Lithos 112S (2009) 843–851

Contents lists available at ScienceDirect

Lithos

j ourna l homepage: www.e lsev ie r.com/ locate / l i thos

The Carolina kimberlite, Brazil — Insights into an unconventional diamond deposit

Lucy Hunt a,⁎, Thomas Stachel a, Roger Morton a,b, Herman Grütter c, Robert A. Creaser a

a Department of Earth and Atmospheric Sciences, University of Alberta, Edmonton, AB, Canada T6G 2E3b Sola Resource Corp, 120-3442 118th Ave. SE, Calgary, AB, Canada T2Z 3X1c BHP Billiton World Exploration Inc, #800 Four Bentall, 1055 Dunsmuir Street, Vancouver, BC, Canada V7X 1L2

⁎ Corresponding author.E-mail address: lchunt@ualberta.ca (L. Hunt).

0024-4937/$ – see front matter. Crown Copyright © 20doi:10.1016/j.lithos.2009.04.018

a b s t r a c t

a r t i c l e i n f o

Article history:Received 10 September 2008Accepted 20 April 2009Available online 3 May 2009

Keywords:Clifford's ruleAmazon CratonEclogiteClinopyroxeneGeothermobarometryTransient heating

The diamondiferous Carolina kimberlite (Rondônia State, Brazil) is located within Proterozoic basementrocks (1.8 to 1.2 Ga) of the Amazon Craton. This “unconventional” post-Archean setting is consistent with alack of harzburgitic (G10) garnets in heavy media concentrate from the kimberlite. Diamonds from Carolinahave high nitrogen contents and in part highly negative carbon isotopic values suggesting derivationpredominantly from eclogitic portions of the underlying lithospheric mantle. This is consistent with theabundance and chemistry of eclogitic garnet xenocrysts, which make up 13% of the garnets analysed: justover half of the eclogitic garnets classify as Group I (N0.07 wt.% Na2O), which is considered to be anindication of good diamond potential. Based on nitrogen contents and aggregation states, the majority of theCarolina diamonds indicate time averaged residence temperatures between 1100 and 1150 °C (at 1.5 Gamantle residence). Platelet degradation was noted in the majority of diamonds, suggesting that their mantlesource was affected by a transient heating event.Geothermobarometry on clinopyroxene grains derived from both surficial samples and kimberlite coreindicates two distinct model geotherms: a hot “Somerset Island type” geotherm (44 mW/m2), and a colder“Slave type” geotherm (38 mW/m2). Grains from the kimberlite drill core exclusively reflect the lower modelgeotherm, whereas clinopyroxenes from surficial samples depict both gradients. Given the Triassic age(230 Ma, Rb–Sr model age on phlogopite) of the Carolina kimberlite, it is speculated that a youngergeneration of Cretaceous–Tertiary kimberlites in the Pimenta Bueno area may represent the source of“hotter” mantle xenocrysts seen in surficial samples. The implied change in geotherm reflects a large scale,possibly plume related, heating episode occurring between the two kimberlite events (i.e. between theJurassic and Cretaceous) that may relate to the opening of the South Atlantic.

Crown Copyright © 2009 Published by Elsevier B.V. All rights reserved.

1. Introduction

Historically, diamond exploration has been guided by Clifford'srule (Clifford, 1966; Jelsma et al., this issue), which states thatdiamondiferous kimberlites are restricted to areas of the Earth's crustthat have been stable for at least 2 Ga. Clifford (1966) recognized twotectono-metallogenic units in Africa containing distinct mineralassemblages, with primary diamond deposits being restricted tocratonic areas that have been stable throughout younger periods oftectonism. This empirical rule was subsequently modified by Janse(1994) who concluded that primary diamond deposits are restrictedto cratons with an age N2.5 Ga (“Archons”).

Clifford's rule has recently been challenged through the discoveryof primary diamondiferous deposits in terranes with metamorphicages b2 Ga. These include the Argyle lamproites, emplaced within thePaleoproterozoic Halls Creek Mobile Zone along the SE edge of theKimberley craton (Jaques et al., 1986; Luguet et al., this issue), and the

09 Published by Elsevier B.V. All rig

Buffalo Hills kimberlites in northern Alberta (Canada), located on thePaleoproterozoic Buffalo Head Terrane (Hood and McCandless, 2004).

The diamondiferous kimberlites in the Pimenta Bueno area ofBrazil provide the strongest challenge yet, since they are located in aprovince of the Amazon craton stabilized≤1.8 Ga (Fig. 1), without anyindications for an Archean crustal pre-history. We have studieddiamonds and mantle xenocrysts from the Carolina kimberlite in thePimenta Bueno area to determine if this unconventional setting isassociated with unusual geochemical signatures. Whilst alluvialdeposits near Pimenta Bueno were known for some time, explorationfor the primary source of these diamonds only began in 1974, with theCarolina kimberlite being found in 2002.

2. Geological setting

The South American platform of Brazil is underlain by an assemblageof two cratons, the Amazon and the São Francisco (Fig.1). The Rondôniakimberlite province, including the Carolina pipe, is located on theAmazon cratonwithin basement rocks of Proterozoic age (1.8 to 1.2 Ga).The ages of these crustal rocks have been tightly constrained by anumber of studies (Tohver et al., 2005 and references therein). Little

hts reserved.

Fig. 1. Geological setting of the Carolina kimberlite. Inset adapted from Tassinari and Macambira (1999).

844 L. Hunt et al. / Lithos 112S (2009) 843–851

work has been carried out to establish the age of the underlyinglithospheric mantle. A Paleoproterozoic age of lithospheric mantle maybe inferred from the observation that crust andmantle in cratonic areasgenerally remain coupled since their formation (Pearson, 1999).Maunula (2006) has shown the presence of harzburgitic (G10) garnetsin concentrates from Pimenta Bueno kimberlites in the same tectonicregion. This observation documents the presence of strongly depleteddomains in the subcratonic lithospheric mantle which may have hadArchean protoliths.

A large continental-scale lineament, referred to as the 125°AZ,trends NW–SE across a large portion of the South American platform,stretching 3000 km from Rio de Janeiro to Rondônia. The lineamentshows a strong spatial association with the occurrence of ultramaficand alkaline intrusions, comprising kimberlites (including Carolina),kamafugites, ugandites and carbonatites (Herz, 1977; Svisero andHaralyi, 1983; Gibson et al., 1995).

Thirty-two kimberlites occur to the east of the Carolina propertywith cross-cutting relationships and limited age dating indicating

Table 1Results of the rubidium strontium model age dating of phlogopite samples.

Sample Weight(g)

Rb(ppm)

Sr(ppm)

87Rb/86Sr

87Sr/86Sr

+/−2SE Model age(0.705)

CAD-04x-A 0.00480 717.0 9.85 225.8 1.45017 0.00003 232.0CAD-04x-B 0.00277 672.5 12.90 159.0 1.22755 0.00007 231.1

845L. Hunt et al. / Lithos 112S (2009) 843–851

Carboniferous (~360–300 Ma) ages (Maunula, 2006). The majority ofthe kimberlites are moderately well-preserved and typically of crateror diatreme facies (Evans, 2005; Masun and Scott Smith, 2008).Structurally, the Rondônia kimberlite province shows an associationwith the Palaeozoic Pimenta Bueno- and Colorado-grabens which arealigned with the 125°AZ lineament. Approximately 200 km furthersouth-east along the lineament lie a number of Cretaceous kimberlitesof the Juina area, Mato Grosso (Heaman et al., 1998; Andreazza et al.,2008). The Juina Field is one of a number of Cretaceous alkalinevolcanic centres developed along the lineament attributed to thepassage of the Brazilian continent over the Trindade Plume (Bulanovaet al., 2008b).

3. Samples and analytical methods

3.1. Samples

Twenty-five diamond samples ranging from 3.0 to 31.7 mg wereprovided for destructive analysis, a further 230 diamondswere obtainedfor non-destructive nitrogen analysis (~3–50 mg). The diamonds wererecovered from the Carolina kimberlite pipe during exploration bulksampling. Before analysis, the samples were cleaned in an ultrasonicbath using petroleum ether to remove surface contamination.

Diamonds collected for destructive analysis were crushed in a steelbreaker and fragments were selected for FTIR (Fourier TransformInfrared) analyses to quantify nitrogen concentrations and character-ise nitrogen defects. These fragments were then further analysed forcarbon isotopic composition.

Garnet and clinopyroxene xenocrysts were selected from heavymineral concentrate derived from both kimberlite drill core andsurficial samples collected from panned concentrate from the top ofthe Carolina kimberlite.

3.2. Analytical techniques

Nitrogen concentrations and aggregation states in the Carolinadiamonds were determined using a Thermo-Nicolet FTIR spectro-meter coupled with a Continuum microscope equipped with a KBrbeam splitter. Sample spectra (650 to 4000 cm−1) were collected for250 s with a spectral resolution of 4 cm−1, and, after subtraction of apure Type II diamond spectrum, converted to absorption coefficients.The spectra were deconvoluted into A, B and D components usingsoftware provided by David Fisher (Diamond Trading Company (DTC)Research Centre, Maidenhead, UK). Nitrogen concentrations, inatomic ppm, were calculated from absorption coefficient values at1282 cm−1 for the A- and B centre (Boyd et al., 1994, 1995).Transmission analysis on all but 25 samples was carried out on wholediamonds. Due to large sample thickness the nitrogen peak reachedsaturation for some samples and calculated nitrogen concentrationsfor these diamonds represent minimum values; these analyses werenot included in the final data set.

Hydrogen related absorbance at 3107 cm−1 and platelet peakintensities were quantified by determining the areas beneath therespective peaks on spectra converted to absorption coefficients (i.e.normalized to a sample thickness of 1 cm). The peak areas weremeasured using the diamond spectrum as a base line.

Nitrogen detection limits are 10–20 ppm, depending largely on thespectral quality of the sample. Accuracy of the reported nitrogenconcentrations and aggregation states is generally within 10–20%relative (1σ) (Banas et al., 2007).

Carbon isotopic analyses were performed on a subset of 25 stones.Inclusion-free diamond fragments (0.7–1.9 mg) were combustedtogether with 1.0–2.0 g CuO in an evacuated and sealed silica glasstube for a minimum of 10 hours at 950 °C. Carbon isotope ratios (δ13C)were measured using a Finnigan MAT 252 mass spectrometer, witha total analytical precision better than ±0.1‰ (1σ, see Donnelly

et al., 2007). Data are reported in standard per mil notation relativeto the V-PDB standard.

The silicate xenocrysts were embedded in Araldite® epoxy resinand final polish was achieved using 0.05 μm alumina suspension on apolishing wheel. The major and minor element compositions weredetermined by wavelength-dispersion spectrometry (WDS) on a JEOLJXA-8900 Superprobe using silicate, oxide and metal standards. Theresults reported represent averages of multiple spot analyses.Operating with an accelerating voltage of 20 kV and a beam currentof 20 nA, counts for each element were collected for 30–60 seconds(peak) and for half the time on each peak side.

The age of the kimberlite was determined using phlogopitemacrocrysts taken from kimberlitic drill core. Samples were pickedafter careful inspection to ensure no alteration or leaching hadoccurred. Due to the pervasive nature of tropical weathering in thearea, representative samples were analysed using a Rigaku GeigerflexPower Diffractometer with a Co tube and a graphite monochromatorto determine the purity of a powdered bulk sample. Absence ofalteration along cleavage boundaries was verified using EMPAtransects over small portions of the phlogopite grains. The analyticalprocedures for Rb–Sr phlogopite dating and sequential leachingtechniques are outlined in Creaser et al. (2004).

4. Results

4.1. Kimberlite phlogopite macrocryst Rb–Sr model age

Fresh phlogopite macrocryst samples were analysed to determinethe age of kimberlite emplacement. The samples show high Rb and lowSr contents typical of kimberlitic phlogopite. Both analyses yieldidentical model ages within their quoted uncertainties of ±1% of232.0 and 231.1+/−2.3 Ma (Table 1). A Triassic (~Ladinian) emplace-ment age is thus implied for the Carolina kimberlite.

4.2. Diamonds

4.2.1. Diamond characteristicsThe subset of 25 diamonds for destructive analysis was characterised

based on their colour, morphology and surface features. The diamondsshowed a range of colour from colourless to yellow, pink, brown andgrey. Grey diamonds were dominant comprising 48% of the sample set,colourless diamondsmadeup23%,with the remaining29% consisting ofyellow (16%), brown (10%), and pink (3%). Brown and pink colourationindicates plastic deformation of the diamond (Harris, 1987), otherfeatures indicative of deformation, such as deformation lines ondodecahedral faces were, however, absent in the majority of diamonds.

Morphologically, Carolina diamonds examined in this studycomprise irregular crystals, octahedra, rounded dodecahedra andoctahedral macles. Irregular shapes dominated and included frag-ments and aggregates. Where possible, irregular stones were classedaccording to residual octahedral and dodecahedral faces. Dodecahe-dral crystals made up 32% of the sample set, with another 26%comprised of irregular dodecahedra. A further 26% were classified asirregular with no original morphology being identified. The breakagesurfaces of these stones often showed no evidence of secondaryresorption indicating breakage post kimberlite emplacement, mostlikely during recovery. The remaining 16% comprised octahedralstones with 3% of those being octahedral macles.

Fig. 2. The carbon isotopic values for the Carolina diamonds. A mode at −5‰ is seen,with a tail stretching to much more negative values.

846 L. Hunt et al. / Lithos 112S (2009) 843–851

Surface features are related to the morphology of the diamond,with certain features found on dodecahedral faces, and others foundexclusively on octahedral faces. As a consequence of the abundanceof dodecahedral stones in the sample set, features restricted tothese faces dominate and mainly consist of terraces and hillocks ofvarying forms. Ruts and corrosion sculptures were also noted on moreresorbed faces.

Graphite was commonly observed within the diamonds studied,occurring mostly as small thin flakes. The mineral inclusion

Fig. 3. Histogram of nitrogen contents of Carolina diamonds. Insert shows nitrogencontents for diamonds with eclogitic inclusions from worldwide sources (data base ofStachel and Harris, 2008). Based on overall higher total nitrogen contents in eclogiticrelative to peridotitic diamonds, eclogitic sources appear likely for the mostly nitrogenrich Carolina diamonds.

assemblage in this small selection of 25 diamonds was limited, withonly sulphides being visible in a few (6) samples.

4.2.2. Carbon isotopic valuesThe diamond samples showed a range in carbon isotopic composi-

tion from−22.4 to−3.7‰ (Fig. 2), with a high proportion (68%) of thediamonds having mantle-like carbon isotopic signatures (δ13C near−5‰), analogous to diamonds worldwide. Worldwide, peridotiticdiamonds show a narrow range in δ13C (about −10 to −2‰), whilsteclogitic diamonds have isotopic compositions that may extend to verynegative (around−38‰) values (Kirkley et al.,1991; Cartigny, 2005). Inthis limited sample set there was no correlation observed betweencarbon isotopic composition and morphology or colour.

4.2.3. Nitrogen content and aggregation stateNitrogen is themain impurity in diamond (Kaiser and Bond,1959),

substituting for carbon atoms in the crystal lattice. All the diamondsstudied from the Carolina pipe contained measurable nitrogen, withconcentrations ranging from 40 to 2120 at. ppm, and consequently areclassified as Type I (Fig. 3). Worldwide, lithospheric inclusion bearingdiamonds without detectable nitrogen (Type II: ≤10 ppm nitrogen)comprise 20% of all samples (Stachel, 2007). The lack of Type IIdiamonds at Carolina is unexpected, particularly in view of theirabundance from the Machado River alluvials 35 km downstream(Bulanova et al., 2008a).

Type I diamondsmaybe further subdividedbasedon theaggregationstate of their nitrogen impurities (Evans et al., 1981). Diamonds withsingly substituted nitrogen are classified as Type Ib and are very rare innature. Diamondswith aggregated nitrogen (Type Ia) reflect a temporalevolution from the single substitutional state through aggregation intopairs (Type IaA) to four nitrogen atoms surrounding a vacancy (TypeIaB) (Davies, 1976; Evans et al., 1981). Intermediate aggregation states(10–90% of nitrogen in B centres) are termed Type IaAB.

Aggregation of nitrogen occurs during mantle residence anddepends on time, total nitrogen content and temperature. At mantletemperatures the transition from Type Ib to IaA occurs very rapidly

Fig. 4. Time averaged mantle residence temperatures determined from nitrogencontent and aggregation state. Nitrogen aggregation (%B) is expressed as the relativeproportion of nitrogen in the B centre— determined as atomic percent of nitrogen in theB centre divided by the sum of the atomic percents of the nitrogen in the A and Bcentres. Isotherms based on 1.5 Ga mantle residence time (isotherms for 3 Ga mantleresidence time are shown as thin grey lines) are calculated after Leahy and Taylor(1997) and Taylor et al. (1990).

Fig. 5. A) A plot of the intensity of the platelet peak (I(B′)) versus the percentage ofnitrogen aggregated into the B component. The solid line represents the linearrelationship between platelet peak intensity and %B for regular diamonds (Woods,1986). B) The same plot separating samples according to their time averaged mantleresidence temperatures. Arrows indicate increasing platelet degradation with increas-ing temperature.

Fig. 6. Plot showing the relationship of the 3107 cm−1 hydrogen peak area and totalnitrogen content for Carolina diamonds.

847L. Hunt et al. / Lithos 112S (2009) 843–851

whilst aggregation from Type IaA to IaB generally takes hundreds ofmillions to billions of years (Evans et al., 1981). Ninety-three percentof the Carolina diamonds classify as Type IaAB, indicating inter-mediate aggregation states. This is common amongst diamonds fromother localities (Stachel, 2007). Type IaA diamonds comprise 6% of thisdiamond population, with just 0.5% classifying as Type IaB.

Time averaged mantle residence temperatures can be determinedfor an assumedmantle residence time using the aggregation state andconcentration of nitrogen (Evans and Harris, 1989; Taylor et al., 1990).These temperatures normally show acceptable agreement withinclusion geothermobarometry (e.g. Leahy and Taylor, 1997). Whilstthe exact residence time of the Carolina diamonds is unclear, it hasbeen shown that mantle residence temperatures are fairly insensitiveto the approximated residence time (Leahy and Taylor, 1997).Calculated for a mantle residence of 1.5 Ga, the majority of Carolinadiamonds indicate residence temperatures between 1100 and 1150 °C(Fig. 4).

4.2.4. Platelet peakWith the aggregation of nitrogen from the A to B centre comes the

concurrent growth of B′ centres, with an absorbance peak between1358 and 1380 cm−1. These centres are defects within the crystallattice caused by platelets, the composition of which has been heavilydebated in the literature. Most authors now regard platelets asinterstitial carbon atoms which are likely displaced in the diamond

lattice with the aggregation of nitrogen (Woods, 1986). The nitrogen,detected in variable amounts, likely reflects a contaminant rather thanan essential component (Goss et al., 2003 and references therein).

Previous work (Woods, 1986) has determined a linear relation-ship between the intensity of the peak (I(B′)) and the relative proportionof nitrogenpresent in the B centre (%B) for regular specimens. A subset ofirregular specimens was also identified where this relationship wasnot observed and for these diamonds ‘catastrophic platelet degradation’(Woods, 1986) was proposed as a consequence of short lived heatingevents (Evans et al., 1995) and possible strain (Woods, 1986).

For the 222 samples with spectra suitable for quantitative analysisof nitrogen there was no measurable platelet peak component in 11%of the stones. In addition, the intensity of the peak component of allbut 10 of the samples, when plotted against %B, fell below the linearrelationship proposed for regular diamonds (Fig. 5A).

4.2.5. 3107 cm−1 hydrogen peakA number of peaks in the absorbance spectrum of diamond have

been attributed to hydrogen impurities within the diamond lattice(Woods and Collins, 1983). In this study, only the intensity of thehydrogen peak at 3107 cm−1 was determined. The weaker hydrogenpeaks at 1405, 2786, 4169 and 4469 cm−1 are only observable at highintensities of the 3107 cm−1 peak. The 3107 cm−1 peak likely relatesto stretching vibrations of CfCH2 (vinylidene) bonds (Woods andCollins, 1983). The hydrogen peak intensity (measured as a peak area)varied from 0 to 86 cm−1 although most (91% of diamonds) fell in therange between 0 and 10 cm−1 (Fig. 6).

For Argyle diamonds a linear relationship was observed betweennitrogen content and the 3107 cm−1 hydrogen absorption peak(Iakoubovskii and Adriaenssens, 2002). Such a correlation is notvisible for Carolina diamonds (Fig. 6). Iakoubovskii and Adriaenssens(2002) suggested that relatively nitrogen rich and hydrogen poordiamonds (as observed for Carolina) could relate to a purely nitrogen-rich growth environment, as opposed to an ammonium-rich growthmedium for the Argyle stones.

4.3. Xenocrysts

4.3.1. Garnet xenocrystsGarnet xenocrysts from coarse concentrate derived from the

surface of the kimberlite (n=349) and from drill core (n=147)have been analysed. Electron microprobe analyses (EPMA) show thegarnets to be derived predominantly from peridotitic sources (87%),which are exclusively lherzolitic (G9) in paragenesis (Fig. 7). Grütteret al. (2004) showed that lherzolitic garnets plotting close to the G10/

848 L. Hunt et al. / Lithos 112S (2009) 843–851

G9 divide are typically associated with cool cratonic lithosphericmantle (G9A), whereas garnets removed towards the Ca rich sideaway from the divide are characteristic of off-craton kimberlites(G9B). Peridotitic garnets derived from the surface of the Carolinakimberlite typically plot in the G9B field whereas garnets taken fromthe drill core predominantly plot in the G9A field (Fig. 7).

Thirteen percent of the garnets belong to the eclogitic suite. Eclogiticgarnets are distinctlymore abundant in the drill core (28%) compared tosurficial samples (9%) and derive from distinct sources with 96% of thegarnets fromthekimberlite core classifying asGroup I (N0.07wt.%Na2O;McCandless and Gurney, 1989) as opposed to only 28% of concentrategarnets. High sodium(N0.07wt.%Na2O) in garnet has been identified asa characteristic of diamond facies eclogites and hence is considered anindication for increased diamond potential of the host kimberlite(Gurney, 1984; Gurney et al., 1993).

Fig. 8. Plot of Cr2O3 versus Al2O3 for clinopyroxenes with the compositional fields ofRamsay (1992). Themajority of grains plot in the “on craton garnet peridotite” field. Thexenocrysts are separated based on whether they were derived from kimberlitic drillcore or surficial samples.

Fig. 7. Plots of Cr2O3 versus CaO in garnet with fields for garnet classes (G1 to G12) afterGrütter et al. (2004). G9A garnets are typically associated with cool cratoniclithospheric mantle, whilst G9B garnets are characteristic of off-craton kimberlites(Grütter et al., 2004). A) Garnets from kimberlitic drill core. B) Garnets from surficialsamples.

4.3.2. Clinopyroxene xenocrystsAll but 4 of 129 clinopyroxene xenocrysts recovered from surficial

samples and drill core fall within the ‘on-craton’ garnet peridotite fieldin the Cr2O3/Al2O3 discrimination diagram of Ramsay (1992) (Fig. 8).The majority of clinopyroxenes from the kimberlite have lowerAl2O3 wt.% than those from the surface concentrate.

Fig. 9. P–T estimates based on the clinopyroxene thermobarometer of Nimis and Taylor(2000). Data from the Slave Craton, Kirkland Lake and Somerset Island are shown forreference to indicate low, intermediate and high cratonic geotherms. Different symbolsare used to separate P–T estimates based on clinopyroxene derived from kimberlite coreand from surficial samples. Geotherms after conductive models of Chapman and Pollack(1977).

849L. Hunt et al. / Lithos 112S (2009) 843–851

For clinopyroxene from garnet lherzolites, pressure and tempera-ture conditions of last equilibration may be estimated using the singlecrystal geothermobarometer of Nimis and Taylor (2000). This semi-empirical thermobarometer satisfies experimental and phase-com-patibility constraints to within acceptable error for a wide variety ofperidotitic bulk compositions and a broad range of pressures andtemperatures (Grütter and Moore, 2003; Grütter, this issue).

Using well studied Canadian localities for comparison (Grütter andMoore, 2003), the calculated pressure–temperature conditions reflecttwo distinct geotherms: a hot “Somerset Island type” geotherm(44 mW/m2), and a colder “Slave type” geotherm (38 mW/m2)(Fig. 9). Clinopyroxenes from kimberlite core exclusively fall on thecolder model geothermwhilst surficial clinopyroxenes predominantlydelineate the hotter geotherm.

5. Discussion

5.1. Diamond source region

Using the Cr2O3/Al2O3 discrimination diagramof Ramsay (1992), theclinopyroxene xenocrysts from the Carolina kimberlite indicate deriva-tion from cratonic garnet peridotites (Fig. 8). Garnet xenocrysts,however, are atypical for diamond prospective cratonic source regions.The complete absence of harzburgitic (G10) garnets is unusual for adiamondiferous kimberlite (Fig. 7), but is consistent with the post-Archean setting of the area (c.f. Griffin et al., 1999, 2003; Hood andMcCandless, 2003). This is believed to relate to secular cooling of theearth, with mantle potential temperatures after the Archean being toolow to allow for the high degrees of melt extraction required to createharzburgitic–dunitic residues (e.g. Stachel et al., 1998; Walter, 1998;Shirey et al., 2004). The absence of harzburgitic garnets thus is notsuggestive of a decoupled origin of crust and mantle beneath theCarolina kimberlite. It may be taken to imply that the underlyinglithospheric mantle has a similar age as the basement rocks of the area,dated at 1.8–1.55 Ga (e.g. Tassinari and Macambira, 1999).

The absence of G10 garnets (Gurney, 1984) suggests that Carolinadiamonds are not derived from peridotitic sources but rather points toan eclogitic diamond population. Eclogitic garnet xenocrysts are presentat Carolina (13% of garnets) and are particularly abundant in kimberlitecore (28% of garnets). Eclogitic garnets from kimberlite core almostexclusively plot in thefieldofdiamond facies eclogites (N0.07wt.%Na2O,Gurney,1984;McCandless andGurney,1989);whilst in surficial samplesa large (72%) proportionof eclogitic garnets is Napoor andhence relatedto sources with low diamond potential. The notion of an eclogiticdiamond population at Carolina is consistent with the occurrence ofhighly negative carbon isotopic values and overall high nitrogencontents (e.g. Cartigny, 2005; Stachel, 2007). This conforms with thepresence of an eclogitic diamond population in the Machado Riveralluvials downstream, as identified by Bulanova et al. (2008a) fromdiamond inclusion studies.

5.2. Thermal evolution of the source region

The nitrogen contents and aggregation states of the diamondsindicate that, for a mantle residence of 1.5 Ga, the majority of Carolinadiamonds resided at temperatures between 1100 and 1150 °C (Fig. 4).The observed range in values from ~1050–1300 °C is consistent withdiamondpopulations elsewhere (Taylor et al.,1995). Theobserved rangein temperatures likely relates to sampling of diamonds by the kimberliteat different depths in the lithospheric mantle, with deeper diamondsresiding at higher temperatures. The small range in temperatures forthe majority of diamonds indicates preferred sampling over a restricteddepth range from sources with a uniform thermal history (e.g. PointLake kimberlite, Canada: Taylor and Milledge, 1995).

Whilst the diamonds appear to have resided at fairly uniformconditions, the observation of common platelet degradation among

Carolina diamonds is evidence for thermal perturbations and/or strainaffecting the diamond source regions prior to kimberlite emplacement.Almost nowork has been carried out on determining the affect of strain,and relatively little is known about the temperatures required forcatastrophic platelet degradation to occur. Experimentally, Evans et al.(1995) showed that platelets become unstable at temperatures ofbetween2500 and 2700 °C. The number of transient events that affectedthe diamond source regions and their timing, e.g. a possible temporalrelationship with the initiation of kimberlite magmatism, cannot beconstrained from the nitrogen data. However any heating events musthave been short lived or occurred just prior to eruption of the kimberliteelse the aggregation state of the diamonds would be higher. Evaluatingplatelet degradation in conjunction with nitrogen based mantleresidence temperatures clearly shows that strong platelet degradationcorrelates with high residence temperatures and, by inference, withgreat depths of origin (Fig. 5B). This is consistentwith theheat source forplatelet degradation being upwelling asthenosphericmelts, and/orwithincreased strain with depth.

The largest change to the diamond source region occurred after theeruption of the Carolina kimberlite. Evidence of a much larger heatingevent postdating emplacement of the Carolina kimberlite is providedby the clinopyroxene geothermobarometry: whilst clinopyroxenesfrom the Carolina kimberlite itself imply a “Slave type” geotherm(38 mW/m2) typically associated with stable cratonic areas, P–T datafor grains collected from the surface define a second fairly hot“Somerset Island type” geotherm (44 mW/m2), indicative of massivemelt infiltration and hence reduced diamond potential (Fig. 9). Thisimplies that in surficial samples, clinopyroxene from younger volca-nic sources associated with a higher geothermal gradient are mixedwith material derived from Carolina itself. This heating event is alsoevidenced by the garnet xenocrysts. Garnets from the kimberlite coreplot predominantly in the G9A field of Grütter et al. (2004), indicativeof a cool cratonic geotherm, whilst xenocrysts taken from the surfaceshow a shift to more Ca rich values (G9B field) typical for off-cratonkimberlites and a hotter geotherm. Given the Triassic age of theCarolina kimberlite, it is speculated that a younger generation ofCretaceous–Tertiary kimberlites in the Pimenta Bueno area couldrepresent this second source of clinopyroxene and garnet xenocrysts.These may or may not be related to the Cretaceous kimberlites of theJuina area, Mato Grosso which lie 200 km south-east along the same125°AZ lineament (Heaman et al., 1998; Andreazza et al., 2008).

Maunula (2006) noted a similar difference for samples from theVaaldiam Pepper and Tumeleiro kimberlite provinces and those taken~10 kmdownstream. Alluvial garnet samples showed amore enrichedmantle signature, whilst the clinopyroxene geotherm for the alluvialgrains was elevated by comparison to those grains from thekimberlites. Maunula (2006) speculated the presence of yet undis-covered kimberlites from which the alluvial grains were sampling.

By implication, in the time span between the Triassic and theCretaceous heating of lithospheric mantle beneath the south-easternAmazon Craton occurred. Similar observations have previously beenmade e.g. for the Sao Francisco Craton, where thermal erosion of thedeep mantle lithosphere occurred in a time window between 95 and85 Ma (Read et al., 2004; Grütter, this issue). Lithosphere extensionand incipient rifting during the opening of the South Atlantic orimpact of a plume have been suggested as possible causes for thethermal erosion of the lithospheric root beneath the Sao FranciscoCraton, but the evidence is inconclusive (Read et al., 2004).

During the Cretaceous South America underwent major crustalevolutionwith the breakup of Gondwanaland and the associated plumeactivity. In the Early Cretaceous, the impact of the Tristan and St Helenaplume heads beneath Gondwana is believed to have been responsiblefor the extensive Paraná–Etendeka magmatism (White and Mckenzie,1989) during the final stages of supercontinent break-up and opening ofthe SouthAtlantic. In addition, large-volumemafic, potassicmagmatismin south-east Brazil between the Amazon and São Francisco cratons

850 L. Hunt et al. / Lithos 112S (2009) 843–851

(Gibson et al., 1995) has also been linked to plume activity in the lateCretaceous. Gibson et al. (1995) proposed that the Alto ParanaíbaProvince alkaline magmatism was caused by impingement of theTrindade plume on the base of the lithosphere. Any of these events mayhave transported sufficient heat into the lithospheric mantle beneaththe Pimenta Bueno area to cause the observed increase in geothermalgradient from 38 to 44 mW/m2 surface heat flow.

6. Conclusions

With its setting in a basement province≤1.8 Ga in age, the Carolinakimberlite represents an unconventional diamond deposit. Althoughsome G10 garnets have been reported from neighbouring kimberlites, acomplete absence of harzburgitic garnet xenocrysts at Carolina isconsistent with this post-Archean setting. Eclogitic garnet xenocrystswith high sodium contents (Na2ON0.7 wt.%), indicative of the presenceof diamond facies eclogite at depth, combined with high nitrogencontents and in part highly negative carbon isotopic values of thediamonds themselves, suggest derivation of Carolina diamonds mainlyfrom eclogitic source regions. Mantle storage of the diamonds occurredat temperatures that are otherwise typical for Archean mantle sources(1100–1150 °C). One or more short lived heating events during mantlestorage are revealed by platelet peak degradation in the majority ofdiamonds.

A typical cratonic model geotherm (~38 mW/m2 surface heat flow)is implied by clinopyroxene xenocrysts derived directly from theCarolina kimberlite. Clinopyroxenes sourced from the top of thekimberlite, likely derived from both local and regional sources, mainlyfall along a hotter 44 mW/m2 model geotherm. This suggests that aftereruption of the Triassic Carolina kimberlite a large scale heating eventaffected the underlying lithospheric mantle and its previously existingdiamond source regions. This heating eventmay relate to intense plumeactivity affecting the Amazon Craton throughout the Cretaceous.

Acknowledgements

The authors would like to thank Sergei Matveev for all his help withthe electron microprobe, as well as Karlis Muehlenbachs for the use ofhis stable isotope lab and equipment. L.H. acknowledges support fromthe Albert Ingenuity Fund. T.S. received funding through an NSERCDiscovery Grant and the Canada Research Chairs programme.

References

Andreazza, P., Kaminsky, F.V., Sablukov, S.M., Belousova, E.A., Tremblay, M., Griffin, W.L.,2008. Kimberlitic sources of super-deep diamonds in the Juina area, Mato GrossoState, Brazil. Extended Abstract No. 9IKC-A-00004, 9th International KimberliteConference, Frankfurt.

Banas, A., Stachel, T., Muehlenbachs, K., McCandless, T.E., 2007. Diamonds from the BuffaloHeadHills, Alberta: formation in a non-conventional setting. Lithos 93 (1–2),199–213.

Boyd, S.R., Kiflawi, I., Woods, G.S., 1994. The relationship between infrared-absorptionand the A defect concentration in diamond. Philosophical Magazine. B 69 (6),1149–1153.

Boyd, S.R., Kiflawi, I., Woods, G.S., 1995. Infrared-absorption by the B-nitrogen aggregatein diamond. Philosophical Magazine. B 72 (3), 351–361.

Bulanova, G.P., Smith, C.B., Kohn, S.C., Walter, M.J., Gobbo, L., Kearns, S., 2008a. MachadoRiver Brazil, Brazil — a newly recognised ultradeep diamond occurrence. ExtendedAbstract No. 9IKC-A-00233, 9th International Kimberlite Conference, Frankfurt.

Bulanova, G.P., Smith, C.B., Walter, M.J., Blundy, J., Gobbo, L., EIMF, Kohn, S.C., 2008b.Proto-kimberlitic ultra-deep diamonds from Collier 4 kimberlite pipe, Juina, Brazil.Extended Abstract No. 9IKC-A-00227, 9th International Kimberlite Conference,Frankfurt.

Cartigny, P., 2005. Stable isotopes and the origin of diamond. Elements 1 (2), 79–84.Chapman, D.S., Pollack, H.N., 1977. Regional geotherms and lithospheric thickness.

Geology 5 (5), 265–268.Clifford, T.N., 1966. Tectono-metallogenic units and metallogenic provinces of Africa.

Earth and Planetary Science Letters 1 (6), 421–434.Creaser, R.A., Grütter, H., Carlson, J., Crawford, B., 2004. Macrocrystal phlogopite Rb–Sr

dates for the Ekati property kimberlites, Slave Province, Canada: evidence formultiple intrusive episodes in the Paleocene and Eocene. Lithos 76 (1–4), 399–414.

Davies, G., 1976. A-nitrogen aggregate in diamond— its symmetry and possible structure.Journal of Physics C 9 (19), L537–L542.

Donnelly, C.L., Stachel, T., Creighton, S., Muehlenbachs, K., Whiteford, S., 2007.Diamonds and their mineral inclusions from the A154 South pipe, Diavik DiamondMine, Northwest Territories, Canada. Lithos 98 (1–4), 160–176.

Evans, B.T., 2005. Technical Report on the Diamond Exploration Potential of the CarolinaProperty, Rondônia State. Sola Resources Corporation., Brazil.

Evans, T., Harris, J.W., 1989. Nitrogen aggregation, inclusion equilibration temperaturesand the age of diamonds. In: Ross, J. (Ed.), Kimberlites and Related Rocks. In: GSASpecial Publication, vol. 14. Blackwell, Carlton, pp. 1001–1006.

Evans, T., Qi, Z., Maguire, J., 1981. The stages of nitrogen aggregation in diamond. Journalof Physics C 14 (12), L379–L384.

Evans, T., Kiflawi, I., Luyten, W., Vantendeloo, G., Woods, G.S., 1995. Conversion of plateletsinto dislocation loops and voidite formation in Type IaB diamonds. Proceedings ofthe Royal Society of London. Series A 449 (1936), 295–313.

Gibson, S.A., Thompson, R.N., Leonardos, O.H., Dickin, A.P., Mitchell, J.G., 1995. The LateCretaceous impact of the Trindade mantle plume — evidence from large-volume,mafic, potassic magmatism in SE Brazil. Journal of Petrology 36 (1), 189–229.

Goss, J.P., Coomer, B.J., Jones, R., Fall, C.J., Briddon, P.R., Oberg, S., 2003. Extended defectsin diamond: the interstitial platelet. Physical Review B 67 (16), 165208.

Griffin, W.L., O Reilly, S.Y., Ryan, C.G., 1999. The composition and origin ofsubcontinental lithospheric mantle. In: Fei, Y., Bertka, C.M., Mysen, B.O. (Eds.),Mantle Petrology: Field Observations and High Pressure Experimentation: Atribute to Francis R. (Joe) Boyd. In: Special Publication. The Geochemical Society,Houston, pp. 13–45.

Griffin, W.L., O Reilly, S.Y., Abe, N., Aulbach, S., Davies, R.M., Pearson, N.J., Doyle, B.J.,Kivi, K., 2003. The origin and evolution of Archean lithospheric mantle.Precambrian Research 127 (1–3), 19–41.

Grütter, H.S. 2009, this issue. Pyroxene xenocryst geotherms: Techniques andapplication. Proceedings of the 9th International Kimberlite Conference. Lithos112S, 1167–1178.

Grütter, H., Moore, R.O., 2003. Pyroxene geotherms revisited — an empirical approachbased on Canadian xenoliths. Extended Abstract — 8th International KimberliteConference, CD-ROM, p. 272.

Grütter, H.S., Gurney, J.J., Menzies, A.H., Winter, F., 2004. An updated classificationscheme for mantle-derived garnet, for use by diamond explorers. Lithos 77 (1–4),841–857.

Gurney, J.J.,1984. Acorrelationbetweengarnets anddiamonds inkimberlites. In:Glover, J.E.,Harris, P.G. (Eds.), Kimberlite Occurrences and Origins: A Basis for Conceptual Modelsin Exploration: Geology Department and University Extension, University of WesternAustralia, Special Publication, vol. 8, pp. 143–166.

Gurney, J.J., Helmstaedt, H., Moore, R.O., 1993. A review of the use and application ofmantle mineral geochemistry in diamond exploration. Pure and Applied Chemistry65 (12), 2423–2442.

Harris, J.W., 1987. Recent physical, chemical, and isotopic research of diamonds. In:Nixon, P.H. (Ed.), Mantle Xenoliths. InJohn Wiley and Sons.

Heaman, L.M., Teixeira, N.A., Gobbo, L., Gaspar, J.C., 1998. U–Pb mantle zircon ages forkimberlites from the Juina and Paranatinga provinces, Brazil. VIIth InternationalKimberlite Conference, Extended abstracts, Cape Town, pp. 322–324.

Herz, N., 1977. Timing of spreading in South-Atlantic — information from Brazilianalkalic rocks. Geological Society of America Bulletin 88 (1), 101–112.

Hood, C., McCandless, T., 2003. Systematic variations in xenocryst mineral compositionat the province scale, Buffalo Hills kimberlites, Alberta, Canada. 8th InternationalKimberlite Conference, Victoria (BC), Canada, p. 3.

Hood, C.T.S., McCandless, T.E., 2004. Systematic variations in xenocryst mineralcomposition at the province scale, Buffalo Hills kimberlites, Alberta, Canada. Lithos77 (1–4), 733–747.

Iakoubovskii, K., Adriaenssens, G.J., 2002. Optical characterization of natural Argylediamonds. Diamond and Related Materials 11 (1), 125–131.

Janse, A.J.A., 1994. Is Clifford's rule still valid? Affirmative examples from around theworld. In: Meyer, H.O.A., Leonardos, O.H. (Eds.), Diamonds: Characterization,Genesis and Exploration: CPRM — Special Publication Jan/94, Brasilia, pp. 215–235.

Jaques, A.L., Lewis, J.D., Smith, C.B., 1986. The kimberlites and lamproites of WesternAustralia. Geological Survey of Western Australia Bulletin 132 268 pp.

Jelsma, H., et al. 2009, this issue. Proceedings of the 9th International KimberliteConference. Lithos 112S, 155–165.

Kaiser, W., Bond, W.L., 1959. Nitrogen, a major impurity in common type-I diamond.Physical Review 1 (4), 857–863.

Kirkley, M.B., Gurney, J.J., Otter, M.L., Hill, S.J., Daniels, L.R., 1991. The application of Cisotope measurements to the identification of the sources of C in diamonds — areview. Applied Geochemistry 6 (5), 477–494.

Leahy, K., Taylor, W.R., 1997. The influence of the Glennie domain deep structure on thediamonds in Saskatchewan kimberlites. Russian Geology and Geophysics 38 (2),481–491.

Luguet, A., et al. 2009, this issue. An integrated petrological, geochemical and Re–Osisotope study of peridotite xenoliths from the Argyle lamproite, Western Australiaand implications for cratonic diamond occurrences. Proceedings of the 9thInternational Kimberlite Conference. Lithos 112S, 1096–1108.

Masun, K.M., Scott Smith, B.H., 2008. The Pimenta Bueno kimberlite field, Rondônia,Brazil: tuffisitic kimberlite and transitional textures. Journal of Volcanology andGeothermal Research 174 (1–3), 81–89.

Maunula, T., 2006. Technical report on the Pimenta Bueno diamond project, Brazil.Report to Vaaldiam Resources Ltd. www.vaaldiam.com/mining/pimenta-bueno.html.

McCandless, T.E., Gurney, J.J., 1989. Sodium in garnet and potassium in clinopyroxene:criteria for classifying mantle eclogites. In: Ross, J. (Ed.), Kimberlites and RelatedRocks: Their Mantle/crust Setting, Diamonds, and Diamond Exploration SpecialPublication — Geological Society of Australia, vol. 2, pp. 827–832.

851L. Hunt et al. / Lithos 112S (2009) 843–851

Nimis, P., Taylor, W.R., 2000. Single clinopyroxene thermobarometry for garnetperidotites. Part I. Calibration and testing of a Cr-in-Cpx barometer and anenstatite-in-Cpx thermometer. Contributions to Mineralogy and Petrology 139 (5),541–554.

Pearson, D.G., 1999. Evolution of cratonic lithospheric mantle: an isotopic perspective.In: Fei, Y., Bertka, C.M., Mysen, B.O. (Eds.), Mantle Petrology Field Observations andHigh-pressure Experimentation: A Tribute to Francis, R. (Joe) Boyd, TheGeochemical Society Special Publications, Houston, TX, pp. 57–78.

Ramsay, R.R., 1992. Geochemistry of diamond indicator minerals. PhD Thesis, Universityof Western Australia, Perth.

Read, G. Grütter H., Winter, S., Luckman, N., Gaunt, F., Thomsen, F., 2004. Stratigraphicrelations, kimberlite emplacement and lithospheric thermal evolution, QuiricóBasin, Minas Gerais State, Brazil. Lithos 77 (1–4), 803–818.

Shirey, S.B., Richardson, S.H., Harris, J.W., 2004. Integrated models of diamondformation and craton evolution. Lithos 77 (1–4), 923–944.

Stachel, T., 2007. Diamond. Mineralogical Association of Canada Short Course Series 37,1–22.

Stachel, T., Harris, J.W., 2008. The origin of cratonic diamonds — constraints frommineral inclusions. Ore Geology Reviews 34 (1–2), 5–32.

Stachel, T., Viljoen, K.S., Brey, G., Harris, J.W., 1998. Metasomatic processes in lherzoliticand harzburgitic domains of diamondiferous lithospheric mantle: REE in garnetsfrom xenoliths and inclusions in diamonds. Earth and Planetary Science Letters 159(1–2), 1–12.

Svisero, D.P., Haralyi, N.L.E., 1983. The kimberlitic province of the Upper Paranaiba,Minas-Gerais. Anais Da Academia Brasileira De Ciencias 55 (1), 137–138.

Tassinari, C.C.G., Macambira, M.J.B., 1999. Geochronological provinces of the AmazonianCraton. Episodes 22, 174–182.

Taylor, W.R., Milledge, H.J., 1995. Nitrogen aggregation character, thermal history andstable isotope composition of some xenolith-derived diamonds from Roberts Victorand Finsch. Sixth International Kimberlite Conference (Novosibirsk), ExtendedAbstracts, pp. 620–622.

Taylor, W.R., Jaques, A.L., Ridd, M., 1990. Nitrogen-defect aggregation characteristics ofsome Australasian diamonds; time–temperature constraints on the source regionsof pipe and alluvial diamonds. American Mineralogist 75 (11–12), 1290–1310.

Taylor, W.R., Gurney, J.J., Milledge, H.J., 1995. Nitrogen aggregation and cathodolumi-nescence characteristics of diamonds from the Point Lake kimberlite pipe, SlaveProvince, NWT, Canada. Sixth International Kimberlite Conference (Novosibirsk),Extended Abstracts, pp. 614–615.

Tohver, E., van der Pluijm, B.A., Mezger, K., Scandolara, J.E., Essene, E.J., 2005. Two stagetectonic history of the SWAmazon craton in the late Mesoproterozoic: identifying acryptic suture zone. Precambrian Research 137 (1–2), 35–59.

Walter, M.J., 1998. Melting of garnet peridotite and the origin of komatiite and depletedlithosphere. Journal of Petrology 39 (1), 29–60.

White, R., Mckenzie, D., 1989. Magmatism at rift zones — the generation of volcaniccontinental margins and flood basalts. Journal of Geophysical Research 94 (B6),7685–7729.

Woods, G.S., 1986. Platelets and the infrared-absorption of type-Ia diamonds.Proceedings of the Royal Society of London. Series A 407 (1832), 219–238.

Woods, G.S., Collins, A.T., 1983. Infrared-absorption spectra of hydrogen complexes intype-I diamonds. Journal of Physics and Chemistry of Solids 44 (5), 471–475.