Trace Element Discrimination Diagrams for the

Tectonic Interpretation of Granitic Rocks

by JULIAN A. PEARCE*, NIGEL B. W. HARRIS AND

ANDREW G. TINDLE
Department of Earth Sciences, The Open University, Milton Keynes, MK76AA, Bucks,
England
(Received 2 February 1984; in revisedform 19 April 1984)

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on April 29, 2014

ABSTRACT
Granites may be subdivided according to their intrusive settings into four main groups—ocean ridge
granites (ORG), volcanic arc granites (VAG), within plate granites (WPG) and collision granites
(COLG)—and the granites within each group may be further subdivided according to their precise
settings and petrological characteristics. Using a data bank containing over 600 high quality trace
element analyses of granites from known settings, it can be demonstrated using ORG-normalized
geochemical patterns and element-SiO2 plots that most of these granite groups exhibit distinctive trace
element characteristics. Discrimination of ORG, VAG, WPG and syn-COLG is most effective in
Rb-Y-Nb and Rb-Yb-Ta space, particularly on projections of Y-Nb, Yb-Ta, Rb-(Y + Nb) and
Rb—(Yb + Ta). Discrimination boundaries, though drawn empirically, can be shown by geochemical
modelling to have a theoretical basis in the different petrogenetic histories of the various granite groups.
Post-collision granites present the main problem of tectonic classification, since their characteristics
depend on the thickness and composition of the lithosphere involved in the collision event and on the
precise timing and location of magmatism. Provided they are coupled with a consideration of
geological constraints, however, studies of trace element compositions in granites can clearly help in the
elucidation of post-Archaean tectonic settings.

INTRODUCTION
Trace element discrimination diagrams have been in use for some time as a means of
fingerprinting the tectonic setting of eruption of basic volcanic rocks from the geological
record (e.g. Pearce & Cann, 1973; Floyd & Winchester, 1975; Pearce, 1975; Wood et al.,
1979; Winchester & Floyd, 1977; Shervais, 1982). There are, however, many areas where the
only exposed products of a magmatic/tectonic event are plutonic rocks, particularly granites
(sensu lato). Our aim is to extend trace element discrimination diagrams to this rock type
which we define for the purposes of this paper as any plutonic igneous rock containing more
than 5 per cent of modal quartz.
There are two main reasons why granites have received less attention than basalts as
tectonic indicators. The more important has been the difficulty of sampling granites of known
setting: by the time they are exposed at the surface, unambiguous geological evidence for the
tectonic setting at the time of intrusion is often difficult to obtain. The second reason is the
more complicated petrogenetic history of granites, which can make their chemical
compositions difficult to interpret: crystal accumulation, involvement of continental crust,
redistribution and loss of elements by volatile fluxing and crystallization of trace element-rich
minor phases are generally of little relevance to basalt genesis yet can obscure the important
geochemical features of granites (e.g. Hanson, 1978). These problems are, however,

* Present address: Department of Geology, University of Newcastle, Newcastle-upon-Tyne, England.

[Jouraml of Pttrolojy, Vol. 25, Pin 4, pp. 956-983, 19841
 TRACE ELEMENTS OF GRANITIC ROCKS 957

counterbalanced in part by the generally much lower degree of alteration suffered by granites
so that the mobile, as well as the immobile, elements (Cann, 1970) can often be used in their
discrimination. Of course, a number of effective schemes for the classification of granites do
already exist and have been shown to have tectonic connotations. Both Peacock's (1931)
alkali-lime index and Shand's (1951) subdivision into 'peraluminous', 'metaluminous' and
'peralkaline' are still used as indicators of the major element characteristics of granites and
have led to the commonly-held (though somewhat simplistic) assumptions that 'calc-alkaline'
granites are the products of volcanic arc magmatism, that 'alkaline' and 'peralkaline' magmas
are associated with within plate settings and that 'peraluminous' granites result from the
anatexis of sedimentary rocks, mainly during continent collision. Streckeisen's (1976)
classification based on modal compositions has also been shown to provide information on
tectonic setting (Lemeyre and Bowden, in press), while Debon & Le Fort (1982), building on

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on April 29, 2014

earlier work by La Roche (1978), have published a set of'characteristic minerals' diagrams
which incorporate the information on tectonic setting inherent in both the chemical and
mineralogical classifications. The subdivision of granites into S-type and I-type (Chappell &
White, 1974; White & Chappell, 1977) has also been extended from its initial genetic concept
to a tectonic indicator in which S-type granites are the assumed product of continent collision
and I-type granites are the assumed product of cordilleran and post-orogenic uplift regimes
(e.g. Beckinsale, 1979; Pitcher, 1983). In accordance with this change in emphasis, A-type
(Collins et al., 1982) and M-type (White, 1979) granites were defined to encompass the
granites of anorogenic and oceanic arc settings respectively. The latter can also be extended
to include the oceanic plagiogranite group of Coleman & Peterman (1975) which covers the
sodic granites of ophiolite complexes that are assumed to have formed at mid-ocean ridges.
Useful though the above classification schemes are, they suffer from major defects as
indicators of past tectonic settings. The mineralogical and major element classification
schemes generally make poor discriminants because they rely on only a few variables, none
chosen for the purpose of tectonic classification. The classification into S-type, I-type, A-type
and M-type granites is difficult to apply because there are no well-defined boundaries between
the granite types and because the simple correlations between these granite types and tectonic
settings do not always hold, as we demonstrate later in this paper. We therefore adopt the
opposite approach to the classification of granites, starting with granites of known tectonic
setting and investigating the extent to which they can be discriminated on the basis of their
geochemical and mineralogical characteristics. The granite data bank set up for this purpose
contains some 600 analyses which are divided on a priori grounds into ocean ridge, volcanic
arc, within-plate and collision settings. The elements considered are the XRF trace elements,
K, Rb, Sr, Y, Zr and Nb (and sometimes Ce, Ba and Th) and the INAA trace elements, Ba,
the rare earth elements, Hf, Ta and Th; information on mineralogical and major element
characteristics has also been recorded. About a third of the analyses were carried out at the
Open University (Potts et al., 1981; 1984) and the remainder have been carefully checked
to ensure comparability with these data. This paper first looks in detail at the tectonic
settings represented by the granites in the data bank, then attempts to devise diagrams for
relating geochemistry to tectonic setting, and finally looks at the petrogenetic rationale for,
and limitations of, these diagrams.

TECTONIC CLASSIFICATION OF GRANITIC ROCKS

Table 1 lists the locations of the granites on which the data bank is based. They have been
divided into four main groups: ocean ridge, volcanic arc, within plate, and collision granites.
Each group has then been further divided into tectonic and petrologic sub-groups where
 TABLE 1
Sample locations for the granitic rock analyses used in this paper
Ocean ridge granites {ORG) Volcanic arc granites (VA G) Within plate granite {WPO) Collision granites (COLG)

{a) 'Normal' ocean Corsica (10) (Beccaluva el al.. (a) Oceanic. Late Intrusive Complex. (a) Intra- Nigeria (45) (Bowden & (a) Continent- Hercynian belt: S.W. England
rulgcs (associated 1977; Ohnenstetter & mainly Oman (17*) (Alabaster M continental Turner. 1974; Bowden & continent (6*) (AUerton « al.. 1980);
with N-type OhnenstettCT. 1980): thokiilic al.. 1982): Little Port ring com- Whitley. 1974; Imeokpana, collision: N Portugal (21) (Albuquerque.
MORB) Tujcany. Italy (9*) (G. Sern arcs complex, Newfoundland (6) plexesond 1982. 1983): Sabaloka syn-tectomc 1971. 1978): NW Spain (12)
& J. Pcarcc unpubl. data): (Malpas, 1979). Canyon graben complex. Sudan (6*) (Hams (Capdcvila 1969: Capdcvila &
Indian Ocean (2) (Engcl & Mountain. OTegon (2) rial.. 1983): Viaietta, 1970): Barouise
Fisher. 1975) (Gerlach « a/_ 1981) Bayuda desert. Sudan (10) Massif. Pyrenees (7*) (this
(/>) Anomalous Mid-Atlantic Ridge, 45° N (8*) (/>) Oceanic. Jamaica (20*) (Isaacs, 1975 (Curtis. 1979); Oslo graben paper). Himalayan belt:
ocean ridges (Aumcnto. 1969.Aldiss, mainly cak- and this paper): SW Pacific (54*) (Neumann el al.. 1977; Manaslu (8) ( U Fort. 1981)
(associated with 1981) alkoline arci (44) (Mason & MacDonald, Khalu rl al.. 1978: this paper) (7) (Le Fort. 1981; Vidal el aL
T o r E type 1978): Aleutians (21) (Pcrfit 1982: Cocherie, 1978). Llagoi
MORB) rial.. 1980: Hill era/., 1981) Kangri belt (JO*) (this paper). -a
(c) Back arc basin Sannicnto complex. Chile (8) (r) Acuvc Antarctic Peninsula (13) Bhutan (23) (Dietrich & m
ridges (Saunders el al.. 1979); Bay continental (Saunders et o/_ 1980); Gansser. 1981); Gabug. TibeJ
of Islands complex. margin Central Chile. 26° S (20*) (3*) (this paper). SE Asia:
Newfoundland (7) (Malpas. (Baldwin* Pearce, 1982 Yunnan (2*) (this paper) n
1979): Smartville compkx. and this paper); Central (6) Attenuated Skaergaard (2*) (this paper): {b) Continent- Hercynian belt: Querigut, Pyrenees m
SicTra Nevada (7*) (AMiss. Chile, 22° S (40*) (G continental E and SE Greenland (15) continent (IO)(Fourcade& Allegre.
ded from http://petrology.oxfordjournals.org/ at Aston University on April 29, 2014

1978) Rogers, unpubl. data); Chile crust (B. Upton, unpubl. data). collision. 1981 ):Gahcia (5) (Cocherie.
33-34° S (17) (Lopez-Escobar Mull. Scotland (20*) (Walsh late post- 1978): Barousse Massif,
rial. 1979). Peru (13*) (A era/. 1979) tectonic Pyrenees (2*) (this paper).
Bussell. unpubl. data): Alpine belt Vedrette di Ries,
Tuolumne batholith. Sierra Eastern Alps (21) (Bellieni el al..
Nevada (22) (Batemtn & 1981): Novate and Bergell. Italy
Chappell. 1979) (3*) (this paper): Adamclk). Italy
(d) Supra-subduction Troodos Massif. Cyprui (27*) (17) (Dupuy Ma/.. 1982)
lone (fore-arc (Aldiss. 1978);Semail (c) Oceanic Ascension Is. (4*) (this paper); (c) Continent- Oman (2*) (this paper), Masirah
basin) ridge Nappe. Oman (5*) islands Reunion (3) (Fisk el al. in arc collisions: Is. (8) (AbtxXts. 1978): Fano
(Alabaster el aL 1982): press) syn-orogenic granite. Greece (2*) (this paper):
Antalya complex. Turkey (8) Bolivia (2*) (this paper)
(Cocherie. 1978). Vourinos
complex. Greece (4*) (this
paper)

Figures in parentheses give the number of analysej per location. Asterisks indicate Open University analyses
 TRACE ELEMENTS OF GRANITIC ROCKS 959

possible. The principal geological and petrological characteristics of each granite type, and
the rationale behind the groupings, are further discussed below.

Ocean ridge granites

Although a few samples of granitic rocks have been recovered directly from the deep ocean
basins by dredging (Engel & Fisher, 1975; Aumento, 1969; Aldiss, 1981; Sharaskin et al., in
press), most of our knowledge of these granites comes from ophiolite complexes, where they
typically occur as small, localized bodies in the uppermost parts of plutonic units. Coleman &
Peterman (1975) and Coleman & Donato (1979) proposed that the general descriptive term,
'oceanic plagiogranite', be used for these rocks, but as this term implies the inclusion of
granites from intra-oceanic settings other than ocean ridges, the more specific term, 'ocean

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on April 29, 2014

ridge granite', is used here. This group therefore encompasses only granitic rocks that either
belong to the oceanic crust itself or form part of layered ophiolite sequences, and specifically
excludes intrusive bodies that cut across these sequences and which could therefore have
island arc or oceanic island affinities.
It is also useful to try to define more precisely the type of ridge at which a given ocean ridge
granite was formed. This has been done for the granites in Table 1 by combining evidence
from their geological setting with the evidence provided by the geochemical characteristics of
their associated basalts. The initial grouping is into subduction-unrelated (a and b in Table 1)
and subduction-induced (c and d in Table 1) ridges. Of the former, we have described the
ridges as 'normal' if the principal volcanic product is an N-type MORB and 'anomalous' if
the product is an E- or T-type MORB, i.e. enriched in incompatible trace elements (Sun &
Nesbitt, 1977; Wood, 1979). Geological evidence demonstrates that only the Corsican,
Apennine and some Iranian ophiolite plagiogranites formed at ridges unrelated to subduction,
and geochemical evidence classes all these ridges as 'normal' (see references in Table 1). Our
only data from an 'anomalous' ridge come from the plagiogranites from 45° N on the
Mid-Atlantic Ridge, one of the classic localities for eruption of E-type MORB. No data are
yet available, however, on plagiogranites from 'anomalous' oceanic ridges in the vicinity of
major sub-oceanic mantle plumes such as Bouvet or Iceland.
The subduction-induced ridges can also be described as 'normal' if the principal volcanic
product is an N-type MORB, but as 'supra-subduction zone (SSZ)' if the product is of island
arc tholeiite or boninite character. Of the plagiogranites used in this study, those from the
Sarmiento, Bay of Islands and Smartville complexes are taken from ophiolites whose
geological characteristics (e.g. overlying volcaniclastic sediments) indicate a back-arc setting
of formation yet whose basalts have similar geochemical characteristics to basalts erupted at
ridges of 'normal' type. By contrast, the plagiogranites of the Troodos, Semail, Antalya and
Vourinos complexes are taken from ophiolites whose geochemical characteristics indicate an
origin at a ridge of SSZ type (e.g. Pearce et al., in press). There is some geological evidence
(Gealey, 1980) that these plagiogranites may have occupied a fore-arc setting similar to that
of the plagiogranites recently dredged from the inner wall of the Tonga trench (Sharaskin et
al., in press).
The plagiogranites of all tectonic sub-groups have hornblende as the dominant
ferromagnesian mineral, plot as quartz diorites or tonalites on the Streckeisen diagram, and
may be metaluminous or peraluminous. The only significant exception is a quartz monzonite
from the Indian Ocean described by Engel & Fisher (1975) which, being a narrow vein, has
not been included in the data bank. However, variation between the sub-groups can be
detected using the Peacock index, plagiogranite suites from normal and anomalous
subduction-unrelated ridges classifying primarily as alkali-calcic, those from normal back-arc
960 J. A. PEARCE ET AL.

ridges classifying as calc-alkalic and those from SSZ ridges classifying as calcic. This appears
to be a significant difference, despite the small sample size.
Volcanic arc granites
Volcanic arc granites can vary in setting from oceanic to continental and in composition
from tholeiitic through calc-alkaline to shoshonitic (Peccerillo & Taylor, 1976). At one end of
the spectrum are those associated with primitive tholeiitic oceanic arcs. The only available
data for this sub-group come from the Canyon Mountain complex, the whole of which
appears to have tholeiitic island arc affinities, and from the small intrusions (<10 km across)
and sill complexes which cut ophiolite complexes of subduction-related origin such as the
Semail Nappe, Bay of Islands and the Sarmiento complex (see references in Table 1) and
which crystallized in the magma chambers that fed the upper lava units of these complexes. In

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on April 29, 2014

these cases the geological evidence is consistent with, but not conclusive proof of, an island
arc origin and the geochemistry of associated basalts has therefore been used to confirm this
interpretation. Data on granites from active or recently-active oceanic arcs were available
from regions of rapid erosion such as the SW Pacific and the Caribbean and also from the
accretionary prism of Alaska. These do include rocks of the tholeiitic series but calc-alkaline
rocks are more common.
At the other end of the spectrum are granites intruded at active continental margins. All
data on this group come from the granites of the Western Americas belt, and particularly
from the Western Cordillera of the Andes where Jurassic-Cretaceous composite batholiths
are exposed along the present coastline and smaller intrusions of Cretaceous and Tertiary age
are exposed inland. All these intrusions belong to the calc-alkaline, high K calc-alkaline or
shoshonitic series as defined by Peccerillo & Taylor (op. cit.).
In selecting samples and analyses of volcanic arc granites for the data bank, care was taken
to include only those that resulted from subduction of oceanic crust. Granites such as those of
the Eastern Cordillera of the Andes, which are thought to be the product of arc-continent
collision events rather than simple subduction (Bourgois & Janjou, 1981), are assigned to the
collision rather than to the volcanic arc group. In addition, some granite suites, such as the
Tertiary Himalayan granites described by Honeggar et al. (1982), have not been included in
the data bank because of the difficulty in distinguishing between subduction-related and
collision-related intrusions.
The granites from the volcanic arc setting vary significantly and systematically in their
major element and mineralogical characteristics. Granites from the oceanic, tholeiitic arcs
plot predominantly in the quartz diorite and tonalite fields on a Streckeisen diagram, have
hornblende as the principal ferromagnesian mineral, and are 'calcic' according to Peacock's
classification. Granites from the calc-alkaline arcs (whether oceanic or continental) plot
primarily in the quartz diorite, quartz monzonite, tonalite and granodiorite fields of the
Streckeisen diagram, have hornblende and biotite as their characteristic ferromagnesian
minerals, and generally belong to 'calc-alkalic' suites according to Peacock's classification.
Granites from the high K calc-alkaline and shoshonitic series of active continental margins
plot predominantly in the fields of quartz monzonite, granodiorite and granite (s.s.) on the
Streckeisen diagram, have biotite ± hornblende as the dominant ferromagnesian minerals and
belong to calc-alkalic and alkali-calcic suites according to Peacock's classification. All of the
granites of sub-groups (a) to (c) can vary from metaluminous to slightly peraluminous and
from M-type (in sub-groups (a) and (£)) to I-type (in sub-groups (b) and (c)).
Within plate granites
Within plate granites can be technically subdivided according to the nature of the crust
into which they were intruded. Those intruded into continental crust of close to normal
 TRACE ELEMENTS OF GRANITIC ROCKS 961

thickness (i.e. the granites from Nigeria, Sudan and the Oslo graben in Table 1) have been
assigned to one sub-group (a), those introduced into strongly attenuated continental crust (i.e.
the Tertiary granites from Greenland and Scotland in Table 1) have been assigned to a second
sub-group (b), while those intruded into oceanic crust (i.e. the granites of Ascension Island
and Reunion in Table 1) have been assigned to a third sub-group (c). The dividing line
between sub-groups (a) and (b) is arbitrarily defined as the appearance of associated dyke
swarms, which corresponds to a beta-value (unattenuated divided by attenuated lithosphere)
of about 5 (e.g. Dewey, 1982); the dividing line between sub-groups (b) and (c) is taken as the
shelf edge.
The majority of the granites in the data bank (most from sub-groups (a) and (c) and some
from (b)) plot in the quartz syenite, granite (s.s.) and alkali granite fields on the Streckeisen
diagram, belong to 'alkalic' suites according to Peacock's alkali-lime index and, by definition,

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on April 29, 2014

fit the description of A-type granites. On other criteria, they can be highly variable: their
ferromagnesian phases range from sodic amphiboles ± sodic pyroxenes to biotites ± sodic
amphiboles and they vary from peraluminous to peralkaline in character. Most of the
sub-group (b) samples, however, belong to calc-alkalic suites, contain calcic (rather than
sodic) amphiboles and pyroxenes, sometimes together with olivine, and are metaluminous
according to Shand's index.

Collision granites
Granites are the major magmatic products of most, if not all, collision belts, and may be
subdivided tectonically according to the type of collision involved (continent—continent,
continent-arc, arc-arc) and to the temporal relationship with the major deformation event
(syn-collision, post-collision). Most of the granites in the data bank are from continent-
continent collision zones and divide almost equally between the syn-tectonic and post-tectonic
groups (see Harris et al., in press, for a more detailed description of sample locations). The
vast majority come from the Hercynian, Himalayan and Alpine belts where the relationships
between magmatism and tectonism are best understood. Three examples of granites intruded
during arc-continent collision are also included: the Upper Cretaceous granites of Oman and
Masirah Island which are related to the emplacement of the Semail ophiolite of Oman; the
Jurassic Fano granite of Northern Greece, which is related to the emplacement of the
Guevgueli ophiolite complex (Bebien, 1982); and the Miocene granites of Bolivia, which were
intruded during collision between the South American continent and the Western Cordillera
of the Andes (Bourgois & Janjou, 1981). Of these, the Oman and Masirah Island granites are
treated as post-tectonic and the Greek and Bolivian granites as syn-tectonic.
The syn-tectonic granites typically plot within the granite (s.s.) field on the Streckeisen
diagram, are muscovite-bearing and peraluminous and exhibit most of the features associated
with S-type granites. The post-tectonic granites most commonly contain biotite ± hornblende
as ferromagnesian minerals, plot in the same region as volcanic arc (sub-group (c)) granites
on the Streckeisen diagram, belong to calc-alkalic suites, are metaluminous to slightly
peraluminous, and exhibit most of the characteristics of I-type granites. However, it should be
noted that granites of S-type (e.g. the Alpine Novate granite) and A-type (e.g. the Hercynian
Grandes Rousses Massif) can also be intruded in this setting, although they are not
well-represented in the data bank at the present time.

TRACE ELEMENT C H A R A C T E R I S T I C S OF THE GRANITE TYPES

Table 2 lists the geochemical analyses of some representative granites from the four main
tectonic settings and their respective sub-settings. For comparability, we have listed analyses
which have SiO2 values as close as possible to 70 per cent for the localities they represent. To
962 J. A. PEARCE ET AL.

TABLE 2
Representative analyses of granites from the various tectonic settings and sub-settings
Ocean ridge granites Sub-group Volcanic arc granites

Sub-group (a) (A) (c) (<0 (o) (A) (c) W)

Tuscany SmarttJtlle MAB.45" N Troodot Oman Little Port Jamaica Chile

S.0, 7213 69-07 72-47 75-33 SiO, 70-1 69-50 68-43 74-50
TiO, 0-23 0-51 0-33 0-23 TiO, 0-53 0-16 0-33 016
Al,Oj 14-60 14-41 1417 12-92 AI,O, 1200 14-60 14-44 12-52
Fe,O, 2-08 4-82 3 16 3-65 Fe,O, 4-85 2-43 3-35 100
MnO 0 02 008 0-08 005 MnO 013 006 006 001
MgO 0-60 0-52 1-39 0-37 MgO 0-80 •01 1-35 0-07
CaO 038 2-65 1-48 2-66 CaO 306 1-72 2-36 0-29
Na,0 8-78 801 5-55 518 Na,O 4-41 5-33 3-70 3-56
K.,0 010 0-04 0-24 0 14 K,0 0-28 1-22 313 5-51

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on April 29, 2014

P,O, 0-29 0-09 006 004 P,O, 0 II 0-07 0-12 0-01
LOI 0-78 n.a. 100 [La. LOI 2-55 1-94 3-21 0-94

Total 99-99 100-20 99-83 100-57 Total 99-42 9804 100-48 98-57

Rb <2 <2 <2 <2 Rb <2 20 63 169

Sr 19 105 89 130 Sr 200 274 210 93
Ba n.a. 89 149 137 Ba 56 81 750 331
Zr 432 512 (285) 93 Zr 99 73 104 184
Y 129 88 (55) 35 Y 44 19 10 30
Nb 20 14 (42) 5 Nb 2 6 9 17
Th 2-80 1-8 (3-4) 1-1 Th 0-32 ni 4-30 20-33
Ta 2 10 1-4 3-2 (0-35) Ta 013 [LC 0-79 1-75
Hf 16-95 11-8 7-6 2-7 Hf 3 03 n.a 2-89 5-73
La 23-4 n.a. 43-8 n.a. La 4-0 n.a 17-7 40-1
Ce 76 0 56-9 81 7 14 9 Ce 10-7 n.s 31-6 85-8
Nd 40-8 42-0 35-1 10-6 Nd 10 3 1LI 15-4 36-2
Sm 100 11-6 7-7 3-6 Sm 3-7 n.i 2-2 8-4
Eu 1 45 2-62 1-62 1 48 Eu 1-11 n.a 0-67 0-76
Gd 11-0 13-1 8-8 4-4 Gd 5 0 na. n-a. n.a.
Tb 2-27 2 6 1-3 0 94 Tb 0-97 [La. 0-30 1 15
Yv 15-03 1106 6-9 4-26 Yb 4-54 n.a. 1-37 312

Within plate granites

Sub-group (i) (a) (A) (A) (c)

Oslo R(ft Sabaloka Skaergaard Mull Ascension
Islands

SiO, 73-16 70-41 76-02 73-26 71-61

TrO, 0-26 0-31 0-29 0-33 019
AI,O, 12 28 1301 12 60 12-55 11-68
2 72 3-79 1-95 3-73 4 48
MnO 0-20 005 002 008 014
MgO 0-10 • 012 0-08 008 017
CaO 0-23 1-34 006 1-07 0-13
Na,0 6-14 2-90 4-07 405 5-50
K,0 4 58 5-46 4-40 4-47 4 68
P,O, 0-04 006 n.d. 0-03 nA.
LOI (La. 1-74 0-76 008 008

Total 99-72 9919 99-25 99-73 98-77

Rb 251 127 97 187 94

Sr 6 40 75 59 1
Ba (80) 195 545 970 53
Zr 827 509 258 520 1089
Y 55 84 57 103 92
Nb 226 56 24 21 168
Th (3) 18 906 18 03 24
Ta (15) 3-0 1-54 211 16
Hf (21) 15-5 7-40 14-86 42
La 72 196-9 39-4 53-4 90-6
Ce 139 4040 96-4 119-4 274-4
Nd njt 1824 38-6 580 122-2
Sm 9-8 29-2 7-7 14-2 17-2
Eu 1-38 1-84 106 2-36 209
Gd [La. n.a. ni. 14-96 nx
Tb 1-30 316 1-21 2-36 3-98
Yb 9-3 7-33 400 10-18 16-50
 TRACE ELEMENTS OF GRANITIC ROCKS 963

TABLE 2 (continued)
Collision granites
Sub-group (a) («) (a) (ft) (a) (ft) (6) (<•)
Yunnan Gabug SW Novalt Barousse Vedrttte Queripil Oman Bolivia
(SW China) (Tibet) England (Italian (Pyrenees) dtRtes (Pyrenees)
Alps) (EAIps)

SiO, 7318 74-56 71-73 76-20 73-89 70-69 72-3 72-3 65-63
TiO ; 0 25 018 0-25 0-24 007 021 018 0-26 0-66
AI,O, 15 03 13 34 14 55 14-21 15-80 15-31 1415 14-11 14-98
Ftp, 1 14 1-63 2-17 0-28 0-26 2-19 1-91 2-33 4-73
MnO 001 005 0 03 0-03 000 0-07 0-04 010 001
MgO 0-45 0-44 0 46 Oil 0-20 0-48 0-37 219 2-40
CaO 1 49 1-58 0-69 105 114 2-76 1 33 2-63 1-64
Na,0 3-40 2-92 2-52 3 61 4-37 3-29 3-49 2-70 3-54
K,0 5-54 4-45 5-50 4-71 3 88 410 4-56 3 38 5-83
P,O, 010 003 0-23 008 016 008 nju 015 001

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on April 29, 2014

LO1 0-25 0-81 1-36 0-23 0 76 0 62 1-39 nj. n.a-

Total 100-84 10000 99-38 100-55 100-63 99-80 99-72 10017 99-43

Rb 210 243 488 104 108 171 238 76 416

Sr 198 75 96 148 156 339 80 117 213
Ba 705 389 n.a. 484 260 n.a. 320 201 559
Zr 165 99 80 47 24 92 (160) 108 132
Y 7 25 13 23 15 (15) (22) 20 24
Nb 6 16 13 4 10 (16) (14) 8 12
Th 23-42 34-92 19-6 3-22 (1) 8-4 14-0 11-1 14-2
Ta 0-6 2-6 1-55 189 1-8 2-6 2-45 0-9 1-4
Hf 4-86 3-95 3-12 1-45 0 76 3-2 4-2 2-7 3-6
La 32-7 31-7 33-8 7-5 6-6 20 23-7 35-3 45-1
Ce 54 7 56-3 70-3 14-0 96 36 44 73-2 66-1
Nd 191 22-3 33-5 7-0 IU 16 23 28-1 29-2
Sm 3-0 4 8 7-4 2-5 1-3 3 6 51 5-2 6-2
Eu 0-72 0-57 0-76 0-35 0-66 0-88 9 53 1-17 1-21
Gd nj. n.a. n.a- 3-1 [U. n.a. QJ. njt 5-3
Tb 0-2 0-82 0 55 0-54 0-27. 0-35 0-95 0-62 0-73
Yb 06 3-53 105 2-48 113 1-5 2-3 1-64 213

All analyses carried out at the Open University except Vedrette di Ries (Bellieni el al, 1981), Little Port
Complex, Newfoundland (Malpas, 1979) and Querigut (Fourcade & AUegre, 1981). n.a. = not analysed;
n.d. = not detected; estimated values in brackets. For details of sub-groups, see Table 1.

illustrate the key features of these analyses we have plotted them as geochemical patterns in
Fig. 1. Since it proved difficult to compare patterns using the strings of elements and the
normalizing factors that are in common use for depicting basalt compositions, a normalizing
factor more appropriate to granitic composition was sought. The factor chosen is a
hypothetical ocean ridge granite (ORG), calculated as the product of fractional crystal-
lization of average N-type MORB by application of the Raleigh fractionation law, assuming a
value of F (the proportion of residual liquid) of 0-25 (Table 3): the precise value is not, of
course, critical since the aim is to compare patterns rather than investigate absolute element
abundances. The elements used to make up the patterns were restricted to those that behave
incompatibly during fractionation of MORB to acid composition; thus Ti, P, Eu and Sr,
normally an integral part of basalt patterns, are not included. For simplicity Ce, Sm and Yb
are used to represent the rare earth elements; complete chondrite-normalized rare earth
patterns are depicted in many of the references in Table 1 and have not therefore been
reproduced here. The ordering of the elements is broadly according to their relative
incompatibility during MORB genesis (which increases from Yb to Rb), and K2O has been
added to the LHS of the pattern.
The normalizing composition thus represents the composition a granite would have had
were it: (i) ultimately derived from convecting upper mantle unaffected by any mantle
enrichment event; (ii) derived from a basalt parent by fractional crystallization of a simple
964 J. A. P E A R C E ET AL.

TABLE 3

MORB values and bulk distribution coefficients (D) used to calculate ORG (ocean ridge
granite) normalizing values

Element MORB D ORG

K20 o-io(%;1 0-01 0-4

Rb 10 001 4
Ba 12-5 0-01 50
Th 0-20 0-01 0-8
Ta 0-20 01 0-7
Nb 3-0 01 10
Ce 10-0 0-1 35
Hf 2-4 005 9

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on April 29, 2014

Zr 90 005 340
Sra 3-3 0-25 9
Y 30 0-4 70
Yb 3-4 0-4 80

MORB values are those of Pearce et al.

(1981) except that K 2 O, Rb and Ba are
reduced to those recommended by Melson
(1978) based on analysis of fresh glass, and
that Nb has been reduced from 3-5 to 3-0.
Values are in p.p.m. unless stated, and have
been rounded-off in view of uncertainties
inherent in the calculation. Values of D are
mosdy estimated from Arth (1976) and
Pearce & Norry (1979); F = 0-25; the
Rayleigh equation (CL = C o x /•">-») was
used in the calculation.

plagioclase—olivine-clinopyroxene-magnetite assemblage; (iii) unaffected by crustal melting

or assimilation or by volatile-dominated processes. Deviations from a flat pattern will then
reflect deviations from this simple genetic history and should thus vary systematically from
setting to setting.
The typical patterns for ocean ridge granites are shown in Fig. la. As might be expected
from the choice of normalizing factor, the patterns from 'normal' oceanic ridges (i.e. (a) and
(c)), whether subduction-induced or not, are flat and have normalized values close to unity for
most elements. They do, however, exhibit strong negative anomalies in K 2 O and Rb, which
may be attributed to the loss of these elements in a volatile phase or to alteration. It is also
possible that the values of these elements in MORB (which have proved difficult to obtain due
to alteration and low abundances) were overestimated. The pattern for the granite from the
anomalous ridge segment (MAR 45 ° N) deviates from this pattern by having relatively high
abundances of Th, Ta, Nb and Ce. These can.be explained in terms of selective enrichments
of the mantle source regions from which their parent basalts were derived (e.g. Wood, 1979).
The supra-subduction ridge granite from the Troodos Massif contains markedly lower
contents of incompatible elements of high ionic potential compared with the normalizing
composition; it also contains high values of Th and Ba and low values of K 2 O and Rb relative
to other elements. Its high ratio of Th to Ta is more characteristic of granites of the island arc
tholeiite series than of other types of ocean ridge granite, a predictable feature in view of the
boninite/island arc tholeiite character of its associated basaltic rocks (Pearce et al., in press).
Patterns typical of volcanic arc granites have been plotted in Fig. \b. Like volcanic arc
 TRACE ELEMENTS OF GRANITIC ROCKS 965

— Tincany (a) — Oman (a)

A
MAR 48° N (b) • A —— Newfoundland (a)
SmarMto (c) 10- v \ ~ — Jamaica (b)
\ \
Troodo* (d) ': » —-- Chlla (c)
i 1
•1
•. 1

'\
\
\
1.0-

(b)

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on April 29, 2014

V-—'
0.1-
R b B * T h T ( N b C « H ) Z r S n Y Yb K.O Rb Ba Th T i Nb C * HI Zr 8 m Y Yb

— Atcantlon I*, (c)

— Olio Rltt (a)

K.O R b B » T h T « N b C m Z r 8 m Y Yb K,0 Ftt) Bt Th T i Nb Ce HI Zr 8m Y Yb

TIb«t (a) Vedratta <S Rim (b)

Yunnan (a) -•••• OiMrigut (b)
- 8.W. Entfand (a) Novate (b)
Barouaas (a) Oman (d)

K.O H b B « T l l T « N b C « H » ZrSltlY Yb K.O R b B < T h T ( N b C « H f Z r 8 m Y Yb

Fio. 1. Ocean ridge granite (ORG) normalized geochemical patterns for the representative analyses, given in
Table 2, of ocean ridge granites (Fig. la), volcanic arc granites (Fig. 16), within plate granites (Fig. lc), within
plate granites (attenuated continental lithosphere) (Fig. Id), syn-collision granites (Fig. le) and post-collision
granites (Fig. If). For discussion, see text; for normalizing values, see Table 3.
966 J. A. PEARCE ET AC

basalts, the granites are characterized by enrichments in K, Rb, Ba, Th and (in calc-alkaline
and shoshonitic series) Ce and Sm relative to Ta, Nb, Hf, Zr, Y and Yb. A further significant
feature is the low value of Y and Yb relative to the normalizing composition. Of the four
patterns represented, the most distinctive is the granite from the late intrusive complex in
Oman, which was chosen to represent the most primitive possible plagiogranite of the island
arc tholeiitic series. This pattern exhibits low values of all elements and is the only pattern
from any setting in which the part of the pattern from Ta to Yb has a positive slope.
Nevertheless, the selective enrichment in Ba and Th relative to these elements is still evident.
The remaining patterns are broadly similar in shape, differing mainly in the absolute
abundances of the elements.
The patterns for within plate granites are shown in Figs. \c and Id. They can be divided
into three broad types. The first, exemplified by the Ascension Island and Oslo Rift patterns,

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on April 29, 2014

is marked by high and approximately equal normalized abundances of Rb, Th, Ta and Nb, a
feature which also characterizes most within plate basalts and which can be explained in
terms of derivation from incompatible-element enriched mantle (e.g. Pearce, 1982). Bearing in
mind the likely absence of major crustal involvement in the genesis of these granites
(Jacobsen & Wasserberg, 1978; Weiss, 1983), their patterns could be described as 'mantle
dominated'. They contrast markedly with the second type of pattern, exemplified by the
Sabaloka intrusion, in which Rb and Th are significantly enriched relative to Nb and Ta and
in which Ce and Sm are enriched relative to their adjacent elements. Such selective
enrichments can be attributed to crustal involvement, a conclusion consistent with isotope
evidence (Harris et ai, 1983), and this type of pattern could therefore be described as
'crust-dominated'. It will be apparent, however, that these two types of pattern do have
common features, namely a large negative Ba anomaly and, Ba apart, a general increase in
normalized abundance from Yb to Rb. The third type of pattern, exemplified by Mull and the
Skaergaard complex, exhibits the high ratios of Rb and Th relative to Ta and Nb that we
consider characteristic of 'crust-dominated' patterns (see also Thirlwall & Jones, 1983) but
differ from the other patterns in having no large negative Ba anomaly and in their flatter
(more MORB-like) trends from Ta to Yb. Despite their large intra-group variability, however,
the within plate granite patterns are distinctive in their combination of: (i) values of Hf to Yb
that are close to the normalizing value; and (ii) high values of K, Rb and Th.
The patterns for the collision granites are shown in Figs, le and / . It is evident that,
superficially at least, most of the patterns resemble those from volcanic arc granites of
calc-alkaline type. There are, however, some distinctive characteristics, notably the
exceptionally high contents of Rb in many of the patterns from the syn-collision granites and
the very low contents of Ce, Zr, Hf and Sm in the patterns from the Novate and Barousse
intrusions.
Finally, it should be noted that the granite patterns depicted in Fig. 1, do not entirely
represent the intrusive suites from which they are derived. Crystal accumulation, particularly
at more basic compositions than those shown, can sometimes change the pattern shape
significantly: for example, some ocean ridge granites can accumulate zircon giving patterns
with positive anomalies in Zr, Hf, Y and Yb. Fractional crystallization can, under certain
circumstances, have a major effect on the shapes of the patterns: for example, the negative Ba
anomalies in the within plate granites of Fig. \c are not apparent at intermediate
compositions. Crustal contamination can be variable, leading to all gradations between
mantle-dominated and crust-dominated patterns. Furthermore, the volatile phase may play a
significant role in distorting the patterns of porphyritic granites and microgranites. Despite
these caveats, however, the patterns shown can be considered as representative of the great
majority of analyses in the data bank.
 TRACE ELEMENTS OF GRANITIC ROCKS 967

TRACE ELEMENT-SiO 2 VARIATION DIAGRAMS

Further information on the applicability of selected trace elements to granite dis-
crimination, and particularly on the sensitivity of these elements to fractional crystallization,
is provided by the trace element-SiO2 variation diagrams in Fig. 2. From these, it can be seen
that:
(i) Y and Yb are generally more abundant in 'normal' ocean ridge and within plate granites
compared with volcanic arc granites for the complete range of SiO2 values (56-80 per
cent) under consideration (Fig. la, b). 'Supra-subduction' ocean ridge granites plot, as
expected, in the volcanic arc rather than 'normal' ocean ridge field.

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on April 29, 2014

1000:
WPO + OftO U - c )
on <P
WPQ + ORO U - c ) °
S "Dl °

100: a Drttj ° ° ' i s .? " • • •• Yb

ppm
Y
ppm !
:
3 ••• ' I '•*
1 0
•

. VAQ + COLO + OHO (d) ,

\ ' . %

VAO + COt.0 + ORO (d)

1 0-

M eo

1000-
•yn-COtO

100-

Rb
ppm
- " * '
* *
10- ' >. ' -

VAQ
• •

• • • • •

1 0- —i—?—i—i—i—i—i—i-

1000 • 0
WK torn u>) WPO +0Boa>)
i aD Bo B
10- •yn-COLO
a tP o
100 . d
°/° ° o o So c P Q Ta
Nb
Dpm
n
o (to Q X3m

10-
rft.
10
•?•; . * •V • h v *• i

10
VAQ + coto + ORO lied)
. . . . 0 1 •
VAO + ORO
• •

M SO 80 M
8IO2 wt% (0 wt%

FIG. 2. SiO2 variation diagrams for Rb, Y, Yb, Ta and Tb based on analyses in the data bank (see Table 1). Open
circles represent ORG, closed circles represent VAG, open squares represent WPG and closed squares represent
COLG. For clarity ORG (d), which have the same Y values as VAG, have not been plotted in Figs, la and 16;
?nd post-COLG, which are generally transitional between syn-COLG and VAG, have not been plotted in
Figs, la1 and 1 /
968 J. A. PEARCE ET AL.

(ii) Rb forms an almost perfect discriminant between ocean ridge and within plate granites
(Fig. 2c) and between volcanic arc and syn-collision granites (Fig. 2d). In the latter
diagram, however, post-orogenic granites overlap the fields of volcanic arc and
syn-orogenic granite.
(iii) Nb and Ta are generally more enriched in within plate granites than in other granite
types (Figs. 2e, f). The principal exceptions are granites intruded in within plate settings
in areas of attenuated continental lithosphere (see Table 1) which overlap the fields of
other granite types. This overlap is comparable to that described previously by Pearce &
Gale (1979) for Nb-SiO 2 covariations. It is also evident that the overlap is greater on the
Ta-SiO 2 than the Nb-SiO 2 diagram owing to the relatively high Ta/Nb ratios in
syn-collision granites (Harris et al., in press).
Of equivalent diagrams for the other elements studied (not reproduced here through lack of

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on April 29, 2014

space) it was observed that:
(i) Ba effectively discriminates between ocean ridge and within plate granites in the same
way as Rb (cf. Fig. 2c) but only for SiO2 values <65 per cent Above this value the
extreme sensitivity of Ba to both biotite (and K-feldspar) crystallization and crustal
contamination causes significant overlap between all magma types.
(ii) K2O discriminates between ocean ridge and within plate granites over the whole range of
SiO2 values (as did Rb in Fig. 2c) but is less effective than Rb in discriminating
between volcanic arc and syn-orogenic granites (cf. Fig. Id).
(iii) Th, Ce and Sm achieve the same type of discrimination as Rb, but with a much greater
degree of overlap between the granite types.
(iv) Zr and Hf provide a discrimination that is comparable to that achieved by Y and Yb for
SiO2 values less than about 68 per cent. At higher values of SiO2, however, Zr and Hf
can decrease in concentration as a consequence of zircon fractionation and (in some
within plate granites) crustal assimilation, and this causes a significant overlap between
the granite types. Syn-collision granites generally (though not always) have lower values
of Zr and Hf than other granite types and so Rb/Zr and Rb/Hf ratios are more effective
than Rb alone in separating syn-collision from volcanic arc granites (Harris et al., in
press).

DISCRIMINATION IN R b - Y - N b AND R b - Y b - T a SPACE

It is apparent from Figs. 1 and 2 that Rb, Y (or Yb) and Nb (or Ta) are likely to be the
most effective elements (of those studied) for the tectonic discrimination of granites. It is also
apparent, however, that 'supra-subduction zone' ocean ridge and post-collision granites
cannot be distinguished from volcanic arc granites on this basis. We therefore restrict our
analysis initially to the 'normal' ocean ridge (OR), within plate (WP), volcanic arc (VA) and
syn-collision (syn-COL) granite groups' and then consider the problems of classification of
post-orogenic granites and 'supra-subduction zone' ocean ridge granites in the subsequent
section. We have also elected to carry out the discrimination empirically rather than
statistically in order to produce diagrams that can also be understood in petrogenetic terms.
Fig. 3 shows the simple projections in Y—Nb (Fig. 3a) and Yb—Ta (Fig. 3b} space, which
represents the projection most likely to be alteration-independent. The significant feature of
both diagrams is the virtually complete separation of the VA and WP granites, a separation
that cannot be achieved quite so effectively by Y, Nb, Yb or Ta alone (cf. Fig. 2). The
discrimination between WP and OR granites is marred by an overlap zone between WP
granites from attenuated continental lithosphere and OR granites from anomalous ridge
segments. Of course, this overlap is not of great practical significance since these two granite
types have strongly contrasting geological characteristics. The discrimination between VA
 TRACE ELEMENTS OF GRANITIC ROCKS 969
100-j

1000-
°D

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on April 29, 2014

(a) (b)

FIG. 3. Nb-Y (Fig. 3a) and Ta-Yb (Fig. 36) discriminant diagrams for syn-collision granites (syn-COLG),
volcanic arc granites (VAG), within plate granites (WPG) and ocean ridge granites (ORG). The dashed line
represents the upper compositional boundary for ORG from anomalous ridge segments. Note that post-collision
granites can plot in all but the ORG fields, and that supra-subduction zone ocean ridge granites plot in the VAG
field: see Figs. 6 and 7 and accompanying text. Symbols as in Fig. 2. Co-ordinates of the discriminant boundaries
are: for the Y-Nb diagram: VAG/WPG = (1,2000) to (50,10); VAG/ORG = (40,1) to (50,10); ORG/WPG =
(50,10) to (1000,100) 0RG(6)/WPG = (25,25) to (1000,400); for the Yb-Ta diagram: syn COLG/WPG =
(55,20) to (3,2); syn-COLG/VAG = (0-1,0-35) to (3,2); VAG/WPG = (3,2) to (5,1); VAG/ORG = (5,0-05) to
(5,1); WPG/ORG = (5,1) to (100,7); WPG/ORG(ft) = (3,2) to (100,20).

and syn-COL granites is achieved almost completely on the Yb-Ta diagram but hardly at all
on the Y—Nb diagram.
The inclusion of Rb should, according to Fig. 2, separate the VA from syn-COL granites
and OR from WP granites. Furthermore, a diagonal axis in Nb—Y space can be seen to be
effective in separating VA from WP granites. Since the latter is equivalent to a discriminant
axis of equation Nb + Y = k, we can predict that a graph of Rb against (Y + Nb) should
carry most of the discriminating power of these three elements while retaining the advantage
(over statistically-derived discriminant functions) of easier petrogenetic interpretation. The
resulting diagram is shown in Fig. 4a and reveals virtually no overlap between the designated
magma types. The equivalent diagram, Rb against (Yb + Ta), is given in Fig. 4b and shows
comparable discrimination. In all diagrams, discriminant boundaries have been drawn by eye
and the details of these are given in the figure captions.

PETROGENETIC SIGNIFICANCE OF THE DISCRIMINANT BOUNDARIES

It can be demonstrated graphically that the boundaries drawn on the various
discrimination diagrams do have a petrogenetic basis for the most part. Fig. 5 illustrates the
case for the Rb-(Y + Nb) diagram. The parameters used in the modelling are listed in Table
4. These are based as far as possible on petrographic, bulk chemistry and isotopic constraints.
However, it should be emphasized that our objective is not to produce precise pathways but
to explain the rationale behind the discrimination. The effects of varying the different
parameters in Table 4 are thus discussed only when they significantly influence conclusions
related to the discrimination.
Fig. 5a illustrates the possible petrogenetic pathways for ocean ridge granites of normal
and anomalous type. In constructing these pathways, mantle heterogeneities are assumed to
970 J. A. P E A R C E ET AL.

1000 - •yn-COLO.
• /

:
m i
L_/"u ?.,"•• TO n /
/
100 - ° /
• * * n

Rb :
ppm;
: /
' /
10 - /
; ORQ
o
\
VAQ o

i O O CD O

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on April 29, 2014

•

1.0 1.0 -
10 100 10OO 1.0 10 100
(a) Y + Nb ppm (b) Yb + Ta ppm
FIG. 4. Rb-(Y + Nb) (Fig. 4a) and Rb-{Yb + Ta) (Fig. 46) discriminant diagrams for syn-collision (syn-COL),
volcanic arc (VA), within plate (WP) and normal and anomalous ocean ridge (OR) granites. Note that
post-collision granites can plot in all but the OR fields and that supra-subduction zone ocean ridge granites plot in
the VA field: see Figs. 6 and 7 and accompanying text. Symbols as in Fig. 2. Co-ordinates of the discriminant
boundaries are: for the Rb-(Y + Nb) diagram: syn COLG/VAG = (2,80) to (55,300); syn COLG/WPG =
(55,300) to (400,2000); VAG/WPG = (55,300) to (51-5,8); VAG/ORG = (51-5,8) to (50,1); ORG/WPG =
(51-5,8) to (2000,400): for the Rb-(Yb + Ta) diagram: syn COLG/VAG = (0-5,140) to (6,2000); syn
COLG/WPG = (6, 200) to (50,2000); VAG/WPG = (6,200) to (6,8); VAG/ORG = (6,8) to (6,1); ORG/WPG
= (6,8) to (200,400).

affect Rb and Nb equally but Y scarcely at all (see, for example, Wood, 1979; Sun & Nesbitt
1977; Pearce, 1982). The composition of the mantle source can then be represented by a
curved line (a-b) through a bulk mantle composition, BM. The latter is derived from the
primordial mantle values of Nb (0-6 p.p.m.) and Y (5 p.p.m.) from Wood's (1979)
compilation, and a value of Rb (0-3 p.p.m.) that is lower than the primordial mantle
composition by an amount (50 per cent) consistent with likely depletions due to crustal
growth and with Nb/Rb ratios in recent ocean island basalts. A MORB primary magma,
produced by 15 per cent partial melting of this source, leaving a residue of composition
ol67.5opx22.3cpx10, will then lie along the line c-d. Fractional crystallization of the primary
magma involving olivine, pyroxene, plagioclase ± magnetite and amphibole will then cause
the magma composition to move to higher Rb and (Nb + Y) values along the vectors
indicated.
The pathway for the Tuscany suite is based on a source composition (MT) close to the
N-type MORB source of Wood (1979) whereas the pathway for the 45° N MAR suite is
based on a source composition (ME) close to Wood's E-type MORB source. Partial melting
of these sources yields primary magmas at P T and P E and subsequent closed system
fractional crystallization gives basic magma compositions at BT and B E , intermediate
compositions at IT and IE and acid compositions at A T and AE. To explain the low Rb values
in the final products volatile degassing or sub-sea floor alteration may be invoked to take the
compositions to A^ and A E . It should be noted that, were open- rather than closed-system
fractional crystallization to have been considered, then B, I and A would probably be
displaced to the right of the values shown; a lower degree of partial melting or lower values
of F would also cause these points to be displaced to the right.
From these models, it can be seen that the left-hand boundary of the OR granite field can
 TRACE ELEMENTS OF GRANITIC ROCKS 971

1000 1000 -a Nigeria

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on April 29, 2014

10 100 100 1000

(a) Y + Nb ppm (b) Y + Nb ppm

1000 -a • yn-COLQ

10 1000

(C) Y + Nb ppm (d) Y + Nb ppm

Fie. 5. Petrogenetic modelling of the Rb-(Y + Nb) diagram to demonstrate the theoretical basis for the
discriminant boundaries. See Table 4 and text for details.

be defined petrogenetically as a line with the same shape as P T -P E and passing between B T
and IT and between BE and I E . This corresponds well with the line empirically drawn on Fig.
4. It can also be seen that the upper boundary should theoretically lie along B E -A E . The
empirically-drawn line, which passes through A E , thus represents a good estimate of the
upper limit of the ORG group, especially since loss of Rb can be expected to take most OR
granites well below this line.
Fig. 5b illustrates possible petrogenetic pathways for within plate granites. These are
 TABLE 4
Parameters used to compute the petrogenetic pathways in Fig. 5
B-l I-A Contaminant _
Source (pj)jn.) Partial melting P-B Fraclionallon Fractionalion Fraclionallon (p-p.m.) >
Rb Nb Y Residue F Assemblage F Assemblage F Assemblage F Rb Nb Y •v
m
(a) Tuscany 0-12 018 5 C
PO-I°PCIJ OI O-» 015 c p 0 j°lo aP'o < 0-5 Pin jjcpj-jmto , 0-2 P o .,cp 0 ,mt|,., 0-5 —
(b) MAR 45° N: 0-4 0-6 5 CPo-lOPoijO'o-l 015 CPj jOl^jplj , 0-5 Pin 6 c p 0 , m t , , 0-2 Pin,Cp O jmt o ., 0-5 —
(c) Ascension Is: 1 1-5 5 015 CPj.jOlj.jplj , 0-5 0-2 0-5 —
o
Pi »^-Po-im^o i Pin i T o J mt o-i m
C
ded from http://petrology.oxfordjournals.org/ at Aston University on April 29, 2014

(d) Skye: 0-4 0-6 5 Pn°Pni0lH 015 •cpj.jOlo^plj., 0-5 Pin t c p 0 jitit^., 0-2 P o • c Po-j m t o i 0-5 •200 8 20
(c) Nigenn: 5 7-5 5 CPj. | opj.jjOl 0 ., 015 Cp0 jOlj.jPlj , 0-5 *P'o , , c p j j a m , ,mtj 0J 0-2 P O-35amO jbinlllt,, 0 j 0-5 •200 8 20
0-36 018 5 OPj jjOl^, 0-3 c 0-5 0-2 p o .jjam 0 .jCp 0 .jmt o 0 3 0-5 —
(0 L. Port Complex: Po j°lo jPlo< Pin jjarn^ ,cp 0 jmtj 0 J
(R) Alaska: 0-96 018 5 °Pojj°le-7 0-3 cPa-filz )P'« 4 0-5 Pin j ) am o . 4 mt o . O j 0-2 p o.jSam0 jbi^jmt, 0 3 0-5 —
(h> Chile: 3-2 0-6 5 OPj j , O l r , 0-3 TO-JOIOJP'H 0-5 P>n jjam^mtoj, 0-2 P o-js am n ibio-jmtu-oj 0-5 —

cp = clinopyroxene: op = orthopyroxenc: ol = olivine; pi = plagioclase; am = amphibole; mt = magnetite. Asterisk refers to cases where combined assimilation-fractionation has been
invoked.
 TRACE ELEMENTS OF GRANITIC ROCKS 973

assumed to be derived from mantle sources that are enriched in incompatible elements relative
to the bulk mantle composition and also to be derived from primary magmas that lie along
the 15 per cent partial melting line. In the model for the Ascension Island suite (which is
assumed to contain no significant crustal component), fractionation trends similar to those
followed by OR granites could relate the granite compositions, at A A , back to a primary
magma P A and in turn back to a mantle source, M A . In the model for the Skye suite a more
depleted source, M,, has been invoked from isotopic evidence (Thirlwall & Jones, 1983). In
addition, crustal assimilation by ultrabasic-basic magma at the base of the crust (Thompson
et al., 1982) has been modelled by combined assimilation—fractionation between P8 and B,.
Closed system fractional crystallization has then been invoked to take the composition from
B, to I, and from I, to A,. It will be noted that, unless the mantle source is of unusual
composition, crustal assimilation is necessary to explain how the compositions of the Skye

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on April 29, 2014

granites lie above the line c-d. Finally, the petrogenetic pathway for the Nigerian granite suite
is based on an enriched mantle source with combined assimilation-fractionation at
intermediate and acid compositions, again in accordance with isotopic constraints.
From these pathways, it can be seen that the lower discriminant boundary of the WP
granite group should theoretically lie along B A -A A , being the trend for an oceanic within plate
granite derived from a relatively unenriched mantle source. The empirically-drawn line lies
very close to this boundary. The left-hand boundary will theoretically pass between B and I
for mantle-dominated suites but could pass through A if extensive crustal assimilation were
to take place. The empirically-drawn line, which is an extrapolation of the left-hand ORG
boundary, is thus consistent with theoretical considerations. The upper and right-hand
boundaries of the within plate granite field need not, of course, be constrained.
Fig. 5c illustrates petrogenetic pathways for volcanic arc granites of oceanic/tholeiitic,
oceanic/calc alkaline and continental/calc alkaline types, the first exemplified by the Little
Port Complex of Newfoundland (Malpas, 1979), the second by the granites intruding the
accretionary prism of Alaska (Perfit et al., 1980; Hill et al., 1981), and the third by granites
from Central Chile (Baldwin & Pearce, 1982). In each case it is necessary additionally to
consider a vertical vector which represents the selective introduction of Rb from the
subduction zone into the mantle source region. The length of the vector and the composition
of the mantle prior to addition of this subduction component can be estimated from the
geochemical patterns of the basic rocks from similar settings (Pearce, 1983): the values used
are listed in Table 4. Thus it can be seen that the Newfoundland mantle source (EN) is
obtained by a three-fold enrichment of Rb from a N-type MORB source (M N ); the Alaska
mantle source (EA) is obtained by an eight-fold enrichment of Rb from a N-type MORB
source (MA); and the Chile mantle source (E c ) is obtained by an eight-fold enrichment of Rb
from a relatively enriched (bulk mantle) source, M c . The remainder of the pathways have
been modelled as before, but taking into account petrological evidence for more partial
melting, earlier crystallization of magnetite and a greater role of amphibole fractionation than
in the genesis of the ocean ridge and within plate granites. The most significant features in all
three pathways are the fractionation vectors for intermediate and acid compositions which,
due to the importance of amphibole as a crystallizing phase, are almost vertical. It is these
vectors that prevent a significant degree of overlap between the VA and WP granites since
this vector also lies sub-parallel to the discriminant boundary. It is also significant that the net
effect of a high degree of partial melting and early magnetite crystallization is to reduce the
lengths of the vectors from E to P and B to I respectively, and this may be important in
defining the right-hand boundary when amphibole is of lesser importance than has been
assumed in these models. The upper boundary is determined by the maximum Rb
concentration that can be achieved from a combination of source enrichments, partial melting
974 J. A. PEARCE ET AL.

and fractional crystallization. It could theoretically lie above that drawn, but only for
magmas derived from unusually enriched mantle sources, for those enriched by influx of
Rb-rich fluids during a late stage in their crystallization history or for those that have suffered
a high degree of crustal contamination.
Pathways for syn-tectonic granites cannot be rigorously modelled since the source is
known to be the continental crust, which has a less predictable composition than the mantle
and which contains residual phases for which distribution coefficients are not available.
Moreover, element transfer in the fluid phase is known to be significant in magma genesis (e.g.
Le Fort, 1981) and this too cannot be modelled with any degree of confidence. The pathways
drawn in Fig. Sd for the Barousse granite and the Yunnan granite are thus simply guides to
the type of variation that can be produced in this setting. In both cases, magma genesis is
assumed to involve volatile-induced melting of the hot bases of thrust sheets and to involve

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on April 29, 2014

two variables: the partial melting trends for the crustal source; and the addition of Rb-rich
(Y + Nb)-poor fluids. The sources taken were a quartz diorite, Q, and a greywacke, G,
respectively, since empirical data are available on their melting behaviour in the upper crust
(Tindle & Pearce, 1983): values of KD are above unity for Rb in both cases due to residual
biotite and KD is about unity for (Y + Nb) in the greywacke but above unity for (Y + Nb) in
the quartz diorite due to residual minor phases and amphibole. Approximate melting trends,
annotated according to degree of melting, are also shown, and demonstrate that simple partial
melting of these sources cannot yield high-Rb magmas of syn-collision type unless the volatile
component is added. It can thus be inferred that the lower boundary of the syn-collision
granite group is a purely arbitrary one and that granites could plot below this line if they were
derived from Rb-poor crust or if the fluid component were small. The right-hand boundary is
constrained almost entirely by the composition of the crustal source and not by the melting
behaviour of the source (since Y and Nb are compatible elements during crustal melting) or
by the fluid phase. Since most crustal rocks lie within the volcanic arc field on this diagram, it
is valid to have derived the boundary between the syn-COL and WP granites by near-linear
extrapolation of the VAG/WPG boundary.
Our modelling of the other diagrams is not presented here owing to lack of space, but it will
be noted that similar pathways can be drawn on the Rb—(Ta + Yb) diagram. On the Nb-Y
and Ta-Yb diagrams, the degree of depletion of the mantle source can be shown to control the
ORG-WPG discriminant boundaries and the combination of higher degrees of melting, early
crystallization of magnetite and amphibole fractionation in VA granites can be shown to
prevent overlap between these and the OR and WP granites. The discrimination of syn-COL
and VA granites on the Yb—Ta diagram can probably be attributed to the introduction of Ta
in the fluid phase, a process that does appear to increase Ta/Nb ratios (Harris et al., in press).

DISCRIMINATION OF OTHER GRANITE TYPES

As we have already discussed, two types of setting present problems in the construction
and interpretation of granite discriminant diagrams: post-collision granites, and plagio-
granites from SSZ ocean ridges. We now consider each of these in turn.

Post-collision granites
This group represents a major problem in all tectonic-geochemical classifications of
granites. The probable reason is that, unlike granites from other settings, post-collision
granites cannot be explained in terms of a single, well-defined mantle or crustal source. It has
been demonstrated that they can result both from melting of the lower crust due to thermal
relaxation following collision and from melting of the upper mantle (which may be of 'within
 TRACE ELEMENTS OF GRANITIC ROCKS 975

plate' or 'arc' composition) due to the adiabatic decompression that accompanies

post-collision uplift and erosion (England & Thompson, in press; Harris et al., in press). In
addition, renewed subduction after collision could generate 'normal' volcanic arc magmas in a
post-collision setting (Colman-Sadd, 1982). The affinity of the resulting granite may therefore
depend on the geometry of the collision event which in turn controls the nature of the crust
and mantle that undergo melting.
The Rb and (Y + Nb) ranges for some typical post-collision granites have been plotted in
Fig. 6. They most commonly lie near the top of the VA granite field, as exemplified by the
zoned Adamello pluton (Dupuy et al, 1982) and the granites of Oman and Masirah Island.
However, they may also span compositional fields, plotting in both the VAG and syn-COLG
fields, as in the case of the Querigut pluton (Fourcade & Allegre, 1981) or in the WPG and
syn-COLG fields as in the case of the Hercynian Grandes-Rousses granite (Oliver et al.,

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on April 29, 2014

1983). In all these granites, variable mixtures of mantle- and crust-derived magmas are
required to explain the observed variations.
Although geochemical discrimination of this group could not be achieved in the projections
presented in this paper, there are projections that permit partial separation. The Ta-Hf-Rb
diagram of Harris et al. (op cit.) was the most effective simple discriminant we could devise,
but is based on a limited sample set and more data are required before it can be fully evaluated.
Thus, at present, geological criteria and the zonation of intrusions are the best indicators of a
post-orogenic setting. However, if the setting can be defined as 'post-collision' on a priori
grounds, then the diagrams published here can be used in evaluating the nature and polarity
of the collision event.

syn-COLG
1000-
/
QrandesRousses "
•
jy WPG .

100- „ Querlflut C_-j^O /

Rb ; AdamelloV /
ppm; /
Oman
10-
' ORG
VAG

1.0-
10 100 1000
Y + Nb ppm
FIG. 6. Distribution of some post-collision granites on the Rb-(Y + Nb) discriminant diagram of Fig. 4.
Data sources are given in Table 1.

SSZ ocean ridge plagiogranites

These granites only present a problem in that they are chemically indistinguishable in most
cases from granites of the island arc tholeiite series—a predictable observation since their
mantle source regions may be identical. Like post-orogenic granites, therefore, the group
cannot be uniquely characterized on the basis of their geochemistry. However, if a priori
976 J. A. PEARCE ET AL.

evidence exists for an ocean ridge setting, the diagrams presented can be used 'in reverse' to
evaluate the type of ocean ridge at which they formed. This discrimination of ocean-ridge
granites is illustrated in Fig. 7 for the Y-Nb projection.

-
\ WPG /
100- \ s
X
Nb : VAG
+
\
N.
X^
MAR 48° N (b)X_
/
rslca <a)
ppm; syn-COLG

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on April 29, 2014

\. ^^y-jr^Sarmlen o (c)
10-
z Antalya (d)—/
r s<V /J/^Tuscany( a)
AM
/^\\ r V 1
-/ Smartvllle (c)
Vourlnos (d)-f 7 ^ L>77
Jf / / - Oman (d)
v
TroodO8 (d)—"^
1 ORG
1.0- 1 i i i i ni| ' i HI

10 100 1000
Y ppm
FIG. 7. Distribution of data from the four sub-groups of ocean ridge granites on the Y-Nb discriminant diagram
of Fig. 3. Data sources are given in Table 1.

PROBLEMS OF APPLICATION
The techniques presented in this paper are subject to the usual set of precautions in their
application to the study of granites of unknown affinities. These are listed below:

Sampling and analytical procedures

The diagrams presented are ideally designed to classify fresh, non-porphyry, non-cumulate,
non-aplitic intrusive rocks containing visible free quartz and collection of such rocks should
be the aim of any sampling procedure. The number of samples required will clearly depend on
the variability of the granite suite under investigation. The normal accuracy achieved by XRF
or INAA techniques is assumed but it will be noted that, when the abundances of the key
elements are close to detection limits, the discrimination does not in any case depend on
having accurate values of the elements concerned.

Alteration
The effects on the discriminant diagrams of hydrothermal and supergene alteration have
been assessed for two of the granites in the data bank: the syn-COL granites of SW England
(Alderton et al, 1980) and the VA granites from the El Salvador porphyry copper intrusive
centre in Chile (Baldwin & Pearce, 1982). The chemical changes that accompany the
different alteration types, as determined by analysis of fresh-altered granite pairs, are
illustrated for the Rb-(Yb + Ta) diagram in Fig. 8a.
It is apparent that Rb-enrichment is a likely consequence of K-silicate and sericitic
alteration (due to the growth of secondary biotite and muscovite respectively) and
Rb-depletion a likely consequence of chloritization and argillic alteration (due to breakdown
of feldspar and mica). By comparison, (Y + Nb) shows only small variations: both Y and Nb
are reduced during K-silicate alteration owing to 'dilution' by added quartz and Y is gained
 TRACE ELEMENTS OF GRANITIC ROCKS 977

1000- 1000-

100

Rb
ppm KEY

10-
•it«f«Uon (8.W. i
lltwiUon (B S«lv>dor )
UXttzitton (S*ud AribW

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on April 29, 2014

1.0- 1.0
100 1000

(a) Yb + Ta ppm (b) Y + Nb ppm

FIG. 8. Illustration of potential problems in the classification of granites. Fig. Sa shows the possible effects of
volatile induced magma chamber zonation and alteration on Rb and (Yb + Ta) compositions. In the former, the
tail of the arrow gives the composition of the early and the arrow-head the composition of the late eruptive
sequence for the Bishop Tuff (BT), Lava Creek Tuff (LCT), Valley-of-Ten-Thousand-Smokes (VTTS), Devine
Canyon Tuff (DCT) and Mazama Climax (MC) as taken from Hildreth (1981). Alteration types are K-silicate
(K-sil), sericitic (Ser), argillic (Arg), chloritization (Chi), tourmalinization (Tour) and albitization (Alb). Fig. 86
shows cumulation trends for the Rb-(Y + Nb) and Nb-Y diagrams based on distribution coefficients (Arth, 1976;
Pearce & Norry, 1979; Condie, 1978) for intermediate-acid compositions. Trends show effects of addition of 50
per cent of the cumulate mineral (except for minor phases where direction only is given).

during chloritization since it can be accommodated in the chlorite structure. An unusual

exception is tourmalinization which can cause depletion of most elements due to the
breakdown of both major and minor phases, including even zircon. However, all the vectors
drawn are based on the complete destruction of the primary minerals and the effects of less
intense alteration will be correspondingly smaller. Alteration by alkali-rich fluids of the type
seen in WP alkali complexes (Harris & Marriner, 1980; Harris, 1981) may increase Rb, Y
and Nb, as also indicated in Fig. 8a: although the precise vector cannot be drawn from the
data available, it is clear that this type of alteration would not prevent the correct
identification of WP granites.

Crystal accumulation
This study has been restricted to samples that do not show obvious cumulate layering.
However, "crystal accumulation can occur by processes such as filter pressing which may
leave no textural evidence, and so has to be considered as a possible cause of
misclassifications. Approximate cumulation trends have been drawn on the Rb-(Y + Nb)
and Y-Nb diagrams in Fig. 8ft for a variety of minerals for magma of intermediate-acid
composition. It is apparent, therefore, that, on these diagrams:
(i) WP and OR granites whose values of Y, Nb and Rb are 'diluted' by accumulation of
plagioclase feldspar may plot in the VAG field;
(ii) VA and syn-COL granites that have accumulated ferromagnesian and minor phases may
plot in the WPG or ORG field.
978 J. A. PEARCE ET AL.

Crustal contamination
The effects of crustal contamination have already been treated as an integral part of the
relationship between granite chemistry and tectonic setting. However, unusual degrees of
contamination are possible, usually at the margins of intrusions and in areas unusually rich in
enclaves. In these cases, the composition of the contaminant is likely to lie in the VAG or
syn-COLG fields, as shown in Fig. 5d. Extensive contamination could therefore cause a
WP granite to be misclassified as VAG or syn-COLG or, rarely, a VA granite to be
misclassiiied as syn-COLG.

Volatile fluxing
This process has also been treated as an integral part of the discrimination. However, in
extreme cases, misleading results can occur. Aplites are the most obvious examples, often

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on April 29, 2014

showing a strong depletion in many trace elements due to the loss of volatile-rich aqueous
fluids from their parent magmas (conversely, many pegmatites are likely to be enriched in
these elements). However some microgranites, and porphyritic granites containing aplitic
groundmasses, may also have suffered trace element depletion (Hudson & Arth, 1983;
Baldwin & Pearce, 1982). It is thus recommended that medium- to coarse-grained
equigranular rocks be used where possible in the classification of granite suites.
Further evidence for volatile-induced geochemical variations has been compiled by
Hildreth (1981) from a study of acid tuff sequences in the western United States. Hildreth's
data have been plotted on the Rb-(Yb + Ta) diagram in Fig. 8a, and show a marked
depletion in Rb, and a smaller depletion in Yb and Ta, in late compared to early eruptions. If
Hildreth is correct in interpreting these variations as due to thermogravitational diffusion of
volatile complexes rather than to crystal-liquid fractionation (cf. Michael, 1983), it must be
concluded that 'anomalous' chemical compositions can be produced by transport of elements
as volatile complexes even in water-undersaturated magmas. Thus care may be needed in the
interpretation of volatile-rich granites, especially if their compositions lie close to a
discriminant boundary.

Age
Since the diagrams are based on analyses of Phanerozoic rocks they cannot automatically
be applied to rocks of Precambrian age. In detail, the main differences between magma
genesis in the Archaean and at the present day are likely to be:
(i) The mantle evolution line in Fig. 5a will be displaced to higher Rb values, and very
slightly higher (Y + Nb) values,
(ii) The degree of partial melting may be greater,
(iii) The mantle may be less fractionated than at the present-day so that generation of
magmas from very enriched and very depleted sources would be less common,
(iv) Crustal melting may be more important in magma genesis in all environments.
The net effect on the diagrams should be to blur the syn-COLG/VAG discrimination and
displace the (VAG + syn-COLG)/(WPG + ORG) boundary to lower values of Y, Yb,
(Y + Nb) and (Yb + Ta). However, the precise effect cannot be known until further studies
have been made on granites of varying ages.

Complex tectonic regimes

As we emphasized in the previous section, the fields on the discriminant diagrams strictly
reflect source regions (and melting and crystallization histories) rather than tectonic regimes.
For syn-COL, VA, 'normal' OR and WP granites source regions and tectonic regimes show a
clear correlation. However, post-COL granites can plot in the syn-COLG, VAG or WPG
 TRACE ELEMENTS OF GRANITIC ROCKS 979

fields and OR granites can plot in the VAG or ORG fields according to their precise setting.
Thus, the discrimination is not perfect between all groups and interpretations must be made
with this caveat in mind.
A further cause for caution concerns the complex nature, not only of collision settings, but
of many apparently simple tectonic regimes. We have already pointed out, for example, that
many ophiolite complexes not only contain granites that formed at ocean ridges but also
granites associated with subduction and collision events; and that many volcanic arcs not
only contain granites that resulted from the subduction of oceanic lithosphere but also
granites associated with back-arc and inter-arc collision and rifting events. We would
therefore expect most granite provinces in the geologic record to contain intrusions which,
despite their almost similar ages and locations, plot in different fields on the discriminant
diagrams.

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on April 29, 2014

SUMMARY
(1) Granites may be classified according to the tectonic setting into which they are
intruded (ocean ridge, volcanic arc, within plate or collision settings). Ocean ridge granites
can be subdivided into subduction-related and subduction-unrelated on tectonic criteria and
into normal, anomalous and SSZ on the basis of the chemistry of their associated basalts.
Volcanic arc granites can be subdivided into intraoceanic and intracontinental on tectonic
criteria and tholeiitic, calc-alkaline and shoshonitic on geochemical criteria. Within plate
granites can be subdivided into intraoceanic, intracontinental and attenuated continental
lithosphere on tectonic criteria. Collision granites can be subdivided into continent-continent,
arc-continent and arc-arc according to the type of collision and into syn-collision and
post-collision according to the time of emplacement.
(2) Geochemical patterns (normalized to a hypothetical ocean ridge granite) reveal
systematic trace element differences between many of the tectonic groups and sub-groups.
Particularly significant are: high Y and heavy REE concentrations in ocean ridge and within
plate granites; high Nb and Ta in many within plate granites; and high Rb in syn-collision
granites. Trace element—SiO2 variation diagrams confirm that these features hold over the
SiO2 range studied (56-80 per cent).
(3) Discrimination diagrams in Rb-Nb-Y and Rb-Ta-Yb space effect almost complete
separation of syn-collision volcanic arc, within plate and (non-SSZ) ocean ridge granites.
Petrogenetic modelling demonstrates that the observed discrimination can be explained by:
the derivation of ocean-ridge granites from generally-depleted mantle sources coupled with
loss of Rb during magma evolution; the derivation of within plate granites from enriched
mantle sources; amphibole fractionation during the genesis of volcanic arc granites; and the
volatile-induced enrichment in Rb (and Ta) during the genesis of syn-collision granites.
(4) Post-collision granites can plot in the volcanic arc, within plate or syn-collision fields
depending on the relative proportions of mantle- and crust-derived magmas and on the
enrichment history of the mantle concerned. Although difficult to distinguish on geochemical
grounds, if a priori evidence is available for their origin, the diagrams can be used to provide
further information on the collision event. Similarly, plagiogranites from supra-subduction
ocean ridges cannot simply be distinguished geochemically from those from oceanic, tholeiitic
arcs. However, if an ocean-ridge origin can be deduced on a priori evidence, the diagrams can
be used to identify the setting of the ridge. For these reasons, the diagrams presented in this
paper should be regarded as guides to the nature of the source regions of granite magmas
which, in conjunction with geological constraints, can add information on the tectonic setting
of intrusion of Phanerozoic, and probably Proterozoic, non-cumulate, non-aplitic granitic
intrusions.
 980 J. A. PEARCE ET AL.

ACKNOWLEDGEMENTS
We would like to express our thanks to the following for donation of samples or
unpublished data, without which this study could not have been carried out: S. Agrell
(Ascension Isles), T. Alabaster (Oman), D. Aldiss (Troodos and Smartville), J. Baldwin
(Chile), A. Bussell (Peru), P. Curtis (Sudan), M. Fisk (Reunion), N. Grant (Bolivia), M.
Isaacs (Jamaica), W. Jun-Wen and C. Zhong-Li (Tibet), E.-R. Neumann (Oslo Graben), R.
Oliver (Alps), G. Rogers (Chile), G. Serri (Tuscany), R. Thorpe (Mull), B. G. J. Upton
(Greenland), E. A. Vincent (Skaergaard). We are also grateful to J. S. Watson, P. J. Potts and
O. W. Thorpe for analytical assistance, to M. Leggett for typing the manuscript and to A.
Autran, A. Cocherie, J. G. Fitton, P. Le Fort, L. Le Bel, R. Macdonald, M. Piboule and those
mentioned above for helpful discussions. J. A. Pearce would additionally like to thank the
Royal Society of London and Academia Sinica for financing a visit to the Guiyang Institute

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on April 29, 2014

of Geochemistry where this study was started, the Institut Dolomieu in Grenoble for an
associate professorship during which time the study was completed, and Prof. Tu, Dr. Cao,
M. Piboule, R. Oliver and others for their help during these periods.

REFERENCES
Abbots, I. L., 1978. High potassium granites in the Masirah ophiolites of Oman. Geol. Mag. 115,415-25.
Alabaster, T., Pearce, J. A., & Malpas, J., 1982. The volcanic stratigraphy and petrogenesis of the Oman ophiolite
complex. Contr. Miner. Petrol. 81, 168-83.
Albuquerque, C. A. R., 1971. Petrochemistry of a series of granitic rocks from Northern Portugal. Bull. geol. Soc.
Am. 82, 2783-98.
1978. Rare elements in 'Younger' granites, Northern Portugal. Lithos, 11, 215-29.
Alderton, D. H. A., Pearce, J. A., & Potts, J. A., 1980. Rare earth element mobility during granite alteration:
evidence from south-west England. Earth planet. Sci. Lett. 40, 149-65.
Aldiss, D. T., 1978. Granitic rocks of ophiolites. Ph.D. thesis, Open University.
1981. Plagiogranites from the ocean crust and ophiolites. Nature, 289, 577-8.
Arth, J. G., 1976. Behaviour of trace elements during magmatic processes—a summary of theoretical models and
their applications. J. Res. US Geol. Surv. 4,41-7.
Aumento, F., 1969. Diorites from the Mid Atlantic Ridge at 44° N. Science, 165, 1112-3.
Baldwin, A. J., & Pearce, J. A., 1982. Discrimination of productive and non-productive porphyritic intrusions in
the Chilean Andes. Econ. Geol. 77, 664-74.
Bateman, P. C , & Chappell, B. W., 1979. Crystallization, fractionation and solidification of the Tuolumne
intrusive series, Yosemite National Park, California. Bull. geol. Soc. Am. 90,465-82.
Bebien, J., 1982. L'association ignee de Guevgueli (Macedoine Grecque). PhD. thesis, Nancy University.
Beccaluva, L , Ohnenstetter, D., & Venturelli, G., 1977. The trace element geochemistry of Corsican ophiolites.
Contr. Miner. Petrol. 64, 177-89.
Beckinsale, R. D., 1979. Granite magmatism in the tin belt of south-east Asia. In: Atherton, M. P., & Tarney, J.
(eds), Origin of Granite Batholiths: Geochemical Evidence. Orpington, Kent: Shiva Publishing Ltd., 34-44.
Bellieni, G., Peccerillo, A., & Poli, G., 1981. The Vendrette di Ries (Rieserferner) plutonic complex: petrological
and geochemical data bearing on its genesis. Contr. Miner. Petrol. 78, 145-56.
Bourgois, J., & Janjou, D., 1981. Subduction oceanique, subduction continentale et surrection andine: Pexemple de
Perou septentrional. C.R. Acad. Sci. Paris, 293, 859-64.
Bowden, P., & Turner, D. C , 1974. Peralkaline and associated ring-complexes in the Nigeria-Niger Province,
West Africa. In: Sorensen, H. (ed.), The Alkaline Rocks. New York: J. Wiley & Sons, 330-51.
& Whitley, J. D., 1974. Rare earth patterns in peralkaline and associated granites. Lithos. 7, 15-21.
Cann, J. R., 1970. Rb, Sr, Y, Zr and Nb in some ocean floor basaltic rocks. Earth planet. Sci. Lett. 10, 7-11.
Capdevila, R., 1969. Le metamorphisme regional progressif et les granites dans le segment hercynien de Galice
nord-orientale. Ph.D. thesis, Montpellier University.
& Vialette, Y., 1970. Estimation radiometrique de Page de la deuxieme phase tectonique hercynienne en
Galice moyenne (Nord-ouest de l'Espagne). C.R. Acad. Sci. Paris, 270, 2527-30.
Chappell, B. W., & White, A. J. R., 1974. Two contrasting granite types. Pacific Geol. 8, 173-74.
Cocherie, A., 1978. Geochimie des terres rares dans les granitoides. Ph.D. thesis. Rennes University.
Coleman, R. G., & Peterman, Z. E., 1975. Oceanic plagiogranite, J. geophys. Res. 80, 1099-1108.
& Donato, M. M., 1979. Oceanic plagiogranite revisited. In: Barker, F. (ed.), Trondhjemites and Related
Rocks. Amsterdam: Elsevier Scientific Publishing Company, 149-68.
Collins, W. J., Beams, S. D., White, A. J. R., & Chappell, B. W., 1982. Nature and origin of A-type granites with
particular reference to Southeastern Australia, Contr. Miner. Petrol. 10, 189-200.
Colman-Sadd, S. P., 1982. Two stage continental collision and plate driving forces. Tectonophysics, 90, 253-82.
 T R A C E E L E M E N T S OF G R A N I T I C R O C K S 981

Condie, K. C., 1978. Geochemistry of Proterozoic granite plutons from New Mexico, U.S.A. Chem. Geol. 21,
131-49.
Curtis, P. A. S., 1979. The geology of the Sultaniyat Ring Complex, Bayuda Desert, Sudan. Ph.D. thesis,
Portsmouth Polytechnic.
Debon, F., & Le Fort, P., 1982. A chemical-mineralogical classification of common plutonic rocks and
associations. Trans. R. Soc. Edinb. Earth Sciences, 73, 135-49.
Dewey, J. F., 1982. Plate tectonics and the evolution of the British Isles. J. geol. Soc. Lond. 139, 371-412.
Dietrich, V., & Gansser, A., 1981. The leucogranites of the Bhutan Himalayas. Schweiz miner, petrogr. Mitt. 61,
177-202.
Dupuy, C , Dorstall, J., & Fratte, M., 1982. Geochemistry of the Adamello Massif (Northern Italy). Contr.
Miner. Petrol. 80, 41-8.
Engel, C. G., & Fisher, R. L., 1975. Granitic to ultramafic complexes of the Indian Ocean ridge system, Western
Indian Ocean. Bull. geol. Soc. Am. 86, 1553-78.
England, P., & Thompson, A. B., 1984. Pressure-temperature-time paths of regional metamorphism. Part I. Heat
transfer during the evolution of regions of thickened continental crust. J. Petrology, 25, 894-928.
Fisk, M. R., Upton, G. C. J., & White, W. M., in press. Chemistry of Reunion Island volcanic rocks and minerals:

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on April 29, 2014

I. The nature of the mantle and fractionation processes.
Floyd, P. A., & Winchester, J. A., 1975. Magma type and tectonic setting discrimination using immobile elements.
Earth planet. Set. Lett. 27, 211-18.
Fourcade, S., & AUegre, C. J., 1981. Trace element behaviour in granite genesis: a case study. The calc-alkaline
plutonic association from the Querigut Complex (Pyrenees, France). Contr. Miner. Petrol. 76, 177-95.
Gealey, W. K., 1980. Ophiolite obduction mechanism. In: Panayiotou, A. (ed.), Ophiolites. GeoL Survey DepL,
Cyprus, 228-43.
Gerlach, D. C , Ave Lallemant, H. G., & Leeman, W. P., 1981. An island arc origin for the Canyon Mountain
ophiolite complex, Eastern Oregon, U.S.A. Earth planet. Sci. Lett. 53, 255-65.
Hanson, G. N., 1978. The application of trace elements to the petrogenesis of igneous rocks of granitic
composition. Ibid. 26—43.
Harris, N. B. W., 1981. The role of fluorine and chlorine in the petrogenesis of a peralkaline complex from Saudi
Arabia. Chem. Geol. 31, 303-10.
& Marriner, G. F., 1980. Geochemistry and petrogenesis of a peralkaline granite complex from the Midian
Mountains, Saudi Arabia. Lithos, 13, 325-37.
Duyverman, H. J., & Almond, D. C , 1983. The trace element and isotope geochemistry of the Sabaloka
igneous complex, Sudan. / . geol. Soc. Lond. 140, 245-56.
Pearce, i. A., & Tindle; A. G., in press. Geochemical characteristics of collision zone magmatism. In:
Shackleton, R. M., Ries, A. C , & Coward, M. P. (eds.), Collision Tectonics. Geol. Soc. Lond. Spec. Publ.
Hildreth, W., 1981. Gradients in silicic magma chambers: implications for lithospheric magmatism. J. geophys.
Res. 86, 10153-92.
Hill, M., Morris, J., & Whelan, J., 1981. Hybrid granodiorites intruding the accretionary prism, Kodiak, Shumagin
and Sanak Islands, S.W. Alaska. Ibid. 86, 10569-90.
Honeggar, K., Dietrich, V., Frank, W., Gansser, A., Thoni, M., & Trommsdorf, V., 1982. Magmatism and
metamorphism in the Ladakh Himalayas (the Indus—Tsangpo suture zone). Earth planet. Sci. Lett. 60,
253-93.
Hudson, T., & Arth, J. G., 1983. Tin granites of the Seward Peninsula, Alaska. Bull. geol. Soc. Am. 94, 768-90.
Imeokparia, E. G., 1982. Geochemistry and relationship to mineralisation of granitic rocks from the Afu Younger
Granite Complex, Central Nigeria. Geol. Mag. 119, 39-56.
1983. Geochemical aspects of the evolution and mineralization of the Amo Younger granite complex
(Northern Nigeria). Chem. Geol. 40, 293-312.
Isaacs, M. C. I., 1975. Comparative geochemistry of selected Jamaican intrusive rocks. M.Sc. thesis, Leeds
University.
Jacobsen, S. B., & Wasserburg, G. J., 1978. Nd and Sr isotopic study of the Permian Oslo rift. 4th Int. Conf.
Geochron. Cosmochron. Isotopic Geol. U.S. Geol. Surv. Open-File Rept. 78-701, 194-6.
Khali], S. E., Neef, J. A., & Brunfelt, A. O., 1978. Trace element abundances of the Holter Kollen pluton complex,
Oslo area, Norway. Chem. Geol. 22, 121-55.
La Roche, H., 1978. La chimie des roches presentee et interpretee d'apres la structure de leur fades mineral dans
Pespace des variables chimiques: fonctions specifiques et diagrammes qui s'en deduisenL Application aux
roches ignees. Chem. Geol. 21, 63-87.
Le Fort, P., 1981. Manaslu leucogranite: a collision signature of the Himalaya: a model for its genesis and
emplacement. J. geophys. Res. 86, 10545-68.
Lemeyre, J., & Bowden, P., in press. Plutonic type series: discrimination of various granitoid series and related
rocks by their modal composition.
Lopez-Escobar, L., Frey, F. A., & Oyarzun, J., 1979. Geochemical characteristics of Central Chile (33°-34° S)
granitoids. Contr. Miner. Petrol. 70,439-50.
Malpas, J., 1979. Two contrasting trondhjemite associations from transported ophiolites in Western
Newfoundland: initial report. In: Barker, F. (ed.), Trondhjemites and Related Rocks. Amsterdam: Elsevier
Scientific Publishing Company, 465-87.
982 J. A. P E A R C E ET AL.

Mason, D. R., & McDonald, J. A., 1978. Intrusive rocks and porphyry copper occurrences of the Papua-New
Guinea-Solomon Islands region: a reconnaissance study. Econ. Geol. 73, 857-77.
Melson, W. G., 1978. Chemical stratigraphy of Leg 45 basalts: electron probe analyses of glasses. In: Melson,
W. G., & Rabinowitz, P. O. (eds.), Initial Rep. Deep Sea drill. Proj. 45, 507-12.
Michael, P. J., 1983. Chemical differentiation of the Bishop Tuff and other high silica magmas through
crystallization processes. Geology, 11, 31—4.
Neumann, E. R., Brunfelt, A. O., & Finstad, K. G., 1977. Rare earth elements in some igneous rocks in the Oslo
rift, Norway. Lilhos, 10, 311-19.
Ohnenstetter, M., & Ohnenstetter, D., 1980. Comparison between Corsican albitites and oceanic plagiogranites.
Arch. Sc. Geneve. 33, 201-20.
Oliver, R. A., Vittoz, P., de Boisset, T., Vivier, G., & Kerr, S. A., 1983. Bimodal magmatism in the Grandes
Rousses Massif, Western French Alps. J. Open Univ. Geol. Soc. 4, 228-33.
Peacock, M. A., 1931. Classification of igneous rock series. J. Geol. 39, 65-7.
Pearce, J. A., & Cann, J. R., 1973. Tectonic setting of basic volcanic rocks investigated using trace element
analyses. Earth planet. Sci. Lett. 19, 290-300.
1975. Basalt geochemistry used to investigate past tectonic environments on Cyprus. Tectonophysics, 25,

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on April 29, 2014

41-67.
& Norry, M. J., 1979. Petrogenetic implications of Ti, Zr, Y and Nb variation in volcanic rocks. Contr.
Miner. Petrol. 69, 33-47.
& Gale, G. H., 1979. Identification of ore-deposition environment from trace element geochemistry of
associated igneous host rocks. In: Volcanic Processes in Ore Genesis. Spec. Publ. Inst. Min. Metall. and
Geol. Soc. London, 14-24.
Alabaster, T., Shelton, A. W., & Searle, M. P., 1981. The Oman ophiolite as a Cretaceous arc-basin
complex: evidence and implications. Phil. Trans. R. Soc. Lond. A300, 299-317.
1983. The role of subcontinental lithosphere in magma genesis at destructive plate margins. In:
Hawkesworth, C. J., & Norry, M. J. (eds), Continental Basalts and Mantle Xenoliths. Nantwich: Shiva
Publications, 230-49.
1982. Trace element characteristics of lavas from destructive plate boundaries. In: Thorpe, R. S. (ed.),
Andesites. New York: John Wiley & Sons, 525-48.
— Lippard, S. J., & Roberts, S., in press. Characteristics and tectonic significance of supra-subduction zone
ophiolites. In: Kokelaar, P., & Howells, M. (eds.), Marginal Basins. Geol. Soc. Spec. Publ.
Peccerillo, A., & Taylor, S. R., 1976. Geochemistry of Eocene calc-alkaline volcanic rocks from the Kastamonu
area, Northern Turkey. Contr. Miner. Petrol. 58, 63-81.
Perfit, M. R., Brueckner, H., Lawrence, J. R., & Kay, R. W., 1980. Trace element and isotopic variations in a
zoned pluton and associated volcanic rocks, Unalaska Island: Alaska: a model for fractionation in the
Aleutian calc-alkaline suite. Contr. Miner. Petrol. 73, 69-87.
Pitcher, W. S., 1983. Granite: typology, geological environment and melting relationships. In: Atherton, M. P., &
Gribble, C. D. (eds.), Migmatites, Melting and Metamorphism. Nantwich: Shiva Publications, 277-87.
Potts, P. J., Thorpe, O. W., & Watson, J. S., 1981. Determination of the rare-earth element abundances in 29
international rock standards by instrumental neutron activation analysis: a critical appraisal of calibration
errors. Chem. Geol. 34, 331-52.
Webb, P. C , & Watson, J. S., 1984. Energy dispersive X-ray fluorescence analysis of silicate rocks for
major and trace elements. X-ray Spectrom. 13, 2-15.
Saunders, A. D., Tarney, J., Stern, C. R., & Dalziel, I. W. D., 1979. Geochemistry of Mesozoic marginal basin
floor igneous rocks from southern Chile. Bull. geol. Soc. Am. 90, 237-58.
Tarney, J., & Weaver, S. D., 1980. Transverse geochemical variations across the Antarctic Peninsula:
implications for calc-alkaline magma genesis. Earth planet. Sci. Lett. 46, 344-60.
Shand, S. J., 1951. Eruptive Rocks. New York: J. Wiley.
Sharaskin, A. Ya., Pustchin, I. K., Zlobin, S. K., & Kolesov, G. M., 1983. Two ophiolite sequences from the
basement of the Northern Tonga arc. Ofioliti. 8,411-30.
Shervais, J. W., 1982. Ti-V plots and the petrogenesis of modern and ophiolitic lavas. Earth planet. Sci. Lett. 59,
101-18.
Streckeisen, A., 1976. To each plutonic rock its proper name. Earth Sci. Rev. 12, 1-33.
Sun, S. S., & Nesbitt, R. W., 1977. Chemical heterogeneity of the Archaean mantle, composition of the Earth and
mantle evolution. Earth planet. Sci. Lett. 35,429-48.
Thirlwall, M. F., & Jones, N. W., 1983. Isotope geochemistry and contamination mechanics of Tertiary lavas from
Skye, Northwest Scotland. In: Hawkesworth, C. J., & Norry, M. J. (eds.), Continental Basalts and Mantle
Xenoliths. Nantwich: Shiva Publications, 186-208.
Thompson, R. N., Dickin, A. P., Gibson, I. L., & Morrison, M. A., 1982. Elemental fingerprints of isotopic
contamination of Hebridean Palaeocene mantle-derived magmas by Archaean sial. Contr. Miner. Petrol. 79,
159-68.
Tindle, A. G., & Pearce, J. A., 1983. Assimilation and partial melting of continental crust: evidence from the
mineralogy and geochemistry of autoliths and xenoliths. Lithos, 16, 185-202.
Vidal, P., Cocherie, A., & Le Fort, P., 1982. Geochemical investigations of the origin of the Manaslu leucogranite
(Himalaya, Nepal). Geochim. cosmochim. Ada. 46, 2279-92.
 TRACE ELEMENTS OF G R A N I T I C ROCKS 983

Walsh, J. N., Beckinsale, R. D., Skelhorn, R. R., & Thorpe, R. S., 1979. Geochemistry and petrogenesis of
Tertiary granitic rocks from the Island of Mull, Northwest Scotland. Contr. Miner. Petrol. 71, 99-116.
Weiss, D., 1983. Pb isotopes in Ascension Island rocks: oceanic origin for the gabbroic to granitic plutonic
xenoliths. Earth planet. Sci. Lett. 62, 273-82.
White, A. J. R., & Chappell, B. W., 1977. Ultrametamorphism and granitoid genesis. Tectonophys. 43, 7-22.
1979. Sources of granite magmas. Geol. Soc.Am. Ann. Gen. Meeting, 539.
Winchester, J. A., & Floyd, P. A., 1977. Geochemical discrimination of different magma series and their
differentiation products using immobile elements. Chem. Geol. 20, 325—43.
Wood, D. A., 1979. A variability veined sub-oceanic upper mantle-genetic significance for mid-ocean ridge basalts
from geochemical evidence. Geology, 7, 499-503.
Joron, J.-L., & Treuil, M., 1979. A re-appraisal of the use of trace elements to classify and discriminate
between magma series erupted in different tectonic settings. Earth planet. Sci. Lett. 45, 326-36.

Downloaded from http://petrology.oxfordjournals.org/ at Aston University on April 29, 2014