Notice: Wiley Online Library will be unavailable on Saturday 30th July 2016 from 08:00-11:00 BST / 03:00-06:00 EST / 15:00-18:00 SGT for essential maintenance. Apologies for the inconvenience.
 Acid-sulfate alteration of basalt by SO2-bearing volcanic vapors has been proposed as one possible origin for sulfate-rich deposits on Mars. To better define mineralogical signatures of acid-sulfate alteration, laboratory experiments were performed to investigate alteration pathways and geochemical processes during reaction of basalt with sulfuric acid. Pyroclastic cinders composed of phenocrysts including plagioclase, olivine, and augite embedded in glass were reacted with sulfuric acid at 145 °C for up to 137 days at a range of fluid : rock ratios. During the experiments, the phenocrysts reacted rapidly to form secondary products, while the glass was unreactive. Major products included amorphous silica, anhydrite, and Fe-rich natroalunite, along with minor iron oxides/oxyhydroxides (probably hematite) and trace levels of other sulfates. At the lowest fluid : rock ratio, hexahydrite and an unidentified Fe-silicate phase also occurred as major products. Reaction-path models indicated that formation of the products required both slow dissolution of glass and kinetic inhibitions to precipitation of a number of minerals including phyllosilicates and other aluminosilicates as well as Al- and Fe-oxides/oxyhydroxides. Similar models performed for Martian basalt compositions predict that the initial stages of acid-sulfate alteration of pyroclastic deposits on Mars should result in formation of amorphous silica, anhydrite, Fe-bearing natroalunite, and kieserite, along with relict basaltic glass. In addition, analysis of the experimental products indicates that Fe-bearing natroalunite produces a Mössbauer spectrum closely resembling that of jarosite, suggesting that it should be considered an alternative to the component in sulfate-rich bedrocks at Meridiani Planum that has previously been identified as jarosite.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Observations made at Mars over the last decade by orbiting spacecraft and landers have documented the widespread occurrence of sulfate-rich rocks and soils across the planet [e.g., Squyres et al., 2004, 2007; Bibring et al., 2005, 2006; Gendrin et al., 2005; Arvidson et al., 2005; Murchie et al., 2007]. The mineralogy, areal extent, and geologic setting of these deposits vary considerably among different sites, suggesting a variety of formation conditions and alteration histories have been involved in their formation. Among the sulfate minerals that have been inferred to be present on Mars based on spectral interpretations are various hydrated Mg-, Fe- and Ca-bearing sulfates, including kieserite, gypsum, szomolnokite, epsomite, and members of the alunite-jarosite mineral group [e.g., Klingelhöfer et al., 2004; Bibring et al., 2005, 2006; Gendrin et al., 2005; Milliken et al., 2008; Murchie et al., 2007; Bishop et al., 2009].
 While a variety of mechanisms have been proposed to explain the origin of Martian sulfate deposits in different settings, it is likely that at least some of the deposits were formed through alteration of basaltic crust at elevated temperatures. Indeed, the OMEGA Team has suggested that “sulfur-rich fluid circulation and the alteration of volcanic ashes into sulfates” may be a major contributor to the globally distributed sulfate deposits [Gendrin et al., 2005]. The Mars Exploration Rover (MER) Spirit has observed sulfate accumulations that suggest acid-sulfate alteration in fumarolic or hot springs environments within Gusev Crater [Squyres et al., 2007; Morris et al., 2008]. In addition, McCollom and Hynek [2005, 2006] have proposed that the layered sulfate deposits observed at Meridiani Planum were formed through alteration of volcanic ash by SO2-bearing vapors under hydrothermal conditions. Other scenarios that have been proposed for the origin of various sulfate-rich deposits on Mars include precipitation during evaporation of acidic, sulfate-bearing groundwater [Squyres et al., 2004, 2009; McLennan et al., 2005; Clark et al., 2005; Wang et al., 2006], low-temperature weathering of basalts by an “acid fog” formed from SO2 in the atmosphere [Banin et al., 1997; Zolotov and Mironenko, 2007; Schiffman et al., 2006; Tréguier et al., 2008], acid weathering of glacial deposits [Niles and Michalski, 2009], oxidation of pyrite or other sulfide minerals [Zolotov and Shock, 2005; Dehouck et al., 2012], and impact processes [Knauth et al., 2005].
 Each of the proposed scenarios for the formation of sulfate deposits appears geologically and chemically feasible for at least some settings, and it seems likely that each of the proposed processes has played a role in formation of some of the diverse sulfate-rich deposits observed around the planet, just as sulfate minerals on Earth arise from different processes in different locations. Interpretation of the geologic process responsible for the formation of specific deposits on Mars will require the development of criteria to differentiate among possible formation mechanisms. Accordingly, a detailed understanding of the geochemical and mineralogical processes leading to the deposition of sulfate minerals in a variety of settings is needed in order to differentiate among possible origins for Martian sulfate deposits. Since many of the proposed mechanisms involve liquid water, understanding the origin of specific sulfate-rich deposits will have substantial implications for assessing the habitability and astrobiological potential of the sites where they occur [e.g., Squyres and Knoll, 2005; Tosca et al., 2008a].
 In an effort to constrain mineralogical and chemical alteration pathways in volcanic fumarole environments for application to Mars, we have initiated a study of acid-sulfate alteration of basalt at Cerro Negro volcano, Nicaragua (CN) [Hynek et al., 2011]. At this site, recently erupted basaltic cinders are actively undergoing alteration by SO2-bearing, steam-rich volcanic vapors in fumarolic environments. As part of this study, we present here the results of laboratory experiments of acid-sulfate alteration of CN basalt conducted to aid in the interpretation of alteration pathways at this and other fumarolic sites. Also presented in this report are numerical geochemical models developed to facilitate interpretation of the experimental results, as well as a discussion of the implications of the experimental results for hydrothermal acid-sulfate alteration on Mars. A detailed description of the mineralogical and chemical composition of the acid-sulfate altered deposits at Cerro Negro, using the results of the present study to interpret reaction pathways, will be the subject of a subsequent contribution.
2 Experimental and Analytical Methods
 Laboratory experiments were performed by reacting fresh cinders of CN basalt with 1.0 M sulfuric acid (H2SO4) at 145°C in stainless steel reactors lined with Teflon inserts (Parr® acid-digestion vessels). A series of eight experiments were conducted for this study, as summarized in Table 1. In order to investigate the impact of reaction parameters on the characteristics of secondary minerals, the experiments encompassed a range of heating durations, fluid : rock ratios (F:R), and grain size. This initial set of experiments was meant only to explore the impact of these various factors on alteration mineralogy, and more comprehensive and systematic studies will be required to confirm the trends observed.
Table 1. Summary of Reaction Parameters for Laboratory Experiments
 The basalt cinders used in the experiments were collected from a recent eruption at Cerro Negro (1995) on a part of the volcano where there was no fumarolic activity. The cinders appeared fresh, with no evidence of alteration or weathering. The basalt is composed of phenocrysts of plagioclase, augite, olivine, and opaque minerals embedded within a glassy matrix (Figures 1a and 1b). The phenocrysts have a bimodal size distribution, with one population ranging in size from ~300 µm to >4 mm, and the other much smaller with diameters mostly 50 µm or less. Plagioclase is, by far, the most abundant phenocryst phase present. We did not undertake a quantitative assessment of the relative proportions of the basalt components, but La Femina et al.  estimated that basalt cinders of the 1999 eruption were composed of 26% plagioclase, 9% augite, 5% olivine, 0.4% opaques, and ~59% glassy groundmass with microcrystals. The bulk chemical composition of the basalt cinders used in this study is listed in Table 2 along with electron microprobe (EM) analyses of glass, plagioclase, and augite in fresh cinders. The plagioclase in the experimental materials is anorthitic (~An92) and the augite corresponds to a composition of ~ En42Wo37Fs21. Olivine is relatively Fe-enriched, with compositions of ~ Fo75. Relative to Martian basalts, bulk CN cinders are enriched in Al and depleted in Fe and Ti (Table 2) [Lodders, 1998; McSween et al., 2006, 2008]. Typical plagioclase compositions in Martian basalts are less calcic (~An60) than the phenocrysts at CN, but average Fe contents of olivine and augite are comparable [e.g., McSween et al., 1996, 2006; Koeppen and Hamilton, 2008; Papike et al., 2009].
aComposition determined by X-ray fluorescence by Activation Laboratories, Inc.
bComposition of basaltic shergottite QUE 94021 from Lodders (1998).
cGlass and mineral compositions measured by EM analysis. Representative composition for glass inclusion is from location shown in Figure 19.
 The experiments were performed at 145°C to simulate near-surface fumarolic environments, and because it was thought that this temperature would allow reactions to proceed at a sufficiently rapid rate that they could be studied at laboratory time scales. Nevertheless, this choice of this temperature is somewhat arbitrary. At Cerro Negro and other fumarolic environments, acid-sulfate alteration occurs over a broad range of temperatures. In areas of vigorous active gas discharge, temperatures are around 100°C at the surface but increase to several hundred °C within a few centimeters beneath the surface. On the fringes of the main discharge site where gas flows are less vigorous, surface temperatures range down to ambient air temperatures. Even in these areas, however, gas discharge may well have been more vigorous, and temperatures higher, in the past than they are at present.
 In most cases, cinders with diameters between 0.3 and 1 cm were used in the experiments (Figure 1), but in one experiment (ADSU8) several large cinders were hand-ground in a porcelain mortar and pestle, and then wet sieved to obtain a 53–212 µm grain size fraction. The solid reactants were placed in the bottom of the Teflon reaction vessel and then a solution of sulfuric acid solution was added, completely immersing the solids. In one experiment (ADSU11), hydrochloric acid (HCl) was included in addition to H2SO4 in the starting solution. During sample loading, no effort was made to purge air from the reactor, so that the reactor headspace contained air at the start of the reaction, consistent with exposure of CN deposits to air during alteration. In most cases, there was a nominal loss of mass in the reaction vessel during the experiments (0.3–0.6 g), indicating the reaction vessels remained sealed during the experiments (Table 1). However, in two experiments (ADSU6 and ADSU8), there was a more significant loss of mass (3.4 and 1.4 g, respectively), which likely occurred as water vapor escaped through the vessel seal.
 Following reaction, the vessel was cooled on the lab bench and then opened. An aliquot of the reacted fluid was immediately removed by plastic pipette and transferred unfiltered to HDPE plastic bottles for later analysis. Preliminary attempts to remove fine particulates from the fluid by filtration resulted in precipitation of solids on the filter, so this procedure was not employed for subsequent samples. The solid products were then separated from the remaining fluid and allowed to dry at room temperature in plastic weigh boats. After air drying in the lab for about 24 h, the solids were stored in sealed glass bottles at ambient laboratory conditions until analyzed. Fluid aliquots were analyzed for major element chemistry by Inductively Coupled Plasma-Atomic Emission Spectroscopy (ICP-AES). Total sulfate concentration in the fluid at the end of one experiment (ADSU7) was analyzed with a Hach portable spectrometer using a kit supplied by the manufacturer. For a couple of experiments, aliquots of the fluid were evaporated to dryness and the resulting precipitates collected for analysis.
 Solid reaction products were routinely characterized by Scanning Electron Microscopy (SEM) equipped with Electron-dispersive X-Ray Spectroscopy (EDS). For these analyses, reacted solids were mounted on aluminum stubs using double-sided carbon tape and analyzed uncoated. A portion of the solid products from several experiments were also embedded in epoxy and polished to expose a cross-section of the cinders for inspection by SEM and EDS. The SEM analyses were conducted on a JEOL 6480LV in either secondary electron or backscattered electron mode, with an accelerating voltage of 15 kV and spot size of 60. Most of the EDS analyses were performed using a Noran System 6, but this instrument was replaced during our study, and later analyses were conducted using an Oxford Instruments collector and processed using INCA software.
 Additional compositional data for selected samples were obtained with electron microprobe (EM) analysis using a JEOL Superprobe JXA 8900 R at the University of Kiel (Germany), equipped with five wavelength-dispersive spectrometers. Minerals and the glassy rock matrix were analyzed with an accelerating voltage of 15 kV. Glass and most minerals were analyzed using a 5 µm beam and a beam current of 10 nA. Owing to the small grain size, a fully focused 1 µm beam was used for measurements of alunite-jarosite group minerals, but to limit beam damage a beam current of only 10 nA was used. Both synthetic and natural mineral and glass standards were used, including barite as a standard for sulfur in sulfate minerals. Raw data were corrected using the CITZAF method [Armstrong, 1995]. Selected samples were also analyzed by X-ray diffraction (XRD) with Co Kα radiation using a Terra instrument (inXitu Incorporated, Campbell, CA).
 To provide additional constraints on Fe-bearing reaction products, selected samples were analyzed by Mössbauer spectroscopy and magnetic susceptibility analysis. Mössbauer spectra were measured at room temperature using a conventional constant-acceleration spectrometer in transmission geometry with a 57Co/Rh source. An α-Fe foil at room temperature was used to calibrate isomer shifts and velocity scale. The magnetic hyperfine parameters including isomer shift (IS) and quadrupole splitting (QS) were fit using the NORMOS program [Brand, 1987], which assumes a distribution of hyperfine parameters during spectral fitting. Magnetic hysteresis properties at room temperature were measured using a vibrating sample magnetometer (Princeton Measurements Corporation) with a maximum applied field 1T. Hysteresis parameters (saturation magnetization, Ms; saturation remanence, Mr; coercivity, Bc) were determined after the loops were corrected for high-field slopes using the approach to saturation fitting described in Jackson and Solheid . High-temperature susceptibility measurements were made using a variable temperature AC Susceptometer (KLY-2 KappaBridge). Susceptibility was measured during heating-cooling cycles up to 700°C in flowing argon to reduce the effects of oxidation. Curie temperatures were determined from the derivative of the susceptibility-temperature curves.
 For comparison with reaction products, alunite-jarosite group minerals with compositions intermediate between natroalunite and natrojarosite were synthesized using methods adapted from Brophy et al. . Briefly, the minerals were synthesized by combining Na-, Al- and ferric Fe-sulfate salts in varying proportions and heating them in the presence of 0.1 M sulfuric acid at 145°C for 2–6 days in the same Teflon-lined reactors used for the acid-sulfate alteration experiments. The proportions of starting materials used in the syntheses are summarized in Table 3. Following reaction, the minerals were removed from the reactor, rinsed with ethanol, dried in a fume hood, and then stored in glass vials at ambient room conditions until analyzed. The chemical compositions of the resulting minerals were assessed by both EDS and wet chemical methods (Table 3). For the wet chemical analysis, the minerals were completely dissolved in a combination of HCl and HNO3 and resulting elements quantified by ICP-AES.
Table 3. Parameters for Synthesis of Natroalunite-Natrojarosite Solid Solutions, With Analysis of Their Chemical Compositions by EDS and Wet Chemical Methods
bFinal pH measured at 25°C; since the values listed were outside the calibrated range for our pH meter (2–10), these values should be regarded as approximations. Fe# = 100 × Fe/(Fe + Al).
 Numerical geochemical models to evaluate fluid speciation and interpret reaction pathways were performed with the computer program EQ3/6 [Wolery and Jarek, 2003]. The thermodynamic databases currently supplied with EQ3/6 (as well as other similar programs) only include data for a limited number of sulfate minerals, and many of these have data only for 25°C which restricts model simulations to low temperature environments. Therefore, in order to broaden the capability to evaluate acid-sulfate alteration in hydrothermal environments, a customized database was constructed for this study that expands the number of sulfate minerals considered and incorporates parameters to allow calculations at elevated temperatures. A description of the database is provided as an appendix, and an electronic version is available by request from the corresponding author. Owing to limited availability of thermodynamic data, the database encompasses only a portion of all known sulfate minerals. Nevertheless, the expanded database includes the primary sulfate minerals that were found in the experiments and in field studies at Cerro Negro (T. M. McCollom et al., manuscript in preparation), as well as the major sulfate phases that have been identified in spectroscopic studies of martian sulfate deposits, including kieserite, gypsum, szomolnokite, epsomite, and members of the alunite-jarosite group [e.g., Klingelhöfer et al., 2004; Bibring et al., 2005, 2006; Gendrin et al, 2005; Milliken et al., 2008; Murchie et al., 2007; Bishop et al., 2009].
3 Experimental Results
 The following subsections describe the solid reaction products generated in the experiments as well as the composition of fluids obtained at termination of the experiments. With only a couple of exceptions, most of the experiments yielded virtually identical solid products, so results are described collectively. However, results from two experiments that yielded notably different reaction products, ADSU8 and ADSU12, are discussed separately. An overview of the solid reaction products found in the experiments is provided in Table 4.
Table 4. Overview of Secondary Products Observed in Experimentsa
Product phases observed
aTentatively identified phases followed by question mark.
Si-rich gel, alunogen?, kieserite?, unidentified Mg- and Al-sulfates
Unidentified Mg-, Al-, Fe- and Ca-sulfates
3.1 Overview of Solid Reaction Products
 Following reaction, the solids were typically enshrouded in an opaque, gray-white to light-yellow gel (Figure 1). When dried, this gel formed a white coating on the surface of the solids. Because this coating interfered with analysis of other alteration products on the cinder surfaces, a portion of the reacted solids from most experiments was washed with a stream of ethanol soon after opening the reaction vessel to remove the gel. Except where noted, the images of cinders shown here are from these rinsed samples. Samples of the gel were also dried and retained separately for analysis. Beneath the gel, the cinders remained cohesive and largely intact (e.g., Figures 1c and 1d), although they were somewhat more friable than the original cinders. In several experiments, white- to light yellow-toned crusts were also found among the reaction products and lining reactor walls.
 Inspection of the reacted cinders by SEM/EDS revealed formation of several new phases. Typical overviews of the surfaces of the reacted cinders are shown in Figure 2. By far, the most abundant crystalline phases observed following the reaction were Ca-sulfates, identified as anhydrite, and minerals of the alunite-jarosite group, both of which were widespread in all of the experiments. The latter had variable chemical composition but, for the most part, corresponded to Fe-rich natroalunite. The only other crystalline products to be found with regularity were spheroidal deposits that were determined to by EDS to be composed predominantly of Fe and O, indicating they were iron-oxide or oxyhydroxide minerals (hereinafter referred to as Fe-oxides/oxyhydroxides, or FeOx). Fan-shaped deposits were widely observed on the altered surfaces of the basalt cinders, but these appeared to be dried gel based on their appearance and close compositional resemblance to the gel phase (e.g., Figure 2d). Several other apparently crystalline phases were observed sporadically in the reacted samples, but these occurred in only trace amounts. In many cases, these rare phases appeared to be similar to phases observed in samples of dried gel, suggesting they precipitated during drying, either from crystallization of the gel itself or by evaporation of fluid contained within the gel. Notably, phyllosilicates were never detected in any of the samples by SEM or other methods.
 Inspection of the cinders also revealed that while plagioclase, augite, and olivine had each reacted extensively during the experiments, the glass showed no chemical or physical evidence of reaction. Remnants of igneous phenocrysts were only occasionally observed on the surfaces of the reacted cinders, and many phenocrysts within the interiors of the cinders were also partially to completely eroded and replaced by secondary minerals (Figure 3). However, other phenocrysts within the interiors remained pristine, apparently because they were protected by surrounding glass. In contrast, no etching or other evidence of reaction was observed in the glass where it was exposed at the surface of cinders, and boundaries of the glass with other phases remained sharply defined. The composition of the glass phase in the reacted cinders was essentially identical to that of the fresh basalt (Table 2), even where measured immediately adjacent to exterior surfaces or to phenocrysts that had been completely replaced by amorphous silica. In a couple of instances, we measured the chemical compositions of glass inclusions within relict plagioclase phenocrysts that had been completely replaced by amorphous silica, and even these inclusions maintained their original composition (Table 2).
 No crystalline silica (SiO2) phase was found during examination of the reacted cinders by SEM. However, amorphous deposits composed predominantly of Si and O were found in all of the experiments (Figures 2 and 3), and Si and O were the major components of the opaque gel that was ubiquitous on the cinder surfaces. Amorphous silica deposits were especially apparent in the interiors of cinders, where this phase was widely observed replacing igneous phenocrysts (Figures 2g, 2h and 3).
 Analysis of the bulk reacted cinders by XRD provided only limited information on secondary products since the diffraction patterns were dominated by remnant plagioclase and other igneous phases. Therefore, we focused on analysis of dried gel and crusts formed in the experiments to identify secondary products (Figure 4). For most samples, the only peaks in the diffractograms that could be confidently identified were anhydrite and natroalunite (Figure 4). Alunogen was identified in a couple of samples, but this phase appeared to originate from drying of the gel phase following the experiments. In many instances, additional peaks were present in the diffractograms that could not be confidently assigned to any phase in our database (Figure 4). No diffraction lines for crystalline SiO2 phases are evident during XRD analysis of the reacted cinders, although there appears to be a vaguely defined broad hump centered at 2Θ ≈ 25° in many of the diffractograms that may be attributable to hydrated amorphous silica.
3.2 Characteristics of Solid Reaction Products
 Two morphotypes of Ca-sulfate crystals were observed in the experimental products. The first type was prismatic crystals up to >250 µm in length, often with striations along the elongated axis (Figures 2b and 2d). Ca-sulfates of this morphotype were widely disseminated across the surface of altered cinders, either as individual crystals or as small clusters. Based on XRD and EDS analyses, this phase was identified as anhydrite. The second morphotype occurred as smaller (≪1 to ~20 µm), blocky crystals (Figure 5). This morphotype was found in dense aggregates that were usually observed encrusting the surfaces of grains that appeared to be relict phenocrysts (Figure 5). In polished grain mounts, blocky Ca-sulfate was observed in close association with relict plagioclase, indicating that the blocky form of the mineral formed during alteration of plagioclase. Since XRD analysis provided no evidence for gypsum or other Ca sulfates, we infer that this morphotype is also anhydrite. Elements other than Ca, S, and O were not observed during EDS analyses of either morphotype, so these appear to be pure minerals.
 Minerals from the alunite-jarosite group were predominantly found in the reaction products as small (<1–20 µm), pseudocubic crystals (Figure 6). Most occurrences of this phase could be classified into one of three categories: (1) individual or small clumps of crystals disseminated across the surfaces of reacted basalt cinders (Figures 2, 6a–6c), (2) large clusters of individual crystals on the surfaces of cinders (Figure 6d), or (3) dense aggregates forming white- to light mustard yellow-colored monomineralic crusts (Figures 6e and 6f). Representative chemical compositions of this phase as determined by EDS from the various experiments are shown in Figure 7 and listed in Table 5. Members of the alunite-jarosite mineral group have the idealized general formula AB3(SO4)2(OH)6, where the A site is most commonly occupied by the monovalent ions K+, Na+, and H3O+ (hydronium) while the B site is occupied primarily by Al3+ and Fe3+ [Stoffregen et al., 2000]. Members of this group with Al > Fe are classified in the alunite subgroup, while those with Fe > Al are in the jarosite subgroup. Natural samples exhibit considerable solid solution mixing in the A site, but occupation of the B site has been reported to be limited to compositions near the Al or Fe end-members [Dutziac and Jambor, 2000; Stoffregen et al., 2000, and references therein]. By convention, measured compositions of minerals in this group are normalized to molecular formula by assuming two SO4 per formula unit (p.f.u.) [Stoffregen et al., 2000], and that convention has been followed for this study.
Table 5. Representative Compositions of Alunite-Jarosite Group Minerals in the Experiments as Determined by EDS
“b.d.” = below detection (<0.01 atom%).
aAtoms per formula unit calculated assuming 2 S per formula unit.
 As shown in Figure 7, the alunite-jarosite group minerals precipitated in the experiments contained variable amounts of both Al and Fe in the B site, with Al > Fe in most samples. The A site in all of the analyzed experimental products was dominated by Na, with little K and no Ca detected during EDS analyses (Table 5). In most cases, a deficiency in Na relative to the ideal formula for this mineral group suggests that hydronium occupied a substantial portion of the A sites. Thus, the compositions of this phase in most experiments correspond to natroalunite, although the minerals in experiment ADSU7, as well as some in experiment ADSU10, were sufficiently enriched in Fe that they would be classified as natrojarosite (i.e., Fe > Al; Figure 7). For brevity, however, we will refer to this phase as natroalunite for the remainder of this report.
 As evident in Figure 7, the natroalunites exhibited substantial variability in the relative amounts of Al and Fe occupying the B site, both within and between experiments. A summary of the range and average values for occupation of the B site, reported as the relative proportion of the Fe-bearing end-member (Fe# = 100 × Fe/[Al + Fe]), is provided in Table 6. While there is some overlap in the distribution of compositions for individual experiments, there is a considerable range in average compositions among the experiments. Minerals from experiments ADSU7 and ADSU10 are especially enriched in Fe while those formed in experiments ADSU4, ADSU8, and ADSU12 contain only limited amounts of Fe, with ADSU5 and ADSU6 having intermediate compositions. Despite these compositional differences, the crystal habits and distributions of this phase were the same in all of the experiments. Considerable compositional variability was also observed within individual natroalunite crystals, which often displayed zoning when viewed in cross-section (Figures 3 and 8). An illustrative example is shown in Figure 8, which shows variability in Al, Fe, Na, and K across several relatively large natroalunite crystals occupying a gas vesicle. The cores and rims of the crystals are enriched in Fe and K, while Al is relatively enriched in the interim.
Table 6. Relative Fe Contents of Natroalunite/Natrojarosite Reaction Products Measured by EDS and EMPA
 Because mineral compositions measured by EDS are generally regarded as semi-quantitative, we performed additional analyses to further evaluate the composition of the natroalunites. The small grain size and restricted distribution of the minerals (i.e., primarily confined to outer cinder surfaces) made measurements of compositions in polished cross-sections problematic. However, the composition of natroalunite was measured by EM analysis for experiment ADSU6, with representative results shown in Figure 7 and Table 7. The EM results show Fe# values ranging from 29 to 45, in good agreement with the relative proportions of Fe and Al measured by EDS (Figure 7 and Table 6). Although the EM analyses yielded a somewhat narrower range of compositions, the EM data represent a much smaller proportion of the reacted cinders than the EDS analyses. The sums of Al plus Fe determined by EM, however, were substantially greater than those measured by EDS and were close to the value of 3 expected for the ideal stoichiometry of alunite-jarosite group minerals (Figure 7), indicating that the EDS measurements systematically underestimated the amount of Al and Fe relative to S. In addition, the sum of Na + K p.f.u. in the minerals as measured by EM is less than one, indicating that the A site is partially occupied by H3O+ to achieve charge balance. However, volatilization of Na during EM analyses (Figure 8b) may account for some of the low total. The EM data also suggest that the relative proportions of Na, K, and H3O+ correlate with the Fe content, with Na increasing while K and H3O+ decrease with increasing amount of Fe (Figure 9).
Table 7. Representative Electron Microprobe Analyses for Natroalunite and Larger FeOx Spheroids (Data From Experiment ADSU6)
Atoms per formula unit (based on 14 O)
Na + K
Al + Fe
 To further assess the relative accuracy of the compositions determined by EDS, we synthesized a suite of minerals with variable Fe contents intermediate between natroalunite and natrojarosite, and then measured their compositions by both EDS and wet chemical methods (Table 3). The synthetic minerals closely resembled the natroalunite produced in the basalt alteration experiments in both size and crystal habit (Figure 10). The relative proportions of Fe and Al in the synthetic minerals determined by EDS agreed closely with the results of wet chemical methods, with values of Fe# agreeing to within 10% (Table 3). Conversely, in many cases the elemental ratios of Na, Fe, and Al to S determined by EDS are lower than those determined by wet chemistry, by up to 40% for Na and ~20% for Fe and Al. Similar trends are apparent the EM analyses for natroalunite in experiment ADSU6 (Figure 7), indicating that the EDS measurements accurately reflect the relative proportions of Fe and Al in the natroalunites (i.e., Fe#), but underestimate the ratios of cations to sulfur in molecular formulas by up to 40%.
 Comparison with XRD patterns for the synthetic natroalunite-natrojarosite solid solution series provides additional support for the accuracy of the relative proportions of Fe and Al measured by EDS (Figure 11). The diffraction patterns of the synthetic crystals show a systematic variation with Fe# in the minerals, corresponding to changes in the unit cell structure [Stoffregen et al., 2000]. For example, Figure 11a shows a series of homologous diffraction peaks occurring in the interval between 28° and 31° that decrease in diffraction angle and intensity with increasing Fe content. Figure 11b shows the corresponding peaks for natrolaunite from two experiments, where it can be seen that the diffraction angles for the reaction products align with synthetic minerals with similar Fe# as determined by EDS.
 Two distinctive groups of Fe-oxide/oxyhydroxide phases were present in the experiments. Both groups exhibited a predominance of Fe and O in EDS measurements, but the two groups displayed distinct differences in their crystal habit. One group consisted of submicron-sized, spherical- to oblate-shaped particles, observed primarily in dense patches on the surfaces of reacted cinders (Figures 12a, 12b, and 12g). These patches were found sparsely distributed in most experiments, but appeared to be more common in experiments ADSU8 and ADSU 12. The other group of FeOx consisted of larger spheroids (typically ~2–5 µm diameter) (Figures 12c–12i). This phase was observed both as individual spheroids and in clusters with botryoidal morphology. In many cases, these larger spheroids were found closely associated with natraolunite, often growing perched on top of pseudocubic crystals from that group (Figures 12c, 12f–12h). The larger spheroids were observed in all experiments, but appeared to be particularly abundant in the longest running experiment, ADSU6, as well as in experiments ADSU8 and ADSU12. Both types of FeOx deposits could be found in most of the experiments. The contrast in size and morphology of the two FeOx phases can be seen clearly in Figure 12g, where examples of both morphotypes occur adjacent to one another. Elemental analysis of the larger spheroids by EM and EDS indicated they were composed predominantly of Fe and O, along with several atom% Si and small amounts of Al, Ti, and S (Table 7). However, owing to the relatively small size of the spheroids, it is possible that some of these other elements may represent underlying or adjacent materials, and spheroids viewed in cross-section often exhibited dark inclusions (e.g., Figure 12i) suggesting additional phases enclosed within the spheroids may be responsible some elements other than Fe.
 On the surfaces of reacted cinders, dried gel formed amorphous blocky materials or fan-shaped deposits, and similar structures were observed in dried gel separated from the reacted cinders (Figures 2 and 13). The chemical composition of dried gel samples as measured by EDS consisted predominantly of Si and O, along with minor and highly variable amounts of Al, Mg, and S. Since the gel likely contained significant amounts of fluid when removed from the experiments, some of latter elements were likely derived from evaporation of liquid trapped within the gel. We infer that the gel itself was largely composed of Si and O, possibly with a small amount of Al, while most of the Mg and S as well as some of the Al were left behind following evaporation of fluid contained within the gel.
 A number of crystalline sulfate phases were apparent in samples of dried gel separated from the reacted cinders (Figure 13). Although prismatic anhydrite crystals and pseudocubic natroalunite were frequently observed embedded within dried gel, these minerals most likely precipitated during the experiments rather than during drying of the gel. However, other phases, including several Mg- and Al-bearing sulfates, appeared to have precipitated during evaporation of fluids trapped within the gel (e.g., Figures 13b and 13d). Similar phases were also observed in solids precipitated during evaporation of the ethanol used to rinse the gel from reacted cinders (Figures 13e–13h). Based on morphology and elemental composition from EDS analysis, we tentatively identify these phases to include pickeringite (Figures 13d and 13g), alunogen (Figure 13h), and kieserite (Figure 13e). Since these phases were only very rarely observed on the surfaces of cinders after removal of the gel, we infer that they precipitated during evaporation of the fluid phase.
 Although no crystalline silica phase was found in any of the experimental products, many amorphous silica-rich deposits were observed on the surface of reacted cinders that were visibly and compositionally distinct from dried gel (Figure 14). Analysis of the deposits by EDS indicated they were composed primarily of Si and O, although small amounts of Al were sometimes present. These silica-rich deposits were frequently found closely associated with blocky Ca-sulfates on the surfaces of cinders (Figure 14). This association suggests that this is a secondary phase replacing plagioclase and other igneous phenocrysts.
 A variety of other Mg-, Al-, Fe-, and Na-bearing sulfate phases were sporadically observed on the surfaces of reacted cinders, a few examples of which are shown in Figure 15. Although the identities of some of these phases can be inferred, we did not make a concerted effort to definitively characterize these minerals since they occurred in only trace amounts and were not observed in all samples. One Na- and Al-bearing sulfate phase with platy crystal habit was observed in small amounts in several samples (Figure 15a and 15b). Based on ratios of 1:1:2 for Na, Al, and S, respectively, as well as a small diffraction peak at 2Θ = 20.8° in a couple of the experimental products where it was relatively more abundant, we tentatively identified this phase as tamarugite. Based on elemental composition and morphology, other phases inferred to be present in trace amounts include kieserite and alunogen (Figure 15). Note that, although these sporadic phases were found on the surfaces of reacted cinders, it is unclear whether they formed during heating of the experiments, precipitated during cooling, or were deposited during evaporation of residual fluids left on the cinders or trapped within the gel.
3.3 Mössbauer and Magnetic Susceptibility Analyses
 To further characterize Fe-bearing phases, Mössbauer (MB) analysis was performed on the solid reaction products of three experiments (ADSU5, ADSU6, and ADSU10). The resulting spectra are shown in Figure 16, with derived MB parameters listed in Table 8. The results indicate that a large proportion of the Fe in the solids from each of the experiments remained in remnant igneous phenocrysts (augite and olivine). However, Fe-bearing secondary phases were also evident in the spectra. Most prominent of these is a phase with fitting parameters that corresponds closely with reported parameters for minerals in the jarosite subgroup (Figure 17) [e.g., Morris et al., 2006]. This signal was found even in the experiments that appeared to contain only Fe-bearing natrolaunite (ADSU5, ADSU6), suggesting that the Fe in the natroalunite was responsible for this signal. Since we were unable to find any literature reports of Mössbauer parameters for Fe-bearing natroalunite to confirm this, we obtained spectra for several synthetic Fe-bearing natroalunites with Fe# ranging from 10 to 25 (Figure 16 and Table 8). The resulting spectra and fitting parameters closely resemble those of the phase observed in the experiments, as well as published values of fitting parameters for natural and synthetic jarosite subgroup minerals (Figure 17). The Mössbauer results indicate that Fe-bearing natroalunite accounts for 20–30% of the Fe in the reaction products (Table 8). Since FeO comprises about 10 wt% of the cinders (Table 2), the MB results indicate that natroalunite comprises up to ~3 wt% of the reacted cinders.
Table 8. Room Temperature Magnetic Hyperfine Mössbauer Parameters for Experimental Products and Synthetic Fe-Bearing Natroalunite
Bhf—magnetic hyperfine field, QS—quadrupole splitting, IS—isomer shift and % relative proportion. Errors are quoted in parenthesis.
 The Mössbauer spectrum for experiment ADSU6 also displayed a low-intensity sextet with broad peaks that was difficult to fit. We tentatively identified the phase as hematite (Table 8), but maghemite would also provide an alternative fit with similar χ2 statistics. This experiment appeared to have a relatively large abundance of the larger FeOx spheroids compared to other experiments, indicating that this phase may be the source of the ferric oxide signal in the MB spectrum. A Mössbauer signal corresponding to hematite/maghemite was not clearly evident in the spectra for the other samples, even though the larger FeOx spheroids were observed among the reaction products for these experiments. However, it may be that the spheroids were not sufficiently abundant in these other experiments to produce a detectable signal during Mössbauer analysis, or that the particular particles selected for analysis may have lower abundance of this phase than the samples examined by SEM. In ADSU6, the Mössbauer data indicate that the ferric oxide phase accounts for about 10% of the Fe in the solid reaction products, suggesting hematite may comprise around 1 wt% of the altered cinders in this experiment.
 The iron oxide mineralogy was characterized further by measuring the temperature dependence of magnetic susceptibility. Thermomagnetic analysis on ADSU5, ADSU6, and ADSU10 all show irreversible behavior during heating-cooling cycles indicating mineralogical changes on heating above 400°C. Two Curie Temperatures (Tc) were observed on heating at 350° C and 450°C indicating two different magnetic phases. There was no indication in the thermomagnetic curves for pure magnetite (Tc = 580°C), hematite (Tc = 680°C), or goethite (Tc = 120°C). The Tc at 350°C is sharp and well defined in all three samples with the heating-cooling curves reversible when heating only to 400°C. This magnetic phase likely corresponds to primary titanomagnetite from the original basalt, and the Curie temperature indicates a titanomagnetite composition (Fe3-xTixO4) with x ~ 0.35 [Lattard et al., 2006], which would be consistent with EDS measurements of primary titanomagnetite that indicate a molar Ti:Fe ratio ~1:9. The higher Tc is also indicative of titanomagnetite, but with a more Fe-enriched composition (x ≈ 0.2). This phase is likely a secondary product formed during the course of the experiment. The relative proportion of the two magnetic phases in each sample vary with the duration of reaction, with the higher Tc phase increasing from 32% to 68% of the total susceptibility as the reaction time progressed from 7 days (ADSU10) to 129 days (ADSU6). Total titanomagnetite concentration based on Ms, including both phases, is less than 1% by weight.
 On cooling from 700°C, the Tc at 350°C is partially recovered but the higher Tc has shifted to ~500°C. In addition, ADSU5 and ADSU6 showed a reduction in susceptibility associated with the two magnetic phases, while ADSU10 showed an increase in susceptibility for the higher Tc phase. All three samples also showed a paramagnetic-like increase in susceptibility (~1/T) from 300°C until room temperature where susceptibility values return to within 10% of their preheated values. The changes on cooling result from mineralogical alteration during exposure to high temperatures during the susceptibility experiment. The high Tc phase seems to be most unstable during heating, and the increase in Curie temperature indicates a change in composition. The appearance of a paramagnetic phase also suggests that an additional Fe-bearing phase has been produced on cooling. Sextets corresponding to titanomagnetite were not detected during MB analysis for any of these samples, but these phases occurred in such low concentrations (<1 wt%) that that they would have been below the detection limits for Mössbauer spectroscopy (~1% of total Fe). Thermomagnetic analysis did not detect any clear evidence for hematite or maghemite, but some of the irreversibility between the heating and cooling curves could potentially be attributable to the inversion of strongly magnetic maghemite to weakly magnetic hematite at temperatures >300°C.
3.4 Cross-Sections of Reacted Cinders
 Reacted cinders from several experiments were examined in cross-section in polished grain mounts to provide additional constraints on alteration pathways (Figures 3 and 18). Although remnant primary igneous minerals were only very rarely observed on the surface of reacted cinders, inspection of cross-sections indicated that remnant crystalline igneous phases were widespread in cinder interiors. While some larger phenocrysts were extensively to completely replaced by alteration minerals, others exhibit only a relatively thin alteration rim. Furthermore, although some smaller phenocrysts are completely altered, others show little or no evidence of alteration. Often, pristine phenocrysts occur immediately adjacent to other phenocrysts that are completely replaced by amorphous silica (e.g., Figures 3b and 3d). Pheoncrysts that are unaltered generally appear to be completely enclosed by glass, but others that appear to be similarly enclosed in the cross-sections were extensively altered. In contrast to the extensive alteration of igneous phenocrysts, the glass phase showed no visual or chemical evidence of alteration in cross-sections, and glass melt inclusions within altered phenocrysts showed no evidence of reaction or dissolution, even when completely surrounded by secondary replacement phases (e.g., Figures 18f and 18g). The chemical composition of the glass remained essentially identical to that of pristine CN basalt, even when measured immediately adjacent to exterior surfaces or completely surrounded by altered phenocrysts (Table 2).
 Figure 18 shows several typical examples of partially altered and replaced phenocrysts. Both plagioclase and augite were replaced predominantly by amorphous silica and anhydrite, with natroalunite and other phases only rarely observed among the replacement products (Figures 18d–18i). Although it is difficult to distinguish in SEM images, EDS analysis indicated that the amorphous Si-rich deposits replacing some large plagioclase phenocrysts comprised two separate phases, one composed primarily of Si and O, with small amounts of Al and S, while the other is composed of variable amounts of Si, Al, Na, S, and O (Figure 18h). Neither of these phases contained more than a trace of Ca. Amorphous silica replacing augite and olivine is composed predominantly of Si and O, with only small amounts of Al and S, and no detectable Ca, Mg, Na, or Fe.
 In partially replaced phenocrysts of both plagioclase and augite, amorphous silica was found immediately adjacent to the interface with the remnant igneous phase, with anhydrite filling cracks and voids created by volume loss during replacement (Figures 18d–18i). Alternating layers of amorphous silica and anhydrite are often interleaved (Figures 3c and 18d). In larger plagioclase phenocrysts, alteration resulted in a network of fractures that are occupied primarily by anhydrite and an apparently amorphous Na- and Al-bearing sulfate phase (Figures 18g and 18h). Anhydrite crystals were also found in dense clusters on the periphery of larger altered plagioclase phenocrysts exposed at cinder surfaces (Figure 18i), which apparently correspond to the blocky form of anhydrite observed on the external surfaces of reacted cinders (Figure 5).
 Figure 19 compares the chemical composition of the Si-bearing phases formed during alteration of plagioclase with that of pristine plagioclase in a partially altered phenocryst. One phase, represented by points 2 and 3, are depleted in Si and Ca relative to plagioclase and enriched in S, Na, and Fe. The other phase, represented by points 4 and 5, is highly enriched in Si and depleted in all other elements. In more extensively altered phenocrysts, the first phase is absent, and the Si-rich phase is the primary replacement phase that remains.
 Curiously, while natroalunite was only sporadically observed replacing plagioclase and augite phenocrysts, it was observed in several instances as a major phase replacing larger olivine phenocrysts, where it was found intergrown with amorphous silica (Figures 18b, 18c, 18j–18l). Thus, natroalunite was found replacing the only primary igneous silicate mineral that contained neither Al nor Na. A good example is shown in Figures 15b and 15c, where elemental maps show correlated enrichments of Na, Al, and S in areas depleted in Si and Mg, corresponding to areas where natroalunite has precipitated during replacement of a partially altered olivine phenocryst. Note that even though it is replacing an Fe-rich mineral (olivine), the natroalunite in this instance has low Fe levels. Similarly, Figures 15j–15l show extensive intergrowth of natroalunite and amorphous silica replacing what appears to have been a large olivine phenocryst.
 In cross-section, most occurrences of natroalunite and FeOx were found deposited on exterior surfaces of the cinders, or lining gas vesicles in the cinder interiors (Figures 3 and 18). As was observed during examination of reacted cinder surfaces, larger spheroidal FeOx was often found closely associated with natroalunite (e.g., Figures 3d and 8). However, dense aggregations of spheroids were also observed that had no evident association with natroalunite.
3.5 Experiments ADSU8 and ADSU12
 Except for the variations in chemical compositions for natroalunites noted above, the secondary phases formed during acid-sulfate alteration were very similar in most of the experiments. However, the products of two experiments, ADSU8 and ADSU12, were sufficiently different to warrant separate description. In contrast to the other experiments which were performed with centimeter-sized basalt cinders, ADSU8 was conducted with cinders that had been pulverized and sieved to the 53–212 µm size fraction. The intent of this treatment was to investigate how grain size and increased surface area might impact alteration products and reaction pathways. Experiment ADSU12 contained the lowest initial fluid : rock ratio of the experiments (F:R ≈ 1.1) (Table 1).
 Experiment ADSU8 produced reaction products that were, for the most part, similar to the other experiments but, as might be expected for the smaller grain size and increased exposure of mineral surface area, the overall amount of products appeared to be significantly greater. This experiment had particularly extensive development of natroalunite and FeOx, including both the submicron and larger spheroidal morphologies (Figure 20). Most surfaces of fractured cinders from this experiment were covered with a layer of natroalunite or FeOx (Figure 20a). Coincident with the relatively extensive development of FeOx, the natroalunite from this experiment had among the lowest Fe contents of all experiments (Figure 7). In contrast to other experiments, several partially corroded phenocrysts were observed in this experiment, which were not protected within glass (e.g., Figure 20d). In addition, whereas the surfaces of relict phenocrysts in other experiments were generally coated with secondary phases (e.g., Figure 5), the surfaces of these phenocrysts were largely free of alteration phases, indicating that dissolved elements were re-deposited elsewhere in the system rather than in the immediate vicinity of the dissolving phase.
 Experiment ADSU12 produced several significant differences from the other experiments in terms of secondary products. When this experiment was opened after 42 days of reaction, no separate fluid phase remained in the reactor, although the solids glistened when removed from the reactor indicating a thin coating of liquid remained. The solids at the top of the reaction products were coated with dark, rusty red-colored deposits not observed in other experiments (Figure 21a). Solids lower in the reaction vessel had a visual appearance similar to the other experiments, with the solids enshrouded in a coating of a translucent, light gray gel.
 Examination of the red-colored deposits by SEM/EDS revealed that the cinder surfaces were covered by widespread spheroidal FeOx deposits, interspersed with natroalunite, amorphous silica, and anhydrite (Figures 21b–21d). The morphology of the FeOx deposits resembled the larger spheroids observed in other experiments (Figure 12), but accumulations of this phase were much more dense and extensively distributed in experiment ADSU12, particularly in the top layer of reacted cinders. A smaller but significant fraction of Fe was also present in another secondary phase primarily composed of Fe, Si, and O, with an Fe:Si ratio of ~2. This phase occurred as fibrous plates interlocked into spherical rosettes found both disbursed among other phases (Figures 21c and 21e) and in monomineralic crusts (Figure 21f). Under reflected light, the phase appeared to have a brick red coloration.
 Another major difference in the secondary reaction products from ADSU12 was a fairly widespread occurrence of magnesium sulfates, which were only sporadically observed in other experiments. The Mg sulfates were found in a variety of morphologies, including globular deposits that sometimes had a hexagonal profile (Figure 21g), massive deposits that morphed into columnar structures (Figure 21h), and short, thick fibrous structures (Figure 21i). Analysis of these phases by EDS indicated that Mg was the only cation present.
 Dense accumulations of natroalunite were present throughout this experiment, both covering the surfaces of reacted cinders (Figure 6d) and forming dense monomineralic crusts (Figures 21d and 21e). The natroalunite had similar morphology to that observed in other experiments, but contained one of the lowest Fe contents of all experiments (Figure 7). Other major secondary phases in this experiment included anhydrite and amorphous silica, similar in morphology and relative abundance to other experiments. Unlike experiments other than ADSU8, a number of partially decomposed olivine, augite, and plagioclase were observed among the ADSU12 products, including some crystals >0.5 mm in diameter (e.g., Figure 21g).
 Analysis of minerals from the top layer of the reaction products from ADSU12 by XRD produced peaks attributable to hexahydrite, anhydrite, and natroalunite (Figure 4), with the hexahydrite signal evidently corresponding to one or more of the Mg sulfate morphologies observed in the reacted cinders. Peaks in the XRD pattern at 2Θ = 35.7° and 33.2° appear to be attributable to hematite, although an expected peak at ~54° for this phase does not appear. Although not definitive, this indicates that the large FeOx spheroids produced in this and other experiments are probably hematite. The XRD pattern also included a prominent peak at 2Θ = 27.8° that may correspond to the Fe-Si-rich phase, but this peak could not be associated with any phase in our diffraction library.
3.6 Fluid Compositions
 The compositions of fluids sampled at termination of the experiments are summarized in Table 9. It should be borne in mind that these compositions represent quenched samples that were not filtered, so it is possible that they may have been altered by reactions during cooling of the experiments to room temperature or by dissolution and precipitation of solids during storage. There was not enough fluid remaining at termination of experiment ADSU12 to collect for measurement, so this experiment is not included in the table.
Table 9. Fluid Compositions at Termination of Experiments
cCalculated in situ pH at 145°C (see text for calculation methods). b.d. = below detection limit (~0.01 mmol/kg). “-” = not determined. ΣSO4 is total dissolved sulfate.
 With the exception of experiments performed with powdered basalt (ADSU8), the fluid compositions were very similar among the different experiments despite differences in duration of reaction, fluid:rock ratio, and initial fluid composition (i.e., inclusion of HCl in ADSU11). The fluids are dominated by dissolved Al and Mg, with the relative proportions of dissolved elements decreasing according to the sequence Al > Mg ≫ Fe > Si ≥ Ca > Mn > K. The only element that showed much variation among the experiments (excluding ADSU8) is Na, with the range of measured Na concentrations spanning more than three orders of magnitude (0.015–22 mmol/kg). Despite the extensive reaction of the fluid with the basalt cinders to form secondary products, the final pH remained strongly acidic in each of the experiments. While measured room-temperature pH values for the experiments were in the range of 1.0–1.6, in situ pH values at experimental conditions were likely to have been somewhat higher, with estimated values ranging from 2.0 to 2.6 (Table 9) (see below for a description of estimation methods). The experiment performed with finely powdered basalt (ADSU8) had a similar composition to the other experiments, except that the concentration of dissolved Al was lower by several orders of magnitude, and the final pH was slightly higher.
 For a couple of experiments, aliquots of the fluid were separated from the solids and evaporated to dryness at room temperature. Several apparently crystalline phases were found to precipitate during evaporation of the fluid, including Mg-, Al-, Fe- and Ca-sulfates (Figure 22). Most of these phases included only a single major cation, but one mixed Al-Mg sulfate was observed (Figure 22b). While natroalunite was found among the precipitates from the dried fluids (e.g., Figure 22a), it is not clear whether this phase represents minerals formed during evaporation or were precipitated during the experiments and inadvertently removed with the fluid sample. However, the compositions of natroalunite in the dried fluid samples were compositionally similar to those found among the solid reaction products, suggesting that they formed during the experiments. Owing to limited amounts of material available, we did not analyze these products by XRD or make further efforts to identify the sulfate phases precipitated from the evaporated fluids.
 Numerical models were used to evaluate the distribution of aqueous species and saturation states of minerals in the fluids. These models also allowed the in situ pH of the experiments to be estimated. The calculations were performed with the program EQ3, which uses thermodynamic data together with the measured elemental composition of the fluid to calculate the equilibrium distribution of aqueous species and saturation state of minerals [Wolery and Jarek, 2003; see also Bethke, 2008]. For the calculations, the fluids were first speciated at 25°C using the elemental composition and room temperature pH measured for each experiment, with charge balance achieved by adjusting the total sulfate concentration. In the second step, the fluid was respeciated at the experimental temperature of 145°C using the calculated sulfate concentration at 25°C and adjusting the pH to achieve charge balance. In the absence of actual measurements, the oxygen fugacity was set to 10−10 for these calculations to reflect a likely depletion of O2 within the reaction vessels during the reactions.
 An example of the output of the fluid speciation and mineral saturation calculations is shown in Table A2 for experiment ADSU5. All of the other experimental fluids showed similar trends, and so are not shown. Listed in the table is the saturation index (SI), defined as SI = log (Q/K) with Q the reaction quotient and K the equilibrium constant. Minerals with SI between −0.4 and 0.4 are considered to be saturated within the accuracy of the thermodynamic data, and those with SI > 0.4 are supersaturated with respect to the fluid.
 The estimated in situ pH values for the experiments calculated using this procedure are listed in Table 9. In all experiments, the calculated in situ pH values were nearly a unit higher than the pH measured at room temperature. In large part, this is attributable to dissociation of HSO4−, which is more stable at elevated temperatures and is calculated to be the most abundant species in the fluids, and would be expected to partially dissociate to H+ and SO42− during cooling.
 At room temperature, the fluids were found to be saturated with respect to gypsum and supersaturated with respect to several phases including quartz, amorphous silica, hematite, nontronite, and the Fe-sulfates jarosite and ferricopiapite, but undersaturated with respect to natroalunite. At experimental temperature, the fluid was saturated with respect to amorphous silica, gypsum, and kieserite, but supersaturated with respect to a number of phases including anhydrite, natroalunite, Fe-oxides and oxyhydroxides, and a number of Si- and Al-bearing minerals including several phyllosilicates. Saturation with respect to amorphous silica is consistent with its occurrence in the experimental products, but the calculated supersaturation with respect to anhydrite and natroalunite is inconsistent with their presence among the reaction products. Possible explanations for this discrepancy include partial dissolution of these minerals during cooling, inaccurate estimation of in situ pH, and inaccuracies in thermodynamic data. The supersaturation of phyllolicates and other Si-, Al-, and Fe-bearing phases indicates that precipitation of these phases was kinetically inhibited during the experiments.
4 Reaction Path Models
 In order to provide a framework for interpretation of the experimental results, numerical geochemical models were constructed to examine acid-sulfate alteration of Cerro Negro basalt under the experimental conditions. The models were constructed to simulate the reaction of sulfuric acid with Cerro Negro basalt using the computer program EQ6, version 8.0 [Wolery and Jarek, 2003; see also Bethke, 2008]. EQ6 is computationally equivalent to other computer programs that have been used to study weathering and the origin of sulfate-rich deposits on Mars, including Geochemist's Workbench (GWB), GEOCHEQ, and CHESS [e.g., Elwood Madden et al., 2004; Schiffman et al., 2006; McAdam et al., 2008; Berger et al., 2009]. However, EQ6 was employed for this study because, unlike other programs, it allows solid solutions to be considered for mineral phases, which is essential for examination of compositional variations in minerals like those of the alunite-jarosite group.
 The model calculations were restricted to the major rock-forming elements in the system Na-K-Mg-Ca-Fe-Al-Si-O-S-H. Based on data from La Femina et al. , Cerro Negro basalt was assumed to be composed of four separate reactive components consisting of 26.0 wt% plagioclase, 9.3 wt% augite, 5.3 wt% olivine, and 59.4 wt% glass. The glass composition for the models was estimated by subtracting the individual mineral components from the measured bulk composition of the basalt used in the experiments, and assuming a molar Fe3+/Fetotal of 0.07. For consistency with experimental conditions, calculations were performed for a temperature of 145°C at steam saturation pressure, and for a water : rock mass ratio of 4.
 The model calculations are performed by starting with a 1 M H2SO4 solution and incrementally adding the basalt components in appropriate proportions. As the basaltic components dissolve, the composition of the solution evolves, and minerals precipitate out of solution as they become saturated. Some precipitates may re-dissolve into solution as conditions change, so that minerals can appear only as transients as the reaction progresses. The outcome of the models is a prediction of the minerals that should precipitate during acid-sulfate alteration as a function of the amount of rock reacted. For some models, dissolution or precipitation of phases was suppressed in order to produce results that were more consistent with the experimental observations. A summary of the assumptions in the five models presented here is given in Table 10.
Table 10. Summary of Input Constraints for Numerical Models
Minerals precluded from precipitation
Olivine, augite, plagioclase, glass
Olivine, augite, plagioclase
Olivine, augite, plagioclase
Si-bearing minerals (except amorphous silica)
Olivine, augite, plagioclase, O2
Si- and Al-bearing minerals (except amorphous silica and Al-sulfates)
Olivine, augite, plagioclase, O2
Si- and Al-bearing minerals (except amorphous silica and Al-sulfates), hematite, goethite
 An initial baseline model (Model #1) was constructed with the assumption that all basalt components dissolve at the same rate and that alteration products instantaneously precipitate to equilibrium with the fluid. That is, it is assumed that there are no kinetic inhibitions to mineral dissolution or precipitation so that complete thermodynamic equilibrium is maintained throughout the reaction. Results of this model are shown in Figures 23a and 23b. In the early stages of reaction, the model predicts that the product minerals will be composed of quartz, natroalunite, anhydrite, and kaolinite, with a trace amount of hematite. As more of the basalt dissolves, the quartz and kaolinite are replaced by a mixture of phyllosilicates including monmorillonite and nontronite, with lesser amounts of magnesium beidellite, chabazite, and pyrite. The amount of natroalunite in the alteration assemblage also decreases as the fraction of basalt reacted increases and as Na and Al are incorporated into other phases. Among the solid solutions, nontronite and montmorillonite are dominated (>97 mol%) by the Mg end-member, while the predicted composition of natroalunite contained a considerable fraction of the alunite component that increased with the extent of reaction, with compositions ranging from about (Na0.8K0.2)Al3(SO4)2(OH)6 to (Na0.6K0.4)Al3(SO4)2(OH)6. For complete reaction, the final alteration mineral composition is predicted to include, in decreasing abundance, monmorillonite, nontronite, anhydrite, K-bearing natroalunite, chabazite, beidellite, and trace hematite.
 During reaction with basalt in Model #1, the fluid pH gradually evolved from strongly acidic to nearly neutral (neutral pH at 145°C is ~5.9) (Figure 23b). As reactions progress, more than half of the sulfate is predicted to precipitate out of solution into mineral phases. In the early stages of reaction where acidic conditions prevail, Al, Fe, Mg, and Ca are all predicted to be present in the fluid at elevated concentrations of >50 mm. With increasing extent of reaction and increasing pH, Mg continues to increase to above 200 mm, while Al and Fe decrease to concentrations below 1 mm. The concentration of Na also increases to more than 100 mm with increasing extent of reaction as natroalunite partially dissolves in favor of other Na- and sulfate-bearing phases. Si and K are present at relatively low levels (<3 mm) throughout the reaction.
 Several discrepancies are immediately apparent between the predictions of Model #1 and the outcome of the experiments. First, although the model presumes that each of the basalt components dissolve at the same rate, the crystalline igneous phases in the experiments reacted extensively while there was no evidence for any reaction of the glass, indicating that glass dissolution at the experimental conditions was much slower than that of the crystalline phases. Second, quartz, phyllosilicates, and other crystalline silicate minerals were not observed in the experiments. Instead, Si released from dissolution of the reactant minerals primarily formed an amorphous gel, indicating that precipitation of crystalline SiO2 and other silicate-bearing phases was kinetically inhibited on the time scale of the experiments. Third, although natroalunite was predicted as a major alteration product in the model, the predicted composition of this phase included more K and much less Fe and H3O+ than was observed in the experiments. Fourth, in the baseline model, nearly all of the Fe remained in the ferrous state, but the formation of Fe-bearing natroalunite in the experiments requires oxidation of Fe since this element occurs as FeIII in natroalunite. Fifth, while the final fluid in the models was only mildly acidic, the fluid in most experiments remained strongly acidic (pH < 2) after reaction.
 In order to investigate the impact of various kinetic factors on alteration pathways, a series of additional, non-equilibrium models were generated. Since there was no evidence for reaction of glass in the experiments, the first of these models (Model #2) assumed that glass dissolution did not occur on the time scale of the experiments, while plagioclase, olivine, and augite are all allowed to dissolve. Although it would be ultimately desirable to include relative reaction rates for dissolution of these phases, relative rates under the experimental conditions could not be easily extracted from the experimental results, and it was assumed that these phases dissolved at the same rate. All other parameters remained as in Model #1. Results of this model are shown in Figures 23c and 23d. Mineral alteration products predicted for this model include quartz, anhydrite, kaolinite, natroalunite, and trace amounts of nontronite and hematite. The predicted composition of natroalunite was essentially the pure end-member [NaAl3(SO4)2(OH)6] throughout the reaction, while nontronite was a mixture of the H- and Mg-bearing end-members (~65–85 mol% H, 15–35 mol% Mg). In contrast to the baseline model, the predicted fluid composition remained strongly acidic (final pH = 1.5), with high (>50 mm) concentrations of Al, Mg, and Fe through most of the reaction (Figure 23d). The concentration of Ca was also high when the fraction of rock reacted was low, but it decreased with more extensive reaction. Predicted concentrations of Si and Na remained relatively low throughout the reaction.
 Since Model #2 still produced crystalline phases not observed in the experiments, in the third model (Model #3) it was assumed that, in addition to the prohibition of glass dissolution, precipitation of all Si-bearing minerals was kinetically inhibited on the time scale of the experiments, leaving amorphous silica as the only SiO2-bearing phase as a proxy for the Si-rich gel that formed in the experiments. The results of this calculation are shown in Figures 23e and 23f. Predicted solid reaction products included amorphous silica, anhydrite, diaspore, and natroalunite, with a trace amount of hematite. Again, the predicted composition of natroalunite was essentially the pure end-member. The predicted fluid composition closely resembled that of Model #2, with the pH remaining acidic and high (>50 mm) concentrations of Al, Mg, and Fe throughout most of the reaction.
 These two models (#2 and #3) exhibit two remaining major discrepancies with the experimental results. First, the predicted natroalunite composition in both models contains virtually no Fe. Furthermore, while almost all of the Fe in the models remained in the ferrous state, incorporation into natroalunite requires oxidation to the ferric state. Second, the predicted mineral assemblage includes diaspore, but no equivalent crystalline Al oxide or oxyhydroxide was observed in the experiments. The first discrepancy is an indication that significant iron oxidation occurred during the experiments that had not been accounted for in the models. It is likely that this oxidation came about through reaction of FeII from the rocks with O2 included in the reactor headspace and initial fluid, or diffused through the walls of the reactor. Alternatively, reaction of FeII from the rocks may be oxidized by water to produce H2, which then diffused out of the reaction vessel. The second discrepancy may be attributable to a kinetic inhibition of precipitation of Al-bearing phases other than Al sulfates, or it may be indicative of inaccuracies in the thermodynamic properties of Al species at low pH which are relatively poorly constrained [e.g., Tagirov and Schott, 2001].
 Accordingly, the fourth model (Model #4) followed the same assumptions as Model #3, but included O2 as an additional reactant and precluded precipitation of Al-bearing minerals in addition to the Si-bearing minerals (other than amorphous silica). For the calculation, an amount of O2 estimated to be contained in the reactor headspace was included as an additional reactant (=45 nmol O2/g rock). Results of Model #4 are shown in Figures 23g and 23h. Predicted mineral products for this model included amorphous silica, anhydrite, natroalunite, and hematite. As in the previous models, the natroalunite was essentially the pure end-member composition throughout the reaction. The predicted fluid composition was fairly similar to the previous models, except that, with the inhibition to Al-bearing minerals (other than Al Sulfates), the concentration of Al rose throughout the reaction to a final value of 370 mmol/kg. The amount of O2 added was only sufficient to oxidize about 60% of the total Fe in the system, with almost all of the ferric Fe precipitating as hematite while the remaining ferrous Fe was entirely dissolved in the fluid.
 Although some Fe oxides/oxyhydroxides were observed in the experiments, these were relatively rare. Additionally, the predicted precipitation of hematite in the models apparently precluded partitioning of Fe into natroalunite, as observed in the experiments. Therefore, a final model (Model #5) was carried out following the same set of assumptions as in Model #4, but with the additional constraint that precipitation of hematite, goethite, and magnetite was suppressed. Results of this model are shown in Figures 23i and 23j. Predicted secondary minerals for this model included amorphous silica, anhydrite, and natroalunite, in addition to the unreacted glass. In contrast to previous models, the natroalunite in this model contained significant amounts of Fe and H3O+, with predicted compositions of about (Na0.6H3O+0.4)(Al1.8Fe1.2)(SO4)2(OH)6 throughout the reaction. The fluid composition closely resembled that of Model #4, with concentrations of dissolved Al and Mg becoming highly elevated as reactions progressed. The pH of the fluid becomes moderately acidic with increasing extent of reaction, with a final predicted in situ pH of 3.2.
5.1 Alteration Reactions and Secondary Mineralogy
 During the experiments, acid-sulfate alteration of the basalt cinders resulted in rapid decomposition of plagioclase, augite, and olivine that were exposed to acid and formation of secondary products dominated by sulfate minerals and a Si-rich gel phase. Conversely, the glass phase showed no evidence of alteration on the time scale of the experiments, and crystalline igneous phases protected within the glass were likewise unaltered. Decomposition of the exposed igneous phases was extensive, even in the shortest duration experiment (7 days; ADSU10), indicating decomposition rates of igneous silicates were very rapid under the experimental conditions wherever the minerals were in communication with the fluid.
 Despite differences in duration of heating and initial fluid : rock ratio (Table 2), the same limited suite of secondary phases dominated the reaction products in all of the experiments (i.e., amorphous silica, anhydrite, natroalunite, and FeOx). Although we did not undertake quantification of the reaction products, the relative proportions of amorphous silica, anhydrite, and natroalunite appeared to be roughly the same in most experiments. However, the relative abundance of the larger FeOx spheroids appeared to be higher in the longest running experiment (ADSU6), the experiment with small grain size (ADSU8), and the experiment at the lowest fluid : rock ratio (ADSU12). The addition of HCl to the initial fluid in experiment ADSU11 did not appear to affect the outcome of the experiment, since the products were the same as those generated in other experiments that did not include HCl.
 In contrast to the uniformity in the suite of secondary minerals formed during alteration, there was substantial variation in the composition of the natroalunites among the different experiments (Figure 7 and Table 6). As shown in Figure 24a, there is a general trend of increasing Fe content of the natroalunites with increasing fluid : rock ratio, suggesting that this variable may have been a factor in regulating their composition. This trend may be attributable to a larger fraction of the Al released from dissolution of plagioclase and augite remaining in solution at higher fluid : rock ratios rather than precipitating as natroalunite, thereby decreasing the amount of Al relative to Fe in the precipitating natroalunite. The Fe content of the natroalunites did not have any clear dependence on duration of heating (Figure 24b), suggesting that the compositions may have remained more or less constant as reactions progressed.
 Two additional phases were observed in ADSU12 that were not found in any other experiments: hexahydrite and an unidentified Fe-Si-rich phase. In ADSU12, essentially all of the water was taken up by precipitation of hydrated minerals during the experiment, leaving only a thin coating of liquid on the solids. As a consequence, while the Mg released during decomposition of olivine and augite remained in solution in the other experiments, the lack of water in ADSU12 led to precipitation of the Mg in hexahydrite (and possibly other Mg-sulfate phases; Figures 21g–21i). The limited availability of water probably also contributed to the formation of the Fe-Si-rich phase, as well as the greater development of FeOx, but the mechanisms responsible are not immediately apparent.
 The numerical models indicate that congruent dissolution of the basalt and attainment of stable thermodynamic equilibrium among the reaction products would have resulted in an alteration mineral assemblage dominated by phyllosilicates, accompanied by anhydrite, Fe-poor natroalunite, and trace amounts of hematite (Figure 23a). The lack of phyllosilicates among the reactions products indicates that the experiments remained far from equilibrium despite a significant degree of reaction. Two factors appear to contribute to the absence of phyllosilicates and greater proportion of sulfate minerals relative to the expected equilibrium assemblage. First, there was clearly a strong kinetic inhibition to dissolution of the glass under the experimental conditions, and no evidence that leaching of the glass made any contribution to alteration reactions. Even without the contribution of elements from the glass phase, the results of Model #2 indicate that Si- and Al-bearing phases such as kaolinite and dickite would still have been expected to precipitate during alteration from the elements supplied by the dissolving phenocrysts (Figure 23c), which is supported by the supersaturation of the fluids with respect to phyllosilicates and other phases (e.g., Table A2). Therefore, a second requirement to explain the absence of phyllosilicates is a kinetic inhibition to precipitation of crystalline aluminosilicate phases. Additional kinetic inhibitions to precipitation of Al- and Fe-oxides/oxyhydroxides are also required to account for the scarcity of minerals such as diaspore and goethite and to allow for the presence of high Fe contents in alunite-jarosite group minerals. Schiffman et al.  similarly found that suppression of nontronite, hematite, and goethite precipitation was required to model the formation of amorphous sililca-jarosite rock coatings produced during acid-sulfate alteration of basalt in Kilauea volcano. However, their model assumes that dissolution of glass is the source of elements for the formation of secondary products and precipitation of phyllosilicates other than nontronite is allowed in the later stages of reaction, so their overall model produces significantly different predictions than our preferred model (#5; Figure 23i).
 When all of these factors are taken into account, the reaction path model (#5) provides results that are in close agreement with the reaction products. The model also predicts a fluid composition with absolute and relative elemental abundances that are comparable to those observed in the experiments. Furthermore, although we did not determine the quantities of reaction products, the relative amounts of solid products produced in the experiments appeared to be consistent with the proportions predicted in the model, especially if the gel phase is considered to be a component of the amorphous silica abundance.
 The persistence of glass in the experiments would appear to be in disagreement with current experimentally derived models for the relative dissolution rates of basaltic components under strongly acidic conditions, which predict that glass should dissolve faster than olivine, plagioclase, and pyroxene at pH below ~2 [e.g., Brantley, 2008; Hausrath et al., 2008]. On the other hand, field-based studies of hydrothermal alteration of andesites by acid-sulfate-chloride fluids indicate that andesitic glass dissolves more slowly than pyroxene and plagioclase under natural conditions [Rowe and Brantley, 1993; Africano and Bernard, 2000]. Since glass dissolution rates have been found to increase exponentially with decreasing SiO2 content [e.g., Wolff-Boenisch et al., 2004], basaltic glass might be expected to react more rapidly than andesitic glass. Yet, the results of this study indicate that the glass still reacted more slowly than the crystalline phases. Glass dissolution has also been found to decrease with increasing Al concentrations under acidic conditions [Oelkers and Gislason, 2001], and the highly elevated concentrations of Al in the experiments may have contributed to inhibition of glass reactivity. Furthermore, experimental dissolution studies for glass and other phases have mostly been studied at much lower temperatures than the present experiments, and relative dissolution rates may change with increasing temperature. Although our data do not allow for a rigorous quantitative evaluation of reaction kinetics, it is apparent that reaction of primary igneous silicates exposed to acid-sulfate fluids proceeded very rapidly at the temperature of the experiments, achieving extensive to complete reaction of exposed minerals on time scales of weeks to a few months. On the other hand, the glass phase will evidently require much longer time scales to react under the same conditions, and igneous phenocrysts protected by the glass would also be expected to persist for much longer times.
 Reaction of the each of the primary silicate phases with the acidic fluid resulted in extensive mobilization of cations, eventually leaving behind a residue composed predominantly of amorphous SiO2 (Figures 14 and 18). In the case of plagioclase, Ca is the first element to become depleted, followed by Na and Al (Figure 19). This trend is similar to the stepwise depletion of Ca, Na, and Al reported for alteration of plagioclase phenocrysts in natural fumarolic environments [e.g., Papike, 1992; Spilde et al., 1993; Africano and Bernard, 2000]. In the natural samples, however, the silica-rich amorphous phase that replaces plagioclase has been observed to be enriched in Cl and F as well as S, leading to suggestions that hydrochloric and hydrofluoric acids played a primary role in the dissolution. Since our experiments show similar features in the absence of these other acids, it appears that the alteration trends are a consequence of the acidic nature of the medium, irrespective of the nature of the acid anion. The intermediate phase in the conversion of plagioclase to amorphous silica includes Fe in addition to S (Figure 19), requiring these elements to be mobilized into the vicinity of the reacting plagioclase. Some of the Ca mobilized during alteration reprecipitates locally as anhydrite in void spaces created by decomposition of the primary plagioclase (Figure 18), or on the surfaces of relict phenocrysts (Figure 5). Conversely, Na and Al were predominantly transported beyond the immediate vicinity of the decomposing plagioclase crystals, and either precipitated as natroalunite elsewhere in the system or remained dissolved in the fluid.
 Alteration of augite also resulted in replacement by amorphous silica and anhydrite (Figures 18e and 18f), although anhydrite was less abundant in deposits replacing augite than for plagioclase. In the process, Mg and Fe were mobilized beyond the boundaries of the original phenocrysts. Whereas the amorphous silica replacing plagioclase seems to have formed by successive depletion of the solid in other cations, there was usually a gap between the eroding surface of augite and replacement products, suggesting that the Si dissolved into solution and then reprecipitated as amorphous silica. Alteration of augite and replacement by amorphous SiO2 in the experiments also resembles that observed in natural samples [Spilde et al., 1993; Africano and Bernard, 2000].
 Olivine phenocrysts also undergo extensive removal of Mg and Fe during alteration and replacement by amorphous SiO2 (Figures 18b and 18j). During olivine replacement, however, the amorphous phase is accompanied in many instances by euhedral natroalunite, requiring transport of Na, Al, and S into the vicinity of the dissolving olivine (e.g., Figures 18b and 18k). It seems counterintuitive that natroalunite should precipitate as a replacement product of olivine but not of plagioclase or augite, since the latter minerals are the primary sources of Na and Al during alteration. Nevertheless, olivine dissolution must produce a local increase in pH or other change in fluid chemistry that induces precipitation of the natroalunite. A particularly puzzling aspect of these precipitates is that, even though olivine dissolution is one of the primary sources of Fe during alteration, the natroalunite replacing olivine has relatively low Fe contents (as reflected, for example, by the dark gray shading for this phase observed in backscattered electron images; Figures 18b and 18k).
 The scarcity of Mg-bearing sulfates in the experiments (except for the experiment performed at the lowest fluid : rock ratio, ADSU12) indicates that essentially all of the Mg derived from dissolution of augite and olivine remained in solution, reflecting the high solubility of Mg-sulfate salts. Provisional mass-balance calculations (not shown) indicated that a large fraction of the Al released from the dissolution of plagioclase and augite also remained in solution, while almost all of the Si, Na, and Ca released from the minerals were removed by precipitation of secondary phases in most experiments. Two exceptions were experiments ADSU7 and ADSU10, which had elevated concentrations of Na and also had the highest measured concentrations of dissolved Fe (Table 9). The solids generated in these experiments appeared to contain somewhat smaller amounts of natroalunite than the other experiments, which may account for a greater fraction of the Na and Fe remaining in solution. Alternatively, the higher Na concentrations could reflect dissolution of small amounts of natroalunite during cooling of the experiments.
 Although the larger FeOx spherules observed in the experimental products could not be definitively identified, this phase appears to be hematite. Mössbauer analysis of experiment ADSU6 suggests the presence of several percent hematite (Figure 16 and Table 8), and diffraction peaks that may correspond to hematite were present in the XRD analysis of the top layer of the products from ADSU12, where the spherules were particularly abundant (Figures 21b and 21d). The spherules also bear a close morphological resemblance to small hematite spherules observed in acid-sulfate altered basaltic tephra deposits on Mauna Kea [Morris et al., 2005] and to those produced during thermal decomposition of hydronium jarosite at 150–200°C in laboratory experiments [Golden et al., 2008]. The common association of the larger spherules with Fe-bearing natroalunite in our experiments (e.g., Figures 8, 12g, and 12h) suggests the possibility that some spherules may have formed through alteration of the underlying natroalunite. Specifically, decomposition of Fe-bearing natroalunite (or natrojarosite) to more Al-rich compositions accompanied by release of Fe may have provided a local source of Fe for precipitation of the spherules, a process that may be similar to the formation of hematite from hydronium jarosite in previous experiments [Golden et al., 2008]. If this is the case, the spherules may reflect an inherent instability of Fe-rich intermediates in the natroalunite-natrojarosite solid solution.
 Since only ferric Fe can occupy the B site of alunite-jarosite group minerals [Stoffregen et al., 2000] and the FeOx produced in the experiments also presumably contains primarily ferric Fe, formation of these phases required extensive oxidation of the Fe dissolved from augite and olivine. This oxidation may have occurred through reaction with O2 from air contained within the initial experimental charge, or with O2 diffused through the walls of the Teflon reaction vessel during heating. Alternatively, the oxidation could have occurred through reaction of ferric Fe with H+ to produce H2, followed by diffusion of H2 through the reactor walls (i.e., 2Fe2+ + 2H+ → 2Fe3+ + H2).
 In sum, our results indicate that the earliest stages of acid-sulfate alteration of pyroclastic basalt in volcanic fumaroles will result in rapid decomposition of the igneous silicates and a secondary mineral assemblage composed predominantly of amorphous silica (or perhaps an equivalent such as opal), anhydrite, and Fe-rich natroalunite, together with minor Fe-oxides/oxyhydroxides and essentially unaltered glass. At lower temperatures, gypsum may replace anhydrite as the Ca-sulfate phase. In environments with more limited amounts of water than were present in our experiments and relatively closed-system conditions, additional Mg-sulfates and other phases may precipitate, similar to those observed in experiment ADSU12. Evaporation of the fluids following reaction, either locally or elsewhere, would result in precipitation of additional Mg-, Al-, and Fe-bearing sulfates. Conversely, during alteration in open-system environments, the high solubilities of Mg- and Al-sulfate salts under strongly acidic conditions, as reflected by the high abundance of Al and Mg in the experimental fluids, should allow preferential removal of these elements during the early stages of alteration. In any event, the results indicate that formation of phyllosilicates is very limited during acid-sulfate alteration, at least during the initial stages of reaction.
 Although a detailed description of the mineralogy and chemical composition of acid-sulfate altered rocks at Cerro Negro is beyond the scope of this report, the primary phases produced in the experiments closely parallel those observed in the natural system [Hynek et al., 2011; McCollom et al., manuscript in preparation]. Alteration of CN basalt cinders by SO2-bearing steam in haloes surrounding active fumaroles results in an alteration assemblage dominated by amorphous silica and gypsum, with minor amounts of alunite-jarosite group minerals and small amounts of Fe-oxides/oxyhydroxides in some samples (Figure 25). Phyllosilcates and crystalline SiO2 phases were not found in samples collected near the fumaroles. The alunite-jarosite group minerals have a similar size and pseudocubic morphology to those produced in the experiments and exhibit a range of composition intermediate between natroalunite and natrojarosite [McCollom et al., 2012]. In the most extensively altered samples at CN, minerals from the alunite-jarostite group and Fe-oxides/oxydroxides are absent and the glass exhibits extensively depletion in all cations other than Si. Consequently, the results of the experiments appear to be most applicable to the earliest stages of alteration at this site, with more extensive leaching of the glass taking place over longer reaction times than was investigated in the experiments.
5.2 Fe-Bearing Natroalunite
 The Fe-rich natroalunites produced in many of the experiments have chemical compositions that are outside of the range of compositions that have been reported for alunite-jarosite group in natural settings. Published compositions of natural minerals within this group lie predominantly near either the Fe or Al end-members, reflecting limited mixing at the B site (Figure 7) [Stoffregen et al., 2000; Papike et al., 2006a]. Reports of members of the jarosite subgroup with Fe# as low as 80 are not unusual, but reports of alunite or natroalunite with an Fe# >5 are very rare [Stoffregen et al., 2000]. In a literature search performed for this study, the most Fe-rich compositions found were from Brophy et al. , who reported a hydrothermal alunite with Fe# equal to 13 from Utah, and Alpers et al. , who reported a single analysis of alunite with an Fe# of ~8 that had precipitated in an acid-saline lake sediment. Alunite subgroup minerals in the literature from hypogene or supergene hydrothermal environments either had no Fe abundance reported or had only trace amounts of Fe present [e.g., Stoffregen and Alpers, 1992; Vasconcelos et al., 2004; Papike et al., 2006a, 2006b]. Scott  reported considerable solid solution mixing of Fe and Al in the B site for some structurally related minerals, but this mixing was only observed when there was substantial (>55%) substitution of PO43− or other trivalent anions in the sulfate site as well as >80% substitution of a divalent cation in the A site. To our knowledge, sulfate-dominated minerals in the alunite-jarosite group with Fe# between 15 and 80 have not previously been reported from natural samples. Nevertheless, we have found minerals from this group with a range of intermediate Fe-Al compositions in recently altered volcanic deposits at Cerro Negro [McCollom et al., 2012, and manuscript in preparation].
 It is not yet clear whether the compositional divide between the alunite and jarosite subgroups observed in natural samples is attributable to environmental or thermodynamic factors [Stoffregen et al., 2000]. In contrast to natural samples, minerals with a full range of Al-Fe compositions intermediate between the alunite and jarosite subgroups are readily synthesized in the lab (i.e., this study, Table 3, as well as Brophy et al.  and Härtig et al. ), demonstrating that formation of minerals with intermediate compositions is at least possible. Stoffregen et al.  suggest that an immiscibility gap between the Fe and Al members of this mineral group appears to be unlikely and attribute the paucity of minerals with intermediate compositions to differences in the hydrolysis constants for dissolved Fe and Al species. Alternatively, Papike et al. [2006b, 2007] suggest that the compositional gap reflects the oxidation state during deposition, with more oxidizing conditions allowing precipitation of ferric Fe-bearing jarosite subgroup minerals, while more reducing conditions, where ferrous Fe predominates, preclude formation of jarosites in favor of members of the alunite subgroup.
 The experimental results suggest alunite-jarosite group minerals with a range of Al and Fe contents may form in the early stages of acid-sulfate alteration of basaltic rocks. The lack of literature reports of minerals with intermediate Al-Fe compositions may indicate that such minerals are unstable over the long term and will decompose in favor of more Al- and Fe-rich compositions over time. Such an inference may be supported by the close spatial association of co-occurring alunite and jarosite subgroup phases in some acid-sulfate altered rocks [e.g., Zimbelman et al., 2005; Papike et al., 2006a, 2006b].
 It is also possible, however, that the intermediate Fe-Al compositions measured in the reaction products may be an artifact of very fine-scale intergrowth of separate Fe- and Al-rich phases that were not resolved by the EDS and EM analyses. Members of this group can exhibit oscillatory compositional zoning at very fine scales (<1 µm), which has been attributed either to a changing chemical environment during precipitation or to exsolution from minerals with mixed composition as the system cooled [e.g., Deyell and Dipple, 2005; Papike et al., 2006a, 2006b, 2007; Desborough et al., 2010]. Compositional zoning was observed in many of the alunite-jarosite group minerals produced in our experiments, although at the scale of the measurements the zones maintained intermediate compositions between the Fe and Al end-members with no evidence of oscillatory zoning between discrete end-members (Figures 3b and 8). Nevertheless, in other studies, minerals from this group whose bulk compositions appear to represent intermediate solid-solution compositions have been found on closer inspection to be composed of very fine-scaled mixtures of end-members [Stoffregen and Alpers, 1992; Desborough et al., 2010].
 In addition to these fine-scale variations, natural samples of alunite-jarosite group minerals commonly exhibit considerable compositional variability among individual crystals within a single locality [e.g., Juliani et al., 2005; Deyell and Dipple, 2005; Papike et al., 2006a, 2006b]. These trends are generally considered to reflect local and temporal variability in ambient fluid compositions during mineral precipitation. The experimental products exhibit similar compositional variability, suggesting a response to spatial and temporal differences in local chemical conditions within the reaction vessel, which may be attributable to changes in relative contributions of elements from dissolution of the igneous phases as the experiments proceeded. Variation in the oxidation state within the reaction vessel as the experiments evolved may have played an additional role, since the ferrous Fe originally present in the igneous minerals must be oxidized to ferric Fe before it can fit into the crystal structure of alunite-jarosite group minerals.
 The paucity of K in the alunite-jarosite group minerals produced in the experiments is a reflection of the lack of K-bearing crystalline phases in the initial rock, and the limited release of K from the glass. Acid-sulfate alteration of igneous rocks containing potassium feldspars under similar conditions might be expected to generate minerals with more significant amounts of the K-bearing end-members (i.e., alunite and jarosite), which may account for the formation of fine-grained psuedocubic alunite rather than natroalunite during reaction of sulfur-bearing gases with andesitic rocks in stratovolcanoes [Zimbelman et al., 2005]. The inferred presence of H3O+ in the structure of the natroalunite formed in the experiments would be consistent with the widespread occurrence of this component in minerals of this group that are formed at moderately low temperatures (≲200°C) [Ripmeester et al., 1986; Stoffregen et al., 2000].
 It has typically been inferred that minerals of the alunite-jarosite group with fine-grained, pseudocubic morphology form in low temperature environments (20–40°C) [Stoffregen and Alpers, 1992; Stoffregen et al., 2000; Deyell and Dipple, 2005]. Nevertheless, minerals in this group that closely resemble the size and morphology of those produced in the experiments are found in acid-sulfate altered basalt deposits at Cerro Negro (Figure 25) as well as other sites of acid-sulfate alteration of volcanic deposits [Morris et al., 1996; Zimbelman et al., 2005; Deyell et al., 2005]. These occurrences, together with the experimental results, demonstrate that fine-grained psuedocubic crystals of this mineral group are not restricted to low temperature environments, but can form at higher temperatures in fumaroles and other hydrothermal settings. Based on S and O isotope systematics, Zimbelman et al.  inferred that some fine-grained pseudocubic alunite in stratovolcanoes may have formed at temperatures as high as 350°C, and Deyell et al.  inferred temperatures up to 140°C for pseudocubic alunite in a hydrothermal deposit from South America.
5.3 Comparisons With Prior Experimental Studies
 Several previous laboratory studies have examined acid-sulfate alteration of basalt in the context of understanding the formation of martian sulfate deposits [Banin et al., 1997; Tosca et al., 2004; Hurowitz et al., 2005; Golden et al., 2005, 2008]. Most of these studies, however, were designed to simulate weathering of Martian rocks under “acid fog” conditions rather than hydrothermal environments and consequently were performed at much lower temperatures than our experiments. This consideration, as well as other differences in experimental design and reaction conditions, makes detailed comparisons of our results with those of previous studies somewhat difficult to interpret. Nevertheless, since comparisons among experimental studies can potentially provide useful insights into how reaction pathways might vary under different conditions and suggest directions for future study, in this section we briefly compare our results with those of prior studies.
Tosca et al.  examined the weathering of synthetic Martian basalt by acidic solutions in laboratory experiments performed at 25°C. In their experiments, crystalline and glassy basalts were reacted with mixed solutions of sulfuric and hydrochloric acids of varying strengths for 14 days, with the residual fluid evaporated from the reaction vessel prior to examination of the solid products. Reaction of the crystalline basalt with strongly acidic solutions, followed by evaporation, resulted in precipitation of amorphous silica along with several Fe-, Mg- and Ca-bearing sulfates that included hexahydrite, rhomboclase, gypsum, anhydrite, and several unidentified phases.
 The suite of secondary products reported by Tosca et al.  resembles the minerals observed in our experiments following evaporation of reacted fluids (Figure 22), with the exception that those authors did not report any natroalunite or other Al-bearing phases among the products, even though plagioclase was a major component of the basalts they used in their experiments. However, the plagioclase in their basalt reactant had substantially lower anorthite contents than those of CN basalt (An17-30 versus An90), and reaction rates of plagioclase have been found to increase with increasing anorthite content [e.g., Casey et al., 1991; Spilde et al., 1993; Africano and Bernard, 2000]. Consequently, the lack of Al-bearing sulfates in the experiments of Tosca et al.  as well as relatively low Al concentrations in the fluids following reaction can potentially be accounted for by limited reaction of plagioclase in their experiments.
 The basalt glass experiments of Tosca et al.  produced results that are strikingly different from our studies. In their experiments, reaction of a low-silica basalt glass (~48.7 wt% SiO2) with a fluid containing 1 M H2SO4 and 0.25 M HCl resulted in complete dissolution of the glass in less than 24 h and formation of a thick gel phase. Solid phases present after evaporation of the reacting fluid included amorphous silica along with a number of Fe-, Mg-, Al-, and Ca-bearing sulfates including hexahydrite, alunogen, tamarugite, gypsum, and anhydrite. The extent of reaction decreased with decreasing concentration of the acids. Nevertheless, reaction of the glass with dilute acid solutions still resulted in extensive leaching of the surfaces of the glass, leaving behind a SiO2-enriched layer up to 10 µm thick. More limited reaction was observed for a glass with a higher silica content (57.9 wt% SiO2), but experiments reacting this glass with high acid concentrations still resulted in some degree of dissolution and formation of an etched, silica-rich rind on the reacted grains and precipitation of small amounts of secondary phases including amorphous silica, gypsum, and some unidentified sulfates.
 In contrast to the results of Tosca et al. , we observed no evidence for alteration of the glass phase in our experiments, which were performed with comparable acid concentrations but much higher temperatures. The reason for this apparent discrepancy is not immediately apparent. Since basaltic glass has been observed to dissolve more slowly with increasing SiO2 content at acidic conditions [e.g., Wolff-Boenisch et al., 2004], it is possible that differences in SiO2 content among the glasses used in the experiments may have contributed to the observed differences in rates. However, the SiO2 content of CN glass (~53 wt%; Table 2) is intermediate between the glass compositions investigated by Tosca et al. , and those authors still observed some reaction with the higher silica glass. In addition, the higher temperatures and longer reaction times employed in our experiments would be expected to produce more extensive reaction, other factors being equal. Thus, it is not clear that differing SiO2 contents can fully explain the differences in glass reaction. Other factors that may have played a role are the higher Fe and lower Al contents of the synthetic glasses used by Tosca et al. , and substantially higher Al contents of the reacting fluid in our experiments, produced by rapid dissolution of plagioclase, which could have interfered with surface reactions involved in glass dissolution [Oelkers and Gislason, 2001].
 In a set of batch experiments, Hurowitz et al.  reacted synthetic crystalline basalt with mixed solutions of sulfuric and hydrochloric acids at 76°C for 14 days. The basalt was synthesized to have a composition similar to that of the basaltic shergottite “Los Angeles” and was composed of 40% plagioclase (~An52), 49% augite, 5% titanomagnetite, 4% other phases, and trace amounts of glass. After reaction, no new phases were identified among the solid reaction products by XRD, but two Fe-rich phases were identified by SEM/EDS on one experiment performed at a high water : rock ratio (=100). Amorphous silica along with several sulfate minerals including anhydrite, alunogen, rhomboclase, and ferrinatrite was identified by XRD analysis of residues left after evaporation of the reacted fluid. This suite of minerals resembles those produced during evaporation of the reacted fluid in our experiments (Figure 22), although Hurowitz et al.  did not report any Mg-sulfates. No alunite-jarosite group minerals were observed by Hurowitz et al. , nor was any anhydrite, gypsum, or amorphous silica identified on the surface of the reacted basalt. However, it is possible that these phases may have been present at levels that were undetectable by XRD.
Golden et al.  performed a series of experiments reacting basalt with sulfuric acid under mild hydrothermal conditions, including experiments intended to simulate vapor phase and open-system reactions. Most directly relevant to the present study were batch experiments in which either olivine-rich or plagioclase-rich basalt were reacted with sulfuric acid at 145°C. Hydrogen peroxide was added to the initial solution to simulate oxidizing conditions at the Martian surface. After 6 days reaction of the olivine-rich basalt, the olivine had completely dissolved, and reaction products included amorphous silica, gypsum, hydronium jarosite, and a hydrated Mg-sulfate. Glass contained in the basalt was etched on the surface, but retained its overall structure. Amorphous silica was the only reported product for reaction of plagioclase-rich basalt under the same conditions.
 The products of olivine-rich basalt batch experiments of Golden et al.  are similar to those observed in our experiments, although gypsum rather than anhydrite was identified as the primary Ca-sulfate phase and hydronium jarosite was produced rather than natroalunite or natrojarosite. The latter can potentially be explained by a larger proportion of olivine relative to plagioclase in their basalt reactant compared to that used in our experiments, or to a more limited reaction of plagioclase. It is worth noting that our experiment with a reaction time most comparable to the batch experiments of Golden et al. (ADSU10) produced some of the most Fe enriched alunite-jarosite group minerals among the experiments (Figure 7). Based on our study, precipitation of a Ca-sulfate phase (gypsum or anhydrite) and natroalunite would also have been expected in the plagioclase-rich basalt experiments of Golden et al. , but only amorphous silica was reported. The apparent absence of these other minerals in their experiments could potentially be accounted for by slower dissolution of plagioclase, which in their basaltic reactant had a low anorthite content (An46) compared to that used in our experiments (An90).
6 Implications for Acid-Sulfate Alteration on Mars
 It is likely that acid-sulfate alteration of basalt in hydrothermal environments has played a prominent role in the geology of the Martian surface. Basaltic volcanism has been widespread on the planet and, given the relatively high sulfur contents of Martian basalts [Lodders, 1998; McSween et al., 2006, 2008], crystallization of Martian lavas must have released significant amounts of sulfur-rich volatiles, particularly SO2. Reaction of volatile sulfur compounds with water in magmatic vapors or with groundwater would inevitably produce acidic fluids that would then go on to interact with the surrounding and overlying basalts, either in vapor or liquid form. At the surface, environments for discharge of sulfur-rich volcanic fluids would include fumaroles, solfataras, and hot springs.
 The experimental results indicate that the earliest stages of acid-sulfate alteration on Mars should result in rapid decomposition of primary igneous silicate minerals, including olivine, pyroxene, and plagioclase, while any glass phase present reacts only slowly or not at all, at least initially. Elements released by alteration would precipitate as sulfate minerals and amorphous silica, with little or no formation of phyllosilicates. Likely early products of the process are Ca-sulfates (anhydrite or gypsum, depending on temperature), alunite-jarosite group minerals, and Fe-oxides/oxyhydroxides. If alteration occurs under circumstances with limited fluid mobility or low water : rock ratios (which seems likely for Mars), additional sulfates will likely precipitate from the fluid, including more soluble Mg-, Fe-, and Al-sulfate salts.
 Of course, there are significant differences in elemental composition between the terrestrial basalt studied here and Martian basalts, and these differences would be expected to effect the relative abundance and chemical composition of products generated during acid-sulfate alteration. In particular, Martian basalts are enriched in Fe and depleted in Al relative to the Cerro Negro basalt used in the experiments (Table 2) [Lodders, 1998; McSween et al., 2006, 2008]. In order to explore the potential consequences of these compositional differences on alteration mineralogy, geochemical models were conducted to investigate acid-sulfate alteration of Martian basalt under conditions similar to those of the experiments.
 The models were performed using the same set of constraints as Model #5 above, except that the composition of the initial rock was modified to approximate Martian basalt. Martian pyroclastic basalt was assumed to comprise 18 wt% plagioclase, 24 wt% pyroxene, 13 wt% olivine, 5 wt% magnetite, and the remainder present as glass. The composition is based on the Barnhill class rocks observed in Gusev Crater by the Mars Exploration Rover Spirit that are thought to represent a pyroclastic basalt whose chemical and mineral composition has been modified to some extent by acid-sulfate alteration [Squyres et al., 2007; Schmidt et al., 2009]. Based on deconvolution of Mini-TES spectra, Schmidt et al.  estimate that the rocks are presently composed of 40% glass, 25% secondary silicates, 10% plagioclase, 10% olivine, 5% pyroxene, and 5% sulfates. Mössbauer analyses indicate that Fe in rocks of the Barnhill class is presently distributed among the following components: olivine (~16%), pyroxene (~21%), nanophase oxides (~28%) (npOx), magnetite (~29%), and hematite (~7%) [Morris et al., 2008]. The relative abundance of Fe-bearing phases in the pristine rocks prior to alteration was estimated by starting with the bulk chemical composition of the Barnhill class rock “Posey Manager” [Schmidt et al., 2009] and distributing the total amount of Fe in the rocks among the minerals identified by Mössbauer. For the purposes of the models, the npOx is presumed to represent a component of devitrified glass rather than an alteration phase, while the hematite is presumed to be a secondary phase not present in the pristine rock. The amount of plagioclase in the pristine rock was calculated with the assumption that it contained 60% of the total Al in the rock, with the remainder in the glass. In the absence of more definitive constraints, primary igneous phases were assumed to have typical Martian compositions, including Fo70 for olivine, En65Wo10Fs25 for pyroxene, and An60 for plagioclase [e.g., McSween et al., 1996, 2006; Koeppen and Hamilton, 2008; Papike et al., 2009]. McSween et al.  calculated normative mineral compositions for Barnhill class rocks, but these values were not used in the models because they do not take into account the presence of a glass phase.
 As for our final kinetic model for CN basalt, the Mars model assumed that glass is inhibited from reaction, and that there are kinetic inhibitions to precipitation of aluminosilicate minerals including phyllosilicates, as well as Al- and Fe-bearing oxides/oxyhydroxides. In addition, magnetite was assumed to be unreactive during alteration. It was further presumed that the alteration took place under oxidizing conditions with sufficient O2 present to oxidize all Fe released by dissolving minerals to Fe(III). Accordingly, the fluid in the models was assumed to maintain equilibrium with O2 in the atmosphere at present levels of 10−5 bar. Many of the sulfate-rich deposits observed on Mars were likely deposited early in the history of the planet, and there are few constraints on the levels of O2 present at that time. However, the sulfate deposits at Meridaini Planum and elsewhere contain abundant ferric Fe-bearing minerals that indicate deposition took place under relatively oxidizing conditions.
 The model predicts that alteration of Martian basalt under conditions similar to the experiments would result in the formation of amorphous silica, anhydrite, natroalunite, kieserite, and, with extensive reaction, small amounts of brucite (Figure 26a and 26b). Through most of the reaction, the natroalunite in the model is predicted to include Fe, with an Fe# ≈ 15. The fluid contains high concentrations of Mg, Fe, and Al, and evaporation of the fluid would result in precipitation of additional Mg-, Fe, and Al-sulfates. As in Model #5 for CN basalt alteration, this model for acid-sulfate alteration of Martian basalt assumes precipitation of Fe oxides and oxyhydroxides is kinetically inhibited. If this assumption is relaxed, the model predicts the addition of hematite along with a trace of pyrite to the alteration assemblage (Figures 26c and 26d). Under this set of assumptions, the composition of natroalunite is predicted to be essentially the pure end-member, as ferric Fe precipitates as hematite instead of substituting for Al in natroalunite. The occurrence of a Mg-sulfate (kieserite) in the predicted secondary mineral assemblage for Martian basalt is largely a consequence of the higher Mg:Ca ratio in the pyroxene relative to the augite used in the CN basalt.
 If these models are representative, the results suggests that the earliest stages of alteration of Martian pyroclastic deposits under acid-sulfate conditions would result in a mineral assemblage that includes amorphous silica (or perhaps a crystalline equivalent such as opal), Ca-sulfate (anhydrite or gypsum), Fe-bearing natroalunite, kieserite, and, possibly, hematite, along with remnant glass. Igneous phases protected within the glass would also be expected to be preserved at this stage. Evaporation of the reactant fluid, or reaction under lower fluid : rock ratios, would result in precipitation of additional Mg-, Al-, and Fe-sulfates. Comparison of the Martian model results with those for Cerro Negro compositions indicates that the principal consequence of the difference in rock composition may be the addition of kieserite or other Mg-sulfates to the alteration mineral assemblage. The only one of our experiments to produce significant amounts of Mg-sulfates was ADSU12, and hexahydrite rather than kieserite was observed, suggesting the possibility that hexahydrite may precipitate in preference to kieserite during acid-sulfate alteration. However, this requires further investigation.
 Of course, our models represent only one set of conditions, and more work will be required to determine how the predicted assemblage might change with variation in parameters such as rock composition, temperature, and fluid : rock ratio. In addition, longer reaction times might dissolve elements from the glass phase that would cause additional changes in the alteration mineralogy. While the models would also be improved by addition of relative reaction rates for the various reacting components, this will require development of relative dissolution rate data for minerals and glass at elevated temperatures. One critical uncertainty in extrapolation of the CN models to Martian conditions is whether the high concentrations of Al resulting from plagioclase dissolution are essential to the inhibition of glass dissolution. If Al is the inhibiting factor for glass dissolution and lower anorthite contents of plagioclase on Mars limit Al concentrations, more extensive glass dissolution may occur and result in a different suite of mineral products. Further experimental work and model development will be needed to address this issue.
Berger et al.  used a somewhat similar set of kinetic constraints to model alteration of Martian rocks by sulfuric acid solutions at 0°C. In their models, dissolution of the crystalline components of Martian basalt, represented by the rock Adirondack from Gusev Crater, were constrained by relative rate laws, such that olivine reacted more rapidly than plagioclase and pyroxene (glass was not considered as a component). Precipitation of chalcedony, goethite, and sulfates was assumed to be instantaneous, while a kinetic rate law was applied to precipitation of clay minerals. The models predict an alteration assemblage that includes chalcedony, jarosite, gypsum, epsomite, and goethite, along with remnant igneous phases. This alteration assemblage is roughly similar to that produced by our Mars model, although there are some differences in the predicted sulfate minerals. Although these differences could arise in part from discrepancies in the thermodynamic data and modeling methods, they suggest the possibility that there may be temperature variations in the sulfate phases that precipitate during acid-sulfate alteration of Martian basalt, which could potentially be used to infer temperatures during formation of the deposits. However, more experimental work coupled to further refinement of numerical models will be required to evaluate this possibility.
 Other efforts to model reaction of Martian basalts with sulfuric acid solutions have not considered kinetic inhibitions to precipitation of saturated phases [e.g., Zolotov and Mironenko, 2007; Tréguier et al., 2008; McAdam et al., 2008]. In models where glass is considered as a component of Martian basalt, it is typically assumed glass would either decompose at comparable rates to other phases or would decompose much faster than primary igneous minerals such as plagioclase, pyroxene, and olivine [e.g., Zolotov and Mironenko, 2007; McAdam et al., 2008; Hausrath et al., 2008]. As a consequence of these assumptions, these models generally predict alteration assemblages dominated by phyllosilicates and fluids that trend to neutral pH as reactions progress. As with our preliminary models, these other models do not reproduce the products observed in the laboratory experiments.
 Kieserite and Ca-sulfates have been inferred to be present in sulfate-rich deposits in many locations on Mars based on observations by both orbiters and landers [e.g., Klingelhöfer et al., 2004; Bibring et al., 2005, 2006; Gendrin et al., 2005; Milliken et al., 2008; Murchie et al., 2007; Bishop et al., 2009]. Natroalunite has apparently not yet been reported to occur in Martian sulfate deposits, although members of the jarosite subgroup have been inferred to be present in some deposits based on space-based spectral observations [Gendrin et al., 2005; Milliken et al., 2008; Murchie et al., 2007; Bishop et al., 2009]. Jarosite group minerals have also been identified in the layered sulfate deposits at the Opportunity landing site on Meridiani Planum based on Mössbauer spectra [Klingelhöfer et al., 2004; Morris et al., 2006].
 The identification of jarosite at Meridiani Planum is considered one of the key findings of the Opportunity rover and has been widely used to infer environmental conditions that were present during formation of the deposits. For instance, based on constraints provided by the stability of jarosite, it has been inferred that conditions at the time of formation of the sulfate-bearing bedrock at Meridiani were mildly to strongly acidic, with pH estimated below ~4.5 [e.g., Elwood Madden et al., 2004; Tosca et al., 2005; Papike et al., 2006a, 2006b]. Jarosite formation also requires moderately to highly oxidizing conditions, since the Fe is present in ferric form [Brown, 1971; Burns, 1987].
 The interpretation that jarosite is present in the sulfate deposits at Meridiani Planum is based largely on analysis of the Mössbauer spectra of the rocks [Klingelhöfer et al., 2004; Morris et al., 2006]. However, the Mössbauer analyses of our experimental products and synthetic minerals indicate that Fe-bearing natroalunite produces a spectral signature that closely resembles that of minerals from the jarosite subgroup (Figures 16 and 17). This signal was observed in natroalunite with Fe contents as low as Fe# = 10 (Natro8). These results indicate that the Fe-bearing members of the alunite subgroup could potentially account for the peaks in the Mössbauer spectra from Meridiani that have previously been ascribed to jarosite, and models for the formation of these deposits should be expanded to include this members of this subgroup.
 Inclusion of Fe-bearing natroalunite or other members of the alunite subgroup among the possible candidates for interpretation of the Mössbauer spectra at Meridiani would considerably expand the range of possible conditions during formation of the deposits. Members of the alunite subgroup can be stable to much higher pH levels (7.5 or higher) [Brown, 1971; Hladky and Slansky, 1981; Burns, 1987], which would allow for less strongly acidic conditions at the site. Fe-bearing natroalunite may also require less oxidizing conditions during formation, since it can precipitate at lower activities of Fe3+ than jarosite and can even precipitate from solutions where Fe2+ is the predominant form of dissolved iron [Brown, 1971]. Moreover, the presence of Fe-bearing members of the alunite subgroup would expand the range of possible scenarios for the formation of deposits to include those where jarosite does not occur, such as the acid-sulfate alteration of basalt that is the subject of this study. Because jarosite is a relatively unstable mineral, its persistence in the Meridiani outcrops would also place constraints on conditions following the initial formation of the deposits [Tosca et al., 2008b; Elwood Madden et al., 2009, 2012]. However, if the phase seen in the Mössbauer spectra is instead an Fe-bearing member of the alunite subgroup, a much broader range of post-depositional conditions would allow their persistence to the present day.
 Another distinctive characteristic of the layered sulfate bedrock at Meridiani Planum is the presence of large hematite spherules embedded within deposits [Squyres et al., 2004; Klingelhöfer et al., 2004]. One proposed pathway for the formation of the hematite spherules involves incongruent dissolution of a jarosite precursor [Morris et al., 2005; Tosca et al., 2008b; Golden et al., 2008; Elwood Madden et al., 2012]. In one such scenario, the precursor jarosite is proposed to be a product of acid-sulfate alteration of basalt [Morris et al., 2005; Golden et al., 2008]. The close association of many of the larger FeOx spherules produced in our experiments with Fe-bearing natroalunites may provide further support for this scenario (Figures 12f–12h). In the experiments, Fe-rich natroalunite may have become unstable as conditions within the reaction vessel evolved, releasing Fe3+ to solution that precipitated as FeOx. In the process, the underlying natroalunite would have evolved to less Fe-enriched compositions. Although the FeOx spherules found in our experiments are much smaller than those that occur in Meridaini Planum, they are similar in size and morphology to hematite spherules observed in acid-sulfate altered basaltic tephra on Earth [Morris et al., 2005] and to spherules generated during thermal decomposition of hydronium jarosite in laboratory experiments [Golden et al., 2008].
 The presently available evidence indicates that temperatures above 70–100°C may be required for the formation of gray hematite with the morphological, chemical, and spectral characteristics of the hematite “blueberries” observed at Meridiani Planum [Catling and Moore, 2003; Golden et al., 2008]. More detailed studies will be required to determine the characteristics of the spheroidal FeOx that formed in our experiments, but it seems likely that they have properties similar to the hematite spherules produced by thermal decomposition of hydronium jarosite [Golden et al., 2008]. The latter have many morphological and spectral similarities to the Martian blueberries, with the exception of a smaller size. Should further research confirm that elevated temperatures are required for the formation of the blueberries, it would be difficult to rationalize their formation within the low-temperature evaporite scenario that has been proposed as the source of the deposits by the MER team [Squyres et al., 2004, 2009; McLennan et al., 2005; Clark et al., 2005; Wang et al., 2006], but may instead point to a hydrothermal origin for the sulfate- and hematite-bearing deposits [McCollom and Hynek, 2005, 2006]
7 Concluding Remarks
 Volcanic fumaroles are highly heterogenous environments where environmental conditions such as temperature, pH, and fluid fluxes can vary substantially with depth and proximity to conduits of volcanic vapors. The experiments described herein encompass only a small portion of the full range of conditions that occur in such environments, and further work is required to determine the degree to which the secondary products we observed are representative of the broader system. Nevertheless, the results should provide a basis for refined models of acid-sulfate alteration of basalts on Mars and Earth and provide new constraints for the interpretation of the origin of sulfate deposits on Mars.
Appendix A: Thermodynamic Data Used in the Numerical Models
 The thermodynamic database used in the numerical models was a modified version of the “data0.ymp.R2” database supplied with the EQ3/6 program, version 8.0. The data0.ymp.R2 database was compiled for evaluation of the Yucca Mountain Nuclear Waste Repository. It was chosen as the core database for this study because it includes a broad range of aqueous species and solid phases, and because it has been extensively vetted for data quality and internal consistency as documented by Wolery and Jove-Colon, 2004. The data0.ymp.R2 database includes a broader range of mineral salts and low temperature silicates than previous databases, such as the data0.com database that was provided with earlier versions of EQ3/6. Data0.com also forms the core of the thermo.com.v8.r6 + .dat database supplied with recent versions of Geochemist's Workbench.
 Despite the expanded number of minerals included in the data0.ymp.R2 database, coverage of sulfate salts remains very limited. Therefore, the database was modified to include additional sulfate minerals needed to study fluid-rock interactions in acid-sulfate environments. Although a large number of sulfate minerals are known to occur in geologic systems [see, for example, Hawthorne et al., 2000], thermodynamic data are presently available for only a small number of these. As a consequence, inclusion in the database was limited to a small subset of Mg-, Fe-, Ca-, and Al-bearing sulfate minerals. However, since the major sulfate-bearing phases that occur during acid-sulfate alteration are included in this group, the database should allow adequate approximation of geochemical reactions for such systems. The sulfate minerals added to the database are listed in Table A1, along with the thermodynamic parameters used for each phase. In addition to the phases listed in the table, existing thermodynamic data in data0.ymp.R2 were adopted for several other sulfate-bearing phases including anhydrite, gypsum, bassanite, arcanite, ettringite, mirabilite, monosulfate, morenosite, and K-alum.
 The thermodynamic data adopted for this study as listed in Table A1 were compiled from a number of different sources. The thermodynamic parameters recommended by Stoffregen et al.  for alunite and natroalunite and by Drouet and Navrtosky  for jarosite and natrojarosite were adopted for the database. Although the latter study also recommends data for hydronium jarosite, those data appear to be precluded by the subsequent study of Majzlan et al. [2004a], whose values were used in the calculations. Data for magnesium sulfate minerals were adopted from the recent compilation of Grevel and Majzlan , while those for iron sulfates were derived primarily from the compilation of Hemingway et al.  and references therein. Majzlan et al.  provide more recent estimates of thermodynamic parameters for a few of the Fe-sulfate minerals; however, because the parameters provided in that study are for non-stoichiometric mineral compositions, it is problematic to make direct comparisons with the data of Hemingway et al. . Ackermann et al.  also provide recent estimates of thermodynamic parameters for kornelite and paracoquimbite, and their data for kornelite show some minor discrepancies with the estimates of Hemingway et al. . Although these more recent studies suggest that some aspects of the Hemingway et al.  compilation could be improved, the data are not yet sufficient to make a comprehensive revision of thermodynamic data for the Fe-sulfate minerals. For the present, we elected to use the Hemingway et al. values to maintain consistency among the suite of Fe-sulfate minerals in the database.
 Heat capacity data required to perform calculations at temperatures above the standard state of 25°C were available for only a few of the minerals listed in Table A1. For these minerals, heat capacities at elevated temperatures can be calculated using Maier-Kelley parameters available from the literature, as given in the table. For the other mineral phases, heat capacities at elevated temperatures were estimated using the van't Hoff approximation.
 The Maier-Kelley parameters recommended by Stoffregen et al.  for the alunite-jarosite group were used in the models; however, some of these values appear to be somewhat problematic. Actual measurements of heat capacities are available for only a few minerals in this group, including alunite [Kelley, 1960], hydronium jarosite [Majzlan et al., 2004a, 2004b], and non-stoichiometric jarosite [Drouet and Navrtosky, 2003]. Estimates of Maier-Kelley parameters have been made by Stoffregen and Cygan  for natroalunite and Stoffregen  for jarosite and natrojarosite. In both cases, the estimates are based on the corresponding Maier-Kelley coefficients for alunite [Kelley, 1960] using the estimation scheme described by Helgeson et al. . As shown in Figure A1, there appear to be some inconsistencies with the measured and estimated values for heat capacities among the members of this mineral group. The measured values for non-stoichiometric jarosite from Drouet and Navrtosky  follow a temperature trend that is similar to the estimated heat capacities for stoichiometric jarosite and natrojarosite as well as the trend for alunite that is based on measured values, but the jarosite data reported by Drouet and Navrotsky have considerably higher absolute values. At the same time, the measurements for hydronium jarosite from Majzlan et al. [2004a, 2004b] have similar absolute values to the estimates for jarosite and natrojarosite, but follow a somewhat different trajectory. These discrepancies suggest that there could be some problem with either the estimated or measured values but, in the absence of sufficient data to provide greater insight into the relative accuracy of the data and reasons for the apparent discrepancies, it is difficult to evaluate the reasons for these discrepancies. Accordingly, the recommended values were accepted for this study as reported, but further work on the heat capacities for this mineral group is clearly warranted. In any event, because the change in heat capacity makes only a relatively small contribution to the Gibbs energy of the minerals over the temperature range applicable to the present models, the potential discrepancies in heat capacities would have a relatively minor impact on the outcome of the models.
 The data0.ymp.R2 database allows solid solutions for a number of phases, all of which were retained in the model database. Additional silicate solid solutions included in the present database included serpentine (chrysotile and greenalite end-members), montmorillonite, and nontronite. For the latter two, each of the Ca-, Mg-, Na-, K-, and H-bearing end-members present in the data0.ymp.R2 database was included in the solid solution. Among the sulfates, solid solution was allowed only for the alunite-jarosite group. The other sulfate minerals included in the database exhibit little or no compositional variation within the chemical system used in the model, so solid solutions were not warranted. In contrast, minerals in the alunite-jarosite group exhibit considerable variation from end-member compositions in both natural occurrences and in our experiments, indicating that a solid solution model was appropriate. In all cases, ideal mixing models were used for the solid solutions.
 Sulfate minerals in the alunite-jarosite group have the general formula AB3(SO4)2(OH)6 where the A site is predominantly occupied by the monovalent ions K+, Na+, and H3O+ (hydronium) while the B site is occupied primarily by Al3+ and Fe3+ (Table A1) [Stoffregen et al., 2000]. Substitution of S by P or As can also occur, but this has no relevance to the present study and so was not considered in the model. Typically, the Al-bearing members (alunite subgroup) and Fe-bearing members (jarosite subgroup) are considered separately in solution models, since most natural samples tend to be either Al- or Fe-rich with limited substitution in the B site [Stoffregen et al., 2000]. However, this separation appears to be attributable more to the local geochemical environment rather than any factors that would preclude mixing of Al and Fe on the B site. Furthermore, the natroalunite produced in the experiments and observed at the field site [McCollom et al., 2012] exhibits considerable mixing on the B site. Accordingly, the alunite-jarosite group was represented in the models by a single five-member ideal site-mixing model that incorporated both the alunite and jarosite group minerals, using the end-member compositions listed in Table A1, excluding hydronium alunite.
 We initially considered inclusion of hydronium alunite in the solid solution model, but solution models that included this component produced unsatisfactory results. Since no published thermodynamic data are available for hydronium alunite, we estimated parameters for this end-member by assuming that the K+ for H3O+ substitution in alunite is thermodynamically equivalent to the same substitution in jarosite. That is, it was assumed that the standard Gibbs energy (ΔGr°), enthalpy (ΔHr°), and entropy (ΔSr°) of reaction, as well as the heat capacity of reaction (ΔCp,r), are all equal to zero for the reaction:
 Based on this assumption, thermodynamic parameters for hydronium alunite could be estimated from the corresponding data for the other phases, as listed in Table A1. When this mineral was included in the reaction path models, hydronium alunite always dominated the alunite-jarosite solid solutions, greatly predominating over the K- and Na-bearing members. Since this result did not appear to be geologically realistic, we concluded that the estimated values were probably not accurate and excluded the hydronium alunite endmember from the final models.
A1 Fluid Speciation Calculations
 An example of the results of the fluid speciation and mineral saturation state calculations performed with EQ3 is given in Table A2.
Table A1. Thermodynamic Parameters for Sulfate Minerals Added to the Database for Use in Numerical Models
Saturation index (SI) equal to log (Q/K), where Q is the reaction quotient and K is the equilibrium constant. Values of SI near one indicate saturation, while values above one indicate supersaturation and values below one indicate undersaturation. Only minerals with SI > −0.5 are shown.
 This work was partially supported by funds from the NASA Astrobiology Institute, NASA Exobiology program grant NNX07AU06G, and NASA Mars Fundamental Research Program grant NNX12AI02G. Tom McCollom gratefully acknowledges the support of the Hanse Wissenschaftskolleg during preparation of the manuscript. The authors appreciate the helpful comments of Paul Niles and an anonymous reviewer on the manuscript. The Institute for Rock Magnetism is supported by grants from the Instruments and Facilities Program, Division of Earth Science, National Science Foundation. This is publication 1201 of the Institute for Rock Magnetism. Niels Jöns was funded through DFG-Research Center/Excellence Cluster “The Ocean in the Earth System.”