Disk-resolved observations of Phobos acquired by OMEGA at a range of lighting and viewing geometries were fit with the Hapke photometric function to solve for the single particle phase function and single scattering albedos from 0.4 to 2.5 μm. Single scattering albedos were recovered from CRISM observations of Phobos using the OMEGA derived single particle phase function and are similar to those from OMEGA data. Both the ubiquitous red unit and the blue unit around the crater Stickney exhibit a smooth red-sloped spectrum, with a steeper continuum in the redder unit. Single scattering albedos retrieved from CRISM measurements of Deimos are similar to those for the red unit on Phobos. Retrieval of single scattering albedos from OMEGA data at 2.8 to 5.0 μm has greater uncertainty, but results in this wavelength range are also consistent with a smooth, red-sloped spectrum. Phobos' and Deimos' low reflectances, lack of mafic absorption features, and red spectral slopes are incompatible with even highly space weathered chondritic or basaltic compositions. These results, coupled with similarities to laboratory spectra of Tagish Lake (possible D-type asteroid analog) and CM carbonaceous chondrite meteorites, show that Phobos and Deimos have primitive compositions. If the moons formed in situ rather than by capture of primitive bodies, primitive materials must have been added to the Martian system during accretion or a late stage impact.
 Mars' moons Phobos and Deimos are small, irregularly shaped, low-density bodies with low albedos and spectra similar to D-type asteroids common in the outer main asteroid belt and outer solar system [e.g.,Burns, 1992; Murchie and Erard, 1996; Rivkin et al., 2002, and references therein]. The compositions of Phobos and Deimos are poorly understood, although proposed formation mechanisms for the moons can help constrain their probable make up. There are two main classes of models to explain the origin of the Martian moons: capture or formation in Mars orbit. Capture of dark, low-density objects from the outer solar system [Burns, 1978; Hartmann, 1990] would explain Phobos' and Deimos' low densities and albedos, but this scenario is considered unlikely given that dynamical models for capture require specific conditions, including aerodynamic drag by an early Mars proto-atmosphere [Hunten, 1979; Sasaki, 1990]. If they are captured bodies, the moons' orbits are also difficult to explain dynamically because both are circular; Phobos spirals toward Mars, while Deimos spirals out. In his review of orbital constraints on the moons' origins, Burns concluded that formation in situ was favored dynamically. In situ formation hypotheses include formation by co-accretion with Mars [Safronov et al., 1986] and formation resulting from a giant impact on Mars [Craddock, 2011].
 Together, these origin hypotheses result in Phobos and Deimos having one of three likely compositions: (1) primitive, undifferentiated compositions dominated by phyllosilicates with C-containing phases, similar to CM-type carbonaceous chondrites [Brearley and Jones, 1998] or to the more primitive Tagish Lake meteorite, commonly thought to be a compositional analog to D-type asteroids [Brown et al., 2000; Hiroi et al., 2001]; (2) compositions resembling pre-differentiated, bulk Mars, which is expected to be mainly composed of ordinary chondrites with a minor carbonaceous chondrite component and dominated by olivine and pyroxene minerals [Lodders and Fegley, 1998]; or (3) basaltic compositions similar to the Martian crust and upper mantle, also dominated by olivine and pyroxene minerals.
 In this paper, we evaluate the compositions of Phobos and Deimos using disk-resolved imaging spectrometer measurements of Phobos covering 0.4 to 5.0 μm acquired by the Observatoire pour la Mineralogie, l'Eau, les Glaces et l'Activité (OMEGA) onboard Mars Express (MEx) [Bibring et al., 2004], and measurements of both moons covering 0.4 to 2.7 μm acquired by the Compact Reconnaissance Imaging Spectrometer for Mars (CRISM) on Mars Reconnaissance Orbiter (MRO) [Murchie et al., 2007]. Although CRISM data extend to 3.9 μm, we restrict our CRISM analyses to shorter wavelengths pending resolution of radiometric calibration artifacts at longer wavelengths [Murchie et al., 2009]. OMEGA observations of Phobos acquired at a range of lighting and viewing conditions for Phobos are used to retrieve single scattering albedos and single particle phase functions using the Hapke photometric function [Hapke, 1993]. These functions are then used to simulate single scattering albedo values for CRISM observations to directly compare the two data sets and to determine if Deimos has similar spectral properties as Phobos. Finally, we compare the moons with telescopic and laboratory spectra in order to assess proposed compositions and discuss implications.
2. Overview of Data Sets
 OMEGA is a suite of three spectrometers: a visible (V) channel (0.35 μm to 1 μm), a short wavelength infrared (SWIR-C) channel (1 μm to 2.5 μm), and a medium wavelength infrared L (SWIR-L) channel (∼2.8 μm to 5 μm) [Bibring et al., 2004]. OMEGA acquired ten disk-resolved observations of Phobos over a period from May 2004 through January 2011 (Table 1), six of which contain high-quality data in both the V and C channels. The remaining four observations are either V-only or were taken late in the instrument's lifetime and contain a number of degraded spectral channels within the C channel. The OMEGA measurements of Phobos have spatial sampling from 120 m/pixel to 2200 m/pixel and provide broad coverage of both the sub- and anti-Mars hemispheres of Phobos (Figure 1). In addition, High Resolution Stereo Camera (HRSC) images acquired in conjunction with the OMEGA observations provide morphologic context for the OMEGA observations (Table 1, Figure 1).
Table 1. Disk Resolved OMEGA and CRISM Observations of Phobos and Deimosa
Avg. Phase Angle
HRSC/HiRISE Context Image
All OMEGA and CRISM Phobos and Deimos disk-resolved observations. Observations in italics were not used in this study due to insufficient wavelength coverage, lower quality data, or low spatial resolution. The observations cover a range of phase angles and have varying spatial resolutions. Incidencei, emergence e, and phase angle α for the OMEGA data were computed using a prototype version of the Navigation and Ancillary Information Facility (NAIF) SPICE Toolkit that includes a Digital Shape Kernel (DSK) system in conjunction with a Phobos shape model. The photometric angles i, e, and α for the CRISM data set were calculated using P. Thomas' shape model and the SciBox toolkit [Thomas, 1993; Choo et al., 2012].
Acquired at a separate time than CRISM observation but still used for context imaging.
 CRISM is a suite of two high spatial resolution spectrometers, a visible/near-infrared (VNIR) channel from 0.4 to 1.0 μm and an infrared (IR) channel from 1.0 to 3.9 μm, both have spectral resolutions of ∼6.55 nm/channel [Murchie et al., 2007]. The CRISM Phobos data set consists of three observations acquired in succession on 23 October 2007 at an average phase angle of 41°, spatial sampling of 350 m/pixel, and covering approximately the same region in the western side of Phobos' sub-Mars hemisphere (Table 1, Figure 2) [Murchie et al., 2008]. HiRISE observations provide high-resolution spatial context for the CRISM data, although the HiRISE data were not acquired in coordination with the CRISM data (Table 1, Figure 2) [Thomas et al., 2011].
 Previous work shows that there are two color or spectral units exposed on Phobos, a red and a blue unit [e.g., Murchie and Erard, 1996; Rivkin et al., 2002]. The red unit covers most of the surface of the moon and is characterized by low overall albedo and a red spectral slope. The blue unit is exposed on and near the 9 km diameter crater Stickney and has spectral properties similar to the ubiquitous red unit, but with a less red spectral slope. Regions of interest (ROIs) within these two units and in Stickney crater are shown in Figures 3 and 4.
 To utilize both the CRISM and OMEGA data sets over as much of their spectral range as possible, we need to explicitly model the spectrophotometry and both the reflected solar and emitted thermal radiation. For example, Figure 3 shows averaged radiance spectra from OMEGA data between 0.4 to 5.0 μm covering portions of the same region on the sub-Mars hemisphere of Phobos. For illustrative purposes, scattered solar irradiance is plotted assuming a Lambert Albedo of 0.05, and thermal emission is shown for a selection of surface kinetic temperatures within the range predicted using models byKuzmin and Zabalueva  for the times and locations of the OMEGA observation. These relatively simple approximations to the spectral radiances for Phobos demonstrate that the transition between solar scattered light and emission dominated spectra occurs longward of ∼2.7 μm. At these longer wavelengths spectrophotometric models must account for the presence of both scattered solar irradiance and thermal emission. It is thus appropriate to begin the analyses using the OMEGA V and C data (0.4 to 2.5 μm) and to then separately turn attention to the L data (2.8 to 5.0 μm).
Figure 4 shows I/F spectra (radiance/(solar irradiance/π)) for OMEGA and CRISM data of approximately the same regions on Phobos. The spectra from the CRISM observation are brighter than spectra from the OMEGA observations because the CRISM data were acquired at a lower phase angle (41°) than the OMEGA observation (63°), and the surface of Phobos is known to be backscattering and thus darkens with increasing phase angle [Simonelli et al., 1998]. Both sets of spectra brighten nearly monotonically with increasing wavelength and exhibit a large thermal emission at longer wavelengths, consistent with radiance spectra shown in Figure 3. Marsshine is present in the sub-Mars facing data for OMEGA data (CRISM observations were taken on the night side of Mars); based on analyses byGoguen et al. , Marsshine in the OMEGA observations is only at most few percent of the radiance and to first order can be ignored.
 CRISM collected for the first time spatially resolved imaging spectrometer measurements of Deimos in June 2007 [Murchie et al., 2008]. Three successive data sets were acquired covering the sub-Martian hemisphere of Deimos at an average phase angle of 22° with a spatial sampling ∼1.2 km/pixel (Figure 5). Because these observations were taken when Deimos was over the night side of Mars, Marsshine has a negligible contribution to the measured radiance. Figure 6 shows the average I/F spectrum from the middle CRISM Deimos observation. There is very little spectral variation from pixel to pixel within this observation except for spatially varying thermal emission. The I/F increases nearly monotonically with wavelength and is similar in magnitude to the I/F spectra of Phobos. As noted previously, here we restrict retrieval of spectrophotometric properties for CRISM data to wavelengths <2.7 μm; inclusion of longer wavelength data in Figure 6 is illustrative and not meant to be quantitative.
3. Retrieval of Spectrophotometric Properties
Overview: The broad phase angle coverage of the six OMEGA Phobos observations (Table 1) allows us to derive spectrophotometric parameters for the V and C channels. OMEGA I/F, equivalent to radiance factor r, may be modeled using Hapke's equation for bidirectional radiance and directional emissivity expressed as a function of incidence angle i, emission angle e, phase angle α, and kinetic temperature T, in the following manner for a given wavelength:
where w is the average single scattering albedo, μ is the cosine of the emergence angle, μe and μ0e are the effective cosines of the emergence and incidence angles, respectively, B(α) models the opposition-effect,p(g1) is the single particle phase function, H is an approximation for the isotropic multiple scattering function, and S is the shadowing function defined by a macroscopic roughness parameter, θ [Hapke, 1993]. In addition, β0 is the Planck irradiance at a given temperature T and J is the solar irradiance at the solar distance at the time of observation [Hapke, 2005].
 The one-term Henyey-Greenstein single particle phase function is dependent only on phase angle and one asymmetry parameter,g1, as
In this form, negative values of g1are associated with backscattering material and positive values indicate forward scattering material. We choose to fit the data using the one-term Henyey-Greenstein function rather than a more complex phase function in order to keep the number of parameters to be fit at the minimum possible. The lack of low phase angle observations did not allow us to constrain parameters associated with the opposition surge,B0 and h, so we adopted the best fit values derived from Viking clear-filter data of 5.7 and 0.072, respectively [Simonelli et al., 1998]. This selection has a minimal effect on the derived best fit parameters as all parameters are derived using phase angles greater than 35 degrees. The opposition surge is most influential for observations at very small phase angles, less than ∼15 degrees.
Solar-Dominated Regime: Our approach was first to solve equation (1) over the spectral region dominated by scattered solar radiation, i.e., using data from the V and C channels. For the six OMEGA observations we searched for the values of w, g1, and θ that would provide the best global average fit for the surface of Phobos at selected wavelengths. A total of 38 spectra were selected from each observation (the maximum number allowed given the observation with the fewest available spectra), and the IDL subroutine MPFIT was used to find the combination of w, g1, and θ values that best fit all of data at each selected wavelength. The incidence and emergence angles for each spectrum used in the fit are presented in Figure 7, phase angles for the six observations are listed in Table 1. Note that these fits are for all OMEGA data and thus should be most relevant to the aerially dominant red unit surfaces.
 MPFIT is a nonlinear, least squares fitting function based on the Levenberg-Marquardt algorithm [Moré, 1978; Markwardt, 2009]. This algorithm searches through a space of chi-square values beginning with a set of initial parameter guesses and finds the parameter combination associated with a local minimum in this chi-square parameter space by following the steepest gradient in the space. Because different values of initial conditions may converge on local minima chi-square values rather than the absolute minimum, we ran the models multiple times using systematically varying initial values to ensure the resultant best fit parameters truly represented the absolute minimum chi-square value rather than a local minimum. The absolute best fit parameters are those with the smallest chi-square values.
 Since the roughness parameter, θ, should be independent of wavelength, this value was determined by first letting it vary freely at each selected wavelength. We found an average best fit θ value of 14° over all wavelengths and subsequently held θfixed at this value while each wavelength was re-run with onlyw and g1 free to vary. We did not attempt to constrain temperature as the contribution of thermal emission is negligible in the solar dominated wavelength regime, i.e., β0(T,λ) ∼ 0 for all reasonable surface temperatures. One-sigma uncertainty values were assigned to the best fitw and g1 values using a routine described by [Cord et al., 2003] and adopted by [Johnson et al., 2006]. In this method, the parameter of interest is held fixed in a series of values moving away from the best fit value while other parameters are allowed to vary freely. The one-sigma confidence level occurs when the resulting chi-square of the best fit model has doubled from its original value. In our case, the opposition surge parametersB0 and h and roughness parameter θ were assumed to be fixed, so the only parameters allowed to vary were either w or g1. For a well-constrained parameter the chi-square value increases quickly as that parameter is forced to move away from its best fit value. Conversely, changing the values of a poorly constrained parameter will result in only a gradual increase of chi-square values. For a more detailed description and validation of this method see [Cord et al., 2003] and [Johnson et al., 2006].
Figure 8shows the wavelength-dependent, global average best fit values for single scattering albedo for Phobos at selected wavelengths. Single scattering albedos increase monotonically with wavelength, beginning with values of ∼0.05 at 0.45 μm and increasing to ∼0.15 at 2.5 μm. The results at shorter wavelengths are consistent with previous photometric studies based on clear-filter Viking data centered at 0.54 μm and Hubble Space Telescope observations from 0.41 to 1.042 μm [Simonelli et al., 1998; Cantor et al., 1999]. The best fit asymmetry parameter was found to be wavelength independent, with a value of −0.28 ± 0.3, indicative of a backscattering surface. The phase function is most sensitive to changing g1 values for phase angles ≤50° and ≥100° [Helfenstein and Veverka, 1989] whereas four of the six OMEGA observations analyzed were made at phase angles between 50° to 100° (Table 1). Thus the large uncertainty in estimating g1is not unexpected. We note that regolith-covered planetary surfaces have been modeled using values ofg1 between −0.20 and −0.40 [e.g., Clark et al., 2002a], providing assurance that our results are reasonable.
 The ability of the Hapke Function to model the OMEGA and CRISM data for the Phobos red unit is shown in Figure 9. Here spectra of similar regions in both data sets have been recast to single scattering albedo by solving equation 1 using measured I/F values, known viewing geometries, and the OMEGA derived best fit phase function and roughness parameter. The similarity in the single scattering albedo spectra indicates that the two instruments provide comparable measurements of sunlight scattered from the surface and that the photometric model represents how light is scattered as a function of lighting and viewing geometries to within stated formal uncertainties. Given that we have derived a global spectrophotometric model of Phobos from OMEGA data that also works for CRISM data and that previous photometry studies have suggested Phobos and Deimos have similar photometric behavior [Thomas et al., 1996; Simonelli et al., 1998], we can therefore also recast the CRISM Deimos data into single scattering albedo values. Results indicate that Deimos is indeed similar to the red unit that dominates the surface of Phobos.
Mixed Solar and Thermal Regimes:The addition of a thermal emission component introduces another unknown (i.e., surface kinetic temperature) into the Hapke Function and makes it impossible to continue to solve for global-average Hapke parameters in OMEGA data at >2.6 μm in the L channel because of pixel-to-pixel temperature variations. These difficulties are evident in inspection of the radiance spectra inFigure 3in which emission from Phobos produces spectra that are shallower than spectrally neutral gray body emitters. Either the single scattering albedos increase in magnitude in the wavelength range of the L channel or the OMEGA pixels sample surfaces with spatially non-uniform temperatures. In order to work with OMEGA data in this region, we assume that the global-average wavelength-independent Hapke parameters found at shorter wavelengths remain constant up to 5 μm. This simplification reduces the unknown quantities to single scattering albedo and surface temperature, assuming that each pixel's radiance can be modeled as resulting from a single surface kinetic temperature.
 We begin our retrieval of single scattering albedos with estimates of the surface temperatures expected in the OMEGA observations using a thermal model of the surface of Phobos [Kuzmin and Zabalueva, 2003]. This model takes into account the ellipsoidal shape of Phobos, eclipses of Phobos by Mars, reflected thermal radiation from Mars, and the absence of internal heat sources. The model uses an average Phobos albedo of 0.07 (consistent with our average retrieval in the solar-dominated region of the spectrum), a surface regolith density of 1100 kg m−3, and temperature-dependent thermal conductivity and specific heat equations that are analogous to lunar regolith. Emission is assumed to be blackbody and directionally uniform. The approach we used was to solve for spectral single scattering albedo and examine the retrieved spectra with the realization that albedos need to be within the bounds between 0 and 1. Single scattering albedo retrievals were found to be particularly noisy around the thermal cross over where solar reflected radiance and thermal emission become equal. The exact wavelength depends on the surface kinetic temperature and lighting and viewing geometries, with a typical value of ∼4.5 μm. Retrievals longward of the cross over wavelength were very noisy indeed (Figure 10). The inability to retrieve precision spectra in the mixed solar and thermal regime is not surprising given the uncertainties in the model used and the complex shape of Phobos as compared to the shape model used to compute the thermal emission. Our best estimate is that the single scattering albedos continue to increase in magnitude (i.e., redden with a increasing wavelength) to at least ∼4.5 μm, but this result is notional and we do not attempt quantitative analyses of the spectral range from 2.6 to 4.5 μm.
4. Analysis of Phobos and Deimos Spectra
 Single scattering albedo data for Phobos and Deimos are shown in Figure 11 for the blue and red units and for the average of the CRISM Deimos observation. OMEGA data for this particular observation, ORB0756_0, show a peculiar concave downward V spectrum that becomes more pronounced toward the upper left of the data frame. We suspect this is a calibration artifact. Otherwise the results for the two instruments are similar indeed and consistent with the limited earlier observations [e.g., Murchie and Erard, 1996; Rivkin et al., 2002] in that the two moons have very low single scattering albedos and red slopes.
4.1. Search for Spectral Absorption Features
 Results from the Phobos-2 Imaging Spectrometer for Mars (ISM) were interpreted as indicating the presence of olivine and pyroxene on Phobos [Murchie and Erard, 1996; Gendrin et al., 2005], and these minerals are expected to be present if Phobos and Deimos have chronditic or basaltic compositions. We have therefore specifically searched for these features in the CRISM and OMEGA single scattering albedo spectra by generating the standard mafic mineral summary parameter maps that are commonly used in analysis of OMEGA and CRISM observations of Mars [Mustard et al., 2005; Loizeau et al., 2007; Pelkey et al., 2007]. These summary parameters are sensitive to the broad absorptions around 1 and 2 μm caused by iron related crystal field transitions in both minerals. The resulting parameter maps designed to highlight the presence of olivine and low- and high-calcium pyroxene show little or no indication for any of these minerals, and we conclude that there is no evidence for mafic mineral absorption features above a few percent, the level of instrument noise.
 In addition, examination of the retrieved L channel single scattering albedos does not indicate discernible water or metal-OH vibrational features at 2.8–3.1 μm stronger than a few percent, although this conclusion is tempered by the inability to model the single scattering albedos with high fidelity and the fact that the OMEGA C and L channel boundary occurs between 2.5 and 2.7 μm. We have also searched in Phobos and Deimos spectra for evidence of water-related combination bands at 1.92 μm and associated wavelengths, for water or OH combination bands at 1.4 μm, and have not found any evidence for either of these features.
 Initial analyses of the CRISM Phobos and Deimos I/F observations revealed the presence of a broad, shallow absorption feature centered near 0.65 μm in the redder unit of Phobos and on Deimos [Murchie et al., 2008]. We leave a thorough discussion of this CRISM-based detection to future work in which we expand on the initial discovery and place limits on the possible causes of this broad feature.
4.2. Comparisons to Laboratory Spectra
 Given our knowledge of Phobos and Deimos spectrophotometry, we can model the spectra to simulate the lighting and viewing conditions as they would appear for laboratory spectra. The bidirectional reflectance measurements provided in the RELAB database may be expressed using the Hapke model for radiance coefficient as [Hapke, 1993]
The ability to model OMEGA and CRISM data as laboratory spectra allows us for the first time to directly compare the slope and absolute magnitudes of the Phobos and Deimos spectra to lab data without the need for scaling, and this is an important diagnostic tool given the lack of obvious mineralogical absorption features in the moons' spectra.
Figure 12 (top left) shows CRISM blue unit spectra from Stickney's interior and ejecta modeled as bidirectional radiance coefficients at i = 30° along with three meteorite analogs that were also observed at i = 30°. We found analogs by searching through libraries of dark, relatively featureless spectra, including highly space weathered lunar material, CI chondrites, and CM chondrites. By visual inspection, the best match meteorite analogs have primitive, CM carbonaceous chondrite-like compositions (Murchison and Mighei) and an ultraprimitive, D-asteroid analog composition (Tagish Lake). All three are close matches to the slope and absolute magnitude of the Phobos blue unit spectrum. The closest meteorite analog is 0–63 μm particles from Murchison, a typical CM chondrite, that have been heated to 700°C for one week [Hiroi et al., 1993]. Before heating, Murchison was dominantly composed of Mg-serpentine and Fe-bearing serpentine-group mineral cronstedtite and sulfide mineral tochilinite, with smaller amounts of olivine, clinoenstatite, and pyrrhotite [Bland et al., 2004]. As a result of heating beyond 600°C, the serpentines began to dehydrate and recrystallize as olivine while iron-rich orthopyroxenes and tochilinite had been converted to troilite [Hiroi et al., 1993]. The featureless heated Murchison spectrum has a similar slope to spectra from Stickney's interior and ejecta, and in particular, has a similar absolute magnitude with the Stickney ejecta spectrum.
 Another closely matched spectral candidate to the blue unit is for 0–45 μm bulk particles from the CM chondrite Mighei. This meteorite has a similar composition to the unheated Murchison but a higher Fe/(Fe + Mg) ratio [Cloutis et al., 2011]. The Mighei spectrum has a similar brightness and slope to the Phobos blue unit, although it does show absorption features near 0.4 μm and 0.7 μm that are not observed in the Phobos blue spectra. The steep drop off near 0.4 μm is found in many unheated CM chondrite spectra and is attributed to phyllosilicates, magnetite, or organics [Cloutis et al., 2011]. The fact that this feature is not observed in the CRISM spectra may be due to unreliable data shorter than 0.44 μm. The shallow, broad absorption centered near 0.7 μm is attributed to mixed valence Fe-bearing serpentine group phyllosilicates [Cloutis et al., 2011].
 The Tagish Lake meteorite has been interpreted as a sample from the ultra-primitive D-type asteroids from the outer solar system [Hiroi et al., 2001]. This meteorite is a carbon-rich, aqueously altered unusual carbonaceous chondrite with high concentrations of presolar grains and carbonate minerals [Hiroi et al., 2001]. The spectrum of Tagish Lake has an absolute reflectance consistent with the spectrum from Stickney's interior and no visible absorption features, although its slope deviates at the longer wavelengths and is not as good a fit as the CM chondrite spectra.
 The closest match analog spectra to the Phobos red unit and Deimos in terms of slope and absolute magnitude are also CM chondrites (Figure 12, top right). Unlike the Phobos blue unit, no laboratory analogs were found that perfectly match the slope and absolute magnitude of red unit spectra, although a few are quite close. Particles <40 μm from a bulk Mighei sample were a good match, as well as particles <125 μm from the CM Chondrite Cold Bokkeveld. Like Mighei, Cold Bokkeveld is large CM chondrite with little terrestrial alteration. The two samples have similar modal volume of the abundant mineral phases Mg-serpentine and Fe-cronstedtite [Cloutis et al., 2011], and these phases are evident in the broad, shallow features near 0.4 μm and 0.7 μm. These features are not observed in the CRISM and OMEGA spectra.
4.3. Effects of Space Weathering
 In the case of airless bodies like Phobos and Deimos, any comparison to laboratory analogs based on the magnitude and slope of spectra is confounded by the fact that we are unable to fully quantify the role space weathering has played in modifying Phobos and Deimos. Space weathering is defined by Clark et al. [2002b] as processes that modify the traits (including spectral properties) of an airless body's surface from analogous traits of the body's inherent bulk materials. The processes we expect to cause space weathering on Phobos and Deimos include micrometeorite impact and solar wind ion implantation and sputtering. Analysis of lunar samples has shown that these processes form reduced, submicroscopic iron particles (SMFe) infused throughout mineral grains and glasses [e.g., Pieters et al., 2000; Hapke, 2001; Noble et al., 2001; Taylor et al., 2001; Noble et al., 2007; Lucey and Noble, 2008] The addition of SMFe particles influences the VNIR spectral properties of a surface by reducing the strengths of absorption features and modifying spectral slope and magnitude. SMFe particles smaller than ∼50 nm both darken and redden the surface while particles larger than ∼50 nm just darken the surface [Noble et al., 2007; Lucey and Noble, 2008; Lucey and Riner, 2011].
 While space weathering can darken a surface, it is difficult to explain the dark nature of Phobos and Deimos through space weathering alone if the inherent compositions of the moons are similar to ordinary chondrites (similar to bulk Mars) or basaltic or ultramafic material (similar to Martian crust and upper mantle). The most mature, darkest lunar mare spectra collected by the Moon Mineralogy Mapper (M3) [Green et al., 2011] are brighter than Phobos and Deimos when corrected to the same viewing geometries (i = α = 30°, e = 0°) (Figure 12, bottom left). Even in the intense space weathering environment at Mercury, average I/F spectra collected from the Mercury Atmospheric and Surface Composition Spectrometer (MASCS) [Izenberg et al., 2012] are substantially brighter that Phobos and Deimos I/F spectra modeled at the same viewing geometry (i = e = 45°, α = 90°) (Figure 12, bottom right). Although the space-weathering environment around Phobos and Deimos is poorly understood, it is difficult to argue that Phobos and Deimos would have a more intense space weathering environment than Mercury. The moons of Mars are intrinsically dark and are better fit by spectra from primitive meteorite analogs than by space weathered basaltic soils (Figure 12).
5. Summary and Conclusions
 Disk-resolved observations of Phobos from six OMEGA observations acquired at a range of lighting and viewing geometries were analyzed with the Hapke photometric function to solve for the single particle phase function and single scattering albedos from 0.4 to 2.5 μm, the spectral region covered by OMEGA V and C channel data. Single scattering albedos recovered from CRISM observations of Phobos using the OMEGA single particle phase function are similar to those retrieved from OMEGA data for the same areas, both for the ubiquitous red unit and for the blue unit exposed in and around the crater Stickney. In addition, CRISM-based single scattering albedos retrieved for Deimos are similar to those for the dominant red unit on Phobos. Retrieval of single scattering albedos for OMEGA L data (2.8 to 5.0 μm) is problematic at this time due to the presence of both reflected solar and thermal emission signatures together with the complex shape of Phobos, although we infer that the single scattering albedo spectra continue to have low values and red slopes.
 The low single scattering albedos and red slopes are consistent with previous studies that indicate primitive compositions for the two moons. Mafic mineral absorptions are absent, as are OH or H2O features. Recasting the spectra for Phobos and Deimos to laboratory lighting and viewing conditions shows that the spectra are very similar to laboratory spectra of Tagish Lake (D-type asteroid meteorite analog) and CM carbonaceous chrondrite meteorites. The spectral properties of Phobos and Deimos are inconsistent with (1) a chondritic composition similar to bulk Mars or (2) a basaltic composition similar to differentiated Martian crust and upper mantle. Even extensive space weathering does not appear to be sufficient to darken and redden chondritic or basaltic composition materials to produce the spectral characteristics of the two moons.
 Phobos and Deimos could be primitive objects from the outer solar system captured into Martian orbit, although dynamical models for capture require special and improbable conditions [Hunten, 1979; Sasaki, 1990]. Conversely, formation of bodies with primitive materials is not currently accounted for in in situ formation models [Safronov et al., 1986; Craddock, 2011]. Thus, future dynamical models of the formation of the moons should specifically account for the primitive character of these bodies. In addition to improved origin models, the next logical step in definitively understanding the origin, evolution, and composition of Phobos and Deimos is in situ analysis of composition or sample return from these satellites.
 We thank NASA/JHUAPL for support on this work and the engineers and scientists associated with the CRISM operations center. Further, we thank the team of OMEGA scientists and engineers for the acquisition, processing, and availability of the OMEGA data. Bruce Hapke and Beth Clark provided constructive comments that greatly enhanced the clarity of the manuscript. We thank Nancy Chabot for reviewing an early draft of this paper leading to an improved version of this manuscript. This research utilizes spectra available from the NASA RELAB facility at Brown University and spectra provided by Noah Izenberg, Rachel Klima, and Debra Lorin. This work was supported by CRISM APL/JPL contract 104149 and AAF was funded by an NSF Graduate Student Research Fellowship, Grant DGE-1143954.