Data returned from the gamma-ray spectrometer onboard the Mercury Surface, Space Environment, Geochemistry, and Ranging (MESSENGER) spacecraft have been interpreted to say that Mercury is a volatile-rich planet (elevated K/Th and K/U), which is important given its heliocentric distance. The MESSENGER X-ray spectrometer provided chemical information from the surface of Mercury which we used to calculate an average surface composition for the regions analyzed. The high S abundance and low FeO abundance of the surface indicates that the oxygen fugacity of the Mercurian interior is very reducing (−6.3 to −2.6 logfO2units below the iron-wüstite buffer). At these low oxygen fugacities, elements that are typically considered lithophile can become more siderophile or chalcophile. We review available metal/silicate partitioning data for K and U to show that Mercury's volatile inventory is still an open question, and additional experiments investigating metal/silicate partitioning at the conditions of Mercury's core formation are needed.
 In planetary science, there are several important geochemical ratios that can reveal information regarding the formation, bulk chemistry, and evolution of a terrestrial planetary body. One of the most important and commonly used ratios are those of moderately volatile large-ion lithophile (LIL) elements, such as K, Rb, and Cs, to refractory LIL elements such as Th, U, and some of the largest rare earth elements [McLennan, 2003, and references therein]. Importantly, LIL elements behave incompatibly during melting and crystallization, so LIL element ratios remain fairly constant during typical igneous processes, including planet-wide differentiation. Consequently, the mean LIL element ratio of a terrestrial body's surface is often used as a first-order estimate of a bulk planet's level of volatile-depletion relative to CI carbonaceous chondrites that represent Solar System elemental abundances.
 The volatile depletion of a planet is widely believed to be correlated to temperature of formation due to the stepwise accretion of planetary materials, so it has commonly been assumed that the volatile depletion of the terrestrial planets would increase as their heliocentric distance decreased due to removal of volatiles from the inner Solar System during the Sun's T Tauri phase [Albarède, 2009]. Elemental data that has been obtained from Earth, Mars and Venus is somewhat consistent with this view (Figure 1). Volatile-depletion can also be caused by globally significant secondary processes. For example, the Moon experienced a giant impact and very high temperature of formation, which is believed to be at least partially responsible for it being the most volatile-depleted planetary body analyzed to date (Hartmann and Davis  and Figure 1). It would be expected that Mercury is more volatile-depleted than Earth, Mars, and Venus due to its heliocentric distance, and if Mercury was involved in a giant impact [Benz et al., 1988; Smith, 1979], it would either enhance the volatile-depletion or have no effect; however, recent results from the Gamma-ray spectrometer (GRS) onboard the Mercury Surface, Space Environment, Geochemistry, and Ranging (MESSENGER) spacecraft that is currently orbiting Mercury indicate that it has relatively high K/Th and K/U ratios (5200 ± 1800 and 12800 ± 4300, respectively [Peplowski et al., 2011]). These data show Mercury has a K/Th ratio most similar to Mars and have been interpreted to say that Mercury is as volatile-rich as the other terrestrial planets [Peplowski et al., 2011] (Figure 1). This result holds profound implications for our understanding of volatile-depletion and mixing processes in the early Solar System and Solar nebula, and it will likely affect our understanding of Mercury's formation conditions. Alternatively, Mercury is in many ways an end-member planet, and one should approach making inferences about its geochemistry based on historical knowledge of elemental behavior with extreme caution. In the present study, we consider the newly obtained GRS data for Mercury, along with data from the X-ray spectrometer (XRS) on MESSENGER, to re-examine the volatile-element inventory of Mercury.
2. Composition of the Mercurian Surface from MESSENGER
 The composition of a planet's surface can be highly informative in deciphering the thermal and magmatic evolution of that planet. The MESSENGER spacecraft is presently providing the first quantitative constraints on the elemental composition of the Mercurian surface. In the present section, we synthesize what MESSENGER has directly measured from XRS and GRS. Subsequently, we infer upper limits on the oxidation state of the Mercurian interior during core formation to shed light on elemental partitioning during the formation of Mercury's core, which can potentially alter the K/Th and K/U ratios used to infer Mercury's volatile-rich nature.
2.1. Elements Directly Measured by MESSENGER Instruments
 The XRS instrument on MESSENGER relies on X-ray emissions induced on the surface of Mercury by the incident solar flux (seeSchlemm et al.  for a comprehensive overview). To date, elemental ratios are only available for portions of the Mercurian surface and these regions are outlined in Nittler et al. [2011, Figure 1]. Nittler et al.  reported an average Si abundance of ∼25 wt.% for the regions of interest (ROIs), and we have adopted this value for calculating the average composition of these ROIs. Relative elemental abundances for each of the ROIs are presented in Nittler et al. , and these values were weighted (by surface area of each ROI) and averaged to determine the average surface composition for all of the ROIs (an un-weighted average is also presented inTable 1). The elemental abundances for each of the ROIs (presented as wt.% oxides) along with our calculated average compositions are presented in Table 1. The potassium value is the northern hemisphere average obtained from GRS reported by Peplowski et al. .
Table 1. Compositions (in wt.%) Computed for Each of the ROIs, along with Several Average Estimates for an Average Surface Compositiona
−O ≡ S
Values are normalized to 99 wt.% if Fe included and 96 wt.% if excluded. T – All Fe measured by XRS is assumed to be FeO. Weighted average was weighted proportionally to surface area of each ROI in Figure 1 from Nittler et al. . Equally weighted average gave equal weight to each ROI from Nittler et al. .
Equally Weighted Average
2.2. Estimation of Oxygen Fugacity of the Mercurian Interior
 The oxygen fugacity (fO2) of the Mercurian interior can be constrained by a number of methods. An estimate for the oxygen fugacity of the Mercurian interior was recently presented by Zolotov et al.  using the surface abundances of sulfur measured by the MESSENGER XRS instrument. Zolotov et al.  used the fO2-fS2relations for sulfide-silicate equilibria to constrain thefO2 of the Mercurian interior to be at most 6.3 to 5.5 log fO2units below the iron-wüstite buffer (ΔIW). These low values require ≤0.10 mol.% FeO in the silicate portion of Mercury, which would require that much of the Fe recently measured on the Mercurian surface by XRS is Fe metal or sulfide. The sulfur content of the surface can also be used to estimate oxygen fugacity using the solubility relationship of sulfur in silicate liquids as a function offO2, which was suggested previously by Zolotov . Based on available experimental data for S solubility in silicate liquids [Beermann et al., 2011; Berthet et al., 2009; Holzheid and Grove, 2002; Malavergne et al., 2007a; Mavrogenes and O'Neill, 1999; McCoy et al., 1999; Moune et al., 2009], and assuming the sulfur abundances at the Mercurian surface are representative of S abundances dissolved in the silicate portion of Mercurian magmas, the oxygen fugacity of those magmas would range from approximately −5.0 to −3.0ΔIW (Figure 2).
 An absolute upper limit on the fO2 for mercury can be established if it is assumed that all the Fe on the surface is Fe2+ bonded to oxygen (i.e., FeO). Based on the average Fe content of the ROIs from Table 1, the Mercurian surface has approximately 2.53 wt.% Fe, which would equate to 3.26 wt.% FeO. This value is likely representative of a combination of primary and secondary crustal material at Mercury's surface [Denevi et al., 2009; Taylor and McLennan, 2008]. A primary crust would likely be representative of the latest stages of magma ocean crystallization on Mercury and would represent iron enrichment relative to the bulk mantle [Brown and Elkins-Tanton, 2009; Riner et al., 2011, 2010, 2009]. The secondary crust would be representative of volcanic products resulting from partial melting of the Mercurian mantle, and the FeO content of the surface would be greater than or equal to the value for the mantle source region because FeO has a source/melt partition coefficient greater than or equal to one [Robinson and Taylor, 2001]. Assuming Mercury's surface is dominated by secondary crust would provide us with the absolute upper limit on the FeO content of Mercury's mantle (i.e., 3.26 wt% FeO). If Mercury's core is predominately Fe0, the oxygen fugacity during core formation is buffered by the reaction Fe + O2 = FeO, allowing us to determine the upper limit on the fO2 at the time of core formation relative to the IW buffer (as in Hillgren et al. ). We assume that the Mercurian core has at least 80 mol% Fe, which is consistent with the likely presence of significant Si and S in Mercury's core due to the low fO2 and high pressure of core formation [e.g., Fei et al., 2011; Malavergne et al., 2010; Riner et al., 2008]. Using these mantle and core FeO and Fe contents, the oxygen fugacity of the Mercurian interior at the time of core formation, calculated according to
where γ is the activity coefficient and X is the mole fraction, is estimated to be −3.0 to −2.6ΔIW. The lower value for this range was computed assuming the activity coefficients for FeO and Fe are both unity. The upper value for this range was computed using an activity coefficient for FeO of 1.3 [O'Neill and Eggins, 2002] and an activity coefficient for Fe0 of 0.85 (calculated using the activity coefficient calculator described in Wade et al. ). Importantly, this is a generous upper limit since some of the Fe on the present day surface may have been emplaced as Fe metal, emplaced as Fe sulfide, and/or delivered to the surface via impactors. Furthermore, much of the observed surface may represent primary crust with Fe enriched relative to the bulk mantle. Based on all the criteria for estimating the fO2 of the Mercurian interior, it can be defined by the range −6.3 to −2.6ΔIW.
3. Is Mercury a Volatile-Rich Planet?
 As discussed above, the elevated K/Th and K/U ratios measured by GRS for the Mercurian surface implies that Mercury is volatile-rich compared to some of the other terrestrial planets [Peplowski et al., 2011]. However, Mercury is the most reduced of the terrestrial planets at −6.3 to −2.6ΔIW, and elements that typically behave as lithophile elements (i.e., K, Th, and U) may behave differently in an oxygen-starved magmatic system like Mercury. If these elements deviate from purely lithophile behavior, conclusions about Mercury's volatile inventory based on LIL element ratios may be inherently flawed.
3.1. Metal/Silicate Partitioning of K, Th and U
 A number of experimental studies investigated metal/silicate partitioning of potassium at elevated temperatures and pressures. Studies were conducted over a range of temperatures (1650–2200°C), pressures (1–15 GPa), fO2's (−4.6 to −1.7ΔIW), and bulk compositions relevant for planetary cores (i.e., <20 wt.% S in the metal) [Bouhifd et al., 2007; Chabot and Drake, 1999; Corgne et al., 2007; Mills et al., 2007]. These studies show that the S content of the metallic phase had the largest effect on the metal/silicate partition coefficient for K, which typically ranged from 0.0003 to 0.009. This range in DKmetal/silicate would result in very little K in the Mercurian core, even with the reducing conditions of Mercurian core formation (Figure 3). In contrast, the DKmetal/silicatefor very S-rich metallic liquids (i.e., FeS) are at least an order of magnitude greater than DKmetal/silicatefor S-poor metallic liquids (i.e., <20% S,Figure 3).
 There are presently very limited amounts of experimental data available for metal/silicate partitioning of U and Th at conditions appropriate for Mercurian core formation. In fact, experimental data are currently only available for U [Malavergne et al., 2007b]. This data indicates that U becomes more chalcophile and/or siderophile under highly reducing conditions consistent with −6 to −3ΔIW at high pressure (5–20 GPa). These conditions are consistent with those established here for the Mercurian interior. Consequently, at the lower range of oxygen fugacities estimated for Mercury (−6.3ΔIW), a significant portion of U could have partitioned into the core, leaving the bulk silicate depleted in U (Figure 3). Although data are not currently available for Th, there is empirical evidence from enstatite chondrites to suggest Th also behaves as a chalcophile element at low oxygen fugacities [Malavergne et al., 2010]. This is further supported by elevated Th and U abundances in the mineral oldhamite (CaS), which is present in highly reduced enstatite chondrite and enstatite achondrite (aubrite) meteorites [Lodders and Fegley, 1993].
3.2. Consequence of Metal/Silicate Partitioning on Mercury's K/Th and K/U Ratios
 If we assume a homogenous Fe-rich core on Mercury, available experimental metal/silicate partitioning data indicate that nearly all K would have partitioned into the silicate portion of Mercury (Figure 3). During this same process, U and possibly Th could have been distributed among the metal and silicate portions of Mercury (as originally suggested by Malavergne et al. ), fractionating them from K. In fact, if Mercury's core formed at the lower range of oxygen fugacity presented here (−6.3ΔIW), ≥10% of the U in Mercury could have partitioned into the core [Malavergne et al., 2007b] (Figure 3). This would imply that Mercury is more volatile depleted than the present-day surface K/U and K/Th ratios imply. Alternatively, recent results from MESSENGER gravity data indicate the presence of an FeS layer at the interface between the core and mantle on Mercury [Smith et al., 2012]. The existence of an FeS layer would allow for substantial K, U, and perhaps Th to have been removed from the silicate portion of Mercury and stored in the FeS layer (Figure 3). Consequently, this FeS layer would have stored a substantial amount of heat producing elements, and could have been a significant heat source for secondary magmatism on Mercury at the core-mantle boundary. This new finding, along with the accumulating evidence for highly reducing conditions on Mercury make a strong case that the use of LIL element ratios are not currently a reliable indicator of the volatile-element inventory on Mercury. Furthermore, additional experimental constraints on the partitioning of K, U, and especially Th under Mercurian conditions are needed before the volatile inventory of Mercury can be fully addressed.
 We would like to thank Mikhail Zolotov and Stuart Ross Taylor for insightful and constructive reviews of the manuscript, and we thank Michael Wysession for the editorial handling of the manuscript. We would also like to thank the MESSENGER science team for all of their hard work in conducting the mission and processing the returned data. F.M.M. would like to acknowledge support from the NASA Cosmochemistry program (NNX11AG76G) during this project. M.A.R. acknowledges financial support from the NASA Planetary Mission and Data Analysis program (NNH10ZDA001N) during this project. Additionally, K.E.V.K. gratefully acknowledges support from a graduate fellowship from the New Mexico Space Grant Consortium.
 The Editor thanks Stuart Ross Taylor and Mikhail Zolotov for their assistance in evaluating this paper.