| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Journal of Petrology Volume 42 Number 12 Pages 2215-2229 2001
© Oxford University Press 2001
The Spatial Distribution of Garnets and Pyroxenes in Mantle Peridotites: Pressure–Temperature History of Peridotites from the Kaapvaal Craton
DEPARTMENT OF EARTH, ATMOSPHERIC AND PLANETARY SCIENCES, MASSACHUSETTS INSTITUTE OF TECHNOLOGY, CAMBRIDGE, MA 02139, USA
Received August 22, 2000; Revised typescript accepted May 30, 2001
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONSAMPLE SELECTIONTEXTURAL ANALYSIS METHODRESULTS AND INTERPRETATIONCONCLUSIONSREFERENCES |
|---|
We present a new method for textural analysis of mineral associations that uses digital back-scattered electron and X-ray images obtained with the electron microprobe to determine the spatial properties of minerals on a two-dimensional surface of the rock at different scale lengths. We determine modal amounts and average grain sizes of each mineral in the thin section without resorting to ellipsoidal approximations of grain boundaries, and investigate the spatial relationship of mineral pairs. The method is used to characterize nine mantle xenoliths exhumed within kimberlite pipes in South Africa and to test whether the pyroxenes are spatially correlated with the garnets. The spatial association of these minerals is used to develop a model for the evolutionary history of the Kaapvaal peridotites. The observed distributions can be explained by a two-stage model. In stage 1, harzburgitic residues are produced by large extents of partial melting at shallow depths (
60–90 km) and high temperatures (1300–1400°C). The melting process leading to this depletion occurs in the garnet stability field where garnet, clinopyroxene and olivine are consumed and orthopyroxene and liquid are produced. The Kaapvaal sample suite shows modal and compositional variations consistent with a progressive melt depletion event. In stage 2, the residuum is dragged down to greater depths by mantle corner flow adjacent to a subducted slab. The most depleted harzburgites descend to 140–160 km depth and are cooled. The least depleted harzburgites end up at shallower depths. The resulting stratigraphy is the opposite of what would be expected for a preserved mantle melt column and is consistent with inversion of the melt column as it was dragged around the wedge corner and cooled by the subducted slab. The cooling process causes clinopyroxene and garnet to exsolve from the orthopyroxene. Therefore, the depleted cratonic peridotites of the Kaapvaal preserve a temperature–pressure path consistent with an origin in an Archean subduction zone. KEY WORDS: craton; mantle xenoliths; microprobe; P–T history; spatial analysis; peridotite
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONSAMPLE SELECTIONTEXTURAL ANALYSIS METHODRESULTS AND INTERPRETATIONCONCLUSIONSREFERENCES |
|---|
Mantle xenoliths exhumed within kimberlite pipes in South Africa may provide clues to the formation and evolution of the depleted mantle that makes up the Kaapvaal craton. One hypothesis that has been proposed is that the peridotite originally resides deep in the mantle (at depths of perhaps 300 or 400 km) and is transported upwards into the cratonic lithosphere (
180 km depth), where re-equilibration occurs and clinopyroxene (cpx) unmixes from the garnet phase (Haggerty & Sautter, 1990
). An alternative idea is that the garnet lherzolites were originally high-temperature harzburgites that originated at depths between 100 and 250 km and that both the cpx and garnet subsequently exsolved from the Al-rich orthopyroxene (opx) when the rocks cooled and re-equilibrated (Cox et al., 1987). If exsolution of minerals has indeed occurred, then a spatial relationship between grains of minerals that were formerly dissolved in each other should be apparent in the present-day xenoliths. In particular, the first model implies a spatial relationship between cpx and garnet for rocks of pyroxenite or eclogitic compositions, whereas the second implies that cpx, opx and garnet should be correlated.
The question of whether two minerals are spatially related arises frequently in petrologic studies. Despite the availability of digital images of thin sections, textural studies described in the literature are typically crystal-oriented rather than pixel-oriented (Cashman & Ferry, 1988; Kretz, 1993; Jerram et al., 1996). These analyses generally take a ‘nearest neighbor approach’, measuring the distance between the center of a grain and the center of its nearest neighbor, to determine whether the grains are randomly distributed, clustered, or ordered (Kretz, 1969, 1993; Carlson et al., 1995; Jerram et al., 1996; Miyake, 1998). The most sophisticated of these involve fitting ellipsoids to a digital image of the grains, whereas others simply project the thin section on a screen and trace the crystals by hand. Recently, studies of digital images have used Markov-chain analysis to look for the same sort of clustering or anti-clustering of minerals (Kruse & Stünitz, 1999). A prior spatial analysis of Kaapvaal peridotites, by Cox et al. (1987), specifically tested the hypothesis that two different minerals were spatially associated and found a strong correlation between the presence of garnet and cpx with opx, but their method required labor-intensive manual outlining, identification and counting of crystals.
We describe a new technique for textural analysis in which the digital image of a thin section is statistically analyzed over a range of spatial scales. By analyzing the pixel data, we determine the modal amounts and average crystal size for each mineral type and examine whether any of the minerals are more closely associated spatially than would be expected if they were randomly distributed. This method may be applicable to different textural studies in which average crystal sizes need to be determined or the relationship between mineral types, and in particular their spatial relation as a result of metamorphism or metasomatism, needs to be assessed. We apply this statistical method to nine mantle xenoliths exhumed within kimberlite pipes in South Africa to determine which, if any, of the constituent minerals are correlated. In this way, we test the two hypotheses regarding the history of South African mantle xenoliths and, by implication, the structure and evolution of the Kaapvaal craton.
|
SAMPLE SELECTION |
|---|
|
TOPABSTRACTINTRODUCTIONSAMPLE SELECTIONTEXTURAL ANALYSIS METHODRESULTS AND INTERPRETATIONCONCLUSIONSREFERENCES |
|---|
We selected three garnet peridotite xenoliths from the Bulfontein kimberlite pipes (Kimberley, South Africa), four from the Jagersfontein pipes (130 km SSE of Kimberley), one from the Premier Pipes (25 km NE of Pretoria, South Africa) and one from the Letseng pipes in northern Lesotho (Fig. 1). Several criteria were applied in selecting the sample suite. All samples contained dominantly olivine and opx with the coarse texture that characterizes low-temperature peridotites (Boyd, 1987
). Peridotite nodules (25 cm x 25 cm x 10 cm) were cut into multiple slabs (3–5 slabs each 1 cm thick) and samples were chosen from nodules that had a uniform distribution of garnet and cpx in all slabs. Samples that contained clots of garnet and cpx were avoided, as were samples that contained veins of phlogopite or samples that were serpentinized. We also avoided samples that appeared to have undergone metasomatic modification. A suite of representative low-temperature peridotites that spanned the range of modal olivine percentages observed by Boyd (1989) were supplied by F. R. Boyd.
|
Mineral compositions and modal analysis
From each slab we prepared 2 inch x 3 inch polished thin sections for chemical and modal analysis. We obtained a back-scattered electron image (brightness corresponding to atomic number) as well as X-ray concentration maps of Si, Al, Ca and Mg) of each thin section using the four wavelength-dispersive spectrometer-JEOL JXA-733 Superprobe electron microprobe at MIT. We used an accelerating voltage of 15 kV, a beam current of 100 nA and a dwell time of 5 ms per spot. By comparing the various images, we were able to identify each pixel in the image as olivine, orthopyroxene, clinopyroxene or garnet. Eight of the samples we mapped and analyzed are shown at the same scale in Fig. 2. The image resolution ranged between 62 and 200 µm/pixel and the areas imaged were approximately 3 cm x 5 cm. Depending on the image resolution and the area mapped, the image collection time varied between 12 and 16 h per sample.
|
Chemical compositions
Chemical compositions of the minerals were obtained for five of the samples with the same electron microprobe at MIT using wavelength-dispersive spectrometry. Phases were analyzed at an accelerating voltage of 15 kV, a beam current of 10 nA, a beam diameter of 1 µm and typical counting times between 20 and 40 s per element. Data were reduced with the CITZAF program (Armstrong, 1995) using the atomic number correction of Duncumb & Reed, Heinrich’s tabulation of mass absorption coefficients and the fluorescence correction of Reed. A minimum of 10 measurements were made for each mineral on every thin section. The average and standard deviation of those measurements are summarized in Table 2. Mineral compositions for the other four samples were provided by F. R. Boyd of the Geophysical Laboratory at the Carnegie Institution of Washington.
|
|
Classification as low-temperature peridotites
Garnet peridotites are broadly classified as either low- or high-temperature in origin, based on estimated equilibration temperatures (Boyd, 1987, 1989). Xenoliths classified as low-temperature have been interpreted as representing lithospheric samples from the upper mantle, originating from depths of <150 km, whereas the high-temperature xenoliths have been interpreted as representing asthenospheric samples from depths of 175 km and greater (Nixon & Boyd, 1973; Harte, 1983; Boyd, 1987, 1989; Finnerty & Boyd, 1987; Nixon, 1987; Brey & Köhler, 1990; Herzberg, 1993). The high-temperature xenoliths tend to have a sheared or deformed texture (Nixon & Boyd, 1973; Bouillier & Nicolas, 1975; Harte, 1977, 1983; Boyd, 1987; Herzberg, 1993) making them poor candidates for a spatial study such as ours, whereas the low-temperature xenoliths have presumably retained their original spatial configuration, making them ideal candidates for a study of unusual mineral clustering.
Table 1 lists the modal proportions and olivine mg-numbers [atomic MgO/(FeO + MgO)] for the samples we analyzed with low- and high-temperature averages for comparison, and Table 2 gives the mineral compositions. All nine xenoliths have olivine mg-numbers >91·5, a coarse, granular texture with no preferred dimensional orientation or other obvious signs of shearing, and estimated equilibration temperatures of between 800 and 1100°C and pressures of 30–55 kbar [Table 1, using the thermometer of Brey & Köhler (1990)] consistent with a classification of low-temperature peridotites (Nixon & Boyd, 1973; Harte et al., 1975; Harte, 1983; Boyd, 1987, 1989; Finnerty & Boyd, 1987). In addition, the bulk-rock, major element geochemistry of the samples is typical of low-temperature peridotites from South Africa, which tend to be depleted in FeO, TiO2, CaO and Al2O3 relative to higher-temperature xenoliths (Nixon & Boyd, 1973; Harte 1977, 1983; Nixon, 1987; Boyd, 1989; Herzberg, 1993).
The oxide compositions of our samples are also consistent with low-temperature xenoliths. Figure 3 shows the weight per cent oxides for the individual minerals of the xenoliths normalized to the average for the corresponding minerals of 13 low-temperature garnet lherzolites from southern Africa analyzed in a previous study by Cox et al. (1987). One of the rocks (JAG90-72) has 2–3 times as much TiO2 as the average 13 low-temperature rocks do; however, it is still within the range of what is found for low-temperature xenoliths (Boyd, 1987). The only way in which some of the rocks we analyzed differ from the low-temperature classification is in their modal concentrations of opx and cpx. Three of the rocks (JAG90-72, JAG93-8 and K12) contain unusually low amounts of opx and one (JAG90-72) has very high amounts of cpx, much more akin to high-temperature xenoliths than low-temperature ones. Except for this variability in modal amounts, these rocks are, otherwise, very typical of low-temperature garnet peridotites from kimberlite pipes in South Africa. Moreover, they span almost the entire range of modal olivine and olivine mg-numbers that is found in the Kaapvaal craton xenoliths (Fig. 4) so they should provide a representative and potentially contrasting set of low-temperature samples.
|
|
) by Boyd (1987)
|
TEXTURAL ANALYSIS METHOD |
|---|
|
TOPABSTRACTINTRODUCTIONSAMPLE SELECTIONTEXTURAL ANALYSIS METHODRESULTS AND INTERPRETATIONCONCLUSIONSREFERENCES |
|---|
To test for a preferential spatial correlation between two minerals, we employ a box-counting method (described in the spatial analysis section below) on a series of different scales, to determine some of the statistical properties of the rock. We compare these results with those obtained from analysis of computer-simulated rocks that contain randomly distributed crystals with the same modal proportions and average grain sizes as determined in our analysis of the real rock. In this way, we can test whether there is a statistically unusual spatial association between any of the mineral pairs in the real rocks.
Spatial analysis
We divide the digital image of the thin section into squares. We choose the square size to have a total number of pixels Nn = 22n, where n is initially the largest possible integer allowed by the image size. Figure 5 shows the back-scattered electron image and X-ray maps for an example thin section (FRB347) as well as the final digital image (the final identification of each pixel as opx, cpx, garnet or olivine). In this case, we begin with n = 7 (corresponding to a square 128 pixels by 128 pixels in size with N7 = 16 384) and four such squares. We then calculate the number of squares that contain a large fraction of each of the minerals. To make this determination, we define a correlation condition
 |
is the fraction of the box the crystal must occupy. In our analysis, we chose = 0·75, 0·80 and 0·875.
|
Figure 6 shows a plot of these results on a ternary diagram for the case of a three-mineral system; however, the garnet–opx–cpx–olivine system has four components, so in our analyses we actually plot the results on a tetrahedron (not shown) that is composed of four ternary diagrams joined on each side. The distance from each corner is equivalent to [1 – (nx/Nn)] and plots as a line near each of the three corners defining a region we call a. The boxes that satisfy the condition defined in equation (1) are represented as points that lie within the corner regions a.
|
We then count the number of boxes that contain a large proportion of any two minerals. These points plot in the sidebars (regions b). The correlation condition for this determination is
 |
As n increases, the percentage of boxes plotted in each of the corners Pnx will gradually decrease. When the box size exceeds the average crystal size, Pnx will drop abruptly. A plot of Pnx as a function of box size (Fig. 7) will show a break in slope, which indicates that this critical size has been reached. For equant minerals, this size will correspond to the average crystal size 
. In some cases, the break in slope is less evident and so we used the additional criterion Pm 
0·25Mx, where Mx is the modal proportion of mineral x.
|
Given the above results, we can create a thin section of a synthetic rock containing the same modal proportions and average grain sizes as the real rock, but with randomly distributed crystals. In this way, the synthetic rocks match the real rocks for n = 0, and the average number appearing in each of the ternary diagram corners is the same as the real rock for all n. We then test whether significantly more points occur in the sidebars of the real rocks than in the randomly distributed synthetic rocks.
Synthetic rock calculations
In general, average crystal sizes differ between minerals in the rocks we examined. Therefore, we create the synthetic thin section in two stages. First, we lay out a grid of the same size as the largest average crystal size. Because the larger crystals will fill the boxes completely whereas the smaller ones will not, we renormalize the modal proportions to account for the various size differences. The renormalized modal amount for each mineral is
 |
is the average crystal size of mineral x. Using a random number generator, we fill the boxes according to the Rx proportions with the appropriate sizes nx. This will produce a partially filled rock with the same modal proportions as the real rock. The second stage is to go back and fill the remaining blank spaces according to the modal proportions Mx of the rock. In this way, the final rock produced has appropriate modal proportions and representative crystal sizes, and the minerals are randomly distributed. Figure 8 shows an example of a synthetic rock that was generated according to this scheme (both stages of its creation) and the real rock it represents.
|
We create 1000 synthetic thin sections for each of the real rocks imaged and then analyze each of the synthetic rocks in exactly the same manner as the real rock. Therefore, for each n, we obtain a distribution of synthetic results for the three corners of the ternary diagram and the three sidebars. Because we have created the synthetic rocks using the average crystal sizes, the results plotting in the corners of the ternary diagram for the real rock are well within the range of 95% of the results obtained for the synthetic rocks for all n. We focus on the sidebars to see if statistically significant differences arise. For example, when examining the garnet–pyroxene bar, we look at the ternary diagram for box size n corresponding to the for garnet and pyroxene. If differs for the two minerals, we take the average of the two and if they differ by just one order of magnitude, then we consider both . When the value in the sidebar of the real rock exceeds 95% of the expected range found in the synthetic rocks, we conclude that the two minerals are spatially associated.
|
RESULTS AND INTERPRETATION |
|---|
|
TOPABSTRACTINTRODUCTIONSAMPLE SELECTIONTEXTURAL ANALYSIS METHODRESULTS AND INTERPRETATIONCONCLUSIONSREFERENCES |
|---|
We found statistically significant relationships between cpx, opx and garnet in eight of the samples and between cpx and opx in the remaining sample, as summarized in Table 3. On the basis of these spatial relationships, it is possible that eight of the nine xenoliths we analyzed (JAG90-72, JAG93-8, UX497, K1, FRB347, FRB850, FRB1350 and FRB4265) were originally two-phase rocks consisting of primary olivine and opx (harzburgites) and that later in the sample’s history the garnet and cpx unmixed from the opx. Assuming this history, we can reconstruct what the composition of these ‘original’ orthopyroxenes would have been using the microprobe analyses of the weight per cent oxides and relative densities of the garnet, cpx and opx.
|
There are several sources of error in making such a reconstruction, including errors in the chemical analyses and errors in the estimates of modal abundances. We have calculated formal errors for the measurements of weight per cent oxides and for the modal amounts imaged in the thin sections and found both to be negligible. The largest source of error is in the assumption that the modal amounts contained in the particular slice of rock we analyzed are representative of the bulk rock. To estimate the size of this error, we subdivide the thin sections into 10 strips and calculate the modal abundances of each mineral for every strip. We then use a Student’s t distribution to determine the error bounds on the modal amounts with 95% confidence. Table 4 lists the chemical compositions of the reconstructed orthopyroxenes and the average estimated errors in these reconstructions given our modal error estimates.
|
The reconstructed Kaapvaal orthopyroxenes are chemically similar to the orthopyroxenes that are produced in both anhydrous and hydrous peridotite melting experiments. In Fig. 9a the compositions of pyroxenes from liquids saturated with a harzburgite (oliv + opx) residue are plotted along with the Kaapvaal reconstructed orthopyroxenes. The anhydrous melting experiments span a range of pressures from 2·2 GPa (Parman, 2001) to between 3 and 6 GPa (Walter, 1998). Over this pressure range there is a systematic decrease in the Ca and Al proportions that is dominantly correlated with an increase in melt fraction. The effect of increasing pressure under anhydrous conditions is to increase the amount of melting required to exhaust cpx and garnet from the residual assemblage (Walter, 1998). Therefore, both higher temperatures and higher melt fractions are required to reach oliv + opx saturation at higher pressures. The effect of increasing H2O is to increase the melt fraction and depress the temperature at which high melt fractions are stable with an oliv + opx residue (Parman, 2001). Also shown in Fig. 9 are the orthopyroxenes in equilibrium with olivine and a Barberton komatiite liquid (24 wt % MgO, 6·5 wt % H2O) at 2·4 GPa. The low Al in the opx is a consequence of the higher temperature of oliv + opx saturation in this liquid compared with the hydrous peridotite melts.
|
Most of the Kaapvaal reconstructed orthopyroxenes parallel the trend defined by the experimental data. The compositions of reconstructed pyroxenes from Cox et al. (1987) also parallel the trend defined by our sample suite. The exception is the cpx + garnet-rich sample JAG90-72, which plots at much higher Ca contents than the experimental data. Sample JAG90-72 will not be considered further in the discussion, and we assume that it formed by a different process than the one discussed below. When the remaining reconstructed orthopyroxenes are compared with the experimentally produced orthopyroxenes, the Ca and Al contents of the anhydrous opx compositions match the reconstructed compositions most closely. However, the differences between the hydrous and anhydrous Ca contents are small and our assumption that opx was the only pyroxene present initially in these samples may be artificially inflating the reconstructed Ca contents to higher values than were actually present in the original, high-temperature protolith (see below). There is an inferred 500°C difference between the anyhydous and hydrous orthopyroxenes containing the lowest Al (Fig. 9b), so that temperature variations have a small effect on the pyroxene Ca content. The important first-order observation is that neither an anhydrous nor hydrous melting origin for the Kaapvaal orthopyroxenes can be excluded.
Kinzler & Grove (1999) explored the compositional trends that would be expected in a peridotite residue as it underwent a near-fractional melting event beginning in the garnet stability field. Using experimental data from Kinzler (1997) and Walter (1998), they calculated the reaction coefficients for the melt reaction and the composition of residues expected for melting until the exhaustion of garnet and cpx. Their model extracts melt over the pressure range of 3·7–1·5 GPa and is reproduced in Fig. 10. The plot of modal olivine vs olivine mg-number (Fig. 10a) shows an enlargement of the low-T peridotite data plotted in Fig. 3, and Fig. 10b shows the evolution of the opx, subcalcic cpx and garnet in the residue as melt is extracted. Kinzler & Grove (1999) demonstrated that fractional melting can produce part of the trend that extends through the center of the modal olivine vs mg-number variation (Fig. 10a). The extension of that trend to more olivine-rich compositions is not explained by fractional melting and requires an alternative process (Boyd et al., 1997; Kelemen et al., 1998). Boyd et al. (1997) proposed that the extension of the modal variations to high opx is a consequence of a later-stage metamorphic process. Kinzler & Grove (1999) proposed that olivine enrichment samples are formed by passage of a melt that crystallizes olivine after the fractional melting event has occurred.
|
When the Kinzler & Grove (1999) model predictions are compared with the reconstructed Kaapvaal opx compositions, some interesting systematics emerge (Fig. 10b). The trend of increasing modal opx with increasing degree of melt extraction predicted by those workers is shown as the continuous line. The trend of the more depleted Kaapvaal peridotites (olivine mg-number >92·5) follows the prediction closely. Sample FRB4265 is the exception and it plots on the olivine-poor side of the predicted fractional melting trend (Fig. 10a). The less depleted samples (with olivine mg-number <92) show a trend of increasing modal opx with decreasing mg-number, which is the opposite of the modeled trend. However, the Kinzler & Grove (1999) model predicts that in this part of the mg-number–modal olivine space a sub-calcic cpx is present in the residue. Therefore, our initial assumption that opx was the only residual pyroxene may be incorrect for these low mg-number samples.
The Kaapvaal samples that plot at the low mg-number, low modal percentage olivine end of the Kinzler & Grove (1999) model trend (FRB1009 and FRB1350) plot at the high-Ca and high-Al end of the opx trend in Fig. 9a, consistent with lower extents of melting. It should be noted that the least depleted Kaapvaal opx (UX497) also plots in the appropriate part of mg-number–modal olivine space. The Kaapvaal reconstructed opx with the lowest Ca and Al contents plots at the end of the Kinzler & Grove (1999) model trend, where the most depleted residue would be expected to plot.
Another prediction of the Kinzler & Grove (1999) model is that the fractional melting residue over the range of olivine mg-number spanned by the Kaapvaal peridotites should have contained both opx and cpx as residual phases. Only at the end of the model residue trend (where it coincides with FRB347) would cpx be exhausted from the residue. Therefore, the modeled Ca contents of the reconstructed Kaapvaal orthopyroxenes could be artificially high because some of the cpx that was recombined should have remained apart. Melting experiments suggest that the high-temperature protolith would have contained cpx with 9–10 wt % CaO and 7–9 wt % Al2O3 (Walter, 1998; Parman, 2001). Our statistical model does not test the presence of mixtures of opx + cpx + garnet that exsolved from both high- and low-Ca pyroxenes. It is also possible that the reconstructed residue was more similar to the lower pressure, lower temperature trend predicted by Kinzler & Grove (1999) and can be accounted for by a shallow anhydrous or hydrous melting process.
Another interesting correlation is the depth from which the samples were derived (Fig. 11). The most depleted samples (FRB347 and FRB4265) come from the greatest depths whereas the least depleted samples (FRB1350, FRB1009 and UX497) all come from the shallowest depths. The Cox et al. (1987) sample suite shows a trend that parallels the trend defined by our sample suite, but it does not extend to the low-temperature, undepleted end member defined by our samples. This relationship is not completely followed by all of the samples because undepleted samples JAG93-8 and JAG90-72 have final equilibration depths equivalent to FRB347, but there are no depleted samples in our dataset that come from shallow depths. If these mantle samples were derived from a melt extraction column that was left behind after near-fractional melting, one would expect that the deepest mantle residues should be the most fertile and the shallowest mantle residues would be the most depleted. The opposite correlation is observed.
|
Our conclusions are basically similar to those of Cox et al. (1987), who found a spatial relationship between grains of opx, cpx and garnet in 10 low-temperature xenoliths from the Bulfontein and Mothae pipes and compared the reconstructed opx compositions of those rocks with the Al2O3 and CaO isopleths of Yamada & Takahashi (1984). In their analysis, Cox et al. (1987) concluded that the peridotites were formed as residual products of melting at temperatures near the dry peridotite solidus of Kushiro (1973) in the pressure range of 4–8 GPa. However, a comparison of their 10 reconstructed orthopyroxenes (Fig. 9) with more recent mantle melting experiments (Walter, 1998; Parman, 2001) shows they are also consistent with large extents of melting as discussed above.
Following Cox et al. (1987), we have plotted the final equilibration temperatures and pressures of our mantle samples on a pressure–temperature grid derived from the geothermobarometer of Brey & Köhler (1990) (Fig. 11). All of them plot in one section of the figure between temperatures of 770 and 1130°C and pressures of 2·7–5·3 GPa (100–165 km depth). For comparison, we have also calculated the final equilibration temperatures of the 10 rocks analyzed by Cox et al. (1987) and found that they too plot in the same temperature range as our samples (Fig. 11). Distinctly different from these final rocks, the reconstructed pyroxenes have compositions that are consistent with dry melting experiments that would place them at 1300–1500°C and pressures of 2–3 GPa (60–90 km depth) or experimentally produced orthopyroxenes in equilibrium with a hydrous mantle melt from 2 GPa (Parman, 2001).
Together, these pieces of information can be reconciled with a scenario in which large degrees of shallow mantle melting (60–90 km) produce a harzburgitic residuum that is subsequently pushed down to greater depths (160 km) as a result of cratonic formation. At these greater depths, the high-aluminum opx is no longer stable and the garnet and cpx exsolve, forming the garnet peridotite rocks that are then erupted as kimberlite xenoliths (Fig. 12). We see no evidence that the xenoliths in our sample suite originated at greater depths (300–400 km), as suggested by Haggerty & Sautter (1990), as none of the rocks we analyzed showed a relationship between cpx and garnet or opx and garnet alone. However, we cannot rule this scenario out, as it is possible that majorite would decompose to the same residual products. It should be realized that the ultradeep samples appear distinctly different in texture and mineralogy from the low-temperature harzburgites and that they constitute 10% of the nodule population at the Jagersfontein pipe (Sautter et al., 1991). Therefore, these ‘deeper’ samples may be representative of another process that is also preserved in the Kaapvaal peridotite suite. Our results are also consistent with previous studies that suggest the low-temperature xenoliths were formed as residual products of extensive amounts of melting (>20%) in the Archean at shallow depths (100 km) (Boyd et al., 1997; Keleman et al., 1998; Walter, 1998).
|
The thermal history inferred for these mantle samples could have been imposed by mantle flow that occurs in a subduction zone. In modern subduction zones, anhydrous and hydrous mantle melting results in extreme depletion of the mantle lithosphere, and subduction zone melts are extracted from shallow mantle depths (Baker et al., 1994). After this mantle is depleted it could be entrained in the flow around the wedge corner, where it would undergo cooling as it descended into the mantle above the subducted slab (Davies & Stevenson, 1992; Kincaid & Sacks, 1997). The presence of the most depleted samples at the greatest depths is consistent with this model. If the melt column that was left from adiabatic decompression melting in the back-arc was dragged around the mantle wedge corner and cooled against the subducted slab, the column may have been inverted (Fig. 12) as it was cooled and dragged to greater depths. If the subduction process shut down or the trench migrated, the depleted mantle would record an initial, hot, shallow equilibration followed by a later, cooler, deeper re-equilibration event. Thus, the depleted Kaapvaal harzburgites may provide a record of Archean subduction processes, and may form a complement to komatiite magmas that have also been proposed to record Archean subduction-zone processes (Parman et al., 1997).
|
CONCLUSIONS |
|---|
|
TOPABSTRACTINTRODUCTIONSAMPLE SELECTIONTEXTURAL ANALYSIS METHODRESULTS AND INTERPRETATIONCONCLUSIONSREFERENCES |
|---|
We have developed a new technique for 2-D textural analysis that uses the raw pixel data of scanning back-scattered electron and X-ray images to determine modal amounts, average crystal sizes and whether any of the minerals are more closely associated spatially than would be expected if they were randomly oriented.
We apply this technique to thin sections of nine, low-temperature garnet peridotites exhumed within kimberlite pipes in South Africa. Eight of the nine rocks we analyzed show a spatial relationship between opx, cpx and garnet. By recombining these three minerals, we infer that the original rock was a harzburgite that originated at 100 km depth from large amounts of mantle melting. The rocks were subsequently pushed down to greater depths in the creation of the craton and the cpx and garnet then exsolved when they re-equilibrated.
|
ACKNOWLEDGEMENTS |
|---|
We are deeply indebted to Professor Maarten de Wit of the University of Cape Town and to DeBeers for assistance in collecting the Kimberley samples. Professor Steven Haggerty at the University of Massachusetts and Dr Joe Boyd at the Carnegie Institution of Washington generously provided the other xenolith samples for this study. Thoughtful comments from Ben Harte, Martin Drury and a third, anonymous, reviewer greatly improved this paper. We also appreciate the efforts of Professor T. H. Jordan, who assisted in the development of the textural analysis method. Support for this research was provided by NSF Grant EAR-9526702 under the Continental Dynamics Program.
|
FOOTNOTES |
|---|
*Corresponding author. e-mail: rsaltzer{at}quake.mit.edu
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONSAMPLE SELECTIONTEXTURAL ANALYSIS METHODRESULTS AND INTERPRETATIONCONCLUSIONSREFERENCES |
|---|
Armstrong, J. T. (1995). CITZAF—a package of correction programs for the quantitative electron microbeam X-ray analysis of thick polished materials, thin-films and particles. Microbeam Analysis 4, 177–200.
Baker, M. B., Grove, T. L. & Price, R. (1994). Primitive basalts and andesites from the Mt. Shasta region, N. California: products of varying melt fraction and water content. Contributions to Mineralogy and Petrology 118, 111–129.
Bouillier, A. M. & Nicolas, A. (1975). Classification of textures and fabrics of peridotite xenoliths from south African kimberlites. Physics and Chemistry of the Earth 9, 467–475.
Boyd, F. R. (1987). High- and low-temperature garnet peridotite xenoliths and their possible relation to the lithosphere–asthenosphere boundary beneath southern Africa. In: Nixon, P. H. (ed.) Mantle Xenoliths . New York: John Wiley, pp. 403–412.
Boyd, F. R. (1989). Compositional distinction between oceanic and cratonic lithosphere. Earth and Planetary Science Letters 96, 15–26.[Web of Science]
Boyd, F. R. & Mertzman, S. A. (1987). Composition and structure of the Kaapvaal lithosphere, southern Africa. In: Mysen, B. A. (ed.) Magmatic Processes: Physiochemical Principles. Geochemical Society, Special Publication 1, 13–24.
Boyd, F. R. Pokhilenko, N. P., Pearson, D. G., Mertzman, S. A., Sobolev, N. V. & Finger, L. W. (1997). Composition of the Siberian cratonic mantle: evidence from Udachnaya peridotite xenoliths. Contributions to Mineralogy and Petrology 128, 228–246.
Brey, G. P. & Köhler, T. (1990). Geothermobarometry in four-phase lherzolites II. New thermobarometers, and practical assessment of existing thermobarometers. Journal of Petrology
31, 1353–1378.
Carlson, W. D., Denison, C. & Ketcham, R. A. (1995). Controls on the nucleation and growth of porphyroblasts: kinetics from natural textures and numerical models. Geological Journal 30, 207–225.
Cashman, K. V. & Ferry, J. M. (1988). Crystal size distribution (CSD) in rocks and the kinetics and dynamics of crystallization. Contributions to Mineralogy and Petrology 99, 401–415.
Cox, K. G., Smith, M. R. & Beswetherick, S. (1987). Textural studies of garnet lherzolites: evidence of exsolution origin from high-temperature harzburgites. In: Nixon, P. H. (ed.) Mantle Xenoliths . New York: John Wiley, pp. 537–550.
Davies, J. H. & Stevenson, D. J. (1992). Physical model of source region of subduction zone volcanics. Journal of Geophysical Research 97, 2037–2070.
Finnerty, A. A. & Boyd, F. R. (1987). Thermobarometry for garnet peridotites: basis for the determination of thermal and compositional structure of the upper mantle. In: Nixon, P. H. (ed.) Mantle Xenoliths . New York: John Wiley, pp. 381–402.
Haggerty, S. E. & Sautter, V. (1990). Ultradeep (greater than 300 kilometers), ultramafic upper mantle xenoliths. Science
248, 993–996.
Harte, B. (1977). Rock nomenclature with particular relation to deformation and recrystallisation textures in olivine-bearing xenoliths. Journal of Geology 85, 297–288.
Harte, B. (1983). Mantle peridotites and processes—the kimberlite sample. In: Hawkesworth, C. J. & Norry, M. J. (eds) Continental Basalts and Mantle Xenoliths . Nantwich, UK: Shiva, pp. 46–91.
Harte, B., Cox, K. G. & Gurney, J. J. (1975). Petrograph and geological history of upper mantle xenoliths from the Matsoku kimberlite pipe. Physics and Chemistry of the Earth 9, 447–506.
Herzberg, C. T. (1993). Lithosphere peridotites of the Kaapvaal craton. Earth and Planetary Science Letters 120, 13–29.
Jerram, D. A., Cheadle, M. J., Hunter, R. H. & Elliott, M. T. (1996). The spatial distribution of grains and crystals in rocks. Contributions to Mineralogy and Petrology 125, 60–74.
Keleman, P. B., Hart, S. R. & Bernstein, S. (1998). Silica enrichment in the continental upper mantle via melt/rock reaction. Earth and Planetary Science Letters 164, 387–406.[Web of Science]
Kincaid, C. & Sacks, I. S. (1997). Thermal and dynamical evolution of the upper mantle in subduction zones. Journal of Geophysical Research 102, 12295–12315.[Web of Science]
Kinzler, R. J. (1997). Melting of mantle peridotite at pressures approaching the spinel to garnet transition: application to mid-ocean ridge basalt petrogenesis. Journal of Geophysical Research 102, 853–874.
Kinzler, R. J. & Grove, T. L. (1999). Origin of depleted cratonic harzburgites by deep fractional melt extraction and shallow olivine cumulate infusion. In: Gurney, J. J. (ed.) Proceedings of the 7th International Kimberlite Conference, Capetown, Vol 1. Red Roof Design, . 1, pp. 437–443.
Kretz, R. (1969). On the spatial distribution of crystals in rocks. Lithos 2, 39–66.
Kretz, R. (1993). A garnet population in Yellowknife schist, Canada. Journal of Metamorphic Geology 11, 101–120.
Kruse, R. & Stünitz, H. (1999). Deformation mechanisms and phase distribution in mafic high-temperature mylonites from the Jotun Nappe, southern Norway. Tectonophysics 303, 223–249.
Kushiro, I. (1973). Partial melting of garnet lherzolites from kimberlite at high pressures. In: Nixon, P. H. (ed.) Lesotho Kimberlites . Cape Town: Cape and Transvaal Printers Ltd, pp. 249–299.
Miyake, A. (1998). Monte Carlo simulation of normal grain growth in 2- and 3- dimensions: the lattice-model-independent grain size distribution. Contributions to Mineralogy and Petrology 130, 121–133.
Nixon, P. H. (1987). Kimberlitic xenoliths and their cratonic setting. In: Nixon, P. H. (ed.) Mantle Xenoliths . New York: John Wiley, pp. 215–239.
Nixon, P. H. & Boyd, F. R. (1973). Petrogenesis of the granular and sheared ultrabasic nodule suite in kimberlites. In: Nixon, P. H. (ed.) Lesotho Kimberlites . Cape Town: Cape and Transvaal Printers Ltd, pp. 48–56.
Parman, S. W. (2001). Petrology and geochemistry of high degree mantle melts. Ph.D. thesis, Massachusetts Institute of Technology, Cambridge, MA.
Parman, S. W., Dann, J. C., Grove, T. L. & de Wit, M. J. (1997). Emplacement conditions of komatiite magmas from the 3·49 Ga Komati formation, Barberton Greenstone Belt, South Africa. Earth and Planetary Science Letters 150, 303–323.[Web of Science]
Sautter, V., Haggerty, S. E. & Field, S. (1991). Ultradeep (>300 kilometers) ultramafic xenoliths: petrological evidence from the transition zone. Science
252, 827–830.
Walter, M. J. (1998). Melting of garnet peridotite and the origin of komatiite and depleted lithosphere. Journal of Petrology 39, 29–60.
Yamada, H. & Takahashi, E. (1984). Subsolidus phase relations between coexisting garnet and two pyroxenes at 50 to 100 kbar in the system CaO–MgO–Al2O3–SiO2. In: Kornprobst, J. (ed.) Kimberlites, II. The Mantle and Crust–Mantle Relationships, Proceedings of the Third International Kimberlite Conference. New York: Elsevier, pp. 247–255.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  D.G. PEARSON and N. WITTIG Formation of Archaean continental lithosphere and its diamonds: the root of the problem Journal of the Geological Society, September 1, 2008; 165(5): 895 - 914. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. W. Carlson and R. O. Moore Age of the Eastern Kaapvaal mantle: Re-Os isotope data for peridotite xenoliths from the Monastery kimberlite South African Journal of Geology, June 1, 2004; 107(1-2): 81 - 90. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. W. Parman, T. L. Grove, J. C. Dann, and M. J. de Wit A subduction origin for komatiites and cratonic lithospheric mantle South African Journal of Geology, June 1, 2004; 107(1-2): 107 - 118. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||