## Introduction

The Hesperian and Amazonian tectonics of Mars are well characterized owing to plentiful exposures. However, the pre-Noachian and Noachian rocks are obscured by extensive resurfacing in many regions. Where exposed, tectonics is seen to be locally complex1,2 as expected from remote geochemistry and geophysics3. In this paper, we report on two large shear zones exposed in a deep erosional window that formed in the brittle-ductile domain of the crust early in Mars history and located in the area between the northern Valles Marineris trough system and the planetary dichotomy boundary (PDB).

The Valles Marineris trough system is exposed atop a broad topographic dome that is superimposed on the Tharsis rise4. The dome’s large free-air gravity anomaly5 is consistent with an upward flexural response of crust thickened between the Late Noachian and Early Hesperian. Trough nucleation6,7 has been interpreted as the consequence of differential loading of the thickening Tharsis volcanic pile over the pre-existing step in basement topography that has been associated with the Borealis basin, which may have formed by an oblique impact event early in Mars history, probably around 4.5 Ga4,8,9. Extrusive volcanic activity associated with the early growth of Tharsis could have led to differential (north-south) loading of the crust across the PDB, producing a circumferential flexural uplift some 500–1000 km wide4, growth of large-scale igneous dikes6,10, dilational normal faulting11,12, and subsidence of crustal blocks parallel to them4,6,7. During the Hesperian, the troughs were partly filled by the Interior Layered Deposits (ILD), consisting of sequences of sulfate-rich layers formed by chemical evolution in an acidic environment of atmospherically-deposited ice, dust, and sulfur-rich volcanics13,14, leading to continued or renewed subsidence. The ILD were then locally eroded down by several kilometers. The reported shear zones are located in areas where ILD erosion attains ~6 km vertically. However, the exposed surface is not the surface of the Hesperian plateau. We show that it is a deeper crustal level in which complex deformation and intrusive activity is apparent, opening a window into the proto-Valles Marineris history.

## Results and discussion

### Identification of a basement unit in northern Valles Marineris

The deepest parts of eastern Ophir and Hebes Chasmata (Fig. 1) display complex terrains that have not been characterized before. These terrains are fully covered by MRO/HiRISE images. Mapping at the HiRISE scale (Fig. 1 and Supplementary Figs. 1,2) reveals highly deformed light-toned rocks. Most of them form a massive unit, with fracture patterns reminiscent of crystalline rocks. This rock unit also contains dikes and other intrusions. The layering of distinct colors indicates that its composition is not uniform, and suggests mechanical contrasts that may influence the deformation style. Although generally light-toned, the morphology of these rocks is distinct from that of the nearby ILD, also light-toned. Supplementary Figs. 37 illustrate this diversity of terrains. These terrains were not identified in earlier geologic maps of Valles Marineris, in which they were attributed to Amazonian landslide-related or other surficial processes (see Supplementary Discussion).

Nonlinear unmixing15 of hyperspectral data from MRO/CRISM (see Methods section) applied to three regions of interest (ROI) where the basement rock is well exposed (see Supplementary Methods and Supplementary Data 116) shows that the rock is mafic and contains olivine, pyroxenes, and plagioclases, associated with various sulfates (Fig. 2). It gives a primary rock composition similar to one of the bright rocks exposed in other places of Valles Marineris17 and at the outer Hellas impact ring structure18. Should the retrieved primary mineral abundances from the analyzed CRISM data be representative of the exposed rock, from conventional igneous rock nomenclature, it would be of olivine-gabbronorite (ROI1) or troctolite (ROI2 and ROI3) composition. In that case, the primary mineral assemblage gives a magmatic origin, with copiapite, jarosite, and szomolnokite additionally suggesting hydrothermal alteration of sulfides. A similar assemblage of sulfates found in Noctis Labyrinthus was also interpreted as the result of hydrothermal activity19. Anhydrite, identified in veins in Gale crater20 and interpreted as originating from basaltic rock weathering, could alternatively be of primary origin21, storing magmatic sulfur prior to breaking down and releasing it to the hydrothermal sulfates22. Kieserite may come from the nearby ILD23, either transported by wind, or present in some small ILD outliers on the chasma floor. The difference between the composition of ROI1 and ROI2-3 confirms that the exposed basement does not have a uniform composition; in fact, HiRISE image analysis suggests a larger diversity of rock types, as illustrated by vertical banding (Supplementary Fig. 7). Gabbronorite and troctolite composition, layering, and the presence of minerals coming from alteration of sulfides make a geological assemblage that characterizes some magmatic cumulates on Earth, e.g., at the Longwood Range in New Zealand24.

The retrieved mineralogical composition is markedly distinct from the ILD composition23. Aside from compositional information, the thickness of the dikes, tens of meters, and the presence of other intrusive bodies at the surface also indicate that the observed surface is exhumed, with probable kilometers of crust removed by erosion prior to ILD deposition, like dikes of similar thickness exposed on terrestrial continental shields7. Chasma formation was therefore driven not only by tectonic extension, but also substantial erosion, that the upper Hesperian ILD age25 constraints to not later than lower Hesperian. The age of the basement rocks themselves must be older, perhaps Noachian or pre-Noachian. Both in the Ophir and Hebes Chasmata, this basement is discontinuously exposed, partly blanketed by low albedo deposits, including mottled terrain, a smooth mantle of probably airfall origin, and dune fields (Supplementary Figs. 13). This generally smooth material, devoid of large impact craters, explains why earlier mappers considered it to be of the Amazonian age (see Supplementary Discussion).

### Brittle-ductile shear zones and fault megabreccia in the basement unit

We now focus on the deformation affecting the light-toned, mafic basement unit underlying the dark mantle terrains. Structural analysis (Fig. 3 and Supplementary Figs. 1, 2) reveals S-C and S-C-C’ structures that point to right-lateral kinematics, with the main shears (C) oriented N055E (Ophir) and N065E (Hebes). Their formation implies shearing in a brittle-ductile regime. In the Hebes shear zone, fractures having walls separated laterally by up to one hundred meters are observed on a HiRISE image. They contain rock fragments of longer dimensions, up to several tens of meters. HiRISE color information reveals at least two different rock compositions (greyish and yellowish). The fragments are incorporated into a bright matrix bounded by the fracture walls. We interpret them as a megabreccia that testifies to brittle deformation.

The minimum shear zone width is estimated to be ~1 km in Ophir Chasma and 3 to 3.5 km in Hebes Chasma (Fig. 1 and Supplementary Figs. 1,2). Using scaling relationships to evaluate shear zone displacement from their measured width, and then shear zone length from displacement (see Methods section), displacements along these structures may range between a few tens to several hundreds of kilometers, implying, in turn, horizontal shear zone extents of several hundred to perhaps several thousand kilometers (Supplementary Fig. 8).

The relative timing of brittle and brittle-ductile deformation could not be determined from mapping. For instance, cross-cutting relationships between tectonic structures formed in the brittle field (e.g., megabreccia) and the ones formed in the brittle-ductile field (e.g., sigmoidal fabric), or evidence of tectonic structure reactivation, were not observed. However, none of these deformation patterns are seen to extend in the ILD nor on the plateau surrounding the troughs, indicating that they formed earlier than the Hesperian plateau surface and ILD1. The Hebes shear zone extends west and, to a lesser extent, north on the Hesperian plateau as a narrow graben system (Fig. 1a and Supplementary Fig. 9). It may suggest reactivation of the shear zone in extension over as much as ~900 km, up to Noctis Labyrinthus, after the deposition of the early Hesperian Tharsis volcanics.

The formation of brittle-ductile shear zones several kilometers below the current surface requires that the Martian crust was wet and weak, in agreement with several works (see Methods section). The available water sources may have included abundant Noachian groundwater26, crustal serpentinization in relation to PDB formation, then deserpentinization and water release to the surrounding rocks in response to the thermal events associated with the formation of early Tharsis27, as well as juvenile water from Tharsis magmas. The heat advected by the fluids contributed to warm the surrounding crust, leading to a variable and locally steep thermal gradient during periods of active magmatism28, perhaps generating crustal strength contrasts at local scales.

### Shear zone significance

Our findings and the constraints they yield for an understanding of the evolution of the Valles Marineris region before the chasma opening are the following: (1) The shear zones are located near the PDB, and their trend is parallel (Hebes) or nearly parallel (Ophir) to it (Fig. 4). (2) The Hebes shear zone is associated with megabreccia in a bright matrix within thick fractures (Fig. 3c), and the Ophir shear zone basement unit includes megabreccia facies (Supplementary Figs. 4-6). (3) They are also located in the transitional area between the weakly magnetized lowlands crust in the north and the intensely magnetized highlands crust in the south, thought to mark the margin of the Borealis basin29 (Fig. 4 and Supplementary Fig. 10). This transition was modified by Tharsis thermal events30, which started in the Noachian after the Martian dynamo shut down, erasing most of the magnetization in the shear zone area (Supplementary Fig. 11). (4) The trend of the shear zones and the Borealis impact edge is parallel to the orientation of elongated crustal magnetic anomalies north of Coprates Chasma (Fig. 4). It was inferred that the impact occurred around 4.5 Ga, while the early dynamo was active31, strengthening the interpretation of a geological connection between them. (5) The westernmost radial magnetic anomalies north of Coprates Chasma are aligned with several crustal ridges of volcanic25 or tectonic17 origin up to 1 km high, formed in pre-Noachian to early Noachian terrain (Fig. 4). (6) Circumferential magnetic anomalies are observed around some impact craters and basins on Earth, where they result from the formation of magnetic minerals by impact-induced long-lasting hydrothermal activity32,33. (7) We note that the crustal ridges display channels restricted to the Noachian or pre-Noachian terrains, indicating water flows (Supplementary Fig. 12). The ridges also display surface fractures of similar, magnetic anomaly parallel orientation. (8) The NE-trending tectonic fabric is aligned with a NE-trending dike swarm on Candor Chasma walls (Fig. 4, Supplementary Figure 12), revealing intrusive magmatism along these magnetic anomalies. (9) Stress trajectories in the Valles Marineris region in response to PDB flexural loading by Tharsis using a thin-shell approach4 are adapted to the tectonic deformation observed today. In spite of its simplicity (rock failure, friction, and long-term crustal viscous relaxation, are not taken into account), the model yields several major results. Tensile deviatoric stress is predicted to be high in the crustal portion of the Valles Marineris area between the PDB and Coprates Chasma. The tension axis is approximately parallel to the magnetic lineations (Fig. 4, with detailed analysis in Supplementary Fig. 13): the tectonic and magnetic fabrics are parallel. A major implication is that the magnetic anomaly patterns correspond to tectonic features. This inference is consistent with the observations made at the crustal ridges and the orientation of the Candor Chasma dike swarm. (10) In the Ophir and Hebes shear zone areas, the orientation of the principal stress trajectories in the flexural model4 is in agreement with the observed right-lateral shear zone kinematics. The expected extensional regime is also consistent with an extension perpendicular to chasma orientation. (11) South of Coprates Chasma, the stress trajectories are no longer aligned with the magnetic anomalies, which perhaps indicates that magnetization is controlled by a crustal structure or composition distinct from those prevailing north of Coprates Chasma.

These elements suggest a model of crustal evolution in Valles Marineris until the opening of the main chasmata (Fig. 5). It explains not only the formation and kinematics of the shear zones, but also a series of geological features, the formation of which has remained elusive. They include the crustal ridges, the chasma-perpendicular dike swarm in Candor Chasma, the peculiar orientation of some narrow grabens between Syria Planum and Hebes Chasma, and the linear nature of magnetic anomalies north of Coprates Chasma.

The relationships between the shear zone, PDB, and anomalies are explained here by basin tectonics34 applied to the Borealis basin. Formation of the Borealis basin at ~4.5 Ga generated circumferential normal faults filled with megabreccia, and crustal warming propagated from the molten crust in the impact area9. Hydrothermal circulation started along the fault pathways (Fig. 5a), mobilizing the water present in the crust, and perhaps also from the impactor. The megabreccia in the Ophir Chasma basement unit may have formed at that time, too, as an impact megabreccia, like the one interpreted around the Hellas basin18, as well as the Candor Chasma dike swarm as an impact-related ring dike system.

The duration of hydrothermal activity was shorter than the duration of the dynamo (Fig. 5b). It is poorly constrained, but for a Hellas size impact basin, it was estimated to be in the order of 10 My35. Uncertainty arises from parameters such as post-impact crustal permeability and water abundance, as well as from the difficulty of modeling fine-scale processes in response to such a large-scale event. The four times larger Borealis basin size, based on gravity and topography8, may have much lengthened the duration of hydrothermal activity. The dynamo activity lasted until 4.1 Ga or even later in the late Noachian31, leaving at least 400 My for the hydrothermal activity to gradually stop and record the succession of polarity inversions while cooling to the Curie temperature.

The location and orientation of the Ophir and Hebes shear zones suggest that they may have been segments of ring normal faults. Similar to the other impact basins, many other ring faults should have fractured the crust in response to the impact. The observed elongated magnetic anomalies of alternating polarity may underline these structures. The hydrothermal minerals detected in the basement of the Ophir Chasma shear zone would therefore be the exposed equivalent to currently buried hydrothermal minerals that generated the observed magnetic anomalies.

In the late Noachian, adjustment of the Borealis basin load should have been achieved, through reactivation of the basin faults, decreasing the topographic difference between the highlands and lowlands (Fig. 5c). The geologic context of the Ophir and Hebes chasmata dramatically changed under the combined influence of the growth of the Tharsis load36 and the major PDB crustal discontinuity. The flexural loading model of Tharsis over the PDB4 predicts strike-slip reactivation along the basin ring structures in the Ophir and Hebes chasmata, resulting in the observed shear zones (Fig. 4).

The distinct tectonic domain south of Coprates Chasma is coherent with its interpretation as the northern limit of a major Hesperian and Noachian deformation system on the Thaumasia plateau and the Coprates rise37. These early Tharsis events may have deformed the crustal magnetic fabric, which would not reflect the orientation of the Borealis basin rim anymore.

The lower Hesperian marks the beginning of flood volcanism in the Tharsis region, specifically centered at Syria Planum6,25. The lava flows loaded the crust in the Valles Marineris region prior to the opening of the chasmata (Fig. 5d), with the consequence of increasing the vertical principal stress. Chasma opening by crustal extension may simply result from the permutation between the largest and intermediate principal stresses due to crustal loading by the lavas (Fig. 5d–f). Flood lava emplacement occurred via an intrusive system including a huge dike swarm6 of mafic composition38, partly cancelling crustal magnetization. New magnetization could have occurred upon dike cooling in response to the external or to the nearby crustal magnetic field. However, the external field is not stable and unlikely to produce consistent patterns all over the dikes. A new magnetization related to nearby magnetization would be much less intense than the original one and is unlikely to be detected from orbit.

The shear zones identified in the Ophir and Hebes chasmata can therefore be understood in a scenario in which an inherited crustal fabric is reactivated (Fig. 5f, stage 1), additional crustal loading increases the differential stress (stage 2), ending up with nucleation of the main chasma faults in tension (stage 3). The edge of the Borealis basin west of Hebes Chasma is nearly parallel to the Valles Marineris chasmata, which explains why the Hebes shear zone was there reactivated in tension, forming some narrow grabens.

It is apparent from this geological context that the hydrothermal activity identified in the Ophir and Hebes shear zone basement may have been widespread. Coprates Chasma, as a major tectonic and magnetic boundary, may represent the southern end of this hydrothermal system in the highlands crust. Evidence of hydrothermal activity at the PDB in the Valles Marineris region suggests that the weak, 10 km thick crustal layer found at the InSight landing site39, also located at the PDB area, might be a similarly altered mafic basement rock.

Brittle-ductile shear zones are one of the contexts where metalliferous mineralizations can be abundant. Our results show that the Ophir and Hebes shear zones are of exceptional, if not unique, interest to search for such mineralizations. Mesothermal gold deposits on Earth (30% of mined gold) are observed in secondary structures of exhumed brittle-ductile shear zones and their shallow extension to the surface40. Au is drained in terrestrial shear zones by hydrothermal fluids, and after a complex path, concentrates in cataclastic quartz veins41. The mineralogically expressed Au is associated with As- paragenesis of Sb, W, Bi, Mo, and Pb, and in the quartz veins, with Pb, Ag, and Cu.

Assuming that the primary minerals identified using CRISM reflect the full basement rock composition, the shear zones could be sites to prospect for Ti-V oxides, which are found in terrestrial gabbronorite–troctolite layered intrusions42. At different depth levels, stratabound Ni-Cu sulfides and chromite may have settled43, as well as platinum-group minerals, especially including Pt, Pd, and Au24. Although the deep source of metals accumulating in the planetary crust is loosely constrained, the sheared, fluid-charged Noachian and pre-Noachian basement rocks, as well as the neighboring intrusions, may be the most favorable environments to discover such deposits.

The influence of the Borealis basin on the tectonics and hydrothermal activity in the Valles Marineris region, and the implications for mineralizations, open a wealth of new questions of global planetary interest. Did the crustal fabric imposed by the impact influence the evolution of the Tharsis crust in other areas? Did it play a role in the genesis of the large fracture zone that guided the NE-SW orientation of many volcanic edifices that formed on the Tharsis dome? At the opposite, the Tharsis-generated stress field was shown to control the stress system of the whole planet44. Did it reactivate other PDB-related structures? Are there other sites where PDB-related hydrothermal activity can be found? Can it be identified elsewhere, without the exceptional exposure conditions in Valles Marineris? The obliquity of the Borealis basin impactor9 should have generated a crustal fracturing asymmetry45, did it affect the distribution of basin-generated mineralizations? Hydrothermal minerals were detected in the Nili Fossae and Nilosyrtis highlands46, next to the PDB, which also display breccia ridges47. Do these features testify to PDB-related geologic events as well? Were the reactivated basin-related radial faults around the Hellas and Isidis basins48 also prone to similar mineralizations? Finally, was the duration of hydrothermal activity associated with the Borealis basin long enough to stabilize the environmental conditions and favor the emergence of life? Radiogenic heating might have contributed to maintain a high geothermal gradient for some time49, making the northern Valles Marineris a preferential region to investigate early life in situ.

## Conclusions

Our results show that the influence of the Borealis impact on the later evolution of Mars may have been underestimated (Fig. 6). Large-scale shearing of the pre-Valles Marineris basement rock testifies to major crustal deformation unrecognized before that shall be taken into account in deciphering the early Mars’ evolution. From a stress field point of view, this event is consistent with, and prefigures Valles Marineris opening, but does not prefigure the current Valles Marineris geometry. The shears probably initiated as normal faults at the time of the Borealis basin impact, as soon as ~4.5 Ga, the estimated age of the basin, while the early Martian dynamo was active31. This very long tectonic activity, combined with deep exhumation, the mafic composition of the sheared basement, and hydrothermal circulation, provide one of the most promising environments for lode deposition and concentration of metals, perhaps of cut-off grade.

The idea that deep faults such as from impact basins were the locus of hydrothermal activity and played a role in crustal magnetization35,50 is also validated by our analysis. The recognition of early fluid-assisted deformation and identification of alteration products associated with early hydrothermal systems suggest that at least some fundamental Martian faults may have been rheologically weak, and, therefore, susceptible to reactivation and seismic activity, from early Noachian time and perhaps into the present in the current stress field, controlled by the Tharsis load44. If still reactivated nowadays, InSight should be able to detect it in a quiet seismic environment. The marsquake of August 25, 2021, measured by InSight in Valles Marineris, of body wave magnitude 4.1-4.251, is consistent with such a reactivation as much as the recognized deep chasma border faults11,12,52. The deep Valles Marineris troughs should be considered a high-priority target for innovative future exploration missions. It would enable a better understanding of the early Martian dynamics, as well as offer the grounds for in situ identification of rare resources.

## Methods

### Geologic mapping

Geologic analysis was fully conducted at the HiRISE scale (25 m pixel-1) in Ophir Chasma. In Hebes Chasma, only ~75% of the shear zone area is covered with HiRISE images. Gaps were filled with CTX images (6 m pixel-1). Conventional geologic mapping was conducted with an emphasis on structural analysis. Kinematic analysis of S-C fabric is from e.g., Platt and Vissers53, following the nomenclature established by ref. 54, and as modeled by ref. 55. Kinematic analysis of S-C-C’ fabric is after ref. 56, who also provide a review of works on this fabric.

### Generation of HiRISE DTM using photogrammetry and photoclinometry

Photoclinometry (or Shape-from-Shading, SfS) refers to the process of inferring 3D surface slopes from the intensity variations of a single shaded image. When combined with photogrammetry, as in ref. 57, SfS can be used to resolve 3D details from a given shaded image and merge it with the coarse 3D information inherited from an initial DEM. In this work, we apply the HDEM method57,58 to HiRISE images (1 m pixel-1) based on HiRISE stereo-derived DEMs (3 m pixel-1), and with a coarse-to-fine strategy, to further refine the DEM results (1 m pixel-1). HDEM SfS takes a single image and a co-registered coarse DEM as inputs, and iteratively refines the coarse input DEM via the minimization of a total cost function. The total cost function comprises three terms. First, the image irradiance term expresses the goodness of fit between the observed image versus its modeling from surface slopes. This modeling combines a surface atmospheric radiative transfer scheme and a realistic spatially varying bidirectional reflectance distribution function. Second, we have the integrability constraint regularization term that connects slopes to the terrain height field. Third, the photogrammetry constraint regularization term ensures the inheritance of large scales from the initial photogrammetrically generated DEM in the refined version. In this work, the spatial BRDF model results from the linear mixing of two endmember terrains, mafic, and sulfate-rich, which spectrophotometric properties are extracted from a near-coincident CRISM multi-angular acquisition sequence (FRT, HRL, HRS, EPF) using the MARS-Reco method59. The HDEM numerical scheme is repeatedly executed until a pre-set maximum number of iterations is reached or until the total cost function for the image irradiance equation, integrability constraint, and photogrammetry constraint converges.

### Determination of basement mineralogy

We analyzed each ROI with the LinMin algorithm15, based on Primal-Dual Interior-Point Optimization60. This method assumes a linear mixture with positivity and sum-lower-than-one constraints:

$${{{{{\bf{Y}}}}}} \sim {{{{{\bf{A}}}}}}.{{{{{\bf{S}}}}}}\,{{{{{\rm{with}}}}}}\,0\, \, < \, {{{{{\bf{A}}}}}} \, < \, 1\,{{{{{\rm{and}}}}}}\,0 \, < \, \Sigma {{{{{\bf{A}}}}}} \, < \, 1$$
(1)

with Y, the ROI spectra, A the abundance (that are estimated), and S the endmember spectra (assumed to be known). The spectra S are from spectral libraries and have been measured in laboratory conditions, as described in Supplementary Methods. We also included flat and slope spectra, as described in ref. 15, to reduce the effect of aerosols and discrepancies between laboratory and remote sensing conditions (grain size, geometry, aerosols in the atmosphere).

In order to improve the fit, we also introduce a nonlinear unmixing based on the fact that grain size changes can be simply modeled by S^e + C, with e an exponent factor, specific to each spectrum, and C a constant value. This new model can be formulated as:

$${{{{{\bf{Y}}}}}} \sim {{{{{\bf{A}}}}}}.{{{{{\bf{S}}}}}} ^{\wedge} {{{{{\bf{e}}}}}}\,{{{{{\rm{with}}}}}}\,0 \, < \, {{{{{\bf{A}}}}}} \, < \, 1\,{{{{{\rm{and}}}}}}\,0 \, < \, \Sigma {{{{{\bf{A}}}}}} \, < \, 1$$
(2)

We solved this problem using the Nelder-Mead Simplex Method61 to estimate the optimum e, including the LinMin algorithm15, by minimizing the reconstruction RMS. Unfortunately, this nonlinear procedure is much more computationally expensive than LinMin and not possible on a full CRISM cube. Nevertheless, for the ROI spectra selected for this work, this method is perfectly fitted. One has to note that retrieved abundances may differ from actual mineral abundance in the rock. Nevertheless, from the extensive experiments conducted, the detection of minerals from observed spectral features is robust. There is thus high confidence in the mineral assemblage for all ROIs.

A detailed description of the mineralogical analysis leading to the results presented in Fig. 2 is provided in Supplementary Methods. A synthesis of the mineral detections from spectral unmixing experiments is given in Supplementary Table 6. The spectral unmixing outputs are provided in Supplementary Data 116 as plots and ASCII files.

### Displacement scaling

Displacement scaling relates the width (w) or length (L) of tectonic structures (such as faults and shear zones) against their maximum displacement (D), along the whole range of rheological conditions in the crust. It is used as one of the tools to constrain the mechanical properties of the fractures and their host rock62,63,64,65. A broad agreement has been found over the years that for most tectonic structures, both (w) and (L) linearly scale with (D), irrespective of rock composition. However, from measurements reported in the literature, the possibility exists that for shear zones, D/L follows a power law, with an exponent of 1.565,66. In displacement scaling, the size of datasets available in the literature is not large enough to infer statistics, and D/w and D/L scaling cannot be used to predict one of the three parameters from the others accurately. Here we use D/w and D/L plots to evaluate the order of magnitude of displacement at the Ophir and Hebes shear zones from the minimum measured width, inferred from the exposures, then to evaluate the order of magnitude of shear zone length, from the displacement.

In the first step, the width of the Ophir and Hebes shear zones are measured using the spacing major shears that bound the zone of most intense deformation on HiRISE images (see Supplementary Figs. 1 for the Ophir shear zone and 2 for the Hebes shear zone). In Ophir Chasma, the width of the exposed shear zone is 1 km. In Hebes Chasma, the width is between 3 (“Hebes1” in Supplementary Fig. 8) and 3.5 km (“Hebes2”). D/w scaling is done based on a dataset of shear zones of similar dimensions 60. We calculate the average D/w value for this dataset: 63, and assume that the same value applies to the Ophir and Hebes shear zones. One order of magnitude is added as a vertical error bar.

In the second step, this (D) range is used to evaluate the length of the shear zones. Three datasets for shear zones are available in the literature62,66, among which a continental shear zone dataset is for shear zones of dimensions comparable to the Ophir and Hebes shear zones, which we use in this work67. We consider both linear (Supplementary Fig. 8b) and power law scaling with an exponent of 1.566,67 (Supplementary Fig. 8c). The average D/L for the continental shear zone dataset is 0.172. Assuming a linear relation between (D) and (L), the scaling line D = 0.172 L is used as a proxy for the Martian shear zones. Using the (D) range calculated earlier, a (L) range is calculated for the Ophir shear zone by projecting the (D) range along the scaling line. Two (L) ranges are calculated for the Hebes shear zone, corresponding to the 3 and 3.5 km measured widths. A horizontal error of one order of magnitude is attributed to the (L) ranges. As a second option, it is assumed that the continental shear zone dataset follows the power law of exponent 1.5. The (D) ranges are then projected onto this scaling line, and similarly, a horizontal bar of one order of magnitude is applied to the corresponding (L) range.

### Determination of potential causes for brittle-ductile crustal deformation at shallow depth in Ophir and Hebes Chasmata

Although we know that wet rheology is more likely for Mars68, we test it first with dry mafic (diabase) rheology. Assuming that deformation occurred at a depth equal to the current chasma floor depth, ~8500 m below the closest plateaus after MGS/MOLA data, the geothermal gradient required to generate brittle-ductile deformation would be 59–65 °C km-1, assuming brittle-ductile behavior at the failure/creep transition temperature of 500–550 °C, depending on plagioclase abundance69. Another composition, such as olivine, or diorite70 would only marginally change the conclusions. A quartz rheology71 would make the brittle-ductile behavior even easier to reach, but is not considered here because of the mineralogical information provided by the CRISM data processing (Fig. 2).

Since the Hesperian lava pile covering the Valles Marineris plateau did not exist when the shear zones developed, and part of the floor depth was attained by tectonic stretching, the floor might have been much closer to the topographic surface, perhaps decreasing the required depth by two. This factor is considered as a maximum, because the Noachian pile is exposed at the plateau level at the same elevation as the Hesperian lava pile 430 km away from the Ophir shear zone, at the eastern edge of Candor Chasma25, with no major change of stratigraphic levels reported along the chasma walls between eastern Candor Chasma and eastern Ophir Chasma. Increasing the required geothermal gradient accordingly yields 120 °C km-1. This value is 5–10 times the average gradient calculated in the Martian highlands during Noachian72,73,74,75 and for the Hesperian extensional deformation observed around Tharsis76, and four times the gradient found during Noachian rifting south of Valles Marineris, at Thaumasia Planum, up to 33 °C km-177. However, it was suggested recently that radiogenic heating78 may also have participated in substantially increasing the thermal gradient during the Noachian49.

The temperatures of 500–550 °C required to make diabase flow may actually not be needed, implying that very high thermal gradients may not be required. There is multiple evidence that crustal strength on terrestrial continents does not exceed ~50 MPa, with ductile creep occurring already at 6 km depth79. Equivalently, viscous relaxation of impact basins indicates a weak lower crust, with moderate heat flow80. Assuming a mean terrestrial crustal density of 2.7 g cm−3, the corresponding lithostatic pressure would normally be 158 MPa, implying stress reduction by a factor of 3 to 4. On Mars, assuming a basaltic crust of density of 2.9 g cm−3, 50 MPa would be located at 4.7 km depth, and if reduced by 3–4, ductile creep would occur as shallow as 1.2–1.6 km below the surface.

The first reason for such a strength decrease is the presence of free water in the crust71, which may be testified to by the CRISM detections of hydrated hydrothermal sulfates (szomolnokite and copiapite have respectively 1, and 20 molecules of water in their structure, and jarosite has three hydroxyl groups). The weakening effect of water is major whatever the creep regime, dislocation, or diffusion81. Even less than 0.1% of water79 may decrease rock strength by a factor of 2 to 369. Wet diabase82,83 is therefore much weaker than dry diabase69, but perhaps more importantly, the hydrothermal paths not only locally dramatically increase heat flow, but also their pervasive penetration in the crust has a substantial warming effect that adds to the background heat flow84.

Water may originate from infiltration through Noachian runoff channels85 as well as volcanic outgassing86. The water of metamorphic origin may also weaken the crust; in particular, the correspondence between the edge of the crust showing magnetic anomalies29 and the geometry of the exposed shear zones (Fig. 4) suggests a genetic link. One of the explanations for crustal magnetization is serpentinization during the Noachian27,87. Serpentinization may have weakened the shear zone area in three possible ways. First, during the magnetization itself, mobilizing free water, and incorporating it into mafic minerals transformed to chrysotile27,87, reducing crustal strength88 and shifting the portion of the crust exposed in the shear zone areas from the brittle toward the ductile domain. Weakening may proceed to such a point that locally serpentinite may rise through fractures by diapirism89, providing a potential explanation for intrusions mapped on Supplementary Figs. 1,2, as an alternative to magmatic intrusions. In that case, the whole magnetized crust being serpentinized, the weak rheology of the shear zone is not an exception, the observation of brittle-ductile structures at this depth is made possible by the denudation of Valles Marineris in the eastern Ophir and western Hebes Chasmata. Secondly, shear zone area weakening may have occurred during the process of crustal serpentinization because the troughs are located near the edge of the Borealis basin, a crustal zone of density contrast where intense fluid circulation is expected to have occurred. The rheology of the shear zones in Valles Marineris would be another tectonic consequence of crustal magnetization. Thirdly, if the shear zones formed later than the crustal magnetization event, crustal weakening could have corresponded to crustal deserpentinization. For instance, under the effect of magmatic heat release at the earliest stages of the Tharsis-Syria Planum magmatic events by the end of Noachian or earliest Hesperian6, water from chrysotile structure could have been releasing west of eastern Ophir Chasma. It is also possible that partial deserpentinization occurred owing to crustal heating from below, which would explain magnetization depth as shallow as 5–10 km90. Whatever the case, the magnetization of the Martian crust by serpentinization, or deserpentinization, is likely to have involved a substantial amount of water, up to 1 km of global equivalent layer27, which should have contributed to weakening the rheology in the shear zone areas.

Finally, crustal rheology strongly depends on the strain rate; taking the viscous component of the Martian lithosphere rheology into account would decrease lithospheric strength and shallow the brittle-ductile transition depth79.

In summary, a simple dry crustal rheology cannot explain the observation of brittle-ductile deformation in the Ophir and Hebes shear zones. Several mechanisms have been identified which may explain the formation of brittle-ductile structures at the observed paleo-depth. Crustal hydration is one, and it was noted that in addition to the dramatic and well-known effect of a very small quantity of free water on the rheology of the crust, for which there is evidence in the Noachian, hydrothermal circulation may have multiplied the geothermal gradients by several times at the scale of the exposed shear zone outcrops, a few tens of km2 in size. Structural water related to crustal serpentinization or deserpentinization (or other metamorphic reactions) may provide another source of rheological weakening, perhaps in connection with the initiation of the Tharsis-Syria Planum magmatism. All these processes are not exclusive, instead, they may have been added and combined, resulting in brittle-ductile deformation in Ophir and Hebes Chasmata.