Cryogenic origin of fractionation between perchlorate and chloride under modern martian climate

The high perchlorate (ClO4−) to chloride (Cl−) ratios observed at the Phoenix landing site, northern polar region of Mars, have been puzzling since detection. However, a lack of understanding of perchlorate-chloride-water systems under cryogenic conditions makes it difficult to assess ClO4−/Cl− ratios during deliquescence-related processes. Here we quantitatively evaluate ClO4−/Cl− fractionation in deliquescence-induced brines of magnesium- and calcium-perchlorate-chloride salt mixtures under subzero conditions, by measuring solubility data and constructing temperature-dependent thermodynamic models. We find that under specific relative humidity (RH) and temperature (T) conditions, deliquescence of perchlorate-chloride mixtures may form brines with fractionated ClO4−/Cl− signatures. Appropriate RH-T, water-limited conditions, and aeolian processes are required to produce and preserve the elevated ClO4−/Cl− signatures in soils. Under the present climate, the north polar region can support ClO4−/Cl− fractionation and potentially enrich perchlorate for longer periods on global Mars. This highlights the uniqueness of Mars’ arctic environment and its implications for modern habitability. High perchlorate/chloride ratios at the Phoenix landing site in the Martian northern polar region could have formed by specific relative humidity and temperature conditions coupled with dust transport, according to experiments and thermodynamic modeling.

P erchlorate (ClO 4 − ) and chlorate (ClO 3 − ) have important implications for the habitability of Mars, such as aqueous activities [1][2][3] , redox capabilities [4][5][6] , and bioavailability 7 . Understanding the distribution and speciation of oxychlorine species on Mars is also critical for understanding the chlorine cycle on the Martian soil-atmosphere interface 8 , and critical for the detection of organic signals, and hazards and in situ resources for future human exploration 9 . To date, perchlorate or oxychlorine species (ClO x − ; equivalent to ClO 4 − + ClO 3 − ) have been detected on Mars in surface soils, sedimentary rocks, and Martian meteorites [10][11][12][13][14][15][16] . Natural perchlorate and chlorate are also observed in (hyper)arid environments on Earth (e.g., Atacama, Antarctic Dry Valleys) and in the surface samples of the moon and chondritic meteorites, indicating a ubiquitous presence of oxychlorine species across the inner solar system 17,18 .
High ClO 4 − /Cl − and ClO 4 − /Cl total ratios found in Phoenix soils might be formed primarily via direct production of ClO 4 − from Cl − (i.e., generation mechanisms) or some concentration pathways of already formed ClO 4 − (i.e., enrichment mechanisms). However, the generation mechanisms are unlikely because Martian and terrestrial natural samples rarely contain ClO 4 − / Cl total molar ratios higher than 0.2, and simulation experiments producing ClO x − species from chloride (via photochemistry, electrochemistry, and ozone oxidation) have not reported any ClO 4 − /Cl total molar ratios higher than 0.1 [21][22][23][24] . Enrichment mechanisms such as deliquescence of perchlorate salts (i.e., moisture-absorption and dissolution of the salts and eutectic melting of salt-ice mixtures) have been proposed to produce brines that may translocate and concentrate perchlorate 1,[25][26][27] . However, a lack of understanding of perchlorate-chloride-water ternary systems under cryogenic conditions makes it difficult to evaluate the ClO 4 − /Cl − signatures in such brines.
Here we quantitatively evaluate ClO 4 − /Cl − fractionation in deliquescence-induced brines of magnesium-and calciumperchlorate-chloride salt mixtures under subzero conditions, by measuring solubility data (233.15 K to 273.15 K) and constructed phase diagrams (180 K to 300 K) with the aid of temperature-dependent thermodynamic models. We find that under specific relative humidity (RH) and temperature (T) conditions, deliquescence of perchlorate-chloride mixtures may form brines with fractionated ClO 4 − /Cl − signatures. Appropriate RH-T, water-limited conditions, and aeolian processes are required to produce and preserve the elevated ClO 4 − /Cl − signatures in soils. Under the present climate, the north polar region can support ClO 4 − /Cl − fractionation and potentially enrich perchlorate for longer periods on global Mars. Our work highlights that the high ClO 4 − /Cl − ratios detected in the Phoenix surface soil are produced by local environmental conditions and surface processes and may not be generalized to global Mars. It also indicates the uniqueness of Mars' arctic environment and its implications for modern habitability. In addition, the high ClO 4 − /Cl − signatures might be further used as a chemical signature to infer deliquescence in a liquid water-limited environment for Mars.

Results and discussion
Thermodynamic constraints on deliquescence of perchloratechloride-water systems. We focus on the ternary systems Mg(ClO 4 ) 2 + MgCl 2 + H 2 O and Ca(ClO 4 ) 2 + CaCl 2 + H 2 O, which are the two most plausible systems suggested for soluble salts in Phoenix soils [28][29][30] Fig. 1 (Supplementary Data 1). For a specific salt mixture, the intersections of the deliquescence surface and the lower-and upper-T boundaries (i.e., dark cyan and dark red planes, respectively, in Fig. 1) constrain the RH-T conditions allowing for ClO 4 − /Cl − fractionation to occur. Deliquescence of perchlorate salt is preferred between the two boundaries while partial chloride salt remains solid, producing a brine with elevated ClO 4 − /Cl − ratios compared to its parent salt mixture. Above the upper T boundary, both perchlorate and chloride salts can fully deliquesce; under the lower T boundary, both perchlorate and chloride salts remain solids. Neither of the latter two cases would produce fractionated ClO 4 − /Cl − signatures.
Note that we did not include chlorate (ClO 3 − ) in our experiments and model, although it is an essential component of oxychlorine species that may be present at both Phoenix site 30 and Gale Crater 9,31 on Mars, with equal or higher abundances than perchlorates 24 . The rationales for not involving chlorate in this work are mainly because (i) general lack of water activity, thermal and cryogenic solubility data for binary chlorate systems; (ii) the absence of ternary solubility data on chlorate systems; and (iii) experimental complexities to include Cl − , ClO 4 − and ClO 3 − in one system. Consequently, the subsequent discussion and conceptual model in the following sections are limited to the ClO 4 − -Cl − -H 2 O system without considering the potential interference of chlorate. Future systematic experimental and modeling work on binary and ternary chlorate systems for cryogenic conditions is necessary to obtain fundamental insights.
Deliquescence and potential perchlorate/chloride fractionation at the Phoenix site versus Gale Crater. Combining the thermodynamic constraints and in situ RH-T measurements, we can further constrain the deliquescence behavior of Mg-and Ca-perchloratechloride salt mixtures at the Phoenix site and Gale Crater on Mars. At the Phoenix site, the surface RH-T conditions recorded by the Thermal and Electrical Conductivity Probe (TECP; recalibrated data) 32 can support the deliquescence of Mg-and Ca-perchloratechloride salt mixtures (  33 , we speculate that the RH-T conditions of Phoenix also likely support at least partial deliquescence and ClO 4 − /Cl − fractionation in the latter cases. At Gale Crater, the Rover Environmental Monitoring Station (REMS) onboard the Curiosity rover recorded RH-T data. We examined the in situ RH-T data of the first four Martian years ( Figure S17), which account for all the available data at the time of writing. Since the REMS data record the RH-T conditions at 1.6 m above the ground, surface brightness temperatures (T surf )   measured by the Ground Temperature Sensor (GTS) were also processed for comparison. We find that the REMS and GTS datasets show no difference in terms of supporting deliquescence at Gale Crater, but the GTS data contain higher uncertainties that would propagate to the derived RH data (Supplementary Figs. 18 and 19). Therefore, we use the REMS data to discuss the Gale Crater environments (Fig. 2b). The variation in RH-T conditions along the traverse of the Curiosity rover is primarily due to the change in ground properties (e.g., thermal inertia and porosity) from loosened regolith to solid bedrock, as the rover started to climb the Mount Sharp (sol 1800 onwards) 34 Fig. 17). The O 2 and HCl evolution patterns during pyrolysis of the aeolian samples "Rocknest" and "Gobabeb" suggest that the cations of the oxychlorine species are likely Na, K and Ca 9 . We find that the absence of ClO 4 − /Cl − fractionation in these samples during the first three Martian years might further rule out Ca-perchlorate. The lack of pyrolysis analyses for oxychlorine in soil/aeolian samples after "Gobabeb" (elevation For reference, we also analyzed relevant data from McMurdo Dry Valleys (MDV), Antarctica. The MDV is considered an excellent analog site for the northern plains of Mars because of the arid conditions, intense UV solar radiation, diurnal freeze/thaw cycles, sparse snowfall, cryogenic processes, low mean temperatures, and Mars-like surface topography 35    In addition to the north polar region, the Acidalia-Chryse and Utopia regions also appear to be favorable, to a lesser extent, to ClO 4 − /Cl − fractionation, as well as a few large basins (e.g., Hellas Planitia and Argyre Planitia), where Ca-salt mixtures could fractionate. The positive topography (e.g., highlands, Olympus Mons), in turn, would not support ClO 4 − /Cl − fractionation even at locations within the mid-latitude regions. Figure 4 shows that the Ca-salt mixture can result in deliquescence and the consequent ClO 4 − /Cl − fractionation for a longer time (maximum~8% of a Mars year or~50 sols per Mars year) and in a more extensive region including equatorial latitudes compared to the Mg-salt mixtures. However, we speculate that Ca-perchlorates-chlorides are unlikely to be the predominant salt phases on the Martian surface. Ca 2+ release by the weathering of Martian basalts is lower than that of Mg 2+ 37 and Ca 2+ tends to precipitate as sulfates rather than chlorides [38][39][40] . The subsequent oxidation of chlorides (e.g., with cations Mg, Na, K) via photochemistry, electrochemistry, or ozone [23][24][25]41 would likely produce perchlorates sharing the same cations as their parent chlorides.
A conceptual model for the formation and preservation of high perchlorate/chloride ratios on the Martian surface. Three essential factors are crucial for the formation and preservation of fractionated ClO 4 − /Cl − signatures. First, the RH-T conditions should be suitable for deliquescence to occur. Second, the water source should be limited in quantity since any substantial liquidwater interactions would dissolve both ClO 4 − and Cl − and obliterate the fractionation signatures. Third, subsequent separation of the perchlorate-enriched brines and chlorideenriched salts is necessary for the ClO 4 − /Cl − signatures to be detected. In this scenario, the transient brines formed under water-limited conditions would be primarily confined onto the particle surface or intergranular pores, and thus, their spatial translocation would be substantially constrained. We suggest that translocation of the salts with soil/dust particles via aeolian processes may be a more practical way to form a locally homogeneous ClO 4 − /Cl − signature, as detected in Phoenix soils.
We propose a conceptual model for forming and preserving high ClO 4 − /Cl − ratios on the present Martian surface (Fig. 5). (1) Aeolian processes (e.g., dust storms/dust devils) continuously transfer dust/soil particles coated with perchlorate-chloride salt mixtures (ClO 4 − /Cl total ratios presumably at the global-average level) to some region (e.g., the Mars arctic region). Dust is likely to be the primary carrier of perchlorates and chlorides 42 , given the high efficiency in converting chlorides to perchlorates on the dust particles 23,24 . (2) Under appropriate RH-T conditions (e.g., the Phoenix landing site), deliquescence of the perchloratechloride salt mixture by interaction with water-vapor or water-ice produces transient brines containing fractionated ClO 4 − /Cl − signatures. The specific ratios depend on the parental salts and the RH-T conditions. (3) Dust/soil particles cemented with brines would be more resistant to aeolian translocation (e.g., during the impact of saltating particles) than dry ClO 4 − -depleted dust/soil particles. In this way, perchlorates preferentially accumulate while chlorides are more likely to be transported locally or regionally.
The texture of the Phoenix soils supports this conceptual model well-a geologically young soil (for the current RH-T conditions to remain applicable), with active reworking (by dust deposition/ lifting) and without interaction with extensive liquid water. The cohesive and cloddy nature of the soils indicates an interaction between ice/absorbed water and salty soil in place 43 . The soil profile demonstrates homogeneity in terms of soil microstructure, and no substantial difference in soil particles is observed above and below the ice table 44 . Microscopic analysis of the Phoenix soil particles suggests that the soil is aeolian in origin 45 , and the particles may have undergone long-distance translocation (>500 km) via air-fall dust processes or saltation 46 . The three soil samples "Rosy Red", "Sorceress 1", and "Sorceress 2", composing the topmost~5 cm of the soil profile, each contain evenly distributed ClO 4 − concentrations and high ClO 4 − /Cl − ratios.
As a result of strong winds induced by baroclinic transient waves occurring at high northern latitudes in the autumn and winter seasons, most regional-scale dust storms originate from the northern plains [47][48][49][50][51] . Frequent local and regional-scale dust storms take place between latitudes~30-80°N 49 . These dust storms enable regional or global circulation of dust across latitudes and facilitate material exchange in and out of the north polar region. Dust storm simulations (e.g., for MY25 and MY34) indicate the general north polar region and the Phoenix site to be an active region in terms of dust activity 50 , which may facilitate dust/soil translocation in that region. Interestingly, the Acidalia-Chryse and Utopia Planitiae regions, which also stand out in Fig. 4, are known to be "storm tracks" where regional-scale dust storms develop 50 , which acts in favor of the proposed scenario in these regions.
The essential factors required for the formation and preservation of fractionated ClO 4 − /Cl − signatures in turn impose environmental constraints on the site where high ClO 4 − /Cl − signatures are detected. At the Phoenix site, for instance, we estimated that the deposition of an~5-cm thick dust layer requires~1.5 kyr to 13.7 kyr (Earth years) by using the rate of dust settling measured during the Phoenix mission lifetime ("fiducials" data of 0.05 ± 0.04 μm per sol, representing any random patch on the lander deck 52 ). Additionally, by using the north polar dust sedimentation rates (0.6~1 mg cm −2 yr −1 ) 49 and assuming a bulk density of 1.2 g cm −3 of the dust, we obtained a consistent estimated 6-10 kyr (Earth years) for a 5-cm dust accumulation. This estimated time-scale is consistent with the time proposed for ground ice formation by diffusive equilibrium (hundreds to thousands of years) 53 and the formation of a shallow ice table within~100 kyr under the current Martian climate regime 54 . Therefore, the active dust processes and the specific RH-T conditions may have lasted for approximately a few to tens of kyr at the Phoenix site.
Our study indicates that environmental conditions and surface processes operate cooperatively to produce and preserve fractio- formation of a lubricating brine in the basal ice layer and facilitate the flow of the northern ice cap 36 . The enrichment of perchlorates in the north polar region also offers implications for modern habitability. On the one hand, the (meta)stable brines that were formed are likely high in salinity and low in water activity, which are not habitable to any known terrestrial life 55,56 . On the other hand, concentrated perchlorates can facilitate water adsorption due to their hygroscopic characteristics and result in the regolith holding a substantial fraction of the water inventory 57 . Such water-rich regolith may shelter against radiation doses and provide safe habitats underground for potential extant life 58 . A thermostatic water bath with a refrigerating machine (Lauda Command and DLK45, Germany) was used to perform the cryogenic phase equilibria experiments ( Supplementary Fig. 1). The temperature uncertainty of the whole thermostat system was controlled within 0.05 K. A 250 mL flat-bottom glass tube with a PTFE stopper and overhead mechanical stirrer was used as the container (i.e., solubility cells in Supplementary Fig. 1). The reaction mixtures were prepared by weighing ultrapure water and the appropriate amount of perchlorate and chloride reagents to fill 1/4 of the reaction container. The container with the reaction mixture was lowered into a water bath until the water covered most of the container. The experiments were conducted at 273.15 K, 263.15 K, 248.15 K, and 233.15 K, respectively. The mixtures were equilibrated for five days at 273.15 K and for at least ten days at 263.15 K, 248.15 K, and 233.15 K, due to the slow rates in reaching equilibrium at subzero temperatures. Throughout the experiments, the reaction mixtures were stirred thoroughly.

Methods
As a conventional analysis method used by physical chemists in thermodynamic experiments, the chemical compositions of the supernatants filtered using 0.45 μm syringe filters were determined by mass titration of Mg 2+ , Ca 2+ , and Cl − , and subsequent subtraction of ClO 4 − according to charge balance. Cations Mg 2+ and Ca 2+ were titrated by EDTA solution calibrated using ZnO and CaCO 3 primary standards, respectively. Anion Cl − was titrated by AgNO 3 solution calibrated using a NaCl primary standard. The relative uncertainties of the chemical analysis were within 0.3% for Mg 2+ , Ca 2+ and Cl − , and 0.6% for ClO 4 − .
The chemical compositions of wet solids (i.e., directly scooped out of the reaction container without further processing) were also determined using the same analytical methods as for the supernatants. Solid phases in equilibrium with the solution at 273.15 K were identified using an X-ray diffractometer (XRD; X'Pert PRO, 2006 PANalytical with Cu Kα radiation and λ = 0.15406 nm) at room temperature ( Supplementary Fig. 2). For solid phases in equilibrium with the solution at temperatures lower than 273.15 K, Schreinemakers' wet residues method 59 was used instead of XRD analysis to avoid possible melting or decomposition of hydrated salts at room temperature. The Schreinemakers' wet residues method was a classical method in the phase diagram experimental determination for the salt-water system and its principle is illustrated in Supplementary Fig. 3. The overall relative uncertainties of the liquid and wet residual compositions in the experiments were within 2%. The details of the experimental methods and results are available in Supplementary Notes 1 and 2, respectively.
Thermodynamic modeling methods and results validation. The ISLEC software, i.e., Institute of Salt Lakes Equilibrium Calculator, was used in this study to model the solubility isotherms and generate the phase diagrams for the ternary Mg(ClO 4 ) 2 + MgCl 2 + H 2 O and Ca(ClO 4 ) 2 + CaCl 2 + H 2 O systems. The theoretical background, basic equations and implementation of the software can be found in a recent work 60 . The software adopted a CALPHAD type (i.e., Computer Coupling of Phase Diagrams and Thermochemistry) thermodynamic framework and a Pitzer-Simonson-Clegg (PSC) excess Gibbs energy model to describe the non-ideality of the aqueous phase. Unlike the conventional solubility modeling method, the CALPHAD framework allowed the thermodynamic model for the aqueous system to be parameterized using various types of thermodynamic data, particularly the thermal data (e.g., enthalpy of dilution, enthalpy of solution, and heat capacity of aqueous solution and solids), to enhance the reliability of the temperature-dependent models. The PSC model was an advanced version of the classical Pitzer model updated by the Pitzer's group 61 . The ISLEC software used the Gibbs energy minimization (GEM) approach for solving the phase equilibria. The ISLEC software has successfully modeled extensive binary systems, ternary systems, and multicomponent systems down to~200 K (see Code Statement for accessible link). A detailed description, parameters, and equations implemented by the ISLEC software for the studied ternary systems Mg(ClO 4 ) 2 + MgCl 2 + H 2 O and Ca(ClO 4 ) 2 + CaCl 2 + H 2 O are summarized in Supplementary Note 3. The experimental data collected from the literature for regressing the temperaturedependent binary perchlorate parameters are shown in Supplementary Table 3 Phase diagram calculation. Deliquescence occurs when the relative humidity (RH) in the gas-phase environment is at or above the mutual deliquescence relative humidity (MDRH) of a salt mixture. The MDRH data points therefore belong to the deliquescence surfaces of a salt mixture 62 . In a salt-solution model, RH can be derived using the water activity (a w ). At a given temperature, RH = 100 • a w = 100 • p w /p w 0 , where p w and p w 0 are the vapor pressure of the saturation solution and pure water, respectively. When a w is of a solution in equilibrium with the salt, the calculated RH is MDRH. The a w can be calculated using equation S16 in the Supplementary Information. The deliquescence surface data in an x-T-RH coordinate system were generated using the C codes (i.e., islec_mars_mdrh_mg.c and islec_mars_mdrh_ca.c; see Code Statement for accessible link). The x was calculated from the equilibrium molality of perchlorate and chloride in aqueous solution. These codes can be compiled using the GCC compiler on the Linux platform or the Dev-Cpp compiler on the Windows platform. The awk script can be used to format the output data produced by islec_mars_mdrh_ca for further visualization. These output data were then visualized with Gnuplot software for 2D and 3D plots. Detailed instructions can be found in the README file included in the source code package (link provided in Code Statement).
Mapping global Mars for the regional environments that may support ClO 4 − / Cl − fractionation. The regions that meet the RH-T surf requirements for ClO 4 − /Cl − fractionation, i.e., where environmental conditions overlap with the phase diagram of a specific salt mixture and can produce brine with ClO 4 − /Cl total > 0.2 (Supplementary Fig. 17  LMD General Circulation Model 63 . It should be noted that the data used for plotting the map are numerical model outputs, not measured climate data of Mars. The LMD GCM simulates water and dust transport, including the microphysics and radiative effects of water ice clouds, dust/ice sedimentation and condensation/ sublimation of water/ice at the surface [63][64][65][66][67] . The model has also been used to predict and analyze the night-time near-surface relative humidity on Mars 68 . The model is continuously confronted with available observations to validate its results. Specifically, the LMD GCM model resolves the equations of the atmospheric circulation on a 3D global grid covering the Martian atmosphere and includes the main physical parametrizations responsible for the climatological forcing, such as radiative transfer, small-scale turbulence 63 or daytime thermal convection 69 . Martian topography is based on MOLA (Mars Orbiter Laser Altimeter) data 70 , whereas surface properties of the Martian soil (thermal inertia and albedo) were taken from TES observations 71 , except for the permanent ice of the North Pole, the main source of water released in the atmosphere, which is prescribed by the model 65 . Radiative transfer takes into account the emission and absorption of CO 2 gas in infrared as well as the emission, absorption, and scattering of airborne dust 64 and water ice 66 . The model simulates water and dust transport, including dust and ice sedimentation and condensation/sublimation of water ice at the surface 63,67 . Water ice cloud microphysics have been introduced in the model to account for the interaction between dust and water condensing in the atmosphere 65 , including the processes of nucleation, ice growth and scavenging, and thereby allowing the representation of supersaturation. Although the dust spatial distribution is explicitly predicted by the model, the dust column opacity is rescaled to match the observed opacities compiled as dust scenarios for each Martian year 72 using a semiinteractive scheme developed 73 . The simulation used for the mapping has been produced with the 'climatology' dust scenario, which is derived from the available observations of dust from multiple years without global planet-encircling dust storms and is therefore representative for a typical Martian year.
The grid resolution was set to a standard value of 5.625°× 3.75°longitudelatitude (64 × 48), with 32 vertical layers (up to an altitude~120 km), a dynamical step of 1.5 min and a physical time step of 15 min. The water mass mixing ratio (mmr) obtained by the model at~2.2 m was converted to H 2 O partial pressure (p H2O ) according to: p H2O = 44/18 × mmr × p surf , and then converted to relative humidity RH at the surface: RH surf = p H2O /p sat (T surf ), where p sat (T surf ) is the saturation vapor pressure of water ice calculated at the surface temperature. The total yearly duration that each grid cell spends in the appropriate RH-T surf space was computed using 48 model outputs per day (30 min apart), 669 days/year, and by summing up all the time periods during which both the RH and T surf variables remained within the limits of the space domain given by the equations in Supplementary Table 7. The precise timings of domain crossing were calculated by interpolating the RH and T surf variables between each output. The time fraction of the year obtained at each grid point was mapped with a Robinson projection, using 15 contour levels (with IDL 8.5), on top of MOLA topographic contours.

Code availability
The ternary aqueous chemistry codes (C programs for deliquescence surface calculations; awk script for data formatting; gnuplot script for graph plotting) used in this study are available at Science Data Bank, https://doi.org/10.11922/sciencedb.00911. The softwares used in this study are accessible at following links provided. The awk: www.gnu.org/ software/gawk; the GCC compiler: gcc.gnu.org; the Dev-Cpp compiler: sourceforge.net/ projects/orwelldevcpp. The Gnuplot: www.gnuplot.info. The ISLEC software and models can be accessed from a PHP based Web interface with a single point calculation mode (www.islec.net/islec-web/islec-web-mg-cl-clo4; www.islec.net/islec-web-ca-cl-clo4). The LMD General Circulation Model is available on request from the LMD team.