 REANALYSIS OF WET CHEMISTRY LABORATORY DATA WITH IMPLICATIONS FOR PARENT SALT ASSEMBLAGES AT THE PHOENIX SITE.  J. D. Toner 1, D. C. Catling 1, and B. Light 2 1 University of Washington, Dept. Earth & Space Sciences, Seattle, WA 98195, USA, 2 Polar Science Center, Applied Physics Laboratory, University of Washington, Seattle, Washington, USA. (e-mail: toner2@uw.edu)   Introduction: We have investigated data from the Wet Chemistry Laboratory (WCL) experiment aboard the Mars Phoenix Lander using some improvements to original analyses, including better noise filtering [1,2,3].  Our goal is to characterize errors and trends in the WCL data and ultimately improve our understanding of possible soluble salt compositions and parent salt assemblages on Mars.  Previously, the WCL data has been analyzed using Fourier filtering to remove high frequencies [2].  We use Kalman smoothing, which can estimate both the true signal state and associated error without removing data [4]. WCL Analysis Methods: Our analysis focuses on data from Rosy Red because it is the least noisy of the three WCL soil analyses.  We analyzed temperature and Ion Selective Electrode (ISE) data using the methods described in [2,3] with modifications as follows: Kalman Smoothing. We used the KFAS package [4] in the statistical program R.  KFAS was chosen over other programs because it can account for missing data points, which are common in the WCL data.  Furthermore, KFAS can estimate the important 'process noise' and 'signal noise' parameters. Calibration Data. Errors associated with the WCL analysis include errors among the five Earth-based ISE calibrations.  To better characterize these errors, we reanalyzed calibration data to determine the error in the ISE calibration slopes.  The final ISE calibration slope we use is the weighted mean of the five Earth-based calibration slopes with associated unbiased standard deviations.  Errors in the two-point pH sensor calibrations are determined by propagating errors in the individual pH measurements to the slope.  Ionic Strength. The conductivity sensor, used for measuring ionic strength failed during the Rosy Red analysis.  We calculate ionic strength by assuming an initial ionic strength and iteratively refine this value by using the resulting concentrations to calculate a new ionic strength until convergence. Debye-Hückel Model. The Debye-Hückel ion-association model, used to determine activity coefficients from ionic strength, is temperature dependent.  We incorporate this temperature dependence into our WCL data reduction. Ion-Pairs. The WCL ISE sensors measure the activity of unpaired ions, but MgSO4 0 and CaSO4 0 ion-pairs can be significant [3].  We determine the concentration of ion-pairs in PHREEQC using the phreeqc.dat aqueous database by adding ions until the unpaired species equals that determined in the WCL analysis. Calibration/Leaching Solution Ions. We subtract ion concentrations present in the WCL calibration solution from the final results.  This correction, which was previously neglected, has a significant effect on low concentration ions such as Cl and NH4. WCL Analysis Results: Ion concentrations resulting from analysis of the Rosy Red WCL experiment using the methods outlined above are presented in Table 1 (including ion-pairs).  Compared to values reported in [2,3], Na, K, and NH4 values are similar, Ca, Cl, ClO4, and SO4 values are significantly lower, and Mg is significantly higher.  In general, our analysis has smaller random errors.  Over the duration of the Rosy Red WCL analysis, all ion concentrations, with the exception of Cl due to leakage of BaCl2 reagent, remain within error bars of the values in Table 1 (Fig. 1).  This suggests that the WCL solution rapidly attained equilibrium and that sparingly soluble salts are of minor importance.  Fig. 1. Concentrations and temperature in Rosy Red.  Vertical lines indicate calibrant and sample addition;  bold dots indicate calibration and analysis intervals. pH, pCO2, and Alkalinity: The pH measured in Rosy Red (Fig. 2) is a function of solution alkalinity and aqueous CO2, which is influenced by CO2(g) in the WCL headspace.  Using PHREEQC, we find that the pH of Rosy Red before sample addition is consistent with a headspace pCO2 of ~8.4 mbar, and that drawer open/close events have a minor effect on 0 510 15 20 25 0.001 0.01 0.1 110 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 Tem per at ure ( °C ) C on ce ntr at io n ( mM ) Local Mean Solar Time Na⁺ K⁺ NH₄⁺ Ca²⁺ Mg²⁺ Cl⁻ ClO₄⁻ Temp Table 1. Total Solute Concentrations in  Rosy Red (mM) Ca 0.16±0.06 Mg 3.42±1.20 Na 1.40±0.41 K 0.33±0.08 NH4 0.07±0.03 Cl 0.39±0.13 SO4 3.87±1.16 ClO4 1.90±0.62 the balance of CO2 in the WCL cell.  Upon sample addition, the pH rapidly increases to ~7.7 (Fig. 2), likely due to CO2 consumption from carbonate dissolution.  The apparently highly soluble nature of this carbonate suggests a hydrated MgCO3 mineral, not calcite.  This is consistent with a carbonate that decomposed at low temperature in the Thermal Evolved Gas Analysis (TEGA) experiment [5].  Assuming that minimal exchange with headspace CO2(g) occurred during the initially rapid increase in pH, we calculate in PHREEQC that the WCL solution alkalinity is ~0.7 meq.   After sample addition, the pH of Rosy Red steadily decreases from 7.7 to 7.1 over several hours.  Although calcite is thought to be present in Phoenix soils [6], this decrease in pH indicates that pH is not controlled by calcite dissolution because calcite dissolution should increase pH.  Instead, a mass transfer model suggests that the pH decrease is consistent with a CO2(g) flux from the WCL headspace into the solution.    Fig. 2. Rosy Red pH after signal processing. Possible WCL Solution Compositions: Soluble ions in the Phoenix soil imply parent salt compositions [1,2,3,6,7].  The relative proportion of ions strongly influences what salts precipitate from solution during freezing or evaporation.  Due to uncertainties in the ion concentrations inferred from WCL (Table 1), we examine the spread of all possible solution compositions to determine variations in the predicted parent salts.  Seven ions were measured during the Rosy Red analysis (Ca, Mg, Na, K, Cl, SO4, and ClO4).  If we assume that there are five possible values for each ion (the measured value, ±σ, and ±σ/2), then there are 57 or 78125 unique ionic combinations possible.  For all of the 78125 possible solutions we calculate (1) alkalinity from charge balance, (2) pH upon addition to the WCL calibration solution, and (3) ion-pair corrections in PHREEQC.  From these possible solutions, we use the following two criteria to determine if a solution is plausible: (1) Alkalinity > 0 and (2) pH=7.72±0.27 (pH measured in the analysis interval).  Based on these criteria, only ~1000 out of the 78125 possible solutions qualify as plausible.  'Chemical Divides'. We evaluate the parent salts of plausible Rosy Red solutions using a 'Chemical Divide' model [8] that is based on equilibria in FREZCHEM (Fig. 3).  In this model, we evaluate the evolution of solution compositions during freezing by considering the relative solubility of salts and ratios of cations to anions.  For example, if Ca<Alkalinity, then precipitation of calcite will consume Ca early during freezing and the solution will evolve towards a Ca-depleted composition. Using the model outlined in Fig. 3, we find that CaCO3, hydrated MgCO3, KClO4, MgSO4.11H2O, and NaClO4.2H2O precipitate from all solutions during freezing.  Other salts that are sometimes precipitated include NaCl.2H2O and Na2SO4.10H2O.  These salts have not been previously considered as possible parent salts.  Our model predicts that Ca(ClO4)2.6H2O does not form during freezing, although we are mindful that the presence of this salt has been inferred from a transient ISE signal [9]. Conclusions: Several improvements have been made to previous analyses of WCL data, including an estimation of alkalinity.  We conclude that CaCO3, hydrated MgCO3, KClO4, MgSO4.11H2O, and NaClO4.2H2O are very likely present in the Phoenix soil.  Other possible salts are Na2SO4.10H2O, MgCl2.12H2O, NaCl.2H2O, and Mg(ClO4)2.8H2O. Acknowledgements: Funding from NASA Mars Data Analysis grant #NNX10AN66G. References: [1] Hecht M. H. et al. (2009) Science, 325, 5936, 64-67.  [2] Kounaves S. P. et al. (2010) J. Geophys. Res., 115, E00E10.  [3] Kounaves S. P. et al. (2010) Geophys. Res. Lett., 37, L09201.  [4] Koopman S. J. and Durbin J. (2001) J. American Stat. A., 92, 1630-38. [5] Sutter B. et al. (2012) Icarus, 218, 290-296. [6] Boynton W. V. et al. (2009) Science, 325, 61. [7] Marion G. M. et al. (2010) Icarus, 207, 675-686. [8] Drever J. I. (1982) The Geochemistry of Natural Waters. [9] Kounaves, S. P. et al. (2012), AGU Fall Meeting. 5 67 810:00 12:00 14:00 16:00 pH Local Mean Solar Time Sample Addition Table 2. % of plausible solutions that precipitate a given salt phase. Salt Phase % CaCO3 100 Hydr. MgCO3 100 KClO4 100 MgSO4.11H2O 100 NaClO4.2H2O 100 MgCl2.12H2O 76.3 Mg(ClO4)2.8H2O 57.4 Na2SO4.10H2O 23.6 NaCl.2H2O 18.9  WCL Solution K < ClO4 and Ca <Alkalinity? Mg > Alkalinity?   Mg > SO4?     Mg < Cl?    Mg > Cl?   Mg-Na-ClO4 Na-Cl-ClO4 Mg-Na-Cl-SO4ClO4   Mg-Na-ClClO4   Na-Cl-SO4ClO4   Mg-Na-Cl-SO4ClO4-HCO3   Mg > SO4? Mg < SO4? Fig. 3. Chemical Divide model of freezing brines.   
