Chemical Geology 511 (2019) 123–151

Contents lists available at ScienceDirect

Chemical Geology
journal homepage: www.elsevier.com/locate/chemgeo

Invited research article

Petrological, mineralogical and geochemical peculiarities of Archaean T

cratons☆
Balz S. Kambera,b, , Emma L. Tomlinsona
⁎

a
Department of Geology, School of Natural Sciences, Trinity College Dublin, Ireland
b
School of Earth, Environmental and Biological Sciences, Queensland University of Technology, Brisbane, Australia

ARTICLE INFO ABSTRACT

Editor: Klaus Mezger The most outstanding features of Archaean cratons are their extraordinary thickness and enduring longevity.
Keywords: Seismically, Archaean cratonic fragments are sharply-bounded deep roots of mechanically strong, cold litho-
Craton spheric mantle, clearly distinguishable from non-cratonic lithosphere. Rhenium-depletion of deep cratonic xe-
Komatiite nolith whole rocks and sulphide inclusions in diamond indicate that melting was broadly coeval with formation
Harzburgite of the overlying proto-cratonic crust, which was of limited mechanical strength.
Deep melting A very important process of proto-cratonic development was vertical crustal reorganisation that eventually
Bi-modal yielded a thermally stable, cratonised crust with a highly K-U-Th-rich uppermost crust and much more depleted
Picrite gap
deeper crust. Clastic sedimentary rocks available for geochemical study are predominantly found in the youngest
Isotope anomalies
parts of supracrustal stratigraphies and over-represent the highly evolved rocks that appeared during cratoni-
sation. Vertical crustal reorganisation was driven by crustal radiogenic heat and emplacement of proto-craton-
wide, incubating and dense supracrustal mafic and ultramafic volcanic rocks. Statistical analysis of these cover
sequences shows a preponderance of basalt and a high abundance of ultramafic lavas with a dearth of picrite.
The ultramafic lavas can be grouped into Ti-enriched and Ti-depleted types and high pressure and temperature
experimental data indicate that the latter formed from previously depleted mantle at temperatures in excess of
1700 °C.
Most mantle harzburgite xenoliths from cratonic roots are highly refractory, containing very magnesian
olivine and many have a high modal abundance of orthopyroxene. High orthopyroxene mode is commonly
attributed to metasomatic silica-enrichment or a non-pyrolitic mantle source but much of the excess silica re-
quirement disappears if melting occurred at high pressures of 4–6 GPa. Analysis of experimental data demon-
strates that melting of previously depleted harzburgite can yield liquids with highly variable Si/Mg ratios and
low Al2O3 and FeO contents, as found in komatiites, and complementary high Cr/Al residues. In many harz-
burgites, there is an intimate spatial association of garnet and spinel with orthopyroxene, which indicates for-
mation of the Al-phase by exsolution upon cooling and decompression. New and published rare earth element
(REE) data for garnet and orthopyroxene show that garnet has inherited its sinusoidal REE pattern from the
orthopyroxene. The lack of middle-REE depletion in these refractory residues is consistent with the lack of
middle- over heavy-REE fractionation in most komatiites. This suggests that such pyroxene or garnet (or pre-
cursor phases) were present during komatiite melting. In the Kaapvaal craton, garnet exsolution upon significant
cooling occurred as early as 3.2 Ga and geobarometry of diamond inclusions from ancient kimberlites also
supports cool Archaean cratonic geotherms. This requires that some mantle roots have extended to 300 to
possibly 400 km and that early cratons must have been much larger than 500 km in diameter.
We maintain that the Archaean-Proterozoic boundary continues to be of geological significance, despite the
recognition that upper crustal chemistry, as sampled by sedimentary rocks, became more evolved from ca. 3 Ga
onwards. The boundary coincides with the disappearance of widespread komatiite and marks the end of for-
mation of typical refractory cratonic lithosphere. This may signify a fundamental change in the thermal structure
of the mantle after which upwellings no longer resulted in very high temperature perturbations. One school of
thought is that the thermal re-ordering occurred at the core-mantle boundary whereas others envisage Archaean
plumes to have originated at the base of the upper mantle. Here we speculate that Archaean cratonic roots may
contain remnants of older domains of non-convecting mantle. These domains are potential carriers of isotope

☆
Invited 30 June 2017
⁎
Corresponding author at: School of Earth, Environmental and Biological Sciences, Queensland University of Technology, Brisbane, Australia.
E-mail address: balz.kamber@qut.edu.au (B.S. Kamber).

https://doi.org/10.1016/j.chemgeo.2019.02.011
Received 25 February 2018; Received in revised form 16 January 2019; Accepted 7 February 2019
Available online 13 February 2019
0009-2541/ © 2019 Elsevier B.V. All rights reserved.
B.S. Kamber and E.L. Tomlinson
anomalies and their base could have constituted a mechanical and thermal boundary layer. Above laterally
extensive barriers, emerging proto-cratons were protected from the main mantle heat loss. The eventual collapse
of these mechanical barriers terminated very high temperature upwellings and dismembered portions of the
barrier were incorporated into the cratonic mantle during the final Neoarchaean ‘superplume’ event. The sur-
viving cratons may therefore preserve biased evidence of geological processes that operated during the
Archaean.
1. Introduction
The word kratogen was first used in the German geological literature
to distinguish stable continental crust from orogenic crust that had been
involved in mountain-forming processes. The term derives from the
Greek word krátos, meaning strength or power, and was later shortened
to Kraton before being accepted into English usage as craton. There is
wide agreement that a craton represents a strong, thick and coherent
piece of continental lithosphere that has maintained long-term tectonic
stability and traditionally, cratons have been viewed as nuclei of con-
tinents. Before the availability of vast quantities of seismic data and
their computational analysis, cratons were geologically and topo-
graphically recognised as forming flat shield areas comprised of ex-
posed crystalline basement or stable platform sediment successions,
typically of early Precambrian age.
Global seismic tomography is one powerful method for differ-
entiating the roots of cratons because the mantle portion of cratons is
characterised by very fast travel speeds of seismic waves. The seismic
architecture of the lithosphere and surrounding asthenosphere can be
defined by comparing the SH body-wave velocity to a standard model
(e.g. Grand, 1994). Cratonic lithosphere roots below most exposed
Archaean shields have an average SH body-wave velocity 2–3% faster
than reference models (e.g. Begg et al., 2009) whereas Proterozoic li-
thospheric mantle is characterised by SH body-wave velocity only 1%
faster (e.g. Lebedev et al., 2009). Faster travel speeds are primarily the
result of substantially lower temperature than adjacent convecting
asthenosphere and only secondarily the result of compositional differ-
ence (e.g. Griffin, 1999), the maximum likely contribution of which
being 1/4 of the actual observed velocity difference (Schutt and Lesher,
2006).
In their analysis of the lithospheric structure of the African con-
tinent, Begg et al. (2009) demonstrated that it is possible to clearly
differentiate and visualise the Archaean cratonic mantle roots from the
rest of the African continent. Fig. 1 (annotated and reproduced from
Begg et al.'s (2009)Fig. 10) shows a view from the deep mantle tran-
sition zone in the southern Atlantic, looking upward in a NE direction
towards Africa. The volumes of red material (showing SH body-wave
velocity > 1.9%) clearly delineate the sharply bounded roots of the
three broad Archaean cratons within Africa: West Africa, Congo (in-
cluding Tanzania), and southern Africa (mostly the Kalahari craton).
On the global shear speed structure maps of the upper mantle of
Schaeffer and Lebedev (2013), most cratonic mantle roots are seen
extending to at least 150 km depth (their fig. 11) with deviations of SH
body-wave velocity as high as +6%, requiring ambient temperature
of < 1000 °C. Because density of mantle peridotite is an inverse func-
tion of temperature, cratonic roots at such low temperatures can only
maintain neutral buoyancy relative to oceanic asthenosphere due to
their extremely refractory chemistry, essentially exhausted of spinel
and clinopyroxene, depleted in garnet and rich in low density Mg-rich
olivine (e.g. Schutt and Lesher, 2006). Thus, the buoyancy of cratons
and the fast SH body-wave velocity in cratonic mantle roots together
are very strong evidence for extreme (40% or more) melt depletion in
Archaean cratonic roots.
A very pleasing completely independent confirmation of this geo-
physical picture of Archaean cratonic mantle roots has emerged from
petrological and geochemical study of mantle xenoliths. Multiple lines
of evidence show that deep cratonic xenoliths, representing samples
124
Chemical Geology 511 (2019) 123–151
from between ca. 80 and 250 km root depth (e.g. Kopylova and Caro,
2004), mostly experienced very high degrees of melt extraction. Per-
haps the best line of evidence is the composition of the predominant
peridotite phase olivine, whose Mg content is most strongly influenced
by the extent of melt depletion (e.g. Baker and Stolper, 1994). Com-
pared to olivines from oceanic, abyssal peridotites with average for-
sterite contents of 90.5 to 91.0%, those in ancient cratonic harzburgites
are much more magnesian, averaging between 92.5 and 93.0% for-
sterite component (e.g. Pearson and Wittig, 2008). Available experi-
mental data indicate that these compositions require that the host
peridotites experienced extraction of 35–45% liquid (see Bernstein
et al., 2007 and Pearson and Wittig, 2008) through repeated melting
events.
After such extensive extraction of melt, the residue ends up vastly
depleted in incompatible elements. Because Re is much more in-
compatible than Os, ancient harzburgitic peridotites develop retarded,
very unradiogenic Os-isotope compositions (e.g. Walker et al., 1989)
that can be used to calculate a Re-depletion age. When initially dis-
covered, this demonstrated that at least for the Kaapvaal craton, melt
extraction must have occurred prior to 2 Ga. In more recent studies,
peridotitic sulphide inclusions in diamonds were analysed and these
have yielded even older Re-depletion ages. For example: 3.1 to 3.5 Ga
for the Siberian craton (Pearson et al., 1999); 3.4 Ga for the southern
Zimbabwe craton (Smith et al., 2009); and ca. 3.3 Ga for the Kaapvaal
craton (e.g. Shirey and Richardson, 2011). In the Slave craton, Aulbach
et al. (2004) retrieved monosulphide solution inclusions with Me-
soarchaean Re-depletion ages from peridotitic olivine xenocrysts. A
subset of the analysed inclusions form a 3.4 Ga isochron, attesting to the
extreme antiquity of the deep residual lithosphere there. Finally, Ar-
chaean mafic to ultramafic intrusions from the Zimbabwe craton con-
tain extremely unradiogenic Os that Naegler et al. (1997) interpreted to
Southern
Africa
Congo
West Africa
Fig. 1. Three-dimensional representation of tomography of the African litho-
spheric mantle (reproduced from Begg et al., 2009). The viewer is positioned in
the mantle transition zone below the southern Atlantic ocean and is looking
upwards in a NE direction towards Africa (continental outline as a white line).
The up to 250–275 km deep roots of the three broad cratonic regions of Africa
are labelled and defined as red coloured volumes in which the SH body-wave
velocity is > 1.9% than the standard model. Blue colours show volumes of
unusually slow SH body-wave velocities, mainly portraying the upwelling of
hot mantle below the East Africa rift system.
B.S. Kamber and E.L. Tomlinson Chemical Geology 511 (2019) 123–151

indicate the beginning of cratonic root depletion from 3.8 Ga onwards,
coincident with the oldest known zircon crystals known from this
craton (Bolhar et al., 2017).
In summary, the overwhelming petrological evidence from harz-
burgite xenoliths combined with Re-Os isotope systematics indicates
ancient melt extraction events consistent with the presently cool and
refractory nature of Archaean cratonic roots inferred from seismology.
The lithospheric mantle root of some cratons was subsequently lost or
eroded (e.g. Miller et al., 2012) and has in many places been modified
(e.g. Griffin et al., 2003) and cannot be directly compared with the root
that was in place during the Archaean.
Complementary observations are made on the proto-cratonic sur-
face where Archaean mafic rocks are clearly much more magnesian
than those erupted after 2.5 Ga (e.g. Keller and Schoene, 2012). Albeit
not the volumetrically dominant component of greenstone belts (e.g.
Sproule et al., 2002), komatiite with between ca. 20–35 wt% MgO is
known from most cratons and where greenstone belt stratigraphy is
well understood (e.g. from komatiite-related Ni-ore exploration; Lesher
and Barnes, 2009), and multiple episodes of komatiite emplacement
over periods lasting ca. 50 Ma are evident (e.g. Thurston et al., 2008).
There is an on-going debate regarding the type of melting that caused
komatiite formation, ranging from dry, very hot and deep ‘plume’ type
melting (Berry et al., 2008) to aqueous fluid fluxing in (Parman et al.,
1997) and away from subduction zones (Sobolev et al., 2016) with
intermediate models envisaging more complex fluid fluxing (Herzberg,
2016). Regardless of the melting mechanism, komatiite requires ad-
vanced stages of peridotite melting that consumed olivine and ortho-
pyroxene (Nesbitt et al., 1979), an observation that at least qualitatively
complements the highly refractory nature of cratonic harzburgite. The
strong agreement about the general (albeit not total) disappearance of
komatiites at the end of the Archaean is also consistent with the fact
that, on average, Archaean cratonic roots have faster SH body-wave
velocities (e.g. Schaeffer and Lebedev, 2013) than Proterozoic mantle
lithosphere and that Archaean harzburgites are much more strongly
depleted than Proterozoic equivalents (e.g. Griffin et al., 2003).
The relatively sudden (in the context of the vast Precambrian time
span) demise of komatiite production at the Archaean-Proterozoic (A-P)
boundary was accompanied by other notable disruptions in secular
trends, which collectively may have relevance for understanding how
Archaean continental lithosphere differed from younger continents.
These include: the evolution of the maximum MgO content of highly
magnesian magmas, which remained almost constant throughout the
Archaean but then dropped (Campbell and Griffiths, 1992); several
general changes in the nature of mafic and felsic magmas emplaced
towards the end of the Archaean (Keller and Schoene, 2012); pro-
nounced changes in the composition of clastic (Condie, 1993) and hy-
drogenous sediments (Konhauser et al., 2009) across the A-P boundary;
a jump in the O-isotope composition of zircon (Spencer et al., 2014);
and a sharp shift in triple-O-isotope composition of clastic sedimentary
rocks (Bindeman et al., 2018). Condie (1993) demonstrated that the
Archaean land surface had a much higher abundance of komatiite and
basalt than would be estimated from outcrop areas at the present-day
exposure level. Archaean clastic sedimentary rocks have very high re-
lative abundance of Co, Ni and Cr whereas Proterozoic sedimentary
rocks do not significantly differ from Phanerozoic equivalents (Fig. 2a,
b). Most recently, Tang et al. (2016) used the Ni/Co ratio in clastic
sedimentary rocks as a proxy for MgO content with which it was pos-
sible to also show the dramatic transition from mafic to felsic upper

4
3.5
Archaean
Proterozoic
A 10 B
3 Phanerozoic

2.5
Co/Th
La/Sc

2
1
1.5
1 UPPER CONTINENTAL
CRUST
0.5 Archaean
post-Archaean
0 0.1
0 0.2 0.4 0.6 0.8 1 1 10 100
Th/Sc Cr/Th
12 12
C D
10 10
Tang et al. (2016)
8 8 Barrie (2005)
Average Ni/Co

Average Ni/Co

Feng and Kerrich (1990)

6 6

4 4

2 2

0 0
0 500 1000 1500 2000 2500 3000 3500 0 500 1000 1500 2000 2500 3000 3500
Age (Ma) Age (Ma)

Fig. 2. Panel (A) and (B) redrawn in colour from Figs. 18 and 19 of Condie (1993). In both panels, the coloured symbols show shale compositions. Phanerozoic and
Proterozoic shales overlap in composition but Archaean shales are much richer in compatible elements Sc, Co, and Cr. The Phanerozoic and Proterozoic UCC average
composition overlaps with shale data but Archaean shales are more compatible element-enriched than UCC as a whole. Panels (C) and (D) show secular evolution of
UCC Ni/Co ratio reconstructed from clastic sedimentary rock compilation of Tang et al. (2016). Panel (C) plots average values with standard errors and panel (D)
same as (C) with additional data from the Abitibi greenstone belt (Feng and Kerrich, 1990; Barrie, 2005).

125
B.S. Kamber and E.L. Tomlinson Chemical Geology 511 (2019) 123–151

continental crust (UCC) during the Neoarchaean (see Fig. 2c). Unless these discrepancies and apparent mutual inconsistencies can
Notwithstanding the harmonious overall picture of Archaean con- be resolved, there is limited hope that the full geological and tectonic
tinental build-up and mantle melting, it turns out that the reconciliation significance of Archaean cratons can be reconstructed. The aims of this
of the seismic information from cratonic mantle roots with petrology of paper are to review discrepancies that have long been known within the
Archaean harzburgite and komatiite fails in terms of persistent devia- specialist disciplines for a wider geochemical readership, to synthesise
tion of mineralogical and geochemical detail from model prediction mutual inconsistencies and to provide perspectives on possible common
(e.g. Walter, 1998; Bédard, 2006; Pearson and Wittig, 2008). Further- solutions.
more, current models for Archaean craton formation struggle to achieve
mass balance and have limited success explaining the co-occurrence of
2. Brief review of some pertinent problematic aspects of Archaean
komatiite with the more dominant tholeiitic basalts. There is the ad-
mantle melting and lithospheric mantle mineralogy
ditional issue that Archaean cratonic mantle roots were apparently al-
ready sufficiently cold to permit diamond formation (Boyd et al., 1985;
2.1. Incompatible element mass balance
Ballard and Pollack, 1988; Kaminsky et al., 2002; Miller et al., 2012)
yet for the most part, the Archaean crust lacked mechanical strength
Jordan (1975, 1978) developed pioneering models for cratonic
and experienced repeated re-melting events. Thus, the Archaean con-
stabilisation, in which he first clearly demonstrated the need for de-
tinental lithosphere apparently existed as a proto-cratonic composite of
pleting the mantle portion of stable shield lithosphere through basalt
a strong deep mantle root with a weak crustal lid before complete li-
extraction into the crust. The resulting root possesses the required
thospheric stabilisation (i.e. cratonisation) that typically, but not ex-
physical characteristics to make it more refractory, less dense and
clusively, happened towards the end of the Archaean.
rheologically distinct from the surrounding asthenosphere. Since these

94 Fig. 3. Empirical and experimental data plotting

94
modal abundance of olivine in residual peridotite vs.
A OCEANIC TREND B OFF CRATON forsterite component in olivine. Panels (A) to (C)
93 93 redrawn from Boyd (1989) showing the oceanic
olivine Mg-number

XENOLITHS
olivine Mg-number

trend (A), off-craton harzburgites (B) and Kaapvaal

92 92 craton harzburgites (C). Panel (D) shows residue
mineralogy of the experiments conducted on fertile
91 91 peridotite by Walter (1998) colour-coded according
to experimental pressure. The cross symbols show
90 90 the position of the oceanic trend of Boyd (1989).
Note that the y-axis scale extends to much more
magnesian olivine. Panel (E) shows the compilation
89 89 of empirical peridotite compositions of Bernstein
et al. (2007). Panel (F) shows the same experimental
100 90 80 70 60 50 40 100 90 80 70 60 50 40 data as panel (D) but at the same scale as Boyd's
Modal olivine wt% Modal olivine wt% (1989) figures and the position of sample 17BSK051
studied herein (orange square symbol). See text for
94 96 explanations.
3 GPa
C 95 4 GPa
4.5-5 GPa
93
olivine Mg-number
olivine-Mg-number

94 6 GPa
7 GPa
Boyd '89 trend
92 93
92
91
91
90 90
KAAPVAAL LOW-T
89 GARNET PERIDOTITES 89
D
88
100 90 80 70 60 50 40 100 90 80 70 60 50 40
Modal olivine wt% Modal olivine vol%
94

94 93
olivine Mg-number
olivine Mg-number

93
92
92 Ubekendt
Wiedemann 91
91 Sarfartoq
Wyoming
90 90
Tanzania
Kaapvaal
89
E Slave
Siberia
89 F
88
100 90 80 70 60 50 100 90 80 70 60 50 40
Modal olivine vol% Modal olivine vol%

126
B.S. Kamber and E.L. Tomlinson
early models, many lines of evidence have emerged supporting the view
that the continental crust and the depleted mantle portion of cratonic
lithosphere share a very long common history, including melt depletion
events coeval with crust formation pulses (e.g. Pearson et al., 1995;
Naegler et al., 1997). However, the elegance of the basalt extraction
hypothesis is compromised by the problem of mass balance.
The issue is that the preserved cratonic mantle root alone cannot
nearly have supplied all the incompatible trace elements present in the
Archaean crust (e.g. Arndt et al., 2002; Carlson et al., 2005) as can be
illustrated with the example of the highly incompatible radioactive heat
producers U, Th and K. In average continental crust these are 62 ± 2
times enriched over primitive mantle. This ratio is quite precisely
known from heat flow data and continental composition estimates
(Rudnick and Gao, 2003), primitive mantle U and Th concentrations
and terrestrial Pb-isotope composition (Sun and McDonough, 1989),
and the canonical Earth K/U ratio (Jochum et al., 1983). The enrich-
ment ratio is somewhat lower (45 ± 2) for average continental crust of
Archaean age as the crustal column generates a lower average modern
heat flow of 41 mW m−2 on cratonic provinces (Michaut et al., 2009).
Nevertheless, a 165 km deep cratonic mantle root (yielding a 200 km
deep craton) only contains 5.8 times the mass of a 35 km thick crust,
not the required 45 ± 2 times. This means that unless the precursor
mantle was much more enriched than primitive mantle, the un-
differentiated rock mass contained in a craton could only have provided
ca. 15–20% of the inventory of the very highly incompatible elements.
Therefore, Archaean crust formation must also have tapped into
magmas that originated from the convecting asthenospheric mantle
(Arndt et al., 2002). Such magmas could have travelled through the
growing Archaean continental lithosphere (e.g. Carlson et al., 2005)
and/or been sourced via lateral accretion (Wyman et al., 2002; Wyman,
2018). Regardless, the discussion of Archaean mantle source regions
must distinguish between the ultra-depleted cratonic roots and the
more modestly depleted convecting asthenosphere.
2.2. Mismatch between predicted and observed mineral modes in cratonic
harzburgites
At relatively low pressures (1–3 GPa), melting experiments of fertile
mantle compositions predict that the residue after 25–30% melt ex-
traction should be devoid of clinopyroxene and Al-phase (spinel and/or
garnet) and consist of > 80% olivine (Baker and Stolper, 1994; Walter,
1998) and that at 40% degrees of melting, the residue should be nearly
mono-mineralic dunite. At higher pressures, garnet can remain in the
residue at up to 40–50% melt extraction. Because the Fo-content of
olivine increases approximately linearly with the extent of melt
Subcalcic garnet present
ZIMBABWE
S N
No subcalcic garnet CRATON
Mobile Belt
BELT
OPO
LIMP
Lithosphere
KHEIS
KAAPVAAL
CRATON
NAM
AQUA-NATAL BE
LT Asthenosphere
0 300
CAPE FOLD BELT
km
Fig. 4. Map showing kimberlite occurrences with and without sub-calcic garnet macrocrysts and lithospheric cross-section with interpretative areas of garnet
formation and metasomatic alteration. Simplified and re-drawn in colour from Nixon et al.'s (1987) Figs. 261 and 267.
127
Chemical Geology 511 (2019) 123–151
extraction (Pearson and Wittig, 2008), a strong correlation between
modal olivine content and Fo-component is expected. This relationship
was prominently described for oceanic peridotites by Boyd (1989) and
is shown in Fig. 3a. Boyd (1989) next compared this oceanic trend with
data from post-Archaean subcontinental lithosphere brought to the
surface by off-craton kimberlites. The comparison confirmed that, to a
first order (Fig. 3b), the data followed the oceanic trend. However,
harzburgites from the Archaean Kaapvaal craton (Fig. 3c) were found to
deviate markedly from the trend, plotting at higher Fo values (92–93%)
but containing only 55 to 80% modal olivine instead of the expected
dunitic abundance (> 90%). Essentially, the harzburgites are much
richer in orthopyroxene than expected and the resulting rock has an
unusually high Si/Mg ratio (as enstatite is the most Si-rich of the
common mantle minerals).
This mismatch between observed and predicted mineralogy is very
strongly expressed in Kaapvaal harzburgites but it is also found in many
other Archaean cratonic harzburgites with < 60% modal olivine, for
example from the Slave, Siberian and Wyoming cratons (e.g. Boyd et al.,
1997; Eggler et al., 1987; Ionov et al., 2010; Kopylova et al., 1999;
Kopylova and Caro, 2004; Newton et al., 2015; Regier et al., 2018;
Solovjeva et al., 1994). Xenolith populations of only few kimberlites,
particularly from the North Atlantic craton (Greenland; e.g. Bizzarro and
Stevenson, 2003; Bernstein et al., 2006), are dominated by dunite as
expected from Boyd's (1989) trend, but with generally high Fo-compo-
nent. On average, high orthopyroxene abundance is more common and
has been very widely discussed in the literature (e.g. Bernstein et al.,
2007) without finding wide petrological consensus. Proposed explana-
tions include: the high Si/Mg ratio was inherited from the precursor
mantle that may have differed from pyrolite due to crystal fractionation
in a magma ocean (Herzberg et al., 1988); the high modal abundance of
orthopyroxene reflects melt extraction at much greater depth (> 8 GPa)
than experienced by oceanic harzburgites (Canil, 1992; Herzberg, 1993);
density sorting in plume melt residues (Arndt et al., 2002); later addition
of Si, possibly from subduction zone metasomatism (Kesson and
Ringwood, 1989) or serpentinisation (Canil and Lee, 2009); late Proter-
ozoic within plate melt infiltration of hydrous melt (Wasch et al., 2009);
reaction between depleted harzburgite either with foundering eclogite
restite (Bédard, 2006) or eclogite-derived melts (Aulbach et al., 2011);
and originally oceanic origin at a shallow very hot ridge (Rollinson,
2010). As can be appreciated, these various scenarios have substantially
different ramifications for the early Earth.
A second unexpected feature of Archaean cratonic harzburgite is
that it commonly contains garnet. High pressure experiments (e.g.
Takahashi, 1986; Canil, 1991; Walter, 1998) show that at pressures
exceeding 6 GPa, minor residual garnet can persist in harzburgite even
Craton
Archaean crust
200 km
on
Z
eo
f1
.4 G
a en Zone of Archaean gt harzburgite
richme
nt
100 km
B.S. Kamber and E.L. Tomlinson
after 40% melt extraction. However, it appears that garnet is not a rare
accessory but a common phase in the Archaean cratonic mantle cross-
section (Griffin et al., 1999, 2002) and this is at odds with the widely-
held model that the refractory lithosphere grew by orogenic stacking of
residues from low pressure melting (e.g. Bernstein et al., 2007; Canil,
2004; Rollinson, 2010; Lee and Chin, 2014). Indeed, Nixon et al. (1987)
demonstrated that the common occurrence of garnets (with a char-
acteristic sub-calcic chemistry) was restricted to xenocrysts and garnet-
bearing xenoliths from kimberlites that intruded through Archaean
cratonic lithosphere (Kaapvaal and Zimbabwe) whereas off-craton
kimberlites erupted through Proterozoic geology were devoid of such
garnets (see Fig. 4). It is possible that the overabundance of orthopyr-
oxene and the dominance of garnet-bearing harzburgite are related
through exsolution of garnet from orthopyroxene (e.g. Cox et al., 1987)
and/or that garnet grew in response to metasomatism (e.g. Stachel
et al., 1998; Pearson and Wittig, 2008). Regardless, Sm/Nd and Lu/Hf
isotope data show that garnet growth in many cratons occurred during
the Archaean (e.g. Shu et al., 2013; Koornneef et al., 2017).
2.3. Seismic evidence for garnet in Archaean lithosphere and lithospheric
buoyancy constraints
In the early long-range seismic refraction profiles, such as those
conducted across continental North America (e.g. Green and Hales,
1968), it was noticed that there is an increase in impedance at depths
between 60 and 80 km. Because impedance is the product of seismic
velocity and density and since temperature is unlikely to decrease with
depth, Hales (1969) proposed that this observation represented the
spinel to garnet transformation. Many seismic studies have since found
similar increases in impedance (e.g. Revenaugh and Jordan, 1991;
Woods et al., 1991) but the observation is not generally accepted to
reflect the spinel to garnet transition globally, although the observed
extent of velocity change fits well with experimental constraints on the
phase transition (Webb and Wood, 1986).
Whereas the detection of the impedance increase can reasonably be
expected in the more fertile oceanic mantle, where it occurs near the
modelled depth of 50–60 km, Revenaugh and Jordan (1991) com-
mented that in continental lithosphere, basalt depletion should strongly
diminish the strength of the reflector by reducing modal abundances of
spinel and garnet. This should be particularly true for cratonic litho-
sphere where multiple lines of evidence suggest extensive melt ex-
traction. However, it appears that the impedance increase is just as
prominent in the lithosphere of several Archaean cratons, including the
Baltic Shield, Western Australia and the Congo craton (Lebedev et al.,
2009). Indeed, in the original work by Hales (1969) the strongest ve-
locity increase was found in the Lake Superior region, underlain by
Archaean lithosphere. In comparison with the oceanic realm, the in-
crease in seismic velocity in cratonic regions occurs from the Moho
down to depths of 100–150 km and is less sharp. Lebedev et al. (2009)
explained this with the survival of spinel to greater depth due to in-
crease in the Cr2O3 activity in the refractory residual harzburgite. Ex-
perimental data support stability of spinel to greater depth in Cr-rich
bulk compositions, but these studies (e.g. Klemme, 2004; Ziberna et al.,
2013; Ziberna and Klemme, 2016) also predict very low overall modal
spinel abundances. Thus, an additional explanation for the increase of
the shear wave velocity below cratons could be the appearance of ultra-
depleted, sub-calcic and calcic garnet, which is often found to have
equilibrium pressures in the region of 3–5 GPa (e.g. Gibson et al., 2013;
Gibson, 2017), corresponding to depths of ca. 90–150 km. The
80–90 km range is also the depth for which Griffin et al. (2002) report
the shallowest appearance of Cr-rich garnet macrocrysts in a number of
Archaean cratons. Regardless of the importance of spinel, the seismic
evidence apparently supports the empirical petrological evidence for
garnet being an important constituent phase of many Archaean cratonic
roots.
Although cratonic lithosphere is commonly thought of as being
128
Chemical Geology 511 (2019) 123–151
buoyant, this is only true in terms of composition but due to its much
lower temperature than asthenosphere, the cratonic root is, on average,
only barely positively buoyant (e.g. Rudnick, 1994). In fact, Schutt and
Lesher (2006) calculated that residual cratonic peridotite formed above
110 km may be negatively buoyant by the time it has cooled to the
cratonic geotherm. These authors suggested that this could imply an
inherent tendency for relatively shallow depleted residues to undergo
vertical downward motion. A final poorly understood aspect of cratonic
buoyancy is that there is a lack of evidence for proto-cratons having
ever stood substantially elevated above sea level during their formation
(Galer and Mezger, 1998; Bindeman et al., 2018). This is surprising
particularly for cratonic fragments that formed rapidly (e.g. many on
the subprovinces of the Superior craton; e.g. Thurston et al., 2008). At
the solidus, the cratonic root density becomes substantially lower than
that of fertile mantle (e.g. Ziberna and Klemme, 2016) and this would
buoy the craton, yet entire stratigraphies of basalt and komatiite were
apparently emplaced in water (e.g. Houle et al., 2009) into environ-
ments with limited topography. This is also true for many of those
demonstrably deposited ensialically through evolved continental crust
(e.g. Chauvel et al., 1985, 1993; Blenkinsop et al., 1993). Many of these
apparent inconsistencies may stem from the fact that in the Archaean,
proto-cratons had not yet reached their mechanical rigidity, with much
reduced strength in the middle and lower crust (Ashwal et al., 1987;
Kramers et al., 2001; Benn and Kamber, 2009).
2.4. Experimental vs. empirical constraints on komatiite major element
chemistry
At least since the experiments by Takahashi and Scarfe (1985) it has
been clear that high degree, high temperature melting of fertile mantle
at 150–200 km depth can produce liquids that are as magnesian as
komatiite. Because komatiite is defined as an ultramafic volcanic rock
with MgO > 18 wt% and depleted in other oxides (e.g. Arndt et al.,
2008), Takahashi and Scarfe (1985) were technically correct in relating
their experimental melts to komatiite. However, there remain very
significant issues with the widely held opinion that komatiites are very
high degree melts. Arndt (1977a) pointed out that since mantle melts
are mostly buoyant relative to refractory olivine, it would be physically
impossible for early melts not to escape. Schmeling and Arndt (2017)
showed that mantle melts are less dense than the main harzburgite
phase olivine to a depth of 230 km, making it implausible that liquids
would have pooled at 50–230 km source depths until 30–40% of
melting. By contrast, it is possible to form liquids with > 30 wt% MgO
at modest degrees of melting from more refractory peridotite that had
previously been depleted or melting at greater depth (e.g. 14 GPa;
Takahashi, 1986). At mantle transition zone depths, relative densities
permit large degree melts to accumulate before eventual escape
(Schmeling and Arndt, 2017).
There are also issues regarding the finer details of major element
composition of experimental versus empirical komatiite compositions
(e.g. Walter, 1998). Many of these problems may relate to the inability
of small-scale experiments to simulate the natural process of melting in
large-scale thermal upwellings (e.g. Robin-Popieul et al., 2012).
Herzberg (2016) provided a recent discussion on Al-undepleted ko-
matiite liquids. It was found that that at a given parental MgO content
the observed SiO2 contents for classic Neoarchaean Abitibi greenstone
belt komatiites are too low by comparison with experimental liquids,
certainly outside analytical error. The diagrams from Herzberg (2016)
most relevant to this mismatch are reproduced in Fig. 5. Relative to an
accumulated fractional melting liquid with 30 wt% MgO, the observed
liquid line of descent (LLD) has ca. 0.8 wt% too little SiO2 (Fig. 5a).
More significantly, in terms of FeO vs. MgO systematics, the empirical
data favour a batch melt (Fig. 5b), which is ca. 1.7 wt% higher in SiO2
at MgO 30 wt% than the empirical LLD (Fig. 5a). If these komatiites are
representative, then either the melting models are in error, the source of
komatiite was not pyrolitic, melting did not occur at dry conditions
B.S. Kamber and E.L. Tomlinson Chemical Geology 511 (2019) 123–151

50 show a surprising lack of strong fractionation. Walter (1998) speculated

that the mantle of at least some of the cratons (e.g. Kaapvaal) could
A Modelled
komatiite LLDs have had a major and trace element composition different from pyr-
48 olite. The melting depth conundrum was further developed by Canil
(2004), who pointed out that the high Cr/Al ratio of cratonic harz-
burgites was difficult to reconcile with melt extraction deeper than
SiO2 (wt%)

46
Alexo and Pyke Hill 3 GPa. One proposed solution was that the cratonic lithosphere grew by
komatiites stacking and thickening.
44 Studying FeO/MgO systematics of cratonic peridotites, Lee and Chin
(2014) proposed that most Archaean mafic magmas could be sourced
Primary magma (volatile-free) Olivine by melting mantle between 2 and 4 GPa at temperatures of 1500 to
42 Batch Melting Mg# 94.5 1650 °C. Because nearly all their studied peridotites yielded higher
Accumulated Fractional Melting equilibration pressures and lower equilibration temperatures, these
authors also advocated vertical foundering and orogenic thickening as
40 the key process for craton growth. However, Lee and Chin (2014) ex-
0 10 20 30 40 50 60
plicitly filtered against orthopyroxene-rich peridotite samples by (p.
MgO (wt%)
277) “excluding those samples with anomalously high SiO2 for a given
14
Mg#”, on the basis that these were considered to implicitly reflect
B Accumulated metasomatic Si-addition.
12 Fractional Melting In summary, there are clearly recognisable patterns in most pro-
posals that attempt to reconcile experimental with empirical petrolo-
L LDs

L + Ol gical and seismic observations regarding Archaean subcontinental li-

thosphere formation. Many of these relate to the minerals
FeO (wt%)

10 L + Ol + Opx

Primary orthopyroxene and garnet both of which are apparently more abundant
Magmas in Archaean cratonic lithosphere than available experimental con-
straints permit. Whereas many ancient cratonic harzburgites are too
8 L + Ol

rich in SiO2 for their MgO content or Mg#, many komatiites seem to
Batch Melting

Olivine Mg#
6 94.0
95.0
4 paper is to explore whether there could be a common solution to the
0 10 20 30 40 50
MgO (wt%)
Fig. 5. Comparison of empirical (Alexo and Pyke Hill) komatiite major element
chemistry with petrological predictions reproduced from Herzberg (2016).
Panel (A) shows MgO vs. SiO2 compositional space in which the observed LLD
(open circles whole rocks and solid grey circles melt inclusions), caused by
olivine crystallisation, plots at lower SiO2/MgO than either batch (red cross) or
accumulated fractional (green cross) melting of the melting model predictions.
Panel (B) shows MgO vs. FeO compositional space in which the observed LLD of
the classic Abitibi greenstone belt komatiites strongly favour batch over accu-
mulated fractional melting of a fertile peridotite (KR4003).
(Herzberg's (2016) preferred option), or a combination of several fac-
tors. Herzberg (2016) also explored to what extent the experimental
liquids of Walter's (1998) experiments could match the empirical
Neoarchaean Abitibi greenstone belt komatiites in MgO vs. SiO2 space
(his Fig. 3a). This comparison revealed that certain liquids from the
highest pressure range runs (6 and 7 GPa) coincided with the empirical
LLD but at MgO contents of 27–28 wt%, i.e. lower than the presumed
30 wt% MgO parental melt deduced from melt inclusions (Sobolev
et al., 2016).
In addition to the MgO/SiO2 mismatch, Walter (1998) commented
on a paradox concerning the depth of melting. When plotting empirical
data and experimental melting contours in CaO/Al2O3 vs. Al2O3 space
(in which olivine crystallisation causes predictable vertical LLD),
Walter (1998) found (his Fig. 7a) that early Archaean komatiites could
be generated at 8–10 GPa with up to 30% melting and that Neoarch-
aean komatiites at 7 GPa with 30 to > 50% melting. However, these
samples have flat chondrite-normalised rare earth element (REE) pat-
terns, which apparently precluded garnet as a stable phase in the
source. Walter (1998) commented (p. 49): “This creates a paradox”; in
that major elements (CaO/Al2O3) require garnet on the solidus yet REE
do not show middle (M) and light (L) enrichment over heavy (H) REE
expected from residual garnet. In terms of Sm/Yb or Gd/Yb, komatiites
suffer from the opposite problem. Finally, while major element sys-
tematics require garnet in the source of many komatiites, their REE
patterns do not show the expected deficit in HREE. The main aim of this
60 issues of Archaean craton formation.
3. Datasets, analysis and samples
3.1. Datasets and data analysis
We revisited four datasets relevant to Archaean craton formation.
Firstly, we reanalysed major and compatible trace elements systematics
of Archaean igneous rocks previously studied by Keller and Schoene
(2012) and Kamber (2015). Specifically, we used the pre-compiled files
from the GEOROC database for the following cratons: Baltic shield;
Dharwar craton; Kaapvaal craton; North Atlantic craton; North China
craton; Slave craton; Superior craton; Tanzania craton, Western Aus-
tralian cratons; and Zimbabwe craton. These were selected to include
cratons for which geochemical and petrological datasets also exist for
peridotites, because they host classic komatiite exposures and for which
statistically significant datasets exist. Each dataset was filtered to only
include extrusive igneous rocks for which whole rock analyses are re-
ported. The analysis of the distribution of MgO and Ni contents of Ar-
chaean volcanic rocks was conducted with kernel density estimation
(KDE), which is a robust method to make statistical inferences on po-
pulation distributions of a finite data sample. The method lends itself to
estimate the number of modes in a population (Silverman, 1981) and is
commonly used to analyse detrital U/Pb zircon and other age popula-
tions (e.g. Petrus et al., 2016). The programme used in this study was
DensityPlotter (Vermeesch, 2012).
A second dataset that was revisited separately is that for komatiites
and komatiitic basalts from the Neoarchaean Abitibi and Pontiac
Subprovinces (Superior Province) reported by Sproule et al. (2002).
This is a very comprehensive dataset, including analyses for the classic
Pyke Hill and Alexo samples. Representative analyses were tabulated in
Sproule et al. (2002) and only these are found in GEOROC. Here we
used the full dataset (Mike Lesher, pers. comm. 2017) to test whether,
within one greenstone belt, temporal trends existed in terms of Mg/Si
systematics (e.g. Herzberg, 2016).

129
B.S. Kamber and E.L. Tomlinson Chemical Geology 511 (2019) 123–151

A B
Superior craton Western Australian
(n=2845) cratons (n=2008)

n n

MgO wt% MgO wt%

C D
Kaapvaal craton Baltic shield
(n=1623) (n=1658)

n n

MgO wt% MgO wt%

E F
Dharwar craton North Atlantic craton
(n=1091) (n=702)

n n

MgO wt% MgO wt%

G Zimbabwe, Tanzania, Slave & H All cratons

North China cratons (n=732)

n
n

MgO wt% MgO wt% (log scale)

(caption on next page)

130
B.S. Kamber and E.L. Tomlinson
Fig. 6. Magnesium (MgO wt%) distribution in Archaean metavolcanic rocks from six cratons (panels A–F) for which there are > 500 compiled analyses; a composite
of four cratons with fewer data (panel G) and for the combined dataset (panel H, note logarithmic x-scale). Each panel shows the raw data as histograms of MgO wt%
re-calculated to anhydrous total of 100 wt%. Superimposed on the histograms is a kernel density function (parameters listed in Table 1) as a dark blue curve with
light blue fill. Linear scale diagrams highlight the dominant basaltic population at MgO 5–9 wt%, the prominent evolved population with ca. 1 wt% MgO, and the
much broader ultramafic population (MgO 25–30 wt%). In all panels, the paucity of rocks with MgO of ca. 18 wt% is evident. The composite figure (panel H) with
logarithmic x-scale accentuates the gap between the basaltic and ultramafic populations.
Global deep-time geochemistry datasets for shale are a rich source
of information regarding the evolution of the uppermost crust (e.g. Smit
and Mezger, 2017; Bindeman et al., 2018). Here we revisited the shale
geochemistry compilation reported by Tang et al. (2016) in their sup-
plemental materials S1 and added data for Neoarchaean argillite
(Barrie, 2005), which is a very fine-grained sedimentary rock con-
taining reduced C. This rock type is extensively studied by exploration
geochemists due to its utility in vectoring volcanogenic massive sul-
phide (VMS) deposits but these data rarely find their way into peer-
reviewed publications.
The final datasets revisited are the experimental ultramafic liquid
compositions and modal abundances reported in Walter (1998) and
Takahashi et al. (1993). Additional highly valuable experiments have
been reported (e.g. Takahashi and Scarfe, 1985; Takahashi, 1986; Canil,
1991, 1992; Takahashi et al., 1993; Baker and Stolper, 1994; Longhi,
2002; Schutt and Lesher, 2006) but those conducted at high pressure
did not report full modal abundances of reaction products. We com-
pared their phase relationships to those of Walter (1998) and Takahashi
et al. (1993) to test for consistency and representativeness.
3.2. Samples
As explained in the introduction and preceding section, the minerals
orthopyroxene and garnet hold a critical role in Archaean craton for-
mation (e.g. Gibson, 2017). In 2015 and 2017, we collected cratonic
peridotites from the Bultfontein pans, Kimberley, Republic of South
Africa (28.739155°S, 24.818094°E) with the aim of better documenting
the mutual phase relations (e.g. Cox et al., 1987; Saltzer et al., 2001)
between garnet and orthopyroxene and to better determine their in-
teresting trace element geochemistry (Wasch et al., 2009). Here we
report data for two representative samples. Sample BP002 is dominated
by a spectacular orthopyroxene megacryst of which a single crystal
fragment measuring 15 × 12 × 7 cm is preserved within a granular
garnet harzburgite. Tomlinson et al. (2018) reported a crystallographic
and geochemical study for this sample. The other sample is a coarse-
grained garnet harzburgite (17BSK051) with a high modal abundance
of orthopyroxene typical of harzburgites from the Kimberley pipes (e.g.
Boyd, 1989; Grégoire et al., 2003; Simon et al., 2007; Wasch et al.,
2009). Analytical details can be found in the Electronic Supplement.
4. Results
4.1. The composition of the uppermost Archaean continental crust
Three prominent studies have investigated the temporal evolution
of the composition of the continental crust with large lithogeochemistry
compilation datasets (Keller and Schoene, 2012; Tang et al., 2016; Smit
and Mezger, 2017). They have in common that they analysed time
series of averaged geochemical parameters (e.g. MgO content, Ni/Co,
Cr/Zn and Cr/U) of lithological groups such as mafic and felsic igneous
and sedimentary rocks. By contrast, the purpose of the analysis per-
formed here was to study the modal distribution of volcanic rocks that
were erupted onto the various cratons in the Archaean.
Fig. 6 shows histograms for MgO wt% distribution in volcanic rocks
on 6 cratons for which > 500 datapoints are available in GEOROC (pa-
nels A–F) and a combined histogram for the Zimbabwe, Tanzania, Slave
and North China cratons (panel G) for which there are smaller datasets.
For each dataset, MgO contents were re-calculated to a volatile-free basis
131
Chemical Geology 511 (2019) 123–151
for those samples (> 90%) for which full major element data and loss on
ignition are reported. The metadata are available in Table_S1. Data for
the few (≪1%) rocks with > 50 wt% MgO were excluded from further
analysis on account of clearly being cumulates. Superimposed onto the
histograms are KDE curves (blue solid lines). Sizes of sample populations,
histogram bin widths and fitting bandwidths (which determine KDE
sensitivity) are listed in Table 1.
A qualitative look at the histograms and the KDEs reveals a common
pattern. There is a prominent population of evolved felsic volcanic
rocks with MgO between 0.7 and 2 wt% and then a very prominent
population with MgO between 5 and 10 wt%. These are separated by a
strong gap. A compositional gap at similar MgO is well known in ocean
island rocks (Chayes, 1963) and arc systems (e.g. Reubi and Blundy,
2009) where it has been attributed to differentiation and re-melting
within the crust.
In all cases, the 5 to 10 wt% MgO basaltic population contains the
largest number of samples. Moving towards higher MgO contents, the
next common pattern is the paucity of samples with MgO in the 15 to
20 wt% range. Beyond 18 wt% MgO, all cratons show one or two broad
populations of ultramafic volcanic rocks. With the exception of the
Baltic shield data, where a population around 25 wt% MgO is quite
prominent, clear ultramafic populations are not easily obvious. One
reason it is difficult to test whether there is a dominant population in a
linear histogram of MgO is that the ultramafic samples are naturally
distributed over a large number of bins. This is because in an ultramafic
parent rich in MgO, relatively limited olivine and pyroxene removal or
accumulation will cause rapid change in MgO of several wt%, i.e.
several histogram bins. By contrast, the same extent of crystal fractio-
nation from a much lower initial MgO content in a felsic or basaltic
magma will only spread the data across one or two bins. This is better
appreciated when the data are subjected to a logarithmic translation of
the x-axis values. The translation results in unevenly sized bin widths
and separates data with low MgO content into more bins. By contrast,
the ultramafic rocks are binned into fewer categories. The resulting
histogram and KDE for the combined data from all cratons considered
are shown on Fig. 6H. This clearly exposes the gap in erupted volcanic
rocks with MgO between 15 and 20 wt% and emphasises the ultramafic
population centred around ca. 28 wt% MgO. To our knowledge, this has
not explicitly been reported before but has been hypothesised to exist
based on MgO vs. siderophile element distributions by Arndt (1991).
The KDE method can also be used to statistically estimate the
number of modes in a population (Silverman, 1981) but the method
depends on the errors of each measurement. Because GEOROC does not
report precision and accuracy of individual data, reasonable errors were
assigned to data as follows: from 0.1 to 1.0 ± 1; > 1.0–5.0 ± 0.2;
> 5.0 ± 0.5 wt% MgO. When applied to the combined dataset from
all cratons (see Fig. 7a), this method finds four prominent populations:
one felsic with 1.0 wt% MgO accounting for 18.1% of the data; one
dominant basaltic with 6.0 wt% MgO containing 50.6% of the data; and
two ultramafic at 21.2 and 35.7 wt% MgO, respectively, collectively
accounting for 31.3% of the data. Similar calculations were performed
for all individual datasets displayed on Fig. 6 (panels A–G) and results
are listed in Table 1. The preferred number of populations varied be-
tween three and five but in all cases, there is a felsic population at
1.2 ± 0.59 wt% accounting for 18.6 ± 7.0% of the data and a domi-
nant basaltic population at 6.3 ± 1.4 wt% MgO containing roughly
half of the data (47.9 ± 7.6). The next important population is ultra-
mafic with 28.3 ± 2.4 wt% MgO making up 21.9 ± 4.1% of the data.
B.S. Kamber and E.L. Tomlinson Chemical Geology 511 (2019) 123–151

Table 1
Summary of population results of KDE analysis.
Number of Histogram bin KDE KDE populations
samples width band-
width Felsic Tholeiitic High-Mg basaltic to Ultramafic Highly
picritic ultramafic

MgO wt% % MgO wt% % MgO wt% % MgO wt% % MgO wt %

%

Baltic shield 1658 0.80 0.35 1.2 12.4 6.0 35.4 13.5 16.5 25.1 26.5 37.1 9.2
Dharwar craton 1091 1.28 0.31 1.0 27.7 8.6 53.1 29.5 19.2
Kaapvaal craton 1623 1.10 0.40 0.8 9.9 4.8 44.8 13.0 25.7 32.0 19.6
North Atlantic craton 702 0.80 0.30 2.5 21.4 8.1 55.0 16.9 16.9 36.5 6.7
Superior craton 2845 0.93 0.40 1.0 26.1 5.8 53.5 27.6 20.4
Western Australian cratons 2009 1.00 0.38 0.8 12.7 5.3 40.8 15.4 16.2 29.0 17.8 43.3 12.5
Zimbabwe, Tanzania, Slave 732 1.00 0.40 1.1 19.8 5.9 52.6 26.7 27.6
and North China cratons
All cratons 10,660 1.00 0.35 1.0 18.1 6.0 50.6 21.2 18.8 35.7 12.5
Averages of populations 1.2 18.6 6.3 47.9 14.7 18.8 28.3 21.9 39.0 9.5
Standard deviation 0.6 7.0 1.4 7.6 1.8 4.6 2.4 4.1 3.8 2.9

It is found in all datasets but the North Atlantic craton. A high-Mg
basalt population with 14.7 ± 1.8 wt% is found on four cratons and on
these it collectively contains 18.8 ± 4.6% of the data. Finally, a highly
ultramafic population with 39.0 ± 3.8 wt% MgO (likely dominated by
olivine cumulate rocks) is also found on four cratons, there containing
9.5 ± 2.9% of the data.
It is not the intention to ascribe specific petrologic significance to
these statistical analyses but the study of the histograms shows that the
dominant Archaean volcanic rock was basalt and the second-most im-
portant volcanic rock type was ultramafic. The broadly komatiitic rocks
are almost certainly over-represented in global datasets due to the
understandable scientific interest in these rocks. However, the im-
portant finding of the paucity of picritic compositions (reflected in the
absence of an identified population in the 15–20 wt% MgO range)
cannot be an artefact of overly keen interest in komatiite, which would
be difficult to distinguish from altered picrite or komatiitic basalt in the
field or the core shed. Finally, the analysis also shows that Archaean
basalt had nearly indistinguishable average MgO content and dis-
tribution as during the Phanerozoic (e.g. Kamber, 2015).
Because Ni content increases strongly with degree of melting and
hence Mg content (Arndt, 1977b), the Ni concentrations of volcanic
rocks of the same cratons was also compiled. Many published Ni ana-
lyses are performed by XRF on ignited powder and therefore con-
centrations were not corrected for loss on ignition. A total of 9288
analyses were compiled (for metadata see Table_S1) and of these 37 had
Ni concentration > 5000 ppm and were not further considered. Due to
the huge range in Ni concentrations (i.e. 0.1 to 5000 ppm), the linear
histograms show very little detail and instead a logarithmic translation
was performed (Fig. 7b). This very clearly shows two dominant popu-
lations representing basalt with ca. 200 ppm Ni and komatiite with ca.
2000 ppm Ni. The felsic population is spread over many bins with a
minor peak at ca. 20 ppm. Thus, the Ni distribution shows a similar
pattern as the MgO distribution, highlighting the dearth of picritic
compositions.
The final analysis was a comparison between rocks 2.5 to 2.9 Ga in
age and those emplaced earlier. This was motivated by the proposal
that there was no sharp A-P boundary but that geology instead changed
in a transitional fashion from ca. 3 Ga onwards (e.g. Shirey and
Richardson, 2011). More recently, Tang et al. (2016) and Smit and
Mezger (2017) suggested that the UCC had started to become more
felsic from 2.9 Ga onwards (see Fig. 2c) prompting an interest to study
the chemistry of volcanic rocks emplaced either side of 2.9 Ga. For this
comparison, data from the Kaapvaal, Zimbabwe, Dharwar, Superior and
Western Australian cratons were separated according to preferred age
stated in the original publications. These cratons expose a higher pro-
portion of Meso- and Palaeoarchaean rocks than most others.
The comparison shows, both in terms of MgO (Fig. 7c, d) and Ni
(Fig. 7e, f), that there is no obvious distinction between the two age
groups, suggesting that the basaltic mode always dominated and that an
equally ultramafic komatiite mode persisted right through the Archaean
Eon. This is consistent with the observation of time-series of maximum
inferred MgO contents of komatiites, showing no clear trend towards
lower MgO with decreasing age (e.g. Campbell and Griffiths, 1992;
Kamber, 2010). There is a small indication that picrite has become
somewhat more common since 2.9 Ga but we note that this could be an
artefact of the lower density of data for the > 2.9 Ga age group.
4.2. Comparison of compositions of komatiites with predictions of melting
models
4.2.1. A comparison of komatiite parental melt composition with
experimental liquids and calculated melts of previously depleted mantle
It is widely argued that many observed komatiite liquid composi-
tions are not easily reconciled with experimental liquids, calling for a
variety of processes, such as fluid-induced melting (e.g. Herzberg,
2016) or ultra-deep melt segregation (e.g. Robin-Popieul et al., 2012) to
explain the discrepancies. Komatiites are traditionally classified into Al-
depleted and Al-undepleted based on CaO/Al2O3 and Al2O3/TiO2 sys-
tematics (Nesbitt et al., 1979). In partly serpentinised samples, Al2O3/
TiO2 is more robust and samples with values < 15 are classed as Al-
depleted, whereas those with 15 > Al2O3/TiO2 < 25 are termed Al-
undepleted. Rarer samples with Al2O3/TiO2 > 25 are termed Ti-de-
pleted. In view of the importance of the Al2O3/TiO2 ratio, we revisited
the data from the Takahashi et al. (1993) and Walter (1998) experi-
ments relating to the Al2O3 and TiO2 systematics of experimental li-
quids as a function of melt fraction. These datasets provided chemical
analyses for all experimental phases as well as modal abundance esti-
mates. Importantly, their phase relations are compatible with more
experiments that do not report modal abundance.
The first thing to note is that whereas Al-uptake into the melt is
complex (Fig. 8a), Ti essentially behaves as an incompatible element
with strong initial enrichment in low degree melts (Fig. 8b), regardless
of pressure (see also Longhi, 2002). As expected for an incompatible
element, at all degrees of partial melting, the Ti content in the liquid is
higher than in the starting lherzolite. The two highest melt fraction
liquids of Takahashi et al. (1993) are an apparent exception to this
trend (Fig. 8b) but this is likely an artefact of rounding low con-
centration values to too few significant digits in the published data
table. Regardless, when the Al2O3/TiO2 ratio is plotted against Al2O3
(Fig. 8c) and TiO2 (Fig. 8d) it is evident that TiO2 exerts dominant
control over the ratio (see also Robin-Popieul et al., 2012). The im-
portant inference arising from the experimental data is that an

132
B.S. Kamber and E.L. Tomlinson Chemical Geology 511 (2019) 123–151

A B All cratons
All cratons
(n=10660) (n=9751)

n n

MgO wt% Ni (ppm)

C 2.5-2.9 Ga samples D
>2.9 Ga samples
(n=1539)
(n=647)

n n

MgO wt% MgO wt%

E 2.5-2.9 Ga samples F >2.9 Ga samples

(n=1618) (n=585)

n n

Ni (ppm) Ni (ppm)

Fig. 7. Magnesium (MgO wt%) and Ni (ppm) distribution in Archaean metavolcanic rocks. Panel (A) shows linear histogram of a MgO on all studied cratons with four
vertical lines denoting the preferred populations from kernel density function statistics. Panels (B) to (F) all show data after logarithmic translation. Panel (B) displays
Ni distribution in all compiled Archaean metavolcanic rocks, highlighting the dominant basaltic population at ca. 200 ppm and the ultramafic population centred
over ca. 2000 ppm. Panels (C) and (D) show a comparison in MgO distribution between 2.5 and 2.9 and > 2.9 Ga metavolcanic samples. Panels (E) and (F) show the
same kind of comparison for Ni distribution.

originally lherzolitic mantle source that became partially depleted (e.g.
via extraction of 10–20% melt) will have a substantially higher Al2O3/
TiO2 ratio than primitive mantle (ca. 22) or the N-MORB source (ca.
25). If such a source was later remelted, it would yield liquids with very
low Ti contents (< 0.2 wt%) and very high Al2O3/TiO2 ratios.
There is currently a lack of experimental komatiitic compositions
with the very low TiO2 concentrations found in some natural komatiites
(e.g. Wilson, 2003). This very likely stems from a lack of successful
experiments from refractory (harzburgitic) starting compositions. To
gain a first-order understanding of what liquids from refractory mantle
could be we calculated the differential melt compositions from isobaric
series of the Walter (1998) and Takahashi et al. (1993) experiments
assuming mass conservation (Table_S2). The validity of these calcula-
tions depends on the quality of the chemical data for the two batch melt
liquids, the accuracy of the melt fraction estimate, and the stability of
the phases at the liquidus. The results show (Fig. 8e) that all the

133
B.S. Kamber and E.L. Tomlinson Chemical Geology 511 (2019) 123–151

16 2.0
A Takahashi et al. (1993) B
14 Walter (1998)
Charge lherzolites 1.5
12

TiO (wt%)
Al O (wt%)

10

2
1.0
3

8
2

6
0.5

2 0.0
0 20 40 60 80 100 0 20 40 60 80 100
Melt fraction (%) Melt fraction (%)
18 2.0
C D
16

1.5
14
Al O (wt%)

TiO (wt%)
12
3

2
1.0
2

10

8
0.5

4 0.0
0 10 20 30 40 50 60 0 10 20 30 40 50 60
Al O /TiO Al O /TiO
2 3 2 2 3 2

8 50
E <20% fertile mantle melts F
7 >20% fertile mantle melts
49
6 Refractory mantle melts
SiO (wt%)

5 48

4
2
n

47 Batch melt f>

3
Batch melt f<
2 Calculated
46 fractional melt

0 45
0 20 40 60 80 100 120 140 160 180 18 20 22 24 26 28 30 32 34
Al O /TiO MgO (wt%)
2 3 2

60 60
G Batch melt
Accumulated
H 3 GPa
4,4.6&5 GPa
fractional melt
6 GPa
Takahashi et al.
(1993) liquids 6.5 GPa
55 Walter (1998)
55 7 GPa
liquids
SiO (wt%)

SiO (wt%)

50 50
2

45 45

40 40
0 10 20 30 40 50 60 0 10 20 30 40 50 60
MgO (wt%) MgO (wt%)

(caption on next page)

134
B.S. Kamber and E.L. Tomlinson Chemical Geology 511 (2019) 123–151

Fig. 8. Selected geochemical systematics of high-P experiments relevant to komatiite petrogenesis. Panels (A) and (B) show dependence of Al2O3 and TiO2 con-
centration, respectively, on extent of melting for experiments performed by Takahashi et al. (1993) and Walter (1998) on slightly different lherzolite starting
compositions (also shown). Panels (C) and (D) show dependence of the Al2O3/TiO2 ratio of liquids on Al2O3 and TiO2 content, respectively. Same symbols as panel
(A). Panel (E) is a histogram of Al2O3/TiO2 ratios in low (< 20%) and higher (> 20%) degree liquids from fertile lherzolites (same data sources). Also shown are
reconstructed hypothetical partial melts from refractory, previously melted harzburgite listed in Table_S2. See text for explanation. Panel (F) shows SiO2 vs MgO
relationships for three examples of calculated melts from refractory mantle. Blue and green colours are from 3 and 7 GPa experiments of Walter, purple from 4.6 GPa
experiment of Takahashi et al. (1993). In each case a higher and lower degree batch melt from fertile mantle is shown with diamond and circle symbols. The
hypothetical refractory melt is represented by the square symbol and corresponds to the composition of the liquid required to move from the lower degree to the
higher degree melt along the tie line. The distances reflect the lever rule to preserve mass. Panel (G) shows the high-Mg basalts, picrites and komatiites from (Sproule
et al., 2002) in SiO2 vs MgO space with superimposed model compositions (batch and accumulated fractional melts from Herzberg (2016)) and liquids of experiments
by Takahashi et al. (1993) and Walter (1998). Panel (H) shows the same data and model but with calculated hypothetical refractory melts (see examples in panel F)
from previously depleted mantle. Symbols are colour coded for pressure. Note that refractory mantle melts can reach very high MgO values with highly variable SiO2
content, reflecting exhaustion of pyroxene

biotite
A B C granite

Timiskaming biotite-amphibole
2687-2675 * granodiorite
Porcupine
2696-2690 * biotite
granodiorite
Blake River
2701-2697
biotite
Kinojevis granodiorite
2702-2701

Tisdale
2710-2703

Kidd-Munro
white and grey
2719-2711 granodorite

Stoughton-
Roquemaure
2723-2720 grey orthogneiss

Deloro
2730-2724 amphibole - biotite
tonalite

Pacaud
2750-2735
V S SiO 2

Conglomerate Intermediate / felsic Sproule et al. (2002)

volcanic rocks
Herzberg (2016)
Turbidite Mafic volcanic rocks

Iron formation Ultramafic volcanic rocks

* Feng + Kerrich (1990)
Barrie (2005)

Fig. 9. Stratigraphic column (panel A) of the Abitibi greenstone belt after Ayer et al. (2002). Seven volcanic dominated assemblages are unconformably overlain by
two sediment-dominated assemblages at the top. Column (B) shows approximate position of mafic to ultramafic metavolcanic samples interrogated by Sproule et al.
(2002) and Herzberg (2016) as well as sedimentary samples described by Feng and Kerrich (1990) and Barrie (2005). Column (C) shows evolutionary trend of
plutonic samples with time recorded in the Kenogamissi complex (after Benn, 2004), starting with tonalite and grey gneisses, moving to granodiorite, biotite-bearing
granodiorite and ending with proper granite.

135
B.S. Kamber and E.L. Tomlinson Chemical Geology 511 (2019) 123–151

16
100
A (n = 2048) B Al-undepleted
Al-depleted
Ti-depleted
14 Batch melt
Accumulated
80 fractional melt
Continental crust
12

FeO (wt%)
60
10
n

40
8

20 6

0 4
12 16 20 24 28 32 36 40 44 48 0 10 20 30 40 50 60
MgO (wt%) MgO (wt%)
60 3.5
C Al-undepleted
Al-depleted Al-undepleted D
Ti-depleted 3.0 Al-depleted
Batch melt Ti-depleted
55 Accumulated
fractional melt
Continental crust
2.5
SiO (wt%)

n
(Sm/Yb)
2.0
50
2

1.5

1.0
45

0.5

40 0.0
0 10 20 30 40 50 60 0 10 20 30 40 50
MgO (wt%) Al O /TiO
2 3 2

60 60
E Kidd-Munro
Stoughon-Roquemaure
F Tisdale
Blake River
Pacaud Baby Group
55 Batch melt 55 Batch melt
Accumulated Accumulated
fractional melt fractional melt
SiO (wt%)

SiO (wt%)

50 50
2

45 45

40 40
0 10 20 30 40 50 60 0 10 20 30 40 50 60
MgO (wt%) MgO (wt%)

16 16
G H
14 14

12 12
FeO (wt%)

FeO (wt%)

10 10

8 8

6 6

4 4
0 10 20 30 40 50 60 0 10 20 30 40 50 60
MgO (wt%) MgO (wt%)

(caption on next page)

136
B.S. Kamber and E.L. Tomlinson
Fig. 10. Major element and selected REE systematics in Abitibi and adjacent greenstone belt high-Mg basalts, picrites and komatiites (all data from Sproule et al.,
2002). Panel (A) shows a histogram of MgO (wt%) content of all least-altered samples. Panels (B) and (C) show MgO (wt%) vs FeO (wt%) and SiO2 (wt%) systematics,
respectively. The samples are grouped into Al-undepleted, Al-depleted and Ti-depleted suites according to the Al2O3/TiO2 criteria > 25, 15–25, < 15. Shown for
reference are the position of bulk continental crust (from Rudnick and Gao, 2003) and the preferred modelled parental komatiite melts by Herzberg (2016) also
shown on Fig. 5. Panel (D) plots Al2O3/TiO2 versus CI-chondrite (after Anders and Grevesse, 1989) normalised Sm/Yb. Panels (E) and (F) show MgO (wt%) vs SiO2
(wt%) systematics separated into 6 stratigraphic assemblages. Panels (G) and (H) show MgO (wt%) vs FeO (wt%) systematics separated into 6 stratigraphic as-
semblages (same symbols as panels (E) and (F)).
hypothetical melts from variably refractory mantle have higher Al2O3/
TiO2 than the starting lherzolite. Hypothetical melts from very highly
depleted mantle have very low TiO2 concentrations, in many cases re-
sulting in negative apparent TiO2 values. These are artefacts of the
cumulative uncertainties that could be removed if the Ti concentrations
of high degree melts were analysed with dedicated long-acquisition
time electron microprobe analyses (e.g. Donovan et al., 2011). Never-
theless, the low Ti concentrations (0.1–0.2 wt%) and the high Al2O3/
TiO2 ratios (25–200) compare well with the ranges reported for Ti-
depleted komatiites of the Abitibi greenstone belt (Sproule et al., 2002)
and in so-called Al-enriched komatiites of the Barberton greenstone belt
(e.g. Wilson, 2003; Robin-Popieul et al., 2012). Thus, at least some
komatiites could have resulted from repeated 10–20% melting of a
progressively more refractory source, overcoming the issue of requiring
high-degree melts that would be too buoyant not to escape (Arndt,
1977a; Schmeling and Arndt, 2017).
We adopted the same approach to explore whether melts from
variably refractory peridotite could help to resolve the SiO2 vs. MgO
mismatch documented by Herzberg (2016). Fig. 8f shows three ex-
amples for 3, 4.6 and 7 GPa. The 3 GPa example is the composition of
the melt required to take the 24% melt (experiment 30.14 of Walter,
1998) to 53% (experiment 30.11). It has 33.6 wt% of MgO and 45.7 wt
% of SiO2. The 4.6 GPa example shows the composition of the melt
required to take the 22% liquid (experiment 1750 °C of Takahashi et al.,
1993) to 55% (experiment 1800 °C). The 7 GPa example is the com-
position of the melt required to take the 16% melt (experiment 70.07 of
Walter, 1998) to 47% (experiment 70.09). As can be seen by comparing
panels (G) and (H) of Fig. 8, the hypothetical melts extend to higher
MgO and more variable SiO2 than the experimental liquids from fertile
mantle and qualitatively illustrate that melting refractory peridotite
could be one potential explanation for the Si/Mg and Al/Ti systematics
of Archaean komatiites and picrites. Takahashi (1986) suggested that
the range in SiO2 values could reflect the presence of absence of pyr-
oxene at the liquidus but these findings require further experimental
and petrological exploration.
4.2.2. An extended comparison of Abitibi and Pontiac Subprovince
komatiites with melting experiments of fertile mantle
The Neoarchaean Abitibi greenstone belt in Ontario and Quebec is a
classic Subprovince of the Superior craton in which komatiites and
mafic meta-volcanic rocks have been studied extensively. Because the
greenstone belt apparently did not form onto pre-existing evolved crust,
most of its igneous rocks preserve least contaminated mantle radiogenic
isotope signatures (e.g. Ketchum et al., 2008) and are considered ‘ju-
venile’. An outstanding feature of the Abitibi greenstone belt is that
across the nearly 700 km width of exposure, it displays a coherent
vertical stratigraphy (e.g. Thurston, 2002) constrained with a very large
number of high-precision U/Pb zircon dates from felsic volcanic rocks
(e.g. Corfu, 1993). The huge wealth of information pertaining to the
stratigraphic relationships is not widely appreciated by the academic
community, probably because most of the data support mineral ex-
ploration and are published in reports of geological survey organisa-
tions. Berger et al. (2011) explained that high-precision U/Pb zircon
dates allowed the 55 rock assemblages recognised in 1991 to be sim-
plified into the 9 assemblages shown in Fig. 9. The new stratigraphy
comprises 7 assemblages dominated by volcanic rocks and an un-
conformably overlying set of two sedimentary rock assemblages. The 7
137
Chemical Geology 511 (2019) 123–151
assemblages dominated by volcanic rocks are separated by condensed
horizons (Thurston et al., 2008) composed of iron formations, chert and
minor fine-grained siliciclastic sedimentary rocks (argillites).
The quality of stratigraphic control allows for an accurate assess-
ment of petrological and potential temporal variety of komatiite within
a greenstone belt. Komatiite occurs in four of the 7 lower assemblages
(Fig. 9). The most widely known komatiites (Pyke Hill and Alexo) be-
long to the Kidd-Munro assemblage and have been extensively studied.
Analysing major element systematics of these two occurrences,
Herzberg (2016) established that batch melting produces a closer fit to
observed MgO vs. FeO systematics than accumulated fractional melting
(Fig. 5b) and furthermore, that compared to model predictions, the
empirical data show a deficit in SiO2 relative to MgO.
Here we revisit the much larger dataset of Sproule et al. (2002) who
interrogated major and trace element geochemistry of 2466 chemical
analyses of high-Mg basalt, picrite and komatiite. Their sample suite
includes representatives from all four ultramafic occurrences within the
Abitibi greenstone belt as well as two younger occurrences in adjacent
greenstone belts to the south (Pontiac Subprovince). They are age
equivalent to the Blake and Timiskaming assemblages of the Abitibi
greenstone belt and their chronologic positions are shown in Fig. 9. To
achieve their wide stratigraphic and geographic coverage, Sproule et al.
(2002) also analysed variably altered samples. After using the excessive
loss on ignition criterion of Sproule et al. (2002) to avoid samples with
strongly modified major element chemistry, 2048 analyses remained.
We first tested whether this dataset also showed the relative dearth
of picritic samples with MgO between 15 and 20 wt% by plotting the
anhydrous rock MgO content in a histogram (Fig. 10a). As can be seen,
the dominant population of data occurs around an MgO content of ca.
28 wt% with another prominent mode at ca. 40 wt% MgO, probably
dominated by olivine-rich cumulates. The basaltic mode is not within
the range of compositions studied by Sproule et al. (2002) whose da-
taset excluded most samples with MgO < 12 wt%. However, the pau-
city of samples with MgO of 15–20 wt% is nonetheless clearly visible,
confirming the picture from the preceding global analysis (Fig. 7).
In the context of this treatment, the most interesting phenomenon of
the Abitibi greenstone belt ultramafic rocks is their great chemical
variability. If the conventional komatiite classification criteria are ap-
plied to all rocks in Sproule et al. (2002) dataset (not just komatiites),
76% classify as Al-undepleted, 17% as Al-depleted and 7% as Ti-de-
pleted types. Panels (B) and (C) of Fig. 10 show similar FeO vs. MgO
and SiO2 vs. MgO diagrams as Fig. 5. In this much larger dataset, it is
evident that for any given MgO value there is a wide range in FeO
(4–5 wt%) and SiO2 (5–7 wt%) content. Contamination with con-
tinental crust and alteration are unlikely to explain the full spread in
FeO vs. MgO space. Some of the most Fe-rich samples may contain
excess accumulated chromite with or without olivine removal (see also
Fig. 4b of Sproule et al. (2002)) but even allowing for this process, a
wide range in Mg# remains. The FeO, SiO2 and MgO systematics do not
correlate in any way with the extent of Al- and Ti-depletion. In SiO2 vs.
MgO space (Fig. 10c), the more elevated SiO2 values observed in some
samples could be explained with continental contamination or post-
emplacement chert infiltration (e.g. Thurston et al., 2012) but the sub-
silicic samples remain unexplained, confirming the finding of Herzberg
(2016) regardless of whether the melts are better approximated by
batch or accumulated fractional melting models.
A smaller subset of samples also reports REE data for which CI-
B.S. Kamber and E.L. Tomlinson
normalised Sm/Yb ratios were calculated (Sm was preferred over Gd as
more samples had values for the former). When plotted versus the
Al2O3/TiO2 ratio (Fig. 10d), the paradox explained by Walter (1998)
becomes apparent. Namely, only the most Al-depleted samples show
somewhat elevated Sm/Yb ratios whereas for samples with Al2O3/TiO2
from 10 to 50, there is no correlation, which is unexpected if garnet
controlled the degree of Al-depletion or enrichment (e.g. Nesbitt et al.,
1979).
The final four panels of Fig. 10 show the same data but now colour-
coded for stratigraphic assemblage to test whether the large petrolo-
gical variability could correlate with age. However, there is no age
progression in FeO vs. MgO and SiO2 vs. MgO systematics. Smaller
datasets (e.g. Pacaud) show tighter systematics but in the three domi-
nant subsets (Kidd-Munro, Stoughton-Roquemaure and Tisdale), the
full range of major element chemistries is represented.
In summary, high-Mg basalts, picrites and komatiites from the
Abitibi and adjacent Pontiac Subprovince represent very wide petro-
logical variety, including melts from more fertile mantle and from re-
fractory sources. These diverse komatiites occur at several stratigraphic
levels but are volumetrically much inferior to the dominant basalts
(Fig. 9). Over the reasonably narrow ca. 60 Ma time window, multiple
mantle sources melted, likely at quite different depths.
4.3. Garnet and orthopyroxene textural relationships and REE systematics
in Archaean orthopyroxene-rich cratonic harzburgite
The early studies on the relationships between Archaean cratonic
harzburgites and komatiites established a paradigm of very deep
(5–8 GPa) melting (e.g. Takahashi and Scarfe, 1985; Cox et al., 1987;
Boyd, 1989; Canil, 1992; Herzberg, 1993). This has since been replaced
with the currently prevailing idea (e.g. Canil, 2004; Lee and Chin, 2014;
Regier et al., 2018) that proto-cratonic lithosphere thickened by
stacking of harzburgite, which represents shallow melt residues
(1–3 GPa). The paradigm shift was precipitated with the realisation that
the minor and REE systematics of komatiites, Archaean cratonic garnets
(Stachel et al., 1998), and of harzburgites (Canil, 2004) favoured melt
extraction in the spinel stability field, implying subsequent thickening
of the cratonic root. The high Cr/Al ratio of many cratonic harzburgites
is quoted as very strong evidence against melt extraction deeper than
3 GPa, so much so that deeper melting is no longer discussed as a viable
alternative by some authors (e.g. Regier et al., 2018). However,
Tomlinson et al. (2018) offered observations that warrant a new look at
the old paradigm of deep melting. These authors used phase relation-
ships in a very coarse-grained orthopyroxene megacryst to postulate
that most garnet in Archaean harzburgite was not originally present but
formed by exsolution from orthopyroxene. In other words, during
proto-craton formation the harzburgite was much hotter and only
consisted of olivine and orthopyroxene, even though it technically re-
sided in the stability field of garnet defined for pyrolitic mantle. Harz-
burgitic compositions can become garnet-free at the deep solidus and
the Cr/Al systematics of experimental olivine and orthopyroxene re-
sidues (e.g. Walter, 1998) are compatible with the observed high Cr/Al
many cratonic harzburgites at pressures of 3–7 GPa, provided the re-
sidue is highly depleted. Below we provide more data on the spatial
association of garnet and orthopyroxene that has long been argued to
contain evidence about the cooling trajectory (e.g. Canil, 1991) of
harzburgites and also document much-needed REE concentration data
for harzburgitic orthopyroxene.
4.3.1. The spatial relationship between garnet and orthopyroxene in garnet
harzburgites
Cox et al. (1987) first systematically studied the spatial relationships
of olivine, clinopyroxene, orthopyroxene and garnet in coarse-grained
cratonic lherzolites. Their observations revealed one consistent phase
relationship, namely a close spatial association of garnet and ortho-
pyroxene. Because garnet was only found in association with
138
Chemical Geology 511 (2019) 123–151
orthopyroxene, they reasoned that garnet represented the Tschermak
component of an original, higher-temperature orthopyroxene. Thus,
garnet was a phase that exsolved from a parental higher-Al orthopyr-
oxene upon cooling. Considering the experimental finding of Kushiro
(1973), Cox et al. (1987) speculated that the more aluminous ortho-
pyroxene precursors could have originated at pressures ranging from 4
to 8 GPa.
Using SEM-BSE, Saltzer et al. (2001) confirmed that the close spatial
association of garnet and orthopyroxene also exists in more refractory
harzburgites. They further discovered that those peridotites that had
equilibrated at the greatest depth along the cratonic geotherm contain
the most magnesian olivine. This observation does not in itself con-
strain the cooling trajectory, which Saltzer et al. (2001) suggested to
have involved compression, a conclusion derived from comparison with
experiments conducted on fertile peridotite compositions.
Tighter constrains on the cooling trajectory can be derived from
analysis of spectacular examples of very coarse-grained orthopyroxenes
that exhibit textural evidence for garnet exsolution. Gibson (2017)
documented a range of garnet exsolution styles from very coarse-
grained orthopyroxenes from the Tanzania and Kaapvaal cratons. A
similar exceptionally well-preserved specimen of a large orthopyroxene
single crystal (size of preserved relic 15 × 12 × 7 cm) from the Kim-
berley fields was studied by Tomlinson et al. (2018). Thermobarometry
inferred equilibration at 4.4 GPa. By reconstituting the exsolved garnet
(and, in some of the samples studied by Gibson (2017) additional
clinopyroxene) into the orthopyroxene host, the Tschermak component
of the original precursor pyroxene can be calculated and plotted as an
isopleth into the garnet-orthopyroxene-olivine stability field of a model
pseudosection. Provided that orthopyroxene formed on the solidus and
that it was the only phase formed on the solidus, the original depth of
formation can be obtained as the intersection of the isopleth with the
solidus. In the cases of very large euhedral macrocrysts this assumption
seems warranted and using this approach, Tomlinson et al. (2018)
calculated that the precursor pyroxene of their studied sample formed
at 5.9 GPa and 1750 °C, implying decompression upon cooling onto the
cratonic geotherm. Tomlinson et al. (2018) noted that orthopyroxene
with this Tschermak component could theoretically also exist on the
very low pressure solidus (0.3 GPa) within the plagioclase stability field
but this is not considered a likely formation depth of any peridotites
from the Kaapvaal craton (see also Lee and Chin, 2014).
It could be argued that the spectacular coarse-grained orthopyrox-
enes and their exsolved garnets are exotic curiosities with limited sig-
nificance for the overall formation of the cratonic mantle. However, an
important aspect of the specimen studied by Tomlinson et al. (2018) is
that the megacryst is seen to disaggregate along its edges and that
fragments of it, with garnet inclusions, are found in the enclosing
harzburgite. Garnet and orthopyroxene in the harzburgite have exactly
the same major and trace element composition as in the megacryst. The
orthopyroxene carries substantial internal strain, a feature also noted
by Gibson (2017), who attributed this to the mechanism of garnet ex-
solution but that could additionally have been caused during transport
within the cratonic root. Considering that the peridotites apparently
experienced vertical movement over tens of km, orthopyroxene dis-
aggregation may have been a common process. This raises the possi-
bility that a substantial proportion of garnet found in cratonic garnet
harzburgites could have originated via exsolution, even if in most cases
of granular peridotites, the exsolution process cannot be directly proven
(Tomlinson et al., 2018). For example, Wasch et al. (2009) described
conspicuous clots of coarse-grained orthopyroxene and garnet with
some garnet visibly exsolved as linear bands within the orthopyroxene.
In Fig. 11, a series of section scans illustrates the progression from
an example of an orthopyroxene megacryst with clear garnet exsolution
lamellae to a ‘normal’ granular garnet harzburgite, with no obvious
exsolution indication of the garnet. The intermediate examples are
progressively less obvious cases for exsolution highlighting examples of
garnet grains fully encased in orthopyroxene and relic alignment trails
B.S. Kamber and E.L. Tomlinson
Fig. 11. Transmitted light section scans of five garnet peridotites from the
Bultfontein pans (Kimberley, Kaapvaal craton) arranged by decreasing petro-
graphic evidence for garnet exsolution (note that these are over-thick thin
sections). From the top, sample BP002 (Tomlinson et al., 2018, this study) with
regular garnet lamellae in orthopyroxene; 17BSK069d with semi-regular garnet
lamellae and lines of equant garnet in a single large (32 mm) orthopyroxene;
CLA18 with large (~10 mm) orthopyroxene crystals, where garnet elongate and
aligned is located within or adjacent to the orthopyroxene; granular peridotite
CLA05, in which garnet occurs as inclusions within and adjacent to coarse
(< 5 mm) orthopyroxene; and garnet harzburgite 17BSK051 (this study) in
which garnet grains are spatially associated with coarse orthopyroxene.
of garnet. A common feature of all samples (and most harzburgites from
Kimberley) is the very coarse-grained nature of the orthopyroxene,
often exceeding 10–15 mm in diameter. To further test the possibility
that garnet in ordinary harzburgite originated by exsolution from for-
merly more aluminous orthopyroxene (e.g. Canil, 1991), we used au-
tomated phase mapping with large format SEM-EDS detectors to test
139
Chemical Geology 511 (2019) 123–151
Fig. 12. Transmitted light scan of garnet harzburgite 17BSK051 (A) and three
phase maps. Panel (B) shows a composite map with olivine in green, ortho-
pyroxene in blue, garnet in red and phlogopite in purple. Panels (C) and (D)
show isolated maps of garnet with olivine and orthopyroxene, respectively. See
text for explanation.
the spatial relationship between garnet and orthopyroxene. The studied
sample, 17BSK051, is a typical Kimberley field cratonic granular
harzburgite, consisting of olivine, orthopyroxene and minor garnet with
late accessory phlogopite. Garnet-orthopyroxene-olivine thermo-
barometry (Harley, 1984; Brey and Köhler, 1990) indicates equilibra-
tion at 954 ± 5 °C and 3.8 ± 0.3 GPa, on the typical cratonic geo-
therm.
The fully-quantitative phase maps were analysed with the EDS
system software tool ‘AutoPhaseMap’, which provides area and mean
compositional information on identified phases. Shown together with a
transmitted light thin section scan (Fig. 12) is a composite phase map
and two isolated phase maps highlighting garnet, one alongside olivine,
the other alongside orthopyroxene. These maps qualitatively show the
close spatial association between garnet and orthopyroxene and the
general paucity of shared grain boundaries between olivine and garnet.
In two cases, garnet is nearly completely enclosed in orthopyroxene. In
terms of area fraction of the three main phases, olivine occupies 68%,
orthopyroxene 29% and garnet 3.0%. The mean Mg# of olivine from
the phase map is 93.1, comparing very well with the mean of five spot
analyses (93.2; Table 2). In the olivine mode vs. olivine Mg# diagram
(Fig. 3f), the sample plots close to the 7 GPa residue trajectory (Walter,
1998) and is similar in mode and olivine composition to many of the
Kaapvaal garnet harzburgites originally described by Boyd (1989).
B.S. Kamber and E.L. Tomlinson Chemical Geology 511 (2019) 123–151

Table 2
Mineral compositions and REE concentrations (ppm).
Sample MINM25-35 diopside BP002 17BSK051

(wt%) Certified This study ± garnet ± orthopyroxene ± garnet ± orthopyroxene ± olivine ±

(n=4) (n=15) (n=18) (n=5) (n=5) (n=5)

SiO2 55.34 55.27 0.25 42.09 0.10 58.16 0.11 41.55 0.14 58.08 0.10 41.07 0.06
TiO2 0.00 0.00 0.00 0.02 0.02 0.01 0.02 0.02 0.05 0.00 0.00 0.00 0.00
Al2O3 0.62 0.61 0.04 22.47 0.19 0.73 0.02 20.81 0.25 0.79 0.04 0.00 0.00
Cr2O3 0.00 0.00 0.00 3.58 0.19 0.27 0.03 5.29 0.18 0.35 0.02 0.00 0.00
MgO 17.76 17.83 0.14 22.23 0.09 35.87 0.11 21.40 0.20 36.96 0.10 52.73 0.11
FeO 0.83 0.86 0.03 6.32 0.06 3.99 0.05 6.54 0.06 4.22 0.07 6.86 0.04
NiO 0.00 0.00 0.00 0.03 0.02 0.12 0.03 0.00 0.00 0.03 0.07 0.42 0.03
MnO 0.00 0.02 0.04 0.27 0.02 0.07 0.02 0.28 0.02 0.04 0.05 0.06 0.05
CaO 24.80 24.93 0.15 4.40 0.05 0.33 0.02 5.24 0.09 0.42 0.02 0.00 0.00
Na2O 0.41 0.38 0.01 0.04 0.03 0.07 0.02 0.00 0.00 0.08 0.08 0.00 0.00
K2O 0.00 0.00 0.00 0.02 0.01 0.01 0.01 0.00 0.00 0.00 0.00 0.00 0.00
Total 99.76 99.89 101.45 99.62 101.13 100.96 101.138

Sample BP002 garnet BP002 kimberlite veins

Kimberlite contaminated Least contaminated

La 0.431 0.569 0.371 0.139 0.233 0.124 0.080 0.072 0.042 0.104 64.20 52.49
Ce 1.812 3.059 2.491 1.472 1.202 1.546 1.560 1.015 1.094 1.637 78.66 93.11
Pr 0.653 0.978 0.878 0.628 0.530 0.668 0.663 0.558 0.612 0.653 6.474 11.99
Nd 5.785 7.472 6.751 5.367 5.282 5.530 5.464 5.461 5.660 5.260 20.61 49.63
Sm 2.878 3.395 3.054 2.746 2.832 2.727 2.754 2.844 2.834 2.696 2.711 7.941
Eu 1.063 1.162 1.018 0.997 1.063 0.982 0.972 1.072 1.033 0.951 0.763 1.745
Gd 2.975 3.073 2.582 2.748 3.021 2.711 2.670 3.019 2.861 2.595 1.539 4.844
Tb 0.364 0.352 0.298 0.332 0.374 0.315 0.314 0.370 0.344 0.301 0.134 0.403
Dy 1.503 1.391 1.151 1.355 1.552 1.264 1.258 1.528 1.432 1.203 0.490 1.274
Ho 0.182 0.172 0.132 0.161 0.187 0.150 0.148 0.181 0.170 0.143 0.063 0.150
Er 0.272 0.242 0.202 0.243 0.273 0.226 0.220 0.271 0.253 0.215 0.105 0.246
Tm 0.026 0.026 0.024 0.025 0.027 0.023 0.023 0.027 0.026 0.023 0.010 0.024
Yb 0.180 0.185 0.186 0.172 0.178 0.165 0.171 0.180 0.182 0.160 0.044 0.098
Lu 0.045 0.045 0.048 0.043 0.042 0.041 0.042 0.044 0.043 0.040 0.005 0.011

Sample BP002 orthopyroxene

Kimberlite contaminated Least contaminated

La 0.0663 0.0064 0.7464 0.0374 1.2149 0.0067 0.0805 0.0091 0.0102 0.0065 0.0112 0.0058 0.0059
Ce 0.2103 0.0472 2.1948 0.2694 3.3054 0.0484 0.2305 0.0519 0.0581 0.0483 0.0527 0.0432 0.0453
Pr 0.0248 0.0106 0.2117 0.0399 0.2645 0.0107 0.0229 0.0101 0.0110 0.0109 0.0114 0.0094 0.0101
Nd 0.1130 0.0691 0.7589 0.1401 0.9251 0.0692 0.1175 0.0683 0.0681 0.0656 0.0662 0.0603 0.0645
Sm 0.0261 0.0217 0.1080 0.0189 0.1311 0.0247 0.0300 0.0253 0.0255 0.0246 0.0233 0.0218 0.0252
Eu 0.0076 0.0065 0.0230 0.0042 0.0314 0.0077 0.0076 0.0068 0.0078 0.0070 0.0072 0.0070 0.0070
Gd 0.0176 0.0152 0.0483 0.0095 0.0671 0.0177 0.0231 0.0171 0.0176 0.0186 0.0159 0.0171 0.0175
Tb 0.0016 0.0013 0.0030 0.0007 0.0060 0.0017 0.0018 0.0016 0.0016 0.0017 0.0016 0.0015 0.0016
Dy 0.0046 0.0043 0.0081 0.0023 0.0238 0.0056 0.0054 0.0053 0.0050 0.0053 0.0057 0.0050 0.0051
Ho 0.00054 0.00031 0.00067 0.00034 0.00271 0.00058 0.00060 0.00050 0.00049 0.00046 0.00049 0.00051 0.00053
Er 0.00061 0.00050 0.00080 0.00036 0.00427 0.00060 0.00103 0.00058 0.00048 0.00049 0.00054 0.00059 0.00062
Tm 0.00003 0.00002 0.00007 0.00003 0.00038 0.00005 0.00003 0.00002 0.00006 0.00003 0.00003 0.00001 0.00005
Yb 0.00021 0.00025 0.00026 0.00026 0.00149 0.00020 0.00033 0.00027 0.00027 0.00026 0.00030 0.00037 0.00022
Lu 0.00004 0.00009 0.00004 0.00006 0.00019 0.00003 0.00009 0.00004 0.00006 0.00006 0.00001 0.00008 0.00004

Sample 17BSK051 garnet 17BSK051 orthopyroxene

Least contaminated Strongly contaminated

La 0.355 0.020 0.026 0.019 0.019 0.020 0.023 0.017 0.132 0.0550 0.1360
Ce 1.110 0.430 0.398 0.395 0.382 0.403 0.468 0.316 0.693 0.1220 0.2540
Pr 0.224 0.174 0.173 0.178 0.176 0.176 0.190 0.144 0.209 0.0200 0.0251
Nd 2.730 2.801 2.690 2.850 2.893 2.785 2.820 2.380 2.890 0.0800 0.1460
Sm 3.540 3.400 3.510 3.610 3.520 3.430 3.260 3.210 3.450 0.0308 0.0419
Eu 1.090 1.088 1.068 1.052 1.070 1.055 0.991 1.020 1.092 0.0071 0.0123
Gd 3.660 3.600 3.610 3.730 3.710 3.620 3.220 3.270 3.570 0.0221 0.0420
Tb 0.448 0.424 0.451 0.460 0.439 0.446 0.412 0.420 0.435 0.0020 0.0048
Dy 3.230 3.120 3.149 3.150 3.170 3.099 2.740 2.840 3.010 0.0114 0.0116
Ho 0.395 0.373 0.375 0.384 0.387 0.383 0.341 0.343 0.381 0.0010 0.0016
Er 1.154 1.087 1.132 1.141 1.126 1.131 0.976 1.001 1.093 0.0017 0.0022
Tm 0.102 0.094 0.096 0.097 0.099 0.098 0.087 0.091 0.100 0.0002 0.0014
Yb 0.704 0.685 0.683 0.704 0.687 0.660 0.624 0.615 0.683 0.0140 0.0019
Lu 0.091 0.085 0.089 0.095 0.091 0.089 0.081 0.083 0.092 0.0014 0.0007

140
B.S. Kamber and E.L. Tomlinson Chemical Geology 511 (2019) 123–151

Sample 17BSK051 orthopyroxene

Somewhat contaminated Least contaminated

La 0.0133 0.0175 0.0142 0.0123 0.0082 0.0079 0.0060 0.0053 0.0069 0.0049 0.0086
Ce 0.0447 0.0406 0.0460 0.0375 0.0440 0.0355 0.0292 0.0314 0.0307 0.0266 0.0290
Pr 0.0094 0.0080 0.0081 0.0081 0.0071 0.0075 0.0064 0.0069 0.0066 0.0061 0.0071
Nd 0.0750 0.0639 0.0734 0.0785 0.0561 0.0589 0.0559 0.0665 0.0675 0.0664 0.0738
Sm 0.0296 0.0320 0.0325 0.0333 0.0273 0.0272 0.0370 0.0324 0.0349 0.0337 0.0356
Eu 0.0064 0.0075 0.0076 0.0076 0.0078 0.0099 0.0089 0.0068 0.0071 0.0067 0.0066
Gd 0.0173 0.0181 0.0187 0.0191 0.0211 0.0206 0.0270 0.0176 0.0188 0.0180 0.0191
Tb 0.0017 0.0034 0.0018 0.0023 0.0027 0.0050 0.0024 0.0023 0.0028 0.0014 0.0026
Dy 0.0108 0.0092 0.0100 0.0089 0.0084 0.0083 0.0078 0.0103 0.0086 0.0075 0.0085
Ho 0.0009 0.0011 0.0012 0.0010 0.0009 0.0011 0.0009 0.0013 0.0015 0.0012 0.0013
Er 0.0020 0.0023 0.0038 0.0029 0.0018 0.0019 0.0016 0.0020 0.0026 0.0015 0.0016
Tm 0.0003 0.0003 0.0017 0.0007 0.0001 0.0011 0.0003 0.0003 0.0009 0.0002 0.0002
Yb 0.0017 0.0014 0.0047 0.0027 0.0013 0.0021 0.0023 0.0007 0.0018 0.0017 0.0014
Lu 0.0022 0.0002 0.0004 0.0004 0.0002 0.0003 0.0002 0.0002 0.0002 0.0005 0.0002

4.3.2. Rare earth element systematics of selected garnets and
orthopyroxenes in garnet harzburgites
Notwithstanding the wide agreement that the key phase for deci-
phering the Si/Mg harzburgite issue is orthopyroxene, the trace element
and isotopic arguments have largely been made from analyses of garnet
(e.g. Griffin et al., 1999; Stachel et al., 2004; Gibson et al., 2013; Shu
and Brey, 2015). One main reason for focusing geochemical studies on
garnet is that it has much higher partition coefficients than orthopyr-
oxene or olivine for most trace elements (e.g. Sun and Liang, 2013),
meaning that even in highly depleted peridotites, in situ analytical
methods can quantify trace element concentrations. As shown here and
elsewhere (e.g. Canil, 1992; Canil, 2004; Gibson, 2017; Tomlinson
et al., 2018), many garnets found in highly depleted harzburgite formed
upon cooling by exsolution from originally more aluminous higher
temperature orthopyroxene that existed on the solidus. Therefore, there
is major petrological interest in the trace element geochemistry of or-
thopyroxene.
Of all measured trace elements in harzburgitic garnet, the REE are
most frequently cited as requiring metasomatism to explain the unusual
systematics. The key observation is that most analysed garnets do not
show the simple steep CI-normalised REE pattern expected from ex-
periments (e.g. Sun and Liang, 2013). Instead, at least half the studied
garnets, including many diamond inclusions (e.g. Stachel et al., 1998,
2004), show some form of sinusoidally-shaped CI-normalised REE
pattern with a ‘hump’ centred over the heavier LREE or the lightest
MREE but with exceedingly low La and Ce concentrations (e.g. Gibson
et al., 2013). This pattern is widely considered to imply metasomatic
addition of LREE.
Here we report full REE data (Table 2) for garnet and co-existing
orthopyroxene for the BP002 megacryst studied by Tomlinson et al.
(2018) and for the phase-mapped granular harzburgite 17BSK051. The
CI-normalised patterns for both phases (Fig. 13: garnet: panels a and b
and orthopyroxene: panels c and d) for both samples show the MREE-
enrichment. For the analyses least compromised by kimberlite con-
tamination and alteration (see also supplemental information), the
HREE concentrations are below 0.5 ppb, with those of Tm and Lu below
0.1 ppb. Despite the large scatter in individual orthopyroxene HREE
analyses (Fig. 13c), the pooled data from least contaminated analyses
result in a smooth average patterns (Fig. 13g and h) characterised by
near-symmetric parabola with an apex over Sm in Onuma-style dia-
grams. Both garnet and orthopyroxene patterns also show a secondary
upswing in the heaviest HREE. For BP002, garnet/orthopyroxene ap-
parent distribution coefficients (not shown) rise monotonically from 10
to 800 from La to Lu, as predicted from experimental data. In
17BSK051, garnet/orthopyroxene apparent distribution coefficients
(not shown) rise monotonically from 2 to 500 from La to Er and stay
around 500 for the remaining HREE. This could be an artefact of the
uncertainties on the HREE concentrations for the orthopyroxene but a
similar observation was also made by Wasch et al. (2009). Regardless,
in both studied samples, garnet and orthopyroxene have sinusoidal REE
patterns, strongly supporting the argument that garnet exsolved from
orthopyroxene. The curious MREE-enrichment was therefore inherited
from the precursor phase, implying that if metasomatism had occurred
it would have predated the exsolution of garnet (Tomlinson et al.,
2018). Alternatively, the relative enrichment in the MREE is unrelated
to metasomatism and inherent to high Si/Mg harzburgite formation.
5. Discussion
Two of the most distinguishing features of Archaean cratons are the
existence of highly refractory, strongly ultramafic harzburgite in the
mantle root and the widespread occurrence of ultramafic lavas at the
proto-cratonic surface. When combined, these mutually related rock
types could be thought of as ultramafic ‘sandwich’ layers of Archaean
cratons. In this final section, we discuss the importance of mafic-ul-
tramafic resurfacing for proto-cratonic crustal architecture and stabili-
sation and analyse the types of mantle source regions that could have a
formed the erupted lavas and residues found as xenoliths.
5.1. The role of mafic-ultramafic lids as conductive incubators
Several prominent papers have suggested that the A-P boundary did
not represent a relatively sudden change in style of geology but that a
more familiar geology began to emerge much earlier and more gradu-
ally, from ca. 3 Ga (e.g. Shirey and Richardson, 2011; Keller and
Schoene, 2012; Dhuime et al., 2012; Tang et al., 2016; Smit and
Mezger, 2017). Nearly all the supporting lines of evidence for this
proposal are based on interpretation of geochemical and/or radiogenic
isotope data. Several authors have argued that from ca. 3 Ga onwards,
plate tectonics (Tang et al., 2016) or Wilson cycles (Shirey and
Richardson, 2011) started operating. This geochemical view of a gra-
dual transition from the Mesoarchaean to the Palaeoproterozoic con-
trasts with the traditional geological evidence, which is in favour of a
much sharper end of the Neoarchaean. For example, Bédard (2018)
pointed out that there is no obvious difference in the geology of
Neoarchaean and Mesoarchaean or even older terranes, with the only
exception of the appearance of K-rich granitoids at ca. 3 Ga. Here we
propose that the seemingly incompatible lines of evidence can be re-
conciled by considering stratigraphic context, which is typically lost in
large time-series compilations.
This is illustrated with the example of the Abitibi greenstone belt,
which has the best defined laterally coherent stratigraphy. Similar
mafic-rock dominated cover sequences have been documented across
other cratons, including the Zimbabwe (Wilson et al., 1995), the
Eastern Dharwar (Jayananda et al., 2013), the East Pilbara (Hickman
and Van Kranendonk, 2012), the Northeastern Superior (e.g. Maurice
et al., 2009) and the Kalgoorlie terrane of the Yilgarn (Hayman et al.,
2015). As illustrated earlier (Fig. 9), the stratigraphy of the Abitibi

141
B.S. Kamber and E.L. Tomlinson Chemical Geology 511 (2019) 123–151

Fig. 13. CI-chondrite-normalised REE pat-

A 10 B terns (Anders and Grevesse, 1989) (panels
10 A–F) and Onuma-type diagrams (panels G
and H). Panels (A) and (C) show garnet and
CI-normalised

orthopyroxene data for megacryst BP002

CI-normalised
1 and (B) and (D) equivalent data for gran-
ular harzburgite 17BSK051. Green coloured
1 symbols denote analyses least affected by
0.1 contamination and/or alteration. Red co-
loured symbols show variably compromised
analyses. In BP002, intergranular (black in
transmitted light) contaminant (likely kim-
0.1 0.01
10 10 berlite) was also analysed and this was used
to calculate mixing models with average
C D most depleted garnet (panel E) and ortho-
pyroxene (panel F). Very good coincidences
1 1 with average contaminated analyses by
admixing between 0.1% and 0.12% con-
CI-normalised taminant. Panel (G) shows precursor phase
CI-normalised

(blue) REE concentrations calculated from

0.1 0.1
modal relative abundances of garnet (red)
and orthopyroxene (green) in an Onuma-
type diagram for BP002. Panel (G) shows
0.01 0.01 mean most depleted garnet (red) and or-
thopyroxene (green) of 17BSK051 in an
Onuma-type diagram.

0.001 0.001

100 F
100 E
10
CI-normalised
CI-normalised

10 1

0.1
1 contaminant
depleted garnet
0.01
contaminated garnet
mixing model (0.1%)
0.1 0.001
La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu

10
G 10
H
CI-normalised

CI-normalised

1 1

0.1 0.1

0.01 0.01

0.001 0.001
1.15 1.1 1.05 1 1.15 1.1 1.05 1

Ionic radius ( ) Ionic radius ( )

greenstone belt is made up of 7 lower assemblages dominated by vol-
canic rocks and two upper assemblages of sediment-dominated rocks
deposited into successor basins (e.g. Ayer et al., 2002). Studies seeking
to infer crustal evolution from the chemistry of sediment rely on faithful
representativeness of the preserved sedimentary rocks, reflecting the
full stratigraphy of the supracrustal strata.
However, in the Abitibi greenstone stratigraphy, sedimentary rocks
of all types are only common in the uppermost two assemblages but
exceedingly sparse within the 7 lower assemblages. There they occur
between volcanic units as conformable condensed horizons (Thurston
et al., 2008) composed of thin iron formation, chert, exhalite and so-
called argillite, which is a very-fine grained sediment containing
reduced C. Of these, only argillite has an appreciable siliciclastic
component that could potentially be of use to the global geochemical
approach proposed by Keller and Schoene (2012), Tang et al. (2016)
and Smit and Mezger (2017). However, due to the precipitated authi-
genic component (Si, Fe) and the sequestration of transition metals
from the water column into argillites, it is impossible to compare their
chemistry with bona fide clastic metasediments. For example, the ar-
gillites of the Kidd-Munro and Kinojevis assemblages studied by Barrie
(2005) in the Holloway area, have high average Co/Th and Cr/Th ra-
tios, which would indicate a high mafic to ultramafic source component
(Fig. 2b) but their average Ni/Co ratio is only 2.7 ± 0.5, much lower
than coeval siliciclastic sediments (Fig. 2d) compiled by Tang et al.

142
B.S. Kamber and E.L. Tomlinson
(2016). This is not because the sediment source of the argillites is more
felsic but because of the higher concentration of Co than Ni in VMS-type
hydrothermal fluids. For many greenstone belts, the desired clastic se-
dimentary rocks sampling the main mafic-ultramafic magmatic units
simply do not exist and are therefore not represented in global datasets.
Instead, the datasets are dominated by and skewed towards the
sand-, silt- and mudstones deposited into the late successor basins.
These formed in rifts (e.g. Van Kranendonk et al., 2010), coevally with
emplacement of potassic magmas (e.g. Mueller et al., 1994) post-dating
major mafic and ultramafic volcanism. The chemistry of these sedi-
mentary rocks is thus not representative of the bulk of the greenstone
belt. In the Abitibi greenstone belt, Feng and Kerrich (1990) studied the
chemistry of rift-hosted Porcupine and Timiskaming assemblage sedi-
mentary rocks (the two youngest assemblages) and the Ni/Co of the
various subunits from their study is shown in Fig. 2d. The mean of all
units (4.9 ± 0.5) confirms the trend identified by Tang et al. (2016)
from global compilations (the mean of their 2.7 Ga data is 5.6 ± 0.35).
We propose that these sediment chemistries inform about the progress
of intra-crustal differentiation and cratonic stabilisation rather than
secular changes in crustal makeup or plate tectonics. Strong upward
vertical movement of the radioactive heat producing incompatible
elements K, U and Th in granitoid melts was accompanied by down-
ward foundering of supracrustal lithologies, including of Timiskaming
age (Lin et al., 2013). Because successor basins and their sediments only
started forming as the crust was gaining mechanical stability (e.g.
Bédard, 2018), there is an inherent bias towards preserving detritus
from these more evolved sources. This limitation does not narrow the
significance of the global analyses of sediment databases (e.g. Tang
et al., 2016; Smit and Mezger, 2017) but complicates the tectonic in-
terpretation of resulting temporal trends.
Our proposal is supported by studies that focus on the chemistry of
the magmatic rather than on sedimentary rocks. Moyen and Laurent's
(2017) analysis of the mafic igneous rock database concluded that
(p.116–117): “the result is surprisingly disappointing, as rocks
throughout the Archaean show exactly the same geochemical pattern,
from the oldest rocks known at Isua or Nuvvuagittuq to the latest Ar-
chaean e.g. in China or Zimbabwe – i.e. over about 1.5 Ga of geological
history”. With respect to the ultramafic rocks, the present analysis of
the MgO-distribution of Archaean volcanic rocks also lacks evidence for
a change in the chemistries of lavas that erupted between the Meso- and
Neoarchaean (Fig. 7). Furthermore, the maximum MgO of Archaean
komatiites does not show a secular trend either (e.g. Kamber, 2010) and
there is no evidence for a secular decrease of depletion extent within
Archaean harzburgite suites but there is a strong change after the A-P
boundary (e.g. Boyd, 1989; Griffin, 1999). We regard these as very
important observations, because mafic and ultramafic magmas have
potential to inform about changes in mantle melting regime.
By contrast, there is very strong evidence from many cratons (e.g.
Smithies et al., 2009; Laurent et al., 2014) for a temporal and strati-
graphic evolution of granitoid magmatism, expressed broadly as a trend
towards more potassic compositions and shallower emplacement (for
much more detail, see Moyen and Laurent, 2017). This trend is also
observed in the Abitibi greenstone belt (Fig. 9c) where the oldest plu-
tonic rocks are grey gneisses of diorite-tonalite composition, maturing
with time to granodiorites and biotite-bearing granodiorites (e.g. Benn,
2004), before reaching granitic compositions that typify shallow late
intrusions (e.g. Petrus et al., 2016). Laurent et al. (2014) interpreted the
first appearance of sanukitoids at ca. 3 Ga to mark the onset of proper
plate collision. However, the alternative explanation is that the evolu-
tion of felsic magmas is the reflection of progressive differentiation of
the crust (Bédard, 2018).
There are two main factors that drove crustal differentiation in the
Archaean. The first is the substantially higher radioactive heat pro-
duction within the crust and the second is the repeated emplacement of
thick supracrustal mafic-ultramafic lids (e.g. Van Kranendonk et al.,
2015; Wiemer et al., 2018). The resulting incubation accelerated
143
Chemical Geology 511 (2019) 123–151
intracrustal reorganisation (Sandiford et al., 2004), with a general
vertical rise of the incompatible heat producing elements in granitoids
and granites. Accordingly, the late intrusion of high-K, shallow lopo-
lithic intrusions (e.g. Frei et al., 1999) marks the mechanical stabili-
sation of the crust (e.g. Kamber, 2015; Bédard, 2018), completing the
process of cratonisation.
The idea of vertical foundering or sagging of denser and colder
greenstones between granitoid diapirs, also known as ‘dome and basin’
or ‘balloon’ tectonics (e.g. Ramsay, 1989), is not universally accepted
(e.g. Kimura et al., 1993). However, thermal and viscoelastic mechan-
ical modelling of inverted density profiles (Robin and Bailey, 2009;
Thebaud and Rey, 2013) shows that it remains a plausible mechanism.
Furthermore, field investigations (e.g. Blenkinsop et al., 1993; Ayer
et al., 2002; Johnson et al., 2016; Wiemer et al., 2016) on various
cratons have questioned the scepticism towards ‘dome and basin’ tec-
tonics that arose from the LITHOPROBE project (e.g. Clowes et al.,
1992). It is perhaps unfortunate that ‘dome and basin’ tectonics has
been termed ‘sagduction’ by some (e.g. Chardon et al., 1996) as this
intracrustal process (e.g. Bouhallier et al., 1995) is not an alternative to
subduction, which describes movements (that include a vertical com-
ponent) of one plate relative to another. Crustal reorganisation through
radioactive heat and conductive incubation is not unique to the Ar-
chaean (Sandiford and McLaren, 2002) but is much less common in the
post-Archaean geological record. Because radioactive heat production
decayed smoothly with time, it alone cannot explain the existence of
the A-P boundary, which more likely represents the combination of
several factors. Here we argue that after 2.5 Ga, the inability to com-
monly produce low viscosity high-Mg lavas as during the Archaean
resulted in the reduction of extent of mafic-ultramafic resurfacing, i.e.
the deposition of thinner ensialic greenstone blankets. It is difficult to
reliably estimate the original thickness of supracrustal blankets because
in most cases, the dense rocks have foundered and are now preserved as
highly deformed remnants at mid-crustal or deeper levels (e.g.
Bouhallier et al., 1995; Johnson et al., 2016) or could have been lost as
foundered eclogite (e.g. Johnson et al., 2014). Furthermore, greenstone
stratigraphies were not deposited with uniform thickness across the
proto-cratons. Variations in thickness of volcanic units may reflect
proximity to major volcanic centres, topography at the proto-cratonic
surface and syn-depositional deformation (e.g. Berger et al., 2011).
Observed stratigraphic thicknesses of rapidly emplaced volcanic units
(‘groups’ or ‘assemblages’) range from 3 to 5 km in the Eastern Yilgarn
(Hayman et al., 2015) and the Eastern Dharwar craton (Jayananda
et al., 2013) to ca. 10 km in the East Pilbara terrane (Hickman and Van
Kranendonk, 2012) and Abitibi greenstone belt (Berger et al., 2011).
Full greenstone belt thicknesses are much greater, reaching up to
20–25 km (Ayer et al., 2002; Van Kranendonk et al., 2007). In numer-
ical models, deposition of 5–15 km thick incubating lids of mafic-ul-
tramafic cover sequences drives the observed crustal reorganisation.
We argue that the markedly reduced post-Archaean conductive in-
cubation together with reduced crustal heat production contributed to
the observed change in geological style. In post-Archaean terranes that
did experience strong basaltic to picritic volcanic resurfacing (e.g. the
Palaeoproterozoic Flin Flon (Canada) and the Ashanti (West Africa)
greenstone belts; Perrouty et al., 2012), the crustal architecture has an
Archaean flavour but komatiite is largely absent and the prevailing
theory is that these Palaeoproterozoic terranes formed via arc-accretion
(e.g. Lucas et al., 1996).
5.2. Melting of fertile, highly refractory or metasomatised mantle?
Turning from the ultramafic rocks at surface to those in the litho-
spheric root, the various possible petrological models that could have
left many cratonic peridotites with high enstatite/olivine (i.e. high Si/
Mg) ratio have been reviewed by Herzberg (1993). At the time, the
preferred explanations ranged from melting residues of a non-pyrolitic
mantle, metamorphic segregation (e.g. Boyd, 1989) to cumulate
B.S. Kamber and E.L. Tomlinson
crystallisation from a high-SiO2 magma ocean. One alternative model,
metasomatic addition of SiO2 in subduction zones (Kesson and
Ringwood, 1989) was less popular on account of the high fluid/rock
ratio required (e.g. Canil, 1992) and the mismatch of observed with
predicted CaO, Al2O3 and SiO2 systematics (Herzberg, 1993). With the
discovery of the sinusoidal REE patterns in cratonic harzburgite garnets
(Stachel et al., 1998) the pendulum swung in favour of metasomatic
addition of LREE and SiO2 (e.g. Simon et al., 2003, 2007; Bernstein
et al., 2007; Pearson and Wittig, 2008; Wasch et al., 2009). However, a
comprehensive survey of the O-isotope composition of variably Si-en-
riched cratonic peridotites has documented a uniform mantle parentage
(Regier et al., 2018), strongly arguing against the subduction zone
metasomatic model.
An aspect common to most of the quoted studies is that they com-
pare cratonic harzburgite olivine modes and forsterite contents with the
empirical trend from depleted modern oceanic peridotites (Fig. 3c)
originally proposed by Boyd (1989). However, Boyd's trend was never
scrutinised with experimental data. Comparison with the experimental
melting residues of Walter (1998) shows a strong pressure effect
(Fig. 3d). None of the isobaric experimental residues reproduce the
(hand-drawn?) concave trend of Boyd (1989). At pressures of 5 GPa or
higher, the experimental olivines are highly magnesian and exist in a
residue containing 25–35% modal combined pyroxene and garnet
(Walter, 1998). It is recognised that this comparison is limited to one
experimental dataset because most other relevant high-pressure studies
did not tabulate estimates for modal abundance. We note, however,
that in terms of phase relationships and stability fields, Walter's (1998)
experiments are compatible with those conducted at lower pressure and
those that did not report full modal abundances (e.g. Takahashi, 1986;
Canil, 1991, 1992; Takahashi et al., 1993; Baker and Stolper, 1994;
Longhi, 2002; Schutt and Lesher, 2006). There is thus no indication that
the modes from these experiments would be very different. We agree
with Canil (2008) that the mismatch between observed and expected
cratonic harzburgite mineralogy is strongest in the Kaapvaal craton but
harzburgites with < 60% modal olivine also exist from the Slave, Si-
berian and Wyoming cratons (Eggler et al., 1987; Ionov et al., 2010;
Kopylova et al., 1999; Kopylova and Caro, 2004; Newton et al., 2015;
Solovjeva et al., 1994). The relatively low modal abundance of olivine
could mainly indicate that melt extraction occurred at high pressure but
further in-depth analysis will be needed for a firmer conclusion. In
terms of the variability in olivine mode at a given forsterite content
(Fig. 3f), we note again the extremely coarse-grained nature of cratonic
harzburgites (e.g. Figs. 11 and 12) and propose that some of the modal
variability could stem from minor mineral segregation and from un-
representativeness of thin sections.
One aspect of the cratonic Si/Mg issue that has not yet been ex-
plored experimentally is the possibility that high-Mg harzburgites could
be residues of multiple melt extraction events. Arndt (1977a) com-
mented (p. 205) that “ultrabasic magmas probably form by a sequential
melting process and are derived from a residuum composed of re-
fractory minerals and trapped liquid left by previous episodes of partial
melting and magma extraction”. Our analysis of the experimental li-
quids of Walter (1998) shows that re-melting of a previously depleted
source can produce a much wider range of Si/Mg in the liquids than
single batch melting (Fig. 10g and h) as well as more extreme Fe/Mg
and Al/Mg variability as seen in some komatiites. More experimental
data on refractory starting compositions and modelling of melt-residue
interaction are certainly needed to better understand the Si/Mg sys-
tematics of komatiites and harzburgites (e.g. Herzberg, 2016), parti-
cularly in light of the new O-isotope data (Regier et al., 2018) that
argue against subduction zone metasomatism.
There are strong indications from the geochemistry of komatiites
and komatiitic basalts that variably fertile sources were being melted
(e.g. Sproule et al., 2002). Although the source fertility of komatiites is
conventionally determined via Al (Nesbitt et al., 1979), experimental
data show that the Ti concentration may be a simpler measure to assess
144
Chemical Geology 511 (2019) 123–151
whether a source had been previously depleted (see also Sproule et al.,
2002; Robin-Popieul et al., 2012). This is because Ti remains in-
compatible across the entire melting depth interval relevant to cratonic
lithosphere (Longhi, 2002). The archetypes of melts from ultra-depleted
sources are from Commondale (Wilson, 2003) but komatiites with
Al2O3/TiO2 > 25 have also been reported from localities outside the
Kaapvaal craton (e.g. Sproule et al., 2002). This supports the proposal
of Arndt (1977a) and suggests that additional experimental constraints
are required for harzburgitic starting compositions. The low-Ti nature
of many komatiites is shared (to some extent) by modern boninites but
this common trait is, in our view, insufficient to justify the subduction
zone origin proposed by Parman et al. (1997).
In terms of REE systematics, one key question of relevance to proto-
craton formation is whether the sinusoidal REE pattern arose from
metasomatism post-dating the original melt-extraction (e.g. Stachel
et al., 2004) or whether it is an inherent feature of cratonic mantle
related to its formation by high degree melting (e.g. Tomlinson et al.,
2018). There is no doubt that cratonic lithosphere has experienced
multiple episodes of metasomatic infiltration, particularly in the base
(e.g. Nixon et al., 1987; Foley, 2008) but there is very strong evidence
that the sinusoidal REE pattern is ancient and not related to Neoarch-
aean metasomatism. The following are the strongest lines of evidence.
Firstly, the Hf-isotope composition of garnets with sinusoidal REE
patterns are extreme, reaching to spectacular present-day ε(Hf) of up to
+1000 (Shu and Brey, 2015) with correspondingly ancient Archaean
model ages. Shu et al. (2013) discovered that garnet separates (with
sinusoidal REE patterns) from Kaapvaal craton kimberlites yield Lu/Hf
regression lines of 3.2 Ga, suggesting that the sinusoidal REE pattern
has to be at least as old. This age is substantially older than the
Neoarchaean cratonisation. Secondly, Koornneef et al. (2017) suc-
ceeded in dating diamond inclusion garnet populations with strongly
sinusoidal REE patterns to 2.95 Ga with 147Sm/143Nd. The majority of
these garnets have very low Sm/Nd ratios, consistent with the hump in
the REE pattern over Nd and Sm. Thirdly, as shown by earlier studies
(Zhang et al., 2001; Simon et al., 2007; Wasch et al., 2009) and the new
data on the BP002 megacryst (Tomlinson et al., 2018; this study), the
REE of orthopyroxene also show a humped pattern, implying that the
entire protolith peridotite carried this signature. Finally, Stachel et al.
(2006) reported on diamond inclusions hosted in 2.7 Ga volcaniclastic
rocks and lamprophyre dykes in the Wawa greenstone belt. The inclu-
sion suite contains garnets very similar to those in common cratonic
harzburgite and the nitrogen aggregation of the diamonds suggests
residence times of tens to hundreds of Ma prior to crustal emplacement.
Collectively, this evidence suggests that harzburgite with sinusoidal
REE patterns existed in the Meso- and Neoarchaean cratonic mantle
root.
This finding is relevant to komatiite geochemistry in two respects.
Firstly, garnet and orthopyroxene are the main carriers of the REE in
advanced melt residues between 4 and ca. 8 GPa. Depending on the
magnitude of the hump over the heavier LREE (Nd-Gd), the Sm/Yb or
Gd/Yb ratio of a melt in equilibrium with these phases will either be
slightly sub- or supra-chondritic. The important conclusion is that if the
komatiite source had a sinusoidal REE pattern, there will only be lim-
ited variability in Sm/Yb or Gd/Yb ratios in liquids. This is one possible
solution to the paradox identified by Walter (1998), namely that major
elements of many komatiites require a garnet-bearing source but REE
do not show the expected large Sm/Yb fractionation. As the new ex-
amples of garnet from exsolved harzburgite orthopyroxene (Tomlinson
et al., 2018; this study Figs. 11 and 12) show, melt extraction at > 3
GPa is not necessarily in conflict with the high Cr/Al ratios of residues,
provided they were originally bi-mineralic olivine-orthopyroxene rocks.
Secondly, a source with sinusoidal REE patterns has the potential to
evolve to diverging Nd and Hf isotope compositions. Because the apex
of the hump in the REE pattern is frequently around Sm or Nd, the
resulting Sm/Nd does no differ much from chondrite and Nd-isotope
ratios will evolve to mildly supra- or even sub-chondritic values. By
B.S. Kamber and E.L. Tomlinson
contrast, the Lu/Hf ratio is mostly strongly superchondritic and Hf-
isotope ratios rapidly evolve to positive ε(Hf) values. We propose that
this could explain the Nd-Hf isotope systematics of the Commondale
komatiites (Hoffmann and Wilson, 2017) and the decoupled Nd-Hf ar-
rays of other Kaapvaal craton komatiites (Blichert-Toft et al., 2015).
It is beyond the scope of this paper to fully discuss the various
proposals for the origin of the sinusoidal REE pattern and its possible
relation to Si enrichment (e.g. Burgess and Harte, 2004). If the sinu-
soidal REE pattern can be attributed to metasomatism with (e.g. Gibson
et al., 2013) or without previous melt depletion it has to be Archaean in
age and, in some cases, clearly predates craton stabilisation. Just as
with the Si/Mg systematics, there remain significant gaps in experi-
mental knowledge for REE behaviour in peridotite at the conditions
relevant to cratonic melting. For example, pseudo-sections for un-
metasomatised fertile peridotite predict that the phase assemblage at
the 5 GPa solidus is orthopyroxene-free, consisting only of olivine,
clinopyroxene and garnet (e.g. Fig. 1 of Jennings and Holland, 2015).
Thus, due to the expansion of the stability field of pigeonitic clin-
opyroxene above 1500 °C and pressures > 4.2 GPa (Takahashi, 1986),
the peridotite that is melting is technically called a garnet wherlite, not
a lherzolite. To our knowledge, no relevant experimental data exist for
the REE partitioning of this type of pyroxene. Until the necessary ex-
periments have been conducted, it will remain unclear if a sinusoidal
REE patterns could result in residues of multiply melted normal peri-
dotite without metasomatic REE addition.
5.3. Possible causes for the demise of very high-temperature mantle
upwellings and the relevance for the Archaean-Proterozoic boundary
There remains much uncertainty regarding the A-P boundary. In our
view, the geological evidence in favour of a boundary outweighs the
geochemical indications for earlier (e.g. 3 Ga) planetary changes in
tectonic style. If we accept the geochronological confirmation of the
original postulation of the A-P boundary in the 19th century (Logan,
1857) as more than a fortuitous coincidence, the boundary must have
major geological significance.
Most ideas seeking to explain the A-P boundary advocate a re-
organisation of the mantle temperature structure and/or heat loss me-
chanism. Two schools of thought are most prominent. Firstly, a change
within or above the core-mantle thermal boundary layer may have
occurred and this affected the type of thermal upwellings that travelled
through the mantle. For example, Campbell and Griffiths (2014) at-
tributed the A-P boundary to piling subducted oceanic plates into D″,
forming an insulating layer from which modern-style plumes began
emerging since the Palaeoproterozoic. Likewise, Bédard (2018) en-
visaged mantle ‘overturn upwelling zones’ that episodically transported
heat and material from the lower mantle towards the surface.
The second group of ideas has identified the mantle transition zone
as a possible location where major change may have occurred at the A-P
boundary. In these models, the base of the transition zone was a much
stronger thermal boundary layer in the Archaean than thereafter. The
hot mantle upwellings that caused komatiite magmatism could have
originated at this boundary layer. Davies (2008) proposed that
throughout the Archaean, subducted slabs piled up in the transition
zone, forming a mechanical barrier that impeded heat loss from the
lower mantle. According to his model, episodic collapse of the barrier
occurred every 100–150 Ma and manifested as pulses of high tem-
perature melting events in the upper mantle. From the A-P boundary
onwards, slabs began to penetrate the lower mantle and the transition
zone barrier stopped building up. Rino et al. (2004) advocated a similar
model and they regarded the greater thickness of the subducted Ar-
chaean oceanic crust as key to the neutral buoyancy of slabs in the
transition zone. Breuer and Spohn (1995) argued that the A-P boundary
marked the change from 2-layer to single layer mantle convection and
envisaged a pronounced flush instability that caused a global
Neoarchaean igneous ‘super event’.
145
Chemical Geology 511 (2019) 123–151
In our view, all these models contain attractive aspects to explain
the A-P boundary and it is noted that the Archaean pile-up of slabs in
the transition zone is not mutually exclusive with post-Archaean slab
pile-up in D″. However, none of the current models has addressed the
wider evidence from the crustal and mantle portions of cratonic litho-
sphere. We regard the following four key observations as most relevant
for such a model:
i) All cratons either still have or have formerly had deep mantle roots
with geotherms corresponding to an average present-day heat flow
of 41 mW m−2 and strong concentration of radioactive heat pro-
duction within the uppermost crust (e.g. Michaut et al., 2009).
Importantly, relatively cool cratonic mantle geotherms apparently
also existed in the Archaean (Boyd et al., 1985). In the Kaapvaal
craton and elsewhere, there is evidence that the Neoarchaean and
early Proterozoic geotherms were surprisingly cool, only mildly
more elevated than today's (around 45–50 mW m−2; e.g. Nimis,
2002). This requires that in the Archaean, conductive heat loss
extended to greater depth than the base of the current cratonic roots
and/or that the original cratons were larger in extent than the
fragments preserved today (Ballard and Pollack, 1988). This re-
quirement is illustrated on Fig. 14. It is widely agreed that most
cratons show evidence for post-Archaean base erosion (e.g. Foley,
2008; Xu and Qiu, 2017) supporting the notion that they were
originally deeper. Rare xenoliths from the Kaapvaal craton contain
majoritic garnet relics from 300 to 400 km depth (Haggerty and
Sautter, 1990), deeper than the present-day base of the lithosphere.
Regardless of preferred proto-craton thickness, strongly elevated
lower crustal temperatures are unavoidable (Fig. 14).
ii) Throughout the Archaean, there is no evidence for secular change
in komatiite Mg-content (Kamber, 2010; Campbell and Griffiths,
2014) and mafic magmas in general have remained chemically
unchanged (e.g. Moyen and Laurent, 2017). But there was a sharp
drop in MgO content of ultramafic magmas in the Palaeoproter-
ozoic. By contrast, basaltic magmas have persisted through geolo-
gical time with very similar mean MgO content as far back as the
Palaeoarchaean (Kamber, 2015 and this study).
iii) Archaean crustal rocks, and particularly komatiites and associated
fractionated basalts, preserve evidence for very ancient element
fractionation of Sm/Nd, Hf/W and Sm/Lu (e.g. Caro et al., 2003;
Touboul et al., 2012). The resulting 142Nd and 182W anomalies and
the 143Nd/176Hf divergence are more pronounced in Palaeo- and
Mesoarchaean rocks (e.g. Bennett et al., 2007; Rizo et al., 2011) but
still found in Neoarchaean samples (e.g. Debaille et al., 2013;
Puchtel et al., 2016). It follows that the mantle domains harbouring
such anomalies must have escaped convective mixing until the
Neoarchaean and earliest Palaeoproterozoic.
iv) Although in many cratons the crust and the mantle portion ex-
perienced coeval events (melting, depletion, reworking, re-enrich-
ment, etc.) the two reservoirs are not complementary. The pre-
served cratonic mantle root alone cannot have supplied all the
incompatible trace elements present in the crust (e.g. Carlson et al.,
2005). Therefore, crust formation must have depleted greater vo-
lumes of mantle that never were or no longer are spatially asso-
ciated with the craton.
One possible solution to the last two observations could be that
proto-cratonic mantle formed in multiple ways. This could include in-
itial shallow melt extraction from asthenosphere that only produced a
relatively thin refractory root, maybe due to effective dripping of
crustal restite residues (e.g. Bédard, 2006; Rollinson, 2010; Johnson
et al., 2014). There is no a priori necessity that these proto-cratons grew
from unusually hot mantle. Indeed, the tholeiitic lavas on many cratons
could have been sourced from relatively shallow mantle (e.g. Wyman
et al., 2002; Smithies et al., 2009). By contrast, the rocks that now
constitute the deep cratonic mantle root may have originated in an
B.S. Kamber and E.L. Tomlinson Chemical Geology 511 (2019) 123–151

0 Fig. 14. Geotherms calculated by Ballard and Pollack (1988) for the Kaapvaal
craton. Panels (A) and (B) depict central cratonic geotherms from models
‘Present 2’ (P2) and ‘Archaean 2’ (A2), respectively. (A) In model P2, the re-
100 duction in central cratonic heat flow to 40 mW m−2 was achieved with a li-
thospheric root of 220 km depth and a reduced internal lithospheric heat pro-
duction. (B) Model A2 shows the same 400 km diameter craton but at 3.2 Ga
with 2.5× higher internal heat production and the mantle potential tempera-
200
ture 200 °C hotter than at present (stippled line is model P2 for reference). The
Depth (km)

higher internal heat production results in a Moho temperature 300 °C hotter

than in the present-day model and the higher mantle potential temperature
300 elevates the mantle root geothermal overall. To achieve a similar geotherm as
P2 with the higher internal heat production would require dropping the 3.2 Ga
mantle potential temperature 200 °C relative to present day. Panel (C) shows an
400 alternative model for the Archaean (A2′), in which the craton is 1.5× larger in
diameter than A2 and the root extents to 500 km instead of 220 km. This results
in similar temperatures as present-day at depths of 120–180 km, the interval
500 where most studied xenoliths derive from but maintains a high Moho tem-
perature.
A
600
0 400 800 1200 1600 2000 unrelated part of the mantle – here termed ancient non-convecting
o
Temperature ( C) domains (ANCD).
0 One plausible way of producing ANCD early in Earth history is
crystallisation and crystal segregation in magma seas (e.g. Arndt et al.,
2002; Hansen, 2015). The geotherm after crystallisation of a global
100 magma ocean is very different from the adiabat and almost certainly
leads to instability (e.g. Elkins Tanton et al., 2003; Kramers, 2007) that
could have yielded domains of ANCD. An additional possibility is that
200 giant impacts caused more local, shallower magma seas and their
crystallisation products (cumulates) may have persisted (e.g.
Depth (km)

Hofmeister, 1983) as ANCDs surrounded by asthenosphere. Alter-

300 natively, stratified crystallisation products of magma seas can experi-
ence gravitational instability and may founder (e.g. Elkins Tanton et al.,
2002) and could survive as non-pyrolitic ANCDs at depth. We note that
400 cumulate formation during fractional crystallisation from a magma
ocean has been suggested as one possible cause for the short-lived
radiogenic isotope anomalies recorded in Archaean rocks (Caro et al.,
500 2005). Finally, very deep mantle melting from lower mantle upwellings
could have formed refractory mantle zones not originally associated
B with a proto-craton.
600
0 400 800 1200 1600 2000
In Fig. 15, we illustrate that if ANCDs occupied laterally extensive
o areas in the mantle transition zone, they could have impeded heat
Temperature ( C)
0 transfer from the lower mantle. Heat would eventually escape, either
through partly eroded ANCDs or along their margins as upwellings that
triggered high temperature melting in the asthenosphere. Nascent
100 continental lithosphere that drifted over areas covered by ANCDs at
depth would be mostly protected from the extensive resurfacing that
affected most of the remaining Earth. The nascent continental plates
200 could have been similar to the ribbon continents proposed by Bédard
(2018). Alternatively, they could have formed as arc-like collages (e.g.
Wyman et al., 2002) that occasionally experienced emplacement of
Depth (km)

300 much hotter and deeply sourced magmas through the ANCD.
With time, the accumulated effects of thermal upwellings weakened
the ANCDs to the point of collapse, marking the A-P boundary.
400 Remnants of the ANCDs were mostly recycled back into the astheno-
sphere but locally, became incorporated into the cratonic roots. The
accompanying decompression and cooling could explain mineral reac-
500 tion textures (including exsolutions) in garnet harzburgites and pyr-
oxenites (e.g. Haggerty and Sautter, 1990). The dismembered deep
C parts of ANCDs would have carried with them transition zone minerals
600 that could survive as diamond inclusions. The majority of transition
0 400 800 1200 1600 2000 zone garnet-structured diamond inclusions are of mixed peridotitic/
Temperature (oC) pyroxenitic chemistry (Kiseeva et al., 2013) and only a minority has an
eclogitic parentage. If these inclusions are samples from the deep
ANCDs, their chemistry favours the idea that the barriers were mainly

146
B.S. Kamber and E.L. Tomlinson
B and evolution of cratonic lithosphere from the
A Mesoarchaean, ancient non-convecting do-
Continental crust Asthenosphere
Shallow harzburgite Magma sea cumulates
Deep residue Magma ocean cumulates
composed of pyroxenite progenitors rather than subducted oceanic
crust but it is possible that slab pile-up as envisaged by Rino et al.
(2004) and Davies (2008) added to the mechanical and thermal barrier.
The initial Hf-isotope composition of Kaapvaal craton harzburgite
garnets presented by Shu et al. (2013) are radiogenic and require that
the precursor material separated from chondritic mantle at least a few
hundred million years before 3.2 Ga. This constraint on timing is con-
sistent with the idea that ANCDs could have formed in giant impact
basins (e.g. Hansen, 2015) during the late bombardment (Bottke et al.,
2012).
A model in which cratonic lithosphere grew from above and was
fused with deep remnants could also explain why one class of harz-
burgites (those without high modal abundance orthopyroxene) appears
to have mainly foundered after melt extraction (Canil, 2004; Lee and
Chin, 2014) whereas others, like BP002, appear to have moved upwards
(Haggerty and Sautter, 1990; Tomlinson et al., 2018). We also note that
some cratons may have more thoroughly depleted lower roots and more
lherzolitic shallower portions (e.g. Griffin et al., 2003). This observa-
tion could be explained by re-enrichment of formerly more strongly
depleted harzburgite with fluids from depth (e.g. Gibson et al., 2013).
Alternatively, the working model advanced here would envisage the
less depleted zone to represent the area of fusion of shallower with
deeper residues. It corresponds in depth both with the impedance in-
crease found by Green and Hales (1968) and a 50 km thick zone of
pronounced anisotropy with horizontal azimuths that could represent
frozen-in drift (Sodoudi et al., 2013) of the shallow nascent continental
plates.
6. Concluding remarks
Regardless of the validity of competing models for craton formation,
the general difficulties of sourcing sufficient highly incompatible ele-
ments for cratonic crust (e.g. Bédard, 2006; Michaut et al., 2009) and
the requirement for cool temperatures in the Archaean roots (e.g.
Nimis, 2002) remain. With respect to the former, Kramers and
Tolstikhin (1997) pointed out that the formation of a global Hadean
protocrust could have led to an initial overall depletion of the entire
mantle. When the Hadean crust was later comprehensively destroyed,
the Eoarchaean asthenosphere could have ended up with an elevated
inventory of incompatible elements if crustal recycling was restricted to
147
Chemical Geology 511 (2019) 123–151
Fig. 15. Schematic model illustrating growth
Mesoarchaean (A) and the latest Neoarchaean
C (B) into the post-Archaean (C). In the
mains (ANCD) existed in the upper mantle as
remnants of cumulates that crystallised from
magma oceans and seas and from very deep
melting in hot upwellings. The ANCDs are not
envisaged to have been a global barrier but
where they were extensive, heat flow from the
lower mantle was impeded. A nascent craton
with a modest mantle root is shown forming
above an ANCD. Occasional upwelling of heat
began to erode through the ANCD and lower-
mantle derived melts may have metasomatised
the cumulates. By the late Neoarchaean, the
NACDs had been dislodged by repeated pulses
of upwellings form the lower mantle. The ma-
jority of the barrier was recycled back into the
asthenosphere but occasional fragments and
melt residues from the NACD became entrained
into the Neoarchaean ‘superplumes’ which
fused this material with the shallower proto-
cratonic root. In post-Archaean events, the base
of the lithosphere began to erode and shallow.
a depth of 670 km. Melting of such enriched asthenosphere to form
Archaean cratons could go a long way towards alleviating the mass
imbalance between cratonic root and crust for highly incompatible
elements. Furthermore, depending on the mass and composition of the
Hadean crust, the upper mantle may also have ended up with a slightly
elevated Si/Mg ratio with higher modal abundance of orthopyroxene.
Concerning the requirement for surprisingly cool temperatures in
the Archaean cratonic roots, it is worth noting that very limited
knowledge exists about the original dimension of Archaean cratons (e.g.
Bleeker, 2003). The model of Ballard and Pollack (1988) assumed re-
latively modest diameters of proto-cratons (400–600 km) informed by
the then prevailing notion that cratons contained small ancient nuclei
onto which additional lithosphere accreted over time (e.g. Kusky,
1998). However, even in the Zimbabwe craton, where the evidence for
a small original nucleus was once thought strongest, newer zircon U/Pb
dates (e.g. Horstwood et al., 1999; Bolhar et al., 2017) and Re-depletion
constrains on the cratonic root (Smith et al., 2009) now suggest that the
craton had attained its full preserved size from the beginning. Fur-
thermore, detailed comparison of crustal stratigraphies, radiogenic
isotope characteristics, dyke swarm analysis and palaeomagnetic data
imply that the Zimbabwe craton could have been part of the same
craton as the Yilgarn (Smirnov et al., 2013). There is the additional
possibility that this Zimgarn craton could itself only be a fragment of an
even larger original craton (e.g. Bleeker, 2003) composed of additional
cratons (including the Slave, Wyoming and North China) that all share
the so-called high μ Pb-isotope signature (Kamber et al., 2003; Kamber,
2015). In terms of Archaean geodynamics, this opens the possibility
that craton formation and preservation may not have been a ubiquitous
process but that only two or three large early cratons with thick roots
may have formed in environments that favoured their preservation and
that they bias our understanding of how early terrestrial geology op-
erated.
Supplementary data to this article can be found online at https://
doi.org/10.1016/j.chemgeo.2019.02.011.
Acknowledgments
This contribution started as a keynote lecture at the 2017
Goldschmidt meeting in the session “Evolution of the Continental Crust
and Mantle Lithosphere” and we would like to thank the convenors
B.S. Kamber and E.L. Tomlinson
Roberta Rudnick, Nick Arndt and Sarah Brownlee for the invitation. We
thank Review Papers Coordinator Tim Horscroft for the invitation to
write this paper, Klaus Mezger for editorial handling and Nick Arndt
and Dante Canil for their thoughtful reviews. Our sincere gratitude goes
to the following individuals: Sergei Lebedev for interesting discussions
on craton formation; Mike Lesher for giving access to the Abitibi
Subprovince komatiite database and discussions on ultramafic mag-
matism; John Ayer and Phil Thurston for advice on greenstone belt
stratigraphy; Jock Robey for facilitating collection of Kimberley peri-
dotites; Paul Guyett for SEM-EDX analysis of garnet harzburgite; Mags
Duncan for helping with data trawling and compilations; Elaine Cullen
for re-drafting figures from cited publications; and Claire Kamber for
drafting Figs. 9 and 15. The geochemistry facility at TCD is supported
by grant SFI/RI/3227.
References
Anders, E., Grevesse, N., 1989. Abundances of elements: meteoritic and solar. Geochim.
Cosmochim. Acta 53, 197–214.
Arndt, N.T., 1977a. Ultrabasic magmas and high-degree melting of the mantle. Contrib.
Mineral. Petrol. 64 (2), 205–221.
Arndt, N.T., 1977b. Partitioning of nickel between olivine and ultrabasic and basic ko-
matiite liquids. In: Carnegie Institution of Washington Yearbook. 76. pp. 553–557.
Arndt, N.T., 1991. High Ni in Archean tholeiites. Tectonophysics 187 (4), 411–419.
Arndt, N.T., Lewin, É., Albarède, F., 2002. Strange partners: formation and survival of
continental crust and lithospheric mantle. Geol. Soc. Lond., Spec. Publ. 199 (1),
91–103.
Arndt, N., Lesher, C., Barnes, S., 2008. Komatiite. Cambridge University Press, Cambridge
(458 p.).
Ashwal, L., Morgan, P., Kelley, S., Percival, J., 1987. Heat production in an Archean
crustal profile and implications for heat flow and mobilization of heat-producing
elements. Earth Planet. Sci. Lett. 85 (4), 439–450.
Aulbach, S., Griffin, W.L., Pearson, N.J., O'Reilly, S.Y., Kivi, K., Doyle, B.J., 2004. Mantle
formation and evolution, Slave Craton: constraints from HSE abundances and Re–Os
isotope systematics of sulfide inclusions in mantle xenocrysts. Chem. Geol. 208 (1),
61–88.
Aulbach, S., Stachel, T., Heaman, L.M., Creaser, R.A., Shirey, S.B., 2011. Formation of
cratonic subcontinental lithospheric mantle and complementary komatiite from hy-
brid plume sources. Contrib. Mineral. Petrol. 161 (6), 947–960.
Ayer, J., Amelin, Y., Corfu, F., Kamo, S., Ketchum, J., Kwok, K., Trowell, N., 2002.
Evolution of the southern Abitibi greenstone belt based on U-Pb geochronology:
autochthonous volcanic construction followed by plutonism, regional deformation
and sedimentation. Precambrian Res. 115 (1–4), 63–95.
Baker, M.B., Stolper, E.M., 1994. Determining the composition of high-pressure mantle
melts using diamond aggregates. Geochim. Cosmochim. Acta 58 (13), 2811–2827.
Ballard, S., Pollack, H.N., 1988. Modern and ancient geotherms beneath southern Africa.
Earth Planet. Sci. Lett. 88 (1–2), 132–142.
Barrie, C., 2005. Geochemistry of exhalites and graphitic argillites near VMS and gold
deposits. In: Ontario Geological Survey Miscellaneous Release Data. 173. pp. 126.
Bédard, J.H., 2006. A catalytic delamination-driven model for coupled genesis of
Archaean crust and sub-continental lithospheric mantle. Geochim. Cosmochim. Acta
70 (5), 1188–1214.
Bédard, J.H., 2018. Stagnant lids and mantle overturns: implications for Archaean tec-
tonics, magmagenesis, crustal growth, mantle evolution, and the start of plate tec-
tonics. Geosci. Front. 9 (1), 19–49.
Begg, G.C., Griffin, W.L., Natapov, L.M., O'Reilly, S.Y., Grand, S.P., O'Neill, C.J., Hronsky,
J.M.A., Djomani, Y.P., Swain, C.J., Deen, T., Bowden, P., 2009. The lithospheric ar-
chitecture of Africa: seismic tomography, mantle petrology, and tectonic evolution.
Geosphere 5 (1), 23–50.
Benn, K., 2004. Late Archaean Kenogamissi complex, Abitibi Subprovince, Ontario,
Canada: doming, folding and deformation-assisted melt remobilisation during syn-
tectonic batholith emplacement. Trans. R. Soc. Edinb. Earth Sci. 95, 297–307.
Benn, K., Kamber, B.S., 2009. In situ U/Pb granulite-hosted zircon dates, Kapuskasing
Structural Zone, Ontario: a late Archean Large Igneous Province (LIP) as a substrate
for juvenile crust. J. Geol. 117 (5), 519–541.
Bennett, V.C., Brandon, A.D., Nutman, A.P., 2007. Coupled Nd-142-Nd-143 isotopic
evidence for Hadean mantle dynamics. Science 318 (5858), 1907–1910.
Berger, B., Bleeker, W., van Breemen, O., Chapman, J., Peter, J., Layton-Matthews, D.,
Gemmell, J., 2011. Results from the Targeted Geoscience Initiative III Kidd–Munro
Project. In: Ontario Geological Survey, Open File Report, pp. 6258.
Bernstein, S., Hanghøj, K., Kelemen, P.B., Brooks, C.K., 2006. Ultra-depleted, shallow
cratonic mantle beneath West Greenland: dunitic xenoliths from Ubekendt Ejland.
Contrib. Mineral. Petrol. 152 (3), 335.
Bernstein, S., Kelemen, P.B., Hanghøj, K., 2007. Consistent olivine Mg# in cratonic
mantle reflects Archean mantle melting to the exhaustion of orthopyroxene. Geology
35 (5), 459–462.
Berry, A.J., Danyushevsky, L.V., O'Neill, H.S.C., Newville, M., Sutton, S.R., 2008.
Oxidation state of iron in komatiitic melt inclusions indicates hot Archaean mantle.
Nature 455 (7215), 960–U42.
Bindeman, I., Zakharov, D., Palandri, J., Greber, N.D., Dauphas, N., Retallack, G.,
148
Chemical Geology 511 (2019) 123–151
Hofmann, A., Lackey, J., Bekker, A., 2018. Rapid emergence of subaerial landmasses
and onset of a modern hydrologic cycle 2.5 billion years ago. Nature 557 (7706), 545.
Bizzarro, M., Stevenson, R.K., 2003. Major element composition of the lithospheric
mantle under the North Atlantic craton: evidence from peridotite xenoliths of the
Sarfartoq area, southwestern Greenland. Contrib. Mineral. Petrol. 146 (2), 223–240.
Bleeker, W., 2003. The late Archean record: a puzzle in ca. 35 pieces. Lithos 71 (2–4),
99–134.
Blenkinsop, T.G., Fedo, C.M., Eriksson, K.A., Martin, A., Nisbet, G.E., Wilson, J.F., 1993.
Ensialic origin for the Ngezi Group, Belingwe greenstone belt, Zimbabwe. Geology
21, 1135–1138.
Blichert-Toft, J., Arndt, N.T., Wilson, A., Coetzee, G., 2015. Hf and Nd isotope systematics
of early Archean komatiites from surface sampling and ICDP drilling in the Barberton
Greenstone Belt, South Africa. Am. Mineral. 100 (11−12), 2396–2411.
Bolhar, R., Hofmann, A., Kemp, A.I., Whitehouse, M.J., Wind, S., Kamber, B.S., 2017.
Juvenile crust formation in the Zimbabwe Craton deduced from the O-Hf isotopic
record of 3.8–3.1 Ga detrital zircons. Geochim. Cosmochim. Acta 215, 432–446.
Bottke, W.F., Vokrouhlicky, D., Minton, D., Nesvorny, D., Morbidelli, A., Brasser, R.,
Simonson, B., Levison, H.F., 2012. An Archaean heavy bombardment from a desta-
bilized extension of the asteroid belt. Nature 485 (7396), 78–81.
Bouhallier, H., Chardon, D., Choukroune, P., 1995. Strain patterns in Archean dome-and-
baisan structures – the Dharwar craton (Karnataka, South-India). Earth Planet. Sci.
Lett. 135 (1–4), 57–75.
Boyd, F.R., 1989. Compositional distinction between oceanic and cratonic lithosphere.
Earth Planet. Sci. Lett. 96 (1–2), 15–26.
Boyd, F., Gurney, J., Richardson, S., 1985. Evidence for a 150–200-km thick Archaean
lithosphere from diamond inclusion thermobarometry. Nature 315 (6018), 387.
Boyd, F., Pokhilenko, N., Pearson, D., Mertzman, S., Sobolev, N., Finger, L., 1997.
Composition of the Siberian cratonic mantle: evidence from Udachnaya peridotite
xenoliths. Contrib. Mineral. Petrol. 128 (2–3), 228–246.
Breuer, D., Spohn, T., 1995. Possible flush instability in mantle convection at the
Archaean-Proterozoic transition. Nature 378, 608–610.
Brey, G., Köhler, T., 1990. Geothermobarometry in four-phase lherzolites II. New ther-
mobarometers, and practical assessment of existing thermobarometers. J. Petrol. 31
(6), 1353–1378.
Burgess, S.R., Harte, B., 2004. Tracing lithosphere evolution through the analysis of
heterogeneous G9–G10 garnets in peridotite xenoliths, II: REE chemistry. J. Petrol. 45
(3), 609–633.
Campbell, I.H., Griffiths, R.W., 1992. The changing nature of mantle hotspots through
time - implications for the chemical evolution of the mantle. J. Geol. 100 (5),
497–523.
Campbell, I.H., Griffiths, R.W., 2014. Did the formation of D″ cause the Archaean-
Proterozoic transition? Earth Planet. Sci. Lett. 388, 1–8.
Canil, D., 1991. Experimental evidence for the exsolution of cratonic peridotite from high-
temperature harzburgite. Earth Planet. Sci. Lett. 106 (1–4), 64–72.
Canil, D., 1992. Ortho-pyroxene stability along the peridotite solidus and the origin of
cratonic lithosphere beneath southern Africa. Earth Planet. Sci. Lett. 111 (1), 83–95.
Canil, D., 2004. Mildly incompatible elements in peridotites and the origins of mantle
lithosphere. Lithos 77 (1–4), 375–393.
Canil, D., 2008. Canada's craton: a bottoms-up view. GSA Today 18 (6), 4.
Canil, D., Lee, C.-T.A., 2009. Were deep cratonic mantle roots hydrated in Archean
oceans? Geology 37 (7), 667–670.
Carlson, R.W., Pearson, D.G., James, D.E., 2005. Physical, chemical, and chronological
characteristics of continental mantle. Rev. Geophys. 43 (1).
Caro, G., Bourdon, B., Birck, J.L., Moorbath, S., 2003. Sm-146-Nd-142 evidence from Isua
metamorphosed sediments for early differentiation of the Earth's mantle. Nature 423,
428–432.
Caro, G., Bourdon, B., Wood, B.J., Corgne, A., 2005. Trace-element fractionation in
Hadean mantle generated by melt segregation from a magma ocean. Nature 436
(7048), 246–249.
Chardon, D., Choukroune, P., Jayananda, M., 1996. Strain patterns, decollement and
incipient sagducted greenstone terrains in the Archaean Dharwar craton (south
India). J. Struct. Geol. 18 (8), 991.
Chauvel, C., Dupré, B., Jenner, G.A., 1985. The Sm-Nd age of Kambalda volcanics is
500 Ma too old! Earth and planet. Sci. Lett. 74, 315–324.
Chauvel, C., Dupré, B., Arndt, N.T., 1993. Pb and Nd isotopic correlation in Belingwe
komatiites and basalts. In: Bickle, M.J., Nisbet, E.G. (Eds.), The Geology of the
Belingwe Greenstone Belt, Zimbabwe. Geol. Soc. Zimbabwe Spec. Publ., Rotterdam,
pp. 167–174.
Chayes, F., 1963. Relative abundance of intermediate members of the oceanic basalt-
trachyte association. J. Geophys. Res. 68 (5), 1519–1534.
Clowes, R.M., Cook, F.A., Green, A.G., Keen, C.E., Ludden, J.N., Percival, J.A., Quinlan,
G.M., West, G.F., 1992. LITHOPROBE – new perspectives on crustal evolution. Can. J.
Earth Sci. 29 (9), 1813–1864.
Condie, K.C., 1993. Chemical composition and evolution of the upper continental crust:
contrasting results from surface samples and shales. Chem. Geol. 104, 1–37.
Corfu, F., 1993. The evolution of the southern Abitibi greenstone belt in light of precise U-
Pb geochronology. Econ. Geol. Bull. Soc. Econ. Geol. 88 (6), 1323–1340.
Cox, K., Smith, M., Beswetherick, S., 1987. Textural studies of garnet lherzolites: evidence
of exsolution origin from high-temperature harzburgites. In: Mantle Xenoliths, pp.
537–550.
Davies, G.F., 2008. Episodic layering of the early mantle by the ‘basalt barrier’ me-
chanism. Earth Planet. Sci. Lett. 275 (3–4), 382–392.
Debaille, V., O'Neill, C., Brandon, A.D., Haenecour, P., Yin, Q.Z., Mattielli, N., Treiman,
A.H., 2013. Stagnant-lid tectonics in early Earth revealed by Nd-142 variations in late
Archean rocks. Earth Planet. Sci. Lett. 373, 83–92.
Dhuime, B., Hawkesworth, C.J., Cawood, P.A., Storey, C.D., 2012. A change in the
B.S. Kamber and E.L. Tomlinson
geodynamics of continental growth 3 billion years ago. Science 335 (6074),
1334–1336.
Donovan, J.J., Lowers, H.A., Rusk, B.G., 2011. Improved electron probe microanalysis of
trace elements in quartz. Am. Mineral. 96 (2–3), 274–282.
Eggler, D.H., McCallum, M., Kirkley, M., 1987. Kimberlite-transported nodules from
Colorado-Wyoming; a record of enrichment of shallow portions of an infertile li-
thosphere. Geol. Soc. Am. Spec. Pap. 215, 77–90.
Elkins Tanton, L.T., Van Orman, J.A., Hager, B.H., Grove, T.L., 2002. Re-examination of
the lunar magma ocean cumulate overturn hypothesis: melting or mixing is required.
Earth Planet. Sci. Lett. 196 (3–4), 239–249.
Elkins Tanton, L.T., Parmentier, E., Hess, P., 2003. Magma ocean fractional crystallization
and cumulate overturn in terrestrial planets: implications for Mars. Meteorit. Planet.
Sci. 38 (12), 1753–1771.
Feng, R., Kerrich, R., 1990. Geochemistry of fine-grained clastic sediments in the Archean
Abitibi greenstone belt, Canada: implications for provenance and tectonic setting.
Geochim. Cosmochim. Acta 54 (4), 1061–1081.
Foley, S.F., 2008. Rejuvenation and erosion of the cratonic lithosphere. Nat. Geosci. 1 (8),
503.
Frei, R., Blenkinsop, T.G., Schönberg, R., 1999. Geochronology of the late Archaean Razi
and Chilimanzi suites of granites in Zimbabwe: implications for the late Archaean
tectonics of the Limpopo Belt and Zimbabwe Craton. S. Afr. J. Geol. 102 (1), 55–63.
Galer, S.J.G., Mezger, K., 1998. Metamorphism, denudation and sea level in the Archean
and cooling of the Earth. Precambrian Res. 92 (4), 389–412.
Gibson, S.A., 2017. On the nature and origin of garnet in highly-refractory Archean li-
thospheric mantle: constraints from garnet exsolved in Kaapvaal craton orthopyr-
oxenes. Mineral. Mag. 81 (4), 781–809.
Gibson, S.A., McMahon, S.C., Day, J.A., Dawson, J.B., 2013. Highly Refractory
Lithospheric Mantle beneath the Tanzanian Craton: evidence from Lashaine Pre-
metasomatic Garnet-bearing Peridotites. J. Petrol. 54 (8), 1503–1546.
Grand, S.P., 1994. Mantle shear structure beneath the Americas and surrounding oceans.
J. Geophys. Res. Solid Earth 99 (B6), 11591–11621.
Green, R., Hales, A., 1968. The travel times of P waves to 30° in the central United States
and upper mantle structure. Bull. Seismol. Soc. Am. 58 (1), 267–289.
Grégoire, M., Bell, D., Le Roex, A., 2003. Garnet lherzolites from the Kaapvaal Craton
(South Africa): trace element evidence for a metasomatic history. J. Petrol. 44 (4),
629–657.
Griffin, W., 1999. The composition and origin of subcontinental lithospheric mantle. In:
Mantle Petrology: Field Observation and High Pressure Experimentation: A Tribute to
Francis R. (Joe) Boyd, pp. 13–45.
Griffin, W.L., Fisher, N.I., Friedman, J., Ryan, C.G., O'Reilly, S.Y., 1999. Cr-pyrope garnets
in the lithospheric mantle. I. Compositional systematics and relations to tectonic
setting. J. Petrol. 40 (5), 679–704.
Griffin, W.L., Fisher, N.I., Friedman, J.H., O'Reilly, S.Y., Ryan, C.G., 2002. Cr-pyrope
garnets in the lithospheric mantle 2. Compositional populations and their distribution
in time and space. Geochem. Geophys. Geosyst. 3, 35.
Griffin, W.L., O'Reilly, S.Y., Abe, N., Aulbach, S., Davies, R.M., Pearson, N.J., Doyle, B.J.,
Kivi, K., 2003. The origin and evolution of Archean lithospheric mantle. Precambrian
Res. 127 (1–3), 19–41.
Haggerty, S.E., Sautter, V., 1990. Ultradeep (greater than 300 kilometers), ultramafic
upper mantle xenoliths. Science 248 (4958), 993–996.
Hales, A., 1969. A seismic discontinuity in the lithosphere. Earth Planet. Sci. Lett. 7 (1),
44–46.
Hansen, V.L., 2015. Impact origin of Archean cratons. Lithosphere 7 (5), 563–578.
Harley, S.L., 1984. An experimental study of the partitioning of Fe and Mg between garnet
and orthopyroxene. Contrib. Mineral. Petrol. 86 (4), 359–373.
Hayman, P.C., Thébaud, N., Pawley, M.J., Barnes, S.J., Cas, R.A., Amelin, Y., Sapkota, J.,
Squire, R.J., Campbell, I.H., Pegg, I., 2015. Evolution of a ∼2.7 Ga large igneous
province: a volcanological, geochemical and geochronological study of the Agnew
Greenstone Belt, and new regional correlations for the Kalgoorlie Terrane (Yilgarn
Craton, Western Australia). Precambrian Res. 270, 334–368.
Herzberg, C.T., 1993. Lithosphere peridotites of the Kaapvaal craton. Earth Planet. Sci.
Lett. 120 (1–2), 13–29.
Herzberg, C., 2016. Petrological evidence from komatiites for an early Earth carbon and
water cycle. J. Petrol. 57 (11–12), 2271–2288.
Herzberg, C., Feigenson, M., Skuba, C., Ohtani, E., 1988. Majorite fractionation recorded
in the geochemistry of peridotites from South Africa. Nature 332 (6167), 823–826.
Hickman, A.H., Van Kranendonk, M.J., 2012. Early Earth evolution: evidence from the
3.5–1.8 Ga geological history of the Pilbara region of Western Australia. Episodes 35
(1), 283–297.
Hoffmann, J.E., Wilson, A.H., 2017. The origin of highly radiogenic Hf isotope compo-
sitions in 3.33 Ga Commondale komatiite lavas (South Africa). Chem. Geol. 455,
6–21.
Hofmeister, A.M., 1983. Effect of a Hadean terrestrial magma ocean on crust and mantle
evolution. J. Geophys. Res. Solid Earth 88 (B6), 4963–4983.
Horstwood, M.S.A., Nesbitt, R.W., Noble, S.R., Wilson, J.F., 1999. U-Pb zircon evidence
for an extensive early Archean craton in Zimbabwe: a reassessment of the timing of
craton formation, stabilization, and growth. Geology 27 (8), 707–710.
Houle, M.G., Prefontaine, S., Fowler, A.D., Gibson, H.L., 2009. Endogenous growth in
channelized komatiite lava flows: evidence from spinifex-textured sills at Pyke Hill
and Serpentine Mountain, Western Abitibi Greenstone Belt, Northeastern Ontario,
Canada. Bull. Volcanol. 71 (8), 881–901.
Ionov, D.A., Doucet, L.S., Ashchepkov, I.V., 2010. Composition of the lithospheric mantle
in the Siberian craton: new constraints from fresh peridotites in the Udachnaya-East
kimberlite. J. Petrol. 51 (11), 2177–2210.
Jayananda, M., Peucat, J.-J., Chardon, D., Rao, B.K., Fanning, C., Corfu, F., 2013.
Neoarchean greenstone volcanism and continental growth, Dharwar craton, southern
149
Chemical Geology 511 (2019) 123–151
India: constraints from SIMS U–Pb zircon geochronology and Nd isotopes.
Precambrian Res. 227, 55–76.
Jennings, E.S., Holland, T.J., 2015. A simple thermodynamic model for melting of peri-
dotite in the system NCFMASOCr. J. Petrol. 56 (5), 869–892.
Jochum, K.P., Hofmann, A.W., Ito, E., Seufert, H.M., White, W.M., 1983. K, U and Th in
mid-ocean ridge baslat glasses amd heat production, K/U and K/Rb in the mantle.
Nature 306 (5942), 431–436.
Johnson, T.E., Brown, M., Kaus, B.J.P., VanTongeren, J.A., 2014. Delamination and re-
cycling of Archaean crust caused by gravitational instabilities. Nat. Geosci. 7 (1),
47–52.
Johnson, T.E., Brown, M., Goodenough, K.M., Clark, C., Kinny, P.D., White, R.W., 2016.
Subduction or sagduction? Ambiguity in constraining the origin of ultramafic–mafic
bodies in the Archean crust of NW Scotland. Precambrian Res. 283, 89–105.
Jordan, T.H., 1975. The continental tectosphere. Rev. Geophys. 13 (3), 1–12.
Jordan, T.H., 1978. Composition and development of the continental tectosphere. Nature
274 (5671), 544.
Kamber, B.S., 2010. Archean mafic-ultramafic volcanic landmasses and their effect on
ocean atmosphere chemistry. Chem. Geol. 274 (1–2), 19–28.
Kamber, B.S., 2015. The evolving nature of terrestrial crust from the Hadean, through the
Archaean, into the Proterozoic. Precambrian Res. 258, 48–82.
Kamber, B.S., Collerson, K.D., Moorbath, S., Whitehouse, M.J., 2003. Inheritance of early
Archaean Pb-isotope variability from long-lived Hadean protocrust. Contrib. Mineral.
Petrol. 145, 25–46.
Kaminsky, F.V., Sablukov, S.M., Sablukova, L.I., Shchukin, V.S., Canil, D., 2002.
Kimberlites from the Wawa area, Ontario. Can. J. Earth Sci. 39 (12), 1819–1838.
Keller, C.B., Schoene, B., 2012. Statistical geochemistry reveals disruption in secular li-
thospheric evolution about 2.5 Gyr ago. Nature 485 (7399), 490–U100.
Kesson, S.E., Ringwood, A.E., 1989. Slab mantle interactions. 2. The formation of dia-
monds. Chem. Geol. 78 (2), 97–118.
Ketchum, J.W.F., Ayer, J.A., Van Breemen, O., Pearson, N.J., Becker, J.K., 2008.
Pericontinental crustal growth of the Southwestern Abitibi Subprovince, Canada-U-
Pb, Hf, and Nd isotope evidence. Econ. Geol. 103 (6), 1151–1184.
Kimura, G., Ludden, J.N., Desrochers, J.P., Hori, R., 1993. A model of ocean-crust ac-
cretion for the Superior province, Canada. Lithos 30, 337–355.
Kiseeva, E.S., Yaxley, G.M., Stepanov, A.S., Tkalčić, H., Litasov, K.D., Kamenetsky, V.S.,
2013. Metapyroxenite in the mantle transition zone revealed from majorite inclusions
in diamonds. Geology 41 (8), 883–886.
Klemme, S., 2004. The influence of Cr on the garnet–spinel transition in the Earth's
mantle: experiments in the system MgO–Cr2O3–SiO2 and thermodynamic modelling.
Lithos 77 (1–4), 639–646.
Konhauser, K.O., Pecoits, E., Lalonde, S.V., Papineau, D., Nisbet, E.G., Barley, M.E., Arndt,
N.T., Zahnle, K., Kamber, B.S., 2009. Oceanic nickel depletion and a methanogen
famine before the Great Oxidation Event. Nature 458 (7239), 750–U85.
Koornneef, J.M., Gress, M.U., Chinn, I.L., Jelsma, H.A., Harris, J.W., Davies, G.R., 2017.
Archaean and Proterozoic diamond growth from contrasting styles of large-scale
magmatism. Nat. Commun. 8 (1), 648.
Kopylova, M., Caro, G., 2004. Mantle xenoliths from the southeastern Slave craton: evi-
dence for chemical zonation in a thick, cold lithosphere. J. Petrol. 45 (5), 1045–1067.
Kopylova, M., Russell, J., Cookenboo, H., 1999. Petrology of peridotite and pyroxenite
xenoliths from the Jericho kimberlite: implications for the thermal state of the mantle
beneath the Slave craton, northern Canada. J. Petrol. 40 (1), 79–104.
Kramers, J.D., 2007. Hierarchical Earth accretion and the Hadean Eon. J. Geol. Soc. Lond.
164, 3–17.
Kramers, J.D., Tolstikhin, I.N., 1997. Two terrestrial lead isotope paradoxes, forward
transport modelling, core formation and the history of the continental crust. Chem.
Geol. 139, 75–110.
Kramers, J.D., Kreissig, K., Jones, M.Q.W., 2001. Crustal heat production and style of
metamorphism: a comparison between two Archaean high grade provinces in the
Limpopo Belt, southern Africa. Precambrian Res. 112, 149–163.
Kushiro, I., 1973. Partial melting of garnet lherzolites from kimberlite at high pressures.
In: Nixon, P.H. (Ed.), Lesotho Kimberlites. Cape and Transvaal Printers Ltd, Cape
Town, pp. 249–299.
Kusky, T.M., 1998. Tectonic setting and terrane accretion of the Archean Zimbabwe
craton. Geology 26 (2), 163–166.
Laurent, O., Martin, H., Moyen, J.-F., Doucelance, R., 2014. The diversity and evolution of
late-Archean granitoids: evidence for the onset of “modern-style” plate tectonics
between 3.0 and 2.5 Ga. Lithos 205, 208–235.
Lebedev, S., Boonen, J., Trampert, J., 2009. Seismic structure of Precambrian lithosphere:
new constraints from broad-band surface-wave dispersion. Lithos 109 (1), 96–111.
Lee, C.-T.A., Chin, E.J., 2014. Calculating melting temperatures and pressures of peri-
dotite protoliths: implications for the origin of cratonic mantle. Earth Planet. Sci. Lett.
403, 273–286.
Lesher, C., Barnes, S., 2009. Komatiite-associated Ni–Cu–(PGE) deposits. In: New
Developments in Magmatic Ni–Cu and PGE Deposits. 290. Geological Publishing
House, Beijing, pp. 27–120.
Lin, S., Parks, J., Heaman, L.M., Simonetti, A., Corkery, M.T., 2013. Diapirism and sag-
duction as a mechanism for deposition and burial of “Timiskaming-type” sedimentary
sequences, Superior Province: evidence from detrital zircon geochronology and im-
plications for the Borden Lake conglomerate in the exposed middle to lower crust in
the Kapuskasing uplift. Precambrian Res. 238, 148–157.
Logan, W.E., 1857. On the division of Azoic rocks of Canada into Huronian and
Laurentian. Proc. Am. Assoc. Adv. Sci. 1857, 44–47.
Longhi, J., 2002. Some phase equilibrium systematics of lherzolite melting: I. Geochem.
Geophys. Geosyst. 3 (3), 1–33.
Lucas, S., Stern, R., Syme, E., Reilly, B., Thomas, D., 1996. Intraoceanic tectonics and the
development of continental crust: 1.92–1.84 Ga evolution of the Flin Flon Belt,
B.S. Kamber and E.L. Tomlinson
Canada. Geol. Soc. Am. Bull. 108 (5), 602–629.
Maurice, C., David, J., Bédard, J.H., Francis, D., 2009. Evidence for a widespread mafic
cover sequence and its implications for continental growth in the Northeastern
Superior Province. Precambrian Res. 168 (1), 45–65.
Michaut, C., Jaupart, C., Mareschal, J.C., 2009. Thermal evolution of cratonic roots.
Lithos 109 (1–2), 47–60.
Miller, C.E., Kopylova, M.G., Ryder, J., 2012. Vanished diamondiferous cratonic root
beneath the Southern Superior province: evidence from diamond inclusions in the
Wawa metaconglomerate. Contrib. Mineral. Petrol. 164 (4), 697–714.
Moyen, J.-F., Laurent, O., 2017. Archaean tectonic systems: a view from igneous rocks.
Lithos 302–302, 99–125.
Mueller, W., Donaldson, J.A., Doucet, P., 1994. Volcanic and tectono-plutonic influences
and sedimentation in the Archean Kirkland basin, Abitibi greenstone belt, Canada.
Precambrian Res. 68 (3–4), 201–230.
Naegler, T.F., Kramers, J.D., Kamber, B.S., Frei, R., Prendergast, M.D.A., 1997. Growth of
subcontinental lithospheric mantle beneath Zimbabwe started at or before 3.8 Ga: Re-
Os study on chromites. Geology 25 (11), 983–986.
Nesbitt, R., Sun, S.-S., Purvis, A., 1979. Komatiites; geochemistry and genesis. Can.
Mineral. 17 (2), 165–186.
Newton, D.E., Kopylova, M.G., Burgess, J., Strand, P., 2015. Peridotite and pyroxenite
xenoliths from the Muskox kimberlite, northern Slave craton, Canada. Can. J. Earth
Sci. 53 (1), 41–58.
Nimis, P., 2002. The pressures and temperatures of formation of diamond based on
thermobarometry of chromian diopside inclusions. Can. Mineral. 40 (3), 871–884.
Nixon, P., Van Calsteren, P., Boyd, F., Hawkesworth, C., 1987. Harzburgites with garnets
of diamond facies from southern African kimberlites. In: Mantle Xenoliths. John
Wiley, Chichester, pp. 523–533.
Parman, S.W., Dann, J.C., Grove, T.L., de Wit, M.J., 1997. Emplacement conditions of
komatiite magmas from the 3.49 Ga Komati Formation, Barberton Greenstone Belt,
South Africa. Earth Planet Sci. Lett. 150, 303–323.
Pearson, D., Wittig, N., 2008. Formation of Archaean continental lithosphere and its
diamonds: the root of the problem. J. Geol. Soc. 165 (5), 895–914.
Pearson, D.G., Carlson, R.W., Shirley, S.B., Boyd, F.R., Nixon, P.H., 1995. Stabilisation of
Archaean lithospheric mantle: a Re-Os isotope study of peridotite xenoliths from the
Kaapvaal craton. Earth Planet. Sci. Lett. 134, 341–357.
Pearson, D., Shirey, S., Bulanova, G., Carlson, R., Milledge, H., 1999. Re-Os isotope
measurements of single sulfide inclusions in a Siberian diamond and its nitrogen
aggregation systematics. Geochim. Cosmochim. Acta 63 (5), 703–711.
Perrouty, S., Aillères, L., Jessell, M.W., Baratoux, L., Bourassa, Y., Crawford, B., 2012.
Revised Eburnean geodynamic evolution of the gold-rich southern Ashanti Belt,
Ghana, with new field and geophysical evidence of pre-Tarkwaian deformations.
Precambrian Res. 204, 12–39.
Petrus, J., Kenny, G., Ayer, J., Lightfoot, P., Kamber, B., 2016. Uranium–lead zircon
systematics in the Sudbury impact crater-fill: implications for target lithologies and
crater evolution. J. Geol. Soc. 173 (1), 59–75.
Puchtel, I.S., Touboul, M., Blichert-Toft, J., Walker, R.J., Brandon, A.D., Nicklas, R.W.,
Kulikov, V.S., Samsonov, A.V., 2016. Lithophile and siderophile element systematics
of earth's mantle at the Archean-Proterozoic boundary: evidence from 2.4 Ga ko-
matiites. Geochim. Cosmochim. Acta 180, 227–255.
Ramsay, J.G., 1989. Emplacement kinematics of a granite diapir: the Chindamora bath-
olith, Zimbabwe. J. Struct. Geol. 11 (1–2), 191–209.
Regier, M., Miškovic, A., Ickert, R.B., Pearson, D.G., Stachel, T., Stern, R.A., Kopylova, M.,
2018. An oxygen isotope test for the origin of Archean mantle roots. Geochem.
Perspect. Lett. 9, 6–10.
Reubi, O., Blundy, J., 2009. A dearth of intermediate melts at subduction zone volcanoes
and the petrogenesis of arc andesites. Nature 461 (7268), 1269–U103.
Revenaugh, J., Jordan, T.H., 1991. Mantle layering from ScS reverberations: 3 the Upper
Mantle. J. Geophys. Res. Solid Earth 96 (B12), 19781–19810.
Rino, S., Komiya, T., Windley, B.F., Katayama, I., Motoki, A., Hirata, T., 2004. Major
episodic increases of continental crustal growth determined from zircon ages of river
sands; implications for mantle overturns in the Early Precambrian. Phys. Earth
Planet. Inter. 146 (1–2), 369–394.
Rizo, H., Boyet, M., Blichert-Toft, J., Rosing, M., 2011. Combined Nd and Hf isotope
evidence for deep-seated source of Isua lavas. Earth Planet. Sci. Lett. 312 (3–4),
267–279.
Robin, C.M.I., Bailey, R.C., 2009. Simultaneous generation of Archean crust and sub-
cratonic roots by vertical tectonics. Geology 37 (6), 523–526.
Robin-Popieul, C.C., Arndt, N.T., Chauvel, C., Byerly, G.R., Sobolev, A.V., Wilson, A.,
2012. A new model for Barberton komatiites: deep critical melting with high melt
retention. J. Petrol. 53 (11), 2191–2229.
Rollinson, H., 2010. Coupled evolution of Archean continental crust and subcontinental
lithospheric mantle. Geology 38 (12), 1083–1086.
Rudnick, R., 1994. Northern Tanzanian peridotite xenolith: a comparison with Kaapvaal
peridotites and evidence for carbonatite interaction with ultra-refractory residues. In:
Proc. 5th Int'l Kimberlite Conference. CPRM, pp. 336–353.
Rudnick, R.L., Gao, S., 2003. Composition of the continental crust. In: Treatise on
Geochemistry. 3. pp. 1–64.
Saltzer, R.L., Chatterjee, N., Grove, T.L., 2001. The spatial distribution of garnets and
pyroxenes in mantle peridotites: pressure-temperature history of peridotites from the
Kaapvaal craton. J. Petrol. 42 (12), 2215–2229.
Sandiford, M., McLaren, S., 2002. Tectonic feedback and the ordering of heat producing
elements within the continental lithosphere. Earth Planet. Sci. Lett. 204, 133–150.
Sandiford, M., Van Kranendonk, M.J., Bodorkos, S., 2004. Conductive incubation and the
origin of dome-and-keel structure in Archean granite-greenstone terrains: a model
based on the eastern Pilbara Craton, Western Australia. Tectonics 23 (1).
Schaeffer, A., Lebedev, S., 2013. Global shear speed structure of the upper mantle and
150
Chemical Geology 511 (2019) 123–151
transition zone. Geophys. J. Int. 194 (1), 417–449.
Schmeling, H., Arndt, N., 2017. Modelling komatiitic melt accumulation and segregation
in the transition zone. Earth Planet. Sci. Lett. 472, 95–106.
Schutt, D., Lesher, C., 2006. Effects of melt depletion on the density and seismic velocity
of garnet and spinel lherzolite. J. Geophys. Res. Solid Earth 111 (B5).
Shirey, S.B., Richardson, S.H., 2011. Start of the Wilson cycle at 3 Ga shown by diamonds
from subcontinental mantle. Science 333 (6041), 434–436.
Shu, Q., Brey, G.P., 2015. Ancient mantle metasomatism recorded in subcalcic garnet
xenocrysts: temporal links between mantle metasomatism, diamond growth and
crustal tectonomagmatism. Earth Planet. Sci. Lett. 418, 27–39.
Shu, Q., Brey, G.P., Gerdes, A., Hoefer, H.E., 2013. Geochronological and geochemical
constraints on the formation and evolution of the mantle underneath the Kaapvaal
craton: Lu–Hf and Sm–Nd systematics of subcalcic garnets from highly depleted
peridotites. Geochim. Cosmochim. Acta 113, 1–20.
Silverman, B.W., 1981. Using kernel density estimates to investigate multimodality. J. R.
Stat. Soc. Ser. B Methodol. 97–99.
Simon, N.S., Irvine, G.J., Davies, G.R., Pearson, D.G., Carlson, R.W., 2003. The origin of
garnet and clinopyroxene in “depleted” Kaapvaal peridotites. Lithos 71 (2–4),
289–322.
Simon, N.S., Carlson, R.W., Pearson, D.G., Davies, G.R., 2007. The origin and evolution of
the Kaapvaal cratonic lithospheric mantle. J. Petrol. 48 (3), 589–625.
Smirnov, A.V., Evans, D.A., Ernst, R.E., Söderlund, U., Li, Z.-X., 2013. Trading partners:
tectonic ancestry of southern Africa and western Australia, in Archean supercratons
Vaalbara and Zimgarn. Precambrian Res. 224, 11–22.
Smit, M.A., Mezger, K., 2017. Earth's early O2 cycle suppressed by primitive continents.
Nat. Geosci. 10 (10), 788.
Smith, C.B., Pearson, D.G., Bulanova, G.P., Beard, A.D., Carlson, R.W., Wittig, N., Sims,
K., Chimuka, L., Muchemwa, E., 2009. Extremely depleted lithospheric mantle and
diamonds beneath the southern Zimbabwe Craton. Lithos 112, 1120–1132.
Smithies, R.H., Champion, D.C., Van Kranendonk, M.J., 2009. Formation of Paleoarchean
continental crust through infracrustal melting of enriched basalt. Earth Planet. Sci.
Lett. 281 (3–4), 298–306.
Sobolev, A.V., Asafov, E.V., Gurenko, A.A., Arndt, N.T., Batanova, V.G., Portnyagin, M.V.,
Garbe-Schönberg, D., Krasheninnikov, S.P., 2016. Komatiites reveal a hydrous
Archaean deep-mantle reservoir. Nature 531 (7596), 628–632.
Sodoudi, F., Yuan, X.H., Kind, R., Lebedev, S., Adam, J.M.C., Kastle, E., Tilmann, F., 2013.
Seismic evidence for stratification in composition and anisotropic fabric within the
thick lithosphere of Kalahari Craton. Geochem. Geophys. Geosyst. 14 (12),
5393–5412.
Solovjeva, L., Vladimirov, B., Dneprovskaya, L., Maslovskaya, M.N., Brandt, S.B., 1994.
Kimberlites and Kimberlite-like Rocks: The Upper Mantle Beneath the Ancient
Platforms. Nauka, Novosibirsk (256 pp.).
Spencer, C.J., et al., 2014. Proterozoic onset of crustal reworking and collisional tectonics:
reappraisal of the zircon oxygen isotope record. Geology 42 (5), 451–454.
Sproule, R.A., Lesher, C.M., Ayer, J.A., Thurston, P.C., Herzberg, C.T., 2002. Spatial and
temporal variations in the geochemistry of komatiites and komatiitic basalts in the
Abitibi greenstone belts. Precambrian Res. 115, 153–186.
Stachel, T., Viljoen, K.S., Brey, G., Harris, J.W., 1998. Metasomatic processes in lherzolitic
and harzburgitic domains of diamondiferous lithospheric mantle: REE in garnets from
xenoliths and inclusions in diamonds. Earth Planet. Sci. Lett. 159 (1–2), 1–12.
Stachel, T., Aulbach, S., Brey, G.P., Harris, J.W., Leost, I., Tappert, R., Viljoen, K.S., 2004.
The trace element composition of silicate inclusions in diamonds: a review. Lithos 77
(1–4), 1–19.
Stachel, T., Banas, A., Muehlenbachs, K., Kurszlaukis, S., Walker, E.C., 2006. Archean
diamonds from Wawa (Canada): samples from deep cratonic roots predating crato-
nization of the Superior Province. Contrib. Mineral. Petrol. 151 (6), 737.
Sun, C.G., Liang, Y., 2013. The importance of crystal chemistry on REE partitioning be-
tween mantle minerals (garnet, clinopyroxene, orthopyroxene, and olivine) and ba-
saltic melts. Chem. Geol. 358, 23–36.
Sun, S.-S., McDonough, W.F., 1989. Chemical and isotopic systematics of oceanic basalts:
implications for mantle composition and processes. In: Saunders, A.D., Norry, M.J.
(Eds.), Magmatism in Ocean Basins. Geol. Soc. Spec. Publ., London, pp. 313–345.
Takahashi, E., 1986. Melting of a dry peridotite KLB-1 up to 14 GPa: implications on the
origin of peridotitic upper mantle. J. Geophys. Res. Solid Earth 91 (B9), 9367–9382.
Takahashi, E., Scarfe, C.M., 1985. Melting of peridotite to 14 GPa and the genesis of
komatiite. Nature 315 (6020), 566–568.
Takahashi, E., Shimazaki, T., Tsuzaki, Y., Yoshida, H., 1993. Melting study of a peridotite
KLB-1 to 6.5 GPa, and the origin of basaltic magmas. Philos. Trans. R. Soc. Lond. A
Math. Phys. Eng. Sci. 342 (1663), 105–120.
Tang, M., Chen, K., Rudnick, R.L., 2016. Archean upper crust transition from mafic to
felsic marks the onset of plate tectonics. Science 351 (6271), 372–375.
Thebaud, N., Rey, P.F., 2013. Archean gravity-driven tectonics on hot and flooded con-
tinents: controls on long-lived mineralised hydrothermal systems away from con-
tinental margins. Precambrian Res. 229, 93–104.
Thurston, P.C., 2002. Autochthonous development of Superior Province greenstone belts?
Precambrian Res. 115 (1–4), 11–36.
Thurston, P.C., Ayer, J.A., Goutier, J., Hamilton, M.A., 2008. Depositional gaps in Abitibi
greenstone belt stratigraphy: a key to exploration for syngenetic mineralization.
Econ. Geol. 103 (6), 1097–1134.
Thurston, P.C., Kamber, B.S., Whitehouse, M., 2012. Archean cherts in banded iron for-
mation: insight into Neoarchean ocean chemistry and depositional processes.
Precambrian Res. 214, 227–257.
Tomlinson, E.L., Kamber, B.S., Hoare, B.C., Stead, C.V., Ildefonse, B., 2018. An exsolution
origin for Archean mantle garnet. Geology 46 (2), 126–129.
Touboul, M., Puchtel, I.S., Walker, R.J., 2012. W-182 evidence for long-term preservation
of early mantle differentiation products. Science 335 (6072), 1065–1069.
B.S. Kamber and E.L. Tomlinson
Van Kranendonk, M.J., Hugh Smithies, R., Hickman, A.H., Champion, D., 2007. Secular
tectonic evolution of Archean continental crust: interplay between horizontal and
vertical processes in the formation of the Pilbara Craton, Australia. Terra Nova 19 (1),
1–38.
Van Kranendonk, M.J., Smithies, R.H., Hickman, A.H., Wingate, M.T.D., Bodorkos, S.,
2010. Evidence for Mesoarchean (similar to 3.2 Ga) rifting of the Pilbara Craton: the
missing link in an early Precambrian Wilson cycle. Precambrian Res. 177 (1–2),
145–161.
Van Kranendonk, M.J., Smithies, R.H., Griffin, W.L., Huston, D.L., Hickman, A.H.,
Champion, D.C., Anhaeusser, C.R., Pirajno, F., 2015. Making it thick: a volcanic
plateau origin of Palaeoarchean continental lithosphere of the Pilbara and Kaapvaal
cratons. Geol. Soc. Lond., Spec. Publ. 389 (1), 83–111.
Vermeesch, P., 2012. On the visualisation of detrital age distributions. Chem. Geol. 312,
190–194.
Walker, R., Carlson, R., Shirey, S., Boyd, F., 1989. Os, Sr, Nd, and Pb isotope systematics
of southern African peridotite xenoliths: implications for the chemical evolution of
subcontinental mantle. Geochim. Cosmochim. Acta 53 (7), 1583–1595.
Walter, M.J., 1998. Melting of garnet peridotite and the origin of komatiite and depleted
lithosphere. J. Petrol. 39 (1), 29–60.
Wasch, L.J., van der Zwan, F.M., Nebel, O., Morel, M.L.A., Hellebrand, E.W.G., Pearson,
D.G., Davies, G.R., 2009. An alternative model for silica enrichment in the Kaapvaal
subcontinental lithospheric mantle. Geochim. Cosmochim. Acta 73 (22), 6894–6917.
Webb, S.A.C., Wood, B.J., 1986. Spinel-pyroxene-garnet relationships and their depen-
dence on Cr/Al ratio. Contrib. Mineral. Petrol. 92 (4), 471–480.
Wiemer, D., Schrank, C.E., Murphy, D.T., Hickman, A.H., 2016. Lithostratigraphy and
structure of the early Archaean Doolena Gap greenstone belt, East Pilbara Terrane,
Western Australia. Precambrian Res. 282, 121–138.
151
Chemical Geology 511 (2019) 123–151
Wiemer, D., Schrank, C.E., Murphy, D.T., Wenham, L., Allen, C.M., 2018. Earth's oldest
stable crust in the Pilbara Craton formed by cyclic gravitational overturns. Nat.
Geosci. 11 (5), 357.
Wilson, A.H., 2003. A new class of silica enriched, highly depleted komatiites in the
southern Kaapvaal Craton, South Africa. Precambrian Res. 127 (1–3), 125–141.
Wilson, J.F., Nesbitt, R., Fanning, C.M., 1995. Zircon geochronology of Archaean felsic
sequences in the Zimbabwe craton: a revision of greenstone stratigraphy and models
for crustal growth. In: Coward, M.P., Ries, A.C. (Eds.), Early Precambrian Processes.
Geol. Soc. Spec. Publ., London, pp. 109–126.
Woods, M.T., Lévêque, J.J., Okal, E.A., Cara, M., 1991. Two-station measurements of
Rayleigh wave group velocity along the Hawai'ian Swell. Geophys. Res. Lett. 18 (1),
105–108.
Wyman, D., 2018. Do cratons preserve evidence of stagnant lid tectonics? Geosci. Front. 9
(1), 3–17.
Wyman, D.A., Kerrich, R., Polat, A., 2002. Assembly of Archean cratonic mantle litho-
sphere and crust: plume-arc interaction in the Abitibi-Wawa subduction-accretion
complex. Precambrian Res. 115 (1–4), 37–62.
Xu, W., Qiu, N., 2017. Heat flow and destabilized cratons: a comparative study of the
North China, Siberian, and Wyoming cratons. Int. Geol. Rev. 59 (7), 898–918.
Zhang, H., Menzies, M.A., Gurney, J.J., Zhou, X., 2001. Cratonic peridotites and silica-
rich melts: diopside-enstatite relationships in polymict xenoliths, Kaapvaal, South
Africa. Geochim. Cosmochim. Acta 65 (19), 3365–3377.
Ziberna, L., Klemme, S., 2016. Application of thermodynamic modelling to natural mantle
xenoliths: examples of density variations and pressure–temperature evolution of the
lithospheric mantle. Contrib. Mineral. Petrol. 171 (2), 16.
Ziberna, L., Klemme, S., Nimis, P., 2013. Garnet and spinel in fertile and depleted mantle:
insights from thermodynamic modelling. Contrib. Mineral. Petrol. 166 (2), 411–421.