Calibration of the Mars Pathfinder alpha proton X-ray spectrometer


  • C. Nicole Foley,

    1. Laboratory for Astrophysics and Space Research, University of Chicago, Chicago, Illinois, USA
    2. Enrico Fermi Institute, University of Chicago, Chicago, Illinois, USA
    3. Department of the Geophysical Sciences, University of Chicago, Chicago, Illinois, USA
    4. Currently at the Geology Department, Field Museum of Natural History, Chicago, Illinois, USA.
    Search for more papers by this author
  • Thanasis E. Economou,

    1. Laboratory for Astrophysics and Space Research, University of Chicago, Chicago, Illinois, USA
    2. Enrico Fermi Institute, University of Chicago, Chicago, Illinois, USA
    Search for more papers by this author
  • Robert N. Clayton,

    1. Enrico Fermi Institute, University of Chicago, Chicago, Illinois, USA
    2. Department of the Geophysical Sciences, University of Chicago, Chicago, Illinois, USA
    3. Department of Chemistry, University of Chicago, Chicago, Illinois, USA
    Search for more papers by this author
  • William Dietrich

    1. Laboratory for Astrophysics and Space Research, University of Chicago, Chicago, Illinois, USA
    Search for more papers by this author


[1] The chemical compositions of Martian rocks and soils examined with the alpha proton X-ray spectrometer (APXS) during the Mars Pathfinder 1997 lander mission were not previously fully determined. Preliminary chemical results included major element abundances determined by the incomplete calibration of the X-ray mode. The data collected from the alpha and proton detectors were not previously analyzed due to significant atmospheric contributions to the spectra. The backup instrument of the Pathfinder alpha proton X-ray spectrometer flight instrument has been used to complete the instrument calibration under simulated Martian conditions at the University of Chicago. An APXS instrument has been used to create a Pathfinder calibration library to test the accuracy of all three instrument modes under simulated Martian conditions. This calibration library has been tested on a number of geologic standards. We have also corrected for instrument differences between the laboratory and flight units. Significant chemical results, somewhat different from those initially reported by Rieder et al. [1997a] and by Wänke et al. [2001], are reported by Foley et al. [2003] as a result of this reanalysis of the Pathfinder APXS data.


1. Introduction

[2] Previous calibration of the APXS X-ray mode, resulting in the preliminary results reported by Rieder et al. [1997a], was based on calibration curves that did not compensate for instrumental differences between the laboratory and flight instruments. Furthermore, the α and proton modes were not previously calibrated under conditions of Martian atmospheric composition and temperature. Two differences between the laboratory and flight instruments were compensated for during the calibration. One is the presence of a protective film held by a grid in the laboratory instrument between the sources and the sample. This film protects the instrument and samples from curium source contamination. However, it also alters the ratio of α-particles/X-rays incident on the sample, which affects the X-ray generation. The other difference is a fixed measurement distance during laboratory calibration, but various measurement distances for the Mars Pathfinder sample analyses. In particular, the Pathfinder measurement distance varied from 4 to 15 mm further than the laboratory calibration distance. This distance change affects the computed α-mode abundances.

[3] The APXS has three instrumental modes that rely on the bombardment of a sample by α-particles and X-rays from the decay of 244Cm sources. The α-mode measures the energy and intensity of back-scattered α-particles from a sample, which enables measurement of all major and some minor rock-forming elements, except for hydrogen. The α-mode is capable of independently measuring C, O, Si, Ca, and Fe abundances. The proton-mode measures the intensity of protons emitted from (α, p) reactions in the sample, which enables measurement of Na, Mg, Al, Si, S, and N. The X-ray mode measures the intensities of X-rays produced from α-particles and X-rays striking the sample and enables measurement of abundances of elements which have X-ray energies ranging from 1 to 15 KeV. This enables measurement of major and minor rock-forming elements ranging from Na through Ni in atomic number. The three systems can be used independently, but it is the combination of all three modes that provides in some cases complementary and in other cases redundant information that results in an accurate and complete chemical analysis.

1.1. Alpha Mode

[4] The decay of 244Cm to 240Pu, within the nine sources of the APXS, yields 1.62 × 109 alpha particles/sec, of which 76.7% have an energy of 5.806 MeV, and 23.3% have an energy of 5.766 MeV [Lederer et al., 1967]. Alpha particles which scatter elastically from the sample at angles between 149° and 175° strike the silicon α-detector with an energy that is dependent on both the atomic mass of the element from which they scattered and the scattering angle.

[5] The physics of elastic scattering from target atoms was described in detail by Rutherford et al. [1930], who first used α-scattering experimental data from Geiger and Marsden [1909] to determine the size of atomic nuclei using Coulomb scattering principles [Rutherford, 1911]. From the conservation of momentum and energy, the energy of the scattered α depends upon both the scattering angle and the atomic mass of the target atom. The scattered α energy, E, is related to the initial α energy, Eo, the atomic mass, A, and the scattering angle, θ, by [Rutherford et al., 1930]

display math

[6] This equation becomes

display math

at a scattering angle of 180°. Because the collection of scattered α particles is limited to angles near 180°, the energies of the collected particles vary predominantly as a function of atomic mass, A. The energy, E, is converted to a channel number and plotted versus the intensity in the α spectra.

[7] Turkevich [1961] proposed for the first time, to use equation (1) to obtain compositional information of unknown samples and used this technique to obtain the chemical composition of the lunar surface on Surveyor missions in the late 1960s. The same technique became the basis for similar instruments used in several planetary missions that also includes the alpha proton X-ray spectrometer on the Mars Pathfinder mission in 1997. The now common Rutherford backscattering technique that is used in material sciences and many other applications also has its beginnings in Turkevich's ideas.

[8] The intensity of elastically scattered alphas is determined by the Rutherford cross section as well as the sample composition. The cross section for α back-scattering, which determines the relative probability of scattering for different elements at various angles, has also been described by Rutherford et al. [1930]. They derived the relationship between the cross section at a particular energy, σ(E) in units of cm2/(nuclei. steradian), from Coulomb interactions to be a function of atomic number (Z), energy in MeV (E), and back-scattering angle (θ) as

display math

where 5.184 × 10−27 is a unit conversion factor. This relationship shows that elastic scattering from higher Z elements is more efficient than that from lower Z elements for the same scattering angle. As a result, in the α mode the sensitivity for detection of higher Z elements is much higher than for lower Z elements.

[9] Inelastic α-scattering also occurs, particularly from lower Z elements, most significantly carbon and oxygen, as noted by Bittner and Moffat [1954] and Ferguson and Walker [1940], respectively. The cross sections for inelastic scattering are more variable with scattering angle than those for elastic scattering, and will be described in further detail in section 3.

[10] The collected α back-scattered intensity in a channel (therefore at a particular energy), theoretically predicted from the work of Rutherford et al. [1930] and described by Patterson et al. [1965] is

display math

in which xz is equal to the atom fraction of element z within the sample, Iz is equal to the intensity of the pure sample element z for 100% abundance of each element measured from a standard sample of known chemical composition, and Zz is the atomic number of the element z within the sample. The denominator of equation (4) is similar to a matrix correction for the intensity of the sample. This equation enables one to compute the chemical abundances, xz values, from the intensities measured by the α mode. Also, it enables one to compute the predicted α spectra for a sample with the composition determined from the X-ray mode, assuming stoichiometric oxygen. This equation predicts intensities that are similar to those experimentally measured by Patterson et al. [1965] and Economou et al. [1970]. Economou et al. [1970] and the current Pathfinder APXS alpha spectra find better matches to experimental data with a change in the power to which Z is raised from 0.5 to 0.6.

[11] The intensity of a measured element therefore depends on the abundance of that element (equation (4)), while the maximum energy (endpoint) at which this intensity is measured depends on the atomic number of the element (shown by equation (2)). For an infinitely thin sample, α particles lose no energy traversing through the sample. Therefore the α spectrum of intensity versus energy for a measured element is a vertical line. For a thick target, α-particles from the surface as well as from depths up to tens of μm are back-scattered to the detector. The actual maximum depth from which alphas return can be estimated from previous experimental data compiled by Ziegler [1977]. Alpha particles scattering from deeper in the sample lose energy traversing the sample. Therefore the α spectrum of intensity versus energy for an element within a homogeneous thick target is rectangular in shape. Equations which describe energy loss of α particles with depth are given by Patterson et al. [1965].

1.2. Proton Mode

[12] Some sample nuclei undergo (α, p) reactions, and therefore release protons and sometimes gamma-rays. This nuclear reaction can occur when the kinetic energy of the α particle exceeds the Q value, for which the general nuclear reaction X(a, b)Y is defined as [Tipler, 1987, p. 389]

display math

in which the m values are the masses of the individual atoms and nuclear particles, and c is the speed of light.

[13] Some elements have multiple energies at which (α, p) reactions occur. Since the α particles lose energy as they traverse the sample, some of these proton reactions occur within the sample, rather than at the sample surface. The elements that produce significant proton signals for the APXS include nitrogen, sodium, magnesium, aluminum, silicon, and sulfur. Minor proton signals for phosphorus, chlorine, and potassium are also observed. The intensities of the proton signals from a given element vary as a function of mean sample composition and concentration of the element, with approximately the same intensity dependence as that for the alphas described in the previous section.

1.3. X-Ray Mode

[14] Incident alphas and X-rays from the 244Cm cause sample atoms to emit characteristic X-rays. X-rays from the lower-Z elements from sodium through calcium are generated predominantly by α-particle bombardment, while those from the higher-Z elements are generated predominantly by X-ray bombardment. The relative excitation of characteristic X-rays from alphas and X-rays for the Pathfinder APXS is discussed further in section 3. Characteristic X-rays are generated from depths up to tens of μm by α and X-ray fluorescence, respectively. However, the average sampling depth by the X-rays varies with atomic number due to absorption of outgoing X-rays within the sample. In particular, low-Z X-rays, from sodium and silicon, come from the top few μm while higher-Z X-rays, from iron, come from depths of up to many tens of μm.

2. Laboratory Calibration of the APXS

2.1. Laboratory APXS Experimental Setup

[15] All calibration results discussed in this paper were completed in the Laboratory for Astrophysics and Space Research at the University of Chicago. The laboratory instrument was operated inside a stainless steel vacuum chamber that has two compartments. The upper compartment houses the APXS instrument and it is kept in a constant pressure. The lower chamber has an intake and exhaust valve that can regulate the pressure and introduce gas into the chamber. It can be sealed off from the upper one and provides a way to pressurize the lower chamber during sample changing without affecting the pressure of the upper chamber. Samples are raised to the APXS sensor head when the pressures on both chambers are equalized.

[16] Samples were prepared onto cylindrical aluminum trays that have radii of 75 mm and depths of 1.5 mm. Most analyzed samples, including standard oxides and/or pure elements and geostandards, are fine-grained homogeneous powders which are compacted into sample trays, giving them approximately uniform porosity and evenness. Samples were baked at 115° C for a minimum of two hours before being analyzed in order to remove adsorbed water. The sample on an aluminum tray was loaded into the lower sample chamber, which was then evacuated. Once at vacuum (10−4 Torr), the lower chamber was filled with a simulated Martian atmosphere of 6–8 mbars and then the gate between the lower and upper chambers was opened. The chamber pressure was maintained at 6–8 mbars with a gas mixture identical in composition to that determined for Mars by Owen et al. [1977]. The sample stage was then raised to the cylindrical base of the APXS, which is 42 mm from the base of the collimators. Laboratory data collection time ranged from 10 to 60 hours per sample.

[17] Many standard and sample measurements were completed under ∼vacuum and under Martian conditions before a break in the protective film of the laboratory instrument led to instrument contamination from the Cm sources and therefore inoperability. Many experiments done under ∼vacuum and done with the Muses C-N AXS instrument were therefore also utilized during this calibration, instead of making additional measurements with the inoperable laboratory APXS. Nevertheless, as discussed in the following sections, ample measurements were performed to assure calibration accuracy for application to analyze the Martian Pathfinder samples.

2.2. X-Ray Mode Laboratory Results

2.2.1. Spectral Characteristics

[18] The characteristic X-ray spectral peaks are nearly gaussian with small low-energy tails caused by incomplete charge collection within the X-ray detector. The X-ray spectra were first analyzed with the program AXIL, distributed by Canberra Packard. This program uses a least squares fitting technique with gaussian peaks, which automatically deconvolves overlapping peaks, on an adjustable-order polynomial background. The parameters of the gaussian peaks, such as the peak center positions and the full-width at half-maximum (fwhm), vary as a function of resolution and were therefore determined for each spectrum. The average fwhm for the laboratory instrument ranged from 280 eV for sodium to 310 eV for iron. X-ray spectra were analyzed with AXIL to remove the background and to compute the peak areas, which were then utilized to compute the sample concentrations.

2.2.2. Method of Converting Peak Areas to Sample Concentrations

[19] Concentrations within samples were determined using software written by O. F. Prilutsky, from the Space Research Institute of the Russian Academy of Sciences. This program takes into account the excitation from both the α-particles and the X-rays from 244Cm source as well as X-ray absorption and enhancement (fluorescence) within the sample. The program therefore makes ZAF corrections like modern EDS systems and electron probes.

[20] In order to calculate elemental abundances, this program uses values for X-ray intensities from standards These intensities, normalized to the same collection time, were first collected for characteristic X-ray peaks from standards. The standards are fine-powdered oxides or pure elements having greater than 99.9% chemical purity. The standards used to measure the respective elements within parentheses include NaCl (Na, Cl), MgCO3 (Mg), Al2O3 (Al), SiO2 (Si), CaHPO4 (Ca, P), S (S), K2CO3 (K), TiO2 (Ti), Cr (Cr), MnO2 (Mn), and FeSO4 (Fe). The proportionality of the count rate to the concentration of each element in the standard, instrument geometry, as well as experimentally determined outgoing X-ray matrix effects were then used to measure the composition of the samples being analyzed. A certain oxidation state for each element was assumed, and therefore oxygen was calculated using stoichiometry. This technique is very similar to that used conventionally in scanning electron microscopy. It differs by its incorporation of the fraction of characteristic X-rays generated by alphas and X-rays within a given volume of sample, and the unique geometry parameters of the APXS described by Rieder et al. [1997b].

[21] By measuring the compositions of a number of geostandard powders under Martian conditions, the accuracy of the X-ray mode was determined. The geostandards analyzed include AGV-1 (andesite), BCR-1 (basalt), G-1 (granite), G-2 (granodiorite), and DTS-1 (dunite). The accepted compositions of these reference standards are listed in a compilation by Govindaraju [1994]. Figure 1 illustrates the range of relative percent deviations from conventional values attained by the X-ray mode. The elements sodium, phosphorus, potassium, and iron show reproducible differences between X-ray and conventional analyses. These differences are not a function of neighboring element concentrations, as they would be with peak interference. Instead the differences are systematic and may be due to inaccuracies within the standard library due to slight changes in measurement distance between different standard measurements. Nevertheless, they can be corrected for by multiplying measured values by a correction factor determined from Figure 1. The scatter in the laboratory measurements, given by the standard deviation of the relative percent differences between laboratory and conventional values, is taken as the average laboratory measurement error (counting statistical error is insignificant for these runs because of long counting times). Although this scatter is significant for some elements, such as magnesium, it is representative of the precision of this technique in the measurement of abundances above detection limits. G-2, for example, having 0.75 wt% MgO [Govindaraju, 1994], has a high enough concentration to be used to infer the accuracy of measuring the MgO content. The assigned analytical uncertainties of Pathfinder measurements were determined by combining the statistical errors for each run as well as the average laboratory error. These average laboratory errors and the correction factors are listed in Table 1. Repeated X-ray measurements of geostandards were the same within statistical error, assuring that instrument precision was within statistical error.

Figure 1.

Laboratory X-ray mode analyses under Martian conditions compared with conventional analyses. Points plotted are from analyses of USGS geostandards AGV-1 (analyzed twice), BCR-1, DTS-1, G-2, and G-1 whose abundances have been well determined by a multitude of analytical techniques [Govindaraju, 1994, and references therein]. Some X-ray mode abundances are systematically high or low, having abundances which are reproducibly outside a range of 10 relative% from conventional, for a number of geostandards. Such X-ray mode analyses are circled.

Table 1. Average Laboratory Errors and Correction Factors for X-Ray Measured Elementsa
ElementAverage Laboratory Errors, 1σ (relative%)Correction Factors
  • a

    Correction factors are values multiplied by analyses to correct for systematic differences between conventional and laboratory abundances. The symbol (–) means no correction factor needed, and n.d. means not determined.


[22] Because of the limited number of analyses under Martian conditions of samples containing sulfur, chlorine, manganese, and chromium, accuracy for these elements was determined from samples examined under vacuum. The uncertainty for these elements under vacuum is twenty relative percent, and none have detectable systematic deviations from conventional values among samples examined. However, only one sample was analyzed for chlorine, from which a relative deviation of +20% from the accepted value was obtained. Thus measured chlorine in unknowns may be approximately 20 relative percent too high.

[23] Peak asymmetries from neighboring elements can have significant effects on the analysis of some elements. Al, Si, S, Ca, and Fe had peak asymmetries which contributed to Mg, Al, P, K and Mn, respectively. For example, the peak asymmetry for sulfur must be corrected because it contributes to the phosphorus peak. If uncorrected, the calculated phosphorus abundance will be higher in samples containing more sulfur. Peak asymmetries were corrected for and tested by analyzing samples free of the element to which the peak asymmetry was contributing. The element and its contribution to its low energy neighboring peak were fit with gaussian peaks. The relative proportion of counts contributing to the lower-energy neighboring element was computed and subtracted in further analyses.

[24] The analyses of the geostandards illustrates the ability of the X-ray mode to measure the relative abundances of Na, Mg, Al, Si, P, S, Cl, K, Ca, Ti, Mn, Cr, and Fe with the accuracy listed in Table 1. The Pathfinder data, analyzed by Foley et al. [2003] for these elements, were also corrected by the same correction factors listed in Table 1. Error bars for the Pathfinder analyses were determined by the average laboratory errors as well as from Pathfinder counting statistics errors.

[25] As a demonstration of the accuracy of the calibration technique, a number of samples have been analyzed and compared with conventional sample abundances. This will be illustrated in conjunction with the α-proton mode calibration in section 3.

2.3. Alpha and Proton Mode Laboratory Results

[26] The combined α-proton mode can measure C, N, O, Na, Mg, Al, Si, (S, P, Cl as a group), (K, Ca as a group), Ti, and (Mn, Cr, Fe, Ni as a group). During laboratory measurements of accuracy and correction factors, as well as Pathfinder data analyses, the α-proton mode relied on the X-ray mode for Mg/Si, S/Si, P/Si, Cl/Si, K/Ca, Mn/Fe, and Cr/Fe ratios. With these values from the X-ray mode, the α-proton mode then gives the abundances of the remaining measurable elements. Like the X-ray mode, the α and proton modes rely on spectra of fine-grained oxide standards for their calibration libraries.

[27] Unlike the X-ray mode, both the α and proton spectra contain atmospheric signals from the carbon and oxygen in the Martian atmospheric CO2, which must first be removed before sample analyses. Once the atmospheric contributions are removed, the α and proton signals were analyzed by a least squares fit to yield a result which is partially dependent on the X-ray mode because it relies on the X-ray mode for Mg/Si, S/Si, S/Si, K/Ca, Mn/Fe, and Cr/Fe ratios. Furthermore, the α-mode accuracy is affected more significantly by changes in sample texture than the X-ray mode due to sharpness of the α-spectra, which are more pronounced with solid and polished samples. The correction for textural effects, the removal of the atmospheric signals, and the overall accuracy for the α and proton modes are described in the following two sections.

2.3.1. Alpha Mode in the Laboratory Spectral Characteristics

[28] The α spectra for thick, chemically homogeneous targets, with ≥20 micrometer depth chemical homogeneity, are generally rectangular in shape for the analyzed energy region. For example, assume that a homogeneous slab of Fe2O3 is examined. The highest energy alphas scattered from both iron and oxygen nuclei are from the surface of the sample. Alphas also back-scatter from the top tens of μm within the sample, with the same relative intensities as those scattered at the surface, but at lower returning energies due to α-particle energy loss as it traverses through the sample. The resulting α-spectrum for Fe2O3, therefore consists of two rectangles, one from the α-scattering from iron and the other from oxygen. The rectangular-shaped contributions from sample elements, for AGV-1, are schematically illustrated in Figure 2.

Figure 2.

Alpha-mode laboratory spectrum of an andesite standard, AGV-1, under Martian conditions. The atmospheric signals from the carbon and oxygen of carbon dioxide appear as peaks at low channels. The carbon atmospheric peak is centered at approximately channel 39, while the oxygen atmospheric peaks are centered at approximately channels 60 and 88. Using the atmospheric subtraction technique described within this work and by Foley et al. [2000], the atmosphere-free spectrum for AGV-1 has been computed and is plotted here as well. The α-spectrum for AGV-1 under vacuum is also shown for comparison. Contributions from each element are shaded.

[29] There is some deviation from the expected rectangular shape for elements having resonant scattering of the α-particles at some energies. This occurs due to either geometrical effects caused by changes in scattering angles at different measurement distances or to textural differences between the standard and measured samples. Geometrical effects were considered because of the variations of measurement distance in the Pathfinder analyses, to be discussed in section 3.2. Textural effects were overcome by analyzing standards having textures similar to those of the analyzed samples. Since the resonant features are in the calibration library as well as the analyzed samples, correction is not a problem if the approximate texture of the analyzed sample is known. Resonant features are subdued in powdered samples and are more pronounced in polished samples and rocks, perhaps because of less feature smoothing due to uniform α-particle energy loss in smooth samples. These effects are minor for resonant features in sodium, aluminum, and magnesium alpha spectra. However, there is a significant effect for silicon. As illustrated in Figures 3a and 3b, powders and rocks or polished targets have different resonant features centered at channels 85 and channels lower than 60, due to differences in silicon resonances from textural effects. This change necessitates a change of silicon standard used when examining either rocks or soils. Rocks are more accurately analyzed with the polished standard, while soils are more accurately analyzed with the fine powder standard.

Figure 3.

(a) Comparison of the α-mode spectra for solid and fine-powdered silicon dioxide glass. Solid silicon dioxide glass is referred to as sputter target. The solid glass spectrum shows sharper resonance feature than the fine powder spectrum at approximately channel 88 as well as a higher intensity in the oxygen region. The α-spectra differences with texture emphasize the importance of using standards with textures similar to samples analyzed. (b) Comparison of α-mode laboratory spectra of the natural rock surface and fine-powdered rock of an olivine-bearing tholeiitic basalt, “My2C”, from Joy Crisp (JPL). The natural rock surface shows a significantly sharper resonance feature at approximately channel 88 as well as a higher intensity in the oxygen region due to sharper resonances within that region. Atmospheric Spectral Contributions

[30] Using a simulated Martian gas mix, of 95.32% CO2, 2.7% N2, and 1.6% Ar [Owen et al., 1977; Owen, 1992], we have studied the effect of the Martian atmosphere on the α spectra. There are two general effects, which are illustrated in Figure 2. The first effect is a shift in the endpoints for each element relative to the endpoints under vacuum, due to the energy loss of the α particles in the gas before and after scattering from the sample. The second effect is the contribution to the α-spectra predominantly from carbon and oxygen of the CO2 gas, with only a small α signal from atmospheric nitrogen and no detected α signal from argon. The atmospheric contributions are peaks, due to the low density of the atmosphere, and partially visible in the spectrum collected under Martian conditions in Figure 2. As shown there, only the carbon atmospheric peak is clearly discernible in sample spectra.

[31] In order to evaluate the Martian-condition spectra, a technique was developed for removing this atmospheric signal [Foley et al., 2000]. We have modeled the contribution of the Martian atmosphere to the α spectra in the laboratory using a beryllium target analyzed with the flight-duplicate of the Pathfinder APXS. Since α-particles scattered from beryllium nuclei have energy below the instrument detection threshold, there is no detected signal from the beryllium target. Only the Martian atmospheric signal at the Pathfinder sample distance is detected. The α-spectra were collected for various pressures covering the number density of atmosphere on the Martian surface during the Pathfinder measurements. These spectra, shown in Figure 4, consist of three distinct peaks from atmospheric CO2, the dominant Martian atmospheric constituent. The two lowest-energy peaks are due predominantly to resonantly scattered α particles from atmospheric carbon and oxygen nuclei, respectively. The highest-energy peak is due to low-angle scattering from oxygen nuclei close to the alpha detector.

Figure 4.

Laboratory α-spectra of beryllium at nominal Pathfinder sample distance, under Martian atmosphere of varying pressures. The α-mode cannot detect beryllium because its signal is below the instrument's energy threshold. Therefore beryllium serves as a blank, which permits the examination of the Martian atmospheric signals at nominal distance. All three peaks are from atmospheric carbon dioxide. The first peak, at the lowest channel, is from the carbon and the center peak is from the oxygen. The flatter peak, having the lowest amplitude and located at the higher channel, is due to near-detector low-angle and forward scattering from oxygen nuclei. The amplitudes, widths, and centers of the peaks co-vary linearly with pressure.

[32] The two lowest-energy peaks shown in Figure 4 were fit with an extreme amplitude function, equation (6), with parameters of peak amplitude (ao), center (a1), and width (a2).

display math

The highest-energy peak, also shown in Figure 4, was fit with a complementary error function (equation (7)). These functions provide good fits to the data, as illustrated in Figure 4.

display math

[33] The amplitudes, centers, and widths of the carbon and oxygen peaks vary linearly with number density. Since the carbon peak is the only one directly measurable in the α spectra of laboratory geostandards as well as Pathfinder samples, it was used to infer the parameters of the other two atmospheric peaks. Most of the atmosphere parameters can be computed from the measurable parameters of the carbon peak as illustrated in Figure 5. The high-energy oxygen width, however, is approximately constant with change in number density.

Figure 5.

Relationships between peak parameters from fitting measured α-mode Martian atmospheric peaks with the functions described in the text. Error functions are derived from the least squares fitting routine of Williamson [1968]. Because only the atmospheric carbon peak parameters of a sample spectrum can be measured, as shown previously in Figure 2, these relationships allow the Martian atmospheric contributions of the two oxygen atmospheric peaks to be calculated and subtracted. Oxygen1 is the low energy oxygen peak, which has a larger and more gaussian-like shape than oxygen2, which is the high energy oxygen peak. The dashed boxes indicate the observed values for the carbon atmospheric signal and the inferred oxygen atmospheric values for the Pathfinder APXS data. The relationship between the oxygen amplitudes had to be derived from runs under CO2 rather than Martian atmosphere because of better counting statistics of those runs. Nevertheless, the derived relationships fit the Martian atmospheric peaks as well and are used to accurately analyze geostandards run under Martian conditions.

2.3.2. Proton Mode in the Laboratory

[34] The proton spectra have several peaks due to (α, p) reactions occurring within samples at different energies for each proton-producing element. The computed and/or measured values for the proton library spectra of Na2O, MgO, Al2O3, SiO2, and S are shown in Figure 6. This proton library consists of the collected intensities from analyzing standards of chemically pure and homogeneous fine-grained simple compounds (NaCl, MgO, Al2O3 and SiO2) and elemental sulfur. There is no proton signal from the oxygen within these oxides. The proton intensity is proportional to the concentration of the proton emitters: Na, Mg, Al, Si, and S. The spectrum for Na2O is computed using the proton spectrum of NaCl for which there is no significant chlorine contribution.

Figure 6.

Proton-mode laboratory spectra. The multiple spectral signals at different channels, proportional to different energies, are due to proton release from sample nuclei (and Martian atmosphere in the case of nitrogen) from (α, p) reactions. The Na2O curve was calculated from measurement of NaCl.

[35] The only significant effect of the Martian atmosphere on the proton spectra is to introduce a small contribution from the 2.7% of atmospheric nitrogen. This nitrogen contribution was measured by pointing the APXS toward the Martian sky to measure only atmosphere during a period of comparable atmospheric number density to the Pathfinder sample measurements. The proton spectrum of Martian atmosphere is shown in Figure 6 and was used to correct Pathfinder sample measurements.

2.3.3. Alpha-Proton Accuracy in the Laboratory

[36] The α and proton data were merged and analyzed as one set of data and therefore we refer to their results as coming from the α-proton mode. Merging the α and proton data means creating a larger single spectrum in which the α mode data are in the first ∼160 channels and the proton data are in the last ∼200 channels of the new spectrum. The two modes were merged because some elements have lower and less distinctive intensities in the α-mode and higher and/or more distinctive intensities in the proton mode and vice versa. For example, the sodium signal for the α-mode is typically difficult to decipher from the large silicon signal in basaltic to andesitic samples, but is easily seen in the proton mode. The silicon and sodium α-spectra are at similar energies with a much greater intensity from silicon than sodium. In contrast, the proton signals from silicon and sodium are at different energies. Because of this, the accuracy of sodium measurement is better for the proton mode than for the α-mode. The other element contributing to the sodium region of the proton-spectrum is aluminum. In order to assure that sodium is calculated accurately from the proton mode, the Al/Si ratio in the alpha-proton mode is set from the X-ray data. The merging of the α and proton spectra was done for each calibration standard as well as for each analyzed sample. A least squares program was then used to calculate the abundances in the unknown using the library, consisting of merged α and proton spectra from homogeneous fine-powdered oxides and/or elements. The least squares program fits the analyzed spectrum with the intensities measured in the standards and converts the measured intensities to sample concentrations using equations described in section 1.1. Accuracy of Major and Minor Elements

[37] We determined the accuracy and precision of the α-proton mode by analysis of geostandards. Repeated α-proton measurements of geostandards were the same within statistical error, assuring that instrument precision was within statistical error. Figure 7 illustrates the range of relative percent deviations of α-proton mode data from conventional values for the same geostandards that were analyzed with the X-ray mode. Only those elements independently measured by the α-proton mode are illustrated in Figure 7. Some elements (sodium, aluminum, potassium, calcium, and iron) show reproducible systematic differences between α-proton and conventional analyses. These differences may be due to inaccuracies in the library calibration values due to slight variations in distance of measurement for the various library standards. The Pathfinder data were corrected by multiplying by a correction factor determined from Figure 7. Error bars for the Pathfinder analyses include the errors of these factors and the counting statistics errors listed in Table 2.

Figure 7.

Laboratory α-proton mode independent analyses under Martian conditions compared with conventional analyses. Points plotted are from analyses of USGS geostandards AGV-1 (analyzed twice), BCR-1, DTS-1, G-2, and G-1 whose abundances have been well determined by a multitude of analytical techniques [Govindaraju, 1994, and references therein]. Some α-proton mode abundances are systematically high or low, having abundances which are reproducibly outside a range of 10 relative% from conventional, for a number of geostandards. Such α-proton mode analyses are circled.

Table 2. Average Laboratory Errors and Correction Factors for α-Proton Measured Elementsa
ElementLaboratory Error, 1σ (relative%)Correction Factors
  • a

    Correction factors are multiplied by analyses to correct for systematic differences. The symbol (–) means no correction factor needed, n.d. means not determined, and (*) means the element is not measured independently by the α-proton mode.

Iron121.11 Detection of Water in Samples

[38] The series of plots shown in Figure 8 demonstrates the α-mode detection of sample-bound water. Since geostandards containing water were examined only under vacuum conditions, their raw spectra had to be shifted to match Martian-condition endpoints in order to test the Martian-condition library. This also necessitated a change in the magnesium proton library value which was used in the combined α-proton analyses, which is different from the vacuum value due to a difference in the bombarding α particle energy.

Figure 8.

Series of α-mode laboratory spectra showing the detection of excess oxygen, present in samples having greater than 1 wt% water. (a) The raw spectrum minus atmospheric signals for a basalt standard, BCR-1, and a spectral fit using the α-program with the library calibration values assuming the oxidation states as listed by Govindaraju [1994]. No excess oxygen was detected. (b) (1) The raw spectrum for a serpentinized peridotite standard, PCC-1, shifted to match the spectral endpoints of the Martian library as discussed in the text; (2) the synthetic spectrum from the α-program for an anhydrous PCC1 assuming oxidation states as listed in Govindaraju [1994]; (3) the α-program fit which includes oxygen bound as water. There is a significant excess of oxygen signal when sample water is present. (c) A plot similar to a and b, for the Murchison meteorite sample. Oxidation states used in the synthetic spectra are from Jarosewich [1971], as is the water in the spectrum of hydrous Murchison. There is a significant signal of excess oxygen when sample water is present.

[39] The raw α-spectra for geostandards were compared with calculated α-spectra assuming stoichiometric oxygen, excluding sample-bound water from their individual analyses. Also, the raw α-spectra were compared with the α-spectral fit from the α-proton program including all measured elements. This was done to observe the magnitude of the oxygen signal due to sample-bound water in geostandards. Figure 8a shows the raw laboratory spectrum for BCR-1, minus atmosphere, compared with the α-spectrum assuming stoichiometric oxygen excluding water. Because BCR-1 contains less than 1 wt% water [Govindaraju, 1994], the raw and the synthetic spectra for BCR-1 are very similar. Figure 8b shows the raw and synthetic spectra for PCC-1 standard. Because PCC-1 contains 4.71 wt% water [Govindaraju, 1994], the raw spectrum for PCC-1 is much higher in the oxygen region than the calculated α-spectrum excluding water. A calculated spectrum including water matches the raw spectrum well. Figure 8c shows the raw laboratory spectrum and the synthetic spectra calculated with and without water, but including carbonate, for the Murchison CM2 chondrite (abundances for Murchison are from Jarosewich [1971]). The large differences between the anhydrous synthetic spectra of PCC-1 and Murchison and their raw spectra and the general average oxygen error (described in the next section) give a water detection limit of ∼1 wt%. Oxygen Accuracy and Inferred Water Accuracy

[40] The α-mode determines sample oxygen content by using library spectra of oxygen derived from many measured oxides. Measured oxide spectra include MgO, Al2O3, SiO2, P2O5, CaO, TiO2, and Fe2O3 (although a measured Fe spectrum was used to analyze sample Fe content because of the variation in the Fe oxidation state within samples). Simulated oxide spectra include Na2O, K2O, and pure O. The calculated library spectrum for pure oxygen, for example, is derived from the general equation for α intensity (equation (4)) and the measured intensities of Fe2O3 and Fe, TiO2 and Ti, and SiO2 and Si. Its spectrum is an average oxygen value acquired from these oxides and elements. In particular, the pure oxygen in the library enables the determination of any oxygen excess over and above the required amount from stoichiometry of all the elements in the library not in oxide form. The only major element whose library value is not an oxide is iron, while minor elements whose library values are not oxides are carbon, nitrogen, and sulfur. Therefore any oxygen excess yields the total oxygen bound with hydrogen, carbon, nitrogen, sulfur, and iron in a given sample. Samples having low abundances of these elements do not test the accuracy of the residual oxygen library standard. This is the case for most samples analyzed in the laboratory under Martian conditions, shown in Figure 7. In particular, the measurements shown in Figure 7 do not demonstrate the accuracy of inferring sample-water content.

[41] In order to test the oxygen value in the calibration library, samples run under vacuum conditions which contain more of these elements were shifted to match the spectral endpoints under Martian conditions and were then analyzed with the Martian library. As shown in Table 3, shifted spectra of AGV-1 and DTS-1 both yield results of accuracy comparable to those attained under Martian conditions. This comparable accuracy is convincing evidence that vacuum spectra may be shifted and analyzed with the Martian library to further test the Martian library. PCC-1 and Murchison are both samples containing appreciable amounts of water, and have been analyzed under vacuum conditions. By shifting PCC-1 and Murchison to the spectral endpoints for Martian conditions, both samples were examined in order to determine the accuracy of oxygen measurement with samples containing water. The results listed in Table 3 show that the APXS is able to measure the water content in samples for which the oxidation states of sample elements are known, with the level of accuracy predicted from the propagation of errors for each oxide, yielding a bulk oxygen relative error of approximately 1 relative%. In order to compute the abundances listed in Table 3, data from both the X-ray and α-proton modes were utilized using the same routine as used to acquire the final results for the Pathfinder abundances [Foley et al., 2003]. This demonstrates the accuracy of our calibration routine to analyze diverse samples, which have varying amounts of sample-bound water, with the APXS.

Table 3. APXS α Analyses Under Vacuum Showing Accuracy of α-Inferred Water Contenta
Oxide, wt%DTS-1AGV-1PCC-1Murchison
Convα/pΔ%Convα/pΔ%Convα/pΔ%(at %)Convα/pΔ%
  • a

    Asterisk (*) denotes oxides that rely on ratios from the X-ray mode. Δ values are relative percent deviations from conventional. Oxide wt% values are normalized to 100% without volatiles. Some trace elements are not listed. Bulk oxygen at%, O atom%, is compared with the bulk conventional analyses, including volatiles normalized without hydrogen. The bound water content, H2O+, is computed using the bulk oxygen content and the conventional oxidation states determined for each sample. Analyses for DTS-1, AGV-1, and PCC-1 are from Govindaraju [1994]. The analysis for Murchison is from Jarosewich [1971] normalized to 100%, including volatiles without hydrogen. Murchison values are converted to atomic percent abundances due to the multiple oxidation states of carbon, sulfur, and iron.

O at%57.0556.40−1.162.3261.75−0.9160.0960.550.8Fe9.2710.2010.01
H2O+ wt%0.00−0.700.78−0.364.715.2812.1Ni0.550.50−7.87
          H2O+ wt%8.9510.5117.43

[42] The α-mode oxygen measurements, shown in Figure 9, yield an accuracy at the one σ level of 1 relative percent both for geostandards containing <1 wt% water and for samples containing abundant water. Hydrogen is the only rock-forming element that cannot be detected directly by the APXS. However, its presence is inferred by excess oxygen above that which can be bound with all other detected elements in their highest oxidation states.

Figure 9.

Accuracy of laboratory α-mode bulk oxygen analyses for geostandards and Murchison meteorite. Atomic% from the α-mode is plotted versus the atomic% of the conventional oxygen analysis normalized on a hydrogen-free basis to compare with the α-mode results (because the α-mode cannot measure hydrogen). This plot shows geostandards containing <1 wt% water, including AGV-1, BCR-1, G-1, G-2, and DTS-1 all measured under Martian conditions, and samples containing water, as labeled, measured under vacuum. The bulk oxygen measurement by the α-mode has a 1σ error of one relative percent calculated from the analyses of multiple geostandards. The errors for conventional analyses are average 1σ errors for total oxygen calculated by propagating the reported errors for each oxide and converting it into the total oxygen error. Carbon Detection Limit

[43] Also important for understanding the volatile content of the Martian soils and rocks is measurement of sample carbon. Carbon is detectable under vacuum by the α mode if present above 0.3 wt%, as demonstrated in Figure 10. The contribution of sample carbon to the α-spectra is in channels 15–45 for laboratory samples examined in vacuum. Under Martian conditions carbon, if present, will appear in channels 15–32 with a similar detection limit, as a rectangle-like feature on top of the rest of the spectrum. The detection limit under Martian conditions is thought to be similar to those under vacuum because of the opposing effects of increased sample carbon signal due to increased mean scattering angle with increased measurement distance for Pathfinder relative to the laboratory and increased interference by atmospheric carbon signal (also due to increased mean scattering angle).

Figure 10.

Laboratory α-spectra showing carbon detection down to 0.3 wt% carbon under vacuum. The carbon abundance proceeded by “∼” is an informational value (meaning its accuracy is not as well known as a recommended value), while other carbon abundances are recommended values. The Murchison and Allende carbon abundances are from Jarosewich [1990]. The BCS301 carbon abundance is from Bureau of Analyzed Standards, Ltd. The G-1 carbon abundance is from Govindaraju [1994]. Overall Alpha-Proton Mode Accuracy and Importance

[44] The calibration of the α-proton mode has enabled the semi-independent measurement of C, O, Na, Al, Si, Ca, and Fe in laboratory geostandards. The α-proton mode may function as a check of the X-ray mode results for the Pathfinder samples for major rock-forming elements. Acquiring the sodium abundance from the α-proton mode is extremely important, since the statistical error for it is significantly better than that from the X-ray mode during Pathfinder measurements, as discussed further by Foley et al. [2003]. Furthermore, the α-proton mode can yield measurements of the amounts of carbon and inferred sample-bound water which may also be present in some Pathfinder samples. For the Pathfinder analyses [Foley et al., 2003], the α-proton and X-ray results (with Na2O from the α-proton mode) are first compared on a water-free basis in order to verify agreement between the modes. As discussed further in the chemical results paper [Foley et al., 2003], the X-ray mode abundances for all elements other than sodium and oxygen (which are acquired from the α-proton mode) are then used to compute the preferred sample abundances. The X-ray mode is primarily used for most abundances because of its lower statistical error for most elements, as discussed in the results paper [Foley et al., 2003].

3. Further Verification of Accuracy for Alpha-Proton-X-Ray Combined Modes

[45] Using the calibration technique described in sections 1 and 2, the compositions of a number of samples, in addition to those previously mentioned, have been determined to verify the accuracy of the α-proton-X-ray modes (Table 4). Although samples listed in Table 4 were obtained under ∼vacuum, instead of under Martian conditions, the same calibration technique was utilized to analyze the samples and these analyses are therefore still approximately representative of the overall accuracy of the APXS.

Table 4. Examples of APXS Analyses With the Calibrated α-Proton and X-Ray Modes Compared With Conventional Valuesa
wt%My2C PowderMy2C RockAGV-1 USGS Standardatm%BCS 301 Standard
APXS1σ Err.XRFAPXS1σ Err.APXS1σ Err.Conv.Err.APXS1σ Err.Conv.
  • a

    Key: All oxide wt% analyses are normalized to 100% without water for comparison. The APXS values from BCS 301 are normalized to 100% with 1.13 atm% hydrogen. My2C is an olivine tholeiite prepared by Joy Crisp (JPL) as an approximately homogeneous powder. However, it is not a geostandard, and some deviations between the XRF analyses from Bondar Clegg, Inc. (normalized to 100%) and the APXS analyses may be due to slight heterogeneities within the sample (especially between the rock and the powder). The sulfur abundance for My2C is from SEM analyses at the University of Chicago. AGV-1 abundances are from Govindaraju [1994], while BCS 301 (a British iron-ore standard) abundances are from the Bureau of Analyzed Samples.

H2O0.10.5n.d.0.71.01−0.40.510.78n.d.Hn.d. 1.13
S0.040.010.03*  Si3.450.072.88
Cl0.0  P0.250.040.26
Cr2O30.  Ca8.540.609.42

[46] The sample abundances in Table 4 are for an olivine-bearing tholeiitic basalt (“My2C” from Joy Crisp at JPL) analyzed as a rock and a powder, a fine-powdered USGS andesitic standard (AGV-1), and a fine-powdered carbonate-rich British standard (BSC 301). As will be done with the Pathfinder samples, the X-ray mode abundances (which have lower statistical errors associated with them) are utilized for all major and minor measured elements except for sodium, whose abundance is acquired from the α-proton (which has lower statistical error). Also, the oxygen and carbon abundances acquired from the α-proton mode are used to determine the amount of sample water and carbonate. These results (Table 4) verify that for samples of varying composition as well as texture (rock and fine powder), the APXS chemical abundances mostly agree with the conventional values within 1σ error. Since My2C is not a certified homogeneous standard, some slight deviations may be due to sample heterogeneity. In particular, since the rock is a natural sample, heterogeneity on the μm-scale (which is the measurement scale of the APXS) is expected. However, these minor differences would importantly not, for example, change the igneous classification of the samples or affect the computed sample-water abundances.

4. Corrections for Differences Between the Laboratory and Flight Instruments

4.1. Protective Film and Grid Covering the Curium Sources

[47] One difference between the laboratory and flight instruments is the presence of a protective thin alumina/VYNS (∼1200 Å alumina/organic polymer) film supported by a stainless steel grid in the laboratory instrument. This protective grid is directly in front of the curium sources and is used in the laboratory as a barrier to protect the instrument chamber and samples from curium contamination. The steel grid is thicker than the range of the α-particles and therefore prevents a fraction of the α-particles from reaching the sample, but causes an insignificant absorption of the X-rays from the curium sources. Therefore relatively fewer alphas will scatter or undergo (α, p) reactions in the presence of the protective grid.

[48] Absorption by the grid causes the α and proton spectral signals for laboratory standards to be relatively lower than those in the Pathfinder samples. Because all the standard intensities are reduced by the same factor, this does not affect the calculation of the absolute sample abundances for the α and proton modes. The X-ray mode Pathfinder abundances, however, are affected by the change in the proportion of alphas and X-rays producing X-rays in the sample. Characteristic X-rays from the sample are excited by both alphas and X-rays from the curium sources, as previously discussed in section 1.3. With a decreased α beam, there is less characteristic X-ray production from the sample. The proportion of generation of sample X-rays by α-particles and X-rays has been determined experimentally with the laboratory APXS by analyzing pure oxide targets with and without a 28 μm mylar foil over the curium sources. This mylar absorbs all of the α-particles, and thus prevents them from interacting with the sample. By comparing the laboratory X-ray yield with and without the mylar foil, the ratios of the X-rays produced by alphas and by the X-rays from the 244Cm source were calculated with results shown in Figure 11. X-ray yields were measured for sodium, silicon, chlorine, titanium, and iron. Other yields were approximated from the best-fit curve to the experimental data shown in Figure 11.

Figure 11.

Fraction of sample characteristic X-rays generated by α-particles for different atomic numbers. The plot shows that elements lower than manganese in atomic number have most of their characteristic X-rays generated by α-particles, while elements higher than manganese in atomic number have most generated by X-rays. The curve used to infer the fraction for nonmeasured elements is an exponential fit to the data points.

[49] The standard library, which contains the intensities corresponding to a certain atomic percent of each element, therefore had to be modified by the difference in α intensity on Mars, due to the absence of the grid on the flown instrument. The fraction of characteristic X-rays generated by α-bombardment was used to increase the reference library values by an appropriate amount. Since the grid reduces the intensity of α-particles by 15%, its surface area, it reduces the amount of X-ray generation by α-particles by 15%. The new library values to apply to the Pathfinder data for the standard intensities are therefore increased by 15% of the fraction of characteristic X-rays produced by alphas.

[50] Wänke et al. [2001] has also corrected the original abundances acquired by Rieder et al. [1997a] to compensate for the presence of the protective grid and film. The details of their corrections are not described; however, the Wänke et al. [2001] Pathfinder X-ray mode results are compared with ours in the accompanying paper of Foley et al. [2003].

4.2. Changes in Measurement Distance

[51] The Pathfinder measurement distances differed from the laboratory calibration distance. The effects of distance change on APXS measurements have been examined using both theoretical and experimental techniques with the assumption that the sample surface was perpendicular to the axis of the sensor head of the APXS.

4.2.1. Theoretical Spectral Changes Due to Distance General Effects on X-Ray and α Abundances

[52] Alpha spectra show nonproportional effects, those which cause a change in the signal of an element relative to the other elements detected, while there is no significant nonproportional effect on the X-ray spectra for samples run at different distances. The only nonproportional effect on the X-ray spectra is due to slightly increased absorption of lower energy X-rays in the atmospheric CO2. The equation used to predict the relative changes in intensity of X-rays traversing through matter is [Goldstein et al., 1992]

display math

where I is the final intensity, Io is the initial intensity, μ/ρ is the mass absorption coefficient, x is the distance traversed, and ρ is the density. The magnitude of this nonproportional change is small. As an example, these changes were calculated for a vertical measurement distance change of 15 mm, assuming a Martian surface atmospheric pressure and temperature of 6.8 mbars and 230 K. These conditions represent the average of the pressures and temperatures measured by the meteorological sensors on the Pathfinder lander. The results indicate that Na would decrease by 6 relative%, Si would decrease by 1.5 relative%, Ca would decrease by 0.15 relative%, and Fe would not change. The precise distance from the measured samples to the X-ray detector at the Pathfinder landing site is not known. Therefore corrections for these potential systematic intensity changes have not been made. These changes in intensity are significantly below the analytical uncertainties for the Pathfinder samples, which will be discussed in the final chemical results paper [Foley et al., 2003]. The largest effect on the X-ray intensities is the proportional intensity reduction with increased sample distance due to a 1/r2 intensity reduction with distance. This proportional change in X-ray intensities does not affect the relative abundances computed with the X-ray data.

[53] The four factors potentially affecting α-scattering with distance are changes in the number of striking α-particles going from the curium to the detectable sample, number of α-particles going from the sample to the detector, angular dependence of Rutherford scattering, and angular dependence of resonant scattering. Of those processes, only resonant scattering changes cause nonproportional changes in the α-spectra and hence changes in the apparent relative abundances from the α-mode. The other three processes may cause the intensities of the signal from each element to change, in proportion with one another, with increased distance. The changes in Rutherford and resonant scattering are due to changes in the α-scattering angles with distance. While the distribution of scattering angles from a sample was defined for the laboratory APXS by maintaining a distance of 42 mm from the collimator base to the sample, changes in measurement distance affect the angular distributions of α-particles back-scattered from the sample into the detector. Resonant Scattering

[54] Elements of atomic number >20 scatter via elastic Rutherford scattering alone for alphas having energies near 5.8 MeV [Patterson et al., 1965]. Alphas of this energy can overcome the Coulomb barrier for Z < 20, allowing inelastic non-Rutherford scattering to occur [Patterson et al., 1965]. Once the Coulomb-barrier is overcome, the alphas interact with and scatter from within the nucleus. As the α energy is reduced upon traversing a sample, some elements near atomic number equal to and less than 14, such as silicon, may undergo resonant scattering.

[55] This non-Rutherford, resonant nuclear scattering for alphas with energies less than or equal to 5.8 MeV causes the large atmospheric signals and also contributes to the α-spectra of solid samples. The probability of resonant nuclear scattering of α particles, also referred to as the nuclear scattering cross section, varies with the α-particle energy and scattering angle. The distribution of sample scattering angles does not vary with fixed instrument geometry. For fixed instrument geometry, the same resonances are present in the oxide and/or element library standards as are in the analyzed geostandards. However, change in the measurement distance affects the distribution of sample-scattering angles. This changes the scattering cross section and the intensity of the α sample-signal for such elements.

4.2.2. Experimental Quantification of Distance Effects

[56] Ideally, the effect of distance on the intensities of the α-spectra would be directly measured with the Pathfinder laboratory instrument. However, the Pathfinder laboratory instrument suffered a break in the protective grid upon pressurization of the sample chamber leading to contamination by curium. Therefore a similar instrument initially developed for the MUSES-CN asteroid mission, called the AXS, was used to determine the distance effects.

[57] The same sample preparation and instrument chamber described in section 2.1 for the APXS was used to analyze samples with the AXS at various distances from a nominal of 11.25 mm. AXS analyses show no significant change in elemental abundances in the X-ray mode with distance, as predicted from the proportional decrease in X-ray signals due the 1/r2 law. Furthermore, the analyses reveal that the primary process affecting the proportional decrease in α signal with distance is the 1/r2 law, while resonant scattering significantly affects the nonproportional spectral change of some elements.

[58] The predictability of distance measurement with the 1/r2 behavior of the Fe α-signal is essential because the Mars Pathfinder samples were analyzed at unknown distances. The alpha signal is used, rather than the X-ray, because the alpha detector is located in the center of the instrument with the detector parallel to the sample surface for a flat nontilted sample, and its signal is therefore more representative of that coming from the average sample distance. Theoretically Predicted Changes in α-Intensity That Are Not Experimentally Observed

[59] Changes in the number of sample-incident alphas and Rutherford scattering are theoretically predicted to cause a 5 relative percent increase in measured α intensity at a distance of 20 mm from nominal. This systematic effect was not experimentally observed for samples BCR-1 and the standard glass sample as shown in Figures 12a and 12b. The calculated change in the iron intensity of a sample, using only the 1/r2 law, is the same as the measured change within ±9 relative percent (Figure 12a). Furthermore, errors associated with measured distances, calculated by repeat measurements, are on average 10 relative percent. Therefore the calculated (by 1/r2) and measured distances are the same within their respective errors (Figure 12b). For the Pathfinder instrument, theoretically predicted changes in the number of incident alphas and Rutherford scattering may cause a +7 relative percent change in measured α-intensity at a distance of 15 mm from nominal. This theoretical change of +7 relative percent in intensity is similar to that predicted for the MUSES AXS instrument of +5 relative percent. Since the calculated change of +5 relative percent was not detected experimentally, it is reasonable to assume that the +7 relative percent change is also not detectable for the Pathfinder APXS. Therefore, for both the MUSES AXS and the Pathfinder APXS it is assumed that the distance can be calculated using the 1/r2 behavior of the iron alpha intensity.

Figure 12.

(a) AXS measured laboratory iron intensity change with measurement distance compared with the calculated change using the 1/r2 law. (b) Comparison of AXS measured and calculated measurement distances. Because measured and computed intensity changes with distance are approximately the same, distances can be computed using the 1/r2 law. Computed distances for BCR-1 and a standard glass sample are within an average of ±9 relative% of measured distances. Measured distance errors are approximately ±10 relative percent and are calculated by comparing the iron α-intensities for a sample in different analyses. Resonant Scattering

[60] AXS analyses revealed that carbon, oxygen, and silicon are the major elements affected by distance change, as illustrated in Figures 13a–13c. The magnitude of the change in the carbon signal with distance is illustrated in the α-spectra of titanium carbide at two distances shown in Figure 13a. The scattered-α intensity for titanium is from Rutherford scattering alone. Therefore the normalization of the two spectra to the titanium levels shows the relative increase due to non-Rutherford scattering in the carbon signal. The changes in the oxygen and silicon signal with distance are illustrated in the silicon dioxide spectra of Figure 13b. Since the silicon dioxide spectra are normalized to a part of the silicon spectrum due to Rutherford scattering alone (near channel 160), differences shown in Figure 13b reveal that the more distant silicon spectrum has a resonant feature in channels 170 through 200, corresponding to a silicon scattering resonance at approximately 5.2 MeV, and the entire oxygen signal increases due to oxygen resonant scattering. The evidence for lack of non-Rutherford contribution near channel 160 of the silicon spectrum is illustrated in the spectrum for pure silicon shown in Figure 13c. As illustrated, the more distant spectrum for silicon is within the error bars of the spectrum collected at the initial distance (at all channels less than 170 with some slight deviation in channels 50–100 within measurement error) normalized only to compensate for the 1/r2 law.

Figure 13.

(a) TiC AXS α-spectra, normalized to Ti level, showing a large nonproportional increase in carbon counts with measurement distance. (b) SiO2 AXS α-spectra, normalized to Si counts, showing both oxygen and silicon nonproportional increases with measurement distance. (c) Si AXS α-spectra, normalized to the calculated silicon level in a nonresonant scattering region (near channel 160), showing that significant nonproportional increases in silicon occur only in some channels centered near channel 190.

[61] Furthermore, Coban at al. [2000] also experimentally observed and theoretically predicted no significant non-Rutherford scattering for silicon with α-particles having energies less than 5.8 MeV apart from the angle-dependent resonance at 5.2 MeV. Previous experimental observations and theoretical calculations for significant large-angle resonant α-scattering variation for α energies less than or equal to 5.8 MeV for carbon, oxygen, and silicon are discussed by Bittner and Moffat [1954], John et al. [1969], and Coban et al. [2000], respectively.

[62] The intensities of the resonant scattering of carbon, oxygen, and silicon increase with larger scattering angles, which are more prevalent for more-distant samples. This intensity change with distance must be taken into account in calculating elemental abundances from the α-spectra. Because no carbon was detected in the Pathfinder rocks and soils as discussed by Foley et al. [2002, 2003], increases in carbon intensity with increased measurement distance are not as pertinent as those in oxygen and silicon.

[63] The AXS analyses demonstrate no significant effects of inelastic scattering for the pertinent distance change for other major rock-forming elements including sodium, magnesium, aluminum, and calcium. The lack of changes in resonant scattering for magnesium, aluminum, and calcium are illustrated in Figures 14a–14c, respectively. Each of these figures shows the spectrum for the element or oxide measured at two distances. Each figure also shows the spectrum calculated from the spectrum of the closer run reduced by the intensity change predicted due to the 1/r2 law. As shown, there is no significant difference between the measured spectra and those calculated from the 1/r2 law alone. The small difference observed for magnesium is within error of measurement and does not significantly affect analyses. If there had been a resonant nuclear scattering change for the change of distance, there would be a difference between the predicted and measured spectra, as was seen in the spectra for silicon in Figure 13c.

Figure 14.

(a) Measured AXS α-spectra of Mg at the nominal distance +10 mm, and at +15 mm compared with the calculated +15 mm spectrum for Mg, assuming 1/r2 decrease in Mg signal with distance. The systematic difference between these two spectra may be accounted for by errors in measurement distance of approximately 10 relative%, as discussed previously. (b) AXS α-spectra of Al at the nominal measurement distance and at +10 mm compared with the Al spectrum for +10 mm, calculated assuming only 1/r2 decreasing Al signal with distance. The laboratory spectrum for Al +10 mm and the calculated spectrum for +10 mm are within 2σ of one another and may also be accounted for by measurement distance uncertainty. (c) AXS α-spectra of CaCO3 at the nominal distance and at +10 mm compared with the spectrum for +10 mm, calculated assuming only 1/r2 decrease in Ca signal with distance. The Ca region of the spectrum for +10 mm and the calculated spectrum for +10 mm are within 2σ of one another. The differences in the C and O regions are due to resonant scattering changes for those elements.

[64] Since both oxygen and silicon α-signals change with distance, a technique for calculating their changes was developed which could be applied to the analyses of the Mars Pathfinder data. In particular, new library values appropriate for the α-signal at different measurement distances have been computed using oxide samples and tested using BCR-1. The change in the silicon signal with distance was determined by examining a standard glass sample containing approximately 46 wt% SiO2, 16 wt% MgO, 11 wt% Na2O, and 25 wt% Fe2O3, as well as a SiO2 sputter target. The percent change in silicon resonant scattering signal within channels 169–200 with sample distance is shown in Figure 15a. In the absence of specific cross sections as a function of angle for silicon, the experimental curve shown in Figure 15a was used to deduce the change in the silicon α-signal with distance.

Figure 15.

Change in the α-intensity of the resonant feature of silicon with distance, as measured with the AXS for silica and a standard glass sample. An exponential fit was used because the relative change from nominal approaches a constant at greater measurement distance, at which the scattering angle distribution no longer changes.

[65] The inelastic scattering cross section for oxygen varies with α-energy and scattering angle. John et al. [1969] examined this variation for α-particles having energy from 5 to 10 MeV. These data together with the distribution of α-scattering angles with sample distance were used to calculate the percent increase in the oxygen signal due to resonant scattering.

[66] This angular distribution change was computed using a program that created arrays of vectors from the 244Cm sources to the surface of the sample and returning through the collimators to the detector. All vectors hitting the walls of the collimator were excluded. The number of scattering α-particles seen by the detector at angles in 1° increments was then converted to a relative percent of detector-striking α-angles for each distance.

[67] The calculated distributions of α-scattering angles for three distances for both the Pathfinder and the AXS instruments are illustrated in Figure 16. The nominal distances, which are the closest, show the largest contributions from lower-angle scattering and are broadest. As distance increases, the average scattering angle approaches 180° and the distribution narrows. The distribution of scattering angles is then combined with the cross-section data of John et al. [1969] to determine the distances for which both the Pathfinder and the AXS will have identical oxygen intensities for the same sample.

Figure 16.

Calculated distributions of back-scattering angles for AXS and Pathfinder APXS with measurement distance.

[68] The oxygen intensity for the Pathfinder APXS at its nominal distance is equal to that of the AXS at its nominal distance plus 9.7 mm as shown by the dashed lines in Figure 17. Similarly, APXS plus 15 mm is equivalent to AXS plus 18.3 mm. Alpha spectra for BCR-1 for the Pathfinder and AXS plus approximately 10 mm, shifted only for gain correction, are plotted in Figure 18 to illustrate the equivalence of the α-spectra at that distance.

Figure 17.

Comparison of theoretical oxygen α-intensity changes with distance for MUSES AXS and Pathfinder APXS. Each point is a calculated value from the computer model. Exponential fits were used to infer values between computed ones. The Pathfinder and MUSES AXS instruments have equivalent oxygen intensities at measurement distances for which their sum of the product (of the angular cross sections and distributions) are equivalent. For example, the Pathfinder instrument in its nominal position yields the same oxygen intensity as the MUSES AXS at its nominal distance plus 9.7 mm.

Figure 18.

BCR-1 α-spectra for AXS +10 mm and for Pathfinder at nominal distance. Spectra are not shown below channel 40 because of electronic threshold differences in that region.

[69] The change in the oxygen signal with distance, shown in Figure 19, was calculated by examining the laboratory α-spectra for SiO2 and computer modeling results. The latter were used so that the AXS instrument results can be later applied to analyses of data collected with the APXS instrument. The laboratory results agree within 1σ error with the computer program results. The offset between the two curves may be due to the larger uncertainties, up to 8 relative percent, of the John et al. [1969] measurements at low scattering angles. This means the initial value for the cross sections could, for example, be too high and therefore the ratios of the more distant oxygen intensities to the initial oxygen intensity could each be too low. Alternatively, the initial laboratory spectrum for silicon dioxide may be different by its statistical error as well. Nevertheless, the computed relative changes in intensity for the measured and calculated values will not be affected by this type of error. The slopes of the two curves in Figure 17 are within 2 percent of one another over the distance range of interest, between +10 to +20 mm from the base of the cylinder. Therefore the offset between the laboratory and computed curves does not significantly affect the accuracy of the calculation of the change in oxygen signal using the computed curve.

Figure 19.

Experimental and theoretical oxygen intensity change with distance for the AXS. The offset between the two curves may be due to increased error in the cross sections from John et al. [1969] at smaller angles. The slopes of the curves, are within 2 relative% of one another so that the oxygen change may be calculated using this technique within 2 relative%.

[70] The calculated and experimental distance effects in oxygen and silicon, respectively, were then used to correct the α-library. The Pathfinder α-library was used to analyze BCR-1 with the AXS at distances ranging from AXS nominal +10 mm to AXS nominal +20 mm. The AXS spectra for these runs are illustrated in Figure 20, which shows the increases of oxygen and silicon intensity over the tested range. Table 5 shows the effect of distance on the chemical abundances computed using the original Pathfinder library, the Pathfinder library modified for the change in oxygen, and the Pathfinder library modified for the change in both oxygen and silicon. The analysis at AXS nominal plus 10 mm, corrected for both oxygen and silicon resonance effects, should be equivalent to the Pathfinder APXS at nominal calibration distance. The results are that: 1.) analyses of BCR-1 at the calculated Pathfinder nominal distance yields chemical abundances that are the same within the average laboratory errors as conventional values; 2.) sample oxygen content and the inferred sample water content increase with distance when the uncorrected Pathfinder library values are used; 3.) a library change in the oxygen intensities of standards alone by the predicted amounts would yield progressively lower oxygen and higher silicon with increasing distance; and 4.) changes in the library values for both oxygen and silicon by the predicted amounts yield chemical results that are equivalent to conventional values within the average laboratory measurement errors listed in Table 1.

Figure 20.

BCR-1 AXS α-spectra for various distances, showing oxygen and silicon inelastic scattering intensity changes with distance. Spectra are normalized to counts/100,000 sec at channel ∼160 (a spectral region created only from Rutherford scattering). The spectral changes at the edges of the oxygen and silicon signals are due to the narrower distribution of α-scattering angles with distance.

Table 5. Chemical Abundances for AXS Laboratory Measurements of BCR-1 at Various Distances From Nominal Computed Using the OPL, the OPL Modified for the Change in Oxygen, and the OPL Modified for the Change in Both Oxygen and Silicona
BCR-1 Sample AnalysesOPLOPL With Oxygen ChangeOPL With Oxygen and Silicon Change
Element, at%Conventional+10 mm+15 mm+20 mm+10 mm+15 mm+20 mm+10 mm+15 mm+20 mm
  • a

    Conventional values are from Govindaraju [1994]. All values are in atom% except for the last row of wt% water values. OPL stands for original Pathfinder library. Elements denoted with the symbol * are determined from X-ray ratios.

Water (wt%)0.751.−1.9−

4.2.3. Application of AXS Distance Experiments to Pathfinder APXS

[71] As illustrated in Table 5, the changes in oxygen and silicon α-spectra with distance are correctly compensated using the AXS laboratory experiments and calculations. Furthermore, as shown in Figure 22, after corrections for the oxygen and silicon, the correct amount of sample water is calculated for BCR-1. Figure 21 also shows that the inferred water content of Pathfinder samples, having abundances of silicon and oxygen similar to BCR-1, should have at most a 3 wt% positive bias, if uncorrected for distance variation during measurements. This is evident because the greatest measurement distance for a Pathfinder sample, as discussed below, is equivalent to an AXS measurement distance of 18 mm.

Figure 21.

The inferred water content in BCR-1 before and after distance corrections. Analyses at distances greater than nominal which are uncorrected for the changes in the oxygen and silicon library values with distance predict too much sample water. After corrections for silicon and oxygen intensity changes with distance, sample water abundance is the same as conventional within measurement error. All distances used were calculated from the 1/r2 decrease in α-intensity with distance, having errors of approximately ±9 relative percent. The 1σ error for the wt% water was calculated by averaging the combined statistical and laboratory error for each sample.

[72] The distances for Pathfinder measurements were determined using the 1/r2 dependence of the iron signal in the α-spectra, as discussed by Foley [2002] in more detail. The estimates of measurement distance and the effects on the oxygen and silicon concentrations for each Pathfinder sample are listed in Table 6. The maximum increase for the oxygen concentration is 12 relative percent, while that for silicon is 7 relative percent. As shown previously using this technique to analyze BCR-1 at various distances, the accuracy of this technique for analyzing the Pathfinder samples is the same as measurements made at the calibration distance within the average laboratory errors listed in Table 1. The error for the inferred water content after correcting for distance, σ(H2O) is then

display math

where σ(H2Oo) is the initial error in the inferred water content based on counting statistics and laboratory accuracy, and σ(H2Odave) is the average error in the inferred water content associated with the distance correction. The average σ(H2Od) value for the Pathfinder samples was taken to be 0.2 wt%, which is the difference between the inferred water content at the initially calculated distance and that calculated 2 mm further (because ±2 mm is the approximate distance error).

Table 6. Approximations for Measurement Distance and the Effects on the Oxygen and Silicon Library Values for Each Pathfinder Samplea
Pathfinder SamplePathfinder Distance, mmEquivalent AXS Distance, mmOxygen Relative % IncreaseSilicon Relative % Increase
  • a

    Pathfinder and AXS distances are from the center of the cylinder bases of the respective instruments to the center of the sample. The AXS distance is the equivalent distance at which the AXS has the same oxygen signal as the Pathfinder instrument.

A-2, Soil by lander14.417.911.97.0
A-3, Barnacle Bill rock9.415.18.65.0
A-4, Soil by Yogi11.916.510.36.0
A-5, Dark soil by Yogi6.413.46.23.0
A-7, Yogi rock10.
A-8, Scooby Doo rock/soil7.414.07.14.0
A-10, Soil by Lamb4.312.24.42.5
A-15, Mermaid Dune soil4.612.44.72.7
A-16, Wedge rock6.913.76.74.0
A-17, Shark rock14.518.011.97.0
A-18, Half Dome rock5.813.15.83.0

[73] The distances computed with the iron intensity from the α-mode generally correlate with the atmospheric carbon peak areas for the Pathfinder spectra collected at similar atmospheric number densities, as shown in Figure 22. The carbon atmospheric peak area increases with atmospheric number density and with distance. Hence the general variation of atmospheric carbon peak area with distance, as shown in Figure 22, supports the general accuracy of our distance calculations.

Figure 22.

The atmospheric carbon peak area (counts/sec) versus the calculated measurement distance based on the iron intensity of the α-mode. All carbon peak areas are for approximately the same measurement temperature of −70°C or were computed for the measured sample near −70°C. Atmospheric carbon peak area error bars are 1σ statistical uncertainty alone. Distance error is 9 relative percent, on the basis of laboratory accuracy of the distance calculation.

[74] Three sensors, located on the bumper of the Pathfinder APXS cylinder, were intended to make contact with the analyzed Martian samples. This bumper is 72 mm in diameter and was approximately flush with the outside edge of the Pathfinder cylinder [Blomquist, 1995]. The APXS cylinder was designed to be tilted a maximum of 20° relative to the sample surface during sample measurement [Blomquist, 1995]. All samples analyzed made contact with one or two sensors, except for soils A-2 and A-9 for which no sensors made contact. This means that all samples making contact were either not perpendicular to the APXS sensor head and/or have an irregular surface. The Pathfinder rocks were observed to have pitted textures with ventifacts that are centimeters in size [Bridges et al., 1999]. These ventifacts may then have centimeter-scale depths as well. It is therefore probable that some of the deviation from the nominal measurement distance for the rocks is due to the pitted texture of the rocks. Furthermore, all measured soils were mixtures of pebbles and/or clods with fine soil [McSween et al., 1999; Bell et al., 2000] and therefore effectively have a centimeter-scale texture. Whether the distance variation is due primarily to the sample texture or to the tilt of the sample relative to the plane of the Pathfinder cylinder, the average distance recorded by the iron signal of the α-mode still yields the applicable measurement distance. Thus the distances listed in Table 6 were used to determine the Pathfinder α-mode abundances.

5. Conclusions

[75] The laboratory calibration of the APXS, under simulated Martian conditions has been completed. Corrections for differences between the laboratory and Martian conditions have also been completed. These differences include the absence of a protective film and supportive grid on Mars, the variable measurement distance on Mars, and the atmospheric contributions on Mars. The grid corrections enable computation of more accurate Pathfinder X-ray abundances. The distance and atmospheric corrections enable computation of accurate α-proton Pathfinder abundances. This calibration routine and corrections may now be utilized to analyze the Mars Pathfinder rock and soil data with all three instrument modes, as discussed by Foley et al. [2003]. Results attained with this new calibration are somewhat different from the initial X-ray results reported by Rieder et al. [1997a] which did not correct for the presence of the grid. Our new Pathfinder X-ray results [Foley et al., 2003] are more similar to those of Wänke et al. [2001] which also corrected for the presence of the grid during laboratory measurements. Yet, results attained using this calibration [Foley et al., 2003] are more comprehensive than previously reported abundances because the alpha-proton mode Pathfinder results, for both carbon and oxygen in particular, are also determined.


[76] Many University of Chicago scientists have greatly assisted in the completion of this project: Frank and Pasquale DiDonna, who assisted in the development and maintenance of the APXS; Ellen LaRue, who ran some of the laboratory samples and helped to maintain the proper chamber conditions for the laboratory APXS; Andrew Davis, Roy Lewis, Ian Steele, and Alfred Anderson, who provided insightful discussions. Many thanks, also, to Joy Crisp from JPL who sent samples for analyses and gave descriptions of the measurement duration and deployment position of the Pathfinder APXS. Ted Foley significantly assisted with the development of a program to compute the distance effects on sample composition for the APXS. We are thankful for the insightful reviews of Hap McSween and Dick Morris. This work was supported by the Department of the Geophysical Sciences of the University of Chicago and by NASA grants *NAG5-11043(TEE) and *NAG5-3986(RNC).