Lithos 46 Ž1999.

535–551

Aluminium saturation in I- and S-type granites and the

characterization of fractionated haplogranites
)
B.W. Chappell
ARC Key Centre for the Geochemistry and Metallogeny of the Continents (GEMOC), Department of Geology,
Australian National UniÕersity, Canberra, ACT 0200, Australia
Received 27 January 1998; accepted 29 June 1998

Abstract

Granites of the Lachlan Fold Belt resulted from partial melting of the crust. In most cases, fusion involved mainly quartz
and feldspar, producing felsic melts. Varying degrees of separation of those melts from the unmelted source rock Žrestite.
were responsible for much of the compositional variation seen in the granites of the belt. Less commonly, melting occurred
at higher temperatures forming more mafic melts, such as for the I-type Boggy Plain Supersuite and the S-type Koetong
Suite. Hence, the felsic haplogranites of the Lachlan belt dominantly formed initially as primary melts that separated from
restite and less often by the fractionation of more mafic melts. Source rocks of the I- and S-type granites were undersaturated
or oversaturated in Al, respectively, and the more mafic granites share that characteristic with their source. As the magmas
of the Boggy Plain Supersuite evolved progressively by fractional crystallization, the rocks trended towards saturation in Al,
to eventually form a mode close to Al saturation. Other felsic I-type magmas, formed directly by partial melting, were
generally more oversaturated in Al, as were the corresponding S-type melts derived from peraluminous source rocks. In an
unfractionated state, there are some overlaps in the degree of Al saturation in these magmas produced by partial melting.
However, when extended fractional crystallization of these felsic partial melts took place, an almost complete separation in
Al-saturation developed between I-type and more peraluminous S-type melts. Because apatite is soluble in peraluminous
melts, P became progressively more abundant in the S-type melts as they fractionated. This led to contrasts in the
abundances of P and of elements such as Y, the rare earth elements, and Th, between the strongly fractionated I- and S-type
granites. Hence, such granites can easily be distinguished from each other. q 1999 Elsevier Science B.V. All rights reserved.

Keywords: Granite; Lachlan; S-type granite; I-type granite; Fractional crystallization

1. Introduction Ž1974. first proposed the I- and S-type granite1

subdivision. Later contributions ŽChappell and White,
It was at an earlier granite symposium in South
America ŽSantiago, Chile. that Chappell and White

1
The term ‘granite’ is used here in a general sense to include
)
Tel.: q61-2-6249-2062; Fax: q61-2-6249-5544; E-mail: all coarse-grained igneous rocks comprising quartz and feldspars,
bruce.chappell@anu.edu.au the ‘granitoid’ of some authors.

0024-4937r99r$ - see front matter q 1999 Elsevier Science B.V. All rights reserved.
PII: S 0 0 2 4 - 4 9 3 7 Ž 9 8 . 0 0 0 8 6 - 3
536 B.W. Chappellr Lithos 46 (1999) 535–551

1984, 1992. provided further details of the I- and 2. Aluminium saturation and the aluminium
S-type subdivision of granites in the Lachlan Fold saturation index (ASI)
Belt ŽLFB. of southeastern Australia. The two groups
are based on features of granites that can be ob- Na and Ca are removed in solution during weath-
served or measured, and which were inferred to ering. Those elements therefore have lower abun-
relate to features of the source rocks. Recognition of dances in those sedimentary rocks that can be par-
the two types focused attention on the role of source tially melted to produce a granite than they have in
rock compositions in determining many of the chem- unweathered igneous rocks. More mafic granites,
ical compositional features of granites, which in turn being closer to source rock compositions than felsic
either determine, or are related to, the mineralogy of granites, therefore have lower abundances of Na and
granites additional to the essential quartz and Ca in S-type relative to I-type granites. Hence, one
feldspars. of the most useful ways of discriminating the two
The I- and S-type subdivision has been widely, granite types is by use of the molar ratio Al 2 O 3r
but not universally, accepted as a useful concept. It ŽNa 2 O q K 2 O q CaO.. This ratio has been referred
might be expected to be an oversimplification, since to as ArCNK by Clarke Ž1981. and the aluminium
in natural systems, one could expect a continuous saturation index ŽASI. by Zen Ž1986.. ArCNK has
gradation in the compositional properties of source been widely used but use of ASI is preferred because
rocks. However, the view can be taken that source the name stresses the importance of the concept of
rocks should be unweathered or not sufficiently Al saturation. The terms metaluminous and peralu-
weathered to affect their chemical composition Žig- minous of Shand Ž1927., corresponding to CIPW
neous., or else weathered, or derived from material normative diopside Ždi. or corundum ŽC., are like-
weathered, to a sufficient extent for that to be ex- wise valuable. The boundaries between those pairs of
pressed in the composition Žsedimentary., with prob- terms correspond to ASI s 1 if the CaO value is
ably only a small region of overlap. At least for the corrected for apatite. In this paper, the ASI values
LFB, such a relatively sharp subdivision is supported are not corrected for apatite.
by the strongly bimodal character of the derived Miller Ž1985. proposed that peraluminous granites
granites for many elements in those granites that are be referred to as strongly peraluminous ŽPs. when
closest to the source rocks in composition. Also, the they contain a mineral more aluminous than biotite
two types of sources are probably fundamentally and weakly peraluminous ŽPw. when they do not.
different in the sense that the igneous sources were This is a useful mineralogical distinction and will be
generally infracrustal, and the sedimentary ones broadly followed here, although the differences in
supracrustal ŽChappell and White, 1984.. Al-saturation recognized in this paper are based on
This paper is concerned principally with examin- chemical compositions alone. Miller Ž1985. argued
ing the degree of Al-saturation in the granites of the for the use of mineralogical criteria partly on the
LFB, both in rocks that have and have not undergone basis of avoiding chemical analytical problems,
fractional crystallization. The degree of Al-satura- which is mostly valid. He noted that more than
tion, in both I- and S-type granites is traced through 3–4% C may be reported in norms when the primary
from the most mafic to the most felsic and strongly mineralogy would not be consistent with so much
fractionated compositions. The clear separation be- excess Al 2 O 3 . Here, we consider granites with up to
tween the I- and S-type granites that is evident at the about 4% C, including WG57 and VB140 in Table 1
most mafic compositions, also manifests itself in the and with the average Cornubian granite ŽCNG. in
most fractionated rocks. The often profound miner- that table containing 3.3% C. All of the strongly
alogical differences between the two granite types, peraluminous rocks considered here have modes
often related to Al-saturation, will not be exa- consistent with that feature, with common mus-
mined here. Chappell and White Ž1992. have pro- covite, and perhaps also andalusite or cordierite.
vided information on those mineralogical and Stoichiometric muscovite contains 25.6% C, so that
other more general aspects of the I- and S-type mineral is generally a major contributor to very high
groups. C contents. However, the use of Al-saturation based
 B.W. Chappellr Lithos 46 (1999) 535–551 537

Table 1
Compositional data for some felsic granites
G223 BB21 TB5 TB130 ZB107 DCC34 KB102 WG61 WG57 VB140 CNG VB98 AB141
SiO 2 75.30 76.56 73.79 76.31 76.48 75.58 73.81 70.54 73.72 72.48 72.35 73.49 77.12
TiO 2 0.15 0.19 0.20 0.06 0.14 0.03 0.30 0.51 0.05 0.25 0.26 0.12 0.12
Al 2 O 3 13.08 12.60 12.88 12.40 11.81 12.50 13.22 14.44 14.79 14.50 14.52 14.54 11.79
Fe 2 O 3 0.28 0.43 0.38 0.32 1.05 0.46 0.62 0.39 0.17 0.29 0.30 0.32 0.40
FeO 1.06 0.43 1.41 0.84 0.73 0.47 1.65 2.63 0.52 1.33 1.56 0.75 0.64
MnO 0.06 0.02 0.03 0.02 0.04 0.02 0.04 0.05 0.05 0.03 0.06 0.04 0.02
MgO 0.25 0.21 0.21 0.05 0.03 0.05 0.60 0.79 0.07 0.47 0.41 0.21 0.04
CaO 1.10 0.97 1.18 0.60 0.32 0.55 1.68 1.54 0.40 0.66 0.79 0.64 0.34
Na 2 O 3.45 3.52 3.00 3.50 4.21 3.95 2.97 2.91 4.00 2.80 2.96 3.28 3.07
K 2O 4.75 4.16 5.17 4.91 4.84 4.50 4.05 4.78 4.11 5.20 5.12 4.72 5.12
P2 O5 0.05 0.03 0.06 - 0.01 0.01 - 0.01 0.09 0.20 0.42 0.33 0.25 0.29 0.02

Trace elements (ppm)

Rb 288 214 349 805 230 745 187 275 945 419 483 496 248
Cs 13 4 15 27 5 35 10 18 44 33 49 79 9
Sr 160 99 82 6 7 3 89 89 15 65 74 37 38
Ba 420 545 330 4 56 4 495 360 2 220 176 86 520
Zr 126 115 181 140 470 142 139 206 27 99 113 46 171
Nb 14 17 13 30 24 39 8 19 47 20 17 20 19
Y 26 26 56 158 64 165 40 36 12 14 18 12 94
La 18 40 51 46 58 17 28 43 5 21 29 8 62
Lu 0.49 0.50 0.89 2.6 1.00 3.1 0.57 0.53 0.14 0.14 0.19 0.10 1.30
Cr 2 -1 2 -1 -1 -1 8 9 -1 9 8 3 -1
Ga 16 12 19 30 20 19 15 19 26 20 24 20 20
Sn 6 3 5 44 7 19 6 6 39 25 16 47 12
Th 37 20 50 79 28 44 19 25 3 14 15 6 28
U 13 4 3 30 7 6 4 5 21 11 13 5 6
ASI 1.021 1.045 1.016 1.019 0.927 a 1.011 1.073 1.132 1.258 1.268 1.226 1.246 1.052
C 0.38 0.61 0.35 0.24 0a 0.13 1.11 2.16 4.04 3.86 3.27 3.57 0.63
Qb 37.1 41.0 37.1 37.5 34.7 36.1 42.2 36.0 37.3 38.6 37.3 38.6 40.7
ab b 32.1 32.3 28.5 31.6 35.4 35.6 29.6 29.8 36.5 26.7 28.4 30.6 27.4
or b 30.8 26.7 34.4 30.9 29.9 28.3 28.2 34.2 26.2 34.7 34.3 30.8 31.9

Sample Unit Batholith Latitude Longitude

X X
‘Unfractionated’ I-type granites G223 Bendemeer New England c 30852.80 S 151813.25 E
X X
BB21 Maffra Berridale 36832.10 S 148858.55 E
X X
‘Fractionated’ I-type granites TB5 Coles Bay Bassian 42804.90 S 148817.95 E
X X
TB130 The Hazards Bassian 42809.50 S 148818.90 E
X X
ZB107 Nallawa Yeoval 32836.70 S 148830.85 E
X X
DCC34 O’Briens Ck Herbertonc 17859.85 S 144802.55 E
X X
‘Unfractionated’ S-type granite KB102 Happy Jacks Kosciuszko 36803.50 S 148829.60 E
X X
‘Fractionated’ S-type granites WG61 Interview Taswegia 41837.70 S 144854.55 E
X X
WG57 Sandy Cape Taswegia 41825.35 S 144844.75 E
X X
VB140 Granya Wagga 36807.50 S 147819.20 E
CNG Mean value Cornubianc
X X
VB98 Mt Flakney Wagga 35813.50 S 147820.05 E
X X
‘Unfractionated’ A-type granite AB141 Mumbulla Bega 36835.50 S 149855.15 E

C is the percent of CIPW normative corundum.

a
Sample ZB107 contains 1.52% normative acmite.
b
Normative Q, ab, and or are recalculated to a total of 100%.
c
Not part of the LFB.
538 B.W. Chappellr Lithos 46 (1999) 535–551

on chemical analyses should not be made without a quartz ŽQ., albite Žab. and orthoclase Žor., and min-
knowledge of the petrography of the rocks, so that eralogically by similar abundances of quartz, K-
one is not simply measuring a mass of clay minerals feldspar and Na-plagioclase, with minor amounts of
produced by hydrothermal alteration. Of course, a other minerals. To a first approximation, such gran-
major element analysis alone can be used to confirm ites have compositions that are independent of their
that the proportions of normative Q, ab, and or are precursor materials and separation into I- or S-type,
close to those of a minimum temperature melt ŽTable at least using major element and mineral abundances,
1 and Fig. 7., and it is important that it be done. is often not possible, unless the granite is part of a
Zen Ž1986. considered in some detail the produc- suite that extends to more mafic Žor, as this paper
tion of peraluminous melts by the fractional crystal- shows, more fractionated. compositions where com-
lization of originally metaluminous melts. The devel- positional distinctions emerge. Because haplogranites
opment of peraluminous compositions by the frac- are dominated by quartz and feldspars ŽASI s 1., the
tional crystallization of hornblende ŽASI - 0.5. had ASI also converges to values close to one for all of
previously been proposed by Cawthorn and Brown those rocks, and the values for I- and S-type granites
Ž1976.. Zen Ž1986. pointed out that such a process overlap; this is shown for the granites of the LFB in
would be inhibited by the precipitation of feldspar Fig. 1. This figure shows that the S-type granites are
ŽASI s 1. while the melt remains metaluminous, but always oversaturated in Al ŽASI ) 1.. The I-type
for the same reason the melt would become progres- granites may be either metaluminous or peralumi-
sively more peraluminous with the removal of nous, and they overlap the S-type granites. Rocks of
feldspar, once it became saturated in Al. Zen Ž1986. the I-type Boggy Plain Supersuite ŽWyborn et al.,
noted that the generation of peraluminous composi- 1987., which are either cumulates or derived liquid
tions by fractional crystallization in that way is compositions ŽSection 4.4., are not included in Fig.
inefficient and that larger bodies of peraluminous 1. The I-type granites represented in Fig. 1, generally
magma require a more efficient mechanism for their correspond to magma compositions that were either
derivation. Such large bodies of magma can be
produced by partial melting of peraluminous source
rocks ŽS-type.. However, peraluminous felsic I-type
granites are common in many regions, and they
comprise 41% of all I-type granites of the LFB.
Clearly, another mechanism for producing peralumi-
nous I-type compositions is required, and it is pro-
posed here that these weakly peraluminous composi-
tions result largely from partial melting rather than
fractional crystallization ŽSection 5.3.. Those compo-
sitions may in turn be modified by the precipitation
of feldspar so that the final rocks may be more
peraluminous ŽFigs. 3 and 4..

3. Overlap between I- and S-type granite compo-

sitions
Fig. 1. Histograms of ASI values for 1025 I-type granites exclu-
sive of the Boggy Plain Supersuite, and 764 S-type granites, from
A major difficulty in applying the I- and S-type the LFB. ASI values of the I-type granites range from 0.774 to
subdivision is that more felsic granites converge 1.154, and the S-type granites from 1.001 to 1.997. Three S-type
granites with ASI values of 1.872, 1.882, and 1.997 are not
towards the minimum temperature composition ŽTut- shown. The I-type data include seven samples for which SiO 2 -
tle and Bowen, 1958.. Such haplogranites are charac- 57% Žgabbroic diorites. but there are none for which SiO 2 - 53%
terised by approximately equal amounts of normative Žgabbros..
 B.W. Chappellr Lithos 46 (1999) 535–551 539

melts, or melts containing entrained restite crystals components as a solid residue, which may have
ŽChappell et al., 1987.. Almost half Ž46.5%. of those disengaged from the melt at or near the source, or
1025 I-type compositions, exclusive of the Boggy later. In both of these cases, the felsic melts have the
Plain Supersuite, have ASI values greater than 1.0, distinctive major element compositions that reflect
and therefore overlap with some S-type granites. The the equilibrium between quartz, feldspars, and a
importance of peraluminous I-type granites in the hydrous melt during their formation as observed by
LFB, or elsewhere, is perhaps not generally realized, Tuttle and Bowen Ž1958..
and it shows that it is inadvisable to refer to those
granites in a group as metaluminous, in contrast to
the ‘metasediment-derived’ or even the ‘per-
4.2. Major element compositions of haplogranites
aluminous metasediment-derived types’, as Wall et
al. Ž1987. have done. In proposing the I- and S-type
classification, Chappell and White Ž1974. were aware Analyses of 13 haplogranites, mostly from the
that I-type granites may be peraluminous, which is LFB, are listed in Table 1. These analyses show the
partly why they suggested terms not intrinsically limited range in major element compositions for
related to Al-saturation. such rocks, regarding P as a trace element. Such
rocks are very siliceous Ž73–77% SiO 2 .. Except for
the strongly fractionated S-type granites, they con-
4. Petrogenetic framework tain low Al 2 O 3 Ž- 13%., and they have very low to
low contents of Mg, Ca, and the transition elements.
4.1. The importance of the haplogranites Na and K have high and relatively constant abun-
dances. These rocks are dominated by four elements,
The I- and S-type subdivision must be examined Si, Al, Na and K, which do not vary greatly in
in terms of other compositional properties of granites amount, and the variation in abundances of the nor-
that reflect on their origin. The most important of mative minerals Q, ab, and or, that incorporate those
those features is that the most felsic granites, or four elements are likewise restricted ŽTable 1..
haplogranites, almost universally have the composi- The analyses in Table 1 include four unfraction-
tions of low-temperature hydrous silicate melts in ated rocks, two I-type ŽG233 and BB21., one S-type
equilibrium with quartz and feldspar ŽTuttle and ŽKB102., and one A-type ŽAB141.. All the other
Bowen, 1958.. That study established the magmatic rocks show compositional effects of fractional crys-
character of granites and while Bowen himself did tallization. The pairs TB5 and TB130 ŽI-type., and
not regard granite magmas as likely to be primary, WG61 and WG57 ŽS-type., represent the least and
he noted that studies showing that granite is a late- most evolved granites available from two suites that
crystallizing residuum of fractional crystallization underwent extreme fractional crystallization. ZB107
imply that it would likewise be an early product of is a rock that apparently fractionated from an origi-
selective fusion of appropriate material ŽBowen, nally rather mafic metaluminous composition, that is
1947.. Thus, the simple haplogranites can form in itself slightly peralkaline. DCC34 is a strongly frac-
two ways, both of which are seen among the granites tionated representative of what is certainly one of the
of the LFB. best and most extensive examples worldwide of
Some felsic granites of the LFB were produced by highly evolved granites, the Carboniferous I-type
the removal by fractional crystallization of mafic granites of northern Queensland ŽChampion and
components from less felsic silicate melts, so that the Chappell, 1992.. VB140 is a fairly strongly fraction-
more mafic rocks represent cumulates, and the more ated S-type granite from the Wagga Batholith of the
felsic ones liquids. This is illustrated by the I-type LFB, which is strikingly close in major element and
Boggy Plain Supersuite and the S-type Koetong Suite. many trace element abundances, to the average Cor-
More commonly in the LFB, haplogranites formed nubian granite CNG of Chappell and Hine Žunpub-
under conditions where only the felsic components lished data.. VB98 is a more strongly fractionated
of the source rocks were fused, leaving the mafic S-type granite from the Wagga Batholith.
540 B.W. Chappellr Lithos 46 (1999) 535–551

4.3. Trace element compositions of haplogranites in the LFB by fractional crystallization is seen only
in rocks of the Koetong Suite ŽChappell, 1996b.,
In contrast to the major element compositions, although some other felsic S-type granites that are
many trace elements can vary widely in concentra- not associated in the field with more mafic rocks
tion in felsic rocks, with some of those variations might also have formed in such a way. For Koetong,
illustrated in Table 1. Those trace elements that at SiO 2 contents greater than about 69%, the abun-
occur in feldspars Že.g., Rb, Sr, Ba. and those that dance of elements such as Rb, Nb and Sn begin to
are major components of accessory minerals Že.g., P, increase rapidly in abundance with increasing SiO 2
Zr, Nb, Y, Ce, Sn, Th, U. may vary widely in content, while others including Sr, Y and Th start to
abundance both within and between rock suites. decrease, consistent with fractional crystallization.
Those that are found in mafic minerals Že.g., Cr. are The haplogranites of the Boggy Plain Supersuite and
invariably low in amount. Those trace elements that the Koetong Suite were therefore initially produced
occur in feldspars and accessory minerals, and which by fractional crystallization of more mafic melts. In
may as a result have abundances that vary with, and that respect, they are unusual among the felsic gran-
reflect, the degree of fractional crystallization of ites of the LFB.
such minerals from a felsic melt, are a major focus
of this discussion. In a sense, all granites are frac- 4.5. Production of haplogranites by partial melting
tionated rocks, in that they have compositions that
differ from other igneous rocks or average crust. More commonly, it was partial melting of the
However, in this paper, the term fractionated is used crust, of earlier igneous or sedimentary components,
to refer to those felsic granites that Ža priori. show that produced the LFB haplogranites, at or close to
the compositional effects of fractional crystallization minimum temperature melt compositions Žnot H 2 O-
of feldspars and accessory minerals; those that do saturated.. Evidence for these rocks representing pri-
not are said to be unfractionated. Examples of both mary magmatic compositions, includes the absence
groups are provided in Table 1 and the different Žor at least rarity. of cumulate rocks and the absence
behaviour of some trace elements in fractionated I- of mafic rocks analogous to those of the Boggy Plain
and S-type granites will be considered in detail in Supersuite. For these I-type granites, the minimum
later sections. SiO 2 content is 54.4% and for only 2.7% of all
analyzed 1025 samples is the SiO 2 below 60%. For
4.4. Production of haplogranites by fractional the S-type granites, the minimum SiO 2 content is
crystallization 63.4%. The relative absence of mafic rocks not only
implies that cumulate rocks were not formed, but
The I-type Boggy Plain Supersuite ŽWyborn et also that the more mafic compositions corresponding
al., 1987., which comprises approximately 5% of all to molten magmas that might have been the source
exposed granites of the LFB, has compositional fea- of these granites through fractional crystallization,
tures which show that the range of rocks was pro- are uncommon. Further evidence against a process of
duced by fractional crystallization from relatively fractional crystallization in producing these rocks,
mafic liquids, so that the more mafic rocks represent comes from the lack of inflexions in concentrations
cumulates and the more felsic granites and related of elements in variation diagrams, and the character-
volcanic rocks generally represent liquid composi- istic linear variations of many elements. Also, Chap-
tions. Evidence for this includes the wide range in pell Ž1996b. has shown, by modelling variations of
composition Ž45.34 to 77.32% SiO 2 ., inflexions of some trace elements in the I-type Glenbog and
elements on variation diagrams, and the occurrence Moruya suites, and the S-type Bullenbalong Suite,
of rocks with cumulate chemical features, such as that variation in those suites cannot have resulted
high Ca and Al, or Ca and Mg, or Cr, or extremely from the development of cumulate rocks by frac-
low abundances of K, P, Rb, Zr, etc., which were tional crystallization; similar calculations show that
components of a trapped interstitial melt. Direct likewise, those suites cannot represent liquid lines of
evidence for the production of S-type haplogranites descent resulting from fractional crystallization. Fur-
 B.W. Chappellr Lithos 46 (1999) 535–551 541

thermore, processes of magma mixing and hybridiza- Chappell and White Ž1974. nominated a value for
tion are localized and insignificant on a pluton scale the ASI significantly greater than 1.0 as the bound-
and the popular model that some granite suites result ary between the I- and S-types, despite the fact that a
from the interaction of mantle melts and crust cannot value of 1.0 might be arbitrarily taken as a boundary
be sustained by the data for the LFB ŽChappell, between unweathered and weathered source rocks. In
1996a.. part, they did that because of the observed values for
the Berridale Batholith. However, they also recog-
4.6. Production of more mafic granites nized more generally that very felsic granites of
I-type suites may be weakly peraluminous, while
The bulk of the haplogranites of the LFB acquired more mafic compositions within that suite are meta-
that property as a primary feature by the partial luminous. The defining sample in that regard is
melting of quartz and two feldspars. Some more G223 of Table 1, a very felsic granite from the New
mafic magmas, such as those that produced the England Batholith ŽChappell, 1978.. This sample is
Jindabyne Suite ŽHine et al., 1978. or the Koetong associated in the Bendemeer pluton with rocks hav-
Suite, formed as a result of melting to higher temper- ing SiO 2 contents down to 65.7% SiO 2 and contain-
atures, such that there was an absence of quartz, ing up to 7% hornblende. G223 contains 3% biotite
K-feldspar or H 2 O in the residue. Sometimes, but and small amounts of hornblende, less than one
not generally, those magmas underwent fractional small crystal per thin-section. Despite the presence
crystallization. In the extreme case of the Boggy of hornblende, G223 is peraluminous; it contains
Plain Supersuite, melting at high temperatures pro- 0.38% C and has an ASI of 1.021. This rock is, or is
duced a hot completely molten magma that yielded a very close to, a minimum temperature melt produced
wide range of rocks by fractional crystallization. from metaluminous source rocks. It shows empiri-
However, in most cases, the more mafic granites of cally that minimum temperature granite melts, formed
the LFB acquired that character by the entrainment in equilibrium with such source rocks, are slightly
of crystals of unmelted source material Žrestite. in a Al-oversaturated. This is in accord with experimental
low temperature melt ŽChappell et al., 1987; Chap- studies showing that peraluminous melts can be pro-
pell, 1996b.. duced by partial melting of mafic rocks Že.g., Helz,
1976; Ellis and Thompson, 1986.. The composition
of another I-type granite representing a partial melt
5. Aluminium saturation in granites of the LFB composition, rock BB21 from the Berridale Batholith,
is also shown in Table 1. This is slightly more
5.1. Aluminium saturation in I- and S-type granites peraluminous than G223, with a C content of 0.61%
and an ASI of 1.045. This rock contains a small
Chappell and White Ž1974. noted that the distinc- amount of muscovite Ž0.3%. that is possibly sub-
tive mineralogical features of I- and S-type granites solidus, rather than hornblende but its I-type charac-
correlate with differences in the abundance of Al ter is confirmed by an initial 87 Srr86 Sr ratio ; 0.705.
relative to Na, K and Ca in the two types, which can The composition of an S-type granite close to a
now be expressed as the ASI. Chappell and White minimum temperature melt is also given in Table 1.
Ž1974. proposed that the boundary between the two This rock KB102 is significantly more saturated in
types be drawn at an ASI of 1.1. At that time, the Al than the two I-type granites discussed earlier,
distinction between the two types was largely based with 1.11% C and an ASI value of 1.073. This is to
on observations of the Berridale Batholith, although be expected in a composition resulting from partial
analogous differences had been observed elsewhere melting in equilibrium with peraluminous minerals,
in the LFB. From the current LFB database, the which show an appreciable solubility in felsic melts.
range in ASI for the I-type granites of the Berridale However, it must be noted that there are felsic S-type
Batholith is from 0.902 to 1.092 with only one granites in the LFB which are less peraluminous than
sample greater than 1.065, and for the S-type gran- KB102, just as there are felsic I-type granites than
ites from 1.044 to 1.248. are more peraluminous than BB21.
542 B.W. Chappellr Lithos 46 (1999) 535–551

5.2. Aluminium saturation in the Boggy Plain Super-

suite

As noted earlier, the data represented in Fig. 1

show the degree of overlap in ASI between the I-
and S-type granites of the LFB. A similar diagram
was provided by Chappell and White Ž1992., but that
figure included data for the Boggy Plain Supersuite,
which are here plotted separately in Fig. 2. The
cumulate character of at least some of the mafic
rocks of that supersuite is confirmed by some very
low vales of ASI, down to 0.178 for a clinopyrox-
ene-rich cumulate, and 61 of the Boggy Plain rocks
have ASI values less than the lower limit of that
parameter of 0.774 for the other I-type granites of
the LFB shown in Fig. 1. The data represented in
Fig. 2 are a composite of values for rocks formed as
Fig. 3. Histograms of the most felsic ŽFe-poor. 10% of I- and
cumulates and from the derived liquids. For the S-type granites of the LFB having Sr contents greater than 50
zoned and intensively analyzed Boggy Plain pluton ppm. The I-type granites are exclusive of the Boggy Plain Super-
Ž34 km2 ; Wyborn, 1983., almost all granites are suite.
cumulate rocks and their ASI values extend up to
0.943; two rocks from that pluton that probably
correspond to fractionated liquid compositions have degree of Al-saturation that developed was restricted
ASI values of 0.990 and 1.024. Elsewhere in the ŽASI - 1.07. relative to those other peraluminous
supersuite, the ASI values extend from low values in I-type granites of the LFB ŽASI - 1.16. that formed
mafic cumulate rocks continuously up to a maximum directly by partial melting, followed in some cases
ASI of 0.95, overlapping slightly with melt composi- by fractional crystallization. Fig. 2 shows that frac-
tions a few of which extend down to 0.90. Some tional crystallization from initial mafic melts in the
fractionated melts were later enriched in crystals of Boggy Plain Supersuite resulted in a distinct peak at
K-feldspar for example, and are therefore cumulate an ASI close to one for the derived liquids. For one
rocks of a different kind. This supersuite provides an component of this supersuite, the most strongly frac-
excellent example of the development, by fractional tionated rocks include some peralkaline composi-
crystallization, of peraluminous melts from a metalu- tions, including sample ZB107 in Table 1; that ZB107
minous starting composition ŽZen, 1986., but the has true peralkaline character is confirmed by the
high Zr content of 470 ppm for a fractionated rock.

5.3. Aluminium saturation in felsic granites

Histograms of ASI values for felsic I- and S-type

granites from the LFB, exclusive of the Boggy Plain
Supersuite, are shown in Figs. 3 and 4. In Fig. 3, the
most felsic 10% of both types with Sr ) 50 ppm are
represented, and in Fig. 4, all granites with Sr - 50
ppm are shown. The separation based on Sr content
has been made to assess the effects of fractional
Fig. 2. Histogram of ASI values for 259 rocks of the I-type Boggy
Plain Supersuite from the LFB. The rocks include 26 gabbros
crystallization of haplogranite compositions, during
Ž - 53% SiO 2 . and 27 gabbroic diorites ŽSiO 2 s 53–57%.. Three which Sr would be removed by the separation of
ASI values at 0.178, 0.383, and 0.587 are not shown. feldspars. No single Sr content can perfectly separate
 B.W. Chappellr Lithos 46 (1999) 535–551 543

slightly but distinctly more peraluminous, centred at

close to ASI s 1.06. These data show that the less
peraluminous I-type haplogranites may become
slightly more peraluminous with fractionation,
whereas those initially more Al-saturated do not
change. There is no tendency for the Al-saturation of
the I-type granites to be driven to high values by the
fractional crystallization of feldspar. These data point
to the moderating effect of biotite precipitation in
this regard.
The fractionated S-type granites ŽFig. 4. have a
very similar range in Al-saturation to the most felsic
unfractionated rocks ŽFig. 3., but in contrast, the
fractionated granites show a distinct concentration of
values in the interval from ASI s 1.11 to 1.19. For
these rocks, their slightly more peraluminous compo-
sition when unfractionated ŽFig. 3. is apparently
sufficient for the precipitation of feldspars, with
Fig. 4. Histograms of I- and S-type granites of the LFB with Sr
contents less than 50 ppm. The I-type granites are exclusive of the ASI s 1, to drive them further into the peraluminous
Boggy Plain Supersuite. field. A comparison of Figs. 3 and 4 suggests that at
the same time, the most strongly peraluminous un-
fractionated melts ŽASI s 1.2 to 1.3 in Fig. 3. re-
fractionated from unfractionated granites, since the main largely unaffected by the fractional crystalliza-
value must depend on the initial Sr content of the tion, and may be driven to slightly less peraluminous
melt, which can vary significantly, and the condi- compositions. This effect could result from the pre-
tions under which the fractionation occurred. How- cipitation of a strongly Al-saturated mineral, such as
ever, the value of 50 ppm should achieve a separa- andalusite from the most peraluminous melt compo-
tion into compositions that have undergone fractiona- sitions.
tion, and those which have been mildly fractionated, There is very little overlap between the abun-
or not fractionated at all. dances of fractionated I- and S-type granites in Fig.
These compositions that are represented in Fig. 3 4. Only 13% of the fractionated S-type compositions
are taken to represent partial melts from metalumi- have an ASI less than 1.11, and 5% of the I-type
nous and peraluminous sources, with those melts at granites are above that value. This almost complete
most being only weakly fractionated. While there is separation between the degree of Al-saturation of
overlap between the two granite types, the data fractionated representatives of the two granite types,
represented in Fig. 3 broadly reflect the contrasting becomes very significant in considering the evolu-
Al-saturation of the sources. The S-type composi- tion of many trace elements during fractional crystal-
tions are all peraluminous, and for the I-type granites lization of the those two groups ŽSections 6 and 7..
that is dominantly the case Ž82% of total.. The felsic The separation also supports the division of the LFB
I-type granites are mainly between ASI s 0.98 and granites into the I- and S-types.
1.08, with a suggestion of bimodal maxima at ASI ;
1.00 and ASI ; 1.04. The S-type granites comprise a 5.4. Aluminium saturation in mafic granites
broad range of ASI values with no suggestion of a
single preferred composition. Chappell and White Ž1992. have previously shown
The range of Al-saturation in the fractionated that at the most mafic compositions, there is no
I-type granites, shown in Fig. 4, is almost identical overlap in ASI between the I- and S-type granites of
to that of the unfractionated equivalents in Fig. 3. the LFB. This is confirmed by the histograms in Fig.
However, the most abundant compositions are 5, which use a slightly larger database and exclude
544 B.W. Chappellr Lithos 46 (1999) 535–551

is invalid. Collins Ž1996. has made the interesting

argument that with a mixture of contrasting sources,
the granites derived from source rocks where either
component is dominant will be more common than
those derived from subequal proportions of the two
end-members. If that were the case, it would account
for the bimodality of the I- and S-type granites, as
Collins Ž1996. pointed out. However, the argument
that those granites were formed from mixtures of
contrasting sources does not stand scrutiny when the
isotopic and chemical data are looked at in conjunc-
tion. It can be shown that the changes in chemical
composition that would be required by the isotope
mixtures proposed by Keay et al. Ž1997. are not
Fig. 5. Histograms of the most mafic ŽFe-rich. 10% of I- and those that are observed. A critical element is Ca,
S-type granites of the LFB. The I-type granites are exclusive of which has an exceptionally low abundance in the
the Boggy Plain Supersuite. Three S-type granites with ASI values Ordovician sedimentary rocks of the LFB, averaging
of 1.873, 1.882, and 1.997 are not shown. 0.31% in 38 analyzed samples. For the I-type gran-
ites of the Bega Batholith, Ca has the highest abun-
dances in those suites which on simple isotopic
I-type rocks of the Boggy Plain Supersuite, which arguments would have contained the largest amounts
since the mafic rocks of that supersuite are cumu- of that low-Ca sediment in their source materials.
lates, means that they include a disproportionately Again, for the S-type granites of the Bullenbalong
large number of the most mafic granites. By exclud- Suite, the Ca content is too high, relative to that
ing most of the cumulate rocks, these mafic rocks sediment, to be accounted for by the presence of
generally represent magma compositions. If that mafic material in the source rocks in the amount
magma includes retained restite crystals, the view of required by the isotope data alone. The presence of a
this author, then the correlation between those com- more Ca-rich sedimentary source rocks at depth is
positions and source compositions is clear. If the inferred.
magmas represent liquid compositions, they will
again relate to source rock compositions, but less
precisely. In either case, the Al-saturation in a mafic 6. Compositional contrasts between fractionated
granite, less or greater than one, will correlate with I- and S-type granites
that property in the source rocks. Hence, the I- and
S-type subdivision is clear. The fact that the ASI 6.1. Contrasting behaÕiour of major and trace
values of I- and S-type granites converge, and even- elements during fractionation
tually overlap in felsic rocks, is of no consequence as
far as the I-S division is concerned, since a degree of Major element compositions show only slight
Al-oversaturation is an intrinsic property of the most variation when melts of haplogranite composition
felsic granitic melts. The overlap in the ASI parame- undergo continuing fractional crystallization. This
ter for felsic I- and S-type granites is to be expected can be seen by reference to Table 1, which includes
and therefore in no way invalidates the subdivision. both unfractionated haplogranites ŽG223, BB21,
These general conclusions have been reported KB102. and others that have undergone strong to
previously ŽChappell and White, 1992.. It is there- extreme fractionation. This situation arises because
fore surprising that Keay et al. Ž1997. concluded that the concentrations of Si, Al, Na and K are set by the
‘‘S- and I-type granites appear to be mixtures of, largely static equilibrium between quartz, feldspars
rather than unique products from, contrasting and the minimum temperature melt. However, Al
sources’’, implying that the I- and S-type separation decreases slightly with progressive I-type melt frac-
 B.W. Chappellr Lithos 46 (1999) 535–551 545

tionation, but increases for S-type fractionation be- values in the felsic granites, ranging from 0.02 to
cause of increasing saturation in Al Žincreasing C., in 0.42%. There is a concentration of points in the
the melt. The elements Ti, Fe, Mn, Mg and Ca, are low-P2 O5 part of the triangle, extending from the
already at low levels before fractional crystallization more mafic to intermediate compositions. These are
of the haplogranite melt occurs, and in absolute dominantly the granites of the Bullenbalong Super-
amounts decrease only slightly. The trace transition suite in which the compositional variation resulted
elements have low abundances in these rocks. Most from separation of a lower-P2 O5 melt such as KB102
other trace elements undergo significant changes in in Table 1, from a magma containing higher-P2 O5
abundance with increasing fractionation, and in fact mafic restite components. Rocks with higher P con-
provide the main evidence for the degree of fraction- tents reflect the role of fractional crystallization, as
ation that has occurred. Those elements that are in the more felsic parts of the Koetong Suite ŽVB140
incorporated in feldspars always behave consistently, and VB98 in Table 1.. The granites with highest
so that Rb, Cs, and Ga increase in abundance with P2 O5 contents are the most strongly fractionated of
fractionation, while Sr and Ba decrease. Some other the S-type granites, but are not at the highest SiO 2
elements are fairly constant in behaviour, so that Nb, values. This is because of the increasing abundance
Sn, and U, for example, increase with fractionation, of C in the most felsic rocks, so that in the extreme
while Zr decreases in abundance except in peralka- case about 4% C dilutes the other components, and
line liquids. Other elements, notably P, Y, the rare consequently, the rock fractionated to a lower SiO 2
earth elements, and Th, show varying patterns of content.
behaviour, governed by the accessory minerals that The higher P2 O5 abundances in the S-type gran-
precipitate, which are in turn largely determined by ites are consistent with their more peraluminous
the composition of the melt, and specifically by its compositions. Montel et al. Ž1988. demonstrated that
degree of saturation in Al. apatite is soluble in peraluminous melts. Wolf and
London Ž1994. have shown experimentally that the
solubilities of apatite in a haplogranite melt, and
6.2. Contrasting behaÕiour of P in I- and S-type equivalent P2 O5 abundances, increase linearly with
granites ASI, up to ; 0.63% of P2 O5 in a melt of ASI s 1.3.
The highest P2 O5 content measured in a fractionated
The contrasting patterns of behaviour of P in I- S-type granite of the LFB is 0.42% in two samples,
and S-type granites of the LFB are shown in Fig. 6. one of which is WG57 in Table 1, which has an ASI
The I-type granites show a band of compositions that value of 1.26 and a C content of 4.04%. Using the
decrease with increasing SiO 2 , converging on values data of Wolf and London Ž1994., this would suggest
less than 0.01% P2 O5 between 75 and 77.5% SiO 2 , that the fractionated S-type granites were never satu-
showing that the concentration of P in the I-type rated in P, so that the element behaved incompatibly
haplogranite melts is always very low. It is the and increased in abundance. The average CNG in
strongly fractionated I-type haplogranites that have Table 1, which contains 3.27% C, is correspondingly
the very low P contents Žcompare TB130 with the high in P2 O5 Ž0.25%.. The contrasting behaviour of
preceding three analyses in Table 1.. The decrease in P2 O5 during the fractional crystallization of I- and
P2 O5 contents with increasing SiO 2 for the I-type S-type granites will be considered further in Section
granites in Fig. 6 is consistent with the decreasing 7.2.
solubility of P in more siliceous melts ŽHarrison and
Watson, 1984., but for these rocks it is thought to be
due to the presence of different proportions of low-P 6.3. Contrasting behaÕiour of rare earth elements in
felsic melt and restite that includes apatite crystals. I- and S-type granites
The analogous plot for S-type granites shows a
strikingly different distribution, with the data lying in The concentrations of La and Y in I- and S-type
a triangular area extending from about 0.15% P2 O5 granites of the LFB are also plotted in Fig. 6. As a
in the most mafic rocks, to both higher and lower broad body of data, there is more scatter in the I-type
546 B.W. Chappellr Lithos 46 (1999) 535–551

Fig. 6. Harker diagrams for P2 O5 , La and Y for I-type Žfilled squares. and S-type Žopen circles. granites of the LFB. Because there are no
S-type granites with less than 63% SiO 2 , I-type granites below that SiO 2 level are not plotted. Data from the Boggy Plain Supersuite are not
included.

granites, particularly for La, as might be expected which lie above an otherwise well-defined band of
since the I-type rocks were derived from a wider compositions. These are perhaps cumulate composi-
range of source materials. However, for Y in the tions, or may represent magmas from a source en-
S-type granites, there are individually four analyses riched in Y by sedimentary processes.
 B.W. Chappellr Lithos 46 (1999) 535–551 547

For La in the I-type granites, the general gradient ture of fractionated I- and S-type granites from
is broadly slightly positive with increasing SiO 2 , the LFB.
whereas for the S-type rocks it is very slightly
negative. Above 73% SiO 2 , corresponding to frac- 6.4. Comparison of Õery felsic I-type and A-type
tional crystallization of haplogranite liquids, La de- granites
creases markedly in the S-type granites, but to a
A comparatively high abundance of the heavy
lesser degree in the I-type compositions, where some
rare earths is a characteristic of the A-type granites
rocks retain high La levels. This change in behaviour
of the LFB ŽKing et al., 1997.. The analysis of a
of La in the S-type granites was determined by the
very felsic A-type granite, AB141 in Table 1, illus-
fractional crystallization of monazite from the felsic
trates the similarities between those rocks and the
liquids, in contrast to its earlier separation as a restite
strongly fractionated I-type granites for major ele-
phase in the more mafic rocks. Monazite occurs in
ments and traces such as Y, Lu and Ga. However,
some felsic I-type granites of the LFB, and it has
unlike those I-type granites, these are primary or
been used successfully for U–Pb dating of those
close to primary features of AB141, as that rock is at
rocks ŽWilliams et al., 1983.. The lower La contents
most only slightly fractionated. Rb and Cs are much
of some I-type haplogranites show that it was possi-
lower than in rocks such as TB130 and DCC34,
bly, sometimes, a fractionating phase despite the
while Sr and particularly Ba, are higher. King et al.
very low P contents of those liquids, or else allanite
Ž1997. noted that many felsic, highly fractionated
was a fractionating phase.
I-type granites have compositions that overlap those
The different behaviour of Y in felsic I- and
of the intrinsically felsic A-type granites for some
S-type granites is, like P2 O5 , one of the most distinc-
elements. For unfractionated to mildly fractionated
tive features of those two types of haplogranites ŽFig.
I-type rocks, such as samples BB21 and TB5 in
6.. For the LFB, Y has remarkably similar abun-
Table 1, the distinction from the A-type granites,
dances in both groups up to 75% SiO 2 , and it
including ones less felsic than AB141, is very clear.
continues at the same level for the S-type rocks. For
At the other extreme, A-type granites that underwent
the I-type granites it increases dramatically, by a
strong feldspar fractionation would be difficult to
factor of around 5 times for some samples. Since
separate from similarly fractionated I-type granites.
these are liquid compositions, it means that Y is
behaving as a strongly incompatible element, imply-
ing that no Y-bearing accessory minerals were being 7. Contrasting patterns of evolution of trace ele-
fractionated. For the fractionating S-type haplogran- ments in the Freycinet and Interview River suites
ites, the bulk partition coefficient for Y was very
close to one, implying the consistent presence of an The I-type Freycinet Suite and the S-type Inter-
Y-bearing accessory mineral. view River Suite of the LFB provide examples of
Concentrations of Y, readily determined by X-ray extreme evolution of felsic granites, dominated by
spectrometry, can be used as indicators of middle to fractional crystallization. These suites are exposed on
heavy rare earth element abundances. Values for Lu the eastern and western coasts of Tasmania, respec-
are given for the 13 haplogranites in Table 1. For the tively. The Freycinet Suite comprises four plutons,
three unfractionated granites, concentrations for that Coles Bay Ž; 30 km2 ., The Hazards Ž11 km2 .,
heaviest rare earth element are close to 0.5 ppm Freycinet Ž50 km2 . and Schouten Island Ž28 km2 ., in
Ž; 13 times chondrites.. For the fractionated I-type which all rocks that have been examined are frac-
granites, the Lu abundances are significantly higher, tionated, shown by Rb contents in the range from
up to 2.6 and 3.1 ppm for the strongly fractionated 343 to 825 ppm for 21 analyzed rocks. The Interview
rocks TB130 and DCC34. In contrast, Lu in the three River Suite includes the Interview River Ž66 km2 .,
fractionated S-type granites from the LFB is substan- Sandy Cape Ž1.7 km2 . and Three Hummock Ž18.4
tially lower than the unfractionated values, in the km2 . units, with Rb varying between 275 and 950
range 0.10 to 0.14 ppm. This marked contrast ppm. For both suites, individual plutons show a
between the abundance of Lu is a general fea- restricted range of Rb contents, hence, degree of
548 B.W. Chappellr Lithos 46 (1999) 535–551

crystallization of an I-type melt. Likewise, for sam-

ples WG61 and WG57 from the S-type Interview
Suite.

7.1. Aluminium saturation in the Freycinet and Inter-

Õiew RiÕer suites

Normative C contents are plotted against Rb for

the I-type Freycinet Suite and S-type Interview River
Suite in Fig. 8. Both suites had undergone a degree
Fig. 7. Q–ab–or diagrams for the I-type Freycinet Suite and the of fractional crystallization before even the least
S-type Interview River Suite. The totals of those three components evolved compositions represented in that figure were
range from 88.9 to 95.9% for Freycinet, and from 82.6 to 92.7% produced. For Freycinet, the degree of Al-saturation
for Interview River, for which the lower totals are partly due to
some rocks being less felsic, but also to normative C contents
does not change significantly with further fractiona-
between 2.16 and 4.14%, compared with a maximum of 1.27% for tion, while for Interview River it commences at a
Freycinet. The position of the minimum temperature at 100 MPa high value Ž; 2% C. and increases dramatically to
is shown ŽTuttle and Bowen, 1958.. an even higher level Ž; 4% C.. The data for Fr-
eycinet are consistent with the continuing separation
of a slightly peraluminous assemblage of crystals,
fractionation, relative to the whole suite. With one comprising quartz q feldspars ŽASI s 1. and a small
exception ŽTB121 from Schouten Island., all of the amount of biotite, from a melt with about the same
rocks that have been analyzed fall very close to the degree of Al-saturation as the bulk crystals, so that
minimum temperature melt composition ŽTuttle and the ASI of the melt does not change systematically
Bowen, 1958. ŽFig. 7., which shows that they repre- with fractionation. For Interview River, separation of
sent liquid compositions and confirms the dominant a similar group of crystals to the above, from an
role of fractional crystallization. Water is one com- initially more peraluminous melt, leads to the melt
ponent that would be concentrated by such extreme evolving to even more peraluminous compositions.
fractionation, but it is clear that the amount of low The latter parts of that evolution can be seen in Fig.
temperature alteration of feldspars was very re- 8. Both of the distinct trends in Fig. 8 resulted from
stricted, although it is noteworthy that all granites of the continuing separation of dominantly quartz and
the Freycinet Suite are reddened. Examination of feldspars from haplogranite melts, with the S-type
later figures will show that the samples from the melt being initially slightly more peraluminous.
Freycinet Suite are compositionally less cohesive
than those from the Interview River Suite, which
may represent a degree of alteration, and perhaps, a
wider range of source compositions.
Sawka et al. Ž1990. have compared the fraction-
ated S- and I-type granites of western Tasmania, and
some of the features discussed here can also be seen
in their data. However, the Freycinet Suite has been
chosen here to compare with the those S-type gran-
ites because it is more strongly fractionated, and in
that regard, very similar to the Interview Suite.
Analyses TB5 and TB130 in Table 1 are represen-
tative of the least and most fractionated rocks from
Fig. 8. Plot of normative C against Rb for the I-type Freycinet
the Freycinet Suite, and a comparison of those two Suite Žfilled squares. and S-type Interview River Suite Žopen
compositions shows the most significant composi- circles.. Four samples from Freycinet plotted at zero C, contain up
tional changes associated with extreme fractional to 0.38% normative di.
 B.W. Chappellr Lithos 46 (1999) 535–551 549

fractionated granites of both I- and S-type ŽTable 1..

Those major and trace elements that occur in mafic
minerals, already low in any felsic melts, are quickly
purged from the evolving system. While the major
element components of alkali feldspars change very
little in abundance during prolonged fractionation,
trace elements that occur in feldspars either increase
or decrease rapidly in amount, depending on whether
their bulk rock partition coefficients are less ŽGa, Rb,
Cs. or greater ŽSr, Ba, Eu. than one Že.g., McCarthy
and Hasty, 1976.. Excluding peralkaline rocks, other
Fig. 9. Plot of P2 O5 against Rb for the I-type Freycinet Suite elements that always increase in abundance in evolv-
Žfilled squares. and S-type Interview River Suite Žopen circles.. ing felsic granites are F, Nb, Sn and U, while Zr
decreases.
7.2. EÕolution of phosphorus abundances in the Fr-
eycinet and InterÕiew RiÕer suites 7.4. Trace elements Õarying in a contrasting way in
the Freycinet and InterÕiew RiÕer suites
The general behaviour of P in the granites of the
LFB has been considered in Section 6.2, with P2 O5 The contrasting behaviour of P in the progres-
increasing in abundance in the fractionated S-type sively fractionated haplogranites of the Freycinet and
granites and decreasing in the I-type rocks. This is
shown for the specific example of the Freycinet and
Interview River suites in Fig. 9. The two suites show
a remarkably clear divergence in P abundances. For
the Freycinet Suite, P abundances start at a low level
Žmaximum of 0.06% P2 O5 down to less than 0.01%
P in some of the least fractionated rocks. and de-
crease further, so that the five rocks containing more
than 500 ppm Rb all contain 0.01% P2 O5 or less.
This is in accord with the data of Harrison and
Watson Ž1984., showing the low solubility of P in
very felsic Žand not strongly peraluminous. melts.
For Interview River, P increases in a close to linear
relationship with Rb, from 0.20% to 0.42% P2 O5 . As
noted in Section 6.2, P2 O5 concentrations in the
Interview River Suite do not exceed the maximum
solubility of P determined by Wolf and London
Ž1994., so the trend for that element does not corre-
spond to a line of saturation in P2 O5 . Rather it
represents the increasing concentration of an element
not being removed in precipitating minerals.

7.3. Trace elements Õarying in a similar way in both Fig. 10. Plots of Y and Th against Rb for the I-type Freycinet
the Freycinet and InterÕiew RiÕer suites Suite Žfilled squares. and S-type Interview River Suite Žopen
circles.. The sample plotted as a star is TB121, the sample from
the Freycinet Suite which plots away from the main group of
In contrast to P, and other elements to be dis- compositions on the Q–ab–or diagram in Fig. 6. One sample from
cussed in Section 7.4, there are many elements whose Freycinet, containing 403 ppm Rb and 122 ppm Th, is not plotted
abundances evolve in a similar fashion in strongly for Th.
550 B.W. Chappellr Lithos 46 (1999) 535–551

Interview River suites has already been discussed. Ž5. Most haplogranites of the LFB attained that
Two other elements with distinct patterns of be- character not by fractional crystallization, but rather
haviour, Y and Th, are shown in Fig. 10. For the by partial melting of quartz and feldspar in the crust.
Interview River Suite, both elements show a very In the unfractionated state, there are some overlaps
regular trend and decrease with fractionation. This in ASI values between these I- and S-type composi-
can probably be ascribed to their removal by incor- tions.
poration in precipitating phosphate phases, since Ž6. There is close to a complete separation of ASI
these melts contain abundant P ŽFig. 9.. Excluding values between fractionated I- and S-type granites,
sample TB121 Žstar in Fig. 10., which is not a with a boundary between most of the two sets of
minimum temperature melt composition and is there- data at ASI s 1.1. With increasing fractionation, I-
fore altered, there is a broad but regular increase in type melts stay close to just saturated, while the
Y with fractionation for the Freycinet Suite. A simi- S-type melts become progressively more peralumi-
lar but less consistent variation exists for Th. The nous, up to ; 4% C in the LFB.
variation in Y is consistent with the that of the Ž7. As a consequence, there is a significant con-
fractionated I-type granites of the LFB as a whole, centration of P in the S-type haplogranite melts,
shown in Fig. 6. Clearly, no significant Y- or Th- which increases with fractionation, whereas it falls to
bearing mineral was present in the mineral assem- low levels in the I-type melts.
blage that fractionated to produce these composi- Ž8. Many elements vary in a similar way during
tions. the strong fractional crystallization of both I- and
S-type haplogranites. However, some do not. The
contrast in Al-saturation, and the dramatic difference
8. Conclusions
in P abundances between the two types critically
Ž1. Source rocks of I-type granites are undersatu- determine the accessory minerals that precipitate and
rated in Al, while those of S-type granites are over- are removed during fractional crystallization. Ele-
saturated because of the loss of Na and Ca during ments that occur in phosphate minerals, such as Y,
weathering. The more mafic I- and S-type granites the rare earth elements, and Th, therefore show
have degrees of Al-saturation that directly reflect strongly contrasting patterns of evolution between
that feature in their source rocks. the highly fractionated I- and S-type granites.
Ž2. Felsic granites of all types Žhaplogranites. are Ž9. Highly fractionated granites comprise only a
close to saturated in Al, and generally slightly over- very small fraction of the components of the original
saturated. This is due to the convergence of granites source rocks, yet they have distinct properties ac-
to the minimum temperature melt composition, with cording to whether those sources were igneous or
the normative minerals Q, ab, and or present in sedimentary. The degree of Al-saturation of the melt
approximately equal amounts and together dominat- was established during the initial equilibrium be-
ing the mineral assemblage. tween melt and residual source rocks. This acted as a
Ž3. Hence, while the more mafic granites have ‘genetic code’ during prolonged fractionation of those
ASI values and associated mineralogical features that melts, resulting in distinctive end products.
are distinctly I- or S-type, the more felsic ones, when Ž10. Highly evolved I-type granites are the ex-
unfractionated, do not. treme product of progressive stages of high tempera-
Ž4. Among the I-type granites of the Lachlan belt, ture fractionation of the Earth, at least on the scale of
the subordinate Boggy Plain Supersuite provides an tens of cubic kilometers.
example of the development of felsic granites through
fractional crystallization. This process produced a
mode in the ASI values close to 1.0. Among the
S-type granites, the rocks of the Koetong Suite also Acknowledgements
evolved through fractional crystallization from a melt
containing ; 69% SiO 2 , leading to strongly peralu- Samples of the Interview River Suite were col-
minous compositions in that case. lected by Drs. Doone Wyborn and Wayne Sawka.
 B.W. Chappellr Lithos 46 (1999) 535–551 551

This paper has benefited from reviews by David Harrison, T.M., Watson, E.B., 1984. The behaviour of apatite
Manning and a second reviewer, and by Allan White during crustal anatexis: equilibrium and kinetic considerations.
Geochim. Cosmochim. Acta 48, 1468–1477.
and Penny King. This is publication number 151 in Helz, R.T., 1976. Phase relations of basalts in their melting ranges
the Australian Research Council Key Centre for the at PH 2 O s 5 kb: Part II. Melt compositions. J. Petrol. 17,
Geochemical Evolution and Metallogeny of Conti- 139–193.
nents. Hine, R., Williams, I.S., Chappell, B.W., White, A.J.R., 1978.
Contrasts between I- and S-type granitoids of the Kosciusko
Batholith. J. Geol. Soc. Aust. 25, 219–234.
Keay, S., Collins, W.J., McCulloch, M.T., 1997. A three-compo-
nent Sr–Nd isotopic mixing model for granitoid genesis,
References
Lachlan Fold Belt, eastern Australia. Geology 25, 307–310.
King, P.L., White, A.J.R., Chappell, B.W., Allen, C.M., 1997.
Bowen, N.L., 1947. Magmas Bull. Geol. Soc. Am. 58, 263–280. Characterization and origin of aluminous A-type granites from
Cawthorn, R.G., Brown, P.A., 1976. A model for the formation the Lachlan Fold Belt, southeastern Australia. J. Petrol. 38,
and crystallization of corundum-normative calc-alkaline mag- 371–391.
mas through amphibole fractionation. J. Geol. 84, 467–476. McCarthy, T.S., Hasty, R.A., 1976. Trace element distribution
Champion, D.C., Chappell, B.W., 1992. Petrogenesis of felsic patterns and their relationship to the crystallization of granitic
I-type granites: an example from northern Queensland. Trans. melts. Geochim. Cosmochim. Acta 40, 1351–1358.
R. Soc. Edinburgh, Earth Sci. 83, 115–126. Miller, C.F., 1985. Are strongly peraluminous magmas derived
Chappell, B.W., 1978. Granitoids from the Moonbi district, New from pelitic sedimentary sources? J. Geol. 93, 673–689.
England Batholith, eastern Australia. J. Geol. Soc. Aust. 25, Montel, J.M., Mouchel, R., Pichavant, M., 1988. High apatite
267–283. solubility in peraluminous melts. Terra Cognita 8, 71.
Chappell, B.W., 1996a. Magma mixing and the production of Sawka, W.N., Heizler, M.T., Kistler, R.W., Chappell, B.W., 1990.
compositional variation within granite suites: evidence from Geochemistry of highly fractionated I- and S-type granites
the granites of southeastern Australia. J. Petrol. 37, 449–470. from the tin–tungsten province of western Tasmania. In: Stein,
Chappell, B.W., 1996b. Compositional variation within granite H.J., Hannah, J.L. ŽEds.., Ore Bearing Systems; Petrogenesis
suites of the Lachlan Fold Belt: its causes and implications for and Mineralizing Processes. US Geol. Surv. Spec. Pap. 246,
the physical state of granite magma. Trans. R. Soc. Edinburgh, 161-179.
Earth Sci. 87, 159–170. Shand, S.J., 1927. Eruptive Rocks. Murby, London.
Chappell, B.W., White, A.J.R., 1974. Two contrasting granite Tuttle, O.F., Bowen, N.L., 1958. Origin of granite in the light of
types. Pacific Geol. 8, 173–174. experimental studies in the system NaAlSi 3 O 8 –KAlSi 3 O 8 –
Chappell, B.W., White, A.J.R., 1984. I- and S-type granites in the SiO 2 –H 2 O. Geol. Soc. Am. Mem. 74.
Lachlan Fold Belt, southeastern Australia. In: Keqin, X., Wall, V.J., Clemens, J.D., Clarke, D.B., 1987. Models for grani-
Guangchi, T. ŽEds.., Geology of Granites and Their Metallo- toid evolution and source compositions. J. Geol. 95, 731–749.
genic Relations. Science Press, Beijing, pp. 87–101. Williams, I.S., Compston, W., Chappell, B.W., 1983. Zircon and
Chappell, B.W., White, A.J.R., 1992. I- and S-type granites in the monazite U–Pb systems and the histories of I-type magmas
Lachlan Fold Belt. Trans. R. Soc. Edinburgh, Earth Sci. 83, Berridale Batholith, Australia. J. Petrol. 24, 76–97.
1–26. Wolf, M.B., London, D., 1994. Apatite dissolution into peralumi-
Chappell, B.W., White, A.J.R., Wyborn, D., 1987. The impor- nous haplogranitic melts: an experimental study of solubilities
tance of residual source material Žrestite. in granite petrogene- and mechanisms. Geochim. Cosmochim. Acta 56, 4127–4145.
sis. J. Petrol. 28, 1111–1138. Wyborn, D., 1983. Fractionation processes in the Boggy Plain
Clarke, D.B., 1981. Peraluminous granites. Can. Miner. 19, 1–2. zoned pluton. PhD Thesis, Australian National University
Collins, W.J., 1996. Lachlan Fold Belt granitoids: products of Žunpublished..
three-component mixing. Trans. R. Soc. Edinburgh, Earth Sci. Wyborn, D., Turner, B.S., Chappell, B.W., 1987. The Boggy
87, 171–181. Plain Supersuite: a distinctive belt of I-type igneous rocks of
Ellis, D.J., Thompson, A.B., 1986. Subsolidus and partial melting potential economic significance in the Lachlan Fold Belt.
reactions in the quartz-excess CaOqMgOqAl 2 O 3 qSiO 2 q Aust. J. Earth Sci. 34, 21–43.
H 2 O system under water-excess and water-deficient condi- Zen, E., 1986. Aluminum enrichment in silicate melts by frac-
tions to 10 kb: some implications for the origin of peralumi- tional crystallization: some mineralogic and petrographic con-
nous melts from mafic rocks. J. Petrol. 27, 91–121. straints. J. Petrol. 27, 1095–1117.