A large fraction of atmospheric mineral aerosol is composed of clays, such as smectites. Smectites are known to expand upon addition of water, and thus their properties in the atmosphere may vary strongly with relative humidity (RH). Here we report on the adsorption of water to Na-montmorillonite, a smectite clay. We probe the water uptake under conditions representative of the Earth's troposphere, as well as those relevant for the surface of Mars, of which montmorillonite is a proposed component. Using a vacuum chamber equipped with transmission Fourier transform infrared spectroscopy, we find that Na-montmorillonite contains 10% water by mass at temperatures from 212 K to 231 K at 50% RH. Surprisingly, the water uptake by Na-montmorillonite is almost as great as that of deliquesced ammonium sulfate. We also find that although water adsorption to Na-montmorillonite depends strongly on RH, there is not a strong dependence on absolute temperature. In addition, we find the time required for the clay to become saturated with water decreases with increasing water vapor pressure and is much shorter than suggested in previous studies. We discuss the implications of these results for the Earth's troposphere and the potential role of montmorillonite in the Martian hydrologic cycle.
 Mineral particles are commonly emitted into Earth's atmosphere by eolian forces and are transported on global scales. Plumes of mineral dust are persistent features visible in satellite images [Prospero and Lamb, 2003] and have been estimated to account for ∼45% of the total atmospheric aerosol load [Caquineau et al., 2002]. The optical properties of mineral particles have been studied [Sokolik, 2002] and numerical simulations of their transport conducted [Colarco et al., 2003], but the level of understanding of their impact on climate remains very low [Intergovernmental Panel on Climate Change (IPCC), 2001]. Mineral particles have also been studied as potential ice nuclei [DeMott et al., 2003; Sassen et al., 2003]. The resulting enhancement in ice nuclei concentrations by mineral particles may have an indirect effect on climate through the modification of cloud properties. Mineral particles may also impact Earth's atmosphere by serving as reactive surfaces [Usher et al., 2003]. Support for this assertion has been obtained in field studies which have detected significant amounts of NO3− and SO4= associated with dust events [Jordan et al., 2003]. In addition, individual particles have been found to contain sulfate, nitrate, water-soluble organic acids and mineral components [Lee et al., 2002].
 Modeling studies have predicted clays to be an important fraction of airborne dust [Claquin et al., 1999; Sokolik and Toon, 1999]. Montmorillonite, a smectite clay, is believed to be a component of the clay fraction. The smectite group of clays has the unique ability to expand its structure, thereby increasing its surface area upon addition of water. Mineral dust is found from the Earth's surface up to an altitude of about 10 km, where temperatures are commonly near 220 K. Although well characterized at 298 K [Cases et al., 1992; Hall and Astill, 1989; Xu et al., 2000], the adsorption of water onto clays under low-temperature conditions representative of the upper troposphere is not well understood. The first purpose of this study is to characterize the water content and specific surface area (SSA) of Na-montmorillonite under conditions representative of the Earth's troposphere.
 In addition to its importance on Earth, water vapor is also an important and highly variable trace gas in the Martian atmosphere [Titov, 2002]. The variability is due to a global water cycle, with water subliming from the summer pole, moving equatorward, forming clouds and re-condensing on the winter pole. A component of the hydrologic cycle could be exchange between the regolith, or soil, of Mars and the atmosphere [Tokano, 2003]. The presence of clays on Mars was first suggested from elemental analysis of Martian soil collected by Viking [Baird et al., 1976; Toulmin et al., 1977] and infrared spectra of atmospheric dust from Mariner 9 [Hunt et al., 1973; Toon et al., 1977]. Satellite measurements of the surface of Mars by the Infrared Space Observatory (ISO) Short Wavelength Spectrometer (SWS) and the infrared mapping spectrometer (ISM) on Phobos 2 are consistent with the presence of hydrous clays such as montmorillonite [Bibring and Erard, 2001]. Measurements of airborne particles from the ISO SWS also support the presence of montmorillonite on Mars [Fedorova et al., 2002]. However, data from the Thermal Emission Spectrometer on Mars Global Surveyor does not seem consistent with the presence of smectite clays [Christensen et al., 2001]. Considering factors such as the variation of infrared spectra of smectite group clays depending on cation substitution [Bishop et al., 2002], the ultimate importance of clay on Mars is not yet established.
 Laboratory studies have been performed to determine the potential role of montmorillonite in a global water cycle on Mars. Zent et al.  reported that under conditions representative of the Martian surface, 211 K and 32% RH with respect to water, the interlayer sites of Na-montmorillonite are not available for adsorption. They also report that the 1/e loading time for saturation of the clay with water is too long for clays to significantly affect the Martian diurnal water cycle. However, it has recently been suggested that the interlayer space of these samples may have been collapsed [Bish et al., 2003]. Here we report on the water uptake by Na-montmorillonite clay under Martian conditions for samples that have been prepared in a manner to avoid irreversible collapse of the interlayer space. We address the relevance of these results to the variability in water vapor on Mars.
2.1. Vacuum Chamber
 Experiments were performed in a high vacuum, stainless steel chamber described previously in the literature [Frinak et al., 2004]. Transmission Fourier Transform Infrared (FT-IR) spectroscopy was used to probe the condensed phase products. The main chamber housed an ionization gauge and an MKS Baratron capacitance manometer for measurement of total chamber pressure. The chamber was pumped through an escape orifice using a 210 L/s turbomolecular pump (Pfeiffer). Samples were supported on a silicon wafer attached with stainless steel clips to a gold-coated copper mount. Indium foil was placed between the silicon wafer and the mount to improve thermal contact. The entire assembly was attached to a liquid-nitrogen-cooled cryostat. Heating against the liquid nitrogen occurred on the back side of the sample support inside the vacuum jacket to ensure that the temperature of the mount, wafer, and sample was as uniform as possible. Two type-T thermocouples were attached to the mount to monitor the temperature. Temperature could be controlled from 100 K to 300 K using a Eurotherm temperature controller. Samples were isolated from the gas flow by covering them with a Teflon cup. Using an O-ring, the Teflon cup sealed against the stainless steel vacuum jacket, allowing a constant signal of water to be established without exposing the sample to the reactant gas.
2.2. Sample Preparation
 The Na-montmorillonite sample was SWy-2 obtained from the Source Clays Repository of the Clay Minerals Society. Table 1 lists the elemental composition of SWy-2 [Mermut and Cano, 2001]. SWy-2 samples were prepared from a water slurry of 25 mL of water and 60–100 mg of the clay. Using a Pasteur pipet, several drops of the slurry were deposited onto the silicon wafer. Samples were dried under ambient conditions, and then placed in the vacuum chamber. The chamber was then evacuated for 24 to 48 hours until a total chamber pressure of approximately 2 × 10−7 Torr was achieved. Sample masses were typically around 1 mg, as measured under ambient conditions upon removal from the vacuum chamber following the experiments.
 While the Teflon cup was open, water adsorbed to the sample was monitored by transmission FT-IR spectroscopy. Spectra were acquired using a Nicolet 740 FT-IR spectrometer equipped with an MCT-B detector. For all experiments, 32 scans were collected at a resolution of 4 cm−1. Condensed phase water in the clay was observed by ratioing to a background of dry SWy-2 collected prior to exposure. Water content at saturation was quantified using Beer's Law as follows. First, the base e absorption coefficient for bulk water (cm−1) [Downing and Williams, 1975] was divided by the number density of water, 3.34 × 1022 molec/cm3, to yield an absorbance cross section a (cm2/molec). Then a probed coverage θ (molec/cmprobed2) was calculated from
where A is the base 10 absorbance. Assuming a uniform water coverage over the entire sample, the probed coverage was multiplied by the geometric surface area of the sample, 2 cm2, to determine the total number of water molecules adsorbed over the entire sample. The number of adsorbed molecules was then converted to a mass of water and divided by the measured weight of the sample to give the clay water content in units of mgwater/gclay. The water spectral features at 3350 cm−1 and 1640 cm−1 were independently used for the calculation of adsorbed water and were in excellent agreement with each other. Reported values for clay water content are the average from these two peaks.
 A typical experiment consisted of cooling to the desired temperature and collecting a background scan. The Teflon cup was then closed, isolating the sample from the chamber. Afterward, a constant flow of water (Fisher Scientific, HPLC grade) was established. The time to achieve a constant water flow varied from 10 min to 1 hour, depending on the pressure of water vapor. Once a constant water flow was established, the cup was retracted, exposing the sample. Samples were allowed to equilibrate with the water from 1 to 4 hours. It was possible to perform multiple experiments on the same sample. After an experiment, samples were heated to 295 K and evacuated overnight. By the following day, the sample was dry and the removal of the adsorbed water was confirmed in the FT-IR spectrum. Experiments were repeated under similar conditions on the same sample to ensure reproducibility of measured water content on a previously exposed sample.
Figure 1 shows FT-IR spectra of SWy-2 collected before and after water exposure at 212 K. These spectra are in excellent agreement with published spectra collected at room temperature [Bishop et al., 1994; Madejová and Komadel, 2001]. Table 2 summarizes the prominent features and their peak assignment based on IR measurements collected for SWy-2 containing adsorbed water [Madejová and Komadel, 2001]. After the adsorption of water, there is a slight change in the Si-O features near 1100 cm−1. The vertical lines in the inset of Figure 1 show the peak position before and after adsorption of water to the sample. This shift has been documented in the literature [Yan et al., 1996], but only at significantly higher water content at room temperature. The shoulder at 1120 cm−1 appears sharper and more resolved as the peak near 1050 cm−1 becomes sharper and shifts to lower wave numbers as compared with the dry spectrum. The biggest changes in the spectrum as RH increases occur around 3350 cm−1 and 1640 cm−1. These peaks represent the stretch and bend of liquid water, respectively, and include both surface and interlattice adsorbed water.
Figure 2 shows a representative uptake curve of water adsorption on SWy-2 as seen in the FT-IR spectrum at 1640 cm−1. This data was collected at 222 K as 1.6 mg of SWy-2 was exposed to water at 29% RH. On the basis of a single exponential fit through zero, the signal reaches a plateau at 0.0607 absorbance units with a 1/e loading time of approximately 10 min. The absorbance of 0.0607 at saturation was then used to calculate the clay water content as described previously.
Figure 3 shows the clay water content as a function of RH for a series of experiments performed at 222 K. It can be seen that the clay water content increases markedly with increasing RH. At approximately 50% RH at 222 K, the SWy-2 contains 10% water by mass. These values are in excellent agreement with room temperature, gravimetric studies [Cases et al., 1992; Hall and Astill, 1989; Xu et al., 2000]. They are also in excellent agreement with room temperature studies conducted by Zent et al. . This agreement suggests that water content of Na-montmorillonite is strongly dependent on RH, but independent of temperature. The error bars represent the variability in the data as determined from experiments performed under similar conditions. In addition to comparing our low-temperature results to published room temperature data, we explicitly studied the temperature dependence. Figure 4 shows the resulting clay water content as a function of temperature for experiments performed at 30% RH. It can be seen that the saturated value for the clay water content is independent of temperature from 212 K to 231 K. This is not surprising given the close agreement between the clay water content we measure at low-temperature and previous measurements at room temperature. The only point of disagreement is the measurement by Zent et al.  at 211 K and 32% RH. They determined a clay water content of 11 mgwater/gclay, which is approximately a factor of 5 lower than our values under similar conditions. The variability in these measurements may be due to sample preparation and will be discussed in more detail below.
 The water content of Na-montmorillonite is dependent on RH because of the ability of the clay to swell. The swelling of Na-montmorillonite takes place in discrete stages and is well characterized at room temperature in the literature both experimentally using X-ray diffraction techniques [Cases et al., 1992] and using models [Hensen and Smit, 2002]. Briefly, as SWy-2 is exposed to water vapor below 20% RH, the water adsorbs predominately to external surfaces of the tactoids, or particles. As the RH increases toward 20%, the tactoid size decreases from a dry size of about 20 clay layers to the final tactoid size of about six layers resulting in an SSA increase of less than a factor of 4 [Cases et al., 1992]. From 20% to 48% RH, the water molecules solvate the interlayer cations causing the structure to expand until reaching monolayer coverage, internal and external, at 48% RH [Hensen and Smit, 2002; Newman, 1983]. Beyond 48% RH, multilayer coverage, internal and external, occurs. However, when Na-montmorillonite is heated to 473 K at ambient pressure, prior to dehydroxylation, the interlayer space irreversibly collapses [Ertem, 1972; Fripiat et al., 1960; Mooney et al., 1952]. The samples studied by Zent et al.  were heated to 383 K under vacuum which may have led to the irreversible collapse of the interlayer whereas our samples were not heated above 295 K under vacuum. If only the external surface was available for adsorption, the value reported in the literature for the saturated water content at 32% RH would be consistent with the present study [Zent et al., 2001]. The factor of 5 discrepancy could simply be accounted for by the differences in the accessible SSA of the two samples. Our results indicate, contrary to the conclusions of Zent et al. , that the interlayer space of Na-montmorillonite is capable of expanding and the SSA will increase with RH under Martian conditions.
 We have determined the SSA of our samples at three values of RH, and these are listed in Table 3. The dry SSA of 320 cm2/mg in Table 3 is from N2 BET measurements provided by the Source Clays Repository which we also independently confirmed. The remaining values were estimated from water uptake measurements by converting the molecules of water adsorbed, as calculated using Beer's Law, to a coverage by assuming that each water molecule covered 0.106 nm2 [Newman, 1983]. The coverage of water was then divided by the measured sample mass to give an SSA (cm2/mg). These estimates are restricted to RH values where the assumption of monolayer coverage is expected to be accurate. At 20% RH where we expect the tactoid size to have reached its minimum and the water to not yet penetrate the interlayer space, we determine an external SSA of 1200 cm2/mg. This estimate is consistent with the reported external SSA of 1200 cm2/mg at 16% RH at 295 K [Cases et al., 1992]. At 48% RH, where we expect one monolayer of adsorbed water on the external and interlayer space [Newman, 1983], we determine an SSA of 3300 cm2/mg from FT-IR measurements. The SSA at 48% RH is a factor of 3 larger than the external surface area alone and is also a full order of magnitude larger than the dry SSA. Our SSA estimates are consistent with the range of SSA values reported in the literature and summarized in Table 3 [Cases et al., 1992].
 Given the adsorption regimes defined above, at 48% RH we expect our monolayer coverage on a 0.8 mg sample of SWy-2 to consist of 2.4 × 1018 molecules of water. On the basis of similar experiments at 20% RH, we attribute 1.0 × 1018 of those water molecules to be adsorbed externally. The remaining 1.4 × 1018 water molecules are believed to reside in the interlayer space. Using the interlayer cation exchange capacity of 70 cmol/kg reported by Mermut and Lagaly , we calculate four water molecules associated with each interlayer cation at monolayer coverage. On the basis of the literature, we expect the solvation of monovalent cations by three water molecules and bivalent cations by more than three water molecules [Sposito and Prost, 1982]. The interlayer cation in SWy-2 is predominately sodium, but it also contains some calcium. Thus, determining four water molecules solvating each interlayer cation is consistent with a sample containing both monovalent and divalent species.
 Our data suggest that even at low temperature, SWy-2 can swell and incorporate additional water into its structure as the RH increases. However, to assess the atmospheric relevance of this process, the timescale for water uptake must be considered. Figure 5 shows the 1/e loading time to reach saturation as a function of water vapor pressure. It can be seen that the 1/e loading time decreases with increasing water vapor pressure. At 222 K and 10% RH, which corresponds to 4.5 mTorr of water, the 1/e loading time is 24 min. However, at the same temperature, when the water vapor pressure is increased to 30 mTorr, which is 61% RH, the 1/e loading time is less than 4 min. These results suggest that saturation of the clay surface with water may occur quickly under temperatures representative of the troposphere. Thus clays transported globally in the Earth's troposphere are likely to become quickly saturated with water.
 A similar trend of decreasing 1/e loading time with increasing water vapor pressure was reported in the study by Zent et al. . Although the trends are similar, the actual values for the loading times are quite different. Zent et al.  found the timescale to reach saturation was in excess of 48 hours and the reported 1/e loading time was 21.6 hours at 211 K and 32% RH. Our SWy-2 samples under similar conditions of 212 K and 29% RH have a 1/e loading time of less than 1 hour. We have suggested that the interlayer space of the samples in the Zent et al.  study was collapsed due to sample preparation. It has been proposed that the collapsed interlayer may account for the long time to reach saturation [Bish et al., 2003]. However, a specific mechanism explaining the longer adsorption time has not been determined at this time. In addition to measuring higher water contents under Mars-like conditions, we also find the timescale for the adsorption of water onto Na-montmorillonite to be significantly shorter than previous studies [Zent et al., 2001].
Figure 6 compares the water content measured for SWy-2 at 222 K with a well-characterized, deliquescent species known to be important in the troposphere, ammonium sulfate. The circles in the plot represent the values from Figure 3 converted to a mole ratio using the molecular weight of 360 g/mole for montmorillonite reported in the literature [Mermut and Lagaly, 2001]. The dashed and solid lines represent the deliquescence and efflorescence branches, respectively, of the ammonium sulfate growth curve [Tang and Munkelwitz, 1994]. It can be seen from this plot that at RH less than 37%, ammonium sulfate is dry while SWy-2 contains approximately 5% water by mass. Thus clays could offer a much wetter environment for additional heterogeneous chemistry than ammonium sulfate at low RH. Although ammonium sulfate may contain more water at RH values above 37%, this water content is only possible after the deliquescence point of 80% RH has been reached. Depending on prior processing of the particle, clays may also contain significantly more water than ammonium sulfate from 37% to 80% RH. This again points to the potential importance of clays as a medium for atmospheric heterogeneous chemistry.
4. Implications for Earth and Mars
 We have shown the saturated water content of SWy-2 changes significantly with RH. We find the water content increases with increasing RH independent of absolute temperature. Around 50% RH at 222 K, SWy-2 contains approximately 10% water by mass. Under these conditions, this water content is achieved with a 1/e loading time of less than 6 min. Our results show the 1/e loading time decreases with increasing water pressure, and the timescale suggests this process is potentially important for both the Earth's troposphere and the Mars global water cycle.
 In addition to enhanced water content, the optical properties of SWy-2 vary with RH as shown in the FT-IR spectrum. This observation may be important to modeling studies which use only one set of optical constants for a particular mineral dust component. It may be important to consider that during transport from the arid source regions to areas of higher RH, the properties of the mineral dust itself, especially dust containing smectite clay, will change dramatically.
 Previous studies have investigated montmorillonite as a potential component of the Martian water cycle. One study concluded that under Mars-like conditions, the interlayer space was not available for water adsorption and the timescales for equilibration were too long for the interaction to be of any significance in a diurnal water cycle [Zent et al., 2001]. Our study, using a different sample preparation method, suggests that the interlayer space is indeed available on a much shorter timescale. If present on Mars, montmorillonite could be a reservoir for bulk water participating in the distribution of water vapor by releasing water to the atmosphere in warmer daytime conditions, and re-adsorbing water vapor under cooler nighttime temperatures [Tokano, 2003]. Also, regions of biological interest have often been linked to possible locations of liquid water. If present on Mars, regions of regolith containing montmorillonite could serve as water-bearing substrates capable of sustaining life [Bish et al., 2003].
 This work was supported by National Science Foundation under grant ATM-0137261.