Download Crystallization sequences in the Muskox intrusion and other layered

Ceoohimicaet CosmochimicaActa, 1975,Vol. 39, pp. 991 to 1020. PergamonPress. Printedin NorthernIreland
Crystallization sequences in the Muskox intrusion and other layered
intrusions-II.
Origin of chromitite layers and similar
deposits of other magmatic ores
T. N. IRVINE
Geophysical
(Received
Laboratory,
Carnegie Institution
of Washington,
Washington,
D.C. 20008, U.S.A.
20 July
1974; accepted
in. revised form
30 September
1974)
Abstract-A
mechanism
of origin for chromite-rich
layers in stratiform
ultramafic-gabbroic
intrusions is proposed whereby the layers are precipitated
on occasions when the basic parental
magma of the intrusion is suddenly extensively
contaminated
with granitic liquid melted from
salic roof rocks. It is inferred that the increase of silica and alkalies in the basic liquid should
cause it to become more polymerized
with a lower frequency
of octahedral sites, so that on
continued
crystallization,
Crs+ is preferentially
expelled
(into chromite)
owing to its large
The feasibility of this process is demonstrated
by
octahedral crystal-field
stabilization
energy.
experimental
data on forsterite-picrochromite
crystallization
relations in the system K,O-&IgOCrsOs-SiO,, and its apparent applicability
to magmas is illustrated through a comparison of the
differentiation
patterns of Cr and Wi in the Muskox intrusion.
The granitic melt is produced
because most of the crystals formed in the intrusion accumulate on its floor, leaving the roof
rocks to be continuously exposed to the high temperature of the basic magma. Between episodes
of contamination,
the melt tends to accumulate on top of the basic magma and to remain separate
If not assimilated, it eventually
resolidifies
as
because of its low density and high viscosity.
granophyre.
In the Muskox intrusion there are two chromite-rich layers, each occurring in a stratigraphic
unit showing the layer sequence, peridotite-chromitite-orthopyroxenite.
This sequence is
explained in a model in which the basic magma is contaminated
while coprecipitating
olivine
and minor chromite.
A period follows when chromite precipitates alone, and then, because the
liquid is enriched in silica, orthopyroxene
crystallizes instead of olivine. Variations on the model
are described that simulate the main layer sequences involving
chromitite
in the Stillwater,
Great Dyke and Bushveld intrusions.
Evidence of contamination
is found in the concentrated
chromite crystals in the form of small spherical, composite silicate inclusions, rich in alkalies,
apparently representing trapped droplets of the contaminant
granitic melt in various stages of
assimilation.
It is suggested that the same type of contamination
mechanism may also yield
concentrated
deposits of magnetit,e and of immiscible sulphide liquid.
INTRODIJCTION
ONE OB the most fascinating occurrences of the mineral chromite is as thin concentrated layers in stratiform ultramafic-gabbroic
intrusions. These layers are especially
common in the Bushveld Complex and Great Dyke in southern Africa and in the
Stillwater Complex in Montana, bodies in which dozens of examples ranging from
a few inches to about 15 ft in thickness have been traced for distances of tens of
miles.* The layers are of interest both as major ore deposits and as a remarkable
phenomenon suggestive of important igneous processes.
* In the Stillwater Complex, there are several zones of chromitite layers ranging from a few
inches to about 15 ft in thickness that have been traced for 15 miles and one that extends for
almost 30 miles (JACKSON, 1963). In the Bushveld Complex, the Leader and Steelpoort chromite
layers or ‘seams,’ which are respectively
1 and 4 ft thick, have been traced together with several
thinner seams for more than 40 miles (CAMERON and DESBOROUGH, 1969), and they are roughly
991
992
T. N.
IRVINE
Itis generally agreed that the layers are deposits of chromite crystals settled from
magma, and some occurrences show strong evidence of having accumulated under
the influence of currents (e.g. CAMERON and DESBOROUGH, 1969). There have been
various suggestions as to the mechanism of chromite enrichment--that
it was concentrated by current sorting or that it was precipitated preferentially in response to
changes of pressure, water content, or oxygen fugacity in the magma-but
none of
the advocated processes would appear to explain the variety of features and associations shown by chromitite layers, and none has been developed in terms of the
overall crystallization history of an intrusion.
In the present paper an attempt is made to outline and substantiate a mechanism
of origin for chromitite layers based mainly on two occurrences in the Muskox
intrusion in the CanadianNorthwest Territories. The Muskox chromite-rich layers
are unimposing in comparison with many of the occurrences referred to above. One
is everywhere less than an inch thick; the other at its best consists of only a 4-in.
unit of concentrated chromite in a total zone of chromite enrichment of about 1 ft.
These layers, however, have been traced in outcrop for about 12 miles, and from
drill-hole intersections it is apparent that their area1 extent is at least 40 square
miles. They are similar to the Bushveld, Great Dyke, and Stillwater chromitite layers
in various details, and perhaps most important, they are contained in an intrusion
of convenient size that is well preserved structurally, fully exposed from floor to
roof in cross-section, and remarkably systematic in its differentiation.
The postulated
origin for the chromite-rich layers is that they precipitated on occasions when their
parental magma deviated from its normal course of crystallization owing to extensive
contamination by granitic melt derived from the roof of the intrusion. The paper is,
in some respects, a sequel to a previous contribution dealing with the crystallization
relations of olivine, pyroxene and plagioclase in the Muskox and other layered
intrusions (IRVINE,1970a).
GEOLOGY OF THE MUSKOX INTRUSION
The Muskox intrusion was mapped and first described by SMITE (1962) and has been the
subject of numerous subsequent publications
(see IRVINE and BARAUAR, 1972). As exposed it is
a north-northwesterly
trending body, about 74 miles long, crossed in the middle by the Coppermino River (Fig. 1). South of the river it appears as a vertical dyke, 500-1700 ft wide, that
(footnote cont’d)
correlative
with the ‘Main chromite seam,’ which extends discontinuously
many times as far.
Lesser concentrations
of chromite in layers only an inch or so in thickness are ubiquitously
associated with the platiniferous
Merensky reef, which has been traced for about 80 miles in the
eastern part of the complex and 120 miles in the western part (cf. WAGER and BROWN, 1968).
In the Great Dyke, WORST (1960)
has distinguished 31 chromitite layers, ranging from 1 to 18 in.
in thickness, extending along large segments of the 330-mile length of the body.
One example,
about 10 in. thick, was traced for 73 miles; another with an almost constant 4-in. thickness wa,y
followed for 55 miles up one side of the dyke and for almost as far back down the other. The
Stillwater and Great Dyke chromitite layers occur mainly with layers of peridotite or dunite in
more or less systematic succession with layers of orthopyroxenite.
The Bushveld chromitite
is
principally
associated with orthopyroxenite
and anorthosite
in sections of very complicated
stratigraphy.
The BushveldComplex
also contains layers of magnetite, comparable in appearance
and extent to the chromitite
layersand similarly inter&ratified
with anorthositic
rooks, that
would seem to require a similar explanation.
Crystallization
sequences
in the Muskox
intrusion
and other layered
intrusions-II
993
LEGEND
Diabose dykes
not shown
and basic
Coppermine
~111s
River
bcsalr
Dolomite
Granophyre
Gronophyre-rich
Two-pyroxene
;
/m
gabbros
Oilvine
gabbros
Picrtt!c
websten!e
Webaterite,
OIlvine
1 I-
qobbros
orthopyroxenite
clinopyroxenlte
[x
Gwnit~c
[la
Melavolconic
rocks
m
Metosedimentory
rocks
rocks
Geologic boundary
(defined.
showtng dip. opproxvnote;
assumed)
Fault (defined: assumed)
”
‘.’
,’
..-..r,ti
/
t 2
0
3 4 Smiler
t---a--+
:.n
.- A-2 ;i: i:,
I
0
4
6 kIlometers
________
._ -_._
Fig.
1. Generalized
geological map of the Muskox intrusion,
of the main diamond drill holes.
showing
the locations
apparently
projects beneath the main body of the intrusion to the north and so is called the
feeder dyke. The main body, which is estimated to have been emplaced at a depth of less than
5000 ft (IRVINEand BARAGAR, 19’72, p. IO), is funnel-shaped
in cross-section with lower ~valls
dipping inward, generally at 25-35°, and roof inclined gently to the north.
It has been tilted
about 5* to the north and consequently
has eroded so that its deepest levels are exposed in the
south and its outcrop width gradually increases northward
(reaching a maximum of about 7.5
miles where it plunges beneath its roof). Further north, the intrusion can be traoed beneath its
roof rocks and younger cover for at least another 20 miles on the basis of an neromagnetic
anomaly, which then merges with a major gravity anomaly that continues nort~l~~~esterly for
about 150 miles. The exposed rooks, therefore, appear to represent only the southern extremity
of a much larger plutonic complex.
The feeder dyke is composed mainly of bronzite gabbro but along most sections of its length
contains either one or two parallel internal zones of picrite.
The gabbro is locally chilled along
T. N. IRVINE
994
the dyke walls, and its composition, which is equivalent to silica-saturated tholeiitic basalt,
appears to be approximately representative of the parental magma of the intrusion (IRVINE,
1970a).
The main body of the intrusion comprises two marginal zones, a layered series, and a granophyrio roof zone. The marginal zones line the inward-dipping footwall contacts and are generally
400-700 ft thick. They grade inward (or upward) from bronzite gabbro at the contact through
pierite and feldspathic peridot&s to peridotite.
The layered series consists of 42 layers of 18 different rook types and has been divided into
26 cyuhc units, these being repeated stratigraphic divisions characterized by specific litholo~o~
sequences or chemical trends (Figs. 2-5). The layers range in thickness from IO to 1100 ft,
:YcLIc
UNITS
. . . . ..~.WESSTERITE
-~.....ORTHOPYROXENITE
PERIDOTITE
~~~--clLI”INE
CL~NOPYROXENITE
C. . . . . .TROCT&lTIC
,........,........
PER,DOT,TE
OL,V,NE
GABBRO
QL,“,NE
CLINOPYROXENITE
_____.................
. ...
. . . . . .. . .
_,..,..._..._.................
MARGINAL
. . . ..PEAIDOTITE
7
OLlVlNE
.. BRONZiiE
BRCNZIiE
GA9BRO
GABBRO
CRYsTALLiNE
j ZoNE
J
BASEMENT
COMPLEX
Fig. 2. Drill-hole sections of the Muskox intrusion, showing the main cyclic units
and the two chromite-rich layers. mt, magnetite zone; il-mt, ilmenite-magnetite
zone, Diabase dykes have been omitted for clarity.
totaling about 6000 ft. They extend between the marginal zones and dip gently to the north (at
about 5”). Some have been traced in outcrop for as far as 15 miles, and it is evident from drillhole intersections that many of them have an area1 extent in excess of 100 square miles.
The rocks of the layered series are entirely cumulates, and the series ranges generally from
dunite at the base through various pyroxenites and peridotite to two-pyroxene gabbro at the top,
The oyclic units fall into three general classes, distinguished in principle by different sucoessions
of rock layers. These sequences developed because the cumulus minerals precipitated in close
accordance with the crystallization orders of their parental liquids and so represent phase layeting
(cf. HESS, 1960). The repetition of units reflects the repeated influx of relatively undifferentiated
‘new’ liquid into the intrusion (IRVINE and SMITH, 1967; IRYINE, 197Oa).
The characteristic rocksuccessions and corresponding crystallization orders for the main types
of cyclic units are summarized in Table 1. The order in which the olasses are listed is essentially
their order of occurrence in the layered series; hence the differences between them arise because
Crystallization sequences in the Muskox intrusion and other layered intrusions-II
995
Table 1. Rock layer sequences and crystallization orders characteristic of the principal cyclic
units in the Muskox intrusion.
Crystallization order
Rock sequence
Class
1
Dunite; olivine clinopyroxenite;
olivine gebbro
01; epx; pl; opx
II
Dunite; olivine clinopyroxenite
websterite
01; cpx; opx; pl
III
Peridotite; orthopyroxenite;
two-pyroxene gabbro
01; opx; cpx; pl
Abbreviations:
01, olivine; cpx, dinopyroxene;
LAYER
NO.
ROCK
TYPE
websterite;
opx, orthopyroxene;
PPM
PPH
Ni
Cr
MODlL
pl, plagioclase.
*
Chromite
CYCLlC
UNIT
Fig, 3. Data from Muskox drill-hole 18618, showing the stratigraphic position of
the ohromite-rich layers in their xespeotive cyclic units, together with data on the
abundances of I%, Cr and chromite in the rocks.
T. N. IRVINE
PPM
ROCK
L*YER
NO.
Cr
VJ
ii
nL %
Chromite
s
91
G
8’
CL
?
$5
).
. .
5.
.
,+
5
Fig. 4. Emission spectrographic and modal data from part of the Muskox South
drill hole, showing cyclic variations in the abundances of Ni, Cr and ohromite.
orthopyroxene periodically advanced in the orystallization order of the magma. The advance
appears to have occurred because the magma became increasingly contaminated with salic
material melted from the intrusion roof rocks (IRVINE, 1970a).
The diagnostic chemical trends in the cyclic units are upward trends of decrease in MgO/FeO
ratio (IRVINE and BAIIAOAR, 1972, Fig. 9) and Ni content (Figs. 3 and 4). These are essentially
fractional orystallization trends for the cumulus mafic minerals (i.e. they represent cryptic
layer&g; cf. WAUER and BROFVN,1968), but they show also in the whole-rock analyses.
The periodio infusions of ‘new’ liquid indicated by the repetition of cyclio units were apparently
introduced laterally by flow between the accumulating layers and the roof contact. (The feeder
dyke does not out the layered series, so the magma could not have come from directly below.)
The chemical trends in the cyclic units provide evidence that some of the infusions were very
large, amounting to practically the whole volume of the exposed part of the intrusion, so it is
presumed that the magma originated from a major feeder system or reservoir in the area of the
large gravity anomaly to the north and moved southward, pushing the residual, ‘old’ liquid
ahead of it and eventually to the surface as volcanic eruptions (IRVINE and SETH, 1967;IRVINE
and BARAOAR, 1972, pp. 21-22).
The granophyric roof zone is an irregular unit up to about 1000 ft in thickness, consisting of
granophyre and granophyrio gabbro transitional with the top of the layered series. The granophyre aonsists mainly of a micrographic intergrowth of quartz and K-feldspar with variable
minor amounts of biotite and ihnenite. Evidence described in a later se&ion indicates that it
was largely crystallized from roof-rock melt.
997
Crystallization sequences in the Muskox intrusion and other layered intrusions-II
ROCK
TYPE
LAYER
NO.
Cr,O, in chromite,
wt % CYCLIC
UNIT
OLIVINE
FABER0
OLIVINE
CLINOPX
_____-
l_-.y.
,_ .
. k-e
.. _
-.\
..:,
. ,. .
‘-
d’- *..:’1,
_.
__&
z
r
___a_
__-
--
..
.-_
.”
.
1
L
.-.
-.‘r
.:
“.
.
..z
I
.
,
-I.
../
.
.
.
-..
_.
.:
‘.‘“..
J..
:j. .
y:
-i;
DUNITE
._A
”
-.
_ .
\ .:-
A.
I$.I
.
-
F-“‘-------_
$1 -*.
_
A
-;j.
._..
l. ..:....--*
.
P
3L
CLINOPX
._.
---
.
.
1
,: 1-L-_-
.I_.
*. . .
-.
.-.Z..
/.
.
.*
.I.
.
*-_
_;3pLl_L
Fig. 5. Electron microprobe data for the same part of the Muskox South drill hole
as in Fig. 4, showing the concentrations of Ni in olivine and Ni and Cr,Os in
chromite.
CHROMI~E IN THE MUSEOX INTRUSION
The distribution and crystallization relations of chromite in the intrusion have
been described by IRVINE (1967) and IRVINE and SMITH (1969), so only critical
features are reviewed here.
Except for the two concentrated layers, almost all the chromite in the intrusion
is disseminated in amounts of l-3 per cent in olivine cumulates that are variously
classed as dunite, peridotite, feldspathic peridotite or picrite, depending on their
content of postcumulus materials. These rocks make up about two thirds of the
cross-section of the intrusion and in the layered series form 22 layers with an aggregate
thickness of about 4000 ft (Fig. 2). They contain chromite almost throughout.
Most of the chromite occurs as subhedral crystals, 0.5-0.15 mm in diameter,
situated either individually or in small clusters between the larger cumulus grains of
olivine. Occasional small euhedral crystals are trapped in the olivine as early formed
inclusions.
These features, together with the observation that the modal ratio of
chromite to olivine is approximately
that in which picrochromite
and forsterite
coprecipitate in the system MgO-Cr,Oa-SiO,
(KEITH, 1964), suggest that the two
998
T. N.
IRVINE
minerals precipita~d simultaneously by fractional crystallization, an inte~retation
that is confirmed by other data described below. The chromite in the concentrated
layers is similar to the disseminated material but commonly is recrystallized to
coarser
grain sizes where the crystals ac~urnul~~d in close contact with one another.
The layers also contain small but conspicuous concentrations of sulphids minerals,
principally pyrrhotite, chalcopyrite and pentlandite, occurring in part as distinctive
ellipsoidal globules, l-10 mm in length, that would appear to represent accumulations
of immiscible sulphide liquid.
Other main features of the distribution of chromite relate to the cyclic units and
are illustrated by the data in Figs. 3 and 4. Three points are noted:
(1) In Fig. 4, the modal abundance of the chromite disseminated in dunite shows
a cyclic variation similar to that of Ni, particularly in units 5 and 6. Considering
that the Ni variation is essentially due to fractional crystallization of the olivine in
the dunite, it is apparent that the chromite must have coprecipitated with the olivine.
(2) The two chromite-rich layers (Fig. 3) are both situated between layers of
peridotite and orthopyroxeni~ and have a definite place in the cyclic repetition. In
the units in which they occur the rock sequence is peridotite-chromitite-orthopyroxenite-websterite.
(3) Chromite tends to be absent in the rocks with cumulus pyroxene. Thus in
several cyclic units in Figs. 3 and 4, its modal abundance abruptly drops practically
to zero in passing upward from peridotite to orthopyroxenite or from dunite to
olivine clinopyroxenite, even though the amount of chromium in the rocks shows a
relatively smooth transition across the contact. The ~scontin~ties are att~butgd to
magmatic reaction relations between chromite and the pyroxenes such that, when
pyroxene began to crystallize, it could accommodate all the chromium the liquid
could supply, so chromite stopped forming.
These obse~ations can be further explained by the schematic phase diagram in
Fig. 6, illustrating a composite fractional crystallization path for the Muskox liquid.
The key features of the diagram are (1) the curved cotectic boundary between the
olivine and chromite liquidus fields; (2) the reaction or dist,ribution point b at the
~tersection of the olivine, orthop~oxene, and chromite fields; and (3) the very low
concentrations of Cr,O, in the liquid as compared with the pyroxene and chromite.
The curvature of the cotectic causes the modal ratio in which olivine and chromite
coprecipitate to gradually decrease, ostensibly from 4: 96 to 1: 99 as in Muskox cyclic
units 5 and 6 (Fig. 4). When the liquid reaches the dist~bution point, the combination of reaction boundaries requires that both minerals stop forming as pyroxenc
begins to precipitate, as apparently happened, for example, in Muskox cyclic unit 19
(Fig. 3). And the phase relations show how the condition that the liquid contains
less Cr,O, than the pyroxene can lead to a chromite + pyroxene reaction relationship.
The value of 0.08 wt. y0 Cr,O, indicated for the initial Muskox liquid comes from
analyses of the chilled margin. The value of 0.03 per cent for the point st which
pyroxene begins to crystallize was obtained by subtraet~g the amount of Cr,O, t-hat
would have been removed in the ohvine-chromite cumulates fractionated from the
liquid by that stage (see IRMNE and SMITH, 1969). These data are tenuous in
themselves but are compatible with analyses reported for chromite-saturated olivine
tholeii~s of overall composition similar to the chilled margin (e.g. YODEE and TILLEY,
1962, TabIe 2; EVANSand WRIGHT, 1972, Table 2).
Crystallization sequences in the Mu&ox
intrusion and other layered
intrusions--l1 N9
Chramites,(-4O%Cr203)
b-c
o-b
CUMULATE
Minerals
Orthopyroxen~
Olivine -(Chromite)
SEQUENCE
Rocks
Oriho~yrox~nite
Dunite, peridotiie
Fig. 6. Schematic projeation illustrating the apparent phase relations of olivine,
orthopyroxene and chromite in the Muskox intrusion as indicated by petrographic
data. The inset triangle shows the plane of projection, which is part of the more
extensive join (Mg, lYe)O-Cr,Os-SiO,. Liquidus relations are modeled after those in
the system MgO-Cr,O,--SiO, (KEITH, 1954) but have been drawn to simulate the
Mu&ox data, and the main diagram has been distorted for clarity. Boundary
curves with double arrows are reaction boundaries.
Two further points directly pertinent to the origin of &omit&e layers can also
be illustrated by means of Fig. 6. The first, a matter of principle, is that fraotional
crystallization by it.seZfcannot yield a concentrated deposit of chromite once olivine
or pyroxene {or any other silicate mineraI) has begun to form (cf. IRVIZNE and SXITH,
1969, p. 93). Thus an explanation of the situation of the Muskox chromite-rich
layers in the middle of cyclic units above layers of peridotite requires some additional
process. The other point, a feature that was first suspected on the basis of the model,
is that addition of siliceous material to liquids on the olivine-chromite cotectic should
tend to shift their compositions into the chromite primary phase field. Figure 6 is
not definitive in this regard inasmuch as it is only a schematic projection, and in fact
even the system BlgO-Cr,O,SiO, (KIWII, 1954) does not show any clear ~dication
of the suggested effect of silica. There is, however, a theoretical line of reasoning
concerning the partitioning of Cr and Ni in silicate melts that leads to the same
prediction, and the effect has been substantiated by an experimental study described
below.
FRACTIONATION OB NICKEL AND
CXEOMIUM IN SILICATE MELTS
The data in Fig. 4 show very similar whole-rock differentiation patterns for nickel
and chromium. (Nickel shows the better developed trends of depletion within cyelic
15
T. N. IRVINE
1000
There is an important
units, but the trends of chromium are obviously comparable.)
difference, however. In the dunite layers, the nickel was almost entirely precipitated
in olivine, which originally made up about 90 per cent of the rock. The olivine is now
completely serpentinized in the upper two units, but in the lower units its Ni variation
chromite, which
closely parallels that of the whole rock; and the coprecipitated
survived the serpentinization without apparent change in composition, exhibits the
same type of Ni variation in all four units (Fig. 5). By contrast, the chromium in the
dunite is almost entirely contained in chromite of constant Cr,O, content (Fig. 5) ;
therefore, its whole-rock variations reflect the modal variations of this mineral (Fig.
4). Thus the difference is that nickel was precipitated as a trace element in a major
mineral (olivine), whereas the chromium was precipitated as a major element in an
accessory or trace mineral (chromite).
Why, then, the similarity of differentiation pattern?
The one thing the two
elements do have in common petrologically is that they have originated from the
same batches of liquid. It is suggested, therefore, that the similarity is primarily a
reflection of the properties of the liquid. In particular, it seems likely that because
both elements have exceptionally
large octahedral site preference energies due to
crystal-field stabilization (cf. BURNS, 1970, Table 6.2), they probably came from much
the same type of site in the liquid. Following an argument used by BURNS and FYFE
(1964, 1967) in discussion of Ni partitioning in magmas, it is suggested that with the
increased polymerization
of the liquid relating to the joint enrichment of silica and
alkalies and the decrease of liquidus temperature caused by fractional crystallization
of the olivine and chromite, there was a marked reduction in the number of octahedral
Thus the remainsites in the liquid that were energetically favorable for occupancy.
ing Cr3+ and Ni2+ in the liquid, along with other octahedrally coordinated ions such
as Mg2+ and Fea+, were required more and more frequently, through the following
combination of equilibria, either to enter tetrahedral sites or to transfer to octahedral
sites in the crystals.
(M~+odahedrsi)~iqni
\\
(Mnfootnhedral)~ystsla,
!l
//
(Mn+~t.tetrahadral)Liqnid
where M”+ = Cr3+, Nisf, Mg2+, Fez+.
Because of their very large octahedral site preference energies, the partitioning
of Cr3+ and Ni2+ would be strongly biased in favor of the crystals;
hence their
equilibria were increasingly shifted to the right, with the effect that they were
preferentially expelled from the liquid at similar rates. The Ni2+ ions substituted for
Mg2+ and Fe2+ in the minerals, whereas the Cr3+ ions effectively controlled the precipitation of the chromite.
The postulated shift of the nickel equilibria is representative
of an effect
that apparently is large enough to reverse the preferred direction of nickel partitioning between olivine and liquid for basaltic liquids as compared with olivine
melts (cf. BURNS and FYFE, 1964, 1967). The shift should be most closely reflected
in a coefficient in which the partitioning
is defined relative to magnesium,
Crystallization
sequences
in the Muskox
intrusion
and other layered
intrusions-II
1001
K = (Ni/Mg) oliviDe/(Ni/Mg)Liquid,
inasmuch as Mg 2+ is similar in size to NP+ but,
not being a transition metal ion, shows no crystal-field stabilization. The shift,
however, is also represented in the distribution ratio, D = (Ni in olivine)/(Ni in
liquid), and it is expedient for present purposes to consider the partitioning in this
form. Figure 7 illustrates data on the variation of D for tholeiitic liquids similar to
60 8
6
I
1 , 1 , 1 , 1 ,
(u
.5
.z
0
E 20
e
G
CL
i
.$
IO
3
(Ni
in olivine)/(Ni
in initial
liquid)
Fig. 7. Variation
curves for Ni in olivine formed by fractional
crystallization,
computed for the situation that the distribution
coefficient,
D, varies with the
amount of normative
olivine in the liquid as shown in the inset diagram.
The
value of D for the systemMgaSiO,-NisSiO,
is estimated fromthe data of RINGWOOD
(1956); the value for Kilauea tholeiite is from H~~LI and WRIOHT (1967). The
other data points are from olivine orystallized experimentally
from melted Kilauea
lava.9 spiked with small amounts of NiO.
The experiments
were run in a gasmixing furnace at oxygen fugacities
approximately
equivalent
to the quartzfayalite-magnetite
buffer.
Products
were quenched and analyzed
by electron
microprobe.
Negative amounts of normative olivine are computed equivalents of
normative
quartz.
The curve drawn through the data points has the equation,
D = -0.91 + 2.969 x 10-3(100 -X0,)
+ 3.248 x 104(100 - XoJ2, where X0,
is the percentage of normative olivine.
Compare with the Ni data in Fig. 5.
the Muskox magma and shows the kinds of fractional crystallization trends that
should develop in olivine as a result of this variation. The tendency of the fractionation curves to be convex upward with respect to Ni content is diagnostic. Comparison with Fig. 4 reveals that the olivine in Muskox cyclic units 4 and 5 does in fact
show this type of Ni variation* (as does the chromite in cyclic units Pi’); hence it
* It will be noted that the olivine in cyclic units 4 and 5 shows the type of Ni variation that
would be expected if it had been fractionated
from liquid initially containing about 15 per cent
normative
olivine.
The data on which the fractionation
curves are based are still provisional,
so no firm conclusion can be drawn in this regard, but it may be noted that, although the Muskox
chilled margin is just saturated in silica, the intrusion may well at times have contained more
primitive liquid with IO-15 per cent normative
olivine.
1002
T. N. &VINE
appears that the inferred shift of equilibria did indeed occur in the Muskox
magma.
In the case of chromium, the shift of equilibria should bear on the proportion of
chromite precipitated and, therefore, should be reflected in the trend of the olivinechromite cotectic. The trend should be such that the addition of constituents that
act to polymerize silicate melts, such as silica and alkalies, should tend to shift the
liquid compositions from the cotectic into the chromite liquidus field. As noted
above, this is the same prediction that arose from the phase diagram model in Fig. 6.
It prompted the experimental study that will now be described.
FORSTERITE-PICROCHROMITE CRYSTALLIZATION ON THE JOIN
MgO.SiO,-Cr,O,-K,O*6SiO,
The specific objective of this investigation was to test whether compositions in the
forsterite liquidus field on the join MgO*SiO,(enstatite)-Cr,O, could be transposed
to the picrochromite field by the join addition of SiOz and K,O. The experiments
were performed by 1-atm quenching methods. Starting materials were prepared
from weighed quantities of dried MgO, cristobalite, Cr,O,, and a powdered crystalline
aggregate of K,O*bSiO, composition. These were thoroughly mixed, fused at
temperatures 30-100°C above their liquidus, and cooled quickly. The experiments
were run at temperatures controlled to f2”C, usually for 1 hr. Run products were
identified optically, and in selected cases were analyzed by electron microprobe.
A prime consideration in choosing the system was to avoid solid solution in the
crystalline phases in order to concentrate on the effects relating to changes in liquid
composition. This goal was evidently satisfied, because the forsterite showed a
maximum of only 0.5 wt.% Cr203, and the picrochromite, only traces of silica. The
join enstatite-Cr,O, was selected as a starting line because it has forsterite and
picrochromite on the liquidus at temperatures low enough to be reached conveniently.
The composition K,O*6SiO, was used because it is rich in silica and yet can be mixed
with enstatite in amounts up to almost 70 wt,% and still have compositions in
the forsterite liquidus field (Fig. 8). From the latter feature, a large working range
in which to examine the forsterite-picrochromite cotectic was anticipated. It should
be noted t,hat the investigated join is not ternary and can show only a line of
intersection or trace of the cotectic surface as developed in the quaternary system
K,O-MgO-Cr,O,-SiO,.
The experimental results (Fig. 9) show that the addition of K,O.6SiO, will
indeed transpose particular compositions from the forsterite liquidus to the picrochromite field, and the phase diagram prepared from the data indicates that it should
shift any composition on the cotectic trace into the spine1 field.* The system is
* It should be noted that this result is not incompatible with the earlier observation that the
addition of silica alone to liquids on the forsterito-picrochromite
cotectic in the system
MgO-Cr,O,-SiO,
(KEITH, 1954) does not apparently shift their compositions into the picrochromite field. In fact, there is a possibility within the framework of existing data that tho trend of
Cr variation in liquids on the cotectic even reverses as their normative olivine content and
temperature decrease in much the same way as the direction of Ni partitioning reverses in Fig. 7,
the trend initially being toward increased concentrations of Cr,03 before turning toward lower
concentrations as in Fig. 9.
Crystallization sequencesin the Muskox intrusion and other layered intrusions--II
1003
W
(Mgo.C_r20.)
Fig. 8. Liquidus relationsonthe joinKzO-2MgO-SiO,SiO,
(after ROEDDER,1951)
superimposedon relations for the join 2MgO*SiO,-MgOX$Os-SiO,
(after KEITH,
1954, shown by light da.shed lines with phase names in parentheses). Liquidus
data for composition on the linefrom point a (located at 99 % MgO*SiO,, 1% Cr,O,f
to IC,O*GSiO,are shown in Fig. 9. (Note: K,O.GSiO, is not a compound.)
extremely simple compositionally in comparison with magmas, but the observed
effect may be expected to have broader significance inasmuch as it is the analogue of
an effect of crystal field stabilization that apparently is extremely important in
determining the behavior of nickel in magmas, and because Cra+ shows an even
stronger preference for octahedral sites owing to crystal-field stabilization than
NP+ (cf. BURNS, 1970,Table 6.2). The effect is fundamental to the mechanism of
origin for chromitite layers that will now be described.
FORMATIONOF CIE~OMITITE
LAYERS
A model of the chemical aspects of the me~ha~sm as applied to the Muskox
layers is illustrated in Fig. 10. It is seen that the fractionation path of a liquid
initially crystallizing olivine and minor cbromite can be altered through contamination by siliceous material so that the liquid subsequently precipitates first chromite
and then orthopyroxene, giving the rock sequence peridotite-chromititorthopyroxenite observed in the lower parts of the cyclic units ~onta~g
the chromiterich layers (Fig. 3). (The succeeding websterite layers could be simulated by including CaO as a component so that two pyroxenes could precipitate; IRVINEand SMITH,
1967.) It is noted that the model also accords with the interpretation that contamination was responsible for the periodic advances of orthopyroxene in the crystallization orders in the three classes of cyclic units listed in Table 1.
Figure 1 I shows variations on the model that simulate the main layer sequences
involving ohromitite in the Stillwater, Great Dyke and Bushveld eomplexes. The
succession of cumulates produced in A is essentially that in the type Stillwater cyclic
unit (JACKSON,1961, I’ig. 11). The sequence in B. either as a whole or in parts,
T. N. IRVINE
1004
Kfl6SiOp
KzO
1300
,
“2’3
./3’.
MgO.Si02
,’
._”
Weight
percent
K20.6Si0i
Fig. 9. Phase relations for part of the join MgOSiOs-CrsOs-K,O*BSiO, at 1 atm.
Boundary lines on the right face (0% CrsO,) are from ROEDDER (1951); the
liquidus on the join MgO*SiO,-CrsO, is consistent with data of KEITH (1954). It
is seen that compositions in the forsterite liquidus field can be transposed to the
picrochromitefield by ‘contamination’ with K,O+SiO,.
Compare with Figs. 6, 10
and 11. Abbreviations: Fo, forsterite; Pr, protoenstatite; PC, picrochromite; L,
liquid; ss, solid solution.
matches various sections of Great Dyke stratigraphy (cf. WORST, 1960, Plate ll),
especially the unit described in detail by BICHAN (1969). And the scheme in C,
coupled with some means for periodically replenishing the liquid, would give an
alternation of chromitite and pyroxenite as observed in the lower part of the Bushveld Complex (e.g. CAMERON and DESBOROUGH, 1969).* It is not possible at this
* In the model in Fig. 1Ic, chromite is shown to be succeededby orthopyroxenein the crystallization history of the Bushveld magma. It should be noted in this regard that, although it has
been amply demonstrated that chromite and orthopyroxene frequently accumulated together
in the Bushveld complex (e.g. CAMERONand DESBOROUQH,
1969), it has not been shown that
they crystallized together.
Crystallization sequences in the Muskox intrusion and other layered intrusions-II
Liquid”~
original
boundaries
magma
CUMUL’ATE
Minerals
of
,SEQUENCE
Rocks
Orlhopyroxene
Orthopyrdxenite
c-d
Chromite
Chromitite
a-b
Olivine-(Chromite)
Peridotite
d-e
Fig. 10. Phase diagram model ilIustrat~g the formation of the lower parts of the
Muskox cyclia units containing chromite-rich layers by a oombination of fractional
crystallization and contamination. The liquidus relations are the same as in Fig. 6.
For simplicity, the contamination is shown as a distinct event from the crystallization. In reality, however, chromite probably crystallized while the siliceous
contaminant was being assimilated, in which case the magma would follow an
arcuate path through the chromite field.
STiLLWATER
GREAT
DYKE
EUSHVELD
A
e-f
Orthopyroxene
q-h
Orthopyroxene
d-e
c-d
o-b
Olivine _ (Chromite)
Chromife
Oiiwne -(Chromife)
f-g
e-f
c-d
b-c
Olivine-(Chromite)
Chromite
Olivine. (Chr0mi:ef
ChremltB
b-c
a-b
Orthopyroxene
Chromate
Fig. II. Variations on the model in Fig. 10 illustrating the formation of the more
common stratigraphic sequences involving chromitite layers in the Stillwater,
Great Dyke and Bushveld complexes.
1006:
1006
T. N. IRVME
time to account for the association of anorthosite with some of the Bushveld chromitite, but the mechanism does offer an explanation for the exceptional abundance of
o~hop~oxeni~
and reIative dearth of affihated olivine cumulates in this complex.
An attempt to define the main physical processes that appear to be required by
the mechanism is shown in Fig. 12. Diagram A depicts ‘normal’ co~~tions during
Fig. 12. Diagrams illustr&ing the physical processes postulated for the formation
of ahromitite layers. It is assumed that the intrusion roof comprises a subs~~t~~
proportion of rocks such as pelitic schists snd granite that will yield salic melt at
temperatures well below the fiquidus of the basic magma. The tendency of RsU
to accumulate at tba top of intrusions would facilitate the melting process. In
diagram A, abcdeis s possible tom~r~t~re path for the basic magma as it moves
in the convection path ABCDE;
crystallization should ideally begin at D and
continue to E. Branch cf exemplifies the temperature path if the liquid were to
descend s cooling wall; branch cg should obtain if the magma cooled to its liquidus
while moving along the roof. The paths in diagram B illustrate possible effects of
contamination.
For further discussion, see text.
accumulation of the layered series ; B and C portray events relating to episodes of
contamination. The temperature and crystallization paths illustrated for the
convecting basic magma in A and B are similar to those described in detail by IRVME
~~97~b). &&ice it to note here that the ~on~e~tio~ currents play an integral role
in both the e~sta~~zation of the magma and the deposition of the crystals.
The critical feature in Fig. 12A is that the roof contact of the intrusion is continuously exposed to the high temperature of the basic magma owing to the fact that
Crystallization
sequences in the Muskox intrusion and other layered intrusions-II
1007
the minerals formed in the body entirely accumulate on its floor. Thus salic rocks
along the contact, such as pelitic schists or granite, would be expected to undergo
melting, yielding granitic liquid that would tend to float on top of the basic magma
and remain separate because of its low density and high viscosity.
Heat transfer
models indicate that large amounts of granitic melt can be produced in this way
(IRVINE,1970b), and in the Muskox intrusion and Bushveld Complex, melt of this
type appears to be represented by thick zones of granophyre (see section on roof-rock
melting below).
In Fig. 12B, the basic magma is contaminated by mixing with a large quantity
of the salic melt. Subsequent events could be very complicated, but in general the
effect of the contamination should be to lower the actual temperature of the basic
liquid (because the salic melt would generally be somewhat cooler) and to reduce ita
liquidus temperature (by freezing-point depression)-and,
of course, it should cause
chromite to crystallize alone. The dissolution of the salic melt might be expected to
take place gradually, and in fact there is evidence that some of the Muskox chromite
may even have grown in an emulsion-like mixture of t’he two liquids (see section on
inclusions in chromite below).
Just what would cause an abrupt episode of magma mixing is a matter of speculation, but a strong possibility is that it is touched off by sagging or shifting of the
intrusion roof in response to tectonic activity or to major movements of magma in
some distant part of the system. Another possibility, illustrated in Fig. 12C, is that
the mixing occurs during the influx of new magma, for example, in advance of the
formation of a new cyclic unit. If the new magma flowed along the roof as shown
(because it was hotter and therefore less dense than the old magma), it would be
subject to contamination at an early stage when chromite was most likely to precipitate. But any chromitite layers produced should probably appear at the base of
cyclic units,* rather than at an intermediate level as in the Muskox intrusion.
The scheme in Fig. 12 embodies several features that appear important to the
explanation of the lateral extent of chromitite layers. Because the melting of the
roof rocks is a natural consequence of the fact that the minerals that formed in the
intrusion accumulate on its floor, it should be possible for the melt to form (and spread)
under large areas of the roof and so be available as a contaminant to large parts of the
intrusion. Because the contamination starts at the roof, there is maximum opportunity for its effects to be dispersed through the basic magma before this magma
convects down to the floor where the chromite is finally deposited.
And if the
contamination occurred during the infusion of fresh magma as in Fig. 12C, the lateral
flow would tend to spread its effects.
This last possibility would appeas to be of
particular interest in the Bushveld Complex and Great Dyke, inasmuch as they
contain exceptionally extensive chromitite layers, many of which occur at the base
of cyclic, or otherwise repeated stratigraphic units (e.g. WORST, 1960; CAMERON and
DESBOROUQH, 1969).
It might be noted in this regard that the cyclic units themselves, just like chromitite layers, tend to be sharply defined and constant in stratigraphy over long distances.
* An exception might arise if the new magma carried a large proportion of suspended olivine
that could settle out before chromite began to crystallize alone.
T. N.
1008
IRVINE
The implication is that the changes in magma composition leading to their formation
are propagated through an intrusion with considerable efficiency. The fact that the
units can be extremely well differentia~d in terms of both phase layering and cryptic
layering also implies that they formed from magma that was very well mixed. Indeed,
the impression is that the mixing was considerably more rapid than the crystallization,
so that even relatively small changes in liquid composition were immediately
reflected in the composition of the cumulates.
Another feature of the scheme in Fig. 12, pertinent to the thickness and stratigraphic defi~tion of c~omiti~ layers, is that the oonta~nation process ~en~u
s&&o calls for only the blending of liquids, not assimilation of solids, so heat
requirements should be minimal. Thus, given the necessary quantity of granitic
melt, an intrusion could become intensely contaminated almost instantaneously in
terms of its life history.
S~PORTI~~ E~IDENOEAXD D~scussroa
Inchsionns in chromite
Some of the strongest evidence in support of the proposed contamination
mechanism comes, remarkably, from features directly visible in the ohromite itself.
It has been found that many of the chromite grains in the Muskox chromite-rich
layers contain roughly spherical, silicate-rich inclusions, lo-100 pm in diameter,
that generally comprise several minerals and so would appear to represent droplets
of trapped liquid (Fig. 13). Examination of the inclusions by reflection microscopy
(they are rarely large enough to be seen in transmitted light in ordinary thin sections)
and by electron microprobe has shown that they variously consist of combinations
of orthop~oxene, a chromian titanian phlogopi~ and its sodium analogue, sodic
plagioclase, pyrrhotite, chal~opy~~, rutile and probably K-feldspar and quartz.
Of more immediate interest, however, are the levels of silica and alkalies. ‘Bulk’
electron microprobe analyses (Fig. 14) show SiO, values up to 86 wt.%, and total
alkalies commonly range from 5 to 11 per cent, higher than would ordinarily be
expected of basaltic liquid. Several inclusions even gave analyses resembling
granite: the one in Fig. 13A, for example, showed 72% SiO,, 19% AlzOs, 5% K,O
and 4% N&,0. Thus it would appear that the inclusions probably represent droplets
of contaminant granitio melt that were trapped in the chromite while in various
stages of mixing with its basaltic parental liquid. If so, then they constitute evidence,
as mentioned earlier, that the concentrated chromite grains actually grew in an
emulsion-like mixture of the two liquids.
Another type of inclusion in Muskox chromite of interest in this regard is rutile.
Some of this mineral, as just mentioned, is found in the silicate inclusions (Fig. 13C),
but more commonly it occurs as small, clearly primary, rodlike crystals scattered
through the dense population of chromite grains in the chromite-rich layers. The
significance of these inclusions relates to the fact that in the system MgO-TiO,-SiO,
at low pressures, rutile is not compatible with forsterite but will ~oprecipita~ w&h
pyroxene of enstatite composition (MASSAZZA
and SIRCRIA,1958). If these relationships extend to olivine and orthopyroxene such as occur in the peridotite and
orthopyroxenite layers adjoining the Muskox chromite-rich layers, then the presence
Crystallization
sequences in the Muskox intrusion and other layered intrusions-II
LLL’-46-w
0
100
20
SIO,,
1009
wt
%
Fig. 14. Electron microprobe data from spheroidal silicate inclusions in chromite
from the Muskox chromite-rich layers. Note that the scale for (Na,O + K,O) is
divided by 2.
of the rutile inclusions
would imply that the Muskox magma was considerably
enriched in silica just before the concentrated
chromite was precipitated,
as in
Fig. 10.
Inclusions in chromite of the types just described have also been found in the
Bushveld and Stillwater chromitite layers. The silicate-rich inclusions in Bushveld
chromite (MCDONALD, 1965) consist principally of orthopyroxene,
clinopyroxene,
biotite and plagioclase, of which only orthopyroxene
is associated as a cumulus
phase; those in Stillwater chromite are mainly olivine and biotite (JACKSON, 1961,
1966). In both intrusions the biotite was noted as distinctive, suggesting that concenMcDonald postulated that the inclusions represent
tration of alkalies is exceptional.
droplets of silicate melt trapped in the chromite as it crystallized from an immiscible
chromite-rich
liquid, but JACKSON (1966) argued that it is unlikely that such a
By the present hypothesis, the chromite
liquid would form in a basic magma.
would crystallize from the basic magma, and the inclusions would represent contaminant salic melt in various stages of mixing and assimilation, combined in some cases
with small amounts of associated crystalline material (e.g. olivine, pyroxene and
rutile) .
The occurrence of rutile in the Bushveld Complex as inclusions in chromite has
been illustrated by MCDONALD (1965, Fig. 4), and according to E. N. Cameron
(personal communication,
1974) it may also be associated as a discrete cumulus
phase. Rutile inclusions in chromite in Stillwater chromitite have been noted by
The fact that rutile is not a usual primary
Jackson (personal communication,
1969).
phase in low-pressure ultramafic and gabbroic rocks suggests the possibility that,
like the chromite, it too may have been precipitated in response to contaminat’ion.
1010
T. N.
IRVINE
No conclusive data are yet available in this regard, but it would appear significant
that the cations in the rutile structture occur entirely on octahedral sites, and that
Cr,C, may enter the structure in relatively large concentrations (e.g. Cr,O, contents
up to 7 wt.% have been recorded in natural rutiles; MEYER, 1975). The rutile in
Muskox chromite shows l-5-2%
Cr,O,. Crystal-field stabilization of the Cr3+ ion
might therefore have been a critical factor in the formation of the mineral, in which
case contamination could have been the controlling mechanism.
Evidence of roof-rock melting and contamination
If the proposed mechanism of origin for chromitite layers is valid, then the
intrusions in which the layers occur should show independent evidence of roof-rock
melting and contamination.
In the Muskox intrusion, the basic magma was emplaced
along and just beneath an unconformity
between a basement complex of pelitic
schists, metamorphosed acidic volcanic rocks, and potassic granite, and a relatively
undeformed cover of quartz-rich sandstone. The roof consists partly of hornfelsed
schist, partly of granite, and partly of quartzite metamorphosed from the sandstone.
The granophyre at the top of the intrusion everywhere includes fragments of quartzite, regardless of the type of roof rock, and many of the fragments have been rounded
and embayed by assimilation. In the northeast (Fig. 1) the granophyre is transitional
with an extensive unit of ‘contact breccia’ consisting of closely packed blocks and
fragments of quartzite up to 5 ft on a side cemented by some lo-20 per cent granophyre. Xenoliths of schist and granite are less common and are restricted to areas
in which these rocks form the roof. The schists in the roof are locally permeated by
small granitic veinlets, and some of the granite shows coarse, graphic textures that
could represent recrystallized melt. Below the granophyre in the layered series and
marginal zones, recognizable
country rock material is extremely rare, the only
occurrences being a few blocks of quartzite in the uppermost pyroxenite and gabbro
layers. In view of these relations, it is not difficult to imagine the granophyre as
representing magma formed mainly by melting of the schist and granite that floated
at the top of the intrusion together with a residue of quartzite fragments, which
survived because they were so refractory.
This interpretation appears to be supported by chemical and isotopic relations.
The granophyre is extremely pot’assic and so is not the kind of rock that would be
expected to form as the end product of crystallization differentiation of normal
tholeiitic magma. On the other hand, the granite and all the schist samples that have
been analyzed from the roof show very high K,O/Na,O
ratios and so, on partial
fusion, should yield a highly potassic salic melt (IRVINE and BARAGAR, 1972, pp.
28-32). The Rb-Sr and Sr-isotope relations of the granophyric rocks suggest an age
of approximately
1700 m.y., similar to typical K-Ar ages from the basement schists
and granites and much older than the 1200-1250 m.y. age indicated for the intrusion
itself by K-Ar, Rb-Sr, and paleomagnetic data on the ultramafic and gabbroic rocks
(D. H. Loveridge, T. N. Irvine, and R. K. Wanless, unpublished data). Apparent
initial s7Sr/86Srratios for the ultramafic and gabbroic rocks range from O-705 to 0.709,
and a 1700-m.y. isochron fitted to the granophyre data projects to the same range.
Such high ratios are indicative of extensive crustal contamination (e.g. FATJREand
HURLEY, 1963).
Crystallization sequences in the Muskox intrusion and other layered intrusions-II
1011
The Bushveld Complex presents a similar picture but on a grander scale. The
complex is enormous, with a layered series about 25,000 ft thick (WILLEMSE, 1969;
WAGER and BROWN, 1968). It has highly varied stratigraphy, featuring hundreds of
layers ranging from orthopyroxenite,
peridotite, anorthosite and chromitite in the
lower levels, through norite and two-pyroxene
gabbro at intermediate levels, to
magnetite, anorthosite and olivine ferrodiorite toward the top. There is no upper
border zone of mafic rocks that could have protected the roof; and the remaining
roof rocks, which are mainly felsite (i.e. a salic volcanic rock) and highly metamorphosed pelitic sediments, show strong evidence of partial melting (WAGER and BROWN,
1968; WILLEMSE, 1969; VON GRUENEWALDT, 1972). Thegranophyreinthecomplex,
together with associated granite, occurs both at the roof contact and as sills in the
felsite and is locally as thick as 7000 ft. IRVINE (1970b) showed by a simple thermal
model that the basic intrusion could have melted some 3000-3500 ft of liquid of
granitic composition, and VON GRUENEWALDT (1972) has supported the model and
argued for conditions that would lead to even more melting. In view of the size of
the complex, one can readily visualize countless events in which the melt might have
mixed with the basic magma and modified its crystallization order, thereby causing
many stratigraphic complications.
As in the Muskox intrusion, this interpretation appears to be supported by Rb-Sr
and Sr-isotope relations. Data obtained by DAVIES et al. (1970) on the granitic and
dioritic rocks yielded an isochron indicating an age of 1954 m.y. and an initial
%Sr/%Sr ratio of 0.716. Data on the gabbroic and ultramafic rocks gave initial ratios
of 0.7069-0.7090,
based on the same age. (The latter rocks gave only very low
a’Rb/YSr ratios, and so no reliable independent age could be established for them.)
The high initial ratios are strongly indicative of contamination,
and DAVIES et al.
(1970, p. 589) remarked that, to explain the data, it was necessary to postulate
extensive differentiation after contamination, which is essentially the nature of the
contamination mechanism proposed here.
The steeply dipping layered rocks of the Stillwater Complex have been unroofed
through a combination of erosion and deformation, and there is no mention of xenoliths, assimilation, or contamination in the available descriptions of the body (e.g.
HESS, 1960; JONES et aE., 1960; JACKSON, 1961). BONINI et ab. (1969) reported a
large gravity anomaly adjoining the stratigraphically
upper side of the complex,
which they ascribed to an extension at depth, and they suggested that the exposed
area of the body might represent as little as 10 per cent of its total extent. If such
is the case, then some roof rock might still be available at depth to sampling by
drilling. Data so far obtained from the complex on Rb-Sr and Sr-isotope relations
(STEUBER and MURTHY, 1966; KISTLER et al., 1969; FENTON and FAURE, 1969)
have not yielded a definitive age, and none of the suggested possibilities is compatible
with a U-Pb age of 2750 m.y., reported by NUNES and TILTON (1971). For an age
of 2750 m.y., four of the investigated samples show nominal initial *Sr/*6Sr ratios of
about 0.701, a value that could be mantle derived. Most samples, however, give
values of about O-703, and two would have much higher ratios. FENTON and FAURE
(1969) suggested that the two high values were due to serpentinization,
a reasonable interpretation, but it is apparent that the problem warrants further study with
a view to possible effects of contamination.
1012
T. N. IRVINE
The Great Dyke also has been unroofed by erosion, and there seems no possibility
that it has any extensions at depth. It does, however, include various country rock
xenoliths, principally of greenstone, hornblende gneiss, banded ironstone (quartzmagnetite rock), quartzite, graywacke and granite, plus a few blocks of exotic
ultramafic rocks and chromitite representative of older ultramafic bodies in the
region (WORST, 1960). The xenoliths are not abundant but apparently are widespread in the gabbroic cumulates that form the upper part of the complex. Interestingly, however, the one mentioned occurrence in the ultramafic cumulates is a block
of granite, about 300 ft long and 80 ft wide, situated between the upper two pyroxenite layers, which is a part of the section that also contains two chromitite layers.
Worst noted rocks similar to the xenoliths among the wall rocks of the intrusion,
but he considered the xenoliths to have come from the roof, implying that the body
did not have a complete upper border zone to protect the roof, at least in the later
stages of layered series accumulation. Worst also remarked that among the sedimentary xenoliths, quartz& appeared to survive the best, and indeed most of the
xenoliths are types that would largely withstand melting in a basic or moderately
ultrabasic magma. The main exception is the granite, and some of it is described as
being cut by ‘acid veins’ and showing granophyric or graphic intergrowths of quartz
and feldspar, features that might be related to melting. Worst’s summary of the
regional geology indicates that granite and metasedimentary schists are the predominant rocks adjoining the dyke, so if the roof had the same constitution, one could
infer from the preserved features of the dyke that large amounts of these rocks had
been melted, leaving only the xenolithic residue of more refractory units listed
above.
DAVIES et al. (1970) obtained a Rb-Sr isochron for the Great Dyke giving an age
of 2532 m.y. and an initial ratio of 0.7025. The latter value indicates that there has
not been much contamination, but so too does the overall lithology of the intrusion,
which is dominated by olivine cumulates. In fact, the Great Dyke is so like the lower
part of the Muskox intrusion as to suggest that it too may have been open repeatedly
on a large scale to loss as well as addition of magma. In such a case contamination
might not be cumulative in the intrusion and so might leave only a limited stratigraphic record that would be difficult to detect.
Role of oxidation
One of the first questions that is raised by the proposed origin for chromitite
layers is, How much contamination is required to form a given layer? It is evident
that a large amount of magma must be affected. For example, if half the chromium in
the liquid could be precipitated, then to form 1 ft of chromitite containing 40 per cent
Cr,O, would require 1000 ft of liquid like the Muskox chilled margin containing
O-08 per cent Cr,O,. But whether this much precipitation would require 5 per cent
contamination or 50 per cent is unknown. This aspect of the problem must be tackled
by experiments on natural materials.
The answer, however, also depends on the role of oxygen. The presentation above
has emphasized the effects of silica and alkalies, but chromite crystallization can be
critically dependent on the oxidation state of the magma, and the contamination
mechanism does offer a convenient means for introducing oxygen (e.g. as H,O in the
granitic melt). An evaluation of this factor requires information on the oxidation
state of the concentrated chromite, particularly as compared with that ofdisseminated
Crystallization
sequenoes in the Muskox intrusion and other layered intrusions-II
1013
chromite
that may have coprecipitated
with silicate minerals at an immediately
preceding stage.
In an attempt to obtain this information
the author has made electron
microprobe analyses of chromites through the Muskox layered series leading up
to and including the chromite-rich layers. From the data, values for the oxidation
ratio Fea+/(Fe 2+ + Fe3+) have been calculated on the basis of the spine1 R,O,
stoichiometry, and estimates of the oxygen fugacities at which the mineral may have
equilibrated have been made following a method outlined by IRVINE(1965). The fo,
results are numerically reasonable in terms of experimental data obtained by HILL
and ROEDER (1974) on chromite crystallization in tholeiitic magmas, but they are
based on several assumptions and so are regarded a,s significant only in the respect
that they show that the fo, values derived in this way correlate closely with the
iron oxidation ratio (Fig. 15). The data show some systematic variations with
tOOFe3?(Fc?fcF~+)
Log(fo,)
16
,100
1200
:
,300
: 56~.
,’
mu4 Olivlneclln!xwEsmui%
.’. :
.”
‘:
,400
;4:
f
.’
.z
;
T-i
g
Feldspathic
peridotite
:
Feldspathic
perldotite
.
B
m.
..
.
. .
B
-
..,.
.
.
e
,500
=
.”
1600
Dunite
1700
Y
. .
iQO0
.
--
., Per’dotite
Lii
I
‘20..
%%?I+
JAi-L-L
Chromite-rich
layers
Fig. 15. Plot of the oxidation ratio of iron in Muskox chromite in the North drillhole section leading up to and including the two chromite-rich layers, computed
from electron microprobe analyses on the basis of spine1 R,O, stoichiometry.
Also
shown are log forvalues estimated from the analyses on the assumption that the
chromite equilibrated with olivine and orthopyroxene at 1300’K (1027%), by a
method outlined by IFNINE (1965).
1014
T. N. IRVINE
stratigraphic height, but there is no clear indication that the concentrated chromite
is more oxidized than the disseminated grains.
A principal problem with this type of approach is that the chromite may have
changed in composition after accumulation by reaction with intercumulus liquid and
associated silicate phases. Reaction with the liquid might explain, for example, the
slightly higher oxidation ratio of the chromite in the peridotite and feldspathic
peridotite as compared with that in the dunite (Fig. 15) inasmuch as they are
essentially distinguished from the dunite by larger proportions of intercumulus
materials. A further complication is that there is evidence in the Muskox intrusion
that intercumulus liquid filtered upward through the layered series for hundreds of
feet, reacting continuously en route (Irvine, unpublished data). Thus it must be
concluded that, if increases info, did contribute to the precipitation of the Muskox
chromite-rich layers, either they were too small to be clearly reflected in the composition of the chromite or their record has been obliterated by later events. Other
intrusions may be more definitive in this regard.
APPLICATIONOF THE CONTAMINATION
MECHANISM
TO OTHER TYPES OF MAQMATICORE DEPOSITS
If the contamination mechanism is viable, then it should also pertain to magmatic
ore deposits other than just chromite, especially layered deposits of magnetite and
sulphides.
Mugnetite
The magnetite layers of the Bushveld Complex are obvious candidates for consideration inasmuch as they are very similar in appearance and lateral extent to
chromitite layers (e.g. WILLEMSE,1969; MOLYNEUX,1970). They occur in association with anorthosite layers at a relatively high stratigraphic level in the complex, just
where olivine reappears as a cumulus phase after being absent through about 13,000 ft
of section. At this level, magnetite has been a subordinate (disseminated) cumulus
phase for about 600 ft, and the mafic silicates have become moderately rich in iron,
their Mg/(Mg + Fe) ratios averaging about 0.5. The general mineral association is
closely represented in the systems MgO-FeO-Fe,O,-SiO, (MUANand OSBORN,1956),
CaMgSi,O,-Mg,SiO,-FeO-E‘e,O,-SiO,
(PRESNALL,1966) and CaAl,Si,O,-Mg,Si04FeO-Fe,O,-SiO, (ROEDERand OSBORN,1966), in all of which oxidation is a factor
of prime concern ; thus we are led naturally to the possibility that the magnetite
layers were precipitated because the magma was suddenly oxidized. The process is
entirely feasible in terms of the physical chemistry, and given that the contamination
mechanism provides a means for introducing oxygen, it is perhaps unnecessary to
look further for a mechanism of origin, only for substantiating evidence. On the
other hand, the association of the magnetite with anorthosite is not obviously accounted for by oxidation, and it might be that there are other factors relating to
contamination that are more critical to the formation of both rock types. The
problem warrants much more investigation.
Sulphide minerals occur in small but conspicuous concentrations in the Muskox
chromite-rich layers (SMITH,1962 ; CHAMBERLAIX,
1967) and in some of the chromitite
layers of the Stillwater Complex (PAGE and JACKSON,1967). In the Bushveld
Crystallizationsequencesin the Mu&ox intrusionand other layeredintrusions---II 1015
complex, sulphides and chromite are distinctive phases in the important platiniferous
Merensky Reef (e.g. LIEBENBERQ, 1970). It is generally believed that the sulphides
were precipita~d as an immiscible sulphid~oxide liquid, and the association with
the chromitite strongly suggest’s that the liquid was concentrated by the same processes as the chromite.
Information on the solubility of sulphide in magmas and on the determining
factors is still very limited. The best available data for basaltic liquids come from a
study by SKINNER and PECK (1969), which indicated that a Kilauea tholeiite became
saturated with sulphide liquid at 0.038 wt.% S and 1065°C. NALDRETT (1973, p. 9)
has quoted a value of 0.16 per cent S for saturation of a picritic liquid at 1450°C. It
has long been known from studies of slags that sulphide solubility in silicate melts
increases with the Fe0 content and temperature of the melt and decreases with its
SiO, content and oxygen fugacity (e.g. VOGT, 1919; RICHARDSON and FINCHAM,
1954). From these relations, RICHARDSON and FINCHAM (1954) argued that the
volubility is essentially determi~~edby the bonding of sulphur with the Fez+ in the
melt, and more recent investigators have generally supported t’his interpretation (e.g.
MACLEAX, 1969; HAUCHTON and ROEDER, 1971; HAUGHTON and SKINNER, 1972).
All the above-mentioned factors are basically represented in the system FeS-FeOFe,O,--SiO,, and the join FeS-FeO-SiO, is of parhicular interest in the present
context
In this join, as seen in Fig. 16A, certain liquid compositions in the fayalite
liquidus field can be shifted by the addition of silica into a two-liquid field where they
will exsolve or precipitate a sulphide-rich liquid. Translating this effect into a
model diagram for tholeiitic magmas (Fig. 16B), one sees that it could yield sequences
of cumulates with concentrations of sulphides at the same stratigraphic levels as
were described earlier for chromitite layers (Figs. 10 and 11). It appears possible
that the effect in the experimental system is due, at least in part, to relative differences in the crystal-field stabilization energies of Fe%+in oxide, silicate and sulphide
structures (cf. BURNS, 1970, Tables 6.2, 6.5 and 7.1), in which case the effect for
nickel and copper sulphides should probably be more pronounced, Ni2+ and Cu2+
showing much greater stabilization. This possibility may be explained by means of
the following equilibria :
(Ma+
octahedral1Silicate liquid -
(Mnfeetabedral)Snlphide liquid
wn+ tetrahedral 1Silicate liquid -
octabearal) Mafiasilicate mineral
where pi:fn+= Xi”*+,CWf, Pea+, Mg2f.
Thus, as the frequency of octahedral sites in a cooling basic silicate liquid just
saturated in olivine or pyroxene and sulphide liquid is reduced through contamination
by salic material, these equilibria should shift strongly to the right for NW, Cu2+and
Fez+, causing them to be preferentially expelled into the other phases. But if the
sulphur in the silicate melt has dissolved largely by bonding to these ions (as apparently it does for at least Fe2+), then it too should preferentia~y exsolve ; hence the
precipitation of sulphide liquid to the exclusion of the silicate minerals is expected.
It would appear that this mechanism for precipitating sulphide liquid from basic
magmas may be particularly important in the case of the Sudbury nickel ores.
T. S. IRVINE
1016
Fe0
wt%
FeS
SiO,
d-e
c-d
a-b
CUMULATE SEQUENCE
Orthopyroxene-(Sulphide liquid)
Sulphide liquid
Olivine - (Sulphide liquid)
ulphides
Oxides)
Olivine
Not to scale
Sulphide
(Oxides)
Fig. 16. (A) Liquidus relations for the join FeS-FeO-SiO,, after MACLEAN
(I969),
showing how certain liquids in the fayalite primary phase field can be forced to
precipitate a sulphide-rich liquid through the addition of silica. (B) Sohematic
representation of similar relations for tholeiitic magmas, showing how a combination of fractionation and oon~min&tion might produce & concentration of
sulphides euch as occurs in the Muskox chromita-rich layers, Compare with Figs.
IO and II.
Over the years, the concentration of these ores around the generally inward-dipping
footwall contact of the oval-shaped Sudbury Irruptive hss repeatedly been attributed
to gravitational
accumulation of immiscible sulphide liquid; and recent studies (e.g.
DIETZ, 1964; FRENCH,1968) have show~l that the Irruptive was probably emplaced
in response to the impact of a Iarge meteorite, having been intruded beneath an
extensive mass of shattered and partly melted material that was evidently formed
by impact from thick units of quartz&! and argillite and an underlying complex of
granitic gneiss, an ideal situation for salic contamination. Another important recent
discovery (SOUGH,PODOLSKYet cd., 1969; NALDRETTand KULLERUD,1967) is that
the ores are Iargely contained in a ‘sub-layer’ of small noritic, gabbroic, and dioritic
intrusions that are also rich in xenoliths, some of these consisting of country rock
schists, and many of distinctive, cumulate-type ultramafic and gabbroic rocks
ranging from dunite through peridotite, orthopyroxenite and websterite, to olivine
gabbro and norite. The main part of the Irruptive, by contrast, consists of thick,
Crystallization sequences in the Muskox intrusion and other layered intrusions-11
1017
rudely stratiform layers of noritic and gabbroic cumulates overlain by a similarly
thick unit of micropegmatite
(granophyre).
No clear age distinction has been
established between the mafic cumulates and the micropegmatite, but there is some
opinion that the sub-layer intrusions are younger than the norite (NALDRETT et al.,
1973, p. 211).
Working from the base summarized above, NALDRETT et al. (1973) proposed a
tentative sequence of events leading to formation of the sulphide ores. The following
is a modified sequence, based on the hypothesis that the sulphides were precipitated
by the contamination mechanism.
As the basic magma of the Irruptive was emplaced, olivine was a principal
mineral to crystallize, and accordingly there developed a complex zone of olivinebearing cumulates in the lowest parts of the funnel-shaped floor. But with the convection accompanying
emplacement and crystallization,
the magma was quickly
contaminated with large amounts of partly melted salic material from its roof, and
as a result the ensuing period was marked by wholesale precipitation of sulphide
liquid and by the erratic but progressive appearance of orthopyroxene
in place of
olivine in the crystallization order. The sulphide liquid presumably collected mainly
in pools on top of the olivine-bearing
cumulates, but as accumulation of silicates
resumed, it would tend to seep to some extent into interstices between the settled
crystals. Eventually conditions stabilized, and the main units of norite and gnbbro
were formed. During this period, orthopyroxene
was joined in the crystallization
order, first by plagioclase and later by augite, magnetite and apatite. However,
because of the high viscosity acquired by the magma through contamination,
the
fractionation of the crystals was very imperfect, and the cumulates that formed were
poorly layered and retained large amounts of pore liquid as are now evidenced by
abundant interstitial granophyre in the norite and gabbro (cf. NALDRETT et al., 1970).
In the meantime, the roof of the intrusion was further melted by the heat from the
basic magma and underwent some gravitative differentiation in itself, distinguished
particularly by the segregation of refractory quartzite fragments, which floated
upward to form a quartzite breccia with granophyre matrix at the present roof
contact of the Irruptive (cf. STEVENSON, 1963) .* The siliceous melt eventually
resolidified as the micropegmatite.
At some time toward the end stages of solidification, perhaps in response to a
major episode of subsidence in the center of the Irruptive (an event that appears to
be indicated by the lopolithic form of the body), the sulphide liquid together with
substantial amounts of basic silicate magma (possibly representing intercumulus
liquid that had been preserved in the hot, deep part of the structure) were squeezed
upward along the footwall contact to form the ore-bearing sub-layer intrusions.
Carried in these intrusions were pieces of the deep ultramafic and gabbroic cumulates,
plus fragments of country rock schists picked up en route.
The key features of this sequence of events in the present context are that (1) it
relates both the sulphide ores and the exceptional abundance of orthopyroxene in the
Irruptive to the meteorite impact, and (2) it allows for the development of Ni-rich
sulphide in that the sulphide liquid is precipitated at an early stage before the basic
magma is depleted of Ni through extensive fractionation of olivine and pyroxene.
* It appears that this quartzite breccia, is almost identical with the contact breccia at the top
of the Muskox intrusion.
1018
T. N. IRVINE
If it should prove that the sub-layer intrusions are older than the norite, the events
would have to be rearranged, but these features could be retained.
A question that is raised that points to a possible test of the mechanism is whether
any chromium was precipitated with the sulphides. GASPAERMI and~ALDRETT(1972)
have reported Cr,O, in concentrations of S-13 wt.% in magnetite in the norite,
enough to suggest that the magma should have been capable of crystallizing chromite
at an earlier stage. It is unlikely that chromite would form once orthopyroxene
began to crystallize, but earlier precipitation might be expected. To the author’s
~owledge, no one has recorded chromite in the ores as other than an accessory phase
in some of the ultramafic xenoliths, but some of the ores are rich in what appears to
be primary magnetite in which chromium has been detected (HAWLEY, 1962), and
perhaps an intensive search would reveal substantial concentrations. On the other
hand, it should be borne in mind that the extensive precipitation of sulphide might
tend to suppress the crystallization of chromite, for exampIe, by depleting the magma
in Fez+.
CoNoLnsIoNs
From this account, two inferences of considerable general significance may be
drawn. One is that the differences between the chemical structures of basic and
salic silicate melts would appear to be extremely important in determining the
precipitation of magmatio ores, especially the ores of transition metals. The other
inference is that the compositional characteristics of the cumulates that collect at
the bottom of a layered intrusion can be strongly dependent on the events that
occurred at its top. The proposed contamination mechanism of origin for chromitite
layers and similar ore deposits combines these two factors, and an attempt has been
made to illustrate its capab~ities and potential, and some possible implications. An
important attribute of the mechanism that might be noted is that it should frequently
be amenable to testing in a variety of ways by field, chemical, mineralogical and
isotopic methods.
Acknowledgements-Muchof the information presentedon the Muskox intrusionstems from work
done at the GeologicalSurvey of Canada, and the writer is very grateful for the support provided.
Survey colleagues to whom particular thanks are due are C. II. S~H,
D. C. FINL)LAY,R. K.
WANLESS and W. D. LOVERIDGE.
At the GeophysicalLaboratory, the electron microprobework
was facilitated through instruction by C. G. HADIDIACOSand L. W. FINGER,and the experimental work, through help from I. KUS~IRO, R. J. ARCULUS,R. H. MC~AE;J;ISTER
and E. F.
OSBORN. The manuscript has been considerably improved as a result of comments and suggestions by R. J. ARCULUS,P. M. BELL, E. F. OSBORN,D. RUMBLE,D. MSHAW and H.S.YODER, Jr.
REFERENCES
BICHANR. (1969) Origin of chromiteseams in the Hartley Complex of the Great Dyke, Rhodesia,
Econ. Geol. Monogr. 4, 95-113.
BONINIW. E., HIGGINSE. and COONLEYL. M. (1969)
Original size of classic Stillwater Complex
and possible controls on regional Larimide tectonics (abstract). Geol.Sot. Amer. Spec. Paper
121,31-32.
BURNS R. G. (1970) ~~~~e~~log~cal
d~~~~c~~~~s of Crystal F&Ed Freon, 224 pp. Cambridge
University Press.
BW~NS R. G. and FYFE W. S. (1964) Site preferenceenergy and selected uptake of transitionmetal ions during magmatic crystallization. Science144, 1001-1003.
BURNS R. G, and FYFE W. S. (1967) Crystal-field theory and the geochemistry of transition
elements. In Researches in Gfeochernistry (editor P. H. Abelson), Vol. 2, pp. 259-285. Wiley.
Crystallization sequences in the Muskox intrusion and other layered intrusions-II
1019
CAMERONE. N. and DESBOROUGH
G. A. (1969) Occurrence and characteristics of chromite
deposits--eastern Bushveld Complex. Eco~. Gd. Monogr. 4, 23-40.
CEAMBERL~NJ. A. (1967) Sulfides in the Muskox intrusion. Can. J. Earth Soi. 4, 101-153.
DAVIES R. D., ALLSOPH. L., EBLANK A. J. and ~MANToN
W, I. (1970) &-isotope studies on
various layered mafie intrusionsin southernAfrica. Ceol.XOC.S. Africa Spew. Fubl.1,576-593.
DEITZ II. (1964) Sudbury structure as an astrobleme. J. Be& 72,412-434.
EVANSB. W. and WRIGHTT. L. (1072) Composition of liquidus ohromitefrom the 1959 (Kilauea
Iki) and 1965 (Makaopuhi)eruptionsof Kilauea volcano, Hawaii. Amer. Mineral. 57,217-230.
FAURE G. and HURLEY P. M. (1963) The isotopic composition of strontium in oceanic and
continental basalts. Application to the origin of igneous rooks. J. Petrol. 4, 31-50.
FENTOWM. D. and FAUREG. (1969) The age of the igneous rocks of the Stillwater Complex of
Montana. Bull. CTeoZ.
Sot. Amer. SO, 1599-1604.
FRENCHB. M. (1969) Sudbury structure, Ontario: some petrographicevidence for an origin by
meteorite impact. In Shook Metamorphism of Natural Materials, (editors B. M. French and
N. M. Short), pp. 383-412. Mono Book Corp., Baltimore, Md.
GASPARRIN~
E. and NAL~RETT A. J. (1972) Magnetite and ilmenite in the Sudbury Nickel
Irruptive. Icon Geol. 67, 605-621.
H~I
T. A. and UPRIGHT
T. L. (1967) The fractionation of nickel between olivine and augite as
a geothermometer. Cieochim.Cosmochim.Acta 31, 877-884.
HATJGKXTOND. R. and ROEDERP. L. (1971) Iron and sulphur in basaltic liquids (abstract). Cara
Miraeral. 11,574-575.
HAU~HTOND. R. and SPINNERB. J. (1972) Chemistry of sulfur in melts. Int. Cfeol. Congr.,
24th, &lontreal, Vol. of Abstracts, p. 420.
Hiwm~
J. E. (1962) The Sudbury ores: their mineralogy and origin. Co% ~~n$~a~. 7, I-207.
Hxss H, II. (1960) Stillwater Igneous Complex, Montana; a quantitative mineralogicalstudy.
Ueol. Sot. Amer. Mem. 80,230 pp.
HILL R. and ROEDERP. L. (1974) The orystallization of basalt as a function of oxygen
fugacity. J. Geol. 82, 709-730.
IRVINET. N. (1965) Chromian spine1as a petrogeneticindicator. Part 1, Theory. Can. J. Earth
Sci. 2, 648-672.
&-NE
T. N. (1967) Chromian spine1 as a petrogenetictindicator, Part 2. Petrologic applications. Gala.S. Earth Sci, 4, 71-103.
IRVINE T. N. (1970a) Crystallization sequences in the Muskox intrusion and other layered
intrusions, I. Olivine-pyroxene-plagioclase relations. Geol. Sot. S. Africa, Spec. Publ. 1,
441-4’76.
IRVINET. N. (1970b) Heat transfer during solidificationof layered intrusions,I. Sheets and sills.
Can J. Earth. Sci. 7, 1031-1061.
IR~IXE T. N. and BARAQARW. R. A. (1972) Muskox intrusion and Coppermine River lavas,
Northwest Territories, Canada. 1st. Geol. Congr., 24th, MontreaE, Et;eld Excursion A29
Bwidebook, 70 pp.
IRVINE‘I. N. and SMITHC. H. (1967) The ultramafic rooks of the Muskox intrusion. In UEtra.
ma& and Related Rocks (editor P. J. Wyllie), pp. 38-49. Wiley.
IRVINET. N. and SMITHC. H. (1969) Primary oxide mineralsin the layered seriesof the Muskox
intrusion. Econ. Geol. illonogp. 4, 76-94, 1969.
JACKS~XE. D. (1961) Primary textures and mineral associationsin the ~tramafi~ zone of the
Stillwater Complex, Montana. U.S.Ceol.
Suw. Prof. Paper 358, 106 pp.
JACKSONE. D. (1963) Stratigraphic and lateral variations of chromite compositions in the
Stillwater Complex. Mineral. Sot. Amer. Spec. Paper 1,46-54.
JACKSONE. D. (1966) Liquid immiscibility in ohromite seam formation-a discussion. Econ.
Geol. 61, 777-780.
JONESW. R., PEOPLESS. W. and HOWL~WDA. L. (1960) Igneous and tectonic structuresof the
Stillwater Complex, Montana. U.S. Geot.Ssrv. B&l. lO’sl-IX, 281-340.
KEITH r/I. L. (1954) The system MgO-Cr,Os-SiO,. J. Amer. Ceram. Sot. 37, 490-496.
KISTLER 3%.W., OBRADOVICH
J. D. and JACKSON
E. D. (1969) Isotopic ages of rocks and minerals
from the Stillwater Complex, Montana. S. (reophys. Res. 74, 3226-3237.
LIEBENBER~L. (1970) The sulphidesin the layered sequence of the Bushveld Igneous Complex.
GeoE.Boc. S. Africa Spec. Publ. 1, 108-207.
1020
T. N. IRVINE
W. H. (1969) Liquidus phase relations in the FeS-FeO-FesO,-SiO,
system, and their
application in geology. Econ. Geol. 64, 865-884.
MASSAZZA
F. and SIRCHIA E. (1968) Il sistema MgO-SiOs-TiOs:
nota II-Gli equilibi alto stato
solid0 e alla fusione. C&m. Irad. (Milan) 40, 460-467.
MCDONALD J. A. (1965) Liquid immiscibility as one factor in chromitite seam formation in the
Bushveld Igneous Complex. Ecola. Geol. 60, 1674-1685.
MEYER H. 0. A. (1975) Chromium and the genesis of diamond. Geochim. Cosmochim. Acta 39,
929-936.
MOLYNEUX T. G. (1970) The geology of the area in the vicinity of Magnet Heights, Eastern
Transvaal, with special reference to the magnetic iron ore. Geol. Sot. S. Africa Spec. Publ. 1,
228-241.
MUAN A. and OSBORN E. F. (1956) Phase equilibria at liquidus temperatures in the system
MgO-FeO-Fe,O,-SiO,.
J. Amer. Ceruna. Sot. 39, 121-140.
NALDRETT A. J. (1973) Nickel sulphide deposits-their classifioation and genesis, with special
emphasis on deposits of volcanic association. Can. Inst. Mining Met. Trans. 76,183-201.
NALDRETT A. J., BRAY J. G., GASPARRINI E. L., PODOLSKY T. and RUCKLIDC+EJ. C. (1970)
Cryptic variation and the petrology of the Sudbury Nickel Irruptive. Econ. Geol. 65, 122-155.
NALDRETT A. J., GREENMANL. and HEWINS R. W. (1973) The Main Irruptive and the sub-layer
at Sudbury, Ontario. Int. Geol. Congr., 24th, Montreal, Section. 4, pp. 206-214.
NALDRETT A. J. and KULLERUD G. (1967) A study of the Strat’hcona mine and its bearing on
the origin of the nickel-copper ores of the Sudbury district, Ontario. J. PetroE. 8, 453-531.
NUNES P. D, and TILTON G. R. (1971) Uranium-lead ages of minerals from the Stillwater
Igneous Complex and associated rocks, Montana. Bull. Geol. Sot. Amer. 82, 2231-2250.
PAGE N. L. and JACKSONE. D. (1967) Preliminary report on sulfide and platinum-group minerals
in the chromitites of the Stillwater Complex, Montana. U.S. Geol. Stbrv. Prof. Paper 575-D,
D123-D126.
PRESNALL D. C. (1966) The join forsterite-diopside-iron oxide and its bearing on the crystallization of basaltic and ultramafic magmas. Amer. J. Sci. 294,753-809.
RICHARDSONR. D. and FINCHAM C. J. B. (1954) Sulfur in silicate and aluminate slags. J. Iron
Steel Inst. London 179,4-14.
RINCWOOD A. E. (1956) Melting relationships of Ni-Mg olivines and some geochemical implications. Geochim. Cosmochim. Acta 10, 297-303.
ROEDDER E. (1951) The system K,O-MgO-SiO,.
Amer. J. Sci. 249,81-130.
ROEDER P. L. and OSBORN E. F. (1966) Experimental data for the system MgO-FeO-FeaO,CaAl,Si,O,,-SiO, and their petrologic implications. Amer. J. Sci. 264, 428-480.
SKINNER B. J. and PECK D. L. (1969) An immiscible sulfide melt from Hawaii. Econ. Geol.
Monogr. 4, 310-320.
SMITH C. H. (1962) Notes on the Muskox intrusion, Coppermine River area, District of Mackenzie. Geol. Surv. Can. Paper 61-25, 16 pp.
SOUCHB. E., PODOLSEY T. and geological staff of the International Nickel Co. of Canada, Ltd.
(1969) The sulfide ores of Sudbury: their particular relation to a distinctive inclusion-bearing
facies of the Nickel Irruptive. Econ Geol.. 65, 122-155.
STEUBER A. M. and MURTRY V. R. (1966) Strontium isotope and alkali element abundances in
ultramafic rocks. Geochim. Cosmochim. Acta 30, 1243-1260.
STEVENSONJ. S. (1963) The upper contact phase of the Sudbury micropegmatite. Can. Mineral.
MACLEAN
7,413-419.
VOQT, J. H. L. (1919) Die sulfid: Silikatschmelzliisungen. Kristiania Vid.-Selsk. I. Mat.-Nat.
Klasse, No. I., 132 pp.
VON GRUENEWALDT G. (1972) The origin of the roof-rocks of the Bushveld Complex between
Tauteshoogte and Paardekop in the eastern Transvaal. Trans. Geol. Sot. S. Africa 75,121-129.
WAGER L. R. and BROWN G. M. (1968) Layered Igneow Rocks, 688 pp. Oliver & Boyd.
WILLEMSE J. (1969) The geology of the Bushveld Igneous Complex, the largest repository of
magmatic ore deposits in the world. Econ. Geol. Monogr. 4, l-22.
WORST B. G. (1960) The Great Dyke of Southern Rhodesia. Bull. Geol. Swv. S. Rhodesia 47,
234 pp.
YODER H. S., JR. and TILLEY C. E. (1962) Origin of basalt magmas: an experimental study of
natural and synthetic rock systems. J. Petrol. 3, 342-532.