A Spectroscopic Study of Mars-analog Materials with Amorphous Sulfate and Chloride Phases: Implications for Detecting Amorphous Materials on the Martian Surface

The Chemistry and Mineralogy X-ray diffraction (XRD) instrument aboard the Curiosity rover consistently identifies amorphous material at Gale Crater, which is compositionally variable, but often includes elevated sulfur and iron, suggesting that amorphous ferric sulfate (AFS) may be present. Understanding how desiccating ferric sulfate brines affect the spectra of Martian material analogs is necessary for interpreting complex/realistic reaction assemblages. Visible and near-infrared reflectance (VNIR), mid-infrared attenuated total reflectance (MIR, FTIR-ATR), and Raman spectra, along with XRD data are presented for basaltic glass, hematite, gypsum, nontronite, and magnesite, each at three grain sizes (<25, 25–63, and 63–180 μm), mixed with ferric sulfate (+/−NaCl), deliquesced, then rapidly desiccated in 11% relative humidity or via vacuum. All desiccated products are partially or completely XRD amorphous; crystalline phases include starting materials and trace precipitates, leaving the bulk of the ferric sulfate in the amorphous fraction. Due to considerable spectral masking, AFS detectability is highly dependent on spectroscopic technique and minerals present. This has strong implications for remote and in situ observations of Martian samples that include an amorphous component. AFS is only identifiable in VNIR spectra for magnesite, nontronite, and gypsum samples; hematite and basaltic glass samples appear similar to pure materials. Sulfate features dominate Raman spectra for nontronite and basaltic glass samples; the analog material dominates Raman spectra of hematite and gypsum samples. MIR data are least affected by masking, but basaltic glass is almost undetectable in MIR spectra of those mixtures. NaCl produces similar FTIR-ATR and Raman features, regardless of analog material.


Introduction
Throughout the traverse of the Curiosity rover, the CheMin instrument has consistently detected high abundances (15-73 wt%) of amorphous (by X-ray diffraction, XRD, hereafter referred to simply as amorphous) material in the sediment and sedimentary rocks of Gale crater (Bish et al. 2013;Blake et al. 2013;Dehouck et al. 2014;Vaniman et al. 2014;Treiman et al. 2016;Yen et al. 2017;Rampe et al. 2020a).The amorphous compositions between samples are variable but typically includes significant amounts of SO 3 , FeO T , and SiO 2 (Morris et al. 2016;Achilles et al. 2017Achilles et al. , 2020;;Morrison et al. 2018;Rampe et al. 2018Rampe et al. , 2020b;;McAdam et al. 2020;Smith et al. 2021Smith et al. , 2022)).Comparisons using mass-balance calculations based on major elements of the calculated amorphous component (Morris et al. 2013;Dehouck et al. 2014;Vaniman et al. 2014) are inconsistent with any single amorphous material, suggesting multiple phases.For example, samples with abundant FeO T may suggest the presence of nanophase Fe-oxides and/or amorphous Fe-sulfates, while the enrichment of SiO 2 suggests the presence of opal-A and/or rhyolitic glass (Rampe et al. 2020a).Smith et al. (2022) found that in the drilled samples, ∼20-90 wt% of the bulk, SO 3 is in the amorphous component, which highlights the significance of amorphous sulfur-bearing phases including Mg-, Fe-, and Casulfates or sulfites.
The prevalence of amorphous materials in Gale crater and their estimated compositions have led to several hypotheses on their origin.While some of the amorphous components could include primary silicate glasses, there are likely secondary phases present.These secondary materials could result from (and thus point to) different pH, temperature, and temporal formation conditions (Black & Hynek 2018;Smith et al. 2021), thus determining their composition is an important component of understanding Mars surface chemistry.To this end, several experimental and terrestrial studies have investigated the formation of secondary amorphous materials relevant to Mars.Analysis of modern subglacial sediments from the Three Sisters Volcanic Complex, Oregon, found multicomponent amorphous material enriched in SiO 2 (∼50-60 wt%) and FeO T (∼15 wt%; Smith et al. 2018;Rampe et al. 2022).Dehydration experiments first performed on epsomite by Vaniman et al. (2004) and on melanterite and epsomite by Wang & Zhou (2014) resulted in amorphization of the original Fe-and Mgsulfates.Diverse amorphous Fe-and Cl-rich nanospheres were found precipitated from an Fe-rich brine at the terminus of Taylor Glacier in Antarctica (Sklute et al. 2022), suggesting that in cold, dry climates, amorphous phases may be prevalent on Earth.Metastable amorphous aluminosilicate phases were also found in the Lake Lewis basin as a result of the alteration of lacustrine clays by brines (English 2001).A popular theory is that brines contribute to the amorphous component on Mars, either by the alteration of rock or sediment or through the evaporation products of brine dehydration (e.g., Sklute et al. 2015Sklute et al. , 2018a;;Yen et al. 2017;Achilles et al. 2020;Bristow et al. 2021;Smith et al. 2021).Dehydration of sulfate-enriched brines may explain the enrichment of sulfur in Gale crater amorphous materials.Due to the detection of jarosite, which requires a low-pH environment to precipitate, it is hypothesized that acidic alteration occurred in Gale crater (e.g., Rampe et al. 2017).The mixture of acids that can form in sulfur-enriched brines could also have provided the acidic alteration environment hypothesized for this region.
To understand the prevalence of brine-derived amorphous materials on Mars and the brines from which they form, amorphous phase compositions must be inferred from the elemental abundances observed in Martian data (see Berger et al. 2022), and then relative stabilities of potential brines must be investigated at Mars-relevant temperatures and relative humidities (Chevrier & Altheide 2008).Sulfate brines containing Fe and Cl are of interest in such investigations because of the prevalence of these elements in the Martian dust (Berger et al. 2016) and the high Fe content of the native basalt sourcing the brine cations (Black & Hynek 2018).Chloride deposits have been identified throughout the southern highlands of Mars by the Mars Odyssey Thermal Emission Imaging System, and spectra of these salt deposits are consistent with the presence of 10-25 wt% halite (Osterloo et al. 2008;Glotch et al. 2016).In Gale crater, localized enrichment in chlorine (up to 15 wt% Cl) is correlated with high sodium, indicating the presence of halite (Thomas et al. 2019).Furthermore, rapid desiccation of a ferric sulfate brine under low-pressure and lowrelative humidity (RH) Mars conditions produces an amorphous solid, both alone and in multicomponent brines that include various chloride species (Sklute et al. 2015(Sklute et al. , 2018a)).Because Fe-sulfate brines are projected to be stable over much of the Martian surface and subsurface (Chevrier & Altheide 2008), form amorphous solids on dehydration, and contain cations enriched in the amorphous fraction, they are a plausible starting point for past or present amorphous mineral formation, and thus are the basis for the present study.
Identifying and quantifying amorphous materials with spectroscopy is complicated, particularly in spectral mixtures (Sklute et al. 2015(Sklute et al. , 2018a;;Sheppard et al. 2022).It has been shown that in simple mixtures, amorphous materials can be masked by crystalline materials (Sklute et al. 2015(Sklute et al. , 2018a;;Black & Hynek 2018;Sheppard et al. 2022), and thus be significantly underestimated in spectral data.Masking in the visible and nearinfrared (VNIR) can be attributed to the nonlinearity of spectral mixing across these wavelengths (Hapke 1981;Singer 1981), which leads to more absorbing materials dominating the spectrum, especially at shorter wavelengths (Wellington 2018).Masking in the mid-infrared (MIR) is caused by similar nonlinearity, which is more pronounced for fine-grained (and likely amorphous) components due to multiple scattering (Hapke 1981;Kraft et al. 2003).Masking of amorphous phases in Raman spectra can be caused by sharp crystalline peaks swamping lower parabolic contributions and fluorescence obscuring identification (Black & Hynek 2018).Nonlinear mixing is seen in Raman data as well due to differences in Raman cross sections (Haskin et al. 1997).
Simple mechanical mixtures of amorphous and crystalline components are crucial to our understanding of limits of detection, but they are not the only type of mixture likely to be encountered on Mars.Mixtures resulting from in situ alteration (deliquescence, dissolution, and reprecipitation) must also be considered.For these mixtures, crystalline materials may participate in the mechanism that forms the amorphous material.For instance, if RH fluctuations or subsurface RH buffering leads to a saturated brine, elements can mobilize from crystalline phases, and those phases could be altered (e.g., Achilles et al. 2020;Bristow et al. 2021).Upon re-desiccation, the result would be a complicated assemblage of primary crystalline materials, crystalline precipitates, and amorphous components, either as independent phases, coatings, or both.If brine cycling does or did play a role in Martian surface chemistry, creating and analyzing these potential real-world, in situ products with significant amorphous content is essential to the identification of these cycles.These data will also aid in the interpretation of the relationship between brine cycling to the data obtained by the instrument suite on board both Curiosity and Perseverance, as well as spectra obtained from orbit.This study focuses on simple brine and analog material mixtures and thus fills the experimental gap between the limit of detection studies investigating simple mechanical mixtures (see Sheppard et al. 2022) and Mars-simulant reaction studies focusing on complicated final assemblages (see Tosca et al. 2004).
Here we present a detailed analysis of mixtures of amorphous materials with crystalline solids formed from deliquescence-driven saturation of ferric sulfate brines in contact with representative Martian analog minerals and glass: Hawaiian basaltic glass, gypsum, hematite, nontronite, and magnesite (with and without NaCl).Mixtures of these materials using <25, 25-63, and 63-180 μm size fractions are analyzed by XRD and VNIR reflectance, Fourier transform MIR attenuated total reflectance (FTIR-ATR), and Raman spectroscopies.All data are included in the supplement and at doi:10.5281/zenodo.7671724.This paper is the first in a project that seeks to understand and characterize spectra of common components of Martian rocks and regolith in mixtures with relevant amorphous solids, such as amorphous ferric sulfate (AFS), to facilitate improved interpretation of remote and in situ spectroscopic data.

Representative Mars-analog Material Preparation
Mars-analog mixtures were created from five representative Martian materials: Hawaiian basaltic glass (abbreviated as Bas), gypsum (Gyp), hematite (Hem), nontronite (Non), and magnesite (Mag).Hawaiian basaltic glass was collected as fresh glass from Kilauea by Tim Orr (USGS, HVO; Mackie et al. 2017).Gypsum was purchased as large crystal selenite (without provenance information) from a mineral dealer.A hematite kidney ore from Cornwall, England was purchased from a mineral dealer.Nontronite is Clay Mineral Society source clay nontronite NAu-1 (Keeling et al. 2000).Magnesite was also purchased (without provenance information) from a mineral dealer.All were initially analyzed spectrally and by XRD to confirm phase identification.A slight quartz contamination in the hematite was reduced by hand-picking out pieces with visible quartz crystals.NAu-1 appears to be a mixture of clays, but it was used without alteration for consistency because it is a standard.Magnesite and gypsum were confirmed pure at the resolution of XRD and infrared spectroscopy.The basaltic glass is primarily XRD amorphous, whereas the gypsum, hematite, and magnesite are clearly crystalline.While nontronite is crystalline, NAu-1ʼs finegrained nature and irregular stacking on its c-axis cause its peaks to broaden and add a substantial baseline swell.We, therefore, group it along with our XRD amorphous materials in this study because we find that it is not identifiable from its XRD pattern alone in a mixture.All materials were processed into powders using a ring and puck mill and dry sieved into <25, 25-63, and 63-180 μm grain sizes.

Ferric Sulfate Preparation
The foundation for each sample mixture in the study was anhydrous ferric sulfate (Fe 2 (SO 4 ) 3 ), also known as mikasaite.It was created from the dehydration of ferric sulfate hydrate (Fe 2 (SO 4 ) 3 • xH 2 O) at 300°C for 24 hr, with a ramp rate of 60°C every 15 minutes in small batches (8-10 g each) to ensure homogenous dehydration of each batch.The product was stored in an anaerobic chamber at room temperature, purged with dry N 2 to prevent rehydration, and each batch was made no more than 3 days before it was used.

Assembling Sample Mixtures
Experimental samples started as mixtures of two or three dry powder components with equal masses, with the resulting volume percentages being variable.All 60 mixtures (Table 1) include 0.300 g of a representative Mars-analog material, 0.300 g of mikasaite, and either no NaCl or 0.300 g of NaCl.
Samples that include NaCl in the starting mixture are designated with "N" in the sample name (e.g., Bas2VN).

Hydration Phase
Each prepared mixture was placed in a small weighing dish and secured to a small plastic stand with double-stick tape.The stand was placed in ∼0.5 cm of double-deionized water in a 0.4 oz airtight container with the dish ∼1.5 cm above the water surface.In the sealed container, the atmosphere reaches equilibrium with the water at 92% relative humidity at room temperature, about 25°C.A separate, identical container with just water and a humidity sensor was used to monitor the relative humidity.There was never direct contact between the sample mixtures and water reservoirs, and the samples hydrated through deliquescence only.The container was kept at room temperature, which was also continuously monitored.The small volume of the container minimized the effect of any atmospheric gradient that would cause heterogeneity in relative humidity or temperature within the sealed containers.
Powder mixtures began to deliquesce almost immediately as the relative humidity inside the containers equilibrated.Preliminary experiments had shown that all of the analog materials were mostly covered by solution after 4 days in ∼92% relative humidity.However, to ensure that the solutions reached saturation, the 60 samples were allowed to deliquesce for 10 days.This extended hydration period also allowed chemical reactions more time to occur in the sample after reaching saturation.The short experimental timescales allowed for the analysis of short cycle spectral changes.After 10 days of deliquescence, the samples were completely saturated and mostly liquid with nonsoluble analog material at the bottom of each saturated brine.

Desiccation Phase
Thirty of the samples were dehydrated in an applied vacuum engineering cylindric laboratory vacuum chamber (∼10 −3 mbar; Edwards 2-stage vacuum pump; see Table 1).The hydrated samples were transferred into glass vials with loosened caps, allowing for the air and water to escape without the loss of solids.Once under vacuum, the liquid samples began to rapidly bubble, and evaporation rate was adjusted so that agitated fluid moved no more than half way up the vial.This simulates boiling due to low atmospheric water vapor pressure on the Martian surface and proceeds more rapidly at room temperature than low-relative humidity desiccation.
The 30 other samples were desiccated under low-relative humidity (RH) in the same plastic dish in which they were hydrated (Table 1).Sample dishes on plastic stands were transferred into identical airtight containers, with lithium chloride (LiCl) powder to act as a RH buffering salt, which quickly stabilized the RH inside the container to ∼11% at equilibrium; this is consistent with daytime RH observed on the Martian surface (Savijärvi 1995;Harri et al. 2014).Separate identical containers with only LiCl and a humidity sensor were used to monitor RH.Containers remained sealed from the start of desiccation to the removal of the samples for spectroscopy.
Throughout both desiccation methods, various precipitates formed, and all solutions turned into either a thin glassy film or a heterogeneous crust.All samples were thoroughly desiccated after 10 days in either a vacuum or low-RH container.They were then transferred into a Vacuum Atmospheres Company anaerobic chamber and prepared for XRD and spectroscopic analysis by lightly grinding with an agate mortar and pestle until they were homogenous powders.Each technique was performed on a separate aliquot of powdered sample.The photographic progression of hydration and desiccation of identical samples can be seen at https://www.lionsandlamms.com.

XRD Analysis
XRD patterns were acquired on a Rigaku SmartLab II X-ray Diffractometer in the laboratory of Dr. Kevin Kittilstved at The University of Massachusetts, Amherst.XRD patterns were taken from 10-60 2θ using CuKα radiation, with a 0.01°step size, sampling at 3°minute −1 on a zero-background holder in Bragg-Brentano geometry.XRD under ambient atmosphere was run only when RH in the laboratory could be maintained <35%.Each sample was exposed to the ambient atmosphere for ∼42 minutes and checked for signs of rehydration after analysis.While Cu-Kα XRD sources can lead to fluorescence in iron-rich samples, this Rigaku instrument has been used extensively to study nanophase iron oxides (see Sklute et al. 2018a) without notable baseline swell or loss of signal.Therefore, we believe that our results are not significantly affected by Fe-fluorescence.
XRD analysis was first performed using the Rigaku SmartLab II (RSL2) software, ICDD PDF2 database (2019 edition).All patterns were also fit using the free Match!software from crystal impact, with the crystallography open database (COD) as a reference library.Certain phases did not as readily appear as positive identifications in Match! as in RSL2.When mineral identifications were forced, they could be located in the COD library, with the exception of basic ferric sulfate (FeOHSO 4 ), which was manually added from the American Mineralogist Crystal Structure Database.Differences between the fitting algorithms used by the two pieces of software clearly affect the results.For both Match! and RSL2, trace phases had to be manually added based on a priori knowledge of peak locations (based on other fits) because those phases rarely met the threshold required for positive identification.Each phase, automatically identified or manually added, was then inspected for reasonableness.Amorphous percent was determined using the Match!software's degree of crystallinity (DOC) function, which subtracts the pattern of the sample holder from the pattern of the sample and then assesses the percent area in peaks versus amorphous background.The results of this calculation are the percentage of the pattern that is in nonpeaks versus peaks, which is not necessarily equivalent to wt. or mol%, especially if the crystalline and amorphous phases have differing compositions.

VNIR Spectroscopy
Bidirectional VNIR reflectance spectra were taken with an ASD FieldSpec 4 Hi-Res spectrometer, with incidence at 30°a nd emission angle of 0°.Samples for VNIR were illuminated through a 1000 μm silicon fiber with an Ocean optics HL-2000-HP tungsten halogen light source.VNIR spectra were taken inside a fully enclosed black box using matte black sample cups.Three spectra were taken for each sample, each an average of 240, 136 ms integrations.Prior to each sample, a dark current and white reflectance (Spectralon) spectrum were acquired.VNIR spectra were recorded under ambient atmosphere only when RH in the laboratory could be maintained <35%.

Raman Spectroscopy
Raman spectra of all samples were acquired on a Bruker BRAVO Raman spectrometer, which uses both 758 and 852 nm lasers, through the bottom of the glass vials filled with N 2 .The glass vials do not have a significant Raman signature, and we confirmed that they did not contribute to the Raman spectra in the regions of interest.Each spectrum was collected over 100 ten second integrations, with 2 cm −1 spectral resolution from 300-3200 cm −1 .With this step size, band assignments within ±2 cm −1 of a Raman peak in the spectra of our experimental products are considered.Baseline removal was performed on all Raman spectra using a concave rubber- Notes.All 60 starting sample mixtures include 0.300 g of a representative Mars-analog material, 0.300 g of mikasaite, and either no NaCl or 0.300 g of NaCl.
a Samples that were vacuum-desiccated have a "V" in the sample name, while samples that were RH-desiccated have an "R" in the sample name.
b "Basalt" is shorthand for "basaltic glass." band correction in the Bruker OPUS software (bin size 64; iterations dependent upon sample).

MIR Spectroscopy
FTIR-ATR spectra were acquired on a Bruker Alpha FTIR spectrometer (RT-DLATGS detector; KBr beam splitter) using a Platinum TM diamond attachment.Spectra were collected from 360-4000 cm −1 with spectral resolution of 4 cm −1 .Each spectrum was an average of 128 scans (Mertz phase correction; Norton-Beer Strong Apodization).Samples were analyzed in the anaerobic chamber.Continuum removal was performed using a concave rubber-band correction in the Bruker OPUS software (bin size 64; iterations dependent upon sample).

Reaction Products
The basaltic glass analog used in this study is unique compared to our other analog materials because it is 100% glass, which is also XRD amorphous.During the 10 days of deliquescence, no visual signs of dissolution of the basaltic glass were observed (e.g., no color or noticeable volume changes).Thus, it was expected that spectra of dehydrated basaltic glass-and FS-containing products would indicate mixtures of Bas+AFS/FS, with various levels of FS hydration.Similarly, the basaltic glass-containing samples that also included NaCl were expected to also have halite or potentially other chlorine salts.Basaltic glass mixtures were the only samples that became rimmed with oily bright yellow fluid, indicating additional solution phases, possibly chlorine species.After the liquids were stirred and allowed to desiccate, they each formed a solid chip interpreted as basaltic glass grains cemented together with AFS (Figure 1(A)) for NaCl-free samples and cemented along with euhedral NaCl crystals by FS/AFS in NaCl-bearing mixtures (Figure 1(B)).The basaltic glass grains naturally separated from the brine despite the powders being initially well mixed, likely due to the significant density difference between basaltic glass grains and the deliquescing FS and NaCl, Hematite-containing samples also showed no visible signs of hematite dissolution during the 10 days of deliquescence.Like the basaltic glass samples, the hematite separated from the brine, leading to a hematite-rich base layer intercalated with and overlain by the saturated FS or FS + NaCl brine.However, by the end of the 10 days of deliquescence, a thin metallic-like sheen was observed floating on top of the brines of all hematite-containing samples.It is unknown whether this substance originates from impurities within the natural hematite sample, or if this is a product of a reaction between hematite and the FS (+ NaCl) brines.After the hematite-containing liquids were stirred and allowed to desiccate, they each formed a solid chip interpreted as a base layer of hematite grains cemented with and overlain by AFS for NaCl-free samples (Figure 1(C)) and cemented along with euhedral NaCl crystals by FS/AFS in NaCl-bearing samples (Figure 1(D)).
Gypsum in Gyp2V, Gyp2R, Gyp2VN, and Gyp2RN visibly dissolved during the 10 days of deliquescence.Despite the extensive dissolution, it was anticipated that there would be substantial Ca-sulfate in the dehydrated product due to its incredibly low solubility, particularly in saturated brine (Reiss et al. 2021).By the end of the 10 days of desiccation, gypsumcontaining mixtures without NaCl displayed patches of opaque material in a transparent matrix interpreted as gypsum cemented by AFS (Figure 1(E)).NaCl-containing samples appeared as a homogenous opaque solid (Figure 1(F)).
The nontronite-containing experimental mixtures did show signs of reaction during 10 days of deliquescence, including subtle color changes and material segregation.After desiccation, the NaCl-free sample formed a fractured, visually homogenous solid (Figure 1(G)) while the NaCl-bearing samples formed a complex solid chip composed of several materials of varied texture, including several light-yellow crystals visible to the naked eye (Figure 1(H)).The solids formed from the desiccation of NaCl-free nontronite mixtures were the hardest and most difficult to grind, demonstrating AFS's potential as an effective cementing agent for certain minerals.
Magnesite-containing mixtures were by far the most visibly reactive in this study.This result was anticipated due to the neutralization potential carbonate would have on the sulfuric acid resulting from the deliquescence of mikasaite.We observed bubbling throughout the 10 days of deliquescence, although the bubbling significantly slowed after just a few days.These were also the only samples not to form a liquid after 10 days of deliquescence.The NaCl-free samples formed a mixture of viscous gel and solid with an orange, sponge-like texture, and NaCl-containing samples formed a porous, iridescent mud.After 10 days of desiccation, NaCl-free samples formed a brittle solid filled with holed where gas escaped the mixture (Figure 1(I)).NaCl-bearing mixtures formed a brittle, hole-ridden, hollow shell (Figure 1(J)).

X-Ray Diffraction
All samples were a mixture of crystalline and amorphous phases, with amorphous phases composing >70% of each bulk sample, consistent with results of previous similar experimental studies (Ling & Wang 2010;Sklute et al. 2015Sklute et al. , 2018a)).There is no trend linking amorphous percent and desiccation method.Basaltic glass-containing samples without NaCl were ∼98% amorphous, and with NaCl were ∼90% amorphous.Hematitebearing samples were 96% amorphous without NaCl, with NaCl, the DOC varied by grain size and dehydration method but was between 84% and 90%.Nontronite-containing samples were determined to be ∼97% amorphous when not containing NaCl and 88% amorphous with NaCl in the mixture.Gypsumbearing samples without NaCl were ∼85% amorphous versus ∼71% amorphous when NaCl was in the mixture.Finally, magnesite samples were determined to be between 89% and 94% amorphous in NaCl-free samples and between 73% and 90% amorphous when NaCl was included.The DOC for each sample is reported as amorphous percentage in Table A1 in the Appendix.
Low DOC values (high amorphous percent) in basaltic glass and nontronite mixtures were expected because the starting material for these samples was also either amorphous (for basaltic glass) or displayed a swelled baseline (for nontronite).However, based upon DOC for the other samples, the 50-66 wt% of crystalline phases in the mixtures do not seem to be reflected in the relative areas of the peaks versus the amorphous baseline swell.By these numbers, it appears that gypsum is the least reactive of the phases, but all samples seem to undergo a degree of transformation that reduces the size of scattering domains (for crystalline samples), leading to broader peaks and lower DOC.Clearly, this is not a rigorous crystallinity assessment.Our use of XRD in this study was to assess if samples still appeared XRD amorphous after hydration/ dehydration with the analog material present, and to identify crystalline phases.A complete Rietveld refinement and quantitative determination of amorphous abundance and crystallite size is outside the scope of this work.However, our data are available in the supplementary information and upon request for any researchers who wish to apply a more rigorous methodology.
The crystalline phases identified by XRD for each experiment are summarized in Table 2 and are listed in detail in Table A1.XRD data show that metasideronatrite (Na 2 Fe(SO 4 ) 2 (OH)•H 2 O) is a preferred product when Na + is present in the solution.However, for hematite samples, it only appears when the samples are vacuum dehydrated.When metasideronatrite is selected as a component of the mixture for RH sample, Match! places its abundance at 0%.The most interesting transformation, however, occurs in the magnesite samples, where natrojarosite (NaFe 3 (SO 4 ) 2 (OH) 6 ) is the preferred Na + -bearing phase.Due to jarosite's identification on Mars (Klingelhöfer et al. 2004), and its indication of a wet, acidic weathering environment (McLennan et al. 2005;Elwood Madden et al. 2009;Pritchett et al. 2012), the fact that it arises in our mixtures only when we have a partial neutralization by carbonate may have implications for Martian surface chemistry.Some of the observed phases are due to contaminants in the starting material that were not apparent in the pure end-member XRD patterns but are clearly present in the final products while others can be explained by cations mobilized by the acid brine.Quartz was a known impurity in the hematite sample; it seems concentrated in the 25-63 μm grain size fraction.Quartz is also a trace phase in the magnesite sample and a prominent minor phase in the nontronite standard.Gypsum is precipitated in both our nontronite and our magnesite samples.Acid mobilization of Ca 2+ from the interlayer site in nontronite or trace calcite/dolomite contaminants in magnesite samples likely combines with sulfate in solution to form this phase.Magnesite mixtures also precipitated hexahydrite (MgSO 4 • 6H 2 O), but only in the slowly dried low RH samples, with the exception of the large-grained Mag1RN, which did not show hexahydrite.In all cases, there are insufficient crystalline phases containing Fe 3+ and SO 4 2-to compensate for what went into solution, indicating that AFS is present in all mixtures.

Grain Size Dependence of Spectral Results
The prepared grain sizes of the analog materials did not make a significant difference in the spectra of the final dehydrated products (see Figures A11-A20).For the samples including gypsum, nontronite, and magnesite, most or all of the analog material visibly dissolved or reacted during the 10 days at 92% RH along with the FS, or FS+NaCl.In these cases, the analog grain size may have had an effect on the rate of deliquescence and dissolution from available reactive surface area, but even these samples appeared to cease reacting long before 10 days at 92% RH was over; thus, the grain size does not seem to be a factor.For the samples with basaltic glass or hematite, the analog material did not noticeably dissolve into the FS or FS+NaCl brines.For these samples, VNIR reflectance does not vary systematically with starting grain size; thus, grain size does not seem to be a controlling factor in final spectral character.During desiccation, grains of basaltic glass or hematite became cemented together, forming a singular mass of grains in a glassy-amorphous matrix.These masses were ground by-hand in a consistent manner in preparation for spectral analysis.The scattering domains or "apparent" grain size are likely of agglomerates of cemented grains.Figure 2 compares basaltic glass mixtures with NaCl at all grain sizes (an analog material that did not dissolve) and gypsum mixtures for all grain sizes (an analog material that did dissolve).Due to the similarity between grain size products, this section details results for the 20 25-63 μm samples only.
There are roughly five crystal field transitions in these spectra attributable to Fe 3+ .AFS, basaltic glass mixtures without NaCl, and gypsum mixtures without NaCl have an absorption near 370 nm ( 6 A 1 → 4 E).Gypsum and basaltic glass mixtures with NaCl, along with all magnesite samples, instead have an absorption near 400 nm ( 6 A 1 → 4 T 2 ).Another band between 420 and 430 nm occurs in AFS with and without NaCl, basaltic glass and gypsum without NaCl, and all nontronite mixtures ( 6 A 1 → 4 E 4 A 1 ).A band between ∼515 and 565 nm (2( 6 A 1 → 4 T 1 ) is seen in AFS, hematite, and magnesite with and without NaCl, and in basaltic glass, gypsum, and nontronite without NaCl.
Finally, the most diagnostic band occurs between 796 and 1050 nm and is highly dependent on the mixture.This last region has potential usefulness for remote sensing identifications, particularly in the pure AFS and AFS with NaCl, where Fe 3+ absorptions occur at shorter wavelengths than almost any common Fe 3+ -bearing mineral (discussed further in Section 4.4).For gypsum, nontronite, and magnesite samples, the visible absorption in the 796-1050 nm range shifts to longer wavelengths when NaCl is present in the mixture.For basaltic glass samples, the trend is reversed, and hematite samples show no evidence of AFS in this spectral region.The shape and position of the VIS maxima (Figure 4) are variable but appear to be dependent upon cations other than Fe 3+ .For instance, in AFS, basaltic glass, and gypsum samples, the presence of NaCl appears to narrow and shift the VIS maximum to longer wavelengths, likely due to a loss of the absorption near 545 nm and increased absorption at shorter wavelengths.Nontronite and magnesite samples have a less clear trend, likely due to the prevalence of other cations that may compete for coordination with the iron octahedra in the alteration products.
Hydration features are also ubiquitous in this data set.AFS and gypsum with and without NaCl display a combined H 2 O/OH (ν 1 + ν 2 + ν 3 ) transition at 1168 and 1180 nm, respectively.All samples have hydration features between 1432 and 1468 nm, assigned to H 2 O/OH (2ν 1 + ν 3 ).Nontronite mixtures display a characteristic doublet in this region, where gypsum mixtures display its typical triplet.AFS, basaltic glass, and gypsum mixtures with and without NaCl, have a H 2 O/OH + SO 4 overtone absorption at 1780 for AFS and basaltic glass and 1750 nm for gypsum.Finally, H 2 O/OH (ν 2 + ν 3 ) bands between 1920 and 1990 nm are observable in all mixtures, manifested either as a single asymmetric absorption or a doublet.There are also typical OH combination triplet bands in gypsum between 2178 and 2276 nm.In basaltic glass, hematite, and one magnesite (all with NaCl), along with all nontronite mixtures, there is a single absorption near 2285 nm.These hydration bands are important because when they occur in combination with the Fe 3+ features, they may be diagnostic of AFS mixtures.In addition, some workers use hydration features in this region to be diagnostic of various clay and sulfate minerals, but these data show that many compounds may have absorptions in this region.

Raman Spectroscopy
Raman spectra of all samples from the 25-63 μm size fractions are shown in Figure 5. Features and assignments, where known, are detailed in Table 4 for each group of samples and summarized here.Assignments are based on the prior work by Berenblut et al. (1971), Lewis & Farmer (1986) All samples have some variation of peaks arising from SO 4 .A prominent and diagnostic set of peaks from 983-1094 cm −1 arises from SO 4 ν 1 .In basaltic glass, nontronite, and Mag2RN, there is a sharp feature at 983-994 cm −1 ; its location is inconsistent with any published Fe-sulfate but is comparable to the ν 1 feature of Na-sulfate (Qiu et al. 2019).All samples excluding basaltic glass and gypsum without NaCl have some combination of the sharp features near 1009 cm −1 .All samples excluding hematite, gypsum, and nontronite with NaCl have a weak feature near 1036 cm −1 .Basaltic glass and nontronite samples prepared with NaCl have three overlapping SO 4 ν 3 bands: 1114-1118 cm −1 , 1135-1154 cm −1 , and 1198-1200 cm −1 assigned to SO 4 ν 3 ; the 1115 cm −1 feature appears in gypsum samples with NaCl, and the 1135 cm −1 band also occurs in all gypsum samples with and without NaCl.
All four magnesite spectra show a feature at 328 cm −1 , which is a CO 3 E g libration (Dufresne et al. 2018;Kim et al. 2021) seen in all magnesite-bearing samples including the initial magnesite.The other two Raman peaks of CO 3 in magnesite, E g (ν 4 ) symmetric bending at 738 cm −1 and A 1g (ν 1 ) symmetric stretching at 1094 cm −1 (Kim et al. 2021), are also present.It is apparent that despite the vigorous reaction between the carbonate and sulfuric acid during the 10 days of deliquescence, some magnesite either persisted or re-precipitated into the final dehydrated products.
A nearby feature at 332 cm −1 is seen in the other four samples (basaltic glass, hematite, nontronite, and gypsum) mixed with NaCl and is assigned to FeCl 4 - (Sitze et al. 2001).
At slightly longer wavenumbers are two minor peaks at 368 and 418 cm −1 , found only in nontronite, that are likely Si-O-Fe vibrations (Frost & Kloprogge 2000) since they are also seen in the initial nontronite.Both Non2VN and Non2RN also show a Fe-O symmetric stretching feature (Sklute et al. 2018b) or possibly an Si-Fe-OH vibration (Frost & Kloprogge 2000), at 388 cm −1 .Basaltic glass, hematite, and nontronite all show subtle bands at 388-410 cm −1 , likely due to Fe-O or Fe-OH (Oh et al. 1998); basaltic glass has an additional feature at 420 cm −1 .Hematite and gypsum have small Fe-O bands at 468 cm −1 , likely related to AFS.Raman spectra of the hematite-containing samples are all remarkably similar and include the same Fe-(OH) bands at 494 and 610 cm −1 , also seen in the spectrum of the initial hematite.
Key takeaways from the Raman data include the following: 1.All NaCl mixtures contain a peak or shoulder near 332 cm −1 ; this could serve as an indicator of materials that have formed from Fe-and Cl-containing brines.This assessment would be corroborated by the presence of Ferelated features in the region around 490 cm −1 .2. The <900 cm −1 region should display peaks from metal cations and sulfate v 2 and v 4 vibrations in the components of the mixtures.However, while the end-member characteristics are apparent in some spectra (e.g., basaltic glass, nontronite, magnesite, and hematite), AFS is not apparent in the gypsum mixtures in this spectral range and incredibly muted in hematite mixtures.This discrepancy is likely due to the difference between the Raman cross sections of the component phases.3.In the region associated with sulfate absorptions (∼980-1100 cm −1 ), spectral masking due to differences in Raman cross section are even more apparent.While the AFS-associated broad sulfate feature at 1036 cm −1 is observable in all magnesite and basaltic glass mixtures, along with gypsum and nontronite without NaCl, it is much weaker in the hematite, gypsum, and nontronite samples without NaCl at the same relative abundance in the starting mixture.In mixtures where this feature is not observable (e.g., Non2RN, Non2VN, Gyp2VN, Hem2RN, and Hem2VN), there is instead a sharp peak   Notes.
a "V" indicates samples that were vacuum-desiccated, "R" samples were RH-desiccated.Samples that included NaCl in the starting mixture are designated with "N." ∼1010 cm −1 , which is attributed to the sulfate V 1 vibration in sulfates containing Na + .4. These results show that spectral masking is prevalent in AFS-bearing mixtures and varies by spectral region and phase assemblage.Most notable is that seen in hematite mixtures without NaCl, where the sulfate peak near 1010 cm −1 or 1036 cm −1 has significantly reduced intensity, potentially leading to underestimation of sulfate abundance in hematite-bearing deposits.

FTIR Spectroscopy
FTIR-ATR spectra are shown in Figure 6, and feature positions and designations are listed in Table 5. Assignments are from White (1971) In the MIR, spectra of all samples are dominated by features from SO 4 and CO 3 .In addition, Fe-O vibrations at 393, 442, and 458 cm −1 occur in hematite prepared without NaCl.AFS, basaltic glass, and nontronite samples without NaCl all have an absorption near 450 cm −1 , which could be due to Fe-O or SO 4 .For basaltic glass and nontronite, this feature also appears in the starting analog material, although shifted slightly.
All samples prepared with NaCl have a distinct absorption at 508-512 cm −1 .The addition of NaCl to the mixture also leads to fine structure in the 593-668 cm −1 range that varies by material but appears to group by major cations; AFS, basaltic glass, hematite, and nontronite samples with NaCl all have similar structure in this region, whereas gypsum and magnesite samples with NaCl are distinct.Features in this range, which are seen in all samples to varying degrees and densities, are likely due to SO 4 ν 4 modes, which has absorptions between 588-597, 650-655, and 666-668 cm −1 (Ling & Wang 2010), although these features do overlap with OH-deformation absorptions (Cornell & Schwertmann 2003).Many spectra also show a familiar series of ν 4 vibrations of SO 4 at 608, 619, and 633 cm −1 (Ling & Wang 2010); however, these features could also be attributed to Fe-O or Fe-OH (Oh et al. 1998;Cornell & Schwertmann 2003) for these samples.Weak features at 800 and 817 cm −1 seen in hematite with NaCl and all nontronite samples are likely due to absorbed surficial H 2 O.
All samples with NaCl display a similar set of nine close ν 1 and ν 3 SO 4 vibrational mode absorptions (972-976, 993-1009, 1031, 1051-1052, 1100-1115, 1125-1128, 1201, 1205-1209, and 1262 cm −1 ).With the exception of magnesite samples, variation in feature position and relative depth in this region appears to be due to admixing with absorptions present in the starting material rather than shifts unique to the mixtures.All samples with and without NaCl display a shallow H 2 O bending absorption at 1620-1627 cm −1 ; gypsum samples, with and without NaCl, also show the other half of this H 2 O bending doublet at 1682 cm −1 .
All four magnesite-bearing samples show features at 473, 414, and 379 cm −1 that are CO 3 lattice modes (Zhuravlev & Atuchin 2021), as well as CO 3 ν 2 and ν 4 at 748 and 887 cm −1 , respectively (Kim et al. 2021).Mag2V and Mag2VR show a    minimum at 1125 cm −1 but that peak is only a shoulder for the low-RH-dehydrated samples.Conversely, Mag2R and Mag2RN show minima at 1080 and 1095 cm −1 , respectively.This difference suggests that rate of desiccation plays a role in the final deformation of the sulfate tetrahedra within the magnesite-bearing sample products; vacuum desiccation is much more rapid than low-RH desiccation.The feature in magnesite at 1095 cm −1 is a ν 1 vibration of CO 3 (White 1971;Canterford et al. 1984) that is an extremely weak feature in the initial magnesite spectrum.Blending of CO 3 ν 1 and SO 4 ν 3 features likely explains the variations in this spectral region compared to the other samples in this study.Finally, the ν 3 vibration of CO 3 is seen in all four magnesite spectra as a broad absorption at ∼1450 cm −1 and a shoulder at 1415 cm −1 (White 1971;Canterford et al. 1984), which is shifted from the starting material.
The MIR region has significant diagnostic potential for these mixtures; the spectra appear, for the most part, to be mixtures of spectral elements from their various phases.The exception to this is the basaltic glass samples, in which the spectrally bland glass shape is overprinted by stronger features.The region below ∼900 cm −1 is a complex overlay of sulfate and metaloxygen vibrational absorptions.Mixtures without NaCl all display a similarly shaped broad band envelope between ∼900 and 1300 cm −1 , whereas all samples with NaCl instead have a series of nine more distinct features.In both cases, this region may prove useful in identifying AFS with and without NaCl in mixtures.Ling & Wang (2010) also showed that AFSs can hold between 5 and 11 structural waters, which correlates with the position of its SO 4 ν 1 Raman peak, with lower wavenumbers indicating an increase in the number of structural waters.All of our desiccated products without NaCl show the SO 4 ν 1 peak between 1030 and 1040 cm −1 , which would imply a varying number of structural waters between samples.We cannot extrapolate water contents in our samples using the Ling & Wang (2010) hydration analysis because the hydration feature is shifted relative to those reports.Qualitatively, vacuumdesiccated samples have less water than RH-desiccated samples.For example, the SO 4 v 1 is at 1036 cm −1 for Hem2R and 1040 cm −1 for Hem2V.In addition, the AFS component of all mixtures appeared to have less water than the pure AFS, whose peak is at 1026 cm −1 (Table 4), although this could be due to admixing of sulfate-bearing phases.

Amorphous Ferric Sulfate
Another trend observed throughout the spectra of this study is that the H 2 O/OH absorptions in the VNIR tend to be sharper for the vacuum-desiccated samples, indicating the rate of water loss is important in final spectral attributes, even when the XRD-derived phases appear the same.The rapid low-pressure boiling of the vacuum desiccation process both acts as a selfmixing process and leads to more rapid and thorough dehydration (as evidenced by the Raman SO 4 v 1 peak position).We posit this combination of factors favors specific molecular orientations, which would lead to a narrowing of peaks with respect to the RH-dehydrated samples where the more gradual dehydration allows for more extensive molecular reorientation.

Effect of Sodium Chloride
The inclusion of NaCl in the starting sample mixture leads to a more complex final product after desiccation.Aqueous Na +  A2.Most library spectra are from minerals in Darby Dyarʼs personal collection, and those data are also included as part of the supplement.Other reference spectra are from the PDS. and Cl − facilitate a much more complex geochemical environment, even if Na + or Cl − are not obviously represented in the final products.Although the hydration states of our products were not determined in this study, NaCl likely had some effect on these hydration states due to lowering of the activity of water.Preliminary pH analysis of the solutions showed that mixtures with NaCl formed a lower pH fluid on hydration than those without NaCl.At saturation, pure FS brine (which forms AFS after desiccation) had a pH of 2.5-3; NaCl + FS brine had a pH < 2. There are a few possible explanations for this decrease in pH.First, it is possible that some Cl − and H + combine to form HCl. Second, it is possible that our pH meter does not function properly with these high concentrations that result from deliquescence.Third and most likely, the increased ionic strength and ion concentration from including NaCl in the mixture allows for more interactions with ions from the other materials in the mixture.This allows for more dissolution of the FS and Mars-analog materials, which contribute to the acids that lower the pH, such as sulfuric acid.Reaction products in the mixtures with NaCl may only precipitate at lower pH.Alternatively, these phases may be stabilized, either directly or indirectly, by the ions in solution.Na + and Cl − can electrostatically coordinate with mineral and proto-mineral surfaces; the charge at mineral surfaces is a controlling factor in mineral growth and transformation (Stumm et al. 1992).The pH-lowering effect can also impact dissolution of the analog material, which could also increase the concentrations and complexity of ions in solution, and thus create more complex final products.We anticipate that these saturated solutions around minerals display pH gradients and microenvironments due to surface charges and cation affinities.These environments will change throughout the desiccation process.It is, therefore, difficult to assess the exact pH conditions driving this complex precipitation chemistry.
Even though NaCl-bearing solutions have increased chemical complexity, the products are incredibly similar; the same sequence of nine SO 4 vibrations between 1300 and 900 cm −1 is seen in FTIR-ATR spectra of all sample mixtures containing NaCl.The band positions and shape of this band envelope are almost identical to those in AFS+NaCl without an analog material.The distortion of the SO 4 tetrahedra in these samples causes the ν 1 and ν 3 vibrations to split into the nine features we consistently seen in this range (Ling & Wang 2010).We conclude that the NaCl dissolved in the brine is the primary cause of this distortion.The slight deviations in relative depths and positions (<±4 cm −1 ) of the features within this range are likely due to admixing of spectral features from the analogs as well as the small changes in bonding conditions caused by analog-derived elements.It is clear from the XRD patterns, and suggested by the spectra, that some of the Na + and Cl − is incorporated into other phases, potentially both crystalline and amorphous; however, the majority of the dissolved NaCl appears to re-crystallize in most of the desiccated products, often forming millimeter-sized euhedral crystals.

Additional Phases
In addition to AFS, we must also acknowledge the potential for additional amorphous phases.Our basaltic glass sample is 100% amorphous; thus, it is no surprise that samples such as Bas2R and Bas2V are consistent with an entirely amorphous solid of basaltic glass and AFS.
The spectra of half of the samples (NaCl-containing and some others) suggests the presence of FeCl 4 -.The common Raman feature at ∼332 cm −1 is most likely due to symmetrical stretching of FeCl 4 -tetrahedra (cited at 334 cm -tetrahedra and then precipitate as a distinct amorphous phase or within the desiccated AFS structure.Its absence in XRD patterns could be because it is XRD amorphous, below the limit of detection, or so unstable in minimally humid air that it does not persist throughout the analysis; amorphous ferric chloride was the least stable phase against rehydration in Sklute et al. (2018a).In addition to NaCl-bearing mixtures, the feature at 332 cm −1 is observed in all magnesite-bearing samples.However, all magnesite-bearing samples also show a CO 3 librational feature at ∼325-330 cm −1 , which also appears in pure magnesite.With our spectral resolution, these two features are indistinguishable.In natural mixed samples observed at decreased spectral resolution, the CO 3 feature could not be distinguished from the FeCl 4 -feature.This fact has important implications for in situ Raman spectra of complicated Martian materials, which may have carbonate materials coated/mixed with amorphous solids containing Fe-Cl bonds.While many of the spectral features observed are explained by mixtures of AFS or AFS+NaCl with our representative analog materials, there are new features across all samples that suggest the formation of additional phases during the deliquescence-desiccation process.Many phases consistent with these features were identified within the crystalline component of our samples via XRD (Tables 2, A1).The spectra of many of our products that contained NaCl in the starting mixture show a close resemblance to that of sideronatrite (Na 2 Fe(SO 4 ) 2 (OH)•3H 2 O) or metasideronatrite (Na 2 Fe(SO 4 ) 2 (OH)•H 2 O).While metasideronatrite was one of the minerals detected as part of the crystalline component in our samples using XRD (Table 2), it is identified as a minor or trace phase in most samples (Table A1), and it is difficult for us to determine how pervasive it is in our bulk products due to their overall predominantly amorphous nature.Metasideronatrite reversibly hydrates to, and is very spectrally similar to sideronatrite; thus, we cannot distinguish between these two in our spectra but conclude that the overall bonding structure of a phase in our samples is similar to (meta)sideronatrite, and that Fe 3+ , Na + , and structural H 2 O/OH distort the sulfate tetrahedra in our samples in a similar manner to that found in (meta)sideronatrite.Even the basaltic glass-and hematitebearing samples, in which we did not expect to see any brineanalog reaction, showed evidence of (meta)sideronatrite in FTIR spectra, even though hematite shows almost no (meta) sideronatrite by XRD.We infer that the formation of (meta) sideronatrite is a result of a reaction between the dissolved Fe 2 (SO 4 ) 3 and NaCl, with the spectral similarities between the spectra of AFS+NaCl and all spectra that included these two components, regardless of analog material.Interestingly, the product of an Fe 2 (SO 4 ) 3 and NaCl brine alone (AFS+NaCl) is X-ray amorphous (Sklute et al. 2018a), yet still shows some spectral similarities to a mineral such as (meta)sideronatrite.However, when desiccated alongside an analog powder, the brine more clearly precipitates out some amount of crystalline (meta)sideronatrite.The analog powders likely provide a nucleation surface for the (meta)sideronatrite, and contribute cations to the fluid that increase supersaturation, thus facilitating its crystallization during desiccation.The exact details of this surface chemistry require further study, and would vary for each analog material.The experimental samples containing nontronite and NaCl in the starting material (Non2VN & Non2RN) are the most similar to (meta)sideronatrite, which reflects the 10%-20% (of the crystalline material) being identified as metasideronatrite by XRD.It is likely that sodium is acid mobilized from the nontronite, pushing the solution chemistry toward supersaturation for (meta)sideronatrite, or similar material.

Diagnostic Features
All of our experimental samples are highly amorphous or nanophase, as is evidenced by the lack of sufficient crystalline phases in the XRD patterns to account for the sample chemistry, along with baseline swell indicative of amorphous materials.The position of the baseline swell is mixture dependent, but often with one broad rise <20°2θ and one between 25°and 35°2θ.
In the VNIR, the Fe 3+ absorption at 799 nm is an indication of the AFS.The position of this feature tends toward lower wavelengths in AFS-containing mixtures than in many common Fe-bearing samples (Figure 10).The shape and position of the VIS maxima are also diagnostic to the mixtures for NaCl-, gypsum-, nontronite-, and magnesite-bearing samples.However, for basaltic glass and hematite mixtures, these VIS features are nearly identical to those of pure Martian analogs and may easily be mistaken for them.This potential problem is most pronounced in hematite mixtures, where a 50:50 wt% mixture of AFS and hematite looks like hematite with shallow hydration features, without even a slight shift in the VIS bands to distinguish among them.
In Raman spectra, all NaCl-bearing samples have a peak ∼332 cm −1 , likely due to FeCl 4 -, along with a sharp peak ∼1010 cm −1 , attributed to sulfate v 1 in Na-bearing sulfates, like (meta)sideronatrite.For Raman, spectral masking of the amorphous component is most pronounced in the region dominated by sulfate features (∼980-1100 cm −1 ), and is greatest in hematite-and gypsum-containing samples, where the analog material is spectrally dominant, as well as in nontronite-containing samples, where the sulfate is spectrally dominant.
FTIR spectra display the clearest mixing of end-member spectral components.The region <900 cm −1 preserves evidence of metal-O vibrations from all species in most cases, although they may be obscured due to complex sulfate absorptions.The position of features and shape of the band envelope in the sulfate SO 4 v 1 and v 3 absorption region between 900 and 1300 cm −1 in all mixtures with and without NaCl mimics that for pure AFS or AFS+NaCl, respectively; peak positions and depths shift slightly with different analog materials, either due to admixing of spectral components or changes to polyhedron coordination.

Importance of Multitechnique Spectroscopy
It has been apparent throughout this study that each spectral technique provides a unique perspective on each group of samples.The combination of XRD, VNIR, Raman, and FTIR-ATR provides a comprehensive view of the molecular structure and composition of these desiccation products.Often a bonding environment is more apparent in one spectroscopic method over another.VNIR spectra are useful for observing water/ hydroxyl vibrations and iron.Raman and FTIR spectra are correlated, and often observe the same vibrational transitions of a molecule.Yet due to their different selection rules, the intensities of each bonding environment observed often vary  A2; abbreviations are the same as in Figure 7. Wherever possible, reference spectra are form the same samples as in Figure 7; spectra from minerals in Darby Dyarʼs personal collection are included as part of the supplement.Other reference spectra are from RRUFF. between techniques.Using only a single technique provides a narrow view of the nature of the sample, and may lead to misinterpretation.For example, with VNIR alone, one might interpret the nontronite-bearing samples to only contain nontronite.This is especially true if the sample was recorded at decreased spectral resolution and the sample composition was initially completely unknown.When Non2RN and Non2VN are analyzed by Raman and FTIR, it becomes clear that the sample is far more complex, and there are likely multiple phases present.The most extreme examples of this among our samples are hematite-and basaltic glass-bearing mixtures.VNIR and Raman spectra of hematite-bearing samples fail to indicate substantial sulfate; only FTIR records this phase.However, when NaCl is present in the mixture, the main hematite absorptions shift in the FTIR spectra, potentially leading to its misidentification unless these properties are known.FTIR masking is even more pronounced in basaltcontaining samples.For these mixtures, VNIR spectra indicate basaltic glass-dominated mixtures, but Raman and FTIR spectra indicate an almost pure sulfate.The Raman cross section of a phase is critical, and there is a lack of research on interpreting Raman spectra of mixtures relevant to Mars while taking the Raman cross section into account.

Implications
Analog studies create an important bridge between wellcharacterized systems where emplacement conditions are known and poorly characterized systems with unknown emplacement conditions.Similarities among overlapping analyses in analog and target locations permit extrapolation of composition and formation conditions.On Mars, such comparisons are possible only if the spectral libraries used to draw such analogies are comprehensive.This study shows that amorphous materials have a wide variety of potentially diagnostic features, and thus must be included in spectral libraries and modeling.Additionally, this study has implications for Mars Sample Return, highlighting the importance of maintaining stable Martian atmospheric conditions during the return to Earth and subsequent analysis to preserve samples that may contain materials such AFS, which can change hydration state or phase with changing humidity.A2; abbreviations are the same as in Figure 7. Wherever possible, reference spectra are from the same samples as in Figures 7-8; spectra from minerals in Darby Dyarʼs personal collection are included as part of the supplement.Other reference spectra are from RRUFF.The amorphous component of the Martian rocks and soils is unlikely to be a single phase in any location and will vary among locations based on local rock chemistry and transformational history.As noted in the results and discussions above, many features in amorphous samples mimic and/or reproduce similar features in mixtures with crystalline phases.To this end, it is important to keep in mind that spectroscopic features only record the bond energies of molecular groups or atoms, and do not, without interpretation, identify minerals or other compounds.While specific features, alone or in combination, can be indicative of certain minerals, they more truly relate to the local energetic structure around a group of atoms.Thus, when simple or complex brines rapidly desiccate into amorphous or poorly crystalline mixtures, it is possible to have similar energetic environments.For example, the short-range atomic order in AFS is like that in ferricopiapite (Sklute et al. 2015).Thus, previous assignments of spectral features to crystalline phases may need to be reconsidered in light of the results of this study.
Complex mixtures including amorphous or poorly crystalline phases also highlight the nonlinearity of spectral mixing, even using techniques and at wavelengths where linear mixing can often be assumed.For example, intimate mixtures of largegrained solids can linearly be modeled from end members (Ramsey & Christensen 1998;Ye et al. 2021), but when grain sizes are decreased, this assumption is no longer valid (Ye et al. 2021).If amorphous materials were present as coatings or if they were created through extensive mechanical degradation (top-down rather than bottom-up amorphization), they would also be fine grained, thus invalidating the large-grain-size assumption.Furthermore, amorphous solids frequently display inhomogeneous line broadening of vibrational features (Ovchinnikov & Wight 1995), which can complicate identification and unmixing.In the case of VNIR spectra, linear mixing is generally only valid for "checkerboard" mixtures but not intimate mixtures (Hapke 1981).Raman scattering intensity is dependent on the Raman cross section of each phase in a mixture, among other factors (Haskin et al. 1997).Should there be a substantial difference between the Raman cross sections of two phases, one can easily mask the other.Amorphous and poorly crystalline materials can exacerbate this issue as they often display low signal and large fluorescence (e.g., clays; Muñoz-Iglesias et al. 2022), which can confound identification.Finally, it is unknown how nano-or microscale portions of a crystalline phase may appear spectrally and with XRD coated in an amorphous material or trapped in an amorphous matrix.It is counterintuitive, but well known in clay literature, that even X-rays are attenuated by coatings (see Hurst et al. 1997;Schroeder 2018).

Conclusion
We present spectra of desiccation products of ferric sulfate and sodium chloride brines in mixtures with various materials representative of Martian rocks and regolith, with an overall goal of facilitating more robust interpretations of spectra of Martian surface materials.All desiccation products formed an amorphous solid, or a heterogeneous amorphous and crystalline mixed solid.The desiccated products of ferric sulfate and sodium chloride brines have a significant effect on the spectra and overall detectability of relevant Martian analog materials, depending on the technique used.The VIS region is useful for distinguishing samples with AFS, with a diagnostic band between 796 and 1050 nm.This region has potential usefulness for remote sensing identifications, as this Fe 3+ absorption occurs at shorter wavelengths than almost any common Fe 3+ -bearing mineral.However, due to nonlinear mixing (as seen, for example, in our hematite mixtures) VNIR spectroscopy may not detect AFS well on Mars.Raman and MIR-ATR spectra are very complementary when characterizing mixtures with AFS.The missing information due to spectral masking of AFS, or phases coated or altered by sulfate brines, in one technique is often corroborated by the other.Combining Raman and MIR-ATR techniques in situ on Mars would provide a well-rounded perspective of the phases present.Phases identified from the XRD and spectral data provide clues to some of the reactions that may be occurring within the analogs when exposed to deliquescent ferric sulfate brines or ferric sulfate + NaCl brines.Spectral identification of some of these mixtures could provide clues to brine compositions that may have once occurred in Martian materials analyzed by rover instruments.In the MIR, complex spectra indicate a complicated series of distortions of SO 4 tetrahedra.In the NIR, structural H 2 O/OH features dominate the spectra, usually indicative of the representative analog material, if that material has structural H 2 O/OH.Otherwise, the NIR spectra more closely resemble AFS or AFS+NaCl.In the VIS, the bonding environment around Fe 3+ within the desiccated sample is the primary factor.Notably, the VIS absorption ∼800 nm in pure AFS and AFS + NaCl is much lower than that seen in any other common materials and is highly diagnostic.However, this feature position is not preserved in mixtures with other cation absorptions in this region.We provide all spectra and XRD patterns as a supplement to this paper and at doi:10.5281/zenodo.7671724so that other researchers may include these phases and mixtures in their future interpretations of Martian rocks and regolith.Note.Degree of crystallinity was determined prior to baseline subtraction.Crystalline phases listed as wt% of crystalline component after tight baseline subtracted as shown in Figures A1-A10.Amorph% = amorphous percent from Match! DOC calculation.Qtz = quartz, Hem = hematite, Gyp = gypsum, Mag = magnesite, Non = nontronite, Hex = hexahydrite, Meta-sid = metasideronatrite, Na-jar = sodium jarosite, Un-IDed = unidentified peak area.If a phase was selected as matching, but 0% was found, this is denoted with "L." without NaCl, Bas2R, and Bas2V are nearly identical.These two spectra show four Fe 3+ crystal field transition absorptions at 370, 423, 540, and 1050 nm in Bas2R and 370, 423, 540, and 1010 nm (see Table 3) in Bas2V.The broad 6 A 1 -4 T 1 transition in these two spectra deviates from its position at 1040 nm in the initial basaltic glass spectrum.Both spectra have a weak SO 4 ν 1 band at 992 cm −1 , a sharp SO 4 ν 1 band at 1011 cm −1 , and another weak SO 4 ν 1 band at ∼1036 cm −1 , more prominent in the spectra of Bas2RN.At longer wavelengths, both spectra show three overlapping SO 4 ν 3 bands: 1118, 1154, and 1200 cm −1 .
A.1.3.FTIR-ATR FTIR-ATR spectra of the basalt-containing samples (Figure 6(A)), regardless of the mineral present or presence of NaCl, show a shallow H 2 O bending absorption at 1625 cm −1 .This feature is very slightly shifted to higher wavenumbers in the Mag-bearing samples, as discussed above.All NaCl-containing samples, regardless of the mineral present in the mixture, show an identical series of nine features within a broad range between 1300 and 900 cm −1 , which is a complex spectral region in these samples due to SO 4 .The ν 3 asymmetric stretching vibration mode of SO 4 shows strong MIR absorptions that split into several bands due to the distortion of SO 4 tetrahedra (Ling & Wang 2010).The ν 1 symmetric stretching mode also has several absorptions in this wavenumber range (Ling & Wang 2010) caused by SO 4 tetrahedral distortion.These modes contribute to the nine absorptions seen in NaClbearing samples: 1262 (ν 3 ), 1209 (ν 3 ), 1201 (ν 3 ), 1125 (ν 3 ), 1115 (ν 3 ), 1052 (ν 1 ), 1031 (ν 1 ), 993 (ν 1 ), and 972 (ν 1 ) cm −1 .The relative depth of these features does vary slightly between reaction products with different representative analogs, but their positions are consistent.The complex but persistent nine features across all NaCl-containing samples show that the distortion of the SO 4 tetrahedra is primarily dependent on chemical or structural conditions caused by the inclusion of NaCl, which deliquesced into the original brine alongside the ferric sulfate.
Spectra of samples without NaCl show a similar persistence across analog types within 1300-900 cm −1 , but with less spectral complexity and slightly more variation.For samples without NaCl, spectra of gypsum-and magnesite-containing samples show the most variation, and the samples containing basaltic glass, hematite, and nontronite show the least variation.There is an SO 4 ν 3 vibrational mode around ∼1205 cm −1 , another ν 3 mode, or possibly SO 4 ν 1 , around ∼1 cm −1 , and a ν 1 mode around ∼1005 cm −1 .The exact position of these features varies slightly for each analog material and desiccation method.
For our basalt-containing samples without NaCl, these three features are at 1205, 1100, and 1009 cm −1 .
Absorptions below 900 cm −1 vary greatly among spectra of samples containing different representative analogs.Vibrations below 900 cm −1 can include H 2 O librations, OH deformations, SO 4 ν 2 symmetric bending, and ν 4 asymmetric bending modes, and several Fe-O or Fe-OH vibrations.In the basalt-containing samples with NaCl, Bas2VN, and Bas2RN, a weak feature at 666 cm −1 is likely a ν 4 mode of SO 4 .Strong SO 4 ν 4 modes occur at 650 and 593 cm −1 (Ling & Wang 2010), although the former might be an OH-deformation absorption (Cornell & Schwertmann 2003).Three additional weak features occur at 633, 619, and 608 cm −1 , possibly related to Fe-O vibrations, but most likely SO 4 ν 4 modes.The strong feature at 510 cm −1 is most likely an Fe-O out-of-plane bending absorption (Cornell & Schwertmann 2003), but could also possibly be a ν 2 symmetric bending mode of SO 4 (Makreski et al. 2005;Lane 2007) or an O 2− displacement (Lewis & Farmer 1986), and varies slightly between Bas2VN and Bas2RN.This feature is located at 510 cm −1 for the vacuum dehydrated sample, and 508 cm −1 for the low-RH-dehydrated sample.Both Bas2VN and Bas2RN show weak features at 479 and 467 cm −1 , which are either ν 2 modes of SO 4 or Fe-O vibrations.The last major feature of these two spectra is a broad Fe-OH asymmetric stretch absorption centered at ∼405 cm −1 (Blanch et al. 2008).For the two samples without NaCl, Bas2V, and Bas2R, the spectra below 900 cm −1 are simpler.There are two very slight shoulders at ∼655 and ∼625 cm −1 , on the edge of a broad absorption at 588 cm −1 .These three features are likely ν 4 The VNIR spectra of Hem2V, Hem2R, Hem2VN, and Hem2RN are shown in Figure 3(B).
Most features in all four sample spectra are present in either the starting hematite or AFS, evidence that there was no significant reaction between the brines and the hematite.All four spectra show crystal field transition absorptions around 550, 665, and 885 nm, and are identical to those in the starting hematite spectrum.All four sample spectra share a minor H 2 O/OH vibration around 1432 nm.One difference between the spectra of reaction products with and without NaCl is that the H 2 O/OH (ν 2 + ν 3 ) vibrations are more pronounced in the samples containing NaCl.In samples without NaCl in the starting mixture, Hem2R and Hem2V, individual hydration features are not as apparent, and there is a broader H 2 O/OH vibration feature centered around 1950 nm in each.However, there are two distinct absorptions at 1933-1922 nm in both Hem2RN and Hem2VN, with an additional feature at 1985 nm only in Hem2VN.These two samples show an OH combination around 2280, and Hem2VN also shows a slight feature at 2173 nm.These features could be due to Fe-OH vibrations or an S-O bending fundamental (Bishop & Murad 2005;Cloutis et al. 2006), but they are not seen in either the hematite or AFS spectra.Because this feature is not seen in any of the starting material, nor the AFS or AFS+NaCl sample spectra, there was likely some formation of a hydroxyl-bearing product in the reaction products of hematite + FS + NaCl.

A.2.2. Raman
Raman spectra of the hematite-containing samples (Figure 5(B)) are all remarkably similar, sharing the same bands at 410, 494, and 610 cm −1 seen in the spectra of the initial hematite.There are only slight differences in the spectra of the reaction products that contained NaCl and those that did not.Two small bands, an FeCl 4 symmetric stretching vibration at 332 cm −1 and SO 4 ν 1 vibrational mode at 1010 cm −1 , are present in Hem2VN and Hem2RN, while Hem2V and Hem2R instead show a slight broad SO 4 ν 1 band centered at 1040 cm −1 .There are slight similarities between the sample spectra and the spectrum of AFS, despite FS composing 50% of the starting mixture for Hem2R and Hem2V, and 33% of the starting mixture of Hem2VN and Hem2RN.The very slight features at 1010 and ∼1040 cm −1 are the only SO 4 vibrational modes seen in the spectra of our samples, and there is a slight Fe-O band at 468 cm −1 likely related to AFS.The slight broad feature at 812 cm −1 is an H 2 O librational mode (Ling & Wang 2010), likely from surficial H 2 O absorbed onto the hematite.The lack of strong SO 4 features, despite the significant sulfate composition of the reaction products, is due to the large difference between the Raman cross section of hematite and Fe-sulfate.Hematite has a much higher Raman cross section than Fe-sulfate; thus, its features dominate these spectra.

A.2.3. FTIR-ATR
Most of the features seen in the FTIR-ATR spectra of the hematite-containing samples, (Figure 6(B)) are not unique to these particular samples.When NaCl is in the starting mixture, Hem2VN and Hem2RN, features include the shallow H 2 O bending absorption at 1625 cm −1 , nine close ν 1 and ν 3 SO 4 vibrational modes (1262,1209,1201,1125,1115,1052,1031,993, and 974 cm −1 ), and SO 4 ν 4 modes at 650, 666, and 593 cm −1 .Also shown in Hem2VN and Hem2RN are the three slight SO 4 ν 4 mode or Fe-O features at 633, 619, and 608 cm −1 .The 512 cm −1 and 458 cm −1 features could be SO 4 ν 4 and ν 2 modes, respectively, and the feature at 400 cm −1 could be an Fe-O or Fe-OH feature.Alternatively, based on the proximity of the three Fe-O bands seen in the original hematite spectrum at 520, 437, and 391 cm −1 , these three hematite bands may have shifted, either as a result of the FS coating on the hematite, or a surface reaction of the hematite with the acidic brine that caused a distortion of the hematite structure.For the two samples without NaCl, Hem2V and Hem2R, there is again an H 2 O bending absorption at 1625 cm −1 , as well as the familiar SO 4 ν 3 vibrational mode ∼1205 cm −1 , another ν 3 mode, or possibly SO 4 ν 1 , ∼1095 cm −1 , and a ν 1 mode ∼1005 cm −1 .A shoulder at ∼630 cm −1 and an absorption at 585 cm −1 are both interpreted to be ν 4 modes of SO 4 .Features at 528, 442, and 393 cm −1 have nearly identical spectral shapes and are assigned despite slight shifts to Fe-O bands found as those in the initial hematite spectrum.A slight feature at 800 cm −1 , also seen on the initial hematite spectrum, is likely due to absorbed surficial H 2 O, as inferred in the Raman spectra.

A.3.2. Raman
Raman spectra of Gyp2V, Gyp2R, Gyp2VN, and Gyp2RN (Figure 5(C)) show all of the bands of our initial gypsum, with features at 414 (ν 2 ), 493 (ν 2 ), 620 (ν 4 ), 670 (ν 4 ), 1008 (ν 1 ), and 1135 (ν 3 ) cm −1 in all of the Gyp-bearing sample spectra (Berenblut et al. 1971).This implies the persistence of gypsum through the deliquescence and desiccation process, as expected despite visible dissolution of gypsum.As in the hematite- bearing samples (Figure 5(B)), gypsum is a strong Raman scatterer; the lack of strong SO 4 features originating specifically from the AFS, despite its high abundance in the reaction products, is due to the large difference between the Raman cross section of gypsum and AFS.The only shared feature between the spectra of AFS and the experimental samples is the Fe-O band at 468 cm −1 .Several additional bands in experimental sample spectra not seen in the spectra of either gypsum or AFS vary slightly, depending on desiccation method.At lower Raman shifts, Gyp2VN and Gyp2RN show a familiar FeCl 4 feature at 332 cm −1 (Sitze et al. 2001) common to all of our sample spectra that included NaCl in the starting mixture.Two different features at 538 and 520 cm −1 are either Fe-O bands (Lewis & Farmer 1986) or ν 2 modes of SO 4 .At 1115 cm −1 , the two NaCl-containing-sample spectra show a slight SO 4 ν 3 mode, a feature not seen in the spectrum of AFS.Spectra of the two samples without NaCl, Gyp2V, and Gyp2R, show a shoulder at 1040 from SO 4 ν 1 symmetric stretching, but otherwise only resemble the spectrum of our initial gypsum.

A.3.3. FTIR-ATR
As in the Raman spectra of these samples, the FTIR-ATR spectra of Gyp2V, Gyp2R, Gyp2VN, and Gyp2RN all show gypsum features (Figure 6(C)): the H 2 O bending doublet at 1682 and 1620 cm −1 , as well as the two ν 4 vibration mode of SO 4 in gypsum at 668 and 597 cm −1 .The strong and broad SO 4 feature of gypsum causes variations in the general ninefeature sequence between 1300 and 900 cm −1 .Eight are still seen in NaCl-containing spectra Gyp2VN and Gyp2RN, with absorptions at 1262, 1209, 1201, 1115, 1052, 1031, 995, and 973 cm −1 .The ν 3 mode of SO 4 at 1125 cm −1 is not apparent in any of our experimental samples containing gypsum.The equivalent ν 3 feature in our initial gypsum at 1107 cm −1 is only seen in the spectrum of Gyp2V despite the presence of gypsum in all samples; it seems to have blended seamlessly with the common feature at 1115 cm −1 .This may also cause the 1115 cm −1 feature to be slightly deeper relative to the other SO 4 features in this spectral region, compared to the basaltic glass or hematite-containing samples.The spectra of Gyp2VN and Gyp2RN show SO 4 ν 4 modes at 650 cm −1 (Ling & Wang 2010), which could possibly be an OH-deformation absorption (Cornell & Schwertmann 2003).In the spectrum of Gyp2VN, this feature has nearly equivalent depth to the feature at 666 cm −1 , while this feature is almost a shoulder in Gyp2RN.Both Gyp2RN and Gyp2VN show a strong feature at 510 cm −1 that is seen in other samples with NaCl in the starting mixture, and is most likely an Fe-O out-of-plane bending absorption (Cornell & Schwertmann 2003), but again could also possibly be a ν 2 symmetric bending mode of SO 4 (Makreski et al. 2005;Lane 2007) or an O 2− displacement (Lewis & Farmer 1986).A feature at 407 cm −1 is either a ν 2 symmetric bending mode of SO 4 (Ling & Wang 2010), or a Fe-OH asymmetric stretching vibration (Blanch et al. 2008).It is much stronger in Gyp2VN, like the other features.The two samples without NaCl, Gyp2V, and Gyp2R, have much simpler spectra.Both spectra have SO 4 shoulders at ∼1220 and ∼1135 cm −1 from ν 3 vibrations, and 1040 cm −1 from ν 1 vibrations.Both spectra also have an absorption at 458 cm −1 and a shoulder at 420 cm −1 , equivalent to the two ν 2 vibrations of SO 4 in the spectrum of our initial gypsum.VNIR spectra of the four nontronite-containing samples, Non2V, Non2R, Non2VN, and Non2RN exhibit features from both starting material spectra, with a few minor exceptions (Figure 3(D)).All spectra show shallow absorption at ∼650 nm, a small Fe 3+ crystal field transition ∼425-430 nm, an additional shoulder at ∼440 nm, and a slight shoulder around ∼515 nm.The broad Fe 3+ crystal field transition in the spectrum of AFS at 793 nm is shifted to longer wavelengths in our samples and varies with analog material.In the spectra of Non2V and Non2R, its minimum is centered around ∼870 nm.In Non2VN and Non2RN, the minimum is centered around 916-932 nm.The related feature in the initial nontronite has a minimum at 930 nm.All four experimental spectra share an H 2 O/OH vibration doublet at 1432 and 1466 nm.The longerwavelength H 2 O/OH features show more variation.Non2V and Non2R show an absorption at 1935 nm, similar to that seen in AFS.Non2VN and Non2RN both show two features, at 1920 and 1990 nm.These features for the NaCl-containing samples are not seen in any of the starting material.They do match up well, however, with features seen in sideronatrite (Figure 7).Finally, all four spectra show a small OH combination absorption at 2285 nm, possibly due to Fe-OH vibrations or S-O bending fundamental (Bishop & Murad 2005;Cloutis et al. 2006).This feature is seen in the initial nontronite spectrum, but also in the spectrum of sideronatrite, as well as many of our other experimental samples.The spectrum of the initial nontronite does not have many significant peaks in bulk Raman using these wavelengths, but several of the few that are present are seen in the spectra of both samples with and without NaCl in the starting mixture.Spectra of Non2V and Non2R both have peaks at 368, 392, 418, 490, 684, 816, and 888 cm −1 .Overlap between Si-O vibrations in nontronite and hydrated sulfates make interpreting bond assignments in this range difficult.The slight peaks at 368 and 418 cm −1 are likely Si-O-Fe vibrations (Frost & Kloprogge 2000) as they are found in our samples and the initial nontronite.The feature at 392 cm −1 could be an additional Si-O-Fe vibration, or an Fe-O symmetric stretching feature (Sklute et al. 2018b).The broad peak at 490 cm −1 could be an Si-O-Fe peak shifted from the 508 cm −1 peak in the spectrum on nontronite (Frost & Kloprogge 2000), but due to its broad shape, it is more likely a ν 2 vibration of SO 4 (Ling & Wang 2010) in AFS, or a combination of these peaks.The weak peak at ∼684 cm −1 is likely an Al-Fe-OH translation (Frost & Kloprogge 2000).The two features at 816 and 888 cm −1 could be due to H 2 O librations but are likely due to Fe-Fe-OH and Al-OH deformations, respectively (Frost & Kloprogge 2000).At 1010 cm −1 , both Non2V and Non2R show a feature from ν 1 vibrations of SO 4 .A stronger SO 4 ν 1 vibration in Non2V and Non2R varies in position, at 1036 and 1033 cm −1 , respectively.For the spectra of samples with NaCl in the starting mixture, Non2VN and Non2RN, we see several more prominent peaks.Mag2RN and Mag2VN show this feature at 650 nm.In the spectrum of Mag2VN, there is also a slight OH combination feature at 2285 nm, as seen in many of the other sample spectra.For H 2 O/OH absorptions, Mag2R and Mag2V show a (2ν 1 + ν 3 ) vibration at 1444 nm, and a (ν 2 + ν 3 ) vibration at 1940 nm.Mag2RN shows these vibrations at 1468 and 1935 nm, and Mag2VN shows them at 1446 and 1939 nm; they appear sharper for the two spectra desiccated in a vacuum.In the spectra of Mag2R and Mag2RN, these features are slightly broader, possibly due to additional H 2 O vibrations at slightly longer wavelengths.Aside from a potentially related broad H 2 O/OH feature around 1950 nm, none of the features of the initial magnesite powder are seen in our experimental sample spectra, but XRD and other analyses show that it is indeed present.

A.5.2. Raman
In the Raman spectra of Mag2V, Mag2R, Mag2VN, and Mag2RN (Figure 5(E)), most spectral features are seen in both samples with and without NaCl in the starting mixture.All four sample spectra show a feature at 328 cm −1 , which is a CO 3 E g libration (Dufresne et al. 2018;Kim et al. 2021) seen in all magnesite-bearing samples including the initial magnesite.The other two Raman peaks of CO 3 in magnesite, E g (ν 4 ) symmetric bending at 738 cm −1 and A 1g (ν 1 ) symmetric stretching at 1094 cm −1 (Kim et al. 2021), are also present.Despite the vigorous reaction between the carbonate and sulfuric acid during the 10 days of deliquescence, some magnesite either persisted or re-precipitated into the final dehydrated products.
The broad feature at 418 cm −1 seen in each spectrum is either an SO 4 ν 2 vibration or possibly a lattice vibration of a hydrated MgCl 2 (Shi et al. 2020).The feature at 496 cm −1 could either be a ν 2 mode of SO 4 (Ling & Wang 2010) or an Mg-O stretching vibration (Shi et al. 2020).The peak at 626 cm −1 is a ν 4 vibration of SO 4 and is only prominent in Mag2VN and Mag2RN.The broad feature seen in all spectra at ∼810 cm −1 is an H 2 O libration.The sharp feature at 1009 cm −1 and the weak feature seen at ∼1036 cm −1 in each spectrum is due to ν 1 vibrations of SO 4 , although the position of the latter of these varies within ±6 cm −1 .Mag2RN has an additional sharp feature at 983 cm −1 , also a ν 1 vibration of SO 4 .The location of this ν 1 peak is inconsistent with any published Fe-sulfate but is comparable to the ν 1 feature of Na-sulfate (Qiu et al. 2019).

A.5.3. FTIR-ATR
The FTIR-ATR spectra of the magnesite-bearing samples are shown in Figure 6(E).H 2 O bending features in the FTIR-ATR are broad; they are shifted closer to 1635 cm −1 for Mag2R and Mag2V and shifted to 1645 cm −1 for Mag2RN.In the spectrum of Mag2VN, this feature is a doublet at 1652 and 1625 cm −1 .The ν 3 vibration of CO 3 has a broad absorption at ∼1450 cm −1 and a shoulder at 1415 cm −1 (White 1971;Canterford et al. 1984).Spectra vary in the region from 1300-900 cm −1 depending on desiccation method.Mag2V and Mag2VR show a minimum at 1125 cm −1 but that peak is only a shoulder for the low-RH-dehydrated samples.Conversely, Mag2R and Mag2RN show minima at 1080 and 1095 cm −1 , respectively.A shoulder at 1095 cm −1 in the vacuum-desiccated samples

Figure 1 .
Figure 1.Reaction products for the compositions and grain size indicated.Photographs taken at day 5 of dehydration for identical samples to those used in this study.Euhedral salt crystals are visible as raised three-pronged ridges and crosses embedded in the sample pucks.

Figure 2 .
Figure 2. Uncorrected VNIR (A), corrected Raman (B), and uncorrected FTIR (C) spectra for all three grain sizes of basaltic glass mixtures with NaCl (where there was no observable dissolution) and gypsum mixtures (where there was extensive dissolution).Data are shown not offset (solid lines) and offset (dashed lines) for comparison.

Figure 4 .
Figure 4. VNIR spectra of each sample, along with AFS and AFS+NaCl, with plot bounds focused on the VIS maxima.Dashed lines are to aid the eye in visualizing changes in the VIS maxima.

Figures 7 -
Figures 7-9 summarize the spectra of our experimental products in comparison with their individual components and other relevant phases.Overall, most of the features seen in our sample spectra are related to Fe-O, SO 4 , or H 2 O/OH bonds also expressed in the spectra of AFS or AFS+NaCl.Each resulting product in this study formed a heterogeneous solid containing a significant amorphous component.In all cases, it is likely that this amorphous component includes Fe 2 (SO 4 ) 3 • 5H 2 O, as our desiccation products, particularly those without NaCl in the mixture, have similar Raman and MIR spectral features as the amorphous pentahydrous ferric sulfate in Ling & Wang (2010).For example, absorptions in the FTIR-ATR spectra of Bas2V and Bas2R match all three H 2 O and seven SO 4 vibrational modes of amorphous pentahydrous ferric sulfate (Ling & Wang 2010).Ling & Wang (2010) also showed that AFSs can hold between 5 and 11 structural waters, which correlates with the position of its SO 4 ν 1 Raman peak, with lower wavenumbers indicating an increase in the number of structural waters.All of our desiccated products without NaCl show the SO 4 ν 1 peak between 1030 and 1040 cm −1 , which would imply a varying number of structural waters between samples.We cannot extrapolate water contents in our samples using theLing & Wang (2010) hydration analysis because the hydration feature is shifted relative to those reports.Qualitatively, vacuumdesiccated samples have less water than RH-desiccated samples.For example, the SO 4 v 1 is at 1036 cm −1 for Hem2R and 1040 cm −1 for Hem2V.In addition, the AFS component of all mixtures appeared to have less water than the pure AFS,

Figure 7 .
Figure 7. VNIR spectra of vacuum dehydrated products plotted with starting materials and library spectra for comparison.Spectra are offset by the amount indicated.Ferricop = ferricopiapite, Metasid = metasideronatrite, Amaran = amarantite, Botryo = botryogen, Rhomb = rhomboclase, Najar = Najarosite.All spectral sample names are included in TableA2.Most library spectra are from minerals in Darby Dyarʼs personal collection, and those data are also included as part of the supplement.Other reference spectra are from the PDS.

Figure 8 .
Figure 8. Raman spectra of vacuum dehydrated products plotted with starting analog materials and library spectra for comparison.Spectra are normalized and offset for clarity.All spectral sample names are included TableA2; abbreviations are the same as in Figure7.Wherever possible, reference spectra are form the same samples as in Figure7; spectra from minerals in Darby Dyarʼs personal collection are included as part of the supplement.Other reference spectra are from RRUFF.

Figure 9 .
Figure 9. FTIR spectra of vacuum dehydrated products plotted with starting analog materials and library spectra for comparison.Spectra are scaled and offset for clarity.All spectral sample names are included TableA2; abbreviations are the same as in Figure7.Wherever possible, reference spectra are from the same samples as in Figures7-8; spectra from minerals in Darby Dyarʼs personal collection are included as part of the supplement.Other reference spectra are from RRUFF.

Figure 10 .
Figure 10.Comparison of offset continuum removed VNIR spectra highlighting the shorter wavelength Fe 3+ absorption in AFS and AFS+NaCl.
Figure 5(A) shows normalized Raman spectra of Bas2V, Bas2R, Bas2VN, and Bas2RN.Spectra of reaction products that included NaCl are more complicated and show far more bands than those that did not include NaCl.At lower wavenumbers, Bas2V and Bas2R have a weak band around ∼332 cm −1 and a broad Fe-O band or SO 4 ν 2 symmetric

Figure A1 .
Figure A1.XRD patterns of FS + basaltic glass reaction products showing backgrounds used during phase matching along with reference spectra from the Match!COD library.

Figure A2 .
Figure A2.XRD patterns of FS + basaltic glass + NaCl reaction products showing backgrounds used during phase matching along with reference spectra from the Match!COD library.

Figure A3 .
Figure A3.XRD patterns of FS + hematite reaction products showing backgrounds used during phase matching along with reference spectra from the Match!COD library.

A
.3.Gypsum-containing samples A.3.1.VNIR VNIR spectra of Gyp2V, Gyp2R, Gyp2VN, and Gyp2RN (Figure 3(C)) share gypsum's characteristic bands due to H 2 O vibrations, with the largest deviations only ±5 nm: a triplet at 1446, 1490, and 1538 nm, and another at 2178, 2217, and 2268 nm.The 2276 H 2 O vibration in Gyp2VN is shifted slightly toward longer wavelengths, possibly indicating the influence of an OH combination feature, as seen in the spectra of several other NaCl-bearing samples.All four experimental spectra show two H 2 O/OH (ν 2 + ν 3 ) vibrations around 1945 and 1985 nm, an absorption at ∼1750 nm due to combinations of SO 4 2-and OH − or H 2 O vibrations (Bishop & Murad 2005; Cloutis et al. 2006), and a shallow Fe triplet at ∼1180, 1206, and 1231 nm.Gyp2R shows Fe 3+ crystal field transition absorptions at 370, 425, 548, and 802 nm.Gyp2V shows these transition absorptions at 375, 425, 550, and 796 nm.Gyp2RN and Gyp2VN both exhibit crystal field transitions at ∼455 and ∼900 nm likely related to the presence of NaCl in the starting mixture.

Figure A4 .
Figure A4.XRD patterns of FS + hematite + NaCl reaction products showing backgrounds used during phase matching along with reference spectra from the Match!COD library.

Figure A5 .
Figure A5.XRD patterns of FS + gypsum reaction products showing backgrounds used during phase matching along with reference spectra from the Match!COD library.
Figure 5(D)  shows the Raman spectra of Non2V, Non2R, Non2VN, and Non2RN along with the spectra for our original nontronite sample and AFS.The spectrum of the initial nontronite does not have many significant peaks in bulk Raman using these wavelengths, but several of the few that are present are seen in the spectra of both samples with and without NaCl in the starting mixture.Spectra of Non2V and Non2R both have peaks at368, 392, 418, 490, 684, 816, and 888 cm −1 .Overlap between Si-O vibrations in nontronite and hydrated sulfates make interpreting bond assignments in this range difficult.The slight peaks at 368 and 418 cm −1 are likely Si-O-Fe vibrations(Frost & Kloprogge 2000) as they are found in our samples and the initial nontronite.The feature at 392 cm −1 could be an additional Si-O-Fe vibration, or an Fe-O symmetric stretching feature(Sklute et al. 2018b).The broad peak at 490 cm −1 could be an Si-O-Fe peak shifted from the 508 cm −1 peak in the spectrum on nontronite(Frost & Kloprogge 2000), but due to its broad shape, it is more likely a ν 2 vibration of SO 4(Ling & Wang 2010) in AFS, or a combination of these peaks.The weak peak at ∼684 cm −1 is likely an Al-Fe-OH translation(Frost & Kloprogge 2000).The two features at 816 and 888 cm −1 could be due to H 2 O librations but are likely due to Fe-Fe-OH and Al-OH deformations, respectively(Frost & Kloprogge 2000).At 1010 cm −1 , both Non2V and Non2R show a feature from ν 1 vibrations of SO 4 .A stronger SO 4 ν 1 vibration in Non2V and Non2R varies in position, at 1036 and 1033 cm −1 , respectively.For the spectra of samples with NaCl in the starting mixture, Non2VN and Non2RN, we see several more prominent peaks.

Figure A6 .
Figure A6.XRD patterns of FS + gypsum + NaCl reaction products showing backgrounds used during phase matching along with reference spectra from the Match!COD library.

Figure A8 .
Figure A8.XRD patterns of FS + nontronite + NaCl reaction products showing backgrounds used during phase matching along with reference spectra from the Match!COD library.

Figure A9 .
Figure A9.XRD patterns of FS + magnesite reaction products showing backgrounds used during phase matching along with reference spectra from the Match!COD library.

Figure A10 .
Figure A10.XRD patterns of FS + magnesite + NaCl reaction products showing backgrounds used during phase matching along with reference spectra from the Match!COD library.

Figure A12 .
Figure A12.VNIR spectra of FS + hematite (+ NaCl) reaction products showing of all three grain sizes along with starting hematite of all three grain sizes.Spectra are offset by amount indicated.Darker colors in each sequence indicate larger grain sizes.

Figure A13 .
Figure A13.VNIR spectra of FS + gypsum (+ NaCl) reaction products showing all three grain sizes along with starting gypsum of all three grain sizes.Spectra are offset by amount indicated.Darker colors in each sequence indicate larger grain sizes.

Figure A14 .
Figure A14.VNIR spectra of FS + nontronite (+ NaCl) reaction products showing all three grain sizes along with starting nontronite of all three grain sizes.Spectra are offset by amount indicated.Darker colors in each sequence indicate larger grain sizes.

Figure A15 .
Figure A15.VNIR spectra of FS + magnesite (+ NaCl) reaction products showing all three grain sizes along with starting magnesite of all three grain sizes.Spectra are offset by amount indicated.Darker colors in each sequence indicate larger grain sizes.

Figure A16 .
Figure A16.FTIR (left) and Raman (right) spectra of all three grain sizes of FS + basaltic glass (+ NaCl) reaction products.

Table 2
Crystalline XRD Phases Identified by Match!Software from Desiccation Products after Baseline Removal (Including Amorphous Swell) Note.Percentages listed in Table A1 and plotted in Figures A1-A10.

Table 4
Raman Features and Assignments a

Table 5
Mid-IR Features and Assignments a (Sitze et al. 20012001).Additionally, there is a weaker feature at 388 cm −1 in the Raman spectra of AFS+NaCl and several other NaClcontaining samples.This feature is also consistent with the presence of FeCl 4 -tetrahedra(Sitze et al. 2001) but overlaps

Table A1
Phases Identified by XRD for Each Experimental Sample