Experimental and Analytical Studies of Partial Melting in Planetesimals and the Martian Mantle by Max Collinet B.Sc

Experimental and analytical studies of partial melting in planetesimals and the Martian mantle by Max Collinet B.Sc. Geological Sciences, 2009 M.Sc. Geological Sciences, 2011 University of Liège, Belgium

Submitted to the Department of Earth, Atmospheric, and Planetary Sciences in partial fulfillment of the requirements for the degree of

Signature of Author: Department of Earth, Atmospheric, and Planetary Sciences October 30, 2019

Certified by: Timothy L. Grove Robert R. Shrock Professor of Earth and Planetary Sciences Thesis Supervisor

Accepted by: Robert D. van der Hilst Schlumberger Professor of Earth and Planetary Sciences Department Head

2 Experimental and analytical studies of partial melting in planetesimals and the Martian mantle by Max Collinet B.Sc. Geological Sciences, 2009 M.Sc. Geological Sciences, 2011 University of Liège, Belgium

Submitted to the Department of Earth, Atmospheric, and Planetary Sciences on October 30th, 2019 in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Geology

Abstract

Planetesimals and planetary embryos, the building blocks of planets, started to melt within a few million years of the formation of the solar system. This thesis explores, through experiments and the analysis of meteorites, the magmatic processes that affected those early-formed bodies. Chapter 1 presents low-pressure experiments that simulate the onset of melting of planetesimals made of different chondritic materials (H, LL, CI, CM and CV). H, LL and CI compositions, melted at lower temperature and produced partial melts with higher SiO2, Al2O3 and alkali element concentrations compared to CM and CV compositions. They formed unique trachyandesite achondrites upon crystallization. In Chapter 2, the experiments are compared to primitive achondrites, distinct groups of meteorites that represent the melting residues “left behind” within planetesimals. Cumulative evidence from trachyandesite achondrites and primitive achondrites suggests that the planetesimals that accreted in the inner solar system were not depleted in alkali elements relative to the composition of the sun’s photosphere. Chapter 3 is a detailed study of ureilites, the largest group of primitive achondrites. Twelve ureilites were analyzed to determine the chemical composition and relative proportions of olivine and pyroxene. Those analyses, together with additional experiments, constrain the initial Mg/Si ratio of the ureilite parent body. The experiments are used to develop a new geothermometer, based on the partitioning of Cr between olivine and pyroxene, which demonstrates that ureilites are residues of incremental melting. Chapter 4 is the first of two chapters describing igneous processes on Mars, a planet sometimes referred to as a planetary embryo due to its small size and early accretion age. It describes a high-pressure experimental study of the partial melting of the primitive Martian mantle and discusses the origin of rocks from the Martian crust. Finally, chapter 5 is a study of Fe-Mg isotopic fraction in the olivine of the “enriched” shergottite Northwest Africa 1068. The composition and crystallization history of the parental melt, which represents a melt extracted from the Martian mantle, are constrained by modeling diffusion and crystal growth simultaneously.

Thesis Supervisor: Timothy L. Grove Title: Robert R. Shrock Professor of Earth and Planetary Sciences

4 Acknowledgments

I have many people to thank for their help, support and encouragement, without which none of this work would have been possible. First, I want to thank my advisor and mentor, Tim Grove. Tim showed me how fun and exciting experimental petrology can be. By working with him, I learned how to perform careful experiments, interpret my results and communicate them orally and in writing. But more importantly, I learned that many worthwhile experimental projects are not flawlessly executed to prove a preconceived idea. They start with the identification of an important question, evolve with successive failures and have to be carried out with perseverance and a lot of curiosity. I am grateful for his unwavering kindness, patience and trust that I repeatedly tested by breaking equipment, misplacing tools and, of course, overtightening valves. I look forward to our future discussions and continued collaboration. Next, I want to thank Oli Jagoutz, the chair of my thesis committee and advisor for my second general project. Oli has always offered his guidance and has kindly listened to me talk about obscure meteorites while I was pushing back the redaction of our own project. I am also grateful to the other members of my thesis committee, Rick Binzel, Ben Weiss and Tim McCoy, for their precious comments and advice. The discussion we had during my defense was inspirational and motivates me to address many other questions related to the early history of the solar system and the differentiation of planetesimals. There are so many other members of the EAPS community, past and present, that I should acknowledge. I know that I will inevitable forget many and present my apologies to those. I thank Neel Chatterjee, for teaching me to use the microprobe. Mira Parsons and Heather Queyrouze, for their support and friendship. Taylor Perron, for his advice during my general exam. Matej Pec, for interesting discussions about melt migration. François Tissot, for sharing is knowledge of cosmochemistry. To my office mates, lab mates and other graduate student peers, Stephanie Brown, Alex Mitchell, Ben Mandler, Ben Klein, Niya Grozeva, Jean-Arthur Olive, Mike Eddy, Annie Bauer, Billy Shinevar, Marjorie Cantine, Maya Stokes, Eva Golos, Patrick Beaudry and Susana Hoyos, thank you for our numerous conversations and for sharing your experience with me. You all greatly helped me to navigate graduate school at MIT. While at MIT, I continued to receive the help and support of previous advisors and mentors. I thank Bernard Charlier, Olivier Namur and Jacqueline Vander Auwera for teaching my first petrology classes in Liège, introducing me to fundamental research, and for remaining in touch during those six years. I thank Etienne Médard and Bertrand Devouard for introducing me to meteorite research during my master at Clermont-Ferrand. I thank Francois Holtz, Harald Behrens and Stefan Weyer for hosting me for six months at the mineralogy institute of Hanover. To my parents, Monique and Roger, thank you for your constant encouragement and love and for fostering my curiosity from a very young age by taking me to the Alps and to the Astronomy club of Spa. To my sister, Odile, thank you for showing me the way and inspiring me with your discipline and dedication in everything you undertake. Your visits to Boston with Antoine and Léone were so much fun. To my wife Nina, thank you for keeping me grounded and supporting me relentlessly during the most difficult times of my PhD. Thank you for exploring New England with me and making me feel at home here. I love you and look forward to discover new places together very soon.

6 Table of Contents

Abstract ...... 3 Acknowledgments ...... 5 Table of Contents ...... 7 Introduction ...... 11 Chapter 1: Widespread production of silica- and alkali-rich melts at the onset of planetesimal melting ...... 17 Abstract ...... 17 1. Introduction ...... 18 2. Experimental and analytical methods ...... 20 2.1. MHC-pressure vessel experiments ...... 20 2.2. Gas mixing furnace experiments ...... 20 2.3. Starting materials ...... 21 2.4. Electron microprobe analyses ...... 21 3. Results ...... 22 3.1. Approach of equilibrium ...... 22 3.2. Mineral compositions and proportions ...... 24 3.3. Temperatures and reactions of melting ...... 24 3.4. Composition of experimental melts ...... 26 4. Discussion ...... 27 4.1. Comparison with previous experiments ...... 27 4.2. Evidence of alkali- and silica- rich melts in the meteorite record ...... 29 4.3. Physical properties of partial melts and mechanism of melt extraction ...... 33 5. Conclusion...... 35 6. References ...... 35 Figures and tables ...... 40 Supplementary material ...... 59 Chapter 2: Formation of primitive achondrites by partial melting of alkali-undepleted planetesimals in the inner solar system ...... 77 Abstract ...... 77 1. Introduction ...... 78 2. Mineral compositions and proportions in experiments...... 80

3. Experimental constraints on the fO2 of melting of primitive achondrites ...... 81 4. Major-element composition of parent bodies and their partial melts ...... 83

7 4.1. Acapulcoites-lodranites ...... 83 4.2. Winonaites ...... 84 4.3. Brachinites and brachinite-like achondrites ...... 85 4.4. Ureilit ...... 87 5. Location of alkali-undepleted planetesimals ...... 88 6. Timescale of alkali- and silica- rich magmatism ...... 90 7. Implications for the distribution of alkali elements in the early-solar system ...... 92 8. Conclusion...... 94 9. References ...... 94 Figures and tables ...... 101 Supplementary Material...... 111 Chapter 3: Incremental melting in the ureilite parent body: initial composition, melting temperatures and melt compositions ...... 113 Abstract ...... 113 1. Introduction ...... 114 2. Methods ...... 116 2.1. Experimental approach ...... 116 2.2. EPMA analyses...... 117 2.3. Calculation of the mode of ureilites and experiments ...... 118 3. Results ...... 119 3.1. Experiments on chondritic residues ...... 119 3.2. Mineral composition and petrographic description of ureilites ...... 120 4. Discussion ...... 121 4.1. The initial composition of the UPB ...... 121 4.2. Literature estimates of ureilite equilibration temperatures ...... 123 4.3. New thermometer based on the partitioning of Cr between olivine and low-Ca pyroxene ...... 125 4.4. Anomalous low-temperature ureilites ...... 127 4.5. Temperatures and extents of melting recorded by pigeonite-olivine ureilites ...... 127 4.6. Incremental melting and composition of “late-stage” silicate melts ...... 128 4.7. Peak temperatures of ureilites and thermal history of the UPB ...... 131

4.8. Variable fO2 during melting, heterogeneity of the UPB and location of the different mantle reservoirs ...... 133 5. Conclusions ...... 135 6. References ...... 136

8 Figures and tables ...... 141 Supplementary material ...... 158 ...... 182 Chapter 4. Melting of the Primitive Martian Mantle at 0.5-2.2 GPa and the Origin of Basalts and Alkaline Rocks on Mars ...... 183 Abstract ...... 183 1. Introduction ...... 184 2. The primitive mantle of Mars ...... 185 3. Experimental and analytical methods ...... 185 3.1. Starting material and experimental conditions ...... 185 3.2. Analytical techniques ...... 187 4. Results ...... 187 4.1. Attainment of equilibrium ...... 187 4.2. Phase assemblages ...... 188 4.3. Mineral compositions...... 188 4.4. Melt compositions ...... 189 4.5. Melting behavior ...... 190 5. Discussion ...... 192 5.1. Position of the solidus ...... 192 5.2. Comparison with the melting of terrestrial peridotites ...... 193 5.3. Comparisons with martian volcanic rocks ...... 194 5.4. The Mg-number of martian mantle reservoirs ...... 198 6. Conclusion...... 199 7. References ...... 200 Figures and tables ...... 204 Supplementary Material...... 213 Chapter 5. Crystallization history of enriched shergottites from Fe and Mg isotope fractionation in olivine megacrysts ...... 221 Abstract ...... 221 1. Introduction ...... 222 2. Analytical methods ...... 224 3. Results ...... 225 3.1. Olivine major element compositions ...... 225 3.2. Composition in minor and trace elements ...... 226 3.3. Fe and Mg isotopic composition ...... 227

9 4. Discussion ...... 228 4.1. Nomenclature and origin of olivine megacrysts in shergottites ...... 228 4.2. Origin of the chemical zoning in olivine megacrysts of NWA 1068 ...... 229 4.3. Diffusion-crystallization models ...... 232 4.4. Timescales of NWA 1068 crystallization ...... 239 4.5. Parental melt of NWA 1068 and implications for the origin of enriched shergottites .240 5. Conclusions ...... 242 6. References ...... 242 Figures and tables ...... 249 Supplementary material ...... 262

10 « En constatant entre les météorites et les masses profondes de notre globe, des liens d’une intimité surprenante, nous arrivons ainsi, non-seulement à dévoiler les phases les plus reculées de l’histoire de notre propre globe, mais encore à faire ressortir la parenté mutuelle des différentes parties de l’Univers. C’est ainsi que la Géologie, prise à un large point de vue, se rattache intimement à l’Astronomie physique, et que, si elle en reçoit des lumières, de son coté, elle contribue à l’éclairer et à la compléter. »

“In observing links of surprising intimacy between the meteorites and the profound masses of our globe, we arrive here, not only to unveil the most distant phases in the history of our own globe, but also to highlight the mutual relationship of the parts of the universe. Thus, Geology, in the broadest sense, is deeply linked to physical Astronomy, and while it receives its light, it also contributes to illuminating and completing it.”

- Gabriel Auguste Daubrée, 1879

Introduction

While meteorites have been collected and recognized as “rocks falling from the sky” by people of many civilizations, western scientists, from Aristotle to the scientists of the enlightenment, considered any claim of their existence to be pure superstition. Ernst Chladni (1794) was the first to suggest publicly that some rocks, such as the pallasite Krasnojarsk, had an extraterrestrial origin. It was not until 1803 and the fall of the L’Aigle meteorite, which produced 3000 fragments and was observed by many witnesses, that the scientific community accepted the existence of meteorites. The scientific value of meteorites was rapidly recognized by geologists. Henry C. Sorby, the inventor of the petrographic microscope, realized that the spherical components contained in some meteorites were made of quenched silicate melt and described them as “drops of fiery rain”. Those meteorites and the glassy spheres that they contain are now called chondrites and chondrules, respectively. Gabriel Auguste Daubrée noticed that most meteorites were made of minerals that are relatively rare in the continental crust: olivine, pyroxene and metal. He reasoned that, if they were abundant in meteorites, they might also be abundant in the Earth’s interior. He proposed that Earth mantle was made of peridotite and that the core was made of iron. On the other hand, because meteorites contain no quartz, he suggested that the parent bodies of meteorites were geologically “less evolved” than the Earth and that the formation of granite and gneiss on Earth were in some way connected to the presence of oceans. Daubrée was also one of the first experimental petrologist. In an attempt to understand the formation conditions of meteorites, he

11 melted olivine and pyroxene in the presence of either coal or hydrogen and compared the abundance of minerals in his experiments to different meteorites. During the first half the 20th century, meteorites played a central role in the development of geochemistry and cosmochemistry. Victor Goldschmidt used meteorites to define lithophile, siderophile and chalcophile elements in his classification of the chemical elements. Following other scientists such as Daubrée, he supposed that meteorites, as a whole, represent the material of which terrestrial planets are made. After analyzing the composition of the different components of meteorites, he weighed silicates, metals and sulfides in relative proportions 10-2-1 to calculate a “cosmic abundance” of the elements. To estimate the composition of the solar system, other geochemists analyzed bulk chondrites as "these objects are so obviously a heterogeneous mixture of materials from many sources and hence may be a proper mixture in themselves" (Suess and Urey, 1956). Since the 1960s and early 1970s, improvements in absorption spectroscopy have led to the recognition that Ivuna-type or type 1 carbonaceous chondrites (CI) are the closest in composition to the sun’s photosphere. In the 1950s, Clair Patterson used the Pb isotopic composition of the Canyon Diablo meteorite, an iron meteorite, to calculate the age of the Earth and of the solar system (4.55 ± 0.07 Ga). This value is within error of the currently accepted age of the solar system as represented by the time of condensation of Ca Al-rich inclusions (CAIs) in CV chondrites (4.567/4.568 Ga). Chondrites are fragments of planetesimals, which are the first planetary bodies that accreted in the solar system (i.e. 50-1000 km in diameter), some of which are now persevered in the asteroid belt. They are solid remnants of the solar nebula, are heterogeneous and are made of a mixture of micrometric to centimetric components that formed at different temperatures. Chondrites vary significantly in composition and likely accreted in different sectors of the protoplanetary disk. CI chondrites are similar in composition to the sun’s photosphere but are enriched in water relative to all other chondrites and were formed in the outer solar system. Other group of carbonaceous chondrites (e.g. CM, CO, CV, CK) have higher concentrations of CaO,

Al2O3 and lower concentrations of Na2O, K2O. Enstatite chondrites are thought to have formed closer to the Earth, contain no water and are rich in SiO2. Iron is mostly present as metal in enstatite chondrites and as oxide in carbonaceous chondrites. Ordinary chondrites are in many ways intermediate in composition between CI and enstatite chondrites. They also have variable bulk iron contents (FeO + Fe in H> L>LL).

12 Chondrites are heterogenous and compositionally diverse because the chemical elements, and their host minerals, condensed at different temperatures in the protoplanetary disk. During cooling, refractory elements (i.e. Ca, Al, Ti) condensed first, followed by the main elements (i.e. Fe, Mg, Si), moderately volatile elements (i.e. Na, K, Ga) and, at low temperature (<390 ºC), highly volatile elements (i.e. O, C, N). The order in which elements condensed with decreasing temperature is known as the condensation sequence. Close to the sun, the protoplanetary disk was initially too dense and too hot for volatile elements to condense. Thus, the chondritic material that formed terrestrial planets should have been different from the water-rich and oxidized CI chondrites. The presence of chemical gradients in the protoplanetary disk is supported by many observations. Refractory elements are abundant in CAIs, which are thought to have formed close to the sun. The composition of asteroids, giant planets, and comets suggest that the concentration of highly volatile elements (i.e. O, C, N) was low in the inner solar nebula but high beyond the “snow line” in the outer nebula. The average redox state of the Earth during accretion appears to have been intermediate between the one of Mercury (poor in FeO) and Mars (rich in FeO). Also, the Martian mantle appears to be richer in moderately volatile elements than the Earth’s mantle. However, the condensation sequence of the elements in the solar nebula does not explain important chemical features of meteorites and planets. The crust of Mercury is rich in Na and S, two elements that are expected to condense at relatively low temperature. Enstatite chondrites, sometimes considered as analogous to the precursor material of Mercury and the Earth, are actually enriched in alkali elements (Na and K) relative to refractory elements (Ca and Al). CK chondrites, the chondrites that are the most depleted in alkali elements and the most enriched in refractory elements are also the most oxidized and are thought to have formed in the outer solar system. Some of those anomalies can be explained by the transport of solid particles over great distances as predicted by dynamical models and evidenced by the diversity of chondrite components. Another reason that equilibrium condensation models fail to predict the composition of terrestrial planets is that planets grew from the merging of planetesimals and planetary embryos by increasingly energetic collisions. Those impacts produced the lunar magma ocean as well as magma oceans on other planetary bodies. Some volatile elements have been lost by evaporation during that stage. However, smaller bodies like the asteroid Vesta, the parent body of howardite- eucrite-diogenite (HED) meteorites, are even more depleted in volatile elements than terrestrial

13 planets. Vesta is believed to have experienced a stage of magma ocean that could be responsible for this depletion. Impacts are an obvious source of energy to generate magma oceans on planets but might not be sufficiently energetic to melt planetesimals to a large extent. Vesta and the parent bodies of many meteorites called “achondrites”, which do not contain chondrules, are known to have experienced high internal temperatures and partial melting. The heat source is thought to be the radioactive decay of 26Al to 26Mg with a half-life of 0.72 Ma. Significant excesses of the daughter isotope have been measured in CAIs and 26Al appears to have been homogeneously distributed in the solar system. Planetesimals that accreted earlier than ~2 half-lives of 26Al would have received enough heat to melt. This process represents an intermediate stage of planetary accretion and differentiation. An increasing number of meteorites called “primitive achondrites” appear to have recorded the onset of melting in planetesimals but little is known about the initial composition of their parent bodies, the mode of melting and the composition of the silicate melts. The purpose of my thesis is to combine chemical analyses of meteorites and experimental studies to improve our understanding of the igneous processes that pre-dated the formation of planets and their influenced on the chemical evolution of the solar system. In Chapter 1, I present melting experiments of a spectrum of chondritic compositions (H, LL, CI, CM and CV) that simulate the partial melting of planetesimals under different redox conditions (i.e. oxygen fugacities; fO2). I show that the proportion of alkali elements relative to major refractory elements (i.e. Ca and Al) is the parameter that controls the temperature at which melting begins and the composition of the partial melts and mineral residues. I then compare the results of these experiments to two types of achondrites: trachyandesite achondrites inferred to represent partial melts (Chapter 1) and primitive achondrites inferred to represent the melting residues of different planetesimals (Chapter 2). I find that all of these meteorites likely derive from planetesimals that were not depleted in alkali elements relative to the sun’s photosphere and that were similar in composition to the parent bodies of H, LL and CI chondrites. This suggests that the depletion in moderately volatile elements of asteroids, planetary embryos and terrestrial planets is a result of early igneous processes rather than being a primary feature of the chondritic precursors. In Chapter 3, I describe how the parent body of ureilites, which are the most abundant group of primitive achondrites, continued to melt following the exhaustion of plagioclase.

14 The bulk composition of the Earth and Mars can be estimated from element correlations in igneous rocks, and by making several assumptions based on geochemical and geophysical constraints. Dreibus and Wänke used Martian meteorites (shergottites) to calculate the bulk composition of the primitive Martian mantle. The Martian mantle is depleted in alkali elements relative to CI chondrites but to a lesser extent than the Earth mantle. In Chapter 4, I describe melting experiments of the primitive Martian mantle and show that they also produce melts relatively rich in SiO2 and alkali elements. Experimental melts are similar in composition to Northwest Africa 7533, a unique Martian meteorite, and several rocks analyzed at the surface of Mars by rovers. While critical to constrain the composition of the primitive mantle, shergottites, including the so-called “enriched” subgroup, were derived from distinct mantle reservoirs that melted later in Mars’ history, about 200 million years ago. In Chapter 5, I use Fe-Mg isotope analyses of large olivine “megacrysts” to constrain the composition of the parental melt of enriched shergottites. I present a model of simultaneous crystal growth and diffusion that reproduced the element and isotope profiles in olivine and is used to estimate the crystallization time. Chapters 1 and 2 are submitted manuscripts, chapter 3 is a manuscript in preparation and chapter 4 and 5 are published articles. The pronoun “we” is used to acknowledge the contributions of all co-authors.

Chapter 1: Collinet M. and Grove T. L. (A) Widespread production of silica- and alkali-rich melts at the onset of planetesimal melting, under review (submitted to Geochimica et Cosmochimica Acta). Chapter 2: Collinet M. and Grove T. L. (B) Formation of primitive achondrites by partial melting of alkali-undepleted planetesimals in the inner solar system, under review (submitted to Geochimica et Cosmochimca Acta). Chapter 3: Collinet M. and Grove T. L. (C) Incremental melting in the ureilite parent body: initial composition, melting temperatures and melt compositions, in preparation for Meteoritics and Planetary Science. Chapter 4: Collinet M., Médard E., Charlier B., Vander Auwera J., Grove T. L. (2015), Melting of the primitive martian mantle at 0.5–2.2 GPa and the origin of basalts and alkaline rocks on Mars, Earth Planetary and Science Letters 427, 83-94. Chapter 5: Collinet M., Charlier B., Namur O., Oeser M., Médard E. and Weyer S. (2017), Crystallization history of enriched shergottites from Fe and Mg isotope fractionation in olivine megacrysts, Geochimica and Cosmochimca Acta 207, 277- 297.

16 Chapter 1: Widespread production of silica- and alkali-rich melts at the onset of planetesimal melting

Abstract

We present the results of melting experiments on a suite of carbonaceous and ordinary chondritic compositions (CV, CM, CI, H and LL) performed at low pressure (0.1 to 13.1 MPa) and over a range of oxygen fugacity (log fO2 – (log fO2)IW = -2.5 to -1 and +0.8, IW being the iron-wustite buffer). These experiments constrain the composition of partial melts (F = 5-25 wt.%) of chondritic planetesimals. Most experiments (∆ IW -2.5 to -1) were conducted in Molybdenum-Hafnium Carbide pressure vessels, which prevented the loss of alkali elements from the melt. The results show that all planetesimals not significantly depleted in moderately volatile elements relative to the sun’s photosphere (e.g. CI, H and LL compositions) produced low-degree melts (<15 wt.%) rich in SiO2, Al2O3 and alkali elements, regardless of the fO2. Despite their high apparent viscosities (104-5 Pa.s), such low-density partial melts (2400-2500 kg/m3) were mobilized and, upon crystallizing, formed rocks containing up to 80 vol.% of plagioclase An10-30 (i.e. oligoclase) such as the trachyandesite achondrites ALM-A, GRA 06128/9, NWA 6698 and NWA 11575, originating from three different parent bodies, the “albitic clasts” of polymict ureilites and some “alkali-silica-rich” inclusions in non-magmatic iron meteorites. We suggest that silica- and alkali- rich melts were widespread in small bodies of the early solar system but that much evidence was erased by subsequent stages of melting, planetary differentiation and later surface impact cratering.

17 1. Introduction

The partial melts of planetesimals have long been assumed to be basaltic in composition

(~50 wt.% SiO2, low concentrations of alkali elements) because the first meteorites recognized as the crystallization products of silicate melts were the meteorites from Vesta (Howardites-Eucrites- Diogenites; HED). The initial interpretation of the basaltic eucrites was that they were low degree partial melts of a primitive undifferentiated chondritic parent body (Stolper, 1977; Consolmagno and Drake, 1977). Later work by Jurewicz et al. (1993) showed that melting chondrites under a range of oxygen fugacity (fO2) conditions (∆IW -1 to +2) could produce basaltic melts if alkali elements were lost simultaneously. However, the evolution of thought on the origin of HED and other groups of achondrites (e.g. angrites) is that they are not low-degree partial melts, but instead that they represent crystallization products in parent bodies that experienced large degrees of melting, efficient core formation and complex igneous differentiation (Righter and Drake, 1997; Greenwood et al., 2005; Weiss et al., 2008; Mandler and Elkins-Tanton, 2013). Similarly, magmatic-iron meteorites are generally thought to have formed in planetesimals that melted to a large extent (e.g. Kruijer et al., 2014). But then, what is the composition of the first silicate melts of primitive chondritic materials? The evidence for removal of small melt fractions (< 15 wt.%) in planetesimals is abundant and clear. Ureilites, acapulcoites-lodranites, brachinites, winonaites and several ungrouped primitive achondrites represent the mantle of partially differentiated planetesimals (Scott et al., 1993; Keil, 2014; Hunt et al., 2017; Keil and McCoy, 2018). However, the melts that were extracted to form these residues have been difficult to identify. Over the last decade, a group of miscellaneous achondrites containing a large proportion of sodic plagioclase (i.e. oligoclase) has been discovered. It is currently composed of Graves Nunatak (GRA) 06128/9 (Day et al., 2009; Shearer et al., 2010), the Almahata Sitta sample ALM-A (Bischoff et al., 2014), Northwest Africa (NWA) 6698 (Bunch et al., 2011) and NWA 11575 (Agee et al., 2018). Their whole-rock compositions are trachyandesitic and characterized by high concentrations of SiO2, Al2O3 and alkali elements (Fig. 1a). The oxygen isotope compositions of “trachyandesite achondrites” suggest that they originated in at least three different parent bodies (Fig. 1b). GRA 06128/9 and ALM-A could represent the crystallization products of silicate melts that were extracted from

18 brachinites and ureilites, respectively. NWA 6698/11575 could have been produced by melting a planetesimal with affinities for LL-chondrites. In this study we demonstrate that these unique and unusual achondrites crystallized from the low-degree partial melts of chondritic materials. They are the “missing links” that relate primitive achondrites that have lost silicate melts to the initial chondritic composition of their parent bodies. We present an extensive set of experiments performed in Molybdenum Hafnium Carbide pressure vessels (MHC-PV), which virtually eliminates the issue of alkali element losses that have plagued previous experimental studies. The experimental results reported here allow us to explore in detail the melting behavior of chondritic materials with different concentrations of moderately volatile elements (MVE; i.e. Na, K) and refractory lithophile elements (i.e. Al and Ca) under different fO2 conditions. We express the relative abundances of MVE and refractory elements in planetesimals as the NaK#: (Na+K)/(Ca+Na+K) × 100 in atom.% and show that it is the primary parameter controlling the composition of low-degree melts. All planetesimals with CI- H- or solar-like NaK# (50) produced low-density, high-viscosity melts that were rich in alkalis and silica regardless of the intrinsic fO2. On the other hand, planetesimals with lower NaK# (e.g. CM and CV-like) produced andesitic to basaltic partial melts under relatively oxidizing conditions (∆IW > -1). Experimental results are first examined in the context of previous studies (section 4.1). Partial melts are then compared to trachyandesite achondrites to constrain their formation conditions (section 4.2). Next, we propose that the albite/oligoclase-rich clasts found in polymict ureilites and some “silica-alkali-rich inclusions” of non-magmatic iron meteorites crystallized from the initial melts of chondritic materials with high NaK# (section 4.3). Finally, we estimate the physical properties of low-degree melts and review possible mechanisms of melt extraction (section 4.4). In a companion paper (Chapter 2), we re-evaluate the bulk composition and the melting behavior of the parent bodies of primitive achondrites. We show they were not depleted in alkalis (NaK# = 50) prior to melting and produced melts rich in SiO2, Al2O3 and alkali elements.

19 2. Experimental and analytical methods

2.1. MHC-pressure vessel experiments

Experiments were conducted in the MIT Experimental Petrology Laboratory in a sealed Molybdenum Hafnium Carbide pressure vessel (MHC-PV) inserted in a vertical Deltech furnace, following the same overall design as Walker and Grove (1993) and Singletary and Grove (2006). The experiments were run with up to six graphite capsules containing about 10 mg of different chondritic starting materials. Open graphite capsules with no lids were positioned vertically in a single Pt outer capsule, which was loosely crimped but not sealed. This arrangement was possible because of the small temperature gradient present in the base of the MHC pressure vessel (< 3 oC over the length of the Pt capsule). The Pt capsule was left open to allow for equilibration with CO gas, the pressure medium, which was used to set the fO2 to the C-CO buffer. At low pressure, the partial pressure of CO and the position of the C-CO buffer relative to the iron wustite (IW) buffer is highly sensitive to small changes in both pressure and temperature (French and Eugster, 1965).

Under our experimental conditions (1063-1300 ºC and 2-13 MPa), the fO2 varies from IW -0.8 to IW -2.3 (Table 1). About half of the experiments were first heated 40-130 ºC above the final temperature for two hours to increase the average grain size and facilitate electron-microprobe analysis of the run products. The experiments were then kept at the final temperature and pressure for up to 8 days (Table 1). The experiments were terminated by quickly pulling the pressure vessel out of the furnace, inverting it and hitting it with a wrench to ensure that the capsule dropped into the water- cooled head of the pressure vessel.

2.2. Gas mixing furnace experiments

Three additional experiments were conducted in a 1-atm gas mixing furnace at a higher fO2

(IW +0.5/0.8), conditions that cannot be attained at the C-CO buffer in a MHC-PV. The fO2 was fixed with a mixture of H2-CO2 and monitored with a ZrO2-CaO electrolyte cell. We used low gas flow rates (~0.1 ml/s) that have be shown to minimize Na-losses at the QFM buffer (IW +3.5) (Tormey et al., 1987; Juster et al., 1989). About 50 mg of the starting material was mixed with elvanol and attached on metallic loops made of Pt70Fe30 alloy to limit Fe exchange between the metal and the silicate material (Grove, 1982). To terminate experiments, an electric current was

20 run through the thin Pt wire holding the sample in the furnace, causing it to drop in a water bowl and quench.

2.3. Starting materials

All experiments were performed from synthetic equivalents of H, LL, CI, CM and CV chondrites. The bulk compositions of Lodders and Fegley (1998) were first renormalized without H, C, N and trace elements. Sulfur was then subtracted as FeS from the bulk compositions to avoid its reaction with the MHC alloy of the pressure vessel. While the sulfur concentration at sulfide saturation in silicate melts is large at very reducing conditions (fO2 < IW -4; Namur et al., 2016) and oxidizing conditions (fO2 > IW + 5; Jugo, 2009), it is at a minimum (200-600 ppm) at our experimental fO2 (IW -2.3 to IW +0.8; Righter et al., 2009). Therefore, subtracting sulfides from the bulk compositions has virtually no effect on the composition of silicate melts and silicate phase relations. Experiments simulate the batch melting of dry chondritic materials, which assumes that no hydrous minerals were present at the time of melting and that degassing was efficient during thermal metamorphism (e.g. Fu and Elkins-Tanton, 2014). MHC-PV experiments, but not gas- mixing experiments, are C-saturated by design. The starting compositions, reported in Table 2, were synthetized by mixing high purity oxides, carbonates (Na2CO3 and K2CO3) and Fe metal. The mixes were ground for 4 hours in an automatic mortar, decarbonated and pre-conditioned for three days in a gas-mixing furnace at 960 ºC and IW +2.5. The five starting compositions vary in terms of three main chemical parameters: the NaK# defined as the ratio (Na+K)/(Na+K+Ca)×100, the Mg# defined as the ratio Mg/(Mg+Fe) ×100 and the Mg/Si ratio, all in atomic percent. H, LL and CI chondrites have a higher NaK# (47- 50) than CM (33) and CV (25) chondrites. CI and LL chondrites have a higher Mg# (69-70) than CV, CM and H chondrites (58-62). Finally, ordinary chondrites (H and LL) have lower Mg/Si ratios than carbonaceous chondrites (CI, CM and CV).

2.4. Electron microprobe analyses

The experimental products were analyzed with the MIT Electron Probe Micro Analyzer (EPMA) JEOL-JXA-8200 SuperProbe. Crystalline phases were analyzed with a focused beam of 10 nA, 15 kV, and 40 sec counting times (20 sec for backgrounds), except plagioclase, which was analyzed with a defocused beam of 2 to 5 µm and lower current intensities (5 nA).

21 Glasses were analyzed with a 10 µm beam where glass pools were sufficiently large. Na was counted first, for 8 sec (4 sec for backgrounds) to limit Na-losses during analysis. For experiments performed at low temperature (<1150 ºC) and in which glass pools were small (Fig. 2), the beam diameter, current intensity and Na counting times were lowered to 5-2 µm, 5 nA and 6-4 sec. A few experiments with large glass pools were analyzed with varying current properties, counting times and beam diameters to ensure that all analysis settings produced consistent measurements. The phase proportions in experiments (i.e. modal composition in weight percent) were calculated by mass balancing the average EPMA analyses of the different phases against the bulk starting composition with the weighted linear regression function fitlm in Matlab.

3. Results

3.1. Approach of equilibrium

Several criteria show that experiments approached the conditions of thermodynamic equilibrium very closely. First, olivine and pyroxene have homogenous Mg# within individual $%&'(%&) experiments. We calculate the Fe-Mg exchange coefficients (K" , the Fe/Mg ratio in olivine over Fe/Mg in the glass) and compare them to the values predicted with the thermodynamic model $%&'(%&) of Toplis (2005). The K" in most experiments are within ± 0.03 of the predicted values (Table $%&'(%&) 3). Low-temperature experiments performed under reducing conditions display lower K" than experiments performed at higher temperature and more oxidizing conditions (Fig. 3). This is in agreement with previous experiments that contained olivine and melts rich in SiO2 and/or alkali elements (Toplis, 2005, and references therein).

For MHC-PV experiments, we calculate the target fO2 based on the position of the C-CO buffer at the experimental pressure and temperature (French and Eugster, 1965). We then compare the target fO2 to the one of a simple olivine-silica-iron (OSI) system (Nitsan, 1974) calculated as a function of the activity of fayalite in olivine in experiments (Williams, 1972) and assuming a value of one for silica activity. We find that the target fO2 is within ± 0.1 ΔIW of the fO2 of the

OSI system if the metal contains less than 10 wt.% Ni (Fig 4a). The target fO2 only deviates noticeably from the OSI system under oxidizing conditions, when the fraction of metal in experiments is small (< 5 wt.%) and the concentration of Ni in metal becomes significant (> 20

22 wt.%, Fig. 4b). This is due to the (erroneously) fixed Fe activity of 1 in the metal that is used to calculate the fO2 of the OSI system. In reality, the increasing Ni content in metal (Fig 4b) lowers the Fe activity and the actual fO2 is slightly more oxidizing than the fO2 predicted using the OSI system. Small deviations can arise due to minor experimental uncertainties including in the measurement of temperature (± 10 ºC) and pressure (± 0.3 MPa). In a few early experiments, the temperature was first brought 200-250 ºC above the final temperature for two hours and then dropped suddenly (in contrast to a super heat of 40 to 130 oC, see section 2.1). These were not included in this study but are discussed in the supplementary material. At such large degrees of superheating plagioclase nucleation was suppressed. Plagioclase nucleation delays in undercooled melts are well documented in the literature (e.g. Grove and Bence, 1979). After changing the ∆T from 200-250 ºC to ~40 to 130 ºC, plagioclase nucleated easily in all experiments. The phase relations in experiments that were first heated 100 ºC above the final temperature are consistent with duplicate isothermal experiments. We refer to the supplementary material for an extended discussion of those issues and a description of the additional precautions that were followed to perform successful experiments that show a close approach to equilibrium. Mass balance calculations indicate that alkali-losses in MHC-PV experiments were small.

In Table 4, we report the difference in Na2O contents between the recombined bulk composition and the target composition. MHC-PV experiments performed on H and LL compositions show the largest apparent loss of Na2O (average ∆ Na of 12 and 13 wt.%). However, other factors probably contribute to this apparent Na-loss such as the mobility of Na2O during EPMA analysis and uncertainties in the relative proportions of glass and plagioclase (Table 4). We interpret ∆ Na values as maximum alkali losses and show in section 4.1 that alkali-losses in MHC-PV experiments are probably less than 10 wt.% relative and much lower than in previous studies. On the other hand, we identified significant Na losses in experiments performed in the 1-atm gas- mixing furnaces. Mass balance calculations suggest that the relative loss could be of the order of 30 wt.% at low temperature (1075-1085 ºC) and as high as 40% at 1118 ºC. While Na losses can be minimized at QFM (∆IW + 3.5; Juster et al., 1989; Tormey et al., 1987), they are unavoidable at ∆IW +1 and lower fO2 where Na volatilization is faster. Therefore, the results of gas-mixing experiments are interpreted conservatively below.

23 3.2. Mineral compositions and proportions

MHC-PV experiments performed on all five starting compositions contain olivine, augite and orthopyroxene, plagioclase, Fe-Ni metal (Fig. 2, 5 and Table 4, 5) and traces of Cr-rich spinel (usually too small to be analyzed, Table S6 in the supplementary material) near the solidus. All five compositions contain around 10-12 wt.% plagioclase at the solidus (Fig. 6a), which is consistent with X-ray diffraction measurements of equilibrated ordinary chondrites (Dunn et al., 2010). The composition of plagioclase near the solidus varies as a function of the NaK# (Fig. 6b).

CI, H and LL compositions contain sodium-rich plagioclase (i.e. oligoclase, An20) while CM and

CV compositions contain intermediate plagioclase (i.e. andesine, An40 and labradorite, An55, respectively). Between the solidus and the temperature at which plagioclase melts out, the anorthite content increases continuously in all compositions (Fig. 6b). At IW-1, near-solidus experiments contain less than 10 wt.% pyroxene for all but one (LL) composition. Most of the pyroxene is augite but traces of orthopyroxene are also present (Table 4). The LL composition, which has the highest Si/(Fe+Mg) ratio, contains more than 20 wt.% orthopyroxene and 5 wt.% augite. The Mg# of pyroxene at IW -1 varies from 69 to 75, and the Mg# of olivine is slightly lower (61-70), with the most magnesian silicates found in the LL composition (Table 5). Between IW -1 and IW -2, as Fe2+ is progressively reduced to Fe0, the proportion of pyroxene near the solidus increases rapidly while the proportion of olivine decreases.

The Mg# of olivine and pyroxene also increase progressively with decreasing fO2 and reach 94 in olivine at IW -2.5 (Fig. 4a). At the most reducing conditions, orthopyroxene is more abundant (30- 40 wt.%) and experimental charges contain less augite (< 5 wt.%) near the solidus. At higher temperature, augite and orthopyroxene are replaced by a single pigeonite in the residual assemblage (> 1130 ºC) and orthopyroxene becomes the only pyroxene remaining above 1170- 1180 ºC (Fig. 5). Detailed compositional information of all phases is provided in the supplementary material (Table S2-S8).

3.3. Temperatures and reactions of melting

The position of the solidus is calculated by linear regression of the melt fractions (F) as a function of the experimental temperature and fO2 (see supplementary material). Because melting begins at low temperature in high-NaK# compositions (H, LL and CI), sub-solidus phase assemblages were only determined for the low-NaK# compositions (CM and CV). At IW -1.5, the

24 solidus temperature of H and LL compositions is estimated to be 1045 ± 10 ºC and the solidus of

CI is slightly lower (1035 ± 10 ºC). At the same fO2, the solidii of CM and CV chondrites, characterized by lower NaK# of 33 and 25, are located at 1090 ± 10 ºC and 1120 ± 10 ºC, respectively. The solidus temperature also increases by 30 to 50 ºC between IW-1 and IW-2 for all five starting compositions due to the increase in Mg# of ferromagnesian silicates (Fig. 5). With increasing temperature, plagioclase is the first phase to disappear. Plagioclase is fully consumed after 12-14 wt.% melting in H and LL compositions, 15-16 wt.% melting in CI, 17-19 wt.% melting in CM and ~25 wt.% melting in CV. Because the relative proportions of orthopyroxene, pigeonite and augite evolve rapidly with F and fO2, they were grouped together to calculate the melting reaction coefficents:

H: 0.72 plag + 0.43 px = 1 Liq + 0.15 oliv

LL: 0.78 plag + 0.43 px = 1 Liq + 0.21 oliv

CM: 0.56 plag + 0.67 px = 1 Liq + 0.23 oliv

CV: 0.41 plag + 0.76 px = 1 Liq + 0.22 oliv

Melting reactions are calculated by linear regression of the proportions of plagioclase, total pyroxene and olivine as a function of the melt fraction and forsterite content in olivine, which is a proxy for the fO2. While the fO2 has a strong effect on the relative proportion of olivine and pyroxene in the residues (section 3.2), it does not seem to significantly influence the coefficients of pyroxene and olivine in melting reactions. All phases are consumed (or produced in the case of olivine) in the same proportions between IW -1 and IW -2. After the disappearance of plagioclase, melting consumes only pyroxene while producing some olivine and py decreases more rapidly per unit of melt produced. At IW -1, pyroxene melts out completely after 15-25 wt.% of melting in all compositions but LL.

The melting reactions are calculated under the assumption that the Fo content in olivine remains constant during the first 15 wt.% of melting. Due to the partitioning of Fe-Mg between $%&'(%&) *+(%&) ferromagnesian silicates and the liquid (K" and K" = 0.25-0.35) this implies that a small fraction of metal (~0.1 unit) is consumed along with plagioclase and pyroxene. Alternatively, melting reaction coefficients can be calculated at constant fO2 (∆IW) or constant metal fraction.

25 All techniques provide similar results and we refer to the supplementary material for a brief discussion of the associated uncertainty.

3.4. Composition of experimental melts

The NaK# of the starting material is the primary parameter influencing the composition of partial melts. As discussed in the previous section, the compositions with high NaK# (LL, H and CI) have lower solidus temperatures, contain sodium-rich plagioclase, and are characterized by melting reactions with higher plagioclase coefficients. In consequence, their low-degree melts are rich in SiO2, Al2O3, K2O and Na2O but poor in MgO and CaO (low CaO/Al2O3 ratios; Fig. 7). On the other hand, chondritic compositions with lower NaK# (CM and CV) are characterized by higher solidus temperatures, intermediate plagioclase, and melting reactions with lower plagioclase coefficients. They produce melts with lower concentrations of SiO2, Al2O3, alkali elements and higher concentrations of MgO and CaO.

The fO2 at which melting occurs primarily impacts the FeO content of silicate melts. In H/LL experimental charges, the FeO content of low-degree melts increases from 3 to 5-7 and 10 wt.% between ∆ IW -2, -1 and +0.8. As a result, SiO2 contents decrease simultaneously from 67 to 64-62 and 60 wt.%, respectively. At a given melt fraction, melts of low-NaK# compositions (CM and CV) have higher FeO contents (3-4 to 16 wt.%). Similarly, as FeO contents increase,

SiO2 contents decrease from 60 wt.% to <50 wt.% between ∆ IW -2 and +0.8, although alkali losses are large at the latter fO2 (gas-mixing furnace, section 3.1 and 4.1). At low fO2, melts are slightly higher in MgO for a given melt fraction due to the higher solidus temperature. The P2O5 contents of melts are also a strong function of the fO2. They are highest in low-degree melts of

CM/CV compositions at IW -1 (Fig. 7f). Under more reducing conditions, all melts have low P2O5 contents and P is almost exclusively present as small phosphides or as P-C-rich metallic melts that could not be analyzed precisely (Fig. S3). Differences in P2O5 contents between the melts of low and high-NaK# compositions result from the higher solubility of P5+ in depolymerized melts with high concentrations of non-bridging oxygens (high NBO/T), where P is in tetrahedral coordination (Ryerson, 1985).

26 4. Discussion

4.1. Comparison with previous experiments

All previous melting experiments on chondritic compositions and aimed at constraining the partial melting of planetesimals were conducted in gas-mixing furnaces, at fO2 conditions under which alkali elements are volatile and are lost in the continuous flow of gas. Experiments performed by Jurewicz et al. (1993, 1995) are essentially free of moderately volatile elements. Feldstein et al. (2001) performed time series experiments at 1200 ºC and observed that 50 wt.% of the bulk Na2O, K2O and P2O5 had been lost within 1 h. Similarly, Gardner-Vandy et al. (2013) completely lost the alkali elements from their experiments. Finally, while the use of evacuated silica tubes by McCoy et al. (1999) seems to have slowed down evaporative losses, their experimental melts above 1300 ºC are also essentially free of alkali elements. Jurewicz et al. (1993) were aware of the rapid loss of alkali elements but interpreted their experimental melts as representing the partial melts of planetesimals that were depleted in MVE either prior to or during melting. Those early studies, which were highly influential for modeling melt migration, thermal evolution and differentiation in planetesimals, are responsible for the view that the partial melt of planetesimals were basaltic in composition. However, none of them are relevant for the onset of melting in planetesimals with chondritic concentrations of MVE, whether it is low (CV-like), moderate (CM-like) or high (H-, LL-, CI- or solar-like). The view that the partial melts of planetesimals were basaltic began to shift following the discovery of the first trachyandesite achondrite, GRA 06128/9, which contains 80 vol.% plagioclase An13 and motivated new experimental studies that attempted to retain MVE during experiments (Usui et al., 2015; Lunning et al., 2017). Usui et al. (2015) observed that melts rich in silica and alkali elements could be produced by melting a H chondrite at IW -1 but not at IW or IW +2. They estimated that alkali losses were small in their experiments but our own experiments performed on a H chondrite composition suggest otherwise. Our MHC-PV experiments contain a plagioclase that is much more Na-rich than the plagioclase in the experiments of Usui et al. (2015). Even in our gas-mixing experiments that lost 25-40 wt.% of Na and K, the plagioclase is more Na- rich than in their experiments (Fig. 6b). This indicates that Usui et al. (2015) underestimated alkali- losses and that the composition of their experimental melts at IW/IW+2 are impacted by this experimental artefact. Our experiments in gas-mixing furnaces suggest that melting an H-like chondritic material at IW +0.8 produce melts relatively rich in SiO2 (59 wt%) and alkalis (5.5

27 wt.%) even with the substantial loss of Na that we sustained, which is possibly as high as 25-40 wt.% relative. Another important issue with the experiments of Usui et al. (2015) is that they contain little plagioclase at IW and no plagioclase at IW -1, even at 1050 ºC. By comparison, our experiments on the H composition show that plagioclase is stable in the residue up to 1110 ºC at IW -1. Experiments performed during the development phase of this study, with an initial ∆T of 200-250 ºC identical to the ∆T used by Usui et al. (2015), highlighted plagioclase nucleation issues (see supplementary material). Therefore, the absence of plagioclase in many of the experiments of Usui et al. (2015) must result from a nucleation delay induced by undercooling (Gibb, 1974; Grove and Bence, 1979). Lunning et al. (2017) report a few melts relatively rich in alkali elements and silica produced by melting an R chondrite. The short duration and low temperature of their experiments (4 h and 1080-1120 ºC) appear to retain some alkali elements although precise mass balance calculations cannot be performed. Because their partial melts at IW/IW+1 were distinct from the melts of Usui et al. (2015) at IW/IW+2 (which have suffered alkali-losses), Lunning et al. (2017) proposed that “non-equilibrium melting”, a term used in reference to the short duration of their experiments, was more likely to yield alkali- and silica-rich melts under relatively oxidizing conditions. Our experiments show that an H composition and all high-NaK# compositions, which also include R chondrites, produce melts rich in alkali elements and silica over a wide range of fO2 conditions, including at IW +0.8. We also note that if the radioactive decay of 26Al is the main heat source driving melting in planetesimals (e.g. Hevey and Sanders, 2006; McCoy et al., 2006), the rapid heating of a chondrite is likely not analogous to the melting process that affected planetesimals. Even if partial melts were rapidly extracted from the interior of planetesimals (i.e. fractional or incremental batch melting), they were still initially produced in a phase assemblage in thermodynamic equilibrium. Experiments performed at higher pressure in piston-cylinder apparatus (1 GPa, IW +1) have shown that the first partial melts of dry and fertile spinel-lherzolites are also rich in SiO2 and alkalis in the Earth mantle (NaK# = 17; Baker et al., 1995) and the martian mantle (NaK# = 27;

Collinet et al., 2015). However, due to the lower NaK# of terrestrial planets, melts rich in SiO2 and alkalis are limited to the first 2-3 % (primitive Earth mantle) or the first 5 % of melting (primitive Martian mantle) as opposed to the first 15-20 % of melting for high-NaK# chondrites (50).

28 4.2. Evidence of alkali- and silica- rich melts in the meteorite record

4.2.1. Trachyandesite achondrites As brachinites, acapulcoites-lodranites and ureilites were growing in number in the meteorite collection (50, 165 and 550 samples at the time of writing), several workers recognized that many “primitive achondrites” had lost a silicate melt containing a plagioclase-component (Bild and Wasson, 1976; Warren and Kallemeyn, 1992; McCoy et al., 1997). However, it was not clear what the partial melts of these residues looked like. These missing melts were assumed to be basaltic, resembling eucrites erupted on the surface of Vesta. Therefore, the discovery of the first trachyandesite achondrites GRA 06128/9, that appear to derive from the same parent body as brachinites (Day et al., 2012, 2009; Shearer et al., 2008; Fig. 1b), came as a surprise. This meteorite was first referred to as “evolved” (Day et al., 2009), by analogy with extrusive terrestrial rocks which are produced by extensive fractional crystallization. Here, we review the mineralogy and bulk composition of GRA 06128/9, and of three other trachyandesite achondrites that have been subsequently discovered. We show that they represent near-liquid compositions similar to the partial melts produced by melting chondritic materials with high NaK# (e.g. CI, H and LL).

GRA 06128/9 contains 80 vol.% plagioclase An13, 9 vol.% olivine Fo41 and 9 vol.% pyroxene (both orthopyroxene and augite). ALM-A (Bischoff et al., 2014) was found in the Almahata Sitta strewn field, produced by the disintegration of asteroid 2008 TC3 and could represent the partial melt that was extracted from ureilites. It has a consistent O isotope signature and a bulk Mg# of 61 identical to low-degree melts in equilibrium with Fo85 olivine. ALM-A contains 70 vol.% plagioclase An10-30 (An15 in average), no olivine but a larger proportion of augite (20 vol.%), some pigeonite (5 vol.%) and traces of Ni-poor metal and quartz-normative glass rich in K2O. NWA 6698 (Bunch et al., 2011) and NWA 11575 (Agee et al., 2018) have oxygen isotope compositions that overlap with the field of LL chondrites (Fig. 1b). They have not yet been described in detail but contain 70 wt.% oligoclase (An17.4 in NWA 11575) and abundant pigeonite and sub-calcic augite. NWA 11575 also contains 5 wt.% of a mixture of a silica polymorph and potassic feldspar, which likely crystallized from a residual liquid. Another silica-rich achondrite, NWA 11119 (Srinivasan et al., 2018), differs significantly from trachyandesite achondrites, and from our experimental melts, in that it is depleted in alkali-elements and contain more abundant free-silica.

29 The bulk compositions of trachyandesite achondrites are similar to those of the partial melts produced in melting experiments of CI, H and LL chondritic compositions (Fig. 8). In a Total Alkali Silica diagram (Fig. 8a), ALM-A and NWA 11575 plot very close to the partial melts produced by 10 to 15 wt.% of melting of high-NaK# compositions. On the other hand, GRA

06218/9 is significantly poorer in SiO2 for a given concentration of alkalis. The concentrations of

Al2O3 and alkalis of all three trachyandesite achondrites are controlled by the quantity of oligoclase that they contain and are identical to experimental melts from high-NaK# compositions (Fig. 8b). In addition, NWA 11575 and GRA 06128 have the same concentrations of CaO while ALM-A has a slightly higher CaO content (Fig. 8c). NWA 11575 has a Na2O/K2O in the range of experimental melts (5-10) while both GRA 06128/9 and ALM-A are characterized by much higher ratios (20-

30) as a result of their low concentrations of K2O (Fig. 8d). In pseudo-ternary space, the composition of experimental melts, expressed as mineral components, only overlap with NWA 11575 (Fig. 9). Both ALM-A and GRA 06128 show a deficit in the quartz component relative to the partial melts of CI, H and LL compositions. In addition, GRA 06128/9, which contains 9 vol.% of olivine (12 wt.%), shows an excess in the olivine component and ALM-A, which contains 20 vol.% of augite (24 wt.%), shows an excess in the clinopyroxene component. To better evaluate how trachyandesite achondrites deviate from true liquid compositions, we compare their modal compositions to the expected crystallization products of experimental melts, estimated by using the algorithm MELTS (Ghiorso and Sack, 1995) to simulate equilibrium crystallization (Table 6). Low-degree melts (< 15 wt.%) of high-NaK# compositions (i.e. H, LL, CI) would crystallize 65-78 wt.% of oligoclase (80 % by volume) with an average composition of

An15-25, which is identical to the plagioclase of trachyandesite achondrites in proportion and composition. All such experimental melts are quartz normative and MELTS produces only minor olivine (0.5-2 wt.%), which is much less than the 12 wt.% present in GRA 06218/9. Low-degree melts are also expected to crystallize 5-12 wt.% of quartz/tridymite and are characterized by significant normative orthoclase. NWA 11575 contains a few percent of a SiO2 polymorph associated with K-feldspar (Agee et al., 2018). In contrast, ALM-A and particularly GRA 06128/9 are depleted in SiO2 (Fig. 8a, 9) and K2O (Fig. 8d) relative to low-degree melts. Finally, low- degree melts are not expected to crystallize more than 10-15 wt.% of augite, a proportion that is significantly smaller than the amount of augite in ALM-A (24 wt.%).

30 NWA 11575, which is closest in composition to the partial melts of high-NaK# chondrites

(Fig. 8-9) and contains free SiO2 and K-feldspar, could represent a true liquid composition. However, GRA 06128/9 and ALM-A probably represent partial cumulates containing accumulated olivine (GRA 06128/9) or augite (ALM-A) and having lost a residual liquid rich in

SiO2 and K2O. If we remove half of the olivine in GRA 06128/9 and half of the augite in ALM-A, and add back 30 wt.% of a residual melt estimated by equilibrium crystallization with MELTS (Table S1), we find that the resulting compositions are nearly identical to the ones of NWA 11575 and low-degree melts of high-NaK# chondrites (Fig. 8 and 9). Equilibrium crystallization of the partial melts of CV and CM chondrites (low NaK#) do not produce a large fraction of oligoclase. A fully crystallized low-degree partial melt (F = 5 wt.%) of a CM chondrite would produce 55 wt.% plagioclase An23 but slightly larger fractions of melt

(F = 10 wt.%) would only produce An35-40 plagioclase (andesine; Table 6). Of course, it could be argued that oligoclase is produced by fractional crystallization. In the absence of trends of igneous differentiation and considering that nearly-identical rocks were produced on different parent bodies, this scenario is currently unsupported (also see next paragraph). In Chapter 2, we demonstrate more explicitly that the parent bodies of primitive achondrites had a high NaK#, and therefore produced melts rich in SiO2 and alkalis.

4.2.2. Plagioclase-rich clasts in ureilites Before the trachyandesite ALM-A was recovered and identified as having crystallized in the ureilite parent body, several groups of plagioclase-rich felsic clasts had been observed in polymict ureilites (Cohen et al., 2004). A majority of them (48 wt.%) contain sodic plagioclase

(An0-25) with the other half split between a “labradoritic” group and a “calcitic” group that covers the rest of the plagioclase solid solution (Goodrich et al., 2017). Oxygen isotopes suggest that at least both the “albitic” and “labradoritic” groups are genetically related to the UPB (Kita et al., 2004). Experimental melts of high-NaK# compositions would crystallize abundant oligoclase

An15-25 and could produce some albite An0-10 by fractional crystallization. For this reason, we infer that the dominant group of clasts with sodic plagioclase are derived from the low-degree melts of a chondritic composition with high NaK# and that those melts were similar to the parental melt of ALM-A. If the UPB had been depleted in alkali elements (e.g. CM or CV), the population of “labradoritic” and “calcic” clasts should be dominant. Instead, those subsidiary groups likely

31 represent subsequent increments of melt extraction from a residue that became increasingly depleted in alkali-elements over time (Chapter 3).

Another felsic clast found in a polymict ureilite contains oligoclase, tridymite and a SiO2 and K2O-rich glass (Beard et al., 2015). Contrary to the clasts discussed above, this one did not originate on the UPB. It has a positive Δ17O that overlap with silica-rich inclusions in IIE and IVA iron meteorites. Although it comes from an unknown parent body, it could have crystallized from low-degree melts produced by melting high-NaK# chondritic material.

4.2.3. Silica-rich inclusions in iron meteorites Differentiated silica-rich inclusions are abundant in silicate-bearing or “non-magmatic” iron meteorites, which were not formed by fractional crystallization of a metallic melt and are therefore not thought to represent the core of planetesimals (Chabot and Haack, 2006). One of those group, IAB meteorites, likely comes from the same parent body as winonaites (Clayton and Mayeda, 1996). The Δ17O of silicates in IIE irons is identical to H chondrites or slightly lower, similar to highly reduced “HH” ordinary chondrites (McDermott et al., 2016), suggesting that, like ordinary chondrites, the parent body of IIE meteorites had a high NaK# of 50. Low-degree melts

(< 15 wt.%) of such planetesimals would have been rich in SiO2, Al2O3 and Na2O and very close in composition to our experimental melts of the H/LL compositions produced at IW -2 (CHS 43- 44, Table 3). The formation processes of IIE iron meteorites and the inclusions that they contain are complex and still debated. IIE irons are characterized by different ages of high-temperature equilibration, which indicate that they could have been heated by radioactive decay of 26Al soon after the accretion of the parent body but also by several later impacts. They appear to have experienced several events of partial melting, crystallization and mixing (Bogard et al., 2000; Ruzicka, 2014; McDermott et al., 2016; Wasson, 2017). Low-degree melts of high-NaK# compositions would crystallize a large volume of plagioclase An15-20 (50-75 wt.%; Table 6). Such low-degree melts could account for some of the most felsic silicate inclusions in IIE meteorites (e.g. Miles) and IAB meteorites (e.g. Udei Station), which contain up to 60 vol.% of plagioclase and up to 80 vol.% locally in “poikilitic areas” of Udei-3B (Ruzicka and Hutson, 2010). Some silicate inclusions also contain a mesostasis made of alkali-feldspar (antiperthite) and trydimite (Ruzicka and Hutson, 2010; Ruzicka, 2014). However, the majority of inclusions contain larger proportions of orthopyroxene and clinopyroxene and probably represent crystallization products of silicate melts produced by larger extent of melting

32 (20-30 wt.%). The Na-rich nature of plagioclase (< An18) in all differentiated silicate inclusions suggests that they formed by equilibrium crystallization of batch melts (Table 6). Melts produced after several increments of melt extraction would have crystallized intermediate plagioclase An30-

70 similar to the labradoritic clasts of polymict ureilites.

4.3. Physical properties of partial melts and mechanism of melt extraction

The physical properties of the low-degree melts of alkali-undepleted chondritic materials are radically different from those of the melts produced in earlier experiments (e.g. Jurewicz et al., 1993) or from those of basaltic liquids formed in planetesimals at higher degrees of melting and during the crystallization of magma oceans. Due to their high concentrations of SiO2, Al2O3 and alkali elements, and low concentrations of FeO and MgO, low-degree melts from H, LL and CI compositions are characterized by low densities (2420 to 2500 kg/m3) and high viscosities (104-5 Pa.s; Fig. 10 and Table 3). The density and viscosity of silicate melts were calculated using the models of Lange and Carmichael (1990) and Giordano et al. (2008), respectively. So far, studies of melt extraction in planetesimals assumed overwhelmingly that low-degree melts were basaltic and characterized by viscosities of 100-2 Pa.s and densities higher than 2800 kg/m3. Those models have to be re-evaluated in detail. Here, we discuss briefly some of the likely outcomes of adopting more realistic rheologies of alkali- and silica- rich melts. Partial melts can only migrate upward if they are buoyant relative to the solid matrix. Fu and Elkins-Tanton (2014) suggest that the density of partial melts was primarily controlled by their

FeO content and, therefore, the intrinsic fO2 of their parent bodies. In reality, while the effect of fO2 becomes significant for high degrees of melting (> 20 wt.%), the bulk NaK# of planetesimals is the dominant control on the buoyancy of partial melts near the solidus. Low-degree melts from high-NaK# planetesimals are always positively buoyant relative to both the mantle (3200-3400 kg/m3) and the crust, even if the macro-porosity of the latter is large (2800-3000 kg/m3). The only case for which low-degree melts could be denser than the crust are planetesimals poor in alkali elements (i.e. CM and CV compositions) that melted at a fO2 more oxidizing than IW -1. As discussed in the companion manuscript (Chapter 2), there is no evidence that such planetesimals were present in the inner solar system. While driven by buoyancy, melt extraction is also influenced by the permeability of the solid matrix. In two-phase flow theory, the permeability is related to the ability of the solid matrix

33 to compact and expel the melt it contains (e.g. McKenzie, 1985, 1984). Simple scaling analyses based on two-phase flow theory have been used to discuss whether melt migration is rapid relative to the rate of heat production by 26Al decay (Taylor et al., 1993; Lichtenberg et al., 2019) and whether removal of 26Al along with the melt could prevent the formation of magma oceans in planetesimals (e.g. Wilson and Keil, 2017). In the supplementary material, we reproduce this exercise using the rheology of SiO2 and alkali-rich melts and find that the melt extraction velocities of low-degree melts (F < 15 wt.%) are three orders of magnitude smaller than the extraction velocities of basaltic melts (2.3-230 m/Ma vs. 1.1-1100 km/Ma, respectively). During one half- life of 26Al (0.72 Ma), the melt would only migrate over a negligible distance compared to the radius of even the smallest planetesimals. This approach suggests that, despite their lower density, low-degree melts rich in alkali and silica would have been nearly impossible to extract due to their higher viscosity. If this was true, with sufficient 26Al present at the time of accretion, partial melting would have progressed as batch melting until the formation magma oceans. This is inconsistent with the occurrence of trachyandesite achondrites (section 4.2.1) and primitive achondrites derived from parent bodies not depleted in alkalis relative to the sun’s photosphere (Chapter 2). Therefore, the mechanism of melt extraction must have been more complex than a simple porous flow. On Earth, silicic magma with densities as high as 108-9 Pa.s could be efficiently separated from crystal mushes due to the buildup of a high internal excess pressure which result from the exsolution of H2O-rich gas during crystallization (Sisson and Bacon, 1999) and/or from the injection of additional CO2-H2O rich fluids (Bachmann and Bergantz, 2003). In planetesimals, most of the water is expected to degas prior to melting (Fu and Elkins-Tanton, 2014) but large excess pressures can develop due to the heating of trapped gas such as CO/CO2 or Cl (Muenow et al., 1992) and the production of CO by oxidation of C and reduction of FeO in Fe metal (i.e. smelting; Warren and Kallemeyn, 1992; Wilson et al., 2008). Even in the absence of a gas phase, low-degree melts rich in alkali and silica form a very important density contrast with the solid matrix (ρs/ρl = 1.32 vs. 1.1 for basaltic melts). When melting begins, the solid-liquid medium might not be able to accommodate this large increase in volume due to its high viscosity and large differential stresses could have appeared (Wilson and Keil, 2012). Because the excess pressure (> 100 MPa) would have been larger than the tensile strength of the matrix, small cracks would have formed and rapidly developed into veins and dykes of increasing sizes that drained partial melts

34 more efficiently than by simple porous flow (Sleep, 1988). This process allows rapid extraction of basaltic melts (1 Pa.s) with transit times of the order of years or months (Wilson et al., 2008) and could also account for the extraction of SiO2 and alkali-rich melts. More complex models of melt migration in planetesimals are needed to explain the formation of trachyandesite achondrites and primitive achondrites.

5. Conclusion

We show experimentally that the partial melts produced in planetesimals that were not depleted in alkali elements relative to the sun’s photosphere (e.g. CI, H and LL chondrites) were rich in SiO2, Al2O3 and alkali elements, regardless of the fO2 of melting. Low-degree partial melts can form trachyandesite achondrites by simple equilibrium crystallization, potentially with minor crystal accumulation (0-12 wt.%) and extraction of a residual liquid. The existence of trachyandesite achondrites originating from three parent bodies suggests that alkali-elements were not lost by evaporation simultaneously with the initial stage of silicate melting. Despite their high viscosity, low-degree melts were extracted from the interior of several planetesimals but might have migrated only locally in other parent bodies (e.g. IAB-winonaites and IIE irons). In the companion manuscript (Chapter 2), we show that the parent bodies of all primitive achondrites were alkali-undepleted and that many lost low degree melts (<15 wt.%) that were rich in alkalis,

Al2O3 and SiO2. Therefore, the formation and extraction of alkali- and silica- rich melts recorded by trachyandesite achondrites could have been widespread in the inner solar-system.

6. References

Agee C. B., Habermann M. A. and Ziegler K. (2018) Northwest Africa 11575: Unique Ungrouped Trachyandesite Achondrite. 49th Lunar Planet. Sci. Conf. 19-23 March, 2018, held Woodlands, Texas LPI Contrib. No. 2083, id.2226 49. Bachmann O. and Bergantz G. W. (2003) Rejuvenation of the Fish Canyon magma body: A window into the evolution of large-volume silicic magma systems. Geology 31, 789–792. Baker M. B., Hirschmann M. M., Ghiorso M. S. and Stolper E. M. (1995) Compositions of near-solidus peridotite melts from experiments and thermodynamic calculations. Nature 375, 308–311. Beard A. D., Downes H. and Chaussidon M. (2015) Petrology of a nonindigenous microgranitic clast in polymict ureilite EET 87720: Evidence for formation of evolved melt on an unknown parent body. Meteorit. Planet. Sci. 50, 1613–1623. Bild R. W. and Wasson J. T. (1976) The Lodran meteorite and its relationship to the ureilites. Mineral. Mag. 40, 721–735. Bischoff A., Horstmann M., Barrat J.-A., Chaussidon M., Pack A., Herwartz D., Ward D., Vollmer C. and Decker S. (2014) Trachyandesitic volcanism in the early Solar System. Proc. Natl. Acad. Sci. U. S. A. 111, 12689–92.

35 Bogard D. D., Garrison D. H. and McCoy T. J. (2000) Chronology and petrology of silicates from IIE iron meteorites: evidence of a complex parent body evolution. Geochim. Cosmochim. Acta 64, 2133–2154. Bunch T. E., Irving A. J., Wittke J. H. and Rumble D. (2011) Northwest Africa 6698: A High Temperature “Dioritic” Achondrite with Oxygen Composition in the LL Chondrite Field. MetSoc 74 74, 5224. Chabot N. L. and Haack H. (2006) Evolution of Asteroidal Cores. Meteorites early Sol. Syst. II, 747–771. Clayton R. N. and Mayeda T. K. (1996) Oxygen isotope studies of achondrites. Geochim. Cosmochim. Acta 60, 1999–2017. Cohen B. A., Goodrich C. A. and Keil K. (2004) Feldspathic clast populations in polymict ureilites: Stalking the missing basalts from the ureilite parent body. Geochim. Cosmochim. Acta 68, 4249– 4266. Collinet M., Médard E., Charlier B., Vander Auwera J. and Grove T. L. (2015) Melting of the primitive martian mantle at 0.5–2.2 GPa and the origin of basalts and alkaline rocks on Mars. Earth Planet. Sci. Lett. 427, 83–94. Consolmagno G. J. and Drake M. J. (1977) Composition and evolution of the eucrite parent body: evidence from rare earth elements. Geochim. Cosmochim. Acta 41, 1271–1282. Dauphas N. (2017) The isotopic nature of the Earth’s accreting material through time. Nature 541, 521– 524. Day J. M. D., Ash R. D., Liu Y., Bellucci J. J., Iii D. R., McDonough W. F., Walker R. J. and Taylor L. A. (2009) Early formation of evolved asteroidal crust. Nature 457, 179–182. Day J. M. D., Walker R. J., Ash R. D., Liu Y., Rumble D., Irving A. J., Goodrich C. A., Tait K., McDonough W. F. and Taylor L. A. (2012) Origin of felsic achondrites Graves Nunataks 06128 and 06129, and ultramafic brachinites and brachinite-like achondrites by partial melting of volatile-rich primitive parent bodies. Geochim. Cosmochim. Acta 81, 94–128. Dunn T. L., Cressey G., McSween H. Y. and McCoy T. J. (2010) Analysis of ordinary chondrites using powder X-ray diffraction: 1. Modal mineral abundances. Meteorit. Planet. Sci. 45, 127–138. Feldstein S. N., Jones R. H. and Papike J. J. (2001) Disequilibrium partial melting experiments on the Leedey L6 chondrite: Textural controls on melting processes. Meteorit. Planet. Sci. 36, 1421–1441. French B. M. and Eugster H. P. (1965) Experimental control of oxygen fugacities by graphite-gas equilibriums. J. Geophys. Res. 70, 1529–1539. Fu R. R. and Elkins-Tanton L. T. (2014) The fate of magmas in planetesimals and the retention of primitive chondritic crusts. Earth Planet. Sci. Lett. 390, 128–137. Gardner-Vandy K. G., Lauretta D. S. and McCoy T. J. (2013) A petrologic, thermodynamic and experimental study of brachinites: Partial melt residues of an R chondrite-like precursor. Geochim. Cosmochim. Acta 122, 36–57. Ghiorso M. S. and Sack R. O. (1995) Chemical mass transfer in magmatic processes IV. A revised and internally consistent thermodynamic model for the interpolation and extrapolation of liquid-solid equilibria in magmatic systems at elevated temperatures and pressures. Contrib. to Mineral. Petrol. 119, 197–212. Gibb F. G. (1974) Supercooling and the crystallization of plagioclase from a basaltic magma. Mineral. Mag. 39, 641–653. Giordano D., Russell J. K. and Dingwell D. B. (2008) Viscosity of magmatic liquids: A model. Goodrich C. A., Treiman A. H. and Boyle S. (2017) Melt formation and evolution on the ureilite parent body, as shown by feldspathic clasts in polymict ureilites. Lunar Planet. Sci., 1196. Greenwood R. C., Franchi I. A., Jambon A. and Buchanan P. C. (2005) Widespread magma oceans on asteroidal bodies in the early Solar System. Nature 435, 916–918. Grove T. L. (1982) Use of FePt alloys to eliminate the iron loss problem in 1 atmosphere gas mixing experiments: Theoretical and practical considerations. Contrib. to Mineral. Petrol. 78, 298–304. Grove T. L. and Bence A. E. (1979) Crystallization kinetics in a multiply saturated basalt magma - an experimental study of Luna 24 ferrobasalt. Lunar Planet. Sci. Conf. Houston, Texas 10th, 439–478.

36 Grove T. L. (1993) Corrections to expressions for calculating mineral components in “Origin of calc- alkaline series lavas at medicine lake volcano by fractionation, assimilation and mixing” and “Experimental petrology of normal MORB near the kane fracture zone: 22°-25°N, mi. Contrib. to Mineral. Petrol. 114, 422–424. Hevey P. J. and Sanders I. S. (2006) A model for planetesimal meltdown by 26 Al and its implications for meteorite parent bodies. Meteorit. Planet. Sci. 41, 95–106. Hunt A. C., Benedix G. K., Hammond S. J., Bland P. A., Rehkämper M., Kreissig K. and Strekopytov S. (2017) A geochemical study of the winonaites: Evidence for limited partial melting and constraints on the precursor composition. Geochim. Cosmochim. Acta 199, 13–30. Jugo P. J. (2009) Sulfur content at sulfide saturation in oxidized magmas. Geology 37, 415–418. Jurewicz A. J. G., Mittlefehldt D. W. and Jones J. H. (1993) Experimental partial melting of the Allende (CV) and Murchison (CM) chondrites and the origin of asteroidal basalts. Geochim. Cosmochim. Acta 57, 2123–2139. Jurewicz A. J. G., Mittlefehldt D. W. and Jones J. H. (1995) Experimental partial melting of the St. Severin (LL) and Lost City (H) chondrites. Geochim. Cosmochim. Acta 59, 391–408. Juster T. C., Grove T. L. and Perfit M. R. (1989) Experimental constraints on the generation of FeTi basalts, andesites, and rhyodacites at the Galapagos Spreading Center, 85°W and 95°W. J. Geophys. Res. Solid Earth 94, 9251–9274. Keil K. (2014) Brachinite meteorites: Partial melt residues from an FeO-rich asteroid. Chemie der Erde - Geochemistry 74, 311–329. Keil K. and McCoy T. J. (2018) Acapulcoite-lodranite meteorites: Ultramafic asteroidal partial melt residues. Chemie der Erde 78, 153–203. Kita N. T., Ikeda Y., Togashi S., Liu Y., Morishita Y. and Weisberg M. K. (2004) Origin of ureilites inferred from a SIMS oxygen isotopic and trace element study of clasts in the Dar al Gani 319 polymict ureilite. Geochim. Cosmochim. Acta 68, 4213–4235. Kruijer T. S., Touboul M., Fischer-Gödde M., Bermingham K. R., Walker R. J. and Kleine T. (2014) Protracted core formation and rapid accretion of protoplanets. Science 344, 1150–4. Lange R. L. and Carmichael I. S. E. (1990) Thermodynamic properties of silicate liquids with emphasis on density, thermal expansion and compressibility. Rev. Mineral. Geochemistry 24, 25–64. Lichtenberg T., Keller T., Katz R. F., Golabek G. J. and Gerya T. V. (2019) Magma ascent in planetesimals: Control by grain size. Earth Planet. Sci. Lett. 507, 154–165. Lodders K. and Fegley B. (1998) The planetary scientist’s companion., Oxford University Press. Lunning N. G., Gardner-Vandy K. G., Sosa E. S., McCoy T. J., Bullock E. S. and Corrigan C. M. (2017) Partial melting of oxidized planetesimals: An experimental study to test the formation of oligoclase- rich achondrites Graves Nunataks 06128 and 06129. Geochim. Cosmochim. Acta 214, 73–85. Mandler B. E. and Elkins-Tanton L. T. (2013) The origin of eucrites, diogenites, and olivine diogenites: Magma ocean crystallization and shallow magma chamber processes on Vesta. Meteorit. Planet. Sci., n/a-n/a. McCoy T. J., Dickinson T. L. and Lofgren G. E. (1999) Partial melting of the Indarch (EH4) meteorite: A textural, chemical, and phase relations view of melting and melt migration. Meteorit. Planet. Sci. 34, 735–746. McCoy T. J., Keil K., Clayton R. N., Mayeda T. K., Bogard D. D., Garrison D. H. and Wieler R. (1997) A petrologic and isotopic study of lodranites: Evidence for early formation as partial melt residues from heterogeneous precursors. Geochim. Cosmochim. Acta 61, 623–637. McCoy T. J., Mittlefehldt D. W. and Wilson L. (2006) Asteroid Differentiation. Meteorites early Sol. Syst. II, 733–745. McDermott K. H., Greenwood R. C., Scott E. R. D., Franchi I. A. and Anand M. (2016) Oxygen isotope and petrological study of silicate inclusions in IIE iron meteorites and their relationship with H chondrites. Geochim. Cosmochim. Acta 173, 97–113. McKenzie D. (1985) The extraction of magma from the crust and mantle. Earth Planet. Sci. Lett. 74, 81– 91.

37 McKenzie D. (1984) The Generation and Compaction of Partially Molten Rock. J. Petrol. 25, 713–765. Muenow D. W., Keil klaus and Wilson L. (1992) High-temperature mass spectrometric degassing of enstatite chondrites: Implications for pyroclastic volcanism on the aubrite parent body. Geochim. Cosmochim. Acta 56, 4267–4280. Namur O., Charlier B., Holtz F., Cartier C. and McCammon C. (2016) Sulfur solubility in reduced mafic silicate melts: Implications for the speciation and distribution of sulfur on Mercury. Earth Planet. Sci. Lett. 448, 102–114. Nitsan U. (1974) Stability field of olivine with respect to oxidation and reduction. J. Geophys. Res. 79, 706–711. Righter K. and Drake M. J. (1997) A magma ocean on Vesta: Core formation and petrogenesis of eucrites and diogenites. Meteorit. Planet. Sci. 32, 929–944. Righter K., Pando K. and Danielson L. R. (2009) Experimental evidence for sulfur-rich martian magmas: Implications for volcanism and surficial sulfur sources. Earth Planet. Sci. Lett. 288, 235–243. Available at: https://www.sciencedirect.com/science/article/pii/S0012821X09005615#fig1 [Accessed May 17, 2019]. Ruzicka A. (2014) Silicate-bearing iron meteorites and their implications for the evolution of asteroidal parent bodies. Chemie der Erde 74, 3–48. Ruzicka A. and Hutson M. (2010) Comparative petrology of silicates in the Udei Station (IAB) and Miles (IIE) iron meteorites: Implications for the origin of silicate-bearing irons. Geochim. Cosmochim. Acta 74, 394–433. Ryerson F. . (1985) Oxide solution mechanisms in silicate melts: Systematic variations in the activity coefficient of SiO2. Geochim. Cosmochim. Acta 49, 637–649. Schmitz B., Yin Q. Z., Sanborn M. E., Tassinari M., Caplan C. E. and Huss G. R. (2016) A new type of solar-system material recovered from Ordovician marine limestone. Nat. Commun. 7, 11851. Scott E. R. D., Taylor G. J. and Keil K. (1993) Origin of ureilite meteorites and implications for planetary accretion. Geophys. Res. Lett. 20, 415–418. Scott E. R. D., Krot A. N. and Sanders I. S. (2018) Isotopic Dichotomy among Meteorites and Its Bearing on the Protoplanetary Disk. Astrophys. J. 854, 164. Shearer C. K., Burger P. V., Neal C. R., Sharp Z., Borg L. E., Spivak-Birndorf L., Wadhwa M., Papike J. J., Karner J. M., Gaffney A. M., Shafer J., Weiss B. P., Geissman J. and Fernandes V. A. (2008) A unique glimpse into asteroidal melting processes in the early solar system from the Graves Nunatak 06128/06129 achondrites. Am. Mineral. 93, 1937–1940. Shearer C. K., Burger P. V., Neal C., Sharp Z., Spivak-Birndorf L., Borg L., Fernandes V. A., Papike J. J., Karner J., Wadhwa M., Gaffney A., Shafer J., Geissman J., Atudorei N.-V., Herd C., Weiss B. P., King P. L., Crowther S. A. and Gilmour J. D. (2010) Non-basaltic asteroidal magmatism during the earliest stages of solar system evolution: A view from Antarctic achondrites Graves Nunatak 06128 and 06129. Geochim. Cosmochim. Acta 74, 1172–1199. Singletary S. and Grove T. L. (2006) Experimental constraints on ureilite petrogenesis. Geochim. Cosmochim. Acta 70, 1291–1308. Sisson T. W. and Bacon C. R. (1999) Gas-driven filter pressing in magmas. Geology 27, 613. Sleep N. H. (1988) Tapping of melt by veins and dikes. J. Geophys. Res. Solid Earth 93, 10255–10272. Srinivasan P., Dunlap D. R., Agee C. B., Wadhwa M., Coleff D., Ziegler K., Zeigler R. and McCubbin F. M. (2018) Silica-rich volcanism in the early solar system dated at 4.565 Ga. Nat. Commun. 9, 3036. Stolper E. (1977) Experimental petrology of eucritic meteorites. Geochim. Cosmochim. Acta 41, 587–611. Taylor G. J., Keil K., McCoy T., Haack H. and Scott E. R. D. (1993) Asteroid Differentiation: Pyroclastic Volcanism to Magma Oceans. Meteoritics 28, 34–52. Toplis M. J. (2005) The thermodynamics of iron and magnesium partitioning between olivine and liquid: criteria for assessing and predicting equilibrium in natural and experimental systems. Contrib. to Mineral. Petrol. 149, 22–39. Tormey D. R., Grove T. L. and Bryan W. B. (1987) Experimental petrology of normal MORB near the Kane Fracture Zone: 22°–25° N, mid-Atlantic ridge. Contrib. to Mineral. Petrol. 96, 121–139.

38 Usui T., Jones J. H. and Mittlefehldt D. W. (2015) A partial melting study of an ordinary (H) chondrite composition with application to the unique achondrite Graves Nunataks 06128 and 06129. Meteorit. Planet. Sci. 50, 759–781. Walker D. and Grove T. (1993) Ureilite smelting. Meteoritics 28, 629–636. Warren P. H. and Kallemeyn G. W. (1992) Explosive volcanism and the graphite-oxygen fugacity buffer on the parent asteroid(s) of the ureilite meteorites. Icarus 100, 110–126. Wasson J. T. (2017) Formation of non-magmatic iron-meteorite group IIE. Geochim. Cosmochim. Acta 197, 396–416. Weiss B. P., Berdahl J. S., Elkins-Tanton L., Stanley S., Lima E. A. and Carporzen L. (2008) Magnetism on the Angrite Parent Body and the Early Differentiation of Planetesimals. Science (80-. ). 322, 713–716. Williams R. J. (1972) Activity-composition relations in the fayalite-forsterite solid solution between 900° and 1300°C at low pressures. Earth Planet. Sci. Lett. 15, 296–300. Wilson L., Goodrich C. A. and Van Orman J. A. (2008) Thermal evolution and physics of melt extraction on the ureilite parent body. Geochim. Cosmochim. Acta 72, 6154–6176. Wilson L. and Keil K. (2017) Arguments for the Non-existence of Magma Oceans in Asteroids. In Planetesimals (eds. L. T. Elkins-Tanton and B. P. Weiss). Cambridge University Press, Cambridge. pp. 159–179. Wilson L. and Keil K. (2012) Volcanic activity on differentiated asteroids: A review and analysis. Chemie der Erde - Geochemistry 72, 289–321.

39 Figures and tables

(a) 10 trachy- trachyte andesite

8 basaltic trachyandesite 6

dacite 4 andesite basalt

Na 2 O + K (wt.%) 2 eucrites 0 50 55 60 65 SiO2 (wt.%) 6 R eucrites (b) GRA 06128 NWA 11575 LL ALM A 4 NWA 11119 L H s

2 ureilite

δ 17 O (‰, SMOW) brac - ap r CCAM Terrestrial FL ac lod 0 0 2468 δ 18O (‰, SMOW)

Figure 1. (a) Total alkali-silica diagram: trachyandesite achondrites are trachy-andesitic in composition and contrast with basaltic achondrites (e.g. eucrites) and the single andesitic achondrite NWA 11119 (Srinivasan et al., 2018), which contains An87 and free SiO2. Light gray area is the range of calculated bulk compositions from Day et al. (2009). (b) Oxygen isotopes compositions: trachyandesite achondrites display different O isotopes and must have formed on different parent bodies. References for trachyandesite achondrites: Agee et al., 2018; Bischoff et al., 2014; Day et al., 2009. References for chondrites and primitive achondrites compiled in: Dauphas, 2017; Schmitz et al., 2016; Scott et al., 2018.

oliv

plag glass

aug

oliv

plag

aug subsolidus

CHS 35 (1079 ºC) - CM 10 µm CHS 35 (1079 ºC) - H 10 µm

plag oliv

glass oliv opx glass plag opx

CHS 44 (1161ºC) - CM 10 µm CHS 42 (1085 ºC) - H 10 µm

Figure 2. Backscattered electron images of experimental charges.

41 90 exp. MHC-PV exp. 1-atm 80 LL H 30 H CM Melt fraction (wt. %) CI 25 70 CM CV 60 20

KD = 0.35 50 15 Mg # of melt 40 10

5 30 KD = 0.25

20 60 65 70 75 80 85 90 95 Mg# of olivine (Fo content) Figure 3. Forsterite content in olivine as a function of the Mg # of the silicate melt in experiments. The Forsterite content (i.e. Mg2SiO4/(Mg2SiO4+Fe2SiO4) in mol.%) is equivalent to the Mg# (i.e. MgO/(MgO+FeO) in mol.% or Mg/(Mg+Fe) in atom.%). Low-degree melts (0-15 ,-(./ wt.%) in equilibrium with Fo85-95 olivine display lower K" (0.25-0.3) than low-degree melts in equilibrium with Fo60-75 olivine (0.3-0.35). The Mg# of silicate melts and olivine increase with decreasing experimental fO2 (∆ IW +0.8 at 1-atm; ∆ IW –1 to –2.5 in MHC-PV), also see Fig. 4. Color inside of symbol denotes the extent of melting or melt fraction of experiments (F, in wt.%).

42 (a) (b) 1 this study: Exp. (literature): 0.5 LL chondrite Jurewicz et al. 1995 (H-LL) H chondrite Usui et al. 2015 (H) Gardner-Vandy al. 2013 (R) CI chondrite IW) 0 CM chondrite -0.5 CV chondrite

-1

-1.5

-2 oxygen fugacity ( -2.5 60 70 80 90 0 20 40 60 Fo in olivine (mol.%) Ni concentration in metal (wt.%)

Figure 4. Forsterite content in olivine (a) and Ni concentration in metal (b) in experiments as a function of the experimental fO2, calculated with the CCO-buffer (French and Eugster, 1965) or measured (gas-mixing experiments; ∆IW > +0.5). The fO2 of the OSI equilibrium (Nitsan, 1974) is shown for comparison at 1100 ºC (black curve in (a)). Note that the OSI equilibrium slightly underestimates the actual experimental fO2 (CCO buffer) at more oxidizing conditions due to the increase in Ni activity (and decrease in metal Fe activity) of the metal.

43 37 37 LL 35 H CI -1 64 64 33 33 px out IW) 38 39 39 p px o i -1.2 g ut

63 o 63 63 u 57 p t -1.4 ig 68 pig ou

o 31 71 p u 57 46 60 l t 42 a 42 p g l -1.6 a 54 t o g u p 40 40 t o 70 l 44 a u 48 62 g o -1.8 t

41 u sol 41 t 43 s 43 27 o 27 i dus l -2 id 62 so u 59 s lid

oxygen fugacity ( 65 65 -2.2 u s 58 oliv 37 1100 1150 1200 1250 35 +pig CV 64 CM -1 +plag Temperature (¼C) 33 33 +metal

IW) Anorthite 38 39 38 39 -1.2 4 px p 70 63 i 63 content out g 6 o px u phases in in plag t out -1.4 60 (mol.%) s experiments: o 42 pi l id g u -1.6 p o s l opx aug 40 u a 50 g 44 t 44 -1.8 50 ou 50 pig plag t 41 40 43 27 43 + oliv s -2 p oli l + metal a d g u 30 oxygen fugacity ( out ± Cr-spinel -2.2 s

20 1050 1100 1150 1200 1250 1100 1150 1200 1250 Temperature (¼C) Temperature (¼C)

Figure 5. Temperature-oxygen fugacity phase diagrams of MHC-PV experiments performed on the five starting compositions. Experiments are identified by their number (e.g. 37 = CHS 37). The solidus and plagioclase-out temperatures are higher in CM and CV compositions and increase with decreasing fO2. Note the difference in plagioclase composition (An content) between high-NaK# compositions (LL, CI, CI) and low-NaK# compositions (CM and CV) and the decrease of An content with increasing temperature. Minor plagioclase in CHS_58 was too small to be analyzed.

44 (a) 12 MHC-PV exp. IW -0.8 / -2.2 10 LL H 8 CI CM 6 CV plag (wt.%) 4

0 05101520 melt fraction F (wt.%) (b) 80 gas mixing exp. 70 CM, IW +0,8; this study H; IW +0.8; this study 60 H; IW; Usui et al. 2015

40 loss of alkali

An content of plag 30 elements

10 1000 1050 1100 1150 Temperature (¼C)

Figure 6. (a) Plagioclase proportion in experiments as a function of the melt fraction. All compositions have 10-11 wt.% plagioclase at the solidus. Plagioclase melts out after 12-15 wt.% of melting in high-NaK# compositions (CI, H and LL), 17-18 wt.% of melting in CM and 25 wt.% of melting in CV. (b) An content of plagioclase as a function of the experimental temperature. The An content is higher for CV and CM compositions due to the depletion of alkali elements (and higher CAI proportions). The An content of plagioclase decreases linearly with decreasing temperature for all compositions until the solidus is reached. Under the solidus, the An content might continue to decrease slightly. All MHC-PV experiments show consistent trends and lower An contents relative to gas-mixing experiments, indicating that alkali-losses are much lower in MHC-PV experiments.

45 + 0.8 -1 -1.5 -2 -2.5

oxygen fugacity (Δ IW) in (a-c) and (f) Experimental melts 70 14 12 (c) (a) IW LL CM (b) -2 -1 H CV IW 12 10 H CI 65 IW 10 -1 CM 8 IW-2 IW-1 8 60 (wt.%) CM 6 2 -2 6 IW

FeO (wt.%) H CM MgO (wt.%) SiO IW-1 2 4 IW- -1 4 IW -2 55 IW H 2 IW-2 2 IW-1 50 0 0 H (d) (e) (f) IW-1 7 CM H 2.5 6 1.5 2 5 CM

4 1 (wt.%) 1.5 5 O (wt.%) O (wt.%) 2 O 2 2 K Na 3 P 1 2 0.5 0.5 1 IW -2 0 0 0 (g) (h) (i) chondritic ratio 10 16 H 0.8

14 (wt.%) 0.6

CM CM 3 O (wt.%) CM 2

3 6 O 2 12 0.4 CaO (wt.%) Al 4 CaO/Al

10 0.2 2 H H plag out plag out

8 0 0 0102030010203001020 30 Melt fraction (wt.%) Melt fraction (wt.%) Melt fraction (wt.%)

Figure 7. Composition of experimental melts as a function of the melt fraction. The influence of the fO2 is largest for FeO, P2O5 and SiO2 contents. The experimental fO2 is specified for SiO2,

FeO, MgO and P2O5 contents (color inside of symbols) and the limits of the shaded areas represent the composition of melts at IW -1 and IW -2. At a given melt fraction, more reducing melts represent a higher temperature (reflected by the higher MgO concentrations). Partial melts from low-NaK# compositions (CV and CM) are poorer in SiO2, alkali elements, and Al2O3 but richer in FeO, MgO and CaO than the partial melts from high-NaK# compositions (LL, H and CI).

(a) (b) (c)

trachyte olg olg 10 trachyandesite An 15-20 An 15-20 all trachyandesite achondrites 8 basaltic trachy- aug andesite A ALM-A L O (wt.%) M

2 6 -A dacite o 40 O + K 4 v F 2 oli Na 2 basalt andesite pig cpx ALM-A aug ALM-A opx 0 45 50 55 60 65 101214161820 2 4 6 8 10 12 SiO (wt.%) Al O (wt.%) CaO (wt.%) (d) 2 2 3 30 Exp. melts LL CM Trachyandesite GRA 06128 corr. achondrites ALM A 25 H CV corr. CI NWA 11575 20

O (wt.%) Melt fraction F (wt.%) 2

K 15

O / 51525 2 10 Na 5

0 0 0.5 1 1.5 2 K O (wt.%) 2 Figure 8. Comparison of trachyandesite achondrites and experimental melts. Trachyandesite achondrites, and especially NWA 11575, are similar in composition to experimental melts from high-NaK# compositions (H, LL and CI). White hexagons represent the mineral phases of trachyandesite achondrites. Concentrations in alkali elements and Al2O3 (b) are controlled by the large volume of plagioclase An15-20 or oligoclase (olg). The black line connects oligoclase to the origin (0 % alkalis and Al2O3). ALM-A and GRA 06128/9 are slightly poorer in SiO2 (a) and

K2O (d) than experimental melts and ALM-A is also richer in CaO (c). The estimated parental melt compositions (corr.; brown crosses) of trachyandesite achondrites are identical to experimental melts (see text).

47 plag plag 80 60 40 cpx 2 3 3 trachyandesite achondrites 1 3 ALM-A corr. 80 GRA 06128 corr. oliv qtz NWA 11575 Exp. melts LL plag cpx H 60 CI CM CV 40 qtz oliv qtz

Figure 9. Composition of experimental melts and trachyandesite achondrites expressed in oxygen mineral components (Tormey et al., 1987)(Grove, 1993) and projected in pseudo-ternary diagrams. NWA 11575 is closest to the composition of experimental melts. Once half of the olivine of GRA 06128/9 (6 wt.%) is subtracted (1), half of the cpx of ALM-A (12 wt.%) is subtracted (2), and 30 wt.% of a residual melt (Table S1, column 11) rich in SiO2 and K2O is added (3), the corrected compositions (corr.; brown crosses) are identical to the experimental melts of high NaK# compositions (LL, H and CI). Such compositions could represent the parental melts of trachyandesite achondrites.

48 oxygen fugacity (∆ IW) -1 -1.2 -1.4 -1.6 -1.8 -2 -2.2

2800 (a) 2750 )

3 2700

2650

2600

2550

melt density (kg/m 2500

2450 MHC-PV exp. 1-atm IW -0.8 / -2.2 gas-mix (b) LL IW +0,8 H CM CI 5 H CM CV , Pa.s)

melt viscosity (log 2

1 0102030 melt fraction (wt.%)

Figure 10. (a) Melt densities calculated with Lange and Carmichael (1990) and (b) melt viscosities calculated with Giordano et al. (2008) as a function of the melt fraction and fO2 (gray scale). Melts from high-NaK# compositions (LL, H and CI) are characterized by low densities and high viscosities. Open red and blue symbols correspond to the three experiments at IW +0.8 (gas-mixing furnace).

49 Table 1. Experimental conditions exp # device P (MPa) T (ºC) fO2 ∆ IW t (h) ∆ T* CHS 4 MHC-PV 5.2 1185 -13.74 -1.34 72 0 CHS 6 MHC-PV 5.2 1166 -13.58 -1.44 72 0 CHS 27 MHC-PV 5.6 1244 -13.49 -2.08 72 0 CHS 31 MHC-PV 7.2 1202 -13.49 -1.56 144 0 CHS 33 MHC-PV 5.2 1112 -14.32 -1.18 216 120 CHS 35 MHC-PV 4.5 1079 -14.67 -1.04 168 130 CHS 37 MHC-PV 4.1 1063 -14.85 -0.98 192 130 CHS 38 MHC-PV 3.4 1086 -14.83 -1.30 144 120 CHS 39 MHC-PV 5.7 1140 -14.05 -1.30 96 90 CHS 40 MHC-PV 2.4 1108 -14.94 -1.81 144 120 CHS 41 MHC-PV 2.4 1142 -14.80 -2.00 120 80 CHS 42 MHC-PV 2.1 1085 -15.30 -1.65 192 130 CHS 43 MHC-PV 2.8 1163 -14.50 -2.06 72 60 CHS 44 MHC-PV 3.7 1161 -14.30 -1.86 144 60 CHS 46 MHC-PV 6.2 1191 -13.68 -1.61 72 40 CHS 48 MHC-PV 4.1 1194 -14.02 -1.98 72 0 CHS 50 MHC-PV 7.2 1248 -13.26 -1.89 48 0 CHS 54 MHC-PV 3.8 1129 -14.46 -1.55 144 0 CHS 57 MHC-PV 3.8 1132 -14.30 -1.44 72 90 CHS 58 MHC-PV 1.7 1136 -15.07 -2.27 96 0 CHS 59 MHC-PV 1.3 1100 -15.51 -2.19 96 100 CHS 60 MHC-PV 2.8 1104 -14.89 -1.63 120 110 CHS 61 MHC-PV 13.1 1301 -12.53 -1.77 9 0 CHS 62 MHC-PV 6.4 1192 -13.60 -1.87 24 0 CHS 63 MHC-PV 9.1 1202 -13.30 -1.38 72 0 CHS 64 MHC-PV 13.1 1202 -13.00 -1.08 48 0 CHS 65 MHC-PV 3.1 1201 -14.21 -2.27 52 0 CHS 66 MHC-PV 3.4 1250 -13.88 -2.53 42 0 CHS 67 MHC-PV 13.3 1274 -12.65 -1.59 20 0 CHS 68 MHC-PV 5.2 1159 -12.49 -1.53 100 0 CHS 69 MHC-PV 5.9 1301 -13.21 -2.44 9 0 CHS 70 MHC-PV 3.8 1167 -14.22 -1.84 96 0 CHS 71 MHC-PV 7.9 1216 -13.34 -1.59 90 58 CH 5 1-atm 0.1 1073 -12.93 0.76 48 120 CH 8 1-atm 0.1 1076 -13.08 0.56 48 120 CH 10 1-atm 0.1 1118 -12.32 0.71 48 90 *∆ T (ºC) of initial 2 h step - final T (ºC), 0 = isothermal MHC-PV: Molybdenum Hafnium Carbide - pressure vessel 1-atm: 1-atm gas mixing furnace

50 Table 2. Starting materials of target composition H LL CI CM CV SiO2 37.63 43.29 40.78 37.34 36.83 TiO2 0.12 0.12 0.13 0.16 0.16 Al2O3 2.22 2.43 2.86 2.89 3.48 Cr2O3 0.51 0.59 0.69 0.55 0.56 FeO 30.80 21.80 21.78 28.53 27.75 MnO 0.31 0.36 0.44 0.33 0.22 MgO 23.78 26.66 28.17 25.62 26.00 CaO 1.78 2.10 2.30 2.41 2.82 Na2O 0.89 0.99 1.20 0.61 0.50 K2O 0.10 0.10 0.12 0.07 0.05 P2O5 0.26 0.22 0.39 0.30 0.28 NiO 1.48 1.00 1.13 1.20 1.25 Total 99.9 99.7 100.0 100.0 99.9

NaK# 49.2 47.5 50.1 33.0 25.5 Mg/Si 0.94 0.92 1.03 1.02 1.05 Mg# 57.9 68.6 69.7 61.5 62.5

Compositions modified from Lodders and Fegley (1998), see text

51 Table 3. Composition of experimental melts

1 2 3 4 5 log T (ºC) Δ IW SiO2 TiO2 Al2O3 Cr2O3 FeO MnO MgO CaO Na2O K2O P2O5 NiO total n F KD KD T05 ρ µ6 CI CHS 69 1301 -2.44 61.2 0.39 9.05 0.76 4.41 0.52 12.6 7.07 3.62 0.32 0.05 0 99.55 4 31.3 0.32 0.32 2.53 1.74 CI CHS 67 1274 -1.59 57.3 0.34 8.05 0.75 12.1 0.48 11.0 6.12 3.30 0.26 0.27 0 99.38 6 34.1 0.33 0.33 2.64 1.52 CI CHS 66 1250 -2.53 62.1 0.45 12.0 0.56 2.88 0.45 8.72 7.68 4.70 0.39 0.07 0 99.9 5 23.4 0.28 0.29 2.50 2.47 CI CHS 63 1202 -1.38 58.0 0.44 10.7 0.53 10.3 0.36 6.91 7.66 4.36 0.35 0.42 0.02 99.8 6 27.3 0.30 0.31 2.60 2.35 CI CHS 64 1202 -1.08 55.1 0.45 9.81 0.48 14.1 0.35 6.50 8.33 3.68 0.28 1.09 0.00 99.2 4 28.1 0.29 0.33 2.67 2.06 CI CHS 65 1201 -2.27 62.7 0.49 13.9 0.44 3.22 0.29 6.00 6.94 5.57 0.47 0.00 0 99.33 3 20 0.27 0.28 2.40 3.13 CI CHS 62 1192 -1.87 59.8 0.49 12.6 0.38 6.92 0.30 6.32 7.23 5.26 0.43 0.21 0.02 98.3 5 22.4 0.30 0.29 2.54 2.73 CI CHS 70 1167 -1.84 61.7 0.50 14.7 0.29 5.15 0.22 4.57 6.15 5.92 0.61 0.22 0 100.6 2 18.5 0.27 0.28 2.50 3.39 CI CHS 68 1159 -1.53 58.9 0.47 12.6 0.36 8.41 0.24 5.07 7.39 5.24 0.54 0.81 0 100.6 5 22.6 0.29 0.29 2.56 2.99 CI CHS 57 1132 -1.44 59.4 0.49 13.6 0.15 8.26 0.22 4.19 7.13 5.24 0.58 0.70 0.05 100.5 7 20.2 0.29 0.27 2.47 4.22 CI CHS 60 1104 -1.63 62.4 0.52 16.0 0.09 5.05 0.17 3.09 4.13 6.78 0.82 0.86 0.06 98.4 5 14.4 0.29 0.30 2.55 3.39 CI CHS 59 1100 -2.19 65.5 0.45 17.6 0.06 2.64 0.13 2.41 3.10 6.98 0.99 0.12 0.04 101.4 4 11.2 0.23 0.25 2.42 4.83

H CHS 61 1301 -1.77 58.9 0.38 8.34 0.55 9.70 0.37 12.5 5.75 3.16 0.30 0.06 0.02 100.0 5 28.3 0.34 0.33 2.60 1.46 H CHS 69 1301 -2.44 61.3 0.44 9.78 0.60 4.47 0.37 11.8 7.16 3.62 0.42 0.02 0.00 99.2 4 22.4 0.34 0.32 2.53 1.81 H CHS 65 1201 -2.27 63.3 0.55 14.19 0.31 3.39 0.15 5.6 6.36 5.53 0.61 0.03 0.00 100.4 3 14.8 0.26 0.27 2.40 3.25 H CHS 27 1244 -2.08 60.9 0.50 12.8 0.41 5.87 0.30 7.17 7.15 4.32 0.48 0.03 0.04 100.0 4 16.5 0.27 0.30 2.53 2.48 H CHS 63 1202 -1.38 58.2 0.52 11.6 0.39 10.9 0.29 6.61 7.09 3.87 0.37 0.20 0.02 100.0 3 18.8 0.30 0.32 2.61 2.42 H CHS 62 1192 -1.87 60.3 0.48 13.7 0.34 7.01 0.24 5.61 6.59 5.10 0.51 0.11 0.03 97.4 4 14.2 0.26 0.29 2.53 2.88 H CHS 46 1191 -1.61 58.3 0.45 11.8 0.36 10.5 0.25 6.05 7.00 4.58 0.43 0.21 0.00 101.1 6 18.8 0.31 0.31 2.59 2.51 H CHS 43 1163 -2.06 63.4 0.57 15.7 0.17 3.32 0.18 4.56 5.47 5.67 0.95 0.02 0.00 100.1 6 12.1 0.27 0.27 2.46 3.75

52 H CHS 44 1161 -1.86 62.5 0.60 15.1 0.21 4.71 0.20 3.88 6.26 5.61 0.79 0.07 0.01 100.5 13 13.1 0.25 0.27 2.49 3.69 H CHS 41 1142 -2.00 63.5 0.58 16.5 0.11 3.92 0.18 3.53 4.80 6.23 1.08 0.14 0.04 98.6 6 9.9 0.24 0.26 2.46 4.01 H CHS 39 1140 -1.30 58.6 0.56 13.1 0.18 10.2 0.17 4.22 6.55 4.79 0.79 0.93 0.01 100.0 5 16.6 0.30 0.31 2.58 3.20 H CHS 57 1132 -1.44 58.7 0.53 14.1 0.19 8.84 0.17 4.04 7.27 5.09 0.59 0.46 0.03 101.4 8 15.6 0.30 0.30 2.56 3.33 H CHS 33 1112 -1.18 58.5 0.58 14.0 0.07 10.5 0.14 3.35 6.33 4.61 0.79 1.02 0.01 100.5 7 12.7 0.31 0.31 2.58 3.57 H CHS 40 1108 -1.81 64.5 0.52 16.5 0.09 3.67 0.15 3.14 4.01 6.02 1.15 0.27 0.02 100.9 4 7.9 0.29 0.26 2.45 4.54 H CHS 42 1085 -1.65 66.5 0.50 16.2 0.07 3.31 0.09 2.08 2.70 6.38 1.84 0.33 0.00 101.3 4 5.0 0.25 0.25 2.42 5.09 H CHS 35 1079 -1.04 62.6 0.55 15.3 0.07 7.62 0.11 2.21 3.68 5.30 1.51 1.05 0.01 98.0 4 5.3 0.32 0.29 2.50 4.59 H CH 8* 1076 0.59 59.5 0.51 14.7 0.07 10.6 0.13 2.86 5.91 4.69 0.56 0.57 0.00 100.3 7 13.4 0.32 0.32 2.58 4.04 H CH 5* 1073 0.79 59.3 0.53 14.8 0.05 10.3 0.13 2.89 6.00 4.55 0.95 1.14 0.08 100.7 6 7.7 0.34 0.32 2.76 2.45 1 2 3 4 Original total of n analyses, compositions in the table are normalized to 100; melt fraction; KD Fe-Mg oliv-liq; KD predicted with Toplis et al. (2005) 5melt density in g/m3 (Lange and Carmichael, 1990); 6log of melt viscosity in Pa.s (Giordano et al., 2008)

52 Table 3. Composition of experimental melts (continued)

1 2 3 4 5 log T (ºC) Δ IW SiO2 TiO2 Al2O3 Cr2O3 FeO MnO MgO CaO Na2O K2O P2O5 NiO total n F KD KD T05 ρ µ6

LL CHIS 27 1244 -2.08 60.8 0.47 12.2 0.40 6.08 0.29 7.45 7.08 4.70 0.50 0.06 0.01 99.7 7 18.9 0.30 0.30 2.53 2.40 LL CHIS 63 1202 -1.38 57.1 0.42 10.6 0.40 12.0 0.28 7.18 7.50 3.98 0.35 0.34 0.02 99.2 4 21.7 0.31 0.32 2.63 2.20 LL CHIS 64 1202 -1.08 56.0 0.39 9.82 0.36 14.6 0.28 6.75 7.60 3.77 0.31 0.54 0.00 100.1 5 24.9 0.31 0.33 2.67 2.06

LL CHIS 43 1163 -2.06 62.7 0.45 15.6 0.21 3.99 0.18 4.80 6.24 5.48 0.56 0.04 0.00 101.4 7 15.3 0.27 0.27 2.48 3.59 LL CHIS 41 1142 -2.00 64.5 0.46 16.3 0.16 3.48 0.16 3.52 4.56 6.01 0.90 0.13 0.02 99.6 3 11.6 0.27 0.26 2.45 4.14 LL CHIS 39 1140 -1.30 58.3 0.36 13.6 0.19 9.57 0.17 4.72 6.65 5.06 0.51 0.84 0.04 100.3 4 17.4 0.33 0.30 2.57 3.09 LL CHIS 33 1112 -1.18 60.6 0.46 15.4 0.07 8.04 0.14 3.20 5.69 4.99 0.69 1.02 0.00 99.7 4 14.9 0.30 0.30 2.53 3.88 LL CHIS 40 1108 -1.81 66.0 0.45 16.8 0.10 3.66 0.14 3.03 3.93 6.87 0.83 0.38 0.03 98.4 5 10.9 0.28 0.26 2.44 4.49 LL CHIS 42 1085 -1.65 66.9 0.40 16.0 0.03 3.61 0.08 2.32 2.66 6.77 1.60 0.00 0.00 98.5 2 5.5 0.26 0.26 2.42 4.97 LL CHIS 37 1063 -0.98 64.5 0.39 16.7 0.07 5.14 0.08 2.02 3.24 6.37 1.11 0.84 0.01 100.9 13 7.6 0.32 0.27 2.46 5.01

CM CHIS 50 1248 -1.89 56.5 0.65 12.3 0.52 8.09 0.39 9.64 9.27 2.37 0.21 0.08 0.03 100.5 3 22.7 0.30 0.33 2.62 1.86 CM CHIS 27 1244 -2.08 57.6 0.67 12.8 0.47 6.63 0.35 8.96 8.97 3.11 0.31 0.02 0.03 99.4 10 21.4 0.31 0.32 2.58 2.06 CM CHIS 63 1202 -1.38 53.6 0.67 11.9 0.47 13.0 0.32 8.32 8.95 2.25 0.21 0.29 0.02 98.9 5 23.9 0.33 0.33 2.69 1.88 CM CHIS 64 1202 -1.08 50.9 0.62 10.5 0.39 18.0 0.31 8.20 8.37 1.90 0.21 0.64 0.00 99.8 4 27.9 0.35 0.34 2.77 1.56 CM CHIS 43 1163 -2.06 59.2 0.79 15.9 0.26 4.34 0.23 6.19 8.36 4.20 0.64 0.05 0.00 100.3 8 10.4 0.29 0.29 2.53 3.19 CM CHIS 44 1161 -1.86 57.9 0.69 15.7 0.31 6.28 0.26 5.38 9.74 3.84 0.43 0.16 0.03 99.5 8 12.7 0.26 0.30 2.57 3.03 CM CHIS 39 1140 -1.30 53.7 0.61 13.8 0.21 11.8 0.22 5.52 9.58 3.14 0.46 0.97 0.03 99.1 6 15.0 0.32 0.32 2.66 2.59 CM CHIS 33 1112 -1.18 55.7 1.05 13.3 0.15 11.2 0.17 4.01 7.53 4.27 1.05 1.89 0.04 100.4 6 5.2 0.33 0.31 2.61 3.28 CM CH 10* 1118 0.75 51.4 0.74 12.3 0.12 16.5 0.24 5.46 10.6 1.87 0.27 0.52 0.71 99.0 1 12.0 0.33 0.34 2.76 2.45

53 CV CHIS 50 1248 -1.89 55.8 0.46 12.6 0.58 8.2 0.21 10.33 9.59 1.90 0.19 0.07 0.11 100.8 5 27.3 0.33 0.33 2.63 1.72 CV CHIS 63 1202 -1.38 53.7 0.48 12.1 0.47 12.4 0.20 8.60 9.48 1.94 0.17 0.34 0.00 99.6 5 28.1 0.34 0.34 2.69 1.87 CV CHIS64 1202 -1.08 49.4 0.55 12.6 0.41 17.0 0.21 7.25 10.3 1.65 0.16 0.71 0.00 99.8 4 27.1 0.29 0.33 2.77 1.41 CV CHIS 39 1140 -1.30 53.4 0.60 13.1 0.15 12.5 0.13 5.76 9.64 3.04 0.37 1.26 0.02 98.8 4 9.6 0.34 0.33 2.67 2.53 CV CHIS 33 1112 -1.18 55.2 0.90 12.5 0.16 11.8 0.12 3.99 7.93 3.52 0.90 2.98 0.00 100.5 2 3.9 0.30 0.32 2.62 3.29 1 2 3 4 Original total of n analyses, compositions in the table are normalized to 100; melt fraction; KD Fe-Mg oliv-liq; KD predicted with Toplis et al. (2005) 5melt density in g/m3 (Lange and Carmichael, 1990); 6log of melt viscosity in Pa.s (Giordano et al., 2008)

53 Table 4. Phase proportions in experiments exp T (ºC) ∆ IW F oliv opx pig aug plag metal ∆ Na1 CI CHS 69 1301 -2.44 31.29 0.6 40.2 0.6 9.3 1 17.6 0.4 -4 CHS 67 1274 -1.59 34.1 1.3 55.3 1.4 8.8 0.3 -3 CHS 66 1250 -2.53 23.36 0.8 35.5 0.9 20.9 1.4 18.5 0.1 -6 CHS 71 1216 -1.59 26* 54.4* 7.7* 10.4* CHS 64 1202 -1.08 28.1 1.3 67.6 1.9 2.9 1.1 -14 CHS 63 1202 -1.38 27.3 0.7 61.4 2.9 2.3 3.6 8.4 0.5 1 CHS 65 1201 -2.27 20 1.1 38.3 4.1 23 4.8 17.1 0.2 -5 CHS 62 1192 -1.87 22.4 0.6 48.3 2.5 16.6 3.5 12.6 0.4 0 CHS 70 1167 -1.84 18.54 0.7 47.5 0.9 19.7 1.3 12.4 0.1 -4 CHS 68 1159 -1.53 22.55 1.3 63.3 1.6 5.6 2.5 7 0.7 -2 CHS 58 1136 -2.27 x x x x x CHS 57 1132 -1.44 20.9 0.6 59.6 1.8 13.7 1.9 6.0 0.6 -6 CHS 54 1129 -1.55 x x x x CHS 60 1104 -1.63 14.4 0.6 58.0 2.1 7.5 2.3 8.7 0.9 1.5 0.5 8.3 1.2 -5 CHS 59 1100 -2.19 11.2 1.0 44.4 8.0 18.0 9.0 7.4 2.8 2.9 0.9 14.3 0.8 -11 H CHS 69 1301 -2.44 22.4 0.6 24.3 0.6 24.1 1.0 27.2 0.5 -6 CHS 61 1301 -1.77 28.3 0.7 34.4 5.2 12.9 6.5 22.7 0.7 1 CHS 27 1244 -2.08 16.5 0.5 25.8 1.6 30.8 1.5 26.3 0.4 -20 CHS 63 1202 -1.38 18.8 0.6 42.2 4.4 19.0 5.5 19.2 0.8 -15 CHS 65 1201 -2.27 14.8 1.4 23.9 5.6 32.8 6.6 26.6 1.2 -5 CHS 48 1194 -1.98 x x x x CHS 62 1192 -1.87 14.2 0.5 29.8 2.8 32.1 3.9 23.5 0.4 -11 CHS 46 1191 -1.61 18.8 0.3 46.7 1.5 15.8 1.7 18.4 0.3 -13 CHS 43 1163 -2.06 12.1 0.7 28.3 3.6 31.0 3.6 27.4 0.7 -18 CHS 44 1161 -1.86 13.1 0.4 31.9 3.3 29.7 4.1 24.4 1.3 -13 CHS 41 1142 -2.00 9.9 1.2 29.0 6.1 29.2 10.4 4.3 4.4 1.7 1.0 25.2 0.8 -14 CHS 39 1140 -1.30 16.6 0.5 55.2 1.9 13.8 2.3 13.9 1.0 -9 CHS 57 1132 -1.44 15.6 0.4 56.3 2.4 13.1 2.0 13.2 0.6 -14 CHS 33 1112 -1.18 12.7 1.1 65.4 2.2 11.4 2.5 1.6 8.5 0.3 -19 CHS 40 1108 -1.81 7.9 1.1 31.4 4.0 30.2 5.5 3.0 1.2 3.7 1.4 23.5 0.3 -17 CHS 42 1085 -1.65 5.0 0.7 34.8 5.3 28.0 7.5 3.5 0.7 5.6 2.0 22.4 1.0 -11 CHS 35 1079 -1.04 5.3 0.7 73.8 4.4 1.8 4.9 6.3 1.0 6.1 0.7 6.0 0.7 -11 CH 8* 1076 0.56 13.4 79.0 4.4 1.4 0.4 -27 CH 5* 1073 0.76 11.1 3.6 82.1 2.0 1.0 3.6 1.9 2.1 1.0 1.0 -29 CHS 37 1063 -0.98 5.0 1.6 76.0 1.0 0.5 7.0 1.1 6.2 4.8 0.4 -16 Numbers in italic represent the standard errors 1 Maximum Na-loss, see text for detail [(Na2O in recombined bulk composition) / (Na2O in target compostion * 100)] - 100 x: melt not analyzable *mass balance calculated with quenched modified melt

54 Table 4. Phase proportions in experiments (continued) exp T (ºC) ∆ IW F oliv opx pig aug plag metal ∆ Na1 LL CHS 27 1244 -2.08 18.9 0.3 29.2 0.8 35.9 0.8 15.8 0.0 -8 CHS 64 1202 -1.08 24.9 0.7 60.1 3.8 12.0 3.3 2.7 1.0 -8 CHS 63 1202 -1.38 21.7 1.0 48.8 3.9 22.2 3.8 7.1 1.1 -12 CHS 31 1202 -1.56 18.2 0.7 43.2 4.3 27.8 5.1 10.4 0.3 -13 CHS 43 1163 -2.06 15.3 0.8 26.4 4.4 42.5 5.4 3.7 14.9 0.5 -14 CHS 41 1142 -2.00 11.6 1.6 29.8 7.1 38.0 11.4 3.6 3.9 1.2 1.2 15.1 0.7 -14 CHS 39 1140 -1.30 17.4 0.7 52.4 2.5 24.4 2.9 4.8 0.4 -8 CHS 33 1112 -1.18 14.9 2.2 55.9 3.8 13.3 12.0 6.8 1.4 0.2 2.8 3.0 0.4 -21 CHS 40 1108 -1.81 10.9 0.7 33.3 2.0 36.3 3.0 5.3 0.6 0.7 0.7 13.1 0.8 -9 CHS 38 1086 -1.30 6.8 2.4 49.7 3.1 25.3 4.1 7.1 1.5 4.9 1.7 5.9 0.2 -21 CHS 42 1085 -1.65 5.5 0.9 36.9 7.2 35.7 10.2 3.0 0.9 6.3 2.7 11.9 0.5 -13 CHS 37 1063 -0.98 7.6 2.2 57.3 3.8 21.4 6.8 6.8 1.4 4.6 2.8 2.3 0.2 -6 CM CHS 50 1248 -1.89 22.7 0.5 39.8 4.0 14.9 4.9 22.0 0.6 -11 CHS 27 1244 -2.08 21.4 0.5 35.8 2.0 19.9 2.6 22.8 0.2 9 CHS 64 1202 -1.08 27.9 0.6 65.2 0.9 7.0 0.1 -15 CHS 63 1202 -1.38 23.9 0.4 53.4 0.9 8.0 1.0 14.3 0.0 -12 CHS 6 1185 -1.44 x x x x CHS 4 1166 -1.34 x x x x CHS 43 1163 -2.06 10.4 0.8 33.4 0.7 28.4 0.9 3.4 0.5 24.3 0.3 -2 CHS 44 1161 -1.86 12.7 0.7 41.1 1.0 21.4 1.2 2.7 0.4 21.2 0.0 1 CHS 41 1142 -2.00 0.0 33.4 29.4 4.0 10.2 22.8 0.1 CHS 39 1140 -1.30 15.0 0.9 62.3 0.4 8.7 0.8 2.6 0.5 11.1 0.0 -1 CHS 33 1112 -1.18 5.2 0.3 74.3 0.4 0.4 6.7 0.2 7.5 0.6 5.7 0.8 -1 CHS 40 1108 -1.81 2.2 37.5 25.4 6.3 7.8 20.5 0.4 CHS 38 1086 -1.30 CHS 42 1085 -1.65 0.0 40.0 2.1 23.3 3.0 6.6 0.2 10.3 0.8 19.2 0.3 CHS 35 1079 -1.04 0.0 76.9 1.8 0.9 0.4 6.6 1.9 9.8 0.3 5.0 0.8 CHS 37 1063 -0.98 0.0 76.9 3.8 1.3 0.6 6.9 4.3 10.4 0.2 4.3 0.2 CH 10* 1118 0.71 12.0 80.4 1.3 4.6 1.0 0.2 -40 CV CHS 50 1248 -1.89 27.3 0.1 46.6 0.9 4.9 1.2 19.5 0.2 2.5 CHS 64 1202 -1.08 27.1 0.3 64.0 1.4 8.4 0.1 -11 CHS 63 1202 -1.38 28.1 0.5 57.0 1.5 0.5 1.9 14.0 0.8 8 CHS 43 1163 -2.06 0.0 31.3 31.0 3.1 10.7 23.3 0.6 CHS 44 1161 -1.86 0.0 37.6 28.8 2.2 10.4 20.5 0.9 CHS 39 1140 -1.30 9.6 3.5 62.4 2.8 12.1 4.1 6.8 2.3 9.6 0.1 12 CHS 33 1112 -1.18 3.9 0.9 74.1 1.3 1.6 8.0 9.5 0.6 5.3 0.2 11 CHS 38 1086 -1.30 66.2 3.0 3.5 3.8 7.7 0.3 11.5 1.0 10.4 0.3 Numbers in italic represent the standard errors 1 Maximum Na-loss, see text for detail [(Na2O in recombined bulk composition) / (Na2O in target compostion * 100)] - 100 x: melt not analyzable *mass balance calculated with quenched modified melt

55 Table 5. Summary of phase compositions olivine opx pig aug plag metal exp T ∆ IW Fo Cr2O3 CaO En Wo En Wo En Wo An Ni CI CHS 69 1301 -2.44 94.1 0.70 0.21 91.4 0.8 n.a CHS 67 1274 -1.59 83.1 0.75 0.22 n.a CHS 66 1250 -2.53 95.1 0.69 0.26 90.7 3.4 n.a CHS 71 1216 -1.59 83.9 0.77 0.28 82.4 3.0 n.a CHS 64 1202 -1.08 74.1 0.58 0.28 30.6 CHS 63 1202 -1.38 80.1 0.64 0.29 77.7 4.6 11.6 CHS 65 1201 -2.27 92.5 0.67 0.33 86.34 5.93 n.a CHS 62 1192 -1.87 84.6 0.68 0.31 78.1 5.8 9.8 CHS 70 1167 -1.84 85.4 0.59 0.30 77.39 8.38 n.a CHS 68 1159 -1.53 78.9 0.48 0.34 72.24 8.24 n.a CHS 58 1136 -2.27 89.2 0.54 0.27 82.6 5.0 53.9 37.7 7.9 CHS 57 1132 -1.44 75.9 0.40 0.31 69.7 8.1 20.0 CHS 54 1129 -1.55 78.1 0.41 0.36 71.3 9.2 13.3 CHS 60 1104 -1.63 79.0 0.32 0.25 77.6 5.2 58.4 29.5 32.5 n.a. CHS 59 1100 -2.19 87.6 0.41 0.23 82.4 5.6 58.5 31.9 29.5 9.8 H CHS 69 1301 -2.44 93.2 0.55 0.22 93.1 0.9 n.a. CHS 61 1301 -1.77 86.9 0.55 0.20 86.4 1.6 7.6 CHS 27 1244 -2.08 89.1 0.52 0.24 86.1 3.3 4.6 CHS 63 1202 -1.38 78.2 0.59 0.26 77.2 3.5 7.0 CHS 65 1201 -2.27 91.9 0.47 0.25 87.5 4.4 n.a. CHS 48 1194 -1.98 87.8 0.53 0.24 83.4 4.4 5.3 CHS 62 1192 -1.87 84.4 0.53 0.25 78.9 4.6 6.7 CHS 46 1191 -1.61 76.8 0.56 0.30 75.7 4.0 8.2 CHS 43 1163 -2.06 90.0 0.41 0.22 82.9 6.7 CHS 44 1161 -1.86 85.6 0.48 0.24 79.8 6.3 4.8 CHS 41 1142 -2.00 87.1 0.38 0.23 81.2 4.6 79.8 6.3 70.1 18.9 43.9 5.5 CHS 39 1140 -1.30 70.9 0.36 0.29 68.6 7.5 9.0 CHS 57 1132 -1.44 73.0 0.31 0.37 68.6 7.3 10.0 CHS 33 1112 -1.18 64.6 0.17 0.37 62.4 10.0 41.4 17.8 CHS 40 1108 -1.81 84.2 0.34 0.24 79.4 3.5 53.7 36.3 38.1 5.3 CHS 42 1085 -1.65 81.6 0.29 0.22 77.0 4.6 53.5 33.0 28.4 6.3 CHS 35 1079 -1.04 61.6 0.22 0.30 69.2 2.8 48.2 32.3 28.7 23.3 CH 8* 1076 0.56 59.0 0.12 0.30 44.4 34.1 72.4 CH 5* 1073 0.76 58.7 0.14 0.31 41.9 38.5 42.1 75.6 CHS 37 1063 -0.98 60.8 0.20 0.29 64.7 3.3 45.6 34.5 24.7 30.0

56 Table 5. Summary of phase compositions (continued) olivine opx pig aug plag metal exp T ∆ IW Fo Cr2O3 CaO En Wo En Wo En Wo An Ni LL CHS 27 1244 -2.08 87.9 0.47 0.29 85.3 3.6 6.8 CHS 64 1202 -1.08 73.0 0.50 0.24 72.7 4.1 24.8 CHS 63 1202 -1.38 77.7 0.52 0.27 76.2 3.7 12.5 CHS 31 1202 -1.56 80.1 0.50 0.26 78.4 4.4 9.7 CHS 43 1163 -2.06 89.0 0.42 0.25 83.3 4.2 CHS 41 1142 -2.00 86.8 0.40 0.23 81.5 4.9 67.4 24.1 41.7 6.3 CHS 39 1140 -1.30 72.6 0.40 0.31 70.0 7.0 16.2 CHS 33 1112 -1.18 70.5 0.27 0.32 72.4 4.1 64.2 12.9 35.0 32.4 CHS 40 1108 -1.81 83.8 0.31 0.23 79.0 4.9 31.1 7.8 CHS 38 1086 -1.30 72.6 0.23 0.25 74.2 4.7 57.7 23.8 30.6 17.9 CHS 42 1085 -1.65 81.4 0.27 0.28 77.2 5.4 53.5 33.0 30.9 7.7 CHS 37 1063 -0.98 68.5 0.14 0.24 73.0 4.2 47.7 32.7 23.0 42.3 CM CHS 50 1248 -1.89 87.7 0.53 0.22 85.6 3.0 5.8 CHS 27 1244 -2.08 88.6 0.50 0.28 85.8 3.4 7.4 CHS 64 1202 -1.08 70.0 0.44 0.31 16.2 CHS 63 1202 -1.38 77.4 0.47 0.29 75.4 4.2 7.7 CHS 6 1185 -1.44 74.7 0.49 0.29 70.9 4.1 CHS 4 1166 -1.34 71.7 0.38 0.31 73.3 6.1 CHS 43 1163 -2.06 89.9 0.40 0.28 81.8 7.3 61.8 30.9 60.7 5.2 CHS 44 1161 -1.86 85.4 0.53 0.24 79.1 7.2 59.5 5.6 CHS 41 1142 -2.00 87.3 0.35 0.27 81.9 4.6 57.3 34.5 49.3 5.6 CHS 39 1140 -1.30 72.2 0.32 0.34 66.4 10.7 58.9 10.3 CHS 33 1112 -1.18 65.6 0.23 0.37 50.0 30.6 49.3 21.6 CHS 40 1108 -1.81 84.5 0.32 0.23 49.5 36.1 48.1 5.9 CHS 38 1086 -1.30 75.8 0.27 0.23 71.5 7.6 54.4 32.2 46.5 0.0 CHS 42 1085 -1.65 81.9 0.30 0.27 77.3 4.1 60.7 26.6 47.1 5.1 CHS 35 1079 -1.04 64.7 0.21 0.30 73.0 2.7 44.2 40.2 48.0 28.1 CHS 37 1063 -0.98 64.0 0.22 0.35 68.6 2.9 44.7 36.9 43.2 31.0 CH 10* 1118 0.71 63.1 0.18 0.39 55.0 20.9 71.4 72.2 CV CHS 50 1248 -1.89 87.3 0.52 0.22 85.3 3.2 6.1 CHS 64 1202 -1.08 72.6 0.40 0.30 15.4 CHS 63 1202 -1.38 78.4 0.51 0.28 75.8 4.2 8.9 CHS 43 1163 -2.06 89.5 0.46 0.29 84.0 4.3 54.5 39.0 66.0 6.1 CHS 44 1161 -1.86 84.3 0.45 0.29 79.1 6.0 55.2 35.6 67.7 7.0 CHS 39 1140 -1.30 70.6 0.32 0.36 64.8 12.0 57.7 23.2 61.7 12.4 CHS 33 1112 -1.18 66.5 0.00 0.32 48.7 29.7 58.3 24.1 CHS 38 1086 -1.30 71.2 0.27 0.30 73.1 4.9 45.3 33.6 54.9 12.7

57 Table 6. Crystallization products of experimental melts MELTS F T Δ IW Qtz Plag An av An max FPS HCP Wo HCP LCP Wo LCP Oliv CI CHS 64 28.1 1202 -1.08 3.1 40.9 20 54 80 21.6 42 27.5 6.5 3.1 CI CHS 63 27.3 1202 -1.38 3.8 45.9 17 52 81 23.4 40 21.0 8.5 2.5 CI CHS 62 22.4 1192 -1.87 3.1 54.8 16 53 94 22.2 40.5 14.5 8.5 1.7 CI CHS 60 20.2 1104 -1.63 5.9 72.6 14 35 99 3.7 41 12.9 3.5 1.5 CI CHS 57 14.4 1132 -1.44 5.4 58.6 18 51 99 17.9 39.5 13.4 9 2.1 CI CHS 59 11.2 1100 -2.19 9.5 78.2 17 35 100 9.4 3.5 0.4

H CHS 61 28.3 1301 -1.77 5.4 34.7 19 52 60 16.4 42 39.0 3.5 1.8 H CHS 46 18.8 1191 -1.61 3.9 50.6 19 54 90 19.6 42 22.2 6.5 2.0 H CHS 63 18.8 1202 -1.38 6.2 46.7 25 56 86 16.4 40 26.3 9 2.1 H CHS 39 16.6 1140 -1.30 4.4 55.4 20 49 93 11.2 39.5 22.2 9 1.4 H CHS 27 16.5 1244 -2.08 7.9 51.7 25 55 87 19.0 42 16.5 3.5 1.6 H CHS 57 15.6 1132 -1.44 3.1 58.5 22 54 98 17.2 39.5 15.1 9 1.5 H CHS 62 14.2 1192 -1.87 4.7 57.3 21 52 92 15.7 40 17.0 8.5 1.5 H CHS 44 13.1 1161 -1.86 7.7 64.5 20 51 99 15.5 40.5 8.2 8.5 1.1 H CHS 33 12.7 1112 -1.18 6.3 57.6 25 49 98 5.7 39 24.2 9 1.2 H CHS 43 12.1 1163 -2.06 9.0 67.4 20 48 96 10.9 44 10.8 2.5 0.6 H CHS 41 9.9 1142 -2.00 4.9 70.6 19 45 100 7.3 41.5 10.6 3.5 0.8 H CHS 40 7.9 1108 -1.81 9.8 71.2 19 40 100 2.5 41.5 12.2 3.5 0.7 H CHS 35 5.3 1079 -1.04 11.8 65.7 18 32 98 17.7 0.5 1.2 H CHS 42 5.0 1085 -1.65 12.3 74.8 13 28 99 0.7 43.5 10.0 2.5 0.3

LL CHS 64 24.9 1202 -1.08 2.6 41.4 19 53 79 21.3 41.5 30.3 6.5 2.5 LL CHS 63 21.7 1202 -1.38 LL CHS 27 18.9 1244 -2.08 5.1 51.2 19 54 86 21.7 42 16.6 3.5 1.2 LL CHS 31 18.2 1202 -1.56 7.1 52.7 22 54 94 19.2 42 17.5 6.5 1.7 LL CHS 39 17.4 1140 -1.30 3.3 57.4 21 50 96 11.1 39.5 22.8 9 1.4 LL CHS 43 15.3 1163 -2.06 3.3 44.3 20 54 80 20.8 40 27.5 9 2.1 LL CHS 33 14.9 1112 -1.18 9.5 63.5 26 46 96 0.2 39.5 22.8 9 1.1 LL CHS 41 11.6 1142 -2.00 8.8 69.2 21 43 99 5.2 42 11.4 3.5 0.6 LL CHS 40 10.9 1108 -1.81 8.7 74.7 16 37 100 3.3 41.5 11.2 3.5 0.5 LL CHS 37 7.6 1063 -0.98 10.0 70.7 15 30 99 12.9 0.5 0.6 LL CHS 42 5.5 1085 -1.65 2.4 72.2 8 29 98 6.3 41.5 8.1 3 0.4

CM CHS 64 27.9 1202 -1.08 1.0 34.5 49 67 77 12.6 42.5 46.4 6.5 2.4 CM CHS 63 23.9 1202 -1.38 4.1 39.9 49 68 80 16.4 42.5 35.1 6.5 2.0 CM CHS 50 22.7 1248 -1.89 7.6 41.5 48 69 81 18.9 41 27.9 8 2.0 CM CHS 27 21.4 1244 -2.08 6.6 46.2 39 64 83 20.6 41 22.3 7.5 1.7 CM CHS 39 15.0 1140 -1.30 2.9 49.5 40 65 98 16.3 40 25.6 8.5 1.8 CM CHS 44 12.7 1161 -1.86 6.4 58.0 39 67 99 21.7 40.5 11.1 8.5 1.6 CM CHS 43 10.4 1163 -2.06 6.7 60.6 35 62 99 16.7 44.5 13.5 2.5 0.8 CM CHS 33 5.2 1112 -1.18 3.2 54.6 23 50 95 7.9 39.5 24.7 8.5 1.6

CV CHS 63 28.1 1202 -1.38 5.1 39.0 55 72 83 16.5 43.0 35.9 6.5 1.2 CV CHS 50 27.3 1248 -1.89 CV CHS 64 27.1 1202 -1.08 38.8 61 74 94 15.9 42.5 40.5 6.5 2.0 CV CHS 39 9.6 1140 -1.30 2.9 47.2 40 65 97 14.9 40.0 29.7 9.0 0.9 CV CHS 33 3.9 1112 -1.18 6.8 48.4 29 49 86 32.0 8.5 0.8 An ave: Anorthite content of plagioclase after complete equilibrium crystallization An max: Anorthite content of plagioclase at plagioclase saturation FPS: residual melt fraction at plagioclase saturation HCP: fraction of high calcium pyroxene (augite or sub-calcic augite) LCP: fraction of low calciium pyroxene (opx or pigeonite) Wo: Wollastonite content of HCP and LCP

Supplementary material

1. Plagioclase nucleation delays in preliminary experiments, duplicate experiments and other precautions As discussed in section 4.1 of the manuscript, approaching the conditions of thermodynamic equilibrium in melting experiments of chondritic materials has been challenging for decades, in large part due to alkali-losses. An additional problem is that at the low temperatures (<1150 ºC), at which low-degree partial melts form (<15 wt.%), crystals are small (<< 5 µm) and the melt wets grain boundaries without forming pools sufficiently large (5-10 µm) to analyze accurately. A technique used to solve this problem is to first heat the chondritic material to a much higher temperature (e.g. 1300;(Usui et al., 2015)) and cool it rapidly to allow the crystals to grow from the melt and, eventually, equilibrate at the final experimental temperature. Experiments performed with an initial heating step above the final temperature (i.e. “superheat”) exhibit a larger average grainsize and contain analyzable pools of glass (Fig S1a). However, a concern inherent to this technique is that some of the phases that would be in thermodynamic equilibrium at the final temperature cannot nucleate during rapid cooling between the high and low temperature steps. In particular, plagioclase nucleation delays due to undercooling are known to occur in silicate melts e.g. (Gibb, 1974)(Grove and Bence, 1979). Melt compositions and phase assemblages in experiments affected by nucleation delays are not meaningful for constraining the melting processes of planetesimals. To ensure that experiments performed with an initial superheat are in equilibrium at the final temperature, duplicate experiments performed at the same conditions but at isothermal conditions are needed. We identified plagioclase nucleation delays in some of our preliminary experiments and these were not included in the study (i.e. not included in the main manuscript or any tables/figures). We report them here because we suspect that the absence of plagioclase in the experiments of (Usui et al., 2015) result from the same kinetic limitations that we discovered in the course of our study. Figure S1 shows three experiments performed with a superheat of 200-250 ºC. The first one, CHS 23_4, was performed on the LL composition and was first heated to 1338 ºC for one hour before being equilibrated at 1081 ºC and IW -1.2 for eight days. Despite the long duration, plagioclase did not nucleate in CHS 23_4. The experiments CHS 35_2 and 35_5 (H and LL compositions) equilibrated under identical conditions (1079 ºC and IW -1.05) but with a smaller superheat (130 ºC instead of 250 ºC) do contain plagioclase and melt in equilibrium with olivine,

59 two pyroxenes and metal (Fig. 2 in the main manuscript). Experiment CHS 33 shows that plagioclase is stable in H and LL compositions at least up to 1112 ºC at IW -1.2. CHS 25_1 and CHS 29_1 are two experiments performed on the CM composition with high superheat (220 ºC and 200 ºC) showing evidence that plagioclase nucleation was delayed. The CM composition contains plagioclase at least up to 1140 ºC at IW -1.3 (CHS 39). However, CHS 25, despite having been equilibrated at 1079 ºC for seven days, contains only skeletal plagioclase (Fig S1b) in lieu of massive plagioclase crystals (CHS 35; Fig.2). At even lower temperature, in CHS 29 (1063 ºC), the plagioclase is completely absent. A

100 µm

Figure S1. Preliminary experiments highlighting plagioclase nucleation delays. A: CHS 23_4 (LL), 1081 ºC, IW -1.2. B: CHS 25_1 (CM), 1079 ªC, IW -0.9. C: CHS 29_1, 1062 ºC, IW -0.9 (see text for detail).

All experiments presented in the main manuscript and in all tables and figures were either conducted isothermally or with smaller superheat of 40-130 ºC (Table 1). These experiments did not suffer from plagioclase nucleation delays. Several duplicate experiments conducted isothermally at the same conditions ensured that phase assemblages were in thermodynamic equilibrium in all experiments. For examples, CHS 28 was conducted isothermally at 1096 ºC and IW -1.3. While not all phases could be analyzed quantitatively due the small grainsize, the presence of olivine, orthopyroxene, augite, plagioclase, chromite, metal and glass was confirmed by energy dispersive spectrometry in the H and LL compositions (Fig. S2). Such a phase assemblage is consistent with the ones in CHS 33 and CHS 35 (1112 ºC and 1079 ºC), which were heated first to 1240 ºC and also contain plagioclase. Experiments CHS 54_4 and CHS 57_1 (1129 ºC and 1132 ºC; CI composition), show that plagioclase melts out at or slightly under 1130 ºC at IW -1.5. While CHS 54 was conducted isothermally and CHS 57 was first heated to 1220 ºC for two hours, none of them contain plagioclase and the compositions of their olivine and pigeonite are nearly identical. All experiments conducted with moderate super heat (40-130 ºC) were also initially kept at a higher pressure than the final pressure to maintain a constant fO2 relative to the IW buffer. The pressure was released precisely at the end of the higher temperature step to attain the same target fO2 at the final (lower) temperature. We find that if the pressure was not adjusted to compensate for the pressure sensitivity of the CCO buffer (French and Eugster, 1965), the fO2 was too reducing during the high temperature step and the experimental products were still slightly heterogenous in terms of Mg# at the end of experiments (inherited Mg-rich cores).

oliv opx glass

aug plag

Figure S2. (previous page) Texture and phase assemblage of low-temperature isothermal experiment CHS 28_3 (LL composition).

2. Melting reaction coefficients Until plagioclase melts out of the residual assemblages, melting reactions consume plagioclase and pyroxene to produce olivine and liquid in all chondritic compositions. To calculate the melting reaction coefficients in section 3.3, we performed linear regressions of the fraction of different phases as a function of the Fo content in olivine and the melt fraction. Alternatively, the coefficients of melting reactions can be calculated as a function of the metal fraction or the fO2 (∆ IW). We show below how the results of the three methods differ for the LL composition:

Constant Fo: 0.65 plag + 0.38 px + 0.13 metal = 1 liq + 0.18 oliv

Constant ∆ IW: 0.54 plag + 0.57 px + 0.24 metal = 1 liq + 0.43 oliv

Constant metal: 0.79 plag + 0.26 px = 1 liq + 0.08 oliv

3. Metal phases in experiments

phosphide

FeNi metal C poor

FeNi metal C poor / C rich

Figure S3. Metal phases in experiment CHS 43_2 (H composition, 1163 ºC, IW -2.1). The texture and composition of FeNi metal suggest that liquid (C-rich) and solid metal (C-poor) were present at the experimental temperature (FeC eutectic = 1149 ºC). Phosphide is ubiquitous in experiments and particularly abundant at low fO2, explaining the low P2O5 contents of experimental melts.

4. Experimental melt compositions as a function of temperature

70 Experimental melts 18 IW -1.5 ± 0.25 LL CM 16 65 H CV CI 14

60 (wt.%) 3 O

2 12 SiO2 (wt.%) Al

55 10

50 8 1.05 1.1 1.15 1.2 1.25 1.3 1.05 1.1 1.15 1.2 1.25 1.3 3 Temperature (¼C) 10 Temperature (¼C) 103

12 12

10 10

8 8

6 6 CaO (wt.%) MgO (wt.%)

4 4

2 2 1.05 1.1 1.15 1.2 1.25 1.3 1.05 1.1 1.15 1.2 1.25 1.3 3 Temperature (¼C) 10 Temperature (¼C) 103

7 35

6 30 gou

25 t 5 20 4 15 Na2O (wt.%)

3 Melt fraction (wt.%) 10

2 5

1 0 1.05 1.1 1.15 1.2 1.25 1.3 1.05 1.1 1.15 1.2 1.25 1.3 3 Temperature (¼C) 10 Temperature (¼C) 103 Figure S1. The partial melt compositions of all high NaK# starting materials (H, LL and CI) are nearly identical at a given temperature. At a given temperature, the same melt represents a larger degree of melting in the CI composition (by ~ 5 wt.%) compared to the H and LL composition.

5. Extraction of partial-melts: two-phase flow scaling analysis

In two-phase flow theory, the characteristic length scale of melt migration is called the compaction length, with larger lengths representing more efficient compaction:

&(ζ + 4η⁄3) δ = % # µ where ζ and η are the effective bulk and shear viscosities of the matrix, µ is shear viscosity of the melt and & is the permeability, which is a function of the grain size, 0, and of the porosity, Φ, and is expressed in the original formulation of (McKenzie, 1984) as: & = 02Φ3/ 1000 The characteristic velocity of melt extraction can be understood as the relative velocity of the melt in the part of the solid-liquid medium that is not affected by compaction: & (1 − Φ) ∆ρ g 7 = 8 µ Φ where ∆ρ is the density difference between the solid and the liquid and g is the surface acceleration of gravity. McKenzie (1985) also shows that the time => needed for the melt fraction to decrease by a factor of l/e in the compacting layer is independent of the matrix viscosity if the compaction length is much smaller than the height of the partly molten region, h: ℎ => = 78 (1 − Φ) For a 100 km planetesimal (g=0.1 m/s2), with an average grainsize of 1 mm to 1 cm, a porosity of 0.1, and a picritic basaltic melt rheology (∆ ρ = 400 kg/m3, µ = 1-10 Pa.s), the compaction length is 0.3-10 km, the melt extraction velocity is 1.1-1100 km/Ma and => is 0.1 to 10 Ma, a time-scale which, with the right combination of parameters (large grain size and low viscosity), can be shorter than the half-life of 26Al (0.72 Ma). (Lichtenberg et al., 2019) modelled this process numerically in 1-D with a viscosity of 1 Pa.s and showed that melt can either be extracted to the surface and form sills or be retained in planetesimal interiors and form magma oceans. However, if a trachy-andesitic rheology is used (∆ ρ = 800 kg/m3, µ = 104 Pa.s) to represent low-degree partial melts on high NaK# planetimals, the compaction length, the melt extraction velocity and => become 10-100 m, 2.3-230 m/Ma and 500-5000 Ma, respectively.

Table S1. Composition of trachyandesite achondrites and NWA 11119 Res. NWA NWA GRA 06128/9 ALM-A Liq 11575 11119 1 2 3 4 5 6 7 8 9 10 11 12 13 SiO2 56.1 55 57.35 58.35 59.98 63.04 59.46 60.63 61.81 64.45 73.25 61.69 61.37 TiO2 0.1 0.12 0.03 0.03 0.04 0.30 0.66 0.36 0.24 0.46 1.19 0.18 Al2O3 14.9 14.8 15.72 16.00 17.02 15.72 14.51 14.66 16.59 15.39 11.40 14.87 19.05 Cr2O3 0.06 0.08 0.03 0.03 0.03 0.03 0.28 0.25 0.16 0.13 0.01 0.11 0.17 FeO 9.96 11.8 9.31 9.47 6.96 6.26 5.51 4.35 3.35 3.48 3.92 8.50 1.49 MnO 0.16 0.14 0.13 0.13 0.10 0.16 0.27 0.26 0.20 0.23 0.35 0.17 0.22 MgO 4.61 4.15 3.75 3.82 2.96 2.36 4.76 4.83 3.53 2.80 0.36 2.83 4.52 CaO 5.37 4.86 4.85 4.03 4.28 3.87 7.22 7.27 5.88 5.10 2.53 4.85 12.08 Na2O 6.81 6.74 7.25 7.36 7.83 6.69 6.52 6.58 7.38 6.35 2.91 6.37 0.89 K2O 0.22 0.22 0.26 0.26 0.28 0.90 0.29 0.31 0.35 0.95 2.96 0.61 0.04 P2O5 1.56 1.21 1.32 0.52 0.55 0.65 0.51 0.48 0.55 0.65 0.99 total 99.9 99.1 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Mg# 45.2 38.5 41.8 41.8 43.1 40.2 60.6 66.4 65.2 58.9 37.3 84.4

66 Fo of olivine in equilibrium (KD = 0.27) 71.3 84.1 68.8 95

1 modal recombination Day et al. 2009 2 modal recombination Day et al. 2009 3 modal recombination based on Day et al. 2012 4 Same as 3 with 1.5 wt.% of phosphate subtracted 5 Same as 4 with half of olivine subtracted (6 wt.%) 6 Possible parental melt, same as 5 with residual melt added; after 70 wt.% of crystallization of the melt in CHS35_2 ICP-AES, Bischoff et al. 7 (2014) modal recombination based on Bischoff et al. 8 (2014) 9 Same as 8 with half of cpx subracted (12 wt.%) 10 Possible parental melt; same as 9 with residual melt added; after 70 wt.% of crystallization of the melt in CHS35_2 11 Residual liquid subtracted from 5 and 9 to obtain parental melt compositions (70 wt% crystallization of melt in CHS 35_2) 12 Shock melt vein (proxy for bulk composition), EPMA, Agee et al. (2018) 11 modal recombination Srinivasan et al. (2018)

Table S2. Composition of glass in experiments

exp SiO2 TiO2 Al2O3 Cr2O3 FeO MnO MgO CaO Na2O K2O P2O5 total n CI CHS 69 61.2 0.1 0.39 0.01 9.05 0.1 0.76 0.01 4.41 0.1 0.52 0.05 12.62 0.1 7.07 0.1 3.62 0.02 0.32 0.01 0.05 0.03 99.5 4 CHS 67 58.6 0.1 0.46 0.01 10.77 0.1 0.61 0.03 9.46 0.1 0.40 0.05 6.73 0.1 8.15 0.0 4.36 0.11 0.34 0.01 0.24 0.08 100.0 4 CHS 66 57.3 0.1 0.34 0.01 8.05 0.2 0.75 0.03 12.09 0.0 0.48 0.03 11.02 0.0 6.12 0.1 3.30 0.10 0.26 0.01 0.27 0.03 99.4 6 CHS 71 62.1 0.3 0.45 0.02 11.97 0.1 0.56 0.05 2.88 0.1 0.45 0.01 8.72 0.3 7.68 0.2 4.70 0.12 0.39 0.01 0.07 0.09 99.9 5 CHS 64 55.1 0.7 0.45 0.03 9.81 0.1 0.48 0.04 14.09 0.1 0.35 0.02 6.50 0.2 8.33 0.2 3.68 0.22 0.28 0.01 1.09 0.09 99.2 4 CHS 63 58.0 0.3 0.44 0.02 10.68 0.2 0.53 0.02 10.31 0.1 0.36 0.02 6.91 0.1 7.66 0.1 4.36 0.11 0.35 0.01 0.42 0.01 99.8 6 CHS 65 62.7 0.1 0.49 0.00 13.89 0.1 0.44 0.02 3.22 0.1 0.29 0.03 6.00 0.1 6.94 0.2 5.57 0.12 0.47 0.02 99.3 3 CHS 62 59.8 1.4 0.49 0.03 12.60 0.3 0.38 0.04 6.92 0.1 0.30 0.01 6.32 0.1 7.23 0.0 5.26 0.14 0.43 0.01 0.21 0.03 98.3 5 CHS 70 61.7 0.4 0.50 0.02 14.70 0.29 0.01 5.15 0.1 0.22 0.11 4.57 0.1 6.15 0.1 5.92 0.06 0.61 0.03 0.22 0.01 100.6 2 CHS 68 58.9 0.4 0.47 0.01 12.58 0.2 0.36 0.01 8.41 0.1 0.24 0.03 5.07 0.0 7.39 0.3 5.24 0.10 0.54 0.02 0.81 0.15 100.6 5 CHS 57 59.4 0.5 0.49 0.05 13.55 0.4 0.15 0.02 8.26 0.2 0.22 0.01 4.19 0.1 7.13 0.3 5.24 0.22 0.58 0.03 0.70 0.06 100.5 7 CHS 60 62.4 1.0 0.52 0.06 15.99 0.2 0.09 0.06 5.05 0.3 0.17 0.03 3.09 0.1 4.13 0.1 6.78 0.19 0.82 0.03 0.86 0.07 98.4 5 CHS 59 65.5 1.1 0.45 0.04 17.58 0.2 0.06 0.04 2.64 0.2 0.13 0.03 2.41 0.3 3.10 0.1 6.98 0.36 0.99 0.09 0.12 0.05 101.4 4 H CHS 69 61.3 0.3 0.44 0.01 9.78 0.1 0.60 0.01 4.47 0.1 0.37 0.04 11.84 0.1 7.16 0.1 3.62 0.15 0.42 0.01 0.02 0.02 99.2 4 CHS 61 58.9 0.1 0.38 0.05 8.34 0.1 0.55 0.03 9.70 0.1 0.37 0.03 12.47 0.1 5.75 0.1 3.16 0.14 0.30 0.00 0.06 0.02 100.0 5 CHS 27 60.9 0.5 0.50 0.06 12.84 0.1 0.41 0.03 5.87 0.2 0.30 0.03 7.17 0.2 7.15 0.1 4.32 0.18 0.48 0.02 0.03 0.05 100.0 4 CHS 63 58.2 0.3 0.52 0.02 11.60 0.2 0.39 0.02 10.87 0.1 0.29 0.02 6.61 0.1 7.09 0.1 3.87 0.11 0.37 0.01 0.20 0.01 100.0 3 CHS 65 63.3 0.1 0.55 0.00 14.19 0.1 0.31 0.02 3.39 0.1 0.15 0.03 5.61 0.1 6.36 0.2 5.53 0.12 0.61 0.02 0.03 100.4 3 CHS 62 60.3 0.8 0.48 0.01 13.70 0.1 0.34 0.03 7.01 0.1 0.24 0.00 5.61 0.1 6.59 0.0 5.10 0.16 0.51 0.01 0.11 0.04 97.4 4 67 CHS 46 58.3 0.1 0.45 0.02 11.85 0.1 0.36 0.02 10.54 0.1 0.25 0.01 6.05 0.1 7.00 0.1 4.58 0.10 0.43 0.01 0.21 0.01 101.1 6

CHS 43 63.4 0.8 0.57 0.03 15.66 0.2 0.17 0.04 3.32 0.0 0.18 0.02 4.56 5.47 0.0 5.67 0.17 0.95 0.04 0.02 0.02 100.1 6 CHS 44 62.5 0.4 0.60 0.02 15.11 0.2 0.21 0.04 4.71 0.2 0.20 0.03 3.88 0.1 6.26 0.2 5.61 0.12 0.79 0.03 0.07 0.03 100.5 13 CHS 41 63.5 0.4 0.58 0.04 16.49 0.3 0.11 0.03 3.92 0.3 0.18 0.04 3.53 0.2 4.80 0.0 6.23 0.22 1.08 0.06 0.14 0.05 98.6 6 CHS 39 58.6 0.3 0.56 0.02 13.08 0.1 0.18 0.01 10.16 0.1 0.17 0.01 4.22 0.0 6.55 0.1 4.79 0.19 0.79 0.02 0.93 0.10 100.0 5 CHS 57 58.7 0.5 0.53 0.05 14.09 0.4 0.19 0.02 8.84 0.2 0.17 0.01 4.04 0.1 7.27 0.3 5.09 0.22 0.59 0.03 0.46 0.06 101.4 8 CHS 33 58.5 0.7 0.58 0.04 14.05 0.5 0.07 0.03 10.50 0.6 0.14 0.01 3.35 0.2 6.33 0.2 4.61 0.09 0.79 0.05 1.02 0.11 100.5 7 CHS 40 64.5 0.4 0.52 0.22 16.47 0.3 0.09 0.03 3.67 0.0 0.15 0.03 3.14 0.1 4.01 0.1 6.02 0.20 1.15 0.06 0.27 0.08 100.9 4 CHS 42 66.5 1.7 0.50 0.06 16.19 0.3 0.07 0.04 3.31 0.0 0.09 0.02 2.08 2.70 0.0 6.38 1.84 0.05 0.33 0.03 101.3 4 CHS 35 62.6 1.4 0.55 0.00 15.27 0.3 0.07 0.01 7.62 0.2 0.11 0.00 2.21 0.2 3.68 0.2 5.30 0.25 1.51 0.02 1.05 0.16 98.0 4 CH 8* 59.5 0.8 0.51 0.01 14.70 0.2 0.07 0.00 10.58 0.1 0.13 0.00 2.86 0.0 5.91 0.4 4.69 0.07 0.56 0.00 0.57 0.00 100.3 7 CH 5* 59.3 1.6 0.53 0.12 14.81 0.2 0.05 0.02 10.33 0.5 0.13 0.03 2.89 0.1 6.00 0.6 4.55 0.12 0.95 0.06 1.14 0.16 100.7 6 numbers in italic are the standard errors, total is the original total but all compositions are normalized to 100, n is the number of analyses *1-atm experiment

Table S2. Composition of glass in experiments (continued)

exp SiO2 TiO2 Al2O3 Cr2O3 FeO MnO MgO CaO Na2O K2O P2O5 total n LL CHIS 27 60.8 1.1 0.47 0.02 12.16 0.1 0.40 0.04 6.08 0.1 0.29 0.03 7.45 0.2 7.08 0.2 4.70 0.26 0.50 0.01 0.06 0.03 99.7 7 CHIS 63 56.0 1.0 0.39 0.02 9.82 1.1 0.36 0.04 14.56 0.4 0.28 0.02 6.75 2.6 7.60 0.7 3.77 0.12 0.31 0.54 0.04 100.1 5 CHIS 64 57.1 0.1 0.42 0.09 10.57 0.1 0.40 0.05 12.01 0.2 0.28 0.03 7.18 0.2 7.50 0.2 3.98 0.06 0.35 0.01 0.34 0.08 99.2 4 CHIS 31 59.8 0.9 0.49 0.02 12.62 0.1 0.33 0.03 8.68 0.2 0.24 0.01 5.45 0.2 7.25 0.1 4.59 0.09 0.40 0.01 0.24 0.06 100.0 5 CHIS 43 62.7 0.6 0.45 0.04 15.59 0.3 0.21 0.04 3.99 0.2 0.18 0.01 4.80 0.2 6.24 0.1 5.48 0.12 0.56 0.03 0.04 0.06 101.4 7 CHIS 41 64.5 1.6 0.46 0.01 16.33 0.2 0.16 0.03 3.48 0.16 0.04 3.52 0.6 4.56 0.1 6.01 0.07 0.90 0.02 0.13 0.00 99.6 3 CHIS 39 58.3 0.1 0.36 0.01 13.60 0.1 0.19 0.02 9.57 0.2 0.17 0.01 4.72 0.3 6.65 0.2 5.06 0.14 0.51 0.02 0.84 0.05 100.3 4 CHIS 33 60.6 1.2 0.46 0.01 15.45 0.2 0.07 0.02 8.04 0.2 0.14 0.02 3.20 0.1 5.69 0.0 4.99 0.11 0.69 0.04 1.02 0.16 99.7 4 CHIS 40 66.0 0.4 0.45 0.22 16.80 0.3 0.10 0.03 3.66 0.0 0.14 0.03 3.03 0.1 3.93 0.1 6.87 0.20 0.83 0.06 0.38 0.08 98.4 5 CHIS 42 66.9 1.4 0.40 0.00 15.99 1.1 0.03 0.04 3.61 0.08 0.03 2.32 2.66 0.2 6.77 0.09 1.60 0.03 98.5 2 CHIS 37 64.5 0.7 0.39 0.03 16.72 0.3 0.07 0.03 5.14 0.2 0.08 0.01 2.02 0.3 3.24 0.1 6.37 0.43 1.11 0.03 0.84 0.11 100.9 13 CM CHS 50 56.5 0.2 0.65 0.02 12.26 0.2 0.52 0.01 8.09 0.3 0.39 0.02 9.64 0.6 9.27 0.4 2.37 0.06 0.21 0.01 0.08 0.02 100.5 3 CHS 27 57.6 1.4 0.67 0.01 12.78 0.1 0.47 0.03 6.63 0.1 0.35 0.02 8.96 0.2 8.97 0.1 3.11 0.19 0.31 0.02 0.02 0.02 99.4 10 CHS 64 50.9 0.2 0.62 0.03 10.46 0.2 0.39 0.03 18.04 0.2 0.31 0.02 8.20 0.1 8.37 0.1 1.90 0.15 0.21 0.01 0.64 0.10 100.1 7 CHS 63 53.6 0.2 0.67 0.01 11.89 0.1 0.47 0.03 13.01 0.0 0.32 0.02 8.32 0.1 8.95 0.1 2.25 0.05 0.21 0.01 0.29 0.03 98.9 5 CHS 43 59.5 1.0 0.83 0.18 15.26 0.6 0.27 0.07 4.44 0.5 0.21 0.03 6.17 0.3 8.14 0.3 4.39 0.20 0.73 0.07 0.07 0.04 100.2 3 CHS 44 57.6 0.8 0.69 0.03 15.46 0.2 0.31 0.04 6.24 0.0 0.26 0.02 5.35 9.68 0.0 3.82 0.17 0.43 0.04 0.16 0.02 100.1 8 CHS 39 53.7 0.6 0.61 0.04 13.75 0.2 0.21 0.04 11.81 0.2 0.22 0.01 5.52 0.0 9.58 0.1 3.14 0.20 0.46 0.02 0.97 0.16 99.1 6 CHS 33 55.9 0.8 1.03 0.04 13.47 0.3 0.16 0.07 11.10 0.6 0.18 0.03 3.93 0.2 7.38 0.7 4.17 0.03 1.09 0.17 1.91 0.40 100.7 4 68 CHS 40 67.5 0.4 0.97 0.22 15.33 0.3 0.05 0.03 3.19 0.0 0.12 0.03 2.14 0.1 3.32 0.1 5.53 0.20 1.94 0.06 0.17 0.08 99.4 6

CH 10* 50.5 1.4 0.82 0.07 11.76 0.6 0.17 0.08 17.06 0.7 0.26 0.07 5.85 0.4 10.91 0.3 1.86 0.02 0.26 0.01 0.58 0.27 98.5 3 CV CHS 50 55.8 0.1 0.46 0.02 12.60 0.1 0.58 0.05 8.21 0.2 0.21 0.02 10.33 0.4 9.59 0.2 1.90 0.10 0.19 0.01 0.07 0.03 0.1 5 CHS 64 49.4 0.4 0.55 0.02 12.62 0.1 0.41 0.07 17.00 0.2 0.21 0.00 7.25 0.1 10.29 0.3 1.65 0.12 0.16 0.71 0.01 4 CHS 63 53.7 1.0 0.48 0.02 12.13 1.1 0.47 0.04 12.44 0.4 0.20 0.02 8.60 2.6 9.48 0.7 1.94 0.12 0.17 0.34 0.04 0.0 5 CHS 39 53.4 1.1 0.60 0.08 13.14 0.4 0.15 0.04 12.50 0.5 0.13 0.03 5.76 0.4 9.64 0.4 3.04 0.30 0.37 0.04 1.26 0.15 0.0 4 CHS 33 55.2 0.8 0.90 0.06 12.52 0.2 0.16 0.01 11.75 0.4 0.12 0.03 3.99 0.1 7.93 0.13 3.52 0.03 0.90 0.02 2.98 0.03 0.0 2 numbers in italic are the standard errors, total is the original total but all compositions are normalized to 100, n is the number of analyses *1-atm experiment

Table S3. Composition of olivine in experiments

exp T ∆ IW SiO2 Al2O3 Cr2O3 FeO MnO MgO CaO NiO total n CI CHS 69 1301 -2.4 40.5 0.2 0.04 0.01 0.70 0.03 5.8 0.1 0.56 0.03 52.1 0.1 0.21 0.01 100.2 4 CHS 67 1274 -1.6 38.4 0.1 0.01 0.01 0.75 0.02 16.0 0.1 0.46 0.03 44.1 0.2 0.22 0.01 0.04 101.1 7 CHS 66 1250 -2.5 40.4 0.2 0.03 0.01 0.69 0.01 4.9 0.1 0.58 0.05 53.1 0.1 0.26 0.01 0.04 100.1 4 CHS 71 1216 -1.6 39.0 0.1 0.01 0.01 0.77 0.01 15.1 0.1 0.55 0.03 44.2 0.1 0.28 0.01 101.3 6 CHS 64 1202 -1.1 36.6 0.3 0.58 0.13 23.8 0.1 0.48 0.02 38.2 0.1 0.28 0.02 0.11 100.8 8 CHS 63 1202 -1.4 38.3 0.1 0.64 0.04 18.5 0.1 0.45 0.02 41.8 0.1 0.29 0.01 0.04 99.8 5 CHS 65 1201 -2.3 40.4 0.4 0.10 0.18 0.67 0.03 7.4 0.1 0.54 0.02 50.6 0.8 0.33 0.07 0.03 100.4 5 CHS 62 1192 -1.9 38.3 0.1 0.68 0.03 14.7 0.1 0.50 0.02 45.4 0.3 0.31 0.01 0.08 99.3 5 CHS 70 1167 -1.8 39.0 0.8 0.03 0.03 0.59 0.01 13.9 0.2 0.54 0.02 45.6 0.2 0.30 0.04 100.5 7 CHS 68 1159 -1.5 38.8 0.3 0.01 0.02 0.48 0.01 19.3 0.1 0.51 0.03 40.5 0.2 0.34 0.02 0.04 101.1 5 CHS 58 1136 -2.3 40.3 0.0 0.05 0.01 0.54 0.01 10.3 0.0 0.60 0.05 47.8 0.2 0.27 0.04 100.2 3 CHS 57 1132 -1.4 37.5 0.3 0.40 0.05 22.2 0.0 0.39 0.01 39.2 0.3 0.31 0.05 0.05 99.2 5 CHS 54 1129 -1.6 37.8 0.3 0.05 0.03 0.41 0.02 20.3 0.2 0.54 0.04 40.5 0.3 0.36 0.01 0.05 100.4 4 CHS 60 1104 -1.6 38.3 0.4 0.02 0.01 0.32 0.02 19.4 0.2 0.56 0.03 41.0 0.4 0.25 0.02 0.07 100.3 8 CHS 59 1100 -2.2 39.7 0.3 0.04 0.02 0.41 0.01 11.9 0.1 0.54 0.02 47.1 0.4 0.23 0.05 101.2 6 H CHS 69 1301 -2.4 40.5 0.2 0.02 0.01 0.55 0.03 6.7 0.1 0.42 0.03 51.5 0.1 0.22 0.01 100.9 5 CHS 61 1301 -1.8 39.3 0.2 0.03 0.02 0.55 0.02 12.6 0.1 0.43 0.04 46.9 0.3 0.20 0.02 0.03 100.7 9 CHS 27 1244 -2.1 40.3 0.3 0.05 0.01 0.52 0.02 10.5 0.1 0.46 0.06 47.9 0.3 0.24 0.00 0.05 101.3 5 CHS 63 1202 -1.4 37.9 0.2 0.59 0.02 20.2 0.4 0.37 0.06 40.6 0.1 0.26 0.01 0.05 100.1 5

69 CHS 65 1201 -2.3 40.1 0.2 0.04 0.02 0.47 0.02 8.0 0.1 0.39 0.03 50.7 0.4 0.25 0.02 0.07 101.1 6

CHS 48 1194 -2.0 39.8 0.1 0.01 0.01 0.53 0.01 11.7 0.2 0.42 0.04 47.2 0.2 0.24 0.02 0.05 100.9 6 CHS 62 1192 -1.9 38.2 0.1 0.53 0.03 15.0 0.1 0.38 0.02 45.6 0.3 0.25 0.01 0.06 99.5 4 CHS 46 1191 -1.6 38.4 0.2 0.03 0.02 0.56 0.02 21.1 0.1 0.41 0.04 39.1 0.1 0.30 0.02 0.10 0.04 100.7 7 CHS 43 1163 -2.1 40.6 0.1 0.04 0.01 0.41 0.02 9.6 0.1 0.47 0.03 48.5 0.0 0.22 0.02 0.07 100.8 4 CHS 44 1161 -1.9 39.8 0.2 0.01 0.00 0.48 0.01 13.6 0.1 0.48 0.03 45.3 0.2 0.24 0.01 0.01 100.2 7 CHS 41 1142 -2.0 40.0 0.4 0.07 0.04 0.38 0.04 12.3 0.1 0.45 0.05 46.4 0.3 0.23 0.02 0.07 101.3 7 CHS 39 1140 -1.3 37.3 0.2 0.01 0.01 0.36 0.02 26.0 0.1 0.25 0.03 35.6 0.1 0.29 0.02 0.07 101.3 6 CHS 57 1132 -1.4 37.8 0.3 0.31 0.05 24.2 0.0 0.42 0.01 36.8 0.3 0.37 0.05 0.01 99.4 3 CHS 33 1112 -1.2 36.0 0.3 0.02 0.02 0.17 0.01 31.1 0.3 0.36 0.04 31.9 0.2 0.37 0.02 0.08 100.5 10 CHS 40 1108 -1.8 39.6 0.5 0.13 0.01 0.34 0.02 14.8 0.8 0.44 0.03 44.3 0.3 0.24 0.00 0.06 101.4 6 CHS 42 1085 -1.7 38.6 0.4 0.10 0.04 0.29 0.04 17.2 0.1 0.43 0.05 43.0 0.3 0.22 0.02 0.05 100.2 9 CHS 35 1079 -1.0 35.7 0.3 0.01 0.01 0.22 0.01 33.3 0.2 0.33 0.03 30.0 0.2 0.30 0.02 0.09 0.02 101.0 7 CHS 37 1063 -1.0 35.2 0.3 0.01 0.01 0.20 0.02 34.0 0.2 0.40 0.03 29.6 0.2 0.29 0.02 0.11 0.03 102.0 7 CH 8* 1076 0.6 35.7 0.2 0.01 0.01 0.12 0.01 35.0 0.1 0.30 0.03 28.2 0.1 0.30 0.02 0.66 0.06 101.1 8 CH 5* 1073 0.8 35.6 0.4 0.01 0.01 0.14 0.04 34.9 0.2 0.22 0.02 27.8 0.2 0.31 0.01 0.90 0.05 101.3 6 numbers in italic are the standard errors, total is the original total but all compositions are normalized to 100, n is the number of analyses

Table S3. Composition of olivine in experiments (continued) ∆ exp T SiO Al O Cr O FeO MnO MgO CaO NiO total n IW 2 2 3 2 3 LL CHIS 27 1244 -2.1 40.7 0.4 0.03 0.01 0.47 0.02 11.4 0.0 0.41 0.00 46.6 0.7 0.29 0.02 0.03 99.6 4 CHIS 63 1202 -1.1 37.2 0.1 0.50 0.04 24.5 0.1 0.38 0.02 37.1 0.1 0.24 0.01 0.07 0.04 100.8 7 CHIS 64 1202 -1.4 37.6 0.2 0.52 0.02 20.7 0.4 0.37 0.06 40.4 0.1 0.27 0.01 0.05 99.4 5 CHIS 31 1202 -1.6 39.8 0.2 0.02 0.01 0.50 0.04 18.1 0.1 0.38 0.01 40.9 0.4 0.26 0.01 100.8 3 CHIS 43 1163 -2.1 40.3 0.9 0.02 0.01 0.42 0.02 10.6 0.7 0.42 0.07 47.8 0.6 0.25 0.03 0.08 0.05 101.0 4 CHIS 41 1142 -2.0 39.7 0.2 0.05 0.01 0.40 0.05 12.6 0.3 0.25 0.14 46.6 0.3 0.23 0.03 0.06 101.1 6 CHIS 39 1140 -1.3 37.7 0.2 0.04 0.03 0.40 0.01 24.6 0.2 0.24 0.03 36.6 0.3 0.31 0.00 0.03 99.8 3 CHIS 33 1112 -1.2 37.1 0.3 0.01 0.01 0.27 0.02 26.5 0.5 0.28 0.03 35.5 0.4 0.32 0.02 0.05 100.3 5 CHIS 40 1108 -1.8 38.8 0.6 0.05 0.04 0.31 0.02 15.3 0.1 0.48 0.06 44.6 0.2 0.23 0.02 0.09 0.09 100.7 4 CHIS 42 1085 -1.7 39.5 0.2 0.15 0.08 0.27 0.04 17.2 0.2 0.35 0.04 42.1 0.3 0.28 0.03 0.04 100.9 5 CHIS 37 1063 -1.0 36.7 0.3 0.07 0.01 0.14 0.02 28.1 0.2 0.38 0.03 34.2 0.2 0.24 0.02 0.08 0.02 100.8 4 CM CHS 50 1248 -1.9 40.4 0.1 0.04 0.01 0.53 0.01 11.6 0.1 0.39 0.04 46.7 0.1 0.22 0.02 0.04 101.0 5 CHS 27 1244 -2.1 39.5 0.3 0.03 0.01 0.50 0.02 11.0 0.1 0.44 0.06 48.2 0.3 0.28 0.00 0.02 98.0 3 CHS 64 1202 -1.1 36.5 0.2 0.44 0.01 27.0 0.1 0.35 0.03 35.3 0.2 0.31 0.03 0.06 101.3 7 CHS 63 1202 -1.4 37.8 0.1 0.47 0.04 20.9 0.2 0.35 0.03 40.1 0.3 0.29 0.02 0.03 99.8 5 CHS 43 1163 -2.1 40.4 0.9 0.05 0.01 0.40 0.02 9.7 0.7 0.38 0.07 48.7 0.6 0.28 0.03 0.02 99.7 11 CHS 44 1161 -1.9 39.2 0.2 0.01 0.01 0.53 0.00 13.9 0.1 0.38 0.04 45.6 0.2 0.24 0.01 0.01 99.7 5 CHS 41 1142 -2.0 39.8 0.3 0.10 0.08 0.35 0.02 12.1 0.1 0.44 0.03 46.7 0.2 0.27 0.03 0.06 101.0 7 70 CHS 39 1140 -1.3 37.6 0.2 0.04 0.02 0.32 0.02 25.0 0.1 0.24 0.03 36.4 0.2 0.34 0.03 0.04 101.0 6 CHS 33 1112 -1.2 36.3 0.3 0.01 0.01 0.23 0.03 30.3 0.2 0.35 0.03 32.5 0.2 0.37 0.02 0.02 100.4 11 CHS 40 1108 -1.8 38.9 0.3 0.20 0.07 0.32 0.02 14.7 0.1 0.45 0.03 45.0 0.6 0.23 0.01 0.05 100.2 5 CHS 38 1086 -1.3 38.4 0.2 0.04 0.02 0.27 0.02 22.0 0.1 0.29 0.03 38.5 0.2 0.23 0.03 0.04 100.6 3 CHS 42 1085 -1.7 39.3 0.3 0.06 0.03 0.30 0.01 16.8 0.4 0.40 0.01 42.9 0.2 0.27 0.04 0.05 100.7 5 CHS 35 1079 -1.0 36.2 0.3 0.01 0.01 0.21 0.03 30.9 0.2 0.36 0.03 31.8 0.2 0.30 0.02 0.05 101.1 6 CHS 37 1063 -1.0 35.7 0.4 0.01 0.01 0.22 0.01 31.5 0.2 0.38 0.03 31.4 0.2 0.35 0.06 0.09 0.02 101.5 7 CH 10* 1118 0.7 36.1 0.1 0.03 0.03 0.18 0.04 31.7 0.3 0.32 0.04 30.5 0.4 0.39 0.02 0.74 0.10 101.2 8 CV CHS 50 1248 -1.9 40.5 0.1 0.05 0.01 0.52 0.02 12.0 0.1 0.23 0.03 46.4 0.1 0.22 0.02 0.03 100.9 4 CHS 64 1202 -1.1 36.6 0.5 0.40 0.02 25.1 0.1 0.23 0.01 37.3 0.2 0.30 0.02 0.04 101.1 8 CHS 63 1202 -1.4 37.8 0.2 0.51 0.02 20.1 0.2 0.19 0.05 41.1 0.1 0.28 0.02 0.02 99.7 5 CHS 43 1163 -2.1 39.4 0.9 0.15 0.46 0.06 10.3 1.1 0.24 0.02 49.2 1.1 0.29 0.03 0.02 99.7 2 CHS 44 1161 -1.9 38.9 0.5 0.45 0.02 15.0 1.1 0.19 0.03 45.1 0.7 0.29 0.06 0.06 100.0 3 CHS 39 1140 -1.3 37.4 0.4 0.04 0.01 0.32 0.04 26.2 0.3 0.12 0.02 35.4 0.1 0.36 0.09 0.06 101.5 4 CHS 33 1112 -1.2 36.4 0.4 0.03 0.02 0.01 29.7 0.4 0.20 0.04 33.0 0.5 0.32 0.04 0.04 100.2 6 CHS 38 1086 -1.3 37.4 0.2 0.27 0.04 25.9 0.1 0.18 0.03 35.9 0.4 0.30 0.07 0.07 100.7 5 numbers in italic are the standard errors, total is the original total but all compositions are normalized to 100, n is the number of analyses

Table S4. Composition of orthopyroxene in experiments

exp T ∆ IW SiO2 TiO2 Al2O3 Cr2O3 FeO MnO MgO CaO Na2O K2O Total n CI CHS 69 1301 -2.44 57.7 0.1 0.06 0.03 0.19 0.04 0.80 0.02 5.3 0.6 0.43 0.01 35.1 0.3 0.40 0.01 0.02 0.02 0.01 0.00 100.6 4 CHS 66 1250 -2.53 57.1 0.9 0.10 0.01 0.30 0.03 0.99 0.05 4.0 0.6 0.63 0.03 34.9 0.7 1.84 0.19 0.06 0.03 0.00 0.00 100.1 4 CHS 71 1216 -1.59 56.0 0.3 0.05 0.01 0.32 0.05 1.07 0.05 9.7 0.3 0.55 0.02 30.7 0.7 1.56 0.37 0.10 0.02 0.01 0.00 101.2 8 CHS 63 1202 -1.38 55.2 0.3 0.01 0.02 0.42 0.02 1.04 0.02 11.7 0.2 0.51 0.02 28.7 0.3 2.37 0.16 0.07 0.03 99.8 5 CHS 62 1192 -1.87 54.6 0.4 0.10 0.02 0.59 0.05 1.11 0.05 10.7 0.2 0.57 0.03 29.2 0.2 3.01 0.24 0.13 0.04 0.01 0.01 100.1 6 CHS 60 1104 -1.63 55.0 0.3 0.09 0.03 0.56 0.09 0.81 0.09 11.4 0.4 0.56 0.02 28.8 0.4 2.68 0.30 0.14 0.05 0.00 0.01 101.6 11 CHS 59 1100 -2.19 55.6 0.5 0.08 0.02 0.72 0.20 1.00 0.12 8.0 0.3 0.63 0.03 30.8 0.5 2.94 0.14 0.16 0.05 0.02 0.01 100.4 8 H CHS 69 1301 -2.44 58.3 0.3 0.05 0.02 0.22 0.05 0.63 0.03 4.1 0.0 0.40 0.02 35.8 0.2 0.49 0.07 0.00 0.00 0.00 0.00 100.6 4 CHS 61 1301 -1.77 57.1 0.2 0.01 0.01 0.24 0.05 0.67 0.01 8.1 0.3 0.41 0.01 32.6 0.3 0.85 0.11 0.05 0.04 100.6 5 CHS 27 1244 -2.08 56.0 0.2 0.10 0.02 0.39 0.03 0.76 0.05 7.3 0.1 0.47 0.01 33.2 0.1 1.76 0.08 0.01 0.01 0.01 0.01 100.4 4 CHS 63 1202 -1.38 54.8 0.2 0.08 0.04 0.44 0.03 0.89 0.03 12.8 0.2 0.45 0.03 28.7 0.2 1.82 0.07 0.02 0.02 100.3 5 CHS 65 1201 -2.27 57.1 0.2 0.11 0.01 0.46 0.03 0.83 0.03 5.5 0.1 0.50 0.05 33.2 0.1 2.34 0.09 0.08 0.02 0.00 0.00 99.9 3 CHS 48 1194 -1.98 56.5 0.5 0.12 0.02 0.51 0.02 0.86 0.03 8.1 0.2 0.49 0.01 31.1 0.2 2.29 0.11 0.05 0.02 0.02 0.01 100.9 8 CHS 62 1192 -1.87 54.5 0.6 0.11 0.03 1.00 1.07 0.90 0.09 11.0 0.4 0.43 0.03 29.5 1.9 2.38 0.40 0.21 0.41 0.02 0.06 99.8 8 CHS 46 1191 -1.61 54.5 0.5 0.08 0.01 0.45 0.04 0.88 0.03 13.5 0.3 0.42 0.02 28.1 0.4 2.08 0.26 0.05 0.01 0.01 0.01 100.6 8 CHS 41 1142 -2.00 55.5 0.4 0.12 0.02 0.59 0.06 0.77 0.04 9.5 0.4 0.60 0.09 30.5 0.5 2.38 0.19 0.10 0.04 0.01 0.00 100.7 7 CHS 40 1108 -1.81 54.9 0.5 0.10 0.06 0.49 0.08 0.74 0.03 11.4 0.7 0.61 0.04 29.8 0.3 1.82 0.04 0.10 0.03 0.00 0.01 100.9 5 CHS 42 1085 -1.65 54.5 0.5 0.09 0.01 0.47 0.05 0.82 0.04 12.4 0.6 0.46 0.05 29.0 0.6 2.41 0.34 0.08 0.05 0.00 0.01 100.2 7

71 CHS 35 1079 -1.04 53.2 0.8 0.07 0.03 0.39 0.08 0.62 0.06 18.3 0.7 0.38 0.02 25.5 0.9 1.45 0.19 0.07 0.01 0.01 0.01 101.3 4 CHS 37 1063 -0.98 52.5 0.16 0.56 0.43 20.7 0.42 23.5 1.69 0.11 0.00 100.1 1

numbers in italic are the standard errors, total is the original total but all compositions are normalized to 100, n is the number of analyses

Table S4. Composition of orthopyroxene in experiments (continued)

exp T ∆ IW SiO2 TiO2 Al2O3 Cr2O3 FeO MnO MgO CaO Na2O K2O Total n LL CHIS 27 1244 -2.08 56.0 0.2 0.07 0.02 0.36 0.03 0.76 0.05 7.6 0.1 0.43 0.01 32.8 0.1 1.91 0.08 0.06 0.01 0.01 0.01 100.5 4 CHIS 63 1202 -1.08 54.5 0.2 0.02 0.01 0.35 0.05 0.76 0.05 15.2 0.1 0.32 0.02 26.8 0.2 2.08 0.09 0.03 0.02 0.01 0.00 101.0 7 CHIS 64 1202 -1.38 54.9 0.3 0.04 0.03 0.43 0.09 0.81 0.04 13.3 0.2 0.40 0.02 28.2 0.4 1.91 0.38 0.03 0.02 0.01 0.01 100.2 6 CHIS 31 1202 -1.56 55.2 0.9 0.07 0.01 0.51 0.05 0.93 0.05 11.4 0.4 0.45 0.02 29.2 0.6 2.28 0.09 0.07 0.05 0.01 0.01 99.6 4 CHIS 43 1163 -2.06 56.1 0.7 0.07 0.01 0.57 0.12 0.82 0.02 8.3 0.4 0.44 0.02 31.3 0.5 2.20 0.32 0.11 0.07 0.01 0.01 100.6 8 CHIS 41 1142 -2.00 55.4 1.0 0.11 0.04 0.75 0.24 0.76 0.06 9.1 0.5 0.53 0.11 30.6 0.6 2.58 0.04 0.15 0.08 0.02 0.01 100.4 6 CHIS 33 1112 -1.18 54.1 0.7 0.03 0.02 0.35 0.05 0.53 0.03 15.6 0.8 0.38 0.02 26.9 0.7 2.10 0.51 0.04 0.03 101.0 6 CHIS 40 1108 -1.81 55.6 0.3 0.07 0.03 0.54 0.11 0.66 0.03 10.7 0.4 0.64 0.02 29.5 0.6 2.53 0.21 0.07 0.03 0.00 100.4 5 CHIS 42 1085 -1.65 55.0 0.2 0.09 0.04 0.53 0.06 0.70 0.05 11.7 0.3 0.41 0.01 29.1 0.5 2.84 0.04 0.12 0.03 0.01 0.01 100.1 7 CHIS 37 1063 -0.98 54.2 0.7 0.06 0.02 0.38 0.19 0.52 0.04 15.1 0.4 0.38 0.01 27.1 0.4 2.17 0.18 0.10 0.05 0.00 0.00 102.0 4 CM CHS 50 1248 -1.89 56.5 0.2 0.08 0.02 0.64 0.13 0.68 0.05 7.7 0.1 0.42 0.02 32.5 0.3 1.59 0.11 0.02 0.03 0.01 0.00 99.7 5 CHS 27 1244 -2.08 55.6 0.2 0.11 0.02 0.62 0.03 0.79 0.05 7.4 0.1 0.42 0.01 33.1 0.1 1.85 0.08 0.07 0.01 0.00 0.01 100.2 4 CHS 63 1202 -1.38 54.6 0.2 0.08 0.04 0.72 0.03 0.78 0.03 13.5 0.2 0.37 0.03 27.9 0.2 2.15 0.07 0.01 0.02 0.01 100.4 3 CHS 43 1163 -2.06 55.1 0.6 0.19 0.03 0.95 0.16 1.02 0.06 7.4 0.9 0.49 0.04 30.9 0.7 3.83 0.80 0.07 0.03 0.00 0.00 100.8 14 CHS 41 1142 -2.00 55.2 0.6 0.19 0.07 0.85 0.12 0.85 0.04 9.0 0.9 0.63 0.03 30.8 0.4 2.41 0.25 0.08 0.03 0.00 0.01 100.3 6 CHS 40 1108 -1.81 54.6 0.2 0.09 0.03 0.67 0.28 0.83 0.02 11.9 0.7 0.52 0.03 29.8 0.6 1.57 0.30 0.01 0.05 0.00 0.01 99.5 5 CHS 42 1085 -1.65 53.8 0.4 0.16 0.04 0.74 0.20 0.75 0.14 12.6 0.6 0.40 0.03 29.3 0.6 2.17 0.30 0.10 0.03 0.01 0.01 99.8 6 CHS 35 1079 -1.04 53.1 0.7 0.11 0.03 0.82 0.41 0.68 0.08 16.2 1.3 0.37 0.03 27.2 1.4 1.42 0.21 0.03 0.05 0.01 0.00 101.1 3 CHS 37 1063 -0.98 51.7 1.3 0.16 0.61 0.13 0.53 19.2 0.1 0.40 25.9 0.2 1.51 0.71 0.03 0.29 0.02 100.1 2 72 CV CHS 50 1248 -1.89 56.5 0.1 0.07 0.01 0.69 0.11 0.71 0.03 7.8 0.2 0.25 0.02 32.3 0.2 1.66 0.17 0.01 0.00 100.0 5

CHS 63 1202 -1.38 54.3 0.3 0.03 0.04 1.07 0.38 0.89 0.04 13.2 0.7 0.24 0.00 28.1 0.5 2.14 0.20 0.02 0.02 100.3 2 CHS 43 1163 -2.06 55.5 0.6 0.14 0.90 0.44 0.80 8.0 0.1 0.29 0.01 32.0 0.2 2.28 0.25 0.05 0.22 0.00 0.01 100.7 6 CHS 44 1161 -1.86 54.5 1.4 0.15 0.05 1.12 0.24 0.89 0.09 10.0 0.7 0.30 0.03 29.8 0.8 3.14 0.72 0.07 0.04 0.01 0.01 100.0 6 CHS 38 1086 -1.30 55.7 0.9 0.10 0.02 1.23 0.01 0.81 0.05 13.8 0.6 0.20 0.00 25.7 3.7 2.40 0.44 0.05 0.01 0.01 0.01 100.3 2 numbers in italic are the standard errors, total is the original total but all compositions are normalized to 100, n is the number of analyses

Table S5. Composition of pigeonite in experiments

exp T ∆ IW SiO2 TiO2 Al2O3 Cr2O3 FeO MnO MgO CaO Na2O K2O Total n CI CHS 65 1201 -2.27 56.7 0.2 0.09 0.01 0.44 0.03 1.10 0.03 5.2 0.1 0.60 0.05 32.7 0.1 3.12 0.09 0.11 0.02 0.00 0.00 100.3 10 CHS 70 1167 -1.84 55.1 0.7 0.10 0.02 0.62 0.12 1.13 0.01 9.4 0.4 0.60 0.02 28.6 0.3 4.32 0.15 0.13 0.05 0.00 0.00 100.8 6 CHS 68 1159 -1.53 55.0 0.1 0.09 0.03 0.46 0.04 0.84 0.02 12.6 0.6 0.54 0.01 26.2 0.3 4.15 0.01 0.11 0.02 0.01 0.00 100.2 5 CHS 58 1136 -2.27 56.2 0.5 0.14 0.03 0.74 0.01 1.11 0.01 6.9 0.2 0.65 0.02 30.6 0.4 3.58 0.07 0.12 0.03 0.01 0.00 100.1 5 CHS 57 1132 -1.44 54.2 0.5 0.10 0.03 0.64 0.09 0.79 0.04 14.3 0.4 0.53 0.01 25.2 0.4 4.10 0.37 0.14 0.03 0.01 0.00 100.8 4 CHS 54 1129 -1.55 54.3 0.5 0.11 0.02 0.63 0.15 0.74 0.04 12.7 0.4 0.55 0.02 26.1 0.5 4.68 0.40 0.12 0.05 0.01 100.6 7 H CHS 43 1163 -2.06 55.7 0.6 0.15 0.03 0.65 0.12 0.84 0.06 7.0 0.4 0.57 0.04 31.4 0.5 3.56 0.55 0.12 0.03 0.00 0.00 101.0 13 CHS 44 1161 -1.86 55.5 1.0 0.12 0.01 0.43 0.05 0.85 0.05 9.3 0.4 0.55 0.03 29.9 0.4 3.29 0.18 0.11 0.01 0.01 0.00 99.2 6 CHS 39 1140 -1.30 53.6 0.5 0.10 0.02 0.61 0.10 0.77 0.02 15.6 0.3 0.46 0.02 25.0 0.5 3.79 0.25 0.10 0.04 0.01 0.01 101.4 8 CHS 57 1132 -1.44 54.2 0.5 0.13 0.03 0.70 0.09 0.71 0.04 15.5 0.4 0.42 0.01 24.7 0.4 3.67 0.37 0.09 0.03 0.00 0.00 100.1 13 CHS 33 1112 -1.18 52.8 0.6 0.13 0.03 0.76 0.14 0.65 0.06 17.7 0.7 0.37 0.02 22.5 0.5 4.99 0.69 0.13 0.04 101.1 6 LL CHIS 39 1140 -1.3 54.0 0.3 0.06 0.03 0.60 0.11 0.70 0.03 15.0 0.4 0.39 0.02 25.8 0.6 3.61 0.21 0.12 0.03 101.2 7 CHIS 33 1112 -1.18 53.0 0.7 0.15 0.02 0.95 0.05 0.73 0.03 14.8 0.8 0.37 0.02 23.3 0.7 6.54 0.51 0.15 0.03 101.0 3 CM CHS 44 1161 -1.86 55.6 0.3 0.14 0.02 0.49 0.05 0.98 0.04 9.1 0.1 0.48 0.02 29.4 0.1 3.71 0.11 0.06 0.03 0.02 0.01 99.2 5 CHS 39 1140 -1.3 52.6 0.2 0.23 0.02 1.68 0.12 1.07 0.06 14.7 0.3 0.39 0.03 23.9 0.5 5.37 0.64 0.10 0.00 0.00 101.7 7 CHS 38 1086 -1.3 52.5 1.1 0.38 0.06 2.23 0.86 0.82 0.18 13.5 0.7 0.44 0.03 26.0 1.0 3.87 0.47 0.21 0.24 0.02 0.03 99.8 10 CV CHS 39 1140 -1.30 52.5 0.4 0.18 0.02 1.89 0.22 1.10 0.13 14.8 0.4 0.28 0.03 23.2 0.6 5.95 0.77 0.09 0.05 0.00 0.00 101.3 7 numbers in italic are the standard errors, total is the original total but all compositions are normalized to 100, n is the number of analyses 73

Table S6. Composition of chromite in experiments

exp T (ºC) ∆ IW SiO2 TiO2 Al2O3 Cr2O3 FeO MnO MgO CaO NiO Total n H CHS 35 1079 -1.04 1.9 2.8 10.0 46.1 29.4 0.3 6.1 0.1 0.1 96.9 2 CH 8* 1076 0.56 5.7 1.2 10.6 46.5 28.2 0.3 4.5 0.1 97.0 1 CM CHS 35 1079 -1.04 4.0 4.5 10.1 41.6 30.5 0.3 8.1 0.1 0.1 99.2 1 CH 10* 1118 0.71 1.4 1.3 19.4 40.7 28.7 0.2 7.2 0.0 2.0 101.0 2 small chromite crystals could not be analyzed without contamination from the surrounding silicates

Table S7. Composition of augite in experiments

exp T ∆ IW SiO2 TiO2 Al2O3 Cr2O3 FeO MnO MgO CaO Na2O K2O Total n CI CHS 58 1136 -2.27 53.5 0.1 0.29 0.04 1.46 0.34 1.25 0.24 5.3 1.2 0.39 0.05 19.0 0.4 18.4 0.7 0.41 0.08 0.02 0.01 101.5 4 CHS 60 1104 -1.63 52.8 0.7 0.35 2.06 0.39 1.74 7.5 0.1 0.47 0.01 20.3 0.1 14.3 0.1 0.38 0.31 0.01 0.03 100.7 2 CHS 59 1100 -2.19 52.6 0.6 0.40 0.02 2.70 0.30 1.91 0.28 5.9 0.1 0.49 0.02 20.2 0.1 15.3 0.6 0.49 0.02 0.02 0.00 101.6 3 H CHS 41 1142 -2.00 53.2 0.4 0.35 0.02 1.38 0.06 1.17 0.04 7.2 0.4 0.51 0.09 26.1 0.5 9.8 0.2 0.34 0.04 0.00 0.00 99.6 1 CHS 40 1108 -1.81 53.8 0.4 0.54 5.26 0.51 1.41 5.5 0.33 16.5 15.6 0.87 0.21 99.5 1 CHS 42 1085 -1.65 52.2 0.5 0.45 1.65 0.05 1.42 8.5 0.40 18.8 16.1 0.46 0.01 102.0 1 CHS 35 1079 -1.04 52.2 0.8 0.40 0.03 1.63 0.08 1.22 0.06 11.9 0.7 0.27 0.02 16.6 0.9 15.4 0.2 0.42 0.01 0.01 0.01 101.3 6 CHS 37 1063 -0.98 50.8 1.1 0.51 0.08 2.36 0.47 1.29 0.19 12.2 0.6 0.26 0.01 15.7 0.4 16.5 0.9 0.47 0.07 0.01 0.02 100.3 7 CH 8* 1076 0.56 50.8 0.8 0.41 0.01 2.65 0.21 0.00 12.9 0.1 0.22 0.00 15.1 0.0 17.5 0.4 0.33 0.07 0.00 98.8 2 CH 5* 1073 0.76 50.5 0.8 0.48 0.05 3.13 0.50 1.30 0.08 11.5 0.3 0.19 0.01 13.9 0.3 18.6 0.3 0.36 0.05 0.00 100.8 5 LL CHIS 41 1142 -2.00 53.2 0.34 1.59 1.17 5.6 0.45 24.9 12.4 0.39 0.01 101.3 1 CHIS 42 1085 -1.65 52.2 0.45 1.65 1.42 8.5 0.40 18.8 16.1 0.46 0.01 102.0 1 CHIS 37 1063 -0.98 51.8 0.4 0.39 0.05 1.96 0.34 1.01 0.26 12.0 0.9 0.35 0.03 16.4 1.3 15.6 1.2 0.41 0.10 0.01 0.00 101.2 5 CM CHS 43 1163 -2.06 53.1 0.3 0.55 0.08 2.86 1.52 1.18 0.16 4.6 0.1 0.39 0.04 21.8 1.6 15.2 0.4 0.35 0.27 0.01 0.01 101.1 3 CHS 41 1142 -2.00 51.3 0.6 0.72 0.08 3.88 0.94 1.46 0.10 5.1 0.5 0.35 0.03 20.0 2.4 16.7 1.0 0.46 0.23 0.01 0.01 102.0 4 CHS 33 1112 -1.18 50.6 1.1 0.46 0.05 2.81 0.33 1.47 0.20 12.0 1.1 0.29 0.01 17.4 0.8 14.8 1.3 0.26 0.03 100.4 8 CHS 40 1108 -1.81 48.2 0.2 0.69 0.03 4.68 0.28 2.06 0.02 9.0 0.7 0.33 0.03 17.3 0.6 17.5 0.3 0.30 0.05 0.01 0.01 97.8 1 CHS 38 1086 -1.30 51.4 0.3 0.62 0.02 3.71 0.57 1.65 0.36 8.1 0.8 0.37 0.07 18.6 0.9 15.3 1.7 0.28 0.01 0.00 0.00 100.2 2 CHS 42 1085 -1.65 49.2 0.4 0.48 0.04 2.54 0.20 1.74 0.14 8.5 0.6 0.38 0.03 22.8 0.6 13.9 0.3 0.31 0.03 0.00 0.01 100.9 1 74 CHS 35 1079 -1.04 49.9 0.4 0.70 0.06 4.68 0.59 1.97 0.30 9.2 0.2 0.21 0.02 14.6 0.6 18.4 0.7 0.34 0.04 0.01 0.00 100.9 8

CHS 37 1063 -0.98 49.5 1.6 0.60 0.20 4.09 0.88 1.53 0.23 11.1 0.9 0.24 0.02 15.2 0.6 17.4 0.8 0.34 0.03 0.00 0.00 100.6 5 CH 10* 1118 0.71 51.2 0.4 0.25 0.06 1.62 0.20 0.77 0.08 15.5 0.9 0.29 0.02 19.8 0.6 10.5 1.1 0.09 0.01 0.02 101.4 4 CV CHS 43 1163 -2.06 52.8 0.71 2.52 1.04 4.1 0.27 19.2 19.1 0.24 101.0 1 CHS 44 1161 -1.86 52.5 0.54 2.87 1.28 5.7 0.21 19.3 17.3 0.25 0.02 100.8 1 CHS 39 1140 -1.30 51.8 0.7 0.30 0.05 2.26 0.36 1.09 0.14 12.1 0.6 0.26 0.02 20.5 0.7 11.5 0.9 0.12 0.02 0.00 0.00 100.6 5 CHS 33 1112 -1.18 50.9 1.2 0.42 0.06 2.40 0.29 0.97 0.14 13.5 0.2 0.22 0.02 17.0 0.3 14.4 0.5 0.18 0.04 100.5 5 CHS 38 1086 -1.30 49.9 1.0 0.73 0.25 3.64 1.02 1.01 0.36 12.8 1.5 0.22 0.05 15.5 0.4 16.0 1.0 0.27 0.03 0.01 0.01 100.2 3 numbers in italic are the standard errors, total is the original total but all compositions are normalized to 100, n is the number of analyses

Table S8. Composition of plagioclase in experiments original exp T ∆ IW SiO TiO Al O FeO MnO MgO CaO Na O K O n 2 2 2 3 2 2 total CI CHS 60 1104 -1.63 59.45 0.7 23.68 0.4 1.17 0.07 0.05 0.01 1.25 0.12 6.68 0.11 7.56 0.31 0.16 0.03 98.3 4 CHS 59 1100 -2.19 60.53 0.4 23.16 0.0 1.07 0.04 0.00 0.01 1.41 0.16 5.93 0.28 7.69 0.21 0.21 0.06 100.3 2 H CHS 41 1142 -2.00 56.89 0.4 26.22 0.7 1.09 0.03 0.03 0.05 0.62 0.48 8.84 0.20 6.16 0.10 0.14 0.01 100.8 3 CHS 33 1112 -1.18 57.64 1.2 25.38 0.8 1.53 0.61 0.73 0.72 8.22 0.10 6.33 0.45 0.16 0.02 99.2 3 CHS 40 1108 -1.81 59.64 0.2 24.04 0.2 1.07 0.06 0.04 0.01 1.26 0.07 7.31 0.45 6.43 0.37 0.21 0.01 99.0 1 CHS 42 1085 -1.65 60.73 0.3 23.80 0.4 0.98 0.11 0.02 0.02 0.33 0.17 5.85 0.24 7.88 0.37 0.41 0.02 100.5 4 CHS 35 1079 -1.04 60.51 0.5 0.07 0.03 23.26 0.4 1.41 0.18 0.52 0.22 5.91 0.54 7.91 0.26 0.32 0.02 100.5 8 CHS 37 1063 -0.98 60.47 1.6 23.02 1.3 1.77 0.51 0.88 0.55 5.11 0.59 8.27 0.36 0.49 0.15 99.9 2 CH 5* 1073 0.76 56.82 0.6 25.88 0.2 1.63 0.16 0.01 0.00 0.53 0.20 8.55 0.05 6.33 0.08 0.24 101.1 2 LL CHIS 41 1142 -2 56.82 26.43 0.80 0.01 0.70 8.58 6.54 0.13 95.1 1 CHIS 33 1112 -1.18 59.34 0.0 24.60 0.3 0.98 0.05 0.52 0.25 7.16 0.27 7.26 0.18 0.14 0.00 98.6 2 CHIS 40 1108 -1.81 60.04 0.9 24.24 0.6 0.78 0.12 0.03 0.02 0.59 0.38 6.42 0.48 7.76 0.33 0.14 0.07 100.1 2 CHIS 42 1085 -1.65 60.79 22.64 1.34 0.07 2.18 5.77 6.96 0.25 102.8 1 CHIS 37 1063 -0.98 62.78 1.6 21.81 1.3 1.19 0.51 0.57 0.55 4.76 0.59 8.59 0.36 0.30 0.15 99.2 8 CM CHS 43 1163 -2.06 52.88 0.2 28.67 0.1 1.24 0.12 0.01 0.01 0.56 0.28 12.22 0.04 4.32 0.15 0.10 0.00 100.6 2 CHS 44 1161 -1.86 53.39 0.3 28.83 0.2 0.91 0.07 0.02 0.01 0.33 0.12 11.98 0.37 4.47 0.20 0.07 0.01 100.5 4 CHS 41 1142 -2 56.90 24.60 1.87 1.96 9.24 4.87 0.57 100.1 1 CHS 39 1140 -1.3 53.76 0.2 27.93 1.0 1.64 0.13 0.01 1.00 11.67 0.67 4.45 0.40 0.08 95.9 2 75 CHS 33 1112 -1.18 55.46 0.6 26.41 0.2 1.53 0.45 0.89 0.47 9.98 0.30 5.54 0.25 0.20 0.02 99.4 5

CHS 40 1108 -1.81 56.38 0.9 26.46 0.6 1.00 0.12 0.03 0.02 0.68 0.38 9.63 0.48 5.55 0.33 0.27 0.07 99.3 6 CHS 38 1086 -1.3 56.57 0.9 26.42 0.4 0.97 0.17 0.02 0.47 0.40 9.41 0.15 5.71 0.14 0.44 0.01 100.3 6 CHS 42 1085 -1.65 55.70 0.3 25.19 0.4 2.08 0.11 0.03 0.02 1.84 0.17 9.26 0.24 5.51 0.37 0.39 0.02 99.8 2 CHS 35 1079 -1.04 56.14 0.5 0.05 0.03 25.77 0.4 1.53 0.18 0.59 0.22 9.75 0.54 5.55 0.26 0.45 0.02 99.7 5 CHS 37 1063 -0.98 56.51 0.4 24.80 0.3 2.11 0.18 0.04 1.37 0.22 8.74 0.38 6.03 0.20 0.49 0.06 98.7 3 CH 10* 1118 0.71 49.86 0.3 27.58 0.3 3.73 0.20 0.05 0.02 2.24 0.25 13.52 0.19 2.96 0.08 0.06 0.00 99.2 2 CV CHS 43 1163 -2.06 52.16 0.6 29.38 0.4 0.82 0.11 0.03 0.01 0.63 0.19 13.20 0.25 3.70 0.22 0.08 0.01 100.2 4 CHS 44 1161 -1.86 51.08 0.6 30.13 0.3 1.06 0.21 0.05 0.04 0.49 0.20 13.58 0.29 3.53 0.13 0.08 0.02 100.0 5 CHS 39 1140 -1.30 52.81 0.2 29.30 0.0 0.89 0.21 0.01 0.01 0.29 0.04 12.43 0.06 4.21 0.09 0.07 0.01 100.6 2 CHS 33 1112 -1.18 54.00 1.6 28.03 0.5 0.84 0.31 0.31 0.09 12.02 0.84 4.65 0.58 0.14 0.03 98.9 5 CHS 38 1086 -1.30 54.48 0.5 27.19 0.4 1.37 0.14 0.75 0.23 11.10 0.38 4.91 0.14 0.20 0.02 101.0 4 numbers in italic are the standard errors, total is the original total but all compositions are normalized to 100, n is the number of analyses

Chapter 2: Formation of primitive achondrites by partial melting of alkali-undepleted planetesimals in the inner solar system

Abstract

Acapulcoites-lodranites, ureilites, brachinites and winonaites are the main groups of primitive achondrites. They are variably depleted in incompatible lithophile elements (Al, Na, K and rare earth elements) and siderophile/chalcophile elements relative to chondrites and are interpreted as the residual mantle of planetesimals from which silicate melts and sulfide/metal melts were extracted. Primitive achondrites recorded the onset of planetesimal melting and a newly identified group of “trachyandesite” achondrites (e.g. ALM-A and GRA 06128) suggests that the partial melts were rich in SiO2 and alkali elements. Here, we use a series of melting experiments conducted with various chondritic compositions (CV, CM, CI, H and LL) to constrain (1) the oxygen fugacity (fO2) and temperature at which melting occurred, (2) the bulk composition of the initial chondritic material and (3) the composition of the average silicate melts produced in the different parent bodies. Primitive achondrites melted at different and variable fO2: ∆IW -0.5/-1.0 for brachinites, ∆IW -1.3/-2.5 for ureilites, ∆IW -1.6/-2.7 for acapulcoites/lodranites and ∆IW -

2.5/-3.0 for winonaites (with ∆IW = log fO2 – (log fO2)IW; IW being the iron-wustite buffer). Primitive achondrite parent bodies had initial Mg/Si and (Na+K)/(Ca+Na+K) ratios similar to ordinary chondrites, CI chondrites and the sun’s photosphere. We infer that all parent bodies with oxygen isotope composition and Cr-Ti-Ni nucleosynthetic anomalies characteristic of the “non- carbonaceous” group of meteorites were not initially depleted in Na2O and K2O. The depletion of moderately volatile elements in other parent bodies (e.g. Vesta, angrites parent body) and planets does not seem to result from the incomplete condensation of the solar nebula and could have occurred during large extents of partial melting and differentiation that affected small planetary bodies.

1. Introduction

Primitive achondrites were initially defined as having bulk chemical compositions similar to chondrites but textures indicative of high temperature equilibration and limited extents of igneous processing. This definition was based on the description of Acapulco (Palme et al., 1981), Winona (Bild, 1977) and Brachina (Nehru et al., 1983), the type specimens of traditional primitive achondrite groups. Prinz et al. (1983) included several winonaites-IAB irons, Acapulco, Lodran and Brachina in their brief review of primitive achondrites and argued that silicate partial melting had been limited. However, most brachinites and brachinite-like achondrites described subsequently were depleted in Al2O3 and light rare earth elements (LREE) and many samples are completely devoid of plagioclase (Warren and Kallemeyn, 1989; Keil, 2014). Lodranites, which are derived from the same parent body as acapulcoites, also lack plagioclase (Keil and McCoy, 2018). Various models have been proposed to explain the formation of primitive achondrites but all major groups are now widely believed to represent melting residues that have lost a fraction of silicate melt and Fe-Ni-S eutectic melt (Keil, 2014; Hunt et al., 2017; Keil and McCoy, 2018). The nature of ureilites has been highly debated but they are increasingly interpreted as the residues of partial melting of a large (>100 km in radius) planetesimal (Goodrich, 1992; Warren and Kallemeyn, 1992; Scott et al., 1993; Walker and Grove, 1993; Mittlefehldt et al., 1998). While the detailed petrogenesis of ureilites remains controversial, the view that they represent melting residues has strengthened with time (Barrat et al., 2016; Goodrich et al., 2013b; Goodrich et al., 2007; Warren, 2012). Therefore, following Weisberg et al. (2006), we treat them as the most abundant group of primitive achondrites. Other workers consider that the absence of plagioclase in all samples and the textural evidence of igneous processing in ureilites justifies grouping them with differentiated achondrites in classification schemes (e.g. Krot et al., 2013). It could be argued that the use of the term primitive achondrites to refer to asteroidal melting residues is improper but there is little justification for discriminating between ureilites, lodranites and brachinites on this basis. The parent bodies of winonaites, acapulcoites-lodranites, brachinites and ureilites all reached internal temperatures that were in excess of the FeNiγ-FeS cotectic curve (940-960 ºC; Kullerud, 1963) and the silicate solidus (1040-1080 ºC; Chapter 1) but that were too low to produce extensive melting and complete differentiation. We propose that the term primitive achondrites be used to refer to all these meteorites, regardless of the precise fraction of melt extracted from the residues.

All groups of primitive achondrites display a heterogeneity in oxygen isotopes (∆17O = δ17O – 0.52 δ18O; Clayton and Mayeda, 1988), likely inherited from nebular processes such as CO self-shielding (Clayton, 2002), of the same magnitude as groups of chondrites (Greenwood et al., 2017). In contrast to differentiated achondrites, like eucrites and angrites, oxygen isotopes were not re-homogenized at the scale of the parent bodies of primitive achondrites. The variable ∆17O values are also in agreement with their origin as residues of partial melting and the absence of sustained and interconnected networks of silicate melt allowing for rapid O diffusion. Primitive achondrites are also distinct from differentiated achondrites in their relative concentrations of Fe, Mn and Mg. Primitive achondrites show Fe variations at constant Mn/Mg ratios interpreted as resulting from the heterogenous distribution of Fe or variable intrinsic fO2 in their starting materials. On the other hand, differentiated meteorites such as eucrites and Martian meteorites form trends of near constant Fe/Mn ratios with increasing Fe/Mg, which are indicative of igneous differentiation (Goodrich and Delaney, 2000). In this manuscript, we continue to discuss the results of an extensive set of experiments performed in Molybdenum Hafnium Carbide pressure vessels (MHC-PV) to characterize the melting behavior of chondritic planetesimals. The experimental and analytical methods are described in Chapter 1. In that first paper, we show that the partial melts of high-NaK# chondrites (e.g. CI, H and LL chondrites; NaK# = (Na+K)/(Na+K+Ca) in atom.%) likely represent the parental melts of trachyandesite achondrites as well as smaller clasts rich in sodic plagioclase (oligoclase, plagioclase with a anorthite content of 10 to 30) found in polymict ureilites and some plagioclase-rich inclusions in non-magmatic irons. In this paper, we compare experimental phase assemblages to primitive achondrites and show that their parent bodies, including the ones of brachinites and ureilites, were not initially depleted in alkalis (NaK# = 50, as in CI, H and LL chondrites) and therefore produced melts rich in SiO2, Al2O3 and alkali elements by melting at various fO2 (∆IW -0.5/-3.0). Based on the evidence provided by trachyandesite achondrites and primitive achondrites, we propose that most planetesimals that accreted in the inner solar system were not depleted in moderately volatile elements (MVE, e.g. Na and K), relative to the sun’s photosphere (NaK# = 50, identical to CI, H and LL chondrites).

2. Mineral compositions and proportions in experiments

Melting experiments of synthetic CI, H, LL, CM and CV compositions are extensively described in the companion manuscript (Chapter 1), which includes a description of mineral assemblages and tables of mass balance calculations and mineral compositions. Here, we remind the reader of the phases that are in equilibrium in experimental charges as a function of the bulk chondritic composition, temperature and fO2 (Table 1, Fig. 1) and highlight important trends that, as discussed in the next sections, are also observed in primitive achondrites. The super-solidus MHC-PV experiments performed on all five starting compositions contain olivine, Fe-Ni metal and, in most cases, at least one pyroxene: orthopyroxene and augite at low temperature (1050-1150 ºC), pigeonite at moderate temperature (1130-1180 ºC) and orthopyroxene at high temperature (>1200 ºC). Plagioclase is only present at low temperature (<1160 ºC in CV experiments and <1130 ºC in H, LL and CI experiments). The FeO content of ferromagnesian silicates decreases with decreasing fO2 (the Mg#, Mg/(Mg+Fe) in atom.%, increases with decreasing fO2). For example, olivine has a forsterite (Fo) content of 75 at ∆IW - 1.3 and a Fo content of 95 at ∆IW -2.5 (Fig. 2a). As the total concentration of FeO + Fe of the bulk compositions is constant, the reduction of Fe2+ to Fe0 leads to an increase of the fraction of metal in experiments (Fig. 3) and, as a result, the Ni content of the metal is progressively diluted (Fig. 2b). The solidus temperature and the temperatures of phase disappearance increases with the Mg# (i.e. plag-out, pig-out, and particularly, px-out; Fig. 1). In experiments, the proportion of pyroxene relative to olivine or “py”, defined as the total fraction of pyroxene divided by the sum of pyroxene and olivine (px/(px+oliv) in wt.%), decreases with increasing extent of melting, as pyroxene is consumed by melting reactions (Chapter 1). The py also increases with decreasing fO2 as the bulk SiO2/(FeO+MgO) ratio increases in response to 2+ the reduction of Fe to Fe metal (Fig. 4a). The Cr2O3 content of olivine increases with temperature from 0.2-0.3 wt.% at 1100 ºC to 0.6-0.8 at 1250 ºC between ∆IW -1 and -2.5 (MHC-PV experiments). The NiO content of olivine is negligible (<0.1 wt.%) at conditions more reducing than ∆IW -1, regardless of the temperature, but is high (0.6-0.9 wt.%) at ∆IW +0.6/+0.8 (Fig. 2c).

3. Experimental constraints on the fO2 of melting of primitive achondrites

The composition of chondrites and primitive achondrites show that bulk Fe contents (i.e. Fe + FeO) varied both across and within planetesimals (Lodders and Fegley, 1998; Goodrich and Delaney, 2000), possibly as a result of small differences in the condensation temperature of metal and silicates (Ebel and Grossman, 2000) or metal-silicate fractionation in the nebula (e.g. Wurm et al., 2013). Among primitive achondrites, the within-group variability of bulk Fe contents could result from the migration of FeNi(-C) melts during melting (Goodrich et al., 2013; McCoy et al.,

2006, 1997b). The FeO content of silicates (their Mg#), which is thought to reflect the fO2 of equilibration, also varies across and within groups of equilibrated chondrites and primitive achondrites. It does not always correlate with the bulk Fe content. Forming FeO-rich silicates by direct condensation of the solar nebula is non-trivial due to the very low O/H ratio of the gas (fO2 < ∆IW -6). High “dust-to-gas ratios” (100-1000×) with dust rich in water ice might be necessary to increase the O/H and O/C ratios sufficiently to reach a fO2 of ∆IW -1/-3 (Ebel and Grossman, 2000; Fedkin and Grossman, 2016). FeO-rich (type II) chondrules in carbonaceous chondrites (Mg# 50-90) could have formed in this type of environment (Tenner et al., 2018). It is possible that the fO2 of melting of primitive achondrites was not “inherited” from nebular processes but instead was set in the parent bodies (Ringwood, 1961). One possibility is that graphite controlled the fO2 (C-CO buffer; French and Eugster, 1965) at low pressure (<10-15 MPa), a process sometimes called “smelting” and which leads to more reducing conditions (higher Mg#) at higher temperature (Walker and Grove, 1993; Rubin, 2007). Sanders et al. (2017) infer that low- temperature fluid alteration can increase the fO2 of parent bodies prior to melting. We will not re- evaluate those different models here but we will demonstrate that the Fo content of olivine, which varies across and within primitive achondrite groups from Fo65 to Fo95, is directly controlled by the fO2 conditions during melting, regardless of its origin. Brachinites and brachinite-like achondrites are known to form at more oxidizing conditions than acapulcoites-lodranites and winonaites based on their FeO-rich olivine and pyroxene (Mg# = 60-80 vs. 85-95 and 93-98) and Ni-rich metal (15-60 wt.%; Burkhardt et al., 2017; Day et al.,

2019, 2012; Keil, 2014). Ureilites are thought to form under fO2 conditions intermediate between those of brachinites and winonaites. However, attempts to constrain their precise fO2 of melting has been entangled in the debate over primary smelting (Goodrich et al., 2013b; Mittlefehldt et al., 1998; Warren, 2012). Some opponents of the smelting model have cast doubt over the existence

81 of a correlation between py and Fo content in ureilites. The experiments show that melting a chondritic material at different fO2 indeed produce a correlation between py and Fo in the residues, as long as the extent of melting is relatively constant (Fig 4a). Primitive achondrites as a whole form a trend that is nearly identical, albeit with significant scatter (Fig. 4 b-d). Ordinary chondrites (petrologic type 6-7) also form a trend of increasing py with Fo content but displaced towards higher py. The positive correlation of Fo content and py across primitive achondrite groups results from: (1) varying fO2 conditions that control the Fo content in olivine and, to first order, the relative proportion of pyroxene and olivine (py); (2) the lower py relative to ordinary chondrites results from the loss of silicate melt enriched in a pyroxene component (Chapter 1); and (3) the scatter results from slight variations in the Mg/Si of the parent bodies and in the extent of melting experienced by individual samples, but also in part from analytical uncertainties. All primitive achondrites contain FeO-bearing olivine and at least traces of metal, which implies that their fO2 can be calculated to first order by using the OSI equilibrium (Fig. 2).

Following this approach, the fO2 of melting of acapulcoites-lodranites, containing 10 vol.% of metal on average, and winonaites, containing 15 vol.% of metal on average, is calculated to be IW -2.2 ± 0.3 and IW -2.7 ± 0.3, respectively (Benedix et al., 2005; Keil and McCoy, 2018). In agreement with those calculations, our experiments performed between IW -1.8 and IW -2.4 contain olivine Fo85-94, a range which overlaps with the composition of olivine in acapulcoites- lodranites (Fo85-96) but only marginally with the composition of olivine in winonaites (Fo93-98).

The fO2 of equilibration of primitive achondrites with lower proportions of metal (trace amounts to 5 vol.%) is slightly more controversial. Brachinites, brachinite-like achondrites and chromite-bearing ureilites (Mg# in silicates = 60-80) are sometimes thought to have melted between IW +1 and IW +3 (Day et al., 2019; Goodrich et al., 2013a; Goodrich et al., 2017), a range which corresponds to the intrinsic fO2 of carbonaceous chondrites. However, equilibration of brachinites and ureilites above the IW buffer is inconsistent with their NiO contents in olivine. Our experiments show that the NiO content of olivine is of the order of 1 wt.% at IW +0.8 (Fig 2c). With the exception of a couple of ungrouped samples with carbonaceous chondrite affinities (e.g. NWA 8186; Srinivasan et al., 2016, NiO=0.8 wt.%), olivines in all primitive achondrites, including brachinites and chromite-bearing ureilites, have Ni contents under 1000 ppm and generally close to the detection limit of standard EPMA analyses. Under oxidizing conditions, Ni is compatible in olivine and its concentration is not expected to decrease with retrograde metamorphism. The

82 oxidized R and CK chondrites contain olivine with 0.25 and 0.5 wt.% NiO, respectively (Righter and Neff, 2007; Bischoff et al., 2011). In CK chondrites, NiO contents do not vary significantly with temperature and are identical in samples of petrologic types 4 to 6. Therefore, the low NiO content of olivine in the large majority of primitive achondrites, including brachinites and ureilites, indicate that they melted at fO2 conditions more reducing than IW.

Ureilites with the lowest forsterite contents in olivine (Fo75-80) melted at IW -0.5/-1.5, depending on the precise Ni and Fe activities in the metal phase in equilibrium (0.4-0.6 and 0.1-

0.9, respectively; Fig. 2b). Ureilites with olivine Fo80-95 melted under conditions progressively more reducing, consistent with the OSI buffer (e.g. IW -2.1 for Fo90 ureilites). For brachinites, equilibration at fO2 IW -0.5/-1 is in agreement with the thermodynamic calculations of Gardner- Vandy et al. (2013), NiO contents in olivine (Fig. 2c) and XANES measurements of V in spinels (Righter et al., 2016).

4. Major-element composition of parent bodies and their partial melts

Because most primitive achondrites cooled relatively slowly, they recorded low (< 1000 ºC) two-pyroxene and olivine-chromite temperatures (Benedix et al., 2005; Keil, 2014; Keil and McCoy, 2018), well below the silicate solidus. The composition of these phases cannot be directly compared to those of melting experiments. Ureilites represent an important exception. They are believed to have cooled 5-10 orders of magnitude faster, as fast as 20 ºC/h (Miyamoto, 1985;

Takeda et al., 1989) and are the only primitive achondrites with Cr2O3 and CaO contents in olivine that match the ones of melting experiments (Fig. 5). To constrain the melting behavior of the other parent bodies, we compare the relative proportion of pyroxene and olivine (py) and the composition of plagioclase between primitive achondrites and experiments. The average composition of the silicate melt produced in the different parent bodies is reported in Table 2.

4.1. Acapulcoites-lodranites

Acapulcoites and lodranites are the archetypes of primitive achondrites. They contain orthopyroxene (30-50 vol.%), olivine (25-40 vol.%), variable proportions of metal (5-15 vol.%), some clinopyroxene, plagioclase, sulfide and other accessory phases (Keil and McCoy, 2018). The presence of relic chondrules (e.g. Schrader et al., 2017) suggest that some acapulcoites are essentially “metachondrites” that did not lose a significant fraction of silicate melt. On the contrary, many lodranites lack plagioclase, are depleted in rare earth elements (REE) relative to 83 chondrites, and are thought to have lost up to 15 wt.% of silicate melt (McCoy et al., 1997a; Floss, 2000). The distribution of HSE and the occurrence of large cm-sized metal veins suggests that both FeS and FeNi melts migrated and likely pooled deeper within the parent body (McCoy et al., 1997b; McCoy et al., 2006; Dhaliwal et al., 2017). The most primitive acapulcoites, including Acapulco (Keil and McCoy, 2018), and a single chondrite of petrologic type 4, Grove Mountains (GRV) 020043 (Li et al., 2018; McCoy et al., 2019), can be used to constrain fairly accurately the initial bulk composition of the parent body. Those meteorites form a cluster on the py vs. Fo diagram (Fig. 4b) that overlaps with the experimental trends of sub-solidus ordinary chondrites. Accordingly, the molar Mg/Si of the parent body of acapulcoites and lodranites was likely identical to the one of H chondrites (0.95). In addition, the NaK# of the most primitive acapulcoites, which averages at 49 (Keil and McCoy, 2018), is identical to the NaK# of GRV 020043 and the NaK# of ordinary chondrites. Therefore, the H composition is a good analog for the composition of the parent body of acapulcoites- lodranites. It melted under slightly more reducing conditions (IW -2.2 ± 0.3) than the intrinsic fO2 of H chondrites. Silicates started to melt at 1090 ± 10 ºC and plagioclase disappeared from residues at 1150 ± 10 ºC, after 12-14 wt.% of melting (Fig. 1). Because, most lodranites are devoid of plagioclase, they probably lost more than 14 wt.% of silicate melt and reached peak temperatures in excess of 1150 ºC and potentially as high as 1200-1240 ºC (e.g. H composition in experiments CHS 27, 62 and 65). The partial melt extracted from acapulcoites-lodranites had high SiO2 (61-65 wt.%),

Al2O3 (13-16.5 wt.%) and Na2O (4.5-6.5 wt.%) concentrations and low FeO (3-6 wt.%), MgO and CaO (2-7 wt.%) concentrations (Table 2; see also Chapter).

4.2. Winonaites

Winonaites, which are cogenetic with IAB iron meteorites, have been long recognized as primitive achondrites having experienced very little melt extraction but showing some textures indicative of partial melting (Benedix et al., 1998; Benedix et al., 2005; Hunt et al., 2017), such as fine-grained pockets enriched in plagioclase and high-Ca pyroxene. The bulk composition of most primitive samples should thus represent the bulk composition of the parent body. Based on the meteorites Fortuna, Pontlyfni, HaH 192 and NWA 1463 analyzed by Hunt et al. (2017), we calculate an average NaK# of 53, Mg/Si of 0.95 and a Mg# of 59. Therefore, in terms of major

84 elements, this composition is also nearly identical to an H chondrite and the parent body of acapulcoites-lodranites. This is at odds with the conclusions of Hunt et al. (2017) who suggest a precursor material with carbonaceous affinities, similar to CM chondrites. Our experiments on the CM composition (NaK# of 33, Mg/Si ratio of 1.03) show that, if this was the case, the plagioclase would be much more calcium rich (An40) near the solidus (Fig. 4c) than the plagioclase in winonaites (An15), which is characteristic of chondrites with no depletion in MVE (Fig. 4b). Some earlier studies have argued for a bulk composition intermediate between ordinary and enstatite chondrites, based on their noble gas compositions and low fO2 (Benedix et al., 1998). However, a lower Mg/Si ratio (<0.9) would result in a higher py than the py of winonaites.

Winonaites appear to have melted at more reducing conditions (log fO2 < IW -2.5) than lodranites and most samples melted to a much lower extent (< 3 wt.%) based on the chondritic nature of plagioclase (proportion and composition) and based on bulk REE patterns (Hunt et al., 2017). Locally, melting could have been more important (20-30 wt.%) and produced the silicate inclusions rich in pyroxene and plagioclase of some IAB irons (e.g. Caddo County and Udei Station; Benedix et al., 2000; Ruzicka and Hutson, 2010). Larger degrees of partial melting could also account for the poikilitic textures with oikocrysts of high-Ca pyroxene containing chadacrysts of olivine and low-Ca pyroxene observed in the winonaite Tierra Bianca (Benedix et al., 1998). Due to the high NaK# of the parent body, the low-degree partial melts would have been rich in

SiO2, Al2O3 and alkalis.

4.3. Brachinites and brachinite-like achondrites

Brachinites are achondrites consisting mostly of olivine (71-96 vol.%), and clinopyroxene (up to 15 vol.%). Most samples contain trace amounts of Fe-Ni metal, Fe-sulfide, chromite and phosphate and about half of them contain plagioclase (traces to 10 vol.%) and minor orthopyroxene (Keil, 2014). Brachinite-like achondrites (BLA) were defined as meteorites with similar oxygen isotope compositions and a mineralogy also dominated by olivine but with a higher forsterite content (>70) than brachinites. BLA can also contain a larger fraction of orthopyroxene (Day et al., 2012). Both groups of meteorites appear to share identical Mg/Mn ratios and sample the same nebular reservoir (Goodrich et al., 2017; Day et al., 2019) but whether they actually formed on the same parent body remains debatable.

Among brachinites and BLA, Brachina represents the most primitive sample. It displays a smaller grain-size, contain sodic-plagioclase (An20) and has chondritic REE concentrations (Nehru et al., 1983). A plagioclase composition of An20 is nearly identical to the near-solidus plagioclase of CI, H and LL compositions (Fig. 4b-d). Therefore, the initial composition of the brachinites parent body was likely not significantly depleted in MVE (NaK# of 45-50). A NaK# similar to the one of the CM composition (33) would have stabilized andesine (An40-50) in lieu of oligoclase (Fig. 4c). If the parent body of brachinites had a bulk Mg/Si similar to ordinary chondrite (0.95), plagioclase would have melted out of the residue when py was between 0.1 and 0.2, which is slightly higher than the pyroxene fraction in brachinites (Fig. 4c). If the parent body had a Mg/Si closer to the one of carbonaceous chondrites (1.03), the residue would have contained plagioclase until a py of 0.03-0.1, which is closer to the average proportion of pyroxene in brachinites (Fig. 4d). We thus infer that the Mg/Si ratio of the brachinite parent body was higher than the one of ordinary chondrites, acapulcoite-lodranite and winonaite parent bodies and in the range 1-1.03, which is identical to the Mg/Si ratio proposed by Dauphas et al. (2015) based on Si isotopes. Upon melting at IW -1, a chondritic material with high NaK# and relatively high Mg/Si ratio would produce abundant alkali- and silica- rich melts (Table 2). The trachy-andesite GRA 06128/9 probably represents the crystallization products of such melts (Chapter 1). Melting would have started at 1040 ± 10 ºC and plagioclase disappeared at 1105 ± 10 ºC, after 14 ± 2 wt.% of melting. As the melt fraction increased, the py decreased and the anorthite content in the residual plagioclase increased from An20 to An40, a stage that could be recorded by brachinites with minor amounts (1-3 wt.%) of andesine (An35-40; e.g. NWA 4874 and NWA 4969). Gardner-Vandy et al. (2013) and Lunning et al. (2017) argue that brachinites and GRA 06128/9 could have formed by melting a R-like parent body under more oxidizing conditions (IW/IW+1). In detail, the relatively low Mg/Si ratio of R chondrites (0.96), similar to ordinary chondrites, is inconsistent with the low proportions of pyroxene in brachinites. R chondrites are also characterized by a higher intrinsic fO2 (see section 3) and, as acknowledged by Lunning et al. (2017), different oxygen isotopes relative to brachinites. However, R chondrites also have a high NaK# (44; Jarosewich, 2006). If they had melted, they would have produced partial melts at least as rich in silica and alkali elements as the melts of experiments performed in gas-mixing furnaces (IW +0.8) on the H composition (Table 2).

4.4. Ureilites Ureilites are the largest group of primitive achondrites (550 samples) and appear to have suffered relatively large extents of partial melting and melt extraction. No sample approaching the composition of the primitive chondritic material has been found to date. Most ureilites contain two silicate phases – olivine and pyroxene (usually pigeonite) – with homogeneous cores but more Mg-rich rims, abundant graphite (~3 vol.%) and metal (1 to 6 vol.%). For comparison, acapulcoites-lodranites and winonaites contain only traces of graphite, usually significantly less than 1 wt.% (Benedix et al., 2005; Keil and McCoy, 2018), and brachinites contain no graphite (e.g. Goodrich et al., 2010b). Unlike other primitive achondrites, ureilites were cooled rapidly following the disruption and destruction of the parent body (20 ºC/h; Takeda et al., 1989). Because, unlike lodranites and brachinites, the phase assemblage of ureilites was not significantly re-equilibrated during retrograde metamorphism, it can be directly compared to the mineral assemblage of experiments. Ureilites are the only primitive achondrites displaying Cr2O3 and CaO contents identical to the ones of olivine in melting experiments between 1150 and 1300 ºC (0.4-0.8 and 0.2-0.4, respectively as opposed to < 0.1 wt.% in other groups; Fig. 5). Experiments performed on high-NaK# compositions (H, LL and particularly CI) produced a ureilite-like phase assemblage composed of olivine, pigeonite and metal in equilibrium with liquid between 1130- 1180 ºC (Fig. 1). The phase assemblage contains augite and orthopyroxene at lower temperature (<1130 ºC) and only orthopyroxene at higher temperature (>1180 ºC). In CM and CV experiments, olivine-pigeonite-metal assemblages resembling ureilites are never formed because plagioclase melts out of the residues at the same temperature (CM) or above the temperature at which pigeonite is replaced by orthopyroxene (CV; Fig. 1). Goodrich (1999), Goodrich et al. (2007) and Singletary and Grove (2006) observed that high CaO/Al2O3 ratios were necessary to stabilize pigeonite in melting residues. Goodrich (1999) and Goodrich et al. (2007) argued that the initial composition of the ureilite parent body (UPB) had a CaO/Al2O3 ratio 2.5 times higher than the chondritic value of 0.8 (also see discussion in Goodrich et al., 2013b). However, there is no evidence in the meteorite record that Ca and Al can be fractionated from each other to any significant extent by nebular or low-temperature processes in parent bodies. On the other hand, our experiments show that Al can be preferentially extracted relative to Ca by partial melting. Experiments performed on high-NaK# compositions produce partial melts with low CaO/Al2O3 ratio (Table 2; Chapter 1) that, once extracted, leave behind

87 residues with high CaO/Al2O3 ratios in which pigeonite is stable. On the other hand, the extraction of partial melts from chondritic materials depleted in alkali elements (CM and CV compositions) do not fractionate Al from Ca sufficiently to stabilize pigeonite. Therefore, the super-chondritic

CaO/Al2O3 ratio of ureilites is acquired during melting of a parent body with high initial NaK#

(>45) and a chondritic CaO/Al2O3 (as envisioned by Kita et al., 2004). Like the parent bodies of acapulcoites-lodranites, winonaites and brachinites, the parent body of ureilites was not depleted in MVE and produced low-degree melts rich in alkalis and silica. This increases the likelihood that the trachyandesite ALM-A (Bischoff et al., 2014) and some of the clasts found in polymict ureilites (Cohen et al., 2004) represent the crystallization products of low-degree partial melts produced in the UPB (Chapter 1). Experiments with H, LL and CI compositions contain pigeonite-olivine-metal residues between 1130-1180 ºC, but orthopyroxene replaces pigeonites at higher temperature. However, many ureilites are thought to have reached peak temperatures of 1200-1280 ºC (Mittlefehldt et al., 1998) and REE patterns suggest that melting was incremental to near-fractional (Barrat et al., 2016) while our experiments simulate a process of batch melting. We performed additional experiments to constrain the initial composition of the UPB and its melting behavior in more detail. They are the subject of a separate manuscript (Chapter 3).

5. Location of alkali-undepleted planetesimals

Meteorites display mass-independent fractionation of oxygen isotopes (Clayton et al., 1976), expressed as the difference relative to the terrestrial fractionation line (∆17O; Fig. 6). Variations in ∆17O are thought to result from nebular processes such as CO self-shielding during photodissociation (Clayton, 2002). This process predicts that solid condensed close to proto-sun (i.e. Calcium Aluminum-rich Inclusions or CAIs) inherited its composition in O isotopes while at increasing heliocentric distances the wavelengths responsible for the dissociation of C16O were attenuated to a much greater extent than those responsible for the dissociation of C17O and C18O. Heavy O radicals were then preferentially condensed as silicates or water ice, such as the cosmic symplectites of the carbonaceous chondrites Acfer 094 (Sakamoto et al., 2007; Tenner et al., 2015). While bulk chondrites and primitive achondrites display mass independent fractionation of O isotopes, they vary within a much smaller range of ∆17O than individual chondrite components and cannot be inferred to originate from a precise heliocentric distance, in accordance with the idea

88 that the accretion disk was a highly dynamical environment (e.g. Heinzeller et al., 2011). Some groups of primitive achondrites, such as ureilites (∆17O = -1 ± 1) and acapulcoites-lodranites (∆17O = -1 ± 0.4) display mass-independent fractionation between samples of the same magnitude as groups of chondrites (Greenwood et al., 2017; Fig. 12). This is one of the arguments supporting a melting residue origin. More recently, isotopic nucleosynthetic anomalies have been identified at the scale of bulk meteorites for elements such as Cr (Trinquier et al., 2007), Ti (Trinquier et al., 2009), Ni (Steele et al., 2012), Ca (Chen et al., 2011) and Mo (Kruijer et al., 2017). They result from the heterogenous distribution of pre-solar grains, formed in massive stars, that never mixed completely in the solar nebula (Dauphas and Schauble, 2016). When ∆17O is plotted against nucleosynthetic anomalies (e.g. ε54Cr), meteorites form two well defined groups (Fig. 7). Carbonaceous chondrites and achondrites (CC) have higher concentrations of neutron-rich isotopes (e.g. 54Cr and 50Ti) than the Earth, Mars and a second group of meteorites called “non-carbonaceous” (NC), which includes enstatite and ordinary chondrites. Those two groups are thought to represent distinct reservoirs of the nebula (Warren, 2011) and they would have been separated from each other when Jupiter opened a gap in the disk and a pressure maximum appeared just beyond its orbit (Kruijer et al., 2017; Desch et al., 2018). According to this model, the NC reservoir would represent everything that condensed within the orbit of Jupiter and the CC reservoir would represent everything that condensed beyond its orbit. Carbonaceous chondrites enriched in refractory elements and characterized by low NaK# (e.g. CV, CM) would have formed in the pressure trap just outside of the orbit of Jupiter where CAIs accumulated by aerodynamic drag. CI chondrites would have accreted further outward. Because this “isotopic dichotomy” affects iron meteorites and differentiated achondrites that were accreted rapidly (0.5-2.3 Ma) as well as chondrites that were accreted later (2-4 Ma; e.g. Sugiura and Fujiya, 2014; Kruijer et al., 2017), the CC and NC reservoirs were long lived and remained separated for 4-5 Ma, possibly until giant planets migrated and mixed NC and CC material in the asteroid belt (Warren, 2011; Walsh et al., 2011; Scott et al., 2018). The trachyandesite achondrites (Chapter 1) and primitive achondrites are all part of the NC reservoir (Fig. 7) and, therefore, are assumed to have formed in the inner disk, inside the orbit of Jupiter. Their exact location of accretion is still undecided but rocks from the Earth/Moon, Mars (shergottites) and Vesta (eucrites) from a trend of increasing depletion in neutron-rich isotopes

(lower ε54Cr and ε50Ti) with increasing heliocentric distance within the NC reservoir. If this trend is indicative of the location of accretion, ordinary chondrites, primitive achondrites and other differentiated meteorites such as HEDs and angrites would have all formed further out than Mars. Ureilites could have formed the furthest from the sun (Yamakawa et al., 2010) while still technically in the NC reservoir. This could explain why they share many geochemical and petrological features with carbonaceous chondrites, including high C abundances (Goodrich et al., 2015). Alternatively the UPB could have accreted extremely early, <0.5 Ma before CAIs, and before the two reservoirs were firmly separated (Goodrich et al., 2015; van Kooten et al., 2017). Based on Figure 7, the parent bodies of angrites and brachinites could have accreted very close to each other. All the parent bodies of the NC reservoir for which the initial bulk composition can be estimated were not depleted in MVE relative to refractory elements (NaK# = 50; Fig. 7). The parent bodies of enstatite chondrites, and probably aubrites, had an even higher NaK# (60) and are thought to have formed closer to the sun, possibly in the Earth neighborhood. Those observations, along with the high concentration of alkali elements in Mercury (Peplowski et al., 2014), suggest that the inner solar system was not depleted in MVE relative to refractory elements at the time of planetesimals accretion. Because many of the planetesimals that accreted in the inner solar system were not depleted in alkali elements, they first produced low-degree melts rich in alkalis and silica upon melting.

6. Timescale of alkali- and silica- rich magmatism

The crystallization of the trachyandesite achondrite ALM-A, assumed to derive from the UPB (Bischoff et al., 2014), and plagioclase-rich clasts in ureilites (Goodrich et al., 2010a; van

Kooten et al., 2017) have been dated at 4-6 Ma after the time of CAIs condensation (∆tCAI) with the short-lived isotope systems 26Al-26Mg and 53Cr-53Mn. GRA 06128/9, which is assumed to have crystallized on the brachinite parent body, is even younger (2.3 ± 0.3 Ma ∆tCAI; Shearer et al., 2010). To our knowledge, no crystallization ages are available for NWA 6698/11575 (Bunch et al., 2011; Agee et al., 2018). The time of magmatic activity can also be estimated with 182Hf-182W model ages of the melting residues (i.e. the primitive achondrites). Budde et al. (2015) calculated a 182Hf-182W model age for the extraction of melts from the UPB at 3.3 ± 0.7 Ma ∆tCAI and used it to calculate an

26 26 accretion age of 1.4-1.7 Ma ∆tCAI. Based on Al- Mg model ages of FeO-rich ureilites representing melting residues, van Kooten et al. (2017) and Baker et al. (2012) infer that the UPB could have accreted as early as 0.6 Ma ∆tCAI, at the same time as many iron meteorites, as suggested by Lee et al. (2009) based on earlier 182Hf-182W systematics. Acapulcoites-lodranites are 182 182 characterized by slightly older Hf- W model ages (5.5 ± 1.0 Ma ∆tCAI) than ureilites, and their parent body has been suggested to have accreted between 1.5 and 2 Ma ∆tCAI (Touboul et al., 2009). Also using the Hf-W system, Burkhardt et al. (2017) dated an event of melt extraction from

182 182 two brachinite-like achondrites, NWA 5400 and NWA 5363, at 2.2 ± 0.8 Ma ∆tCAI. A Hf- W model age for Brachina suggests that it cooled under 1000 ºC at 4 ± 0.9 Ma ∆tCAI (Wadhwa et al., 1998). The brachinite parent body, assuming that GRA 06128, brachinites and brachinite-like achondrites do indeed come from the same parent body, is the only one for which 182Hf-182W model ages are consistent with 26Al-26Mg model ages. Melting, melt extraction and subsequent crystallization appear to have occurred at 2-4 Ma ∆tCAI. The age of accretion and early magmatic activity on the UPB is still uncertain. The accretion ages of the UPB and acapulcoites-lodranites parent (1.4-1.7 and 1.5-2.0 ∆tCAI, respectively; Budde et al., 2015; Touboul et al., 2009) are calculated with temperature evolution models of planetesimals heated by the radioactive decay of 26Al. However, such models are dependent on the mechanisms of melt extraction, which removes 26Al from the interior of planetesimals (Moskvitz and Gaidos, 2011; Neumann et al., 2014; Lichtenberg et al., 2019) and could be too sensitive to modeling assumptions. For example, current models could underestimate the efficiency of melt extraction and removal of 26Al (Chapter 1), and, therefore, could overestimate accretion ages. Despite the remaining uncertainties, those age constraints suggest that the parent bodies of primitive achondrites were affected by igneous processes at the same time as the angrite parent body (APB; 4 Ma ∆tCAI; Tissot et al., 2017) and Vesta (2.5-4 Ma ∆tCAI; Hublet et al., 2017). Magmatic activity on some planetesimals with carbonaceous chondrite affinities, such as the parent bodies of NWA 2976 (5 Ma ∆tCAI; Bouvier et al., 2011) and NWA 6704 (5-5.5 Ma ∆tCAI; Amelin et al., 2019; Sanborn et al., 2019) seems to have occurred during the same period in the outer solar system.

7. Implications for the distribution of alkali elements in the early-solar system

It is striking that planetesimals that accreted in the NC reservoir, the inner solar system, could have been affected by such contrasting styles of magmatism simultaneously. For example, angrites and brachinites have nearly identical isotopic compositions (Fig. 7) but one parent body produced silica-undersaturated melts almost devoid of alkali elements while the other produced melts rich in alkalis and silica. It could be argued that the parent bodies of primitive achondrites, even if they formed in the same region of the solar nebula and were active at the same time, accreted later (1-2 Ma ∆tCAI) than Vesta, the angrites parent body (APB) and the parent bodies of

NC irons meteorites (0.5-1 Ma ∆tCAI; Sugiura and Fujiya, 2014; Kruijer et al., 2017). As discussed in section 6, this is not clearly supported by current isotope systematics but cannot be ruled out either. Assuming that Vesta and the APB accreted earlier, their low NaK# and depletion in MVE could result from the incomplete condensation of the solar nebula (Albarede, 2009; Palme and O’Neill, 2014). However, IIAB and IC irons, which are also part of the NC reservoir have the oldest Hf-W model ages of core formation (0.8 and 0.3 ± 0.5 Ma) among iron meteorites (Kruijer et al., 2017) but also the highest concentrations of Ga (Chabot and Haack, 2006), a siderophile MVE with a condensation temperature identical to Na (Lodders, 2003). Therefore, we consider it more likely that, even if Vesta and the APB accreted before primitive achondrites, the bulk composition of their precursor materials was not significantly depleted in MVE. It is possible that Vesta, the APB and, more generally, the planetesimals that accounted for most of the mass of terrestrial planets had high NaK# and formed alkali- and silica- rich melts when they first melted.

Several ungrouped primitive achondrites dominated by FeO-rich olivine (Fo64-72) have recently been described: Tafassasset (Gardner-Vandy et al., 2012), NWA 8186 (Srinivasan et al., 2016) and LEW 88763 (Day et al., 2015). Unlike the main groups of primitive achondrites, they have affinities for the CC reservoir (Fig. 6-7; Sanborn et al., 2019). Tafassasset has a near- chondritic composition in major elements and is only slightly depleted in REE relative to CI chondrites. However, the average composition of plagioclase (An32) is lower in Na than in all equilibrated chondrites from the NC reservoir. LEW 88763 is also characterized by chondritic concentrations of REE, suggesting that no silicate melt was extracted. Based on the plagioclase composition (An30) and the bulk NaK# (39), the level of depletion in MVE seems to be intermediate between CI (NaK# = 50) and CM chondrites (NaK# = 33). NWA 8186 is a primitive

92 achondrite candidate with affinities for CV/CK. Because the composition of the plagioclase (An48-

51) is fairly homogeneous and identical to the composition of plagioclase in our sub-solidus experiments performed on a CV chondrite composition, we posit that NWA 8186 did not reach the temperature of the solidus and that no melt has been extracted. Its parent body, possibly the same as the one of CV or CK chondrites, was depleted in alkali elements. All the parent bodies of the CC reservoir, with the exception of CI chondrites, were variably depleted in alkali elements. On the other hand, all the parent bodies from the NC reservoir were not initially depleted in alkali elements. When they melted, they produced low-degree partial melts rich in alkalis, Al2O3 and SiO2 that, due to their low density (Chapter 1), were efficiently extracted from the interior of at least four planetesimals (the parent body of NWA 11575/6698, the UPB, the brachinite PB and the acapulcoite-lodranite PB). Because Al is a lithophile element, melt migration would have removed 26Al, altered the thermal history of planetesimals and influenced the formation, or lack thereof, of magma oceans (Scott et al., 1993; Hevey and Sanders, 2006; Neumann et al., 2012; Neumann et al., 2014; Wilson and Keil, 2017; Lichtenberg et al., 2019). Some planetesimals, such as Vesta (e.g. Mandler and Elkins-Tanton, 2013), likely experienced a magma ocean stage. Low-degree melts rich in alkalis and Al2O3 could have been retained and allowed melting to continue until the formation of magma oceans. Alternatively, low- 26 degree melts could have been extracted if the small quantity of Al2O3 (including Al) remaining in the residue after the exhaustion of plagioclase was sufficient to melt the parent body further

(0.3-0.4 wt.% Al2O3, corresponding to ~20 wt.% of the initial chondritic concentration). The depletion of alkali-elements in parent bodies could have started with the extraction of low-degree melts. Partial melts and their crystallization products could have been affected by pyroclastic eruptions (e.g. Warren and Kallemeyn, 1992; Wilson and Keil, 2012) or crustal erosion from impactors (e.g. Boujibar et al., 2015; Carter et al., 2018). The depletion of MVE could have continued during magma ocean stages (e.g. Norris and Wood, 2017; Young et al., 2019). All those processes, and probably others, influenced the final, alkali-depleted, composition of terrestrial planets but the precursor chondritic materials of planetesimals from the inner solar system do not appear to have been depleted in MVE.

8. Conclusion

We show that the parent bodies of all primitive achondrites were alkali-undepleted (i.e. similar to H, LL, CI chondrites and the sun’s photosphere). Despite melting at contrasting fO2, they produced up to 15 wt.% of silica- and alkali-rich melts. The extraction of low-degree melts from those parent bodies resulted in two complementary types of achondrites: trachyandesite achondrites representing the partial melts and primitive achondrites representing the residues. All the parent bodies from the “non-carbonaceous” reservoir for which we can reconstruct the initial chondritic composition were alkali-undepleted relative to refractory elements (Ca, Al) and the sun’s photosphere. We infer that the chondritic precursors of Vesta and the angrite parent body had similar alkali element contents (NaK# of 50) and became depleted in moderately volatile elements following complex igneous processes. Even if the depletion of MVE appeared rapidly in planetesimals and planetary embryos, it does not seem to be a primary feature and do not result from the incomplete condensation of the solar nebula.

9. References

Agee C. B., Habermann M. A. and Ziegler K. (2018) Northwest Africa 11575: Unique Ungrouped Trachyandesite Achondrite. 49th Lunar Planet. Sci. Conf. 19-23 March, 2018, held Woodlands, Texas LPI Contrib. No. 2083, id.2226 49. Albarede F. (2009) Volatile accretion history of the terrestrial planets and dynamic implications. Nature 461, 1227–1233. Amelin Y., Koefoed P., Iizuka T., Fernandes V. A., Huyskens M. H., Yin Q.-Z. and Irving A. J. (2019) U-Pb, Rb-Sr and Ar-Ar systematics of the ungrouped achondrites Northwest Africa 6704 and Northwest Africa 6693. Geochim. Cosmochim. Acta 245, 628–642. Baker J. A., Schiller M. and Bizzarro M. (2012) 26Al–26Mg deficit dating ultramafic meteorites and silicate planetesimal differentiation in the early Solar System? Geochim. Cosmochim. Acta 77, 415– 431. Barrat J.-A., Jambon A., Yamaguchi A., Bischoff A., Rouget M.-L. and Liorzou C. (2016) Partial melting of a C-rich asteroid: Lithophile trace elements in ureilites. Geochim. Cosmochim. Acta 194, 163– 178. Benedix G. K., Lauretta D. S. and McCoy T. J. (2005) Thermodynamic constraints on the formation conditions of winonaites and silicate-bearing IAB irons. Geochim. Cosmochim. Acta 69, 5123–5131. Benedix G. K., McCoy T. J., Keil K., Bogard D. D. and Garrison D. H. (1998) A petrologic and isotopic study of winonaites: evidence for early partial melting, brecciation, and metamorphism. Geochim. Cosmochim. Acta 62, 2535–2553. Benedix G. K., McCoy T. J., Keil K. and Love S. G. (2000) A petrologic study of the IAB iron meteorites: Constraints on the formation of the IAB-Winonaite parent body. Meteorit. Planet. Sci. 35, 1127–1141. Bild R. W. (1977) Silicate inclusions in group IAB irons and a relation to the anomalous stones Winona and Mt Morris (Wis). Geochim. Cosmochim. Acta 41, 1439–1456. Bischoff A., Horstmann M., Barrat J.-A., Chaussidon M., Pack A., Herwartz D., Ward D., Vollmer C. and Decker S. (2014) Trachyandesitic volcanism in the early Solar System. Proc. Natl. Acad. Sci. U. S.

A. 111, 12689–92. Bischoff A., Vogel N. and Roszjar J. (2011) The Rumuruti chondrite group. Geochemistry 71, 101–133. Boujibar A., Andrault D., Bolfan-Casanova N., Bouhifd M. A. and Monteux J. (2015) Cosmochemical fractionation by collisional erosion during the Earth’s accretion. Nat. Commun. 6, 8295. Bouvier A., Spivak-Birndorf L. J., Brennecka G. A. and Wadhwa M. (2011) New constraints on early Solar System chronology from Al–Mg and U–Pb isotope systematics in the unique basaltic achondrite Northwest Africa 2976. Geochim. Cosmochim. Acta 75, 5310–5323. Budde G., Kruijer T. S., Fischer-Gödde M., Irving A. J. and Kleine T. (2015) Planetesimal differentiation revealed by the Hf–W systematics of ureilites. Earth Planet. Sci. Lett. 430, 316–325. Bunch T. E., Irving A. J., Wittke J. H. and Rumble D. (2011) Northwest Africa 6698: A High Temperature “Dioritic” Achondrite with Oxygen Composition in the LL Chondrite Field. MetSoc 74 74, 5224. Burkhardt C., Dauphas N., Tang H., Fischer-Gödde M., Qin L., Chen J. H., Rout S. S., Pack A., Heck P. R. and Papanastassiou D. A. (2017) In search of the Earth-forming reservoir: Mineralogical, chemical, and isotopic characterizations of the ungrouped achondrite NWA 5363/NWA 5400 and selected chondrites. Meteorit. Planet. Sci. 52, 807–826. Carter P. J., Leinhardt Z. M., Elliott T., Stewart S. T. and Walter M. J. (2018) Collisional stripping of planetary crusts. Earth Planet. Sci. Lett. 484, 276–286. Chabot N. L. and Haack H. (2006) Evolution of Asteroidal Cores. Meteorites early Sol. Syst. II, 747–771. Chen H.-W., Lee T., Lee D.-C., Shen J. J.-S. and Chen J.-C. (2011) 48 Ca Heterogeneity in Differentiated Meteroites. Astrophys. J. 743, L23. Clayton R. N. (2002) Self-shielding in the solar nebula. Nature 415, 860–861. Clayton R. N. and Mayeda T. K. (1988) Formation of ureilites by nebular processes. Geochim. Cosmochim. Acta 52, 1313–1318. Clayton R. N., Onuma N. and Mayeda T. K. (1976) A classification of meteorites based on oxygen isotopes. Earth Planet. Sci. Lett. 30, 10–18. Cohen B. A., Goodrich C. A. and Keil K. (2004) Feldspathic clast populations in polymict ureilites: Stalking the missing basalts from the ureilite parent body. Geochim. Cosmochim. Acta 68, 4249– 4266. Dauphas N., Poitrasson F., Burkhardt C., Kobayashi H. and Kurosawa K. (2015) Planetary and meteoritic Mg/Si and variations inherited from solar nebula chemistry. Earth Planet. Sci. Lett. 427, 236–248. Dauphas N. and Schauble E. A. (2016) Mass Fractionation Laws, Mass-Independent Effects, and Isotopic Anomalies. Annu. Rev. Earth Planet. Sci. 44, 709–783. Dauphas N. (2017) The isotopic nature of the Earth’s accreting material through time. Nature 541, 521– 524. Day J. M. D., Corder C. A., Assayag N. and Cartigny P. (2019) Ferrous oxide-rich asteroid achondrites. Geochim. Cosmochim. Acta. Day J. M. D., Corder C. A., Rumble D., Assayag N., Cartigny P. and Taylor L. A. (2015) Differentiation processes in FeO-rich asteroids revealed by the achondrite Lewis Cliff 88763. Meteorit. Planet. Sci. 50, 1750–1766. Day J. M. D., Walker R. J., Ash R. D., Liu Y., Rumble D., Irving A. J., Goodrich C. A., Tait K., McDonough W. F. and Taylor L. A. (2012) Origin of felsic achondrites Graves Nunataks 06128 and 06129, and ultramafic brachinites and brachinite-like achondrites by partial melting of volatile-rich primitive parent bodies. Geochim. Cosmochim. Acta 81, 94–128. Desch S. J., Kalyaan A. and Alexander C. M. O. (2018) The Effect of Jupiter’s Formation on the Distribution of Refractory Elements and Inclusions in Meteorites. Astrophys. J. Suppl. Ser. 238, 11. Dhaliwal J. K., Day J. M. D., Corder C. A., Tait K. T., Marti K., Assayag N., Cartigny P., Rumble D. and Taylor L. A. (2017) Early metal-silicate differentiation during planetesimal formation revealed by acapulcoite and lodranite meteorites. Geochim. Cosmochim. Acta. Dunn T. L., Cressey G., McSween H. Y. and McCoy T. J. (2010) Analysis of ordinary chondrites using powder X-ray diffraction: 1. Modal mineral abundances. Meteorit. Planet. Sci. 45, 127–138.

Ebel D. S. and Grossman L. (2000) Condensation in dust-enriched systems. Geochim. Cosmochim. Acta 64, 339–366. Fedkin A. V. and Grossman L. (2016) Effects of dust enrichment on oxygen fugacity of cosmic gases. Meteorit. Planet. Sci. 51, 843–850. Available at: http://doi.wiley.com/10.1111/maps.12627 [Accessed August 8, 2019]. Floss C. (2000) Complexities on the acapulcoite-lodranite parent body: Evidence from trace element distributions in silicate minerals. Meteorit. Planet. Sci. 35, 1073–1085. French B. M. and Eugster H. P. (1965) Experimental control of oxygen fugacities by graphite-gas equilibriums. J. Geophys. Res. 70, 1529–1539. Gardner-Vandy K. G., Lauretta D. S., Greenwood R. C., McCoy T. J., Killgore M. and Franchi I. A. (2012) The Tafassasset primitive achondrite: Insights into initial stages of planetary differentiation. Geochim. Cosmochim. Acta 85, 142–159. Gardner-Vandy K. G., Lauretta D. S. and McCoy T. J. (2013) A petrologic, thermodynamic and experimental study of brachinites: Partial melt residues of an R chondrite-like precursor. Geochim. Cosmochim. Acta 122, 36–57. Goodrich C. A. (1999) Are ureilites residues from partial melting of chondritic material? The answer from MAGPOX. Meteorit. Planet. Sci. 34, 109–119. Goodrich C. A. (1992) Ureilites - A critical review. Meteoritics 27. Goodrich Cyrena A, Ash R. D., Van Orman J. A., Domanik K. and McDonough W. F. (2013) Metallic phases and siderophile elements in main group ureilites: Implications for ureilite petrogenesis. Geochim. Cosmochim. Acta 112, 340–373. Goodrich C. A. and Delaney J. S. (2000) Fe/Mg–Fe/Mn relations of meteorites and primary heterogeneity of primitive achondrite parent bodies. Geochim. Cosmochim. Acta 64, 149–160. Goodrich C. A., Hartmann W. K., O’Brien D. P., Weidenschilling S. J., Wilson L., Michel P. and Jutzi M. (2015) Origin and history of ureilitic material in the solar system: The view from asteroid 2008 TC3 and the Almahata Sitta meteorite. Meteorit. Planet. Sci. 50, 782–809. Goodrich C. A., Hutcheon I. D., Kita N. T., Huss G. R., Cohen B. A. and Keil K. (2010a) 53Mn–53Cr and 26Al–26Mg ages of a feldspathic lithology in polymict ureilites. Earth Planet. Sci. Lett. 295, 531–540. Goodrich C. A., Kita N. T., Spicuzza M. J., Valley J. W., Zipfel J., Mikouchi T. and Miyamoto M. (2010b) The Northwest Africa 1500 meteorite: Not a ureilite, maybe a brachinite. Meteorit. Planet. Sci. 45, 1906–1928. Goodrich C. A., Kita N. T., Sutton S. R., Wirick S. and Gross J. (2017) The Miller Range 090340 and 090206 meteorites: Identification of new brachinite-like achondrites with implications for the diversity and petrogenesis of the brachinite clan. Meteorit. Planet. Sci. 52, 949–978. Goodrich C. A., Van Orman J. A. and Wilson L. (2007) Fractional melting and smelting on the ureilite parent body. Geochim. Cosmochim. Acta 71, 2876–2895. Goodrich Cyrena A., Sutton S. R., Wirick S. and Jercinovic M. J. (2013a) Chromium valences in ureilite olivine and implications for ureilite petrogenesis. Geochim. Cosmochim. Acta 122, 280–305. Goodrich Cyrena A., Wilson L., Van Orman J. A. and Michel P. (2013b) Comment on “Parent body depth-pressure-temperature relationships and the style of the ureilite anatexis” by P. H. Warren (MAPS 47:209-227). Meteorit. Planet. Sci. 48, 1096–1106. Greenwood R. C., Burbine T. H., Miller M. F. and Franchi I. A. (2017) Melting and differentiation of early-formed asteroids: The perspective from high precision oxygen isotope studies. Chemie der Erde - Geochemistry 77, 1–43. French B. M. and Eugster H. P. (1965) Experimental control of oxygen fugacities by graphite-gas equilibriums. J. Geophys. Res. 70, 1529–1539. Heinzeller D., Nomura H., Walsh C. and Millar T. J. (2011) Chemical evolution of protoplanetary disks - The effects of viscous accretion, turbulent mixing, and disk winds. Astrophys. J. 731, 115. Hevey P. J. and Sanders I. S. (2006) A model for planetesimal meltdown by 26 Al and its implications for meteorite parent bodies. Meteorit. Planet. Sci. 41, 95–106.

Hublet G., Debaille V., Wimpenny J. and Yin Q.-Z. (2017) Differentiation and magmatic activity in Vesta evidenced by 26Al-26Mg dating in eucrites and diogenites. Geochim. Cosmochim. Acta 218, 73–97. Hunt A. C., Benedix G. K., Hammond S. J., Bland P. A., Rehkämper M., Kreissig K. and Strekopytov S. (2017) A geochemical study of the winonaites: Evidence for limited partial melting and constraints on the precursor composition. Geochim. Cosmochim. Acta 199, 13–30. Jarosewich E. (2006) Chemical analyses of meteorites at the Smithsonian Institution: An update. Meteorit. Planet. Sci. 41, 1381–1382. Keil K. (2014) Brachinite meteorites: Partial melt residues from an FeO-rich asteroid. Chemie der Erde - Geochemistry 74, 311–329. Keil K. and McCoy T. J. (2018) Acapulcoite-lodranite meteorites: Ultramafic asteroidal partial melt residues. Chemie der Erde 78, 153–203. Kita N. T., Ikeda Y., Togashi S., Liu Y., Morishita Y. and Weisberg M. K. (2004) Origin of ureilites inferred from a SIMS oxygen isotopic and trace element study of clasts in the Dar al Gani 319 polymict ureilite. Geochim. Cosmochim. Acta 68, 4213–4235. van Kooten E. M. M. E., Schiller M. and Bizzarro M. (2017) Magnesium and chromium isotope evidence for initial melting by radioactive decay of 26Al and late stage impact-melting of the ureilite parent body. Geochim. Cosmochim. Acta 208, 1–23. Krot A. N., Keil K., Scott E. R. D., Goodrich C. A. and Weisberg M. K. (2013) Classification of Meteorites and Their Genetic Relationships. In Treatise on Geochemistry: Second Edition pp. 1–63. Kruijer T. S., Burkhardt C., Budde G. and Kleine T. (2017) Age of Jupiter inferred from the distinct genetics and formation times of meteorites. Proc. Natl. Acad. Sci. U. S. A. 114, 6712–6716. Lee D.-C., Halliday A. N., Singletary S. J. and Grove T. L. (2009) 182Hf–182W chronometry and early differentiation of the ureilite parent body. Earth Planet. Sci. Lett. 288, 611–618. Li S., Yin Q.-Z., Bao H., Sanborn M. E., Irving A., Ziegler K., Agee C., Marti K., Miao B., Li X., Li Y. and Wang S. (2018) Evidence for a Multilayered Internal Structure of the Chondritic Acapulcoite- Lodranite Parent Asteroid. Geochim. Cosmochim. Acta. Lichtenberg T., Keller T., Katz R. F., Golabek G. J. and Gerya T. V. (2019) Magma ascent in planetesimals: Control by grain size. Earth Planet. Sci. Lett. 507, 154–165. Lodders K. (2003) Solar System Abundances and Condensation Temperatures of the Elements. Astrophys. J. 591, 1220–1247. Available at: http://stacks.iop.org/0004-637X/591/i=2/a=1220 [Accessed November 26, 2018]. Lodders K. and Fegley B. (1998) The planetary scientist’s companion., Oxford University Press. Lunning N. G., Gardner-Vandy K. G., Sosa E. S., McCoy T. J., Bullock E. S. and Corrigan C. M. (2017) Partial melting of oxidized planetesimals: An experimental study to test the formation of oligoclase- rich achondrites Graves Nunataks 06128 and 06129. Geochim. Cosmochim. Acta 214, 73–85. Mandler B. E. and Elkins-Tanton L. T. (2013) The origin of eucrites, diogenites, and olivine diogenites: Magma ocean crystallization and shallow magma chamber processes on Vesta. Meteorit. Planet. Sci., n/a-n/a. McCoy T. J., Carlson W. D., Nittler L. R., Stroud R. M., Bogard D. D. and Garrison D. H. (2006) Graves Nunataks 95209: A snapshot of metal segregation and core formation. Geochim. Cosmochim. Acta 70, 516–531. McCoy T. J., Corrigan C. M., Dickinson T. L., Benedix G. K., Schrader D. L. and Davidson J. (2019) Grove Mountains (GRV) 020043: Insights into acapulcoite-lodranite genesis from the most primitive member. Geochemistry, 125536. McCoy T. J., Keil K., Clayton R. N., Mayeda T. K., Bogard D. D., Garrison D. H. and Wieler R. (1997a) A petrologic and isotopic study of lodranites: Evidence for early formation as partial melt residues from heterogeneous precursors. Geochim. Cosmochim. Acta 61, 623–637. McCoy T. J., Keil K., Muenow D. W. and Wilson L. (1997b) Partial melting and melt migration in the acapulcoite-lodranite parent body. Geochim. Cosmochim. Acta 61, 639–650. Mittlefehldt D. W., McCoy T. J., Goodrich C. A. and Kracher A. (1998) Non-chondritic meteorites from asteroidal bodies. Rev. Mineral. Geochemistry 36, 4.1-4.195.

Miyamoto M. (1985) Cooling history of some antarctic ureilites. Proceding Sixt. lunar Planet. Sci. Conf. 90, D116–D122. Moskvitz N. and Gaidos E. (2011) Differentiation of planetesimals and the thermal consequences of melt migration. Meteorit. Planet. Sci. 46, 903–918. Nehru C. E., Prinz M., Delaney J. S., Dreibus G., Palme H., Spettel B. and Wänke H. (1983) Brachina: A new type of meteorite, not a chassignite. J. Geophys. Res. 88, B237. Neumann W., Breuer D. and Spohn T. (2012) Differentiation and core formation in accreting planetesimals. Astron. Astrophys. 543, A141. Neumann W., Breuer D. and Spohn T. (2014) Differentiation of Vesta: Implications for a shallow magma ocean. Earth Planet. Sci. Lett. 395, 267–280. Norris C. A. and Wood B. J. (2017) Earth’s volatile contents established by melting and vaporization. Nature 549, 507–510. Palme H. and O’Neill H. S. C. (2014) Cosmochemical Estimates of Mantle Composition. Treatise on Geochemistry, 1–39. Palme H., Schultz L., Spettel B., Weber H. ., Wänke H., Michel-Levy M. C. and Lorin J. . (1981) The Acapulco meteorite: Chemistry, mineralogy and irradiation effects. Geochim. Cosmochim. Acta 45, 727–752. Peplowski P. N., Evans L. G., Stockstill-Cahill K. R., Lawrence D. J., Goldsten J. O., McCoy T. J., Nittler L. R., Solomon S. C., Sprague A. L., Starr R. D. and Weider S. Z. (2014) Enhanced sodium abundance in Mercury’s north polar region revealed by the MESSENGER Gamma-Ray Spectrometer. Icarus 228, 86–95. Prinz M., Nehru C. E., Delaney J. S. . and Weisberg M. (1983) Silicates in IAB and IIICD Irons, Winonaites, Lodranites an Brachina: a Primitive and Modified-Primitive Group. LUNAR Planet. Sci. XIV, 616–617. Righter K. and Neff K. E. (2007) Temperature and oxygen fugacity constraints on CK and R chondrites and implications for water and oxidation in the early solar system. Polar Sci. 1, 25–44. Righter K., Sutton S. R., Danielson L., Pando K. and Newville M. (2016) Redox variations in the inner solar system with new constraints from vanadium XANES in spinels. Am. Mineral. 101, 1928–1942. Ringwood A. E. (1961) Chemical and genetic relationships among meteorites. Geochim. Cosmochim. Acta 24, 159–197. Rubin A. E. (2007) Petrogenesis of acapulcoites and lodranites: A shock-melting model. Geochim. Cosmochim. Acta 71, 2383–2401. Ruzicka A. and Hutson M. (2010) Comparative petrology of silicates in the Udei Station (IAB) and Miles (IIE) iron meteorites: Implications for the origin of silicate-bearing irons. Geochim. Cosmochim. Acta 74, 394–433. Sakamoto N., Seto Y., Itoh S., Kuramoto K., Fujino K., Nagashima K., Krot A. N. and Yurimoto H. (2007) Remnants of the Early Solar System Water Enriched in Heavy Oxygen Isotopes. Science (80- . ). 317, 231 LP – 233. Sanborn M. E., Wimpenny J., Williams C. D., Yamakawa A., Amelin Y., Irving A. J. and Yin Q.-Z. (2019) Carbonaceous achondrites Northwest Africa 6704/6693: Milestones for early Solar System chronology and genealogy. Geochim. Cosmochim. Acta 245, 577–596. Sanders I. S., Scott E. R. D. and Delaney J. S. (2017) Origin of mass-independent oxygen isotope variation among ureilites: Clues from chondrites and primitive achondrites. Meteorit. Planet. Sci. 52, 690–708. Schmitz B., Yin Q. Z., Sanborn M. E., Tassinari M., Caplan C. E. and Huss G. R. (2016) A new type of solar-system material recovered from Ordovician marine limestone. Nat. Commun. 7, 11851. Schrader D. L., McCoy T. J. and Gardner-Vandy K. (2017) Relict chondrules in primitive achondrites: Remnants from their precursor parent bodies. Geochim. Cosmochim. Acta 205, 295–312. Scott E. R. D., Krot A. N. and Sanders I. S. (2018) Isotopic Dichotomy among Meteorites and Its Bearing on the Protoplanetary Disk. Astrophys. J. 854, 164. Scott E. R. D., Taylor G. J. and Keil K. (1993) Origin of ureilite meteorites and implications for planetary

accretion. Geophys. Res. Lett. 20, 415–418. Shearer C. K., Burger P. V., Neal C., Sharp Z., Spivak-Birndorf L., Borg L., Fernandes V. A., Papike J. J., Karner J., Wadhwa M., Gaffney A., Shafer J., Geissman J., Atudorei N.-V., Herd C., Weiss B. P., King P. L., Crowther S. A. and Gilmour J. D. (2010) Non-basaltic asteroidal magmatism during the earliest stages of solar system evolution: A view from Antarctic achondrites Graves Nunatak 06128 and 06129. Geochim. Cosmochim. Acta 74, 1172–1199. Singletary S. J. and Grove T. L. (2003) Early petrologic processes on the ureilite parent body. Meteorit. Planet. Sci. 38, 95–108. Singletary S. and Grove T. L. (2006) Experimental constraints on ureilite petrogenesis. Geochim. Cosmochim. Acta 70, 1291–1308. Srinivasan P., McCubbin F. M. and Agee C. B. (2016) Assessing the Formation of Ungrouped Achondrite Northwest Africa 8186: Residue, Crystallization Product, or Recrystallized Chondrite? Lunar Planet. Sci. Conf. 47, 1620. Steele R. C. J., Coath C. D., Regelous M., Russell S. and Elliott T. (2012) Neutron-Poor Nickel Isotope Anomalies in Meteorites. Astrophys. J. 758, 59. Sugiura N. and Fujiya W. (2014) Correlated accretion ages and ε 54 Cr of meteorite parent bodies and the evolution of the solar nebula. Meteorit. Planet. Sci. 49, 772–787. Takeda H., Mori H. and Ogata H. (1989) Mineralogy of augite-bearing ureilites and the origin of their chemical trends. Meteoritics 24, 73–81. Tenner T. J., Nakashima D., Ushikubo T., Kita N. T. and Weisberg M. K. (2015) Oxygen isotope ratios of FeO-poor chondrules in CR3 chondrites: Influence of dust enrichment and H2O during chondrule formation. Geochim. Cosmochim. Acta 148, 228–250. Tenner T. J., Ushikubo T., Nakashima D., Schrader D. L., Weisberg M. K., Kimura M. and Kita N. T. (2018) Oxygen Isotope Characteristics of Chondrules from Recent Studies by Secondary Ion Mass Spectrometry. In Chondrules Cambridge University Press. pp. 196–246. Tissot F. L. H., Dauphas N. and Grove T. L. (2017) Distinct 238U/235U ratios and REE patterns in plutonic and volcanic angrites: Geochronologic implications and evidence for U isotope fractionation during magmatic processes. Geochim. Cosmochim. Acta 213, 593–617. Touboul M., Kleine T., Bourdon B., Van Orman J. A., Maden C. and Zipfel J. (2009) Hf–W thermochronometry: II. Accretion and thermal history of the acapulcoite–lodranite parent body. Earth Planet. Sci. Lett. 284, 168–178. Trinquier A., Birck J. and Allegre C. J. (2007) Widespread 54 Cr Heterogeneity in the Inner Solar System. Astrophys. J. 655, 1179–1185. Trinquier A., Elliott T., Ulfbeck D., Coath C., Krot A. N. and Bizzarro M. (2009) Origin of Nucleosynthetic Isotope Heterogeneity in the Solar Protoplanetary Disk. Science (80-. ). 324, 374– 376. Wadhwa M., Shukolyukov A. and Lugmair G. W. (1998) 53Mn-53Cr systematics in Brachina: A record of one of the earliest phases of igneous activity on an asteroid. Lunar Planet. Sci. Conf. XXIX 29, #1480 (abstract). Walker D. and Grove T. (1993) Ureilite smelting. Meteoritics 28, 629–636. Walsh K. J., Morbidelli A., Raymond S. N., O’Brien D. P. and Mandell A. M. (2011) A low mass for Mars from Jupiter’s early gas-driven migration. Nature 475, 206–209. Warren P. H. (2012) Parent body depth–pressure–temperature relationships and the style of the ureilite anatexis. Meteorit. Planet. Sci. 47, 209–227. Warren P. H. (2011) Stable isotopes and the noncarbonaceous derivation of ureilites, in common with nearly all differentiated planetary materials. Geochim. Cosmochim. Acta 75, 6912–6926. Warren P. H. and Kallemeyn G. W. (1989) Allan Hills 84025 - The second brachinite, far more differentiated than brachina, and an ultramafic achondritic clast from L chondrite Yamato 75097. Proc. 19th Lunar Planet. Sci. Conf., 475–486. Warren P. H. and Kallemeyn G. W. (1992) Explosive volcanism and the graphite-oxygen fugacity buffer on the parent asteroid(s) of the ureilite meteorites. Icarus 100, 110–126.

Weisberg M. K., McCoy T. J. and Krot A. N. (2006) Systematics and Evaluation of Meteorite Classification. Meteorites early Sol. Syst. II, 19–52. Wilson L. and Keil K. (2017) Arguments for the Non-existence of Magma Oceans in Asteroids. In Planetesimals (eds. L. T. Elkins-Tanton and B. P. Weiss). Cambridge University Press, Cambridge. pp. 159–179. Wilson L. and Keil K. (2012) Volcanic activity on differentiated asteroids: A review and analysis. Chemie der Erde - Geochemistry 72, 289–321. Wurm G., Trieloff M. and Rauer H. (2013) Photophoretic separation of metals and silicates: The formation of mercury-like planets and metal depletion in chondrites. Astrophys. J. 769, 78. Yamakawa A., Yamashita K., Makishima A. and Nakamura E. (2010) Chromium isotope systematics of achondrites: Chronology and isotopic heterogeneity of the inner solar system bodies. Astrophys. J. 720, 150–154. Young E. D., Shahar A., Nimmo F., Schlichting H. E., Schauble E. A., Tang H. and Labidi J. (2019) Near-equilibrium isotope fractionation during planetesimal evaporation. Icarus.

100

Figures and tables

p p p l l a a l 37 37 a g g g o H CI o LL 35 o u 64 u 64 u -1 t t 33 33 px out t IW) p px o 38 39 39 pi i -1.2 g g u 63 t 63 o 63 in u 57 p t pi p ig i -1.4 g 68 pig ou g in o 31 i n 71 u 57 46 60 42 t 42

-1.6 54 t 40 40 44 70 -1.8 48 62 41 solidus 41 43 s 43 27 o 27 l -2 id 62 so u 59 s lid

oxygen fugacity ( 65 65 -2.2 u s 58

p i 37 g i ureilite-like 35 CM n CV -1 64 residues: 33 33 pig-oliv-metal IW) 38 39 38 39 -1.2 4 px p 63 i 63 out g 6 o px u p t out -1.4 ig

i s n o 42 pi l id g u -1.6 p o s l 40 u a opx aug g 44 t 44 50 o 50 -1.8 u t 41 pig plag 43 27 43 s -2 p CI oli + oliv l a d g u + metal oxygen fugacity ( s out CM -2.2 ± Cr-spinel H-LL

1050 1100 1150 1200 1250 1100 1150 1200 1250 1100 1150 1200 1250 Temperature (¼C) Temperature (¼C) Temperature (¼C) Figure 1. Temperature-oxygen fugacity phase diagrams of MHC-PV experiments performed on five starting compositions: LL, H, CI, CM and CV. Experiments are identified by their number (e.g. 37 = CHS 37). See Chapter 1 for experimental procedures, conditions, detailed phase compositions and proportions of experimental phases. See Table 1 for a summary. The gray shaded areas highlight the conditions under which a “ureilite-like” phase assemblage made of olivine, pigeonite and metal is stable. Ureilites can only form by melting a high NaK# chondritic material (e.g. H, LL or CI).

101

Experiments (this study) Exp. (literature) (c) 2.5 LL chondrite CM chondrite Jurewicz et al. 1995 (H-LL) H chondrite CV chondrite Usui et al. 2015 (H) 2 IW CI chondrite Gardner-Vandy al. 2013 (R) 1.5 (a) (b) 1

0.5

IW) 0

-0.5 Brachinites BLA Ureilites -1 all primitive -1.5 achondrites A-L -2 oxygen fugacity ( -2.5 60 70 80 90 0 20 40 60 0 0.5 1 1.5 2 2.5 Fo in olivine (mol.%) Ni concentration in metal (wt.%) NiO concentration in olivine (wt.%) Figure 2. Forsterite content in olivine (a), Ni concentration in metal (b) and NiO content in olivine (c) in experiments as a function of the fO2, calculated with the CCO-buffer (French and Eugster, 1965) or measured (gas-mixing experiments; ∆IW > +0.5). The black curve in (a) represents the fO2 of the OSI equilibrium at 1100 ºC (Nitsan, 1974). The colored boxes represent the range of olivine compositions (Fo (a) and NiO (c)) in primitive achondrites, which constrain their fO2 of melting. BLA: Brachinite-like achondrites. A-L: acapulcoites-lodranites.

102

95 (a) Ureilites (b) 90 1-atm gas-mix MHC-PV exp. IW +0,8 IW -0.8 / -2.2 A-L 85 CM LL H IW +0.8 H 80 BLA CI 75 CM CV

Fo in olivine 70 Brach. 65

55 051015200 20 40 60 80 metal fraction (wt.%) Ni concentration in metal (wt.%) Figure 3. (a) Metal fraction in experiments as a function of the forsterite content in olivine. (b) Ni concentration of the metal as a function of Fo content. For each composition, the metal fraction increases linearly with increasing Fo content in olivine (~15 wt.% metal for 25 Fo units), as a consequence of the reduction of Fe2+ in Fe0. Simultaneously, Ni content in the metal decreases by dilution. The colored boxes represent the range of primitive achondrites (see Fig. 2). No obvious correlations of the metal fraction with Fo content are observed within the groups. Metal fraction is highly variable in acapulcoites and lodranites. The extremities of the darker A- L box (x axis) represent the average metal fractions for acapulcoites (left) and lodranites (right; Keil and McCoy, 2018).

103

IW -1 IW -1.5 IW -2 IW -2.5 IW -1 IW -1.5 IW -2 IW -2.5

(a) experiments (b) 0.8 0.8 GRV 020043 70 LL 0.7 0.7 Watson 012 (H7) H Anorthite in plag solid. 60 0.6 CI CM/CV 0.6 CM/CV sub- CM -solid. 0.5 0.5 50 CV LL sub H6 L6 ag out 0.4 0.4 pl LL6 40 0.3 olid. id. 0.3 py = px/(px+oliv) sub-s H 0.2 H CI sub-sol 0.2 30 0.1 increasing F 0.1 20 60 70 80 90 100 60 70 80 90 100 (c) (d) 0.8 0.8 (b) (c) (d) primitive achond. 0.7 0.7 Brach. 0.6 0.6 BLA Ureil. 0.5 0.5 Acap. 0.4 0.4 Lodr. out t Win. g ou 0.3 0.3 ag pla pl type 6 H-L-LL py = px/(px+oliv) chondrites 0.2 0.2 CM CI type 4 “acapulco 0.1 0.1 chondrite”

60 70 80 90 100 60 70 80 90 100 Fo in olivine Fo in olivine Figure 4. Relative proportions of olivine and pyroxene in experiments, ordinary chondrites and primitive achondrites as a function of the forsterite content in olivine. The fraction of total pyroxene (opx+pig+aug) relative to olivine is expressed as py: pxtot/(pxtot+oliv) in wt. %. For primitive achondrites, py was calculated by normalizing modal compositions by the phase densities. Most phase proportions and compositions are from (Day et al., 2012; Hunt et al., 2017; Keil, 2014; Keil and McCoy, 2018; Singletary and Grove, 2003). Sub-solidus curves for the LL, H, CI, CM and CV compositions and An contents (color) are calculated by linear regression of the olivine, pyroxene and melt fraction and plagioclase composition in experiments (see supplementary material). In experiments and natural samples, high Fo contents and high py correspond to more reducing conditions of equilibration/melting (lower fO2). In experiments, but not always in natural samples (e.g. ureilites), high Fo contents and high py are also associated with larger metal fractions (Fig. 3). Ordinary chondrites (Dunn et al., 2010) have py identical to H and LL sub-solidus curves calculated from experiments. All primitive achondrites have lower py, indicating that they lost a silicate melt with a pyroxene-rich component during melting.

104

0.5 ureilites 0.4 NWA 3153

0.3 brachina

0.2 CaO (wt.%) MHC-PV exp. IW -0.8 / -2.5 1-atm LL other brachinites gas-mix H 0.1 IW +0.6 / +0.8 CI acapulcoites-lodranites CM H CM winonaites CV 0 0 0.2 0.4 0.6 0.8 Cr O (wt.%) 2 3

Figure 5. Concentration of minor elements in olivine (CaO and Cr2O3) in experiments (symbols) and primitive achondrites (gray contours). Note that only ureilites coarsely match the composition of experiments. The other groups were cooled too slowly and re-equilibrated during retrograde metamorphism. Ureilites data from the literature of Chapter 3 and reference therein. All other achondrites data from Goodrich et al. (2017).

105

70 6 R 6 R LL 6 L 6 LL H 6 4 60 6 chondrites + prim. achond.type EH/EL 6 L Brachina H Win tes 7531 ili lith.B An content in plag Acap re 50 2 u brac O ap- r ac od 40 17 l LEW CM CCAM δ 0 11561 Earth, Moon, EC, aubTaf R C 3100 V C ungr. NC PA 30 -2 ungr. CC PA angr 3133 eucr 10503 K 20 C GRA 06128 -4 8186 NWA 11575 CO ALM A 10 -2 0 2 4 6 8 18 δ O Figure 6. Oxygen isotope compositions of primitive achondrites that have not lost a significant fraction of melt (e.g. Acapulco, Brachina, Winona) and equilibrated chondrites (type 6) containing plagioclase. Most primitive meteorites have a plagioclase rich in Na2O (An10-20) and only a few samples with affinities for CR and CK-CV chondrites have a plagioclase more calcic than An30. The anorthite content can be used as a proxy for the bulk NaK# and extent of depletion in MVE (Fig. 7). Taf= Tafassasset, LEW = LEW 88763. Numbers refer to NWA meteorites. Most oxygen isotope data are from the compilations of Dauphas (2017), Schmitz et al. (2016) and Scott et al. (2018). NWA 7531 lith. B is from Bunch et al. (2013).

106

2 011 2976 Taf 6704 CC 1.5 3100 CI 60 2994 CH 3133 10503 (if equilibrated near-solidus) CR CB

25 An content in plag 1 CM 50 CV CO Cr 8186 33 40

54 0.5 CK ε NC EC

Au initial bulk NaK# 0 5363 R 30 5400 Mars Bra H GRA L LL -0.5 Ac An 20 HED 50 Ur Win -1 -5 -4 -3 -2 -1 0 1 2 3 17 O Figure 7. Oxygen isotope compositions (∆17O = δ17O – 0.52 δ18O (Clayton and Mayeda, 1988), mass independent fractionation relative to the Earth) and ε54Cr of meteorites illustrating the dichotomy between the carbonaceous (CC) and non-carbonaceous reservoirs (NC). Terrestrial planets are part of the NC reservoir, which is therefore assumed to represent the inner solar system. The Earth-Moon system is at the origin [0,0]. All parent bodies for which the initial composition can be estimated in the NC reservoir were rich in Na2O and not depleted in MVE relative to the sun’s photosphere and CI chondrites. Isotopic data from the compilations of Dauphas and Schauble (2016), Schmitz et al. (2016), Scott et al. (2018). See references therein.

107

Table 1. Summary of experiments exp T ∆ IW F py CaO/Al O olivine opx pig plag metal 2 3 wt.% (of liq, wt.%) Fo Cr2O3 CaO NiO En Wo En Wo An Ni

C CHS 69 1301 -2.44 31.3 0.19 0.78 94.1 0.70 0.21 91.4 0.8 n.a

I CHS 67 1274 -1.59 34.1 0.00 0.76 83.1 0.75 0.22 0.04 n.a

CHS 66 1250 -2.53 23.4 0.37 0.64 95.1 0.69 0.26 0.04 90.7 3.4 n.a

CHS 71 1216 -1.59 0.12 83.9 0.77 0.28 82.4 3.0 n.a

CHS 64 1202 -1.08 28.1 0.00 0.85 74.1 0.58 0.28 0.11 30.6

CHS 63 1202 -1.38 27.3 0.04 0.72 80.1 0.64 0.29 0.04 77.7 4.6 11.6

CHS 65 1201 -2.27 20.0 0.38 0.50 92.5 0.67 0.33 0.03 86.3 5.9 n.a

CHS 62 1192 -1.87 22.4 0.26 0.57 84.6 0.68 0.31 0.08 78.1 5.8 9.8

CHS 70 1167 -1.84 18.5 0.29 0.42 85.4 0.59 0.30 77.4 8.4 n.a

CHS 68 1159 -1.53 22.6 0.08 0.59 78.9 0.48 0.34 0.04 72.2 8.2 n.a

CHS 58 1136 -2.27 89.2 0.54 0.27 82.6 5.0 53.9 37.7 7.9

CHS 57 1132 -1.44 20.2 0.19 0.53 75.9 0.40 0.31 0.05 69.7 8.1 20.0

CHS 54 1129 -1.55 78.1 0.41 0.36 0.05 71.3 9.2 13.3

CHS 60 1104 -1.63 14.4 0.22 0.26 79.0 0.32 0.25 0.07 77.6 5.2 58.4 29.5 32.5 n.a.

CHS 59 1100 -2.19 11.2 0.36 0.18 87.6 0.41 0.23 82.4 5.6 58.5 31.9 29.5 9.8

H CHS 69 1301 -2.44 22.4 0.50 0.73 93.2 0.55 0.22 0.00 93.1 0.9 n.a.

CHS 61 1301 -1.77 28.3 0.27 0.69 86.9 0.55 0.20 0.03 86.4 1.6 7.6

CHS 27 1244 -2.08 16.5 0.54 0.56 89.1 0.52 0.24 0.05 86.1 3.3 4.6 108 CHS 63 1202 -1.38 18.8 0.31 0.61 78.2 0.59 0.26 0.05 77.2 3.5 7.0

CHS 65 1201 -2.27 14.8 0.58 0.45 91.9 0.47 0.25 0.07 87.5 4.4 n.a.

CHS 48 1194 -1.98 87.8 0.53 0.24 0.05 83.4 4.4 5.3

CHS 62 1192 -1.87 14.2 0.52 0.48 84.4 0.53 0.25 0.06 78.9 4.6 6.7

CHS 46 1191 -1.61 18.8 0.25 0.59 76.8 0.56 0.30 0.10 75.7 4.0 8.2

CHS 43 1163 -2.06 12.1 0.35 90.0 0.41 0.22 0.07 82.9 6.7

CHS 44 1161 -1.86 13.1 0.48 0.41 85.6 0.48 0.24 0.01 79.8 6.3 4.8

CHS 41 1142 -2.00 9.9 0.54 0.29 87.1 0.38 0.23 0.07 81.2 4.6 70.1 18.9 43.9 5.5

CHS 39 1140 -1.30 16.6 0.20 0.50 70.9 0.36 0.29 0.07 68.6 7.5 9.0

CHS 57 1132 -1.44 15.6 0.19 0.52 73.0 0.31 0.37 0.01 68.6 7.3 10.0

CHS 33 1112 -1.18 12.7 0.15 0.45 64.6 0.17 0.37 0.08 62.4 10.0 41.4 17.8

CHS 40 1108 -1.81 7.9 0.51 0.24 84.2 0.34 0.24 0.06 79.4 3.5 53.7 36.3 38.1 5.3

CHS 42 1085 -1.65 5.0 0.48 0.17 81.6 0.29 0.22 0.05 77.0 4.6 53.5 33.0 28.4 6.3

CHS 35 1079 -1.04 5.3 0.10 0.24 61.6 0.22 0.30 0.09 69.2 2.8 48.2 32.3 28.7 23.3 CHS 37 1063 -0.98 5.0 0.08 60.8 0.20 0.29 0.11 64.7 3.3 45.6 34.5 24.7 30.0

CH 8* 1076 0.56 13.4 0.05 0.40 59.0 0.12 0.30 0.66 44.4 34.1 72.4 CH 5* 1073 0.76 11.1 0.04 0.41 58.7 0.14 0.31 0.90 41.9 38.5 42.1 75.6

py = pyroxene/(pyroxene+olivine) see Chapter 1 for detailed phase compositions and proportions

108

Table 1. Summary of experiments (continued) exp T ∆ IW F py CaO/Al O olivine px1 px2 plag metal 2 3 wt.% (of liq, wt.%) Fo Cr2O3 CaO NiO En Wo En Wo An Ni

LL CHS 27 1244 -2.08 18.9 0.55 0.58 87.9 0.47 0.29 0.03 85.3 3.6 6.8

CHS 64 1202 -1.08 24.9 0.17 73.0 0.50 0.24 0.07 72.7 4.1 24.8

CHS 63 1202 -1.38 21.7 0.31 0.71 77.7 0.52 0.27 0.05 76.2 3.7 12.5

CHS 31 1202 -1.56 18.2 0.39 0.77 80.1 0.50 0.26 78.4 4.4 9.7

CHS 43 1163 -2.06 15.3 0.62 0.57 89.0 0.42 0.25 0.08 83.3 4.2

CHS 41 1142 -2.00 11.6 0.58 0.40 86.8 0.40 0.23 0.06 81.5 4.9 67.4 24.1 41.7 6.3

CHS 39 1140 -1.30 17.4 0.32 0.28 72.6 0.40 0.31 0.03 70.0 7.0 16.2

CHS 33 1112 -1.18 14.9 0.31 0.49 70.5 0.27 0.32 0.05 72.4 4.1 64.2 12.9 35.0 32.4

CHS 40 1108 -1.81 10.9 0.56 0.37 83.8 0.31 0.23 0.09 79.0 4.9 31.1 7.8

CHS 38 1086 -1.30 6.8 0.39 0.23 72.6 0.23 0.25 0.07 74.2 4.7 57.7 23.8 30.6 17.9

CHS 42 1085 -1.65 5.5 0.51 0.17 81.4 0.27 0.28 0.04 77.2 5.4 53.5 33.0 30.9 7.7

CHS 37 1063 -0.98 7.6 0.33 0.19 68.5 0.14 0.24 0.08 73.0 4.2 47.7 32.7 23.0 42.3

CM CHS 50 1248 -1.89 22.7 0.27 0.76 87.7 0.53 0.22 0.04 85.6 3.0 5.8

CHS 27 1244 -2.08 21.4 0.36 0.70 88.6 0.50 0.28 0.02 85.8 3.4 7.4

CHS 64 1202 -1.08 27.9 0.00 0.80 70.0 0.44 0.31 0.06 16.2

CHS 63 1202 -1.38 23.9 0.13 0.75 77.4 0.47 0.29 0.03 75.4 4.2 7.7

CHS 6 1185 -1.44 74.7 0.49 0.29 73.9 4.1 n.a.

109 CHS 4 1166 -1.34 71.7 0.38 0.31 73.3 6.1 n.a.

CHS 43 1163 -2.06 10.4 0.46 0.53 89.9 0.40 0.28 0.02 81.8 7.3 61.8 30.9 60.7 5.2

CHS 44 1161 -1.86 12.7 0.34 0.62 85.4 0.53 0.24 0.01 79.1 7.2 59.5 5.6

CHS 41 1142 -2.00 0.0 0.50 87.3 0.35 0.27 0.06 81.9 4.6 57.3 34.5 49.3 5.6

CHS 39 1140 -1.30 15.0 0.12 0.70 72.2 0.32 0.34 0.04 66.4 10.7 58.9 10.3

CHS 33 1112 -1.18 5.2 0.08 0.57 65.6 0.23 0.37 0.02 50.0 30.6 49.3 21.6

CHS 40 1108 -1.81 2.2 0.46 84.5 0.32 0.23 0.05 49.5 36.1 48.1 5.9

CHS 38 1086 -1.30 0.0 0.00 75.8 0.27 0.23 0.04 71.5 7.6 54.4 32.2 46.5 0.0

CHS 42 1085 -1.65 0.0 0.43 81.9 0.30 0.27 0.05 77.3 4.1 60.7 26.6 47.1 5.1

CHS 35 1079 -1.04 0.0 0.09 64.7 0.21 0.30 0.05 73.0 2.7 44.2 40.2 48.0 28.1 CHS 37 1063 -0.98 0.0 0.10 64.0 0.22 0.35 0.09 68.6 2.9 44.7 36.9 43.2 31.0

CH 10* 1118 0.71 12.0 0.02 0.86 63.1 0.18 0.39 0.81 55.0 20.9 71.4 72.2

CV CHS 50 1248 -1.89 27.3 0.09 0.76 87.3 0.52 0.22 0.03 85.3 3.2 6.1

CHS 64 1202 -1.08 27.1 0.00 0.82 72.6 0.40 0.30 0.04 15.4

CHS 63 1202 -1.38 28.1 0.01 0.78 78.4 0.51 0.28 0.02 75.8 4.2 8.9

CHS 43 1163 -2.06 0.0 0.52 89.5 0.46 0.29 0.02 84.0 4.3 54.5 39.0 66.0 6.1

CHS 44 1161 -1.86 0.0 0.45 84.3 0.45 0.29 0.06 79.1 6.0 55.2 35.6 67.7 7.0

CHS 39 1140 -1.30 9.6 0.16 0.73 70.6 0.32 0.36 0.06 64.8 12.0 57.7 23.2 61.7 12.4

CHS 33 1112 -1.18 3.9 0.10 0.63 66.5 0.00 0.32 0.04 48.7 29.7 58.3 24.1 CHS 38 1086 -1.30 0.0 0.15 71.2 0.27 0.30 0.07 73.1 4.9 45.3 33.6 54.9 12.7

py = pyroxene/(pyroxene+olivine) see Chapter 1 for detailed phase compositions and proportions

109

Table 2. Average silicate melts extracted primitive achondrites 1 2 brach 3 urei 4 lodr 5 6 7 SiO2 59.5 60.0 60.9 63.8 65.5 63.3 62.1 TiO2 0.51 0.53 0.50 0.52 0.45 0.55 0.45 Al2O3 14.7 14.3 14.6 15.8 17.6 14.2 12.0 Cr2O3 0.07 0.11 0.13 0.17 0.06 0.31 0.56 FeO 10.6 8.92 7.45 3.77 2.64 3.39 2.88 MnO 0.13 0.16 0.16 0.17 0.13 0.15 0.45 MgO 2.86 3.44 3.62 4.01 2.41 5.61 8.72 CaO 5.91 5.88 5.74 5.21 3.10 6.36 7.68 Na2O 4.69 4.99 5.52 5.93 6.98 5.53 4.70 K2O 0.56 0.87 0.81 0.86 0.99 0.61 0.39 P2O5 0.57 0.94 0.65 0.14 0.12 0.03 0.07 NiO 0 0.01 0.02 0.02 0.04 0 0 total1 100.3 99.7 99.9 100.0 101.4 100.45 99.9

CaO/Al2O3 0.40 0.41 0.39 0.33 0.18 0.45 0.64 T 1080 1115 1123 1149 1100 1200 1250 6.5 ± 19 ± F (wt.%) 13.4 12.8 ± 2 14.8 ± 2 12 ± 2 2† 14.8 2† log fO2 (Δ IW) 0.6 -1.2 -1.5 -2.0 -2.2 -2.3 -2.5

Kd oliv-liq 0.32 0.30 0.29 0.27 0.23 0.26 0.28 ρ (g/cm3) 2.58 2.53 2.52 2.45 2.42 2.40 2.50 log µ (Pa.s) 4.04 3.71 3.73 3.92 4.83 3.25 2.47 1: melt produced by melting H chondrite at fO2 more oxidizing than IW (CH 8, for comparison with 2)

2: melt extracted from the brachinites parent body (CHS 33, 35, 39, 57; H, LL and CI) 3: melt extracted from the UPB; dominant FeO-rich region (CHS 33, 39, 42, 57, 60 and 70; H, LL and CI)

4: melt extracted from the lodranites parent body (CHS 40, 41, 43, 44 and 65, H and LL) 5-7: melt composition that could have formed in the winonaites PB; not extracted (CHS 59, CHS 65, CHS 66)

Melts of H, LL and CI compositions are identical at a given temperature (supplementary material) † melt fraction estimated from the difference in F between CI and H compositions (-5 wt.%; Fig. S1)

110

Supplementary Material

1. Linear regressions of solidus temperatures and An contents in plagioclase used in figures

The temperatures of the solidus of the H, LL and CM compositions were calculated by linear regression of the experimental temperature as a function of the melt fraction and either the fO2 or the Fo content in olivine (for experiments with F < 15 wt.%). For CI and CV experiments, due to the limited number of near-solidus experiments, the solidus temperature was estimated at

Fo 75-80 (IW -1.5) only and assumed to share the same fO2 dependence than H, LL and CM compositions.

H : T(ºC) = (F (in wt. %) + 1.269 – 0.0513 * ∆ IW) / 0.001291 (1)

LL : T (ºC) = (F (in wt.%) + 1.58 – 0.0743 * ∆ IW) / 0.001622 (2)

CM: T (ºC) = (F (in wt.%) + 2.069 – 0.0691 * ∆ IW) / 0.002 (3)

Alternatively:

H : T(ºC) = (F (in wt. %) + 1.116 + 0.00206 * Fo) / 0.001226 (4)

CM: T (ºC) = (F (in wt. %) + 2.204 + 0.00292 * Fo) / 0.002223 (5)

Equations (1-5) work best between IW -1 and IW -2 and for a melt fraction of 0 (solidus) to 15.

CI: T (ºC) = (F (in wt. %) + 1.539 + 0.002964 * Fo) / 0.001737 (6)

CV: T (ºC) = (F (in wt. %) + 3.158 + 0.01093 * Fo) / 0.003537 (7)

Equations (6-7) should only be used with Fo 75-80 and F = 0-15.

To calculate An content as a function of the Fo content and py (i.e. px/(oliv+px)), we first parametrize An as a function of the experimental temperature (Fig 6b in main manuscript):

H and CI: An = 0.2401 * T – 231 (8)

CM: An = 0.1553 * T – 121.9 (9)

Next, we parametrize the total pyroxene and olivine fraction as a function of the Fo content in olivine and the melt fraction:

H: px = 0.01005 * Fo – 0.4755 * F – 0.4755 (10)

111

oliv = -0.01736 * Fo + 0.1741 * F + 1.784 (11)

LL: px = 0.00888 * Fo – 0.449 * F – 0.297 (12)

oliv = -0.01567 * Fo + 0.22 * F + 1.627 (13)

CM: px = 0.01104 * Fo – 0.73 * F – 0.621 (14)

oliv = -0.0185 * Fo + 0.25 * F + 1.945 (15)

Combining equations (1-15) gives An as a function of py and Fo for H, LL and CM compositions (Fig. 10 in main manuscript).

H: An = -185 + 4.19 * Fo – 841.3 * py + 8.449 * Fo * py – 264.1 * py2

CI: An = -126.3 + 2.507 * Fo – 607.4 * py + 6.446 * Fo * py – 192.7 * py2 CM: An = -25.15 + 1.225 * Fo – 196.4 * py + 1.878 * Fo * py – 58.69 * py2

112

Chapter 3: Incremental melting in the ureilite parent body: initial composition, melting temperatures and melt compositions

Abstract

Ureilites are carbon-rich ultramafic achondrites that have been heated above the silicate solidus, do not contain plagioclase, and represent the melting residues of an unknown planetesimal (i.e. the ureilite parent body, UPB). Melting residues identical to pigeonite-olivine ureilites (representing 80 % of ureilites) have been produced in batch melting experiments of chondritic materials not depleted in alkali elements relative to the sun’s photosphere (e.g. CI, H, LL chondrites), but only in a relatively narrow range of temperature (1120-1180 ºC). However, many ureilites are thought to have formed at higher temperature (1200-1280 ºC). New experiments, described in this study, show that pigeonite can subsist at higher temperature (up to 1280 ºC) when CI and LL chondrites are melted incrementally and while partial melts are progressively extracted. The melt productivity decreases dramatically after the exhaustion of plagioclase with only 5-9 wt.% melt being generated between 1120 and 1280 ºC. The relative proportion of pyroxene and olivine in experiments are compared to 12 ureilites, analyzed for this study, together with ureilites described in the literature to constrain the initial Mg/Si ratio of the UPB (0.98-1.05). Experiments are also used to develop a new thermometer based on the partitioning of Cr between olivine and low-Ca pyroxene that is applicable to all ureilites. The equilibration temperature of ureilites increases with decreasing Al2O3 and Wo contents of pyroxene and decreasing bulk REE concentrations. The UPB melted incrementally, at different fO2, and did not cool significantly (0-30 ºC) prior to its disruption. It remained isotopically heterogenous but the initial concentration of major elements (SiO2, MgO, CaO, Al2O3, alkali elements) was similar in the different mantle reservoirs.

113

1. Introduction

Ureilites are the largest group of ultramafic achondrites (550 samples at the time of writing) and the second largest group of achondrites after howardites-eucrites-diogenites. Most ureilites are composed of olivine and pigeonite with a significant fraction of metal (<5 wt.%), graphite (~3 wt.%), and traces of sulfide concentrated along silicate grain boundaries. Fewer samples contain orthopyroxene and augite instead of or in addition to pigeonite (e.g. Goodrich et al., 2001; Takeda et al., 1989). With the exception of brecciated ureilites (i.e. polymict ureilites), they are completely devoid of plagioclase. Their bulk composition is depleted in incompatible lithophile elements and chalcophile elements relative to all chondrites. While they were once described as ultramafic igneous cumulates (Berkley et al. 1980; Goodrich et al. 1987), most ureilites are now recognized as residues of partial melting representing the mantle of a planetesimal that lost abundant FeNi-sulfide eutectic melts (e.g. Barrat et al., 2015; Goodrich et al., 2013a; Warren et al., 2006) and silicate melts (e.g. Barrat et al., 2016; Kita et al., 2004; Scott et al., 1993; Warren and Kallemeyn, 1992). The Ureilite Parent Body (UPB) was violently disrupted while its internal temperature was still high (1150-1300 ºC; e.g. Sinha et al., 1997; Takeda et al., 1989). Because cooling was rapid following the destruction of the UPB (2-20 ºC/h; Goodrich et al., 2001; Herrin et al., 2010; Miyamoto, 1985; Takeda et al., 1989), the high temperature mineral assemblage has been preserved. Ureilites represent “quenched melting residues”, the only ones in the meteorite record, and document how planetesimals melted in the early solar system. However, the initial bulk composition of the UPB, the melting conditions and the composition of the melts that were extracted have proved challenging to decipher. No chondritic sample representing the primitive material has been found and remnants of the associated silicate melts are extremely scarce (Bischoff et al., 2014; Chapter 1; Kita et al., 2004). It is clear that the UPB was and remained isotopically heterogenous during partial melting until it was disrupted. Ureilites have heterogeneous oxygen isotope compositions and form a trend of mass-independent fraction (i.e. slope 1 in the three-isotope plot) which correlates with the Mg# (MgO/(MgO+FeO) in mol.%) of ferromagnesian silicates (Clayton and Mayeda 1988; Greenwood et al. 2017). The forsterite content of olivine cores (Fo, equivalent to the Mg#) varies between 74 and 96 with increasing 16O contents. In addition, ureilites are characterized by heterogeneities in C (Barrat et al. 2017), Cr (van Kooten et al. 2017; Zhu et al. 2019) and noble gas isotopes (Broadley et al. 2019), which also appear to correlate with O isotope compositions.

114

The initial composition of the UPB in major elements has long been unknown. The abundance of graphite and the oxygen isotopes point to a connection with carbonaceous chondrites (Clayton and Mayeda 1988; Goodrich et al. 2015) while nucleosynthetic anomalies in Cr, Ti and Ni are more akin to ordinary chondrites and characteristic of the “non-carbonaceous group” (Warren 2011). We have recently described batch melting experiments of a representative spectrum of chondritic materials (LL, H, CI, CM, and CV) and showed that pigeonite-olivine residues resembling ureilites can be produced by melting high NaK# compositions such as CI, H and LL chondrites (50; with NaK# = (Na+K)/(Na+K+Ca) in atom.%). The UPB could not have been depleted in alkali elements because low NaK# compositions (CV and CM) produce orthopyroxene-olivine residues (Chapter 2). Batch melting experiments of H, LL and CI chondrites produce pigeonite-olivine residues in a relatively narrow temperature interval, between 1120 and 1180 ºC, while many ureilites are thought to have formed at 1200-1270 ºC (Sinha et al. 1997; Takeda et al. 1989). Here, we investigate whether pigeonite can subsist at higher temperature when CI and LL chondrites are melted incrementally and while partial melts are progressively extracted. Experimental results are compared to 12 ureilites that were analyzed for this study and ureilites described in the literature to constrain the initial composition of the UPB and its Mg/Si ratio in particular. Then, we develop a new thermometer based on the partitioning of Cr between olivine and Low-Calcium Pyroxene (LCP; orthopyroxene or pigeonite) that can be used to constrain the temperature of equilibration

$%&'()"* of all ureilites. The D"# thermometer is used to distinguish between ureilites representing simple melting residues and ureilites representing cumulates or the products of more complex igneous processes. We compare the former (most pigeonite-olivine ureilites) to our experiments to constrain melting processes and estimate the composition of the last silicate melts that were in equilibrium with ureilites. We also use the new temperature of equilibration of ureilites to discuss the thermal history of the UPB and the nature of the heterogeneity in fO2 and oxygen isotopes of the UPB.

115

2. Methods

2.1. Experimental approach

All experiments were performed in the MIT Experimental Petrology Laboratory in a Molybdenum Hafnium Carbide pressure vessel (MHC-PV) placed in a Deltech vertical furnace following the same overall approach described in Walker and Grove (1993) and Singletary and Grove (2006). One key difference is that several starting compositions (~10 mg each) were equilibrated simultaneously in individual graphite capsules positioned vertically in a single Pt outer capsule. The temperature gradient is small at the base of the MHC-PV (< 3 oC over the length of the Pt capsule) and all experimental charges were equilibrated under the same P-T-fO2 conditions. The outer Pt capsule was loosely crimped but not sealed to allow for equilibration with the CO pressure medium and set the fO2 to the CCO buffer (French and Eugster 1965). In all but two experiments, the temperature was fixed for the duration of the experiments. In the other two experiments (CHS 46 and 71), the temperature was first raised 20-50 ºC above the final temperature for two hours. After seven hours to five days, experiments were terminated by pulling the MHC-PV out of the furnace, inverting it and hitting it with a wrench to ensure that the capsule dropped into the water-cooled head of the pressure vessel and quench. Experiment conditions are summarized in Table 1. This study is a complement to a larger series of melting experiments of chondritic materials (Chapter 1 and 2). In previous manuscripts, we reported on experimental charges that contain strictly chondritic bulk compositions (H, LL, CI, CM and CV) and simulate the batch melting of planetesimals. In this study, we describe distinct experimental configurations that simulate incremental melting processes. The batch melt compositions of Chapter 1 were mathematically subtracted from the CI and LL compositions to obtain melting residue analogs. These new compositions were synthetized by mixing high purity oxides, silicates (CaSiO3), carbonates

(Na2CO3, K2CO3) and Fe metal (Grove and Bence 1977). Starting materials were mixed for four hours in an automatic mortar and pestle and pre-conditioned for three days in a gas-mixing furnace at 1000 ºC and QFM –1 (IW + 2.5). These starting compositions (CIR1, CIR2, CIR3 and LLR1, LLR2, Table 2) represent melting residues of either a LL or a CI composition from which increasing fractions of batch melts were removed (11, 15.5, 19.5 wt.% and 8.5, 16 wt.%, respectively).

116

During incremental melting, successive or “instantaneous” melt compositions are expected to vary as melting progresses and become rapidly depleted in incompatible elements (e.g. K2O,

Na2O) relative to batch melts. However, the composition of the aggregate melt (i.e. the sum of all melt increments) and the composition of the residue are virtually identical to the composition of the melt and residue produced by batch melting, after the same total degree of melting (e.g. see pMELTS simulations in the supplementary material; Ghiorso et al., 2002). Accordingly, CIR1-3 and LLR1-2 compositions can also be understood as the melting residues from which the melt was extracted as successive increments of any size. The LL and CI compositions were selected because, like the initial composition of the UPB (Chapter 2), they are not depleted in alkali-elements relative to the sun’s photosphere (NaK# = 50). Yet, they are characterized by contrasting Mg/Si (0.92 and 1.03, respectively) and Ca/Si ratios (0.052 and 0.06). In addition, CI and LL compositions are relatively close in Mg# (68.5-70) to the most FeO-rich ureilites (74), after removal of sulfur as FeS. Two additional compositions, RCa and RCa2, represent arbitrary “near-chondritic” compositions depleted in Al2O3, Na2O and K2O but enriched in CaO. They were included to produce two-pyroxene assemblages that could be compared to orthopyroxene-augite ureilites.

2.2. EPMA analyses

The JEOL-JXA-8200 SuperProbe electron probe micro analyzer (EPMA) of the MIT Electron Microprobe Facility was used to acquire quantitative analyses of experimental run products as well as olivine and pyroxene in 12 ureilites. Silicates were analyzed with a focused beam, a voltage of 15 kV and a current intensity of 10 nA. All elements were measured for 40 s (20 s for backgrounds) except Na, which was counted first for 8 s (4 s for backgrounds). In many experiments, the small quantity of glass (<5 wt.%) wetting grain boundaries, but rarely pooling, precluded reliable measurement of its composition. When attempted, the analysis of experimental glasses was done with a 5 nA current intensity, a spot size of 2-4 µm and a counting time of 4 s for Na. Elemental X-ray maps of Si, Mg, Fe, Ca and Al were acquired for 10 of the 12 ureilites. The maps were performed with a beam size of 5-10 µm, a step size of 15-25 µm, a voltage of 15 kV, a current intensity of 30 nA and a counting time of 40 ms. Elemental intensities were acquired simultaneously with five wavelength dispersive spectrometers (WDS).

117

2.3. Calculation of the mode of ureilites and experiments

Elemental X-ray maps were compiled with MultiSpec© and used to create automatically 50 clusters of pixels in 5-dimensions with the algorithm ISODATA. Pixel clusters were then manually grouped into classes representing the different phases: (1) Fe-metal, sulfide and oxides, (2) olivine, (3) low-Ca pyroxene and, if present, (4) augite. Graphite, the residual porosity and all ambiguous clusters were grouped as background. The number of pixels in each class, representing the modal composition of ureilites in vol.%, were counted with ImageJ. This treatment was applied to the 10 ureilite thin sections mapped for this study and to eight additional ureilites studied by Singletary and Grove (2003) using their original X-ray maps. The phase proportions in experiments were estimated by mass balancing the composition of the different phases against the bulk composition of the starting material with the function fitlm in Matlab©. However, because the chemical composition of most melts could not be analyzed, we first had to estimate their composition by altering the composition of experimental batch melts at the same temperature and fO2 conditions (Chapter 1). The concentrations of SiO2, Al2O3 and alkali elements were progressively lowered while the concentrations of CaO and, to a lesser extent, FeO and MgO were slightly increased. Five to ten multiple linear regressions were performed and the one with the lowest sum of squared residuals was selected. In practice, we find that as long as a melt is included in mass balance calculations, the relative proportions of olivine and pyroxene are not very sensitive to small variations in the estimated melt composition. However, when no melt is included, mass balances tend to overestimate the proportion of pyroxene relative to olivine by ~ 5 wt.% (or 5 py unit; where py is the ratio px/(px+oliv)*100 in wt.%) even in experiments that apparently contain only a few percent of liquid (e.g. LLR2 in CHS 47, 51). Finally, to confirm that this method provides reasonable estimates of py, we calculated phase proportions in vol.% by image analysis of 10 backscattered electron (BSE) images for two experimental charges: LLR2 in CHS 47 and CIR2b in CHS 63. Modal compositions in vol.% were converted in wt.% by multiplying the phase proportions by the densities of olivine and pyroxene of appropriate composition. The py calculated by images analyses are consistent (within ~3 py units) with the ones calculated by mass balance: 23.1 (vs. 21.2) and 42 (vs. 44.6) for CHS 47 and CHS 63, respectively.

118

3. Results

3.1. Experiments on chondritic residues

Most experimental charges performed using the LLR1-2 and CIR1-3 bulk compositions (chondritic residues) contain olivine, orthopyroxene or pigeonite, glass and metal (Fig. 1). Only one charge, CIR2 in CHS 65 (1201 ºC, IW -2.3), has minor augite in addition to pigeonite. Experimental runs using CIR1 and LLR1, the least refractory residues, contain more melt than CIR2-3 and LLR2 runs at a given temperature. We calculate the aggregate melt fraction (i.e. the total amount of melt produced from the initial chondritic composition), by adding the melt fraction in CIR1-3 or LLR1-2 experiments to the melt fractions associated with the different starting compositions (Table 2). At a given temperature, aggregate melt fractions are significantly smaller (15-22 wt.%) in re-melted residues than in batch melting experiments (up to 35 wt.%; Fig. 2). The main effect of removing partial melts as melting progresses is to considerably lower the melt productivity at higher temperature. Because melt removal slows down the melting process considerably, the pigeonite (i.e. LCP with Wo content greater than 5, when defined based on composition) can subsist in experimental residues up to ~1280 ºC, 80 ºC higher than in batch melting experiments (Fig. 3). The bulk CIR2 composition corresponds to the residue immediately following the disappearance of plagioclase (F=15.5 wt.%; Chapter 1) and produces pigeonite with a high wollastonite content (Wo) at the highest temperature (1280 ºC). LLR1-2 charges contain a pigeonite with lower Wo content and pigeonite disappears at ~1220 ºC (as opposed to 1170 ºC in batch melting experiments of H and LL compositions). The pigeonite in LLR1-2 charges has a lower Wo content because py is significantly larger than in CIR1-3 charges (Fig. 4). The larger proportion of pyroxene relative to olivine in experimental charges with LLR1-2 compositions results from their lower bulk Mg/Si ratios (Table 2). In other words, in LLR1-2 charges, similar bulk CaO concentrations are “diluted” into a larger volume of pyroxene than in CIR1-3 charges. As melting progresses, pyroxene is consumed and py decreases along with the Wo content in the remaining pyroxene. However, at a given total extent of melting, there is ~ 20 wt.% less pyroxene in CI residues, with a higher Wo content, than in LL residues. The Mg# of olivine and pyroxene increases with decreasing fO2 conditions from 75 at IW -1.3 to 95 at IW -2.5. In parallel, the fraction of Fe metal increases by 12 wt.% (Fig. S1).

119

Experimental charges with CaO-rich bulk compositions that do not correspond to known chondritic residues (DCa1 and DCa2) contain olivine, pigeonite, augite, glass and metal. No attempt was made to estimate the melt fractions in those experiments. They are used in section 4.3 and 4.7, together with other experimental charges, to constrain the temperature of equilibration of ureilites and the thermal history of the UPB. The detailed composition of experimental runs products are reported in the supplementary material (Table S1-S5).

3.2. Mineral composition and petrographic description of ureilites

Among the 12 ureilites analyzed for this study, eight are US Antarctic samples provided by the Meteorite Working Group. Several of them have been described to various extents in the literature: ALHA 82106 (Takeda et al. 1989), EET 96293 (Singletary and Grove 2003), LAP 03721 (paired to LAP 03587; Warren and Rubin, 2010), EET 90019 (Warren 2012). Four additional samples from Northwest Africa were acquired from private collectors and have not been previously described in detail. The phase compositions and the phase proportions are reported in Table 3 and the BSE images of six samples are shown in Fig. 5. Three of the nine ureilites that we studied are orthopyroxene-augite-olivine ureilites (EET 96293, NWA 11754 and ALHA 82106), one is an orthopyroxene-pigeonite-olivine ureilite (NWA 5555) and the remaining eight are pigeonite-olivine ureilites. EET 90019 and NWA 4852 are some of the most FeO-poor pigeonite-olivine ureilites (Fo89.4 and Fo87.7, respectively). We observed a single 80-µm pigeonite in section 15 (0.6 cm2) of DOM 08012. Chromite was only found in MIL 07447 as two 50-µm spherical crystals: one zoned and enclosed in olivine and the other relatively homogenous and enclosed in pigeonite (Fig. S2). In addition to the grain boundary metal and sulfide, spherules of variable size (1-20 µm), made of C-rich metal (possibly cohenite), C-poor metal, sulfides, and frequently all three, occur as inclusions in pigeonite and olivine (Fig. S2). Most samples also display the traditional 10-50 µm rims of MgO-rich olivine containing sub-µm metal blebs and produced by reduction during the disruption of the UPB (i.e. “secondary smelting”, see section 4.8). In four of the eight pigeonite-olivine ureilites (NWA 11755, LAP 03721, MIL 090076 and EET96042), the pyroxene has been extensively smelted (Fig. 5e-f), probably due to a brief episode of shock heating that followed the impact (Warren and Rubin 2010). In those samples, but not in the other four, olivines contain sub-µm metal inclusions in their cores (Fig. S4). Unlike the metal-

120 sulfide spherules mentioned in the previous paragraph, they are homogenously distributed and exclusively sub-µm. NWA 11755 (Fig. 5e) displays particularly clear evidence of this reduction process, including the same bimodal porosity affecting “relic” pigeonite cores and reduced FeO- poor pyroxene as in LAR 04315 (Warren and Rubin 2010). Because the pyroxene was almost completely smelted and is porous, we did not attempt to estimate the pre-impact phase proportions of NWA 11755. NWA 11754 has a large fraction of pyroxene, including abundant augite, and a large fraction of metal that fills fractures and silicate grain boundaries (Fig. 5d). NWA 11754 is in every way identical to Hughes 009 and other members of the “Hughes cluster”, a group of ureilites interpreted as igneous cumulates rather than melting residues (Goodrich et al., 2001, 2009). EET 96293 has the same Fo content in olivine as Hughes 009 and, while it contains less augite than Hughes 009/NWA 11754, it contains a similar fraction of total pyroxene (Fig. 6). ALHA 82106 is a unique orthopyroxene-augite-olivine ureilite that contains wavy exsolution lamellae of augite in orthopyroxene (Fig. 5c). Because ALHA 82106 also contains coarse-grained augite, its silicate phase assemblage at peak temperature is believed to have been augite-pigeonite-olivine (Takeda et al. 1989). The element maps used to calculate the modal composition of ten of the twelve ureilites and eight of the ureilites analyzed by Singletary and Grove (2003) are available in the supplementary material. All but three ureilites form an overall trend of increasing py with Fo in olivine (Fig. 6; R2 = 0.74). Only the two anomalous “Hughes cluster” samples and MET 01083, which displays a poikilitic texture (Fig. S11), were excluded. The trend formed by ureilites in Fig. 6 has the same slope as residues with constant Mg/Si ratios, having lost the same aggregate melt fraction at various fO2. It is also parallel to similar trends formed by ordinary chondrites and other groups of primitive achondrites (Chapter 2).

4. Discussion

4.1. The initial composition of the UPB

Ureilites have long been thought to be related to carbonaceous chondrites due to their high graphite content. In support of this view, Clayton et al. (1976) and Clayton and Mayeda (1988) showed that ureilites form a trend of mass-independent fractionation in three-isotope plots and are coarsely aligned with CV chondrites (Greenwood et al., 2017). However, the discovery of

121 nucleosynthetic anomalies in Cr, Ni and Ti isotopes in bulk ureilites has questioned the connection between ureilites and carbonaceous chondrites (Qin et al. 2010; Warren 2011). On a ∆17O - ε54Cr diagram, ureilites, which have higher concentrations of neutron-rich isotopes (e.g. 54Cr and 50Ti) relative to carbonaceous chondrites, are part of the “non-carbonaceous” (NC) group along with enstatite and ordinary chondrites. Goodrich et al. (2015) inferred that ureilites still formed in the same region of the nebula as carbonaceous chondrites (greater heliocentric distance) but at a time when nucleosynthetic anomalies were distinct (~0.5 Ma after CAIs). However, the NC and CC reservoirs are thought to have been present and isolated from each other very early (<1 Ma after CAIs), presumably by a proto-Jupiter, based on the Mo isotopes of NC and CC iron meteorites (Kruijer et al. 2017). If this is true, and if the UPB accreted relatively late (1.4-1.7 Ma after CAIs) as suggested by Budde et al. (2015), then the UPB would have accreted within the orbit of proto- Jupiter together with the other NC parent bodies. However, the accretion time of the UPB is still debated and van Kooten et al. (2017) argued that the UPB was partly differentiated as early as 0.6 Ma after CAIs. Based on a presumed relationship between Si isotopes and the Mg/Si ratio of parent bodies, Dauphas et al. (2015) calculated a value of 1.00 ± 0.08 for the UPB, but the uncertainty overlaps with both carbonaceous and ordinary chondrites. Therefore, at present, isotopic systems do not provide strong constraints on the initial major element composition of the UPB. In a previous study (Chapter 2), we have shown experimentally that the initial composition of the UPB was characterized by a high NaK# (50). High NaK# chondrites (H, LL, CI) produce pigeonite-olivine residues upon melting, contrary to CM (33) and CV (25) chondrites. Goodrich et al. (2007) suggested that the UPB was initially enriched in CaO relative to Al2O3. We showed that this was not the case, but rather that the UPB had a chondritic CaO/Al2O3 and became depleted in Al2O3 following the extraction of low-degree melts. The assumption that the precursor material of FeO-poor ureilites contained more CAIs than the one of FeO-rich ureilites (Goodrich et al., 2013a; Singletary and Grove, 2006) appear to be no longer justified. Any significant addition of CAIs to FeO-poor ureilites would drive down the NaK# towards the value of CM (33) and CV chondrites (25), which do not form pigeonite-olivine residues by partial melting. The new incremental melting experiments and additional analyses of ureilites can be used to constrain the initial Mg/Si ratio of the UPB in more detail. Warren (2011) argued that, despite being carbon-rich, ureilites could derive from “non-carbonaceous” chondritic materials with Mg/Si ratios resembling ordinary chondrites (0.92-0.95) but lower Fe/Si ratios. Goodrich and

122

Wilson (2013) proposed that the Mg/Si ratio of the UPB could have been even lower than the one of ordinary chondrites and close to the one of enstatite chondrites (0.74-0.88). The olivine and pigeonite of CIR2 experimental charges are identical, both in terms of chemical composition and relative proportions (py), to the olivine and pigeonite in ureilites (Fig. 4 and 6). On the other hand, LLR1-2 experimental charges contain an overabundance of low-Ca pyroxene, with relatively lower Wo contents (4-6.5) compared to ureilites (up to Wo 12). Therefore, the Mg/Si ratio of the LL starting composition (0.92) places a firm lower limit on the initial Mg/Si ratio of the UPB. Any composition with a lower ratio would contain no pigeonite at all (Wo <5). The Mg/Si ratio of the CI composition (1.03) is likely much closer to the average initial ratio of the UPB. The positive correlation between the Fo content in olivine and the py of ureilites displays some scatter that allows for a certain degree of heterogeneity in the initial Mg/Si ratio (i.e. 0.98-1.05). However, the slope of the Fo-py trend, which is identical to experimental trends of constant aggregate melt fraction, suggest that, if present, slight variations in the Mg/Si ratio were independent of the Fo content in olivine. Therefore, ureilites with Fo75 and Fo95 olivine were characterized by the same overall initial Mg/Si ratio. If the initial Fe/Mg of the UPB had been homogenous, high-Fo ureilites, that formed under more reducing conditions (IW -2 / -2.5) should also contain more metal, unless some of the metal was extracted from the residues along with FeS (Barrat et al. 2015), in which case highly siderophile element concentrations (HSE) would be expected to be lower (Warren and Huber 2006;

Warren et al. 2006). Ureilites with Fo85 and Fo95 olivine should contain 6 wt.% and 12 wt.% more 0 Fe than ureilites with Fo75 olivine (Fig. S1). Because neither metal fractions nor siderophile element concentrations correlate with Fo content in olivine, the UPB is often thought of as initially heterogenous in Fe/Mg. However, the siderophile element concentrations of ureilites are still not well understood despite extensive studies. It remains possible that the UPB was heterogenous in siderophile elements or that a fraction or exogenous metal overprinted siderophile element concentrations (Goodrich et al., 2013a).

4.2. Literature estimates of ureilite equilibration temperatures

Estimating the temperature of equilibration (TE) of individual ureilite samples has been a central issue in interpreting their petrogenesis. As most ureilites only contain pigeonite and olivine, two-pyroxene (Lindsley 1983; Lindsley and Andersen 1983; Sack and Ghiorso 1994) and olivine-

123 chromite thermometry (Evans and Frost 1975; Wlotzka 2005) can only be applied to relatively small sub-groups of ureilites. Most two-pyroxene temperatures that have been calculated for orthopyroxene-augite or pigeonite-orthopyroxene ureilites are in the range 1200 – 1270 ºC (Chikami et al. 1997; Hiroshi Takeda 1989; Sinha et al. 1997; Takeda et al. 1989; Weber et al. 2003) while olivine-chromite temperatures are much lower 1040 – 1060 ºC (Goodrich et al. 2014). Singletary and Grove (2003) made the first attempt at estimating the temperature of pigeonite-olivine ureilites by calibrating an empirical thermometer based on the MgO and CaO concentrations of pigeonite. Our batch and incremental melting experiments of chondritic compositions can be used to evaluate its utility. Out of 22 experimental charges containing olivine, pigeonite, metal and melt, only five have a calculated temperature within ± 20 ºC of the experimental temperature. Five experiments have a calculated temperature within ± 50 ºC and the last ten experiments are within ± 100 ºC. The pigeonite thermometer of Singletary and Grove (2003) overestimates the actual temperatures of most of our experiments (Fig. S8). Pyroxene phase diagrams (Lindsley and Andersen 1983; Sack and Ghiorso 1994), previous experiments on basaltic compositions (Grove and Juster 1989) and the experiments of this study show that the minimum temperature at which pigeonite is stable increases with the Mg#. The maximum temperature of pigeonite stability (i.e. melting temperature in simple systems) is also a function of the Mg#: 1409

ºC for a pigeonite En90Wo10 but probably ~1350 ºC for the pigeonite of FeO-rich ureilites

(En74Wo10; Presnall, 1995; Warren and Rubin, 2010). Therefore, if permitted by the bulk composition of the system, pigeonite with given MgO and CaO concentrations (or En and Wo contents) can be stable over a temperature interval of 150-250 ºC. The apparent temperature dependence of MgO and CaO concentrations in pigeonite used by Singletary and Grove (2003) reflected the restricted temperature and compositional range of the melts produced in earlier experiments. The UPB was initially rich in alkalis (Bischoff et al., 2014; Chapter 2; Kita et al.,

2004) but the absence of Na2O in ureilite pyroxenes did not allow Singletary and Grove (2003) to include this important element in the calibration of their single-pyroxene thermometer. Therefore, the thermometer of Singletary and Grove (2003) should be replaced by the thermometer described in the next section.

124

4.3. New thermometer based on the partitioning of Cr between olivine and low-Ca pyroxene

The compatibility of Cr in olivine and LCP strongly decreases with increasing temperature in batch melting experiments of chondrites (Fig. 7 a-b). At 1100 ºC, the Cr2O3 content of olivine

)"*(%&+ LCP liq is less than half of the Cr2O3 content of coexisting LCP. However, the D"# (Cr2O3 /Cr2O3 $%&'(%&+ in wt.%) decreases faster with increasing temperature than the D"# and the Cr2O3 content of olivine becomes nearly identical to that of the coexisting LCP at 1300 ºC. Therefore, with

$%&'(%&+ )"*(%&+ $%&'()"* increasing temperature, the ratio of D"# and D"# (i.e. D"# ) increases with increasing $%&'(%&+ )"*(%&+ $%&'()"* temperature and follows a linear relationship (Fig. 7 c-d). The D"# , D"# and D"# do not seem to be influenced by the fO2 over the range of IW -1.3 and IW -2.5 (Fig. 7c).

$%&'()"* Because the melt does not have to be analyzed to calculate the D"# , we can use all of our experimental charges, including residues with only a few percent of melt, to calibrate a $%&'()"* D"# thermometer. In practice, we use a subset of 60 charges for which Cr2O3 contents were the most accurately determined (Table S6). The 60 experimental charges include all ten starting compositions (Table 1): H, LL and CI (31), LLR1-2 and CIR1-3 (23) and RCa1-2 (6). All experimental charges, whether they contain one LCP (orthopyroxene or pigeonite) or one LCP and augite, visible traces of chromite, and regardless of the experimental fO2 fall on the same linear

$%&'()"* trend with an accuracy of ±15 ºC (Fig. 7 c-d). The D"# also appears to be largely independent of the wollastonite content in LCP (Fig. 7d). Nonetheless, a few experiments with very low Wo content in pyroxene (<3), not observed in ureilites, could be characterized by slightly higher $%&'()"* D"# and were not included in the linear regression. $%&'()"* The independence of the D"# to fO2 is not intuitive because the average valence state 3+ 2+ of Cr decreases with decreasing fO2. Most Cr is present as Cr at QFM +1 (~IW +4.5) but Cr predominates at conditions more reducing than IW -1 (Hanson and Jones 1998; Papike et al. 2005) and Cr0 becomes the dominant oxidation state under IW -5/-6 (Holzheid and O’Neill 1995). Cations with contrasting charge and size such as Cr2+ and Cr3+ would be expected to partition differently in the M1 and M2 octahedral sites of the olivine and pyroxene crystal structures. 3+ 2+ Presumably, this is not the case in olivine, for which the partition coefficients DCr and DCr appear to be identical and largely independent of the fO2 (Papike et al. 2005), as suggested by our experimental results. In pyroxene, Cr3+ should fit easily in the smaller M1 site and can be charged

125 balanced by Al3+ in the tetrahedral site (Tschermak substitution) while Cr2+ would only fit in the $%&'()"* larger M2 site. Therefore, D"# could be expected to decrease with fO2, a behavior that has been shown experimental between IW +1 and IW -1 (Karner et al. 2007) but that should become negligible under IW -1 as most Cr is already in the Cr2+ state, as suggested by our experiments (Fig. S6). The predominance of Cr2+ in ureilites is also suggested by XANES analyses (calculated $%&'()"* valence of 2.0-2.2; Goodrich et al., 2013b; Sutton et al., 2017). In summary, the D"# thermometer can be used at conditions more reducing than IW -1 without any fO2 dependent term, $%&'()"* as the influence of fO2 is undetectable. The D"# thermometer is specifically designed for ureilites but is likely not directly applicable to more oxidizing systems in its current form. $%&'()"* The D"# thermometer can be used to calculate the temperature of equilibration (TE) of all ureilites, regardless of their mineral assemblage, as long as they contain olivine and at least one LCP. It can be directly applied to 75 samples with detailed EPMA analyses of both olivine and LCP (Table S8). Their TE cover a large range of temperature, 1054-1276 ºC, which is identical

$%&'()"* to the range of previous TE estimates (section 4.2). However, the D"# TE offer a new highly reliable representation of the temperature distribution in the UPB and provide important new constraints on the petrogenesis of ureilites.

Ureilites with low Fo content in olivine (Fo75-80) have the most contrasting TE (1054-1276

ºC) while ureilites with high Fo content in olivine (Fo90-95) have more restricted and higher TE of

1180-1250 ºC (Fig. 8). Histograms of the TE of ureilites with either Fo74-85 or Fo87-96 olivine display overlapping modes at 1180-1220 ºC. In other words, FeO-poor and FeO-rich ureilites are characterized by identical average TE. Nevertheless, eight of the nine samples with TE higher than

1240 ºC are FeO-rich. This result is in stark contrast with the positive correlation between TE and olivine Fo contents that was predicted by the pigeonite thermometer of Singletary and Grove (2003). $%&'()"* All ureilites with two pyroxenes show a D"# equilibration temperature that is lower than the minimum temperature of pigeonite stability of Sack and Ghiorso (1994). The few ureilite samples containing euhedral chromite, such as NWA 766, LEW 88774, and NWA 3109, display low olivine-chromite temperatures (1037, 1042 and 1058 ºC; Goodrich et al., 2014) and low

$%&'()"* D"# temperatures (1054, 1085 and 1141 ºC) that correlate with the Cr# (i.e. Cr/(Cr+Al) in mol.%) of chromites: 0.65, 0.71 and 0.76, respectively.

126

4.4. Anomalous low-temperature ureilites

Ureilites containing orthopyroxene and augite, characterized by low TE (open symbols in Fig. 8), display poikilitic textures with large (up to 15 mm) orthopyroxene oikocrysts enclosing rounded olivine and augite crystals: LEW 88774 (Chikami et al. 1997), MET 78008 and Y 74130 (Takeda et al. 1989), RaS 517 (Rosén et al. 2019) and HaH 064 (Weber et al. 2003). Many such samples contain unusually large fractions of pyroxene (e.g. LEW 88774, the “hughes cluster” samples, RaS 517) or, more rarely, unusually little pyroxene (MET 01083). In addition, the Al2O3 concentrations of their LCP are significantly higher (1.3-2.2 wt.%) than in other ureilites (0.3-1.0 wt.%; Fig 9a). The “hughes cluster” samples traditionally contain a large fraction of augite (> 30 vol.%) and their bulk Ca/Mg ratios are super-chondritic (Goodrich et al., 2001, 2009). Due to their large pyroxene fractions and poikilitic textures, some anomalous ureilites have been interpreted as cumulates or older residues that reacted with silicate melts (Downes et al. 2008;

Goodrich et al. 2009). All of the aforementioned samples are characterized by lower TE than the temperature of plagioclase disappearance in chondritic materials (Fig. 8). If they represented simple melting residues quenched at peak temperature following the destruction of the UPB, they would be expected to contain plagioclase. Yet, like all monomict ureilites, they are completely devoid of plagioclase and must reflect a more complex petrogenesis. Anomalous low-temperature ureilites are the only samples that do not plot on the trend of increasing py with increasing Fo content in olivine (Fig. 6b). RaS 517 was recently described as containing primary melts and recording the ureilite anatexis (Rosén et al. 2019). Due to its very low TE (1072 ºC) and high augite fraction, this scenario does not seem viable. Instead, the silicate melt could represent a residual liquid crystallizing in a superficial intrusion or magma conduit. Some brachinite-like meteorites display similar coarse poikilitic textures and are inferred to form by partial crystallization of the melt during melt migration (Goodrich et al. 2017a).

4.5. Temperatures and extents of melting recorded by pigeonite-olivine ureilites

The anomalous samples described in the previous section are in part responsible for the once popular view that ureilites were igneous cumulates (Berkley et al. 1980; Goodrich et al. 1987). However, since the 1990s, pigeonite-olivine ureilites have largely been regarded as melting residues (Scott et al. 1993; Warren and Kallemeyn 1992). If FeO-rich pigeonite-olivine ureilites

(Fo76-82), which represent 70-80 % of all ureilites, are melting residues from a relatively

127 homogeneous chondritic material, the samples that equilibrated at high temperature should have experienced larger extents of melting than the samples that equilibrated at lower temperature. The chemical composition and petrologic features of pigeonite-olivine ureilites should correlate with

TE. Several such correlations are indeed observed. In ureilites with Fo76-82 olivine, the Al2O3 content of LCP decreases linearly with increasing TE from 1 wt.% at 1150 ºC to 0.35 wt.% at 1276

ºC (Fig. 9b). The Wo content in LCP decreases in parallel, albeit with more scatter, from Wo10.5-

12 to Wo4.5-5 (Fig. 9c). Finally, Barrat et al. (2016) have recently analyzed the rare earth element (REE) concentrations of a large number of ureilites. They identified two main patterns: (1) “group B” with steeper LREE profiles and high HREE concentrations, representing MgO-rich ureilites

(Fo89-95), and (2) “group A” with less steep LREE profiles and more variable bulk REE depletions, representing FeO-rich ureilites. The REE concentrations, as well as the Zr concentrations of the ureilites of “group A” (bulk leachates) correlate negatively with TE (Fig. 10).

The ureilites of “group B” (Fo89-95 olivine) could represent the melting residues, produced under more reducing conditions, of a chondritic material with initial NaK# and Mg/Si ratios similar to the chondritic material that produced FeO-rich ureilites (section 4.1). In addition, FeO-poor ureilites were equilibrated over the same range of temperature as FeO-rich ureilites (1180-1220 ºC; Fig. 8). Because the melting temperature of ferro-magnesian silicates increases with the Mg#, FeO-poor ureilites should represent residues that melted to a lower extent than the FeO-rich ureilites equilibrated at the same temperature. A lower bulk degree of melting could explain in part the higher REE concentrations of group B ureilites (Fig. 10).

4.6. Incremental melting and composition of “late-stage” silicate melts

As discussed in section 4.1, a chondritic composition with a high NaK# (50) and a relatively homogenous Mg/Si ratio (0.98-1.05) can produce ureilite-like residues by partial melting

(F=15-24 wt.%) under variable fO2 conditions. In batch melting experiments, pigeonite melts out at 1170-1200 ºC while it persists up to 1280 ºC in incremental melting experiments. The ureilite samples with the highest TE (Goalpara, 1276 ºC, and Dingo Pup Donga, 1267 ºC) have a LCP with low Wo content (4.7-4.6, i.e. orthopyroxene). However, many FeO-rich samples with TE between

1240 and 1260 ºC contain Wo6.5-10 pigeonite: RC 027, Y 74123, NWA 11755, Y 82100 and GRO

95575 (Table S8). Residues containing pigeonite (Wo6.5-10) at 1250 ºC cannot form by batch

128 melting of any chondritic material (Fig. 3) and, instead, suggest that silicate melts were extracted in several steps. The composition of the first silicate melts produced in ureilites have been constrained by batch melting experiments of high NaK# chondritic materials (Chapter 1). Low-degree partial melts (< 15 wt.%), rich in alkali elements, Al2O3 and SiO2 were likely extracted from their source and formed the ALM-A trachyandesite (Bischoff et al. 2014) and the albite/oligoclase-rich clasts of polymict ureilites (Cohen et al., 2004; Kita et al., 2004). If pigeonite-ureilites melted incrementally, the melts that were produced in later melting stages would have been radically different in composition from the melts of batch melting experiments. Because the melt productivity drops significantly following the extraction of alkali-rich melts (~5 wt.% between 1120 and 1250/1280 ºC; Fig. 2), ureilites would have contained very little melt at the time of the destruction of the parent body. Similarly, LLR2 and CIR2-3 experimental charges contain very little melt (2-5 wt.%) that could not be analyzed successfully. These “late-stage” melts are expected to be depleted in alkali elements relative to batch melts. Following the disappearance of plagioclase, Na concentrations decrease rapidly with successive increments of melting (see pMELTS simulations in supplementary material).

Pigeonite-olivine ureilites with Fo75-82 olivine and high TE (1200-1260 ºC) have the highest concentrations of CaO in olivine (0.36-0.44; Fig. 11b). The partition coefficient of CaO between oliv-liq olivine and liquid (DCa = CaOoliv/CaOliq) decreases with increasing temperature (Fig 11a), and can be parametrized to calculate the CaO content of the late-stage melts that were in equilibrium with specific samples (Fig. 11c). At low temperature (<1170 ºC), the CaO content of the melts in equilibrium with ureilites were identical to batch melts of high NaK# chondrites (H, LL and CI). However, at higher temperature, the late-stage melts were increasingly CaO-rich (8-12 wt.%; Fig. 11d) while the CaO content of batch melts plateaus at 7-8 wt.% and then decreases from 1220 ºC. The high CaO content of late-stage melts cannot result from batch melting and supports an origin of ureilites as residues of incremental melting. )"*(%&+ The Al2O3 content of late-stage melts could be estimated based on the D,% . In practice, )"*(%&+ due to the low Al2O3 content of pyroxene in experiments, the D,% is difficult to calculate precisely. In batch melting experiments (Chapter 1), and the experiments of Singletary and Grove )"*(%&+ (2006), the D,% varies between 0.04 and 0.07 at 2-13 MPa for a pigeonite with Wo6-10 but we

129

$-.(%&+ could not identify systematic variations. In Collinet et al. (2015), the D,% strongly increases with pressure but is constant over a 150 ºC temperature interval (i.e. 0.11 at 0.5 GPa, 0.2 at 1.0

GPa and 0.3 at 2.0 GPa). Therefore, we estimate the Al2O3 concentrations of late-stage melts with pig-liq a constant DAl of 0.055 and find that, within the ± 20 wt.% relative uncertainty, they are indistinguishable from the Al2O3 concentrations of batch melts. Our best estimate of the composition of a late-stage melt, produced at 1240 ºC by incremental melting and in equilibrium with olivine Fo79, is reported in Table 4. The MgO content is estimated as a function of the $%&'(%&+ temperature and is used to constrain the FeO content, assuming a K0,23(45 of 0.31, from batch melting experiments. Late-stage melts such as the one reported in Table 4 could easily have crystallized the labradorite-rich clasts observed in polymict ureilites (Cohen et al., 2004; Goodrich et al., 2017b; Kita et al., 2004). The behavior of melting of the UPB has been disputed based on differing interpretations of REE patterns (Goodrich et al., 2007, 2013c; Warren, 2012; Warren and Huber, 2006; Warren and Kallemeyn, 1992). Most recently, Barrat et al. (2016) used REE melting models to argue that ureilites formed by continuous/dynamic melting, a specific case of incremental melting during which a fixed melt fraction, referred to as the critical melt fraction, is retained in the residue. We reproduced their calculations using the modal proportions and melting coefficients of Chapter 1 and find that the REE patterns of FeO-rich ureilites can be explained by 15-24 wt.% of continuous melting with a 2 wt.% critical melt fraction (Fig. 12a), a range that is identical to the aggregate melt fractions of experimental charges containing ureilite-like residues (section 4.1, Fig. 4). In agreement with their more restricted TE, FeO-poor ureilites can be explained by 16-18 wt.% of continuous melting. In detail, REE melting models are non-unique and the REE patterns can be reproduced by different incremental melting processes, such as the complete removal of larger melt increments (e.g. 5-10 wt.%; Fig. S10). On the other hand, near-fractional melting would produce stronger depletions in LREE (Fig. 11c), unless melting was faster than LREE diffusion in pyroxene (Goodrich et al. 2007). Because all ureilites represent melting residues that melted enough to exhaust plagioclase (F > 15.5 wt.%), and because REE melting models produce model- dependent results (Fig. S10), it is impossible to constrain uniquely the size of the melt increments. We infer that they were in the range 2-5 wt.%.

130

4.7. Peak temperatures of ureilites and thermal history of the UPB

Ureilites are sometimes described as having a two-stage cooling history: an initial episode of slow-cooling followed by an episode of rapid-cooling simultaneous with the destruction of the $%&'()"* UPB (e.g. Goodrich, 1992). It could be argued that the TE of ureilites, estimated with the D"# thermometer, does not represent the peak temperature. Cr diffuses relatively fast in olivine and in orthopyroxene. Diffusion coefficients are only one order of magnitude slower than the ones of Fe- Mg interdiffusion (Dohmen et al. 2007; Ganguly et al. 2007; Ito and Ganguly 2006). If main group

26 $%&'()"* ureilites had stopped melting and started cooling following the extinction of Al, the D"# thermometer would have recorded a lower final temperature rather than the peak melting temperature. Only a handful of samples have been previously suggested to record two distinct temperatures of equilibration and, therefore, a slow-cooling event. In the next paragraphs, we use $%&'()"* our experiments and the D"# thermometer to re-evaluate whether some ureilites, and more generally the UPB, went through a stage of slow cooling. As long as pigeonite-olivine ureilites remained in the field of pigeonite stability, a stage of slow-cooling could be undetectable. However, assuming that they did cool slowly, it could not have been over a large temperature interval. Many pigeonite-olivine ureilites with Wo6.5-10 were last equilibrated at 1250 ºC, a temperature which is close to the highest temperature at which we observe pigeonite (Wo5.5-7) in incremental melting experiments (1274 ºC). At higher temperature

(1300 ºC), the pigeonite is replaced by orthopyroxene in experiments. In addition, if the final TE had been significantly lower than the peak temperatures, the correlation of TE with bulk REE concentrations and Al2O3 contents in pyroxene might have been erased. This is not the case. If many ureilites had cooled slowly across the minimum temperature of pigeonite stability, many ureilites would contain inverted pigeonite characterized by a host phase of orthopyroxene and augite lamellae. Only one ureilite displays this texture: ALH 82106 (Takeda et al., 1989; Fig.

5c). This unique sample has a TE of 1245 ±15 ºC (Fig. 8), just under the minimum temperature of pigeonite stability. One of our experiments (CHS 66; RCa1) performed at 1250 ºC contains olivine

(Fo95.3 vs. Fo95.5 in ALH 82106), augite and a pigeonite that is almost identical to the bulk unmixed pyroxene of ALH 82106 (opx+augite lamellae; Wo10.5, Table S7). This experiment suggests that ALH 82106 reached its peak temperature at 1250 ± 10 ºC and only had to cool by a few to 20 ºC to form inverted pigeonite. We posit that if the UPB had cooled slowly by more than 0-30 ºC below its peak temperature, ALH 82106 should not be the only ureilite containing inverted pigeonite.

131

To our knowledge, only one other ureilite displays clear signs of slow cooling: the highly anomalous sample LEW 88774 (Chikami et al. 1997), which contains coarse ~50 µm intergrown lamellae of orthopyroxene and augite. Based on the bulk composition of the pyroxene (Wo15-20; sub-calcic augite), Chikami et al. (1997) argued that the pyroxene progressively recrystallized during slow cooling, starting at 1280 ºC. The model of Sack and Ghiorso (1994), which agrees closely with our experimental results (Fig. S5), suggests that a bulk pyroxene with a composition of En66Wo15 is stable at 1180-1200 ºC as opposed to 1280 ºC (Chikami et al. 1997). LEW 88774

$%&'()"* might have cooled slowly from 1200 ºC to 1100 ºC, the D"# temperature. Because LEW 88774 is a highly unusual pyroxenite (py = 76) it could represent a cumulate formed within an outer layer of the UPB and its thermal history is likely not representative of the conditions recorded by most ureilites in the deeper interior of the UPB. Several other “anomalous low-temperature ureilites” (section 4.4), are thought to have equilibrated at two distinct temperatures. For example, Goodrich et al. (2001) suggest that Hughes 009, containing orthopyroxene and augite (Mg# of 88-89), cooled from 1250 ºC to 1050 ºC. One of our experiments places an upper limit to the peak temperature of Hughes 009. CIR1-CHS 48 contain pigeonite (Mg# 88) at 1194 ºC, which is close to the minimum temperature of pigeonite stability at that Mg# (1190 ºC; Sack and Ghiorso, 1994). Because Hughes 009 contain no exsolution lamellae but “standard” anhedral crystals of orthopyroxene and augite, it probably reached a peak temperature lower than the temperature of pigeonite stability (1190 ºC), while the $%&'()"* final temperature recorded by the D"# thermometer is 1167 ± 15 ºC. Those two temperatures suggest that Hughes 009 cooled, at most, over 30 ºC instead of 200 ºC. Using a similar reasoning, and comparing the ureilite HaH 064 to experiment CI-CHS 54 performed at 1129 ºC (Chapter 1) we infer that its peak temperature was in the range 1125 ºC (Sack and Ghiorso 1994) – 1089 ºC $%&'()"* (D"# TE) rather than 1200 ºC (Weber et al. 2003) and that, therefore, it did not cool significantly. In any case, ALH 82106 is the only sample that has recorded an episode of slow cooling at high temperature and in the region of the UPB representing the residual mantle. It cooled by only ~ 20 ºC. The bulk of evidence suggests that the ureilites that represent melting residues were maintained very close to their peak temperature, and might have been actively melting, at the time of the disruption of the UPB.

132

4.8. Variable fO2 during melting, heterogeneity of the UPB and location of the different mantle reservoirs

The metal and carbon polymorphs present along silicate grain boundaries have occasionally been interpreted as being added by the impactor that disrupted the UPB (Boynton et al. 1976; Day et al. 2017; Wasson et al. 1976). However, the weakly shocked ureilite ALHA 78019 (Berkley and Jones 1982) contains euhedral crystals of graphite intergrown with metal and sulfide. In addition, mixed cohenite-metal-sulfide inclusions occur in the silicates of many ureilites (Goodrich et al., 2013a; Fig. S2) and strongly suggest that graphite and metal were in equilibrium with silicates during partial melting. The presence of metal in equilibrium with olivine places strict constraints on the fO2 of the system (OSI equilibrium; Nitsan, 1974). A composition of Fo75-95 in olivine and Fe90-95Ni10-5 in the metal at 1100-1300 ºC correspond to ΔIW -1.3 to -2.5, which is identical to our experimental range of fO2. In experiments, the fO2 relevant to the formation of ureilites is imposed by the C-CO buffer. Its position relative to the IW buffer is strongly dependent on both the temperature and the pressure. If the temperature is increased or the pressure is decreased, Fe2+ is progressively reduced to Fe0. Because ureilites contain abundant graphite, it has long been assumed that the C-CO buffer imposed the fO2 of equilibration of ureilites, like in our experiments, and was responsible for the variable Fo content of olivine cores (Fo75-95) between samples (e.g. Berkley et al., 1980; Berkley and Jones, 1982). In metallurgy, the process by which Fe metal is produced by heating FeO ores in the presence of C is called “smelting”. By analogy, a primary control of the C-CO buffer on the composition of silicate cores (i.e. Mg#) in ureilites has been called smelting by Walker and Grove (1993). It is now called “primary smelting” or “equilibrium smelting” in the recent ureilite literature, by opposition to “secondary smelting” or “disequilibrium smelting”, which refer to the reduction event associated with the disruption of the parent body. Despite minor semantic disagreement (Warren and Huber 2006), disequilibrium smelting has not been a source of contention. It is widely recognized as having produced the ubiquitous 10-50 µm wide rims of olivine that are increasingly FeO-poor outward and contain small metal blebs and free SiO2 (Fig. 5a), following the pressure drop and rapid cooling at 2-20 ºC/h caused by the destruction of the UPB (Herrin et al. 2010; Miyamoto 1985; Takeda et al. 1989). In most cases, the pyroxene is less affected by this process, probably due to the slower rate of Fe-Mg interdiffusion. However, a few samples display pyroxene smelted to different extents (Fig. 5e-f and Fig. S3) associated with

133 olivine crystals containing homogenously distributed sub-µm metal inclusions (Fig. S4). Such samples are interpreted as briefly shock-heated by the impact before being rapidly cooled (Warren and Rubin 2010). Equilibrium smelting, which would control the composition of olivine cores, has been more controversial. It has been invoked in the context of cumulate and residue models alike and encompasses different processes (Goodrich et al. 1987, 2007; Singletary and Grove 2006; Sinha et al. 1997; Walker and Grove 1993). In Goodrich et al. (1987), equilibrium smelting is envisioned as affecting the parental melts of ureilites as they migrate upwards and decompress before pooling at various depths. In the model of Singletary and Grove (2006), smelting affects partly molten diapirs heated more rapidly than the surrounding mantle due to their higher CAIs contents, and higher 26Al concentrations. In the model of Goodrich et al. (2007), primary smelting is driven by an increase in temperature at constant pressure and happens simultaneously with partial melting. $%&'()"* In detail, simple primary smelting models appear to be inconsistent with the new D"# thermometer. For example, the model of Goodrich et al. (2007) was based on the assumption that ureilites, regardless of their Mg#, reached peak temperatures of 1250-1275 ºC and that FeO-rich ureilites (Fo75-80) melted deeper in the parent body than FeO-poor ureilites, as a consequence of the pressure sensitivity of the C-CO buffer (Goodrich et al., 2013c; Warren, 2012). However, it is now clear that the equilibration temperatures of FeO-rich pigeonite-olivine ureilites span a 150 ºC interval (1130-1280 ºC). The Fo content of olivine in our experiments, which is a function of the temperature and of the CO pressure, is used to calculate the pressure of equilibration of ureilites, assuming that the fO2 is primarly controlled by the C-CO buffer (Fig. 13a and Fig. S9). Because FeO-rich ureilites formed over a large range of temperature, they would also have melted over a large range of pressure (4-14 MPa), which would overlap with the pressure of equilibration of FeO-poor ureilites (2-8 MPa; Fig 13a). If melting occurred at higher pressure (> 15 MPa), where graphite can be stable with olivine of various compositions (Fo74-95) over the same temperature interval (1130-1280 ºC), the Fo $%&'()"* content in olivine and the TE based on the D"# would not be sufficient to constrain the pressure of ureilite equilibration (Walker and Grove 1993; Warren and Huber 2006; Warren and Kallemeyn 1992). Nevertheless, the temperature interval over which pigeonite-olivine ureilites equilibrated (1130-1280 ºC) could represent a radial temperature gradient within the UPB. The slope of such a temperature profile is unconstrained as it would be influenced by many parameters

134

(e.g. Neumann et al., 2012). In any case, the identical equilibration temperatures of FeO-poor and FeO-rich ureilites could indicate that they melted at the same depth (Fig. 13b).

The origin of the intrinsic fO2 of meteorite parent bodies is an active area of research. The formation of FeO in the nebula setting is non-trivial and could require high dust-to-gas ratios and dust rich in H2O (e.g. Fedkin and Grossman, 2016). The correlation between Mg# of silicates and ∆17O in ureilites is observed in various other meteorite groups and could be connected to the distribution of ice and/or water vapor, either in the accretion disk itself or within parent bodies (e.g. Tenner et al., 2015; Sanders et al., 2017). The FeO-poor and FeO-rich reservoirs of the UPB appear to have melted at similar pressure-temperature conditions but at distinct intrinsic fO2, regardless of the pressure at which melting occurred and of the precise influence of graphite, CO and other gas species. Therefore, despite being isotopically heterogeneous (Barrat et al. 2017; Clayton and Mayeda 1988; Greenwood et al. 2017), the UPB was likely not stratified.

5. Conclusions

The large majority of ureilites represent the residual mantle of a chondritic planetesimal from which 15-24 wt.% of silicate melts were extracted. The UPB was heterogenous in isotope compositions and intrinsic fO2 (IW -1.3/-2.5) but relatively homogenous in most major elements (e.g. NaK#, Mg/Si ratio). All anomalous ureilites with poikilitic textures and large fractions of pyroxene, sometimes interpreted as cumulates, formed at relatively low temperature (1060-1170 ºC) and experiments confirm that they do not represent simple melting residues. One of them, LEW 88774 cooled slowly over 100 ºC (section 4.7). Anomalous low-temperature ureilites probably formed at shallower depths than pigeonite-olivine ureilites. Pigeonite-olivine ureilites represent the large majority of ureilites and were equilibrated at higher temperature (1150-1280 ºC) and are melting residues. They melted incrementally and experienced several events of melt extraction (2-5 wt.%). Over this interval of temperature, the melt productivity was low (5 to 9 wt.%) as a result of the prior extraction of melts rich in SiO2, Al2O3 and alkali elements at lower temperatures (<1120 ºC). Because the temperature of last equilibration of ureilites correlates with petrological and chemical features representative of the total extent of melting (Wo and Al2O3 contents of pyroxene and bulk REE concentrations), melting in the UPB was ongoing or had just stopped at the time of its disruption. Despite being heterogenous, the parent body was not stratified

135 in terms of the Mg# of silicates, intrinsic fO2, or oxygen isotopes as ureilites with different Mg# equilibrated at the same mean temperature (1200 ºC).

6. References

Barrat J. A., Rouxel O., Wang K., Moynier F., Yamaguchi A., Bischoff A., and Langlade J. 2015. Early stages of core segregation recorded by Fe isotopes in an asteroidal mantle. Earth and Planetary Science Letters 419:93–100. Barrat J.-A., Jambon A., Yamaguchi A., Bischoff A., Rouget M.-L., and Liorzou C. 2016. Partial melting of a C-rich asteroid: Lithophile trace elements in ureilites. Geochimica et Cosmochimica Acta 194:163–178. Barrat J.-A., Sansjofre P., Yamaguchi A., Greenwood R. C., and Gillet P. 2017. Carbon isotopic variation in ureilites: Evidence for an early, volatile-rich Inner Solar System. Earth and Planetary Science Letters 478:143–149. Berkley J. L., Taylor G. J., Keil K., Harlow G. E., and Prinz M. 1980. The nature and origin of ureilites. Geochimica et Cosmochimica Acta 44:1579–1597. http://www.sciencedirect.com/science/article/pii/0016703780901192. Berkley J. L., and Jones J. H. 1982. Primary igneous carbon in ureilites: Petrological implications. Journal of Geophysical Research: Solid Earth 87:A353–A364. http://dx.doi.org/10.1029/JB087iS01p0A353. Bischoff A., Horstmann M., Barrat J.-A., Chaussidon M., Pack A., Herwartz D., Ward D., Vollmer C., and Decker S. 2014. Trachyandesitic volcanism in the early Solar System. Proceedings of the National Academy of Sciences of the United States of America 111:12689–92. http://www.ncbi.nlm.nih.gov/pubmed/25136108 (Accessed December 22, 2017). Boynton W. V, Starzyk P. M., and Schmitt R. A. 1976. Chemical evidence for the genesis of the ureilites, the achondrite Chassigny and the nakhlites. Geochimica et Cosmochimica Acta 40:1439–1447. Broadley M. W., Bekaert D. V., Marty B., Yamaguchi A., and Barrat J. A. 2019. Noble Gas Variations in Ureilites Demonstrate Heterogenous Volatile Distribution in the Early Solar System. LPI Contributions 2157:6469. Budde G., Kruijer T. S., Fischer-Gödde M., Irving A. J., and Kleine T. 2015. Planetesimal differentiation revealed by the Hf–W systematics of ureilites. Earth and Planetary Science Letters 430:316–325. Chikami J., Mikouchi T., Takeda H., and Miyamoto M. 1997. Mineralogy and cooling history of the calcium-aluminum-chromium enriched ureilite, Lewis Cliff 88774. Meteoritics & Planetary Science 32:343–348. Clayton R. N., Mayeda T. K., Onuma N., and Shearer J. 1976. Oxygen isotopic composition of minerals in the Kenna ureilite. Geochimica et Cosmochimica Acta 40:1475–1476. Clayton R. N., and Mayeda T. K. 1988. Formation of ureilites by nebular processes. Geochimica et Cosmochimica Acta 52:1313–1318. http://www.sciencedirect.com/science/article/pii/0016703788902864. Cohen B. A., Goodrich C. A., and Keil K. 2004. Feldspathic clast populations in polymict ureilites: Stalking the missing basalts from the ureilite parent body. Geochimica et Cosmochimica Acta 68:4249–4266. Collinet M., Médard E., Charlier B., Vander Auwera J., and Grove T. L. 2015. Melting of the primitive martian mantle at 0.5–2.2 GPa and the origin of basalts and alkaline rocks on Mars. Earth and Planetary Science Letters 427:83–94. Dauphas N., Poitrasson F., Burkhardt C., Kobayashi H., and Kurosawa K. 2015. Planetary and meteoritic Mg/Si and variations inherited from solar nebula chemistry. Earth and Planetary Science Letters 427:236–248. Day J. M. D., Corder C. A., Cartigny P., Steele A., Assayag N., Rumble Iii D., and Taylor L. A. 2017. A

136

carbon-rich region in Miller Range 091004 and implications for ureilite petrogenesis. Geochimica et Cosmochimica Acta 198:379–395. Dohmen R., Becker H.-W., and Chakraborty S. 2007. Fe–Mg diffusion in olivine I: experimental determination between 700 and 1,200°C as a function of composition, crystal orientation and oxygen fugacity. Physics and Chemistry of Minerals 34:389–407. http://dx.doi.org/10.1007/s00269- 007-0157-7. Downes H., Mittlefehldt D. W., Kita N. T., and Valley J. W. 2008. Evidence from polymict ureilite meteorites for a disrupted and re-accreted single ureilite parent asteroid gardened by several distinct impactors. Geochimica et Cosmochimica Acta 72:4825–4844. Evans B. W., and Frost B. R. 1975. Chrome-spinel in progressive metamorphism—a preliminary analysis. Geochimica et Cosmochimica Acta 39:959–972. Fedkin A. V., and Grossman L. 2016. Effects of dust enrichment on oxygen fugacity of cosmic gases. Meteoritics & Planetary Science 51:843–850. French B. M., and Eugster H. P. 1965. Experimental control of oxygen fugacities by graphite-gas equilibriums. Journal of Geophysical Research 70:1529–1539. Ganguly J., Ito M., and Zhang X. 2007. Cr diffusion in orthopyroxene: Experimental determination, 53Mn–53Cr thermochronology, and planetary applications. Geochimica et Cosmochimica Acta 71:3915–3925. Ghiorso M. S., Hirschmann M. M., and Reiners P. W. 2002. The pMELTS: A revision of MELTS for improved calculation of phase relations and major element partitioning related to partial melting of the mantle to 3 GPa. Goodrich C. A., Jones J. H., and Berkley J. L. 1987. Origin and evolution of the ureilite parent magmas: Multi-stage igneous activity on a large parent body. Geochimica et Cosmochimica Acta 51:2255– 2273. Goodrich C. A. 1992. Ureilites - A critical review. Meteoritics 27. Goodrich C. A., Fioretti A. M., Tribaudino M., and Molin G. 2001. Primary trapped melt inclusions in olivine in the olivine-augite-orthopyroxene ureilite Hughes 009. Geochimica et Cosmochimica Acta 65:621–652. Goodrich C. A., Van Orman J. A., and Wilson L. 2007. Fractional melting and smelting on the ureilite parent body. Geochimica et Cosmochimica Acta 71:2876–2895. Goodrich C. A., Fioretti A. M., and Van Orman J. 2009. Petrogenesis of augite-bearing ureilites Hughes 009 and FRO 90054/93008 inferred from melt inclusions in olivine, augite and orthopyroxene. Geochimica et Cosmochimica Acta 73:3055–3076. Goodrich C. A., and Wilson L. 2013. Non-Basaltic Magmatism On the Ureilite Parent Body. abtract, Workshop on Planetseimal Formation and Differentiation #8018. Goodrich C. A., Ash R. D., Van Orman J. A., Domanik K., and McDonough W. F. 2013a. Metallic phases and siderophile elements in main group ureilites: Implications for ureilite petrogenesis. Geochimica et Cosmochimica Acta 112:340–373. Goodrich C. A., Sutton S. R., Wirick S., and Jercinovic M. J. 2013b. Chromium valences in ureilite olivine and implications for ureilite petrogenesis. Geochimica et Cosmochimica Acta 122:280–305. Goodrich C. A., Wilson L., Van Orman J. A., and Michel P. 2013c. Comment on “Parent body depth- pressure-temperature relationships and the style of the ureilite anatexis” by P. H. Warren (MAPS 47:209-227). Meteoritics and Planetary Science 48:1096–1106. Goodrich C. A., Harlow G. E., Van Orman J. A., Sutton S. R., Jercinovic M. J., and Mikouchi T. 2014. Petrology of chromite in ureilites: Deconvolution of primary oxidation states and secondary reduction processes. Geochimica et Cosmochimica Acta 135:126–169. Goodrich C. A., Hartmann W. K., O’Brien D. P., Weidenschilling S. J., Wilson L., Michel P., and Jutzi M. 2015. Origin and history of ureilitic material in the solar system: The view from asteroid 2008 TC3 and the Almahata Sitta meteorite. Meteoritics & Planetary Science 50:782–809. http://dx.doi.org/10.1111/maps.12401. Goodrich C. A., Kita N. T., Sutton S. R., Wirick S., and Gross J. 2017a. The Miller Range 090340 and

137

090206 meteorites: Identification of new brachinite-like achondrites with implications for the diversity and petrogenesis of the brachinite clan. Meteoritics & Planetary Science 52:949–978. Goodrich C. A., Treiman A. H., and Boyle S. 2017b. Melt formation and evolution on the ureilite parent body, as shown by feldspathic clasts in polymict ureilites. Lunar and Planetary Science 1196. Greenwood R. C., Burbine T. H., Miller M. F., and Franchi I. A. 2017. Melting and differentiation of early-formed asteroids: The perspective from high precision oxygen isotope studies. Chemie der Erde - Geochemistry 77:1–43. Grove T. L., and Bence A. E. 1977. Experimental study of pyroxene-liquid interaction in quartz- normative basalt 15597. Lunar and Planetary Science Conference Proceedings 8:1549–1579. Grove T. L., and Juster T. C. 1989. Experimental investigations of low-Ca pyroxene stability and olivine- pyroxene-liquid equilibria at 1-atm in natural basaltic and andesitic liquids. Contributions to Mineralogy and Petrology 103:287–305. http://dx.doi.org/10.1007/BF00402916. Hanson B., and Jones J. H. 1998. The systematics of Cr3+and Cr2+partitioning between olivine and liquid in the presence of spinel. American Mineralogist 83:669–684. Herrin J. S., Zolensky M. E., Ito M., Le L., Mittlefehldt D. W., Jenniskens P., Ross A. J., and Shaddad M. H. 2010. Thermal and fragmentation history of ureilitic asteroids: Insights from the Almahata Sitta fall. Meteoritics & Planetary Science 45:1789–1803. http://dx.doi.org/10.1111/j.1945- 5100.2010.01136.x. Holzheid A., and O’Neill H. S. C. 1995. The Cr2O3 oxygen buffer and the free energy of formation of Cr2O3 from high-temperature electrochemical measurements. Geochimica et Cosmochimica Acta 59:475–479. Ito M., and Ganguly J. 2006. Diffusion kinetics of Cr in olivine and 53Mn–53Cr thermochronology of early solar system objects. Geochimica et Cosmochimica Acta 70:799–809. Karner J. M., Papike J. J., Sutton S. R., Shearer C. K., McKay G., Le L., and Burger P. 2007. Valence state partitioning of Cr between pyroxene-melt: Effects of pyroxene and melt composition and direct determination of Cr valence states by XANES. Application to Martian basalt QUE 94201 composition. American Mineralogist 92:2002–2005. Kita N. T., Ikeda Y., Togashi S., Liu Y., Morishita Y., and Weisberg M. K. 2004. Origin of ureilites inferred from a SIMS oxygen isotopic and trace element study of clasts in the Dar al Gani 319 polymict ureilite. Geochimica et Cosmochimica Acta 68:4213–4235. van Kooten E. M. M. E., Schiller M., and Bizzarro M. 2017. Magnesium and chromium isotope evidence for initial melting by radioactive decay of 26Al and late stage impact-melting of the ureilite parent body. Geochimica et Cosmochimica Acta 208:1–23. Kruijer T. S., Burkhardt C., Budde G., and Kleine T. 2017. Age of Jupiter inferred from the distinct genetics and formation times of meteorites. Proceedings of the National Academy of Sciences of the United States of America 114:6712–6716. Lindsley D. H., and Andersen D. J. 1983. A two-pyroxene thermometer. Journal of Geophysical Research 88:A887. Lindsley D. H. 1983. Pyroxene thermometry. American Mineralogist 68:477–493. Miyamoto M. 1985. Cooling history of some antarctic ureilites. Proceding of the sixteen lunar and planetary science conference 90:D116–D122. http://adsabs.harvard.edu/full/1985LPSC...16..116M. Neumann W., Breuer D., and Spohn T. 2012. Differentiation and core formation in accreting planetesimals. Astronomy & Astrophysics 543:A141. Nitsan U. 1974. Stability field of olivine with respect to oxidation and reduction. Journal of Geophysical Research 79:706–711. http://dx.doi.org/10.1029/JB079i005p00706. Papike J. J., Karner J. M., and Shearer C. K. 2005. Comparative planetary mineralogy: Valence state partitioning of Cr, Fe, Ti, and V among crystallographic sites in olivine, pyroxene, and spinel from planetary basalts. American Mineralogist 90:277–290. Presnall D. C. 1995. Phase Diagrams of Earth-Forming Minerals. American Geophysical Union (AGU). pp. 248–268 Qin L., Rumble D., Alexander C. M. O. D., Carlson R. W., Jenniskens P., and Shaddad M. H. 2010. The

138

chromium isotopic composition of Almahata Sitta. Meteoritics and Planetary Science 45:1771– 1777. Rosén Å. V., Pape J., Hofmann B. A., Gnos E., and Guillong M. 2019. Quenched primary melt in Ramlat as Sahmah 517 – Snapshot of ureilite anatexis in the early solar system. Geochimica et Cosmochimica Acta 246:1–20. Sack R., and Ghiorso M. 1994. Thermodynamics of multicomponent pyroxenes: II. Phase relations in the quadrilateral. Contributions to Mineralogy and Petrology 116:287–300. http://dx.doi.org/10.1007/BF00306498. Sanders I. S., Scott E. R. D., and Delaney J. S. 2017. Origin of mass-independent oxygen isotope variation among ureilites: Clues from chondrites and primitive achondrites. Meteoritics & Planetary Science 52:690–708. Scott E. R. D., Taylor G. J., and Keil K. 1993. Origin of ureilite meteorites and implications for planetary accretion. Geophysical Research Letters 20:415–418. http://dx.doi.org/10.1029/93GL00474. Singletary S. J., and Grove T. L. 2003. Early petrologic processes on the ureilite parent body. Meteoritics & Planetary Science 38:95–108. Singletary S., and Grove T. L. 2006. Experimental constraints on ureilite petrogenesis. Geochimica et Cosmochimica Acta 70:1291–1308. http://www.sciencedirect.com/science/article/pii/S0016703705009130. Sinha S. K., Sack R. O., and Lipschutz M. E. 1997. Ureilite meteorites: Equilibration temperatures and smelting reactions. Geochimica et Cosmochimica Acta 61:4235–4242. Sutton S. R., Goodrich C. A., and Wirick S. 2017. Titanium, vanadium and chromium valences in silicates of ungrouped achondrite NWA 7325 and ureilite Y-791538 record highly-reduced origins. Geochimica et Cosmochimica Acta 204:313–330. Takeda H., Mori H., and Ogata H. 1989. Mineralogy of augite-bearing ureilites and the origin of their chemical trends. Meteoritics 24:73–81. http://dx.doi.org/10.1111/j.1945-5100.1989.tb00947.x. Tenner T. J., Nakashima D., Ushikubo T., Kita N. T., and Weisberg M. K. 2015. Oxygen isotope ratios of FeO-poor chondrules in CR3 chondrites: Influence of dust enrichment and H2O during chondrule formation. Geochimica et Cosmochimica Acta 148:228–250. Walker D., and Grove T. 1993. Ureilite smelting. Meteoritics 28:629–636. Warren P. H., and Kallemeyn G. W. 1992. Explosive volcanism and the graphite-oxygen fugacity buffer on the parent asteroid(s) of the ureilite meteorites. Icarus 100:110–126. Warren P. H., Ulff-Møller F., Huber H., and Kallemeyn G. W. 2006. Siderophile geochemistry of ureilites: A record of early stages of planetesimal core formation. Geochimica et Cosmochimica Acta 70:2104–2126. https://www.sciencedirect.com/science/article/pii/S0016703706000469 (Accessed November 27, 2018). Warren P. H., and Huber H. 2006. Ureilite petrogenesis: A limited role for smelting during anatexis and catastrophic disruption. Meteoritics & Planetary Science 41:835–849. Warren P. H., and Rubin A. E. 2010. Pyroxene-selective impact smelting in ureilites. Geochimica et Cosmochimica Acta 74:5109–5133. http://www.sciencedirect.com/science/article/pii/S0016703710003042. Warren P. H. 2011. Stable isotopes and the noncarbonaceous derivation of ureilites, in common with nearly all differentiated planetary materials. Geochimica et Cosmochimica Acta 75:6912–6926. Warren P. H. 2012. Parent body depth–pressure–temperature relationships and the style of the ureilite anatexis. Meteoritics & Planetary Science 47:209–227. http://dx.doi.org/10.1111/j.1945- 5100.2011.01320.x. Wasson J. T., Chou C.-L., Bild R. W., and Baedecker P. A. 1976. Classification of and elemental fractionation among ureilites. Geochimica et Cosmochimica Acta 40:1449–1458. Weber I., Bischoff A., and Weber D. 2003. TEM investigations on the monomict ureilites Jalanash and Hammadah al Hamra 064. Meteoritics & Planetary Science 38:145–156. Wlotzka F. 2005. Cr spinel and chromite as petrogenetic indicators in ordinary chondrites: Equilibration temperatures of petrologic types 3.7 to 6. Meteoritics & Planetary Science 40:1673–1702.

139

Zhu K., Moynier F., Schiller M., Wielandt D., Larsen K. K., van Kooten E. M. M. E., and Bizzarro M. 2019. Chromium isotopic constraints on the origin the ureilite parent body. 82nd Annual Meeting of The Meteoritical Society 2019 2019:6323.

140

Figures and tables a) CHS 63_2 pig b) CHS 46_2 1202 ºC 1191 ºC CIR2 LLR1

oliv

pig 10 µm 10 µm

Figure 1. Backscattered images of experimental charges containing ureilite-like residues (olivine-pigeonite-metal). (a) CI residue with py of 21. (b) LL residue with py of 39.5. py = px/(px+oliv)*100 in wt.%.

40 plag out 35 lting

30 batch me

10 CI CIR1

aggregate melt fraction (wt.%) 5 CIR2

0 1050 1100 1150 1200 1250 1300 Temperature ¼C

Figure 2. Aggregate melt fraction as a function of the experimental temperature in batch and incremental melting experiments (IW -1.4/-1.7). The melt productivity is low following the extraction of 15 wt.% melt (plag-out temperature).

141

95 plag out 12 plag out H 90 LL Fo content in olivine 10 LLR1 85 LLR2 8 80

6 75 CI 4 Wo content in LCP CIR1 70 CIR2 2 CIR3 65

0 60 1100 1150 1200 1250 1300 1100 1150 1200 1250 1300 Temperature ¼C Temperature ¼C

Figure 3. Composition of low Ca pyroxene (LCP) as a function of the temperature in batch melting experiments (shaded areas; Chapter 1) and incremental melting experiments. (a) ordinary chondritic compositions. (b) CI chondritic compositions.

142

aggregate melt fraction

10 15 20 25

(a) (b) 12 H LL 10 LLR1 LLR2 8

Wo content in LPC in content Wo 4

0.7 (c) (d)

CI 0.6 CIR1 0.5 CIR2 CIR3 0.4 15 0 0.3 2

0.2

= px/(px+oliv) in wt.% = px/(px+oliv) py 15 0.1 20

0 70 75 80 85 90 95 70 75 80 85 90 95 Fo content in olivine Fo content in olivine

Figure 4. Composition (a-b) and fraction of pyroxene (c-d) as a function of the Fo content in olivine in batch and incremental melting experiments of ordinary (a-c) and CI (b-d) chondritic compositions. The Fo content in olivine and py increase with decreasing fO2. From Fo75 (IW -

1.3) to Fo95 (IW -2.5), ~12 wt.% Fe metal is produced (Figure S1). Wo content and py decrease with increasing aggregate melt fraction (colored bar). The lines represent the modal and chemical compositions of residues for a given degree of melting (Chapter 2).

143 a) EET 90019 b) MIL 07447

pig metal + C oliv pig

oliv

100 µm 100 µm c) ALHA 82106 d) NWA 11754 opx aug oliv

aug opx

oliv

1 mm 100 µm e) NWA 11755 f) MIL 090076 pig pig FeO-poor px

FeO-poor px

oliv

100 µm 100 µm

Figure 5. Backscattered images of a subset of the ureilites analyzed for this study. See text for detail. All samples display reduction rims in olivine. Pyroxene is almost completely “smelted” in NWA 11755 (e). It displays a coarse porosity in the FeO-poor area and a fine porosity in the FeO-rich area. MIL 090076 displays a similar texture but the pyroxene cores are homogeneous and free of porosity (f). NWA 11754 is characteristic of the “Hughes cluster” (d).

144

15 pig a) oliv pig-opx

exp. oliv

C hughes 10 cluster ALHA aug-opx 82106 oliv

xp.

L L aug+opx anom. 5 Wo content in LCP

this study Havero SG03 + this study

0 other studies b) LEW 88774 “Hughes cluster” 0.7

0.6 QUE 93336 0.5 7349

0.4 ave this study p. ex 15 LL 0.3 MET 20 01083

py = px/(px+oliv) in wt.% 0.2 15 ALHA xp. 0.1 I e 77257 C 20

70 75 80 85 90 95 Fo content in olivine

Figure 6. Composition (Wo content) and fraction of pyroxene (py) as a function of the Fo content in olivine in ureilites. All X-ray maps used to calculate the modal compositions are available in the supplementary material. Leaving out the Hughes cluster samples and the anomalous sample MET 01083, ureilites form of similar trend of increasing py with Fo (dotted line, equation R2=0.74). The solid lines (b) and shaded areas (a) represent the experimental trends of Fig. 4: CI residues (light gray) and LL residues (dark gray).

145

1300 (a) (b)

1200

Exp. T (¼C) 1100

0 0.5 1 1.5 012 oliv-liq LCP-liq log (DCr ) log (DCr ) -3 = -6.51×10 -3 T + 8.77 = -9.54×10 T +12.46 -1 (c) 1300

1250 -1.5 log f O 2 - ∆ IW

1200

1150 -2

Experimental temperature (¼C) 1100

1050 -2.5

12 (d) 1300 oliv-LCP T (¼C) = 913.9 + 465.6 D Cr R2 = 0.97 10 1250 Wo conetnt in LCP

excluded from 8 linear regression 1200 Wo < 3

6 1150 4

Experimental temperature (¼C) 1100 2

1050 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 oliv-LCP D Cr

$%&'()"* Figure 7. Calibration of the D"# thermometer. (a-b) different temperature dependence of the Cr partition coefficients between olivine (a) or low Ca pyroxene (b) and silicate melt in batch melting experiments of a CI composition (fO2: IW -1.3/-2.2; Chapter 1). (c-d) Linear relationship $%&'()"* between the D"# and the experimental temperature of 60 charges with various bulk compositions (supplementary material and see text for detail). The linear relationship is used as a simple mineral thermometer that is independent of the fO2 (within the experimental range) and the Wo content. 146

30 20 10 Fo74-85 Fo87-96 1300 94 SG 1260 22 21 20 1220 19

Cr 18 CM L 1180 I-H-L 17 10 C 12-15 16 ureilite types: 1140 9 g-out batch pig-oliv pla 11 pig-opx-oliv 6 8 aug-opx-oliv 1100 7 5 4 Hughes cluster 2 3 other anom.

Temperature based on D oliv-lcp (¼C) 1060 1 70 75 80 85 90 95 10 20 1020 10 20 Forsterite content in olivine

Figure 8. Equilibration temperature (TE) of 75 ureilites as a function of the Fo content in olivine.

All pigeonite-olivine ureilites have a higher TE (within ±15 ºC uncertainty) than the minimum temperature of pigeonite stability of Sack and Ghiorso (1994). Two-pyroxene ureilites have a lower TE than pigeonite-olivine ureilites at a given Fo content in olivine (or Mg# in pyroxene, see supplementary material). FeO-rich ureilites (e.g. Fo77-80) cover a large range of temperature (210 ºC). The temperature at which plagioclase melts out during batch melting of CI, H, LL and

CM chondrites is shown for reference (green lines). Ureilites with Fo74-85 and Fo87-96 olivine are equilibrated at similar temperatures (mode at 1180-1220 ºC). Open symbols are considered “anomalous” in the sense that they do not represent simple melting residues. Key samples: [1]NWA 766 [2]MET 78008 [3]Y 74130 [4]RaS 517 [5-6]LEW 88774 [7]HaH 064 [8]NWA 7349 [9]LAP 02382 [10]Havero [11]EET 96293 [12]FRO 90054 [13]EET 96314 [14]NWA 11754 [15] Hughes 009 [16]MIL 091004 [17]MET01083 [18]NWA 5555 [19-20]ALHA 82130 [21]ALHA 84136 [22]ALHA 82106. See text and supplementary material for references.

147

2.5 (a) plag-out 2 batch CI-H-LL ureilite types: pig-oliv 2 3 pig-opx-oliv 1 aug-opx-oliv 4 Hughes cluster 7 1.5 anomalous 6 5 (b) all experiments except RCa1/2 1 concentration in LCP (wt.%) 3

O 0.5 2 Al

1050 1100 1150 1200 1250 1300 1350

Temperature D Cr oliv-LCP (¼C) (b) (c) 1.2 Fo 76.3-82.3 Fo 76.3-82.3 12 1 10 0.8

in LCP (wt.%) 0.6 3 O 2

R2=0.92 Wo content in LCP 6 R2=0.47

Al 0.4 (0.33)

1100 1200 1300 1100 1200 1300 Temperature (¼C) Temperature (¼C)

Figure 9. (a) Concentration of Al2O3 in low-Ca pyroxene (orthopyroxene or pigeonite) as a

$%&'()"* function of the D"# temperature (ureilites) or the experimental temperature (+). See caption of Fig. 8 for sample numbers. (b) Same as (a) but only keeping FeO-rich pigeonite-olivine

$%&'()"* ureilites. (c) Wo content in LCP as a function of the D"# temperature of FeO-rich pigeonite-olivine ureilites (same samples as in (b). The R2 value in parentheses characterizes the linear regression without the three open symbols.

148

a) b) 1280 0 5 10 MET 0108 1260 EET 83225 1240 -1 10 1220 Y 791538 ALH 82130 1200

10-2 1180 1160

-3 LAR 4315 10 Temperature ¼C (D Cr) 1140 LAP 3587 Y 790981 REE Concentration (CI norm) 1120

-4 1100 10 -3 -2 -1 La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu 10 10 10 Zr concentration (CI norm) c) Barrat et al. 2016 d) 1280 Group A ureilites 1280 Group B ureilites 1260 1260 smelted px ureilites 1240 (Warren and Rubin 1240

Cr 2010) 1220 1220

1200 1200

1180 1180

1160 1160 1140

Temperature ¼C (D ) 1140

1120 1120

1100 1100 10-4 10-3 10-2 10-1 0 0.1 0.2 0.3 0.4 0.5 0.6 La concentration (CI norm) Dy concentration (CI norm)

Figure 10. (a) Rare earth element concentrations of ureilites from Barrat et al. (2016). We excluded one sample from group A (LAP 03587) and one from group B (LAR 04315) as their pyroxene is thought to have been shock-melted (Warren and Rubin, 2010), a process that probably lowered the bulk concentration of incompatible elements. (b-d) Bulk incompatible

$%&'()"* element concentrations (leachates) as a function of the D"# temperature. Incompatible element concentrations of Group A samples correlate negatively with TE. Yamato 790981 (open symbol) might represent an anomalous low-temperature sample.

149

CaO concentration in oliv (wt.%)

0.25 0.3 0.35 0.4 (a) (b) 1300 1300 CI, H, LL batch exp. 1260 1260

1220 1220 oliv-LCP (¼C)

1180 Cr 1180 MET 01083 (poikilitic) 1140 1140 “hughes cluster”

1100 1100 “px-rich and T based on D Experimental temperature (¼C) poikilitic ureilites 1060 1060 ureilites” 0.02 0.04 0.06 0.08 0.1 70 75 80 85 90 95 100 DCa oliv-liq Forsterite content in olivine

(c) (d) 14 16 exp. H calculated CI pMELTS co n batch LL ureilite melts t. 1 12 melts CI Fo 76-84 14 % CM frame in (b) CV 10 12 con t. 3 %

8 10 ch t ba ba tc V h C batch 6 M 8 C h c t

ba CaO concentration (wt.%) CaO concentration (wt.%)

,CI 4 L 6 L H,

1050 1100 1150 1200 1250 1300 1350 1050 1100 1150 1200 1250 Temperature (¼C) Temperature (¼C)

Figure 11. (a) Temperature dependence of Ca partition coefficients between olivine and liquid in CI, H and LL batch melting experiments (Chapter 1). (b) Concentration of CaO in olivine as a function of the Fo content and the temperature (colored bar). (c) Estimated concentration of CaO in the “late-stage melts” that were last equilibrated with FeO-rich ureilites and comparison with batch melts (IW -1.3/-1.7) of H, LL, CI, CM and CV chondrites at the same temperature. The “late-stage melts” of ureilites show a CaO-enrichment trend that is characteristic of incremental melting. (d) pMELTS simulation (CI, IW -1.5) illustrating how instantaneous melts, produced by incremental melting (e.g. continuous melting, 1 and 3 wt.% critical melt fraction) become CaO- rich with increasing temperature relative to batch melts. Batch melt compositions and melting temperatures are not consistent with experiments in detail.

150

(a) 0 F= 7 10 ut plag o

0.5 -1 6.3 + 10 F= 1

F = 16.3 10-2

F= 24.2 + 0.5 10-3 REE concentrations (CI norm.) (CI REE concentrations Continuous melting .5 Φ = 0.02 24 = F -4 10 (b)

out plag

.3 -1 0.1 15 10 3 + F = 15. F= F = 18.3

10-2

-3

REE concentrations (CI norm.) (CI REE concentrations 10 Continuous melting Φ = 0.02

-4 10 (c) F = 10 t ou plag 10-1

F=15.2 10-2

-3

REE concentrations (CI norm.) (CI REE concentrations 9 10 . 9 . Fractional melting 18

= (equilibrium)

= F F 10-4 La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu Figure 12. Rare earth element modeling using the data and the same overall approach as Barrat et al. (2016) but with the melting coefficients and phase proportions of Chapter 1. The blue field represents the “Group A” and the red field the “Group B” see Figure 10. Following Barrat et al. (2016), merrillite is stable in the residue and melts out simultaneously with plagioclase. (a) dynamic melting at IW -1.5 reproducing the REE patterns of FeO-rich ureilites, with a 0.02 critical melt fraction until the last increment, where either all the melt is removed (black lines) or 0.015 (75 wt.% of the final melt fraction) is removed (blue lines). (b) same as (a) but with a pyroxene-rich starting composition (IW -2) and 0.001 of the final melt fraction (5 wt.%) retained. (c) fractional melting. REE melting model results are non-unique, see supplementary material.

151

Forsterite content in olivine

60 70 80 90 1300 a) CO pressure control

1260

1220 Cr

1180

ureilite types: pig-oliv 1140 Temperature D oliv-LCP (¼C) pig-opx-oliv aug-opx-oliv

1100 051015 Pressure CO (MPa)

1300 b) temperature control (arbitrary profile) 1260

1220 Cr

1180

ureilite types: pig-oliv 1140 Temperature D oliv-LCP (¼C) pig-opx-oliv aug-opx-oliv

1100 >15 Minimum pressure (MPa) Figure 13. (a) Pressure of equilibration of ureilites assuming that the fO2 is primarily controlled by the C-CO buffer. The pressure of equilibration (PC-CO) is fixed by the Fo content of olivine of $%&'()"* individual samples and their D"# temperature of equilibration (TE; Fig. S9). (b) If the actual pressure of equilibration of all ureilites is larger than 15 MPa, the PC-CO of (a) is no longer relevant and the fO2 is not controlled by the C-CO buffer. The intrinsic fO2 of ureilites is variable

(IW -1.3/-2.5) and is higher than the fO2 of the C-CO buffer. Graphite is stable and there are no precise constraints on the pressure of equilibration (PE). The various TE of ureilites can still be interpreted as representing a “geotherm” (a temperature profile). Although the slope of the geotherm is unconstrained, ureilites with different Fo content in olivine but identical TE could be assumed to have equilibrated at the same pressure.

152

conductively cooled lid melt erupted and/or crystallized in intrusions ? anomalous low-T ureilites mantle residue, FeO-rich (IW -1.3) ureilites mantle residue, FeO-poor (IW -2.5) ureilites FeS (+ FeNi-C core) ? Figure 14. Possible origin of different ureilites in the UPB (boxes 3 to 5). The large diversity of ureilites in terms of TE and extent of melting could be interpreted as a vertical distribution of temperature in the UPB. Ureilites equilibrated at higher temperature could have originated from deeper in the UPB. If this is correct, because the TE of ureilites with Fo75-85 and Fo90-95 olivine overlap (Fig. 13), the UPB would be heterogenous but not stratified. The anomalous low-T ureilites could represent cumulates or derive from melt-rock reactions in magma conduits.

153

Table 1: summary of experimental conditions exp # P (CO) T int log fO2 Δ IW t (h) ∆ T* starting compositions per experiment MPa ºC (CCO) ºC 1 2 3 4 5 CHS 46 6.2 1191 -13.7 -1.61 72 20 LLR1 H CHS 47 6.6 1200 -13.6 -1.63 72 0 CIR1 LLR1 LLR2 CHS 48 4.1 1194 -14.0 -1.98 72 0 CIR1 LLR1 H CHS 49 10.0 1248 -13.0 -1.62 24 0 CIR1 CIR3 LLR1 LLR2 CHS 50 7.2 1248 -13.3 -1.89 27 0 CIR1 CIR2a CIR3 CHS 51 8.1 1217 -13.3 -1.58 68 0 CIR2a CIR1 CIR3 LLR1 LLR2 CHS 53 12.1 1248 -12.8 -1.47 46 0 CIR2a CIR1 CIR3 CHS 63 9.1 1202 -13.3 -1.38 72 0 CIR2b CI H CHS 65 3.1 1201 -14.2 -2.27 52 0 CIR2b CIR1 RCa CI H CHS 66 3.4 1250 -13.9 -2.53 42 0 CIR1 CIR2b RCa CI CHS 67 13.3 1274 -12.6 -1.59 20 0 CIR2b CIR1 CI CHS 68 5.2 1159 -12.5 -1.53 100 0 CIR2b RCa2 RCa CI CHS 69 5.9 1301 -13.2 -2.44 9 0 RCa RCa2 CIR2b H CI CHS 70 3.8 1205 -14.2 -1.84 96 0 RCa2 RCa CI

154 CHS 71 7.9 1216 -13.3 -1.59 90 58 LLR1 CIR1 CIR2b CI

fO2 calculated based on the CCO buffer (extended formulation, French and Eugter, 1965) and expressed relative to

the IW buffer (Huebner, 1971) *T (ºC) of initial 2 h step - final T (ºC); 0 = isothermal Experimental charges with starting chondric compositions (italic) are described in Chapter 1

154

Table 2. Experimental starting compositions and bulk ureilite compositions CI* CIR1 CIR2a CIR2b CIR3 PCA 82506 North Haig H* LL* LLR1 LLR2 DPD RCa1 RCa2 F= 11 15.5 15.5 19.5 mmict (pmict) 8.5 16 mmict n.a. n.a. SiO2 40.8 39.7 37.1 38.1 38.3 42.0 40.3 37.6 43.5 41.6 40.3 41.8 37.8 41.5 TiO2 0.13 0.09 0.07 0.05 0.05 0.04 0.09 0.12 0.12 0.09 0.05 0.09 0.07 0.06 Al2O3 2.86 1.15 0.65 0.37 0.38 0.10 0.20 2.22 2.44 1.20 0.39 0.37 0.85 1.26 Cr2O3 0.69 0.60 0.79 0.60 0.62 0.82 0.73 0.51 0.59 0.63 0.65 0.81 0.65 0.88 FeO 21.8 22.4 24.6 22.7 23.3 19.2 18.4 30.8 21.9 23.5 24.5 21.2 21.5 19.3 MnO 0.44 0.36 0.36 0.36 0.37 0.41 0.44 0.31 0.36 0.38 0.39 0.36 0.34 0.30 MgO 28.2 32.1 32.4 34.2 34.1 36.3 37.9 23.8 26.8 28.8 30.7 33.2 33.9 30.4 CaO 2.30 1.87 1.99 2.00 1.34 1.00 1.38 1.78 2.11 1.96 1.41 1.54 3.58 5.14 Na2O 1.20 0.54 0.26 0.24 0.25 0.03 0.09 0.89 0.99 0.56 0.26 0.07 0.23 0.21 K2O 0.12 0.03 0.00 0.01 0.01 0.01 0.03 0.10 0.10 0.03 0.01 0.03 0.05 0.08 P2O5 0.39 0.20 0.32 0.19 0.19 0.04 0.23 0.26 0.22 0.21 0.20 0.21 0.18 0.16 NiO 1.13 1.04 1.31 1.10 1.12 0.10 0.15 1.48 1.01 1.09 1.18 0.32 1.04 0.93

Mg/Si 1.03 1.21 1.30 1.34 1.33 1.29 1.40 0.94 0.92 1.03 1.14 1.18 1.33 1.09 (Ca/Mg)CI 1.01 0.72 0.76 0.73 0.49 0.34 0.45 0.93 0.98 0.84 0.57 0.58 1.31 2.10 (Al/Mg)CI 1.00 0.35 0.20 0.11 0.11 0.03 0.05 0.92 0.90 0.41 0.13 0.11 0.25 0.41 155 Mg# 69.7 71.9 70.1 72.9 72.3 77.1 78.6 57.9 68.6 68.7 69.1 73.6 73.8 73.8

*Average chondritic starting compositions of Lodders and Fegley, renormalized without FeS or volatile elements; starting compositions of batch melting experiments (Chapter 1) F= estimated melt fraction extracted from residues relative to chondritic composition CI and LL, respectively mmict and pmcit = bulk composition of monomict and polymict ureilites from Jarosewich (2006), DPD: Dingo Pup Donga

155

Table 3. Mineralogy of ureilites analyzed for this study py px2/pxtot oliv px1 (LCP) px2 (aug or opx) chromite Fo Cr2O3 CaO Wo Cr2O3 Al2O3 Wo Cr2O3 Al2O3 Fe# Cr# Antartic LAP 03721 9.1 75.1 0.61 0.32 9.3 1.22 0.92 EET 90019 36.7 89.4 0.64 0.34 9.0 1.04 0.81 ALHA 82106 49.5 19.0 95.5 0.57 0.28 4.8 0.80 0.46 36.1 0.87 0.83 MIL 090076 29.0 78.8 0.72 0.35 10.6 1.36 0.93 EET 96293 57.0 3.5 87.1 0.48 0.29 4.8 1.07 1.22 37.2 1.33 1.87 DOM 08012 77.6 0.71 0.36 7.1 1.08 0.51 EET 96042 15.1 81.5 0.84 0.36 9.1 1.37 0.73 MIL 07447 25.6 82.2 0.71 0.34 10.5 1.25 0.89 0.48 0.77

Northwest Africa NWA 5555 39.0 21.1 90.9 0.64 0.31 8.7 1.03 0.55 4.8 0.95 0.57 NWA 4852 41.8 87.5 0.64 0.31 8.4 1.04 0.75 NWA 11754 60.5 33.6 87.5 0.51 0.28 4.9 0.98 1.23 36.5 1.19 1.91

156 NWA 11755 78.3 0.73 0.36 8.2 1.01 0.44 py is the ratio (pyroxene total)/(pyroxene total + olivine)*100, wt.% of percent relative to olivine;

calculated from modal composition (X-ray maps) and phase densities px2/pxtot is the ratio px2/(px1+px2) for full EPMA analyses and analytical uncertaities, see Table S2

156

Table 4. Average composition of melt extracted 1 2 1104 ºC 1240 ºC

SiO2 62.44 55.7 TiO2 0.52 0.3 ±0.1 Al2O3 15.99 9 ±2 Cr2O3 0.09 0.65 ±0.15 FeO 5.05 12.1 ±1.5 MnO 0.17 0.55 ±0.1 MgO 3.09 8 ±1 CaO 4.13 11.5 ±1 Na2O 6.78 2 ±1 K2O 0.82 0.05 ±0.05 P2O5 0.86 0.1 ±0.1 1 Aggregate melt extracted immediately prior to plagioclase exhaustion (Chapter 1) 2 Average composition of "late-stage" melt see text for detail

157

Supplementary material

15 LL H CI 10

H CI

metal fraction (wt.%) 5 LL CIR1 LLR1 CIR2 LLR2 CIR3 0 70 75 80 85 90 95 Fo content in olivine Figure S1. Metal fraction in experiments as a function of the forsterite content in olivine.

158

(a) (b) MIL 07447, 10 MIL 07447, 10

oliv pig

(a) Cr-rich spinel Cr-rich spinel

10 µm 10 µm (c) (d) MIL 07447, 10 DOM 08012, 15 sulfde FeNi

C-rich iron sulfde FeNi

pig oliv 10 µm 10 µm (e) (f) EET 96042, 60

C-rich iron cohenite? FeNi / C-rich iron? sulfde FeNi pig sulfde sulfde 10 µm

pig 10 µm

Figure S2. (a) zoned chromite inclusion in olivine with metal blebs surrounding the outer Mg- rich rims, indicative of “disequilibrium smelting”. (b) a nearly identical chromite inclusion but in pigeonite instead of olivine is not zoned/smelted. The compositions at the centers of both chromite inclusions are identical. (c and e) Composite C-rich metal (cohenite?), FeNi C-poor metal and sulfide spherules (see (Goodrich et al., 2013)). (d) inclusions of immiscible sulfide and metal. Presumably both phases were liquid, which is consistent with TE of DOM 08012 (1220 ºC, 70 ºC above the FeC eutectic). (f) smaller inclusions in MIL 07447. Even the trails of µm to sub-µm inclusions in pigeonite can contain FeNi metal and sulfide.

159

NWA 11755 EET 96042 pig

glass smelted px glass smelted px SiO2 SiO2

Figure S3. Additional BSE images of disequilibrium pyroxene smelting. Similar textures affecting the rims of homogenous pyroxene crystals can be observed in MIL 090076 and LAP 03721. Also see (Warren and Rubin, 2010).

Figure S4. (next page) sub-µm metal inclusions in the olivine cores of ureilites that suffered pyroxene smelting (EET 96042, LAP 03721, MIL 090076, NWA 11755), whether pyroxene smelting is limited to the rims or is extensive (NWA 11755). The last two pictures represent the homogeneous cores of two samples that do not display pyroxene smelting (MIL 07447 and DOM 08012). Also see Figure 5 of the main manuscript.

160

EET 96042 NWA 11755

LAP 03721 MIL 090076

MIL 07447 DOM 08012

Figure S4. (caption on previous page)

161

Wo a) 1250 ¼C b) 1200 ¼C

Di Hd 1000 1000 ¼C ¼C

1100 ¼C 1100 ¼C

En Fs 1200 ¼C 1200 ¼C

c) 1160 ¼C d) 1130-1070 ¼C

2-px experiments 1000 1000 (aug-opx / aug-pig) ¼C ¼C

1100 ¼C 1100 ¼ pig-oliv experiments C 12 00 ¼C 1200 ¼C aug-opx-oliv ureilites pig-oliv ureilites

Temperature (¼C) 1300 1250 1200 1150 1100 1050

Figure S5. Composition of pyroxene in experiments and ureilites compared to the phase diagram of (Sack and Ghiorso, 1994). The shaded areas represent the pyroxene assemblage in all other primitive achondrites (lodranites, brachinites, acapulcoites), which, contrary to ureilites cooled oliv-LCP slowly. The color code represents the experimental temperature or the DCr temperature of equilibration (for ureilites).

162

7 te augi Karner et al. 2007 6 1170 ºC

px-liq Cr 5

D ite geon 4 pi

3 batch melting opx H-LL-CI 2 1190-1205 ºC 1 -2.5 -2 -1.5 -1 -0.5 0 0.5 1 oxygen fugacity ( IW)

Figure S6. Oxygen fugacity dependence of the partitioning coefficients of Cr between pyroxene and silicate melt extrapolated from Karner et al. (2007) experiments at 1170 ºC. Our experiments

(Chapter 1), do not show a clear fO2 dependence at more reducing conditions.

163

1300

1260 22 21 20 1220 19 18 CM -LL 1180 CI-H 17 10 16 12-15 ureilite types: 1140

Temperature (¼C) 9 pig-oliv -out batch plag 11 pig-opx-oliv 1100 6 aug-opx-oliv 7 5 Hughes cluster 4 3 2 other anom. 1060 1 70 75 80 85 90 95 100 Mg# in LCP Figure S7. Equilibration temperature (TE; based on the partitioning of Cr) of 75 ureilites as a function of the Mg# in low-Ca pyroxene. The black line represents the minimum temperature of pigeonite stability from (Sack and Ghiorso, 1994)(compare with Figure 8 in the main manuscript). Key samples: [1]NWA 766 [2]MET 78008 [3]Y 74130 [4]RaS 517 [5-6]LEW 88774 [7]HaH 064 [8]NWA 7349 [9]LAP 02382 [10]Havero [11]EET 96293 [12]FRO 90054 [13]EET 96314 [14]NWA 11754 [15] Hughes 009 [16]MIL 091004 [17]MET01083 [18]NWA 5555 [19-20]ALHA 82130 [21]ALHA 84136 [22]ALHA 82106.

164

1300 100

1250 Mg # in pyroxene 90

1200 85

-20 ºC ºC 0 2 + 80 C ºC º 50 00 1150 + 1 + 75 experimental temperature (¼C)

1100 70 1100 1150 1200 1250 1300 1350 SG03 calculated temperature (¼C) Figure S8. Offset between calculated temperature, using the thermometer of (Singletary and Grove, 2003), and the experimental temperature of pigeonite-olivine-liquid-metal experiments of this study. Offsets are large and generally overestimate the temperature (by up to 100 ºC) with no systematics. The pigeonite temperature of (Singletary and Grove, 2003) was calibrated in Na2O- free melts, unlike the partial melts shown to be in equilibrium with ureilites. This thermometer oliv-LCP should be replaced with the DCr thermometer calibrated in this study (see main manuscript).

165

Forsterite content in olivine

60 70 80 90 1300 a) experiments

1260

1220

1180 Exp. Temperature (¼C)

1140

1100 051015 Exp. Pressure (MPa) 1300 b) ureilites

1260

1220 Cr

1180

ureilite types: pig-oliv 1140 Temperature D oliv-LCP (¼C) pig-opx-oliv aug-opx-oliv

1100 051015 Pressure CCO (MPa)

Figure S9. (a) Relationship between experimental pressure and temperature and the forsterite content in olivine of experiments. A subset of the experiments (CI composition) are used to calculate a linear regression of the pressure as a function of the temperature and Fo content in olivine: P (MPa) = -62.3 – 0.02993 Fo + 0.07828 T (ºC). (b) Pressure of equilibration of ureilites assuming that the fO2 is controlled by the CCO buffer (primary smelting). The pressure is oliv-LCP calculated from the temperature of equilibration (DCr ) and the Fo content of olivine in ureilites. 166

Incremental melting: 2 increments Incremental melting: 6 increments melt extracted at plagioclase exhaustion (F= 15) melt extracted at F= 5, 10, 15, 16, 20 with or without merrillite with merrilite i2 F=5 0 0 10 F = 10 10 i3 F=10 plag out t 15 ou ag 4 F= -1 pl i 10 10-1 F= 16 + 0.5

6 1 -2 = 10 10-2 F .3 0 20 2 = = F= 22.5 + 0 F F -3 -3 i5 .5 2 10 10 2 = 5 30 F 2 = = F F i6 -4 10 10-4 La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu

Dynamic/Continuous melting Dynamic/Continuous melting Φ = 0.02 Φ = 0.02 with merrillite without merrillite

0 F= 7 0 10 out 10 plag F= 7 ut g o pla .5 -1 0.5 -1 + 0 10 16.3 + 10 15.6 F= F=

-2 F = 16.3 10 10-2 .3 F = 15.6 4.8 + 0 0.5 F= 2 24.2 + -3 F= 10 10-3 8 .5 . 24 = = 24 F F -4 10 10-4 La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Li La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu Figure S10. Additional REE models. Blue field represent the “group A” ureilites and red field represent the “group B” ureilites from Barrat et al. (2016). Continuous melting equations are from (Zou, 1998).

167

Incremental melting: 2 increments Incremental melting: 2 increments melt extracted at plagioclase exhaustion (F= 15) melt extracted at plagioclase exhaustion (F= 15) with merrillite without merrillite

0 10 F=10 100

F=14 F=10 out -1 lag -1 g out 10 p 10 pla

-2 10 10-2 F = 20 F = 20

F = 30 F = 30 -3 10 10-3

-4 10 10-4 La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu Incremental melting: 6 increments Incremental melting: 6 increments melt extracted at F= 5, 10, 15, 16, 20 melt extracted at F= 5, 10, 15, 16, 20 with merrilite without merrilite i2 F= 5 0 10 i3 F= 10 100 F = 5 out plag 10 i4 = -1 -1 F 10 10 0 15 = 2 F F = i5 20 -2 F = 10 10-2

5 2 = -3 F -3 10 i6 10 5 2 = F

-4 10 10-4 La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu Dynamic/Continuous melting Dynamic/Continuous melting Φ = 0.02 Φ = 0.02 with merrillite without merrillite

0 100 10 out plag

.1 5.3 -1 + 0 1 -1 ut 10 .3 F = 10 o 15 ag = .3 pl F 18 F =

-2 -2 3 15. 10 10 = .3 F 8 1

= F -3 10-3 10

-4 10-4 10 La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu La Ca Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu Fractional melting with merrillite Fractional melting without merrillite

0 100 F = 10 10 t g ou pla -1 -1 10 10 0 1 = F 15.2 = F ut -2 10-2 10 lag o p

9 -3 -3 2 . 9 . . 10 9 10 . 9 5 . 18 1 5 18 15 1 = =

= F = F =

F F F -4 10-4 10 La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu Figure S10. (continued)

168

Figure S11. RGB (FeMgCa) image of anomalous (poikilitic) sample MET 01083 Blue=augite, dark green=opx, light green=olivine. pMELTS simulations (1): melting of a LL chondrite at IW -1 The melting temperatures, temperatures of phase disappearance, the composition of batch melts and the composition of mineral phases calculated with pMELTS deviate significantly from experimental results. This is not surprising as pMELTS is not optimized to simulate the melting of chondritic material at low pressure. Nevertheless, pMELTS simulations are valuable for visualizing some of the relative differences between batch and fractional melting. For near- fractional melting, the melt is continuously extracted from the residue while maintaining a 0.2 wt.% critical melt fraction (i.e. retained melt fraction). The composition of cumulates and aggregate melts at a given total melt fraction are identical (except for K2O contents) for the different melting behaviors (right column). Following the exhaustion of plagioclase (F = 17 wt.%), the melt productivity essentially drops to zero in near-fractional melting. Note that the concentration of alkali elements in instantaneous near-fractional melts becomes negligible after the exhaustion of plagioclase and the concentration of Al2O3 decreases as well. At the same time, the CaO contents of instantaneous melts increase sharply (right column). However, due to the low melt productivity, this transition is stretched over ~100 ºC where very little melt is actually produced (left column). The equivalence of melting residues between the different melting behavior shows that increasing fractions of batch melts can be extracted from a chondritic

169 composition to simulate incremental melting, which is how we calculate CIR1-3 and LLR1-2 compositions (Table 2 of the main manuscript). Following the exhaustion of plagioclase and the extraction of low-degree melts, ureilites essentially stopped melting, not just because they became depleted in 26Al, but also because the melt productivity became negligible over a significant temperature interval and pigeonite remained stable. The few percent of “late-stage melts” of ureilites, produced after the exhaustion of plagioclase were enriched in CaO, depleted in alkali elements and probably slightly depleted in Al2O3.

50 batch 40 incremental aggregate 30 fract. aggregate 20 fract. instantaneous

10 melt fraction (wt.%) 0 1100 1200 1300 temperature (¼C) 0.12 0.12

0.1 0.1

0.08 0.08

0.06 0.06

0.04 0.04

0.02 0.02 K2O content of residue (wt.%) K2O content of residue (wt.%)

0 0 1100 1150 1200 1250 1300 05101520 Temperature (¼C) melt fraction (wt.%)

2 2

1.5 1.5

1 1

0.5 0.5 K2O content of melt (wt.%) K2O content of melt (wt.%)

0 0 1100 1150 1200 1250 1300 05101520 Temperature (¼C) melt fraction (wt.%)

170

50 batch 40 incremental aggregate 30 fract. aggregate 20 fract. instantaneous

10 melt fraction (wt.%) 0 1100 1200 1300 temperature (¼C)

1 1

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2 Na2O content of residue (wt.%) Na2O content of residue (wt.%) 0 0 1100 1150 1200 1250 1300 05101520 Temperature (¼C) melt fraction (wt.%)

8 8

6 6

4 4

2 2 Na2O content of melt (wt.%) Na2O content of melt (wt.%)

0 0 1100 1150 1200 1250 1300 05101520 Temperature (¼C) melt fraction (wt.%)

171

50 batch 40 incremental aggregate 30 fract. aggregate 20 fract. instantaneous

10 melt fraction (wt.%) 0 1100 1200 1300 temperature (¼C) 2.5 2.5

2 2

1.5 1.5

1 1

0.5 0.5 Al2O3 content of residue (wt.%) Al2O3 content of residue (wt.%) 0 0 1100 1150 1200 1250 1300 05101520 Temperature (¼C) melt fraction (wt.%) 14 14

12 12 10

8 10

6 8 4 6 2 Al2O3 content of melt (wt.%) Al2O3 content of melt (wt.%)

0 4 1100 1150 1200 1250 1300 05101520 Temperature (¼C) melt fraction (wt.%)

172

50 batch 40 incremental aggregate 30 fract. aggregate 20 fract. instantaneous

10 melt fraction (wt.%) 0 1100 1200 1300 temperature (¼C) 2.5 2.2

2 2 1.8

1.5 1.6

1 1.4 1.2 0.5 1 CaO content of residue (wt.%) CaO content of residue (wt.%) 0 0.8 1100 1150 1200 1250 1300 05101520 Temperature (¼C) melt fraction (wt.%)

10 9.5

8 8.5

6 7.5

4 6.5 CaO content of melt (wt.%) CaO content of melt (wt.%) 6

2 5.5 1100 1150 1200 1250 1300 05101520 Temperature (¼C) melt fraction (wt.%)

173 pMELTS simulations (2): melting of a CI chondrite at IW -1.5

40 batch melt 30 fractional melting (aggregate melt) 20 fractional melting (instantaneous melt) 10 continuous melting (instantaneous melt) melt fraction (wt.%) 0 1100 1200 1300 1400 temperature (¼C) 6 6

5 5

4 4

3 3

2 2 Na2O wt.% in liquid Na2O wt.% in liquid 1 1

0 0 1100 1150 1200 1250 1300 05101520 Temperature (¼C) melt fraction (wt.%) 14 14

12 12

10 10

8 8 CaO wt.% in liquid CaO wt.% in liquid

6 6

4 4 1100 1150 1200 1250 1300 05101520 Temperature (¼C) melt fraction (wt.%)

174

Table S1 modal composition of experiments comp. exp # T (ºC) Δ IW total px oliv liq agg. liq metal py Fo Wo px1 Wo px2 LLR1 CHS 46 1191 -1.61 31.5 48.4 8.0 15.8 9.0 39.5 78.7 6.3 LLR1 CHS 47 1200 -1.63 35.7 42.1 9.9 17.5 11.0 45.9 81.3 5.5 LLR1 CHS 48 1194 -1.98 42.1 32.6 7.5 15.3 16.2 56.4 88.4 5.7 LLR1 CHS 49 1248 -1.62 34.8 39.1 12.1 19.6 12.6 47.1 83.9 4.0 LLR1 CHS 51 1217 -1.58 33.3 43.7 10.4 18.0 11.1 43.2 80.9 4.7 LLR1 CHS 71 1216 -1.59 34.8 41.0 10.7 18.4 12.1 45.9 82.9 4.1

LLR2 CHS 47 1200 -1.63 37.3 46.4 2.8 18.0 12.1 44.6 81.6 5.2 LLR2 CHS 49 1248 -1.62 37.0 43.1 4.6 19.4 14.0 46.2 84.2 4.1 LLR2 CHS 51 1217 -1.58 33.5 50.1 3.3 18.4 11.6 40.1 80.6 5.3

CIR1 CHS 47 1200 -1.63 16.8 62.7 10.6 21.4 8.5 19.1 81.3 7.2 CIR1 CHS 48 1194 -1.98 27.4 49.0 9.3 19.6 14.5 33.8 88.4 6.7 CIR1 CHS 49 1248 -1.62 14.0 62.1 12.4 22.9 9.2 16.4 83.9 4.0 CIR1 CHS 50 1248 -1.89 21.9 51.2 11.2 21.9 14.2 28.0 88.4 5.0 CIR1 CHS 51 1217 -1.58 15.0 64.2 10.6 21.4 8.7 16.9 81.2 6.4 CIR1 CHS 53 1248 -1.47 19.8 59.1 10.5 20.9 8.4 20.1 81.0 4.3 CIR1 CHS 65 1201 -2.27 31.3 41.8 9.5 20.1 17.4 40.8 92.6 6.5 CIR1 CHS 66 1250 -2.53 33.4 37.3 9.5 20.2 18.9 45.3 94.6 4.7 CIR1 CHS 67 1274 -1.59 21.0 52.1 13.5 24.0 11.8 26.8 85.4 6.8 CIR1 CHS 71 1216 -1.59 18.1 59.2 11.1 21.4 10.1 21.5 83.4 5.1

CIR2a CHS 50 1248 -1.89 22.6 57.1 3.8 19.2 14.8 28.4 88.9 9.5 CIR2a CHS 51 1217 -1.58 17.8 69.5 2.3 17.9 8.5 20.4 81.0 10.4 CIR2a CHS 53 1248 -1.47 16.5 70.6 3.7 19.1 7.4 19.0 80.1 10.3 CIR2a CHS 53 1248 -1.47 17.8 68.4 3.9 19.3 8.1 20.6 80.9 9.8 CIR2b CHS 63 1202 -1.38 17.5 65.4 3.5 19.0 10.0 21.1 80.0 11.5 CIR2b CHS 65 1201 -2.27 30.2 43.0 3.0 18.5 19.8 41.2 92.8 7.9 34.6 CIR2b CHS 66 1250 -2.53 30.0 41.1 5.0 19.9 21.3 42.2 95.0 7.0 CIR2b CHS 67 1274 -1.59 17.3 57.6 8.6 23.4 13.4 23.1 84.5 5.2 CIR2b CHS 68 1159 -1.53 16.5 67.1 2.5 18.0 10.2 19.7 79.9 8.9 CIR2b CHS 69 1301 -2.44 21.1 45.3 9.8 24.4 20.5 31.8 94.5 5.0 CIR2b CHS 71 1216 -1.59 17.2 61.6 5.5 20.3 12.8 21.8 83.8 8.1

CIR3 CHS 49 1248 -1.62 19.9 63.0 4.6 22.8 11.0 21.0 83.6 5.6 CIR3 CHS 50 1248 -1.89 23.9 55.8 3.8 22.1 14.7 27.0 88.3 5.6 CIR3 CHS 51 1217 -1.58 16.6 69.0 3.5 21.8 9.0 16.4 81.3 6.8 CIR3 CHS 53 1248 -1.47 17.9 67.6 4.2 22.4 8.9 17.9 81.1 5.6

175

Table S2 olivine composition in experiments T comp. exp # (ºC) Δ IW SiO2 Cr2O3 FeO MnO MgO CaO P2O5 NiO total n LLR1 CHS 46 1191 -1.61 38.7 0.1 0.56 0.01 19.5 0.2 0.32 0.02 40.5 0.1 0.31 0.01 0.06 0.02 0.08 0.07 101.0 6 LLR1 CHS 47 1200 -1.63 38.5 0.1 0.55 0.02 17.4 0.1 0.37 0.02 42.6 0.2 0.30 0.01 0.06 0.02 0.13 0.03 101.5 5 LLR1 CHS 48 1194 -1.98 40.4 0.5 0.54 0.01 11.1 0.2 0.37 0.03 47.2 0.3 0.28 0.04 0.04 0.02 0.06 0.04 101.2 5 LLR1 CHS 49 1248 -1.62 39.0 0.3 0.56 0.02 15.2 0.2 0.37 0.05 44.4 0.2 0.30 0.01 0.05 0.04 0.08 0.06 101.7 4 LLR1 CHS 51 1217 -1.58 39.0 0.2 0.57 0.02 17.7 0.4 0.36 0.03 41.9 0.4 0.31 0.01 0.00 0.00 0.11 0.09 101.3 5 LLR1 CHS 71 1216 -1.59 38.9 0.9 0.57 0.03 16.1 0.1 0.40 0.04 43.7 0.3 0.31 0.01 0.00 0.00 0.00 0.00 101.4 6

LLR2 CHS 47 1200 -1.63 38.8 0.2 0.59 0.02 17.1 0.2 0.44 0.02 42.7 0.2 0.30 0.01 0.02 0.01 0.01 0.02 101.3 7 LLR2 CHS 49 1248 -1.62 39.1 0.2 0.61 0.03 14.8 0.1 0.46 0.04 44.5 0.2 0.30 0.02 0.03 0.02 0.08 0.05 101.0 6 LLR2 CHS 51 1217 -1.58 39.5 0.6 0.60 0.02 17.8 0.3 0.44 0.02 41.3 0.9 0.29 0.02 0.00 0.00 0.01 0.03 100.9 6

CIR1 CHS 47 1200 -1.63 38.7 0.1 0.56 0.01 17.4 0.1 0.32 0.02 42.5 0.1 0.32 0.01 0.03 0.02 0.06 0.07 101.7 6 CIR1 CHS 48 1194 -1.98 40.1 0.2 0.53 0.01 11.1 0.1 0.35 0.02 47.4 0.2 0.27 0.01 0.03 0.01 0.10 0.07 100.9 5 CIR1 CHS 49 1248 -1.62 39.0 0.1 0.56 0.02 15.2 0.1 0.37 0.03 44.4 0.2 0.30 0.01 0.05 0.03 0.08 0.06 101.7 4 CIR1 CHS 50 1248 -1.89 40.7 0.1 0.52 0.02 11.0 0.1 0.34 0.03 47.1 0.1 0.29 0.01 0.03 0.02 0.00 0.00 100.2 4 CIR1 CHS 51 1217 -1.58 39.1 0.1 0.59 0.03 17.4 0.1 0.32 0.02 42.2 0.1 0.33 0.01 0.00 0.00 0.06 0.05 101.4 5 CIR1 CHS 53 1248 -1.47 37.8 0.6 0.61 0.04 17.9 0.1 0.34 0.04 43.0 0.2 0.31 0.01 0.00 0.00 0.02 0.04 100.7 4 176 CIR1 CHS 65 1201 -2.27 40.4 0.1 0.48 0.03 7.29 0.1 0.33 0.03 51.1 0.2 0.27 0.01 0.00 0.00 0.04 0.02 100.6 6

CIR1 CHS 66 1250 -2.53 40.4 0.3 0.55 0.02 5.35 0.1 0.37 0.04 53.0 0.2 0.28 0.02 0.00 0.00 0.04 0.01 100.5 7 CIR1 CHS 67 1274 -1.59 38.0 1.1 0.57 0.02 14.2 0.2 0.36 0.03 46.5 0.2 0.40 0.02 0.00 0.00 0.03 0.02 99.9 7 CIR1 CHS 71 1216 -1.59 39.1 0.2 0.57 0.03 15.6 0.1 0.35 0.04 44.0 0.1 0.33 0.02 0.00 0.00 0.00 0.00 101.2 7 numbers in italic are standard errors

Table S2 olivine composition in experiments

comp. exp # T (ºC) Δ IW SiO2 Cr2O3 FeO MnO MgO CaO P2O5 NiO total n

CIR2a CHS 50 1248 -1.89 40.7 0.3 0.50 0.02 10.5 0.2 0.40 0.02 47.5 0.1 0.39 0.05 0.02 0.02 0.00 0.00 100.1 3 CIR2a CHS 51 1217 -1.58 38.9 0.1 0.60 0.02 17.6 0.1 0.40 0.02 42.0 0.2 0.37 0.02 0.00 0.00 0.02 0.04 101.3 5 CIR2a CHS 53 1248 -1.47 38.1 0.3 0.66 0.00 18.6 0.0 0.40 0.01 41.8 0.1 0.45 0.02 0.00 0.00 0.01 0.01 101.5 3 CIR2a CHS 53 1248 -1.47 38.1 0.3 0.59 0.02 17.9 0.1 0.41 0.02 42.4 0.4 0.47 0.01 0.00 0.00 0.03 0.05 101.2 5 CIR2b CHS 63 1202 -1.38 38.1 0.3 0.68 0.02 18.6 0.3 0.28 0.03 41.8 0.2 0.38 0.02 0.00 0.00 0.07 0.05 99.7 5 CIR2b CHS 65 1201 -2.27 40.4 0.3 0.66 0.04 7.06 0.1 0.39 0.04 51.0 0.2 0.33 0.04 0.00 0.00 0.07 0.02 101.1 6 CIR2b CHS 66 1250 -2.53 40.7 0.4 0.70 0.02 4.99 0.1 0.39 0.02 52.8 0.4 0.38 0.08 0.00 0.00 0.05 0.02 100.8 7 CIR2b CHS 67 1274 -1.59 38.7 0.1 0.73 0.02 14.7 0.1 0.38 0.03 45.1 0.2 0.34 0.01 0.00 0.00 0.04 0.02 101.0 5 CIR2b CHS 68 1159 -1.53 38.9 0.1 0.47 0.03 18.5 0.1 0.38 0.02 41.3 0.2 0.36 0.05 0.00 0.00 0.03 0.02 101.4 7 CIR2b CHS 69 1301 -2.44 40.6 0.2 0.74 0.02 5.44 0.0 0.44 0.04 52.4 0.2 0.35 0.02 0.00 0.00 0.00 0.00 100.8 7 CIR2b CHS 71 1216 -1.59 39.1 0.3 0.77 0.02 15.2 0.2 0.41 0.04 43.9 0.7 0.43 0.12 0.00 0.00 0.00 0.00 101.4 7

CIR3 CHS 49 1248 -1.62 38.9 0.2 0.59 0.03 15.47 0.2 0.41 0.02 44.2 0.1 0.34 0.00 0.03 0.01 0.02 0.04 100.8 3 CIR3 CHS 50 1248 -1.89 40.4 0.2 0.56 0.01 11.14 0.1 0.44 0.03 47.0 0.5 0.33 0.02 0.02 0.01 0.07 0.06 99.4 3 CIR3 CHS 51 1217 -1.58 39.0 0.1 0.59 0.01 17.30 0.2 0.42 0.01 42.3 0.3 0.34 0.03 0.00 0.00 0.00 0.00 101.4 5 CIR3 CHS 53 1248 -1.47 38.4 0.0 0.64 0.02 17.64 0.1 0.42 0.01 42.5 0.1 0.36 0.00 0.00 0.00 0.04 0.05 101.2 3 177 RCa CHS 65 1201 -2.27 40.6 0.4 0.49 0.02 6.97 0.1 0.38 0.03 51.1 0.4 0.34 0.08 0 0 0.02 0.01 100.0 8

RCa CHS 66 1250 -2.53 40.8 0.3 0.52 0.03 4.69 0.1 0.37 0.02 53.2 0.4 0.40 0.09 0 0 0.03 0.02 100.6 9 RCa CHS 68 1159 -1.53 38.6 0.7 0.49 0.01 17.70 0.2 0.32 0.02 42.3 0.3 0.40 0.01 0 0 0.04 0.01 100.8 7 RCa CHS 69 1301 -2.44 41.2 0.3 0.53 0.01 5.65 0.1 0.37 0.05 51.7 0.2 0.46 0.02 0 0 0.05 0.02 100.2 6 RCa CHS 70 1205 -1.84 38.9 0.2 0.52 0.01 15.23 0.2 0.35 0.04 44.5 0.2 0.34 0.01 0 0 0.00 0 101.6 3

RCa2 CHS 68 1159 -1.53 39.2 0.2 0.40 0.02 17.7 0.1 0.36 0.04 42.0 0.2 0.39 0.09 0.0 0 0.03 0.01 101.4 7 RCa2 CHS 69 1301 -2.44 41.3 0.2 0.71 0.04 5.15 0.1 0.33 0.03 51.9 0.4 0.48 0.06 0.0 0 0.03 0.02 100.3 6 RCa2 CHS 70 1205 -1.84 39.1 0.2 0.53 0.03 14.0 0.0 0.41 0.04 45.5 0.1 0.31 0.03 0.0 0 0 0 101.4 4 numbers in italic are standard errors

177 Table S3 low Ca pyroxene composition in experiments T comp. exp # (ºC) Δ IW SiO2 TiO2 Al2O3 Cr2O3 FeO MnO MgO CaO P2O5 total n LLR1 CHS 46 1191 -1.61 54.6 0.8 0.09 0.02 0.47 0.10 0.91 0.04 12.7 1.2 0.37 0.02 27.6 0.5 3.24 0.43 0.11 0.04 101.0 17 LLR1 CHS 47 1200 -1.63 55.2 0.3 0.09 0.02 0.38 0.06 0.87 0.03 11.3 0.2 0.35 0.01 28.8 0.3 2.84 0.13 0.08 0.02 101.0 8 LLR1 CHS 48 1194 -1.98 56.5 0.5 0.11 0.03 0.44 0.10 0.86 0.05 7.66 0.2 0.39 0.02 31.0 0.6 2.98 0.28 0.08 0.03 100.7 14 LLR1 CHS 49 1248 -1.62 55.7 0.7 0.07 0.02 0.28 0.06 0.85 0.02 10.2 0.2 0.34 0.02 30.4 0.6 2.12 0.20 0.08 0.05 100.4 8 LLR1 CHS 51 1217 -1.58 55.1 0.3 0.05 0.02 0.34 0.05 0.85 0.04 11.2 0.2 0.38 0.01 29.5 0.8 2.47 0.07 0.09 0.03 99.8 3 LLR1 CHS 71 1216 -1.59 56.0 0.2 0.08 0.02 0.28 0.06 0.89 0.04 10.6 0.1 0.37 0.01 29.5 0.3 2.11 0.34 0.12 0.02 101.1 5

LLR2 CHS 47 1200 -1.63 55.4 0.5 0.10 0.02 0.34 0.11 0.90 0.02 11.2 0.5 0.44 0.01 28.7 0.4 2.68 0.06 0.09 0.03 101.0 6 LLR2 CHS 49 1248 -1.62 56.2 0.3 0.05 0.01 0.15 0.04 0.84 0.03 9.72 0.1 0.42 0.02 30.2 0.6 2.14 0.12 0.09 0.04 101.1 6 LLR2 CHS 51 1217 -1.58 55.3 0.2 0.04 0.01 0.22 0.05 0.83 0.03 11.0 0.3 0.43 0.00 29.3 0.1 2.75 0.14 0.12 0.02 99.8 2

CIR1 CHS 47 1200 -1.63 55.4 0.5 0.08 0.02 0.42 0.05 0.90 0.03 11.2 0.3 0.35 0.01 28.0 0.4 3.69 0.18 0.09 0.03 101.1 10 CIR1 CHS 48 1194 -1.98 56.2 0.7 0.11 0.01 0.49 0.05 0.92 0.03 7.56 0.1 0.37 0.02 30.7 0.4 3.49 0.12 0.08 0.01 100.7 9 CIR1 CHS 49 1248 -1.62 55.7 0.3 0.07 0.02 0.28 0.06 0.85 0.04 10.2 0.3 0.34 0.01 30.4 0.3 2.12 0.16 0.08 0.02 100.4 8 CIR1 CHS 50 1248 -1.89 56.5 0.6 0.07 0.02 0.32 0.07 0.70 0.02 7.64 0.5 0.35 0.01 31.7 0.2 2.66 0.16 0.06 0.04 99.9 10 CIR1 CHS 51 1217 -1.58 55.3 0.1 0.09 0.03 0.33 0.05 0.90 0.03 10.9 0.1 0.35 0.01 28.8 0.4 3.32 0.25 0.08 0.01 99.9 6 CIR1 CHS 53 1248 -1.47 54.9 0.8 0.06 0.03 0.28 0.05 0.84 0.03 11.3 0.1 0.33 0.01 30.0 0.2 2.28 0.21 0.06 0.02 99.4 5

178 CIR1 CHS 65 1201 -2.27 56.6 0.2 0.12 0.01 0.51 0.07 0.86 0.03 5.15 0.2 0.40 0.09 32.8 0.4 3.47 0.16 0.09 0.02 100.4 7 CIR1 CHS 66 1250 -2.53 57.2 0.5 0.11 0.01 0.26 0.03 0.76 0.03 4.07 0.6 0.40 0.02 34.6 0.1 2.55 0.01 0.08 0.04 100.8 4 CIR1 CHS 67 1274 -1.59 55.9 0.2 0.04 0.02 0.11 0.01 0.76 0.02 8.94 0.1 0.40 0.02 30.1 0.5 3.59 0.42 0.07 0.03 101.0 10 CIR1 CHS 71 1216 -1.59 56.1 0.4 0.08 0.02 0.28 0.04 0.89 0.04 10.1 0.1 0.33 0.02 29.4 0.2 2.65 0.16 0.09 0.04 101.0 7

numbers in italic are standard errors

178 Table S3 low Ca pyroxene composition in experiments T comp. exp # (ºC) Δ IW SiO2 TiO2 Al2O3 Cr2O3 FeO MnO MgO CaO P2O5 total n CIR2a CHS 50 1248 -1.89 56.3 0.4 0.06 0.01 0.24 0.04 0.70 0.04 7.23 0.2 0.43 0.02 29.9 0.3 4.96 0.20 0.10 0.02 99.9 11 CIR2a CHS 51 1217 -1.58 55.0 0.4 0.08 0.02 0.32 0.04 0.89 0.03 10.9 0.4 0.39 0.02 27.0 0.6 5.32 0.46 0.12 0.05 100.2 6 CIR2a CHS 53 1248 -1.47 55.4 0.5 0.05 0.02 0.17 0.01 0.83 0.04 10.8 0.2 0.38 0.01 27.0 0.6 5.30 0.43 0.07 0.04 100.7 8 CIR2a CHS 53 1248 -1.47 55.4 0.3 0.03 0.02 0.13 0.02 0.81 0.04 10.5 0.3 0.38 0.02 27.6 0.3 5.08 0.45 0.09 0.02 100.6 8 CIR2b CHS 63 1202 -1.38 54.3 0.5 0.11 0.03 0.44 0.06 1.13 0.04 10.9 0.3 0.42 0.02 26.6 0.5 5.92 0.56 0.10 0.03 99.7 4 CIR2b CHS 65 1201 -2.27 55.6 1.0 0.12 0.02 0.65 0.05 1.09 0.02 5.28 0.3 0.51 0.02 32.4 0.4 4.22 0.26 0.12 0.04 99.4 2 CIR2b CHS 66 1250 -2.53 56.8 0.7 0.07 0.02 0.33 0.09 0.99 0.03 3.71 0.2 0.47 0.01 33.8 0.9 3.73 0.23 0.09 0.02 101.0 14 CIR2b CHS 67 1274 -1.59 55.6 0.4 0.05 0.02 0.20 0.01 0.97 0.02 9.55 0.1 0.39 0.01 30.4 0.1 2.72 0.12 0.08 0.04 100.4 8 CIR2b CHS 68 1159 -1.53 54.8 0.5 0.12 0.02 0.47 0.04 0.84 0.03 11.8 0.3 0.43 0.04 26.8 0.3 4.52 0.15 0.14 0.03 99.7 5 CIR2b CHS 69 1301 -2.44 57.3 0.3 0.04 0.02 0.16 0.03 0.92 0.03 4.24 0.3 0.45 0.02 34.2 0.3 2.69 0.12 0.04 0.02 101.0 7 CIR2b CHS 71 1216 -1.59 55.8 0.7 0.06 0.02 0.31 0.07 1.12 0.03 9.59 0.2 0.41 0.03 28.4 0.4 4.14 0.35 0.11 0.03 101.2 12

CIR3 CHS 49 1248 -1.62 55.7 0.5 0.04 0.02 0.17 0.02 0.84 0.03 10.2 0.1 0.40 0.02 29.6 0.3 2.91 0.20 0.08 0.03 100.7 8 CIR3 CHS 50 1248 -1.89 56.3 0.5 0.06 0.02 0.25 0.06 0.76 0.03 7.75 0.4 0.44 0.02 31.4 0.3 2.94 0.10 0.14 0.06 100.0 7 CIR3 CHS 51 1217 -1.58 55.8 1.2 0.08 0.02 0.32 0.09 0.84 0.03 10.9 0.2 0.41 0.01 28.1 0.7 3.46 0.26 0.13 0.03 100.1 4 CIR3 CHS 53 1248 -1.47 55.6 0.2 0.06 0.01 0.23 0.07 0.85 0.02 11.1 0.2 0.40 0.02 28.8 0.4 2.91 0.26 0.06 0.03 100.7 15

179 RCa CHS 65 1201 -2.27 56.5 0.3 0.05 0.01 0.72 0.07 0.84 0.04 4.68 0.2 0.51 0.04 32.0 0.4 4.53 0.38 0.09 0.04 100.5 5 RCa CHS 66 1250 -2.53 57.1 0.2 0.02 0.02 0.39 0.04 0.74 0.04 3.25 0.1 0.44 0.03 32.6 0.4 5.38 0.40 0.08 0.02 100.9 9 RCa CHS 68 1159 -1.53 55.1 0.3 0.05 0.01 0.96 0.26 0.92 0.08 10.4 0.2 0.34 0.02 25.8 0.7 6.23 0.87 0.16 0.09 100.1 12 RCa CHS 69 1301 -2.44 56.5 0.7 0.03 0.01 0.12 0.01 0.66 0.04 5.28 1.2 0.39 0.01 32.3 0.6 4.69 0.37 0.05 0.02 99.8 5 RCa CHS 70 1205 -1.84 55.6 0.3 0.05 0.02 0.81 0.11 0.89 0.04 9.44 0.3 0.34 0.01 27.8 0.4 4.97 0.20 0.13 0.05 100.4 8

RCa2 CHS 69 1301 -2.44 57.4 0.1 0.01 0.01 0.14 0.03 0.87 0.03 4.26 0.5 0.33 0.03 32.4 0.3 4.77 0.09 0.03 0.02 99.4 6

numbers in italic are standard errors

179 Table S4 augite composition in experiments comp. exp # T (ºC) Δ IW SiO2 TiO2 Al2O3 Cr2O3 FeO MnO MgO CaO Na2O total n CIR2b CHS 65 1201 -2.27 54.4 0.11 0.98 0.86 2.90 0.34 22.5 17.8 0.19 101.1 1 RCa CHS 65 1201 -2.27 54.2 0.2 0.12 0.02 1.28 0.09 0.82 0.02 2.78 0.1 0.33 0.02 21.7 0.2 18.5 0.28 0.18 0.02 100.8 7 RCa CHS 66 1250 -2.53 55.2 0.3 0.08 0.04 0.75 0.13 0.66 0.03 2.29 0.1 0.33 0.04 23.4 1.2 17.0 1.59 0.16 0.06 101.3 11 RCa CHS 68 1159 -1.53 53.4 0.9 0.09 0.02 1.44 0.09 1.14 0.09 6.93 0.2 0.24 0.01 20.4 0.3 16.1 0.66 0.24 0.02 100.3 2 RCa CHS 69 1301 -2.44 55.2 0.0 0.15 0.01 0.97 0.25 1.01 0.11 3.43 0.0 0.36 0.04 25.7 0.3 13.1 0.32 0.09 0.04 100.8 3 RCa CHS 70 1205 -1.84 53.0 0.0 0.10 0.01 1.60 0.11 1.17 0.04 5.92 0.0 0.24 0.02 19.5 0.5 18.3 0.53 0.21 0.00 101.3 8 RCa2 CHS 68 1159 -1.53 52.9 0.1 0.22 0.01 1.50 0.06 1.09 0.02 6.39 0.1 0.24 0.02 18.0 0.6 19.2 0.44 0.27 0.05 100.3 3 RCa2 CHS 69 1301 -2.44 55.7 0.1 0.09 0.01 0.81 0.08 1.37 0.09 3.21 0.2 0.35 0.00 25.6 0.1 12.8 0.06 0.09 0.01 100.3 4 RCa2 CHS 70 1205 -1.84 52.8 0.8 0.18 0.03 1.58 0.09 1.19 0.11 5.80 0.2 0.25 0.01 20.4 0.0 17.6 0.10 0.20 0.05 101.4 5 numbers in italic are standard errors

Table S5a Olivine SiO2 TiO2 Al2O3 Cr2O3 FeO MnO MgO CaO Total n Fo LAP 03721 38.1 0.3 0.01 0.01 0.00 0.00 0.63 0.08 22.8 0.15 0.40 0.02 38.8 0.5 0.33 0.04 101.1 17 75.2 ol-pig EET 90019 40.0 0.2 0.01 0.02 0.00 0.00 0.64 0.04 10.3 0.11 0.45 0.03 48.7 0.2 0.34 0.07 100.4 10 89.4 ol-pig NWA 4852 39.5 0.3 0.01 0.01 0.03 0.01 0.64 0.03 12.0 0.15 0.45 0.03 47.3 0.3 0.31 0.01 100.2 15 87.5 ol-pig

180 MIL 090076 38.4 0.1 0.03 0.02 0.04 0.01 0.73 0.06 19.9 0.21 0.43 0.05 41.6 0.1 0.36 0.03 101.5 7 78.8 ol-pig MIL 07447_10 39.3 0.3 0.01 0.01 0.03 0.01 0.71 0.02 16.8 0.07 0.41 0.05 43.7 0.1 0.34 0.01 101.3 9 82.2 ol-pig

DOM 08012_15 39.1 0.1 0.03 0.02 0.02 0.01 0.71 0.03 20.6 0.15 0.44 0.03 39.9 0.1 0.36 0.02 101.1 5 77.6 ol-pig EET 96042_60 39.3 0.1 0.00 0.01 0.04 0.01 0.84 0.05 17.4 0.10 0.43 0.03 42.9 0.1 0.36 0.02 101.4 10 81.5 ol-pig NWA 11755 38.3 0.3 0.02 0.02 0.01 0.01 0.73 0.04 20.3 0.09 0.40 0.03 41.0 0.2 0.36 0.02 101.2 13 78.3 ol-pig NWA 5555 40.3 0.1 0.01 0.01 0.02 0.01 0.64 0.03 8.9 0.07 0.44 0.04 49.4 0.2 0.31 0.02 100.1 14 90.9 ol-pig-opx ALHA 82106 40.8 0.3 0.01 0.01 0.03 0.01 0.58 0.02 4.5 0.08 0.53 0.03 53.8 0.2 0.28 0.02 100.5 6 95.5 ol-opx-aug NWA 11754" 39.7 0.1 0.02 0.01 0.03 0.01 0.51 0.02 12.0 0.06 0.53 0.04 47.2 0.2 0.29 0.01 100.3 12 87.5 ol-opx-aug EET 96293" 39.3 0.1 0.01 0.01 0.03 0.01 0.48 0.05 12.3 0.10 0.56 0.03 46.9 0.1 0.29 0.02 101.4 10 87.1 ol-opx-aug "Hughes cluster samples

180 Table S5b Pigeonite SiO2 TiO2 Al2O3 Cr2O3 FeO MnO MgO CaO Na2O Total n Mg# Wo LAP 03721 54.0 0.7 0.05 0.01 0.93 0.21 1.22 0.22 13.3 0.32 0.40 0.02 25.8 1.1 5.27 1.19 0.07 0.02 101.1 14 77.6 10.2 EET 90019 56.3 0.3 0.12 0.02 0.81 0.02 1.04 0.06 6.3 0.12 0.48 0.03 31.0 0.2 4.77 0.07 0.07 0.03 100.9 10 89.8 9.0 NWA 4852 56.3 0.2 0.11 0.02 0.75 0.01 1.04 0.04 7.3 0.10 0.47 0.02 30.3 0.2 4.36 0.11 0.05 0.03 100.6 11 88.1 8.4 MIL 090076 54.7 0.3 0.10 0.01 0.95 0.07 1.39 0.05 11.5 0.11 0.42 0.01 27.0 0.2 5.54 0.10 0.10 0.04 101.7 16 80.7 10.6 MIL 07447_10 55.0 0.4 0.08 0.02 0.89 0.01 1.25 0.03 9.8 0.09 0.45 0.02 26.9 0.1 5.28 0.09 0.03 99.7 9 83.1 10.5 DOM 08012_15 55.6 0.1 0.04 0.03 0.51 0.03 1.08 0.04 12.3 0.03 0.40 0.02 26.5 0.0 3.56 0.02 0.04 0.02 100.2 3 79.3 7.1 EET 96042_60 55.5 0.1 0.11 0.01 0.73 0.02 1.37 0.01 10.2 0.11 0.46 0.01 27.0 0.1 4.58 0.04 0.11 0.03 100.1 7 82.5 9.1 NWA 11755 55.4 0.4 0.04 0.02 0.44 0.11 1.01 0.09 11.7 0.33 0.41 0.02 27.7 0.4 4.28 0.31 0.04 0.02 101.0 24 80.8 8.2 NWA 5555 56.5 0.5 0.12 0.01 0.55 0.02 1.03 0.03 5.5 0.06 0.47 0.02 30.9 0.1 4.53 0.05 0.04 0.02 99.7 10 90.9 8.7 ALHA 82106* 57.2 0.16 0.54 0.82 2.7 0.45 33.7 5.72 0.08 101.3 n.a. 95.6 10.5 * average non-inverted pigeonite (orthopyroxene+augite exsolutions)

Table S5c Orthopyroxene SiO2 TiO2 Al2O3 Cr2O3 FeO MnO MgO CaO Na2O Total n Mg# Wo NWA 5555 57.2 0.9 0.13 0.01 0.57 0.02 0.95 0.04 5.4 0.04 0.42 0.01 32.9 0.1 2.53 0.02 0.02 0.02 100.2 8 91.6 4.8

181 ALHA 82106 57.6 0.3 0.13 0.01 0.47 0.01 0.81 0.02 3.0 0.08 0.47 0.03 36.2 0.2 2.62 0.02 0.03 0.02 101.2 5.00 95.6 4.8 NWA 11754" 56.0 0.5 0.15 0.01 1.21 0.02 0.96 0.04 7.3 0.11 0.51 0.02 31.0 0.1 2.52 0.02 0.01 0.01 99.5 12 88.4 4.9

EET 96293" 56.3 0.2 0.13 0.02 1.24 0.02 1.09 0.02 7.4 0.08 0.52 0.02 32.1 0.2 2.53 0.03 0.05 0.02 101.5 10.00 88.5 4.8 "Hughes cluster samples

Table S5d augite SiO2 TiO2 Al2O3 Cr2O3 FeO MnO MgO CaO Na2O Total n Mg# Wo ALHA 82106 55.4 0.1 0.28 0.01 0.85 0.03 0.88 0.03 1.8 0.05 0.36 0.01 23.0 0.1 18.92 0.05 0.30 0.03 101.8 11.00 95.7 36.1 NWA 11754" 53.2 0.5 0.28 0.02 1.91 0.04 1.20 0.03 4.2 0.05 0.39 0.02 20.5 0.1 18.25 0.09 0.20 0.02 100.0 5 89.7 36.5 EET 96293" 53.7 0.0 0.27 0.03 1.90 0.06 1.35 0.05 4.2 0.07 0.40 0.02 20.4 0.1 18.78 0.03 0.23 0.03 101.3 5.00 89.6 37.2 "Hughes cluster samples

181 Table S5e Chromite in MIL 07447 oxides SiO2 TiO2 Al2O3 Cr2O3 FeO MnO MgO CaO NiO ZnO Total Comment 0.08 0.74 12.10 59.50 17.79 0.52 11.16 0.03 0.00 0.35 102.3 inclusion in pig 0.10 0.70 12.01 59.25 17.83 0.45 11.20 0.04 0.02 0.33 101.9 inclusion in pig 0.06 0.70 11.94 58.94 18.23 0.43 10.97 0.05 0.01 0.33 101.7 inclusion in pig 0.10 0.71 12.34 59.54 17.68 0.45 11.30 0.00 0.02 0.39 102.5 inclusion in oliv (core) 0.07 0.74 12.35 59.45 16.93 0.41 11.89 0.00 0.02 0.39 102.3 inclusion in oliv 0.08 0.74 12.32 60.07 14.22 0.47 13.94 0.00 0.01 0.33 102.2 inclusion in oliv 0.06 0.72 12.51 59.92 12.64 0.56 14.92 0.00 0.02 0.45 101.8 inclusion in oliv (rim)

atoms Si Ti Al Cr Fe Mn Mg Ca Ni Zn Total Comment 0.002 0.018 0.454 1.500 0.474 0.014 0.531 0.001 0.000 0.008 3.002 inclusion in pig 0.003 0.017 0.453 1.498 0.477 0.012 0.534 0.001 0.001 0.008 3.004 inclusion in pig 0.002 0.017 0.452 1.498 0.490 0.012 0.526 0.002 0.000 0.008 3.007 inclusion in pig 0.003 0.017 0.462 1.494 0.469 0.012 0.535 0.000 0.001 0.009 3.002 inclusion in oliv (core) 0.002 0.018 0.462 1.491 0.449 0.011 0.562 0.000 0.001 0.009 3.005 inclusion in oliv 0.003 0.017 0.455 1.488 0.373 0.013 0.651 0.000 0.000 0.008 3.008 inclusion in oliv 0.002 0.017 0.461 1.480 0.330 0.015 0.695 0.000 0.001 0.010 3.011 inclusion in oliv (rim) 182

182 Chapter 4. Melting of the Primitive Martian Mantle at 0.5-2.2 GPa and the Origin of Basalts and Alkaline Rocks on Mars

Abstract

We have performed piston-cylinder experiments on a primitive martian mantle composition between 0.5 and 2.2 GPa and 1160 to 1550 ºC. The composition of melts and residual minerals constrain the possible melting processes on Mars at 50 to 200 km depth under nominally anhydrous conditions. Silicate melts produced by low degrees of melting (< 10 wt.%) were analyzed in layers of vitreous carbon spheres or in micro-cracks inside the graphite capsule. The total range of melt fractions investigated extends from 5 to 50 wt.%, and the liquids produced display variable SiO2 (43.7-59.0 wt.%), MgO (5.3-18.6 wt.%) and

Na2O+K2O (1.0-6.5 wt.%) contents. We provide a new equation to estimate the solidus temperature of the martian mantle: T (ºC) = 1033 + 168.1 P (GPa) - 14.22 P2 (GPa), which places the solidus 50 ºC below that of fertile terrestrial peridotites. Low- and high-degree melts are compared to martian alkaline rocks and basalts, respectively. We suggest that the parental melt of Adirondack-class basalts was produced by ~25 wt.% melting of the primitive martian mantle at 1.5 GPa (~135 km) and ~1400 °C. Despite its brecciated nature, NWA 7034/7533 might be composed of material that initially crystallized from a primary melt produced by ~10-30 wt.% melting at the same pressure. Other igneous rocks from Mars require mantle reservoirs with different CaO/Al2O3 and FeO/MgO ratios or the action of fractional crystallization. Alkaline rocks can be derived from mantle sources with alkali contents (~0.5 wt.%) similar to the primitive mantle.

183 1. Introduction

Remote analyses by rovers and orbiting spacecraft as well as the compositions of shergottites, the most abundant group of martian meteorites, suggested that the surface of Mars is mainly covered by iron-rich tholeiitic basalts (McSween et al., 2009). Further studies of the infrared spectral data (Carter and Poulet, 2013; Christensen et al., 2005; Wray et al., 2013), new meteorite discoveries (Agee et al., 2013; Humayun et al., 2013), recent analyses by the Mars Science Laboratory mission (Stolper et al., 2013; Sautter et al., 2014), and re-interpretation of Pathfinder analyses (Foley et al., 2003) have also highlighted the presence of alkaline rocks, silicic rocks, and anorthosites. These occurrences indicate that the martian crust exhibits a larger compositional variability than initially thought. Part of this variability could result from magmatic differentiation (McSween et al., 2003; Rogers and Nekvasil, 2015), but a better knowledge of the range of potential primary melts is necessary to understand the effects of processes such as fractional crystallization in more detail. In order to evaluate the extent of compositional variability arising from partial melting of the mantle, we have performed nominally anhydrous experiments over a broad range of pressure (0.5-2.2 GPa) and temperature (1160-1550 ºC) on the most widely accepted composition of the primitive martian mantle (Dreibus and Wanke, 1984; 1985). This study extends earlier experiments performed on similar compositions to a larger pressure range (Agee and Draper, 2004; Bertka and Holloway, 1994b) and melting degree (Matsukage et al., 2013), and provide a comprehensive understanding of potential melting products. In particular, low-degree melts, which are of considerable importance to constrain alkaline magmatism on Mars, have not yet been documented. Our new array of experimental melts enables us to provide new equations to calculate the solidus temperature of the primitive martian mantle and the melt fraction as a function of pressure and temperature between 0 and 2.5 GPa. Improved accuracy on the position of the solidus is critical to reconstruct the geodynamic and thermochemical evolution of Mars (Baratoux et al., 2011; Breuer and Spohn, 2006; Plesa et al., 2014). In addition, the compositions of our experimental melts are compared to crustal martian igneous rocks and serve as a reference for discussing the melting regimes in the mantle and the relative importance of differentiation processes and source heterogeneities. Pressure and temperature conditions of melting are estimated from the natural samples that are similar in composition to the experimental melts. Finally, when the mineralogy and texture of martian rocks are known to some extent, our experimental melt compositions can

184 help determine if they represent silicate liquids, cumulates or rocks affected by secondary alteration.

2. The primitive mantle of Mars

Models have been developed to estimate the composition of the bulk silicate portion of Mars (see Taylor, 2013, for a recent review). These models are based on both geochemical (Dreibus and Wanke, 1984; 1985; Lodders and Fegley Jr, 1997; Morgan and Anders, 1979; Sanloup et al., 1999) and geophysical arguments (Khan and Connolly, 2008). They converge towards a single composition that displays key differences from the bulk silicate Earth: lower Mg# (75-78) and higher incompatible (Na, K, P) and compatible (Cr, Mn) volatile elements (Table 1). This composition represents the primitive mantle of Mars (hereafter referred to as PMM) unaffected by magmatic processes such as magma ocean fractional crystallization and crust formation. We selected the composition from the model of Dreibus and Wanke (1984) for a number of reasons: (1) it is based on comparisons between shergottites and CI chondrites with carefully chosen ratios of elements that behave similarly in magmatic systems; (2) of all the geochemical models, it is the closest to the physically constrained composition of Khan and Connolly (2008); (3) while enriched in volatile elements relative to the bulk silicate Earth, it is characterized by less extreme alkali contents than other models (Lodders and Fegley Jr, 1997; Sanloup et al., 1999) and provides a composition of the primitive mantle after the potential syn-accretion volatile depletion (Albarede, 2009; Halliday and Porcelli, 2001); (4) it is similar to compositions used for earlier experimental studies of the melting of the martian mantle (Agee and Draper, 2004; Bertka and Holloway, 1994a,b; Matsukage et al., 2013), allowing direct comparisons.

3. Experimental and analytical methods

3.1. Starting material and experimental conditions

The starting material was synthesized from dried high purity oxides (SiO2, TiO2, Al2O3,

Cr2O3, Fe2O3, MnO, MgO) and pre-conditioned silicate mixes (CaSiO3, NaAlSi3O8, KAlSi3O8, 3+ 3+ Fe2SiO4). Fe was added as Fe2SiO4 and Fe2O3 (0.2 mol.% of Fe ). The amount of Fe was set to match the speciation of Fe in a melt between the redox buffers CCO and IW under the experimental conditions (Holloway et al., 1992; Médard et al., 2008). Series of 5 to 8 experiments were conducted at each pressure from 0.5 to 2.0 GPa in 0.5

185 GPa increments and two additional experiments were performed at 2.2 GPa. This pressure range corresponds to a depth of 50 to 200 km in the planet’s interior. Experiments were performed in piston cylinder apparatus at the “Laboratoire Magmas et Volcans” (LMV) in Clermont-Ferrand and at the Massachusetts Institute of Technology (MIT). The experimental temperature (1160-1550 ºC) was measured close to the hotspot with type C (LMV) and D (MIT) thermocouples and kept constant during the duration of the experiment (24-96 hours). The temperature reproducibility was monitored to be within ±10 ºC. A mass of 8-15 mg of the starting material was introduced into a graphite capsule inside a platinum outer capsule, and dried at 400 °C for 12 hours before being welded shut to ensure near- anhydrous conditions. The water content is thought to not exceed 0.02-0.05 wt.% from past experiments using similar drying procedures (Laporte et al., 2004; Médard et al., 2008), equivalent to the water content of the martian mantle estimated from apatite compositions (McCubbin et al., 2012). At LMV, experiments were conducted in end-loaded 12.7 mm (2 GPa) and non-end-loaded 19.1 mm (< 2 GPa) pressure vessels. The experimental assembly was composed of crushable MgO spacers, an inner Pyrex sleeve, a graphite furnace, an outer Pyrex sleeve and a NaCl cell wrapped in lead foil. Such an assembly has very low friction: calibration of the 19.1 mm assembly against the melting point of salt resulted in a negligible correction (maximum 3%), whereas the 12.7 mm assembly has a friction correction ≤ 5%, as determined from the quartz-coesite transition. At MIT,

12.7 mm vessels were used, a BaCO3 cell replaced the salt cell as pressure medium and the experimental assembly did not contain Pyrex. This assembly has been found to have no friction correction through calibration against the reaction kushiroite = anorthite + gehlenite + corundum (Hays, 1966). Pressures are thought to be accurate to within ± 0.05 GPa. Obtaining a glass that represents the equilibrium liquid is non-trivial for experiments at low degrees of melting. During quench, crystal growth modifies the composition of melts preserved in small pockets. Therefore, from 0.5 to 1.5 GPa, we used a thin layer of vitreous carbon spheres (Fig. 1A) to isolate some of the liquid from the crystals (Wasylenki et al., 2003). Because vitreous carbon transforms to graphite needles at higher pressure (2.0-2.2 GPa), we optimized the design of graphite capsules to induce micro-cracks in the lid (Fig. 1B) and extract the melt while limiting iron loss to the Pt outer capsule (Laporte et al., 2004). Melt extraction is helped by the small temperature gradient (10 ºC) inherent to our experimental design. For both extraction techniques, we oriented the capsule so that the liquid traps were located at the hotspot.

186 3.2. Analytical techniques

Experimental run products were analyzed by electron microprobe, using a Cameca SX-100 at LMV and a JEOL JXA-8200 Superprobe at MIT. A selection of experiments was measured on both microprobes to ensure inter-laboratory consistency and results were found to be identical within analytical errors. At LMV, the beam conditions were set to 15 kV and 15 nA for silicate and spinel analyses. For glasses, the intensity was reduced to 8 nA and the beam defocused to 10 µm to limit Na loss. Peak and background counting times were 20–40 s per element, and matrix corrections were performed with the PAP model (Pouchou and Pichoir, 1984). At MIT, a current intensity of 10 nA and a voltage of 15 kV were used for crystals and glasses. To analyze glasses, the beam was defocussed to 10 µm and Na was first counted for 5 s. The CITZAF package was used for matrix correction. Analytical standards include well-characterized synthetic and natural glasses and minerals including diopside (Si, Ca, Mg), oligoclase (Na, Al), hematite (Fe), rutile (Ti), chromite (Cr), rhodochrosite (Mn), orthoclase (K) and apatite (P). Phase proportions in experiments were determined by mass balance calculation from the chemical compositions of the different minerals and the glass. We used the linear regression algorithm LIME (Krawczynski and Olive, 2011), which prevents negative proportions and takes analytical uncertainties into account.

4. Results

4.1. Attainment of equilibrium

Experimental textures are consistent with equilibrium conditions (Fig. 1A). The grain boundaries are visible even for low degrees of melting (5-10 wt.%) and highlight the presence of interstitial melt throughout run products, a prerequisite to equilibrate melt traps with crystalline phases. No zoning in ferromagnesian silicates was detected. Only one mineral phase, the sub-calcic augite, exhibits a significant amount of chemical variability (Fig. 2). Equilibrium between melt and olivine is usually tested by the exchange coefficient between

Fe and Mg (KD = ratio of Fe/Mg in olivine over Fe/Mg in the melt,). Experimental KD’s are in the range 0.30-0.38 (Table 2), in good agreement with the model of Toplis (2005) and the classic value of 0.35 generally accepted for martian compositions (Filiberto and Dasgupta, 2011). Fe loss has been estimated by mass balance calculations using all the elements except Fe. Phase proportions are multiplied by the FeO content of each phase and the sum is compared to the

187 FeO content of the starting material. Relative Fe variations compared to the starting material are between +5 % and -7 %, within uncertainties of the mass-balance calculations. The final test for equilibrium is based on the quality of mass balance calculations performed from all the elements (including Fe). The sum of squared residuals is low in most cases (<0.3) and can reach slightly higher values (up to 1.3) for low-degree melts or when the melt formed dendrites during the quench. Overall, the combination of those tests and the fact that melt compositions and phase proportions form consistent trends (section 4.3 and 4.4) indicates that experiments are close to thermodynamic equilibrium. Graphite capsules set an upper limit to the experimental oxygen fugacity at the CCO buffer

(Holloway et al., 1992). The average fO2 of this type of experiment is estimated at CCO-0.8 (~IW+0.5; Médard et al., 2008). However, redox conditions likely vary slightly (possibly by one log unit) with pressure, temperature and melt fraction. These variations in redox conditions are unlikely to have noticeable effects on phase relations. An important exception could be the stability of Cr-rich spinel, as both Cr2+ and Cr3+ are present under the experimental redox conditions. Since Cr3+ is more compatible in spinel, the stability limit of spinel at high temperature provided in this study is imperfectly constrained and will depend on oxygen fugacity.

4.2. Phase assemblages

Just above the solidus, four (1.0-2.2 GPa) to five (0.5 GPa) crystalline phases are in equilibrium with experimental melts. The mineral assembly is composed of olivine, orthopyroxene, sub-calcic augite, spinel and plagioclase at 0.5 GPa. When plagioclase is present, it quickly melts out, and clinopyroxene is the next phase to disappear. At 0.5 and 1.0 GPa, orthopyroxene coexists with sub-calcic augite at low temperature and with pigeonite at higher temperature, highlighting the presence of a pigeonite-augite solvus (Fig. 2). Eventually, pigeonite melts out as the melt fraction increases to around 20 wt.%. At 1.5-2.2 GPa, sub-calcic augite is completely consumed after 10-15 wt.% melting and pigeonite is no longer stable. Spinel is the next phase to melt out from the residue when the melt fraction reaches about 25 wt.%. At higher temperature, orthopyroxene and olivine compose the residual mineral assembly.

4.3. Mineral compositions

Mineral compositions evolve with pressure and temperature (Supplemental Table 2). The CaO content of the clinopyroxene decreases as the temperature increases. At 0.5 and 1.0 GPa, it

188 drops from 12 wt.% (sub-calcic augite) to 4 wt.% (pigeonite). The Mg# of olivine and orthopyroxene also evolve with temperature. The Mg# continuously increases from 74 (olivine) and 77 (orthopyroxene) at low temperature (5 wt.% of melt) to 83-85 at high temperature (50% of melt). Orthopyroxene and sub-calcic augite are richer in Al2O3 closer to the solidus and at higher pressure. Finally, the Al2O3 content of spinel decreases with temperature while the Cr2O3 content increases.

4.4. Melt compositions

Experimental melt compositions are strongly dependent on pressure at low degrees of melting. They vary from basanite (2.2 GPa), to nepheline normative hawaiite (1.5 GPa), to mugearite (1.0 GPa), and benmoreite (0.5 GPa) but converge in the field of basalts for higher degrees of melting (Fig. 4). The silica content decreases with increasing pressure from 59.0 wt.% at 0.5 GPa to 43.7 wt.% at 2.2 GPa for the near-solidus experiments (Fig. 5, Table 2). SiO2 decreases with increasing degree of melting at low pressure (0.5-1.0 GPa) and slightly increases with increasing degree of melting at high pressure (2.0-2.2 GPa). The Al2O3 content decreases with increasing temperature at all pressures but is higher in the lower-pressure melts. However, at 0.5

GPa and low temperature, the melt is in equilibrium with plagioclase and contains less Al2O3 than melts in experiments at 1.0 GPa. FeO, MgO and MnO are more abundant at high pressure and increase with increasing temperature. CaO behaves as a compatible component and increases in the melt as the melting progresses. CaO reaches a maximum before clinopyroxene disappears and is progressively diluted afterwards. The alkalis are incompatible in residual minerals and decrease with increasing degree of melting. They are slightly more abundant in low-pressure melts as the compatibility of Na in pyroxene is higher at high pressure. Conversely, while titanium and phosphorus are also incompatible, they are more abundant in high-pressure melts. The low P2O5 content of low-pressure liquids reflects a higher compatibility of phosphorus in olivine at low pressure. Distribution coefficients for P between olivine and melt decrease from 0.045 at 1.0 GPa

(DW18) to 0.018 at 2.2 GPa (C547). Cr2O3 is compatible and increases in the melt with increasing degree of melting. At low temperature, high-pressure melts are slightly enriched in Cr2O3 compared to low-pressure melts which are close to the detection limit of EPMA analysis. Water contamination cannot be entirely avoided in nominally anhydrous experiments (e.g. Laporte et al., 2004; Médard et al., 2008). Water contents were not directly measured in our experimental glasses. However, the totals of EPMA analyses are close to 100 wt.% (Table 2) and

189 are uncorrelated to the melt fractions. This is consistent with small amounts of H2O, not exceeding the contamination expected from previous experiments using the same drying procedure. The average total of analyses performed on the 7 low-degree melts (<10 wt.% of melt) is 99.3 wt.% and suggest that the total amount of dissolved volatiles is < 1wt.%. A maximum value of 0.02-0.05 wt.% H2O in the bulk (Laporte et al., 2004; Médard et al., 2008) would translate to a maximum of

0.5-1.0 wt% H2O in the lowest degree melts. In addition to water contamination, the dissolution of a small amount of carbon is ubiquitous in experiments in equilibrium with graphite. The maximum amount of CO2 expected in our experiments would be in the range 0.05-0.2 wt.% CO2 in the melt 3+ if most Fe is reduced and the fO2 reached the CCO buffer (Holloway et al., 1992). Projection diagrams (Fig. 6) from calculated mineral components (Grove, 1993) highlight the pressure effect on the composition of melts. Melts move away from the quartz apex and towards the olivine and clinopyroxene apices with increasing pressure. As the degree of melting increases, melts get closer to the clinopyroxene apex until clinopyroxene disappears from the melting residue. With only orthopyroxene and olivine being consumed, melts then move towards the olivine-quartz join. Finally, melts are aligned with the bulk composition and the olivine apex when orthopyroxene melts out. Melt compositions are in good agreement with experiments from Bertka and Holloway

(1994b). However, their lower temperature experiment is ~2 wt.% poorer in SiO2. This is likely due to the omission of key minor elements (K, P, Mn, Cr) in their starting material but could also reflect a lower H2O contamination (Supplemental Fig. 2). Experiments from Matsukage et al. (2013) are anomalous compared to our experiments and those from Bertka and Holloway (1994b). In particular, we believe that the composition of 1.0 GPa melts by Matsukage et al. (2013) is affected by quench modifications as shown by the low olivine/melt – Fe/Mg KD’s (0.21-0.25 instead of ~0.35 in this study). Note that the thermodynamic calculator pMELTS considerably underestimates SiO2 contents and overestimate FeO contents of our melts (~5 wt.% offsets; Supplemental Fig. 4). Similar shortcomings have also been observed for terrestrial compositions (e.g. Gaetani, 1998).

4.5. Melting behavior

The proportions of residual mineral phases are dependent on the degree of melting (or the temperature) and the pressure (Fig. 7). At 0.5-1.0 GPa and low temperature, the olivine proportion first increases with increasing degree of melting while the proportions of pyroxene and spinel

190 diminish. The relative proportions of clinopyroxene and orthopyroxene evolve in a complex manner. The amount of sub-calcic augite slightly increases while its wollastonite content drops. As a result, the CaO content of the liquid increases with temperature despite the increase of the clinopyroxene fraction. Eventually, sub-calcic augite transforms into pigeonite when the solvus is reached (Fig. 2) and the proportion of clinopyroxene falls off rapidly at higher temperature. A melting reaction can be calculated if we approximate the clinopyroxene fraction to decrease linearly with increasing temperature. The following expression represents a peritectic melting reaction and is valid at 0.5 and 1.0 GPa.

0.47 (±0.15) orthopyroxene + 0.80 (±0.20) clinopyroxene + 0.08 (±0.03) spinel = 0.35 (±0.10) olivine + 1.00 melt

Above 1.0 GPa, olivine is a reactant and orthopyroxene is produced as long as clinopyroxene is present. In addition, the augite-pigeonite solvus likely becomes metastable and/or sub-calcic augite disappears before the solvus can be reached. The melting reactions can thus be approximated by peritectic reactions producing orthopyroxene instead of olivine at 1.5 GPa:

1.60 (±0.3) clinopyroxene + 0.22 (±0.05) olivine + 0.08 (±0.03) spinel = 0.90 (±0.20) orthopyroxene + 1.00 melt and at 2.0 GPa:

2.10 (±0.30) clinopyroxene + 0.23 (±0.05) olivine + 0.17 (±0.06) spinel = 1.50 (±0.30) orthopyroxene + 1.00 melt

Following the disappearance of clinopyroxene, melting reactions turn into eutectic reactions that can be expressed by a single equation, which is insensitive to pressure:

0.70 (±0.15) orthopyroxene + 0.30 (±0.08) olivine = 1.0 melt

The experimental pressure and temperature have a considerable effect on the composition of silicate melts coexisting with olivine, orthopyroxene and clinopyroxene. The silica activity of the liquid drops with pressure while the activities of MgO and FeO increase (Supplemental Fig. 1; Sack and Ghiorso, 1989; Hirschmann et al., 1998). Accordingly, the liquid in equilibrium with martian peridotites becomes enriched in olivine component as the pressure increases. Minor elements such as K, Na and H can also influence major elements concentrations by changing the melt structure and activity coefficients (Gaetani and Grove, 1998; Hirschman 1998). In particular,

191 the concentration of minor incompatible elements in low-degree melts could contribute to increase the SiO2 contents at low pressure (Supplementary material). Pressure and temperature conditions also have an effect on the stability of silicates as observed from the phase diagram (Fig. 3). At 2.0-2.2 GPa, the stability field of clinopyroxene is decreased and it melts out faster (10-15 wt.% of melt) than at lower pressure (20 wt.% of melt). On the contrary, the stability of orthopyroxene increases, as illustrated by the orthopyroxene-out curve, which enter regions of higher melt fraction at higher pressure.

5. Discussion

5.1. Position of the solidus

We use the melt fraction in each experiment, determined by mass-balance calculation, to estimate the position of the solidus of the martian mantle. The solidus temperature at each pressure is given by the zero intercept of the melt fraction as a linear function of the temperature. Then, the four solidus temperatures are fitted with first- and second-order polynomial equations. The following linear approximation is sufficient to predict the temperature of the solidus between 0.5 and 2.2 GPa, the range of pressure investigated.

TS (ºC) = 1054 – 129.3 P (GPa)

The martian solidus is 50 ºC lower than the solidus of a terrestrial peridotite over this pressure range (Hirschmann, 2000). This difference likely reflects the lower Mg# and higher alkali content of the Dreibus and Wänke (1984) composition. Compared to our results, Bertka and Holloway (1994a) estimated the solidus to be 30 ºC higher on average. This offset might result from the absence of K2O in their starting material and the difficulty of observing melt at low temperature without liquid traps. The pressure and temperature conditions of the solidus of the terrestrial mantle follow a second-order polynomial relationship up to 10 GPa (Hirschmann, 2000). We adjusted the coefficients of our own second-order fit to follow the curve of Hirschman (2000) and maintain a ~50 ºC offset between martian and terrestrial peridotites up to around 8 GPa (Supplemental Fig. 3):

2 Ts (ºC) = 1033 + 168.1 P (GPa) – 14.22 P (GPa)

We also performed a second order polynomial regression of the melt fraction as a function of the

192 pressure and temperature based on a selection of experiments. Experiment DW16 (1.5 GPa, 1450 ºC) is characterized by anomalous MgO/temperature and melt fraction/temperature ratios relative to the general trend (Fig. 7). It must have equilibrated at a higher temperature (>1500 ºC) and was excluded. We obtained the following expression for the melt fraction as a function of the pressure and temperature:

F = – 149.2 – 56.93 P – 2.442 P2 + 0.1439 T + 0.0321 P T

F is the melt fraction in weight percent; P is the pressure in GPa; T is the temperature in ºC. This function is plotted on the P-T phase diagram (Fig. 3) and can be compared to melt proportions in Supplemental Table 1. We believe that meaningful extrapolations can be performed up to 5 GPa (Supplemental Fig. 3).

5.2. Comparison with the melting of terrestrial peridotites

The melting behavior of the PMM and its pressure dependence share key characteristics with the melting of fertile terrestrial peridotites. First, the decrease in SiO2 with increasing pressure has been recognized by O'Hara (1968) and Takahashi and Kushiro (1983) as an important characteristic of peridotite melting. Peritectic reactions characterized by olivine production at low pressure and orthopyroxene production at high pressure are similar to those described by Kinzler and Grove (1992) and Kinzler (1997). However, the transition from olivine-production to orthopyroxene-production occurs at lower pressure (1.0-1.5 GPa) for the Fe-rich martian mantle compared to terrestrial compositions (~1.9 GPa; Kinzler, 1997). Our experiments showcase a CaO maximum that coincides with clinopyroxene disappearance for a degree of melting of 15 to 20 wt.%, similar to that reported by Baker and Stolper (1994). Compared to the Earth’s mantle, the principal difference in phase stability is the presence of pigeonite at a relatively high pressure (1.0 GPa) during partial melting because the pigeonite-augite solvus is stabilized to lower temperatures. The presence of a pigeonite-subcalcic augite solvus at high pressure and just above the solidus temperature results from the high Fe content of the martian mantle (Bertka and Holloway, 1993).

The high Cr2O3 content of the PMM might also stabilize spinel to higher temperature. For a given degree of melting, martian melts are richer in FeO, Cr2O3, alkalis and P2O5 and poorer in TiO2,

Al2O3 and CaO compared to their terrestrial equivalents.

193 5.3. Comparisons with martian volcanic rocks

The compositions of a large number of volcanic rocks analyzed in Gusev crater, and several new samples from the collection of martian meteorites, are similar to our experimental melts. Conversely, shergottites and the few volcanic rocks analyzed by the Curiosity rover in Gale crater present significant differences. In the following section, we compare these rocks to our experimental melts (Fig. 8) to discuss their origin and crystallization history.

5.3.1. Gusev plains: Adirondack-class basalts The Adirondack-class basalts were rapidly recognized as potential primary basalts derived from the martian mantle given their high MgO content (Mg# = 76.5) and abundant olivine phenocrysts (McSween et al., 2006b). The experimental study by Monders et al. (2007) confirmed that the average composition of the Adirondack basalts could represent a primary mantle melt as it is multiply saturated with orthopyroxene and olivine at 1.0 GPa and 1320 ºC. Adirondack-class basalts are similar to our experimental melts produced between 1.0 and 2.0 GPa for 15-25 wt.%. This range in melt fraction corresponds to the transition from a lherzolitic to a harzburgitic residue, and Adirondack basalts plot along the junction between orthopyroxene- olivine and clinopyroxene-orthopyroxene-olivine cotectic curves on projection diagrams (Fig. 6). Overall, we observe that the best match with Adirondack basalts is a liquid produced by 25 wt.% melting (1380 ºC) at 1.5 GPa (135 km depth) that crystallized 5 wt.% of olivine. Monders et al. (2007) proposed source conditions of 1.0 GPa and 1320 ºC from the position of the olivine- orthopyroxene multi-saturation point (MSP) in their experiments. These different conditions of melting result from the higher FeO/MgO ratio of Adirondack basalts compared to our experimental melts. The difference between the FeO/MgO ratios might reflect a source slightly richer in iron or the fractionation of a small amount of olivine from the parental melt of the Adirondack-class basalts. An addition of 5 wt.% of olivine to a basaltic composition can move the MSP by up to 0.5 GPa and 90 ºC towards higher pressure and temperature (Médard et al., 2004). Accordingly, the source conditions proposed by Monders et al. (2007) represent lower-pressure/temperature limits, while our experimental conditions (1350-1400 ºC; 1.5 GPa) likely represent upper- pressure/temperature limits. Our experiments confirm that Adirondack-class basalts are derived from a deep (~ 90–135 km) and primitive mantle reservoir.

194 5.3.2. Columbia Hills: alkali basalts In the Columbia Hills, Spirit analyzed the first alkali basalts on Mars. Irvine, Backstay and Humboldt Peak seem weakly affected by secondary processes and are usually referred to as alkaline volcanic rocks or alkali basalts (McSween et al., 2006a; Ming et al., 2008). They are aphanitic float rocks that likely contain olivine and pyroxene. Columbia Hills rocks share key characteristics with experimental melts produced at low degrees of melting (8 to 12 wt.%) and plot very close to them on projection diagrams (Fig. 6). However, they are not identical in composition to the experimental melts. The most striking difference is the lower CaO contents of alkaline rocks compared to our melts (Fig. 8). Significant chemical diversity is also observed among the rocks from the Columbia Hills. Strictly speaking, Irvine is a sub-alkaline basalt that resembles Adirondack-class basalts but is much richer in K2O, slightly richer in Na2O and poorer in CaO and Al2O3. Conversely, Backstay contains more Al2O3 and has a higher Mg# (53.3) than Adirondack-class basalts (50.1), Irvine (46.2) and Humboldt Peak (49.7). The lower CaO content of alkali basalts from Columbia Hills compared to our experimental melts could imply that they are not primary basalts and that their low CaO content was produced by crystallization of a mineral assemblage containing Ca-rich pyroxene. However, the high variability of their Al2O3 concentrations (8.5 to 13.5 wt.%) for a limited range of MgO contents (8.5 to 9.8 wt.%) challenges this scenario. Plagioclase must have appeared after different degrees of crystallization to produce residual liquids with variable Al2O3 contents. Irvine and Backstay are saturated with olivine on the liquidus (e.g. Nekvasil et al., 2009). Therefore, the crystallization of a large amount of plagioclase should be accompanied by significant olivine fractionation, which is inconstant with constant and elevated MgO contents. Invoking a refractory mantle source to account for the lower CaO content of Columbia Hills rocks is equally unsatisfactory. Our low- degree melts highlight that mantle sources affected by a previous melting event should be more depleted in Al2O3 than in CaO relative to the PMM and could not produce a melt with an Al2O3 content as high as that of Backstay. In addition, a depleted mantle will start to melt at a higher temperature and, as long as clinopyroxene is present, the CaO/MgO ratio is not expected to change significantly. If Humboldt Peak, Irvine and Backstay all represent liquid compositions, a more complex scenario is necessary to account for their low CaO content and their variable Al2O3 contents. Mantle reservoirs with distinct Al2O3 contents but similar CaO contents in close proximity are expected from the crystallization and overturn of a magma ocean (Debaille et al.,

195 2009; Elkins-Tanton et al., 2005; Scheinberg et al., 2014). Another possibility, which could be related to the former, is that the mantle source of Columbia Hills rocks was re-fertilized by variable amounts of low-degree melts. Finally, if Columbia Hills rocks represent cumulates rather than liquid compositions, the low CaO contents could reflect a late (post-cumulus) appearance of Ca- rich pyroxenes relative to plagioclase (Francis, 2011).

5.3.3. Gale crater: mugearite and other alkaline rocks In early 2013, with the first results obtained at Gale crater by Curiosity (Stolper et al., 2013) and the description of a new meteorite type – the regolith breccias 7034/7533 (see next section) – alkaline magmatism was shown to be ubiquitous on Mars and of greater importance than previously thought (McSween et al., 2009). Numerous rocks analyzed at Gale crater displayed extreme contents of alkali elements, particularly potassium. However, only a few samples have been argued to be of volcanic origin. Jake Matijevic (Jake_M), analyzed with both the APXS and ChemCam instruments, was described as the first martian mugearite. It is characterized by low MgO and FeO contents and high Al2O3 and alkali contents, and was interpreted as the product of extensive fractional crystallization of an alkali-basalt based on comparisons with terrestrial analogs (Stolper et al., 2013). Other potential volcanic rocks from Gale crater include the different rock types from the Bradbury Rise locality. These samples were only analyzed with the ChemCam instrument (laser-induced breakdown spectroscopy), and precise bulk rock compositions are not available. However, chemical analyses highlighted an important compositional diversity ranging from basaltic to evolved and alkali-rich compositions similar to Jake_M (Sautter et al., 2014; Schmidt et al., 2014). Alkali-basalts and tholeiitic basalts might coexist in the same locality, as also observed in Gusev Crater. Jake_M is significantly richer in alkalis compared to our experimental melts and also has a higher K2O/Na2O ratio. Both of these characteristics might point towards extremely low degrees of melting (2-3 wt.%). The lowest-degree melts that we were able to measure range between 5 and 8 wt.% of melting. Thus, experimental trends must be extrapolated in order to discuss the potential origin of Jake_M as a primary melt. Between 0.5 and 1.0 GPa, lower degree melts from the PMM would have lower FeO, MgO and CaO contents compared to Jake_M. On the other hand, higher pressure melts (~1.5 GPa) would have similar abundances of all the major elements except SiO2, and would also have much higher P2O5 and TiO2 contents. Jake_M is therefore more likely derived from melts that have undergone fractional crystallization. In order to maintain a high Al2O3 content,

196 the crystallization should have occurred at depth or in the presence of water, when the appearance of plagioclase is delayed (Stolper et al., 2013). The extreme alkali and Al2O3 contents of Jake_M are easier to produce from a low degree melt (~10 wt.%) produced at a pressure higher than 1.0 GPa. Metasomatism is often invoked to enrich the source of Jake_M in potassium (Schmidt et al., 2014; Stolper et al., 2013). Alternatively, we suggest that low-degree melts, produced from a mantle not anomalously enriched, can fractionate and eventually crystallize to form rocks similar to Jake_M. Experiments performed from the composition of Backstay illustrate how residual liquids rich in Al2O3 and poor in MgO can be derived from martian alkali basalts (Nekvasil et al., 2009).

5.3.4. NWA 7034/7533: martian crust-like meteorites The paired meteorites NWA 7034 and 7533 are regolith breccias that have been argued to be more representative of the bulk martian crust than shergottites (Agee et al., 2013). They have a chondritic CaO/Al2O3 ratio and are rich in alkali elements. The bulk-rock composition of NWA 7034 (Agee et al., 2013) and the composition of the basaltic clast VI from NWA 7533 (Humayun et al., 2013) are close to our low degree melts (Fig. 8). In particular, unlike alkali basalts from

Columbia Hills, they are not depleted in CaO and have lower K2O/Na2O ratios. A vitrophyric clast in NWA 7034 is closer in composition to Humphrey (Adirondack-class) and thus to our higher- degree melts (Udry et al., 2014). The average composition of basalt clasts analyzed by Santos et al. (2015) is also within experimental trends with the exception of TiO2, which is more than twice as abundant (2.3 wt.%). Other components of NWA 7034 exhibit low CaO/Al2O3 (~0.5) ratios for relatively high MgO contents compared to experimental melts (McCubbin et al., 2015). The complex texture of NWA 7034/7533 and the diversity of its components make it difficult to argue that it represents a primitive melt or even a silicate liquid. The two clasts displaying magmatic textures, the basaltic clast (VI) of Humayun et al. (2013) and the vitrophyre of Udry et al. (2014), are impact melts contaminated by a CI chondrite as suggested by their high Ni and Ir contents. However, the chondritic contaminant is unlikely to have disturbed the composition in major elements. If no fractional melting or crystallization were associated with the impact, magmatic clasts could represent the composition of their protoliths. Olivine must be added to the two clasts to bring their FeO/MgO ratios closer to those of experimental melts (Fig. 8). When ~8 wt.% of olivine is added to the composition of clast VI, it becomes almost identical to the melt of experiment DW10, which corresponds to a degree of melting of ~10 wt.% (1300 ºC) at 1.5 GPa.

197 When ~12 wt.% of olivine is added to the composition of the vitrophyre, it approaches the composition of another experimental melt (DW04), which corresponds to a degree of melting of ~30 wt.% (1400 ºC) at 1.5 GPa. The basaltic material that was incorporated in the regolith breccias might thus be derived from the PMM and represent near-primary melts formed at a depth of ~135 km.

5.3.5. Shergottites: high CaO/Al2O3 basalts Shergottites are the most abundant group of martian meteorites and the only group that contains olivine-bearing basalts. Some shergottites are believed to represent primitive melts (Gross et al., 2013; Gross et al., 2011; Musselwhite et al., 2006; Peslier et al., 2010). However, they exhibit a significant depletion in aluminum when compared to virtually all other analyses of martian rocks (McSween et al., 2009). They plot close to our melts produced at high degrees of melting (25-50 wt.%) on projection diagrams (Fig. 6). Shergottites also appear enriched in the clinopyroxene component and poorer in the plagioclase component relative to experimental melts. This reflects the high CaO/Al2O3 ratio that is characteristic of shergottite bulk-rock compositions. The low

Al2O3 content of shergottites seems in better agreement with the melting of a refractory mantle source. The melting is more likely to have occurred at high pressure (> 2 GPa) in order to produce the picritic compositions (i.e. high MgO contents) that are characteristic of shergottites at more realistic degrees of melting (< 30 wt.%).

The influence of water has been invoked to account for the relatively high SiO2 content of some shergottites (Balta and McSween, 2013). However, our experiments show that liquids rich in

MgO (15-20 wt.%) with 48 to 52 wt.% of SiO2 are produced under nominally anhydrous conditions (< 0.5 wt.%).

5.4. The Mg-number of martian mantle reservoirs

Most recent geochemical and geophysical models of the martian mantle converge towards a Mg# of 75-77 (Taylor, 2013). The starting composition used for our experiments is on the low end with a Mg# of 75 (Dreibus and Wanke, 1984). From 7 to 25 wt.% melting, melts have Mg# ranging from 50 to 55 and are in equilibrium with olivine containing 74 to 78 mol.% forsterite (Fo). Olivine megacrysts in LAR 06319 and NWA 1068, the most primitive enriched shergottites, have Fo contents that reach 77 mol.% (Basu Sarbadhikari et al., 2009). After 15 to 25 wt.% melting, the residual assemblages (Mg# 77-79) contain an olivine with Fo76-78, similar to the olivines that would

198 have initially crystallized from LAR 06319, NWA 1068 and the parental melt of the Adirondack basalts. This indicates that the mantle sources of enriched shergottites and Adirondack-class basalts might have a similar FeO/MgO ratio to that of the PMM.

The primitive depleted shergottites contain olivine as magnesian as Fo86 (Usui et al., 2008), which could be found in the residue of the PMM only after more than 50 wt.% melting. Such extensive melting is invoked for the production of komatiites on Earth but is unlikely for depleted shergottites due to their relatively young crystallization age. In addition, as discussed in the previous section, the high CaO/Al2O3 ratio of shergottites is not in agreement with our experimental melts. In order to produce melts with 18-20 wt.% of MgO in equilibrium with Fo86 olivine at a low degree of melting (8-15 wt.%) and at low pressure (1.2 GPa; Musselwhite et al., 2006), the mantle source must have a Mg# in the range 82-85. From the study of shergottites and Adirondack basalts, we expect mantle sources on Mars to vary in composition by at least 7 units of Mg# (75-82). To produce this level of heterogeneity from the PMM, parts of the mantle must have been affected by more than 50 wt.% melting during one or several events. However, such refractory reservoirs are expected to produce melts with lower concentration of incompatible elements relative to the primitive depleted shergottites (Gross et al., 2011; Usui et al., 2008). The crystallization of a magma ocean might more easily produce mantle sources with elevated Mg# (Elkins-Tanton et al., 2003).

6. Conclusion

Low degree melts produced from the primitive martian mantle between 0.5 and 2.2 GPa cover a wide compositional range from benmoreites and mugearites at low pressure to hawaiites and basanites at higher pressure. Melt compositions converge towards basalt at higher extents of melting. Due to its large diversity, the experimental dataset can be directly compared to most martian rocks that could represent primitive silicate liquid compositions. Among those, the Adirondack-class basalts and basaltic clasts from NWA 7034/7533 are close in composition to experimental trends and they can be inferred to derive from the PMM. However, other mantle reservoirs and the influence of fractional crystallization are necessary to retrace the crystallization history of other igneous rocks from the martian crust. The alkaline rocks from Columbia Hills are depleted in CaO relative to melts from the PMM, and we suspect that Al2O3 is not distributed homogenously in their mantle source. Jake_M might be derived from an alkali-rich melt from the

199 PMM but fractional crystallization must have occurred to explain its extreme K2O content.

Shergottites have higher CaO/Al2O3 compared to melts from the PMM, and the depleted sub-group must derive from a more magnesian source with a Mg# of 82-84. Experimental melts from a nominally dry PMM cannot account for the composition of the entire collection of martian basalts and alkaline rocks, but rather represent a benchmark for the study of other igneous processes and mantle heterogeneities.

7. References

Agee, C.B., Draper, D.S., 2004. Experimental constraints on the origin of Martian meteorites and the composition of the Martian mantle. Earth Planet. Sci. Lett. 224, 415-429. Agee, C.B., et al., 2013. Unique Meteorite from Early Amazonian Mars: Water-Rich Basaltic Breccia Northwest Africa 7034. Science 339, 780-785. Albarede, F., 2009. Volatile accretion history of the terrestrial planets and dynamic implications. Nature 461, 1227-1233. Baker, M.B., Stolper, E.M., 1994. Determining the composition of high-pressure mantle melts using diamond aggregates. Geochim. Cosmochim. Acta 58, 2811-2827. Balta, J.B., McSween, H.Y., 2013. Water and the composition of Martian magmas. Geology doi:10.1130/G34714.1. Baratoux, D., Toplis, M.J., Monnereau, M., Gasnault, O., 2011. Thermal history of Mars inferred from orbital geochemistry of volcanic provinces. Nature 472, 338-341. Barrat, J.A., Jambon, A., Bohn, M., Gillet, P., Sautter, V., Gopel, C., Lesourd, M., Keller, F., 2002. Petrology and chemistry of the picritic shergottite North West Africa 1068 (NWA 1068). Geochim. Cosmochim. Acta 66, 3505-3518. Basu Sarbadhikari, A., Day, J.M.D., Liu, Y., Rumble, D., Taylor, L.A., 2009. Petrogenesis of olivine- phyric shergottite Larkman Nunatak 06319: Implications for enriched components in martian basalts. Geochim. Cosmochim. Acta 73, 2190-2214. Bertka, C.M., Fei, Y., 1997. Mineralogy of the Martian interior up to core-mantle boundary pressures. J. Geophys. Res. 102, 5251-5264. Bertka, C.M., Holloway, J.R., 1993. Pigeonite at solidus temperatures: Implications for partial melting. J. Geophys. Res.: Solid Earth 98, 19755-19766. Bertka, C.M., Holloway, J.R., 1994a. Anhydrous partial melting of an iron-rich mantle 1: Subsolidus phase assemblages and partial melting phase-relations at 10 to 30 kbar. Contrib. Mineral. Petrol. 115, 313-322. Bertka, C.M., Holloway, J.R., 1994b. Anhydrous partial melting of an iron-rich mantle 2: Primary melt compositions at 15 kbar. Contrib. Mineral. Petrol. 115, 323-338. Breuer, D., Spohn, T., 2006. Viscosity of the Martian mantle and its initial temperature: Constraints from crust formation history and the evolution of the magnetic field. Planet. Space Sci. 54, 153-169. Carter, J., Poulet, F., 2013. Ancient plutonic processes on Mars inferred from the detection of possible anorthositic terrains. Nature Geosci 6, 1008-1012. Christensen, P.R., et al., 2005. Evidence for magmatic evolution and diversity on Mars from infrared observations. Nature 436, 504-509. Debaille, V., Brandon, A.D., O/'Neill, C., Yin, Q.Z., Jacobsen, B., 2009. Early martian mantle overturn inferred from isotopic composition of nakhlite meteorites. Nature Geosci 2, 548-552. Dreibus, G., Wanke, H., 1984. Accretion of the Earth and the inner planets. Proc. 27th International Geological Congress 11, 1-20. Dreibus, G., Wanke, H., 1985. Mars, a volatile-rich planet. Meteoritics 20, 367-381.

200 Elkins-Tanton, L.T., Parmentier, E.M., Hess, P.C., 2003. Magma ocean fractional crystallization and cumulate overturn in terrestrial planets: Implications for Mars. Meteorit. Planet. Sci. 38, 1753- 1771. Elkins-Tanton, L.T., Zaranek, S.E., Parmentier, E.M., Hess, P.C., 2005. Early magnetic field and magmatic activity on Mars from magma ocean cumulate overturn. Earth Planet. Sci. Lett. 236, 1- 12. Filiberto, J., Dasgupta, R., 2011. Fe2+-Mg partitioning between olivine and basaltic melts: Applications to genesis of olivine-phyric shergottites and conditions of melting in the Martian interior. Earth and Planetary Science Letters 304, 527-537. Foley, C.N., Economou, T., Clayton, R.N., 2003. Final chemical results from the Mars Pathfinder alpha proton X-ray spectrometer. J. of Geophys. Res.: Planets 108, 8096. Francis, D., 2011. Columbia Hills — An exhumed layered igneous intrusion on Mars? Earth and Planetary Science Letters 310, 59-64. Gaetani, G.A., 1998. Igneous Petrology. Science 282, 1834-1835. Gaetani, G.A., Grove, T.L., 1998. The influence of water on melting of mantle peridotite. Contrib. Mineral. Petrol. 131, 323-346. Gross, J., Filiberto, J., Herd, C.D.K., Daswani, M.M., Schwenzer, S.P., Treiman, A.H., 2013. Petrography, mineral chemistry, and crystallization history of olivine-phyric shergottite NWA 6234: A new melt composition. Meteorit. Planet. Sci. 48, 854-871. Gross, J., Treiman, A.H., Filiberto, J., Herd, C.D.K., 2011. Primitive olivine-phyric shergottite NWA 5789: Petrography, mineral chemistry, and cooling history imply a magma similar to Yamato- 980459. Meteorit. Planet. Sci. 46, 116-133. Grove, T., 1993. Corrections to expressions for calculating mineral components in “Origin of calc-alkaline series lavas at medicine lake volcano by fractionation, assimilation and mixing” and “Experimental petrology of normal MORB near the kane fracture zone: 22°-25°N, mid-atlantic ridge”. Contrib. Mineral. Petrol. 114, 422-424. Halliday, A.N., Porcelli, D., 2001. In search of lost planets – the paleocosmochemistry of the inner solar system. Earth Planet. Sci. Lett. 192, 545-559. Hays, J.F., 1966. Stability and Properties of the Synthetic Pyroxene CaAl2SiO6. Am. Mineral. 51, 1524- 1529. Hirschmann, M.M., 2000. Mantle solidus: Experimental constraints and the effects of peridotite composition. Geochem. Geophy. Geosy. 1, 1042. Hirschmann, M.M., Baker, M.B., Stolper, E.M., 1998. The Effect of Alkalis on the Silica Content of Mantle-Derived Melts - NMR, Raman, and infrared spectroscopy. Geochim. Cosmochim. Acta 62, 883-902. Holloway, J.R., Pan, V., Gudmundsson, G., 1992. High-Pressure Fluid-Absent Melting Experiments in the Presence of Graphite - Oxygen Fugacity, Ferric Ferrous Ratio and Dissolved CO2. Eur. J. Mineral. 4, 105-114. Humayun, M., et al., 2013. Origin and age of the earliest Martian crust from meteorite NWA 7533. Nature 503, 513-516. Khan, A., Connolly, J.A.D., 2008. Constraining the composition and thermal state of Mars from inversion of geophysical data. J. Geophys. Res. 113, E07003. Kinzler, R.J., 1997. Melting of mantle peridotite at pressures approaching the spinel to garnet transition: Application to mid-ocean ridge basalt petrogenesis. J. of Geophys. Res.: Solid Earth 102, 853-874. Kinzler, R.J., Grove, T.L., 1992. Primary Magmas of Mid-Ocean Ridge Basalts 1. Experiments and Methods. J. Geophys. Res. 97, 6885-6906. Krawczynski, M.J., Olive, J.L., 2011. A new fitting algorithm for petrological mass-balance problems. American Geophysical Union, Fall Meeting 2011, abstract #V53B-2613. Laporte, D., Toplis, M., Seyler, M., Devidal, J.-L., 2004. A new experimental technique for extracting liquids from peridotite at very low degrees of melting: application to partial melting of depleted peridotite. Contrib. Mineral. and Petrol. 146, 463-484.

201 Lodders, K., Fegley Jr, B., 1997. An Oxygen Isotope Model for the Composition of Mars. Icarus 126, 373-394. Matsukage, K.N., Nagayo, Y., Whitaker, M.L., Takahashi, E., Kawasaki, T., 2013. Melting of the Martian mantle from 1.0 to 4.5 GPa. J. Mineral. Petrol. Sci. 108, 201-214. McCubbin, F.M., Hauri, E.H., Elardo, S.M., Vander Kaaden, K.E., Wang, J., Shearer, C.K., 2012. Hydrous melting of the martian mantle produced both depleted and enriched shergottites. Geology 10.1130/G33242.1. McCubbin, F.M., et al., 2015. Composition of fine-grained bulk matrix and protobreccia clast matrix in Northwest Africa 7034: Implications for the composition of the martian crust. Lunar and Planet. Sci. Conference 46th, 1723. McSween, H.Y., Grove, T.L., Wyatt, M.B., 2003. Constraints on the composition and petrogenesis of the Martian crust. J. of Geophys. Res.: Planets 108, 5135. McSween, H.Y., et al., 2006a. Alkaline volcanic rocks from the Columbia Hills, Gusev crater, Mars. J. of Geophys. Res.: Planets 111, E09S91. McSween, H.Y., et al., 2006b. Characterization and petrologic interpretation of olivine-rich basalts at Gusev Crater, Mars. J. Geophys. Res.: Planets 111, E02S10. McSween, H.Y., Taylor, G.J., Wyatt, M.B., 2009. Elemental Composition of the Martian Crust. Science 324, 736-739. Médard, E., McCammon, C.A., Barr, J.A., Grove, T.L., 2008. Oxygen fugacity, temperature reproducibility, and H2O contents of nominally anhydrous piston-cylinder experiments using graphite capsules. Am. Mineral. 93, 1838-1844. Médard, E., Schmidt, M., Schiano, P., 2004. Liquidus surfaces of ultracalcic primitive melts: formation conditions and sources. Contrib. Mineral. Petrol. 148, 201-215. Ming, D.W., et al., 2008. Geochemical properties of rocks and soils in Gusev Crater, Mars: Results of the Alpha Particle X-Ray Spectrometer from Cumberland Ridge to Home Plate. J. Geophys. Res.: Planets 113, E12S39. Misawa, K., 2004. The Yamato 980459 olivine-phyric shergottite consortium. Antartic Meteorite Res. 17, 1-12. Monders, A.G., Médard, E., Grove, T.L., 2007. Phase equilibrium investigations of the Adirondack class basalts from the Gusev plains, Gusev crater, Mars. Meteorit. Planet. Sci. 42, 131-148. Morgan, J.W., Anders, E., 1979. Chemical composition of Mars. Geochim. Cosmochim. Acta. 43, 1601- 1610. Musselwhite, D.S., Dalton, H.A., Kiefer, W.S., Treiman, A.H., 2006. Experimental petrology of the basaltic shergottite Yamato-980459: Implications for the thermal structure of the Martian mantle. Meteorit. Planet. Sci. 41, 1271-1290. Nekvasil, H., McCubbin, F.M., Harrington, A., Elardo, S., Lindsley, D.H., 2009. Linking the Chassigny meteorite and the Martian surface rock Backstay: Insights into igneous crustal differentiation processes on Mars. Meteorit. Planet. Sci. 44, 853-869. O'Hara, M.J., 1968. The bearing of phase equilibria studies in synthetic and natural systems on the origin and evolution of basic and ultrabasic rocks. Earth-Science Reviews 4, 69-133. Peslier, A.H., Hnatyshin, D., Herd, C.D.K., Walton, E.L., Brandon, A.D., Lapen, T.J., Shafer, J.T., 2010. Crystallization, melt inclusion, and redox history of a Martian meteorite: Olivine-phyric shergottite Larkman Nunatak 06319. Geochim. Cosmochim. Acta 74, 4543-4576. Plesa, A.C., Tosi, N., Breuer, D., 2014. Can a fractionally crystallized magma ocean explain the thermochemical evolution of Mars? Earth Planet. Sci. Lett. 403, 225-235. Pouchou, J.L., Pichoir, F., 1984. New model for quantitative x-ray microanalysis. Part I: Application to the analysis of homogeneous samples. Recherche Aerospatiale (English Edition), 13-38. Rogers, A.D., Nekvasil, H., 2015. Feldspathic rocks on Mars: Compositional constraints from infrared spectroscopy and possible formation mechanisms. Geophy. Res. Lett., 2015GL063501.

202 Sack, R., Ghiorso, M., 1989. Importance of considerations of mixing properties in establishing an internally consistent thermodynamic database: thermochemistry of minerals in the system Mg2SiO4-Fe2SiO4-SiO2. Contrib. Mineral. Petrol. 102, 41-68. Sanloup, C., Jambon, A., Gillet, P., 1999. A simple chondritic model of Mars. Phys. Earth Planet. In. 112, 43-54. Santos, A.R., Agee, C.B., McCubbin, F.M., Shearer, C.K., Burger, P.V., Tartèse, R., Anand, M., 2015. Petrology of igneous clasts in Northwest Africa 7034: Implications for the petrologic diversity of the martian crust. Geochim. Cosmochim. Acta. 157, 56-85. Sautter, V., et al., 2014. Igneous mineralogy at Bradbury Rise: The first ChemCam campaign at Gale crater. J. of Geophy. Res.: Planets 119, 2013JE004472. Scheinberg, A., Elkins-Tanton, L.T., Zhong, S.J., 2014. Timescale and morphology of Martian mantle overturn immediately following magma ocean solidification. J. of Geophys. Res.: Planets 119, 2013JE004496. Schmidt, M.E., et al., 2014. Geochemical diversity in first rocks examined by the Curiosity Rover in Gale Crater: Evidence for and significance of an alkali and volatile-rich igneous source. J. of Geophys. Res.: Planets 119, 64-81. Stolper, E.M., et al., 2013. The Petrochemistry of Jake_M: A Martian Mugearite. Science 341, 1239463. Takahashi, E., Kushiro, I., 1983. Melting of a dry peridotite at high pressures and basalt magma genesis. Am. Mineral. 68, 859-879. Taylor, G.J., 2013. The bulk composition of Mars. Chemie der Erde - Geochemistry 73, 401-420. Toplis, M.J., 2005. The thermodynamics of iron and magnesium partitioning between olivine and liquid: criteria for assessing and predicting equilibrium in natural and experimental systems. Contrib. Mineral. Petrol. 149, 22-39. Udry, A., Lunning, N.G., McSween Jr, H.Y., Bodnar, R.J., 2014. Petrogenesis of a vitrophyre in the martian meteorite breccia NWA 7034. Geochim. Cosmochim. Acta 141, 281-293. Usui, T., McSween, H.Y., Floss, C., 2008. Petrogenesis of olivine-phyric shergottite Yamato 980459, revisited. Geochim. Cosmochim. Acta 72, 1711-1730. Wasylenki, L.E., Baker, M.B., Kent, A.J.R., Stolper, E.M., 2003. Near-solidus Melting of the Shallow Upper Mantle: Partial Melting Experiments on Depleted Peridotite. J. Petrol. 44, 1163-1191. Wray, J.J., Hansen, S.T., Dufek, J., Swayze, G.A., Murchie, S.L., Seelos, F.P., Skok, J.R., Irwin Iii, R.P., Ghiorso, M.S., 2013. Prolonged magmatic activity on Mars inferred from the detection of felsic rocks. Nature Geosci 6, 1013-1017. Zipfel, J., et al., 2011. Bounce Rock—A shergottite-like basalt encountered at Meridiani Planum, Mars. Meteorit. Planet. Sci. 46, 1-2.

203 Figures and tables

Figure 1. Back-scattered electron images of experimental run products. A. Experiment DW 17, performed at 1.5 GPa, 1270 ºC. The liquid is extracted in a layer of vitreous carbon spheres. B. Experiment DW 31, performed at 2 GPa, 1350 ºC. The liquid is extracted into a micro-crack in the lid of the graphite capsule.

orthopyroxene 1350 clinopyroxene

1300 1.0 GPa

1250 Temperature (º C) (º Temperature

1200 0.5 GPa

5 10 15 20 25 30 Wollastonite (mol%) Figure 2. Composition range of pyroxene at 0.5 and 1.0 GPa and possible evolution of the pyroxene phase diagram. The pigeonite-augite solvus is reduced at 1.0 GPa and becomes metastable at higher pressures. Horizontal bars represent the dispersion of microprobe analyses for each phase. Vertical bars represent the ±10 ºC temperature reproducibility. The central symbol is the average as reported in supplemental Table 2. The wollastonite content is Ca/(Ca+Fe+Mg)×100 in moles.

204

Figure 3. Phase diagram based on experimental results. Solid lines represent the solidus and phase boundaries. The dot-dash line is the solidus for terrestrial peridotite (Hirschmann, 2000). Melt fractions are calculated by linear regression. Experiment DW 33 (in grey, 1450ºC – 1.5 GPa) was equilibrated at a higher temperature (~1500 ºC) than the nominal temperature as suggested by its high MgO content.

205

10 0.5 GPa 1.0 GPa Benmoreite 1.5 GPa 8 Mugearite 2.0 GPa Hawaiite Tephrite 2.2 GPa Basanite sh. Shergottites 6 NWA 7034 ba. sp. Spirit rocks/soils ad. Adirondack 4 Basalt sp. Humboldt Peak andesite

Na 2 O + K (wt %) Backstay ad. bo. Basaltic bo. Bounce rock 2 andesite ba. Bathrust Inlet sh. Jake_M 0 Rocknest 40 45 50 55 60 SiO2 content (wt %)

Figure 4. Total Alkali-Silica (TAS) diagram comparing experimental melts with martian rocks. Modified from Stolper et al. (2013) and Schmidt et al. (2014). The solid grey line separates alkaline (up) and sub-alkaline (down) fields. Dotted lines are visual aids illustrating the compositional trends at each pressure.

206 20

9 16

2.2 12 1.5 -2.0 2.2 7 2.0 8 1.5 MgO (wt.%) 0.5 GPa CaO (wt.%) 1.0 1.0 1.0 GPa 1.5 GPa 4 0.5 0.5 2.0 GPa 5 2.2 GPa

1.0 58 0.5 16 0.5 1.5 54 1.0 2.0 (wt.%)

(wt.%) 12 3 2 2.2 SiO 50 Al 2 O

1.5 8 46 2.0

2.2 4 10 20 30 40 50 10 20 30 40 50 Melt fraction (wt.%) Melt fraction (wt.%) Figure 5. Evolution of the key major elements in experimental melts as a function of melt fraction andpressure. Lines are simple visual aids illustrating compositional trends at each pressure. Vertical and horizontal bars represent one standard deviation uncertainties.

207 2.2

1.5 2.0

1.0 Plag 60 80 Ol 1.5 1.0 Primitive martian 2.0 0.5 2.2 mantle

80 20 2.2 GPa melts 0.5 2.0 GPa melts 1.5 GPa melts 1.0 GPa melts 40 60 40 0.5 GPa melts

ol + melt 60 40 60 ol + opx + melt ol + opx + pig + sp + melt ol + opx + pig + sp + plag +melt

80 20 80 NWA 7533 - clast VI Gusev basalts NWA 7034 - Basalt. c. Adk class (3), Irvine NWA 7034 - bulk Backstay, Humboldt P. Bounce Shergottite 20 40 60 80 Jake_M Cpx Qz melts Figure 6. Projection diagrams in the clinopyroxene-plagioclase-quartz-olivine tetrahedron after Grove (1993). The composition of Martian rocks are from Agee et al. (2013), Barrat et al. (2002), Basu Sarbadhikari et al. (2009), Gross et al. (2011), Gross et al. (2013), Humayun et al. (2013), McSween et al. (2006b), Ming et al. (2008), Misawa (2004), Santos et al. (2015), Stolper et al. (2013) and Zipfel et al. (2011).

208 Temperature (ºC) 0.5 GPa 60 ol

40 aug-pig cpx out cpx opx 20 Melt fraction (wt.%) cpx liq 0 1150 1250 1350 1.0 GPa 60 ol

40 opx aug-pig cpx out cpx

20 Melt fraction (wt.%) cpx liq 0 1200 1300 1400 1.5 GPa 60 ol

40 cpx out cpx opx

Melt fraction (wt.%) 20 cpx liq 0 1250 1350 1450 2.0 GPa 60 ol

40 cpx out cpx opx 20 Melt fraction (wt.%) cpx liq 0 1350 1450 1550 Temperature (ºC)

Figure 7. Evolution of phase proportions with temperature: melts (open circles), clinopyroxene (triangles), orthopyroxene (squares), olivine (diamonds). Spinel is not plotted due to its low proportions (< 2 wt.%). The dashed lines represents the temperatures of sub-calcic augite to pigeonite transition and clinopyroxene disappearance. Vertical error bars are 1 s of mass balance calculations. Horizontal error bars are the ± 10 ºC uncertainties on experimental temperatures.

209 MgO (wt.%) MgO (wt.%) 4 6 8 10 12 14 16 18 20 4 6 8 10 12 14 16 18 20 60 18 1 0.5 .0 -2. 2 0.5 56 14 1.0 52 10 (wt.%) 3

48 O SiO 2 (wt.%) 1.5 Al 2 6 2.0 44 2.2 2

2.0 20 10

0.5 9 16 8 0 . .2 2 2 .5 7 1

FeO(wt.%) 12 CaO (wt.%) CaO 6

1.0

5 . 8 5 0

1 1 0. 8 0 . . 0 5- 5 1 1.0 .5 2 . 2 0.8 0 .2 1 .5 2 .0 6 -2 0.6 .2 2 (wt.%) 4 TiO 0.4 Na 2 O +K (wt.%) 2 0.2

Melts 0.3 2. 2.5 1 2 . 5 2.2 GPa 1 2. 2 .0 2 0 . . 0 0 2 2 2.0 GPa . 1.0 1 5 .5 0.2 1.5 GPa 0 1.5 .5 1.0 GPa 0.5 GPa P 2 O 5 (wt. %) 1

K 2 O/Na O (wt.%) 0.1

0.5

2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20 +5% +10% olivine MgO (wt.%) MgO (wt.%) NWA 7034 BC Jake_M Humboldt Peak Adirondack NWA 7034 B NWA 7034 PB Backstay Irvine NWA 7533 clast VI NWA 7034 M NWA 7034 V

210 Figure 8. (previous page) Variation diagrams of experimental melts and key martian igneous rocks. Color lines represent Martian rocks with the left extremities being the bulk rock compositions, the intermediate lines being the bulk rock + 5% of olivine and the right extremities the bulk rock +10% of olivine. Equilibrium olivine was added assuming a KD Fe-Mg of 0.35 (Filiberto and Dasgupta, 2011; this study) and increments of 0.1 wt.%. NWA 7034 M and PB correspond to the composition of the Matrix and Proto-Breccia clasts from McCubbin et al. (2015). NWA 7034 V is the vitrophyre from Udry et al. (2014). NWA BC 7034 is the average of basaltic clasts from Santos et al. (2015). The TiO2 content (2.3 wt.%) is out of the field. The compositions of other martian rocks are from Agee et al. (2013), Humayun et al. (2013), McSween et al. (2006b), Ming et al. (2008) and Stolper et al. (2013).

Table 1. Model compositions of the primitive martian mantle and starting composition for experiments.

DW 84*,a LF 97b BF 97c S 99d KC 08e

SiO2 44.4 45.4 43.7 47.5 44

TiO2 0.14 0.14 - 0.1 -

Al2O3 3.02 2.89 3.13 2.5 2.5

Cr2O3 0.76 0.68 - 0.7 - FeO 17.9 17.2 18.7 17.7 17 MnO 0.46 0.37 - 0.4 - MgO 30.2 29.7 31.5 27.3 33 CaO 2.45 2.35 2.49 2 2.2

Na2O 0.50 0.98 0.5 1.2 -

K2O 0.04 0.11 - - -

P2O5 0.16 0.17 - - -

Mg# 75.0 75.5 75.0 73.3 77.6 Total 100.03 100.01 100.01 99.4 98.7 aDreibus and Wanke (1985), bLodders and Fegley Jr (1997), cBertka and Fei (1997), dSanloup et al. (1999), eKhan and Connolly (2008)

211

Table 2. Experimental melt compositions

Exp # T P n SiO2 TiO2 Al2O3 Cr2O3 FeO MnO MgO CaO Na2O K2O P2O5 Sum KD KD (T) DW35 1350 0.5 6 50.3(4) 0.30(4) 7.9(2) 0.80(5) 17.1(1) 0.53(2) 14.8(3) 6.16(9) 1.53(9) 0.15(1) 0.41(7) 99.5 0.37 0.34 DW27 1300 13 51.4(5) 0.39(6) 9.7(2) 0.58(7) 15.7(2) 0.46(5) 11.6(4) 7.4(3) 2.0(3) 0.24(3) 0.47(8) 99.1 0.35 0.34 DW26 1250 15 52.3(3) 0.49(5) 11.9(2) 0.27(5) 13.7(3) 0.37(4) 8.9(2) 8.0(2) 3.1(2) 0.27(4) 0.61(8) 98.9 0.33 0.33 DW28 1200 14 54.6(9) 0.57(6) 14.5(7) 0.11(4) 10.0(1.0) 0.26(7) 6.2(7) 8.1(3) 4.3(3) 0.46(7) 0.92(23) 98.6 0.33 0.32 DW34 1160 7 59.0(8) 0.73(4) 15.1(2) 0.03(3) 7.3(4) 0.21(3) 3.93(12) 5.6(3) 5.7(2) 1.0(1) 1.2(2) 100.3 0.30 0.35 DW16* 1450 1.0 6 48.9(1.0) 0.25(8) 5.5(3) 1.0(1) 20.1(4) 0.58(4) 17.2(1.6) 5.0(4) 0.98(6) 0.11(4) 0.31(5) 99.5 0.30 0.34 DW09 1385 7 49.7(3) 0.35(6) 8.36(8) 0.69(6) 17.8(2) 0.47(7) 13.64(16) 6.9(1) 1.50(7) 0.19(2) 0.42(5) 99.0 0.35 0.35 DW03 1350 8 49.4(2) 0.45(4) 9.98(6) 0.40(4) 16.8(2) 0.45(3) 11.64(8) 8.0(1) 2.05(9) 0.17(2) 0.54(4) 98.9 0.35 0.34 DW02 1300 8 52.5(6) 0.45(7) 12.0(1) 0.24(8) 13.2(3) 0.37(4) 8.64(10) 8.93(10) 2.79(8) 0.26(6) 0.61(9) 99.0 0.35 0.35 DW01 1250 10 52.3(4) 0.57(5) 14.4(3) 0.09(4) 11.5(2) 0.31(6) 6.9(2) 8.2(2) 4.4(3) 0.44(4) 0.88(6) 99.3 0.33 0.32 B1330 1240 6 52.1(5) 0.62(4) 16.3(2) 0.08(4) 10.1(3) 0.31(2) 5.69(15) 8.00(15) 5.25(16) 0.53(3) 0.98(9) 99.0 0.30 0.31 DW07 1220 8 53.0(5) 0.70(2) 15.9(2) 0.07(2) 9.4(1) 0.28(4) 6.19(9) 7.47(13) 4.67(12) 0.75(3) 1.55(18) 99.1 0.38 0.32 DW18 1200 5 53.2(7) 0.67(4) 17.02(14) 0.04(1) 8.8(3) 0.23(3) 5.3(3) 6.3(2) 5.73(10) 0.95(10) 1.72(3) 99.5 0.36 0.31 DW33* 1450 1.5 8 47.6(3) 0.24(3) 5.9(2) 0.86(5) 20.1(2) 0.62(2) 18.1(7) 5.1(3) 1.17(10) 0.10(1) 0.23(4) 99.4 0.34 0.34

212 DW04* 1400 18 47.8(7) 0.37(3) 8.4(4) 0.63(9) 19.3(7) 0.53(10) 14.0(1.0) 6.9(5) 1.5(5) 0.20(7) 0.46(9) 99.5 0.35 0.34 DW14 1350 5 47.4(2) 0.56(2) 10.27(13) 0.34(1) 17.1(3) 0.43(3) 11.35(5) 8.85(10) 2.39(5) 0.32(2) 0.75(12) 98.7 0.34 0.33

DW10 1300 8 47.4(1.0) 0.76(8) 12.8(3) 0.16(4) 14.7(2) 0.39(3) 9.12(12) 8.75(11) 4.2(2) 0.51(4) 1.18(10) 98.8 0.34 0.32 DW17 1270 5 47.1(4) 0.94(6) 14.56(13) 0.08(3) 13.7(3) 0.35(4) 7.8(4) 7.6(1) 5.06(9) 0.90(5) 2.0(2) 99.0 0.33 0.31 DW25* 1550 2.0 20 47.4(3) 0.27(6) 5.8(2) 0.86(7) 20.0(4) 0.53(5) 18.6(3) 5.0(2) 1.1(3) 0.13(4) 0.30(6) 99.1 0.35 0.35 DW23* 1500 12 47.1(5) 0.33(2) 6.9(3) 0.74(6) 20.2(7) 0.51(8) 16.4(8) 5.8(4) 1.4(3) 0.15(2) 0.40(5) 98.5 0.34 0.35 DW20 1450 9 45.9(6) 0.45(2) 8.5(3) 0.53(8) 19.4(2) 0.52(4) 14.4(5) 7.7(2) 1.8(2) 0.22(4) 0.57(6) 99.7 0.35 0.34 DW29 1375 8 46.4(7) 0.77(4) 10.8(1) 0.27(5) 16.6(3) 0.43(3) 10.8(1) 8.7(2) 3.61(15) 0.48(3) 1.2(1) 98.2 0.35 0.33 DW31 1350 8 45.2(8) 0.99(2) 11.6(3) 0.20(2) 16.8(2) 0.49(3) 10.1(3) 8.19(8) 4.0(2) 0.72(3) 1.63(9) 99.9 0.34 0.32 C556 1400 2.2 3 44.5(6) 0.63(5) 10.5(3) 0.24(2) 17.9(2) 0.52(3) 11.7(6) 8.8(2) 3.20(7) 0.44(2) 1.48(4) 99.9 0.34 0.33 C547 1350 4 43.7 (1.0) 1.1(2) 11.1(3) 0.14(4) 16.(3) 0.46(2) 10.6(6) 8.2(2) 4.1(4) 0.96(3) 2.8(1) 98.7 0.36 0.32 Temperature (T) is in ºC and Pressure (P) is in GPa. All oxide concentrations are in wt.% and compositions are normalized to 100. Sum is the original un-

normalized EPMA average total. The digits in brackets represent one standard deviation from the set of analyses averaged for each experiment. n is the number of

EPMA analyses averaged. KD and KD (T) are experimental and modeled (Toplis, 2005) Fe-Mg exchange coefficients between olivine and melt. KD’s are

calculated with the assumptions that experimental melts are anhydrous and have a Fe3+/Fetot ratio of 0.02. *liquids analyzed as a mix of glass and quench

dendrites

212 Supplementary Material

0.5 A 0.5 GPa 0.45 1.0 GPa 1.5 GPa 0.4 2.0 GPa aSiO 2 0.35

0.3

0.9 γ SiO 2 0.8

0.7

1100 1200 1300 1400 1500 1600 Temperature (ºC)

Supplemental Figure 1. Silica activities (A) and silica activity coefficients (B) of experimental melts.

Silica activities are calculated from the equation:

Mg2Si2O6 (in opx) = Mg2SiO4 (in olivine) + SiO2 (in liquid)

+/ /14 & &,+,$ &,+/ &,/14 :.+ × :0123 ΔG#$% = µ)% − µ.+ − µ0123 = 56 ln 9 +,$ < :)%

/14 /14 /14 :0123 = =0123>0123

&,+,$ &,+/ &,/14 +/ +,$ /14 With µ)% , µ.+ , µ0123, :.+, :)% and >0123 calculated from the Melts on-line supplemental calculator

(Ghiorso and Sack, 1995).

213 5 10 15 20 5 10 15 20 60 18

16 56 14

52 12 2 O 3 (wt.%) Al SiO 2 (wt.%) 48 10 8 44 6

20 12

10 16 8

0.5 (wt.%) CaO FeO (wt.%) FeO 12 BH /1.5 1.0 6 M /1.0 1.5 M /2.5 2.0 8 2.2 4

5 10 15 20 5 10 15 20 MgO (wt.%) MgO (wt.%)

Supplemental Figure 2. Comparison of experimental melts (0.5-2.2 GPa) with previous studies performed from the composition of Dreibus and Wanke (1985). BH/1.5 are experiments by Bertka and Holloway (1994) performed at 1.5 GPa. M/1.0 and M/2.5 are experiments performed by Matsukage et al. (2013) at 1.0 and 2.5 GPa, respectively. The experiment from BH with the lower MgO content (JFR4) is a sandwich experiment while all the others are direct melting experiments.

214 The high SiO2 concentrations of our low-degree experimental melt at 0.5 and 1.0 GPa (Supplemental figure 2) reflect the decrease in the activity coefficient of SiO2 (Supplemental figure 1), At low pressure and low degree of melting, high concentrations of alkalis depolymerize the silicate melts by breaking Si-

O-Si bonds and decreasing the silica activity coefficient, thus increasing the SiO2 concentrations (Laporte et al., 2014; Hirschmann et al., 1998). Melts obtained by Bertka and Holloway (1994) in K-free experiments contain less alkalis and have lower silica contents. A slightly higher water contamination in our experiments (0.5-1.0 wt.%) cannot be completely ruled out and could contribute to lower activity coefficients (Gaetani and Grove, 1998). This effect is more limited at high pressure (> 1.5 GPa) probably due to change in melt structure and/or alkali/silica interactions (Hirschmann et al. 1998). On the contrary, P is a network former that increases the degree of polymerization of melts and thus the activity coefficient of silica (Toplis et al., 1994), which in turn should decrease the silica concentration. However, this effect is not clearly visible in our experiments and may be less important than previously suggested.

Supplemental Figure 3. Summary of experiments performed from the composition of the primitive martian mantle and in which the silicate melt was analyzed. The black symbols represent experiments from the literature: Matsukage 13 (Matsukage et al., 2013), BH 94 (Bertka and Holloway, 1994), AD 04 (Agee and Draper, 2004). The black dotted lines are the first and second order polynomial fit of the solidus. The solid line is the second order polynomial modified to preserve the 50 ºC difference with the solidus of terrestrial peridotite (Hirschmann, 2000) at higher pressure; Ts (ºC) = 1033 + 168.1*P (GPa) – 14.22*P^2 (GPa). The red dashed line is the polynomial fit of (Filiberto and Dasgupta, 2011) based on BH 94 and AD 04 experiments.

215 20 2.0 5 1. 9 16 .5 1.0 0

.0 .5 1 12 0 1.5 7 2.0 8 MgO (wt.%) 0.5 GPa CaO (wt.%) 1.0 GPa 1.5 GPa 4 2.0 GPa 5

58 16 0. 5

54 0.5 1.0 (wt.%)

(wt.%) 12 3 2 1.5 SiO 50 Al 2 O 1.0 2.0 8 46 1.5

4 2.0 24

1.5 1500 2.0 1.5 1.0 20 1400

(¼C) 1.0

0.5 0.5 16 1300 FeO (wt.%)

12 Temperature 1200

8 1100

10 20 30 40 50 10 20 30 40 50 Melt fraction (wt.%) Melt fraction (wt.%) Supplemental Figure 4. Comparison of experimental melts with pMELTS simulations (lines) at 0.5 to 2 GPa and FMQ-3. pMELTS produces major offsets by overestimating FeO, MgO and underestimating

SiO2 and Al2O3. Offsets are small at 0.5 GPa but increase with the increasing pressure.

216 Supplemental Table 1. Experimental conditions and phase proportion

Exp # P (GPa) T (ºC) t (h) Phase proportions (wt%) melt 1 σ olivine 1 σ opx 1 σ pig 1 σ sp 1 σ plag 1 σ RSS ΔFe DW35 0.5 1350 18 41.4 0.9 53.1 0.8 5.7 1.2 ------0.44 -5.9 DW27 0.5 1300 30 29.3 0.5 57.9 0.5 12.3 0.8 - - 0.4 - - - 0.12 -2.2 DW26 0.5 1250 30 22.7 1.3 58.9 1.5 17.6 2.3 - - s - - - 0.29 -3.1 DW28 0.5 1200 48 16.1 1.7 60.4 9.5 11.4 12.7 11.2 4.2 1.0 0.2 - - 0.05 -2.3 DW34 0.5 1160 72 7.2 6.7 57.7 4.0 21.5 11.8 9.6 2.1 1.6 1.2 2.1 1.0 1.27 -7.9 DW16 1.0 1450 18 51.3 0.4 48.7 0.4 ------0.06 -0.1 DW09 1.0 1385 24 33.1 0.3 54.5 0.3 12.0 0.4 ------0.03 -0.8 DW03 1.0 1350 36 25.9 0.4 56.5 0.3 17.6 0.6 ------0.11 1.7 DW02 1.0 1300 47 18.7 0.4 58.6 0.4 15.5 1.9 1.9 1.8 0.8 0.1 - - 0.01 0.1 DW01 1.0 1250 54 12.2 0.8 56.3 0.9 11.9 3.0 2.8 4.2 1.3 0.2 0.03 0.1 B1330 1.0 1240 24 11.2 2.5 56.2 2.6 19.7 6.1 5.1 4.9 1.2 0.7 - - 0.42 -5.0 DW07 1.0 1220 48 8.4 1.1 54.0 1.2 26.4 2.4 1.9 2.5 1.3 0.3 - - 0.08 2.4 DW18 1.0 1200 73 4.5 2.0 51.9 2.5 31.4 4.3 2.7 3.9 1.4 0.6 - - 0.29 4.8 DW33 1.5 1450 20 53.1 0.8 47.7 0.7 ------1.14 5.4

217 DW04 1.5 1400 28 28.2 0.6 51.6 0.4 19.7 0.8 ------0.07 2.7 DW14 1.5 1350 70 22.8 0.7 53.6 0.6 23.6 0.9 - - s - - - 0.14 -2.1

DW10 1.5 1300 72 9.5 0.9 52.7 0.6 28.1 1.7 1.7 2.1 1.1 0.2 - - 0.04 -2.0 DW17 1.5 1270 73 6.9 1.6 53.0 5.4 23.9 6.6 1.6 1.0 1.0 0.2 - - 0.01 1.1 DW25 2.0 1550 22 53.1 0.5 45.1 0.4 5.3 0.6 ------0.04 -1.5 DW23 2.0 1500 24 28.2 0.4 47.8 0.2 13.6 0.5 ------0.03 -0.4 DW20 2.0 1450 48 22.8 0.6 47.2 0.5 25.3 0.7 - - s - - - 0.09 -2.5 DW29 2.0 1375 72 9.5 1.4 52.6 1.2 29.4 2.1 - - s - - - 0.14 0.4 DW31 2.0 1350 71 6.9 0.9 52.8 0.5 20.8 2.5 16.5 2.8 0.9 0.2 - - 0.04 1.8 C556 2.2 1400 47 11.8 1.7 46.1 1.0 29.1 3.9 11.0 4.3 0.7 0.5 - - 0.18 -5.0 C547 2.2 1350 51 5.1 1.5 46.3 0.8 32.7 3.5 14.4 2.5 1.1 0.2 - - 0.10 -3.6 s: present in experiment but not analyzed due to small crystal sizes. t(h): duration of experiments in hours. One standard deviation from linear regressions are calculated with LIME (Krawczynski and Olive, 2011). RSS is the residual sum of squares from mass balance calculations. ΔFe is the Fe loss (-) or gain (+) relative to the starting material and estimated from mass balance calculations excluding FeO.

217 SiO2 TiO2 Al2O3 Cr2O3 FeO MnO MgO CaO Na2O K2O n Sum DW35 ol 38.35 (15) bdl 0.06 (2) bdl 18.59 (6) 0.47 (2) 42.3 (3) 0.19 (1) na na 3 100.81 opx 54.8 (1.1) 0.03 (1) 0.71 (13) 0.97 (3) 11.6 (2) 0.37 (1) 30.7 (4) 0.74 (2) bdl bdl 5 99.47 DW27 ol 38.3 (3) 0.02 (2) 0.03 (2) 0.47 (5) 19.7 (3) 0.42 (4) 40.2 (4) 0.23 (2) na na 10 100.12 opx 55.5 (4) 0.03 (3) 0.81 (10) 0.85 (6) 12.3 (2) 0.41 (5) 28.6 (4) 1.4 (2) 0.04 (2) bdl 12 99.94 sp 0.45 0.49 12.8 55.10 20.7 0.48 9.9 bdl na na 1 97.11 DW26 ol 38.8 (2) bdl 0.05 (3) 0.42 (5) 20.4 (3) 0.41 (5) 39.5 (3) 0.24 (4) na na 5 100.65 opx 54.8 (3) 0.07 (4) 1.6 (4) 0.98 (14) 12.6 (2) 0.43 (5) 27.8 (3) 1.7 (2) 0.06 (2) bdl 5 100.24 DW28 ol 38.4 (3) 0.02 (2) 0.17 (14) 0.36 (15) 21.19 (13) 0.43 (3) 39.0 (2) 0.32 (5) na na 5 99.65 opx 54.7 (9) 0.09 (3) 1.7 (6) 0.74 (20) 12.6 (3) 0.37 (1) 27.7 (4) 1.9 (3) 0.11 (5) bdl 5 99.71 pig 54.4 (6) 0.10 (3) 1.7 (3) 0.66 (8) 13.2 (3) 0.39 (4) 25.1 (7) 4.2 (3) 0.14 (2) bdl 5 100.95 sp 1.6 (3) 0.7 (2) 25.6 (3.0) 39.0 (3.0) 21.2 (1.2) 0.31 (1) 11.4 (8) 0.2 (1) na na 3 100.93 DW34 ol 37.6 (2) 0.02 (1) 0.07 (2) na 22.8 (1) 0.55 (3) 38.7 (3) 0.24 (1) na na 6 101.07 opx 54.4 (3) 0.04 (2) 1.8 (5) 0.74 (22) 12.3 (8) 0.38 (5) 28.2 (5) 2.1 (2) 0.04 (2) bdl 4 100.30 pig 52.7 (5) 0.15 (4) 2.5 (3) 0.9 (3) 10.1 (9) 0.44 (2) 19.3 (9) 13.4 (1.6) 0.45 (12) bdl 5 100.57 sp 0.7 (4) 2.78 (4) 48.9 (8) 11.0 (2) 21.18 (5) 0.38 (2) 15.1 (3) bdl na na 3 101.60 plag 56.3 (5) bdl 27.3 (4) bdl 0.39 (5) bdl 0.13 (3) 9.7 (6) 6.0 (4) 0.17 (3) 4 101.30 DW16 ol 39.5 (3) 0.03 (3) 0.04 (2) 0.56 (10) 15.5 (3) 0.32 (2) 43.8 (2) 0.14 (4) na na 5 100.84 DW09 ol 38.95 (12) bdl 0.06 (1) 0.49 (4) 19.2 (3) 0.42 (3) 40.7 (3) 0.19 (4) na na 4 100.48 opx 55.4 (3) 0.03 (1) 1.26 (3) 1.16 (3) 11.63 (8) 0.38 (3) 29.0 (1) 1.00 (1) 0.09 (3) bdl 4 100.92 DW03 ol 38.8 (2) 0.01 (1) 0.05 (1) 0.36 (3) 20.6 (2) 0.43 (5) 39.4 (2) 0.27 (3) 0.02 (2) na 8 100.41 opx 54.7 (3) 0.05 (2) 1.7 (2) 1.09 (11) 12.4 (2) 0.39 (4) 28.0 (3) 1.51 (8) 0.07 (2) bdl 10 100.43 DW02 ol 38.7 (2) bdl 0.07 (2) 0.34 (4) 21.4 (3) 0.47 (6) 38.7 (2) 0.29 (2) na na 5 100.29 opx 54.2 (7) 0.09 (3) 2.5 (7) 1.0 (2) 12.5 (8) 0.40 (6) 27.1 (7) 2.16 (15) 0.07 (3) bdl 17 100.35 pig 54.3 (4) 0.08 (3) 1.7 (4) 0.9 (1) 12.5 (3) 0.47 (6) 25.7 (3) 4.14 (11) 0.11 (5) bdl 6 100.52 sp 0.29 (5) 0.54 (3) 28.1 (8) 38.5 (1.5) 20.7 (2) 0.36 (2) 11.45 (5) 0.07 (1) na na 2 97.87 DW01 ol 38.7 (3) 0.02 (2) 0.05 (15) 0.17 (3) 22.0 (3) 0.45 (4) 38.4 (3) 0.29 (3) na na 7 99.77 opx 53.8 (3) 0.13 (2) 3.0 (4) 0.9 (1) 12.8 (3) 0.40 (5) 26.7 (3) 2.3 (2) 0.10 (4) bdl 12 100.20 pig 53.9 (6) 0.12 (10) 2.2 (8) 0.7 (2) 13.1 (6) 0.47 (4) 24.1 (8) 5.1 (5) 0.25 (5) bdl 9 100.18 sp 0.20 (4) 0.54 (5) 30 (3) 35 (4) 21.7 (3) 0.40 (3) 11.6 (4) 0.07 (2) na na 3 97.65 B1330 ol 38.5 (2) bdl bdl na 21.5 0.47 (7) 39.3 (1) 0.19 (3) na na 5 99.31 opx 53.5 (4) 0.09 (3) 3.4 (2) 0.84 (19) 13.0 (1) 0.45 (4) 25.8 (3) 2.67 (13) 0.22 (10) 0.01 (1) 5 100.72 pig 52.8 (5) 0.13 (1) 3.0 (6) 1.1 (1) 11.8 (9) 0.49 (4) 21.7 (1.5) 8.7 (2.1) 0.36 (11) 0.01 (1) 8 100.60 sp 0.53 (12) 0.50 (1) 31.8 (1.8) 33.6 (2.9) 21.2 (7) 0.41 (5) 11.9 (3) 0.10 (1) na na 2 98.97 DW07 ol 38.2 (3) bdl 0.10 (7) 0.15 (5) 22.9 (2) 0.47 (4) 37.8 (3) 0.29 (4) na na 5 99.79 opx 53.5 (4) 0.12 (6) 2.9 (5) 0.86 (6) 13.4 (2) 0.42 (2) 26.3 (4) 2.4 (2) 0.13 (1) bdl 5 99.93 pig 52.6 (6) 0.17 (3) 3.1 (7) 1.0 (2) 11.3 (7) 0.46 (4) 20.6 (1.5) 9.8 (2.0) 0.42 (13) bdl 6 99.56 sp 0.32 0.65 32.9 31.9 22.3 0.35 11.5 0.07 na na 1 97.02 DW18 ol 38.49 (5) 0.02 (1) 0.06 (5) 0.14 (3) 23.4 (3) 0.45 (10) 37.13 (16) 0.27 (4) na na 5 100.02 opx 52.9 (6) 0.13 (4) 3.5 (6) 0.8 (3) 13.5 (4) 0.42 (3) 26.4 (2) 2.2 (2) 0.13 (5) bdl 8 100.48 pig 52.3 (8) 0.23 (6) 3.9 (1.1) 1.1 (2) 10.6 (8) 0.43 (4) 19.0 (1.2) 11.7 (1.7) 0.64 (15) bdl 15 100.14 sp 0.26 0.54 33.0 30.9 22.5 0.41 11.7 0.29 0.35 0.04 1 97.49 DW33 ol 39.2 (5) bdl bdl bdl 16.9 (1) 0.41 (2) 43.3 (4) 0.14 (2) na na 5 99.20 DW04 ol 39.1 (3) bdl 0.09 (3) 0.38 (3) 19.7 (4) 0.41 (5) 39.9 (6) 0.26 (2) na na 5 100.07 opx 54.4 (3) 0.04 (1) 2.30 (16) 1.24 (13) 11.9 (3) 0.39 (5) 28.2 (6) 1.35 (9) 0.07 (2) bdl 15 100.50 DW14 ol 38.8 (2) bdl 0.09 (2) 0.32 (2) 20.5 (2) 0.45 (1) 39.5 (1) 0.30 (4) na na 5 100.31 opx 54.1 (3) 0.04 (2) 2.41 (5) 1.07 (7) 12.3 (4) 0.44 (2) 27.5 (4) 2.0 (1) 0.11 (3) bdl 4 100.03 DW10 ol 38.8 (6) bdl 0.06 (1) 0.11 (1) 21.7 (2) 0.46 (1) 38.35 (18) 0.30 (1) na na 4 100.37 opx 51.8 (5) 0.07 (2) 3.8 (6) 1.0 (1) 12.3 (3) 0.39 (6) 25.3 (5) 2.3 (1) 0.19 (3) bdl 6 102.72 pig 52.5 (7) 0.13 (4) 4.3 (6) 1.3 (4) 11.2 (7) 0.45 (3) 20.5 (7) 8.9 (1.0) 0.66 (9) bdl 6 100.04 sp 0.34 0.49 31.4 32.9 22.8 0.39 11.7 0.01 0.03 na 1 98.15 DW17 ol 37.9 (2) 0.02 (1) 0.08 (3) 0.14 (3) 22.87 (14) 0.47 (4) 38.2 (2) 0.22 (20 na na 5 101.15 opx 52.8 (2) 0.12 (1) 4.1 (2) 0.88 (17) 13.23 (14) 0.41 (4) 26.1 (3) 2.14 (4) 0.20 (7) bdl 4 100.68 pig 52.2 (3) 0.17 (5) 4.2 (5) 1.2 (2) 11.9 (5) 0.47 (4) 20.9 (1.0) 8.3 (1.0) 0.68 (15) bdl 6 100.36 sp 0.27 (3) 0.50 (3) 33.9 (2.8) 30.2 (3.0) 22.5 (1) 0.41 (1) 12.1 (4) 0.08 (4) 0.02 (1) na 3 99.59 DW25 ol 39.6 (2) 0.02 (3) 0.07 (1) 0.45 (4) 16.4 (2) 0.35 (6) 43.0 (3) 0.15 (3) na na 10 100.50 opx 56.1 (3) 0.02 (2) 1.25 (9) 0.91 (5) 10.0 (2) 0.31 (4) 307 (2) 0.72 (5) 0.06 (2) bdl 10 100.04 DW23 ol 39.3 (4) bdl 0.08 (1) 0.46 (3) 18.1 (2) 0.34 (5) 41.6 (5) 0.18 (2) na na 5 99.60 opx 55.0 (6) 0.03 (1) 1.7 (2) 1.05 (9) 10.9 (2) 0.33 (5) 29.8 (5) 1.1 (2) 0.09 (2) bdl 9 100.19 DW20 ol 387 (1) bdl 0.10 (2) 0.31 (4) 19.6 (1) 0.44 (6) 40.5 (1) 0.25 (4) na na 4 100.98 opx 54.5 (4) 0.02 (2) 2.33 (2) 1.2 (1) 11.7 (2) 0.37 (3) 28.5 (1) 1.33 (9) 0.12 (1) bdl 4 100.65 DW29 ol 38.6 (2) 0.02 (3) 0.09 (3) 0.20 (4) 21.6 (2) 0.43 (5) 38.8 (1) 0.28 (3) na na 10 99.59 opx 53.6 (2) 0.07 (2) 3.17 (14) 1.3 (2) 12.4 (2) 0.39 (5) 26.7 (2) 2.17 (7) 0.23 (4) bdl 10 99.77 DW31 ol 38.4 (3) 0.02 (2) 0.14 (5) 0.22 (4) 22.3 (3) 0.51 (3) 38.0 (4) 0.30 (5) na na 7 100.98 opx 53.2 (6) 0.09 (3) 4.0 (7) 1.03 (8) 12.7 (3) 0.43 (5) 26.2 (4) 2.2 (2) 0.24 (3) bdl 8 101.17 pig 52.9 (5) 0.11 (3) 4.1 (8) 1.15 (5) 12.2 (2) 0.52 (2) 21.9 (7) 6.3 (5) 0.75 (10) bdl 5 101.29 sp 0.80 0.54 34.3 29.2 22.6 0.36 12.0 0.08 0.08 na 3 100.57 C556 ol 37.7 (2) bdl 0.07 (2) na 21.4 (1) 0.40 (3) 40.2 (1) 0.19 (3) na na 4 100.96 opx 53.40 (6) 0.07 (2) 4.46 (15) 1.00 (2) 12.9 (1) 0.44 (1) 25.5 (15) 2.01 (6) 0.23 (3) bdl 4 99.06 pig 52.5 (2) 0.07 (1) 4.45 (4) 1.17 (6) 11.71 (16) 0.50 (1) 21.5 (2) 7.2 (3) 0.80 (4) bdl 4 101.22 sp 0.7 (5) 0.36 (2) 30.0 (8) 32.8 (3) 23.3 (1) 0.36 (8) 12.4 (1) 0.14 (12) na na 2 100.93 C547 ol 37.3 (3) na 0.15 (12) na 22.70 (12) 0.49 (4) 39.2 (2) 0.18 (3) na na 5 101.47 opx 52.4 (5) 0.11 (5) 4.6 (6) 1.10 (10) 12.9 (4) 0.39 (2) 26.2 (9) 2.0 (5) 0.33 (11) bdl 4 101.01 pig 51.8 (5) 0.17 (3) 5.3 (4) 1.36 (14) 11.3 (3) 0.45 (3) 20.0 (4) 8.4 (5) 1.17 (7) 0.00 4 100.43 sp 0.24 0.43 32.2 31.8 23.4 0.36 12.1 0.07 na na 1 100.62

218 Supplemental Table 2. Mineral compositions Experiments ordered as in Supplemental Table 1: from low to high pressures and high to low temperatures. All oxide concentrations are in wt.% and compositions are normalized to 100. Sum is the original total of analyses before normalization. n is the number of EPMA analyses averaged. The digits in brackets represent one standard deviation from the set of analyses averaged for each experiment. na: not analyzed. bld: below detection limit.

219 220 Chapter 5. Crystallization history of enriched shergottites from Fe and Mg isotope fractionation in olivine megacrysts

Abstract

Martian meteorites are the only samples available from the surface of Mars. Among them, olivine-phyric shergottites are basalts containing large zoned olivine crystals with highly magnesian cores (Fo 70-85) and rims richer in Fe (Fo 45-60). The Northwest Africa 1068 meteorite is one of the most primitive “enriched” shergottites (high initial 87Sr/86Sr and low initial ε143Nd). It contains olivine crystals as magnesian as Fo 77 and is a major source of information to constrain the composition of the parental melt, the composition and depth of the mantle source, and the cooling and crystallization history of one of the younger magmatic events on Mars (~180 Ma). In this study, Fe-Mg isotope profiles analyzed in situ by femtosecond-laser ablation MC- ICP-MS are combined with compositional profiles of major and trace elements in olivine

56 megacrysts. The cores of olivine megacrysts are enriched in light Fe isotopes (! FeIRMM-14 = -0.6

26 to -0.9 ‰) and heavy Mg isotopes (! MgDSM-3 = 0 to 0.2 ‰) relative to megacryst rims and to

56 26 the bulk martian isotopic composition (! Fe = 0±0.05 ‰, ! Mg = -0.27±0.04 ‰). The flat forsterite profiles of megacryst cores associated with anti-correlated fractionation of Fe-Mg isotopes indicate that these elements have been rehomogenized by diffusion at high temperature. We present a 1-D model of simultaneous diffusion and crystal growth that reproduces the observed element and isotope profiles. The simulation results suggest that the cooling rate during megacryst core crystallization was slow (43±21 ºC/year), and consistent with pooling in a deep crustal magma chamber. The megacryst rims then crystallized 1 to 2 orders of magnitude faster during magma transport towards the shallower site of final emplacement. Megacryst cores had a forsterite content 3.2±1.5 mol% higher than their current composition and some were in equilibrium with the whole-rock composition of NWA 1068 (Fo 80±1.5). NWA 1068 composition is thus close to a primary melt (i.e. in equilibrium with the mantle) from which other enriched shergottites derived.

221 1. Introduction

Martian meteorites are the only rock samples available from Mars. Together with data obtained from spacecraft observations and rover missions, they hold key information on the formation and evolution of the planet. Shergottites, which are basalts or gabbros, are the most common type of martian meteorites. Among shergottites, the olivine-phyric group (Goodrich, 2002) is the most primitive, and some samples are arguably representative of liquids produced by partial melting of the mantle and not significantly affected by fractional crystallization or assimilation (Basu Sarbadhikari et al., 2009; Gross et al., 2011; Musselwhite et al., 2006; Peslier et al., 2010). These samples record magmatic processes that occurred over a large range of pressure and provide constraints on the depth and composition of their mantle sources and the shallower conditions of crystallization. Some olivine-phyric shergottites are strongly (e.g. Aoudjehane et al., 2012; Shirai and Ebihara, 2004; Zipfel et al., 2000) or moderately (Filiberto et al., 2012; Gross et al., 2013) depleted in light rare earth elements (LREE) relative to heavy rare earth elements (HREE). Other shergottites are not depleted in LREE and are characterized by higher initial 87Sr/86Sr and lower initial ε143Nd. Shergottites from this last group are referred to as “enriched” (e.g. Barrat et al., 2002; Basu Sarbadhikari et al., 2009; Peslier et al., 2010). Shergottites have been the subject of continuous discussions (see the review by Jones, 2015) concerning their crystallization age (Bouvier et al., 2009; Moser et al., 2013), the origin of the enriched source (Debaille et al., 2008; Peters et al., 2015) and how they relate to other martian rocks (Filiberto and Dasgupta, 2011; Schmidt and McCoy, 2010). The detailed crystallization history of olivine-phyric shergottites has remained elusive due to the controversial origin of the olivine megacrysts 0.3-3 mm in diameter. These megacrysts are either interpreted as xenocrysts added to a basaltic melt (Barrat et al., 2002; Filiberto et al., 2010; Herd, 2006; Shearer et al., 2006) or phenocrysts that potentially accumulated in the magma (Balta et al., 2013, 2015; Filiberto et al., 2010; Shearer et al., 2013). Iron and magnesium zoning profiles across olivine crystals have been used to constrain the cooling rates of terrestrial magmas (Costa and Chakraborty, 2004; Costa et al., 2008; Costa and Dungan, 2005) based on the assumption that chemical zoning mainly results from diffusion. For shergottites, it is sometimes assumed that olivine zoning results purely from rapid crystal growth in an evolving liquid. In such cases, the Fe-Mg zoning is used to reconstruct the order of appearance of different pyroxenes (orthopyroxene, pigeonite and augite) relative to olivine (Gross et al., 2013;

222 Peslier et al., 2010). Finally, other authors modelled the combined effect of crystal growth and diffusion to constrain the cooling rates of shergottites (Mikouchi et al., 2001; Miyamoto et al., 2010). However, diffusion, crystallization or a combination of both, usually lead to non-unique profiles of Fe-Mg in olivine (see section 4.2.1). As the origin of olivine megacrysts in shergottites is controversial, constraining the cooling rate or the crystallization sequence from the forsterite content (Fo; equivalent to Mg# = [Mg/(Mg+Fe2+)]×100 with concentrations expressed in mol.%) of olivine crystals alone remains speculative. In this work, we study zoning in olivine megacrysts of Northwest Africa (NWA) 1068. This olivine-phyric shergottite, together with Larkman Nunatak (LAR) 06319, is one of the few picritic basalts from the enriched group and, presumably, the closest in composition to a primary melt. The composition of olivine in NWA 1068 was first measured by Barrat et al. (2002) who performed individual electron microprobe (EMP) analyses and profiles in selected megacrysts and in groundmass crystals. They reported a compositional range from Fo42 to Fo72. Herd (2006) later measured olivine crystals reaching Fo74 and Shearer et al. (2008) reported a composition of Fo75 in NWA 1110, a meteorite paired with NWA 1068. In the Earth’s mantle, olivine is typically close to

Fo90. The martian mantle is richer in FeO (Dreibus and Wänke, 1985; Taylor, 2013) and the composition of mantle olivine should be in the range Fo75-Fo85 (Fo85 corresponds to the source of “depleted” shergottites; Musselwhite et al., 2006). Olivine megacrysts in NWA 1068 (Fo 74-77) are at the lower end of this range and thus represent ideal witnesses of the deep structure and plumbing systems of a recent (185±11 Ma; Shih et al., 2003) volcanic system on Mars. In addition to element profiles, we measured the first Fe-Mg isotope profiles in shergottites to further constrain the origin and cooling history of olivine megacrysts. Recent data on Fe and Mg isotopes in olivine have revealed that diffusive isotope fractionation can exceed equilibrium isotope fractionation at magmatic temperatures (> 1000 ºC) by an order of magnitude (Dauphas et al., 2010; Oeser et al., 2015; Sio and Dauphas, 2016; Sio et al., 2013; Teng et al., 2011; Weyer and Seitz, 2012). Accordingly, Fe-Mg element zoning in olivine when coupled with Fe-Mg isotopic zoning allows distinguishing between chemical zoning produced by crystal growth and diffusion and to constrain cooling rates more reliably (Oeser et al., 2015; Richter et al., 2016; Sio et al., 2013; Sio and Dauphas, 2016, Teng et al., 2011). Therefore, by studying simultaneously major and trace element contents and Fe-Mg isotope fractionation, we can observe the effect of both diffusion and crystal growth on the composition of

223 NWA 1068 olivine megacrysts. We then model isotope and element profiles to place quantitative constraints on the crystallization history of NWA 1068 and other related enriched shergottites. In particular, we provide estimates on both the residence time of NWA 1068 parental melt in its magma chamber and the cooling rate during magma ascent and crystallization of the groundmass.

2. Analytical methods

Olivine megacrysts were analyzed from 2 thin sections of NWA 1068 with surface areas of 3.5 cm2 and 0.8 cm2. We refer to those sections as A and B, respectively. The major (Fe, Mg, Si) and minor/trace (Ca, Mn, Cr, Ti, P, Ni) elements were first measured by wavelength dispersive spectrometry (WDS) with two Cameca SX100 EMP at the “Laboratoire Magmas et Volcans” (LMV) in Clermont-Ferrand and at the Leibniz Universität Hanover (LUH). A beam current of 15 nA and a voltage of 15 kV were used for major elements. The current intensity and counting time were increased to 100 nA and 300 s (as opposed to 20 s) for trace elements. Major element maps of olivine crystals in section A (Fig. 1) were also acquired at LMV with a current of 100 nA, a 20 kV voltage (instead of 15 kV), a counting time of 40 ms, and a step of 4 µm.

Additional trace elements (V, Co, Sc, Ni, Cu, Zn, Cr and Zr) were measured by femtosecond-laser ablation (fs-LA) of the sample with a Spectra-Physics Solstice system (see Oeser et al., 2014 and Lazarov and Horn, 2015 for details) connected to a fast scanning sector field ICP-MS (ThermoScientific Element XR®) at the Institute of Mineralogy, LUH. Sample ablation was performed by spot analyses with a beam diameter of 40 µm and a step width of 50 µm on average (Fig. 1). Laser repetition rate was 8-10 Hz. Each analysis started with a 30 second background acquisition, followed by an ablation interval of 40-60 seconds. External calibration of the acquired data was performed using the USGS reference glasses BCR-2G and BIR-1G with the preferred values reported in the GeoReM database (Jochum et al., 2005). 29Si was used for internal standardization. Data reduction was performed with the Lotus-based spreadsheet program LAMTRACE (Jackson, 2008), which also automatically calculates the lower limit of detection (LLD) for each analysis using the algorithm developed by Longerich et al. (1996).

Fe- and Mg isotope ratios were measured in situ using the same laser ablation system but connected to a ThermoFinnigan Neptune multicollector ICP-MS (MC-ICP-MS) operated in a high mass resolution mode (Weyer and Schwieters, 2003) with sample-standard bracketing following the exact procedure described in Oeser et al. (2014, 2015). Analyzing simultaneously the 60Ni/58Ni

224 ratio of a Ni standard solution (NIST SRM 986) with the ablation aerosol allowed to monitor the instrumental mass bias for Fe isotopes (Poitrasson and Freydier, 2005). Using this technique, the long-term reproducibility has been shown to be better than 0.13 ‰ for both δ56Fe and δ26Mg. The within-session reproducibility, which is more appropriate for defining the uncertainties of the measured d-values for a single isotope profile, is better than 0.10 ‰ (2 SD), based on replicate analyses of silicate reference glasses (Oeser et al., 2014). Ablation of olivine crystals was performed along lines of 150-200 µm in length parallel to the crystal rim with a laser spot size of ~40 µm (Fig. 1). Bracketing standards were the silicate reference glasses BHVO-2G and GOR132- G for Mg isotope analyses and BCR-2G for Fe isotope analyses. As the Fe and Mg isotopic compositions of these reference materials relative to IRMM-14 and DSM-3 are known (Oeser et al., 2014), we report isotope measurements as δ56Fe and δ26Mg relative to IRMM-14 and DSM-3, respectively (equations [1] and [2]).

*+ *2 $% (),-./013 (),-./01 ! "# = '*+ *2 − 1; × 1000 [1] ()45667823 ()4566782

D+ D2 A% BC,-./013 BC,-./01 ! ?@ = ' D+ D2 − 1; × 1000 [2] BCEF67G3 ()EF67G

As Fe-Mg diffusion in olivine is anisotropic (Buening and Buseck, 1973; Dohmen et al., 2007), knowledge of the crystallographic orientation is necessary to constrain the Fe-Mg diffusivity. We obtained the crystal lattice orientation by electron backscatter diffraction (EBSD) measurements carried out at the Ruhr University Bochum (RUB) with a SEM Zeiss LEO 1530 Gemini Field Emission at an acceleration voltage of 20 kV, a working distance of 25 mm, and a Nordlys EBSD detector tilted by 70º.

3. Results

3.1. Olivine major element compositions

In sections A and B, we observed three megacrysts with core compositions in the range of

Fo74-75 and one glomerocryst composed of four olivine megacrysts, the largest of which has a composition of Fo77 (Fig. 2). Fo77 is the most Mg-rich olivine composition documented for NWA

1068 and paired meteorites. A few megacrysts have core compositions as low as Fo70 but crystal cores are usually in the range Fo70-74 (Fig. 3 and 4). There appears to be a continuous spectrum of

225 core compositions from Fo70 to Fo77 rather than different populations of megacrysts. The Fo profiles of megacryst cores are relatively flat and suggest that the variability in Fo does not primarily result from a thin section cutting effect. Megacryst rims, usually 100-200 µm thick, show a steep zoning characterized by an outward decrease of the forsterite content down to Fo57-47. The composition of the outer rim at the interface with the groundmass slightly varies from one megacryst to another (Fig. 3). Some forsterite profiles in the rim of megacrysts appear to be slightly “concave up” (e.g. B1, B2) while others are not (e.g. A1). Zoning is preserved within the glomerocryst and the Fo content drops significantly between the cores of adjacent twinned crystals (Fig. 2). Short-wavelength oscillatory zoning has been recently observed in some nakhlites and attributed to crystal growth processes (Jambon et al., 2016). No oscillatory zoning was found in the olivine megacrysts from NWA 1068 despite an extensive search.

3.2. Composition in minor and trace elements

The outer 100-200 µm of the rims with strong Fe-Mg zoning are also zoned for minor element concentrations. MnO increases outward from ~0.5 to ~0.75 wt.%, Ni decreases from ~700 ppm to ~400 ppm and Cr decreases from ~600-800 ppm to ~200 ppm (Fig. 3). Occasionally, Cr contents as high as 1200 ppm were measured by LA-ICP-MS in olivine cores but they likely result from the ablation of micrometric chromite crystals, which are abundant in NWA 1068 olivine megacryts (Barrat et al., 2002; Shearer et al., 2008). Similarly, V contents seem to decrease from core (~20 ppm) to rim (~12 ppm) and outlier analyses up to 50 ppm are interpreted as resulting from contamination by spinel (Table 1). Zn appears to increase from core to rim but also varies significantly between megacrysts. Sc slightly increases outward from ~6 ppm to 9-10 ppm. P contents vary between 30 and 1200 ppm but the P-zoning was not studied in detail and we refer to Shearer et al. (2013) who highlighted a complex oscillatory P-zoning. CaO contents stay relatively homogeneous throughout megacrysts (0.2-0.25 wt.%) with some excursions towards much higher CaO contents (up to 1 wt.%). Those higher CaO values result from the presence of Ca-rich carbonates precipitated within microcracks in the hot desert environment in which NWA 1068 was collected. The typical 40 µm spot-size used for LA-ICP-MS analyses makes some contamination by carbonates unavoidable. Unlike most trace elements, Ti is only zoned in the outermost 50±10 µm of the megacrysts where the Ti content suddenly rises from ~30 ppm in core to 150-300 ppm at the megacryst-groundmass interface. Both EMP and LA-ICP-MS analyses are consistent with a

226 Ti content of ~30-40 ppm in megacryst cores (Fig. 3), less than half of what was measured by Shearer et al. (2013) in NWA 1183. The concentrations of other trace and minor elements are consistent with previous analyses performed on NWA 1110/1183 that are both paired with NWA 1068 (Shearer et al., 2008, 2013).

3.3. Fe and Mg isotopic composition

Megacryst cores are homogeneous in isotopic compositions within analytical uncertainty and variations only appear in the ~100 µm rims of the megacrysts (Fig. 4). The Mg isotope ratios are consistent in all the megacryst cores analyzed and vary in a narrow range of 0.0 to +0.2 ‰ in

26 δ MgDSM-3. The Fe isotope ratios of megacryst cores, while homogeneous, are more variable 56 (δ FeIRMM-14 of -0.6 to -1.0 ‰) from one crystal to another. Fe and Mg isotope ratios in olivine megacryst cores are significantly fractionated relative to the bulk isotopic composition of martian meteorites. The δ56Fe of virtually all the martian meteorites analyzed are identical and

56 unfractionated relative to the standard with δ FeIRMM14 of 0 ± 0.05 (Anand et al., 2006; Poitrasson et al., 2004; Sossi et al., 2016; Wang et al., 2012; Weyer et al., 2005). Early studies of Mg isotope ratios by Norman et al. (2006), Wiechert and Halliday (2007), and Chakrabarti and Jacobsen (2010) 26 observed a large range of δ MgDSM-3 (-0.1 to -0.57 ‰) for martian meteorites. The light measurements (-0.57 ‰) of Chakrabarti and Jacobsen (2010) have been shown to result from analytical artifacts (Teng et al. 2015). A recent and more exhaustive study (Magna et al., 2017) 26 shows that 31 martian meteorites have bulk δ MgDSM-3 ranging between -0.11 and -0.32 ‰ and 26 that all olivine-bearing samples have a δ MgDSM-3 of -0.27±0.04 ‰. As the analysis of Fe-Mg isotopes by fs-LA-MC-ICP-MS appears to be largely matrix-independent (Oeser et al., 2014; Steinhoefel et al., 2009), we also performed two raster-mode measurements in the groundmass of NWA 1068. We obtained a δ56Fe (relative to IRMM-14) of -0.03 ‰ and a δ26Mg of -0.41 ‰. The δ56Fe of NWA 1068 groundmass is fully consistent with the martian average of the literature. The δ26Mg of the groundmass is also consistent, within the ±0.13 ‰ analytical uncertainty, to the recent average δ26Mg of olivine-bearing martian meteorites (Magna et al., 2017). It is however possible that the groundmass of NWA 1068 is slightly lighter (-0.4 ‰) than this average composition. Heavier olivine cores (0.0 to 0.2 ‰) and a lighter groundmass (-0.4 ‰) could result in a bulk value closer to the average value of -0.27±0.04 ‰.

227 Olivine megacryst cores have heavier Mg isotopes and lighter Fe isotopes relative to all bulk martian meteorites and to the groundmass of NWA 1068. Their δ26Mg are higher by 0.3-0.5 ‰ and their δ56Fe are lower by 0.6 to 1.0 ‰ (Fig. 4). On the contrary, the rims of olivine megacrysts converge toward the average isotopic compositions of the literature (martian meteorites) and NWA 1068 groundmass (δ26Mg ~ -0.4 or -0.25 ‰ and δ56Fe ~0.0 ‰). The Mg isotope profiles of some olivine megacrysts are more complex and are characterized by a spike of heavier isotopes in the intermediate rim (50-100 µm from the outer rim). This spike typically reaches higher values of δ26Mg (0.3-0.4 ‰) than those observed for megacryst cores and can reach a δ26Mg of 0.9 ‰ (A4 and B7, Fig. 4).

4. Discussion

4.1. Nomenclature and origin of olivine megacrysts in shergottites

The origin of olivine megacrysts in olivine-phyric shergottites is controversial as reflected by the complex and sometimes inconsistent nomenclature used in the literature. Numerous authors used the term xenocryst to refer to the megacrysts (Barrat et al., 2002; Filiberto et al., 2010; Goodrich, 2003; Herd, 2006; Koizumi et al., 2004; McSween and Jarosewich, 1983; Shearer et al., 2006; Wadhwa et al., 2001) while others preferred the term phenocryst (Balta et al., 2013, 2015; Basu Sarbadhikari et al., 2009, 2011; Filiberto and Dasgupta, 2011; Gross et al., 2011, 2013; Koizumi et al., 2004; Liu et al., 2016; Mikouchi et al., 2001; Musselwhite et al., 2006; Peslier et al., 2010; Shearer et al., 2008, 2013; Taylor et al., 2002; Usui et al., 2008). Xenocrysts are generally defined as disequilibrium crystals with no genetic relationship to the magma. Phenocrysts are equilibrium crystals that formed within the magma and reached a large size as a consequence of low nucleation rates. In the recent literature, the term phenocryst is generally preferred due to the fact that the trace element contents (Ni, Co, Y) of megacryst cores (Shearer et al., 2008), the composition in major elements and REE of melt inclusions (Basu Sarbadhikari et al., 2009, 2011; Shearer et al., 2008), and the isotopic systematics (Shafer et al., 2010) are all consistent with a common origin of the groundmass and olivine megacrysts. However, megacryst cores would have formed during an early stage of crystallization and would have been later mobilized by the melt that crystallized the groundmass. Several authors have suggested that when the magma crystallized near the surface, it contained the megacrysts from more than one liquid aliquot such that most olivine-phyric shergottites contain >10 wt.% of “accumulated phenocrysts” also referred to as

228 “antecrysts”. This line of reasoning is based on two main arguments: (1) most bulk rock compositions of olivine-phyric shergottites are too rich in Mg compared to a liquid in equilibrium with megacryst cores (e.g. Filiberto and Dasgupta, 2011); (2) the crystal size distribution (CSD; Marsh, 1988) of olivine is kinked around 1.5 mm due to an excess of large crystals (e.g. Balta et al., 2015; Basu Sarbadhikari et al., 2009; Ennis and McSween, 2014). Different authors have slightly different definitions of phenocrysts, antecrysts, accumulated phenocrysts and xenocrysts. Large crystals of different origins can also locally coexist in a single rock. Therefore, the crystallization history of the large olivine crystals relative to the groundmass cannot be summarized by a single term. As we describe the crystallization history of the large olivine crystals in NWA 1068, we will continue to refer to them as olivine megacrysts, a term that is free of any petrogenetic connotation.

4.2. Origin of the chemical zoning in olivine megacrysts of NWA 1068

Olivine megacrysts in NWA 1068 were first interpreted as xenocrysts derived from disrupted cumulates, possibly co-genetic with the groundmass (which some authors would call “antecrysts”), but occasionally showing disequilibrium textures (Fig. 7 from Barrat et al., 2002).

The forsterite content of the cores was described as flat (at ~Fo72) and the zoning of the rim as resulting chiefly from diffusion driven by contact with the Fe-rich melt that later formed the groundmass. Shearer et al. (2008, 2013) confirmed that the megacrysts were genetically related to the groundmass by analyzing minor elements in olivine (P, Ni, Co, Y) and in olivine melt inclusions (REE). However, Filiberto et al. (2010) and Filiberto and Dasgupta (2011) argued that olivine megacrysts were partly accumulated based on the Mg# of the bulk-rock. Herd (2006) showed that the olivine megacrysts in NWA 1068 also crystallized at different redox conditions (FMQ-2.5) than the groundmass (FMQ+0.5). While this observation was first inferred to point toward a “xenocrystic origin” of megacrysts, it was later shown that it is also consistent with closed system crystallization and result from the incompatible behavior of Fe3+ (Peslier et al., 2010) and/or the degassing of volatile elements (Balta et al., 2013; Shearer et al., 2013). Despite the increased understanding of the crystallization conditions of NWA 1068, the origin of the Fe-Mg zoning in megacrysts and the timescale of crystallization remain ambiguous. The initial interpretation of forsterite profiles by Barrat et al. (2002) and Mikouchi and Miyamoto (2002) was that only the olivine rims were affected by diffusion as megacrysts cores are nearly

229 homogeneous in composition. When an olivine crystal is placed in a melt that is enriched in Fe compared to the liquid from which the olivine crystallized, Fe-Mg interdiffusion will progressively bring the olivine towards a more Fe-rich equilibrium composition (Fig. 5). As heavy isotopes diffuse slightly slower than the light ones (Jambon, 1980; Richter et al., 1999), the inward replacement of Mg by Fe (decrease in Fo content) also enriches the olivine crystals in 26Mg relative to 24Mg and depletes them in 56Fe relative to 54Fe (Oeser et al., 2015; Sio et al., 2013; Teng et al., 2011). Diffusion first lowers the Fo content of the rims while leaving the core unaffected. No isotope fractionation is observed in the core at this stage (Fig 5; Oeser et al., 2015; Richter et al., 2016). As the olivine crystal continues to equilibrate, Fe-Mg isotope fractionation progressively reaches the core. The Fo profile of the core is no longer flat but forms a smooth (bell-like) curve (Fig. 5; Sio et al., 2013). In NWA 1068, the cores of olivine megacrysts have relatively flat Fo and Ni profiles and only the rims display a strong zoning (rims and outer rims in Fig. 3). However, the cores of megacrysts are also characterized by the anti-correlated isotope fractionation that is characteristic of diffusion (Fig. 4). Those combined observations are incompatible with diffusion alone accounting for the isotope and element profiles of NWA 1068 megacrysts. Diffusion of Fe and Mg in olivine is anisotropic and 4-8 times faster along the crystallographic c-axis (Dohmen et al., 2007). Crystallographic orientations obtained from EBSD measurement were used to calculate the diffusivity of Fe-Mg at a fixed temperature, pressure, redox condition and starting olivine composition (Dohmen et al., 2007; Dohmen and Chakraborty, 2007; also see section 4.2.1). We find no simple correlation between the diffusivity along each Fo profile and the chemical gradient of the rims (Fig. 6). If the Fo zoning resulted purely from diffusion, the chemical gradient would be expected to be weaker for directions at low angle of the c-axis (larger diffusivities). Yet, there are clear signs that olivine megacrysts were affected by diffusion. The presence of Fe-Mg anti-correlated isotope fractionation in the core indicates that diffusion influenced the composition of megacryst cores. In addition, oscillatory growth zoning (Jambon et al., 2016) is only preserved by the slowest diffusing element (P; Shearer et al. 2013) and was erased by diffusion for other, faster diffusing, elements (Fig. 3). Important evidence suggests that the strong zoning of the megacryst rims actually results partly from crystal growth in a melt of evolving composition. Forsterite zoning is preserved between the crystals of the olivine glomerocryst (Fig. 2). This indicates that olivine of Fe-rich composition was crystallizing when the crystals grew and got attached to each other. The small

230 olivine crystals within the groundmass have the same composition as the megacryst outer rims. The liquid from which the groundmass crystallized was thus saturated in olivine and both the small crystals and megacryst rims formed simultaneously. Shearer et al. (2013) interpret complex P-maps of NWA 1183 megacrysts as highlighting different stages of crystal growth. The last one would have formed the outer ~50 µm of the olivine megacrysts. We postulate that the sharp Ti enrichment in the outermost 40-60 µm of megacryst rims in NWA 1068 (see outer rims, Fig. 3) results from the same final stage of crystal growth. The sudden rise in Ti concentration could result from entrapment in the interfacial region (e.g. Watson, 1996) or from the development of a diffusive boundary layer in the silicate melt, as suggested by Shearer et al. (2013) for P. Explaining the Ti enrichment without some type of kinetic disequilibrium would require an increase in the partitioning of Ti between olivine and the melt or in the concentration of Ti in the residual melt by at least one order of magnitude. The profiles in other minor and trace elements are more similar to the Fo profiles with zoning affecting the outer 200 µm of megacrysts. Like Fe and Mg, these minor elements are mainly divalent, more compatible in olivine relative to Ti, and characterized by faster diffusivities relative to Ti4+ (Richter et al., 2003). We suggest that the Ti zoning preserved the initial size of a crystal overgrowth (outer rims, Fig. 3), which formed during a shallower and faster stage of crystallization. Atoms of the same element can also move within the crystal lattice in the absence of chemical gradient. This process, known as self- or tracer- diffusion (e.g. Chakraborty et al., 1994), would erase the anti-correlated Fe-Mg isotope fractionation over time if the megacrysts were maintained at high pressure and temperature. The presence of Fe-Mg isotope fractionation in the megacryst cores of NWA 1068 indicates that, while preceding the crystallization of the groundmass, the megacryst cores are not derived from the mantle or ancient crustal cumulates produced in an older magmatic system. Instead, as suggested by the composition of minor elements (Shearer et al., 2008, 2013), olivine megacrysts must be co-genetic with the groundmass. In summary, olivine megacrysts in NWA 1068 were formed by the crystallization of a melt genetically related to the melt that formed the groundmass. The isotope fractionation of megacryst cores indicates that their composition was progressively enriched in Fe by diffusion and that they likely grew slowly in a melt of evolving composition. The 50 µm outer rims crystallized faster than the rest of the megacryst. The Fo profiles of olivine megacrysts thus result from the simultaneous influence of crystal growth and diffusion.

231 4.3. Diffusion-crystallization models

In order to place quantitative constraints on the cooling rate of the different stages of NWA 1068 crystallization described in the above section and previously proposed in the literature (Filiberto et al., 2010; Shearer et al., 2013), we describe a model of simultaneous diffusion and crystal growth with variable successive cooling rates. We first illustrate how more simple models are inadequate to reproduce the complex chemical and isotope profiles of NWA 1068 olivines. Finally, we discuss the uncertainties associated with the model of combined diffusion and crystal growth.

4.3.1. Simple diffusion and crystallization models The methods used to simulate crystal growth and diffusion are first described separately. These simple models illustrate that crystallization or diffusion alone cannot reproduce the isotope and element profiles of olivine crystals in NWA 1068. However, once combined, they constitute the framework of the model of simultaneous growth and diffusion, which provides quantitative constraints on the cooling history of NWA 1068 (section 4.2.2). For the simple-diffusion model, we use the one-dimension diffusion equation with the diffusion coefficient (D in m2/s) varying as a function of the mole fraction of FeO or MgO (C) in olivine and the position (x) along the profile: HI(K,M) HO HI(K,M) HDI(K,M) = + Q [3] HM HK HK HKD The diffusion equation was solved numerically using the finite difference method (e.g. Costa et al., 2008). The Fe-Mg diffusivity was calculated, parallel to the c-axis, using the following expression (Dohmen et al., 2007; Dohmen and Chakraborty, 2007):

WX T/% AYTYYYd(eRTY*)∙f∙TY7+ Q = 10RS,AT ∙ V D [ ∙ 10](^_1RY.T) ∙ exp V− [ ()RBC TY7Z gh [4] where fO2 is the oxygen fugacity and P is the pressure (both in Pascals), T is the temperature in

Kelvin and XFe is the mole fraction of fayalite. The diffusivity of 56Fe (~92 % of total Fe isotopes) and 24Mg (~77 % of total Mg isotopes) are set equal to the Fe-Mg inter-diffusion coefficient (equation 4). The diffusivity of other isotopes is calculated using the following empirical formula (Richter et al., 1999) which is a modified version of Graham’s law for molten oxides and crystals:

232 i O8 BD = V [ [5] OD B8 where D1 and D2 are the diffusivity and M1 and M2 the masses of two isotopes of the same element. The exponent β is 0.5 in a gas but is known to be smaller in olivine crystals and in the range 0.08- 0.16 for Mg and 0.16-0.27 for Fe (Oeser et al., 2015; Sio et al. 2013). Results from Oeser et al. (2015) point towards the lower values and both studies are consistent with the estimated β-Fe/β- Mg ratio (≈ 2; Van Orman and Krawczynski, 2015). β-Mg and β-Fe were varied in a range of 0.08- 0.28 and 0.12-0.32, respectively. The initial composition of the olivine crystal is homogeneous (i.e. no chemical zoning), based on the assumption that megacrysts were fully equilibrated. To set this initial composition, the bulk composition of NWA 1068 is adjusted by subtracting or adding olivine incrementally until the bulk composition corresponds to the parental melt of the olivine of interest (e.g. Mg# = 47.1, in the case study of Fig. 7A). The parental melt of each olivine is thus assumed to vary from the NWA 1068 bulk composition by adding or subtracting an olivine component only. The parental melt is selected when it reaches equilibrium conditions (using a KD Fe-Mg olivine-liquid of 0.35; Filiberto and Dasgupta, 2011; Toplis, 2005) with the daughter olivine (e.g. Fo 71.8, Fig. 7A). The composition of the groundmass parental melt is calculated using the same technique, by removing olivine from the bulk composition of NWA 1068 until a composition in equilibrium with megacrysts outer rims is reached. The composition of the groundmass parental melt is then used to calculate the temperature of the system as the MgO content of olivine-saturated basaltic liquids varies linearly with the temperature (e.g. Sugawara, 2000). We used a set of unpublished experiments performed at LMV in a 1-atm gas mixing furnace from the bulk composition of LAR 06319 / NWA 1068 to parameterize the following linear relationship:

T (ºC) = MgOpqrsqt (wt. %) × 23.15 + 1022 [6]

The experiments were conducted at fO2 conditions 2.5 log-unit under the fayalite-magnetite-quartz (FMQ) buffer. This equation is consistent (within ± 30 ºC) with the results of published 1-atm experiments performed on similar bulk compositions and under similar redox conditions (Filiberto et al., 2010; Filiberto et al., 2008; Monders et al., 2007). The Fe-Mg diffusivity is then calculated using equation [4] and the temperature of equation [6]. The pressure is set to 0.1 MPa and the fO2 is set at the FMQ buffer, the crystallization conditions of NWA 1068 groundmass (Herd, 2006). The diffusion of Fe and Mg in basaltic melts is several orders of magnitude faster than Fe-Mg

233 interdiffusion in olivine (Zhang et al. 2010). Therefore, we only model diffusion in olivine megacrysts and consider that diffusion in the melt is infinitely fast. To match the forsterite zoning of olivine megacrysts, the diffusion process has to be interrupted before the core is affected. In consequence, no isotope fractionation is generated in the core (Fig 7A), in disagreement with the strong isotope fractionation relative to the bulk shergottites measured in megacryst cores. For the other simple model, we assume that olivine crystallization is infinitely fast and that the effect of diffusion is insignificant. The zoning in Fo content appears spontaneously by crystal growth. Olivine grows from a fixed volume of evolving liquid that is progressively depleted in MgO relative to FeO. The chosen volume of liquid (density of 2.8 g/cm3) is converted into a mass. Like in the previous model, the composition of the initial liquid is obtained by removing incrementally olivine from NWA 1068 bulk composition until it is in equilibrium with a specific megacryst core. The starting temperature is calculated using equation [6]. The temperature is then decreased and, at each step, the MgO content of the residual liquid is recalculated. The amount of olivine crystallized at each increment is calculated by mass balance from the amount of olivine needed to accommodate the change of MgO content of the liquid. The mass transferred from the liquid to the olivine is converted into a volume assuming an average density of 3.6 g/cm3 for olivine. Finally, the volume is converted into the corresponding segment length and added to Fe and Mg profiles. The crystal growth rate is controlled by the mass balance calculation but can be approximated by a simple growth law. If a constant mass of material is added to a sphere per temperature increment, the growth rate depends on to the radius of the sphere (R): tg T th = { [7] tM gD tM C scales with the mass of olivine added per increment of temperature, which depends on the initial volume of liquid. The model described here slightly deviates from equation [7]. As the mass of residual liquid is progressively transferred to olivine, the mass of olivine necessary to accommodate the linear decrease in MgO content of the liquid (and the decrease in temperature) is ~40 % smaller at the end of crystallization than at the beginning. Therefore, C is not strictly a constant but also decreases as crystallization advances. The growth law also implies that crystal growth only occurs when the temperature is decreasing, which is not necessarily the case. Undercooling, surface energy

234 minimization and diffusion-limited transport are a few examples of processes that can induce crystal growth at constant temperature (also see Sio and Dauphas 2016). As the cooling rate is first assumed to be infinitely fast, diffusion is insignificant and no isotope fractionation can develop in olivine megacrysts (Fig. 7B). The equilibrium fractionation between melt and olivine is assumed to be negligible as well as potential fractionation due to crystal growth kinetics. Equilibrium fractionation of Fe isotopes can occur in the melt between Fe2+ and Fe3+ and with changes in oxygen fugacity but is expected to be negligible in a basaltic melt under reducing conditions and is thus not taken into account (Dauphas et al., 2014). Like the simple diffusion model, this model cannot reproduce the isotope profiles of NWA 1068 olivine megacrysts.

4.3.2. Simultaneous crystal growth and diffusion model As discussed in section 4.1 and illustrated in the above section, forsterite and isotope profiles of NWA 1068 olivine megacrysts cannot be reproduced by simple diffusion or crystal growth models. This is not a surprise considering the complex history of NWA 1068, which formed during several stages of crystallization. Diffusion and crystal growth are the two endmember processes that lead to crystal zoning in a closed system. When olivine megacrysts crystallize slowly, both processes are likely to have a significant influence on the final zoning. For this reason, both simple models described in section 4.2.1 are combined in a crystallization and diffusion model. The simultaneous crystallization and diffusion model is divided in three stages characterized by different cooling rates (CR). During the first stage, the temperature is decreased slowly (CR1) to allow megacryst cores to be affected by diffusion during their growth. The second stage is characterized by a much faster cooling rate and corresponds to the crystallization stage during which the groundmass and the outermost 40-60 µm rim of olivine megacrysts crystallize (CR2). The third and final stage is the final cooling of the sample after crystallization stopped and until the temperature drops below 700 ºC, where diffusion becomes negligible (CR3). The cooling rate of this last stage is chosen to be equal to or greater than CR2. The fO2, which is one of the key variables in the diffusivity equation [4], is changed from FMQ-2.5 to FMQ between stage 1 and stage 2 to account for the change of fO2 recorded by NWA 1068 (Herd, 2006). During stages 2 and 3, the cooling rates are chosen to be sufficiently fast to preserve a sub-homogenous forsterite profile in the megacryst core but slow enough to allow diffusion to affect the outer 100-200 µm of the crystal.

235 The initial composition of the parental melt is calculated by adding an arbitrary olivine component to the melt in equilibrium with the olivine megacryst. Then, the initial temperature is calculated from the MgO content of this liquid (equation [6]). The initial volume of liquid is adjusted such that ~20% of the liquid has crystallized at the end of stage 1. This implies that all megacrysts, regardless of their size, crystallize from the same fraction of melt. An olivine fraction of 20% is close the actual modal composition of NWA 1068. Once the model parameters are initialized, the temperature is progressively decreased following the 3 fixed cooling rates by increment of time (10 s) and the MgO content of the liquid is continuously adjusted. The fO2 is maintained 2.5 log units under the FMQ buffer (or on the FMQ buffer for stages 2 and 3) and recalculated to reflect the change in temperature (Huebner, 1971; O'Neill, 1987): ]Y%Ç$.f eRT log ~O = 82.76 − + 0.092 − 10.6199 × ln Ö + 0.00484 A h h [8] with fO2 and P in bars and T in Kelvin. The composition and amount of olivine crystallized are calculated to accommodate the change in MgO of the liquid (equation [6]). For each time increment, the diffusivity of each isotope along the c-axis is recalculated (equation [4] and [5]) to account for the changing Fo content, temperature, and redox conditions (equation [8]). The volume of residual liquid between stages 1 and 2 can be decreased while keeping the composition unchanged. Modifying the volume of residual liquid could be justified by the onset of pyroxene crystallization in the groundmass. As pyroxene and smaller olivine crystals nucleate and start to form the groundmass, they compete with the growth of megacrysts. This model is more realistic than simple diffusion and crystal growth model and produces profiles that match the observations much better (Fig. 7C, 8, EA4). In particular, the relatively flat and fractionated isotope profiles in olivine megacrysts can now be reproduced. The cooling rate of each stage provides a duration estimate for the main crystallization steps. The model can also be used to constrain the initial composition of olivine that first crystallized from the parental melts and thus constrain the composition of the parental melt itself. The model parameters used for each simulation are listed in Table 2. Table 3 summarizes the key outputs of the model for several

236 simulations that produced a good fit. The corresponding profiles are available in Fig. 7C, Fig. 8 and Fig. S2.

4.3.3. Model uncertainties and limitations The model of simultaneous crystallization and diffusion described above contains many different parameters that can be independently adjusted (Table 2). Changing several parameters simultaneously can balance the effect of each modification in such a way that the simulation outputs are very similar (Table 3). The isotope and forsterite profiles of each megacryst can thus be reproduced with slightly different parameters and each simulation does not lead to entirely unique results, especially once the analytical uncertainties are taken into account. In Tables 2 and 3, the input parameters and the results of a set of simulations that reproduce reasonably well the Fo and isotope profiles are presented for 7 different olivine megacrysts. The parameter space that can reproduce the isotope and Fo profiles of megacryst cores is fairly limited. The cooling rate during core crystallization (CR1) can only be changed by a factor of ~5, while modifying other parameters, to maintain a strong isotope fractionation in megacryst cores. To illustrate this, we include, in Tables 2-3, two simulations (olB2_bf and olB2_bs) that only differ from olB2_b (Fig. 6C) by the cooling rate CR1. The cooling rate was either increased or decreased by one order of magnitude. In both cases, the isotope fractionation in the megacryst core is much lower than what was analyzed. The main uncertainty concerns the relative timescale of rim crystallization (CR2) vs. diffusion-only cooling (CR3). For example, a faster cooling rate for the diffusion-only stage can be counterbalanced by a slower cooling rate for rim crystallization. The most conservative assumption is to keep both cooling rates equal and calculate the elapsed time between the onset of rim crystallization and the final emplacement, defined when the temperature drops under 700 ºC. The lack of constraints on the unfractionated composition of Mg isotopes (δ26Mg -0.25 or - 0.4 ‰, see section 3.3) is another source of uncertainty. However, most of this uncertainty is reported on the β and does not affect significantly other model parameters. A starting δ26Mg of - 0.4 ‰ and a β-Mg of 0.24-0.21 (dark green in Fig. 7, 8) produce the same profiles as a δ26Mg of - 0.25 ‰ and a β-Mg of 0.17-0.15 (light green in Fig. 7, 8). It is interesting that in this last case the β-Fe/β-Mg is ~2 and equal to the ratio of previous studies of in-situ Fe-Mg isotope measurements in olivine (Sio et al., 2013; Oeser et al., 2015; Van Orman and Krawczynski, 2015).

237 In order to obtain a deeper understanding of model uncertainties for stage 1, the parameter space has been randomly explored for the crystallization of 3 olivine megacryst cores (olB1, olB2 and olB7). The initial composition of the core, β-Fe, β-Mg and the cooling rate were all randomly sampled. The equilibrium Mg isotopes composition was fixed to either δ26Mg -0.25 or -0.4 ‰. About 4500 simulations were performed, 122 of which were considered to be successful (Table 4). Successful simulations produced a final Fo content equal, within ±0.5 Fo, to the composition of megacryst cores and a final δ26Mg / δ56Fe within ±0.12 ‰ of their isotopic compositions. Only the simulations that satisfied simultaneously those 3 criteria were selected. The average and standard deviation of each model parameter for the different sets of simulations are reported in Table 4. This analysis also constrains more quantitatively the uncertainty associated with the model outputs. Overall, the results are consistent with the simulations shown in Fig. 7 and 8 and summarized in Table 2 and 3, for which the parameters were manually adjusted. The parameters associated with the crystallization of megacryst rims and subsequent cooling (stages 2 and 3) were not extensively explored and therefore remain poorly constrained. Due to the large beam size (40 µm), only 1 to 2 measurements could be performed in megacryst rims. Isotope fractionation in megacryst rims cannot be sufficiently resolved to justify the random sampling of the large number of parameters associated with stages 2 and 3. Stages 2 and 3 have a minimal influence on the chemical and isotopic composition at the center of olivine megacrysts (Fig. 7C and Fig. S1) and can be ignored to constrain the parameters relevant for stage 1. The model also assumes that crystal growth is not interrupted between crystallization of the core and the rim and that the temperature decreased linearly during both crystallization stages. The implicit assumption is that no replenishment of the magma chamber significantly changed the composition of the residual melt or increased the temperature of the magma. The megacrysts were never partly resorbed and diffusion never paused (equilibrium) or reversed (bulk transfer of Mg from the liquid to the olivine). If the crystallization history deviated significantly from those assumptions, the modeled durations calculated here would represent lower limits. The complex P (and Al) zoning of olivine megacrysts that is observed in an increasing number of olivine-phyric shergottites (Ennis and McSween, 2014; Liu et al., 2016; Peslier et al., 2010; Shearer et al., 2013) is sometimes interpreted as evidence for initial dendritic growth (Ennis and McSween, 2014; Welsch et al., 2014) although solute trapping resulting from growth rate changes and the slow diffusion of P in the melt are also possible (e.g. Shearer et al., 2013; Watson and

238 Müller, 2009). There is a possibility that olivine dendrites could have formed a skeleton that would have been subsequently filled as olivine continued to crystallize (e.g. Shea et al., 2015). In such cases, olivine megacrysts do not fully crystallize in a concentric manner and melt inclusions can easily be entrapped. The progressive filling of a dendritic skeleton would most likely results in more effective diffusion, as the liquid-crystal interface would have larger total surface area. This source of uncertainty is difficult to quantify but it is clear that any deviation of the olivine megacrysts from the morphology of a perfect sphere would make diffusion more efficient and shorten the duration of megacryst core crystallization (stage 1). Due to the absence of correlation between the diffusivity and chemical zoning (Fig. 6), the crystallographic orientation was not taken into account to calculate diffusivities in the model. The diffusivity is calculated along the c-axis and we consider that the highest diffusivity is the one controlling the composition of megacryst cores. This could also lead to underestimate the duration of stages 1 to 3. A 3-D model would be necessary to test this model assumption.

4.4. Timescales of NWA 1068 crystallization

The simulation results of the diffusion and crystallization model confirm that megacryst cores have been rehomogenized and that the current forsterite content is 3.2 (±1.3) mol% lower than the nuclei that grew to form olivine megacrysts (2σ on olB1, Table 4). Sio and Dauphas (2016) recently reinterpreted the results of Sio et al. (2013) with a model of simultaneous diffusion and crystallization that share many similarities with the model described here. Interestingly, they also concluded that olivine megacrysts from Kilauea Iki lava lake were rehomogenized and that the Fo content of one olivine core could have been lowered by ~3 units of Fo. It also appears that the most magnesium-rich olivine megacrysts in NWA 1068 were also the richest in Fo before rehomegenization (Fig. 9). Prior to rehomogenization, olivine had a range of Fo contents (74.5- 78), which suggests that they started to crystallize at different times and that nucleation has initially been continuous. All the successful simulations require a relatively slow cooling rate of 0.05-0.13 ºC/day (Table 2-4) during the crystallization of megacryst cores. The megacryst rims crystallized much faster with a cooling rate of 5-10 ºC/day (Table 3). However, this cooling rate is more difficult to constrain and, depending on the model assumptions, the total uncertainty could be larger than one order of magnitude. According to our model, the total time elapsed between the beginning of

239 olivine nucleation and the final emplacement of the magma is 1.8-6.4 years (2σ on olB1, Table 4), which is similar to the crystal residence time in many terrestrial basalts (e.g. de Maisonneuve et al., 2016). If crystallization started near the base of the crust (40-85 km on Mars; Genova et al., 2016; Zuber, 2001), we can calculate the average ascent rate of NWA 1068 magma: 1.9×10-4 to 1.5×10-3 m/s (0.7-5.4 m/h). Those values are low relative to average magma ascent rates (e.g. Rutherford, 2008; Wilson and Head, 1981) and are probably indicative that the ascent was not continuous but, instead, that the magma pooled in the martian crust. As the magma traveled upward, some megacrysts collided, stuck together due to surface tension, and intergrew to form glomerocryts (Fig. 2). This process is known as synneusis (e.g. Vance, 1969) and is characteristic of magmatic turbulence during transport or convection in a magma chamber. The pyroxene crystals with the highest Mg concentration (Mg# 71) are not in equilibrium with the most Mg-rich olivine but with Fo68 (assuming a Kd olivine-pigeonite of 1.2; Longhi and

Pan, 1989). Fo68 corresponds approximately to the composition of olivine megacrysts before the outermost part of the rim (~50 µm) starts to crystallize in our model (Fig. 7C). Diffusion continually affected the outer rims and the olivine of the groundmass as they crystallized, lowering their Fo contents. On the other hand, we propose that the initial composition of pyroxene was mostly preserved as Fe-Mg interdiffusion coefficients in pyroxenes are 2 orders of magnitude lower than in olivine (e.g. Dohmen et al., 2016). Therefore, pyroxene and olivine from the groundmass started to crystallize simultaneously with megacryst outer rims. Plagioclase quickly joined the crystallizing assemblage and the viscosity of the magma increased drastically as the melt fraction dropped under 50% (Marsh, 1981). The time elapsed during the second and third stages of the model (100±40 days) likely encompasses the final step of ascent and an episode of stationary cooling when the viscosity reached a critical value.

4.5. Parental melt of NWA 1068 and implications for the origin of enriched shergottites

The parental melts of several olivine-phyric shergottites have been estimated using the “phenocryst matching test” (BVSP, 1981; Filiberto and Dasgupta, 2011). This test assumes that phenocryst cores represent the composition of the first crystal that formed in the parental melt. As the Fe-Mg equilibrium between silicate melts and olivine is known (Filiberto and Dasgupta, 2011; Toplis, 2005), the Mg# of the parental melt can be predicted (Fig. 10). If the Mg# of the bulk rock is equal to this prediction, the bulk rock composition is assumed to be representative of the parental

240 melt. While some shergottites have been recognized as possible parental melts (e.g. Mussewhilte et al. 2006; Gross et al., 2011), NWA 1068 has been suspected to contain up to 20% of accumulated olivine (Filiberto et al., 2010), which represents the total fraction of olivine in NWA 1068 (Barrat et al., 2002). However, this test is only valid if we assume perfect fractional crystallization with no diffusive chemical diffusion during crystal growth. Our simulations show that even the most Mg- rich megacrysts were rehomogenized by 3.2±1.3 Fo units. In consequence, accumulation of olivine is no longer required to reach Fe-Mg equilibrium and NWA 1068 could represent the composition of the parental melt (Fig. 10). The olivine compositions in equilibrium with primary martian mantle melts vary from Fo76 for melting of the primitive martian mantle (Collinet et al., 2015) to Fo85 for the mantle source of the depleted shergottites (Musselwhite et al., 2006). The parental melt of NWA

1068 could have been in equilibrium with Fo80±1 olivines and would represent a primary melt of a mantle source with an intermediate Mg#. Alternatively, the parental melt could have been affected by early crystallization of olivine, which was removed from the magma. It this case, NWA 1068 would derive from a mantle source with an even higher Mg#, potentially equal to the Mg# of depleted shergottites. If the parental melt of NWA 1068 is near-primary, it should be saturated with mineral phases characteristic of a mantle residue. Experiments performed from the whole-rock composition of NWA 1068 (Filiberto et al. 2010) have identified a multiple saturation point (MSP) where olivine and orthopyroxene appear on the liquidus of NWA 1068 at 1.7 GPa and 1520 ºC. Under those conditions, the primitive martian mantle is expected to contain only olivine and orthopyroxene

(Collinet et al. 2015). The CaO/Al2O3 of NWA 1068 is super-chondritic and rules out that the parental melt derives from the primitive martian mantle by equilibrium melting. Nonetheless, the parental melt of NWA 1068 could be near-primary and derived from a refractory mantle source which was affected by several events of melting. The meteorite LAR 06319 displays many similarities with NWA 1068. Both are enriched olivine-phyric shergottites, have the same bulk composition (Barrat et al., 2002; Basu Sarbadhikari et al., 2009), the same range of Fo content (~72-77) in olivine megacryst cores and rims (~50-58), and both crystallized over the same range of redox conditions (Herd, 2003, 2006; Peslier et al., 2010). Megacrysts in LAR 06319 were probably rehomogenized (Balta et al., 2013) and affected by a similar cooling history. If this is the case, the bulk rock composition of LAR 06319 is also likely to reach Fe-Mg equilibrium once the composition of megacryst cores is corrected for

241 diffusion. The main difference is the larger average crystal size of LAR 06319 groundmass relative to NWA 1068, which suggest that both shergottites may have crystallized at slightly different conditions but in the same overall environment. We thus infer that both NWA 1068 and LAR 06319 originated from similar parental melts and that the olivine megacrysts crystallized under similar conditions. Just as Yamato 980459 and NWA 5789 likely represent parental melts from which other depleted shergottites are derived (Gross et al., 2011; Rapp et al., 2013), there is a strong possibility that the whole group of enriched shergottites originates from a parental melt similar in composition to LAR 06319 and NWA 1068. Whether the enriched signature is already acquired in the mantle source of this parent melt or was overprinted later is out of the scope of this study.

5. Conclusions

The anti-correlated fractionation of Fe and Mg isotopes indicates that olivine megacrysts in NWA 1068 formed during an early, 2 to 6 year-long, stage of crystallization and were progressively rehomogenized by the evolving liquid in which they crystallized. However, the olivine megacrysts did not significantly accumulate and the bulk rock composition of NWA 1068 is likely representative of the parental melt. During a second, ~100 day-long, stage of crystallization, the residual melt formed the groundmass and megacryst rims as the magma was transported to its shallower site of emplacement. NWA 1068 and LAR 06319 likely share the same source and crystallization history. Their bulk compositions are identical and they are good candidates for the parental melt from which the other enriched shergottites are derived. This parental melt could be near-primary and in equilibrium with a refractory martian mantle (Mg# > 80) composed of olivine and orthopyroxene.

6. References

Anand, M., Russell, S.S., Blackhurst, R.L., Grady, M.M., 2006. Searching for signatures of life on Mars: an Fe-isotope perspective. Philosophical Transactions of the Royal Society B: Biological Sciences 361, 1715-1720. Aoudjehane, H.C., Avice, G., Barrat, J.-A., Boudouma, O., Chen, G., Duke, M.J.M., Franchi, I.A., Gattacceca, J., Grady, M.M., Greenwood, R.C., Herd, C.D.K., Hewins, R., Jambon, A., Marty, B., Rochette, P., Smith, C.L., Sautter, V., Verchovsky, A., Weber, P., Zanda, B., 2012. Tissint Martian Meteorite: A Fresh Look at the Interior, Surface, and Atmosphere of Mars. Science 338, 785-788.

242 Balta, J.B., Sanborn, M., McSween, H.Y., Wadhwa, M., 2013. Magmatic history and parental melt composition of olivine-phyric shergottite LAR 06319: Importance of magmatic degassing and olivine antecrysts in Martian magmatism. Meteoritics & Planetary Science, 1359-1382. Balta, J.B., Sanborn, M.E., Udry, A., Wadhwa, M., McSween, H.Y., 2015. Petrology and trace element geochemistry of Tissint, the newest shergottite fall. Meteoritics & Planetary Science 50, 63-85. Barrat, J.A., Jambon, A., Bohn, M., Gillet, P., Sautter, V., Gopel, C., Lesourd, M., Keller, F., 2002. Petrology and chemistry of the picritic shergottite North West Africa 1068 (NWA 1068). Geochimica Et Cosmochimica Acta 66, 3505-3518. Basu Sarbadhikari, A., Day, J.M.D., Liu, Y., Rumble, D., Taylor, L.A., 2009. Petrogenesis of olivine- phyric shergottite Larkman Nunatak 06319: Implications for enriched components in martian basalts. Geochimica Et Cosmochimica Acta 73, 2190-2214. Basu Sarbadhikari, A., Goodrich, C.A., Liu, Y., Day, J.M.D., Taylor, L.A., 2011. Evidence for heterogeneous enriched shergottite mantle sources in Mars from olivine-hosted melt inclusions in Larkman Nunatak 06319. Geochimica Et Cosmochimica Acta 75, 6803-6820. Bouvier, A., Blichert-Toft, J., Albarede, F., 2009. Martian meteorite chronology and the evolution of the interior of Mars. Earth and Planetary Science Letters 280, 285-295. Buening, D.K., Buseck, P.R., 1973. Fe-Mg lattice diffusion in olivine. Journal of Geophysical Research 78, 6852-6862. Bunch, T.E., Irving, A.J., Wittke, J.H., Rumble, D., III, Korotev, R.L., Gellissen, M., Palme, H., 2009. Petrology and composition of Northwest Africa 2990; a new type of fine-grained, enriched, olivine-phyric shergottite. 40th Lunar and Planetary Science Conference, abstract 2274. BVSP, 1981. Experimental petrology of basalts and their source rocks. Basaltic Volcanism on the Terrestrial Planets. Pergamon Press, Inc, New York, 1286 pp. Chakrabarti, R., Jacobsen, S.B., 2010. The isotopic composition of magnesium in the inner Solar System. Earth and Planetary Science Letters 293, 349-358. Chakraborty, S., Farver, J.R., Yund, R.A., Rubie, D.C., 1994. Mg tracer diffusion in synthetic forsterite and San Carlos olivine as a function of P, T and fO2. Phys Chem Minerals 21, 489-500. Collinet, M., Médard, E., Charlier, B., Vander Auwera, J., Grove, T.L., 2015. Melting of the primitive martian mantle at 0.5–2.2 GPa and the origin of basalts and alkaline rocks on Mars. Earth and Planetary Science Letters 427, 83-94. Costa, F., Chakraborty, S., 2004. Decadal time gaps between mafic intrusion and silicic eruption obtained from chemical zoning patterns in olivine. Earth and Planetary Science Letters 227, 517-530. Costa, F., Dohmen, R., Chakraborty, S., 2008. Time Scales of Magmatic Processes from Modeling the Zoning Patterns of Crystals. Reviews in Mineralogy and Geochemistry 69, 545-594. Costa, F., Dungan, M., 2005. Short time scales of magmatic assimilation from diffusion modeling of multiple elements in olivine. Geology 33, 837-840. Dauphas, N., Roskosz, M., Alp, E.E., Neuville, D.R., Hu, M.Y., Sio, C.K., Tissot, F.L.H., Zhao, J., Tissandier, L., Médard, E., Cordier, C., 2014. Magma redox and structural controls on iron isotope variations in Earth's mantle and crust. Earth and Planetary Science Letters 398, 127-140. Dauphas, N., Teng, F.-Z., Arndt, N.T., 2010. Magnesium and iron isotopes in 2.7 Ga Alexo komatiites: Mantle signatures, no evidence for Soret diffusion, and identification of diffusive transport in zoned olivine. Geochimica et Cosmochimica Acta 74, 3274-3291. de Maisonneuve, C.B., Costa, F., Huber, C., Vonlanthen, P., Bachmann, O., Dungan, M.A., 2016. How do olivines record magmatic events? Insights from major and trace element zoning. Contributions to Mineralogy and Petrology 171, 1-20. Debaille, V., Yin, Q.Z., Brandon, A.D., Jacobsen, B., 2008. Martian mantle mineralogy investigated by the 176Lu–176Hf and 147Sm–143Nd systematics of shergottites. Earth and Planetary Science Letters 269, 186-199. Dohmen, R., Becker, H.-W., Chakraborty, S., 2007. Fe–Mg diffusion in olivine I: experimental determination between 700 and 1,200°C as a function of composition, crystal orientation and oxygen fugacity. Phys Chem Minerals 34, 389-407.

243 Dohmen, R., Chakraborty, S., 2007. Fe–Mg diffusion in olivine II: point defect chemistry, change of diffusion mechanisms and a model for calculation of diffusion coefficients in natural olivine. Phys Chem Minerals 34, 409-430. Dohmen, R., Ter heege, J.H., Becker, H.-W., Chakraborty, S., 2016. Fe-Mg interdiffusion in orthopyroxene. American Mineralogist 101, 2210-2221. Dreibus, G., Wanke, H., 1985. Mars, a volatile-rich planet. Meteoritics 20, 367-381. Ennis, M.E., McSween, H.Y., 2014. Crystallization kinetics of olivine-phyric shergottites. Meteoritics & Planetary Science 49, 1440-1455. Filiberto, J., Chin, E., Day, J.M.D., Franchi, I.A., Greenwood, R.C., Gross, J., Penniston-Dorland, S.C., Schwenzer, S.P., Treiman, A.H., 2012. Geochemistry of intermediate olivine-phyric shergottite Northwest Africa 6234, with similarities to basaltic shergottite Northwest Africa 480 and olivine- phyric shergottite Northwest Africa 2990. Meteoritics & Planetary Science, 1256-1273. Filiberto, J., Dasgupta, R., 2011. Fe2+-Mg partitioning between olivine and basaltic melts: Applications to genesis of olivine-phyric shergottites and conditions of melting in the Martian interior. Earth and Planetary Science Letters 304, 527-537. Filiberto, J., Musselwhite, D.S., Gross, J., Burgess, K., Le, L., Treiman, A.H., 2010. Experimental petrology, crystallization history, and parental magma characteristics of olivine-phyric shergottite NWA 1068: Implications for the petrogenesis of "enriched" olivine-phyric shergottites. Meteoritics & Planetary Science 45, 1258-1270. Filiberto, J., Treiman, A.H., Le, L., 2008. Crystallization experiments on a Gusev Adirondack basalt composition. Meteoritics & Planetary Science 43, 1137-1146. Genova, A., Goossens, S., Lemoine, F.G., Mazarico, E., Neumann, G.A., Smith, D.E., Zuber, M.T., 2016. Seasonal and static gravity field of Mars from MGS, Mars Odyssey and MRO radio science. Icarus 272, 228-245. Goodrich, C.A., 2002. Olivine-phyric martian basalts: A new type of shergottite. Meteoritics & Planetary Science 37, B31-B34. Goodrich, C.A., 2003. Petrogenesis of olivine-phyric shergottites Sayh Al Uhaymir 005 and elephant moraine A79001 lithology A. Geochimica Et Cosmochimica Acta 67, 3735-3772. Gross, J., Filiberto, J., Herd, C.D.K., Daswani, M.M., Schwenzer, S.P., Treiman, A.H., 2013. Petrography, mineral chemistry, and crystallization history of olivine-phyric shergottite NWA 6234: A new melt composition. Meteoritics & Planetary Science, n/a-n/a. Gross, J., Treiman, A.H., Filiberto, J., Herd, C.D.K., 2011. Primitive olivine-phyric shergottite NWA 5789: Petrography, mineral chemistry, and cooling history imply a magma similar to Yamato- 980459. Meteoritics & Planetary Science 46, 116-133. Herd, C.D.K., 2003. The oxygen fugacity of olivine-phyric martian basalts and the components within the mantle and crust of Mars. Meteoritics & Planetary Science 38, 1793-1805. Herd, C.D.K., 2006. Insights into the redox history of the NWA 1068/1110 martian basalt from mineral equilibria and vanadium oxybarometry. American Mineralogist 91, 1616-1627. Huebner, J.S., 1971. Buffering Techniques for Hydrostatic Systems at Elevated Pressures, in: Ulmer, G.C. (Ed.), Research Techniques for High Pressure and High Temperature. Springer Berlin Heidelberg, Berlin, Heidelberg, 123-177. Jackson, S.E., 2008. LAMTRACE data reduction software for LA-ICP-MS., in: Sylvester, P. (Ed.), Laser ablation-ICP-mass spectrometry in the Earth sciences: Current practices and outstanding issues. Mineralogical Association of Canada Short Course Series, 305–307. Jambon, A., 1980. Isotopic fractionation: A kinetic model for crystals growing from magmatic melts. Geochimica et Cosmochimica Acta 44, 1373-1380. Jambon, A., Sautter, V., Barrat, J.A., Gattacceca, J., Rochette, P., Boudouma, O., Badia, D., Devouard, B., 2016. Northwest Africa 5790: Revisiting nakhlite petrogenesis. Geochimica et Cosmochimica Acta 190, 191-212. Jochum, K.P., Willbold, M., Raczek, I., Stoll, B., Herwig, K., 2005. Chemical Characterisation of the USGS Reference Glasses GSA-1G, GSC-1G, GSD-1G, GSE-1G, BCR-2G, BHVO-2G and BIR-

244 1G Using EPMA, ID-TIMS, ID-ICP-MS and LA-ICP-MS. Geostandards and Geoanalytical Research 29, 285-302. Jones, J.H., 2015. Various aspects of the petrogenesis of the Martian shergottite meteorites. Meteoritics & Planetary Science 50, 674-690. Koizumi, E., Mikouchi, T., Monkawa, A., Miyamoto, M., 2004. Origin of olivine in megacrysts and the groundmass crystallization of the Dar al Gani 476 shergottite. Antarctic meteorite research 17, 84- 96. Lazarov, M., Horn, I., 2015. Matrix and energy effects during in-situ determination of Cu isotope ratios by ultraviolet-femtosecond laser ablation multicollector inductively coupled plasma mass spectrometry. Spectrochimica Acta Part B: Atomic Spectroscopy 111, 64-73. Liu, Y., Baziotis, I.P., Asimow, P.D., Bodnar, R.J., Taylor, L.A., 2016. Mineral chemistry of the Tissint meteorite: Indications of two-stage crystallization in a closed system. Meteoritics & Planetary Science, n/a-n/a. Longerich, H.P., Jackson, S.E., Gunther, D., 1996. Inter-laboratory note. Laser ablation inductively coupled plasma mass spectrometric transient signal data acquisition and analyte concentration calculation. Journal of Analytical Atomic Spectrometry 11, 899-904. Longhi, J., Pan, V., 1989. The parent magmas of the SNC meteorites. Procedings of the 19th Lunar and Planetary Science Conference, 451-464. Magna, T.H., Y.; Teng, F.-Z.; Mezger, K., 2017. Magnesium isotope systematics of SNC meteorites: evidence for crystal fractionation or surface process? 48th Lunar and Planetary Science Conference, abstract 1795. Marsh, B.D., 1981. On the crystallinity, probability of occurrence, and rheology of lava and magma. Contributions to Mineralogy and Petrology 78, 85-98. Marsh, B.D., 1988. Crystal size distribution (CSD) in rocks and the kinetics and dynamics of crystallization. Contributions to Mineralogy and Petrology 99, 277-291. McSween, H.Y., Jarosewich, E., 1983. Petrogenesis of the Elephant Moraine A79001 meteorite: Multiple magma pulses on the shergottite parent body. Geochimica Et Cosmochimica Acta 47, 1501-1513. Mikouchi, T., Miyamoto, M., 2002. Olivine cooling rate of the northwest africa 1068 shergottite. 33th Lunar and Planetary Science Conference, abstract 1346. Mikouchi, T., Miyamoto, M., McKay, G.A., 2001. Mineralogy and petrology of the Dar al Gani 476 martian meteorite: Implications for its cooling history and relationship to other shergottites. Meteoritics & Planetary Science 36, 531-548. Miyamoto, M., Mikouchi, T., Satake, W., Koizumi, E., Kaiden, H., 2010. The cooling rate of several olivine-phyric shergottites on the basis of Fe-Mg zoning in olivine. 41st Lunar and Planetary Science Conference, abstract 1143. Monders, A.G., Medard, E., Grove, T.L., 2007. Phase equilibrium investigations of the Adirondack class basalts from the Gusev plains, Gusev crater, Mars. Meteoritics & Planetary Science 42, 131-148. Moser, D.E., Chamberlain, K.R., Tait, K.T., Schmitt, A.K., Darling, J.R., Barker, I.R., Hyde, B.C., 2013. Solving the Martian meteorite age conundrum using micro-baddeleyite and launch-generated zircon. Nature 499, 454-457. Musselwhite, D.S., Dalton, H.A., Kiefer, W.S., Treiman, A.H., 2006. Experimental petrology of the basaltic shergottite Yamato-980459: Implications for the thermal structure of the Martian mantle. Meteoritics & Planetary Science 41, 1271-1290. Norman, M.D., Yaxley, G.M., Bennett, V.C., Brandon, A.D., 2006. Magnesium isotopic composition of olivine from the Earth, Mars, Moon, and pallasite parent body. Geophysical Research Letters 33, L15202. O'Neill, H.S.C., 1987. Quartz-fayalite-iron and quartz-fayalite-magnetite equilibria and the free energy of formation of fayalite (Fe2SiO4) and magnetite (Fe3O4). American Mineralogist 72, 67-75. Oeser, M., Dohmen, R., Horn, I., Schuth, S., Weyer, S., 2015. Processes and time scales of magmatic evolution as revealed by Fe–Mg chemical and isotopic zoning in natural olivines. Geochimica et Cosmochimica Acta 154, 130-150.

245 Oeser, M., Weyer, S., Horn, I., Schuth, S., 2014. High-Precision Fe and Mg Isotope Ratios of Silicate Reference Glasses Determined In Situ by Femtosecond LA-MC-ICP-MS and by Solution Nebulisation MC-ICP-MS. Geostandards and Geoanalytical Research 38, 311-328. Peslier, A.H., Hnatyshin, D., Herd, C.D.K., Walton, E.L., Brandon, A.D., Lapen, T.J., Shafer, J.T., 2010. Crystallization, melt inclusion, and redox history of a Martian meteorite: Olivine-phyric shergottite Larkman Nunatak 06319. Geochimica Et Cosmochimica Acta 74, 4543-4576. Peters, T.J., Simon, J.I., Jones, J.H., Usui, T., Moriwaki, R., Economos, R.C., Schmitt, A.K., McKeegan, K.D., 2015. Tracking the source of the enriched martian meteorites in olivine-hosted melt inclusions of two depleted shergottites, Yamato 980459 and Tissint. Earth and Planetary Science Letters 418, 91-102. Poitrasson, F., Freydier, R., 2005. Heavy iron isotope composition of granites determined by high resolution MC-ICP-MS. Chemical Geology 222, 132-147. Poitrasson, F., Halliday, A.N., Lee, D.-C., Levasseur, S., Teutsch, N., 2004. Iron isotope differences between Earth, Moon, Mars and Vesta as possible records of contrasted accretion mechanisms. Earth and Planetary Science Letters 223, 253-266. Rapp, J.F., Draper, D.S., Mercer, C.M., 2013. Anhydrous liquid line of descent of Yamato-980459 and evolution of Martian parental magmas. Meteoritics & Planetary Science, 1780-1799. Richter, F., Chaussidon, M., Mendybaev, R., Kite, E., 2016. Reassessing the cooling rate and geologic setting of Martian meteorites MIL 03346 and NWA 817. Geochimica et Cosmochimica Acta 182, 1-23. Richter, F.M., Davis, A.M., DePaolo, D.J., Watson, E.B., 2003. Isotope fractionation by chemical diffusion between molten basalt and rhyolite. Geochimica et Cosmochimica Acta 67, 3905-3923. Richter, F.M., Liang, Y., Davis, A.M., 1999. Isotope fractionation by diffusion in molten oxides. Geochimica et Cosmochimica Acta 63, 2853-2861. Rutherford, M.J., 2008. Magma Ascent Rates. Reviews in Mineralogy and Geochemistry 69, 241-271. Schmidt, M.E., McCoy, T.J., 2010. The evolution of a heterogeneous Martian mantle: Clues from K, P, Ti, Cr, and Ni variations in Gusev basalts and shergottite meteorites. Earth and Planetary Science Letters 296, 67-77. Shafer, J.T., Brandon, A.D., Lapen, T.J., Righter, M., Peslier, A.H., Beard, B.L., 2010. Trace element systematics and (super 147) Sm/ (super 143) Nd and (super 176) Lu/ (super 176) Hf ages of Larkman Nunatak 06319; closed system fractional crystallization of an enriched shergottite magma. Geochimica Et Cosmochimica Acta 74, 7307-7328. Shea, T., Lynn, K.J., Garcia, M.O., 2015. Cracking the olivine zoning code: Distinguishing between crystal growth and diffusion. Geology 43, 935-938. Shearer, C.K., Aaron, P.M., Burger, P.V., Guan, Y., Bell, A.S., Papike, J.J., 2013. Petrogenetic linkages among fO2, isotopic enrichments-depletions and crystallization history in Martian basalts. Evidence from the distribution of phosphorus in olivine megacrysts. Geochimica et Cosmochimica Acta 120, 17-38. Shearer, C.K., Burger, P.V., Papike, J.J., Borg, L.E., Irving, A.J., Herd, C., 2008. Petrogenetic linkages among Martian basalts: Implications based on trace element chemistry of olivine. Meteoritics & Planetary Science 43, 1241-1258. Shearer, C.K., McKay, G., Papike, J.J., Karner, J.M., 2006. Valence state partitioning of vanadium between olivine-liquid: Estimates of the oxygen fugacity of Y980459 and application to other olivine-phyric martian basalts. American Mineralogist 91, 1657-1663. Shih, C.-Y.N., L. E.; Wiesmann, H.; Barrat, J. A., 2003. Age and Petrogenesis of Picritic Shergottite NWA1068: Sm-Nd and Rb-Sr Isotopic Studies. 34th Lunar and Planetary Science Conference, abstract 1439. Shirai, N., Ebihara, M., 2008. Constraints on the magmatism of Mars inferred from chemical compositions and radiogenic isotopic compositions of shergottites. Meteoritics & Planetary Science 43, Suppl., A144.

246 Sio, C.K.I., Dauphas, N., 2016. Thermal and crystallization histories of magmatic bodies by Monte Carlo inversion of Mg-Fe isotopic profiles in olivine. Geology, doi:10.1130/G38056.1 Sio, C.K.I., Dauphas, N., Teng, F.-Z., Chaussidon, M., Helz, R.T., Roskosz, M., 2013. Discerning crystal growth from diffusion profiles in zoned olivine by in situ Mg–Fe isotopic analyses. Geochimica et Cosmochimica Acta 123, 302-321. Sossi, P.A., Nebel, O., Anand, M., Poitrasson, F., 2016. On the iron isotope composition of Mars and volatile depletion in the terrestrial planets. Earth and Planetary Science Letters 449, 360-371. Steinhoefel, G., Horn, I., von Blanckenburg, F., 2009. Matrix-independent Fe isotope ratio determination in silicates using UV femtosecond laser ablation. Chemical Geology 268, 67-73. Sugawara, T., 2000. Empirical relationships between temperature, pressure, and MgO content in olivine and pyroxene saturated liquid. Journal of Geophysical Research: Solid Earth 105, 8457-8472. Taylor, G.J., 2013. The bulk composition of Mars. Chemie der Erde - Geochemistry 73, 401-420. Taylor, L.A., Nazarov, M.A., Shearer, C.K., McSween, H.Y., Cahill, J., Neal, C.R., Ivanova, M.A., Barsukova, L.D., Lentz, R.C., Clayton, R.N., Mayeda, T.K., 2002. Martian meteorite Dhofar 019: A new shergottite. Meteoritics & Planetary Science 37, 1107-1128. Teng, F.-Z., Dauphas, N., Helz, R.T., Gao, S., Huang, S., 2011. Diffusion-driven magnesium and iron isotope fractionation in Hawaiian olivine. Earth and Planetary Science Letters 308, 317-324. Teng, F.-Z., Li, W.-Y., Ke, S., Yang, W., Liu, S.-A., Sedaghatpour, F., Wang, S.-J., Huang, K.-J., Hu, Y., Ling, M.-X., Xiao, Y., Liu, X.-M., Li, X.-W., Gu, H.-O., Sio, C.K., Wallace, D.A., Su, B.-X., Zhao, L., Chamberlin, J., Harrington, M., Brewer, A., 2015. Magnesium Isotopic Compositions of International Geological Reference Materials. Geostandards and Geoanalytical Research 39, 329- 339. Toplis, M.J., 2005. The thermodynamics of iron and magnesium partitioning between olivine and liquid: criteria for assessing and predicting equilibrium in natural and experimental systems. Contributions to Mineralogy and Petrology 149, 22-39. Usui, T., McSween, H.Y., Floss, C., 2008. Petrogenesis of olivine-phyric shergottite Yamato 980459, revisited. Geochimica Et Cosmochimica Acta 72, 1711-1730. Usui, T., Sanborn, M., Wadhwa, M., McSween, H.Y., 2010. Petrology and trace element geochemistry of Robert Massif 04261 and 04262 meteorites, the first examples of geochemically enriched lherzolitic shergottites. Geochimica Et Cosmochimica Acta 74, 7283-7306. Van Orman, J.A., Krawczynski, M.J., 2015. Theoretical constraints on the isotope effect for diffusion in minerals. Geochimica et Cosmochimica Acta 164, 365-381. Vance, J.A., 1969. On synneusis. Contributions to Mineralogy and Petrology 24, 7-29. Wadhwa, M., Lentz, R.C.F., McSween, H.Y., Crozaz, G., 2001. A petrologic and trace element study of Dar al Gani 476 and Dar al Gani 489: Twin meteorites with affinities to basaltic and lherzolitic shergottites. Meteoritics & Planetary Science 36, 195-208. Wang, K., Moynier, F., Dauphas, N., Barrat, J.-A., Craddock, P., Sio, C.K., 2012. Iron isotope fractionation in planetary crusts. Geochimica et Cosmochimica Acta 89, 31-45. Watson, E.B., 1996. Surface enrichment and trace-element uptake during crystal growth. Geochimica et Cosmochimica Acta 60, 5013-5020. Watson, E.B., Müller, T., 2009. Non-equilibrium isotopic and elemental fractionation during diffusion- controlled crystal growth under static and dynamic conditions. Chemical Geology 267, 111-124. Welsch, B., Hammer, J., Hellebrand, E., 2014. Phosphorus zoning reveals dendritic architecture of olivine. Geology, doi:10.1130/G35691.1. Weyer, S., Anbar, A.D., Brey, G.P., Münker, C., Mezger, K., Woodland, A.B., 2005. Iron isotope fractionation during planetary differentiation. Earth and Planetary Science Letters 240, 251-264. Weyer, S., Schwieters, J.B., 2003. High precision Fe isotope measurements with high mass resolution MC-ICPMS. International Journal of Mass Spectrometry 226, 355-368. Weyer, S., Seitz, H.M., 2012. Coupled lithium- and iron isotope fractionation during magmatic differentiation. Chemical Geology 294–295, 42-50.

247 Wiechert, U., Halliday, A.N., 2007. Non-chondritic magnesium and the origins of the inner terrestrial planets. Earth and Planetary Science Letters 256, 360-371. Wilson, L., Head, J.W., 1981. Ascent and eruption of basaltic magma on the Earth and Moon. Journal of Geophysical Research: Solid Earth 86, 2971-3001. Zhang, Y., Ni, H., Chen, Y., 2010. Diffusion Data in Silicate Melts. Reviews in Mineralogy and Geochemistry 72, 311-408. Zipfel, J., Scherer, P., Spettel, B., Dreibus, G., Schultz, L., 2000. Petrology and chemistry of the new shergottite Dar al Gani 476. Meteoritics & Planetary Science 35, 95-106. Zuber, M.T., 2001. The crust and mantle of Mars. Nature 412, 220-227.

248 Figures and tables

Figure 1. Element maps of olivine A1 and A2 showing the position of LA-ICP-MS analyses. Element profiles are in orange and isotope profiles in light grey. The stars represent the EBSD analyses. RGB image with Fe in red, Mg in green and Ca in blue.

249 B A

[100] [010] [001]

C 75

55 Forsterite content content Forsterite

50 0 200 400 600 800 1000 1200 1400 1600 Distance from left rim (µm) Figure 2. Glomerocryst of olivine B4. [A] Back scattered image with the position of EBSD analyses (colored circles) and one EMP profile through the crystal (blue line). [B] Stereographic projection of the crystallographic orientation for each of the 3 crystals in [A] (colored circles). Crystals red and green are twinned on their face (001) and crystals green and grey are twinned on their face (100). [C] Forsterite profile across the glomerocryst (blue line in A).

250 A1 B1 B2 76 76 72 72 72 68 68 outer rim 68 64 64 outer rim outer rim rim core 60 rim core 64 60 rim core 56 56 60 EMPA LA-ICP-MS Forsterite content 52 52

800 1000 1000

600 800 800

Ni (ppm) 600 600 400 400 400 200

120 160 160

120 80 120 80 80 Ti (ppm) 40 40 40

0.6 0.6 0.6

0.4 0.4 0.4 Ca (wt.%) 0.2 0.2 0.2

0 0 0 0 200 400 600 800 0 100 200 300 400 500 0 100 200 300 400 500 Distance from rim (µm) Distance from rim (µm) Distance from rim (µm) Figure 3. Chemical profiles of major (forsterite content) and minor elements in olivines A1, B1 and B2 from EMP and LA-ICP-MS measurements. The outer rim is defined by the sharp increase in Ti and the rim is defined as the part of the crystal characterized by a steep Fo profile.

251 A1 75 75 1 0 70 70 0.8 -0.2 0.6

65 -0.4 δ56 Fe 65 0.4 δ 26 0.2 Mg -0.6 60 60 0 -0.8 -0.2 Forsterite mol.% 55 -1 55 -0.4

-1.2 -0.6 50 50 0 50 100 150 200 250 300 350 0 50 100 150 200 250 300 350 A2 75 75 1 0 70 0.8 70 -0.2 0.6

-0.4 δ56 Fe 65 0.4 δ 26 65 0.2 Mg -0.6 60 0 Forsterite mol.% -0.8 -0.2 60 -1 55 -0.4

-1.2 -0.6 55 50 0 50 100 150 200 250 300 350 0 50 100 150 200 250 300 350 75 75 A4 1 0 0.8 70 70 -0.2 0.6

65 -0.4 65 0.4 δ56 Fe δ 26

0.2 Mg -0.6 60 60 0 -0.8

Forsterite mol.% -0.2 55 -1 55 -0.4

-1.2 -0.6 50 50 0 50 100 150 200 250 300 0 50 100 150 200 250 300 B1 75 75 1 0 70 70 0.8 -0.2 0.6 δ56 Fe δ

-0.4 0.4 26 Mg 65 65 -0.6 0.2 0 -0.8

Forsterite mol.% -0.2 60 60 -1 -0.4

-1.2 -0.6 55 55 0 50 100 150 200 250 300 350 0 50 100 150 200 250 300 350 Distance from rim (µm) Distance from rim (µm)

252 75 75 B2 1

0 0.8 70 -0.2 70 0.6

0.4 -0.4 δ56 Fe δ26 Mg 65 65 0.2 -0.6 0

Forsterite mol.% -0.8 -0.2 60 60 -1 -0.4

-1.2 -0.6 55 55 0 50 100 150 200 250 300 350 0 50 100 150 200 250 300 350

75 75 B5 1 0 0.8

70 -0.2 70 0.6

-0.4 0.4 δ26 Mg δ56 Fe 65 65 0.2 -0.6 0 -0.8 Forsterite mol.% -0.2 60 60 -1 -0.4

-1.2 -0.6 55 55 005050 100 150 200 250 0 50 100 150 200 250 300

75 75 B7 1 0 0.8 70 70 -0.2 0.6

-0.4 65 0.4 δ26 Mg δ56 Fe 65 0.2 -0.6 60 0 -0.8 Forsterite mol.% -0.2 60 -1 55 -0.4

-1.2 -0.6 55 50 0 50 100 150 200 250 300 350 0500 50 100 150 200 250 300 Distance from rim (µm) Distance from rim (µm) Figure 4. Forsterite and isotope profiles in olivine A1, A2, B1, B2, B5 and B7, all the megacrysts for which isotope measurements are available. Error bars are 2 standard deviations (0.10 ‰). The grey horizontal bar represent the average isotopic composition for martian rocks from the literature and in-situ analysis of NWA 1068 groundmass (see text for detail).

253 75 3

2 70 1

65 0 or δ26 Mg

-1 δ 56 Fe

Forsterite content content Forsterite 60 -2

2 70 1

65 0 or δ26 Mg

-1 δ 56 Fe

Forsterite content content Forsterite 60 -2

0 100 200 300 400 500 600 distance from rim (µm) Figure 5. Evolution of Fo content and Fe-Mg isotope fractionation during simple diffusion in olivine (see section 4.2.1 for model description). Isotope fractionation affects the rim first and the core remains chemically homogeneous. When isotope fractionation reaches the core, the Fo profile is curved.

254 14 A1 A2 12 A4 B1 10 B2 B5 8 B7 6

decreasing angle with c-axis Forsterite gradient gradient Forsterite 4 89º 23º

(Forsterite content at 75 µm - 25 µm) content (Forsterite x10-16 1 2 3 4 5 6 7 8 9 10 Fe-Mg difusivity (m2/s)

Figure 6. Comparison between the difference in forsterite content at 75 µm and 25 µm from the megacryst rims (forsterite gradient) and the Fe-Mg diffusivity along each profile based on the crystallographic orientation. The diffusivities are calculated following Dohmen et al., (2007) and Dohmen and Chakraborty (2007) for a temperature of 1390 ºC, a pressure of 1 bar, an average Fo content of 70, a fO2 at the QFM buffer and considering that the diffusivity is 6 times higher along the c-axis compared to crystallographic axes a and b.

255 A difusion only BCcrystallization only Simultaneous difusion and crystallization Liquid 1 Liquid 1 Liquid 1 1 Mg#=47.1 Mg#=47.7 Mg#=53.4 olivine 1 olivine 1 olivine 1 Fo71.8 Fo72.3 Fo76.6

Liquid 2 T =1314-1220 ºC Mg#=31.3 t = ~6.6 years Rim: Fo68.7 2 T = 1170 ºC 20% crystallization of initial liquid

T =1170 ºC T= 1110 ºC T =1220-700 ºC t = ~64 days t= N.A. t = ~6.6 years 3 + 108 days 0% crystallization 29% crystallization 35% crystallization of (open system) of initial liquid initial liquid

75 75

70 70

65 overgrowth 65

60 60 Forsterite content (%) content Forsterite 55 55

0.2 0.2 Fe 6 0 0

−0.2 −0.2

−0.4 Mg or δ5 −0.4 26 δ −0.6 −0.6

−0.8 −0.8

0 100 200 300 400 500 600 0 100 200 300 400 500 600 0 100 200 300 400 500 600 Distance from rim (µm) Figure 7. Schematic representation of the three models, resulting forsterite and isotope profiles and comparison with profiles measured in olivine B2. The 2 shades of green bracket the uncertainty on the unfractionated composition of Mg isotopes. Light green: β-Mg = 0.15, initial δ26Mg = -0.25. Dark green: β-Mg = 0.21, initial δ26Mg = -0.4. [A] Simple diffusion model. [B] Simple crystallization model. [C] Model of simultaneous diffusion and crystallization. The dotted lines represent the profiles at an intermediate stage, before the outer rim (overgrowth) crystal. The case study in C correspond to simulation olB2_b from table 2-3.

256 A2 B1 A1

75 75

70 70

65 65

60 60

Forsterite content (%) content Forsterite 55 55

0.6 0.6

0.4 0.4

0.2 0.2

0 0

-0.2 -0.2

-0.4 -0.4

δ26 Mg or δ56 Fe -0.6 -0.6

-0.8 -0.8

-1.0 -1.0

010020030040050060070001002003004005006007000100200300400500600700 Distance from rim (µm)

Figure 8. Example of simulations of the simultaneous diffusion and crystallization model for three other olivine crystals (A2, B1 and A1). The simulations correspond to the rows of table 2-3 appearing in bold. The dotted lines are the unfractionated, average isotopic composition for Mars.

75 A1 B1 A2 B2 74 A4 B5 B7 73

Forsterite content (mol.%) content Forsterite 71

70 74 75 76 77 78 79 initial Forsterite content (mol.%) Figure 9. Comparison of the initial forsterite content of olivine as predicted by the model (x-axis) and the actual current forsterite content measured. The result of 2 or 3 simulations are represented for the 7 olivine megacrysts for which isotopic measurements are available. The color bars represent the 1 σ of the simulations in Table 4 (random search).

257 70 DaG 479 SaU 005 NWA 5789 RBT 04262 65 Yamato NWA 6234 980459 Dhofar 019 60 difusion NWA 1068 LAR 06319

whole rock Mg# (mol.%) whole rock 50 NWA 2990 accumulation

85 80 75 70 65 60 Forsterite content (mol.%) Figure 10. Representation of the Fe-Mg equilibrium between whole-rock composition and the composition of olivine megacrysts for depleted (blue diamond) and enriched (red squares) shergottites. The dashed lines encompass Kd Fe-Mg olivine-pyroxene values in the range 0.34- 0.36 (solid line is 0.35). NWA 1068 and LAR 06319 are represented as rectangles to account for the variable composition of megacryst cores. The orange extension of those rectangles towards the left (equilibrium conditions) represents the effect of diffusion. When diffusion is taken into account, no accumulation of megacryst is required to reach the equilibrium conditions. Data from Basu Sarbadhikari et al. (2009), Bunch et al. (2009), Goodrich (2003), Gross et al. (2011, 2013), Peslier et al. (2010), Taylor et al. (2002), Usui et al. (2008, 2010) and Wadhwa et al. (2001).

258 core σ rim Sc olA1 6.72 0.36 8.2 olA2 5.74 0.3 8.7 olA3 7 0.9 7.9 olA4 7.5 0.3 9.7 olB1 6.1 0.3 8.9 olB2 8.4 1.4 9.2 olB3 6.9 0.2 7.1 V olA1 18.1 3.2 14.6 olA2 17.5 2.1 11.6 olA3 19.1 3.5 8.2 olA4 23 2.4 16.5 olB1 23 1.9 20.4 olB2 17 2.1 13.9 olB3 23.5 4.2 9.6 Cr olA1 630 122 283 olA2 756 103 274 olA3 771 147 152 olA4 756 153 251 olB1 1025 240 448 olB2 986 220 261 olB3 1025 304 198 Co olA1 101 6 118 olA2 85.7 2 91 olA3 94 4.5 91.7 olA4 96.5 0.3 98.1 olB1 112 0.7 115 olB2 104 3 107 olB3 100 1 97 Ni olA1 633 30 418 olA2 606 15 379 olA3 611 89 345 olA4 728 39 446 olB1 1085 75 659 olB2 647 14 512 olB3 621 13 434 Cu olA1 10.8 0.6 6.7 olA2 10.3 0.7 6.4 olA3 10.7 0.8 10.4 olA4 11.3 1 10.2 olB1 11.75 0.2 9 olB2 9.7 0.5 7.3 olB3 8.3 0.5 5.7 Zn olA1 88 22 104 olA2 69 3 95 olA3 93 16 98 olA4 151 6 146 olB1 68.5 3 96 olB2 88 11.4 88 olB3 70 7 76 Table 1 Composition of trace elements in olivine megacrysts (ppm). Core compositions are the average of 3 to 5 analyses and σ is the standard deviation on those analyses. Rim compositions represent a single analysis.

259 Table 2. Model parameters.

CR1 CR2 CR3 β-Fe β-Mg ∅total Rim (∅/2) ∅sph ∅sph2 Mg# liq1 ºC/s ºC/s ºC/s µm µm µm µm mol.% olA1_a 7.0E-07 6.0E-05 6.0E-05 0.3 0.25 / 0.17 670 37.5 1000 860 54.2 olA1_b 8.5E-07 8.0E-05 8.0E-05 0.31 0.24 / 0.17 700 50 1000 940 54.2

olA2_b 9.0E-07 1.5E-05 1.0E-03 0.25 0.22 / 0.16 700 40 1050 920 54.0 olA2_a 1.0E-06 2.0E-05 1.0E-03 0.25 0.22 / 0.16 700 40 1050 920 53.8 olA2_c 8.0E-07 5.0E-05 1.0E-04 0.25 0.22 / 0.16 690 35 1060 900 54.0 olA4_a 1.1E-06 1.2E-04 1.2E-04 0.31 0.24 / 0.17 570 55 770 770 56.1 olB1_a 8.0E-07 8.0E-05 8.0E-05 0.3 0.25 / 0.17 750 70 1060 1040 54.4 olB1_b 7.0E-07 4.5E-05 4.5E-05 0.28 0.23 / 0.16 750 45 1100 1000 55.3

olB2_a 9.0E-07 5.0E-05 5.0E-05 0.22 0.18 / 0.13 660 40 960 930 53.4 olB2_b 7.0E-07 5.5E-05 5.5E-05 0.25 0.21 / 0.15 660 50 960 920 53.4 olB2_bs 7.0E-08 5.5E-05 5.5E-05 0.25 0.21 / 0.15 660 50 960 920 53.4 olB2_bf 7.0E-06 5.5E-05 5.5E-05 0.25 0.21 / 0.15 660 50 960 920 53.4 olB2_c 7.0E-07 6.0E-05 6.0E-05 0.28 0.21 / 0.15 660 60 960 940 53.1

olB5_a 1.2E-06 6.0E-05 6.0E-05 0.29 0.25 / 0.17 520 50 760 760 50.9 olB5_b 9.0E-07 4.0E-05 4.0E-05 0.29 0.25 / 0.17 520 40 800 740 50.9 olB5_c 6.0E-07 4.0E-05 4.0E-05 0.29 0.25 / 0.17 520 40 740 730 53.7

olB7_a 1.1E-06 7.0E-05 7.0E-05 0.28 0.26 / 0.18 570 35 840 790 52.2 olB7_b 5.5E-07 5.5E-05 5.5E-05 0.29 0.25 / 0.17 570 35 860 750 53.3 olB7_c 8.0E-07 9.0E-05 9.0E-05 0.29 0.27 / 0.18 570 50 860 790 51.1 Min 5.5E-07 1.5E-05 4.0E-05 0.22 0.18 / 0.13 520 35 740 730 50.9 Max 1.2E-06 1.2E-04 1.0E-03 0.31 0.27 / 0.18 750 70 1100 1040 56.1 Input parameters for a set of simulations (see text for detail). The rows in bold correspond to the simulations illustrated in Fig 7, 8. The first value of β-Mg (0.18-0.27) and a starting δ26Mg -0.4 ‰ were used for the simulations but the second value of β-Mg (0.13-0.18) gives similar results with a starting δ26Mg of -0.25 ‰ (see dark and light green lines, Fig. 7, 8). ∅sph is the radius of the sphere of melt (of composition Mg# liq1) from which the crystal grows for stage 1 (CR1). ∅sph2 is the radius of the growth medium for stages 2 and 3 (CR2, CR3).

260 Table 3. Model outputs. 26 56 26 56 Mg# liq2 Fo core1 Fo core2 ΔFo Fo rim δ Mg core δ Fe core δ Mg rim δ Fe rim F res t core t rim t diff t rim tot mol.% mol.% mol.% mol.% mol.% wt.% years days days days olA1_a 28.4 77.2 72.8 4.3 53.8 0.15 -0.81 0.82 -1.26 0.52 7.1 21 78 99 olA1_b 28.4 77.2 73.1 4.0 53.7 0.15 -0.88 0.61 -1.08 0.58 6.0 15 59 74

olA2_b 30.2 77.1 73.4 3.7 55.9 0.11 -0.72 0.30 -0.71 0.56 5.4 76 15 92 olA2_a 29.9 76.9 73.5 3.4 55.5 0.10 -0.71 0.27 -0.67 0.56 4.9 57 5 62 olA2_c 31.5 77.1 73.4 3.6 57.4 0.07 -0.82 0.40 -0.67 0.55 5.9 22 49 71 olA4_a 27.2 78.5 72.9 5.6 51.6 0.15 -0.84 0.66 -1.05 0.59 4.3 39 13 52 olB1_a 30.1 77.3 73.9 3.5 55.6 0.09 -0.75 0.45 -0.90 0.62 5.6 16 60 76 olB1_b 31.9 78.0 73.8 4.1 57.7 0.10 -0.79 0.50 -1.03 0.58 7.4 25 109 134

olB2_a 31.7 76.6 72.4 4.2 57.4 0.06 -0.67 0.25 -0.71 0.64 5.8 18 98 116 olB2_b 30.8 76.6 72.3 4.3 56.4 0.05 -0.64 0.40 -0.84 0.63 6.6 20 88 108 olB2_bs 30.9 76.6 69.5 7.2 56.5 -0.26 -0.17 0.34 -0.75 0.63 66.0 20 88 108 olB2_bf 29.7 76.6 75.9 0.7 55.1 -0.21 -0.34 0.45 -0.91 0.62 0.7 20 867 887 olB2_c 28.8 76.4 72.5 3.9 56.7 0.00 -0.64 0.34 -0.88 0.65 5.9 19 81 100

olB5_a 29.8 74.7 71.2 3.6 55.2 0.13 -0.69 0.58 -0.97 0.68 3.2 18 80 98 olB5_b 30.8 74.7 70.8 3.9 56.4 0.12 -0.67 0.64 -1.07 0.65 4.2 25 122 147

261 olB5_c 30.6 76.9 70.9 5.9 56.3 0.10 -0.64 0.68 -1.11 0.64 8.2 27 121 147

olB7_a 29.7 75.8 71.6 4.2 55.2 0.26 -0.82 0.64 -0.94 0.62 4.5 13 68 82 olB7_b 29.3 76.6 71.5 5.1 54.9 0.09 -0.64 0.77 -1.17 0.56 8.3 22 86 108 olB7_c 28.1 74.9 71.3 3.6 53.3 0.12 -0.62 0.72 -0.98 0.62 4.7 14 52 66 Min 28.1 74.7 70.8 3.4 53.3 0.00 -0.88 0.25 -1.26 0.52 3.2 13 5 52 Max 31.9 78.0 73.9 5.9 57.7 0.26 -0.62 0.77 -0.67 0.68 8.3 76 122 147 Mg# liq2: Final composition of the liquid. Fo core 1: Initial composition of olivine. Fo core 2: Final composition of megacryst cores. Fo rim: Final composition of megacryst rims. δ26Mg/ δ56Fe core: isotope fractionation reached at the center of olivine megacrysts. δ26Mg/ δ56Fe rim: maximal isotope fractionation reached in megacryst rims (variable position, see figure 7, 8) F res: fraction of residual liquid (1 – ol crystallized). t core: time elapsed during core crystallization. t rim: time elapsed during rim crystallization (overgrowth) t diff: time elapsed during diffusion only phase t tot: t rim + t diff. Profiles for simulations in bold are in Fig. 7 and 8. Other profiles are in Fig. EA4.

Table 4. Average of model parameters and results from random sampling δ26Mg start # of runs CR1 (ºC/s) σ t (years)* t (years)** β-Fe σ β-Mg σ Δ Fo σ range min 7.00E-08 0.12 0.08 0 max 3.00E-06 0.32 0.28 8.7 olB2 -0.4 37 1.20E-06 4.29E-07 2.1 – 4.3 1.6 – 9.8 0.269 0.021 0.214 0.030 3.160 0.944 olB2 -0.25 31 1.25E-06 4.61E-07 1.9 – 3.9 1.5 – 9.7 0.279 0.024 0.147 0.036 3.072 0.900 olB1 -0.25 28 1.20E-06 3.41E-07 2.2 – 3.9 1.8 – 6.4 0.296 0.017 0.160 0.039 3.140 0.647 olB7 -0.25 33 1.59E-06 4.88E-07 1.7 – 3.2 1.4 – 5.7 0.273 0.022 0.148 0.034 3.523 0.767 Range min/max : Parameter space sampled for all megacrysts. *duration of core crystallization calculated for 1 standard deviation on CR1. ** duration calculated for 2 σ on CR1. Δ Fo is the difference between the initial and final forsterite content at the center of the megarcryst.

261 Supplementary material

0.1

-0.1

-0.2

-0.3 Fe

6 -0.4 δ5 -0.5

-0.6

-0.7

-0.8

-0.9 100 200 300 400 500 600 Distance from rim (µm)

Figure S1. Model of Fe isotope profile for olB2_b (simulation from Table 2 and 3). Isotope fractionation due to stage 1 (blue) and stage 2 and 3 (orange) and bulk isotope fractionation (red). The isotopic composition at the center of the megacryst results mainly from the core crystallization (stage 1).

262

263

264

265

Figure S2. Isotope and element (Fo content) profiles of olivine megacrysts. The simulation numbers (e.g. OlA1_a) correspond to the simulations of Table 2 and 3 (input and output).

266 A2 A3 76 76 72 72 68 64 68

outer rim 60 EMPA LA-ICP-MS outer rim 64 56 rim 52 Forsterite content core rim core 60 48

1000 800 800 600

Ni (ppm) 600 400 400

120 160

80 120

Ti (ppm) 80 40 40

0.6 0.6

0.4 0.4 Ca (ppm) 0.2 0.2

0 0 0 100 200 300 400 0 100 200 300 400 Distance from rim (µm) Distance from rim (µm)

267 A4 B3

72 72 68 68 64 outer rim outer rim 64 60 EMPA LA-ICP-MS 56 Forsterite content 60 rim core rim core 52

1200 800 1000

600 800

Ni (ppm) 600 400 400 200

160 160

120 120

Ti (ppm) 80 80

40 40

0.6 0.6

0.4 0.4 Ca (ppm) 0.2 0.2

0 0 0 50 100 150 200 250 300 0 50 100 150 200 250 300 Distance from rim (µm) Distance from rim (µm) Figure S3. Additional profiles of major and minor elements in olivines from EMP and LA-ICP- MS measurements. The outer rim is defined by the sharp increase in Ti and the rim is defined as the part of the crystal characterized by a steep Fo profile.

268