Using the modified Gaussian model to extract quantitative data from lunar soils

Transcription

1 Click Here for Full Article JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 111,, doi: /2006je002721, 2006 Using the modified Gaussian model to extract quantitative data from lunar soils Sarah K. Noble, 1,2 Carlé M. Pieters, 1 Takahiro Hiroi, 1 and Lawrence A. Taylor 3 Received 23 March 2005; revised 23 May 2006; accepted 7 June 2006; published 22 November [1] The Lunar Soil Characterization Consortium (LSCC) has examined and characterized a suite of lunar soils with a wide range of compositions and maturities. The purpose of this study is to compare the Vis/NIR spectral properties of these lunar soils with their petrologic and chemical compositions using the modified Gaussian model (MGM) to obtain quantitative data about the character of relatively weak near-infrared absorption bands. Useful compositional information can be extracted from high-quality soil spectra using the MGM. The model had some difficulty fitting absorption bands in the 2 mm region of the lunar spectrum, but bands in the 1 and 1.2 mm regions provided physically realistic results. The model was able to distinguish high-ca and low-ca pyroxenes in the LSCC suite of lunar soils in the appropriate relative abundance. In addition, unexpected insights into the nature and causes of absorption bands in lunar soils were identified. For example, at least two distinct absorption bands are required in the 1.2 mm region of the spectrum, and neither of these bands can be attributed to plagioclase or agglutinates, but are found instead to be largely due to pyroxene. Citation: Noble, S. K., C. M. Pieters, T. Hiroi, and L. A. Taylor (2006), Using the modified Gaussian model to extract quantitative data from lunar soils, J. Geophys. Res., 111,, doi: /2006je Introduction [2] The Moon is the only airless body from which we have direct geologic samples, though even those samples are restricted to a few isolated locales on the lunar nearside. The U.S. Apollo and Soviet Luna samples are an invaluable resource which can provide ground truth for our remotely sensed data. By understanding in detail the relationship of laboratory spectra of these samples to their maturity, composition, and chemistry, we can test the abilities and limitations of applying quantitative deconvolution methods to comparable spectral resolution remotely sensed data, such as that which will be obtained by missions like Japan s SELENE and India s Chandrayaan-1. [3] The Lunar Soil Characterization Consortium (LSCC) was formed to fully characterize the physical, chemical, and spectral properties of a suite of lunar soils [e.g., Taylor et al., 1999a, 1999b, 2000a, 2000b, 2000c, 2001a, 2001b, 2002, 2003; Pieters et al., 2000, 2002; Pieters and Taylor, 2003]. Soil samples selected for the consortium study were chosen to represent the widest variety of compositions and maturities from the lunar environment. The soils are listed in Table 1 along with their maturity (as indicated by I s /FeO) and the measured modal abundance of selected mineral and glass phases. 1 Brown University, Providence, Rhode Island, USA. 2 Now at NASA Johnson Space Center, Houston, Texas, USA. 3 University of Tennessee, Knoxville, Tennessee, USA. Copyright 2006 by the American Geophysical Union /06/2006JE002721$09.00 [4] The soils were wet sieved (triply distilled water) into several size fractions and reliable data on mineral phases were acquired. The consortium has concentrated on the mm and mm size fractions, because the optical properties of these sizes bear the greatest resemblance to those of the bulk soil [Pieters et al., 1993]. The <10 mm fraction is included in the consortium study because of its importance in understanding the effects of space-weathering processes [see Noble et al., 2001], but is not included in the MGM analyses because mineral bands are too weak to provide reliable information, largely because of the effects of abundant nanophase metallic Fe (npfe 0 ). The spectra of the finer fractions, which have high surface-to-volume ratios, tend to be more strongly affected by space weathering than larger grain sizes because nanophase-fe-bearing rims (a major product of space weathering) are surface correlated. The optical effects of nanophase-fe-bearing rims include a reddening and darkening of the spectrum, as well as an attenuation of the characteristic absorption bands [Pieters et al., 1993; Hapke, 2001]. In the <10 mm size fraction, the absorption bands have been severely reduced such that useful compositional data can no longer be extracted. Likewise, the mm size fraction is more weathered than the mm fraction because of its higher surface-to-volume ratio, and thus has weaker absorption bands. 2. Methods [5] In order to quantify relatively weak mineral absorption bands in lunar soil spectra, we have chosen to utilize the modified Gaussian model (MGM) developed by 1of17

2 Table 1. Basic Characteristics of the Lunar Soils Used in This Study Sample Relab ID Morris I s /FeO a LSCC I s /FeO Plagioclase Agglutinitic Glass Total Pyx Low-Ca: High-Ca Pyx mm Mare LR-CMP : LR-CMP : LR-CMP : LR-CMP : LR-CMP : LR-CMP : LR-CMP : LR-CMP : LR-CMP : mm Highland LR-CMP : LR-CMP : LR-CMP : LR-CMP : LR-CMP : LR-CMP : LR-CMP : LR-CMP : LR-CMP : LR-CMP : mm Mare LR-CMP : LR-CMP : LR-CMP : LR-CMP : LR-CMP : LR-CMP : LR-CMP : LR-CMP : LR-CMP : mm Highland LR-CMP : LR-CMP : LR-CMP : LR-CMP : LR-CMP : LR-CMP : LR-CMP : LR-CMP : LR-CMP : LR-CMP :33 a Bulk <250 mm I s /FeO data from Morris [1978]. Sunshine et al. [1990, 1999]. The MGM is a method for deconvolving spectra into individual absorption bands. On the basis of crystal field theory [Burns, 1993], the MGM been demonstrated to be a physically realistic model of electronic transition absorption bands. There has been substantial success in applying the MGM to modeling minerals and simple mixtures [Sunshine and Pieters, 1993, 1998]. The model has also been applied to Martian meteorites and even remote data from Mars [Sunshine et al., 1993; Mustard and Sunshine, 1995]. Applying the MGM to lunar soils requires special consideration because the lunar continuum, which is largely a result of spaceweathering processes, must first be accurately modeled and removed. [6] We have developed a method of continuum removal specifically for lunar soil samples [Hiroi et al., 2000], utilizing a three-component solution consisting of a linearin-energy term (c 1 ), a linear-in-wavelength term (c 1 ), and an offset (c 0 ), as shown in equation (1) below. This threeterm continuum allows more realistic solutions than the earlier method of double-linear removal [Hiroi and Pieters, 1998]. CðlÞ ¼ c 1 =l þ c 0 þ c 1 l l ¼ wavelength c 1 ; c 0 ; c 1 ¼ constants This method of continuum removal is compatible across a large array of compositions and maturities. Ueda et al. [2002] attempted to further account for space-weathering effects by adding an additional term to this continuum removal method, on the basis of Hapke s space-weathering model [Hapke, 2001]. However, it was found that this modified equation only works well for very small degrees of space weathering, and thus it is not generally appropriate for lunar soils. [7] In MGM analysis, each absorption band is fit to a modified Gaussian shape which can be fully described by three terms: band center, band strength, and bandwidth ð1þ 2of17

3 Figure 1. Example deconvolution of the mm fraction of highland soil 61221: (a) Three-band deconvolution; arrows indicate systematic errors illustrating the need for additional bands. (b) Five-band deconvolution; persistent error near 1.25 mm is marked with an arrow. (c) Six-band deconvolution; note that the error has been reduced. The bands are labeled one through six as referred to throughout the text. (FWHM), as shown in Figure 1a. The starting values for the continuum are derived using contact points at 0.75, 1.5, and 2.6 mm for each sample. The same starting parameters are given for the center and widths of absorption bands for every sample. The fitting program allows additional constraints to be placed on the range of centers, widths, and strengths for each band. The model simultaneously fits both the continuum and the absorption bands, while minimizing the RMS error. Additional bands are added as needed until the residual is small and there are no systematic features (Figures 1b and 1c). The resulting band parameters (center, strength, and width) provide quantitative information about the best fit for each absorption band that can then be directly compared to measured compositional data. [8] Lunar soils are composed of several minerals, as well as volcanic glasses and weathering products. Typical spectra of the more common components of lunar soils are shown in Figure 2. Lunar spectra tend to be dominated by pyroxene, which has very prominent absorption features centered near 1 and 2 mm, as well as a weaker band near 1.2 mm [e.g., Sunshine et al., 1990]. The locations of the 1 mm and 2 mm features are highly dependent on the pyroxene composition: High-Ca pyroxenes have bands centered at longer wavelengths than low-ca pyroxenes [Adams, 1974; Burns, 1993; Cloutis and Gaffey, 1991]. Iron content also plays a role in band location, with the 1 and 2 mm bands moving to longer wavelengths and the 1.2 mm band strengthening with increasing iron [Adams, 1974; Klima et al., 2005]. The 1.2 mm 3of17

4 additional band is required because of the presence of ilmenite-rich pyroclastic glass. The final continuum and Gaussian parameters are compiled in Tables 4a, 4b, 5a, and 5b. With six (or seven) bands, the residual errors have been reduced such that no systematic structure is apparent, as can be seen in Figure 4. Figure 2. Representative reflectance spectra of the major components of lunar soils. The following samples from the RELAB database were used: olv: PO-CMP-026, plag: SW-CMP-012, glass: LS-CMP-035A, cpx: LS-CMP-009, opx: SB-RGB-006, aggl: LS-CMP-045, ilm: JB-JLB-308. region is particularly complex with several soil components expressing absorption bands in the region. Plagioclase feldspar has a band at about 1.2 mm[adams, 1975; McCord et al., 1981]. Olivine has a compound band centered near 1.1 mm [e.g., Burns, 1993; Sunshine et al., 1990]. Glasses often have broad absorption bands in the same general region [e.g., Mao et al., 1978]. The glass sample in Figure 2 is a quenched glass created from an Apollo 17 impact melt breccia [Tompkins et al., 1996]. Compared to the quenched glass, the spectrum of agglutinates (hand picked from Apollo 11 soil 10084) is very dark and almost featureless [Pieters et al., 1993]. Agglutinates and agglutinitic glasses are rich in nanophase and larger blebs of metallic iron that act to reduce their albedo. Ilmenite and other opaque minerals also darken the spectra throughout the region of interest. 3. Results [9] Initially, each soil was fit with only three bands centered at approximately 1, 1.28, and 2 mm, using the initial parameters listed in Table 2. As demonstrated in Figure 1a, these three-band fits produced unacceptable systematic errors, indicating that additional bands are necessary [e.g., Sunshine et al., 1990]. This was not unexpected, since the compositional data [Taylor et al., 2001a] indicate that both high-ca and low-ca pyroxenes are present in all of the soils. The model was then run using five bands: a band near 1.28 mm, plus two bands near 1 mm and two near 2 mm to account for the high-ca and low-ca pyroxene. An example of a five-band fit is shown in Figure 1b. With five bands, the errors have been considerably reduced. [10] However, it was noticed that even with the five-band fit, there was still a consistent error in the results for all samples between approximately 1.2 and 1.3 mm. The recurring error is highlighted in several examples in Figure 3. A sixth band is required to eliminate this error. The initial parameters for the six-band fits are listed in Table 3 and an example is shown in Figure 1c. In addition, a seventh band was needed at 0.6 mm for the four Apollo 17 soils. This 4. Discussion 4.1. One and Two mm Bands (Pyroxenes) [11] Different compositions of pyroxene appear to contribute to all six of the bands identified in our study. In Figure 5, graphs are shown of the total pyroxene content versus the band strength for each of the six bands. A strong correlation is expected between the total pyroxene content and the 1 mm and 2 mm bands (bands 1, 2, 5, and 6), and in fact, all four bands do show a correlation with pyroxene content, though the correlation with band 5 is weaker than the others. In addition, it appears that the 1.2 mm bands (bands 3 and 4) are correlated to pyroxene content as well. These 1.2 mm bands are discussed in further detail in the next section. Some of the scatter seen in these graphs relates to differences in degree of space weathering among the samples. Because the products of space weathering result in an attenuation of absorption bands, more mature soils will tend to have weaker bands, regardless of pyroxene content. Also, the different compositions of pyroxenes will effect which band they correlate with; as is discussed further below, bands 1 and 5 should be dominated by orthopyroxene and bands 2 and 6 will be more strongly influenced by clinopyroxene. [12] Adams [1974] illustrated the regular relationship between the band centers of the 1 mm versus the 2 mm band for a suite of pyroxenes and demonstrated how the wavelengths of the bands shift as a result of differing Ca 2+ and Fe 2+ contents. In Figure 6, we have superimposed the band center results from our MGM deconvolutions onto his original data along with similar data from other studies [Adams, 1975; Hazen et al., 1978; Cloutis and Gaffey, 1991]. The modeled bands for the LSCC samples fall roughly within the same range as the previous works for pyroxenes. Specifically, the results of band 1 versus band 5 (the low-ca pyroxene bands) for both mare and highland soils fall in the middle of the previous data for orthopyroxene. The results for the mare soils of band 2 versus band 6 (the high-ca pyroxene bands) also fall squarely within the previous data. [13] By contrast, band centers for the majority of the highland high-ca pyroxene bands (2, 6) fall slightly to the left of the data cloud. Specifically, for the highland soils in our study, either band 2 (at 1 mm) is shifted to slightly longer wavelengths or band 6 (at 2 mm) is shifted to slightly shorter wavelengths than was the case for the mare soils. Adams [1974] suggested that the addition of olivine would cause samples to fall above this l-l trend. However, in our Table 2. Starting Parameters for Three-Band Deconvolutions Modified Gaussians Center Constraints Width Constraints mm/± mm/± mm/± mm/± mm/± mm/± of17

5 Figure 3. Errors from five-band deconvolutions for several mare and highland soils. The persistent error near mm can be identified in all of the soils and indicates that an additional band is necessary. suite, both the highland and mare samples are relatively low in olivine, containing only about 1 5 vol.%. In addition, the highland samples generally contain slightly less olivine than the mare samples. Thus this is unlikely to be the cause of the observed deviation. The addition of Fe-bearing glass may also result in band centers that fall off the pyroxene trend. However, there is nothing in our data that suggests that the highland soils would be preferentially affected by the presence of glass. In fact, it is the four Apollo 17 mare soils that contain high concentrations of pyroclastic beads. It is those soils that might be expected to be most influenced by glass absorptions, but the mare soils all fall on the expected trend. Rather than band 2 (near 1 mm) being shifted to longer wavelengths, it is more likely that band 6 (near 2 mm) is shifted to shorter wavelengths for the highland soils. The low-pyroxene contents of the highland soils make them particularly difficult to fit. Furthermore, because of the method of continuum removal and band fitting utilized for these soils, we have found that the longestwavelength band (6) is the most difficult band from which to obtain consistent and reasonable results. While compositional differences cannot be ruled out, the shift seen in the highland soils is likely to be an artifact of the approach. Table 3. Starting Parameters for Six- (or Seven-) Band Deconvolutions Modified Gaussians Center Constraints Width Constraints mm/± mm/± mm/± mm/± mm/± mm/± mm/± mm/± mm/± mm/± mm/± mm/± (if needed) 0.60 mm/± mm/± of17

6 Table 4a. Mare Error and Continuum Parameters Continuum Parameters Sample RMS c( 1) c(0) c(1) mm E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E mm E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E-01 [14] The combined strength of the four 1 mm and 2 mm pyroxene bands (bands 1, 2, 5, and 6) is compared to total pyroxene content in Figure 7. As expected, there is a clear trend of increasing band strength with increasing pyroxene content. The standard deviation of a linear regression fit to this plot is ±5.17%. As in Figure 5, some of the scatter in this graph is due to differences in degree of space weathering. Though it may seem counterintuitive, there is also a tendency for highland soils to actually have stronger pyroxene bands than mare soils for a given pyroxene content. This is not due simply to differences in degree of weathering, but rather relates to the fact that highland soils contain fewer opaques, such as ilmenite, which tend to reduce spectral contrast overall. [15] One of the goals of this study was to test whether the model could predict the relative abundance of low-ca and high-ca pyroxene for lunar soils. Since using only three bands produced such poor fits to the data (Figure 1a), it is clear that multiple bands are required to account for the pyroxene diversity. In determining soil mineralogy and composition, the pyroxenes were classified into four categories: orthopyroxene, pigeonite, Mg-rich clinopyroxene (augite), and Fe-rich clinopyroxene [Taylor et al., 1999a]. A pyroxene quadrilateral with the locations of these pyroxene types is shown in Figure 8. In order to relate the mineralogy to the spectral analysis, in some cases, we have combined the orthopyroxene and pigeonite measurements into a low-ca pyx category and the Mg-rich and Fe-rich pyroxenes into a high-ca pyx category. This grouping on the basis of Ca content necessarily neglects the role of iron in determining band center, but some simplifying assumption was necessary to reduce the data for comparison with the pyroxene-related bands. We have chosen to concentrate on calcium differences because they are more important for determining band centers than iron. [16] If we assume that the strength of bands 1 and 2 is related largely to the calcium content of the pyroxene, then the ratio of the strengths should be directly proportional to the ratio of low-ca to high-ca pyroxene in a given soil [Sunshine and Pieters, 1993]. The same should also be true of bands 5 and 6. Comparison of the ratio of low-ca to high-ca pyroxene to the ratio of the strength of band 1 to band 2, as well as of band 5 to band 6, are shown in Figures 9a and 9b, respectively. [17] For the 1 mm region (Figure 9a), the mm and the mm fractions of mare soil, as well as the mm fraction of highland soils, show strong correlations between band strength and composition. The mm fractions of highland soils are, however, considerably more scattered. The combination of low total pyroxene contents and the tendency for weathering products to increase in the finer fractions likely contributes to the poor fit for those soils. If the highland mm fraction is ignored, the remaining soils have an R 2 of 0.74 and a standard deviation of ±0.92. [18] In contrast to the results in the 1 mm region, no simple relationship is seen between composition and band strength in the 2 mm region (Figure 9b). As evidenced by Figure 6, our model has been shown to have difficulty in fitting the 2 mm region, thus it is disappointing, but not entirely surprising that such a poor relationship is observed The 1.2 mm Region [19] Interpretation of the 1.2 mm region of lunar soil spectra is complex. At least two distinct bands appear to be required to adequately fit the LSCC soil data, as indicated by the consistent error obtained when only one band is utilized (Figure 3). With two bands, we have eliminated the systematic errors in our fits (Figure 4); however, given the complexity of this region, it is entirely possible that multiple components are represented by these bands, including ferrous iron in the M1 site of orthopyroxene, pigeonite, and clinopyroxene, as well as iron in plagioclase. [20] The two bands identified by our model are labeled bands 3 and 4. Band 3 is a stronger narrower band at shorter wavelengths ( mm). Band 4 is a wider, weaker band at longer wavelengths ( mm). The two bands are highly correlated in their band strength (R 2 = Table 4b. Highland Error and Continuum Parameters Continuum Parameters Sample RMS c( 1) c(0) c(1) mm E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E mm E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E E-02 6of17

7 Table 5a. Mare Soils Deconvolution Results Sample Band 1 Band 2 Band 3 Band 4 Band 5 Band 6 Band 7 Center Width Strength Center Width Strength Center Width Strength Center Width Strength Center Width Strength Center Width Strength Center Width Strength mm mm Table 5b. Highland Soils Deconvolution Results Sample Band 1 Band 2 Band 3 Band 4 Band 5 Band 6 Center Width Strength Center Width Strength Center Width Strength Center Width Strength Center Width Strength Center Width Strength mm mm of17

8 Figure 4. Errors from six-band deconvolutions for several mare and highland soils. Systematic structure has been eliminated. 8of17

9 Figure 5. Total pyroxene content versus band strength for each of the six bands. For this and all subsequent plots: (open symbols) highland soils, (closed symbols) mare, (diamonds) mm size fractions, and (squares) mm fraction. In addition: (black) mature soils, (medium gray) submature soils, and (light gray) immature soils. 9of17

10 Figure 6. Plot of location of the center of the 1 mm band versus the center of the 2 mm band. The data from this study are superimposed on data from previous studies: Adams Adams [1974, 1975]; Hazen Hazen et al. [1978]; C&G Cloutis and Gaffey [1991]. 0.75) (Figure 10). This correlation suggests that the cause(s) of the bands is not independent. [21] A band at roughly 1.2 mm in lunar soils has often been attributed to plagioclase [e.g., McCord et al., 1981]. However, Klima et al. [2005] have clearly documented this feature in Fe-Mg pyroxenes. Similarly, using the MGM model, Sunshine et al. [1993] found that the 1.2 mm band in a Martian meteorite was due solely to the pyroxene in the sample and was not related to plagioclase. Our data confirm that this association with pyroxene is also the case for lunar soils. As illustrated in Figure 11, plagioclase content of the soil has almost no effect on the strength of band 3 or band 4, even for highland samples that are greater than 60 vol.% plagioclase. The centers of bands 3 and 4 (Figures 11c and 11d) do show slight correlations with plagioclase, suggesting that plagioclase may have some impact on this region, but it is clearly not a major contributor. [22] Agglutinitic glasses often constitute a significant fraction of lunar soils, up to 70 vol.% of our most mature samples. However, they have no influence on either the strength or location of bands 3 or 4, as can be seen in Figures 12a and 12b). Certain lunar soils are rich in volcanic glasses, some of which also have bands in the 1.2 mm region. For example, in our suite, the four Apollo 17 soils contain a significant fraction of volcanic glass beads. As was previously reported [Noble et al., 1999], this glass does appear to have some influence on the strength of bands in the 1.2 mm region in these soils (Figures 12c and 12d). [23] As illustrated in Figures 5c and 5d, total pyroxene content is correlated to the strength of bands 3 and 4. In Figure 13, we have separated the pyroxenes into their four classes and compared them each to the strength of bands 3 and 4: Orthopyroxene content shows a very weak correlation for highland soils, but no obvious relationship for mare soils; pigeonite, on the other hand, is well correlated to the strength of band 3 and, to a lesser extent, band 4 for both mare and highland soils; Mg-rich clinopyroxene shows no relationship with either band; and Fe-rich clinopyroxene, though constituting only a very small percentage of the total pyroxene, is surprisingly well correlated to the strength of both bands 3 and 4. [24] Olivine is a minor component of the soils in our study, composing no more than 5 vol.% of any of our samples. However, the absorption bands in olivine extend into the 1.2 mm region (Figure 1); thus it is possible that olivine could be contributing to one or both of the bands. Figures 14a and 14b illustrate that olivine appears to be very weakly correlated to band strength of both bands 3 and 4. There is no correlation between olivine content and band center for bands 3 and 4 (Figures 14c and 14d). The weak correlation with band strength is likely a consequence of the correlation between olivine content and pyroxene content and probably does not represent a true relationship between the 1.2 mm band and olivine, though the possibility cannot be fully eliminated Maturity [25] I s /FeO is a standard measurement of maturity (length of exposure time) for lunar soils. The parameter, I s, represents the relative strength of the ferromagnetic resonance (due to Fe 0 particles between about 4 and 33 nm) of a soil, which is then normalized by the total FeO content of the soil to account for compositional differences [Morris, 1978]. The I s /FeO for each of the size fractions of the LSCC soils were measured. Maturity, as indicated by I s /FeO, has little effect on the band center or widths of the deconvolved soils. Space weathering is known to reduce the strength of absorption bands [e.g., Pieters et al., 1993; Fischer and Pieters, 1994], thus it is not surprising that the band strength of nearly all the bands are anticorrelated to I s /FeO. Figure 15 shows plots of I s /FeO versus band strength for each of the six bands. Five of the six bands show essentially the same Figure 7. Combined band strength of bands 1, 2, 5, and 6 versus total pyroxene content. (open symbols) highland soils, (closed symbols) mare. In addition: (black) mature soils, (medium gray) submature soils, and (light gray) immature soils. R 2 = 0.79, s = ±5.17%. 10 of 17

11 Figure 8. Pyroxene quadrilateral showing the location of the average composition for the four pyroxene categories for each soil. trend: Immature soils may have strong or weak bands, but mature soils (those with high-i s /FeO values) never have strong bands. Band 4 (Figure 15d) does not follow this trend. Here we see that even mature soils can have strong band 4s; this is particularly true among the mare soils, which generally have a stronger band 4 than the highland soils. Though the previous graphs indicate that pyroxene is at least partially responsible for band 4, Figure 15d suggests that an additional soil component is also contributing to the band, and that component is either unaffected by space weathering, or it is actually created in the weathering process. 5. Conclusions [26] Dissecting the modeling of a suite of well-characterized lunar soils has demonstrated both the capabilities and the limitations of this model utilizing the MGM method of spectral deconvolution as a quantitative tool for lunar soils. The model does a good job eliciting data from the 1 and 1.2 mm regions. Low-pyroxene contents and high degrees of space weathering result in lower accuracy in predictions of pyroxene composition in the 2 mm region. [27] As a method of spectra deconvolution, the modified Gaussian model was able to identify both high-ca and Figure 9. Ratio of low-ca to high-ca pyroxene compared to the ratio of (a) the band strength of band 1 versus band 2 and (b) the band strength of band 5 versus band 6. The preponderance of low-ca pyroxene in the highlands is readily identified. Figure 10. The strength of band 3 versus the strength of band 4. The bands are highly correlated (R 2 = 0.75). 11 of 17

12 Figure 11. Plagioclase content versus properties of the 1.2 mm band (bands 3 and 4): (a) Strength of band 3. (b) Strength of band 4. (c) Center of band 3. (d) Center of band of 17

13 Figure 12. Agglutinate and glass content versus the 1.2 mm band (bands 3 and 4): (a) Strength of band 3 versus agglutinate content. (b) Strength of band 4 versus agglutinate content. (c) Volcanic glass content versus strength of band 3 for the Apollo 17 soils. (d) Volcanic glass content versus strength of band 4 for the Apollo 17 soils. 13 of 17

14 Figure 13. The 1.2-mm band (bands 3 and 4) versus the four categories of pyroxenes. 14 of 17

15 Figure 14. Olivine content versus the 1.2 mm band (bands 3 and 4): (a) Strength of band 3 versus olivine content. (b) Strength of band 4 versus olivine content. (c) Center of band 3 versus olivine content. (d) Center of band 4 versus olivine content. 15 of 17

16 Figure 15. Band strength versus maturity as measured by I s /FeO for each of the six bands. low-ca pyroxene in a suite of lunar soils with a range of weathering degrees. In addition to the well-known pyroxene absorption bands near 1 and 2 mm, it was discovered that at least two distinct 1.2 mm bands (bands 3 and 4) are required by the data. Plagioclase and agglutinates have been eliminated as possible causes of the 1.2 mm bands. Rather, pyroxenes appear to be largely responsible for those bands. The I s /FeO data suggest that there is also an additional unidentified component influencing the strength of band 4, probably related to space weathering. 16 of 17

17 [28] The strong correlation between band strength and pyroxene content (Figure 7) suggests that for high-quality spectra data, the MGM does allow a predictive capability, even for complex soils. Even more encouraging, beyond simply estimating total pyroxene content, the MGM may allow us to assess the ratio of high-ca to low-ca pyroxene from high-quality lunar soil spectra. [29] Several simplifying assumptions were made in this study, including the combining of four types of pyroxene into just two classes. The analysis presented here is a first step in understanding and unlocking the potential of MGM deconvolution for lunar soils. Our model only captures a small part of the physical complexity of these samples. Further work exploring this complexity may help resolve many of the issues encountered and lead to improvements in the predictive capabilities of the model. [30] Acknowledgments. Thank you to Jessica Sunshine for her encouragement and helpful suggestions. This manuscript benefited greatly from thorough reviews by Paul Lucey and Ed Cloutis. NASA support (NNG05-GG15G, C.M.P.; NNG05-GG41G, L.A.T.) is gratefully acknowledged. Support of Oak Ridge Associated University through the NASA postdoctoral program to S.K.N. is appreciated. RELAB is a multiuser facility supported under NAG References Adams, J. B. (1974), Visible and near-infrared diffuse reflectance spectra of pyroxenes as applied to remote sensing of solid objects in the solar system, J. Geophys. Res., 79, Adams, J. B. (1975), Interpretation of visible and near-infrared diffuse reflectance spectra of pyroxenes and other rock forming minerals, in Infrared and Raman Spectroscopy of Lunar and Terrestrial Materials, edited by C. Carr, pp , Elsevier, New York. Burns, R. G. (1993), Mineralogical Applications of Crystal Field Theory, 2nd ed., Cambridge Univ. Press, New York. Cloutis, E. A., and M. J. Gaffey (1991), Pyroxene spectroscopy revisited: Spectral-compositional correlations and relationship to geothermometry, J. Geophys. Res., 96, 22,809 22,826. Fischer, E. M., and C. M. Pieters (1994), Remote determination of exposure degree and iron concentration of lunar soils using VIS-NIR spectroscopic methods, Icarus, 111, Hapke, B. (2001), Space weathering from Mercury to the asteroid belt, J. Geophys. Res., 106, 10,039 10,073. Hazen, R. M., P. M. Bell, and H. K. Mao (1978), Effects of compositional variation on absorption spectra of lunar pyroxenes, Proc. Lunar Planet. Sci. Conf. 9th, Hiroi, T., and C. M. Pieters (1998), Modified Gaussian deconvolution of reflectance spectra of lunar soils, Proc. Lunar Planet. Sci. Conf. [CD-ROM], 29, Hiroi, T., C. M. Pieters, and S. K. Noble (2000), Improved scheme of Modified Gaussian deconvolution for reflectance spectra of lunar soils, Proc. Lunar Planet. Sci. Conf. [CD-ROM], 31, Klima, R. L., C. M. Pieters, and M. D. Dyar (2005), Pyroxene spectroscopy: Effects of major element composition on near, mid and farinfrared spectra, Proc. Lunar Planet. Sci. Conf. [CD-ROM], 36, Mao, A. L., P. M. Bell, and H. K. Mao (1978), Crystal field spectra of Luna 24 glass, Proc. Lunar Planet. Sci. Conf. 9th, abstract, McCord, T. M., R. N. Clark, B. R. Hawke, L. A. McFadden, P. D. Owensby, C. M. Pieters, and J. B. Adams (1981), Moon: Near-infrared spectral reflectance, a first good look, J. Geophys. Res., 86, 10,883 10,892. Morris, R. V. (1978), The surface exposure (maturity) of lunar soils: Some concepts and I s /FeO compilation, Proc. Lunar Planet. Sci. Conf. 9th, Mustard, J. F., and J. M. Sunshine (1995), Seeing through the dust: Martian crustal heterogeneity and links to the SNC meteorites, Science, 267, Noble, S. K., C. M. Pieters, L. A. Taylor, L. P. Keller, R. V. Morris, D. S. McKay, and S. J. Wentworth (1999), The optical properties of the finest fraction of lunar soils: Initial results and implications for weathering processes, Proc. Lunar Planet. Sci. Conf. [CD-ROM], 30, Noble, S. K., C. M. Pieters, L. A. Taylor, R. V. Morris, C. C. Allen, D. S. McKay, and L. P. Keller (2001), The optical properties of the finest fraction of lunar soil: Implications for space weathering, Meteorit. Planet. Sci., 36, Pieters, C. M., and L. A. Taylor (2003), Systematic global mixing and melting in lunar soil evolution, Geophys. Res. Lett., 30(20), 2048, doi: /2003gl Pieters, C. M., E. M. Fischer, O. Rode, and A. Basu (1993), Optical effects of space weathering: The role of the finest fraction, J. Geophys. Res., 98, 20,817 20,824. Pieters, C. M., L. A. Taylor, D. S. McKay, S. J. Wentworth, R. V. Morris, and L. P. Keller (2000), Spectral characteristics of lunar mare soils, Proc. Lunar Planet. Sci. Conf. [CD-ROM], 31, Pieters, C. M., L. A. Taylor, and the LSCC (2002), The perplexing role of TiO 2 in the evolution of lunar soils, Proc. Lunar Planet. Sci. Conf. [CD-ROM], 33, Sunshine, J. M., and L. A. McFadden (1993), Reflectance spectra of the Elephant Moraine A79,001 meteorite: Implications for remote sensing of planetary bodies, Icarus, 105, Sunshine, J. M., and C. M. Pieters (1993), Estimating modal abundances from the spectra of natural and laboratory pyroxene mixtures using the modified Gaussian model, J. Geophys. Res., 98, Sunshine, J. M., and C. M. Pieters (1998), Determining the composition of olivine from reflectance spectroscopy, J. Geophys. Res., 103, 13,675 13,688. Sunshine, J. M., C. M. Pieters, and S. F. Pratt (1990), Deconvolution of mineral absorption bands: An improved approach, J. Geophys. Res., 95, Sunshine, J. M., C. M. Pieters, S. F. Pratt, and K. S. McNaron-Brown (1999), Absorption band modeling in reflectance spectra: Availability of the modified Gaussian model, Proc. Lunar Planet. Sci. Conf. [CD-ROM], 30, Taylor, L. A., A. Patchen, R. V. Morris, D. Taylor, C. M. Pieters, L. P. Keller, D. S. McKay, and S. J. Wentworth (1999a), Chemical and mineralogical characterization of the 44 20, 20 10, and <10 micron fractions of lunar mare soils, Proc. Lunar Planet. Sci. Conf. [CD-ROM], 30, Taylor, L. A., C. M. Pieters, R. V. Morris, L. P. Keller, D. S. McKay, A. Pachen, and S. J. Wentworth (1999b), Integration of the chemical and mineralogical characteristics of lunar soils with reflectance spectroscopy, Proc. Lunar Planet. Sci. Conf. [CD-ROM], 30, Taylor, L. A., R. V. Morris, C. M. Pieters, A. Patchen, L. P. Keller, D. Taylor, S. J. Wentworth, and D. S. McKay (2000a), Chemical characterization of lunar mare soils, Proc. Lunar Planet. Sci. Conf. [CD-ROM], 31, Taylor, L. A., R. V. Morris, L. P. Keller, C. M. Pieters, A. Pachen, D. Taylor, S. J. Wentworth, and D. S. McKay (2000b), Major contributions to spectral reflectance opacity by non-agglutinitic, surface-correlated nanophase iron, Proc. Lunar Planet. Sci. Conf. [CD-ROM], 31, Taylor, L. A., A. Pachen, D. Taylor, R. V. Morris, C. M. Pieters, L. P. Keller, S. J. Wentworth, and D. S. McKay (2000c), Mineralogical characterization of lunar mare soils, Proc. Lunar Planet. Sci. Conf. [CD- ROM], 31, Taylor, L. A., C. M. Pieters, L. P. Keller, R. V. Morris, and D. S. McKay (2001a), Lunar mare soils: Space weathering and the major effects of surface-correlated nanophase Fe, J. Geophys. Res., 106, 27,985 27,999. Taylor, L. A., J. Cahill, A. Pachen, C. M. Pieters, R. V. Morris, L. P. Keller, and D. S. McKay (2001b), Mineralogical and chemical characterization of lunar highland regolith: Lessons learned from mare soils, Proc. Lunar Planet. Sci. Conf. [CD-ROM], 32, Taylor, L. A., A. Pachen, J. Cahill, C. M. Pieters, R. V. Morris, L. P. Keller, and D. S. McKay (2002), Mineralogical and glass characterization of Apollo 14 soils, Proc. Lunar Planet. Sci. Conf. [CD-ROM], 33, Taylor, L. A., C. M. Pieters, A. Pachen, D. H. Taylor, R. V. Morris, L. P. Keller, and D. S. McKay (2003), Mineralogical characterization of lunar highland soils, Proc. Lunar Planet. Sci. Conf. [CD-ROM], 34, Tompkins, S., C. M. Pieters, and G. Ryder (1996), Spectral variations among crystalline and glassy Apollo 17 impact melt, Proc. Lunar Planet. Sci. Conf. [CD-ROM], 27, Ueda, Y., T. Hiroi, C. M. Pieters, and M. Miyamoto (2002), Expanding the modified Gaussian model to include the space weathering effects: Estimation of the weathering degrees of pulse-laser treated olivine samples, Proc. Lunar Planet. Sci. Conf. [CD-ROM], 33, S. K. Noble, NASA JSC, Houston, TX 77058, USA. (sarah.k.noble@ nasa.gov) C. M. Pieters and T. Hiroi, Brown University, Providence, RI 02912, USA. L. A. Taylor, University of Tennessee, Knoxville, TN 37996, USA. 17 of 17