Introduction

Oxygen fugacity (hereafter referred to as fO2) is a critical physico-chemical parameter that impacts the evolution and potential habitability of terrestrial planets. Initial fO2 conditions of different planetary bodies are primarily controlled by the nature of the accreted materials (e.g., various groups of chondrites, bodies with more or less ice, and differentiated bodies, and so on), which can be further modified by processes such as mantle outgassing during the accretion and differentiation stage, and as well mantle equilibration with the nebular gas. Specifically, the initial fO2 conditions of planetary mantle reservoirs following planet formation influence primary planetary-scale differentiation processes, including the degassing behavior of volatiles such as hydrogen, carbon, nitrogen, and oxygen1,2,3,4,5 and the partitioning behavior of other elements between distinct chemical reservoirs during core-mantle and mantle-crust differentiation6,7. Thus, differences in initial fO2 conditions have a profound impact not only on compositional evolution of a planet’s oceans and atmospheres, but also on its thermal and geodynamical evolution1,2,3,4,6,7.

Existing fO2 estimates of planetary materials rely on petrological experiments and studies of phase equilibrium relations during magma crystallization such as mineral equilibrium assemblages8,9, partitioning of redox-sensitive transition metals between minerals and melts10,11,12, and the valence states of redox-sensitive transition metals in minerals13,14. Based on these approaches, a wide variability of fO2 conditions in Solar System materials spanning ~14 orders of magnitude has been identified14. Despite this, there remain controversies on the fO2 conditions of the mantle reservoirs right after core-mantle differentiation for terrestrial planets, in particular that for Mars8,9,10,15,16,17,18,19. This is because the primordial fO2 conditions in the mantles of large rocky bodies are hard to determine, e.g., these records can be overprinted by later fluid/melt metasomatism in the mantle20 or the fO2 records of mantle-derived magmas can be readily modified by assimilation and degassing processes during magma ascent into the planetary crusts18,21. Thus, a full understanding of the initial fO2 conditions of planetary mantle reservoirs requires a proxy that is insensitive to modification by secondary planetary and/or magmatic processes.

The stable isotope geochemistry of titanium is an emerging tool that can be used to determine the primary fO2 conditions of planetary mantle reservoirs. Titanium has two valence states, namely Ti3+ and Ti4+, and below ΔIW + 0 (where ΔIW represents a difference of fO2 of a system relative to the iron-wüstite buffer in a logarithmic unit), the proportion of Ti3+ in basaltic silicate melts increases with progressively reducing conditions and approaches Ti3+/Ti4+ = 1 at ~ΔIW–6.122. In addition, first-principles calculations reveal that the presence of Ti3+ can lead to Ti stable isotope fractionation of a sub per mil magnitude in 49Ti/47Ti ratio between pyroxene/pyrope and mafic silicate melts, with the Ti3+ hosted by silicate minerals being enriched in the light Ti isotopes23. This is in contrast with Ti4+, which exhibits insignificant isotope fractionation between silicate minerals and mafic silicate melts within an analytical uncertainty of ±0.015‰, as corroborated by studies of Earth’s mantle and mantle-derived rocks24,25,26, In a Ti3+-bearing system, the extracted basaltic silicate melt from partial melting processes preferentially uptakes heavy Ti isotopes, leaving behind a residual mantle reservoir that is isotopically lighter. At the same degree of melt extraction, the magnitude of the light Ti isotope enrichment of the residual mantle may vary with fO2, as prospected for the changing proportion of Ti3+ in the melt. Compared to the fO2 of the melt, the Ti stable isotopic composition of a magma is much less sensitive to modification by assimilation and degassing processes, offering an opportunity to determine via Ti isotopes the initial fO2 conditions of planetary mantle reservoirs.

We recently developed protocols for ultra-high-precision Ti isotope measurements using a next-generation multi-collector inductively-coupled-plasma mass spectrometer (MC-ICP-MS), namely the Neoma MC-ICP-MS. This approach allows for the concomitant determination of the mass-dependent Ti isotope fractionation and nucleosynthetic variations in meteorites via a 47Ti-49Ti double-spike protocol26. The external reproducibility of the δ49Ti (the per mil deviation in the 49Ti/47Ti ratio relative to OL-Ti standard) and ε50Ti (the per ten thousand deviation of the mass-bias-corrected 50Ti/47Ti ratio relative to the OL-Ti standard) values using our method are ±0.010‰ and ±0.15 epsilon, respectively26. This enables the identification of the probable minute Ti isotopic variability produced during magmatic processes in planetary mantle reservoirs, such as the mantle of the ureilite parent body, the Vesta mantle, and the martian mantle. Here, we intend to compare the Ti isotopic composition of meteorites with models of Ti isotope fractionation to quantify fO2 conditions of the planetary mantles.

First, we focus on ureilites (n = 9) and diogenites (n = 8). Ureilites are understood to represent mantle residues of the ureilite parent body (UPB) after melt extraction at fO2 conditions of ~ΔIW–3.3 to ~ΔIW–1.527,28. It is noteworthy that melt extraction from UPB mantle has been dominated by graphite29,30,31, leading to more reducing conditions relative to that of core formation for UPB set by high FeO contents of ureilites (~ΔIW–1) and thus plausibly larger Ti isotope fractionation due to the enhanced Ti3+/Ti4+ ratios in the melts during melt extraction in UPB mantle. Note that graphite metasomatism in ureilites after disruption of the UPB may further result in changes in olivine Mg# and impose at the microscale much more reducing conditions in ureilites31. However, this should have negligible effects on the Ti isotopic composition of ureilites on the whole-rock scale as the removal of silicate melts was supposedly very limited during this process. In contrast, diogenites are predominantly magma cumulates formed on the asteroid Vesta, and have slightly elevated fO2 values of ~ΔIW–2 to ~ΔIW–0 relative to ureilites14,32,33,34. Both these types of achondrites represent products from melt-mineral segregation in the presence of Ti3+ and, as such, are suited to evaluate the potential of Ti isotopes as a tracer for fO2 during melt extraction in planetary mantle reservoirs and during subsequent magma differentiation.

Second, we apply Ti isotopes to constrain the initial fO2 of the martian mantle. Previous studies of SNC meteorites have provided fO2 estimates ranging from ~ΔIW–1 to ~ΔIW + 48,9,10,15,16,17,18. Importantly, the fO2 records of SNC meteorites broadly correlate with their geochemical features (e.g., La/Yb, 87Sr/86Sr, and 143Nd/144Nd ratios), with the depleted samples being relatively reduced and the enriched ones more oxidized8,9,10,17,18. While it has been often inferred that the highly varying fO2 values of SNC meteorites may represent the signals in the mantle sources8,15,16,17,18, there remains a possibility that martian mantle may have a single, reducing fO2 condition and the more oxidizing records are secondary during magma migration in the crust. As such, we have selected for study martian meteorites with various geochemical characteristics, i.e., 17 shergottites (four depleted, three intermediate, and ten enriched), one chassignite, and three nakhlites.

Results

Our new measurements are reported in Fig. 1 together with previously published data for the same meteorite groups. Ureilites have low δ49Ti values that vary from –0.105 ± 0.019‰ to –0.038 ± 0.019‰ compared to the chondrite average of +0.053 ± 0.005‰26, except for EET 87517 that has a δ49Ti value of +0.052 ± 0.004‰ (Fig. 1). We note that EET 87517 is also characterized by a higher ε50Ti value of –1.35 ± 0.03 relative to the other ureilites (–2.17 ± 0.24 to –1.63 ± 0.34) as well as a heavily oxidized interior such as the absence of iron metal and high Fe3+/(Fe2+ + Fe3+) ratio of 0.41935. Diogenites have δ49Ti compositions that mostly cluster around or below the chondrite average, with Bilanga and NWA 4223 having the lowest δ49Ti values of –0.039 ± 0.004‰ and –0.024 ± 0.007‰, respectively (Fig. 1).

Fig. 1: Titanium isotopic results of ureilites, diogenites and martian meteorites.
figure 1

Data of rock standards and meteorite samples from this study are provided in Tables S1, S2 in the Supplementary Materials. The chondrite average from ref. 26 and the literature data of martian meteorites from ref. 36 are shown in gray for comparison. Error bars represent the 2se uncertainties.

Full size image

Our new data for martian meteorites show a range of δ49Ti values (–0.023 ± 0.007‰ to +0.138 ± 0.011‰), consistent within uncertainty with the earlier work36 (Fig. 1). However, our high-precision data allows us to resolve a systematic difference in δ49Ti between martian meteorite groups that correspond to geochemical enrichment, where the δ49Ti values from depleted shergottites, intermediate shergottites, enriched shergottites to nakhlites/chassignite tend to increase and show larger inter-group variability (Fig. 1).

Discussion

Modeling Ti isotope fractionation during melt extraction in planetary mantle

Although there are a number of processes, in particular in magmatic systems on Earth, that can alter the Ti isotope composition of magmas and their residues, these mechanisms are not suited to account for the δ49Ti variation in ureilites, diogenites, and shergottites. This is because on Earth, crystallization and separation of Ti-rich minerals (e.g., magnetite, ilmenite, and rutile) during processes such as fractional crystallization during magma migration and partial melting of crustal precursors are the major drivers of Ti isotope fractionation between magmatic rocks24,25,26,37,38,39,40. The formation of Fe-Ti oxides during magma differentiation, however, requires fO2 conditions that exceed the typical range of fO2 in ureilites, diogenites, and shergottites, and/or very evolved melt composition with MgO <5 wt%. While crystallization of armalcolite near the iron-wüstite buffer can also induce large Ti isotope fractionation41, the saturation of armalcolite requires specific melt composition resembling that of high-Ti lunar mare basalts (i.e., TiO2 contents over 10 wt%) that can be hardly approached in most ultramafic to mafic partial melts from other terrestrial planetary mantles.

Instead, for planetary objects like the parent bodies of ureilites and diogenites, and Mars, the relevant magmatic systems are dominated by ultramafic and mafic lavas, and felsic rocks such as those commonly present in Earth’s continental crust are scarce or absent. Considering the negligible Ti isotopic fractionation of Ti4+ between silicate minerals and mafic silicate melts based on Ti isotope data of Earth’s igneous rocks24,25,26 and by first-principles calculations23, the δ49Ti variations observed for ureilites and diogenites that have melting residue or cumulate origins require the presence of some amount of Ti3+ in the magma systems. The magnitude of Ti3+-induced Ti isotope fractionations during melt extraction from a planetary mantle can be modeled by combining: (i) the experimental calibration of Ti3+ and Ti4+ proportions in basaltic melts against fO222 and (ii) the equilibrium Ti isotopic fractionation factors (i.e., 103lnα) of Ti3+ and Ti4+ in silicate minerals and basaltic melts that can be derived from first-principles calculations and studies of Earth’s igneous rocks23,24,26 (Fig. 2; see demonstration in Methods).

Fig. 2: Modeling Ti isotopic effects from Ti3+ during melt extraction in planetary mantle.
figure 2

A Valence states of Ti in basaltic melts and silicate minerals in equilibrium derived according to ref. 22. B Calculated equilibrium Ti isotope fractionation factors of basaltic melts and silicate minerals at varying fO2 according to ref. 22,23,24,26. C Calculated Ti isotope fractionations in the extracted basaltic melts and melt-extracted residues in planetary mantle at varying fO2 following two fractional melting scenarios (i.e., fTi = 0.20 and fTi = 0.35).

Full size image

It is worth noting that in addition to fO2, the Ti3+/(Ti4++Ti3+) ratio in the silicate melt in reality can be also a function of melt composition22,42,43,44. Therefore, some knowledge of the composition of the melt is needed before translating the Ti3+/(Ti4+ + Ti3+) ratio (and hence Ti isotopic variations) into fO2. In this study, we adopt the calibration of the Ti3+/Ti4+ ratio against fO2, i.e, Ti3+/(Ti4++Ti3+) = 0.411 at ~ΔIW–5.5, for the silicate melts at equilibrium with major mineral phases of planetary mantles (e.g., olivine, orthopyroxene, and clinopyroxene)22, containing SiO2 (47.52–54.79 wt%), TiO2 (1.06–1.69 wt%), Al2O3 (10.03–21.12 wt%), FeO (0.19–0.65 wt%), MgO (15.08–15.73 wt%), and CaO (14.71–21.56 wt%). Notably, the melt products from the previously compiled calibration experiments42 can contain TiO2 contents over 25 wt%, which may have led to a non-ideal dependence of Ti3+/(Ti4+ + Ti3+) on fO244. Promisingly, consistent Ti3+/(Ti4+ + Ti3+)-fO2 relations, i.e., Ti3+/(Ti4+ + Ti3+) = 0.359–0.426 at ~ΔIW–5.5, have been obtained for the silicate melts at anorthite-diopside eutectic (or by addition of forsterite, enstatite, or wollastonite, i.e., SiO2 = 49.15–54.09 wt%, TiO2 = 1.02–1.06 wt%, Al2O3 = 6.25–15.44 wt%, MgO = 4.27–21.66 wt%, CaO = 14.31–37.43 wt%) in ref. 44 (Supplementary Fig. S1). A remaining caveat is that all the above-mentioned experimental melts22,44 contain minor to no iron (FeO ≤0.65 wt%). Nonetheless, the consistency despite the wide chemical variability of the melt products for calibration22,44 underscores the robustness of the used Ti3+/(Ti4+ + Ti3+)-fO2 calibration22 to model melt depletion processes in terrestrial planetary mantles.

Due to the higher partition coefficients of Ti3+ relative to Ti4+ in silicate minerals (e.g., D (Ti3+) = 3.27 × 10–1 ± 2.58 ×  10–2 and D (Ti4+) = 9.54 × 10–2 ± 2.52 × 10–3 for orthopyroxene; D (Ti3+) = 1.10 × 10+0 ± 7.82 × 10–2 and D (Ti4+) = 2.87 × 10–1 ± 4.74 × 10–3 for clinopyroxene)22,43, at equilibrium the Ti3+/(Ti3+ + Ti4+) ratios in silicate minerals are higher than those in basaltic melts (Fig. 2A). Integrating with existing Ti isotope fractionation factors of Ti3+ and Ti4+ in silicate minerals and melts, basaltic melts have higher 103lnα (for the 49Ti/47Ti ratio) values than silicate minerals (Fig. 2B), meaning that extraction of basaltic melts from planetary mantle would preferentially uptake the heavier Ti isotopes and leave behind isotopically lighter melt-extracted residues. Furthermore, equilibrium Ti isotope fractionation factor between basaltic melts and melting residues, i.e., 103lnα (melt-residue), shows an inverted “U”-shaped dependence on fO2, which increases from +0.004‰ at ~ΔIW + 3.5 to the maximum of +0.134‰ at ~ΔIW–5 and declines at fO2 below ~ΔIW–5. As consequence, the Ti isotope fractionations in the extracted basaltic melts from fractional melting processes follow an inverted “U”-shaped dependence on fO2, with the opposite recorded in the melt-extracted residues (Fig. 2C). This is because the maximum extent of isotope fractionation occurs while Ti3+/(Ti3+ + Ti4+) ratio approaches ~0.35 at ~ΔIW–5, where difference of the mean Ti valence between the melt and clino/orthopyroxene reaches maximal (i.e., 3.66 versus 3.36). It is worth noting that the magnitude of Ti isotope fractionation in the residual mantle would be larger at higher degrees of melt extraction, in other words, with lower Ti fraction (fTi) remaining in the melt-extracted residues (Fig. 2C).

Stable Ti isotopic variation as a proxy for initial fO2 conditions in planetary mantle reservoirs

The modeling results can be further compared to the δ49Ti data of ureilites and diogenites for verification. For a chondritic precursor, a partial melting event extracting ~80% Ti into a basaltic melt (i.e., fTi = 0.20, in other words a partial melting degree of ~20% at DTi = 0.15) via basaltic melt can produce isotope fractionations of up to ~0.22‰ in δ49Ti for the residual mantle reservoir at the fO2 conditions relevant to ureilites and diogenites (≤ΔIW + 0; Fig. 2C)14, which can account for the lowest δ49Ti values of ureilites (Fig. 3). With respect to diogenites that have cumulate origins, the same systematics may still hold, given that orthopyroxene and clinopyroxene from a reducing, Ti3+-bearing basaltic melt should be enriched in Ti3+ (and therefore the light Ti isotopes) relative to the melt (Fig. 3). Such Ti isotope fractionation between cumulates and melts may decrease or diminish in the case that the melts have become more oxidized (in other words containing much less or no Ti3+ in the melts) due to degassing or wall-rock assimilation during magma migration. In this scenario, the cumulates would inherit the Ti isotopic composition of the silicate melts, which may be able to account for the elevated δ49Ti values in most of the diogenite samples that plot above the modeled melting residue curve (Fig. 3). In comparison, the chondritic-like δ49Ti value (+0.052 ± 0.004‰) for the anomalous ureilite (i.e., EET 87517) is in line with the highly oxidized nature of this sample, which most likely reflects admixing of oxidized carbonaceous chondrite materials to the ureilite parent body. This is further corroborated by the elevated ε50Ti value of this sample (–1.35 ± 0.03) relative to other ureilites (–2.17 to –1.63), as carbonaceous chondrites are characterized by the high ε50Ti values of +2 to +545. The minimum δ49Ti values defined by ureilites (i.e., –0.105 ± 0.019‰), diogenites (i.e., –0.039 ± 0.004‰), and the ~3.8 Ga Isua metabasalts broadly agree with the curve predicting isotopic effects from Ti3+ during melt extraction in planetary mantle reservoirs (Fig. 3). Remarkably, the mean valences of Ti in the clinopyroxene and orthopyroxene from ureilite Y-791538 are 3.64 ± 0.05 and 3.68 ± 0.05, respectively46. Such redox states of Ti can be translated into Ti3+/(Ti3+ + Ti4+) ratios of 0.32-0.36 that are well consistent with the predicted Ti3+/(Ti3+ + Ti4+) ratios in clinopyroxene and orthopyroxene at ~ΔIW–3.0 to ~ΔIW–2.5 by the model in this study (0.30–0.36), representing independent evidence for Ti3+ in ureilites. This establishes that Ti isotopes can serve as a sensitive tracer for fO2 conditions during planetary mantle-crust differentiation.

Fig. 3: Titanium isotopic effects from Ti3+ as recorded by ureilites and diogenites.
figure 3

The δ49Ti data of the ~3.8 Ga Isua metabasalts are from ref. 26. Modeling results for melt extraction from a chondritic precursor with the presence of Ti3+ are shown for comparison, where the red solid and red dashed curves represent partial melts and melting residues, respectively, after melt extraction at fTi = 0.2 (see Method for the modeling details). The shaded red, blue, and orange boxes cover the ranges of δ49Ti and fO2 for ureilites, diogenites, and Earth’s upper mantle in the early Archean, respectively. Error bars represent the 2se uncertainties.

Full size image

The primary redox state of the martian mantle constrained by Ti isotopes

Abundant chemical and Sr-Nd-Hf isotope systematics of martian meteorites suggest the existence of multiple chemical reservoirs in the martian mantle, which are usually referred to as depleted, intermediate, and enriched mantle sources of Mars8,47,48,49,50,51. The Sm-Nd model ages further imply that these mantle reservoirs have been separated from each other during a mantle differentiation event that can be dated back to as early as 4504 ± 6 Ma52. The δ49Ti values of shergottites are positively correlated with the (La/Yb)N values (where “N” denotes a normalization to the same ratio in CI-type chondrites). Thus, similar to the trend observed for ureilites and diogenites, resolvable Ti isotope fractionations arising from Ti3+ appear to have occurred during the ~4.5 Ga mantle differentiation event on Mars (Fig. 4). Based on the simulation of Ti isotopic effects from Ti3+ (Fig. 2C), the lowest δ49Ti value of –0.023 ± 0.007‰ for depleted shergottite Dhofar 019 corresponds to a fO2 condition of ΔIW–1.0 ± 0.2 at the use of model of Ti isotope fractionation at fTi = 0.2 (Fig. 4). It is worth noting that the estimated fO2 condition would be more reducing (i.e., ΔIW–1.4 ± 0.2) if adopting a larger fTi value of 0.25 for the melt extraction model, in other words lower degree of Ti extraction from the mantle source. Nonetheless, depleted shergottites represent partial melts and their mantle sources are predicted to be characterized by even lower δ49Ti values after taking into account the Ti isotopic effects from magma generation, e.g., Δ49Tipartial melt–source of +0.020‰ under an fO2 condition close to ΔIW–1 at fTi = 0.2 (Fig. 4). This implies more reducing conditions (≤ΔIW–1) during the ~4.5 Ga mantle differentiation event on Mars.

Fig. 4: Titanium isotope fractionation occurring during the ~4.5 Ga mantle differentiation event on Mars.
figure 4

The chondrite average from ref. 26 and the literature data of martian meteorites from ref. 36 are shown in gray for comparison. The (La/Yb)N values of the martian meteorites have been calculated using the literature chemical data in the Martian Meteorite Compendium originally compiled by Charles Meyer and updated and revised by K. Righter in 2017, after a normalization onto the chemical data of chondrites in ref. 75. The red dashed arrows indicate the inferred δ49Ti values of the magma sources for the depleted shergottites after correction of the Ti isotopic effects during magma generation as modeled in Fig. 2C. Error bars represent the 2se uncertainties.

Full size image

After correction of Ti isotopic effects from magma generation with a Δ49Tipartial melt–source value of +0.020‰, mantle sources of the four depleted shergottites in this study (i.e., NWA 4527, SaU 005, DaG 735, and Dhofar 019) have δ49Ti values ranging from –0.043‰ to –0.014‰, which based on the previous simulation defines a fO2 condition of ~ΔIW–1.6 to ~ΔIW–0.8 during the formation of the depleted mantle source of Mars at ~4.5 Ga (Fig. 5). This fO2 range is in line with the core formation conditions (~ΔIW–1.25) of Mars inferred from Ni, Co, Mo, W, and P abundances in the martian mantle53 or that (~ΔIW–1.7) inferred based on the FeO content of the martian mantle54. Importantly, the newly defined fO2 condition of ~ΔIW–1.6 to ~ΔIW–0.8 for the ~4.5 Ga martian mantle also agrees with the lowest fO2 values (~ΔIW–1) ever reported for martian meteorites, which have crystallization ages typically younger than ~1.4 Ga (Fig. 5)8,9,10,16,17,18,19. This consistency implies that the reducing mantle (≤ ΔIW–1) formed from the early differentiation of the martian mantle likely resided for an extended time period in the deep martian interior. Considering that the ancient martian crust has been greatly oxidized (up to ΔFMQ + 4) during impact-induced remelting at ~4.3-4.4 Ga36,55,56, the predominantly oxidized nature of most previously studied shergottites and nakhlites8,9,10,16,17,18 is likely secondary, reflecting admixing of crustal materials into their shallow mantle sources via impacts36,57 or, alternatively, into evolving magmas by assimilation during migration and ascent in the crust8,16.

Fig. 5: Redox evolutions of Mars constrained by Ti isotopic results from this study.
figure 5

The previously recommended fO2 values of core formation for Mars53,76 and martian meteorites8,9,10,16,17,18 are shown for comparison. Modeling results for the melting residues in martian mantle, after melt extraction at fTi = 0.20 and varying fO2 conditions in the presence of Ti3+, are shown as a red solid curve (see Methods for the modeling details). The red dashed arrows indicate the inferred δ49Ti values of the magma sources for the depleted shergottites after correction of the Ti isotopic effects during magma generation as modeled in Fig. 2C, which further define the blue section of the red model curve for melting residues in the martian mantle, translating into fO2 conditions of ~ΔIW–1.6 to ~ΔIW–0.8 during prior melt depletion.

Full size image

Implications for the evolution and potential habitability of Mars

The primary fO2 condition of the martian mantle at ~4.5 Ga as constrained by Ti isotopes (i.e., ~ΔIW–1.6 to ~ΔIW–0.8) contrasts with that of Earth’s upper mantle, which has been shown to be oxidized (i.e., ~ΔFMQ + 0) since 4.3–4.4 Ga58,59. It has been posited that Earth’s upper mantle may have been oxidized to its modern state during and/or just after magma ocean solidification4,60,61. A deep magma ocean develops a vertical gradient in fO2 at constant Fe3+/Fe2+ ratio and, as a result, the surface conditions can be up to 4–5 orders of magnitude more oxidizing than at the base of the terrestrial magma ocean4,62,63. Another related mechanism to enhance fO2 in the upper mantle is iron disproportionation reaction in the solid state, arising from the preference of Fe3+ over Fe2+ for bridgmanite, which operates after core-mantle differentiation60,61. The reducing conditions of the martian mantle at ~4.5 Ga, as implied by δ49Ti results of shergottites in this study, is consistent with those inferred from core-mantle differentiation, excluding noticeable iron disproportionation in the martian mantle.

Based on recent planetary outgassing models2, the Earth’s primordial atmosphere would be predominantly composed of CO2 and N2 after water condenses. The implied early martian mantle fO2 conditions from our data, on the other hand, suggest that as early as ~4.5 Ga, Mars’ atmosphere from mantle outgassing would have higher H2/H2O and CO/CO2 ratios at a given temperature. Such an early reducing atmosphere stands in contrast to the oxidizing state of the modern CO2-dominated martian atmosphere64,65, meaning that Mars may have gone through a transition in atmospheric composition in its geological history. Such a transition can occur at the cooling of an isochemical atmosphere2, and/or may also relate to the delivery of water-rich asteroidal objects that induced impact remelting and subsequent oxidation of the ancient crust of Mars36,57 and/or atmospheric hydrogen escaping through time57. Early martian crustal water reservoirs acquired a high D/H ratio of ~2–4 before ~4.1 Ga, possibly corresponding to this oxidation process, whereas later changes in D/H ratio might be due to secular hydrogen escape over the following 4 billion years66.

Overall, our new results here suggest that Mars likely has time periods of a few hundred million years with reducing atmospheres, which can facilitate prebiotic chemistry forming complex organic molecules relevant for early life67,68. The shift in composition of the Martian atmosphere may have contributed to transitional phases in which increased CO2-H2 collision-induced absorption may sufficiently elevate the greenhouse effect to support liquid water on the surface of Mars ~3.7 billion years ago57,69,70. Such transition seems to be coeval with the oxidation of the ancient martian crust, which was later sampled by parent magmas of SNC meteorites8,9,10,15,16,17,18. Early reducing atmospheric conditions would more easily facilitate habitable conditions of Mars with a mild climate and available liquid water. Future missions that obtain samples containing the earliest geologic records during the transition of atmospheric/crustal oxidation state on Mars would advance our understanding of the origin of life during the geological evolution of terrestrial planets like Earth and Mars.

Materials and methods

Samples

The samples analyzed in this study include nine ureilites (i.e., DaG 319, GRO 95575, ALH 78019, ALH 84136, EET 87517, LEW 85328, MET 78008, PCA 82506, and EET 96042), eight diogenites (i.e., MET 00436, NWA 1461, NWA 4215, NWA 4223, NWA 5480, NWA 1821, Tatahouine, and Bilanga), and a set of martian meteorites that comprise four depleted shergottites (i.e., NWA 4527, SaU 005, DaG 735, and Dhofar 019), three intermediate shergottites (i.e., NWA 5029, NWA 480, and NWA 1950), ten enriched shergottites (i.e., JaH 479, NWA 2975, NWA 7258, NWA 7320, Shergotty, NWA 2990, NWA 4468, NWA 856, NWA 1068, and NWA 6963), one chassignite (i.e., NWA 2737), and three nakhlites (i.e., NWA 5790, NWA 817, and NWA 998).

Sample dissolution and chromatographic purification of Ti

Powders of meteorite samples (~100 mg for each) or reference materials (i.e., AGV-2, BHVO-2, BCR-2, and BIR-1) were weighed into Savillex beakers and digested using mixtures of 22 M HF and 14 M HNO3 acids with a volume ratio of 2:1 at 120 oC for 4 days. The samples were evaporated to dryness and re-dissolved in ~5–10 ml 6 M HCl at 120 oC. This process was repeated several times to decompose fluorides formed from HF dissolution until clear solutions were achieved. Afterwards, sample aliquots containing ~6 µg Ti were taken and properly mixed with a prepared 47Ti-49Ti double spike, where Ti concentrations of sample solutions were determined in advance via a pre-spiked protocol to ensure consistent sample-spike mixing ratio between samples and standards26. The spiked sample aliquots were evaporated to dryness and re-dissolved in 6 M HCl overnight to achieve sample-spike equilibration. Titanium purification was carried out following a three-step protocol using AG1x8 (200–400 meshes) and DGA resins26, which was modified after the earlier methods25,71. To remove potential trace amounts of Ca and Cr, the spiked OL-Ti standards and the final Ti cuts of meteorites and reference materials have been passed once more through the DGA columns before treatment in 14 M HNO3 at 120 oC for destruction of the resin particles and organics inherited from column chemistry and final storage in 0.5 M HNO3 + 0.01 M HF acids.

Neoma MC-ICP-MS

Titanium isotopic compositions of the samples after purification were measured via the Thermo Fisher Scientific Neoma MC-ICP-MS, following an established sample introduction protocol26. An actively cooled membrane desolvation component with an APEX HF desolvating nebulizer from Elemental Scientific was used to stabilize the signals, and a sapphire injector was adopted to suppress oxide and silicon fluoride formation. In addition, N2 gas at a rate of a few mL/min was added to enhance the sensitivity. Such instrumental settings return an intensity of ~15 V on 48Ti+ for a solution of ~600 ppb Ti at an uptake rate of ~50 µL/min using medium mass resolution mode. The newly designed Neoma MC-ICP-MS has an increased mass dispersion (~20%) comparing to the earlier generation of instruments (~14%), which allows monitoring simultaneously ion species ranging from 43Ca+ to 53Cr+ in a single cup configuration26 and thus avoids the use of a dynamic mode for data acquisition. Titanium isotopic data of the samples were measured following the previously described mass spectrometry settings and data acquisition strategies26. Importantly, monitoring signal intensities of 44Ca+, 51V+, and 53Cr+ simultaneously with 46Ti+, 47Ti+, 48Ti+, 49Ti+, and 50Ti+ permits a high-precision correction of isobaric interferences, after which a concomitant and high-precision derivation of δ49Ti and ε50Ti values of samples can be achieved from the double-spike measurements26. As demonstrated by repeated experiments on chondrites (i.e., NWA 5697, NWA 530, NWA 1232, NWA 4428, and NWA 1563) and reference materials (i.e., BHVO-2, BCR-2, and AGV-2), the method provides long-term external reproducibilities of ± 0.010‰ and ± 0.15 epsilon for δ49Ti and ε50Ti, respectively26.

Modeling of Ti isotopic effects from Ti3+ during partial melting of a chondritic precursor

  1. I.

    Quantifying proportions of Ti3+ and Ti4+ in silicate melts and minerals at varying fO2

    As earlier parameterized22,43, redox reaction of Ti in silicate melts follows:

    $${{Ti}}^{3+}{O}_{1.5}\left({melt}\right)+\frac{1}{4}{{{\rm{O}}}}_{2}={{Ti}}^{4+}{O}_{2}\left({melt}\right)$$
    (1)

    Petrological experiments in ref. 22 provide an empirical calibration of Ti3+ and Ti4+ proportions in basaltic silicate melts against fO2 (Fig. 2A):

    $${K}^{{\prime} }_{\hom }=\frac{\left[{{TiO}}_{2}\left({melt}\right)\right]}{\left[{{TiO}}_{1.5}\left({melt}\right)\right]}\times {\left(f{{{\rm{O}}}}_{2}\right)}^{-\frac{1}{4}}$$
    (2)
    $${\left({K}^{{\prime} }_{\hom }\right)}^{-1}=1.24\times {10}^{-4}\pm 1.31\times {10}^{-5}$$
    (3)

    This calibration was established for the silicate melts of basaltic to andesitic compositions, i.e., SiO2 = 47.52–54.79 wt%, TiO2 = 1.06−1.69 wt%, Al2O3 = 10.03–21.12 wt%, FeO = 0.19–0.65 wt%, MgO = 15.08–15.73 wt%, and CaO = 14.71–21.56 wt%22. This allows the derivation of the Ti3+ proportion in a basaltic silicate melt at a chosen fO2. While partitioning coefficients of Ti3+ and Ti4+ between minerals (e.g., orthopyroxene) and basaltic silicate melts have been experimentally determined, e.g., \({D}^{{{Ti}}^{3+}}({opx}-{melt})=0.327\) and \({D}^{{{Ti}}^{4+}}\left({\mathrm{opx}}-{\mathrm{melt}}\right)=0.0954\)22\(,\) Ti3+ and Ti4+ proportions in orthopyroxene at equilibrium with basaltic silicate melts can be estimated via:

    $${X}^{{{Ti}}^{3+}}\left({opx}\right)=\frac{{X}^{{{Ti}}^{3+}}\left({melt}\right)\times {D}^{{{Ti}}^{3+}}\left({opx}-{melt}\right)}{\left[1-{X}^{{{Ti}}^{3+}}\left({melt}\right)\right]\times {D}^{{{Ti}}^{4+}}\left({opx}-{melt}\right)+{X}^{{{Ti}}^{3+}}\left({melt}\right)\times {D}^{{{Ti}}^{3+}}\left({opx}-{melt}\right)}$$
    (4)
  2. II.

    Modeling of Ti isotopic effects from Ti3+ during melt extraction

In order to quantify the Ti isotopic effects during melt extraction in the presence of Ti3+, it requires the Ti isotope fractionation factors between silicate minerals and silicate melts to be calibrated either by analyses of natural and experimental samples or by first-principles calculations. Here, we integrate existing constraints from literature:

  1. i)

    First-principles calculations show that there can be some Ti isotope fractionations between Ti4+-dominated orthopyroxene/clinopyroxene and Ti3+-dominated orthopyroxene/clinopyroxene23. The equilibrium Ti isotope fractionation factor (i.e., 103lnα of 49Ti/47Ti) of a phase can be formalized as:

    $${10}^{3}{ln}\alpha ={ax}+b{x}^{2}+c{x}^{3}$$
    (5)

    where x = 106/T2 with T as temperature in Kelvin.

    For Ti3+ substitution of Mg2+ + Si4+ = Ti3+ + Al3+ (cpx or opx), there are:

    a = –0.89292, b = 3.012 × 10−2, c = –9.991 × 10−4 for Mg15TiCa16Si31AlO96 (clinopyroxene) with Ti/(Ti+Si) = 1/32;

    a = –0.91643, b = 3.060 × 10−2, c = –1.015 × 10−3 for Mg31TiSi31AlO96 (orthopyroxene) with Ti/(Ti+Si) = 1/32.

    For Ti4+ substitution of Si4+ = Ti4+ (opx), there are:

    a = –0.02689, b = 8.900 × 10−4, c = –3.100 × 10−5 for Mg32Si31TiO96 (orthopyroxene) with Ti/(Ti+Si) = 1/32;

    a = –0.04235, b = 1.170 × 10−3, c = –3.900 × 10−5 for Mg64Si63TiO192 (orthopyroxene) with Ti/(Ti+Si) = 1/64.

    Overall, isotope fractionation factors of either Ti3+ or Ti4+ turn out to be identical between orthopyroxene and clinopyroxene23. For systems containing only Ti4+, Ti-rich oxides (e.g., rutile, ilmenite, and magnetite) show 103lnα values lower by ~0.7–1.0‰ relative to those of orthopyroxene and clinopyroxene at 1000 K23. However, the crystallization of rutile, ilmenite, and magnetite usually requires fO2 conditions exceeding typical fO2 conditions in ureilites, diogenites, and shergottites or very evolved melt composition with MgO <5 wt.%. Therefore, we consider in this study mainly the Ti isotope fractionation factors (i.e., 103lnα) of Ti3+ and Ti4+ in orthopyroxene and clinopyroxene, respectively, at 1473 K:

    $${10}^{3}{ln}\alpha \left({{{Ti}}^{3+}}_{{opx}/{cpx}}-{{{Ti}}^{4+}}_{{opx}/{cpx}}\right)=-0.450{{\textperthousand }}$$
    (6)
  2. ii)

    Studies of Earth’s mantle and mantle-derived rocks, which are characterized by fO2 conditions with predominantly Ti4+, show that at equilibrium there is negligible Ti isotope fractionation between orthopyroxene/clinopyroxene and basaltic silicate melts, that is:

$${10}^{3}{ln}\alpha \left({{{Ti}}^{4+}}_{{opx}/{cpx}}-{{{Ti}}^{4+}}_{{melt}}\right)=+0.000{{\textperthousand }}$$
(7)

While neither experimental nor theoretical constraint on Ti3+ in silicate melt exists, we can assume for Ti3+:

$${10}^{3}{ln}\alpha \left({{{Ti}}^{3+}}_{{opx}/{cpx}\,}-{{{Ti}}^{3+}}_{{melt}}\right)=+0.000{{\textperthousand }}$$
(8)

Afterwards, we can combine Eqs. 2 to 4 and 6 to 8 to calculate the equilibrium Ti isotope fractionation factor between mineral and melt at varying fO2:

$${10}^{3}{ln}\alpha \left({mineral}-{melt}\right)= \, \left[{X}^{T{i}^{3+}}\left({opx}/{cpx}\right)-{X}^{T{i}^{3+}}\left({melt}\right)\right] \\ \times {10}^{3}{ln}\alpha \left({{{Ti}}^{3+}}_{{opx}/{cpx}}-{{{Ti}}^{4+}}_{{opx}/{cpx}}\right)$$
(9)

The Ti isotopic effects of Ti3+ and Ti4+ during partial melting of a chondritic precursor can be further quantified via a Rayleigh process:

$${\Delta }^{49}{{Ti}}_{{melting}\,{residue}-{chondritic}\,{precursor}}=-{10}^{3}{ln}\alpha \left({mineral}-{melt}\right)\times {ln}\left({f}_{{Ti}}\right)$$
(10)

where fTi represents the Ti fraction remaining in the melting residue after melt extraction, e.g., fTi = 0.2 adopted in Figs. 2C, 3, 5. The modeled δ49Ti value of the melting residue should be:

$${\delta }^{49}{{Ti}}_{{melting}\,{residue}}={\delta }^{49}{{Ti}}_{{chondrite}}+{\triangle }^{49}{{Ti}}_{{melting}\,{residue}-{chondritic}\,{precursor}}$$
(11)

where the chondrite average has been adopted here, i.e., δ49Ti = +0.053‰26. Overall, integrating Eqs. 1 to 11 provides the modeling curves for Ti isotopic effects from Ti3+ during melt extraction in Figs. 2C, 3, 5.

  1. III.

    Key variables to the model

  1. i)

    Equilibrium Ti isotope fractionation factors

    We note that \({10}^{3}{ln}\alpha ({{{Ti}}^{3+}}_{{opx}/{cpx}}-{{{Ti}}^{4+}}_{{opx}/{cpx}})\) and \({10}^{3}{ln}\alpha ({{{Ti}}^{4+}}_{{opx}/{cpx}}-{{{Ti}}^{4+}}_{{melt}})\) can be estimated based on first-principles calculations and studies of Earth’s mantle and mantle-derived rocks, respectively. In comparison, \({10}^{3}{ln}\alpha ({{{Ti}}^{3+}}_{{opx}/{cpx}}-{{{Ti}}^{3+}}_{{melt}})\) is relatively uncertain. Nonetheless, choosing instead a \({10}^{3}{ln}\alpha ({{{Ti}}^{3+}}_{{opx}/{cpx}}-{{{Ti}}^{3+}}_{{melt}})\) value of –0.050‰ in Eq. 8 would only lead to shifts of ≤0.011‰ in the predicted δ49Ti values of the residual mantle after melt extraction at fTi value of 0.2 and fO2 conditions of ≥ ΔIW–3. Such insensitivity of the predicted Ti isotope fractionation in the residual mantle to the chosen \({10}^{3}{ln}\alpha ({{{Ti}}^{3+}}_{{opx}/{cpx}}-{{{Ti}}^{3+}}_{{melt}})\) value arises from a fact that proportion of Ti3+ remains low (≤14%) in basaltic silicate melts at fO2 conditions of ≥ΔIW–3.

    We also have to point out that the \({10}^{3}{ln}\alpha ({{{Ti}}^{3+}}_{{opx}/{cpx}}-{{{Ti}}^{4+}}_{{opx}/{cpx}})\) value of –0.450‰ at 1473 K was chosen based on existing first-principles calculations on Ti3+ substitution of Mg2+ + Si4+ = Ti3+ +  Al3+ for orthopyroxene or clinopyroxene23, which is largely true for most planetary magma systems. This Ti3+ substitution however can be suppressed in an Al-poor and Ti-rich system (e.g., the parent magmas of lunar mare basalts), which may lead to distinct Ti isotopic effects. For instance, petrological experiments using a mare basalt-like starting material (i.e., SiO2 = 43.06 wt%, TiO2 = 22.06 wt%, Al2O3 = 5.25 wt%, MgO = 22.47 wt%, and CaO = 8.09 wt%) provide slightly positive \({10}^{3}{\mathrm{ln}}\alpha \left({\mathrm{mineral}}-{\mathrm{melt}}\right)\) values of +0.050 ± 0.021‰ to +0.057 ± 0.037‰ at fO2 conditions of ΔIW–1 to ΔIW + 041. As such, more experimental or theoretical calibrations would be required if aiming to trace fO2 conditions of the Moon with Ti isotopes.

  2. ii)

    Remaining Ti fraction in melting residue (fTi)

    Assuming a Rayleigh process for melt extraction, the magnitude of Ti isotope fractionation left in the residual mantle would be controlled by \({10}^{3}{\mathrm{ln}}\alpha \left({\mathrm{mineral}}-{\mathrm{melt}}\right)\) and fTi (Eq. 10). For the chosen equilibrium Ti isotope fractionation factors of Ti3+ and Ti4+ in minerals and silicate melts (Eqs. 6 to 8), a fTi value of 0.2 provides the modeling results that fit best with the δ49Ti data from ureilites and diogenites (Fig. 3). Such a fTi value is well consistent with a fractional melting process of a chondritic precursor with an average DTi value of 0.15 and a partial melting degree of 0.2. For fTi = 0.35, the Ti isotope fractionation in the residual mantle is predicted to be weaker (Fig. 2C), which cannot account for the lowest δ49Ti values observed for ureilites (down to –0.105 ± 0.019‰) and diogenites (down to –0.039 ± 0.004‰).

  3. iii)

    Chondrite δ49Ti average

Recently, a Ti isotope method has been developed via Neoma MC-ICP-MS for concomitant determination of δ49Ti and ε50Ti in meteorites at external reproducibilities of ±0.010‰ and ±0.15 epsilon, respectively26. Using this method, the δ49Ti average of chondrites has been recommended to be +0.053 ± 0.005‰ (2se, n = 22), including 22 chondrites for the average (i.e., 1 CI, 1 CV, 6 CM, 2 CO, 1 CH, 2 CK, 4 CR, 1 EH, 3 L, and 1 LL). However, a different δ49Ti average (+0.023 ± 0.009‰, 2se, n = 20) for chondrites was later proposed based on measurements of 10 ordinary chondrites by three laboratories72. According to Eq. 11, a decrease of 0.030‰ in the adopted chondrite δ49Ti average can lead to an equivalent downward shift of the modeling curves in Figs. 2C, 3, 5. Afterwards, a higher fTi value may be required to reduce the Ti isotope fractionation from melt extraction to account for the δ49Ti data of ureilites, diogenites, and martian meteorites. Such a shift is much weaker if combining the two datasets for the estimate of chondrite δ49Ti average, i.e., +0.044 ± 0.006‰ (2se, n = 32).

We notice that for the same ordinary chondrite samples (e.g., Estacado and Richardton) in ref. 72, the δ49Ti values obtained by the three included laboratories can differ by up to 0.068‰ (Supplementary Fig. S2), manifesting large analytical uncertainties associated with their published δ49Ti dataset. In conjunction with the 2sd values of ±0.024‰ to ± 0.049‰ in δ49Ti for their geostandards, the δ49Ti data of ordinary chondrites from ref. 72 are three- to four-fold less precise compared to those reported in ref. 26 (Supplementary Fig. S2). Overall, the δ49Ti average of +0.053 ± 0.005‰ from ref. 26 seems to be the best estimate for chondrites so far if taking into account the following additional lines of evidence:

  1. a)

    The ~3.8 Ga Isua metabasalts in ref. 26 show a δ49Ti average of +0.048 ± 0.005‰ (2se, n = 5), and this average has been reproduced by ref. 73 (+0.046 ± 0.016‰, 2se, n = 11). The studied ~3.8 Ga Isua metabasalts in ref. 26 have MgO ≥7.05 wt%, FeOT ≤13.21 wt%, and TiO2 ≤1.518 wt%, which are unlikely to reach Fe-Ti oxide saturation under fO2 conditions relevant to Earth’s upper mantle (~ΔFMQ + 0)74. Also, the ~3.48 Ga Barberton komatiites and basaltic komatiites provide consistent δ49Ti average values of +0.044 ± 0.009‰ (2se, n = 4) and +0.048 ± 0.008‰ (2se, n = 4). All these averages of Earth’s early Archean mantle-derived rocks are well consistent with the δ49Ti average of +0.053 ± 0.005‰ for bulk chondrites from ref. 26 or the δ49Ti estimate of +0.044 ± 0.006‰ for bulk chondrites by combining the chondrite data from refs. 26, 72.

  2. b)

    The oxidized ureilite sample EET 87517 in this study provides a δ49Ti value of +0.052 ± 0.004‰, which, as modeled in Fig. 2C, likely records negligible isotopic effects from Ti3+ and would most likely represent the δ49Ti value of its chondritic precursor. Six of the studied diogenites from the Vesta asteroid show δ49Ti values clustering around +0.053 ± 0.005‰. In addition, the studied shergottites define a positive correlation between δ49Ti and (La/Yb)N and the enriched shergottites having (La/Yb)N values of ~0.75–1.15 show a δ49Ti average of +0.043 ± 0.015‰ (2se, n = 10). All these suggest that Mars and the parent bodies of ureilites and diogenites have bulk δ49Ti values that are in line with the recommended δ49Ti average for chondrites in ref. 26 (i.e., +0.053 ± 0.005‰).

  1. iv)

    Influence of model variables on the constrained initial fO2 of the martian mantle

Note that the depleted shergottites have been characterized by δ49Ti values that are transitional between those of diogenites and ureilites. Accepting that the enrichments of the light Ti isotopes in diogenites, ureilites and depleted shergottites arise from the presence of Ti3+ in the magma systems, the fO2 conditions during early mantle differentiation of Mars at ~4.5 Ga would range between that of ureilites (~ΔIW–3.3 to ~ΔIW–1.5) and diogenites (~ΔIW–2 to ~ΔIW–0), irrespective of the above-mentioned model variables.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.