We studied the low-temperature properties of sodium and magnesium perchlorate solutions as potential liquid brines at the Phoenix landing site. We determined their theoretical eutectic values to be 236 ± 1 K for 52 wt% sodium perchlorate and 206 ± 1 K for 44.0 wt% magnesium perchlorate. Evaporation rates of solutions at various concentrations were measured under martian conditions, and range from 0.07 to 0.49 mm h−1 for NaClO4 and from 0.06 to 0.29 mm h−1 for Mg(ClO4)2. The extrapolation to Phoenix landing site conditions using our theoretical treatment shows that perchlorates are liquid during the summer for at least part of the day, and exhibit very low evaporation rates. Moreover, magnesium perchlorate eutectic solutions are thermodynamically stable over vapour and ice during a few hours a day. We conclude that liquid brines may be present and even stable for short periods of time at the Phoenix landing site.
 NASA Phoenix landed on the surface of Mars on May 25, 2008, in the Vastitas Borealis plains at 68.22°N 234.25°E, a region dubbed “Green Valley”, and performed an extensive set of physical and chemical analyses aimed at characterizing the history of water and potential habitability of the Martian subsurface. While the direct observation of ice was an important discovery, a big surprise was the detection of the perchlorate ion ClO4− by the MECA (Microscopy, Electrochemistry, and Conductivity Analyzer) instrument [Hecht et al., 2009; Kounaves et al., 2009].
 Perchlorate is a strongly oxidized ion (Cl in the +7 oxidation state) and is extremely rare on the surface of the Earth: its natural occurrences are limited to hyper arid environments such as the Atacama Desert in Chile where perchlorate is present as a few percent associated with nitrate deposits [Ericksen, 1981]. Although nitrates have not been detected, perchlorates are part of a saline alkaline paragenesis dominated by chlorine [Kounaves et al., 2009], and may constitute a significant component in other regions where chlorides have been identified [Osterloo et al., 2008]. In this case, the determination of perchlorate thermodynamic and kinetic properties would help elucidate the reactivity of the soil at the Phoenix landing site, and also the potential presence of liquid water [Renno et al., 2009].
 Therefore, we present a study of the stability of perchlorate solutions on Mars. We calculated the thermodynamic stability diagrams of two Martian perchlorates relevant to Mars [Marion et al., 2009]: sodium (NaClO4) and magnesium (Mg(ClO4)2), in order to determine the conditions suitable for the formation of liquid solutions. We investigated the evaporation rates of these solutions as a function of temperature and concentration, to evaluate their residence times on the surface.
 Solutions were made using D.I. water and sodium perchlorate (NaClO4.H2O, EMD product #SX0694-1) at concentrations 20, 30, 40 and 55 wt% (±0.5%), and magnesium perchlorate (Mg(ClO4)2, Alfa Aesar product #11636) at concentrations 20, 30, 40 and 49 wt% (±0.5%). Evaporation rates of the perchlorate solutions were measured in our 0.6 m3 Mars simulation chamber [Chevrier and Altheide, 2008; Sears and Chittenden, 2005; Sears and Moore, 2005] under 7 ± 0.01 mbar of pure CO2 at temperatures ranging from 255 K to 275 K (Table S1 of the auxiliary material). The atmosphere was first evacuated from the chamber by lowering the pressure under 0.06 mbar. The chamber was then filled with pure CO2 to atmospheric pressure and cooled below 273 K. Once filled with CO2, the chamber was opened and the sample placed on a loading analytical balance. The platform supporting the balance set-up was then lowered into the chamber, the lid sealed, and the pressure pumped down to 7 mbar. Once the pressure reached 7 mbar, we waited 20 min to accommodate the instruments to the new conditions. Then we started recording mass, temperature (chamber and sample), pressure, and relative humidity (Table S1). The pressure was maintained at 7.00 ± 0.01 mbar during the whole experiment and the atmosphere was continuously exchanged to maintain the relative humidity (RH) below 1%.
3. Stability Diagrams and Eutectic Determination
 Vapour pressure of liquid solutions are controlled by temperature and composition. Therefore, we calculated the stability diagram of Mg and Na-perchlorates as a function of concentration and temperature. Such diagrams are also useful to determine the hydration state of the salt depending on temperature. Perchlorate solubility data are scarce and most thermodynamic parameters (Gibbs free energy and enthalpy) are unknown. Therefore, we used theoretical calculations and adjusted the thermodynamic parameters to obtain the best fit of existing data. We determined the eutectic temperature and concentration for each salt, by calculating the theoretical intercept between the liquid-ice liquidus (concentration below the eutectic) and the liquidus of the salt hydrate (concentration above the eutectic). The activities of ions and water were calculated using the Pitzer ion interaction model [Pitzer, 1991].
 We determined the evaporation rates of perchlorate solutions exposed to 7.00 ± 0.01 mbar of CO2 at temperatures between 255 K and 275 K (Figure 2). Evaporation rates were calculated by dividing the mass loss data by the surface area of the sample and its density (Table S1). Evaporation rates of NaClO4 solutions range from 0.07 to 0.49 mm h−1, and from 0.06 to 0.29 mm h−1 for Mg(ClO4)2 brines. The error on the evaporation rate results from the error on the regression line (<5%), the incertitude of the balance (0.01 g), and the error on the measurement of the density. The 1σ error is generally below 10% and we use this value as a standard error. Our data show that the evaporation rate is dependent on the temperature of the liquid. For example, the 40 wt% NaClO4 solution drops from 0.37 mm h−1 at 267 K to 0.11 mm h−1 at 256.5 K, i.e. a factor 3 over a range of 10 K (Figure 2a). Similarly the 49 wt% Mg(ClO4)2 shows a factor 3 decrease over 12 K. This is in agreement with our previous studies [Chevrier and Altheide, 2008; Sears and Chittenden, 2005], and is mostly related to the temperature dependency of the saturation vapour pressure of water, which controls the density gradient between the surface and the atmosphere [Ingersoll, 1970].
 As we previously observed for other salts [Chevrier and Altheide, 2008], the evaporation rate also depends on the salt concentration: at 266 K, the 20 wt% NaClO4 solution evaporates at 0.49 mm h−1 and at 0.14 mm h−1 for the 55 wt% sample. Magnesium perchlorate solutions show similar behavior: at 264 K, the 20 wt% Mg(ClO4)2 solution evaporates at 0.29 mm h−1, compared to 0.07 mm h−1 for 49 wt% (Figure 2b).
 At martian surface conditions, the mono- and dihydrate sodium perchlorates and the magnesium perchlorate hexahydrate are the most relevant phases. Magnesium perchlorate is highly hygroscopic, but this is valid only for the trihydrate form [Smith et al., 1924; Willard and Smith, 1922]. Its temperature of formation (∼420 to 434 K) is too high for the surface of Mars, even in the equatorial regions, and it is probable that the hexahydrate is the most stable form. Other hydration states, 4 and 2, may be possible in low humidity conditions [Krivtsov et al., 1989]. Large variations of humidity could induce variations in hydration state of magnesium perchlorate, in a similar way as for magnesium sulfate [Vaniman et al., 2004]. However, further work on the effect of humidity would be necessary to determine the stability on these hydrates.
 To determine the formation conditions of liquids at the Phoenix landing site, we compared the eutectic temperatures of each perchlorate to the temperatures during the coldest and warmest days [Hudson et al., 2009]. Results show that sodium perchlorate is liquid for about 8 hours only during the warmest day, whereas magnesium perchlorate remains liquid for nearly 18 hours during the warmest day and 12 hours during the coldest (Figure 3a). Thus, the presence of metastable liquids at the Phoenix landing site is possible.
 Using our model [Chevrier and Altheide, 2008], we calculated the evaporation rates of saturated perchlorate solutions at the Phoenix landing site during the coldest and warmest summer days [Hudson et al., 2009], using smoothed temperature data (Figure 3a). The evaporation rates are also dependent on the total pressure and on the humidity in the atmosphere. We used the water vapour measured by the Thermal and Electrical Conductivity Probe TECP onboard Phoenix [Hudson et al., 2009] and averaged the data using a Gaussian fit (Figure 3b), allowing us to calculate the average humidity at any time of the day. Using the LMD Mars Global Circulation Model GCM, we determined the total pressure [Forget et al., 1999] to be ∼8 mbar at LS 90–120, almost identical to the values measured by Phoenix [Taylor et al., 2009]. Contrary to the total pressure, the water partial pressure determined from the LMD-GCM is very different from the measured value. The GCM values appear nearly constant at ∼0.3 Pa, while the measured values show variations of two orders of magnitude (Figure 3b). Since the GCM does not include regolith-atmosphere coupling, this suggests that the humidity is controlled by the regolith.
 Integration of the evaporation rates over a full day shows that ∼80 μm of Mg(ClO4)2 or NaClO4 would evaporate during the warmest day (Figure 3c). The difference in the water activity, 0.69 for NaClO4 and 0.54 for Mg(ClO4)2, results in a slightly higher saturation pressure on NaClO4 brine than on a brine of Mg(ClO4)2 (Figure 3b). Sodium perchlorate is frozen for a large part of the day, while magnesium perchlorate remains liquid for much longer, and thus could evaporate for longer durations (Figure 3c). On the coldest day, NaClO4 is completely frozen (Figure 3a), while Mg(ClO4)2 remains frozen during almost half the day. The cumulative evaporation rate of Mg(ClO4)2 is only 4 μm (Figure 3c), due to the strong dependency of the evaporation rate on temperature.
 The effect of humidity on the evaporation rate is almost negligible; evaporation of the solutions in a completely dry atmosphere results in a difference of only ∼1 μm per day. Most of the evaporation occurs during the warmer periods, when the water pressure on the surface is several orders of magnitude higher than the atmospheric water vapour. Since the evaporation rate is a linear function of the humidity, but an exponential function of the temperature (through the saturation pressure), the humidity does not show large enough variations (Figure 3b) to affect the evaporation rate.
 The more significant effect of humidity occurs during the evening or the early morning, when the equilibrium water vapour on the surface of the liquid drops under the atmospheric humidity (Figure 3b). For Mg(ClO4)2, the liquid temperature is above the freezing temperature (206 K) but under the temperature at which the equilibrium water vapour is above the atmospheric vapour (∼215 K, Figure 3c), allowing the liquid to be thermodynamically stable between 6 and 7 PM and between 8 and 10 AM. Interestingly, these periods correspond exactly to the observations by Phoenix of large variations of dielectric permittivity [Zent et al., 2009]. Moreover, the liquid is paradoxically stable during the coldest day, since high temperatures during the warmest day result in continual evaporation (Figure 3c). For sodium perchlorate, the frozen period is much longer than the liquid stability period (Figure 3c), and thus liquid is never stable with respect to ice or vapour, but only frozen or evaporating.
 The presence of liquids in the regolith, even in low concentrations could have had significant effects on the stickiness of the soil, depending on the day (warm/cold) and the hour. Salts stable at very low-temperatures often show high viscosities at temperatures close to their eutectics [Chevrier et al., 2009], as verified for strontium perchlorate [Pestova et al., 2005]. Yet, application to magnesium perchlorate remains to be verified by experimental measurements.
 We have studied the stability of liquid solutions of sodium and magnesium perchlorates on the surface of Mars, using a combination of experimental measurements and thermodynamic and kinetic calculations. Magnesium and sodium perchlorate liquids are metastable during the day, due to low eutectic temperatures, although magnesium perchlorate is more stable due to its much lower eutectic temperature, 206 K versus 236 K for sodium perchlorate. Despite the very low evaporation rates of liquids, which possibly allow them to last several days, their metastability at longer timescales requires a replenishing process such as seasonal high humidity or deliquescence during periods of high obliquity. Indeed, during limited periods of time in the evening and early morning, eutectic solutions of magnesium perchlorate are thermodynamically stable and could actually form and reabsorb the water lost during the evaporation periods, explaining the humidity swings. In contrast, sodium perchlorate eutectic solutions are never stable. Therefore, the liquid indirectly observed by Phoenix, if present may be a eutectic solution of magnesium perchlorate.
 We thank Derek Sears, Giles Marion and Michael Hecht for the discussion on the stability and nature of perchlorates at the Phoenix landing site. We also thank Chris McKay and an anonymous reviewer for helping improve our manuscript.