Earth and Planetary Science Letters The vanadium isotopic composition of lunar basalts

We present the ﬁrst high-precision vanadium (V) isotope data for lunar basalts. Terrestrial magmatic rock measurements can display signiﬁcant V isotopic fractionation (particularly during (Fe,Ti)oxide crystallisation), but the Earth displays heavy V (i.e. higher 51 V/ 50 V) isotopic compositions compared to meteorites. This has been attributed to early irradiation of meteorite components or nucleosynthetic heterogeneity. The Moon is isotopically-indistinguishable from the silicate Earth for many refractory elements and is expected to be similar in its V isotopic composition. Vanadium isotope ratios and trace element concentrations were measured for 19 lunar basalt samples. Isotopic compositions are more variable ( ∼ 2.5 (cid:2) ) than has been found thus far for terrestrial igneous rocks and extend to lighter values. Magmatic processes do not appear to control the V isotopic composition, despite the large range in oxide proportions in the suite. Instead, the V isotopic compositions of the lunar samples are lighter with increasing exposure age ( t e ). Modelling nuclear cross-sections for V production and burnout demonstrates that cosmogenic production may affect V isotope ratios via a number of channels but strong correlations between V isotope ratios and t e ∗ [Fe]/[V] implicate Fe as the primary target element of importance. Similar correlations are found in the latest data for chondrites, providing evidence that most V isotope variation in chondrites is due to recent cosmogenic production via Fe spallation. Contrary to previous suggestions, there is no evidence for resolvable differences between the primary V isotopic compositions of the Earth, Moon, chondrites and Mars.


Lunar origin
The most widely-accepted theory of lunar origin is the socalled "giant impact" model whereby a smaller planetary object struck the proto-Earth in its final stages of accretion leading to an impact-generated disk from which the Moon formed (Cameron and Ward, 1976;Hartmann and Davis, 1975). The model satisfies constraints on the geodynamical properties, planetary densities, and broad chemical features of the Earth-Moon system. In early numerical simulations of the giant impact most of the Moonforming material derives from the impactor, with only a small contribution from Earth (e.g. Canup, 2004). Extra-terrestrial bodies typically have distinct isotopic compositions (Goderis et al., and the Moon. These include high-energy impacts (Canup, 2012;Cuk and Stewart, 2012), a hit-and-run model (Reufer et al., 2012), and a multiple-impact scenario (Rufu et al., 2017). Compared to the canonical model, each of these scenarios invokes restricted dynamic conditions prior, or subsequent, to impact.
Models of an impactor with Earth-like isotopic composition (Dauphas et al., 2014), and models of inner Solar System homogeneity have also been proposed (Mastrobuono-Battisti et al., 2015). These studies argue against the assumption of an isotopically-distinct impactor. However, a consensus on the best model of lunar formation has yet to be reached.

Chemical and isotopic variations in lunar basalts
Lunar basalts are primarily divided by TiO 2 contents into verylow-Ti basalts (VLT) (<1.5 wt% TiO 2 ), low-Ti basalts (1.5-6 wt% TiO 2 ) and high-Ti basalts (>6 wt% TiO 2 ). Chemically-distinct regions of the lunar mantle must exist to explain the compositional diversity of mare basalts (Papike et al., 1976), with high-Ti basalts requiring significant amounts of ilmenite in their sources. Lunar mantle heterogeneities arise from the differentiation of the lunar magma ocean (LMO), a global melting of the lunar mantle as a result of accretion, during which differentially-dense regions of the mantle inefficiently mixed as crystallisation proceeded (Elkins-Tanton et al., 2011). Stable isotope studies of lunar basalts have not only provided evidence of the isotopic kinship between the Moon and the Earth, but also the extent of lunar mantle heterogeneity, the provenance of lunar sample subsets and the nature of lunar differentiation processes. For example, high-Ti basalts show light Mg (Sedaghatpour et al., 2013), and heavy Fe and Ti isotope ratios (Liu et al., 2010;Wiesli et al., 2003). These isotopic differences are thought to relate to crystallisation of ilmenite during LMO differentiation, or partial melting during formation of high-Ti and low-Ti parental melts. In contrast, Cr isotope ratios show mass-dependent fractionations correlating with indices of magmatic differentiation, but no difference between high-Ti and low-Ti basalts (Bonnand et al., 2016). Fractionations in Cr isotopes result from crystallisation of Cr-spinel and Cr-bearing ülvospinel. Post-crystallisation processes can modify isotopic compositions further (in a mass-independent fashion), and lead to difficulty in identifying the original isotopic signatures of the samples. For example, small excesses in 53 Cr and 54 Cr (Mougel et al., 2018) and depletions in ε 50 Ti (Zhang et al., 2012) have been found in mare basalts, reflecting cosmogenic processes at the lunar surface.

Vanadium
Vanadium occurs in multiple oxidation states (V 2+ , V 3+ , V 4+ , V 5+ ) and behaves as a moderately incompatible element in terrestrial igneous systems. Previous studies have taken advantage of the strongly redox-dependent behaviour of V to investigate oxidation states of mantle and melts throughout Earth's history (e.g. Canil, 1997;Lee et al., 2003;Lee et al., 2005Lee et al., , 2004Mallmann and O'Neill, 2009;Wood et al., 2008). Vanadium is depleted in the Earth's mantle compared to CI chondrites (McDonough and Sun, 1995), and its abundance relative to siderophile elements has been used to argue for changes in mantle oxidation during Earth's accretion and core formation (Wood et al., 2008). The lunar and terrestrial upper mantles are estimated to have similar V contents (79 ppm and 81 ppm, respectively; Seifert and Ringwood, 1988). Vanadium is incompatible in plagioclase, and so is unaffected by anorthite crystallisation during LMO differentiation. Differing oxygen fugacity ( f O 2 ) of planetary mantles can affect the partitioning behaviour of V. For example, at low lunar f O 2 , 1 to 2 log units below the Iron-Wüstite buffer (IW-1 to IW-2) most V is present in lunar magmas as V 3+ . Vanadium is more compatible in lunar olivine, pyroxene, and spinel than for terrestrial analogues, where V 4+ is the major species (Papike et al., 2005). The multiple valence states of V over a wide range of f O 2 makes it attractive as a potential tracer of planetary magmatic processes.

Vanadium isotopes
Vanadium has two stable isotopes: 50 V (0.25%) and 51 V (99.75%) and has been of cosmochemical interest for decades, due to the hypothesis that 50 V could be generated by cosmic ray irradiation processes (Stauffer and Honda, 1961). Advances in high precision multiple collector inductively-coupled-plasma mass spectrometry (MC-ICPMS) and the development of more effective separation procedures (e.g. Nielsen et al., 2011;Wu et al., 2016) now allow V isotope determination to precisions useful for geological applications (∼0.2h). Vanadium isotopes are reported in per mil (h) deviation relative to an Alfa Aesar (AA) V solution standard introduced by Nielsen et al. (2011): Terrestrial magmatic samples show a range of ∼2.3h in δ 51 V.
This variability has been interpreted in terms of equilibrium fractionation Prytulak et al., 2013;Prytulak et al., 2017;Wu et al., 2018, Fig. 1). The bulk silicate Earth (BSE) is estimated at δ 51 V AA = −0.7 ± 0.2h based on analyses of bulk peridotites (Prytulak et al., 2013), although more recent work suggests a lighter BSE (δ 51 V = −0.91 ± 0.05, 2SD; . Metal-silicate experiments suggest terrestrial core formation does not significantly fractionate V isotopes (Nielsen et al., 2014). Much of the variation in magmatic samples has been proposed to result from crystallisation of (Fe, Ti) oxides during magmatic differentiation (Prytulak et al., 2017;Sossi et al., 2018;Wu et al., 2018). The similarity in whole rock δ 51 V variation with respect to indicators of magmatic evolution in two cogenetic lava suites with disparate redox conditions (Hekla, Iceland and Anatahan, Mariana arc) suggests crystallisation of (Fe, Ti) oxides and the coordination difference between melt and mineral are dominant in driving isotope fractionation rather than δ 51 V variation directly reflecting changes in oxidation state (Prytulak et al., 2017). Recent experimental work directly determining V magnetite-melt isotope fractionation factors ( 51 V magnetite-melt ) at variable, controlled redox conditions yields 51 V magnetite-melt consistent with those needed to explain the variation seen in ∼ closed system crystallisation and resulting Rayleigh isotope fractionation (Sossi et al., 2018). Interestingly, Sossi et al. (2018) also observed a strong positive relationship between the magnitude of 51 V magnetite-melt and f O 2 . Initial analyses of chondrites, achondrites and martian meteorites returned a restricted range of δ 51 V compositions (δ 51 V AA = −1.7 ± 0.2) uniformly-lighter than BSE (Nielsen et al., 2014). This was suggested to be due to incorporation of irradiated components in meteorites. However, more recent analyses of chondrites with  (Davis et al., 2018). The BSE estimate is indicated by the blue bar (Prytulak et al., 2013). Terrestrial basalts from multiple settings shows a much more restricted range than lunar basalts (Prytulak et al., , 2013(Prytulak et al., , 2017Wu et al., 2018). The measured V isotope range (∼2.3h) of terrestrial magmatic rocks is displayed by the grey dashed line, with the heaviest V isotope samples thought to result from oxide fractionation (Prytulak et al., 2017). been attributed to 50 V production during intense early solar cosmic ray (SCR) irradiation, in agreement with mathematical predications for CAI formation close to an early, active Sun (Gounelle et al., 2001;Lee et al., 1998). This suggests cosmogenic reactions are viable mechanisms for production of light δ 51 V. One body for which δ 51 V data is lacking is the Moon. As a refractory element, V is expected to be isotopically-similar in the Moon and the Earth. Data from the Moon may also provide constraints on the origin of the difference in δ 51 V observed between chondrites and the Earth.

Sampling approach
Nineteen unbrecciated lunar basalts from Apollo missions 11, 12, 14, 15, and 17 were selected based on geochemical and mineralogical variability (Table 1). Samples were processed in sufficient quantities for both V isotope and trace element measurements. The samples include both low-Ti and high-Ti basalts, with reported TiO 2 contents ranging from 2.2 wt% (15535) to 12.9 wt% (70215) ( Table S7) and have reported modal ilmenite contents from 0.5% (12018) to 17% (70215 , Table S5). An Apollo 11 Type A (high-K) basalt (10017), enriched in K, REEs and other incompatible trace elements (Tables S7, S9, Neal and Taylor, 1992), and a high-Al/low-K Apollo 14 basalt (14053) were included. Samples were extracted from the interior of their parents, with three exceptions. The exteriors of the parent samples 10017 and 12054 were sampled and a fragment of a glass-splash coating from sample 12054 was also analysed. The glass was emplaced after a nearby impact or an eruptive event and has an Apollo 12 ilmenite basalt composition (similar to its host rock, Hartung et al., 1978). Interior aliquots of samples were obtained from a maximum depth of 2.5 cm within their parent samples (except 12063-343 which was located between 0.5-5 cm).

Sample preparation
Details of procedures for sample preparation, dissolution, ionexchange chromatography and elemental and isotopic analyses can be found in the Supplementary Text. Briefly, sample chips were washed in ∼18.2 M ·cm ultra-pure water (MQ H 2 O), centrifuged and dried at low temperature. Samples were crushed to homogenise any intra-sample geochemical variation using an agate pestle and mortar. Separate fractions of these powders were weighed for trace element determinations (∼10 mg) and vanadium isotope analyses (20-100 mg). Samples were digested in a multistage hotplate dissolution procedure.

Trace element analysis by ICPMS
Trace element analyses were performed using a PerkinElmer ® NexION ® 350D quadrupole ICPMS (at the Department of Earth Sciences, University of Oxford) on ×10000 dilutions of digested sample solutions. A Re-Rh-In internal standard addition (for monitoring instrumental drift) was performed using an in-line ESI prep-FAST auto-dilution system. The relative standard deviation was <5%, except for Li (7.9%). United States Geological Survey (USGS) reference material, BHVO-2, was measured alongside the lunar sample suite to assess accuracy.

Vanadium isotope analysis by MC-ICPMS
Chemical isolation of V was performed using methods outlined in Wu et al. (2016). Sample aliquots containing 5-10 μg of V  Table S7). a, b Gd isotopes and exposure ages compiled from published values (Supplementary Text, Tables S3, S6). Exposure ages for 12054 (independently-calculated) and 71546 (estimated from other samples in the same rake) are less reliable than for the other samples in the suite (Section 4.5, Supplementary Text, Table S6).
were passed twice through cation-exchange columns (using Bio-Rad ® AG50W-X12 200-400 mesh resin) to remove the majority of matrix elements (Fe, Ti, Al, Ca). An anion-exchange column (using Bio-Rad ® AG1-X8 200-400 mesh resin) was then used to separate V from Cr, and repeated in scaled-down form to remove any remaining Cr.
Quantitative yield of V must be obtained to exclude possible isotopic fractionation during incomplete elution in ion-exchange chromatography. The yield was found to be 100 ± 15% at the end of the procedure (by comparing MC-ICPMS beam intensities to Quad ICP-MS measurements). Cuts were collected before and after the vanadium elution steps in the cation-exchange columns to ensure the calibration of V elution (Wu et al., 2016). Total V blanks were <0.3 ng, which is insignificant compared to the amount of V processed. Samples returned 49 Ti/ 51 V and 53 Cr/ 51 V ratios <0.00001 , confirming negligible contributions of 50 Cr and 50 Ti to the 50 V signal.
Vanadium isotopes were analysed on a Plasma II (Nu Instruments) MC-ICPMS in the MAGIC laboratories at the Department of Earth Sciences and Engineering, Imperial College London. Measurements were performed at solution concentrations of 600-800 ng g −1 . Samples were aspirated into a Nu Instruments DSN-100 to desolvate before injection into the plasma. Masses 49, 50, 51, 52 and 53 were monitored to measure V isotopes and to correct for Ti and Cr interferences on mass 50 (Table S1).
Pseudo-high-resolution mode (M/ M > 5000) was used to resolve polyatomic interferences (particularly 40 Ar 12 C + from 52 Cr + ) from the isotopes of interest. Faraday cups were equipped with 10 11 resistors, except for cups with low-intensity signals (m/z: 49, 50, 53), which were equipped with 10 12 resistors. This amplifies the resultant signal by a factor of 10 and improves signal/noise ratios. Each sample analysis was bracketed by two measurements of an Alfa Aesar (AA) V solution, and a well-characterised secondary standard from BDH chemicals (BDH, Nielsen et al., 2011). The AA standard defines zero (Section 1.4). BDH has been measured (n = 1329) to have a composition of δ 51 V AA = −1.17 ± 0.17h (Prytulak et al., 2017). Measurements of BDH in this study gave an average δ 51 V AA = −1.21 ± 0.22h (n = 121). United States Geological Survey (USGS) reference materials BCR-2, BHVO-2, and BIR-1a were processed through the same ion-exchange procedures and δ 51 V values were found to be in agreement with literature values (Table S2).

Major and trace elements
Comparison of ICPMS elemental concentrations to previous analyses of lunar basalts shows that samples lie within expected ranges for their mare basalt subtype (Table 1, Figs. 3c, S2, S3). Sample sizes in this study are small (up to 125 mg) compared to the majority of literature chemical analyses (Table S7), which were typically performed on much larger aliquots (>1 g) and would be expected to be more representative of the geochemistry of parent samples (e.g. Ryder and Schuraytz, 2001). However, obtaining major and trace element concentrations and V isotope ratios on the same aliquots allows internally-consistent interpretation of isotopic variations. Major element concentrations of samples lie within 5% of published values, with minor exceptions. Samples 12018 (−9%), 12054-150 (+6.3%), and 70255 (−9.8%) show TiO 2 contents that differ by more than 5% from published values. Sample 12054-146 differs in K 2 O by +8%, and sample 70255 differs in MnO by −6.3%.
Chondrite-normalised REE patterns of the lunar samples match those of previous analyses and show the characteristic patterns expected for their sub-groups (Fig. S2). A detailed discussion of the trace element behaviour of the sample suite is beyond the scope of this study, however, elemental concentrations from this study and previously-reported values can be found in Tables S4, S7 &

Magmatic effects on V and δ 51 V
Under terrestrial magmatic conditions V behaves relatively incompatibly through magmatic evolution except during crystallisation of (Fe, Ti) oxides into which V 3+ partitions readily (O'Neill and Navrotsky, 1984). On Earth, the most extreme δ 51 V arises from ∼ closed system crystallisation of isotopically light (Fe, Ti) oxides, which drives residual melts to heavier compositions (up to δ 51 V AA ∼+1h; Prytulak et al., 2017). Given the high modal proportions of (Fe, Ti) oxides (particularly ilmenite) in lunar basalts, it is tempting to assume that the large variability recorded in the lunar sample suite is a result of ilmenite-melt fractionation. However, oxygen fugacity is much lower on the Moon (IW-1 to IW-2) than in terrestrial melts, which not only affects the speciation and partitioning behaviour of V (V 3+ > V 4+ , on Earth V 4+ > V 3+ ), but also the crystallising mineral phases (e.g. from lack of available Fe 3+ ). Analyses of minerals in lunar basalts show V present in spinel (<8000 ppm, Papike et al., 2005), armalcolite (<3000 ppm) and ilmenite (<2500 ppm, El Goresy et al., 1974, pyroxene (<500 ppm), and olivine (<80 ppm - Papike et al., 2005). Likewise, partition coefficients for V at low f O 2 , show V to be compatible in spinel, armalcolite, ilmenite and pyroxene (in decreasing order of compatibility, Karner et al., 2006;Klemme et al., 2006;Papike et al., 2005). Under lunar conditions, V appears more compatible than in terrestrial settings. The abun-dances of minerals in samples is highly-variable (Table S5) and crystal fractionation may control the observed range of δ 51 V.
Although V varies within the suite with varying major element compositions (Fig. 3c, Fig. S3) and mineralogy (e.g. ilmenite, pyroxene Fig. S4) there is no systematic relationship between δ 51 V and TiO 2 or V (Fig. 3a, b) or between δ 51 V and modal proportions of the major hosts for V (spinel, ilmenite, pyroxene, Fig. S4). Likewise, samples previously studied for mass-dependent isotopes (e.g. Cr, Fe, Fig. S5) do not show correlations with δ 51 V (Bonnand et al., 2016;Liu et al., 2010;Wiesli et al., 2003). Apollo 11 Type-A basalt 10017 is >1h lighter than the other basalts (Table 1, Fig. 2). Although the origin of the Apollo 11 Type-A basalts remains contested, the high-abundances of K, REEs, and other incompatible trace elements suggest involvement of a KREEP-like component (Neal and Taylor, 1992). The light δ 51 V in this sample may reflect the extreme composition of this KREEP-y reservoir. However, KREEP is thought to represent the last vestiges of LMO prior to crystallisation. Prior removal of oxides from the LMO should leave this reservoir enriched in heavy δ 51 V (i.e. opposite to the direction of δ 51 V offset observed, assuming similar mineral-melt fractionations to terrestrial settings). The lack of evidence for systematic magmatic-driven δ 51 V variability in this suite is further highlighted by the 0.84h isotopic difference between samples 15535 and 15556 (Fig. 2). These samples are genetically-related by olivine fractionation (Mason et al., 1972), which is unlikely to account for the large δ 51 V difference observed.

Reduced magmatic δ 51 V fractionation at lower f O 2
Considering the magnitude of isotopic fractionation associated with (Fe, Ti)oxide crystallisation on Earth (e.g. 51 V magnetite-melt = −0.4 to −0.5h in Hekla and Anatahan lavas, Prytulak et al., 2017), the lack of correlation of δ 51 V with TiO 2 or ilmenite or oxide content is notable, particularly since the range in abundances of (Fe, Ti) oxides in lunar basalts is much larger than in terrestrial samples. Wu et al. (2018) recently proposed a smaller terrestrial 51 V mineral-melt of −0.15h. However, this factor was derived from a MORB-suite that is not cogenetic and has documented evidence of magma-mixing (indeed, their isotope-mixing calculations fit the δ 51 V variation observed just as well as their reduced mineral-melt V isotope fractionation factor). A recent experimental study of V isotope fractionation between magnetite and hydrous granitic melt at 800 • C and controlled f O 2 conditions (from FMQ-1 to FMQ+5) suggests 51 V magnetite-melt reduces with decreasing f O 2 (Sossi et al., 2018). Extrapolating these results to lunar f O 2 and temperatures (1/T 2 dependence; Schauble, 2004) yields a 51 V oxide-melt ∼−0.2h (assuming lunar oxides fractionate δ 51 V similarly to magnetite, which is not present at lunar f O 2 ). For lunar basalts, a reduced fractionation factor at lower f O 2 may explain a lack of variation between δ 51 V and other geochemical parameters but cannot explain the wide range and light δ 51 V of the suite.

Cosmogenic irradiation
Exposure ages have been previously-determined for most of the samples in this suite (Table 1, Table S6). To allow the most consistent cross-sample comparisons, exposure ages calculated using 81 Kr-Kr and 38 Ar (where 81 Kr-Kr data are absent) were chosen over other techniques due to their widespread adoption and intercalibration (e.g. Hohenberg et al., 1978; see Supplementary Text, Table S6). The range of sample δ 51 V is lighter and more variable at longer exposure ages (t e , Fig. 4), implicating modification of primitive isotopic compositions by exposure-related processes.
Samples exposed at the lunar surface may be affected by meteorite sputtering, surficial contamination, solar wind implantation, or irradiation from high-energy particles derived from solar Samples analysed in this study lie in the regions expected for their mare basalt subtype (c). A strong correlation exists between V and TiO 2 in the high-Ti basalts. Larger, solid symbols represent data from the samples in this study. Semi-transparent symbols are a compilation of literature-reported values for V and TiO 2 in lunar basalts (Table S8).
There is no similar correlation between TiO 2 and δ 51 V (a) or between V and δ 51 V (b). Sample 10017 (two Apollo 11 high-K basalts) has δ 51 V values which lie outside of the range of the other mare basalts.  Table S6. Symbols as in Fig. 2. and galactic cosmic rays (SCRs and GCRs). If exposure-related isotopic variability is achieved after crystallisation (as indicated by Fig. 4), sputtering, surficial contamination and solar wind implantation processes are unlikely to explain the observed δ 51 V variability. All but three samples are picked from the interiors of their parents and would be shielded from these weakly-penetrating processes. Thus, cosmogenic irradiation from SCRs and GCRs is most likely. Cosmogenic effects have been observed in lunar samples for neighbouring Ti and Cr isotope systems, with negative anomalies in ε 50 Ti (Zhang et al., 2012) and positive anomalies in ε 54 Cr and ε 53 Cr (Mougel et al., 2018) attributed to neutron-capture processes. It is possible that similar processes operate on V isotopes.
The flux of energetic particles in SCRs and GCRs consists mainly of protons and alpha-particles (with minor fluxes of heavier nuclei). These primary particles can react with sample material to produce a cascade of secondary energetic particles (e.g. protons, neutrons, alpha-particles, 3 He, deuterium, muons, kaons, and pions) which can then further interact with sample material. Secondary neutrons are of particular interest for a number of elements, particularly those having isotopes with large neutroncapture cross-sections (e.g. gadolinium, Gd, where neutron-capture cross-sections for 157 Gd 156 Gd; Eugster et al., 1970). The 158 Gd/ 157 Gd ratio is sensitive to the flux of thermal neutrons generated during cosmic-ray reactions. Sample δ 51 V shows a negative correlation with 158 Gd/ 157 Gd for five samples with published Gd isotope data (Fig. 5, Table S3), although the strength of this relationship may be enhanced by the position of sample 10017, which lies far outside the range of δ 51 V and 158 Gd/ 157 Gd for the other samples. This correlation operates in a similar sense to that between δ 51 V and exposure age (i.e. the most-exposed samples have the lightest δ 51 V). However, three different samples (15016, 15556, 70215) have been measured for ε 50 Ti (another neutron-capture-sensitive proxy, Zhang et al., 2012) and show similar trends (Fig. S6).  (Prytulak et al., 2013) and the terrestrial Gd isotope value from Dubois et al. (1992). A linear regression is calculated on the lunar samples. References for the Gd isotope data can be found in Table S3. Symbols as in Fig. 2.

Cross-sections
Earlier modelling of cosmogenic V isotope effects focused on production of 50 V from solar proton and He-nuclei interactions with CAI-precursor target materials (Gounelle et al., 2001;Lee et al., 1998). However, only a select handful of reaction pathways and target elements (Ti, Cr, and 51 V) were considered. Under lunar conditions, other target elements (e.g. Fe, Mn) may be important. Secondary particle fluxes (e.g. neutrons, Fig. 5, Fig. S6) are clearly present in these samples and should be considered. Cosmogenic reactions affecting 51 V must also be considered to ascertain the resultant effects on δ 51 V.
Cross-sections for reactions involving the production or burnout of both 50 V and 51 V via proton-, alpha particle-and neutroninteractions with lunar sample compositions were calculated using TALYS-1.9 (Koning et al., 2008). This is the most recent iteration of the TALYS nuclear-modelling code and one of the most accurate codes for energies <200 MeV, being regularly evaluated and updated against experimental data. Although the energies of incident SCR and GCR particles regularly exceed this upper energy limit, this energy range should provide insights into both spallation-and capture-processes, and the relative importance of different target elements.
Cross-sections were calculated for nuclear reactions between target elements Ti, V, Cr, Mn and Fe (assumed to be present in natural isotope proportions) and incident neutrons (at energies 10 −11 MeV-200 MeV), protons, and alpha-particles (energies 0.1 . To allow comparison between the reactions (n > 1000), cross-section data was sub-divided into three categories: burnout of V isotopes (where target V isotope, j, forms another resulting element, R), internal production (target V isotope, i, produces the other V isotope, j -i.e. 50 V → 51 V or vice versa), and external production (where nat Ti, nat Cr, nat Mn, or nat Fe produces V isotope, j). All other factors being constant, burnout and internal production are controlled by the abundance of target V, and external production is controlled by the abundance of the non-V target element. The sums of the cross-sections, j σ X,k , for burnout, internal and external production were calculated for each target, X , at different energies of incident particle, k (where k = proton, neutron, or alpha-particle; Fig. S10).
Reactions producing radioactive isotopes 51 Sc, 51 Ti, 51 Cr, 51 Mn, 51 Fe included in the sum of cross-sections for 51 V production (these isotopes are assumed to decay to before further interaction with energetic particles). Cross-sections for reaction pathways where the product nucleus is the same as the target nucleus (e.g. 50 V(n,n) 50 V, or 51 V(p,n) 51 Cr which decays to 51 V) were excluded.
The compiled cross-sections for each target ( j σ X, k ) were adjusted to reflect atomic abundances of target elements relative to V in samples via: (1) where [X]/[V] is the ratio of target isotope or element relative to V. Total concentration-adjusted cross-sections ( j σ X,k,adjusted ) of different particle pathways are compared for their effects on 50 V and 51 V (Fig. 6, where the "average" lunar sample composition is used).
Internal production displays lower cross-sections than external production in all scenarios, indicating that non-V target elements likely dominate cosmogenic production. Above ∼50 MeV, Fe displays the largest cross-sections for production of both V isotopes in proton and alpha-particle reactions (Fig. 6b, c, e, f). Between 5-50 MeV, cross-sections for production of 50 V from protons with nat Ti (Fig. 6b) is higher than for production of 51 V (Fig. 6e). Burnout of either V isotope becomes increasingly-important at lower proton and alpha-particle energies (<5 MeV). At neutron energies >1 MeV, Cr and Fe appear to be most important for production of 51 V and 50 V (Fig. 6a, d). At neutron energies >50 MeV, Fe is the most important target element for production of either V isotope. Below 1 MeV, there are numerous pathways towards production of 51 V (via neutron interaction with Ti, Cr, 50 V), but no pathways for 50 V production. This has important implications: if thermal neutron capture is most significant amongst the array of possible reaction pathways, it is impossible to reconcile with the direction of cosmogenic isotope shift. Thus, charged-particle reactions or high-energy neutrons must contribute a significant amount to the total cosmogenic V production. In general, production cross-sections for 51 V appear larger than for 50 V (Fig. 7a, b, c) -however, a direct comparison must be carried out to ascertain the resulting effect on sample δ 51 V.

Contribution to sample δ 51 V
The total 51 V/ 50 V in irradiated samples can be displayed as follows: where n( j V) 0 is the initial number of atoms of isotope, j, and n( j V) cosm is the number of atoms of isotope, j (i.e. 51 V, 50 V), produced through cosmogenic reactions: The resulting isotopic effect is: Assuming a natural isotopic ratio ( n ( 51 V 0 ) n ( 50 V 0 ) = 400) substituting for the terms in Equation (2) gives: . 6. The sum of adjusted cross-sections ( j σ X,k , y-axes) for reactions between an incident particle, k, (k = neutron, proton or alpha particle) and nuclei (X) to either produce or burn out j V (where j is the V-isotope of interest -i.e. 50 V, 51 V) at different particle energies (x-axis). Reactions are broken down into three categories: burnout (light blue) of j V to another isotope, R; internal production (black) of j V from i V ( j = 50, 51; i = 51, 50); external production of j V from target elements nat Ti (gold); nat Cr (green); nat Mn (purple); nat Fe (red). Cross-sections are adjusted to the average composition of the suite via [X]/[V] (where [X] is the atomic abundance of the target element or isotope, [V] is the atomic abundance of V). Targets are assumed to have natural isotopic abundances. Cross-section curves are displayed for each incident particle (neutrons a, d; protons b, e; alpha-particles c, f) and its effect on each isotope of V ( 50 V: a, b, c; 51 V: d, e, f). Within each panel, the curves represent reactions which add to the abundance of isotope j V, except the burnout curves (light blue) which will decrease the abundance of j V. Note the different energy scales for neutrons and protons/alpha-particles.
For ease of comparison between the total adjusted cross-sections a new variable, the "excess total 51 V cross-section" is defined: This is more useful than a direct cross-section ratio ( 51 σ total-adjusted / 50 σ total-adjusted ) since it allows direct inspection of the isotopic effects of particular pathways at different energies and can account for instances where the sum of concentration-adjusted cross-sections for an isotope is ≤0 (e.g. where burnout is more significant than production). For all the particle reactions considered, it appears that only neutron capture (<1 × 10 −3 MeV) can produce excesses in ε 51 σ total (Fig. 7d, g). There are negative excursions in ε 51 σ X for proton and alpha-particle reactions with Ti between 20-100 MeV ( Fig. 7e, f), and smaller negative excursions in ε 51 σ total for Fe reactions with protons >50 MeV. Similarly, high-energy neutrons display negative ε 51 σ X for Fe (>10 MeV, Fig. 7d). The magnitude of the cosmogenic isotope effect is expected to vary with sample compositions (i.e. changing [X]/[V] ratio, Fig. 7g, h, i). High-Ti samples are expected to show a stronger ε 51 σ response to interactions with protons and alpha-particles when compared to average sample or low-Ti sample compositions. However, at lower neutron energies (<10 −3 MeV), high-Ti samples would be expected to display isotopically-heavier cosmogenic signatures when compared to low-Ti samples (Fig. 7g). It appears that for protons and alpha-particles, Fe and Ti are the primary contributors to V isotope anomalies for the energy ranges investigated. Burnout of V and internal production are relatively unimportant in affecting the final isotopic ratio.

Reconciling with chemical compositions
Plotting sample δ 51 V against exposure age over V concentration (t e /[V]) yields a strong linear correlation (Fig. 8a). This correlation is stronger than the relationship between δ 51 V and  , and alpha-particles (c) and targets of average lunar basalt composition at different energies are displayed. Cross-sections for production of 51 V is higher than 50 V, except for alpha-particle reactions between 20-50 MeV. The excess 51 V production cross-section for different target elements (ε 51 σ X , see Section 4.7) varies across the range of incident particle energies and particle compositions (d, e, f). The total excess 51 V production cross-section (ε 51 σ total ) changes in intensity with changing target chemical compositions (g, h, i). For all particles (except neutrons below 10 −5 MeV) there are significant negative ε 51 σ deviations. Thermal neutrons (≤2.5 × 10 −8 MeV) show positive ε 51 σ deviations (g), as only 51 V production is viable the target elements investigated (Fig. 6a, 6d). Average and low-Ti basalts show reduced ε 51 σ total deviations compared to high-Ti basalts reflecting variation in [X]/[V] between samples. At high-energies (>50 MeV) negative deviations in ε 51 σ are seen from the effect of Fe as a target. Key displayed in a), d), g).
important in controlling the resultant δ 51 V after cosmogenic exposure than the concentration of the target element -i.e. the target element concentration doesn't vary much compared to [V]. Both Mn and Fe are relatively restricted in variability compared to V, Cr, and Ti (Table S4, S7). However, the concentration-adjusted cross-sections for Mn are orders of magnitude lower than for Fe (Fig. 6), and Fe is therefore more likely to be the primary target element. Iron is relatively invariable within this suite (RSD ∼8%). A lone high-energy (600 MeV and 21 GeV) experimentalstudy of V-production from proton-induced Fe-spallation (Perron, 1976) indicates that production cross-sections are still significant at high-energies, and the ratio of 51 V/ 50 V produced is low (<1, and this ratio is lower at higher proton energies). Thus, Fe appears to remain an important target for production of light δ 51 V for higher-energies particle fluxes (e.g. as seen in GCRs) than directly calculated here.
The weaker correlation with t e × [Ti]/[V] (despite strong ε 51 σ depletion at certain energy intervals, Fig. 7e, f) could be due to a number of reasons: 1) the real energetic particle flux is stronger for parts of the energy spectrum where cross-sections with Fe are strongest (typically energies >50 MeV), or 2) thermal neutron capture on Ti at low energies is quantitatively-important -the production of 51 V here would counteract production of lower 51 V/ 50 V from Ti reactions and protons/alpha-particles.  (Fig. S7). The fact that most samples fall on the same trendline indicates that pre-irradiation δ 51 V variation is restricted, despite highly-variable magmatic histories. A linear regression (R 2 = 0.95) is calculated to intercept at δ 51 V = −1.05 ± 0.14 (95% confidence intervals In summary, thermal neutrons -although clearly present in sufficient fluxes to impart significant 158 Gd/ 157 Gd excesses (Fig. 5) and ε 50 Ti depletions (Fig. S6) -cannot be the sole cause of δ 51 V alteration due the lack of viable pathways for neutron capture production of 50 V (Fig. 6). Significant fluxes of higher-energy neutrons (>1 MeV, Fig. 7d, f), protons, or alpha-particles (>5 MeV) must have been present in the samples during their periods of exposure at the lunar surface. The most plausible target element is Fe (Fig. 8a, b). Although SCRs have significantly weaker penetrativeability than GCRs, both processes may be important for these samples obtained from shallow depths (Section 2).

Pre-irradiation compositions and implications for the solar system
A linear regression is calculated on the data in Fig. 8b yielding a slope of −2.091 × 10 −6 , and a y-intercept of −1.05 (R 2 = 0.95).
Samples with less-reliable exposure ages (12054, 71546, Supplementary Text, Table S6) or where only literature chemical data exists (12054-13, 70215, 71546) were excluded from this calculation, but their inclusion only slightly affects the resulting regression (slope = −2.092 × 10 −6 , R 2 = 0.95). The coincidence of high-Ti and low-Ti basalts upon the same regression line indicates that, despite different mantle sources and magmatic histories, high-Ti and low-Ti basalts have similar pre-exposure δ 51 V. The two samples with the shortest exposure age (14053-305, 74275-355) have similar δ 51 V values despite their significant geochemical differences.
Exclusion of the unusually-light sample, 10017, (Section 4.3) results in a similar regression with slope −1.989×10 −6 and intercept −1.06. Thus, the δ 51 V range across all samples can largely be attributed to cosmogenic processes post-crystallisation. The intercept of Fig. 8b defines the pre-irradiation average δ 51 V for the suite. Pre-irradiation lunar basalt composition is estimated to be δ 51 V AA = −1.05 ± 0.14h (95% CI). Although fractionation during partial melting cannot be excluded, and so an estimate of the bulk lunar mantle δ 51 V cannot be confidently made, the convergence of high-Ti and low-Ti suites suggests δ 51 V may be relatively uniform in the basalt source regions.
Thus, it appears that recent cosmogenic exposure can explain δ 51 V variations in chondrites in Nielsen et al. (2019). This is not related to differential incorporation of irradiated components (e.g. CAIs), but to irradiation processes operating after the breakup of Early studies of V isotope ratios in geological materials (prior to the development of MC-ICPMS) were hindered by relatively poor analytical precision, but studies of iron meteorites (e.g. Stauffer and Honda, 1961) reported samples with remarkable 51 V/ 50 V ratios (ranging from ∼ 5 to 150). In light of the relationship between δ 51 V and t e * [Fe]/[V] seen in lunar samples and chondrites, and the apparent importance of Fe as a target element, these extreme ratios appear to be in accordance with the chemical compositions of iron meteorites where [V] typically is <1 ppm.  Fig. 8b), strongly suggesting the variation observed in recent chondrite data (Nielsen et al., 2019;Xue et al., 2018) is also due to irradiation of samples. Pre-irradiation values overlap with estimates for the BSE (Prytulak et al., 2013;Qi et al., 2018) and Mars (Davis et al., 2018) suggesting that δ 51 V is relatively uniform between planetary and chondrite parent bodies measured thus far. Symbols as in Fig. 1  The pre-irradiation δ 51 V calculated for the Moon (δ 51 V AA = −1.05 ± 0.14h, 95% CI) is in accordance with the value obtained for chondrites (δ 51 V AA = −1.06 ± 0.06h, 95% CI) and Mars (δ 51 V AA = −1.05 ± 0.20h, 2SD, Davis et al., 2018). These values are similar to the composition of BSE (δ 51 V AA = −0.7 ± 0.2h) from Prytulak et al. (2013) and largely overlap with a more recent estimate from Qi et al. (2018) (δ 51 V AA = −0.91 ± 0.05, 2SD, Fig. 9). In contrast to initial studies (Nielsen et al., 2014) it appears that the δ 51 V of planetary bodies (Earth, Moon, Mars) and chondrites are remarkably similar.

Conclusions
1. Vanadium is more compatible during melting, and has higher partition coefficients in crystallising mineral phases, on the Moon compared to the Earth. Therefore, multiple phases can host V in significant quantities during lunar fractional crystallisation.
2. There is no correlation between sample δ 51 V and mineralogy or chemical composition. High-Ti (excluding 10017) and low-Ti basalts span a similar range in δ 51 V. Lunar basalts are mostly isotopically lighter than terrestrial basalts, but also show a high degree of variability, spanning a range equal to the most extreme products of magmatic differentiation processes on Earth (Prytulak et al., 2017). 3. Terrestrial crystallisation of (Fe, Ti) oxides strongly fractionates V isotopes, but despite large variations in oxide content in the lunar suite, there is little evidence for oxide crys- 4. Vanadium isotopes correlate with exposure age and neutronsensitive Gd isotopes, with longer-exposed samples having lighter δ 51 V (Fig. 4). Calculations of reaction cross-sections protons, alpha-particles and neutrons indicate that Ti and Fe are most important potential targets for cosmogenic alteration of δ 51 V. However, neutron-capture alone cannot drive the isotopic variability observed.
5. The correlation observed between δ 51 V and t e * [Fe]/[V] suggests Fe is the primary target element of importance for cosmogenic V isotope effects, and similar pre-exposure δ 51 V for high-Ti and low-Ti basalts (δ 51 V AA = −1.05 ± 0.14h, 95% CI, Fig. 8b). The similarity in V isotope ratios between lunar basalts and the BSE adds further weight to the growing body of evidence that the Moon is the Earth's isotopic twin. 6. The co-linearity of the lunar and meteorite correlations for V isotopes versus t e * [Fe]/[V] implicates cosmogenic effects as the dominant cause of solar system V isotope variability. There is no longer evidence for a significant difference between the BSE and other reservoirs. 7. The magnitude of cosmogenic isotope effects seen in vanadium isotope ratios is much larger than those seen in Cr and Ti isotopes. Vanadium isotopes therefore have the potential to be a useful proxy for recording exposure histories in extraterrestrial materials.