Economic Geology

Vol. 71, 1976, pp. 1131-1158

Ultramafic and Related Mafic Rocks: Their Classification and

Genesiswith SpecialReferenceto the Concentrationof

NickelSulfidesandPlatinum-Group
Elements
A. J. NALDRETT
ANDL. J. C^
Abstract

Ultramafic and relatedmafic rockscan convenientlybe classifiedin terms of the tectonic environmentin which they were eraplaced.Among thosebodieseraplacedin an
orogenic environmentone can distinguishsynvolcanicbodies,bo',h intrusive and extrusive,eraplacedbeforethe peak of orogenesis;alpine bodies(ophiolites) eraplaced

duringorogenesis;
and Alaskan-type
complexes
eraplaced
in the waningstagesof orogenesis.Large strafiformcomplexes
(Bushveldtype), plateau-type
sills (Dufek type),
and medium-and small-sizedpicritic intrusionsare characteristicof a cratonic environ-

ment. Synvolcanic
bodiesbelongto two distinctmagmasuites:(1) the highlymagnesiankomatiiticsuite,characterized
by spinifex-textured
ultramaficflows,a low FeO/
(FeO + MgO) ratiofor a givenAI2Oacontent,
andrelativelylow TiO2 contents;
and

(2) a lessmagnesian
tholeiiticsuite,lackingspinilextextureandhavinga higher

FeO/(FeO + MgO) ratio andTiO2content.

The Alaskan
complexes
differfromeachof theothereconomically
interesting
types

in the alkalicnatureof theirmagma(Irvine, 1974)andthereforethe totalabsence

of
orthopyroxene
in the rocks. Stratiformcomplexes
suchas Bushveldand Stillwaterare

at theotherendof thespectrum.Theyhavea highnormafive

orthopyroxene
to clinopyroxeneratio and crystallizein this order: olivine,orthopyroxene,
and plagioclase
plusaugiteor plagioclase
and thenaugite. Magmasresponsible
for someof the other
important
bodies
fall between
thesetwoendmembers,
although
theyare all tholeiiticand

containnormafiveorthopyroxene.

Platinum-group
element
concentrations
occuras discrete
platinum-group
mineralsor
as solidsolutions
in sulfides,
arsenides,
and sulfarsenides
associated
with synvolcanic

bodies,
stratiform
complexes,
andpicriticintrusions.AlpineandAlaskan-type
rocks
containonlyverylow gradeconcentrations
of sulfides,
butplatinum-group
element
concentrations
occuras placerdeposits
derivedfrom Alaskan-type
bodiesand,to a lesser
extent, from alpine bodies.

In the sulfideoresof the Merensky,

Pechenga,
Sudbury,
and Noril'skdeposits,
the
Pt/(Pt+ Pd) ratio decreases
systematically
with an increasing
Cu/(Cuq-Ni) ratio.
Sincethe latterratio is thoughtto increase
with progressive
differentiation
of the host
silicate
magma,
it is suggested
thatthetrendof thedecreasing
Pt/(Pt q-Pd) ratiois
alsorelatedto differentiation.
Oresassociated
withkomatiites
are majorexceptions
to

thetrendexhibited
by thetholeiites,
thekomatiite
oreshavingbotha low Cu/(Cuq-

Ni) anda low Pt/(Ptq- Pd) ratio. It is suggested

that this is dueto the fact that

komatiites
originate
at unusually
greatdepthsin the mantle,depthsat whichmantle
sulfides
(richin Pd) haveaccumulated
andarehenceincorporated
in themagmas.
Introduction

stresson magmaticprocesses
and, following a discussion of the classification scheme in which distinc-

SINCEplatinumgroupelements
areassociated
almost
are pointedout,
exclusively
with ultramafic
andmaficrocks,the aim tive featuresof the differentclasses
of this paper is to review the different classesof

the composition
and differentiation
trendsof magmas

theserocks,laying particularstresson thosethat responsiblefor many of the important sourcesof

are the sourceof concentrations
of platinum-groupplatinum-groupelementsare compared.
Data on the platinum-groupelementcontent of
elements.The primaryconcentration
of theseelementsis broughtaboutin two main ways,either different classes of ultramafic rocks have been comafter the rock has cooled,by weathering
and the piled from the literature. Data on a number of suldevelopment
of placers,or at the magmaticstage fide depositshave alsobeengatheredfrom the Iiterathrough the formation and concentrationof an im-

ture and throughthe kindnessof a numberof mining

misciblesulfidetacit. This paperlays particular companies. It is the hope that this review and corn1131

1132

A. J. NALDRETT

AND L. J. CABRI

TXBLE 1. Classification of Ultramafic and Related Mafic Bodies

Class

Examples

Remarks

A. Bodiesemplacedin active orogenicareas

1. Bodiescontemporaneous
with eugeosynclinal
volcanism
(i) Tholeiitic suite
(a) Picritic subtype
Munro-Dundonald

area

(Naldrett & Mason, 1968; McRae, 1969;

Arndt, 1975, 1976)

Kakagi Lake (Ridler, 1966)

Pechenga (Gorbunov, 1968)
(b) Anorthositic subtype
Dord Lake Complex (Allard, 1970)
Bell River Complex (Sharpe,1965)
Kamiskotia Complex,Timmins, Ont.

Gravity-differentiated flows,
cappedby hyaloclastitewith
13-15 wt% MgO (seeTable 2,
cols. 12 & 14). Gravity differentiated

sills.

Some examplesof this classare

conformable,others appear to be
discordant. They are possibly
differentiation

chambers for over-

lying volcanism.
(ii) Komatiitic suite

Munro-Dundonald

area, Ont.

(Naldrett & Mason, 1968; Pyke etal.,

1973; Arndt etal., in prep.)
Eastern Goldfields, Australia

(McCall & Leishman,1971; Nesbitt, 1971;

Hallberg & Williams, 1972; Williams, 1972;

Simple flows, spinilex-capped

flows, differentiated flows, and

differentiated sills. Composition
of flows ranges from peridotite to
basalt and of sills from dunire to

anorthositic gabbro.

Naldrett & Turner, 1977)

2. Alpine-type bodies
(i) Large obducted sheets
New Caledonia (Guillon, 1975)

Papua-NewGuinea (Davies,1968,1971)
(ii) Ophiolite complexes
Vourinos (Moores, 1969)

Troodos(Gass,1967, 1968; Moores& Vine,

1971)

Bay of Islands(Dewey& Bird, 1971;Irvine

& Findlay, 1972)

CanyonMountain (Thayer & Himmelberg,

1968)

EasternTownships,Quebec(Lamarche,1972;
Laurent, 1975)

(iii) Deformedophiolitecomplexes
andclasticblocksin mlangeterranes
Twin Sisters,Washington(Ragan, 1967)
Vermont serpentinites(Jahns,1967)
Iran (Gansser, 1959)

FranciscanSeries(Hamilton, 1969; Hsu, 1968)

(iv) Possiblediapirs

Mr. Albert, Quebec(MacGregor,1962)

3. Alaskan-type complexes
Intrusions of Alaska and British Columbia,

includingDuke Island (Irvine, 1974),Union

Bay (Ruckmick& Noble, 1959), Tulameen
(Findlay, 1969)
Intrusionsof Urals (Vorobeyevaet al., 1962)

B. Bodiesemplacedin non-orogenicareas
4. Large stratiformly layeredcomplexes
Bushveld(Hall, 1932; Wager & Brown, 1968;
Visser &von Gruenewaldt, 1970)

Stillwater (Hess, 1960; Jackson,1961; Bowes

et al., 1973)

Muskox (Irvine & Smith, 1967)

Duluth (Taylor, 1964; Phinney,1970)
Kiglapait (Morse, 1969)
Sudbury (Naldrett et al., 1970, 1972)
5. Sillsand sheetsequivalentto floodbasalts
Palisadessill (F. Walker, 1940; K. Walker,
1969)

Insizwa-Ingeliintrusion,SouthAfrica (Maske,
1966)

Dufek Intrusion,Antarctica (Ford & Boyd,

1968; Ford, 1970; Himmelberg & Ford,
1976)

Generally occur in areas in

which extrusion of flood basalts
has occurred. The sills are chem-

ically similar to the extruded

basalts.

ULTR.4M.4FIC

AND

REL,TED

31.4FIC

1133

ROCKS

T,BLE 1.-- (Continued)

Class
6. Medium-

Examples
and small-sized

Remarks

intrusions

Skaergaard (Wager & Brown, 1968)

Rhum (Wager & Brown, 1968)
Noril'sk-Talnakh area (Godlevskii, 1959;
Zolutuchin & Vasil'ev, 1967)

7. Alkalic ultramafic rocksin ring complexesand kimberlite pipes

pilation will provide a backgroundhelpful to the

reader of the other papersin this number.
In the courseof our compilation,we noted that,

grained pyroxenite capping the flows, or as vent

brecciasprobablyrelated to the sills. A typical sill
of this subclass,such as the Dundonald sill, consists
when viewed in relation to the nature of their host
of a basal,320-m-thicklayer of peridotiteoverlainby
rock and the Cu/Ni ratio of the ores in which they 130 m of augitepyroxenitecappedby up to 260 m
occur, platinum and palladium concentrationsex- of gabbro within which zones of granophyre up to
hibit variationswhich are explicableon the basisof 60 metersthick occurtoward the top (Naldrett and
a genetic mode which is discussed.
Mason, 1968). Average analysesof samplesfrom
these
zonesare given in Table 2. A similar stratigUltramafic
and Related
Mafic
Rocks:
raphy
is observedin the lower portion of flows of
Their Classification
and Some of the
this class,although in these the gabbro grades upMagma Types Involved
ward into a fine-grainedpyroxenite cappedby a
Ultramafic and related mafic rocks can be grouped pyroxenitehyaloclastite(Arndt, 1975, 1976). A pipeconvenientlyin terms of the tectonicsettingin which like body of hyaloclastite(see Ta'ble2 for composithey were eraplaced. Naldrett (1973) presenteda tion) cuts through komatiitic flows overlying the
tabulation of the major classesof ultramafic bodies Dundonald sill at the Alexo mine, brecciatingthem
alongtheselines. In Table 1 we presentan updated and merging with a stratiform hyaloclastitelayer
version of this scheme. It can be seen that the
which is itself cappedby further komatiitic flows,
initial subdivisionis betweenthosebodieseraplaced thus proving the contemporaneityof the two types
in an environmentthat is either actively undergoing of magmatism. The compositionsof hyaloclastites
or is shortly to 'becomeinvolved in orogenesisand of this type (Table 2, columns 12 and 14) are
those emplacedin non-orogenic,commonlystable thought to representthe initial compositionof the
cratonic areas.
magmaresponsible
for flows and sills They contain
10
to
20
percent
normative
olivine and can be deBodies eraplacedin an orogencsetting
scribedas olivine tholeiiteor picrite.
One can recognizethree broadtypesof ultramafic
Tholelites--subclass b (anorthositic): These
bodiesin this setting:thosethat are contemporane- bodiesconsistpredominantlyof anorthosite,gabbroic
ous with early "eugeosynclinal"volcanism; alpine anorthosite, or anorthositicgab'broand are characbodies,most of which are eraplacedduring the com- terized by layers rich in titaniferous magnetite.
pressivephaseof orogenesis;and Alaskan-typeultra- Ultramafic rocks are relatively uncommonand are
marlcsemplacedat a late stage,largely during nplift. usually restrictedto layers rich in cumuluspyroxenite. One of the most completelystudied,although
1. S'ynvolcancbodies
most altered,examplesof this classis the Dor Lake
Experience in Archean and Proterozoic greenComplex (Allard, 1970; Allard et al., 1972). The
stone belts has shown that synvolcanicultramafic
general conformityof the layering in the body to
rocks f.ormparts of bodieswhich fall into two main
folding in the enclosingrocks suggeststhat it was
classes,tholelites and komatiites, and that the thointrudedprior to olding.However, althoughit is
leiites constitutetwo distinctive subtypes,(a) one
poorly exposed,the Kamiskotia complex, 24 km
picritic and (b) the other anorthositic. Since the
west of Timmins, Ontario, seems to be more oi a
anorthositicsubtypecontainsa very small propordiscordantplug.
tion of ultramaficrocks,we will focuslargely on the
Komatiites:
Some confusion has arisen over exkomatiiticand picritic tholeiitictypes.
actly
what
constitntes
a komatiite. The name was
Tholelites--subclass
a (picritic): Ultramaficrocks
coined
by
Viljoen
and
Viljoen
(1969) ftr r(ckswell
lelongingto this class occur as basal accumulations
to differentiated sills or flows. They also occur developedin the Komati River area of the Barber-

either as zonesof hyaloclastite,

or aphaniticor fine- ton Mountain
land. Certainaspects
of theirinitial

1134

.4. J. N.4LDRETT .4ND L. J. C,4BRI

definitionand in particularthe interpretationof it monly 20 to 30, weight percentMgO and alsocarryformulatedby Brooksand Hart (19741 haveproved ing 20 to 30 percentolivine phenocrystsin suspento be somewhat too restrictive and Arndt et al.
sion. Settlingand accumulation
of the phenocrysts,
(in prep.) haveproposeda redefinitionof the koma- coupled with filter pressing,compaction,and postiitic igneoussuite,maintainingthat it is as distinc- sibly some adcumu!usgrowth, are responsiblefor
tive and deservingof the samestatusas the tholeiitic, the peridotitesand dunites (Naldrett and Turner,
calcalkalic, or alkalic suites.
19771.
In someflows (and perhapsin somenear-surface
Membersof the komatiiticsuiterange from dunite
(>40 wt percentMgO calculatedon an anhydrous intrusions),platy and skeletalgrowthof olivineand
clinopyroxenecrystals characterizesthe overlying
basis) throughperidotite (30 to 40 percentMgO),
have settledand is
pyroxeneperidotite(20 to 30 percentMgO), pyrox- melt from whichthe phenocrysts
enite (12 to 20 percentMgO), and magnesianbasalt known as spinilextexture (Pyke et al., 19731. In
(10 to 12 percent MgO) to basalt (<10 percent other flows, the settling of phenocrystsappearsto
MgO). Typical analysesfor eachof thesetypes are have been retarded,in which casea zone of spinifex
given in Table 2. Both extrusive and intrusive
variantsare usuallypresentin a given area with the
former predominant. The ultramafic mem'bersof
the suite are interpreted as the result of emplacemerit of a liquid containingup to 35, but more corn-

texture is lacking, although skeletal overgrowthsto

phenocrystsand interstitial skeletal pyroxene are

present. Becauseof the texture,whichis suggestive

of fairly rapidcooling,zonesof spinifexare regarded
as approximatingthe liquid portion of a given

Txm, E 2. Typical Analysesof Ultramafic and Related Mafic Rocks (Recalculatedto eliminate HOq-CO.q-S?
or loss on ignition and with P20 omitted.)
Hyaloclastites

Typical komatiites
'1

SiO
AhOa

TiO
FeOs
FeO
MnO

MgO
CaO
Na20

K20
Cr,-Os
Total
Fe as FeO

40.2
0.29
0.02
7.89
n.d.
0.39
51.6
0.04
0.01
<0.0l
n.d.
n.d.
100.44
7.10

Dundonald sill

10

11

12

13

14

44.0
5,27
0.27
12.2
n.d.
0.22
32.5
5.48
0.32
i).04
n.d.
n,d.
100.30
10.95

47.6
6.19
I).33
10,4
n.d.
0.22
27.7
6.82
I).19
0 01
n.d.
n.d.
99.46
9.38

48.9
10.3
0.5
12.3
n.d.
0.23
16.5
9.78
1.18
0.13
n.d.
n.d.
99.82
11.07

51.0
12.6
0.59
12.6
n.d.
0.4
10.68
11).54
2.35
0.13
n.d.
n.d.
100.70
11.31

52.2
14.8
0.80
11.52
n.d.
0.2
7.95
11).17
2.44
0.28
n.d.
n.d.
100,36
10.36

42.0
1.l t
0.20
15.5
n.d.
I}.23
40.4
I).46
0.0
0.0
1.11/
n.d.
1t}1.00
13.93

52.2
3.11
0.46
1.19
6.71
0.17
16.0
17.4
0.52
0.0
0.43
n.d.
98.19
7.78

48.7
13.1
1.85
4.46
11.9
0.22
5.63
10.5
2.98
i}.4,t
0.01
n.d.
09.78
15.92

55.4
13.7
1.58
2.32
13.4
I).22
3.13
5.62
4.ll
i).36
0.03
n.d.
99,87
15.49

64.7
15.1
0.92
0.81
5.2l
0.13
2.47
4.64
5.18
i).69
0.04
n.d.
99.80
5.94

50.0
9.36
1.15
15.6
n.d.
0.27
1,t.3
9.46
2.16
0.09
0,13
n.d.
101.52
14.04

51.9
9.77
1.17
14.63
n.d.
0.2l
9.36
10.5
3,95
0.03
n.d.
n.d.
101.52
13.17

50.3
7.93
0.07
16.4
n.d.
0.25
15.2
10.2
1.00
0.25
n.d.
n.d.
102.5(}
14.78

Pechenga

SiO=
AIsO,
TiO-,
FeOa
FeO
MnO

MgO
CaO
Na=O
K20
CrOs

P=O
Total
ZFe as FeO

Noril'sk

15

16

17

18

19

41.1
3.52
1.30
1.67
13.5
0.17
35.3
1.83
0.13
0.17
n.d.
n.d.
98.69
15.0

45.3
5.11
2.43
1.7,t
15.1
0.19
12.05
17.4
0.39
0.50
n.d.
n.d.
100.20
16.66

46.3
12.5
5.36
2.19
13.3
0.14
6.18
10.1
2.30
1.81
n.d.
n.d.
100.18
15.27

54.5
15.(}
2.14
2.81
13.6
0.20
2.42
6.23
4.05
2.54
n.d.
n.d.
103.49
16.13

49.2
15.8
1.66
3.52
8.62
(}.19
6.91
10.6
2.41
0.54
0.03
0.29
99.77
11.79

20

45.1
16.0
0.93
4.09
9.84
0.20
12.5
9.42
1.29
0.78
0.52
0.20
100.87
13.52

21

41.3
10. i
0.63
2.64
11.9
0.20
24.5
6.66
0.74
0.38
1.07
0.12
100.24
14.28

Sudbury
22

23

24

25

26

27

28

44.9
14.9
0.65
2.28
9.54
0.19
16.8
8.79
1.38
0.31
0.04
0.14
99.92
11.59

48.5
16.2
1.12
2.91
7.88
0.17
8.67
11.9
1.87
0.39
0.03
0.12
99.76
10.50

50.7
13.8
1.86
5.43
8.34
0.20
5.59
10.7
2.37
0.49
n.d.
0.48
99.96
13.23

55.3
18.6
I}.52
(}.79
6.06
0.11
6.47
8.02
2.64
0.98
n.d.
0.19
99.68
6.77

56.9
16,0
0.66
3.27
6.24
0.10
4.28
7.96
3.23
1.43
n.d.
0.24
100.31
9.18

51.7
15.5
2.11
3.12
9.28
0.11
4.64
8.01
2.80
1.38
n.d.
1.21
99.86
12.09

61.8
14.2
I}.07
1.78
7.81
0.18
1.58
3.76
3.09
3.25
n.d.
0.42
98.84
9.41

Bushveld

SiO=
Ab. Oa
TiO_
Fe=Os
FeO
MnO

MgO
CaO
Na20
K=O
Cr=O,
P=O
Total
,Fe as FeO

and

related rocks

Stilhvater

29

30

31

32

33

34

35

36

37

38

39

40

43.8
1.27
0.27
1.40
11.3
0.11
36.5
1.43
0.21
0.32
3.40
tr
101.01
12.56

55.5
2.05
0.10
0.15
9.68
0.15
31.0
0.97
0.10
tr
0.63
0.2
100.53
9.82

52.7
5.31
0.21
2.31
11.9
0.16
22.1
5.42
0.64
0.16
0.16
0.09
101.16
13.98

52.5
17.0
0.22
0.68
6.78
0.14
9.06
11.6
1.89
0.37
0.00
0.11
100.35
7.39

56.6
13.3
1.75
5.42
9.29
0.26
0.57
6.35
3.51
2.60
0.00
0.36
100.01
14.17

62.4
13.2
0.95
3.34
7.06
0.15
0.43
4.50
3.76
3.15
0.00
0.2
100.04
10.97

40.1
2.64
0.12
1.48
12.0
0.19
39.1
1.40
0.19
0.03
2.43
0.01
99.69
13.33

49.02
4.4l
0.10
1.23
9.94
0.18
30.5
3.05
0.41
0.02
0.74
0.01
99.61
11.05

55.1
6.27
0.08
0.24
7.91
0.18
23.1
5.52
0.44
0.02
0.70
0.01
99.57
8.13

51.7
14.2
0.14
0.50
6.59
0.13
16.7
9.12
0.84
0.01
0.23
0.00
100.16
7.04

50.3
19.8
0.14
0.21
4.40
0.09
10.7
13.1
1.01
0.01
0.19
0.00
99.95
4.59

50.6
17.2
0.19
I).23
5.02
0.12
11.3
14.0
1.18
0.08
0.13
0.01
100.06
5.23

ULTR.4MAFIC

AND

REL.4TD

MAFIC

1135

ROCKS

TABZE 2-- (Continued)

Explanation
Rock & location
1. Dunitie komatiite
2. Peridotitic komatiite

(spinifex texture)

3. Pyroxene peridotitic konatiite (spinifex texture)

4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.

Pyroxenitic komatiite
Magnesian basaltic komatiite
Basaltic komatiite
Basal peridotite, Dundonald sill
Augite cumulate, Dundonald sill
Oide-rich gabbro, Dundonald sill
Nornal gabbro, Dundonald sill
Granophyric gabbro, Dundonald sill
Hyaloclastite, Theo's flow
Aphanitic pyroxenite. Theo's flow
Hyaloclastite, Alexo area
Peridotite, Pechenga area
Pyroxenite, Pechenga area
Melagabbro, Pechenga area
Gabbro, Pechenga area
Contact gabbro (lower), Noril'sk II
Taxitic gabbro, Noril'sk II
Picritic gabbro, Noril'sk II
Olivine-biotite gabbro, Noril'sk II
Olivine gabbro, Noril'sk II
Gabbro-diorite, Noril'sk II
South Range noritc, Sudbury
Felsic norire, Sudbur

Average of analyses

Reference

234, 253, 272

106, 108

Naldrett
Naldrett

59, 170, 171

Naldrett & Turner (1977

55, 159, 160, 166, 180

116, 139, 177
154, 157, 162, 163, 168
204. 319. 334, 335, 340,436
306, 307, 313, 314, 318
301, 308, 316, 317
304, 305, 309, 31 l, 312. 315
302, 303, 310
P9-156
P9-158

& Turner (1977)

& Turner (1977)

Naldrett & Turner (1977)

Naldrett & Turner (1977)
Naldrett & Turner (1977)
Naldrett & Mason (1968)
Naldrett & Mason (1968)
Naldrett & Mason (1968)
Naldrett & Mason (1968)
Naldrett & Mason (1968)
Arndt (19i5)
Arndt (19i5)
Naldrett, unpublished data
Gorbunov (1968), table 8
Gorbunov (1968), table I0
Gorbunov (1968), table I 1
Gorbunov (1968), table 11
Zolutuchin & Vasil'ev (1967),
Zolutuchin & Vasil'ev (1967),
Zolutuchin & Vasil'ev (1967),
Zolutuchin & Vasil'ev (1967),
Zolutuchin & Vasil'ev (1967),
Zolutuchin & Vasil'ev (1967),
Collins (1934)
Collins (1934)

14, 17, 18, 21. 25

64, 65, 66
68, 69, 70
71, 72, 73

1-4
5-11
12-14

30. Bronzitite, Bushveld

31. Merensky Reef, Bushveld

15-18
19-29
33-35
89W, 88W, 87W
50W. 31C, 3(1C
51W, 52W, 53W, 27C
26C, 25C
IV
I, II, III
I, VIII

32. Norite, Bushyeld

33. Ferrogabbro, Bushyeld

34. Granophyric gabbro, Bushveld

35. Dunire, Stillwater

I to V
I

II, III, IV
25, 26, 27

Hall (1932, p. 304), table 28

Hall (1932, p. 307), table 29

36. Harzburgite, Stillwater

28, 29, 30, 31, 32

Boweset al. (1973)

37.
38.
39.
40.

33
41, 42, 43
35, 36, 38, 40
37, 38

Bowes et al,
Bowes et al.
Bowes et al.
Bowes et al.

27. Oxide-rich gabbro, lower part, Sudbury

28. Oxide-rich gabbro, upper part. Sudburs'
29. Harzburgite, Bushveld

Bronzitite, Stillwater
Marie noritc, Stillwater
Noritc, Stillxvater
Gabbro, Stillwater

table 8 '
table 8
table 8

table 8
table 8
table 8

Collins (1934)
Collins (1934)
Hall (1932, p. 327). table 37
Hall (1932, p. 315). table 32
Hall (1932, p. 319), table 33

Hall (1932. p. 307). table 29

Bowes et al. (1973)
(1973)
(1973)
(1973)
(1973)

magma. Using this criterion,flows with spinilex this is done for the Munro rocks and for cumulate
containingabout 23 weight percentMgO would rocks of the Dundonald sill. The distinction that the
havebeenextrudedat around1,425C(see Shima diagramgivesbetweenthe two magmatypesis very

andNaldrett,1975)andthosewithup to 35 percent clear; komatiites plot above the sloping line and
tholeiites below it. Arndt et al. point out that
MgO at around1,650C(Greenet al., 1975).
The pyroxenitic and basaltic members of the

komatiiticsuiteappearto be the resultof fractional

crystallization
of a moremagnesian
komatiiteliquid.
The pyroxenites
are characterized
by a varietyof
spinilextexturesmadeup of highlyacicular(almost
needle-like)skeletalgrainsof clinopyroxene.
Skeletal texturesare eitherpoorlydeveloped
or lacking
in the basalts.

anotherdistinguishing
feature is the relativelylow
TiO2 contentof komatiiteswith respectto tholeiites,
as illustratedin the TiOe-MgO plot of Figure 2A.
Althoughthe cumulusperidotitesandpyroxenitesof
the Dundonaldsill plot within the komatiite field,
the diagram is very effectivein distinguishingthe
gabbroicvariants of the sill, and also extrusive tho-

leiites of undifferentiatedand differentiatedtype,

Arndt (1975, 1976) describesa differentiatedfrom

the komatiites.

komatiiticflow, 120 m thick. This consistsof a basal

If komatiitesin other parts of the world are conzone of cumulusperidotiteoverlainby pyroxenite sidered,those describedby Naldrett and Turner
(dominantlyaugite,althoughsomehypersthene
is (1977) from the Yakabindiearea of Western Auspresent)and thengabbrowhichgradesupwardinto tralia exhibit a similar behavior to the Munro rocks
a quartz gabbro. The flow is cappedby a 30-m- on both the AIO vs FeO/(FeO + MgO) (Fig.
thickzoneof spinilex-textured
olivineandpyroxene. lB) and TiO2 vs MgO (Fig. 2B) plots. Naldrett

As Arndt et al. (in prep.) emphasize,

thisflowis im-

and

Turner

have

also demonstrated

that

basaltic

portant becauseit provides a direct link between komatiitesat Yakabindie are characterizedby high
ultramafic and basaltic komatiites.
contentsof Ni (average 146 ppm, range 100 to 181
Arndt et al. (in prep.) pointout that komatiitesin ppm) and Cr203 (average417 ppm, range 210 to
Munro Township, Ontario, can be distinguished 800 ppm) in comparisonwith tholeiites from the
conveniently
from tholelitesby plottingAlcOaagainst samearea (averageNi = 76, range25 to 127 ppm;
the FeO/(FeO + MgO) content of the rock, in averageCr208 -- 138,range< 10 to 336 ppm). Analwhichtotal iron is calculatedas FeO. In Figure 1A yses reported by Nesbitt (1971) from the Scotia

1136

.4. J. N.4LDRETT

,4ND L. J. C.4BRI

( 12

."_';'
oo PICRITIC
THOLElITE
.'.' ?o.
+*

oTheo's
KOM&TIITE

. * *

02

0.

,
05

,
o

0.7

0.

0.9

+ PerdohhcKomatutes(Av.)

+GelS)'

Barberton

Gel(O+ Basalhc Komohite (Av.)

Barberton-

."B,
/ +Bad

Geluk high AI

Gel(d)+ Bosolhc Komotdte (Av.)

* S,mple
flows,
Munro
Twp

0.4

tHtes-Yakablnde

Koreabites
-- Scoha

Z +Gel(u)
o ".t---

Barberton--Geluk

low AI

Bad+ Basalhc Komatiite Bodplaos

Barb+ 8asalhc Komatute, Barberton

,
0

FeO/(FeO + MgO)

0-1

02

0.3

04

05

o.6

o.7

OB

0.9

.0

FeO/(FeO + MgO)

Fro. 1. Variationof A12Oa

with FeO/(FeO-FMgO) wt percentratio in rocks{rom: A,

theAbitibigreenstone
belt,andB, selected
areasin SouthAfricaandWesternAustralia.

Notethat totalFe in the rockis calculated

as FeO andall analyses
are for rocksrecalculated

to exclude HO and CO or loss on ignition.

of deep-sea
sediments
including
radiolarian
area of Western Australia (80 km north of Kal- capping
goorlie)plot togetherwith thosefrom Yakabindie. cherts.
In someplaceseraplacement
has been accomHowever, in contrastto rocksfrom CanadaandWestplished
with
a
minimum
of
disruption
andthe full
ern Australia, some of the Barberton komatiites,
sequence
is well preserved
but in others
in particularthe basaltickomatiitesof the Badplaas ophiolitic
and Barbertontype and the low-A1variant of the the ultramafic rocks occur as isolated blocks within
the Geluktype,plot belowthe slopingline in Figure a mlange.
1A. Arndt (1976) discussesthe unusual com3..41askan-typeultramaficcomplexes
positionsof these rocks, at presentknown to occur
only in the Barbertonarea, and suggestssomereasons for them.

Complexes
of this typedrawtheir namefroma
seriesof distinctive
bodiesexposed
alongthe interior

A word of cautionis requiredat this stageto the islands and mainland of southeast Alaska. More
effectthat because
a givenrockpassesthe chemical than30 occuralongthe550-kinlengthof theAlaskan
mostof themin a zoneabout40 km
screensoutlinedhere,it is not necessarily
komatiitic. panhandle,
have outcrop
In addition to the chemical criteria, other features wide. The eight largestcomplexes
areas
ranging
from
5
to
30
square
kin.
Someof the
that we have mentioned,such as the presenceof

showa roughconcentric
zoning
lavas,possiblesubvolcanic
feedersrich in olivine, largercomplexes
consists
of a dunitic
and the occurrence
of spinilextexture within some which,whenbestdeveloped,
by successive
shellsof olivineclinorepresentative
samples,shouldbe presentto ensure coresurrounded
pyroxenite,
magnetite-rich
clinopyroxenite,
andhornwith reasonable
certaintythat one is lookingat a
blendite. As a group,Alaskan-typecomplexes
are
suite of komatiites.
2. ,41pineullramafic rocks

distinguished
fromalpineultramafics
or stratiform
intrusionsin havinga highly calcicclinopyroxene,

no orthopyroxene
or plagioclase,
abundanthornblende,moreiron-richchromite,
andmagnetite,
the
last
mineral
occurring
in
concentrations
which
in
manyauthorities
that alpinebodiesare portionsof
ophiolite
complexes,
eraplaced
in thesolidstatedur- some cases are of economic interest.
ingtectonism.Ophiolites
themselves
areinterpreted Similar bodies are known in the Urals, southas portionsof the oceaniccrustand uppermantle centralBritish Columbia,and Venezuela(see Table
studied
istheDukeIsland
andarecharacterized
by a basalzoneof metamorphic- 1). Themostextensively
described
by Irvine(1974),whoalsoprotexturedperidotiteoverlainby a sequence
of cumu- Complex,
of bodies
latesrangingfrom peridotiteto gabbro,a sequence videsthe mostup4o-dategeneralaccount
of
this
type.
of pillowlavasand associated
feederdikes,and a
The differenttypesof alpineultramaficbodiesare

listed in Table 1. Here we take the view held by

ULTR./tM.4FIC

AND

RELATED

Bodieseraplacedin a nonorot7enic
environment

M.4FIC

ROCKS

1137

25

Apart from distinguishing

intrusionsof plateau
basaltmagmaand alkalic intrusions,the classification of bodiesin a nonorogenicenvironment,as

A
2.0

THOLElITES

n Dundonald
Sill

shown in Table 1, is more a matter of size rather

P, Theo'sFlow

than of fundamental
differencein type of intrusion
or magma. Furthermore,there is a gray zone between what one might regardas a nonorgenicand
an orogenicenvironment. For example,Hunter
(1974) and Naldrettand Turner (1977) haverelatedkomatiitic
volcanism
to riftingof a preexisting
sialic plate. In these two casesthere is no evidence

o Basalts

KOMATIITES

() Simpleflows,Munro

+ Fred's
Flow
(differentiated)

O5

that the rifting proceeded

far enoughto form fresh

oceanic
crust. Rifting was,however,succeeded
by
compressive
foldingof the rockseraplacedin the
postulatedrift zones,whichis why we haveclassed
the ultramaficbodiesthemselvesin the orogenic

co

group.

It is a debatable
pointwhetherthe tectonicsetting
of the Archean komatiites at the time of their era-

2.5

placement
differedfundamentally
from,for example,
the settingof the sequence
of Permo-Triassicvolcanics and intrusions of the Noril'sk area which are

..
+

Komotiites

Yokobindie

Komotiites

Scotia

o Komot.tes
o Tholeiifes

20

locatedwithina major graben(Stairnov,1966);

Barberton (overages)
Yakabndie

o
o

from the Duluthcomplexand Keewanawan

lavasof

Oo

the Lake Superior area which are related to the

North Americanmid-continental
rift zone; or from
the lavas and intrusions of East Greenland and West

Scotlandflankingand relatedto the rifting of the

North Atlantic. All that is lackingin thesethree

0/

casesis the final compression.

In drawingup the classification,

we haveplaced
importanceon the existenceor absenceof the final

e ' .. .-/'

0 .5-

// . . ,,o,,

compresslye
phase,although
we realizethatthedeepseatedtectonicprocesses
responsible
for the rifting
of the unstable Archean crust onto which the koma-

50

tiiteswereextruded
andthoseresponsible
for rifting,
evenincipient
riftingof younger
cratons,
maynotbe
so very different. In a later section,differences
or

40

.. . +'..'

. /

30

20

I0

Wt. percent MgO

Fro. 2. Wt percent TiO vs wt percent MgO for: A,

B, selected

rocks from the Abitibi greenstonebelt, and

similarities
in thecomposition
of themagma
respon- areas
in South Africa and Western Australia.

sible for the different classesof ultramafic rocks are

discussed.

4. Lar#e stratiformcomplexes

5. Sills of plateau(flood) basalt

Bodiesof this classincludethe very largeDufek
Intrusionin Antarctica(6 to 7 km thick,morethan

8,000km2 in area) and the InsizwaComplexof

As shownin Table 1, this classincludesbodies SouthAfrica (Table 1). They are associated
with

suchas the Bushveld,

Stillwater,Muskox,andSud- tholeiiticbasaltsof the plateautypewhichare charburycomplexes.The overallcompositions
of these acterizedby a higher K20 content than other thocomplexes
tend to be maficrather than ultramafic, leiitic types. Even mafic rocks from the intrusions
althoughultramaficzonesare presentin eachof exhibita ratherhigh(1 wt percent)KuOcontent.
themexceptSudbury.Evenat Sudbury,numerous
ultramaficinclusions
rangingfrom duniteto olivine

6. and 7.

Other classes

melagabbro
occurin the ore (Rae, 1975) and are
These includenedium-and small-sizedintrusions
thought
to havebeenderived
froma hidden
layered and alkalicrockssuchas kimberlitesand carbonatite
intrusion(Naldrettet al., 1972).
complexes.

1138

.4. J. NALDRETT

AND L. J. CABRI

T^BLE 3. Field Criteria for Recognition of Ultramafic Rocks

as Komatiites
Criteria
1

Sudbury

Bushveld
Noril'sk
Pechenga
Western Australia

+
dd-

+
+
+

Abitibi (Marbridge, Alex,

Langmuir)
Ungava nickel belt
Manitoba nickel belt

+
d-

1. Should be part of a sequenceof lavas and shallow intrusions.

2. Some representativesshould be allvine rich.

3. Spinifex texture is present.
+

= Criteria fulfilled.

= Criteria

not fulfilled.

The nature o] the magma involved in bodiesimportant with respectto nickel sulfide and platinumgroup elementconcentrations
The bodies to be discussed in this section include

be considered. In the Bushveld, Stillwater, and

Sudburyareasdirectlyassociated
volcanicsare lack-

ing, as is spinilextexture; despitethe presenceof

ultramaficrocksassociated
with eachbody,the average compositionof the exposedsequenceof rock is
not ultramafic, so that we have excluded these com-

plexesfrom further consideration

as komatiitic. The
Alaskan-typebodiesfulfill the criteria of the presence
of olivine-richrocks,overallultramaficcomposition,
and, in somecases,of relatedvolcanicrocks (Irvine,
1974). As will be shownlater, however,the magma
involvedappearsto havebeendistinctlyalkalic(critically undersaturated),and the possiblekomatiitic
nature of these rocks will not be considered further.

Figure 3 illustratesrocksfrom sevenof the districts (E.asternGoldfields,Abitibi, Manitoba, Ungava, Pechenga,Noril'sk, and Shangani) discussed
aboveon an AlcOavs FeO/(FeO + MgO) diagram,
while Figure 4 illustratesthe variationin TiO2 with
MgO. In the latter diagram,the field of the Munro
and Yakabindiekomatiites (taken from Fig. 2) is
outlinedby the dashedline. Only rocksassociated
with nickel mineralization and exclusive of the Yakabindie and Munro areas are shown for the Eastern

host rocks to nickel sulfidesin (1) the Eastern

Goldfieldsof WesternAustralia,(2) the Abitibi belt Goldfieldsand Abitibi belt; theseplot abovethe line
of Canada, (3) the Manitoba nickel belt, (4) the in Figure 3 and within the dashedoutlineof Figure

Cape Smith-WakehamBay belt in Ungava, (5) 4, confirmingtheir komatiiticnature. Analysesof

fromthe minesof the Thompson
areaof the
the Sudburydistrict, (6) the Pechengaarea of the samples
Kola Peninsula,(7) the Noril'sk districtof Western Manitoba belt are not availablefor publication,but
Siberia,and (8) the Shanganimine,Rhodesia,and Stephenson(1972) presentsanalysesof ultramafic
host rocksto platinum-groupelementconcentrations and mafic rocks of the belt in which the most ultrain (9) the BushveldComplex,(10) the Stillwater mafic variants are essentiallyidentical, except for
their lack of sulfides,with ultramaficrocks associComplex,and (11 ) Alaskan-typecomplexes.
We will first examine the rocks from .both oro-

genicareasto seeif they appearto be komatiitesand

thennotevariationsin their pathof fractionalcrystallization which bear on the nature of the initial

ABtTIBI

T - Texmon Mne
16

14

Possible komatiitic

L- Longmuir
Deposits
Mme

nature

0 0

A -Alexo

M - Marbridge Mine
E GOLDFELDS

magma.

K - Kombaldo

S-

SpargovlHe

R- Ravensthorpe

- Momtobo Nckel Belt

As statedabove,field criteria for komatiitic mag-

U - Ungova

+- Pechenga '

matisminclude(1) the presence

of extrusiverocks
withinan igneoussuite,(2) the presence
of olivinerich rockswithinthe suite,and (3) the presence
of

0 - NOrll'sk

spinilextextures. Field characteristics

bearingon
the possible
komatiiticnatureof rocksfrom eight
of the areas listed above are summarized in
Table 3. Rocks from the Eastern Goldfields,

A A u u

Abitibi belt, and Manitoba nickel belt (W. V.

Peredery,pers. commun.,1975, has reportedthe

presence
of spiniex-textured
rocksin two localities
withinthis'belt)meetall threecriteria. Thosefrom
Ungava,Pechenga,
Noril'skandShangani
fulfilltwo

/
o

o4

o6

07

08

0'9

FeO/(FeO + MgO)

3. Variation of Al.oOwith FeO/(FeO + MgO) for

of the criteria but lack spinilextextures. Since theFro.
host rocks of certain nickel sulfide orebodies,excluding
spinilextextureis not an absolute
requirement,
the thoseplottedin Figure 1. The methodof presentingthe

possibility
thatthe rocksare komatiites
stillhasto

analysesis the same as in Figure 1.

ULTR.4M.4FIC

4ND REL.4TED

M.4FIC ROCKS

1139

TaaLE 4. Sequenceof Cumulus Phases

ated with the ores. These analysesalso plot above

the line in Figure 3 and either within or very close

to the outlinedfield in Figure 4, reinforcingthe case

Komatiites

ol, ol +cpx (+opx),

Dundonald sill

ol, cpx, plag+cpx-t-mt,

for the existence of komatiites

Pechenga

ol, cpx, cpx -t-mr, plag -t-cpx -J-mt

Noril'sk

not known

in this belt.

The

cpx +plag (+ol)

plag-t-cpx(-t-hyp?)

analysesof Wilson et al. (1969) of ore-associated Bushveld

opx -J-ol. opx -J-ol+chr, plag -t-opx -t-chr -J-cpx( -t-oi),
1)lag-t-cpx (d-opx), plag -t-cpx d-mt, plag -t-cpx -t-mr -t-ol
rocksfrom Ungava plot abovethe line in Figure 3
Stillwater
ol, ol -t-opx, ol -t-opx-t-chr, plag +opx, plag -J-cpx-t-opx,
and abovebut closeto the outlinedfield in Figure 4.
1)lag+cpx -t- (opx d-ol). plag -t-cpx d-opx
plag -J-opx,plag -t-opx-J-cpx,plag -t-cpx, plag d-cpxd-rnt
In this casealso,we are confidentthat theserepresent Sudbury
ol d-chr, ol -J-cpx, cpx -t-mt, cpx -t-mt +plag
a seriesof komatiites. Viljoen et al. (1976) give Duke Island
two average analysesof rocks from the ultramafic
host rock at Shangani which also fulfill all of the magnetite,ilmenite, apatite, anorthite, albite, and
chemicalcriteria necessaryfor them to be classified orthoclase have been subtracted. In essence the
as komatiitic.
diagramis an expanded,iron-bearingversionof the
Average analysesof Pechengaperidotite,pyrox- diopside-forsterite-silica
system.
enite,gabbro,and leucogabbro
(Table 2) calculated Figure 5A illustratesaverage analysesof the
from the data of Gorbunov(1968) mostlyplot well Bushveld,Stillwater,and Sudburycomplexes
plotted
below the line in Figure 3, while their extremely on the diagram; the phaserelationsare thoseestihigh TiO2 contentsare brought out in Figure 4, matedby Irvine; they are includedas referencelines
indicatingthat thesebodiesare far from beingkoma- and are not necessarilydirectly applicableto the
tiitic. Averageanalysesof rocksfrom the Noril'sk complexesplottedon this or other diagramsof this
II deposit(listed in Table 2) plot abovethe line in sequence.The sequenceof principalcumulusminFigure 3, but Figure 4 illustratestheir relatively erals is listed in Table 4. It must be stressed at this
high TiO2 content, ruling them out in our viexv stagethat crystallization
hasbeena complexprocess

from consideration

as komatiites.

in eachof theseintrusions,and numerouscyclical

units are present in both Bushveldand Stillwater,

Cao.5( MgFe) o.0- ( MgFe) O-ociO2variation

resultingperhapsfrom repetitivecrystallization
due
to freshinfluxesof magmaor the periodicconvective

Irvine (1974) has pointed out the usefulnessof

looking at both cumulus and non-cumulusrock as overturnof existingmagma.
they plot on the Ca0.(MgFe)0.50-(MgFe)O-SiO2
Considering Bushveld and Stillwater first, the
molecular diagram once the cation norm has been positioningof the rock averageson Figure 5A,
calculatedand the proportionsof elementspresentin coupledwith the informationon cumulusphasesin
30+

ABJTIBI

T - Texmonf

Mine

L -- Langmmr Deposits
, - Alexo

25

Mine

M - Marbridge Mine
E GDLDFIELDS
K -- Karnbalda

s - Spargowlle
R -- Ravensthorpe

2O

ManHoba

Nickel

Bell

u - Ungava
+ - Pechenga
o - Norll'sk
o-

Table 4, indicatesan orderlycrystallization

sequence
from peridotitcto bronzititeto norite and then gabbro. This is represented
approximately
by a liquid
fractionatingalong the paths 1 (peridotite), 2
(bronzitite,norite), and 3 (gabbro), althoughsome
modificationof this liquid path and henceof the
phaserelationsindicatedis necessary
to accountfor
the co-precipitation
of olivineand orthopyroxenereported in each intrusion. Although Naldrett et al.

(1970) have demonstrated

that crypticvariationis
presentin the SudburyNickel Irruptive and that the
samesequenceof cumulusphasesis presentin all
exposedportionsof the intrusions,the positioning
of the averagesof the Sudburyrocksshowsno sys-

Shangam

tematic variation.
o //
//

0.5

u u u / .....

50

40

/---o

' J/

/11

o//

u TM '
u

30

20

10

Wt. percent Mg0

Fro. 4. Wt percent TiO vs wt percent gO

rks

for the host

of certain nickel sulfide orebodies.

This is attributed to the altered

nature of the rocksat Sudbury and to the importancethat contamination

hasplayedin their development.

As demonstratedby Irvine (1974), the Duke

Island rocks exhibit a very differenttrend. Their
fractionation
is from duniteandperidotite(area I in
Fig. 5B) to olivineand hornblende
pyroxenite(II)
to hornblendite(III), with the liquidmovingaway
from olivine and then away from silica along the
alkalicfractionationtrend 1-2-3 in Figure 5B.

1140

A. J. N./ILDRETT AND L. J. CABRI

B / ' Mole
percen

(MgFe)O

Ol

opxs soln

SO2

(MgFe)O

ol

Cao.5(MgFe)0.50

C
/

(MgFe)O

01

opxssoln

SlO2

Cao 5(MgFe)0.50

opxs$oln

D /
Mole
percent
/

el;u
'd'
':Lpositions)

S102

(MgFe)O

01

Mole
perce
MO

opx$$oln

Norfl'$k

Si02

Cao.5(MgFe)0.50

Cao.5(MgFe)0.50

;:':u::;;:
Z

(MgFe)O

01

opxs son

SO2

(MqFe)O

01

opxs soln

-- E Pechengo

SIO?

ULTRM,,IFIC
The komatiitic

fractionation

A'D REL./ITED 3IFIC

trend is illustrated

in

ROCKS

1141

donaldsill(VI) providessupportfor our contention

Figure 5C in which a seriesof liquid compositions, of a cogenetic

origin of the magmafor flowsand sills
estimatedfor the Yakabindie rocks by Naldrett and of this type.
Table 4 showsthat the Pechengarocks consistof
Turner (1977), are plotted. As expected,the initial
liquid plots much closer to olivine than that esti- a sequenceof cumulusphasessimilar to those of
mated for other intrusions. The fractionation trend
the Dundonald sill, although Figure 5E demonindicatedon the diagram is one of olivine removal stratesthat the pyroxenites(B) and melagabbros(C)
followedby removal of olivine plus clinopyroxene, are distinctlypoorerin SiO2 whenprojectedin this
consistentwith the mineralogyobservedin the flows way than their Dundonald equivalents. The
and given in Ta'ble4.
chengasequencecan be explainedby a liquid fracAverage rock compositionsfrom Noril'sk (Table tionatingalong the path a-b-c.
2) are plottedin Figure 5D. Smirnov(1966) argues
Figure 5F providesa comparison
of all the liquid
that the Noril'sk igneous bodies are feeders for fractionation trends that we have estimated. The
volcanoesand that the rocks present have not dif- Bushveld and Stillwater trends lie at one end of the
ferentiated in situ but are the result of successive
spectrum and Duke Island at the other, the last
surgesof magma. Whether this interpretationis named actually falling on the orthopyroxene-absent
correct or not, Figure 5 indicatesthat the rocks are side of the olivine-diopsidejoin (analogousto that
the result of fractionalcrystallizationwhich appears used by Yoder and Tilley, 1962, to divide alkalic
to have beensimply that of olivine removalfollowed from olivine tholeiite magmas). All of the remainby the removal of clinopyroxeneand orthopyroxene. ing magmasfall on the orthopyroxene-present
side
In Figure 5E, averagecompositions
for cumulus of this join and ultimately show silica enrichment,
rocks from the ore-bearing intrusions of the Pe- althoughthe Pechengaliquid lies closeto the join
chengadistrict are comparedwith those from the and seems to have some hesitation about the direcDundonaldsill, .bothsillsbeingsynchronous
with the tion in which to fractionate.
early volcanismof an orogenicbelt. With respectto
Irvine (1970) has shown that contaminationby
the Dundonaldsill, averagerock compositions
plotted re!sic material has altered the compositionof the
on the figure, consistingof peridotite (I), clino- Muskox magma sufficientlyso that its initial tenpyroxenite (II), magnetite gabbro (III), gabbro dencyto fractionatealong a trend similar to that of
(IV), and granophyricga'bbro(V), agree with the the Dundonald sill has been replaced.by one more
akin to that shown'by Bushveldand Stillwater. He
sequenceof cumulusphasesol, cpx, plag + cpx +
rot, and plag + cpx(+ opx?). Fractionation of a has suggestedthat the Bushveld and Stillwater
and hence
liquid alongthe path 1-2-3-4 can explain this rock magmasmay owe their initial composition
sequence.The fact that the compositionof the hyalo- crystallizationpath to similar contamination.With
clastite (column 12 in Table 2) overlyingthe .dif- referenceto the Bushveldin particular,E. N. Camferentiatedtholeiiticflow in Munro Township lieson eron (pers. commun.)haspointedout that the vast
the estimatedliquid fractionationtrend for the Dun- size of this intrusion, the fact that it has been JrFro. 5. Ca0.5(MgFe)o.O-(MgFe)O-SiO plots for ultramafic and mafic rocks from important igneousareas. The method of plotting the analysesis discussedin the text and by
Irvine (1974). The phase boundariesare approximate and are those estimated by Irvine.
Because their positions are influenced by variables such as total pressure, Pa_,, Pco2, and
chemical variations in the magma compositions,the phase boundariesdo not apply necessarily
in each of the casesillustrated. In Figure 5A the areas labeled I, II, and III cover the average compositionsof cumulus dunites plus peridotites, bronzites plus riorites, and gabbros, respectively, from the Bushveld and Stillwater igneous complexes. A liquid fractionating along
the path 1-2-3 can account for this sequenceof cumulates. Similarly in Figure 5B, taken
directly from Irvine, a liquid fractionating along the path 1-2-3 can account for the cumulus
dunites and peridotites (I), pyroxenites (II), and hornblendites(III) of the Duke Island
Conplex. The points in Figure 5C represent successiveliquid compositionsestimated for

komatiiticlavas of the Yakabindiearea (Naldrett and Turner, 1977). In Figure 5D, a liquid
fractionatingalong the path 1-2 can accountfor the sequenceof picrites and gabbrosfrom
Noril'sk plottedon the diagram. In Figure 5E, averagecumulusperidotites(I), clinopyroxenites(II), magnetite-richgabbros(III), and gabbros(IV) from the Dundonaldsill can be
accountedfor by a magma fractionatingalong the path 1-2-3-4 and endingup at the compositionof the granophyricgabbroin the sill (V). The initial magmacomposition
is thought

to be the sameas a hyaloclastite

cuttingandthenbecoming
interstratified
with lavasoverlying
the sill (VI). The sequence
of averageperidotites(A), clinopyroxenites
(B), and gabbros
(C andD) fromthe Pechenga
areacanbe explained
by a liquidfractionating
alongthe path
a-b-c. In Figure5F the fractionation
pathsfor differentliquidsare compared.Clearlythe

cotectic
linesin the diagramwouldneedmodification
for eachof the igneous
systems
under
consideration
exceptDukeIsland(for whichIrvinedrewthe linesin the first place),but
the relative geometryof the boundarieswould be similar in each case.

1142

,,1. J. NLDRETT
4,5

i?

ND

L. J. CBRI

deposits,whichwill requirelarge-scale
mining and
processing
with the productionof a very high proportion of tailings if they are to becomeeconomic.
The depositsof the Noril'sk district are associated
with the medium- and small-sized intrusions of class

6. Alpinebodiesof class2, Alaskan-type

complexes
of class3, and carbonatitesand kimberlites included

in class7 are unlikelyhostsfor nickelsulfideson the

basisof past experience.
Genesisof Sulfide-bearingMagmas

The relationshipbetweennickel sulfidedeposits

and differenttypesof ultramaficrocksis explicable
to somedegree on the basisof what we know about

Fro. 6. Nickel metal in past productionplus reserves associatedwith different classesof host rock (modified after
Naldrett, 1973). The key to the classesis given in Table 1.

the genesisof the deposits. The main criterionfor

the formationoa rich concentration
of magmatic
sulfidesis that the host magma should.besaturated
in sulfur and in additioncontaina reasonably
high
proportionoimmisciblesulfidedropletswhichcan
then settle rapidly to form a continuous,thick sulfide blanket.

If this criterion is not met and sulfides

rupted at five separatecenters,and the very similar

crystallizationhistory exhibitedby each of the five
centerswould require sucha degreeof homogenization following contaminationas to make this suggestion seem unlikely. Although assimilation of
wall rocks has undoubtedlyplayed a role in modifying magma compositions,
we are of the opinion
that this role hasbeenminor (with the exceptionof
$udbury where heat generatedduring meteoriteimpact may have provideda reservoirof molten felsic
magma [Peredery, 1972]) and that differencesin

segregateslowly as the crystallization of silicates

proceeds,thenthe sulfidesare likely to becomemixe(l
with the settlingsilicatecrystalsto form a disseminated zone olittle economicsignificance.
The productionof a high proportionof immisci-

world and the classof ultramaficrock to which they

belong. Apart from the uniquepositionof Sudbury
in termsof world sulfidenickel (shown by the crosshatchedportion of the columnlabeledclass4.5 in
Figure 6), komatiitic magmatism,class l(ii), has
the greatestproportionof associatedmineralization
and includesthe ores of Western Australia; Thompson, Manitoba; Ungava, Quebec; and Shangani,
Rhodesia. Tholeiitic sills of the picritic type, labeled
class 1(i) in this figure, are important primarily
becauseof the depositsof the Pechengadistrict.
Depositsnear the basalcontactof the Stillwaterand
Duluth complexes,also under class 4, have great
potentialif ever environmentalconsiderations
are

inducedfelsic melt may explain both the felsic character othe Sudbury Nickel Irruptive and the rich
concentrationsof magmatic sulfides.

ble sulfidesat a givenstagein the historyof a magmatic body can be broughtabout in a number of
ways. One of these, assimilatingsulfur from an
external source,has been shownto be important in
two of the districtsmentionedabove--Duluth(Mainwaring and Naldrett, 1974) and Noril'sk (Godlevliquid compositionsillustrated in Figure 5F reflect skii and Grinenko,1963)--in both of which the suldifferencesoriginatingin the sourceregion of the fur is extremelyrich in the heavyisotope.
magmasinvolved.
A secondpossi'ble
methodis to reducethe soluRelationship of Nickel-Copper Sulfide Mineraliza- bility of sulfur in a given magmaabruptlyby oxidizing or reducingthe magma or altering its comtion to Classes of Ultramafic
Rocks
position in some other way. Although theoretical
Figure 6, modifiedafter Naldrett (1973), indi- possibilities,none of these mechanismshas been
catesthe relationshipbetweenknown reservesplus shown to apply in a known case, although Irvine
past productionof nickel in the main depositsof the (1975) hassuggested
that assimilation
of an impact-

Komatiitic

ores are associated with

bodies of

highly magnesianundifferentiatedmagma,many of
whichhavebeenemplacedas flows20 to 100 m thick.
Evidence

of assimilation

of either felsic material

or

sulfides is lacking, the flows show no evidence of

having differentiated,and the ore usually occursas

basal accumulations

which must have settled at an

early stageto avoid becomingtrapped in the fairly

rapidly coolingmagma of the flows. These factors
lead to the conclusionthat the komatiitic magmas
that gave rise to orebodiescarried excessmantlewhen they were emoercome,but theyconstitute
lowgrade,.disseminatedderived sulfidesin suspension

ULTRAMAFIC

AND RELATED

MAFIC ROCKS

1143

placed,andit is thesesulfides
that settledoutrapidly
to form the ores,becoming
"riffledout" and trapped
as the flow magmaspassedacrossirregularitiesin
the underlyingtopography(Naldrett and Arndt,
1976). Naldrett (1973) hassuggested
that the sulfide-richnature of many komatiiticmagmasis re-

ioo?

latedto their deviationfrom deep,sulfide-rich

zones
within the mantle. A modifiedversionof this suggestionis given 'below,where we first discusscurrent

GEOTHEtMS

Postulated
200

Zoneof $utfdeMettin9

300
500

Figure 7 illustratesthe relationshipbetweenthe

Oceamc

{Clark ?mlwood,t964]

ideasconcerning
thegenesis
of magmas
in themantle.
Petragenesis
of maficand ultramaficrocks

,rchean

Present Oceanic

I000

1500

2000

2500

TemperotureC

degreeof meltingof averagemantlematerial and the

Fro. 7. A depth-temperaturediagram showing the relageothermas it might have appliedin the Archean, tionship between estimatesof the modern and ancient oceanic
and melting relations of possiblemantle material
usingdata from Green (1975). Mantle materialat geotherm
tpyrolite II +0.2 wt % HO) after Clark and Ringwood
the depth indicatedby point A would have been at (1964) and Green (1975) and drawn to illustrate the gena temperature sufficientlyabove the solidusto result

eration of komatiitic magmas.

in its beingapproximately
5 percentmolten. Any
slight local perturbationmight have causeda mass
of this material,perhapsseveraltens of kilometers
in diameter,to moveupwardas a diapir. Thermal
conductivities
of rocksare so low that given any
reasonablerate of movementthis uprise into cooler
surroundings
would have beenessentiallyadiabatic,
along a curve such as ABCD. The tendencyto
rise would then have .beenreinforcedby the hot
material's expandingas it moved upward into a
lower pressureenvironmentand thereforeassuming
a lower densitythan its surroundings.By the time
the diapir had reachedpoint B, 25 to 30 percent
would have been liquid and this liquid, of basaltic
composition,would have separatedfrom the unmeltedresiduumto rise to the surfacealonga nonadiabaticpathsuchasBF (dueto the smallervoltime
of liquid and the thinner conduitalong which the
low-viscosityliquid would have traveled, both factors leadingto an interchangeof heat betweenliquid

superimposed
on Green's(1975) dataon the melting
of mantlematerial. Note that the postulatedArchean geothermintersectsthe zone of melting at a
depthof about 100 kin. Above this depth. sulfides
in a mantle representedby this geotherm will be
solid,belowit they will be liquid. In this regionthe
mantle consistsof approximately10 percentmelt.
We suggestthat, given a relatively static situation
in areas away from major convectiveoverturn. the
denseliquid sulfideswill percolatedownward, displacingthe lessdensebasalticpore fluid and depleting in sulfur a zone in the mantle below 100 km.

As Naldrett originally suggested,this downward

percolationmay finally be arrestedby the sulfides'
solidifyingas the steepeninggeothermcrossesback
to the low-temperatureside of the sulfideliquidus
at greater depth. However, in the light of ideas
concerningsteepgeothermsin the Archean,we now
and surroundings). Continuedrise of the unmelted think it more likely that the downwardmovementis
residuumof the diapir from ]3 to C would have re- arrestedby a decreasingproportionof mantle pore
sulted in its undergoingfurther melting, perhaps fluid as the geothermleavesthe 10 percentmelting
as much as an additional 30 percent at C. (The contourin Figure 7 and,with increasing
depth.penecontourson Figure 7 are no longer directly appli- trates a region of a progressivelyIoxverdegree of
cablesincethey refer to undepletedmantle and not partialmelting. Sulfideswill tendto becometrapped
to the refractory, once-meltedmaterial that we en- as the porosityand hencepermeabilityof the rocks
visage traveling the path BC.) Separationof a decreases. As can be seen from this figure, the
second,much more magnesianliquid at C would degreeof melting (and thus volume of pore fluid)
have given rise to intrusionor extrusionof a high- decreases
below 5 percentin the very zone that we
temperatureultramaficmagmaat point E.
have suggestedis the source region of our most
magnesiankomatiiticmagmas(near point A). We
Concentrationof sulfidesin tl,e sourceregion of suggest,therefore, that the Archean mantle became
komatiites
zonedwith respectto sulfidesin this way, with sulfides
enrichedat depthscorresponding
to the source
In his 1973paper,Naldrett indicatedthe rangeof
pressnreand temperattireunder which natural nickel-

bearing sulfideswould melt. In Figtire 7 this is

region of komatiitic magmas,thus accountingfor the

relationshipbetween such magmasand ores.

1444

.4. J. NALDRETT AND L. J. CABRI

Platinum-Group Elements and Gold

Concentrations in rocks

Publisheddata on the platinum-groupelement

concentrationsin rocks from alpine and Alaskan
bodiesandthe Stillwaterand Bushveldlayeredcomplexes are summarizedin Table 5 and presented
graphicallyin Figure 8. Whole-rock and mineralseparateanalysesare alwayslikely to be influenced

'by the samplescontainingnonrepresentative

impurities, and this is especiallytrue for platinumgroup elementanalysesof ultramaficrocksand their

mineral separatesbecausethe platinum-groupelements are so irregularly dispersedamong the silicates. Another problem is that little is known re-

garding the solubilitiesof the numerousplatinumgroupmineralsthat may occurin a particularrock,

so that the resultsof chemicaldecomposition
thought
to dissolveonly an oxide (e.g., chromite) or the
silicates,and leave the platinum-groupmineralsas
unreacted residue, must be treated with caution as

pointed out by Crocket (1969, p. F-6). Crocket

also points out that fire assayor wet chemicalconcentration,followed by emissionspectrographic
or
spectrometric
analysisto overcomesensitivityproblems, causeslossesby adsorption,coprecipitationin
the cupellationstage of the fire assay,and/or contaminationby the use of large quantitiesof reagents.
In the case of alpine ultramafics, the average
platinum-group
elementvaluefor dunitesandperidotires from the Urals is within a factor of 2.5 of that

from Mr. Albert, Quebec,the Os contentbeingvery

ALPINE
COMPLEXES
I ALASKAN-TYPE
COMPLEXES
DUNTEB

CHROMITE-

OUNTEB

I I[ I I J[ cM
I II

PERiDOTITE

RCH ROCKS

PERIDOTTE

YR

CHROMITE-

PI......II

RCH ROCKS

Rh+Pd

Ru+Rh

Ir

Os+lr

Os+lr

+Pt

Pd + Pt

Pd+Pt

+Au

Fro. 8. Ratios of certain platinum-group elements and

gold in alpine- and Alaskan-type ultramafic complexes.The
numbers at the top of the columns refer to line numbers
in Table

5 A and B.

REVIEW

OF MINERALOGY

OF MERENSKY

REEF

1245

A. Y. NALDRETT AND L. .F.CABRI

ULTRAMAFIC

AND

RELATED

MAFIC

ROCKS

1141

dO

1148

A. J. N.dLDRETT

.4ND L. J. CzBI?I

similar in the two areas. Reported gold valuesvary

widely. We regardthe level of about2 pp'breported
for Mt. Albert and for somesamplesfrom the Urals
as the most realistic. The Pt/(Pt + Pd) ratio is
high, averaging about 0.8 for the Urals. The Pt/

crease in lhe Pt/(Pt + Pd) ratio. Von Gruenewaldt (1976) reportsthat the Pt/(Pt + Pd) ratio
in sulfidesjust below the main magnetiteseam of
the Bushveld (at a point at which the iron concentration in the Bushveld magma must surely have

(Pt + Ir) ratio is also high, averaging0.87. This

contrastswith the data of Cabri and Harris (1975),
whohaveshownthat placersrelatedto alpinedeposits
are characterizedby a predominanceof Os-Ir and
Ir-Os alloymineralsandlow valuesof the Pt/(Pt +
Ir + Os) ratio. Cousinsand Vermaak (1976) suggest that the discrepancybetweensourcerocks and
placersis due to the differentialmobility of the different platinum-groupelementsduring weathering
and transportation,but while this may well account
for variationin the ratio involvingPd, in our opinion
it has yet to be demonstrated
that Pt, Ir, and Os

been very high) is 0.45 (Table 5C, line 9), significantlylower than the value of 0.7 in the underlying MerenskyReef (Table 5D, line 3). The Pt/
(Pt + Pd) ratio of the Johns-ManvilleCompany
platinum-groupelementprospectin the bandedzone
of the Stillwater Complexis reportedas 0.22 (Table
5D, line 7), markedlylower than the averagevalue
of 0.64 reportedby Page et al. for the underlying
bronzititememberor the averageof 0.69 reportedby
them for the bandedand upper zones. I4. K. Conn
(pers. commun., 1976) reports that the Pt/(Pt +
Pd) ratio of a large sampleof subeconomic
but platratios are affected.
Chromite concentrations and
inum-group element-enrichedmaterial from the
samplesrich in disseminated
chromiteare generally upper part of the bronzititememberis 0.5 (Table
higher in platinum-group
elements,particularlyPt, 5C, line 4), significantly
lower than averageof Page
and the limited data suggestthat the Pt/(Pt + Pd) et al. for this unit.
ratiomay be distinctlyhigherin chromitethanin the
Later in this paper,we suggestthat the dominant
rock as a whole.
trend with fractional crystallizationof a tholeiitic
Platinum-group element and gold levels in melt is one of decreasein the Pt/(Pt + Pd) ratio.
Alaskan-type
complexes
(Table5B) are of the same We do this on the basisof comparingthe composiorder as thosereportedfrom alpinerocks,although tion of massive sulfide concentrations that have
the Pt/(Pt + Pd) ratio appearsto be lower in the segregatedfrom silicatemelts that ,had reacheddifformer. Pyroxenitesfrom Alaskan-typecomplexes ferent stagesin their fractionalcrystallizationat the
displayevenlower Pt/(Pt + Pd) ratios. As with time of sulfide segregation.

alpinecomplexes,
chromite-rich
rocksand chromite The highly variable data emergingfor rocks and
concentrationsand segregationsfrom Alaskan-type zones of relatively weak sulfide disseminationin
complexesare much richer in platinum-groupele- stratiformcomplexeslead us to suspectthat the platof cumulusrocks
ments than the chromite-poorrocks and, as with inum-groupelementconcentration
the rocks themselves,may have a lower Pt/(Pt + forming part of a large stratiform complex may be
susceptible
to changeafter the accumulationo the
Pd) ratio than chromitein alpinecomplexes.
Absolute concentrationsin the layered complexes rocks themselves. Processesresponsiblefor this
of platinum-group
are highlyvaria.ble,
depending
to someextenton the changemay include interc,hange
portionof sulfidein the samples
analyzed.Cousins elements with ascending intercumulus liquids
and Vermaak (1976) emphasize
that the Pt/(Pt + squeezedout of the underlyingcrystalcumulatepile
madeby Vermaak durPd) ratio increases
upwardin the BushveldCom- as it compacts(a suggestion
periodat the Denver PlatinumSymplex (Table5C, lines7-8; Table5D, line 3). They ing a discussion
contrast this with the trend observed in other intruposium) and interchangewith deuteric solutions
sions,where differentiationis one of silica enrich- circulatingwithin the consolidatingpile. As a rement (referredto by themas the Sudburytrend) sult of processessuchas these,the platinum-group
and where the Pt/(Pt + Pd) ratio decreases
with
crystallization. They attribute the behavior at

element concentration

in the cumulus rocks would

not be related directly to the compositionof the

magma from which they crystallized. Massive,parthan silica enrichment is the dominant differentia- ticularly basal, accumulations
of sulfide would be
to changesof this sort. Action trend, suggestingthat the more siderophile much less susceptible
natureof Pt as comparedto Pd causesit to follow cordingly,in developingour subsequentargument,
iron and be retained in the magma. An in.creasein we rely on data from depositsof this type.
Bushveld to the fact that iron enrichment rather

the Pt/(Pt +Pd) ratio with differentiationis also

in sulfide ores
suggested
by the data of Page et al. (1972) con- Concentration,v
Absolute values of platinum-groupelementsand
cerningthe StillwaterComplex(Table 5C, lines1,
2, 3, and 5). At bothStillwaterand Bushveld,other Au concentrations in sulfide ores and sulfide-rich
data are inconsistentwith a continuousupward in- rocksare given in Table 5D. Thesedata havebeen

ULTR,,IiiI/IFiC ,4ND REL/t7'ED ,'II,4FIC ROCKS

CRATONIC

OROGENIC
OM AT II TE

SHANGANI

ABITIBI MANIBRIDGERAGLAN

1149

LARGE

THOLE lITE

PECHENGA KANICHEE

MERENSKY

$TRATIFORM

COMPLEXES

STILLWATER

SUDBURy

INTRUSIONS
1
NORIL'SK

O6

04
O2
O0

O8

1'
1'
1'
1'
10
O6

04

O2

O0

O8
O6

Ir

04

O2
O0

O8

O6

Os + Ir+

04

Pt

O2
O0
O8

O6

04

oz

oo

O8

Au

06
04

Pd+
Pt+Au.j

.FT,

oz

Fr6. 9. Ratiosof certainplatinum-groupelementsand gold in sulfideoresfrom

different igneousassociations.

taken from a wide variety of sourcesand are of

variable quality. In many caseswe have taken productionfiguresfor total platinum-groupelements(as
at Sudbury) or estimatesof the concentrationof
total platinum-groupelementsin a depositor district
(as at Noril'sk) and used the ratios to proportion

is that some of the results are for sulfide concen-

trates,and the processof ore beneficiation

may have
concentratedsome platinum-groupminerals selectively over others.

Despitethesereservations,
we feel that our data
show some interestingtrends which are probably
the different elements to total this amount. While this
meaningful,particularlywhen one is considering
is the only approachavailableto us, it bringswith it ratios or interpreting differencesin concentrations
a number of major drawbacks.One of these is that of an order of magnitudeor more. In supportof
in a typical magmaticNi-Cu ore, ore dressingand this, considerthe three setsof figuresgiven in Table
refining may be gearedprimarily to elementsother 5D for the Merensky Reef and consistingof (1)
than platinum-groupelementsand only a proportion the averageof a seriesof direct analysesfrom a
of the total platinum-groupelementsrecovered,with singlecoreat theWesternPlatinum(Lonrho) mine;
the resultthat our figuresrepresentminimum values. (2) an estimateof the ratios at Western Platinum
A seconddrawbackis exemplifiedby Noril'sk, where calculated on the basis of an assumed overall conour estimated concentration for the ore is frown Newcentrationof metals; and (3) a similar approachto
man (1973) and is an overall estimatefor the dis- that of (2) abovefor the Rustenburgmine. Despite
trict; on the other hand, our metal ratios are from

the fact that variationsin grade ratio, and especially

Smirnov (1966) and are for differentore typeswith

no guide as to the proportion that each ore type

in platinum-group
mineralogy(cf. 7Brynardet al.,
1976; and Vermaak and Hendriks, 1976), are

contributesto the whole. We have applied Newman's grade to each of the ore types and have thus
comeup with concentrations
for eachelementbut in
doingthis .havemadethe unjustifiedassumptionthat
the grade of platinum-groupelementsdoesnot vary
from ore type to ore type. We have done this be-

known to occurthroughout the reef, the three sets

of figuresare remarkablysimilarwith the exception

of the gold concentration
givenfor Rusten'burg.
Two
sets of ratios for the platinum-groupelementsand
Am, one from Cousins(1973) and the other from
the InternationalNickel Companyof CanadaLimited
causethe figuresprovidea roughmeansfor compari- (R. A. Alcock,pers. commun.,1975), were availson with other districts, but we have expressedour ableto us for the Sudburydistrictand are alsovery
reservationsby placing the figures in parentheses similar.
in Table 5D. A third reservation about Table 5D
In general,where differentsets of figureswere

1150

,4. J. N`4LDRETT

AND

ratio Au/(Au + Pt +Pd), ranging from 0.02 tu

0.15, showsno systematic
variationbetweendeposits.
The Pt/(Pt +Pd) ratio decreasesamongdeposits
with tholeiiticaffinitiesfrom a high of 0.55 at Pechenga,through0.49 at Sudbury,to a low at Noril'sk
(0.27 to 0.009) and Kanichee(0.15). The komatiitic oresare characterizedby intermediatevaluesof
this ratio, rangingfrom 0.27 at Manibridgeto 0.46

09

O8

07

0.6

PechengoeN
05

at Shangani.

, Sudbury

The Pt/(Pt +Pd) ratio is plotted against the

Cu/(Cu + Ni) ratio in Figure 10. The deposits
associatedwith tholeiitesdisplay an ahnost linear
trend of decreasingPt/(Pt +Pd) with increasing

04

f Komatdtes
,/
03

_I.._.__
l///

L. J. CABR[

Cu/(Cu + Ni), 'but the komatiiticdepositscluster

as a groupin a totally differentpart of the diagram.
Our explanation
for this observation
is relatedto the
genetichistoriesof the differenttypesof depositand

'x
o
rll
'sk
N,.'
----Nx
.Kon,chee

02

is discussed in a later section.

O0

Genesisojt platinum-group
elementconcentrations
oo

ol

o.a

o3

o4

oo

o.6

o?

Before the relationshipbetween platinum-group

elementcontentand the genesisof komatiitescan
be discussed,
we must first discussthe ?,'enesis
of

Cu/(Cu+Ni)

lrmo.10. The relationshipbetweenPt/(Pt+ Pd) and Cu/

(Cu + Ni) wt percentratiosin certainsulfideores.

platinum-groupelementconcentrations
in general.
The four main stepsin the formationof a concentra-

availablefor a singledepositor district,we havese- tion are illustrated in Figure 11. The ultramafic
lectedonesetas our preference.This hasbeendone and mafic rockswith which platinum-groupelement
on the basis of the nature (mill head, concentrate, deposits
areassociated
aretheresulteitherof partial

production
figure,individualsample)and numberof meltingof the mantleand the introductionof the
the samples
coveredby the figuresand on the basis melts themselves into the crust or of the direct emof our estimateof how representative
they are likely placement
of solidmantlematerialinto the crustas
to be of the depositas a whole.The preferredfigtires an alpineultramaficbody.
are indicatedby checkmarksin the table.
All sulfide-associated
platinum-group
elementdepositsoccurwith rocksintroduced
as melts. The
Fariation in the concentrationof platinu-groupele- platinum-group
elementcontentof the deposits
rementsand Au in sulfide-richoreswith the nature of

flects the amounts of these elements in the silicates,

the host rock

oxides,and sulfidesin the sourceregion in the

in whichtheseminerals
Before proceedingwith this section,we should mantleand the proportions
in the melts.
remind the reader that the Sudbury, Noril'sk, and were incorporated
On risingup into the crust,a silicatemelt will
Pechengaoresare associated
with tholeiiticmagmas.
and the concentration
of individual
platMarbridge,Mani'bridge,Raglan,and Shanganiare crystallize
elements
in the residual
meltwill change,
associated with komatiites. Hawke and James inum-group
(1975) statethat the hostrock to the Kanicheede-

posithasa tholeiiticaffinity.We excludethe Vlakfonteinnickelpipesfrom consideration

becausethey
are thoughtto havebeenemplacedinto the Bush-

' anddllferentiatlonJ"-

sulhdes
or

Red,sfr,buhon by
hydrothermalprocesses

veld Complexat least500 m.y. after it crystallized

(Vermaak,1976); in our opinionthis raisesuncertaintiesaboutthe natureof the magmawith which

theywereassociated
or, if thisis assumed
to be the
Bushveldmagma,uncertaintiesabout changesinducedin their metal ratiosduring the remobilization
of the ores.

Manlie
Melhncl
Incorporohoflof J
sllicOtes, oxides
ond su)fmdes

Sufficient data on the ratios of elements (illus-

tratedin Fig. 9) are availableonly for Pt, Pd, and

Au for meaningfulcomparisons
to be made. The

Fro. 11. Importantprocesses

in the distributionandconcentration of platinum-groupelements.

ULTRAMAFICAND RELiTED M.4FIC ROCKS

1151

dependingon how muchof eachelementis removed

in the crystallizingsilicateand oxide minerals. At
somestageduring crystallization,dropletsof liquid
sulfide wilI segregatefrom the melt. The sulfides
viI1 act as a collectorof platinum-groupelements

Duluth

Nord'sk

from throughout the magma and, coalescingand

settlingunder gravity, perhapsto form a sulfide-rich
zone, will be the main means of concentratingthese

Stlllwaler

--

Sudbury

elements.

As is emphasizedin Figure 11, although the associationwith sulfide is marked, it is clear that in
manycasesredistributionof platinum-groupelements
has occurredafter crystallizationof the main sulfide
minerals as a result of exsolution

AlexoKarnbalda
E Goldfields

and the action of

other elementssuch as Te, Bi, Sb, and As as collectors,or as a result of the action of hydrothermal
solutions,or as a result of the combinedprocesses.

$tumpfl (1974; Stumpfl and Tarklan, 1976) has

emphasized
the hydrothermaleffect. Cabri and Laflareroe(1976) concludethat muchof the platinumgroupelementcontentat Sudburyis presentin discrete platinum-groupelement-richminerals which
would not have existed at magmatic temperatures
and must have formed during cooling. For example,
micheneriteis the main Pd-bearing mineral at Sudbury, and its upper stability limit has been deter-

mined as 501C (Hoffman, 1975).

When we considerstepsII and III in Figure 11,
we find it profitableto inquireas to the stagein the
fractionalcrystallizationof a given silicatemelt at
which an immisciblesulfide liquid segregated. Becauseof the tendencyfor Ni to substitutemuchmore

readily than Cu in early mafic minerals,the Cu/

(Cu + Ni) ratio of the sulfidesprovidesone of the
bestmethodsof assessing
this. Wilson and Anderson (1959) drew attentionto the relationshipbetween the Cu/Xi ratio of an ore and the character
of its host rock, and we also illustrate this for a
wide rangeof depositsin Figure 12. The trend of
increasingCu/Ni with decreasingmarlcity is clear
and this, together with the relationship illustrated
in Figure 10, leadsus to hypothesizethat the norreal trend of the Pt/(Pt + Pd) ratio is to decrease
with decreasingmarlcityor, to extendthe hypothesis

slightly,to decreasewith increasingdifferentiation

of the host rock. This may come about as the result of Pt beingincorporatedpreferentiallywith respectto Pd in early forming silicatesand oxides,
leadingto a decreasein the ratio in the magma and
hence to a decreasein the sulfideswhen they segregate. Some supportfor this hypothesiscomesfrom
the observationthat pyroxenitesin the Alaskan-type
intrusions have lower Pt/(Pt + Pd) ratios than
dunitesand peridotites,althoughthe availabledata
are too meagerfor great relianceto be placedon this
argmnent. The data of Yushko-Zakharovaet al.

--Manitoba

Perseverance
I

30

20

Io

Wi perceni MgO In hqmd

Dumte

Peridotlie

Pyroxemte

Gabbra

Fro. 12. Relationship

betweenCu/(Cu+ Ni) ratio of a

sulfide ore and the nature of its host magma or rock.

Where quantitativeestimatesof the MgO contentof the

liquidportionof the hostmagmaare available,the deposits
are plottedas dotsagainstwt percentMgO. Where only
the host rock type is known, the depositsare plotted as
horizontal bars. The Pechengamagma compositionwas ob-

tainedby making a weightedaverageof the different rock

types in the host intrusionsusing the proportionsgiven by
Gorbunov (1968). This gives a very magnesian liquid.
Some of the gabbroic portion of the intrusions may have
been squeezedout of the magma chambers,continuingto
surfaceas flows, thus biasing our estimate. This uncertainty
is shown by a dashed line. Although the Sudbury ores are
noxv directly associatedwith norite, Naldrett et al. (1972)
considerthat they may have segregatedat an earlier stage
from a more mafic magma and then have been remobilized
and emplaced in younger norites. The uncertainty about
the compositionof this magma is also shown by a dashed
line.

(1967) indicate that the Pt/(Pt+

creases from 0.68 in ultramafic

Pd) ratio de-

to 0.67 in mafic to

0.50 in telsicrocks. The studiesof Chyi and Crocket

(1976) on the distribution of Pt and Pd at the
Strathconamine, Sudbury, also suggestthat Pt is
relativelyoxyphilecomparedwith Pd. It shouldbe
stressedat this stage that a decreasein the ratio
does not imply an absolutedecreasein Pt. The
absoluteamounts of both Pt and Pd may increase
with increasingdifferentiation,Pd increasingmore
rapidlythan Pt. The controlof boththeir ratio and
their respectiveconcentrationswill depend on the
concentrationin the initial magma and the concen-

tration in the crystallizingsilicatesand oxides.Gijbelset al. (1974 and 1976) haveattemptedto model
fractionation trends quantitatively by .determining
Os, Ru, and Ir concentrations
in mineral separates.
Their models, which suggestthat absolute concentrations shoulddecreasewith fractionation,are open
to questionbecauseof uncertaintiesas to the purity
of their mineral separates.

The questionremainswhy the komatiiticoresare

a majorexception
to the trendbasedon thetholelites.

1152

.4. J. NALDRETT AND L. J. CABRI

TABLE6. RoughEstimatesof Concentrationof Ni, Cu (wt%), Platinum-GroupElements,and Au (ppm)
in Original Sulfide Liquid

Merensky Reef

Ni

Cu

10.9

4.65

10.9

4.65

4.5

4.0

9.0
7.6
4.6
6.5
10.0
9.75

10.8
10.9
2.45
0.38
0.75
1.20

Ru

21

Rh

17

Pd

Os

Ir

Pt

Au

Ref.

109

2.7

252

31

18

102

4.3

258

21

Stillwater

(Johns-Manville)
Sudbury

0.08

0.09

600-2,000
1.09

0.03

0.03

160-550
1.04

0.37

*
15

18.0
15.2
0.6
0.36
0.95
1.3

1.6
1.7
0.2
0.18
0.24
0.10

6
6
5
16
16
23

Noril'sk,

Picritic gabbro
Taxitic gabbro
Pechenga
Marbridge
Manibridge
Shangani

1.7
1.4
0.08
0.26
0.20

47.8
42.1
0.5
0.92
2.6
2.0

* Basedon the figuresgivenin Table 5D, togetherwith a field estimateof the sulfidecontentof the ore (H. K. Conn, pers.
commun., 1976).
N.B. The referencesare the original data on ore grade as listed in Table 5.

\Ve suggestthat this is due to the high proportion

of mantle sulfide which, as stressedbefore, we believehasbeenincorporated
in ore-bearing
komatiitic
magmas. The platinum-groupelementsin this sulfide will exert a strong influenceon the platinumgroup dement contentof the magma as a whole.
This contrastswith the case of depositsassociated

ment concentrations in sulfides associated with differ-

ent host magmas is required.

Estimatesof the platinum-groupelementconcentrations in sulfide melts

The concentrations
given in Table 5D apply to
sulfide-bearing
oresbut not to the composition
of the
with relativelysulfide-poortholeiiticmagmas;in original sulfide-oxidemelt responsiblefor the ores.
thesethe proportionof platinum-group
elementsin In most casesthis melt would have first crystallized
the silicatesand oxides entering the host silicate as a nickel- and copper-bearinghigh-temperature
solidsolutionor Mss)
magmawas the factorcontrollingthe compositionpyrrhotitephase(monosulfide
of the sulfidewhich uItimatelysegregated
from it. and magnetite. On cooling,the Mss would have exIt is generallyacceptedthat tholelitesare the con- solvedthe presentnickel,copper,andplatinum-group
sequence
of a muchmorelimiteddegreeof melting minerals. The total platinum-groupelementcontent
of the mantle than komatiites. Spinel, which we would be the sum of these elements in solid solution
postulate
to'berelativelyrichin Pt, constitutes
a rela- in the various sulfide minerals and magnetite, and
tively largeproportionof this limitedpartialmelt, present as separateminerals. Table 6 presentsthe
thus accounting
for the high Pt/Pd ratio.* The resultsof an attemptto estimatethe platinum-group
reasonableness
of an explanation
of this typecanbe element content of the high-temperatureMss and
testedin a crudeway byestimatingconcentrationshencethe sulfidemelt. For most depositsthis has
and partitioncoefficients
for mantlematerialsand been done by estimatinga value for the nickel conthemagmas
derived
frolnthem. Thisis donein the tent of the original Mss (basedon a knowledgeof
lastsection
of thispaper. In orderto do this,how- the deposit), determining the factor necessaryto
bring the nickel grade of the ore up to this estimate,
ever,someinsightinto absoIute
platinum-group
ele- and then applyingthe samefactor to the platinum* This postulateis madeon the basisof the strongcor- group elements. In effect,we are using the nickel
relation observedbetween platinum and chromite concen- contentof the ore to provide a guide to the proportrations,as illustrated
in Table5 andin the work of Razin
et al. (1965) and Razinand Khomenko
(1969.). Cousins tion of sulfidesin the orebody at the present time
andVermaak(1976) discuss
the correlation,
payingparticu- and then increasing the platinum-group elements
lar attentionto whether platinum and other platinum-group by the factor requiredto give their concentrationin
elements substitute within the structure of the chromite or
of 100 percentsulfides. In doing
are presentas micron-sized
inclusions
of other minerals. a liquid composed
They conclude
that this questionis very muchan openone. this, we are ignoring the small amount of magnetite
We suggestthat at magmatictemperatures
the spinelstructure may be receptiveto platinum-groupelements,par- (generallylessthan 10 percent;seeNaldrett, 1969)
ticularly platinum,but that theseelementsmay exsolveto presentin most ores, but the error introducedin
some extent during cooling. This suggestionexplainsthe this way is small. For Noril'sk, Smirnov's (1966)
associationbetweenplatinum-groupelementsand chromite
and the presence
of inclusions
of platinum-group
minerals, data, and for the Merensky Reef, Liebenberg's
but it cannotexplain the presenceof someplatinum-group (1970) data, on the Ni, Cu, and S contentsof the
elementsas very small "negativecrystals"as describedby
Hendriks and Peyerl and Hendriks and Kinloch in the un-

ores were used to calculate the Ni

content of the

published
company
reportscitedby Cousins
andVermaak. sulfidephasein orderto determinethe factor.

ULTR,qM.4FIC .4ND REL.4TED 2[.4FICROCKS

1153

If the deficienciesof our data are borne in mind, time because

of the difikully of makingpure concenit appearsfrom Table 6 that in sulfidemeltsgiving trates. As better data becomeavailable,the approach
riseto Ni-Cu ores,the "normal"rangeof Pt is 0.36 usedhere can be refinedto test the feasibilityof the
to 1.6 ppm and Pd 0.5 to 2.6 ppm. The concentra- hypothesismore thoroughly.Our modelis developed
tion of both Pt and Pd in the Noril'sk sulfide melt

as follows'

was10 to 20 timesthehighest"normal"vaiuesthat

1. Peridotitic komatiites are taken to represent

we quoteabove. If one assumes
that all platinumvery completemelts of mantle material and the comgroupelementsiu the MerenskyReef were collected
by a sulfideliquid, this liquid contained160 times positionof komatiiticoresto representsulfideliquids
the highest "normal" level of Pt and 40 times the in equilibriumwith thesemelts. The Marbridge sulhighest"normal"levelof Pd. If the salneassump- fides, containing920 ppb Pd and 360 ppb Pt, are
tion weremadefor the Johns-Manville
propertyat typicalof suchoresand thesefigureswill be used.
2. Since alpine ultramarie rocks are regarded as
Stillwater,this depositwouldhavecontained100 to
350 times as much Pt and 230 to 770 times as much
Pd as our "normal"

levels.

There is no theoreticalreasonto suppose

that the
partition coefficientsgoverningthe distributionof
platinum-groupelelnentsbetweensilicateand sulfide
melts in the Bushveldor Stillwater magmasshould
havebeenvery differentfrom thoseoperativeat, for
example,Sudbury. Two alternativespresentthemselves. Either the magmaresponsible
for the MereriskyReef wasapproximately160 timesenrichedin
Pt and that at Stillwater over 230 times enriched in

Pd with respectto the magmafroln which the Sudbury ores segregated,or the mechanismfor concentrating platinum-groupelementsat Merenskyand
Stillwaterwas differentfrom that whichoperatedat
Sudbury and at most other massive Ni-Cu sulfide
deposits.* We cannot make a firm choice between

the alternatives,but as emphasizedpreviously,we

suspectstronglythat the simplehypothesisof sulfide
segregationmay be an inadequateexplanationior
the concentration
of platinum-groupelementsin the
MereriskyReef andthe StillwaterComplex. A satisfactoryexplanationmay developas our knowledge
of other platinum-group element-rich zones of
weakly disseminatedsulfidesin igneousbodies increases.

2[odel to illustrate possiblemechanismfor concen-

tratingplatinum-group
elementsin differentmagma
types

This model is Jutendedto illustratehow the hypothesis presented earlier could work and what
someof the necessaryrequirementswould be. It is
basedon a numberof assumptions
concerningthe
concentrationof platinum-groupelementsin certain

fragmentsof the upper mantle, the averageconcentration of platinum-groupelementsin Ural alpine

dunite plus peridotite provides the best estimate
amongthe data availableto us of the platinum-group
element

content

of the mantle

and therefore

of the

platinum-groupelement content of the sourcerock

for magmas derived from the mantle (18 ppb Pd,
73 ppb Pt). The rocksrepresented
'by this average
probably include some sulfide (Guillon, 1975, has
pointedout that the New Caledonianophiolitecontains modest amountsof sulfide), although no information

on the sulfide content

is available

in the

original references. We will showthat 0.5 percent

sulfideor greater would, on the basisof our model,
result in a negative content of Pd for the silicate
plus oxide fractionof the mantle. Our modelworks
bestif we assume0.35 percentsulfideto be present
on average in the Ural rocks, equivalentto about
0.12 weight percent sulfur. Some of the Pt and
much of the Pd in the figures quoted above will be
lockedup as sulfide. We will calculateexactly how
much in step 4 below.
3. In order to developthe model,we must assume
partition coefficients
for Pd and Pt, definedas
(concentration of metal M in
sulfide liquid in mantle)
DMSL/KO _(concentration of metal I in
coexisting komatiitic melt)

These have been selectedby a processof trial and

error to 'be20 and 4 for Pd and Pt, respectively,to
ensure the success of the model.

Further

discussion

of this selectionprocessis given in step 10 below.

Supportfor the more chalcophile
nature of Pd with
respect to Pt stems from Goldschmidt's(1954)
minerals and the concentration of sulfides in certain
analysescoinparing the partitioning of platinumrocks. Reliable concentrationdata are lacking and, group elementsbetweentroilite and Fe-Ni alloy in
for minerals,are likely to remain lackingfor some chondritic meteorites. Although both Pd and Pt
* It is relevant to point out also that the Pt and Pd sul- prefer the metallic phase, Pd partitionstwice as
fides cooperiteand braggite, with or without vysotskite, are readily as Pt into the sulfide. It is not unreasonable
important platinum-groupminerals at Merensky, Stillwater. to suggestthat where oxide-sulfidepartitioning is
and the Lac des Isles deposit in Ontario in contrast to the concerned,both elementswill favor the sulfide, Pd
platinum-group minerals usually present in massive Ni-Cu
favoring it more so than Pt.
sulfide deposits.

1154:

.4. J. N./tLDREIT

rtiV]_) ]_,.J.

4. At this stagewe can investigatewhat will happen during the nearly completemelting of a sulfurrich portion of the mantle to give rise to a sulfiderich komatiitic magma. Let us assume that the
mantlein questioncontains1.5 percentsulfidedueto
a processof enrichmentsuchas that suggestedabove.
We don't know the compositionof this sulfide; all
we know from step 1 aboveis the compositionthat
it has once the mantle has been melted and equi-

and that they are not saturatedat the moment of

emplacement,
containingperhaps0.05 weight percent sulfur (equivalentto 0.15 percentsulfide) ac-

librium has been reached between sulfide and silicate

in equilibriumwith the komatiiticmagmabut will be

quiredfrom their sourceregion. The data of Haughton et al. (1974) indicatethat magmasof this type
containingabout 10 to 14 weight percent total Fe
calculated

as FeO

would

become

saturated

with

about 0.08 to 0.1 weight percent sulfur. The compositionof this sulfidewill not be that of the liquid

nelts. Partitioning between the solid phases,or that of the sulfide occurring in alpine peridotire,
betweensolidsilicatesplus oxidesand liquid sulfides, whose compositionwe have calculatedin step 4
is likely to ,be different,probablywith Pd and Pt above.
6. In accordance
with data presentedin Table 5,
tendingto favor the sulfideto an evengreater degree
than is the caseof partitioningwith a silicatemelt. let us assume that much of the Pd and Pt in the
Making the not unreasonableassumption(Shima sourceregionof tholeiiticmagmasis presentin spinel

andNaldrett,1975) that0.5 percentsulfidedissolves in which the Pd and Pt concentrations are 80 and

in the silicatemelt, we find that the concentrationof 800 ppb respectively[this gives Pt/(Pt + Pd)Pd and Pt in the sulfides in the unmelted

mantle

0.91, which is the same as that of the chromite re-

materialand in the silicateand oxide phasesin the

portedon line 11 of Table 5A].

same material can be calculated from a mass balance

7. Spinel lherzolites,consideredto be representative of the averagemantle,containabout5 percent

equationof the type,

{).015 }\[su q- 0.985 [-\ltD - 0.0035 M sU-]

= 0.0l M s' q- 0.99 51Ko

26.7 ppb Pt to be accounted

for by t.hesilicatemin-

w h ere

erals. (Cf. calculationin step 4, above.)

concentration

of element

in sulfides in

unmelted mantle, 0.015 is the weight fraction

of

sulfide

in

the

sulfide-enriched

mantle, and 0.0035 the weight fraction of

sulfide in the Ural dunire plus peridotite.
'.\IUD

concentration

of the element

in Ural

dunire plus peridotire and 0.985 the

weight fraction of this rock in the sulfideenriched zone in the mantle.
concentration
of element
M

in

sulfide

liquid in mantle and 0.01 the weight fraction of sulfide liquid coexisting with the
komatiitic melt representing completely
[KOM

spinel. If we assumethis amount to be present,

spinelwill contribute4 ppb Pd and 40 p?b Pt to the
total in our model mantle, leaving 2.8 ppb Pd and

melted mantle.
concentration
of element

in komatiitic

melt in mantle = M s + D s/

8. Let us assume that the mantle source under-

goes25 percentpartial melting. This is approximatelythe degreeof meltingrequiredto producea

tholeiitic basalt magma (Green and Ringwood,

1967). Let us alsomakethe reasonable

assumption
that all of the spinel dissolvesin the partial melt.
The concentrations
of Pd and Pt in the partial melt
will consistof the contributionfrom spinel(wt frac-

tion= 0.2) plusthat from sulfideacquiredfromthe

sourceregion(wt fraction: 0.0015) plusthat from
silicates(wt fraction= 0.7965). Thesewill amount
to 23.0ppbPd and 184ppbPt. Usingthepartition
coefficient,
D:sr'/s, given in step3 above,the sulfide in equilibriumwith this magmawould contain

460ppbPd and736ppbPt [Pt/(Pt + Pd) -- 0.62],

valuesthat are fairly closebothin absolute
leveland
ratio to those of the Pechenga ores.

Solution of this equationgives 3,203 ppb Pd and

1,800ppb Pt in the sulfides(M su) and 6.8 ppb Pd
and 66.7 ppb Pt in the silicatesplus oxidesof unmeltedmantle (M U)- 0.0035 MSV). If we had assumedthe weight fraction of sulfidein Ural dunire
peridotireto be 0.005, (M ur'- 0.005 M sU) would
havehad a value --0.5 ppb Pd.
5. Now in consideringtholeiitic magmas such as
those responsiblefor the Bushveld,Sudbury, Pechenga,and Noril'sk intrusions,let us assumethat
thesemagmasare generatedin zonesin the mantle

9. Suppose
that the earlysilicateminerals(olivine
and pyroxene) to crystallizefrom this tholeiitic
magmacontainPd andPt at concentrations
similar

less rich in sulfide than is the case for the komatiites

fractionalcrystallization
of silicatesfrom a tholeiitic

to those estimated for the mantle source region

(givenin step7 above). Thenthepartitioncoefficients,

(metal in tholeiitic nagma)

(metal in silicatescrystallizing
from the tholeiitic magma)

are 8.2 and 6.9 for Pd and Pt, respectively. Thus,

ULTRAMAFICAND RELATEDMAFIC ROCKS

magma would both increasethe absolutelevels of
Pd and Pt in the magmaand also decreasethe Pt/
(Pd + Pt) ratio. In this way it is possibleto account for the trends observedin passingfrom Pechengato Sudbury to Noril'sk. The presenceof perhaps 10 percentof a spinel suchas chromiteamong
the early crystallizingphaseswould both retard the
overallenrichmentin the melt in platinum-groupelementsand causea morepronounced
decreasein the
Pt/Pd ratio.
10. A valid criticism of the above model is that

1155

depthsin the mantleat whichsulfideshad alsobecome concentrated.

3. Rich concentrations of sulfides are associated

with tholeiiticmagmasat Sudbury,Pechenga,and

Xoril'sk. In addition,a great deal of nickelis contained within large volumesof tholeiiticrocks as
moderate
to weaklydisseminated
sulfide. Examples
of theseincludethe depositsnear the baseof the
Duluth,Bushveld,and Stillwatercomplexes.Zones

of weaklydisseminated
sulfidemanythousands
of
feet above the base of the Bushveld and Stillwater

the partitioncoefficients
for Pd and Pt betweensilicate and sulfideliquids are too small. However, so

complexes
alsocontainlargereserves
of nickeland
copperbut are remarkable
primarilyfor their high

long as we hold to our assumptionthat the platinumgroup elementlevels in komatiitic sulfidesare those
of a sulfideliquid in equili.briumwith a nearly complete melt of mantle material closeto Ural dunite +
peridotitein composition,
we are constrainedto using

concentrationsof platinum-groupelements.

coefficients of the order that we have selected.

This

4. Assumingthat most Ni-Cu ores originated

from the segregation
of an immiscible
sulfideliquid
from a silicatemagma,the "normal" rangeof concentrationof Pt and Pd in this sulfideliquid appears
to have been 0.3 to 1.6 and 0.5 to 2.6 ppm, respec-

is assumed
to be
can be demonstratedby rearranging the equation tively. If the samemechanism
solelyresponsible
for the concentration
of platinumgiven in step 4 aboveto read
0.0116

M sU

group elementsin the MereriskyReef and upper

bandedzoneof the StillwaterComplex,the Pt and

(0.01 M sL -t

0.99 I sL

D.xis/o

in the sulfideliquid were 2 to 3

0.985 M urn) Pd concentrations

ordersof magnitude
greater. On thisbasisalonewe
question
the modelthat supposes
simplesulfidesegThe expressionon the right-hand side of the equaregation
to
be
solely
responsible
for concentrating
tion becomes
negativeand thereforeimpossible
when
platinum-group
elements
in
these
two
deposits.
Dersn/Km exceeds 5.7 or Dm s/r'u exceeds 107.
These valuescan be raised slightly by assumingthat
more than ! percentfree sulfidecoexistswith the
mantle melt, but the proportionsof sulfidefound in
komatiitic rocks suggeststhat even 1 percent is perhaps a generousestimate.
Conclusions

In our opinion the most obvious conclusionto

emergefrom this paperis the necessityof determining the nature of the igneoushost rock of a deposit
and the relationshipof the ore to the host rock (i.e.,
concentratedsulfide segregationor weakly disseminated sulfides)when trying to understandplatinumgroup elementdistribution.
Some other significantpoints are:

5. The availabledata suggestthat in sulfideores

associatedwith tholeiitesthe Pt/(Pt + Pd) ratio

decreases
withincreasing
Cu/(Cu + Ni) ratiowhich.
in turn, is an indicationof an increasingdegreeof
differentiation

of the host rock.

6. Ores of komatiitic associationdiffer markedly

in havingbothlow Pt/(Pt + Pd) andlow Cu/(Cu

+ Ni) ratios. This is consistent
with the suggestion that the komatiitesare particularlyrich in juvenile,mantle-derived
sulfur,with the resultthat they
contained immiscible sulfides at the time of their

eraplacement.
Acknowledgments

We are mostgratefulto Dr. A. R. Grahamof

Nickel Mines Limited for releasing

1. The major nickel-sulfidedepositsare associated Falconbridge
information
on
the platinum-groupelementcontent
with a wide range of rock types, all of them either
tholeiitic

or komatiitic.

2. Komatiites are particularly favorablehostsfor

suchores, includingthe depositsof Western Australia, the Manitoba nickel belt, the Ungava nickel

belt, Shanganiand probablyother depositsin southern Africa, and small depositsin the Abitibi greenstone belt of Canada. Many of the komatiite-associateddepositsare small and of high grade. It is
suggested
that the association
is due to peridotitic
komatiitemagrnasgeneratedat greater than normal

of the Manibridge,Marbridge,and Raglanproperties,to Dr. R. A. Alcockof the International

Nickel

Company
for information
onplatinum-group
element
ratiosat Sudbury,and to Mr. R. C. Stavelyfor informationon the Kanicheedeposit.The comments
of
Professor
J. H. Crocketof McMasterUniversityand
Messrs. A. Green and E. L. Hoffman of the Uni-

versityof Torontoweremosthelpful.
Mr. F. Jurgeneitis responsible
for the drafting
and Mr. B. O'Donovanfor the photography.Sup-

1156

,4. J. NALDRE'I'TAND L. J. CABRI

port from Professor Naldrett's National Research

Grant No. A4244 is acknowledged.

A. I. N.
DEPARTMENT OF GEOLOGY
UNIVERSITY OF TORONTO

TORONTO,
ONT^RO,C^NADAMSS 1A1
L. J. C.
CANADA CENTRE FOR MINERAL AND ENERGY
TECHNOLOGY

555 BOOTH STREEt

OTr^W^, OT^RIO, C^NXD^

K1A OG1
June 29, 1976
REFERENCES

Allard, G. O., 1970, The Dore Lake complex of Chibougamau, Quebec--a metamorphosedBushveld-type layered
complex: Symposiumon the Bushveld and Other Layered
Intrusions: Geol. Soc. South Africa Spec. Pub. 1, p. 477-

and Vermaak, C. F., 1976, The contribution of southern African ore depositsto the geochemistryof the platinum group metals: Ecos. GoI., v. 71, p. 287-305.
Crocket, J. H., 1969, Platinum metals. Sections in Handbook of geochemistry: Berlin, Springer-Verlag, v. 2,
sec. 1.

1974, Gold. Section in Handbook of geochemistry:

Berlin, Springer-Verlag, v. 2, sec.4, p. 79-1 to 79-0-1.
--and
Chyi, L. L., 1972, Abundancesof Pd, It, Os and

Au in an alpineultramariepluton: Internat. Geol. Cong.,

24th, Montreal, Proc., sec. 10, p. 202-209.
Crocket,J. H., Teruta, Y., and Garth, J., 1976,The relative
importanceof sulfides,spinels, and platinoid minerals as
carriers of Pt, Pd, Ir, and Au in the Merensky Reef at
Western PlatinumLimited,near Marikana, SouthAfrica:
Ecoa. Ggo., v. 71, p. 1308-1323.
Davies, H. L., 1968,Papuanultramaficbelt: Internat. Geol.
Cong.,23rd, proc., sec. 1, p. 209-220.
1971, Peridotite-gabbro-basalt
complex in Eastern
Papua: An overthrust place of oceanicmantle and crust:

AustralianBur. Mineral Resources

Rep. 128,48 p.

Dewey,J. F., and Bird, J. M., 1971,Originand emplacement of the ophiolite suite: Appalachianophiolitesin

Newfoundland:Jour. Geophys.Res., v. 76, no. 14, p.

491.

Duquette, G., Latulippe, M., and van de Walle, M.,

1972, Precambrian geology and mineral depositsof the
Noranda-Val d'Or and Matagami-Chibougamau greenstonebelts,Quebec: Internat. Geol. Cong.,24th, Montreal,

3179-3206.

Dodin,D. A., Solomina,

L. A., and Shatkov,V. A., 1973,
Someaspectsof the geochemistry
of platinummetalsand

goldand its application

in prospecting
for Noril'sk-type
coppernickel ore deposits,in The North-Siberian nickelbearingregionand its economicoutlooks: USSR Scienti-

Field Guide to Excursions A41 and C41.

Arndt, N. T., 1975, Ultramarie rocks of Munro Township

and their volcanicsetting: Unpub. Ph.D. Thesis, Uni-

fic ResearchInst. of Arctic Geology,p. 108-115(in

Russian).
versity of Toronto.
Findlay, D.C., 1969, Origin of the Tulameen ultramafic1976, Ultramafic lavas of Munro Township; economic
gabbro complex, southern British Columbia: Canadian
and tectonic implications, in Strong, D. F., ed., MetalJour. Earth Sci., v. 6, p. 399-426.

logeny and plate tectonics: Geol. Assoc. Canada Spec.

Paper 14, p. 617-657.
Naldrett, A. J., and Pyke, D. R., 1976 (in prep.),
Komatiitic rocks from Munro Township, Ontario.
Borisenko, L. F., Miller, A.D., and Fisher, E. I., 1972,

Fominykh,V. G., and Khvostova,V. P., 1970, Platinum

contentof Ural dunite: Akad. Nauk SSSR Doklady,v.
191,no. 2, p. 443-445(Englishtranslation,
p. 184-186).
Ford, A. B., 1970,The development
of the layeredseries
and the cappinggranophyreof the Dufek intrusionsof

The abundanceof gold in ultrabasicrocks: Geokhimiya,

Antarctica:Symposium
on the Bushveld
and otherlayno. 2, p. 188-196 [English translation (1972) Geochemisered
intrusions:Geol.Soc.SouthAfrica Spec.Pub.1, p.
492-510.
try Internat., p. 121-128].
Bowes, D. R., Skinner, W. R., and Skinner, D. L., 1973,
and Boyd,W. W., Jr., 1968,The Dufek intrusion,a
Petrochemistryof the Stillwater Igneous Complex, Monmajor stratiform gabbroicbody in the Pensacolamountana: Geol. Soc. South Africa Trans., v. 76, p. 153-163.
tains,Antarctica: Internat.Geol. Cong.,23rd Prague,
Brooks, C., and Hart, S. R., 1974, On the significanceof
Proc., sec.2, p. 213-228.
komatiite: Geology, v. 2, p. 107-110.
Gansser,A., 1959, AusseralpineOphiolitprobleme:Geol.
Brynard, H. J., de Villiers, J.P. R., and Viljoen, E. A..
Helvetica, v. 52, p. 659-680.

1976,A mineralogical
investigation
of the MerenskyReef
at the Western Platinum mine, near Marikana, South
Africa: Eco.Gv.oL., v. 71, p. 1299-1307.
Cabri, L. J., and Harris, D.C., 1975, Zoning in Os-Ir
alloys and the relation of the geologicaland tectonic environment of the source rocks to the bulk Pt:Pt q-Ir 4-

Os ratio for placers: CanadianMineralogist,v. 13, p.

266-274.

Cabri, L. J., and Laflamme,J. H. G., 1976,The mineralogy

of the platinum-group elements from some copper-nickel
depositsof the Sudburyarea, Ontario: EcoN. Gv.oL., v.
71, p. 1159-1195.

Gass,I. G., 1967,The ultrabasicvolcanicassemblage

of the
Troodosmassif,Cyprus,in Wyllie, P. J., ed., Ultramarie

and relatedrocks: New York, JohnWiley and Sons,p.

121-134.

--

1968,Is the Troodosmassifof Cyprusa fragmentof

Mesozoicoceanfloor?: Nature,v. 220,p. 39.

Gijbels,R., Millard, I-I. T., Jr., Desborough,

G. A., and
Bartel, A. J., 1974, Osmium, ruthenium, iridium and

uranium in silicates and chromite from the eastern Bushveld Complex, South Africa: Geochim. et Cosmochim.
Acta, v. 38, p. 319-337.

P., andZels,J., 1976,Geochemistry

Chyi, L. L., and Crocket,J. H., 1976,Partition of platinum, Gijbels,R., Henderson,
of sometrace elements
in mineralseparates
from Rhum,
palladium, iridium, and gold among coexisting minerals
with special
emphasis
on iridium: Ecoa. GLory.,
v. 71, p.
from the deep ore zone, Strathconamine, Sudbury,On1364-1370.
tario: Ecoa'. GEou.,v. 71, p. 1196-1205.

Clark, S. P., and Ringwood,A. E., 1964, Density distribution and constitution
of the mantle: Rev. Geophys.,
v. 2,
p. 35-88.

Godlevskii,M. N., 1959,Traps and ore-bearingintrusions

of the Noril'skregion: Gosgeoltekhizdat,
Moscow,69 p.
(in Russian).

Collins,W. H., 1934, Life history of the Sudburynickel ---- and Grinenko,L. N., 1963,Somedata on the isotopic
compositionof sulfur in the sulfidesof the Noril'sk deirruptive (1): Royal Soc. [Canada]Trans., sec. 4, v.
posit: Geochemistry,
no. 1, p. 35-41.
28, p. 123-172.
V. M., 1954,Geochemistry:
London,Oxford
Cousins,
C. A., 1964,The platinumdeposits
of the Merensky Goldschmidt,
Univ. Press, 730 p.
Reef, in Haughton,S. H.. ed., The geologyof someore
depositsin southernAfrica: Geol. Stsc.South Africa, Gorbunov,
G. I., 1908,Geologyand origin,f the copperJohannesburg,v. 2, p. 225-237.

--

1973,Noteson the geochemistry

of the platinumgroup

nickel sulfideore depositsof Pecbenga(t'etsamt):

Nedra,Moscov,352 p. (in Russian).

elements:
Geol.Soc,SouthAfricaTrans.,v. 76,p. 77-81. Green,D. H., 1975,Genesis
of Archeanperidotltic
magmas

ULTR.4MAFIC

AND RELATED 3IAFIC ROCKS

and constraints on Archeart geothermal gradients and

tectonics: Geology, v. 3, no. 1, p. 15-18.
and Ringwood, A. E., 1967, The genesis of basaltic
magma: Contr. Mineralogy Petrology, v. 15, p. 103-190.
Green, D. H., Nicholls, L. A., Viljoen, M., and Viljoen, R.,
1975, Experimental demonstration of the existence of
perdotitic liquids in earliest Archeart magmatism: Geology, v. 3, no. 1, p. 11-14.

Guillon, J.-H., 1975, Les massifsperidotitiquesde NouvelleCaledonie type d'appareil ultra basique stratiforme de
chaine rcente: Memoires Orstom no. 76, O.R.S.T.O.M.,
Paris, 120 p.
Hall, A. L., 1932, The Bushyeld Igneous Complex of the
central Transvaal: Geol. Survey South Africa Mere. 28,
560 p.
Hallberg, J. A., and Williams, D. A. C., 1972, Archean
marie

and

ultramarie

rock

associations

in

the

Eastern

Goldfields region, Western Australia: Earth Planet. Sci.

Letters, v. 15, p. 191-200.
Hamilton, W., 1969, Mesozoic California and the underflow

of Pacific mantle: Geol. Soc. America Bull., v. 80, p.

2409-2429.

Haughton, D. R., Roeder, P. L., and Skinner, B. J., 1974,

Solubility of sulfur in mafic magmas: Ecoa. GEo..,v. 69,
p. 451-467.
Hawke, D., and James, R., 1975, The Kanichee layered intrusion: Petrology and geochemistry: Geol. Soc. Amer-

1157

Dept. Energy, Mines and Resources,Ottawa, Canada, v.

42, no. 3, p. 65-69.
Laurent, R., 1975,Occurrenceand origin of the ophiolitesof
southernQuebec, Northern Appalachians: Canadian Jour.
Earth Sci., v. 12, p. 443-455.
Liebenberg, Leon, 1970, The sulphidesin the layered sequence of the Bushyeld Igneous Complex: Geol. Soc.
South Africa Spec. Pub. 1, p. 108-207.
MacGregor, I. D., 1962, Geology, petrology and geochemistry of the Mount Albert and associated ultramafic bodies

of central GaspS, Quebec: Unpub. M.Sc. thesis, Queen's

University, Kingston, Ontario.
Mainwaring, P. R., and Naldrett, A. J., 1974, Genesisof
Cu-Ni sulfidesin the Duluth Complex: Geol. Soc. America
Abstracts with Programs, v. 6, no. 7, p. 854-855.
Maske, S., 1966, The petrographyof the Ingeli mountain
range: Annals Univ. Stellenbosch,v. 41, ser. A, no. 1,
109 p.
ffcCall, G. J. H., and Leishman, J., 1971, Clues to the
origin of Arcbean eugeosynclinalperidotitesand the nature
of serpentinization:Geol. Soc. Australia Spec. Pub. 3, p.
281-300.

McRae, N. D., 1969, Ultramarie intrusions of the Abitibi

area, Ontario: CanadianJour. Earth Sci., v. 6, p. 281-304.

Mertie, J. B., Jr., 1969, Economicgeologyof the platinum

metals: U.S.

Geol. Survey Prof. Paper 630, 120 p.

Moores, E. M., 1969, Petrology and structure of the

Vourinos ophiolitic complex of northern Greece: Geol.

ica Abstracts with Programs, North Central Section, v.
Soc. America Spec. Paper 118, 74 p.
7, no. 6, p. 777.
and Vine, F. J., 1971, The Troodos massif, Cyprus,
Hawley, J. E., 1962, The Sudbury ores, their mineralogy
and other ophiolites as oceanic crust, evaluation and imand origin: CanadianMineralogist, v. 7, p. 1-207.
plications: Royal Soc. [London] Philos. Trans., A, p.
Hess, H. H., 1960, Stillwater IgneousComplex,Montana:
443-466.
Geol. Soc. America Mem. 80, 230 p.
Morse, S. A., 1969, The geology of the Kiglapait layered
Himmelberg, G. R., and Ford, A. B., 1976, Pyroxenes of
intrusion,Labrador: Geol. Soc.AmericaMere. 112, 146 p.
the Dufek intrusion, Antarctica: Jour. Petrology, v. 17, Naldrett, A. J., 1969, A portion of the system Fe-S-O bep. 219-243.
tween 900 and 1,080C and its application to sulfide ore
Hoffman, E. L., 1975, Phase relations of michenerite+
magmas: Jour. Petrology, v. 10, p. 171-201.
merenskyite in the Pd-Bi-Te system; structural refine1973, Nickel sulfide deposits--Their classificationand
ment of chalcanthite and a descriptionof a new Pb-Sn-S
genesis,with special emphasison depositsof volcanic asmineral: Unpubl. M.Sc. Thesis, McGill University.
sociation: Canadian Inst. Mining Metallurgy Trans., v.
76, p. 183-201.
Hsu, K. J., 1968, Principles of m61angesand their bearing
and Arndt, N. T., 1976, Volcanogenicnickel deposits
on the Franciscan-Knoxvilleparadox: Geol. Soc. America -Bull., v. 79, p. 1063-1074.
with someguidesfor exploration: Soc. Mining Engineers
Trans., v. 260, p. 13-15.
Hunter, D. R., 1974, Crustal development
in the Kaapvaal
Craton. 1. The Archean: PrecambrianResearch,v. 1, Naldrett, A. J., and Mason, G. D., 1968, ContrastingArcheanultramaficigneousbodiesin Dundonaldand Clergue
p. 259-294.
townships,Ontario: CanadianJour. Earth Sci., v. 5, p.
Irvine, T. N., 1970, Crystallization sequencein the Muskox
intrusion and other layered intrusions: Geol. Soc. South

Africa Spec. Pub. 1, p. 441-476.

1974, Petrology of the Duke Island Ultramarie Complex, southeastern Alaska: Geol. Soc. America Mere.
138, 240 p.
-1975, Crystallization sequencesin the Muskox intrusion

111-143.

Naldrett, A. J., and Turner, A. R., 1977,The geologyand

petrogenesisof a greenstone belt and related nickel sulfide
mineralization at Yakabindie, Western Australia: Precambrian Research (in press).

Naldrett,A. J., Bray, J. G., Gasparrini,E. L., Podolsky,T.,

and Rucklidge, J. C., 1970, Cryptic variation and the
and other layeredintrusions-II. Origin of chromitelayers
petrologyof the SudburyNickel Irruptive: Ecoa. GoI.,
and similar depositsof other magmatic ores: Geochim.
v. 65, p. 122-155.
et Cosmochim.Acta, v. 39, p. 999-1020.
Naldrett, A. J., Hewins, R. H., and Greenman,L., 1972,
and Findlay, D.C., 1972,Alpine peridotitewith parThe main Irruptive and the sub-layerat Sudbury,Ontario:
ticular referenceto the Bay of Islandsigneouscomplex:
Internat. Geol. Cong., 24th Montreal, proc., sec. 4, p.
Earth PhysicsBr. Pubs., Dept. Energy, Mines, and Re206-214.
sources,Ottawa, Canada,v. 42, no. 3, p. 97-128.
Nesbitt, R. W., 1971, Skeletal crystal forms in the ultraIrvine, T. N., and Smith, C. H., 1967,The ultramaficrocks
mafic rocks of the Yilgarn block, Western Australia:
of the Muskox intrusion,NorthwestTerritories, Canada,
Evidence for an Archeart ultramafic liquid: Geol. Soc.
in Wyllie, P. J., ed., Ultramafic and related rocks: New
Australia Spec. Pub. 3, p. 331-350.
York, John Wiley and Sons,p. 38-49.
Newman, S.C., 1973, Platinum: Inst. Mining Metallurgy
Jackson,E. D., 1961, Primary textures and mineral asTrans., v. 82, p. A52-A65.
sociations in the ultramafic zone of the Stillvater ComPage, N.J., Riley, L. B., and Haffty. J., 1972, Vertical and
plex, Montana: U.S. Geol.SurveyProf. Paper358, 106
lateral variation of platinum, palladium,and rhodium in
p.
the Stillwater Complex,Montana: Ecoa. Gon., v. 67, p.

Jahns,R. H., 1967,Serpentinites

of the Roxburydistrict,
Vermont, in Wyllie, P. J.. ed., Ultramafic and related

rocks: New York, JohnWiley and Sons,p. 137-160.

Johns-Manville
Company
Ltd., 1975,Pressrelease,
Dc. 17,
Denver, Colorado.

Lamarche,R. Y., 1972,Ophiolitesof southernQuebec,in

The ancient
oceanic
lithosphere:
EarthPhysics
Br. Pubs.,

915-923.

Petedefy, W. V., 1972,Chemistryof fluidal glassesand melt

bodies in the Onaping formation: Geol. Assoc. Canada
Spec. Paper 10, p. 49-59.
Phinney, W. C., 1970, Chemical relations between Keewanawan lavas and the Duluth Conplex. Mhmesota: Geol. Soc,

America Bull., v. 81, p. 2487-2495.

1158

.4. J. N.4LDRETT .4ND L. J. C.4BRI

Pyke, D. R., Naldrett, A. J., and Eckstrand, O. R., 1973,

Arcbean ultramafic flows in Munro Township, Ontario:
Geol. Soc. America Bull., v. 84, p. 955-977.
Rae, D. R., 1975, Inclusionsin the sublayerfrom Strathcona
mne, Sudbury, and their significance: Unpub. M.Sc.
thesis, University of Toronto.
Ragan, D. M., 1967, The Twin Sisters dunite, Washington,
in Wyllie, P. J., ed., Ultramafic and related rocks: New
York, John Wiley and Sons, p. 160-167.
Raicevic, D., and Cabri, L. J., 1976, Mineralogy and concentration of Au and Pt from alluvial grayel from the
Tulameen

River

area

in

British

Columbia:

Canadian

Mountain, Oregon: Internat. Geol. Cong., 23rd, Proc.,

ser. 1, p. 175-186.
Vermaak, C. F., The nickel pipesof Vlakfontein and vicinity,
western Transvaal: Ecoa. GFoL.,v. 71, p. 261-286.
and Hendriks, L. P., 1976, A review of the mineralogy
of the Merensky Reef, with specificreference to new data
on the precious-metalmineralogy: Ecoa. GFoL.,v. 71, p.
1244-1269.

Viljoen, M. J., and Viljoen, R. P, 1969, Evidence of the

existence of a mobile extrusive peridotitic magma from
the Komati Formation of the Onverwacht Group: Geol.
Soc. South Africa Spec. Pub. 2, p. 87-112.
Viljoen, M. J., Bernasconi, A., van Coller, N., Kinloch, E.,
and Viljoen, R. P., 1976, The geology of the Shangani
nickel deposit, Rhodesia: Ecoa. Gvo.,v. 71, p. 76-95.
Visser, D. J. L., and yon Gruenewaldt, G., eds., 1970, Symposium on the Bushveld Igneous Complex and other layered intrusions: Geol. Soc. South Africa Spec. Pub. 1, p.

Inst. Mining Metallurgy Bull., v. 69, no. 770, p. 111-119.

Razin, L. V., and Khomenko, G. A., 1969, Accumulationof
osmium, ruthenium and other platinum group metals in
chrome spinel in platinum-bearing dunites: Geochemistry
Internat., v. 6, p. 546-557.
Razin, L. V., Khvostov, V. P., and Novikov, V. A., 1965,
Platinum metals in the essential and accessory minerals of
763.
ultramafic rocks: Geochemistry Internat., v. 2, p. 119-131. yon Gruenewaldt, G., 1976, Sulfides in the Upper Zone of
Ridler, R. H., 1966, Petrographic study of a Crow Lake
the eastern Bushveld Complex: Ecoa. Gvo[., v. 71, p.
1324-1336.
ultrabasic sill, Keewatin volcanic belt, northwestern Ontario: Unpub. M.A.Sc. thesis, University of Toronto.
Vorobeyeva, O. A., Samoilova, N. V., and Sheshnikova, E.
Ruckmick, J. C., and Noble, J. A., 1959, Origin of the ultraV., 1962, Gabbro-pyroxenite-dunite belt of the central
mafic complex at Union Bay, southeasternAlaska: Geol.
Urals: Inst. Geol. Ore Deposits Trans., Petrog. Mineral.
Soc. America Bull., v. 70, p. 981-1017.
Geochem.,v. 65, 319 p. (in Russian).
Sharpe, J. I., 1965, Field relations of Matagami sulfide Wager, L. R., and Brown, G. M., 1968, Layered igneous
masses bearing on their disposition in time and space:
rocks: London, Oliver and Boyd, 588 p.
Canadian Inst. Mining Metallurgy Trans., v. 68, p. 265Walker, Frederick, 1940, Differentiation of the Palisades
278.
diabase,New Jersey: Geol. Soc. America Bull., v. 51, p.
1059-1105.
Shima, H., and Naldrett, A. J., 1975, Solubility of sulfur in
an ultramafic melt and the relevance of the system Fe-S-O:
Walker, K. R., 1969, The Palisades sill, New Jersey, a reEcoN. Gvo..,v. 70, p. 960-967.
investigation: Geol. Soc. America Spec. Paper 111, 178 p.
Stairnov, M. F., 1966, The structure of Noril'sk nickel- Williams, D. A. C., 1972, Archean ultramafic, mafic and asbearing intrusions and the genetic types of their sulfide
sociated rocks, Mt. Monger, Western Australia: Geol.
ores: All-Union
Scientific Res. Inst. Mineral
Raw MaSoc. Australia, Jour., v. 19, pt. 2, p. 163-188.
terials (VIMS), Moscow. 60 p. (in Russian).
Wilson, H. D. B., and Anderson, D. T., 1959, The comSobolev, S. F., Anoshin, G. N., and Perezhogin, G. A.,
position of Canadian sulfide ore deposits: Canadian Inst.
1973, Distribution of gold in rocks of ultrabasic and basic
Mining Metallurgy Trans., v. 62, p. 327-339.
formations in Ural: AN SSSR Izv, Ser. Geol. 10, p.
Wilson, H. D. B., Kilburn, L. C., Graham, A. R., and
68-78 [English translation (1974) Internat. Geology Rev.,
Ramlal, K., 1969, Geochemistry of some Canadian nickeliv. 16, no. 7, p. 800-809].
ferrous ultrabasic intrusions: Ecor. GFoL.MoN. 4, p. 294309.
Stephenson,J. F., 1972, Geology of the Ospwagon Lake
Area: Pub. Manitoba Mines Branch, p. 16-25.
Yoder, H. S., Jr., and Tilley, C. E., 1962, Origin of basalt
Stumpfl, E. F., 1974. The genesis of platinum deposits:
magmas: An experimental study of natural and synthetic
further thoughts: Minerals Sci. Eng., v. 6, p. 120-141.
rock systems: Jour. Petrology, v. 3, p. 342-532.
and Tarkian M., 1976, Platinum genesis: new minYushko-Zakharova, O. Ye., Ivanov, V. V., Razina, I. S., and
eralogical evidence: EcoN. GEOL.,v. 71, p. 1451-1460.
Chernayayev, L. A., 1967, Geochemistry of platinum
Taylor, R. B., 1964, Geology of the Duluth gabbro complex
metals: Geokhymiya, no. 11, p. 1381-1394 (in Russian).
near Duluth, Minnesota: Minnesota Geol. Survey Bull.,
Zolutuchin, V. V., and Vasil'ev, U. R., 1967, Particular
v. 44, 63 p.
characteristicsof formation of some trap intrusives of the
southeast Siberian platform: Moscow, Nauka, 231 p.
Thayer, T. P., and Himmelberg, G. R., 1968, Rock suc(in Russian).
cession in the alpine-type mafic complex at Canyon