Professional Documents
Culture Documents
Sun & McDonough (1989) PDF
Sun & McDonough (1989) PDF
Notes
SUMMARY: Trace-element data for mid-ocean ridge basalts (MORBs) and ocean
island basalts (OIB) are used to formulate chemical systematics for oceanic basalts. The
data suggest that the order of trace-element incompatibility in oceanic basalts is Cs ~ Rb
(-~ Tl) = Ba(= W) > Th > U ~ Nb = Ta ~ K > La > Ce = Pb > Pr ( ~ Mo) ~- Sr >
P --~ Nd (> F) > Zr = Hf = Sm > Eu ~ Sn ( ~ Sb) ~ Ti > Dy ~ (Li) > Ho = Y > Yb.
This rule works in general and suggests that the overall fractionation processes operating
during magma generation and evolution are relatively simple, involving no significant
change in the environment of formation for MORBs and OIBs.
In detail, minor differences in element ratios correlate with the isotopic characteristics
of different types of OIB components (HIMU, EM, MORB). These systematics are
interpreted in terms of partial-melting conditions, variations in residual mineralogy,
involvement of subducted sediment, recycling of oceanic lithosphere and processes
within the low velocity zone. Niobium data indicate that the mantle sources of MORB
and OIB are not exact complementary reservoirs to the continental crust. Subduction of
oceanic crust or separation of refractory eclogite material from the former oceanic crust
into the lower mantle appears to be required. The negative europium anomalies observed
in some EM-type OIBs and the systematics of their key element ratios suggest the
addition of a small amount (~<1% or less) of subducted sediment to their mantle sources.
However, a general lack of a crustal signature in OIBs indicates that sediment recycling
has not been an important process in the convecting mantle, at least not in more recent
times (~<2 Ga). Upward migration of silica-undersaturated melts from the low velocity
zone can generate an enriched reservoir in the continental and oceanic lithospheric
mantle. We propose that the HIMU type (eg St Helena) OIB component can be
generated in this way. This enriched mantle can be re-introduced into the convective
mantle by thermal erosion of the continental lithosphere and by the recycling of the
enriched oceanic lithosphere back into the mantle.
That much good ensues and that the science is greatly advanced by the
collision of various theories cannot be doubted. Each party is anxious to
support opinions by facts. Thus, new countries are explored and old districts re-
examined; facts come to light that do not suit either party; new theories spring
up; and in the end, a greater insight into the real structure of the earth's surface
is obtained.
Henry Thomas De l~t Beche (1830)
From SAUNDERS,A. D. & NORRY,M. J. (eds), 1989, Magmatism in the Ocean Basins, 313
Geological Society Special Publication No. 42, pp. 313-345.
Downloaded from http://sp.lyellcollection.org/ at Niedersachsische Staats Und Universitatsbibliothek on
March 7, 2012
bombardment and lithosphere subduction. Vig- arc basalts (eg Saunders et al. 1980). Recycling
orous convection in the early Earth's mantle of this mantle wedge or the overlying lithosphere
would quickly stretch and mix this oceanic litho- could introduce subduction-zone-related chemi-
sphere with the convective mantle. Some of this cal and isotopic characteristics into the convec-
lithosphere may have been subducted into the ting asthenosphere. For example, the recycled
lower mantle. The magnitude of these activities oceanic crust might show a positive niobium
decreased as the Earth's heat engine slowed anomaly relative to lanthanum (ie La/Nb > 0.95
down. Two-layer mantle convection may have (the chondritic ratio), but La/Sm < chondrite).
eventually been initiated as a result of a density The upward migration of highly incompat-
barrier imposed by temperature-dependent ible-element-enriched CO2-rich melts from the
phase transformation at 670 km (eg Sawamoto low velocity zone (LVZ) beneath the continen-
1987) or cumulation of refractory subducted tal and oceanic lithosphere (eg kimberlites, car-
lithosphere at the base of the upper mantle bonatites, nephelinites) can be a continuous or
(Ringwood & Irifune 1988). However, it is poss- episodic process responsible for local enrich-
ible that whole-mantle convection has continued ments of incompatible trace elements (eg Green
throughout Earth's history (Gurnis & Davies 1971). Although such enrichment processes
1986). should not be limited to the Archaean and early
If the mantle differentiation scheme discussed Proterozoic, the more vigorous melting possible
above is realistic then the high 3He/4He ratios in the Precambrian mantle would promote such
(1> 20 x atmospheric) observed in some enrichment processes. This type of enrichment
OIBs do not indicate an origin from a primordial can be superimposed upon earlier enrichment
unfractionated lower mantle. These ratios may events related to subduction zone processes.
instead result from the cooling of the outer Such enriched lithospheric mantle commonly
core. Similarly, the 129Xe anomaly observed in has a refractory major-element composition,
some MORB samples (All/~gre et al. 1987) is reflecting residual products of earlier melting
unlikely to result from the early (4.4 Ga) separ- event(s) (eg Frey & Green 1974).
ation of the upper mantle from the lower mantle. Recycling of the continental lithosphere by
The stabilization of the Archaean continental delamination (McKenzie & O'Nions 1983) or
crust was accompanied by the formation of thermal erosion, and the subduction of oceanic
thick refractory harzburgitic lithosphere. This lithosphere (Hofmann & White 1982; Ringwood
subjacent mantle formed as a residual product 1982), introduce compositionally dis-
of mantle melting. Occurrences of diamonds of tinctive sources into the convecting mantle. Such
Archaean age from kimberlite pipes from the material could be preserved as coherent entities
Kaapvaal craton, South Africa (Kramers 1977; through billions of years of convection. Thermal
Richardson et al. 1984), suggest the early exist- reactivation of enriched ancient continental
ence of Archaean continental lithosphere up to lithosphere by plume activity or rifting may also
about 200 km thick. Such lithosphere could play an important role in the generation of
have been developed through underplating of some continental basalts. Additionally, it may
residual mantle related to intraplate magmatic also cause thermal erosion of the continental
activity and/or depletion of the mantle wedge lithosphere and incorporate it into the convec-
above a subduction zone. ting mantle.
Studies of modern island arc basalts suggest There are different opinions regarding the
that subduction of the oceanic lithosphere, with fate of the subducted oceanic lithosphere. It
or without associated sediment, can continu- could have been continuously stretched and
ously modify the chemical, isotopic and miner- eventually well mixed back into the convective
alogical composition of the overlying mantle upper mantle (eg All~gre & Turcotte 1986).
wedge. The release of fluids and hydrous silicate Alternatively, because of the density barrier at
melts from the down-going altered oceanic crust 670 km caused by phase transformations, the
can result in the enrichment of incompatible subducted oceanic crust and refractory harzbur-
elements in the mantle wedge above the sub- gite lithosphere beneath it may buckle and ac-
duction zone. Stabilization of phlogopite in the cumulate at the base of the upper mantle and
mantle wedge provides a site for the introduced form a boundary layer (eg Ringwood 1982;
alkalis and barium (eg Wyllie & Sekine 1982) Ringwood & Irifune 1988). However, sub-
and it has been suggested that the formation of duction of the oceanic lithosphere into the lower
titanate minerals (such as sphene and rutile) in mantle appears to be indicated by geophysical
the subducted slab may have played an import- observations (Creager & Jordan 1984). In
ant role in the depletion of high field strength Ringwood's model, the megalith accumulates at
elements (eg titanium, niobium, tantalum) in the base of the upper mantle and warms up
Downloaded from http://sp.lyellcollection.org/ at Niedersachsische Staats Und Universitatsbibliothek on
March 7, 2012
the trace-element abundances of the primitive of mineral/melt Kd for these elements during
mantle composition. This type of plot, estab- oceanic basalt genesis is the presence of residual
lished by Sun et al. (1979), has been called a clinopyroxene and garnet (Fujimaki et al. 1984;
'spidergram' by Thompson et al. (1983) and an Jochum et al. 1986; Watson et al. 1987).
'extended Coryell-Masuda diagram' by On the basis of these chemical systematics,
Hofmann et al. (1986). We prefer that the diag- when accurate analytical data are available for
ram be called a mantle normalizing diagram. only a few critical trace elements (eg Y, Ti, Zr,
This method of normalization is similar to the Sr, P, Nb, K, Rb), the concentrations of other
practice of chondrite normalization for the rare elements on this plot can be estimated quite
earth elements. However, as Newsom et al. accurately. In this respect it is somewhat sur-
(1986) and Hofmann et al. (1986) have pointed prising to observe that phosphorus and neo-
out, there are potential problems in such nor- dymium behave very similarly during MORB
malization plots if the relative incompatibility and OIB genesis (P/Nd = 74+ 13) (McDonough
of elements changes with time. They found that et al. 1985), since these elements substitute into
the Ce/Pb and Nb/U ratios observed in OIB and different crystal sites (phosphorus substitutes
MORB are nearly constant, 25+5 and 47+10 for silicon whereas neodymiun substitutes for
respectively, but quite different from the primi- calcium). Additionally, it is suggested that large
tive silicate mantle ratios (9 and 30), indicating deviations from Nb/Ta = 17 and Zr/Hf = 36 in
that processes operating in the early history of MORB and OIB are most probably due to
the Earth (~> 2 Ga) caused fractionation analytical errors or contamination of niobium
of cerium from lead and niobium from and tantalum from the tungsten carbide mill
uranium. Rubidium and caesium show a similar during sample preparation. Fractionation of
behaviour. Oceanic basalts have nearly constant Nb/Ta in island arc basalts due to residual
Rb/Cs ratios of about 80, which is distinct from titanates during magma generation is possible,
the primitive mantle value of 20+10 but reliable data are needed to evaluate this
(McDonough et al. 1987). Taking these obser- possibility.
vations into account we have calculated the Neodymium and hafnium isotope studies of
normalizing mantle values for lead (used in Fig. mantle-derived volcanic rocks indicate that
1 and other figures of this paper) on the basis of there is a consistent behaviour between S m - N d
a Ce/Pb ratio of 25 and for caesium on the basis and L u - H f throughout geological time
of an Rb/Cs ratio of 80. The primitive mantle (Patchett 1983). Since Sr -~ Pr, Pb ~ Ce and Hf
values used here are given in Table 1, with the Sm on the mantle-normalized plot (Fig. 1)
element concentrations in C1 carbonaceous and samarium, praseodymium and cerium are
chondrites and the three different types of close to neodymium, the mixing of different
oceanic basalts shown for comparison. mantle components for MORB and OIB (with
Figure 1 presents the primitive-mantle- different mantle-normalized patterns) is ex-
normalized patterns (except for Cs and Pb) for pected to generate linear trends in plots of
typical N-type and E-type MORB and an oceanic isotope ratios, eg 143 Nd/ 144Nd versus 87Sr/86Sr,
143
alkali basalt (OIB) with 87Sr/86Sr ~ 0.7035. The Nd/--144 Nd versus 176Hf/ 177 Hf and 875r/86Sr
data suggest that the order of trace-element versus 2~176 The isotope data in the
incompatibility in oceanic basalts is Cs = Rb literature are consistent with this (eg Zindler
(~ T1) ~ Ba (-~W) > Th > U ~ Nb = Ta ~ K et al. 1982; White 1985).
> La > Ce ~ Pb > Pr ( ~ Mo) = Sr > P = Nd The chemical regularities observed in basalts
(>F)>Zr=Hf~Sm>Eu~Sn(~Sb) can be used to obtain information on the com-
Ti> Dy ~ (Li) > Ho = Y > Yb. This order is position of the source and the processes and
mainly based upon the depletion pattern of N- mineralogy involved in magma generation. The
type MORB and the systematics of concentration relatively constant but low concentrations of
ratios observed in E-type MORB and OIB (Sun potassium, coupled with high concentrations of
& Hanson 1975a, 1976; Sun & Nesbitt 1977; niobium and lanthanum in some highly under-
Sun et al. 1979; Sun 1980; Hofmann et al. 1986; saturated alkali basalts, kimberlites and car-
Newsom et al. 1986; Ryan & Langmuir 1987). bonatites, suggests that potassium in the melts
The relative incompatibilities of these elements could be buffered by residual amphibole,
appear constant within oceanic basalts suggesting phlogopite and clinopyroxene (under very high
that the fractionation processes are simple and pressure) during source enrichment processes
involve no drastic change in elemental behaviour. and/or magma generation (eg Sun & Hanson
The systematic behaviour of some elements, eg 1975a; Clague & Frey 1982). Such residual
Nb-Ta, Sr-Pr, Z r - H f - Sm, T i - E u 3+, potassium-bearing minerals could also cause
C e - P b , N b - U , indicate that the main control depletions of rubidium, caesium, and barium in
Downloaded from http://sp.lyellcollection.org/ at Niedersachsische Staats Und Universitatsbibliothek on
March 7, 2012
a The compositions of C1 chondrite and primitive mantle are from McDonough & Sun (in prep.)
Values for N-type and E-type MORB and OIB are based on a literature survey and internal consistency of
elemental ratios.
b For mantle-normalized diagrams, the recommended normalizing values for lead and caesium are 0.071 and
0.0079, respectively.
these magmas relative to their neighbouring initial small degrees of mantle melting or during
elements on a mantle normalizing diagram. De- a metasomatic enrichment of the source region.
creasing ratios of Ba/Nb, K/Nb, Rb/Nb, Cs/Nb If phlogopite is indeed the mineral holding back
and Rb/Sr with increasing lanthanum and nio- potassium, rubidium, and caesium in the source
bium abundances are c o m m o n in alkali basalts, regions of silica-undersaturated nephelinites,
nephelinites and kimberlites, whereas tholeiites non-micaceous kimberlites and melilitites (Sun
commonly do not show such relative depletions. & Hanson 1975a, b; Clague & Frey 1982), then
This suggests that phlogopite does not play a the near constant Ba/Rb (~12) and Rb/Cs (~80)
role during tholeiitic magma generation. More- ratios observed in most OIBs requires that
over, since phlogopite is not an early liquidus barium and caesium have similar phlogopite/
mineral in most silica-undersaturated basaltic melt Kd values relative to rubidium (Philpotts
magmas, it acts as a residual phase during the & Schnetzler 1970; Ku e h n e r et al. 1981) or that
Downloaded from http://sp.lyellcollection.org/ at Niedersachsische Staats Und Universitatsbibliothek on
March 7, 2012
Alkali Basalt
(~o.7o.~5)
10
;.u
N-typeMORB z 1 0 ~ ~
lit i i i l I I I I I I I I I I
LaCe PrNdSmEuGd Tb DtHo Y ErTmYoLu
' ' ' i i i i i i i i i i i i ! i i i i i i
Cs R b B a T h U Nb K L a Ce P b P r S r P Nd Zr Sm Eu Ti Dy Y Yb Lu
Tl W Ta Mo F Hf Sn Li Ho
Fro. 1. Primitive mantle (modified) normalized abundance patterns for moderately to highly incompatible
elements in average N-type and E-type MORB and oceanic alkali basalts with 87Sr/S6Sr~ 0.7035. See Table 1
for normalizing values.
Kd are large for barium, rubidium and caesium Modification of the 'mantle-normalized' rule
when phlogopite is stable. Studies of uranium and exceptions to the rule
disequilibrium series isotopes in OIBs indicate
Among OIB samples with La 1> 50 ppm, de-
that 20% uranium or more has been held back
coupling and lowering of titanium, zirconium,
in the source, assuming that thorium is per-
strontium, phosphorus, potassium and thorium
fectly incompatible (eg Oversby & Gast 1968;
abundances from their corresponding rare earth
Newman et al. 1984). Therefore the variable
elements (REEs) is common. A new set of
(by more than a factor of 2) K/U ratios observed
relationships is developed for such rocks (Table
in OIBs could be due to these combined effects
2). These features may be due to factors such as
in addition to original source variations.
the saturation of minor minerals, changes in the
Finally, constant elemental ratios suggest that
residual mineral proportions (eg an increase in
zone refining melting or the chromatographic
garnet abundance with depth), a change in
column model of Navon & Stolper (1987) and
mineral/melt Kd, mantle metasomatism and/or
magma chamber processes such as mixing and
melting processes (eg if zone refining becomes
extensive fractionation and replenishment (as
important). For example, nephelinites from
discussed by O'Hara & Mathews 1981), which
Oahu Island have Ti/Eu ratios less than half of
could fractionate incompatible elements effi-
that commonly observed in alkali basalts (3000
ciently, are not dominant processes operating
versus approximately 6000+700 respectively)
during the generation of primitive oceanic ba-
salts. and they are also depleted in niobium, tantalum,
a In general we are not sure whether at low degrees of partial melting these ratios are reflecting source
mineralogy or subsequent lower-pressure fractionation effects.
Downloaded from http://sp.lyellcollection.org/ at Niedersachsische Staats Und Universitatsbibliothek on
March 7, 2012
zirconium and hafnium. It is very likely that 1982; White 1985; Hart et al. 1986; Zindler &
these depletions are due to saturation of a ti- Hart 1986; Allrgre et al. 1987). Different
tanate mineral in the source region (Clague & approaches have been used to estimate the
Frey 1982). minimum number of end-members required.
Furthermore, an examination of the chemical Zindler et al. (1982) suggested that on a three-
and isotopic compositions of group I and II dimensional plot of lead, strontium and
kimberlite samples (Kramers et al. 1981; Smith neodymium isotopes most oceanic basalts fall
1983; Muramatsu 1985; Muramatsu & on or near a mantle plane defined by average N-
Wedepohl 1985; Smith et al. 1985) reveals that type MORB (2~176 ~ 18.5 87Sr/86Sr
in many respects group I kimberlites are com- 0.7029, eNd = + 13), St Helena (~ 20.8, 0.7029,
parable with Bouvet type OIB, whereas group +5) and Kerguelen Island (~ 18.4, 0.7054,
II kimberlites are more akin to Dupal type OIB - 1 ) . White (1985) subdivided oceanic basalts
(Le Roex 1986). However, exceptions to these into five distinct groups on isotope plots: MORB
general comparisons are that both groups of group (including Iceland, Galapagos, Easter
kimberlites have higher La/Ce ratios, lower Islands), St Helena group (including Austral,
Sr/Nd ratios and commonly show zirconium and Comores, Ascension), Kerguelen group (in-
hafnium depletions relative to samarium. These cluding Gough, Tristan da Cunha), Society
differences are most probably due to melting group (including Marquesas, Samoa, Sao
under very different conditions (eg variable Miguel) and Hawaii island group. Most Society
clinopyroxene/garnet ratios, P - T conditions group samples clearly fall above the mantle
and/or Kd values). Detailed study of the fine plane of Zindler et al. (1982) with higher 87Sr/
features of the mantle-normalized diagram may 86Sr ratios at the same lead and neodymium
provide useful information regarding the rela- isotope ratios. Zindler & Hart (1986) used
tive importance of garnet and clinopyroxene, in different terminology to define the charac-
addition to the effects of other factors. It may teristics of mantle end-members: DMM (de-
also give insights into the possible process of pleted MORB mantle), PREMA (prevalent
melting of stretched veins of eclogite during mantle, including Iceland, Galapagos, the
OIB generation. Easter Islands of White's MORB group and the
There are several reports of tholeiites and Hawaiian islands), HIMU (long-term high U/Pb
alkali basalts that are enriched in REEs and in the source, including St Helena, Tubuai,
yttrium relative to other incompatible elements Mangaia), EM1 (enriched mantle type 1 for the
(eg western Victoria, Australia (Frey et al. 1978; Kerguelen type) and EM2 (enriched mantle
McDonough et al. 1985), Norfolk Island (Green type 2 for the Society Island type). Their EM1
1978) and the Hawaiian islands (Fodor et al. samples from the Walvis Ridge clearly fall below
1987)). Some samples have negative cerium the mantle plane of Zindler et al. (1982) with
anomalies and concave-downward REE patterns lower 87Sr/86Sr at the same lead and neodymium
with relative enrichment of the middle REEs. isotope ratios (Fig. 2).
Recently, detailed microprobe studies by
Nagashima et al. (1986), Fodor et al. (1987) and , r !
Price (1987, pers. comm.) have detected very
small amounts of REE-rich carbonates and
phosphates of secondary origin. It is therefore
likely that post-magmatic alteration is respon-
sible for REE and yttrium enrichment in these 5 ~ EMOA
unusual samples. TIES
~Nd
Different mantle processes have been pro- potassium-bearing residual minerals, as dis-
posed for the generation of the different isotope cussed earlier.
groups and end-members. Since isotope vari- In Fig. 3 a representative H I M U - t y p e sample
ation is a function of parent/daughter abundance (St Helena: 2882) and an EM-type sample
ratio and isolation time, the same mantle pro- (Gough island: G i l l ) (Table 4) are used to
cess operating at different times and with dif- illustrate the differences in mantle-normalized
ferent degrees of elemental fractionation will patterns for these two types of mantle source.
result in variable but coherent lead, strontium These patterns are considered representative of
and neodymium isotopic compositions. Con- their respective types on the basis of these plus
sequently, the 'end-members' (or components) other high quality data having good consistency
proposed by Zindler & Hart (1986) would not with other samples from the same islands. This
be expected to have fixed isotopic compositions. statement is also supported by data of critical
It is not surprising to find that the different element ratios presented in Table 3. The H I M U -
mantle processes responsible for generating a type mantle with low (0.7027-0.7030)87Sr/86Sr
variety of isotope groups also create distinct and very radiogenic 2~176 is charac-
chemical signatures. Therefore the variations in terized by a low Rb/Sr ratio (0.01-0.03) and a
the mantle-normalized patterns for different high 238U/2~ ratio (i>30); such mantle ma-
mantle components may reflect the effects of terials tend to have low K/U and K/Nb ratios
the different processes and environments, as (related to potassium, rubidium and barium de-
with isotopic data. A comparison of Nb/Pb, pletions). The mantle-normalized patterns of
Ce/Pb, Nb/U, K/U, K/Nb and Rb/Cs ratios, H I M U - t y p e mantle (St Helena) display a con-
ratios which encompass elements considered to tinuous decrease in the abundances of the ele-
behave similarly during generation of oceanic ments niobium to caesium, similar to typical
basalts, for different OIB end-members is pre- M O R B . In contrast, for EM-type OIB with
sented in Table 3. The contrast between samples 87Sr/86Sr i>0.7040 and lower 2~176 ratios
of the H I M U (or St Helena) and E M (or Gough) than St Helena type mantle (<18.6 for EM1,
type is quite clear and there are consistent 18.6-19.7 for EM2), the mantle-normalized
differences, especially for Nb/Pb and Ce/Pb pattern shows increased abundances of
ratios, in other OIB samples. Ratios involving elements more incompatible than niobium
potassium may be affected by the retention of (Cs, Rb, Ba, Th, U, K) and lead compared with
potassium (rubidium, barium and caesium) in the St Helena type. Because these two patterns
TABLE 3. Contrasts o f some element ratios observed in N-type M O R B and HIMU-type and EM-type
ocean island basalts
a Low K/La suggesting potassium has been held back? Data sources: Sun & Hanson (1975a); White et al.
(1979); Hofmann & White (1983); Hofmann et al. (1986); Newsom et al. (1986); Palacz & Saunders (1986);
Weaver et al. (1986); this paper.
Downloaded from http://sp.lyellcollection.org/ at Niedersachsische Staats Und Universitatsbibliothek on
March 7, 2012
(Fig. 3) have similar REE abundances, the two have high Ix values (~20). This correlation must
distinctive patterns from caesium to niobium be related to the source character and partial-
cannot be due solely to different degrees of melting conditions. Several mechanisms have
partial melting. Moreover, such OIBs often have been proposed to generate the isotopic and
barium spikes and high La/Nb ratios (some- chemical characteristics of different types of
times higher than the chondritic ratio). This OIB mantle sources. The HIMU and low 87Sr/
latter point is illustrated in Fig. 4 which shows a 86Sr character St Helena-type OIBs have been
linear correlation between Ba/Nb and La/Nb attributed to the recycling of altered oceanic
among OIBs (Weaver et al. 1986). These two crust into the OIB source region (eg Zindler et
ratios also correlate with 87Sr/86Sr in general. al. 1982; Palacz & Saunders 1986), mantle en-
'Exceptions' are known for other EM-type richment through melt migration from the LVZ
samples (eg group I kimberlites and Tutuila), into oceanic and continental lithosphere (eg
which have Ba/Nb ~< 8 and La/Nb ~< 0.8, and Sun 1980; Hart et al. 1986) or derivation of
such differences most probably reflect the mul- OIBs from the lower mantle which has lost lead
tiple origins of EM-type source regions. How- by continuous core formation (eg Vidal & Dosso
ever, we note that a general trend of low 87Sr/ 1978). However, Newsom et al. (1986) showed
868r, La/Nb and Ba/Nb to high 87Sr/86Sr, La/Nb that St Helena basalts have the same molyb-
and Ba/Nb exists. denum (highly chalcophile) to praseodymium
Lamproites from western Australia (Jaques (highly lithophile) abundance ratios as MORBs
et al. 1988) may also represent extremely and other OIBs, effectively arguing against con-
enriched EM2 type. They have a highly fraction- tinuous core formation. The EM-type character
ated mantle-normalized pattern (Fig. 5), es- is generally considered to be related to processes
pecially from niobium to rubidium. These samples associated with convergent margins, such as the
have low strontium and phosphorus abun- recycling of sediments with oceanic crust (eg
dances, a negative europium anomaly of up to Hawkesworth et al. 1979; Cohen & O'Nions
20% and high lead and zirconium abundances 1982; White 1985; Weaver et al. 1986). The
(similar to some western US minettes; Alibert formation of these OIB sources from remobiliz-
et al. 1986), possibly reflecting a sedimentary ation or delamination of metasomatized en-
component added to their source (eg Thompson riched continental lithosphere has also been
et al. 1984). Often, but not always, they have proposed (eg Richardson et al. 1982; McKenzie
depletions of niobium relative to lanthanum. & O'Nions 1983; Hawkesworth et al. 1986).
Multiple origins for EM-type mantle is likely,
including superposition of different processes.
The generation of different mantle Menzies (1983) suggested that EM1 could be
types and the niobium issue related to enrichment by the introduction of
CO2-rich silicate melt (eg kimberlite, nephelin-
OIBs with high K20 (t>2%) do not always have ite) whereas the high Rb/Sr, 878r/86Sr character
low U/Pb (~t) values (~<16) but it is generally of EM2 is similar to some modern island arc
true that OIBs with low KzO (-~1%) always basalts (eg White 1985). Hart (1984) and Zindler
100
S t H e l e n a (2882) ~ \
1 i i i i ! i | i | i i i i ,,i i | i !
Rb Ba Th U K Nb La Ce Pb Sr Nd P Sm Eu Ti Dy Er Yb
FIG. 3. A comparison of the mantle-normalized patterns of basalts from St Helena (2882), with a low 87Sr/86Sr,
and Gough Island (G 111), with a high STSr/86Sr. See Table 1 for normalizing values. Data from Table 3.
Downloaded from http://sp.lyellcollection.org/ at Niedersachsische Staats Und Universitatsbibliothek on
March 7, 2012
Neodymium and strontium isotope data for Chain 43(104-16 and 104-18) and St Helena samples are from
White & Hofmann (1982) (see their CH43 106-16 and 106-18 sample numbers); however, for 104-16 we
report the strontium isotope ratio for an acid-leached sample (which was normalized relative to 0.70800 for the
E&A standard).
& Hart (1986) emphasized that EM2 is almost convection patterns. Hawkesworth et al. (1986),
exclusively restricted to the southern hemi- on the other hand, favour the idea that the EM-
sphere. It could have originated from the type mantle in the southern Atlantic (Gough,
delamination of subduction-zone-related Walvis Ridge, Tristan da Cunha) has a shallow
metasomatized continental lithosphere of origin due to the thermal remobilization and
G on d aw a n a or the recycling of the mantle detachment of the continental lithosphere of
wedge above the subduction zones. However, southern Africa from South America during the
they do not preclude the earlier suggestion by breakup of Gondawana. More recently, Hart et
Hart (1984) that this regional EM-type anomaly al. (1986) pointed out that H I M U and EM1
in the sourthern hemisphere existed since the mantle types are often spatially related and
Archaean and might be controlled by global share the characteristic of lying below the mantle
Downloaded from http://sp.lyellcollection.org/ at Niedersachsische Staats Und Universitatsbibliothek on
March 7, 2012
10000
Kimberley Lamproites, W Australia ( E M 2?)
i ,.+
j ,o zo6/zo4,b-17~ t~, -~
87/86 Sr - 0.714 CelPb - 13
-.%
ENd ~ .13
FIG. 5. Mantle-normalized diagram of EM2-type Miocene iamproites from W Kimberley and 1.2 Ga
lamproites from E Kimberley, western Australia. Data source" Jaques et al. (in press). See Table 1 for
normalizing values.
Downloaded from http://sp.lyellcollection.org/ at Niedersachsische Staats Und Universitatsbibliothek on
March 7, 2012
19 10
MOST 9 525A
GOUGH ~ WALVIS RIDGE
SAMPLES~ ~N 0 7050 10446
1 i i i i i i i i h i k i i i i i i
Rb Ba Th U K Nb La Ce Sr Nd P Sra Zr gu Ti Y Yb
Ba/Nb ~WALVIS RIDGE 0.7040
f "=~/TRISTAN da CUNHA
FIG. 7. Comparison of mantle-normalized patterns
11 ~ 0.7051 for two MORB samples from 45~ from the Mid-
PRIMITIVE Atlantic Ridge. These basalts have similar lead,
MANTLE I KOOLAU strontium and neodymium isotope compositions but
--~- ITHOLEIITES different La/Nb ratios which can only be attributed to
(/ ------T~.ASCENSIO N ] 0.7040 melting. See Table 1 for normalizing values.
7 "~[~"" 0.7028
i30A ST HELENA high La/Nb ratios (1.6), similar to those found
0.7029 I N-MORB 0.7025
I I I I I in the Honolulu series, are also found in some
0.6 0.8 1.0 1.2 Koolau tholeiites (Fig. 8), although these two
La/Nb Oahu basalt series are derived from isotopically
FIG. 6. Positive correlation between Ba/Nb and different sources (Roden et al. 1984). In contrast
La/Nb ratios of basalts from St Helena, Ascension, with the Oahu case, EM samples from Walvis
Tristan da Cunha and Walvis Ridge (modified from Ridge and Gough (Fig. 4) with high La/Nb
Weaver et al. 1986). Data for Koolau tholeiites, N- ratios have Ti/Eu ratios close to the chondritic
type MORB, group II kimberlites and sediments are ratio, suggesting that a titanate mineral does
added for comparison. not play a role in the generation of these basalts.
We therefore agree with Weaver et al. (1986)
that the high La/Nb ratios of these EM-type
from Haleakala, Maui, also show a consistent basalts reflect a mantle source characteristic,
variation in La/Nb ratios with degree of partial possibly resulting from the involvement of sub-
melting: the basanites representing the lower ducted sediments. This conclusion is further
degrees of partial melting have lower La/Nb supported by the high Ba/Nb, 87Sr/86Sr (Fig. 6)
ratios than the tholeiites which are derived by and 2~176 commonly observed in EM-type
larger degrees of partial melting. However, in OIBs and typical of sedimentary material. How-
contrast to the Loihi basalts, the tholeiitic and ever, we would emphasize that sediment sub-
alkalic basalts from Haleakala are derived from duction, along with the oceanic crust, is not a
isotopically different mantle sources ( C h e n & unique solution for the generation of EM-type
Frey 1985). mantle. The addition of a fluid or melt derived
In addition to reflecting the source character, from an altered subducted oceanic crust, with
high La/Nb ratios in OIBs can be produced if a or without sediment, into the overlying mantle
residual titanate mineral is present during open- wedge or adjacent continental lithospheric
system processes in the source (eg metasomatic mantle may also generate high Ba/Nb and La/Nb
enrichment or partial melting) or during magma ratios.
differentiation. Nephelinites and basanites of
the Honolulu series from Oahu are good
examples (Fig. 8). They show relative depletions
Mid-ocean ridge basalt source connection of the
in titanium, niobium, tantalum, zirconium and
HIMU-type mantle
hafnium and have Ti/Eu ratios (=3000) less
than half those normally observed in OIBs HIMU-type OIBs are characterized by very low
87
(about 60(~, close to the chondritic ratio ap- Sr/86 Sr ratios (as low as 0.7027) similar to
proximately 7800). Clague & Frey (1982) some N-type MORBs, high 2 0 6 Pb/ 204 Pb ratios
suggested that a residual titanate mineral is (~21.0) but low inferred Th/U ratio in the
responsible for the niobium depletion relative source (approximately 3.3, see below). It is now
to lanthanum. Additionally, it is interesting that generally accepted that the HIMU-type mantle
Downloaded from http://sp.lyellcollection.org/ at Niedersachsische Staats Und Universitatsbibliothek on
March 7, 2012
Koolau Tholeiite
(0.7043, +1.9)
"~ Rib I I I I I I I I I I I I I I I i , , I I I I I
Ba Th K N b La Ce N d R b Ba Th K N b La Ce N d R b Ba Th K N b La Ce N d
FXG.8. Comparison of mantle-normalized patterns of Rb, Ba, Th, K, Nb, La, Ce and Nd for selected Hawaiian
basalts. The samples used include Koolau tholeiite (WW 9980), Honolulu nephelinite (68 Pb 2), Honomanu
tholeiite (C122), Hana basanite (65-11), Loihi tholeiite (24-7) and Loihi basanite (17-7). Data sources:
Clague & Frey (1982); Frey & Clague (1983); Roden et al. (1984); Staudigel et al. (1984); Chen& Frey (1985).
See Table I for normalizing values.
was originally derived from a MORB-type com- mantle, as suggested by Zartman & Doe (1981),
ponent more than a billion years ago; however, to calculate a two-stage lead isotopic evolution.
there is disagreement regarding the processes Their version II lead isotopic growth curves for
responsible. the depleted mantle are consistent with the
On lead isotope ratio plots (Fig. 9) each available lead isotope data for Archaean and
H I M U island exhibits a small range of isotopic Proterozoic komatiites and basalts derived from
variation. To estimate time of isolation for the depleted mantle, as well as Phanerozoic ophio-
HIMU-type mantle, we use the lead isotopic lites. Feasible solutions to our calculations are
composition evolution of the depleted upper evaluated by comparing the calculated IX2values
FI6.9. Simplified 2~176 versus 206pb/2O4pbplot for oceanic basalts from mid-ocean ridge (MORB),
Walvis Ridge (W), Hawaiian Islands (H), Kerguelen (K), Gough (G), Tristan da Cunha (T), Society Islands
(S), St Paul's Rocks (St P), Ua Pou (U), Ross Island (R), Canary Islands (C), Bouvet (B), St Helena (St H),
Tubuai (Tu) and Mangaia (M). Data sources: Sun (1980); Cohen & O'Nions (1982); Richardson et al. (1982);
Roden et al. (1984); Vidal et al. (1984); Palacz & Saunders (1986); Stille et al. (1986); Dupuy et al. (1987);
Wright & White (1987).
Downloaded from http://sp.lyellcollection.org/ at Niedersachsische Staats Und Universitatsbibliothek on
March 7, 2012
ST HELENA 2882 /
206pb/204pb = 20.90 /
2~176 - 15.79 /
2~ = 40.13 J
hi,3 = Measured238U/204pb-" 28 # ~
/ x
25
/ N o fractionation of
/ U/Pb during magma
generation
2.
2O
~_ I . I
3.5
~,~ 3.0
2.6 -
~
I I
16.0
',d"
r red
~-- 15.8
o-
15.6
I I
3 2 l
Ga
FIG. 10. Model calculations for the lead isotope evolution of a St Helena sample (2882) assuming that it was
derived from a depleted upper mantle MORB-type source some time ago. When an isolation time is assigned,
specific values of ~ and K2 are required to match the measured 206 Pb/204 Pb and 208 Pb/204 Pb ratios. At 2.0 Ga,
model calculations give the best solution to match the measured 2~176 and p values.
Downloaded from http://sp.lyellcollection.org/ at Niedersachsische Staats Und Universitatsbibliothek on
March 7, 2012
328 S . - s . S u n & W . F. M c D o n o u g h
It is important to point out that if St Helena, tholeiites have high Nb/La, Ce/Pb, U/Pb and
Tubuai, Mangaia and Ua Pou samples are not Hf/Lu ratios. Dupuy et al. suggested that the
derived from pure H I M U end-members but mantle source of the Ua Pou tholeiites contains
either (i) contain an EM-type component with residual oceanic crust formed by dehydration
low 2~176 but high 2~176 (eg Duncan and partial melting during subduction.
et al. 1986), or (ii) are derived from a recently These unusual Ua Pou tholeiites with
enriched high-~t reservoir (eg St Paul's Rock; concave-downward R E E patterns (Fig. 11) of-
Roden et al. 1984) followed by mixing with fer a rare opportunity to evaluate the recycled
depleted asthenosphere (Fig. 9), then theage of oceanic crust model. They have high nickel
the H I M U isolation could be considerably (155-550 ppm), TiOa ( 3 . 7 % - 4 . 1 % ) and ce-
younger. Such an interpretation may apply to rium (61-78 ppm), comparable with those of
much of the non-HIMU OIB lead isotope data other H I M U OIBs considered in this study.
(see later). Consequently, based on the arguments pre-
Model calculations for the 2~176 evol- sented earlier, it would be reasonable to expect
ution of St Helena samples (Fig. 10) yield a more than 20?/0 fractionation of the La/Nb ratio
Th/U ratio of 3.30+0.05 in the mantle source during magma generation. The measured La/Nb
starting at 2.0 Ga ago. This value is lower than ratio (0.85) of these tholeiites is higher than
the measured Th/U ratios in St Helena samples those of other H I M U OIBs (St Helena, 0.69;
(3.5-3.8) (Weaver et al. 1986; this paper), Ascension, 0.65; Mangaia, 0.77) but similar to
consistent with the fractionation of thorium that of the EM-type OIB of Tristan da Cunha
from uranium during OIB magma generation. (0.86). After a 20% correction for the partial-
Similar conclusions are reached through 23~ melting effect (based on studies of Hawaiian
232Th studies on young OIB (eg Oversby & nephelinites discussed earlier), the mantle
Gast 1968; Newman et al. 1984). source for Ua Pou tholeiites has La/Nb = 1.06,
Recently, Dupuy et al. (1987) presented a which is greater than the chondritic ratio (0.96),
strong case for a connection between the re- ie it actually shows niobium depletion relative
cycled oceanic crust and HIMU-type OIB to lanthanum. The lack of a positive niobium
sources on the basis of their study of Ua Pou anomaly relative to lanthanum in their mantle
Island, Marquesas Archipelago. They empha- source weakens the argument of a recycled
sized that Ua Pou tholeiites are highly relatively oceanic crust model for the Ua Pou tholeiites,
depleted in most incompatible elements (barium, or at least makes it unnecessary.
rubidium, thorium, uranium, potassium, nio- The high Hf/Lu ratios (22-28) observed in
bium and light REEs) and have mantle- the Ua Pou tholeiites have been used by Dupuy
normalized patterns (Fig. 11) showing mirror et al. to support the recycled oceanic crust
images of those for island arc basalts, ie these model. However, since hafnium is much more
100
UaPou6597
e~
9 (0.70289, +4.6,19.86,15.54)
La Ce Nd S m Eu Tb Yb Lu
1 I I I I I I I I I I I I I I I | I I
Rb Ba Th U K Nb La Ce Pb Sr Nd P Sm Z r Eu Ti Y u
Fro. 11. Mantle-normalized diagram of two HIMU-type OIB samples from Ua Pou. They have concave-
downward patterns in the light REE region. Data from Liotard et al. (1986) and Dupuy et al. (1987). See Table
1 for normalizing values.
Downloaded from http://sp.lyellcollection.org/ at Niedersachsische Staats Und Universitatsbibliothek on
March 7, 2012
incompatible than lutetium during OIB gener- At present, there are no experimental data to
ation, this argument may not be valid. A com- show how these trace elements (and the La/Nb
parison of Srn/Hf is more appropriate because ratio discussed earlier) will be fractionated
these elements are similarly incompatible (Fig. during partial melting in the garnetite stability
1). In fact, the Sm/Hf ratio of Ua Pou tholeiites field. However, upward migration of COs-rich
(1.43) is equal to that of the primitive mantle silicate melts such as kimberlite, carbonatite
(Table 1) suggesting that hafnium has not been and nephelinite from the LVZ at the top of the
fractionated from samarium by processes in- depleted asthenosphere could be an effective
volved in the generation of HIMU-type mantle. mechanism to achieve the HIMU characteristics
Although we agree with Dupuy et al. that the (including Th/U) and real geological obser-
very high Ce/Pb ratios of Ua Pou tholeiites vations are available to support this idea (eg
(38-78) are very unusual for OIBs (25_5), Green 1971; Kay 1979; Menzies & Murthy 1980;
these ratios are also much higher than those of Sun 1980; Hart et al. 1986). These CO2-rich
other HIMU-type OIBs shown in Table 2. We silica-undersaturated melts have low Rb/Sr
do not have a good explanation for this anomaly. ratios but high U/Pb ratios (see earlier) and
The neodymium isotope data presented by match the characteristics of the HIMU type.
Duncan et al. (1986) and Dupuy et al. (1987) Thus, the HIMU character can be fully satisfied
require that the mantle source of Ua Pou by this melting process, since thorium is more
tholeiites has an Sm/Nd ratio lower than chon- incompatible than uranium, which is more in-
drite, after isolation from the MORB-type compatible than lead, and residual phlogopite
source about 1 billion years ago. Combining can hold back more rubidium from the melt
this information with niobium and REE data, during small degrees of partial melting. The
we suggest that this mantle has a concave- choice between oceanic and continental litho-
downward REE pattern with niobium depletion sphere for the HIMU component will be evalu-
similar to the pattern of many Hawaiian ated later.
tholeiites (eg Fig. 8).
In order to make a correct choice between
Subduction zone processes and EM-type mantle
the two processes proposed for generation of
the HIMU-type mantle, ie through recycling of EM2-type mantle with high Rb/Sr and S7Sr/8%r
altered oceanic crust or migration of silicate ratios is commonly considered to be a product
melt from the depleted mantle into the litho- of the migration of water-rich fluids or melts
sphere, critical evaluation of the two processes enriched in potassium, rubidium and barium
is required. Some MORB and OIB tholeiites (ie associated with subduction zones (eg Menzies
Iceland) have light-REE-depleted or flat pat- 1983; White 1985; Zindler & Hart 1986),
terns with positive niobium kicks on the mantle- whereas the chemical and isotopic character of
normalized diagram (La/Nb less than the EMl-type mantle, with high Ba/Nb and La/Nb
chondritic value). These data support the idea ratios, has been interpreted by Weaver et al.
of recycling oceanic crust in the MORB and (1986) as the product of sediment subduetion.
OIB sources and decreasing La/Nb through the In this section we point out some weaknesses of
subduction zone dehydration/partial-melting the subduction-zone-related models, emphasize
process. However, if the oceanic crust recycling the need for alternative mechanisms and pro-
model is correct for the HIMU character, then pose more definite tests for sediment involve-
alteration and subduction of this crust (dehy- ment in OIB sources.
dration and partial melting) must decrease the As discussed earlier, high Ba/Nb and La/Nb
La/Nb ratio by 20% or more and the U/Pb ratio ratios observed in many EM-type OIBs are not
by a factor of 2 without significantly (>20%) necessarily diagnostic of sediment involvement
affecting the Rb/Cs, Rb/Sr, Th/U, Ce/Pb and in their source. Except for Gough, Walvis
Nb/U ratios. This does not seem likely. Taking Ridge and Aitutaki (Table 3), there is a general
Th/U as an example, a model calculation of the lack of strong trace-element indication of
lead isotope evolution of the HIMU-type St sediment recycling. Mixing as little as 1% of
Helena sample (2882, Fig. 10) requires that the pelagic sediment (with Pb=25 ppm, Pb/Ce=0.6)
Th/U actually increases slightly relative to the with subducted oceanic crust (or 0.2% for
MORB source at about 2.0 Ga ago. Further- MORB-type peridotite mantle) is expected to
more, dehydration and partial melting is an change the Nb/U, Ce/Pb and Rb/Cs ratios con-
effective 'depletion' mechanism. In order to siderably from the nearly constant values (47+_10,
generate a St Helena-type OIB source character 25+5, 80+20) for MORB and OIB (Hofmann
an enrichment process after subduction, such as & White 1983; Hofmann et al. 1986; Newsom
megalith melting (Ringwood 1982), is required. et al. 1986). For example, the Ce/Pb ratio could
Downloaded from http://sp.lyellcollection.org/ at Niedersachsische Staats Und Universitatsbibliothek on
March 7, 2012
be reduced to about 14. Furthermore, a mixture mixture of sediment and oceanic crust. It is
of about 0.5% average shale (Taylor & more likely to be derived from an enriched
McLennan 1985, Table 2.9) and a model OIB mantle wedge or lithosphere above the sub-
source (based on the OIB data in Table 1) duction zone, with some input from the sub-
would produce a source capable of generating ducted sediments.
an EM-type OIB with consistent key element Secondly, a subduction zone environment is
ratios as found in Table 3 (eg Nb/U=42, not required for at least some EMl-type mantle.
Ce/Pb=20, Nb/Pb= 11, K/U=13 000, K/Nb=275, Enrichment can be achieved by silicate melt
Nb/Th= 11). Additionally, the introduction of migration from mantle plumes with less depleted
such a small sediment component 1.0-2.0 Ga character than a MORB source (eg Bouvet,
ago would also be consistent with the variation Hawaii or OIBs lying isotopically closer to the
in strontium and neodymium isotopic compo- primitive mantle). EMl-type mantle generated
sitions observed in EM-type OIB (eg Bouvet in this way will have higher Th/Nb and Th/Ta
versus Gough islands). ratios than the HIMU-type mantle, whereas
However, if sediment subduction is invoked Ba/Nb, Rb/Nb and La/Nb ratios could be similar
to explain the isotope evolution and the devel- to the H I M U type owing to a low alkali charac-
opment of negative europium anomalies in some ter controlled by residual potassium-bearing
EM-type mantle (see below), then some modifi- minerals and the fact that niobium is more
cation of this simple mixing model is required. incompatible than lanthanum. These characters
In this connection, it is interesting to note that are shared by some group I kimberlites with
the ratios Nb/U ~ 34, Ce/Pb ~ 13, Rb/Cs ~ 123 878r]86Sr = 0.704-0.705 and some E M 2 0 I B s
(Fig. 5) of Miocene lamproites from western such as Sao Miguel (White et al. 1979; Le Roex
Australia (Jaques et al. 1988) are not very differ- 1986). Le Roex (1986) emphasized that the
ent from values for MORB and OIB mentioned trace-element and isotope data for group I
above. Isotope and trace-element characteristics kimberlites cannot be explained by the sediment
(including a 20% negative europium anomaly) subduction model. Instead, they are probably
of these lamproites (extreme EM2 types) evolved from less depleted or close to primitive
strongly suggest the involvement of sediments mantle material.
in their mantle source (Nelson et al. 1986; Jaques Although the discussion presented above
et al. 1988). Since the EM-type OIBs in Table 3 sounds negative towards the connection be-
require much less sediment involvement than tween sediment subduction and the formation
these lamproites, the lack of a large amount of of EM-type mantle, we do believe that sediment
variation in Nb/U, Ce/Pb and Rb/Cs ratios may subduction has left an important fingerprint on
not be a strong argument against involvement the geochemistry of at least some EM-type OIBs
of a minor amount of sediment (2% or less) in (eg Walvis Ridge, Gough, Society Island,
some EM-type mantle. However, mixing of Samoan Islands). Further detailed geo-
small amounts of pelagic sediment with oceanic chemical studies of both EM1 and EM2 end-
crust is not capable of generating the 2~176 members are required to establish this case
characteristics of the EMl-type mantle in the S firmly. We suggest that accurate estimation of
Atlantic. Using lead isotope evolution curves the europium anomaly in OIBs can be of special
for the depleted mantle and upper crust (sedi- importance for the evaluation of the effect of
ment source) given in Zartman & Doe (1981), a the subduction of sediment and altered oceanic
model calculation for sediment-oceanic crust crust on the OIB source as well as onfO2 during
mixing at 1.5-2.0 Ga with 50% of the lead magma generation. Since plagioclase, which has
contributed by the 1% subducted sediment (as a high Kd for Eu 2+, is neither a residual mineral
suggested by Weaver et al. 1986) requires a during alkali basalt genesis nor an early liquidus
Th/U ratio of 5.0 to generate the observed mineral during magma differentiation, eu-
2~176 ratio. This value is much higher ropium anomalies measured in primitive alkali
than the value (approximately 3.7) estimated basalts reflect the combined effects of source
for the Tristan da Cunha source. In general, character and fO2 during magma generation.
most oceanic basalts (EM, MORB, HIMU) Sun & Hanson (1975a) suggested that the
come from source regions with Th/U ratios of 2 % - 5 % positive europium anomaly commonly
3.8 or less, as has been demonstrated by 23~ observed in OIBs is due to the existence of
23aTh studies, or by a reduction of 15% +5% of Eu 2+ (->10%) in addition to Eu 3+ under low
the measured Th/U of fresh basalts to correct f O 2 mantle conditions and the fact that Eu 2+
for the Th/U fractionation during melting (eg (~Sr) is more incompatible than Eu 3+ and its
Oversby & Gast 1968). In our opinion, EM1- neighbouring Sm 3+ and Gd 3+. Sediments de-
type OIB in the S Atlantic does not represent a rived from various terranes have different
Downloaded from http://sp.lyellcollection.org/ at Niedersachsische Staats Und Universitatsbibliothek on
March 7, 2012
amounts of negative europium anomaly (eg This is consistent with the absence of the Dupal
MacLennan & Taylor 1981), with up to 35% anomaly in the Pacific mid-ocean ridge of the
europium depletion for modern crustal rocks. southern hemisphere (Macdougall & Lugmair
The available accurate REE data indicate that 1986; White et al. 1987). Furthermore,
HIMU OIBs have a 4 % - 6 % positive europium Archaean and Proterozoic Dupal-like anomalies
anomaly whereas E M 2 0 I B s from Samoan and observed in komatiites and galena and feldspar
Society islands show up to a 4 % - 6 % negative in granites of the southern continents (Africa,
europium anomaly (eg Sun & Hanson 1975a; Australia, India), where there is abundant early
Newsom et al. 1986; Palacz & Saunders 1986; Archaean crust, are likely to be a result of
McDonough & Sun, unpublished data), con- crustal contamination. For example, some late
sistent with the sediment subduction model. In Archaean komatiites (and associated NiS miner-
contrast, EM-type OIBs from Gough and Tristan alization) of the Yilgarn Block, western
da Cunha have a 2 % - 5 % positive europium Australia, have been contaminated by early
anomaly despite our belief that sediment has Archaean felsic crust which resulted in Dupal-
been involved in their mantle source. Either the type lead anomalies (Sun et al., in press). In
sediment which was subducted and ultimately contrast, Hawkesworth et al. (1986) suggested
incorporated into their source has a small nega- that the Dupal anomaly in the southern Atlantic
tive europium anomaly (ie from Archaean (ie Walvis Ridge, Gough, Tristan da Cunha and
terranes) and/or more probably the sediment the Discovery Table Mounts) may be of shallow
effect has been compensated by an increase of origin and due to the thermal reactivation and
the Eu2+/Eu3+ ratio during source modification detachment of the metasomatized lithosphere
(eg dehydration and partial melting of the sub- of southern Africa and South America before
ducted sediments) and magma generation under the breakup of Gondwana. The continental
more reducing environments than for other lithospheric delamination model of McKenzie
OIBs. To test further the significance of the & O'Nions (1983) is attractive because there is
negative europium anomaly in some primitive the opportunity to generate a variety of mantle
OIBs, we propose that samples with a strong types in the continental lithosphere within close
sediment lead isotope signature are most suit- proximity through subductioti-zone-related and
able, because modern sediments have a distinct unrelated processes. The superposition of the
and strong negative europium anomaly. products of different processes is possible. It
also allows long-term isolation of the enriched
mantle within the lithosphere before delami-
Continental lithosphere connection of the Dupal nation and recycling back into the convecting
anomaly and the low neodymium array mantle.
More recently, Hart et al. (1986) also seem to
Hart (1984) pointed out that EM-type OIBs are favour an ultimate origin for the Dupal anomaly
concentrated in the southern hemisphere, are from the continental lithosphere. They pointed
globe encircling in extent, and are centred at out that EM1- and HIMU-type OIBs often
latitude 30~ He named this isotopically anom- show close association in occurrence and share
alous region the Dupal anomaly re#on and the same character of long-term low Rb/Sr ratio
suggested that it may have been in existence compared with OIBs within the mantle array on
in the southern hemisphere since Archaean an S7Sr/S6Sr versus end plot (Fig. 2). HIMU-and
times. In terms of the model, the occurrences of EMl-type OIBs have lower eNd values than the
EM-type OIBs in the northern hemisphere, eg mantle array. There appears to be a lower limit
the Azores, Oahu Island (Koolau tholeiites) on this plot for OIB data defined by connecting
and the Japan Sea, and numerous occurrences the HIMU and EM1 end members. Hart et al.
of EM-type continental alkali basalts, tholeiites (1986) named this limiting line a 'LoNd' (low
and kimberlites have to be considered as excep- end) array. (Since HIMU and EM1 are charac-
tions to the general rule or due to crustal con- terized by having lower 87Sr/86Sr values due to
tamination. Later, Zindler & Hart (1986) lower time-integrated Rb/Sr ratios than samples
suggested that the enhanced rate of Pangean within the mantle array, it might be more ap-
subduction into the Dupal mantle (Anderson propriate to call the array a 'LoSt' (low 87Sr/
1982) may account for localization of EM2-type S6Sr) array.) They suggested that this LoNd
OIBs in the southern hemisphere. If this inter- array was developed within the subcontinental
pretation is correct, then EM2 anomalies in the lithosphere through the introduction of silicate
southern hemisphere are a relatively young sub- melts of different chemical and isotopic com-
duction-related phenomenon surrounding the positions (p. 1556): ' . . . EM1 end member being
Pangean continents, but are not circum-global. slightly modified primitive mantle, and the
Downloaded from http://sp.lyellcollection.org/ at Niedersachsische Staats Und Universitatsbibliothek on
March 7, 2012
332 S . - s . S u n & W . F. M c D o n o u g h
HIMU end member being metasomatically pro- very minor in modern times, (ii) metasomatized
duced, possibly during subduction episodes pre- continental lithosphere formed at 2 Ga or less is
ceding lithosphere accretion.' It is interesting to abundant on the worldwide scale and dominates
note that many of the Mesozoic and Cenozoic the scene, or (iii) delamination of continental
carbonatites of worldwide occurrence, not lim- lithosphere is sufficient but not necessary for
ited to the southern hemisphere, also fall on or OIB generation, and survival times of 2 Ga or
close to this LoNd array (eg Bell & Blenkinsop less in the convecting mantle are possible
1987; Dudas et al. 1987; Nelson et al. 1987). (Gurnis & Davies 1986). The last option will be
This is also true for some group 1 kimberlites further discussed in later sections.
(Kramers et al. 1981; Smith 1983). A major drawback to the continental litho-
The association of HIMU- and EM-type OIBs sphere delamination model is that there is a
in some ocean island or island groups of the general lack of OIB samples showing strong
southern Pacific (Vidal et al. 1984; Palacz & subduction-zone-related chemical and isotopic
Saunders 1986; Dupuy et al. 1987) and Walvis characteristics, which are commonly found in
Ridge (Fig. 6) does not offer a firm argument many continental mafic and ultramafic volcanic
for identifying their parentage (eg continental rocks. As shown in Figs 5 and 12, relative
lithosphere). Alternative mechanisms can be depletion of titanium, phosphorus, niobium,
suggested, eg mixing of different types of mantle tantalum and strontium, but extreme enrich-
components within the convective upper ment of lead and the presence of up to 20%
mantle. In addition, we would like to empha- negative europium anomalies, reflecting sedi-
size that the regional but isolated occurrences ment subduction effects (McLennan & Taylor
of HIMU (eg St Helena, Ascension, Cape 1981), have been observed in many post-orogenic
Verde, New England Seamounts, Canary and intraplate mafic to ultramafic peralkaline
Islands, Ahaggar, Cameroon Line, Guadalupe, volcanics (group II kimberlites of southern
SW Pacific islands, eastern Australia, South Africa and lamproites from western Australia)
Island, New Zealand, and Ross Island and Mary which are considered to be derived from the
Byrd Land, Antarctica) or EM (Gough, Tristan continental lithosphere modified by subduction
da Cunha, Kerguelen, Samoa) without an ac- zone processes (eg Duncan et al. 1984;
companying LoNd counterpart are quiet com- Thompson et al. 1984; Nelson et al. 1986; Jaques
mon. The regional distribution tendency of the et al. 1988). Among these samples there is often
Dupal anomaly (S Atlantic, Indian and SW a positive correlation between the size of the
Pacific Oceans) and the HIMU anomaly (some negative europium anomaly and the 87Sr/86Sr
of which are probably of shallow asthenospheric ratio (and Sr/Nd ratio) which is thought to
origin) also requires serious consideration in reflect the effect of subducted sediment, with
any genetic model. high 875r/86Sr, large negative europium anom-
The long isolation time (up to 2 Ga) required alies (about 30% but variable depending on the
by the isotope data of OIBs has been a major source rocks) and low Sr/Nd ratios (eg Alibert
consideration of many people in favouring the et al. 1986). In this connection, negative eu-
model of continental lithosphere delamination. ropium anomalies observed in Samoan and
In this respect, it is surprising to see that the Society alkalic basalts can be considered
OIB members of the LoNd array have well- as a good indication of sediment involvement in
defined isotopic correlations on strontium versus their mantle sources.
neodymium and lead versus strontium plots Detailed chemical and isotopic studies of
(Hart et al. 1986). Since the extent of low ENd Mesozoic Karoo basalts of southern Africa,
values below the main array is a function of Parana basalts of South America, African
Rb/Sr fractionation and isolation time, in ad- kimberlites and modern Dupal OIBs in the
dition to other variables such as Sm/Nd ratio, a S Atlantic (eg Smith 1983; Smith et al. 1986;
well-regulated mantle LoNd array is not ex- Duncan et al. 1984; Hawkesworth et al. 1984,
pected. Hart et al. (1986) pointed out that there 1986; Le Roex 1986) have offered an opportunity
are many EMl-type continental peralkaline to evaluate different models suggesting their
rocks (eg Leucite Hills, Smoky Butte) which fall source relationships and mechanisms of magma
well below the LoNd array. The lack of an generation. Based on hot-spot trace reconstruc-
obvious influence of the low neodymium array tions and geochemical data, Le Roex (1986)
by these extremely low ENd components from concluded that group I kimberlites are not
the continental lithosphere formed in the Dupal type (represented by Walvis Ridge and
Archaean (eg Dudas et al. 1987) may imply that Gough Island) but are related to Bouvet and
(i) delamination of continental lithosphere with Marion type hot-spots, whereas group II
Archaean isotopic signatures is not feasible or kimberlites are related to the Dupal islands.
Downloaded from http://sp.lyellcollection.org/ at Niedersachsische Staats Und Universitatsbibliothek on
March 7, 2012
C h e m i c a l a n d i s o t o p i c s y s t e m a t i c s o f o c e a n i c basalts 333
100
~9 - r-- ~ [ [ \ ~ S r i = 0.7095
9= .
S r i = 0.7050 ~ Nd i = 4 . 2
FIG. 12. Mantle-normalized diagram of two selected least-fractionated basalts from southern Karoo and
central Karoo. They show strong signatures of sediment and subduction-zone-related processes. Data from
Duncan et al. (1984). See Table 1 for normalizing values 9
Some clinopyroxene megacrysts from the group signature of sediment subduction combined with
I kimberlites (eg the Kimberley pool) actually subduction zone processes (ie depletion in nio-
have HIMU-type isotopic compositions (eg bium, strontium, phosphorus and titanium, a
Kramers 1977). The group II kimberlites have negative europium anomaly and an enrichment
higher 87Sr/86Srratios (0.707-0.710), lower ~Nd in lead). Alternatively, this crustal signature
values (--5 to --12), and less radiogenic lead may be derived from sediment subduction be-
isotopic compositions (2~176 ~ 17.2- neath Gondwana.
17.9) compared with the high P - T i (HPT) A magnification of this subduction-zone-
basalts of Parana and northern Karoo as well as related character is shown in the least fraction-
the Dupal islands of the southern Atlantic ated Mesozoic southern Karoo basalts (Fig. 12)
(Hawkesworth et al. 1986). As shown in Fig. 13, with about a 15% negative europium anomaly.
northern Karoo basalts have mantle-normalized These basalts have more radiogenic lead and
diagrams (especially for Ce/Pb and La/Nb) simi- strontium isotopic compositions but lower end
lar to but still not quite the same as Dupal values than group II kimberlites and the high
OIBs. If we accept the asthenospheric origin of phosphorus, titanium northern Karoo and
Dupal OIBs and at least a partial asthenospheric Parana basalts (Smith 1983; Mantovani et al.
origin for the group II kimberlites, it seems 1985; Hawkesworth et al. 1986). Additionally,
reasonable to suggest that the geochemical dif- the isotopic and chemical characteristics of
ference between them and the northern Karoo southern Karoo basalts are shared by the rela-
basalts is due to the involvement of the con- tively contemporaneous Ferrar dolerites of the
tinental lithosphere (including crustal contami- Trans-Antarctic mountains and the Tasmanian
nation) of southern Africa, which has a strong dolerites (Hergt 1987). Hergt (1987) has shown
1000
Northern Karoo Basalts
I
S r i = 0.7051 E Nd i = .6.8
FIG. 13. Mantle-normalized diagram of two selected least-fractionated basalts from northern Karoo. They
have similar 87Sr/86Srratios but lower end values than Dupal-type OIB from the S Atlantic Ocean. Data from
Duncan et al. (1984). See Table 1 for normalizing values.
Downloaded from http://sp.lyellcollection.org/ at Niedersachsische Staats Und Universitatsbibliothek on
March 7, 2012
0.7045
J HaleatalaI []
: Koolau I [] []
Tholeiitic Basalls 0.7040
0.7040
Basalts o non $ Haleakala
[] Kauai
Transitional Bnsalts u m Loihi
~n b IP Transitional
0.7035 0.7035 []
[] Basalts
M ~
Hawa~an
~tan~~ l J Alkalic Basalts
q~ AIkallc Basalts Hawaiian Basalts []
Basalts o n
[] n []
0.7030 MORB 0.7030 []
Source
[] Trendfor a small degree of melting of a MORB souse
["-'1 Trendfor partial melting effect >
0.7025 | I I I I I I I
Fro. 14. Plots of (a) S7Sr]S6Sr versus La/Nb and (b) 879r/S6Sr versus Ba/La for selected Hawaiian basalts
illustrating the combined effects of mixing of two types of mantle sources and different degrees of partial
melting. The trend of changing La/Nb and Ba/La ratios caused by different degrees of partial melting is
indicated by the arrow direction.
Downloaded from http://sp.lyellcollection.org/ at Niedersachsische Staats Und Universitatsbibliothek on
March 7, 2012
crust has an enriched pattern which is in part a Despite all uncertainties involved in this calcu-
mirror image of the depleted M O R B pattern. lation, it is clear that a major discrepancy be-
This complementary relationship between the tween the calculated and observed La/Nb ratios
crust and upper-mantle M O R B source is com- requires special attention. We have argued that
monly accepted as evidence for a mass balanced the lack of an obvious positive niobium anomaly
system. In Table 5 we use the primitive mantle in the average M O R B and OIB sources and
composition presented in Table 1 and estimates continental lithosphere requires a mechanism
of crustal abundances by Taylor & McLennan to fractionate niobium from lanthanum or pre-
(1985) to perform some simple calculations. To ferentially withdraw former oceanic crust with
satisfy crustal abundances of highly incompat- low La/Nb ratios from the upper mantle. This
ible elements (eg, barium, rubidium, caesium, can be achieved by subduction of oceanic litho-
thorium and uranium), 25%--30% of the sphere (Creager & Jordan 1984) or subtraction
volume of the upper mantle must be primitive of refractory eclogitic material from the mega-
mantle. Using higher estimates of element lith into the lower mantle and/or at the base of
abundances in the crust, up to 50 % primitive the upper mantle (Ringwood & Irifune 1988).
mantle is required. Furthermore, if the Preferential withdrawal of niobium from the
continental lithosphere (150-200 km thick) is upper mantle can also reduce the discrepancy
globally or locally enriched in incompatible ele- between the calculated and measured Nb/U
ments, then depletion of an even higher portion ratio of the depleted mantle shown in Table 4.
of the convecting mantle is required. Evidence A plot of La/Ta versus La/Sm ratios for MORBs
for enriched continental lithosphere comes from from the literature, along with our estimates o f
studies of continental basalts and ultrapotassic the primitive mantle, is shown in Fig. 15. It is
rocks (eg Carlson 1984; Fraser et al. 1986; quite clear that transitional-type MORBs and
Nelson et al. 1986). These rocks always show some N-type M O R B s with light-REE-depleted
strong crustal geochemical characteristics (rela- to slightly enriched patterns ( L a / S m ~ 0 . 8 - 1 . 8 )
tive depletions of titanium, niobium, phos- have tantalum (and niobium) anomalies (La/Ta
phorus and strontium on a m a n t l e - n o r m a l i z e d < 17). These results are consistent with the idea
diagram, high La/Nb (>> 1) and large negative of recycling and resorption of the subducted
europium anomalies) and isotopic systematics former oceanic crust into the asthenospheric
suggestive of crustal recycling in the early mantle source for these MORBs.
Precambrian.
According to the calculations in Table 5, the
Formation of heterogeneous mid-ocean ridge
residual M O R B source in the upper mantle
basalt sources
should have a fertile major- and trace-element
composition, similar to those of the fertile ultra- It appears that recycled former oceanic crust is
mafic xenoliths studied by Jagoutz et al. (1979). not simply resorbed and mixed into neighbour-
TABLE 5. Mass balance calculation assuming complementary relationship between the bulk
continental crust and M O R B source (upper mantle)
Rb U Nb La Ce Pb Nd Sr Sm AI% Ca%
~ -MORB ~ Ta
Anomaly
in the variation of chemical and isotopic charac-
teristics in different ocean basins (eg the exist-
ence of the Dupal anomaly in the Indian Ocean).
A veined and plum-pudding type of upper
mantle with various scales of old and young
heterogeneity can be generated this way.
T- M O R B - ~ . ~ . ~
R = ReducedEnvironment
A
S S S S A P = Phlogopite held back
A A A
S S = Sediment Addition
A
100 .tt A = Arc Signature
:~ ~o
P P P A
S P
A
g
1 ' ' ' , , i , , , l l i l , i , l l l i l ,
7s Rb Ba Th U Nb K La Ce Pb Pr Sr P N d Z r S m E u TiDy Y YbLu
T1 W Ta Mo Hf Sn Li 1to
1~. 16. Schematic description of the effects of various factors which can modify the mantle-normalized pattern
of OIBs. For some elements enrichments or depletions may be due to a combination of factors; however, by
using several different elements it may be possible to distinguish the relative roles of multiple effects. For
example sediment addition (S) and arc signature (A) will both cause increases in caesium, rubidium, barium,
potassium and lead; however, strontium is generally enriched in the arc signature, whereas it is generally
depleted in the sediment addition.
studies of element ratio systematics (eg the core - mantle boundary, as a number of
Ba- Rb-Cs, Nd-Sr, Nb-La, P-Nd, Ti-Eu, geophysicists seem to prefer, then there are
S m - Z r ) and the effect of changing mineralogy three possible ways in which to explain these
on Kd under different pressure and temperature constant ratios: (i) the recycled oceanic litho-
conditions. Such differences in mineral assem- sphere which sank to the c o r e - m a n t l e bound-
blage and melting conditions may also be re- ary about 2 Ga ago becomes the plume source,
sponsible for two types of light R E E patterns (ii) the upper and lower mantle are well mixed
observed in OIBs: (i) the predominant pattern and (iii) mantle material that arises from the
showing a continuous increase from heavy R E E s core - mantle boundary only initiates the plume
to light R E E s with La/Ce = 0.46 ( + 1 0 % ) and activity, whereas the main chemical and isotopic
(ii) a less common type showing concave- characteristics of OIBs are derived from shal-
downward light R E E patterns as in Ua Pou lower mantle (1000 km or less).
tholeiites (Fig. 11) and most Hawaiian It has been shown by some geophysicists (eg
tholeiites. see a review by Hager & Gurnis 1987) that the
If whole-mantle convection was important in global gravity pattern is consistent with whole-
the early Earth's history, then we suggest that mantle convection, with viscosity increasing
the lower mantle composition would also have with depth across the phase transformation
been involved in crust formation and mantle ( 1 0 - 1 0 0 times), and that oceanic lithosphere
differentiation ( M O R B - O I B ) . The total ab- subduction can continue into the lower mantle
sence of OIBs with certain primitive mantle (Creager & Jordan 1984). It has also been shown
ratios (eg Rb/Cs ~ 20, Nb/U -~ 30, Ce/Pb ~ 9) that the geoid pattern reflects plate tectonic
suggests that the different OIB-type sources patterns of the last 100-200 Ma (eg Hager &
have been previously well mixed with the de- Gurnis 1987). This whole-mantle convection
pleted M O R B reservoir (Hofmann et al. 1986). model has the property of faster convective
If some OIB sources are indeed derived from mixing in the upper mantle than the lower
Downloaded from http://sp.lyellcollection.org/ at Niedersachsische Staats Und Universitatsbibliothek on
March 7, 2012
amount of continental lithosphere available for In conclusion, in order to improve our inter-
recycling and a general lack of crustal and pretation of the chemical and isotope data, it is
subduction zone signatures in OIBs, we favour important to integrate a geochemical approach
the idea that the H I M U source character (es- with other approaches based on information
pecially those with 87Sr/S6Sr=0.7026-0.7028) is derived from high-pressure experimental petro-
generated by recycling of oceanic lithosphere. logy, the tectonic history of the Earth's crust,
Due to the problem of mantle mixing, no geophysical observations and numerical model-
unique isotope isolation age can be established ling. With a better understanding of the petro-
for E M l - t y p e mantle with confidence, whereas logical and geophysical processes involved in
trace-element and isotope data of the EM2-type the dynamic, chemical and isotopic properties
mantle with 2~176 _> 19.0 seems to suggest of the continental lithosphere we shall have a
the involvement of recently subducted sedi- more realistic flux input for the 'multi-box'
ments. A range of source isolation ages from a models currently used in numerical calculations
few hundred million years to 2 billion years for the chemical and isotopic evolution of the
therefore seems likely for the EM-type re- Earth.
servoirs.
The three-dimensional geometrical relation- ACKNOWLEDGEMENTS: Constructive suggestions and
ship between the different OIB-type sources comments from F. A. Frey, C. J. Hawkesworth, J.
and the depleted upper mantle asthenosphere Hergt, B. M. Jahn, A. L. Jaques, M. T. McCulloch,
( M O R B source) cannot be clearly established Z. A. Palacz, R. Price, R. L. Rudnick, A. D.
as yet. However, it is evident that post-erosional Saunders and W. M. White are gratefully acknowl-
edged. S.-S. Sun publishes with the permission of the
alkali basalts of the Hawaiian islands and
Director, Bureau of Mineral Resources, Geology and
Samoan islands were derived from shallower Geophysics, Australia.
depths, whereas many OIB sources are derived
from greater depths in the form of mantle
plumes or isolated blobs.
References
ALIRERT, C., MICHARD, A. & ALBAR~.DE, F. nature of the subcontinental mantle. Geochimica
1986. Isotope and trace element geochemistry of et Cosmochimica Acta 48, 2357-2372.
Colorado plateau volcanics. Geochimica et CHEN, C-Y & FRE~', F. A. 1985. Trace element and
Cosmochimica Acta 50, 2735-2750. isotopic geochemistry of lavas from Haleakala
ALL~GRE, C. J. & TURCOTrE, D. L. 1986. Impli- Volcano, East Maui, Hawaii: implications for
cations of a two-component marble-cake mantle. the origin of Hawaiian basalts. Journal of Geo-
Nature 323, 123-127. physical Research 90, 8743-8768.
, HAMELIN,B., PROVOST,A. & DUPR~, B. 1987. CLAGUE, D. A. & FREY, F. A. 1982. Petrology and
Topology in isotopic multispace and origin of trace element geochemistry of the Honolulu
mantle chemical heterogeneities. Earth and Volcanics, Oahu: implications for the oceanic
Planetary Science Letters 81,319-337. mantle below Hawaii. Journal of Petrology 23,
ANDERSON, O. L. 1982. Hotspots, polar wander, 447-504.
Mesozoic convection and the Geoid. Nature 297, COHEN, R. S. O'NIONS, R. K. 1982. Identification of
391-393. recycled continental meterial in the mantle from
ARMSTRONG, R. L. 1968. A model for the evolution Sr, Nd and Pb isotope investigations. Earth and
of strontium and lead isotopes in a dynamic Planetary Science Letters 61, 73-84.
Earth. Reviews of Geophysics 6, 175-199. Cox, K. G. 1978. Flood basalts, subduction and the
- - 1981. Radiogenic isotopes: the case for crustal break-up of Gondwanaland. Nature 274, 47-49.
recycling on a near-steady-state no-continental CREAGER, K. C. & JORDAN, T. H. 1984. Slab pen-
growth earth. Philosophical Transactions of the etration in the lower mantle. Journal of Geo-
Royal Society of London 301,443-472. physical Research 89, 3031-3049.
BELL, K. & BLENKINSOP,J. 1987. Nd and Sr isotopic DAVIES, G. F. 1981. Earth's neodymium budget and
compositions of East African carbonatites: impli- structure and evolution of the mantle. Nature
cations for mantle heterogeneity. Geology 15, 290, 208-213.
99-102. DUD,~S, F. O. CARLSON, R. W., & EGGLER, D. H.
BROOKS, C., JAMES, D. E . & HART, S. R. 1976. 1987. Regional middle Proterozoic enrichment
Ancient lithosphere: its role in young continental of the subcontinental mantle source of igneous
volcanism. Science 193, 1086-1094. rocks from central Montana. Geology 15, 22-25.
CARLSON, R. W. 1984. Isotopic constraints on DUNCAN,A. R., ERLANK,A. J. & MARCH,J. S. 1984.
Columbia River flood basalt genesis and the Regional geochemisty of the Karoo igneous
Downloaded from http://sp.lyellcollection.org/ at Niedersachsische Staats Und Universitatsbibliothek on
March 7, 2012