- Email: [email protected]
The metal content of the eucrite parent body: constraints from the partitioning behavior of tungsten
Geochimica et Cosmochimica Acta Vol. 46, pp. 2483-2489 © Pergamon Press Ltd. 1982. Printed in U.S.A.
0016-7037/82/122483-07503.00/0
The metal content of the eucrite parent body: constraints from the partitioning behavior of tungsten HORTON E. NEWSOM* and MICHAEL J. DRAKE~f Lunar and Planetary Laboratory, University of Arizona, Tucson, AZ 85721 (Received October 8, 1981; accepted in revised form September 2, 1982) Abstract--The solid metal/silicate melt partition coefficient for W has been determined experimentally to have a value of 25 + 5 at 1190"C and an oxygen fugacity of 10-134, the temperature and oxygen fugaeity conditions at which eucritic basalts formed. Given this partition coefficient, scenarios for the metal content and evolution of the euerite parent body (EPB) are constructed to explain the reduction by a factor of 30, relative to the chondrites, of the W/La ratio in the eucrites. A possible model for the early geologic history of the EPB begins with accretion of a parent body, ebondritie in composition with respect to nonvolatile siderophile and lithophile elements. The solid metal content was between 2% and 10%, which is within the range observed in the ordinary cbondrites. Subsequent heating of the EPB caused the metal phase to separate and become isolated from the silicate phases before the degree of partial melting of the silicates reached 4% to 5%. Equilibrium partitioning of most of the W into the solid metal phase at low degrees of partial melting reduced the W/La ratio in the remaining silicates. Continued partial melting of the silicates generated primary eucritic magmas which recorded the reduced W/La ratio. INTRODUCTION THE EUCRITES are basaltic achondrite meteorites which probably formed on an asteroidal sized parent body shortly after the formation of the solar system. Knowledge of the chemical and mineralogical composition of the eucrite parent body (EPB) would permit us to address several interesting questions such as (a) the homogeneity and internal structure of asteroidal sized bodies, (b) the relationship between the basaltic meteorite parent bodies and the parent bodies of undifferentiated (chondritic) meteorites, and (c) the geochemical behavior of chemical elements during basalt formation under conditions more reducing than are generally found on the Earth at present. A reasonable agreement exists among different investigators regarding the silicate composition of the eucrite basalt source regions except for the FeO-content (Consolmagno and Drake, 1977; Morgan et al., 1978; Dreibus and W~ake, 1980). Estimates of the abundance and distribution of metal vary widely, however, with reported metal contents of 13% --- 3 4% (Morgan et al., 1978), >20% (Dreibus and W~/nke, 1980), and 37 +__5% (Palme and Rammensee, 1981). Two critical observations constrain the metal content of the EPB. First, the low W/La ratios in the eucrites relative to the chondrites suggest that W has been segregated into a metal phase (Rammensee and W//nke, 1977). Second, the positive correlation between W and La in the eucrites suggests that metal was removed from the eucrite source region prior to * Also Department of C-eosciences,current address: MaxPlanck-Institut fur Chemic, Abteilung Kosmochemie, Saarstrasse 23, D-6500 Mainz, West Germany. ~"Also Department of Planetary Sciences.
the igneous events which generated the eucrite magmas (Palme and Rammensee, 1981). Even with these constraints, estimates of metal abundances in the EPB depend critically on the assumptions concerning its early evolution. This paper reexamines the question of the abundance and distribution of metal in the EPB and addresses the question of the early igneous history of that object. We report solid metal/silicate melt partition coefficients for W determined experimentally under conditions appropriate to eucrite petrogenesis. We utilize these data to examine metal-silicate fractionation in the EPB. First, however, we review the behavior of W in nebular and planetary processes. W/LA SYSTEMATICS DURING NEBULAR CONDENSATION AND IGNEOUS DIFFERENTIATION The behavior of W during nebular condensation is incompletely understood. At high temperatures W is associated with refractory noble metals which are embedded in refractory Ca-Al-rich silicate or oxide inclusions (W~inkeet al., 1974;Palme and Wlotzka, 1976). At lower temperatures W appears to behave primarily as an oxidized species, as indicated by the low concentration of W in metallic nickeliron in unequilibrated ordinary chondrites and by the approximately constant W/La ratios in ordinary chondrites of widely varying metal content. This conclusion is supported by the work of Rambaldi and Cendales (1977) who suggested, on the basis of the approximately constant W/Fe ratio in the ordinary chondrites, that W and Fe were substantially present in oxidized form during metal-silicate fractionation. W/inke et al. (1973) reported that the Earth, Moon, and EPB have W/La ratios which correspond to depletions of W relative to La by factors of approximately 19 compared to chondrites (Fig. 1). More recent data show that primary eucritic magmas are slightly more depleted in W relative to La, by a factor of 31 ± 8 (Palme and Rammensee, 1981). In view of the evidence for the behavior of W primarily as an oxidized species in nebular condensates available for
2483
2484
H.E. Newsom and M. J. Drake I000
r
• I00
,
i
3/
"eL
/
.PA .ST
/ 50
ez / • .[a
/
"YA
Sc • .Ju Ju" "%c
2o I0 0.1
s0
"*co "ST8 o'
i8
_o~/ I
I ~ 7 1 I LO ppm
" I
}J° I I0
FIG. 1. W and La concentrations for the eucrites. The reference line labeled Moon is the average W/La ratio for lunar samples. The eucrite and lunar data are from Palme and Rammensee (198 l) and references therein. The C 1, H, L and LL chondrite data are a combination of data from Mason (1979), Palme et al. (1981), and Rambaldi et al. (1979). The eucrite abbreviations are: JU = Juvinas, SC = Sioux County, PA = Pasamonte, IB = Ibitira, ST = Stannern, BO = Bouvante, BE = Bereha, YA = Yamato 74450, SD = Serra de Mage, CA = Cachari, NL = Nuevo Laredo, MC = Moore County, HA = Haraiya, AH = Allan Hills 77005, BI = Binda, JO = Jonzac, MA = Macabini, PO = Pomozdino.
planetary body assembly, this reduction in the W/La ratio is generally attributed to partitioning of W into a metal phase after planetary assembly. The geochemical behavior of W depends critically on the presence or absence of Fe-metai. During mineral-melt fractionation, both W and La behave as lithophile incompatible elements. In the presence of Fe-metal, however, W behaves as a compatible element because of its siderophile character. Hence the depletion of W in the Earth, Moon and EPB is intimately associated with metal-silicate fractionation. With the aid of solid metal/silicate melt partition coefficients, D(W), appropriate to eucritic basalt petrogenesis, the get)logic history of metal in the EPB may be deduced. The partitioning behavior of W depends critically on the degree of partial melting of the EPB. For example, if the silicates are totally molten the solid metal/silicate melt partition coefficient is 25. But if the bulk silicate mineral/silicate melt partition coefficient for W is 0.01, i.e., W behaves as an incompatible element in silicate systems like La, then for an unmelted EPB the solid metal/solid silicate partition coefficient is 2500. For a partially molten EPB, the metal/ silicate partition coefficient lies between these extremes, having a higher value at lower degrees of partial melting. Consequently less metal is required to achieve the observed depletion of W at low degrees than at high degrees of partial melting. EXPERIMENTAL PROCEDURES Solid metal/silicate melt partitioning experiments for W were performed in a vertical muffle tube furnace equipped with a gas mixing apparatus. Oxygen fugacity in the experiments was controlled using mixtures of H2 and CO2, and was actively monitored with a system using a solid ceramic electrolyte oxygen concentration cell, modified after a design by Williams and Muffins (1976). The fugacity probe was calibrated against the iron-wustite (IW) buffer. Pure Fe wires were held at oxygen fugacities either just above or just below the IW buffer for one day. The oxidized wire increases its diameter and crumbles when touched while the reduced wire remains ductile. Myers (1981) has redetermined the equilibrium constant for the IW buffer and places the buffer
approximately 0.2 log units more oxidizing at 1190°C than data from Robie et al. (1978). Our expefienee with the fugacity probe calibration procedure supports Myers' results. In addition, there is close agreement between the fOe determined from the fugacity probe in our experiments and the fO2 calculated using Myers' thermodynamic data from the mole fractions of Fe and FeO in metal and silicate in the experimental charges. For these reasons we believe that all of the experiments listed in Table I were run at an oxygen fugacity within _+0.1 log units of -13.4. Temperature was monitored continuously during the experiments with a P h 0 o - ~ o thermocouple. The thermocouple was calibrated against the melting point of gold wire (1063°C). Thermocouple accuracy was better than -+3°C. Samples were placed in alumina crucibles suspended in the muffle tube. We attempted to evaluate the temperature inside the crucibles. An experiment was run using the broken electrical circuit method, in which gold wire was suspended within a crucible but above an experimental charge. This test indicated that temperatures inside the crucible may be no more than 7°C lower than the temperature of the hot spot recorded by the thermocouple outside of the crucible. The experimental charges consist of a mixture of pure Fe metal (325 mesh) and an oxide mixture which forms the silicate melt (composition listed in Table 1). The W is added either as WO3 powder or as W metal (100 mesh). The ratio of metal to silicate is one to one by volume. During the run the solid metal and molten silicate mixture forms a pillar ( ~ 10 m m high, ~ 5 mm diameter) separated from the walls of the crucible. Interior slices of the experimental run products were analyzed using an ARL SEMQ electron microprobe. A pure metal standard was used for W. A ZAF correction program was used for determining the W content of the metal and silicate phases. A typical counting time of 40 seconds on the WMa peak resulted in approximately 1.3% counting statistics for each metal analysis. A minimum of fifteen metal phase analyses was used for each partition coefficient determination. Typical counting times of 300 to 500 seconds (depending on W concentration) on the WMa peak were used to obtain counting statistics of approximately 6% for each silicate analysis. In general a minimum of eight silicate phase analyses was used for each D (except for MS7d, MS-9c where four analyses were obtained on each phase). A small additional correction of 0.025 to 0.04 wt% was added to the concentration of W in the silicate phase because an interference near the WMo peak resulted in an over-estimation of background intensity. This correction was obtained by peak counting for long periods of time on W-free glass of the experimental composition. An addition of 0.02% to the concentration of W in the metal phase was also necessary and was determined in a similar manner. EXPERIMENTAL RESULTS T h e e x p e r i m e n t a l r u n p r o d u c t s are a n i n t i m a t e m i x t u r e o f m e t a l a n d silicate o n a 100 m i c r o n size scale. T h e analyses listed o n T a b l e 1 a n d plotted in Fig. 2 were m e a s u r e d o n a d j o i n i n g m e t a l a n d silicate areas to m i n i m i z e possible p r o b l e m s associated w i t h a n imperfect a p p r o a c h to e q u i l i b r i u m . T h e r u n p r o d ucts are generally h o m o g e n e o u s w i t h i n each phase, b u t very small local areas w i t h h i g h e r W c o n c e n t r a tions d o exist in s o m e samples a n d m a y represent t h e a - s t r u c t u r a l phase ( H a n s e n , 1958). These isolated W e n r i c h e d areas are visible petrographically as areas consisting largely o f m e t a l with m i n o r silicate. M o s t o f the visible area o f the runs, however, has m e t a l a n d silicate in a ratio o f a p p r o x i m a t e l y 1 to 1.
Eucrite parent body
2485
TABLE 1 RESULTS OF RETAL-SILICATE PARTITIONING EXPERIMENTS FOR W
EXPERIMENT MS5 MS8 MS9 MSll MSIIa MS7d HS9c
DURATION I N DAYS i0 6 12 8 13 8 9
D(W) 25.3 18.4 29.3 21.1 30.3 47.8 45.8
+ 3.8 + 2.0 ~ 2.2 + 1.4 _~ 2.8 + 7.2 + 7.5
W IN ~ T A L WT% W
W IN SILICATE WT% W
STARTING W
1.95 4.29 4.42 4.16
0.077 +_ 0.011
W metal W03 W metal WO 3 both W metal W metal
+ 0.09 + 0.20 _~ 0.05 + 0.08
0.233 + 0.024 0.151 _~ 0.011 0.197 + 0.012
4.33 ~ 009
0.143 _~ 0.013
3.25 + 0.06 3.30 ~ 0.06
0.068 + 0,010 0.072 ~ 0.012
All experiments were run at ll90°C and at log fO2 - -13.4. W-free silicate composition i s
II v t . %
The typlcal
CaO, 13 ~ : : . g AI203, 50 wt.Z 8102,
7 wt.% ME0, 19 wt.Z FeO.
Five experiments (5, 8, 9, 11, 1 la) define a W partition coefficient of 25 + 5 at approximately 1190°C and log fO2 = - 13.4, conditions approximating the eucritic liquidus (Stolper, 1975, 1977). Two types of reversals were run in an attempt to demonstrate equilibrium partitioning. The first type is illustrated by experiments 5, 8, 9 and 11. In experiments 5 and 9, W was present in the starting mixture in metallic form. In experiments 8 and 11, W was originally present in oxidized form as WO3. As expected, experiments starting with W metal have slightly higher solid metal/silicate melt partition coefficients than experiments starting with WO3. In the second type of experiment, the charges (7d, 9c) were first run for 14 days at a much lower oxygen fugacity than that appropriate to the eucritic liquidus, resulting in most of the W dissolving in the metal phase. In these experiments, W could not be detected in the silicate phase. We estimate that D ( W ) > 100. The same charges were then rerun at the oxygen fugacity apropriate to eueritic liquidus conditions with the result that the partition coefficients (Fig. 2) approached values obtained in the first set of experiments but did not reach the equilibrium value of D(W). The larger analytical error may also be a consequence of the failure of these experiments to reach equilibrium. Additional support for the establishment of equilibrium at D(W) = 25 comes from the small difference between experiments 11 and 11 a. Experiment 11 was run for 8 days. A slice was then removed and analyzed, while the remainder of the charge (1 la) was run for an additional 13 days. The values of D(W) for 11 and 11 a are 21 and 30 respectively indicating only a small change in D(W). The variability among the different experiments in Table 1 may be due to a slight variability in the actual oxygen fugacity in each run, to which D(W) is very sensitive in this range of temperature and oxygen fugacity. Because of the limited sensitivity of the electron microprobe the experiments were doped with W to very high concentrations compared to natural systems. Rammensee and W~inke (1977), however, found no deviations from Henry's Law behavior over a wide range in W concentrations. Our experiments showed no significant deviation over a factor of two in concentration (Table 1), although our experiments
were at higher W concentrations than Rammensee and W~inke's. Extrapolation of the results of Rammensee and W~'nke to 1190°C gives a value of D(W) = 17, in acceptable agreement with our results. DISCUSSION
Nature of metal/silicate fractionation in the EPB Petrogenetic processes that account for the depletion of W relative to La in the eucrites fall into three main categories (Table 2). First, there are processes involving equilibrium between liquid metal and silicates during high degrees of partial melting or total melting of the EPB mantle, followed by fractional crystallization to produce the eucrites. Second, there are processes involving equilibration and separation of solid metal and solid silicates in the EPB mantle prior to the melting of the silicates to produce eucrites. Third, there are processes involving equilibration and separation of solid metal and silicates during low degrees of partial melting of the EPB mantle, with additional melting producing the eucrites. The possibility of the presence of a S- or P-bearing metallic liquid must also be considered in the second and third classes of processes. Within each category the possibility of a heterogeneous or homogeneous distribution of metal must be considered. The petrogenetic model that is most reasonable must be com70
, INITIAL STATE of W:
60
..... 1t
W - metal
• .
.
.
.
.
.
.
.
.
.
.
0 W03
.
.
.
.
.
.
W - metal ÷ WO 3
50 [3 40
Run first at low fo 2 Rerun at eucrlte conditions
D(W) 50
20
IO
.05
.
Wt.%
0.15
0.2
0.25
W in silicate
FIG. 2. Solid metal/silicate melt partition coefficient for W, D(W), versus wt% W in the silicate phase. Data from Table 1. Error bars arc 1~.
H. E. Newsom and M. J. Drake
2486
TABLE 2 SUMMARYOF MODELS FOR METAL-SILICATE SEPARATION AND EUCRITE MAGMAFORMATION HETAL-SILICATE EQUILIB. MODEL
MAGMA FORMATION
• Total melting
fract, cryst•
• Subsolldus separation
partial melt
partial melt
L Partial melt low F
METAL CONSISTENT METAL DISTRIBUTION WITHCo CONTENT homogen• or heterogen.
yes
25-38%
heterogen•
no
homogen
yes
I0-50% 24%
heterogen• homogen•
no yes
3-25%a 2-I0%
aThis model requires the f u l l range of metal contents from 3% to 25% metal.
patible with the siderophile element data, in addition to being physically plausible.
Equilibrium between metal and silicate during extensive melting This model is the basis for the calculations of Palme and Rammensee (1981) and assumes that equilibrium is maintained between metal and silicates at high degrees of partial melting or total melting of the EPB. Because metal would settle gravitationally to form a metal core before total melting is accomplished, the initial distribution of metal is unimportant. However, equilibrium must be maintained between the core and the liquid silicate despite the large distance for diffusion and the possibility that a layer of dense residual crystalline silicate (e.g., olivine) could isolate the core. Rapid convection may therefore be required in both the silicate melt and in the metal core. A potent heat source is also required for this model. Several sources, for example, A12~and electromagnetic heating (Sonett and Reynolds, 1979) have been proposed, but it is problematical that they operated during the required time interval. The bulk metal content required to reduce the W/ La ratio during extensive melting may be calculated by the formula: x = (a - I)/(D + a - 1) where D is the solid metal/silicate melt partition coefficient and a is the depletion factor for the W/La ratio (Rammensee and W~nke, 1977). Palme and Rammensee (1981) obtain a solid metal content of 37 + 5% for the EPB by assuming a depletion factor of 31 + 8 and D(W) = 53. The value of D(W) = 53 used by Palme and Rammensee (1981) is for a temperature of 1300°C (appropriate to total melting of the silicates) and for log fOe = -12.55. At lower but still substantial degrees of partial melting of the silicates the amount of metal required is less because of the incompatible nature of W (Newsom and Drake, 1982). Tungsten is excluded from solid silicates so that during partial melting almost all of the W is forced into the solid metal and liquid silicate phases. For example, the equilibrium solid metal content required at 15% melting of the silicates (F = 0.15) is approximately 25%. The range of solid
metal contents required by extensive (F = 0.15-1.0) melting of the EPB is 25%-37%, which is outside the range of the ordinary chondrites (0.3%- 19%; Wasson, 1974)• A major problem with extreme melting of the EPB is the requirement that the eucritic magmas were formed by fractional crystallization from a single magma. Early models did explain the eucrites as residual liquids from fractional crystallization of an unsampled primary magma (Mason, 1962, 1967; Bunch, 1975). Stolper (1977), however, showed that the major and trace element compositions of the eucrites were inconsistent with formation from a single primary magma by fractional crystallization. Stolper (1977) and Consolmagno and Drake (1977) showed that the main group (e.g., Juvinas) and the Stannern trend of eucritic magmas could be produced by equilibrium partial melting of a single primitive source region. The Nuevo Laredo trend appears to require fractional crystallization, possibly of a main group magma. Equilibrium crystallization could also produce magmas identical to those produced by equilibrium partial melting, but this mechanism is physically improbable (Stolper, 1977). More complicated multiple stage melting or crystallization schemes are probably excluded by the ancient crystallization ages of the eucrites, at 4.54 _+ 0.02 b.y. (Basaltic Volcanism, 1981)• Because total melting of the EPB in the absence of a demonstrably adequate heat source is implausible, coupled with a physically improbable mechanism for generating eucritic magmas, we consider extensive melting of the EPB to be an unlikely possibility. The metal content calculated by Palme and Rammensee (1981) for total melting is probably an overestimate of the true metal content of the EPB.
Subsolidus separation of metal In this model, metal separates from the silicates before the silicates melt. The separation of metal is governed by the abundance and size-frequency distribution of the metal, the amount of time available, the EPB gravity, and the viscosity of the EPB mantle (Stevenson, 1981). Perhaps the most critical param-
Eucrite parent body eter is viscosity. Viscosity varies tremendously from a solid unmelted mantle to a completely molten magma ocean. The high viscosity of an unmelted EPB mantle could possibly be reduced by formation of a sulfur-bearing metallic liquid at temperatures above 1000°C, allowing the remaining solid metal to separate more easily (Newsom and Drake, 1979). The metal-silicate partitioning data from ordinary chondrites provide a means for understanding the partitioning behavior of W at subsolidus temperatures (Rambaldi and Cendales, 1977). All three classes of ordinary chondrites have members with partition coefficients greater than 100 (type 6). We adopt 100 as a reasonable value for the metal/silicate partition coefficient during the subsolidus separation, because higher values may be affected by contamination of the silicates by metal. The possible presence of W-bearing minor non-metal phases in the subsolidus assemblage would also tend to prohibit partition coefficients much greater than 100. The metal/silicate partition coefficient values deduced from chondrites are applicable to the EPB since the eucrite source region is essentially chondritic in a temperature-oxygen fugacity plot calculated by the method of Williams (197 l). Note that the T-fO2 curves for the ordinary chondrites in Williams (197 l) are incorrect due to an arithmetic error (pers. commun., Williams, 1979). For subsolidus separation of metal to account for the reduced W / L a ratios, 24% metal is required, assuming a partition coefficient of 100 and a depletion factor of 31. The metal distribution must be relatively homogeneous to produce the relatively constant W / L a ratios observed in the eucrites. A heterogeneous metal content would require an unlikely covariance of partition coefficient with metal content. Metal contents from 10% to 50% would require partition coefficients of 320 to 35. A major difficulty with subsolidus separation of metal is that of achieving complete metal separation from the silicates. Some metal segregation may begin at subsolidus temperatures, but complete separation of metal from silicates is unlikely prior to partial melting of silicates and concomitant reduction of mantle viscosity. This process is therefore also considered to be unlikely.
Metal separation at low degrees of partial melting For equilibration and separation of solid metal from silicate at low degrees of partial melting we can apply our experimentally determined solid metal/silicate melt partition coefficients, D(W), because the actual behavior of W during partial melting in the presence of metal can be calculated. We assumed that D(W) is independent of melt composition, noting that the experimental eucrite composition should be appropriate to the calculations. We have calculated the W and La concentrations in a silicate melt for a range in the degree of partial melting, and with the
2487 I000
' ~
'
'
~
OlW) ,25
C)"MODEL iO0
•
COMPOSITION
50
.~o
" . •
10% 15% ZO% 25%
20
t
I0 O, I
'
=
I
f
F • 0,15 o.~ I I I LO ppm
t 0.04
=
IO
FIG. 3. Tungsten and La concentrations in eucritic liquids assuming equilibrium partial melting. Starting compositions for different amounts of metal and melting proportions for the eucrite source region are based on the work of Consolmagno and Drake (1977). The calculated La concentrations in the eucritic liquids are essentially independent of metal content and are indicated by arrows for different degrees of partial melting (F). The source region compositions are as follows: metal x%, olivine 85% - x%, plagioclase 5%, clinopyroxene 5%, orthopyroxene 5%. The nonmoral melting proportions are 40% plagioclase, 30% clinopyroxene, 30% orthopyroxene. Silicate mineral-silicate melt partition coefficients for W are assumed to be the same as for La and are: olivine, 0.01; low-calcium pyroxene, 0.006; high-calcium pyroxene, 0.07; plagioclase, 0.14. (Consolmagno and Drake, 1977). The degree of partial melting is the fraction of silicate melt divided by the fractions of metal + solid silicate + silicate melt. Dots are eucrite compositions from Fig. I. assumption of different amounts of solid metal in the source region (Fig. 3). Because no experimental solid silicate/silicate melt partition coefficients for W are known, this partition coefficient for W is assumed to be identical to that for La. Heterogeneous metal distribution. The results of the calculations plotted in Fig. 3 allow us to examine two possibilities for metal separation at low degrees of partial melting. The first possibility is the situation where metal is present in the source regions during the partial melting event that produced the eucrites (Newsom and Drake, 1979). As illustrated by Fig. 3, this possibility requires that the source regions contained variable solid metal contents (3% to 25%) because the presence of metal in the source regions buffers the W content of the melt. The substantial heterogeneity in the distribution of metal within the EPB deduced from this model is in contrast to the relative homogeneity of metal distribution in each class of ordinary chondrites. In addition, covariance between the metal content and the degree of partial melting is required in order to produce the relatively constant W / L a ratios observed in the eucrites. These observations raise problems which do not arise with the alternative below. Homogeneous metal distribution. The alternative is that the metal was homogeneously distributed and separated at very low degrees of partial melting of the cucrite source regions before the eucritic magmas were generated. The degree of partial melting represented by the eucritic magmas is between 4% and
2488
H.E. Newsom and M. J. Drake
15% based on modeling of rare earth element abundances, assuming initially chondritic abundances of the REE (Consolmagno and Drake, 1977). Separation of solid metal before partial melting of the silicates reached 4% could have reduced the W/La ratio in the silicate source region to the present eucrite value. Subsequent partial melting or crystallization in the absence of metal would not significantly alter the W/La ratio because both elements are almost identically incompatible in the absence of metal. For example, if the metal separated out just as the degree of partial melting represented by Stannern and Bouvante was reached, a solid metal content of 3% in the source region is required to achieve the necessary W/ La ratios (Fig. 3). Note that the line labeled 3% metal in Fig. 3 represents the silicate liquid composition. The corresponding W and La concentrations in the solid silicates yield the same W/La ratio, at much lower absolute abundances. Figure 3 dearly indicates that at low degrees of partial melting the eucritic W/La ratio is reached with small metal contents. At low metal contents, however, another problem becomes important. Anders (1964) showed that a definite relationship exists between the amount of metal in ordinary chondrites and the concentration of nickel in the metal. Chondrites with 3% metal have approximately 30% Ni in the metal. In order to remain in the range offO2 for the eucrites as determined by Stolper (1977) the FeO content of the silicates must decrease if the Ni content of the metal increases, assuming that fO2 is controlled by the Fe-FeO buffer. The possible range in oxygen fugacity for the eucrites (log/O2 = - 1 3 . 3 _+0.2) determined by Stolper (1977) leaves room, however, for most of the proposed estimates of the FeO content for the EPB, which vary from I 1% (Vizgirda and Anders, 1976) to 28% (Consolmagno and Drake, 1977). Only broad limits can actually be placed on the metal content if it were homogeneously distributed because of experimental uncertainties in determining D(W) and the uncertainties in the fO2 during eucrite formation. Nevertheless, a relatively small metal content, perhaps a maximum of 10%, is predicted. Metal contents much lower than 2% are unlikely because very high values of D(W) are required, along with an extremely small degree of partial melting. In this case, problems arise in that the chondritic model for the relationship between metal content and Ni concentration would indicate relatively oxidizing conditions such that D(W) would not be large enough to explain the W/La depletion. We conclude that equilibration and separation of homogeneously distributed metal at low degrees of partial melting is the mechanism most consistent with eucritic W/La ratios. A metal content for the EPB between 2% and 10% is inferred. COMPARISON WITH AN INDEPENDENT ESTIMATE OF THE EPB METAL CONTENT We have calculated a metal content of 2% to 10% for the EPB assuming equilibrium of W and La
among solid metal, solid silicates and liquid silicates at low degrees of partial melting. This metal content may be compared with an independent estimate. Morgan et al. (1978) estimated the metal content of the EPB to be 13% by assuming that the refractory siderophile element Ir was introduced solely in a high temperature condensate. By estimating the amount of Ir in this condensate and the amount of the condensate in the EPB, Morgan et al. calculated the total amount of Ir in the source region, assuming that the Ir concentration of the EPB metal was the same as that in the IIE iron meteorites. Given uncertainties in assumptions in this work and the work of Morgan et al., the agreement is satisfactory. CONCLUSIONS The metal content of the eucrite parent body obtained by geochemical calculations is highly model dependent. We conclude that metal-silicate separation at very low degrees of partial melting is most consistent with observed siderophile element abundances and experimentally determined partition coefficients. A metal content of 2% to 10% is calculated for this mechanism using our experimentally determined solid metal/silicate melt partition coefficient for W of 25 +_ 5. These conclusions are consistent with the following four stages in the geologic history of the eucrite parent body: 1. During the formation of the solar system the EPB accreted as a "raisin cake" planetary body with a generally homogeneous distribution of metal and silicate. The planetary body was approximately chondritic with respect to nonvolatile lithophile and siderophile elements, and had a solid metal content between 2% and 10%, consistent with the range observed in the ordinary chondrites. The redox state of the EPB was similar to the ordinary chondrites. This planetary body was subsequently heated by an unknown mechanism. 2. Metallic core formation occurred, possibly with some metal separation beginning prior to partial melting. During the early stages of partial melting of the EPB mantle (F < 0.04) metal separation essentially went to completion in the eucritic magma source regions. The separation of solid metal reduced the W/La ratio in the source region silicates by a factor of approximately 3 I. 3. Continued partial melting of the metal-depleted mantle generated eucritic magmas with trace element concentrations which broadly reflect a memory of the core formation event, but which also contain overprints due to mantle heterogeneities and fractionation of mafic phases. These magmas crystallized at or close to the surface. 4. Impact events brecciated and eventually ejected the eucritic meteorites from the EPB. Some fraction of the ejecta subsequently impacted the Earth. Acknowledgments--Discussions with and/or reviews by H.
W~nke, J. W. Morgan, J. Jones, W. Rammensee, H. Palme, E. Stolper, R. Williams, D. S. Burnett, and J. Ganguly were
Eucrite parent body valuab~. This work was s u p p o ~ e d b y N A S A g r a n t N S G 7576.
~FE~N~S Anders E. (1964) Origin, age and composition of meteorites. Space Sci. Rev. 3, 583-714. Basaltic Volcanism Study Project ( 1981 ) Basaltic Volcanism on the Terrestrial Planets. Pergamon Press, Inc. New York. 1286 pp. Bunch T. E. (1975) Petrography and petrology of basaltic achondrite polymict breccias (howardites). Proc. Lunar Sci. Conf. 6th, 469-492. Consolmagno G. J. and Drake M. J. (1977) Composition and evolution of the eucrite parent body: evidence from rare earth elements. Geochim. Cosmochim. Acta 41, 1271-1282. Dreibus G. and W~'nke H. (1980) The bulk composition of the eucrite parent asteroid and its bearing on planetary evolution. Z. Natu(forsch. 35a, 204-216. Hansen M. (1958) Constitution of Binary Alloys. McGraw Hill, New York, 1305 pp. Mason B. (1962) Meteorites. John Wiley and Sons, Inc. New York. 274 pp. Mason B. (1967) The Bununu meteorite, and a discussion of the pyroxene-plagioclase achondrites. Geochim. Cosmochim. Acta 31, 107-115. Mason B. (1979) Data of geochemistry, sixth edition, chapter B. Part i. Meteorites. U.S. GeoL Surv. Prof. Paper 440-B-1. Morgan J. W., Higuchi H., Takahashi H. and Hertogen J. (1978) A "chondritic" eucrite parent body: inference from trace elements. Geochim. Cosmochim. Acta 42, 2738. Myers J. ( 1981) Thermodynamic properties of minerals in the system Fe-Si-O. EOS 62, 1070. Newsom H. E. and Drake M, J. (1979) Metal depletion in the eucrites: evidence for a core or for a heterogeneous mantle in the eucffte parent body (abstract). In Lunar and Planetary Science X, pp. 910-912, The Lunar and Planetary Science Institute, Houston. Newsom H. E. and Drake M. J. (1982) Constraints on the Moon's origin from the partitioning behavior of tungsten. Nature 297, 210-212. Palme H. and Wlotzka X. (1976) A metal particle from a Ca, Al-rich inclusion from the meteorite Allende. Earth Planet. Sci. Left. 33, 45-60. Palme H. and Rammensee W. (1981) The significance of
2489
W in planetary differentiation processes: evidence from new data on eucrites. Proc. Lunar Planet. Sci. Conf. 12B, 949-964. Palme H., Suess H. E. and Zeh H. D. (1981) Abundances of the elements in the solar system. In Landholt-BOrnstein VI, 2, pt. a., p. 257-252, Springer-Verlag, Berlin. Rambaldi E. R. and Cendales M. (1977) Tungsten in ordinary chondrites. Earth Planet. Sci. Lett. 36, 372-380. Rambaldi E. R., Wb;nke H. and Larimer J. W. (1979) Interelement refractory siderophile fractionation in ordinary chondrites. Proc. Lunar Planet. Sci. Conf. lOth, 9971010. Rammensee W. and W/inke H. (1977) On the partition coefficient of tungsten between metal and silicate and its beating on the origin of the moon. Proc. Lunar Sci. Conf. 8th, 399-409. Robie R. A., Hemingway B. S. and Fisher J. R. (1978) Thermodynamic properties of minerals and related substances at 298.15K and 1 Bar (105 pascals) pressure and at higher temperatures. U.S. Geol. Surv. Bull. 1452, 456 pp. Sonctt C. P. and Reynolds R. T. (1979) Primordial heating of asteroid parent bodies. In Asteroids (ed. T. Gehrels), p. 822-848. University of Arizona Press, Tucson. Stevenson D. J. (1981) Models of the Earth's core. Science 214, 611-619. Stolper E. (1975) Petrogenesis of eucrite, howardite and diogenite meteorites. Nature 258, 220-222. Stolper E. (1977) Experimental petrology of eucritic meteorites. Geochim. Cosmochim. Acta 41, 587-611. Vizgirda J. and Anders E. (1976) Composition of the Eucrite Parent Body (abstract). In Lunar Science Vllp. 898-900. The Lunar and Planetary Institute, Houston. Wgnke H., Baddenhausen H., Dreibus G., Jagoutz E., Kruse H., Palme H., Spettel B. and Teschke F. (1973) Multielement analyses of Apollo 15, 16 and 17 samples and the bulk composition of the Moon. Proc. Lunar Sci. Conf. 4th, 1461-1481. W/inke H., Baddenhausen H., Palme H. and Spettel B. (1974) On the chemistry of the Allende inclusions and their origin as high temperature condensates. Earth Planet. Sci. Left. 23, 1-7. Wasson J. T. (1974)Meteorites. Springer-Verlag, Berlin. Williams R. J. (1971) Equilibrium temperatures, pressures, and oxygen fugacities of the equilibrated chondrites. Geochim. Cosmochim. Acta 35, 407-411. Williams R. J. and Mullins O. (1976) A system using solid ceramic oxygen electrolyte cells to measure oxygen fugacities in gas-mixing systems. NASA Tech. Mere. X58167.
0016-7037/82/122483-07503.00/0
The metal content of the eucrite parent body: constraints from the partitioning behavior of tungsten HORTON E. NEWSOM* and MICHAEL J. DRAKE~f Lunar and Planetary Laboratory, University of Arizona, Tucson, AZ 85721 (Received October 8, 1981; accepted in revised form September 2, 1982) Abstract--The solid metal/silicate melt partition coefficient for W has been determined experimentally to have a value of 25 + 5 at 1190"C and an oxygen fugacity of 10-134, the temperature and oxygen fugaeity conditions at which eucritic basalts formed. Given this partition coefficient, scenarios for the metal content and evolution of the euerite parent body (EPB) are constructed to explain the reduction by a factor of 30, relative to the chondrites, of the W/La ratio in the eucrites. A possible model for the early geologic history of the EPB begins with accretion of a parent body, ebondritie in composition with respect to nonvolatile siderophile and lithophile elements. The solid metal content was between 2% and 10%, which is within the range observed in the ordinary cbondrites. Subsequent heating of the EPB caused the metal phase to separate and become isolated from the silicate phases before the degree of partial melting of the silicates reached 4% to 5%. Equilibrium partitioning of most of the W into the solid metal phase at low degrees of partial melting reduced the W/La ratio in the remaining silicates. Continued partial melting of the silicates generated primary eucritic magmas which recorded the reduced W/La ratio. INTRODUCTION THE EUCRITES are basaltic achondrite meteorites which probably formed on an asteroidal sized parent body shortly after the formation of the solar system. Knowledge of the chemical and mineralogical composition of the eucrite parent body (EPB) would permit us to address several interesting questions such as (a) the homogeneity and internal structure of asteroidal sized bodies, (b) the relationship between the basaltic meteorite parent bodies and the parent bodies of undifferentiated (chondritic) meteorites, and (c) the geochemical behavior of chemical elements during basalt formation under conditions more reducing than are generally found on the Earth at present. A reasonable agreement exists among different investigators regarding the silicate composition of the eucrite basalt source regions except for the FeO-content (Consolmagno and Drake, 1977; Morgan et al., 1978; Dreibus and W~ake, 1980). Estimates of the abundance and distribution of metal vary widely, however, with reported metal contents of 13% --- 3 4% (Morgan et al., 1978), >20% (Dreibus and W~/nke, 1980), and 37 +__5% (Palme and Rammensee, 1981). Two critical observations constrain the metal content of the EPB. First, the low W/La ratios in the eucrites relative to the chondrites suggest that W has been segregated into a metal phase (Rammensee and W//nke, 1977). Second, the positive correlation between W and La in the eucrites suggests that metal was removed from the eucrite source region prior to * Also Department of C-eosciences,current address: MaxPlanck-Institut fur Chemic, Abteilung Kosmochemie, Saarstrasse 23, D-6500 Mainz, West Germany. ~"Also Department of Planetary Sciences.
the igneous events which generated the eucrite magmas (Palme and Rammensee, 1981). Even with these constraints, estimates of metal abundances in the EPB depend critically on the assumptions concerning its early evolution. This paper reexamines the question of the abundance and distribution of metal in the EPB and addresses the question of the early igneous history of that object. We report solid metal/silicate melt partition coefficients for W determined experimentally under conditions appropriate to eucrite petrogenesis. We utilize these data to examine metal-silicate fractionation in the EPB. First, however, we review the behavior of W in nebular and planetary processes. W/LA SYSTEMATICS DURING NEBULAR CONDENSATION AND IGNEOUS DIFFERENTIATION The behavior of W during nebular condensation is incompletely understood. At high temperatures W is associated with refractory noble metals which are embedded in refractory Ca-Al-rich silicate or oxide inclusions (W~inkeet al., 1974;Palme and Wlotzka, 1976). At lower temperatures W appears to behave primarily as an oxidized species, as indicated by the low concentration of W in metallic nickeliron in unequilibrated ordinary chondrites and by the approximately constant W/La ratios in ordinary chondrites of widely varying metal content. This conclusion is supported by the work of Rambaldi and Cendales (1977) who suggested, on the basis of the approximately constant W/Fe ratio in the ordinary chondrites, that W and Fe were substantially present in oxidized form during metal-silicate fractionation. W/inke et al. (1973) reported that the Earth, Moon, and EPB have W/La ratios which correspond to depletions of W relative to La by factors of approximately 19 compared to chondrites (Fig. 1). More recent data show that primary eucritic magmas are slightly more depleted in W relative to La, by a factor of 31 ± 8 (Palme and Rammensee, 1981). In view of the evidence for the behavior of W primarily as an oxidized species in nebular condensates available for
2483
2484
H.E. Newsom and M. J. Drake I000
r
• I00
,
i
3/
"eL
/
.PA .ST
/ 50
ez / • .[a
/
"YA
Sc • .Ju Ju" "%c
2o I0 0.1
s0
"*co "ST8 o'
i8
_o~/ I
I ~ 7 1 I LO ppm
" I
}J° I I0
FIG. 1. W and La concentrations for the eucrites. The reference line labeled Moon is the average W/La ratio for lunar samples. The eucrite and lunar data are from Palme and Rammensee (198 l) and references therein. The C 1, H, L and LL chondrite data are a combination of data from Mason (1979), Palme et al. (1981), and Rambaldi et al. (1979). The eucrite abbreviations are: JU = Juvinas, SC = Sioux County, PA = Pasamonte, IB = Ibitira, ST = Stannern, BO = Bouvante, BE = Bereha, YA = Yamato 74450, SD = Serra de Mage, CA = Cachari, NL = Nuevo Laredo, MC = Moore County, HA = Haraiya, AH = Allan Hills 77005, BI = Binda, JO = Jonzac, MA = Macabini, PO = Pomozdino.
planetary body assembly, this reduction in the W/La ratio is generally attributed to partitioning of W into a metal phase after planetary assembly. The geochemical behavior of W depends critically on the presence or absence of Fe-metai. During mineral-melt fractionation, both W and La behave as lithophile incompatible elements. In the presence of Fe-metal, however, W behaves as a compatible element because of its siderophile character. Hence the depletion of W in the Earth, Moon and EPB is intimately associated with metal-silicate fractionation. With the aid of solid metal/silicate melt partition coefficients, D(W), appropriate to eucritic basalt petrogenesis, the get)logic history of metal in the EPB may be deduced. The partitioning behavior of W depends critically on the degree of partial melting of the EPB. For example, if the silicates are totally molten the solid metal/silicate melt partition coefficient is 25. But if the bulk silicate mineral/silicate melt partition coefficient for W is 0.01, i.e., W behaves as an incompatible element in silicate systems like La, then for an unmelted EPB the solid metal/solid silicate partition coefficient is 2500. For a partially molten EPB, the metal/ silicate partition coefficient lies between these extremes, having a higher value at lower degrees of partial melting. Consequently less metal is required to achieve the observed depletion of W at low degrees than at high degrees of partial melting. EXPERIMENTAL PROCEDURES Solid metal/silicate melt partitioning experiments for W were performed in a vertical muffle tube furnace equipped with a gas mixing apparatus. Oxygen fugacity in the experiments was controlled using mixtures of H2 and CO2, and was actively monitored with a system using a solid ceramic electrolyte oxygen concentration cell, modified after a design by Williams and Muffins (1976). The fugacity probe was calibrated against the iron-wustite (IW) buffer. Pure Fe wires were held at oxygen fugacities either just above or just below the IW buffer for one day. The oxidized wire increases its diameter and crumbles when touched while the reduced wire remains ductile. Myers (1981) has redetermined the equilibrium constant for the IW buffer and places the buffer
approximately 0.2 log units more oxidizing at 1190°C than data from Robie et al. (1978). Our expefienee with the fugacity probe calibration procedure supports Myers' results. In addition, there is close agreement between the fOe determined from the fugacity probe in our experiments and the fO2 calculated using Myers' thermodynamic data from the mole fractions of Fe and FeO in metal and silicate in the experimental charges. For these reasons we believe that all of the experiments listed in Table I were run at an oxygen fugacity within _+0.1 log units of -13.4. Temperature was monitored continuously during the experiments with a P h 0 o - ~ o thermocouple. The thermocouple was calibrated against the melting point of gold wire (1063°C). Thermocouple accuracy was better than -+3°C. Samples were placed in alumina crucibles suspended in the muffle tube. We attempted to evaluate the temperature inside the crucibles. An experiment was run using the broken electrical circuit method, in which gold wire was suspended within a crucible but above an experimental charge. This test indicated that temperatures inside the crucible may be no more than 7°C lower than the temperature of the hot spot recorded by the thermocouple outside of the crucible. The experimental charges consist of a mixture of pure Fe metal (325 mesh) and an oxide mixture which forms the silicate melt (composition listed in Table 1). The W is added either as WO3 powder or as W metal (100 mesh). The ratio of metal to silicate is one to one by volume. During the run the solid metal and molten silicate mixture forms a pillar ( ~ 10 m m high, ~ 5 mm diameter) separated from the walls of the crucible. Interior slices of the experimental run products were analyzed using an ARL SEMQ electron microprobe. A pure metal standard was used for W. A ZAF correction program was used for determining the W content of the metal and silicate phases. A typical counting time of 40 seconds on the WMa peak resulted in approximately 1.3% counting statistics for each metal analysis. A minimum of fifteen metal phase analyses was used for each partition coefficient determination. Typical counting times of 300 to 500 seconds (depending on W concentration) on the WMa peak were used to obtain counting statistics of approximately 6% for each silicate analysis. In general a minimum of eight silicate phase analyses was used for each D (except for MS7d, MS-9c where four analyses were obtained on each phase). A small additional correction of 0.025 to 0.04 wt% was added to the concentration of W in the silicate phase because an interference near the WMo peak resulted in an over-estimation of background intensity. This correction was obtained by peak counting for long periods of time on W-free glass of the experimental composition. An addition of 0.02% to the concentration of W in the metal phase was also necessary and was determined in a similar manner. EXPERIMENTAL RESULTS T h e e x p e r i m e n t a l r u n p r o d u c t s are a n i n t i m a t e m i x t u r e o f m e t a l a n d silicate o n a 100 m i c r o n size scale. T h e analyses listed o n T a b l e 1 a n d plotted in Fig. 2 were m e a s u r e d o n a d j o i n i n g m e t a l a n d silicate areas to m i n i m i z e possible p r o b l e m s associated w i t h a n imperfect a p p r o a c h to e q u i l i b r i u m . T h e r u n p r o d ucts are generally h o m o g e n e o u s w i t h i n each phase, b u t very small local areas w i t h h i g h e r W c o n c e n t r a tions d o exist in s o m e samples a n d m a y represent t h e a - s t r u c t u r a l phase ( H a n s e n , 1958). These isolated W e n r i c h e d areas are visible petrographically as areas consisting largely o f m e t a l with m i n o r silicate. M o s t o f the visible area o f the runs, however, has m e t a l a n d silicate in a ratio o f a p p r o x i m a t e l y 1 to 1.
Eucrite parent body
2485
TABLE 1 RESULTS OF RETAL-SILICATE PARTITIONING EXPERIMENTS FOR W
EXPERIMENT MS5 MS8 MS9 MSll MSIIa MS7d HS9c
DURATION I N DAYS i0 6 12 8 13 8 9
D(W) 25.3 18.4 29.3 21.1 30.3 47.8 45.8
+ 3.8 + 2.0 ~ 2.2 + 1.4 _~ 2.8 + 7.2 + 7.5
W IN ~ T A L WT% W
W IN SILICATE WT% W
STARTING W
1.95 4.29 4.42 4.16
0.077 +_ 0.011
W metal W03 W metal WO 3 both W metal W metal
+ 0.09 + 0.20 _~ 0.05 + 0.08
0.233 + 0.024 0.151 _~ 0.011 0.197 + 0.012
4.33 ~ 009
0.143 _~ 0.013
3.25 + 0.06 3.30 ~ 0.06
0.068 + 0,010 0.072 ~ 0.012
All experiments were run at ll90°C and at log fO2 - -13.4. W-free silicate composition i s
II v t . %
The typlcal
CaO, 13 ~ : : . g AI203, 50 wt.Z 8102,
7 wt.% ME0, 19 wt.Z FeO.
Five experiments (5, 8, 9, 11, 1 la) define a W partition coefficient of 25 + 5 at approximately 1190°C and log fO2 = - 13.4, conditions approximating the eucritic liquidus (Stolper, 1975, 1977). Two types of reversals were run in an attempt to demonstrate equilibrium partitioning. The first type is illustrated by experiments 5, 8, 9 and 11. In experiments 5 and 9, W was present in the starting mixture in metallic form. In experiments 8 and 11, W was originally present in oxidized form as WO3. As expected, experiments starting with W metal have slightly higher solid metal/silicate melt partition coefficients than experiments starting with WO3. In the second type of experiment, the charges (7d, 9c) were first run for 14 days at a much lower oxygen fugacity than that appropriate to the eucritic liquidus, resulting in most of the W dissolving in the metal phase. In these experiments, W could not be detected in the silicate phase. We estimate that D ( W ) > 100. The same charges were then rerun at the oxygen fugacity apropriate to eueritic liquidus conditions with the result that the partition coefficients (Fig. 2) approached values obtained in the first set of experiments but did not reach the equilibrium value of D(W). The larger analytical error may also be a consequence of the failure of these experiments to reach equilibrium. Additional support for the establishment of equilibrium at D(W) = 25 comes from the small difference between experiments 11 and 11 a. Experiment 11 was run for 8 days. A slice was then removed and analyzed, while the remainder of the charge (1 la) was run for an additional 13 days. The values of D(W) for 11 and 11 a are 21 and 30 respectively indicating only a small change in D(W). The variability among the different experiments in Table 1 may be due to a slight variability in the actual oxygen fugacity in each run, to which D(W) is very sensitive in this range of temperature and oxygen fugacity. Because of the limited sensitivity of the electron microprobe the experiments were doped with W to very high concentrations compared to natural systems. Rammensee and W~inke (1977), however, found no deviations from Henry's Law behavior over a wide range in W concentrations. Our experiments showed no significant deviation over a factor of two in concentration (Table 1), although our experiments
were at higher W concentrations than Rammensee and W~inke's. Extrapolation of the results of Rammensee and W~'nke to 1190°C gives a value of D(W) = 17, in acceptable agreement with our results. DISCUSSION
Nature of metal/silicate fractionation in the EPB Petrogenetic processes that account for the depletion of W relative to La in the eucrites fall into three main categories (Table 2). First, there are processes involving equilibrium between liquid metal and silicates during high degrees of partial melting or total melting of the EPB mantle, followed by fractional crystallization to produce the eucrites. Second, there are processes involving equilibration and separation of solid metal and solid silicates in the EPB mantle prior to the melting of the silicates to produce eucrites. Third, there are processes involving equilibration and separation of solid metal and silicates during low degrees of partial melting of the EPB mantle, with additional melting producing the eucrites. The possibility of the presence of a S- or P-bearing metallic liquid must also be considered in the second and third classes of processes. Within each category the possibility of a heterogeneous or homogeneous distribution of metal must be considered. The petrogenetic model that is most reasonable must be com70
, INITIAL STATE of W:
60
..... 1t
W - metal
• .
.
.
.
.
.
.
.
.
.
.
0 W03
.
.
.
.
.
.
W - metal ÷ WO 3
50 [3 40
Run first at low fo 2 Rerun at eucrlte conditions
D(W) 50
20
IO
.05
.
Wt.%
0.15
0.2
0.25
W in silicate
FIG. 2. Solid metal/silicate melt partition coefficient for W, D(W), versus wt% W in the silicate phase. Data from Table 1. Error bars arc 1~.
H. E. Newsom and M. J. Drake
2486
TABLE 2 SUMMARYOF MODELS FOR METAL-SILICATE SEPARATION AND EUCRITE MAGMAFORMATION HETAL-SILICATE EQUILIB. MODEL
MAGMA FORMATION
• Total melting
fract, cryst•
• Subsolldus separation
partial melt
partial melt
L Partial melt low F
METAL CONSISTENT METAL DISTRIBUTION WITHCo CONTENT homogen• or heterogen.
yes
25-38%
heterogen•
no
homogen
yes
I0-50% 24%
heterogen• homogen•
no yes
3-25%a 2-I0%
aThis model requires the f u l l range of metal contents from 3% to 25% metal.
patible with the siderophile element data, in addition to being physically plausible.
Equilibrium between metal and silicate during extensive melting This model is the basis for the calculations of Palme and Rammensee (1981) and assumes that equilibrium is maintained between metal and silicates at high degrees of partial melting or total melting of the EPB. Because metal would settle gravitationally to form a metal core before total melting is accomplished, the initial distribution of metal is unimportant. However, equilibrium must be maintained between the core and the liquid silicate despite the large distance for diffusion and the possibility that a layer of dense residual crystalline silicate (e.g., olivine) could isolate the core. Rapid convection may therefore be required in both the silicate melt and in the metal core. A potent heat source is also required for this model. Several sources, for example, A12~and electromagnetic heating (Sonett and Reynolds, 1979) have been proposed, but it is problematical that they operated during the required time interval. The bulk metal content required to reduce the W/ La ratio during extensive melting may be calculated by the formula: x = (a - I)/(D + a - 1) where D is the solid metal/silicate melt partition coefficient and a is the depletion factor for the W/La ratio (Rammensee and W~nke, 1977). Palme and Rammensee (1981) obtain a solid metal content of 37 + 5% for the EPB by assuming a depletion factor of 31 + 8 and D(W) = 53. The value of D(W) = 53 used by Palme and Rammensee (1981) is for a temperature of 1300°C (appropriate to total melting of the silicates) and for log fOe = -12.55. At lower but still substantial degrees of partial melting of the silicates the amount of metal required is less because of the incompatible nature of W (Newsom and Drake, 1982). Tungsten is excluded from solid silicates so that during partial melting almost all of the W is forced into the solid metal and liquid silicate phases. For example, the equilibrium solid metal content required at 15% melting of the silicates (F = 0.15) is approximately 25%. The range of solid
metal contents required by extensive (F = 0.15-1.0) melting of the EPB is 25%-37%, which is outside the range of the ordinary chondrites (0.3%- 19%; Wasson, 1974)• A major problem with extreme melting of the EPB is the requirement that the eucritic magmas were formed by fractional crystallization from a single magma. Early models did explain the eucrites as residual liquids from fractional crystallization of an unsampled primary magma (Mason, 1962, 1967; Bunch, 1975). Stolper (1977), however, showed that the major and trace element compositions of the eucrites were inconsistent with formation from a single primary magma by fractional crystallization. Stolper (1977) and Consolmagno and Drake (1977) showed that the main group (e.g., Juvinas) and the Stannern trend of eucritic magmas could be produced by equilibrium partial melting of a single primitive source region. The Nuevo Laredo trend appears to require fractional crystallization, possibly of a main group magma. Equilibrium crystallization could also produce magmas identical to those produced by equilibrium partial melting, but this mechanism is physically improbable (Stolper, 1977). More complicated multiple stage melting or crystallization schemes are probably excluded by the ancient crystallization ages of the eucrites, at 4.54 _+ 0.02 b.y. (Basaltic Volcanism, 1981)• Because total melting of the EPB in the absence of a demonstrably adequate heat source is implausible, coupled with a physically improbable mechanism for generating eucritic magmas, we consider extensive melting of the EPB to be an unlikely possibility. The metal content calculated by Palme and Rammensee (1981) for total melting is probably an overestimate of the true metal content of the EPB.
Subsolidus separation of metal In this model, metal separates from the silicates before the silicates melt. The separation of metal is governed by the abundance and size-frequency distribution of the metal, the amount of time available, the EPB gravity, and the viscosity of the EPB mantle (Stevenson, 1981). Perhaps the most critical param-
Eucrite parent body eter is viscosity. Viscosity varies tremendously from a solid unmelted mantle to a completely molten magma ocean. The high viscosity of an unmelted EPB mantle could possibly be reduced by formation of a sulfur-bearing metallic liquid at temperatures above 1000°C, allowing the remaining solid metal to separate more easily (Newsom and Drake, 1979). The metal-silicate partitioning data from ordinary chondrites provide a means for understanding the partitioning behavior of W at subsolidus temperatures (Rambaldi and Cendales, 1977). All three classes of ordinary chondrites have members with partition coefficients greater than 100 (type 6). We adopt 100 as a reasonable value for the metal/silicate partition coefficient during the subsolidus separation, because higher values may be affected by contamination of the silicates by metal. The possible presence of W-bearing minor non-metal phases in the subsolidus assemblage would also tend to prohibit partition coefficients much greater than 100. The metal/silicate partition coefficient values deduced from chondrites are applicable to the EPB since the eucrite source region is essentially chondritic in a temperature-oxygen fugacity plot calculated by the method of Williams (197 l). Note that the T-fO2 curves for the ordinary chondrites in Williams (197 l) are incorrect due to an arithmetic error (pers. commun., Williams, 1979). For subsolidus separation of metal to account for the reduced W / L a ratios, 24% metal is required, assuming a partition coefficient of 100 and a depletion factor of 31. The metal distribution must be relatively homogeneous to produce the relatively constant W / L a ratios observed in the eucrites. A heterogeneous metal content would require an unlikely covariance of partition coefficient with metal content. Metal contents from 10% to 50% would require partition coefficients of 320 to 35. A major difficulty with subsolidus separation of metal is that of achieving complete metal separation from the silicates. Some metal segregation may begin at subsolidus temperatures, but complete separation of metal from silicates is unlikely prior to partial melting of silicates and concomitant reduction of mantle viscosity. This process is therefore also considered to be unlikely.
Metal separation at low degrees of partial melting For equilibration and separation of solid metal from silicate at low degrees of partial melting we can apply our experimentally determined solid metal/silicate melt partition coefficients, D(W), because the actual behavior of W during partial melting in the presence of metal can be calculated. We assumed that D(W) is independent of melt composition, noting that the experimental eucrite composition should be appropriate to the calculations. We have calculated the W and La concentrations in a silicate melt for a range in the degree of partial melting, and with the
2487 I000
' ~
'
'
~
OlW) ,25
C)"MODEL iO0
•
COMPOSITION
50
.~o
" . •
10% 15% ZO% 25%
20
t
I0 O, I
'
=
I
f
F • 0,15 o.~ I I I LO ppm
t 0.04
=
IO
FIG. 3. Tungsten and La concentrations in eucritic liquids assuming equilibrium partial melting. Starting compositions for different amounts of metal and melting proportions for the eucrite source region are based on the work of Consolmagno and Drake (1977). The calculated La concentrations in the eucritic liquids are essentially independent of metal content and are indicated by arrows for different degrees of partial melting (F). The source region compositions are as follows: metal x%, olivine 85% - x%, plagioclase 5%, clinopyroxene 5%, orthopyroxene 5%. The nonmoral melting proportions are 40% plagioclase, 30% clinopyroxene, 30% orthopyroxene. Silicate mineral-silicate melt partition coefficients for W are assumed to be the same as for La and are: olivine, 0.01; low-calcium pyroxene, 0.006; high-calcium pyroxene, 0.07; plagioclase, 0.14. (Consolmagno and Drake, 1977). The degree of partial melting is the fraction of silicate melt divided by the fractions of metal + solid silicate + silicate melt. Dots are eucrite compositions from Fig. I. assumption of different amounts of solid metal in the source region (Fig. 3). Because no experimental solid silicate/silicate melt partition coefficients for W are known, this partition coefficient for W is assumed to be identical to that for La. Heterogeneous metal distribution. The results of the calculations plotted in Fig. 3 allow us to examine two possibilities for metal separation at low degrees of partial melting. The first possibility is the situation where metal is present in the source regions during the partial melting event that produced the eucrites (Newsom and Drake, 1979). As illustrated by Fig. 3, this possibility requires that the source regions contained variable solid metal contents (3% to 25%) because the presence of metal in the source regions buffers the W content of the melt. The substantial heterogeneity in the distribution of metal within the EPB deduced from this model is in contrast to the relative homogeneity of metal distribution in each class of ordinary chondrites. In addition, covariance between the metal content and the degree of partial melting is required in order to produce the relatively constant W / L a ratios observed in the eucrites. These observations raise problems which do not arise with the alternative below. Homogeneous metal distribution. The alternative is that the metal was homogeneously distributed and separated at very low degrees of partial melting of the cucrite source regions before the eucritic magmas were generated. The degree of partial melting represented by the eucritic magmas is between 4% and
2488
H.E. Newsom and M. J. Drake
15% based on modeling of rare earth element abundances, assuming initially chondritic abundances of the REE (Consolmagno and Drake, 1977). Separation of solid metal before partial melting of the silicates reached 4% could have reduced the W/La ratio in the silicate source region to the present eucrite value. Subsequent partial melting or crystallization in the absence of metal would not significantly alter the W/La ratio because both elements are almost identically incompatible in the absence of metal. For example, if the metal separated out just as the degree of partial melting represented by Stannern and Bouvante was reached, a solid metal content of 3% in the source region is required to achieve the necessary W/ La ratios (Fig. 3). Note that the line labeled 3% metal in Fig. 3 represents the silicate liquid composition. The corresponding W and La concentrations in the solid silicates yield the same W/La ratio, at much lower absolute abundances. Figure 3 dearly indicates that at low degrees of partial melting the eucritic W/La ratio is reached with small metal contents. At low metal contents, however, another problem becomes important. Anders (1964) showed that a definite relationship exists between the amount of metal in ordinary chondrites and the concentration of nickel in the metal. Chondrites with 3% metal have approximately 30% Ni in the metal. In order to remain in the range offO2 for the eucrites as determined by Stolper (1977) the FeO content of the silicates must decrease if the Ni content of the metal increases, assuming that fO2 is controlled by the Fe-FeO buffer. The possible range in oxygen fugacity for the eucrites (log/O2 = - 1 3 . 3 _+0.2) determined by Stolper (1977) leaves room, however, for most of the proposed estimates of the FeO content for the EPB, which vary from I 1% (Vizgirda and Anders, 1976) to 28% (Consolmagno and Drake, 1977). Only broad limits can actually be placed on the metal content if it were homogeneously distributed because of experimental uncertainties in determining D(W) and the uncertainties in the fO2 during eucrite formation. Nevertheless, a relatively small metal content, perhaps a maximum of 10%, is predicted. Metal contents much lower than 2% are unlikely because very high values of D(W) are required, along with an extremely small degree of partial melting. In this case, problems arise in that the chondritic model for the relationship between metal content and Ni concentration would indicate relatively oxidizing conditions such that D(W) would not be large enough to explain the W/La depletion. We conclude that equilibration and separation of homogeneously distributed metal at low degrees of partial melting is the mechanism most consistent with eucritic W/La ratios. A metal content for the EPB between 2% and 10% is inferred. COMPARISON WITH AN INDEPENDENT ESTIMATE OF THE EPB METAL CONTENT We have calculated a metal content of 2% to 10% for the EPB assuming equilibrium of W and La
among solid metal, solid silicates and liquid silicates at low degrees of partial melting. This metal content may be compared with an independent estimate. Morgan et al. (1978) estimated the metal content of the EPB to be 13% by assuming that the refractory siderophile element Ir was introduced solely in a high temperature condensate. By estimating the amount of Ir in this condensate and the amount of the condensate in the EPB, Morgan et al. calculated the total amount of Ir in the source region, assuming that the Ir concentration of the EPB metal was the same as that in the IIE iron meteorites. Given uncertainties in assumptions in this work and the work of Morgan et al., the agreement is satisfactory. CONCLUSIONS The metal content of the eucrite parent body obtained by geochemical calculations is highly model dependent. We conclude that metal-silicate separation at very low degrees of partial melting is most consistent with observed siderophile element abundances and experimentally determined partition coefficients. A metal content of 2% to 10% is calculated for this mechanism using our experimentally determined solid metal/silicate melt partition coefficient for W of 25 +_ 5. These conclusions are consistent with the following four stages in the geologic history of the eucrite parent body: 1. During the formation of the solar system the EPB accreted as a "raisin cake" planetary body with a generally homogeneous distribution of metal and silicate. The planetary body was approximately chondritic with respect to nonvolatile lithophile and siderophile elements, and had a solid metal content between 2% and 10%, consistent with the range observed in the ordinary chondrites. The redox state of the EPB was similar to the ordinary chondrites. This planetary body was subsequently heated by an unknown mechanism. 2. Metallic core formation occurred, possibly with some metal separation beginning prior to partial melting. During the early stages of partial melting of the EPB mantle (F < 0.04) metal separation essentially went to completion in the eucritic magma source regions. The separation of solid metal reduced the W/La ratio in the source region silicates by a factor of approximately 3 I. 3. Continued partial melting of the metal-depleted mantle generated eucritic magmas with trace element concentrations which broadly reflect a memory of the core formation event, but which also contain overprints due to mantle heterogeneities and fractionation of mafic phases. These magmas crystallized at or close to the surface. 4. Impact events brecciated and eventually ejected the eucritic meteorites from the EPB. Some fraction of the ejecta subsequently impacted the Earth. Acknowledgments--Discussions with and/or reviews by H.
W~nke, J. W. Morgan, J. Jones, W. Rammensee, H. Palme, E. Stolper, R. Williams, D. S. Burnett, and J. Ganguly were
Eucrite parent body valuab~. This work was s u p p o ~ e d b y N A S A g r a n t N S G 7576.
~FE~N~S Anders E. (1964) Origin, age and composition of meteorites. Space Sci. Rev. 3, 583-714. Basaltic Volcanism Study Project ( 1981 ) Basaltic Volcanism on the Terrestrial Planets. Pergamon Press, Inc. New York. 1286 pp. Bunch T. E. (1975) Petrography and petrology of basaltic achondrite polymict breccias (howardites). Proc. Lunar Sci. Conf. 6th, 469-492. Consolmagno G. J. and Drake M. J. (1977) Composition and evolution of the eucrite parent body: evidence from rare earth elements. Geochim. Cosmochim. Acta 41, 1271-1282. Dreibus G. and W~'nke H. (1980) The bulk composition of the eucrite parent asteroid and its bearing on planetary evolution. Z. Natu(forsch. 35a, 204-216. Hansen M. (1958) Constitution of Binary Alloys. McGraw Hill, New York, 1305 pp. Mason B. (1962) Meteorites. John Wiley and Sons, Inc. New York. 274 pp. Mason B. (1967) The Bununu meteorite, and a discussion of the pyroxene-plagioclase achondrites. Geochim. Cosmochim. Acta 31, 107-115. Mason B. (1979) Data of geochemistry, sixth edition, chapter B. Part i. Meteorites. U.S. GeoL Surv. Prof. Paper 440-B-1. Morgan J. W., Higuchi H., Takahashi H. and Hertogen J. (1978) A "chondritic" eucrite parent body: inference from trace elements. Geochim. Cosmochim. Acta 42, 2738. Myers J. ( 1981) Thermodynamic properties of minerals in the system Fe-Si-O. EOS 62, 1070. Newsom H. E. and Drake M, J. (1979) Metal depletion in the eucrites: evidence for a core or for a heterogeneous mantle in the eucffte parent body (abstract). In Lunar and Planetary Science X, pp. 910-912, The Lunar and Planetary Science Institute, Houston. Newsom H. E. and Drake M. J. (1982) Constraints on the Moon's origin from the partitioning behavior of tungsten. Nature 297, 210-212. Palme H. and Wlotzka X. (1976) A metal particle from a Ca, Al-rich inclusion from the meteorite Allende. Earth Planet. Sci. Left. 33, 45-60. Palme H. and Rammensee W. (1981) The significance of
2489
W in planetary differentiation processes: evidence from new data on eucrites. Proc. Lunar Planet. Sci. Conf. 12B, 949-964. Palme H., Suess H. E. and Zeh H. D. (1981) Abundances of the elements in the solar system. In Landholt-BOrnstein VI, 2, pt. a., p. 257-252, Springer-Verlag, Berlin. Rambaldi E. R. and Cendales M. (1977) Tungsten in ordinary chondrites. Earth Planet. Sci. Lett. 36, 372-380. Rambaldi E. R., Wb;nke H. and Larimer J. W. (1979) Interelement refractory siderophile fractionation in ordinary chondrites. Proc. Lunar Planet. Sci. Conf. lOth, 9971010. Rammensee W. and W/inke H. (1977) On the partition coefficient of tungsten between metal and silicate and its beating on the origin of the moon. Proc. Lunar Sci. Conf. 8th, 399-409. Robie R. A., Hemingway B. S. and Fisher J. R. (1978) Thermodynamic properties of minerals and related substances at 298.15K and 1 Bar (105 pascals) pressure and at higher temperatures. U.S. Geol. Surv. Bull. 1452, 456 pp. Sonctt C. P. and Reynolds R. T. (1979) Primordial heating of asteroid parent bodies. In Asteroids (ed. T. Gehrels), p. 822-848. University of Arizona Press, Tucson. Stevenson D. J. (1981) Models of the Earth's core. Science 214, 611-619. Stolper E. (1975) Petrogenesis of eucrite, howardite and diogenite meteorites. Nature 258, 220-222. Stolper E. (1977) Experimental petrology of eucritic meteorites. Geochim. Cosmochim. Acta 41, 587-611. Vizgirda J. and Anders E. (1976) Composition of the Eucrite Parent Body (abstract). In Lunar Science Vllp. 898-900. The Lunar and Planetary Institute, Houston. Wgnke H., Baddenhausen H., Dreibus G., Jagoutz E., Kruse H., Palme H., Spettel B. and Teschke F. (1973) Multielement analyses of Apollo 15, 16 and 17 samples and the bulk composition of the Moon. Proc. Lunar Sci. Conf. 4th, 1461-1481. W/inke H., Baddenhausen H., Palme H. and Spettel B. (1974) On the chemistry of the Allende inclusions and their origin as high temperature condensates. Earth Planet. Sci. Left. 23, 1-7. Wasson J. T. (1974)Meteorites. Springer-Verlag, Berlin. Williams R. J. (1971) Equilibrium temperatures, pressures, and oxygen fugacities of the equilibrated chondrites. Geochim. Cosmochim. Acta 35, 407-411. Williams R. J. and Mullins O. (1976) A system using solid ceramic oxygen electrolyte cells to measure oxygen fugacities in gas-mixing systems. NASA Tech. Mere. X58167.