Accepted Manuscript

Kimberlite and carbonatite dykes within the Premier diatreme

root (Cullinan Diamond Mine, South Africa): New insights to
mineralogical-genetic classifications and magma CO2 degassing

Ashish Dongre, Sebastian Tappe

PII: S0024-4937(19)30169-0
DOI: https://doi.org/10.1016/j.lithos.2019.04.020
Reference: LITHOS 5049
To appear in: LITHOS
Received date: 11 February 2019
Accepted date: 20 April 2019

Please cite this article as: A. Dongre and S. Tappe, Kimberlite and carbonatite dykes
within the Premier diatreme root (Cullinan Diamond Mine, South Africa): New insights
to mineralogical-genetic classifications and magma CO2 degassing, LITHOS,
https://doi.org/10.1016/j.lithos.2019.04.020

This is a PDF file of an unedited manuscript that has been accepted for publication. As
a service to our customers we are providing this early version of the manuscript. The
manuscript will undergo copyediting, typesetting, and review of the resulting proof before
it is published in its final form. Please note that during the production process errors may
be discovered which could affect the content, and all legal disclaimers that apply to the
journal pertain.
 ACCEPTED MANUSCRIPT

Kimberlite and carbonatite dykes within the Premier diatreme

root (Cullinan Diamond Mine, South Africa): New insights to

mineralogical-genetic classifications and magma CO2

degassing

T
IP
Ashish Dongre1, * and Sebastian Tappe1#

CR
1
Deep & Early Earth Processes (DEEP) Research Group, Department of Geology, University of

US
Johannesburg, Auckland Park 2006, South Africa
AN
*Corresponding author
M

Ashish Dongre

andongare@unipune.ac.in; andongrey@gmail.com
ED

Tel. +91 9922410132; Fax +91 20 25695373

PT

ORCID iD: 0000-0002-0216-9753

Now at: Department of Geology, Savitribai Phule Pune University, Pune 411007, India
CE

# sebastiant@uj.ac.za
AC

Tel. +27 011-5594716; Fax +27 011-5594702

ORCID iD: 0000-0003-1224-5155

 ACCEPTED MANUSCRIPT

Abstract

The ca. 1153 Ma Premier kimberlite pipe on the Kaapvaal craton has been intruded by

late-stage kimberlite and carbonatite magmas forming discrete 0.5 to 5 m wide dykes

within the lower diatreme. On the basis of petrography and geochemistry, the fresh

kimberlite dykes represent archetypal monticellite phlogopite kimberlite of Group-1

T
affinity. Their mineral compositions, however, show marked deviations from trends that

IP
are typically considered as diagnostic for Group-1 kimberlite in mineralogical-genetic

CR
classification schemes for volatile-rich ultramafic rocks. Groundmass spinel

compositions are transitional between magnesian ulvöspinel (a Group-1 kimberlite

US
hallmark feature) and titanomagnetite trends, the latter being more diagnostic for
AN
lamproite, orangeite (formerly Group-2 kimberlite), and aillikite. The Premier kimberlite

dykes contain groundmass phlogopite that evolves by Al- and Ba-depletion to

M

tetraferriphlogopite, a compositional trend that is more typical for orangeite and aillikite.
ED

Although high-pressure cognate and groundmass ilmenites from the Premier

hypabyssal kimberlites are characteristically Mg-rich (up to 15 wt.% MgO), they contain
PT

up to 5 wt.% MnO, which is more typical for carbonate-rich magmatic systems such as
CE

aillikite and carbonatite. Manganese-rich groundmass ilmenite also occurs in the

Premier carbonatite dykes, which are largely devoid of mantle-derived crystal cargo,
AC

suggesting a link to the kimberlite dykes by fractionation processes involving

development of residual carbonate-rich melts and fluids. Although mineralogical-genetic

classification schemes for kimberlites and related rocks may provide an elegant

approach to circumvent common issues such as mantle debris entrainment, many of the
 ACCEPTED MANUSCRIPT

key mineral compositional trends are not as robust for magma type identification as

previously thought.

Utilizing an experimentally constrained CO2-degassing model, it is suggested that the

Premier kimberlite dykes have lost between 10 and 20 wt.% CO 2 during magma ascent

T
through the cratonic lithosphere, prior to emplacement near the Earth’s surface.

IP
Comparatively low fO2 values down to -5.6 ∆NNO are obtained for the kimberlite dykes

CR
when applying monticellite and perovskite oxybarometry, which probably reflects

significant CO2 degassing during magma ascent rather than the original magma redox

US
conditions and those of the deep upper mantle source. Thus, groundmass mineral
AN
oxybarometry may have little value for the prediction of the diamond preservation

potential of ascending kimberlite magmas.

M
ED

After correction for olivine fractionation and CO2-loss, there remains a wide gap

between the primitive kimberlite and carbonatite melt compositions at Premier, which
PT

suggests that these magma types cannot be linked by variably low degrees of partial
CE

melting of the same carbonated peridotite source in the deep upper mantle. Instead,

fractionation processes produced carbonate-rich residual melts/fluids from ascending

AC

kimberlite magma, which led to the carbonatite dykes within Premier pipe.

Keywords: Hypabyssal kimberlite, kimberlite-carbonatite relationships, magma redox

compositions, magma degassing, kimberlite magma evolution, kimberlite classification,

mineral chemistry
 ACCEPTED MANUSCRIPT

1. Introduction

The ca. 1153 Ma Premier kimberlite pipe in South Africa is a classic example of a large

(~32 hectares) diamondiferous diatreme structure that comprises both coherent

magmatic and volcaniclastic kimberlite infill, testifying to a complex emplacement history

(Bartlett, 1994; Tappe et al., 2018a). Premier Mine, which has operated intermittently

T
since 1903 and was renamed to Cullinan Diamond Mine in 2003, holds several records;

IP
for example, the largest ever discovered gem-quality rough diamond (the 3,106 carats

CR
Cullinan stone from 1905) and the highest recovery rate of large gem diamonds with

more than 800 stones above 100 carats in weight (up to end Financial-Year-2018;

US
personal communication by A. Rogers of Petra Diamonds Ltd). Premier is one out of
AN
only seven kimberlite pipes in the world holding the status of a Tier-1 diamond deposit,

with reserves worth more than 20 billion US$ in contained revenue (De Wit et al., 2016).
M
ED

Numerous studies have discussed the geology of the iconic Premier kimberlite pipe

including the petrology and geochemistry of the various identified kimberlite varieties
PT

such as the volcaniclastic ‘Grey’ and hypabyssal ‘Piebald’ kimberlite units (Wagner,
CE

1914; Allsopp et al., 1967; Bartlett, 1994; Field et al., 2008; Wu et al., 2013). Of

particular petrological interest, however, have been late-stage carbonate-rich dykes that
AC

cut the main kimberlite pipe infill, because these enigmatic dykes were considered

critical in the debate surrounding the origins of carbonatites and kimberlites, including

possible relationships between these magma types (Robinson, 1975; Mitchell, 1979;

Gaspar and Wyllie, 1984). The early work by Wagner (1914) suggested that these

carbonate-rich dykes represent hydrothermally or metasomatically overprinted

 ACCEPTED MANUSCRIPT

kimberlite. Daly (1925) identified several petrographic features such as intercumulus

textures that led him to suggest an igneous origin of the carbonate-rich dykes at

Premier kimberlite pipe. Subsequent petrographic and geochemical work on newly

exposed carbonate-rich dykes in the expanding underground operations of the mine

further supported a high-temperature magmatic origin (Deines and Gold, 1973;

T
Sheppard and Dawson, 1975; Robinson, 1975). Gradually, the debate shifted to

IP
whether or not such carbonate-rich dyke rocks that are intimately associated with

CR
kimberlites should be called ‘carbonatites’, even though their petrogenesis is distinctly

different from that of carbonatites in alkaline igneous complexes (Mitchell, 1979; Gaspar

US
and Wyllie, 1984). Although this debate strongly influenced models of kimberlite and
AN
carbonatite magma petrogeneses (Dalton and Presnall, 1998), there is now an

increasing consensus that carbonatites can have diverse origins in the upper mantle
M

(Mitchell, 2005; Bell and Simonetti, 2010; Jones et al., 2013). It has been widely
ED

recognized that carbonatite magmas are associated with many different types of

mantle-derived carbonated silicate melts, and that a range of processes (e.g., liquid
PT

immiscibility, fractional crystallization, flow differentiation) operating in diverse tectonic

CE

settings can provide links between co-occurring silicate-dominated and carbonate-rich

rock types (Woolley and Kjarsgaard, 2008; Tappe et al., 2017; Schmidt and
AC

Weidendorfer, 2018; Abersteiner et al., 2019).

We recently collected a new sample suite of late-stage dyke rocks cutting through Grey

volcaniclastic kimberlite pipe infill at the 645, 717, 763, and 839 m underground levels of

Cullinan Diamond Mine. This sample suite comprises carbonate-rich dykes (hereafter
 ACCEPTED MANUSCRIPT

referred to as carbonatite dykes) similar to the magnetite-serpentine-calcite dykes

described by Robinson (1975) and, importantly, hitherto unrecognized bona fide

kimberlite dykes that have an emplacement age of 1139.8 ±4.8 Ma (Tappe et al.,

2018a). Field evidence together with this high-precision U/Pb perovskite age suggest

coexistence of kimberlite and carbonatite magmas beneath the Premier cluster several

T
millions of years after the main pulses of pipe formation at 1153.3 ±5.3 Ma (Tappe et al.,

IP
2018a).

CR
In this contribution we discuss newly collected mineral and bulk rock major element data

US
for the coexisting kimberlite and carbonatite dykes from Premier pipe to better constrain
AN
their nature and evolution. We also report equivalent data for a few representative

samples of the volcaniclastic ‘Grey’ and hypabyssal ‘Piebald’ kimberlite varieties to

M

enable comparisons with the dyke rocks. One of our findings shows that the major
ED

element compositional trends of groundmass spinel and phlogopite in the kimberlite

dykes deviate appreciably from the respective mineral compositions in archetypal

PT

Group-1 magmatic kimberlites from other clusters on the Kaapvaal craton (cf., Mitchell,
CE

1995). This observation demonstrates that the widely applied mineralogical-genetic

classification of kimberlites and related rocks has significant caveats that need to be
AC

taken into consideration during both research and diamond exploration programs, which

rely on robust magma type identification for its geologic and economic implications.
 ACCEPTED MANUSCRIPT

2. Background

The Archean and Paleoproterozoic basement domains of the Kalahari craton (Kaapvaal

and Zimbabwe cratons plus surrounding mobile belts) in southern Africa host more than

1000 archetypal or Group-1 kimberlite occurrences (Jelsma et al., 2004; Griffin et al.,

2014; Tappe et al., 2018b), with the majority of clusters falling into five eruption

T
episodes: 1850-1650 Ma (e.g., Kuruman), 1200-1100 Ma (e.g., Premier), 600-500 Ma

IP
(e.g., Venetia), 250-220 Ma (e.g., Jwaneng), 110-70 Ma (e.g., Orapa and Kimberley). In

CR
contrast, Group-2 kimberlite magma eruptions were apparently confined to the

Kaapvaal craton and its margins during the time period between 200 and 115 Ma

US
(Phillips et al., 1998; Giuliani et al., 2015). The areal and temporal restriction of the
AN
Group-2 kimberlites, along with their ultramafic potassic alkaline nature, has prompted

renaming to ‘orangeites’ (Mitchell, 1995), or ‘lamproites variety Kaapvaal’ (Mitchell,

M

2006), in order to set them apart more rigorously from archetypal or Group-1 kimberlites
ED

that will hereafter be simply referred to as ‘kimberlite’ (unless specified otherwise).

Kimberlites are volatile-rich ultramafic magmas of mainly sublithospheric origin

PT

(Kjarsgaard et al., 2009), and they do not have the compositional traits of potassic melts
CE

derived from strongly K-metasomatized cratonic mantle lithosphere, such as

diamondiferous lamproites/ orangeites and ultramafic lamprophyres (Mitchell, 2006;

AC

Tappe et al., 2008; Giuliani et al., 2015; Shaikh et al., 2019).

The 1153.3 ±5.3 Ma Premier kimberlite pipe is with ~32 hectares surface area (400×860

m) the largest known diatreme structure in South Africa (Field et al., 2008; Tappe et al.,

2018a). It has an average diamond grade of 40 carats per hundred tons of kimberlite
 ACCEPTED MANUSCRIPT

ore, which makes Premier the richest body within a cluster of 11 Mesoproterozoic pipes

(de Wit et al., 2016), although the grade itself is far from being world-class compared

with other Tier-1 diamond deposits. Premier pipe was the first major diamondiferous

kimberlite discovery in the former Transvaal Province of South Africa near Pretoria in

1902, and open-pit mining commenced in 1903. The open pit was abandoned in 1932

T
after reaching a depth of 189 m, and since 1950 mining continues underground. In

IP
2003, Premier Mine was renamed to Cullinan Diamond Mine - after Thomas Cullinan

CR
who was the discoverer and first owner of the deposit - on the occasion of the mine

centenary. In 2007/2008, De Beers ownership (dating back to 1926) was transferred to

US
Petra Diamonds, and an ambitious underground mine extension program was
AN
completed in 2015, the first portion of the so-called Cut-C between 750 and 1000 m

depths below surface. At this depth level the Premier kimberlite pipe still has a footprint
M

of ~5 hectares, similar to the surface area of the ca. 90 Ma ‘Big Hole’ kimberlite pipe in
ED

Kimberley, South Africa. Besides its impressive inventory of exceptionally large and

pure high-value Type-IIa/b diamonds (Moore, 2009; Smith et al., 2016), the Premier
PT

kimberlite pipe is also renowned for its mantle-derived peridotite xenoliths (coarse and
CE

sheared varieties), which have been key in constraining Kaapvaal craton lithosphere

evolution in the vicinity of the 2056 Ma Bushveld Complex (Danchin, 1979; Gregoire et
AC

al., 2005; Viljoen et al., 2009).

The Premier kimberlite pipe intrudes sedimentary rocks of the Neoarchean to

Paleoproterozoic Transvaal Supergroup, as well as igneous rocks of the ca. 2056 Ma

Bushveld Complex at depth (Fig.1). The currently most robust age determination for
 ACCEPTED MANUSCRIPT

explosive magmatic activity at Premier is 1153.3 ±5.3 Ma (Tappe et al., 2018a), and this

pooled U/Pb perovskite age for several discrete volcaniclastic kimberlite units further

improves previous age constraints on Premier pipe emplacement (Allsopp et al., 1967;

Kramers and Smith, 1983; Wu et al., 2013). It is important to note, however, that late-

stage kimberlite dykes that cut the volcaniclastic units yielded a pooled U/Pb perovskite

T
age of 1139.8 ±4.8Ma, which suggests that the magmatic plumbing system was active

IP
for several millions of years beneath the volcanic field (Tappe et al., 2018a). A peculiar

CR
feature of the 1153.3 ±5.3 Ma Premier kimberlite pipe is the occurrence of a 75 m thick

1115 ±15 Ma gabbro sill (Allsopp et al., 1967) that is probably part of the 1112-1108 Ma

US
Umkondo large igneous event on and around the Kaapvaal craton (de Kock et al.,
AN
2014). The presence of the gabbro sill continues to create challenges during mining.

Wagner (1914) noted the presence of a large ‘floating reef’ of Paleoproterozoic

M

Waterberg quartzite that is now removed due to mining activity. Entrainment of

ED

Waterberg quartzite, a stratigraphic unit that has been entirely eroded from the Cullinan

area after 1153 Ma, suggests that the top 300 m of the original pipe are missing; i.e.,
PT

there is no record of crater- and bedded upper diatreme-facies kimberlite materials at

CE

Premier (de Wit et al., 2016).

AC

Detailed geological work identified the presence of different kimberlite varieties at

Premier pipe (Robinson, 1975; Bartlett, 1994), and their extent is now confirmed to at

least 1000 m depth below surface (Field et al., 2008). These varieties include massive

volcaniclastic kimberlites (formerly known as tuffisitic kimberlite breccias), such as the

‘Grey’ and ‘Brown’ kimberlite units, which represent typical lower diatreme infill in many
 ACCEPTED MANUSCRIPT

South African kimberlite pipes (Clement and Skinner, 1985; Sparks, 2013). The true

nature of the ‘Black/Green’ kimberlite unit is more uncertain and has recently been

regarded as transitional between hypabyssal magmatic and volcaniclastic (Skinner and

Marsh, 2004; de Wit et al., 2016). The western-central part of Premier pipe is occupied

by a magmatic plug, the so-called ‘Piebald’ kimberlite unit, from which we included a

T
few samples into our study for comparative purposes (note that ‘Piebald’ and ‘Black’

IP
kimberlite varieties represent discrete geological units). Wagner (1914) assigned the

CR
coherent magmatic Piebald kimberlite material to his ‘basaltic kimberlite’ group of

southern Africa (as opposed to the ‘micaceous kimberlite’ group), which decades later

US
became ‘Group-1 kimberlites’ based on geochemical and Sr-Nd-Pb isotopic constraints
AN
(Smith et al., 1985). Late-stage carbonatite dykes with sharp contacts against the host

volcaniclastic kimberlite pipe infill were reported by Wagner (1914) and Daly (1925)
M

from the open pit of Premier Mine, and their occurrence within both Grey and Black
ED

kimberlite units at the 500 m underground level was investigated in more detail by

Robinson (1975). We have studied further new occurrences of such carbonatite dykes
PT

from Cullinan Diamond Mine, where they co-occur with the 1139.8 ±4.8 Ma kimberlite
CE

dykes, first reported in Tappe et al. (2018a).

AC

3. Methodology

Kimberlite and carbonatite dykes were sampled from the 645, 717, 763, and 839 m

underground levels of Cullinan Diamond Mine (Gauteng Province, South Africa) (Fig.1)

in 2015 and 2016. These dykes are typically 0.5 to 5 m wide, but reach 10 m thickness

in places, intruding the massive volcaniclastic kimberlite units of the lower diatreme of
 ACCEPTED MANUSCRIPT

Premier pipe. A total of 14 individual dykes were sampled, out of which 6 are bona fide

kimberlites and 8 are carbonatite dykes. After rock cutting, the samples were cleaned

under running water, contamination-screened and chipped to small fragments using a

plastic-lined hammer followed by milling to powder in agate at the University of

Johannesburg (UJ). A representative polished thin section was prepared for each dyke

T
sample at the UJ.

IP
CR
Quantitative mineral analyses were conducted with a Cameca SX100 electron probe

micro-analyzer (EPMA) at the UJ. An acceleration voltage of 20 kV, a beam current of

US
20 nA, and a beam diameter of 1 µm were used. Counting times for non-volatile
AN
elements varied from 12 to 40 s on peak. During phlogopite analysis, Ba, F and Cl were

measured for 60, 50 and 30 s on peak, respectively, with a 3 µm defocused beam (15
M

kV, 20 nA). The electron microprobe instrument was calibrated using natural jadeite
ED

(Na), olivine (Mg), almandine (Al), diopside (Si), orthoclase (K), wollastonite (Ca),

rhodonite (Mn), hematite (Fe), barite (Ba), fluorite (F), and halite (Cl), as well as
PT

synthetic pure oxides for Ti, Cr, and Ni. All these elements were measured on their X-
CE

ray Kα lines and matrix corrections were done with the ‘X-PHI’ method, which is a φ(

ρz) type correction method.

AC

Bulk rock major element compositions were analyzed using a PANalytical MagiX PRO

X-ray fluorescence (XRF) spectrometer at the UJ. Sample powders were dried at 105

°C, and loss on ignition (LOI) was determined after heating for 30 minutes at 930°C in

air. The concentrations of SiO2, TiO2, Al2O3, Fe2O3, MnO, MgO, CaO, Na2O, K2O, NiO,
 ACCEPTED MANUSCRIPT

Cr2O3 and P2O5 were determined by XRF after borate fusion of the dried sample

powders. Detection limits are approximately 0.05 wt.%. Calibration of the XRF

spectrometer instrument was conducted using mixtures of pure metal oxides, and

accuracy and precision of our method were monitored using certified reference

materials (e.g., BE-N) and in-house kimberlite and carbonatite standards

T
(Supplementary File 1). Bulk rock CO2 concentrations were determined at ACME

IP
Analytical Laboratories Ltd in Vancouver (Canada) by liberation of CO2 gas from the

CR
powders in a reaction with 15% HClO4 and photo-coulometry analysis. The lower limit of

detection is approximately 0.02 wt.% CO2.

US
AN
4. Results

4.1. Petrography and mineral compositions

M

All studied kimberlite dykes and the investigated material from the Piebald kimberlite
ED

plug of Premier pipe show weakly macrocrystic textures imparted by the presence of

mostly olivine and occasionally rare ilmenite macrocrysts >1 mm in diameter. The
PT

kimberlite groundmass is dominated by olivine microphenocrysts, as well as a

CE

mesostasis consisting of monticellite, phlogopite, spinel, ilmenite, perovskite, apatite,

calcite, and serpentine. The carbonatite dykes show a fine-grained aphanitic texture and
AC

consist mainly of calcite, serpentine and titanomagnetite, plus rare olivine

microphenocrysts that are mostly serpentinized. Ni-Fe sulfide, apatite, ilmenite,

perovskite, and andradite garnet crystals typically <50 µm in size form volumetrically

minor but widespread constituents of the carbonatite dykes from Premier kimberlite

pipe. Further petrographic detail for the kimberlite and carbonatite dyke samples from
 ACCEPTED MANUSCRIPT

Premier pipe, plus for the few additional kimberlite samples from other units of the pipe,

is provided in Table 1 and Supplementary Table 1. The complete dataset of electron

microprobe mineral analyses is provided in Supplementary Table 2.

4.1.1. Olivine

T
In all studied Premier kimberlite dykes, as well as the Piebald kimberlite unit,

IP
subrounded olivine macrocrysts and euhedral to subhedral olivine microphenocrysts are

CR
common (Fig.2a, b). Fresh olivine suitable for EPMA analysis, however, is only present

in the kimberlite dykes, although serpentine also partially replaces these olivine crystals.

US
We did not find any compositional difference between macrocrystal and phenocrystal
AN
olivine from the Premier kimberlite dykes. Olivine compositions vary from Fo93 to Fo85.

The NiO and MnO contents vary from 0.3 to 0.5 wt.% and 0.07 to 0.16 wt.%,
M

respectively. The Cr2O3 contents are <0.15 wt.%. Olivine in the Premier kimberlite
ED

dykes follows a general evolutionary trend of correlated decreasing NiO and forsterite

contents (Fig.3a). However, this trend is unlikely to record fractional crystallization,

PT

because no relationship exists between the MnO and forsterite contents (Fig.3b). The
CE

cores of olivine macrocrysts show compositional similarity with olivine crystals from the

coarse granular and sheared peridotite xenoliths recovered from Premier kimberlite pipe
AC

and many other kimberlite occurrences worldwide (Fig.3a, b). This suggests that a large

proportion of macrocrystal olivine cores is derived from disaggregated mantle peridotite

wall-rocks as a consequence of dynamic kimberlite magma ascent (Bussweiler et al.,

2015; Howarth and Taylor, 2016; Lim et al., 2018). Peridotite-derived olivine xenocrysts

served as nucleation points for true magmatic olivine growth (Kamenetsky et al., 2008;
 ACCEPTED MANUSCRIPT

Pilbeam et al., 2013; Giuliani, 2018), as reflected by the correlated NiO and forsterite

relationship (Fig.3a). Olivine crystals (macro- and phenocrysts) from the Premier

kimberlite dykes are inclusion-free, and the apparent lack of pronounced compositional

zoning may be due to the fact that the olivine margins and rims are mostly replaced by

serpentine.

T
IP
4.1.2. Monticellite

CR
Monticellite can be an important groundmass phase in kimberlites and related rocks. In

these silica-poor ultramafic magma types it typically crystallizes after spinel and

US
perovskite (with some temporal overlap), but prior to the crystallization of groundmass
AN
calcite and late-stage primary serpophitic serpentine (Mitchell, 1986; Soltys et al.,

2018a). Monticellite is typically absent from more evolved calcite kimberlites, because
M

its crystallization is suppressed at high CO2 activities (Yoder, 1979), and replacement
ED

by calcite during late-stage magma crystallization is common. At Premier pipe,

monticellite occurs only in the kimberlite dykes, and it appears to be absent from the
PT

Grey volcaniclastic and Piebald hypabyssal kimberlite units, as well as from the
CE

carbonatite dykes. The monticellite crystals in the kimberlite dykes are <100 µm across,

uniformly distributed within the groundmass, and they have anhedral to subhedral
AC

shapes (Fig.2c, d). The monticellite crystals do not contain mineral inclusions and they

do not show significant compositional zoning. Another mode of occurrence are

monticellite overgrowths and rims on serpentinized olivine microphenocrysts in the

groundmass of the Premier kimberlite dykes (Fig.2d), which suggests reactions

between the original olivine crystals and residual volatile-rich liquids (Tappe et al., 2009;
 ACCEPTED MANUSCRIPT

Abersteiner et al., 2018). Monticellite crystals - together with groundmass spinel and

perovskite - are also poikilitically enclosed by phlogopite indicating their formation

shortly prior to, or contemporaneously with, groundmass phlogopite (Fig.4a). The

analyzed monticellite crystals have compositions close to the CaMgSiO4 end-member

(Fig.3c); however, in some cases notable solid solution toward CaFeSiO4 (9 to 22

T
mol.%) is observed, which can be utilized as an oxygen barometer (Le Pioufle and

IP
Canil, 2012) to estimate the redox conditions during kimberlite magma solidification (see

CR
Section 5.2.1).

4.1.3. Phlogopite
US
AN
Phlogopite crystals in the kimberlite dykes and the Piebald kimberlite unit are confined

to the groundmass; i.e., phlogopite macrocrysts and megacrysts are rare or absent.
M

Kimberlite dyke CIM15-75 shows abundant corroded and distorted/kinked phlogopite

ED

microcrysts (100-300 µm) (Fig.4c), which may represent mantle-derived antecrysts from

previous failed kimberlite magmatic pulses at depth. These phlogopite microcrysts

PT

typically contain <0.2 wt.% TiO2, an additional criterion suggested by Giuliani et al.
CE

(2016) for type kimberlite from Kimberley to devise a xenocrystic origin. In contrast, the

groundmass of the other studied kimberlite dykes contains euhedral to subhedral

AC

phlogopite aggregates typically <100 µm across, approaching 2 wt.% TiO2. Phlogopite

microphenocrysts (100-500 µm), i.e. crystals that directly formed from the cooling

kimberlite dyke magma, may contain poikilitic inclusions of monticellite, perovskite, and

spinel (Fig.4a, b). Elongated magnetite crystals may occur along phlogopite cleavage

planes (Fig.4a, b).

 ACCEPTED MANUSCRIPT

Phlogopite cores in the kimberlite dykes are variably enriched in Al 2O3 (up to 16 wt.%)

and FeOT (up to 15 wt.%) (Fig.5a,b). The TiO2 contents range from 0.01 to 1.7 wt.%.

The Grey volcaniclastic kimberlite samples are characterized by phlogopite

compositions with slightly elevated Al 2O3 (9.1-13.4 wt.%) and TiO2 (1- 3 wt.%), and low

T
FeOT (4.5-7.7 wt.%). These compositions are reminiscent of groundmass phlogopite

IP
from orangeites or Group-2 kimberlites. In contrast, phlogopite from hypabyssal Piebald

CR
kimberlite shows extreme enrichment in Al 2O3 (14-20 wt.%), but low TiO2 (0.7-2.2 wt.%)

and FeOT (2-6 wt.%), which is typical for archetypal or Group-1 kimberlites from

US
southern Africa (Mitchell, 1995; Giuliani et al., 2016). Importantly, phlogopite crystals
AN
from the Premier kimberlite dykes do not show compositions typical for Group-1

kimberlites. Three individual kimberlite dykes (CIM15-72, -74 and -80) show
M

development of tetraferriphlogopite rims on groundmass phlogopite, which is a more

ED

characteristic feature of aillikites (Tappe et al., 2005, 2006; Hutchison et al., 2018), as

well as orangeites and certain types of lamproites (Jaques et al., 1989; Mitchell and
PT

Bergman, 1991; Fritschle et al., 2013). The Al2O3 and TiO2 contents of groundmass
CE

phlogopite in these Premier kimberlite dykes vary from 0.9 to 6 wt.% and 0.07 to 1 wt.%,

respectively.
AC

The groundmass of the Piebald hypabyssal kimberlite sample CIM15-85 contains

abundant small (<20 µm across) elongated clusters of chloritized phlogopite flakes

(Fig.4d) that are highly enriched in BaO (up to 17 wt.%). Groundmass phlogopite from

the Premier kimberlite dykes also shows Ba enrichment with up to 6 wt.% BaO. Mitchell
 ACCEPTED MANUSCRIPT

(1995) reported that groundmass phlogopite in kimberlites typically has elevated Ba

contents reaching up to 20 wt.% BaO. The high Ba contents correspond typically to

increased alumina contents indicating development of solid solution toward kinoshitalite,

which is a Ba end-member of the trioctahedral mica group. Phlogopite cores from the

Premier kimberlite dykes and Piebald kimberlite plug are Ba-rich and this enrichment is

T
linked to slightly increased alumina contents, which demonstrates the presence of a

IP
considerable kinoshitalite component in groundmass phlogopite (Fig.6a, b). The BaO

CR
contents also show an increase with F contents in all kimberlite samples and facies

studied from Premier (Fig.6b). The coupled substitution between K++ Si4+and Ba2++ Al3+

US
is illustrated in Fig.6c, and it shows a 1:1 to 2:1 correspondence between K and Ba for
AN
groundmass phlogopite in the Premier kimberlite samples.
M

4.1.4. Spinel group minerals

ED

Spinel is a ubiquitous primary mineral in the groundmass of all studied kimberlite dykes
PT

from Premier pipe. It occurs as euhedral to subhedral crystals often together with
CE

perovskite. In a few cases spinel is overgrowing perovskite grains, and it also occurs as

inclusions in perovskite (Fig.7a, Supplementary Fig.2). Spinel grains also form poikilitic
AC

inclusions <50 µm across in groundmass phlogopite (Fig.7b). Magnetite blades can

commonly be observed along the cleavage planes of groundmass phlogopite crystals

(Fig.4a, 7c). Spinel (mainly magnetite and titanomagnetite) is abundant in the Premier

carbonatite dykes, where it occurs as discrete crystals <50 µm across.

 ACCEPTED MANUSCRIPT

Spinels from the Premier kimberlite dykes show a wide compositional range with

FeT2+/(FeT2++Mg) ratios between 0.2 and 0.9 (Fig.8a), straddling the boundary region

between two prominent magmatic trends known from kimberlites and related rocks

(Mitchell, 1986): the magnesian ulvöspinel trend (i.e., magmatic Trend-1) and the

titanomagnetite trend (i.e., magmatic Trend-2). Such transitional groundmass spinel

T
compositions have been reported more recently from a number of kimberlite and aillikite

IP
dyke and sill occurrences on the Canadian-Greenland Shield (Tappe et al., 2006, 2009,

CR
2014; Nielsen et al., 2009; Zurevinski and Mitchell, 2011; Sarkar et al., 2018), and their

potential significance will be discussed in Section 5.1. Groundmass spinel crystals from

US
the Premier kimberlite dykes have Cr2O3 and TiO2 contents of up to 43 wt.% and 16
AN
wt.%, respectively, and their MnO contents vary from 0.2 to 5.8 wt.%. Rare primary

spinel grains from Grey volcaniclastic kimberlite of the Premier pipe also show a wide
M

range in FeT2+/(FeT2++Mg) ratios between 0.3 and 1, with an affinity toward the
ED

titanomagnetite trend or magmatic Trend-2 (Fig.8a). Spinels from the carbonatite dykes

of Premier pipe have predominantly magnetite and titanomagnetite compositions with

PT

up to 25 wt.% TiO2. Our data for these newly discovered carbonatite dykes extend the
CE

known range of spinel compositions for Premier, because only relatively low TiO2 (0.6-

2.7 wt.%) and low MgO (4-9 wt.%) magnetite grains had been reported previously for
AC

the carbonatite dykes (Mitchell, 1979; Gaspar and Wyllie, 1984).

4.1.5. Ilmenite
 ACCEPTED MANUSCRIPT

Ilmenite was identified in the Premier kimberlite dykes and Piebald kimberlite unit, and it

appears to be rare in the carbonatite dykes. Ilmenite crystals in the kimberlite dykes are

anhedral and <500 µm across. Individual grains appear to be fragments of larger

ilmenite xenocrysts/macrocrysts, and in some cases they are overgrown by spinel and

perovskite (Supplementary Fig.1). Polycrystalline, disaggregated ilmenite macrocrysts

T
are also observed in the kimberlites dykes. These crystals show the effects of

IP
deformation and recrystallization in ascending kimberlite magma and, similar to some of

CR
the kinked phlogopite microcrysts described in Section 4.1.3, are probably representing

high-pressure antecrysts derived from failed kimberlite material at mantle depth (see

US
Castillo-Oliver et al., 2017). For the Premier carbonatite dykes, we only found rare <50
AN
µm large ilmenite grains and narrow ilmenite overgrowths on magnetite grains in

CIM16-12, which suggests a primary magmatic origin, as was also proposed by Gaspar
M

and Wyllie (1984) for the ilmenite mode of occurrence in a different suite of carbonatite
ED

dyke samples from Premier.

PT

Ilmenite from the Premier kimberlite dykes and Piebald kimberlite unit contains up to 2.5
CE

wt.% Cr 2O3 and minor amounts of Al 2O3 (<2.1 wt.%). The MgO and MnO contents are

elevated reaching up to 15 wt.% and 5 wt.%, respectively. Wyatt (1979) reported zoned
AC

ilmenite grains from volcaniclastic kimberlite units and the carbonatite dykes of Premier

pipe, emphasizing their Mn-rich margins. Extremely high MnO contents between 12.2

and 17.5 wt.% were confirmed for ilmenite crystals from the Premier carbonatite dykes

by Gaspar and Wyllie (1984). Groundmass ilmenite from the carbonatite dyke CIM16-12

analyzed here has somewhat lower MnO contents between 6.4 and 8 wt.%. Although
 ACCEPTED MANUSCRIPT

the different suites of carbonatite dyke samples mentioned above have a very similar

mineralogy (e.g., calcite, serpentine, magnetite, perovskite, ilmenite and apatite; see

Section 4.1), it must be noted that our samples were taken from new underground

exposures at much greater depths compared to the previous studies (cf., Wyatt, 1979;

Gaspar and Wyllie, 1984).

T
IP
In the FeTiO3-MgTiO3-MnTiO3 ternary diagram (Fig.8b), the ilmenite grains from the

CR
Premier kimberlite dykes and Piebald kimberlite unit analyzed here straddle the

boundary between kimberlitic (i.e., Mg-rich) and carbonatitic (i.e., Mn-rich)

US
compositions. In addition to the conspicuously Mn-rich nature of ilmenite from the
AN
Premier kimberlite dykes, the Nb2O5 content is also elevated reaching up to 1 wt.%,

which is more typical for ilmenite from carbonatites (Mitchell, 1978; Carmody et al.,
M

2014; Castillo-Oliver et al., 2017).

ED

4.1.6. Perovskite
PT

Perovskite is a ubiquitous accessory groundmass phase in the late-stage kimberlite

CE

dykes from Premier pipe ranging in diameter from 30 to 120 µm. It occurs in two

parageneses: (1) as discrete crystals in the groundmass, and (2) in an overgrowth to

AC

intergrowth association with groundmass spinel (Supplementary Fig.2). Discrete

groundmass perovskite crystals are subhedral to anhedral in shape (i.e., rounded and

abraded). They are compositionally relatively homogeneous and no major element

concentration differences are observed between perovskite parageneses. Minor

quantities of FeOT (0.7-2.3 wt.%) and Al2O3 (0.1-1 wt.%) are present, and given the
 ACCEPTED MANUSCRIPT

relatively low totals of the EPMA analyses, appreciable quantities of Na, Sr and LREE

(e.g., Ce) may also be present but were not analyzed. Niobium content does not exceed

0.85 wt.% Nb2O5. Although perovskite crystallizes mainly after olivine and spinel, and

prior to groundmass phlogopite and monticellite, the intergrowth with anhedral spinel

crystals indicates early crystallization during groundmass formation (Malarkey et al.,

T
2010; Soltys et al., 2018a).

IP
CR
4.1.7. Carbonates

Carbonate minerals in magmatic kimberlites occur generally as fine-grained

US
groundmass crystals, up to cm-long laths, complex segregations, secondary veins, and
AN
late crystallizing phases replacing olivine and other groundmass constituents (Mitchell,

1986; Castillo-Oliver et al., 2018). In most of the late-stage Premier kimberlite dykes
M

and in the hypabyssal Piebald kimberlite unit primary groundmass carbonate is present.
ED

Carbonate is the most abundant constituent of the carbonatite dykes at Premier pipe

(>30 vol.% as per definition of carbonatite by Mitchell, 2005). For both the kimberlite and
PT

carbonatite dykes the main identified carbonate phase is calcite. The groundmass of the
CE

kimberlite dykes and the hypabyssal Piebald kimberlite unit shows abundant interstitial

calcite crystals <100 µm across, and texturally most of these crystals appear to be of
AC

primary origin formed during solidification of residual kimberlite melt, which is also

supported by elevated Sr contents of up to 0.9 wt.% SrO (Supplementary Table 2).

However, some coarser anhedral calcite grains appear to replace olivine, monticellite,

and serpentine in the kimberlite groundmass. The main compositional difference

between calcite from the kimberlite and carbonatite dykes at Premier pipe is the slightly
 ACCEPTED MANUSCRIPT

higher Mg content of the latter reaching up to 1.4 wt.% MgO (Fig.9a). By using CL-

imaging techniques, Castillo-Oliver et al. (2018) observed some minor replacement of

original mm-sized calcite crystals in two Premier carbonatite dykes by finer secondary

calcite. The large calcite crystals show slightly elevated Mg contents of up to 0.55 wt.%

MgO, as well as high Sr concentrations of up to 1 wt.% SrO (Castillo-Oliver et al., 2018),

T
which is overall very similar to our results for calcite from the Premier carbonatite dykes

IP
(up to 0.8 wt.% SrO; Supplementary Table 2).

CR
4.1.8. Andradite garnet

US
Primary magmatic andradite garnets are generally absent from the groundmass of
AN
kimberlites from worldwide occurrences (Mitchell, 1986; Tappe et al., 2005), an

observation that is also confirmed for the fresh kimberlite dykes from Premier pipe
M

investigated here. However, garnet crystals rich in andradite component are present in
ED

the mosaic-textured groundmass of the carbonatite dykes, where they occur as

alteration or reaction rims on calcite crystals. These garnets have high CaO (34-36
PT

wt.%), FeO (17-24 wt.%) and TiO2 (up to 11 wt.%) contents, corresponding to 65 to 87
CE

mol.% of andradite component. However, Ti contents are lower than in the schorlomite-

and kimzeyite-rich magmatic garnets observed in aillikites and orangeites from

AC

worldwide occurrences (Fig.9b) (Tappe et al., 2004, 2006, 2009; Dongre et al., 2016). In

the present study we also observed andradite-rich garnet in Grey volcaniclastic

kimberlite sample CIM15-83, and the mode of occurrence suggests formation during

interaction between late-stage carbonate-rich fluids and unconsolidated kimberlite

tephra; i.e., a secondary origin. Although the secondary andradite crystals encountered
 ACCEPTED MANUSCRIPT

in the carbonatite dykes and volcaniclastic kimberlites from Premier pipe are small (<50

µm) and have irregular porous textures, the acquired EPMA results are fully

stoichiometric (Supplementary Table 2). The EPMA data confirm garnet crystal

chemistry, with some structurally bound OH- molecules, i.e., hydrogarnet composition.

Garnet crystallography was also verified for similar-quality secondary andradite grains

T
from orangeites of the Dharwar craton in India by Raman spectroscopy (Dongre et al.,

IP
2016).

CR
4.2. Whole-rock geochemistry

US
AN
Major- and minor element compositions of the kimberlite and carbonatite dykes from

Premier pipe are listed in Table 2 and displayed in Figure 10 (selected bulk kimberlite
M

analyses have recently been reported in Tappe et al., 2018a). The kimberlite dykes are
ED

relatively SiO2-poor (31.1-36.1 wt.%) and the carbonatite dykes contain 10 to 12 wt.%

SiO2 (Fig.10). A wide range of MgO is observed for the carbonatite dykes (10-30.6
PT

wt.%), whereas the kimberlite dykes show a much more restricted MgO range between
CE

31.1 and 32.4 wt.%. Higher abundances of calcite in the carbonatite dykes are clearly

the source of elevated CaO ranging from 20 to 36 wt.%, which is much higher than in
AC

the kimberlite dykes (4.5-14.3 wt.% CaO). The TiO2 (1.2-1.8 wt.%) and Al2O3 (0.8-1.2

wt.%) contents are low in the carbonatite dykes compared with the kimberlite dykes

(2.2-2.6 wt.% TiO2 and 1.9-2.5 wt.% Al2O3). The hypabyssal Piebald kimberlite sample

CIM15-85 is plotted alongside the kimberlite dyke samples from Premier pipe and they

show very similar major element compositions that are diagnostic of an archetypal
 ACCEPTED MANUSCRIPT

Group-1 kimberlite magma lineage (e.g., Fig.11a). The CO2 contents vary widely in the

carbonatite dykes with maximum concentrations of 26.3 wt.%, whereas the kimberlite

dykes contain only 0.3 to 4 wt.% CO2. The few Grey volcaniclastic kimberlite samples

from Premier pipe have higher SiO2 (44.7-46.7 wt.%) and Al 2O3 (4.2-4.5 wt.%) contents

compared with the magmatic kimberlite samples studied here. The Grey volcaniclastic

T
kimberlite samples have lower MgO (23-25 wt.%) and CaO (6.7-7.3 wt.%) compared

IP
with the kimberlite dykes and hypabyssal Piebald kimberlite unit. The kimberlite dykes

CR
have a Contamination Index close to 1 (1.0-1.2; C.I. =

[SiO2+Al2O3+Na2O)/(MgO+2*K2O]), which indicates that only little or no crustal

US
basement material has been incorporated into the kimberlite magma during ascent
AN
(Kjarsgaard et al., 2009).
M

5. Discussion
ED

5.1. Pitfalls in the application of mineralogical-genetic classifications

The classification and subdivision of volatile-rich ultramafic rocks that intrude cratons
PT

and surrounding mobile belts rely on the identification of primary mineral assemblages
CE

(i.e., petrography) and their compositional trends, as determined by electron microprobe

analysis. This approach dates back to Dawson (1980) and has been developed by R.H.
AC

Mitchell in a series of monographs dedicated to kimberlites, lamproites and orangeites

(Mitchell, 1986; Mitchell and Bergman, 1991; Mitchell, 1995). The approach is

commonly referred to as ‘petrogenetic classification’ or ‘mineralogical-genetic

classification’ (Woolley et al., 1996; Mitchell, 2006; Mitchell and Tappe, 2010; Scott

Smith et al., 2013). Besides petrographic observations, detailed analysis of the

 ACCEPTED MANUSCRIPT

groundmass mineral compositions in volatile-rich ultramafic rocks is a critical step

toward developing a sound understanding of the nature and evolution of their parental

magmas, because the presence of abundant mantle-derived fragments (i.e., xenoliths

and xenocrysts) tends to obscure subtle major element geochemical differences

between these economically important rock types. Furthermore, primitive volatile-rich

T
magmas (e.g., with abundant H2O, CO2, F, Cl) tend to solidify as highly variable and

IP
diverse mineral assemblages as a function of P-T-fO2, so that mineral compositional

CR
trends are often the only means by which distinct parental magma types can be

identified (Rock, 1991). Tappe et al. (2005) pointed out, however, that ideally

US
mineralogical and geochemical methods should be applied in concert, and that
AN
investigations restricted to a single hand-sample bear the risk of failure to identify

diagnostic minerals and mineral compositional trends for petrogenetically related

M

volatile-rich ultramafic rock occurrences.

ED

Before we can address potential genetic relationships between the kimberlite and
PT

carbonatite dykes from Premier pipe at Cullinan Diamond Mine, it is important to clarify
CE

their nature that leads us to suggest the rock terminology used herein. Based on

petrography and bulk rock major element compositions, the Premier kimberlite dykes
AC

are identified herein as microporphyritic to weakly macrocrystic monticellite phlogopite

hypabyssal kimberlites of clear Group-1 affinity (Fig.11a). Although this result comes as

no surprise provided that Premier pipe has been previously described as a Group-1

kimberlite based on the petrography and geochemical compositions, including Sr-Nd-Hf

isotope ratios, of the Black and Piebald units (Bartlett, 1994; Nowell et al., 2004;
 ACCEPTED MANUSCRIPT

Woodhead et al., 2009; Wu et al., 2013), groundmass spinel and phlogopite from the

newly discovered late-stage kimberlite dykes follow compositional trends that are

different from those generally devised for archetypal kimberlites in mineralogical-genetic

classification schemes (cf., Mitchell, 1995). For example, the spinel compositions are

transitional between magmatic Trend-1 and Trend-2, and they tend to be too Mg-poor

T
for ‘typical’ groundmass spinel from archetypal kimberlites (Fig.8a). This may be due to

IP
relatively late Cr-spinel crystallization together with groundmass phlogopite and

CR
monticellite (Fig.7), or even together with late olivine, phases that compete for and

consume MgO (Roeder and Schulze, 2008). Early liquidus chromite of Trend-1 affinity

US
may have been lost from the Premier kimberlite dyke system due to onset of crystal
AN
fractionation at much greater depths than the current outcrop level (see Abersteiner et

al., 2019). The relatively low Cr contents of the spinel crystals, less than 43 wt.% Cr2O3,
M

would be in support of this notion, together with the observation that such spinel grains
ED

form poikilitic inclusions in groundmass phlogopite (Fig.7b). This suggestion is further

supported by an observation from West Greenland, where the fresh Majuagaa

PT

kimberlite dyke contains abundant primary magnesian ulvöspinel crystals (Trend-1) set
CE

in a groundmass of calcite, with little or no phlogopite and monticellite (Nielsen and

Sand, 2008). Transitional groundmass spinel compositions between magnesian

AC

ulvöspinel (Trend-1) and titanomagnetite (Trend-2) have been reported more recently

from a number of kimberlite and aillikite dyke and sill occurrences from across the

Canadian-Greenland Shield (Birkett et al., 2004; Tappe et al., 2006, 2009, 2014;

Nielsen et al., 2009; Zurevinski and Mitchell, 2011; Sarkar et al., 2018). Groundmass

spinel with Trend-2 compositional evolution has also been reported from magmatic
 ACCEPTED MANUSCRIPT

kimberlite units of the Letseng pipe in Lesotho (Stamm et al., 2018), which is a ‘classic’

Group-1 kimberlite occurrence on the Kaapvaal craton. This prompts the question of

how diagnostic and useful spinel compositions actually are for discrimination between

kimberlites and related volatile-rich ultramafic rock types in mineralogical-genetic

classification schemes (cf., Mitchell, 1995; Tappe et al., 2005).

T
IP
Groundmass phlogopite compositions from the late-stage kimberlite dykes at Premier

CR
are relatively alumina-poor and evolve by Al- and Ba-depletion toward

tetraferriphlogopite (Fig.5), which is uncommon for Group-1 kimberlites (Mitchell, 1995).

US
Group-1 kimberlites typically contain alumina-rich phlogopite in the groundmass that
AN
evolves by Al- and Ba-enrichment to kinoshitalite-rich mica compositions. Although

alumina-poor phlogopite compositions are typical for orangeites and lamproites, these
M

more micaceous potassic ultramafic rock types typically contain groundmass phlogopite
ED

with much higher Ti contents than those from the Premier kimberlite dykes (Fig.5a). In

contrast, the hypabyssal Piebald kimberlite plug from Premier pipe contains
PT

groundmass phlogopite that compositionally satisfies the ‘accepted’ criteria for Group-1
CE

kimberlites in mineralogical-genetic classification schemes; i.e., this phlogopite is

alumina-rich and evolves by Al- and Ba-enrichment toward kinoshitalite end-member

AC

composition (Fig.6). Clearly, there is no single solution to the classification issue that

can be obtained from the iconic Premier kimberlite pipe. Recently, Stamm et al. (2018)

also reported tetraferriphlogopite compositions from magmatic units of the Letseng

kimberlite pipe in Lesotho, another classic Group-1 kimberlite occurrence on the

 ACCEPTED MANUSCRIPT

Kaapvaal craton (Fig.1b). This should prompt the question of when exceptions may

become the rule.

Ilmenite grains <500 µm in size from the Premier kimberlite dykes and Piebald

kimberlite unit have elevated MgO and MnO contents reaching up to 15 wt.% and 5

T
wt.%, respectively. Extremely high MnO content between 12.2 and 17.5 wt.% was

IP
reported for groundmass ilmenite from the late-stage carbonatite dykes at Premier

CR
(Gaspar and Wyllie, 1984), and our ilmenite data for new carbonatite dyke discoveries

(6.4-8 wt.% MnO) fall close to the elevated Mn contents of the ‘kimberlitic’ ilmenite

US
grains. The ilmenite grains from the kimberlite dykes compositionally straddle the
AN
boundary between kimberlitic (i.e., Mg-rich) and carbonatitic (i.e., Mn-rich) ilmenites

(Fig.8b). In addition to their conspicuously Mn-rich nature, the Nb2O5 content is also
M

elevated reaching up to 1 wt.%, which is more typical for ilmenite from carbonatites
ED

(Mitchell, 1978; Carmody et al., 2014; Castillo-Oliver et al., 2017). Although this

observation suggests a genetic link between the late-stage kimberlite and carbonatite
PT

dykes at Premier pipe, which will be discussed in Section 5.2.3, it introduces further
CE

ambiguity when searching for answers in mineralogical-genetic classification schemes

devised for volatile-rich ultramafic rocks including carbonatites (cf., Mitchell, 1995;
AC

Woolley et al., 1996; Tappe et al., 2005).

The late-stage carbonatite dykes from Premier pipe contain >30 vol.% of primary

magmatic carbonate (mainly Mg-bearing calcite; see Fig.9a), which justifies

classification of these rather unusual rocks as carbonatites using the revised lowered
 ACCEPTED MANUSCRIPT

modal carbonate cut-off criterion by Mitchell (2005). A primary magmatic nature of the

carbonate component in these dykes was also suggested by Deines and Gold (1973)

and Sheppard and Dawson (1975) on the basis of 13C/12C determinations, which yielded

mantle-like compositions at around -5‰ 13C. Besides the textural evidence for the

primary magmatic nature of the large calcite grains in the Premier carbonatite dykes

T
mentioned in Section 4.1.7, Castillo-Oliver et al. (2018) additionally reported low initial

IP
87
Sr/86Sr (0.7022-0.7026) for these carbonates, determined by LA-MC-ICPMS analysis.

CR
These authors pointed out that the Sr isotopic compositions of the magmatic carbonates

US
from the carbonatite dykes overlap with those of groundmass carbonates (0.7022-

0.7027; Castillo-Oliver et al., 2018) and perovskites (0.7025-0.7030; Woodhead et al.,

AN
2009; Wu et al., 2013) from the Black kimberlite variety at Premier pipe, strongly

supporting melt derivation from a common source in the upper mantle. Much debate
M

surrounded the nature and nomenclature of these carbonatite dykes from Premier pipe
ED

in the past, when they were variably referred to as ‘calcite-kimberlites’ and ‘calcareous

kimberlites’, or simply as ‘carbonate-rich dykes’ (Mitchell, 1986; and references therein).

PT

However, it has become clearer, especially after the discovery of bona fide kimberlite
CE

dykes at Premier pipe (Tappe et al., 2018a; this study), that these calcite – serpentine –

magnetite rocks lack a typical kimberlite mineral assemblage including the mantle-
AC

derived xenocrystic component.

In summary, if the Premier Group-1 kimberlite dykes were encountered in a diamond

exploration program at an early stage without sufficient geological and petrological

context, then rigorous application of the existing mineralogical-genetic classification

 ACCEPTED MANUSCRIPT

schemes by non-experts would have suggested that they are Group-2 kimberlites or

aillikites, which represent lithosphere-sourced/influenced and generally less

economically important magma types compared with Group-1 kimberlite magmas from

mainly sublithospheric sources. This finding illustrates that the widely applied

mineralogical-genetic classification of kimberlites and related rocks has more caveats

T
than is probably appreciated within the community, and that it should only be applied in

IP
conjunction with detailed petrographic observations (Mitchell and Tappe, 2010). In the

CR
case of the Premier ultramafic dykes, an affinity to Group-2 kimberlites (orangeites) can

be ruled out based on the common presence of groundmass monticellite (Mitchell,

US
1995). However, petrographic distinctions between the Premier Group-1 kimberlite
AN
dykes and typical aillikites are less clear-cut, because both rock types can contain

monticellite (Rock, 1991; Tappe et al., 2009). Although the absence of primary
M

groundmass clinopyroxene, as at Premier, is a hallmark feature of Group-1 kimberlites,

ED

clinopyroxene is not an essential mineral in aillikites (Tappe et al., 2005), even though it

may occur in appreciable abundances (Tappe et al., 2004, 2006; Hutchison et al.,
PT

2018). Caution must be exercised when using the mineralogical compositions of

CE

volatile-rich and alkaline ultramafic rocks for identification of their parental magma

lineage, which should be ideally done on large and representative sample suites,
AC

preferably in combination with bulk rock geochemical analyses. The importance of this

exercise should not be underestimated, because both research and diamond

exploration programs rely on robust magma type identification for its geologic and

economic implications.
 ACCEPTED MANUSCRIPT

5.2. Petrology

5.2.1. Groundmass monticellite oxybarometry

The oxygen fugacity (fO2) of kimberlite magmas can play an important role in affecting

the quantity and quality of the entrained diamond cargo (Fedortchouk et al., 2005).

T
Elevated fO2 conditions can promote diamond resorption, which diminishes the diamond

IP
grade and value within a given kimberlite magma batch or unit. Although the fO2

CR
conditions of many magma types can be readily calculated by a variety of techniques,

kimberlites offer only a small number of approaches by which fO2 can be evaluated in a

US
meaningful manner. For example, an oxygen barometer was developed for kimberlites
AN
based on partitioning of Fe3+ between groundmass perovskite and liquid (Bellis and

Canil, 2007). More recently, Le Pioufle and Canil (2012) proposed an oxygen barometer
M

for kimberlite magmas based on the Fe content of monticellite in equilibrium with

ED

kimberlite liquid. We prefer this monticellite oxybarometer over its perovskite

counterpart, because one can safely assume that all Fe in monticellite is

PT

accommodated as Fe2+ (Le Pioufle and Canil, 2012). Moreover, it is relatively

CE

straightforward to collect high-quality stoichiometric mineral chemical data for

monticellite (Supplementary Table 2), whereas guaranteeing quality EPMA data for
AC

perovskite is more challenging provided the large number of existing solid solution end-

members including exotic components (see Section 4.1.6). Textural observations from

the Premier kimberlite dykes, such as monticellite and perovskite crystals being

poikilitically enclosed in groundmass phlogopite (Fig.4a), predict that similar fO2 values

should be recorded by these two phases, assuming comparable data quality.

 ACCEPTED MANUSCRIPT

Application of the monticellite oxybarometer to the Premier kimberlite dykes yields

∆NNO values that range from -5.60 to -0.27 (Supplementary Table 3), which is almost

identical to the range obtained from groundmass perovskite (-4.53 to +0.18 ∆NNO).

These kimberlite magma oxygen fugacity estimates for Premier pipe are generally

T
comparable to other magmatic kimberlites from South Africa (Dutoitspan; -5.3 to -1.1

IP
∆NNO; Ogilvie-Harris et al., 2009), India (Narayanpet; -3.2 to -1.9 ∆NNO; Chalapathi-

CR
Rao et al., 2012), and Arctic Canada (Somerset Island; -4.3 to zero ∆NNO; Canil and

Bellis, 2007), as summarized by Le Pioufle and Canil (2012). They are also in good

US
agreement with monticellite-based oxygen fugacity estimates for calcite kimberlite dykes
AN
from Tikiusaaq in West Greenland (-3.3 to +1.2 ∆NNO), as calculated herein based on

the data given in Tappe et al. (2009, 2017). However, the Premier kimberlite dykes
M

record some of the lowest ∆NNO values (at around -5.60) in the global kimberlite
ED

database (Fig.11b). Although these apparently low fO2 conditions for the Premier

kimberlite dykes must have been favorable for diamond preservation during magma
PT

ascent and cooling, the actual significance of such oxygen barometry data remains
CE

difficult to establish in light of the very high fO2 conditions that are recorded by the Lac

de Gras kimberlites (∆NNO values of up to +6; Canil and Bellis, 2007), which host two
AC

major diamond deposits (i.e., Ekati and Diavik mines) on the central Slave craton in NW

Canada (Fig.11b).

The comparatively low fO2 values obtained for the Premier kimberlite dykes may instead

reflect significant CO2 degassing and volatile loss during magma ascent (Le Pioufle and
 ACCEPTED MANUSCRIPT

Canil, 2012), which would be in good agreement with the relatively low CO 2 contents

(<4 wt.%) of the kimberlite dykes (Fig.10f). The main obstacle with this interpretation is

the presence of tetraferriphlogopite rims on groundmass phlogopite crystals in the

Premier kimberlite dykes, which is typically thought to record fairly abrupt and strong

oxidation of the late-stage liquids in order to explain the high Fe 3+/Fe2+ of these alumina-

T
poor micas (McCormick and Le Bas, 1996).

IP
CR
5.2.2. Estimates of CO2 degassing from Premier kimberlite magma

US
Several petrological studies suggest that on the order of 10 wt.% CO2 are typically lost
AN
from primitive kimberlite magmas upon ascent to the Earth’s surface (Brooker et al.,

2011; Moussallam et al., 2016; Soltys et al., 2018b; Sun and Dasgupta, 2019). In an
M

attempt to quantify the amount of CO2-loss from the intruding Premier kimberlite
ED

magma, we have utilized an experimental fit between melt CO 2 and SiO2 contents

based on the peridotite-melting-experiment database of Massuyeau et al. (2015), in

PT

which all melt compositions are in equilibrium with olivine and orthopyroxene. The
CE

database reveals the following empirical relationship: CO2 = -0.85 × SiO2 + 42 (Fig.12),

with all concentrations expressed as wt.% (see Supplementary Figure 3 for derivation of
AC

the equation). Sun and Dasgupta (2019) obtained a similar parameterization, but their

study aimed at understanding reaction processes of carbonate melts in the deep upper

mantle. In any case, the first step to estimate the amount of CO2-loss is to correct the

measured kimberlite major element composition (except for CO2) for olivine

accumulation, which is done using a KDol-liqFe-Mg of 0.45, as guided by recent CO2-H2O-

 ACCEPTED MANUSCRIPT

bearing peridotite melting experiments (Sokol et al., 2013; Stamm and Schmidt, 2017).

Although this correction assumes that all olivine is magmatic, the difference in Mg/Fe

ratios between magmatic (e.g., Mg# ~86-90) and mantle-derived xenocrystic (e.g., Mg#

~90-92) olivine components in the Premier kimberlite dykes - and in general - is small

(Fig.3). This means that the correction accounts crudely for the entire ‘excess’ olivine

T
population that typically shifts parental kimberlite magmas to higher MgO contents (e.g.,

IP
Soltys et al., 2018b). The CO2-loss is then constrained by anchoring near-primary

CR
kimberlite melt compositions at 22 wt.% MgO (i.e., at the lower end of the range for

parental Group-1 kimberlite magmas as proposed by Kjarsgaard et al., 2009; Fig.10), so

US
that the original melt CO2 content can be determined from the empirical CO2-SiO2 fit
AN
(Fig.12). Further details of this method are provided in Massuyeau et al. (2019,

submitted).
M
ED

Applying this approach to the Premier kimberlite dykes suggests a minimum of 10 wt.%

CO2-loss, but up to 20 wt.% CO2-loss is possible (Fig.12). The effect of olivine

PT

accumulation on CO2 degassing is relatively minor at Premier. The calculated near-

CE

primary kimberlite melt compositions for Premier, prior to CO2-loss and anchored at 22

wt.% MgO, contained between 23 and 30 wt.% SiO2 (clustering at ~24 wt.% SiO2;
AC

Fig.12), which is very similar to ultrafresh quenched-undegassed calcite kimberlite

dykes from Tikiusaaq in West Greenland (Tappe et al., 2017). The carbonatite dykes

from Premier were not corrected for olivine accumulation and a similar approach was

used to correct for CO2 degassing. The calculated carbonatite melt compositions prior

to CO2-loss contained approximately 10 wt.% SiO2. Projection of the corrected CO2

 ACCEPTED MANUSCRIPT

abundances for the carbonatite dykes suggests that a minimum of 5 wt.% and a

maximum of 30 wt.% CO2-loss has occurred (Fig.12). The meaning of these values,

however, remains uncertain provided that the carbonatite dykes were probably far

removed from primary liquid compositions (see Section 5.2.3).

T
5.2.3. Kimberlite - carbonatite link at Premier pipe: a preliminary model

IP
CR
The exercise of estimating CO2-loss in Section 5.2.2 demonstrates that both kimberlite

and carbonatite dykes at Premier have lost at least 10 and 5 wt.% CO2, respectively,

US
which highlights the fact that the parental magmas underwent some form of
AN
differentiation. This magma differentiation, including decarbonation, may have occurred

within the cratonic mantle lithosphere down to ~120 km depth (e.g., Stone and Luth,
M

2016; Aulbach et al., 2017), and it does not necessarily correspond to the eruption
ED

process and volatile degassing (e.g., CO2 + H2O) at surface. In CO2 – SiO2 parameter

space there is a wide gap between the calculated near-primary kimberlite and
PT

carbonatite melt compositions for Premier pipe (Fig.12). This simple observation
CE

suggests that a primary melting continuum of carbonated peridotite, where the only

changing variable is the degree of partial melting (cf., Dalton and Presnall, 1998;
AC

Agashev et al., 2008), cannot provide a link between the carbonatite and kimberlite

dykes. Immiscible separation of carbonate-rich from silicate-rich liquids remains an

option by which the kimberlite and carbonatite dykes may be related. However, there is

no textural evidence, such as the presence of ocelli and carbonate segregations, to

support the operation of two-liquid immiscibility at Premier pipe. Also, miscibility gaps
 ACCEPTED MANUSCRIPT

are very small for carbonated silicate melts with low SiO 2 contents (Brooker and

Kjarsgaard, 2011), which renders this process unlikely to yield carbonatite liquid from

kimberlite magma (Kamenetsky and Yaxley, 2015). Links between coeval kimberlite and

carbonatite intrusives via fractionation processes have been proposed for the rifted

margin of the North Atlantic craton in West Greenland (Tappe et al., 2017). For this

T
region, it was demonstrated that carbonatite sheets are more fractionated in terms of

IP
trace element compositions and stable C-Li isotope ratios compared with co-occurring

CR
kimberlite dykes, and that efficient removal of silicate (e.g., olivine and phlogopite) and

oxide (e.g., spinel and ilmenite) crystals from kimberlitic magma can lead to the

US
observed small volumes of carbonatite liquid. On a much smaller scale, an evolution of
AN
kimberlitic melt toward carbonatitic liquid compositions by means of crystal fractionation

has also been observed in detailed studies of primary melt inclusions in magmatic
M

minerals of kimberlites (Giuliani et al., 2017; Abersteiner et al., 2019). Although the
ED

mineralogy of the Premier carbonatite dykes is primarily calcite – serpentine –

titanomagnetite, rare olivine phenocryst pseudomorphs and ilmenite crystals have been
PT

identified, and they may represent a portion of a fractionating high-pressure mineral

CE

assemblage that was ultimately derived from evolving kimberlite magma.

AC

An insightful constraint on a fractionation relationship between the Premier kimberlite

and carbonatite dykes may come from the observed ilmenite compositions, which have

significantly elevated Mn contents (Fig.8b). Elevated Mn in ilmenite from kimberlite was

suggested to indicate rather fractionated and carbonate-enriched kimberlite magma

compositions during late-stage crystallization (Tompkins and Haggerty, 1985). Recently,

 ACCEPTED MANUSCRIPT

Castillo-Oliver et al. (2017) suggested that Mn-rich ilmenites in hypabyssal kimberlites

from NE Angola formed by intensive metasomatism induced by carbonate-rich melts

and fluids at a late-stage during kimberlite magma emplacement. The fact that both

kimberlite and carbonatite dykes at Premier contain ilmenite grains (macrocrysts-

antecrysts and groundmass crystals) with conspicuously high Mn contents suggests an

T
important role for fractionation processes by which carbonate-rich residual melts/fluids,

IP
as probably represented by the carbonatite dykes, developed from the evolving

CR
kimberlite magma.

US
At present, a petrogenetic link between the Premier kimberlite and carbonatite dykes via
AN
crystal fractionation - and perhaps some form of filter-pressing and fluid mobility - is

preferred, but this preliminary model awaits further testing by geochemical means,
M

including trace element and isotope analyses.

ED

6. Summary and conclusions

PT
CE

The 1153.3 ±5.3 Ma Premier kimberlite pipe has been intruded at 1139.8 ±4.8 Ma by

late-stage kimberlite and carbonatite magmas forming discrete 0.5 to 5 m wide dykes
AC

within the lower diatreme. On the basis of petrography and bulk rock major element

compositions, the kimberlite dykes represent archetypal Group-1 kimberlite. Their

mineral compositions, however, show marked deviations from trends that are

considered diagnostic for Group-1 kimberlite in mineralogical-genetic classification

schemes for volatile-rich ultramafic rocks. For example, groundmass spinel

 ACCEPTED MANUSCRIPT

compositions are transitional between magnesian ulvöspinel (a hallmark feature of

Group-1 kimberlite) and titanomagnetite trends, the latter being more diagnostic for

aillikite, orangeite and lamproite. The Premier kimberlite dykes contain groundmass

phlogopite that is much lower in alumina than considered diagnostic for archetypal

kimberlite. It evolves by Al- and Ba-depletion to tetraferriphlogopite, a compositional

T
trend that is more typical for orangeite and aillikite. Although ilmenite macrocrysts-

IP
antecrysts from the Premier kimberlite dykes are characteristically Mg-rich, they contain

CR
up to 5 wt.% MnO, which is more typical for carbonate-rich magmatic systems such as

aillikite and carbonatite. Mn-rich ilmenite also occurs in the groundmass of the Premier

US
carbonatite dykes, which suggests a petrogenetic link to the kimberlite dykes by late-
AN
stage fractionation processes involving development of residual carbonate-rich melts

and fluids. Although mineralogical-genetic classification schemes for discrimination

M

between kimberlite groups and the various alkaline ultramafic rock suites provide an
ED

elegant approach that circumvents the common issues of magma contamination by

mantle-derived debris as well as heteromorphic reactions caused by fluctuating volatile

PT

contents, many of the ‘well-established’ mineral compositional trends may not be as

CE

robust and diagnostic as previously thought.

AC

Utilizing an experimentally constrained CO2 degassing model, it is suggested that the

Premier kimberlite dyke magmas have lost between 10 and 20 wt.% CO2 during ascent,

whereas CO2-loss for the carbonatite dykes was between 5 and 30 wt.%. Comparatively

low fO2 values down to -5.6 ∆NNO are obtained for the kimberlite dykes when

monticellite oxybarometry is applied. This probably reflects significant CO2 degassing

 ACCEPTED MANUSCRIPT

during magma ascent rather than the redox conditions of the parental magma and its

ultimate mantle source. If correct, then monticellite oxybarometry – and groundmass

mineral-based fO2 estimates in general - have little value in assessing the preservation

potential of diamonds in ascending kimberlite magmas.

T
After correction for olivine accumulation and CO2-loss, there still remains a wide gap

IP
between the primitive kimberlite and carbonatite melt compositions at Premier, which

CR
suggests that these magma types cannot be linked by variably low degrees of partial

melting of the same carbonated peridotite source in the deep upper mantle. Rather, an

US
important role is ascribed to fractionation processes by which carbonate-rich residual
AN
melts/fluids developed from mantle-derived evolving kimberlite magma.
M

Acknowledgments
ED

This study was supported by the National Research Foundation of South Africa through

IPRR and DST-NRF CIMERA grants to ST. Additional funding was provided by the
PT

Geological Society of South Africa via a REI grant to ST. We thank Petra Diamonds Ltd,
CE

in particular Andrew Rogers, Anton Wolmarans and Theo Phahla, for facilitating

sampling at Cullinan Diamond Mine in 2015/2016. Special thanks go to Drs. Christian

AC

Reinke and Willem Oldewage for maintenance and management of functioning EPMA

and XRF laboratories at the University of Johannesburg. Malcolm Massuyeau is

sincerely thanked for help with the CO2 degassing model. Discussions with Jock Robey,

Andrew Rogers, Craig Smith, Fanus Viljoen, and Bruce Kjarsgaard are gratefully
 ACCEPTED MANUSCRIPT

acknowledged. Highly constructive comments by the journal reviewers Andrea Giuliani

and Dima Kamenetsky, as well as by editor Andrew Kerr were much appreciated.

References

T
Abersteiner, A., Kamenetsky, V. S., Goemann, K., Giuliani, A., Howarth, G. H., Castillo-Oliver,

IP
M., Thompson, J., Kamenetsky, M., and Cherry, A., 2019, Composition and

CR
emplacement of the Benfontein kimberlite sill complex (Kimberley, South Africa):

Textural, petrographic and melt inclusion constraints: Lithos, v. 324, 297-314.

US
Abersteiner, A., Kamenetsky, V.S., Pearson, D.G., Kamenetsky, M., Goemann, K., Ehrig, K.,

Rodemann, T., 2018. Monticellite in group-I kimberlites: Implications for evolution of

AN
parental melts and post-emplacement CO2 degassing. Chemical Geology 478, 76-88.
M

Agashev, A.M., Pokhilenko, N.P., Takazawa, E., McDonald, J.A., Vavilov, M.A., Watanabe, T.,

Sobolev, N.V., 2008. Primary melting sequence of a deep (> 250 km) lithospheric mantle
ED

as recorded in the geochemistry of kimberlite–carbonatite assemblages, Snap Lake

PT

dyke system, Canada. Chemical Geology 255, 317–328

Allsopp, H.L., Burger, A.J., Van Zyl, C., 1967. A minimum age for the Premier kimberlite pipe
CE

yielded by biotite Rb/Sr measurements, with related galena isotopic data. Earth and

Planetary Science Letters 3, 161–166.

AC

Aulbach, S., Sun, J., Tappe, S., Höfer, H. E., and Gerdes, A., 2017. Volatile-rich metasomatism

in the cratonic mantle beneath SW Greenland: Link to kimberlites and mid-lithospheric

discontinuities: Journal of Petrology 58 (12), 2311-2338.

Bartlett, P.J., 1994. Geology of the Premier diamond pipe. In: C.R. Anhaeusser (Editor)

Proceedings XVth CMMI Congress, South African Institute of Mining and Metallurgy,

Symposium Series S14-3, 201-214.

 ACCEPTED MANUSCRIPT

Bell, K., Simonetti, A., 2010. Source of parental melts to carbonatites: critical isotopic

constraints. Mineralogy and Petrology 98, 77-89.

Bellis, A.J., Canil, D. 2007. Ferric iron in CaTiO 3 perovskite as an oxygen barometer for

kimberlitic magmas I: experimental calibration. Journal of Petrology 48, 219–230.

Birkett, T.C., McCandless, T.E., Hood, C.T., 2004. Petrology of the Renard igneous bodies:

host rocks for diamond in the northern Otish Mountains region, Quebec. Lithos 76, 475–

T
IP
490.

CR
Brooker, R. A., Kjarsgaard, B.A., 2011. Silicate–Carbonate Liquid Immiscibility and Phase

Relations in the System SiO2–Na2O–Al2O3–CaO–CO2 at 0·1–2·5 GPa with

US
Applications to Carbonatite Genesis. Journal of Petrology 52 (7-8), 1281 - 1305.

Brooker, R.A., Sparks, R.S.J., Kavanagh, J.L., Field, M., 2011. The volatile content of
AN
hypabyssal kimberlite magmas: Some constraints from experiments on natural rock

compositions. Bulletin of Volcanology 73, 959-981.

M

Bussweiler, Y., Foley, S.F., Prelević, D., Jacob, D.E., 2015. The olivine macrocryst problem:
ED

new insights from minor and trace element compositions of olivine from Lac de

Gras kimberlites, Canada. Lithos 220–223, 238–252.

PT

Canil, D., Bellis, A.J., 2007. Ferric iron in CaTiO 3 perovskite as an oxygen barometer for
CE

kimberlite magmas II: applications. Journal of Petrology 48: 231-252

Carmody, L., Taylor, L.A., Thaisen, K.G., Tychkov, N.S., Bodnar, R.J., Sobolev, N.V.,
AC

Pokhilenko, L.N., Pokhilenko, N.P., 2014. Ilmenite as diamond indicator mineral in the

Siberian Craton: a tool to predict diamond potential. Economic Geology 109, 775–783.

Castillo-Oliver, M., Giuliani, A., Griffin, W.L., O’Reilly, S.Y., 2018. Characterization of primary

and secondary carbonates in hypabyssal kimberlites: an integrated compositional and

Sr-isotopic approach. Mineralogy and Petrology. doi.org/10.1007/s00710-018-0626-3

 ACCEPTED MANUSCRIPT

Castillo-Oliver, M., Melgarejo, J.C., Galí, S., Pervov, V., Gonçalves, A.O., Griffin, W.L., Pearson,

N.J., O'Reilly, S.Y., 2017. Use and misuse of Mg- and Mn-rich ilmenite in diamond

exploration: A petrographic and trace element approach, Lithos 292–293, 348–363.

Chalapathi Rao, N.V., Paton, C., Lehmann, B., 2012. Origin and diamond prospectivity of

Mesoproterozoic kimberlites from the Narayanpet field, Eastern Dharwar Craton,

T
IP
southern India: insights from groundmass mineralogy, bulk‐chemistry and perovskite

CR
oxybarometry. Geological Journal 47(2-3), 186-212

Clement, C.R., Skinner, E.M.W., 1985. Geochemical character of southern African kimberlites: a

US
new approach based on isotopic constraints. Transactions Geological Society of South

Africa 88(2), 267-280.

AN
Dalton, J.A., Presnall, D.C. 1998. The continuum of primary carbonatitic–kimberlitic melt

compositions in equilibrium with lherzolite: data from the system CaO–MgO–Al2O3–

M

SiO2–CO2 at 6 GPa. Journal of Petrology 39, 1953–1964

ED

Daly, R.A., 1925. Carbonate dykes of the Premier Diamond Mine, Transvaal. The Journal of

Geology 3, 659-684
PT

Danchin, R.V., 1979. Mineral and bulk chemistry of garnet lherzolite and garnet harzburgite
CE

xenoliths from the Premier mine, South Africa. In: Boyd, F.R., Meyer, H.O.A. (Eds.), The

Mantle Sample: Inclusions in Kimberlites and other Volcanics. Proceedings of the 2nd
AC

International Kimberlite Conference, vol. 2. American Geophysical Union, Washington,

pp. 104-126.

Dawson, J.B., 1980. Kimberlites and their xenoliths. Springer-Verlag, Berlin, Germany, 252p.

De Kock, M. O., Ernst, R., Söderlund, U., Jourdan, F., Hofmann, A., Le Gall, B., Bertrand, H.,

Chisonga, B. C., Beukes, N., Rajesh, H. M., Moseki, L. M., & Fuchs, R., 2014. Dykes of

the 1.11 Ga Umkondo LIP, Southern Africa: Clues to a complex plumbing system.

Precambrian Research 249, 129–143.

 ACCEPTED MANUSCRIPT

De Wit, M., Bhebhe, Z., Davidson, J., Haggerty, S.E., Hundt, P., Jacob, J., Lynn, M., Mar-shall,

T.R., Skinner, C., Smithson, K., Stiefenhofer, J., Robert, M., Revitt, A., Spaggiari, R.,

Ward, J., 2016. Overview of diamond resources in Africa. Episodes 39, 199–237.

Deines, P., Gold, D.P., 1973. The isotopic composition of carbonatite and kimberlite carbonates

and their beraring on the isotopic composition of deep seated carbon. Geochimica et

Cosmochimica Acta 37, 1709-1733.

T
IP
Dongre, A.N.; Viljoen, K.S.; Chalapathi Rao, N.V.; Gucsik, A., 2016. Origin of Ti-rich garnets in

CR
the groundmass of Wajrakarur field kimberlites, southern India: Insights from EPMA and

Raman spectroscopy. Mineralogy and Petrology 110, 295–307.

US
Fedortchouk, Y., Canil, D., Carlson, J.A., 2005. Dissolution forms in Lac de Gras diamonds and

their relationship to the temperature and redox state of kimberlite magma. Contributions
AN
to Mineralogy and Petrology 150, 1725–1745.

Field M., Stiefenhofer J., Robey J. and Kurszlaukis S., 2008. Kimberlite-hosted diamonds in
M

Southern Africa: A review. Ore Geology Reviews, 34, 33-75.

ED

Fritschle, T., Prelevic, D., Foley, S.F., Jacob, D.E., 2013. Petrological characterization of the

mantle source of Mediterranean lamproites: Indications from major and trace elements
PT

of phlogopite. Chemical Geology 353, 267-279.

CE

Gaspar, J.C., Wyllie, P.J., 1984. The alleged kimberlite-carbonatite relationship: evidence from

ilmenite and spinel from Premier and Wesselton mines and the Benfontein sill, South
AC

Africa. Contributions to Mineralogy and Petrology 85(2), 133-140.

Giuliani, A., 2018. Insights into kimberlite petrogenesis and mantle metasomatism from a review

of the compositional zoning of olivine in kimberlites worldwide. Lithos 312-313, 322-342.

Giuliani, A., Phillips, D., Kamenetsky, V.S., Goemann, K., 2016. Constraints on kimberlite

ascent mechanisms revealed by phlogopite compositions in kimberlites and mantle

xenoliths. Lithos 240–243, 189–201.

 ACCEPTED MANUSCRIPT

Giuliani, A., Phillips, D., Woodhead, J.D., Kamenetsky, V.S., Fiorentini, M.L., Maas, R., Soltys,

A., Armstrong, R.A., 2015. Did diamond-bearing orangeites originate from MARID-

veined peridotites in the lithospheric mantle?. Nature Communication 6, 1–10.

Giuliani, A., Soltys, A., Phillips, D., Kamenetsky, V.S., Maas, R., Goemann, K., Woodhead, J.

D., Drysdale, R.N., Griffin, W.L., 2017. The final stages of kimberlite petrogenesis:

petrography, mineral chemistry,melt inclusions and Sr-C-O isotope geochemistry of the

T
IP
Bultfontein kimberlite (Kimberley, South Africa). Chemical Geology 455, 342–356.

CR
Gregoire, M., Tinguely, C., Bell, D.R., le Roex, A.P., 2005. Spinel lherzolite xenoliths from the

Premier kimberlite (Kaapvaal craton, South Africa): Nature and evolution of the shallow

US
upper mantle beneath the Bushveld complex. Lithos 84 (3-4), 185-205.

Griffin, W.L., Batumike, J.M., Greau, Y., Pearson, N.J., Shee, S.R., O'Reilly, S.Y., 2014.
AN
Emplacement ages and sources of kimberlites and related rocks in southern Africa: U–

Pb ages and Sr–Nd isotopes of groundmass perovskite. Contributions to Mineralogy

M

and Petrology 168, 1032–1045.

ED

Howarth, G.H., Taylor, L.A., 2016. Multi-stage kimberlite evolution tracked in zoned olivine

from the Benfontein sill, South Africa. Lithos 262, 384–397.

PT

Hutchison, M.T., Faithfull, J.W., Barfod, D.N., Hughes, J.W., Upton, B.G.J., 2018. The mantle of
CE

Scotland viewed through the Glen Gollaidh aillikite. Mineralogy and Petrology 112(1),

115-132.
AC

Jaques, A.L., Haggerty, S.E., Lucas, H., and Boxer, G.L., 1989. The mineralogy and petrology

of the Argyle (AK1) lamproite pipe, Western Australia. Kimberlites and Related Rocks,

Vol. 1, Their Composition, Occurrence, Origin, and Emplacement (Eds. J. Ross and

others), Geological Society of Australia Special Publication 14, p.153-169.

Jelsma, H.A., de Wit, M.J., Thiart, C., Dirks, P.H.G.M., Viola, G., Basson, I.J., Anckar, E., 2004.

Preferential distribution along transcontinental corridors of kimberlites and related rocks

of southern Africa. South African Journal of Geology 107, 301–324

 ACCEPTED MANUSCRIPT

Jones, A.P., Genge, M., Carmody, L., 2013. Carbonate melts and carbonatites. In: Hazen, R.M.,

Jones, A.P., Baross, J.A. (Eds.), Carbon in Earth. Mineralogical Society America,

Chantilly, pp.289–322.

Kamenetsky, V.S., Kamenetsky, M.B., Sobolev, A.V., Golovin, A.V., Demouchy, S., Faure, K.,

Sharygin, V.V., Kuzmin, D.V., 2008. Olivine in the Udachnaya-East Kimberlite (Yakutia,

Russia): types, compositions and origins. Journal of Petrology 49, 823–839.

T
IP
Kamenetsky, V.S., Yaxley, G.M., 2015. Carbonate–silicate liquid immiscibility in the mantle

CR
propels kimberlite magma ascent. Geochimica et Cosmochimca. Acta 158, 48–56.

Kjarsgaard, B.A., Pearson, D.G., Tappe, S., Nowell, G.M., Dowall, D.P., 2009. Geochemistry of

US
hypabyssal kimberlites from Lac de Gras, Canada: Comparisons to a global database

and applications to the parent magma problem. Lithos 112S, 236–248.

AN
Kramers, J.D., Smith, C.B., 1983. A feasibility study of U/Pb and Pb/Pb dating of kimberlites

using groundmass mineral fractions and whole-rock samples. Chemical Geology 1,

M

23–38.
ED

Le Pioufle, A., Canil, D., 2012. Iron in monticellite as an oxygen barometer for kimberlite

magmas. Contributions to Mineralogy and Petrology 163(6). 1033-1046

PT

Lim, E., Giuliani, A., Phillips, D., and Goemann, K., 2018, Origin of complex zoning in olivine
CE

from diverse, diamondiferous kimberlites and tectonic settings: Ekati (Canada), Alto

Paranaiba (Brazil) and Kaalvallei (South Africa): Mineralogy and Petrology, v. 112, p.
AC

539-554.

Malarkey, J., Pearson, D.G., Kjarsgaard, B.A., Davidson, J.P., Nowell, G.M., Ottley, C.J.,

Stammer, J., 2010. From source to crust: tracing magmatic evolution in a kimberlite and

a melilitite using microsample geochemistry. Earth and Planetary Science Letters 299,

80–90.

Massuyeau, M., Gardès, E., Morizet, Y., Gaillard, F., 2015. A model for the activity of silica

along the carbonatite-kimberlite-melilitite-basanite melt compositional joint. Chemical

 ACCEPTED MANUSCRIPT

Geology 418, 206-216.

Massuyeau, M., Tappe, S., and Viljoen, S. K., 2019, A thermodynamic model with CO 2 and

H2O for the lithosphere-asthenosphere boundary beneath thick continental roots,

submitted.

McCormick, G.R., Le Bas, M.J., 1996. Phlogopite crystallization in carbonatitic magmas from

Uganda. Canadian Mineralogist 34, 469-478.

T
IP
Mitchell, R.H., 1978. Manganoan magnesian ilmenite and titanian clinohumite from the

CR
Jacupiranga carbonatite, Sao Paulo, Brazil. American Mineralogist 63, 544-547.

Mitchell, R.H. 1979. The alleged kimberlite–carbonatite relationship: additional contrary

US
evidence. American Journal of Science 279, 570-589.

Mitchell, R.H., 1986. Kimberlites: mineralogy, geochemistry and petrology. Plenum

AN
Press, New York. 442 pp.

Mitchell, R.H., 1995. Kimberlites, orangeites and related rocks. Plenum Press, New York, pp.
M

410
ED

Mitchell, R.H., 2005. Carbonatites and carbonatites and carbonatites. Canadian Mineralogist

43, 2049–2068.
PT

Mitchell, R.H., 2006. Potassic magmas derived from metasomatized lithospheric mantle:
CE

Nomenclature and relevance to exploration for diamond bearing rocks. Journal of the

Geological Society of India 67, 317-327.

AC

Mitchell, R.H.., Bergman, S.C. 1991. Petrology of lamproites, Plenum Press, New York, 408 pp.

Mitchell, R.H., Tappe, S., 2010. Discussion of “Kimberlites and aillikites as probes of the

continental lithospheric mantle”, by D. Francis and M. Patterson. Lithos 109, 72–80

Moore, A.E., 2009. Type II diamonds: Flamboyant Megacrysts? South African Journal of

Geology 112(1), 23-38.

Moussallam, Y., Morizet, Y., Gaillard, F., 2016. H2O–CO2 solubility in low SiO2-melts and
 ACCEPTED MANUSCRIPT

the unique mode of kimberlite degassing and emplacement. Earth and Planetary

Science Letters 447, 151–160.

Nielsen, T.F.D., Jensen, S.M., Secher, K., Sand, K.K., 2009. Distribution of kimberlite and

aillikite in the Diamond Province of southern West Greenland: a regional perspective

based on groundmass mineral chemistry and bulk compositions. Lithos 112, 358–371.

Nielsen, T.F.D., Sand, K.K., 2008. The Majuagaa kimberlite dike, Maniitsoq region, western

T
IP
Greenland: constraints on an Mg-rich silicocarbonatitic melt composition from

CR
groundmass mineralogy and bulk compositions. Canadian Mineralogist 46, 1043–

1061.

US
Nowell, G.M., Pearson, D.G., Bell, D.R., Carlson, R.W., Smith, C.B., Kempton, P.D., Noble,

S.R., 2004. Hf isotope systematics of kimberlites and their megacrysts: new constraints
AN
on their source regions. Journal of Petrology 45, 1583–1612.

Ogilvie-Harris, R.C., Field, M., Sparks, R.S.J., Walter, M.J., 2009. Perovskite from the
M

Dutoitspan kimberlite, Kimberley, South Africa: implications for magmatic processes.

ED

Mineralogical Magazine 73 (6), 915-928.

Phillips, D., Machin, K. J., Kiviets, G. B., Fourie, L. F., Roberts, M. A., and Skinner, E. M. W.,
PT

1998. A petrographic and Ar-40/Ar-39 geochronological study of the Voorspoed

CE

kimberlite, South Africa: Implications for the origin of Group II kimberlite magmatism:

South African Journal of Geology, v. 101, no. 4, p. 299-306.

AC

Pilbeam, L.H., Nielsen, T.F.D., Waight, T.E., 2013. Digestion fractional crystallization (DFC): an

important process in the genesis of kimberlites. Evidence from olivine in the Majuagaa

Kimberlite, Southern West Greenland. Journal of Petrology 54, 1399–1425.

Robinson, D.N., 1975. Magnetite–serpentine–calcite dykes at Premier Mine and aspects of

their relationship to kimberlite and to carbonatite of alkalic carbonatite complexes.

Physics and Chemistry of the Earth 9, 61-70.

Rock, N.M.S., 1991. Lamprophyres. Blackie & Son, Glasgow. 285 pp.
 ACCEPTED MANUSCRIPT

Roeder, P. L., and Schulze, D. J., 2008, Crystallization of groundmass spinel in kimberlite:

Journal of Petrology, v. 49, no. 8, p. 1473-1495.

Sarkar, C., Kjarsgaard, B.A., Pearson, D. G.,Heaman, L. M, Locock, A. J., Armstrong, J. P.,

2018. Geochronology, classification and mantle source characteristics of kimberlites and

related rocks from the Rae Craton, Melville Peninsula, Nunavut, Canada. Mineralogy

and Petrology 112, 653-672.

T
IP
Schmidt, M.W., Weidendorfer, D., 2018. Carbonatites in oceanic hotspots. Geology 46(5), 435-

CR
438.

Scott-Smith, B.H., Nowicki, T.E., Russell, J.K., Webb, K.J., Mitchell, R.H., Hetman, C.M.,

US
Harder, M., Skinner, E.M.W. and Robey, J.V.A., 2013. Kimberlite Terminology and

Classification. In Pearson D.G. et al. (editors), Proceedings of the 10th International

AN
Kimberlite Conference, Vol.2, Special Issue of the Journal of Geological Society of India,

1-17.
M

Shaikh, A.M., Patel, S.C., Bussweiler, Y., Kumar, S.P., Tappe, S., Ravi, S., Mainkar, D., 2019.
ED

Olivine trace element compositions in diamondiferous lamproites from India:

Proxies for magma origins and the nature of the lithospheric mantle beneath the Bastar
PT

and Dharwar cratons. Lithos 324-325, 501-518.

CE

Sheppard, S. M. F., Dawson, J. B., 1975. Hydrogen, carbon and oxygen isotope studies of

megacryst and matrix minerals from Lesothan and South African kimberlites. Physics
AC

and Chemistry of the Earth 9, 747-763.

Skinner, E.M.W., Marsh, J.S., 2004. Distinct kimberlite pipe classes with contrasting eruption

processes. Lithos 76, 183–200.

Smith, C.B., Gurney, J.J., Skinner, E.M.W., Clement, C.R., Ebrahim, N., 1985. Geochemical

character of Southern African kimberlites: a new approach based on isotopic constraints.

Transactions of Geological Society of South Africa 88, 267–280.

 ACCEPTED MANUSCRIPT

Smith, E., Shirey, B.S., Nestola, F., Bullock, E., Wang, J., Richardson, S.H., Wang, W., 2016.

Large gem diamonds from metallic liquid in Earth’s deep mantle. Science 354,1403–

1405

Sokol, A.G., Kupriyanov, I.N., Palyanov, Y.N., 2013. Partitioning of H 2O between olivine and

carbonate–silicate melts at 6.3 GPa and 1400 °C: Implications for kimberlite formation.

Earth and Planetary Science Letters 383, 58–67.

T
IP
Soltys, A., Giuliani, A., and Phillips, D., 2018a, Crystallisation sequence and magma evolution of

CR
the De Beers dyke (Kimberley, South Africa): Mineralogy and Petrology, v. 112, p. 503-

518.

US
Soltys, A., Giuliani, A., Phillips, D., 2018b. A new approach to reconstructing the composition

and evolution of kimberlite melts: a case study of the archetypal Bultfontein kimberlite
AN
(Kimberley, South Africa). Lithos 304-307, 1-15.

Sparks, R.S.J., 2013. Kimberlite volcanism. Annual Review of Earth and Planetary Sciences 41,
M

497–528,
ED

Stamm, N., Schmidt, M.W., 2017. Asthenospheric kimberlites: volatile contents and bulk

compositions at 7 GPa. Earth and Planetary Science Letters 474, 309–321.

PT

Stamm, N., Schmidt, M. W., Szymanowski, D., von Quadt, A., Mohapi, T., and Fourie, A., 2018.

Primary petrology, mineralogy and age of the Letšeng-la-Terae kimberlite (Lesotho,

CE

Southern Africa) and parental magmas of Group-I kimberlites: Contributions to

AC

Mineralogy and Petrology 173 (76). doi.org/10.1007/s00410-018-1502-1

Stone, R. S., and Luth, R. W., 2016. Orthopyroxene survival in deep carbonatite melts:

Implications for kimberlites: Contributions to Mineralogy and Petrology 171(7), 9.

Sun, C. G., and Dasgupta, R., 2019. Slab-mantle interaction, carbon transport, and kimberlite

generation in the deep upper mantle: Earth and Planetary Science Letters 506, 38-

52.

Tappe, S., Dongre, A., Liu, C.Z., Wu, F.Y., 2018a. ‘Premier’ evidence for prolonged kimberlite
 ACCEPTED MANUSCRIPT

pipe formation and its influence on diamond transport from deep Earth. Geology 46(10),

843-846.

Tappe, S., Foley, S.F., Jenner, G.A., Heaman, L.M., Kjarsgaard, B.A., Romer, R.L., Stracke, A.,

Joyce, N., Hoefs, J., 2006. Genesis of ultramafic lamprophyres and carbonatites at Aillik

Bay, Labrador: a consequence of incipient lithospheric thinning beneath the North

Atlantic craton. Journal of Petrology 47, 1261–1315.

T
IP
Tappe, S., Foley, S.F., Jenner, G.A., Kjarsgaard, B.A., 2005. Integrating ultramafic

CR
lamprophyres into the IUGS classification of igneous rocks: rational and implications.

Journal of Petrology 46, 1893–1900.

US
Tappe, S., Foley, S.F., Kjarsgaard, B.A., Romer, R.L., Heaman, L.M., Stracke, A., Jenner, G.A.,

2008. Between carbonatite and lamproites: diamondiferous Torngat ultramafic

AN
lamprophyres formed by carbonate-fluxed melting of cratonic MARID-type metasomes.

Geochimica et Cosmochimica Acta 72, 3258–3286.

M

Tappe, S., Jenner, G.A., Foley, S.F., Heaman, L.M., Besserer, D., Kjarsgaard, B.A., Ryan, B.,
ED

2004. Torngat ultramafic lamprophyres and their relation to the North Atlantic Alkaline

Province. Lithos 76, 491–518.

PT

Tappe, S., Kjarsgaard, B.A., Kurszlaukis, S., Nowell, G.M., Phillips, D., 2014. Petrology and
CE

Nd-Hf isotope geochemistry of the Neoproterozoic Amon kimberlite sills, Baffin Island

(Canada): evidence for deep mantle magmatic activity linked to supercontinent

AC

cycles. Journal of Petrology 55, 2003–2042.

Tappe, S., Romer, R.L., Stracke, A., Steenfelt, A., Smart, K.A., Muehlenbachs, K., Torsvik, T.H.,

2017. Sources and mobility of carbonate melts beneath cratons, with implications for

deep carbon cycling, metasomatism and rift initiation. Earth and Planetary Science

Letters 466, 152-167.

Tappe, S., Smart, K.A., Torsvik, T.H., Massuyeau, M., de Wit, M.C.J., 2018b. Geodynamics of

kimberlites on a cooling Earth: Clues to plate tectonic evolution and deep volatile cycles.
 ACCEPTED MANUSCRIPT

Earth and Planetary Science Letters 484, 1-14.

Tappe, S., Steenfelt, A., Heaman, L.M., Simonetti, A., 2009. The newly discovered Jurassic

Tikiusaaq carbonatite-aillikite occurrence, West Greenland, and some remarks on

carbonatite-kimberlite relationships. Lithos 112, 385–399.

Tompkins, L.A., Haggerty, S.E., 1985. Groundmass oxide minerals in the Koidu kimberlite

dikes, Sierra Leone, West Africa. Contributions to Mineralogy and Petrology 91, 245 –

T
IP
263.

CR
Viljoen, F., Dobbe, R., Smit, B., 2009. Geochemical processes in peridotite xenoliths from the

Premier diamond mine, South Africa: evidence for the depletion and refertilisation of

US
subcratonic lithosphere. Lithos 112, 1133-1142

Wagner, P.A., 1914. The diamond fields of Southern Africa. 2nd Edition 1971. C Struik (Pty) Ltd,
AN
Cape Town, 355 p.

Woodhead, J., Hergt, J., Phillips, D., and Paton, C., 2009, African kimberlites revisited: In situ
M

Sr-isotope analysis of groundmass perovskite: Lithos, v. 112, p. 311-317.

ED

Woolley, A.R., Kjarsgaard, B.A., 2008. Paragenetic types of carbonatite as indicated by the

diversity and relative abundances of associated silicate rocks: Evidence from a global
PT

database. Canadian Mineralogist 46(4), 741-752.

CE

Woolley, A.R., Bergman, S.C., Edgar, A.D., Le Bas, M.J., Mitchell, R.H., Rock, N.M.S., Scott-

Smith, B.H., 1996. Classification of lamprophyres, lamproites, kimberlites and the

AC

kalsilitic, melilitic and leucitic rocks. Canadian Mineralogist 34, 175-186.

Wu, F.-Y, Mitchell, R.H., Li, Q., Sun, J., Liu, C. and Yang, Y-H., 2013. In situ U-Pb age

determination and Sr-Nd isotope analysis of perovskite from the Premier (Cullinan)

kimberlite, South Africa. Chemical Geology 353, 83-95.

Wyatt, B. A., 1979. Manganoan ilmenite from the Premier kimberlite. Kimberlite Symposium II.

Cambridge, Extended Abstracts.

Yoder, H. S., Jr., 1979. Melilite-bearing rocks and related lamprophyres, in Yoder, H. S., Jr., ed.,
 ACCEPTED MANUSCRIPT

The evolution of the igneous rocks: Fiftieth anniversary perspectives: Princeton, N.J.,

United States, Princeton University Press, p. 391-411.

Zurevinski, S.E., Mitchell, R.H., 2011. Highly evolved hypabyssal kimberlite sills from Wemindji,

Quebec, Canada: insights into the process of flow differentiation in kimberlite magmas.

Contributions to Mineralogy and Petrology 161, 765–776.

T
IP
CR
Figure captions

Figure 1: (A) Reconstructed geological cross-section of the Premier kimberlite pipe,

US
Cullinan Diamond Mine, South Africa (modified from Tappe et al., 2018a). Samples

investigated during this study were taken from various underground mining levels as
AN
indicated in red. (B) Map of South Africa showing the position of the Kaapvaal craton
M

and surrounding tectonic belts, as well as the locations of major Group-1 (blue) and
ED

Group-2 (red) kimberlite clusters (modified after De Wit et al., 2016). The ca. 1153 Ma

Premier/Cullinan kimberlite pipe on the central Kaapvaal craton is marked with a blue
PT

asterisk.
CE

Figure 2: Back-scattered electron (BSE) images showing: (A) Rounded to

AC

subhedral/euhedral olivine macrocrysts and microcrysts set in monticellite-dominated

groundmass of kimberlite dyke CIM15-74; (B) Close-up of a several mm-sized unzoned

olivine macrocryst in kimberlite dyke CIM15-76. Potential olivine overgrowth margins

and rinds are probably eradicated by serpentinization (dark grey); (C) Mosaic of

groundmass monticellite crystals with incipient alteration (light grey) along their margins
 ACCEPTED MANUSCRIPT

in kimberlite dyke CIM15-72; (D) Monticellite overgrowth on a serpentinized olivine

microcryst in kimberlite dyke CIM15-74. Ol – olivine, Mtc – monticellite.

Figure 3: (A) NiO and (B) MnO versus Mg-number (Mg#) for olivine macrocrysts and

microcrysts in various Premier kimberlite dykes. Fields for olivine megacrysts from

T
kimberlites and olivine from mantle-derived peridotite xenoliths from cratons worldwide

IP
are adopted from Giuliani (2018). Fields for olivine from mantle-derived sheared and

CR
granular peridotite xenoliths from the Premier kimberlite pipe are based on our own

unpublished data (S. Tappe, unpublished). (C) Compositional variation of groundmass

US
monticellite from the Premier kimberlite dykes in the ternary system forsterite (Mg2SiO4),
AN
monticellite (CaMgSiO4) and kirschsteinite (CaFeSiO4). The field for monticellite

compositions from global magmatic kimberlites is based on data from Mitchell (1986)
M

and various other literature sources referred to in the main text.

ED

Figure 4: Back-scattered electron (BSE) images showing: (A) Poikilitic inclusions of

PT

spinel, perovskite and monticellite in a large plate of groundmass phlogopite in

CE

kimberlite dyke CIM15-74; (B) Cr-spinel crystal included in a phlogopite groundmass

plate suggesting simultaneous growth in kimberlite dyke CIM15-74; (C) Elongated

AC

microphenocrysts of phlogopite in close textural relationship with groundmass spinel

and perovskite in kimberlite dyke CIM15-75; (D) Clusters of high-Ba phlogopite flakes

(bright) in close textural relationship with serpophitic serpentine in the groundmass of

the Piebald kimberlite plug unit (sample CIM15-85). Phl - phlogopite, Spl - spinel, Mag -

magnetite, Prv - perovskite, Mtc - monticellite, Srp - serpentine, Cb - carbonate.

 ACCEPTED MANUSCRIPT

Figure 5: Phlogopite compositional trends for the kimberlite dykes and Piebald

kimberlite plug variety of Premier pipe. Data for selected Grey volcaniclastic kimberlite

samples from Premier pipe are also shown: (A) Al2O3 versus TiO2, and (B) Al2O3 versus

FeOT variations of groundmass phlogopite. Devised fields for phlogopite from magmatic

T
kimberlite, orangeite and lamproite are adopted from Mitchell (1995). Although the

IP
Piebald hypabyssal kimberlite plug variety contains groundmass phlogopite that has

CR
alumina-rich compositions typical for Group-1 kimberlite, the Premier kimberlite dykes

contain groundmass phlogopite that is generally alumina-poor and evolves by Al-

US
depletion toward tetraferriphlogopite (TFP), which is more typical for orangeite and
AN
lamproite.
M

Figure 6: (A) Groundmass phlogopite compositions from various Premier kimberlite

ED

varieties projected into the ternary system Al-Mg-FeT on an atomic basis (cations per 22

oxygen atoms). Total Fe expressed as Fe2+. The phlogopite macrocryst field for
PT

kimberlites is adopted from Mitchell (1995). EAST - eastonite, SID - siderophyllite, PHL -
CE

phlogopite, ANN - Annite, TFP – tetraferriphlogopite. (B) Fluorine versus BaO, and (C)

K versus Ba variations of groundmass phlogopite from various Premier kimberlite

AC

varieties. Note the coupled and elevated F and Ba concentrations of phlogopite from the

Piebald kimberlite plug variety.

Figure 7: Back-scattered electron (BSE) images showing: (A) Close textural

relationship between groundmass spinel and perovskite in kimberlite dyke CIM15-72

 ACCEPTED MANUSCRIPT

(SIMS U/Pb perovskite age reported in Tappe et al., 2018a); (B) Cr-spinel crystals

included in a phlogopite groundmass plate suggesting simultaneous growth in kimberlite

dyke CIM15-72; (C) Intergrowth between groundmass spinel and phlogopite in

kimberlite dyke CIM15-75; (D) Abundant atoll-textured spinel in the groundmass of

kimberlite dyke CIM15-75. Phl - phlogopite, Spl - spinel, Prv – perovskite.

T
IP
Figure 8: (A) Ti/(Ti+Cr+Al) versus FeT2+/(FeT2++Mg) variations of groundmass spinel

CR
crystals from various Premier kimberlite varieties and the carbonatite dykes.

Groundmass spinel from the kimberlite dykes is transitional between the magnesian

US
ulvöspinel and titanomagnetite trends, whereas spinel from the carbonatite dykes
AN
follows the titanomagnetite trend. Spinel crystals from the Grey volcaniclastic kimberlite

variety have similar Cr-rich core compositions compared with spinel cores from the
M

kimberlite dykes, but they show distinctly different rim compositions approaching the
ED

magnetite end-member. Fields for spinel trends are adopted from Mitchell (1986) and

(1995). (B) Ilmenite macrocryst-antecryst and groundmass compositions from the

PT

Premier kimberlite and carbonatite dykes projected into the ternary system geikielite
CE

(MgTiO3) - ilmenite (FeTiO3) - pyrophanite (MnTiO3) on an atomic basis. Note the Mn-

rich nature of both kimberlitic and carbonatitic ilmenite from Premier pipe. Fields for
AC

ilmenite from global kimberlites and carbonatites (Mitchell, 1995), and from Premier pipe

(Wyatt, 1979; Gaspar and Wyllie, 1984) are shown for comparison.

Figure 9: (A) CaO versus MgO variations for groundmass calcite from the Premier

kimberlite and carbonatite dykes. Note the slightly elevated Mg contents of calcite in
 ACCEPTED MANUSCRIPT

carbonatites, as also observed in the data from two Premier carbonatite dykes reported

in Castillo-Oliver et al. (2018). (B) Compositions of secondary andradite garnet grains

from the Premier carbonatite dykes projected into the ternary system TiO 2-CaO-FeOT

on an oxide wt.% basis. The compositional field for andradite garnet from orangeites

and ultramafic lamprophyres is based on data reported in Tappe et al. (2004, 2006,

T
2009) and Dongre et al. (2016).

IP
CR
Figure 10: Bulk rock major element variation diagrams for various Premier kimberlite

varieties and the carbonatite dykes. Reference data for global magmatic kimberlites are

US
from the compilation downloadable in Tappe et al. (2017). Selected reconstructed
AN
parental kimberlite melt estimates are taken from Kjarsgaard et al. (2009) and Soltys et

al. (2018b).
M
ED

Figure 11: (A) Bulk rock TiO2 versus K2O variation of the Premier kimberlite dykes and

selected Grey volcaniclastic kimberlite samples. The fields for South African kimberlites
PT

and orangeites, as well as the divide between them are after Smith et al. (1985). The
CE

field for North American kimberlites is based on compilations in Kjarsgaard et al. (2009)

and Tappe et al. (2017). (B) Estimated log fO2 oxygen fugacity conditions (ΔNNO;
AC

derived from groundmass monticellite and perovskite) for the Premier kimberlite dykes

(this study) compared with cratonic mantle lithosphere, various non-cratonic mantle-

derived magma types, and global kimberlites (adapted from Canil and Bellis, 2007; Le

Pioufle and Canil, 2012). Data sources are as follows: Dutoitspan kimberlite (Ogilvie-

Harris et al., 2009); Narayanpet kimberlite (Chalapathi Rao et al., 2012); Lac de Gras
 ACCEPTED MANUSCRIPT

and Somerset Island kimberlites (Le Pioufle and Canil, 2012); Tikiusaaq kimberlite

dykes (Tappe et al., 2009, 2017). Prv - perovskite, Mtc – monticellite.

Figure 12: Bulk rock CO2 versus SiO2 variation of Premier kimberlite and carbonatite

dykes. The blue open circles represent the compositions of kimberlite dykes corrected

T
for olivine accumulation and CO2-loss assuming a near-primary liquid composition of 22

IP
wt.% MgO. For the carbonatite dykes, no correction was done for olivine accumulation.

CR
The empirical inverse relationship between CO 2 and SiO2 contents of primitive mantle-

derived melts is based on the experimental database for liquids in equilibrium with

US
olivine and orthopyroxene of Massuyeau et al. (2015); that is, for melts derived from
AN
CO2-fluxed peridotites such as kimberlites. The database for global magmatic

kimberlites (grey circles) is from Tappe et al. (2017). Note the pronounced
M

compositional gap between the kimberlite and carbonatite dykes from Premier before
ED

and after the olivine and CO2 correction procedure. For further details see Section 5.2.2

of the main text.

PT
CE
AC
 ACCEPTED MANUSCRIPT

Table 1: Details of the kimberlite and carbonatite dykes at Cullinan Diamond Mine

Dyke Macroscopic
No. Rock type Mineralogy Depth
sample features
Monticellite, Olivine,
Dark grey/ black, Phlogopite, Spinel,
1 CIM-15-72 Kimberlite 645 m
macrocrystic Perovskite, Carbonate,
Ilmenite
Monticellite, Olivine,
Light grey, fine
2 CIM-15-74 Kimberlite Spinel, Perovskite, 645 m

T
grained
Phlogopite

IP
Spinel, Olivine
(serpentinised),
Dark grey/ black,

CR
3 CIM-15-75 Kimberlite Phlogopite, Perovskite, 645 m
fine grained
Carbonate, Ilmenite,
Apatite

US
Olivine, Monticellite,
Dark grey,
4 CIM-15-76 Kimberlite Phlogopite, Perovskite, 645 m
macrocrystic
Spinel, Ilmenite
Monticellite, Olivine
AN
(serpentinised),
Dark grey/
5 CIM-15-80 Kimberlite Phlogopite, Spinel, 645 m
macrocrystic
Perovskite, Ilmenite,
M

Carbonate
Phlogopite, Carbonate,
CIM-16- Light grey, weakly
ED

6 Kimberlite Spinel, Serpentine, 717 m

004 macrocrystic
Perovskite, Ilmenite
Light grey, fine Carbonate, Andradite,
7 CIM-15-70 Carbonatite 824 m
PT

grained Altered olivine

Dark grey, fine Altered olivine,
8 CIM-15-71 Carbonatite 645 m
grained Carbonate, Andradite
CE

Light grey, fine Carbonate, Altered

9 CIM-15-73 Carbonatite 645 m
grained olivine, Andradite
Light grey, fine Carbonate, Andradite,
10 CIM-15-77 Carbonatite 645 m
AC

grained Serpentine
Medium grey, fine Carbonate, Andradite,
11 CIM-15-78 Carbonatite 645 m
grained Spinel
Dark grey, fine Carbonate, Serpentine,
12 CIM-15-79 Carbonatite 645 m
grained Andradite, Spinel
CIM-16- Medium grey, Carbonate, Serpentine,
13 Carbonatite 763 m
006 mosaic textured Andradite, Spinel
Medium grey/
Carbonate, Serpentine,
brown, mosaic
14 CIM-16-12 Carbonatite Andradite, Ilmenite, 839 m
textured to
Spinel
segregationary
 ACCEPTED MANUSCRIPT

Table 2: Bulk chemical compositions (XRF results) of the dykes and other units of Premier
kimberlite

Rock
type Kimberlite dykes Carbonatite dykes
CIM15- CIM15- CIM15- CIM15- CIM15- CIM16- CIM15- CIM15- CIM15- CIM15- CIM
Sample 72 74 75 76 80 004 71 73 77 78 79
SiO2 31.70 33.96 31.64 31.11 31.83 36.09 12.30 11.61 12.15 12.68 12

T
TiO2 2.19 2.38 2.21 2.23 2.31 2.63 1.81 1.57 1.74 1.73 1.
Al2O3 1.97 2.49 1.95 2.05 2.05 2.23 1.11 1.27 1.01 0.97 1.

IP
Fe2O3 9.75 10.45 10.18 9.30 10.03 10.40 13.40 11.38 12.81 13.54 12
MgO 31.09 31.36 32.36 31.23 31.54 31.41 24.00 26.32 30.62 29.45 27

CR
MnO 0.16 0.17 0.17 0.16 0.17 0.16 0.27 0.34 0.39 0.37 0.
CaO 10.07 9.22 7.05 14.29 9.00 4.49 23.19 22.59 20.27 20.98 20
K2O 1.23 1.77 0.13 0.67 1.23 0.25 n.d. 0.05 n.d. n.d. n.

US
Na2O 0.00 0.00 0.00 0.00 0.00 0.00 n.d. n.d. n.d. n.d. n.
Cr2O3 0.24 0.24 0.23 0.20 0.21 0.22 n.d. n.d. n.d. n.d. n.
AN
P2O5 0.13 0.23 0.24 0.14 0.13 0.29 0.39 1.07 2.85 3.42 0.
NiO 0.16 0.16 0.17 0.17 0.17 0.16 n.d. n.d. n.d. n.d. n.
BaO 0.07 0.10 0.07 0.07 0.07 0.06 0.07 0.06 0.07 0.11 0.
M

LOI 11.85 8.30 14.18 8.48 11.62 12.23 23.52 23.96 17.65 16.85 22
Sum 100.67 100.82 100.71 100.11 100.42 100.78 100.06 100.23 99.56 100.10 10
ED

C.I. 1.0 1.1 1.0 1.0 1.0 1.2

CO2 2.8 0.53 4.01 0.32 2.3 2.43 13.69 12.93 3.51 3.56 10
PT

Rock
type Piebald Volcaniclastic kimberlite
CIM15- CIM16- CIM16- CIM16-
CE

Sample 85 008 009 013

SiO2 37.13 44.72 46.72 45.17
TiO2 2.14 1.61 2.07 2.00
AC

Al2O3 3.17 4.20 4.46 4.34

Fe2O3 8.70 8.81 8.93 9.12
MgO 27.63 24.96 23.06 23.96
MnO 0.14 0.13 0.15 0.14
CaO 10.37 6.77 6.91 7.27
K2O 0.68 0.97 1.18 1.08
Na2O 0.00 0.90 0.85 0.66
Cr2O3 0.15 0.14 0.14 0.17
P2O5 0.37 0.31 0.28 0.29
NiO 0.15 0.14 0.13 0.12
 ACCEPTED MANUSCRIPT

BaO 0.13 0.07 0.07 0.09

LOI 9.33 6.07 4.91 5.83
Sum 100.14 99.89 99.95 100.31
C.I. 1.4 1.9 2.1 2.0
CO2 0.56 n.a. n.a. n.a.

n.d. = not detected, or means below the XRF detection limit; n.a.= not analyzed; LOI = loss on
ignition

T
C.I. = Contamination Index (C.I. = SiO2+Al2O3+Na2O)/(MgO+2*K2O)

IP
CR
US
AN
M
ED
PT
CE
AC
 ACCEPTED MANUSCRIPT

 Kimberlite and carbonatite dykes at Premier pipe linked by fractionation

processes

 Kimberlite magma lost 20% CO2 upon ascent through SCLM unrelated to pipe

formation

 Group-1 kimberlites at Premier contain Mn-rich ilmenite and tetraferriphlogopite

T
 Archetypal kimberlite dykes have mineral compositions deviating from Group-1

IP
trends

CR
 Mineralogical classifications of kimberlites show caveats that require revisions

US
AN
M
ED
PT
CE
AC
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6
Figure 7
Figure 8
Figure 9
Figure 10
Figure 11
Figure 12
Figure 13
Figure 14