on the high temperatures involved and equilibrium thermodynamic considerations (Richet et al., 1977; Busigny and Bebout, 2013). The assumption that nitrogen isotopic fractionation is chiefly a low-temperature, often biogenic phenomenon, contradicts large δ15N variations reported for syngenetic minerals in a mantle xenolith (Yokochi et al., 2009). Furthermore, recent experiments suggest that non- equilibrium, kinetic effects are important for isotopic fractionation at high temperature. During

progressive NH3 decomposition, the N2 produced is isotopically lighter by ~17‰, evolving up to ~40‰, with no temperature dependence at 600–800 °C (Li et al., 2009). The isotopic

fractionation factor (αN2–NH3 ) calculated from these results was of 0.983 ± 0.002, which differs from theoretical predictions due to kinetic effects (Li et al., 2009). Therefore, isotopic + fractionation in the mantle during the oxidation of stored NH4 to N2 in partial melts should be + considered as a possibility. If N2 liberated by oxidation of NH4 has a similar sense and magnitude of isotopic fractionation as in the experiments of Li et al. (2009) noted above, then the δ15N value of -5‰ in diamonds does not represent the parental source of the diamond forming + media, provided NH4 oxidation does not locally go to completion. The mantle source would be 15 + isotopically heavier, with δ N>-5‰ (Figure 4-6). The oxidative liberation of stored NH4 to

mobile N2 may be accompanied by isotopic fractionation, which could help account for the observed isotopic mismatch between nitrogen in the surface and mantle reservoirs. Further experimental work should be conducted to constrain the role of kinetic isotopic fractionation during processes that liberate N from the mantle (Busigny and Bebout, 2013). Ultimately, isotopic fractionation could add flexibility into geodynamic models of deep N cycling models (Marty and Dauphas, 2003).

4.6 Summary and conclusions Nitrogen is generally considered a trace component in the Earth's mantle. Mantle volatiles sampled from mid-ocean ridge basalts or preserved as fluid inclusions in xenoliths are dominated

by CO2, with only traces of nitrogen. Here we describe CO2–N2 fluid inclusions with 40 mol%

N2 and silicate melt inclusions with ~0.1 wt% dissolved N2 in mantle-derived diamonds from 3 different cratons. The diamonds are octahedrally-grown, which is the most common growth habit and rarely reported to contain fluid inclusions. The inclusions have C/N ratios <1, more than two orders of magnitude below typical mantle values. We propose that nitrogen can become concentrated to high degrees by processes related to diamond growth. A growing diamond does not readily consume the nitrogen available, as evidenced by the high concentrations of N2 that

remain in inclusions. This suggests incompatible behaviour. Residual N2 from this process represents a concentrated nitrogen flux escaping the convecting mantle. Hidden nitrogen fluxes like this might be significant in counteracting the large apparent net influx by subduction. Based on nitrogen and carbon speciation in the inclusions, we propose that redox processes control the + liberation of nitrogen from the convecting mantle, by the oxidation of NH4 in silicates to mobile

N2, concurrent with a parallel oxidation process releasing carbonate melt from the asthenosphere. Isotopic fractionation accompanying changes in nitrogen speciation could help account for the apparent isotopic mismatch between surficial and mantle nitrogen.

5 Implications of metallic iron for diamonds and nitrogen in the sublithospheric mantle4

5.1 Introduction Most recent experiments on the Fe speciation in the mantle suggest that the mantle below ~250 km contains metallic Fe (Ballhaus, 1995; Frost et al., 2004; Rohrbach et al., 2007; Rohrbach et al., 2011). Here, we explore the implications of this metallic Fe for the behavior of mantle N based on new empirical and experimental data on its distribution between high- pressure mantle phases. Specifically, we propose an explanation for the origin of the low N concentrations that typify diamonds from the sublithospheric mantle. The uppermost ~200 km of the mantle, sampled by xenoliths, massifs, and mantle melts,

has a moderately oxidized character, with oxygen fugacities (fO2) lying within a few log units of

the fayalite-magnetite-quartz buffer (FMQ; O2 + 3Fe2SiO4 ↔ 2Fe3O4 + 3SiO2)(Frost and

McCammon, 2008; Stagno et al., 2013). Garnet peridotite xenoliths show decreasing fO2 with depth, down to 3 log units below FMQ at 6 GPa in the deepest lithosphere of the Slave (Canada) and Kaapvaal (South Africa) cratons (Stagno et al., 2013). Theoretical calculations and

experiments show that at ~250 km (8±1 GPa), the fO2 reaches the iron-wüstite buffer (IW; O2 + 2Fe ↔ 2FeO), where (Fe, Ni) metal becomes stable (Ballhaus, 1995; Frost et al., 2004; Rohrbach et al., 2007; Rohrbach et al., 2011) (Figure 5-1). At this depth, majoritic garnet and subcalcic pyroxene incorporate appreciable Fe3+, to the extent that Fe2+ will disproportionate into Fe3+ and Fe0 (Rohrbach et al., 2011). As the Fe3+ is taken up by silicate minerals, the complementary Fe0 forms a stable metal phase, increasing its mode from the ~250 km saturation depth to ~0.5 wt% at 660 km, i.e. at the base of the transition zone (Rohrbach et al., 2011; Rohrbach and Schmidt, 2011). Below 660 km, MgSi-perovskite incorporates Fe3+, permitting a Fe0 content of ~1 wt% throughout the lower mantle (Figure 5-1) (Frost et al., 2004). No rock samples from below ~200

4 This chapter has been accepted for publication in the Canadian Journal of Earth Sciences as of March 2014. Smith, E.M., and Kopylova, M.G., Implications of metallic iron for diamonds and nitrogen in the sublithospheric mantle. Additional, unpublished data has been provided in appendices. 82

km are available as xenoliths, but metal-saturation of the deeper mantle may be confirmed by diamonds. Diamonds derived from the sublithospheric mantle, including the lower mantle, sometimes contain inclusions of native Fe, Ni, Fe-Ni alloys (Figure 5-2), and Fe carbides (Hayman et al., 2005; Bulanova et al., 2010; Gurney et al., 2010; Kaminsky and Wirth, 2011). These inclusions testify to the presence of Fe0 in the sublithospheric mantle.

∆log fO v Opx 2 LITHOSPHERIC Depth relative 2+ 3+ Fe ↔Fe DIAMONDS (km) to IW buffering [N] = 235 ppm mantle HP avg Non-con Garnet Cpx 200 ~250 km Olivine Cpx Proportion of Fe0↔Fe2+ S Type II diamonds buffering D N e Type I “Asthenosphere”

MO 400 tl A n I a Wadsleyite D m

C Type II I Majorite g n R i t E [N]avg = 56 ppm

Ringwoodite zone ransition 600 ec PH T v S n o O C H

MgSi-perovskite IT L

Metallic Fe B Mineral 800 SU Fe-periclase (wt. %) proportions (wt.%) [N] = 17 ppm mantle Lower 20 40 60 80 CaSi-perovskite -1.5 0 +3 0.1 0.5 1 avg

Figure 5-1. Schematic mantle section, summarizing mineralogy, fO2 characteristics, and diamond nitrogen content. Mineralogy and phase changes are based on peridotitic "pyrolite" mantle composition (Irifune and Tsuchiya, 2007). Relative fO2, compared to the IW buffer, changes character with depth beyond ~250 km as the mantle becomes saturated in metallic Fe (Rohrbach and Schmidt, 2011; Stagno et al., 2013). The metal saturation depth and base of cratonic lithosphere are dashed, as these are not fixed depths. Sublithospheric diamonds, grown in this metal-saturated environment, have low average nitrogen contents ([N]avg) and high proportions of Type II diamonds. The increase in metallic Fe up to ~1 wt% in the lower mantle is further correlated with lower [N]avg and a greater proportion of Type II's. Diamond characteristics for lithospheric (17% Type II, n=917), asthenospheric/transition zone (63% Type II, n=65), and lower mantle (81% Type II, n=155) diamonds are calculated from the same data sources as Figure 5-3.

10 μm

Figure 5-2. Electron backscatter image of metallic Fe-Ni blebs (bright) coating the periphery of a ferropericlase ((Mg,Fe)O) inclusion, liberated from a lower mantle diamond of the Rio Soriso river deposits, Brazil (after Hayman et al., 2005).

If the mantle below ~250 km does have a stable Fe0 component, this has significant implications for the behavior of mantle N. Partitioning of N into ambient metallic phases could explain why diamonds from the sublithospheric mantle have low N contents. Moreover, given the marked solubility of N in metallic Fe, especially compared to other mantle minerals, (Fe,Ni) metal could be a major mantle N reservoir, even though it may constitute no more than ~1 wt% of the lower mantle.

5.2 Siderophile character of N The presence of metallic Fe in the mantle will affect the behaviour of N, because N is moderately siderophile. Owing to its small atomic size, N can occupy the interstitial lattice sites in metallic Fe, which allows it to dissolve in molten and crystalline iron as well as form Fe- nitride compounds (Ringwood and Hibberson, 1991). The solubility of N in metallic Fe is up to several orders of magnitude higher than in silicate melt, depending on pressure, temperature, and

fO2 (Hashizume et al., 1997; Miyazaki et al., 2004; Roskosz et al., 2013). In experiments with

silicate melt of simplified CI chondrite composition and (Fe, Ni) metal at 2200–2700 °C, and fO2 near IW, N solubility varies from a few ppm to a plateau of 0.6–0.8 wt% in the silicate, and 0.035 wt% to 18 wt% in the metal, as pressure increases from atmospheric up to 18 GPa (a depth of ~530 km) (Roskosz et al., 2013). At pressures above 1 GPa, the partition coefficient between the metal and silicate melts is 18±10, which has been shown to be capable of drawing considerable N into the Earth’s metallic core during segregation of the early Earth (Marty, 2012;

Roskosz et al., 2013). This siderophile behaviour for N is also supported from first principles molecular dynamic modelling (Zhang and Yin, 2012). During diamond growth, the same tendency of N to partition into metallic Fe phases is evident. Synthetic diamonds grown in (Fe,Ni) metal contain less N than those synthesized in silicate–carbonate melts, due to stronger partitioning of N into the metal. Experiments to grow diamond by reactions between metallic Fe and carbonate sealed in a single vessel have shown this tendency clearly. Palyanov et al. (2013) conducted experiments to simulate the redox interaction between mantle with metallic Fe and subducted, carbonate-bearing crust.

Experimental runs used a Fe or Fe3C pellet embedded in (Mg,Ca) carbonate, within a platinum capsule. Run products revealed diamond growth both in the metal-rich and carbonate-rich regions, on either side of the redox front. Notably, the diamonds in contact with metallic Fe contained only 100–200 ppm N, while those in the carbonate melt portion contain 1000–1500 ppm N (Palyanov et al., 2013). Other synthetic diamonds grown either in (Fe,Ni) alloy or silicate–carbonate melts show similar results (Pal'yanov et al., 2002a; Spivak et al., 2008; Babich

et al., 2012). This implies that the diamond-melt distribution coefficient (DN = mole fraction N in diamond / mole fraction N in growth medium) is lower for metallic Fe than for silicate or carbonate melts. Partitioning of N between synthetic diamond and metallic Fe can be constrained from experiments, which typically run at pressures of 5–7 GPa. It should be noted that absolute N concentration of synthetic diamonds should not be expected to coincide with natural diamonds, because it depends on the bulk N content of the starting materials. Synthetic diamonds are often grown in Fe or (Fe,Ni) metal, to which the addition of N-bearing phases has allowed some experiments to produce diamonds with higher N concentrations. In one set of experiments, up to

0.5 wt% Ba(N3)2 was added to the starting materials to synthesize high-N diamonds, resulting in diamonds containing up to 1600–2400 ppm N (Yu et al., 2008). Even in experiments with high nickel contents in the metal (Ni0.7Fe0.3), which will significantly reduce its N solubility (Simmons, 1996), the N concentration of the metal remains higher than the synthesized diamonds (Palyanov et al., 2010). A more extreme experiment, with a catalyst consisting entirely

of Fe3N, which effectively contains 250,000 ppm N, yielded diamonds with 1350–3300 ppm N

(Borzdov et al., 2002). These experiments in Fe3N imply DN values of 0.005–0.013. Relative 85

incompatibility of N in synthetic diamonds compared to metallic Fe (i.e. DN<1) is further illustrated by the fact that slowing the diamond growth rate, which allows the N partitioning to approximate equilibrium more closely, yields diamonds with lower N concentration (Spivak et al., 2008). Relative incompatibility of N is also supported by the fact that the morphologically dominant diamond faces have higher N concentrations (Burns et al., 1999). Although successive growth layers accumulate slowly on dominant crystal faces, growth steps on dominant faces advance more rapidly than they do on subordinate faces, so dominant faces have less opportunity to reject incompatible impurities during growth (Sunagawa, 1984). Natural diamonds also provide evidence of N partitioning into Fe0 phases. For example, a lower mantle diamond from Juina, Brazil was found to contain inclusions of Fe-carbides (Fe0 with interstitial C0) with appreciable N contents. The N concentration was 73,000–91,000 ppm in some inclusions, while the host diamond contained merely 44 ppm N (Kaminsky and Wirth, 0 2011). These values give a very low DN of 0.0005 between the diamond and the Fe phase. Higher pressure in the lower mantle compared to laboratory diamond synthesis will achieve

higher N solubility in the metal and lead to lower DN values (Roskosz et al., 2013). The N contents of other reported metallic Fe inclusions in sublithospheric diamonds have not been analyzed. During routine analysis with X-ray spectroscopy in an electron microscope or electron microprobe, N is among the light elements that are generally not analyzed, so it is plausible that high N has gone unnoticed in other studied metallic Fe or Fe-carbide inclusions.

5.3 Consequences of metallic Fe in the sublithospheric mantle

5.3.1 Low N content of sublithospheric diamonds If the mantle below ~250 km is saturated in metallic Fe, and N tends to partition into this metal, a difference in N content between lithospheric and sublithospheric diamonds should be expected. Such a contrast is indeed observed and has long been an unexplained curiosity (Figure 5-3). Lithospheric diamonds, which comprise the majority of mined diamonds, grow in the subcontinental lithospheric mantle of old, thick continental regions (cratons) at depths of 140– 200 km (Boyd and Gurney, 1986; Meyer, 1987). The top of the depth interval corresponds to the graphite–diamond transition, while the bottom corresponds to the lower boundary of cratonic lithosphere. The depth of origin is inferred from studies of diamond inclusions and 86

diamondiferous xenoliths. Lithospheric diamonds contain garnet, clinopyroxene, orthopyroxene, olivine, Mg-chromite, rutile, coesite, as well as sulfides, and these correspond to peridotitic and eclogitic host rocks (Stachel and Harris, 2008).

a) Lithospheric, Peridotidic 250 Type II

y 200 c n e 150 qu e r

F 100

b) Lithospheric, Eclogitic 60

Type II y c

n 40 e qu e r F 20

c) Sublithospheric 200

y 150 c n e Type II

qu 100 e r F 50

0 0 250 500 750 1000 1250 1500 N concentration (atomic ppm) Figure 5-3. Nitrogen concentration distributions for inclusion-bearing diamonds. Peridotitic (a) and eclogitic (b) lithospheric diamonds have more N than sublithospheric (c) diamonds. Data sources: peridotitic and eclogitic (Stachel and Harris, 2009); sublithospheric (Deines et al., 1991; Stachel and Harris, 1997b; Davies et al., 1999a; Hutchison et al., 1999; McDade and Harris, 1999; Stachel et al., 2002; Hayman et al., 2005; Tappert et al., 2005a; Tappert et al., 2005b; Tappert et al., 2009; Bulanova et al., 2010; Palot et al., 2012). Sublithospheric data compilation given in Appendix J: Nitrogen in sublithospheric diamond.

In contrast, sublithospheric diamonds are distinguished by higher-pressure mineral inclusions, derived from greater depths. These diamonds make up a small and variable proportion (~1%) of mine production (Gurney et al., 2010). Below 250 km, the first higher- pressure mineral inclusion is majoritic garnet which becomes increasingly stable through the deeper mantle to 410 km depth (Figure 5-1). Diamonds hosting majorite are referred to as “asthenospheric” (Stachel et al., 2005). At the transition zone (410–660 km), first wadsleyite, then ringwoodite are formed, which are high-pressure polymorphs of olivine, (Mg, Fe)2SiO4. Although they are destabilized and reverted upon decompression, wadsleyite and ringwoodite inclusions can still be recognized by geochemical means and are interpreted as sourced from the transition zone (Kaminsky, 2012). Below 660 km, in the lower mantle, ferropericlase, MgSi- perovskite, and CaSi-perovskite are the principal inclusions phases (Kaminsky, 2012). Metallic Fe inclusions have also been reported, mainly in sublithospheric diamonds (Gurney et al., 2010), although limited occurrences of metallic Fe inclusions in lithospheric diamonds are also known (Bulanova, 1995; Davies et al., 1999b). A more detailed overview of sublithospheric diamond inclusions is given by Kaminsky (2012). Sublithospheric diamonds, including those formed in the “asthenosphere”, transition zone, and lower mantle, are markedly depleted in N compared to those from the lithosphere (Davies et al., 1999a; Hutchison et al., 1999; Kaminsky et al., 2001; McCammon, 2001; Stachel et al., 2002; Stachel et al., 2005; Bulanova et al., 2010; Palot et al., 2012). Substitutional N is the most common impurity in diamond and is routinely analyzed with infrared spectroscopy. Figure 5-3 shows the N content of diamonds with mineral inclusions, used to identify host rock paragenesis. On average, the N content of lithospheric diamonds is 150–200 ppm for peridotitic and 300–400 ppm for eclogitic diamonds (Cartigny, 2005; Stachel and Harris, 2009). Taken together, the lithospheric diamonds in Figure 5-3 average 253 ppm (90% are within 0–700 ppm). Type II diamonds, those with N contents below the detection limit of infrared spectroscopy (~10 ppm), make up ~24% of peridotitic diamonds and ~10% of eclogitic diamonds (Stachel and Harris, 2009; Gurney et al., 2010). Not shown in Figure 5-3 are fibrous diamond coats/cuboids. This less common, dendritic growth habit of lithospheric diamond, typically has N contents within 600– 1400 ppm (Cartigny, 2005).

In sharp contrast to lithospheric diamonds, the average for sublithospheric diamonds lies at 33 ppm (90% are within 0–100 ppm) for the data compiled in Figure 5-3 (Appendix J: Nitrogen in sublithospheric diamond). Moreover, sublithospheric diamonds are dominated by Type II's, which make up 73% of the 243 samples in Figure 5-3. Subdivision of sublithospheric diamonds into lower mantle and asthenospheric/transition zone further resolves this pattern. Figure 5-1 shows that lower mantle diamonds contain less nitrogen (81% Type II) than those from the asthenosphere/transition zone (63% Type II) (Deines et al., 1991; Hutchison et al., 1999; Hayman et al., 2005; Palot et al., 2012). It is also worth noting the crystal lattice configuration of N in sublithospheric diamonds tends to be highly aggregated to B centres (McCammon, 2001; Hayman et al., 2005; Tappert et al., 2009). The subset of sublithospheric diamonds in Figure 5-3 with detectable N are, on average, 90% aggregated to B centres. The high aggregation state is consistent with the high temperatures (Taylor et al., 1990) of the sublithospheric mantle. In comparison, lithospheric diamonds have a range of aggregation states, averaging about 40% B centres, and only 5% of these reach beyond the 90% B mark (Stachel, 2007). The proposed presence of metallic Fe throughout most of the sublithospheric mantle and its demonstrated ability to scavenge N presents a straightforward solution for the low N content of sublithospheric diamonds. Figure 5-1 illustrates this inverse correlation between the metal content of the ambient mantle and the N content of diamonds. We propose that metallic Fe acts to trap N during diamond growth. N partitions into the Fe-carbide or metallic Fe phase and little is taken up in the diamond lattice. This partitioning is demonstrated not only in synthetic diamonds (Borzdov et al., 2002; Yu et al., 2008; Palyanov et al., 2010; Babich et al., 2012;

Palyanov et al., 2013), but is also manifested directly in the lower mantle diamond with a DN of 0.0005 between the diamond and the included Fe-carbonitride phase described by Kaminsky and Wirth (2011). Thus the characteristically low N content of sublithospheric diamonds may well be a consequence of diamond growth in a mantle setting with metallic Fe, in a system buffered by Fe0↔Fe2+. Conversely, lithospheric diamonds grow in the absence of metallic Fe, in host rocks whose oxygen fugacity is generally governed by Fe2+↔Fe3+ buffering. This model also explains the slightly lower N content in lower mantle diamonds, sourced from below 660 km, compared to “asthenospheric” and transition zone diamonds (Deines et al., 1991; Hutchison et al., 1999; Hayman et al., 2005; Palot et al., 2012). The 660 km depth 89

corresponds to the onset of ferropericlase stability and an increase in metallic Fe content (Rohrbach and Schmidt, 2011). This increase in metal content is responsible for a boost in the capacity of the ambient mantle to trap N during diamond formation (Figure 5-1). Occasionally, metallic Fe may also occur in the lithosphere. Native (Fe, Ni) inclusions have rarely been found in the centre of lithospheric diamonds (Bulanova, 1995; Davies et al., 1999b). Some of these diamonds may also exhibit low N contents, like the N-free polycrystalline diamond aggregate with garnet, Fe-carbide, (Fe, Ni) metal, and FeS inclusions reported by Jacob et al. (2004). However, these occurrences are seen as small, anomalous perturbations in the

overarching pattern of Fe oxidation state, fO2 buffering, and diamond formation described above (Figure 5-1).

5.3.2 Model of diamond growth in the sublithospheric mantle Rohrbach and Schmidt (2011) outlined a model of redox melting and freezing in the deep mantle based on the response of C to the imposed oxygen fugacity. In the presence of Fe0, mobile carbonate melts are reduced and immobilized. The carbon is initially incorporated into the metallic phase, making Fe-carbides, but converted to diamond when additional C can no longer be accommodated by the Fe0 at hand (Rohrbach and Schmidt, 2011). Diamond growth in this metal-saturated system occurs by the reduction of carbonates, imposed by the ambient mantle oxygen fugacity that is buffered by Fe0↔Fe2+ reactions (Palyanov et al., 2013). Diamond growth will cease when either all the available carbonate is reduced or the local Fe0 is exhausted. If the Fe0 is exhausted, Fe should be present as Fe2+ and Fe3+ in silicates and ferropericlase, and C will be present as diamond (Rohrbach and Schmidt, 2011; Palyanov et al., 2013). Any additional carbonate entering this mantle region will not be reduced. Instead, it will remain mobile and may percolate upward to be reduced/immobilized upon meeting Fe0 (Rohrbach and Schmidt, 2011). Metallic Fe, plus Fe-carbide, should always be present during the active growth of sublithospheric diamond. When carbonatitic diamond-forming melt with traces of N percolates through the metal- saturated sublithospheric mantle, N will partition into the metallic Fe or Fe-carbide phases. The effectiveness of this process is ensured by the low melt/solid ratio and the restriction of both metallic Fe (Rohrbach et al., 2011) and the melt to interstices and grain boundaries. If the

diamond-forming melt is equilibrated with an equal mass of Fe0, the N concentration of the

diamond forming melt may be reduced by a factor of 18±10, using the experimental DN for metallic Fe and silicate melt (Roskosz et al., 2013). A reduction in N concentration of this magnitude for the diamond-forming melt agrees with the factor of 8 decrease in average N concentration of diamonds from in sublithospheric diamonds compared to lithospheric diamonds (Figure 5-3). As long as there is Fe0 available to reduce carbonate to diamond, there are Fe0-bearing phases to take up N preferentially. As the Fe0 is consumed, we should expect decreasing ability of metallic Fe or Fe-carbide to uptake N, due to a combination of N-saturation and oxidative consumption of these phases. This expected behaviour is consistent with the actual N zonation recorded in successive growth layers of diamond. Sublithospheric diamonds tend to show higher N concentrations in the rim of the crystal than the core, opposite to the typical trend exhibited by lithospheric diamonds (Davies et al., 1999a; Bulanova et al., 2010). In addition to the effect of partitioning of N into metallic Fe, the incorporation of N into diamond will also depend on diamond growth rate as well as other factors affecting the N content of the melt, such as the crystallization of other minerals (Cartigny et al., 2001; Babich et al., 2012). If the metallic Fe content of the sublithospheric mantle provides favourable conditions for the growth of N-free diamonds, a sublithospheric origin may tentatively be assigned to those Type II (N-free) diamonds that do not contain mineral inclusions. Particularly, there is a special subset of Type II diamonds that are often large, inclusion-free, highly-resorbed, generally anhedral, and of high-quality, whose paragenesis remains enigmatic (Gurney and Helmstaedt, 2012). Some of the largest and most valuable diamonds, like the famous Cullinan diamond, belong to this variety (Janse, 1995). Often, they can be visually identified with confidence based on their unique characteristics (Bowen et al., 2009). The generally large size of these Type II diamonds is demonstrated for the Letšeng mine, where they increase in proportion from 18% of 2 carat (0.4 g) diamonds to 68% of diamonds larger than 10.8 carats (2.16 g)(Bowen et al., 2009). Characteristics of these unique diamonds may be consistent with growth in the sublithospheric mantle, although they differ somewhat from Type II diamonds with sublithospheric inclusions (Moore, 2009). Nevertheless, the redox buffering capacity by Fe0↔Fe2+ in metal-saturated mantle is large (Rohrbach and Schmidt, 2011), providing a 91

favourable environment to reduce a large amount of carbonate in one location and grow large diamonds. For example, 1 kg of mantle with 1 wt% Fe0 could potentially reduce ~1 g of C to diamond (Rohrbach and Schmidt, 2011). Admitting the affinity of N for metallic Fe, the key to producing the unique, large, relatively inclusion-free Type II variety of diamond may be a setting buffered by Fe0↔Fe2+. If grown as described here, the 621.2 g Cullinan diamond (Janse, 1995) would require the reducing capacity of ~620 kg of mantle with 1 wt% Fe0. Considering a 3.5–5.5 kg/m3 mantle density range, the Cullinan would therefore require a redox reaction within 110–180 m3 of mantle. This demands reactant/product transport distances of at least 3.0–3.5 m. The relatively high temperatures afforded by the sublithospheric mantle may ensure high diffusion rates, to promote the growth of large crystals. High temperatures, combined with low diamond N concentrations, would also permit significant dislocation mobility (DeVries, 1975; Field, 1992a) to help account for the predominance of bulk plastic deformation features (Gurney and Helmstaedt, 2012) and polygonized dislocation networks (De Corte et al., 2006) in these unique Type II diamonds. Aside from their reflection of N behaviour, sublithospheric diamonds record ambient mantle assemblages as well as local diamond-forming processes, that is, modification of the mantle by the redox reactions. Localized Fe to FeO reactions accompanying diamond growth may help explain the prevalence of ferropericlase and its variable, Fe-rich composition among lower mantle diamond inclusions (Palyanov et al., 2013), like those described by Hayman et al. (2005) and Kaminsky (2012), compared to peridotitic “pyrolite” models for lower mantle assemblages.

5.3.3 Metallic Fe as a host of mantle nitrogen The proposed ~1 wt% metallic Fe content in the mantle and its ability to incorporate N has wide implications for the overall N budget of the mantle. Two longstanding problems surrounding mantle N are the nature of the phase(s) that host N and the reason for the high N/36Ar ratio of the mantle compared to the atmosphere. The first problem has been approached + from the perspective of an oxidized upper mantle, where N is expected to occur as NH4 substituting for K in silicates (Marty and Dauphas, 2003). However, it is not clear what minerals

should host the bulk of mantle N and there are deficiencies in attempting to account for mantle N + as being hosted in NH4 -bearing silicates alone. For example, clinopyroxene has been shown + experimentally to be capable of hosting up to 500–1000 ppm NH4 in the upper mantle, but this provides a capacity for merely 1012 mol of N (Goldblatt et al., 2009; Watenphul et al., 2010). This figure is dwarfed by the estimated 3.6±1.8×1020 mol of N residing in the bulk mantle (Marty, 2012). Other experiments have demonstrated modest N solubility in silicate minerals at 3.5 GPa, with up to 10 ppm in forsteritic olivine and up to 100 ppm in enstatite (Li et al., 2013). The dominant lower mantle mineral phases, MgSi-perovskite, ferropericlase, and CaSi- perovskite, present no obvious N storage solutions. However, given the known extreme N solubility in (Fe,Ni) metal, up to ~12 wt% for Fe-rich (Fe,Ni) alloys (Roskosz et al., 2013), the metallic Fe content of the mantle may serve as a significant host for mantle N. Experimental data are not available to allow quantification of the distribution of N between the metallic Fe and other mantle minerals at sublithospheric conditions. However, if N in the dominant mantle minerals remains incompatible during partial melting, and we know the metal- silicate melt distribution coefficient to be 18±10 (Roskosz et al., 2013), the distribution coefficient between metallic Fe and other mantle minerals will be >>18±10. Considering a bulk mantle N concentration of 1 ppm (Marty, 2012), a ~1 wt% Fe0 content in the mantle could readily dissolve the entire N budget of the mantle (Frost et al., 2004; Rohrbach et al., 2007; Roskosz et al., 2013). If significant amounts of N are dissolved in metallic Fe phases, and we observe N escaping

the mantle as N2 gas (Chapter 4; Marty, 1995; Smith et al., 2014), it is implied that the liberation of N from upwelling mantle is controlled by oxygen fugacity and the destabilization of Fe0. Metallic Fe not only provides a storage solution for mantle N, it also lends insight into the notably higher N/36Ar ratio of the mantle compared to the atmosphere. The difference is often

explained in the framework of N recycling by subduction. Current models contend that both N2 and 36Ar were largely degassed from the early mantle (Marty, 1995; Miyazaki et al., 2004), and + + that the present mantle N is of recycled origin, subducted as NH4 in micas and stored as NH4 substituting for K+ in high-pressure silicate minerals (Marty and Dauphas, 2003; Miyazaki et al., 2004; Goldblatt et al., 2009; Watenphul et al., 2009; Marty, 2012).

The recognition of Fe0 in the mantle prompts a reconsideration of this popular view. Compared with a purely silicate system, N and Ar will behave differently in the presence of Fe0 because N is siderophile (Hashizume et al., 1997; Bouhifd et al., 2010) and Ar is not (Matsuda et al., 1993). As an alternative, the fractionation of N/36Ar could be have been accomplished by retention of N in metallic Fe in the mantle early in Earth history, while 36Ar was degassed to the atmosphere. Therefore Fe0 in the mantle provides a facility to retain N so that extensive N recycling is not required. An early establishment of the N budget of the mantle, as opposed to accumulating recycled N over time, is supported by several lines of evidence. Firstly, N isotopes incorporated in old lithospheric, cratonic diamonds match present-day upper mantle values (Cartigny, 2005), indicating that this reservoir was established prior to 3.5–3.2 Ga (Gurney et al., 2010). Secondly, the early establishment of mantle N would be in agreement with the 4.0 Ga 40K–40Ar age of the mantle (Halliday, 2013). These elements are linked because N correlates with 40Ar, which is a radioactive decay product of 40K. Specifically, N/40Ar ratios between mid-ocean ridge basalt, ocean island basalt mantle samples and the atmosphere are similar and reflect comparable long- term N/K in the mantle (Marty and Dauphas, 2003). Thus, if bulk mantle 40K and 40Ar contents are 4.0 Ga old, an early establishment of mantle N rather than buildup over geologic time via + subduction of NH4 (and K) is preferred. If we accept that N was retained in the mantle in metallic Fe, this may provide a convenient explanation for the low 15N/14N that characterizes the upper mantle (δ15N ≈ -5‰, 15 15 14 15 14 where δ N = [( N/ N)sample /( N/ N)air-1]×1000)(Marty and Dauphas, 2003) in comparison to the atmosphere. N taken up by the metal should be isotopically light and thus induce isotopic fractionation (Dauphas and Marty, 1999). The retention of isotopically light N in metallic Fe eliminates the need to explain this signature with extensive, early subduction of uniquely-low δ15N, biogenic, Archean N (Marty and Dauphas, 2003), which is not well supported by evidence (Cartigny and Ader, 2003; Kerrich and Jia, 2004).

5.4 Summary and conclusions Recently published experimental data suggest that the Earth's mantle below ~250 km is saturated in metallic Fe, reaching a concentration of ~1 wt% Fe0 throughout the lower mantle

and buffering oxygen fugacity by Fe0↔Fe2+ reactions. Metallic Fe in the mantle bears important implications for N, which behaves as a moderately siderophile element. Here, we propose that the trapping of N in (Fe, Ni) metal and Fe-carbides during diamond growth may account for the characteristically low N content of diamonds from the sublithospheric mantle. This model may also explain the origin of especially valuable, large, anhedral, flawless Type II (N-free) diamonds, like the Cullinan, that comprise a minor part of world diamond production. Partition

coefficients of N between diamond and metallic Fe within DN = 0.0005–0.013 have been demonstrated in high-P experiments for diamond synthesis and in a natural sample of a N-poor lower mantle diamond with inclusions of Fe-nitrocarbide. More N is incorporated into diamond if it grows in the lithosphere, where there is no ambient Fe0 to trap N. As a broader implication, the recognition of Fe0 in the mantle and its affinity for N suggests metallic Fe should be a major host of mantle N. Retention of primordial mantle N in metallic Fe could explain the high N/36Ar and low 15N/14N ratios of the mantle compared to the atmosphere.

6 Concluding discussion

6.1 Volatiles involved in fibrous and octahedral diamond formation Water is the dominant volatile phase in the fibrous diamonds investigated here, as well as those reported in the literature. Peak heights of the OH-stretching modes in FTIR spectra show the range of water content in the fibrous diamond fluids is typically 10–25 wt%, or 50–500 ppm, in terms of the bulk diamond (Weiss et al., 2010; Weiss et al., 2011). A high water content is indirectly supported by findings of hydrous carbonates and hydrous silicates (Kopylova et al., 2010; Smith et al., 2011a) as well as hydrous phosphates (Kopylova et al., 2010) in daughter minerals crystallized from the fluid. Carbon in the fluid is almost completely in the form of carbonate, with only minor amounts of molecular CO2 (Navon et al., 1988; Zedgenizov et al.,

2009). The N2 content of these fluids is unknown, because the techniques normally used to study

fibrous diamond fluids are insensitive to N2. The Neoarchean Wawa samples demonstrate that the major and trace element compositions, and volatile contents, of fibrous diamond fluids are not only consistent across many cratons worldwide, but are also consistent through time to as far back as 2.70 Ga. These hydrous fluids contrast sharply with octahedrally-grown diamonds. Despite the scarcity of reports on fluids in octahedrally-grown diamond, samples were recovered that represent three different cratons. Water was not identified among the volatile components of these diamonds. Instead, the inclusions are unexpectedly rich in N2. This is only the second time

mantle volatile compositions reaching 100 mol% N2 have ever been reported for any mantle sample, with the first being in xenoliths from the Canary Islands (Andersen et al., 1995). The

remainder of the volatile budget is CO2, which makes up <9 mol% in the Congo and Kaapvaal craton diamond samples and 60±4 mol% in the Siberian diamonds. In addition to volatiles, the octahedrally-grown diamonds from the Congo and Kaapvaal cratons contain former silicate melt, making the overall composition of these fluid inclusions different from fibrous diamonds. This finding alone is significant. The composition of the silicate melt inclusions in the Democratic Republic of Congo sample described in Chapter 4 have unusually low silica contents and high Mg numbers for a melt. The bulk melt composition is ultramafic. Such compositions are usually found in depleted, 96

refractory restites, and it is challenging to explain how a melt could be close in composition to olivine. Two ideas were briefly mentioned in Chapter 4. The first, and most straightforward way to achieve this composition, would be by melting dunite. This would require very high temperatures, within 1560–1890 °C (Blatt and Tracy, 1996). The most energetically feasible way to do this would be to have melting occur in a relatively small, localized region, which would have to be done quickly to maximize retention of heat and avoid incorporating silica by causing low-degree melting in neighbouring rocks. A second, perhaps more realistic method to achieve the ultramafic melt composition would be to begin with a Mg-rich carbonate-bearing melt with higher silica content than what is currently preserved in the inclusion. Reduction of the carbonate to diamond would drive up the ratio of (Mg, Fe)O to SiO2 in the melt, thereby decreasing its relative silica content. This reaction is analogous to the EMOD buffer (enstatite + magnesite ↔ olivine + diamond) (Wyllie et al., 1983; Luth, 1993). Decreasing the relative silica content to make an ultramafic melt would also shift the liquidus and solidus to higher temperatures. Although the temperature and pressure of the melt have not changed, it will then find itself below the solidus. However, crystallization is not instantaneous and the melt may become metastable. Decreasing the silica content makes a metastable, "supercooled" melt by chemical means rather than by cooling. This metastable ultramafic melt was then trapped as inclusions in the diamond. The melt was trapped relatively quickly, before having a chance to crystallize. A lack of visible crystal grains in the former melt supports this scenario by demonstrating the melt had a low density of nucleation sites. This is similar to what happens in komatiites, where melt is allowed to become undercooled due to the scarcity of nucleation sites (Fowler et al., 2002).

To summarize, contrasting concentrations of H2O, N2, and silicate melt in octahedrally- grown and fibrous diamond suggests that different types of fluids equilibrated with these two different types of diamond. As a consequence, the saline-carbonatitic-silicic ternary diagram commonly used for fibrous diamond fluids (Figure 1-5) does not encompass all diamond-related fluids (or melts). Furthermore, the disparity in water content between fibrous and octahedrally- grown diamond may indicate that water is the governing factor for triggering fibrous diamond growth as opposed to octahedral growth (Figure 1-3).

6.2 Water as a trigger for fibrous diamond habit The controls over diamond habit that cause fibrous diamond growth as opposed to smooth, layer-by-layer octahedral growth remain an open question. The presence of hydrous fluid in

fibrous diamonds and CO2–N2 fluid dissolved in silicate melt in octahedrally-grown diamonds suggests that water content among the volatile load of diamond-forming media may be a controlling factor for diamond habit. Elevated water activity has been shown to cause {111} dendritic growth of synthetic diamond (Kanda, 1985; Palyanov et al., 2012). These experiments are compelling, even though they are conducted in metal catalysts and therefore not directly applicable to nature. During diamond synthesis experiments, Palyanov et al. (2012) found that increases in water content from 0 to 0.2 wt% caused diamond morphology transitions from octahedra, to platelike (oblate) octahedra, to {111} dendrites. Further increases in water content up to 0.4 wt% result in rhombic dodecahedra and {110} dendrites. With the addition of even more water, diamond nucleation and growth cease and graphite is produced. Palyanov et al. (2012) offered the explanation that hydroxyl (C–OH) and carboxyl (HO–C=O) groups act as impurities on diamond faces. To expand on this, experiments on wet chemical oxidation of diamond surfaces, show the density of covalently-bonded carboxyl (HO–C=O) groups varies significantly with crystallographic orientation (Wang et al., 2012). The density of these functional groups is lowest for {111} surfaces and higher for {100} surfaces (Wang et al., 2012). During the growth of a smooth octahedral {111} face, the advancing growth steps, which advance perpendicular to {111}, may become saturated with water-related groups so that the addition of new diamond to the step is inhibited. The modified diamond surface inhibits growth within {111} surfaces, but still allows heterogeneous nucleation on {111} surfaces. Figure 6-1 reveals that this effect can be observed in the dendritic diamond from Palyanov et al. (2012). The limited development of {111} faces is similar to diamond resorption studies that show water deteriorates octahedral, {111} faces by attacking their edges (Fedortchouk et al., 2007). It is therefore not surprising that water activity in the growth medium may influence diamond habit.

<111> growth and branching water inhibits development of {111} face

≈{110} {111}

Figure 6-1. Synthetically grown dendritic diamond. The dentrites grow and branch along <111> vectors. Each dendrite would terminate in a {111} face if not for the effect of water on crystal growth. The closeup shows dendrites terminate with pseudo-{110} surfaces, reminiscent of resorbed, dodecahedroid morphology (Figure 1-1). Note the roughly hexagonal axial cross section of the dendrites. SEM image modified from Palyanov et al. (2012).

In addition to inhibiting advancement of growth steps within {111} faces, water may cause other effects during natural diamond growth. The addition of water is expected to increase carbon solubility, which would accelerate growth and nucleation kinetics (Davis et al., 2000; Sokol and Pal'yanov, 2008). Water could lower the crystal-fluid interfacial energy, alter desolvation mechanisms, and make the growth interface more diffuse. All these factors could promote dendritic growth (Sunagawa, 1984; De Yoreo and Vekilov, 2003). Dendrite propagation relies on lowering the energy barrier for nucleation relative to the energy barrier for growth. The energy barrier for nucleation is proportional to α3/σ2, where α is the interfacial energy and σ is supersaturation (De Yoreo and Vekilov, 2003). Interfacial energy expresses the energy difference between atoms inside the diamond and those at the interface. It is always positive and acts to destabilize nuclei. Lowering this term has a large effect toward decreasing the nucleation barrier and an even larger effect toward increasing the nucleation rate. Nucleation rate is proportional to e^(−Aα3/σ2), where A is a factor that depends on many parameters (De Yoreo and

Vekilov, 2003). This expression shows that nucleation rate increases drastically with decreasing interfacial energy. For diamond growth from an oxidized medium, the diamond surface should become oxygenated. An oxygenated diamond surface has been shown experimentally to have an increased polar component of its surface energy, making it more hydrophilic and increasing its wettability by hydrous media (Ostrovskaya et al., 2002). Diamond-forming media with more water will therefore be expected to have a superior ability to wet the diamond surface. This effect will lower the interfacial energy, thereby lowering the energy barrier for nucleation and boosting

the nucleation rate. A drier carbonate or CO2-bearing medium will be less effective at wetting the diamond surface during growth and a lower nucleation rate is expected. The combination of observed fluid inclusions and experimental/theoretical considerations suggests that fibrous diamond habit is the result of its growth from a more hydrous medium compared to octahedrally- grown diamond. During progressive diamond growth from the fluid phase, the oxygen fugacity may change within a locally closed environment. Such a progressive evolution of fO2 during diamond formation has been observed in experiments with carbonate and metallic Fe (Palyanov et al., 2013). When diamond crystallizes from carbonatitic fluid reacting with the reduced mantle

(Cartigny et al., 1998a; Frost and McCammon, 2008), fO2 in the fluid will drop according to the reducing effect of the host rocks. As oxygen fugacity drops, the interfacial energy will increase slightly as the diamond surface becomes less oxygenated and more hydrophobic. The redox reaction between the fluid and host rocks may also slow, decreasing the supersaturation, σ. Either effect would act to lower the nucleation rate, and may help explain why diamond fibres often thicken (Lang, 1974) and the frequency of fluid inclusions wanes (Navon et al., 1988; Kopylova et al., 2010) toward the periphery of fibrous diamond growth layers. A decreasing nucleation rate would also explain why N concentrations tend to decrease (Boyd et al., 1994), if N uptake is controlled kinetically (Cartigny et al., 2001).

6.3 Temporal evolution of water content in fluid Fibrous diamond commonly occurs as a coat over an octahedrally-grown core. The reverse situation, with a octahedral overgrowth on a fibrous diamond core, is much rarer. Interestingly,

100

growth habit does not change gradually or seem to oscillate back and forth between octahedral and fibrous. The evolution in habit is unidirectional, usually from octahedral to fibrous. If water content of diamond-forming fluid imposes control over growth habit, as argued in Section 6.2, then these characteristics could indicate a tendency for water activity in the growth medium to be either low or high, with limited variability between these modes. Low and high water activities would cause crystallization of octahedrally-grown and fibrous diamond, respectively, and intermediate activities would be repressed. The common occurrence of fibrous overgrowths would then indicate that late-stage diamond-forming fluids are generally more hydrous. The tendency for water activity to be either high or low could be explained by the supercritical or subcritical behaviour of diamond-forming fluids. As a supercritical fluid, a

generic carbonate-silicate-CO2-H2O medium would have an essentially unlimited capacity for water. For subcritical conditions, a carbonate-silicate melt fraction would be immiscible with a hydrous fluid phase (Figure 6-2) (Wyllie and Ryabchikov, 2000). Thus, the onset of immiscibility divides the carbonate and water into separate phases, causing a shift in the nature of the carbonate-bearing medium from a wet supercritical fluid to a drier carbonate-bearing melt. In contrast, the onset of miscibility upon entering supercritical conditions would cause a sudden increase in water bearing capacity of the carbonate-bearing melt. The distinction between supercritical and subcritical conditions depends on the position of the second critical end point (2CP), whose pressure and temperature conditions depend on the chemical composition of the system. Estimates in a lherzolite-CO2-H2O system, for example place the 2CP above 7.5 GPa (Wyllie and Ryabchikov, 2000), whereas in a basalt–H2O system the solidus terminates at a 2CP between 5 and 6 GPa (Kessel et al., 2005). For diamond growth in the subcontinental lithospheric mantle, within 4.5–6.5 GPa, deciding whether or not the diamond-forming medium is supercritical is not straightforward. It depends on bulk composition of the metasomatic growth medium, and not necessarily the host rock (c.f. Stachel and Harris, 2008), as the infiltrating metasomatic agent may not be in equilibrium with the host rock. For diamond growth from carbonatitic media, it is appropriate to consider the 2CP of carbonatite. In

the system CaO-SiO2-CO2-H2O, the 2CP lies at 3.25 GPa and 515 °C (Boettcher and Wyllie, 1969). This compositional system is not necessarily expected to be in equilibrium with mantle host rocks, which might be expected to be much more Mg-rich. It is, however, worth considering 101

because mantle metasomatic reactions like diamond formation are an expression of disequilibrium, as the system attempts to achieve equilibrium. Also, published experimental work on the 2CP of other carbonatitic melt compositions could not be found. Nonetheless, it is immediately clear the 2CP in carbonatitic systems is lower compared to peridotitic or eclogitic assemblages. This means that infiltrating carbonatitic media should themselves be supercritical, even though they may be below the 2CP of the host rock. The supercritical carbonate-bearing fluid will become subcritical if either 1) the fluid pressure or temperature decreases below its 2CP, which is unlikely for carbonatitic compositions with low 2CP's, or 2) if its composition evolves, for example by dissolving adjacent host rock or by crystallizing diamond, such that the 2CP of the fluid increases above ambient pressure and temperature. In other words, increasing the silica content of the supercritical fluid causes exsolution of a hydrous phase.

a b

melt melt + vapour e e r r

tu melt + tu

ra minerals ra minerals + melt + vapour empe empe ock ock

T S T r OLI D U S r ee ee r r -f -f r rock + vapour r apou apou v v

Solid components Volatile components Solid components Volatile components Bulk composition Bulk composition

Figure 6-2. Isobaric phase diagrams, for a generalized rock system (e.g. lherzolite-CO2-H2O). These diagrams show the distinction in phase behaviour for pressures below (a) and above (b) the second critical end point. For rocks in (a), heating results in melting at the solidus. For rocks in (b) pressure is beyond the second critical endpoint (i.e. beyond the termination of the solidus), so there is no temperature at which melting begins, unless the rock is completely free of volatiles. Heating causes continuous dissolution into the supercritical fluid, and there is no distinction between melt and vapour. After Wyllie and Ryabchikov (2000).

102

I propose that these mechanisms cause the majority of diamond-forming fluids to become subcritical carbonate-bearing melts, with limited water activity. These media form octahedrally- grown diamond. The water solubility in this carbonate-bearing melt would not be as high as carbonatitic melt (>10 wt %; Keppler, 2003), because if the melt truly was carbonate-rich it would remain a supercritical fluid at the pressure and temperature (P-T) conditions of diamond formation. If these mechanisms can be repressed so that the diamond-forming fluid remains supercritical, or if conditions can evolve from subcritical to supercritical, then higher water activities may permit fibrous diamond to form. This would augment the model in Section 6.2, such that fibrous diamond habit is triggered by high water activity, and high water activity depends on achieving supercritical conditions in the diamond-forming medium. In this framework, fibrous diamond would crystallize from supercritical carbonatitic fluids while octahedrally-grown diamond would crystallize from subcritical silicate-carbonate melts. At subcritical conditions, the fluid phase could potentially form diamond as well, if it contains sufficient CO2. Some support for miscible and immiscible behaviour in diamond-forming environments comes from Rege et al. (2010), who argued for immiscible separation of phases as the most viable mechanism to fractionate Ta from Nb and Hf from Zr to achieve the non- chondritic Ta/Nb and Hf/Zr ratios present in some fibrous diamond fluids. For common fibrous coats over octahedrally-grown diamond cores, the interpretation that the coats are much younger (Boyd et al., 1987) presents the convenient option the core and coat formed from completely distinct fluids. There is a large time gap, and the fibrous-diamond forming fluids may be generated and delivered in an entirely independent fashion from the core. However, the supercritical/subcritical idea described here allows us to entertain the possibility that there was no significant time gap between core and coat (Araújo et al., 2009). Undeniably, the interface between octahedral cores and fibrous coats is sharp and immaculate, suggesting no significant time gap of millions or billions of years during which resorption could have attacked the octahedral faces. The arrival of fresh, oxidized, fibrous diamond-forming fluids might be expected to cause some resorption before it can be reduced and begin crystallizing new diamond (Boyd et al., 1994), but fibrous coats are never reported to grow on resorbed diamonds. Araújo et al. (2009) have argued for continuous diamond growth from core to rim, noting that the diamond N content and Ba content change in the outermost 103

octahedral layers in a way that is sympathetic to the subsequent overgrowth of fibrous diamond. The diamond characteristics appear to anticipate the impending fibrous overgrowth. Cathodoluminescence images reveal regular, concentric, uninterrupted growth layers. Despite lower levels of N aggregation data in coats, these observations suggest that growth has occurred continuously within an evolving system, rather than in distinct events, from distinct fluids, separated in time. If growth really is continuous, the supercritical/subcritical model for controlling water activity, would provide an elegant explanation for the sudden, dramatic change in diamond growth habit. The onset of fibrous diamond growth may then reflect a transition to supercritical conditions where more water is allowed to coexist with the carbonate in the growth medium. The transition could be caused by an increase in fluid pressure, to exceed the 2CP, or more likely, by a shift in bulk composition of the subcritical melt/fluid that lowers the P-T conditions of the 2CP. A compositional shift could be the result of fractional crystallization or the ingress of a new pulse of carbonatitic fluid.

6.4 Behaviour of carbonate-bearing media in eclogite: Separating CO2

The Siberian diamonds discussed in Chapter 4 demonstrate that a free CO2 fluid can be

separated within an eclogitic diamond-forming environment (e.g. Luth, 1993). The studied CO2–

N2 inclusions are hosted in healed fractures, but the NE Siberian alluvial diamonds also contain primary carbonatitic microinclusions trapped in subgrain boundaries (Ragozin et al., 2009; Logvinova et al., 2011). The carbonatitic inclusions contain daughter minerals of Ba-Sr and Ca- Fe carbonates, K-Ba phosphates, Ti-Si and Ti-Al phases, and apatite (Logvinova et al., 2011), giving a bulk composition resembling to the silicic–low-Mg carbonatitic trend in fibrous diamond (Figure 1-5). Gas chromatography of these diamonds suggests the carbonatitic inclusions contain water as their dominant volatile, although the analysis was a mixture of primary and secondary inclusions, and may have incorporated volatiles adsorbed within surface- reaching fractures (Logvinova et al., 2011). The difference in inclusion style between the

carbonatitic and CO2–N2 fluids, being either along subgrain boundaries or healed fractures, implies the fluids were trapped at different times. Carbonatitic fluid inclusions were trapped

during growth, and CO2–N2 fluids were trapped later.

104

The concurrence of diamond, carbonate, and CO2 indicates fluid evolution near the conditions where the stability fields of these phases meet. This point is the intersection of the

dolomite–coesite–diopside–diamond (DCDD) fO2 buffer and the C–CO2 (CCO) fO2 buffer

(Figure 6-3) (Luth, 1993). Within an eclogite, release of CO2 fluid from carbonate could be encouraged by a temperature increase or by increasing the activity of dolomite relative to diopside. The latter would shift the DCDD curve to higher fO2 (Figure 6-3) and destabilize

carbonate at point "X", converting it to CO2 without changing P, T, or fO2.

Figure 6-3. Carbon phase changes with oxygen fugacity and temperature, at 5 GPa. Shaded regions show the stability field of carbonate, diamond, and CO2 in carbonated eclogite. The DCDD reaction is dolomite + coesite ↔ diopside + diamond. The CCO reaction is diamond ↔ CO2. The DCDV reaction is dolomite + coesite ↔ diopside + CO2. The noted shift in DCDD is caused by a relative increase in the dolomite/diopside activity ratio, which also shifts the DCDV reaction to lower temperatures. Relative carbonate/silicate activities can thus destabilize carbonate at point "X" and convert it to CO2 without changing P, T, or fO2. Reduction of carbonate (or CO2) to diamond is accomplished by lowering fO2 of the diamond-forming medium by reacting with more reduced host rocks. For reference, the long dashed lines mark the EMOD and EMFDD buffers. Note that this diagram is calculated for solid rocks without molten components, but is still an instructive guide. After Luth (1993).

Finding carbonatitic and CO2–N2 fluids trapped in eclogitic, non-fibrous diamonds is a significant observation as it has implications for the interpretation of C isotopes. Eclogitic, non- fibrous diamond is distinguished from other diamonds, as it is the only diamond variety with a 13 significant tail towards low δ C values (Figure 6-4a). Progressive loss of CO2 from a carbonatitic fluid was proposed by Cartigny et al. (1998a) as a means to fractionate 13C/12C and 13 produce the low δ C values in eclogitic diamond (Figure 6-4b). Liberation of a free CO2 fluid

105

with high δ13C should be permitted in eclogite host rocks, where there is no olivine that would

retain CO2 as carbonate (Cartigny et al., 1998a). Carbonatitic and CO2–N2 fluids in Siberian

diamonds lend support to this theory by proving that CO2 liberation does, in fact, occur in eclogitic diamond-forming environments. Further support comes from the 40 mol% N2

concentration in the CO2–N2 fluid inclusions, as it suggests the liberated CO2 carries N2 with is as it separates from carbonatitic fluids. This result is predicted by the sharp, progressive decrease 13 in worldwide diamond N concentration toward lower δ C values (Cartigny et al., 2001). As N2

enters the free fluid phase, it dilutes the CO2 and lowers its activity. Depending on how much the

CO2 activity decreases, there should be a shift of the DCDV curve to the left and a shift of the

CCO curve downward in Figure 6-3. Thus, partitioning of N2 into the fluid phase could

encourage further CO2 liberation.

a b Peridotitic diamonds (n = 1357) 3000 . ppm) . t a tion (

a 2000 t r e n c on

Eclogitic diamonds c (n = 997)

ogen 1000 r it N

-35 -25 -15 -5 -25 -20 -15 -10 -5 δ13C (‰) δ13C (‰) Figure 6-4. Carbon isotopic characteristics of eclogitic diamonds. a) Histograms showing the different distributions of carbon isotopic composition in peridotitic and eclogitic diamonds, after Cartigny (2005). b) Trend of decreasing maximum nitrogen concentration for lower δ13C values in diamonds. Cartigny et al. (2001) argue the limit sector curve reflects a process of isotopic fractionation by the escape of CO2. Worldwide data, of all parageneses (n ≈ 1200), after Cartigny et al. (2001).

106

Liberation and escape of CO2 is clearly indicated by fluid inclusions in diamonds and must therefore be considered an isotopic fractionation mechanism, even if other sources like subducted organic carbon might contribute to the low δ13C values of eclogitic diamonds. Although δ13C values have not been measured in the fluids, the associated isotopic fractionation has a sound basis from theory, experiments, and other geological settings (Javoy et al., 1986; Galimov, 1991).

The liberation of a free CO2 fluid is normally prohibited in peridotitic mantle, because

CO2 immediately reacts with olivine to make carbonates (Luth, 1993). However, the presence of

water should allow some portion of dissolved molecular CO2 to subsist, as seen in FTIR spectra of fibrous diamond (Figure 3-2). If so, examples of δ13C fractionation between carbonate and

CO2 may be present not only in eclogitic host rocks, but also in water-rich fibrous diamond- forming systems. Fibrous diamonds from Internationalnaya, Russia, exhibit a positive correlation between δ13C and the water/carbonate ratio (Zedgenizov et al., 2009). Given the high

water content of these fluids, there is likely to be dissolved molecular CO2 in the water in addition to the carbonate ions. Isotopic fractionation may then yield dissolved molecular CO2 with higher δ13C than the carbonate. When this hydrous-silicic-carbonatitic fluid eventually forms fibrous diamond, the diamond δ13C signature will be a function of the proportion of 13 carbonate to CO2 (hydrous) involved. In this case the high- δ C signature of CO2 does not escape from the system, as is proposed for in the model for octahedrally-grown eclogitic

diamond. Instead, the CO2 remains a miscible component in this supercritical carbonatitic fluid.

6.5 Implications for subduction signatures in the mantle

As discussed in Section 6.4, the finding of CO2–N2 fluid inclusions in eclogitic diamonds

strongly suggests CO2 liberation in eclogites is a real phenomenon in the mantle, and the ensuing C isotopic fractionation cannot be dismissed. Thus the conspicuosly low δ13C values that typify eclogitic diamonds (Figure 6-4) cannot be automatically assigned to subducted organic material (e.g. Sobolev and Sobolev, 1980; McCandless and Gurney, 1997). Cartigny (2005) also argued against recycled C because the large proportion of recycled carbonate accompanying low-δ13C organic matter should produce a greater proportion of higher δ13C values within -5–0‰ and homogenizing effects should bar extreme isotopic variability (cf. Figure 6-4). In general, caution

107

must be used when trying to use subduction to explain mantle signatures. It is often tempting to explain mantle signatures in terms what we know from studying the surface, because our knowledge of mantle geology and processes is comparatively limited. Mantle N isotopes, for example have mostly been approached with such a bias, relying on subduction of surficial signatures to explain δ15N variability (e.g. Marty and Dauphas, 2003). Motivated by my + observations of N in fluid inclusions, however, I proposed the effects of NH4 (Chapter 4) or metallic Fe (Chapter 5) may provide viable mechanisms for 15N/14N fractionation in the mantle. To expand on this point, diamond fluid inclusions described in Chapter 3 also yielded signatures that imitate familiar surficial signatures, but can be explained by mantle processes. Fluid inclusions in Archean fibrous diamonds from Wawa were found to have Cl-rich compositions and positive Eu anomalies. These could be interpretted as the signatures of seawater (Izraeli et al., 2001) and plagioclase breakdown, derived from subducted oceanic crust. However, Burgess et al. showed that mantle fluids could account for the Cl in saline fibrous diamond fluids (Burgess et al., 2009). Furthermore, as illustrated in Chapter 3, the Eu anomalies can be produced by the fluid, by virtue of its high Cl content as it percolates through mantle rocks. Fractionation of Eu from other rare earth elements is caused by its strong complexes with Cl− (Flynn and Burnham, 1978), an effect also seen in hydrothermal seafloor vents (Craddock et al., 2010). These results imply the imprint of subduction-related geochemistry in cratonic materials is often overestimated. In the absence of strong O and S isotopic evidence for a surficial origin of mantle material, within eclogites for example (e.g. Javoy et al., 1986), we should always consider mantle processes that may imitate familiar surficial geochemical signatures.

6.6 On the incompatibility of nitrogen There has long been debate over the compatibility of N in the diamond lattice (Section 1.2.8). My data on the N-rich fluid inclusions in octahedrally-grown diamonds (Chapter 4) supports N incompatibility and thus contributes to this currently unresolved issue. Generally decreasing N content from core to rim in many diamonds has been interpreted as depletion of the N in the fluid due to incorporation of N in the diamond, and a compatible relationship is inferred

(KN>1) (Boyd and Pillinger, 1994; Thomassot et al., 2007; Stachel et al., 2009; Wiggers de Vries 108

et al., 2013). However, this interpretation is not supported by other lines of evidence. Cartigny et al. (2001) argue that N is incompatible (KN<1) during growth and that its uptake is controlled by the kinetics of diamond growth. Diamond that grows slowly approaches equilibrium and favours

low N concentrations, according to its incompatibility (KN<1), whereas faster growth admits higher N concentrations as the diamond concedes to kinetic effects (Cartigny et al., 2001). Kinetic control over N uptake would explain the observation that fibrous diamond, which is believed to be the product of rapid crystallization (Sunagawa, 1984), has considerably higher average N concentrations than octahedrally-grown diamond (Cartigny, 2005). Furthermore, N incompatibility and kinetic control of N uptake has been demonstrated experimentally by a positive correlation between growth rate and the N concentration of diamonds grown in (Fe,Ni) metal (Babich et al., 2012). Along similar lines of reasoning, N incompatibility agrees with the observation of higher N concentrations in morphologically dominant {111} growth sectors compared to {100} sectors in natural cubo-octahedral diamonds (Sunagawa, 1984; Bulanova et al., 2002; Cartigny et al., 2003). Although morphologically dominant faces are the slowest growing faces, the growth steps within those faces advance more quickly, making it more likely for incompatible elements to be incorporated in dominant faces (Sunagawa, 2005). With kinetics imparting a significant influence, the resulting N concentration in a given diamond is a result of competition between equilibrium and kinetic effects. In Chapter 5, the difference in N concentration of lithospheric and sublithospheric diamond was argued to be caused by equilibrium partitioning in the absence or presence of metallic Fe, respectively. Both populations are still subject to kinetic effects of N uptake, but they are being drawn toward different equilibrium partitioning values. The incompatibility of N in diamond is intensified in the presence of metallic Fe, due to partitioning into the metallic phase, which has the effect of decreasing the overall concentration of N taken up by diamond (Figure 5-3). The N distributions in Figure 5-3 exemplify the competition between equilibrium and kinetic effects. The distributions have modes at low N concentrations where diamonds have approached equilibrium according to N incompatibility, and tails trending off toward higher N concentrations where kinetic effects have hindered equilibrium. The fluid inclusions in octahedrally-grown diamonds reported in Chapter 4 have high N concentrations that further support an incompatible relationship for N in diamond. All the 109

inclusions measured have C/N ratios below 0.8, which, if they have evolved from typical mantle values of 535±224 (Marty and Zimmermann, 1999), suggest that they are residual after >99% of the C to diamond. Thus, the high N concentrations in the fluid inclusions persist after most of the C has been precipitated as diamond. Furthermore, the N concentrations of 0.1 wt% in melt

inclusions and 40 mol% in CO2–N2 fluids, are higher than should be expected for carbonatitic melts or C-O-H-N fluids with mantle C/N ratios of 535±224 (Chapter 4). Incompatible behaviour

of N between the diamond and fluids (KN<1) is the simplest explanation for these observations,

at least for diamond formation from carbonate (and CO2). This is especially true for the CO2–N2 fluid with 40 mol% N, as diamonds are never found to contain anywhere near such high N concentrations. Measured N concentrations, and C/N ratios, in natural diamonds agree with N incompatibility and are consistent with kinetically controlled N incorporation. The highest expected diamond N concentration should then be governed by the C/N ratio of the growth medium (Cartigny et al., 2001). Such a relationship is demonstrated in Figure 4-5, for 917 measured cratonic diamonds, assuming diamond-forming fluids begin with typical mantle-like C/N ratios like the convecting MORB-source mantle at 535±224 (Marty and Zimmermann, 1999). The N-rich end of the cratonic diamond population effectively terminates at the C/N ratio of 535±224. Considering kinetics is especially important for diamond growth compared to the classic understanding of crystal growth from a magma. Whereas crystal growth in a magma depends on cooling, diamond growth by carbonate reduction is a redox reaction, whose rate slows inevitably as reactants are consumed and products accumulate. It should be expected that diamond growth rate slows during the course of an episode of growth. If N is incompatible, this slowing will naturally manifest itself as rimward-decreasing N contents in the diamond. Consequently, instances of rimward-decreasing N contents should not be interpreted as steady growth in a closed system, where N is being partitioned into diamond and depleted from the growth medium (e.g. Stachel et al., 2009; Wiggers de Vries et al., 2013). This misinterpretation casts doubt on recent modelling that has attempted to describe trends in δ13C that are observed along with decreasing N, which are inferred to record Rayleigh isotopic fractionation (Thomassot et al., 2007; Stachel et al., 2009; Wiggers de Vries et al., 2013). There 110

are two problems. Firstly, rimward-decreasing N contents in diamonds should not be used to infer closed-system behaviour. Secondly, the incompatible behaviour of N and kinetic control of it uptake mean that the diamond N concentration is not a direct function of the C remaining in the growth medium. In other words, the diamond N concentration is not described adequately by

a partition coefficient (KN) alone because of the competing effect of kinetics. The models attempting to account for covariations in diamond δ13C and N content describe the N/C ratio and δ13C as varying directly with the fraction of C remaining in the fluid, as shown by the following equation (Cartigny et al., 2001; Thomassot et al., 2007): 13 13 Equation 6-1: ln(N/C) = ln(N/C)0 + [(δ C−δ C0)/∆C] × (KN−1) 13 where (N/C) and δ C are for the instantaneous growth medium, with initial values of (N/C)0 and 13 δ C0. The isotopic fractionation factor between the growth medium and diamond, ∆C, is about

−3.5‰ for carbonate or CO2 (Richet et al., 1977; Stachel et al., 2009). Based on Equation 6-1, Stachel et al. (2009) explained the range of cratonic peridotitic diamond δ13C and N content (Figure 6-5a) by calling upon isotopically distinct methane and carbonatitic diamond-forming fluids, respectively, to explain diamonds δ13C values that fall below and above the modal δ13C value near -5‰ (Figure 6-5a). However, the two problems noted above suggest the Rayleigh model in Equation 6-1 may be inappropriate for diamond crystallization, at least for carbonate- bearing diamond-forming media. An alternative explanation for the variability of δ13C and N concentration in peridotitic diamonds can be made based on the incompatible behaviour of N in diamond and the influence of kinetic factors for N incorporation and δ13C fractionation (Figure 6-5b). If we adopt the view of N incompatibility, then the peak in N versus δ13C space is not a starting point that diamond δ13C values evolve away from as argued by Stachel et al. (2009) and shown by their modelling in Figure 6-5a. Instead, the peak represents disequilibrium N incorporation due to rapid diamond growth. As growth slows, it will then tend toward equilibrium N partitioning according to the growth environment. The data in Figure 6-5a clearly show that higher N concentrations are associated with a restricted range of δ13C, implying that rapid growth inhibits the effect of δ13C fractionation. Rapidly-grown fibrous diamonds also show this effect of having high N contents and relatively restricted values of δ13C centred on -5‰ (Cartigny, 2005). In the model proposed here, the modal diamond δ13C value of -5‰ is the product of subdued isotopic fractionation, 111

rather than the first-crystallized diamond from efficient, Rayleigh-type isotopic fractionation. Thus rapid diamond growth should produce diamond δ13C characteristics that more closely resemble the diamond forming medium (Figure 6-5b). It is therefore not surprising that diamonds with high N concentrations should converge on the expected mantle δ13C value near -5‰. An effect of muted δ13C fractionation with increasing diamond growth rate has, in fact, been shown experimentally (Reutsky et al., 2012).

a c Measured trends in individual diamonds b 5

Proposed behaviour based on N incompatibility

Rapid

Moderate

Slow

3 1 4

Figure 6-5. Variations in diamond N concentration versus δ13C. a) Worldwide dataset for peridotitic diamonds, shown with models for the co-evolution of N concentration and δ13C in fluid and precipitating diamond, with N behaving as a compatible element (after Stachel et al., 2009). The upper model uses CH4, in which the growing diamond depletes the fluid of 13C and N, and evolves toward lower δ13C values. The lower model uses a carbonate-based growth medium, in which the growing diamond depletes the fluid of 12C and N, and evolved toward higher δ13C values. b) Alternative model to (a) based on N incompatibility. Rapid diamond growth leads to high N concentrations and diminished δ13C fractionation. At slower growth rates less N is incorporated and δ13C fractionation is more pronounced. The arrows show Rayleigh-type δ13C fractionation for diamond growth from carbonate or CO2. c) Measured trends in cratonic diamonds. The evolution toward higher δ13C values and diminishing N concentrations agree with the proposed model in (b). Trends 1-5 are from Wiggers de Vries et al. (2013) and trend 5 is from Smart et al. (2011).

In contrast, diamonds with low N contents will reflect slow diamond growth that approaches equilibrium and promotes δ13C fractionation. The notion of closed-system behaviour can now be revisited on the basis of data presented in Chapter 4. Low C/N ratios in the 112

inclusions suggest that 99.8% of C can be precipitation as diamond, which would be difficult to achieve in an open system where fluids are permitted to leave the site of diamond growth. A locally closed system would then permit Rayleigh isotopic fractionation of δ13C and explain some of the δ13C variability observed, especially the progressive trends in δ13C in the growth layers of individual diamonds (Smart et al., 2011; Wiggers de Vries et al., 2013). When given the opportunity to fractionate, the first diamond grown from a carbonatitic fluid would be light, evolving toward heavier values (Stachel et al., 2009) as shown by the “Slow” growth trend in Figure 6-5b. Fractionation will produce a range of diamond with δ13C values lower and higher than the initial parental δ13C value. This model can thereby account for N concentrations and the range of δ13C below and above the central -5‰ value for cratonic peridotitic diamonds. This proposed model can explain diamond δ13C and N concentration trends measured within single diamonds by Wiggers de Vries et al. (2013) and Smart et al. (2011) (Figure 6-5c). Diamonds with lower N contents begin their growth path at lower values of δ13C, as expected for slower growth rates and more pronounced C isotopic fractionation. In the old model (Figure 6-5a), these different δ13C starting points requires a range of initial fluid δ13C. Taking kinetics into account eliminates this need by allowing the efficacy of thermodynamic isotopic fractionation to vary with growth rate. As the diamond-forming reaction slows with time, the incompatibility of N will produce a trend of decreasing diamond N concentrations. Also, the trends tend toward higher δ13C values, which is consistent with Rayleigh-type isotopic fractionation from a carbonatitic fluid (Stachel et al., 2009) It is therefore proposed that N incompatibility leads to a simpler explanation of diamond δ13C and N content variations, without having to invoke both methane and carbonatitic diamond forming fluids of isotopically distinct character (e.g. Stachel et al., 2009) (Figure 6-5a). The open or closed character of diamond-forming systems should be re-evaluated in consideration of N incompatibility before undertaking further modelling of diamond δ13C trends. Open systems are likely to be widespread in mantle processes. Thus, assumptions using equilibrium considerations are not always appropriate, as equilibrium is not attained. The open or closed character of fluid or melt related processes is intimately tied to interconnectivity between grain boundaries, and therefore, dihedral angles (Litasov and Ohtani, 2007). However, these angles vary greatly with composition of the fluid and the solid phase, and become more complex 113

upon considering supercritical fluids. Dihedral angles are thought to be low for carbonatitic fluids in the mantle, giving them mobility along grain boundaries (Hunter and McKenzie, 1989). The vast array of metasomatic processes involving carbonatitic fluids may be predisposed to be open and dynamic systems.

6.7 Future studies My research has opened up new prospective directions in diamond research and mantle petrology. Ideas for new studies that stem from my findings are summarized below.

• Until now, there have been no reports of bona fide fluid inclusions in any octahedrally- grown diamonds beyond a unique Siberian alluvial population (Chapter 4). I am the first to find and analyse fluid inclusions in octahedrally-grown diamond from additional localities. Surprisingly, several examples of fluid inclusions hosted in healed fractures were identified from a relatively small set of diamonds (Chapter 4). Thus, the most exciting area of future research will be a focussed search for more octahedrally-grown diamonds containing fluids inclusions. The search would require initial optical examination for prospective diamonds, followed by diamond polishing to make optical windows and more thorough exploration of the interior of each diamond with a petrographic microscope. Inclusions would be analysed using Raman spectroscopy to identify their composition. Cathodoluminescence imaging would be a valuable complimentary tool to reveal growth history and the style of any fracturing/healing of the diamond. • I found extreme N concentrations in fluid inclusions from octahedrally-grown diamond (Chapter 4), and this discovery should provoke inquiry into the N content in fibrous diamond fluid inclusions. Fibrous diamonds are reported to have high concentrations of N in the diamond lattice based on FTIR (Boyd et al., 1987; Cartigny, 2005). Methods normally used to study fibrous diamond fluids, however, are insensitive to N. The sub- micron size of the inclusions prohibits Raman spectroscopy, but gas chromatography of vacuum-crushed samples would allow measurement of the volatile species in the fluid.

After analysing the total amount of different volatiles, the original concentration of N2 in the fibrous diamond-forming fluid could be calculated by comparing the ratio of total 114

released N2 to water with FTIR measurements that give the ratio of water to daughter mineral species. • My own research (Chapter 4) and many cathodoluminescence images from the literature (e.g. Taylor and Anand, 2004) reveal healed internal fractures in diamonds. Inclusions hosted along healed fractures underscore the need to understand why or how diamonds fracture in the mantle. Dislocation mobility in diamond increases markedly above 900 °C (DeVries, 1975), which is why plastic deformation is common in diamond. Fracturing would require stresses that not only exceed the requirements for dislocation movement, but also exceed the brittle failure envelope. Investigating existing diamond deformation literature and potentially conducting high-pressure–high-temperature experiments would help constrain the conditions responsible for diamond fracturing in the mantle. • My model on the crucial role of water on the growth habit of diamond implies that the transition from octahedral-growth to fibrous growth in coated diamonds could proceed without a break in time, as a result switching from subcritical to supercritical conditions. The switch depends on the relationship between the pressure and temperature of the diamond-forming medium and its own second critical end point, which, in turn, changes with composition. The current literature on coated diamonds explains the core and coat as being produced by two distinct events, from two distinct fluids separated by a time gap. However, several authors have noted chemical changes in diamond properties leading up to, and sympathetic with, the coat interface (see Section 6.3) (Boyd et al., 1987; Boyd et al., 1994; Araújo et al., 2009). This idea deserves further investigation, in the form of systematic core-to-rim profiles of coated diamonds using µFTIR and SIMS, paying special attention to changes that occur just before the core/coat interface. The proposed new model of growth of coated diamond within an evolving system can be further constrained by a systematic experimental study of the aggregation kinetics in fibrous diamond. • The low N aggregation state of fibrous diamond has been used as an argument for very short mantle residence times, and therefore a close temporal and genetic relationship between fibrous diamond and kimberlite (Boyd et al., 1992). However, this interpretation makes the critical assumption that the initial lattice distribution and aggregation kinetics of

115

N in fibrous diamond are the same as octahedrally-grown diamond. The only existing models of aggregation kinetics (e.g. Taylor et al., 1990; Taylor et al., 1996) are based on octahedrally-grown natural diamond and other non-fibrous synthetic diamonds. Aggregation behaviour might be different in fibrous diamond. In fact, in a study of rare fibrous diamond cores with octahedrally-grown diamond rims Zedgenizov et al. (2006) noted that the N aggregation in the rims paradoxically indicated they were older than the fibrous cores. Howell et al. (2012) have also noted inconsistent aggregation behavior between octahedral, gemmy cuboid, and outer fibrous growth layers of a coated diamond. A systematic experimental study of the aggregation kinetics in fibrous diamond would allow us to make a more confident and accurate assessment of the timing of fibrous diamond growth. • The proposed new model of growth of coated diamond within an evolving system would benefit from having a better understanding of the total diamond growth time. It is currently not known whether diamond growth is something that tends to occur on the scale of minutes, years, or more. Octahedrally-grown diamonds are commonly analysed with FTIR to determine N concentration and aggregation state. These data are then plotted in a [N] versus %B diagram with several isotherms/isochrons on it and an inference is made about the temperature and/or duration mantle storage (Figure 6-6). However, there is a potentially powerful technique concealed within this established framework. It may be possible to determine the length of time taken to grow a diamond. Even if the timeframe is poorly constrained, it would be the only such estimate. The technique would involve taking many measurements of N concentration and aggregation state in a core-to-rim profile. The difference in extent of aggregation upon comparing an older layer (1) and a younger layer (2) would be a function of mantle storage temperature (T), time (t, time of residence at temperature T), and N concentration between 1 and 2 (Figure 6-6). When layer 1 grows, it has a period of time, ∆t, before layer 2 grows, during which its N will aggregate. When layer 2 grows, the aggregation of 1 and 2 become systematically paired because they are subject to the same t and T thereafter.

116

100 °C 1 2

400 Ma at

1600 Ma at 1065 °C 3200 Ma at 1050 °C 145 °C 1 10 °C 11

400 Ma at 1600 Ma at 1175 °C 140 °C 3200 Ma at 1095 °C 1 12 °C 11 1170 °C 155 °C 2 400 Ma at 1 1600 Ma at 400 Ma at 1205 °C 3200 Ma at 1600 Ma at 3200 Ma at Nitrogen concentration, [N] (ppm) 1 400 Ma at 1265 °C 1600 Ma at 1230 °C 3200 Ma at 1215 °C

Aggregation state (%B)

Figure 6-6. Nitrogen concentration and aggregation diagram. A diamond with a certain nitrogen concentration, [N], will have an aggregation state 0%B when it grows. Its %B will increase as a function of temperature, time. The curves shown (Taylor et al., 1990) are isochrons/isotherms. They are curved because higher values of [N] evolve more rapidly to higher %B values. The star points show an inner (1) and outer (2) growth layer that grew with no significant time gap. Star points 1 and 2 spent 400 Ma in the mantle at 1145 °C, or 1600 Ma at 1110 °C, or 3200 Ma at 1095 °C, etc. The circle points show an inner (1) and outer (2) growth layer separated by a significant time gap, during which point 1 was given a head start on aggregation.

Aggregation state, %B, is a function of time, temperature, and N concentration (e.g. Taylor et al., 1990). It can be rearranged to solve for mantle residence time, t, and the new function written out for layers 1 and 2:

Equation 6-2: t1 = f(T1, [N]1, %B1)

Equation 6-3: t2 = f(T2, [N]2, %B2) Layer 1 has spent longer in the mantle than layer 2, by an amount of time (∆t) equivalent to t1 − t2. We can now write:

Equation 6-4: ∆t = f(T1, [N]1, %B1) − f(T2, [N]2, %B2)

FTIR measurements provide [N]1, %B1, [N]2, and %B2. Then, by using an assumed temperature, it should be possible to gauge the time difference between different layers of 117

growth. It would also be possible to calculate ∆t with non-constant temperature. This innovative way to assess the time gap between distinct diamond growth episodes would show if models of the continuous or pulsed growth are more feasible. • Evolution of diamond-forming media, specifically with respect to water content, is central to my growth model of coated diamond. Changes in fluid composition during the growth can be seen in fibrous diamond fluid inclusions, and the evolution of the fluid could better traced if those changes are coupled with carbon isotopes. I propose to study fibrous diamond δ13C variations as a function of fluid composition in fibrous diamond. Several Brazilian diamonds have shown as association between heavy δ13C in fibrous diamonds with silicic fluids, but lighter δ13C in diamonds with carbonatitic fluids (Shiryaev et al., 2005). Diamonds from Internationalnaya exhibit a positive correlation between δ13C and the water/carbonate ratio (Zedgenizov et al., 2009). These observations may show Rayleigh isotopic fractionation of carbon during diamond crystallization from a carbonate- bearing medium. As the carbonate is reduced to diamond, the remaining fluid evolves to higher δ13C values and decreases in carbonate content. An alternative explanation is that

there is isotopic fractionation between carbonate and molecular CO2 dissolved in water. 13 The carbonate will be lower in δ C than the CO2. When this hydrous-silicic-carbonatitic fluid eventually forms diamonds, the diamond δ13C signature will be a function of the

proportion of carbonate to CO2 (hydrous) involved.

118

References Afanasyev, V.P., Agashev, A.M., Orihashi, Y., Pokhilenko, N.P. and Sobolev, N.V., 2009. Paleozoic U- Pb age of rutile inclusions in diamonds of the V–VII variety from placers of the northeast Siberian platform. Doklady Earth Sciences, 428(7): 1151-1155. Akagi, T. and Masuda, A., 1988. Isotopic and elemental evidence for a relationship between kimberlite and Zaire cubic diamonds. Nature, 336(6200): 665-667. Allègre, C.J., Staudacher, T. and Sarda, P., 1987. Rare gas systematics: formation of the atmosphere, evolution and structure of the Earth's mantle. Earth and Planetary Science Letters, 81(2-3): 127- 150. Allen, D.E. and Seyfried, W.E., Jr. , 2005. REE controls in ultramafic hosted MOR hydrothermal systems: An experimental study at elevated temperature and pressure. Geochimica et Cosmochimica Acta, 69(3): 675-683. Anand, M., Taylor, L.A., Misra, K.C., Carlson, W.D. and Sobolev, N.V., 2004. Nature of diamonds in Yakutian eclogites: views from eclogite tomography and mineral inclusions in diamonds. Lithos, 77(1-4): 333-348. Andersen, T., Burke, E.A.J. and Neumann, E.R., 1995. Nitrogen-rich fluid in the upper mantle: fluid inclusions in spinel dunite from Lanzarote, Canary Islands. Contributions to Mineralogy and Petrology, 120(1): 20-28. Andersen, T. and Neumann, E.R., 2001. Fluid inclusions in mantle xenoliths. Lithos, 55(1-4): 301-320. Araújo, D.P., Griffin, W.L., O'Reilly, S.Y., Grant, K.J., Ireland, T., Holden, P. and van Achterbergh, E., 2009. Microinclusions in monocrystalline octahedral diamonds and coated diamonds from Diavik, Slave Craton: Clues to diamond genesis. Lithos, 112(Supplement 2): 724-735. Arima, M., Kozai, Y. and Akaishi, M., 2002. Diamond nucleation and growth by reduction of carbonate melts under high-pressure and high-temperature conditions. Geology, 30(8): 691-694. Arima, M., Nakayama, K., Akaishi, M., Yamaoka, S. and Kanda, H., 1993. Crystallization of diamond from a silicate melt of kimberlite composition in high-pressure and high-temperature experiments. Geology, 21(11): 968-970. Babich, Y.V., Feigelson, B.N., Yelisseyev, A.P. and Chepurov, A.I., 2012. Nitrogen incorporation in octahedral diamonds grown in the Fe-Ni-C system. Geochemistry International, 50(2): 179-184. Ballhaus, C., 1995. Is the upper mantle metal-saturated? Earth and Planetary Science Letters, 132(1-4): 75-86. Bau, M., 1991. Rare-earth element mobility during hydrothermal and metamorphic fluid-rock interaction and the significance of the oxidation state of europium. Chemical Geology, 93(3-4): 219-230. Bercovici, D. and Karato, S.-i., 2003. Whole-mantle convection and the transition-zone water filter. Nature, 425: 39-44. Blatt, H. and Tracy, R.J., 1996. Petrology: Igneous, Sedimentary, and Metamorphic. W.H. Freeman and Co., New York. Boettcher, A.L. and Wyllie, P.J., 1969. The system CaO-SiO2-CO2-H2O−III. Second critical end-point on the melting curve. Geochimica et Cosmochimica Acta, 33(5): 611-632. Borzdov, Y., Pal'yanov, Y., Kupriyanov, I., Gusev, V., Khokhryakov, A., Sokol, A. and Efremov, A., 2002. HPHT synthesis of diamond with high nitrogen content from an Fe3N−C system. Diamond and Related Materials, 11(11): 1863-1870. Bouhifd, A., Roskosz, M., Jephcoat, A. and Mysen, B.O., 2010. Nitrogen solubility in a molten assemblage of an (Fe, Ni) alloy and a CI chondritic silicate up to 18 GPa. Geochimica et Cosmochimica Acta, 74(11): A109. Bowen, D.C., Ferraris, R.D., Palmer, C.E. and Ward, J.D., 2009. On the unusual characteristics of the diamonds from Letšeng-la-Terae kimberlites, Lesotho. Lithos, 112, Supplement 2(0): 767-774. 119

Boyd, F.R. and Gurney, J.J., 1986. Diamonds and the African Lithosphere. Science, 232(4749): 472-477. Boyd, S.R., Mattey, D.P., Pillinger, C.T., Milledge, H.J., Mendelssohn, M. and Seal, M., 1987. Multiple growth events during diamond genesis: an integrated study of carbon and nitrogen isotopes and nitrogen aggregation state in coated stones. Earth and Planetary Science Letters, 86(2-4): 341- 353. Boyd, S.R. and Pillinger, C.T., 1994. A preliminary study of 15N/14N in octahedral growth form diamonds. Chemical Geology, 116(1-2): 43-59. Boyd, S.R., Pillinger, C.T., Milledge, H.J., Mendelssohn, M.J. and Seal, M., 1992. C and N isotopic composition and the infrared absorption spectra of coated diamonds: evidence for the regional uniformity of CO2-H2O rich fluids in lithospheric mantle. Earth and Planetary Science Letters, 109(3-4): 633-644. Boyd, S.R., Pineau, F. and Javoy, M., 1994. Modelling the growth of natural diamonds. Chemical Geology, 116(1-2): 29-42. Bräuer, K., Kämpf, H., Niedermann, S., Strauch, G. and Weise, S.M., 2004. Evidence for a nitrogen flux directly derived from the European subcontinental mantle in the Western Eger Rift, Central Europe. Geochimica et Cosmochimica Acta, 68(23): 4935-4947. Brookes, E., Comins, J., Daniel, R. and Erasmus, R., 2000. A study of plastic deformation profiles of impressions in diamond. Diamond and Related Materials, 9: 1115-1119. Bulanova, G., Pearson, D., Hauri, E. and Griffin, B., 2002. Carbon and nitrogen isotope systematics within a sector-growth diamond from the Mir kimberlite, Yakutia. Chemical Geology, 188(1): 105-123. Bulanova, G.P., 1995. The formation of diamond. Journal of Geochemical Exploration, 53(1-3): 1-23. Bulanova, G.P., Walter, M.J., Smith, C.B., Kohn, S.C., Armstrong, L.S., Blundy, J. and Gobbo, L., 2010. Mineral inclusions in sublithospheric diamonds from Collier 4 kimberlite pipe, Juina, Brazil: Subducted protoliths, carbonated melts and primary kimberlite magmatism. Contributions to Mineralogy and Petrology, 160(4): 489-510. Bundy, F.P., Hall, H.T., Strong, H.M. and Wentorf, R.H., 1955. Man-Made Diamonds. Nature, 176(4471): 51-55. Bureau, H., Langenhorst, F., Auzende, A.-L., Frost, D.J., Estève, I. and Siebert, J., 2012. The growth of fibrous, cloudy and polycrystalline diamonds. Geochimica et Cosmochimica Acta, 77(0): 202- 214. Burgess, R., Cartigny, P., Harrison, D., Hobson, E. and Harris, J., 2009. Volatile composition of microinclusions in diamonds from the Panda kimberlite, Canada: Implications for chemical and isotopic heterogeneity in the mantle. Geochimica et Cosmochimica Acta, 73: 1779-1794. Burke, E.A.J., 2001. Raman microspectrometry of fluid inclusions. Lithos, 55(1-4): 139-158. Burns, R.C., Hansen, J.O., Spits, R.A., Sibanda, M., Welbourn, C.M. and Welch, D.L., 1999. Growth of high purity large synthetic diamond crystals. Diamond and Related Materials, 8(8-9): 1433-1437. Busigny, V. and Bebout, G.E., 2013. Nitrogen in the silicate Earth: Speciation and isotopic behavior during mineral−fluid interactions. Elements, 9(5): 353-358. Busigny, V., Cartigny, P. and Philippot, P., 2011. Nitrogen isotopes in ophiolitic metagabbros: A re- evaluation of modern nitrogen fluxes in subduction zones and implication for the early Earth atmosphere. Geochimica et Cosmochimica Acta, 75(23): 7502-7521. Cartigny, P., 2005. Stable isotopes and the origin of diamond. Elements, 1: 79-84. Cartigny, P. and Ader, M., 2003. A comment on “The nitrogen record of crust–mantle interaction and mantle convection from Archean to Present” by B. Marty and N. Dauphas [Earth Planet. Sci. Lett. 206(2003) 397–410]. Earth and Planetary Science Letters, 216(3): 425-432. Cartigny, P., Harris, J.W. and Javoy, M., 1998a. Eclogitic diamond formation at Jwaneng: No room for a recycled component. Science, 280(5368): 1421-1424. 120

Cartigny, P., Harris, J.W. and Javoy, M., 2001. Diamond genesis, mantle fractionations and mantle nitrogen content: A study of δ13C-N concentrations in diamonds. Earth and Planetary Science Letters, 185(1-2): 85-98. Cartigny, P., Harris, J.W., Phillips, D., Girard, M. and Javoy, M., 1998b. Subduction-related diamonds?− The evidence for a mantle-derived origin from coupled δ13C-δ15N determinations. Chemical Geology, 147(1-2): 147-159. Cartigny, P., Harris, J.W., Taylor, A., Davies, R. and Javoy, M., 2003. On the possibility of a kinetic fractionation of nitrogen stable isotopes during natural diamond growth. Geochimica et Cosmochimica Acta, 67(8): 1571-1576. Cartigny, P. and Marty, B., 2013. Nitrogen isotopes and mantle geodynamics: The emergence of life and the atmosphere−crust−mantle connection. Elements, 9(5): 359-366. Charlier, B.L.A., Ginibre, C., Morgan, D., Nowell, G.M., Pearson, D.G., Davidson, J.P. and Ottley, C.J., 2006. Methods for the microsampling and high-precision analysis of strontium and rubidium isotopes at single crystal scale for petrological and geochronological applications. Chemical Geology, 232(3-4): 114-133. Chung, F., 1974. Quantitative interpretation of X-ray diffraction patterns of mixtures. I. Matrix-flushing method for quantitative multicomponent analysis. Journal of Applied Crystallography, 7(6): 519- 525. Craddock, P.R., Bach, W., Seewald, J.S., Rouxel, O.J., Reeves, E. and Tivey, M.K., 2010. Rare earth element abundances in hydrothermal fluids from the Manus Basin, Papua New Guinea: Indicators of sub-seafloor hydrothermal processes in back-arc basins. Geochimica et Cosmochimica Acta, 74(19): 5494-5513. Creaser, R.A., Grütter, H., Carlson, J. and Crawford, B., 2004. Macrocrystal phlogopite Rb-Sr dates for the Ekati property kimberlites, Slave Province, Canada: Evidence for multiple intrusive episodes in the Paleocene and Eocene. Lithos, 76(1-4): 399-414. Crisp, J.A., 1984. Rates of magma emplacement and volcanic output. Journal of Volcanology and Geothermal Research, 20(3-4): 177-211. Cullity, B.D. and Stock, S.R., 2001. Elements of X-Ray Diffraction. Prentice Hall. Dalton, J.A. and Presnall, D.C., 1998. The continuum of primary carbonatitic-kimberlitic melt compositions in equilibrium with lherzolite: data from the system CaO-MgO-Al2O3-SiO2-CO2 at 6 GPa. Journal of Petrology, 39(11-12): 1953-1964. Dauphas, N. and Marty, B., 1999. Heavy nitrogen in carbonatites of the Kola Peninsula: A possible signature of the deep mantle. Science, 286(5449): 2488-2490. Davies, R., Griffin, W., Pearson, N., Andrew, A., Doyle, B. and O’Reilly, S.Y., 1999a. Diamonds from the deep: pipe DO-27, Slave Craton, Canada, Proc. 7th Int. Kimberlite Conf., Red Roof Design, Cape Town, pp. 148-155. Davies, R.M., O'Reilly, S.Y. and Griffin, W.L., 1999b. Diamonds from Wellington, NSW; insights into the origin of eastern Australian diamonds. Mineralogical Magazine, 63(4): 447-471. Davis, K.J., Dove, P.M. and De Yoreo, J.J., 2000. The role of Mg2+ as an impurity in calcite growth. Science, 290(5494): 1134-1137. Dawson, J.B., Pinkerton, H., Pyle, D.M. and Nyamweru, C., 1994. June 1993 eruption of Oldoinyo Lengai, Tanzania: Exceptionally viscous and large carbonatite lava flows and evidence for coexisting silicate and carbonate magmas. Geology, 22(9): 799-802. De Corte, K., Anthonis, A., Van Royen, J., Blanchaert, M., Barjon, J. and Willems, B., 2006. Overview of dislocation networks in natural type IIa diamonds. Gems & Gemology, 42: 122-123. De Stefano, A., Lefebvre, N. and Kopylova, M., 2006. Enigmatic diamonds in Archean calc-alkaline lamprophyres of Wawa, southern Ontario, Canada. Contributions to Mineralogy and Petrology, 151(2): 158-173. 121

De Yoreo, J.J. and Vekilov, P.G., 2003. Principles of Crystal Nucleation and Growth. Reviews in Mineralogy and Geochemistry, 54(1): 57-93. Deines, P., 1980. The carbon isotopic composition of diamonds: relationship to diamond shape, color, occurrence and vapor composition. Geochimica et Cosmochimica Acta, 44(7): 943-961. Deines, P., Harris, J.W. and Gurney, J.J., 1991. The carbon isotopic composition and nitrogen content of lithospheric and asthenospheric diamonds from the Jagersfontein and Koffiefontein kimberlite, South Africa. Geochimica et Cosmochimica Acta, 55(9): 2615-2625. Denaix, L., van Oort, F., Pernes, M. and Jongmans, A.G., 1999. Transmission X-ray diffraction of undisturbed soil microfabrics obtained by microdrilling in thin sections. Clays and Clay Minerals, 47(5): 637-646. DeVries, R.C., 1975. Plastic deformation and "work-hardening" of diamond. Materials Research Bulletin, 10: 1193-1200. Dickens, B. and Brown, W.E., 1970. Crystal structure of calcium carbonate hexahydrate at about -120°. Inorganic Chemistry, 9(3): 480-486. Dobrzhinetskaya, L.F., Wirth, R., Yang, J., Hutcheon, I.D., Weber, P.K. and Green, H.W., 2009. High- pressure highly reduced nitrides and oxides from chromitite of a Tibetan ophiolite. Proceedings of the National Academy of Sciences, 106(46): 19233-19238. Dobson, D.P., Jones, A.P., Rabe, R., Sekine, T., Kurita, K., Taniguchi, T., Kondo, T., Kato, T., Shimomura, O. and Urakawa, S., 1996. In-situ measurement of viscosity and density of carbonate melts at high pressure. Earth and Planetary Science Letters, 143(1-4): 207-215. Duan, Z. and Li, D., 2008. Coupled phase and aqueous species equilibrium of the H2O-CO2-NaCl-CaCO3 system from 0 to 250 °C, 1 to 1000 bar with NaCl concentrations up to saturation of halite. Geochimica et Cosmochimica Acta, 72(20): 5128-5145. Dubrovinsky, L., Dubrovinskaia, N., Kantor, I., Nestola, F. and Gatta, D., 2006. High-brilliance X-ray system for high-pressure in-house research: applications for studies of superhard materials. High Pressure Research, 26(2): 137 - 143. Evans, T., 1992. Aggregation of nitrogen in diamond. In: J.E. Field (Editor), The Properties of Natural and Synthetic Diamond. Academic Press Limited, San Diego, CA, pp. 259-290. Fagan, A.J. and Luth, R.W., 2011. Growth of diamond in hydrous silicate melts. Contributions to Mineralogy and Petrology, 161(2): 229-236. Faure, F., Trolliard, G., Nicollet, C. and Montel, J.-M., 2003. A developmental model of olivine morphology as a function of the cooling rate and the degree of undercooling. Contributions to Mineralogy and Petrology, 145(2): 251-263. Fedortchouk, Y., Canil, D. and Semenets, E., 2007. Mechanisms of diamond oxidation and their bearing on the fluid composition in kimberlite magmas. American Mineralogist, 92(7): 1200-1212. Field, J.E. (Editor), 1992a. The Properties of Natural and Synthetic Diamond. Academic Press Limited, San Diego, CA. Field, J.E. (Editor), 1992b. The Properties of Natural and Synthetic Diamond. Academic Press Limited, San Diego, CA, 710 pp. Fischer, T.P., Burnard, P., Marty, B., Hilton, D.R., Furi, E., Palhol, F., Sharp, Z.D. and Mangasini, F., 2009. Upper-mantle volatile chemistry at Oldoinyo Lengai volcano and the origin of carbonatites. Nature, 459(7243): 77-80. Fischer, T.P. and Marty, B., 2005. Volatile abundances in the sub-arc mantle: insights from volcanic and hydrothermal gas discharges. Journal of Volcanology and Geothermal Research, 140(1-3): 205- 216. Flynn, R.T. and Burnham, W.C., 1978. An experimental determination of rare earth partition coefficients between a chloride containing vapor phase and silicate melts. Geochimica et Cosmochimica Acta, 42(6): 685-701. 122

Font, L., Nowell, G.M., Graham Pearson, D., Ottley, C.J. and Willis, S.G., 2007. Sr isotope analysis of bird feathers by TIMS: a tool to trace bird migration paths and breeding sites. Journal of Analytical Atomic Spectrometry, 22(5): 513-522. Fowler, A.D., Berger, B., Shore, M., Jones, M.I. and Ropchan, J., 2002. Supercooled rocks: development and significance of varioles, spherulites, dendrites and spinifex in Archaean volcanic rocks, Abitibi Greenstone belt, Canada. Precambrian Research, 115(1-4): 311-328. Frezzotti, M.L., Ferrando, S., Tecce, F. and Castelli, D., 2012a. Water content and nature of solutes in shallow-mantle fluids from fluid inclusions. Earth and Planetary Science Letters, 351-352(0): 70- 83. Frezzotti, M.L., Peccerillo, A. and Panza, G., 2009. Carbonate metasomatism and CO2 lithosphere− asthenosphere degassing beneath the Western Mediterranean: An integrated model arising from petrological and geophysical data. Chemical Geology, 262(1-2): 108-120. Frezzotti, M.L., Tecce, F. and Casagli, A., 2012b. Raman spectroscopy for fluid inclusion analysis. Journal of Geochemical Exploration, 112(0): 1-20. Frost, D.J., Liebske, C., Langenhorst, F., McCammon, C.A., Trønnes, R.G. and Rubie, D.C., 2004. Experimental evidence for the existence of iron-rich metal in the Earth's lower mantle. Nature, 428(6981): 409-412. Frost, D.J. and McCammon, C.A., 2008. The redox state of Earth's mantle. Annual Review of Earth and Planetary Sciences, 36: 389-420. Galimov, E.M., 1991. Isotope fractionation related to kimberlite magmatism and diamond formation. Geochimica et Cosmochimica Acta, 55(6): 1697-1708. Giardini, A.A. and Melton, C.E., 1975. The nature of cloud-like inclusions in two Arkansas diamonds. The American Mineralogist, 60: 931-933. Goldblatt, C., Claire, M.W., Lenton, T.M., Matthews, A.J., Watson, A.J. and Zahnle, K.J., 2009. Nitrogen-enhanced greenhouse warming on early Earth. Nature Geoscience, 2(12): 891-896. Goldschmidt, V.M., 1913-1923. Atlas der Krystallformen. Carl Winters Universitätsbuchhandlung, Heidelberg, Germany. Gurney, J.J. and Helmstaedt, H.H., 2012. The origins of Type IIa diamonds and their enhanced economic significance. 10th International Kimberlite Conference, Short Abstract No. 10IKC-123: 247-248. Gurney, J.J., Helmstaedt, H.H., Le Roex, A.P., Nowicki, T.E., Richardson, S.H. and Westerlund, K.J., 2005. Diamonds: crustal distribution and formation processes in time and space and an integrated deposit model. In: J.W. Hedenquist, J.F.H. Thompson, R.J. Goldfarb and J.P. Richards (Editors), Economic Geology 100th Anniversary Volume. Society of Economic Geologists, pp. 143-177. Gurney, J.J., Helmstaedt, H.H. and Moore, R.O., 1993. A review of the use and application of mantle mineral geochemistry in diamond exploration. Pure and Applied Chemistry, 65(12): 2423-2442. Gurney, J.J., Helmstaedt, H.H., Richardson, S.H. and Shirey, S.B., 2010. Diamonds through time. Economic Geology, 105(3): 689-712. Guthrie, G.D.J., Veblenb, D.R., Navon, O. and Rossman, G.R., 1991. Submicrometer fluid inclusions in turbid-diamond coats. Earth and Planetary Science Letters, 105: 1-12. Halliday, A.N., 2013. The origins of volatiles in the terrestrial planets. Geochimica et Cosmochimica Acta, 105(0): 146-171. Hammersley, A., Svensson, S., Hanfland, M., Fitch, A. and Häusermann, D., 1996. Two-dimensional detector software: from real detector to idealised image or two-theta scan. High Pressure Research, 14: 235-248. Harlou, R., Pearson, D.G., Nowell, G.M., Ottley, C.J. and Davidson, J.P., 2009. Combined Sr isotope and trace element analysis of melt inclusions at sub-ng levels using micro-milling, TIMS and ICPMS. Chemical Geology, 260(3-4): 254-268.

123

Hashizume, K., Kase, T. and Matsuda, J., 1997. On the siderophile behavior of nitrogen and carbon: Implications for their inventory in the Earth. Kazan, 42: S293-S301. Hayman, P., Kopylova, M. and Kaminsky, F., 2005. Lower mantle diamonds from Rio Soriso (Juina area, Mato Grosso, Brazil). Contributions to Mineralogy and Petrology, 149(4): 430-445. Helmstaedt, H.H., Gurney, J.J. and Richardson, S.H., 2010. Ages of cratonic diamond and lithosphere evolution: Constraints on Precambrian tectonics and diamond exploration. Canadian Mineralogist, 48(6): 1385-1408. Howell, D., O'Neill, C.J., Grant, K.J., Griffin, W.L., Pearson, N.J. and O'Reilly, S.Y., 2012. µ-FTIR mapping: Distribution of impurities in different types of diamond growth. Diamond and Related Materials, 29(0): 29-36. Howell, D., Wood, I.G., Dobson, D.P., Jones, A.P., Nasdala, L. and Harris, J.W., 2010. Quantifying strain birefringence halos around inclusions in diamond. Contributions to Mineralogy and Petrology, 160(5): 705-717. Hunter, R.H. and McKenzie, D., 1989. The equilibrium geometry of carbonate melts in rocks of mantle composition. Earth and Planetary Science Letters, 92(3-4): 347-356. Hutchison, M., Cartigny, P. and Harris, J., 1999. Carbon and nitrogen compositions and physical characteristics of transition zone and lower mantle diamonds from Sao Luiz, Brazil, Proceedings of the 7th International Kimberlite Conference, pp. 372-382. Iakoubovskii, K. and Adriaenssens, G., 2002. Optical characterization of natural Argyle diamonds. Diamond and Related Materials, 11(1): 125-131. Ickert, R.B., Stachel, T., Stern, R.A. and Harris, J.W., 2013. Diamond from recycled crustal carbon documented by coupled δ18O−δ13C measurements of diamonds and theirinclusions. Earth and Planetary Science Letters, 364(0): 85-97. Irifune, T. and Tsuchiya, T., 2007. Mineralogy of the Earth−Phase transitions and mineralogy of the lower mantle. In: G.D. Price and G. Schubert (Editors), Treatise on Geophysics, pp. 33-62. Ishitani, T. and Yaguchi, T., 1996. Cross-sectional sample preparation by focused ion beam: A review of ion-sample interaction. Microscopy Research and Technique, 35(4): 320-333. Iwamori, H., 1998. Transportation of H2O and melting in subduction zones. Earth and Planetary Science Letters, 160(1-2): 65-80. Izraeli, E.S., Harris, J.W. and Navon, O., 1999. Raman barometry of diamond formation. Earth and Planetary Science Letters, 173(3): 351-360. Izraeli, E.S., Harris, J.W. and Navon, O., 2001. Brine inclusions in diamonds: a new upper mantle fluid. Earth and Planetary Science Letters, 187(3-4): 323-332. Izraeli, E.S., Harris, J.W. and Navon, O., 2004. Fluid and mineral inclusions in cloudy diamonds from Koffiefontein, South Africa. Geochimica et Cosmochimica Acta, 68(11): 2561-2575. Jacob, D.E., 2004. Nature and origin of eclogite xenoliths from kimberlites. Lithos, 77(1-4): 295-316. Jacob, D.E., Kronz, A. and Viljoen, K.S., 2004. Cohenite, native iron and troilite inclusions in garnets from polycrystalline diamond aggregates. Contributions to Mineralogy and Petrology, 146(5): 566-576. Janse, A.J.A., 1995. A history of diamond sources in Africa: Part I. Gems & Gemology, 31(4): 228-255. Javoy, M., Pineau, F. and Delorme, H., 1986. Carbon and nitrogen isotopes in the mantle. Chemical Geology, 57(1-2): 41-62. Jeng, H., 2010. On small samples and the use of robust estimators in loss reserving. Casualty Actuarial Society E-Forum, Fall: 1-27. Juster, T.C., Brown, P. and Bailey, S., 1987. NH4-bearing illite in very low grade metamorphic rocks associated with coal, northeastern Pennsylvania. American Mineralogist, 72(5-6): 555-565. Kaminsky, F., 2012. Mineralogy of the lower mantle: A review of 'super-deep' mineral inclusions in diamond. Earth-Science Reviews, 110(1-4): 127-147. 124

Kaminsky, F. and Wirth, R., 2013. New minerals in the primary, deep-seated carbonatitic association. Mineralogical Magazine, 77(5): 1424. Kaminsky, F., Zakharchenko, O., Davies, R., Griffin, W., Khachatryan-Blinova, G. and Shiryaev, A., 2001. Superdeep diamonds from the Juina area, Mato Grosso State, Brazil. Contributions to Mineralogy and Petrology, 140(6): 734-753. Kaminsky, F.V. and Wirth, R., 2011. Iron carbide inclusions in lower-mantle diamond from Juina, Brazil. The Canadian Mineralogist, 49(2): 555-572. Kanda, H., 1985. Effect of metal solvents and water on growth and dissolution morphologies of diamond crystals: unpublished D.Sc. thesis, Tohoku University, Sendai. Kanda, H. and Sekine, T., 2001. High temperature high pressure synthesis of single crystal diamond. In: A.J. Neves and M.H. Nazaré (Editors), Properties, Growth, and Applications of Diamond. INSPEC, The Institution of Electrical Engineers, Lodon, pp. 247-255. Kelemen, P.B., Braun, M. and Hirth, G., 2000. Spatial distribution of melt conduits in the mantle beneath oceanic spreading ridges: Observations from the Ingalls and Oman ophiolites. Geochemistry Geophysics Geosystems, 1(1999GC000012). Keppler, H., 2003. Water solubility in carbonatite melts. American Mineralogist, 88(11-12): 1822-1824. Keppler, H., Wiedenbeck, M. and Shcheka, S.S., 2003. Carbon solubility in olivine and the mode of carbon storage in the Earth's mantle. Nature, 424(6947): 414-416. Kerrich, R. and Jia, Y., 2004. A comment on “The nitrogen record of crust–mantle interaction and mantle convection from Archean to Present” by B. Marty and N. Dauphas [Earth Planet. Sci. Lett. 206(2003) 397–410]. Earth and Planetary Science Letters, 225(3-4): 435-440. Kessel, R., Ulmer, P., Pettke, T., Schmidt, M.W. and Thompson, A.B., 2005. The water-basalt system at 4 to 6 GPa: phase relations and second critical endpoint in a K-free eclogite at 700 to 1400 °C. Earth and Planetary Science Letters, 237: 873-892. Klein-BenDavid, O., Izraeli, E.S., Hauri, E. and Navon, O., 2004. Mantle fluid evolution–a tale of one diamond. Lithos, 77(1-4): 243-253. Klein-BenDavid, O., Izraeli, E.S., Hauri, E. and Navon, O., 2007a. Fluid inclusions in diamonds from the Diavik mine, Canada and the evolution of diamond-forming fluids. Geochimica et Cosmochimica Acta, 71(3): 723-744. Klein-BenDavid, O., Logvinova, A.M., Schrauder, M., Spetsius, Z.V., Weiss, Y., Hauri, E.H., Kaminsky, F.V., Sobolev, N.V. and Navon, O., 2009. High-Mg carbonatitic microinclusions in some Yakutian diamonds–a new type of diamond-forming fluid. Lithos, 112(Supplement 2): 648-659. Klein-BenDavid, O. and Pearson, D.G., 2009. Origins of subcalcic garnets and their relation to diamond forming fluids–Case studies from Ekati (NWT-Canada) and Murowa (Zimbabwe). Geochimica et Cosmochimica Acta, 73(3): 837-855. Klein-BenDavid, O., Pearson, D.G., Nowell, G.M., Ottley, C., McNeill, J.C.R. and Cartigny, P., 2010. Mixed fluid sources involved in diamond growth constrained by Sr–Nd–Pb–C–N isotopes and trace elements. Earth and Planetary Science Letters, 289: 123-133. Klein-BenDavid, O., Wirth, R. and Navon, O., 2006. TEM imaging and analysis of microinclusions in diamonds: A close look at diamond-growing fluids. American Mineralogist, 91: 353-365. Klein-BenDavid, O., Wirth, R. and Navon, O., 2007b. Micrometer-scale cavities in fibrous and cloudy diamonds – A glance into diamond dissolution events. Earth and Planetary Science Letters, 264(1-2): 89-103. Koga, K.T., Van Orman, J.A. and Walter, M.J., 2003. Diffusive relaxation of carbon and nitrogen isotope heterogeneity in diamond: a new thermochronometer. Physics of the Earth and Planetary Interiors, 139(1-2): 35-43. Kopylova, M., Navon, O., Dubrovinsky, L. and Khachatryan, G., 2010. Carbonatitic mineralogy of natural diamond-forming fluids. Earth and Planetary Science Letters, 291: 126-137. 125

Kopylova, M.G., Afanasiev, V.P., Bruce, L.F., Thurston, P.C. and Ryder, J., 2011. Metaconglomerate preserves evidence for kimberlite, diamondiferous root and medium grade terrane of a pre-2.7 Ga Southern Superior protocraton. Earth and Planetary Science Letters, 312(1-2): 213-225. Kumar, M.D.S., Akaishi, M. and Yamaoka, S., 2000. Formation of diamond from supercritical H2O−CO2 fluid at high pressure and high temperature. Journal of Crystal Growth, 213(1-2): 203-206. Lang, A.R., 1974. Space-filling by branching columnar single-crystal growth: An example from crystallisation of diamond. Journal of Crystal Growth, 23(2): 151-153. Lang, A.R. and Walmsley, J.C., 1983. Apatite inclusions in natural diamond coat. Physics and Chemistry of Minerals, 9(1): 6-8. Langford, J.I. and Wilson, A.J.C., 1978. Scherrer after sixty years: A survey and some new results in the determination of crystallite size. Journal of Applied Crystallography, 11(2): 102-113. Li, L., Cartigny, P. and Ader, M., 2009. Kinetic nitrogen isotope fractionation associated with thermal decomposition of NH3: Experimental results and potential applications to trace the origin of N2 in natural gas and hydrothermal systems. Geochimica et Cosmochimica Acta, 73(20): 6282-6297. Li, Y., Wiedenbeck, M., Shcheka, S. and Keppler, H., 2013. Nitrogen solubility in upper mantle minerals. Earth and Planetary Science Letters, 377-378(0): 311-323. Lippmann, F., 1959. Darstellung und kristallographische daten von CaCO3*H2O. Die Naturwissenschaften, 19: 553-554. Litasov, K.D. and Ohtani, E., 2007. Effect of water on the phase relations in Earth's mantle and deep water cycle. Geological Society of America Special Papers, 421: 115-156. Logvinova, A., Wirth, R., Fedorova, E. and Sobolev, N., 2008. Nanometre-sized mineral and fluid inclusions in cloudy Siberian diamonds: new insights on diamond formation. European Journal of Mineralogy, 20: 317-331. Logvinova, A.M., Wirth, R., Tomilenko, A.A., Afanas'ev, V.P. and Sobolev, N.V., 2011. The phase composition of crystal-fluid nanoinclusions in alluvial diamonds in the northeastern Siberian Platform. Russian Geology and Geophysics, 52(11): 1286-1297. Lowenstern, J.B., 2003. Melt inclusions come of age: Volatiles, Volcanoes, and Sorby's Legacy. In: B. De Vivo and R.J. Bodnar (Editors), Melt Inclusions in Volcanic Systems: Methods, Applications and Problems, Developments in Volcanology 5. Elsevier, Amsterdam, pp. 1-22. Luth, R.W., 1993. Diamonds, eclogites, and the oxidation state of the Earth's mantle. Science, 261: 66-68. Luth, R.W., 2005. Mantle volatiles–Distribution and consequences. In: R.W. Carlson (Editor), The Mantle and Core. Elsevier, pp. 319-361. Machado, N., Brooks, C. and Hart, S.R., 1986. Determination of initial 87Sr/86Sr and 143Nd/144Nd in primary minerals from mafic and ultramafic rocks: Experimental procedure and implications for the isotopic characteristics of the Archean mantle under the Abitibi greenstone belt, Canada. Geochimica et Cosmochimica Acta, 50(10): 2335-2348. Malkovets, V.G., Griffin, W.L., O'Reilly, S.Y. and Wood, B.J., 2007. Diamond, subcalcic garnet, and mantle metasomatism: Kimberlite sampling patterns define the link. Geology, 35(4): 339-342. Marty, B., 1995. Nitrogen content of the mantle inferred from N2-Ar correlation in oceanic basalts. Nature, 377(6547): 326-329. Marty, B., 2012. The origins and concentrations of water, carbon, nitrogen and noble gases on Earth. Earth and Planetary Science Letters, 313-314(0): 56-66. Marty, B. and Dauphas, N., 2003. The nitrogen record of crust−mantle interaction and mantle convection from Archean to Present. Earth and Planetary Science Letters, 206: 397-410. Marty, B. and Zimmermann, L., 1999. Volatiles (He, C, N, Ar) in mid-ocean ridge basalts: assesment of shallow-level fractionation and characterization of source composition. Geochimica et Cosmochimica Acta, 63(21): 3619-3633.

126

Marty, B., Zimmermann, L., Pujol, M., Burgess, R. and Philippot, P., 2013. Nitrogen isotopic composition and density of the Archean atmosphere. Science, 342(6154): 101-104. Matsuda, J., Sudo, M., Ozima, M., Ito, K., Ohtaka, O. and Ito, E., 1993. Noble gas partitioning between metal and silicate under high pressures. Science, 259(5096): 788-790. McCammon, C., 2001. Deep diamond mysteries. Science, 293(5531): 813-814. McCandless, T.E. and Gurney, J.J., 1997. Diamond eclogites: comparison with carbonaceous chondrites, carbonaceous shales, and microbial carbon-enriched MORB. Russian Geology and Geophysics, 38(2): 394-404. McDade, P. and Harris, J., 1999. Syngenetic inclusion bearing diamonds from Letseng-la-Terai, Lesotho. In: J.J. Gurney, J.L. Gurney, M.D. Pascoe and S.H. Richardson (Editors), Proceedings of the 7th International Kimberlite Conference. Red Roof Design, Cape Town, pp. 557-565. McDonough, W.F. and Sun, S.S., 1995. The composition of the Earth. Chemical Geology, 120(3-4): 223- 253. McMillan, P., 1984. Structural studies of silicate glasses and melts−Applications and limitations of Raman spectroscopy. American Mineralogist, 69: 622-644. McNeill, J.C.R., 2011. New techniques for trace element and radiogenic isotope measurement of diamonds: Their application to diamond petrogenesis and source tracing, PhD Thesis, Durham University, Durham, U.K., 281 pp. McNeill, J.C.R., Pearson, D.G., Klein-BenDavid, O., Nowell, G.M., Ottley, C. and Chinn, I., 2009. Quantitative analysis of trace element concentrations in some gem-quality diamonds. Journal of Physics: Condensed Matter, 21(36): 364207-364220. Melton, C.E. and Giardini, A.A., 1974. The composition and significance of gas released from natural diamonds from Africa and Brazil. American Mineralogist, 59: 775-782. Melton, C.E. and Giardini, A.A., 1975. Experimental results and a theoretical interpretation of gaseous inclusions found in Arkansas natural diamonds. American Mineralogist, 60: 413-417. Melton, C.E. and Giardini, A.A., 1981. The nature and significance of occluded fluids in three Indian diamonds. American Mineralogist, 66: 746-750. Melton, G.L., McNeill, J., Stachel, T., Pearson, D.G. and Harris, J.W., 2012. Trace elements in gem diamond from Akwatia, Ghana and DeBeers Pool, South Africa. Chemical Geology, 314-317: 1- 8. Meyer, H.O.A., 1987. Inclusions in diamond. In: P.H. Nixon (Editor), Mantle Xenoliths. John Wiley & Sons, Chichester, pp. 501-522. Michard, A., Albarede, F., Michard, G., Minster, J.F. and Charlou, J.L., 1983. Rare-earth elements and uranium in high-temperature solutions from East Pacific Rise hydrothermal vent field (13° N). Nature, 303(5920): 795-797. Miller, C.E., Kopylova, M. and Ryder, J., 2012. Vanished diamondiferous cratonic root beneath the Southern Superior province: Evidence from diamond inclusions in the Wawa metaconglomerate. Contributions to Mineralogy and Petrology. Miller, C.E., Kopylova, M.G. and Smith, E.M., 2013. Mineral inclusions in fibrous diamonds: Constraints on cratonic mantle refertilization. Mineralogy and Petrology. Miyazaki, A., Hiyagon, H., Sugiura, N., Hirose, K. and Takahashi, E., 2004. Solubilities of nitrogen and noble gases in silicate melts under various oxygen fugacities: implications for the origin and degassing history of nitrogen and noble gases in the earth. Geochimica et Cosmochimica Acta, 68(2): 387-401. Moore, A.E., 2009. Type II diamonds: Flamboyant Megacrysts? South African Journal of Geology, 112(1): 23-38. Navon, O., 1991. High internal pressures in diamond fluid inclusions determined by infrared absorption. Nature, 353(6346): 746-748. 127

Navon, O., Griffin, W.L. and Weiss, Y., 2011. "Table" vs "bench": Trace elements in fibrous diamonds. Mineralogical Magazine, 75. Goldschmidt Conference Abstracts: 1528. Navon, O., Hutcheon, I.D., Rossman, G.R. and Wasserburg, G.J., 1988. Mantle-derived fluids in diamond micro-inclusions. Nature, 335(6193): 784-789. Newton, R.C. and Manning, C.E., 2002. Experimental determination of calcite solubility in H2O-NaCl solutions at deep crust/upper mantle pressures and temperatures: Implications for metasomatic processes in shear zones. American Mineralogist, 87(10): 1401-1409. Orlov, Y.L., 1977. The Mineralogy of the Diamond. John Wiley & Sons, Inc., Toronto. Ostrovskaya, L., Perevertailo, V., Ralchenko, V., Dementjev, A. and Loginova, O., 2002. Wettability and surface energy of oxidized and hydrogen plasma-treated diamond films. Diamond and Related Materials, 11(3-6): 845-850. Pal'yanov, N., Sokol, A.G., Borzdov, M. and Khokhryakov, A.F., 2002a. Fluid-bearing alkaline carbonate melts as the medium for the formation of diamonds in the Earth's mantle: an experimental study. Lithos, 60(3-4): 145-159. Pal'yanov, Y., Borzdov, Y., Kupriyanov, I., Gusev, V., Khokhryakov, A. and Sokol, A., 2001. High- pressure synthesis and characterization of diamond from a sulfur−carbon system. Diamond and Related Materials, 10(12): 2145-2152. Pal'yanov, Y.N. and Sokol, A.G., 2009. The effect of composition of mantle fluids/melts on diamond formation processes. Lithos, 112(Supplement 2): 690-700. Pal'yanov, Y.N., Sokol, A.G., Borzdov, Y.M., Khokhryakov, A.F. and Sobolev, N.V., 2002b. Diamond formation through carbonate-silicate interaction. American Mineralogist, 87(7): 1009-1013. Palot, M., Cartigny, P., Harris, J.W., Kaminsky, F.V. and Stachel, T., 2012. Evidence for deep mantle convection and primordial heterogeneity from nitrogen and carbon stable isotopes in diamond. Earth and Planetary Science Letters, 357-358(0): 179-193. Palyanov, Y.N., Bataleva, Y.V., Sokol, A.G., Borzdov, Y.M., Kupriyanov, I.N., Reutsky, V.N. and Sobolev, N.V., 2013. Mantle−slab interaction and redox mechanism of diamond formation. Proceedings of the National Academy of Sciences. Palyanov, Y.N., Borzdov, Y.M., Khokhryakov, A.F., Kupriyanov, I.N. and Sokol, A.G., 2010. Effect of nitrogen impurity on diamond crystal growth processes. Crystal Growth & Design, 10(7): 3169- 3175. Palyanov, Y.N., Borzdov, Y.M., Kupriyanov, I.N. and Khokhryakov, A.F., 2012. Effect of H2O on diamond crystal growth in metal−carbon systems. Crystal Growth & Design, 12(11): 5571-5578. Palyanov, Y.N., Sokol, A.G., Tomilenko, A.A. and Sobolev, N.V., 2005. Conditions of diamond formation through carbonate-silicate interaction. European Journal of Mineralogy, 17(2): 207- 214. Patterson, M.V. and Francis, D., 2013. Kimberlite eruptions as triggers for early Cenozoic hyperthermals. Geochemistry, Geophysics, Geosystems, 14(2): 448-456. Pearson, D.G., Shirey, S.B., Carlson, R.W., Boyd, F.R., Pokhilenko, N.P. and Shimizu, N., 1995. Re-Os, Sm-Nd, and Rb-Sr isotope evidence for thick Archaean lithospheric mantle beneath the Siberian craton modified by multistage metasomatism. Geochimica et Cosmochimica Acta, 59(5): 959- 977. Perrillat, J.-P., Ricolleau, A., Daniel, I., Fiquet, G., Mezouar, M., Guignot, N. and Cardon, H., 2006. Phase transformations of subducted basaltic crust in the upmost lower mantle. Physics of the Earth and Planetary Interiors, 157(1-2): 139-149. Ragozin, A., Shatskii, V. and Zedgenizov, D., 2009. New data on the growth environment of diamonds of the variety V from placers of the northeastern Siberian platform. Doklady Earth Sciences, 425(2): 436-440.

128

Ragozin, A., Shatsky, V., Rylov, G. and Goryainov, S.V., 2002. Coesite inclusions in rounded diamonds from placers of the Northeastern Siberian platform. Doklady Earth Sciences, 384: 385-389. Rege, S., Griffin, W.L., Pearson, N.J., Araujo, D., Zedgenizov, D. and O'Reilly, S.Y., 2010. Trace- element patterns of fibrous and monocrystalline diamonds: Insights into mantle fluids. Lithos, 118(3-4): 313-337. Reutsky, V.N., Borzdov, Y.M. and Palyanov, Y.N., 2012. Effect of diamond growth rate on carbon isotope fractionation in Fe-Ni-C system. Diamond and Related Materials, 21(0): 7-10. Richardson, S.H., Gurney, J.J., Erlank, A.J. and Harris, J.W., 1984. Origin of diamonds in old enriched mantle. Nature, 310(5974): 198-202. Richet, P., Bottinga, Y. and Javoy, M., 1977. A review of hydrogen, carbon, nitrogen, oxygen, sulphur, and chlorine stable isotope enrichment among gaseous molecules. Annual Review of Earth and Planetary Sciences, 5: 65-110. Ringwood, A.E. and Hibberson, W., 1991. Solubilities of mantle oxides in molten iron at high pressures and temperatures: implications for the composition and formation of Earth's core. Earth and Planetary Science Letters, 102(3-4): 235-251. Rohrbach, A., Ballhaus, C., Golla-Schindler, U., Ulmer, P., Kamenetsky, V.S. and Kuzmin, D.V., 2007. Metal saturation in the upper mantle. Nature, 449(7161): 456-458. Rohrbach, A., Ballhaus, C., Ulmer, P., Golla-Schindler, U. and Schönbohm, D., 2011. Experimental evidence for a reduced metal-saturated upper mantle. Journal of Petrology, 52(4): 717-731. Rohrbach, A. and Schmidt, M.W., 2011. Redox freezing and melting in the Earth's deep mantle resulting from carbon-iron redox coupling. Nature, 472(7342): 209-212. Rondeau, B., Fritsch, E., Moore, M., Thomassot, E. and Sirakian, J.F., 2007. On the growth of natural octahedral diamond upon a fibrous core. Journal of Crystal Growth, 304(1): 287-293. Roskosz, M., Bouhifd, M.A., Jephcoat, A.P., Marty, B. and Mysen, B.O., 2013. Nitrogen solubility in molten metal and silicate at high pressure and temperature. Geochimica et Cosmochimica Acta, 121(0): 15-28. Sano, Y., Takahata, N., Nishio, Y., Fischer, T.P. and Williams, S.N., 2001. Volcanic flux of nitrogen from the Earth. Chemical Geology, 171(3-4): 263-271. Scherrer, P., 1918. Göttingen Nachrichten, 2: 96-100. Schrauder, M., Koeberl, C. and Navon, O., 1996. Trace element analyses of fluid-bearing diamonds from Jwaneng, Botswana. Geochimica et Cosmochimica Acta, 60(23): 4711-4724. Schrauder, M. and Navon, O., 1994. Hydrous and carbonatitic mantle fluids in fibrous diamonds from Jwaneng, Botswana. Geochimica et Cosmochimica Acta, 58(2): 761-771. Shahar, A., Bassett, W.A., Mao, H., Chou, I. and Mao, W., 2005. The stability and Raman spectra of ikaite, CaCO3*6H2O, at high pressure and temperature. American Mineralogist, 90(11-12): 1835- 1839. Shirey, S.B. and Richardson, S.H., 2011. Start of the Wilson Cycle at 3 Ga shown by diamonds from subcontinental mantle. Science, 333(6041): 434-436. Shiryaev, A.A., Izraeli, E.S., Hauri, E.H., Zakharchenko, O.D. and Navon, O., 2005. Chemical, optical, and isotopic investigations of fibrous diamonds from Brazil. Russian Geology and Geophysics, 46(12): 1207-1222. Shushkanova, A. and Litvin, Y.A., 2006. Formation of diamond polycrystals in pyrrhotite-carbonic melt: Experiments at 6.7 GPa. Doklady Earth Sciences, 409: 916-920. Shushkanova, A.V. and Litvin, Y.A., 2008. Diamond nucleation and growth in sulfide-carbon melts: an experimental study at 6.0−7.1 GPa. European Journal of Mineralogy, 20(3): 349-355. Simmons, J.W., 1996. Overview: high-nitrogen alloying of stainless steels. Materials Science and Engineering: A, 207(2): 159-169.

129

Smart, K.A., Chacko, T., Stachel, T., Muehlenbachs, K., Stern, R.A. and Heaman, L.M., 2011. Diamond growth from oxidized carbon sources beneath the Northern Slave Craton, Canada: δ13C−N study of eclogite-hosted diamonds from the Jericho kimberlite. Geochimica et Cosmochimica Acta, 75(20): 6027-6047. Smith, E.M. and Kopylova, M.G., 2013. A fresh look at Eu anomalies: The effect of Cl-rich fluids. GAC MAC Winnipeg 2013, Abstract No. 142, (oral presentation). Smith, E.M., Kopylova, M.G., Dubrovinsky, L., Navon, O., Ryder, J. and Tomlinson, E.L., 2011a. Transmission X-ray diffraction as a new tool for diamond fluid inclusion studies. Mineralogical Magazine, 75(5): 2657-2675. Smith, E.M., Kopylova, M.G., Dubrovinsky, L. and Tomlinson, E., 2010a. X-ray diffraction study of the mineral and fluid inclusions in fibrous diamond. Yellowknife Geoscience Forum 2010, (poster presentation). Smith, E.M., Kopylova, M.G. and Dubrovinsky, L.S., 2010b. X-ray diffraction study of the mineralogy of microinclusions in fibrous diamond. GAC MAC Calgary 2010, Geocanada, Abstract No. 271, (oral presentation). Smith, E.M., Kopylova, M.G. and Dubrovinsky, L.S., 2010c. X-ray diffraction study of the mineralogy of microinclusions in fibrous diamond. Geophysical Research Abstracts, 12, European Geosciences Union EGU 2010-4741-1 (oral presentation). Smith, E.M., Kopylova, M.G., Dubrovinsky, L.S. and Tomlinson, E.L., 2010d. X-ray diffraction study of the inclusions in fibrous diamond. Bayerisches Geoinstitut (BGI) Annual Report: 84-85. Smith, E.M., Kopylova, M.G., Frezzotti, M.L. and Afanasiev, V.P., 2013a. Diamond inclusions reveal fugitive mantle nitrogen. Goldschmidt 2013, Abstract ID 1601, (oral presentation - invited speaker). Smith, E.M., Kopylova, M.G., Frezzotti, M.L. and Afanasiev, V.P., 2013b. Nitrogen bubbles in the mantle: Evidence from diamond inclusions. GAC MAC Winnipeg 2013, Abstract No. 141, (oral presentation). Smith, E.M., Kopylova, M.G., Frezzotti, M.L. and Afanasiev, V.P., 2013c. “Vapour” vs. melt inclusions in Siberian placer diamonds. GEM (Geoscience for Energy and Minerals) Geological Survey of Canada, Diamond Project Workshop, Vancouver, (oral presentation). Smith, E.M., Kopylova, M.G., Frezzotti, M.L. and Afanasiev, V.P., 2014. N-rich fluid inclusions in octahedrally-grown diamond. Earth and Planetary Science Letters, 393(0): 39-48. Smith, E.M., Kopylova, M.G., Nowell, G.M., Pearson, D.G. and Ryder, J., 2012a. Archean mantle fluids preserved in fibrous diamonds from Wawa, Superior craton. Geology, 40(12): 1071-1074. Smith, E.M., Kopylova, M.G., Nowell, G.M., Pearson, D.G., Ryder, J. and Afanasiev, V.P., 2012b. The contrast in trace element chemistry and volatile composition between fluid inclusions in fibrous and octahedral diamonds. 10th International Kimberlite Conference, Extended Abstract No. 10IKC-102, (poster presentation). Smith, E.M., Kopylova, M.G. and Ryder, J., 2011b. Fluid inclusions in Archean diamonds from Wawa, Ontario. GAC MAC Ottawa 2011, Abstract No. 95, (oral presentation). Soave, G., 1972. Equilibrium constants from a modified Redlich-Kwong equation of state. Chemical Engineering Science, 27(6): 1197-1203. Sobolev, V.S. and Sobolev, N.V., 1980. New proof of very deep subsidence of eclogitized crystal rocks. Doklady Akademii Nauk SSSR, 250: 683-685. Sokol, A. and Pal'yanov, Y.N., 2008. Diamond formation in the system MgO-SiO2-H2O-C at 7.5 GPa and 1,600 °C. Contributions to Mineralogy and Petrology, 155(1): 33-43. Sokol, A.G., Palyanova, G.A., Palyanov, Y.N., Tomilenko, A.A. and Melenevsky, V.N., 2009. Fluid regime and diamond formation in the reduced mantle: Experimental constraints. Geochimica et Cosmochimica Acta, 73(19): 5820-5834. 130

Sokol, A.G., Tomilenko, A.A., Pal'yanov, Y.N., Borzdov, Y.M., Pal'yanova, G.A. and Khokhryakov, A.F., 2000. Fluid regime of diamond crystallisation in carbonate-carbon systems. European Journal of Mineralogy, 12(2): 367-375. Spetsius, Z.V. and Taylor, L.A., 2002. Partial Melting in Mantle Eclogite Xenoliths: Connections with Diamond Paragenesis. International Geology Review, 44(11): 973-987. Spivak, A.V., Litvin, Y.A., Shushkanova, A.V., Litvin, V.Y. and Shiryaev, A.A., 2008. Diamond formation in carbonate-silicate-sulfide-carbon melts: Raman-and IR-microspectroscopy. European Journal of Mineralogy, 20(3): 341-347. Stachel, T., 2007. Diamond, Mineralogical Association of Canada Short Course 37, Yellowknife, Northwest Territories, pp. 1-22. Stachel, T., Aulbach, S., Brey, G.P., Harris, J.W., Leost, I., Tappert, R. and Viljoen, K., 2004. The trace element composition of silicate inclusions in diamonds: a review. Lithos, 77(1): 1-19. Stachel, T., Brey, G. and Harris, J.W., 2005. Inlcusions in sublithospheric diamonds: glimpses of deep earth. Elements, 1: 73-78. Stachel, T., Harris, J., Aulbach, S. and Deines, P., 2002. Kankan diamonds (Guinea) III: δ13C and nitrogen characteristics of deep diamonds. Contributions to Mineralogy and Petrology, 142(4): 465-475. Stachel, T. and Harris, J.W., 1997a. Diamond precipitation and mantle metasomatism − evidence from the trace element chemistry of silicate inclusions in diamonds from Akwatia, Ghana. Contributions to Mineralogy and Petrology, 129(2-3): 143-154. Stachel, T. and Harris, J.W., 1997b. Syngenetic inclusions in diamond from the Birim field (Ghana)−A deep peridotitic profile with a history of depletion and re-enrichment. Contributions to Mineralogy and Petrology, 127(4): 336-352. Stachel, T. and Harris, J.W., 2008. The origin of cratonic diamonds − Constraints from mineral inclusions. Ore Geology Reviews, 34: 5-32. Stachel, T. and Harris, J.W., 2009. Formation of diamond in the Earth's mantle. Journal of Physics: Condensed Matter, 21(36): 364206. Stachel, T., Harris, J.W. and Muehlenbachs, K., 2009. Sources of carbon in inclusion bearing diamonds. Lithos, 112, Supplement 2(0): 625-637. Stagno, V., Ojwang, D.O., McCammon, C.A. and Frost, D.J., 2013. The oxidation state of the mantle and the extraction of carbon from Earth's interior. Nature, 493(7430): 84-88. Strong, H.M. and Hanneman, R., 1967. Crystallization of diamond and graphite. The Journal of Chemical Physics, 46: 3668. Sunagawa, I., 1984. Morphology of natural and synthetic diamond crystals. In: I. Sunagawa (Editor), Materials Science of the Earth's Interior. Terra Science, Tokyo, pp. 303-330. Sunagawa, I., 2005. Crystals: Growth, Morphology and Perfection. Cambridge University Press, 295 pp. Tackley, P.J., 2012. Dynamics and evolution of the deep mantle resulting from thermal, chemical, phase and melting effects. Earth-Science Reviews, 110(1-4): 1-25. Tappert, R., Foden, J., Stachel, T., Muehlenbachs, K., Tappert, M. and Wills, K., 2009. Deep mantle diamonds from South Australia: A record of Pacific subduction at the Gondwanan margin. Geology, 37(1): 43-46. Tappert, R., Stachel, T., Harris, J., Muehlenbachs, K., Ludwig, T. and Brey, G., 2005a. Diamonds from Jagersfontein (South Africa): messengers from the sublithospheric mantle. Contributions to Mineralogy and Petrology, 150(5): 505-522. Tappert, R., Stachel, T., Harris, J.W., Shimzu, N. and Brey, G.P., 2005b. Mineral inclusions in diamonds from the Panda kimberlite, Slave Province, Canada. European Journal of Mineralogy, 17(3): 423- 440.

131

Taylor, L.A. and Anand, M., 2004. Diamonds: time capsules from the Siberian Mantle. Chemie der Erde - Geochemistry, 64(1): 1-74. Taylor, L.A., Keller, R.A., Snyder, G.A., Wang, W., Carlson, W.D., Hauri, E.H., McCandless, T., Kim, K.-R., Sobolev, N.V. and Bezborodov, S.M., 2000. Diamonds and their mineral inclusions, and what they tell us: A detailed "pull-apart" of a diamondiferous eclogite. International Geology Review, 42(11): 959-983. Taylor, W., Jaques, A. and Ridd, M., 1990. Nitrogen-defect aggregation characteristics of some Australasian diamonds: Time-temperature constraints on the source regions of pipe and alluvial diamonds. American Mineralogist, 75: 1290-1310. Taylor, W.R., Canil, D. and Judith Milledge, H., 1996. Kinetics of Ib to IaA nitrogen aggregation in diamond. Geochimica et Cosmochimica Acta, 60(23): 4725-4733. Thomassot, E., Cartigny, P., Harris, J.W. and Viljoen, K.S., 2007. Methane-related diamond crystallization in the Earth's mantle: Stable isotope evidences from a single diamond-bearing xenolith. Earth and Planetary Science Letters, 257(3-4): 362-371. Thompson, M. and Lowthian, P.J., 2011. Notes on Statistics and Data Quality for Analytical Chemists. Imperial College Press. Tolansky, S. and Sunagawa, I., 1959. Spiral and other growth forms of synthetic diamond: A distinction between natural and synthetic diamond. Nature, 184: 1526-1527. Tomilenko, A.A., Chepurov, A.L., Palʼyanov, Y.N., Pokhilenko, L.N. and Shebanin, A.P., 1997. Volatile components in the upper mantle. Russian Geology and Geophysics, 38(1): 294-303. Tomlinson, E., De Schrijver, I., De Corte, K., Jones, A.P., Moens, L. and Vanhaecke, F., 2005. Trace element compositions of submicroscopic inclusions in coated diamond: A tool for understanding diamond petrogenesis. Geochimica et Cosmochimica Acta, 69(19): 4719-4732. Tomlinson, E., Jones, A. and Milledge, J., 2004. High-pressure experimental growth of diamond using C- K2CO3-KCl as an analogue for Cl-bearing carbonate fluid. Lithos, 77(1-4): 287-294. Tomlinson, E.L., Jones, A.P. and Harris, J.W., 2006. Co-existing fluid and silicate inclusions in mantle diamond. Earth and Planetary Science Letters, 250: 581-595. Tomlinson, E.L., Muller, W. and E.I.M.F., 2009. A snapshot of mantle metasomatism: Trace element analysis of coexisting fluid (LA-ICP-MS) and silicate (SIMS) inclusions in fibrous diamonds. Earth and Planetary Science Letters, 279: 362-372. Touret, J.L., 1992. CO2 transfer between the upper mantle and the atmosphere: temporary storage in the lower continental crust. Terra Nova, 4(1): 87-98. Touret, J.L.R. and Huizenga, J.-M., 2011. Fluids in granulites. Geological Society of America Memoirs, 207: 25-37. Van den Kerkhof, A. and Thiéry, R., 2001. Carbonic inclusions. Lithos, 55(1-4): 49-68. Van den Kerkhof, A.M., 1990. Isochoric phase diagrams in the systems CO2−CH4 and CO2−N2: Application to fluid inclusions. Geochimica et Cosmochimica Acta, 54(3): 621-629. Veksler, I.V. and Lentz, D., 2006. Parental magmas of plutonic carbonatites, carbonate-silicate immiscibility and decarbonation reactions: evidence from melt and fluid inclusions. In: J.D. Webster (Editor), Melt inclusions in plutonic rocks. Mineralogical Association of Canada, pp. 123-149. Walmsley, J.C. and Lang, A.R., 1992. On sub-micrometre inclusions in diamond coat: crystallography and composition of ankerites and related rhombohedral carbonates. Mineralogical Magazine, 56: 533-543. Wang, S.Y., Sharma, S.K. and Cooney, T.F., 1993. Micro-Raman and infrared spectral study of forsterite under high pressure. American Mineralogist, 78(5-6): 469-476.

132

Wang, X., Ruslinda, A.R., Ishiyama, Y., Ishii, Y. and Kawarada, H., 2012. Higher coverage of carboxylic acid groups on oxidized single crystal diamond (001). Diamond and Related Materials, 20(10): 1319-1324. Watenphul, A., Wunder, B. and Heinrich, W., 2009. High-pressure ammonium-bearing silicates: Implications for nitrogen and hydrogen storage in the Earth's mantle. American Mineralogist, 94(2-3): 283-292. Watenphul, A., Wunder, B., Wirth, R. and Heinrich, W., 2010. Ammonium-bearing clinopyroxene: A potential nitrogen reservoir in the Earth's mantle. Chemical Geology, 270(1-4): 240-248. Weiss, Y., Griffin, W.L., Bell, D.R. and Navon, O., 2011. High-Mg carbonatitic melts in diamonds, kimberlites and the sub-continental lithosphere. Earth and Planetary Science Letters, 309(3-4): 337-347. Weiss, Y., Griffin, W.L., Elhlou, S. and Navon, O., 2008. Comparison between LA-ICP-MS and EPMA analysis of trace elements in diamonds. Chemical Geology, 252(3-4): 158-168. Weiss, Y., Griffin, W.L. and Navon, O., 2009a. Trace elements of fibrous diamonds. EGU 2009, Abstract 802, Geophysical Research Abstracts, 11. Weiss, Y., Kessel, R., Griffin, W.L., Kiflawi, I., Klein-BenDavid, O., Bell, D.R., Harris, J.W. and Navon, O., 2009b. A new model for the evolution of diamond-forming fluids: Evidence from microinclusion-bearing diamonds from Kankan, Guinea. Lithos, 112(Supplement 2): 660-674. Weiss, Y., Kiflawi, I. and Navon, O., 2010. IR spectroscopy: Quantitative determination of the mineralogy and bulk composition of fluid microinclusions in diamonds. Chemical Geology, 275(1-2): 26-34. White, S.M., Crisp, J.A. and Spera, F.J., 2006. Long-term volumetric eruption rates and magma budgets. Geochemistry, Geophysics, Geosystems, 7(3): Q03010. Wiggers de Vries, D.F., Bulanova, G.P., De Corte, K., Pearson, D.G., Craven, J.A. and Davies, G.R., 2013. Micron-scale coupled carbon isotope and nitrogen abundance variations in diamonds: Evidence for episodic diamond formation beneath the Siberian Craton. Geochimica et Cosmochimica Acta, 100(0): 176-199. Williams, L.B., Wilcoxon, B.R., Ferrell, R.E. and Sassen, R., 1992. Diagenesis of ammonium during hydrocarbon maturation and migration, Wilcox Group, Louisiana, U.S.A. Applied Geochemistry, 7(2): 123-134. Wirth, R. and Yang, J., 2013. Sources of diamond formation revealed by nano-inclusions in diamond. GAC MAC Winnipeg 2013, Abstract No. 425. Wopenka, B. and Pasteris, J.D., 1986. Limitations to quantitative analysis of fluid inclusions in geological samples by laser Raman microprobe spectroscopy. Applied Spectroscopy, 40(2): 144-151. Wright, R.B. and Wang, C.H., 1973. Density effect on the Fermi resonance in gaseous CO2 by Raman scattering. Journal of Chemical Physics, 58(7): 2893-2895. Wyllie, P.J., Huang, W.L., Otto, J. and Byrnes, A.P., 1983. Carbonation of peridotites and decarbonation of siliceous dolomites represented in the system CaO-MgO-SiO2-CO2 to 30 kbar. Tectonophysics, 100(1-3): 359-388. Wyllie, P.J. and Ryabchikov, I.D., 2000. Volatile components, magmas, and critical fluids in upwelling mantle. Journal of Petrology, 41(7): 1195-1206. Yokochi, R., Marty, B., Chazot, G. and Burnard, P., 2009. Nitrogen in peridotite xenoliths: Lithophile behavior and magmatic isotope fractionation. Geochimica et Cosmochimica Acta, 73(16): 4843- 4861. Yu, R.Z., Ma, H.A., Liang, Z.Z., Liu, W.Q., Zheng, Y.J. and Jia, X., 2008. HPHT synthesis of diamond with high concentration nitrogen using powder catalyst with additive Ba(N3)2. Diamond and Related Materials, 17(2): 180-184.

133

Zedgenizov, D., Ragozin, A. and Shatsky, V., 2007a. Chloride-carbonate fluid in diamonds from the eclogite xenolith. Doklady Earth Sciences, 415(2): 961-964. Zedgenizov, D.A., Harte, B., Facility, E.I.M., Shatsky, V., Politov, A., Rylov, G. and Sobolev, N., 2006. Directional chemical variations in diamonds showing octahedral following cuboid growth. Contributions to Mineralogy and Petrology, 151(1): 45-57. Zedgenizov, D.A., Kagi, H., Shatsky, V.S. and Sobolev, N.V., 2004. Carbonatitic melts in cuboid diamonds from Udachnaya kimberlite pipe (Yakutia): evidence from vibrational spectroscopy. Mineralogical Magazine, 68(1): 61-73. Zedgenizov, D.A., Ragozin, A.L., Shatsky, V.S., Araujo, D., Griffin, W.L. and Kagi, H., 2009. Mg and Fe-rich carbonate-silicate high-density fluids in cuboid diamonds from the Internationalnaya kimberlite pipe (Yakutia). Lithos, 112(Supplement 2): 638-647. Zedgenizov, D.A., Rege, S., Griffin, W.L., Kagi, H. and Shatsky, V.S., 2007b. Composition of trapped fluids in cuboid fibrous diamonds from the Udachnaya kimberlite: LAM-ICPMS analysis. Chemical Geology, 240(1-2): 151-162. Zhang, Y. and Yin, Q.-Z., 2012. Carbon and other light element contents in the Earth's core based on first-principles molecular dynamics. Proceedings of the National Academy of Sciences, 109(48): 19579-19583. Zhang, Y. and Zindler, A., 1993. Distribution and evolution of carbon and nitrogen in the Earth. Earth and Planetary Science Letters, 117: 331-345. Zhang, Z. and Fedortchouk, Y., 2012. Records of mantle metasomatism in the morphology of diamonds from the Slave craton. European Journal of Mineralogy, 24(4): 619-632.

134

Appendices

Appendix A: Diamond sample catalogue The table below summarizes the diamond samples studied and the analyses performed on them. The place of origin for each diamond is indicated in the sample number by its prefix:

• MMZ: Mbuji-Mayi, Democratic Republic of Congo (formerly Zaire) • J and JER: Jericho kimberlite, Nunavut, Canada • W and WA: Wawa metaconglomerate (locally called leadbetter conglomerate), near Wawa, Ontario, Canada • PAN: Panda kimberlite (known to be from Panda, before mining operations started mixing ore material from different kimberlites), Ekati mine, NWT, Canada (diamonds borrowed from Dr. Emma Tomlinson of the Royal Holloway University of London) • Sib: (Siberian) Alluvial workings along the Ebelyakh River, NE Siberian Platform, Russia • DRC: Democratic Republic of Congo (formerly Zaire) • RV: Roberts Victor mine, South Africa • DVK: Diavik mine, NWT, Canada

Samples DRC-F01 and RV-F03 are the most scientifically valuable diamonds studied here, because they contain the first recognized silicate melt inclusions with exsolved N-rich bubbles. The letter "F" in a sample number stands for "fluid" and indicates a non-fibrous sample that was selected as for its potential fluid inclusions.

135

Cleaned Average # of Trace Sr Other in Chapter and/or publication Sample # diameter pieces Morphology Highlights XRD FTIR SEM EPMA elements isotopes Raman methods Polished HF+HNO3 featured in synchrotron Chapter 2, (Kopylova et al., 2010; MMZ-8 4 mm 2 fibrous cuboid  XRF   Smith et al., 2011a) synchrotron Chapter 2, (Kopylova et al., 2010; MMZ-9 4 mm 1 fibrous cuboid  XRF  Smith et al., 2011a) fibrous coated synchrotron Chapter 2, (Kopylova et al., 2010; MMZ-10 4 mm 2 octahedron  XRF   Smith et al., 2011a) fibrous coated synchrotron Chapter 2, (Kopylova et al., 2010; MMZ-11 5 mm 1 octahedron  XRF   Smith et al., 2011a) synchrotron Chapter 2, (Kopylova et al., 2010; MMZ-14 5 mm 1 fibrous cuboid   XRF   Smith et al., 2011a) synchrotron Chapter 2, (Kopylova et al., 2010; MMZ-15 5 mm 1 fibrous cuboid  XRF   Smith et al., 2011a) fibrous coated synchrotron Chapter 2, (Kopylova et al., 2010; MMZ-16 6 mm 1 octahedron  XRF   Smith et al., 2011a) fibrous coated synchrotron Chapter 2, (Kopylova et al., 2010; MMZ-19 5 mm 2 octahedron  XRF   Smith et al., 2011a) fibrous coated synchrotron Chapter 2, (Kopylova et al., 2010; MMZ-22 6 mm 1 octahedron  XRF   Smith et al., 2011a) synchrotron Chapter 2, (Kopylova et al., 2010; MMZ-25 4 mm 1 fibrous cuboid  XRF   Smith et al., 2011a) synchrotron Chapter 2, (Kopylova et al., 2010; MMZ-27 5 mm 1 fibrous cuboid  XRF   Smith et al., 2011a) fibrous coated synchrotron Chapter 2, (Kopylova et al., 2010; MMZ-28 4 mm 1 octahedron  XRF   Smith et al., 2011a) synchrotron Chapter 2, (Kopylova et al., 2010; MMZ-29 4 mm 1 fibrous cuboid  XRF   Smith et al., 2011a) fibrous coated synchrotron Chapter 2, (Kopylova et al., 2010; MMZ-31 5 mm 2 octahedron  XRF   Smith et al., 2011a) fibrous coated synchrotron Chapter 2, (Kopylova et al., 2010; MMZ-75 7 mm 2 irregular core  XRF   Smith et al., 2011a) fibrous coated synchrotron Chapter 2, (Kopylova et al., 2010; MMZ-76 2 mm 1 octahedron  XRF   Smith et al., 2011a) synchrotron Chapter 2, (Kopylova et al., 2010; MMZ-79 2 mm 1 fibrous cuboid  XRF   Smith et al., 2011a) fibrous coated synchrotron Chapter 2, (Kopylova et al., 2010; MMZ-81 3 mm 1 octahedron  XRF   Smith et al., 2011a) fibrous coated synchrotron Chapter 2, (Kopylova et al., 2010; MMZ-85 3 mm 1 octahedron  XRF   Smith et al., 2011a) synchrotron J-300339G 1 mm 1 fibrous cuboid  XRF  Chapter 2, (Smith et al., 2011a) fibrous coat synchrotron Chapter 2, 3, (Smith et al., W1 3 mm 2 (fragment) saline fluid       XRF   2011a, 2012a) synchrotron W2 1 mm 1 fibrous cuboid   XRF   Chapter 2, (Smith et al., 2011a) fibrous synchrotron  W3 1 mm 1 dodecahedroid   XRF   Chapter 2, (Smith et al., 2011a) 136

Cleaned Average # of Trace Sr Other in Chapter and/or publication Sample # diameter pieces Morphology Highlights XRD FTIR SEM EPMA elements isotopes Raman methods Polished HF+HNO3 featured in synchrotron W4 1 mm 1 cuboid   XRF   Chapter 2, (Smith et al., 2011a) irregular twinned synchrotron W5 1 mm 2 fibrous cubes   XRF   Chapter 2, (Smith et al., 2011a) fibrous synchrotron Chapter 2, (Smith et al., 2011a;  W6 1 mm >1 dodecahedroid   XRF   Miller et al., 2013) fibrous coat synchrotron Chapter 2, 3, (Smith et al., W7 2 mm 1 (fragment) saline fluid       XRF   2011a, 2012a) synchrotron W8 1 mm 1 irregular, granular   XRF   Chapter 2, (Smith et al., 2011a) fibrous coat synchrotron Chapter 2, 3, (Smith et al., W9 1 mm 1 (fragment) saline fluid       XRF   2011a, 2012a) W13 1 mm >1 fibrous cuboid (Miller et al., 2013)

W14 1 mm 1 fibrous cuboid      Chapter 3, (Smith et al., 2012a) W15 1 mm >1 fibrous cuboid (Miller et al., 2013) W16 1 mm >1 fibrous cuboid (Miller et al., 2013) W17 1 mm >1 fibrous cuboid (Miller et al., 2013)

W29 1 mm 1 fibrous cuboid       Chapter 3, (Smith et al., 2012a) W30 1 mm 1 fibrous cuboid    W31 1.5 mm 1 fibrous cuboid   W32 1 mm 1 fibrous cuboid   synchrotron W36 2 mm >1 fibrous cuboid  XRF  Chapter 2, (Smith et al., 2011a) W40 1 mm >1 fibrous cuboid (Miller et al., 2013) W41 1 mm >1 fibrous cuboid (Miller et al., 2013)

W50 1 mm 1 fibrous cuboid      Chapter 3, (Smith et al., 2012a) W52 1 mm >1 fibrous cuboid (Miller et al., 2013) W53 1 mm >1 fibrous cuboid (Miller et al., 2013) fibrous coated octahedron, synchrotron Chapter 2, (Tomlinson et al., PAN1 2 mm n/a polished fragment saline fluid XRF   2006, 2009; Smith et al., 2011a) fibrous coated octahedron, synchrotron Chapter 2, (Tomlinson et al., PAN2 2 mm n/a polished fragment saline fluid XRF   2006, 2009; Smith et al., 2011a) fibrous coated octahedron, synchrotron Chapter 2, (Tomlinson et al., PAN3 2 mm n/a polished fragment saline fluid XRF   2006, 2009; Smith et al., 2011a) fibrous coated octahedron, synchrotron Chapter 2, (Tomlinson et al., PAN4 2 mm n/a polished fragment saline fluid XRF   2006, 2009; Smith et al., 2011a) fibrous coated octahedron, synchrotron Chapter 2, (Tomlinson et al., PAN5 2 mm n/a polished fragment saline fluid XRF   2006, 2009; Smith et al., 2011a) 137

Cleaned Average # of Trace Sr Other in Chapter and/or publication Sample # diameter pieces Morphology Highlights XRD FTIR SEM EPMA elements isotopes Raman methods Polished HF+HNO3 featured in fibrous coated octahedron, synchrotron Chapter 2, (Tomlinson et al., PAN6 0.5 mm n/a polished fragment saline fluid XRF   2006, 2009; Smith et al., 2011a) fibrous coated octahedron, synchrotron Chapter 2, (Tomlinson et al., PAN7 2 mm n/a polished fragment saline fluid XRF   2006, 2009; Smith et al., 2011a) fibrous coated octahedron, synchrotron Chapter 2, (Tomlinson et al., PAN8 2 mm n/a polished fragment saline fluid XRF   2006, 2009; Smith et al., 2011a) octahedron, Sib2 2 mm 2 abraded/resorbed CO2-N2 fluid      Chapter 4

octahedral cluster, abraded/resorbed, Sib3 3 mm 2 prolate CO2-N2 fluid      Chapter 4

octahedral cluster, Sib4 2 mm 2 abraded/resorbed CO2-N2 fluid      Chapter 4

octahedral cluster, Sib5 2 mm 2 abraded/resorbed CO2-N2 fluid      

octahedral cluster, CO2-N2 fluid; abraded/resorbed, SiO2 & Al2SiO5 Sib6 3 mm 2 prolate in SEM      Chapter 4

octahedral cluster, Sib7 3 mm 1 abraded/resorbed  

octahedral cluster, abraded/resorbed, Sib8 3 mm 1 prolate coesite in Raman   fluorescent in Sib10 3 mm 1 dodecahedroid Raman   3.66 ct N2 bubble in melt DRC-F01 (6 mm) 1 octahedron inclusions    CL imaging  Chapter 4 octahedron, DRC-F02 1.5 mm 1 irregular/rough   fluorescent in WA-F01 1 mm 1 octahedron Raman   fluorescent in WA-F02 1 mm 1 octahedron Raman   fluorescent in WA-F03 1.5 mm 1 dodecahedroid Raman   fluorescent in WA-F04 1 mm 1 octahedron Raman   138

Cleaned Average # of Trace Sr Other in Chapter and/or publication Sample # diameter pieces Morphology Highlights XRD FTIR SEM EPMA elements isotopes Raman methods Polished HF+HNO3 featured in irregular, granular fluorescent in WA-F05 1 mm 1 octahedron Raman   octahedron, fluorescent in WA-F06 1 mm 1 resorbed Raman  

bubble in melt incl.?; fluorescent WA-F07 1 mm 1 octahedron in Raman   RV-F01 1 mm 1 octahedron  

octahedron, RV-F02 1.5 mm 1 polycentric twinned   octahedron, resorbed N2 bubble in melt RV-F03 1 mm 1 polycentric twin inclusion   Chapter 4 octahedron, RV-F04 1.5 mm 1 resorbed   RV-F05 1 mm 1 octahedron   octahedron, RV-F06 1 mm 1 resorbed   irregular fluorescent in RV-F07 1.5 mm 1 octahedron Raman   octahedron, fluorescent in RV-F08 1 mm 1 resorbed Raman   fluorescent in RV-F09 1 mm 1 octahedron Raman   octahedron, polycentric DVK-F1 4 mm 1 twinned, fractured  

octahedron, DVK-F2 4 mm 1 resorbed, fractured   octahedron, DVK-F3 3 mm 1 polycentric   cubo-octahedral DVK-F4 4 mm 1 fragment   octahedron, DVK-F5 3 mm 1 chipped   octahedron, fractured, with turbid/fibrous DVK-F6 4 mm 1 region in fracture?   JER-F01 1 mm 1 octahedron  

139

Appendix B: X-ray diffraction patterns The following images show the 2-dimensional X-ray diffraction patterns collected using the laboratory-based XRD setup at BGI, Germany, described in Chapter 2. The diamond sample number appears above each pattern, along with the collection time in seconds. Some image labels also designate "rot," for rotation, to indicate the sample was rotated during the collection. Patterns shown have been limited to one per diamond and some phases reported in Chapter 2 may not appear in these images. Some samples patterns only showing diamond reflections, and no additional phases, are omitted.

140

141

142

143

144

145

146

147

148

149

150

151

152

153

154

155

Appendix C: Detailed methods for Chapter 3

Electron microprobe analysis Major element compositions of the fluid inclusions were determined by electron microprobe analysis (EPMA) at the Dept. of Earth, Ocean and Atmospheric Sciences at the University of British Columbia, using a Cameca SX100 with an accelerating voltage of 15 kV and 20 mA current. Analyzed fluid inclusions were unexposed and intact, lying just below a polished and acid-cleaned surface. Approximately 80 microinclusions were analyzed for each the 3 fibrous diamond coat samples, W1, W7, and W9. Analyses with oxide totals below 1% begin to show spurious compositions. Thus, measurements with totals below 1% were rejected in diamonds W7 and W9. The cutoff was raised to 4% in sample W1 to improve data quality, which was allowed by its higher average totals. The average total oxides of all accepted analyses was 4.7%. Although the sub-micron size and high volatile content of the fluid inclusions gives low totals, the accuracy is better than 15% for the major elements present (Schrauder and Navon, 1994; Izraeli et al., 2004; Weiss et al., 2008). Each analysis was renormalized to 100% on a carbon, water and carbonate-free basis. This technique is widely used for determining fluid inclusion composition in fibrous diamond (e.g. Schrauder and Navon, 1994; Izraeli et al., 2001; Izraeli et al., 2004; Klein-BenDavid et al., 2004; Tomlinson et al., 2006; Zedgenizov et al., 2006; Klein-BenDavid et al., 2007a; Weiss et al., 2008; Klein-BenDavid et al., 2009; Zedgenizov et al., 2009; Kopylova et al., 2010).

Infrared spectroscopy Fourier transform infrared (FTIR) spectroscopy was carried out at the Dept. of Earth and Atmospheric Sciences, University of Alberta, using a Thermo Nicolet Nexus 470 FTIR spectrometer with a Continuum infrared microscope. Nitrogen was purged through the system for beam stability. Spectra were collected through 650–4000 cm-1 with 200 scans at a resolution of 4 cm-1. The focused beam footprint is approximately 100 μm across, but imperfections can scatter the beam within the diamond and make the analyzed region more diffuse. Each spectrum was given an appropriate baseline and normalized for sample thickness using a reference Type II diamond spectrum. Spectra were processed using a deconvolution

156

spreadsheet from D. Fisher (Diamond Trading Company) to fit A and B centre components. The detection limit is approximately 5–10 ppm and concentrations typically have an error of 5–10%.

Trace element and Sr isotope analysis Trace elements and Sr isotope characteristics were determined in all 6 diamonds, closely following the methodology described by McNeill et al. (2009) and Klein David et al. (2010). The technique employs off-line laser ablation to accumulate a sample within a sealed vessel. Samples were collected and analyzed at the Dept. of Earth Sciences, Durham University.

Sample collection Diamonds were ultrasonically cleaned in a 1:1 mix of 29 molar (M) ultra-purity (UPA,

triple distilled) HF and 16 M UPA HNO3, followed by ultrasonic cleaning in 6 M UPA HCl, and a rinse in Milli-Q water. Samples were collected from each diamond by off-line laser ablation in a sealed PTFE cell with a laser window, using a UP-213 New Wave Nd:YAG 213 nm laser. Diamonds were ablated for 3 hours in a rectangular raster pattern, liberating about 500 μg of material. One sample was taken from each diamond, except for the largest Wawa diamond (W1), from which one sample was collected from the outermost half of the coat and a second sample (A2) was collected from the inner half. Diamonds were weighed before and after ablation, so that element concentrations could be normalized to the mass of diamond ablated, to give bulk concentrations in each diamond.

Sample chemistry Ablated material was taken into solution by leaching the diamond and ablation cell with 29

M UPA HF and 16 M UPA HNO3 at 120 °C for 24 hours and 1 hour of ultrasonic agitation, augmented by a second leach with 6 M UPA HCl for 120 °C for 24 hours. The leachate was

dried down, then dissolved into up in 10 µl of 16 M UPA HNO3 and diluted to 0.5 M (3%) with UPA water for sample chemistry. All samples were divided into two aliquots, taking 20% by weight for trace element analysis, with the remainder for isotopic analysis. Aliquots for isotopic analysis were dried and

taken up in 3 N UPA HNO3. Sr was isolated into a separate fraction using Sr-spec resin (Charlier et al., 2006; Harlou et al., 2009).

157

Blanks Three total-procedural blanks (TPB) were prepared in the same manner as samples, post laser ablation, to gauge trace element contributions of all acids and materials used. One of three blanks (TPB3) had higher than expected concentrations compared to the other two (TPB1 and TPB2). These blanks were then compared to a larger dataset of 28 TPB’s for the same methodology, generated in the same lab, in the same 9 month period of as the Wawa fibrous diamond analyses, using the same reagents or manufacturer of reagents. Including additional blanks determined in the same laboratory using the same reagents allowed more meaningful statistics to be calculated. Analysis of this dataset using Chauvene’s criterion or a Grubbs test flag TPB3 as an outlier, while TPB1 and TPB2 are consistent with the main population of blanks. There is a clear case for using additional blanks to provide an adequate database for properly estimating the limits of detection and quantification. Robust statistical methods (Huber’s H15 method) (Jeng, 2010) incorporate all blanks, with appropriate weighting for outliers, as suggested by Thompson & Lowthian (2011). Using the combined set of 31 TPB’s, H15 means and standard deviations were calculated to define the limit of detection (LOD = mean TPB + 3σ) and limit of quantification (LOQ = mean TPB + 10σ) (Appendix E: Trace element and Sr isotope analyses).

Trace element mass spectroscopy Trace element aliquots were brought up to a volume of 250 µl and spiked with In to a concentration of 0.1 ppb. Trace element concentrations were determined with a Thermo Scientific Finnigan Element2 ICPMS, measuring the isotopes 49Ti, 85Rb, 88Sr, 89Y, 90Zr, 93Nb, 137Ba, 139La, 140Ce, 141Pr, 143Nd, 147Sm, 151Eu, 157Gd, 159Tb, 161Dy, 166Er, 172Yb, 175Lu, 179Hf, 208Pb, 232Th, and 238U. Each analysis scanned through the range of isotopes 4 times, for a total analysis time of 110 s per sample. Variably dilute solutions of standards AGV-1, BHVO-1, and W-2 were used to calibrate sample concentrations.

Sr isotope mass spectroscopy

The Sr aliquots were dried down, carefully taken up in 1 µl of 16 M UPA HNO3 and loaded on Re filaments with TaF5 activator (Charlier et al., 2006; Font et al., 2007). Sr isotopes

158

were measured using a Thermo Scientific Triton TIMS. Two 0.5 ng and two 1.0 ng NBS987 standards were run along with the samples.

159

Appendix D: Wawa electron microprobe data Fluid microinclusions in Wawa fibrous diamonds, W1, W7, and W9.

total oxide wt% SiO2 TiO2 Al2O3 Cr2O3 FeO MnO MgO CaO SrO BaO Na2O K2O P2O5 Cl SO2 renorm. initial Sample: W1 point

W-55-1 0.63