ORIGINAL PAPER
D. R. Bell Æ M. Gregoire Æ T. L. Grove Æ N. Chatterjee
R. W. Carlson Æ P. R. Buseck
Silica and volatile-element metasomatism of Archean mantle:a xenolith-scale example from the Kaapvaal Craton
Received: 23 June 2004 / Accepted: 30 March 2005 / Published online: 6 September 2005� Springer-Verlag 2005
Abstract Textural evidence in a composite garnet harz-burgite mantle xenolith from Kimberley, South Africa,suggests metasomatism of a severely melt-depleted sub-strate by a siliceous, volatile-rich fluid. The fluid reactedwith olivine-rich garnet harzburgite, converting olivine toorthopyroxene, forming additional garnet and introduc-ing phlogopite, and small quantities of sulfide and prob-able carbonate. Extensive reaction (>50%) formingorthopyroxenite resulted from channelized flow in a vein,with orthopyroxene growth in the surrounding matrixfrom a pervasive grain-boundary fluid. Themineralogy ofthe reaction assemblage and the bulk composition of theadded component dominated by Si and Al, with lesserquantities of K, Na, H, C and S, are consistent withexperimental studies of hybridization of siliceous melts orfluids with peridotite. However, low Na, Fe and Cacompared with melts of eclogite suggest a fluid phase thatpreviously evolved by reaction with peridotitic mantle.Garnet and phlogopite trace element compositions indi-cate a fluid rich in large-ion lithophile (LIL) elements, butpoor in high field-strength elements (HFSE), qualitativelyconsistent with subduction zone melts and fluids. An Os
isotope (TRD) model age of 2.97 ± 0.04 Ga and lack ofcompositional zonation in the xenolith indicate an an-cient origin, consistent with proposed 2.9 Ga subductionand continental collision in the Kimberley region. Theveined sample reflects the silicic end of a spectrum ofcompositions generated in the Kimberley mantle litho-sphere by the metasomatizing effects of fluids derivedfrom oceanic lithosphere. These results provide petro-graphic and chemical evidence for fluid-mediated Si-,volatile- and trace-element metasomatism of Archeanmantle, and support models advocating large-scalemodification of regions of Archean subcontinental man-tle by subduction processes that occurred in the Archean.
Introduction
Ancient continental nuclei, or cratons, are underlain bydeep roots of refractory, chemically buoyant peridotitethat is depleted in basaltic components such as Na, Ca,Al, and Fe compared with the surrounding mantle(Boyd and McCallister 1976; Boyd and Mertzman1987). These roots are believed to play a key role in thelong-term survival of continental crust, which they iso-late from convective instability by a combination ofbuoyancy and mechanical strength (Jordan 1988,Shapiro et al. 1999).
The important physical properties of Archean litho-spheric mantle reflect its unique chemical composition.In addition to the high Mg# [=100 Mg/(Mg+Fe)]typical of Archean cratonic peridotite (Boyd andMertzman 1987; Griffin et al. 1999; Gaul et al. 2000),many samples, in particular those from the Kaapvaalcraton of southern Africa, contain unusually high Sicontents (manifested by high modal orthopyroxene)compared with other mantle samples (O’Hara et al.1975; Boyd 1989; Herzberg 1993). Studies on severalother Archean cratons reveal that not all Archeanmantle is Si-rich (Kelemen et al. 1998; Kopylova andRussell 2000; Walter 2004), suggesting that this charac-
Communicated by J. Hoefs
D. R. Bell (&) Æ P. R. BuseckDepartment of Geological Sciences and Department of Chemistry/Biochemistry, Arizona State University, 871404, Tempe,AZ 85287-1404, USAE-mail: [emailprotected]: +1-480-965-8102
D. R. Bell Æ T. L. Grove Æ N. ChatterjeeDepartment of Earth, Atmospheric and Planetary Sciences,Massachusetts Institute of Technology, Cambridge,MA 02139, USA
M. GregoireObservatoire Midi-Pyrenees, UMR - CNRS 5562,14 Avenue E. Belin, 31400 Toulouse, France
R. W. CarlsonDepartment of Terrestrial Magnetism,Carnegie Institution of Washington, 5241 Broad Branch RoadN.W., Washington, DC 20015, USA
Contrib Mineral Petrol (2005) 150: 251–267DOI 10.1007/s00410-005-0673-8
teristic may be related to specific regional geologicenvironments in the Archean.
The compositional features of cratonic mantle xeno-liths place constraints on the tectonic environment oftheir formation that are important in the current debateover craton root origins in plumes or subduction envi-ronments (e.g., Wyman and Kerrich 2002; Griffin et al.2003; Parman et al. 2004).
Whereas the high Mg# indicates substantial meltdepletion (Boyd 1989), mass balance constraints pre-clude the simple removal of basalt or komatiite to ex-plain the high SiO2 content of some xenoliths, leading toa variety of other proposals for their origin (Boyd andMertzman 1987; Kesson and Ringwood 1989; Herzberg1993; Boyd et al. 1997; Kinzler and Grove 1999; Walter1998; 1999). Furthermore, experimental studies simu-lating mantle melting indicate that the Si-rich composi-tions of xenolith populations from the Kaapvaal,Siberian and Slave cratons are not compatible withresidues of melt extraction under the wide range ofconditions investigated to date (Walter 2004).
Recently, models emphasizing the importance of melt-rock and fluid-rock interaction in explaining variousmantle phenomena, as well as crustal compositions, haveachieved increasing prominence (Kelemen 1990, 1995;Kelemen et al. 1993; Niu et al. 1997, Wagner and Grove1998; Kinzler and Grove 1999). Kesson and Ringwood(1989) proposed ametasomatic solution to the problem ofArchean mantle xenolith compositions, in which silicifi-cation of occurred in the mantle above a subduction zoneby the reaction of olivine with siliceous aqueous fluids toform enstatite. Subsequent variations on this theme(Kelemen et al. 1993, 1998; Rudnick et al. 1994) proposedthat the metasomatic agent was a siliceous melt. Suchproposals have remained controversial because they re-quire large volumes of metasomatic agent, intrudingpervasively on a widespread scale, and because littlepetrographic evidence for such a process operating inArcheanmantle has been presented. Nevertheless, there isevidence of orthopyroxene introduction in xenoliths de-rived from mantle below the Proterozoic Colorado Pla-teau (Smith et al. 1999), and of phlogopite and/orenstatite introduction in massif peridotites that are be-lieved to represent mantle wedge environments (Zanettiet al. 1999; Tomoaki et al. 2003). Additionally, severalstudies of xenoliths from modern subduction-zone envi-ronments have implicatedmetasomatism by volatile-rich,alkaline aluminosilicate melts (MacInnes and Cameron1994; Schiano et al. 1995; Kepezhinskas et al. 1995, 1996;Prouteau et al. 2001; Laurora et al. 2001).
Pervasive reaction of Archean mantle with fluids ormelts is additionally important in its potential implica-tions for the origins of ubiquitous trace elementenrichment observed in samples of Archean mantle(Shimizu 1975; Hoal et al. 1994; Pearson et al. 1995) andof diamond and other volatile-rich minerals (Kesson andRingwood 1989; Bell et al., 2003). Here, we describe thepetrographic and chemical features of a sample of Ar-chean mantle that we interpret as evidence for the
deposition of silica and other mobile components byreaction of a fluid with refractory mantle harzburgite.
Sample selection
This study focuses on a veined subcalcic garnet harz-burgite nodule from the Boshof Road Dump at Kimber-ley, SouthAfrica.Material from this diamondmine dumpis believed to derive from the Bultfontein kimberlite, inwhich the sample studied here occurred as a mantle-de-rived xenolith. This locality is one of few worldwide thatproduces abundant samples of Archean mantle suitablefor textural studies on a large (>25 cm) scale.
The present work arose from the recent collection oflarge xenoliths aimed at exploring the significance oftextural variations among highly refractory harzburgitesin the Bultfontein suite. This collection, which aimed toexclude the products of Mesozoic metasomatism exten-sively studied previously (e.g., Jones et al. 1982; Kramerset al. 1983; Erlank et al. 1987; Gregoire et al. 2002,2003), revealed a number of textural variants and tex-turally heterogeneous samples (Bell et al. 2003). Inaddition to typical coarse-grained, low-temperatureharzburgites, uniformly rich in orthopyroxene, severalsamples with a heterogeneous spatial distribution oforthopyroxene were encountered. Features of suchxenoliths include diffuse veins and irregularly shapedpatches of orthopyroxenite, commonly phlogopite-bearing and sometimes with exsolved garnet, within arelatively orthopyroxene-poor matrix. This study con-cerns the detailed analysis of one such xenolith. SampleBFT137 measures 40·22·15 cm and has the approxi-mate shape of an egg, halved lengthwise. The flat surfaceis assumed to result from breakage of an originallyovoid xenolith in the plane of a vein of phlogopite-bearing garnet orthopyroxenite. This vein is preservedon the flat sample surface as a selvage some 2–3 cmthick. Preferential fracture of xenoliths along variouskinds of veins is commonly observed at Kimberley. Theremainder of the xenolith consists of clinopyroxene-freegarnet harzburgite, with sparsely distributed phlogopite.
Contrasting with such samples are uniformly ortho-pyroxene-poor, olivine-rich harzburgite containingsparse subcalcic garnet and relatively abundant discrete(i.e., non-symplectic) magnesiochromite. One exampleof this group (BFT147) was studied for comparison withthe veined sample. This xenolith measures�20·10·10 cm and is a hom*ogeneous, orthopyroxene-poor garnet harzburgite/dunite with accessory chromiteand large patches of coarse phlogopite.
Analytical methods
Sample BFT137 was sectioned in a plane perpendicularto the orthopyroxenite vein, and perpendicular to thelong axis of the xenolith, and four large (5·7.5 cm)polished thin sections were prepared from the slab
252
shown in Fig. 1. A single large thin section of sampleBFT147 was prepared. The thin sections were mappedchemically on a 10-lm-grid scale for Mg, Al, and Causing the JEOL733 electron microprobe at MIT and themaps analyzed spatially with commercial image pro-cessing software. Full major element chemical analysesof each mineral, including traverses perpendicular to thevein and core-rim were performed on the JEOL JXA-8900 electron probe at the Geophysical Laboratory,Carnegie Inst. of Washington. The electron microprobeanalyses were designed to provide high levels of preci-sion for individual point analyses. Counting times formajor elements (i.e., > 10 wt.% oxide) were typically setto yield <0.2% relative precision based upon countingstatistics. All minerals were analyzed at 15 kV acceler-ating potential, with the following beam currents mea-sured on the Faraday cup: 40 ± 2 nA (orthopyroxene),60 ± 2 nA (garnet), 150 ± 2 nA (olivine, phlogopite).Extended counting times up to 180 s on peak andbackground were used for trace elements.
A wider range of trace elements was determined insitu on selected grains in > 120-mm-thick polishedsections by LA-ICP-MS at the University of CapeTown. The Perkin Elmer Elan 6000 ICPMS instrumentwas coupled to a Cetac LSX-200 laser ablation modulethat uses a 266-nm frequency-quadrupled Nd-YAG
laser. Details of the LA-ICP-MS analytical procedureappear in Gregoire et al.(2002). Bulk compositions forvein and substrate were computed by combining modalproportions from the chemical maps with electronmicroprobe and LA-ICP-MS mineral analyses.
Bulk analyses of Re and Os concentration and iso-tope composition of BFT137 were determined on asample of the matrix immediately adjacent to the vein,extending 5 cm into the matrix from the vein boundary.The analyses were performed at the Carnegie Institutionof Washington, Department of Terrestrial Magnetismusing techniques described by Carlson et al. (1999).
Mineralogy and texture
A macroscopic view of the sample is given in Fig. 1,illustrating the linear selvage (vein) of coarse-grainedorthopyroxenite bounding a matrix of coarse-grainedharzburgite. The boundary of the vein is not sharp, butappears to grade over a distance of one to two centime-ters, giving rise to a somewhat indistinct boundary in thinsection. Modal heterogeneities in other Bultfonteinharzburgite samples have a similarly diffuse character.The modal compositions of the vein and host matrix aregiven in Table 1.
Fig. 1 1-cm-thick slab of sample BFT137, cut at right angles to the vein and the long axis of the xenolith. Orthopyroxene-rich vein occursalong the top edge of sample and the position of four large thin sections, A, B, C and D, used for mineral analyses and bulk compositionalcalculations is illustrated
253
Vein
Besides modally dominant enstatite, the vein region alsocontains garnet, subordinate olivine, abundant coarsephlogopite, and large (up to 0.5 mm diameter) sub-spherical to euhedral sulfide grains. Occasional patcheswith the outline of a primary mineral (Figs. 2a, 2b) butcomprising a fine-grained intergrowth dominated byserpentine are inferred to be pseudomorphs after a car-bonate, as described by Berg (1986). Orthopyroxenegrains are mostly equant without preferred orientationand olivine occurs interstitially and as inclusions to theorthopyroxene. Most garnet grains are rounded, withoccasional irregular to poikilitic grain shapes (Fig. 3).The primary minerals in both vein and host matrixare also affected to various degrees by hydrothermal
Table 1 Modal compositions (volume%)
Mineral Volume%
BFT137 BFT137 BFT147Vein Host matrix
Olivine 29.7 77.9 88.0Enstatite 56.6 18.2 4.5Garnet 6.5 3.6 2.5Cr-spinel – – 2.5Phlogopite 6.7 – 2.4Carbonate* 0.3 0.3 –Sulfide 0.3 – –
*Carbonate inferred from serpentine-brucite-calcite assemblageswith primary grain outlines, described in text. The carbonateabundances inferred from these pseudomorphs were determinedover the entire xenolith and therefore reported here as equal in veinand matrix
Fig. 2 Petrographic features of sample BFT137. a Low-magnifica-tion view of sawn slab surface, illustrating primary-appearingoutlines of serpentine-brucite pseudomorphs (mag) after probablecarbonate (Berg 1986). b Thin section plane-polarized light view ofvein assemblage, including phlogopite (phl), enstatite (opx), olivine(ol), serpentine-brucite pseudomorphs (‘‘mag’’), sulfides (sulf) andveins of serpentine formed during emplacement-related hydrother-mal alteration. c Backscattered electron image of vein assemblage.Sulfide surrounded by olivine (ol) and by phlogopite (phl) partially
altered along cleavage planes (chl). The original sulfide isrecrystallised to an Fe-oxide (mt) and Ni-rich secondary sulfides(pn-mi), with secondary barite formed along one margin of thegrain. d Secondary Ni-sulfide (N) and Fe-oxide (F) assemblages ina pseudomorph after original euhedral sulfide crystal. The oxideportion of the assemblage is located where a prominent serpentinevein (serp) that transgresses the enclosing enstatite (opx) intersectsthe original sulfide
254
processes inferred to be associated with kimberliteemplacement. Olivine is partially replaced (10–20%) bya network of chrysotile. Most sulfide grains are exten-sively oxidized, comprising a mixture of Ni-rich sulfides,magnetite, and fine-grained silicate, occasionally withassociated barite (Fig. 2c). This alteration can in somecases be traced directly to individual veins of serpentine(Fig. 2d) and is common in mantle-derived sulfides(Lorand 1993). Phlogopite contains cleavage-parallelzones of chloritization, apparent as low-atomic numberregions in backscattered electron images (Fig. 2c).
Host matrix
The host region consists of the same minerals, except forsulfide, but with more olivine and less enstatite, garnet,and phlogopite. Carbonate pseudomorphs also occur inthe matrix, but are more difficult to quantify unambig-uously than the other phases as a result of serpentini-zation. Phlogopite in the host is only present in a singlelarge, elongated patch running oblique to the mainorthopyroxenite vein. This region of the matrix is locallyricher in enstatite and appears to be a poorly defined,vein-like structure. Garnets are distributed evenly in thematrix and are rounded to subrounded.
The color-coded Mg map (Fig. 3) highlights the tex-ture of orthopyroxene in the matrix. Orthopyroxenes arecommonly highly irregular in outline, containingnumerous embayments, and commonly completelyenclosing olivine grains. Many orthopyroxene grains areelongate and sinuous, appearing to indicate originalgrowth along grain boundaries, followed by poikiliticengulfment of olivine as the grains grew. Because en-statite growth requires introduction of silica, the texturesuggests growth from a dispersed, grain-boundary
medium. Orthopyroxene shapes in the vein are not assinuous, but there are many instances of rounded olivineinclusions, suggesting that the enstatite vein also grew byreaction and replacement of olivine. The greater ortho-pyroxene abundance in the vein results in the mutualinterference of grains, and their more equant habitsuggests that growth by reaction proceeded to a moreadvanced stage, reflecting extended reaction progress ina region of greater reactant supply. These modal abun-dances and internal textures of the vein suggest a regionof channelized fluid flow. However, the fluid conduitcould have been significantly narrower than the vein it-self, and its original dimensions or form are not evident.
In contrast to BFT137, BFT147 has much less garnet(2.5%) and enstatite (4.5%) and no carbonate pseud-omorphs. However, it contains about 2% magnesi-ochromite. Large phlogopite patches (2%) are present.Enstatite also has a habit suggesting growth from aninter-granular medium, but the grains are small andappear to represent the initial stages of such a process.
Mineral compositions
Major elements
Average mineral compositions determined by electronmicroprobe are given in Tables 2 and 3. Garnet, orth-opyroxene and olivine are internally hom*ogeneous ex-cept for small variations (up to 0.4 wt.%) in CaO andCr2O3 in garnet. A consistent difference of 0.1–0.2 wt.%exists between cores and rims, with zoning to higherCaO and lower Cr2O3 in the outermost 300 microme-ters. This zonation may result from the infiltration ofsmall quantities of kimberlite-derived fluids at, orshortly before, eruption. Variations in the CaO and
Fig. 3 Magnesium map of aportion of BFT137, section B(Fig. 1), determined by WDSanalysis. The gray scaleisproportional to Mgconcentration. Light grey–olivine, medium gray–orthopyroxene, dark grey roundgrains– garnets. Phlogopite hasbeen colored pink and sulfideyellow. The orthopyroxene-richvein with phlogopite occurs inthe upper portion of the section.Orthopyroxene exhibitsamoeboid to sinuous habit,with olivine inclusions. Theposition of the compositionaltraverses (Fig. 4) marked bybroad arrows. Scale bar = 1 cm
255
Cr2O3 content of grain cores also exceed analyticaluncertainty, but do not vary systematically across thexenolith. Phlogopite is hom*ogeneous in its major com-ponents, but certain trace elements show variations thatexceed analytical uncertainties. All minerals are ex-tremely magnesian, with olivine Fo94.0, orthopyroxeneEn94.7, phlogopite Mg# = 95.5 and garnet Mg# =87.9. The garnet contains 5.4 wt.% Cr2O3 and 3.3 wt.%CaO and thus falls well within the G10 (‘‘harzburgitic’’)field defined by Dawson and Stephens (1975) and Gur-ney (1984). All phases are extremely low in TiO2. The
orthopyroxene contains an unusually high concentrationof Na (0.17 wt.% Na2O) for this low Ti content. Ther-mobarometry calculations based upon Fe–Mg exchangebetween coexisting garnet and olivine (O’Neill andWood 1979; O’Neill 1980) and Al content of orthopy-roxene in equilibrium with garnet (Brey and Kohler1990) give equilibration conditions of 3.6 GPa and1000�C.
Analytical traverses perpendicular to the vein inBFT137 reveal that compositions are equilibrated acrossthe xenolith, with no compositional differences betweenminerals in vein and matrix (Fig. 4). The decimeter-scalehom*ogeneity confirms inferences from chemical com-positions that vein formation is ancient and not relatedto metasomatism shortly before eruption.
The minerals in BFT147 (Table 3) are also highlymagnesian, the olivine (Fo94.1) and enstatite (En95.0)marginally more so than in BFT137. Ni in BFT147olivine (0.353 wt.% NiO) is lower than in the olivine ofBFT137 (0.386 wt.% NiO). Relatively low Ni content isa general feature of olivines from the Kimberley ortho-pyroxene-poor harzburgites (Bell et al. 2003). Garnethas somewhat higher Cr (7.5 wt.% Cr2O3) and lower Ca(3.2 wt.% CaO) content than BFT137. The spinel is Cr-rich (61 wt.% Cr2O3). Phlogopite is richer in K and Cl
Table 2 Major and trace element compositions (wt.%) for primaryminerals in BFT137, determined by electron microprobe*
Olivine Enstatite Garnet Phlogopiten = 66 n = 96 n = 104 n = 16
SiO2 41.8 (2) 58.2 (3) 42.5 (3) 41.7 (5)TiO2 <0.005 <0.01 0.023 (7) 0.089 (5)Al2O3 <0.005 0.88 (1) 20.6 (1) 12.9 (2)Cr2O3 0.038 (4) 0.48 (2) 5.36 (10) 1.06 (2)FeO 6.06 (4) 3.62 (4) 5.67 (6) 2.25 (3)MnO 0.080 (6) 0.088 (9) 0.29 (1) 0.017 (4)MgO 52.8 (3) 36.5 (2) 23.2 (2) 26.5 (3)CaO 0.016 (3) 0.35 (2) 3.28 (16) < 0.01Na2O 0.013 (4) 0.18 (1) 0.03 (2) 0.11 (3)K2O n.a. n.a. n.a. 10.1 (3)NiO 0.386 (6) 0.10 (1) n.a. 0.208 (7)P2O5 0.009 (4) n.a. 0.053 (8) n.a.F n.a. n.a. n.a. 0.24 (5)Cl n.a. n.a. n.a. 0.075 (5)BaO n.a. n.a. n.a. 0.07 (2)Rb2O n.a. n.a. n.a. 0.014 (6)Mg# 93.95 (4) 94.73 (5) 87.92 (14) 95.46 (9)Cr# 26.6 (9) 14.8 (3) 5.23 (9)
* Mean and standard deviation (last significant digits) of all anal-yses (n = number of analyses). n.a. not analyzed. Data forphlogopite are derived only from areas that appear fresh in BSEimaging (see text). Analyses of altered phlogopite areas are ex-cluded
Table 3 Major and trace element compositions (wt.%) for primaryminerals in BFT147, determined by electron microprobe
Olivine Opx Garnet Phlogopite Spinel
SiO2 40.5 58.2 41.3 42.0 0.07TiO2 <0.005 0.009 0.025 0.11 0.11Al2O3 0.008 0.78 18.8 12.9 8.70Cr2O3 0.039 0.59 7.49 1.50 60.8FeO* 5.82 3.47 5.6 2.48 15.0MnO 0.092 0.093 0.30 0.021 0.25MgO 52.6 37.2 22.6 26.0 14.1CaO 0.027 0.31 3.18 <0.03 <0.03Na2O 0.018 0.16 0.03 0.09 <0.03K2O n.a. n.a. n.a. 9.71 n.a.NiO 0.353 0.092 0.004 0.21 0.10P2O5 n.a. n.a. 0.07 n.a. n.a.Cl n.a. n.a. n.a. 0.55 n.a.F n.a. n.a. n.a. 0.17 n.a.BaO n.a. n.a. n.a. 0.11 n.a.Mg# 94.1 95.0 87.8 94.7 62.6Cr# 33.7 21.1 7.2 82.4
FeO*: all Fe reported as FeO. n.a. = not analyzed
Fig. 4 a, b Electron microprobe traverses across the xenolithillustrating compositional hom*ogeneity. The two parallel traverses(one represented by open symbols, the other by solid symbols) weredetermined on Sect. B (diamond symbols), continuing onto Sect. D(squares). See Figs. 1 and 3 for location of the traverses
256
than the phlogopite in BFT137. The garnet and spinelexhibit minor variations in Cr content from grain tograin, but no core to rim zonation was noted. Thissample equilibrated at 3.4 GPa and 940�C, withinuncertainty of the conditions calculated for BFT137.
Trace elements
Average LA-ICP-MS analyses for minerals in BFT137are reported in Table 4. The garnets in BFT137 arehom*ogeneous from grain to grain, and in spots analyzedat garnet cores and rims (Fig. 5). The garnet chondrite-normalized REE patterns are broadly similar to thosedetermined for garnets from low-temperature harz-burgites on the Kaapvaal Craton (Shimizu 1975; Hoalet al., 1994; Gregoire et al., 2003) but are notable inhaving HREE concentrations at the low end, and LREEconcentrations at the high end, of the range for Kim-berley samples. The BFT137 garnet patterns differ fromthe sinusoidal patterns skewed towards LREE enrich-ment that have been observed in many subcalcic Cr-pyrope inclusions in diamond (Shimizu and Richardson1987; Stachel et al. 2004), some diamondiferous peri-dotites (Pokhilenko et al. 1993, Stachel et al. 1998;Shimizu et al. 1999), and sample BFT147 in this study
(Fig. 5). In contrast to these, the REE pattern inBFT137 approximates a symmetric bell-shape, i.e.,MREE enriched �10X over LREE and HREE (Fig. 5).The typical diamond-inclusion-type patterns (andBFT147) show a peak in normalized REE abundances atNd, whereas those in the present sample show a peak atSm to Eu. A similar pattern to BFT137, but withsomewhat higher HREE content, is reported by Stachelet al. (1999) in a harzburgitic diamond inclusion garnetfrom Mwadui, Tanzania.
The very low Ti concentration in the BFT137 garnetresults in a marked anomaly in the mantle-normalizedtrace element diagram (Fig. 6), where the elements areplotted in terms of increasing compatibility during ba-salt petrogenesis. However, Zr, Hf, and Nb abundancesare not unusually low in concentration compared withneighboring REE. They are, however, a factor of 2 to 4lower than the abundances in a garnet from a shearedperidotite, and the (La/Nb)N, (Sm/Zr)N, (Sm/Hf)N and(Gd/Ti)N are all substantially higher than for thesheared peridotite example (Fig. 6). The sheared peri-dotite is considered to result from the metasomatic ef-fects of a silicate melt related to the Cr-poor megacrystsuite and to kimberlite (Gurney and Harte 1980; Hopset al.1989). These ratios indicate relative high field-strength element (HFSE) depletion of the BFT137garnet in relation to garnets equilibrated with the kim-berlite-related megacryst parent magma, which has LIL-element to HFSE ratios similar to OIB. The garnet inBFT147 has approximately half the Ti, Zr, and Hfconcentrations of the BFT137 garnet, and approxi-mately twice the U, Nb and Sr concentrations. Othertransition element concentrations of the garnets in thetwo samples are similar.
Phlogopites in BFT137 show greater compositionalvariability than the garnets, probably as a result of sec-ondary alteration. Back-scattered electron imaging of
Fig. 5 Chondrite-normalized REE abundances in BFT137 garnetsand a sible analysis from BFT147, determined by LA-ICP-MS. TheREE pattern of a garnet megacryst from kimberlite (D. Bell and A.Kennedy unpubl. SIMS analysis) is shown for comparison.Chondrite abundances are from Anders and Grevesse 1989
Table 4 Mineral trace element compositions (wt. ppm) determinedby LA-ICP-MS
Sample Garnet Olivine Enstatite Phlogopite GarnetBFT137 BFT137 BFT137 BFT137 BFT147n = 7 n = 2 n= 2 n = 3 n = 2
Sc 99 (17) 1.9 (9) 2.7 (13) 2.7 (9) 88 (3)V 172 (24) 5.9 (2) 38 (14) 86 (13) 170 (11)Ni 37 (4) 2313 (157) 550 (177) 969 (93) 29 (1)Co 31 (4) 91 (6) 34 (11) 31 (4) 30 (1)Ga n.d. n.a. n.a. 9 (2) 0.86 (10)Rb n.a. n.a. n.a. 161 (25) n.a.Sr 0.40 (5) 0.68 (8) 2.3 (6) 8.2 (32) 1.19 (7)Ba n.a. n.a. n.a. 527 (195) n.a.Y 1.8 (6) n.d. n.d. 0.03 (1) 1.53 (6)U 0.054 (23) n.a. n.a. n.a. 0.073 (16)Ti 107 (17) 4.8 (7) 24 (5) 283 (55) 66 (2)Zr 42 (10) 0.19 (9) 0.52 (15) 3.2 (8) 22 (1)Hf 0.76 (16) n.a. n.a. n.a. 0.46 (2)Nb 0.13 (4) 0.15 (0.1) 0.20 (11) 11 (2) 0.28 (4)Pb n.a. n.a. n.a. 1.17 (5) n.a.La 0.14 (1) 0.15 (7)Ce 0.38 (8) 1.53 (12)Pr 0.20 (4) 0.84 (8)Nd 2.64 (30) 8.60 (57)Sm 1.86 (25) 1.47 (1)Eu 0.70 (10) 0.33 (4)Gd 1.78 (36) 0.71 (3)Dy 0.62 (17) 0.33 (5)Ho 0.074 (20) 0.044 (7)Er 0.11 (3) 0.11 (2)Yb 0.11 (3) 0.11 (2)Lu 0.021 (5) 0.023
Mean and standard deviation (in last significant digit) of n analyses.n.d. not detected, n.a. not analyzed
257
phlogopite accompanied by trace element electronmicroprobe analysis reveals zones from which the alkalisand alkaline-earth elements have been leached. Thesemay be a cause of the substantial heterogeneity in Baanalyses of phlogopite by LA-ICP-MS (Table 4). Elec-tron microprobe analyses are less variable than the ICP-MS analyses, presumably because of the higher spatialresolution and greater care (backscattered electronimaging) with which analytical areas may be selected,hence avoiding areas of alteration. Phlogopite in BFT137contains an average of approximately 150 ppm Rb and550 ppm Ba. The Rb content is lower than that of allmetasomatic micas analyzed by Gregoire et al. (2002)using the same analytical procedures, and the Ba contentis higher than all but one sample. The TiO2 content of theBFT137 phlogopite (0.09 wt.%) is distinctly lower thanthat of micas in metasomatic xenoliths and at the low endof primary micas in garnet peridotites. Nb, Zr, and Pbcontents are not distinguishable from metasomaticphlogopite in micaceous xenoliths (Gregoire et al. 2002).The phlogopite of BFT147 is also poor in TiO2
(0.12 wt.%) and contains notably higher Cl concentra-tion (0.55 wt.%) than that of BFT137 (0.08 wt.%).
Whole-rock compositions
The bulk compositions of the vein and matrix in BFT137and of the orthopyroxene-poor sample BFT147 werecomputed by combining the modal proportions andmineral compositions, and are reported in Table 5. Thevein is richer in SiO2, Al2O3, Cr2O3, CaO, Na2O, K2O,H2O and S, and poorer in FeO, MgO and NiO than thematrix. BFT147 has a composition similar to the matrixof BFT137, but is richer in Cr2O3 and poorer in SiO2.Also listed in Table 5 is the composition of BFT147recalculated on a K2O- and H2O-free basis, proposed asa possible pre-metasomatic protolith composition forthis sample. Aspects of the major element compositionsare plotted in Fig 7, where they are compared with the
Kaapvaal data of Boyd and coworkers. Fig. 7a illus-trates the highly Si-rich character of the vein, the rela-tively Si-poor character of the BFT137 matrix, and theexceptionally Si-poor and Mg-rich character of theorthopyroxene-poor sample BFT147. The range ofcompositions determined for four sections of the matrixof BFT137 spans approximately the full range of Mg/Siratios for a given olivine Mg# in Kaapvaal low-temper-ature peridotites, and demonstrates that local modalvariations can be responsible for much of the scatter seenin such plots, as noted by Boyd and Mertzman (1987). Incontrast, the variation in Ca and Al contents (like that ofMg#) accounts for only a small percentage of the vari-ability in the regional data set (Fig. 7b).
If the serpentine-brucite patches, which we interpretas carbonate pseudomorphs, were composed of originalmagnesite (the likely stable carbonate under these P-T-X
Fig. 6 Primitive mantle-normalized trace elementsdiagram comparing the averageBFT137 garnet with the traceelement pattern of a garnetfrom the sheared garnetlherzolite PR89-1 from thePremier kimberlite (Gregoireet al. 2003). Notable differencesinclude lower Nb, Zr, Hf, Ti, Srand HREE in BFT137, butsimilar LREE abundances.Primitive mantle abundancesare from McDonough and Sun(1995)
Table 5 Calculated bulk compositions (wt.%) a
BFT137 BFT137 BFT147 BFT147Vein Host Phl-FREEb
SiO2 50.6 44.5 40.3 40.3TiO2 0.011 0.002 0.006 0.004Al2O3 2.68 0.90 1.06 0.76Cr2O3 0.69 0.30 1.90 1.87FeO 4.39 5.61 5.84 5.91MnO 0.094 0.089 0.098 0.100MgO 39.8 48.8 49.5 50.1CaO 0.41 0.19 0.12 0.12Na2O 0.11 0.033 0.026 0.024K2O 0.66 < 0.01 0.23 < 0.01NiO 0.19 0.32 0.32 0.33P2O5 0.006 0.009 n.a. n.a.H2O
c 0.26 < 0.01 0.095 < 0.01Sd 0.12 < 0.01 < 0.01 < 0.01
a Calculated from mineral analyses and element-map-based modalanalysisb Analysis recalculated on phlogopite-free basisc H2O calculated assuming 4 wt.% H2O in phlogopite, none inother mineralsd S calculated assuming 35 wt.% S in sulfide, none in other min-erals
258
conditions), then BFT137 would have containedapproximately 0.15 wt.% CO2.
Re–Os isotopes
The Re–Os abundances and isotope compositions arepresented in Table 6. The Re and Os concentrations aretypical of Kaapvaal peridotites. The Re abundance of0.116 ppb is close to the average Kaapvaal peridotite,but is higher than most harzburgites from Kimberley(average 0.031 ppb; Carlson and Moore 2004). The lowcOs of �16.3 is typical of Archean subcontinentalmantle and indicates a long time-integrated historyof extreme Re-depletion, consistent with Archeanmelt extraction. The Re-depletion age (TRD) of 2.97
± 0.04 Ga is among the oldest recorded for Kimberleysamples (Walker et al. 1989; Pearson et al. 1995a;Carlson et al. 1999; Simon et al. 2003a, b). However, thedepleted mantle model age (TMA) > 4.5 Ga indicateslate-stage Re addition, probably during emplacement-
Fig. 7 a Fo content of olivinevs Mg/Si of bulk rock and bCaO vs. Al2O3 for BFT137components (the four large thinsections illustrated in Fig. 1,and average vein and matrixconcentrations) and Kaapvaallow-temperature peridotitewhole rock analyses compiledfrom the literature. The arrowsrepresent the trend defined byaverage compositions ofoceanic peridotite suites (Boyd1989)
Table 6 Re–Os isotope data for BFT137
Re 0.116 ppbOs 2.714 ppb
187Re/188Os 0.2045 ± 0.0016187Os/188Os 0.10753 ± 0.00027187Os/188Os90 0.10722 ±cOs90 �16.3 ± 0.3TRD 2.97 ± 0.04TMA 5.52 ± 0.11
259
related hydrothermal serpentinization that resulted insulfide recrystallisation.
Discussion
Textural evidence for metasomatic introductionof orthopyroxene
The petrographic features of enstatite, including its dis-tribution and grain shapes, are evidence that this mineralgrew from a grain-boundary fluid that permeated thexenolith, with a greater fluid supply and degree of reac-tion in the vein. Similar evidence has been presented forfluid-mediated orthopyroxene growth in mantle-derivedxenoliths from the Colorado Plateau (Smith et al.1999;Smith 2000) and the southwestern Pacific (McInnes andCameron 1994). Both these suites of samples derive fromthe mantle in the neighborhood of subduction zones,which are thought to be the source of the metasomatizingfluids (Smith et al. 2000, 2004; McInnes and Cameron1994; McInnes et al. 2001). The orthopyroxenes from theColorado Plateau xenoliths are richer in small olivineinclusions than the Bultfontein sample, and also arecharacterized by a particular Al-poor composition. Incontrast, the Bultfontein orthopyroxene compositionsare typical for low-temperature cratonic peridotites andreflect equilibrium with an aluminous phase (garnet orspinel). These textural and chemical differences are notsurprising given the probability that the vein under dis-cussion is likely to be much older (�2.9 Ga) than theCenozoic to recent age for the other examples. Partialtextural re-equilibration during this time may have re-moved small inclusions and promoted chemical equili-bration with other minerals in the surrounding mantle.Chemical equilibration with garnet or spinel would haveremoved any evidence that the orthopyroxene in BFT137was originally low in Al. The xenolith reveals thatchemical equilibrium has been achieved on a length scaleof at least a decimeter, and may extend further.
In sample BFT137, the enstatite in the vein isaccompanied by volatile-rich phases, including phlogo-pite, sulfide and probable carbonate. The sulfide isconfined to the vein, whereas carbonate and phlogopiteoccur both in matrix and vein. Phlogopite is less abun-dant in the matrix, but the inferred carbonate pseud-omorphs occur in similar proportions in both.Observations of other veined, orthopyroxene-rich sam-ples reveal that these pseudomorphs are sometimesconcentrated in a zone outside of the vein defined bymodal enstatite. The distribution of pseudomorphssuggests that carbon, in contrast to sulfur, is relativelymobile during fluid-peridotite reaction.
The abundances of phlogopite and enstatite are cor-related. In addition, the higher modal abundance ofgarnet in the vein requires introduction of Al. However,garnet grain shapes appear similar in vein and matrix,and distinctive textural evidence for metasomatic garnetgrowth is lacking. There is also no evidence that garnet
exsolved from aluminous enstatite, as inferred for someother samples of Kaapvaal craton peridotite (Cox et al.1987; Canil 1991; Saltzer et al. 2001; Lahaye and Brey2003; Dawson 2004). The preservation of orthopyroxenetextures and lack of evidence for garnet exsolutionsuggest that metasomatic reaction occurred at temper-atures not substantially above those recorded by thexenolith (�1000�C) at the time it was sampled by therising kimberlite magma.
The conclusion that enstatite was introduced into thisxenolith by a mobile phase has important consequencesfor models of cratonic mantle evolution. The Si-richnature of Archean low-temperature peridotite from theKaapvaal craton mantle root has been long-recognizedand its origins much debated (O’Hara et al. 1975;Herzberg 1993; Walter 2004). Although metasomaticmodels have played an important role in this debate(Kesson and Ringwood 1989; Rudnick et al. 1994;Kelemen et al. 1998), the present description represents,to our knowledge, the first report of petrographic evi-dence in support of such a process in the Kaapvaalmantle. Evidence from sample BFT137 indicates thatintroduction of phlogopite, some garnet, and traceamounts of sulfide and carbonate accompanies enstatiteformation, whereas olivine, and possibly Cr-spinel areremoved in the process.
In sample BFT147, which contains only 4.5% en-statite, the textures also suggest metasomatic introduc-tion. Although enstatite distribution is not confined toveins or patches, grains appear to be aligned in lineararrays parallel to their individual direction of elonga-tion. The initial stages of enstatite growth are repre-sented by small, sinuous orthopyroxenes, very differentin habit from the equant, cm-sized crystals that typify Si-rich cratonic peridotite. The sample also containsmetasomatic phlogopite. These minerals may representthe initial stages of infiltration by a volatile-rich fluid, inwhich fluid-rock interaction has been insufficient forsignificant quantities of enstatite growth, or where thecomposition of the fluid was richer in K and poorer in Sias a result of chromatographic fractionation during itspassage through, and reaction with, the mantle (Navonand Stolper 1987). In our model, the phlogopite-bearing,but orthopyroxene-poor assemblages are proposed tooccur distal to zones of concentrated fluid flow, andshould bear additional features of metasomatism byextensively fractionated fluids (e.g., Bodinier et al. 2004).Indeed, the trace element composition of BFT147phlogopite and garnet reflect higher concentrations ofhighly incompatible elements (Cl, U, Nb, LREE and Sr)than the minerals in BFT137.
Chemical reactions accompanying orthopyroxenegrowth
The modal proportions listed in Table 1 indicate that thefollowing net mineralogical reaction took place,depending upon the choice of initial unreacted substrate:
260
BFT137 Matrix D as substrate:1.25 Oliv + Fluid = 1 Opx + 0.076 Gar + 0.17Phlog + 0.0086 Sulf + 0.038 Carb (1)BFT147 as substrate:1.15 Oliv + 0.05 Sp + Fluid = 1 Opx + 0.076 Gar+ 0.13 Phlog + 0.0063 Sulf + 0.028 Carb (2)
In (1) and (2) the reaction coefficients are in terms ofvolume. The coefficient for carbonate was calculatedassuming that all carbonate pseudomorphs in both veinand matrix were introduced during vein formation. Thisis probably a minimum, in that some introduced car-bonate could have been deposited beyond the presentboundary of the xenolith. Note that reaction (1) doesnot contain Cr-spinel (Sp) as a reactant because none isobserved in the matrix to this xenolith. However, thisdoes not mean it was not originally present; in fact it islikely, because our observations on a wider suite ofrefractory harzburgites indicate that Cr-spinel is com-mon in harzburgites and dunites with Fo>93.5. Reac-tions (1) and (2) are similar in their coefficients ofreaction, indicating approximately one for one replace-ment of olivine by orthopyroxene, with other phasesplaying a much lesser role. The amount of fluid in thereaction cannot be constrained because the solute con-centrations are unknown.
Nature of the metasomatic agent
The very high Mg#, extremely low Ti, Ca and HREEcontents of minerals, but presence of volatile-rich phasesand high incompatible element concentrations inBFT137 indicate that a mafic or ultramafic silicate meltis an unlikely metasomatic agent. Reaction with suchmelts is commonly inferred for other mantle xenoliths,where it leads to prominent and distinctive Fe- and Ti-enrichment (Gurney and Harte 1980) and high HREE ingarnet (Griffin et al. 1989; Burgess and Harte 1999).Mass balance considerations suggest that carbonatiticmelt is unlikely to precipitate enstatite in significantquantities, although it may be in equilibrium withharzburgite at � 4 GPa (Lee and Wyllie 2000). The mostlikely candidate for the parent fluid therefore appears tobe a hydrous siliceous fluid or melt, poor in ferromag-nesian components, with minor but geochemically sig-nificant quantities of C and S.
The mineral assemblages and inferred reactions inBFT137 are consistent with the body of experimentalresults for the hybridization of hydrous, silica-rich meltsor fluids with refractory peridotite (Sekine and Wyllie1982; Wyllie et al. 1989; Rapp et al. 1999; Wunder andMelzer 2003). These studies show that enstatite, garnet,and phlogopite are the reaction products over a widerange of temperatures and pressures. However, the ab-sence in BFT137 of Na-rich clinopyroxene, which iscommon in these hybridization experiments, suggests ametasomatic agent poor in Na and Ca. Whether themobile phase was a dense hydrous supercritical fluid(‘‘vapor’’) or a water-bearing siliceous melt is debatable.
Experimental studies, reviewed by Wyllie and Ryabchi-kov (2000), indicate that a wide range of solute con-centrations is possible as volatile-rich upper mantlesystems approach a second critical endpoint in the P–Trange from which our sample originates.
Understanding trace element partitioning as a func-tion of volatile content may help to determine where themetasomatic agent fits in the fluid-melt spectrum. Fromtheir bulk chemical analyses of Kaapvaal xenoliths thatindicated a lack of correlation between Re and Al,Carlson et al. (1999) favored a low-Al fluid rather than asiliceous melt such as a tonalite–trondhjemite for themetasomatic agent. This inference is based upon theassumption that Re, like Fe, would be less soluble in afluid phase than a melt. The low Re content inferred forBFT137 from its low 187Os/186Os ratio is consistent withthis interpretation. Furthermore, Stachel et al. (2004)argued that the REE patterns of harzburgitic garnets inxenoliths and diamond inclusions are qualitatively moreconsistent with fluids than melts.
Composition of the added component
The bulk chemical changes that accompany orthopy-roxene growth were calculated by assuming an appro-priate starting composition and identifying a referencecomponent unaffected by the metasomatic reaction.Two starting compositions were used; the BFT137 ma-trix, and sample BFT147. It is clear in BFT137 fromtextural evidence that some introduction of enstatite intothe matrix occurred, and therefore this calculation yieldsa minimum estimate of the ensuing chemical change.BFT147 was chosen as a pre-metasomatic substratebecause this sample has very similar bulk Mg# toBFT137 (93.8 vs. 93.9 for BFT137), but lacks majorenstatite. Because the phlogopite is of likely metaso-matic origin, the initial protolith composition was cal-culated without the K2O and H2O that are assumed tohave been introduced into the xenolith. The choice ofreference component is restricted in this study by theinference that diffusive re-equilibration of componentsafter vein formation has occurred. The most suitablereference components are, therefore, those fixed bymineral stoichiometry, so long as the modes in vein andmatrix have remained unchanged since their formation.We chose (Mg+Fe), which is approximately fixed bymineral stoichiometry in these low-Ca rocks.
Experimental studies at 1 – 4 GPa support the pre-sumption of low Mg+Fe in upper mantle fluids andsiliceous melts (Sekine and Wyllie 1982; Schneider andEggler 1986; Brenan et al. 1995; Stalder et al.1998).However, the Mg content and Mg/Si of hydrous fluid inequilibrium with peridotite increase as a function ofpressure (Stalder et al. 2001; Mibe et al. 2002), with<Mg/Si � 1 at 5–6 GPa.
The compositions of the added component calculatedusing the two starting compositions are reported inTable 7 and are similar. They show that the principal
261
changes are in the Si and Al content of peridotite, withlesser amounts of added Ca, Na, K, H and S. Carbon isalso added, but its abundance is not listed in table 7because its distribution does not permit a clear distinc-tion to be drawn between matrix and substrate. If allobserved carbonate pseudomorphs were introducedduring metasomatism, then about 4 wt.% CO2 shouldbe added to the composition listed in Table 7.
The Na2O and CaO contents listed in Table 7 areminimum values, because the Na and Ca introducedduring metasomatism may have dissipated beyond theboundaries of the sampled xenolith during subsequentequilibration of mineral compositions with surroundingmantle. However, the overall low CaO and Na2O con-tents of the minerals suggest that this is not a majoreffect. For example, if all the Na in BFT137 was derivedfrom the vein, the amount of Na2O estimated for theadded component (0.4–0.5 wt.%, Table 7) would onlyincrease by � 0.5 wt.%. The Na content of the vein isunlikely to originally have been much higher becausethis would require Na contents in original orthopyrox-ene that exceed values measured in mantle samples.Therefore, we conclude that only limited quantities ofNa and Ca were introduced. In the case of Al, a smallcorrection was made to the nominal estimate based uponmass balance between vein and host, assuming that ex-cess Cr in the vein had migrated into the vein duringequilibration and was balanced by an equimolar quan-tity of Al that migrated out.
The composition of the added component is com-pared in Fig. 8 to those of various melts and fluids fromexperimental studies and of some glass inclusions inxenoliths from a mantle wedge environment. The com-position of the added component is not likely to repre-sent the bulk composition of a fluid or melt phase, or itsanhydrous component, but rather the change in com-position undergone by such a phase during its equili-bration with the mantle. Fig. 8 shows that the addedcomponent is more Si-rich, and Al- and alkali-poor thanmelt inclusions in xenoliths (Schiano et al. 1995) andexperimental siliceous liquids in equilibrium with olivine+ orthopyroxene ± garnet ± phlogopite over a range
of upper mantle pressures (Sekine and Wyllie 1982,Carroll and Wyllie 1989, Johnston and Wyllie 1989,Rapp et al. 1999, Wunder and Melzer 2003). In partic-ular, the added component is poor in Na in comparisonto the experimental liquids and natural melt composi-tions from xenoliths. Natural granitoid magmas andsiliceous liquids produced by melting hydrous basalt areNa-rich and produce a Na-rich reaction assemblage inhybridization experiments (Sekine and Wyllie 1982).
The Na-poor composition of the added component inBFT137 and the lack of clinopyroxene in the reactionassemblage indicate that the parental fluid or liquiddiffered from natural granitoid magmas or the primarymelting products of subducted oceanic crust. A likelyexplanation for the difference is that the fluid or meltthat interacted with BFT137 had already reacted sub-stantially with the mantle. Garnet-phlogopite websteriteassemblages are present (though not abundant) in theBultfontein xenolith suite and attest to more Na-, Ca-and Al-rich reaction products in the general vicinity.Elsewhere, such assemblages have been linked to reac-tion of slab-derived melts (Aulbach et al. 2002). SampleBFT137 may therefore have been metasomatized by anevolved melt or dense hydrous fluid, derived from asiliceous melt that had undergone some prior reactionwith the mantle to increase its K2O/Na2O, SiO2/Al2O3,MgO/FeO and (SiO2+Al2O3)/(CaO+Na2O). Thisreaction would also alter the trace element signature ofthe fluid, leading, for example, to increasing LREE/HREE.
An alternative possibility that could account for thelack of Na and clinopyroxene in the reaction assemblageis that the parental fluid was formed by dehydration of
Fig. 8 Composition of added component projected in the system(Na2O+K2O) – Al2O3 – SiO2, compared with experimental liquidsderived from the equilibration of granitoid compositions withperidotite at 1.5–3.7 GPa (see text for references), and with naturalglass inclusions in peridotite xenoliths (Schiano et al. 1995). Theletters A and B represent the compositions listed in Table 7
Table 7 Calculated composition of added component (wt.%)
A B
SiO2 80 83Al2O3* 13 8.2CaO 1.4 1.4K2O 3.7 3.0Na2O** 0.45 0.40H2O 1.5 1.2SO2 0.68 0.54
1. Assumes BFT137 matrix as protolith, Mg+Fe as referencecomponent2. Assumes BFT147 (phlogopite-free) as protolith, Mg+Fe asreference component* Al2O3 adjusted for post-formation Cr–Al exchange (see text)** Na2O not adjusted for post-formation redistribution in opx andis therefore a minimum estimate
262
phengite in a subducting slab (Schmidt 1996, Domanikand Holloway 1996, Wunder and Melzer 2003), ratherthan by granitoid reaction-differentiation. Such fluidsalso react with peridotite to produce enstatite andphlogopite (Wunder and Melzer 2003). However, thepresence of garnet in addition to phlogopite in thereaction products of BFT137 suggests molar Al2O3/K2Ogreater than that predicted for a fluid resulting fromphengite breakdown to an anhydrous assemblage con-taining garnet and omphacite (Schmidt 1996). In orderto discriminate conclusively between the possible com-positions and genetic origins of the parental metaso-matizing agent it may be necessary to quantify the fullrange of chemical changes found in the Kimberleyxenoliths, and to assess their relative abundances.
Geological setting of Kimberley metasomatism
Recent studies in the western Kaapvaal craton indicatethat the Kimberley area from which our samples derivelies within a zone of Archean subduction and conti-nental collision. Isotopic evidence from crustal zircons,peridotite xenoliths and diamond inclusions suggestsconvergence of eastern and western Kaapvaal blocks at2.88 – 2.92 Ga, accompanied by mantle metasomatism,eclogite and diamond formation, and widespread latergranitoid magmatism (Richardson et al. 2001; Shireyet al. 2002; Poujol et al. 2003; Schmitz et al. 2004). Theevidence in BFT137 for mantle metasomatism by avolatile-rich siliceous fluid, as well as the geochemicalcharacteristics inferred for this fluid, are consistent withthe inferred proximity to a convergent plate boundary.Rb-Sr and Sm-Nd systematics of garnets suggest thatthe trace element enrichment in Kimberley harzburgitesand diamond inclusions is Archean (Richardson et al.1984, 1985), a conclusion supported by the large dis-tances over which post-metasomatic chemical equilib-rium was attained in sample BFT137.
The TRD of 2.97 ± 0.04 Ga for BFT137 is consistentwith the proposed tectono-magmatic history of theKimberley area. This minimum age may reflect meltdepletion in the mantle wedge approximately contem-poraneous with, or immediately prior to, collision.However, this age also permits mantle lithosphere for-mation in an earlier event as proposed by Griffin et al.(2004).
Significance for cratonic mantle petrogenesis
Kesson and Ringwood (1989) suggested that orthopy-roxene-rich peridotites of the Kaapvaal and other Ar-chean cratons were formed by the reaction of silicadissolved in hydrous, slab-derived fluids with olivine toform enstatite. Mibe et al. (1998) demonstrated experi-mentally the feasibility of a porous flow mechanism forhydrous fluid metasomatism of the mantle at pressuresgreater than 3 GPa. This observation provides impor-
tant support for the metasomatic hypothesis. Distinctveins of orthopyroxenite are not common within mantlexenolith suites from the Kaapvaal Craton and a meta-somatic origin therefore requires a dispersed medium toaccount for the relatively uniform distribution of silicaenrichment. The textural evidence in sample BFT137 isconsistent with these proposals, and the chemicalchanges documented in this sample are consistent withthe general sense in which Kaapvaal peridotites deviatefrom melting residue compositions (Walter 2004).
In a variation on this metasomatic theme, Rudnicket al. (1984) and Kelemen et al. (1998) proposed thatsiliceous melts, derived from the melting of a subductedslab could be the metasomatic agent. The textural evi-dence from BFT137 suggests that the metasomatic agenthad very low viscosity. In order for a siliceous melt toattain sufficiently low viscosity to effectively percolatealong grain boundaries as inferred in this particular case,high water content would be required. Complete misci-bility between water-saturated melts and hydrous fluidsfor siliceous systems at pressures relevant to the presentsample have been demonstrated (Shen and Keppler1997; Bureau and Keppler 1999), and their low viscosi-ties verified (Audetat and Keppler 2004). Sekine andWyllie (1982) argued that hydrous melts would evolve towater saturation and evolution of hydrous fluid duringtheir reaction with mantle peridotite. Thus, a continuumof melt-fluid compositions is likely to be generatedduring reactions of slab-derived mobile phases withoverlying mantle, and the chemical and mineralogicalfeatures of individual xenolith samples should reflect thisdiversity.
A range of effects at different extents of fluid-rockreaction is also expected (Navon and Stolper 1987;Harte et al. 1983; Bodinier et al. 2004). In the olivine-rich garnet peridotite BFT147, the lack of orthopyrox-ene growth, but relative abundance of phlogopite with atenfold increase in Cl content over that observed in theenstatite-rich veined sample, provides possible evidenceof such compositional evolution in the fluid. The garnetof this sample is also enriched in the highly incompatibleelements (U, Nb, Sr, LREE) relative to BFT137.
The occurrence of sulfides and carbonate pseud-omorphs in BFT137 suggests that significant quantities ofCO2 and S could have been introduced into the litho-spheric mantle by subduction–derived fluids. Reductionof CO2 and SO2 to sulfide and graphite or diamondprovides a mechanism to oxidize the lithospheric mantle.The restriction of sulfide to the vein itself suggests that itmay have been fixed rapidly by reaction of fluid-derivedSO2 or sulfate with FeO in the matrix minerals.
Although sample BFT137 likely represents one stagein a continuum of metasomatic products, its represen-tativeness for the Kaapvaal cratonic mantle as a wholemay be evaluated by considering that the SiO2/K2Oratio of the added component (using data from Table 7)is 25 ± 3. If it is assumed that the difference in SiO2
content of average Kaapvaal low-temperature peridotite(46.6 wt.% - Boyd 1989) was all derived by metasoma-
263
tism of a protolith resembling the BFT137 matrix, or theolivine-rich sample BFT147, by the reactions discussedabove, then a metasomatic K2O content in the range0.085 – 0.25 wt.% is implied. Such estimates are in therange of average measured K2O contents of Kaapvaalperidotites, but are not easily reconciled with thermalconstraints that require low mantle heat production(Rudnick et al. 1998). This suggests that the sampleexamined here represents a particularly K-rich part ofthe reaction process, supporting arguments above for arelatively evolved fluid and the existence elsewhere ofless K-rich siliceous reaction products.
By studying a wider range of metasomatic reactionproducts and determining their stratigraphic position inthe mantle by thermobarometry, the Kimberley xenolithsuite and others from surrounding mantle offer a po-tential opportunity to reconstruct details of an ancientsubduction zone and provide important constraints onArchean tectonic processes and geochemical cycles.Recognizing and accounting for metasomatic effects inthe bulk compositional trends of cratonic peridotitesshould also permit a more reliable assessment of theconditions under which the primary melt depletion wasacquired. The results of this study are consistent withmodels for cratonic lithosphere generation in a sub-duction zone setting (Kesson and Ringwood 1989, Par-man et al. 2004) but do not address directly the questionof where primary melt depletion occurred. However,indications that metasomatism decreases the Cr/Al ratioof cratonic peridotite strengthens arguments for meltingwithin the stability field of spinel (Kesson and Ringwood1989, Stachel et al. 1998).
Summary
The veined harzburgite xenolith BFT137 from Kimber-ley provides evidence for metasomatic introduction ofSi, Al, K, H, C and S into the cratonic mantle. Themetasomatic agent is inferred to be a hydrous fluid richin silica, which had evolved by prior reaction with sur-rounding mantle. The parent to this fluid originated inoceanic lithosphere, subducted during convergence be-tween eastern and western blocks of the KaapvaalCraton at � 2.9 Ga. Trace element enrichment patternsin garnet are qualitatively similar to others from Ka-apvaal xenoliths and suggest that the pervasive traceelement enrichment of the Kaapvaal mantle is effectedby fluids similar to those inferred here. These resultsimply that subduction processes played an importantrole in modifying the composition of Archean SCLM.
Acknowledgments This work was supported by US National Sci-ence Foundation grants EAR-9526702, EAR-9526840, EAR-0003533 and EAR-0310330. DRB gratefully acknowledges supportat MIT from a Crosby Fellowship. We thank Sam Bowring (MIT)for discussions, Anton le Roex (University of Cape Town) forhosting the ICP-MS analyses in his laboratory, the CarnegieInstitution of Washington for access to electron microprobe facil-
ities, and De Beers Consolidated Mines for field support and per-mission to sample the Kimberley dumps.
References
Anders E, Grevesse N (1989) Abundances of the elements: mete-oritic and solar. Geochim Cosmochim Acta 53: 197–214
Audetat D, Keppler H (2004) Viscosity of fluids in subductionzones. Science 303: 513–516
Aulbach S, Stachel T, Viljoen KS, Brey GP, Harris JW (2002)Eclogitic and websteritic diamond sources beneath the Limpo-po belt – is slab melting the link. Contrib Mineral Petrol 143:56–70
Bell DR, Gregoire M, Grove TL, Chatterjee N, Bowring, SA(2003) Silica and carbon deposition in the Kimberley perido-tites. Extd Abstr 8th International Kimberlite Conf VictoriaCanada, 4 pp (unpaginated CD)
Berg GW (1986) Evidence for carbonate in the mantle. Nature 324:50–51
Bodinier J-L, Menzies MA, Shimizu N, Frey FA, McPherson E(2004) Silicate, hydrous and carbonate metasomatism at Lherz,France: contemporaneous derivatives of silicate melt – harz-burgite reaction. J Petrol 45: 299–320
Boyd FR (1989) Compositional distinction between oceanic andcratonic lithosphere. Earth Planet Sci Lett 96: 15–26
Boyd FR, McCallister RH (1976) Densities of fertile and sterilegarnet lherzolites. Geophys Res Lett 3: 509–512
Boyd FR, Mertzman SA (1987) Composition and structure of theKaapvaal lithosphere. In: Mysen BO (ed) Magmatic processes:Physicochemical principles. Geochemical Society Spec Publ 1:13–24
Boyd FR, Pokhilenko NP, Pearson DG, Mertzman SA, SobolevNV, Finger LW (1997) Composition of the Siberian cratonicmantle: evidence from Udachnaya peridotite xenoliths. ContribMineral Petrol 128: 228–246
Brenan JM, Shaw HF, Ryerson FJ, Phinney DL (1995) Mineral-aqueous fluid partitioning of trace elements at 900�C and20 GPa: constraints on the trace element chemistry of mantleand deep crustal fluids. Geochim Cosmochim Acta 59: 3331–3350
Brey GP, Kohler T (1990) Geothermobarometry in four phaselherzolites II. New thermobarometers and practical assessmentof existing thermobarometers. J Petrol 31: 1353–1378
Bureau H, Keppler H (1999) Complete miscibility between silicatemelts and hydrous fluids in the upper mantle: experimentalgeochemical evidence and geochemical implications. EarthPlanet Sci Lett 165: 187–196
Burgess SR, Harte B (1999) Tracing lithosphere evolution throughthe analysis of heterogeneous G9/G10 garnets in peridotitexenoliths, I: major element chemistry. In: Gurney JJ, GurneyJL, Pascoe MD, Richardson SH (eds) The J.B. Dawson Vol-ume. Proc 7thInternat Kimberlite Conf, Red Roof Design,Cape Town, pp 66–80
Canil D (1991) Experimental evidence for the exsolution of cra-tonic peridotite from high-temperature harzburgite. Earth Pla-net Sci Lett 106: 64–72
Carlson RW, Moore RO (2004) Age of the eastern Kaapvaalmantle: Re-Os isotope data for peridotite xenoliths from theMonastery kimberlite. S Afr J Geol 107: 81–90
Carlson RW, Pearson DG, Boyd FR, Shirey SB, Irvine G, MenziesAH, Gurney JJ (1999) Re-Os systematics of lithospheric peri-dotites: implications for lithosphere formation and preserva-tion. In: Gurney JJ, Gurney JL, Pascoe MD, Richardson SH(eds) The J. B. Dawson Volume Proc 7th Internat KimberliteConf. Red Roof Design, Cape Town, pp 99–108
Carroll MR, Wyllie P J (1989) Experimental phase relations in thesystem peridotite – tonalite – H2O at 15 kbar; implications forassimilation and differentiation processes near the crust mantleboundary. J Petrol 30: 1351–1382
264
Cox KG, Smith MR and Beswetherick S (1987) Textural studies ofgarnet lherzolites: evidence of exsolution origin from high-temperature harzburgites. In Nixon PH (ed) Mantle Xenoliths.John Wiley and Sons, Chichester, pp 537–550
Dawson JB (2004) A fertile harzburgite to garnet lherzolite tran-sition: possible inferences for the roles of strain and metaso-matism in upper mantle peridotites. Lithos 77: 553–570
Dawson JB, Stephens WE (1975) Statistical classification of garnetsfrom kimberlite and associated xenoliths. J Geol 83: 589–607
Domanik KJ, Holloway JR (1986) The stability and compositionof phengitic muscovite and associated phases from 5.5 to11 GPa: implications for deeply subducted sediments. GeochimCosmochim Acta 60: 4133–4150
Erlank AJ, Waters FG, Hawkesworth CJ, Haggerty SE, AllsoppHL, Rickard RS, Menzies M (1987) Evidence for mantlemetasomatism in peridotite nodules from the Kimberley pipes,South Africa. In: Menzies M, Hawkesworth CJ (eds) Mantlemetasomatism. Academic Press, London, pp 221–309
Gaul O, Griffin WL, O’Reilly SY, Pearson NJ (2000) Mappingolivine composition in the lithospheric mantle. Earth Planet SciLett 182: 223–235
Gregoire M, Bell DR, le Roex AP (2002) Trace element geo-chemistry of phlogopite-rich mafic xenoliths: their classificationand their relationship to phlogopite-bearing peridotites andkimberlites revisited. Contrib Mineral Petrol 142: 603–625
Gregoire M, Bell DR, le Roex AP (2003) Garnet lherzolites fromthe Kaapvaal Craton (South Africa): trace element evidence fora metasomatic history. J Petrol 44: 629–657
Griffin WL, Smith D, Boyd FR, Cousens DR, Ryan CG, Sie SH,Suter GF (1989) Trace element zoning in garnets from shearedmantle xenoliths. Geochim Cosmochim Acta 53: 561–567
Griffin WL, O’Reilly SY, Ryan CG (1999) The composition andorigin of the sub-continental lithospheric mantle. In: Fei Y,Bertka CM, Mysen BO (eds) Mantle petrology: field observa-tions and high pressure experimentation: a tribute to Francis R.(Joe) Boyd. Geochemical Soc Spec Publ 6: 13–45
Griffin WL, O’Reilly SY, Abe N, Aulbach S, Davies RM, PearsonNJ, Doyle BJ, Kivi K (2003) The origin and evolution of Ar-chean lithospheric mantle. Precambrian Res 127: 19–41
Griffin WL, Graham S, O’Reilly SY, Pearson NJ (2004) Litho-sphere evolution beneath the Kaapvaal Craton: Re-Os sys-tematics of sulfides in mantle-derived peridotites. Chem Geol208: 89–118
Gurney JJ (1984) A correlation between garnets and diamonds. In:Glover JE, Harris PG (eds) Kimberlite occurrence and origin: abasis for conceptual models in exploration. Geology Dept andUniversity Extension, University of Western Australia Publ 8:143–166
Gurney JJ, Harte B (1980) Chemical variations in upper mantlenodules from southern African kimberlites. Phil Trans RoyalSoc London A297: 273–293
Harte B, Hunter RH, Kinny PD (1993) Melt geometry, movementand crystallization, in relation to mantle dykes, veins, andmetasomatism. Phil Trans R Soc Lond A342: 1–21
Herzberg C (1993) Lithosphere peridotites of the Kaapvaal craton.Earth Planet Sci Lett 120: 13–29
Hoal KEO, Hoal BG, Erlank AJ, Shimizu N (1994) Metasomatismof the mantle lithosphere recorded by rare-earth elements ingarnet. Earth Planet Sci Lett 126: 303–313
Hops JJ, Gurney JJ, Harte B, Winterburn P (1989) Megacrysts andhigh temperature nodules from the Jagersfontein kimberlitepipe. In : Ross J et al (eds) Kimberlites and Related rocks.Volume 2. Their mantle/crust setting, diamonds and diamondexploration. Geol Soc Australia Spec Publ 14: 759–770
Johnston AD, Wyllie PJ (1989) The system tonalite-peridotite-H2Oat 30 kbar, with applications to hybridization in subductionzone magmatism. Contrib Mineral Petrol 102: 257–264
Jones AP, Smith JV, Dawson, JB (1982) Mantle metasomatism in14 veined peridotites from Bultfontein mine, South Africa. JGeol 90: 435–453
Jordan TH (1988) Structure and formation of the continental tect-osphere. Journal of Petrology Special Lithosphere Issue: 11–37
Kelemen PB (1990) Reaction between ultramafic wallrock andfractionating basaltic magma Part I Phase relations, the originof calc-alkaline magma series and the formation of discordantdunite. J Petrol 31: 51–98
Kelemen PB (1995) Generation of high-Mg andesites and thecontinental crust. Contrib Mineral Petrol 120: 1–19
Kelemen PB, Shimizu N, Dunn JT (1993) Relative depletion ofniobium in some arc magmas and the continental crust: parti-tioning of K, Nb, La and Ce during melt/rock reaction in theupper mantle. Earth Planet Sci Lett 120: 111–133
Kelemen PB, Hart SR, Bernstein S (1998) Silica enrichment in thecontinental upper mantle via melt/rock reaction. Earth PlanetSci Lett 164: 387–406
Kepezhinskas PK, Defant MJ, Drummond MS (1995) Na meta-somatism in the island-arc mantle by slab melt – peridotiteinteraction: evidence from mantle xenoliths in the NorthKamchatka arc. J Petrol 36: 1505–1527
Kepezhinskas PK, Defant MJ, Drummond MS (1996) Progressiveenrichment of island arc mantle by melt-peridotite interactioninferred from Kamchatka xenoliths. Geochim Cosmochim Acta60: 1217–1229
Kesson SE, Ringwood AE (1989) Slab – mantle interactions. 2. Theformation of diamonds. Chem Geol 78: 97–118
Kinzler RJ, Grove TL (1999) Origin of depleted cratonic harz-burgite by deep fractional melt extraction and shallow olivinecumulate infusion. In Gurney JJ, Gurney JL, Pascoe MD,Richardson SH (eds) The J.B. Dawson Volume. Proc 7th In-ternat Kimberlite Conf. Red Roof Design, Cape Town, pp 437–443
Kopylova MG, Russell JK (2000) Chemical stratification of cra-tonic lithosphere: constraints from the Northern Slave Craton,Canada. Earth Planet Sci Lett 181: 71–87
Kramers JD Roddick JCM, Dawson JB (1983) Trace element andisotopic studies on veined, metasomatic and ‘‘MARID’’ xeno-liths from Bultfontein. South Africa Earth Planet Sci Lett 65:90–106
Lahaye Y, Brey GP (2003) Scale and timing constraints on chem-ical redistribution between minerals of a composite garnetperidotite/orthopyroxenite. Abstr 8th Internat Kimberlite ConfVictoria, Canada, CD unpaginated
Laurora A, Mazzucchelli M, Rivalenti G, Vannucci R, Zanetti A,Barbier MA, Cingolani C (2001) Metasomatism and melting incarbonated peridotite xenoliths for the mantle wedge: the Go-bernador Gregores case (southern Patagonia). J Petrol 42: 69–87
Lee WJ, Wyllie PJ (2000) The system CaO-MgO-SiO2-CO2 at1 GPa, metasomatic wehrlites, and primary carbonatite mag-mas. Contrib Mineral Petrol 138: 214–228
Lorand JP (1993) Comment on ‘Content and isotopic compositionof sulphur in ultramafic xenoliths from central Asia’ by D. A.Ionov, J. Hoefs, K. H. Wedepohl and U. Wiechert. EarthPlanet Sci Lett 119: 627–634
McDonough WF, Sun S (1995) The composition of the Earth.Chem Geol 120: 223–253
McInnes BIA, Cameron EM (1994) Carbonated, alkaline hybrid-izing melts from a sub-arc environment: mantle wedge samplesfrom the Tabar-Lihir-Feni arc, Papua New Guinea. EarthPlanet Sci Lett 122: 125–141
McInnes BIA, Gregoire M, Binns RA, Herzig PM, HanningtonMD (2001) Hydrous metasomatism of oceanic sub-arc mantle,Lihir, Papua New Guinea: petrology and geochemistry of fluid-metasomatised mantle wedge xenoliths. Earth Planet Sci Lett188: 169–183
Mibe K, Fujii T, Yasuda A (1998) Connectivity of aqueous fluid inthe Earth’s upper mantle. Geophys Res Lett 25: 1233–1236
Mibe K, Fujii T, Yasuda A (2002) Composition of aqueous fluidcoexisting with mantle minerals at high pressure and its bearingon the differentiation of the Earth’s mantle. Geochim Cosmo-chim Acta 66: 2273–2285
Navon O, Stolper EM (1987) Geochemical consequences of meltpercolation: the upper mantle as a chromatographic column. JGeol 95: 285–307
265
Niu Y, Langmuir CH and Kinzler RJ (1997) The origin of abyssalperidotites; a new perspective. Earth Planet Sci Lett 152: 251–265
O’Hara MJ, Saunders MJ, Mercy, ELP (1975) Garnet-peridotiteprimary ultrabasic magma and eclogite; interpretation ofthe upper mantle process in kimberlite. Phys Chem Earth 9: 571–604
O’Neill HS (1980) An experimental study of Fe-Mg partitioningbetween garnet and olivine and its calibration as a geother-mometer: corrections. Contrib Mineral Petrol 72: 337
O’Neill HS, Wood BJ (1979) An experimental study of Fe-Mgpartitioning between garnet and olivine and its calibration as ageothermometer. Contrib Mineral Petrol 70: 59–70
Parman SW, Grove TL, De Wit MJ, Dann JC (2004) A subductionorigin for komatiites and cratonic lithospheric mantle. S Afr JGeol 107: 107–118
Pearson DG, Carlson RW, Shirey SB, Boyd FR, Nixon PH (1995a)Stabilisation of Archaean lithospheric mantle; a Re-Os isotopestudy of peridotite xenoliths from the Kaapvaal Craton. EarthPlanet Sci Lett 134: 341–357
Pearson DG, Shirey SB, Carlson RW, Boyd FR, Pokhilenko NP,Shimizu N (1995b) Re-Os, Sm-Nd, and Rb-Sr isotope evidencefor thick Archaean lithospheric mantle beneath the Siberiancraton modified by multistage metasomatism. Geochim Cos-mochim Acta 59: 959–977
Pokhilenko NP, Sobolev NV, Boyd FR, Pearson DG, Shimizu N(1993) Megacrystalline pyrope peridotites in the lithosphere ofthe Siberian Platform; mineralogy, geochemical peculiarities andthe problem of their origin. Russian Geol Geophys 34: 56–67
Poujol M, Robb LJ, Annhaeusser CJ, Gericke B (2003) A review ofthe geochronological constraints on the evolution of the Ka-apvaal Craton, South Africa. Precambrian Res 127: 181–213
Prouteau G, Scaillet B, Pichavant M, Maury R (2001) Evidence formantle metasomatism by hydrous silicic melts derived fromsubducted crust. Nature 410: 197–200
Rapp RP, Shimizu N, Norman MD, Applegate GS (1999) Reac-tion between slab-derived melts and peridotite in the mantlewedge: experimental constraints at 38 GPa. Chem Geol 160:335–356
Richardson SH, Gurney JJ, Erlank AJ, Harris JW (1984) Origin ofdiamonds in old enriched mantle. Nature 310: 198–202
Richardson SH, Erlank AJ, Hart SR (1985) Kimberlite-bornegarnet peridotite xenoliths from old enriched subcontinentallithosphere. Earth Planet Sci Lett 75: 116–128
Richardson SH, Shirey S B, Harris JW, Carlson RW (2001) Ar-chean subduction recorded by Re-Os isotopes in eclogitic sul-fide inclusions in Kimberley diamonds. Earth Planet Sci Lett191: 257–266
Rudnick RL, McDonough WF, Orpin A (1994) Northern Tanza-nian peridotite xenoliths: a comparison with Kaapvaal perido-tites and inferences on metasomatic interactions. In: MeyerHOA, Leonardos O (eds) Kimberlites, related rocks and mantlexenoliths, Vol 1 Proc 5th Internat Kimberlite Conf CPRMBrasilia pp 336–353
Rudnick RL, McDonough WF, O’Connell RJ (1998) Thermalstructure, thickness and composition of continental lithosphere.Chem Geol 145: 395–411
Saltzer RL, Chatterjee N, Grove TL (2001) The spatial distributionof garnets and pyroxenes in mantle peridotites; pressure-tem-perature history of peridotites from the Kaapvaal Craton. JPetrol 42:2215–2229
Schiano P, Clochiatti R, Shimizu N, Maury RC, Jochum KP,Hoffman AW (1995) Hydrous, silica rich melts in the sub-arcmantle and their relationship with erupted arc lavas. Nature377: 595–600
Schmidt MW (1996) Experimental constraints on recycling ofpotassium from subducted oceanic crust. Science 270: 625–627
Schmitz MD, Bowring SA, de Wit MJ, Gartz V (2004) Neoarcheancontinent collision in the western Kaapvaal craton, southernAfrica and its implications for the stabilization of continentallithosphere. Earth Planet Sci Lett 222: 363–376
Schneider M E, Eggler D H (1986) Fluids in equilibrium withperidotite minerals: implications for mantle metasomatism.Geochim Cosmoschim Acta 50: 711–724
Sekine T, Wyllie PJ (1982) The system granite – peridotite – H2O at30 kbar, with applications to hybridization in subduction zonemagmatism. Contrib Mineral Petrol 81: 190–202
Shapiro SS, Hager B H, Jordan T H (1999) Stability and dynamicsof the continental tectosphere. Lithos 48: 115–133
Shen AH, Keppler H (1997) Direct observation of complete mis-cibility on the albite-H2O system. Nature 385: 710–712
Shimizu N (1975) Rare earth elements in garnets and clinopyrox-enes from garnet lherzolite nodules in kimberlites. Earth PlanetSci Lett 25: 26–32
Shimizu N, Richardson SH (1987) Trace element abundance pat-terns of garnet inclusions in peridotite-suite diamonds. Geo-chim Cosmochim Acta 51: 755–758
Shimizu N, Pokhilenko NP, Boyd FR, Pearson DG (1999) Traceelement characteristics of garnet dunites/harzburgites, hostrocks for Siberian peridotitic diamonds. Gurney JJ, Gurney JL,Pascoe MD, Richardson SH (eds) The P.H. Nixon VolumeProc 7th Internat Kimberlite Conf. Red Roof Design, CapeTown, pp 773–782
Shirey SB, Harris JW, Richardson SH, Fouch MJ, James DE,Cartigny P, Deines P, Viljoen F (2002) Diamond genesis, seis-mic structure, and evolution of the Kaapvaal-Zimbabwe Cra-ton. Science 297: 1683–1686
Simon NSC, Irvine GJ, Davies GR, Pearson DG, Carlson RW(2003a) The origin of garnet and clinopyroxene in ‘‘depleted’’Kaapvaal peridotites. Lithos 71: 289–322
Simon NSC, Carlson RW, Davies G R, Nowell GM, Pearson D G(2003b) Os-Sr-Nd-Hf isotope evidence for the ancient depletionand subsequent multistage enrichment history of the Kaapvaalcratonic lithosphere Extd Abstrs 8th Internat Kimberlite ConfVictoria Canada, CD unpaginated
Smith D (2000) Insights into the evolution of the uppermost con-tinental mantle from xenolith localities on and near the Colo-rado Plateau and regional comparisons. J Geophys Res 105:16,769–16,781
Smith D, Riter JCA, Mertzman SA (1999) Water-rock interactions,orthopyroxene growth, and Si-enrichment in the mantle: evi-dence in xenoliths from the Colorado Plateau, southwesternUnited States. Earth Planet Sci Lett 165: 45–54
Smith D, Connelly JN, Manser K, Moser D, Housh T, McDowellF, Mack LE (2004) Evolution of Navajo eclogites and hydra-tion of the mantle below the Colorado Plateau, southwesternUnited States. Geochem Geophys Geosystems 5 (4): Q04005doi:10.1029/2003GC000675
Stachel T, Viljoen KS, Brey GP, Harris JW (1998) Metasomaticprocesses in lherzolitic and harzburgitic domains of diamond-iferous lithospheric mantle: REE in garnets from xenolithsand inclusions in diamonds. Earth Planet Sci Lett 159: 1–12
Stachel T, Harris JW, Brey GP (1999) REE patterns of peridotiticand eclogitic inclusions in diamonds from Mwadui (Tanzania).In. Gurney JJ, Gurney JL, Pascoe MD, Richardson SH (eds)The NixonVolume Proc 7th Internat Kimberlite Conf, RedRoof Design, Cape Town, pp 829–835
Stachel T, Aulbach S, Brey GP, Harris JW, Leost I, Tappert R,Viljoen KS (2004) The trace element composition of silicateinclusions in diamonds: a review. Lithos 77:1–21
Stalder R, Foley SF, Brey GP, Horn I (1998) Mineral-aqueous fluidpartitioning of trace elements at 900–1200�C and 30–57 GPa:new experimental data for garnet, clinopyroxene, and rutile,and implications for mantle metasomatism. Geochim Cosmo-chim Acta 62: 1781–1801
Stalder R, Ulmer P, Thompson AB, Gunther D (2001) Highpressure fluids in the system MgO-SiO2-H2O under uppermantle conditions. Contrib Mineral Petrol 140: 607–618
Tomoaki M, Arai S, Green DH (2003) Evolution of low-Alorthopyroxene in the Horoman Peridotite Japan; an unusualindicator of metasomatizing fluids. J Petrol 44: 1237–1246
266
Wagner TP, Grove TL (1998) Melt/harzburgite reaction in thepetrogenesis of tholeiitic magma from Kilauea Volcano, Ha-waii. Contrib Mineral Petrol 131: 1–12
Walker RJ, Carlson RW, Shirey SB, Boyd FR (1989) Os, Sr, Nd,and Pb isotope systematics of southern African peridotitexenoliths: implications for the chemical evolution of subconti-nental mantle. Geochim Cosmochim Acta 53: 1583–1595
Walter MJ (1998) Melting of garnet peridotite and the origin ofkomatiite and depleted lithosphere. J Petrol 39: 29–60
Walter MJ (1999) Melting residues of fertile peridotite and theorigin of cratonic lithosphere. In Fei Y, Bertka CM, Mysen BO(eds) Mantle petrology: field observations and high pressureexperimentation: a tribute to Francis R (Joe) Boyd Geochem-ical Soc Spec Publ 6: 225–239
Walter MJ (2004) Melt extraction and compositional variability inmantle lithosphere. In: Carlson RW (ed) The Mantle and Core,
Vol 2 Treatise on Geochemistry (eds HD Holland and KKTurekian), Elsevier-Pergamon, Oxford, pp 363–394
Wunder B, Melzer S (2003) Experimental evidence on phlogopitemantle metasomatism induced by phengite dehydration. Eur JMineral 15: 641–647
Wyllie PJ, Carroll MR, Johnston AD, Rutter MJ, Sekine T, Vander Laan SR (1989) Interactions among magmas and rocks insubduction zone regions: experimental studies from slab tomantle to crust. Eur J Mineral 1: 165–179
Wyllie PJ, Ryabchikov ID (2000) Volatile components, magmas,and critical fluids in upwelling mantle. J Petrol 41: 1195–1206
Zanetti A, Mazzucchelli M, Rivalenti G, Vannucci R (1999) TheFinero phlogopite-peridotite massif: an example of subduction-related metasomatism. Contrib Mineral Petrol 134: 107–122
267
