Small amounts of dissolved salts increase the mobility of mud flows on Mars and other extraterrestrial bodies

Abstract

Sedimentary volcanism is a widespread geological phenomenon on Earth. Similar processes are theorized to occur on extraterrestrial bodies like Mars, potentially representing surface expressions of subsurface liquid water reservoirs. As we are currently missing ground truth, recognizing signs of extraterrestrial mud volcanism is based on comparative analysis with terrestrial mudflows. Despite the ubiquitous presence of salts on planetary surfaces, it remains unclear how different dissolved salts in mud mixtures influence the behavior of erupted mud. Our study integrates laboratory experiments, thermodynamic calculations, and rheometry measurements to investigate how concentrations of NaCl, MgSO₄, Na₂SO₄, and CaSO₄ affect such behavior. The results show that 10 wt.% MgSO₄ and only 2.5 wt.% NaCl maximizes mud propagation, producing ropy patterns and narrow flows, while higher salt concentrations result in sheet-like flows covering wider areas. When the mud is supersaturated in salt, the liquid state is prolonged but propagation is reduced. Hence, mud salinity has a significant influence on mud flow behavior in low-pressure environments, reflecting the balance between anti-freezing effects and viscosity-compositional effects. Our experimental results are limited to bentonite muds, yet the sensitivity to salt type/concentration is also expected for other muds, leading to a much broader range of morphology, longevity and spatial dispersion than previously assumed.

Introduction

Sedimentary volcanism is a geological phenomenon that can create volcano morphologies on the Earth’s surface^1,2,3. Extraterrestrial mud-volcano-like (MVL) structures represent key targets for studying the hydrology and potential habitability of subsurface environments on other planetary bodies. At Mars and Ceres, where surface conditions are hostile to known life^4,5,6, the subsurface might be the most likely habitable environment. Yet only certain salts are particularly suited to capture and preserve organic material, including possible biosignatures^7,8,9,10,11. Thus, any surface expression of subsurface fluids, such as a mudflow, could contain information regarding the habitability of otherwise inaccessible environments. Therefore, identifying the potential morphological signatures of the MVL landforms and unraveling mudflow behavior beyond Earth^{12,13,14,15,16,17,18,19,20,21} is important to understand the nature of aqueous environments within the whole Solar System.

On Earth, mud volcanoes typically occur in petroleum basins where gravitationally unstable, rapidly buried sediments trigger the formation of subsurface diapirs. These are also fueled by hydrocarbons generated in deeper organic-rich source rocks where kerogen maturation occurs¹. The sediments extruded during mud volcano eruptions result in a variety of meter- to kilometer-scale morphologies characterized by complex mudflows of various shapes^1,2,3. Both external and internal forces control the ultimate mud propagation process. External environmental forces include the atmospheric pressure and temperature, the gravity field, and surface-substrate properties, while internal forces include the composition, temperature, and other related physical properties of the mud itself^22,23,24. For example, muds enriched in clay minerals (such as bentonite-based mixtures) have rheological properties that are particularly sensitive to any additional chemical constituents and to temperature variations^25,26,27.

On Mars, even though we do not know their exact mechanism of formation, the existence of MVL has been proposed in many regions, based on morphological similarities with terrestrial mud volcanoes^4,28,29,30 and is currently discussed in context of remote sensing spectroscopic datasets¹⁸. Lacking ground truth, the existence of MVL is ambiguous, and igneous volcanism is often cited as an alternative explanation^12,31,32. The ambiguity also arises from the uncertainty in the physical-mechanical properties of muds under low-pressure conditions, and thus the resulting expected MVL morphologies. Experiments performed under Martian pressure conditions^22,23 have shown that in case of “cold” surfaces (−15 °C), low viscosity mud with salinity 0.5% propagates like pahoehoe lava on Earth²², while in case of “warm” (+20 °C) and unconsolidated surfaces, mud “levitates” and hence can be transported to longer distances²³. These behaviors differ from what is observed on Earth and are caused by the instability of liquid water under the <5–12.4 mbar atmospheric pressure of Mars³³. The fast pressure drop when mud is erupted from the subsurface, causes a sequential phase transition from liquid phase into the vapor and solid almost instantaneously, affecting the subsequent mud behavior^33,34.

As shown previously^22,23, the dynamics of mud cooling—including its phase transitions and flow mechanical properties—depend on external environmental factors such as pressure, temperature, substrate character, amongst others. However, it remains poorly understood how internal properties, such as mud composition, effect mud behavior. On Mars, the presence of a wide range of inorganic salts has been detected^{35,36,37,38,39,40,41,42,43,44,45} (Fig. 1A), meaning that mud slurries likely contain some dissolved ionic solutes. Salts are also a key component of extruded material on Ceres⁴⁶ and at other icy bodies within the Solar System, e.g., Enceladus⁴⁷ and likely Europa⁴⁸. Dissolved salts can dramatically affect both the boiling and freezing point, heat capacity, and rheology of mud flows as observed from the behavior of clay-rich soils on Earth^{49,50,51,52,53,54,55}. Yet, it still remains unknown how different salts and their concentration affect the flow behavior of muds exposed to low-pressure environments. We hypothesize, based on theoretical considerations, that salts should have a significant impact on the combined thermodynamic and thermo-mechanical behavior of mud-brine slurries during their depressurization and should, in turn, impact the flow properties, the final morphologies of the resulting mudflows as well as their dimensions. We also expect that the anti-freeze effect (Fig. 1C), together with a modified rheology, may extend the propagation of mud flows exposed to cold surfaces in a low-pressure environment.

Fig. 1: Overview of salts and water presence on Mars.

A displays distribution of selected salt types across the Martian surface, with regions marked by a gray shade being where spectral methods are ineffective due to dust cover. Yellow circles represent selected sites where mudflows have been previously reported⁴. The red stripe in the north polar region outlines the approximate extent of the gypsum detected in the Olympia Undae dune sea⁴⁵. B shows water ice distribution within the top meter of the Martian surface as inferred from gamma-ray spectroscopy⁸⁹. C displays the eutectic temperature and phase diagram for NaCl brine, highlighting how the presence of dissolved salts can lead to a depression of the freezing point, to a minimum temperature defined by the eutectic. Surface and water maps are modified from the website of the Center for Planetary Sciences (https://planetary-science.org/) and the phase diagram is inspired from data at https://saltsmart.org/. Distribution of salt types is drawn from⁸⁹, Background maps are provided by NASA under a free license.

To test these hypotheses, we conducted environmental simulation experiments to investigate the behavior of mudflows under low atmospheric pressure using different dissolved salts (NaCl, MgSO₄, Na₂SO₄, and CaSO₄) at concentrations between 0.5 and 15 wt.%. We performed complementary experiments to isolate the effect of rapid and slow pressure drops on (a) boiling, (b) evaporative cooling, and (c) freezing of pure binary salt-water solutions.

Results

Mudflow experiments

The experimental procedure adopted that of Brož et al.^22,23 and was also performed in the Large Mars Chamber at the Open University, UK—please see the Methods section for details. During the flow experiments, when the desired pressure was reached (5.9 ± 1 mbar), 500 ml of mud was released onto a pre-cooled inclined sand surface. Once the mud touched the sand, the mud flow started to develop. During this process, the evaporative cooling operating within the mud was accompanied by a conductive heat transfer between the mud and sand. From that moment, the behavior of the individual mud flows varied based on the content and type of dissolved salt (Fig. 2). It should be noted that the salt concentrations quoted in the text represent the theoretical concentrations assuming fully dissolved salt prior to depressurization and do not reflect how this concentration further evolves.

Fig. 2: Plan view of representative mud flows for each salt and the tested concentrations.

The experiments are organized to show important transitions in morphology with increasing salinity (a–c, d–f, g–i, j–l). For example, >2.5 wt.% NaCl (e) typically results in fast and thin flows without any significant change at higher salinity. This means surface features that are typical for 10 wt.% MgSO₄ (b) is also produced by muds with 0.5–2.5 wt.% NaCl (d, e). Thin dashed lines are the boundaries from replicate experiments (scale in I applies to all panels). Note: this study focuses on the processes and generalized outcomes rather than precise dimensions and shapes. Therefore, variation in shape shown by the replicates does not compromise the statistical robustness of our findings. Please also refer to the experimental videos included in the supplementary files (Movies 1–13) and supporting materials: https://doi.org/10.5281/zenodo.12801318 and to Fig. S6. Displayed concentrations are hypothetical if all additive salt is fully dissolved in the volume.

The experiments using MgSO₄ in 0.5, 5, 10 and 15 wt.% concentrations (Fig. 2a–l, Fig. 3a–l) showed that the lowest concentration led to the development of multiple, short, and isolated mud lobes with the rapid formation of an icy-muddy crust. The individual lobes typically flowed slowly and their downslope propagation over the sand was in the form of repeated effusions from broken crust at its toe. The final lengths of the mud flows were from 25 to 40 cm and they show a complex undulating surface geometry (Fig. 2a). Their internal structure consisted of an ice-mud rim and a central region consisting of a cavity containing liquid mud (Fig. 3a). As salinity increased (5–10 wt.%), the length of the mud flow increased (~50–80 cm) and the geometry of the flow changed from multiple, short lobes into one narrow lobe (Fig. 2b). These narrow lobes had a ropy pattern (Fig. 3b) reflecting nonhomogeneous free-surface Couette flow^56,57 altering between solid (icy) slabs and liquid surface, and were also characterized by stepwise-formation of lobes at the toe. The maximum salinity (15 wt.%) led to the shortest (~35–50 cm), widest and sheet-like flows with significantly delayed freezing and formation of an icy crust (Fig. 2c). The flows using MgSO₄ concentrations > 0.5 wt.% were completely solidified after chamber re-pressurization, with trapped bubbles that were larger in size closer to the surface (Fig. 3c). The frozen interior was characterized by vertically-oriented sheets of mud and ice/salt crystals (Fig. 3b).

Fig. 3: Vertical sections through the mud flows showing their internal structure taken after chamber re-pressurization.

Each panel displays frozen mud flow in correspondence with the related salt and its concentration (a–c, d–f, g–i, j–l). The sections are taken at the middle part of the flows and correspond to the same concentrations as also displayed on Fig. 2. Scale is provided by ruler on each panel (a–l).

Mud mixtures containing NaCl (tested at 0.5, 2.5, and 10 wt%; Fig. 2d–f, Fig. 3d–f) led to similar trends as MgSO₄. However, NaCl mixtures had comparable features at lower salt concentrations (Figs. 2e, f, 3e, f). For 2.5–5 wt.% NaCl flows dispersed faster than at the corresponding MgSO₄ concentration and spreading occurred without formation of lobes at the toe (for details, see Supplementary Movies 1–13). The highest concentration of salt led to the slowest movement of mud, but preserved it in a liquid form for the longest period of time.

The Na₂SO₄ mixture was tested at concentrations of 0.5, 5, and 10 wt.%. The lowest concentration led to the formation of short and complex mud flows, represented by narrow single lobes (Figs. 2g, 3g). 10 wt.% Na₂SO₄ led to delayed freezing and larger spreading (i.e., consistent with 2.5–5 wt.% NaCl or 10–15 wt.% MgSO₄). This salinity also produced longer (~40 cm) mud flows with alternation of flat and undulating topography, with a ropy pattern on top of the flow toe (Fig. 2h–i, Fig. 3h–i). At ≥5 wt.%, salt precipitation is visible as a whitish coating on the surface of mudflows, which based on theoretical calculations (Fig. S5) likely forms initially as mirabilite but then may dehydrate to thenardite in the low-pressure conditions at the surface of the flow (see also the internal structure on Fig. 3). Such color changes were also seen for the other muds and vary slightly depending on each salt type and concentration (Figs. 2 and 3).

Finally, we tested concentrations of 0.5, 5 and 10 wt.% CaSO₄. Regardless of the concentration used, the resulting flows featured a system of multiple, narrow, and short lobes (Fig. 2j–l) with superposed channels (Fig. 3i–l). For all concentrations, the mud remained partially liquid in the form of thin sheets, captured inside the solidified lobes (Fig. 3l).

Rheology of mud mixtures

The flow experiments exhibit a wide range of morphologies controlled by phase changes and mechanical properties reflecting the mud rheology. For these reasons, we conducted isolated investigations of the rheological properties of the mud-brine mixtures (see Methods for description of rheometry procedure). The above-described flow experiments represent a complex physical problem, where the pressure and temperature drop changed the proportions of clay-water-salt. To tackle this, we initially reduced the complexity of the rheometry investigations focusing on the initial states of the muds before their depressurization and obtained rheological curves under standard laboratory conditions at 25 °C (Figs. 4, and S3; see Methods and Supplementary texts: notes 1–2 for additional information).

Fig. 4: Thermo-mechanical behavior of the binary solutions and mud suspensions.

a The temperature changes at the bottom of low-concentration binary samples over time. b The temperature changes for the high-concentration samples for fast (pale colors) and slow (bright colors) pressure drops. Dashed lines correspond to the approximate time when the pressure of 20 mbar was reached and the dotted(-dashed) curve(s) represent the pressure drop itself. The boiling corner is an approximation of the onset of boiling (1st phase transition) in the dominant part of a sample volume, associated with a rapid temperature drop. The terminal temperatures denote the freezing (2nd phase transition) of the sample and the restoration of equilibrium below the isolation (ice) layer when boiling is terminated and the temperature stabilizes. c Results of the p-T modeling with projected experimental data of the tube experiments. Solid lines are interfaces between the stable liquid-vapor fields (at 1 bar) and dashed lines represent the H₂O liquidus where ice and brine mixtures are formed. Calculated and experimental data are both low-concentration (pale) and high-concentration (bright) solutions. Dots parallel to the dotted lines correspond to the terminal temperatures when the sample is re-pressurized. d Results of the rheology tests for the bentonite-salt-water suspensions. Displayed are approximated envelopes for the individual fields of the viscosity-shear rate dependence diagram (in cut-detail derived from Fig. S3 in the Supplementary texts). NaCl and MgSO₄ have a unified envelope for all the concentrations due to small variations of the viscosity dependence (the detailed view showing viscosity ranges is provided in the inset). The pure bentonite salt-free suspension that is not shown in this detailed inset has typically viscosity of 1–10 Pa.s for decreasing shear rate in the range of 80–10 s⁻¹ (for raw rheology data and saltless suspension see the Supplementary texts: notes 1 and 2).

All the bentonite suspensions behave as non-Newtonian fluids, exhibiting shear thinning behavior. In addition, two nonlinear dependencies were also seen for the composition itself. The first one reflected the type of salt while the second one was determined by the salt concentration. Typically for deflocculation process, with increasing concentration of salt, viscosity firstly rapidly decreased up to optimal salt concentration and then slightly increased with further increasing of salt content. This trend holds in the case of suspensions containing MgSO₄, NaCl, and CaSO₄ (Figs. 4, and S3).

The less obvious correlation was observed in the case of Na₂SO₄ mud, where only viscosity decrease was observed with increasing salt concentration. For example, the viscosity was almost one order higher for the 0.5 wt.% compared with 10 wt.% (shear rate ~60 s⁻¹). The viscosity of muds containing NaCl, MgSO₄, and CaSO₄ was even half an order lower compared with the 10 wt.% Na₂SO₄ mud. In addition, we expect the viscosity drop of these muds (with rapidly increasing shear rate) further continues and converge to similar final values below 1 × 10⁻²Pa.s.

Depressurization-cooling of binary solutions

We isolated the behavior of the binary (brine) component in the mud mixtures during the pressure drop by performing complementary tube experiments (see the Methods section for details). We performed 31 experiments with pure brines to compare p-T-t paths influenced by (a) various concentrations of a single salt, (b) identical concentrations of various salts, and (c) fast vs. slow pressure drops, to mimic fast/slow ascent of the brine in a column (for 0.5 wt.% and 10 wt.% samples).

The typical evolution of the solution in the tube had 3 stages: (1) accelerated degassing, which was typically observed between ~70 mbar and the boiling point, (2) boiling and cooling, which depended on salt concentration, initial bulk temperature, and rate of the pressure drop starting at ~28–6.5 mbar, and (3) development of an ice lid or bulk freezing and salt precipitation, which mainly occurred in the range of pressures from 6.5 to 1 mbar. The boiling points had differences of 1–2 s (and 2.8 mbar) between the MgSO₄ and NaCl for the smallest (0.5 wt.%) concentration, while the highest (10 wt.%) concentration leads to 6–9 s (and ~7 mbar) difference between the CaSO₄ and NaCl solution (Fig. 4a, b). The slow pressure drop amplified this effect and extended the difference by an additional 2–14 s (Fig. 4b). This effect of various boiling points and corresponding cooling rates was visible in the experiments comparing individual salts with the same concentration (Table S4). Here the maximum difference was found for the NaCl solution (7–10 s) while the smallest difference was found with CaSO₄ solutions (1–2 s).

We observed that the freezing point was more sensitive to the concentration and type of salt than the boiling point. NaCl had the strongest delay of freezing (Fig. 4a, b) with the maximum delay from 0.5 to 10 wt.%, and the minimum delay was for 10 wt.% MgSO₄. In addition, the formation of salt-water crystal mush in the brine column also locally increased the temperature by releasing latent heat, as seen for the Na₂SO₄ temperature curves (Fig. 4b and Supplementary movies 14, 16) but not for other salts.

Thermodynamic modeling and p-T space of binaries

The experiments were supplemented by thermodynamic phase equilibria modeling of binary (brine) solutions. For corresponding setups and techniques see the Methods and for detailed results the corresponding Supplementary texts, Fig. S5.

Calculations predicted that, upon complete freezing, each salt solution should form solid phase assemblages comprising ice and a single salt mineral (Fig. S5 and Table S3). At both 0.5 and 10 wt.%, calculations show that following ice formation, but before salts precipitate, NaCl solutions should retain a greater liquid volume fraction at a given temperature than other salt solutions (Fig. S5). At 10 wt.%, MgSO₄ solutions have the second largest liquid volume fraction until the eutectic temperature is reached, whereas at 0.5 wt.%, Na₂SO₄ solutions instead exhibit the second-largest liquid volume fractions. Calculated phase equilibria (Fig. 4c) show that boiling in the experimental solutions, indicated by the onset of temperature decrease at ~18 °C, occurs soon after pressure drops below the liquid-vapor boundary in all cases. The minimum temperature achieved by the 10 wt.% NaCl experimental solution is at least 5 °C lower than all other experiments and agrees well with the predicted ice formation temperature at ~−6 °C. The 10 wt.% Na₂SO₄ and MgSO₄ solutions achieved the second and third lowest minimum experimental temperatures, consistent with the relative positions of their ice formation temperatures as predicted by thermodynamic calculations.

Discussion

Thresholds in morphology of mudflows

As can be seen from Figs. 2 and 3, the morphological pattern formed by 0.5 wt.% concentration is similar for all the tested salts and corresponds to previously published results²². This indicates that the low-viscosity (highly water-saturated) and low-salinity muds propagate under martian pressure in a Pahoehoe lobe style independent of the type of salt. This is valid when the salt concentration of tested samples is below 2.5 wt.% (variations between 0.5 and 2.5 wt.% were not tested). However, when salt concentrations increase, the flow properties diverge, resulting in contrasting flow morphologies. With one exception; CaSO₄-rich mud (Figs. 2j–l, and 3j–l) also produces flows similar to low MgSO₄ mud experiments presented in earlier study¹⁴, with its mobility only slightly affected by increased salt concentration. Such flow style is likely because this salt has a low solubility and is supersaturated even at 0.5 wt.% which implies that CaSO₄ solutions likely had the same true salinity set by gypsum saturation, with the 10 wt.% solutions exhibiting a denser suspension of precipitated salt (Fig. S5). In contrast, other salts show different behavior and morphological features depending on the salt type and concentration, and higher salt content is also reflected by a progressively altered surface color, indicating the precipitation of secondary salts (Figs. S5, S6). All this suggests that salts can play an important role in shaping mud-volcano-like structures under low-pressure conditions⁴. As such, variations in salt types and concentrations might help to, at least partly, explain the large variability in the shapes of putative Martian mud volcanoes¹² and their spectral variations^58,59.

Anti-freeze effects on mud propagation

The presence of dissolved salts both lowers the ice formation temperature of solutions and ensures that liquid brine can remain even after water has undergone phase changes into vapor and ice. Tube experiments and thermodynamic calculations both support the observation that solutions with lower freezing and boiling points lead to greater flow mobility (Fig. 4c), likely by slowing the rate at which the mud solidifies, or preventing its complete solidification entirely. In the case of NaCl-rich flows, we interpret this greater mobility to be a consequence of (1) the lower ice formation temperature, (2) greater resistance to low-pressure boiling due to a lower saturation vapor pressure, and (3) a greater retained liquid fraction upon ice formation in such salt solutions (Figs. 4c and S5). The anti-freeze effects are further highlighted by the tube experiments, in which boiling of the higher salinity NaCl, MgSO₄, and Na₂SO₄ solutions continued to near the predicted ice formation temperature of the salt solution, below the freezing point of pure water (Fig. 4c). In general, the anti-freeze properties of muds and coarse-grained soils with salt are also widely recognized on Earth^60,61 and the freezing temperature is influenced by factors such as clay content, effective porosity, and salt concentration in the pore fluid^50,52,60,61. However, the freezing point is consistently higher for the entire bulk volume consisting of a certain ratio of clay, ice crystals, specific salt crystals, and brine than for pore-confined brine itself⁵¹. Rearrangements of acting van der Waals forces in between molecules of clay sheets⁶², ice, and salt crystals^{49,62,63,64,65} are expected to change clay facet contacts defining macroscopic bulk properties^62,64 (such as for example porosity, capillarity, etc.) and therefore further affect thermodynamic parameters like Gibbs free energy, enthalpy, and entropy. In our study, this rearrangement can contribute to the delay of mud freezing when exposed to low pressure. On the other hand, the supporting mud cooling experiments (see Supplementary texts: discussion 1–2, Fig. S7) have shown that the cooling trends of mud samples are mutually similar to the cooling of their pure brine component but with freezing temperatures about 1–1.5 °C higher but with the boiling corner at ~15 mbar.

The anti-freeze effects are well documented on Earth⁶⁰, experimentally studied in low pressures⁶⁶, and in our case notably facilitate the extensive spatial dispersion of NaCl muds, leading to the formation of wide, sheet-like flows that remain in a completely liquid state for the longest duration compared to all other tested mud samples. In contrast, muds containing MgSO₄ salts up to 10 wt.%, characterized by narrow, long and slowly propagating lobe-shaped flows, have a very similar texture to rope lavas. Their local propagation resembles that described by ref. ²² for 0.5 wt.% of MgSO₄, although these authors document mud extension more than two or three times smaller in comparison to our 10% concentrations. That generally reflects earlier freezing of MgSO₄ brines and a smaller amount of liquid water in the system (see Fig. S5), causing fast formation of a protective ice-salt-mud crust at the top of the flow. The Na₂SO₄ mud anti-freezing behavior is a “transitional” between CaSO₄ and MgSO₄ mud.

Although our experiments focused on a limited amount of mud (500 ml), they show that notable disparities in the flow appearance, based on anti-freezing properties, should be expected from our terrestrial experience when mud would be exposed to a low-pressure environment(s). Our experiments also suggest that muds with increased salt content can reach significantly longer distances and cover larger areas compared to saltless muds. In addition, results of tube experiments suggest that processes in mudflow close to effusion sites are influenced by rate of brine depressurization, where a faster pressure drop affects volume-crystallization-thermal dynamics. Thus, the entire mud flow can be significantly pre-organized from its initial exposure to low pressure. The rate of pressure drop can be relevant also for effusive cryolavas or slurries with low clay content, which are expected on Ceres or icy moons^47,48. The movement through conduits and extrusion rates can therefore be an important factor in later spatial dispersion on the surface. To illustrate this point, fast depressurization can facilitate the rapid crystallization of smaller rather than larger crystals, leading to the immediate distributed release of latent heat, prolonging liquid state but possibly blocking the further flow. Despite recent studies investigating such nontrivial interplay on microlites formed in Earth’s magmas⁶⁷, this problem remains poorly understood.

Impact of rheology on dynamics of salty mudflows

Despite the fact that our rheological measurements were performed under Earth surface atmospheric pressure, combining experimental observations with molecular theory^{18,44,45,51,52,53} we assess a direct correlation between saline muds and their ability to reach more extensive distances, displaying a complex temporal evolution of flow structure in low-pressure. As it is known^{26,27,51,52,53,54}, intermediate salt concentrations produce conditions that maximize flow propagation. Typically, the NaCl and MgSO₄ muds are able to flow over longer distances, but only if they are not oversaturated by salt, with maximum flow capability at 2.5–10% concentrations. This non-linear trend is attributed to the repulsion effect between mud particles surrounded by salt ions creating a diffuse double-layer^{26,27,68,69,70,} (as illustrated in Fig. 5a) and is also highlighted by viscosity drops and decreases of rheological curves (Figs. 4d and S3). With rapidly increasing salinity, these mudflows are slowed down due to increased viscosity and their surface appearance reflects the typical ropy pattern. However, due to the anti-freeze effect, and the presence of retained brine even after ice formation, the bulk of the mudflow still remains in liquid state for a longer period.

Fig. 5: Interpretation of the rheological and thermodynamical effects on mobility of the mud mixtures.

a shows the increase and decrease of the mud bulk shear velocity as it depends on salt content. With additional salt, the optimum amount of salt leads to surrounding of clay particles by salt ions and results in the formation of a diffusion layer (e.g., between clay tetrahedral structure), which causes repulsion effect between mud particles and therefore mud mobility increases. Simultaneously, higher salt concentration delays freezing of the mixture. Higher than optimum, the concentration of salt overcomes the repulsion effect due to a large amount of ions in full volume of solution (around montmorillonite tetra-octa-edric sandwich) and thus preventing further gliding of the clay sheets, resulting in a slight viscosity increase. Low mechanical mobility keeps the mud in a steady regime and allows faster surface freezing while the internal parts of the flow are liquid for a longer time. b is the interpretation of how the individual salts in various concentrations (sorted from lowest to highest effect) impact the freezing delay during the pressure-temperature change. The mobility potential (estimated from the flow experiments and rheometry for current salt) is displayed by degree of color saturation. The internal molecular structure (without marked distribution of salt molecules) is shown in the small circles which reflect the gradual pressure and temperature drop (logarithmic drop is marked by increasing distance of side arrows). For simplicity, we reduced the montmorillonite tetrahedron-octahedron sandwich, and just tetrahedrons within the insets are shown. Even a large concentration of the CaSO₄ has a small effect on the overall anti-freeze potential while at small concentration of the NaCl allows significant pressure-temperature drop and sustains the liquid state of the suspension for a significantly extended time. c three main contrasting evolutionary trends in dependence on salt type and concentration: 1. multiple short flows that form as parallel lobes, 2. single long-narrow flow - that forms as serial rope-shaped lobes for 10% MgSO₄ and lobe-less gently roped pattern for 2.5 NaCl, and 3. single wide and longer “sheet-shaped” flow. The most effective propagation is observed with a combination of 10% MgSO₄ and 2.5% NaCl, where the synergistic effects of minimum viscosity and lowered freezing point are most pronounced.

The MgSO₄ in its average concentrations (5–10%) is the only salt that develops localized ropes in the form of stacked lobes at the actual flow front. As revealed by the tube experiments, the Na₂SO₄ and CaSO₄ salt additive has less impact on freezing delay, however, increased Na₂SO₄ content leads to a significant viscosity drop and increases the spreading potential. Furthermore, even if Na₂SO₄ prevents freezing similarly to MgSO₄, the same concentration leads to increased viscosity and rapidly reduces mud mobility. Ultimately, the crystallization of secondary minerals within the crust of the mud flow firstly delays freezing due to latent heat release, but later inhibits further fast movement as crystals and mud sheets are effective mechanical obstacles (we refer to the Supplementary texts: discussion 3 for our speculations on this effect applying to the rheological discrepancy of Na₂SO₄ muds). However, due to lack of experimental data, it was impossible to precisely evaluate the direct impact of the documented crystallization on local temperature increases, nor to investigate the microstructure of the formed salt-water-clay crystal mush in the mud volume.

Flow experiments reflect this salt-induced mud viscosity reduction revealing increased mobility and spatially more distributed flows with a simpler surface morphology. This effect holds, however, until mud is (over)saturated by salt. We, therefore, suggest that the transitional regimes in the flows (between the slow and fast motion, and between the simple-flat and complex-undulated topography) are driven by a combination of the anti-freeze effect and non-linear changes in mud bulk rheology with increasing salinity. Further interpretations are impaired by the absence of knowledge regarding temporally dependent factors, such as fluctuations in salinity resulting from water loss through evaporation from the mud. Therefore, it is likely that flow rheology undergoes substantial spatiotemporal evolution, leading to the formation of heterogeneous thickness and porosity in the protective crust composed of ice and crystallized salt (see also insets on Fig. 3). As the salinity increases, the brine component in the shallow parts of the flow may become even more stable for extended periods of time for sulfates and chlorides, as also demonstrated by recent experimental and kinetics models⁷¹.

Moreover, we observe that the rate of pressure drop can have a substantial impact on thermal evolution of brines and muds, suggesting the importance of non-equilibrium thermodynamics, as evidenced by varying cooling effects observed during two contrasting rates of depressurization (Fig. S4 and Table S2). The data also highlights the significance of salt concentration, particularly in low-gravity environments, where a reduced gravitational influence can result in slower mud flow rates and a wider range of viscosities dependent on this flow rate. For instance, in the Martian environment where NaCl and MgSO₄ salts are present^41,42,43,44, similar flow properties to those of NaCl mud on Earth can be achieved with a concentration of MgSO₄ four to six times higher. At smaller bodies within the Solar System this difference could be even higher. However, this study does not delve into the precise impact of gravity on the flow behavior of the mud, and the actual influence of various gravity fields on mud propagation remains uncertain.

Summarizing the previous analysis, the most important observation is the synergistic effect of locally decreased viscosity and anti-freezing effect. Muds with completely distinct salts and their mutually contrasting concentrations (though not necessarily the highest) can achieve a similar degree of maximum spatial dispersion (Fig. 5c). However, the spreading of these muds occurs through two main differing styles: open channel flow and serial lobe-rope propagation. This is best illustrated by the example of MgSO₄ and NaCl mud. Both muds have maximum spreading potential for 10% (MgSO₄) and 2.5% (NaCl) of the salt, but flow differently in the same low-pressure conditions. In contrast, the presence of low-solubility salts, such as CaSO₄, or the precipitation of secondary salts like Na₂SO₄ (mirabilite or thenardite) can result in rapid freezing. This occurs due to the high concentration of precipitation sites within the volume and resultant flows are spatially limited.

Experimental limitations and applicability

Our experiments have certain limitations that may affect their predictive capabilities. The mud suspensions used are simplified, consisting solely of bentonite (montmorillonite) in a fixed ratio of 33% clay and 66% brine containing variable components. Since the molecular processes described depend on both the type and amount of clay, we anticipate that natural mixtures on Mars or Ceres will differ, leading to variations in flow properties.

Phyllosilicates have been documented on both Mars and Ceres^72,73,74. Nontronite and saponite are the primary clay minerals observed on Mars^74,75,76, while serpentine and saponite are found on Ceres^72,77,78. Additionally, the brine on Ceres is likely alkaline⁷⁸, whereas the pH of mud volcanoes on Mars is less understood and may be acidic if volcanic gases are dissolved within. Pure serpentine, saponite, and bentonite have a point of zero charge (PZC) around pH 3–4^79,80. However, the addition of salts and different mineral combinations can influence macroscopic rheological properties through zeta potential and electro-viscous forces, as the PZC of multi-component mixtures varies. For instance, acidic brine compositions could limit interactions, resulting in different flow behaviors compared to our tested mixtures. Thus, strong compositional variations may restrict molecular interactions and documented non-linearities to specific conditions observed in our rheology measurements. However, the liquefying effect of multivalent anions and the opposite effect of protons in bentonites and kaolines with acidic and neutral salt additives is well-documented by ref. ²⁶.

Our supplementary pH measurements (Table S4, Supplementary texts) indicate that all mud-brine suspensions are nearly neutral, with slight alkalinity ranging from pH 7.04 to 7.94 (below Ceres brine values). For alkaline and neutral bentonite mixtures, the drop in mud bulk strength for decreased pH is documented²⁷. The contrasting pH levels relative to clay types are expected to affect the edge faces of clay sheets by adsorbing charges; however, the exact processes and their impacts under varying conditions on Mars and Ceres remain speculative. Therefore, a significant limitation of our results is the chemical composition restricting processes to bentonite or kaolin mud suspensions where liquefying effect is valid²⁶, and which may represent only sample behavioral regimes of real muds on Mars or Ceres.

The combination of multiple salts may also affect the bulk behavior of muds. To explore this effect, we conducted simple experiments with selected salt combinations (NaCl + MgSO₄, NaCl + Na₂SO₄, and MgSO₄ + Na₂SO₄). However, in tested ranges (5–10 wt.% mutual concentrations), these did not exhibit any mud behaviors beyond those reported in the main results (see Supplementary texts: Table S1 and S2 for additional data). Although we did not conduct an extensive investigation into this area, other salt-clay combinations warrant further exploration in future experiments, as they may be relevant to conditions on Mars or Ceres.

Our experiments were conducted with natural samples of bentonite that had non-zero humidity and a particle size ranging from approximately 50 microns to 2 mm (refer to Table S5 for more information). However, the exact particle properties, as well as their distribution in the blended mud bulk before and during the experiment, are unknown. Changes in the quantitative distribution of particles may have a significant effect on freezing due to the reorganization of bubble and crystal nucleation sites. Such effects, along with the varying capacity of different clays to lock water into their structure, may also lead to increased discrepancies between the freezing properties or between the rheological behavior of different muds. Therefore, we suggest an importance of subsequent experiments using industrially processed, dried clays and a systematic investigation of particle size differences.

Last but not least, a limitation is the scaling related to gravity and the dimensions of the process. Our experiments focus on the edges or foreheads of the flows, which restricts their dimensions to a small scale where our results are rather valid. However, as demonstrated by ref. ⁸¹, thicker flows with variable initial velocities (determined by slope) lead to different Reynolds numbers, affecting turbulent-laminar behavior of cryolavas and, consequently, internal crystallization dynamics that can alter spatial extension and surface solidification. Conversely, slow flows are influenced by heat transfer from the substrate, which can increase salt precipitation and further slow down the flow. The velocity effect is then evident in our rheometry results, where low shear rates, especially for low-solubility Na₂SO₄, result in increased bulk viscosities. The impact of decreased gravity on various aspects of discussed processes is far beyond the scope of our study and would require detailed three-dimensional thermal-mechanical-chemical numerical modeling.

Despite these limitations, our work reveals the complexity of brine mud behavior under low pressure and underscores the importance of both salt effects on rheology and phase transitions - brine/mud stabilization. Theoretical MVL structures on Mars or other bodies may therefore result from different regimes of mudflow that differ from those observed on Earth, and be characterized by unique small-scale formations such as long and narrow chains of lobes. In such enclosed structures, the increased longevity of liquid brine (or mud) components^4,5,66,71 can be further enhanced, potentially creating a suitable environment for the evolution of microbial life forms^7,8,9,10,11. Furthermore, the existence of MVL on a planetary surface provides evidence of subsurface liquid and may record information about the composition and possible habitability of such subsurface environments. Therefore, investigating MVL sites should be crucial for prioritizing future exploration missions^82,83,84.

Methods

Experimental setup and processing

Flow experiments

We performed 55 experiments in total (18 modifications, each was triplicated; one experiment failed, see Table S1 in Supplementary texts), during which a 500 ml portion of mud was released from the container situated above the 0.9 × 0.4 m slightly inclined (5°) aluminum tray (see Fig. 6a). The tray was filled with a ~2 cm deep layer of fine-grained (ca. 63–200 μm) silica sand and pre-cooled to average −25 °C. During the experiments, the average ambient pressure was maintained at the value of 5.9 ± 1 mbar (see Table 1 in the Supplementary texts). The tray, chamber and mud reservoir were equipped with 3 thermocouples and each run was documented with three cameras from different (pseudo-orthogonal) directions. Before initiation of the experiment, the pressure drop was maintained under the constant vent conditions (keeping a similar, exponential, drop inside the chamber). The temperature of mud samples in the container with similar initial temperatures (from ca 18.9 to 20.8 °C) thus followed a progressive thermodynamic p-T evolution, reflecting: (1) initial degassing (ca 150–23 mbar), (2) boiling (from ca 20–23 mbar, depending on salt type) and associated evaporative cooling (to ca 5.9 mbar and temperature ca −0.15 °C, average). When the desired pressure was reached, the manual release of the mud sample to a surface was initiated to induce a controlled directional flow. The experiments did not take into account the effect of the weaker gravity field on Mars where gravity acceleration is only ~38% of the surface gravity of Earth.

Fig. 6: Geometrical and technical setup of tray (flow) and tube (cooling) experiments.

Setup of the (a) flow experiments based on experimental setup utilized by refs. ^22,23 and (b) tube experiments with the schematized internal structure of binary solutions. Displayed are also positions of thermocouples, photogrammetry targets and individual cameras.

Experiments were performed by using the low-viscosity (and hence high water content) mud prepared from the clay mixture, which was already utilized by ref. ^14,15,16 The clay base was the bentonite obtained from the claystone named after the “Rokle” locality, mined by the Keramost company (Kadaň, Czech Republic). The bentonite is composed of ~76% montmorillonite, 23% illite, and 1% kaolinite and formed by alteration of pyroclastic rocks. As it is currently unknown, which types of clays might be associated with putative martian sediment mobilization, this type of clay might be at first approximation a suitable analog as pyroclastic deposits as well as water are present on Mars (for overview see Brož et al.⁴). Another component of the mud mixture was binary solution with various (mainly sulfate-based) salt contents. Selected salts represent a kind of compromise between materials that could be easily and safely handled in the laboratory and those expected to be possibly present on Mars or elsewhere in the Solar System. We used the MgSO₄.7H₂O - epsomite (Intra Laboratories, UK), NaCl (APC Pure, UK), Na₂SO₄ (VItal Minerals, UK) and CaSO₄ (Special Ingredients, UK) in concentrations ranging from 0.5 to 15 wt.% (note, that in the case of MgSO₄.7H₂O, the real concentration of MgSO₄ itself is about two times smaller than that of heptahydrate). The preparation of the final mixture had two steps. At first, we mixed deionized water and salt in ratio leading to the desired concentration. We kept the salt dissolved at an ambient temperature of 20 °C for at least 15 h to assure sufficient solution of all the salt grains (typically, the sodium sulfate has a problematic dissolution process). In the second step, we mixed 400 g of the binary solution with 200 g of bentonite using a blender for 5 min to ensure sufficient homogenization of the raw volume, which was then allowed to relax for 30 min under the same conditions (temperature and periodic stirring every 10 min). The exact volume of 500 ml was measured before each experiment, re-stirred for homogenization, and placed into the experimental container. This methodology should be suitable for maintaining an appropriately homogeneous mixture and minimizing any further effects of clay swelling within the mud bulk.

One of the experimental limitations arising from the mixture preparation is the uncertainty of particle properties, such as grain shape and size distribution, within the blended mud samples. We used natural bentonite with basic properties characterized in Table S5 of the Supplementary texts. However, despite knowing the approximate composition of the raw bentonite batch, the blending process altered these attributes, resulting in final mixtures with unknown particle properties (please refer to the note in Table S5 for more information). To control the process of salt-water crystal and bubble nucleation in any subsequent smaller-scale experiments, such data would be important. It should also be noted that using cryo-SEM methods for further investigations of particle properties may change bulk geometry and solid-liquid fraction or cause reorganization of clay particles and other undesired effects. The initial rheology before the pressure drop in the chamber of suspensions was measured at the Institute of Chemical Process Fundamentals of the Czech Academy of Sciences. Details about rheological measurements are provided in the following section. To perform the experiments with the same volume of mud mixture, we used 500 ml of the final mixtures. Each experimental run (see Table S1 in the Supplementary texts) was pre-prepared in a pressurized chamber to ambient laboratory pressure ~1020 mbar and in 18–20 °C air temperature. The first step was to insert the cold (~ −33 °C) copper plates serving as passive cooling above an inclined (5°) metal base plate. Then a 0.4 × 0.9 large aluminum tray infilled with ~3 cm thick frozen sand layer was placed on top of pre-cooled plates. This bed/substrate was prepared in minimum 8 h (usually day) before the experiment by making a layer of a dried natural quartz sand quarried in Southeast England (type Leighton Buzzard; with the average particle diameter ø ∼ 200 μm). The sand layer was then levered to make a flat surface with gently thicker ridges at the sides of the tray with the aim to prevent contact of the mud with the tray walls. After preparation, the tray with the sand was placed in a freezer and cooled down to average −28 °C. As another step we mounted and fixed the tilting container with the mud mixture to the upper wall of the tray and equipped it with a single thermocouple to gain data about cooling during pressure drop in the chamber (initial values before the pressure drop varied in a range of 18–21 °C). Another thermocouple was situated in the lower part of the tray, approximately 1.5 cm deep inside the sand to monitor the temperature of the sand layer, and one in the middle of the chamber in empty space beneath the whole apparatus to monitor the general temperature of the air inside the chamber (we refer to Table S1 for exact ranges of substrate temperatures at point of measurement). We used two web cameras to record the experiment inside the chamber, from top and in the direction of the expected mud stream. Additional movie camera was set outside the chamber with a frontal view through the glass window (see Fig. 6a for the overall experimental setup).

Once the experiment was set, the chamber was hermetized and the controlled pressure drop was started. The final pressure was maintained at a desired value of about 5.9 ± 1 (4.5 ± 1) mbar (see Table S1) with a period of reaching this value during approximately four minutes. Every time the mud self-cooled to the temperature in range from ~−2 °C to ~4 °C during the pressure drop, however, at the moment of release, the mud was still in the liquid state. When the appropriate pressure was reached, the container with the mud was manually tilted by external manipulation lever and the mud to the surface of the sand substrate was released. We tried to minimize the differences in release time for all experiments performed. Each experimental run was triplicated to confirm the reproductivity of observed features and to increase robustness of data for post-processing.

When the experiment was finished, we left the tray with resultant mud flow structure in the chamber for an additional 10–30 min, to assure the mud will be sufficiently frozen for later sectioning and further manipulation. After that, the chamber was pressurized back to ~1020 mbar, opened and the tray was placed out for post-processing, including sectioning. Mud flows of different compositions as well as salinity share some similarities, which allowed us to exclude some concentrations from our experimental set. For example, the 2.5% NaCl mud develops a similar type of flow as the 10% MgSO₄ or 5% Na₂SO₄ and 5% CaSO₄. Therefore, we excluded the 15 wt.% concentrations for all salts, except the MgSO₄, because increasing concentration (typically >10 wt%) leads to a convergence of flow abilities under the same experimental conditions (see also the rheological convergence at Fig. 4d or Fig. S3 in Supplementary texts). For a similar reason, we included 2.5 wt.% concentration for the NaCl, to find the threshold behavior at the lower concentration. Datasets, including the p-T-t data, movies and images, were stored in a high-capacity digital storage device and backuped at online repositories. Primary raw p-T-t data and associated movies are part of the Supporting materials, stored at: https://doi.org/10.5281/zenodo.12801318.

Tube experiments

These simple experiments were designed to document the evaporative cooling and delayed freezing of the brine solutions to approximate phase transitional times, temperatures, and pressures. The solutions were prepared by D.I. water and salt in desired concentration (0.5–10 wt.%). Here we ran 24 experiments (Table 2 and Fig. 6b in the Supplementary texts) to test: (1) simultaneous pressure drop and cooling of different salt-water solutions with the same salt concentration, (2) simultaneous pressure drop and cooling of solutions containing the same salt but of different concentrations and (3) the effect of the rate of depressurization on phase transitions (boiling, freezing). We employed slow (~1 h duration) versus fast (~4.5–10 min duration) pressure drops to finite values (typically 0.5–2 mbar). The pressure minimums are maintained at lower values than for the flow experiments, to achieve the cooling and freezing of high-salinity mixtures over an increased time. In all sets of the experiments we compared four solutions situated in transparent 0.5 m (height) × ~0.05 m (diameter) cylinders. The test cylinders were equipped with a single thermocouple each, positioned ~2 cm above the bottom. Experiments were documented by a single camera situated inside the chamber (Supplementary Movies 14–16).

We excluded the mud mixtures from these experiments to gain a “bridge” between behavior of the bentonite-water-salt systems and water-salt systems, which is implicitly less complex. These experiments serve also as a basis for potential further direct comparison between pressure-induced cooling of muds and binaries. Another reason is simpler investigation of bulk interior due to transparency of such solutions and possibility to better establish the times of boiling and freezing (correlated observations and thermal data) or formation of ice lids. Even if the tube experiments don’t directly simulate the mud mixtures, they do allow us to see relative differences between the different salt solutions.

The four cylinders we used were equipped with a single thermocouple each positioned 2 cm above the bottom. The chamber itself was, similarly as in case of tray experiments, equipped with one thermocouple to have a reference measurement of ambient temperature. Before the experiments were run, we prepared the binary mixtures, following the same procedure like in the case of mud mixtures (at least 12 h for sufficient dissolving and stabilization of bulk temperature). We used 150 ml of prepared mixture which was placed in each tube - following a scheme of various concentrations of a current salt or various salts of the same concentration in further experiments. If we observed a strong temperature variance, typically higher than 0.7 °C, we adjusted the temperature by warm/cold iron stick to the desired reference temperature of ~19 °C. After checking potential leaks and thermocouple proper functions, we hermetized the chamber and started depressurization. We followed two schemes for the pressure drop, the fast (approximately 4 min to reach the boiling point) and slow (approximately 20 min to reach the boiling point). To find the proper correlation between the pressure log, thermal data, and visual observations in the time frame of the experiment, we synchronized all the times and performed two single light impulses visible in the movie (start of the experiment and termination of the experiment). When the experiment was finished, we pressurized the chamber back to ambient laboratory pressure and carefully removed the cylinders for a quick investigation of the internal structure (ice-salt mushy zones, ice lids etc.). However, outside the chamber in the laboratory conditions we were unable to use the finite state of partially frozen solutions in the cylinders for further deep investigations since those ice structures quickly undergo the melting.

Primary raw p-T-t data and associated movies are part of the Supporting materials, stored at: https://doi.org/10.5281/zenodo.12801318.

P-T modeling

Comparing all the results of depressurization cooling, we observed that: the NaCl solution reached 1-1.8 mbar with finite bulk temperatures about ~ −7.1 °C (similar to slow pressure drop) and froze last. In contrast, the MgSO₄ solution froze first when at low concentrations (0.5 wt%) at pressure range ~3.7–5 mbar with bulk temperature ~2.7–3.2 °C. With increased salinity to 10 wt%, the CaSO₄ and Na₂SO₄ froze last (typically at ~ 3.2–4.5 mbar) with significantly higher finite bulk temperatures (3.1–5.7 °C). This effect was correlated with the freezing of the entire bulk instead of protective ice crust formation (see Supplementary Movies 8 and 9). A significant temperature step-increase associated with the Na₂SO₄ sample, caused due to the release of the latent heat during salt/ice crystallization can be seen in Fig. 4b. This effect was well-reproduced through most of the performed Na₂SO₄ experiments and together with the presence of mechanically non-inert crystals probably affects the convective streaming and overall cooling of the sample.

To overcome problems with an interpretation of the exact times of phase transitions tracked by thermocouples inside the tubes, we also employed thermodynamic modeling to find theoretical equilibrium phase fields for binary samples and compared the resultant curves with the experimental measurements (Fig. 4c). Calculations of thermodynamic equilibria were performed using the Pitzer ion-ion interaction approach implemented using the frezchem.dat database within the geochemical modeling program PHREEQC⁸⁵. For a full description of the Pitzer equations for calculating activity coefficients at high ionic strength, and their application within the FREZCHEM model, see Marion et al. ⁸⁶ FREZCHEM was ported to PHREEQC by Toner and Sletten⁸⁷. Due to a paucity of thermodynamic data within the relevant Pitzer databases on ice-water equilibria at low temperatures and sub-ambient pressures, calculations were performed at 1 bar. This will not affect predictions of saturation vapor pressure, which depend only on temperature, but calculated ice formation temperatures (liquidus temperatures) may deviate slightly from true liquidus temperatures. However, we expect this effect to be negligible. The freezing point of pure H₂O is approximately 0.01 °C warmer at 6 mbar than at 1 bar⁸⁸, thus we expect our 1 bar calculations to underestimate true liquidus temperature by a similar degree. Simulations subjected fluids to decreasing temperatures between 20 °C and −10 °C in 0.1 °C decrements, spanning the observed temperature ranges in chamber experiments. Fluid compositions matched those in chamber experiments. The saturation vapor pressures and liquidus temperatures of salt solutions were compared to those of pure H₂O. To ensure numerical consistency between calculations containing salts and pure H₂O, the vapor pressure of pure H₂O was approximated as a function of temperature in PHREEQC using an extremely dilute solution containing 10⁻⁷m Na⁺ and Cl⁻. The difference in saturation vapor pressure between this dilute solution and pure H₂O is 4.4 × 10⁻⁶mbar at 20 °C. It should be noted that the measured minimum temperatures of the solutions are several degrees above the predicted ice formation temperature, as is the case for all other experimental solutions beside 10 wt% NaCl. This discrepancy could in part result from lower predicted liquidus temperatures at 1 bar than the true liquidus temperatures at chamber pressures. It can be further explained by the limited number of temperature measurement devices in the cylinders, where the thermocouples are positioned deeper in the solution. Hence, the measured temperature is expected to be slightly higher than at the upper, freezing level.

Viscosity measurements

Viscosities of bentonite suspensions were measured using viscometers Brookfield LVDV-II+Pro Extra, RVDV2T Extra, and rheometer Brookfield HBDV-III Ultra which differ in maximum spring torques (0.0673, 0.7187, and 5.7496 mNm, respectively). Both apparent flow curves and real flow curves were measured. Apparent flow curves were measured via immersion viscometry using either cylindrical (LV-1, LV-2C, LV-3C, LV-4) or vane (V-72, V-73, V-74, V-75) spindles. Real flow curves were measured using a small sample adapter containing cylindrical chamber SC4-13R/RP and cylindrical spindles (SC4-18, SC4-31). Viscosity measurements were performed at 25 °C using a proper combination of viscometer and spindle. Speed ramps were measured and the time period for each spindle rotation was long enough to obtain steady state. It was inspected by repetitive measurements that sedimentation does not influence results markedly. Both apparent and real viscosity curves coincide very well, therefore results of all independent measurements for each suspension were merged into one viscosity curve. The obtained viscosity curves were fitted in the given range of shear rates via power law model:

$$\begin{array}{ccc}\sigma =K{\dot{\gamma }}^{n} & {{\rm{or}}} & \eta =\frac{\sigma }{\dot{\gamma }}=K{\dot{\gamma }}^{n-1},\end{array}$$

(S1)

where σ, $\dot{\gamma }$, η correspond to shear stress, shear rate and viscosity and K, n are parameters of the power law model, i.e., coefficient of consistency and flow index.

Static yield stress σ_γ of high consistent suspension was measured at 25 °C using rheometer Brookfield HBDV-III Ultra with spindle V-73 under rotation 0.05 rpm. The result is an average of 12 independent measurements. For further details, we refer to the Supplementary texts (Fig. S1–S3). The basic principles and standards for immersion viscometers, including device specifications, are described in the ISO standard (ISO 2555:2018). Similarly, rotational viscometry is defined by the standard ISO 3219-2:2021.

During the rheometry tests, no changes in volume and therefore no bulk composition variations were documented. To approximate the state of the sample before the experiments in the low-pressure chamber, we homogenized the pre-prepared mud samples by gentle stirring and allowing them to relax for approximately one minute before measuring. The cylindrical containers used had a diameter larger than 10 cm, which should be a sufficient approximation of half-infinite space for generally low-viscosity materials, such as our mud samples. Changes in the bulk volume fraction of solid and liquid, as well as any skeletal changes, were not performed due to limitations of the selected methodology and expected effects on the measured samples. The volume of the samples was measured before and after transport to the Czech Republic for rheometry without any documented change.

We would like to note that, in contrast to rheometry tests, during the actual experiments in the low-pressure chamber, the mud samples may lose a small portion of water due to evaporative effects and therefore slightly change in composition.