Magnesium sulfate probably plays a dominant role in the water cycle of Mars away from the polar ice caps through hydration and dehydration reactions. This prominence is due to its abundance, its occurrence in numerous hydration states, and its ability to hydrate and dehydrate rapidly. New experimental studies on the metastable reaction between hexahydrite (MgSO4·6H2O) and starkeyite (MgSO4·4H2O) as a function of temperature and relative humidity, supplemented by recent investigations of the stable reaction between epsomite (MgSO4·7H2O) and hexahydrite and by phase equilibrium calculations, suggest that the most important magnesium sulfate phases involved in the Martian water cycle are MgSO4·11H2O, epsomite, starkeyite, and possibly kieserite (MgSO4·H2O). Hexahydrite is not predicted to be stable on the surface of Mars. During diurnal variations in temperature and relative humidity, 1 kg of MgSO4 can release or remove from the atmosphere 1.5 kg of H2O by cycling between kieserite and MgSO4·11H2O. Despite subequal abundances of calcium sulfate, calcium sulfates are not likely to be important in the water cycle of the planet because of sluggish rates of hydration and dehydration and a more limited range of H2O concentrations per kilogram of CaSO4 (0.00 to 0.26 kg kg−1). Modern or recent erosion on Mars attributed to liquid water may be due to the dehydration of MgSO4·11H2O because of the inferred abundance and likelihood of occurrence of this phase and its limited stability relative to known variations in temperature and relative humidity.
 Investigations of planetary geochemical cycles are hampered by a limited understanding of the actual phases (solid, liquid, or gas) involved in these cycles. The water cycle on Mars is no different. For Martian studies, insights are largely restricted to remotely sensed spectroscopic data from satellites, spectroscopic data from landers or rovers, and meteorites thought to have originated from the planet. Few of these studies yield definitive results as to the mineralogy, let alone the mineral chemistry of the Martian surface. Instead, much of the identity of minerals present near the surface of Mars is inferred in the context of known phase relations on the basis of bulk geochemical data from the Martian regolith and documented variations in temperature and relative humidity, which range in the summer from approximately 175 K and near 100% relative humidity (RH) during the night to approximately 255 K and 5% RH during the day [Savijärvi, 1995]. An additional challenge to this indirect approach is in the difficulty of extrapolating kinetic data on pathway-dependent phase transformations obtained in the laboratory to the lower-temperature conditions found on the planet. In contrast, the extrapolation of stable or metastable equilibria to lower temperatures lacks these challenges because of its independence from pathway constraints.
 Key issues related to the water cycle include the distribution of water and the processes that control its distribution. Feldman et al. [2004a] used neutron spectroscopy to map the distribution of water-equivalent hydrogen on the surface of Mars and found a range from at least 2% to over 9% with an average of 4.8% on a mass basis. They identified three distinct reservoirs: polar ice caps, atmospheric moisture, and solids other than ice in the equatorial regions where ice would be expected to be unstable because of sublimation at surface conditions. Candidate hosts for the equatorial hydrogen include buried ice, water molecules adsorbed on soil particles, and a variety of minerals.
 Numerous researchers for more than 3 decades have highlighted and refined the understanding of the critical role played by magnesium sulfates and, to a lesser extent, calcium sulfates, clays, and zeolites in the Martian water cycle [Clark et al., 1976, 1982; Clark, 1993; Bish et al., 2003; Gendrin et al., 2005; Peterson and Wang, 2006; Vaniman and Chipera, 2006a, 2006b; Wang et al., 2006]. Particular attention has been given to the role of magnesium-sulfate salts in the distribution and cycling of water because of their abundance in Martian soils and rock outcrops [Rieder et al., 2004; Christensen et al., 2004], their occurrence in multiple hydration states ranging from 1 to 11 mol of H2O per mole of MgSO4, their ability to act as both a source and a sink for water during diurnal cycles through their participation in a variety of stable and metastable dehydration and hydration reactions, and the rapid rate of these reactions, at least under ambient conditions at the surface of the Earth [Chou and Seal, 2003; Feldman et al., 2004b; Chipera et al., 2005]. Magnesium-sulfate salts are interpreted to have formed through the combined processes of weathering and evaporation [Clark et al., 1976; King et al., 2004]. Feldman et al. [2004b] attempted to model the variations in the concentrations of water-equivalent hydrogen in terms of hydration and dehydration reactions among magnesium-sulfate minerals with incomplete success. They found that magnesium-sulfate hydration states ranging from kieserite (MgSO4·H2O) to epsomite (MgSO4·7H2O) could explain the range of water-equivalent hydrogen compositions inferred from neutron spectrometry data. However, they also noted a poor correlation between inferred epsomite-hexahydrite (MgSO4·6H2O) stability boundaries mapped on the surface of Mars on the basis of known variations of temperature and relative humidity and the measured water-equivalent hydrogen abundances. This discrepancy led them to suspect other factors including variations in magnesium-sulfate abundances, the presence of other hydrated phases like gypsum or jarosite, and deviations away from the nominal stoichiometries of magnesium-sulfate phases used in their interpretation as causes for the poor correlation. Calcium sulfates may also play a role in the Martian hydrologic cycle because they occur in hydration states ranging from 0 to 2 mol of H2O per mole of CaSO4.
 This paper investigates hydration and dehydration equilibria under liquid-absent conditions between hexahydrite and starkeyite (MgSO4·4H2O) and uses these results with previously published results on the hydration and dehydration equilibria between epsomite and hexahydrite to understand better the phase relations in the system MgSO4-H2O; the humidity buffer technique used in these studies has proven to be superior in precision to other currently available techniques for investigating experimental reversals for hydration and dehydration reactions at low temperatures [Chou et al., 2002; Chou and Seal, 2003, 2005]. These results are used to evaluate stable and metastable equilibria and then these reactions are extrapolated rigorously using thermodynamic principles to conditions found on Mars. The stable and metastable reactions form a basis from which to evaluate the applicability of rate data for magnesium sulfates presented by recent studies to conditions on Mars.
 Seven hydrated magnesium sulfates have been identified as minerals: kieserite (MgSO4·H2O), sanderite (MgSO4·2H2O), starkeyite (MgSO4·4H2O), pentahydrite (MgSO4·5H2O), hexahydrite (MgSO4·6H2O), epsomite (MgSO4·7H2O), and yet-to-be-named MgSO4·11H2O [Hawthorne et al., 2000; Peterson and Wang, 2006]. A recent crystal structure refinement by Peterson and Wang  found that the phase traditionally considered to contain 12 water molecules (MgSO4·12H2O) actually only contains 11 (MgSO4·11H2O). The compositions of these phases represent a wide range of H2O contents per mole of MgSO4 (Table 1). The structures of these minerals result from varying arrangements of SO4 tetrahedra and Mg(O, OH)6 octahedra, with epsomite and MgSO4·11H2O also having extrapolyhedral water. Hexahydrite, epsomite, and MgSO4·11H2O are characterized by unconnected tetrahedra and octahedra. The difference between hexahydrite and epsomite is the presence of an extrapolyhedral water molecule in epsomite; MgSO4·11H2O has five extrapolyhedral waters relative to hexahydrite. Thus the reaction of MgSO4·11H2O to epsomite, or epsomite to hexahydrite and back, merely results from the loss or gain, respectively, of extrapolyhedral water molecule(s). The structure of pentahydrite consists of infinite chains of alternating octahedra and tetrahedra, whereas that of starkeyite consists of finite clusters of octahedra and tetrahedra and that of kieserite consists of an infinite framework of polyhedra. Thus dehydration from hexahydrite to starkeyite to kieserite requires increased polymerization of octahedra and tetrahedra, which may explain the tendency of hexahydrite to form an amorphous phase upon dehydration under low-pressure conditions in the laboratory [Vaniman et al., 2004].
Table 1. Mineral Chemistry of Magnesium and Calcium Sulfates
Kilograms H2O per Kilogram Anhydrous Sulfate, kg H2O kg−1
 Three calcium sulfates have been identified as minerals: anhydrite (CaSO4), bassanite (hemihydrate; CaSO4·0.5H2O), and gypsum (CaSO4·2H2O). Thus the compositions of calcium sulfates have a more limited range of H2O contents per mole of sulfate compared to the magnesium sulfates (Table 1). The structures of anhydrite and gypsum are based on chains of alternating SO4 tetrahedra and CaO8 dodecahedra; the structure of bassanite consists of similar chains, but the calcium is in ninefold coordination [Hawthorne et al., 2000].
3. Previous Studies
 The phase equilibria of various portions of the MgSO4-H2O system have been investigated by numerous researchers at a range of temperatures under both liquid-present and liquid-absent conditions. Hogenboom et al.  evaluated phase equilibria under liquid-present conditions between 230 and 300 K through experimental studies supplemented by literature review (Figure 1). Similarly, Chou and Seal  investigated phase equilibria between epsomite and hexahydrite under liquid-absent conditions between 298 and 318 K, which included a review of published studies. Phase relations and rates of dehydration and hydration reactions among magnesium sulfates have been investigated by Bish et al. , Vaniman et al. , Chipera et al. , Vaniman and Chipera [2006a, 2006b], and Chipera and Vaniman . They found that hydration and dehydration experiments of durations ranging between 20 and 99 d at temperatures between 100 and 348 K typically produced disequilibrium assemblages consisting of multiple magnesium sulfate phases such as hexahydrite, pentahydrite, and starkeyite or an amorphous magnesium sulfate phase.
 Phase relations among calcium sulfate minerals have been investigated by numerous researchers as summarized by Hardie , Blount and Dickson , and Møller , who provided phase equilibria and thermodynamic data from 298 to 398 K for gypsum, anhydrite, and bassanite (hemihydrate). These studies found bassanite to be metastable at all temperature ranges. They also found that the rate of anhydrite nucleation at temperatures below 343 K was prohibitively low with respect to meaningful equilibrium experimental studies.
4. Procedure and Results
 In this study, we evaluated the stable and metastable equilibria of the system MgSO4-H2O under conditions relevant to the surface of Mars through a combination of new experimental studies and the estimation of thermodynamic parameters. We determined the stability boundary between hexahydrite and starkeyite using the humidity buffer technique, which complements our previous experimental study on the reaction between epsomite and hexahydrite [Chou and Seal, 2003]. We also calculated the stability boundary between epsomite and MgSO4·11H2O based on thermodynamic principles.For the reaction
where ΔGr° is the standard Gibbs free energy of reaction for reaction (1), K is the equilibrium constant, R is the gas constant (8.31451 J mol−1 K−1), T is absolute temperature, aH2O is the activity of H2O, fH2O is the equilibrium H2O fugacity (in MPa), and f*H2O is the fugacity of water (in MPa). The standard states for minerals and H2O are pure solids and H2O, respectively, at 0.1 MPa and temperature. Reversals along four humidity buffer curves between 310 and 334 K at 0.1 MPa were obtained for reaction (1) and are plotted in Figure 2; the experimental data are listed in Table 2, and those along the NaBr humidity buffer curve are plotted in Figure 3. The detailed experimental procedures were described previously [Chou et al., 2002], and the duration of each experimental run is more than 45 h, which is much longer than the duration of about 10 h required for the experimental system to reach a steady state humidity controlled by the buffer [Chou et al., 2002, Figure 6]. Note that the equilibrium boundary between hexahydrite and starkeyite determined in this study was based on reversal experiments as shown in Figure 3, where dehydration reaction occurred at temperatures above 316 K as indicated by the loss of total mass of the initial mixture of hexahydrite and starkeyite, whereas hydration reaction occurred at lower temperatures as indicated by the gain of total mass. It should also be emphasized that both solid reactant and product for the hydration/dehydration reaction were present in the starting sample of each run and therefore activation energy required for the nucleation of a new solid phase was minimized. However, careful identification of the reaction products was essential to ensure that the observed mass changes were not caused by the presence of other phases. This was accomplished by both X-ray diffraction and Raman spectroscopy. Powdered samples were analyzed at room temperature with a Scintag X-1 diffractometer, utilizing CuKα radiation, and digital data were collected every 0.02° 2θ, for 0.5 s per step, from 5° to 65°. Pattern processing and mineralogical analyses were done using Jade software (1995–2005), and no solid phases other than hexahydrite and starkeyite were identified in our sample throughout the experiment. This result was further supported by Raman spectroscopy, which can identify phases in exceedingly minor quantities [Wang et al., 2005, 2006]. Selected grains of samples were analyzed by Raman spectroscopy at room temperature without humidity control. A JY/Horiba LabRam HR Raman system was used, with 532.06 nm (frequency doubled Nd:yttrium/aluminum/garnet (YAG)) laser excitation, a 100 × Olympus objective with 0.9 numerical aperture, and ∼20 mW laser power at sample surface with various analysis times and accumulations. Spectra were acquired with a 600 groove mm−1 grating with a spectral resolution of 2 cm−1, and typical results are shown in Figure 4. They were in excellent agreement with those reported by Wang et al. [2006, Figures 3 and 4 and Table 3] for hexahydrite and starkeyite phases but were distinctly different from spectra of other MgSO4-containing phases, including anhydrous MgSO4, kieserite, sanderite, MgSO4·3H2O, pentahydrite, epsomite, MgSO4·11H2O, and MgSO4 aqueous solution [Wang et al., 2006]. The durations required to collect a complete Raman spectrum were no more than 6 min, and there was no indication of any changes in the hydration states of the sample during the measurement. The volume of the sample under investigation was about 1 μm diameter and 5 μm in depth, and the presence of any other phases should have been easily identified.
Values in parentheses were used to bracket the reaction.
Starting material consisted of a mixture of MgSO4·6H2O and MgSO4·4H2O; this was synthesized from reagent grade MgSO4·7H2O (ACROS, lot 013106101).
Humidity Buffer Mg(NO3)2
Humidity Buffer NaBr
Humidity Buffer KI
Humidity Buffer NaNO3
 Experimental results and derived equilibrium constants are summarized in Table 3. Figure 5 shows the relation between ln K and 1/T for reaction (1), and the standard enthalpy of reaction, ΔHr°, was calculated according to the relation
The value of ΔHr° for reaction (1) is 110.01 ± 1.79 kJ mol−1, which is 3.33 kJ mol−1 higher than the value reported by DeKock . From equation (2) the derived standard Gibbs free energy of reaction at 298.15 K is 21.57 ± 0.15 kJ mol−1, which is 3.30 kJ mol−1 lower than the value calculated from the data compiled by DeKock . The entropy of reaction, ΔSr°, was calculated from the relation
and our value of 296.6 ± 6.5 J mol−1 K−1 for ΔSr° is 20 J mol−1 K−1 higher than the value given by DeKock .
Table 3. Derived Equilibrium Constants for Reaction (1) at 0.1 MPa
Based on equation (2); ln K = 2 ln [(f*H2O/0.1) · (%RH)/100].
310.93 ± 0.51
49.09 ± 0.33
−6.874 ± 0.049
315.95 ± 0.72
52.47 ± 0.44
−6.206 ± 0.069
329.58 ± 0.50
63.58 ± 0.46
−4.465 ± 0.046
333.95 ± 0.75
67.23 ± 0.57
−3.946 ± 0.066
 Stable phase equilibria for the system MgSO4-H2O as a function of temperature and RH are shown in Figure 6 as solid lines. Metastable equilibria are shown as dashed lines. Experimental studies and interpretations of phase equilibria data under liquid-saturated conditions at 0.1 MPa suggest that equilibrium dehydration reactions in the system MgSO4-H2O in order of increasing temperature involve the solid phases MgSO4·11H2O, MgSO4·7H2O, MgSO4·6H2O, and MgSO4·H2O [Hogenboom et al., 1995; Peterson and Wang, 2006] (Figure 1). Thus starkeyite is metastable under all conditions, and the reaction between hexahydrite and starkeyite investigated in the present study represents a metastable equilibrium, albeit a reversible one. As shown in Figure 1, the location of the metastable isobaric invariant point E for the assemblage hexahydrite-starkeyite-aqueous solution-vapor is within the stability field of liquid + kieserite. Other key isobaric invariant points along the liquid-saturated curve include the transition of MgSO4·11H2O to epsomite at 275.2 K (point C in Figures 1, 2, and 6), the transition of epsomite to hexahydrite at 321.2 K (point A in Figures 1, 2, and 6), and the transition of hexahydrite to kieserite at 342.2 K (point B in Figures 1, 2, and 6) [Hogenboom et al., 1995; Chou and Seal, 2003]. In Figures 2 and 6 the location of the epsomite-hexahydrite reaction is from Chou and Seal . The location of the hexahydrite-kieserite boundary is anchored by the invariant point and extrapolated to lower temperatures on the basis of the well-defined isobaric invariant point for the assemblage hexahydrite-kieserite-aqueous solution-vapor at 342.2 K (point B in Figures 1 and 2) and the more recent enthalpy data for hexahydrite and kieserite [DeKock, 1986]. This invariant point is further confirmed by the vapor pressure measurements of van't Hoff et al.  (see Figure 2).
 The potential role of magnesium sulfates in the Martian water cycle is further enhanced by consideration of the role of MgSO4·11H2O, a phase not rigorously considered by previous investigators of Martian water cycles, with the exception of the recent papers by Peterson and Wang , Wang et al. , and Freeman et al. . The location of the MgSO4·11H2O-epsomite reaction in the presence of liquid is well established at 275.2 K (point C in Figures 1, 2, and 6) [Peterson and Wang, 2006]. Unfortunately, the location of this reaction in terms of temperature and RH has not been determined experimentally, nor have the enthalpy and heat capacity of MgSO4·11H2O been measured to facilitate the calculation of the stability boundary of this reaction. However, Hemingway et al.  found that the thermodynamic properties of hydrated sulfate minerals, including enthalpy and Gibbs free energy of formation and entropy, can be accurately estimated on the basis of the summation of known properties of component phases (Appendix A). Subsequent experimental studies have confirmed the accuracy of these estimation methods [Chou et al., 2002; Chou and Seal, 2003, 2005]. Thus the location of the MgSO4·11H2O-epsomite stability boundary as a function of temperature and RH can be accurately estimated (Figure 6). This reaction boundary intersects the range of temperature and RH documented on the surface of Mars near the middle of the spectrum of diurnal variations [Savijärvi, 1995], suggesting that MgSO4·11H2O may play an important, if not dominant, role in the daily water cycle on Mars.
 The epsomite-hexahydrite and hexahydrite-kieserite curves intersect around 284 K, thereby limiting the stability field of hexahydrite such that it would not be expected as a stable phase on the surface of Mars. In addition, if instead, metastable equilibria involving starkeyite prevail, then the reactions between hexahydrite-starkeyite and hexahydrite-epsomite would also limit the stability field of hexahydrite to conditions significantly warmer than those of the Martian equatorial surface (above point F in Figure 6). All curves that extend below 273 K are deflected because of the change in the reference state of H2O (f*H2O) in relative humidity calculations; the reference states are vapor pressures of pure water and ice at temperatures above and below 273 K, respectively. Values of f*H2O for ice were calculated from the relation
and the calculated f*H2O values are accurate to better than 0.1% (K.G. Libbrecht, Physical properties of ice, 2007, available at http://www.its.caltech.edu/∼atomic/snowcrystals/ice/ice.htm). An important ramification of the change in standard state of H2O and its effect on the location of the dehydration curves is that the curves intersect the range of surface conditions on Mars more toward the middle of the range rather than at the low-humidity extreme. Thus it would appear likely that dehydration and hydration reactions of magnesium sulfates play a more important role in the diurnal water cycle on Mars than previously appreciated.
 The stable and metastable equilibria described above form a sound basis from which to evaluate potential sets of hydration and dehydration reactions that may be contributing to the water cycle on Mars in terms of participating phases. In addition, several studies have investigated the rates of hydration and dehydration reactions for magnesium sulfates. Vaniman et al.  demonstrated that dehydration and hydration reactions between epsomite and hexahydrite were rapid at 298 K, consistent with the results of Chou and Seal . However, they found that laboratory dehydration of hexahydrite at low humidity (0.5% RH) produced an amorphous magnesium-sulfate phase with approximately 22 wt % H2O (well in excess of the 13 wt % H2O in kieserite) within the time span of a day, which persisted for at least 4 months. Hydration of the amorphous material at 298 K and 31% RH produced a disequilibrium assemblage of hexahydrite, pentahydrite (MgSO4·5H2O), and starkeyite. In contrast, hydration of kieserite formed hexahydrite, followed by epsomite with continued hydration. Therefore short-term transitions in magnesium sulfates display a range of stable and metastable behaviors. Vaniman and Chipera [2006b] and Chipera and Vaniman  obtained similar results over a wider temperature range (243 to 348 K) in experiments ranging in duration from 12 d to up to a year. McCord et al.  found the rates of epsomite dehydration at lower temperatures relevant to the Jovian satellite Europa (< 130 K) to be prohibitively slow under those conditions, but their data suggest that dehydration may occur in reasonable time frames at higher temperatures associated with equatorial Mars (175 to 255 K) to allow epsomite to play an active role in the Martian water budget.
 Improved knowledge of variations in magnesium sulfate abundances or the presence of significant concentrations of other hydrated phases requires evidence from the planet surface, either directly analyzed on the surface or remotely sensed from orbit. Feldman et al. [2004b] modeled the distribution of water-equivalent hydrogen using the assumption of the uniform presence of 10 mass % MgSO4 in the soils. However, SO3 concentrations in soils and rocks are known to vary from 4.1 to 31.7 wt % and from 0.52 to 12.9 wt %, respectively; magnesium and calcium are present in subequal proportions [Clark et al., 1976; Rieder et al., 2004; Gellert et al., 2006]. Furthermore, thermal emission spectrometry data also suggest the presence of calcium sulfates on Mars [Christensen et al., 2004], and Mössbauer spectrometry has identified minor amounts of jarosite [(K,Na,H3O+)Fe3(SO4)2(OH)6] [Klingelhöfer et al., 2004] as other potential sources of water-equivalent hydrogen.
 Of the factors proposed by Feldman et al. [2004b] to explain the poor correlation between the inferred epsomite-hexahydrite and epsomite-kieserite stability boundaries mapped on the surface of Mars and the measured water-equivalent hydrogen abundances, the results of our study provide greater insights into the stoichiometries of hydrated magnesium sulfates potentially involved in the diurnal Martian water cycle. The results of our experiments suggest that the metastable epsomite-starkeyite stability boundary should be considered because it falls within the range of diurnal variations in temperature and relative humidity, albeit near the low relative humidity/high-temperature end of the spectrum (Figure 6).
 The possible occurrence of MgSO4·11H2O in the near-surface environment of Mars is more likely to have a profound effect on the diurnal water cycle than uncertainties related to the presence or absence of kieserite [Peterson and Wang, 2006]. One kilogram of MgSO4 as kieserite contains only 0.15 kg of H2O, whereas 1 kg of MgSO4 as MgSO4·11H2O contains 1.80 kg of H2O. For comparison, 1 kg of MgSO4 as starkeyite contains 0.6 kg of H2O, and 1 kg as epsomite contains 1.05 kg of H2O. Our results indicate that hexahydrite should not be stable on the surface of Mars but 1 kg of MgSO4 as hexahydrite would contain 0.90 kg of H2O. Thus diurnal cycling of H2O with MgSO4 between MgSO4·11H2O and either starkeyite or kieserite would involve 1.20 or 1.65 kg, respectively, of H2O. From a crystal chemical perspective the cycling between MgSO4·11H2O and epsomite is likely to be kinetically favorable because the transition merely involves the removal or addition of extrapolyhedral water molecules. The transitions from epsomite, which should be stable near the surface of Mars, to starkeyite and back, which have not been studied experimentally, may be more difficult because they require disruption of the arrangement of SO4 tetrahedra and Mg(O, OH)6 octahedra.
 Calcium sulfates are less likely to be important sources and sinks of water in the Martian water cycle despite the fact that geochemical and spectral data suggest that they may be equally as abundant as magnesium sulfates [Rieder et al., 2004; Christensen et al., 2004; Langevin et al., 2005]. The extremes of hydration states are gypsum (CaSO4·2H2O) and anhydrite (CaSO4), which only have a difference of 0.27 kg of H2O per kg of CaSO4. This pales in comparison to the mass differences between the magnesium sulfates likely to be found on Mars (Table 1 and Figure 6). In addition, the rates of hydration and dehydration reactions between gypsum and anhydrite are known to be slow, even at saturated conditions near 298 K [Hardie, 1967; Blount and Dickson, 1973]. Thus magnesium sulfates, ranging in hydration state from MgSO4·11H2O to MgSO4·H2O (kieserite), may play the dominant role in the Martian water cycle away from the poles.
 Despite increasing degrees of sophistication in laboratory experiments investigating magnesium-sulfate hydration and dehydration pathways and rates, the level of understanding of these processes on Mars will remain speculative to a certain degree until more detailed data are available from the surface of Mars. For example, this study and Chou and Seal  found significant reaction between hexahydrite and starkeyite and between epsomite and hexahydrite, respectively, in experiments lasting less than a few days when both reactants and products were present in the experimental charges. In contrast, Vaniman et al.  and Chipera and Vaniman  noted slower reaction rates in their experiments, which lacked seed crystals of run products. Likewise, the absence of kieserite in laboratory studies [Vaniman et al., 2004] does not necessarily preclude its involvement in the diurnal Martian water cycle. In fact, recent visible near-infrared hyperspectral imaging has identified kieserite on the Martian surface [Gendrin et al., 2005]. Similarly, textural evidence discussed by Peterson and Wang  suggests that MgSO4·11H2O may play an important role in mass wasting on the surface of Mars despite the fact that laboratory studies by Vaniman and Chipera [2006a, 2006b] suggest that low-temperature, nighttime hydration may be too sluggish to form these phases on a diurnal timescale. The presence of other cations as impurities, such as Ca or Fe, may also favor the stability of some phases over others on the planet depending upon solid solution effects. For example, siderotil (FeSO4·5H2O) is not stable in the pure FeSO4-H2O system, but Jambor and Traill  found that solid solution of Cu was necessary to stable the siderotil structure.
 On the basis of phase equilibria considerations, MgSO4·11H2O may be the most likely source of liquid water on Mars, in the absence of ice, in the context of modern or recent erosional or other geomorphologic processes potentially associated with liquid water on the surface of Mars [Peterson and Wang, 2006]. The abundance and likelihood of occurrence of MgSO4·11H2O, its limited stability relative to known variations in temperature and relative humidity, and the rate with which it dehydrates in the laboratory make it a more likely candidate for sources of large quantities of water than other water- or hydroxyl-bearing phases, i.e., other sulfates such as jarosite and gypsum, or silicate minerals such as clays and zeolites found on the planet [Christensen et al., 2004; Klingelhöfer et al., 2004].
 Magnesium sulfates probably play a dominant role in the water cycle of Mars away from the polar ice caps through hydration and dehydration reactions involving atmospheric moisture during diurnal cycles of heating and cooling. Important phases may include MgSO4·11H2O and epsomite at higher relative humidities during the night with starkeyite or kieserite being important during the day. The involvement of MgSO4·11H2O in the water cycle of Mars provides far greater flexibility in interpreting the less than ideal correlations between measured water-equivalent hydrogen abundances and inferred mineral stabilities than was previously possible. Nevertheless, global variations in the concentration of MgSO4 in the near-surface environment of Mars must also be factored into future analyses because large variations would be expected. In addition, interpretation of the potential role of magnesium sulfates in the Martian water cycle must be approached with caution because of the uncertainties in reaction pathways at the lower temperatures of the Martian surface, which have not yet been investigated experimentally; in the importance of seed crystals in facilitating hydration and dehydration reactions in short time frames; and in solid solution effects expanding the stability fields of some magnesium-sulfate minerals relative to others.
Appendix A:: Calculations for the Phase Boundary Between Epsomite and MgSO4·11H2O
 The phase boundary between epsomite and MgSO4·11H2O was calculated by the following steps:
 1. Obtain the equilibrium constant for the reaction
at the invariant point C. The Pitzer model [Plummer et al., 1988] for the epsomite-saturated solution at 295.15 K indicates the activity of H2O (aH2O) = 0.9577 and the corresponding ln K = −20.385 based on the relation ln K = 4 ln aH2O.
 2. Calculate the enthalpy of formation from elements (ΔHf°) for MgSO4·11 H2O. Using the ΔHf° data compiled by DeKock  for MgSO4·H2O, MgSO4·2 H2O, MgSO4·4 H2O, MgSO4·6 H2O, and MgSO4·7 H2O, the enthalpy contribution for each water of crystallization in hydrated sulfate salts, except for the first water, at 298.15 K was calculated by the relation
where ΔHxw° is the enthalpy contribution for each additional water of crystallization at 298.15 K, ΔHf°, i and ΔHf°, j are the enthalpy of formation from elements at 298.15 K for the hydrous phases i and j, respectively, and n = I − J, where I and J are the stoichiometric coefficients for water in the hydrous phases i and j, respectively. Using all hydrous phase pairs, we obtained the average value for ΔHxw°, = −297.77 kJ mol−1, and ΔHf° for MgSO4·11 H2O was calculated by
 3. Calculate ln K for reaction (2) at various temperatures based on the relation
where ΔHr° = 223.88 kJ mol−1, based on the ΔHf° value of −241.8 kJ mol−1 [Robie and Hemingway, 1995] for water and the ΔHf° values for epsomite and MgSO4·11 H2O described above.
 Nadine Piatak provided X-ray diffraction analysis of run products. The manuscript benefited from thoughtful and constructive reviews from Jeff Grossman, Jane Hammarstrom, David Vaniman, and Alian Wang. Discussions with Ron Peterson provided helpful insights into crystallographic differences among magnesium sulfates. The research was funded by the Mineral Resources Program of the U.S. Geological Survey.