Sobron Et Al., 2012

This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and education use, including for instruction at the authors institution and sharing with colleagues. Other uses, including reproduction and distribution, or selling or licensing copies, or posting to personal, institutional or third party websites are prohibited. In most cases authors are permitted to post their version of the article (e.g. in Word or Tex form) to their personal website or institutional repository. Authors requiring further information regarding Elsevier’s archiving and manuscript policies are encouraged to visit: http://www.elsevier.com/copyright Author's personal copy

Spectrochimica Acta Part B 68 (2012) 1–16

Contents lists available at SciVerse ScienceDirect

Spectrochimica Acta Part B

journal homepage: www.elsevier.com/locate/sab

Extraction of compositional and hydration information of sulfates from laser-induced plasma spectra recorded under Mars atmospheric conditions — Implications for ChemCam investigations on Curiosity rover

Pablo Sobron a,⁎, Alian Wang a, Francisco Sobron b a Department of Earth and Planetary Sciences and McDonnell Center for the Space Sciences, Washington University, St. Louis, MO 63130, USA b Unidad Asociada UVa-CSIC a traves del Centro de Astrobiología, Parque Tecnológico de Boecillo, Parcela 203, Boecillo (Valladolid), 47151, Spain article info abstract

Article history: Given the volume of spectral data required for providing accurate compositional information and thereby in- Received 13 July 2011 sight in mineralogy and petrology from laser-induced breakdown spectroscopy (LIBS) measurements, fast Accepted 2 January 2012 data processing tools are a must. This is particularly true during the tactical operations of rover-based plan- Available online 21 January 2012 etary exploration missions such as the Mars Science Laboratory rover, Curiosity, which will carry a remote LIBS spectrometer in its science payload. We have developed: an automated fast pre-processing sequence Keywords: of algorithms for converting a series of LIBS spectra (typically 125) recorded from a single target into a reli- LIBS Martian sulfates able SNR-enhanced spectrum; a dedicated routine to quantify its spectral features; and a set of calibration Composition curves using standard hydrous and multi-cation sulfates. These calibration curves allow deriving the elemen- Hydration state tal compositions and the degrees of hydration of various hydrous sulfates, one of the two major types of sec- Calibration curve ondary minerals found on Mars. Our quantitative tools are built upon calibration-curve modeling, through the correlation of the elemental concentrations and the peak areas of the atomic emission lines observed in the LIBS spectra of standard samples. At present, we can derive the elemental concentrations of K, Na, Ca, Mg, Fe, Al, S, O, and H in sulfates, as well as the hydration degrees of Ca- and Mg-sulfates, from LIBS spectra obtained in both Earth atmosphere and Mars atmospheric conditions in a Planetary Environment and Analysis Chamber (PEACh). In addition, structural information can be potentially obtained for various Fe- sulfates. © 2012 Elsevier B.V. All rights reserved.

1. Introduction detect and to quantify the concentration of major, minor, and often trace elements present in a sampling target. The accuracy and preci- Laser-induced breakdown spectroscopy (LIBS) is recognized as a sion of the quantification of elemental concentrations based on LIBS powerful tool for detailed geochemical investigations of rocks and measurements are influenced mainly by two factors: the matrix ef- regolith and for exploratory survey in future landed missions on plan- fects within a specific sample and the fluctuations in experimental etary bodies including Mars [1], the Moon [2], and Venus [3,4].A conditions. A good explanation of the matrix effects in LIBS was stand-off LIBS spectrometer is part of the ChemCam instrument [5], given in Cremers and Radziemski [6]. They grouped the matrix effect included in the scientific payload of NASA's Mars Science Laboratory into two kinds: physical and chemical matrix effects. On one hand, (MSL) rover (named Curiosity), that was launched in late 2011. physical matrix effects, related to the physical properties of the target ChemCam will determine the chemical composition of rocks and reg- (e.g., grain size, surface roughness, absorptivity and thermal conduc- olith on Mars at distances ranging from 1 to 7 m. tivity). Physical matrix effects complicate quantitative analysis with The LIBS technique is particularly well suited for planetary explo- LIBS by causing uncontrolled random fluctuations in the emission ration, as it is sensitive to all rock-forming elements, as well as H, C, N, from the plasma. On the other hand, chemical matrix effects, related O, S, P, Fe, and Cl, relevant to the search for habitable environments to the elemental and molecular compositions of the sample. They and for traces of past and present signs of life on Mars. In LIBS, the can result in higher emission from easily ionized elements existing wavelength and the intensity of atomic emission lines are used to in the matrix, i.e., the same concentration of an element in different matrices will result in emission lines with different intensities, thus affecting the accuracy of compositional quantification using LIBS ⁎ Corresponding author at: Space Science & Technology, Canadian Space Agency, 6767 data. As for the fluctuations in experimental conditions, they include Rte. de l'Aéroport, St. Hubert, QC, Canada J3Y 8Y9. Tel.: +1 450 926 5847; fax: +1 450 926 4766. pulse-to-pulse variations in the properties of laser beam (pulse fre- E-mail addresses: [email protected], [email protected] (P. Sobron). quency, pulse width, and energy density), variations in laser-to-

2 P. Sobron et al. / Spectrochimica Acta Part B 68 (2012) 1–16 sample optical path (e.g., accidental scatters), and variations in and electron density of the plasma created by the laser pulse. In CF- sample-to-collection optics optical path (including atmospheric ef- LIBS, local thermodynamic equilibrium (LTE) is assumed. Note that fects). The latter two can be considered physical matrix effects as LTE and other conditions relevant to CF-LIBS have been verified [16]. well, since they are mainly due to random effects in the laser ablation In this paper we describe a methodology for general automated phase and contribute strongly to the standard deviation of LIBS mea- pre-processing and quantitative analysis of LIBS spectral data and re- surement by modifying the collection angles. port its step-by-step application to construct a set of calibration All the factors mentioned above compromise the reproducibility curves for common elements in sulfate matrix under Mars and of LIBS measurements and the accuracy of quantitative calculations Earth atmospheric conditions. We also show the application of the based on LIBS data. To compensate for matrix effects and variations calibration curves to characterize two natural sulfate-bearing sam- in experimental conditions, normalization procedures are commonly ples. At the time this manuscript was prepared, matrix-specific cali- employed [1,6,7]. In most normalization approaches, the intensity of a bration curves for additional matrices relevant to Mars (basalts, measured peak is normalized to that of another peak in the same phyllosilicates, carbonates, etc.) are under construction. spectrum that is known to reflect spectral changes due to matrix effects [1,6]. However, in applications where such internal standard does not exist (e.g., most mineralogical analyses of geological sam- 2. Experimental ples, such as ChemCam measurements on Mars), normalization by the total emission intensity of the spectrum is preferred (e.g., [7]). 2.1. Sulfate standard samples for the construction of calibration curves In the preceding paragraphs we have emphasized those factors that affect quantitative analyses of targets. By quantitative analyses Sulfates were selected for the development of this first set of cali- we mean a series of mathematical operations performed on a LIBS bration curves because they are one of the two major types of second- spectrum that aim to reconstruct the stoichiometry of a target inter- ary minerals found on Mars that may provide potentially habitable rogated by the laser, i.e., to determine the elemental abundances in environments due to their association with ancient aqueous environ- a sample. Most common quantitative analysis methods use calibra- ments in which life might have thrived. In this work we used pure hy- tion procedures to generate calibration curves for one or several ele- drous sulfates with multiple cations synthesized in previous studies ments from the spectra of standard samples in a specific matrix, each by the authors [17,18], and several commercial-grade sulfates. These sample containing precisely known concentrations of the elements of sulfates are used as standards to establish the calibration curves interest [8]. Different calibration curves are often constructed for dif- based on their LIBS spectra obtained under Mars atmospheric condi- ferent matrices. The concentrations of the elements in unknown sam- tions. All of the samples, including their provenance and the atomic ples are determined from the calibration curves by a quantitative fractions of the elements relevant for this work calculated from X- measure of certain spectral features in their spectra. An alternative ray fluorescence measurements (described below), are listed in approach to the quantitative analysis of LIBS spectra is multivariate Table 1. A manual pellet press was used to apply ~4 tons of pressure analysis (MVA), i.e., the simultaneous analysis of more than one sta- to the powdered samples, which were placed in compressible alumi- tistical variable. MVA is widely used and seems to be preferred for num cups (Chemplex Industries, Inc. PelletCups) to form solid bri- the study of materials relevant to Mars exploration [9,10]. The use quettes. The top flat surfaces of the pellets were used for LIBS of artificial neural networks (ANN) has also been suggested as a pow- measurements. erful quantitative approach in LIBS [11–13]. Calibration-free (CF) Samples for this study were analyzed for major and minor ele- techniques have been proposed for processing LIBS spectra without ments in the XRF laboratory at Washington University in St. Louis using matrix-similar standards [14,15]. CF-LIBS essentially derives (under the direction of Rex Couture) using novel procedures (Cou- the composition of the sample from calculations of the temperature ture, 2011; personal communication). The elemental abundances

Table 1 Samples used to construct the calibration curves.

# Name Empirical formula Origin XRF-calculated atomic fractions of relevant elements

KFeCaMgNaAlHSO

3+ 1 Natrojarosite NaFe 3(SO4)2(OH)6 Excalibur Mineral 3.6E-03 0.18 9.7E-04 3.7E-04 0.05 8.9E-04 3.8E-03 0.12 0.65 Corporation

2 Anhydrite CaSO4 Fisher Scientiﬁc 0.17 0.01 0.16 0.66 3+ 3 Ferric sulfate pentahydrate Fe 2(SO4)3·5H2O Synthetic 0.06 0.36 0.08 0.50

4 Kieserite MgSO4·H2O Fisher Scientiﬁc 0.11 0.23 0.11 0.55

5 Thenardite Na2SO4 Fisher Scientiﬁc 0.28 1.1E-03 0.14

6 Potassium sulfate K2SO4 Fisher Scientiﬁc 0.23 0.12 0.12 0.53 3+ 3 7 Ferricopiapite Fe 0.66Fe Synthetic 0.05 0.40 0.07 0.48 + 2(SO4)6(OH)2·20H2O 3+ 8 Magnesiocopiapite MgFe 4(SO4)6(OH)2·20H2O Synthetic 0.04 0.01 0.42 0.06 0.47 3+ 3 9 Aluminocopiapite Al 0.66Fe Synthetic 0.04 0.01 0.41 0.06 0.44 + 2(SO4)6(OH)2·20H2O

10 Gypsum CaSO4·2H2O Fisher Scientiﬁc 0.09 0.32 0.09 0.51

11 Sodium bisulfate NaHSO4·H2O Fisher Scientiﬁc 0.10 0.32 0.09 0.49 monohydrate

12 Alum-(K) KAl(SO4)2·12(H2O) Marysville, UT 0.02 0.03 0.02 0.01 0.12 0.29

13 Bassanite 2(Ca)2(SO4)·H2O Fisher Scientiﬁc 0.14 0.11 0.14 0.61

14 Hexahydrite MgSO4·6H2O Synthetic 0.04 0.52 0.04 0.41

15 Starkeyite MgSO4·4H2O Synthetic 0.06 0.44 0.06 0.41

16 Kieserite MgSO4·H2O Fisher Scientiﬁc 0.11 0.23 0.11 0.55

17 Magnesium sulfate anhydrate MgSO4 Fisher Scientiﬁc 0.16 0.01 0.16 0.66 3+ 18 Rhomboclase HFe (SO4)2·4H2O Synthetic 0.04 0.47 0.05 0.45 3+ 19 Kornelite Fe 2(SO4)3·7H2O Synthetic 0.05 0.38 0.08 0.50 2+ 20 Rozenite Fe (SO4)·4H2O Synthetic 0.04 0.47 0.05 0.45 Author's personal copy

P. Sobron et al. / Spectrochimica Acta Part B 68 (2012) 1–16 3 calculated on the basis of these XRF measurements are reported in baseline removal, spectral normalization, outlier discarding, spectra Table 1. averaging, and resolution enhancement.

2.2. Laboratory LIBS setup in Earth and Mars atmospheric conditions 3.1.1. Baseline removal LIBS spectra feature background signals essentially caused by Our laboratory LIBS setup allows LIBS to conduct measurements bremsstrahlung and recombination processes [21]. Background con- under dual atmospheric conditions: Earth and Mars [19]. The labora- tributes as a form of baseline, although LIBS spectra acquired with a tory features two optical configurations that share the same excita- gated detector such as the one used in this work show relatively tion laser (an actively Q-switched Nd:YAG Minilite-II laser from weak background. In this work we have used the baseline removal Continuum: 7 ns pulse width, max 45 mJ/pulse, 1064 nm wave- routines developed by Sobron et al. [22] to generate a baseline by lin- length) and the same spectrograph (Andor Mechelle 5000) equipped ear interpolation from a certain number of spectral points, which is with an intensified CCD detector (Andor iStar 712). Two identical sets then subtracted point by point from the spectrum. Although original- of front optics sit in the Earth and Mars sides of the laboratory, and ly developed for baseline removal purposes in Raman spectra, these are selectable by a flip mirror. The front optics for LIBS measurements routines generate excellent baselines for the LIBS spectra recorded in Mars conditions was installed inside a recently developed Plane- in this work. However, the implementation of the baseline removal tary Environment and Analysis Chamber (PEACh) [19,20]. The front routines resulted in no significant improvement in the subsequent optics includes a dichroic mirror to direct the 1064 nm laser pulse quantitative calculations performed on the spectra. This issue has into a 5× microscope objective lens assembly (OFR LMH-5X-1064) been recently discussed at length by Tucker et al. [23]. Fig. 1 shows that focuses the laser pulse onto the sample surface for LIBS excita- a single-shot LIBS spectrum of CaSO4·2H2O and the calculated base- tion, with a spot size of ~100 μm diameter. When operating in Mars line in the 420–430 nm region. The spectrum was collected in Earth conditions, the excitation pulse laser beam penetrates through a atmosphere. The interpolated intensity value of the baseline at the quartz fused silica viewport mounted on a lateral wall of the PEACh position 422.67 nm (Ca I emission line) is two orders of magnitude and the LIBS signal collecting optical fiber cable connects to the spec- smaller than the intensity of the Ca emission signal. After applying trometer (outside the PEACh) via a feedthrough mounted on the top the LIBS data baseline removal routine to all the spectra we observed wall of the PEACh. that the ratio spectrum intensity-to-baseline intensity ranges from 10 A bare UV-enhanced optical fiber (230 μm core diameter) posi- to 103 for the emission lines considered for quantification purposes tioned at ~10 mm and 45° from the surface of the sample collects (see Section 3.2.1). We therefore consider that the background is es- the light from the plasma emission. The other end of optical fiber sentially inexistent, mostly thanks to the use of a gated CCD. Chem- for LIBS signal collection is connected to the high-resolution broad- Cam on Curiosity rover will use non-gated detectors, therefore band spectrometer assembly. When using a 50 μm entrance slit, the higher levels of background are expected and baseline removal will spectrometer configuration allows covering the 200–950 nm spectral be necessary. range with a wavelength-dependent spectral resolution ranging from 0.04 to 0.1 nm. For measurements in terrestrial atmosphere, the spec- 3.1.2. Normalization of the spectra tra were collected with a 700 ns delay, and the integration time was In Section 1 we mentioned that most normalization approaches in set to 4 μs. For measurements in Mars atmosphere, the delay and in- LIBS use internal standards to compensate for spectral changes due to tegration times were set to 100 ns and 4 μs, respectively. The spec- matrix effects and variation in experimental conditions. In our partic- trometer was wavelength-calibrated using ten atomic emission lines ular case, finding such standard peaks is not possible as the concen- from an NIST-traceable Hg/Ar lamp. In order to correct the overall trations of all elements vary significantly across the set of samples spectral response of the whole system, especially the nonlinear effi- we analyzed. A more accurate approach seems to be normalizing ciency of the dispersion grating in the Echelle spectrograph and the each spectrum to the total emission integrated intensity, or the total spectral response of detector, a NIST-traceable dual deuterium/quartz area under the spectrum in the 275–850 nm spectral range. Since tungsten halogen lamp was used as an intensity standard. For each of the total collected emission integrated intensity represents approxi- the standard samples used for constructing the calibration curves, an mately the total energy released by the plasma in every shot, this nor- average of five spots across the surface of the sample pellets were an- malization helps, in principle, correcting for pulse-to-pulse variations alyzed; 25 single-shot spectra were recorded at each spot, yielding a total of 125 spectra per calibration target. Dual measurements were made in Earth laboratory atmospheric conditions (1050 mbar of air) and in Mars atmospheric conditions (pressure inside the PEACh first reduced to 0.07 mbar, then raised to 7 mbar of CO2). The delivered energy at the sampling spot was set to 30 mJ for all of the analyzed samples in either atmospheric condition.

3. Automated routines for data processing

3.1. Pre-processing of spectral data

Once the spectral acquisition is finished, the data must be pre- processed to minimize the influence of variable experimental conditions and physical matrix effects. In the following sections we will refer to a series of spectra as a set of typically 125 spectra collected at five spots across the surface of an individual sample. In our LIBS setup, the acquisition software stores single-pulse spectra, hence allowing the manipulation of individual spectra within a series of spectra. We have developed a pulse-to-pulse analysis for the pre- Fig. 1. Single-shot LIBS spectrum of CaSO4·2H2O in Earth atmosphere (blue) and the processing of the LIBS series of spectra that consists of five steps: calculated baseline (green). Author's personal copy

4 P. Sobron et al. / Spectrochimica Acta Part B 68 (2012) 1–16 in the laser energy, spot size, plasma geometry and brightness, collection geometry, and physical matrix effects in our experiments.

3.1.3. Outlier discarding within a series of spectra After normalization, the differences in the spectral features among spectra in the same series are reduced, yet they are noticeable. In our view, these differences can be accounted for by variations in pulse-to- pulse plasma size, or perhaps by plasma initiation by dust particles. These phenomena, among others, result in so-called outlier within a series of spectra. Most spectroscopic techniques, including LIBS, use spectrum arithmetic averaging to generate a single (averaged) spectrum from all of the spectra collected on the same sample regardless of differences between them. Although this approach is generally very effective in the case of, for instance, Raman and infrared spectroscopy, signal variations are often large within LIBS spectral series and may affect quantitative analyses (see discussion above about factors compromising LIBS repeatability). It has been demonstrated that discarding outlier spectra in combination with other data manipulation techniques prior to spectral averaging improves the accuracy of LIBS elemental analysis (e.g., [24]). The pre-processing package we have developed includes a routine to remove outlier spectra based on the Chauvenet criterion [25].To apply this outlier removal criterion to our particular case, we take the values of the intensity of the most intense peak in a (previously baseline-corrected and normalized) spectrum as the data points (dis- criminating variables) in the discrimination algorithm. The intensity maximum is computed for each spectrum in a series of spectra within the 275–850 nm range (a wavelength region that features the element emission lines relevant for this study of hydrous sulfates). The mean (MMean) and standard deviation (MMean) of the intensity maxima within the whole series (125 spectra) is then calculated. A spectrum is considered outlier if the value of its intensity maximum is beyond an interval around MMean defined as: MMean±n×Mσ. The selection of n was made by empirical investigation of different spectral series and its typical value is 0.75. When outliers are found, the corresponding spectra are discarded from the series, and the method is iteratively applied to the remaining spectra in the series until: (i) no outliers are found, or (ii) the number of remaining spectra in the series is less than (typically) 60% of the total number of original spectra. Fig. 2(a) shows the 125 single-shot baseline-corrected and normalized LIBS spectra of CaSO4. The spectra were collected in simulated Mars atmosphere. For visualization purposes only a small spectral region (420–430 nm) is shown. Fig. 2(b) shows the series of spectra of CaSO4 after the first iteration of the outlier removal routine (n=0.75). Note that 50 spectra have been discarded in this first iter- Fig. 2. (a) 125 baseline-corrected and normalized series of LIBS spectra of CaSO4 in ation, and the variability, Mσ (Msigma in the figure), has been re- Mars atmosphere. (b) Series of spectra after the first iteration of the outlier removal duced by a factor of 2.3 (from 0.1881 to 0.0827). If a second routine. (c) Series of spectra of CaSO4 after the second iteration of the outlier removal iteration were applied (Fig. 2(c)), Mσ would be further reduced, as routine. shown in the plot's statistics, but only 33 spectra of the starting 125 spectra would remain for this particular case (CaSO4). However, as discussed below (Section 3.1.4), the SNR of the final averaged spectrum (generated from the remaining spectra after outlier discarding) Table 2 Comparison of pulse-to-pulse variance measured as the relative standard deviation is highly affected by the number of discarded spectra. Therefore, a (RSD=100×Mσ×Mmean) in the intensity maxima of the raw and pre-processed trade-off needs to be considered between the need to remove outliers spectra in laboratory and Mars atmospheres. The number of outlier spectra is in and the necessity to keep enough spectra in a series of spectra for brackets given as a percentage to the total number of raw spectra in a series. averaging. In practice, our rule is to always keep at least 60% of the Samples Raw spectra RSD RSD pre-processed spectra original spectra in a series during the outlier removal process. containing [outlier spectra] The effect of the outlier discarding methodology on the precision Lab Mars Lab Mars and repeatability of the LIBS measurements on the reference samples atmosphere atmosphere atmosphere atmosphere used for constructing the calibration curves is summarized in Table 2. K 58.7 26.8 18.8 [44.6] 24.77 [42.0] The relative standard deviation (RSD) of the intensity maxima in raw Fe 67.7 37.4 18.0 [43.1] 20.97 [44.7] and pre-processed spectra has been calculated by groups of elements Ca 45.0 29.8 12.4 [24.9] 12.94 [34.7] to provide insight on the pulse-to-pulse variation in the emission Mg 51.6 37.6 12.1 [35.0] 17.38 [40.5] properties of different cations in sulfate matrix, and under different Na 71.4 33.0 19.3 [47.5] 11.05 [74.8] Al 55.7 26.5 15.7 [49.4] 24.5 [37.6] atmospheres. The discussion of the phenomenology associated to Author's personal copy

P. Sobron et al. / Spectrochimica Acta Part B 68 (2012) 1–16 5 the differences observed is, however, beyond the scope of this paper Table 3 and the reported values are used as a quality control parameter for Comparison of relative standard deviation (RSD) and signal to noise ratio (SNR) values for the series of LIBS spectra of CaSO after pre-processing. RSD is calculated as de- further quantitative analyses only. The general conclusion is that the 4 scribed in Table 2. SNR is calculated in the averaged spectra after each iteration (up ﬁ outlier discarding process increments signi cantly the precision of to four) as the ratio of the intensity of the emission peaks and the root mean square our LIBS measurements for all of the elements. The number of (RMS) of the spectral response over a wavelength region which is essentially free of remaining (good) spectra per series is in the order of 60% for all ele- emission lines, hence representative of noise alone. ments in both atmospheres, excepting for Na in Mars atmosphere Iteration number Remaining spectraa RSD SNR Ca SNR S (25%). 422.67 nm 545.38 nm

1 75 10.7 344 6.9 2 33 4.9 328 6.1 3.1.4. Spectra averaging 3 13 2.0 306 5.5 Once the outlier spectra are removed from a spectral series the 4 7 0.5 259 6.0 0b 1b – 150 4.4 remaining spectra are averaged. This step results in a single spectrum for each sample that is signal-to-noise enhanced with respect to the a The number of spectra in the raw series is 125. b shot-to-shot individual spectra within a series. Fig. 3 shows an over- SNR calculated for the single-shot spectrum plotted in Fig. 3. lay of a single-shot spectrum of CaSO4 collected in Mars atmosphere and the average spectrum of the same sample in the 420–430 nm region obtained after the pre-processing routines were applied. Pre- 3.1.5. Resolution enhancement processing includes baseline subtraction, normalization and a single To correct for peak distortion effects due to the ﬁnite resolution of iteration for the removal of 50 outlier spectra. The improvement in the spectrometer we treat all of our pre-processed LIBS spectra through the signal-to-noise ratio (SNR) upon spectral averaging is qualitative- a series of routines based on Fourier self-deconvolution (FSD), originally ly noticeable. From a quantitative point of view, Table 3 shows a com- developed by Kaupinnen et al. [26,27]. By way of an example, Fig. 4 3+ parison between precision and signal to noise ratio (SNR) values of shows the LIBS spectrum of natrojarosite [NaFe 3(OH)6(SO4)2]in two emission peaks for pre-processed LIBS series of spectra of CaSO4 Mars atmosphere in the 775–780 nm region before and after FSD. The as a function of the number of iterations for outlier discarding. two intense peaks correspond to oxygen emission lines. The resolution Whereas using a high (≥2) number of iterations yields better RSDs, of the spectrum is notably improved upon FSD (the HWHM of the it implies discarding a great number of the original 125 spectra in 777.19 nm O I peak is reduced from 0.08 to 0.05 nm), allowing the series, which we have observed has a negative effect in later the detection of one additional peak located at 777.08 nm within quantitative analyses. In addition, the SNRs worsen as the number the broad emission of O I peak at 777.19 nm and two additional of remaining spectra in a series decreases. This is not problematic in peaks within the broad emission peak of O at 777.42 nm, located at the case of the Ca I emission peak at 422.67 nm and other peaks 777.55 and 777.72 nm. These three additional self-resolved peaks are where the SNRs are large enough to perform accurate quantitative assigned by NIST's Atomic Spectra Database [28] to Fe I (777.08 nm), calculations even with rather noisy averaged spectra, but it becomes O I (777.55 nm), and Xe III (ambiguous, 777.72 nm) emissions. The an issue for instance for the S emission peak at 545.38 nm shown in 777.72 nm peak is probably an artifact introduced in the spectrum by Table 3 and other low intensity peaks where SNRs are small the FSD procedure (over-deconvolution), given the extremely low Special care has been taken throughout this work to produce aver- probability of detecting Xe at such a high ionization state with our aged spectra from spectral series that contain the maximum number experimental setup. of single-shot spectra while discarding those spectra that, due to the Further than improving the resolution of the LIBS spectra, self- reasons explained above, deviate much from their companion spectra resolution methods provide a way to compare spectra collected in a given recorded series. The semi-arbitrary 60% cutoff value for the with different instrumental set-ups by minimizing the inﬂuence of maximum number of removed outliers allowed per series of spectra the experimental conditions in the broadening and shifting of the is derived from careful empirical inspection of the RSD and SNR statis- emission peaks. In combination with our Mars atmosphere simula- tical values obtained upon processing each of the spectra collected tion capabilities, this might be useful for providing support for Chem- throughout the experiments. Cam data analyses of Martian materials by Curiosity, assuming that

Fig. 4. LIBS spectrum of natrojarosite in Mars atmosphere before (blue) and after Fou-

Fig. 3. Single-shot spectrum of CaSO4 in Mars atmosphere (blue) and the average spec- rier self-deconvolution (green) showing the detection of additional peaks (O, Fe, Xe) trum obtained upon pre-processing (red, from the spectral subset shown in Fig. 2(b)). after spectral resolution enhancement. Author's personal copy

6 P. Sobron et al. / Spectrochimica Acta Part B 68 (2012) 1–16

ChemCam's data preprocessing tools allow generating comparable lines from an element that show major overlap with emission lines LIBS spectra. It is worth noting that a precise independent character- from a different element (e.g., the 308.22/308.37 nm Al/Fe peaks) in ization of the instrumental broadening function is required for our LIBS spectra. Another constrain we have imposed to our database obtaining an adequate performance of the FSD-based methods. is to include only emission lines within the spectral range of Chem- The flowchart in Fig. 5 lists the five pre-processing operations de- Cam (224–327, 384–473, and 494–933 nm) [30], so that our methods scribed above and summarizes their main features and highlights. could eventually provide support to tactical operations of the upcom- ing MSL mission based on ChemCam LIBS data. Table 4 shows the el- 3.2. Quantification of LIBS spectral features ement emission spectral database relevant to the samples studied in this work. An example of the automated peak identification is given in The tools we have developed for the quantification of LIBS spectral Fig. 6, which shows the pre-processed LIBS spectrum of aluminocopia- 3+ features include two steps. The first is an automated set of routines to pite [Al0.66Fe4 (SO4)6(OH)2·20(H2O)] in the 275–850 nm region. The identify the peaks in the pre-processed spectra and to relate them to peaks that match emission lines within our database have been anno- relevant elements in a targeted sample. The second is to use peak- tated with their corresponding element identification. fitting routines to quantify a set of parameters (center wavelength, intensity, width, and Gauss–Lorentz factor) for all of the identified el- 3.2.2. Peak area calculation ement emission peaks for further correlation with the elemental con- The proportional relationship between the intensities of the emis- centrations of the targets. These routines are described in the sion peaks of an element in a LIBS spectrum and the concentration of following two subsections. this element in a sample is the basis for quantitative analysis using LIBS data. Unfortunately, an emission line observed in a LIBS spec- 3.2.1. Line selection trum extends over a range of wavelengths rather than to a single The pre-processed LIBS spectra are compared to a database of wavelength. In addition, under certain circumstances, its center may emission lines built upon the NIST Atomic Spectra Database [28,29] be shifted from its nominal central wavelength. Distortions such as and our own spectral libraries to identify the elements in the interro- line broadening and line shift are caused by local effects during the gated sample. Our routines operate on the search–match principle as formation and evolution of the plasma plume, such as thermal and explained in a previous paper by the authors for Raman spectroscopy pressure broadening as well as by non-local effects such as opacity [22]. To avoid misidentifying elements insofar possible, our database broadening (absorption of electromagnetic radiation as it travels includes multiple emission lines for a given element, each showing through space), self-absorption and macroscopic Doppler broadening minimal or no overlap with the emission lines of other elements. In and shift [31]. The FSD approach mentioned in Section 3.1.5 aims to practice, the later is very difficult to achieve as most geological sam- correct for these line broadening effects, yet the spectral peaks ples have complex compositions that give raise to LIBS spectra popu- show non-zero linewidth after applying the self-deconvolution, due lated with many peaks, most of which overlap with each other. As a to the intrinsic limitations of the method and to the lack of knowledge general rule to avoid this inconvenience we have discarded emission of a function that accounts for the physical and instrumental

Fig. 5. Flowchart of the ﬁve pre-processing operations and summary of their main features. Author's personal copy

P. Sobron et al. / Spectrochimica Acta Part B 68 (2012) 1–16 7

Table 4 Spectral lines of the elements used in this work displayed in wavelength order. Transi- tion type in brackets.

Element Emission lines (nm)

Al (I) 308.2 (I) 309.3 (I) 394.4 (I) 396.2 Ca (II) 315.9 (II) 317.9 (II) 370.6 (II) 373.7 (I) 422.7 (I) 534.9 Fe (II) 278 (I) 404.6 (I) 438.4 (I) 561.6 H (I) 434 (I) 656.3 K (I) 404.7 (I) 693.6 (I) 766.5 (I) 769.9 Mg (I) 285.2 (II) 292.9 (I) 293.7 (I) 517.3 Na (I) 589 (I) 589.6 O (I) 777.2 (I) 777.4 S (II) 414.5 (II) 545.4 (II) 547.4 (II) 551 (II) 556.5 (II) 560.6

broadenings altogether. It is our opinion that the intensity of a peak measured at a single wavelength may not express the true intensity of the emission line associated to it, and that the best correlation between the spectral peaks associated to an element and the concentra- Fig. 7. LIBS spectrum of aluminocopiapite recorded under Mars atmospheric conditions tion of this element in a target (i.e., the most accurate calibration (blue circles) with fitted peaks (cyan), convolution spectrum (dark blue), and residuals curve) is obtained in terms of the integrated intensity over a range (gray). The central peak is contributed by Fe emission lines alone. of wavelengths around the central peak position (i.e., peak area) rather than the intensity at that central peak position. In our approach to the calculation of the integrated intensity of a peak or peak area, the spectral peaks related to elements identified composition upon comparison with our calibration curves (sulfate- upon comparison to the emission lines database are fitted to Voigt bearing samples based on this study). functions by another type of spectral deconvolution: non-linear iter- ative curve fitting based on the Marquardt–Levenberg method [32]. 4. Development of the calibration curves A typical peak-fitting for the spectra used in this work for the quantification of a single Fe emission line is displayed in Fig. 7. Plots of the peak area versus the XRF-calculated elemental atomic Each of the emission lines listed in Table 4 has been carefully ex- fraction have been constructed for selected emission lines of K, Fe, Na, amined and constraints related to the number of bands that compose Ca, Al, Mg, H, O and S in sulfate matrix. The relative standard devia- a peak envelope and their parameters have been determined and tion of the peak fitting and peak area calculation processes is below implemented in the automated peak-deconvolution routines. This ef- 5% for all of the peaks used to generate the calibration curves, except- fort benefits in turn the overall effectiveness of the LIBS data proces- ing for the S peak. Other factors affecting the overall deviations in the sing package by reducing time and computer memory consumption. values of the peak areas are roughly accounted for by the RSDs de- In summary, the result of peak-fitting as applied in this work is scribed in Sections 3.1.4 and 3.2.2.WehaveestimatedtheRSDsof convolution spectral peaks that are composed of a number of peaks the data points in the calibration curves for all elements to be with- whose parameters (center wavelength, peak intensity, FWHM, and in the 5–10% range. The vertical error bars in the calibration curves GLF) are accurately characterized and optimized to fit the original correspond to 10% of the peak area value except where indicated spectral emission peaks with minimum deviation. The derived peak otherwise. In the following subsections, we report the calibration areas of the fitted peaks associated to the elements identified in the curves for the aforementioned elements and discuss their expected LIBS spectra of the calibration standards listed in Table 1 are the ana- accuracy for predicting elemental concentrations in geological sam- lytical signals used to construct our calibration curves. For unknown ples relevant for Mars exploration. The influence of matrix effects samples, the peak areas of the fitted peaks associated to the elements and self-absorption in the calibration curves is also discussed. identified in their LIBS spectra are used to determine the elemental Here we pay attention to the calibration curves in Martian atmosphere only, although the equivalent curves constructed for Earth atmospheric conditions are also available for the same elements, excepting S.

4.1. Ca

Gypsum (CaSO4·2H2O) has been detected on Mars from orbit in northern latitudes [33,34] and in layered deposits near the equator

[35–37]. The occurrence of anhydrite (CaSO4) and bassanite [2(Ca)2(SO4)·(H2O)] at typical Martian cold temperatures has been suggested based on dehydration/rehydration experiments on natural gypsum [38] and from Thermal and Evolved Gas Analyzer (TEGA) measurements on board the Phoenix lander [39].Mawrth Vallis (23°N), one of the ﬁnal candidate landing sites for the Mars Science Laboratory mission, provides access to both phyllosilicates

and sulfates [40], including bassanite [41].Syngenite[K2Ca(SO4)2·(H2O)] is a diagenetic component of marine salt deposits formed after Na–K–

Mg–Ca–Fe–Cl–SO4–H2O-rich brines from which certain Martian mineral assemblages are presumed to have originated [42,43]. However, syngen- ﬁ Fig. 6. Automated peak identiﬁcation in a pre-processed LIBS spectrum of aluminoco- ite has not been identi ed on Mars surface to date. Fig. 8 shows the cal- piapite recorded under Mars atmospheric conditions. ibration curve for Ca developed using three Ca-sulfates with different Author's personal copy

8 P. Sobron et al. / Spectrochimica Acta Part B 68 (2012) 1–16

possible by moving the sample a few microns between laser pulses using a high resolution motorized rotation stage installed inside the PEACh and high resolution linear positioning stages outside, for Mars and Earth conﬁgurations, respectively [19,20]. Since the Ca curve is essentially ﬂat in the 0.09–0.16 Ca atomic fraction region, examining the peak area of the Ca I emission line at

422.67 nm alone is not enough for distinguishing between CaSO4, 2(Ca)2(SO4)·H2O and CaSO4·2H2O. In fact, none of the emission lines considered here for Ca-identiﬁcation purposes offers peak area variations within these three compounds beyond ±10% uncertainty; therefore, H and O emission lines should be examined to determine the hydration state of these Ca-sulfates. The ideal Ca atomic fraction of syngenite (~0.08) is, however, within the quasi-linear region of our calibration curve (Fig. 8).

4.2. Na Fig. 8. Calibration curve for Ca from the LIBS spectra of Ca-sulfates recorded under Mars atmospheric conditions. Na-sulfate minerals have been observed on Mars surface from orbit using VNIR reflectance spectroscopy data provided by the hydration states. The natrojarosite data point was used to mark trace Ca Observatoire pour la Mineralogie, l'Eau, les Glaces et l'Activité concentration. The highly non-linear behavior of the data points at high (OMEGA) instrument on board Mars Express (e.g., [45]). West Candor concentrations of Ca (Ca atomic fraction>0.08) is due to self- Chasma hosts one of the largest sulfate deposits observed on Mars, in- absorption, which can be explained as follows: Laser ablation produces cluding Na-rich sulfate mineralogy [46]. Thenardite (Na2SO4) seems persistent amounts of aerosol above the interrogated sample that sup- to be the most likely Na-sulfate to occur on Mars as a result of evap- plies material into the hot plasma in subsequent pulses. When the con- oration events [47], although mirabilite (Na2SO4·10H2O) and blödite centration of certain species in the plasma exceeds a threshold, i.e., in (Na2SO4·MgSO4·4H2O) often occur in association with thenardite in very high plasma density conditions, the plasma absorbs its own emis- terrestrial evaporite deposits (e.g., [48,49]). Fig. 10 shows the calibra- sion. Wisbrun et al. [44] demonstrated that laser repetition rate have sub- tion curve for Na developed using thenardite, sodium bisulfate mono- stantial influence in the intensity of emission lines in LIBS spectra due hydrate and natrojarosite. The area of the emission line increases essentially to the mentioned self-absorption effect. In addition to monotonously as the atomic fraction of Na in the standard samples influencing the intensity of the observed emission lines in a LIBS spec- increases. Self-absorption is apparent in the Na emission lines in trum, self-absorption causes line-broadening and intensity dips at the Na2SO4. Sallé et al. [50] did not observe self-absorption in the center of the emission lines, as shown in Fig. 9 (pulses #10 and #20). 589.59 nm Na I emission line under Mars atmosphere, but they ana-

The four single-shot spectra of a CaSO4 sample plotted in the ﬁgure lyzed this emission line in non-sulfate-rich natural rocks up to an were collected at the same spot in Earth atmosphere using 35 mJ/pulse Na atomic fraction of ~0.031 (trachyte). The Na atomic fractions in and 1 Hz repetition rate. The emission peak of Ca I at 422.67 nm deterio- the samples used in our calibration curve are much higher (0.05– rates quickly after a few pulses. Spectra collected after 10 shots show an 0.30). Whether a sulfate matrix favors self-absorption of certain Na intense dip at exactly 422.67 nm that can lead the observer, or automat- emission lines in Mars atmosphere, speciﬁcally in the Na atomic frac- ed software, to believe that there are two peaks instead of a single fairly tion range studied here, is unclear given that our calibration curve is symmetrical peak typically associated to the Ca I emission at that wave- constructed with three data points only. Additional Na-rich standard length. Self-absorption of this resonant Ca emission line also occurs in sulfates are needed to provide insight in this issue. Assuming self- Mars atmosphere, although there is relatively low spectral distortion absorption of the Na emission lines considered in this work occurs, compared to Earth atmosphere. In both cases we mitigated it insofar only the ideal atomic fractions of Na in mirabilite and blödite (0.054 and 0.080, respectively) are within the linear region of our Na

Fig. 9. Single-shot LIBS spectra of CaSO4 in Earth atmosphere collected at the same spot after different pulses count showing an intensity dip at the central wavelength and line Fig. 10. Calibration curve for Na from the LIBS spectra of Na-sulfates recorded under broadening as pulses count increases. Mars atmospheric conditions. Author's personal copy

P. Sobron et al. / Spectrochimica Acta Part B 68 (2012) 1–16 9 calibration curve. Nevertheless, mirabilite is not stable even under alunite/jarosite is 0.15 and 0.038, respectively, well within the span normal terrestrial laboratory conditions, not to mention those of of our calibration curve. Mars. 4.4. Al

2+ 4.3. K Alunite, alunogen [Al2(SO4)3·17(H2O)], and halotrichite [Fe Al2(SO4)4·22(H2O)], among other metal-sulfate salts rich in alumi- The presence of alunite [KAl3(SO4)2(OH)6] on Mars has been sug- num, are known to form through hydrothermal alteration gested form orbital measurements with the Compact Reconnaissance [52,59,60] and pyrite oxidation [61–63]. Such Al-bearing sulfates Imaging Spectrometer for Mars (CRISM) on the Mars Reconnaissance may also have formed on Mars through similar processes. Gendrin Orbiter (MRO) [51]. Although not detected on Mars through January et al. [36] have suggested the presence of halotrichite on Mars

2012, alum-(K) [KAl(SO4)2·12H2O] often forms from oxidation of sulfu- from OMEGA observations, whereas Lane et al. [64] speculate that rous gasses in geothermal environments (e.g. [52]). Such environments halotrichite may contribute to some spectroscopic data from the are widespread on Mars and include volcanic and hydrothermal sys- salty soils at Gusev crater obtained with the Mössbauer instrument tems, which are considered promising sites for finding evidence of on board the Spirit rover (Mars Exploration Rovers). As mentioned 3+ past life on the Martian surface [53]. Jarosite [KFe 3(OH)6(SO4)2] above, alunite has been detected from orbit. Fig. 12 shows the cali- was identified by Mössbauer spectrometer on board Opportunity bration curve for Al developed using aluminocopiapite, alum-(K), and rover and it accounted for 10 wt.% of a Meridiani outcrop on Mars natrojarosite. Although the natrojarosite data point features trace [54]. Fig. 11 shows the calibration curve for K developed using alum- atomic fraction of Al, the three data points can be fitted to a straight

(K), natrojarosite, and K2SO4. The sum of the area of the emission line with minimal error. The ideal Al atomic fractions of the Mars- lines increases monotonously as the atomic fraction of K in the standard relevant alunite, alunogen and halotrichite are 0.115, 0.029 and 0.022, samples increases. There is no evidence of self-absorption of the 766.45 respectively. These Al atomic fractions are beyond the span of our cal- and 769.90 nm K I emission lines. Similar to Na, Salle et al. [50] did not ibration curve. However, we can take advantage of the linear behavior observe self-absorption for K under Martian conditions, although the of the curve and extend the calibration to higher Al atomic fractions. highest K atomic fraction used in their work is ~0.027 for trachyte, much smaller that the atomic fraction of K2SO4 (0.23) used in our cali- 4.5. Mg bration curve. It must be noted, however, that reduced pressure in

Mars atmosphere facilitates plasma expansion [55–57]. Rapid plasma Kieserite (MgSO4·H2O) has been identified by OMEGA in many lo- expansion reduces the probability of re-absorption of the emitted pho- cations on Mars [36]. Experimental work [65–69] suggests that highly tons when they reach the edges of the plasma plume; this phenomenon hydrated Mg-sulfates could have formed and persisted under moder- contributes strongly to the self-absorption mechanism [21].Thiscould ate relative humidity and warmer temperature conditions expected be the case in our K calibration curve; although this claim is tentative for Mars during periods of high obliquity [70]. Wang and Freeman due to the lack of reliable reference samples featuring K atomic fractions [71] have recently demonstrated through laboratory experiments that in the 0.04–0.3 range. Equally tentative is fitting the data points to a lin- kieserite, starkeyite (MgSO4·4H2O), a low-temperature MgSO4·7H2O ear calibration curve. One possible explanation for the relatively poor fit phase, and meridinaniite (MgSO4·11H2O) are stable/metastable phases is the low intensity of the K emission peaks in the LIBS spectra of alum- of Mg-sulfate under current T-RH conditions at the Mars surface and (K) and natrojarosite, which increases the uncertainty in the quantifica- subsurface. Fig. 13 shows the calibration curve for Mg developed tion of the peak areas. Additional reliable standards should be used to using magnesiocopiapite, hexahydrite, starkeyite, kieserite and anhy- optimize this regression. The extended concentration range shown in drate MgSO4. The effect of self-absorption previously discussed for the the curve plot embraces all of the possible K concentrations (major Ca calibration curve can be observed here for high Mg concentrations and minor) in sulfate matrix. (Mg atomic fraction>0.04). The samples were moved before the acqui- It has been demonstrated that alum-(K) dehydrates under low sition of each single-shot spectrum to minimize the spectral distortions pressure conditions [58]. Although there is no evidence, to our knowl- due to self-absorption. Although the Mg atomic fractions of all the pos- edge, that dehydrated phases of alum-(K) exist on Mars surface, their sible hydrated phases of Mg-sulfate are within the span of our calibra- K atomic fractions (0.02 to 0.083) would be within the boundaries of tion curve, self-absorption of the Mg I 285.21 nm emission line our calibration curve. The ideal K atomic fraction of syngenite and prevents extraction of the hydration state of the Mg-salts based on

Fig. 11. Calibration curve for K from the LIBS spectra of K-sulfates recorded under Mars Fig. 12. Calibration curve for Al from the LIBS spectra of Al-sulfates recorded under atmospheric conditions. Mars atmospheric conditions. Author's personal copy

10 P. Sobron et al. / Spectrochimica Acta Part B 68 (2012) 1–16

Fig. 13. Calibration curve for Mg from the LIBS spectra of Mg-sulfates recorded under Mars atmospheric conditions. this line alone. Similar to Ca-sulfates, H and O emission lines must be examined for this purpose. Meridianiite is an exception since its ideal Mg atomic fraction is most likely within the quasi-linear region of our calibration curve.

4.6. Fe

Iron sulfates, including szomolnokite (FeSO4·H2O) and other hydroxylated ferric sulfates (possibly copiapite and jarosite) have been detected from orbit in localized areas on Mars surface [37,72]. Jarosite and nonspecific ferric sulfates have been identified in Meridiani Pla- num and Gusev Crater by the Mars Exploration Rovers [54,73]. The 2+ 3+ occurrence of ferricopiapite [Fe 0.66Fe4 (SO4)6(OH)2·20(H2O)], hy- 3+ 3+ dronium jarosite [(H3O)Fe 3(SO4)2(OH)6], fibroferrite [Fe (SO4) 3+ (OH)·5(H2O)], rhomboclase [HFe (SO4)2·4(H2O)] and paracoquim- 3+ bite [Fe 2(SO4)3·9(H2O)] has been suggested based on PanCam spectra analysis of salty soils excavated by Spirit rover at the foot of Husband Hill in Gusev crater [74]. These Fe-rich sulfates are just few of the minerals that can occur on Mars through oxidative weathering of iron sulfides [75], although other alternatives for Fe-rich mineralogy formation have been proposed [76–78]. Given the abundance of Fe-rich sulfates relevant to Mars and their importance we have analyzed eight standard iron sulfates to construct the Fe calibration curves shown in Fig. 14. The values of the sum of the peak areas of the three Fe emission lines selected seem to group into two calibration curves (Fig. 14(b) and (c)). We propose an explanation in terms of the chemical matrix effect associated to different crystalline structures in the Fe-sulfates analyzed. The Fe-calibration curve labeled as number 1 (Fe cal. cur. 1, Fig. 16(b)) was constructed using the data points obtained for ferricopiapite, alu- 3+ minocopiapite, magnesiocopiapite [MgFe 4(SO4)6(OH)2·20(H2O)] 3+ and HFe (SO4)·4H2O. In copiapite-group minerals, chains of [Fe(OH)(H2O)O3] octahedra link through additional [SO4] tetrahedra to form infinite chains [79,80]. These chains are linked together through interstitial water and unconnected [AH2O] octahedra (A=Fe3+,Al3+,Mg2+) that contribute to the complex arrangement of hydrogen bonds present in the structure. (There are subtle struc- Fig. 14. Calibration curves for Fe from the LIBS spectra of Fe-sulfates recorded under Mars atmospheric conditions. tural differences among the three copiapites associated to the occupa- tion of the A site [81]; however, for the purpose of our discussion they 3+ are considered isostructural). HFe (SO4)2·4H2O has an equally (natro)jarosite and the rest of Fe-sulfates whose data points were + complex hydrogen bonding scheme involving interstitial (H5O2 ) used to construct the Fe calibration curve number 2 (Fe cal. cur. 2, dimmers that link finite clusters of (SO4) tetrahedral and (FeO6) octa- Fig. 14(c)). At high atomic fractions (>0.06) the computed sum of hedra [80]. the peak area deviates substantially from this linear behavior, most The structures featuring infinite chains (copiapites) and finite likely due to self-absorption of the three Fe emission lines in the 3+ clusters [HFe (SO4)2·4H2O] are very different than the structure of natrojarosite standard. The structure of jarosite-group minerals is Author's personal copy

P. Sobron et al. / Spectrochimica Acta Part B 68 (2012) 1–16 11

based on infinite [Fe(OH)6(SO4)2] sheets, some of which are held to- This was a-priori expected, as the detection of sulfur in LIBS spectra gether by interstitial cations and hydrogen bonds [80]. The structure of sulfur-bearing materials is often a very challenging task. There of Fe2(SO4)3·7H2O is based on infinite sheets of (SO4) tetrahedra are three main reasons for this: (1) the strongest S lines in LIBS emis- 3+ and (Fe O6) octahedra. Similar to jarosite, these sheets are linked sion are found in the vacuum UV (b200 nm) and the NIR (>900 mm) by hydrogen bonds through (H2O) groups in the interlayer [80,82]. ranges [84,85], while our LIBS setup, as well as ChemCam on Curiosity Slabs featuring the same polyhedral arrangement are found in the rover, is designed to analyze the range between 200 nm and 900 nm. structure of Fe2(SO4)3·5H2O, although there are no (H2O) groups in Only low intensity S lines are observed in the visible region the interlayer, and the Fe octahedra are connected solely via (SO4) (400–600 nm); (2) S lines in this spectral range overlap with Fe emis- tetrahedra [83]. The structure of FeSO4·4H2O is based on finite clus- sion lines (where present) which are notably stronger; and (3) elec- 2+ ters of (SO4) tetrahedra and (Fe O6) octahedral linked solely by hy- tronically excited S in the plasma of S-bearing samples can react drogen bonds within and between adjacent clusters [80]. with oxygen in laboratory atmosphere [85] and, to a lesser extent,

Considering the differences in the complex Fe-rich crystalline with oxygen produced by molecular breakdown of CO2 in Mars atmo- structures of the standards used in this work, it seems appropriate sphere. This phenomenon results in a very poor signal-to-noise ratio to construct two Fe calibration curves to correlate the sum of the of the S lines in the mentioned spectral range, as we showed in a pre- peak areas of the Fe II 259.83, Fe I 259.88 and Fe I 259.93 nm emission vious paper [19]. We can conclude that relatively accurate detection lines with the atomic fraction of Fe in the samples. Aside from this of S in sulfate-bearing samples can be achieved in Mars atmosphere, triplet peak, no other Fe emission lines recorded in this work generat- but not in Earth atmosphere. Nevertheless, quantification of S, ed simple or informative calibration curves. The dual Fe calibration where possible, is subject to large uncertainty. curve we have obtained analyzing these particular Fe emission lines is probably the best way to account for the chemical matrix effects as- 4.8. H sociated to structural arrangements at the molecular level. We suggest that molecular (structural) information can be obtained from The characterization of the hydration state of sulfate minerals and the interrogated sample along with its elemental composition in the salt assemblages is important to reconstruct ancient Martian environ- context of standard calibration curve-based quantitative LIBS ana- ments, which has broad implications for habitability studies. The LIBS lyses. It must be noted that it is not the aim of this work to provide technique is uniquely suited to this purpose, since it is in principle a detailed correlation between the fine structure of Fe-sulfates and sensitive to hydrogen. Fig. 16 shows the calibration curve for H, their spectral features. This is a somewhat unexplored territory, as it which includes all of the standard samples used in this work. The fig- is often assumed that the plasma generation/breakdown process pre- ure shows also a plot of a linear fitted function that represents, how- cludes the characterization of the structural information from the ever, a poor fit. This is due to chemical matrix effects caused by the sample. complexity of hydrogen bonding in the crystal structures of these samples, and to a relatively low sensitivity of our system to H. Note 4.7. S that the steepness of the linear curve is ~1:20. Close inspection of Fig. 16 allows observing that certain minerals tend to group in consis- Fig. 15 shows the calibration curve for S, which includes all of the tent trends where the peak area of the broad H I 656.3 nm line in- standard samples used in this work. We have tentatively outlined two creases monotonously as a function of H atomic fraction. For clarity, possible calibration curves that show very small steepness (~1:500). Fig. 17 shows close-up views of the individual H calibration curves For comparison, the steepness of the calibration curves of Ca, Na, K, we have constructed for Ca-, Mg-, and Fe-rich sulfates. The peak Al, Mg, and Fe (Fe cal. cur. 1) is ~1:10, 1:2, 1:1.7, 1:1, 1:2, and 1:0.5. areas of the analyzed H emission lines increase monotonously as H

Fig. 15. Calibration curves for S based on the LIBS spectra of all the standard sulfate samples recorded under Mars atmospheric conditions. Vertical error bars are plotted on a ±15% basis. Author's personal copy

12 P. Sobron et al. / Spectrochimica Acta Part B 68 (2012) 1–16

Fig. 16. Calibration curve for H based on the LIBS spectra of all the standard sulfate samples recorded under Mars atmospheric conditions. concentration increases in the Ca and Mg-sulfate samples, enabling O emission lines in the LIBS spectra of geological samples, thus en- the extraction of the hydration state of the hydrous CaSO4.nH2O and abling quantitative measurements. This applies to both Earth and MgSO4.nH2O series from their LIBS spectra. This capability can be Mars atmospheres. Finally, it is worth noting that the detection of car- compromised when encountering a mineral mixture in a sampling bon in geological samples in Martian atmosphere is also known to be spot on Mars. However, during the ChemCam investigations by highly compromised by the coupling of atmospheric carbon (from

Curiosity, multiple sampling at one location can help mitigate the CO2 breakdown) and carbon in the samples [86]. More work needs problem. to be done in this direction to enable the unambiguous detection of In the Fe-sulfates, the H calibration curve displayed in Fig. 17(c) organic biomarkers using ChemCam on the MSL rover. does not show a trend consistent with all of the plotted values. The structural differences among the various Fe-sulfates discussed in 5. Application of the standard calibration curves to the Section 4.6 account for most of the chemical matrix effects in the characterization of natural samples samples thought to be responsible for the likely dual calibration curve shown in Fig. 14. They are also likely to be responsible for the 5.1. Natural sulfate from a playa deposit far-from-linear behavior of the H calibration curve for iron sulfates. Natrojarosite, featuring very low amounts of hydrogen, can be readily We have analyzed a sample collected during field work in north- identified from the H calibration curve in Fig. 17(c). Copiapite-group western Qaidam Basin on Qinghai–Tibet Plateau, China [48,87]. Qing- minerals can also be identified from this plot. However, the specific hai paleolake evaporated completely in the Late Pleistocene at Da type of copiapite (Al-, Mg- or Fe-copiapite) cannot be determined as Langtan area creating a playa surface. The evaporative salts from they feature almost identical concentration of hydrogen. For that pur- lakes and playas in this region represent the nearly final stage of pose, it is necessary to extract the concentration of Al, Mg and Fe from evaporation sequence of K, Na, Ca, Mg, Fe, C, B, S, and Cl-bearing the corresponding calibration curves. The same is true for the remind- brines. They may serve as an analog site for studying the precipitation ing hydrous Fe-sulfates. sequence and subsequent dehydration/degeneration of Martian salts. The analyzed sample was a whitish granular aggregate potentially 4.9. O containing several mineral phases. It was pressed into a pellet as described in Section 2.1. The LIBS spectra were recorded in Mars atmo- Fig. 18 shows the calibration curve for O, which includes all of the sphere using the experimental conditions detailed in Section 2.2. standard samples used in this work. There is no correlation between Fig. 19 displays the LIBS spectrum of the sample labeled DL0216. the peak area of the O I 777.19/777.42 nm doublet and oxygen con- The statistics of the spectral pre-processing and the results of the centration. This is because atmospheric oxygen (from the breakdown quantitative analysis using the calibration curves are listed in of CO2 molecules) contributes strongly to the LIBS signal from the Table 5. The quantitative analysis revealed that the sample contains oxygen-containing sulfates. This complicates further quantitative cal- relatively large amounts of Al, relatively low amounts of Na and K, culations. This phenomenon has been investigated by Knight et al. trace amounts of Mg, and below limit of detection amounts of Fe

[57]. They determined that at typical Mars CO2 pressures, certain O and Ca. Cu and Si were also identified although their abundances lines in LIBS spectra of soil samples double their intensity compared were not quantified. The atomic fractions of three potential phases to those obtained for the same sample in Ar atmosphere. Differentia- that we suggest could be present in the analyzed sample are also tion of atmospheric oxygen from oxygen in samples is a very difficult listed in Table 5: natroalunite [NaAl3(SO4)2(OH)6], alum-(K) and task which has to be carried on a sample-specific basis, as plasma cou- natrochalcite [NaCu2(SO4)2(OH)·H2O]. These three minerals are con- pling with the gaseous matrix is most likely determined by the chem- sistent with evaporated hyper-saline lakes and lacustrine environ- ical composition and structure of the sample under investigation. ments [88], oxidation of sulfides in shales and slates [89], and While oxygen detection is possible, further investigation is required oxidation of copper deposits in hyper-arid environments [90], respec- to accurately remove the contribution of atmospheric oxygen to the tively. Note that some of the minerals identified by the authors using Author's personal copy

P. Sobron et al. / Spectrochimica Acta Part B 68 (2012) 1–16 13

Fig. 18. Calibration curve for O based on the LIBS spectra of all the standard sulfate samples used in this study recorded under Mars atmospheric conditions.

Qaidam Basin [92]. Cu-enriched sulfate salts form as a byproduct of the oxidation of such metal-sulﬁde mineral deposits. Although these salts are generally very soluble in water, the extreme aridity of the region favors its preservation. It is likely that trace amounts of Cu- sulfates are mixed with the major mineral phases within our sample thereby giving rise to Cu lines in its LIBS spectrum.

5.2. Natural sulfate from an acid-mine related environment

The second sample we have studied was collected in Rio Tinto, Spain, in the surroundings of Peña del Hierro. Weathered shales associated with mine drainage from oxidized sulfides are common in the area [93]. Rio Tinto, and Peña del Hierro in particular, is a sound example of acid mine drainage (AMD) environments, formed after large amounts of sulfides are exposed to water, air, and often mi- crobes that catalyze the oxidation. Given the association of microor- ganisms such as Acidithiobacillus ferrooxidans with the aqueous oxidation of sulfides in Rio Tinto [94,95], the site is accepted by some as a geo/bio/mineralogical setting on our planet which offers comparable scenarios to those on Mars [96], and where technologies for Martian exploration can be defined and tested [97–99]. Our natural sulfate sample is a coated slate displaying whitish, yellowish and orange colors. The coating was mechanically separated from the slate, powdered and pelletized as described in Section 2.1. The LIBS

Fig. 17. Mineral group-speciﬁc calibration curves for H based on the LIBS spectra of selected sulfates recorded under Mars atmospheric conditions. (a) CaSO4·nH2O hydrated series. (b) MgSO4·nH2O hydrated series. (c) Fe-sulfates.

Raman spectroscopy in the playa deposits in Qaidam Basin include carbonate–gypsum-bearing surface soils, salt–clay-bearing deposits, dehydrated Na-sulfates, hydrous multication-sulfates, carbonates, and chlorites [48]. The presence of phyllosilicates and hydrated sulfate mixtures, known to happen in the ﬁeld site where the sample Fig. 19. LIBS spectrum of a natural sulfate-bearing sample (DL0216) from the Qaidam was collected [91], explains the Si spectral lines present in the LIBS Basin and automated element identiﬁcation (labeled with red triangles) by our proces- spectrum. Cu–S deposits are frequent in the northern rim of the sing software. Author's personal copy

14 P. Sobron et al. / Spectrochimica Acta Part B 68 (2012) 1–16

Table 5 Table 6 Statistics and quantitative analysis of the series of LIBS spectra from playa deposit sam- Statistics and quantitative analysis of the series of LIBS spectra from acid-mine deposit ple DL0216. sample loco6D.

Pre-processing Pre-processing

No. of spectra in raw series: 125 No. of spectra in raw series: 125 No. of iterations for outliers removal: 2 No. of iterations for outliers removal: 2 No. of spectra in final series: 93 No. of spectra in final series: 84 RSD final series: 36.7 RSD final series: 25.1 Quantitative analysis Quantitative analysis

Elementa Area Atomic Natroalunite Alum- Natro Elementa Area Atomic Mg- Al- Na- fractionb K chalcitec fractionb copiapite copiapite jarosite

Mg 0.0013±5.E-05 b0.001d Mg 0.0101±4.E-04 0.0135– 0.01 H 0.0090±4.E-04 0.35– 0.23 0.50 0.17 0.0145 0.45 H 0.0106±4.E-04 0.4–0.5 0.42 0.43 0.23 ±4.E-02 ±5.E-02 O 0.0459±2.E-03 NAe 0.54 0.42 0.56 O 0.0353±1.E-03 NAc 0.46 0.47 0.54 Fe 0.0000±0.E+00 b0.025d Fe 0.0441±2.E-03 0.0292– 0.04 0.04 0.12 S 0.0003±1.E-05 NAe 0.08 0.04 0.11 0.0312 Ca 0.0000±0.E+00 b0.001d ±3.E-03 Na 0.0227±9.E-04 0.053– 0.04 0.06 S 0.0004±2.E-05 NAc 0.06 0.06 0.08 0.054 Ca 0.0007±3.E-05 b0.001d ±5.E-03 Na 0.0017±7.E-05 0.0421– 0.04 Al 0.0145±6.E-04 0.0168– 0.12 0.02 0.0447 0.0185 ±4.E-03 ±2.E-03 Al 0.0020±8.E-05 0.0028– 0.01 K 0.0177±7.E-04 0.0533– 0.02 0.0030 0.0565 ±3.E-04 ±6.E-03 K 0.0000±0.E+00 b0.02d

a Additional elements identified: Cu, Si. a Additional elements identified: Cu. b Given as an interval where the minimum and maximum values are calculated from b Given as an interval where the minimum and maximum values are calculated from the calibration curve using the derived peak areas±5%. the calibration curve using the derived peak areas ±5%. c Ideal Cu atomic fraction is 0.1111. c Not quantifiable at present (see calibration curves discussion, Sections 4.7–4.9). d Atomic fraction is below the limit of detection or beyond the linear regime of the d Atomic fraction is below the limit of detection or beyond the linear regime of the calibration curve. calibration curve. e Not quantifiable at present (see calibration curves discussion, Sections 4.7–4.9).

to account for the observed emission lines related to these elements. spectra were recorded in Mars atmosphere using the experimental Cuprocopiapite [CuFe3+ (SO ) (OH) ·20(H O)], jarosite and gypsum, conditions detailed in Section 2. The LIBS spectrum of the sample la- 4 4 6 2 2 found in the area by the authors in independent studies [100,101],are beled loco6D is plotted in Fig. 20. The statistics of the spectral pre- possible candidates. processing and the results of the quantitative analysis using the calibration curves are listed in Table 6. While Mg, Na and Fe are the major 6. Summary and concluding remarks cations, there is significant abundance of Al and Ca. K and Cu were also identified although their abundance was not quantified. Given We have developed a set of routines for LIBS spectral data proces- the derived atomic fraction we suggest that the sample is potentially sing aimed to determine the elemental composition of sulfate sam- composed of a mixture of magnesiocopiapite [MgFe3+ (SO ) (OH) 4 4 6 2 ples. Elements are identified by their fingerprint spectra and their ·20(H O)], aluminocopiapite, and natrojarosite. The ideal atomic frac- 2 atomic fraction is calculated by using calibration functions con- tions of these three minerals are also listed in Table 6.Thepresenceof structed upon analysis of reference sulfate materials under relevant trace amounts of other Cu-, K-, and Ca-rich sulfates is also suggested atmospheric conditions (Earth and Mars). The calibration curves we have developed for Fe are probably related to the crystalline structure of the Fe-sulfates we have analyzed in this work. The potential for LIBS for providing structural information is a promising possibility that warrants further investigation. In addition, the calibration curves we have constructed allow determining the degree of hydration in hydrous Mg-, Ca-, and, to a certain extent, Fe-sulfates. The routines and the calibration curves have been used to analyze the LIBS spectra of two natural sulfate samples from potential Mars analog sites. Both DL0216 and loco6D samples are good examples of complex mixtures of several mineral phases. We have used our data processing tools to identify most of the elements present in the samples and we have extracted relatively accurate values for the abundances of several of these elements. The characterization of the mineral phases present in the samples is however ambiguous as such mixtures add great complexity to the LIBS spectra. This inconvenience has been minimized by making use of the contextual geological and mineralogical information available for the field sites via published literature and in-situ field investigations. In cases where such information is not available, techniques for mineral identification like laser Fig. 20. LIBS spectrum of a natural sulfate sample (loco6D) from Rio Tinto and automat- Raman spectroscopy and X-ray diffraction should be used to constrain ed element identification (labeled with red triangles) by our processing software. the mineralogy of the samples under investigation. Author's personal copy

P. Sobron et al. / Spectrochimica Acta Part B 68 (2012) 1–16 15

The software we have developed can process 125 LIBS spectra col- [14] A. Ciucci, M. Corsi, V. Palleschi, S. Rastelli, A. Salvetti, E. Tognoni, New procedure for quantitative elemental analysis by laser-induced plasma spectroscopy, Appl. lected at the same spot of a target to yield quantitative compositional Spectrosc. 53 (1999) 960–964. information in less than 30 s when using an Intel® Core™2 Duo [15] P. Yaroshchyk, D. Body, R.J.S. Morrison, B.L. Chadwick, A semi-quantitative E6550 CPU running Microsoft Windows XP. The whole package in- standard-less analysis method for laser-induced breakdown spectroscopy, Spec- trochim. Acta Part B 61 (2006) 200–209. cludes, so far: (1) pre-processing of LIBS spectral data: baseline cor- [16] M. Corsi, V. Palleschi, A. Salvetti, E. Tognoni, Making LIBS quantitative — a critical rection, spectra normalization, outlier discarding, spectra averaging, review of the current approaches to the problem, Res. Adv. Appl. Spectrosc. 1 and resolution enhancement; (2) quantitative spectral analysis algo- (2000) 41–46. rithms for element identification against a database of 100+ atomic [17] W.G. Kong, A. Wang, J.J. Freeman, P. Sobron, A comprehensive spectroscopic study of synthetic Fe2+,Fe3+,Mg2+ and Al3+ copiapite by Raman, XRD, LIBS, spectral lines, computation of peak areas and central wavelengths MIR and vis–NIR, J. Raman Spectrosc. 42 (2011) 1120–1129. upon spectral peak fitting; (3) tools for the derivation of the atomic [18] Z.C. Ling, A. Wang, A systematic spectroscopic study of eight hydrous ferric sul- – fractions of up to eight elements in the interrogated spot of a given fates relevant to Mars, Icarus 209 (2010) 422 433. [19] P. Sobron, A. Wang, A planetary environment and analysis chamber (PEACh) for geological sample (sulfate-bearing samples only based on the calibra- coordinated Raman-LIBS-IR measurements under planetary surfaces environ- tion curves presented in this study). mental conditions, J. Raman Spectrosc. (2011), doi:10.1002/jrs.3017. [20] P. Sobron, A. Wang, A Planetary Environment and Analysis Chamber Equipped with Multiple Spectroscopic Sensors, Lunar and Planetary Institute Science Con- Acknowledgments ference Abstracts, 41, 2010, abstract# 1994. [21] D.A. Cremers, L.J. Radziemski, Handbook of Laser-Induced Breakdown Spectros- copy, John Wiley & Sons, Chichester, West Sussex, England; Hoboken, NJ, 2006. This work was supported by NASA grant #1295053 under the MoO [22] P. Sobron, F. Sobron, A. Sanz, F. Rull, Raman signal processing software for auto- program for ExoMars mission; by NASA grants NNX07AQ34G and mated identification of mineral phases and biosignatures on Mars, Appl. Spec- trosc. 62 (2008) 364–370. NNX10AM89G under the MFRP program for studying hydrous sulfates [23] J.M. Tucker, M.D. Dyar, M.W. Schaefer, S.M. Clegg, R.C. Wiens, Optimization of relevant to Mars; by NASA contract #09-030 under the MIDP program laser-induced breakdown spectroscopy for rapid geochemical analysis, Chem. for planetary instrument development. This work was also partially Geol. 277 (2010) 137–148. supported by a special fund from the McDonnell Center for the Space [24] D. Body, B.L. Chadwick, Optimization of the spectral data processing in a LIBS simultaneous elemental analysis system, Spectrochim. Acta Part B 56 (2001) 725–736. Sciences at Washington University in St. Louis for developing the LIBS [25] W. Chauvenet, A manual of Spherical and Practical Astronomy, 5th ed. Dover and Raman sensing technology in a Planetary Environment and Analy- Publications, New York, 1960. sis Chamber (PEACh). We thank Roger Wiens and Sam Clegg of MSL- [26] J.K. Kauppinen, D.J. Moffatt, H.H. Mantsch, D.G. Cameron, Fourier self- deconvolution — a method for resolving intrinsically overlapped bands, Appl. ChemCam team for providing advices in building the LIBS facility at Spectrosc. 35 (1981) 271–276. WUSTL. The authors thank Rex Couture for conducting the XRF analysis. [27] J.K. Kauppinen, D.J. Moffatt, H.H. Mantsch, D.G. Cameron, Fourier-transforms in the computation of self-deconvoluted and 1st-order derivative spectra of overlapped band contours, Anal. Chem. 53 (1981) 1454–1457. References [28] Y. Ralchenko, A.E. Kramida, J. Reader, N.A. Team, NIST Atomic Spectra Database (ver. 4.0.1), National Institute of Standards and Technology, Gaithersburg, MD, [1] J.R. Thompson, R.C. Wiens, J.E. Barefield, D.T. Vaniman, H.E. Newsom, S.M. Clegg, 2010 [Online]. Remote laser-induced breakdown spectroscopy analyses of Dar al Gani 476 and [29] D.E. Kelleher, W.C. Martin, W.L. Wiese, J. Sugar, J.R. Fuhr, K. Olsen, A. Musgrove, Zagami Martian meteorites, J. Geophys. Res. Planets 111 (2006) E05006. P.J. Mohr, J. Reader, G.R. Dalton, The new NIST atomic spectra database, Phys. [2] R. Shu, H. Qi, G. Lu, D. Ma, Z. He, Y. Xue, Laser-induced breakdown spectroscopy Scr. T83 (1999) 158–161. based detection of lunar soil simulants for moon exploration, Chin. Opt. Lett. 5 [30] N.L. Lanza, R.C. Wiens, S.M. Clegg, A.M. Ollila, S.D. Humphries, H.E. Newsom, J.E. (2007) 58–59. Barefield, C. Team, Calibrating the ChemCam laser-induced breakdown spectrosco- [3] S.M. Clegg, J.E. Barefield, S. Humphries, R.C. Wiens, D. Vaniman, M.D. Dyar, J.M. py instrument for carbonate minerals on Mars, Appl. Optics 49 (2010) C211–C217. Tucker, S.K. Sharma, A.K. Misra, Remote Laser Induced Breakdown Spectroscopy [31] H.R. Griem, Spectral Line Broadening by Plasmas, Academic Press, New York, 1974. (LIBS) Geochemical Investigation under Venus Atmospheric Conditions, AGU [32] D.W. Marquardt, An algorithm for least-squares estimation of non-linear param- Fall Meeting Abstracts, 31, 2009, 07. eters, J. Soc. Ind. Appl. Math. 11 (1963) 431–441. [4] J.L. Lambert, J. Morookian, T. Roberts, J. Polk, S. Smrekar, S.M. Clegg, R.C. Weins, [33] Y. Langevin, F. Poulet, J.P. Bibring, B. Gondet, Sulfates in the north polar region of M.D. Dyar, A. Treiman, Standoff LIBS and Raman Spectroscopy Under Venus Con- Mars detected by OMEGA/Mars express, Science 307 (2005) 1584–1586. ditions, Lunar and Planetary Institute Science Conference Abstracts, 41, 2010, [34] K.E. Fishbaugh, F. Poulet, V. Chevrier, Y. Langevin, J.P. Bibring, On the origin of gyp- abstract# 2608. sum in the Mars north polar region, J. Geophys. Res. Planets 112 (2007) E07002. [5] R.C. Wiens, S. Clegg, S. Bender, N. Lanza, B. Barraclough, R. Perez, S. Maurice, M.D. [35] J.J. Wray, S.L. Murchie, S.W. Squyres, F.P. Seelos, L.L. Tornabene, Diverse aqueous Dyar, H. Newsom, C. Team, Initial Calibration of the ChemCam LIBS Instrument environments on ancient Mars revealed in the southern highlands, Geology 37 for the Mars Science Laboratory (MSL) Rover, Lunar and Planetary Institute Sci- (2009) 1043–1046. ence Conference Abstracts, 40, 2009, abstract# 1461. [36] A. Gendrin, N. Mangold, J.P. Bibring, Y. Langevin, B. Gondet, F. Poulet, G. Bonello, [6] D.A. Cremers, L.J. Radziemski, History and fundamentals of LIBS Laser Induced C. Quantin, J. Mustard, R. Arvidson, S. LeMouelic, Sulfates in Martian layered ter- Breakdown Spectroscopy, in: Andrzej W. Miziolek, Palleschi Vincenzo, I. Schechter rains: the OMEGA/Mars Express view, Science 307 (2005) 1587–1591. (Eds.), Laser Induced Breakdown Spectroscopy, Fundamentals and Applications, [37] J.L. Bishop, M. Parente, C.M. Weitz, E.Z.N. Dobrea, L.H. Roach, S.L. Murchie, P.C. Cambridge University Press, 2006, pp. 1–39. McGuire, N.K. McKeown, C.M. Rossi, A.J. Brown, W.M. Calvin, R. Milliken, J.F. [7] R.S. Harmon, F.C. De Lucia, A.W. Miziolek, K.L. McNesby, R.A. Walters, P.D. Mustard, Mineralogy of Juventae Chasma: sulfates in the light-toned mounds, French, Laser-induced breakdown spectroscopy (LIBS) — an emerging field- mafic minerals in the bedrock, and hydrated silica and hydroxylated ferric sul- portable sensor technology for real-time, in-situ geochemical and environmen- fate on the plateau, J. Geophys. Res. Planets 114 (2009) E00D09. tal analysis, Geochem. Explor. Environ. Anal. 5 (2005) 21–28. [38] D.T. Vaniman, D.L. Bish, S.J. Chipera, Bassanite on Mars, Lunar and Planetary In- [8] C.B. Stipe, B.D. Hensley, J.L. Boersema, S.G. Buckley, Laser-induced breakdown stitute Science Conference Abstracts, 40, 2009, abstract# 1654. spectroscopy of steel: a comparison of univariate and multivariate calibration [39] D.C. Golden, D.W. Ming, B. Sutter, B.C. Clark, R.V. Morris, W.V. Boynton, M.H. methods, Appl. Spectrosc. 64 (2010) 154–160. Hecht, S.P. Kounaves, Sulfur Mineralogy at the Mars Phoenix Landing Site, [9] S. Rosenwasser, G. Asimellis, B. Bromley, R. Hazlett, J. Martin, T. Pearce, A. Zigler, Lunar and Planetary Institute Science Conference Abstracts, 40, 2009, abstract# Development of a method for automated quantitative analysis of ores using 2319. LIBS, Spectrochim. Acta Part B 56 (2001) 707–714. [40] J. Grotzinger, Beyond water on Mars, Nat. Geosci. 2 (2009) 231–233. [10] J.B. Sirven, B. Salle, P. Mauchien, J.L. Lacour, S. Maurice, G. Manhes, Feasibility [41] J.J. Wray, S.W. Squyres, L.H. Roach, J.L. Bishop, J.F. Mustard, E.Z.N. Dobrea, Identifica- study of rock identification at the surface of Mars by remote laser-induced tion of the Ca-sulfate bassanite in Mawrth Vallis, Mars, Icarus 209 (2010) 416–421. breakdown spectroscopy and three chemometric methods, J. Anal. At. Spectrom. [42] A.G. Fairen, A.F. Davila, L.G. Duport, E.R. Uceda, D.S. Lim, R. Amils, C.P. McKay, A 22 (2007) 1471–1480. Cold and Wet Mars, Lunar and Planetary Institute Science Conference Abstracts, [11] J.B. Sirven, B. Bousquet, L. Canioni, L. Sarger, S. Tellier, M. Potin-Gautier, I. Le 39, 2008, abstract# 1453. Hecho, Qualitative and quantitative investigation of chromium-polluted soils [43] P.L. King, D.T. Lescinsky, H.W. Nesbitt, The composition and evolution of primor- by laser-induced breakdown spectroscopy combined with neural networks dial solutions on Mars, with application to other planetary bodies, Geochim. analysis, Anal. Bioanal. Chem. 385 (2006) 256–262. Cosmochim. Acta 68 (2004) 4993–5008. [12] J.B. Sirven, B. Bousquet, L. Canioni, L. Sarger, Laser-induced breakdown spectros- [44] R. Wisbrun, I. Schechter, R. Niessner, H. Schroder, K.L. Kompa, Detector for trace copy of composite samples: comparison of advanced chemometrics methods, elemental analysis of solid environmental-samples by laser-plasma spectrosco- Anal. Chem. 78 (2006) 1462–1469. py, Anal. Chem. 66 (1994) 2964–2975. [13] A. Koujelev, M. Sabsabi, V. Motto-Ros, S. Laville, S.L. Lui, Laser-induced break- [45] M. Zhu, H. Xie, G. Guan, R.K. Smith, E. Stansbery, Mineral and Lithologic Mapping down spectroscopy with artificial neural network processing for material iden- of Martian Low Albedo Regions Using OMEGA Data, Lunar and Planetary Science tification, Planet. Space Sci. 58 (2010) 682–690. Conference Abstracts, 37, 2006, abstract# 2173. Author's personal copy

16 P. Sobron et al. / Spectrochimica Acta Part B 68 (2012) 1–16

[46] N. Mangold, A. Gendrin, B. Gondet, S. LeMouelic, C. Quantin, V. Ansan, J.P. Bibring, [75] R.G. Burns, Ferric sulfates on Mars, J. Geophys. Res. Solid Earth 92 (1987) E570–E574. Y. Langevin, P. Masson, G. Neukum, Spectral and geological study of the sulfate- [76] N.J. Tosca, S.M. McLennan, D.H. Lindsley, M.A.A. Schoonen, Acid-sulfate weather- rich region of West Candor Chasma, Mars, Icarus 194 (2008) 519–543. ing of synthetic Martian basalt: the acid fog model revisited, J. Geophys. Res. [47] J.M. Moore, M.A. Bullock, H. Newsom, M. Nelson, Laboratory simulations of Mars Planets 109 (2004) E05003. evaporite geochemistry, J. Geophys. Res. Planets 115 (2010) E06009. [77] H. Nekvasil, F.M. McCubbin, N.J. Tosca, A. Smirnov, Hydrothermal activity on [48] A. Wang, M. Zheng, F. Kong, P. Sobron, D.P. Mayer, Saline Playas on Qinghai- Mars: can magmatic waters play a viable role? Lunar and Planetary Institute Sci- Tibet Plateau as Mars Analog for the Formation-Preservation of Hydrous Salts ence Conference Abstracts, 39, 2008, abstract# 1828. and Biosignatures, AGU Fall Meeting Abstracts, 12, 2010, 02. [78] F.M. McCubbin, A. Smirnov, H. Nekvasil, J. Wang, E. Hauri, D.H. Lindsley, Hydrous [49] L.A. Hardie, The origin of the recent non-marine evaporite deposit of Saline Val- magmatism on Mars: a source of water for the surface and subsurface during the ley, Inyo County, California, Geochim. Cosmochim. Acta 32 (1968) 1279–1301. Amazonian, Earth Planet. Sci. Lett. 292 (2010) 132–138. [50] B. Salle, J.L. Lacour, P. Mauchien, P. Fichet, S. Maurice, G. Manhes, Comparative [79] L. Fanfani, A. Nunzi, P.F. Zanazzi, A.R. Zanzari, Copiapite problem — crystal-structure study of different methodologies for quantitative rock analysis by Laser- of a ferrian copiapite, Am. Mineral. 58 (1973) 314–322. Induced Breakdown Spectroscopy in a simulated Martian atmosphere, Spectro- [80] F.C. Hawthorne, S.V. Krivovichev, P.C. Burns, The crystal chemistry of sulfate chim. Acta Part B 61 (2006) 301–313. minerals, Rev. Mineral. Geochem. 40 (2000) 1–112. [51] G.A. Swayze, B.L. Ehlmann, R.E. Milliken, F. Poulet, J.J. Wray, R.O. Rye, R.N. Clark, [81] J. Majzlan, R. Michallik, The crystal structures, solid solutions and infrared spec- G.A. Desborough, J.K. Crowley, B. Gondet, J.F. Mustard, K.D. Seelos, S.L. Murchie, tra of copiapite-group minerals, Mineral. Mag. 71 (2007) 553–569. Discovery of the acid-sulfate mineral alunite in Terra Sirenum, Mars, using MRO [82] P.D. Robinson, J.H. Fang, Crystal structures and mineral chemistry of hydrated CRISM: possible evidence for acid-saline lacustrine deposits? AGU Fall Meeting ferric sulphates .2. Crystal structure of paracoquimbite, Am. Mineral. 56 Abstracts, 44, 2008, 04. (1971) 535–539.

[52] R. Martin, K.A. Rodgers, P.R.L. Browne, The nature and significance of sulphate- [83] J. Majzlan, C. Botez, P.W. Stephens, The crystal structures of synthetics Fe2(SO4)3(H2- rich, aluminous efflorescences from the Te Kopia geothermal field, Taupo Volca- O)5 and the type specimen of lausenite, Am. Mineral. 90 (2005) 411–416. nic Zone, New Zealand, Mineral. Mag. 63 (1999) 413–419. [84] B. Salle, J.L. Lacour, E. Vors, P. Fichet, S. Maurice, D.A. Cremers, R.C. Wiens, Laser- [53] J.J. Walker, J.R. Spear, N.R. Pace, Geobiology of a microbial endolithic commu- induced breakdown spectroscopy for Mars surface analysis: capabilities at nity in the Yellowstone geothermal environment, Nature 434 (2005) stand-off distances and detection of chlorine and sulfur elements, Spectrochim. 1011–1014. Acta Part B 59 (2004) 1413–1422. [54] G. Klingelhofer, R.V. Morris, B. Bernhardt, C. Schroder, D.S. Rodionov, P.A. de [85] L. Dudragne, P. Adam, J. Amouroux, Time-resolved laser-induced break- Souza, A. Yen, R. Gellert, E.N. Evlanov, B. Zubkov, J. Foh, U. Bonnes, E. Kankeleit, down spectroscopy: application for qualitative and quantitative detection P. Gutlich, D.W. Ming, F. Renz, T. Wdowiak, S.W. Squyres, R.E. Arvidson, Jarosite of fluorine, chlorine, sulfur, and carbon in air, Appl. Spectrosc. 52 (1998) and hematite at Meridiani Planum from Opportunity's Mossbauer spectrometer, 1321–1327. Science 306 (2004) 1740–1745. [86] A.M. Ollila, J.G. Blank, R.C. Wiens, J. Lasue, H.E. Newsom, S.M. Clegg, A. Cousin, S. [55] A.J. Effenberger, J.R. Scott, Effect of atmospheric conditions on LIBS spectra, Sens. Maurice, Preliminary results on the capabilities of the ChemCam laser-induced Basel 10 (2010) 4907–4925. breakdown spectroscopy (LIBS) instrument to detect carbon on Mars, Lunar [56] Y. Iida, Effects of atmosphere on laser vaporization and excitation processes of and Planetary Science Conference Abstracts 42 (2011), abstract# 2395. solid samples, Spectrochim. Acta Part B 45 (1990) 1353–1367. [87] P. Sobron, J.J. Freeman, A. Wang, Field Test of the Water-Wheel IR (WIR) Spec- [57] A.K. Knight, N.L. Scherbarth, D.A. Cremers, M.J. Ferris, Characterization of laser- trometer on Evaporative Salt Deposits at Tibetan Plateau, Lunar and Planetary induced breakdown spectroscopy (LIBS) for application to space exploration, Institute Science Conference Abstracts, 40, 2009, abstract# 2372. Appl. Spectrosc. 54 (2000) 331–340. [88] C.N. Alpers, R.O. Rye, D.K. Nordstrom, L.D. White, B.-S. King, Chemical, crystallo- [58] A.K. Galwey, G.G.T. Guarini, A kinetic-study of the dehydrations of the alums graphic and stable isotopic properties of alunite and jarosite from acid-hypersa-

KCr(SO4)2.12H2O and KAl(SO4)2.12H2O, Philos. R. Soc. Lond. A Math. 441 line Australian lakes, Chem. Geol. 96 (1992) 203–226. (1993) 313–329. [89] J.L. Jambor, D.K. Nordstrom, C.N. Alpers, Metal-sulfate salts from sulfide mineral [59] T. Minakawa, S. Noto, Botryogen from the Okuki mine, Ehime Prefecture, Japan, oxidation, Rev. Mineral. Geochem. 40 (2000) 303–350. Chigaku Kenkyu 43 (1994) 175–179. [90] O.W. Jarrell, Oxidation at Chuquicamata, Chile, Econ. Geol. 39 (1944) 251–286. [60] R.L. Frost, M. Weier, J. Martinez-Frias, F. Rull, B.J. Reddy, Sulphate efflorescent [91] D.P. Mayer, R.E. Arvidson, A. Wang, P. Sobron, M.P. Zheng, Mapping Minerals at a minerals from El Jaroso Ravine, Sierra Almagrera — an SEM and Raman spectro- Potential Mars Analog Site on the Tibetan Plateau, Lunar and Planetary Institute scopic study, Spectrochim. Acta Part A 66 (2007) 177–183. Science Conference Abstracts, 40, 2009, abstract# 1877. [61] G. Garcia Garcia, The Rio Tinto mines, Huelva, Spain, Mineral. Rec. 27 (1996) [92] D.Q. Zhang, X.Y. Dang, D.X. Li, C.Y. Feng, H.Q. She, J.W. Li, Two types of massive 275–285. sulfide deposits along northern margin of Qaidam basin, Qinghai Province: II. [62] D.K. Nordstrom, C.N. Alpers, Negative pH, efflorescent mineralogy, and conse- Qinglongtan-style VHMS Cu–S deposits (in Chinese with English abstract), quences for environmental restoration at the Iron Mountain Superfund site, Cal- Miner. Deposita 24 (2005) 575–583. ifornia, Proc. Natl. Acad. Sci. U. S. A. 96 (1999) 3455–3462. [93] D.C. Fernández-Remolar, O. Prieto-Ballesteros, D. Gómez-Ortíz, M. Fernández- [63] T. Buckby, S. Black, M.L. Coleman, M.E. Hodson, Fe-sulphate-rich evaporative Sampedro, P. Sarrazin, M. Gailhanou, R. Amils, Río Tinto sedimentary mineral as- mineral precipitates from the Rio Tinto, southwest Spain, Mineral. Mag. 67 semblages: a terrestrial perspective that suggests some formation pathways of (2003) 263–278. phyllosilicates on Mars, Icarus 211 (2011) 114–138. [64] M.D. Lane, J.L. Bishop, M.D. Dyar, P.L. King, M. Parente, B.C. Hyde, Mineralogy of [94] A.I. Lopez-Archilla, I. Marin, R. Amils, Microbial community composition and the Paso Robles soils on Mars, Am. Mineral. 93 (2008) 728–739. ecology of an acidic aquatic environment: the Tinto River, Spain, Microbial. [65] D.L. Bish, J. William Carey, D.T. Vaniman, S.J. Chipera, Stability of hydrous min- Ecol. 41 (2001) 20–35. erals on the martian surface, Icarus 164 (2003) 96–103. [95] R. Amils, E. González-Toril, D. Fernández-Remolar, F. Gómez, N. Rodríguez, C. [66] S.J. Chipera, D.T. Vaniman, D.L. Bish, J.W. Carey, W.C. Feldman, E. Stansbery, Ex- Durán, Interaction of the sulfur and iron cycles in the Tinto River ecosystem, perimental Stability and Transformation Kinetics of Magnesium Sulfate Hy- Rev. Environ. Sci. Biotechnol. 1 (2002) 299–309. drates that may be Present on Mars, Lunar and Planetary Science Conference [96] D.C. Fernandez-Remolar, R.V. Morris, J.E. Gruener, R. Amils, A.H. Knoll, The Rio Abstracts, 36, 2005, abstract# 1497. Tinto basin, Spain: mineralogy, sedimentary geobiology, and implications for in- [67] I.M. Chou, R.R. Seal II, Determination of epsomite–hexahydrite equilibria by the terpretation of outcrop rocks at Meridiani Planum, Mars, Earth Planet. Sci. Lett. humidity-buffer technique at 0.1 MPa with implications for phase equilibria in 240 (2005) 149–167.

the system MgSO4–H2O, Astrobiology 3 (2003) 619–630. [97] C.R. Stoker, L.G. Lemke, H. Cannon, B.J. Glass, S. Dunagan, J. Zavaleta, D. Miller, J. [68] H.H. Emons, H. Jahn, G. Wolf, Investigations of salt-mixed solvent systems .38. Gómez-Elvira, E. Stansbery, Field Simulation of a Drilling Mission to Mars to Crystallization enthalpy of magnesium-chloride in mixed-solvent systems, Search for Subsurface Life, 36th Annual Lunar and Planetary Science Conference, J. Solut. Chem. 19 (1990) 447–455. 36, 2005, abstract# 1537. [69] D.T. Vaniman, D.L. Bish, S.J. Chipera, C.I. Fialips, J. William Carey, W.C. Feldman, [98] F. Rull, G. Klingelhöfer, A. Sansano, I. Fleitcher, P. Sobrón, M. Blumers, A. Magnesium sulphate salts and the history of water on Mars, Nature 431 Lafuente, D. Schmanke, J. Maul, In-situ micro-Raman and Moessbauer spectro- (2004) 663–665. scopic study of evaporate minerals in Rio Tinto (Spain): applications for plane- [70] M.I. Richardson, M.A. Mischna, Long-term evolution of transient liquid water on tary exploration, Conference on Micro-Raman Spectroscopy and Luminescence Mars, J. Geophys. Res. 110 (2005) E03003. Studies in the Earth and Planetary Sciences, 1473, 2009, abstract# 68–69. [71] A. Wang, J.J. Freeman, Pathways and rates of Mg-sulfate dehydration and rehy- [99] V. Parro, L.A. Rivas, J.A. Rodríguez-Manfredi, Y. Blanco, G. de Diego-Castilla, P. dration on Mars, Conference on Micro-Raman Spectroscopy and Luminescence Cruz-Gil, M. Moreno-Paz, M. García-Villadangos, C. Compostizo, P.L. Herrero, The Studies in the Earth and Planetary Sciences, 1473, 2009, abstract# 89–90. SOLID (Signs Of LIfe Detector) Instrument Concept: an Antibody Microarray- [72] K.A. Lichtenberg, R.E. Arvidson, R.V. Morris, S.L. Murchie, J.L. Bishop, D.F. Remolar, Based Biosensor for Life Detection in Astrobiology, EGU General Assembly 2009, T.D. Glotch, E.N. Dobrea, J.F. Mustard, J. Andrews-Hanna, L.H. Roach, Stratigraphy 11, 2009, abstract# 8614. of hydrated sulfates in the sedimentary deposits of Aram Chaos, Mars, [100] P. Sobron, A. Sansano, A. Sanz, T. Acosta, B. Lafuente, F. Rull, F. Sobron, J. Medina, J. Geophys. Res. Planets 115 (2010) E00D17. Raman spectroscopy, a powerful tool for the characterization of hydrated sul- [73] R. Gellert, R. Rieder, J. Bruckner, B.C. Clark, G. Dreibus, G. Klingelhofer, G. Lugmair, fates and acidic water in Rio Tinto (Spain), Conference on Micro-Raman Spec- D.W. Ming, H. Wanke, A. Yen, J. Zipfel, S.W. Squyres, Alpha particle X-ray spectrom- troscopy and Luminescence Studies in the Earth and Planetary Sciences, 1473, eter (APXS): results from Gusev crater and calibration report, J. Geophys. Res. 2009, abstract# 77–78. Planets 111 (2006) E02S05. [101] P. Sobron, A. Sanz, T. Acosta, F. Rull, A Raman spectral study of stream waters [74] J.R. Johnson, J.F. Bell, E. Cloutis, M. Staid, W.H. Farrand, T. Mccoy, M. Rice, A. and efﬂorescent salts in Rio Tinto, Spain, Spectrochim. Acta Part A 71 (2009) Wang, A. Yen, Mineralogic constraints on sulfur-rich soils from Pancam spectra 1678–1682. at Gusev crater, Mars, Geophys. Res. Lett. 34 (2007) L13202.