High‐pressure metamorphic mineralogy of the Martian crust with implications for density and seismic profiles

Here, we calculate the mineralogy of the Martian lower crust and upper mantle as a function of pressure and temperature with depth using four bulk compositions (average crust, Gusev basalt, olivine‐phyric shergottite, and primitive average mantle). We then use this mineralogy to extract rock properties such as density and seismic velocities, describe their changes with varying conditions and geotherms, and make predictions for the crust–mantle boundary. Mineralogically, all compositions produce garnet, orthopyroxene, clinopyroxene in varying proportions at high pressures, with differences in minor minerals (spinel, ilmenite, rutile, and/or K‐feldspar). According to our calculations, the average crust and Gusev basalt compositions have the potential to yield higher densities than the average mantle composition, particularly for thicker crusts and/or colder geotherms. Therefore, recycling of the Martian crust into the mantle could occur through the process of crustal delamination, if not kinetically inhibited. However, our results show that, depending on crustal thickness, the crust may not be easily distinguishable from the mantle in seismic properties.


INTRODUCTION
Investigating the alteration history of the Martian crust has been a major research and exploration focus because alteration minerals can record the history of past fluids that could have been a habitable environment. However, most of the focus has been on low temperature fluids and the minerals produced at near surface conditions. Understanding the alteration history and mineralogy of the lower crust and upper mantle is critical to unravel the internal structure of Mars and processes at the crust-mantle boundary. Therefore, here we calculate the mineralogy of the lower crust and upper mantle as a function of increased pressure and temperature with depth and use that mineralogy to constrain variations in density and seismic properties.
Secondary mineralogy records the P-T-X fluid conditions of alteration. Alteration minerals have been found in the Martian meteorites, at every landing site, and in the crust from orbit (see reviews in Filiberto and Schwenzer 2019). Extensive thermochemical calculations have been conducted to investigate the range of conditions of alteration providing evidence for mostly lower temperature fluids (<200°C) and from nearneutral to acidic fluids (e.g., Griffith and Shock 1997;McAdam et al. 2008;Schwenzer and Kring 2013;Zolotov and Mironenko 2016).
Recent work has expanded investigations of alteration minerals to low-grade metamorphic conditions. Evidence for hydrothermal and low-grade metamorphism comes from both spectral investigations of the Martian crust and detailed analyses of Martian meteorites. Filiberto et al. (2014) found evidence for high temperature fluids in the form of marialite (Na-Cl scapolite) in a melt inclusion in Nakhla, which is consistent with formation from Cl-rich, water-poor fluid (either magma or brine) at a minimum temperature of 700°C. Further evidence for Cl-rich hydrothermal conditions has been found in other nakhlites in the form of Cl-potassic-hastingsite (e.g., Giesting and Filiberto 2016), which forms in medium-grade metamorphic conditions (>400°C). Finally, recent investigations of meteorite NW 7533 found eskolaite encased in chromite-magnetite and suggested that it formed from an active hydrothermal environment in the Martian crust (Liu et al. 2016). From orbit, smectite, chlorite, prehnite, phyllosilicates, the zeolite analcime, opaline silica, and serpentine have been detected (e.g., Ehlmann and Edwards 2014). Some of the mineral assemblages require elevated temperatures of formation (McSween et al. 2015). McSween et al. (2015) and Semprich et al. (2019) used metamorphic phase diagrams and phase equilibria models to constrain formation conditions of the minerals detected from orbit. Their results replicate the low-grade metamorphic mineral phases that have been detected on the Martian surface, which could have either formed by impact or magmatic hydrothermal activity or along a somewhat higher geothermal gradient than for present-day Mars.
Better constraints on the composition and processes in the Martian lower crust are crucial to understand the internal structure and for the determination of crustal thickness and evolution by geodynamic models. However, only one paper has experimentally investigated higher metamorphic grade at one single P-T-X condition. Papike et al. (2013) experimentally produced an eclogite from a QUE 94201 bulk composition at 4 GPa and 1200°C. The mineralogy was 54% omphacite, 35% garnet, 6% ilmenite, 4% SiO 2 -phase, and trace rutile and phosphate (Papike et al. 2013). While this study shows that an eclogitic assemblage can be produced from Martian compositions, the experimental P-T conditions exceed those estimated at the crust-mantle boundary (~0.5 to 2 GPa at 800-1000°C; Babeyko and Zharkov 2000), even when assuming a crustal thickness exceeding 100 km. Only one study calculated the mineralogy, density, and seismic properties of Martian meteorite basalts up to depths corresponding to the crust-mantle boundary assuming that the shergottites represent the average crust (Babeyko and Zharkov 2000). Their results produce a very dense, overly thick crust for most geotherms; however, recent work has shown that the shergottites do not represent average crust (e.g., Filiberto 2017) and the results from these calculations likely do not represent mineralogy or physical properties of the Martian crust. Specifically, Martian meteorites have higher MgO and FeO and lower Na 2 O and K 2 O contents than the average crust (e.g., Filiberto 2017), which would produce an overly dense crust if assumed to represent the entire crust. Furthermore, Babeyko and Zharkov (2000) determine the crustal thickness from the condition that densities at the crustmantle boundary range between 3450 and 3500 kg m À3 . However, neither the crust nor the mantle (Bertka and Fei 1998) is expected to reach densities of 3500 kg m À3 at pressures ≤1 GPa corresponding to crustal thickness estimates~100 km (e.g., Wieczorek and Zuber 2004;Baratoux et al. 2014).
Therefore, there are only few constraints on highgrade metamorphism of lower crust/upper mantle, which are important for our understanding of the Martian interior such as crustal thickness and processes at the crust-mantle boundary. Furthermore, better constraints on the properties of the Martian lower crust and upper mantle as a function of compositions and P-T conditions will be vital for understanding and interpreting the seismic data returned from the recently landed InSight mission (Smrekar et al. 2018). Here, we expect changes in texture and mineralogy and therefore metamorphic reactions within the Martian crust due to an increase in pressure and temperature with depth by burial without invoking any plate tectonic processes such as subduction or orogeny. Since the mineral assemblages are calculated by Gibbs free energy minimization at any given pressure and temperature and therefore without assuming any metamorphic history, our results are also relevant for igneous rocks that have intruded at depth resulting in high-pressure mineral assemblages (e.g., Medaris et al. 2006Medaris et al. , 2018M€ untener and Ulmer 2006). We use the Perple_X software suite (Connolly 2005) to calculate the metamorphic mineralogy of the Martian lower crust and upper mantle for four bulk compositions. We then adopt this mineralogy to extract rock properties such as density and seismic velocities (v s and v p ), describe their changes with P-T-X, and discuss implications for the crustmantle boundary and how our results can be useful for geodynamic and geophysical models and the interpretation of seismic data.

Bulk Compositions
To investigate a range of bulk compositions (Table 1), we chose the average Martian crust (Taylor and McLennan 2009), the surface basalt analyzed at Gusev Crater: Humphrey (McSween et al. 2006), Martian meteorite Yamato 980459 a primitive, mantlederived, olivine-phyric shergottite (an average of data calculated by mineral mode and electron microprobe analyses by Greshake et al. [2004]; and wet chemical analyses from Misawa [2004]), and a calculated mantle composition (Taylor 2013). From here on, these will be referred to as: average crust, Gusev basalt, shergottite, and average mantle, respectively (Table 1). These provide a large range of Mg# to explore. We assume water-free conditions in our calculations (1) to be comparable with experimental work on high-pressure phase relations (Papike et al. 2013); (2) based on studies of the terrestrial crust, pore water is typically squeezed out of the system by compaction and diagenesis (e.g., Heier 1965); and (3) eclogite facies do not typically contain hydrous phases and are relatively anhydrous with between 3000 and 300 ppm H 2 O (e.g., Katayama et al. 2006). We assume all iron as Fe 2+ . The oxygen fugacity of the Martian mantle, as exemplified by the shergottites, is around QFM À3 to +3 (Herd 2003;Righter et al. 2008), which would equate to <5% Fe 3+ . For Humphrey, similar values in the range of À2.3 to À3.6 have been obtained (Schmidt et al. 2013).

Phase Equilibria Modeling
We followed the same approach from Semprich et al. (2019) for using the Gibbs free energy minimization software Perple_X 6.8.6 (Connolly 2005) and an internally consistent thermodynamic dataset Powell 1998, 2011) to calculate metamorphic phase diagrams. We computed phase equilibria for four starting compositions (see next section) in the range 0.1-3 GPa (equivalent to~10 to 255 km depth) and 500-1100°C. Mn and P were not considered due to their relatively low abundances and/ or incomplete set of solid solution models. Cr was considered in the shergottite and average mantle compositions, while K has only been added to the average Martian crust but is negligible for all other compositions. Furthermore, Ti was not considered for the average mantle composition due to its low abundance. For the average crust and Gusev basalt compositions, we used the following solid solution models: olivine (Holland and Powell 1998), omphacite (Green et al. 2007), garnet and orthopyroxene (White et al. 2014), feldspar (Fuhrman and Lindsley 1988) for calculations with K, plagioclase (Newton et al. 1980) for calculations without K, spinel (White et al. 2002), and ilmenite (White et al. 2000). A similar set of solid solution models was used for the shergottite except for clinopyroxene (Holland and Powell 1996), garnet, and spinel (Jennings and Holland 2015). For the average mantle composition, the following solid solutions were used: olivine, orthopyroxene, garnet, spinel (Jennings and Holland 2015), clinopyroxene (Holland and Powell 1996), and plagioclase (Newton et al. 1980).
Densities as well as seismic velocities were extracted from the phase diagrams using the Perple_X application werami according to the approach described by Connolly and Kerrick (2002). Herewith, the compressional-(v P ) and shear-wave (v S ) velocities are described as a function of three thermodynamic properties: the density (q), the adiabatic bulk modulus (K S ), and the shear modulus (µ) in the following equations for an isotropic, homogeneous material (Bina and Helffrich 1992;Karki et al. 2001): Both the density and the adiabatic bulk modulus can be expressed as a function of Gibbs free energy: with N being the molar formula weight and G the molar free energy (Connolly and Kerrick 2002). The shear modulus for all relevant phases is computed as 3K S (1À2m)/(m+1)/2 where the Poisson ratio m is set to 0.35. Due to the lack of detailed textural information, the adiabatic bulk and shear moduli are estimated as the weighted average of upper and lower bounds on the moduli. The Voigt-Reuss-Hill (VRH) method (e.g., Berryman 2013) is used in our calculations where the lower bound (Z-) on the modulus Z of an aggregate is calculated as the harmonic mean on the corresponding moduli on its constituent phases (p):

.p]) and the upper bound (Z + ) is given by
, where x i is the volume fraction of phase i. The aggregate modulus is then calculated as Z = (1 À v) 9 Z À + v 9 Z + , where the weighting factor v is 0.5 (e.g., Connolly and Kerrick 2002).

Physical Parameters of Mars
The upper crustal layer of 8-11 km is expected to show a high grade of porosity and/or alteration, which has recently been inferred by receiver function analysis of seismic data obtained by the seismometer of the InSight lander at Elysium Planitia (Lognonn e et al. 2020). Since our study focuses on consolidated rocks, we start our calculations at 10 km (~0.1 GPa) depth below the high porosity layer. Large quakes or located impacts that could provide constraints on the composition of the lower crust and upper mantle and the location of the crust-mantle boundary have not yet been detected on Mars (Lognonn e et al. 2020). Therefore, estimates of the mean Martian crustal thickness are based on indirect studies such as local relationships between gravity and topography or geochemical observations and range between 30 and 80 km (Zuber et al. 2000;Neumann et al. 2004;Wieczorek and Zuber 2004;Parro et al. 2017). However, crustal thicknesses of up to 100 km are consistent with global geophysical constraints (Sohl and Spohn 1997;Kavner et al. 2001;Gudkova and Zharkov 2004). An upper limit of 150 km may be possible based on the recalculation of the moment of inertia factor (e.g., Sohl et al. 2005) although viscous relaxation arguments suggest that the average crust does not exceed 100 km (Zuber et al. 2000;Nimmo and Stevenson 2001). While our calculated phase diagrams and extracted densities and seismic velocities cover depths in the range of~10 and 250 km, we also compared rock and mineral properties for all compositions at specific depths of 50 and 150 km at temperatures of 700 and 1000°C, relevant for cold and warm geotherms, respectively.
Due to the lack of in situ measurements, estimates of the surface heat flux of Mars are also based on indirect methods depending on several parameters that are currently not well constrained, such as abundance of heat-producing elements, topography, as well as crust and mantle thickness and densities (e.g., McGovern et al. 2002McGovern et al. , 2004Plesa et al. 2016;Parro et al. 2017). The present-day heat flux shows regional differences and is modeled in most regions to vary between 14 and 25 mW m À2 (Parro et al. 2017), 11-50 mW m À2 , and 17.2-49.9 mW m À2 (Plesa et al. 2016). These values correspond to a range of geotherms predominantly varying between 3 and 20 K km À1 ). In accordance with these findings, we assume 5, 7, 10, and 20°C km À1 as examples of possible geothermal gradients. Colder geotherms could be relevant for Pavonis Mons and Elysium Rise, while warmer geothermal gradients have been estimated for Noachis Terra and Northeastern Arabia Terra ).
To estimate depth from pressure, we use the following equation: h = P/qg, with a gravity (g) value for Mars of 3.727 m s À2 and the average crustal density (q) of 2900 kg m À3 , which is within the range of most Mars models (e.g., McGovern et al. 2004;Sohl et al. 2005;Plesa et al. 2016;Parro et al. 2017). Note that the actual density is not constant and varies as a function of pressure, and hence, the depths given in this work are only a rough approximation (Cammarano 2013).

RESULTS
Mineralogically, all samples are similar with a calculated low pressure assemblage of orthopyroxene, clinopyroxene, and olivine dominating the mineralogy with either plagioclase-rich feldspar (average crust composition due to addition of K) or plagioclase and ilmenite (average crust, Gusev basalt, shergottite) or spinel (shergottite and average mantle compositions) as accessory phases. Figure  Olivine proportions in the shergottite at low pressures are~10 vol% with Fa 40 (Fig. 1i) and~59 vol% in the average mantle composition with Fa 25 (Fig. 1m). The average crust and Gusev basalt contain 15-20 vol% clinopyroxene with Di 50-65 (lower values at higher temperature) in the average crust ( Fig. 1b) and Di 60-70 in the Gusev basalt (Fig. 1f). The shergottite contains 22 to 25 vol% clinopyroxene with Di 62-77 (Fig. 1j), while the average mantle contains~4 to 7 vol% clinopyroxene with Di 70-80 (Fig. 1n). Plagioclase in the average crust and Gusev basalt is present up to~50 vol% with An 30-34 and 9-12 vol% in the shergottite and average mantle with An 40-50 .
As the pressure increases in our modeled phase diagrams, the modal amounts of olivine and plagioclase in the average crust and Gusev basalt start to decrease and are being replaced by garnet and omphacitic clinopyroxene (Figs. 1a,1b,1d,1e,1f,1h). These reactions are shifted toward higher pressure with increasing temperature. In the Gusev basalt, shergottite, and average mantle, olivine is stable within all calculated pressure and temperature conditions but decreases with increasing pressure before slightly increasing again in the shergottite and average mantle toward the highest pressure considered (Figs. 1i and  1m). At higher pressures, the remaining plagioclase reacts out for Gusev basalt (Fig. 1g), while the remaining feldspar in the average crust is rich in sanidine component due to the K-bearing composition (Fig. 1c). Plagioclase in the shergottite and average mantle is first replaced by spinel and then by garnet as pressure increases (Figs. 1k, 1o, 1l, and 1p). The highpressure assemblage for all compositions contains garnet, orthopyroxene, and clinopyroxene in varying proportions. The average crust and Gusev basalt contain~45 to 50 vol% clinopyroxene with Di 0-20 , 20-28 vol% garnet with Alm 56-68 , and 14-27 vol% orthopyroxene at the highest pressure of 3 GPa. The shergottite shows 27-30 vol% clinopyroxene (Di 56-74 ), 15-20 vol% garnet (Alm 46-58 ), and 45-49 vol% orthopyroxene, while the average mantle has~11 vol% clinopyroxene (Di 50-60 ), 8-10 vol% garnet (Alm 34-50 ), and 21-24 vol% orthopyroxene. Olivine is stable in the Gusev basalt (~8 to 10 vol%), shergottite (~5 to 6 vol%), and average mantle (~56 vol%) at high pressures. The average crust composition contains feldspar (<4 vol%), ilmenite (<0.3 vol%), and rutile (<0.8 vol%) as accessory phases due to K and Ti being considered in the starting composition; the shergottite contains ilmenite and the average mantle composition shows spinel at the lowest temperatures in the phase diagram. This mineralogy is similar to the experimental work on a QUE 94201 bulk composition at 4 GPa and 1200°C, which produced 54% omphacite, 35% garnet, 6% ilmenite, 4% SiO 2 -phase, and trace rutile and phosphate (Papike et al. 2013). This mineralogy is also comparable to the results of Babeyko and Zharkov (2000) for basaltic shergottites with orthopyroxene, clinopyroxene, and garnet dominating the higher pressure assemblages and only minor amounts of quartz. Due to composition and model conditions, our calculated phase diagrams do not contain quartz at higher pressures.
Modeled densities range from 3100 to 3650 kg m À3 for all bulk compositions, but there are important differences of how density varies with pressure and temperature for each bulk composition (Fig. 2). In Table 2, we have extracted mineral and system properties including modal mineral proportions, densities, and seismic velocities for all compositions at specific conditions of 0.5 GPa and 700°C; 0.5 GPa and 1000°C; 1.55 GPa and 700°C; 1.55 GPa and 1000°C. The lower pressures approximately represent a 50 km thick crust assuming different temperature profiles, while the latter two conditions represent an overly thick crust of approximately 150 km (these depths are shown by the dashed red lines in Fig. 2).
The average crust and Gusev basalt compositions show the lowest densities at low pressures (Figs. 2a and 2b). This is due to the high modal amount of lowdensity feldspar/plagioclase (~2600 to 2650 kg m À3 , Table 2). At pressure conditions of 0.5 GPa (~50 km) and temperatures of 700 and 1000°C, respectively, the assemblage consists of 45-49 vol% feldspar/plagioclase resulting in rock densities in the range of 3100-3174 kg m À3 with variations in density mainly caused by changes in mineral composition (with Fe-bearing endmembers showing higher densities than Mg-bearing ones) and to a lesser extent by expansion and compression due to higher temperatures and pressures (Table 2). With increasing pressure, plagioclase and olivine react to form orthopyroxene, clinopyroxene, and garnet (Figs. 1a-h). This transition starts at higher pressures in the average crust compared to the Gusev basalt and causes significant densification in both compositions (Figs. 2a and 2b). At 1.55 GPa (~150 km) and both temperatures considered, the modal amount of garnet is 16-27 vol% with densities in the range of 3914-4029 kg m À3 (Table 2). At conditions where feldspar/plagioclase is either absent or only present in low proportions, rock densities of the compositions exceed 3500 kg m À3 (Table 2). Only the crustal composition at 1.55 GPa and 1000°C contains a significant amount of feldspar/plagioclase (~23 vol%) and hence yields a lower density of 3357 kg m À3 . Figure 2 also demonstrates that a hot geotherm of 20°C km À1 is located within the feldspar/plagioclaserich mineral assemblage for the average crust and Gusev basalt composition and hence will never exceed a density of~3200 kg m À3 . The 10°C km À1 geotherm, however, is partly (crustal composition) and completely (Gusev basalt) within the garnet stability field, but not yet in the area of significant densification. Even colder geotherms of 7 and 5°C km À1 are well into the garnet stability field and both compositions could show significant densification in regions where the crustal thickness exceeds 100 km. In areas of lower crustal thickness, the cold geotherms will not be hot enough (>700°C) for the rock compositions to equilibrate, and therefore, they will likely be preserved as a metagabbro with lower densities.
The shergottite and average mantle compositions are significantly denser (3320-3370 km m À3 ) at low pressures than the average crust and Gusev basalt (Figs. 2c and 2d; Table 2). However, due to their low Al 2 O 3 content in the whole rock composition, the amount of Al-bearing phases (plagioclase, spinel, and garnet) is significantly lower resulting in less significant densification with pressure (~3418 to 3493 kg m À3 at 1.55 GPa and both temperatures). Geotherms of 20 and 10°C km À1 are completely within the spinel stability field, while only the coldest geotherm of 5°C km À1 is within the garnet stability field.
Seismic properties (v P and v S ) follow the same pattern as mineralogy and density (v P in Fig. 3 and v S in Fig. S1 in supporting information; Table 2). All compositions have similar ranges in v P and v S but show significant differences in how seismic velocities change with pressure. The average crust and Gusev basalt compositions show a more significant increase in seismic velocities with increasing pressure from v P~6 .07-6.31 km s À1 (v S~2 .92-3.03 km s À1 ) at 0.5 GPa to v P~6 .68-7.31 km s À1 (v S~3 .21-3.41 km s À1 ) due to the formation of garnet, clinopyroxene, and orthopyroxene at the expense of plagioclase and olivine. The shergottite and average mantle compositions, however, show more gradual increases in both v P (6.38-6.82 km s À1 at 0.5 GPa; 6.77-7.12 km s À1 at 1.55 GPa) and v S (3.06-3.28 km s À1 at 0.5 GPa; 3.25-3.42 km s À1 at 1.55 GPa) with increasing pressure.
Subtracting densities of the average mantle composition from those of the average crust and Gusev basalt, respectively (Fig. 4), shows at which conditions the density of the average crust and the Gusev basalt exceed that of the average mantle. Both compositions are denser than the average mantle at pressures exceeding 1 GPa (~100 km) at 600°C, but the density crossover is shifted toward higher pressures with increasing temperatures. If the crust can get denser than the mantle depends on several factors including the crustal composition, crustal thickness, and the exact geotherm; thicker crusts and colder geotherms are more favorable for a density crossover.

Implications for Crustal Properties and the Crust-Mantle Boundary
The range of our calculated densities (3100-3650 kg m À3 ) for the lower crust overlaps with that obtained by Baratoux et al. (2014) derived from major element chemistry of Martian meteorites, in situ measurements by rovers, and from surface geochemistry with values in the range of 3100-3700 kg m À3 . However, our diagrams show that densities above 3300 kg m À3 for both the crustal and the basaltic compositions are only attained above~0.7 GPa corresponding to depths of~60 km. Density changes as a function of mineralogy and therefore with pressure and temperature. While lower temperatures are generally more favorable for higher densities since the transition to eclogite is shifted toward lower pressures, slow reaction kinetics will prevent mineral reactions below a certain temperature. Babeyko and Zharkov (2000) assume a "freezing" temperature for dry basalts at~700 to 800°C. As a result, the mineralogy of the rock will not change despite a decrease in surrounding temperatures. Calculating densities for Martian basalts at 1 bar and 25°C, as Baratoux et al. (2014) have done in their study, will therefore yield higher densities than the actual values. For example, measured grain densities (porosity free) of the Zagami basaltic shergottite range  Table 1). Dashed black lines represent approximate geotherms varying from 5 to 20°C km À1 . Dashed red horizontal lines represent crustal depths of 50 and 150 km, at which mineral and rock properties are extracted ( Table 2). The depth scale shown is approximated by the formula: h = P/qg, with a gravity (g) value for Mars of 3.727 m s À2 and an average crustal density (q) of 2900 kg m À3 . (Color figure can be viewed at wileyonlinelibrary.com.)          Baratoux et al. (2014) are above 3300 kg m À3 . While our calculations focus on conditions relevant for the lower crust, the low-pressure assemblage of orthopyroxene, clinopyroxene, olivine, feldspar/ plagioclase, and ilmenite is stable at pressures relevant to surface conditions. If the crust is~50 km thick, the density values will thus not exceed 3100 kg m À3 and will likely be reduced significantly in the upper few kilometers due to impact cratering and increased porosity. However, in areas where the crust exceeds 50 km, higher densities in the lower section of the lower crustal column are likely. Based on their observations, Baratoux et al. (2014) argue for an average crustal density above 3100 kg m À3 , which would be compatible with the mass of Mars and the moment of inertia factor if the average crustal thickness approaches 100 km. In contrast, the majority of geophysical models assume a conservative range of crustal densities within 2700-3100 kg m À3 resulting in crustal thickness estimates of 50 AE 12 km (e.g., Wieczorek and Zuber 2004;Sohl et al. 2005). The thickness of the crust significantly influences rock properties at the crust-mantle boundary. Our results suggest a density contrast at the crustmantle boundary of ≥200 kg m À3 at a crustal thickness of~50 km but <100 kg m À3 at a crustal thickness of 100 km (Fig. 4) with the possibility of a density crossover for basaltic compositions at colder temperatures.

Relevance to Mars
Martian models of crustal thickness and heat flux are dependent on numerous parameters, which are not well constrained and therefore vary considerably. Modeled crustal thicknesses vary predominantly between 30 and 80 km depending on regional  Table 1). (Color figure can be viewed at wileyonlinelibrary.com.) differences but can reach 100 km with an upper limit of 150 km (e.g., Zuber et al. 2000;Gudkova and Zharkov 2004;Neumann et al. 2004;Sohl et al. 2005;Parro et al. 2017). The present-day heat flux also shows considerable regional differences and is modeled to be in the range of 11-50 mW m À2 (e.g., McGovern et al. 2004;Plesa et al. 2016;Parro et al. 2017). Given this wide range of values, we have picked four regions as examples demonstrating how different crustal depths and geotherms can influence crustal densities: Capri Chasma in Valles Marineris, Noachis Terra, as well as Pavonis Mons and Arsia Mons in the Tharsis region. It is also important to note that the heat flux presently modeled in these areas may have changed over geological time and the crust may show different features than described below.
Capri Chasma is located in Valles Marineris, a large canyon system with an estimated crustal depth of 40 km ) and a thermal gradient of <7°C km À1 . The combination of a relatively shallow crust and a cold geotherm does not result in temperatures high enough to cause significant mineralogical changes in the average crust or Gusev basalt compositions until the crust-mantle boundary and there is no significant densification. Due to likely changes in textures with increasing overload pressure, we have labeled these rocks metagabbro (Fig. 5). The average mantle composition at 40 km is in the plagioclase and spinel stability field due to the cold geothermal gradient and will likely not equilibrate until high enough temperatures are reached at much greater depth.
Noachis Terra is an extensive southern landmass west of the Hellas impact basin. The average crustal depth is in the range of 60 km  with a geothermal gradient of >20°C km À1 ). Due to the hot geotherm, the rock compositions will always stay within the plagioclase stability field and no significant densification is expected (Fig. 5). The average mantle composition at the crustmantle boundary is also expected to be within the plagioclase and spinel stability field.
Pavonis Mons and Arsia Mons are both shield volcanoes in the Tharsis region with crustal depth estimates in the range of 80-100 km ) and geothermal gradients of >5°C km À1 for Pavonis and <10°C km À1 for Arsia . The combination of a thick crust and cold geotherm in the case of Pavonis Mons results in temperatures too low to equilibrate the rocks to higher grade mineralogy and therefore a lack of significant densification. A warmer geotherm as estimated for Arsia Mons combined with a crustal depth of up to 100 km, however, allows for pressures and temperatures high enough to change the mineralogy to garnet-bearing granulite. Igneous rocks intruding at these depths will also show this mineralogy. If the crustal depth exceeds 100 km at geotherms between 10 and 7°C km À1 , at least the Gusev basalt has the potential to become denser than the mantle.

Density Overturn and Delamination
Although the average crust and Gusev basalt compositions show a similar mineralogy, the Gusev basalt is enriched in MgO and CaO compared to the average crust, which results in higher modal proportions Fig. 4. Density difference between crust and mantle using the computed densities in Fig. 2. a) The density of the average mantle subtracted from the density of average crust; (b) the density of the average mantle subtracted from the density of the Gusev basalt composition. Negative numbers represent conditions where the mantle is denser than the crustal compositions. Positive numbers imply that the density of the crustal composition exceeds that of the mantle. (Color figure can be viewed at wileyonline library.com.) of garnet. Consequently, the Gusev basalt yields higher densities and seismic velocities even before the transition to eclogite is complete. Hence, small compositional differences can have a significant influence on rock properties and may be of importance for phase transitions and the location of the crust-mantle boundary.
According to our calculations, the crustal compositions considered have the potential to yield higher densities than the average mantle at a crustal thickness exceeding 100 km and for colder geotherms. Therefore, recycling of the Martian crust into the mantle could occur through the process of crustal delamination. However, delamination only takes place if the thermal and compositional parameters are favorable for rheological decoupling (e.g., Ueda et al. 2012). Such a process of gravitational lithospheric instability (or delamination) causing recycling has been proposed for continental regimes in the Earth, Venus, and early Mars (Breuer and Spohn 2003;Elkins-Tanton et al. 2007;Johnson et al. 2014) and may help explain the high potassium of rocks at Gale Crater through mantle metasomatism (Filiberto 2017).
However, a thicker crust not only exposes the root to higher pressures but also increased temperatures. Higher temperatures are less favorable for densification and may therefore prevent delamination (Semprich and Simon 2014), in particular when temperatures above 1000°C are required to ensure low viscosity of the underlying mantle (Kukkonen et al. 2008). Lower temperatures may not be favorable for mineral reactions and the lower crust may therefore be preserved in a metastable state or only partially reacted to eclogite in the absence of water. Fluids play an important role in the Earth's crust (Putnis and John 2010) and their absence may inhibit eclogitization and densification (e.g., Leech 2001). However, it is not clear how much fluid is required to trigger mineral reactions. Experiments have shown that very low water concentrations in the range of ppm are sufficient for diffusion and mineral growth (e.g., Milke et al. 2009). Water contents in apatite suggest that the Martian mantle source region to the shergottites, chassignites, and nakhlites contains 36-290 ppm H 2 O and that the bulk Martian crust has~1410 ppm H 2 O McCubbin et al. 2016). These values may be enough for mineral reactions but not to form waterbearing minerals such as amphiboles.

Relevance to Geodynamic Models and the Interpretation of Seismic Data
Due to the lack of seismic constraints, neither the crust-mantle boundary nor the densities of the lower crust and upper mantle can be determined without uncertainties. We consider only a few options as lower crustal compositions, which are based on meteorite compositions and in situ measurements of surface rocks. While the lower crust is assumed to be basaltic, the chosen compositions may not adequately reflect Mars' lower crust. Furthermore, variations in crustal composition and properties are expected for a planet of Mars' size.
This study provides pressure-temperature and composition-dependent density and seismic velocity grids that can be implemented into geodynamic and geophysical models to constrain the Martian interior structure and to model processes at the crust-mantle boundary. Furthermore, our seismic velocity diagrams can help improve radially symmetric Mars models by providing mineralogically realistic seismic velocities of the lower portion of the crust. In addition, the seismic velocities calculated in this study can facilitate the determination of the composition of the lower crust once InSight receives seismic data from the Martian interior. A crust-mantle boundary at~50 km could be identified by the relatively large differences in seismic velocities between crustal compositions and the mantle in the range of 0.6-0.7 km s À1 and 0.5 km s À1 (P-wave) for the average crust and Gusev basalt, respectively. In case of a thicker crust exposed to pressure~1 GPa, this velocity contrast is 0.5 or less for the average crustal composition and only 0.2-0.4 for the Gusev basalt. As a result, the crust may not be easily distinguishable from the mantle and the identification of a crust-mantle boundary may be difficult.

Our
calculations show that compositional differences of possible Martian lower crust can have a significant influence on densities and seismic velocities despite their similar mineral assemblages. At crustal thicknesses of~50 km, we expect large density and seismic velocity contrasts between the lower crust and the mantle. A thicker crust and therefore higher pressures in combination with relatively cold geotherms will cause a significant densification of the crustal compositions resulting in very little density and velocity contrasts. Therefore, the crust-mantle boundary may be difficult to identify in seismic profiles. The Gusev basalt composition has the potential to become denser than the mantle if the crustal thickness exceeds~100 km. Delamination may be possible, if the temperatures required for rheological decoupling are not too high.