geochronological and geochemical constraints on the ... · the roberts victor mine is located...

Available online at www.sciencedirect.com

www.elsevier.com/locate/gca

Geochimica et Cosmochimica Acta 113 (2013) 1–20

Geochronological and geochemical constraintson the formation and evolution of the mantle underneath

the Kaapvaal craton: Lu–Hf and Sm–Nd systematicsof subcalcic garnets from highly depleted peridotites

Qiao Shu a,b,⇑, Gerhard P. Brey b, Axel Gerdes b, Heidi E. Hoefer b

a China University of Geosciences (Beijing), Xueyuan Road 29, Haidian, Beijing 100083, Chinab Institut fur Geowissenschaften, FE Mineralogie, Goethe Universitat Frankfurt, Altenhoferallee 1, 60438 Frankfurt, Germany

Received 9 June 2012; accepted in revised form 8 March 2013; available online 23 March 2013

Abstract

Subcalcic garnets carry the major inventory of most trace elements of their host harzburgites and are thus proxies of thebulk composition. We used single garnet grains from heavy mineral concentrates from the Kaapvaal craton (Roberts Victorand Lace mine) to determine the major and trace elements and the Sm–Nd and Lu–Hf isotope systematics of these highlydepleted members of the peridotitic suite. The combination of the results with previous work from the Finsch mine (Lazarovet al., 2009a) allowed us to reconstruct the formation and evolution of the mantle underneath the Kaapvaal craton.

Several lines of evidence from major and trace elements suggest that mantle melting was mainly at shallow pressures fol-lowed by subduction into the garnet stability field. A 3.22 Ga metasomatic event underneath the East block occurred in apreviously depleted mantle (eHf = +16) which was sufficiently stabilized by that time to hold a crust with tonalite–trondhje-mite–granodiorite (TTG’s) and greenstone belts. Such high eHf values can be produced within a few hundred million years by25% non-modal fractional melting in the spinel stability field. This is the first prove of a mantle underneath the East blockwith an age similar to a 3.65 Ga crustal age.

Before the amalgamation around 2.88 Ga, oceanic lithosphere was created between the W- and E-block around 2.95 Ga(group RV1 samples from the Roberts Victor) and subducted underneath the W-block. Another mantle portion (group RV2garnets) from Roberts Victor yielded a seeming age of 3.27 ± 0.15 Ga with eHf = +17.6. It actually results from an enrich-ment process in a highly depleted mantle about 2.8–2.9 Ga ago. This may have been the depleted mantle wedge above thesubduction and final collision between the West and the East block. The creation of a cratonic nucleus for the West-blockis unknown until 3.2 Ga when the oldest mantle TRD and oldest crustal zircon ages are reported. Such ages were not directlyobtained from the study of the subcalcic garnets but the high positive eHf value (+25) of the 2.62 Ga enrichment age fromFinsch and the existence of a highly depleted mantle wedge before 2.8–2.9 Ga at Roberts Victor suggests that the depletionoccurred several hundred million years earlier. Subsequently, the 2.6–2.8 Ga old voluminous Ventersdoorp volcanics pouredover a major part of the unified Kaapvaal craton, an event which also may have modified portions of the mantle underneathFinsch. The attachment of the Kheis–Magondi belt to the Kaapvaal craton from the West caused further modification around1.90 Ga. The latest stages of craton-scale metasomatism recorded in the subcalcic garnets lie between 0.9 and 1.3 Ga as can beseen in the Sm–Nd isotope systematic. This age range coincides with the Namaqua–Natal belt orogeny. The present study andprevious work on subcalcic garnets show that they can be excellent recorders of multiple mantle events which can be corre-lated with the tectonomagmatic evolution of the craton.� 2013 Elsevier Ltd. All rights reserved.

0016-7037/$ - see front matter � 2013 Elsevier Ltd. All rights reserved.

http://dx.doi.org/10.1016/j.gca.2013.03.010

⇑ Corresponding author at: Institut fur Geowissenschaften, FE Mineralogie, Goethe Universitat Frankfurt, Altenhoferallee 1, 60438Frankfurt, Germany. Tel.: +49 (0)69 798 40136; fax: +49 (0)69 798 40121.

E-mail address: [email protected] (Q. Shu).


mailto:[email protected]


2 Q. Shu et al. / Geochimica et Cosmochimica Acta 113 (2013) 1–20

1. INTRODUCTION

The timing of growth and amalgamation of Archeancontinental nuclei, and the temporal relationships betweenprocesses in the mantle and the overlying crust are funda-mental questions concerning stabilization of long-livedcratonic blocks by buoyant underlying lithospheric mantlekeels. Most of the evidence for crustal evolution comesfrom U–Pb zircon ages and for the mantle from theRe–Os system (Re depletion TRD and Re model agesTMA) (Walker et al., 1989). For the Kaapvaal craton withthe West-block (W-block) and the East-block (E-block) asthe two major building units (Fig. 1a) there is broad over-lap of the timing of crustal building events and mantle Redepletion (Pearson et al., 1995; Carlson et al., 1999;Menzies et al., 1999; Irvine et al., 2001; Carlson andMoore, 2004; Griffin et al., 2004; Pearson and Nowell,2004; O’Reilly et al., 2008). The amalgamation betweenthe two blocks occurred at around 2.9 Ga whereby theE-block was subducted underneath the W-block in a mod-ern style subduction zone scenario (Schmitz et al., 2004).The TRD’s from both the W- and E-block form a contin-uum between 2.5 and 3.2 Ga with a peak at around2.9 Ga. In contrast crustal growth is episodic. On theW-block, these episodes fall mainly between 3.25 and2.8 Ga (Drennan et al., 1990; Robb et al., 1990; Thomaset al., 1993; Poujol et al., 2002; Schmitz et al., 2004) andoverlap between crust and mantle ages exists. Quite differ-ently, crustal ages on the E-block are older than the TRD’sof the underlying mantle and episodic between 3.65 and3.1 Ga (summaries by Armstrong et al., 1991; de Witet al., 1992; Eglington and Armstrong, 2004). The oldestTRD ages of mantle peridotites from Northern Lesothoand the Monastery mine in the south of the E-block reachonly 3.25 and 3.02 Ga respectively (Carlson and Moore,2004). The existence of older crustal rocks led to sugges-tions that the subcontinental lithospheric mantle (SCLM)was formed after the crust or an older SCLM was replacedduring younger events or was completely reset by metaso-matism (Carlson et al., 1999; Irvine et al., 2001; Shireyet al., 2004; Gao et al., 2008; Griffin et al., 2004).

The disparity on the E-block of crustal and mantle agesmay be due to a sampling problem: (i) sampling localitiesfor mantle and crustal material often do not coincide espe-cially on the E-block; (ii) the crust is sampled laterally whilethe mantle is sampled vertically; (iii) the mantle is sampledmore comprehensively on the W-block with its numerousxenoliths bearing kimberlite diatremes than on the E-block;and (iv) the Archean crust is much better exposed on the E-block than on the W-block. These drawbacks are inherentto the geological situation and they may be overcome onlyin special cases, e.g. where crustal and mantle xenoliths aresampled together by the kimberlite. Insight into whether thecontinuous Re–Os age spectra reflect prolonged craton for-mation or metasomatic overprinting and whether the lackof Re–Os ages corresponding to the old crustal ages ofthe E-block is due to extensive metasomatism or mantlereplacement may be obtained from isochron informationfrom other isotopic systems, such as the Lu–Hf isotope sys-tem. Isochrons have the advantage of providing both age

and initial ratios. The initial isotope ratios indicate specificsof the sources and the petrogenetic history. The Lu–Hf sys-tem has emerged as very promising in that respect (Schereret al., 1997; Blichert-Toft et al., 1999; Ionov and Weiss,2002; Schmidberger et al., 2002; Carlson et al., 2004; Wittiget al., 2007; Lazarov et al., 2009a, 2012b). These authorsused Lu–Hf isotopic compositions from measured or calcu-lated bulk rocks or clinopyroxene separates or single grainsubcalcic garnets (the latter two as representatives of a bulkrock) to derive isochrons which date depletion or enrich-ment events. The correct interpretation of enrichment ordepletion can only be derived from a joint considerationof the isotope ratios and trace element abundances (seeSection 5).

Subcalcic garnets (often presented in CaO–Cr2O3 corre-lation diagrams like Fig. 1b) stem from clinopyroxene-freeharzburgites and are found in such xenoliths, as inclusionsin diamonds and in heavy mineral concentrates from kim-berlites. In the absence of clinopyroxenes, such garnets car-ry the major proportion of the incompatible lithophileelements of the bulk rock and potentially record meltingand enrichment events of the lithospheric mantle. Error-and isochrons from subcalcic garnets and reconstructedbulk rocks were successfully used to unravel multiple deple-tion and enrichment events at Finsch mine, South Africa,by the combined use of the Lu–Hf and Sm–Nd isotope sys-tems (Lazarov et al., 2009a, 2012a,b). The present study ex-tends this approach to other localities on the Kaapvaalcraton (Roberts Victor and Lace mine) to test whether (1)these results can be representative for the craton-wideKaapvaal lithospheric mantle, (2) mantle processes can bedated more precisely and (3) there are age differences inprocesses between and within the cratonic blocks. We haveconcentrated on mantle samples from the older group IIkimberlite generation (early Cretaceous) because thosefrom the younger group I kimberlites (late Cretaceous) ap-pear to be have been metasomatized by the preceding kim-berlite generation as their isotopic compositions wereheavily disturbed and a time span of 30 Mio years wasnot sufficient to reach full re-equilibration within thesemantle portions (Simon et al., 2007).

2. LOCALITIES, SAMPLES, DEPTH OF ORIGIN AND

GEOLOGICAL BACKGROUND

The Finsch mine is located close on the western borderof the W-block with the Kheis–Okwa–Magondi belt(Fig. 1a). Lazarov et al. (2009a) studied the trace elementand isotope geochemistry of the subcalcic garnets thereand derived a Lu–Hf isochron and Sm–Nd errorchrons.In this study, we have used their data set for further evalu-ation (see Section 4). These authors also carried out anextensive trace element and isotope chemistry study ongarnet peridotites from this locality and derived similarconclusions as those drawn from the single grain subcalcicgarnet study (Lazarov et al., 2012a,b).

The Roberts Victor mine is located fairly central in theKaapvaal craton very close to the Colesberg lineament,i.e. close to the collision zone between the E- and W-blocks.Jelsma et al. (2004) place it to the east of the Colesberg lin-

Fig. 1. (a) Structural units of the Kaapvaal craton and color coded fields of exposed Paleo- to Neoarchean crystalline basement, volcanic andmagmatic rocks with age divisions schematic after Eglington and Armstrong (2004), Anhaeusser (2006) with interpretations after Jacobs et al.(2008). Also shown are the locations of the Newlands, Kimberley, Finsch, Roberts Victor and Lace mines. The conditions of origin andchemical parameters of the subcalcic garnets studied from these three localities are shown in the following diagrams as gray squares, bluediamonds and red triangles respectively. (b) Equilibration conditions of the subcalcic garnets calculated as the averages of the Ni-in-garnetthermometers of Canil (1999) and Griffin et al. (1989) and projected onto a 40 mW/m2 conductive geothermal gradient of Chapman andPollack (1977). There is no distinction made into further subgroups because P, T conditions are statistically distributed within the variousgroups. (c) CaO–Cr2O3 correlation diagram of the subcalcic garnets; the symbols are split up into subgroups of the various localities. (d)Subcalcic garnets plotted into a log(Lu/Hf)C1–(Lu/Er)C1 diagram. Symbols are as in (b). These ratios are equal to those of the bulk host rock.Both ratios increase in partial melting residues and plot in quadrant I. Garnets from this quadrant were selected for isotopic work becausethey record the nature of the partial melting process most closely. Model calculations for non-modal fractional melting in the garnet (5 GPa;pink line) and spinel (1.5 GPa; green line) stability are shown for comparison. Partial melting model calculations were carried out usingpartition coefficients Di (for element i) from Green et al. (2000) for the garnet stability field and for the spinel stability field from Kelemen et al.(1993) and Suhr et al. (1998). Melting modes were taken from Kinzler (1997) for spinel peridotite and Walter (1998) for garnet peridotitemelting. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Q. Shu et al. / Geochimica et Cosmochimica Acta 113 (2013) 1–20 3

eament, i.e. on the westernmost border of the E-block butits exact belonging remains ambiguous. The Lace mine issituated slightly to the west of the centre of the E-block(Fig. 1a). About 100 purple to violet garnets with grain sizes>2 mm in diameter were preselected from heavy mineralconcentrates from the Lace mine and about 300 were pres-elected from the Roberts Victor mine. Eighteen subcalcicgarnets were identified from Lace by electron probe microanalysis from their CaO–Cr2O3 relationships and 36 fromRoberts Victor (Fig. 1b; Table A.1). Depths of origin wereestimated from the average temperatures of the Ni in garnet

thermometers of Griffin et al. (1989) and Canil (1999) whichwere projected onto a conductive geothermal gradient of40 mW/m2 (Chapman and Pollack, 1977) (Fig. 1b). Thesubcalcic garnets from Lace mine give a depth range from140 to 185 km (only one sample from 120 km) and thosefrom Roberts Victor from 125 to 175 km. The samples fromthe Finsch mine come from a restricted depth range of150–180 km, a range which is in accord with that obtainedfrom garnet lherzolites from Finsch (Lazarov et al., 2009b).There is no further division in this diagram into the varioussubgroups which were derived from each locality. No further


information would be obtained this way because P and T

conditions are statistically distributed within each group.The depth ranges from the various localities are in agreementwith the empirical Cr-barometer of Grutter et al. (2006). Inthe absence of spinel this barometer gives only minimumpressures. We do not know whether our garnets were coexis-ting with spinel but about half of the higher pressure samplesfrom Finsch and Roberts Victor give similar high pressureswith the Cr-barometer as obtained from the projection ontothe conductive geothermal gradient with the averaged Ni-thermometers. These could be spinel saturated.

Trace elements were measured in all selected garnets byLaser Ablation Inductively Coupled Plasma Mass Spec-trometry (LA ICP MS). With a few exceptions only garnetsplotting in quadrant I on a (Lu/Hf)C1 versus (Lu/Er)C1 dia-gram (Fig. 1d) were analyzed for Sm–Nd and Lu–Hf iso-tope compositions. Quadrant I corresponds to the field ofresidues of partial melting of a peridotitic mantle. The de-gree of chemical overprint of these garnets by reenrichmentprocesses should be low compared to garnets in the otherquadrants. They should provide the best opportunity todate partial melting or re-enrichment processes or both inthe mantle. Seven samples from Lace and 25 from RobertsVictor plot into quadrant I. Twelve of the 25 samples fromRoberts Victor (representative for the spread of Lu/Hf andLu/Er ratios) were analyzed for their isotopic compositionsand also the seven Lace samples from quadrant I and fourLace samples from quadrant II.

3. ANALYTICAL METHODS (DETAILS IN THE

SUPPLEMENTARY FILE – METHOD DESCRIPTION)

3.1. Major element analysis

Analyses of the major and trace elements and the Sm–Ndand Lu–Hf isotopic systems were carried out as described byLazarov et al. (2009a); detailed information is also suppliedin the Supplementary file of the present paper). Major ele-ments were measured in the wave length dispersive mode(WDS) with a JEOL JXA 8900RL electronprobe microana-lyser. Three to six spots on each of several pieces of thecrushed garnet grains were mounted in epoxy, polished andanalyzed to establish compositional homogeneity. For de-tails and accuracy see electronic appendix).

3.2. Trace element analysis

The trace elements were analyzed by Laser ablation ICPMS using a New Wave Research LUV213e ultraviolet Nd-YAG laser coupled with a Finnigan Element 2 at the Goe-the University in Frankfurt. The laser was used at a pulsefrequency of 10 Hz and an energy pulse of 0.6–0.8 mJ (cor-responding to 60–80% laser power) with spot sizes of 60–95 lm. NIST 612 glass was used as a calibration standard.USGS BIR-1 glass (concentrations from Eggins et al., 1997)was the external and Ca from microprobe analysis of theminerals the internal standard. BIR-1 glass and one in-house garnet standard (PN2b) were measured several timeswithin each sequence. The raw data were processed on-lineusing the GLITTER software.

3.3. Sample preparation and isotope analysis

For analysis of the Lu–Hf and Sm–Nd isotopic systemsthe crushed garnet grains were first leached at room temper-ature in 6 M HCl or HNO3 in an ultrasonic bath for abouthalf an hour. The dried down fragments were hand-pickedto optical purity and, if necessary, the acid ultrasonication re-peated and the grains re-picked. Finally, 30–150 mg of garnetseparates were ultrasonicated with MQ H2O and dried priorto spiking with 176Lu–180Hf and 149Sm–150Nd tracers beforedissolution. Our total procedural Hf blank measured for ourmethods were from 25 ± 5 pg for repeat measurements byisotope dilution (ID) ICP MS, which is necessary for mea-surement of the very low Hf concentrations. For those sam-ples with Hf lower than 5 ng, blank corrections were carefullyapplied. Our total procedural Lu blank was 5 ± 2 pg. Theisotope ratios were analyzed in a static mode with a multi col-lector ICP-MS (Finnigan Neptune). The sample amounts ofHf were always more than 3 ng. With our instrumental set upwe are able to measure such low concentrations of Hf (3 ngresult in a signal of �90 mV on 176Hf) with a precision ofaround 1e (on 176Hf/177Hf). Repeated measurements of theHf standard JMC475 produced 176Hf/177Hf of0.282153 ± 0.000014 (2r), which is in good agreement withthe literature (Blichert-Toft et al., 1997; Chu et al., 2002).Replicate digestion and analysis of BHVO-1 gave176Hf/177Hf = 0.283107 ± 0.000013 (2r) (Table A.4) whichis in good agreement with the values reported by Blichert-Toft, 2001 and Bizzarro et al., 2003. Nd-isotope and Sm-ID measurements were also performed by MC ICP-MS.The Nd aliquots always contained more than 10 ng Nd. Suchamounts could be measured with a precision of better than0.5e as determined by replicate standard measurements. Pro-cedure blanks were lower than 70 pg for Nd and 25 pg forSm. Repeated measurements of Nd isotope standardsyielded 143Nd/144Nd = 0.511725 ± 0.000068 (2r) for MerckNd2O3 and 0.5119538 ± 0.000026 for Ames-Nd. This is ingood agreement with literature values (Deckart et al., 2005;Caro et al., 2006). Replicate digestion and analysis ofBHVO-1 yielded 143Nd/144Nd = 0.512938 ± 0.000017 (2r)(Table A.6). Uncertainties of Lu–Hf and Sm–Nd ratiosand isotope compositions are based on replicate BHVO-1standard solutions measurements and are within 2e for176Hf/177Hf and 143Nd/144Nd, and 0.32% for 176Lu/177Hfand 0.54% for 147Sm/144Nd.

4. RESULTS

4.1. Major elements

The major element compositions of the garnets are givenin Table A.1 in the appendix and shown in Fig. 1c in aCaO–Cr2O3 diagram. Our larger-sized garnets (>2 mm)from all three localities give similar ranges in Cr2O3 (2–12 wt.%) with a mean at around 4 wt.% and high and lowCaO contents (from 0.3 wt.% to 4.3 wt.%). It seems thatthe large-sized garnets are restricted in composition com-pared to those found in heavy mineral concentrates usedin diamond exploration (selected grain size usually < 1 mm)and as inclusions in diamonds. These fall mostly into the


range between 2.5 and 15 wt.% Cr2O3 and cover the wholerange of CaO contents in the harzburgitic field (Stachel andHarris, 2008). Few reach 20 wt.% Cr2O3 with low to verylow CaO contents. The Mg#’s from Roberts Victor varyfrom 83 to 89 while garnets from Lace mine vary from 85to 91. The higher average Mg# from Lace (88) indicateshigher average degree of partial melting compared to Rob-erts Victor (86). The garnets are distinguished into groupswhich were derived mainly based on their trace elementcontents. The groups F1 from Finsch and RV1 from Rob-erts Victor, a number of RV2 and RV3 and L2 garnets plotbelow about 2 wt.%. The majority of the group F2, half ofRV2, L1 and L2 plot at higher CaO and subparallel to theharzburgite–lherzolite boundary.

4.2. Trace elements

Trace elements of subcalcic garnets from Roberts Victorand Lace are given in Table A.2 in the appendix and shownin chondrite (McDonough and Sun, 1995) and Vitim garnet(Ionov et al., 2005) normalized REE diagrams and in prim-itive mantle (Hofmann, 1988) normalized spider diagramsin Figs. 2 and 3. The sequence of elements for the spiderdiagrams is taken from the order derived by Hofmann(1988) for decreasing incompatibility except for Zr andHf. The Hofmann sequence reflects the nature of the resid-ual phases for low pressure melting with olivine, orthopy-roxene, ±clinopyroxene and spinel as residual phases. Ifone of these phases is exhausted or if another phase likegarnet appears in the residue, this order will change. Garnetharzburgites from the subcratonic lithosphere are recrystal-lised residua of high to very high degrees of partial melting.The phases present during the partial melting process wereeither ol + high-T opx + Cr-rich spinel (low pressure melt-ing) or ol + high-T opx + low-Cr grt (high pressure melting).For both kinds of residua Zr and Hf will behave more com-patible than indicated in the Hofmann (1988) sequence. Theyhave their place between the REE heavier than Sm and Eu.We chose to place them between Gd and Tb since the relativepartition coefficients of garnet/silicate melt and of orthopy-roxene/silicate melt for Zr and Hf are such that they fallsomewhere between Gd and Dy (e.g. Irving and Frey,1978; Green et al., 2000; van Kan Parker et al., 2010). In addi-tion to our own data, the REE and trace element pattern of agarnet from a primitive garnet peridotite xenolith (313-105)from Vitim near Lake Baikal (Ionov et al., 2005) is alsoshown in Figs. 2 and 3a and c and we have normalized ourgarnet REE data to the REE of the Vitim garnet. Sample313-105 has coexisting ol, opx, cpx, grt and probably smallamounts of spinel and a major and trace element composi-tion close to a primitive mantle. It is used here to visualizethe differences in composition of our subcalcic garnets tothe primitive mantle.

4.2.1. Roberts Victor

The chondrite normalized REE patterns of garnets fromRoberts Victor are typically sinusoidal as they are wellknown, e.g. from subcalcic inclusions in diamonds (Stachelet al., 2004). All garnets have strongly fractionated HREEand minima in the M-HREE segments of the patterns

which vary between Ho and Tb (Fig. 2a–c). However, thereare significant variations in the extent of LREE–MREEenrichment and HFSE contents. They allow the divisionof the garnets selected from Roberts Victor into two groups(RV1 light blue lines; RV2 dark blue lines). This division isbased on generally lower MREE in the RV-1 garnets andlower HFSE element contents (esp. Nb and Ta; Fig. 2c)and can also be seen in the Lu–Hf isotope relationships(see below). The RV2 samples (and two RV1) also are char-acterized by a pronounced peak at Nd with up to 100 timeschondritic contents. This Nd peak is not an analytical arti-fact because LA ICP MS and isotope dilution data are inagreement (see Fig. A.2 in the Appendix). The garnets havetherefore very low (Sm/Nd)C1 ratios, down to 0.05. Thenormalization of the REE to Vitim garnet (Fig. 2b) showsthat the HREE of the Roberts Victor garnets are stronglydepleted and the MREE and LREE enriched in two seg-ments as indicated in the two black lines.

The primitive mantle normalized spider diagrams showthat the garnets are enriched in most incompatible elements,like LREE, U, Nb and Sr compared to Vitim garnet (Fig. 2c).All samples show high U contents and negative Sr anomalies.Zirconium and Hf have similar abundance levels in RV1 withrespect to their neighboring elements Gd and Tb, but there isa negative slope from Dy, Tb, Hf, Zr to Gd in RV2 garnets.Overall, the more incompatible elements of group RV2 sam-ple are more enriched than group RV1.

4.2.2. Lace

The REE and trace element patterns of the garnets fromLace allowed us to divide them into two groups L1 (redlines) and L2 (pink lines; Fig. 3a–c). The four group L1 gar-nets with the highest (Lu/Hf)C1 ratios (red patterns inFig. 3) plot into quadrant I of Fig. 1d. Two of these (L12and L32) show only a weak sinusoidal pattern and highHREE contents (only three times below Vitim garnet).The other two have strong sinusoidal patterns and lowHREE and both have negative Zr and Hf anomalies inprimitive mantle normalized patterns (Fig. 3c). The six L2garnets (pink lines; three garnets from quadrant I and threegarnets from quadrant II in Fig. 1d) have low HREE (4–73times below Vitim garnet), are higher in LREE and havemuch more pronounced sinusoidal patterns with peaks atNdC1 or SmC1 and minima around ErC1 (Fig. 3a). The nor-malization to primitive mantle shows positive Zr and Hfanomalies (Fig. 3c). The normalization of REE to Vitimgarnet shows that all samples are depleted in HREE andthat the MREE and LREE are divided in two segmentsas indicated by the two black lines in Fig. 3b. All Lace sam-ples are higher than the Vitim garnet in the highly incom-patible elements from Sm onwards except for the mostincompatible elements Ba and Rb.

4.3. Isotope systematics

4.3.1. Samarium–neodymium isotope compositions from

Roberts Victor and Lace (Table 1a and b)

There is no significant correlation of the 143Nd/144Ndisotope with the 147Sm/144Nd ratios for both localities

Fig. 2. Subcalcic garnets from Roberts Victor: (a) C1 chondrite normalized REE patterns. Light blue patterns are group RV1 garnets, darkblue patterns are group RV2 garnets. The green pattern with yellow diamonds is that of a garnet from a primitive garnet peridotite 313-105from Vitim in Siberia (see text); (b) Vitim garnet normalized REE patterns; note the depleted HREE which are taken as the signature left frompartial melting and the enrichment of the middle and LREE in two steps as indicated by the black lines; (c) primitive mantle normalized spiderdiagrams of the subcalcic garnets. (For interpretation of the references to color in this figure legend, the reader is referred to the web version ofthis article.)


(Fig. 4; two 900 Ma reference lines are given in this figure).The Roberts Victor samples mostly have lower Sm/Nd ra-tios than the Lace samples but overlap exists. The six RV2garnets (with very pronounced Nd-peaks in their REE pat-terns – Fig. 2a) have high 143Nd/144Nd ratios (dark bluediamonds in Fig. 4) compared to most RV1 garnets. Their143Nd/144Nd ratios increase with Sm/Nd and overlap withand are followed by six Lace samples. The group RV2and six Lace garnets scatter around a 900 Ma reference line(blue dashed line in Fig. 4). One RV1 (RV24) sample whichplots with the RV2 garnets also has a very pronounced Ndpeak and thus the lowest Sm/Nd ratio. Its present-day eNdvalue is �35. One further RV1 garnet (RV93) with the leastoverprint on LREE (and therefore the highest Sm/Nd and143Nd/144Nd ratios amongst the RV1 garnets) also lies closeto the blue dashed reference line. The remaining four RV1garnets have nearly constant and very low 143Nd/144Nd ra-tios but variable 147Sm/144Nd (0.03–0.10). There are fourmore garnets from Lace which plot close to another900 Ma reference line at lower 143Nd/144Nd ratios.

4.3.2. Lutetium–hafnium isotope compositions from Roberts

Victor (Table 2a)

The subcalcic garnets from Roberts Victor are highlycorrelated in a 176Hf/177Hf versus 176Lu/177Hf diagram

(Fig. 5a). On close inspection two correlation lines can bedistinguished: (i) a line from the RV1 garnets (low MREEand low Hf) which yields an isochron of 2.947 ± 0.058 Ga(MSWD – Mean Square Weighted Deviation = 3.4) witheHf = 2.7 ± 1.4; (ii) a second line from the RV2 garnets giv-ing an age of 3.266 ± 0.15 Ga (MSWD = 58) witheHf = 17.6 ± 3.6. In Fig. 5b, we have plotted the modelages, calculated with respect to a primitive mantle, againstthe 176Hf/177Hf ratios. Four of the six RV1 data pointsfrom the 2.947 Ga isochron have model ages close to2.95 Ga in agreement with the isochron age; the two sam-ples with the lowest Lu/Hf ratios have around3.10 ± 0.02 Ga. The model ages of data points from the3.266 Ga isochron are older than 3.31 Ga and linearly in-crease with decreasing 176Hf/177Hf. They increase asymp-totically when the chondritic 176Lu/177Hf ratio isapproached. The relationships between the Hf isotope ra-tios and model ages are used in the discussion to distinguishwhether the isochrons date a partial melting event or both.

4.3.3. Lutetium–hafnium isotope compositions from Lace

(Table 2b)

The Lace garnets (Fig. 6a) give a Lu–Hf errorchron ageof 3.22 ± 0.5 Ga with a high initial of eHf = +16. In themodel age vs. 176Hf/177Hf diagram those garnets with high176Hf/177Hf (which are also those with 176Lu/177Hf ratios

Fig. 3. Subcalcic garnets from Lace: (a) C1 chondrite normalized REE patterns; (b) Vitim garnet normalized REE patterns; note the depletedHREE which are taken as the signature left from partial melting and the enrichment of the middle and LREE in two steps as indicated by theblack lines; (c) primitive mantle normalized spider diagrams. Red lines are L1 and purple are L2 garnets; primitive Vitim garnet is in green.(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)


higher than the primitive mantle) show a broad correlationof increasing model ages with decreasing 176Hf/177Hf. FourL2 garnets (purple triangles) with strong positive Zr–Hfanomalies and with low 176Hf/177Hf isotope ratios plot tothe left and below this correlation (Fig. 6b).

4.3.4. A reexamination of lutetium–hafnium isotope data

from Finsch (Lazarov et al., 2009a)

We have re-examined the Lu–Hf-isotope data of thesubcalcic garnets from Finsch (group F1 and F2 distin-guished by Lazarov et al., 2009a). Six group F1 garnets de-fine an isochron of 2.62 Ga which is about 100 Ma olderthan reported due to the use of the newer Lu decay constant(Amelin and Davis, 2005) (Fig. 7a). It can be seen in Fig. 7bthat all F1 data points have model ages older than 2.62 Ga.The model ages increase with decreasing 176Hf/177Hf firstlinearly and then asymptotically when the Lu/Hf ratioscome close to the chondritic value. Our attention was fur-ther drawn in Fig. 7b to the fact that six F2 garnets givea constant model age close to 1.9 Ga. When these samplesare plotted into an isochron diagram they give an age of1.90 ± 0.19 Ga and, out of necessity (agreement of isochronand model ages), an initial eHf of zero (�0.1 ± 3.2; Fig. 7c).The 176Lu/177Hf ratios of these samples spread from sub-chondritic (four garnets) to superchondritic (two garnets)which indicates that the Lu–Hf system dates a 1.9 Gaenrichment event rather than a partial melting event.

5. DISCUSSION

5.1. Evaluation of isochron diagrams – general remarks and

principles

A reliable isochron is a regression line with a statisticallymeaningful number of data points and a resulting MSWDvalue of around one. The initial characterizes the sourcecommon to all members along the isochron. Due to the spe-cial sampling process of mantle xenoliths through verticalmelt channels, working on mantle samples generally meansthat we deal with material from a huge depth range respec-tively a huge rock volume. Portions of this may have differ-ent origins and histories. It may become problematic toidentify those members amongst the xenoliths or xenocrystswith a common genetic origin. However, the antique natureof Archean peridotites brings two distinct advantages: (a)the antiquity ascertains sufficient time for the isotopic sys-tems to develop into significant correlations for any kindof event and (b) the peridotitic mantle samples are a prioriresidues of medium to very high degrees of partial meltingwhich will provide a good spread in parent/daughter ratiosof a radiogenic system in the residue. Subsequent metaso-matism and possibly second stage melting appears to beinevitable in Archean mantle samples. If an afflicted rockvolume is not completely homogenized by the second eventthe initials will vary outside the analytical uncertainties for

Table 1Sm–Nd isotope compositions of subcalcic garnets from the Roberts Victor and Lace mine.

Sample ID-Sm (ppm) ID-Nd (ppm) 147Sm/144Nd 143Nd/144Nd ±2 S.E. eNd (T = 120 Ma) T CHUR (Ga)

(a) Sm–Nd isotope compositions of subcalcic garnets from the Roberts Victor mineGroup RV1

RV24 0.768 48.480 0.0097 0.510841 0.000006 �32 1.46 ± 0.03RV26 0.405 3.952 0.0635 0.510466 0.000012 �40 2.47 ± 0.05RV31 0.071 1.266 0.0347 0.510515 0.000006 �39 1.99 ± 0.04RV93 0.043 0.148 0.1461 0.512038 0.000065 �11 1.79 ± 0.04RV111 0.181 1.546 0.0724 0.510533 0.000011 �39 2.57 ± 0.05RV123 0.455 2.777 0.1015 0.510501 0.000007 �40 3.39 ± 0.07Group RV2

RV12 4.492 18.259 0.1671 0.511952 0.000008 �13 3.49 ± 0.07RV23 1.080 5.843 0.1136 0.511261 0.000007 �26 2.51 ± 0.05RV54 3.684 12.708 0.1798 0.511837 0.000008 �15 7.06 ± 0.15RV94 0.599 11.757 0.0313 0.511237 0.000007 �25 1.29 ± 0.03RV100 3.851 30.175 0.0764 0.511328 0.000004 �24 1.65 ± 0.03RV124 0.932 34.091 0.0171 0.510945 0.000006 �30 1.43 ± 0.03

(b) Sm–Nd isotope compositions of subcalcic garnets from the Lace mineGroup L1

L12 0.562 1.375 0.2471 0.512401 0.000008 �5 �0.71L32 0.679 1.534 0.2675 0.51238 0.000006 �6 �0.55L14 0.442 0.741 0.2645 0.511561 0.000010 �22 �2.44L31 0.330 1.118 0.1683 0.511047 0.000007 �31 3.99 ± 0.08L26 0.537 2.424 0.1339 0.511835 0.000008 �15 1.94 ± 0.04L28 0.878 2.147 0.1830 0.511881 0.000008 �15 8.21 ± 0.17Group L2

L36 0.655 3.682 0.1075 0.510753 0.000008 �35 8.32 ± 0.17L2 0.881 4.478 0.1209 0.510631 0.000013 �38 3.19 ± 0.07L7 0.646 1.378 0.2872 0.512411 0.000007 �6 7.97 ± 0.17L30 0.781 2.228 0.1843 0.511973 0.000008 �13 �0.38L39 1.356 2.531 0.3241 0.512368 0.000034 �7 �3.22

Abbreviation: ID – isotope dilution.


a resulting correlation in an isochron diagram. The MSWDvalue can be much higher than one but the correlation willnevertheless has an age significance for old events. Dealingwith Archean samples has the advantage that high MSWDvalue can be tolerated.

A linear correlation of isotope ratios in an isochron dia-gram for mantle peridotites may date an enrichment (mixing)process or a partial melting event or a coupling of both. Theinterpretation of an isochron as monitoring an enrichment orpartial melting event can be decided upon from the relativeabundances of the incompatible elements and the relation-ship between the isotope composition of the sample andthe primitive mantle. We discuss here the Lu–Hf isotope sys-tem and introduce the relationship between the isochron ageand the model age (which is calculated with respect to theprimitive mantle) of each data point as a tool to decidewhether an isochron reflects partial melting of a primitivesource, remelting of a depleted source or re-enrichment (withHf). The model age of each data point along an isochron willbe identical to the isochron age for partial melting of a prim-itive mantle and Lu/Hf ratios will always be superchondritic.Model ages will be older than the isochron age for secondstage partial melting and Lu/Hf ratios will be very muchhigher than the chondritic value.

A linear array may be simply due to mixing with thehost kimberlite (or a precursor) at or near the time of kim-berlite eruption. Enrichment (mixing) can be discerned

from a linear relationship between the Hf isotope composi-tion and the reciprocal Hf content (1/Hf) of the membersalong this array. The Hf isotopic composition and the Hfcontent of the host kimberlite (or a similar silicocarbonatit-ic melt in the mantle) must be the lower end of the lineararray. Other isotope ratios like 143Nd/144Nd must also bein agreement with the composition of the enriching agent.

Otherwise, a linear array must be interpreted as datingthe ancient enrichment of a peridotitic restite with Hf andthe calculated model ages are older than the isochron ageand increase with decreasing 176Hf/177Hf isotope ratios.This increase is at first linear and then asymptotic, whenthe chondritic Lu/Hf ratio at the time of enrichment is ap-proached. If the Lu/Hf ratios become subchondritic veryhigh negative model ages are obtained which decrease withdecreasing Lu/Hf. These relationships are shown schemati-cally in Fig. 8a and b for the case that isotopic homogeni-zation of the daughter element occurred at the time ofenrichment (=no correlation of 176Hf/177Hf with 1/Hf).The development of the Hf isotopes in a primitive mantlewith time is shown in Fig. 8a with a gray line. The blackdots mark the time of partial melting and the developmentof the Hf isotopes in the residues up to the time of enrich-ment (green circle). The blue cross marks the compositionof the Hf-metasomatized and isotopically homogenized res-tites. Increasing Hf/Lu give increasingly older model agesuntil they become infinite when the chondritic ratio is

Fig. 4. The combined data from Lace and Roberts Victor for theSm–Nd system with two 900 Ma reference lines: there is nocorrelation of the Nd isotope with the 147Sm/143Nd ratios for bothlocalities. However, the six RV2 garnets with very pronounced Nd-peaks in their REE patterns correlate at high 143Nd/144Nd ratioswith 147Sm/144Nd parallel with a 900 Ma reference isochron with aeNd of �11, overlapped and followed by six Lace samples (with amaximum at Nd). Two of the RV1 samples also plot along thisreference line. The remaining four RV1 garnets have constant, butvery low 143Nd/144Nd ratios with variable 147Sm/144Nd. There arefive garnets from Lace which plot close to a 900 Ma reference lineat lower 143Nd/144Nd ratios (eNd = �27). The symbols are largerthan the errors.


reached. Further increase gives negative values. The iso-chron dates the real age of enrichment. A special case orig-inates when mixing and isotopic homogenization generatesthe 176Hf/177Hf ratio of the primitive mantle at that time.This results in an isochron which seemingly dates partialmelting of a primitive mantle (see the example for the 1.9isochron from Finsch; Fig. 7c). Both subchondritic andsuperchondritic Lu/Hf ratios and trace element composi-tions, however, document an enrichment event.

Mixing without isotopic homogenization results in a var-iability of the 176Hf/177Hf ratios along the vertical axis inFig. 8a and a linear relationship of the 176Hf/177Hf ratioand 1/Hf at the time of mixing. This develops into an iso-chron which yields a maximum possible age for the enrich-ment process. The real age of enrichment may be estimatedby testing for the statistic significance of the correlation ofthe isotope ratios with R2, the Student-t-test, beginning withthe recalculated isotope ratios at the seeming time of enrich-ment and calculate for younger and younger ages until a highlevel of confidence is reached (R2 should be >0.9 for the smallnumber of data). See actual example for the 3.27 Ga “iso-chron” from Roberts Victor below (Fig. 11a and b).

5.2. Partial melting regimes and enrichment processes from

trace element systematics

Subcalcic garnets are the compositional proxies formany elements of their host rocks and protoliths and someof these compatible elements in garnets like Cr and theHREE are useful to infer melting regimes of the protoliths

(e.g. Boyd et al. (1993); Stachel et al., 1998; Klein-Ben Da-vid and Pearson, 2009). The prerequisite is that they are notor were only negligibly modified by metasomatism subse-quent to partial melting. This is valid for chromium eventhough Klein-Ben David et al. (2011) found that Cr maybe mobilized and transported under mantle conditions bysaline solutions (however without significantly affectingthe comparatively high Cr-contents of the bulk mantlerocks). Also, the HREE of mantle peridotites may not soeasily be affected by metasomatism. Any plausible enrichingagent in the mantle has highly fractionated REE patternswith high LREE and low HREE abundances. Positivelysloped HREE patterns of sinusoidal garnets at low abun-dances will thus reflect the original signature of the partialmelting residue since such slopes can only be generated in agarnet-absent melting regime. Because all REE are incom-patible in mantle phases except for garnet low pressuremelting causes REE depletion in the residue from the onsetof low-pressure melting, i.e. all REE will be much morestrongly depleted in such a residue compared to a residuefrom the garnet stability field.

The Cr/Al partition coefficient ratios of (DCr/DAl)grt/melt

and (DCr/DAl)opx/melt are about 1.0 and 2.0, respectively, at

pressures of 5 GPa and more (Bulatov et al., 1991; Caniland Wei, 1992; Stachel et al., 1998). Residual garnets andorthopyroxenes will therefore have low Cr/Al ratios, i.e.Cr-rich garnets (as they are common in the subcratonic man-tle) cannot be generated as residual phases at high pressuresand also not by exsolution from residual high pressure opx.At low pressures in the spinel peridotite stability field, theCr/Al partition coefficient ratios of (DCr/DAl)

spinel/melt and(DCr/DAl)

opx/melt are about 40 and 10 respectively (see refer-ences above) and residual harzburgites would have very highCr/Al ratios. Very Cr-rich garnets can grow from such com-positions upon subduction into the garnet stability field.

The boundary conditions placed by the partition coeffi-cients and resulting element ratios and abundances havebeen used previously to deduce that a major proportionof the garnet peridotites from the Kaapvaal craton are sub-ducted residues of partial melting at low pressures (e.g.Ringwood, 1977; Kesson and Ringwood, 1989; Stachelet al., 1998; Pearson and Nowell, 2002; Wittig et al.,2008; Lazarov et al., 2009a, 2012b). Our new data are in fullaccord with the conclusions of all these authors.

Two models (with modifications) exist for the origin ofresidual subcratonic mantle:

(a) The plume model, where partial melting occurred athigh pressures in the garnet stability field in an upris-ing plume (e.g. Boyd, 1989). Olivine, orthopyroxeneand garnet are residual phases. Since the HREE arecompatible in garnet they diminish only slightly withincreasing degree of partial melting and are retainedin the residue as long as garnet is present. The HREEcontents will only decrease when garnet is exhaustedat very high degrees of partial melting (i.e. at veryhigh temperatures). Garnets exsolving from the resid-ual high temperature orthopyroxenes upon coolingopx via the reaction

Table 2Lu–Hf isotope compositions of subcalcic garnets from the Roberts Victor and Lace mine.

Sample ID-Lu (ppm) ID-Hf (ppm) 176Lu/177Hf 176Hf/177Hf ±2 S.E. eHf (T = 120 Ma) TCHUR (Ga)

(a) Lu–Hf isotope compositions of subcalcic garnets from the Roberts Victor mineGroup RV1

RV24 0.092 0.075 0.1719 0.290817 0.000006 273 3.02 ± 0.07RV26 0.037 0.073 0.0703 0.285001 0.000012 75 3.13 ± 0.07RV31 0.135 0.036 0.5266 0.310794 0.000015 952 2.96 ± 0.06RV93 0.062 0.065 0.1318 0.288261 0.000018 186 2.90 ± 0.06RV111 0.027 0.071 0.0534 0.283971 0.000006 40 3.09 ± 0.07RV123 0.070 0.038 0.2553 0.295319 0.000013 426 2.94 ± 0.06

Group RV2

RV12 0.075 0.196 0.0451 0.284192 0.000005 49 6.05 ± 0.13RV23 0.141 0.081 0.2435 0.296838 0.000008 480 3.47 ± 0.08RV54 0.121 0.256 0.0585 0.284582 0.000008 62 3.70 ± 0.08RV94 0.160 0.070 0.3327 0.301857 0.000009 651 3.31 ± 0.07RV100 0.091 0.127 0.099 0.287279 0.000008 154 3.55 ± 0.08RV124 0.094 0.155 0.084 0.286545 0.000005 129 3.84 ± 0.08

(b) Lu–Hf isotope compositions of subcalcic garnets from the Lace mineGroup L1

L12 0.201 0.157 0.1825 0.293031 0.000005 351 3.56 ± 0.07L14 0.039 0.057 0.0982 0.285480 0.000009 91 2.19 ± 0.04L26 0.070 0.124 0.0801 0.287400 0.000006 160 5.04 ± 0.10L28 0.028 0.103 0.0391 0.283912 0.000007 40 9.48 ± 0.19L31 0.039 0.038 0.1451 0.289892 0.000005 243 3.30 ± 0.07L32 0.198 0.255 0.1103 0.288539 0.000004 198 3.86 ± 0.08L36 0.137 0.213 0.0912 0.285929 0.000004 107 2.84 ± 0.06

Group L2

L2 0.050 1.006 0.0070 0.280946 0.000004 �63 3.61 ± 0.07L7 0.032 0.303 0.0151 0.282708 0.000002 �1 0.19 ± 0.01L30 0.008 0.096 0.0113 0.281572 0.000005 �41 2.86 ± 0.06L39 0.049 0.636 0.0109 0.282197 0.000005 �19 1.39 ± 0.03

Abbreviations: ID – isotope dilution.

10Q

.S

hu

etal./

Geo

chim

icaet

Co

smo

chim

icaA

cta113

(2013)1–20

Fig. 5. (a) Isochrons from subcalcic garnets from Roberts Victor:group RV1 (light blue line) give 2.947 Ga; group RV2 (dark blueline) give 3.266 Ga. (b) The Hf-model ages of the RV1 garnets arearound 2.95 Ga and the same as the isochron age (light bluediamonds connected by black horizontal line). All six RV2 garnetshave higher model ages than their isochron age (=3.266 Ga); themodel ages increase with decreasing 176Hf/177Hf (Lu/Hf). Thesymbols are larger than the errors. (For interpretation of thereferences to color in this figure legend, the reader is referred to theweb version of this article.)

Fig. 6. (a) Isochron diagram for Lace subcalcic garnets: the lacesamples form an errorchron of 3.222 ± 0.5 Ga isochron witheHf = +16. Red triangles are L1 and purple triangles are L2garnets. (b) In a model age versus 176Hf/177Hf diagram, the garnetswith high 176Hf/177Hf (which are also those with 176Lu/177Hf ratioshigher than the primitive mantle) show a broad correlation ofincreasing model ages with decreasing 176Hf/177Hf. Four L2 garnets(purple triangles) with strong positive Zr–Hf anomalies and withlow 176Hf/177Hf isotope ratios plot to the left and below thiscorrelation. The symbols are larger than the errors. (For interpre-tation of the references to color in this figure legend, the reader isreferred to the web version of this article.)


2MgSiOopx3 þMgCrAlSiOopx

6 ¼Mg3ðAlCrÞ2Si3Ogrt12

will still be Cr-poor because of (DCr/Al/DCr/Al)grt/opx of

around two resulting from the low (DCr/DAl)grt/melt and

(DCr/DAl)opx/melt at high pressures (see above). The plume

model therefore does not appear to be viable to explainthe major proportion of the subcratonic mantle.

(b) An upwelling mantle model with partial melting atshallow depths in the spinel stability field beyondthe exhaustion of cpx in a setting similar to midocean ridges (Kesson and Ringwood, 1989). Thiswas followed by subduction to generate Cr-rich gar-nets by the reactions

4MgSiOopx3 þMgðAl;CrÞ2Osp

4 ¼Mg3ðAl;CrÞ2Si3Ogrt12

þMg2SiOol4

and

2MgSiOopx3 þMgCrAlSiOopx

6 ¼Mg3ðAlCrÞ2Si3Ogrt12

An upwelling mantle model appears to be more plausible toexplain the major proportion of the subcratonic mantle asmelting residue also in the light of HREE abundancesand ratios.

Partial melting increases the Lu/Hf and Lu/Er ratios inthe residues to various extents depending on the melting re-gime, the degree of melting and the melting mode. Partialmelting model calculations were carried out using partitioncoefficients Di (for element i) from Green et al. (2000) forthe garnet stability field and for the spinel stability fieldfrom Kelemen et al. (1993) and Suhr et al. (1998). Meltingmodes were taken from Kinzler (1997) for spinel peridotiteand Walter (1998) for garnet peridotite melting. The resultsfor non-modal fractional melting are shown in Fig. 1d.Batch melting (not shown in Fig. 1d) would yield the small-est range of these ratios (least fractionation of Lu from Hf

Fig. 7. The Lu–Hf isotopic systematics for subcalcic garnets from Finsch (data from Lazarov et al., 2009a): (a) six F1 garnets from quadrant Iin Fig. 1d define an isochron of 2.621 Ga; (b) the model ages for these garnets (black squares) are all older than the isochron age and increasewith decreasing 176Hf/177Hf. Six F2 garnets (red squares) from quadrant II in Fig. 1d give constant model age of 1.9 Ga; (c) these define anisochron of 1.907 Ga. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)


and Er) compared to a somewhat higher range for non-modal batch melting. Non-modal fractional melting moststrongly fractionates the elements in the comparison ofthe various melting models. For the same degree of partialmelting, the Lu/Hf ratios are much higher at high pressuresin the presence of ol, opx and grt and, on further melting, ofol and opx in comparison to the ratios obtained in the spi-nel stability field. The comparison of the model calculationwith the element ratios of our subcalcic garnets (=those ofthe bulk rock) shows that single stage non-modal fractionalmelting of primitive mantle component in the garnet stabil-ity field is not a viable scenario for the majority of our sam-ples, but that a number of samples from Lace, RobertsVictor and Finsch follow the trend for melting in the spinelstability field. A number of Roberts Victor garnets alsoform a parallel alignment to the calculated spinel meltingtrend at lower Lu/Hf (Fig. 1d). This trend would be a pos-sible melting trend, if the bulk Hf partition coefficients weresimilar to Ho or Er. Lower (Lu/Hf)C1 ratios in quadrant Ioff the modeling melting trend are caused by small amountsof metasomatic overprint. Samples which plot in the otherquadrants were more severely overprinted and possibly sev-eral times.

Low pressure melting quickly diminishes the HREEcontents in the residue from the onset accompanied by asteady increase in the Cr/Al ratio (see above). Melting in

the presence of garnet would keep the HREE at high levelsand also not cause the fraction between Cr and Al, i.e. notseverely change the Cr/Al ratios. We have plotted in Fig. 9aand b the Lu respectively the Hf contents in the garnetsagainst their Cr# [100 � Cr/(Cr + Al)] and, in Fig. 9a, thecomposition of the Vitim garnet as representative of theprimitive mantle. The F1 (shaded blue) and most of theF2 garnets from Finsch follow a hyperbolic trend (gray linein Fig. 9a) which extends from the Vitim garnet and indi-cates a melting trend for low pressure partial melting. TheRV1 garnets (shaded green) fall onto this trend at higherCr# than the F1 garnets which indicates higher degrees ofpartial melting. A few of the RV2 garnets and half the Lacegarnets also follow this trend. Other RV2 form a furtherhyperbolic trend (stippled line) which is shifted to higherCr#. It may be explained as a trend of second stage meltingof an already depleted mantle. Isotope ratios were notdetermined in these samples. The RV2 garnets in which iso-topes were measured are enclosed by a light blue box with adark rim. The Lu–Cr# relationship suggests that all theinvestigated samples are restites of garnet-absent partialmelting at low pressures. The hafnium contents should alsobe related to Cr# in a hyperbolic trend if they are direct res-tites from depletion. However, the majority of our samplesis shifted to higher Hf values except the RV1 samples withlowest contents. We consider these very low Hf abundances

Fig. 8. (a and b) Reenrichment or partial melting? – the relationships between the Lu–Hf isotopic compositions of the samples (mantleresidues) and the primitive mantle. The relationships are shown schematically in Fig. 8a and b. The development of the Hf isotopes in aprimitive mantle with time is shown in Fig. 8a with a gray line. The black dots mark the time of partial melting and the development of the Hfisotopes in the residues up to the time of enrichment (green circle). The blue cross marks the composition of the Hf-metasomatized andisotopically homogenized restites. Partial melting of a primitive mantle can be recognized if the model age of each data point (calculated withrespect to the primitive mantle) along an isochron is identical to the isochron age. Model ages will be older than the isochron age for secondstage partial melting of the restites and Lu/Hf ratios will be much higher than the chondritic value. Ancient reenrichment with Hf lowers theLu/Hf ratios to various extent and isotopic homogenization of the daughter element may occur at that time (=no correlation of 176Hf/177Hfwith 1/Hf). Increasing Hf/Lu (black bowed arrow) give increasingly older model ages [(indicated by the intersection of the steepest red line ina) with the primitive mantle and the upper red line in b)] until they become infinite when the chondritic ratio is reached (red line marked with1). Further increase gives negative model ages (lower left red line in b). The isochron dates the real age of enrichment. A special caseoriginates when mixing and isotopic homogenization generates the 176Hf/177Hf ratio of the primitive mantle at that time. This results in anisochron which seemingly dates partial melting of a primitive mantle (see the example for the 1.9 isochron from Finsch; Fig. 7c). Bothsubchondritic and superchondritic Lu/Hf ratios and trace element compositions, however, document an enrichment event. Mixing withoutisotopic homogenization results in a variability of 176Hf/177Hf ratios along the vertical axis in (a) and a linear relationship of 176Hf/177Hf ratioand 1/Hf at the time of mixing. This develops into an isochron which yields a maximum possible age for the enrichment process. (Forinterpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)


to be the residual values from partial melting of a primitivemantle portion around 2.95 Ga which is in agreement withour finding that the isochron age of the RV1 garnets is iden-tical to that of the model ages of the members along the iso-chron (Fig. 5a and b). Group F1 and RV2 garnets havehigher values. Accordingly, these garnets yielded a metaso-matic age (Fig. 7a and b).

5.2.1. Roberts Victor mine

The low-Ca garnets from Roberts Victor (0.55–3.7 wt.%CaO) in quadrant I (Fig. 1d) show a large range of Cr#from 8 to 28 [Cr# = 100 � Cr/(Cr + Al)], low to extremelylow HREE contents and a high fractionation of the HREEwith (Lu/Er)C1 up to 11 (Fig. 2a and b).These chemicalparameters indicate an origin of majority of the RobertsVictor garnet harzburgites as restites of partial melting atlow pressures. When normalized to the Vitim garnet theREE patterns have a positive slope from Ho or Er to Lu,a shallow increase from Ho to Ce for four RV1 garnetsand a negative slope in two increments from Ho to Nd orCe for the RV2 and two RV1 garnets (Fig. 2b). An increasein two steps indicates at least a twofold overprint bymetasomatizing agents. The agents which equilibrated withgroup RV2 garnets were rich in LREE, MREE, Sr, Zr–Hfand U, depleted in Nb and Ta and had extremely highLREE/HREE ratios. Based on the low HFSE/LREE ra-tios, high U contents, high Zr/Hf and Nb/Ta ratios we con-clude that the RV samples were metasomatized by (silico-)carbonatitic melts (Yaxley et al., 1991; Ionov et al., 1993;

Rudnick et al., 1993; Brey et al., 2008). The full reasoningand modeling calculations with the whole range of trace ele-ments (using the partition coefficient set provided by Girniset al., 2013) is provided in an accompanying paper by Shuet al. (2003).

5.2.2. Lace mine

Most subcalcic garnets from Lace show a similar rangeof Cr# (Fig. 1b) and HREE patterns and abundances asthose from Roberts Victor which indicates a similar lowpressure origin. Group L1 samples (red lines in Fig. 3) showhigher HREE contents than group L2 samples (pink lines)but both groups are re-enriched in the more incompatibleelements. Group L2 garnets have strongly positive Zr–Hfand group L1 garnets negative Zr–Hf anomalies comparedto the neighboring Gd and Tb. Modeling calculations forthe whole set of trace element data show like for RobertsVictor that the metasomatizing agents were (silico-)carbon-atitic melts (Shu et al., 2003).

5.3. Sm–Nd and Lu–Hf isotope systems – dating enrichment

and partial melting

The present day eHf and eNd values of the garnets fromall three localities show a tremendous range from +952 to�42 for eHf and +19 to �41 for eNd. The developmentof the Hf isotope ratios into very positive eHf values dem-onstrate the antiquity of the partial melting events and thehigh degrees of partial melting involved in the origin of

Fig. 9. (a and b) Diagrams of Cr# vs. Lu content and Cr# vs. Hf content: (a) Cr# vs. Lu (ppm): subcalcic garnets of groups F1 and F2 fromFinsch and group RV1 from Roberts Victor define a correlation of decreasing Lu with increasing Cr# which can be the trend of residues fromlow-pressure partial melting from a primitive mantle (gray line with arrow). (b) Cr# vs. Hf (ppm): a few group F1 garnets from Finsch and thegroup RV1 samples from Roberts Victor have the lowest Hf contents with variable Cr#. They should fall on a trend of residues from low-pressure partial melting from a primitive mantle (gray dashed line with arrow). Group RV1 samples fall on the depletion line in both diagramand have the lowest Lu and Hf contents (the legends of various groups are noted next to the symbols). The blue diamonds with pink rimsrepresent the group RV2 garnets which yielded a 3.26 Ga isochron. (For interpretation of the references to color in this figure legend, thereader is referred to the web version of this article.)

Fig. 10. (a and b) The correlation of the 176Hf/177Hf compositions of the RV2 garnets (3.27 Ga isochron age) at the time of kimberliteeruption (120 Ma) with 1/Hf and the 143Nd144Nd values of the same samples with 1/Nd. The negative linear correlation between 176Hf/177Hfand 1/Hf indicates a mixing process with two components. If we assume the mixing agent to be the kimberlite magma and take the 176Hf/177Hfcompositions of group I and group II kimberlites reported by Nowell et al. (2004), the kimberlite magma should have only around 0.25 ppmHf which is very much lower than the contents in kimberlites of >0.89 ppm. The unradiogenic 143Nd144Nd isotope ratios also precludecontamination by the kimberlite magma as these have more radiogenic 143Nd144Nd ratios (Nowell et al., 2004).


these restites and highly negative eNd to the antiquity of themetasomatizing agents or an ancient enrichment event. Amore detailed discussion of the variation of the eHf andeNd values is given in Shu et al. (2003).

5.3.1. Errorchrons from the Sm–Nd system

Incompatible trace elements indicate a twofold metaso-matic overprint of many samples from Roberts Victor, Laceand Finsch. The imprint on the garnets varies with theamount of the metasomatizing agent and the modal abun-dance of garnet in the rock (garnets of lower modal abun-dances are more affected by the same amount ofmetasomatizing agent). Sm and Nd will always be affected

by each metasomatic event. Because of this multiplicity,no significant correlation could be found in this studywhich would correspond to an isochron. However, theslope of six RV2 garnets, two RV1 samples and six of theLace garnets is roughly parallel to a 900 Ma reference linewith low initial with eNd of �11 (black dashed line inFig. 4). The other samples from both localities also plotparallel to a 900 Ma reference line but with much lower ini-tial with eNd of �27 (gray dashed line in Fig. 4). Such anage corresponds to the last stages of the Namaqua–Natalorogeny (0.9–1.3 Ga; see Fig. 1) which subducted materialunderneath the craton (Thomas et al., 1993; Schmitz andBowring, 2003; Hopp et al., 2008). It may have caused

Fig. 11. (a and b) The development of 176Lu/177Hf with time indiagrams of 176Hf/177Hf vs. 176Lu/177Hf and 1/Hf in RV2 garnets:the correlation of the 176Lu/177Hf (1/Hf) and calculated176Hf/177Hf at different time from back-time calculation: the bluediamonds are the data points measured today which define an ageof 3.27 Ga in an isochron diagram (a). They also define a goodcorrelation with 1/Hf (b) deciphering the correlation in (a) a mixingline. As we preclude the possibility of recent mixing withkimberlite, we use the iterative back-time calculation, starting at3.27 Ga until a good linear correlation of 1/Hf vs. 176Hf/177Hfobtained. At 3.27 Ga, the R2 of the linear correlation is 0.0033. Itquickly rises to 0.9619 for 2.9 Ga and 0.9729 at 2.8 Ga and slowlyincreases to 0.98 until today. Calculations for the isochronrelationship show the same fast increase of R2 over the first 400million years. It is concluded that the mixing process occurred at atime of 2.8–2.9 Ga. (For interpretation of the references to color inthis figure legend, the reader is referred to the web version of thisarticle.)


the last widespread metasomatism in this area of the subcr-atonic mantle before the Mesozoic kimberlite activity. Theextremely low Nd isotope ratios (down to eNd = �41) ofthe lowest Sm/Nd samples show that the enriching agentsstem from a reactivated reservoir with long term low Sm/Nd ratios such as old crust.

5.3.2. Isochrons from the Lu–Hf system

5.3.2.1. Roberts Victor. Two isochrons could be distin-guished amongst the subcalcic garnets which plot intoquadrant I of Fig. 1d:

(a) a 2.95 ± 0.06 Ga isochron with eHf = 2.7 ± 1.4 and(b) a 3.27 ± 0.15 Ga isochron with eHf = 17.6 ± 3.6.

The age of 2.95 Ga for the RV1 garnets is interpreted asdating the partial melting of a close to primitive mantle por-tion at shallow pressures in an upwelling mantle setting.Reasons are (i) that the isochron and the model ages ofthe individual data points agree (see Fig. 5b), (ii) thatCr2O3 contents in garnets can become very high (Fig. 2)and that the abundance levels of the HREE is low to verylow. Melting occurred in some kind of mid ocean ridge set-ting and probably subduction underneath the W-block. Thehost rocks of the RV1 garnets (cpx-free garnet harzburg-ites) represent the subducted, depleted lithosphere whichwere residues from depletion around 2.95 Ga and the corre-sponding melts (possibly the mafic oceanic crust) may bethe ubiquitous eclogites from Roberts Victor (e.g. Jacobet al., 2005). The collision of the W- and E-block occurredat around 2.9 Ga along the Colesberg lineament and west-ward subduction is supported by the occurrence of calcalca-line granitoids on the W-block of similar age alignedparallel to the Colesberg lineament (Schmitz et al., 2004).Also, the Re–Os ages of eclogitic sulfide inclusions in dia-monds is about 2.9 Ga in several localities on the W-block(Shirey et al., 2004). They may record this subduction cou-pled with diamond formation by redox reactions.

Six RV2 garnets define a 3.27 Ga isochron with a highinitial eHf of +17.6 (Fig. 5a). From the trace element pat-terns and abundances (Fig. 2), the “model age vs.176Hf/177Hf” relationship (Fig. 5b), the positive alignmentof the data points in a 176Hf/177Hf versus 1/Hf diagram(Fig. 10a), the high positive eHf and the criteria laid downin the “general remarks and principles section” above weinterpret this age as dating a maximum possible age ofenrichment of an already strongly depleted mantle. The iso-chron in Fig. 5a cannot be a recent mixing line with thekimberlite host or a precursor kimberlite because the176Hf/177Hf ratios of group I and II kimberlites intersectthe mixing line in Fig. 9a (calculated at 120 Ma) at Hf con-centrations of about 0.25 ppm Hf. This is far too low com-pared to the usual >0.89 ppm contents of kimberliticcompositions (Nowell et al., 2004). The 143Nd/144Nd alsorule out recent mixing with kimberlite because they are al-ways lower than those in group I and group II kimberlitesat the time of eruption (Fig. 10b). The enrichment mustthen have been ancient without homogenization of the Hfisotope ratios throughout the afflicted rock volume. Theisochron age is therefore a maximum possible age and thereal age must be estimated by projecting the Lu–Hf isotoperatios back in time (Fig. 11a and b). These figures show thatit takes about 400 Ma (from 3.27 Ga) until a statisticallysignificant correlation exists between the parent/daughterisotope ratios (Fig. 10a) and for the mixing line(Fig. 11b). In other words, the enrichment must have oc-curred sometime between 2.8 and 2.9 Ga in a mantle wedgeby (silico-)carbonatitic melts driven off a subducting slab(corresponding to the oceanic crust generated by the2.95 Ga partial melting).

The protolith of the mantle wedge peridotites were olderspinel harzburgites which (i) originated by high degrees ofpartial melting at shallow depths. The Roberts Victor kim-berlites sampled the metamorphosed residues of the2.95 Ga partial melting event, possibly its subducted and

Fig. 12. (a–e) A model proposed for the formation and evolution of the lithospheric mantle beneath the Kaapvaal craton. Mantle melting andcrust formation is taken as being caused mainly by passive upwelling and subduction related processes. (a) The period between 3.2 and 3.4 Ga:the major part of the continental crust of the E-block already existed. The crust rested on a highly depleted mantle which was metasomatized3.3 Ga ago (Lu–Hf age from subcalcic garnets from the Lace mines). The existence and size of the W-block is enigmatic until 3.2 Ga, fromwhen the oldest mantle TRD ages and crustal zircon ages are reported. (b) During the time span between 3.1 and 2.95 Ga igneous activityincreased on the W-block with TTG’s and the emplacement of greenstone belts. Oceanic crust and depleted mantle was created in between thetwo blocks in a passively upwelling mantle setting at 2.95 Ga (partial melting age from Roberts Victor subcalcic garnets) and subductedunderneath the W-block. (c) At around 2.88 Ga, the two blocks collided along the Colesberg lineament. (d) Between 2.6 and 2.8 Ga, thevoluminous Ventersdoorp volcanics poured out over a major part of the then unified Kaapvaal craton. The 2.62 Ga enrichment at Finsch fallswithin this period and may record pervasive metasomatism in connection with the Ventersdoorp magmatism. (e) During 1.8–2.1 Ga, thecraton was subjected chiefly to modification from the margins by subduction which caused metasomatism in the SLCM. The 1.9 Gaenrichment age from Finsch garnets lies within the period of the attachment of the Kheis–Magondi belt to the Kaapvaal craton. (f) The Sm–Nd isotope system possibly records craton-scale metasomatism, from 0.9 to 1.3 Ga as a result of the Namaqua–Natal orogeny in the south ofthe Kaapvaal craton.



eclogitisized oceanic, an older mantle wedge and membersof the overlying continental crust.

5.3.2.2. Lace. The 10 samples from the Lace mine yield anerrorchron of the 3.27 ± 0.6 Ga which seemingly overlapswith an enrichment age from Roberts Victor (Fig. 6a).However, we have shown above that the real age of enrich-ment of the RV2 peridotites at Roberts Victor was ataround 2.9 Ga. If the errorchron from Lace gives a realage it represents metasomatism at 3.27 Ga of an already de-pleted mantle because there is no correlation of the Hf iso-topes with 1/Hf and model ages increase with decreasing176Hf/177Hf (Fig. 6b). The age of partial melting in the spi-nel peridotite stability field can be placed several hundredmillion years earlier judging from the high positive eHf(+15). We estimate that 25% non-modal fractional meltingin the spinel stability field gives a Lu/Hf ratio of 0.7 whichwould yield the high eHf of +12 within about 300 Ma.Crustal ages on the E-block are periodic from 3.1 Ga on-wards up to 3.65 Ga. According to de Wit et al. (1992)the amalgamation of the E-Block from different small terr-anes occurred at 3.31 Ga. The 3.27 Ga enrichment in themantle underneath Lace may be connected with this event.The reality of the 3.27 Ga errochron would be the firstprove of the existence of mantle portions older than3.2 Ga (=oldest TRD age) underneath the E-block.

5.3.2.3. Finsch. The previous work on subcalcic garnetsfrom Finsch by Lazarov et al. (2009a) had concluded thatthese are products of two melting episodes, the first withhigh to very high degrees of partial melting at around3.6 Ga (model age of one subcalcic garnet) and the second,final depletion at 2.62 Ga and enrichment in between.

The criteria described in the “general remarks and princi-ples” section above (increase of model ages with decreasing176Hf/177Hf, no correlation of 176Hf/177Hf with 1/Hf) leadus to the re-interpretation of the 2.62 age as mostly datingthe re-enrichment (including isotopic homogenization) ofan ultra depleted mantle with eHf = +25 by a metasomatiz-ing agent which simultaneously may have triggered somesmall degrees of partial melting. Depleted upper mantle is as-sumed to have eHf = +8 at 2.62 Ga. The very high eHf showshigh depletion prior to 2.62 Ga but the timing of partial melt-ing cannot strictly be fixed. Published rhenium depletion agesfrom Finsch peridotites lie between 1.1 and 2.8 Ga (Carlsonand Moore, 2004; Griffin et al., 2004). The oldest TRD agefrom the west block is 3.23 Ga (from Newlands; Menzieset al., 1999). If we take this as the time of first melting and as-sume 25% non-modal fractional melting in the spinel stabil-ity field a Lu/Hf ratio of 0.7 results which would yield thehigh eHf of +25 at 2.62 Ga within the 600 Mio years.

Lu–Hf ages derived from subcalcic garnets alone are sys-tematically young because of the coexistence of opx (Laza-rov et al., 2009a) in the rocks. Considering the contents andLu/Hf ratios of orthopyroxene these authors concludedthat the age would increase by about 30 Ma at most. Thislies within the error of the isochron but it would bring upthe Finsch age up to 2.65 Ga. Already the garnet age aloneoverlaps with the late period of the extensive Ventersdoorpvolcanism which lasted from 2.8 to 2.6 Ga and covered a

major portion of the Kaapvaal craton. The ultra depletedFinsch peridotites could not have been the source rocksfor these volcanics. Nevertheless, they may witness this ma-jor event because they were overprinted by the fluids, whichpossibly triggered partial melting in the Ventersdoorp mag-ma source region. These may have been pervasive through-out the Kaapvaal lithosphere.

Our scrutiny of the “Lu–Hf model age versus 176Hf/177Hf”relationship from the Finsch data (Fig. 7b) revealed a furtherisochron hidden in the F2 (Hf-enriched) garnet data set withan age of 1.907 ± 0.19 Ga and eHf� 0 (Fig. 7c). Applyingthe principles from above (Fig. 8b) this may date mixing andisotopic rehomogenization at this time in proportions whichcreated the primitive mantle Lu/Hf isotope ratios just at thattime. The age coincides with the age range reported from the1.85–2.0 Ga old Kheis–Magondi belt immediately to the westof Finsch which was formed by compression and eastwardthrusting (Cornell et al., 1998; Griffin et al., 2003; Eglington,2006; Jacobs et al., 2008).

6. SUMMARY AND CONCLUSIONS

By applying a combined evaluation of major, trace ele-ment and isotope compositions of single grain subcalcicgarnets, which are considered as proxies of the bulk rockcomposition, we have been able to select samples which stillcontain unequivocal evidence of the formation of their pro-toliths and date the timing of depletion and re-enrichment.Several isochrons were retrieved from three kimberlitelocalities from the Kaapvaal craton, namely from Finschon the western border of the W-block with the 1.85–2.0 Ga Kheis–Magondi belt, from Roberts Victor near oron the Colesberg lineament and from Lace, roughly in themiddle of the E-block.

The age information for the mantle underneath theKaapvaal craton from this study, from the published Re–Os age data and the age data set of crustal events is summa-rized in Fig. 12 in a cartoon of successions of mantle melt-ing, modification and crustal growth. Our new model forthe formation and evolution of the SLCM underneath theKaapvaal craton is based on previous work from variousauthors, e.g. that of Schmitz et al. (2004) and Shirey et al.(2004). It describes that mantle melting was mainly at shal-low pressures, followed by subduction. Crust formation be-gan for parts of the E-block at about 3.65 Ga ago (Kroneret al., 1996). A correspondingly old, depleted mantle com-ponent of similar age has not been reported yet from under-neath the E-block of the Kaapvaal craton. However, theeHf initial of the 3.22 Ga enrichment errorchron from Laceis high (+12) and partial melting for the original depletionof this mantle portion should have occurred several hun-dred million years earlier. The depleted mantle underneathparts of the E-block which existed at 3.27 Ga must havebeen sufficiently stabilized to preserve a record of metaso-matism and to hold a crust with TTG’s and greenstonebelts. The existence of the W-block is enigmatic until3.2 Ga, the time of the oldest TRD’s and crustal zirconsages. In the time span between 3.1 and 2.95 Ga igneousactivity increased on the W-block with the formation ofTTG’s and greenstone belts. Before the final collision of


the West and East blocks around 2.88 Ga, a possible midocean ridge like setting existed, as indicated by the unusualhigh abundances of eclogites at Roberts Victor (Jacobet al., 2005). Also, the low pressure melting of an almostprimitive mantle in such a setting is documented by the2.95 Ga old RV1 isochron from Roberts Victor. The oce-anic crust and the depleted residues were injected into apre-depleted mantle underneath the W-block by subduc-tion. The pre-existing depleted mantle portion is deducedfrom the 3.3 Ga isochron from group RV2 garnets whichactually documents enrichment around 2.9 Ga of a depletedmantle (presumably a mantle wedge above a subductingslab). This age coincides with the occurrence of 2.88 to>2.97 Ga calcalkaline plutonism on the W-block (Schmitzet al., 2004). At around 2.88 Ga the E-block collided withthe W-block and was possibly partly subducted (Schmitzet al., 2004) underneath the W-block. A major period ofdiamond formation seems to be connected with this majorevent since eclogitic sulfide inclusions in diamonds fromKimberley (DeBeers pool), Koffiefontein but also fromfar away localities (Jwaneng and and Orapa) yield a Re–Os isochron of 2.9 Ga (Shirey et al., 2004; Shirey and Rich-ardson, 2011). The overlying crust is intruded at that timeby abundant granites. Between 2.6 and 2.8 Ga the volumi-nous Ventersdoorp volcanics covered a major part of thethen unified Kaapvaal craton. The 2.62 Ga enrichment atFinsch falls within this period and may record pervasivemetasomatism, which was related to the Ventersdoorp mag-matism. Metasomatism was renewed at Finsch at 1.9 Gathrough subduction processes connected with the attach-ment of the Kheis–Magondi belt to the Kaapvaal craton.Further metasomatic overprint occurred around 0.9–1.3 Ga (age information indicated by the Sm–Nd systemat Roberts Victor and Lace and at Finsch – Lazarovet al., 2009a) which may be related to the Namaqua–Natalorogeny in the south of the Archean crust.

ACKNOWLEDGMENTS

The authors gratefully acknowledge the continuous help andsupport in the laboratory and through discussions by G.D. Pear-son, S. Weyer, S. Aulbach, J. Heliosch, F. Kneissl, M. Lazarov,A. Neumann, J. Schastok and H.-M. Seitz. Jeff Harris was fantasticcompany in the field through his local knowledge, advice, discus-sions and social skills. Steve West from diamondcorp.plc is thankedfor allowing access to the coarse concentrates from the Lace mine.We highly appreciate the thoughtful comments of the two reviewers(Mark Schmitz and Nadine Wittig) which greatly improved thequality of the manuscript. The project was supported by the Deut-sche Forschungsgemeinschaft (BR 1012/33-1).

APPENDIX A. SUPPLEMENTARY DATA

Supplementary data associated with this article can befound, in the online version, at http://dx.doi.org/10.1016/j.gca.2013.03.010.

REFERENCES

Amelin Y. and Davis W. J. (2005) Geochemical test for branchingdecay of 176Lu. Geochim. Cosmochim. Acta 69, 465–473.

Anhaeusser C. R. (2006) A reevaluation of Archean intracratonicterrane boundaries on the Kaapvaal Craton, South Africa:collisional suture zone? GSA Spec. Pap. 405, 93–210.

Armstrong R. A., Compston W., Retief E. A. and Welker H. J.(1991) Zircon ion microprobe studies bearing on the age andevolution of the Witwatersrand triad. Precambr. Res. 53, 243–266.

Bizzarro M., Baker J. A. and Ulfbeck D. (2003) A new digestionand chemical separation technique for rapid and highlyreproducible determination of Lu/Hf and Hf isotope ratios ingeological materials by MC-ICP-MS. Geostand. Geoanal. Res.

27, 133–145.Blichert-Toft J., Chauvel C. and Albarede F. (1997) Separation of

Hf and Lu for high precision isotope analysis of rock samplesby magnetic sector multiple collector ICP MS. Contrib. Miner.

Petrol. 127, 248–260.Blichert-Toft J., Albarede F. and Kornprobst J. (1999) Lu–Hf

isotope systematics of garnet pyroxenites from Beni Bousera,Morocco: implications for basalt origin. Science 283, 1303–1306.

Blichert-Toft J. (2001) On the Lu–Hf isotope geochemistry ofsilicates. Geostand. Geoanal. Res. 25, 41–56.

Boyd F. R. (1989) Compositional distinction between oceanic andcratonic lithosphere. Earth Planet. Sci. Lett. 96, 15–26.

Boyd F. R., Pearson D. G., Nixon P. H. and Mertzman S. A.(1993) Low-calcium garnet harzburgites from Southern Africa –their relations to craton structure and diamond crystallization.Contrib. Miner. Petrol. 113, 352–366.

Brey G. P., Bulatov V. K., Girnis A. V. and Lahaye Y. (2008)Experimental melting of carbonated peridotite at 6–10 GPa. J.

Petrol. 49, 797–821.Bulatov V., Brey G. P. and Foley S. F. (1991) Origin of low-Ca

high-Cr garnets by crystallization of low-pressure harzburgites.Extended Abstract, Fifth International Kimberlite Conference,

Araxa, Brazil, 2/91. CPRM Special Publication, Rio de Jeneiro,Brazil, pp. 29–31.

Canil D. and Wei K. J. (1992) Constraints on the origin of mantle-derived low-Ca garnets. Contrib. Miner. Petrol. 109, 421–430.

Canil D. (1999) The Ni-in-garnet geothermometer: calibration atnatural abundances. Contrib. Miner. Petrol. 136, 240–246.

Carlson R. W., Pearson D. G., Boyd F. R., Shirey S. B., Irvine G.,Menzies A. H. and Gurney J. J. (1999) Re–Os systematics oflithospheric peridotites: implications for lithosphere formationand preservation. In Proceedings of the Seventh International

Kimberlite Conference. Red Roof Design, Cape Town, SouthAfrica, pp. 99–108.

Carlson R. W., Irving A. J., Schulze D. J. and Carter Hearn B.(2004) Timing of Precambrian melt depletion and Phanerozoicrefertilization events in the lithospheric mantle of the WyomingCraton and adjacent Central Plains Orogen. Lithos 77, 453–472.

Carlson R. W. and Moore R. O. (2004) Age of the EasternKaapvaal mantle: Re–Os isotope data for peridotite xenolithsfrom the Monastery kimberlite. S. Afr. J. Geol. 107, 81–90.

Caro G., Bourdon B. and Birck J. L. (2006) High-precision142Nd/144Nd measurements in terrestrial rocks: constraints onthe early differentiation of the Earth’s mantle. Geochim.

Cosmochim. Acta 70, 164–191.Cornell D. H., Armstrong R. A. and Walraven F. (1998)

Geochronology of the Proterozoic Hartley Basalt Formation,South Africa: constraints on the Kheis tectogenesis and theKaapvaal Craton’s earliest Wilson cycle. J. Afr. Earth Sci. 26,5–27.

Chapman D. S. and Pollack H. N. (1977) Regional geotherms andlithospheric thicknesses. Geology 5, 265–268.

Chu N. C., Taylor R. N., Chavagnac V., Nesbitt R. W., Boella R.M., Milton J. A., German C. R., Bayon G. and Burton K.



http://refhub.elsevier.com/S0016-7037(13)00152-X/h0005









































































(2002) Hf isotope ratio analysis using multi-collector induc-tively coupled plasma mass spectrometry: an evaluation ofisobaric interference corrections. J. Anal. At. Spectrom. 17,1567–1574.

de Wit M. J., de Ronde C. E. J., Tredoux M., Roering C., Hart R.J., Armstrong R. A., Green R. W. E., Peberdy E. and Hart R.A. (1992) Formation of an Archaean continent. Nature 357,553–562.

Deckart K., Bertrand H. and Liegeois J. P. (2005) Geochemistryand Sr, Nd, Pb isotopic composition of the Central AtlanticMagmatic Province (CAMP) in Guyana and Guinea. Lithos 82,289–314.

Drennan G. R., Robb L. J., Meyer F. M., Armstrong R. A. and deBruiyn H. (1990) The nature of the Archaean basement in thehinterland of the Witwatersrand Basin: II. A crustal profile westof the Welkom Goldfield and comparison with the Vredefortcrustal profile. S. Afr. J. Geol. 93, 41–53.

Eggins S. M., Woodhead J. D., Kinsley L. P. J., Mortimer G. E.,Sylvester P., McCulloch M. T., Hergt J. M. and Handler M. R.(1997) A simple method for the precise determination of >40trace elements in geological sample by ICP-MS using enrichedisotope internal standardization. Chem. Geol. 134, 311–326.

Eglington B. M. and Armstrong R. A. (2004) The KaapvaalCraton and adjacent orogens, southern Africa: a geochrono-logical database and overview of the geological development ofthe craton: South African. J. Geol. 107, 13–32.

Eglington B. M. (2006) Evolution of the Namaqua–Natal Belt,southern Africa – a geochronological and isotope geochemicalreview. J. Afr. Earth Sci. 46, 93–111.

Gao S., Rudnick R. L., Xu W. L., Yuan H. L., Liu Y. S., WalkerR. J., Puchtel I. S., Liu X. M., Huang H., Wang X. R. andYang J. (2008) Recycling deep cratonic lithosphere andgeneration of intraplate magmatism in the North China Craton.Earth Planet. Sci. Lett. 270, 41–53.

Girnis A. V., Bulatov V. K., Brey G. P., Gerdes A. and Hoefer H.E. (2013) Trace element partitioning between mantle mineralsand siloco-carbonatite melts at 6–12 GPa and applications tomantle metasomatism and kimberlite genesis. Lithos 160–161,183–200.

Green T. H., Blundy J. D., Adam J. and Yaxley G. M. (2000)SIMS determination of trace element partition coefficientsbetween garnet, clinopyroxene and hydrous basaltic liquids at2–7.5 GPa and 1080–1200 �C. Lithos 53, 165–187.

Griffin W. L., Cousens D. R., Ryan C. G., Sie S. H. and Suter G.F. (1989) Ni in chrome pyrope garnets – a new geothermom-eter. Contrib. Miner. Petrol. 103, 199–202.

Griffin W. L., O’Reilly S. Y., Natapov L. M. and Ryan C. G.(2003) The evolution of the lithospheric mantle beneath theKalahari Craton and its margins. Lithos 71, 215–241.

Griffin W. L., Graham S., O’Reilly S. Y. and Pearson N. J. (2004)Lithosphere evolution beneath the Kaapvaal Craton: Re–Ossystematics of sulfides in mantle-derived peridotites. Chem.

Geol. 208, 89–118.Grutter H., Latti D. and Menzies A. (2006) Cr-saturation arrays in

concentrate garnet compositions from kimberlite and their usein mantle barometry. J. Petrol. 47, 801–820.

Hofmann A. W. (1988) Chemical differentiation of the earth – therelationship between mantle, continental-crust, and oceanic-crust. Earth Planet. Sci. Lett. 90, 297–314.

Hopp J., Trieloff M., Brey G. P., Woodland A. B., Simon N. S. C.,Wijbrans J. R., Siebel W. and Reitter E. (2008) 40Ar/39Ar-agesof phlogopite in mantle xenoliths from South African kimber-lites: evidence for metasomatic mantle impregnation during theKibaran orogenic cycle. Lithos 106, 351–364.

Ionov D. A., Dupuy C., O’Reilly S. Y., Kopylova M. G. andGenshaft Y. S. (1993) Carbonated peridotite xenoliths from

Spitsbergen: implications for trace element signature of mantlecarbonate metasomatism. Earth Planet. Sci. Lett. 119, 283–297.

Ionov D. A. and Weiss D. (2002) Hf isotope composition of mantleperidotites: first results and inferences for the age and evolutionof the lithospheric mantle. In Fourth International Conference

on Orogenic Lherzolites and Mantle Processes, Abstracts,Semani, Japan, pp. 56–57.

Ionov D. A., Blichert-Toft J. and Weis D. (2005) Hf isotopecompositions and HREE variations in off-craton garnet andspinel peridotite xenoliths from central Asia. Geochim. Cosmo-

chim. Acta 69, 2399–2418.Irvine G. J., Pearson D. G. and Carlson R. W. (2001) Lithospheric

mantle evolution of the Kaapvaal Craton: a Re–Os isotopestudy of peridotite xenoliths from Lesotho kimberlites. Geo-

phys. Res. Lett. 28, 2505–2508.Irving A. J. and Frey F. A. (1978) Trace element abundances in

megacrysts and their host basalts: constraints on partitioncoefficients and megacryst genesis. Geochim. Cosmochim. Acta

48, 1201–1221.Jacob D. E., Bizimis M. and Salters V. J. M. (2005) Lu–Hf and

geochemical systematics of recycled ancient oceanic crust:evidence from Roberts Victor eclogites. Contrib. Miner. Petrol.

148, 707–720.Jacobs J., Pisarevsky S., Thomas R. J. and Becker T. (2008) The

Kalahari Craton during the assembly and dispersal of Rodinia.Precambr. Res. 160, 142–158.

Jelsma H. A., de Wit M., Thiart C., Dirks P. H. G. M., Viola G.,Basson I. J. and Anckar E. (2004) Preferential distributionalong transcontinental corridors of kimberlites and relatedrocks of Southern Africa. S. Afr. J. Geol. 107, 301–324.

Kelemen P. B., Shimizu N. and Dunn T. (1993) Relative depletionof niobium in some arc magmas and the continental crust:partitioning of K, Nb, La and Ce during melt/rock reaction inthe upper mantle. Earth Planet. Sci. Lett. 120, 111–134.

Kesson S. E. and Ringwood A. E. (1989) Slab–mantle interactions:2. The formation of diamonds. Chem. Geol. 306(78), 97–118.

Kinzler R. J. (1997) Melting of mantle peridotite at pressuresapproaching the spinel to garnet transition: Application to mid-ocean ridge basalt petrogenesis. J Geophys. Res-Sol Ea. 102,853–874.

Klein-Ben David O. and Pearson D. G. (2009) Origins of subcalcicgarnets and their relation to diamond forming fluids – casestudies from Ekati (NWT-Canada) and Murowa (Zimbabwe).Geochim. Cosmochim. Acta 73, 837–855.

Klein-Ben David O., Pettke T. and Kessel R. (2011) Chromiummobility in hydrous fluids at upper mantle conditions. Lithos

125, 122–130.Kroner A., Hegner E., Wendt J. I. and Byerly G. R. (1996) The

oldest part of the Barberton granitoid greenstone terrain, SouthAfrica: evidence for crust formation at 3.5 and 3.7 Ga.Precambr. Res. 78, 105–124.

Lazarov M., Brey G. B. and Weyer S. (2009a) Time steps ofdepletion and enrichment in the Kaapvaal craton as recordedby subcalcic garnets from Finsch (SA). Earth Planet. Sci. Lett.

27, 1–10.Lazarov M., Woodland A. B. and Brey G. P. (2009b) Thermal state

and redox conditions of the Kaapvaal mantle: a study ofxenoliths from the Finsch mine, South Africa. Lithos 112S,913–923.

Lazarov M., Brey G. B. and Weyer S. (2012a) Evolution of theSouth African mantle – a case study of garnet peridotites fromthe Finsch diamond mine (Kaapvaal craton): Part 1. Inter-mineral trace element and isotopic equilibrium. Lithos 154,193–209.

Lazarov M., Brey G. B. and Weyer S. (2012b) Evolution of theSouth African mantle – a case study of garnet peridotites from







































































































































the Finsch diamond mine (Kaapvaal craton): Part 2. Multipledepletion and re-enrichment processes. Lithos 154, 210–223.

McDonough W. F. and Sun S. S. (1995) The composition of theEarth. Chem. Geol. 120, 223–253.

Menzies A. H., Carlson R. W., Shirey S. B. and Gurney J. J. (1999)Re–Os systematics of Newlands peridotite xenoliths: implica-tions for diamond and lithosphere formation, Cape Town,1998. In Proceedings of the Seventh International Kimberlite

Conference. Red Roof Design, Cape Town, South Africa, pp.566–573.

Nowell G. M., Pearson D. G., Bell D. R., Carlson R. W., Smith C.B., Kempton P. D. and Noble S. R. (2004) Hf isotopesystematics of kimberlites and their megacrysts: new constraintson their source regions. J. Petrol. 45, 1583–1612.

O’Reilly S. Y., Griffin W. L., Pearson N. J., Jackson S. E.,Belousova E. A., Alard O. and Saeed A. (2008) Taking thepulse of the Earth: linking crustal and mantle events. Aust. J.

Earth Sci. 55, 983–995.Pearson D. G., Shirey S. B., Carlson R. W., Boyd F. R.,

Pokhilenko N. P. and Shimizu N. (1995) Re–Os, Sm–Nd, andRb–Sr isotope evidence for thick Archean lithospheric mantlebeneath the Siberian Craton modified by multistage metaso-matism. Geochim. Cosmochim. Acta 59, 959–977.

Pearson D. G. and Nowell G. M. (2002) The continentallithospheric mantle: characteristics and significance as a mantlereservoir. Philos. Trans. R. Soc. Lond. A 360, 2383–2410.

Pearson D. G. and Nowell G. M. (2004) Re–Os and Lu–Hf isotopeconstraints on the origin and age of pyroxenites from the BeniBousera peridotite massif: implications for mixed peridotite–pyroxenite mantle sources. J. Petrol. 45, 439–455.

Poujol M., Anhaeusser C. R. and Armstrong R. A. (2002) Episodicgranitoid emplacement in the Archaean Amalia-Kraaipanterrane, South Africa: confirmation from single zircon U–Pbgeochronology. J. Afr. Earth Sci. 35(1), 147–161.

Ringwood A. E. (1977) Synthesis of pyrope–knorringite solid-solution series. Earth Planet. Sci. Lett. 36, 443–448.

Robb L. R., Davis D. W. and Kamo S. L. (1990) U–Pb ages onsingle detrital zircon grains from the Witwatersrand Basin:constraints on the age of sedimentation and on the evolution ofgranite adjacent to the basin. J. Geol. T98, 311–328.

Rudnick R. L., McDonough W. F. and Chappell B. W. (1993)Carbonatite metasomatism in the northern Tanzanian mantle –petrographic and geochemical characteristics. Earth Planet. Sci.

Lett. 114, 463–475.Shirey S. B. and Richardson S. H. (2011) Start of the Wilson cycle

at 3 Ga shown by diamonds from subcontinental mantle.Science 333, 434–436.

Shirey S. B., Richardson S. H. and Harris J. W. (2004) Integratedmodel of diamond formation and craton evolution. Lithos 77,923–944.

Scherer E. E., Cameron K. L., Johnson C. M., Beard B. L.,Barovich K. M. and Collerson K. D. (1997) Lu–Hf geochro-nology applied to dating Cenozoic events affecting lower crustalxenoliths from Kilbourne Hole, New Mexico. Chem. Geol. 142,63–78.

Schmidberger S. S., Simonetti A., Francis D. and Gariepy C.(2002) Probing Archean lithosphere using the Lu–Hf isotopesystematics of peridotite xenoliths from Somerset Island kim-berlites, Canada. Earth Planet. Sci. Lett. 197, 245–259.

Schmitz M. D. and Bowring S. A. (2003) Constraints on thethermal evolution of continental lithosphere from U–Pb acces-

sory mineral thermochronometry of lower crustal xenoliths,southern Africa. Contrib. Miner. Petrol. 144, 592–618.

Schmitz M. D., Bowring S. A., de Wit M. J. and Gartz V. (2004)Subduction and terrane collision stabilize the western Kaapvaalcraton tectosphere 2.9 billion years ago. Earth Planet. Sci. Lett.

222, 363–376.Shu Q., Brey G. P., Gerdes A. and Hoefer H. E. (2003) Mantle

metasomatism recorded in subcalcic garnet xenocrysts: the

simultaneity of mantle metasomatism, diamond growth and

tectonic and magmatic events in the crust. Geochim. CosmochimActa.

Simon N. S. C., Carlson R. W., Pearson D. G. and Davies G. R.(2007) The origin and evolution of the Kaapvaal cratoniclithospheric mantle. J. Petrol. 48, 589–625.

Suhr G., Seck H. A., Shimizu N., Gunther D. and Jenner G. (1998)Infiltration of refractory melts into the lower most oceaniccrust: evidence from dunite- and gabbro-hosted clinopyroxenesin the Bay of Islands ophiolite. Contrib. Miner. Petrol. 131,136–154.

Stachel T., Viljoen K. S., Brey G. P. and Harris J. W. (1998)Metasomatic processes in lherzolitic and harzburgitic domainsof diamondiferous lithospheric mantle: REE in garnets fromxenoliths and inclusions in diamonds. Earth Planet. Sci. Lett.

159, 1–12.Stachel T., Aulbach S., Brey G. P., Harris J. W., Leost I., Tappert

R. and Viljoen K. S. (2004) The trace element composition ofsilicate inclusions in diamonds: a review. Lithos 77, 1–19.

Stachel T. and Harris J. W. (2008) The origin of cratonic diamonds– constraints from mineral inclusions. Ore Geol. Rev. 34, 5–32.

Thomas R. J., von Veh M. W. and McCourt S. (1993) The tectonicevolution of southern Africa: an overview. J. Afr. Earth Sci. 16,5–24.

van Kan Parker M., Liebscher A., Frei D., van Sijl J., vanWestrenen W., Blundy J. and Franz G. (2010) Experimentaland computational study of trace element distribution betweenorthopyroxene and anhydrous silicate melt: substitution mech-anisms and the effect of iron. Contrib. Miner. Petrol. 159, 459–473.

Walker R. J., Carlson R. W., Shirey S. B. and Boyd F. R. (1989)Os, Sr, Nd and Pb isotope systematics of southern Africanperidotite xenoliths: implications for the chemical evolution ofsubcontinental mantle. Geochim. Cosmochim. Acta 53, 1583–1595.

Walter M. J. (1998) Melting of garnet peridotite and the origin ofkomatiite and depleted lithosphere. J. Petrol. 39, 29–60.

Wittig N., Baker J. A. and Downes H. (2007) U–Th–Pb and Lu–Hfisotopic constraints on the evolution of sub-continental litho-spheric mantle, French Massif Central. Geochim. Cosmochim.

Acta 71, 1290–1311.Wittig N., Pearson D. G., Webb M., Ottley C. J., Irvine G. J.,

Kopylova M., Jensen S. M. and Nowell G. M. (2008) Origin ofcratonic lithospheric mantle roots: a geochemical study ofperidotites from the North Atlantic Craton, West Greenland.Earth Planet. Sci. Lett. 274, 24–33.

Yaxley G. M., Crawford A. J. and Green D. H. (1991) Evidence forcarbonatite metasomatism in spinel peridotite xenoliths fromwestern Victoria, Australia. Earth Planet. Sci. Lett. 107, 305–317.

Associate editor: Steven B. Shirey

























































































































geochronological and geochemical constraints on the ... · the roberts victor mine is located...

Documents