volatile composition of microinclusions in diamonds from ...cartigny/2009-gca-burgessetal.pdf ·...

Author's personal copy

Volatile composition of microinclusions in diamonds fromthe Panda kimberlite, Canada: Implications for chemical

and isotopic heterogeneity in the mantle

Ray Burgess a,*, Pierre Cartigny b, Darrell Harrison a, Emily Hobson a, Jeff Harris c

a School of Earth, Atmospheric and Environmental Sciences, University of Manchester, Oxford Rd, Manchester M13 9PL, UKb IPGP et Universite Paris 7, Laboratoire de Geochimie des Isotopes Stables, Institut de Physique du Globe, UMR CNRS 7047,

4 Place Jussieu, T54-64 E1, 75251 Paris cedex 05, Francec Department of Geographical and Earth Sciences, University of Glasgow, Gregory Building, Glasgow G12 8QQ, Scotland, UK

Received 5 September 2008; accepted in revised form 26 December 2008; available online 12 January 2009

Abstract

In order to better investigate the compositions and the origins of fluids associated with diamond growth, we have carried-out combined noble gas (He and Ar), C and N isotope, K, Ca and halogen (Cl, Br, I) determinations on fragments ofindividual microinclusion-bearing diamonds from the Panda kimberlite, North West Territories, Canada. The fluid concen-trations of halogens and noble gases in Panda diamonds are enriched by several orders of magnitude over typical upper man-tle abundances. However, noble gas, C and N isotopic ratios (3He/4He = 4–6 Ra, 40Ar/36Ar = 20,000–30,000, d13C = �4.5&

to �6.9& and d15N = �1.2& to �8.8&) are within the worldwide range determined for fibrous diamonds and similar to themid ocean ridge basalt (MORB) source value. The high 36Ar content of the diamonds (>1 � 10�9 cm3/g) is at least an order ofmagnitude higher than any previously reported mantle sample and enables the 36Ar content of the subcontinental lithosphericmantle to be estimated at �0.6 � 10�12 cm3/g, again similar to estimates for the MORB source. Three fluid types distin-guished on the basis of Ca–K–Cl compositions are consistent with carbonatitic, silicic and saline end-members identifiedin previous studies of diamonds from worldwide sources. These fluid end-members also have distinct halogen ratios (Br/Cland I/Cl). The role of subducted seawater-derived halogens, originally invoked to explain some of the halogen ratio variationsin diamonds, is not considered an essential component in the formation of the fluids. In contrast, it is considered that largehalogen fractionation of a primitive mantle ratio occurs during fluid–melt partitioning in forming silicic fluids, and duringseparation of an immiscible saline fluid.� 2009 Elsevier Ltd. All rights reserved.

1. INTRODUCTION

Microinclusions in fibrous diamonds provide uniquesamples of deep fluids from the mantle. Investigation ofthe isotopic and chemical composition of these fluids is use-ful for several reasons. First, they improve our understand-ing of the processes of diamond growth and help elucidatethe composition of the upper mantle environment in which

diamonds originate (Navon et al., 1988; Navon, 1999; Sch-rauder and Navon, 1994; Izraeli et al., 2001; Klein-BenDa-vid et al., 2004, 2006, 2007; Zedgenizov et al., 2004, 2007;Tomlinson et al., 2006). Secondly, detailed investigationof the microinclusion fluid compositions in diamonds thatformed during different epochs and in different geographi-cal regions of the world, provides an opportunity to tracethe time-resolved, global evolution of volatile elements inthe mantle (Johnson et al., 2000; Burgess et al., 2002;Klein-BenDavid et al., 2007). Finally, fluids trapped in dia-monds are potentially the best samples to assess to whatextent mantle processes (e.g. metasomatism) are able to

0016-7037/$ - see front matter � 2009 Elsevier Ltd. All rights reserved.

doi:10.1016/j.gca.2008.12.025

* Corresponding author. Fax: +44 161 275 3947.E-mail address: [email protected] (R. Burgess).

www.elsevier.com/locate/gca

Available online at www.sciencedirect.com

Geochimica et Cosmochimica Acta 73 (2009) 1779–1794


produce large chemical (and time-integrated isotopic)variations and to test whether volatiles derived from thesubducted crust are able to penetrate to the base of conti-nental cratons (e.g. Richardson et al., 2001; Cartigny,2005).

Previous investigations have revealed that the isotopiccomposition of volatile elements such as C, N, He andAr, in microinclusion-bearing diamonds from Africa(Jwaneng mine, Botswana and from the DemocraticRepublic of Congo), Siberia (Aikhal mine, Yakutia, Rus-sia) and Canada (Ekati mine, Lac de Gras, North West Ter-ritories, Canada) are similar to present-day upper mantlevalues (Boyd et al., 1992; Cartigny et al., 2003; Johnsonet al., 2000; Burgess et al., 2002; Wada and Matsuda,1998). In contrast, the chemical composition of the fluidsin diamonds are highly variable with three end-membersidentified (summarised in Klein-BenDavid et al., 2007) asfollows: (1) carbonatitic fluid rich in carbonate, K, Ca, Feand Mg; (2) silicic fluid containing water, Al and K; and(3) saline fluid containing high concentrations of Cl andK. Mixtures of carbonatitic and silicic fluids have beenfound in microinclusions in diamonds from southern Afri-ca, Brazil and Siberia (Navon et al., 1988; Schrauder andNavon, 1994; Shiryaev et al., 2005), and in Canadian dia-monds from the Diavik (Klein-BenDavid et al., 2006,2007) and Panda mines (Tomlinson et al., 2006). Saline flu-ids containing high concentrations of Cl (up to 40 wt.%)and K were first documented in cloudy diamonds fromKoffiefontein, South Africa (Izraeli et al., 2001). Subse-quently, Klein-BenDavid et al. (2004, 2006, 2007) andTomlinson et al. (2006) have shown that saline fluids arepresent in microinclusion-bearing diamonds from Diavikand Panda, Canada, where they occur mixed with carbon-atitic fluids. The absence of co-existing saline and silicicfluid microinclusions suggests that these two fluid typesmay be immiscible during diamond growth (Perchuket al., 2002; Navon et al., 2003; Klein-BenDavid et al.,2004, 2006, 2007), which is supported by recent experimen-tal work (Safonov et al., 2007). The saline fluid is associatedwith both peridotitic and eclogitic mineral inclusions in dia-monds from Koffiefontein (Izraeli et al., 2001) and Panda(Tomlinson et al., 2006), and with peridotitic minerals inDiavik diamonds (Klein-BenDavid et al., 2004, 2007).Izraeli et al. (2001) suggested the brine in Koffiefontein dia-monds may have formed by separation of immiscible fluid,or possibly was derived from subducted seawater.

Previously, we have reported high halogen abundancesand variable halogen ratios (Br/Cl and I/Cl) in coated dia-monds from four kimberlites from the Ekati property,North West Territories, Canada (Fox, Grizzly, Koala andLeslie kimberlites: Johnson et al., 2000; Burgess et al.,2002). The highest Br/Cl and I/Cl values measured in someof these diamonds are well above values for other mantlesamples, including Siberian and African diamonds, andare similar to the highest values reported for crustal fluids(Johnson et al., 2000). The origin of the halogen fraction-ation is uncertain, and may be related to crystallisation ofa Cl-bearing phase such as apatite leaving a fluid enrichedin Br and I, recycling of crustal halogens having high Br/Cl and I/Cl values, or melt–fluid partitioning with halogen

fractionation controlled by differences in the ionic radii ofhalide ions (Johnson et al., 2000).

In this study we combine analyses of K, Ca and halogens(Cl, Br and I) determined using the Ar–Ar technique, noblegas (He and Ar), C and N isotopes, and infra-red spectros-copy, to understand the relationships between different fluidtypes in diamonds from the Panda kimberlite, Canada. Thisenables better constraints to be placed on likely scenariosfor generating the large range of halogen ratios observedfor different fluid compositions.

2. EXPERIMENTAL METHODS

2.1. Panda diamond samples

Twelve diamond samples from the Panda kimberlitehave been studied. Panda is a group I kimberlite, and isone of several diamondiferous kimberlites forming the Eka-ti property, North West Territories. The Panda kimberlitehas a phlogopite Rb–Sr eruption age of 53.2 ± 0.3 Ma(2r; Creaser et al., 2004). An overview of Panda kimberlitegeology is given by Nowicki et al. (2004).

The diamond samples analysed in this study were eitherwhole samples or fragments weighing between 0.02 and0.05 g of light or dark grey coated diamonds (fibrous ‘‘coat-s” overgrowing clear octahedral cores) with coat thick-nesses between 0.2 and 0.7 mm (Fig. 1). Only one sample,PEK013/9 differs in having a frosted appearance on the sur-face possibly resulting from partial resorption by the hostkimberlite magma (Robinson et al., 1989). The coats con-tain fluid-bearing microinclusions 61 lm in size (Schrauderand Navon, 1994; Tomlinson et al., 2006), that are toosmall to analyse by conventional fluid inclusion techniques,and even using micro-scale techniques (e.g. electron and ionprobes) it is difficult to avoid the complicating effects ofoverlapping the analysing beam with host diamond matrixor nearby inclusions. Tomlinson et al. (2006) have reportedthe chemical compositions of co-existing silicate macroin-clusions and fluid microinclusions in the fibrous coats ofeight coated diamonds from the Panda kimberlite. Theirstudy revealed that fibrous diamonds with either eclogiticor peridotitic mineral inclusions also contain fluid microin-clusions dominated by H2O, carbonate and KCl. They con-clude that fibrous diamonds grow in the same parageneticenvironments as octahedral diamonds and that the fluidcomponents, possibly originally derived from a subductedslab, were chemically modified during mantle metasoma-tism. The mineral parageneses of the Panda diamonds inthis study were not determined.

Westerlund et al. (2006) obtained a Re–Os isochron ageof 3.52 ± 0.17 Ga for eleven sulfide inclusions in peridotiticoctahedral diamonds from the Panda kimberlite. The age ofthe diamond coat growth is unknown but must have oc-curred after the octahedral diamonds formed. Insight intothe relative timing of coat growth can be obtained fromthe aggregation state of nitrogen in diamond. During dia-mond growth, nitrogen is trapped by substitution for car-bon atoms (Type Ib diamond). The nitrogen thenmigrates to form pairs of atoms (Type IaA) and eventually,

1780 R. Burgess et al. / Geochimica et Cosmochimica Acta 73 (2009) 1779–1794


clusters of four atoms and a vacancy (Type IaB) and plate-lets. The aggregation process is diffusion controlled and thepercentage of the nitrogen B-centres (i.e. clusters) dependsupon initial nitrogen concentration and the integratedtime–temperature history of the diamond (Taylor et al.,1996). Nitrogen aggregation states are expressed as the pro-portion of the B-defect relative to the total nitrogen content(%B). Available data have shown that fibrous diamonds arecharacterised by moderately aggregated nitrogen withoutany development of B-centers (Boyd et al., 1987; Navonet al., 1988). Tomlinson et al. (2006) found only nitrogenin A centres in Panda coats, compared with up to 33% Bnitrogen (absorptions at 1332 and 1175 cm�1) in the coresas well as some platelets (�1370 cm�1). The low aggrega-tion state of nitrogen is a global feature of coated diamondand is usually interpreted to indicate either unusually lowtemperature storage in the mantle, or a relatively shortmantle residence time (e.g. Taylor et al 1996).

Prior to experimental work, Panda diamonds werecleaned in 10% HF acid at room temperature for 48 h to re-move any adhering silicate material, followed by ultrasoniccleaning in acetone and deionised water. A visual inspectionunder a binocular microscope was made to ensure completeremoval of any silicates. Each diamond was then brokeninto a number of fragments, each weighing a few milli-grams, for use in different experiments. Between 2 and 10fragments from each diamond were analysed for noblegases (Ar and He), or irradiated for Ar–Ar isotopic (Ar,K, Ca, Cl, Br, I, U) experiments. Infra-red (IR) spectrawere obtained for most of the diamonds studied and Cand N isotopes were determined for all diamonds. No at-tempt was made to separate interior (core) and exterior (fi-brous coat) portions of the diamond for noble gas workbecause the concentrations in core regions are known tobe relatively low (Turner et al., 1990). Prior to C and N iso-tope determinations and IR spectroscopy, each diamondfragment was examined under a binocular microscope toverify that it was predominantly composed of coat material.Noble gas and Ar–Ar measurements were carried-out at the

University of Manchester, and C and N isotope determina-tions and IR spectroscopy at the University of Paris.

2.2. Fourier transform infra-red spectroscopy, C and N stable

isotope analyses

Diamonds were characterised using Fourier TransformInfra-red (FTIR) micro-spectroscopy to determine nitrogencontent and aggregation state and fluid characteristics. In-fra-red absorption spectra were collected in the range of4000–600 cm�1 at a resolution of 4 cm�1 using a beamdiameter of 100 lm. Measurements were made with a Nico-let 510 attached to a NicPlan microscope using a KBr beamsplitter, an EverGlo source and an MCTA detector.

FTIR spectra were obtained on broken (i.e. unpolished)diamond fragments of moderate size (<2/3 mm3), and be-cause of the irregular shape and small sample size, it wasnot possible to investigate the spatial variation of nitrogencontent or fluid characteristics within each sample. Samplethickness was estimated assuming an 11.94 absorbance at1992 cm�1. Nitrogen contents were measured from the base-line-corrected absorbance at 1282 cm�1 using coefficients of16.5 and 79.4 at. ppm cm�1 for A- and B-centres (Boyd et al.,1994, 1995a). Water and carbonate were determined from thebaseline-corrected absorbance of the O–H stretching(e3420 = 80 L/mol cm�1) and the m3 stretching band of car-bonate using an absorption coefficient (following subtractionfor diamond absorption) of 235 L/mol cm�1 at 1430 cm�1,respectively (see details in Navon et al., 1988; Tomlinson etal., 2006). The presence and nature of silicates can be

investigated from absorption bands at �1000 cm�1

(Tomlinson et al., 2006). Errors in the nitrogen content andaggregation state, water and carbonate contents are esti-mated to be better than 20% and 5% (Boyd et al., 1987),±25 and ±40 ppm (Navon et al., 1988), respectively.

Nitrogen isotopes, nitrogen contents and carbon isotopedeterminations were subsequently made using proceduresdescribed in Boyd et al. (1995b). After online diamondcombustion in an oxygen atmosphere at a temperature of

Fig. 1. Photographs of Panda coated diamond analysed in this study. The examples shown are the most intact diamonds investigated, othersamples were broken fragments containing mixed core and coat material.

Volatile composition of fluid microinclusions in diamonds 1781


1100 �C, nitrogen was separated from carbon dioxide andany nitrogen oxides reduced to N2 using a CaO/Cu mixtureat 600 �C. Nitrogen concentrations were measured with acapacitance manometer with a precision better than 5%and the nitrogen isotopic composition was analyzed witha home-made triple collector static vacuum mass spectrom-eter directly connected to the extraction line. Carbon diox-ide produced by the combustion was analyzed with aconventional dual-inlet gas source mass spectrometer. Iso-topic compositions are expressed using delta notation, de-fined as d13C = [(13C/12Csample)/(

13C/12CPDB) � 1] � 1000and d15N = [(15N/14Nsample)/(

15N/14Nair) � 1] � 1000. Inaddition to blank determination, 40Ar (m/z = 40) was alsomonitored in the mass spectrometer as an indicator of po-tential atmospheric contamination for both sample andblank. For all samples, the blank contribution was below3.5 � 10�9 g N with a corresponding d15N value between�10& and 0&. The accuracy of the measurement, estab-lished on the basis of standard analyses, is better than±0.1& and ±0.5& (2r) for d13C and d15N, respectively(Boyd et al., 1995b) However, the errors reported in Table1 can be larger and take into account a maximum blankcontribution with 15N between �10& and 10&.

2.3. Noble gas analyses

Argon isotopic analysis of neutron-irradiated diamondswas used to determine Ar, K (via 39ArK), Cl (38ArCl), Ca(37ArCa),, Br (80KrBr), I (128XeI) and U (134XeU) concentra-tions in each fragment. Neutron irradiation of samples wascarried-out in position B2W of the SAFARI-1 reactor,NECSA, Pelindaba, South Africa with a fast neutron fluxof 1 � 1018 n/cm2 as determined from Hb3gr flux monitorsincluded in the irradiation. Experimental procedures weresimilar to those described previously in Johnson et al.(2000) and Burgess et al. (2002). Argon was extracted in aresistance furnace using two temperature steps each of30 min duration. A low temperature step at 800 �C wasused to remove adsorbed atmospheric noble gases fromthe samples, followed by a high temperature step at

2150 �C to graphitise the diamond and liberate noble gasesfrom microinclusions.

Isotopic measurements were made using the MS1 massspectrometer which is equipped with a Faraday detectorfor Ar isotope measurements and a channel electron multi-plier for lower abundance Kr and Xe isotope detection.Average furnace blanks at 2150 �C were approximately1.5 � 10�9 cm3 STP 40Ar, 0.2 � 10�12 cm3 STP 84Kr and0.4 � 10�12 cm3 STP 132Xe. Blank corrections are minorfor most noble gas isotopes, the exception is 36Ar wheresamples released between 2 and 8 times blank levels. Argonisotope measurements were corrected for blanks, mass dis-crimination, 37Ar decay and neutron interferences. The lat-ter were determined from CaF2 and K2SO4 to give thefollowing values: (40Ar/39Ar)K = 0.044 ± 0.023; (38Ar/39Ar)K = 0.01243 ± 0.00002; (39Ar/37Ar)Ca = 0.000745 ±0.000003; and (36Ar/37Ar)Ca = 0.000269 ± 0.000007.

Bromine and iodine concentrations were calculated from80KrBr and 128XeI, respectively (Johnson et al., 2000). A com-bined blank and air correction has been applied to 80Kr and128Xe from measured 84Kr and 129Xe, respectively, andassuming air ratios. 84Kr was initially corrected for a small(<1%) contribution from neutron-induced fission of U basedupon 134XeU release and assuming 134XeU/84KrU = 0.02356.With the exception of PEK013/9, blank correctionsamounted to <1% and 2–5% for 80Kr and 128Xe, respectively.For PEK013/9 the blank corrections were higher at 1–3%and 13–44% for 80Kr and 128Xe, respectively. Thermal andresonant neutron production of 128XeI (and 80KrBr) wasmonitored using the Shallowater meteorite which has an ini-tial 127I/129I ratio of 1.125 � 10�4. The main uncertainties inBr and I contents derive from the measurement precision ofm/z 80 and 128 ion beams in the mass spectrometer.

It was not possible to determine Ca (via 37ArCa,

t½ = 35 days) contents of three Panda diamonds (samplesPEK013/5, PEK013/10 and PEK013/11) because they wereanalysed more than 6 months following irradiation, afterwhich 37ArCa had decayed to negligible levels.

Helium and Ar isotopes were extracted from unirradi-ated diamond fragments using a low blank filament furnace

Table 1Carbon isotope, nitrogen and IR data for Panda diamonds (2r error).

Sample Weight(mg)

d13C(&)

d15N(&)

±d15N(&)

N (ppm)combustion

N (ppm)FTIR

%B H2O(ppm)

CO2

(ppm)100 H2O/(H2O + CO2)(molar)

Other mainabsorptions

PEK013/1

0.4249 �6.93 �3.1 +0.6 1136 1193 0 1854 802 85 1049, 1014, 667, 861,837, 714, 687�0.6

PEK013/2

0.6506 �5.34 �8.8 +0.5 1352 n.d. n.d.�0.6

PEK013/4

0.3548 �5.10 �2.3 +0.5 1899 808 13.7�0.6

PEK013/5

0.2348 �5.20 �2.3 +0.6 775 712 27.8�0.7

PEK013/9

0.5226 �5.52 �4.1 +0.6 418 n.d. n.d.�0.7

PEK013/10

0.5682 �6.15 �1.2 +0.5 1872 1347 0 959 621 79 992, 886, 961, 656,844�0.5

PEK013/11

0.3961 �4.53 �1.7 +0.6 1222 1283 0 1886 1262 78 970, 994, 878�0.6



system. This consisted of five tungsten ribbon filamentseach welded to a pair of wires mounted on an 11-pin elec-trical feedthrough flange. Prior to sample loading a currentwas passed through each filament, heating it to hightemperature (>1800 �C) in order to degas any adsorbedatmospheric He and Ar. Diamond fragments were loadedon to the filaments by inserting them into a ‘‘V” shapeformed in the filament and enclosed by heat shields. Sam-ples were heated in single steps of 5 min duration in orderto release noble gases. Purification was achieved using ahot (450 �C) Zr–Al getter (SAES GP50) for 10 min. Argonwas then separated from He at liquid nitrogen temperature(�196 �C) on activated charcoal. He and Ar isotopes wereanalysed in a MAP 215 mass spectrometer using proceduresdescribed in Harrison et al. (2003). High temperatureblanks were determined on an unloaded filament to be(3.6 ± 0.7) � 10�13 cm3 STP 4He and (1.4 ± 0.6) � 10�10

cm3 STP 40Ar. These amounts correspond to <0.1% ofthe amount of He and Ar released from a diamond sample,thus only a minor blank correction has been applied to thedata.

All noble gas and Ar–Ar data are reported at a one stan-dard deviation level of uncertainty unless otherwise stated.Concentrations of He, Ar and halogens were obtained bycombining the amounts released from fragments of eachdiamond and are reported relative to the bulk diamond.Noble gas/halogen and halogen ratios are given as molarvalues.

3. RESULTS AND DISCUSSION

3.1. Nitrogen aggregation, FTIR spectra and carbon and

nitrogen isotopes

Five of the seven investigated samples have type IaAaggregation states which, together with high levels of nitro-gen (average �1360 ppm), are typical of fibrous diamonds(Table 1: Boyd et al., 1987, 1992; Cartigny et al., 2003;

Tomlinson et al., 2006). However, two samples (PEK013/4 and PEK013/5) show significant B-percentage values(up to 28%, Table 1) from which we infer that despite theresults of our visual inspection some core material was pres-ent (e.g. Boyd et al., 1987; Navon et al., 1988; Tomlinsonet al., 2006).

Sufficient material was available from three Panda dia-monds for FTIR spectroscopy (Table 1). These diamondshave water contents of between 78% and 85%, higher thanthe range of 55–68% reported previously by Tomlinsonet al. (2006). Significant peaks in the FTIR spectra ofPEK013/1 and PEK013/11 are from silicate-related absorp-tions. The saline fluid composition described by Klein-Ben-David et al. (2007) is strongly depleted in silica. Thus, thesignificant silicate peak identified in the course of the pres-ent FTIR investigation are likely to relate to silicate inclu-sions, as reported previously by Tomlinson et al. (2006),and also inferred later in Section 3.3 from their apparentCa-enrichment. Samples are characterised by additionalabsorptions bands (see Table 1) and we did not make anyattempt to determine possible phases. Residual internal dia-mond pressure results in a shift of characteristic absorptionbands relative to 1 bar reference spectra and, given thathigh pressure spectral data are still lacking, the identifica-tion of any phase would be speculative.

Carbon and nitrogen isotope compositions show ratherrestricted ranges, with d13C and d15N between �6.9& to�4.5& and �8.8& to �1.2&, respectively (Fig. 2). Carbonand nitrogen isotopic compositions do not appear to corre-late with fluid composition, or any halogen and noble gasisotopic variations. d13C–d15N values are similar to the glo-bal range for fibrous diamond: d13C = �8& to �5& andd15N = �12& to �0& (Cartigny, 2005). The two samplesanalysed that were inferred to contain some non-fibrousdiamond cores (PEK013/4 and PEK013/5) have carbonand nitrogen isotope ratios well within the range for Pandadiamonds (Table 1) indicating that the core material doesnot have a highly unusual composition. In detail however,

δ13 C

(‰)

world-wide range for peridotiticand eclogitic diamonds

range for fibrous & cubic diamonds-10

-20

-30-12 -9 -6 -3 -0 +3 +6 +9 +12 +15

δ15N (‰)

0

Panda

Fig. 2. Carbon and N isotope composition Panda diamonds. Most samples lie within the range determined for fibrous and cubic diamondsfrom worldwide localities (Cartigny, 2005).



it is clear that most (6 out of 7) d15N values fall on the en-riched 15N-side of the fibrous diamond range (Fig. 2). It ispossible that the higher d15N content reflects slight differ-ences in their mantle source compared with Siberian andAfrican cratonic lithospheres, although confirmation of thiswill require further detailed study.

A comparison between the N content determined byFTIR and bulk combustion (n = 5) shows good agreement(within 10%) for three analyses (Table 1). However, a com-parison of these two techniques for the remaining two sam-ples shows differences in N concentrations of up to a factor2.4 (PEK013/4). Large differences in nitrogen contents ofgem quality diamonds are commonly found when compar-ing bulk and FTIR (which is more spatially resolved) tech-niques (e.g. Cartigny et al., 2001), while the cause isunknown one reason could be linked to the difference inspatial resolution of these two techniques especially if thereis significant nitrogen zoning within the diamonds.

3.2. Ar–Kr–Xe analysis of irradiated diamonds

Argon, halogens, K, Ca and U data, as determined bynoble gas analyses of neutron-irradiated diamonds, are pre-sented in Table 2. Halogen data are discussed separately inSection 3.4. Results for Panda diamonds are similar tothose obtained previously from diamonds at the nearbyFox, Grizzly, Koala and Leslie kimberlites (Johnsonet al., 2000; Burgess et al., 2002). 40Ar and K are poorly cor-related in Panda diamonds with all apparent ages >4.5 Gaindicating the ubiquitous presence of excess 40Ar*

(40Ar* = 40Ar � 295.5 � 36Ar). In situ contributions ofradiogenic 40Ar based upon K concentrations are insignifi-cant (<0.1%) assuming a minimum age for coat growth of53 Ma. Even using an upper limit for coat growth givenby the Re–Os age for peridotitic diamonds of 3.52 Ga, givesa maximum of only 5% radiogenic 40Ar in PEK013/9 and<1% in all other diamonds.

40Ar*/Cl values are between 444 and 887 � 10�6 (Fig. 3;Table 2) which is about a factor of two lower than obtainedpreviously for Canadian diamonds, but is similar to coateddiamonds from Siberian and African sources (Johnsonet al., 2000; Burgess et al., 2002). The 40Ar–Cl correlationindicates the presence of a hydrous fluid-bearing phase con-taining noble gases and halogens. The range of K/Cl valuesdetermined for Panda diamonds is between 0.5 and 1.5which is comparable to the range of 0.14–1.44 for inclusionsanalysed by electron microprobe (Tomlinson et al., 2006).The relatively limited variation in K/Cl of Panda diamondsmeans that it is not possible to use 40Ar/K/Cl correlationsto constrain the age of coat growth as was done previouslyfor Siberian diamonds (Burgess et al., 2002).

3.3. Relationship to previously identified fluid types in

diamonds

Determination of Ca, K and Cl (Table 2) means it ispossible to relate fluid compositions of Panda diamondsto the carbonatitic, silicic and saline fluid end-members pre-viously identified in Canadian, African, Brazilian and Sibe-rian diamonds (Schrauder and Navon, 1994; Izraeli et al.,

2001; Klein-BenDavid et al., 2004; Shiryaev et al., 2005;;Klein-BenDavid et al., 2007). Fig. 4 shows K–Ca–Cl datafor individual fragments and the weighted average for eachPanda diamond analysed in this study, as well as publisheddata for other Canadian, African and Siberian diamonds.With the exception of sample PEK013/9, the data fromPanda diamonds form an array between the saline fluidcomposition found in most Canadian and some African(Koffiefontein) diamonds, and a Ca-rich component(Fig. 4). PEK013/9 is the only diamond in this study tolie along the carbonatitic–silicic fluid trend, as defined bytwo other diamonds from Canada (Diavik), and diamondsfrom African (Jwaneng) and Siberian (Aikhal) sources.

Fragments of individual Panda diamonds show consid-erable chemical variation in Fig. 4. Some of the intra-sam-ple variation may result from chemical zoning as proposedfor some other elements in Panda and Diavik diamonds(Klein-BenDavid et al., 2004, 2007; Tomlinson et al.,2006). However, it is also likely that some variation is re-lated to Ca-rich silicate mineral inclusions whose presenceis indicated by FTIR. Previously, Tomlinson et al. (2006)reported abundant silicate inclusions (2–20 lm size) of gar-net, olivine, clinopyroxene and orthopyroxene in the coatsof Panda diamonds. Using the high spatial resolution ofthe electron microprobe employed during their study, thesemineral inclusions could be analysed separately from fluidmicroinclusions. However, the method used here of steppedheating is a bulk extraction method that releases noblegases simultaneously from all types of inclusions in thediamond.

Chlorine contents of silicate inclusions have not been re-ported, however they are expected to be low. Potassium is aminor element in pyroxene (clinopyroxene K2O 60.5 wt.%and orthopyroxene <0.3 wt.%; Tomlinson et al., 2006).Therefore, the K and Cl concentrations of diamonds willbe dominated by fluid microinclusions (Tomlinson et al.,2006). In contrast, the Ca contents of clinopyroxene andgarnet are relatively high between 5 and 22 wt.% CaO(Tomlinson et al., 2006), hence these minerals will have highCa/Cl and Ca/K ratios. Fig. 4 shows the range of composi-tions of silicate inclusions based upon data from Tomlinsonet al. (2006).

During stepped heating of diamonds, the release ofCa-derived 37Ar from any silicate mineral inclusions willskew data towards the Ca apex in Fig. 4. The extent towhich individual fragments are displaced toward the Caapex will be determined mainly by the composition andproportions of silicate inclusions. The trend toward highCa/K shown by the majority of Panda diamonds (Fig. 4)suggests that garnet and clinopyroxene are the mainsource of calcium. This is consistent with the results ofTomlinson et al. (2006) who reported only four orthopy-roxene inclusions that were present in a single Panda dia-mond. Although stepped heating is a bulk analysistechnique, it does not enable partial resolution of thefluid end-member chemical composition. Thus, it is rea-sonable to conclude that saline fluids are present in allof the Panda diamonds analysed except PEK013/9. Inthe latter diamond, it is more likely that a mixture of car-bonatitic and silicic fluids is present.



Table 2Panda diamond Ar–Ar data. Errors are 1r; nd = not determinable i.e error >30%.

Sample Wt (mg) Cl (ppm) Br (ppb) I (ppb) K (ppm) Ca (ppm) U (ppb) Br/Cl � 10�3 M I/Cl � 10�6 M K/Cl M 40Ar*/Cl � 10�6 M

PEK013/

1a 3.9 165 ± 5 515 ± 5 5.6 ± 0.4 182 ± 16 914 ± 243 107 ± 6 1.38 ± 0.04 9.4 ± 0.7 1.00 ± 0.09 696 ± 201b 3.1 216 ± 3 845 ± 7 8.7 ± 0.9 211 ± 25 388 ± 59 141 ± 8 1.73 ± 0.03 11.3 ± 1.2 0.88 ± 0.11 753 ± 11Total 1 7.0 188 ± 3 662 ± 4 7.0 ± 0.5 195 ± 14 680 ± 137 122 ± 5 1.56 ± 0.03 10.3 ± 0.7 0.94 ± 0.07 725 ± 12

2a 3.4 74 ± 1 252 ± 1 2.3 ± 0.3 61 ± 1 119 ± 25 22 ± 4 1.52 ± 0.02 8.7 ± 1.0 0.74 ± 0.01 660 ± 92b 2.8 130 ± 2 503 ± 4 5.8 ± 0.3 146 ± 5 341 ± 91 181 ± 12 1.71 ± 0.02 12.4 ± 0.7 1.01 ± 0.03 655 ± 8Total 2 6.2 99 ± 1 367 ± 2 3.9 ± 0.2 99 ± 2 220 ± 43 94 ± 6 1.63 ± 0.02 10.9 ± 0.6 0.90 ± 0.02 657 ± 6

3a 3.2 216 ± 8 2088 ± 12 28.9 ± 1.5 279 ± 8 537 ± 189 431 ± 22 4.27 ± 0.16 37.2 ± 2.4 1.17 ± 0.05 765 ± 283b 4.3 74 ± 1 353 ± 2 5.2 ± 0.3 94 ± 4 87 ± 37 77 ± 4 2.10 ± 0.03 19.3 ± 1.0 1.14 ± 0.05 835 ± 9Total 3 7.5 135 ± 3 662 ± 4 7.0 ± 0.5 173 ± 4 279 ± 83 228 ± 10 3.59 ± 0.09 31.6 ± 1.6 1.16 ± 0.04 787 ± 20

4a 2.1 129 ± 1 464 ± 3 3.1 ± 0.6 151 ± 7 168 ± 137 50 ± 1 1.59 ± 0.02 6.7 ± 1.3 1.06 ± 0.05 787 ± 94b 2.2 98 ± 1 316 ± 2 4.0 ± 0.2 124 ± 4 nd 61 ± 8 1.42 ± 0.02 11.2 ± 0.5 1.14 ± 0.03 771 ± 94c 4.7 208 ± 3 893 ± 5 9.8 ± 0.4 127 ± 10 244 ± 49 37 ± 20 1.90 ± 0.03 13.1 ± 0.5 0.55 ± 0.05 780 ± 104d 2.4 121 ± 1 516 ± 2 7.3 ± 0.5 87 ± 3 144 ± 10 33 ± 29 1.88 ± 0.02 16.7 ± 0.1 0.65 ± 0.03 770 ± 94e 1.6 17 ± 1 54 ± 1 1.1 ± 0.1 24 ± 7 215 ± 54 68 ± 3 1.41 ± 0.04 17.4 ± 0.2 1.28 ± 0.39 723 ± 23Total 4 12.9 135 ± 1 533 ± 2 6.2 ± 0.2 110 ± 4 163 ± 33 46 ± 9 1.79 ± 0.02 12.6 ± 0.4 0.73 ± 0.03 778 ± 6

Total 5 18.3 136 ± 11 1092 ± 4 14.0 ± 0.7 80 ± 11 nd nd 1.92 ± 0.16 16.2 ± 1.5 0.54 ± 0.08 769 ± 63

6a 4.4 104 ± 1 410 ± 3 4.9 ± 0.2 107 ± 2 200 ± 91 60 ± 5 1.74 ± 0.02 13.0 ± 0.7 0.93 ± 0.02 783 ± 96b 3.3 171 ± 2 787 ± 3 8.1 ± 0.2 179 ± 10 6 0 ± 226 90 ± 5 2.04 ± 0.03 13.3 ± 0.4 0.95 ± 0.05 800 ± 11Total 6 7.7 133 ± 1 571 ± 2 6.3 ± 0.2 138 ± 4 140 ± 110 73 ± 4 1.91 ± 0.02 13.2 ± 0.4 0.94 ± 0.03 793 ± 7

7a 2.6 76 ± 1 412 ± 2 5.4 ± 0.1 128 ± 5 119 ± 75 66 ± 4 2.40 ± 0.03 19.9 ± 0.3 1.52 ± 0.07 822 ± 107b 3.2 84 ± 1 537 ± 3 7.7 ± 0.1 118 ± 3 158 ± 75 65 ± 2 1.45 ± 0.03 25.6 ± 0.5 1.27 ± 0.03 891 ± 11Total 7 5.8 80 ± 1 480 ± 2 6.7 ± 0.1 122 ± 3 140 ± 53 66 ± 2 2.65 ± 0.02 23.1 ± 0.3 1.52 ± 0.04 861 ± 7

Total 8 2.2 160 ± 2 779 ± 3 14.0 ± 0.7 191 ± 3 538 ± 111 139 ± 6 2.15 ± 0.03 24.4 ± 1.3 1.08 ± 0.02 806 ± 9

9a 3.2 12 ± 1 31 ± 1 1.2 ± 0.1 176 ± 2 278 ± 229 5 ± 1 1.11 ± 0.05 27.9 ± 2.0 5.50 ± 0.26 580 ± 249b 2.7 19 ± 1 70 ± 1 2.0 ± 0.1 124 ± 4 257 ± 106 2 ± 1 1.68 ± 0.04 29.4 ± 1.4 6.05 ± 0.22 588 ± 139c 2.8 15 ± 1 43 ± 1 1.3 ± 0.1 109 ± 2 303 ± 113 7 ± 1 1.25 ± 0.09 23.4 ± 2.2 6.47 ± 0.49 428 ± 339d 2.2 19 ± 1 74 ± 1 2.3 ± 0.1 125 ± 4 530 ± 154 9 ± 1 1.72 ± 0.06 32.9 ± 1.6 5.91 ± 0.26 565 ± 259e 3.0 11 ± 1 35 ± 1 1.2 ± 0.1 91 ± 1 172 ± 655 44 ± 5 1.40 ± 0.05 29.4 ± 1.4 7.37 ± 0.25 542 ± 229f 3.5 57 ± 1 185 ± 2 3.2 ± 0.1 200 ± 4 235 ± 44 191 ± 5 1.44 ± 0.02 15.9 ± 0.7 3.19 ± 0.07 609 ± 89g 3.8 17 ± 1 55 ± 2 1.3 ± 0.1 128 ± 3 233 ± 24 17 ± 1 1.46 ± 0.06 22.2 ± 2.3 6.92 ± 0.26 663 ± 209h 2.6 29 ± 1 113 ± 10 2.6 ± 0.1 154 ± 3 473 ± 153 94 ± 3 1.73 ± 0.15 25.5 ± 0.1 4.81 ± 0.11 595 ± 139i 2.8 62 ± 1 294 ± 3 4.4 ± 0.3 190 ± 3 516 ± 119 85 ± 5 2.11 ± 0.04 19.8 ± 1.3 2.78 ± 0.06 631 ± 99j 4.1 24 ± 1 93 ± 1 1.5 ± 0.1 122 ± 2 2 ± 52 87 ± 6 1.75 ± 0.03 18.0 ± 1.1 4.67 ± 0.09 444 ± 7Total 9 30.5 27 ± 1 99 ± 1 2.1 ± 0.1 132 ± 1 277 ± 73 46 ± 1 1.65 ± 0.02 21.6 ± 0.4 4.50 ± 0.05 444 ± 6

Total 10 12.8 21 ± 2 85 ± 1 1.9 ± 0.1 18 ± 2 nd nd 1.78 ± 0.15 15.0 ± 1.6 0.78 ± 0.10 713 ± 56

Total 11 7.5 67 ± 6 406 ± 3 4.9 ± 1.1 36 ± 8 nd nd 2.68 ± 0.22 23.2 ± 2.3 0.48 ± 0.11 887 ± 73

12a 2.2 594 ± 8 3168 ± 11 37.0 ± 2.0 789 ± 24 3471 ± 1404 83 ± 15 2.36 ± 0.03 17.3 ± 1.0 1.20 ± 0.04 817 ± 1112b 1.5 472 ± 8 2875 ± 11 43.1 ± 2.6 646 ± 40 5554 ± 1791 306 ± 17 2.70 ± 0.05 25.5 ± 1.6 1.24 ± 0.08 746 ± 12Total 12 3.7 545 ± 6 3051 ± 8 39.4 ± 1.6 732 ± 22 4306 ± 1106 173 ± 11 2.48 ± 0.03 20.2 ± 0.8 1.21 ± 0.04 793 ± 8

1 Italics were used to distinguish ‘‘totals” from sub-samples (fragments).

Vo

latileco

mp

ositio

no

ffl

uid

micro

inclu

sion

sin

diam

on

ds

1785


0

5000

10000

15000

20000

25000

30000

35000

40Ar

/36Ar

Cl/36Ar (x 106 molar)0 5 10 15 20 25 30 35 40

PEK013/9

Pandabulkfragments

1500

x 10

-6

500 x 10-6

Fig. 3. 40Ar/36Ar versus Cl/36Ar correlation for Panda diamonds (open triangles – fragments, solid symbols – bulk compositions), errors are1r. The correlation indicates that 40Ar and Cl are present within fluid microinclusions in the diamonds, with variation of 40Ar/Cl in the fluidbeing within the range of approximately 500–1500 � 10�6 obtained from fluid microinclusions in diamonds from worldwide localities.

Panda (this study)

Aikhal

Koffiefontein

Diavik

Jwaneng

Panda (Tomlinson et al 2006)

Cl1

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0Ca1

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0

K 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

912

1

27 3

4

8

6

garnet

cpx

opx

Siberia

Africa

Canada

Fig. 4. Chemical composition of mantle fluids in Panda diamonds (open symbols – fragments; filled symbols – bulk values) and publisheddata. Numbers (n) refer to sample numbers PEK013/n. The compositions of silicate macroinclusions (garnet, clinopyroxene [cpx] andorthopyroxene [opx]) are based upon data from Tomlinson et al., (2006). Data sources for diamonds: Diavik (14 samples) – Klein-BenDavidet al., 2004, 2006, 2007; Panda (7 samples) – Tomlinson et al., 2006; Jwaneng (10 samples) – Schrauder and Navon, 1994; Koffiefontein (10samples) – Izraeli et al., 2001; Aikhal (5 samples)- Burgess et al., 2002.



3.4. Halogens

The fluids in Panda diamonds are enriched in halogensby a factor of 10–50 compared to values measured in coateddiamonds from Africa and Siberia (Johnson et al., 2000;Burgess et al., 2002). This finding is consistent with electronmicroprobe measurements of fluid microinclusions byTomlinson et al. (2006) who obtained high Cl contents,averaging 36.5 and 32.5 wt.% in Panda diamonds contain-ing peridotitic and eclogitic mineral inclusions, respectively.Fig. 5a shows the halogen ratios of fragments and bulksamples of Panda diamonds, compared to some other dia-

monds, MORB and seawater. Panda diamonds show arange of Br/Cl = 1.4–4.3 � 10�3 and I/Cl = 9.4–37.2 � 10�6, they are distinct from seawater and extend significantlyabove the MORB value (Br/Cl = 1.03 ± 0.29 � 10�3 [Jam-bon et al., 1995]; I/Cl = 15–25 � 10�6 [see Johnson et al.,2000 for discussion]), however they are within the rangepreviously determined for Canadian diamonds (Johnsonet al., 2000). Some samples such as PEK013/3, show largehalogen ratio variations between fragments (Fig. 5a) consis-tent with chemical zoning within the diamonds.

Fig. 5b shows a plot of Br/Cl versus I/Cl containing allpublished data for African, Canadian and Siberian dia-

Br/Cl (×103 mol)

I/Cl(

×106

mol

)

1 10 10010

100

1000

10000

Africa

SiberiaCanada

Marine pore fluids

Marine brines

MORB

Panda

(

MORBArea shownin Fig 5a

Br/Cl (×103 mol)

I/Cl(

×10

6m

ol)

01 2 3 4 5

10

20

30

40

0

3

12

7

4

62

1

9

8

5

11

10

seawater

MORB

3

12

7

4

62

1

9

8

5

11

10

Africa

SiberiaCanada

Panda

a

b

Fig. 5. (a) Br/Cl versus I/Cl of Panda diamonds (open symbols – fragments, filled symbols – bulk values) are shown enclosed by a solid linefor each diamond. Number (n) refers to the sample number PEK013/n. Some samples show a large range in composition (e.g. PEK013/3)possibly indicative of chemical zoning. The seawater and MORB halogen compositions are shown for reference. (b) Br/Cl versus I/Cl forfluid-bearing diamonds from Panda, African (Jwaneng kimberlite, Botswana and Democratic Republic of Congo), other Canadian (Fox,Grizzly, Koala and Leslie, kimberlites) and Siberian (Aikhal kimberlite) sources (Johnson et al., 2000; Burgess et al., 2002). Halogencompositions from marine pore fluids (Martin et al., 1993; Mahn and Gieskes, 2001) and marine brines (Muramatsu et al., 2001) overlap withthe range of composition in African diamonds, however the saline fluids in Canadaian diamonds form a distinct array possibly reflectingfractionation during formation of an immiscible brine.



monds (Johnson et al., 2000; Burgess et al., 2002). Themajority of Canadian diamonds form a broad array inFig. 5b, characterised by highly variable Br/Cl and I/Cl ra-tios extending from the MORB value to maximum Br/Cl P 63.0 � 10�3 and I/Cl P 1703.5 � 10�6. Tomlinsonet al. (2006) suggested that the diamond coats grew froman original carbonate–H2O–KCl fluid that unmixed intocarbonatitic and saline fluids, with silicic fluids possiblyformed during metasomatic reactions with eclogitic miner-als. This process appears to lead to the formation of salinefluid present with a wide variation of halogen ratios.

The halogen composition Aikhal (Siberia) diamonds ismost similar to MORB, with Br/Cl = 1.74 ± 0.18 � 10�3

and I/Cl = 22.0 ± 3.4 � 10�6 (Burgess et al., 2002), andmay represent the composition of carbonatitic fluid. Jwan-eng and Zaire (African) and a few Canadian diamondsform a sub-vertical array in Fig. 5b with a relatively widerange of I/Cl between 10 and 100 � 10�6 but more re-stricted Br/Cl of between 1 and 3 � 10�3; this may reflectthe composition of carbonatitic and silicic fluidcomponents.

Several possible processes have been considered previ-ously to explain the halogen variations in diamonds, includ-ing mixing, immiscibility and fractionation involving Cl-rich phases (Johnson et al., 2000; Izraeli et al., 2001; Bur-gess et al., 2002; Klein-BenDavid et al., 2007). Based onan expanded data-set of halogen determinations and ourability to establish their relationship to the three end-mem-ber fluid types (carbonatitic, silicic and saline fluids) we fur-ther consider the evidence for subducted halogens andmantle fractionation as the means to explain halogenvariations.

3.4.1. Seawater-derived halogens

Upper crustal processes often lead to extreme chemicaland isotopic fractionation so variations in mantle materialsare often seemingly compatible with subduction recyclingof crustal materials. The biologically-mediated enrichmentof iodine in sediments means that modern marine pore flu-ids in sediment cores, and sediment-derived seafloor brineslocated adjacent to subduction zones, are highly enriched iniodine and Br relative to seawater (Martin et al., 1993;Mahn and Gieskes, 2001; Muramatsu et al., 2001). Datafor modern marine pore fluids overlaps and extends beyondthe range of I/Cl and Br/Cl of silicic fluids in African dia-monds making it feasible that subducted halogens are pres-ent in fluid microinclusions in diamonds (Fig. 5b). The highenrichment of I in ocean sediments (15 ppm; Muramatsuand Wedepohl, 1998) and pore fluids (8–17 ppm;, Martinet al., 1993; Mahn and Gieskes, 2001) compared to the sub-continental lithospheric mantle (SCLM; 0.4 ppb; Burgesset al., 2002) and MORB source (0.8 ppb; Deruelle et al.,1992), means that this element is potentially a sensitive indi-cator of crustal recycling. For example, the iodine abun-dance of the upper mantle of 0.8 ppb can be explained by53 ppm of recycled marine sediment having an I concentra-tion of 15 ppm and assuming no loss during subduction.This is notably similar to the 44 ppm of recycled seawateror marine pore fluid that has been proposed to explainthe 36Ar concentration of the MORB mantle (Holland

and Ballentine, 2006). To increase the mantle I/Cl valueto the approximate range obtained for African diamondsof 10–100 � 10�6, requires only 3–50 ppm of marine sedi-ment with I/Cl of 150–1000 � 10�6 a range that is typicalfor marine pore fluids.

If crustal recycling has influenced the halogen composi-tion of the fluids sampled by diamonds, then it does not ap-pear to have affected their N isotope compositions whichretain upper mantle values. However, halogens and nitro-gen are expected to be strongly fractionated during dia-mond formation because N content is controlled by itspreferential substitution for carbon in the lattice, whereashalogens are present in trapped fluids in microinclusions(the fluid concentration of N is not known but may alsobe significant).

3.4.2. Mantle halogen fractionation

Fractionation of halogens may accompany the processesleading to the enrichment of halogens in mantle fluids.Mechanisms of halogen fractionation in the mantle arespeculative, because the partition coefficients of halogensare unknown at the pressure, temperature and chemicalconditions of diamond growth. However, partition coeffi-cients are expected to be related to ionic radius with fluid/melt partitioning increasing in order of Cl < Br < I. Exper-imental halogen partitioning data are only available for al-bitic and rhyolitic melts and hydrous fluids at significantlylower pressure than diamond formation (Bureau et al.,2000; Bureau and Metrich, 2003). These results indicatethat Br/Cl and I/Cl of the fluid should be higher by factorsof about 2 and 13, respectively, relative to their startingcomposition; this is similar to the level of enrichment foundin hydrous silicic inclusions in African diamonds, but lowerthan in saline fluids in Canadian diamonds. Experiments onphase relations in the chloride–carbonate–silicate systemsupport formation of an immiscible Cl-rich phase and showevidence for K/Cl variation with Si content (Safonov et al.,2007). The K/Cl fractionation is considered to be controlledby complex exchange reactions with components in the melt(Safonov et al., 2007), however the effects of this process onhalogen fractionation is unknown.

Further halogen fractionation may occur during crystal-lisation of Cl-rich phases (e.g. apatite and amphibole). Apa-tite may be stable in the diamond forming region of theSCLM, is a widespread accessory mineral in xenoliths fromlithospheric mantle (O’Reilly and Griffin, 2000), and hasbeen observed as a secondary phase in fluid microinclusionsin diamonds (Tomlinson et al., 2006; Klein-BenDavid et al.,2007). The problem with invoking the crystallisation of Cl-rich phases to fractionate halogens is in explaining the rel-atively constant 40Ar*/Cl ratios of the different fluid typeswhich are all within about a factor of three and have a med-ian value similar to the predicted MORB value (see Section3.6). Similarly, any decrease in salinity of the residual fluidsaccompanying crystallisation of Cl and OH-rich phasessuch as apatite, may lead to a decrease in Br/Cl and I/Clin the residual fluid, the opposite of what is observed inthe most saline fluids. Other more complex diamond forma-tion processes can also be envisaged including multiple orprolonged diamond growth events in the presence of an



unmixing metasomatic fluid, possibly during interactionwith host mantle rocks, or from multiple fluid sources.These more dynamic interactions may help to account forthe large halogen variability in saline fluids in diamonds.

3.5. Helium and argon isotopes

Helium and argon isotope data are presented in Table 3and shown in Figs. 6–8. Sample PEK013/9 gave differentresults from other Panda diamonds, and is considered sep-arately below. He isotope data (Fig. 6) show a narrow spanof 3He/4He values of 5.4–6.4 Ra (R/Ra = the 3He/4He rationormalized to the atmospheric value of 1.4 � 10�6) how-ever there is no apparent relationship between He isotoperatios and fluid composition. The 3He/4He values are withinthe range of values reported previously for diamonds fromAfrican sources (Fig. 6; 3.7–7.0 Ra – Jwaneng, Burgesset al., 1998; 4.0–7.2 Ra – Democratic Republic of Congo,Wada and Matsuda, 1998). The majority also fall withinthe range obtained from many mantle xenoliths derivedfrom the SCLM of 6 ± 1 Ra (Gautheron and Moreira,2002; Dunai and Porcelli, 2002) and are generally lowerthan the MORB value of 8.75 ± 2.14 Ra (Graham, 2002).The 3He/4He values of fluid-bearing diamonds are similarto of the range of between 5.2 and 5.7 Ra obtained by in va-

cuo crushing release of melt inclusions in olivine from theUdachnaya kimberlite (Sumino et al., 2006) although theyare much lower than fresh kimberlites from Greenland ofup to 27 Ra (Tachibana et al., 2006) which may have pre-served He from a deeper mantle source.

The concentration of He is exceptionally high at be-tween 13 and 66 � 10�6 cm3/g and 1–2 orders of magnitudehigher than diamonds containing carbonatitic or silicic flu-ids (Fig. 6). Helium concentration does not correlate with3He/4He ratio indicating that there has not been a signifi-cant post-entrapment contribution of 4He from in situ de-cay of U and Th, or by external a-implantation. In situ4He contributions in diamonds where both 4He (Table 3)and U (Table 2) concentrations have been determined,and assuming 53 Ma as the age of coats growth, are esti-mated to be: 3% in PEK013/1; 2% in PEK013/2; and 1%in PEK013/4. Assuming an upper limit given by the Re–

Os age of 3.52 Ga for peridotitic diamonds (Westerlundet al., 2006), although incompatible with N aggregationdata, gives in situ produced 4He concentrations higher thanthose measured in all the samples except for PEK013/4where it would account for 41% of the total 4He. At presentit is not possible to determine whether the diamonds havequantitatively retained He (young coats) or experiencedHe loss by diffusion (ancient coats) although negligibleHe loss is most likely for diamond (Honda et al., 1987).From hereon, it is assumed that no major addition or lossof He has occurred, and that the He abundances and iso-tope ratios are representative of the fluid trapped at thetime of coat crystallisation.

Intra-sample helium concentrations vary by up to a fac-tor of two (e.g. PEK013/1; Table 3; Fig. 6), which is attrib-uted either to there being variable amounts of fluid-poor(diamond core) and fluid-rich (coat) material present ineach fragment, or to variations of the inclusion populationdensities in fibrous diamonds (e.g. Klein-BenDavid et al.,2007). The He isotopic compositions of different fragmentsare all indistinguishable within two standard deviationssuggesting there is no evidence for He isotope zoning withindiamonds (Table 3).

Argon isotope ratios of Panda diamonds were measuredby filament heating of unirradiated diamonds and resis-tance furnace stepped heating of irradiated samples duringAr–Ar analyses. Comparing results between the two meth-ods indicates variations in Ar concentrations of up to a fac-tor of 10 difference (PEK013/10) which may be attributedto the variable proportions of 40Ar*-poor core and 40Ar*-rich coat in the samples, or to differences in the numbersof microinclusions. Intra-sample 40Ar* variations reach amaximum of a factor of 5 in PEK013/9. From hereon, onlythe results from unirradiated diamonds are considered be-cause they gave consistently higher 40Ar/36Ar values dueto the order of magnitude lower blank associated with fila-ment heating.

Panda diamonds yielded 40Ar/36Ar values in the range17,000–36,000 (Fig. 7, Table 3) but like He, do not showany evidence for isotopic zonation. 40Ar* concentrationsare high at between 26 and 95 � 10�6 cm3/g (Table 3) andare similar to the concentrations obtained previously for

Table 3He and Ar data from Panda diamonds errors are 1r.

Sample Wt (mg) 4He (�10�6 cm3/g) 36Ar (�10�9 cm3/g) 3He/4He (Ra) 40Ar/36Ar 38Ar/36Ar 40Ar*/4He

PEK013/

1a 4.81 27.95 ± 0.22 1.49 ± 0.06 6.4 ± 0.2 27500 ± 1600 0.1851 ± 0.0144 1.54 ± 0.021b 3.07 13.08 ± 0.10 1.37 ± 0.09 6.4 ± 0.2 17300 ± 750 0.1926 ± 0.0107 1.97 ± 0.02Total 1 7.88 22.15 ± 0.14 1.44 ± 0.05 6.4 ± 0.1 23700 ± 950 0.1880 ± 0.0099 1.64 ± 0.01

Total 2 3.39 19.64 ± 0.15 1.19 ± 0.03 6.4 ± 0.2 26700 ± 1900 0.1928 ± 0.0190 1.77 ± 0.02

4a 3.10 39.55 ± 0.31 3.90 ± 0.31 5.7 ± 0.2 23100 ± 1800 0.1819 ± 0.0164 2.34 ± 0.034b 2.70 34.41 ± 0.27 2.50 ± 0.13 5.4 ± 0.2 28100 ± 2400 0.1884 ± 0.0138 2.29 ± 0.03Total 4 5.80 37.16 ± 0.21 3.25 ± 0.18 5.5 ± 0.1 25000 ± 1400 0.1843 ± 0.0117 2.32 ± 0.02

Total 5 5.58 57.09 ± 0.45 2.17 ± 0.02 5.7 ± 0.2 35800 ± 3000 0.1896 ± 0.0156 1.40 ± 0.02

9a 5.28 1.57 ± 0.01 5.49 ± 0.06 3.3 ± 0.1 480 ± 6 0.1877 ± 0.0014 1.08 ± 0.029b 3.50 2.88 ± 0.02 4.98 ± 0.25 2.8 ± 0.1 499 ± 6 0.1898 ± 0.0019 0.52 ± 0.01Total 9 8.78 2.09 ± 0.01 5.29 ± 0.11 3.0 ± 0.1 487 ± 4 0.1895 ± 0.0027 0.78 ± 0.01

Total 10 3.22 65.43 ± 0.51 2.86 ± 0.25 5.6 ± 0.2 31200 ± 2200 0.1852 ± 0.0166 1.41 ± 0.02

Total 11 3.45 65.89 ± 0.52 3.58 ± 0.30 5.6 ± 0.2 26000 ± 1200 0.1890 ± 0.0033 1.44 ± 0.02

Italics were used to distinguish ‘‘totals” from sub-samples (fragments).



fibrous diamonds from nearby kimberlites of Fox, Grizzly,Koala and Leslie (Johnson et al., 2000). Panda diamondsalso contain relatively high concentrations of 36Ar, howeverthere is no obvious correlation with 40Ar/36Ar ratio (Fig. 7).The origin and implications of the high 36Ar content ofthese diamonds is discussed in Section 3.6.

Most Panda diamonds have similar 40Ar*/4He values inthe range 1.1–2.6 (Fig. 8 and Table 3). This range is abovethe upper mantle 40Ar*/4He production ratio of 0.6 (inte-grated over 4.5 Ga assuming K/U = 12,700), and also thepresent-day mantle production ratio of approximately 0.2

(e.g. Burnard, 2001). The 40Ar*/4He value of diamondsare higher than the MORB popping rock (taken to be theleast degassed basalt) value of �0.7 (Moreira et al., 1998).Enhanced 40Ar*/4He values in diamonds appears to be awidespread characteristic of fluid-bearing diamonds. Previ-ously 40Ar*/4He values of between 0.7–1.0 and 1.1–4.4 wereobtained for diamonds from the Democratic Republic ofCongo (Wada and Matsuda, 1998) and Jwaneng (Burgesset al., 1998), respectively (Fig. 8). It is possible that suchhigh 40Ar*/4He ratios are generated in a mantle region hav-ing an exceptionally high K/U of 50,000. However, K and

0.01 0.1 1 10 1004He (x 10-6 cm3/g)

0

1

2

3

4

5

6

7

8

9

10

3 He/

4 He

(Ra)

- Panda- Zaire

SCLM

xx

x

xx x

x - Jwaneng

PEK013/9

Fig. 6. 3He/4He versus 4He concentration of Panda diamonds (open symbols – fragments; filled symbols – bulk values). Published data forDemocratic Republic of Congo cubic diamonds (Wada and Matsuda, 1998) and Jwaneng coated diamonds (Burgess et al., 1998) are shownfor comparison. Note the low He isotope ratio and concentration in PEK013/9 compared to other Panda diamonds analysed.

0.1 1 10 1001/36Ar (× 1/10-9 cm3/g)

0

5000

10000

15000

20000

25000

30000

35000

40000

40Ar

/36Ar

PEK013/9

PandaZaire2πD43

atmosphere

man

tle

Fig. 7. 40Ar/36Ar versus 1/36Ar for Panda diamonds (open symbols – fragments; filled symbols – bulk values). Published data for DemocraticRepublic of Congo cubic diamonds (Wada and Matsuda, 1989), and the MORB popping rock sample 2pD43 (Moreira et al., 1998; Trieloffet al., 2003) are shown for comparison. With the exception of sample PEK013/9, Panda diamonds have high 40Ar/36Ar values characteristic ofthe upper mantle, and about a factor of 100 higher concentration of 36Ar. The compositions of diamonds from other sources (and includingPEK013/9) are broadly consistent with a mixture of Ar from atmospheric and mantle components as indicated by the arrow.



U concentrations of Panda diamonds determined by Ar–Aranalyses (Table 2), indicate that the mantle fluids havemuch lower K/U values that are mostly between 1000and 2000. Lowering of 40Ar*/4He values in MORB sampleshas been related to the extent of magmatic degassing (e.g.Burnard, 2001), thus the higher values in diamonds maybe characteristic of the mantle source where degassing pro-cesses are unlikely to be a major influence.

Sample PEK013/9 differs from other Panda samples inhaving much lower 3He/4He = 3.0 Ra, 40Ar/36Ar � 500and 40Ar*/4He = 0.5–1; the concentrations of He and Arare also about an order of magnitude lower in this diamond(Figs. 6–8, Table 3). 4He produced by in situ decay of U andTh (U = 46 ppb in Table 2, and assuming Th/U = 3.3)within the fluid inclusions since eruption 53 Ma ago,amounts to 11% of the total, giving an initial trapped3He/4He of only 3.4 Ra. The frosted surface appearanceof this diamond suggests it has been partially resorbed inthe host kimberlite, possibly removing a portion of aninclusion-rich outer layer. Some loss of noble gases mayhave occurred during the resorption process. However,the differences in noble gas isotope ratios are more likelyto reflect a difference in the source of fluids trapped inPEK013/9.

3.6. 36Ar content of the subcontinental lithospheric mantle

The 36Ar content of the mantle is an important con-straint for degassing models of the earth’s interior and for-mation of the atmosphere. It is commonly assumed thatnearly all of the 36Ar present in MORB and OIB samplesderives from atmospheric contamination and that measured40Ar/36Ar in basalts are therefore considered to be mini-mum estimates of their mantle source (e.g. Ballentine andBarfod, 2000). Using noble gas data from continental well

gas samples it has recently been suggested that mantle36Ar may be introduced as dissolved air in seawater by sub-duction processes (Holland and Ballentine, 2006). Becauseof their deep origin and low noble gas diffusivities, atmo-spheric contamination should be much less of a problemin diamonds.

The 40Ar/36Ar values of diamonds are high, typically>3000 with some values in excess of 40,000 (Turner et al.,1990). However the 36Ar contents are usually low beingof similar magnitude to the furnace blanks at temperaturesabove 2000 �C required for graphitisation. In contrast,heating experiments on Panda diamonds using a low blankfilament furnace, show they contain unusually high concen-trations of 36Ar, making it possible to place some con-straints on the 36Ar content of their mantle source. Fig. 7is a plot of 1/36Ar versus 40Ar/36Ar for Panda diamondsand published data for diamonds from the DemocraticRepublic of Congo (Wada and Matsuda, 1998) and pop-ping rock sample 2pD43, the most noble gas-rich MORBglass sample yet analysed (Moreira et al., 1998; Trieloffet al., 2003). With the exception of Panda diamonds, alldata form a broad trend between a component having ahigh 36Ar content and low 40Ar/36Ar(36Ar > 1 � 10�9 cm3/g; 40Ar/36Ar � 300) consistent withatmospheric Ar, and a mantle component having a low36Ar content and high 40Ar/36Ar (36Ar < 0.1 � 10�9 cm3/g; 40Ar/36Ar P 30,000), similar to MORB popping rock.It is not yet possible to determine whether the low40Ar/36Ar component is air contamination from furnaceblank or an air-like component derived from the mantlesource.

With the exception of sample PEK013/9, Panda dia-monds have 40Ar/36Ar of 17,000–36,000 and 36Ar concen-trations of >1 � 10�9 cm3/g, which is at least an order ofmagnitude higher than any other diamond or mantle sam-

x

321040Ar*/4He

3 He/

4 He

(Ra)

PandaDRC

MORB

xx

x

x

x Jwaneng

PEK013/9

2

4

6

8

10

Fig. 8. 3He/4He versus 40Ar*/4He for Panda diamonds (open symbols – fragments; filled symbols – bulk values), Democratic Republic ofCongo (DRC) and Jwaneng. The MORB value is the measured value in popping rock (Graham, 2002 and references therein) is assumed torepresent the least degassed mantle sample. Data for Democratic Republic of Congo and Jwaneng diamonds are from Wada and Matsuda(1998) and Burgess et al. (1998), respectively.



ple analysed (Fig. 7). The high 40Ar/36Ar of Panda dia-monds is characteristic of the MORB value, indicating thatthe high 36Ar content is also derived from their mantlesource. Similar concentrations of 36Ar to those found inPanda diamonds were reported previously in other Cana-dian diamonds (Johnson et al., 2000). The level of enrich-ment of 36Ar is at least a factor of 10 higher compared toAfrican and Aikhal diamonds. This is of similar magnitudeto the enrichment of major elements (e.g. K, Cl, Na) in thesaline fluids (Klein-BenDavid et al., 2007), implying thatthe higher noble gas (He and Ar) contents are more likelyto be related to evolution of the saline fluid rather than todifferences in the population density of microinclusions.In Fig. 7, PEK013/9 lies along the array formed by Africandiamonds, again consistent with it having a fluid composi-tion similar to these samples.

The mantle fluids are so Cl-rich that there is the possibil-ity that some of the Ar in brines may have formed in situ

from the neutron reactions: 35Cl(n,c)36Cl(b)36Ar and37Cl(n,c)38Cl(b)38Ar, with the neutrons being supplied byradioactive decay of U and Th in the fluid and surroundingkimberlite. These reactions would generate a 38Ar/36Ar pro-duction ratio of 0.02, and lead to a gradual decrease overtime of the original mantle 38Ar/36Ar value of 0.188. Themeasured 38Ar/36Ar ratios of Panda diamonds are all with-in error of 0.188 (Table 3) showing that neutron-produced36Ar is only a minor component in the fluid and that thecoats are not exceptionally old. Using the measured Uand Cl contents of the diamonds obtained from Ar–Aranalyses (Table 2), it is estimated that the proportion ofCl-derived 36Ar generated since the kimberlite was em-placed 53 Ma ago, is only of the order of 10�17 cm3/g. Thisis an approximate estimate because it neglects the fractionof thermal neutrons captured by other elements, and takesno account of the neutron flux from the surrounding kim-berlite, however it is noted that the U concentration inthe fluid is likely to be 100 times greater than in the kimber-lite (Schrauder et al., 1996; Tomlinson et al., 2006; Zedgen-izov et al., 2007). Despite these limitations, the contributionof nucleogenic 36Ar in the brines is likely to be negligible,and the 36Ar contents of Panda diamonds are, like the hal-ogens, enriched by several orders of magnitude over typicalupper mantle values.

The strong 40Ar*/Cl correlation found in diamonds andmantle xenoliths during our previous studies is the evidencethat 40Ar* is dissolved in Cl-rich fluid (Turner et al., 1990;Johnson et al., 2000; Burgess et al., 2002). Furthermore, ithas been argued that the relatively restricted range of40Ar*/Cl obtained in diamonds from global sources (cur-rently 444–1347 � 10�6), is in good agreement with the va-lue of 900 ± 400 � 10�6 estimated for MORB mantle basedupon independent estimates of 40Ar and Cl concentrationsof the upper mantle (Johnson et al., 2000). This suggests nosignificant fractionation of noble gases and chlorine duringfluid evolution and justifies the use of Cl/36Ar values to esti-mate the 36Ar content in the region of the mantle where thediamonds formed.

The maximum Cl/36Ar of 30 � 106 is obtained duringhigh temperature heating steps of Canadian diamonds(including Panda sample PEK013/12). This value is as-

sumed to be least affected by atmospheric contaminationand therefore most representative of the mantle end-mem-ber. Combining this value with the average Cl content of34.5 wt.% for microinclusions in Panda diamonds (Tomlin-son et al., 2006) leads to a fluid 36Ar concentration of�7 � 10�6 cm3/g. If it is further assumed that the SCLMhas a Cl content of �3 ppm (Burgess et al., 2002) withthe same Cl/36Ar value as diamonds, then a 36Ar contentof �0.6 � 10�12 cm3/g (�1.7 � 109 atoms/g) is obtainedfor the mantle source. This is indistinguishable from esti-mates of the 36Ar content of the MORB mantle sourcebased on He–Ar systematics of �1.5 � 109 atoms/g (Hol-land and Ballentine, 2006).

It is suggested that the similar noble gas isotope compo-sition (He and Ar) and abundance (36Ar) of SCLM andMORB mantle is not a coincidence and is evidence thatthese mantle regions have experienced similar levels of out-gassing. Volatile components may be periodically concen-trated within the SCLM during melting ormetasomatising events, each perhaps marked by the growthof diamonds and formation of kimberlites.

4. CONCLUSIONS

Combined He, Ar, C, N isotope and K and halogensdeterminations have been carried-out on multiple splits ofindividual microinclusion-bearing diamonds from the Pan-da kimberlite, Slave Craton, Canada. The diamonds havehigh Br/Cl and I/Cl values within the range measured pre-viously for Canadian diamonds from other kimberlites inthis area. Despite the presence of silicate macroinclusions,Ca–Cl–K compositions enable the fluids in these diamondsto be reconciled with previously identified types viz carbon-atitic, hydrous silicic and saline end-members. These fluidshave distinctive Br/Cl and I/Cl values which, combinedwith their mantle-like C and N stable isotope compositions,enable source and mantle fractionation process to be eluci-dated. Carbonatitic fluids in Aikhal diamonds show leastvariation of halogen ratios and are most similar to pres-ent-day MORB values. Halogen ratios in African diamondswith silicic fluids trend toward and overlap with the lowerend of the range of I/Cl values in marine sediments. Theamount of marine pore fluid required to account for thehalogen concentrations in the upper mantle is similar tothat needed for 36Ar recycling. However, there are difficul-ties in understanding how large soluble ions like I are sub-ducted, and the range of marine pore fluid I/Cl and Br/Cldoes not explain the much larger halogen ratios in Cana-dian diamonds. Mantle fractionation processes involving,melt–hydrous fluid partitioning and separation of immisci-ble fluids are believed to play a major role in halogen frac-tionation, without the need to invoke crustal recycling.

ACKNOWLEDGEMENTS

We thank De Beers for providing us with diamond samples.This work would not have been possible without the technical assis-tance of Dave Blagburn and Bev Clementson in Manchester. Weare grateful to Greg Holland for discussions and insights that im-proved the quality of interpretation. We thank O. Navon, D. Phil-lips, H. Sumino and M. Trieloff for reviews that led to significant



improvement of the manuscript and J. Matsuda for helpful com-ments and editorial assistance.

REFERENCES

Ballentine C. J. and Barfod D. N. (2000) The origin of air-likenoble gases in MORB and OIB. Earth Planet. Sci. Lett. 180,

39–48.

Boyd S. R., Mattey D. P., Pillinger C. T., Milledge H. J.,Mendelssohn M. J. and Seal M. (1987) Multiple growth eventsduring diamond genesis: an integrated study of carbon andnitrogen isotopes and nitrogen aggregation state in coatedstones. Earth Planet. Sci. Lett. 86, 341–353.

Boyd S. R., Pillinger C. T., Milledge H. J., Mendelssohn M. J. andSeal M. (1992) C and N isotopic composition and the infraredabsorption spectra of coated diamonds: evidence for theregional uniformity of CO2–H2O rich fluids in lithosphericmantle. Earth Planet. Sci. Lett. 109, 633–644.

Boyd S. R., Kiflawi I. and Woods G. S. (1994) The relationshipbetween infrared absorption and the A defect concentration indiamond. Phil. Mag. B69, 1149–1153.

Boyd S. R., Kiflawi I. and Woods G. S. (1995a) Infraredabsorption by the B nitrogen aggregate in diamond. Phil.

Mag. B72, 351–361.

Boyd S. R., Rejou-Michel A. and Javoy M. (1995b) Improvedtechniques for the extraction, purification and quantification ofnanomole quantities of nitrogen gas: the nitrogen content of adiamond. Meas. Sci. Technol. 6, 297–305.

Bureau H., Keppler H. and Metrich N. (2000) Volcanic degassingof bromine and iodine: experimental fluid/melt portioning dataand applications to stratospheric chemistry. Earth Planet. Sci.

Lett. 183, 51–60.

Bureau H. and Metrich N. (2003) An experimental study ofbromine behaviour in water-saturated silicic melts. Geochim.

Cosmochim. Acta 67, 1689–1697.

Burgess R., Johnson L. H., Mattey D. P., Harris J. W. and TurnerG. (1998) He, Ar and C isotope composition of coated andpolycrystalline diamond. Chem. Geol. 146, 205–217.

Burgess R., Layzelle E., Turner G. and Harris J. W. (2002)Constraints on the age and halogen composition of mantlefluids in Siberian coated diamonds. Earth Planet. Sci. Lett. 197,

193–203.

Burnard P. G. (2001) Correction for volatile fractionation inascending magmas: noble gas abundances in primary mantlemelts. Geochim. Cosmochim. Acta 65, 2605–2614.

Cartigny P. (2005) Stable isotopes and the origin of diamond.Elements 1, 79–84.

Cartigny P., Harris J. W. and Javoy M. (2001) Diamond genesis,mantle fractionations and mantle nitrogen content: a study ofd13C–N concentrations in diamonds. Earth Planet. Sci. Lett.

185, 85–98.

Cartigny P., Harris J. W., Taylor A., Davies R. and Javoy M.(2003) On the possibility of a kinetic fractionation of nitrogenstable isotopes during natural diamond growth. Geochim.


Creaser R. A., Grutter H., Carlson J. and Crawford B. (2004)Macrocrystal phlogopite Rb–Sr dates for the Ekati property,Slave Province, Canada: evidence for multiple intrusive epi-sodes in the Paleocene and Eocene. Lithos 76, 399–414.

Deruelle B., Dreibus G. and Jambon A. (1992) Iodine abundancesin oceanic basalts: implications for Earth dynamics. Earth

Planet. Sci. Lett. 108, 217–227.

Dunai T. J. and Porcelli D. (2002) Storage and transport of noblegases in the subcontinental lithosphere. Rev. Mineral. Geochem.

47, 371–409.

Gautheron C. and Moreira M. (2002) Helium signature of thesubcontinental lithospheric mantle. Earth Planet. Sci. Lett. 199,

39–47.

Graham D. W. (2002) Noble gas isotope geochemistry of mid-ocean ridge and ocean island basalts: characterization of mantlesource reservoirs. Rev. Mineral. Geochem. 47, 247–317.

Harrison D., Leat P. T., Burnard P. G., Turner G., Fretzdorff S.and Millar I. L. (2003) Resolving mantle components in oceaniclavas from segment E2 of the East Scotia back-arc ridge, SouthSandwich Islands. In Intra-Oceanic Subduction Systems: Tec-

tonic and Magmatic Processes, vol. 219 (eds. E. D. Larter and P.T. Leat). Geol. Soc. London Special Publication, pp. 333–344.

Holland G. and Ballentine C. J. (2006) Seawater subductioncontrols the heavy noble gas composition of the mantle. Nature

441, 186–191.

Honda M., Reynolds J. H., Roedder E. and Estein S. (1987) Noblegases in diamonds – occurrences of solar-like helium and neon.Geophys. Res. 92, 12507–12521.

Izraeli E. S., Harris J. W. and Navon O. (2001) Brine inclusions indiamonds: a new upper mantle fluid. Earth Planet. Sci. Lett.

187, 323–332.

Jambon A., Deruelle B., Dreibus G. and Pineau F. (1995) Chlorineand bromine abundance in MORB: the contrasting behaviourof the Mid-Atlantic Ridge and East Pacific Rise and implica-tions for the chlorine geodynamic cycle. Chem. Geol. 126, 101–

117.

Johnson L. H., Burgess R., Turner G., Harris J. H. and MilledgeH. J. (2000) Noble gas and halogen geochemistry of mantlefluids in diamond: comparison of African and Canadian stones.Geochim. Cosmochim. Acta 64, 717–732.

Klein-BenDavid O., Izraeli I., Hauri E. and Navon O. (2004)Mantle fluid evolution – a tale of one diamond. Lithos 77, 243–

253.

Klein-BenDavid O., Wirth R. and Navon O. (2006) TEM imagingand analysis of microinclusions in diamonds: a close look atdiamond-growing fluids. Am. Mineral. 91, 353–365.

Klein-BenDavid O., Izraeli E. S., Hauri E. and Navon O. (2007)Fluid inclusions in diamonds from the Diavik mine, Canadaand the evolution of diamond-forming fluids. Geochim. Cos-

mochim. Acta 71, 723–744.

Mahn C. L. and Gieskes J. M. (2001) Halide systematics incomparison with nutrient distributions in sites 1033B and1034B, Saanich Inlet: ODP Leg 169S. Mar. Geol. 174, 323–339.

Martin J. B., Gieskes J. M., Torres M. and Kastner M. (1993)Bromine and iodine in Peru margin sediments and pore fluids –implications for fluid origins. Geochim. Cosmochim. Acta 57,

4377–4389.

Moreira M., Kunz J. and Allegre C. (1998) Rare gas systematics inpopping rock: isotopic and elemental compositions in the uppermantle. Science 279, 1178–1181.

Muramatsu Y. and Wedepohl K. H. (1998) The distribution ofiodine in the earth’s crust. Chem. Geol. 147, 201–216.

Muramatsu Y., Fehn U. and Yoshida S. (2001) Recycling of iodinein fore-arc areas: evidence from the iodine brines in Chiba,Japan. Earth Planet. Sci. Lett. 192, 583–593.

Navon O., Hutcheon I. D., Rossman G. R. and Wasserburg G. J.(1988) Mantle-derived fluids in diamond micro-inclusions.Nature 335, 784–789.

Navon O., Izraeli E. S., and Klein-BenDavid O. (2003) Fluidinclusions in diamonds – the carbonatitic connection. EighthInternational Kimberlite Conference Victoria, Canada, 2003.#FLA_0107 (abstr.).

Navon O. (1999) Diamond formation in the Earth’s mantle. InProceedings of the seventh International Kimberlite Conference(eds. J. J. Gurney, J. L. Gurney, M. D. Pearce, S. H.Richardson). vol. 2, pp. 584–604.



Nowicki T., Crawford B., Dyck D., Carlson J., McElroy R.,Oshust R. and Helmstaedt H. (2004) The geology of kimberlitepipes of the Ekati property, Northwest Territories, Canada.Lithos 76, 1–27.

O’Reilly S. Y. and Griffin W. L. (2000) Apatite in the mantle:implications for metasomatic processes and high heat produc-tion in Phanerozoic mantle. Lithos 53, 217–232.

Perchuk L. L., Safonov O. G., Yapaskurt V. O. and Barton J. M.(2002) Crystal-melt equilibria involving potassium-bearingclinopyroxene as indicator of mantle-derived ultrahigh potassicliquids: an analytical review. Lithos 60, 89–111.

Richardson S. H., Shirey S. B., Harris J. W. and Carlson R. W.(2001) Archean subduction recorded by Re–Os isotopes ineclogitic sulfide inclusions in Kimberley diamonds. Earth

Planet. Sci. Lett. 191, 257–266.

Robinson D. N., Scott J. A., Van Niekerk A., Anderson V. G.(1989) The sequence of events reflected in the diamonds of somesouthern African kimberlites. In Kimberlites and related rocks.Their mantle crust setting. Diamonds and diamond explora-tion, vol. 2. Geol. Soc. Australia, GSA Special Publication 14,pp. 990–1000.

Safonov O. G., Perchuk L. L. and Litvin Y. A. (2007) Meltingrelations in the chloride–carbonate–silicate systems at high-pressure and the model for formation of alkalic diamond-forming liquids in the upper mantle. Earth Planet. Sci. Lett.

253, 112–128.

Schrauder M. and Navon O. (1994) Hydrous and carbonatiticmantle fluids in fibrous diamonds from Jwaneng. Geochim.


Schrauder M., Koeberl C. and Navon O. (1996) Trace elementanalyses of fluid-bearing diamonds from Jwaneng, Botswana.Geochim. Cosmochim. Acta 60, 4711–4724.

Shiryaev A. A., Izraeli E. S., Hauri E. H., Zakharchenko O. D. andNavon O. (2005) Chemical, optical and isotopic investigation offibrous diamonds from Brazil. Russ. Geol. Geophys. 46, 1185–

1201.

Sumino H., Kaneoka I., Matsufuji K. and Sobolev A. V. (2006)Deep mantle origin of kimberlite magmas revealed by neonisotopes. Geophys. Res. Lett. 33, L16318.

Tachibana Y., Kaneoka I., Gaffney A. and Upton B. (2006) Ocean-island basalt-like source of kimberlite magmas from WestGreenland revealed by high 3He/4He ratios. Geology 34, 273–

276.

Taylor W. R., Canil D. and Milledge H. J. (1996) Kinetics of Ib toIaA nitrogen aggregation in diamond. Geochim. Cosmochim.

Acta 60, 4725–4733.

Tomlinson E. L., Jones A. P. and Harris J. W. (2006) Co-existingfluid and silicate inclusions in mantle diamond. Earth Planet.

Sci. Lett. 250, 581–595.

Trieloff M., Falter M. and Jessberger E. K. (2003) The distributionof mantle and atmospheric argon in oceanic basalts glasses.Geochim. Cosmochim. Acta 67, 1229–1245.

Turner G., Burgess R. and Bannon M. (1990) Volatile-rich mantlefluids inferred from inclusions in diamond and mantle xeno-liths. Nature 344, 653–655.

Wada N. and Matsuda J. (1998) A noble gas study of cubicdiamonds from Zaire: constraints on their mantle source.Geochim. Cosmochim. Acta 62, 2335–2345.

Westerlund K. J., Shirey S. B., Richardson S. H., Carlson R. W.,Gurney J. J. and Harris J. W. (2006) A subduction wedge originfor Paleoarchean peridotitic diamonds and harzburgites fromthe Panda kimberlite, Slave craton: evidence from Re–Osisotope systematics. Contrib. Mineral. Petrol. 152, 275–294.

Zedgenizov D. A., Kagi H., Shatsky V. S. and Sobolev N. V. (2004)Carbonatitic melts in cuboid diamonds from the Udachnayakimberlite pipe (Yakutia): evidence from vibrational spectros-copy. Min. Mag. 68, 61–73.

Zedgenizov D. A., Rege S., Griffin W. L., Kagi H. and Shatsky V.S. (2007) Composition of trapped fluids in cuboid fibrousdiamonds from the Udachnaya kimberlite: LAM-ICPMSanalysis. Chem. Geol. 240, 151–162.

Associate editor: Jun-ichi Matsuda


volatile composition of microinclusions in diamonds from ...cartigny/2009-gca-burgessetal.pdf ·...

Documents