sign in sign up
<<
Home , Composition of Mars, Magma

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 111, E03S10, doi:10.1029/2005JE002645, 2006 [printed 112(E3), 2007] Click Here for Full Article

Bulk composition and early differentiation of G. Jeffrey Taylor,1 W. Boynton,2 J. Bru¨ckner,3 H. Wa¨nke,3 G. Dreibus,3 K. Kerry,2 J. Keller,2 R. Reedy,4 L. Evans,5 R. Starr,6 S. Squyres,7 S. Karunatillake,7 O. Gasnault,8 S. Maurice,8 C. d’Uston,8 P. Englert,1 J. Dohm,2,9 V. Baker,2,9 D. Hamara,2 D. Janes,2 A. Sprague,2 K. Kim,2 and D. Drake10 Received 23 November 2005; revised 12 April 2006; accepted 20 April 2006; published 19 December 2006.

[1] We report the concentrations of K, Th, and Fe on the surface, as determined by the gamma ray spectrometer onboard the . K and Th are not uniformly distributed on Mars. K ranges from 2000 to 6000 ppm; Th ranges from 0.2 to 1 ppm. The K/Th ratio varies from 3000 to 9000, but over 95% of the surface has K/Th between 4000 and 7000. Concentrations of K and Th are generally higher than those in basaltic Martian (K = 200–2600 ppm; Th = 0.1–0.7 ppm), indicating that Martian meteorites are not representative of the bulk . The average K/Th in the crust is 5300, consistent with the Wa¨nke-Dreibus model composition for bulk Mars. Fe concentrations support the idea that bulk Mars is enriched in FeO compared to . The differences in K/Th and FeO between Earth and Mars are consistent with the planets accreting from narrow feeding zones. The concentration of Th on Mars does not vary as much as it does on the Moon (where it ranges from 0.1 to 12 ppm), suggesting that the primary differentiation of Mars differed from that of the Moon. If the average Th concentration (0.6 ppm) of the surface is equal to the average of the entire crust, the crust cannot be thicker than about 118 km. If the crust is about 57 km thick, as suggested by geophysical studies, then about half the Th is concentrated in the crust. Citation: Taylor, G. J., et al. (2006), Bulk composition and early differentiation of Mars, J. Geophys. Res., 111, E03S10, doi:10.1029/ 2005JE002645. [printed 112(E3), 2007]

1. Introduction lution of Mars. In this paper we focus on using the concentrations of K, Th, and Fe to test models for the bulk [2] The Mars Odyssey gamma ray spectrometer (GRS) , which has implications for planetary provides the first direct determination of elemental concen- , and to investigate the formation and subsequent trations of the entire [Boynton et al., 2004], differentiation of the planet. including materials at a depth of about one third meter from the surface. Although the spatial resolution is on the order of 500 km, the data allow us to address significant global 2. Methods problems concerning the geochemical and geological evo- [3] The Odyssey GRS and data reduction methods are described by W. V. Boynton et al. (Concentration of H, Si, Cl, K, Fe, and Th in the low and mid latitude regions of 1 Institute of and Planetology, Honolulu, Hawaii, Mars, submitted to Journal of Geophysical Research, 2006, USA. 2Lunar and Planetary Laboratory, University of Arizona, Tucson, hereinafter referred to as Boynton et al., submitted manu- Arizona, USA. script, 2006). In this paper we present three types of data 3Max-Planck-Institu¨t fu¨r Chemie, Mainz, Germany. collected through March 2005. First, we provide maps of 4Institute of Meteoritics, University of New Mexico, Albuquerque, New the concentrations of K, Th, K/Th, and Fe. These maps are Mexico, USA. 5 derived from a 2 base map and smoothed with a 10 boxcar Computer Sciences Corporation, Lanham, Maryland, USA. ° ° 6Department of Physics, Catholic University of America, Washington, filter. They are useful for showing global variations in DC, USA. concentration. The maps are essentially global for K, Th, 7Center for Radiophysics and Space Research, Cornell University, and K/Th, but are restricted to regions of relatively low Ithaca, New York, USA. for Fe (see below). 8Centre d’Etude Spatiale des Rayonnements, Centre National de la Recherche Scientifique/Universite´ Paul Sabatier, Toulouse, France. [4] Second, we use 5° Â 5° binned data for x-y plots. The 9Department of Hydrology and Resources, University of Arizona, 5° Â 5° data have been smoothed with a 10° filter around Tucson, Arizona, USA. each point. (Filtering removes significant levels of noise in 10 TechSource, Santa Fe, New Mexico, USA. the initial data.) For Fe data, we report only those points in regions where H contents are low enough to not interfere Copyright 2006 by the American Geophysical Union. 0148-0227/06/2005JE002645$09.00 in the determination of the Fe concentration. Hydrogen has

E03S10 1of16 E03S10 TAYLOR ET AL.: MARS COMPOSITION AND DIFFERENTIATION E03S10 a high cross section for capturing thermal neutrons, so it can uncertainty is over 3000. On the other hand, other areas significantly affect the flux of thermal neutrons in the upper appear to be distinctly higher or lower, such as the high 30 cm of the Martian surface. To account for the effect of region in and of Valles Marinaris, at least at the 1s H, we use a correction procedure that involves both the level. Details of the variations in K/Th are discussed by measured fluxes of gamma rays from H, Fe, and Si, and the Taylor et al. [2006]. fluxes calculated from a neutron transport–gamma ray [7] Fe concentrations (Figure 3) in the northern plains are production model (Boynton et al., submitted manuscript, higher than in the southern highlands. Nevertheless, con- 2006). We accomplished that by using the H mask described centrations are almost everywhere higher than in typical by Boynton et al. (submitted manuscript, 2006). This terrestrial and generally consistent with the high Fe approach has resulted in reasonable values at equatorial contents of Martian meteorites (discussed further below) latitudes. Because the approach results in uncertain values at and with the inferred high FeO in bulk silicate Mars. higher polar latitudes where the influence of hydrogen Additional data are presented in the following sections, in dominates elemental signatures, the results presented here which we compare our GRS data to the compositions are constrained using a mask based upon both H concen- of Martian meteorites, test models for the bulk composition tration values and described by Boynton et al. (submitted of Mars, and explore the record of the planet’s early manuscript, 2006). The H mask corresponds to roughly plus differentiation. or minus 45° of latitude from the equator. We restrict our analysis of binned data for K and Th to between 75° south 4. Relation of Surface Compositions to the Entire and 75° north latitude; at higher latitudes the concentrations Crust of K and Th are diluted by high water contents, increasing the uncertainty of those measurements. The [8] The GRS measures the composition of the upper few typical uncertainties (relative percent) stemming from tens of centimeters of the dusty Martian surface, yet we are counting statistics for an average point are 5% for K, 10% trying to understand the formation of the entire crust and the for Th, and 5% for Fe. bulk composition of the entire planet. Several factors allow [5] Third, we use summed spectra to compute the con- us to extrapolate our surface measurements to great depth. centrations in specific geologic regions (described below). One is that K and Th have very similar geochemical These spectra involve large counting times (>3 Â 106 s), behavior in igneous systems, as shown by their similar, and because statistical uncertainties vary as the square root and very low (1), -melt distribution coefficients of the counting time, they have correspondingly low statis- [Beattie, 1993; Borg and Draper, 2003; Hauri et al., 1994]. tical uncertainties. The uncertainty varies with the size of Both elements are incompatible and their concentrations in the region, hence with the total counting times in the are not greatly affected by source composition summed spectra. Typical uncertainties are 1% for K, 3– or crystallizing phase, even when garnet is involved. There 5% for Th, and 2–4% for Fe. These are the uncertainties in are interesting exceptions, however. Th is highly compatible the measurements and define the confidence to which we in phosphate [, 1995]. Phosphates form late know the means. It does not reflect the variation in in the of a and are unlikely to be concentrations across the Martian surface. retained in a source region, so probably do not play a role in fractioning K from Th during igneous processes. 3. Results However, in principle, it could be significant if mantle regions were matasomatized by fluids that contained [6] Maps of the distribution of K, Th, K/Th, and Fe are phosphate components. K is compatible in phlogopite presented in Figures 1–3. K and Th are not uniformly [Halliday et al., 1995] and somewhat compatible in amphi- distributed on Mars. The northern plains from about À60° bole [Halliday et al., 1995], so if these phases were present to +180°E are rich in both, though the higher-than-average in the Martian mantle, it could lead to fractionation of K Th region extends much further south into the highlands. from Th. Nevertheless, in general, K/Th in a flow Both are generally medium to low over . There is reflects the ratio in its mantle source region. distinctly higher K and Th in the highlands in the region [9] Another reason for the upper few tens of centimeters À15° to À45°S, and extending from À150° to +165°E. The reflecting the bulk composition of the crust is that the crust region west of Hellas contains average K, but has relatively was constructed by intrusion of magma far below the high Th. The K/Th ratio varies by somewhat over a factor of surface and extrusion of lava on its surface. On Earth, the two, but most of the surface area is between 4500 and ratio of the intrusive to extrusive magma volumes is 5:1 6500 ppm. It is distinctly low west of in [Crisp, 1984] in oceanic (basaltic) regions. If this holds for , in the region where meets Mars, the abundant lava flows that decorate the surface , in western , and in Syrtis (themselves forming thick sequences of ) are accom- Major Planum. K/Th is high in and panied by five times as much magma of similar composi- surrounding region, , and in the Hellas tion. Of course, we see only the uppermost, youngest lava basin. These modest variations may reflect a combination flows, and it is possible that magma compositions changed of bulk Martian K and Th concentrations, igneous processes, with time. Basaltic SNC meteorites argue that there was no and aqueous-related processes, including alteration, erosion, systematic change in magma composition with time. All are and deposition. However, the uncertainties in the K/Th <600 Ma in age, but were produced from a range of mantle measurements make these apparent differences statistically sources [e.g., Borg et al., 1997]. uncertain (Figure 2). For example, Hellas appears to have [10] A final reason for thinking that the upper few tens of elevated K/Th with values up to 10,000, but the 1s centimeters reflects deeper material is that the crust has been

2of16 E03S10 TAYLOR ET AL.: MARS COMPOSITION AND DIFFERENTIATION E03S10

Figure 1. Maps of the distribution of K and Th on Mars, as measured by the Mars Odyssey gamma ray spectrometer. Data have been smoothed using a 10° boxcar filter. The data are displayed over a shaded relief map of Mars, with mission landing sites indicated: V1 and V2, and 2; PF, Pathfinder; M, in ; G, in Crater. churned by impacts. Hartmann et al. [2001] show that the surfaces could have been gardened more or less uniformly surfaces formed before the early would have to depths of 1–2 km [Hartmann and , 2001]. been gardened to at least a few meters. Older surfaces would from the many large basins on Mars would have excavated have been gardened even deeper. The oldest materials from depths of tens of kilometers. Thus impact

3of16 E03S10 TAYLOR ET AL.: MARS COMPOSITION AND DIFFERENTIATION E03S10

Figure 2. Maps of (top) the distribution of K/Th on Mars and (bottom) the one-sigma uncertainty in the measurements. Data have been smoothed using a 10° boxcar filter. The data are displayed over a shaded relief map of Mars, with mission landing sites indicated: V1 and V2, Viking 1 and 2; PF, Pathfinder; M, Opportunity in Meridiani Planum; G, Spirit in Gusev Crater. processes may provide us with an upper surface that reflects fractionate K and Th. In fact, an important question is the the entire crust. extent to which the range in K/Th observed across the [11] The case for the surface representing the entire crust surface of Mars (Figure 2) is due to aqueous processes. is not open and shut, however. Aqueous processes can Equally important, dust blankets much of the surface and

4of16 E03S10 TAYLOR ET AL.: MARS COMPOSITION AND DIFFERENTIATION E03S10

Figure 3. Map of the distribution of Fe on Mars, as measured by the Mars Odyssey gamma ray spectrometer. Data have been smoothed using a 10° boxcar filter. The mapped area encompasses only that portion of Mars in which H content does not dominate corrections. (Such corrections are not necessary for radioactive elements; Figures 1–2.) The data are displayed over a shaded relief map of Mars, with mission landing sites indicated: V1 and V2, Viking 1 and 2; PF, Pathfinder; M, Opportunity in Meridiani Planum; G, Spirit in Gusev Crater. The black line represents the 0-km contour, a reasonable separation between highlands and lowlands.

the fine materials seem to have very similar compositions at 5.7 and 4.4 [Rubin et al., 2000; Sautter et al., 2002], which all the Viking, Pathfinder, Spirit, and Opportunity landing is within the range of most SNC meteorites. SNC meteorites sites [e.g., Gellert et al., 2004; Rieder et al., 2004; Yen et al., contaminated by terrestrial groundwater, such as DaG 476 2005]. The presence of these pervasive fines might blunt and SaU 005, have Th/U < 0.5.) The average surface is differences in bedrock composition from place to place on clearly higher in Th and K than the average SNC . Mars. The strong evidence for the presence of layered Overall, it appears that the Martian meteorites are not chemical and siliciclastic at the Opportunity site representative of the Martian crust, as also reported by [Squyres et al., 2004] suggests that some regions within a analysis of data from the Thermal Emission Spectrometer GRS resolution element are sedimentary in origin, hence onboard [Hamilton et al., 2003]. The may not reflect the composition of the igneous crust. On the source areas for SNC meteorites must either be small other hand, large areas are not dominated by fines or relative to the GRS resolution or are covered with enough chemical sediments, as shown by TES spectra that indicate dust (altered surface material) to mask their chemical the presence of igneous minerals [Bandfield et al., 2000]. signatures. This suggests that the minimum concentrations of K and Th in surface rocks might be lower than the 5. Comparison to Martian Meteorites minimum values we observe in the GRS data set. [13] Recent publications have shown that basaltic SNC [12] Studies of Martian meteorites (also called SNC meteorites (shergottites and -phyric shergottites) meteorites, for Shergottites, , and Chassignites) have distinct trends in their abundance pat- have shed light on Martian magmatic history, nature of the terns such as La/Yb ratios, initial 87Sr/86Sr, 147Sm/144Nd, mantle reservoirs that gave rise to meteorite parent magmas, 180Hf/183W, and 176Lu/177Hf, and fugacity [Borg et Martian bulk composition, and aqueous alteration processes al., 1997, 2003; Herd et al., 2002; Herd, 2003; Goodrich et on Mars [e.g., Bridges et al., 2001; McSween, 2003]. Thus al., 2003; Borg and Draper, 2003; McLennan, 2003]. These the study of Martian meteorites plays an important role in distinct trends were originally interpreted to reflect crustal understanding Mars. However, as Figure 4 shows, SNC assimilation accompanied by fractional crystallization. This meteorites in general have much lower K and Th contents interpretation fell out of favor because it does not explain than does the Martian surface. All but two SNC meteorites the lack of correlation between indices of differentiation (shergottite Los Angeles and NWA 817) have (e.g., Mg/Fe) and the concentrations of incompatible trace lower K than any GRS data points, and about half have elements [e.g., Borg and Draper, 2003]. The alternative lower Th. (Los Angeles and NWA 817 do not appear to be interpretation is that trends reflect the chemical properties of contaminated by on Earth. Their Th/U ratios are the mantle source regions for the basaltic SNC meteorites.

5of16 E03S10 TAYLOR ET AL.: MARS COMPOSITION AND DIFFERENTIATION E03S10

Figure 4. K and Th concentrations in 5° Â 5° smoothed pixels from Mars Odyssey GRS data compared to basaltic (includes olivine phyric) SNC and other SNC meteorites. GRS data collected through March 2005. Uncertainty is shown for a typical data point. SNC data are from Lodders [1998, and references therein] and Meyer [2003, and references therein]. Note that all but two meteorites (basaltic shergottite Los Angeles and nahklite NWA 817) plot substantially below GRS data, implying that the meteorites are not representative of typical Martian surface materials.

End-member cases are (1) an enriched source with high and is similar to basaltic shergottites, it appears that most of the approximately chondritic La/Yb, relatively high oxidation crust formed from undepleted, possibly state (1.5 log units below the -fayalite- rock. This has implications for early differentiation, as [QFM] oxygen fugacity buffer), negative eNd, and high discussed below. initial 87Sr/86Sr, and (2) a depleted source with low La/Yb, reduced (3–4 log units below QFM), positive eNd, 6. Testing Models of Martian Bulk Composition and low Rb, leading to low initial 87Sr/86Sr. Basaltic SNC 6.1. Models for the Bulk Composition of Mars meteorites with intermediate properties formed by melting of sources intermediate to those two end-members. The [15] There are three prominent models for the bulk nakhlites appear to derive from yet a third source region composition of Mars, each derived by a different method. [Foley et al., 2005]. Isotopic data show conclusively that Dreibus and Wa¨nke [1984], Wa¨nke and Dreibus [1988, these distinctive source regions formed early in Martian 1994], and Longhi et al. [1992] estimate the bulk compo- history, approximately by 4.5 Ga [Borg et al., 2003; Foley et sition from element correlations in SNC meteorites, with the al., 2005], and only 20 Ma after -aluminum-rich assumption that refractory elements are present in chondritic inclusions [Harper et al., 1995; Lee and Halliday, 1997; abundances. This model is directly tied to Mars through Blichert-Toft et al., 1999; Kleine et al., 2002; Yin et al., element abundances in SNC meteorites. The Wa¨nke- 2002], perhaps only 12 Ma after CAI formation [Foley et Dreibus model predicts that K/Th in bulk Mars is 5450. al., 2005], and that the mantle has remained unmixed since As we show below, global K/Th is 5300, favoring the then. This implies no widespread homogenization of the Wa¨nke-Dreibus model. Martian mantle or recycling of the crust. [16] Ganapathy and Anders [1974] and Morgan and [14] GRS data place this story into a global context. Anders [1979] modeled Mars as a mixture of chondritic La/Yb correlates with K and Th in basaltic shergottites materials that had been modified by the same limited set of (Figure 5), so K and Th concentrations also correlate with processes that affected chondrites, such as variations in 87 86 condensation temperature and fractionation of metal from eNd, initial Sr/ Sr, and oxygen fugacity. The average crust on Mars contains 3300 ppm K and 0.6 ppm Th (see below silicate. They propose that there are three primary conden- for a complete discussion of crustal averages). Comparing sates from the solar nebula: a high-temperature, refractory- those to basaltic Martian meteorites (Figure 5) suggests that rich condensate; Fe-Ni metal; and magnesian . the average crust ought to have chondritic (or even slightly Morgan and Anders [1979] define a fourth component, super chondritic) La/Yb and be as oxidized as Los Angeles FeS and FeO, which they postulate formed by reaction of and Shergotty (about QFM-1.5). It would have formed from Fe metal with H2S and H2O, respectively. Morgan and undepleted mantle, and, given the isotopic data from the Anders point out that elements of similar volatility do not fractionate during nebular processes, allowing them to use meteorites, the mantle source would have formed very early 36 in Martian history. The lowest Th and K values (Figure 4) four ‘‘index’’ elements (U, Fe, K, and Tl or Ar) to measured by GRS, if representative of the igneous rocks of calculate the abundances of 83 elements in the planet. They the crust (see section 4), suggest that a portion of the crust defined the concentrations of the index elements in Mars, formed from somewhat depleted mantle, but no large areas using gamma ray data from the Soviet orbiter Mars 5, (larger than the 500-km GRS footprint) are composed of thermal models available at the time, the density of the depleted basalts (those with 500 ppm K or less). If the crust mantle and the Martian moment of inertia, and volatile elements present in the atmosphere. Morgan and Anders

6of16 E03S10 TAYLOR ET AL.: MARS COMPOSITION AND DIFFERENTIATION E03S10

Figure 5. La/Yb variations versus K and Th in basaltic shergottites (SNC Martian meteorites). There is a clear trend of increasing K and Th with La/Yb. Chondritic La/Yb (1.5) is shown for comparison. The meteorite data suggest that if basaltic rock dominates the surface of Mars, the average Martian crust (as measured by GRS) has an approximately chondritic REE pattern, though it could be slightly enriched in light REE. Meteorite data sources are from Lodders [1998, and references therein] and Meyer [2003, and references therein].

[1979] predicted a value of 620 for K/Th in bulk Mars, 6.2. Regions Investigated and Their Compositions much lower than our surface value of 5300. [19] We have divided Mars into numerous regions to [17] Third, Lodders and Fegley [1997] focused on fitting investigate the relation of composition and geological the oxygen isotopic composition of Mars, known from SNC province or surface properties. (For more information on meteorites, to mixtures of chondritic meteorites. Their geologic province designations, see Dohm et al. [2005].) calculations led to the estimate that Mars was constructed The regions are usually large, so total counting times are from a mixture of 85% H-chondrites, 11% CV-chondrites, large and analytical uncertainties in concentrations are and 4% CI-chondrites. Lodders and Fegley [1997] predicted small. To understand the global composition of Mars, we a value of 16,000 for K/Th in bulk Mars, much higher than use K, Th, and Fe data from the following regions. our data indicate. (Sanloup et al. [1999] took a similar [20] 1. The entire globe includes north and south polar approach in estimating the composition of Mars but did not regions, which are extremely rich in H (H2O), hence tend to estimate the abundances of K and Th.) dilute the measured concentrations of other elements, but is [18] In testing these models, we will focus on K/Th, the still useful for assessing the K/Th ratio. A modification of ratio of a moderately volatile element to a refractory this is to include the entire area within the H mask. The element, and FeO, which varies among the terrestrial results reported in Table 1 are global for K and Th, and planets and among chondrite groups. We first summarize within the H masked area for FeO. our data for a diverse set of geologic regions on Mars, and [21] 2. The ancient southern highlands are regions of the present a summary of bulk K/Th and FeO in Mars. oldest crust, identified as places where magnetic anomalies

7of16 E03S10 TAYLOR ET AL.: MARS COMPOSITION AND DIFFERENTIATION E03S10

Table 1. K, Th, K/Th, and FeO in Selected Regions on Mars Region Ka, ppm Tha, ppm K/Thb FeOa Global 3300 ± 10 0.62 ± 0.02 5330 ± 220 18.4 ± 0.1 Ancient Southern Highlands 3630 ± 20 0.70 ± 0.02 5220 ± 180 17.6 ± 0.2 Arabia Terra 3540 ± 20 0.72 ± 0.03 4900 ± 220 18.0 ± 0.3 Major volcanic provinces 3010 ± 20 0.55 ± 0.02 5510 ± 230 18.4 ± 0.1 Northern Plains 3860 ± 20 0.71 ± 0.03 5340 ± 250 20.1 ± 0.2 Rocky regions 3630 ± 22 0.69 ± 0.03 5250 ± 210 18.1 ± 0.2 Dusty regions 3290 ± 20 0.64 ± 0.03 5130 ± 210 18.2 ± 0.2 aUncertainty represents 1-s counting statistics for global summed spectrum (total of 2.4 Â 107 s counting time for global average; >3 Â 106 s for other regions). b 2 2 1/2 Uncertainty calculated from (K/Th) [(sK/K) +(sTh/Th) ] , where K and Th are mean concentration of K and Th (columns 2 and 3). are visible. These highlands represent the oldest crust on 0.71 ppm. Most important, the K/Th ratio is quite uniform, Mars. ranging from 4900 to 5500. Dusty and rocky regions are [22] 3. Arabia Terra is a proposed ancient impact basin also indistinguishable in K/Th, suggesting that the presence and possible site of extensive sedimentary deposition. The of dusty materials does not affect our assessment of sediments would have been derived from the older high- the average K/Th on the surface. The ancient highlands lands, and might represent an average composition of those have K/Th essentially the same as the global averages. The highlands. K/Th ratio of the Martian surface and, we infer, for bulk [23] 4. Major volcanic provinces include Tharsis, Elysium, silicate Mars, is clearly between 5000 and 5500. We adopt and Hadriaca/Tyrrhena. Basaltic lavas are probes of mantle the global value of 5300 to represent the Martian K/Th ratio. composition, so the K/Th and Fe in them reflect the mantle [29] Our results are very different from those reported for composition at least since the Middle Noachian for Tharsis the 2 mission [Surkov et al., 1994], which operated [Dohm et al., 2001], for Elysium [e.g., Tanaka for only 12 days. K averages about 4000 ppm in the Phobos et al., 1992], and Late Noachian for Hadriaca/Tyrrhena 2 data set, compared to our average of 3300 ppm (Table 1). [Crown et al., 1992]. Th is much higher in the Phobos 2 data, 2.3 ppm versus our [24] 5 Northern Plains include the entire northern plains, crustal average of 0.6 ppm. This leads to a much lower K/ including the Formation. This deposit Th of 1740 in the Phobos 2 data, compared to our value of makes an interesting contrast to ancient highlands deposits 5300. However, as discussed by Trombka et al. [1992] the and allows an assessment of the extent to which aqueous peaks for Th in the Phobos 2 data are barely above processes might have affected K/Th. background, hence have large statistical uncertainties. [25] 6. Rocky regions combine the places on Mars with [30] In Table 2, we compare our measured K/Th for Mars highest rock abundance, determined on the basis of thermal to those of the three Martian bulk compositions described inertia, TES , and rock abundance models. above, and to the Earth, Vesta, the Moon, and CI carbona- [26] 7. Dusty regions are the dustiest regions on Mars, as ceous chondrites. Our inferred bulk K/Th is very close to quantitatively defined by thermal inertia and albedo. Dusty that calculated in the Wa¨nke-Dreibus model. In contrast, the regions might represent homogenized surface material, model reported by Ganapathy and Anders [1974] and hence may give a good estimate of the global abundance Morgan and Anders [1979] has a much lower K/Th (only of K, Th, and Fe. This is supported by the similarity of 620), which is inconsistent with our value of 5300. How- elemental compositions of soils at the Pathfinder and Viking ever, this does not negate the Ganapathy and Anders [1974] landing sites [e.g., Wa¨nke et al., 2001]. It also follows a long approach used for estimating planetary bulk compositions. tradition of using sediments to determine the composition of Updated compositional data for Martian surface might allow the terrestrial crust [McLennan et al., 1980]. However, we a more reasonable estimate using their approach. do not know the origin of the dusty material, so it might [31] The oxygen model developed by Lodders provide us with the composition of late stage volcanics and and Fegley [1997] requires that K/Th be much higher than altered materials, rather than an average of crustal materials. we find. Lodders and Fegley [1997] recognized that the abundance of alkalis was higher in their model than in the 6.3. K/Th in Mars Wa¨nke-Dreibus model and in Martian meteorites. They [27] The variation of K/Th on Mars is shown in Figures 1, suggested that aqueous leaching in the mantle and hydro- 4, and 6. Most of the Martian surface has K/Th between thermal activity led to the preferential deposition of K and about 4500 and 6500, with a sharp peak between 5000 and other alkalis, as well as halogens, in the Martian crust. 6000, although the total range is from about 3200 to 9000. Several features of the GRS data argue against this inter- Note that there is a tail off to high K/Th. This may reflect pretation. First, the process of alkali and halogen transport fractionation of K from Th, perhaps by aqueous processes. and crustal enrichment must operate surprisingly uniformly [28] K and Th concentrations and the K/Th ratios for the because K does not vary by more than about 50% (relative) regions are given in Table 1. Uncertainties are based solely across the surface (Figures 1 and 4). Second, and more on counting statistics, and so represent our confidence in the importantly, the K/Th ratio does not vary greatly (Figure 6), reported average value, not the range of values measured so if K is enriched by aqueous transport, it is enriched within each region. The range in the mean values is not everywhere on the surface by roughly the same amount. A large. K ranges from 3000 to 3600 ppm, Th from 0.55 to third point comes from mass balances. We estimate in

8of16 E03S10 TAYLOR ET AL.: MARS COMPOSITION AND DIFFERENTIATION E03S10

Figure 6. Distribution of K/Th values on Mars. Data are for 5° smoothed bins between 75° south and 75° north latitude, using data through March 2005. There is a clear peak in the distribution at about 5500. section 6 the percentage of bulk planetary K and Th residing than are the northern plains, and the global average, sum of in the crust, using the Wa¨nke-Dreibus K and Th concen- all rocky and dusty regions, and Arabia are intermediate, trations for the primitive mantle. The percentage is the same although not very different from the southern highlands. for both elements, suggesting no special enrichment process The global average (using the H mask) is 14.3 wt% Fe, for K. Finally, K does not correlate with Cl on the Martian corresponding to 18.4 wt% FeO. The GRS data are very surface [Keller et al., 2006]. If both elements were trans- slightly offset to higher values than those for SNC meteor- ported from the mantle together, one would expect that they ites and surface samples (Figure 7), but the difference is not would correlate, although subsequent aqueous processes significant. Most importantly, our GRS analytical points are might fractionate them. Lodders [2000] suggests that the almost everywhere higher than common terrestrial basalts. lower volatile content of Mars derived from compositions of Mid-ocean ridge basalts, for example, average 10.5 wt% SNC meteorites is only apparent. Higher might be FeO [Melson et al., 1976]. Thus the GRS results appear to in an enriched reservoir not sampled among the SNC confirm the inference from meteorites that Mars is enriched meteorites. This seems unlikely in light of our global data. in FeO compared to Earth. The Wa¨nke and Dreibus [1988, The GRS data set shows that the entire crust has K/Th much 1994] estimate of 17.9 wt % is close to our crustal average lower than the oxygen isotope mixing model predicts. [32] We conclude that the Wa¨nke-Dreibus model is a reasonable estimate of the K/Th concentration in bulk silicate Mars, and we adopt a K/Th value of 5300 for Mars. Table 2. Comparison of K, Th, and K/Th in Mars, Earth, Vesta, This is almost twice as high as reported for Earth (Table 2), the Moon, and Carbonaceous Chondrites confirming that Mars is enriched in moderately volatile elements compared to Earth. It is also clearly enriched K, ppm Th, ppm K/Th Mars (Average Crust)a 3300 0.62 5300 compared to the Moon and the Howardite-Eucrite-Diogenite b (HED) parent body (assumed to be ). Mars Bulk Silicate-WD 305 0.056 5450 Mars Bulk Silicate-Lc 920 0.056 16,400 Although Mars is enriched compared to those bodies, it is Mars Bulk Silicate-MAd 77 0.125 620 still substantially depleted compared to CI chondrites. Earth Continental Cruste 11,000 4.2 2600 Earth Bulk Silicatef 230 0.079 2900 6.4. FeO in Mars Vesta (HED meteorites)g 420 0.40 1050 Moon (high-K KREEP)h 8000 22 360 [33] Our 5° Â 5° Fe data are shown in Figure 7. We have CI Chondritesi 550 0.029 19,000 converted the data from Fe to FeO for easier comparison a with data from Martian meteorites and the Viking, This work. bDreibus and Wa¨nke [1984] and Wa¨nke and Dreibus [1988, 1994]. Pathfinder, and MER landed missions. We present FeO data cLodders and Fegley [1997]. for specific regions in Table 1. There is a clear peak between dMorgan and Anders [1979]. FeO concentrations of 18 and 21 wt%. Higher values for Fe eTaylor and McLennan [1985]. f are confined to the northern lowlands and lower values Average of Taylor and McLennan [1985], Jagoutz et al. [1979], and McDonough and Sun [1995]. occur in the southern highlands (Figure 3). This can also be gKitts and Lodders [1998] and Mittlefehldt and Lindstrom [1993]. seen in the averages for the regions we analyzed for this hWarren and Wasson [1979] and Warren [1989]. paper (Table 1). The ancient highlands are lower in FeO iMcDonough and Sun [1995].

9of16 E03S10 TAYLOR ET AL.: MARS COMPOSITION AND DIFFERENTIATION E03S10

below the QFM buffer [Minitti and Rutherford, 2000]. (In comparison, Minitti and Rutherford [2000] show that MgO decreases faster, resulting in a substantial increase in FeO/MgO.) Thus any magma generated by of a wet Martian mantle and subsequent fractionation of that hydrous magma would lead to lower FeO than in the original mantle . Nekvasil et al. [2003] used experimental data to estimate the type of magma that was parental to the . They sug- gested that the magma was alkaline and hydrous, and showed that fractionation of such a magma leads to a decrease in FeO with increasing crystallization and SiO2. This interpretation suggests that, if anything, the interior of Mars is richer in FeO than the surface. Hydrous partial melting might be part of the cause of lower FeO in the highlands compared to the northern plains. Thus the differ- ence in FeO might reflect lower H2O in the mantle source that melted to produce the highlands lavas and sediments derived from them that compose the northern plains. [35] Surface and near-surface aqueous processes might have caused transport of Fe from the highlands to the lowlands, thus accounting for the dichotomy in Fe concen- trations (Figure 3), a possibility raised by Boynton et al. (submitted manuscript, 2006). This would be particularly likely at the low pH conditions that have prevailed on Mars [e.g., Burns, 1993; Hurowitz et al., 2005]. However, one would expect that the solutions that delivered large amounts of Fe to the northern plains would be brines, hence Cl ought to be enriched in the northern plains, which we do not observe [Keller et al., 2006]. Figure 7. (a) GRS 5° Â 5° smoothed data set for Fe (converted to FeO), which shows a broad peak between 18 6.5. Implications for Planetary Accretion and 21 wt %. (b) Data for Martian meteorites, Viking, [36] There is a clear difference in the abundance of Pathfinder, and Spirit landing sites, which show a similar moderately volatile elements in Mars compared to Earth, result. The data are consistent with Mars having signifi- as shown by K/Th being double that in Earth (Table 2). This cantly more FeO in its mantle than Earth does. Data sources implies that on average Mars accreted from materials that are as follows: SNC meteorites from Lodders [1998, and were richer in moderately volatile elements than the average references therein] and Meyer [2003, and references there- of the materials that formed the Earth. In turn, this suggests in]; Viking from et al. [1982]; Pathfinder from that mixing among planetesimals was not extensive during Bru¨ckner et al. [2003]; Spirit data from Gellert et al. [2004]; the formation of the two planets, as argued on different and Opportunity data from Rieder et al. [2004]. For the grounds by Drake and Righter [2000]. They show that the MER sites, only samples abraded with the Rock Abrasion Earth’s composition is unique: It is different from known Tool are included for rock analyses. chondrites and other planets, for example, in Mg/Si and oxygen isotopic compositions. They conclude that there was not widespread mixing of material throughout the inner and consistent with the Martian moment of inertia [Bertka Solar System during accretion and that accretion zones were and Fei, 1998a, 1998b]. The geophysical data do, however, relatively narrow. This conclusion contrasts with calcula- allow the value to be slightly lower (12–14 wt%), as tions of the dynamics of planet formation, which suggest suggested by high-pressure experiments on chondritic mete- that there was considerable mixing of planetary embryos orites by Agee and Draper [2004] and supported by during the final stage of accretion [Wetherill, 1994; [Borg and Draper, 2003; Draper et Wetherill and Stewart, 1993; Chambers, 2001]. al., 2005]. [37] The Earth-Mars difference in abundances of moder- [34] Petrologic processes might have affected the parti- ately volatile elements does not necessarily imply a simple tioning of FeO into the crust. Its partitioning depends on variation with heliocentric distance. Vesta, often viewed as a oxidation state, water content, and pressure. For example, small , resides in the inner yet hydrous partial melting tends to partition less FeO into melt has much lower K/Th than does Mars, accompanied by than is in the original peridotite that was melting [Gaetani about the same concentration of FeO, 20 wt% [Warren, and Grove, 1998]. This effect is controlled by the temper- 1997]. Venus has roughly the same K/Th as does Earth, but ature of melting rather than by the FeO content (in reality the measurements have high uncertainty and substantial the activity of FeO) in the peridotite. Fractional crystalliza- variation [Surkov et al., 1987]. K/Th is unknown for tion of basaltic magma containing 1–1.5 wt% water causes , and models for its composition vary widely a decrease in FeO with crystallization at oxygen slightly [Taylor and Scott, 2004]; its K/Th will be determined by

10 of 16 E03S10 TAYLOR ET AL.: MARS COMPOSITION AND DIFFERENTIATION E03S10 the MESSENGER mission. The Moon is extremely low in to slightly more than 10 ppm, a factor of 100. K actually volatiles (K/Th < 400, Table 2). This reflects either its varies an equal amount but its low concentration makes it formation by a giant impact or that the giant impactor was less obvious. This reflects the formation of severely depleted in moderately volatile elements. cumulates low in Th and K, and the concentration of [38] The much higher FeO abundance in Mars (18 wt%) residual, KREEP that was available for excavation shows that the planet is more oxidized than Earth (8 wt%). by large impacts and for incorporation into magmas. In Robinson and Taylor [2001] argue that Venus has roughly contrast, Th on the Martian surface ranges from 0.2 to the same FeO as Earth and that Mercury has only 2–3 wt%. 1 ppm, a factor of only 5. K varies from roughly 2000 to The trend even includes Vesta in this case, although it would 6000, a factor of only 3. It is possible that the pervasive appear to be no more oxidized than Mars. The higher FeO presence of dust prevents us from measuring the true in Mars may have been caused by oxidation of metallic minimum values of crustal rocks. The SNC meteorite data by H2O during accretion of Mars, as hypothesized by (Figure 4) certainly show that some rocks have lower K than Wa¨nke and Dreibus [1994]. As discussed by Bertka and our global measurements. Fei [1998b], an oxidation event like this would increase the [42] It is possible that the striking difference between the content of a metallic core. Whatever the cause of the lunar and Martian GRS data is due to the higher resolution variation of FeO in the inner solar system, comparison of of the lunar data. To test this, we summed the lunar data to the compositions of Mars and Vesta indicates that there is have the same spatial resolution as do our 5° Â 5° Odyssey not a simple correlation of FeO and K/Th with heliocentric GRS data (also shown in Figure 8). In this case the highest distance: both bodies have high FeO, but very different Th point in the lunar data is still over 8 ppm and there are K/Th (Table 2). several points in the range 4–8 ppm, showing that there are large areas of the Moon enriched in Th. There is therefore 7. Style of Differentiation no evidence from the GRS data for large occurrences of primary cumulates on Mars or for extensively fractionated, [39] Taylor [1992] has defined three types of planetary KREEP-like materials. crusts. Primary crust forms as the result of planet-wide, [43] Does this mean that there was not a magma ocean on extensive early melting; i.e., from a magma ocean. Second- Mars? Not necessarily. It may mean that the processes in ary crust forms by partial melting of the interior to produce the Martian magma ocean were so different from those in basaltic rocks. This has clearly happened on all the terres- the lunar magma ocean that surface cumulates and highly trial planets and the Moon; the lava plains and large evolved melts were either not produced or were not volcanoes on Mars are clear evidence for the formation of segregated from accompanying cumulus . Theoret- secondary crust on Mars. Secondary crusts are basaltic. ical considerations and detailed modeling [Hess, 2002; Tertiary crust forms by the partial melting and differentia- Elkins-Tanton et al., 2003, 2005; Borg and Draper, 2003] tion of secondary crustal materials and sediments derived indicate that Al will be sequestered in the mantle, depleting from them. The presence of tertiary crust implies recycling the residual magma ocean in Al, and leading to delayed of the crust. Taylor [1992] notes that the terrestrial con- nucleation. Plagioclase, even when it finally tinents may be the only example of tertiary crust. Global does crystallize, might have a higher density than the GRS data suggest that Mars is dominated by secondary magma (especially if H2O is present), and hence sink [Hess, crust, though it does not rule out formation of a primary 2002]. This effectively rules out creation of a plagioclase- crust from a global magma system. If present, tertiary crust rich crust as on the Moon. No other minerals are good is rare. candidates to form a floatation crust on Mars, so we do not 7.1. Did Mars Have A Magma Ocean (Primary Crust)? expect to see regions with very low Th and K as we do in the lunar highlands. In comparing the Moon and Mars we [40] It is likely that the Moon was surrounded by an should keep in mind that we do not yet understand the ocean of magma when it formed; see reviews by Warren complex array of processes that operated in the lunar magma [1985, 2003]. This event produced the original (primary), ocean, the fate of its products, and their final locations in the anorthositic crust in the lunar highlands, a residual magma differentiated Moon. For example, Longhi [2003] proposed rich in , rare earth elements, and that there was a magma ocean, but that the -rich (KREEP) that seems to have concentrated in the Imbrium- crust did not form directly from it. Procellarum region of the nearside [Jolliff et al., 2000], and [44] The limited concentration ranges of K and Th on a density-unstable mantle whose overturn led to melting and Mars could mean that there was less fractional crystalliza- formation of the Moon’s secondary crust. To a great extent, tion and less crystal-melt segregation in a Martian magma the lunar magma ocean set the course for subsequent lunar ocean than in the lunar one. Elkins-Tanton and Parmentier magmatic evolution. Thus it is essential that we determine if [2004] point out that even in the lunar magma ocean most of Mars also had a magma ocean. K and Th concentrations the magma lies between the liquidus and solidus, hence will show a much smaller range on Mars than they do on the have a substantial crystal fraction. At crystal fractions above Moon, suggesting that Mars did not have a magma ocean, a critical value the melt-crystal system begins to behave as a that processes operated differently in it than on the Moon, or solid, inhibiting fractionation. This happens at crystal frac- that the record of it has been erased. tions of about 50% [Van der Molen and Paterson, 1979; [41] We compare global GRS data for the Moon and Mars and Okamura, 1977; Campbell et al., 1978; Marsh, in Figure 8. The data clearly show the large depletion in K 1988], although it can be as low as 25% if networks of lath- in the Moon compared to Mars. Equally striking is the very shaped plagioclase crystals form [Philpotts et al., 1998]. large range in Th on the Moon. It ranges from about 0.1 ppm This important effect led Elkins-Tanton and Parmentier

11 of 16 E03S10 TAYLOR ET AL.: MARS COMPOSITION AND DIFFERENTIATION E03S10

Figure 8. Comparison of GRS data for Mars (our data) and the Moon [Prettyman et al., 2002]. Lunar data are shown for both 60-km resolution elements and in degraded resolution equal to 10° Â 10° resolution elements on Mars (about 500 km). K/Th lines are for reference only, not fits to the data. The Moon clearly has much lower K/Th than Mars does. Nevertheless, it shows a much larger range in K and Th than GRS measurements of the Martian surface do. The lunar range (shown most readily by the Th data) is caused by the presence of large areas containing low-Th rocks (the lunar highlands, composed of cumulate ) and very high Th (highly differentiated rocks called KREEP). Comparison of the high- and low-resolution lunar data shows that the high Th concentrations are not in small, localized locations. These differences in the distribution of trace elements may reflect very different differentiation histories for the two bodies.

[2004] to propose that even the lunar magma ocean did not 7.2. Partitioning of K and Th Into the Crust directly produce the anorthosite crust on the Moon, suggest- [45] An important consideration in thermal evolution is ing it formed shortly after crystallization as overturn occurs how much of the planet’s bulk K, Th, and U was partitioned to remove strong density gradients. Longhi [2003] reaches into the crust versus remaining in the mantle. We show the same conclusion from the perspective of phase equilib- below that about half of the K and Th are fractionated into ria in a lunar magma ocean. By analogy, the Martian magma the crust, leaving enough in the mantle to allow for the ocean may not have segregated crystals and liquid efficiently young ages of basaltic SNC meteorites. Although the extent enough to produce a distinctive layer of KREEP-like of melting during initial Martian differentiation is uncertain, residual melt. The residual melt instead may have crystal- isotopic data from SNC meteorites show unambiguously lized largely in place, perhaps in large part by local that Mars differentiated early. Their apparent depletion in Al equilibrium crystallization. Overturn of the mantle shortly is consistent with a melting event that involved depths of at after the original Martian magma ocean had crystallized or least hundreds of kilometers so that garnet crystallized. This almost crystallized, could have led to a variety of regions led to an early partitioning of K, Th, and U into the crust, containing different amounts of depleted cumulates and which affected the subsequent magmatic evolution of Mars trapped , variable amounts of water (possibly leading [McLennan, 2001; Hauck and Phillips, 2002; Kiefer, 2003]. to variations in oxygen fugacity in the source regions), and [46] The average Th in the crust is 0.6 ppm (Table 2). The an initial crust composed of the lowest density materials. thickness of the crust has been estimated from geophysical Calculations by Elkins-Tanton et al. [2003] suggest that this data. Estimates place limits on the crustal thickness, with original crust would contain about 45 wt% SiO2, low Al2O3, the broad limits of 29 to 115 km (summarized by Wieczorek and only 5–10 wt% FeO. Subsequent magma production and Zuber [2004]). A reasonable range is 33–81 km would have led to higher FeO contents. Elkins-Tanton et al. [Wieczorek and Zuber, 2004], with a nominal value of [2003] predict that FeO increased with depth in Mars after 57 km. Assuming that the bulk Th in Mars is 0.056 ppm À3 overturn, but the upper mantle would still have low Al2O3. (Table 2), crustal density of 2.9 g cm , mantle density of In more detailed calculations, Elkins-Tanton et al. [2005] 3.5 g cmÀ3, and a core radius of 1634 km, then we can suggest that magmas produced by partial melting of over- calculate the percentage of Th partitioned into the crust as a turned magma ocean cumulates would produce magmas function of crustal thickness (Figure 9). The maximum with FeO contents of 11–22 wt.%, in the range observed for thickness of the crust is 118 km, calculated by assuming the Martian surface. For a given percentage of melting that all the Th is in the crust. (inferred from MgO content of the magma), deep partial [47] For the nominal crustal thickness of 57 km, 50% of melting produces magma with higher FeO than shallow the Th is in the crust. If the thickness of the crust is closer to melting. Perhaps the slightly lower FeO content of the the upper limit of 81 km, then 70% of the Th is in the crust. southern highlands compared to the northern plains is the A calculation for K concentrations gives the same answers. result of southern highland basalts being derived from, on These are high values: the concentration of heat-producing average, shallower depths than those composing the north- elements in the terrestrial crust is <40% [McLennan, 2001], ern plains. although the terrestrial crust composes only about 1% of the

12 of 16 E03S10 TAYLOR ET AL.: MARS COMPOSITION AND DIFFERENTIATION E03S10

Figure 9. Calculation of the percentage of Th (hence inferred REE) in the crust of Mars, using assumptions listed in the top left. The maximum crustal thickness is given by the thickness at which all the Th would be in the crust. If the crust is 57 km thick, as inferred from geophysical studies [e.g., Wieczorek and Zuber, 2004], then 50% of the bulk Th is in the crust. Dashed arrow indicates the uncertainty in crustal thickness determined by Wieczorek and Zuber (±24 km). Taking the uncertainties into account, the percentage of bulk Th that is in the crust is in the range 29% to 70%. silicate Earth (the Martian crust makes up about 5% of crust. If so, the processes that produced the dust and other silicate Mars). The consequence of sequestering such a large components did not fractionate K from Th fraction of the heat-producing elements into the crust and (Figure 2) significantly. This problem is discussed in detail putting them there early, is that the rate of crustal production by H. E. Newsom et al. ( of and becomes very small after about 4 Ga [Hauck and Phillips, bedrock in mantled and less mantled terrains with gamma 2002], assuming mantle rheology consistent with the pres- ray data from Mars Odyssey, submitted to Journal of ence of water. Kiefer [2003] calculated that the present Geophysical Research, 2006). volcanic eruption rate is consistent with retaining 43–50% of the radioactive elements in the mantle, in reasonable 7.3. Is There Evidence for Tertiary Crust on Mars? agreement with our estimate assuming a crust 57-km thick. [49] GRS data do not provide compelling evidence for the Norman [1999] calculated the thickness of the crust from a formation of a tertiary crust involving extensive remelting mass balance model based on the rare earth abundances and of the original basaltic crust to yield more evolved magma. Nd isotopic compositions of Martian meteorites. He esti- As shown in Figure 10, terrestrial rocks have a large range mated that the thickness was <45 km, with the most likely in K and Th (Figure 10 is presents averages of suites of value being 20–30 km. He emphasized that on the basis of rocks and so does not show the full range). This is a isotopic data that this crust must have formed early. An consequence of melting of hydrous subducting slabs, melt- initial crust 20–30 km thick may have formed at 4.5 Ga ing of the overlying the slab, and remelting of with significant additions in the next few million years and both igneous and sedimentary rocks of the . only modest subsequent additions. This is consistent with It appears from the very restricted range of K and Th on Martian meteoritic isotopic data, the geologic record placing Mars that formation of tertiary crust was not a widespread most volcanic activity before the end of the Hesperian phenomenon. It might, however, have occurred locally: [Tanaka et al., 1992], and thermal modeling by Hauck Bandfield et al. [2004] reported patches of materi- and Phillips [2002]. Norman [2002] revised his mass als and Christensen et al. [2005] describe in Syrtis balance model by changing the initial Nd isotopic compo- Major. Our data do not indicate widespread evolved rocks, sition for the primitive mantle. This resulted in a thicker suggesting that the northern plains (surface type 2 of crust (85 ± 10 km). He concluded that the crust formed by a Bandfield et al. [2000]) are not . This is discussed multistage process in which a very early crust was 20– in detail by Karunatillake et al. [2006]. 30 km thick and enriched in light REE elements. Subse- quent intrusion and extrusion of magma (light REE depleted 8. Conclusions and undepleted) thickened the crust. [50] The global data we report here lead us to the [48] One caveat to the estimate of K and Th in the crust is the possibility that they are enriched in the mobile dust following conclusions. component, so the uppermost surface does not reflect the [51] 1. Bulk Mars is enriched in moderately volatile concentrations of K and Th in igneous rocks or the bulk elements compared to Earth, but has a much lower K/Th

13 of 16 E03S10 TAYLOR ET AL.: MARS COMPOSITION AND DIFFERENTIATION E03S10

Figure 10. K versus Th for major types of terrestrial rocks compared to GRS data for the Martian surface. The terrestrial crust has a much larger range in composition than the crust of Mars does (at least as sampled by the Odyssey GRS). This difference is caused by the formation of substantial amounts of tertiary crust on Earth. Data sources are as follows: Average igneous and sedimentary rocks were compiled by Condie [1993]; additional oceanic sediments are from Elliott et al. [1997], Gromet et al. [1984], Vroon et al. [1995], Gaillardet et al. [1999], Plank and Langmuir [1998]; MORB and Ocean Island averages are from Sun and McDonough [1989]; Earth bulk crust and continental crust are from Taylor and McLennan [1985]; arc magmas are from Grove et al. [2002], Elliott et al. [1997], and Turner et al. [1996]. ratio than CI chondrites or models based on matching the Beattie, P. (1993), The generation of series disequilibria by partial melting of peridotite: Constraints from partitioning studies, Earth. oxygen isotopic composition of Mars with the oxygen Planet. Sci. Lett., 117, 379–391. isotopic compositions of mixtures of chondrites. Bertka, C. M., and Y. Fei (1998a), Density profile of an SNC model [52] 2. Mars is enriched in FeO compared to Earth. Martian interior and the moment-of-inertia factor of Mars, Earth Planet. Sci. Lett., 157, 79–88. [53] 3. Conclusions 1 and 2 suggest that terrestrial planets Bertka, C. M., and Y. Fei (1998b), Implications of data for formed from relatively narrow accretion zones. the accretion history of the terrestrial planets, Science, 281, 1838–1840. [54] 4. Surface concentrations of K and Th measured by Blichert-Toft, J., J. D. Gleason, P. Telouk, and F. Albarede (1999), The GRS are systematically higher than in SNC meteorites. The Lu-Hf of shergottites and the evolution of the Martian mantle-crust system, Earth. Planet. Sci. Lett., 173, 25–39. meteorites are not representative of Martian surface rocks. Borg, L. E., and D. S. Draper (2003), A petrogenetic model for the origin [55] 5. About half the planet’s inventory of heat-producing and compositional variation of the Martian basaltic meteorites, Meteorol. elements is in the crust (assuming crust is 57-km thick). Planet. Sci., 38, 1713–1731. Borg, L. E., L. E. Nyquist, L. A. Taylor, H. Wiesmann, and C.-Y. Shih 56 [ ] 6. Evidence for a lunar-like magma ocean on Mars is (1997), Constraints on Martian differentiation processes from Rb-Sr lacking in the GRS data set, in spite of meteorite data and Sm-Nd isotopic analyses of the basaltic shergottites QUE 94201, showing extensive, early differentiation. Geochim. Cosmochim. Acta, 61, 4915–4931. [57] 7. GRS data for K and Th do not seem consistent Borg, L. E., L. E. Nyquist, H. Weissman, C.-Y. Shih, and Y. Reese (2003), The age of Dar al Gani 476 and the differentiation history of the Martian with widespread recycling of the crust, long-acting plate meteorites inferred from their radiogenic isotopic systematics, Geochim. , or extensive formation of a tertiary crust on Mars Cosmochim. Acta, 67, 3519–3536. (unless such evidence occurs systematically at smaller Boynton, W. V., et al. (2004), The Mars Odyssey gamma-ray spectrometer instrument suite, Space Sci. Rev., 10, 37–83. scales that the 500-km GRS resolution). Bridges, J., D. C. Catling, J. M. Saxton, T. D. Swindle, I. C. Lyon, and M. M. Grady (2001), Alteration assemblages in Martian meteorites: [58] Acknowledgments. We thank the Mars Odyssey Project for Implications for near-surface processes, Space Sci. Rev., 96, 365–392. financial support. We thank Jeff Gillis-Davis for converting Lunar Pros- Bru¨ckner, J., G. Dreibus, R. Rieder, and H. Wa¨nke (2003), Refined data of pector data to 500-km spatial resolution. This is SOEST Publication 6785 alpha proton X-ray spectrometer analyses of soils and rocks at the Mars and HIGP Publication 1441. Pathfinder site: Implications for surface chemistry, J. Geophys. Res., 108(E12), 8094, doi:10.1029/2003JE002060. Burns, R. (1993), Rates and mechanisms of chemical weathering of ferro- References magnesian silicate minerals on Mars, Geochim. Cosmochim. Acta, 57, Agee, C. B., and D. S. Draper (2004), Experimental constraints on the 4555–4574. origin of Martian meteorites and the composition of the Martian mantle, Campbell, I. H., P. L. Roeder, and J. M. Dixon (1978), Plagioclase buoy- Earth Planet. Sci. Lett., 224, 415–429. ancy in basaltic liquids as determined with a centrifuge furnace, Contrib. Bandfield, J. L., V. E. Hamilton, and P. R. Christensen (2000), A global . Petrol., 67, 369–377. view of Martian surface compositions from MGS-TES, Science, 287, Chambers, J. E. (2001), Making more terrestrial planets, Icarus, 152, 205– 1626–1630. 224. Bandfield, J. L., V. E. Hamilton, P. R. Christensen, and H. Y. McSween Jr. Christensen, P. R., et al. (2005), Evidence for magmatic evolution and (2004), Identification of quartzofeldspathic materials on Mars, J. Geo- diversity on Mars from infrared observations, Nature, 436, 504–509, phys. Res., 109, E10009, doi:10.1029/2004JE002290. doi:10.1038/nature03639.

14 of 16 E03S10 TAYLOR ET AL.: MARS COMPOSITION AND DIFFERENTIATION E03S10

Clark, B. E., A. K. Baird, R. J. Weldon, D. M. Tsusaki, L. Schnabel, and Hauck, S. A., and R. J. Phillips (2002), Thermal and crustal evolution of M. P. Candelaria (1982), Chemical composition of Martian fines, J. Geo- Mars, J. Geophys. Res., 107(E7), 5052, doi:10.1029/2001JE001801. phys. Res., 87, 10,059–10,067. Hauri, E. H., T. P. Wagner, and T. L. Grove (1994), Experimental and Condie, K. (1993), Chemical composition and evolution of the upper con- natural partitioning of Th, U, Pb, and other trace elements between gar- tinental crust: Contrasting results from surface samples and shales, Chem. net, clinopyroxene, and basaltic melts, Chem. Geol., 117, 149–166. Geol., 104, 1–37. Herd, C. D. K. (2003), The oxygen fugacity of olivine-phyric Martian Crisp, J. A. (1984), Rates of magma emplacement and volcanic output, basalts and the components within the mantle and crust of Mars, Meteor- J. Volcanol. Geotherm. Res., 20, 177–211. it. Planet. Sci., 38, 1793–1805. Crown, D. A., K. H. Price, and R. (1992), Geologic evolution of Herd, C. D. K., L. E. Borg, J. H. Jones, and J. J. Papike (2002), Oxygen the east rim of the Hellas basin, Mars, Icarus, 100, 1–25. fugacity and geochemical variations in the Martian basalts: Implications Dohm, J. M., J. C. Ferris, V. R. Baker, R. C. Anderson, T. M. Hare, R. G. for Martian basalt petrogenesis and the oxidation state of the upper man- Strom, N. G. Barlow, K. L. Tanaka, J. E. Klemaszewski, and D. H. Scott tle of Mars, Geochim. Cosmochim. Acta, 66, 2025–2036. (2001), Ancient drainage basin of the Tharsis region, Mars: Potential Hess, P. (2002), Origins of the Martian crust and mantle, paper presented at source for outflow channel systems and putative oceans or paleolakes, Workshop on Unmixing the SNCs: Chemical, Isotopic, and Petrologic J. Geophys. Res., 106, 32,943–32,958. Components of the Martian Meteorites, Lunar and Planet. Inst., Houston, Dohm, J. M., K. Kerry, J. Keller, V. R. Baker, W. Boynton, S. Maruyama, Tex. and R. C. Anderson (2005), Mars geological province designations for Hurowitz, J. A., N. J. Tosca, S. M. McLennan, and the Athena Science the interpretation of GRS data, Lunar Planet. Sci. Conf. [CD-ROM], Team (2005), Experimental basalt alteration at low-pH: Implications for XXXVI, Abstract 1567. weathering relationships on Mars, Proc. Lunar Planet. Sci. Conf., 36th, Drake, M. J., and K. Righter (2000), Determining the composition of the Abstract 2025. Earth, Nature, 416, 39–44. Jagoutz, E., H. Palme, H. Baddenhausen, K. Blum, M. Cendales, G. Dreibus, Draper, D. S., L. E. Borg, and C. B. Agee (2005), Crystallization of B. Spettel, V. Lorenz, and H. Wa¨nke (1979), The abundances of major, a Martian magma ocean and the formation of shergottites source regions: minor, and trace elements in the Earth’s mantle as derived from primi- A less Fe-rich Mars?, Lunar Planet. Sci., XXXVI, Abstract 1429. tive ultramafic nodules, Proc. Lunar Planet. Sci. Conf., 10th, 2031– Dreibus, G., and H. Wa¨nke (1984), Accretion of the Earth and the inner 2050. planets, Proc. Int. Geol. Conf., 27th(11), 1–20. Jolliff, B. L., J. L. Gillis, L. A. Haskin, R. L. Korotev, and M. A. Wieczorek Elkins-Tanton, L. T., and E. M. Parmentier (2004), Consequences of high (2000), Major lunar crustal terranes: Surface expressions and crust-man- crystallinity for the evolution of the lunar magma ocean: Trapped plagi- tle origins, J. Geophys. Res., 105, 4197–4216. oclase, Lunar Planet. Sci. Conf., XXXV, Abstract 1678. Jones, J. H. (1995), Experimental trace element partitioning, in A Handbook Elkins-Tanton, L. T., E. M. Parmentier, and P. C. Hess (2003), Magma of Physical Constants: Rock Physics and Phase Relations, pp. 73–104, ocean fractional crystallization and cumulate overturn in terrestrial pla- AGU, Washington, D. C. nets: Implications for Mars, Meteorit. Planet. Sci., 38, 1753–1771. Karunatillake, S. K., et al. (2006), Composition of northern low-albedo Elkins-Tanton, L. T., P. C. Hess, and E. M. Parmentier (2005), Possible regions of Mars: Insights from the Mars Odyssey gamma ray spectro- formation of ancient crust on Mars through magma ocean processes, meter, J. Geophys. Res., doi:10.1029/2006JE002675, in press. J. Geophys. Res., 110, E12S01, doi:10.1029/2005JE002480. Keller, J., et al. (2006), Equatorial and midlatitude distribution of Elliott, T., T. Plank, A. Zindler, W. White, and B. Bourdon (1997), Element measured by Mars Odyssey GRS, J. Geophys. Res., doi:10.1029/ transport from slab to volcanic front at the Mariana arc, J. Geophys. Res., 2006JE002679, in press. 102, 14,991–15,019. Kiefer, W. S. (2003), Melting in the Martian mantle: Shergottite formation Foley, C. N., M. Wadhwa, L. E. Borg, P. E. Janney, R. Hines, and T. L. and implications for present-day on Mars, Meteorit. Grove (2005), The early differentiation history of Mars from 182W-142Nd Planet. Sci., 38, 1815–1832. isotope systematics in the SNC meteorites, Geochim. Cosmochim. Acta, Kleine, T., C. Munker, K. Metzer, and H. Palme (2002), Rapid accretion 69, 4557–4571. and early core formation on and the terrestrial planets from Gaetani, G. A., and T. L. Grove (1998), The influence of water on Hf-W chronometry, Nature, 418, 952–955. melting of mantle peridotite, Contrib. Mineral. Petrol., 131, 323– Kitts, K., and K. Lodders (1998), Survey and evaluation of eucrite bulk 346. compositions, Meteorit. Planet. Sci., 33, A197–A213. Gaillardet, J., B. Dupre, and C. J. Allegre (1999), Geochemistry of large Lee, D.-C., and A. N. Halliday (1997), Core formation on Mars and differ- river suspended sediments: Silicate weathering or recycling tracer?, Geo- entiated asteroids, Nature, 388, 854–857. chim. Cosmochim. Acta., 63, 4037–4051. Lodders, K. (1998), A survey of shergottites, nahklite and chassigny me- Ganapathy, R., and E. Anders (1974), Bulk compositions of the Moon and teorites whole-rock compositions, Meteorit. Planet. Sci., 33,A183– Earth, estimated from meteorites, Proc. Lunar Sci. Conf., 5th, 1181– A190. 1206. Lodders, K. (2000), An oxygen isotope mixing model for the accretion and Gellert, R., et al. (2004), Chemistry of rocks and soils in Gusev crater composition of rocky planets, Space Sci. Rev., 92, 341–354. from the alpha particle X-ray spectrometer, Science, 305, 829–832. Lodders, K., and B. Fegley Jr. (1997), An oxygen isotope model for the Goodrich, C. A., C. D. K. Herd, and L. A. Taylor (2003), and composition of Mars, Icarus, 126, 373–394. oxygen fugacity in olivine-phyric and lherzolitic shergottites, Meteorit. Longhi, J. (2003), A new view of lunar ferroan anorthosites: Postmagma Planet. Sci., 38, 1773–1792. ocean petrogenesis, J. Geophys. Res., 108(E8), 5083, doi:10.1029/ Gromet, L. P., R. F. Dymek, L. A. Haskin, and R. L. Korotev (1984), 2002JE001941. The ‘‘North American Shale Composite’’: Its compilation, major and Longhi, J., E. Knittle, J. R. Holloway, and H. Wa¨nke (1992), The bulk trace element characteristics, Geochim. Cosmochim. Acta, 48,2469– composition, mineralogy, and internal structure of Mars, in Mars, edited 2482. by H. H. Kieffer et al., pp. 184–208, Univ. of Ariz. Press, Tucson. Grove, T. L., S. W. Parman, S. A. Bowring, R. C. Price, and M. B. Baker Marsh, B. D. (1988), Crystal capture, sorting, and retention in convecting (2002), The role of an H2O-rich fluid component in the generation magma, Bull. Geol. Soc. Am., 100, 1720–1737. of primitive basaltic and andesites from the Mt. Shasta region, McDonough, W. F., and S.-S. Sun (1995), The composition of the Earth, N. California, Contrib. Mineral. Petrol., 142, 375–396. Chem. Geol., 120, 223–253. Halliday, A. N., D.-C. Lee, S. Tommasini, G. R. , C. R. Paslick, J. G. McLennan, S. M. (2001), Crustal heat production and the thermal evolution Fitton, and D. E. James (1995), Incompatible trace elements in OIB and of Mars, Geophys. Res. Lett., 28, 4019–4022. MORB and source enrichment in the sub-oceanic mantle, Earth Planet. McLennan, S. M. (2003), Large- lithophile element fractionation during Sci. Lett., 133, 379–395. the early differentiation of Mars and the composition of the Martian Hamilton, V. E., P. R. Christensen, H. Y. McSween Jr., and J. L. Bandfield primitive mantle, Meteorit. Planet. Sci., 38, 895–904. (2003), Searching for the source regions of Martian meteorites using McLennan, S. M., W. B. Nance, and S. R. Taylor (1980), Rare earth ele- MGS TES: Integrating Martian meteorites into the global distribution ment– correlations in sedimentary rocks, and the composition of of volcanic materials on Mars, Meteorol. Planet. Sci., 38, 871–886. the continental crust, Geochim. Cosmochim. Acta, 44, 1833–1839. Harper, C. L., L. E. Nyquist, B. Bansal, H. Wiesmann, and C.-Y. Shih McSween, H. Y., Jr. (2003), Mars, in Treatise on Geochemistry, vol. 1, (1995), Rapid accretion and early differentiation of Mars indicated by Meteorites, Comets, and Planets, edited by A. Davis, pp. 601–621, Else- 142Nd/144Nd in SNC meteorites, Science, 267, 213–217. vier, New York. Hartmann, W. K., and G. Neukum (2001), Cratering chronology and the Melson, W. G., T. Vallier, T. L. Wright, G. Byerly, and J. Nelen (1976), evolution of Mars, Space Sci. Rev., 96, 165–194. Chemical diversity of abyssal erupted along Pacific, Atlan- Hartmann, W. K., J. Anguita, M. A. de la Casa, D. C. Berman, and E. V. tic, and Indian Ocean -floor spreading centers, in The Geophysics of Ryan (2001), Martian cratering 7: The role of impact gardening, Icarus, the Pacific Ocean Basin and Its Margin, Geophys. Monogr. Ser., vol. 19, 149, 37–53. pp. 351–367, AGU, Washington, D. C.

15 of 16 E03S10 TAYLOR ET AL.: MARS COMPOSITION AND DIFFERENTIATION E03S10

Meyer, C., Jr. (2003), Mars meteorite compendium, Rep.27672,rev.B, Turner, S., C. Hawkesworth, P. van Calsteren, E. Heath, R. Macdonald, and NASA Johnson Space Cent., Houston, Tex. (Available at http://curator. S. Black (1996), U-series and destructive plate margin magma jsc..gov/antmet/mmc/mmc.htm) genesis in the Lesser Antilles, Earth Planet. Sci. Lett., 142, 191–207. Minitti, M. E., and M. J. Rutherford (2000), Genesis of the Mars Pathfinder Van der Molen, I., and M. S. Paterson (1979), Experimental deformation of ‘‘sulfur-free’’ rock from SNC parental liquids, Geochim. Cosmochim. partially-melted , Contrib. Mineral. Petrol., 70, 299–318. Acta, 64, 2535–2547. Vroon, P. Z., M. J. van Bergen, G. J. Klaver, and W. M. White (1995), Mittlefehldt, D., and M. Lindstrom (1993), Geochemistry and of Strontium, neodymium, and lead isotopic and trace-element signatures of a suite of ten Yamato HED meteorites, Proc. Natl. Inst. Polar Res. Symp. the east Indonesian sediments: Provenance and implications for Banda Antarct. Meteorites, 6, 268–292. Arc magma genesis, Geochim. Cosmochim. Acta, 59, 2573–2598. Morgan, J. W., and E. Anders (1979), Chemical composition of Mars, Wa¨nke, H., and G. Dreibus (1988), Chemical composition and accretion Geochim. Cosmochim. Acta, 43, 1601–1610. history of the terrestrial planets, Philos. Trans. R. Soc. London, Ser. A, Nekvasil, H., J. Filiberto, M. Whitacker, and D. H. Lindsley (2003), Mag- 325, 545–557. mas parental to the Chassigny meteorite: New considerations, paper pre- Wa¨nke, H., and G. Dreibus (1994), Chemistry and accretion history of sented at Sixth International Conference on Mars, Lunar and Planet. Inst., Mars, Philos. Trans. R. Soc. London, Ser. A, 349, 285–293. Pasadena, Calif. Wa¨nke, H., J. Bru¨ckner, G. Dreibus, R. Rieder, and I. Ryabchikov (2001), Norman, M. D. (1999), The composition and thickness of the crust of Mars Chemical composition of rocks and soils at the Pathfinder site, Space Sci. estimated from rare earth elements and neodymium-isotopics of Martian Rev., 96, 317–330. meteorites, Geochim. Cosmochim. Acta, 34, 439–449. Warren, P. H. (1985), The magma ocean concept and lunar evolution, Annu. Norman, M. D. (2002), Thickness and composition of the Martian crust Rev. Earth Planet. Sci., 13, 201–240. revisited: Implications of an ultradepleted mantle with a Nd isotopic Warren, P. H. (1989), KREEP: Major element diversity, trace-element uni- composition like that of QUE94201, Lunar Planet. Sci., XXXIII, Abstract formity (almost), in Workshop on Moon in Transition: Apollo 14, KREEP, 1175. and Evolved Lunar Rocks, edited by G. J. Taylor and P. H. Warren, Tech. Philpotts, A. R., J. Shi, and C. Brustman (1998), Role of plagioclase crystal Rep. 89–03, pp. 149–153, Lunar and Planet. Inst., Houston, Tex. chains in the differentiation of partly crystallized basaltic magma, Nature, Warren, P. H. (1997), MgO-FeO mass balance constraints and a more 395, 343–346. detailed model for the relationship between eucrites and diogenites, Plank, T., and C. H. Langmuir (1998), The chemical composition of sub- Meteorit. Planet. Sci., 32, 945–963. ducting and its consequences for the crust and mantle, Chem. Warren, P. H. (2003), The Moon, in Treatise on Geochemistry,vol.1, Geol., 145, 325–394. Meteorites, Comets, and Planets, edited by A. Davis, pp. 559–599, Prettyman, T., W. C. Feldman, D. J. Lawrence, G. W. McKinney, A. B. Elsevier, New York. Binder, R. C. Elphic, O. M. Gasnault, S. Maurice, and K. R. Moore Warren, P. H., and J. T. Wasson (1979), The origin of KREEP, Rev. Geo- (2002), Library least squares analysis of Lunar Prospector gamma ray phys., 17, 73–88. spectra, Lunar Planet. Sci. [CD-ROM], XXXII, Abstract 2012. Wetherill, G. W. (1994), The provenance of the terrestrial planets, Geochim. Rieder, R., et al. (2004), Chemistry of rocks and soils at Meridiani Planum Cosmochim. Acta, 58, 4513–4520. from the alpha particle X-ray spectrometer, Science, 306, 1746–1749. Wetherill, G. W., and G. R. Stewart (1993), Formation of planetary em- Robinson, M. S., and G. J. Taylor (2001), Ferrous oxide in Mercury’s crust bryos: Effects of fragmentation, low relative velocity, and independent and mantle, Meteorit. Planet. Sci., 36, 841–847. variation of eccentricity and inclination, Icarus, 106, 190–209. Rubin, A. E., P. H. Warren, J. P. Greenwood, R. S. Verish, L. A. Leshin, Wieczorek, M. A., and M. T. Zuber (2004), The thickness of the Martian R. L. Hervig, R. N. Clayton, and T. K. Mayeda (2000), Petrology of Los crust: Improved constraints from geoid-to-topography ratios, J. Geophys. Angeles: A new basaltic shergottites find, Geology, 28, 1011–1014. Res., 109, E01009, doi:10.1029/2003JE002153. Sanloup, C., A. Jambon, and P. Gillet (1999), A simple chondritic model of Wright, T. L., and R. T. Okamura (1977), Cooling and crystallization of Mars, Phys. Earth Planet. Inter., 112, 43–54. tholeiitic basalt, 1965 Makaopuhi lava , Hawaii, U.S. Geol. Surv. Sautter, V., J. A. Barrat, A. Jambon, J. P. Lorand, P. Gillet, M. Javoy, J. L. Prof. Pap., 1004, 78 pp. Joron, and M. Lesourd (2002), A new Martian meteorite from Morocco: Yen, A. S., et al. (2005), An integrated view of the chemistry and miner- The nakhlite northwest Africa 817, Earth Planet. Sci. Lett., 195, 223– alogy of Martian soils, Nature, 436, 49–54. 238. Yin, Q., S. B. Jacobsen, K. Yamashita, J. Blichert-Toft, P. Telouk, and Squyres, S. W., et al. (2004), evidence for an ancient aqueous F. Albarede (2002), A short timescale for terrestrial planet formation from environment at Meridiani Planum, Mars, Science, 306, 1709–1714. Hf-W chronometry of meteorites, Nature, 418, 949–952. Sun, S.-S., and W. F. McDonough (1989), Chemical and isotopic systema- tics of oceanic basalts: Implications for mantle composition and pro- ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀ cesses, in in the Ocean Basins, edited by A. D. Saunders V. Baker, W. Boynton, J. Dohm, D. Hamara, D. Janes, J. Keller, K. Kerry, and M. J. Norry, Spec. Publ. Geol. Soc. Am., 42, 313–345. K. Kim, and A. Sprague, Lunar and Planetary Laboratory, University of Surkov, Y. A., F. F. Kirnozov, V. N. Glazov, A. G. Dunchenko, L. P. Tatsy, Arizona, Tucson, AZ 85721, USA. and O. P. Sobornov (1987), Uranium, thorium, and potassium in the J. Bru¨ckner, G. Dreibus, and H. Wa¨nke, Max-Planck-Institu¨t fu¨r Chemie, Venusian rocks at the landing sites of Vega 1 and 2, Proc. Lunar Planet. Postfach 3060, D-55020 Mainz, Germany. Sci. Conf. 17th, Part 2, J. Geophys. Res., 92, E537–E540. D. Drake, TechSource, P.O. Box 31057, Santa Fe, NM 87594-1057, Surkov, Y. A., L. P. Moskaleva, M. Y. Zolotov, V. P. Kharynkova, O. S. USA. Manvelyan, G. G. Smimov, and A. V. Golovin (1994), Phobos-2 data on C. d’Uston, O. Gasnault, and S. Maurice, Centre d’Etude Spatiale des Martian surface geochemistry, Geochem. Int., 31, 50–58. Rayonnements, CNRS/UPS, 9, avenue du Colonel Roche, Boite postale Tanaka, K. L., M. Chapman, and D. H. Scott (1992), Geologic map of the 4346, F-31028 Toulouse Cedex 4, France. Elysium region of Mars, U.S. Geol. Surv. Misc. Invest. Ser. Map, I-2147, P. Englert and G. J. Taylor, Hawaii Institute of Geophysics and scale 1:5,000,000. Planetology, 1680 East-West Road, Honolulu, HI 96822, USA. (gjtaylor@ Taylor, G. J., and E. R. D. Scott (2004), Mercury, in Treatise on Geochem- higp.hawaii.edu) istry, vol. 1, Meteorites, Comets, and Planets, edited by A. Davis, pp. L. Evans, Computer Sciences Corporation, 7700 Hubble Drive, Lanham, 1477–485, Elsevier, New York. MD 20706, USA. Taylor, G. J., et al. (2006), Variations in K/Th on Mars, J. Geophys. Res., S. Karunatillake and S. Squyres, Center for Radiophysics and Space doi:10.1029/2006JE002676, in press. Research, Cornell University, Ithaca, NY 14853, USA. Taylor, S. R. (1992), Solar System Evolution: A New Perspective, 307 pp., R. Reedy, Institute of Meteoritics, University of New Mexico, Cambridge Univ. Press, New York. Albuquerque, NM 87131, USA. Taylor, S. R., and S. M. McLennan (1985), The Continental Crust: Its R. Starr, Department of Physics, Catholic University of America, Composition and Evolution, 312 pp., Blackwell Sci., Malden, Mass. Washington, DC 20064, USA. Trombka, J. I., et al. (1992), Analysis of Phobos Mission gamma ray spectra from Mars, Proc. Lunar Planet. Sci. Conf., 22nd, 23–29.

16 of 16