Tectono‐Magmatic, Sedimentary, and Hydrothermal History of Arsinoes and Pyrrhae Chaos, Mars

Arsinoes and Pyrrhae Chaos are two adjacent chaotic terrains located east of Valles Marineris and west of Arabia Terra, on Mars. In this work, we produced a morpho‐stratigraphic map of the area, characterized by a volcanic bedrock disrupted into polygonal mesas and knobs (Chaotic Terrain Unit) and two nondisrupted units. The latter present a spectral variation, likely associated with hydrated minerals, and they are here interpreted as sedimentary units. The reconstructed geological history of the area starts with the emplacement of the basaltic bedrock, followed by the collapse that caused the formation of the chaotic terrains. Since evidences of volcano‐tectonic activity are widespread across the area (e.g., fissure vents/graben, radial and concentric systems of faults, y‐shaped conjunctions, lava flows, and pit chains), and an intricate system of lava conduits is hypothesized for the occurrence of such features, we propose the possibility that the whole collapse was caused primarily by volcano‐tectonic processes. In a late stage, after the end of the volcano‐tectonic activity, a lacustrine/evaporitic depositional environment could have set, with the deposition of the nondisrupted units. The hydrated minerals found in the periphery of the Chaos could be the result of hydrothermal alteration of the basaltic bedrock.

Chaos, the sedimentary deposits are not observed at all. The lack of studies in this location emphasizes the need to expand our knowledge of this area, providing a broader context on Martian chaotic terrains.

Data, Processing And Tools
The imagery used to perform the geological mapping is provided by the Context Camera (CTX; Malin et al., 2007) and High Resolution Imaging Science Experiment (HiRISE; McEwen et al., 2007) instruments onboard the Mars Reconnaissance Orbiter (MRO). One High Resolution Stereo Camera (HRSC, on board Mars Express spacecraft) image was also used to observe in false color the study area and one HRSC Digital Elevation Model (DEM) was downloaded for the topographic contours in eastern Pyrrhae Chaos, since the area is not covered by CTX stereo pairs. CTX imagery was used as a basemap; in particular, we used a global blended CTX mosaic provided by the Murray Lab (Dickson et al., 2018). HiRISE images were used to observe in detail the stratigraphic contacts in certain areas: Experiment Data Record products were processed and tiled through the USGS software ISIS3 (Gaddis et al., 1997). The data processing was supported by GNU Parallel (Tange, 2011). Variations in thermal inertia were investigated on Java Mission-planning and Analysis for Remote Sensing (JMARS; Christensen et al., 2009) developed by ASU's Mars Space Flight Facility. JMARS is a Geographic Information System (GIS) where different layers can be loaded on a global basemap: it was used to visualize the THEMIS Derived Global Thermal Inertia Mosaic.
DEMs from CTX were computed using Ames Stereo Pipeline (ASP) developed by NASA (Beyer et al., 2018;Moratto et al., 2010): the resulting products are bundle-adjusted to the global topography (Mars Orbiter Laser Altimeter [MOLA]

Table 1 List of Images and Data Cubes Used in This Work
The imagery was then imported into Environmental System Research Institute ArcGIS for the geological mapping. The bedding attitudes were measured using the beta version of LayerTools (Kneissl et al., 2011), kindly provided by Dr. Thomas Kneissl.
The CRISM (Murchie et al., 2007) cubes (S detector-short wavelength channel, and L detector-long wavelength detector) with full resolution (FRT) available in the study area (also listed in Table 1) were processed by means of the software ENVI (with the CRISM Analysis Toolkit extension) where atmospheric corrections and projection were applied, allowing the subsequent visualization and analysis of the spectra.

Scale
The geological map of the area was digitized on ArcGIS at the CTX resolution (∼5 m/px), but the chosen output scale is 1:3.000.000.
Only craters with a diameter larger than 2 km were mapped.

Polygonal Features-Morpho-Stratigraphic Units
The units observed in the study area are five and were mapped as polygons. The units include the three subunits of the Chaotic Terrain unit (Fractured Plains, Knobby Terrain, and High Thermal Inertia Chaotic Terrain) and the nondisrupted units (Light-toned Layered Unit [LLU] and Cap Unit [Cap]). Furthermore, the inner part of the postcollapse craters was mapped as an additional unit: Post Collapse Craters. Following the USGS guideline (2005), we chose warm colors for the Chaotic Terrain subunits given their volcanic nature.
On the other hand, we chose to use cold colors for the nondisrupted units since they were interpreted as sedimentary deposits and green for the craters so that they could be clearly distinguished from the bedrock.

Linear Features-Structural Features and Contacts
The mapped linear features include the crater rims, the contacts between different units and the structural features. The fractures affecting the bedrock and the elongated graben-like depressions were included in a single category called grabens. This choice is mainly due to the coalescence of the structures and the difficulty in distinguishing them. The second group of structural features is represented by the wrinkle ridges. The contacts were divided into certain contacts (continuous lines) and approximate/inferred contacts (dotted lines), for those cases where the mantling covers the contact (e.g., between Cap and LLU) or if the change from one unit to another is transitional (e.g., between Knobby Terrain and High Thermal Inertia Chaotic Terrain).

Surface Features-Pit Chains and Pitted Areas
Pit chains and diffuse pits are widespread throughout the study area. In several parts, the high density of pits did not allow to map individual pit chains, leading to the necessity to introduce a polygonal feature to indicate these areas (Pitted areas).

Symbols
The symbology used for crater rims, grabens, pit chains, wrinkle ridges, certain and approximate contacts was chosen in agreement with the standards recommended by USGS and FGDC for planetary mapping (FGDC, 2006). For pitted areas, we introduced a new symbol.

Morphological Observations and Stratigraphy
The stratigraphic relationships indicate that the oldest unit observed in Arsinoes and Pyrrhae Chaos is the Chaotic Terrain Unit, composed of three subunits: Fractured Plains (characterized by angular flat-topped blocks), the Knobby Terrain (displaying rounded mounds), and the High Thermal Inertia Chaotic Terrain (heavily eroded and characterized by the higher thermal inertia). The Fractured Plains occur predominantly along the rim of the Chaos, while the Knobby Terrain acts as a transition toward the inner parts of the Chaos. The Knobby Terrain subunit is the most extensive material representing the Chaotic Terrain Unit in this area. The Knobby Terrain is in contact with the sharp angular mesas of the Fractured Plains ( Figure 2) and the small mounds of the High Thermal Inertia Chaotic Terrain. The latter occurs only in a small area in the NE part of Arsinoes Chaos, in contact with the Knobby Terrain, and in a crater within Pyrrhae Chaos, in contact with the Fractured Plains. The rounded mounds characterizing the Knobby Terrain are located at lower elevations compared to the mesas of the Fractured Plains ( Figure 2). The scarps bounding the mesas of the Fractured Plains show a depth up to 1 km and the flat-topped blocks appear as irregular polygonal bodies. The mesas of the Fractured Plains subunit show a layering implying multiple depositional events, nevertheless, it is difficult to follow the layers in lateral continuity on the adjacent mesas due to the different erosion and therefore we could not constrain with precision this bedding.
Stratigraphically above the Chaotic Terrain Unit, nondisrupted and layered deposits lie unconformably. In Arsinoes Chaos, the nondisrupted deposits overlying the Chaotic Terrain Unit displays two different morphologies and attitudes. Furthermore, they occur at different elevations showing their stratigraphic relationship. Therefore, they were considered as two distinct units, the LLU and the Cap (Figure 3). At lower elevations, the light-toned deposits are characterized by planar bedding and scalloped surfaces; their morphology resembles the Interior Layered Deposits described by Sowe et al. (2007), Le , and Schmidt et al. (2018). The average attitude of the light-toned layered deposits is ∼S80 E, 20°. The LLU outflanks the mounds of the Knobby Terrain (Figure 4), assuming often a lobate shape and wrapping around the knobs. The overlying sedimentary deposits (Cap) are separated from the LLU by an unconformity: the attitude of the youngest deposits (informally called Cap due to the fact that these deposits "seal" the succession) is ∼N70 W, 3°. The Cap does not seem to be layered or if it is, the bedding is massive and the outcrop visible today represents only one thick bed with a plateau-like morphology. The occurrence of these two nondisrupted units only in some areas of the Chaos leads to the question of whether the current extent represents approximately the original extent or the erosion obliterated a large portion of the deposits that were originally covering the entire Chaos.
In Pyrrhae Chaos, the nondisrupted sedimentary units are not present, although the two adjacent Chaos show a similar depth (in certain areas of Pyrrhae Chaos the depth is even higher than Arsinoes Chaos).
LUZZI ET AL.

Structural Observations
In Arsinoes and Pyrrhae Chaos, as well as in other chaotic terrains, the main structural feature is represented by the deep fractures bounding the polygonal blocks of the Fractured Plains. The fractures are open and show no displacement. Erosion and degradation might have enhanced the opening of these fractures after their formation. The minimum depth of the fractures can be assumed by the height of the mesa from the floor, and it is on the order of few hundred meters, up to 1 km. Additionally, the floor of the chaotic terrain accommodates the nondisrupted units, thus, preventing the determination of the real depth of the blocks. The orientation of fractures bounding the blocks follows two trends that show variations throughout the chaos. The polygonal geometries are due to the irregularly orthogonal disposition of these two trends (Figures 5a and 5b).
A large number of elongated grabens were mapped (Figures 5a and 5b). These confined structures, that may resemble channels at a first glance, are often in coalescence with the orthogonal fractures of the polygonal mesas ( Figure 5c). In addition, grabens are also in coalescence with pit chains and/or occur in areas heavily pitted ( Figure 5d). Considering graben-like depressions and fractures as belonging to the same group of structures, two patterns can be distinguished based on their orientation: one group of structures seems to follow the rim of the Chaos, showing therefore a concentric pattern with respect to the rim of the basin; LUZZI ET AL.
10.1029/2019JE006341 6 of 22 another group shows instead a radial pattern that was likely radiating from the center of the basin outwards, but that is now only visible in proximity of the Chaos rim, due to dust cover and younger overlying units occurring within the basin. When these two sets cross each other, polygonal blocks are defined. The elongated graben-like depressions have a linear or slightly sinuous morphology and do not show any braided system nor meanders. The depth range is between 100 and 400 m (up to 1,000 if we consider those bounding the Chaos), while the length can reach up to 40 km. Moreover, several elongated graben-like depressions display y-shaped bifurcations.
Although the nondisrupted deposits do not have complete lateral continuity, no major faults affecting them were detected, not even at the HiRISE resolution. However, the lack of continuity of these deposits due to erosion means we cannot exclude the existence of younger faults affecting the layered deposits.
Compressive structures such as wrinkle ridges were also observed ∼30 km SE from Arsinoes Chaos ( Figure 6), providing an important clue on the existence of a compressional regional stress prior the formation of the widespread grabens (east to Arsinoes Chaos a wrinkle ridge is cut by a graben). The wrinkle ridges show a typical orientation ∼N-S and are characterized by sinuous/arcuate morphologies.
The last important structural observation concerns two major precollapse craters that were incorporated in the collapse, one in the north-eastern Arsinoes Chaos and another in the south-eastern Pyrrhae Chaos (depicted in Figure 7). The craters are both filled by material affected by the grabens. The infilling has a maximum observable thickness of 1,500 m in the crater located at NE Arsinoes Chaos, while the thickness in SE Pyrrhae Chaos reaches only 1,000 m. The appearance of the infilling deposits seems to not differ from the surrounding basaltic bedrock, suggesting that the impacts predate not only the collapse, but also the last volcanic resurfacing events. The rims of the craters are partially degraded (in both craters, the southern rim is less degraded than the northern rim) and remnants of the ejecta are only visible in the southeastern crater. An exposed lava flow occurs in the crater at SE Pyrrhae ( Figure 8a). The lava flow is darker and less eroded than the surrounding bedrock, less covered by dunes and dust (Figure 8b), and it is overlying all the volcanic deposits, suggesting that it may represent a late volcanic resurfacing event. This last resurfacing event is still predating the collapse as evidenced in Figure 8c, where both the lava flow and the underlying rest of the bedrock are cut by a graben.

Thermal Inertia
The surface temperature of a given area depends on the properties of the exposed materials, but it is also affected by external factors such as dust covering and atmospheric pressure. Diurnal changes in temperature can be detected and described through the thermal inertia, a bulk property of materials defined by the relationship between thermal conductivity, density, and specific heat of the considered material. These properties are different for each material and some of the factors influencing the thermal behavior are for example grain size, cementation, packaging of the grains, and degree of exposure. Therefore, surfaces with different thermal inertia are most likely indicative of a change in composition and/or physical properties. A material with high thermal inertia is able to gather the heat and conduct it beneath the surface during the day, while during the night, the stored heat is released through the surface . In this way, the surface appears cold during the day and warmer during the night. High thermal inertia is typical of consolidated/lithified materials, such as an exposed bedrock, lava flows, indurated, and compact rocks; on the other hand, unconsolidated sands and dust show a low thermal inertia (Fergason et al., 2006). On Mars, thermal information is provided by TES (see e.g., Jakosky et al., 2000), on board Mars Global Surveyor, and THEMIS (see e.g., Christensen et al., 2004), on board 2001 Mars Odyssey. We investigated the thermal inertia visualizing the THEMIS-Derived Global Thermal Inertia Mosaic on JMars. As anticipated in the previous sections, two areas with higher thermal inertia were detected in the study area. The first area occurs in Arsinoes Chaos and corresponds to the most eroded part of the Chaotic Terrain Unit that was identified as the High Thermal Inertia Chaotic Terrain subunit. The differentiation from the Knobby Terrain is transitional and delineating a sharp boundary was not trivial, but the high thermal inertia seems to coincide with the most eroded and peaked knobs. The second area is located in Pyrrhae Chaos and corresponds to the lava flow previously described (Figure 9), located in proximity of a set of grabens. In this case, the contrast is sharp and well-defined and the higher thermal inertia coincides perfectly with the lava flow that differently from the surrounding basaltic bedrock does not present neither mantling nor regolith (except for small dunes within the pits on the surface of the lava flow).

CRISM Spectral Analyses
In the nearby Aram Chaos,   the authors as hematite-bearing, while within other layered deposits, the iron oxide was not found. Moreover, Lichtenberg et al. (2010) provided a stratigraphic and mineralogical characterization of the hydrated sulfates occurring in Aram Chaos, based on CRISM data. The authors identified two sedimentary units: the oldest consisting of monohydrated sulfates intercalated with ferric hydroxy-sulfate or nanophase ferric oxides and the youngest bearing polyhydrated sulfates and crystalline hematite. Monohydrated sulfates were detected by Lichtenberg et al. (2010) observing minor absorptions at 2.1 and 2.4 μm; ferric hydroxy-sulfate was interpreted through absorptions at 2.238 μm, associated with minor absorptions at 1.49, 1.82, and 2.38 μm; polyhydrated sulfates were inferred from the absorptions at 1.9 and 2.4 μm within the youngest unit unconformably lying on the oldest with monohydrated sulfates. Polyhydrated sulfates in association with crystalline gray hematite have been found also by Dobrea et al. (2008) in Aram, Aureum, and Iani Chaos: using the TES and OME-GA datasets, the authors were able to point out a correlation between all these chaotic terrains East of Valles Marineris. Sowe et al. (2012) analyzed CRISM spectra from Aureum Chaos, and, also in this case, the authors were able to identify within the LLUs hydroxylated, monohydrated, polyhydrated sulfates. The hydroxylated sulfates were interpreted by the authors based on the absorptions at 2.23 μm (related to the occurrence of OH), 1.42-1.45 μm and its weak 1.93 and 2.4 μm bands. Monohydrated sulfates (kieserite) were detected through absorptions at 2.12 μm and a major absorption at 2.4 μm, while polyhydrated sulfates were diagnosed based on absorptions at 1.42-1.44 and at 1.92-1.93 μm.
Considering these evidences from the nearby Aram and Aureum Chaos and given the morphologic analogies, we investigate if similar mineralogies can be detected also in Arsinoes Chaos.
LUZZI ET AL.   Only one FRT CRISM cube is available within the inner region of Arsinoes Chaos (frt00008233_07_if164). This cube was analyzed in order to characterize the nondisrupted units overlying the basaltic bedrock. A first distinction of spectrally different terrains was made based on the RGB images ( Figure 10) with combined summary products (Viviano-Beck et al., 2014). First, we examined the minerals revealed by previous works on chaotic terrains. In Figure 10c, the summary parameter for chlorides shows a possible occurrence of hydrated minerals (in yellow). The putative hydrated minerals coincide in extent with the morphologically identified LLU. The RGB composites derived from the parameters that are specifically used to identify hydrated minerals (such as sulfates and phyllosilicates) did not provide a clear distinction of a mineralogical variation. Moreover, the difficulty in finding a spectrum with clear evidences of hydrated minerals is attributed to the detection limit and to the noise affecting the data. Therefore, summary products suggest a mineralogical variation and likely bearing hydrated minerals, but the spectra do not allow the detection of specific minerals. The interpretation of these deposits as sedimentary can only be supported by analogies with the adjacent chaotic terrains and by the morpho-stratigraphic observations.
Other hydrated minerals (CRISM cube frt000196b0_07) were found in the northeastern periphery of Arsinoes Chaos (Figure 11), within a deposit in correspondence of some of the collapse features previously described. Slightly north of this area, we performed also analyses on the bedrock (CRISM cube frt00023790_07 and frt000196b0_07).
The hydrated minerals detected in the CRISM cube frt000196b0_07 are concentrated in the central area of the cube, precisely coinciding in extent with an exhumed deposit slightly different in albedo from the surrounding materials. As shown in Figure 12a, a mineralogical variation is already distinguishable from the infrared false color image where the hydrated minerals appear to be pale green. This first clue is therefore additionally supported by the RGB composites in Figures 12b and 12c. In Figure 12b, the RGB composite PFM is shown for the detection of Fe and Mg in the crystalline structure of the hydrated phyllosilicates, in particular, for significant band depths at 2.3 μm; as a result, Fe-Mg hydrated phyllosilicates are displayed as cyan, allowing to appreciate the mineralogical variation at a first glance. For the ratioed spectrum the numerator is an average of some of the cyan pixels shown in Figure 12b having both band depth at 1.9 and 2.3 micron well above the detection limit, while the denominator is an average of pixels with relatively flat spectra picked from the same column of their respective numerators. The resulting ratioed spectrum is reported in Figure 12, where a comparison with spectra from the CRISM resampled library is provided. Several phyllosilicates sharing similar absorptions (indicative of Fe, Mg-OH) were plotted in order to compare even the weakest absorption to understand the corresponding mineral that may occur in the deposit. We interpreted the ratioed spectrum as smectite, since the ratioed spectrum and the smectite's spectrum share the same absorptions, including the absorption at 2.28 μm that is instead shifted toward >2.3 μm in the other plotted phyllosilicates, due to a bigger concentration of Fe instead of Mg (Clark et al., 1990).
On the same, CRISM cube frt000196b0_07 analyses of the bedrock revealed the occurrence of mafic minerals, in particular pyroxenes with low Ca content. A first identification of the pyroxenes was made by analyzing the summary product LCPINDEX2 (low-Ca pyroxenes; Figure 13). A significant spectrum for the analysis of the bedrock results from the CRISM cube frt00023790_07, where broad absorptions at 1 and 2 μm confirm the occurrence of pyroxenes ( Figure 14) (Viviano-Beck et al., 2014). The asymmetry of the absorption at 1 μm toward longer wavelength could also be associated with the occurrence of a second mafic LUZZI ET AL.    phase as olivine or an high-Ca pyroxene, with variations in breadth due to the amount of Fe and Mg or Ca (Cloutis et al., 1986;King & Ridley, 1987). Pyroxenes and olivine are indicative of basaltic compositions. This observation supports the basaltic nature of the bedrock of the chaotic terrain (Chaotic Terrain Unit).

Morpho-Stratigraphic Map
The observations carried on and described in the previous section led to the creation of a morpho-stratigraphic map of the area (Figure 15).
The units were classified through photo-interpretation, taking into consideration: visible and sharp contacts, transitional contacts, differences in texture, degree of erosion and mantling, morphology, bedding when occurring, relative relationships between units, attitudes, and compositional information where available. The bedrock of the area is identified in the Chaotic Terrain Unit, a basaltic unit subdivided into three subunits: the Chaotic Terrain Fractured Plains (on the map), the Knobby Chaotic Terrain, and the High Thermal Inertia Chaotic Terrain. Because of the volcanic nature of the Chaotic Terrain Unit, the three subunits were mapped with warm colors. An unconformity separates the bedrock from the overlying LLU, while a second unconformity is interposed between the LLU and the overlying Cap that seals the succession. Spectral data showed the possibility that LLU is composed by hydrated minerals, but the detection limit and the low quality of the CRISM cube did not allow a specific classification of the hydrated minerals. On the other hand, hydrated minerals classified as Fe-and Mg-phyllosilicates were found in a small deposit in the northeastern periphery of Arsinoes Chaos, but the limited extent did not allow to represent it on our map at this scale.
Craters in the area with a diameter larger than 2 km were mapped as Postcollapse crater.
LUZZI ET AL.  As for the other features (pit chains, wrinkle ridges, grabens, and crater rims), they were mapped using the standard symbols FGDC. We introduced a new symbol for pitted areas, where the pits were highly coalescent and concentrated making it difficult to map the single feature.
The produced map contributes to fill the gaps regarding Chaotic terrains, but most importantly it highlights the need to expand the availability of maps in this complex area in order to facilitate future works and discussions.
LUZZI ET AL. The band depth at 2.3 μm is significant in correspondence of the hydrated minerals. On the right, the ratioed spectrum is compared to resampled CRISM spectra of Smectite 397S170, Illite LAIL01, and Hisingerite LAHI01 from the spectral library. The black lines highlight the absorptions of the ratioed spectrum at 1.42, 1.92 and 2.28 μm. CRISM cube frt000196b0_07. CRISM, Compact Reconnaissance Imaging Spectrometer for Mars.

Precollapse Events and Floor-Fractured Craters
The information that we gathered on the precollapse stratigraphy is limited to a few observations: (i) the basaltic mesas of the Chaotic Terrain Unit present different layers, suggesting multiple volcanic resurfacing events before the collapse; (ii) the ancient impact craters occurring in NE Arsinoes Chaos and SE Pyrrhae Chaos predate the collapse since they are filled with materials affected by the collapse-related structures; (iii) it is still unclear if the three subunits of the Chaotic Terrain Unit represent just a different lateral erosional evolution of the same material or if the different erosional patterns are due to a lateral mineralogical variations as yet undetected. The craters in NE Arsinoes Chaos and SE Pyrrhae Chaos were likely formed between the multiple eruptive events that emplaced the Chaotic Terrain Unit and afterward they have been embedded into the collapse. The crater in NE Arsinoes Chaos presents a thicker infilling, a more eroded rim and no sign of ejecta: these could be clues of an older age compared to the crater in SE Pyrrhae Chaos that has a thinner infilling and visible traces of the eroded ejecta. We have to keep in mind that both of them are filled by fractured materials; therefore, even though the southeastern crater may be younger, it still predates the collapse and the end of the lava supply that formed the Chaotic Terrain Unit.
This particular type of craters affected by polygonal fractures were classified on Mars by Bamberg et al. (2014) as Floor-Fractured Craters (FFCs). FFCs have been extensively studied on the Moon: the origin of their fractures was attributed to two main processes. One of the discussed processes is the viscous relaxation of the crater topography proposed by Hall et al. (1981) but then proved wrong by Dombard and Gillis (2001); the other proposed process (the most widely accepted) consists of a sill emplacement and consequent inflation (Jozwiak et al., 2012(Jozwiak et al., , 2015Thorey & Michaut, 2014;Wichman & Schultz, 1996). Bamberg et al. (2014) performed an accurate global classification of the martian FFCs, dividing them into two groups: one group of FFCs occur in fluvial areas and were formed (or modified after the formation) by fluvial activity. The second group (that includes our area of study) was attributed to intrusive volcanism: the factors considered by the authors include absence of fluvial morphologies and outflow channels, but also occurrence of volcanic pits, collapsed lava conduits, lava sheets, and basaltic composition of the bedrock. We agree with this interpretation and we propose that processes related to magmatic intrusion could explain on a large scale the collapse of the chaotic terrain itself in absence of aqueous-related surface features.

Bedrock Collapse
The widespread occurrence of grabens leads to question their nature and the possibility that they could be linked to the origin of the collapse. Following Scott and Wilson (2002), grabens are interpreted here as volcanic graben. Similar structures within the Tharsis region, on Olympus Mons and Ascraeus Mons, were interpreted as fissure vents by Mouginis-Mark and Christensen (2005), but also as fluvial-related by Scott and Wilson (1999). Furthermore, grabens similar to that occurring in Arsinoes and Pyrrhae Chaos were mapped and interpreted as vents in Ascraeus Mons (Mars) by Pozzobon et al. (2015). Such vents would have erupted lavas coming from a complex network of dykes, while collapsed pits would be associated with feeder dykes. Dykes were invoked for the formation of grabens and pit chains also in LUZZI ET AL. Pavonis Mons by Montési (2001), where the author describes the possible interactions between dykes and volatiles and the resulting structures based on the degree of interaction. According to Montési (2001), grabens are formed by deep dykes that do not reach the volatile-rich layer but stop at the extensional stage; if the dyke is able to quickly reach the volatile-rich layer, their interaction results into pits; coalescing pit chains would be formed by an intermediate interaction between the dyke and the volatile-rich layer.
In first instance, we rule out the possibility that the grabens may have any connection with fluvial systems: no braided nor meandering patterns were recognized, as well as no tributaries nor typical fluvial depositional morphologies anywhere in the study area. Since outflow channels or other fluvial features are lacking, a volcano-tectonic origin seems more likely, considering the basaltic mineralogies across the area, the presence of y-shaped conjunctions indicative of inflation and at least one evident lava flow spatially associated with the fissures (Figure 8a). Moreover, the grabens have a significant depth (up to 1,000 m), that could be explained by a constant flow (in case of water) and not by a periodic/stagnant river; nevertheless, there are no evidences of such a developed hydrographic system, and the channels terminate abruptly without fluvial deposits nor confluence into other channels. Additionally, we rule out that grabens were formed by groundwater sapping which, as evidenced by Lamb et al. (2006) are unlikely to erode in short time a basaltic bedrock (and if it spanned over a long period then we would expect more hydrated alteration of the basalts). by default because many other factors may influence the formation of the headscarps, and the morphology of the valleys per sē cannot justify such interpretation.
According to Mège and Masson (1996), the grabens occurring in Valles Marineris were formed during Hesperian as a passive rift system, but the authors do not exclude the possibility of an interplay with magmatic processes related to the Tharsis region that occurred in a more recent span of time. Tanaka and Golombek (1989), propose that tension fractures were responsible for the formation of grabens on Mars, particularly, in the area of Valles Marineris. These authors argue that the orientations of grabens and pit chains suggest a structural control that cannot be attributed to volcanic processes since no volcanic deposits occur in correspondence of these structures. Nevertheless, more recent works (with access to younger and better datasets than in 1989) showed the occurrence of widespread basaltic mineralogies within the chaotic terrains , small pitted cones interpreted as of volcanic nature (Meresse et al., 2008), dykes within Valles Marineris (Flahaut et al., 2011;Mège & Gurgurewicz, 2016;Okubo & Schultz, 2005), and lava flows draping the outflow channels (Leone, 2014).
Therefore, based on the mentioned references and on our observations, the grabens that we described and interpreted as volcanic grabens (sensu Scott & Wilson, 2002) may be considered as the result of the collapse of lava conduits, or fissure vents, or due to inflation processes originated from buried magma chambers and/or buried magma intrusions, or as due to the ascent of dykes.
Another surface evidence for such collapses is represented by the extensive pits and pit chains, that at a first glance may resemble small craters but lack in ejecta and raised rims. Pits and pit chains were interpreted in literature as the surface collapse of buried lava conduits (Leone, 2014), as collapse due to magma stoping or as the result of explosive activity, likely due to interactions between magma and H 2 O (Head & Figure 15. Morpho-stratigraphic map of Arsinoes and Pyrrhae Chaos. Okubo & Martel, 1998;Wyrick, 2004), and as the result of the interaction between dykes and volatiles (Montési, 2001). An additional interpretation based on a terrestrial analog was proposed by Ferrill et al. (2011), who observed pit chains in Iceland formed by the interplay of dilational faults, extension fractures, and tectonic caves. The authors suggested a similar combination of mechanisms for the formation of pit chains on Mars. In our study area, extensional faults with clear displacements were not found, the polygonal blocks are different from the horst and graben systems on Earth because arranged in radial and concentric patterns. The pits and pit chains are often coalescent with volcanic grabens, suggesting perhaps extremely unstable conditions of interconnected plumbing systems. Moreover, pit chains that are often coalescent and intersect each other with different orientations suggest excluding a pure tectonic control.
An important topic that we want to mention is the origin of the outflow channels, absent in Arsinoes and Pyrrhae Chaos but occurring in association with other chaotic terrains. Despite the broad consensus of the scientific community on aqueous erosional processes carving the outflow channels (Andrews-Hanna & Phillips, 2007;Baker, 2001;Carr, 1979;Harrison & Grimm, 2008;Hoke et al., 2011;Leask et al., 2007;Meresse et al., 2008;Rodriguez et al., 2005Rodriguez et al., , 2015, recently several authors proposed that the outflow channels might have been carved by lava flows (Jaeger et al., 2010;Leone, 2014;Leverington, 2011). Leverington (2011) lists and discusses the evidence against an aqueous origin of outflow channels: (i) lack of fluvial deposits along the channels, together with deltas and sedimentary shoreline features, while channels seem to be lava draped instead; (ii) incompatibility between the amount of required water and the global mineralogical observations of the unaltered olivine; (iii) paucity of analogs in the solar system of large-sized channels carved by aquifer outburst; and (iv) the required permeability of the megaregolith might not be realistic. In contrast with the aqueous origin of outflow channels, the author highlights how a volcanic origin seems to be the simplest explanation based on the evidences on the martian surface. Koeppen and Hamilton (2008) performed a global analysis of the olivine end-member mineralogies found on the Martian surface: they postulated that the poor alteration of the olivine could be interpreted as due to spatially inhomogeneous wet conditions in the early climate, with very wide dry areas, or to the short-term availability of water on Mars.
Based on our overall observations in our area of study, we speculate that it would be possible that the trigger of the outflow channels could have been magmatic bodies heating the permafrost. The absence of outflow channels in our area of study could be explained by a lateral variation of the permafrost thickness and/or possibly the local fracturing state of the uppermost crust.
Reassessing the extent of the role played by water or lava in forming outflow channels is beyond the aim of this work. The literature reviewed above follows two schools of thought, but we believe that both water and magmatic processes contributed to the final geological setting of chaotic terrains, one prevailing on the other and vice versa at different times.
We observed hydrated mineralogies, especially in the periphery of Arsinoes Chaos, while in the inner Chaos the detection of hydrated minerals is weaker and limited to the RGB composites. The spectra there did not allow the identification of specific phases, probably because their abundance is below the detection limit. Therefore, we confirm that water was present but our interpretation, based on the stratigraphic relationships between the observed geological units and structures, separate the major contributions of volcanic and aqueous processes in two stages: during the first stage, magmatic processes were mainly responsible for the collapse of the chaotic terrain, possibly interacting with volatiles and/or permafrost; in a second stage (postcollapse) water played a major role (perhaps for a short time given the limited extent of the hydrated deposits).

Thermal Inertia
The high thermal inertia of the lava flow occurring in Pyrrhae Chaos could be related to the younger age of the lava flow than the surrounding bedrock. A high thermal inertia is normally observed in lava flows, but the absence of high thermal inertia in the underlying basaltic bedrock (except for the most eroded subunit, the High Thermal Inertia Chaotic Terrain, and for the steep slopes of the mesas, recognizing that mantling debris on slopes may thermally obscure to different extents) is due to the fact that the Chaotic Terrain unit is partially covered by regolith and/or mantling. Our guess is that the basaltic bedrock underlying the lava flow could be older than the overlying lava flow, and therefore more affected by mantling and weathering, while the lava flow is still acting thermally as a rocky surface. The interplay between dust cover and thermal behavior has been already investigated on Mars by Crown and Ramsey (2017), based on a THEMIS IR survey in Arsia Mons. These authors describe the difficulties in discriminating the effect of mantling and albedo from the real thermal inertia, but they were able to identify two types of lava flow: one group with relatively high albedo and large extent, and a second one darker, smooth and smaller, associated with elongated channels and fissures, very similar to the lava flow observed in Pyrrhae Chaos.
Despite the inferred younger age of the lava flow with respect to the surrounding basaltic bedrock, we found that the lava flow predates the collapse since it is cut by a volcanic graben. Thus, the lava flow could be interpreted as the last recognizable volcanic resurfacing event, probably toward the end of the hypothesized multiphase volcanic activity. The end of lava supply could have been responsible for the collapse of the plumbing system generating the volcanic grabens.

Nondisrupted Units
The LLU and the Cap were interpreted as sedimentary deposits first of all by analogy with the nearby chaotic terrains, where deposits sharing the same characteristics of those in Arsinoes Chaos were already interpreted as sedimentary units and spectral analyses were already carried out (Dobrea et al., 2008;Glotch & Rogers, 2007;Lichtenberg et al., 2010;Sowe et al., 2012). Furthermore, we interpreted the nondisrupted units as sedimentary for the following reasons: Lack of volcano-tectonic features, high albedo, planar bedding, more prone to erosion than the surrounding basalts, and mineralogical variation that could be associated with occurrence of hydrated minerals The reason why the NIR absorption features are not as prominent as in other CRISM data from Aureum and Aram Chaos may be due to several factors: a lower abundance of hydrated minerals, absence of areas bereft of hydrated minerals within the CRISM cube (thus the denominator cannot be ideal), lack of totally dustfree areas, features below the detection limit. Especially for the Cap, it was not possible to determine from a compositional point of view the presence of the hydrated silicates detected in other similar sedimentary units because even in the RGB with combined summary products, no significant variation was appreciable.
Nevertheless, the observed variation in the LLU could be consistent with the interpretation of the layered deposits in the adjacent chaotic terrains, except for the high concentration of hematite that was well described in Aram and Aureum Chaos but was not found in Arsinoes Chaos. The hematite in Aram Chaos is considered by  as a fundamental key for a lacustrine interpretation together with the bedding and the closed geometry of the basin and the outflow channels. Despite the lack of hematite (missing or simply below the detection limit) and outflow channels in Arsinoes Chaos, the hypothesis of a lake or evaporitic basin as a depositional environment for the LLU cannot be excluded either. We consider reasonable that a closed and deep basin such as Arsinoes Chaos, that seems to possibly host hydrated minerals, might have been filled by groundwater after the collapse.
It remains uncertain why in Pyrrhae Chaos the sequence ends with the Chaotic Terrain Unit. The depth of the basin is approximately the same as Arsinoes Chaos (even deeper), and their proximity would suggest that the same processes should have acted, but for some reason the deposition of sedimentary units after the collapse did not happen in Pyrrhae Chaos (or the deposits were completely eroded).
The exhumed light-toned deposit found in the north-eastern periphery of Arsinoes Chaos were instead interpreted as related to hydrothermal activity that could have started as soon as the volcanic activity responsible for the collapse has begun to stabilize. This hypothesis is supported by the evidence of smectite within the deposit. In this case, the hydrated minerals are not in a closed basin, but associated with the volcano-tectonic structures that we interpreted as directly related to volcanic collapses. Furthermore, the associations of basaltic minerals (pyroxenes) and hydrated Fe-Mg phyllosilicates (smectite) can be explained by a process of hydrothermal alteration, as summarized by Inoue (1995). According to that author, smectite can be the result of hydrothermal alteration of andesitic to basaltic compositions under neutral or alkaline conditions. On Earth, this has been observed in several geological contexts, including stagnant hydrothermal alteration of caldera deposits (Inoue et al., 1984). The extent of the hydrothermal deposit is limited to 1.3 km, and for scale reasons it was not possible to include it in the geomorphic map. Nevertheless, in the surrounding area light-toned exhumed patches are visible, suggesting an extent of the hydrothermal deposit of at least 10 km.

Conclusion
The morpho-stratigraphic mapping performed in Arsinoes Chaos highlighted the occurrence of two major groups of geomorphic units: the basaltic Chaotic Terrain Unit, further subdivided into three subunits (Fractured Plain, Knobby Terrain, and High Thermal Inertia Chaotic Terrain) represents the bedrock of the area and it is characterized by polygonal irregular mesas and rounded knobs; the second group is composed by two nondisrupted units (LLU and Cap) that were deposited after the collapse of the Chaotic Terrain and lie unconformably on top of the bedrock.
The LLU is characterized by a planar bedding and the scalloped surfaces show high albedo; the Cap seems instead to be a single thick layer (plateau-like) and lies unconformably on top of the LLU. In Pyrrhae Chaos, the nondisrupted units are missing.
The spectral analyses performed in Arsinoes Chaos could not entirely confirm the morpho-stratigraphic evidences for the sedimentary deposits with the presence of hydrated phase similar to the case of Aram and Aureum lying on the bedrock of the Chaotic Terrain Unit. Nevertheless, a mineralogical variation is present and the existence of hydrated minerals below CRISM detection limits cannot be ruled out.
The bedrock was confirmed to be basaltic in composition due to the occurrence of mafic minerals, in particular, low-Ca pyroxenes; the CRISM analyses did not reveal crucial information on the Cap and LLU, but we detected hydrated minerals hosted by a small deposit located in proximity of the collapse-related structure in the northeastern periphery of Arsinoes Chaos and we identified hydrated Fe-Mg phyllosilicates (likely smectite). While the LLU may be explained by a lacustrine/evaporitic environment that could have been established after the collapse and the stabilization of the volcanic activity, the hydrated minerals occurring in the northeastern periphery cannot be explained by such a hypothesis because they are not placed into a closed basin, but they drape the volcano-tectonic structures interpreted as volcanic graben. For this reason, the most likely hypothesis involves a hydrothermal system where hot water rises through the fractures, deposits hydrated minerals, and alters the preexisting basaltic bedrock. The structural evidences of volcano-tectonic activity support this interpretation, suggesting that after the collapse, the volcanic activity may have turned into a hydrothermal environment warming up the groundwater that was infiltrating the fractures and finally reaching the surface with processes of deposition and alteration. Given the lack of evidence for aqueous activity pre and syn-collapse, and the occurrence of widespread volcanic features, we infer the collapse to have been triggered primarily by volcano-tectonic processes, possibly interacting with volatiles and/or permafrost. Further studies will reinforce or rule out this hypothesis.

Data Availability Statement
The processed CRISM cubes and the geopackage of the morpho-stratigraphic map, including vectors and the raster of the basemap was stored in a repository following the FAIR principles (Luzzi et al., 2020). The CTX mosaic is described by Dickson et al. (2018) and since it is a beta version it does not have a DOI yet. A grid shapefile with the tiles and corresponding links can be downloaded at http://murray-lab.caltech. edu/CTX/tiles/beta01/Murray-Lab_CTX-Mosaic_beta01_QuadMap.zip. Alternatively the single CTX images can be downloaded at https://ode.rsl.wustl.edu/mars/indexproductsearch.aspx after setting the adequate parameters (MRO, CTX, Arsinoes Chaos, Pyrrhae Chaos). The HiRISE dataset is included in McEwen et al. (2007). The HRSC dataset is included in Walter and van Gasselt (2014).