Recent seismic data acquired by the InSight lander have revealed seismic discontinuities in the Martian crust that have been interpreted as sharp transitions in porosity or chemical composition. Here we propose an alternative model in which the transition from dry cracks to water-filled cracks could explain the observed seismic discontinuity in the Martian crust. Our model can explain sharp increases in seismic velocity and Vp/Vs at ∼10 km depth with no associated changes in porosity or chemical composition. The present model suggests the local existence of liquid water in the Martian crust, which could potentially serve as a subsurface habitat for life.

The Interior Exploration using Seismic Investigations, Geodesy and Heat Transport (InSight) mission landed on Mars in 2018 and has since detected seismic signals from the planet's interior (e.g., Lognonné et al., 2020). The seismic structure of the Martian crust beneath the InSight landing site is characterized by three layers with seismic discontinuities at ∼10 km and ∼20 km depth (e.g., Kim et al., 2021; Knapmeyer-Endrun et al., 2021; Durán et al., 2022; Joshi et al., 2023). Additional shallow discontinuities have been discussed in recent analyses (e.g., Shi et al., 2023; Carrasco et al., 2023; Drilleau et al., 2023), but here we focus primarily on the discontinuities detected at ∼10 km and ∼20 km depth.

The boundaries found in the crustal layer have been interpreted as distinct changes in porosity or chemical composition, given that these factors have a marked influence on seismic velocity (e.g., Wieczorek et al., 2022; Kilburn et al., 2022). Wieczorek et al. (2022) suggested that the discontinuity at ∼10 km depth reflects compositional variation between upper volcanic materials and lower impact ejecta deposits and that the deeper discontinuity at ∼20 km is due to a transition from porous to nonporous materials in the Martian crust.

A variety of geomorphological features on Mars indicate the existence of a persistent ocean during its early history (e.g., Carr, 1987; Baker et al., 1991), which is in agreement with the widespread occurrence of clay minerals in the Noachian terrains (e.g., Ehlmann et al., 2011). Although there is limited evidence of subsurface water on present-day Mars, recent remote-sensing images show recurring slope lineae that might indicate liquid fluid flow near the surface (McEwen et al., 2011).

If liquid water exists in interstitial spaces in the crust, then the seismic velocity is expected to be higher than that of dry material, as expected from the effective medium theory (e.g., Heap, 2019; Manga and Wright, 2021; Kilburn et al., 2022). Therefore, one possible explanation for a seismic discontinuity in the Martian crust is a transition from a dry layer to a water-rich layer with no associated changes in porosity or chemical composition. We test this hypothesis by examining the seismic velocity of fractured crustal rocks under dry, water-saturated, and frozen conditions and discuss the origin of the seismic discontinuities in terms of the presence or absence of subsurface water on present-day Mars.

In the present experiments, we used diabase samples from Rydaholm in southern Sweden with an equigranular texture and composed mainly of plagioclase and orthopyroxene because this composition is considered to represent a typical Martian crustal rock (e.g., Zuber, 2001). We investigated the effect of cracks on the seismic velocity of the analog rocks under different crack-filling conditions, in which thermal cracks were created by heating the samples to 800 °C in a nitrogen-purged furnace. The porosity of the original samples was <0.2%, and that of the thermally treated samples was increased to ∼1.2%. Following the thermal treatment, scanning electron microscopy revealed an almost isotropic distribution of intra and intergranular cracks (Jayawickrama and Katayama, 2023).

The compressional-wave (P-wave) and shear-wave (S-wave) seismic velocities were measured via a pulse transmission method using piezoelectric transducers with a resonant frequency of 2 MHz. We measured the seismic velocity in one direction due to the nearly isotropic structure and homogeneous crack distribution in the samples. Dry measurements were made after vacuum heating at 100 °C for 24 h, and wet measurements were made after saturating the samples with brine (0.5 mol/L NaCl) for 10 days. The saturated samples were then stored in a freezer at −55 °C, which is below the eutectic point of the brine, with the seismic velocities then measured in the freezer.

The P- and S-wave travel times vary among the analyzed crack-filling phases, with the dry (gas-filled) samples yielding the slowest seismic velocity (Fig. S1 in the Supplemental Material1). The seismic velocities decrease systematically with increasing porosity, although the porosity effect is dependent on the crack-filling phases due to the different elastic moduli for gas, water, and ice (Fig. 1; data are presented in Table S1 in the Supplemental Material). The elastic properties of crack-bearing materials are known to be sensitive to the crack porosity and crack shape according to the effective medium theory (e.g., Kuster and Toksöz, 1974; O’Connell and Budiansky, 1974; Kachanov, 1993). In this study, we applied the effective medium theory of Kuster and Toksöz (1974) because of its applicability to a wide range of porosity with different inclusion phases, including ice, while the models of O’Connell and Budiansky (1974) and Kachanov (1993) are limited to the soft fluid phases such as gas and water.

According to the long-wavelength first-order scattering theory of Kuster and Toksöz (1974), the effective bulk modulus K* and shear modulus μ* are dependent on porosity ϕ and crack aspect ratio α, as given by:
and
where the subscripts m and i refer to the matrix and inclusion moduli, respectively. The coefficients β and ζ are given by:
and

The assumptions involved in this derivation are homogeneity and isotropic crack distribution in the solid matrix, and the crack shape is treated as a spheroidal inclusion with aspect ratio α defined as the ratio of polar to equatorial lengths.

The experimental results for dry and water-saturated samples are well explained by the model with an average crack aspect ratio of ∼6 × 10−3 (Fig. 1), which is nearly consistent with that inferred from terrestrial crustal rocks (e.g., O’Connell and Budiansky, 1974). A similar relationship between seismic velocity and porosity is obtained from the other theoretical models (Fig. S2), and we also explored the relationship with different crack aspect ratios (Fig. S3).

Although the dry and water-saturated cracks significantly reduce the seismic velocity, the ice-filled cracks have little effect on the seismic velocity due to their relatively stiff moduli compared to the gas and water phases. Applying the Hashin-Shtrikman lower bound to the ice-filled cracks yields a relationship similar to that obtained from the Kuster and Toksöz model (1974; Fig. S2), indicating less sensitivity of seismic velocity to crack shape for the ice model.

We show that the seismic velocities of the gas-, water-, and ice-filled cracks are significantly different, consistent with theoretical model calculations. This suggests that a seismic discontinuity may occur at the transition from one crack-filling phase to another. Porosity is expected to decrease with depth on Mars, with possible pore closure occurring at depths where viscous deformation dominates at higher temperatures (Gyalay et al., 2020). In our model, we assumed a simple porosity profile with a constant value down to ∼20 km depth, and seismic velocities were calculated for the different crack-filling phases using a crack aspect ratio of 6 × 10−3 determined in the analog rocks (parameters used for the calculation are listed in Table 1).

Our calculations show that a change from dry cracks to water-filled cracks in the middle crust can lead to a significant increase in seismic velocity, even for constant porosity (Fig. 2). A further increase in seismic velocity is expected at greater depths, where most cracks close due to viscous flow at depths below ∼20 km. Crack density and porosity may change continuously with depth due to compaction and precipitation, but the gradational change with depth does not provide the necessary impedance contrast across the boundary (Fig. S4).

The results also depend on the crack aspect ratio used in the calculation. Previous studies have used relatively large aspect ratios in the range of 0.03–1 (e.g., Heap, 2019; Kilburn et al., 2022). Such models may be appropriate for the brecciated layer with various ejecta fragments at shallow impact craters, whereas the damaged basement is likely to have abundant fractures at deeper levels (e.g., Pilkington and Grieve, 1992). In this study, we show that the fractured analog rocks contain cracks with small aspect ratios (α = 0.005–0.007), and the large velocity variation in the Martian crust can be explained by the thin-crack model.

Recent Bayesian inversion analysis by Wright (2024) suggests the presence of liquid water in the Martian middle crust, similar to our conclusion, although their model requires round pores with a large aspect ratio (α = 0.19 ± 0.18). We have tested such a large aspect ratio, but the calculated S-wave velocities with dry and water-filled pores are almost identical at this aspect ratio (Fig. S5), hindering development of a seismic discontinuity in the crustal layer.

We also note that ice-filled cracks have a small effect on seismic velocity and are unlikely to explain the observed seismic velocity variations beneath the InSight landing site, as Manga and Wright (2021) suggested no cryosphere beneath the landing site. Given the current heat flow on Mars (Nimmo and Tanaka, 2005), the freezing temperature of water occurs at a depth of ∼10 km (Fig. S6). This allows for the existence of liquid water below the middle crust. Such water likely contains chlorides and sulfates (e.g., McEwen et al., 2011), further lowering the eutectic point and prolonging the stability of the subsurface water. Crustal temperature can reach 79 °C at 30 km depth, which slightly changes the bulk modulus of water, but the effect on seismic velocity for the water-filled cracks is limited to <0.6% in these temperature ranges.

The joint inversion of receiver functions and ellipticity of Rayleigh waves has recently shown seismic discontinuities in both P- and S-wave velocities and Vp/Vs increases in the Martian middle crustal layer at ∼10–20 km depth (Carrasco et al., 2023). Our experimental data and theoretical calculations indicate that Vp/Vs can increase with the water-filled cracks, while the dry cracks are predicted to decrease Vp/Vs. The high Vp/Vs anomaly also would be difficult to achieve due to the rounded pores (Fig. S5) or the compositional variations of the crustal rocks (Christensen, 1996). These lines of evidence are consistent with a water-rich fractured layer in the middle Martian crust.

At depths shallower than ∼10 km, the model predicted a significantly lower Vp/Vs than the observed anomaly. When cracks are filled by water, both P-wave and S-wave velocities are increased, but water has a greater effect on Vp compared to Vs. Thus, partial saturation of dry cracks may cause the relatively fast P-wave velocity and high Vp/Vs in the shallow crust. Another possibility is a change in crack shape at shallow depth, where dry cracks with a large aspect ratio can result in relatively high Vp/Vs.

Extrapolation of laboratory data to seismological observations must be done with caution due to differences in frequency range. In the case of fluid-saturated cracks, elastic wave propagation may induce local fluid flow under relaxed conditions, and Gassmann's relation has been applied to the low-frequency seismic data (e.g., Adelinet et al., 2011). However, statistical estimates of heterogeneity in the Oman ophiolite on Earth indicate large-scale unrelaxed conditions for pore fluids in the crust, suggesting that wave-induced fluid flow is unlikely to occur, and thus the effective medium theory can be applicable even at the seismic frequency scale (Akamatsu et al., 2024).

The InSight mission detected surface waves caused by meteorite impacts away from the landing site (Kim et al., 2022). Analysis of these data revealed small velocity variations in the crust, which is in contrast to the layered seismic structure at the InSight landing site. However, surface waves represent an averaged velocity along their travel paths and are less sensitive to imaging seismic discontinuities. The relatively low seismic velocity in the upper crust beneath the landing site is compatible with highly fractured materials that possibly originated within a buried impact crater, as seen in the quasi-circular depression at the landing site (e.g., Pan et al., 2020). Recent analyses have shown the uppermost low velocity layer as ∼1–2 km thick (e.g., Shi et al., 2023; Drilleau et al., 2023), interpreted as impact breccias or ejecta deposits overlying the crustal layer, which may have a higher porosity and different pore geometry compared to those in the deeper basement. Seismic anisotropy with a fast vertically polarized velocity also indicates the presence of highly fractured zones in the upper crust near the InSight landing site (Li et al., 2022). The anisotropic features are not included in our calculations, so the velocity difference between model and observation could be partly due to this effect.

We propose that the seismic discontinuity detected at ∼10 km depth at the InSight landing site may be caused by the presence of water-filled cracks in the middle crustal layer, indicating the possible existence of subsurface water in the Martian crust (Fig. 2). This is not a unique solution to explain the current observational data, but given the sensitivity of electrical conductivity to aqueous fluids, electromagnetic measurements on Mars may be helpful in further testing the above hypothesis. Although liquid water is not stable on the surface of present-day Mars due to the low atmospheric pressure, the increase in lithostatic pressure with depth could allow the presence of subsurface water at temperatures above the eutectic point.

Long-term cycling of water between the atmosphere and near-surface crust may have occurred throughout the geological history of Mars, with the hydraulic head gradient created by latitudinal groundwater mounds driving fluid flow toward the equator (Clifford, 1993). Most groundwater may have been depleted to form the ice-cemented cryosphere (e.g., Weiss and Head, 2017), but subsurface water may have survived locally around impact craters due to the relatively high thermal gradients in the crust. On Earth, numerous microbes are present in the subsurface and receive nutrients through fluid-rock interactions such as serpentinization (e.g., Kelley et al., 2005). If liquid water exists in the interior of present-day Mars, it could provide a habitable environment and support a microbial community.

1Supplemental Material. Figures S1–S6 (velocity calculation) and Table S1 (experimental data). Please visit https://doi.org/10.1130/GEOL.S.26980444 to access the supplemental material; contact [email protected] with any questions.

This study was supported by the Japan Society for the Promotion of Science (20H00200 and 17H06454). We thank Haruna Wakabayashi for preparing the samples used in the measurements. Comments from Doyeon Kim and two anonymous reviewers were helpful in improving the manuscript. We also appreciate the comments and careful handling of our manuscript by the science editor (Tracy Rushmer).