Authors Copy
Authors Copy Earth and Planetary Science Letters, 24(1974)59—70 © North-Holland Publishing Company, Amsterdam Printed in The Netherlands
E
—
METAL—OLIVINE ASSOCIATIONS AND Ni—Co CONTENTS IN TWO APOLLO 12 MARE BASALTS R.H. HEWINS and J.I. GOLDSTEIN Department of Metallurgy and Materials Science, Lehigh University, Bethlehem, Pa. (USA) Original manuscript received June 6, 1974 Returned to author for revision July 18, 1974 Revised version received September 18, 1974 of Olivine crystals in mare basalts 12004,8 and 12022,12 are normally zoned with Cr-poor rims. The Ni content increases. rare 2—10-jim metal inclusions in olivine decreases markedly as Fe/Mg in their immediate olivine hosts The frac Each metal grain appears to have been enclosed by late olivine almost immediately after it crystallized. For the tionation trend for the olivine and metal contrasts with the subsolidus equilibration trend for pallasites. Ni microprobe. by detected can be interfaces metal/olivine Co at and Ni Fe, of equilibration basalts, not even local detection, and Co concentrations range from about 300 ppm in olivine cores to about 70 ppm in rims. The limits of of Mg and at 95% confidence, are 36 ppm (Ni) and 25 ppm (Co). The distribution of Ni and Co in olivine, like that Cr, records the depletion of these elements in the melt. concen Fractional solidification models, using the Ni and Co concentrations of the whole rock, and Ni and Co composi trations of the earhest formed olivine, metal and “opaques” as initial compositions, allow metal and olivine be estab can n crystallizatio of order the Conversely known. is n crystallizatio of the if order predicted tions to be and olivine lished if known olivine and metal compositions are reproduced. Calculated Ni and Co contents for metal liquid the until delayed is precipitation metal only where s concentration observed to correspond these basalts in has crystallized 4—5 wt.% olivine.
1. Introduction One of the most interesting differences between mare basalts and terrestrial basalts is the appearance of metallic nickel iron in the lunar rocks. It was shown by Reid et al. [1] that the composition of the metal in some of the Apollo 12 basaltic rocks depends upon its position in the crystallization sequence of the rock. Early metal, specifically that enclosed in olivine crystals, has a high Ni content while later metal has a lower Ni content. This regular variation in Ni content with paragenetic sequence led to the sugges tion of formation of metal by reduction of the silicate magma, with the wide range of Ni concentration in the metal suggesting fairly continuous crystallization of metal. Brett et al. [2] and Taylor et al. [3] recorded Ni concentrations in olivine in 12004 basalt varying from 0.04 to 0.005 wt.%. Brett et al. [2] attempted to determine temperatures from the compositions of co existing olivine and metal as calculated by Buseck and
Goldstein [4] for pallasites using a Prior’s Rule approach. The Ni—Fe partitioning failed to give meaningful temperatures for the basalts, leading to the conclusion that the metal and olivine crystals were not equilibrated in the solid state. The possibility remained, however, that metal and olivine which crystallized simultaneously in the frac tionation sequence could be identified by probe and petrographic methods; hence, a series of decreasing temperatures corresponding to this sequence could possibly be calculated. The well-characterized thinsections 12004,8 and 12022,12 [1,2, 18], two fairly rapidly crystallized but glass-free basalts, have been reexamined. This paper attempts to trace, with the electron probe, the record of equilibrium between magma, olivine and metal as well as local equilibrium between the olivine hosts and the metal inclusions. Points examined include: (1) the temperature of metal nucleation relative to the crystallization of the other phases; (2) the time of metal incorporation by
Authors Copy 60
R.H. HEWINS AND JJ. GOLDSTEIN
olivine; and (3) the nature of the distribution of Ni and Co across olivine crystals and within olivine adjacent to metal inclusions.
2. Analytical methods Metal grains were analyzed by the electron microprobe for Fe, Ni, Co, Cr, and P. Operating conditions were 20 kilovolts and 0.05 microamps sample current, using as standards pure elements, Fe—Ni alloys, and meteoritic schreibersite. Point analyses were made on olivine for Fe, Cr, and Mg at 15 kilovolts and 0.05 microamps sample current. The standards used were meteoritic olivine from pallasites and elemental Cr. The data were reduced with EMPADR VII [5J, supply. ing the average values for minor elements (TiC2, CaO, MnO) in lunar olivines [61 and calculating the silica content assuming stoichiometry. Attempts to analyze Ni and Co in olivine were made under a variety of conditions. In each case the equations developed by Ziebold [7] were used to determine the detectability limits, those concentra
tions statistically greater than zero at the 95% confi dence level. The detectability limit for an element depends on peak/background ratio, count rate, count ing time and number of replicates, the sensitivity im proving as these factors are increased. Buseck and Goldstein [4] analyzed trace elements, including Ni, in olivine and obtained detectability limits as low as 10 ppm at 30 kilovolts and 0.2 microamps sample current. For the samples studied here 15 kilovolts and 0.2 microamps sample current provided stable operating conditions. F our replicates of the NiK
peak and background readings, taken on each side of the peak, were obtained using 100-second counting times. The detectability limits were calculated to be 36 ppm for Ni and 25 ppm for Co. According to Buseck and Goldstein [4], the precision of each mea surement equals the detectability limit. Thus errors of ± 36 ppm Ni and ± 25 ppm Co apply to all the data. Systematic errors in the present analyses are considered negligible, as Springwater olivine used as an internal standard was found to contain 43 ppm
Ni, compared to the accepted value of 50 ppm [4J. Olivine crystals were traversed by step scanning
100,000
0 0
80,000
60,000 METAL
9.000 ±IS N 0 0
8,000
7,000 z
L’588
695
6,000 CORE Fe75
0
ZONE OF INCLUSIONS
20
40
1
CORE Fe72
60
RIM Fe64
80
100
micron,
Fig. 1. Peak and background counts for Ni as a function of position within an olivine crystal (no. 25) in rock 12004,8. The upper curve is Ni peak counts, with backgrounds at 1.588 A and 1.695 A. Note that all three rise near the inclusion because of continuum fluorescence of Ni in the metal. The Fe and Mg content of the olivine is also plotted.
.
Authors Copy 61
METAL—OLI VINE ASSOCIATIONS IN MARE BASALTS
3. Olivine chemistry
in order to analyze areas of olivine adjacent to metal inclusions and far from them. At each point additional Fe, Cr, and Mg data were collected, followed by Ni and Co peak and background readings, so that the trace element data could be fixed accurately in the zonation framework of each crystal. The supplementary major element data were thus obtained at high count rates but, as the standard data were also, errors in the dead time correction are minimal. In Fig. 1 the values for Ni peak intensity and background intensity on each side of the peak, as a function of distance in olivine 25, rock 12004,8, are displayed. The count rate on the NiK0 line is significantly greater than the mean background count rate. The Ni concentration values in this complex crystal, described below, are as high as 300 ppm. It is important to note that as the Fe content and the average atomic number of the olivine goes up, the background intensity of Ni also increases (Fig. 1). Backgrounds for Ti, Cr, Mn and Ca in these zoned crystals are similarly affected [8]. Because of this increase in Ni background with Fe—Mg content it was necessary to measure Ni backgrounds at each point. It was also necessary to record Co background readings at each point, taking them very close to the Co peak in order to avoid interference from the FeK peak whose intensity varied with the zonation of the crystals.
OLIVINE
Metal inclusions in the olivine crystals in rocks 12004 and 12022 are mainly in the size range 2—10 microns. Rare olivine crystals contain as many as 3 metal inclusions usually arranged close to the rim of the crystal. An apparent exception to this is the metal inclusion in olivine 25, 12004, shown in Fig. 1. This grain occurs well within the divine crystal but in a zone of inclusions. Secondary fluorescence effects are not important, as discussed later. The olivine close to the inclusions is as Fe-rich as the rim of the crystal. This zone therefore represents late growth in what was originally perhaps a U-shaped skeletal crystal. The variation of Cr203 content with mol.% forsterite in olivine crystals, with and without metal, in 12004,8 and 12022,12, is shown in Fig. 2. The well-known trend of decreasing wt.% Cr203 with decreasing forsterite content [6, 8], consistent with partitioning of Cr and Mg preferentially from the liquid into the olivine, is apparent. Metal-free olivine crystals in 12004,8 are zoned with mean compositions Fo733, 0.35% Cr2 03 for cores and Fo690, 0.32% Cr203 for rims. Olivine crystals containing metal in clusions are zoned from Fo705, 0.34% Cr203 to Fo540, 0.28% Cr203. The chemistries of the olivine crystals suggest that metal grains were enclosed by CORES
RIMS
0
o
WITH METAL INCLUSION WITHOUT UE’tAL INCLUSION
.40
.40
.20
.20
0 C’
U
I
-
70
70
60
60
-
Fo
Fig. 2. Variation of Cr2O3 content with mol.% Fo in olivine crystals in 12004,8 (left) and 12022,12 (right). Tie lines join core and rim(s) of the same crystal.
Authors Copy 62
R.H. HEWINS AND J.I. GOLDSTEIN
olivine crystals which nucleated later and continued to crystallize longer than others. The metal inclusions occur close to the rims of the olivine grains but al though they appear to have been enclosed relatively late in the crystallization history of each olivine crystal a significant volume of olivine must have grown after enclosure.
• 12004,e 50
012022,12
20 I
z
4. Metal composition The metallic Ni—Fe in the mare basalts occurs as a single phase and contains significantly more Co than the vast majority of meteoritic metal [1], (see also Fig. 7). Fig. 3 is a plot which compares the Ni content of the metal to the Fo content of the host olivine immediately surrounding each inclusion. This figure shows that in both rocks the early metal (20—30% Ni) is surrounded by the most forsteritic olivine (Fo70); whereas the late metal (around 10% Ni) is enclosed by the most Fe-rich olivine (Fo60). On the other hand there is poor correlation between the Co content of the metal and the Fo content of the olivine. As spinel crystallizes along with olivine and takes up Co, this may be a local disequilibrium effect resulting from strong partitioning of Co from the liquid into spinel. More than one metal inclusion was analyzed in four of the olivine crystals. For these crystals lines are drawn between the analyses (Fig. 3) to show the trend from the early to the late inclusions as judged by the chemistry of the surrounding olivine. Again there is a trend from high-Ni metal in more magnesian olivine to low-Ni metal in more Fe-rich olivine. This relation shows that metal grains forming from the silicate melt were enclosed by growing olivine essen tially in the order in which they formed. It is there fore likely that the liquid crystallized olivine alone (or with spinel) over a composition range equivalent to at least a small temperature interval before the crystallization of metal. The composition of the olivine crystallized in this interval is Fo74_69 (12004) and Fo70_69 (12022). It is also apparent from Fig. 3, that the fractionation trend relating metal and olivine com positions in these basalts is very different from the pallasite equilibrium trend taken from Buseck and Goldstein [4]. This evidence that equilibration of the sub-solidus olivine—metal assemblage did not occur
PALLASITE
z
TREND
10
0
eo
-
70 Fo
Fig. 3. Variation of the Ni content of metal inclusions in olivine with the Fo content of their immediate olivine hosts. Data are for five olivine crystals in 12004 and four in 12022. Tie lines join metal grains within the same olivine crystal. Pallasite data 141 are added.
is consistent with experimental evidence that 12022 represents liquid + olivine + spinel chilled very rapidly from 1170°C [9] and it is inappropriate to use the Prior’s Rule approach to deduce the conditions of equilibration of these olivines. —‘
4
5. Distribution of Ni and Co in olivine
Fig. 4 illustrates the apparent variations of Ni, Co and Fo content in a very small and somewhat asym metrically zoned olivine crystal from rock 12004. Ni and Co concentrations are lowest in the rim which is the most fayalitic. Ni and Co concentrations are rather variable in the remainder of this crystal but appear to be highest close to the metal inclusions. Possible ex planations of the increased Ni and Co intensities in olivine close to the metal include Ni metal—silicate equilibration in the solid state and fluorescence of the Ni in the metal by the X-ray continuum of the olivine as the electron beam approaches the metal in clusion. If an olivine crystal, containing a Ni-rich metal particle reduced from the silicate melt, cooled slowly, it might undergo local equilibration, thus raising the Ni concentration in the olivine close to the metal in clusion. Ni equilibration between metal and olivine
.
Authors Copy 63
METAL—OLIVINE ASSOCIATIONS IN MARE BASALTS TABLE 1 Average compositions of olivine grains containing metal in clusions
-70
t
Fo
METAL_\
Fo
Ni (ppm,
2004,8 OLIVINE 27
Olivine 25
core 71.3 rim 65.9
312 207
245 207
Olivine23
core69.1 rim 62.6
179 68
185 108
Olivine 27
core 68.3 rim 66.8
240 64
184 76
300 200 Co,
eel.
ppm
Co (ppm,
36)
12004,8
±
±
25)
6
C
100
•4
0 300
41
200 Ni, ppm
44
100
4 40
20 0 DISTANCE
microns
fig. 4. F o, Ni and Co variation across a small metal-bearing zoned olivine grain (no. 27), 12004,8. High Ni and Co in tensities adjacent to the metal are due to fluorescence of these elements in the metal by continuum X-rays from the olivine. The error bars in Figs. 4—6 are ± the detectability limit.
has certainly been demonstrated for the very slowly cooled pallasites [4] but appears unlikely in these rapidly crystallized rocks. To invesligate further the possibility of local equi libration, an olivine/metal interface from an equi librated pallasite, Springwater, was examined with the probe. fig. 5 shows the Ni content of Springwater
olivine as a function of distance. No composition gradients were observed and the Ni content is much less than the lunar olivine. This is consistent with sub solidus olivine—metal equilibration as discussed by Buseck and Goldstein [4]. It should be noted that Ni concentration appears to increase within 5—10 microns of the interface between olivine and metal (fig. 5). In the case of this pallasite, there is no reason for the Ni content in the olivine at the metal/olivine interface to rise. The apparent Ni increase at the interface is due to continuum fluorescence effects comparable to those found at an fe/Ni interface by Goldstein and Ogilvie [10]. Trace element data collected very close to metal grains are therefore ineaningless. The Ni and Co data summarized in Table 1 are averages for the cores and most fayalitic rims of three metal-bearing crystals in rock 12004. The uncertainties quoted (derived from the detectability limits) apply equally to all the data. It is clear that in each grain the averaged Ni and Co concentrations decrease with
400 SPRING WATER OLIVINE
300
300 EE
0
Q
200
ç 200
100
.0 ‘.i
METAL
0
00
. 0
20
40
60
80
100
120
140
microns
Fig. 5. Variation of Ni across part of an olivine crystal in the Springwater pallasite. A metal/olivine interface is shown on the right-hand side of the figure. Ni intensities show secondary fluorescence within 5—10 microns of the interface.
72
64
68
60
Fo
fig. 6. Variation of average Ni and Co concentrations with fo content in cores and rims of olivine crystals in 12004,8.
Authors Copy 64
R.H. HEWINS AND J.I. GOLDSTEIN
forstedte content, i.e., from the core to the rim. This variation of Ni and Co concentration with the fo content of olivine is plotted in fig. 6. It appears that both Ni and Co decrease in a fairly coherent manner as the forsterite content of the crystals decreases. This trend indicates that Ni and Co were partitioned from the silicate liquid into the olivine crystals and that the Ni—Co concentrations in the liquid and in the olivine decrease as crystallization proceeded. Any abrupt drop in Ni concentration at the rims of olivine crystals might reflect the onset of metal crystallization.
6. Ni/Co ratios and their significance The Ni/Co ratios of Fe-rich metallic crystals vary from environment to environment and also within any one environment. This variation in Ni and Co is illustrated for lunar rocks and meteorites in Fig. 7. The corresponding changes in Ni/Co ratio are listed in Table 2, with the maximum Ni/Co ratios corre sponding to the highest Ni and Co contents, i.e. the Ni/Co ratio and the Ni and Co concentrations of metal decrease together. Apollo 12 basalts 12004 and 12022 (III and IV) with fairly continuous reduc tion of metal from silicate melt have been contrasted with Apollo 11 basalts (WI) containing high-fe metal crystallized from immiscible sulfide liquid [1, 14]. 6
iv 4
TABLE 2 Range of Ni/Co ratios in metal from different environments plotted in fig. 8 Metal
Ni/Co
Ni/Co
environment
maximum
minimum
33
12
(I) meteoritic (11) 68416 (III) 12004 (IV) 12022 (V)pyroxenephyric basalt (VI) 78501 basalt (VII) Apollo 11 basalt
±
7
20 12 5 5
—
2 0.1
-
4
Ref.
6 4 2 1
[111 this study this study this study [13]
0 0
112] [11]
±
Apollo 17 basaltic metal (VI) is intermediate between the higher-Co Apollo 12 (IV) or the pyroxene-phyric Apollo 15 (V) and the Apollo 11 types [12, 13]. The plotted Apollo 12 rocks, essentially similar chemically and mineralogically [1], show widely different concen trations of Co in metal over a similar range of Ni con centration (fig. 7). This range of metal compositions is slightly wider than that displayed in the various classes of Apollo 15 basaltic rocks, particularly the feldspathic and olivine microgabbros of Dowty et al. (fig. 8 in [13]). The trends on the Ni—Co diagram may reflect variation in the bulk compositions of the systems in which the metal crystallized or variation in some physical condition such as oxygen fugacity. The influence of oxygen fugacity on metal com position in a metal—oxide—silicate assemblage can be evaluated by considering the reaction: M0=M+ 02
I0
30
% Ni Fig. 7. Range of metal compositions in various environments; I = meteoritic; II = 68416 feldspathic basalt; III = 12004; IV = 12022; V Apollo 15 pyroxene-phyric basalt; VI = 78501 ba salt; VII = Apollo 11 basalt. For sources of data, see Table 2.
a
where M is the metal ion. From thermochemical data (table E in [15]) the changes in free energy in this reaction at 1400°K are 41,120 cal/mole for feO, 32,240 cal/mole for CoO and 25,480 cal/mole for NiO. Thus formation of Fe by reduction is more difficult than Co and formation of Co more difficult than Ni and when metallic Fe is formed by partial reduction, Ni and Co. if present, are concentrated in the metallic phase. An important consequence of this oxidation—reduction relationship is Prior’s Rule that the average Ni content of metal in a meteorite increases as the average fe/Mg ratio of associated silicates increases [16]. TheCo content and
4
Authors Copy 65
METAL—OLIVINE ASSOCIATIONS IN MARE BASALTS
Ni/Co ratio of meteorite metal should also increase as the Fe/Mg ratio of associated silicates increases. Any bulk composition would be expected to give metal of lower Ni/Co ratio and higher Fe content if crystallized at lower oxygen fugacity. The trend towards Ni-poor and Fe-rich compositions in the metal of fractionally crystallized basalts [1, 2] thus appears consistent with normal behavior, a decrease of oxygen fugacity as temperature falls, although the Fe/Mg ratio of coexist ing silicates increases as a result of the fractionation. In the case of rocks, the effects of crystallization of other Ni- and Co-bearing phases on the composition of metal must be considered. The Ni/Co ratios of oxides and sulfides, unlike those of metal, are less than 1 [3], while that of olivine is approximately 1 (this study). One way of evaluating the influence of these phases on the metal composition is by developing and using a fractional solidification model.
7. Fractional solidification model A fractional solidification or Rayleigh depletion model assumes diffusion in the solid after crystalliza tion is too slow to permit reequilibration. It should reproduce the sequences of observed mineral composi tions if diffusion maintained a homogeneous parent liquid [171. The success of Butler [8] in explaining the variation of Ti, Cr, Mn and Ca in these olivines by such a model suggests that disequilibrium in the liquid is negligible. Partition coefficients for components of interest can be calculated after collecting information on the compositions of the liquid and the first formed solid phase. Partition coefficients for Ni and Co are defined as: KNi
CNi (solid) =
CNj (liquid)
and KCO
Co (solid) C0 (liquid)
where C is concentration. The first step in the model is to crystallize an arbitrary amount of the primary liquidus phase with its initial composition. Knowing the mass of Ni and Co remaining in the liquid and the mass of the liquid remaining after this crystalliza tion step, the new concentrations in the liquid can be calculated. The composition of the solid phase after the second stage of crystallization is determined from
the new concentrations of Ni and Co in the liquid and the partition coefficients defined above. At some stage it may be appropriate to introduce a second solid phase whose initial composition is also known from analyses, and calculate its partition coefficient using the liquid composition for this stage. In subsequent crystalliza tion steps the concentrations of Ni and Co in the solid phases and remaining liquids may be calculated. The general expression for the concentration of a compo nent in the nth liquid is:
CflL
=
(Cn_l,L
yC0)/ (i —
where C concentration, y amount of phase, L = liquid, I = I to! are solid phases; and the composition of the solid phases are known from the partition co efficient equation. Clearly these crystallization steps involve many degrees of freedom and it is necessary by trial and error to determine a path producing solid phases whose composition and abundance are realistic when compared to measured data. It is also clear that the partition coefficients are model-dependent; i.e., their estimated values depend on the position of the phase concerned in the assumed order of crystalliza tion. However, this has little influence on the sequence of compositions of this phase because its initial com position and the bulk composition of the system are fixed. Early introduction of a phase, when concentra tions in the liquid are high, produces low estimated partition coefficients; late introduction, when concen trations in the liquid are low, produces high partition coefficients. Thus the effects of increasing partition coefficients as the model is modified are counteracted by lower concentrations in the liquid. For mare basalts 12004 and 12022, the Ni and Co contents of each whole rock sample were taken as the initial concentrations of each liquid. The initial composition of each phase (olivine, metal and “opaques”) was taken as the highest measured Ni and Co concentrations for each phase. Of these assumed initial compositions, that of metal is least certain be cause of the variability of Co content at high Ni con centrations. The initial compositions used for calcula tions on 12004 and 12022 are listed in Table 3 along with the source of the data. It is apparent from these compositions that olivine is a sink for both Ni and Co, metal is a sink for Ni and “opaques” are a sink for Co.
Authors Copy 66
R.H. HEWINS AND LI. GOLDSTEIN
o
TABLE 3 Initial compositions used in solidification models 12004 Ni Liquid
67 ppm
Metal
29.7 wt.%
Olivine Opaque
400 ppm —
12022 Co
Ni
50 ppm 2.37 wt.% 300 ppm —
The other important phases in the rock, plagioclase and pyroxene, appear after the crystallization of the
metal—olivine association and are ignored in these calculations. In any case, crystallization of pyroxene will have little effect on the Ni and Co concentrations of the liquid while plagioclase will tend to concentrate Ni and Co in the liquid, thus retarding but not modify ing to any extent the fractionation trends for metal and olivine. Various solidification models involving differ ent orders of crystallization were considered. The models were tested by comparing the calculated and observed compositions and abundances of the phases. These basalts contain about 16% modal olivine as phenocrysts in a matrix rich in pyroxene but contain ing only 12—20% modal plagioclase as fine laths. Opaque minerals, crystallizing in the order chromite, ulvospinel, ilmenite, troilite, are more abundant in 12022 (11%) than in 12004 (5%). Metal content is estimated at 0.03 vol. ¾. (The modal data are taken from Brett et al. [2J.) The amounts of metal in mare basalts are commonly given as
E
—
METAL—OLIVINE ASSOCIATIONS AND Ni—Co CONTENTS IN TWO APOLLO 12 MARE BASALTS R.H. HEWINS and J.I. GOLDSTEIN Department of Metallurgy and Materials Science, Lehigh University, Bethlehem, Pa. (USA) Original manuscript received June 6, 1974 Returned to author for revision July 18, 1974 Revised version received September 18, 1974 of Olivine crystals in mare basalts 12004,8 and 12022,12 are normally zoned with Cr-poor rims. The Ni content increases. rare 2—10-jim metal inclusions in olivine decreases markedly as Fe/Mg in their immediate olivine hosts The frac Each metal grain appears to have been enclosed by late olivine almost immediately after it crystallized. For the tionation trend for the olivine and metal contrasts with the subsolidus equilibration trend for pallasites. Ni microprobe. by detected can be interfaces metal/olivine Co at and Ni Fe, of equilibration basalts, not even local detection, and Co concentrations range from about 300 ppm in olivine cores to about 70 ppm in rims. The limits of of Mg and at 95% confidence, are 36 ppm (Ni) and 25 ppm (Co). The distribution of Ni and Co in olivine, like that Cr, records the depletion of these elements in the melt. concen Fractional solidification models, using the Ni and Co concentrations of the whole rock, and Ni and Co composi trations of the earhest formed olivine, metal and “opaques” as initial compositions, allow metal and olivine be estab can n crystallizatio of order the Conversely known. is n crystallizatio of the if order predicted tions to be and olivine lished if known olivine and metal compositions are reproduced. Calculated Ni and Co contents for metal liquid the until delayed is precipitation metal only where s concentration observed to correspond these basalts in has crystallized 4—5 wt.% olivine.
1. Introduction One of the most interesting differences between mare basalts and terrestrial basalts is the appearance of metallic nickel iron in the lunar rocks. It was shown by Reid et al. [1] that the composition of the metal in some of the Apollo 12 basaltic rocks depends upon its position in the crystallization sequence of the rock. Early metal, specifically that enclosed in olivine crystals, has a high Ni content while later metal has a lower Ni content. This regular variation in Ni content with paragenetic sequence led to the sugges tion of formation of metal by reduction of the silicate magma, with the wide range of Ni concentration in the metal suggesting fairly continuous crystallization of metal. Brett et al. [2] and Taylor et al. [3] recorded Ni concentrations in olivine in 12004 basalt varying from 0.04 to 0.005 wt.%. Brett et al. [2] attempted to determine temperatures from the compositions of co existing olivine and metal as calculated by Buseck and
Goldstein [4] for pallasites using a Prior’s Rule approach. The Ni—Fe partitioning failed to give meaningful temperatures for the basalts, leading to the conclusion that the metal and olivine crystals were not equilibrated in the solid state. The possibility remained, however, that metal and olivine which crystallized simultaneously in the frac tionation sequence could be identified by probe and petrographic methods; hence, a series of decreasing temperatures corresponding to this sequence could possibly be calculated. The well-characterized thinsections 12004,8 and 12022,12 [1,2, 18], two fairly rapidly crystallized but glass-free basalts, have been reexamined. This paper attempts to trace, with the electron probe, the record of equilibrium between magma, olivine and metal as well as local equilibrium between the olivine hosts and the metal inclusions. Points examined include: (1) the temperature of metal nucleation relative to the crystallization of the other phases; (2) the time of metal incorporation by
Authors Copy 60
R.H. HEWINS AND JJ. GOLDSTEIN
olivine; and (3) the nature of the distribution of Ni and Co across olivine crystals and within olivine adjacent to metal inclusions.
2. Analytical methods Metal grains were analyzed by the electron microprobe for Fe, Ni, Co, Cr, and P. Operating conditions were 20 kilovolts and 0.05 microamps sample current, using as standards pure elements, Fe—Ni alloys, and meteoritic schreibersite. Point analyses were made on olivine for Fe, Cr, and Mg at 15 kilovolts and 0.05 microamps sample current. The standards used were meteoritic olivine from pallasites and elemental Cr. The data were reduced with EMPADR VII [5J, supply. ing the average values for minor elements (TiC2, CaO, MnO) in lunar olivines [61 and calculating the silica content assuming stoichiometry. Attempts to analyze Ni and Co in olivine were made under a variety of conditions. In each case the equations developed by Ziebold [7] were used to determine the detectability limits, those concentra
tions statistically greater than zero at the 95% confi dence level. The detectability limit for an element depends on peak/background ratio, count rate, count ing time and number of replicates, the sensitivity im proving as these factors are increased. Buseck and Goldstein [4] analyzed trace elements, including Ni, in olivine and obtained detectability limits as low as 10 ppm at 30 kilovolts and 0.2 microamps sample current. For the samples studied here 15 kilovolts and 0.2 microamps sample current provided stable operating conditions. F our replicates of the NiK
peak and background readings, taken on each side of the peak, were obtained using 100-second counting times. The detectability limits were calculated to be 36 ppm for Ni and 25 ppm for Co. According to Buseck and Goldstein [4], the precision of each mea surement equals the detectability limit. Thus errors of ± 36 ppm Ni and ± 25 ppm Co apply to all the data. Systematic errors in the present analyses are considered negligible, as Springwater olivine used as an internal standard was found to contain 43 ppm
Ni, compared to the accepted value of 50 ppm [4J. Olivine crystals were traversed by step scanning
100,000
0 0
80,000
60,000 METAL
9.000 ±IS N 0 0
8,000
7,000 z
L’588
695
6,000 CORE Fe75
0
ZONE OF INCLUSIONS
20
40
1
CORE Fe72
60
RIM Fe64
80
100
micron,
Fig. 1. Peak and background counts for Ni as a function of position within an olivine crystal (no. 25) in rock 12004,8. The upper curve is Ni peak counts, with backgrounds at 1.588 A and 1.695 A. Note that all three rise near the inclusion because of continuum fluorescence of Ni in the metal. The Fe and Mg content of the olivine is also plotted.
.
Authors Copy 61
METAL—OLI VINE ASSOCIATIONS IN MARE BASALTS
3. Olivine chemistry
in order to analyze areas of olivine adjacent to metal inclusions and far from them. At each point additional Fe, Cr, and Mg data were collected, followed by Ni and Co peak and background readings, so that the trace element data could be fixed accurately in the zonation framework of each crystal. The supplementary major element data were thus obtained at high count rates but, as the standard data were also, errors in the dead time correction are minimal. In Fig. 1 the values for Ni peak intensity and background intensity on each side of the peak, as a function of distance in olivine 25, rock 12004,8, are displayed. The count rate on the NiK0 line is significantly greater than the mean background count rate. The Ni concentration values in this complex crystal, described below, are as high as 300 ppm. It is important to note that as the Fe content and the average atomic number of the olivine goes up, the background intensity of Ni also increases (Fig. 1). Backgrounds for Ti, Cr, Mn and Ca in these zoned crystals are similarly affected [8]. Because of this increase in Ni background with Fe—Mg content it was necessary to measure Ni backgrounds at each point. It was also necessary to record Co background readings at each point, taking them very close to the Co peak in order to avoid interference from the FeK peak whose intensity varied with the zonation of the crystals.
OLIVINE
Metal inclusions in the olivine crystals in rocks 12004 and 12022 are mainly in the size range 2—10 microns. Rare olivine crystals contain as many as 3 metal inclusions usually arranged close to the rim of the crystal. An apparent exception to this is the metal inclusion in olivine 25, 12004, shown in Fig. 1. This grain occurs well within the divine crystal but in a zone of inclusions. Secondary fluorescence effects are not important, as discussed later. The olivine close to the inclusions is as Fe-rich as the rim of the crystal. This zone therefore represents late growth in what was originally perhaps a U-shaped skeletal crystal. The variation of Cr203 content with mol.% forsterite in olivine crystals, with and without metal, in 12004,8 and 12022,12, is shown in Fig. 2. The well-known trend of decreasing wt.% Cr203 with decreasing forsterite content [6, 8], consistent with partitioning of Cr and Mg preferentially from the liquid into the olivine, is apparent. Metal-free olivine crystals in 12004,8 are zoned with mean compositions Fo733, 0.35% Cr2 03 for cores and Fo690, 0.32% Cr203 for rims. Olivine crystals containing metal in clusions are zoned from Fo705, 0.34% Cr203 to Fo540, 0.28% Cr203. The chemistries of the olivine crystals suggest that metal grains were enclosed by CORES
RIMS
0
o
WITH METAL INCLUSION WITHOUT UE’tAL INCLUSION
.40
.40
.20
.20
0 C’
U
I
-
70
70
60
60
-
Fo
Fig. 2. Variation of Cr2O3 content with mol.% Fo in olivine crystals in 12004,8 (left) and 12022,12 (right). Tie lines join core and rim(s) of the same crystal.
Authors Copy 62
R.H. HEWINS AND J.I. GOLDSTEIN
olivine crystals which nucleated later and continued to crystallize longer than others. The metal inclusions occur close to the rims of the olivine grains but al though they appear to have been enclosed relatively late in the crystallization history of each olivine crystal a significant volume of olivine must have grown after enclosure.
• 12004,e 50
012022,12
20 I
z
4. Metal composition The metallic Ni—Fe in the mare basalts occurs as a single phase and contains significantly more Co than the vast majority of meteoritic metal [1], (see also Fig. 7). Fig. 3 is a plot which compares the Ni content of the metal to the Fo content of the host olivine immediately surrounding each inclusion. This figure shows that in both rocks the early metal (20—30% Ni) is surrounded by the most forsteritic olivine (Fo70); whereas the late metal (around 10% Ni) is enclosed by the most Fe-rich olivine (Fo60). On the other hand there is poor correlation between the Co content of the metal and the Fo content of the olivine. As spinel crystallizes along with olivine and takes up Co, this may be a local disequilibrium effect resulting from strong partitioning of Co from the liquid into spinel. More than one metal inclusion was analyzed in four of the olivine crystals. For these crystals lines are drawn between the analyses (Fig. 3) to show the trend from the early to the late inclusions as judged by the chemistry of the surrounding olivine. Again there is a trend from high-Ni metal in more magnesian olivine to low-Ni metal in more Fe-rich olivine. This relation shows that metal grains forming from the silicate melt were enclosed by growing olivine essen tially in the order in which they formed. It is there fore likely that the liquid crystallized olivine alone (or with spinel) over a composition range equivalent to at least a small temperature interval before the crystallization of metal. The composition of the olivine crystallized in this interval is Fo74_69 (12004) and Fo70_69 (12022). It is also apparent from Fig. 3, that the fractionation trend relating metal and olivine com positions in these basalts is very different from the pallasite equilibrium trend taken from Buseck and Goldstein [4]. This evidence that equilibration of the sub-solidus olivine—metal assemblage did not occur
PALLASITE
z
TREND
10
0
eo
-
70 Fo
Fig. 3. Variation of the Ni content of metal inclusions in olivine with the Fo content of their immediate olivine hosts. Data are for five olivine crystals in 12004 and four in 12022. Tie lines join metal grains within the same olivine crystal. Pallasite data 141 are added.
is consistent with experimental evidence that 12022 represents liquid + olivine + spinel chilled very rapidly from 1170°C [9] and it is inappropriate to use the Prior’s Rule approach to deduce the conditions of equilibration of these olivines. —‘
4
5. Distribution of Ni and Co in olivine
Fig. 4 illustrates the apparent variations of Ni, Co and Fo content in a very small and somewhat asym metrically zoned olivine crystal from rock 12004. Ni and Co concentrations are lowest in the rim which is the most fayalitic. Ni and Co concentrations are rather variable in the remainder of this crystal but appear to be highest close to the metal inclusions. Possible ex planations of the increased Ni and Co intensities in olivine close to the metal include Ni metal—silicate equilibration in the solid state and fluorescence of the Ni in the metal by the X-ray continuum of the olivine as the electron beam approaches the metal in clusion. If an olivine crystal, containing a Ni-rich metal particle reduced from the silicate melt, cooled slowly, it might undergo local equilibration, thus raising the Ni concentration in the olivine close to the metal in clusion. Ni equilibration between metal and olivine
.
Authors Copy 63
METAL—OLIVINE ASSOCIATIONS IN MARE BASALTS TABLE 1 Average compositions of olivine grains containing metal in clusions
-70
t
Fo
METAL_\
Fo
Ni (ppm,
2004,8 OLIVINE 27
Olivine 25
core 71.3 rim 65.9
312 207
245 207
Olivine23
core69.1 rim 62.6
179 68
185 108
Olivine 27
core 68.3 rim 66.8
240 64
184 76
300 200 Co,
eel.
ppm
Co (ppm,
36)
12004,8
±
±
25)
6
C
100
•4
0 300
41
200 Ni, ppm
44
100
4 40
20 0 DISTANCE
microns
fig. 4. F o, Ni and Co variation across a small metal-bearing zoned olivine grain (no. 27), 12004,8. High Ni and Co in tensities adjacent to the metal are due to fluorescence of these elements in the metal by continuum X-rays from the olivine. The error bars in Figs. 4—6 are ± the detectability limit.
has certainly been demonstrated for the very slowly cooled pallasites [4] but appears unlikely in these rapidly crystallized rocks. To invesligate further the possibility of local equi libration, an olivine/metal interface from an equi librated pallasite, Springwater, was examined with the probe. fig. 5 shows the Ni content of Springwater
olivine as a function of distance. No composition gradients were observed and the Ni content is much less than the lunar olivine. This is consistent with sub solidus olivine—metal equilibration as discussed by Buseck and Goldstein [4]. It should be noted that Ni concentration appears to increase within 5—10 microns of the interface between olivine and metal (fig. 5). In the case of this pallasite, there is no reason for the Ni content in the olivine at the metal/olivine interface to rise. The apparent Ni increase at the interface is due to continuum fluorescence effects comparable to those found at an fe/Ni interface by Goldstein and Ogilvie [10]. Trace element data collected very close to metal grains are therefore ineaningless. The Ni and Co data summarized in Table 1 are averages for the cores and most fayalitic rims of three metal-bearing crystals in rock 12004. The uncertainties quoted (derived from the detectability limits) apply equally to all the data. It is clear that in each grain the averaged Ni and Co concentrations decrease with
400 SPRING WATER OLIVINE
300
300 EE
0
Q
200
ç 200
100
.0 ‘.i
METAL
0
00
. 0
20
40
60
80
100
120
140
microns
Fig. 5. Variation of Ni across part of an olivine crystal in the Springwater pallasite. A metal/olivine interface is shown on the right-hand side of the figure. Ni intensities show secondary fluorescence within 5—10 microns of the interface.
72
64
68
60
Fo
fig. 6. Variation of average Ni and Co concentrations with fo content in cores and rims of olivine crystals in 12004,8.
Authors Copy 64
R.H. HEWINS AND J.I. GOLDSTEIN
forstedte content, i.e., from the core to the rim. This variation of Ni and Co concentration with the fo content of olivine is plotted in fig. 6. It appears that both Ni and Co decrease in a fairly coherent manner as the forsterite content of the crystals decreases. This trend indicates that Ni and Co were partitioned from the silicate liquid into the olivine crystals and that the Ni—Co concentrations in the liquid and in the olivine decrease as crystallization proceeded. Any abrupt drop in Ni concentration at the rims of olivine crystals might reflect the onset of metal crystallization.
6. Ni/Co ratios and their significance The Ni/Co ratios of Fe-rich metallic crystals vary from environment to environment and also within any one environment. This variation in Ni and Co is illustrated for lunar rocks and meteorites in Fig. 7. The corresponding changes in Ni/Co ratio are listed in Table 2, with the maximum Ni/Co ratios corre sponding to the highest Ni and Co contents, i.e. the Ni/Co ratio and the Ni and Co concentrations of metal decrease together. Apollo 12 basalts 12004 and 12022 (III and IV) with fairly continuous reduc tion of metal from silicate melt have been contrasted with Apollo 11 basalts (WI) containing high-fe metal crystallized from immiscible sulfide liquid [1, 14]. 6
iv 4
TABLE 2 Range of Ni/Co ratios in metal from different environments plotted in fig. 8 Metal
Ni/Co
Ni/Co
environment
maximum
minimum
33
12
(I) meteoritic (11) 68416 (III) 12004 (IV) 12022 (V)pyroxenephyric basalt (VI) 78501 basalt (VII) Apollo 11 basalt
±
7
20 12 5 5
—
2 0.1
-
4
Ref.
6 4 2 1
[111 this study this study this study [13]
0 0
112] [11]
±
Apollo 17 basaltic metal (VI) is intermediate between the higher-Co Apollo 12 (IV) or the pyroxene-phyric Apollo 15 (V) and the Apollo 11 types [12, 13]. The plotted Apollo 12 rocks, essentially similar chemically and mineralogically [1], show widely different concen trations of Co in metal over a similar range of Ni con centration (fig. 7). This range of metal compositions is slightly wider than that displayed in the various classes of Apollo 15 basaltic rocks, particularly the feldspathic and olivine microgabbros of Dowty et al. (fig. 8 in [13]). The trends on the Ni—Co diagram may reflect variation in the bulk compositions of the systems in which the metal crystallized or variation in some physical condition such as oxygen fugacity. The influence of oxygen fugacity on metal com position in a metal—oxide—silicate assemblage can be evaluated by considering the reaction: M0=M+ 02
I0
30
% Ni Fig. 7. Range of metal compositions in various environments; I = meteoritic; II = 68416 feldspathic basalt; III = 12004; IV = 12022; V Apollo 15 pyroxene-phyric basalt; VI = 78501 ba salt; VII = Apollo 11 basalt. For sources of data, see Table 2.
a
where M is the metal ion. From thermochemical data (table E in [15]) the changes in free energy in this reaction at 1400°K are 41,120 cal/mole for feO, 32,240 cal/mole for CoO and 25,480 cal/mole for NiO. Thus formation of Fe by reduction is more difficult than Co and formation of Co more difficult than Ni and when metallic Fe is formed by partial reduction, Ni and Co. if present, are concentrated in the metallic phase. An important consequence of this oxidation—reduction relationship is Prior’s Rule that the average Ni content of metal in a meteorite increases as the average fe/Mg ratio of associated silicates increases [16]. TheCo content and
4
Authors Copy 65
METAL—OLIVINE ASSOCIATIONS IN MARE BASALTS
Ni/Co ratio of meteorite metal should also increase as the Fe/Mg ratio of associated silicates increases. Any bulk composition would be expected to give metal of lower Ni/Co ratio and higher Fe content if crystallized at lower oxygen fugacity. The trend towards Ni-poor and Fe-rich compositions in the metal of fractionally crystallized basalts [1, 2] thus appears consistent with normal behavior, a decrease of oxygen fugacity as temperature falls, although the Fe/Mg ratio of coexist ing silicates increases as a result of the fractionation. In the case of rocks, the effects of crystallization of other Ni- and Co-bearing phases on the composition of metal must be considered. The Ni/Co ratios of oxides and sulfides, unlike those of metal, are less than 1 [3], while that of olivine is approximately 1 (this study). One way of evaluating the influence of these phases on the metal composition is by developing and using a fractional solidification model.
7. Fractional solidification model A fractional solidification or Rayleigh depletion model assumes diffusion in the solid after crystalliza tion is too slow to permit reequilibration. It should reproduce the sequences of observed mineral composi tions if diffusion maintained a homogeneous parent liquid [171. The success of Butler [8] in explaining the variation of Ti, Cr, Mn and Ca in these olivines by such a model suggests that disequilibrium in the liquid is negligible. Partition coefficients for components of interest can be calculated after collecting information on the compositions of the liquid and the first formed solid phase. Partition coefficients for Ni and Co are defined as: KNi
CNi (solid) =
CNj (liquid)
and KCO
Co (solid) C0 (liquid)
where C is concentration. The first step in the model is to crystallize an arbitrary amount of the primary liquidus phase with its initial composition. Knowing the mass of Ni and Co remaining in the liquid and the mass of the liquid remaining after this crystalliza tion step, the new concentrations in the liquid can be calculated. The composition of the solid phase after the second stage of crystallization is determined from
the new concentrations of Ni and Co in the liquid and the partition coefficients defined above. At some stage it may be appropriate to introduce a second solid phase whose initial composition is also known from analyses, and calculate its partition coefficient using the liquid composition for this stage. In subsequent crystalliza tion steps the concentrations of Ni and Co in the solid phases and remaining liquids may be calculated. The general expression for the concentration of a compo nent in the nth liquid is:
CflL
=
(Cn_l,L
yC0)/ (i —
where C concentration, y amount of phase, L = liquid, I = I to! are solid phases; and the composition of the solid phases are known from the partition co efficient equation. Clearly these crystallization steps involve many degrees of freedom and it is necessary by trial and error to determine a path producing solid phases whose composition and abundance are realistic when compared to measured data. It is also clear that the partition coefficients are model-dependent; i.e., their estimated values depend on the position of the phase concerned in the assumed order of crystalliza tion. However, this has little influence on the sequence of compositions of this phase because its initial com position and the bulk composition of the system are fixed. Early introduction of a phase, when concentra tions in the liquid are high, produces low estimated partition coefficients; late introduction, when concen trations in the liquid are low, produces high partition coefficients. Thus the effects of increasing partition coefficients as the model is modified are counteracted by lower concentrations in the liquid. For mare basalts 12004 and 12022, the Ni and Co contents of each whole rock sample were taken as the initial concentrations of each liquid. The initial composition of each phase (olivine, metal and “opaques”) was taken as the highest measured Ni and Co concentrations for each phase. Of these assumed initial compositions, that of metal is least certain be cause of the variability of Co content at high Ni con centrations. The initial compositions used for calcula tions on 12004 and 12022 are listed in Table 3 along with the source of the data. It is apparent from these compositions that olivine is a sink for both Ni and Co, metal is a sink for Ni and “opaques” are a sink for Co.
Authors Copy 66
R.H. HEWINS AND LI. GOLDSTEIN
o
TABLE 3 Initial compositions used in solidification models 12004 Ni Liquid
67 ppm
Metal
29.7 wt.%
Olivine Opaque
400 ppm —
12022 Co
Ni
50 ppm 2.37 wt.% 300 ppm —
The other important phases in the rock, plagioclase and pyroxene, appear after the crystallization of the
metal—olivine association and are ignored in these calculations. In any case, crystallization of pyroxene will have little effect on the Ni and Co concentrations of the liquid while plagioclase will tend to concentrate Ni and Co in the liquid, thus retarding but not modify ing to any extent the fractionation trends for metal and olivine. Various solidification models involving differ ent orders of crystallization were considered. The models were tested by comparing the calculated and observed compositions and abundances of the phases. These basalts contain about 16% modal olivine as phenocrysts in a matrix rich in pyroxene but contain ing only 12—20% modal plagioclase as fine laths. Opaque minerals, crystallizing in the order chromite, ulvospinel, ilmenite, troilite, are more abundant in 12022 (11%) than in 12004 (5%). Metal content is estimated at 0.03 vol. ¾. (The modal data are taken from Brett et al. [2J.) The amounts of metal in mare basalts are commonly given as
