Petrogenesis of main group pallasite meteorites based on relationships among texture, mineralogy, and geochemistry

Main group pallasite meteorites are samples of a single early magmatic planetesimal, dominated by metal and olivine but containing accessory chromite, sulfide, phosphide, phosphates, and rare phosphoran olivine. They represent mixtures of core and mantle materials, but the environment of formation is poorly understood, with a quiescent core–mantle boundary, violent core–mantle mixture, or surface mixture all recently suggested. Here, we review main group pallasite data sets and petrologic characteristics, and present new observations on the low‐MnO pallasite Brahin that contains abundant fragmental olivine, but also rounded and angular olivine and potential evidence of sulfide–phosphide liquid immiscibility. A reassessment of the literature shows that low‐MnO and high‐FeO subgroups preferentially host rounded olivine and low‐temperature P2O5‐rich phases such as the Mg‐phosphate farringtonite and phosphoran olivine. These phases form after metal and silicate reservoirs back‐react during decreasing temperature after initial separation, resulting in oxidation of phosphorus and chromium. Farringtonite and phosphoran olivine have not been found in the common subgroup PMG, which are mechanical mixtures of olivine, chromite with moderate Al2O3 contents, primitive solid metal, and evolved liquid metal. Lower concentrations of Mn in olivine of the low‐MnO PMG subgroup, and high concentrations of Mn in low‐Al2O3 chromites, trace the development and escape of sulfide‐rich melt in pallasites and the partially chalcophile behavior for Mn in this environment. Pallasites with rounded olivine indicate that the core–mantle boundary of their planetesimal may not be a simple interface but rather a volume in which interactions between metal, silicate, and other components occur.


INTRODUCTION
Naturally delivered samples of differentiated planetesimals, formed in the first few million years of the solar system's history, are available to us as nonchondritic meteorites. However, these physical records are overwhelmingly represented by metallic meteoritic sources with far fewer samples of unique mantles and crusts (in the ranges 26-60 versus 4-18 parent bodies, respectively; e.g., Bell et al. 1989;Greenwood et al. 2015). Associated with this is a scarcity of sulfide-rich meteorites that should have crystallized abundant troilite via differentiation of metallic liquids, but these might be highly vulnerable to destruction during transport, atmospheric entry, or during residence on the surface of the Earth (Kracher and Wasson 1982). Alternately, sulfide might have been sequestered deep inside planetesimal cores and be unsuitably located for transport to Earth as meteorites (e.g., Boesenberg et al. 2012).
Although processes of differentiation leading to separate planetesimal cores and mantles are poorly recorded by meteorites, the final stages of core-mantle separation are preserved in a few planetary sources, particularly the various pallasite parent bodies. Pallasite meteorites consist predominantly of mixtures of olivine and metal with accessory troilite, chromite, phosphide, and various phosphate minerals (Boesenberg et al. 2012). Most are geochemically, isotopically, and petrologically related, earning them the designation "main group" pallasite (hereafter PMG following Wasson and Choi 2003). In particular, their coherent triple-oxygen isotope systematics strongly indicate that the PMG originated from just one or possibly two well-sampled parent bodies (Greenwood et al. 2015;Ali et al. 2018). However, just as the number of recognized planetesimal sources of crustal, basaltic rocks has increased in recent years Sanborn and Yin 2014), so have multiple pallasite parent bodies been recognized by significant differences in oxygen isotope composition that cannot be related to each other through mass-dependent fractionation (e.g., Mayeda 1978, 1996). These other bodies are poorly sampled, but illustrate the abundance of pallasiteproducing planetesimals in the early solar system, with at least five distinct sources represented by the various groups or individual pallasites. These include the PMG; the Eagle Station group (PES after Wasson and Choi 2003); and the ungrouped or otherwise anomalous pallasites Milton (Jones et al. 2003), Vermillion, Yamato 8451 (Boesenberg et al. 2000), Choteau (Gregory et al. 2016), Zinder, Northwest Africa 1911 (Bunch et al. 2005), and Northwest Africa 10019 (Agee et al. 2015). These confirm that pallasitic material is frequently produced after partial or complete melting of primitive planetesimal mantles. For molten planetesimals, pallasitic material may represent the default, penultimate end product of differentiation, core-mantle separation, and final metal-silicate equilibration.
The available data for pallasites, long dominated by large petrographic and geochemical studies (Buseck 1977;Buseck and Holdsworth 1977;Scott 1977aScott , 1977b have recently received interesting new contributions from a variety of fields. Improvement in metallographic cooling rates (Yang et al. 2010), constraints on the stability and occurrence of unusual phosphorus-rich olivine (with more than~1 wt% P 2 O 5 and hereafter referred to as "phosphoran olivine"; Boesenberg and Hewins 2010;Fowler-Gerace and Tait 2015), and new paleomagnetic histories (Tarduno et al. 2012;Bryson et al. 2015) have been brought to the literature. Each of these has driven its own, somewhat exclusive model of pallasite petrogenesis. The suggested formation environments include deep, thorough impact mixing zones of olivine and metal (Scott 2007;Yang et al. 2010), a quiescently evolving core-mantle boundary or olivine-metal zone (e.g., Boesenberg et al. 2012), and surface mixing of olivine and metal (Tarduno et al. 2012;Bryson et al. 2015). Recently, experimental work (Solferino et al. 2015) and new petrographic observations (Scott 2017) have indicated that the morphology of pallasite olivine, that is, the macroscale textures and grain size, was controlled by the presence of and their residence time in an Fe-S liquid. Fragmental olivines are therefore likely to have formed from relatively large (~5-15 mm) rounded or angular precursors, which are the most common morphological type observed in pallasites (Scott 1977b). Here, we investigate whether any previously unrecognized correlation exists between olivine morphology, olivine geochemistry, and accessory phase mineralogy in the PMG. Our study focuses on the fragmentally textured pallasite Brahin (e.g., Boesenberg et al. 2012), the relationship between it and the other members of the PMG, and the possible planetesimal environments responsible for the various pallasite subgroups.

MOTIVATION, SAMPLE SELECTION, AND METHODS
Within the PMG, Boesenberg et al. (2012) highlighted the difficulty of identifying coherent subgroups that gather all textural, mineralogical, and geochemical features together. The geochemical information available for olivine is particularly scarce, with Mn being the only trace element that has been collected systematically. The Fe-Mg-Mn systematics of PMG olivine define a subgroup (hereafter referred to as "low-MnO PMG") with MnO contents in olivine lower than~0.25 wt% (e.g., Boesenberg et al. 2012). We will refer to the main cluster of PMG with similar olivine major element compositions of~Fa 12-13 and with~0.30 wt% MnO as the "common subgroup PMG." The low-MnO PMG subgroup includes Brahin, Brenham, Hambleton, Krasnojarsk, Molong, Sericho, Seymchan, and Thiel Mountains (Smith et al. 1983;Mittlefehldt et al. 1998;Wasson et al. 1999;Hsu 2003;Boesenberg et al. 2012Boesenberg et al. , 2018Ali et al. 2018). The olivines in Brenham, Krasnojarsk, and Thiel Mountains seem to be exclusively of the rounded variety (Buseck 1977;Scott 1977b). The olivines in the heavily corroded Molong pallasite (Mingaye 1916) were described as predominantly angular although with other types present (Buseck 1977) but were specifically identified as having potentially originated from a Brenham-like precursor by separation of olivine grains because of the presence of some rounded olivine (Scott 1977b). Similarly, Sericho has been reported to contain rounded and occasionally polyhedral grains (Boesenberg et al. 2018). The olivine in Seymchan is diverse, including rounded, angular, and fragmental grains (Van Niekerk et al. 2007;Yang et al. 2010;Boesenberg et al. 2012). By contrast, Hambleton has been reported to contain angular and abundant fragmental olivine, and is anomalous in its very high sulfide content (Johnson et al. 2006a(Johnson et al. , 2006b). Brahin has been described on the macroscale as fragmental (Buseck 1977) and as containing both highly angular olivine fragments as well as large angular olivine grains (Scott 1977b). Brahin and Hambleton are therefore the only low-MnO PMG that have not been reported to contain some rounded olivine. Brahin is also of interest as one of the few PMG to contain phosphoran olivine (Buseck 1977) and silico-phosphate (Buseck and Holdsworth 1977).
Because Brahin contains geochemically similar olivine but a conspicuous fragmental appearance compared with that of other low-MnO PMG, sections of this pallasite were investigated at the Royal Belgian Institute of Natural Sciences, Brussels, Belgium (RBINS). A piece has been identified measuring~5 9 5 cm (RBINS sample M226) with approximately half containing relatively abundant large rounded olivine, with most~4-5 mm and up to~6 mm diameter, while the rest contains finer angular and fragmental olivine with a wide range of grain sizes that is more typical of this meteorite. This petrographic arrangement has not been reported previously in Brahin and the section was therefore mapped by microscale X-ray fluorescence (micro-XRF) using a Bruker M4. By comparison with other methods (e.g., backscatter electron imaging by scanning electron microscopy), this method enables a larger field of view, requires no sample preparation, and is also nondestructive. Our system is equipped with a 30 W Rh-anode X-ray source with a Be side-window, and two silicon drift detector spectrometers with detector areas of 30 mm 2 , located symmetrically with respect to the sample and X-ray source (energy resolution of 145 eV for Mn K a ). The source was run at an operating voltage of 50 kV and current 600 lA and the sample was held under vacuum (20 mbar). The Bruker M4 Tornado software was used to generate a false color phase map from the obtained fourdimensional "data-block" (x, y, and K a intensity for all detected elements; Fig. 1a). The simplified accessory phase map was manually traced by hand and represents an interpretation map excluding olivine and metal (Fig. 1b).
The aforementioned preferential association of rounded olivine morphology, anomalous (i.e., low-MnO PMG) olivine composition, and phosphorus-rich phases has also prompted us to obtain quantitative major and minor element analyses of olivine in several PMG by electron microprobe analysis (EMPA). We analyzed olivine in Brenham (from the Research School of Earth Sciences, Australian National University); Cumulus Ridge 04071,23 (from the Meteorite Working Group, NASA Astromaterials Curation; hereafter referred to as CMS 04071); Esquel and Imilac (from RBINS, samples M218 and M222); and Brahin, Fukang, and Seymchan (from private dealers). The piece of Seymchan we investigated contains only angular and minor fragmental olivine, but rounded olivine has also been reported (Yang et al. 2010;Kichanov et al. 2018). The other samples in our study are petrographically similar to previous reports for these meteorites. There are few complete EMPA analyses available for olivine from Brahin, CMS 04071, Esquel, Fukang, and Seymchan in the literature, that is, analyses that include MnO. There are more data available for Brenham and Imilac and these serve as a check on our analyses.
Quantitative analysis of olivine cores was made with a JEOL JXA-8500F field-emission electron microprobe (Museum f€ ur Naturkunde, Berlin, Germany) with a fully focused beam of 50 nA current with 15 kV accelerating voltage, utilizing five wavelength-dispersive X-ray spectrometers. New calibrations were made using Astimex and Smithsonian reference materials. K a peak measurements were made for Si, Al, Mg, P (TAP), Ti, Mn, (PETJ), Ni, Ca, (PETH), Cr, and Fe (LIFH) with backgrounds measured on both sides of each peak. Spectrometer arrangement and measurement times are given in Table 1. Matrix effects were corrected using the JEOL inbuilt ZAF routine. These are the same instrumental settings as previously reported in Chernonozhkin et al. (2017).

Micro-XRF Mapping and Petrographic Observations
Our micro-XRF map illustrates a range of olivine morphologies in the studied piece of Brahin (Fig. 1a). All olivine grains are set in a groundmass of metal or accessory phases. Large (up to~6 mm) rounded olivines only exhibit curvilinear margins along surfaces in contact with the groundmass. Large olivine crystals may have some contact with each other and where this occurs, their mutual boundaries are straight; some crystals are sufficiently close-set to one another that they share triple junctions in section (lower middle of Fig. 1a). Some exhibit straight margins that match nearby grains, as though having been invaded by metal (middle of Fig. 1a), a feature most often observed in pallasites dominated by angular olivine or polycrystalline olivine aggregates. Midsized and fine-grained olivine crystals are more irregular in shape and have potentially high aspect ratios, indicative of intragranular fracture or fragmentation. However, almost all midsized and fine-grained olivine have rounded corners to some degree, even where the macroscale morphology is angular. In association with olivine are occasional voids that resulted from plucking of olivine, which were later filled with epoxy resin and are represented in black (Fig. 1a). The bottom left corner was removed during the sample preparation of a previous study and this also appears as black.
In most of this section of Brahin, regions containing only a single accessory phase can be identified; therefore, in Fig. 1b, we present an interpretation map indicating our inferred domain boundaries for each accessory phase. Two large domains in which the only accessory phase is phosphide (upper left and lower right of Figs. 1a and 1b) and one in which it is sulfide (upper right and through the middle of Figs. 1a and 1b) host submillimeter scale grains of each phase that enclose fragmentally textured olivines or adjacent to large olivine grains. Chromite is present in two domains. The first is mostly as a millimeter scale, probably a continuous domain adjacent to olivine. The chromite does not contain distinctive faceted crystal edges, and rather occurs as curved masses that conform to the edge of the olivine (lower left of Figs. 1a and 1b; similar to massive chromite described by Wasson et al. [1999] for Brenham). The second chromite domain in this section is associated with fragmental olivine (upper middle of Figs. 1a and 1b). Although we cannot be sure that the two domains of phosphide and the two domains of chromite were not, respectively, connected in three dimensions, the absence of these phases in other parts of the section indicates strong heterogeneity at the centimeter scale and possible continuity outside of the plane of the section. Rare phosphate is also present as submillimeter grains in a single domain in the lower right, surrounded by the large phosphide region (lower right of Figs. 1a and 1b). An additional phosphate grain occurs in association with the largest chromite.

Olivine Geochemistry by EMPA
Our EMPA analyses of olivine for PMG olivine cores are in good agreement with previously published data (Tables 2 and 3; Smith et al. 1983;Wasson et al. 1999;Lauretta et al. 2006;Danielson et al. 2009;Boesenberg et al. 2012;Ali et al. 2018). We find similar values for the content of the fayalitic component (Fa#, defined as molar Fe/[Fe+Mg] 9 100) for olivine from these pallasites, with relatively low values for Brahin, Esquel, and Seymchan (~Fa 11.5 ); intermediate values for Brenham, CMS 04071, and Imilac (~Fa 12 ); and a higher value for Fukang (Fa 13.7 ). The contents of MnO for Brahin and Brenham olivine are lower than the others (0.21 and 0.20 wt% versus average of 0.27 wt% for the common PMG samples). The MnO content we have found for angular olivine in Seymchan (0.26 wt%) is higher than that reported by Ali et al. (2018;0.22 wt%), yielding molar Fe/Mn values that span the gap between common and low-MnO subgroups. Ali et al. (2018) do not indicate whether their olivine was of the angular, rounded, or fragmental variety, but this large discrepancy is consistent with differing MnO contents in the olivine of the same pallasite according to olivine morphology. We therefore tentatively suggest the possibility that Seymchan could be classified as "transitional" to the common and low-MnO PMG subgroups (Table 3).
Other minor elements are close to or below detection limit, but some are usefully resolved from zero (Table 2). We have found the Cr 2 O 3 , CaO, and Al 2 O 3 contents of PMG olivine to be correlated. In Esquel and Fukang, they are higher than the other pallasites, at 0.05, 0.025, and 0.02 wt%, respectively, while the others have values that decrease until close to or below detection limits (at approximately 0.010, 0.003, and 0.005 wt%, respectively; Table 2 and supporting  information Table S1 in Data S1). We found the P 2 O 5 contents of olivine to be below the quantification limit (with analyses almost exclusively below 0.02 wt% and most below detection limit of~0.011 wt%; Table S1 in Data S1). We did not observe any phosphoran olivine (with~1 wt% P 2 O 5 ) in any of our samples. We do not report values for TiO 2 or NiO because they are all very close to or below detection limit (approximately 0.009 and 0.010 wt%, respectively).
All new data used to calculate the average EMPA compositions in Table 2 are given in Table S1 in Data S1. Our new olivine compositions have been incorporated into the full data set used to construct Fig. 2 (Table 3; Table S2 in Data S1). These analyses, in conjunction with observations from the large slab of Brahin that we investigated (Fig. 1) and our assessment of the literature (Table 3; Fig. 2a), establish a new constraint on pallasite petrogenesis: that pallasites in the low-MnO olivine subgroup of PMG have partly or exclusively "rounded" primary olivine textures. In the next section, we will show that morphologies and compositions of olivine in the various PMG subgroups are also related to their accessory phase mineralogies.

Olivine Morphology and Rounding
There are three macroscale occurrences of olivine in PMG meteorites and the relationships between the various types have only recently become clear. In the first occurrence, relatively large (commonly~5 mm diameter and up to~5-15 mm diameter; Scott 1977b) rounded olivine exhibits a restricted range of grain sizes and often occurs such that grains are partly isolated from each other by metal or other groundmass phases. On a microscale, the olivine margins of many pallasites including those of the other petrographic types also exhibit roundness (e.g., Buseck 1977). In the second occurrence, angular, equiaxed grains may exhibit a larger range in grain size (up tõ 25 mm or even larger; Scott 1977b). Where adjacent to each other, they have been considered "equilibrated" or in "textural equilibrium," sharing triple junctions in cross section (e.g., Buseck 1977;Scott 1977b). Larger assemblages of angular olivine also occur and have been referred to as aggregates and in rare cases as masses or clusters (Scott 1977b;Boesenberg et al. 2012). These may have undergone intergranular fracture and greater or lesser degrees of gentle or more vigorous separation with the space between now filled with metal or an accessory mineral. Finally, fragmental or "highly angular" olivine shards exhibit a wide range of grain sizes and were inferred to form by intragranular fracture. We use the angular, rounded, and fragmental (A, R, F) terminology of Buseck (1977) for the macroscale textures of pallasites and assign each PMG one or more of these signifiers (the full details of our textural assignments are given in Table S3 in Data S1). We have conservatively attributed each signifier even on the basis of only a few reported grains of each type, or a brief mention in the description given in any particular publication. This may be of particular interest for pallasites dominated by angular olivine but with a few rounded olivines, for example, CMS 04071 (Danielson et al. 2009) and Fukang (Lauretta et al. 2006;DellaGiustina et al. 2011DellaGiustina et al. , 2019.
The original form of olivine in pallasites, at the time of olivine-metal mixing (if olivine represents a cumulate) or immediately after olivine grain growth (if refractory olivine and metal remain as restite after high degrees of melting) is of importance for constraining the histories of pallasite parent bodies. Early suggestions to explain pallasite textures favored rounding of initially fragmental or angular olivine, driven by the observation of microscale rounding of olivine in virtually all pallasites (Buseck 1977;Scott 1977b) that might occasionally proceed to high degrees. However, it has since been shown that olivine rounding in various liquids proceeds relatively quickly on geologic timescales, and faster in the presence of a metal-sulfide liquid than under subsolidus conditions (Ohtani 1983;Saiki et al. 2003;Guignard et al. 2012;Solferino et al. 2015). In the presence of substantial sulfide melt, the rounding of olivine on a macroscale is possible over timescales on the order of a few hundred thousand years (Solferino et al. 2015). Along with the occasional occurrence of more than one textural type of olivine in a single pallasite (e.g., Boesenberg et al. 2012), this rules out the possibility of fragmental olivine as a precursor for other types of pallasite olivine (although angular olivine could conceivably become rounded, and ultimately the very earliest form of large olivine might even have been euhedral in the likely case that it formed in the presence of a silicate melt, e.g., Scott 2017). Additionally, these rapid timescales for rounding indicate that separate sources or environments are required for pallasites dominated by rounded olivine, and for those by angular or polycrystalline olivine masses. According to Solferino et al. (2015), it seems that the main control is whether any particular olivine grain was exposed to melt or shared a boundary with other olivine grains. In the former case, olivine would develop conspicuous rounded surfaces where they were exposed to metallic liquid, while in the latter case, they would grow into dense dunitic masses (Scott 1977b). Slight separation of olivine, disintegration of olivine masses, and infiltration of metal occurred before final cooling. The primary, high-temperature silicate products of pallasitic environments are therefore rounded olivine and angular or polycrystalline olivine masses, which we refer to as the primary olivine or texture of pallasites. A combination of both rounded and angular olivine in a single pallasite indicates either a variable olivine-liquid ratio over the scale of sampling (as in the quiescent model of Boesenberg et al. 2012), or that mixing of olivine occurred across geological domains (as in the impact mixing model of Yang et al. 2010). Fragmental olivine is a secondary product formed by brittle deformation at high strain rates (Scott 1977b(Scott , 2017. In the following sections, we will connect these petrographic characteristics of primary rounded and angular olivine with geochemistry and accessory phase mineralogy.

Olivine Geochemistry
Previous PMG subgroupings have been suggested on the basis of olivine and metal geochemistry, and the most commonly used subgroup is that defined by olivine major element composition. The PMG olivine fayalite contents indicate that there are two populations or a possible bimodal distribution. They are dominated by compositions at~Fa 12-13 which include most pallasites such as Admire, Esquel, and Imilac, but with a tail in the distribution toward a small high-FeO population at~F a 16-18 defined by Rawlinna 001 (hereafter Rawlinna), Phillips County, Springwater, Sterley, and Zaisho (Buseck 1977;Scott 1977b;Shima et al. 1980;Buseck and Clark 1984;Ali et al. 2018;Fig. 2). The latter have been designated "anomalous silicate" (PMG-as; e.g., Wasson and Choi 2003; hereafter we refer to them as "high-FeO PMG-as" for consistency with both Wasson and Choi [2003] and our designation of the low-MnO subgroup). This high-FeO olivine is usually of the rounded variety (Kissin 2008(Kissin , 2009. It is also associated with higher Ni contents in the metal, indicating a more oxidizing environment via loss of Fe to the silicate component (Prior 1916;Mason 1963) although there is Olivine morphologies are classified according to: A = angular; R = rounded; F = fragmental. Olivine Fa# = molar 100 9 Fe/(Fe+Mg); MnO in wt% (all compositions from EMPA). Values given are averages calculated from those in the original studies and in this study, except Seymchan (high-MnO value from this study). Data from: Buseck (1977); Buseck and Holdsworth (1977); Scott (1977b); Shima et al. (1980); Smith et al. (1983); Buseck and Clark (1984); Mittlefehldt et al. (1998) Boesenberg et al. (2018); and this study. Full data set and detailed data sources presented in Data S1.
considerable scatter in Ni contents due to kamacitetaenite sampling noise (Scott 1977a;Righter et al. 1990;Wasson and Choi 2003). In addition, some "anomalous metal" compositions have been identified (PMG-am in Wasson and Choi [2003], including Argonia, Brenham, Glorieta Mountain, Huckitta, Krasnojarsk, and Pavlodar; since then, Seymchan has also been attributed to the PMG-am by Van Niekerk et al. [2007]). Similar to PMGas, there is also a strong correlation between rounded olivine textures and anomalous metal compositions (Kissin 2008(Kissin , 2009). Unlike PMG-as, which identifies simple MgO-FeO exchange in olivine and complementary variation in the Fe/Ni of metal, the PMG-am designation combines several trace element peculiarities (e.g., Au-Ir systematics or Ga and Ge concentrations) and we will therefore address their geochemistry in the Trace Element Geochemistry of Olivine and Metal section. The concentrations of MnO in PMG olivine delineate another bimodality or tail in the population to lower concentrations. In most olivine of the PMG, the concentration of MnO is~0.30 wt% but in some, it falls to~0.20 wt%; for PMG with olivine of~Fa 12-13 , we defined these values as the common subgroup and low-MnO PMG, respectively. This variation has been reported in many studies over many years (Lovering 1957;Buseck and Goldstein 1969;Cooper 1974;Kolomeitseva 1975;Takeda et al. 1978;Smith et al. 1983;Mittlefehldt et al. 1998;Mittlefehldt 1999;Hsu 2003;Mittlefehldt and Herrin 2010;Boesenberg et al. 2012;Ali et al. 2018). Decreasing MnO content at steady Fa# results in a small number of pallasites at higher molar Fe/Mn in Fig. 2, including Brahin, which is the main focus of our study. As well as Brahin, the low-MnO PMG includes Brenham, Hambleton, Krasnojarsk, Molong, Sericho, and Thiel Mountains. Practically, one could use a cutoff value of molar Fe/Mn of~48 at normal olivine composition of Fa 12-13 to distinguish the subgroups (olivine from the Southampton pallasite, like Seymchan, may be transitional to the two subgroups at Fe/Mn of 48.5; Kissin et al. [2013]; Table 3). Like the high-FeO PMG-as, there is a high frequency of rounded olivine in the low-MnO PMG. Because some, but not all, of these groups were also assigned to the PMG-am subgroup, Kissin (2008Kissin ( , 2009 suggested that there was a relationship between olivine morphology and unusual metal or olivine compositions. We defined the remaining PMG with normal olivine compositions of~Fa 12-13 and molar Fe/Mn less than~48 as the common subgroup. They contain predominantly angular rather than rounded olivine ( Fig. 2; Table 3). In Fig. 2a, we plot the Fe-Mg-Mn systematics indicate those PMG that have been reported to contain rounded olivine with large circles and others with small squares. Most PMG containing rounded olivine occur in the high-FeO We plot all available data from the literature that are of sufficient quality, and new data from this study. Those PMG containing rounded olivine are indicated in (a) by white circles; those containing farringtonite are indicated in (b) by white squares; all other PMG are indicated by black squares. Farringtonite-bearing pallasites without information for Mn concentrations in olivine are given on a separate scale below each subfigure. Replicates of the same pallasite are indicated by gray fields. Each of the subgroups identified in this study are labeled (common subgroup PMG; low-MnO PMG, and high-FeO PMG-as). Following the method of Scott (1977a), we label selected pallasites with the following abbreviations: Brahin (Bra); Brenham (Br); Cumulus Ridge 04071 (CMS); Esquel (Es); Fukang (Fu); Hambleton (Ha); Imilac (Im); Krasnojarsk (Kr); Molong (Mo); Pavlodar (Pa); Seymchan (Sey); Springwater (Sp); Sterley (Ste); Southampton (Sou); Thiel Mountains (TM); Zaisho (Za). PMG-as and low-MnO subgroups. There are a few exceptions to this pattern, the most important being Pavlodar, which is anomalous in containing rounded olivine of composition Fa 12.5 with unusually high MnO contents for the common group PMG (0.35-0.37 wt% yielding Fe/Mn of~32; Smith et al. 1983;Boesenberg et al. 2012). Information on texture and olivine Fe-Mg-Mn systematics is lacking for some pallasites. For example, the high-FeO PMG Zaisho was first reported to contain fragmental olivine (Shima et al. 1980) and a more detailed study of a small nonrepresentative section contained abundant chromite, with only sparse olivine less than 1 mm in size (Buseck and Clark 1984). This olivine appears to be neither of the relatively large (~5 mm diameter) rounded or angular types, but fragmental grains that had undergone more extensive rounding than could be described "microscale," as occurs in nearly all other pallasites. For a few pallasites, the Mn content of olivine has not been reported, including interesting high-FeO PMG-as such as Rawlinna and Phillips County (Fig. 2). We have compiled textural information and literature EMPA data for the Fe-Mg-Mn systematics of PMG olivine in Table 3 and in Table S2 in Data S1.
Aside from Mn, available PMG minor and trace element data are few and coverage of the various subgroups is patchy. On the scale of whole-grain olivine analyses, the large range in concentration of the refractory lithophile element Sc (0.7-2.1 ppm) in the various PMG has been used to argue for pallasite silicates as olivine restites rather than cumulates (Mittlefehldt 1999(Mittlefehldt , 2005Mittlefehldt and Rumble 2006;Mittlefehldt and Herrin 2010;Boesenberg et al. 2012). However, the large scatter in Sc contents for pallasite olivine could possibly be explained by sampling noise due to intracrystal chemical and diffusion effects, such as correlations between Al, Ti (Hsu 2003) and Sc (McKibbin et al. 2013). The concentration of Sc has a large range within olivine in Brahin (0.9-1.7 ppm; McKibbin et al. 2013), which nearly covers the whole pallasite range. Because only a small number of PMG olivine grains have been investigated in detail for in situ trace element variations, it is clear that a larger study is needed to fully understand these systematics.

Phosphorus-Bearing Minerals
The geochemistry of phosphorus in pallasites is complex but informative, and has resulted in several accessory minerals that we discuss here. Whether coherent subgroups can be identified within the PMG that gather all textural, mineralogical, and geochemical features together is an outstanding question (e.g., Boesenberg et al. 2012), and this is especially the case for the diverse phosphate minerals and phosphoran olivine (with several wt% P 2 O 5 ).
There are two common phosphate minerals in most PMG and other rarer phosphate minerals. The first is merrillite, a Ca-rich anhydrous phosphate equivalent to whitlockite (Ca 9 Na[Fe,Mg][PO 4 ] 7 ), which has been reported under this name in many early pallasite studies (or approximated as Ca 3 [PO 4 ] 2 , e.g., Fuchs 1969). It is equivalent to the phosphates in many other meteorite groups such as ordinary chondrites (e.g., Hughes et al. 2008;Ward et al. 2017 (Dufresne and Roy 1961;Fuchs et al. 1973;Buseck and Holdsworth 1977;Shima et al. 1980;Buseck and Clark 1984;Fig. 2). Wahl (1965) reported farringtonite of a different color to that of Springwater in Huckitta, Marjalahti, Newport, and Pavlodar, but these occurrences do not seem to have been confirmed. The simplified Ca-Mg phosphate phase diagram presented by Ando (1958) included these three main phosphates, indicating that merrillite is a hightemperature refractory phase, that stanfieldite exhibits incongruent (peritectic) behavior and can therefore crystallize via consumption of merrillite and more Mgrich phosphate melt, and that the eutectic lies between farringtonite and stanfieldite. Fuchs (1967Fuchs ( , 1969 suggested that the single-phase field for stanfieldite might be larger than that reported by Ando (1958) and that the coexistence of the Mg-and Ca-bearing endmembers in the same meteorite should indicate disequilibrium. Farringtonite has only been reported in the low-MnO and high-FeO PMG-as, while stanfieldite and merrillite can occur in any subgroup. Excluding breakdown phosphates that probably formed during terrestrial weathering, there are in addition to the three common phosphates at least two occurrences of a silicophosphate phase in Brahin and Springwater (Buseck and Holdsworth 1977) and a similar phase in Brenham reported as calcic-chladniite (Wasson et al. 1999). We have compiled phosphate mineralogy from the literature in Table S4 in Data S1, and summarize this information in Figs. 2b and 3b.
Rare earth trace element studies of phosphates in PMG found patterns typically strongly depleted in light relative to heavy rare earth elements, and generally at moderate concentrations in merrillite, but low to very low concentrations in stanfieldite and farringtonite. However, some phosphates are rich in light rare earth elements and they may have frequent strong depletions in Eu (Davis andOlsen 1991, 1996;Hsu 2003). In conjunction with the phosphate phase equilibria indicating that merrillite is a high temperature refractory phase while stanfieldite and farringtonite are present at the eutectic (Ando 1958), the trace element systematics are generally consistent with extensive fractional crystallization but with a large number of inherited, potentially xenocrystic or locally processed grains (Davis andOlsen 1991, 1996;Hsu 2003). Davis and Olsen (1991) presented trace element data for silico-phosphate in Springwater that strongly indicate it is also of xenolithic nature, with a contrasting light rare earth element-enriched pattern. No trace element data are available for the silico-phosphate in Brahin or the calcic-chladniite in Brenham.
Phosphoran olivine with~0.8-7.4 wt% P 2 O 5 has been reported in several pallasites (Buseck 1977;Buseck and Clark 1984;Wasson et al. 1999;Fowler-Gerace and Tait 2015). It is likely that continuous variation in olivine phosphorus concentration is possible, but in PMG, most normal olivines have low phosphorus contents at less than 100-200 ppm (e.g., McKibbin et al. 2013McKibbin et al. , 2016. The potentially metastable character of this mineral (Boesenberg and Hewins 2010;Boesenberg et al. 2012) and its presence as overgrowths on other olivine grains (Fowler-Gerace and Tait 2015) suggests its formation during rapid cooling of a silicophosphate melt, during the termination of pallasite magmatic evolution. Phosphoran olivine has been found in the low-MnO PMG Brenham and Brahin as well as PMG-as Springwater, Zaisho, and Rawlinna, and therefore preferentially (or perhaps exclusively) occurs in pallasites that do not belong to the common subgroup PMG (Buseck 1977;Buseck and Clark 1984;Wasson et al. 1999). In Brenham, phosphoran olivine occurs in association with chromite-rich regions lacking in metal (Wasson et al. 1999).
In contrast to phosphates and phosphoran olivine, the more common phosphide mineral schreibersite ([Fe, Ni] 3 P) is present in all pallasites (table 4 in Buseck 1977). In iron and pallasite meteorites, schreibersite likely forms at lower temperature by exsolution from metal (e.g., Clark and Goldstein 1978;Yoshikawa and Matsueda 1992). In terms of trace elements, it is similar to kamacite and taenite in being a host for most siderophile elements, with possible exception of low Cu and high Ga and Mo contents (Danielson et al. 2009). Schreibersite is important in buffering oxygen in iron meteorite systems at lower temperature by its reaction relationship with phosphates (Olsen and Fuchs 1967).

Chromite Geochemistry
Chromite is present in PMG at levels from zero to a few percent (e.g., Bunch and Keil 1971;Buseck 1977), but it can be very heterogeneously distributed and occasionally represents several tens vol% depending on sampling heterogeneity (in "massive chromite" regions in Brenham and Seymchan: Francis and Lange 1987;Wasson et al. 1999;Van Niekerk et al. 2007). The major element geochemistry of PMG chromite displays a complex distribution ( Fig. 3   contents). The common subgroup PMG chromites form a noisy correlation between Fe# and Cr# (defined as molar 100 9 Fe/[Fe + Mg] and 100 9 Cr/[Cr + Al], respectively) while the low-MnO pallasites have low Al 2 O 3 contents (chromite in Brenham and Molong has Cr# higher than 98.5, while in Brahin, it is~97). Chromite compositions for high-FeO PMG-as have only been reported for Zaisho and Phillips County. They are also relatively poor in Al 2 O 3 and have the highest Fe# among PMG except for that in the common subgroup (although PMG-am) Pavlodar ( Fig. 3; Table S5 in Data S1). The highest yet reported Al 2 O 3 contents for PMG chromite are for Pallasovka (Sadilenko et al. 2006), Quijingue (Coutinho et al. 1999;Zucolotto 2000), and Fukang (DellaGiustina et al. 2019) at Cr# 77-78. Chromite from the Omolon pallasite, which contains rounded olivine, appears as an outlier to all other pallasites in Fig. 3, perhaps because only compositional data for small chromite inclusions are available (Sharygin et al. 2006).
There are therefore correlations between several characteristics in PMG that have been masked by disparate studies on limited samples, but are clear from a synthesis of the literature (Table 3; Figs. 2 and 3; Data S1). Anomalous olivine geochemistry, either as high-FeO PMG-as or low MnO PMG, is preferentially associated with rounded olivine morphologies, accessory or abundant farringtonite, rare silico-phosphate (or calcic-chladniite), and phosphoran olivine. As well as predominantly rounded olivine morphologies, the low-MnO PMG contain low-Al chromite, while the chromite in common subgroup PMG exhibits a correlation between Mg and Al contents.

PMG Textures and Candidate Liquids to Drive Rounding of Olivine
The driving force for rounding of olivine in pallasites is minimization of interfacial energy between olivine crystals or crystal fragments and the matrix (Scott 1977b;Ohtani 1983), approximated by a reduction in the olivine surface area (Scott 2017). Olivine rounding is independent of grain growth and the latter may be suppressed if grains are isolated from each other, because olivine chemical components are highly insoluble in metal. In this case, the reduction of interfacial energy between grains is achieved at constant grain volume and only through changes in shape (Saiki et al. 2003). However, most rounded olivine in pallasites probably forms part of an interconnected network, while angular olivines exhibit their morphologies precisely because they are in contact with other olivine grains or were in such an arrangement shortly before being separated by intruding metal (Buseck 1977;Spinsby et al. 2008). The growth of large grains proceeds at the expense of small grains, and can occur via Ostwald ripening when a matrix is present (Lifshitz and Slyozov 1961;Wagner 1961) or as "normal" grain growth in a monomineralic system (Atkinson 1988). Both mechanisms are applicable to pallasites, and higher growth rates prevail at high sulfur contents (Gaetani and Grove 1999;Solferino et al. 2015).
With the clear case of large (frequently~4-5 mm and up to~6 mm diameter) rounded olivine in the low-MnO Brahin pallasite and some crystals with partially angular features, in addition to already welldocumented fragmental olivine (Fig. 1), it is worth reevaluating grain growth and rounding in pallasites and potential correspondence among texture, mineralogy, and geochemistry. The presence of rounded or angular olivine grains seems to be partly a function of how close set the olivine grains are to one another, and whether they share mutual boundaries, as observed here for Brahin but also relevant to many other pallasites such as Brenham and Seymchan (the latter containing angular, rounded, and fragmental olivine; Van Niekerk et al. 2007;Kichanov et al. 2018). One particular difficulty in understanding pallasite petrogenesis has been in finding coherent relationships between these characteristics, including the occurrence or absence of accessory phases such as phosphate minerals and phosphoran olivine. The latter is a conspicuous phase which, among meteorites, is almost unique to pallasites (Wasson et al. 1999;Boesenberg and Hewins 2010;Boesenberg et al. 2012;Fowler-Gerace and Tait 2015) with only a few rare occurrences in other types such as Martian meteorites (Goodrich 2003), carbonaceous chondrites (Wang et al. 2006), and IIE iron meteorites (Van Roosbroek et al. 2015, 2017. Phosphoran olivine with P 2 O 5 at the level of several weight percent is present in low-MnO PMG Brahin and Brenham, and in high-FeO PMG-as Rawlinna, Springwater, and Zaisho (Buseck 1977;Scott 1977b;Buseck and Clark 1984;Wasson et al. 1999; Table 3). It seems to be more common in high-FeO PMG-as Springwater (Fowler-Gerace and Tait 2015) and Zaisho (Buseck and Clark 1984) but is absent in common subgroup PMG, indicating that there is no simple relationship to fayalite content (Table 3). Previously, in the absence of known rounded olivine in Brahin, it was concluded by Boesenberg et al. (2012) that "there is no obvious correlation between phosphoran olivine and either olivine composition or texture." Now, with rounded olivine in Brahin, it seems that phosphoran olivine is confined to low-MnO and high-FeO olivine subgroups, which also preferentially exhibit rounded olivine. It is not known from the common PMG subgroup, which is dominated by angular olivine, for example, Admire, Esquel, and Imilac (Table 3).
Phosphoran olivine is potentially metastable, being intermediate in composition between normal, stoichiometric olivine (containing phosphorus at the level of just a few tens or hundreds ppm) and true phosphate minerals in pallasites (Boesenberg et al. 2012). This suggests that the P 2 O 5 component in pallasites should be considered as an incompatible component in a magmatic system, in which P 2 O 5 is hosted in a melt which crystallizes to olivine and one or more phosphates, or to phosphoran olivine as a quench product. Late crystallization of phosphoran olivine, or the more common assemblage of normal, low-P 2 O 5 olivine with one or more phosphate minerals, indicates that the partially molten assemblage of metal, olivine, chromite, troilite, and phosphate, wetted by interstitial immiscible metal-sulfide-phosphide and silico-phosphate liquids, underwent final cooling from magmatic temperatures. This effectively terminated the evolution of the melt (Fowler-Gerace and Tait 2015). The presence of phosphoran olivine overgrowths on round olivines is initially encouraging for an interstitial silicophosphate melt to promote rounding, with a "phosphorus-bearing silicate melt" having been suggested by Boesenberg et al. (2012). Phosphoran olivine is common in PMG-as Springwater and Zaisho and is also present in Rawlinna (Buseck 1977;Buseck and Clark 1984). However, it poses problems because a silicate melt wetting olivine surfaces should enhance the euhedral form of olivine grains, resulting in conventional igneous textures. This is illustrated in Springwater by common phosphoran olivine overgrowths containing internal serrated oscillatory zoning crystallized epitaxially over normal olivine, with later growth layers having smoother and more undulating morphologies, finally terminating against metal or farringtonite (Fowler-Gerace and Tait 2015). Additionally, phosphoran olivine is rare in the low-MnO PMG having only been observed in Brahin and Brenham, and therefore, a silico-phosphate melt was unlikely to have been present in sufficient amounts to drive rounding in such pallasites (Buseck 1977;Buseck and Clark 1984;Wasson et al. 1999). The common occurrence of phosphoran olivine in massive chromite regions of Brenham, but not other parts of the meteorite, is a good indication that phosphoran olivine and rounding are not directly related (Wasson et al. 1999). Phosphoran olivine has not yet been found in Krasnojarsk, Hambleton, Molong, Phillips County, Sericho, Seymchan, Sterley, or Thiel Mountains although it might be expected (Krasnojarsk has instead been reported to contain locally abundant farringtonite). Neither has phosphoran olivine been found in the anomalous PMG-am Pavlodar, which contains conspicuously rounded olivine with a higher MnO content. The high Ir content of the metal in this pallasite suggests a primitive composition that might therefore have had low P contents (Wasson and Choi 2003). It seems likely that the presence of phosphoran olivine in pallasites and former presence of silicophosphate melt are simply correlated with rounded olivine textures, rather than one being the cause of the other.
Better candidates for the melts that enable rounding of olivine are perhaps found in the nonsilicate constituents of pallasites in which an olivine component is highly insoluble, such as liquid metal-sulfide or -phosphide. The rate of rounding and grain growth is strongly enhanced by the presence of a liquid, and in the absence of a liquid, it may only be possible to drive rounding on a microscale over geological timescales (Ohtani 1983). In a sulfide-and phosphide-free system at high temperatures (1400°C; although still subsolidus and under conditions preventing grain growth), olivine-metal mixtures were readily able to generate microscale rounding (Saiki et al. 2003). However, these conditions were only marginally able to achieve rounding rates consistent with rounding larger olivine with diameters of 500 lm over~2 Myr and much longer timescales for larger olivine grains. It is therefore plausible that microscale rounding of angular or fragmental olivine records the trapping and isolation of these grains in solid metal shortly after olivine separation or fragmentation, shortly after nearly complete crystallization of the metallic groundmass (Scott 2017). This would occur as temperature dropped to near the solidus (~950°C; Raghavan 2004;Starykh and Sineva 2012). The experiments of Saiki et al. (2003) also imply that sulfide and phosphide in the liquid play a critical role in enhancing diffusion rates. Solferino et al. (2015) confirmed the effect of liquid metal-sulfide in enhancing rates of both grain growth and rounding. Although they could not unambiguously identify the grain growth mechanism likely for pallasite olivine (probably a mixture of Ostwald ripening and grain-boundary migration) and had higher sulfur contents in their experiments than in natural pallasites, the obtained growth rates are the most applicable to natural systems. At temperatures of~1300°C only relatively short timescales of~0.1 Myr are required to achieve the typical maximum pallasite grain size of~10 mm ( fig. 7a in Solferino et al. 2015). This timescale would probably be slightly longer at lower sulfide content and lower temperatures nearer the solidus, as long as the assemblage contained sufficient melt (~950°C; Raghavan 2004;Starykh and Sineva 2012).
Considering the ubiquity of the phosphide mineral schreibersite in pallasites, another possible liquid responsible for rounding of olivine might be a phosphorusrich metallic liquid. Immiscibility between Fe-sulfide and Fe-phosphide has long been considered for meteorites (Kracher et al. 1977;Ulff-Møller 1998a) guided by early studies of Fe-S-P systems (Sch€ urmann and Sch€ afer 1968;Sch€ urmann and Neubert 1980;Raghavan 1988). However, some petrographic observations of iron meteorites as well as refinements to the phase diagram suggest that schreibersite would usually form only by exsolution under subsolidus conditions and that immiscibility might be difficult to achieve (Clark and Goldstein 1978;Yoshikawa and Matsueda 1992;Chabot and Drake 2000). Practically, any rounding or grain growth effects due to a heterogeneously distributed P-rich liquid metal might be difficult to distinguish from liquid metal-sulfide. Our inference of domains in Brahin containing only a single accessory mineral over the~cm scale (Fig. 1b) suggests that sulfide-phosphide immiscibility indeed did occur in the Brahin region of pallasites and could be responsible for some rounding of olivine. Similar observations have been reported for the sulfide-rich low-MnO pallasite Hambleton (Johnson et al. 2006a(Johnson et al. , 2006b) as well as the common subgroup pallasite CMS 04071 in which some rounded olivine has been reported (Danielson et al. 2009; Table 3). For Brahin, liquid Fe-sulfide or Fe-phosphide may both be responsible for the olivine rounding, but in the section that we studied, large (frequently~4-5 mm diameter and some larger) rounded olivine does not seem to be preferentially associated with either sulfide or phosphide, but rather with chromite (Fig. 1a).
The heterogeneous distribution of accessory chromite and phosphate minerals in most pallasites indicates that these phases are not responsible for rounding, consistent with expectations for solid-state transformations (Ohtani 1983). These phases are present at abundances of only a few percent at most (e.g., farringtonite in Springwater and Krasnojarsk; Buseck 1977) and although chromite and rounded olivine are associated in Brahin in Fig. 1, they are usually heterogeneously distributed with no correspondence between olivine morphology and local mineralogy (e.g., massive chromite in Brenham and Seymchan; Francis and Lange 1987;Wasson et al. 1999;Van Niekerk et al. 2007). The continuous textures in pallasites containing rounded olivine are more consistent with the liquid responsible for rounding being an interconnected, small degree Fe-S melt saturated in crystalline metal, troilite, and chromite (with possible immiscibility generating an additional Fe-P melt) that migrated out of the pallasite assemblages at a late stage in its crystallization and rounding history, as also concluded from the low abundance of sulfide in pallasites (Buseck 1977;Ulff-Møller et al. 1998).

The Distribution of Mn in PMG Olivine and Origin of Low-Degree Melts
The common PMG contain angular olivine and polycrystalline olivine masses as their primary morphologies (Buseck 1977;Scott 1977b). Boesenberg et al. (2012) suggested that common PMG formed by very thorough extraction of silicate melt from the solid, refractory olivine host, resulting in the production of triple junction boundaries between adjacent olivine grains, and ultimately large masses of olivine. The occurrence of frequently disaggregated olivine masses intruded by metal indicates that common PMG are effectively simple mechanical mixtures of olivine and metal. This olivine may be typical of the "middle mantle" of the pallasite parent body intruded by variably evolved metal (Boesenberg et al. 2012) over various depths (Yang et al. 2010), although magmatic temperatures are far above kamacite-taenite exsolution temperatures and this may mask a long interval in pallasite history (Fowler-Gerace and Tait 2015). The common PMG, like most pallasites, contain remarkably little sulfide (e.g., Ulff-Møller et al. 1998) and no indication of a silico-phosphate melt (lacking phosphoran olivine and farringtonite; Table 3). Rather, their potentially xenocrystic merrillite and stanfieldite (Davis andOlsen 1991, 1996;Hsu 2003) in combination with angular or fragmental textures suggest that the common PMG pallasites do not represent quiescent core-mantle boundaries (in agreement with Yang et al. 2010). Rather, these mechanical mixtures cooled through the solidus over rapid timescales or rapidly expelled their final liquids in order to isolate olivine and prevent rounding (Solferino et al. 2015;Scott 2017).
Unlike the common PMG, pallasites of the low-MnO or high-FeO subgroups contain accessory reaction products including farringtonite and phosphoran olivine, and do not seem to be simple mechanical mixtures of core and mantle materials. Some PMG have been reported to contain only stanfieldite (e.g., Molong and Thiel Mountains in the low-MnO subgroup, and Newport in the common subgroup: Buseck and Holdsworth 1977;Davis and Olsen 1996), while Brahin contains stanfieldite, silico-phosphate, and phosphoran olivine (Sonzogni et al. 2009). However, the more diagnostic minerals farringtonite or merrillite have not yet been observed in these pallasites. We anticipate that further phosphate reports in these and other PMG will follow the trend that we have established.
The common PMG and the low-MnO PMG, despite their dissimilar textures, share the same major element composition at~Fa 12 but differing contents of MnO (Mittlefehldt 1999;Hsu 2003;Boesenberg et al. 2012). The rounded olivine in the latter seems to indicate that Mn has some affinity for the liquid responsible for rounding. This liquid was previously suggested to be liquid metal-sulfide (e.g., Ohtani 1983;Solferino et al. 2015) or a silico-phosphate melt, due to the occurrence of phosphoran olivine and the Mgphosphate farringtonite in these pallasites (Boesenberg et al. 2012). The presence of a late, interstitial, Mn-rich melt in pallasites is indicated by Mn-rich rims on olivine in some pallasites. Of the common subgroup PMG, there is slight enrichment of Mn in olivine from Albin (Hsu 2003) but not in others such as Esquel and Imilac (Miyamoto 1997). One report of Mn zoning in Springwater olivine (Zhou and Steele 1993) stands in contrast to other studies (Leitch et al. 1979;Fowler-Gerace and Tait 2015). However, among low-MnO pallasites, enrichment of Mn in olivine rims is quite clear, indicated by vertical trends in plots of olivine Fe/ Mg versus Fe/Mn ( fig. 2a in Boesenberg et al. 2012) and in line profiles approaching olivine margins (McKibbin et al. 2016). The higher MnO in rims of olivine from low-MnO PMG indicates potential late diffusion of Mn into olivine, or late overgrowths. In either case, it implies two stages of equilibration or crystallization of olivine, the first to establish homogeneous interiors (by comparison with common PMG, in the presence of a preferred host for Mn) and the second in which the Mn was redistributed into the olivine rims under different boundary conditions. This model stands in contrast to those in which Mn is considered to be unmodified by repartitioning between olivine and metal or sulfide (e.g., Mittlefehldt 1999). In a series of olivine-metal equilibration and diffusion experiments, Donohue et al. (2018) found very rapid diffusion of Mn in olivine under pallasitic conditions. They inferred that most trace elements in PMG equilibrated across olivine crystals coexisting with metal at relatively high temperatures, in the rangẽ 1300 and~1550°C for Mn in Glorieta Mountain and Albin, and at 1200°C or higher for Cr in most pallasites (the latter approaching Al-in-olivine temperatures of~1000-1290°C from McKibbin et al. [2013] and Prissel et al. [2017]). We conclude that if PMG with normal levels of MnO sample a general starting material for subsequent evolution into low-MnO PMG or other types, then any evidence for early losses or gains of Mn has long since been erased. In the next section, we will use the available literature data to evaluate the role of accessory phases and interstitial melts in controlling the geochemistry of Mn in the PMG environment.

Accessory Phase Controls on Mn and the Petrogenesis of Low-MnO PMG
In order to understand the Fe-Mg-Mn systematics of pallasites, the various hosts for Mn, and ultimately the potential sinks for Mn that might be responsible for the low-MnO PMG, we review here the petrogenesis of those accessory phases (or their sources) that have potential to change the mass balance. These include the various phosphate minerals or their precursor phosphate melt, troilite or its precursor sulfide melt, and chromite.

Phosphate Minerals and (Silico-)Phosphate Melt
The various phosphate minerals in PMG are minor hosts for Mn at variable levels. In the low-MnO pallasites, the least common but potentially voluminous Mg-phosphate farringtonite has been found in Krasnojarsk at an abundance of~1 vol%, while a single occurrence has been reported for in Brenham (Buseck 1977;Wasson et al. 1999). Farringtonite seems to be more abundant in the high-FeO PMG-as, especially in Springwater (~4%) and is also present in other members of that group (Buseck 1977;Buseck and Holdsworth 1977;Buseck and Clark 1984;Fig. 2b). The highly lithophile character of Mg is strongly indicative of farringtonite being produced in large amounts by reaction of olivine or primitive silicate melt with a reservoir of phosphorus and perhaps oxygen. Oxidation of phosphorus that was originally present in the metallic liquid has been suggested in many previous studies, and it was likely present in the pallasitic metallic liquid at the level of 2-4% (e.g., Davis 1982). Similar to the kinetics of rounding and grain growth of olivine, these reactions may have been more likely to occur at higher temperatures in a partially molten state involving phosphorus in liquid Fe-Ni metal coexisting with olivine or ultramafic silicate melt, rather than in the subsolidus state involving phosphorus in metal or schreibersite (as suggested by Davis and Olsen 1991).
The production of significant farringtonite could proceed according to: after Olsen and Fredriksson (1966) and Fuchs (1967) (Equation 1); and modified from Davis and Olsen (1991) (Equation 2). These reactions generate either fayalite or ferrosilite, and the former would be qualitatively consistent with the production of high-FeO PMG-as such as Springwater by conversion of some of the forsteritic component of common PMG olivine to a fayalitic component (although this would not explain higher Ni content in the metal of these pallasites). For Krasnojarsk, the production of 1 vol% farringtonite by this mechanism would probably not go unnoticed in a corresponding enrichment in the fayalite component of olivine, but by comparison with other low-MnO PMG such as Brenham, this does not seem to have occurred. We note, however, that the total area of Krasnojarsk studied in Buseck (1977) is only 38 cm 2 compared to 475 cm 2 for Brenham and 376 cm 2 for Springwater. The high abundance of farringtonite and lack of other phosphates in Krasnojarsk might indicate unrepresentative sampling, and a lower modal proportion of farringtonite production by Equation 1 might not increase the fayalite content of olivine in this pallasite by very much. By contrast, the latter reaction (Equation 2) would require the presence of a silicate melt to carry away the excess pyroxenite component. Alternately, (3) was suggested by Kracher (1983) but results in even more abundant pyroxene for a given amount of farringtonite. The absence of large pyroxene masses in these pallasites suggests that Equation 3 was not important for the production of farringtonite. In support of a pyroxene-free production of farringtonite, Brunet and Laporte (1998) produced synthetic analogs to pallasites (in a Ni-and S-free system at moderate pressure) containing farringtonitic melts coexisting with olivine and metallic melt, with stoichiometry achieved via modification of the olivine composition (presumably via Equation 1). While loss of a pyroxenitic silicate melt is also an option, the correspondence between rounded olivine, anomalous (low-MnO or high-FeO) olivine geochemistry, and the presence of farringtonite and the occasional closely related phosphoran olivine, all suggest that a relatively simple magmatic environment was responsible for the petrogenesis of such pallasites. The MnO contents of farringtonite as obtained by EMPA are available for Krasnojarsk, Springwater, Zaisho, and Rawlinna (Fuchs et al. 1973;Buseck and Holdsworth 1977;Shima et al. 1980;Buseck and Clark 1984;Davis and Olsen 1991). For this data set, Krasnojarsk farringtonite has the lowest MnO and FeO content (0.13 and 2.7 wt%; Fuchs et al. 1973). Farringtonite in Zaisho has the highest reported MnO (0.25 wt%; Shima et al. 1980) while that in Rawlinna has the highest FeO (5.67-5.78 wt%; Buseck and Holdsworth 1977); that is, the FeO content of farringtonite does not correlate with that of olivine. Farringtonite grains in Zaisho are well resolved from each other by EMPA, with MnO and FeO contents spanning the entire reported range for this mineral in PMG (Buseck and Clark 1984). Farringtonite is therefore only a minor host for Mn in the low-MnO PMG such as Krasnojarsk (where its abundance may have been overestimated due to possible heterogeneity) and of even lesser importance in others such as Brenham (where only one occurrence has been reported). It is, however, more important for the high-FeO PMG-as, but does not seem to be well equilibrated either at the scale of a polished section, or within each region responsible for production of the various PMG subgroups.
The MnO contents of the Mg-Ca-phosphate stanfieldite in Springwater and Rawlinna are higher than their coexisting farringtonite grains at 0.45-0.66 and 0.25-0.32 wt%, respectively, although they may also not be at equilibrium due to the coexistence of more than two phosphate minerals in the same meteorites (Buseck and Holdsworth 1977;Davis and Olsen 1991). Unfortunately, stanfieldite has not been observed in Krasnojarsk, and MnO was not reported for this mineral in Zaisho (Buseck and Clark 1984). The MnO contents of stanfieldite grains in other PMG are similar and typically~0.3-0.4 wt% (Buseck and Holdsworth 1977) although with one higher report for Omolon (0.35-0.55 wt%; Sharygin et al. 2006). Stanfieldite is therefore another minor host for Mn, but its lower volumetric abundance in most pallasites (usually much less than 1 vol%; Buseck 1977) suggests that it is not responsible for the depletion of Mn in the olivine of low-MnO PMG. Trace element coverage for stanfieldite is poor, but the possibility that some may be xenocrystic in the same way as merrillite also needs to be further investigated (Davis and Olsen 1991;Hsu 2003).
Finally, of the three main phosphate minerals, the refractory calcic endmember merrillite seems to have the least potential to be an important host for Mn, with levels near or below EMPA detection limits (~0.05 wt% in Buseck and Holdsworth 1977), and with the added complication that such phosphates are likely to be nonequilibrium phases or even xenocrystic. A lack of complete equilibrium is indicated by the presence of farringtonite, as in Springwater and Rawlinna (Ando 1958). The high rare earth element contents of merrillite also indicate a petrogenesis involving a trace elementrich melt (Davis and Olsen 1991;Hsu 2003;Boesenberg et al. 2012). Some of the stanfieldite grains also exhibit such xenocrystic trace element patterns, likely excluding them as an important host for Mn in the low-MnO PMG (Davis and Olsen 1991;Hsu 2003). The occurrence of every pallasite phosphate in one or another low-MnO PMG (merrillite, stanfieldite, farringtonite, and silico-phosphate including potentially distinct calcic-chladniite) indicates that phosphates have not reacted to equilibrium across the entire low-MnO PMG region (Fig. 2b). We therefore conclude that phosphate minerals are probably not the main controlling factors in the Fe-Mg-Mn systematics of PMG and especially those of the low-MnO PMG.
Although phosphate minerals themselves are unlikely to be responsible, a silicate, phosphate, or silico-phosphate melt should also be briefly considered. An olivine/silicate melt partition coefficient for Mn of approximately unity prevails at conditions relevant to pallasites and is not sensitive to the pressures (fugacities) of oxygen (fO 2 ) or sulfur (fS 2 ; e.g., fig. 6 in Gaetani and Grove 1997). The MnO content of phosphoran olivine in low-MnO and high-FeO PMG-as subgroups may serve as a proxy for a silico-phosphate melt. It is, however, very similar to that of coexisting olivine and Mg-bearing phosphates farringtonite and stanfieldite (e.g., 0.26-0.52 wt% : Buseck 1977;Buseck and Clark 1984;Fowler-Gerace and Tait 2015). Extraction of a voluminous silico-phosphate melt therefore does not seem to be an obvious sink for Mn. We will consider a different case of melt escape below.

Troilite and Sulfide Melt
Troilite (FeS) is present in all pallasites at levels of around one or a few percent (e.g., Buseck 1977). It is clear that the present abundance of troilite in pallasites is lower than expected from the geochemistry of the metallic phases, and that liquid sulfide must have somehow been lost (Ulff-Møller et al. 1998). However, the partially chalcophile character of Mn might still be recorded in troilite Mn concentrations or could be inferred from previous work on such systems. There are not many data available for Mn in pallasite troilite but where given they are quite low compared to olivine. Danielson et al. (2009) reported 0.025 wt% Mn in troilite from CMS 04071 using EMPA (along with Cr contents of 0.09 wt%), while Zucchini et al. (2018) report a single analysis of sulfide from Mineo with 0.01 and 0.07 wt% Mn and Cr (their table A3). Such levels are in general agreement with early determinations on separated bulk troilite from Brenham and Springwater, both yielding 0.032 wt% for Mn as well as 0.14 and 0.09 wt%, respectively, for Cr (Lovering 1957), consistent with in situ methods assuming very slight contamination of the bulk troilite sample by another phase. Similar levels of 0.017-0.053 wt% were reported for troilite from the IIIAB iron Cape York (Jochum et al. 1975), which is geochemically very similar to pallasitic metal. Troilite therefore seems to be a poor host for Mn. Considering its low volumetric abundance, it is probably less significant for the subsolidus mass balance of Mn in the low-MnO PMG than most other accessory phases.
Although troilite itself may not host much Mn, the Fe-Ni-sulfide melt from which it crystallized can host higher levels. A number of experimental studies are relevant to the partitioning and Mn budget of pallasites, especially those at low or only moderate pressures and low fO 2 and water contents. In Fe-bearing systems comprising silicate and sulfide melts, the partitioning of partially chalcophile divalent trace elements such as Mn between silicate and sulfide is controlled by reactions such as: and with sulfide-silicate partition coefficients a function of fS 2 0.5 /fO 2 0.5 and not a strong function of pressure (Gaetani and Grove 1997;Li and Aud etat 2012;Kiseeva and Wood 2013). At conditions relevant for pallasites, Mn can exhibit partially chalcophile behavior and partitions into sulfide melt at sufficiently high fS 2 / fO 2 . In the experiments of Drake et al. (1989) in which sulfide was present but fS 2 was not controlled, Mn exhibited very slightly chalcophile around the IW buffer, with a D value of~0.1 increasing to~1 as fO 2 was lowered over the next one to two orders of magnitude. However, these experiments were probably conducted under quite low effective fS 2 . The experiments of Gaetani and Grove (1997) were conducted under controlled fO 2 and fS 2 and also crystallized olivine and chromite. Here, Mn was found to be partially chalcophile, but only at quite low FeO contents in the silicate melt (FeO~3-4 wt% and D values for Mn around unity at log fS 2 0.5log fO 2 0.5 = 4.5, corresponding to very slightly below the IW buffer). This outcome was supported by moderate pressure experiments by Kiseeva and Wood (2013) in which D for Mn of~1 was obtained at FeO contents of~5.0 wt% in silicate melt, that is, still quite low compared to pallasite olivine or likely silicate melts (the partitioning of FeO for olivine-silicate assemblages is close to unity in Fe 2+ -dominated basaltic systems, e.g., Gaetani and Grove 1997;Putirka 2005Putirka , 2016Toplis 2005;Mysen 2007;Matzen et al. 2011). This indicates that D values for Mn were very unlikely to be near to or higher than unity in the PMG environment, and only partially chalcophile behavior is expected.
Nevertheless, the potentially strong influence of sulfide on Mn distribution in olivine-sulfide mixtures is supported by other types of experiments originally intended to assess rates of olivine rounding and grain growth, but which also yielded hints as to the fate of Mn in such systems. Such experiments have been conducted on olivine-sulfide assemblages in the absence of silicate melt, with rather high modal contents of sulfide (~10-30 vol%). Gaetani and Grove (1999) conducted olivinesulfide experiments at similar conditions to those presented in their earlier study on sulfide-silicate melt partitioning (1350°C and various fO 2 and fS 2 in Gaetani and Grove 1997). They reported low MnO contents in olivine indicating variable degrees of extraction, dropping from 0.14 wt% in their starting San Carlos to 0.05-0.13 wt% in their product olivine. The complete Mn inventory was not fully accounted for in the recombined sulfide melt composition obtained from sulfide quench products and may be hosted in small Mn-rich phases. Similarly, the olivine reported in Solferino et al. (2015) lost about half of its Mn inventory, dropping from 0.23 wt% in their starting San Carlos to levels of typically 0.09-0.12 wt% in the products. These experiments indicate that in the absence of chromite and silicate melt, high modal sulfide contents might be able to extract significant Mn from olivine, which presumably crystallizes as heterogeneously distributed alabandite (MnS) or could be taken up by chromite in natural systems.

Chromite
Chromite is a very common accessory phase in PMG but is heterogeneously distributed (mostly~1 but locally up to 30 vol% in section depending on sampling heterogeneity; Buseck 1977;Wasson et al. 1999). The diverse major element chemistry of chromite enables it to track two processes in pallasitic environments. The Al 2 O 3 content records the former presence of silicates, especially plagioclase or a basaltic component, and chromites with variable alumina contents are found in common subgroup PMG that also contain predominantly angular olivine. In low-MnO PMG and high-FeO PMG-as that also contain predominantly rounded olivine, alumina contents are low or very low, strongly suggesting an origin that does not involve, or has much less to do with, an Al-bearing silicate melt. While Cr-Al systematics are preserved due to their slow interdiffusion in chromite (Suzuki et al. 2008), this mineral simultaneously tracks Mg-Fe exchange with olivine on shorter timescales due to relatively fast interdiffusion of these elements in both phases (Van Orman and Crispin 2010;Vogt et al. 2015). During decreasing temperature, the Mg# of chromite falls and that of olivine rises, with zoning especially obvious near rims or the interface between both phases (Wasson et al. 1999). The production of voluminous Mg-bearing and low-Al chromite, by analogy with Equation 1 for phosphates, could proceed according to a simple reaction such as: and the remaining pure chromite component FeCr 2 O 4 could crystallize directly from the metallic melt. As for the P-P 2 O 5 reaction occurring in the presence of xenocrystic phosphates, in the common subgroup PMG, this occurred for Cr in the continuous presence of chromite indicated by some chromites with significant Al 2 O 3 contents (Ulff-Møller 1998b; Wasson et al. 1999).
In this scenario, chromite may be a refractory, restitic phase or contain inherited cores. Other metal-bearing meteorites provide clear evidence for metallic melts evolving toward or finally becoming buffered by P-P 2 O 5 and Cr-Cr 2 O 3 redox reactions. These include troilite nodules in IIIAB iron meteorites associated with schreibersite, chromite, and various phosphate minerals (Olsen et al. 1999) and P 2 O 5 -and Cr 2 O 3 -rich rims on silicate inclusions in the IIE iron Netscha€ evo, which crystallized phosphoran olivine and chromite (Van Roosbroek et al. 2015, 2017. For olivine-chromite Mg-Fe exchange, Wasson et al. (1999) calculated quite low temperatures of 600-800°C, that is, far subsolidus and indicating extensive reequilibration between these minerals during cooling. Geothermometry by Al-in-olivine coexisting with spinel is more likely to preserve crystallization temperatures (Wan et al. 2008;Coogan et al. 2014). Quite low Al contents are present in pallasite olivine, with lower contents for a given chromite composition indicating lower temperatures. McKibbin et al. (2013) used laser ablation inductively coupled plasma mass spectrometry to find 15-30 ppm Al in Brenham and 3-33 ppm in Brahin; similar values of 10-50 ppm were found by Prissel et al. (2017) for Brahin using EMPA. These correspond to temperatures of~1000 to 1290°C (Prissel et al. 2017). However, the lack of much correspondence between Al distributions in olivine and the olivine morphologies also indicates that these distributions do not necessarily relate to the current petrographic arrangement, and that cores may not have been well equilibrated (McKibbin et al. 2013). Decreasing Al contents at olivine margins indicate either late overgrowths at lower temperature or diffusive modification of olivine, consistent with prolonged grain rounding in the presence of a liquid and in equilibrium with chromite (e.g., Solferino et al. 2015). The lower Alin-olivine temperatures of~1000°C are broadly consistent with residence in a Fe-Ni-S melt (eutectic 950°C; Raghavan 2004;Starykh and Sineva 2012) although perhaps not with lower temperatures associated with more complex Ni-and P-bearing melts (Ni-P eutectic at 850-875°C; Doan and Goldstein 1970).
Chromite in PMG is an important host for Mn, of which the concentrations are imperfectly correlated with Fe#. Most have MnO concentrations in the range 0.51-0.84 wt%, but some lower MnO concentrations are found in the cores of chromite in the low-MnO PMG Brenham and Molong (down to 0.30 wt%; Wasson et al. 1999) and a few higher values were reported from some of the common subgroup PMG (1.13, 1.02, and 0.93 wt% in Glorieta Mountain, Marjahlati, and Pavlodar, respectively). Higher Mn combined with lower Al in the rims of Brenham chromites indicates that although olivine-chromite re-equilibration may well occur, the growth of new chromite in the presence of decreasing amounts of silicate melt is also important (Wasson et al. 1999). However, diffusive redistribution of Mn in chromite is very likely to have occurred, because of the likely similar diffusion rates to Fe in chromite (Van Orman and Crispin 2010). Therefore, even though Mn and Al zoning might correlate, they are almost certainly controlled by different diffusion processes.
The IIIAB iron meteorites are geochemically very similar to pallasitic metal (if not in fact genetically related) and serve as a test of the role of olivine in pallasitic systems. By contrast with pallasites, in the absence of silicate phases and lacking a source of Mg and Al, the chromite in IIIAB irons is almost pure chromite (FeCr 2 O 4 ) with a significant Mn content that is presumably derived from the metal-sulfide melt (~0.5-2 wt% MnO; Olsen et al. 1999). Chromite therefore has a high capacity to trace the Mn concentration of the metal-sulfide, but in pallasites, it is also likely to reequilibrate with olivine during slow cooling. By analogy with Mg-Fe exchange, the Mn content of chromite should increase with decreasing temperature and the present MnO contents in pallasite chromite are therefore likely to be upper limits for magmatic temperatures. In any case, some combination of metalsulfide melt and chromite are likely to be responsible for the depletion of Mn in the low-MnO PMG subgroup, which also contains rounded olivine indicative of a considerable residence time in Fe-Ni-S melt.

Modal Mineralogy of Pallasites and Use of Mn as a Tracer of Melt Escape
The partial compatibility of Mn in liquid sulfide (and in chromite, which crystallizes from this liquid) might be reflected in the modal mineralogy of PMG and could potentially be used to track the depletion of sulfide in the pallasite parent body. Although the coarse grain size of pallasites adds sampling noise, most determinations of modal mineralogy have found less than 5 vol% troilite and less than 2 vol% chromite (Buseck 1977). On the basis of troilite/olivine and chromite/olivine ratios, a number of PMG stand out at high values for one or both, while most lie near the origin reflecting high olivine contents ( Fig. 4; in these ratios, we use troilite as a tracer of sulfide and olivine rather than Fe-Ni metal, because Mn is not hosted in metal). Regions of massive chromite from Brenham obviously yield high chromite/olivine ratios, but they also have relatively high troilite/olivine ratios (Wasson et al. 1999). At the other extreme, Phillips County, Hambleton, the chromite-free region of Brenham, Springwater, Krasnojarsk, and Thiel Mountains (all containing rounded olivine) as well as Glorieta Mountain, Newport, and Albin (common subgroup PMG) have high troilite/olivine ratios at low or zero chromite/olivine. Most PMG are near the origin, with Pavlodar (an anomalous member of the common subgroup that is also PMG-am) and Rawlinna (high-FeO PMG-as) having the lowest troilite/olivine (Fig. 4).
The two most sulfide-rich pallasites are noteworthy because of their anomalous olivine compositions which include fragmental olivine. The high-FeO PMG-as Phillips County has a highly heterogeneous metal composition (Wasson and Choi 2003). From its high troilite/metal ratio of~5-10 (Scott 1977b), we conservatively estimate a value for the troilite/olivine ratio of~0.5-1.0 (no olivine abundance is available in the literature). The low-MnO PMG Hambleton is similarly rich in sulfur (Johnson et al. 2006a(Johnson et al. , 2006b. Glorieta Mountain is an unusual pallasite in several respects: as well as a high sulfide content, it has heterogeneous olivine and chromite compositions (relative variation of more than 10% and 9%, respectively; Buseck and Goldstein 1969;Boesenberg et al. 2012); it has been classified as PMG-am for its low Ga, Ge, and Ir contents (Wasson and Choi 2003); it contains more radiogenic W (Quitt e et al. 2005); and it had the lowest cooling rate at low temperatures of all pallasites investigated so far (Yang et al. 2010). Aside from high troilite contents, Albin and Newport are noteworthy only for their similarity to Glorieta Mountain in having heterogeneous chromite compositions (Wasson et al. 1999;Boesenberg et al. 2012; Fig. 3).
To Fig. 4, we have added contour lines indicating the distribution coefficient D Mn which we define for the Mn sink and olivine (i.e., [sulfide or chromite]/olivine), in a simple partitioning model for a starting olivine composition with 0.3 wt% MnO (similar to common subgroup PMG) being modified to yield 0.2 wt% MnO (similar to low-MnO PMG). The identity of the phase that extracts Mn is not immediately important but does have petrological implications: in the case of sulfide, a melt can be removed after extracting Mn from olivine, while in the second, solid chromite is unlikely to be transported very far and would maintain the Mn locally in the newly formed low-MnO PMG region. In Fig. 4, one can see that only in the unlikely case of Mn being highly incompatible in olivine (i.e., highly chalcophile, with D Mn~1 0) is it possible to explain the low-MnO PMG by Mn redistribution between olivine and the sink in approximately their observed abundances (and in any case, such high Mn levels are not observed in either troilite or chromite). If Mn were moderately incompatible in olivine (D Mn~1 -3), then there would be more capacity to produce the low-MnO PMG from the common subgroup PMG. However, Mn is unlikely to have been quite so chalcophile (Gaetani and Grove 1997;Li and Aud etat 2012;Kiseeva and Wood 2013) and the high local abundance of massive chromite in certain parts of Brenham is not representative of the low-MnO PMG region. Large volumes of chromite probably had effective feeding zones across much larger scales of 10-100 cm (Wasson et al. 1999). At more realistic Mn distribution values slightly less than unity (in our definition, moderately compatible in olivine and only slightly chalcophile; Kiseeva and Wood 2013), values for the sulfide/olivine or chromite/olivine ratios of~0.5 become reasonable for extracting sufficient Mn from the common subgroup PMG to produce the low-MnO PMG. Such a ratio probably only exists in the troilite-rich high-FeO PMG-as Phillips County (Scott 1977b). This implies that for realistic, slightly chalcophile behavior for Mn, loss of a substantial amount of sulfide melt might conceivably explain the olivine composition in low-MnO PMG. Sulfide loss might have occurred either by a simple extraction at near-eutectic temperatures (~950-1000°C; Raghavan 2004; Starykh and Sineva 2012) implying a bulk sulfide/olivine ratio of 0.5 for the precursor to the low-MnO PMG. Alternately, percolation of larger volumes of sulfide melt through the low-MnO PMG region could have occurred, and would be necessary if the D Mn value was low (e.g., around~0.1). Percolation could occur by disseminated or channelized flow, the latter of which would act to increase the effective sulfide/olivine ratio locally. From their high troilite contents, Phillips County and Hambleton might be samples of trapped channelized sulfide melt. The very low sulfide abundance in most common subgroup PMG (Buseck 1977;Ulff-Møller et al. 1998) might be consistent with crystallization of metal around angular olivine in that region at high temperature, isolating these olivine grains and preventing rounding beyond the microscale (Scott 2017). Sulfide would then have been available for transport through the low-MnO PMG and high-FeO PMG-as regions to enable macroscale rounding in those areas. A high abundance of sulfide in the PMG parent body of around 10% is expected from the similarity of IIIAB and PMG metal geochemistry, and especially the partitioning behavior of highly siderophile elements that are sensitive to sulfur content in the melt (Chabot 2004;Goldstein et al. 2009;Boesenberg et al. 2012). The lack of troilite in most pallasites, but occasional troilite-rich specimens such as Phillips County, Hambleton, and Glorieta Mountain indicate concentration of sulfide into poorly sampled pockets. The low and very low cooling rates at low temperatures for Glorieta Mountain and Hambleton, respectively (Yang et al. 2010), further indicate that sulfide pockets or a small sulfide inner core might exist inside the PMG parent body (as suggested by Boesenberg et al. 2012). Such a core could be a significant sink for Mn in the PMG. These pallasites might be samples of a drainage channel feeding that core, or less likely could be samples of the core itself. Further evidence for large late structures in the PMG parent body might be indicated by a small compositional gap between chromites of the common subgroup PMG with the highest alumina contents and those of intermediate content, that is, between Ahumada and Otinapa, with the chromites of Glorieta Mountain lying on either side of this gap (Bunch and Keil 1971;Boesenberg et al. 2012;~Cr# 87 in Fig. 3). Interesting zoning in chromite from Otinapa and Newport ( fig. 5 in Boesenberg et al. 2012) combined with their partially fragmental textures (Buseck 1977;Fig. 4. Modal chromite/olivine and troilite/olivine for PMG from Buseck (1977) and Wasson et al. (1999;Brenham massive chromite). Contours for D Mn (for [chromite or sulfide]/olivine) have been superimposed to illustrate required degree of chalcophile behavior to account for Mn concentrations in olivine from low-MnO PMG subgroup. Symbols and labeling of selected pallasites as for Figs. 2a and 3a. Scott 1977b;Haag 2003) might indicate potential for chromites to record transport of grains or liquids in breccia channels, or other more complex histories.

Trace Element Geochemistry of Olivine and Metal
An important question of interest for the PMG parent body, and for the parent bodies of nonchondrite meteorites in general, is whether they typically underwent bulk or fractional style melting (e.g., Boesenberg et al. 2012) and following from this, whether magma oceans of any significant depth formed at their surfaces (e.g., Wilson and Keil 2017). Here, we bring together some aspects of the available trace element data to present a synthesis of pallasite petrogenesis for the PMG parent body.
The simplest co-magmatic relationship between olivine in common subgroup PMG and that of high-FeO PMG-as would be a process of igneous fractionation, by either equilibrium or fractional crystallization (or alternately, melting to leave refractory olivine). However, on the basis of trace element analyses of olivine on various scales, it has been concluded that noisy trends defined by olivine from the common subgroup PMG do not extrapolate to that of the PMG-as (from whole-or intra-grain data using instrumental neutron activation analysis and EMPA; Mittlefehldt 1999Mittlefehldt , 2005Mittlefehldt and Rumble 2006;Mittlefehldt and Herrin 2010;Boesenberg et al. 2012). The degree of fractionation is also unreasonably high for pallasites to represent a cumulate pile at the coremantle boundary (e.g.,~33% from Sc; Mittlefehldt 1999). A simpler relationship was inferred from the mineralogy and mineral chemistry of low-MnO PMG and high-FeO PMG-as, in which some characteristics appear to be related by simple in situ reactions (Wasson and Choi 2003;Yang et al. 2010;Scott 2017).
Here, we would review two points on the relationship between the PMG subgroups from the perspective of lithophile element geochemistry. First, the difference between common (as well as low-MnO) PMG and high-FeO PMG is dictated in a general fashion by oxygen fugacity (Righter et al. 1990), with more magnesian olivine and metal richer in Fe existing at lower oxygen fugacities than more fayalitic olivine and metal poorer in Fe. Olivine and its coexisting metal may therefore have exchanged Fe at an early stage of planetesimal differentiation, simply as a function of available oxygen. In this model, the mantle or pallasite forming region was not fully depleted by melt extraction, because pyroxene or silicate melt must have been present in order to accommodate the additional FeO and prevent w€ ustite production. This pyroxene or silicate melt has subsequently been lost from the PMG region, in a similar fashion to sulfide. In the PMG, those pyroxenes that are present appear to be secondary products crystallized as intergrowths with other accessory phases, and with only very low CaO contents (e.g., Buseck 1977;Boesenberg et al. 2012). This finegrained pyroxene does not record the final widely distributed (and potentially Ca-and Al-bearing) parental silicate melt, but rather subsequent subsolidus reactions (Boesenberg et al. 2012) or very low degree remelting and sulfidation events. The signature of any parental silicate melt has therefore been obscured by interaction of olivine with metal-sulfide melt.
Second, although redox balance might qualitatively explain the high-FeO PMG-as and common (or low-MnO) PMG as silicate-metal reaction products with equivalent bulk compositions, the high abundance of farringtonite in the former complicates the model through addition of a phosphate component (Wasson and Choi 2003;Yang et al. 2010;Scott 2017). Oxidation of phosphorus from metal and repartitioning into a silicate melt is favored at lower temperature; this was previously suggested to have occurred in the solid state (Olsen and Fredriksson 1966) but is probably best enabled by trace oxygen in the metallic melt (Davis 1982;Kracher 1983). In the FeO-rich precursor to the PMG-as, reaction at lower temperature should strongly promote oxidation of phosphorus according to Equation 1 (Olsen and Fredriksson 1966;Fuchs 1967) enabling production of farringtonite and potentially phosphoran olivine. The presence of redox-related phosphate and phosphoran olivine in the absence of significant pyroxene strongly suggests, contrary to the assertion of Kracher (1983), that liquid metal can be the source of both phosphorus and oxygen. Both of these phases are also present in low-MnO pallasites but are rather rare by comparison to high-FeO PMG-as, and they are not known from the common PMG (Table 3; Fig. 2 and Table S4 in Data S1). Nevertheless, the textural, geochemical, and mineralogical correlations identified here indicate that the high-FeO PMG-as and the low-MnO subgroups may be more closely related than previously suspected. Full coverage of major and minor element geochemistry of olivine and chromite, and ultimately systematic lithophile trace element investigations of olivine and phosphate phases in all subgroups, will be required for more concrete interpretations.
A different perspective is available from the trace element geochemistry of metal. Although Ni contents of PMG metal are correlated with FeO content of olivine, replicate analyses of the same pallasites indicate that they are affected by sampling bias related to coarsely crystalline kamacite-taenite exsolution (e.g., Scott 1977a;Wasson and Choi 2003). Some siderophile trace element analyses are also affected due to partitioning between these phases, and additionally into small accessory phases (e.g., Cu as native metal or in Cu-sulfide, and Cr in chromite : Buseck 1977;Wasson et al. 1999). Those siderophile elements that are less fractionated during subsolidus kamacite-taenite exsolution, especially Au and Ir, have been used to evaluate how primitive or evolved the metal might be in a simple model of fractional crystallization with occasional trapping of refractory solid metal (Wasson and Choi 2003). Other siderophile elements bring additional information to this scheme; the "anomalous metal" designation for some pallasites (PMG-am) combines several trace element peculiarities (e.g., Au-Ir systematics or Ga and Ge concentrations). Some are due to nonrepresentative sampling of the metal, while others might represent genuinely anomalous compositions. The PMG-am includes Argonia, Brenham, Glorieta Mountain, Huckitta, Krasnojarsk, Pavlodar (Wasson and Choi 2003), and Seymchan (Van Niekerk et al. 2007). In the following discussion, we will re-evaluate the classification and petrogenesis of the PMG-am.
Following Wasson et al. (1999) in using Au rather than Ni as an indicator of the degree of metal evolution, Wasson and Choi (2003) found that Krasnojarsk and Omolon have higher Co, As, and Sb with lower Ni and Cu than expected, suggestive of simply oversampling kamacite and undersampling taenite. Several of the PMG-am might be explained by sampling bias. In Fig. 5a, we plot Au versus Ir for bulk metal from PMG (most data from Wasson and Choi [2003] with some additional data from Scott 1977a;Lauretta et al. 2006;Sadilenko et al. 2006;Van Niekerk et al. 2007;Kissin et al. 2013). In the same figure, we also plot in situ analyses of discrete phases as crosses (indicating analytical errors) from PMG and the closely related IIIAB iron meteorite group (data from Mullane et al. 2004;Danielson et al. 2009). It can be seen that the phase with the highest concentration of Au in such meteorites is not kamacite or taenite, but probably a tiny accessory phase found so far in fine "plessite" mixtures or inclusions (labeled as taenite with inclusions by Danielson et al. 2009). This provides a possible new explanation for highly anomalous metal compositions found for Argonia and one replicate of Phillips County (Wasson and Choi 2003) as a mixture of mostly refractory Ir-rich and Au-poor metal with Ir-poor and Au-rich "plessite" (top-right of Fig. 5a). We suggest that the possible mixing field should be expanded to encompass the entire PMG data set. This would make production of the PMG and related anomalous IIIAB irons (Treysa quintet) from components similar to those in IIIAB meteorites easier than previously suspected (Wasson and Choi 2003;Wasson 2016). If the in situ analyses of CMS 04071 by Danielson et al. (2009) are representative of PMG, then almost all pallasites seem to contain at least some of this plessite component, with the exception of analyses presented for Southampton, which seem to be practically separates of kamacite and taenite (Kissin et al. 2013; leftmost bulk metal analyses of Fig. 5a). Finally, it seems that the inclusion of a large amount of heterogeneously distributed, Au-rich and Ir-poor plessite or other highly evolved xenocrystic metal in Phillips County may also be consistent with its diverse texture including rounded, angular, and fragmental olivine, suggestive of a breccia channel that filled with troilite. Despite its high troilite content, the rounding of olivine might have been prevented simply by a high cooling rate through intermediate, nearsolidus temperatures.
Other potentially important PMG-am examples are identified by anomalous concentrations of the volatile siderophile elements Ga and Ge. For their Au contents, Huckitta and Brenham have higher concentrations, while in Glorieta Mountain, they are lower; Wasson and Choi (2003) interpreted this to indicate volatility of Ga and Ge at high redox states, potentially induced locally by void space within the parent body. On an in situ basis, only CMS 04071 has been investigated for these elements, with chromite found to be the main host for Ga (Danielson et al. 2009). As for most trace lithophile elements in olivine, more complete coverage will be helpful in understanding the geochemistry of Ga and Ge of pallasites and the potential for tracing gas pressures and void space in the PMG parent body.
The conspicuous PMG-am Pavlodar is unlike the other common subgroup PMG in several characteristics. Pavlodar contains olivine with a Fa-content typical for the common PMG subgroup (Buseck 1977;Scott 1977b;Fa 12.5-12.7 with the normal range being Fa 11-14 ), but this olivine is of the rounded variety and has a higher MnO content (0.35-0.37 wt%, as opposed to 0.32 wt% or lower in most other PMG with similar Fa-contents; Smith et al. 1983;Boesenberg et al. 2012). Pavlodar olivine cores contain Cr-rich exsolution features and therefore high bulk olivine Cr contents (~550-600 ppm from Steele 1994;Mittlefehldt 1999) indicative of a potentially homogeneous Cr content at high temperature that later exsolved at lower temperature. This olivine is hosted in metal with a primitive, Ni-and Au-poor, Ir-and Pt-rich composition with chondritic Ga/Ge, indicating a potential parental metallic liquid (Scott 1977a;Wasson and Choi 2003). Chromite in this pallasite is nearly Al-free (Boesenberg et al. 2012), similar to the low-Al 2 O 3 chromites in low-MnO PMG (of which it is not a member) and also near to pure chromite in IIIAB iron meteorites (Olsen et al. 1999). This indicates that it crystallized almost entirely from liquid metal (Ulff-Møller 1998b;Wasson et al. 1999). Despite these unusual characteristics, the oxygen isotope compositions of Pavlodar are within the normal range for PMG, strongly suggesting origin in the same planetesimal source (Greenwood et al. 2015).
The presence of rounded olivine in Pavlodar is a potentially important environmental indicator. Solferino et al. (2015) summarized the controls on wetting of olivine by metal-sulfide melts, indicating that S-, Ni-, and O-poor melt is the least likely to be interconnected by high dihedral angles. Rounding and grain growth in the case of Pavlodar is therefore unexpected, or may represent the most primitive metallic melt composition capable of generating relatively large rounded olivines (scale of several millimeters; Scott [1977b] and references therein; Steele 1994), with more primitive metallic melts becoming trapped as inclusions in olivine grains or in triple junctions in olivine aggregates. The duration of grain growth and rounding of~0.1 Myr found by Solferino et al. (2015) may therefore be a lower limit for Pavlodar. Yang et al. (2010) identified Pavlodar as a potential "core-mantle" boundary sample in the classical case of quiescent and fully separated core and mantle domains. If the metal in Pavlodar represents the most primitive fully interconnected metallic melt, then its anomalous olivine composition may be consistent with more efficient silicate melt extraction.
Additionally, in situ equilibrium crystallization of primitive metallic melt might have prevented oxidation of phosphorus (via concentration of oxygen in fractionally crystallized metallic melt) and explains the lack of farringtonite and phosphoran olivine that are found in other pallasites with rounded olivine (Scott 2017). Those pallasites with metal compositions that are most similar to Pavlodar also lack farringtonite and phosphoran olivine, consistent with this process (e.g., Argonia, Finmarken, Huckitta, Marjalahti, Pecora Escarpment 91004, and Seymchan). Aside from the lack of phosphate-bearing phases in primitive metal, there is no obvious relationship among the PMG between Ir-Au metal systematics and the other correlations that we have identified in this study, namely olivine texture, olivine composition, and type of accessory phases (Al-free or Al-bearing chromite, the various phosphates, and phosphoran olivine). Boesenberg et al. (2012) highlighted the potential importance of incomplete melting scenarios in modeling pallasite petrogenesis. In low-degree fractional melting of a planetesimal that fails to produce a deep magma ocean or complete metal-silicate separation, it is expected that primitive, refractory, high-Ir metal would be preserved in the middle or upper mantle (Scott 2017). Pallasites with higher Ir than expected for their Fig. 5. a) Au and Ir bulk metal PMG concentrations (Wasson and Choi [2003] with some additional data from Scott 1977a;Lauretta et al. 2006;Sadilenko et al. 2006;Van Niekerk et al. 2007;and Kissin et al. 2013) and in situ analyses of kamacite, taenite, and plessite for PMG (Brenham) and closely related IIIAB irons (Mullane et al. 2004;Danielson et al. 2009). In situ analyses of these materials are indicated by crosses with the lengths of each arm indicating analytical error (in Mullane et al. 2004;Danielson et al. 2009) and labeled in the figure. Following Wasson and Choi (2003), we have added a mixing line between primitive and evolved endmembers, but use more recent measurements from Henbury IIIAB taenite and CMS 04071 plessite as endmembers. The mixing curve is labeled with the percentage of the plessite and taenite component. Sulfide-rich pallasites are indicated by "S" inside the symbol. All other symbols are as for Figs. 2a and 3a, and replicates of the same pallasite are indicated by tie-lines. b) Cloudy zone particle size (Yang et al. 2010) versus bulk Ir analyses (Wasson and Choi 2003;Van Niekerk et al. 2007) for PMG samples. Symbols as for (a). Selected pallasites are labeled following Figs. 2-4, with the following additional abbreviations: Argonia (Ar); Huckitta (Hu); Finmarken (Fi); Pecora Escarpment 91004 (PCA). other metal trace element contents such as Argonia, Finmarken, Marjalahti, and one replicate of Phillips County record a mixture of exactly such refractory, solid metal with more evolved liquid metal compositions (Fig. 5a;Wasson and Choi 2003). The Ir-Au systematics do not require input of foreign metal; all analyses seem to be consistent with mixtures of early and late components. This was suggested to have occurred during catastrophic mixing (Wasson and Choi 2003) but could also have occurred by migration of metal partial melts from different parts of the planetesimal while solid Ir-rich metal grains remained stationary in an olivine framework. In addition, preliminary findings for triple-oxygen isotope systematics show no resolvable difference in mass-independent components between the IIIAB and PMG parent bodies, as recorded by chromites in the former and olivine in the latter (i.e., no obviously foreign material; Franchi et al. 2013). Instead, recent data from Ali et al. (2018) indicate that some relationship may exist between modal proportion of olivine and oxygen isotopes within the PMG. Although this is an area of active and ongoing investigation, at this stage these data suggest a scenario in which foreign metal need not be introduced, but rather that high degrees of melting occurred in which complete melting of metal and silicate was never reached.

Paleomagnetic, Cooling Rate, and Isotopic Constraints on the Evolution and Final Structure of the PMG Parent Body
Several pallasites have been shown to record the former presence of a magnetosphere upon cooling through the Curie temperatures of tiny inclusions in olivine and of microstructures in metal (Tarduno et al. 2012;Bryson et al. 2015;Nichols et al. 2016). These paleofields were suggested to have been imparted by a dynamo present during one or more stages of core evolution, from a possible early thermally driven regime (a fully liquid, convecting core), a compositionally driven regime (a partially molten, crystallizing core), or a late remnant field preserved in the solid core or nearby rocky material. High and decreasing paleofield intensities were recorded in Imilac and Esquel by inclusions in olivine and in "cloudy zone" exsolved nanometer-scale tetrataenite inclusions hosted within ordered Fe 3 Ni metal (~120 lT in the former, and Fig. 6. Olivine molar Fe/Mg, Fe/Mn, and metallographic cloudy zone grain size (as a proxy for cooling rate) for available PMG plotted with symbols from Fig. 5. The scale for cloudy zone grain size has been inverted to reflect more rapid cooling resulting in finer exsolution grain size near the surface of a postulated parent body, and slower cooling resulting in coarser grain size in the interior. Replicates of Seymchan (which occur in both common and low-MnO subgroups) are connected with a dotted line. Sulfide-and iridium-rich pallasites are indicated by "S" and "Ir" inside symbols, respectively.
~80 lT decreasing to the detection limit of~10 lT in the latter; Bryson et al. 2015). By contrast, low or zero paleofield intensities were found for Brenham and Marjalahti from cloudy zone exsolved tetrataenite alone (~10 lT or lower; Nichols et al. 2016). In these investigations, the thermal histories of these meteorites were interpreted in the context of a relatively simple structure comprising a fully segregated liquid core with overlying mantle, leading to a model in which a 200 km radius conductive planetesimal cooled over a period of~250 Myr. Nichols et al. (2016) used metallographic cooling rates from Yang et al. (2010) to infer depths and a chronological order of cooling. Higher cooling rates for Brenham and Marjalahti, from cloudy zone particle sizes of 123 AE 3 and 118 AE 3 nm, respectively, correspond to 6.2 AE 0.9 K/Myr in the former (2 SEs). These were taken to indicate that Brenham and Marjalahti were high in the mantle and reached closure relatively early. Lower cooling rates for Imilac and Esquel, from coarser cloudy zone particle sizes of 143 AE 4 and 157 AE 11 nm, respectively, indicated a position lower in the mantle (although not near the core-mantle boundary) and that they reached closure later. In this model, an early magnetosphere driven by thermal convection predated a period of dynamo quiescence, and Brenham and Marjalahti reach closure in the mid-upper mantle during this inactivity. Subsequently, a crystallization-driven regime took over, powering a strong magnetosphere in which Imilac and finally Esquel reach closure sometime in the next 100 Myr or more (Nichols et al. 2016). The need to thermally isolate the pallasitic region from the outwardcrystallizing core led to the conclusion that pallasites formed in the upper or mid-mantle and that the metal was foreign, with locally high Ir concentrations taken as support (supplementary material for Tarduno et al. 2012).
In the PMG, there are local enhancements of Ir indicating that complete melting of metal never occurred (Boesenberg et al. 2012) and no clear tripleoxygen isotopic evidence yet for large mass-independent variation associated with potentially foreign chemistry or mineralogy (thus far only with olivine content; Ali et al. 2018) nor with chromites in closely related IIIAB metal (indistinguishable from olivine; Franchi et al. 2013;Greenwood et al. 2015). Although this is an area of active investigation, we tentatively suggest an alternative interpretation of the paleomagnetic data. At peak temperatures, the PMG metal was subliquidus and the parent body was already in the compositional (crystallization) driven convective regime. Following Boesenberg et al. (2012), the presence of refractory Irrich metal in the pallasitic mantle suggests that the pallasitic region was thick and overlying a partially molten, inwardly crystallizing core. There is a slight tendency for high-Ir PMG to have only moderately coarse (Argonia, Marjalahti, Huckitta, and Seymchan at 125 nm) or fine (Finmarken 86 AE 2 nm) cloudy zone microstructures, while sulfide-rich pallasites are among those pallasites with the coarsest cloudy zone microstructures and therefore the lowest cooling rates (Glorieta Mountain and Hambleton, at 170 AE 6 and 152 AE 6 nm, respectively). These data are illustrated in Fig. 5b, in which it can be seen that the relationship used by Boesenberg et al. (2012) to argue that metal crystallization in the pallasite region occurred from outside to inside, is in fact more subtle than originally implied because of considerable noisy scatter in cloudy zone grain size or geochemistry of the metal. Nevertheless, the aforementioned relationship appears robust and indicates that late sulfide was mobile and concentrated at greater depths in the PMG parent body (Boesenberg et al. 2012). However, the stepwise nature of cooling in the PMG parent body (e.g., Fowler-Gerace and Tait 2015) indicates that the cooling path of any particular package of pallasitic material at high temperature need not correspond perfectly to its cooling rates at low temperature. Features obtained during magmatic evolution (including crystallization of metallic melts, grain growth and rounding of olivine, and accessory phase mineralogy all being set at~1000°C or higher) formed at least several million years before those obtained during cooling at low temperatures (appropriate for generation of metallographic textures, recording of cooling rates and paleomagnetic fields, as well as fission track dating taking place at~400-600°C). Systematic fission track dating of pallasites seems especially promising for dating the cooling rates obtained from metallographic textures, although from only five pallasites a bimodal distribution of ages has been found (4.37 and~4.2 Ga) that is so far uncorrelated with the size of cloudy zone microstructures (Pellas et al. 1983;Bondar and Perelygin 2003). The large difference in temperature and time between magmatic and metallographic cooling stages in the evolution of the PMG parent body, and the long interval required to approach those low cooling rates, implies a more complex planetesimal history in the intervening period.
Rather than invoking the input of foreign metal into the surface of the PMG parent body, we suggest removing this requirement and instead propose a degree of local geological complexity to explain the scattered tendencies in Fig. 5b. There are possible planetesimal scale features that might explain the presence of cold pallasitic material capable of recording a magnetic field, lying in proximity to its native, partially molten convecting core. This arrangement must achieve thermal isolation of the pallasitic region from the partially liquid core, that is, a steeper geothermal gradient in the metalsilicate mixture, or use local amplification of a very weak magnetic field in the core. Both might occur through the presence of large structures in a highly evolved, complex planetesimal. The simplest cases of thermal isolation involve a thick, continuous silicate mantle to provide insulation (Tarduno et al. 2012;Bryson et al. 2015;Nichols et al. 2016) or residence in a porous megaregolith, although a near-surface impact-generated regolith seems unlikely considering the limited evidence for late shock processing. Any early substructure in olivine crystal lattices may have been erased by subsequent recovery, especially in those pallasites containing rounded olivine (Matsui et al. 1980;Mori 1986;Desrousseaux et al. 1997). Springwater exhibits an early preferential alignment of olivine (Fowler-Gerace et al. 2016) and if this is confirmed in other similar pallasites, it would have important tectonic implications. In order to increase the thermal gradient, more useful types of porosity might include cavities in which Ga and Ge could have become volatile (Wasson and Choi 2003) and emptied breccia channels that might have enabled drainage of sulfide into the interior as the planetesimal adjusted to lower temperature (indicated by low cooling rates for sulfide-rich pallasites; Fig. 5b). Such void space was suggested to be possible via contraction of the metal relative to olivine during cooling (Wasson and Choi 2003). The observation of preferential kink banding deformation in angular olivine rather than rounded suggests that those pallasites containing rounded olivine somehow accommodated small degrees of deformation, perhaps by expulsion of very low degree interconnected melts (Klosterman and Buseck 1973;Fowler-Gerace et al. 2016). These kink bands probably formed at low to intermediate temperatures, perhaps close to or slightly higher than that of kamacite-taenite exsolution (Klosterman and Buseck 1973).
A further scenario that might help reconcile the various pallasite characteristics has been opened by the possibility of immiscibility between sulfide-and phosphide-rich melts, suggested by regions in Brahin in which only a single accessory mineral occurs (Figs. 1a and  1b). Separation of Fe-Ni-phosphide melt by immiscibility may enable partial melts to persist to lower temperatures, until finally reaching the Ni-phosphide eutectic in the range 850-875°C (Doan and Goldstein 1970) or even lower if additional soluble elements are present in the concentrated melt. The weak magnetic field associated with such a tiny inward-crystallizing metal-phosphide core could be locally modified by massive channels of metal (e.g., olivine-free regions in Brenham, Glorieta Mountain, and Seymchan) or sulfide (Glorieta Mountain, Hambleton, and Phillips County) with possible amplification by a factor of 3-200 in proximity to extreme morphologies (Bryson et al. 2015). Core crystallization occurring from the outside toward the inside may therefore generate a weak magnetic field that could be locally enhanced by structures in the planetesimal to be recorded by pallasitic material in appropriate locations, for example, Esquel and Imilac residing in solid metal, perhaps in a dendrite. The additional question of whether the IIIAB irons are closely or more distally related to PMG metal has also been recently addressed by cooling rate information. Although the Ni-rich end of the IIIAB sequence is geochemically very similar to the PMG metal, the latter seems to have crystallized in its parent body from the outside to the inside (Boesenberg et al. 2012;Fig. 5b) and the cooling rates the IIIAB irons are higher than those of the PMG (Yang and Goldstein 2006;Yang et al. 2010). This is inconsistent with a simple planetesimal comprising PMG material overlying a IIIAB core. Therefore, we only suggest that components in IIIAB irons are similar or analogous to those in PMG metal. They may nevertheless be related through formation in an even earlier precursor planetesimal, or represent a mixture of two planetesimals (e.g., Yang et al. 2010), or ultimately have inherited a common trace element inventory from the same part of the solar nebula.
The above short history involving rapid early cooling from high to moderate temperatures is supported by short-and long-lived radiogenic isotope studies that indicate rapid pallasite formation and cooling to subsolidus temperatures (e.g., Al-Mg, Mn-Cr, Hf-W, Ag-Pd, and Re-Os). The isotope systematics of lithophile elements in pallasites diverged from chondritic behavior very early. From the Al-Mg decay system, silicate differentiation and isolation of PMG olivine from aluminous melt occurred at 1.24 (+0.40/ À0.28) Myr after solar system formation (Baker et al. 2012). Similarly, rapid cooling of the PMG parent body is indicated by Mn-Cr isotope systematics, yielding diffusive closure of Mn and Cr in olivine of Omolon and Springwater within~10 Myr of solar system formation (Lugmair and Shukolyukov 1998), and possible cooling to near-solidus conditions and segregation of low-degree Mn-rich melts in Brahin and Brenham within~2.5 to 4 Myr (McKibbin et al. 2016). For the lithophile-siderophile Hf-W decay scheme, W-deficit dating methods have undergone considerable revision due to correction for exposure in space (e.g., Kruijer et al. 2014). Despite this, the available data for the PMG parent body metal-silicate differentiation event seem to have been at a similarly early period to most other iron meteorites, except for the more radiogenic Glorieta Mountain (Quitt e et al. 2005). This may be consistent with the slow cooling rate of this pallasite at low temperature (Yang et al. 2010) and its angular olivine and variable mineral geochemistry (Boesenberg et al. 2012). A very similar timescale is indicated for the PMG by the purely siderophile Pd-Ag system, which indicates metal fractionation and isolation of most PMG within~12 Myr of solar system formation. Lower precision chronological information for Re-Os give ages of~60 AE 40 Myr (Shen et al. 1998;Chen et al. 2002). The Re-Os ages might indicate slow closure or later disturbance for this particular system at relatively low temperature. Taken together, the radiogenic isotope ages are more consistent with rapid cooling, rather than the long timescales of simple cooling models used to interpret paleomagnetic data (~100 Myr or longer: Tarduno et al. 2012;Bryson et al. 2015;Nichols et al. 2016). They are therefore encouraging for a planetesimal model involving refractory metal trapped throughout the mantle and a small crystallizing liquid inner core.

CONCLUSIONS
Among the PMG, there is a strong correlation between olivine morphology, specifically the existence of "macroscale" rounded olivine (diameter~5-15 mm ;Scott 1977b), and anomalous characteristics (e.g., PMG-am and high-FeO PMG-as; Kissin 2008Kissin , 2009) including low MnO contents of olivine (e.g., Boesenberg et al. 2012;McKibbin et al. 2016). In the high-FeO PMG-as and the low-MnO PMG subgroups, the MgO-rich and phosphorus-bearing phases farringtonite and phosphoran olivine occur with low-Al 2 O 3 chromite and rounded olivine. Large (diameter~5 mm or more) rounded olivine morphologies and the presence of farringtonite and phosphoran olivine indicate prolonged interaction between silicate and metallic reservoirs in the PMG parent planetesimal, with phosphorous, chromium, and oxygen sourced from the metallic liquid. These minerals might therefore be indicators of genuine core-mantle boundaries or core-mantle interaction zones. The prior removal of silicate melt at the planetesimal's peak temperature (between eutectic melting at 1050-1100°C and upper limits of 1600-1700°C; Boesenberg et al. 2012) indicates that farringtonite and low-Al 2 O 3 chromite formed by back-reaction during retrograde cooling (tõ 1000°C; e.g., Brunet and Laporte 1998). This occurred via oxidation of phosphorus and chromium, which are siderophile at high temperatures but lithophile at lower, near-solidus temperatures.
Farringtonite and phosphoran olivine are not present in the more populous common subgroup PMG (i.e., most pallasites), which may be simple mechanical mixtures of olivine, Al-bearing chromite, primitive metal xenocrysts, and more evolved liquid metal. Large olivine masses or disaggregated olivine crystals hosted in metal may be typical of the "middle mantle" of the pallasite parent body intruded by variably evolved metal (Boesenberg et al. 2012) over various depths (Yang et al. 2010). However, because magmatic temperatures are far above kamacite-taenite exsolution temperatures, this may mask a long interval in pallasite history, and significant geological noise has been superimposed upon the relationships described above through an erratic, stepwise cooling history (Fowler-Gerace and Tait 2015). The Pavlodar pallasite seems to be an early isolate of highly refractory metal, perhaps separated at an early stage, at peak melting and melt extraction.
The geochemistry of Mn in olivine and chromite, particularly the low concentrations in olivine of the low-MnO PMG and high concentrations in low-Al 2 O 3 chromites, traces the former presence of sulfide melt in pallasites and partially chalcophile nature of Mn. Due to fractional crystallization of the metallic liquid, it is likely that a small liquid sulfide inner core, and possibly a very small liquid phosphide inner core, developed in the PMG parent body due to fractional crystallization and liquid immiscibility. Such a small inner core might generate a weak and locally variable dynamo responsible for paleomagnetic signatures in some pallasites (e.g., Tarduno et al. 2012).
The geochemical behavior of Mn in pallasitic environments may be similar across multiple planetesimals. The PES group, although originating in a different parent body in a very different part of the solar system Mayeda 1978, 1996;Warren 2011), shares a very similar accessory phase mineralogy, including Mn-bearing phosphate minerals and troilite. In the PES Yamato 8451, olivine rims are enriched in Mn and have a lower fayalite content than their cores, suggestive of repartitioning or new crystallization during cooling (Miyamoto 1997). For the low-MnO PMG pallasites, a Mn-rich silico-phosphate liquid was suggested to control the late Cr-isotope evolution of such parent bodies, with 53 Mn-53 Cr decay generating isotopic anomalies in interstitial phases and olivine rims (McKibbin et al. 2016). The partially chalcophile behavior of Mn may provide another mechanism by which the fractionation of Mn on planetary scales could occur, with sequestration of Mn along with sulfide in pockets in the interior or in the inner cores of planetesimals (suggested by sulfide-rich pallasites with low cooling rates, such as Glorieta Mountain and Hambleton; Yang et al. 2010). and data processing; and Maxwell Brown for useful discussions about paleomagnetism. We also thank Richard Windmill, Joseph Boesenberg, Ed Scott, Neva Fowler-Gerace, and one anonymous reviewer for their critical reviews and suggested improvements to the manuscript. This research was supported by Planet TOPERS, an initiative of the Belgian Science Policy Office (BELSPO), and by a fellowship from the Research Foundation-Flanders (Fonds Wetenschappelijk Onderzoek-Vlaanderen; FWO-project 12O9515N) to S. McK., who is currently a postdoctoral fellow of the Alexander von Humboldt Foundation.
Editorial Handling-Dr. Edward Scott