Constraints on the formation of carbonates and low‐grade metamorphic phases in the Martian crust as a function of H2O‐CO2 fluids

Low‐grade metamorphic hydrous minerals and carbonates occur in various settings on Mars and in Martian meteorites. We present constraints on the stability of prehnite, zeolites, serpentine, and carbonates by modeling the influence of H2O‐CO2 fluids during low‐grade metamorphism in the Martian crust using compositions of a Martian basalt and an ultramafic cumulate. In basaltic compositions with 5 wt% fluid, our models predict prehnite in less oxidized, CO2‐poor conditions (≤0.44 mol kg−1 CO2) on warmer geotherms of 20 °C km−1. At fluid‐saturated conditions, epidote and laumontite are replaced by quartz, calcite, chlorite, and muscovite. In ultramafic compositions with 5 wt% fluid, antigorite (serpentine) is stable at CO2‐poor conditions of ≤0.33 mol kg−1, while talc forms at 0.05–0.56 mol kg−1 CO2. At fluid‐saturated conditions, antigorite is replaced by talc and chlorite, and at higher X(CO2) by magnesite and quartz. Our models therefore suggest that prehnite, zeolites, and serpentine have formed in a CO2‐poor environment on Mars implying that fluids during their formation either did not contain high amounts of CO2 or had degassed CO2. Carbonates and potentially talc would have formed in the presence of a CO2‐bearing fluid and therefore at different alteration stages than for prehnite, zeolites, and serpentine either in the same hydrothermal event during which the fluid composition changed gradually due to cooling and precipitation or by separate and successive alteration events with fluids of different compositions.


INTRODUCTION
Carbonate minerals are of special interest in the study of the potential for life on Mars, since they are linked to both the water and inorganic carbon cycles (e.g., Niles et al., 2013). Locations where hydrated phyllosilicates and carbonates have been detected, such as in the Nili Fossae region as well as Leighton and McLaughlin craters, could be particularly suitable for a deep biosphere on Mars, as they suggest a prolonged past presence of alkaline water, carbon, and temperatures above freezing, all of which are critical prerequisites for life (Michalski & Niles, 2010;Michalski et al., 2013;Niles et al., 2013). The Martian subsurface may be a habitable environment for chemolithoautotrophic microorganisms since it is protected from surface radiation, provides increased porosity and permeability due to the lower gravity on Mars, includes possible nutrients and energy from geochemical sources such as serpentinization reactions (Fisk & Giovannoni, 1999;Jones et al., 2011;Michalski et al., 2013;Schrenk et al., 2013), and provides a habitable zone possibly extending up to 14 km below the surface (McMahon et al., 2013).
Better constraints on mineral stabilities and reactions in the Martian upper crust as a function of H 2 O-CO 2 fluid variations are therefore critical to assess the habitability potential of the subsurface. In this study, we use phase equilibria modeling to study the effect of fluid composition on low-grade metamorphic reactions.
While this study is focused on metamorphic subsurface fluids at temperatures above 150°C, which is higher than the known limit of 122°C (at elevated hydrostatic pressure) for the survival of terrestrial hyperthermophilic microbes (Takai et al., 2008), these initially hot fluids could cool down over geological timescales and circulate through the subsurface by geothermal gradient-driven groundwater convection (Travis et al., 2003).
The physical conditions and water abundances required for the formation of OH-bearing low-grade metamorphic phases in Martian basaltic and ultramafic rock compositions have been investigated via phase equilibria modeling assuming the fluid to be pure H 2 O (Semprich et al., 2019). The models estimated the formation of prehnite from basaltic protoliths to require at least 2.5 wt% water in the system. Serpentine forms by alteration of ultramafic compositions at >2.7 wt% H 2 O (Semprich et al., 2019). While water is fundamental for the formation of these characteristic minerals, potential Martian subsurface fluids likely contain other species in addition to H 2 O, such as CO 2 as suggested by the presence of carbonates detected on Mars and in meteorites (e.g., Bandfield et al., 2003;Bridges et al., 2019;Leshin et al., 2013;Morris et al., 2010;Sutter et al., 2012).
Processes suggested for the crystallization of carbonates on Mars include (1) as a product of serpentinization (Amador et al., 2017;Brown et al., 2010, and references therein) although a study of spectral data across Mars only identified serpentine together with carbonates in the Nili Fossae region ; (2) as an alteration product of either olivine or serpentine due to volcanic-or impactrelated hydrothermal activity Ehlmann et al., 2009;Golden et al., 2000;Viviano et al., 2013); (3) as sedimentary (fluvial/lacustrine/aeolian) detrital or in situ deposits (Doran et al., 1998;Goudge et al., 2015;Grotzinger et al., 2005;Horgan et al., 2020); (4) by diagenesis in a low-temperature environment (e.g., ; (5) by fluid evaporation (Boynton et al., 2009;Ruff et al., 2014); and (6) by weathering of olivine-rich rocks exposed at the surface (Bultel et al., 2019). A combination of processes may also be feasible as suggested for the olivine-carbonate mineralogy in Jezero crater (Brown et al., 2020). The lack of widespread carbonate deposits on the Martian surface could also indicate that most alteration processes occurred in the subsurface not in contact with the atmosphere (e.g., Ehlmann et al., 2011). This is corroborated by carbonates being mixed with igneous minerals at Nili Fossae , Leighton crater , and Gusev cater (Morris et al., 2010), suggesting their formation by fluid alteration in the subsurface rather than sedimentary processes . Furthermore, carbonates in meteorites were likely formed during brief subsurface aqueous events (Bridges & Schwenzer, 2012;Bridges et al., 2019;Changela & Bridges, 2011;Valley et al., 1997). Additionally, spectral features in thermal emission spectrometer (TES) data in northeast Syrtis Major and several other regions on Mars are best explained by the presence of carbonate decomposition products, which could indicate an interaction of lavas with subsurface carbonate deposits or devolatilization of initially carbonate-bearing lavas by later impacts of magmatic activity (Glotch & Rogers, 2013). Understanding the role of CO 2 in fluid-rock interaction models and its influence on low-grade metamorphic mineral stability and phase relations can therefore provide better constraints on subsurface fluid conditions. Several thermochemical models have produced carbonates using starting compositions resembling Martian rocks under various temperature conditions (0-300°C), CO 2 fugacities, and water-to-rock ratios (Bridges & Schwenzer, 2012;Bridges et al., 2015;Filiberto & Schwenzer, 2013;Griffith & Shock, 1995, 1997Zolotov & Mironenko, 2016) but at predominantly low pressures and therefore relatively shallow depth. The secondary mineralogy in the nakhlites is best explained by percolating CO 2 -rich hydrothermal fluids with initial temperatures between 150 and 200°C, pH 6-8, and a water-to-rock ratio of ≤300 (Bridges & Schwenzer, 2012). Upon cooling of the fluids, Fe-rich phyllosilicate and subsequently an amorphous gel were precipitated (Bridges & Schwenzer, 2012;Changela & Bridges, 2011). The carbonate-bearing assemblages in nakhlites hence precipitated from rapidly cooling fluids and are likely metastable . The zoned carbonate globules in the orthopyroxenite cumulate ALH 84001, however, likely formed at low temperatures presumably near 18°C as suggested by carbon and oxygen isotope ratios (Halevy et al., 2011). A low temperature <150°C is also suggested by experimental work (Golden et al., 2000(Golden et al., , 2001 and geochemical modeling (Melwani Daswani et al., 2016; although formation conditions still remain relatively unconstrained. Furthermore, groundwater alteration processes at low temperature (~5°C) and atmospheric CO 2 partial pressures between~0.001 and 2 bars have been suggested by geochemical modeling as a possible formation mechanism for carbonate and phyllosilicate alteration in the Nili Fossae region . While the low temperatures match the observed Mg-rich carbonate above phyllosilicate stratigraphy on a smaller scale , the region was likely exposed to multiple alteration events under varying conditions due to extensive cratering including higher temperature fluids as suggested by the presence of prehnite (Ehlmann et al., 2011).
In particular, impact cratering and volcanism can induce large-scale, long-lasting hydrothermal systems on Mars resulting in the formation of the observed alteration minerals (Abramov & Kring, 2005;Costello et al., 2020;Crandall et al., 2021;Filiberto & Schwenzer, 2013;Griffith & Shock, 1995, 1997Schwenzer & Kring, 2013). While most studies of hydrothermal alteration have focused on low-pressure environments and hence shallow depths, we are interested in low-grade metamorphic reactions at deeper levels and low fluid contents, resembling restricted fluid flow in the deeper subsurface. Here, we investigate the effect of variations in H 2 O-CO 2 fluid compositions on a shergottite-like Martian basaltic and an ultramafic cumulate composition.

Model Parameters
Phase diagrams are calculated with the Gibbs free energy minimization software Perple_X 6.8.9 (Connolly, 2005) using an internally consistent thermodynamic data set (Holland & Powell, 1998 following our previous approach (Semprich et al., 2019). MnO, Cr 2 O 3 , and P 2 O 5 are excluded from the calculation because of their low abundance and/or limitation of solid solution models. Fe 2 O 3 is defined by the following existing components: 2 FeO + 0.5 O 2 and is then set to 0 for reducing conditions and to a value corresponding to 10% of the total iron for more oxidizing conditions (resulting in~1.5-2.5 wt% Fe 2 O 3 for the compositions considered). All oxides are normalized to 100% by Perple_X. We use the following solid solution models (Table 1): clinopyroxene (Cpx), orthopyroxene (Opx), olivine (Ol), dolomite (Dol), and magnesite (Mgs) from Holland and Powell (1998); pumpellyite (Pmp), actinolite (Act), and stilpnomelane (Stp) from Massone and Willner (2008); chlorite (Chl) and white mica   Thompson and Waldbaum (1969); plagioclase (Pl) from Newton et al. (1980); ilmenite (Ilm) from White et al. (2000). Talc (Tlc) and brucite (Brc) were assumed to be ideal solutions. Quartz (qz), prehnite (prh), calcite (cal, and aragonite in the low-temperature, high-pressure section of the diagrams), stilbite (stlb), laumontite (lmt), wairakite (wrk), lawsonite (lws), rutile (rt), titanite (ttn), and magnetite (mag) are treated as pure phases. Vesuvianite and garnet (andradite) are excluded since their modeled stabilities exceed those in real rocks and they have not been detected on Mars. Fluid (F) is represented by a C-O-H generic hybrid equation-of-state fluid model (Connolly, 1995) allowing for the following species: H 2 O, CO 2 , CH 4 , H 2 , CO. The internal molecular fluid equation of state for pure H 2 O and CO 2 species is set to a Compensated-Redlich-Kwong equation, which allows the calculation of the volumes and fugacities of H 2 O and CO 2 over a large pressure and temperature range (Holland & Powell, 1991.

Protolith Compositions
We use the starting compositions of Bounce Rock (Zipfel et al., 2011) and a poikilitic shergottite ALHA77005 (Lodders, 1998) as representative Martian mafic and ultramafic rocks, respectively (Table 2). Bounce Rock resembles basaltic shergottite meteorites in texture, mineralogy, and chemistry, in particular lithology B of EETA79001 and QUE 94201 (Zipfel et al., 2011). For both compositions, we assume two different oxidation states (1) all iron is assumed to be divalent FeO; and (2) 10% of the iron is assumed to be trivalent Fe 2 O 3 , and the remainder divalent, which gives 1.55 wt% Fe 2 O 3 and 2.23 wt% Fe 2 O 3 for Bounce Rock and ALHA77005, respectively. The chosen range covers the redox states measured for the majority of Martian basalts (e.g., Herd, 2003;Schmidt et al., 2013).

Fluid Content and Composition
While the fluid content of the deeper Martian subsurface now and at earlier time periods is not known, the present-day estimates by gamma ray spectroscopy of surface rocks range between 1.5% and 7.5 wt% H 2 O in low-and mid-latitude regions (Boynton et al., 2007). A similar near-surface water range of 2-7 wt% (Wernicke & Jakosky, 2021) was estimated based on data obtained by the Mars Odyssey Neutron Spectrometer (MONS; Feldman et al., 2004;Maurice et al., 2011) and the 3 µm absorption of the Observatoire pour la Min eralogie, l'Eau, les Glaces et l'Activit e (OMEGA; Audouard et al., 2014;Milliken et al., 2007). Furthermore, Curiosity's Dynamic Albedo of Neutron (DAN) instrument yields estimates between 1.5 and 3.3 wt% and up to~5 wt% water locally within the upper 60 cm below the surface (Litvak et al., 2016; (Thomas et al., 2020). The percentages for the Martian soil at Gale crater are likely derived from both adsorbed water molecules in the regolith and water structurally bound in minerals (Mitrofanov et al., 2014). While the H 2 O content of adsorbed water in the deeper crust is unknown, phyllosilicates have been excavated by impact cratering from depths of at least 7 km and even up to 17 km for one impact event (Flahaut et al., 2012;Mustard, 2019;Mustard et al., 2008;Sun & Milliken, 2015). This would suggest that at least the upper 10 km of the crust contains structurally bound water in the form of hydrous minerals (Mustard, 2019). Estimates of average subsurface water content range between 0.5 and 3 wt% (Mustard, 2019; Wernicke & Jakosky, 2021) although regional differences may result in locally higher values.
Modeling of low-grade metamorphic phases as a function of water availability has shown that~2.5 wt% H 2 O is required for the formation of prehnite in basaltic rocks and~2.7 wt% for serpentine in ultramafic compositions, while the zeolites laumontite and analcime require~4 wt% and 5.2 wt% H 2 O, respectively (Semprich et al., 2019). To include these important low-grade metamorphic phases in our models, we have therefore chosen to compute phase equilibria at somewhat higher fluid content than the estimated average for the Martian subsurface. The amount of CO 2 in the subsurface is not well constrained. The Martian dust contains 2-5 wt% carbonates according to detections by remote sensing from orbit (Bandfield et al., 2003), and soils contain up to 1 wt% . However, in situ measurements by the Spirit rover in Gusev crater identified carbonates within localized outcrops in abundances of 16-34 wt% (Morris et al., 2010). Based on their relative limited detection, Hu et al. (2015) estimate the amount of carbonates in the crust to range between 1 and 5 wt%. Similar to estimates of the water content, regional variations are likely as shown by the low amounts of carbonates in large parts of Gale crater until the recent detection of siderite (Archer et al., 2020).
Based on measured values and mineralogical restrictions, we use two setups for our calculations including (1) a model where the amount of fluid in the system is restricted to 5 wt% and its composition is varied from H 2 O-dominated to CO 2 -dominated and (2) a fluid-saturated model with the fluid composition defined by X(CO 2 ). In calculations with a saturated fluid, O 2 is set as an unconstrained saturated component to be consistent with C-O-H fluid speciation instead of specifying FeO/Fe 2 O 3 in the starting composition. In this setup, we exclude all COH fluid species other than H 2 O, CO 2 , and O 2 .

Further Model Parameters
Until the availability of in situ measurements, estimates of the surface heat flux of Mars and resulting geothermal gradients are based on indirect methods relying on several poorly constrained parameters such as the abundance of heat-producing elements, topography, crust and mantle thickness and their densities (e.g., McGovern et al., 2002McGovern et al., , 2004Parro et al., 2017;Plesa et al., 2016). The estimated heat flow in Noachian terrains ranges from >35 to 60 mW m À2 corresponding to thermal gradients between >14 and 20°C km À1 (McGovern et al., 2004) and a Noachian geotherm of 12°C km À1 has been calculated for the Nili Fossae region (McSween et al., 2015), which shows an abundance of low-grade minerals and carbonate detections (e.g., Ehlmann et al., 2011). Based on these values, we selected 10, 13, and 20°C km À1 geotherms in all calculated phase diagrams. Mineral modes as a function of fluid composition are only extracted along the 13 and 20°C km À1 geotherms since assemblages formed along these gradients are more likely to contain the characteristic low-grade metamorphic minerals such as prehnite and zeolites. For depth estimates from pressure, we use the following equation: h = P/qg, where the gravity (g) for Mars is assumed to be 3.727 m s À2 and the average crustal density (q) is 2900 kg m À3 , a commonly used value in Mars models (e.g., McGovern et al., 2004;Parro et al., 2017;Plesa et al., 2016;Sohl et al., 2005). However, any depth estimates given in this work should be taken as a rough approximation since the actual density is not constant and varies with the extent of fracturing and of compaction from overburden pressure in the upper crust (Clifford, 1993) and as a function of pressure at depth (e.g., Semprich & Filiberto, 2020).

Mineral Stabilities in Basaltic Compositions with Variation in H 2 O-CO 2
Mineralogy of Bounce Rock with 5 wt% Bulk Fluids Figure 1 shows phase stability diagrams calculated for the composition of Bounce Rock (Table 2) with only divalent iron (Figs. 1a, 1c, and 1e), and 1.55 wt% Fe 2 O 3 representing 10% of FeO tot recalculated to be trivalent (Figs. 1b, 1d, and 1f). The amount of fluid in the bulk composition is set to 5 wt% and increases in CO 2 from top to bottom with 0 wt% CO 2 (Figs. 1a and 1b), 1 wt% CO 2 (Figs. 1c and 1d), and 2 wt% CO 2 (Figs. 1e and 1f). Figure 2 shows mineral modes as a function of CO 2 (mol kg À1 ) in the bulk composition at 200°C on the 13 and 20°C km À1 geotherms (indicated by red circles in Fig. 1).
The prehnite stability field (shown in yellow in Fig. 1) is largest in calculations with divalent iron and pure H 2 O fluids in the temperature range of 150°C tõ 320°C and pressures ≤0.37 GPa (Fig. 1a). In the modeled composition with trivalent iron, prehnite stability is reduced toward lower pressures (≤0.3 GPa) due to the presence of pumpellyite and toward lower temperatures (≤280°C) due to the presence of epidote. Prehnite is stable together with pumpellyite (orange field, Fig. 1) and epidote (green field; Fig. 1). In compositions with as little as 1 wt% CO 2 , prehnite stability is decreased significantly for both reduced and oxidized conditions (Figs. 1c and 1d); however, its stability is still larger in the composition with divalent iron (Fig. 1c). In the composition with trivalent iron, prehnite occurs predominantly with epidote, pumpellyite, or both ( Fig. 1d). Compositions with >2 wt% CO 2 have no prehnite (Figs. 1e and 1f). Prehnite is present in up tõ 20 vol% in the composition with no Fe 2 O 3 at 200°C on both the 20 and 13°C km À1 geotherms (Figs. 2a and 2c) and at lower CO 2 , but is not stable abovẽ 0.44 mol kg À1 (1.9 wt%) CO 2 . In compositions with 1.55 wt% Fe 2 O 3 , prehnite is only present at 200°C on the 20°C km À1 geotherm up to 15 vol% and coexists with pumpellyite at low CO 2 of ≤0.26 mol kg À1 (1.13 wt %) and is not stable >0.44 mol kg À1 CO 2 . At 200°C on the 13°C km À1 geotherm, prehnite is not stable. Hence, our models predict prehnite formation in less oxidized, CO 2 -poor environments and predominantly along warmer geotherms.
Pumpellyite is stable in modeled compositions with divalent and trivalent iron and pure H 2 O and limited CO 2 fluids but absent at higher CO 2 of 2 wt% . In calculations with divalent iron, pumpellyite replaces prehnite above~0.1 GPa at 150°C and at higher pressures with increasing temperature (Figs. 1a and 1c). In compositions with trivalent iron, pumpellyite is stable with prehnite at lower pressures ( Figs. 1b and 1d). In calculations with 1 wt% CO 2 , pumpellyite stability is extended toward higher pressures, since calcite replaces the Ca-bearing lawsonite (Figs. 1c and 1d). Pumpellyite is not stable in compositions with divalent iron at 200°C on the 13 and 20°C km À1 geotherms (Figs. 2a and 2c). In calculations with trivalent iron on the 20°C km À1 geotherm (Fig. 2b), pumpellyite is stable in up to 15 vol% together with prehnite at low CO 2 of ≤0.26 mol kg À1 (1.13 wt%). Pumpellyite is present in amounts of >20 vol% in models with trivalent iron on the 13°C km À1 geotherm, where prehnite is absent (Fig. 2d). According to our models, pumpellyite would be expected in CO 2 -poor conditions at higher pressures and hence on colder geotherms and is more stable in compositions with trivalent iron. Fig. 1. Phase stability fields for the composition of Bounce Rock with fluid limited to 5 wt% in the range of 150-450°C and 0-0.5 GPa using starting compositions with FeO only (a, c, e), and 1.55 wt% Fe 2 O 3 , equivalent to 10% of FeO tot (b, d, f). The CO 2 content in the bulk composition is increased from top to bottom with 0 wt% CO 2 (a, b); 1 wt% CO 2 (c, d), and 2 wt% CO 2 (e, f). Mineral abbreviations (solid solutions capitalized; pure phases lowercase): Act = actinolite; Bt = biotite; cal = calcite; Chl = chlorite, Cpx = clinopyroxene; Dol = dolomite; Ep = epidote; F = COH-Fluid, ilm = ilmenite, Kfs = K-feldspar, lmt = laumontite, lws = lawsonite, Ms = muscovite (white mica), Pl = plagioclase, Pmp = pumpellyite, prh = prehnite, qz = quartz, stlb = stilbite, Stp = stilpnomelane, ttn = titanite, wrk = wairakite. Small fields are not labeled. Colored areas indicate the stability fields of characteristic low-grade minerals relevant for Mars: red = pumpellyite, yellow = prehnite, blue = epidote, orange = pumpellyite + prehnite, green = epidote + prehnite, purple = pumpellyite + prehnite, turquoise = pumpellyite + epidote + prehnite. Textures indicate stabilities of several zeolites. Minerals in bold have been reported as alteration phases in the Nili Fossae area (e.g., McSween et al., 2015), in particular the assemblage prehnite + chlorite, unspecified zeolites (which are represented here as stilbite, laumontite, and wairakite), albite, shown here as Kfs, and Fe/Mg carbonates (here as dolomite). Potential mineral detections are in bold italic including quartz (which is detected as opaline silica), epidote, and muscovite. Solid black lines represent geotherms at 10, 13, and 20°C km À1 , respectively. Red circles on the geotherms represent conditions at which mineral modes are extracted for Epidote is stable at temperatures >250°C in modeled compositions with divalent iron and with 0-1 wt% CO 2 (Figs. 1a and 1c). Epidote stability is shifted toward lower temperatures (150°C) in calculations with trivalent iron at 1 wt% CO 2 (Fig. 1d). In compositions with 2 wt% CO 2 , epidote replaces prehnite, pumpellyite, and zeolites at all temperatures considered (Figs. 1e and 1f). Epidote is stable in the range of 0.28-0.88 mol kg À1 CO 2 (1.2-3.9 wt%) up to 22 vol% in compositions with divalent iron on both geotherms (Figs. 2a and 2c). In calculations with trivalent iron, epidote stability is shifted toward lower CO 2 levels of 0.17 mol kg À1 (0.76 wt%) on both geotherms (Figs. 2b and 2d). Therefore, epidote in our models is stable at higher CO 2 levels and coexists with dolomite and magnesite in contrast to other Ca-Albearing phases such as pumpellyite, prehnite, and zeolites.
The zeolites stilbite, laumontite, and wairakite are stable in the compositions with no CO 2 (Figs. 1a and 1b). Their stability is reduced significantly in compositions with only 1 wt% CO 2 , where stilbite is absent at all oxidation states. Laumontite is absent in the composition with divalent iron and present in a limited P-T range in the composition with trivalent iron, while wairakite is still present in both compositions (Figs. 1c and 1d). At 2 wt% CO 2 (Figs. 1e and 1f), zeolites are absent. Both the 20 and 13°C km À1 geotherms predominantly pass through the laumontite stability field, which is therefore the only stable zeolite occurring in~≤12 vol% at conditions considered in Fig. 2. Laumontite decreases rapidly with an increase in CO 2 and is not stable above 0.22 mol kg À1 (~0.97 wt%) CO 2 .
Calcite forms as soon as CO 2 is added to the fluid composition and occurs in <5 vol% together with laumontite, prehnite, and pumpellyite. At 0.44 mol kg À1 CO 2 , dolomite replaces calcite at <6 vol %. Magnesite forms at~0.6 mol kg À1 (~2.7 wt%) CO 2 and increases in abundance to~10 vol% with CO 2dominant fluids. Actinolite is a ubiquitous phase (≤45 vol%) in all modeled conditions and is only absent where the fluid is pure CO 2 (Fig. 2). Chlorite is present at lower CO 2 in ≤18 vol% and is not stable above~0.77 mol kg À1 (3.42 wt% CO 2 ) on both geotherms and oxidation states (Fig. 2). Micas are present in low amounts <5 vol % at the conditions considered (not shown in Fig. 2 but listed in Table S1 in supporting information). Stilpnomelane is present at low temperatures and low CO 2 . Biotite is present at higher temperatures and CO 2 but in low amounts of <1 vol%. White mica (muscovite) is stable at higher pressures and would be expected on colder geotherms of 10°C km À1 (Fig. 1). Lawsonite is present in the phase diagrams at high pressures, low temperatures, and low CO 2 (Figs. 1a and 1b) but is not expected to occur on Martian geotherms.
Feldspars are present at all modeled pressuretemperature conditions with the K-feldspar solid solution (of predominantly albite composition) stable at lower temperatures and plagioclase at higher temperatures and higher CO 2 (Table S1). K-feldspar abundance is not significantly influenced by an increase in CO 2 in the fluid composition and is present up tõ 17 vol% (not shown in Fig. 2). Plagioclase is stable above~0.8 mol kg À1 (3.5 wt%) CO 2 on both geotherms and increases with an increase in CO 2 up to 24 vol%.
Quartz is present in relatively small amounts of 4 vol% at H 2 O-rich conditions but increases with an increase in CO 2 to~11 vol% (Fig. 2). Clinopyroxene is stable at low CO 2 and again at high CO 2 with up to 30 vol% (Fig. 2). Orthopyroxene forms at high CO 2 of 0.98 mol kg À1 (4.3 wt%) in ≤9 vol% (not shown in Fig. 2). Titanite, ilmenite, magnetite, and a fluid phase are present as accessory phases in ≤2 vol% (see Table S1). Figure 3 shows mineral modes in Bounce Rock at 200°C on the 20 and 13°C km À1 geotherms as a function of X(CO 2 ) at fluid-saturated conditions assuming the fluid phase components H 2 O and CO 2 are always present in sufficient quantity to saturate the system. In calculations for the 20°C km À1 geotherm (Fig. 3a), epidote and clinopyroxene are only stable at very low X(CO 2 ). Laumontite is stable up to X(CO 2 ) 0.00063 (~4.7 wt% H 2 O and 6.7 wt% CO 2 or 1.5 mol kg À1 in the bulk composition). Above this value, the assemblage is dominated by quartz (~35 vol %), calcite (~21 vol%), chlorite (~18 vol%), and muscovite (~14 vol%). K-feldspar is also present iñ 4 vol%. Additional phases include ilmenite and rutile (not shown in Fig. 3). On the 13°C km À1 geotherm (Fig. 3b), laumontite is stable to X(CO 2 )~0.00028 and the assemblage at high X(CO 2 ) is dominated by quartz, calcite, chlorite, and muscovite.

Mineral Stabilities in Ultramafic Compositions with Variation in H 2 O-CO 2
Mineralogy of the Ultramafic Composition with 5 wt% Bulk Fluids Figure 4 shows mineral modes as a function of bulk CO 2 (mol kg À1 ) for the ultramafic composition of ALH 77005 (  (Figs. 4b and 4d). Antigorite (serpentine) is a dominant phase (≤36 vol%) for all modeled conditions at low CO 2 and is not stable above~0.26 mol kg À1 (1.2 wt%) CO 2 on the 20°C km À1 geotherm and 0.33 mol kg À1 (1.4 wt%) CO 2 on the 13°C km À1 geotherm. Talc is stable in the range betweeñ 0.05 mol kg À1 (0.2 wt%) and~0.62 mol kg À1 (2.7 wt %) CO 2 on the 20°C km À1 geotherm and betweeñ 0.09 mol kg À1 (0.4 wt%) and 0.96 mol kg À1 (4.2 wt%) CO 2 on the 13°C km À1 geotherm. Magnesite is the first and only carbonate forming in all compositions at 0.14 mol kg À1 (~0.6 wt%) CO 2 , increasing in abundance with CO 2 up to~11 vol% on the 20°C km À1 geotherm, and at~0.34 mol kg À1 (1.5 wt %) CO 2 on the 13°C km À1 at ≤9 vol% in a pure CO 2 fluid. Actinolite (≤32 vol%) is present at most conditions except at pure CO 2 . Chlorite abundance increases up to~18 vol% at lower CO 2 but decreases at higher CO 2 and is not stable above~0.96 mol kg À1 (4.2 wt%) CO 2 on the 20°C km À1 geotherm and 0.72 mol kg À1 (3.2 wt%) CO 2 on the 13°C km À1 geotherm. Olivine is a stable phase at all conditions but is less abundant where talc is present. Clinopyroxene is present at CO 2 -poor and CO 2 -rich conditions. Orthopyroxene forms at~0.57 mol kg À1 (2.5 wt%) CO 2 on the 20°C km À1 geotherm and reaches~39 vol% at maximum CO 2 . On the 13°C km À1 geotherm, orthopyroxene forms at ≥0.73 mol kg À1 (3.2 wt%) CO 2 at~50 vol% at maximum CO 2 . Spinel forms at higher CO 2 on the 20°C km À1 geotherm and is present at all conditions on the 13°C km À1 geotherm. Ilmenite and magnetite (not shown in Fig. 4) are also stable at most conditions. K-feldspar (not shown) only forms in CO 2rich conditions. Figure 5 shows mineral modes of the ALH77005 composition at 200°C on the 20 and 13°C km À1 geotherms as a function of X(CO 2 ) at fluid-saturated conditions. On the 20°C km À1 geotherm, calcite forms at low X(CO 2 ) but is replaced by dolomite at X(CO 2 ). Antigorite is not stable above~0.0035 X(CO 2 ) or 1.06 mol kg À1 (4.6 wt%) CO 2 in the bulk composition, and the assemblage is dominated by talc (~53 vol%), Fig. 4. Modes (vol%) for selected minerals in the ultramafic starting composition of ALHA77005 with 5 wt% of fluid varying from 0 to 5 wt% CO 2 plotted as a function of CO 2 (mol kg À1 ) in the bulk composition at 200°C on geotherms of 20°C km À1 (a, b), and 13°C km À1 (c, d) assuming only divalent iron (a, c) and 2. 23 wt% Fe 2 O 3 (b, d). Act = actinolite; Atg = antigorite; Chl = chlorite; Cpx = clinopyroxene; Mgs = magnesite; Ol = olivine; Opx = orthopyroxene; Spl = spinel; Tlc = talc. (Color figure can be viewed at wileyonlinelibrary.com.) chlorite (~17 vol%), and dolomite (~10 vol%) with lesser amounts of actinolite (~6 vol%) and magnesite (~4 vol%).

Limitations of the Model
Our calculations assume thermodynamic equilibrium, which may not be achieved everywhere at lowtemperature conditions with slow reaction rates on Mars. For instance, the carbonate assemblages in nakhlite meteorites are likely metastable (e.g., Bridges et al., 2019). However, we consider these phase equilibria calculations as good approximations to model subsurface low-grade metamorphic processes on Mars, since we assume a relatively closed system at depth where fluids are present in the pore space and can, therefore, react with the rocks over geological time scales. It is very likely that Martian basaltic rocks preserve their textures at low-grade metamorphic conditions as observed in terrestrial metabasic volcanic rocks, where alteration phases form in veins and vesicles that are more accessible to fluids (e.g., Cho & Liou, 1987;Starkey & Frost, 1990). Nevertheless, it can be assumed that at least locally the fluid is equilibrated with the rock to form characteristic alteration phases, but these textural relationships may explain why our models predict abundant actinolite, while it is not observed in most spectral studies (see Semprich et al., 2019, for a detailed discussion of this issue).
Our models specifically address the influence of CO 2 on mineral stabilities, and we therefore did not consider Cl, F, or S in our calculations. High-temperature Martian fluids are enriched in Cl as indicated by Clbearing phases in meteorites (e.g., Filiberto et al., 2014;Viennet et al., 2020), and low-grade metamorphic assemblages may therefore also be influenced by these elements. However, thermodynamic data for minerals including halogens are scarce and therefore currently not available in the thermodynamic database for the phases of importance for low-grade metamorphic conditions. Subsurface fluids on Mars may also likely increase in salinity if they are comparable to deep waters on Earth (M€ oller et al., 1997;Nurmi et al., 1988), which is not considered in our calculations. We also do not consider the oxides MnO, P 2 O 5 , and Cr 2 O 3 in the starting composition, which are relatively low and are therefore not expected to significantly influence the calculated mineral assemblages although they could cause minor changes in phase stability fields and mineral abundances.

Effect of Oxidation State on Mineral Stability
In addition to the protolith composition, the oxygen fugacity during metamorphism is also influenced by the fluid, adding further uncertainties. A large range of oxygen fugacities has been reported for Martian igneous rocks (e.g., Herd, 2006;Schmidt et al., 2013). Therefore, we cover most of the range of possible Martian oxygen fugacities by assuming two endmember compositions with only divalent iron and 10% of the total iron as trivalent iron. Uncertainties may arise due to limited thermodynamic data on Fe 3+ members or the lack of solid solution models including Fe 3+ endmembers, which is discussed for prehnite below.
In Bounce Rock, changes in oxidation state mainly affect the stability of prehnite, pumpellyite, and epidote with a decrease in the prehnite stability field toward lower pressures and temperatures in the compositions including trivalent iron (Figs. 1 and 2). While pumpellyite and epidote can be modeled with solid solutions including Fe 3+ endmembers, prehnite is assumed to be a pure phase and therefore does not allow for Fe 3+ substitution in our calculations. As a result, the stability fields of pumpellyite and epidote are increased at more oxidizing conditions. However, in natural prehnite, Fe 3+ can replace octahedral Al according to Ca 2 Al 1-x Fe x (AlSi 3 O 10 )(OH) 2 , where x denotes the amount of moles of substitution. Fe 3+ -rich prehnite occurs in hydrothermal settings with mole fractions of Fe 3+ as high as~0.6 (Freedman et al., 2009;Nagashima et al., 2018;Wheeler et al., 2001). While the overall effect of additional components such as Fe 3+ on modeled prehnite stability is difficult to assess, they are likely not as significant as for phases such as pumpellyite and epidote since the Fe 3+ endmember for prehnite is usually estimated to be~5% (Liou et al., 1983).
Experimental studies show that prehnite and epidote become more enriched in Al with a decrease in oxygen fugacity and an increase in temperature (Liou et al., 1983). It is therefore likely that the prehnite stability in our models in compositions with Fe 2 O 3 is underestimated. Since there is likely less Fe 3+ substitution in prehnite than in pumpellyite, more reducing conditions would likely be more favorable for prehnite stability. Low oxygen fugacities during metamorphism could result in the formation of prehnite instead of pumpellyite, which may explain the absence of pumpellyite detections in spectral studies from orbit. However, the current nondetection of pumpellyite in visible and near-infrared spectra could also be due to its spectrum being similar to Mg-rich chlorite (Ehlmann et al., 2009;Viviano et al., 2013), and pumpellyite may still be present on Mars as suggested by its detection using radiative transfer modeling on OMEGA reflectance spectra .
Since the solid solution models for antigorite and talc (Table 1) contain no endmembers with trivalent iron, the ultramafic composition is less sensitive to oxidation state and there are no major changes in mineralogy although minor differences are recorded in mineral stability and abundances (Fig. 4).

Effect of CO 2 on Mineral Stabilities and Implications for Mars
Our calculations show that even small amounts of CO 2 during low-grade metamorphism change the mineralogy significantly. In the modeled basaltic compositions with limited fluid availability, zeolites, prehnite, and pumpellyite are only stable at low CO 2 , while epidote is stable at higher amounts of CO 2 together with calcite and dolomite. In the modeled fluid-saturated compositions, Ca-Al silicates are replaced by an assemblage dominated by quartz, calcite, chlorite, and muscovite, which is consistent with studies on Earth, where prehnite and the zeolites stilbite and laumontite form at low XCO 2 from basaltic protoliths (Digel & Gordon, 1995;Thompson, 1971). Epidote and wairakite, however, can be stable at higher CO 2 partial pressure (e.g., Gianelli et al., 1998). In regions on Mars where prehnite has been detected, fluids could therefore not have contained significant amounts of CO 2 . This may be due to initially low CO 2 in Martian subsurface fluids that have not been in contact with a CO 2 -rich atmosphere. It is also possible that the fluids were in contact with a CO 2 -poor or "thin" atmosphere (Manning et al., 2006), although a recalculation of the crustal carbonate abundances that have been detected as pCO 2  is in favor of a "thick" atmosphere model (Manning et al., 2006). Alternatively, prehnite may have formed as a result of CO 2 degassing within a hydrothermal system from hot alkali chloride waters as has been suggested for the formation of hydrothermal prehnite on Earth (Wheeler et al., 2001). Chloride-rich fluids have been proposed for Mars based on minerals in melt inclusions in nakhlite meteorites (Filiberto et al., 2014;Giesting & Filiberto, 2016;McCubbin et al., 2013;Viennet et al., 2020). While residual melts are expected to be enriched in halogens compared to the bulk rocks, degassing at shallow depths may release halogens (e.g., Villemant & Boudon, 1999;Wang et al., 2014) and result in chlorine-rich hydrothermal fluids which could have precipitated the low-grade metamorphic assemblages on Mars (e.g., Filiberto et al., 2014).
The carbonates associated with low-grade metamorphic phases may also have formed at the same pressure-temperature conditions from CO 2 -rich fluids before degassing or during periods of higher volcanic activity which could have released magmatic gases and increased the CO 2 in the hydrothermal fluids (e.g., Griffith & Shock, 1995). Alternatively, the carbonates could have formed by later low-temperature alteration either in the shallow subsurface or close to the surface (e.g., Viviano et al., 2013). Higher amounts of carbonates could have potentially formed during early Mars and later been partially dissolved by acidic weathering (Ehlmann et al., 2009) or replaced by clays and oxides in the presence of later circumneutral fluids Piercy et al., 2019). Until better constraints on textural and geological relationships are available, it is not possible to determine the specific process most likely for carbonate formation. Once the samples drilled by Mars 2020 Perseverance are returned to Earth, the mineralogical relationships can be compared to our models to better constrain carbonate formation in Jezero crater.
For the ultramafic model compositions, CO 2enriched fluids result in assemblages dominated by talccarbonate, and quartz-carbonate replacing serpentine. Studies analyzing Compact Reconnaissance Imaging Spectrometer for Mars mineral spectra have suggested that talc could be associated with carbonates detected in the Nili Fossae area (Brown et al., 2010;Viviano et al., 2013). Based on Earth analogs, the presence of talc likely indicates hydrothermal alteration of olivine-rich rocks (e.g., Bjerga et al., 2015;Brown et al., 2005). The fact that olivine seems to be co-located with the talccarbonate unit could suggest that the alteration reaction was limited by the availability of either fluids or heat (Brown et al., 2010). Our models suggest that serpentine, olivine, talc, and very small amounts of magnesite could coexist at ≤0.33 mol kg À1 CO 2 when the amount of fluid is restricted to 5 wt% (Fig. 4) and could, therefore, have formed during a single alteration event. Alternatively, the talc and carbonate assemblage could have formed by subsequent carbonation of serpentine as suggested by Viviano et al. (2013) implying at least two alteration events with different fluid compositions. Similar processes on small scale could explain the mixed-layer serpentine-talc detected by Curiosity, which were likely transported into Gale crater as sedimentary detritus from ultramafic source rocks .
In our models, assemblages dominated by quartz and carbonates are predicted to form from basaltic and ultramafic conditions at X(CO 2 ) of >0.0005 and >0.0035, respectively (Figs. 3 and 5). The alteration of basaltic and ultramafic rocks by CO 2 -bearing fluids at hydrothermal or low-grade metamorphic temperatures (>150°C) could hence produce silica-rich rocks in the Martian subsurface. Although silica-enriched deposits have been reported by the Curiosity rover at Gale crater (Frydenvang et al., 2017;Morris et al., 2016;Rampe et al., 2017;Yen et al., 2017), by the Spirit rover at Gusev crater (Squyres et al., 2008), and by spectral data from orbit (e.g., Milliken et al., 2008), these are mainly composed of amorphous silica, opaline silica, or tridymite and cristobalite rather than quartz, which is the dominant SiO 2 phase in our models. The differences in mineralogy suggest that the observed deposits formed either during low-temperature (<80°C), diagenetic conditions (Frydenvang et al., 2017), by hydrothermal activity potentially in a hot spring environment (Ruff & Farmer, 2016;Squyres et al., 2008;Yen et al., 2021), by chemical weathering  or, contentiously, as detrital accumulations with volcanic precursors (Morris et al., 2016).

Implications for Habitability
While the temperatures considered in our models are too high for the survival of microbial life, deep, metamorphic fluids could either circulate within the crust and be transported to shallower depths and cooler temperatures or remain trapped in the host rock as they cool. Prerequisites for microbial life are a supply of carbon, hydrogen, nitrogen, oxygen, phosphorus, sulfur, specific micronutrients, and an energy source (Fisk & Giovannoni, 1999). While we have not included N, P, and S in our models, they are usually present in low concentrations in Martian meteorites and soils Filiberto et al., 2019;Franz et al., 2017;Grady et al., 1995;Greenwood et al., 2003;Greenwood & Blake, 2006;Kounaves et al., 2014;Nachon et al., 2017;Stern et al., 2015) and can be released during fluid-rock reactions and transported by hydrothermal fluids (Ehrlich, 1996). Where CO 2 -bearing fluids are absent, the carbon required for biomass could have been provided by magmatic carbon in basalts, as reported on the Martian surface and in meteorites (Ming et al., 2014;Sephton et al., 2002;Steele et al., 2012).
Deep subsurface environments are expected to be anoxic and oligotrophic, restricting metabolisms to anaerobic respiration and fermentation (Lovley & Chapelle, 1995). On Earth, subsurface lithoautotrophic microbial ecosystems use a diversity of metabolic mechanisms for energy generation, including H 2 oxidation, methanogenesis, anaerobic methane oxidation, iron and sulfur redox transformation, and nitrate reduction (Nealson et al., 2005;Sar et al., 2019;Stevens & McKinley, 2000). In these environments, the fixation of carbon dioxide is a key metabolic process that drives production. Carbon fixation can be achieved by multiple pathways, including the Wood-Ljungdahl, or reductive acetyl-CoA, pathway (Ragsdale & Pierce, 2008), which has been suggested as one of the earliest metabolisms on Earth (Braakman & Smith, 2012;Fuchs, 2011;Nitschke & Russell, 2013) and is employed by acetogenic bacteria, methanogens, and archaeal sulfate reducers (Smith et al., 2019). Microbes with these metabolisms have been shown to inhabit extreme environments close to the thermodynamic limits of life (Berg, 2011) and could hypothetically survive in the Martian subsurface (Cousins, 2015;Cousins & Crawford, 2011;Nixon et al., 2013).
Sulfate-reducing bacteria reverse the Wood-Ljungdahl pathway to generate metabolic energy by coupling the oxidation of acetate to H 2 and CO 2 and using sulfate as a terminal electron acceptor, reducing it to hydrogen sulfide (Schauder et al., 1988;Spormann & Thauer, 1988).
On Earth, molecular hydrogen serves as an energyrich electron donor for chemolithoautotrophic microbial communities in the subsurface (Nealson et al., 2005;Reith, 2011;Stevens & McKinley, 2000) and could therefore also fuel similar communities on Mars. Possible geogenic hydrogen sources in the Martian subsurface could be radiolysis of water (Lin et al., 2005), friction during seismic events (Kita et al., 1982;Lippmann-Pipke et al., 2011) or at the base of ice sheets (Hirose et al., 2011;Telling et al., 2015), and serpentinization (Hellevang et al., 2011;Onstott et al., 2019;Schulte et al., 2006;Vance & Melwani Daswani, 2020;Westall et al., 2013). Radiolysis has the potential to yield production rates of H 2 comparable to those on Earth due to the increased porosity space at depth (Dzaugis et al., 2018;Tarnas et al., 2018); this process would also create potential electron acceptors, including sulfate via oxidation of sulfides (Lefticariu et al., 2006;Li et al., 2016), which can create an environment to sustain thermophilic sulfate reducers (Chivian et al., 2008;Lin et al., 2006). Serpentinization also produces H 2 , particularly at temperatures of 200-315°C (McCollom & Bach, 2009). However, a consequence of serpentinization is an increase in pH of the reacted fluids (>9), which can play a role in reducing microbial diversity in terrestrial serpentinization settings (Brazelton et al., 2010;Rempfert et al., 2017;Schrenk et al., 2004). The released geogenic H 2 can either react with aqueous or gaseous carbon or with transition metal sulfide catalysts via Fischer-Tropsch-type reduction to produce methane and a variety of hydrocarbons and organic compounds (Jones et al., 2010;McCollom, 2016;Proskurowski et al., 2008;Sherwood Lollar et al., 2006. Abiogenically produced methane and hydrocarbons could potentially diffuse upward and be used by heterotrophic microbes, such as anaerobic methane oxidizers (Marlow et al., 2014;Purkamo et al., 2015). Serpentinite-hosted ecosystems and environments enriched in H 2 that have a high pH on Earth are typically dominated by hydrogen-oxidizing microaerophilic Betaproteobacteria and anoxic, fermenting Clostridia, indicating the viability of these metabolisms under serpentinizing conditions (Brazelton et al., 2010;It€ avaara et al., 2011;Moser et al., 2005;Schrenk et al., 2013).
The shallower Martian subsurface, where temperatures are expected in the range of 10-100°C, could potentially host functional groups shown to occupy neutrophilic and iron-rich environments, such as iron-oxidizing bacteria and archaea, which have been reported to inhabit various terrestrial environments including freshwater (e.g., Gallionella ferruginea, Leptothrix, Sideroxydans), as well as in marine hydrothermal systems (Mariprofundus ferrooxydans; Edwards et al., 2004;Emerson & Moyer, 2010;. Iron-oxidizing microbes have also been identified (e.g., Pseudomonas sp. HerB) at the basalt-ice boundary in lava tube caves using olivine as energy source in low O 2 conditions (Popa et al., 2012).

CONCLUSIONS
Our phase equilibria models using basaltic and ultramafic starting compositions and H 2 O-CO 2 -bearing fluids show that characteristic low-grade metamorphic minerals such as prehnite, zeolite, and serpentine that have been observed spectrally on Mars are only stable at low CO 2 . Higher concentrations of CO 2 will result in assemblages dominated by quartz, calcite, chlorite, and muscovite in the basaltic composition, and by talc-chlorite-dolomite, or magnesite-quartzchlorite-dolomite assemblages in the ultramafic composition. This could imply that the metamorphic fluids were initially CO 2 -poor, or the phases were precipitated after the fluids lost CO 2 by degassing, or carbonates and low-grade metamorphic minerals formed during separate alteration events. Carbonates and talc that have been reported together with metamorphic phases, such as serpentine, could have either formed at similar low-grade metamorphic conditions in the presence of a different, CO 2 -bearing, fluid composition in the crust. Alternatively, they could have formed by lower temperature surface or shallow subsurface alteration processes when the initial low-grade metamorphic assemblage was exposed to the CO 2 -rich atmosphere. reviews and comments, which greatly improved the manuscript, and Dr. Gordon Osinski for editorial handling. Profound thanks go to Dr. Michael Macey for his comments on the habitability section. JS and SPS acknowledge funding from Research England Expanding Excellence in England (E3) fund (grant code 124.18).
Data Availability Statement-The data that supports the findings of this study are available in the supplementary material of this article.