Contributions to Mineralogy and Petrology

, Volume 160, Issue 4, pp 489–510

Mineral inclusions in sublithospheric diamonds from Collier 4 kimberlite pipe, Juina, Brazil: subducted protoliths, carbonated melts and primary kimberlite magmatism

Authors

    • Department of Earth SciencesUniversity of Bristol
  • Michael J. Walter
    • Department of Earth SciencesUniversity of Bristol
  • Chris B. Smith
    • Department of Earth SciencesUniversity of Bristol
  • Simon C. Kohn
    • Department of Earth SciencesUniversity of Bristol
  • Lora S. Armstrong
    • Department of Earth SciencesUniversity of Bristol
  • Jon Blundy
    • Department of Earth SciencesUniversity of Bristol
  • Luiz Gobbo
    • Rio Tinto Desenvolvimentos Minerais Ltda
Original Paper

DOI: 10.1007/s00410-010-0490-6

Cite this article as:
Bulanova, G.P., Walter, M.J., Smith, C.B. et al. Contrib Mineral Petrol (2010) 160: 489. doi:10.1007/s00410-010-0490-6

Abstract

We report on a suite of diamonds from the Cretaceous Collier 4 kimberlite pipe, Juina, Brazil, that are predominantly nitrogen-free type II crystals showing complex internal growth structures. Syngenetic mineral inclusions comprise calcium- and titanium-rich phases with perovskite stoichiometry, Ca-rich majoritic-garnet, clinopyroxene, olivine, TAPP phase, minerals with stoichiometries of CAS and K-hollandite phases, SiO2, FeO, native iron, low-Ni sulfides, and Ca–Mg-carbonate. We divide the diamonds into three groups on the basis of the carbon isotope compositions (δ13C) of diamond core zones. Group 1 diamonds have heavy, mantle-like δ13C (−5 to −10‰) with mineral inclusions indicating a transition zone origin from mafic protoliths. Group 2 diamonds have intermediate δ13C (−12 to −15‰), with inclusion compositions indicating crystallization from near-primary and differentiated carbonated melts derived from oceanic crust in the deep upper mantle or transition zone. A 206Pb/238U age of 101 ± 7 Ma on a CaTiSi-perovskite inclusion (Group 2) is close to the kimberlite emplacement time (93.1 ± 1.5 Ma). Group 3 diamonds have extremely light δ13C (−25‰), and host inclusions have compositions akin to high-pressure–temperature phases expected to be stable in pelagic sediments subducted to transition zone depths. Collectively, the Collier 4 diamonds and their inclusions indicate multi-stage, polybaric growth histories in dynamically changing chemical environments. The young inclusion age, the ubiquitous chemical and isotopic characteristics indicative of subducted materials, and the regional tectonic history, suggest a model in which generation of sublithospheric diamonds and their inclusions, and the proto-kimberlite magmas, are related genetically, temporally and geographically to the interaction of subducted lithosphere and a Cretaceous plume.

Keywords

BrazilCollier 4 kimberliteSublithospheric diamondsInclusionsCarbon isotopesSubductionCarbonatite melt

Introduction

The majority of the world’s diamonds are interpreted to have formed in the subcontinental lithospheric mantle 3.5–1.5 Ga ago (Pearson et al. 1999; Richardson et al. 1984; Rudnick et al. 1993). However, during the last two decades it has been documented on the basis of the mineralogy of inclusions within diamonds, that some diamonds originate from deeper, sublithospheric horizons and provide examples of transition zone and lower mantle compositions (Brenker et al. 2002; Harte and Harris 1994; Hutchison 1997; Kaminsky et al. 2001; Stachel et al. 2000; Stachel et al. 2005; Wilding 1990). Inclusions in ‘ultradeep’ sublithospheric diamonds also provide potential examples of chemical heterogeneity within the deep mantle, possibly linked to subduction of lithosphere, oceanic crust and carbonaceous sediments, and of the interaction of deep mantle reservoirs with partially melted subducted materials (Brenker et al. 2002, 2005, 2007; Tappert et al. 2005a, b, 2009; Walter et al. 2008).

The Juina kimberlite field in Brazil is a well-known source of alluvial sublithospheric diamonds as identified by their properties and mineral inclusions (Araujo et al. 2003; Harte et al. 1999; Hayman et al. 2005; Hutchison et al. 2001; Kaminsky et al. 2001). Diamonds from primary kimberlite sources in the Juina area have been recently described (Andreazza et al. 2008; Bulanova et al. 2008; Kaminsky et al. 2009; Walter et al. 2008). Here, we present a comprehensive study of sublithospheric diamonds and their syngenetic inclusions from the Collier 4 kimberlite pipe. Sublithospheric diamonds have complex and irregular shapes, are broken, resorbed, etched, plastically deformed and contain deep cracks. We therefore examined them in polished plates and sections so that any core–rim primary diamond zonation can be revealed and the syngenetic nature of inclusions demonstrated. We use mineralogical, chemical, and isotopic evidence from the diamonds and their inclusions to document the nature of deep mantle compositions, and to interpret diamond formation and kimberlite emplacement within the regional tectonothermal setting of the Amazon craton.

Geological background

The diamonds are from the Collier 4 kimberlite pipe located within the Juina kimberlite field. This pipe is one of the possible primary kimberlite sources for the Juina sublithospheric inclusion-bearing alluvial diamonds previously described from the Sao Luiz and Rio Sorriso rivers (Harte et al. 1999; Hayman et al. 2005; Hutchison et al. 2001; Kaminsky et al. 2001) (Fig. 1). The pipe has a U/Pb zircon age of 93.1 ± 1.5 Ma (Heaman et al. 1998) and intrudes granitoid gneisses of the Mesoproterozoic Rio Negro–Juruena Mobile Belt. This belt accreted onto the southern edge of the Amazon Craton during coalescence of magmatic arcs between 1.75 and 1.55 Ga in a continental margin setting, preceded by subduction of oceanic crust (de Almeida et al. 2000; Geraldes et al. 2004). Mafic eclogites and garnet lherzolites are the most abundant mantle xenoliths in the Juina kimberlites with estimated temperatures and pressures for the peridotites of 860–1,360°C and 5.1–6.7 GPa, and temperatures of 1,180–1,290°C for the eclogites (Costa et al. 2003). The peridotites give Proterozoic model ages of 1,884–1,114 Ma and three eclogites give ages of 1,166, 1,593 and 1,648 Ma, suggesting a possible link to Mesoproterozoic subduction of oceanic crust during formation of the Mobile Belt. The base of the lithosphere beneath Juina at the southern edge of the Amazon craton is at a depth of ~140 km according to constraints from seismic tomography (Feng et al. 2007).
https://static-content.springer.com/image/art%3A10.1007%2Fs00410-010-0490-6/MediaObjects/410_2010_490_Fig1_HTML.gif
Fig. 1

a Map of Brazil showing the position of the Juina field relative to Mesozoic alkaline volcanics and the speculated Trindade plume track. b Location of the Juina kimberlite field showing the position of the Collier 4 kimberlite in relation to nearby alluvial occurrences of sublithospheric diamonds described by earlier workers

The Juina Field is one of many expressions of Cretaceous alkaline volcanicity (kimberlites, lamproites, kamafugites, carbonatites) developed along the AZ 125° lineament (Bardet 1977; Tompkins 1991), and has been ascribed to the passage of the Brazilian continent over the Trindade Plume (Gibson et al. 1995) (Fig. 1b). The Juina Field is near the north-western end of this 2,500 km long alignment of volcanic and intrusive rocks which show a general decrease of emplacement ages in a south-easterly direction. The offshore Trindade–Vitoria seamount chain ostensibly represents the Tertiary to Quaternary surface expression of the plume (Fig. 1b).

Samples and analytical methods

Samples

Twenty diamonds with mineral inclusions were selected from a Rio Tinto collection for study by optical microscopy. The diamonds are 2–3.5 mm in size, light brown or occasionally white octahedral–dodecahedral transition forms, with rare macles and aggregates (Table 1). Most of these crystals are broken, heavily resorbed, plastically deformed and have internal cracks and deep etch channels (Fig. 2). The diamonds display a dark blue color in photoluminescent (PL) and cathodoluminescent (CL) light, or do not show any luminescence. The diamonds were polished along dodecahedral, cubic, or slightly inclined to the octahedron planes, in order to expose inclusions for analysis and to produce central sections of crystals for study of growth history and carbon isotope composition in core–rim traverses. The spatial location of inclusions in relation to diamond core–rim zonation was recorded.
Table 1

Diamond characteristics, carbon isotope compositions, mineral inclusions and pressure estimates

#

Shape

Sizea

PL color

N ppm (X IaB) core; rim

Type

δ13C ‰

Inclusions

P estimate; methodb

Core

Inter

Rim

J1

O/D/b

3.5

None

Below detection

II

−14.7

−14.1

−12.7

CaSiO3 + CaTiO3 = CaTiSi-Pv; Maj-Gt; Po

>10 GPa < 20 to ~ 7 GPa; ph-rel and Xmaj

J2

O/M/I/b

2.5

None

Below detection

II

−24.6

−23.8

−14.7

CaAlSi + Ky; Fe; Mgt; MgFe-spinel; Maj-Gt; K-Fsp

>15 GPa to 7 GPa; ph-rel and Xmaj

J3

O/D/I/b

3.3

Dark blue

102 (0.29); 585, (0.61)

Ia

−10.1

na

−10.6

Six clinopyroxenes

~5 GPa; Cr-Cpx

J4

O/I/b

1.8

Dark blue

Below detection

II

−5.8

−5.8

−6.6

TAPP phase; Carbonate

Trans. zone? TAPP stability?

J5

O/M/I/b

3.0

None

Below detection

II

−26

−24

−18.2

Po + Magnetite; SiO2

None

J6

O/D

3.2

Blue

Below detection

II

−8

−8

−6.7

Po

None

J7

O/f

2.2

Blue

67 (1.0); 180 (0.95)

Ia

−11.8

−11.8

−12.5

Po

None

J8

O/b

2.0

None

Below detection

II

−9.2

−8.8

−5.9

(Cpx + MgFeAl-phase) = Maj-Gt?

~8 GPa; Xmaj

J9

O/f

2.4

Dark blue

Below detection

II

−13.3

−10.1

−7.5

Three Maj-Gt

Xmaj

J10

O/D

2.5

Blue

Below detection

II

−14.4

−13.4

−13.1

CaSiO3 + CaTiO3 = CaTiSi-Pv; Po

>10 GPa < 20 to ~7 GPa; ph-rel

J11

M/A

1.6

Dark blue

Below detection

II

na

na

na

Black microinclusions

None

J12

M/O/A

2.2

None

Below detection

II

−15

−13

−18

Black microinclusions

None

J13

D

2.7

None

Below detection

II

na

na

na

KAlMgFeTiSi microinclusion

None

J14

O/D/b

3.0

None

Below detection

II

−8.5

−8.5

−6.4

Three CaSi-Pv; SiO2

>~17 GPa; ph-rel

J15

O/I

2.8

Weak blue

Below detection

II

na

na

na

Po

None

J16

D

4.0

Blue

Below detection

II

na

na

na

Po

None

J17

D/b

2.2

Weak blue

170 (1.0)

Ia

na

na

na

SiO2

None

J18

D/b

2.8

Weak blue

Below detection

II

−24.8

−25

−24.8

Fe + FeO

None

J19

M/O/A

1.4

None

na

na

−23.4

−23.4

−22.9

SiO2 + kyanite

None

J20

D/b

2.7

Weak blue

Below detection

II

−5.6

−5.6

−6.4

‘Olivine’; Ca–Mg-carbonate

None

O octahedron, Dod. dodecahedroid, I intergrowth, M macle, A aggregate, b broken, f fragment, PL photoluminescence, Pv perovskite, Maj majorite, Gt garnet, Ky kyanite, Fe native iron, Mgt magnetite, Fsp feldspar, Po pyrrhotite, Cpx clinopyroxene, na not analyzed

aAverage diameter in mm

bEstimated pressures from phase relations (ph-rel), majorite component in garnet (Xmaj), and Cr in Cpx

https://static-content.springer.com/image/art%3A10.1007%2Fs00410-010-0490-6/MediaObjects/410_2010_490_Fig2_HTML.jpg
Fig. 2

Morphology of Collier 4 diamonds. J1 chipped, resorbed and etched white octahedron, J2 white octahedron/macle/intergrowth broken on one side, J9 pale brown resorbed, broken octahedron, J10 white resorbed octahedron/dodecahedroid

Analytical methods

The internal morphology of diamonds was investigated using anomalous birefringence (ABR), photoluminescence (PL) and cathodoluminescence (CL) by imaging of their central sections. ABR patterns of diamond plates were observed using a polarized light microscope. The PL images of diamonds were obtained using a OI-18A UV light source instrument. CL images were taken with a JSM-35 scanning electron microscope (I = 3 nA, U = 20 kV) at the University of Bristol and a Phillips XL30CP scanning electron microscope at the Grant Institute of Earth Sciences, Edinburgh University. Using these images as maps for crystal growth, the spatial distribution of N and H impurities and carbon isotopes were studied in core–rim traverses of the diamonds.

Infrared spectra were collected using a SpectraTech infrared microscope coupled to a Nicolet Nexus Fourier transform infrared spectrometer at the University of Bristol. All spectra were recorded at 4 cm−1 resolution using unpolarised IR light generated by a Globar source, KBr beamsplitter and a liquid nitrogen-cooled MCT detector. Doubly polished diamond slices were placed on a KBr disc, and a background was recorded through the disc. The baselines of the spectra were estimated visually and subtracted using the Nicolet OMNIC software. The nitrogen concentration and degree of aggregation in terms of A, B and D components reported in Table 1 were determined by the procedure of Taran et al. (2006). The results were checked against an alternative method of data reduction using the “FTIR analyzer” program of John Chapman (Rio Tinto Diamonds) and good agreement was found in all cases.

Electron microprobe analysis of exposed and polished mineral inclusions in diamonds was made at the University of Bristol with a Cameca SX100 using a beam current of 20 nA at 15 kV voltage, and a spot size of ~1 micron at the surface. Silicate and oxide standards with conventional PAP data reduction techniques were employed, and replicate analysis of standards yields uncertainties at the 2 and 5% level, respectively, for major and minor elements.

Trace element concentrations in mineral inclusions were determined using a Cameca IMS-4f ion-microprobe at the Edinburgh Ion Microprobe Facility (EIMF). The primary beam was ~11 keV 16O ions (~15 keV net impact energy), with a sample current of 2 nA that corresponds to a spatial resolution of ~15 μm at the sample surface. The secondary ion accelerating voltage of 4,500 V was offset by 75 eV (energy window of 40 eV) to reduce molecular ion transmission. Calibration was performed on glass standards under identical operating conditions. Statistical precision at concentrations >1 ppm is better than 10% relative for all isotopes. Accuracy is better than 10% relative for the rare-earth elements (REE), Ba, Sr, Nb, Zr and Y. Hf, Rb, Th and U are accurate to within 30% relative.

The carbon isotope data were acquired at the EIMF with a Cameca ims 1270, using a ~6 nA primary 133Cs+ beam. Secondary ions were extracted at 10 kV, and 12C- (~2.0 × 107 cps) and 13C- (~1.0 × 109 cps) were monitored simultaneously on dual Faraday cups (L′2 and H′2). Each analysis involved a pre-sputtering time of 60 s, followed by data collection in two blocks of five cycles, amounting to a total count time of 40 s. The internal precision of each analysis is ±0.2 per mil. To correct for instrumental mass fractionation (IMF), all data were normalized to standard, synthetic diamond (SYNAL δ13C −23.92 PDB, B. Harte, personal communication), which was measured throughout the analytical sessions and mounted in the same sample holder as the unknowns.

U/Pb dating of a CaTiSi-Pv inclusion was made by SIMS on a Cameca ims-1270 at EIMF based on U/Pb versus UO/U ratios and using titanite and perovskite standards. The combined use of both standards was needed as the characteristics of the matrix analyzed (principally denoted by the U/oxide ratios) fell between those observed for these two phases.

Results

Diamond internal structures

Previous studies have documented that diamonds of sublithospheric origin are characterized by complex internal structures, plastic deformation and stress, and breakages and resorption (Gaspar et al. 1998; Hayman et al. 2005; Hutchison 1997; Kaminsky et al. 2001). Collier 4 diamonds are also very different from lithospheric diamonds in terms of internal complexity. While a few of the diamonds have weak undisturbed octahedral zonation (Fig. 3a), the majority contain internal cracks in-filled with later diamond generation (Fig. 3b), are broken and resorbed crystals (Fig. 3c) or display very complex growth patterns (Fig. 3d, e). For example, Fig. 3e shows the complex growth structure of diamond J9. The core zone of an octahedral/rounded shape likely formed in a regime of oscillating growth and slight resorption, which is consistent with free growth from a silicate melt/fluid with a low-degree of carbon saturation (Sunagawa 1984). The right side of the core zone was then intensively and selectively resorbed. The intermediate area shows irregular-aggregate growth with features of deformation and stress, which can be attributed to formation in a more viscous environment with a higher degree of carbon saturation; these conditions of growth are likely synchronous with deformation (Grigor’yev and Jabin 1975). Several inclusions of diamond-in-diamond in this zone, including one with cubic shape (Fig. 3e), demonstrate that many diamond seeds were nucleating within close proximity without developing individually into large crystals, but were instead incorporated into the intermediate zone of an enveloping larger crystal (J9). The narrow rim zone of the diamond has fine octahedral zoning with plastic deformation. Many cracks, some internal, others reaching the surface, are present in the stone. It is apparent that Collier 4 diamonds have a very complex history of formation, originating under conditions of alternating growth and intensive resorption, whilst having been subjected to both brittle and plastic deformation similar to sheared mantle xenoliths.
https://static-content.springer.com/image/art%3A10.1007%2Fs00410-010-0490-6/MediaObjects/410_2010_490_Fig3_HTML.jpg
Fig. 3

CL images revealing the internal structure of Collier 4 diamonds. Also shown on the diamond images are δ13C values (‰ relative to PDB) as determined by SIMS analysis. a J2 diamond with weak octahedral zonation and large change in δ13C from core to rim, b weakly zoned J3 stone with deep system of internal cracks, c J6 broken and resorbed crystal, d J8—broken, deformed and resorbed diamond, e J9 diamond with ‘sheared’ zonation described in the text and large change in δ13C from core to rim

Collier 4 diamonds exhibit several different kinds of macroscopic defects and it is not always straightforward to classify inclusions as ‘syngenetic’ and ‘epigenetic’, even in polished sections. The inclusions discussed below are all classified as syngenetic because they have morphology imposed by the diamond-hosts and do not contain cracks connected with the surface. Inclusions identified as epigenetic are not considered further.

Nitrogen content and aggregation

FTIR characteristics of the diamonds are given in Table 1. All but three diamonds studied are N-free (type II), a generally consistent feature of sublithospheric diamonds (Hayman et al. 2005; Hutchison et al. 1999; Kaminsky et al. 2001; Stachel et al. 2001). Diamond J7 has 67 ppm N in the core (100% type 1aB) and 180 ppm N in the rim (95% type 1aB). Diamond J17 has 170 ppm N and very high aggregation (100% 1aB). Both diamonds contain moderate to high amounts of hydrogen as shown by the presence of a peak at 3,107 cm−1 (absorbance 0.3–2.1 cm−1). Diamond J3 is more complex as it is zoned, both in terms of N concentration and aggregation. It contains 102 ppm N in the core (29% type 1aB) and 585 ppm in a rim zone (61% type 1aB). Such zonation is not common for lithospheric diamonds where the core zones typically have higher N content and aggregation than later rim growth. In the case of diamond J3, the co-variations in nitrogen concentration and aggregation can be explained in terms of constant temperature during growth and a single age for the diamond (i.e. rapid growth rate compared with the temporal resolution of the technique), because aggregation rate increases as a function of nitrogen concentration (Taylor et al. 1990).

In our Collier 4 collection (20 diamonds) there is a much higher proportion of type II diamonds (~80%) than found among the 150 diamonds studied by Kaminsky et al. (2009) (~12%). Further, all three of the 1aA–1aB type diamonds from our collection (Table 1) are richer in nitrogen and show higher aggregation than those reported in that study. The reason for this discrepancy is not clear, but it could be a sampling bias in one or both cases, as all the diamonds presented here were selected because they contain inclusions, but most of those reported by Kaminsky et al. (2009) are not described as containing inclusions.

The depths of origin of diamonds J3, J7 and J17 are not well constrained by the mineral inclusions (clinopyroxene and pyrrhotite, Table 1) and could be either lithospheric or sublithospheric. Should these diamonds be lithospheric, then possible mantle residence times of ~2 Ga would produce the observed N aggregation values at temperatures (1,170–1,300°C) typical of the lithosphere. Many diamonds from the Juina area, including most of those from Collier 4 as described below, are sublithospheric in origin and may have originated at considerably higher temperatures. For example, at an ambient mantle temperature of 1,500°C, such as might occur along a normal mantle geotherm at transition zone depths, the observed degrees of aggregation would require ≪1 Ma of mantle residence time for J3 and 1–10 Ma for J7 and J17.

Mineralogy and chemistry of inclusions

Table 1 provides a summary of the mineral inclusions found in the Collier 4 diamonds. Table 2 presents the major element chemistry of the inclusions, and Table 3 gives the trace element chemistry of selected inclusions. In this section, we describe briefly the chemistry and mineralogy of the inclusions, some of which are shown in backscattered electron images in Fig. 4.
Table 2

Major element composition of mineral inclusions as determined by electron probe microanalysis

Diamond #

J1

J1

J1

J1

J2

J2

J2

J2

J2

J2

J3

J3

J3

J3

Inclusion#

1a

1b

1 bulk

3

2-1

2-2

2 bulk

4

5

6

3

4

5

6

Phase

CaTi-pv

CaSi-pv

CaTiSi-pva

Garnet

CaAlSi

Kyanite

CASb

Sp

K-Fsp

Garnet

Cpx

Cpx

Cpx

Cpx

nc

1

1

1

6

1

1

1

1

1

2

2

2

2

2

SiO2

1.35

50.94

18.71

41.35

36.11

35.45

36.29

0.44

50.75

40.59

55.56

55.89

54.35

54.33

TiO2

56.46

0.20

36.77

1.35

0.59

0.08

0.42

0.23

0.48

0.93

0.63

0.50

0.51

0.49

Al2O3

1.17

0.33

0.88

21.64

34.91

60.51

41.99

64.42

19.12

21.84

3.48

3.45

3.47

3.39

Cr2O3

d

0.01

0.05

0.03

0.01

0.03

0.10

0.01

0.04

0.21

0.18

0.27

0.24

FeO

0.52

1.87

0.99

10.51

1.83

0.08

1.36

11.51

2.31

16.42

8.25

8.06

8.21

8.09

MnO

0.02

0.04

0.03

0.30

0.01

0.01

0.01

0.19

0.04

0.38

0.12

0.11

0.12

0.12

MgO

0.00

0.07

0.02

9.03

1.12

0.02

0.80

20.67

1.00

11.17

16.33

16.23

15.60

15.82

CaO

38.08

45.56

40.7

15.63

20.05

0.32

14.21

0.01

0.19

8.61

13.44

13.60

13.49

13.60

Na2O

0.12

0.15

0.13

0.59

1.83

0.06

1.32

0.02

2.16

0.40

0.00

0.00

0.01

0.02

K2O

0.01

0.03

0.02

1.47

0.02

1.00

0.16

16.10

0.00

1.82

2.06

1.84

1.83

NiO

0.02

Total

97.72

99.19

98.23

100.45

97.93

96.55

97.42

97.75

92.16

100.37

99.85

100.08

97.87

97.91

Cations

 Si

0.031

0.996

0.404

3.047

2.448

0.993

2.391

0.011

2.670

3.015

2.009

2.015

2.008

2.006

 Ti

0.972

0.003

0.597

0.075

0.030

0.002

0.021

0.004

0.019

0.052

0.017

0.014

0.014

0.014

 Al

0.032

0.008

0.022

1.879

2.790

1.997

3.261

1.947

1.186

1.912

0.148

0.147

0.151

0.148

 Cr

0.000

0.000

0.000

0.003

0.001

0.000

0.001

0.002

0.000

0.002

0.006

0.005

0.008

0.007

 Fe

0.010

0.031

0.018

0.648

0.104

0.002

0.075

0.247

0.102

1.020

0.250

0.243

0.254

0.250

 Mn

0.000

0.001

0.001

0.019

0.000

0.000

0.000

0.004

0.002

0.024

0.004

0.003

0.004

0.004

 Mg

0.000

0.002

0.001

0.992

0.113

0.001

0.079

0.790

0.078

1.236

0.880

0.872

0.859

0.871

 Ca

0.934

0.954

0.942

1.234

1.457

0.009

1.003

0.000

0.011

0.685

0.521

0.525

0.534

0.538

 Na

0.005

0.006

0.006

0.084

0.240

0.003

0.168

0.001

0.220

0.057

0.128

0.144

0.132

0.131

 K

0.000

0.001

0.001

0.127

0.001

0.084

0.005

1.081

0.000

0.000

0.000

0.000

0.001

 Ni

0.000

Cations

1.984

2.001

1.991

7.980

7.310

3.009

7.083

3.013

5.368

8.004

3.961

3.968

3.965

3.969

O

3

3

3

12

11

5

11

4

8

12

6

6

6

6

Mg#e

0.605

0.521

0.316

0.511

0.762

0.434

0.548

0.779

0.782

0.772

0.777

Ca#e

0.554

0.928

0.915

0.927

0.000

0.121

0.357

0.372

0.376

0.383

0.382

J4

J8

J8

J8

J9

J9

J9

J10

J10

J10

J14

J18

J19

J20

J20

1

1a

1b

1bulk

1

2

3

1a

1b

1 bulk

1

1b

1

3

1

TAPP

Cpx

MgFeAl

Garnetf

Garnet

Garnet

Garnet

CaTi-pv

CaSi-pv

CaTiSi-pvf

CaSi-pv

FeO

Kyanite

Olivine

Carbonate

3

1

1

1

2

5

3

1

1

1

2

1

1

3

1

35.17

47.88

35.88

41.88

42.31

41.59

40.94

1.67

47.89

24.78

51.91

2.05

32.21

40.34

0.59

4.09

0.08

0.36

0.22

1.06

0.95

1.20

54.51

3.02

28.77

0.24g

0.00

0.15

0.00

0.01

19.92

13.39

31.05

22.22

21.45

21.10

21.36

0.13

0.37

0.25

0.06

0.33

66.54

0.01

0.09

0.03

0.02

0.04

0.03

0.07

0.06

0.06

0.01

0.00

0.01

0.00

0.16

0.00

0.02

0.02

23.10

12.75

20.53

16.64

14.90

14.86

13.54

0.12

0.44

0.28

0.08

97.48

0.33

9.26

0.87

0.49

0.42

0.40

0.41

0.30

0.28

0.23

0.00

0.04

0.02

0.02

0.53

0.02

0.17

0.84

15.91

7.12

11.78

9.45

10.05

10.58

9.51

0.01

0.26

0.14

0.03

0.00

0.55

49.12

23.71

0.05

16.14

0.28

8.21

9.48

8.68

11.57

39.78

44.87

42.33

48.29

0.02

0.00

0.14

27.71

0.05

1.92

0.04

0.98

0.85

0.92

0.78

0.47

0.16

0.32

0.00

0.14

0.20

0.02

3.21

0.00

0.05

0.54

0.30

0.01

0.02

0.00

0.00

0.06

0.01

0.02

0.01

0.18

0.00

0.17

0.01

98.81

99.72

100.37

100.04

100.48

99.020

99.190

96.75

97.59

97.17

100.41

100.91

100.00

99.25

57.12

2.718

1.792

2.647

3.112

3.117

3.107

3.064

0.039

0.956

0.533

0.997

0.024

0.877

0.995

0.019

0.238

0.002

0.020

0.012

0.059

0.053

0.068

0.955

0.045

0.465

0.004

0.000

0.003

0.000

0.000

1.814

0.591

2.700

1.946

1.862

1.858

1.884

0.004

0.009

0.006

0.001

0.005

2.136

0.000

0.003

0.002

0.000

0.003

0.002

0.004

0.004

0.004

0.000

0.000

0.000

0.000

0.001

0.000

0.000

0.001

1.493

0.399

1.267

1.034

0.918

0.929

0.847

0.002

0.007

0.005

0.001

0.943

0.008

0.191

0.023

0.032

0.013

0.025

0.026

0.019

0.018

0.015

0.000

0.001

0.000

0.000

0.005

0.000

0.004

0.023

1.833

0.397

1.295

1.047

1.104

1.178

1.061

0.000

0.008

0.004

0.001

0.000

0.022

1.807

1.127

0.004

0.647

0.022

0.653

0.748

0.695

0.928

0.993

0.959

0.975

0.994

0.000

0.000

0.004

0.947

0.007

0.139

0.005

0.141

0.122

0.133

0.113

0.021

0.006

0.013

0.000

0.003

0.011

0.001

0.199

0.000

0.001

0.014

0.008

0.000

0.000

0.000

0.000

0.002

0.000

0.000

0.000

0.003

0.000

0.007

1.867

8.140

3.981

7.984

7.973

7.952

7.975

7.982

2.016

2.005

2.010

1.999

0.985

3.057

3.009

4.212

12

6

12

12

12

12

12

3

3

3

3

1

5

4

6

0.551

0.499

0.506

0.503

0.546

0.559

0.556

0.129

0.513

0.462

0.364

0.000

0.747

0.904

0.980

0.002

0.619

0.017

0.384

0.404

0.371

0.466

1.000

0.992

0.996

0.999

1.000

0.000

0.002

0.456

Pv perovskite, sp spinel, CAS calcium aluminum silica phase, Fsp feldspar, Cpx clinopyroxene, TAPP tetragonal almandine-pyrope phase

aBulk calculated as 65:35 mixture

bBulk based on wide beam analysis encompassing the entire inclusion

cn number of analyses

d‘–’ below detection

eMg# Mg/(Mg + Fe), Ca# Ca/(Ca + Mg)

fBulk calculated as 50:50 mixture

gSIMS analysis

Table 3

Trace element contents of select mineral inclusions as determined by SIMS (ppm)

Diamond#

J1

J1

J2

J3

J3

J4

J8

J9

J10

J14

Inclusion#

1 bulk

3

2-1

1

2

1

1 bulk

1

1 bulk

1

Mineral

Ca(Ti, Si)-pv

Garnet

CAS

Cpx

Cpx

TAPP

Garnet

Garnet

Ca(Ti, Si)-pv

CaSi-pv

Rb

1.2

na

2.0

7.1

na

na

11.8

15.0

2.7

na

Ba

7.0

0.5

1.7

74.1

73.8

0.3

0.3

0.6

6.1

253

Th

85.2

1.5

0.2

9.8

11.0

0.01

0.1

0.1

10.8

0.5

U

63.9

0.8

0.4

27.4

27.8

0.02

0.1

0.2

27.8

111

Nb

353

0.1

0.1

0.2

0.5

21.9

0.0

2.8

243

7.6

La

1,629

10.2

1.4

0.6

3.2

0.04

0.0

0.2

185

134

Ce

9,670

126

1.4

0.4

1.3

0.09

0.0

6.0

142

219

Pr

1,055

25.9

0.4

1.5

6.1

0.02

0.0

2.2

131

25.7

Sr

351

12.6

263

0.4

1.1

0.2

0.3

5.0

59.2

8614

Nd

5,769

190

2.9

2.8

6.7

0.2

0.2

20.9

143

151

Zr

2,766

192

0.6

1.7

2.2

37.7

31.1

69.1

95.1

119

Hf

32.8

4.9

2.4

0.5

0.6

3.3

1.4

2.5

24.7

3.3

Sm

1,281

57.9

0.3

2.0

2.2

0.10

0.2

6.7

166

66.2

Eu

331

15.6

0.3

0.4

0.4

0.01

0.0

1.9

152

21.4

Gd

607

44.2

0.8

2.0

2.7

0.04

0.1

5.5

154

83.7

Tb

69.4

5.7

0.4

0.4

0.4

0.00

0.1

0.9

165

11.2

Dy

329

25.3

2.3

1.4

1.3

0.00

0.6

4.6

172

62.0

Y

973

78.6

5.1

0.2

0.2

0.06

9.6

20.9

117

260

Ho

46.8

3.9

0.2

0.8

0.9

0.01

0.2

0.9

140

9.3

Er

92.4

8.6

0.5

0.2

0.2

0.04

1.8

3.0

116

19.7

Tm

10.4

1.4

0.2

1.6

1.4

0.01

0.5

0.5

101

2.7

Yb

59.5

8.3

2.1

0.0

0.0

0.00

3.9

3.7

87.8

13.6

Lu

8.04

1.3

0.1

0.0

0.1

0.01

0.7

0.6

51.6

1.8

Pv perovskite, CAS calcium aluminum silica phase, Cpx clinopyroxene, TAPP tetragonal almandine-pyrope phase, na not analyzed

https://static-content.springer.com/image/art%3A10.1007%2Fs00410-010-0490-6/MediaObjects/410_2010_490_Fig4_HTML.jpg
Fig. 4

Backscattered electron images of Collier 4 diamond inclusions: a, b two-phase inclusions of CaTiO3 + CaSiO3 in diamonds J1 and J10, c, d Majoritic garnet inclusions in diamonds J1 and J9, e clinopyroxene + MgFeAl-silicate in diamond J8 (former garnet), f TAPP-phase inclusion in diamond J4, g CaAlSi-phase + kyanite (former CAS phase) inclusion in diamond J2, h native iron in diamond J2, i SiO2 + kyanite in diamond J19, j Clinopyroxene in diamond J3, k primary pyrrhotite inclusion in diamond J6, l secondary Fe–Cu–sulfide + Ca–Fe–Mg–Al–silicate inclusion in diamond J8

Ca-rich perovskites

Calcium-rich inclusions with a perovskite stoichiometry (ABO3) occur in three diamonds. In diamond J14 three inclusions of CaSiO3-perovskite (CaSi-Pv) were recovered (together with a SiO2 inclusion), whereas composite inclusions of CaTiO3 + CaSiO3 (hereafter, CaTiSi-Pv) were found in diamonds J1 (core zone) and J10 (intermediate zone) (Fig. 4a, b). The two-phase composite inclusions are similar to those described in Juina alluvial diamonds, which previous workers explained as retrograde exsolution products from a single-phase CaTiSi-Pv originating in the transition zone or lower mantle (Brenker et al. 2005, 2007; Hayman et al. 2005; Kaminsky et al. 2001). Walter et al. (2008) described the composite inclusions in J1 and J10, and also interpreted them as retrograde exsolution products of a former single-phase perovskite which originated at a depth constrained by phase relations to be between about 300 and 700 km.

The primitive mantle-normalized trace element compositions of CaTiSi-Pv inclusions in J1 and J10, and a CaSi-Pv inclusion in J14, are shown in Fig. 5a. Walter et al. (2008) reported on the trace element composition of the J1 inclusion and noted the overall enrichments in incompatible elements as a group, especially the extreme enrichment in the REE and the high field strength elements (HFSE). Incompatible element abundances are elevated in J10 as well, although considerably less so for the light-REE (LREE) than in J1. The CaSi-Pv in J14 diamond is also enriched in the REE but shows conspicuous relative depletions in Nb, Zr, Hf and Ti. Brenker et al. (2007) also report highly elevated REE in CaSi- and CaTiSi-Pv inclusions in Juina diamonds. Overall, the trace element abundances in the J10 and J14 perovskites are similar to those previously reported for CaSi-Pv inclusions in diamonds worldwide (Stachel et al. 2005).
https://static-content.springer.com/image/art%3A10.1007%2Fs00410-010-0490-6/MediaObjects/410_2010_490_Fig5_HTML.gif
Fig. 5

Trace element abundances of the mineral inclusions (Table 3) normalized to primitive mantle (McDonough and Sun 1995): a Bulk Ca(Ti,Si)O3 inclusions in diamond J1 and J10 and CaSiO3 inclusion in J14, b majoritic garnet inclusions in diamonds J1, J8 and J9, c CaAlSi-phase in diamond J2 and TAPP phase in diamond J4, d clinopyroxenes in diamond J3

Garnet

High-Ca, -Ti, and -Na majoritic garnets occur in three diamonds (Fig. 4c, d): as a single inclusion found in the rim zone of diamond J1, as multiple inclusions (3) in diamond J9, and as a partially altered inclusion in diamond J2. The J2 inclusion is associated with a crack that reaches the diamond surface, and about two-thirds of the garnet is altered to a serpentine-like secondary phase. The preserved, fresh portion of the inclusion deeper in the diamond, however, yields a good garnet stoichiometry with a composition very similar to the inclusions in diamonds J1 and J9 (Table 2). All these garnets are of a mafic ‘eclogitic’ affinity as demonstrated by their Ca and Cr contents (Fig. 6) as well as their relatively low Mg#s (Aulbach et al. 2002; Stachel et al. 2005). The majoritic component in these garnets is relatively low. Pressure determined from the systematics of Al + Cr and Si per atomic formula unit (Harte and Cayzer 2007) shows that the inclusions last equilibrated within the range of ~6–11 GPa (Fig. 6). The three garnets in diamond J9 have remarkable differences in their Ca-contents, with Ca#s ranging from 0.37 to 0.47 (Fig. 6). Such changes in chemistry within a single diamond, and in the case of garnets 2 and 3 (Table 2) within the same growth zone, are indicative of large variations in the chemistry of the inclusion growth medium.
https://static-content.springer.com/image/art%3A10.1007%2Fs00410-010-0490-6/MediaObjects/410_2010_490_Fig6_HTML.gif
Fig. 6

Plots showing geobarometry and compositional characteristics of majoritic garnet inclusions in Collier 4 diamonds. a Diagram showing Al + Cr and Si per formula unit (12 oxygens) used as a geobarometer (after Harte and Cayzer 2007). Garnet inclusions indicate pressures of ~6–11 GPa. b CaO versus Cr2O3 plot showing well-defined eclogitic character of the garnet inclusions (fields after Aulbach et al. 2002)

An unusual two-phase composite inclusion was identified in stone J8 (Fig. 4e), with one phase having a clinopyroxene composition and the other a MgFeAl-silicate phase (Table 2). The clinopyroxene is calcium-rich (~45% wollastonite) and has ~17 mol% jadeite component, features signifying a high-pressure origin in a Ca-rich protolith. The Al and Si contents are similar to those in clinopyroxenes in high-aluminous eclogite xenoliths from kimberlites (Spetsius and Serenko 1990), but the Fe content is higher by an order of magnitude. The contents of SiO2 and FeO in the MgFeAl-silicate phase are similar to those found in the tetragonal almandine-pyrope phase (TAPP) in diamond J4 described as “Iron-rich TAPP phase” (Table 2). If the composite inclusion represents unmixing of a single mineral, then on the basis of the exposed proportion of the phases (~50:50) we calculate a bulk composition similar to majoritic garnet inclusions in diamonds J1, J2 and J9 (Table 2). The majorite content of the calculated composition indicates an equilibration pressure of ~6–11 GPa for this inclusion (Fig. 6).

The primitive mantle-normalized trace element compositions of majoritic garnet inclusions in diamonds J1, J8 (bulk) and J9 are shown in Fig. 5b. The trace element chemistry of the J9 majoritic garnet is similar to those summarized by (Stachel et al. 2001). Like the CaTiSi-Pv in the same diamond, the majoritic garnet in J1 is exceptional, containing extreme enrichments in incompatible elements—the highest concentration of REE we are aware of for a garnet inclusion. The composite inclusion in J8 has a trace element abundance pattern resembling that in the TAPP inclusion described below (Fig. 5c).

Iron-rich TAPP phase

An almandine-pyrope-like phase occurs in diamond J4 (Fig. 4f). The inclusion has low SiO2 (35.17 wt.%) and elevated TiO2 (4.09 wt.%) (Table 2). The crystalline structure of the J4 inclusion is not yet determined, but its strongly elevated Ti and paucity of Ca and Na makes it abnormal for an upper mantle garnet. The inclusion in J4 is iron rich (Mg# ~55), but has a stoichiometry closely resembling TAPP. The high Ti content in the inclusion is very similar to one of the TAPP phases reported in alluvial diamonds from Sao Luiz, Brazil (Harris et al. 1997), and to phases found in Rio Soriso alluvial diamonds (Hayman et al. 2005; Kaminsky et al. 2001).

The trace element abundance pattern in the TAPP phase is unique (Table 3), as shown in Fig. 5c. Incompatible elements are uniformly depleted relative to primitive mantle, with the notable exception of the HFSE, which as a group are enriched relative to other incompatible elements by about two orders of magnitude.

Olivine

An inclusion with an olivine stoichiometry occurs in diamond J20 (Table 2). The ‘olivine’ has an Mg# of 90, is low in Ni and Al, but has elevated Na, Ca, and Mn relative to typical upper mantle peridotitic olivine (Fig. 7), features shared by some other ‘olivine’ compositions in sublithospheric diamonds (Brey et al. 2004).
https://static-content.springer.com/image/art%3A10.1007%2Fs00410-010-0490-6/MediaObjects/410_2010_490_Fig7_HTML.gif
Fig. 7

NiO (a), Na2O (b) and CaO (c) versus Mg# for olivine inclusions in worldwide diamonds (after Brey et al. 2004) and from diamond J20 (this study)

CaAlSi-phase + kyanite, KAlSi-phase, MgFeAl-spinel, and native iron

In addition to the garnet described above, diamond J2 contains a unique assemblage of other mineral phases (Tables 1, 2) that comprises an unusual composite inclusion composed of a calcium aluminum silicate phase (CaAlSi) plus kyanite (Fig. 4g), an inclusion with a K-feldspar-like stoichiometry (KAlSi), an MgFeAl-spinel, and native iron (Fig. 4h). Like the garnet inclusion in this diamond the spinel has an uncertain origin as it is adjacent to cracks that reach the diamond surface. However, we note that the spinel in J2 has a composition unlike those found in kimberlites.

The CaAlSi-phase of the composite inclusion has a composition reminiscent of zoisite and lawsonite, but is richer in alumina and poorer in silica than both of these. We analyzed the bulk inclusion using a microprobe spot size that encompassed the composite inclusion and found a bulk composition that is very similar to the synthetic CAS-phase found experimentally to exist at a pressure of ~16–25 GPa (Hirose et al. 2004; Irifune et al. 1994). Thus, the composite inclusion could represent unmixing of a CAS-phase during decompression. If so, this is the first report of a naturally occurring CAS-phase we are aware of.

The KAlSi inclusion is silica deficient relative to K-feldspar, although a low microprobe total was obtained due to the very small inclusion size and surface irregularities (Table 2). Based on the presence of the high-pressure CAS-phase in the same diamond, we suppose that the KAlSi inclusion represents a former K-hollandite phase stable at ~16–30 GPa pressure (Hirose et al. 2004; Irifune et al. 1994).

The trace element composition of the CaAlSi-phase is shown in Fig. 5c. The overall abundance pattern is unfractionated relative to primitive mantle, with the exception of enrichment in U and Sr, and depletion in Zr.

SiO2 ± kyanite

Inclusions of SiO2 are found in five Collier 4 diamonds, either as single phases or together with sulfides or Ca-perovskite (J14). The single 20 μm inclusion of SiO2 in diamond J19 (Fig. 4i) contains a microblock of Al-rich silicate approaching a kyanite composition (Table 2). SiO2 inclusions in Collier 4 diamonds clearly have a primary origin and represent former high-pressure polymorphs, either coesite or stishovite.

Clinopyroxene

Diamond J3, which is one of only three stones in the Collier 4 collection that is zoned in N content and aggregation (Table 1), contains multiple clinopyroxene inclusions, and the compositions of four of these are given in Table 2. The inclusions exposed at the surface by polishing (Fig. 4j) and inside the J3 crystal have imposed diamond morphology and no cracks leading to the surface, hence they are regarded as syngenetic in origin. The clinopyroxenes are aluminian sodian augites. The J3 clinopyroxenes are subcalcic relative to most mantle samples, which implies a high temperature origin. They are distinct from peridotitic clinopyroxenes in their low Mg# and low Cr contents (<0.3%). Compared to clinopyroxene from eclogite xenoliths they have lower Ca# and are low in Na but high in Ti. The inclusions appear to be akin to some clinopyroxene megacrysts found in kimberlites with their relatively high Ti contents (Ramsay 1992), but are distinctly enriched in iron. Overall, the J3 pyroxenes have chemistry atypical of those in peridotitic lithospheric mantle. They are also distinct from clinopyroxene inclusions found in Sao Luiz alluvial diamonds that, based on their association with other mineral inclusions such as majorite garnet and Mg-perovskite, have been interpreted to be sublithospheric, originating in the transition zone or lower mantle (Hutchison 1997).

The average trace element abundances of two clinopyroxene inclusions from J3 are shown on Fig. 5d. The clinopyroxene inclusions display only slight overall element enrichments relative to primitive mantle, and are generally unfractionated. Overall the concentration of REE in clinopyroxene inclusions are similar to those of eclogitic clinopyroxenes, although they are relatively enriched in HREE (Stachel et al. 2004).

Low Ni sulfides and iron phases

Inclusions of low-Ni sulfides are abundant, occurring in seven diamonds (Table 1, Fig. 4k). Four of these are primary pyrrhotites with compositions in the range of 57–60% Fe, 0.3–3% Ni, 0.2–5% Cu and 0.15–0.3% Co (at.%). They are single inclusions or associated in the same diamonds with CaTiSi-perovskite, garnet and SiO2. The sulfides’ association with sublithospheric mineral inclusions indicates that they are part of this paragenesis. The remaining three sulfide inclusions are altered (Fig. 4l) and have the composition of chalcopyrite. Native iron, FeO and magnetite are also found as inclusions (Table 1).

Ca–Fe–Mg-carbonates

Ca–Mg–Fe-carbonates occur mainly as micro-inclusions, but in several diamonds they are large enough for EMP analysis. In diamond J20, which also contains an olivine inclusion, the carbonate has a composition similar to dolomite but with 3.2 wt% Na2O (Table 2). The primary nature of carbonates in sublithospheric Juina alluvial diamonds was demonstrated by Brenker et al. (2007).

Carbon isotope compositions

The Collier 4 diamonds display a wide range of carbon isotopic compositions (δ13C) from ~−5 to −25‰, a spread similar to that reported by Kaminsky et al. (2009) in their Collier 4 collection, and encompassing the major portion of the range observed in the worldwide diamond population and in lherzolite and eclogite xenoliths (~+5 to −40‰) (Cartigny et al. 1998b; Deines 2002). We note further that δ13C measurements in core–rim traverses within some individual crystals varied substantially, indicating multi-stage growth histories.

On the basis of the carbon isotopic compositions of the crystal cores, the diamonds are separated into three distinct groups: a heavy Group 1 (−5 to −10‰), an intermediate Group 2 (−13 to −15‰) and a light Group 3 (−23 to −25‰) (Fig. 8). These groupings may be arbitrary and only a function of the small population of diamonds studied, but the end members defined by Groups 1 and 3 mimic the well-established bi-modality of carbon isotope composition in diamonds, mantle xenoliths and basalts with peaks near ~−5 and −20‰ (Deines 2002). We also note that Kaminsky et al. (2009) reported carbon isotope compositions for twenty diamonds from Collier 4, and of these we would classify ~50% with our Group I, and ~25% each with our Group II and Group III, proportions similar to what we observe in our collection. Finally, these isotopic groups apparently characterize genetically related sets of mineral inclusions, especially in the cases of Groups 2 and 3. For these reasons we will adhere to these groupings in further discussion of the origin of the diamonds and their inclusions.
https://static-content.springer.com/image/art%3A10.1007%2Fs00410-010-0490-6/MediaObjects/410_2010_490_Fig8_HTML.gif
Fig. 8

δ13C (‰ relative to PDB) in Collier 4 diamonds. Shown on the left are observed ranges for various potential carbon sources (carbonate, mantle and organic) (Cartigny 2005; Deines 2002; Schidlowski 1988), a schematic showing the δ13C distribution of worldwide diamonds, and ranges for eclogitic and peridotitic diamonds determined previously (Araujo et al. 2003; Hutchison et al. 1999; Kaminsky et al. 2001; Stachel et al. 2005). Collier 4 diamond δ13C compositions are shown according to their general position within the diamond zonation. Shown on the right are Collier 4 diamond groups based on core δ13C compositions

Heavy Group 1

Group 1 includes diamonds J3, J4, J6, J8, J14 and J20. The carbon isotopic compositions of the diamonds in this group range from ~−5 to −10‰. Core to rim isotopic variations are generally small, not larger than about 4‰, with both increases and decreases observed in δ13C. In comparison with the global diamond population, Group 1 diamonds have carbon isotopic compositions generally akin to ‘mantle’ carbon as characterized by a predominance of compositions centered at δ13C ~−5‰ (Cartigny et al. 1998b). The silicate inclusions in this group are ‘olivine’, CaSi-Pv, Ca-rich majorite garnet, clinopyroxene, and the TAPP phase. Three Collier 4 diamonds containing ferropericlase inclusions reported by Kaminsky et al. (2009) have isotopic composition between ~−2 and −8‰, and we classify these with our Group I as well.

Intermediate Group 2

Included in this group are diamonds J1, J9, and J10. The core δ13C compositions range from about −14 to −15‰ (Fig. 8). Diamonds J1 and J10 become slightly heavier in δ13C in their rims (~−12.7‰). In contrast, J9 becomes progressively heavier from core (−13.3‰) to rim (−7.6‰) in a single traverse (Figs. 3e, 8), covering the entire range of δ13C of Juina alluvial diamonds containing majorite inclusions summarized by Stachel et al. (2005). The inclusions in this group are CaTiSi-Pv (J1 and J10) and Ca-rich majoritic garnets (J1 and J9). Intermediate carbon isotope values have been reported for diamonds from several localities (Cartigny 2005). Kaminsky et al. (2001) also reported a CaTiSi-Pv inclusion-bearing alluvial diamond from the Juina area with a δ13C composition similar to diamonds in this group (~−11‰).

Light Group 3

Included in this group are diamonds J2, J5, J18 and J19. The cores of these crystals are characterized by very light δ13C values (~−23 to −25‰, Fig. 8). However, it is notable that the rim zones in diamonds J2 and J5 have considerably heavier δ13C values (~−18 to −15‰). The silicate mineral inclusions in this group, including eclogitic garnet, CaAlSi-phase, kyanite, KAlSi-phase (former hollandite) and SiO2, are unique in their aluminous and siliceous nature; low-Ni sulfides, native iron, and magnetite are also present. Kaminsky et al. (2009) report on two rutile inclusions in a Collier 4 diamond with a δ13C values of ~−23‰, and we classify this diamond with our Group 3.

U–Pb dating of CaTiSi-Pv

A 206Pb/238U age of 101 ± 7 Ma has been obtained for the CaTiSi-Pv inclusion in diamond J1. This age is ostensibly about 8 Ma older than the age of the kimberlite eruption (93.1 ± 1.5 Ma), but the ages are equivalent within the uncertainty. The 207Pb/206Pb data do not give any constraints other than that the sample is of Phanerozoic age. This young age is unique for a diamond inclusion, and represents the first direct dating of a sublithospheric diamond.

Discussion

In discussing the possible origin of Collier 4 diamonds and their inclusions we refer again to the three groups defined above.

Heavy Group 1

The ‘olivine’ inclusion in diamond J20 contains only 0.03 wt% Al2O3, which is suggestive of an upper mantle origin because higher-pressure polymorphs (wadsleyite and ringwoodite) should contain >0.3 wt% Al2O3 in normal peridotitic mantle compositions (Akaogi and Akimoto 1979). However, as shown in Fig. 7, the J20 ‘olivine’ differs from nearly all olivine inclusions in diamonds that are ascribed to an upper mantle peridotitic paragenesis (Brey et al. 2004); J20 ‘olivine’ is depleted in Ni, but is unusually enriched in Na, Ca and Mn (Table 2) relative to such inclusions. J20 ‘olivine’ is also distinct from a separate group of olivine + ferropericlase inclusions that, according to Brey et al. (2004), originated in the upper mantle (Fig. 7). But J20 ‘olivine’ does have a composition similar to some distinctive inclusions shown on Fig. 7 that are interpreted to have originated in the transition zone or lower mantle. Several of these inclusions are from Juina alluvial diamonds and are associated with CaSi-perovskite (Kaminsky et al. 2001) or ferropericlase, MgSi-perovskite and TAPP (Hayman et al. 2005). The ‘olivine’ in diamond J20 is accompanied by a micro-inclusion of a Ca-rich carbonate phase (Table 2). Brenker et al. (2007) also report on a suite of Juina diamonds with a number of Ca-rich silicate and carbonate inclusions, and olivine occurs in some of these diamonds as well. These authors suggest the inclusions originated in the transition zone. A possible explanation for the origin of J20 ‘olivine’ is that the very high Ca and Na contents indicate the involvement of a low-degree, carbonated partial melt. This could plausibly explain a low Ni content as well if the melt reacted with dunite to form ferropericlase (e.g. Brey et al. 2004). Although we have yet to identify ferropericlase in our Collier 4 collection, Kaminsky et al. (2009) found ferropericlase inclusions in three isotopically heavy Collier 4 diamonds (~−8 to −2‰), which we classify above with our heavy Group 1 diamonds.

The TAPP phase in diamond J4 is Ti- and Fe-rich, indicating a mafic rather than ultramafic protolith. The inclusion occurs in isolation from other mineral phases so we can only estimate a depth of origin by analogy with other TAPP occurrences. The phase relations of TAPP are not experimentally determined, but the association of Mg-rich TAPP with CaSi-Pv, ferropericlase and either former wadsleyite or MgSi-Pv in Juina alluvial diamonds has been taken as evidence for an origin near the transition zone—lower mantle boundary (Harris et al. 1997; Hayman et al. 2005; Hutchison et al. 2001). A possible origin for this phase as a product of a complex retrograde P–T path upon diamond exhumation has also been made on crystallographic grounds (Finger and Conrad 2000). The composite inclusion with a majoritic garnet composition in diamond J8 is also consistent with an origin related to a mafic protolith on the basis of its high Ca and low Cr contents.

The CaSi-Pv in J14 diamond is akin to other perovskite inclusions found in Juina alluvial diamonds that are commonly ascribed to an origin in the deep transition zone or lower mantle in either mafic or ultramafic lithologies (Hayman et al. 2005; Kaminsky et al. 2001); a second inclusion in J14 is SiO2, possibly indicating a mafic lithology in this case. In our collection of Collier 4 diamonds we have not identified any inclusions with MgSiO3-perovskite stoichiometry, although this phase is found in suites of Juina alluvial diamonds (Hayman et al. 2005; Kaminsky et al. 2001). The CaSi-Pv in J14 is enriched in Ba, U, and Sr, and contains LREE elevated ~100 times relative to primitive mantle (Fig. 5a). Wang et al. (2000) noted that such enrichments in CaSi-Pv inclusions would require mantle protoliths with absurdly elevated trace element abundances, and postulated instead a role for low-degree carbonatitic melts in their origin. Likewise, Walter et al. (2008) showed that CaTiSi-Pv inclusions in diamonds J1 and J10 likely crystallized from low-degree carbonated partial melts (see below). We suggest that the CaSi-Pv in J14 may have an origin related to crystallization from a low-degree, Ca-rich but Ti-poor carbonated melt, deep in the transition zone.

The clinopyroxene inclusion in J3 is unique in that it is unlike typical lithospheric clinopyroxene inclusions in diamonds. If the J3 diamond has a young age like that determined for J1 (~100 Ma), then a temperature of ~1,400–1,500°C is implied for diamond formation based on its N aggregation state. On the basis of the Cr content in clinopyroxene barometer (Nimis and Taylor 2000), which may be inappropriate as it assumes equilibration with pyrope garnet, the J3 clinopyroxene yields pressures of ~5–6 GPa at these temperatures, similar to pressures obtained for some of the Collier 4 garnets. This would place the origin of J3 beneath the ~140-km-deep Juina lithosphere (Feng et al. 2007).

As an assemblage the minerals in Group 1 indicate asthenospheric to transition zone depths of origin. In general, the inclusion compositions are inconsistent with an origin in normal peridotitic mantle, but have characteristics indicating a role for subducted mafic lithologies and/or the involvement of low-degree, carbonated partial melt. The pronounced affinity to subducted oceanic crustal lithologies exhibited by the Groups 2 and 3 diamonds as discussed below, is suggestive that the Group 1 diamonds might have an origin related to mafic/ultramafic portions of subducted lithosphere. The heavy isotopic composition is consistent with a mantle source of carbon (Fig. 8).

Intermediate Group 2

CaTiSi-Pv inclusions in diamonds J1 and J10 contain trace elements that are so enriched relative to primitive mantle that, like other Ca-Pv inclusions in diamond, a subsolidus origin is virtually precluded (Walter et al. 2008; Wang et al. 2000). The similarly elevated REE abundances in Ca-Pv and carbonate inclusions in Juina diamonds reported by Brenker et al. (2007) suggest that these inclusions also crystallized from low-degree melts.

Walter et al. (2008) have shown that a liquid coexisting with the CaTiSi-Pv inclusion in diamond J1 has characteristics of a primary carbonate-rich melt derived from subducted oceanic crust; it is depleted in LILE, but not depleted in REE or Ti, Nb and Zr, as shown on Fig. 9a. Also shown is a calculated coexisting liquid for the CaTiSi-Pv inclusion in J10. In comparison to the J1 calculated liquid, the J10 liquid is conspicuously depleted in U and Th but not in Rb, Ba and Sr, and is depleted in the REE as a group, especially the light- to mid-REE. However, Ti, Nb, and Zr abundances in the J10 liquid are similar to those in the J1 liquid, yielding an apparent enrichment in these elements relative to the REE.
https://static-content.springer.com/image/art%3A10.1007%2Fs00410-010-0490-6/MediaObjects/410_2010_490_Fig9_HTML.gif
Fig. 9

a Calculated trace element composition of melts that could coexist with perovskite inclusions in diamonds J1, J10 and J14. b Model fractionated liquid compared to the melt calculated to coexist with the perovskite inclusion in J10. The fractionated liquid is calculated assuming 40% fractional crystallization of CaTiSi-perovskite from an initial liquid with the composition of the calculated liquid coexisting with the CaTiSi inclusion in diamond J1. Mineral/melt partition coefficients used in the calculations are as provided in Supplemental Information in Walter et al. (2008) and are based on the experiments of Corgne and Wood (2002)

The difference between the J1 and J10 calculated liquids is plausibly explained by crystal-liquid fractionation in the deep mantle. CaTiSi-Pv is a liquidus phase in Ca- and Ti-rich carbonated melts (Walter et al. 2008), and this phase has an affinity for U, Th and the REE, but not for Rb, Ba, or Sr (Corgne and Wood 2002). We developed a simple fractional crystallization model assuming a starting liquid composition with trace element characteristics of the J1 liquid, and then removed CaTiSi-Pv assuming perfect fractional crystallization. The resultant liquid after 40% crystallization is shown on Fig. 9b, and given the first-order nature of the model it corresponds remarkably well with the J10 calculated liquid. On the basis of this model we suggest that the CaTiSi-Pv inclusions in diamonds J1 and J10 record crystallization from near-primary and differentiated carbonated melts, respectively. These melts were derived ultimately from subducted oceanic crust in the deep upper mantle or transition zone.

Walter et al. (2008) also calculated coexisiting partial melts for the majoritic-garnet inclusions in diamonds J1 and J9, and noted the distinct fractionation of LREE from HREE (LREE/HREE ~102 to 103) and the large depletions of the HFSE in these calculated liquids, and suggested they might represent the products of fractionation of phases like majorite and ilmenite from low-degree melts. We note that Kaminsky et al. (2009) report a picroilmenite inclusion in a diamond from the Pandrea 6 kimberlite from Juina, which showed a marked enrichment in Zr, Nb, Hf and Ti. Thus, on the basis of the geochemistry of Group 2 inclusions, we suggest that carbonated melts in the transition zone or deep upper mantle differentiate, possibly to more siliceous compositions, as they migrate, crystallize and interact with their host rocks.

The diamonds in Group 2 are defined by intermediate carbon isotope values (Fig. 8). Given that subducted protoliths are implicit in the origin of these diamonds via partial melting, it is tempting to attribute the intermediate isotope values to mixing of isotopically light, subducted biogenic carbon with heavy mantle or subducted carbonate carbon (Fig. 8).

Light Group 3

The diamonds in this group all have cores with very light carbon (δ13C ~−24‰). The important observation for Group 3 diamonds is that the inclusions have compositional characteristics suggesting an origin related to subducted pelagic sediments. On the basis of the experiments of Irifune et al. (1994), the overall mineral inclusion assemblage in the Group, and specifically in diamond J2, of possible CAS-phase, kyanite, KAlSi-phase, garnet, and SiO2 has a striking resemblance to the phase assemblage expected in deeply subducted metasediments (Poli and Schmidt 2002). From the phase relations depicted in Fig. 10, the CAS-phase would have originated in the transition zone at pressures greater than ~15 GPa. Upon exhumation to a lower pressure, ultimately about 7 GPa based on the garnet inclusion in diamond J2, the CAS-phase unmixed to a CaAlSi composition plus kyanite. We would also classify an isotopically light Collier 4 diamond with rutile reported by Kaminsky et al. (2009) in our light Group 3, and recent experiments show rutile stability in subducted sediments to at least 5 GPa (Auzanneau et al. 2006; Hermann and Rubatto 2009).
https://static-content.springer.com/image/art%3A10.1007%2Fs00410-010-0490-6/MediaObjects/410_2010_490_Fig10_HTML.gif
Fig. 10

Phase proportions as a function of pressure for pelagic sediment (after Irifune et al. 1994)

The origin of δ13C variations in Collier 4 diamonds

The origin of carbon isotopic variation in diamonds is controversial and several contrasting interpretations are available, including (1) primordial carbon isotopic heterogeneity preserved in the mantle (Deines 1980; Haggerty 1999; Kirkley et al. 1991), (2) isotopic fractionation (Cartigny et al. 1998a, b, 2001; Deines and Harris 1995; Deines et al. 2001; Javoy et al. 1986; Maruoka et al. 2004; Mysen et al. 2009; Thomassot et al. 2007), and (3) carbon isotopic heterogeneity related to subduction of biogenic carbon (Aulbach et al. 2002; Fitzsimons et al. 1999; Frank 1969; Kirkley et al. 1991; Mohapatra and Honda 2006; Schulze et al. 2004; Tappert et al. 2005a, b, 2009).

Which of these models can best explain the origin of Collier 4 diamonds and their inclusions? Considering both our results and those of Kaminsky et al. (2009), Collier 4 diamonds show an isotopic spread that effectively spans the worldwide diamond distribution, as shown on Fig. 8. Figure 8 also shows that several of the diamonds have apparent core to rim variations in δ13C of up to 8‰. Isotopic increases of up to ~3‰ have been identified as a result of fractionation accompanying diamond crystallization from a carbonated melt or fluid (Thomassot et al. 2007), and this process may account for some of the isotopic variation in Collier 4 diamonds. However, all the inclusions showing large isotopic variations (e.g. ≥5‰) also have chemical characteristics indicative of an origin involving subducted materials; diamond J9 with isotopic variation of ~5‰ is interpreted to have crystallized from carbonated melts derived from subducted oceanic crust, and diamonds J2 and J5 with isotopic variation of ~5 and ~8‰, respectively, have the lightest δ13C and an origin ostensibly related to pelagic sediments.

Our observations could be interpreted as supporting the subducted biogenic carbon model for the origin of Collier 4 diamonds. But subducted mafic crust is also implicit in the fractionation model of Cartigny et al. (1998b) where the source of carbon is not biogenic, but instead large isotopic fractionations are a result of CO2-loss from carbonated melt or fluid. Eclogite, unlike peridotite, is a lithology within which CO2 can coexist with carbonate under certain conditions in the upper mantle (Luth 1996, 1999), so that evidence for subducted eclogitic lithologies alone is not sufficient to differentiate between the models. However, we find it compelling that diamond J2, which contains the clearest evidence for a subducted oceanic sediment protolith, also has exceptionally light carbon (−24‰) in its core. Further support for this interpretation comes from a rutile-bearing diamond from Collier 4 that also has exceptionally light carbon (−24‰) (Kaminsky et al. 2009).

A conceptual petrogenetic model for Collier 4 diamonds and inclusions

The Collier 4 diamonds and their inclusions exhibit a range of physical, chemical, and isotopic characteristics that when viewed holistically provide the basis for a conceptual petrogenetic model. Collier 4 diamonds bear all the ‘trade marks’ of a sublithospheric origin: nearly all the mineral inclusions have chemical characteristics implicating an origin related to subducted protoliths, and several inclusions bear a clear imprint of equilibration with low-degree partial melts. The young age of diamond J1 (101 ± 7 Ma) suggests a connection between diamond generation and the origin of associated Cretaceous alkaline rocks (Fig. 1), which has been linked causatively with either the Trindade plume track (Gibson et al. 1995; Gibson et al. 1999) or to conduction of heat into the lithosphere from a fossil plume head associated with the Tristan da Cunha plume (VanDecar et al. 1995); in either case Cretaceous magmatism is attributed to a mantle plume. We present a speculative conceptual model, a schematic of which is shown in Fig. 11, in which we propose that it is the plume event that initiates melting of carbonated subducted materials in the deep upper mantle and transition zone, provides the upward transportation of diamonds, and ultimately leads to the generation of the kimberlitic melts that exhume the diamonds from below the base of the Amazonian lithosphere.
https://static-content.springer.com/image/art%3A10.1007%2Fs00410-010-0490-6/MediaObjects/410_2010_490_Fig11_HTML.gif
Fig. 11

A conceptual petrogenetic model for the origin of Collier diamonds and their inclusions

Our model requires subduction of oceanic lithosphere beneath Juina, and two possibilities are: (1) subduction along the south-southwest margin of Gondwanaland related to the early Paleozoic to Permo-Triassic Terra Australis Orogen (Cawood 2005; Tappert et al. 2009), and (2) subduction along the southern margin of the Amazon craton during formation of the Mesoproterozoic Rio Negra-Juruena Mobile Belt during magmatic arc formation and accretion (de Almeida et al. 2000). Mafic eclogite xenoliths from the Collier 4 pipe dated at 1,166–1,648 Ma may relate to this event (Costa et al. 2003). Whichever the source, subducted materials must be stored in the transition zone until mobilized in the Cretaceous. Dynamic models suggest that slab penetration into the lower mantle is effectively inevitable, but that ponding of lithosphere at the transition zone—lower mantle boundary can occur depending on the physical details of subduction (e.g., subduction angle, slab temperature, trench dynamics) (Ganguly et al. 2009; Goes et al. 2008).

Diamond formation can conceivably occur within the slab during subduction, as has been suggested for other locations (Tappert et al. 2005a, b, 2009). However, the melt signatures recorded by several inclusions and the young age of the J1 CaTiSi-Pv inclusion provide the basis for relating diamond formation to the interplay between a mantle plume and subducted lithosphere. We note also that subducted carbon of a biogenic origin may be converted to carbonate in the downgoing slab as a consequence of the interaction with oxidized fluids produced by slab dehydration reactions (Poli et al. 2009), and carbonate can remain refractory with respect to slab fluids and melts into the deep mantle (Yaxley and Green 1994).

As shown on Fig. 11, we suppose that subducted slab materials that contain carbonate are subducted into the mantle transition zone (Brenker et al. 2005, 2007). The subducted slab stalls at the base of the transition zone due to the density inversion expected between oceanic lithosphere and peridotitic mantle, and impedance of subduction into a high viscosity lower mantle (Ganguly et al. 2009; Hirose et al. 1999; Ringwood 1991); deflection of some slabs at the base of the transition zone has been imaged by seismic tomography (Lay 1994; Shearer and Masters 1992). Oceanic crust reaches the transition zone (~400 km) at temperatures of ~1,000–1,200°C (Peacock 2003), and melting of carbonated, dehydrated mafic eclogite in the upper mantle or transition zone is unlikely even along the hottest slab geotherm (Dasgupta et al. 2006; Yaxley and Brey 2004). However, heating of carbonated subducted materials to ambient mantle temperatures could lead to melting at deep upper mantle and transition zone depths (Ballhaus et al. 1990; Dasgupta et al. 2006). The correlation between the timing of plume activity and the age of diamond formation (e.g. J1) suggests that heating by the Cretaceous plume initiates melting of carbonated, subducted materials in the transition zone and deep upper mantle.

Diamond crystallization from carbonated melt

Diamond can crystallize via a number of pathways, and here we consider two possibilities that are suggested by the mineral inclusions and the inferred presence of a carbonated melt: (1) carbonate reduction by sulfide, and (2) carbonate reduction by methane.

Low-Ni sulfides occur in several of our Collier 4 diamonds (Table 1) as well as one of the Collier 4 diamonds reported by Kaminsky et al. (2009), and are common in diamonds generally (Bulanova 1995). On the basis of experimental evidence, interaction of carbonate melt with iron sulfide can lead to carbonate reduction by a reaction in which the liberated oxygen dissolves into sulfide melt (Gunn and Luth 2006):
$$ {\text{MgCO}}_{{3({\text{melt}})}} = {\text{MgO}}\left( {\text{ferropericlase}} \right) + {\text{C}}_{{({\text{diamond}})}} + 2{\text{O}}_{2} \left( {{\text{in}}\;{\text{Fe}}{\text{--}}{\text{S}}{\text{--}}{\text{O}}\;{\text{melt}}} \right) $$
where ferropericlase is a by-product of the reaction. As mentioned above, although we have not identified ferropericlase in our Collier 4 collection, several of the diamonds reported on by Kaminsky et al. (2009) have ferropericlase inclusions, and ferropericlase is a common inclusion in Juina diamond suites generally (Harte and Harris 1994; Harte et al. 1999; Hayman et al. 2005).
Another possible diamond crystallization mechanism is reduction of carbonate by reducing fluids. In the deep upper mantle and transition zone, preferential incorporation of ferric iron into garnet generates reducing conditions, which may render methane-rich fluids or melts in the ambient mantle (Ballhaus et al. 1990; Frost and McCammon 2008; Luth 1993; Rohrbach et al. 2007; Woodland and Koch 2003). Carbonated melts from subducted materials may become exposed to these reducing conditions, possibly through mobilization of melts into surrounding mantle, or by invasion of subducted materials by mantle fluids. When carbonated melt and methane-rich fluids interact, the carbonate component can be reduced to crystallize diamond, with H2O as a by-product (Brey et al. 2004; Haggerty 1999; Pal’yanov et al. 1999, 2005; Yamaoka et al. 2002):
$$ {\text{CO}}_{{2({\text{carbonate\;melt}})}} + {\text{CH}}_{{4({\text{fluid,\;melt}})}} = 2{\text{H}}_{ 2} {\text{O}}_{{({\text{fluid,\;melt}})}} + 2{\text{C}}_{{({\text{diamond}})}} $$
H2O is itself highly soluble in carbonated melt (Keppler 2003), and the release of water to the partial melt would tend to increase the melt fraction and make a more siliceous melt (Dasgupta et al. 2006; Falloon and Green 1990; Gudfinnsson and Presnall 2005). This could promote further diamond crystallization and melt hydration through a reaction such as:
$$ {\text{CaCO}}_{{3({\text{melt}})}} + {\text{SiO}}_{{2({\text{melt}}/{\text{solid}})}} + {\text{CH}}_{{4\left( {{\text{fluid}}/{\text{melt}}} \right)}} = {\text{CaSiO}}_{{3({\text{melt}}/{\text{solid}})}} + 2{\text{H}}_{2} {\text{O}}_{{({\text{melt}})}} + 2{\text{C}}_{\text{diamond}} $$

The net effect of this process, in which the product H2O acts to promote further reaction until all the available methane is exhausted, is a melt that becomes decreasingly carbonated, but increasingly hydrated and siliceous.

Many of the Collier 4 diamonds show evidence of resorption and regrowth during complex growth histories (Fig. 3). Two majorite garnets in diamond J9 (2 and 3 in Table 2) occur in the same growth zone yet have demonstrably different chemistry (Ca# varying from ~0.37 to 0.47), indicating rapidly changing melt compositions with which the garnets equilibrated. The large carbon isotopic variations exhibited in several of the diamonds, as discussed above, may also reflect changing melt compositions, due either to isotopic fractionation as a consequence of carbonate reduction, or to a shift from a predominantly subducted carbon source to the progressive influence of mantle-derived carbon in the melts from which the diamonds crystallize.

Our observations indicate that in some cases diamonds formed originally in the deep upper mantle or transition zone and were later transported to shallower depths where inclusion re-equilibration occurred. Many of the diamonds apparently last resided at depths in the region of ~200 km before exhumation by the kimberlite near the base of the lithosphere below Juina at ~140 km (Feng et al. 2007; Heit et al. 2007). We suggest that upward transportation occurred when mantle and subducted materials were entrained in the upwelling plume during the Cretaceous, and were deposited for a relatively short time (<~10 Ma) near the base of the Amazonian lithosphere before being excavated by the kimberlite at ~93 Ma (Heaman et al. 1998). This conjecture is based on the U–Pb age of 101 ± 7 Ma and is also consistent with a short mantle residence time inferred from FTIR N aggregation characteristics in diamond J8. Harte and Cayzer (2007) also suggested a model of upward transport of diamonds beneath Juina based on composite garnet–clinopyroxene inclusions in alluvial diamonds from the Sao Luiz area, and concluded that transport from the transition zone to the base of the lithosphere would have occurred in a timeframe broadly compatible with rates of asthenospheric convection or plume flow.

A possible link between carbonatite and kimberlite melts

The barometric evidence (Table 1) indicates that many Collier 4 diamonds were exhumed by a kimberlite magma from below but close to the base of the Amazonian lithosphere (~140 km). The generation of the Collier 4 primary kimberlite melt must have occurred at this or greater depth. Some diamonds and their inclusions yield ample evidence for the involvement of low-degree carbonated melts in their origin, and trace element abundances in some kimberlites are remarkably similar to those in some carbonatites (Nelson et al. 1988), as well as to liquids that could have coexisted with majoritic garnet inclusions in diamonds from Collier 4 and other deep diamond locations (Keshav et al. 2006; Moore et al. 1991; Walter et al. 2008). This implies a link between carbonated melts released from subducted materials in the deep mantle or transition zone and primary kimberlite melt generation.

We speculate that carbonated melts from subducted materials metasomatize the ambient mantle, and as mantle upwells adiabatically because of plume flow it eventually melts to a sufficient degree to produce primary kimberlite melts. At deep upper mantle pressures, carbonated mantle produces a continuum of melt compositions from carbonatitic (low-Si) to kimberlitic (high-Si) as the degree of melting increases (Dalton and Presnall 1998; Gudfinnsson and Presnall 2005). Additionally, water has the effect of lowering the solidus further such that mixed CO2–H2O melting might occur at temperatures considerably lower (e.g. 150°C) than the solidus in the pure CO2 system (Falloon and Green 1990). Kimberlite melts can be produced by melting of upwelling CO2–H2O bearing mantle even along a mantle adiabat similar to that of MORB mantle (Gudfinnsson and Presnall 2005). Water might become available as mantle upwells, due to the lower storage capacity in the deep upper mantle relative to the transition zone (Hirschmann 2006), or as a consequence of reduction of carbonated melts by methane-rich mantle melts or fluids.

Conclusions

Diamonds and their mineral inclusions from the Collier 4 kimberlite pipe, Juina, Brazil, originated in the sublithospheric mantle. The chemistry of mineral inclusions is consistent with high-pressure phases stable in the deep upper mantle and transition zone. Mineral chemistry also implies an important role for subducted lithologies in the origin of these inclusions, as well as a polybaric, retrograde equilibration history. High-Ca majoritic garnet and former CaTiSi-perovskite likely crystallized from near-primary to differentiated carbonatitic liquids from partial melting of subducted oceanic crust. One inclusion has a mineral assemblage comprising garnet, kyanite, and former K-hollandite and CAS-phases, an assemblage that equates with high-pressure subducted and re-equilibrated pelagic sediment. The carbon isotope composition of Collier 4 diamonds is consistent with an origin involving subducted protoliths. This includes distinct groups of isotopically light diamonds with δ13C values of ~−14 and −24‰, respectively. Large increases in carbon isotopic composition from core to rim up to 8‰ in some diamonds indicate crystallization from varying carbon isotope reservoirs. A U/Pb date on a former CaTiSi-perovskite inclusion in one diamond gives an age of 101 ± 7 Ma, close to the age of the kimberlite volcanism (~93 Ma), suggestive of a link between carbonatitic melt metasomatism and primary kimberlite genesis. We propose that subducted lithosphere, possibly Mesoproterozoic to Triassic in origin, provides a source of carbonated subducted materials in the transition zone. Heating by a Cretaceous plume (e.g. Trinidade plume) initiates melting of carbonated slab materials (e.g. mafic eclogite, sediments). Some diamonds and inclusions crystallize initially from carbonated melts at deep upper mantle and transition zone depths. Subducted materials are transported upwards in a rising plume, and ultimately reside at the base of the Brazilian lithosphere at ~200 km. Protokimberlitic melts may be generated when carbonatite metasomatized mantle is mobilized and melted, possibly in the presence of water, as a consequence of passage of the plume.

Acknowledgments

Rio Tinto Desinvolvimentos Minerais Ltd. is thanked for access to Juina diamond samples and for permission to publish. Special thanks to R. Hinton and J. Craven at the Edinburgh Ion Microprobe Facility (EIMF) for their helpfulness and expertise in data acquisition. Thanks to D. Canil, T. Stachel and M. Schmidt for thoughtful reviews that improved the manuscript. This work was supported by NERC grant NE/E010466/1 to M.J. Walter.

Copyright information

© Springer-Verlag 2010