Lithos 112S (2009) 201–212

Contents lists available at ScienceDirect

Lithos
j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / l i t h o s

Origin of olivine in kimberlite: Phenocryst or impostor?

R.C. Brett a,⁎, J.K. Russell a, S. Moss a,b
a
Volcanology & Petrology Laboratory, Department of Earth & Ocean Sciences, The University of British Columbia, 6339 Stores Road, Vancouver, British Columbia, Canada V6T 1Z4
b
Mineral Deposit Research Unit, Department of Earth & Ocean Sciences, The University of British Columbia, 6339 Stores Road, Vancouver, British Columbia, Canada V6T 1Z4

a r t i c l e i n f o a b s t r a c t

Article history: Kimberlite hosts two populations of olivine that are distinguished on the basis of grain size and morphology;
Received 26 September 2008 the populations are commonly described genetically as xenocrysts and phenocrysts. Olivine xenocrysts or
Accepted 20 April 2009 macrocrysts are thought to derive from disaggregation of mantle xenoliths whereas the smaller, euhedral
Available online 19 May 2009
olivine crystals are presumed entirely cognate to the kimberlite melt. Recent studies of zoning patterns of
euhedral olivine in kimberlite have, however, cast doubt on the actual origins of the smaller olivine crystals.
Keywords:
Orthopyroxene
Here, we elucidate the nature and origins of the textural and chemical zonation that characterize both
Kimberlite populations of olivine: macrocrysts (olivine-I) and euhedral crystals (olivine-II). Specifically, we show that
Olivine both olivine-I and olivine-II feature chemically distinct overgrowths resulting from heterogeneous
Xenocryst crystallization onto pre-existing olivine xenocrysts. Our analysis limits the total volume of olivine crystallized
Phenocryst during transport to ≤ 5% in contrast to previous estimates of ~ 25%. The reduced extent of olivine
Overgrowth crystallization allows for closer reconciliation of crystallized olivine compositions and estimates of Mg#s for
Zoning primitive kimberlite melts. It also places constraints on processes involving orthopyroxene assimilation by
Crystallization
kimberlite melt. If olivine crystallization and orthopyroxene assimilation are coupled, then orthopyroxene
Assimilation
assimilation is limited to ~ 7%. Larger masses of orthopyroxene assimilation (i.e. 25%) are possible only if
Volume
Enthalpy kimberlite magmas originate at super-liquidus (N 100 ºC) conditions and sub-equal amounts of olivine
crystallization occurs.
© 2009 Elsevier B.V. All rights reserved.

1. Introduction
Olivine is the dominant phase in all kimberlite bodies and a
petrogenetic model for kimberlite must, therefore, include a complete
explanation for its origins. The olivine hosted by kimberlite forms two
populations that are distinguished on the basis of grain size and
morphology (Clement et al., 1977; Skinner and Clement, 1979). The
volumetrically important population comprises medium to coarse-
grained, rounded to sub-rounded olivine grains, and are referred to as
macrocrysts (Clement, 1982). The second population features medium
to fine-grained (b0.5 mm), euhedral to subhedral olivine grains and are
commonly designated as phenocrysts. For over 30 years the two
populations of olivine found in kimberlite have been ascribed different
origins, as encapsulated by their labels: the macrocrysts are considered
xenocrystic whilst the phenocrysts are assumed to have crystallized
from the kimberlite melt (Jerram et al., this issue; Field et al., this issue).
Kimberlite from the Diavik Diamond Mine, N.W.T., Canada contains
olivine with rounded to sub-rounded cores featuring distinct overgrowths
of later crystallized olivine. In all samples we observe overgrowths on
both ‘macrocrystic’ and ‘phenocrystic’ olivine crystals. This paper
elucidates the origins of these overgrowths, and explores the implications
⁎ Corresponding author. EOS-UBC, 6339 Stores Road, Vancouver, B.C., Canada V6T
1Z4. Tel.: +1 778 778 3738; fax: +1 604 822 6088.
E-mail addresses: cbrett@eos.ubc.ca (R.C. Brett), krussell@eos.ubc.ca (J.K. Russell),
smoss@eos.ubc.ca (S. Moss).
of their occurrence. Our analysis suggests that virtually all olivine within
kimberlite has a xenocrystic origin and derives from disaggregated
mantle peridotite. This substantially reduces the perceived amount of
olivine crystallization during transport relative to previous estimates
(up to 25%; Clement, 1982; Scott Smith, 1996; Harris et al., 2004;
Mitchell, 2008). The reduced extent of olivine crystallization has
important implications on: i) estimates of primitive kimberlite melt
compositions (e.g. Price et al., 2000; Patterson et al., this issue), and ii)
the extent of orthopyroxene dissolution and melt modification attend-
ing kimberlite ascent (e.g. Mitchell, 1986).
2. Kimberlite sample suite
The Diavik diamond mine is located within the Lac de Gras area of the
Slave craton, approximately 300 km north-east of Yellowknife in the
Northwest Territories, Canada. Four kimberlite pipes are currently in the
Diavik mine plan (A154N, A154S, A418, A21) and are Eocene in age (55–
56 Ma; Heaman et al., 2004). We document the textural and composi-
tional attributes of olivine occurring in kimberlite dykes and volcaniclas-
tic kimberlite units from Diavik. Five samples of coherent kimberlite are
from dykes associated with three of the Diavik pipes (Table 1). In these
samples olivine is not serpentinized, not even along grain margins or
internal factures. Whole rock chemical compositions for the samples are
given in Table A1. In terms of the major constituents (SiO2, MgO, CaO and
FeO (T)), the compositions are similar to other analyses of ‘fresh’

0024-4937/$ – see front matter © 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.lithos.2009.04.030
202 R.C. Brett et al. / Lithos 112S (2009) 201–212

Table 1 3. Petrography
List of samples and descriptions of source, deposit type and olivine content.

Labela Kimberlite Unit Facies Vol.% Reference Samples of coherent kimberlite contain 40–50 vol.% olivine
pipe olivine (Table 1), and on the basis of grain size and morphology the olivine
1 A154 north Late dyke Coherent 41 Moss et al. (2008) constitutes two distinct populations: A) medium to coarse-grained
2 A154 north Late dyke Coherent 40 – (1 mm to 10 mm), rounded to sub-rounded crystals (olivine-I, Fig. 1A),
3 A154 north Late dyke Coherent 50 –
and B) fine-grained (b1 mm), euhedral to subhedral crystals (olivine-
4 A154 south Late dyke Coherent 45 –
5 A21 Late dyke Coherent 40 – II, Fig. 1B). The samples also contain garnet (b2%), clinopyroxene
6 A154 north MVK Volcaniclastic 70 Moss et al. (this issue) (b2%) and orthopyroxene (b1%). The groundmass mineralogy
7 A154 north GK Volcaniclastic 53 Moss et al. (this issue) includes equant oxides consisting of Cr-spinel with overgrowths of
a
1:A154N_08_pet02; 2:A154N_10_12; 3:A154N09_07; 4:A154_35_03; 5:A21_GT03_01; ulvöspinel-spinel that have dark-brown rims of transparent serpen-
6:GTH_75_17_01; 7:A154N_340_GK_B02; MVK = Massive volcaniclastic kimberlite; GK = tine. Groundmass minerals identified petrographically, in order of
Graded kimberlite. decreasing abundance, include: olivine, opaque oxides, monticellite,
apatite and perovskite. These minerals are commonly euhedral and
kimberlites (Dawson, 1994; Price et al., 2000; Caro and Kopylova, 2004). are enclosed by a carbonate–serpentine mesostasis. Occasionally,
Two samples were collected from two distinct volcaniclastic kimberlite carbonate is found to poikilitically enclose groundmass minerals and
deposits (Table 1) described previously by Moss et al. (2008). xenocrysts. In most samples, monticellite is only present within

Fig. 1. Photomicrographs of olivine from Diavik kimberlite deposits, showing: A) xenocrystic olivine (olivine-I) comprising anhedral, rounded to sub angular, inequigranular crystals,
B) “phenocrystic” olivine (olivine-II) identified as smaller (b 1 mm), euhedral to subhedral crystals, C) overgrowth on olivine macrocryst (olivine-I) defined by increase in
concentration of inclusions, D) olivine-II crystal showing euhedral overgrowth on rounded core, E) mantle xenolith containing mm-scale band of polycrystalline recrystallized olivine
grains, F) enlarged view of polycrystalline recrystallized olivine showing characteristic grain size (b 1 mm) and subhedral to euhedral habit.
 R.C. Brett et al. / Lithos 112S (2009) 201–212 203

change the overall shapes of olivine-I grains because the rims are
thin (b0.12 mm) relative to the diameter of the original grain (Figs. 1C
and 3B). In contrast, the shapes of the olivine-II grains are strongly
modified toward euhedrality due to a higher rim:core volume ratio
(Figs. 1D and 3A–C). Fragmented olivine-I is ubiquitous in pyroclastic
samples (6 and 7; Table 1) and makes identification of overgrowth rims
difficult. Additionally, small degrees of serpentinization on the edges of
olivine crystals may obscure overgrowths.
As stated above, the interface of the core–overgrowth boundary is
marked by the presence of numerous, small (b5 µm) spherical fluid

Fig. 2. Photomicrograph of olivine shown in Fig. 1D. Increased magnification shows the
diversity of inclusions in olivine overgrowths including: opaque minerals, apatite,
perovskite, and fluid inclusions. The inset shows planes (P) of fluid inclusions marking
the core–overgrowth boundary.

carbonate-rich portions of the matrix whilst olivine, oxides and

apatite occur in both carbonate-rich and serpentine-rich domains.
Samples of pyroclastic kimberlite comprise mainly individual
olivine crystals that have essentially no selvages of kimberlite
magma (i.e. free olivine grains), small (b10 mm in diameter) juvenile
pyroclasts of kimberlite, and accessory country rock fragments
cemented by a matrix of yellow–brown coloured serpentine. Olivine
is more abundant in the pyroclastic kimberlite deposits than in
samples of coherent kimberlite (Table 1) and the individual olivine
grains are commonly broken (angular). Olivine within juvenile
pyroclasts of kimberlite is relatively fresh and shows little serpenti-
nization. Juvenile pyroclasts contain euhedral, square-shaped opaque
oxides within a matrix of serpentine and minor carbonate.
Rounded to sub-rounded peridotite and dunite microxenoliths are
common as lithic fragments in samples of kimberlite. These mantle-
derived xenoliths commonly contain neoblasts of olivine grains (Fig. 1E)
that result from recrystallization of earlier deformed olivine during
kimberlite transport (Arndt et al., 2006) or from recrystallization in the
mantle before entrainment in the kimberlite (Ross, 1983; Kennedy et al.,
2002). The neoblastic olivine occurs as unstrained, single tabular crystals
up to 1 mm in length and may be completely enclosed by larger, strained
olivine xenocrysts (Fig. 1E). Some large strained xenocrysts (b10 mm) of
olivine feature clusters of small (b0.1 mm), euhedral to subhedral
neoblasts of olivine (Fig. 1F) that are unstrained, lack mineral or fluid
inclusions, and occasionally show weak crystallographic alignment. The
neoblastic olivine grains have grain sizes and crystal habits equivalent to
previously-described kimberlitic olivine phenocrysts.

4. Petrographic zonation in olivine

Olivine (xenocrysts and phenocrysts) in kimberlite can show

complex zoning patterns (Carter Hearn, 2004; Fedortchouk and Canil,
2004; Kamenetsky et al., 2008). Kamenetsky et al. (2008) presented
results on zoning patterns found in olivine classified as phenocrysts
from the Udachnaya-East kimberlite, Yakutia, Russia. Their cores
showed distinct chemical affinities to xenocrystic olivine (i.e. macro-
crysts). On that basis, Kamenetsky et al. (2008) proposed the non-
Fig. 3. False-coloured, back-scattered micrographs of olivine crystals (red colours)
genetic terms olivine-I and olivine-II to avoid connotations with
showing major element zoning and overgrowth morphology. Yellow (Mg-rich
extant xenocryst and phenocryst terminology. For the purpose of this carbonate) and orange (calcite) colours represent carbonate. False colours also identify
research we have adopted their terminology. serpentine mesostasis (black) and spinel microphenocrysts (white). A) Normally zoned
Within our suite of kimberlite samples, olivine-I (Fig.1C) and olivine- olivine-II crystal where overgrowth changes the original shape (core) from anhedral to
II (Fig. 1D) have petrographically visible overgrowths of olivine that euhedral, B) larger normally-zoned olivine-I crystal where the thin overgrowth does not
change the overall shape of original grain, and C) an array of olivine crystals including
are partly defined by the presence of mineral, fluid and melt inclusions homogeneous smaller crystals and crystals that show normal (light red core) or reverse
(Fig. 2). The cores for both types of olivine grains are inclusion free and (darker red core) zoning. Overgrowths are relatively homogeneous from crystal to
commonly rounded to sub-rounded in shape. Overgrowths do not crystal whereas core compositions are diverse.
204 R.C. Brett et al. / Lithos 112S (2009) 201–212

Table 2 Thirty-two trace elements were analyzed by laser ablation

Compositions of olivine-I for coherent kimberlite samples based on N crystals reported inductively coupled plasma mass spectrometry (LA-ICP-MS). Analyses
as: mean (x̄), standard deviation (1s), maximum, and minimum values of mol%
forsterite and NiO content (ppm).
were made with a New Wave UP 213 Nd-YAG 213 nm UV laser ablation
system coupled to a ThermoFinnigan Element2 single collector, high-
1 2 3 4 5 resolution magnetic sector ICP-MS housed at the Department of
Core x̄Fo 91.3 91.3 91.8 91.4 91.8 Geological Sciences, University of Manitoba, Winnipeg, Manitoba,
1s 0.9 0.9 0.9 0.9 1.7
Canada. The laser was operated at 10 Hz, laser flux of approximately
Min 88.9 89.7 89.7 89.4 89.1
Max 92.5 92.9 93 92.9 93.2 6 J/cm2 with a 30–40 µm beam. For line profiles a beam diameter of
N 20 18 22 18 7 15 µm at 20 Hz at a scan speed of 2 µm/s was used. Calibration was
Rim x̄Fo 91.0 91.1 91.2 91 90.8 performed using the National Institute of Standards (NIST) 610 glass
1s 0.4 0.2 0.5 0.3 1 standard, with silicon, previously determined by electron microprobe
Min 90.3 90.1 90.7 90.5 89.4
Max 91.3 91.4 92.1 91.3 92.3
analysis, as an internal standard. The detection limit for LA-ICP-MS is
N 6 8 7 7 8 below 0.1 ppm for most elements and analytical precision is ~ 10%.
Core x̄NiO 3550 3400 3400 3350 3100
1s 550 400 450 550 600
Min 1900 2650 2500 2300 2200 5.2. Variations within and between samples
Max 4600 4100 4000 4250 3800
We consider five different textural types of olivine in each sample:
N 20 18 22 18 7
Rim x̄NiO 1900 1650 2600 1500 2150 i) cores of olivine-I, ii) overgrowths of olivine-I, iii) cores of olivine-II,
1s 850 300 1100 300 500 iv) overgrowths of olivine-II, and v) neoblastic olivine within
Min 950 1050 1350 1100 1700 microxenoliths of peridotite and dunite. Statistical analysis of the
Max 3300 2100 3800 1900 3150
compositional data shows that each of the five textural groups of
N 6 8 7 7 8
olivine has statistically equivalent means. The equivalence permits
aggregation of our data independent of textural attributes, with re-
inclusions (Fig. 2). The fluid inclusions in Fig. 1C and D are expressed sultant statistical properties in terms of Fo and NiO contents sum-
as a dark band marking the core–overgrowth boundary. At higher marized in Table 2 (for olivine-I) and Table 3 (for olivine-II).
magnification (Fig. 2) the dark band is seen to comprise ‘bubbles’ We investigated the compositional differences between olivine-I
aligned in planes and mineral inclusions (Fig. 2, inset). Mineral cores and their overgrowths using the olivine-I datasets for all cores
inclusions are often b1 µm at the core–overgrowth boundary and vs. all overgrowths (Fig. 4A–D; Table 2). In terms of distinguishing
typically increase in size towards the edge of the crystal. Mineral olivine-I overgrowths from their cores, we found the forsterite content
inclusions are identified petrographically as opaque oxides, rutile and and NiO content to be the most effective. The composition of olivine-I
perovskite. On the basis of scanning electron microscopy (SEM), the cores varies from Fo89 to Fo93 and has a mean of Fo92 (Fig. 4A). The NiO
opaque minerals are predominantly magnesium-rich chromites and content in the cores ranges from 1900 to 4600 ppm, and has a mean of
ulvöspinel-spinel solid solution. Larger (b60 µm), irregular-shaped melt 3400 ppm (Fig. 4B). In comparison, the olivine-I overgrowths have
inclusions are trapped by overgrowths, and are dominantly composed of an overlapping but more restricted range in Fo-content (Fo90–92), an
carbonate (Fig. 2). average composition of Fo91 (Fig. 4C), and NiO content spanning 950
The nature of the compositional zoning for several samples is to 3800 ppm. NiO contents of the olivine-I overgrowths show a bi-
illustrated as false coloured, back-scattered micrographs (Fig. 3). Darker modal distribution (Fig. 4D). There is a large peak present at
shades of red show more forsterite rich regions. Olivine grains show ~1500 ppm and a smaller peak at ~ 3500 ppm. The smaller peak
normal zoning (Fig. 3A), reverse zoning (Fig. 3C) or, in many cases, no coincides with the mean core composition and probably results from
zoning (i.e., homogeneous, small grains in Fig. 3C). The compositional blended analyses of overgrowths and cores due to the difficulty of
zoning seen in Fig. 3 corresponds to the overgrowths seen petrographi- sampling thin overgrowth material (typically b10 µm). The differ-
cally (Fig. 1C–D). The overgrowths also show a zoning that commonly ences of forsterite contents between olivine-I rims and cores (Fig. 4C)
terminates as a thin, (b5 µm) dark coloured rim (Fig. 3C). These outer-
most rinds of olivine have extremely high forsterite content (Fo96–98)
and contain up to 2 wt.% CaO, and have been described in other Table 3
kimberlite deposits (Fedortchouk and Canil, 2004; Arndt et al., 2006; Compositions of olivine-II for 7 kimberlite samples based on N crystals reported as:
mean (x̄), standard deviation (1s), maximum, and minimum values of mol% forsterite
Kamenetsky et al., 2008; Scott Smith, pers. com.).
and NiO content (ppm).

5. Olivine composition 1 2 3 4 5 6 7
Core x̄Fo 91.3 91.1 90.6 91.2 91.6 91.2 91
5.1. Analytical methods 1s 0.9 0.6 0.8 0.9 1 0.8 0.5
Min 89.6 90.2 88.7 90.2 90 90.1 90.3
Max 92.7 91.9 91.84 93.5 93.3 93 91.9
Quantitative analysis of olivine for major and minor element N 12 13 14 19 13 15 11
content was performed using the fully automated CAMECA SX-50 Rim x̄Fo 91.3 91.2 91.2 91.2 91.3 91.2 91.4
electron microprobe housed at the Department of Earth and Ocean 1s 0.3 0.4 0.2 0.4 0.4 0.3 0.2
Min 91 90.8 90.9 90.6 90.9 90.6 91.1
Sciences, University of British Columbia, Vancouver, Canada. Analytical
Max 92 92.6 91.7 92.3 92.2 91.6 91.7
conditions were 15.03 kV accelerating voltage, a 20.05 nA primary N 10 17 15 20 14 16 11
electron beam current, and peak and background counting times of Core x̄NiO 3600 3550 3550 3300 3450 3650 3650
60 s and 30 s respectively. We used a 5 µm beam diameter. For the 1s 450 350 450 400 500 500 500
thinnest rims on olivine a beam diameter of 1 µm was used, and count Min 2800 3000 2900 2650 2550 2850 2450
Max 4450 4050 4650 3950 4750 4350 4200
times were increased. The only minor elements detected were nickel
N 12 13 14 19 13 15 11
and calcium. We analyzed seven different samples (Table 1), and Rim x̄NiO 1500 1450 1800 2050 1700 2100 1450
obtained a total of 85 point-analyses of cores of olivine-I grains, 36 of 1s 450 500 850 1000 500 750 350
olivine-I rims, 97 of olivine-II cores, 103 of olivine-II rims, and 60 of the Min 850 800 700 800 1100 1200 850
Max 2400 3050 3750 4150 2600 4100 2150
neoblastic olivine in peridotite or dunite microxenoliths. Appendix C
N 10 17 15 20 14 16 11
(supplementary material) contains all measured olivine compositions.
 R.C. Brett et al. / Lithos 112S (2009) 201–212 205

Fig. 4. Histogram of chemical compositions of olivine as mole fraction Fo and NiO content (ppm). The solid black lines represent model Gaussian distributions based on the calculated
means and variances of each data set.
206 R.C. Brett et al. / Lithos 112S (2009) 201–212

are minor (±0.6 Fo) compared to olivine-I cores (±1 Fo; Fig. 4A),
whereas the deviation of NiO is greater in the rims (±1500 ppm;
Fig. 4B) than in the cores (±1000 ppm) in olivine-I (Fig. 4D).
Our null hypothesis for the olivine-I datasets is:
Ho. The mean compositions of the olivine-I cores and the over-
growths are equal.
Rejection of Ho implies the alternate hypothesis:
Ha. The pooled populations have different means.
If Ha is accepted, then the mean compositions of cores and
overgrowths are chemically distinct, and the two groups of olivine
composition may require different origins. In the case of cores and
overgrowths for olivine-I, Ho is rejected in terms of NiO at the 95%
confidence level (Ha is accepted). This supports the idea that the
overgrowths represent a distinct process relative to that forming the
cores (see Appendix B for methodology and table B1 for the statistical
results).
Compositional differences between olivine-II cores and their over-
growths using the pooled olivine-II dataset are again most evident in
forsterite and NiO contents (Fig. 4E–H; Table 3). Olivine-II core
compositions vary from Fo89–94 with a mean of Fo91 (Fig. 4E), and
their NiO is centered around 3500 ppm (range is 2450 to 4750 ppm; Fig.
4F). The compositional range of olivine-II rims is Fo91–93, with a mean of
Fo91 (Fig. 4G), which is a substantially smaller variability than forsterite
in the cores (Fig. 4E). The standard deviation of forsterite contents for
olivine-II overgrowths is again smaller (±0.3; Fig. 4E) compared with
that of olivine-II cores (±0.8; Fig. 4F). The NiO content in olivine-II
overgrowths ranges from 700 to 4150 ppm with a mean of 1750 ppm
(Fig. 4H). In contrast to Fo-content, the deviation in NiO content is much
greater in the cores (±2800 ppm; Fig. 4F) than in the overgrowths
(±750 ppm; Fig. 4H).
We tested the olivine-II core and overgrowth datasets using the
same method as described above. A statistical analysis of the NiO data
requires rejection (95% confidence level) of the null hypothesis:
Ho. The mean compositions of olivine-II cores and overgrowths are
equal (Table B1).
This implies that the overgrowths to olivine-II are compositionally
distinct from their cores, suggesting that the cores and overgrowths
have distinct origins.
We have also explored the geometry of trace element zoning in
olivine-II crystals from core to rim (Fig. 5). The EMP traverses show
that the NiO content decreases sharply from ~ 3500 (Fig. 5C) and
4000 ppm (Fig. 5D) in the core to ~1500 ppm in the rim. CaO contents
increase at the core–overgrowth interface, and peak at the rim of the
crystal (i.e. at the Fo-rich rind). LA-ICP-MS results show a similar
zonation in NiO content (open circles) to that of the EMP analyses
(Fig. 5C). Movement of the 30-micron laser across the core–rim
boundary is characterized by a marked decrease in NiO content, and
an increase in Cr2O5 and CaO. CaO content (crosses), which is rarely
detectable in olivine, is detected in the overgrowths and the Cr2O5

Fig. 5. Chemical composition profiles for two olivine-II crystals. A) Back-scattered image of olivine-II crystal from sample 1 showing orientation and length of electron microprobe
traverse (white line), B) Back-scattered image of olivine-II crystal from sample 5 showing orientation and length of electron microprobe traverse (white line). The black dot on the
white line marks major element zoning. (C and D) Electron microprobe traverses (solid black lines) show the sharp decrease in NiO content at the core–overgrowth interface. Grey
shades mark the sharp change in Mg and Fe content.
 R.C. Brett et al. / Lithos 112S (2009) 201–212 207

content is higher in rims (closed circles) than in cores (Fig. 5C).

Additional trace elements analyzed by LA-ICP-MS in olivine include
Ca, V, Cr, Co, Ni, Nb, La, and Ce (Table 4). These elements are all above
detection in the rims of olivine-II. Ca, V, Cr, Co, Ni and Nb are all
detectable in both olivine-I and II cores, whereas La and Ce are below
detection limits in cores. Significant increases in Ca, Cr, Nb, La and Ce
contents are evident from core to overgrowth whereas Ni content
drops significantly in the overgrowths.
The compositions of more than 60 neoblasts from three micro-
xenoliths were also measured (Fig. 6B; shaded curves; Appendix C
(supplementary material); samples 1, 3 and 5). The major and minor
element compositions of neoblastic olivine within individual micro-
xenoliths are homogeneous, but vary between xenoliths. The forsterite
content in all samples has a range Fo90–92, however over 90% of the
grains analyzed are Fo91 in composition. The NiO in polycrystalline
olivine is also homogenous in any given nodule, but varies between
samples. The lowest NiO content measured in neoblastic olivine is
2550 ppm, and most analyses are greater than 3000 ppm.

6. Origin of olivine

The results described above suggest that a simple xenolithic vs.

phenocrystic origin does not accurately describe kimberlitic olivine. A
comparison of olivine-I and olivine-II cores and overgrowths is made
using the respective pooled data sets of NiO contents (Fig. 6A). These
data clearly illustrate that the overgrowths on kimberlitic olivines are
chemically distinct from their cores (Fig. 6A, C). More importantly, the
cores of olivine-I and olivine-II have identical NiO contents and the
overgrowths of olivine-I and olivine-II have identical NiO content,
though distinctly different from the cores (Fig. 6A, C).
We have also compared the pooled NiO contents for cores and
overgrowths of olivine-I and olivine-II to the NiO contents of neoblastic
olivine grains from 3 microxenoliths (Fig. 6B). The olivine neoblast
compositions span the full compositional ranges found in olivine-I and
olivine-II cores, but are compositionally distinct from the overgrowths on
olivine-I and olivine-II grains (Fig. 6B). This comparison suggests that
some of the small, non-cored, euhedral olivine grains found in kimberlite
could derive from xenoliths of recrystallized mantle peridotite.
The Fo-contents of olivine cores presented in this study (Fig. 6;
Tables 2 and 3) are fully consistent with a mantle origin, as reported
for olivine in peridotitic mantle from the Slave craton (Fo91–94; Griffin
et al., 1999; Kopylova and Caro, 2004), and from around the world
(Fo90–94; Gaul et al., 2000; Griffin et al., 2003). NiO contents reported
for mantle xenoliths (Sato, 1977; Griffin et al., 1992; Ryan et al., 1996;
De Stefano et al., 2006) and xenocrysts in kimberlite (Fedortchouk and Fig. 6. Differences in olivine compositions summarized in terms of NiO contents (ppm).
Canil, 2004; Arndt et al., 2006; Kamenetsky et al., 2008) typically are A) Compositions of cores and rims of olivine-I (cartoon of rounded crystal) and olivine-
in excess of 3000 ppm. The authors are not aware of studies reporting II grains (cartoon of euhedral crystal). Shading of cartoons denotes part of crystal
analyzed. Solid lines show Gaussian distribution curves for core and rim analyses from
mantle olivine with NiO contents less than 2400 ppm. We therefore
olivine-I grains; dashed lines correspond to analyses from olivine-II grains. B) Pooled
make the interpretation that all of the cores measured in this study compositional datasets for all cores and all rims of olivine grains (cf. panel A) shown as
(1900 to 4750 ppm; x̄ = 3450 ppm) are mantle-derived. This is critical two separate curves. These curves are compared to compositions derived from three
since it implies that a high proportion by mass or volume of euhedral polycrystalline microxenoliths (shaded fields). C) Electron microprobe analyses of
olivine from Diavik kimberlite plotted as mean NiO content (ppm) vs. mean Fo-content
olivine crystals (i.e. those with distinct cores) are, in fact, mantle-
for olivine-I and olivine-II cores and rims (see text). Crosses indicate the 1s variation in
derived and do not result from crystallization of kimberlite magma. each sample set and are compared to the mean analytical uncertainty (1s; upper right).

Table 4
Mean trace element contents (ppm) of olivine by (LA-ICP-MS, see text). The recrystallization processes that produced the neoblastic olivine
(e.g., Passchier and Trouw, 2005) found in microxenoliths of peridotite
Element Ol-II core Ol-I core Ol-II rim
provide an additional source for small euhedral “phenocryst-like”
Ca 261 259 326 crystals. Such grains are distinguished in kimberlite by their euhedral
V 6.57 5.43 6.03
Cr 263 167 392
shape, size (b1 mm), the absence of a discrete overgrowth, and NiO
Co 143 156 143 and Fo-contents indicative of a mantle origin (Fig. 6B).
Ni 3141 3130 2727 Overgrowths on olivine-I and olivine-II are marked by the presence
Nb 0.18 0.22 0.41 of mineral and fluid inclusions and are associated with a major drop in
La b.d.l b.d.l 0.20
NiO content (700 to 4150 ppm; x̄ = 1850). The low NiO contents are
Ce b.d.l b.d.l 0.29
inconsistent with a mantle source and best interpreted as resulting
208 R.C. Brett et al. / Lithos 112S (2009) 201–212

from crystallization of kimberlite melt onto cores of pre-existing and, thus, have to be solved iteratively until there is a fit between the
xenocrystic olivine. current OSD, the model original OSD, and the total amount of new
In summary, we observe texturally and compositionally distinct crystallization. The summation of overgrowth volumes for all crystals
overgrowths on olivine-I (i.e., “macrocrysts”) and olivine-II (i.e., defines the amount of crystallized olivine as a percent of the system (Vx).
“phenocrysts”). The larger grains remain rounded in shape (Fig. 1A) The surface area growth model causes smaller original crystals to
because the overgrowths are thin relative to the diameter of the have thin overgrowths and larger crystals to have thick overgrowths
original olivine crystal. The smaller olivine grains are commonly
idiomorphic because the overgrowths are sufficiently voluminous to
change the shape of the original grain (Fig. 1B). Our analysis shows
that overgrowths exist on all morphologies and sizes of olivine in
kimberlite; there is little evidence to support exclusive homogeneous
nucleation of olivine in kimberlite (i.e. true phenocrysts). Rather, the
smaller idiomorphic olivine population found in kimberlite results
from heterogeneous crystallization of olivine on activated surfaces
offered by pre-existing, rounded, smaller (b1 mm) xenocrysts of
olivine. This suggests that previous estimates of olivine crystallization
in kimberlite melt during transport from the mantle (25%, Scott Smith,
1996; Mitchell, 2008) have been greatly overestimated. Below we
discuss the extent of such olivine crystallization and its consequences.

7. Discussion

7.1. Models for olivine growth

Renewed crystal growth on pre-existing mineral surfaces (hetero-

geneous crystallization) is commonly observed in volcanic rocks (e.g.
Boudier, 1991; Ulianov et al., 2007). For example, Davidson et al. (2005)
suggest that the crystal ‘cargo’ (e.g., larger crystals) in many intermediate
magmas are not true phenocrysts (i.e., derived from the host melt).
Rather, they are crystals inherited by the host magma from other sources
and modified by subsequent crystallization. Previous studies of hetero-
geneous crystallization in silicate melts have suggested that overgrowths
on pre-existing crystals can: (i) be the same thickness regardless of the
size of the original crystal (constant thickness; e.g., Kretz, 1966; Carlson
et al., 1995), (ii) have varying thicknesses resulting from equal volumes of
overgrowth on each original crystal (equal volume; e.g. Tiller, 1991), or
(iii) have varying thicknesses that are proportional to the surface area of
the original crystal (surface area controlled; e.g. Kouchi et al., 1986; Kile
and Eberl, 2003). For any observed crystal size distribution, the rela-
tionship between the original xenocryst size and the thickness of the
overgrowth is different for each of these styles of crystallization. This, in
turn, affects estimates for the amount of crystallized olivine in the system.
We have modeled the three end-member styles of crystallization
with the aim of estimating the amount of olivine that has crystallized
from the kimberlite melt. Our modeling is based on the olivine crystal
size distribution (OSD) for sample 1 by tracing outlines of 1.0 to
3.4 mm olivine in a polished slab (3598 mm2) and adding to that the
abundance of 0.01 to 1.0 mm olivine in a thin section (350 mm2) from
the same sample. For modeling purposes, we have converted the
measured area of each grain to a circle of equivalent area. The OSD
comprises 24,446 particles of olivine representing 44 vol.% of the
sample (c.f. Higgins, 2006). The image analysis methods are fully
described by Moss et al. (this issue) and Holden et al. (in press).
Crystallization models also depend on limits derived from petrog-
raphic observation, including: (i) observed overgrowth thicknesses
never exceed 120 µm, and (ii) the largest euhedral crystals lacking
discrete overgrowths are ~300 µm in diameter. We assume that the
observed OSD represents the combination of original xenolithic olivine,
plus the crystallized overgrowths. The models also allow for a calculated
fraction of newly nucleated and grown olivine crystals; these would be Fig. 7. Models for crystallization of olivine in kimberlite. A) Model output is plotted as
true phenocrysts resulting from homogeneous crystallization. observed radius vs. overgrowth thickness for the surface area controlled model. The
Surface area activated growth assumes that the surface area of grey line marks heterogeneous growth for the optimal solution. Dashed lines show
the original olivine grain controls crystallization. Specifically, the total results for non-feasible solutions. Lengths noted on each line represent the model maxi-
mum radius of the largest phenocryst for each model. Vx is the volume of crystallized
amount of olivine to crystallize is distributed equally over the total olivine. B) Differences between the models are summarized as a bar graph showing
surface area represented by all pre-existing olivine grains. The surface volume the % of heterogeneous vs. homogeneous olivine crystallization. C) Measurements
areas (total and individual) at the onset of crystallization are unknown of mean diameter (mm), and growth thicknesses for 95 olivine grains.
 R.C. Brett et al. / Lithos 112S (2009) 201–212
(Fig. 7A). The thick line represents the growth thickness vs. original
radius for 5% olivine crystallization. This amount of model crys-
tallization is deemed optimal because it produces no over-
growths larger than 120 µm. A consequence of this model is that
some crystals are too small to allow a surface area controlled over-
growth. The 1:1 line represents such crystals, which are treated as
products of homogeneous crystallization. In this model, homogeneous
and heterogeneous OSD components are marked by the vertical
dashed line connecting the dark curve (to the right) and the 1:1 line
(to the left). Surface area controlled growth ascribes small volumes of
olivine to homogeneous crystallization (b0.5%), and most of the
crystallized olivine is attached to pre-existing crystal surfaces (~4.5%;
Fig. 7B).
For comparative purposes we have modeled the effects of greater
extents of olivine crystallization (dashed curves). For 10 vol.% olivine
crystallization the model predicts overgrowth thicknesses to a maxi-
mum of 380 µm. However, overgrowth thicknesses N120 µm are not
supported petrographically. Similarly, smaller amounts of crystal-
lization (i.e. b5%) are not supported because the overgrowth thick-
nesses would be too thin.
Current models for crystallization that produce constant thick-
nesses of overgrowths on pre-existing crystals are based on
growth rates independent of crystal size (Nielsen, 1964; Kirkpatrick,
1981; Lasaga, 1998). We have modeled olivine crystallization in
kimberlite according to this stricture by reducing the size of each
olivine (ro) in the OSD by a constant thickness (rg). The model is
constrained by the observed maximum growth thickness of 120 µm
and imposes a maximum phenocryst diameter of 240 µm. Crystals
b240 µm are, thus, too small to accommodate such a large decrease in
their radius without being consumed completely. This scenario has an
ideal olivine crystallization volume (Vx) of ~10% in the system. The
optimal model of 10 vol.% crystallization suggests that the majority of
olivine crystallization is heterogeneous (N95% of crystallized olivine)
and supports minimal homogeneous growth (Fig. 7B). An additional
scenario involving 15 vol.% of crystallized olivine requires growth
thicknesses of 190 µm on each crystal, and is not supported by
observation. Smaller amounts of crystallization under the constant
thickness stricture (i.e. b10 vol.%) produce only very thin overgrowths
and predict very small phenocryst sizes (bb240 µm).
The constant volume model requires an equal volume of olivine to be
subtracted from each crystal. The constant volume assigned to each
crystal will produce varying ‘overgrowth thicknesses’ depending on the
size of the original seed crystal. Larger crystals have thin overgrowths,
and smaller crystals have thicker overgrowths. Our optimal model
contains 1 vol.% of crystallized olivine, and is chosen because it best
matches the limiting value of observed overgrowth thickness (120 µm).
We identify the grains as bona fide phenocrysts (homogeneous growth)
where the volume subtracted from the crystal is greater than the volume
of the crystal. If the constant volume model is chosen to have greater
extents of olivine crystallization (i.e. 5 vol.%), the model predicts
homogeneous growth for crystals up to 500 µm in radius, and again is
not supported with observation. The optimal model (1% crystallization)
requires small volumes (b0.03 mm3) of olivine added to each crystal and
necessitates sub-equal amounts of homogeneous (0.55%) vs. hetero-
geneous (0.45%) crystallization of olivine (Fig. 7B).
We tested our model results by examining high-resolution BSE
images of coherent kimberlite for an entire thin section. The contrast
of the image was adjusted to amplify forsterite-zoning patterns in
olivine. The image was then printed on a 107 by 193 cm size paper
enabling identification of zoning patterns in olivine down to
0.08 mm diameter. Major and minor axes were measured for all
grains that showed discernable overgrowths in the BSE image. The
thickest and thinnest rim thicknesses were measured for each grain
and averaged. The observed average crystal diameter vs. the apparent
rim thickness is presented for a total of 95 grains (Fig. 7C). The
results show that larger crystals have thicker overgrowths than
209
smaller crystals. These results are permissive only with the surface
area activated growth model and are inconsistent with the constant
thickness and constant volume crystallization styles.
Surface area controlled crystallization is thus our preferred model
because: (i) the model is consistent with observations of maximum
olivine phenocryst sizes and overgrowth thicknesses in kimberlite and,
(ii) this style of crystallization has been shown experimentally to be
dominant in stirred systems (e.g. Eberl et al., 2002; Kile and Eberl,
2003). This implies that a maximum of 5 vol.% of the rock is crystallized
olivine, and that the volume of new-grown olivine is dominantly
present as overgrowths on xenocrystic olivine cores. The reduced
estimate of olivine crystallization departs from the extant paradigm
that kimberlite contains 25% phenocrysts (e.g. Scott Smith, 1996;
Mitchell, 2008).
7.2. Importance for primitive kimberlite melts
Generally, estimates of primitive kimberlite melt compositions are
obtained by isolating and analyzing fine-grained, quenched kimberlite
found at country rock-kimberlite contacts (e.g., chill margins; Price
et al., 2000). A recent example of this approach is afforded by the work
of Kopylova et al. (2007) who measured the bulk composition of
samples of aphanitic (‘macrocryst-free’), vesiculated kimberlite from
the Jericho kimberlite. The bulk composition estimate was based on a
chemical point counting technique, which created a liquid composi-
tion from the proportions and identities of groundmass phases and the
proportion and composition of phenocryst-shaped olivine (Fo91).
Kopylova et al. (2007) considered these measurements as representa-
tive of the composition of the primitive kimberlite melt. On this basis
the reconstructed primitive kimberlite melt composition for the
Jericho kimberlites was predicted to have an Mg# of 87.3. However,
Kopylova et al. (2007) noted that the olivine in equilibrium with this
melt was expected to have a substantially higher forsterite content
(Fo95; assuming a KFe–Mg
D of 0.36; Herzberg and O'Hara, 2002), than
observed in the phenocrystic olivine (Fo91). The discrepancy was
recognized and clearly discussed by the authors, but no explanation
was found.
Our work provides a rationale for partially reconciling Kopylova
et al.'s (2007) results by showing that small, euhedral olivine grains
(olivine-II) comprise both cores of xenocrystic olivine and over-
growths of newly crystallized material. One by-product of Kopylova
et al.'s (2007) image analysis of SEM microphotographs is that they
have well-measured modal proportions (~27% by volume) of pheno-
cryst-shaped olivine. The assumption that all of the observed
phenocryst-shaped olivines are products of melt crystallization causes
an overestimate of the Mg-number of the reconstructed primitive
melt. We suggest instead that crystallized olivine might actually be
closer to 5 vol.%, and that the remaining olivine (22 vol.%) is actually
xenocrystic. Subtracting 22 vol.% Fo94 olivine (i.e. model xenocrysts)
from their melt composition reduces the melt Mg-number from 87.3 to
83.6, and reduces the attendant equilibrium olivine composition from
Fo95 to Fo93.5. This modified estimate represents the magma composi-
tion comprising melt plus 5 vol.% crystallized olivine; the melt would
have an Mg# of 83.6 and be just saturated with ~ Fo93.5 olivine. This
recalculation reduces the gap between model olivine composition and
observed compositions (e.g., Fo93 vs. Fo91).
We have highlighted the effect that modal abundance estimations
have when calculating melt bulk compositions for the samples
studied by Kopylova et al. (2007). This correction stands for several
other studies aimed at the estimation of kimberlite melt chemical
composition (e.g., Price et al., 2000). It should be noted that recalcu-
lation using lower Kolivine-melt
D values (i.e. bb0.25, Canil and Bellis,
2008) elevates forsterite content of crystallized olivine, thereby
causing a larger disparity between predicted and measured olivine
compositions.
210 R.C. Brett et al. / Lithos 112S (2009) 201–212

7.3. The dissolution of orthopyroxene

The observed ratio of mantle-derived orthopyroxene to olivine

macrocrysts is low in kimberlite relative to the ratio of orthopyroxene
to olivine in cratonic mantle xenoliths (e.g. Kopylova et al., 2007). The
relatively low abundance of orthopyroxene in kimberlite has been
ascribed to rapid dissolution of orthopyroxene in the kimberlite melt
during transport (Edgar and Charbonneau, 1993; Ulmer and Sweeney,
2002; Le Roex et al., 2003). Estimates of the amount of orthopyroxene
assimilated by the kimberlite melt are as high as 25 vol.% (Le Roex
et al., 2003; Mitchell, 2008; Kopylova et al., 2007). One hypothesis
is that orthopyroxene dissolution is balanced, in terms of molar
amounts and enthalpy, by concomitant crystallization of olivine
(Mitchell, 1995). This would provide a possible explanation for the
paucity of orthopyroxene xenocrysts in kimberlite, and would permit
large volumes of olivine to crystallize during the rapid ascent of
kimberlite.
We have modeled crystallization of olivine and assimilation of
orthopyroxene using MELTS (Ghiorso and Sack, 1995). We adopt an
average model estimate of kimberlite melt composition and perform
all calculations at 10 kbar. We suppressed all solid phases from crys-
tallizing except for olivine, while forcing assimilation of orthopyrox-
ene. The calculations were performed such that the mass of original
melt and assimilated orthopyroxene summed to 100 g of new melt. Our
petrographic estimates are in vol.% but were converted to mass %
equivalents.
For reference, we have modeled the crystallization path of olivine
without orthopyroxene assimilation (Case 1, thin black line in Fig. 8A).
The model path starts at liquidus conditions (1569 °C) and produces
5 wt.% olivine crystallization after a ~40 °C decrease in temperature,
and requires 200 °C of cooling in order to crystallize 25 wt.% olivine.
Fig. 8B shows temperature vs. the change in system enthalpy for
each reaction path and is contoured for orthopyroxene assimilation
and olivine crystallization. The change in system enthalpy is zero
(ΔH = 0 kJ) prior to crystallization or assimilation. The system loses
enthalpy by cooling (sensible heat) and shedding heats of crystal-
lization (latent heat) in order to crystallize olivine. This crystallization Fig. 8. Models for assimilation of orthopyroxene by kimberlite melt and concomitant
path enriches the residual melt in SiO2 whilst reducing the Mg# of the olivine crystallization. End-member models include (1) olivine crystallization (thin
black curve), (2) the 1:1 molar reaction of orthopyroxene assimilation to olivine
melt.
crystallization (thick grey curve), and (3) assimilation of a fixed amount (27.5 g) of
We model two additional reaction paths that show the effects of orthopyroxene supported by varying amounts olivine crystallization (thick black
orthopyroxene assimilation (Cases 2 and 3 in Fig. 8A). Case 2 involves curve). A) T (°C) vs. olivine crystallization (wt.%); horizontal grey line denotes our
mole for mole assimilation of orthopyroxene and crystallization of model estimate of olivine crystallization (~5 vol.%), B) T (°C) vs. the change in system
enthalpy for each model path contoured for different amounts of orthopyroxene
olivine (thick grey line).
assimilation and olivine crystallization (thin black contours), and C) composition of
melt (SiO2 and Mg#) produced along each model path, normalized to the original melt
Mg2 Si2 O6 fOpxgfMg2 SiO4 fOlg + SiO2 fMeltg ð1Þ



composition (relative per cent). For reference we also show the initial melt liquidus
(dashed vertical line) and the specific masses of each mineral consumed or crystallized
Eq. (1) couples orthopyroxene assimilation to olivine crystalliza- along the model (labeled dots). Arrows denote process direction.

tion and serves as a vehicle to produce melts richer in silica whilst

conserving enthalpy and minimizing cooling. However, based on our
estimate of 5% olivine crystallization, Case 2 only allows for 7.5 g of
orthopyroxene assimilation while producing 5.4 g olivine (i.e. 5 vol.%).
The enthalpy of the system (Fig. 8B) is nearly balanced for this reaction,
and little cooling of the system is required (5 °C). Compositional
modification of the kimberlite melt is only minor and the melt
composition is enriched by +1 wt.% SiO2 and by + 1 in Mg#.
Case 3 imposes a fixed amount (27.5 g) of orthopyroxene assimi-
lation and a variable (0 to 18.5 g) amount of crystallized olivine; this
requires super-liquidus initial conditions (Fig. 8A; thick black line).
If there is no concomitant olivine crystallization, Case 3 requires
the initial system to be ~100 °C super-liquidus. Olivine crystallization
can partially offset the enthalpy reduction due to orthopyroxene
assimilation such that increased olivine crystallization allows for
lower initial magma temperatures (Fig. 8B). The decrease in system
enthalpy is mitigated somewhat by olivine crystallization as
indicated by the contours of olivine content (Fig. 8B). For example,
assimilation of 27.5 g of orthopyroxene causes a 325 kJ reduction in
enthalpy and requires the initial magma to be 100 °C super-liquidus
without olivine crystallization. However if 15 g of olivine crystallizes,
the system enthalpy is reduced only by ~ 215 kJ (Fig. 8B) and the
initial magma need not be super-liquidus (Fig. 8A). Case 3 predicts
the melt to become enriched in SiO2 (~4 wt.%), and increases the
Mg# by ~ 4% (black vertical line; Fig. 8C). This model implies that
~ 13% of total silica in the melt results from assimilation of
orthopyroxene.
In summary, using our estimates of olivine crystallization (5 vol.%),
up to 7% of orthopyroxene can be assimilated if orthopyroxene
assimilation and olivine crystallization are coupled takes place (c.f.
Mitchell, 1995). Higher amounts of orthopyroxene can assimilate only
if it is introduced to the melt at super-liquidus conditions (up to
100 °C; scenario 3). Orthopyroxene assimilation and olivine crystal-
lization are contemporaneous under scenario 2, allowing for olivine
crystallization at great depth, given the known instability of
orthopyroxene over the depth range 100 km to Earth's surface in
 R.C. Brett et al. / Lithos 112S (2009) 201–212 211

kimberlite melt (Luth, in press). If scenario 3 is accepted, significant
quantities of orthopyroxene have to assimilate in order to cool the
melt to reach its liquidus. The time required to cool the melt may allow
the melt to ascend to crustal levels before the melt reaches olivine
saturation.
Appendix B. Whole rock kimberlite compositions
Table A1
Major element (wt.%) whole rock chemical compositions of coherent kimberlite. Major
element (XRF), H2O+ (Furnace-IR), and CO2 (Leco) analyzed at ALS Chemex
laboratories, Vancouver, Canada.

Sample no. 1 2 3 4 5
8. Summary
SiO2 33.28 29.62 30.39 29.92 26.66
TiO2 0.88 1.10 0.75 0.85 1.27
Our investigation shows that olivine in volcaniclastic and Al2O3 1.70 1.53 1.29 1.48 1.68
coherent kimberlite from Diavik comprises Ni- and Mg-rich Cr2O3 0.31 0.38 0.31 0.31 0.35
xenocryst cores with about 120 µm wide rims resulting from FeO(T) 7.91 7.61 7.57 7.47 7.77
MnO 0.16 0.16 0.16 0.16 0.19
kimberlite crystallization. This primary result supports similar
MgO 38.75 34.50 38.51 37.52 36.80
observations made for samples from the Homestead kimberlite, U.S.A., CaO 7.46 9.90 8.83 8.23 11.54
(Carter Hearn, 2004), the Udachnaya-East kimberlite, Russia, (Kame- Na2O b.d.l. 0.05 b.d.l 0.02 0.03
netsky et al., 2008) and the Lac de Gras kimberlite cluster, Canada, K2O 0.30 0.36 0.03 0.08 0.05
P2O5 0.32 0.44 0.39 0.43 0.64
(Fedortchouk and Canil, 2004). Crystallized olivine in kimberlite
H2O 1.77 2.61 3.70 3.46 4.65
comprises approximately 5 vol.% of kimberlite and results from S 0.06 0.08 0.07 0.07 0.09
heterogeneous crystallization. It mainly occurs as rims on xenocrystic CO2 5.75 8.83 4.76 7.47 6.12
derived olivine. A minor component (b0.5%) of olivine crystallization Total 98.65 97.17 96.76 97.47 97.84
occurs as fine-grained euhedral crystals representing homogeneous LOI 7.58 11.40 8.90 11.30 11.60
CO2/H2O 3.25 3.38 1.29 2.16 1.32
crystallization. Crystallization of overgrowths on rounded olivine
xenocrysts traps melt, fluid and solid phases. Our reduced estimates
of crystallized olivine provide a means to modify previous estimates Table B1
of primitive kimberlite melt compositions so that they can be re- Results of students t-test for olivine-I and II cores and rims tested for equal means in
conciled with measured crystallized olivine compositions. If olivine terms of NiO content.

crystallization and orthopyroxene dissolution are strictly coupled, Ol-I core Olivine-I rim Ol–II core Ol–II rim
then the extent of orthopyroxene assimilation during transport of Ol-I core
kimberlite melt through the mantle is restricted to no more than Olivine-I rim Ni
7 vol.%. If large (~ 25 vol.%) amounts of orthopyroxene assimilate Ol-II core – Ni
Ol-II rim Ni – Ni
during transport, then kimberlite melts must begin their transit at
super-liquidus (N100 °C) conditions. Our analysis also suggests that Ni denotes rejection of the null hypothesis (see text).
care must be taken when modes and compositions of phenocryst-
shaped olivine are used to reconstruct primitive kimberlite melt
Appendix C. Supplementary data
compositions. The focus should be on recognizing, isolating and
analyzing true overgrowths, to exclude xenocrystic material mas-
Supplementary data associated with this article can be found, in
querading as phenocryst-shaped grains. Likewise, zoning on large
the online version, at doi:10.1016/j.lithos.2009.04.030.
macrocrysts may be due to olivine crystallization and care must be
taken to distinguish crystallized rims from diffusive profiles that
denote other processes. References

Acknowledgements
This research was funded by an NSERC Collaborative Research
and Development Grant (Kimberlite eruption dynamics: Implications
for diamond distribution in the Diavik kimberlite) held by JKR
and sponsored by Diavik Diamond Mines, Inc. (DDMI). We acknowl-
edge many beneficial conversations with Terence Gordon and Barbara
Scott-Smith. We gratefully acknowledge critical reviews by Lotte
Melchior Larsen, Dante Canil, Ruud Koole and Herman Grutter. We
would also like to thank Professor Mati Rausdupp for guidance with
electron microprobe analysis and Dima Kamenetsky for the high-
resolution BSE images.
Appendix A. Statistical methods
For statistical analysis of the chemical variations within and bet-
ween olivine grains, the concentrations of each element or component
were binned and plotted on a standard histogram. We used Lilliefors'
composite goodness-of-fit test provided in Matlab Statistics toolbox to
evaluate each of our data sets for normality. The Lilliefors' test the null
hypothesis that the data set is normally distributed with an
unspecified mean and standard deviation at a 5% significance level.
The alternative hypothesis is that the data set is not normally
distributed. The test is a 2-sided goodness-of-fit test appropriate for
situations where a fully specified null distribution is unknown.
Arndt, N.T., Boullier, A.M., Clement, J.P., Dubois, M., Schissel, D., 2006. What olivine, the
neglected mineral, tells us about kimberlite petrogenesis. eEarth 1, 15–21.
Boudier, F., 1991. Olivine xenocrysts in picritic magmas: an experimental and micro-
structural study. Contributions to Mineralogy and Petrology 109, 114–123.
Canil, D., Bellis, A.J., 2008. Phase equilibria in a volatile-free kimberlite at 0.1 MPa and
the search for primary kimberlite magma. Lithos 105, 111–117.
Carlson, W.D., Denison, C., Ketcham, R.A., 1995. Controls on the nucleation and growth
of porphyroblasts: kinetics from natural textures and numerical models. Geological
Journal 30, 207–225.
Caro, G., Kopylova, M.G., 2004. The hypabyssal 5034 kimberlite of the Gahcho Kue
cluster, southeastern Slave Craton, Northwest Territories, Canada: a granite-
contaminated group-I kimberlite. Canadian Mineralogist 42, 183–207.
Carter Hearn Jr., B., 2004. The Homestead kimberlite, central Montana, USA: mineralogy,
xenocrysts, and upper-mantle xenoliths. Lithos 77, 473–491.
Clement, C.R., 1982. A comparative geologic study of some major kimberlite pipes in
the Northern Cape and Orange Free State. Ph.D. Thesis. Department of Geology,
University of Cape Town.
Clement, C.R., Skinner, E.M.W., Scott, B.H., 1977. Kimberlite redefined. 2nd International
Kimberlite Conference, Santa Fe, NM, Expanded Abstracts.
Davidson, J.P., Hora, J.M., Garrison, J.M., Dungan, M.A., 2005. Crystal forensics in arc
magmas. Journal of Volcanology and Geothermal Research 140, 157–170.
Dawson, J.B., 1994. Quaternary kimberlitic volcanism on the Tanzania Craton. Con-
tributions to Mineralogy and Petrology 116, 473–485.
De Stefano, A., Lefebvre, N., Kopylova, M., 2006. Enigmatic diamonds in Archean calc-
alkaline lamprophyres of Wawa southern Ontario, Canada. Contributions to Miner-
alogy and Petrology 151, 158–173.
Eberl, D.D., Kile, D.E., Drits, V.A., 2002. On geological interpretations of the crystal
size distributions: constant vs. proportionate growth. American Mineralogist 87,
1235–1241.
Edgar, A.D., Charbonneau, H.E., 1993. Melting experiments on a SiO2-poor, CaO-rich
aphanitic kimberlite from 5–10 GPa and their bearing on sources of kimberlite
magmas. American Mineralogist 78, 132–142.
Fedortchouk, Y., Canil, D., 2004. Intensive variables in kimberlite magmas, Lac de Gras,
Canada and implications for diamond survival. Journal of Petrology 45, 1725–1745.
212 R.C. Brett et al. / Lithos 112S (2009) 201–212
Field, M., et al., 2009, this issue. Variations of olivine abundance and grain size in the
Snap Lake kimberlite intrusion, Northwest Territories, Canada: A possible proxy for
diamonds. Proceedings of the 9th International Kimberlite Conference. Lithos 112S,
023–035.
Gaul, O.F., Griffin, W.L., O'Reilly, S.Y., Pearson, N.J., 2000. Mapping olivine composition in
the lithospheric mantle. Earth and Planetary Science Letters 182, 223–235.
Ghiorso, M.S., Sack, R.O., 1995. Chemical mass transfer in magmatic processes. IV.
A revised and internally consistent thermodynamic model for the interpolation and
extrapolation of liquid–solid equilibria in magmatic systems at elevated tempera-
tures and pressures. Contributions to Mineralogy and Petrology 119, 197–212.
Griffin, W.L., Gurney, J.J., Ryan, C.G., 1992. Variations in trapping temperatures and trace
elements in peridotite-suite inclusions from African diamonds: evidence for two
inclusion suites, and implications for lithosphere stratigraphy. Contributions to
Mineralogy and Petrology 110, 1–15.
Griffin, W.L., Doyle, B.J., Ryan, C.G., Pearson, N.J., O'Reilly, S.Y., Davies, R., Kivi, K., Van
Achterbergh, E., Natapov, L.M., 1999. Layered mantle lithosphere in the Lac de Gras
area, Slave Craton: composition structure and origin. Journal of Petrology 40, 705–727.
Griffin, W.L., O'Reilly, S.Y., Abe, N., Aulback, S., 2003. The origin and evolution of Archean
lithospheric mantle. Precambrian Research 121, 19–41.
Harris, M., le Roex, A., Class, C., 2004. Geochemistry of the Uinjiesberg kimberlite, South
Africa: petrogenesis of an off-craton, group I, kimberlite. Lithos 74, 149–165.
Heaman, L.M., Kjarsgaard, B.A., Creaser, R.A., 2004. The temporal evolution of North
American kimberlites. Lithos 76, 377–397.
Herzberg, C., O'Hara, M.J., 2002. Plume-associated ultramafic magmas of Phanerozoic
age. Journal of Petrology 43, 1857–1883.
Higgins, M.D., 2006. Quantitative textural measurements in igneous and metamorphic
petrology. Cambridge University Press, New York.
Holden, E.J., Moss, S., Russell, J.K., Dentith, M.C., in press. An image analysis method to
determine crystal size distributions of olivine in kimberlite. Computational Geosciences.
doi:10.1007/s10596-008-9115-9.
Jerram, D.A., et al., 2009, this issue. 3D crystal size distributions: A case study on
quantifying olivine populations in kimberlites. Proceedings of the 9th International
Kimberlite Conference. Lithos 112S, 223–235.
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.
Kennedy, L.A., Russell, J.K., Kopylova, M.G., 2002. Mantle shear zones revisited: the
connection between the cratons and mantle dynamics. Geology 30, 419–422.
Kile, D.E., Eberl, D.D., 2003. On the origin of size-dependent and size-independent
crystal growth: Influence of advection and diffusion. American Mineralogist 88,
1514–1521.
Kirkpatrick, R.J., 1981. Kinetics of crystallization of igneous rocks. In: Lasaga, A.C.,
Kirkpatrick, R.J. (Eds.), Kinetics of Geochemical Processes. In: Reviews in Miner-
alogy, vol. 8. Mineralogical Society of America, Washington, D.C., pp. 321–398.
Kopylova, M.G., Caro, G., 2004. Mantle xenoliths from the southeastern Slave Craton:
evidence for chemical zonation in a thick, cold lithosphere. Journal of Petrology 45,
1045–1067.
Kopylova, M.G., Matveev, S., Raudsepp, M., 2007. Searching for parental kimberlite melt.
Geochimica et Cosmochimica Acta 71, 3616–3629.
Kouchi, A., Tsuchiyama, A., Sunagawa, I., 1986. Effect of stirring on crystallization ki-
netics of basalt: texture and element partitioning. Contributions to Mineralogy and
Petrology 93, 429–438.
Kretz, R., 1966. Grain-size distribution for certain metamorphic minerals in relation to
nucleation and growth. Journal of Geology 74, 147–173.
Lasaga, A.C., 1998. Reaction kinetics in geosciences. Princeton University Press,
Princeton, New Jersey. 811 pp.
Le Roex, A.P., Bell, D.R., Davis, P., 2003. Petrogenesis of Group I kimberlites from
Kimberley, South Africa: evidence from bulk rock geochemistry. Journal of
Petrology 44, 2261–2286.
Luth, R.W., in press. The activity of silica in kimberlites, revisited. Contributions to
Mineralogy and Petrology doi:10.1007/s00410-009-0383-8.
Mitchell, R.H., 1986. Kimberlites: Mineralogy, Geochemistry, and Petrology. Plenum
Press, New York.
Mitchell, R.H., 1995. Kimberlites, Orangeites, and Related Rocks. Plenum Press,
New York.
Mitchell, R.H., 2008. Petrology of hypabyssal kimberlites: relevance to primary magma
compositions. Journal of Volcanology and Geothermal Research 174, 1–8.
Moss, S., Russell, J.K., Andrews, G.D.M., 2008. Progressive infilling of a kimberlite pipe at Diavik,
Northwest Territories, Canada: insights from volcanic facies architecture, textures, and
granulometry. Journal of Volcanology and Geothermal Research 174, 103–116.
Moss, S., et al., 2009, this issue. Spatial and temporal evolution of kimberlite magma at
A154N, Diavik, Northwest Territories, Canada. Proceedings of the 9th International
Kimberlite Conference. Lithos 112S, 541–552.
Nielsen, A.E., 1964. Kinetics of Precipitation. Pergamon, New York.
Passchier, C.W., Trouw, R.A.J., 2005. Microtectonics, 2nd ed. Springer.
Patterson, M., et al., 2009, this issue. Kimberlites: Magmas or mixtures? Proceedings of
the 9th International Kimberlite Conference. Lithos 112S, 191–200.
Price, S.E., Russell, J.K., Kopylova, M.G., 2000. Primitive magma from the Jericho pipe,
N.W.T., Canada: Constraints on primary kimberlite melt chemistry. Journal of
Petrology (41) 6, 789–808.
Ross, J.V., 1983. The nature and rheology of the cordilleran upper mantle of British
Columbia: Inferences from peridotite xenoliths. Tectonophysics 100, 321–357.
Ryan, C.G., Griffin, W.L., Pearson, N.J., 1996. Garnet geotherms: pressure–temperature
data from Cr-pyrope garnet xenocrysts in volcanic rocks. Journal of Geophysical
Research 101, 5611–5625.
Sato, H., 1977. Nickel content of basaltic magmas: identification of primary magmas and
a measure of the degree of olivine fractionation. Lithos 10, 113–120.
Scott Smith, B.H., 1996. Undersaturated alkaline rocks: mineralogy petrogenesis, and
economic potential. In: Mitchell, R.H. (Ed.), Mineralogical Association of Canada,
short course series. Manitoba, Winnipeg.
Skinner, E.M.W., Clement, C.R., 1979. Mineralogical classification of southern African
kimberlites. In: Boyd, F.R., Meyer, H.O.A. (Eds.), Proceedings of 2nd International
Kimberlite Conference, Washington D.C. AGU, pp. 129–139.
Tiller, W.A., 1991. The Science of Crystallization — Microscopic interfacial phenomena.
Cambridge University Press, Cambridge.
Ulianov, A., Muntener, O., Ulmer, P., Pettke, T., 2007. Entrained macrocryst minerals as a
key to the source region of olivine nephelinites: Humberg, Kaiserstuhl, Germany.
Journal of Petrology 48, 1079–1118.
Ulmer, P., Sweeney, R.J., 2002. Generation and differentiation of group II kimberlites:
Constraints. from a high-pressure experimental study to 10 GPa. Geochimica et
Cosmochimica Acta 66, 2139–2153.