Contrib Mineral Petrol (2002) 144: 38–56

DOI 10.1007/s00410-002-0374-5

Balz S. Kamber Æ Anthony Ewart Æ Kenneth D. Collerson

Michael C. Bruce Æ Graeme D. McDonald

Fluid-mobile trace element constraints on the role of slab melting

and implications for Archaean crustal growth models

Received: 12 July 2001 / Accepted: 9 May 2002 / Published online: 8 August 2002

Springer-Verlag 2002

Abstract We use published and new trace element data stant. Formation of adakites has therefore always de-
to identify element ratios which discriminate between pended on local plate geometry and not on potential
arc magmas from the supra-subduction zone mantle mantle temperature.
wedge and those formed by direct melting of subducted
crust (i.e. adakites). The clearest distinction is obtained
with those element ratios which are strongly fractionated
during refertilisation of the depleted mantle wedge, ul- Introduction
timately reflecting slab dehydration. Hence, adakites
have significantly lower Pb/Nd and B/Be but higher Nb/ The majority of magmas which form along present
Ta than typical arc magmas and continental crust as a convergent plate boundaries derive from the supra-
whole. Although Li and Be are also overenriched in subduction zone asthenospheric mantle wedge. Melting
continental crust, behaviour of Li/Yb and Be/Nd is of mantle wedge peridotite is triggered by aqueous fluids
more complex and these ratios do not provide unique which originate from prograde dehydration of sub-
signatures of slab melting. Archaean tonalite–trondhje- ducted slab. Composition of typical arc melts can
mite–granodiorites (TTGs) strongly resemble ordinary therefore be quantified by slab dehydration, hydration
mantle wedge-derived arc magmas in terms of fluid- of the overlying mantle wedge, melting of mantle wedge
mobile trace element content, implying that they did not peridotite, and fractional crystallisation/crustal assimi-
form by slab melting but that they originated from lation (e.g. Ayers 1998). Direct melting of the oceanic
mantle which was hydrated and enriched in elements lost lithosphere, on the other hand, is very rare. It is possible
from slabs during prograde dehydration. We suggest where the subducted lithosphere is exceptionally young
that Archaean TTGs formed by extensive fractional and therefore relatively hot (Defant and Drummond
crystallisation from a mafic precursor. It is widely 1990). Gutscher et al. (2000) proposed a model of flat
claimed that the time between the creation and sub- subduction which explains direct slab melting of ap-
duction of oceanic lithosphere was significantly shorter parently older and colder oceanic lithosphere. An al-
in the Archaean (i.e. 20 Ma) than it is today. This dif- ternative model for melting of old oceanic lithosphere
ference was seen as an attractive explanation for the was presented by Yogodzinski et al. (2001) who found
presumed preponderance of adakitic magmas during the that the edge of a torn slab can be exposed to mantle
first half of Earth’s history. However, when we consider flow, resulting in slab ablation, heating, and eventually
the effects of a higher potential mantle temperature on melting. Magmas thought to represent melts from sub-
the thickness of oceanic crust, it follows that the mean ducted slabs are collectively known as adakitic melts
age of oceanic lithosphere has remained virtually con- (Defant and Drummond 1990; following Marsh 1976,
and Kay 1978). They are typically andesitic to dacitic in
composition and characterised by strongly fractionated
B.S. Kamber (&) Æ K.D. Collerson
ACQUIRE Centre, Department of Earth Sciences,
rare-earth-element (REE) patterns. Their depletion in
University of Queensland, Brisbane, the heavy REEs (HREEs) relative to the light REEs
Queensland 4072, Australia (LREEs) is commonly interpreted to reflect residual
E-mail: [email protected] garnet (and amphibole) in the melt source. Kay (1978)
A. Ewart Æ M.C. Bruce Æ G.D. McDonald proposed that the incompatible trace element contents
Department of Earth Sciences, University of Queensland, of adakites required a metabasaltic rather than a peri-
Brisbane, Queensland 4072, Australia dotitic source. Adakites are thus believed to form by wet
Editorial responsibility: T.L. Grove melting of subducted basaltic eclogite.
 39

Based on the similarity in REE patterns between which are preferentially transported into the mantle
Archaean tonalite–trondhjemite–granodiorites (TTGs) wedge. On this premise, we review some geochemical
and modern adakites, Martin (1986) proposed that both features of modern island arc basalts which we identify
could represent direct melting products of subducted as potential discriminators between adakites and ordi-
(eclogite facies) oceanic lithosphere. Martin (1986) and nary arc melts.
many subsequent proponents of Archaean slab melting Element transport from dehydrating slabs to the
speculated that the Earth’s higher heat output in the overlying mantle wedge is ultimately expressed in the
Archaean and, in particular, the higher potential mantle chemical inventory of the continental crust as a whole.
temperature would have led to higher oceanic crust The degree to which elements are enriched in continental
production rate and hence subduction of, on average, crust follows, in general, the order of incompatibility
younger and hotter slabs which would have facilitated found in mid-ocean-ridge basalts (MORBs). This is vi-
direct slab melting, thereby providing the tectonic ra- sualised by the negative slope in a plot of MORB-nor-
tionale for the alleged preponderance of direct slab malised concentration of incompatible elements in
melting in the Archaean. average continental crust arranged in order of decreas-
Since postulation of the adakite–TTG connection, ing enrichment of average MORB (Fig. 1). There are,
our understanding of arc magmatism has improved however, some important exceptions to this rule (Hof-
considerably. Yet modern experimental, observational, mann 1988; Brenan et al. 1998a): Pb, B, and Li are
and geophysical evidence has rarely been used to scru- overproportionally concentrated in continental crust,
tinise the claim for the adakite–TTG connection. In this and contents of the pentavalent elements Nb, Ta, and Ti
paper, we review some of the salient geochemical fea- are lower than expected. There is general agreement that
tures created by material transfer between slab and overenrichment of island arc magmas and the conti-
mantle wedge and verify whether, in true adakites, fin- nental crust as a whole in Pb and B is the result of
gerprints of slab dehydration are indeed suppressed. We preferential transfer of these elements to the mantle
find, based on fluid-sensitive geochemical criteria, that wedge by fluids originating from subducted slabs (e.g.
typical Archaean TTGs are not adakites and that their Ryan and Langmuir 1993; Miller et al. 1994; Brenan
strong LREE/HREE fractionation was not caused by et al. 1995a). The reason behind this phenomenon is that
residual garnet in oceanic eclogite. We demonstrate also the slab mineralogy/fluid partition coefficients of Pb and
that, on average, Archaean oceanic lithosphere was not B are more than one order of magnitude smaller than
significantly younger than at present because of the ef- those of the similarly compatible (in MORB melting) Nd
fects of higher Archaean potential mantle temperature and Be (Brenan et al. 1995a, 1998a, 1998b). However,
on the thickness of melt column (e.g. Galer 1991). due to the continuous nature of dehydration in a sub-
ducted slab (Schmidt and Poli 1998), and due to the
progressive change in slab mineralogy, the extent of
Fingerprints of slab dehydration overenrichment of the fluid in mobile element depends
on a large number of factors. The most straightforward
Stronger HREE depletion of adakites compared to
mantle wedge-derived arc magmas is but one diagnostic
feature of direct slab melting (note that because the
Y/Ho ratio is constant in terrestrial samples, Y-based
ratios are simply another way of expressing HREE de-
pletion). Recent experimental constraints (e.g. Brenan
et al. 1995a; Keppler 1996; Tatsumi and Kogiso 1997;
Brenan et al. 1998a; Stalder et al. 1998), continually
improving understanding of arc magma chemistry (e.g.
Ryan and Langmuir 1988; Woodhead et al. 1998; Ewart
et al. 1998; Tomascak et al. 2000), and accurate deter-
mination of trace element contents of high-pressure
metamorphic rocks and particularly eclogites (Henry
et al. 1996; Becker et al. 2000) have revealed a number of
geochemical features which can be directly linked to
dehydration of subducted slabs, rather than to the hy-
pothesised residual source mineralogy. These geochem-
Fig. 1. N-MORB-normalised (Sun and McDonough 1989) trace
ical fingerprints are based on the observation that element pattern of average continental crust (estimate by Rudnick
dehydration-driven element transport from the slab to and Fountain 1995, except for Nb, Ta, B, and Be which were taken
the overlying mantle wedge is capable of fractionating from McLennan 2001). Elements are arranged in decreasing
element pairs which have very similar solid/melt distri- MORB-melting compatibility (after Sun and McDonough 1989).
MORB Be concentration was calculated using a constant Be/Nd
bution coefficients. If adakites and Archaean TTGs were ratio of 0.05 (Ryan and Langmuir 1988), and B concentration was
direct slab melting products, they would be expected to estimated using a constant B/Be ratio of 3 (Ryan and Langmuir
be less enriched than typical arc magmas in elements 1993)
40

expression of this phenomenon is the observation that higher than that in the corresponding wall rocks.
trace element chemistry of magmas changes systemati- However, at higher metamorphic grade when clinopy-
cally across some arcs at least (Ryan et al. 1995; roxene is part of the mineral assemblage, it can be ex-
Woodhead et al. 1998). These variations most probably pected from experimental data (Brenan et al. 1995a) that
relate to the slab depth which, in turn, determines the Be remains immobile and is retained in the slab. The Be/
volume and chemistry of fluid expelled to the overlying Nd ratio of island arc magmas may thus hold potential
mantle wedge. Nevertheless, for the highly mobile ele- to constrain possible metamorphic dehydration reac-
ments Pb and B, overenrichment can be expected in tions in the subducting slab.
expelled fluids from a range of depths. It is thus rea- The collective depletion of Nb, Ta and Ti is not a
sonable to expect that typical adakites should have conclusive fingerprint for slab dehydration but Jochum
lower Pb/Nd and B/Be ratios than melts which are de- et al. (1989) and Green (1995) have demonstrated that
rived from hydrated mantle wedge. Furthermore, by the the geochemical twins Nb and Ta are depleted in the
time the crustal portion of the slab has metamorphosed continents to surprisingly different degrees. Fig. 1 shows
to garnet amphibolite or eclogite, prograde dehydration that Ta is noticeably less depleted than its geochemical
will have removed substantial fractions of the original twin Nb. Given the near-chondritic Nb/Ta ratio of
Pb and B, such that the Pb/Nd and B/Be ratios of the MORB, this translates to a substantially subchondritic
metamorphosed slab can be expected to be lower than in continental Nb/Ta ratio of ca. 10. Empirical observa-
the hydrated oceanic crustal protolith (Brenan et al. tions and experimental constraints indicate that com-
1995a, 1998a; Kamber and Collerson 2000; Becker et al. plexing in aqueous fluids is a likely process which
2000). fractionates elements of identical charge and near-iden-
Current estimates of the Li content of continental tical effective ionic radius (Bau 1996; Kamber and
crust suggest that, like Pb and B, Li is also preferentially Collerson 2000), implying that the low continental Nb/
transported from slabs to the overlying mantle wedge. Ta ratio reflects preferential transport of Ta over Nb
Indeed, Li/Yb (or Li/Y for sample sets without accurate during slab dehydration. The Nb/Ta ratio is thus an
HREE data) are generally high in those arc magmas equivalent tool to the Pb/Nd ratio (Miller et al. 1994;
where a high, slab-derived fluid flux is implied by other Brenan et al. 1995a) but has the advantage that it is less
geochemical tracers. In the few cases where Li-isotope affected by crystal fractionation (Klein et al. 1997;
data are available, it appears that the excess Li is derived Münker 1998). Again, if adakites and Archaean TTGs
from altered oceanic crust (Chan et al. 1999; Tomascak were direct slab melts, they are predicted to have near-
et al. 2000). However, compared to Pb and B, the be- chondritic (or higher than chondritic) Nb/Ta ratios, re-
haviour of Li in the mantle wedge is more complicated. gardless of collective depletion in Nb, Ta, and Ti.
Firstly, because Li is quite compatible in olivine (Brenan In brief, there is now very strong experimental and
et al. 1995b), melts from Li-enriched mantle wedge empirical evidence (both from island arc volcanics and
peridotite may not necessarily inherit a high Li/Yb ratio metamorphic rocks) that the characteristic, relative en-
from their source. Secondly, release of Li with fluid richment and depletion of average continental crust in
during metamorphic dehydration may be more strongly some elements is due to the transfer and retention of
controlled by mineralogy than release of Pb and B. For elements from subducting slabs into the mantle wedge.
example, Li/Zr ratios of metamorphic veins (interpreted This leaves little doubt that the bulk of continental crust
as having precipitated from synmetamorphic fluids) in has been extracted from the supra-subduction zone
high-P, low-T metamorphic rocks from the Western mantle wedge. If TTGs, which represent a prominent
Alps (Henry et al. 1996) are highly variable. In many group of Archaean magmatic rocks, formed by a dif-
samples, the ratio in the vein is identical to that in the ferent process (i.e. direct fusion of subducted oceanic
wall rock whereas, in other cases, the veins are very crust), we can reasonably expect them to have, at least
strongly enriched in Li. Thirdly, because Li is compared on average, relative trace element contents which are not
to the HREE Yb which is severely depleted in adakites, characteristic of hydrous element transfer. Before we test
it is not clear that Li/Yb ratios in adakites should be this hypothesis with new trace element data for early and
lower than in typical arc melts. late Archaean TTGs, we inspect whether modern direct
A further element which seems to be mildly overen- slab melts (i.e. adakites) confirm these geochemical
riched in continental crust is Be (Fig. 1). Compared to B, predictions.
which has a very similar distribution coefficient in
MORB melting (Ryan and Langmuir 1993), the en-
richment is not nearly as strong. Transfer of Be from Geochemistry of modern adakites
hydrated oceanic crust or from sediment into the mantle
wedge and ultimately into island arc magmas is directly It is widely recognised that the chemistry of rocks which
expressed in the overabundance of the cosmogenic nu- classify as adakites is invariably more complex than
clide 10Be in some island arc volcanics (Brown et al. could be expected from a direct melt derived from gar-
1982). Preferential Be loss during metamorphic dehy- net–amphibolite or eclogite facies MORB (e.g. Stern and
dration is also evident in the data of Henry et al. (1996), Kilian 1996). Several factors are believed to complicate
because the Be/Zr ratio of most synmetamorphic veins is adakite chemistry, the most important of which is that
 41

adakites always occur as volumetrically minor parts database comprising 140 samples (their Table 1) but it
together with island arc volcanics of more typical seems improbable that complete data sets were available
chemistry. The coexistence of melts from the slab itself for all samples, and average Nb and Ta values are ex-
and from the supra-subduction zone mantle wedge is pected to be defined on a much smaller number of
expected because prograde dehydration from relatively samples. Nevertheless, the average adakite Nb/Ta cal-
hot slabs triggers melting of the mantle wedge peridotite. culated from the Drummond et al. (1996) compilation is
This bears the potential of creating hybrid melts in 15.7, which is only slightly lower than modern MORB
which the typical adakite geochemical fingerprint is (16.66; Kamber and Collerson 2000) and clearly higher
obscured. A further factor which complicates adakite than average continental crust (10).
chemistry is the possibility of peridotite assimilation Smith and Leeman (1987) reported Nb, Ta, and Be
upon migration through the mantle wedge. This process (but not Li, Pb, and B) concentration data for adakitic
is inferred on the basis of elevated compatible trace el- dacites and rhyolites from Mt. St. Helens. The median
ement content in some adakites, which would be ex- Nb/Ta ratio of those lavas is 17.3±2.4 (n=10), which is
pected to be depleted in elements like Cr and Ni. It is much higher than that of continental crust. It is also
generally argued that peridotite assimilation should not considerably higher than median Nb/Ta ratios of arc
greatly disturb the original incompatible element ratios. lavas which are derived from previously depleted supra-
However, the incompatible element inventory could be subduction zone mantle wedge which was metasoma-
seriously affected if the melts were not derived from pure tised by slab-derived fluids. For example, the average
MORB but had a contribution from the thin veneer of Nb/Ta ratio of Tongan lavas is 12.9±2.7 (n=21; Tonga
sediment which is sometimes subducted on top of slabs. Island only, data from Ewart et al. 1998), and that of
Because this sediment material inherits the chemical New Britain Island Arc lavas from sectors E and F is
signature of continental crust (Fig. 1), relatively small 14.0±0.9 (n=16; data from Woodhead et al. 1998).
melt contribution from metasediments can effectively Stern and Kilian (1996) presented a combined trace
mimic typical island arc, volcanic incompatible element element and radiogenic isotope data set for adakites
geochemistry even though the melt is derived from the from the south Andean Austral Volcanic Zone (Chile).
slab. Finally, adakite geochemistry can further be After evaluation of radiogenic isotope signatures, they
modified by fractional crystallisation (Romick et al. concluded that only adakitic andesites from Cook Island
1992) and assimilation of crustal material. volcano are pure melts from an eclogitic assemblage
There exists thus a category of rocks which have with MORB composition. These samples (Ck-3-198, Ck-
some adakitic affinity but whose chemical integrity has 1-468, Ck-3-197; Table 1 of Stern and Kilian 1996) have
been seriously compromised by one or more of these the high average Nb/Ta ratio of 17.6±1.8 predicted for
processes. Although crustal contamination or sediment direct slab melts. Adakitic rocks from the other volcanic
contribution to the source can be monitored with radi- centres which, based on radiogenic isotopes (not Nb/Ta
ogenic isotopes (Stern and Kilian 1996), the general lack considerations) were postulated to contain a sediment
of isotope data for many postulated adakite occurrences component, have invariably lower ratios (average of
compromises the interpretation of literature data. De- 11.3±2.83). No Pb, Li, B, and Be concentration data
spite all these complications, we show below that were reported for these rocks.
adakitic rocks still yield, on average, significantly higher High-Mg andesites from the Ryukyu Arc, Japan,
Nb/Ta and lower Pb/Nd, B/Be, and Be/Nd than mantle which share many similarities with adakites (Shinjo
wedge-derived arc magmas and continental crust as a 1999), also have a high median Nb/Ta ratio of
whole. 16.82±2.02. Bruce and Niu (2000a, 2000b) recently
published complete geochemical data for early Permian
adakitic intrusives from the Marlborough Block and
Review of literature data Percy Isles, Eastern Australia. These rocks (samples
MB 20, 73, and 74b) have a high average Nb/Ta ratio of
The majority of geochemical studies on adakites have 20.09±3.04.
focused on REE distribution because it was the strong The only adakites for which available data suggest
HREE depletion which led to their discovery and defi- relatively low Nb/Ta ratios (14.83±1.36) are from the
nition (Kay 1978). Very few complete trace element data Philippines (Prouteau et al. 2000). It is unclear whether
sets for Phanerozoic adakites are available. Therefore, this is due to the use of mixed ICP-MS (Nb) and INAA
we discuss adakite geochemistry in order of specific trace (Ta) analytical techniques, fractionation by Ti phases
element ratios rather than occurrence. (Prouteau et al. 2000), or a non-slab melt origin of these
adakites (Yumul et al. 2000). High Pb/Nd ratios (Yumul
et al. 2000) from Central Luzon back-arc adakites ap-
Adakite Nb/Ta systematics pear to favour the latter explanation.
In summary, the Nb/Ta ratio of adakites and high-
Drummond et al. (1996) published a comprehensive Mg andesites of adakitic affinity is similar to that of
overview of world-wide Cenozoic adakites. Their com- MORB (Fig. 2a) and clearly higher than that of average
posite average Cenozoic adakite is quoted to reflect a continental crust or typical island arc volcanics.
42

older, calc-alkaline lavas which yield a B/Be ratio of

11.3±6.1. In other arcs with strong dehydration signa-
ture, such as the Vanuatu Arc, even higher B/Be of
18.2±12.6 has been recorded (Peate et al. 1997). Higher
still is the average B/Be ratio (36.5±30.2) for Central
American Volcanic arc lavas published by Chan et al.
(1999). Hence, from the limited evidence available to
date, it appears that the geochemical prediction is con-
firmed in that adakites are not nearly as enriched in B as
typical island arc magmas.

Adakite Pb/Nd systematics

Surprisingly few data sets exist in the literature which

contain Nd and Pb concentration measurements for
adakites. The median Pb/Nd ratio of 0.329±0.105 of
high-Mg andesites from the Ryukyu Arc (Japan) is
lower than estimates of typical continental crust which
range from 0.5 (McLennan 2001) to 0.63 (Rudnick and
Fountain 1995). The early Permian adakites from east-
ern Australia also have variable but low Pb/Nd
(0.390±0.405; data from Bruce and Niu 2000a, 2000b).
Late Miocene adakites and Nb-enriched basalts from
Baja California (n=5) have similarly low Pb/Nd ratios
of 0.313±0.323 (Aguillón-Robles et al. 2001). When
compared to lavas from Tonga (0.499±0.280; Tonga
Island only, data from Ewart et al. 1998) and from the
New Britain Island Arc (0.494±0.161; data from sectors
E and F from Woodhead et al. 1998), it appears that
adakites have, on average, consistently lower Pb/Nd
ratio (Fig. 2b), again confirming geochemical predic-
tions.

Adakite Li/Yb systematics

Both in lavas from Tonga (data from Ewart et al. 1998)

and from the New Britain Arc (Woodhead et al. 1998),
the Li/Yb ratio is positively correlated with the Pb/Nd,
Fig. 2a–c. Fluid-sensitive element ratios in normal arc magmas Nb/Ta, and Be/Nd ratios. The correlation is strongest
compared to adakites and lavas of adakitic affinity. Literature (r=0.782) between Li/Yb and Pb/Nd in the Tongan
sources are given in text. a Nb/Ta ratio: normal arc magmas (from data set, which implies that those samples with the
left to right) are Tonga, New Britain Arc, and average continental strongest overenrichment in Li also have strongest ov-
crust; average N-MORB from Sun and McDonough (1989);
adakites are (from left to right) average compiled by Drummond
erenrichment in Pb (and Ta and Be), which supports the
et al. (1996), Ryukyu Arc, Mt. St. Helens, Cook Island, Aleutian interpretation of element abundance in the continental
Arc, East Queensland, and Philippines. b Pb/Nd ratio: normal arc crust (Fig. 1). However, Li/Yb ratios of adakites are
magmas are from Tonga and New Britain Arc; adakites (from left higher than those in typical continental crust. Two
to right) are from Ryukyu Arc, Aleutian Arc, East Queensland, and samples from Panama have a Li/Yb ratio of 36.4±15.67
Baja California. c Be/Nd ratio: normal arc magmas are from New
Britain Arc, Tonga, and Panama; adakites are from Mt. St. Helens, (Defant et al. 1991a, 1991b; Tomascak et al. 2000), and
East Queensland, Panama, and Aleutian Arc three adakitic samples of eastern Australia yield a ratio
of 16.48±19.7. This observation has two plausible ex-
planations. Firstly, the Li/Yb ratio is strongly biased by
Adakite B/Be systematics the adakites’ Yb depletion (due to garnet in their
source). Secondly, because the peridotite source of nor-
To our knowledge, adakites from Panama are the only mal arc basalts contains olivine, Li can be retained in the
occurrence for which combined B and Be data exist slab whereas Li is lost from the garnet–clinopyroxene
(Tomascak et al. 2000). The average B/Be of four source of adakites. Thus, despite the clear overenrich-
adakites is 6.76±1.01, lower than the average of six ment of continental crust in Li, the Li/Yb ratio is not a
 43

suitable discriminator between adakites and mantle average Sr/Y ratio (42) is not as high as that of proper
wedge-derived island arc volcanics. adakites (61; Drummond et al. 1996). Nevertheless, the
average Sr/Y ratio is still much higher than in average
continental crust (13; McLennan 2001). There is a dis-
Be/Nd systematics cernible positive correlation between Sr/Y and La/Yb
ratios in the Aleutian data set which follows the trend of
From the relative abundance in continental crust, we can global arc magmas (Fig. 4). Although the Aleutian
expect that adakites will have a slightly lower Be/Nd samples measured in this study are not as depleted in
ratio than continental crust (Fig. 1). A comprehensive HREE and not as enriched in La and Sr as most
data set (n=26) exists for Mt. St. Helens adakitic lavas adakites, they nevertheless are clearly resolved from
(Smith and Leeman 1987) in which the average Be/Nd more typical arc magmas like those from Tonga (Fig. 4).
ratio is 0.080±0.024 (note that Nd abundance was es- This could suggest that the Tonga-Lau basin represents
timated from Sm concentration with the average Ceno- an end-member oceanic arc system which is poor in al-
zoic adakite Sm/Nd ratio of 0.1564; Drummond et al. kalis and alkaline earths, and very depleted in high-field-
1996). This is identical to the average Be/Nd ratio strength elements after pronounced and ongoing mantle
(0.08±0.03) of three early Permian adakites from east- wedge depletion. Interestingly, some of the samples from
ern Australia (Bruce and Niu 2000a, 2000b), and that of the extreme northern end of the arc system, where the
adakites (0.079±0.058) from Panama (Defant et al. slab is being torn (and is shallowing), seem to be de-
1991a, 1991b; Tomascak et al. 2000). These values are veloping ‘adakitic’ affinity.
indeed lower than average continental crust which, ac- In comparing the Aleutian data (and many Cenozoic
cording to McLennan’s (2001) compilation, has a Be/Nd island arc volcanics of adakitic affinity), we also have to
ratio of 0.094. Equally, the New Britain Arc lavas from consider that their HREE depletion is not as pro-
sectors E–F have a higher Be/Nd ratio (0.104±0.014, nounced as that of Archaean TTGs (Fig. 3). The average
n=16; data from Woodhead et al. 1998). However, lavas Archaean TTG SiO2 content is 70.2 wt%, which is much
from the Island of Tonga (Ewart et al. 1998) have lower higher than that of Cenozoic adakites (63.9; Drummond
Be/Nd (0.075±0.025) than continental crust, as have et al. 1996). Our samples from Adak and nearby islands
Panamese lavas (0.060±0.015) with slab-derived Li have an even lower SiO2 content (51.45 wt%, Table 1). It
(Defant et al. 1991a, 1991b; Tomascak et al. 2000). This is obvious from major-element chemistry alone that
comparison therefore shows (Fig. 2c) that similar rocks with adakitic affinity from the Aleutian Arc, in-
amounts of Be and Nd are lost during slab dehydration cluding the type-locality ‘adakites’, are not pure slab
and, although Be/Nd ratios of adakites appear to be melts. Instead, it has been argued that they represent
low, they do not constitute conclusive evidence for or hybrid magmas derived from melting of eclogitic sub-
against direct slab melting. ducted MORB, a subducted sediment component and a
In summary, a review of the geochemical literature supra-subduction zone component, with less evolved
confirms that Nb/Ta, Pb/Nd, and B/Be, representing REE patterns (e.g. von Drach et al. 1986). Myers and
element pairs with similar MORB melting partition co- Marsh (1987) presented Pb-isotope data (Fig. 5) for la-
efficients but different fluid/solid partition coefficients, vas from Adak and Atka islands which clearly demon-
clearly discriminate between possible direct slab melts strate a mixing relationship. Although the mixing end
and typical arc melts from hydrated mantle wedge per- members could not be uniquely constrained, Myers and
idotite (Fig. 2a, b). The Be/Nd ratio has similar beha- Marsh (1987) speculated that the original melt of Adak
viour but, due to the relative immobility of Be (despite Island lavas could have derived from direct slab melting.
incontrovertible cosmogenic Be evidence), its resolution Here we use their isotope data to complement interpre-
is insufficient (Fig. 2c). Li/Yb systematics of arc magmas tation of our new, comprehensive trace element data
are compromised by the role of garnet (Yb retention) (Table 1). The aim is to explore, within the framework of
and olivine (Li retention) which can lead to chromato- our preceding discussion of adakite trace element syst-
graphic effects (Tomascak et al. 2000). ematics, whether a direct slab component can be iso-
lated.
The Aleutian lavas show a wide spread in Pb/Nd
New geochemical data for Aleutian lavas ratio (Fig. 6a), ranging from 0.16 to 0.55 (sample AD-72
has an even higher ratio of 0.78 which we regard as
In order to substantiate the findings from the literature anomalous). The samples at the higher end of the
survey and to test whether an adakitic component can be spectrum are comparable to typical island arc basalts,
detected in hybrid melts, we report new trace element with Pb/Nd between 0.5 and 0.65 (Fig. 2b) whereas
data for lavas from the Aleutian Arc (Table 1). The many of the Adak and Bogoslof samples have much
samples are from Adak Island (nine samples), Bogoslof lower ratios, similar to adakites (Fig. 2b). There is a
Island (three samples) and Great Sitkin Island (one scattered anti-correlation between the Pb/Nd and La/Yb
sample). None of them classifies as adakite, as they are ratio, indicating that the samples most strongly enriched
not enough evolved (Table 1), their LREE/HREE in Pb also have the lowest La/Yb ratios, whereas those
fractionation is not sufficiently strong (Fig. 3), and their with steeper REE patterns clearly have Pb/Nd ratios
44

Table 1. Chemical compositions of Aleutian Arc lavas (all trace elements by ICP-MS)

Tectonic unit Aleutian Arc

Locality Adak Bogoslof Great Sitkin

Sample ID AD-14 AD-17 AD-36 AD-39 AD-44 AD-49 AD-51 AD-61 AD-72 BOG-3 BOG-5 BOG-6 GS-4
Referencea 1 1 1 1 1 1 1 1 1 2 2 3 1

Major element (wt%)

SiO2 49.93 48.39 50.87 51.39 55.27 56.34 51.98 52.14 57.63 48.09 47.23 47.95 52.39
TiO2 0.90 0.92 0.75 0.84 0.70 0.72 0.64 0.67 0.54 1.29 1.33 1.30 0.89
Al2O3 19.46 17.90 16.12 18.78 18.12 17.17 18.63 19.44 16.13 18.00 17.50 17.97 18.88
Fe2O3 3.24 4.48 4.76 4.96 3.90 4.83 4.28 4.79 3.28 5.45 5.11 5.51 3.12
FeO 6.30 6.30 4.54 4.83 4.11 2.76 3.87 4.25 3.64 4.73 5.59 4.73 6.35
MnO 0.18 0.17 0.16 0.18 0.16 0.14 0.12 0.19 0.15 0.16 0.17 0.16 0.19
MgO 5.07 6.13 6.83 4.38 3.73 4.80 3.86 3.73 5.52 5.30 5.74 5.29 4.39
CaO 10.43 11.83 10.84 10.16 8.53 8.97 8.40 9.79 8.88 11.72 12.17 11.18 9.52
Na2O 3.31 2.98 4.16 3.09 3.68 1.32 2.93 2.92 2.93 2.91 2.91 3.02 3.26
K2O 1.05 1.00 1.34 1.31 1.48 1.89 0.87 0.59 1.22 1.80 1.65 1.83 1.08
P2O5 0.23 0.30 0.19 0.23 0.20 0.24 0.14 0.19 0.15 0.34 0.38 0.37 0.25
Trace element (ppm)
Li 6.98 5.06 9.79 9.06 10.4 12.9 4.50 5.88 13.9 6.08 5.42 5.70 9.49
Be 0.62 0.62 0.68 0.64 0.74 0.89 0.43 0.62 0.66 0.92 0.85 1.03 0.62
Sc 23.0 44.5 36.7 26.2 18.0 25.5 10.8 13.4 27.2 37.7 47.4 31.0 22.9
V 244 340 277 320 193 236 136 134 159 328 361 315 224
Cr 23.3 37.0 125 10.1 7.58 55.1 2.69 3.85 234 24.3 40.9 20.3 5.20
Co 28.5 39.5 34.5 32.9 21.4 21.9 14.5 19.1 23.1 31.9 35.5 31.1 26.1
Ni 13.1 19.7 36.9 10.7 8.88 17.6 4.20 4.84 39.6 19.1 22.0 18.3 9.89
Cu 44.7 70.4 78.8 104 59.9 49.2 31.5 30.3 32.3 89.1 78.5 59.7 107
Zn 72.3 85.3 63.2 79.6 59.4 55.8 45.0 65.3 49.2 72.3 74.1 73.7 72.8
Ga 18.3 18.1 17.5 19.0 17.5 17.5 12.9 18.1 15.7 19.3 18.7 19.7 18.8
Rb 21.7 22.7 19.6 20.7 30.2 29.4 9.67 10.7 27.1 44.2 40.0 44.2 18.0
Sr 451 390 652 564 467 740 353 580 474 754 717 790 488
Y 15.1 16.5 12.0 13.8 14.1 10.3 7.77 9.87 10.2 18.0 18.1 18.7 18.2
Zr 60.8 57.6 58.3 60.6 77.9 57.2 43.1 56.7 53.3 95.1 88.6 104 77.2
Nb 1.70 1.66 1.48 1.59 1.84 2.02 0.79 1.16 1.51 3.07 2.61 3.45 1.84
Sn 0.78 0.97 0.80 0.86 0.68 0.74 0.36 0.49 0.48 1.57 2.04 1.63 0.98
Cs 0.76 0.54 0.81 0.36 0.78 1.08 0.29 0.28 1.62 0.67 1.32 0.51 0.93
Ba 424 455 364 396 537 504 235 286 504 839 726 860 511
La 7.79 7.48 8.46 7.83 9.64 9.91 5.98 6.28 7.64 13.7 12.5 15.5 7.94
Ce 18.0 17.7 19.4 18.6 21.1 21.6 12.8 14.7 16.1 34.4 32.1 37.9 19.3
Pr 2.66 2.69 2.82 2.75 2.92 3.00 1.81 2.15 2.15 5.20 4.97 5.65 2.85
Nd 11.6 12.1 12.2 12.3 12.1 12.7 7.50 9.47 8.76 23.1 22.2 24.5 12.6
Sm 2.83 3.12 2.81 2.99 2.76 2.80 1.68 2.26 2.00 5.15 5.00 5.30 3.25
Eu 0.97 1.03 0.88 0.96 0.89 0.87 0.56 0.85 0.68 1.49 1.51 1.55 1.07
Gd 2.85 3.26 2.54 2.86 2.64 2.32 1.59 2.13 1.89 4.46 4.38 4.54 3.39
Tb 0.44 0.51 0.38 0.43 0.41 0.33 0.24 0.31 0.30 0.63 0.64 0.65 0.54
Dy 2.81 3.22 2.31 2.68 2.58 1.98 1.46 1.91 1.90 3.71 3.75 3.77 3.42
Ho 0.60 0.68 0.49 0.56 0.55 0.41 0.31 0.39 0.40 0.74 0.74 0.76 0.73
Er 1.72 1.89 1.38 1.59 1.59 1.17 0.88 1.11 1.16 1.99 1.99 2.11 2.13
Tm 0.26 0.28 0.21 0.23 0.25 0.18 0.13 0.16 0.18 0.29 0.29 0.31 0.32
Yb 1.68 1.75 1.31 1.52 1.63 1.15 0.87 1.08 1.22 1.83 1.81 1.92 2.08
Lu 0.26 0.27 0.20 0.23 0.26 0.18 0.14 0.17 0.19 0.27 0.27 0.29 0.32
Hf 1.76 1.81 1.75 1.80 2.19 1.88 1.21 1.64 1.65 2.98 2.83 3.11 2.24
Ta 0.10 0.10 0.10 0.10 0.13 0.13 0.06 0.08 0.11 0.20 0.16 0.20 0.12
Pb 4.56 4.63 4.59 6.58 5.17 4.36 2.73 2.57 6.82 3.62 4.07 4.14 7.01
Th 1.98 2.08 2.14 1.94 3.14 2.82 1.46 1.09 2.74 3.12 2.63 3.51 2.16
U 0.89 0.93 0.94 0.89 1.46 1.11 0.65 0.56 1.27 1.39 1.21 1.59 0.92
a
Major-element data from 1, Marsh (1976); 2, Marsh and Leitz (1978); 3, von Drach et al. (1986)

much lower than typical island arc magmas or conti- (Fig. 6b) allows interpretation of a plot of Pb-isotope
nental crust as a whole. Importantly, in our view, the ratio vs. Pb content (Fig. 6c) in terms of source contri-
Pb/Nd ratios of the Adak Island lavas are strongly bution. There is good correlation between Pb content
controlled by the abundance of Pb (Fig. 6b) and not by and Pb-isotope composition, with the Pb-depleted
the abundance of Nd. Although no published Nd con- samples plotting at unradiogenic isotope ratios whereas
centration data exist for the samples for which Myers Pb-rich samples plot at the radiogenic end of the mixing
and Marsh (1987) obtained Pb-isotope data, the good line. Hence, combined chemical and isotopic informa-
correlation between Pb content and Pb/Nd ratio tion for Adak Island lavas reveals that those samples
 45

theless an affinity with the Adak Island lavas in La/Yb

and Sr/Y. It appears that back-arc volcanics show closer
affinities with their associated arc volcanoes than to each
other (Fig. 4). This could suggest that Bogoslof (and
Adak) lavas were derived from a less depleted mantle
wedge than the Niua fo’ou back arcs and their associ-
ated Tonga-Lau arc volcanics.
In summary, careful inspection of complex island arc
lavas for which an adakitic component is indicated by
traditional geochemical fingerprints (i.e. Sr/Y) requires
concentration data for elements which are fluid sensitive
(Pb, Ta, B, Li) and elements with comparable solid/melt
distribution coefficients (Nd, Nb, Be, Yb). Characteri-
sation of the adakitic component is possible in Phan-
erozoic lavas, even in hybrid examples such as the Adak
Fig. 3. Plot of La/Yb versus Yb content after Martin (1986). Island, and should therefore also be feasible in Archaean
Archaean TTGs define a field characterised by high La/Yb ratios at TTGs, provided that fresh samples are used which have
comparatively low Yb contents (shaded, after Martin 1986),
thought to reflect HREE depletion. Post-Archaean granitoids
not experienced metamorphic loss of fluid-sensitive ele-
define a field with comparatively lower La/Yb ratios and higher Yb ments.
contents (nonshaded, after Martin 1986). Chondrite values are from
McDonough and Sun (1995). Average of Cenozoic adakites (dotted
open circle) after Drummond et al. (1996) plots into the overlapping
area of both fields. The majority of other subrecent adakites (small
Geochemistry of Archaean TTGs
open circles) from the Andean Austral Volcanic Zone, Chile (Stern
and Kilian 1996), Solander Island, New Zealand (Reay and Comparison of trace element ratios in adakites vs. Ar-
Parkinson 1997), and Mount St. Helens, USA (Smith and Leeman chaean TTGs must take into account that TTGs are, on
1987) also plot into the area of overlap between Archaean and average, more evolved than adakites and, due to their
post-Archaean granitoids. Five data points for Permian SE
Queensland adakites (closed circles) plot into the Archaean TTG coarse-grained nature, also chemically more heteroge-
field (note that the most HREE-depleted sample plots off-scale). neous on the scale of typical hand specimens. Variations
Island arc volcanics from the Aleutian Islands (large open circles) in ratios of trace elements of similar compatibility can
are only weakly depleted in Yb. Four samples (from Adak Island) therefore amount to >50% in highly evolved plutonic
plot into the area of overlap between Archaean and post-Archaean
granitoids. The remaining samples (five from Adak Island, three samples, even for a suite of samples from a single pluton.
from Bogoslof Island, and one from Great Sitkin Island) plot into For example, the Pb/Nd ratio of the arc-related 375-Ma
the post-Archaean granitoid field Shelburne Pluton (southern Nova Scotia; data from
Currie et al. 1998) varies by 67% (1 standard deviation)
yet the median ratio of 0.94 is typical of arc magmas. An
with the highest La/Yb ratios and the lowest Pb/Nd additional source of chemical variability comes from the
ratios (i.e. those with the strongest adakitic affinity) have possibility of multiple melt sources and assimilation
Pb-isotope characteristics of MORB or the MORB- (both in the mantle and in the crust). Our characterisa-
source mantle. This we interpret as evidence that these tion of TTG geochemistry therefore relies on compari-
melts, albeit hybridised with ordinary wedge-derived son of average values.
melts or contaminated by peridotite (Yogodzinski et al.
1995), were originally of adakitic parentage. For the
purpose of this paper, we therefore suggest that the New geochemical data for early and late Archaean
samples with the lowest Pb/Nd and Be/Nd and highest TTGs
Nb/Ta should be used to estimate the Aleutian adakite
component. We regard the median ratios (Nb/Ta=15.6 We report new trace element data for three sets of Ar-
±1.1; Pb/Nd=0.377±0.174; Be/Nd=0.054±0.011) chaean TTGs (Tables 2, 3). Firstly, for a suite of five
as minima (Nb/Ta) and maxima (Pb/Nd and Be/Nd) 3.80-Ga tonalitic Itsaq gneisses which were collected 6 to
respectively. 20 km south of the Isua greenstone belt (SW Green-
The more evolved Pb-isotope end member could land). The geology of this area has been described by
represent an enriched mantle domain in the supra-sub- Nutman et al. (1999) who concluded that the exposed
duction zone wedge, a sediment component in the slab, gneisses represent the best-preserved early Archaean
or a lithospheric contaminant. An interesting aspect of crust known on Earth. Intrusion ages cluster tightly
the Aleutian data set is shown by the three Bogoslof between 3.79 and 3.81 Ga (Nutman et al. 1999). Sec-
samples which represent the back arc. These samples ondly, for a suite of ten type-locality Amı̂tsoq gneisses
have high Nb/Ta and low Pb/Nd, Li/Yb and Be/Nd from the Godthabsfjord area and the Isua dome in West
ratios, indicating that they derive from a part of the Greenland. These samples belong to a relatively late
mantle which has not experienced as much enrichment suite of intrusions and are widely believed to be 3.65 Ga
driven by dehydration of the slab, but there is never- old (e.g. Whitehouse et al. 1999). They yield 3.65-Ga Pb/
46

Fig. 4. Covariation between

La/Yb and Sr/Y ratio in arc
magmas. In the Tonga-Lau ba-
sin there is a gradational in-
crease in both ratios from
typical MORB values to oce-
anic arc values, with precursor
Lau Basin arc lavas from ODP
site 839 plotting midway be-
tween MORB-like basalts and
proper Tongan arc lavas. The
most evolved lavas in these
ratios are those from the
northern end of Tonga where
the slab is being torn and is
shallowing. Typical adakites
have much higher La/Yb and
Sr/Y ratios. Aleutian arc lavas
of ‘adakitic’ affinity are resolv-
ably more evolved in these
ratios than typical oceanic arc
lavas. Tongan data are from
Ewart et al. (1998), and typical
adakites include those quoted in
Fig. 3, plus data for adakitic
rocks from Ecuador (Bourdon
et al. 2002) and Borneo (Prou-
teau et al. 2001)

Mount Isa Block in NE Australia. The Black Angel

gneisses have interpreted U/Pb ion-probe zircon intru-
sion ages of 2.46 to 2.51 Ga and show juvenile Nd-iso-
tope characteristics (McDonald et al. 1997).
When the median trace element concentrations of
these TTGs are normalised to MORB it is evident that,
unlike adakites, they have very pronounced overen-
richment of Pb, Be and Li, and a distinctively sub-
chondritic Nb/Ta ratio (Fig. 7). The Nb/Ta medians of
the Itsaq gneisses (8.81), the type-Amı̂tsoq gneisses
(13.64), and the Black Angel gneisses (12.67) are in good
agreement with bulk continental crust (10). Similarly,
the Itsaq Gneiss median Pb/Nd is 1.77, that of the type-
Amı̂tsoq gneisses 0.75, and that of the Black Angel
gneisses 0.445, which compare well with average conti-
Fig. 5. Common Pb isotope diagram. Adakites from Adak Island nental crust (0.63; Rudnick and Fountain 1995). The
(squares; data from Myers and Marsh 1987) define a binary mixing
line. According to the Pb-isotope evolution model of Kramers and
median Pb/Nd ratio of 0.75 for the type-Amı̂tsoq
Tolstikhin (1997), the more primitive end member represents gneisses is higher than that of bulk continental crust but
modern MORB-source mantle (circles connected by solid line; the uniquely unradiogenic Pb of these gneisses negates
spacing represents 100-Ma time steps). The more radiogenic end the involvement of older crustal material in their genesis
member is represented by modern ‘erosion mix’ (circles connected (Kamber and Moorbath 1998). Rather, the high Pb/Nd
by broken line; spacing represents 100-Ma time steps). Both end
members could be hosted by the subducted slab if the unradiogenic ratio reflects a similar (slab-) dehydration process which
Pb was derived from the oceanic crust and the radiogenic Pb was is responsible for the continuing Pb enrichment of the
derived from subducted sediment. Alternatively, the radiogenic end continents (Miller et al. 1994). Similar trace element
member could represent enriched mantle or assimilated crust patterns were obtained for 3.80-Ga Itsaq TTGs by
Nutman et al. (1999) who concluded that a substantial
Pb, Sm–Nd, and Rb–Sr regression ages and intercepts role of fluids is implied by the chemistry of these tona-
indicative of juvenile origin. They are not part of the lites. The juvenile mantle origin of the Black Angel
sample suite for which some authors claim ages older gneisses is evident from their radiogenic initial Nd iso-
than 3.65 Ga. Thirdly, for a set of five late Archaean topes (initial
Nd between +3.9 and +5.8) which are
tonalitic to granodioritic Black Angel gneisses from the similar to that of coeval depleted mantle (e.g. Nägler and
 47

are higher than in typical continental crust (Black Angel

gneisses=32; type-Amı̂tsoq gneisses=34; 3.80-Ga Itsaq
gneisses=118). Although this indicates that metamor-
phic, fluid-related light-element loss is evidently not an
issue with these samples, the Li/Yb ratio is not a strong
discriminator for Archaean petrogenesis. For example, if
the median Li concentration is divided by the Yb con-
centration of average continental crust, we obtain Li/Yb
ratios between 11 and 15 (instead of 32–118). This shows
that the Li/Yb ratio is largely controlled by HREE de-
pletion and not by addition of fluid-mobile elements.
Nevertheless, Pb/Nd and Nb/Ta ratios clearly show that
none of these TTGs can be considered to be adakites.
The geochemistry of Itsaq, Amı̂tsoq, and Black Angel
gneiss samples is by no means exceptional. Condie
(1993) demonstrated that the La/Yb ratios of TTGs
decreased from 30 in the Archaean to 17.33 in the Pro-
terozoic and 14.7 in the Phanerozoic, but Nb/Ta and Pb/
Nd ratios have remained virtually constant (10.7, 11.25,
12.3, and 0.86, 0.75, 0.82 respectively), in good agree-
ment with the median Nb/Ta ratio (9.82) of 174 Ar-
chaean, high-Al trondhjemite–tonalite–dacites compiled
by Drummond et al. (1996). Subchondritic Nb/Ta and
overproportional Pb concentration are therefore typical
for Archaean TTGs. We conclude that, at least since ca.
3.8 Ga, element transport by aqueous fluids in the
mantle has been an integral part of continent formation.
Smithies (2000) recently argued, based on differences
in major- and compatible-element contents, that TTGs
could not be analogues of adakites but that they may
have formed through melting of hydrous basaltic ma-
terial at the base of the crust. Kramers (1988) and Ridley
and Kramers (1990) also stressed that whatever mode of
TTG formation is envisaged, partial hydration of mantle
melts and fractionation of amphibole appear unavoid-
able. Nutman et al. (1999) noted that the high relative
abundances of Rb, Cs, Sr, Ba, and Pb made an origin
from direct eclogite melting impossible, and pointed to a
Fig. 6a–c. Selected trace element and Pb-isotope systematics of substantial role of fluids in early crust formation. Our
Aleutian lavas. a Although scattered, there is a discernible anti-
correlation between La/Yb and Pb/Nd in the data set, with the results agree with the findings of these authors and we
exception of a single sample (AD-72; Table 1). b Strong positive conclude that the strong LREE–HREE fractionation of
linear correlation (r2=0.7984) exists between the Pb content (in Archaean TTGs must have an other origin than residual
ppm) and the Pb/Nd ratio of the samples from Adak Island (again garnet in the eclogite source of a subducted oceanic
omitting AD-72). No correlation (r2=0.1839) is found between the
Nd content and the Pb/Nd ratio (not shown). c The Pb-isotope crust.
composition of Adak Island lavas is positively correlated with the
Pb content, indicating that the Pb-poor samples have a MORB-like
Pb-isotope composition and that the Pb-rich samples contain Alternative explanation for strong HREE depletion
crustal Pb. The two samples (of twelve) which have the highest Pb
contents (open circles) do not follow the trend. Data are from
of TTGs
Myers and Marsh (1987)
There are two important observations regarding the
HREE depletion of Archaean TTGs which, in our view,
Kramers 1998). The overenrichment in Pb and the have not been sufficiently considered. These concern the
subchondritic Nb/Ta ratio of the Black Angel gneisses average composition of upper continental crust as ap-
are thus not caused by assimilation of pre-existing crust proximated by sedimentary rocks, and the fact that
but are genuine features of continental crust formation. TTGs are, on average, also overly enriched in LREEs.
The median Be/Nd ratios of all three sets of TTGs Condie (1993) convincingly demonstrated that there is a
(Black Angel gneisses=0.073; type-Amı̂tsoq gneis- mismatch between estimates of upper Archaean crust
ses=0.07; 3.80-Ga Itsaq gneisses=0.19) are within the calculated from relative rock abundance and sedimen-
range of typical continental crust. Median Li/Yb ratios tary rock composition. The relative contribution of
48

Table 2. Chemical compositions of TTGs from the Itsaq Gneiss Terrain, Godthabsfjord, Greenland (all trace elements by ICP-MS)

Tectonic unit Itsaq Gneiss Terrain, Greenland

Locality Godthabsfjord, type-locality Amı̂tsoq Gneiss

Sample ID 155702 155708 155710 155724 155736 155739 155768 155774 155807 155810
a
Major element (wt%)
SiO2 66.49 72.93 51.85 80.12 66.67 69.25 75.93 66.90 70.47 65.17
TiO2 0.79 0.17 1.80 0.05 0.99 0.37 0.27 0.31 0.22 0.47
Al2O3 14.38 14.61 14.85 11.08 13.39 13.43 11.73 15.44 16.08 17.47
Fe2O3 5.50 1.62 14.49 0.51 6.68 6.13 2.60 4.39 2.00 3.77
FeO – – – – – – – – – –
MnO 0.07 0.02 0.16 0.01 0.11 0.05 0.04 0.05 0.03 0.05
MgO 0.88 0.34 3.25 0.05 0.98 0.33 0.60 0.94 0.98 1.99
CaO 3.27 1.71 7.69 1.42 3.74 2.09 2.63 3.22 3.54 4.75
Na2O 3.67 4.04 4.18 2.90 3.11 3.09 4.10 5.37 5.59 5.13
K2O 3.67 4.09 0.94 3.58 3.20 4.44 1.20 1.49 1.34 1.26
P2O5 0.14 0.01 0.23 0.00 0.30 0.04 0.12 0.07 0.03 0.05
Trace element (ppm)
Li 26.5 19.2 20.1 10.4 60.4 69.1 24.1 42.5 41.0 40.2
Be 1.50 0.81 1.29 0.91 1.26 0.72 1.88 2.56 1.23 1.26
Sc 19.3 2.75 29.6 0.15 11.6 7.15 0.76 2.93 2.77 9.16
V 57.3 17.9 327 3.10 48.5 14.5 12.6 28.8 19.9 54.1
Cr 14.9 11.9 3.46 0.47 8.27 11.2 1.63 8.90 25.7 44.9
Co 12.7 4.11 48.0 1.15 10.1 5.02 4.05 10.2 5.23 15.0
Ni 7.84 4.36 19.2 0.71 4.23 3.18 3.14 10.3 10.7 41.5
Cu 16.8 2.47 47.2 1.17 16.8 4.15 2.03 3.73 20.7 8.86
Zn 98.0 35.7 141 10.3 88.4 49.7 37.7 74.9 29.8 53.4
Ga 22.2 19.6 24.1 12.0 21.8 19.4 13.6 20.9 17.8 20.8
Rb 87.6 86.3 6.6 93.5 88.0 130 57.0 94.1 50.3 56.1
Sr 193 194 255 80.1 237 108 189 275 520 426
Y 57.0 6.15 26.6 1.71 28.0 28.7 4.79 3.28 4.30 10.5
Zr 572 258 26.0 8.78 259 432 158 145 83.5 128.6
Nb 10.5 1.06 3.78 0.77 7.78 4.62 2.33 2.46 1.37 2.39
Sn 0.86 0.43 1.66 0.29 1.01 0.50 0.64 0.66 0.62 1.28
Cs 0.72 0.69 0.09 0.43 6.68 4.24 1.05 3.86 1.75 1.65
Ba 532 655 173 203 771 317 93.9 173 209 400
La 36.6 15.2 26.0 7.06 29.5 54.3 33.2 8.00 22.9 15.6
Ce 99.3 23.0 58.2 8.17 71.2 106 62.1 15.0 41.2 37.0
Pr 15.3 2.32 7.79 0.64 10.1 12.8 6.94 1.75 4.61 5.04
Nd 68.9 7.83 31.8 1.66 43.4 46.5 22.5 6.25 15.5 20.4
Sm 16.7 1.30 6.66 0.24 8.89 8.22 2.98 1.09 2.37 4.04
Eu 3.16 0.87 1.87 1.32 2.87 1.27 0.74 0.39 0.81 1.11
Gd 15.9 1.14 6.22 0.22 7.79 6.94 1.71 0.90 1.61 3.19
Tb 2.25 0.17 0.91 0.04 1.06 1.00 0.19 0.12 0.19 0.42
Dy 12.7 1.05 5.40 0.29 5.98 5.98 0.96 0.65 0.94 2.28
Ho 2.45 0.23 1.08 0.07 1.15 1.24 0.18 0.13 0.16 0.42
Er 6.14 0.69 2.93 0.21 3.06 3.42 0.47 0.35 0.41 1.09
Tm 0.79 0.11 0.42 0.03 0.41 0.47 0.07 0.05 0.05 0.15
Yb 4.56 0.73 2.56 0.24 2.54 2.85 0.46 0.34 0.33 0.90
Lu 0.65 0.12 0.39 0.04 0.39 0.42 0.07 0.06 0.05 0.13
Hf 15.1 7.44 1.19 0.39 6.65 11.48 4.06 3.75 2.18 3.19
Ta 0.74 0.11 0.28 0.10 0.57 0.22 0.29 0.08 0.10 0.10
Pb 13.8 20.4 9.11 17.4 12.8 17.4 19.5 13.2 18.1 13.0
Th 0.42 0.46 0.16 0.98 0.52 2.96 6.97 0.98 2.82 0.13
U 0.38 0.32 0.02 0.16 0.25 0.26 0.54 0.23 0.17 0.25
a
Major-element data by ICP-OES

TTGs to average Archaean cratonic areas at their pre- (McLennan and Hemming 1992). This suggests that
sent erosion level is ca. 50%, and that of granites is ca. relative lithological contributions to the present erosion
20% (Condie 1993). Both TTGs and granites are level are not at all representative of original upper Ar-
strongly depleted in HREEs and have average Sm/Nd chaean crust. Condie (1993) concluded that typical up-
ratios of 0.155 and 0.137 respectively (Condie 1993). per Archaean crust prior to erosion to its present level
Estimates of the Sm/Nd ratio of Archaean to early was composed of up to 60% mafic lithologies. This
Proterozoic shales and pelites, on the other hand, are finding is of direct relevance to models for the origin of
considerably higher and range between 0.175 (Condie TTGs as it indicates that, on average, the REE com-
1993), 0.188 (Jahn and Condie 1995), and 0.201 position of upper continental crust has not changed with
 49

Table 3. Chemical compositions of TTGs from the Itsaq Gneiss Terrain, unnamed block south of the Isua Belt, Greenland, and the Black
Angel Gneiss, Australia (all trace elements by ICP-MS)

Tectonic 3.8-Ga Itsaq Gneiss Terrain, Greenland 2.5-Ga Black Angel Gneiss, Australia
unit
Locality Unnamed block south of Isua Belt Mt. Isa Block, North Queensland

Sample ID SM/GR/ SM/GR/ SM/GR/ SM/GR/ SM/GR/ GM94–100 GM95–31 GM95–34 GM95–36A GM95–38
98/1 98/2 98/3 98/4 98/34

Major element (wt%)a

SiO2 – – – – – 61.93 63.16 66.88 65.76 64.23
TiO2 – – – – – 1.05 0.90 0.74 0.79 0.93
Al2O3 – – – – – 15.46 16.03 15.97 16.17 16.25
Fe2O3 – – – – – 7.51 6.20 4.43 4.85 5.65
FeO – – – – – 5.32 4.28 2.62 3.14 3.91
MnO – – – – – 0.09 0.07 0.05 0.06 0.06
MgO – – – – – 3.71 2.21 1.32 1.48 1.68
CaO – – – – – 5.09 4.64 4.18 4.24 4.57
Na2O – – – – – 1.95 2.32 2.71 2.79 2.55
K2O – – – – – 2.82 2.49 2.00 2.01 2.11
P2O5 – – – – – 0.36 0.17 0.13 0.15 0.17
Trace element (ppm)
Li 33.0 23.8 30.8 31.0 28.0 23.9 26.5 16.4 18.9 25.8
Be 1.73 1.23 2.33 1.17 1.45 2.18 2.09 2.48 3.07 1.79
Sc 11.9 6.20 2.64 1.87 1.95 19.6 11.8 7.68 7.64 7.70
V 102 45.8 10.1 7.56 19.8 113 82.8 35.9 40.8 47.2
Cr 44.7 23.0 6.65 2.80 286 47.5 43.3 16.5 15.1 19.1
Co 31.6 28.5 26.6 28.4 29.9 71.5 14.9 9.94 10.3 11.7
Ni 51.2 23.2 11.1 8.82 94.8 40.6 10.1 3.06 3.59 4.19
Cu 47.1 9.49 2.84 3.13 3.37 54.6 18.3 5.15 12.8 6.82
Zn 77.9 54.0 24.8 26.1 44.5 115 80.5 33.5 52.5 73.4
Ga 22.4 19.7 19.5 18.6 16.5 22.1 22.3 21.1 22.4 22.6
Rb 47.9 34.7 29.1 20.6 44.9 148 124 98.3 108 102
Sr 452 371 219 258 297 514 431 382 386 467
Y 10.9 7.60 2.49 2.22 1.83 15.1 12.5 5.86 10.1 7.21
Zr 79.3 151 66.5 81.1 89.9 99.0 224 177 233 257
Nb 4.80 3.65 5.74 1.10 3.02 12.4 11.0 10.6 9.70 9.96
Sn 0.37 0.57 0.73 0.53 0.62 – – – – –
Cs 1.89 0.94 3.54 1.45 2.05 5.54 5.43 3.09 4.21 4.74
Ba 207 238 78.4 112 276 927 600 364 443 505
La 16.9 28.4 3.43 7.96 5.62 55.9 46.7 30.4 40.9 45.8
Ce 38.1 50.3 5.50 13.4 9.22 106 85.7 60.0 77.0 84.1
Pr 5.61 6.67 0.72 1.79 1.14 12.5 9.20 7.12 9.24 10.0
Nd 23.1 23.7 2.65 6.22 4.01 42.5 29.0 23.4 29.9 31.8
Sm 4.72 3.78 0.68 1.04 0.70 8.49 4.27 3.61 4.83 4.63
Eu 1.13 0.94 0.34 0.38 0.46 1.79 1.13 1.07 1.10 1.28
Gd 3.59 2.58 0.68 0.70 0.57 5.69 3.57 2.31 3.25 2.97
Tb 0.45 0.32 0.10 0.08 0.07 0.57 0.51 0.28 0.41 0.34
Dy 2.38 1.64 0.53 0.45 0.37 2.97 2.48 1.32 2.11 1.64
Ho 0.43 0.30 0.09 0.08 0.07 0.52 0.55 0.23 0.39 0.28
Er 1.07 0.76 0.21 0.24 0.17 1.59 1.21 0.55 0.97 0.69
Tm 0.15 0.11 0.03 0.04 0.02 0.18 0.14 0.07 0.12 0.08
Yb 0.88 0.64 0.15 0.27 0.15 1.22 0.91 0.43 0.76 0.53
Lu 0.13 0.09 0.02 0.04 0.02 0.17 0.15 0.06 0.11 0.07
Hf 2.34 3.94 1.98 2.42 2.57 2.62 4.76 4.08 5.16 5.10
Ta 0.47 0.35 0.90 0.43 0.34 1.61 0.84 0.82 0.90 0.71
Pb 13.6 9.15 18.7 11.0 14.4 11.4 14.0 13.7 14.4 12.9
Th 3.20 2.82 0.52 1.71 0.37 18.02 8.97 5.80 8.41 7.74
U 0.74 0.17 0.17 0.25 0.23 2.21 1.16 1.34 1.38 0.93
a
Major-element data by XRF

time. This contradicts the view that TTGs are direct slab overly enriched in LREEs. For example, Condie’s (1993)
melts because clearly, substantial addition of strongly average La concentration of both TTGs and granites
HREE-depleted material to ordinary continental crust dropped from 30 and 50 ppm in the Archaean to 25 and
would have lowered the average Sm/Nd ratio. 40 ppm respectively in the Phanerozoic. At the same
Furthermore, the notion that Archaean TTGs have time, average Yb concentrations increased in these rock
high La/Yb because of HREE depletion is not entirely types from 1.0 and 2.0 to 1.7 and 3.2 respectively. The
correct as it also reflects the fact that many TTGs are change in La content thus contributes 19% (TTGs) and
50

the fractionation series are rarely exposed (e.g. Brophy

1989). In cases where discrete intrusive complexes con-
tain mafic to silicic constituents, fractional crystallisa-
tion models are often successful in providing a magmatic
lineage (e.g. Tate et al. 1999) comparable to those de-
rived from experimental constraints (e.g. Grove and
Baker 1984). Archaean TTGs are even more problematic
than Phanerozoic TTGs because post-magmatic tecto-
nism has complicated field relationships between po-
tential co-magmatic phases, and because the igneous
mineralogy or mineral chemistry is rarely preserved.
Nevertheless, some Archaean greenstone belts con-
tain volcanic successions of mafic to very silicic com-
position. One example for which high-quality trace
Fig. 7. N-MORB-normalised (Sun and McDonough 1989) trace element data is available is the Kalgoorlie Terrane (a
element pattern of Archaean TTGs (medians) compared to average greenstone belt in the Yilgarn craton of Australia) which
continental crust estimate (stippled line, no symbols) of Rudnick contains the 2.68-Ga, dominantly felsic volcanic Black
and Fountain (1995). All three types of TTGs (Black Angel gneisses
as open crosses, 3.81-Ga Itsaq gneisses as closed circles, and 3.65 Flag Group (Morris and Wit 1997). Morris and Wit
type-Amı̂tsoq gneisses as open circles) show strong enrichment in (1997, p. 95) postulated that ‘‘in terms of both volume
Pb, Li, and Be and subchondritic Nb/Ta ratio. They also display relationships and chemistry, it is unlikely that Black
strong LREE/HREE fractionation typical of Archaean TTGs. Flag Group basalts were parental to andesites, or were
Elements are arranged in decreasing MORB-melting compatibility,
and normalisation parameters are identical to those used in Fig. 1 derived by melting of the same source ...’’ and they
stated that the basalts were ‘‘... not considered further in
this paper’’. However, the first argument appears ques-
25% (granites) to the resulting decrease in La/Yb ratio. tionable in view of Condie’s (1993) observation regard-
This observation is illustrated by the REE patterns of ing shale and pelite geochemistry. Morris and Wit’s
the type-Amı̂tsoq gneisses and the Black Angel gneisses (1997) rejection of consanguinity of the Black Flag
(Fig. 7), in that they are more enriched in LREEs than Group volcanics thus depends on their interpretation of
average continental crust. In order for the average La/ geochemistry. They refer to their Fig. 2d which is a plot
Yb ratio of cratonic shales to have remained constant of SiO2 versus Cr concentration (Morris and Wit 1997;
(Condie 1993), a higher proportion of low La/Yb mag- p. 93) showing that some of their andesites have anom-
mas must have been present in the Archaean upper alously high Cr contents. They implicitly conclude that if
crust. In other words, at least in terms of REEs, the bulk basalts were parental, they are expected to have even
of material which formed continental crust has not sig- higher Cr contents, which they do not. However, other
nificantly changed through time. This is also true for the compatible elements in these rocks do not seem to share
major elements and the majority of incompatible ele- this feature. For example, there is a strong negative
ments. Only the abundances of compatible elements correlation between TiO2 and SiO2 (Fig. 2a of Morris
have decreased with time in average upper continental and Wit 1997) for the entire volcanic sequence (including
crust, due to the decline in mafic magmas reaching the the basalts), and Ni contents show a good positive cor-
surface. It is therefore plausible that it was the intensity relation with MgO (r2=0.737; not shown by Morris and
and mode of intracrustal differentiation which changed Wit 1997). Furthermore, if the Black Flag volcanics are
over time, rather than the composition of the parental plotted in Le Maı̂tre et al.’s (1989) classification diagram
melt itself (and consequently the processes which form (Fig. 8a), they yield a well-defined subalkaline trend
that melt). indicative of differentiation. The most evolved samples
Theoretically, the strong HREE depletion of Archa- require removal of between 70–80% solids. Interesting-
ean TTGs (e.g. Fig. 7) could indicate significant frac- ly, the samples also yield a very well-defined negative
tional crystallisation of a mineral with DYbmineral/ correlation in a plot of total Fe versus La/Yb (Fig. 8b),
melt
>>DLamineral/melt during magma differentiation. which could indicate fractional crystallisation of mine-
Between 80 and 90% of fractional crystallisation is re- ral with DHREEmineral/melt>>DLREEmineral/melt, notably
quired to produce residual TTGs from a basaltic parent amphibole (e.g. Arth and Barker 1976; Klein et al.
melt. Drummond et al. (1996, p. 206) regarded such 1997). Morris and Wit (1997) proposed that the basalts
degrees of fractional crystallisation as unlikely because show no clear relationship to the more felsic volcanics in
of ‘‘the lack of high proportions of mafic to intermediate terms of REEs. However, their Fig. 3b (Morris and Wit
igneous rocks’’. However, shale and pelite geochemistry 1997, p. 94) clearly shows that the basalts have lower
shows that the lack of mafic and intermediate rocks is LREE but higher HREE contents than andesites, which
only apparent, and that the ancient exposed crust was in is what would be expected if they were related by frac-
fact dominated by such rocks. It is difficult to quanti- tional crystallisation of amphibole. The REE systemat-
tatively demonstrate the origin of tonalitic rocks by ics of the Black Flag volcanics demonstrate that the
fractional crystallisation because the mafic members of degree of HREE, and even MREE depletion critically
 51
b
Fig. 8. a Volcanic rock classification diagram after Le Maı̂tre et al.
(1989). Black Flag Group volcanics (squares; data from Morris and
Witt 1997) form a succession from (picro-)basalt (field I), basaltic
andesite (field II), andesite (field III), dacite (field IV) to
trachydacite (field V) and rhyolite (field VI). The tight trend
defined by all but one samples represents a typical subakaline
differentiation series. Plotted for comparison are the estimates for
average TTG by Condie (1993; open circle) and by Drummond et al.
(1996; closed circle). b Plot of total Fe (as Fe2O3 wt%) versus La/
Yb ratio of 2.7-Ga Black Flag Group volcanics (same data as in a).
Strong correlation indicates that HREE depletion could be caused
by fractional crystallisation of hornblende and/or clinopyroxene.
c Plot of La/Yb versus Yb content (scale, fields and normalisation
as in Fig. 3). Plotted are the same samples as in a: shaded field
Archaean TTGs, closed circles basalts, open circles dacites, and
dotted open circles rhyolites. Shown for comparison is a fractional
crystallisation trajectory (solid curve) calculated for removal of
mixture of amphibole (75%) and garnet (25%), starting from the
most evolved basalt sample data point. Open circles along this
trajectory are spaced at intervals of 5% remaining liquid from 100
to 25%. Distribution coefficients are from Zack et al. (1997)

differentiation models, the fact remains that rocks of

tonalitic composition can be generated from a mafic
parental magma. Trace element geochemistry of TTGs
has changed over geological time but we believe that
differences have been overplayed in the past.
It is not our aim to discuss in detail how mechanisms
which drive fractional crystallisation could have chan-
ged over time, if only for the fact that surprisingly little
is known about the effects of these processes on Phan-
erozoic magmas. Primary differentiation of modern arc
magmas starts within the mantle wedge, probably at
depths where rising melt-matrix domes encounter the
steep geothermal gradient which is maintained under
the lithosphere by wedge convection (e.g. Tatsumi and
Eggins 1995). It would thus appear that the thermal
structure in the mantle wedge and across the base of the
lithosphere could have an effect on how strongly pri-
mary arc magmas fractionated. Differentiation (and as-
similation) also occurs at the base of and within the
crust, especially in continental arcs. It appears that
structure and thickness of the crust influence the degree
of magma fractionation (Ewart 1979). Furthermore, the
chemistry and oxidation stage of the primary magma
also affect the sequence of phases which crystallise (e.g.
Grove and Baker 1984). Müntener et al. (2001) have
presented evidence that the initial water content of the
basaltic melt within the mantle wedge has a surprisingly
strong influence on the sequence of fractional crystalli-
depends on how strongly differentiated the samples are sation. At water-saturated conditions, garnet and am-
(Fig. 8b). In the La/Yb versus Yb concentration dia- phibole, instead of olivine and plagioclase, crystallise
gram (Fig. 8c) of Martin (1986), the basalts plot into the together with clino- and orthopyroxene. Both amphibole
field of post-Archaean granitoids, dacites and rhyolites and garnet strongly fractionate REEs in the direction
plot into the field of Archaean TTGs whereas the required by TTG. Importantly, Müntener et al. (2001)
andesites plot mainly into the field of overlap. When showed that the garnet-bearing cumulate is denser than
compared to a calculated fractional crystallisation tra- the surrounding mantle and could thus have been recy-
jectory (75% amphibole and 25% garnet), it appears cled, alleviating the mass balance problem of obtaining
that the earlier stages of fractional crystallisation re- continental crust of, on average, andesitic composition
quired more garnet removal, and that the later stages from basaltic parental melt. Finally, for melts which
could have been dominated by amphibole. We therefore were not fully water-saturated, the relative importance
conclude that in spite of the general scepticism towards of cooling vs. ascent to drive crystallisation of magma
52

has to be considered. Blundy and Cashman (2001) re- heat is lost by production and cooling of oceanic plates
cently showed that highly silicic melts can be produced (Stacey 1980), the hypothesis of a very hot Archaean
only at shallow depth. It is likely that all the factors mantle would actually imply a reduced oceanic plate
involved in these processes have changed over time, production rate and hence no preponderance of hot and
possibly in response to cooling of the Earth. Thus, it is young subducted plates. If, on the other hand, it is ac-
premature the reject a TTG origin by differentiation cepted that the cooling rate of the mantle is some
from a basaltic precursor. function of its convection pattern (e.g. Stacey 1980), the
particular temperature of the upper mantle at any given
time is still ill constrained.
Discussion The second aspect that the slab melting model does
not consider is that the ridge driving force, which ulti-
Trace element chemistry of Archaean TTGs requires mately determines the plate generation rate, is largely a
differential transport of fluid-mobile elements to the function of the thermal structure of the oceanic plates.
mantle melt source. The only feasible mechanism by Thus, the cooler the average subducted plate, the higher
which significant volumes of water can be transported the ridge driving force and hence the plate production
sufficiently deep into the mantle is by emplacement of rate, and vice versa. Parsons (1982) concluded from this
relatively dense, cold and hydrated oceanic lithosphere. interrelationship that fluctuations in the areal rate of
Hydrated oceanic lithosphere must have existed since the generation and destruction of oceanic plates would be
establishment of a liquid hydrosphere (dating back to at limited narrowly about a basic state.
least 3.65 Ga; Whitehouse et al. 1999) but this in itself does Thirdly, if the potential upper mantle temperature
not imply operation of subduction in the present form was indeed higher in the Archaean, its effect on the
since 3.65 Ga. However, it seems difficult to transport thickness of oceanic lithosphere produced along
water into the deeper mantle in the absence of some form spreading ridges has to be considered also (McKenzie
of subduction. Because heat loss across the surface and Bickle 1988; Galer 1991). In the following we ex-
boundary layer of a planet without subduction is re- plore the extent to which thicker Archaean oceanic crust
stricted to conduction through a (thick) stable oceanic could compensate the increase in spreading rate (and
lithosphere, an unstable boundary layer would form be- hence decrease in mean age of oceanic lithosphere) re-
low the lithosphere in the uppermost astenosphere quired by considerations of heat dissipation.
(Campbell and Griffiths 1993). Instability from this As noted above, the least constrained parameters are
boundary layer would lead to sinking of cold plumes of the secular change in potential upper mantle tempera-
depleted mantle material but not hydrated crust. We ture and the oceanic crust production rate through time.
therefore agree with Campbell and Griffiths (1993) that For the purpose of testing the direct slab melting hy-
the birth of subduction is marked by creation and pres- pothesis, it is necessary and reasonable to propose that
ervation of the first continental lithosphere at ca. 3.8 Ga. oceanic crust production rate in the Archaean was
The hypothesis that Archaean TTGs represent direct higher than at present. Kramers and Tolstikhin (1997)
slab melts implicitly assumes operation of subduction. adopted an oceanic crust production rate (Ø) which is
The prevalence for slab melting in the Archaean was sufficient to account for post-accretion degassing:
logically linked to the cooling history of the Earth (e.g.
Ø ¼ 33e0:37t  1016 ga1 ð1Þ
Martin 1986). In short, a hotter Archaean Earth dissi-
pated more heat via oceanic crust production, which in where age (t) is time in Ma. This function is displayed in
turn implies higher spreading and subduction rates. Fig. 9a and yields a realistic present-day oceanic crust
Thus, it was hypothesised that on average Archaean production rate of 6.1·1016 g a–1 (Reymer and Schubert
oceanic lithosphere was younger and hotter when sub- 1984). The following calculations clearly depend on the
ducted, such that direct slab melting was more feasible. exact shape of this function but we reiterate that the
Although this logic appears intuitively correct, it ignores purpose of this calculation is to test the hypothesis of
a number of physical constraints on the relationship TTG origin by direct slab melting, in which a higher
between terrestrial cooling and oceanic plate production oceanic crust production rate is implicit. In the next step,
(Galer 1991). we estimate potential shallow mantle temperature
Firstly, it is undisputed that radiogenic heat pro- (TPSM). Galer and Mezger (1998) have given a compre-
duction has declined with time, but the potential upper hensive review of estimated cooling rates for the Earth.
mantle temperature is equally dependent on the cumu- Using a present-day TPSM of 1,350 C and a mean time-
lative heat loss which the Archaean Earth has ex- averaged cooling rate of 45 C Ga–1, we obtain a TPSM
perienced. Hence, although assumed implicitly in the of 1,555 C at time zero. We describe the temporal
slab-melting model, the assumption of a hotter upper evolution of TPSM with a slightly exponential decay
Archaean mantle has not been explicitly quantified. For function (Fig. 9b):
example, some petrologists argue that the Archaean
TPSM ¼ 1; 555  C  e0:031t ð2Þ
mantle was substantially hotter than it is today, based
on petrological arguments relating to anhydrous mantle which gives a present-day TPSM of 1,350 C. We do not
melting to yield komatiite. Since virtually all terrestrial regard this function to be valid for the Hadean era
 53

column (and what thickness the resulting crust attains)

but a general decrease of oceanic crustal thickness with
time seems unavoidable.
The rate of ocean floor production (i.e. spreading
rate) can then be obtained by division of the crust pro-
duction rate (Ø) by the crust thickness (h) for any given
mean density of oceanic crust. For a density of 2.85 g cm–3
we obtain a present rate of ocean floor production of
ca. 3.0 km2 a–1 which, by definition, is very similar to the
estimate of Reymer and Schubert (1984), itself an input
parameter for Eq. (1). The somewhat surprising finding
for the past is that spreading rate has only varied be-
tween ca. 3.0 and 4.5 km2 a–1 through geological time.
These calculations show that oceanic crustal thickness
has to be considered when attempting to estimate past
spreading rates. Thus, we disagree with Martin (1986)
and subsequent proponents that spreading rates were
two to three times higher during the Archaean.
The real issue, however, is to estimate the mean age of
oceanic lithosphere prior to subduction. In view of the
unconstrained length of active margins and ridges in the
past, we assume an equilibrium between oceanic crust
production and destruction (Parsons 1982). We there-
fore estimate the mean age of oceanic lithosphere by
calculating the time it would take to cover half the area
occupied by oceanic basins (T1/2OB), with new oceanic
lithosphere created at the spreading rate shown in
Fig. 9d. For the present situation this estimate predicts a
mean age of oceanic crust of 59 Ma, which is in good
agreement with independent estimates. In order to per-
form this calculation for the past, we first have to esti-
mate the temporal change in area covered by ocean
basins. We used the continental mass versus time curve
of Collerson and Kamber (1999) and a present-day area
occupied by continents and continental margins of
1.48·108 km2. Our estimate (Fig. 9e) assumes constant
continental thickness. If Archaean continents were
thicker (as argued by Galer and Mezger 1998), we would
slightly underestimate the area of ocean basins. Re-
gardless, because Archaean ocean basins covered a
Fig. 9a–f. Time series of calculated a oceanic crust production larger area, the somewhat higher spreading rates are
rate, b potential shallow mantle temperature, c thickness of oceanic
crust, d area of annually generated oceanic crust, e area covered by almost compensated and the mean age of oceanic crust
oceanic lithosphere, and f mean age of oceanic crust (for details of has only varied slightly with time (Fig. 9f). This finding
calculation see text) appears to confirm Parsons’ (1982) prediction that the
plate driving force has not varied strongly with time.
Therefore, if the Archaean mantle was indeed hotter, it
during which subduction may not have operated. Oce- would not imply that at the time of subduction, oceanic
anic crustal thickness (h) can be calculated as a function lithosphere was significantly hotter than it is today.
of temperature (TPSM) using isentropic melting models. Rather, it appears that the most likely mechanism to
In combination with Eq. (2), we obtain oceanic crustal accommodate higher Archaean heat loss was by creation
thickness as a function of time (t). In this calculation we of a thicker oceanic plate.
adopted the results of isentropic batch melting models These basic considerations are too simplistic and the
by Iwamori et al. (1995), using an entropy change (DS) parameters derived in Fig. 9 should not be used as an
of 350 J kg–1 K–1. It is important to note that relative approximation to reality, yet the calculations demon-
change of thickness predicted by any isentropic melting strate that the superficially logical link between pro-
model (Fig. 9c) is much greater than the relative change posed Archaean slab melting and hotter Archaean
in TPSM because of the exponential increase of melt oceanic lithosphere is, in fact, flawed. We emphasise
fraction with TPSM. There are still many uncertainties as that the true issue which governed the thermal struc-
to how efficiently melt is extracted from the mantle ture of subducted Archaean oceanic plates is the
54

question whether thick plates (such as those used in

this calculation) could at all be subducted. Continental References
crust volume versus time curves certainly document
Aguillón-Robles A, Calmus T, Benoit M, Bellon H, Maury RC,
great variability in net crustal growth rates throughout Cotten J, Bourgois J, Michaud F (2001) Late Miocene adakites
the Archaean (e.g. Kramers and Tolstikhin 1997; and Nb-enriched basalts from Vizcaion Peninsula, Mexico:
Collerson and Kamber 1999) which may indicate that Indicators of East Pacific Rise subduction below southern Baja
efficient subduction in the modern sense only became California? Geology 29:531–534
feasible beyond a certain maximum thickness of oce- Arth JG, Barker F (1976) Rare-earth partitioning between horn-
blende and dacitic liquid and implications for genesis of tron-
anic plates. dhjemitic-tonalitic magmas. Geology 4:534–546
Ayers J (1998) l modeling of aqueous fluid – peridotite interaction
in the mantle wedge of subduction zones. Contrib Mineral
Petrol 132:390–404
Conclusion Bau M (1996) Controls on the fractionation of isovalent trace el-
ements in magmatic and aqueous systems – evidence from Y/
A review of Phanerozoic adakite trace element geo- Ho, Zr/Hf, and lanthanide tetrad effect. Contrib Mineral Petrol
chemistry confirms that these melts contain a compo- 123:323–333
Becker H, Jochum KP, Carlson RW (2000) Trace element frac-
nent which is derived by direct melting of subducted tionation during dehydration of eclogites from high-pressure
slabs but pure slab melts do not appear to exist. Nev- terranes and the implications for element fluxes in subduction
ertheless, rocks with adakite affinity can be distinguished zones. Chem Geol 163:65–99
from mantle wedge-derived arc magmas by element ra- Blundy J, Cashman K (2001) Ascent-driven crystallisation of dacite
magmas at Mount St Helens, 1980–1986. Contrib Mineral
tios which are most strongly fractionated by refertilisa- Petrol 140:631–650
tion of the depleted mantle wedge (i.e. reflecting element Bourdon E, Eissen JP, Monzier M, Robin C, Martin H, Cotten J,
fractionation during slab dehydration). Using these fin- Hall ML (2002) Adakite-like lavas from Antisana volcano
gerprints, we conclude that Archaean TTGs are melting (Ecuador): evidence for slab melt metasomatism beneath the
Andean northern volcanic zone. J Petrol 43:199–217
products from the supra-subduction zone mantle wedge. Brenan JM, Shaw HF, Ryerson FJ (1995a) Experimental evidence
TTGs thus inherit typical continental element ratios for the origin of lead enrichment in convergent-margin magmas.
from the refertilised mantle wedge, much in the same Nature 378:54–56
way as modern arc magmas. Brenan JM, Shaw HF, Ryerson FJ, Phinney DL (1995b) Mineral-
A key feature of generation of continental crust is aqueous fluid partitioning of trace elements at 900 C and
2.0 GPa: constraints on the trace element chemistry of mantle
dehydration-related element transport in the mantle. and deep crustal fluids. Geochim Cosmochim Acta 59:3331–
Preferential element transport by fluids requires dehy- 3350
dration due to pressure-induced phase changes, which in Brenan JM, Ryerson FJ, Shaw HF (1998a) The role of aqueous
turn requires environments of low geothermal gradients fluids in the slab-to-mantle transfer of boron, beryllium, and
lithium during subduction: experiments and models. Geochim
in the mantle. It is difficult to imagine tectonic envi- Cosmochim Acta 62:3337–3347
ronments other than subduction zones in which rela- Brenan JM, Neroda E, Lundstrom CC, Shaw HF, Ryerson FJ,
tively cold material is deeply entrained into the mantle Phinney DL (1998b) Behaviour of boron, beryllium, and lithi-
such that dehydration occurs before direct melting. The um during melting and crystallisation: constraints from min-
eral-melt partitioning experiments. Geochim Cosmochim Acta
hypothesis that Archaean TTGs are adakites is not 62:2129–2141
supported by in-depth study of their trace element geo- Brophy JG (1989) Can high-alumina arc basalt be derived from
chemistry nor can it be expected from the Earth’s low-alumina arc basalt? Evidence from Kanaga Island, Aleu-
cooling history. tian Arc, Alaska. Geology 17:333–336
Evidence for a secular decrease in HREE depletion in Brown L, Klein J, Middleton R, Selwyn Sacks I, Tera F (1982) 10Be
in island-arc volcanoes and implications for subduction. Nature
arc-related, highly silicic magmas remains strong but 299:718–720
cannot be interpreted to indicate a prevalence of direct Bruce MC, Niu Y (2000a) Early Permian supra-subduction
slab melting in the Archaean. REE geochemistry of assemblage of the South Island terrane, Percy Isles, New
pelites and shales demonstrates that the degree of HREE England Fold Belt, Queensland. Aust J Earth Sci 47:1077–
1085
depletion of the bulk of the material extracted along arcs Bruce MC, Niu Y (2000b) Evidence for Palaeozoic magmatism
has remained constant through time (Condie 1993). It recorded in the Late Neoproterozoic Marlborough ophiolite,
thus appears that the stronger HREE depletion of an- New England Fold Belt, central Queensland. Aust J Earth Sci
cient arc magmas simply reflects more efficient crustal 47:1065–1076
Campbell IH, Griffiths RW (1993) The evolution of the mantle’s
differentiation. chemical-structure. Lithos 30:389–399
Chan LH, Leeman WP, You CF (1999) Lithium isotopic compo-
Acknowledgements Journal reviewers J. Ryan and K. Condie sig- sition of Central American volcanic arc lavas: Implications for
nificantly improved this manuscript. S.J.G. Galer is thanked for modification of the sub-arc mantle by slab-derived fluids. Chem
discussing oceanic plate production rates. Any remaining short- Geol 160:255–280
comings of this paper are entirely the authors’ responsibility. Ste- Collerson KD, Kamber BS (1999) Evolution of the continents and
phen Moorbath kindly provided the Greenland tonalite samples. the atmosphere inferred from Th-U-Nb systematics of the de-
This research was funded by the Swiss National Science Founda- pleted mantle. Science 283:1519–1522
tion grant 8220-050352 to BSK, and by an ARC Large Grant Condie KC (1993) Chemical composition and evolution of the
(#A00104754) to KDC. Our trace element and isotope facility is upper continental crust: Contrasting results from surface sam-
supported by UQ. ples and shales. Chem Geol 104:1–37
 55

Currie KL, Whalen JB, Davis WJ, Longstaffe FJ, Cousens BL Kramers JD (1988) An open-system fractional crystallization
(1998) Geochemical evolution of peraluminous plutons in model for very early continental crust formation. Precambrian
southern Nova Scotia, Canada – a pegmatite-poor suite. Lithos Res 38:281–295
44:117–140 Kramers JD, Tolstikhin IN (1997) Two terrestrial lead isotope
Defant MJ, Drummond MS (1990) Derivation of some modern arc paradoxes, forward transport modelling, core formation and
magmas by melting of young subducted lithosphere. Nature the history of the continental crust. Chem Geol 139:75–110
347:662–665 Le Maı̂tre RW et al. (1989) A classification of igneous rocks and
Defant MJ, Clark LF, Steward RH, Drummond MS, de Boer JZ, glossary of terms. Blackwell, Oxford
Maury RC, Bellon H, Jackson TE, Restrepo JF (1991a) Marsh BD (1976) Some Aleutian andesites: their nature and
Andesite and dacite genesis via contrasting processes: the ge- source. J Geol 84:27–45
ology and geochemistry of El Valle Volcano, Panama. Contrib Marsh BD, Leitz RE (1978) Geology of Amak Island, Aleutian
Mineral Petrol 106:309–324 Islands, Alaska. J Geol 87:715–723
Defant MJ, Richerson PM, de Boer JZ, Steward RH, Maury R, Martin H (1986) Effect of steeper Archaean geothermal gradient
Bellon H, Drummond MS, Feigenson MD, Jackson TE (1991b) on geochemistry of subduction zone magmas. Geology 14:753–
Dacite genesis via both slab melting and differentiation: Pet- 756
rogenesis of La Yeguada Volcanic Complex, Panama. J Petrol McDonald GD, Collerson KD, Kinny PD (1997) Late Archaean
32:1101–1142 and early Proterozoic crustal evolution of the Mount Isa Block,
Drummond MS, Defant MJ, Kepezhinskas PK (1996) Petrogenesis northwestern Queensland, Australia. Geology 25:1095–1099
of slab-derived trondhjemite-tonalite-dacite/adakite magmas. McDonough WF, Sun SS (1995) The composition of the Earth.
Trans R Soc Edinburgh Earth Sci 87:205–215 Chem Geol 120:223–253
Ewart A (1979) A review of the mineralogy and chemistry of McKenzie D, Bickle MJ (1988) The volume and composition of
Tertiary-recent dacitic, latitic, rhyolitic, and related salic vol- melt generated by extension of the lithosphere. J Petrol 29:625–
canic rocks. In: Barker F (ed) Trondhjemites, dacites, and re- 679
lated igneous rocks. Elsevier, New York, pp 13–122 McLennan SM (2001) Relationships between the trace element
Ewart A, Collerson KD, Regelous M, Wendt JI, Niu Y (1998) composition of sedimentary rocks and upper continental crust.
Geochemical evolution within the Tonga-Kermadec-Lau arc- G-cubed 2:2000GC000109
back-arc systems: the role of varying mantle wedge composition McLennan SM, Hemming S (1992) Sm/Nd elemental and isotopic
in space and time. J Petrol 39:331–368 systematics in sedimentary rocks. Geochim Cosmochim Acta
Galer SJG (1991) Interrelationships between continental freeboard, 56:887–898
tectonics, and mantle temperature. Earth Planet Sci Lett Miller DM, Goldstein SL, Langmuir CH (1994) Cerium/lead and
105:214–228 lead isotope ratios in arc magmas and the enrichment of lead in
Galer SJG, Mezger K (1998) Metamorphism, denudation and sea the continents. Nature 368:514–520
level in the Archaean and cooling of the Earth. Precambrian Morris PA, Witt WK (1997) Geochemistry and tectonic setting of
Res 92:389–412 two contrasting Archaean felsic volcanic associations in the
Green TH (1995) Significance of Nb/Ta as an indicator of geo- Eastern Goldfields, Western Australia. Precambrian Res 83:83–
chemical processes in the crust-mantle system. Chem Geol 107
120:347–359 Münker C (1998) Nb/Ta fractionation in a Cambrian arc/back arc
Grove TL, Baker MB (1984) Phase equilibrium controls on the system, New Zealand: source constraints and application of
tholeiitic versus calc-alkaline differentiation trends. J Geophys refined ICPMS techniques. Chem Geol 144:23–45
Res 89:3253–3274 Müntener O, Kelemen PB, Grove TL (2001) The role of H2O
Gutscher MA, Maury R, Eissen JP, Bourdon E (2000) Can slab during crystallisation of primitive arc magmas under uppermost
melting be caused by flat subduction? Geology 28:535–538 mantle conditions and genesis of igneous pyroxenites: an ex-
Henry C, Burkhard M, Goffé B (1996) Evolution of synmeta- perimental study. Contrib Mineral Petrol 141:643–658
morphic veins and their wallrocks through a Western Alps Myers JD, Marsh BD (1987) Aleutian lead isotopic data: addi-
transect: no evidence for large-scale fluid flow. Stable isotope, tional evidence for the evolution of lithospheric plumbing sys-
major- and trace element systematics. Chem Geol 127:81–109 tems. Geochim Cosmochim Acta 51:1833–1842
Hofmann AW (1988) Chemical differentiation of the Earth: the Nägler TF, Kramers JD (1998) Nd isotopic evolution of the upper
relationship between mantle, continental crust, and oceanic mantle during the Precambrian: Models, data and the uncer-
crust. Earth Planet Sci Lett 90:297–314 tainty of both. Precambrian Res 91:233–252
Iwamori H, McKenzie D, Takahashi E (1995) Melt generation by Nutman AP, Bennett VC, Friend CRL, Norman MD (1999) Meta-
isentropic upwelling. Earth Planet Sci Lett 134:253–266 igneous (non-gneissic) tonalites and quartz-diorites from an
Jahn B-M, Condie KC (1995) Evolution of the Kaapvaal Craton as extensive ca. 3800 Ma terrain south of the Isua supracrustal
viewed from geochemical and Sm-Nd isotopic analyses of in- belt, southern West Greenland: constraints on early crust for-
tracratonic pelites. Geochim Cosmochim Acta 59:2239–2258 mation. Contrib Mineral Petrol 137:364–388
Jochum KP, McDonough WF, Palme H, Spettel B (1989) Com- Parsons B (1982) Causes and consequences of the relation between
positional constraints on the continental lithospheric mantle area and age of the ocean floor. J Geophys Res 87:289–302
from trace elements in spinel peridotite xenoliths. Nature Peate DW, Pearce JA, Hawkesworth CJ, Colley H, Edwards CMH,
340:548–550 Hirose K (1997) Geochemical variations in Vanuatu arc lavas:
Kamber BS, Collerson KD (2000) Role of ‘hidden’ deeply sub- the role of subducted material and a variable mantle wedge
ducted slabs in mantle depletion. Chem Geol 166:241–254 composition. J Petrol 38:1331–1358
Kamber BS, Moorbath S (1998) Initial Pb of the Amı̂tsoq gneiss Prouteau G, Maury RC, Sajona FG, Cotten J, Cotten JL (2000)
revisited: implication for the timing of early Archaean crustal Behavior of niobium, tantalum and other high field strength
evolution in West Greenland. Chem Geol 150:19–41 elements in adakites and related lavas from The Philippines.
Kay RW (1978) Aleutian magnesian andesites: melts from sub- Island Arc 9:487–498
ducted Pacific Ocean crust. J Volcanol Geotherm Res 4:117–132 Prouteau G, Maury RC, Pubellier M, Cotten J, Bellon H (2001) Le
Keppler H (1996) Constraints from partitioning experiments on magmatisme post-collisionel du Nord-Ouest de Bornéo, produit
the composition of subduction-zone fluids. Nature 380:237– de la fusion d’un fragment de croûte océanique ancré dans le
240 manteau supérieur. Bull Soc Géol Fr 172:319–332
Klein M, Stosh H-G, Seck HA (1997) Partitioning of high field- Reay A, Parkinson D (1997) Adakites from Solander Island, New
strength and rare-earth elements between amphibole and Zealand. N Z J Geol Geophys 40:121–126
quartz-dioritic to tonalitic melts: an experimental study. Chem Reymer A, Schubert G (1984) Phanerozoic addition rates to the
Geol 138:257–271 continental crust and crustal growth. Tectonics 3:63–77
56

Ridley JR, Kramers JD (1990) The evolution and tectonic conse- and processes. In: Saunders AD, Norry MJ (eds) Magmatism in
quences of a tonalitic magma layer within Archean continents. ocean basins. Geol Soc Lond Spec Publ, pp 313–345
Can J Earth Sci 27:219–228 Tate MC, Norman MD, Johnson SE, Fanning CM, Anderson JL
Romick JD, Kay SM, Kay RW (1992) The influence of amphibole (1999) Generation of tonalite and trondhjemite by subvolcanic
fractionation on the evolution of calc-alkaline andesite and fractionation and partial melting in the Zarza intrusive com-
dacite tephra from the central Aleutians, Alaska. Contrib plex, Western Peninsular Ranges Batholith, Northwestern
Mineral Petrol 112:101–118 Mexico. J Petrol 40:983–1010
Rudnick RL, Fountain DM (1995) Nature and composition of the Tatsumi Y, Eggins S (1995) Subduction zone magmatism. Black-
continental crust: a lower crustal perspective. Rev Geophys well, Cambridge, USA
33:267–309 Tatsumi Y, Kogiso T (1997) Trace element transport during de-
Ryan JG, Langmuir CH (1988) Beryllium systematics in young hydration processes in the subducted oceanic crust. 2. Origin of
volcanic rocks: implications for 10Be. Geochim Cosmochim chemical and physical characteristics in arc magmatism. Earth
Acta 52:237–244 Planet Sci Lett 148:207–221
Ryan JG, Langmuir CH (1993) The systematics of boron abun- Tomascak PB, Ryan JG, Defant MJ (2000) Lithium isotope evi-
dances in young volcanic rocks. Geochim Cosmochim Acta dence for light element decoupling in the Panama subarc
51:1727–1741 mantle. Geology 28:507–510
Ryan JG, Morris J, Tera F, Leeman DH, Tsvetkov A (1995) Cross- von Drach V, Marsh BD, Wasserburg GJ (1986) Nd and Sr iso-
arc geochemical variations in the Kurile arc as a function of topes in the Aleutians: multicomponent parenthood of island-
slab depth. Science 270:625–627 arc magmas. Contrib Mineral Petrol 92:13–34
Schmidt MW, Poli S (1998) Experimentally based water budgets Whitehouse M, Kamber BS, Moorbath S (1999) Age significance of
for dehydrating slabs and consequences for arc magma gener- ion-microprobe U-Th-Pb zircon data from early Archaean
ation. Earth Planet Sci Lett 163:361–379 rocks of West Greenland – a reassessment based on new com-
Shinjo R (1999) Geochemistry of high Mg andesites and the tec- bined ion-microprobe and imaging studies. Chem Geol
tonic evolution of the Okinawa Trough Ryukyu arc system. 160:204–221
Chem Geol 157:69–88 Woodhead JD, Eggins SM, Johnson RW (1998) Magma genesis in
Smith DR, Leeman WP (1987) Petrogenesis of Mount St. Helens the New Britain Island arc: further insights into melting and
dacitic magmas. J Geophys Res 92:10,313–10,334 mass transfer processes. J Petrol 39:1641–1668
Smithies RH (2000) The Archaean tonalite-trondhjemite-granodi- Yogodzinski GM, Kay RW, Volynets ON, Kolosov AV, Kay SM
orite (TTG) series is not an analogue of Cenozoic adakite. (1995) Magnesian andesite in the western Aleutian Koman-
Earth Planet Sci Lett 182:115–125 dorsky region: implications for slab melting and processes in the
Stacey FD (1980) The cooling Earth: a reappraisal. Phys Earth mantle wedge. GSA Bull 107:505–519
Planet Inter 22:89–96 Yogodzinski GM, Lees JM, Churikova TG, Dorendorf F, Woerner
Stalder R, Foley SF, Brey GP, Horn I (1998) Mineral-aqueous fluid G, Volynets ON (2001) Geochemical evidence for the melting of
partitioning of trace elements at 900–1200 C and 3.0–5.7 GPa: subducting oceanic lithosphere at plate edges. Nature 409:500–
new experimental data for garnet, clinopyroxene, and rutile, 504
and implications for mantle metasomatism. Geochim Cosmo- Yumul GP, Dimalanta CB, Bellon H, Faustino DV, De Jesus JV,
chim Acta 62:1781–1801 Tamayo RA, Jumawan FT (2000) Adakitic lavas in the Central
Stern CR, Kilian R (1996) Role of the subducted slab, mantle Luzon back-arc region, Philippines: lower crust partial melting
wedge and continental crust in the generation of adakites from products? Island Arc 9:499–512
the Andean Austral Volcanic Zone. Contrib Mineral Petrol Zack T, Foley SF, Jenner GA (1997) A consistent partition co-
123:263–281 efficient set for clinopyroxene, amphibole and garnet from
Sun S-S, McDonough WF (1989) Chemical and isotopic syste- laser ablation microprobe analysis of garnet pyroxenites from
matics of oceanic basalts: implications for mantle composition Kakanui, New Zealand. N Jahrb Miner Abh 172:23–41