Economic Geology

Vol. 80, 1985, pp. 1089-1108

The Nature and Origin of Titaniferous Magnetite-Rich Layers in the

Upper Zone of the BushveldComplex: A Review and Synthesis
IVAN M. REYNOLDS

Departmentof Geology,RhodesUniversity,Grahamstown,6140, Republicof SouthAfrica

Abstract

The occurrenceand geologicrelationshipsof Ti magnetite layers indicate that their

genesisis intimatelyrelatedto fractionalcrystallizationprocesses that were responsiblefor
the formation of their silicate host rocks. The exact mechanismby which these oxide-rich
layers form is, however, still debated. Most recent modelsrecognizethat precipitationof
copiousquantitiesof Ti magnetiteis triggeredby episodicincreasesin fo2, but the process
whereby this occursis not well understood.The nature and occurrenceof Ti magnetite-
rich layers in the BushveldComplex is reviewed and the requirementsof any satisfactory
geneticmodelare outlined.The formationandsubsequent texturalevolutionof Ti magnetite
layersis believed to have taken place as follows:
1. Favorable conditionsfor the precipitationof large quantitiesof Ti magnetitewere
createdby a lengthyperiod of fractionalcrystallizationthat resultedin the concentration
of large amountsof Fe, Ti, and V in the residualmagma.
2. Large-scalein situ bottom crystallizationof plagioclaseresultedin a markedincrease
in the total Fe content and densityof the immediatelysurroundingmelt which collectedas
a denselayer on the bottom of the magmachamber.This denseFe-Ti-(V)-enrichedliquid
did not mix with the overlying magmaand formed a stagnantlayer from which copious
amountsof Ti magnetitemay have crystallized.
3. Crystallizationof Ti magnetiteis controlledby the Fe203/FeO ratio of the liquid and
is a functionof fo2, temperature,and the fH2o/fH•ratio. The Fe203 contentof the liquid
will alsoincreaseduringthe crystallization of ilmeniteand Fe+2-bearing silicates(clinopy-
roxene,pigeonitc,and fayaliticolivine). The HzO contentof the residualliquid will alsobe
increasedby the fractionationof anhydroussilicatephases.
4. A complexinterplayof thesefactorsresultedin the precipitationof relatively large
amountsof Ti magnetiterequired for the developmentof ore-rich layers.The bulk of the
Ti magnetite formed by in situ bottom crystallization.The nucleation and growth of
magnetite within the stagnantlayer and gravitative settling of these crystalsover a short
distance(on the order of meters)may have augmentedthe growth of the layers.
5. Variationsin the relativeproportionsof coexistingoxideand silicatephasesmay be
related to differencesin concentrations and diffusionratesof ionsin the liquid immediately
above the crystallizinglayer. This resulted in differencesin the rates of nucleationand
growthof individualphases.
6. Precipitationof abundantmagnetitelowered the densityof the stagnantlayer until it
reached that of the overlying magma.The two liquids then mixed, thus terminating the
cycle of magnetitelayer formationand markinga return to silicate-dominated fractionation.
7. The ore-rich layersoriginallyconsistedof Ti magnetitecrystalsand variableamounts
of silicatecrystalswhichcontaineda certainamountof interstitialresidualliquid.Densification
of theselayersand expulsionof significantamountsof this liquid were accomplished by
annealingat high subsolidus temperatures,possiblyaugmentedby the additionof suitable
materialfrom interprecipitateliquids.
8. Variationsin fo• during subsolidus coolingresultedin the developmentof a wide
range of ulv6spineland ilmenite microintergrowths.

Introduction 1983a, 1985). Theselayersare of economicinterest

asthey are exploitedfor the productionof vanadium
THE occurrence of discrete vanadium-bearingTi andsteel(Willerase,1969;Luyt, 1976)andrepresent
magnetitelayersin the upper zone of the Bushveld potentialresources for the productionof vanadium,
Complex and in the later crystallizingfractionsof high titania materials,iron, and steel (Reynolds,
numerous other layered mafic intrusions is a well 1978, 1983a, 1985). The distributionand geologic
documentedfeature (e.g., Willemse, 1969; Moly- relationships of theseoxide-richlayersclearlyindi-
neux, 1970a and b; Reynolds, 1978, 1979a and b, catethatthey are integralcomponents of the layered

0361-0128/85/418/1089-2052.50 1089
1090 I•',4N M. REYNOLDS

mariesequence andtheir genesis isthereforedirectlyA rare exceptionto this rule is affordedby layer 13
relatedto processesthat were operativeduringlater in the easternBushveldwhich exhibitsa gradational
stagesof fractionalcrystallization.The exact mech- lower contactand a sharpupper contact(Von Grue-
anism by which they developed is, however, still newaldt, 1973). The hanging-walland footwallrocks
debated. The aim of this contribution is to review of the ore-richlayersare commonlyplagioclase-rich
the availableinformationon the nature of layered varietiessuchasanorthositeand troctolite,although
Fe-Ti oxide ores and to present a genetic model gabbroic varieties are also present. The opaque
that can accountfor manyof their primaryfeatures oxide content of the associated silicate rocks is
and subsequentsubsolidusmodifications. highly variable and rangesbetween nil and 50 vol
percent. The anorthositicfootwall rocks of the
Field Relationships thicker and more massiveTi magnetitelayers also
commonly display varying degrees of late stage
Field relationshipsof BushveldTi magnetitelayers alteration and may show local sulfide enrichment.
havebeen describedby Wagner(1928), Hall (1932), Individual ore-rich layers are rarely composed
Coertze (1966), Willemse (1969), Molyneux (1970a entirely of Fe-Ti oxidesandcertainlayers,or portions
and b), and Von Gruenewaldt(1973) and only the of layers, may be particularly silicate rich. The
essential features are summarized here. The Fe-Ti dominantsilicatemineral is plagioclase,which may
oxide ores form discrete layers that vary between be present either as isolatedtabular crystalsor in
0.1 and 10.0 m in thickness and are concordant with larger aggregates.Thesecrystalscommonlyexhibit
the igneouslayeringof the upper zone. The conti- a distinct preferred orientationin which the long
nuity and lateral extent of many of the individual axesof tabularcrystalslie parallel or subparallelto
layersis unfortunatelynot known becauseof poor the lower contactsof the ore-richlayers,thusgiving
exposures.The economicallyimportant Main Mag- rise to an igneous lamination. Inverted pigeonitc
netite layer variesbetween 1.0 and 2.5 m in thickness and/or augitc may also be present as inclusionsin
and exhibits a strike length of approximately200 the magnetitelayers. Narrow lenslike, or more con-
km in the western BushveldComplex and 120 km tinuousplagioclase-richzones,or anorthositicpart-
in the easternBushveldComplex(Willemse, 1969). ingsare alsodevelopedwithinmanyore-richlayers.
The equivalentof this layer exhibitsa strike length For example,the Main Magnetitelayer in the eastern
of approximately100 km in the Potgietersruslimb Bushveldis split into two roughlyequalportionsby
of the complex(Van der Merwe, 1976). a remarkablypersistentnarrow anorthositicparting
Mapping in well-exposedareas of the eastern that varies between 0.15 and 0.3 m in thickness.
BushveldComplexhasrevealedthe presenceof at The uppermostTi magnetitelayer in the samearea
least 30 Ti magnetite layers in the upper zone (layer 21) is a compositehorizon that consistsof a
(Molyneux, 1970a). The most continuousof these large number of relatively thin ore-rich layers that
have been assignednumberswith the prominent are separated by narrow lenses and partings of
Main Magnetite layer being taken as the datum. olivine-bearinganorthositeand troctolite.
Four layers,termed the lower layersand numbered
i to 4, are presentbelow the Main Magnetitelayer; Ore Petrography
those above this level have been numbered 1 to 21
in order of appearancewith increasingstratigraphic Silicate-freeportionsof individualore-richlayers
height (Molyneux,1970a). Certain of the numbered are composedalmost entirely of Ti magnetite and
layers are compositeand consistof two or more do not usually contain more than approximately6
discreteore-richlayersthat are separatedby narrow vol percentgranularilmenite (Molyneux,1970a and
layers of silicate rock. Idealized sectionsand de- b; Reynolds, 1985). The constituentTi magnetite
scriptionsof individuallayers have been published crystalscommonlyexceed 10 mm in diameter. This
by Molyneux(1970a) andVon Gruenewaldt(1973). is three to ten times greater, on average,than the
The majority of these layers are relatively thin and grainsizeof Ti magnetitein silicate-bearingportions
consist of less than 30 cm of fairly massiveTi of the samelayer or in the associatedcountryrocks.
magnetite.The important exceptionsto this are the The silicate-freeores therefore form massivepoly-
Main Magnetite layer and the very thick (10-m), crystalline aggregatesin which the constituentTi
layer 21 which can alsobe regardedas a composite magnetitecrystalsmeet in well-definedtriple junc-
body. tionswith interfacialanglesof 120ø (Fig. 1A). Grain
The lower contactsbetweenindividuallayersand boundariesare often straightto gently curved but
their footwallrocksare generallysharpbut are often may exhibit greater curvaturein areaswhere there
dimpled or gently undulatingon a scaleof several is a pronouncedvariationin grain size (Fig. 1A and
centimeters. The upper contacts,in contrast, are B). Smallpolygonalilmenitecrystalsmaybe located
usuallygradationalin nature and reflect a gradual interstitiallybetweenthe larger Ti magnetitecrystals
decreasein the opaqueoxide content of the rocks. and stringersof very muchsmallerilmenite crystals
 REVIEW OF MAGNETITE-RICHLAYERS 1091

ß 2mm .

FIG. 1. Camera lucida drawing of typical grain boundary relationshipsin Bushveldtitaniferous

iron ores.A. Coarse-grained,well-annealedore from the Main Magnetite layer showingcurved grain
boundariesand triple junctions. B. Ore containing small amountsof interstitial silicate that is located
along grain boundaries and at triple junction points. C. Sample containing a narrow silicate-rich
stringer. Note the abrupt decrease in grain size of the oxides in this area. D. Contact between
massiveore and a narrow layer containing a high proportion of interstitial silicates, illustrating
variationsin grain size and grain boundary relationships.E. Massive,well-annealed ore containing
numbersof clinopyroxenecrystalsthat encloselarge numbersof tiny oxide crystals.F. Ore containing
numbersof coprecipitatedplagioclasecrystalsand showingtypical grain boundaryrelationships.Ti
magnetite = white, ilmenite = black, silicates-- shaded.

may be present along grain boundaries.These ores which is used here without any genetic connotation,
may therefore be classifiedas adcumulusrocks ac- as recommendedby Irvine (1982).
cordingto the terminologyof Wager et al. (1960) Boththe grainsizeof Ti magnetiteanddegreeof
1092 •VAN M. REYNOLDS

curvatureof grainboundariesis highly variableover the immediatevicinity of silicatemineralinclusions

shortdistanceswithin individuallayers(Fig. 1A-F). and the normalpolygonaltexture is not as evident.
These variationscan be related directly to the pres- Well-definedtriple junctionswith interfacialangles
ence of silicatemineralsor ilmenite crystalsand two of 120ø are, however, alwaysdevelopedat mutual
basictypes of relationshipexist. contactsbetween Ti magnetitecrystals.
Ti magnetite in ore-rich layers is characterized
Interstitial phases by the developmentof a wide range of ilmenite-
Small amountsof plagioclase,inverted pigeonite, and/or ulv6spinel-bearingmicrointergrowthsthat
and augitecommonlyoccupysmalltricuspateareas developed during subsoliduscooling.The nature
at triple junction pointsbetweenTi magnetitecrys- andrelativedevelopmentof thesemicrointergrowths
talsor are presentasthin filmsalonggrainboundaries reflect variationsin fo• at various stagesduring
(Fig. lB). This materialmay representthe crystalli- subsolidus cooling (Willemse, 1969; Molyneux,
zation products of entrapped intercumulusliquid 1970aandb; Reynolds,1983a, 1985). The sequential
that was not expelled during adcumulusgrowth developmentof theseintergrowthsand the contem-
processes. The grain size of Ti magnetiteis notably poraneousoxidation-exsolution of ilmenite from
smallerin areascontaininglarger amountsof inter- magnetite-ulvbspinel solid solutionshas been de-
cumulusphases,particularly where these partially scribedby, e.g., Buddingtonand Lindsley(1964),
surroundedoxide crystals(Fig. lB). The presence Duchesne(1970), Haggerty (1976), and Reynolds
of stringersof small ilmenite crystalsalong grain (1983a, 1985).
boundariesalso correlateswith a decreasein grain Exsolved transparentspinelsof pleonaste-hercynite
size. composition are commonin Ti magnetitecrystalsin
The transition between zones of massive, coarse- certainore-richlayers,particularlythosetowardthe
grained Ti magnetite ore and horizonscontaining baseof the upper zone (Willemse,1969; Molyneux,
abundant intercumulus phases is abrupt and is 1970a and b, 1972; Reynolds,1983a, 1985). These
markedby a significantdecreasein grain size (Fig. exsolution bodies also exhibit wide variations in
1C and D). Furthermore, the Ti magnetite crystals morphologyand mode of occurrencewhich have
in these intercumulusareas commonlyexhibit eu- beendescribedby, e.g., Haggerty(1976) andBowles
hedral or subhedral morphologieswhen they are (1976, 1977).
enclosedby silicatephases.In contrast,Ti magnetite In contrast to the situation in ore-rich layers,
crystals invariably adopt polygonal morphologies ilmenite is commonlythe dominantoxide phasein
and show the developmentof well-defined triple rocks containingless than approximately20 vol
junctionsat mutual contactswith other Ti magnetite percentcombinedopaqueoxides(Reynolds,1983a,
crystals.The grain size of Ti magnetitein areas 1985). An antipatheticrelationshipappearsto exist
containingabundantinterstitialphases(0.5-2.0 mm) between ilmenite and Ti magnetite in which the
is generallyin the samesizerangeasthat of dissem- volume of coarse-grainedilmenite decreasesmark-
inated Ti magnetitein associated countryrocks. edly as the total volumeof opaqueoxideincreases.
Poikilitic inverted pigeoniteand/or augite grains Coarse-grained ilmenite is normallyonly presentin
up to 5 mm in size are sparselydistributedwithin minor amountsin the massiveore-rich layers(Reyn-
certain areasof otherwisemassiveTi magnetiteores. olds, 1983a, 1985). Ilmenite crystalsin silicate-rich
These pyroxene grains commonly contain large rocksmay exhibitsubhedralmorphologies, but the
numbersof very muchsmallersubhedralto euhedral associated Ti magnetitegrainsare invariablyanhe-
Ti magnetitecrystalsthat are commonlypresentin dral,exceptin caseswheresmallsinglecrystals
may
chainlikeaggregates(Fig. 1E). Precipitationof host be enclosedby silicateminerals.
pyroxenesaroundthese magnetitecrystalsappears The opaqueoxideaggregates in silicate-rich
rocks
to havetaken placeat a relativelyearly stageduring are locatedinterstitiallybetweenlarge cumulussil-
crystallization icate crystalsand appearto be intercumulusphases.
of the interstitialliquid.Thisprotected
them from further modification during subsequent The intercumulusappearanceof the oxideaggregates
developmentof the monomineraliclayer and they commonlypersistsasthe volumesof oxideincrease
are probablyrepresentativeof the grain size of to in excessof 40 percent. This may, however,
primary Ti magnetitecrystals. simplyreflectthe presenceof extensivepostcumulus
overgrowthson cumulusoxide grains.The oxide
Coprecipitatedphases aggregatesthemselvesalso characteristically show
Ti magnetite also commonlyexhibits a marked the development of polygonalgrainboundaries and
decreasein grain size in the immediate vicinity of 120ø triple junctionanglesat mutualcontactswith
silicate or ilmenite inclusions where the ore exhibits oxides. This is in marked contrast to the irregular
a great range in grain size (Fig. 1F). Ti magnetite grainboundariesthat are developedbetweenoxides
grain boundariesare also more stronglycurved in and silicategrains.
 REVIEW OF MAGNETITE-RICH LAYERS 1093

Chemistry netitccrystalson the consolidated

floorof the magma
chamber.
Few completemajor and minorelementanalyses
of BushyeldTi magnetiteoreshave been published Hall (1932) proposedthat a dense "magnetite"
(Wagner, 1928; Schwellnusand Willemse, 1943; liquid might separatefrom the crystallizingmagma
Molyneux, 1970b, 1972) and virtually nothing is and accumulateon the floor of the magmachamber.
knownabouttheir traceelementgeochemistry. The Any unconsolidated plagioclasecrystalson the floor
behavior of Cr in Ti magnetite concentrateshas of the chamber would float upward through the
been studiedby Cawthornand McCarthy (1980, oxide liquid and become concentratedtoward the
1981) andsimilarstudiesof Cr andV are in progress top of such layers, thereby accountingfor their
(McCarthy et al., 1983; Walsh et al., 1983). plagioclase-rich,gradationalupper contacts.Subse-
The available data indicate that the bulk TiO2 quent crystallizationof thesemagnetiteliquidsthen
contentof the variousore-richlayersand individual gave rise to ore-rich layers. Bateman(1951) elabo-
rated on these ideas and concluded that concordant
Ti magnetite crystalsincreasesprogressivelywith
increasingstratigraphicheight from the baseof the Fe-Ti ores were formed by late gravitativeaccu-
upper zone (+_12%TiO2) to the top of the layered mulationsof oxideliquidson top of the consolidated
sequence(+_20%TiO2) (Schwellnusand Willemse, crystalpile. This liquidmightalsohavebeen injected
1943; Willemse, 1969; Molyneux, 1970b, 1972; into the country rocksto form discordantbodies.
Reynolds, 1983a, 1985). Vanadium exhibits an an- Lister (1966) also concluded that the concordant
tipathetic relationshipto TiO,on a grossscale types formed via the separationand accumulation
(+_2.0%V,• at the baseto 0.2% V,• at the top), of an immiscible Fe-Ti oxide liquid. In contrast,
but Walsh et al. (1983) have shownthat this picture Coertze (1966) postulated that the Bushveld Ti
is more complex in detail. magnetitelayerswere derived from the crystalliza-
Distributionof the minor elementsMg and Mn is tion of an oxideliquid that wasintruded from below
controlledlargely by their preferentialpartitioning and spread laterally along certain lithological con-
tacts.
into ilmenite relative to Ti magnetiteduring crys-
tallizationandsubsequent reequilibration Separationand accumulationof immiscibleFe-Ti
subsolidus
(e.g., Vincent and Phillips, 1954; Neumann, 1974; oxideliquidremainsan attractivehypothesis because
Himmelbergand Ford, 1977). This primary distri- it can explain many of the field and textural rela-
bution patternis, however,complicatedby reequi- tionshipsof theseores.Thishypothesis alsoexplains
librationof theseelementswith coexisting
pyroxene manyof the featuresof apatite-Fe-Tioxideoresand
and olivine(Speidel,1970; Andersenand Lindsley, the existenceof liquid immiscibilityin such phos-
1979; Bishop,1980; Reynolds,1983a). phorous-richsystemsis supportedby experimental
evidence (Philpotts,1967; Kolker, 1982; Reynolds,
Earlier Genetic Models 1983a). Unfortunately,only limited experimental
evidence exists to support the existenceof Fe-Ti
By the turn of the centuryit waswidelyacknowl- oxideliquid immiscibilityin petrologicallyrealistic,
edged that concordantTi magnetitelayers in the phosphorous-poor, ferrobasaltic-dioritic
silicatesys-
Bushveld Complex and other gabbroic intrusions tems(Naslund,1976; Roedder,1978, 1979). Further
represented"magmaticsegregationdeposits"(Mo- investigationsin this direction are clearly needed
lengraaff,1904; Singewald,1912). The earlier mod- and this particular genetic model must therefore
els acknowledgedthe existence of genetic links remain speculative,but neverthelesspotentially im-
between concordantlayered Fe-Ti oxide ores and portant.
their host rocks and either invoked mechanisms
involving the separation and accumulation of Ti Later Genetic Models
magnetitecrystalsto form layers or postulatedthe
existenceof discreteFe-Ti oxideliquidsfrom which Virtually all later genetic modelsinvoke mecha-
the orescrystallized.Du Toit (1918) suggestedthat nismswherebyepisodicincreasesin fo2 triggerthe
suchores might have crystallizedfrom Fe-Ti oxide crystallizationof copiousquantitiesof Ti magnetite
liquidsthat were injectedinto partiallyconsolidatedthat are required for the development of ore-rich
host rocks at a late stage during crystallization. layers.The formulationof thesemodelswaspreceded
Osborne(1928) andDunnandDey (1937) suggested by experimentalstudiesof syntheticFe-bearingsys-
that concordantFe-Ti oresformedby an accumula- tems which indicatedthat fo2 is an importantcon-
tion of late-stageresidual Fe-Ti oxide liquid that trolling factorin the crystallizationof magnetiteand
was filter pressedfrom the surroundingpartially other Fe-bearing phases(e.g., Muan and Osborn,
consolidated rocks.In contrast,Wagner(1928) sug- 1956; Osborn, 1959; Roeder and Osborn, 1966).
gestedthat Ti magnetitelayersformedby the sep- These studiesindicated that the Fe•3/FeO ratios
aration and gravitationalaccumulationof Ti mag- of magmasare, to a large extent, dependentonfo2
1094 IVAN M. REYNOLDS

and that magnetiteprecipitationwould be enhanced in suddencompaction of the semiconsolidated

crystal
by an increasein the amountof Fe+a in the melt. mushtowardthe top of the solidifyingcrystalpile.
Ulmer (1969) and Cameron and Desborough The suddenintroductionof significantamountsof
(1969) demonstratedthat increasesin fo2 might be oxygenthen resultsin the precipitationof copious
an important factor in controllingchromite precipi- amountsof Ti magnetite at or near the base of the
tation to form analogousore-rich layers. Molyneux magmacolumn. Accordingto this model the thick-
(1970a and b) suggestedthat crystallizationof Ti nessof individual ore-rich layers and the Ti mag-
magnetite might take place near the floor of the netite/silicate ratios of associatedrocks are depen-
magmachamberin responseto increasedfo2. Hill dent largely on the volumesof intercumulusliquid
and Roeder (1974) demonstrated experimentally that were expelled. This is, in turn, a function of
that precipitationof Ti magnetitefrom basalticliquid the intensity of the seismicevent, volume, andfo2
is controlledlargely by fo• and that it is alwaysa of the liquid that is availablefor expulsionand the
late crystallizingphase.They alsoconfirmedOsborn's degree to which expulsioncan take place.
(1959) earlier observationsthat marked Fe enrich- The major problem faced by the majority of the
ment of the residualliquid will occur during pro- fo•-triggered hypothesesis that oxidation events
longed fractional crystallizationunder low or de- must have been operative more or less simulta-
creasingfo• conditions. neouslyover the entire width of the magmachamber
Irvine (1975) suggestedthat Ti magnetite-rich to allow for the uniform thickness and lateral con-
layersmight developin responseto episodicoxida- tinuity of individual ore-rich layers. This would
tion eventsthat were causedby extensivecontami- therefore rule out mechanismsinvolvinga localized
nationof the basicparentalmagmaby graniticliquid processsuch as contamination.More widespread
derived from the melting of sialicroof rocks.Irvine processessuchas pressurechangescausedby frac-
(1977) alsosuggestedthat an influx of fresh, undif- turing through to the surface(Ulmer, 1969) would
ferentiated magma into the systemmight produce obviouslybe more important in this respect. Simi-
similarresults.Episodicincreasesin fo2 as the trig- larly, magma-mixingmodelscannot effectivelyac-
gering mechanismfor the precipitationof copious count for the common presence of numerous Ti
amountsof Ti magnetitehave alsobeen appealedto magnetite-richlayerswithin smallvertical intervals
by Reynolds (1979b, 1980, 1981). Klemm et al. and variations in silicate mineral content within
(1982) have speculatedthat Ti magnetitelayer for- individual layers.
mation might result from higherfo• causedby the
presenceof volatiles derived from floor rocks that FactorsInfluencing the Formation of Ti
are undergoingthermal metamorphismbelow the Magnetite-Rich Layers
intrusion.
Fractional crystallization
Reynolds(1985) proposedthat the expulsionof
highly fractionatedinterstitialliquid from the semi- The restrictionof discreteTi magnetitelayersto
consolidatedcrystalpile into the base of the over- the later crystallizingportionsof the BushveldCom-
lying magma column might represent an effective plex and other layered mafic intrusionsclearly in-
mechanismfor the introductionof additionaloxygen dicatesthat extensiveperiods of Fe-Ti-(V) enrich-
into the system. This liquid would be gradually ment of the residualliquid occurredbefore Ti mag-
expelled from the crystal pile during the upward netite precipitation was initiated. This is also
displacement of part of the intercumulus phase supportedby the experimentaldata of Roeder and
during solidificationin responseto compaction,in- Osborn (1966) and Hill and Roeder (1974). The
filtration metasomatism (Irvine, 1978, 1980a), and/ marked degree of late-stageFe enrichment that is
or annealing (Reynolds, 1979a). It was postulated required for the precipitationof copiousamountsof
that the water contentof this liquid mightbe higher Ti magnetiteis normally only encounteredin frac-
than that of the overlyingmagma(due to the crys- tionated basalticrocks of tholeiitic character(e.g.,
tallization of anhydroussilicate phases)and that it Walker and Poldervaart, 1949; Osborn, 1959, 1962;
might have been further increasedby the introduc- Carmichaelet al., 1974), larger differentiatedtho-
tion of additionalwater from the developmentof a leiitic intrusions(Ealesand Robey, 1976), and lay-
meteoric hydrothermalsystemaroundthe intrusion ered maficintrusionsof inferred tholeiitic parentage
during the later stagesof crystallization(Taylor and suchasthe Skaergaardintrusion(Wager and Brown,
Forester, 1979). The requiredfo• increasewas na- 1968). The grosschemicalcompositionof the frac-
ively ascribedto the dissociationof H20 followed tionatingmagmatherefore representsan important
by diffusion-controlledhydrogen loss. Reynolds primarychemicalcontrol.Vincent (1960) suggested
(1985) concludedthat episodicincreasesin the rate that the late appearanceof primary Ti magnetite
of interstitialliquid expulsionmight havebeen trig- and ilmenite in the Skaergaardintrusion might in-
gered by earthquake-generated shockthat resulted dicate that elements such as Fe +a and Ti had to
 REVIEW OF MAGNETITE-RICH LAYERS 1095

become concentrated to certain levels in the residual and Roeder, 1974) and the compositions of copre-
liquid before crystallizationcould occur. This was cipitatingTi magnetite and ilmenite are alsodeter-
supportedby Wager (1960) who demonstrated that mined by theseparameters(Buddingtonand Linds-
levelsof 3.5 wt percentFe203 and 2.8 wt percent ley, 1964). They are invariably late-crystallizing
TiO2 were attainedbeforeTi magnetiteandilmenite phasesand the nucleation of ilmenite in basaltic
were able to coprecipitatefrom the Skaergaard rocksnormallyprecedesthat of Ti magnetite(Evans
magma.CarmichaelandNicolls(1967) subsequently andMoore, 1968; EgglerandBurnham,1973; Eales
confirmedthat relativelyhigh levelsof Fe•O3 (1.2- et al., 1980; Reynolds,1983b). Precipitationof Ti
3.5 wt %) andTiO• (0.9-1.2 wt %) mustbe present magnetitein tholeiiticrocksis usuallyinitiatedafter
in a tholeiiticliquidbeforeTi magnetiteandilmenite a lengthy period of silicate fractionationduring
will coprecipitateasprimaryphases. which no primaryspinelis formed.This corresponds
Eales et al. (1980) have shown that the well- to the "spinelgap" of layeredmaficintrusionsand
defined,late-stageFe-enrichmenttrend of tholeiitic is represented by the main zone of the Bushveld
Karoorocksis alsomarkedby a sympathetic increase Complex.The spinelgap is causedby the cessation
in TiO• and a slight enrichment in V. Zirconium of chromespinelprecipitationat a relatively early
behavesas an almostideally incompatibleelement stagein the crystallizationhistory as a result of Cr
in this suiteandcanbe usedto calculatethe weight depletion which is usually ascribedto the onset of
fraction of residualliquid that each analyzedrock clinopyroxenecrystallization(Irvine, 1967; Hill and
represents(Ealesand Robey, 1976). Consequently Roeder,1974) andis clearlyillustratedby the Karoo
the degreeof crystallizationcan be calculatedfrom tholelites(Fig. 2).
the Zr dataon the assumption of Rayleighfraction- The absenceof Fe-bearing spinelswithin the
ation and a plot of whole-rock levels of TiO•, V, spinelgap hasbeen ascribedto the existenceof low
and Cr againstpercentagecrystallizationis shown fo2 during this stage of crystallization(Hill and
in Figure 2 (Ealeset al., 1980). These calculations Roeder, 1974; Eales et al., 1980). It is however,
indicate that TiO2 levels increasesteadilyin the possiblethat temperaturemay play an important
residualliquid duringfractionalcrystallizationdue role.Completesolidsolutionexistsbetweenchromite
to the lackof spinelor ilmeniteprecipitation between andTi magnetiteat magmatictemperatures(Stevens,
approximately30 and 70 percent crystallization. 1944; Evans and Moore, 1968; Lindsley, 1976;
Vanadiumshowsonly a slightincreasein abundance Eales, 1979) in which Ti magnetite (magnetite-
in the residualliquid over this interval, probably ulv6spinel solid solution) is the low-temperature
due to the incorporation of someV into clinopyrox- component.Depletion of Cr from the systemwill
ene. Both TiO• and V showa rapid depletionfrom thereforeresultin a cessationof spinelprecipitation
the melt beyond 70 percent crystallizationdue to until the temperatureof the residualmelt hasfallen
the onsetof ilmeniteandTi magnetiteprecipitation sufficientlyto intersectthe stabilityfield of essentially
(Ealeset al., 1980; Reynolds,1983b). Cr-free, Ti magnetite. The onset of Ti magnetite
precipitation at the base of the upper zone may
Crystallizationof Ti magnetiteand ilmenite thereforereflect a significantoveralldecreasein the
Crystallization of Ti magnetitefrombasalticliquid temperature of the residual melt. For example,
is controlledlargely by temperatureand fo• (Hill chromite precipitationis usually completedat ap-
proximately 1,150øC under lower fo• conditions
(Hill and Roeder, 1974), whereasilmenite precipi-
tation commencedat 1,070øC and Ti magnetiteat
Crystalhzahon
of CrystaB.
izahon 1,050øC in the Alae lava lake (Peck et al., 1966).
v

200-
Cr'TL--m-•-'g-n
e-•t--e-
Sp,ne[
6•p_o
f_
_C_r
:sj•
i_n
e_l._s
_•J The upward increasein bulk TiO• content of Ti
magnetite in the BushveldComplex may also be
related to progressive fractional crystallization.
ppm TiO2 Magnetiterepresentsthe high-temperaturemember
% of the magnetite-ulvbspinel solid solutionseriesso
lOO
that progressivefractionalcrystallizationmight be

200
//
lOO
, ,r . ,
50
,
•_•.j1
. • ,
expectedto yield an increasein ulv6spinelcontent.
This relationshipis, however,complicatedbecause
the ulv6spinelcontentof magnetitethat is coprecip-
PERCENTAGE CRYSTALLIZED itating with ilmenite is determinedby both temper-
•'•0. 2. A derivative variation diagram for Karoo tholeiites
ature andoxygenfugacity(BuddingtonandLindsley,
showingthe concentrations of TiO2, V, and Cr at variousstages 1964). The ambientfo2 at the time of crystallization
of crystallizationascalculatedfromZr datawith the assumption cannot be determined directly by means of the
of Rayleighfractionation(modifiedfrom Ealeset al., 1980). Buddingtonand Lindsley (1964) geobarometerbe-
1096 IV.4N M. REYNOLDS

causeoriginal liquiduscompositionshave been sig- %-

nificantlymodifiedby subsolidus
reequilibrationand -8
oxidation effects.
Coarse-grained,possiblycoprecipitated,ilmenite -9
is normallyonly a minor constituentof the Bushveld
ore-rich layers and does not contribute significant
TiO2 to bulk chemicalanalysesof such ores. This ø•-10
impliesthat the bulk of the TiO2 reportedfor such
analyses(Schwellnusand Willemse, 1943) or for Ti
magnetiteseparates(Molyneux, 1970b, 1972) must -12

originally have been held in solid solutionas ulv6-

spinel componentsin the primary Ti magnetite -13
/•
phases(Reynolds,1983a, 1985). On the basisof this
,

assumptionit is clear that the compositionsof the -14

primary Ti magnetite precipitates have changed

systematically from approximatelyulv6spine13,•_40at 900 1000 1100
the base of the upper zone to ulv6spine170_7.• near TEMPERATURE
*œ
the top of the upper zone, i.e., over a vertical FIG. 3. A plot of log•Co• vs. T illustrating the approximate
thicknessof approximately2,000 m. The experi- fo• trajectory of the Bushveldresidual magma during crystalli-
mental data of Buddingtonand Lindsley (1964) zation of some2,000 m of cumulates.The postulatedtrajectory
clearly indicatethat this degree of Ti enrichmentof is based on the inferred bulk compositionsof Ti magnetite
coexisting with ilmenite. Magnetite-ulv6spinel compositional
magnetitein the presenceof ilmenite can only occur contours are from Spencer and Lindsley (1981). Ilmenite com-
in responseto a significantdecreasein the fo2 at positional contours are omitted for clarity. T and fo• conditions
constantor decreasingtemperature. at which magnetite precipitation commenced are as follows: A
The temperatureat which primary Ti magnetite -- experimental data of Hill and Roeder (1974), B = direct
crystallization was initiated is not known, but measurementsin Alae lava lake (Peck et al., 1966), C = estimates
by Evans and Moore (1968).
the available data on spinel crystallization in
basalticrockssuggestthat thiswouldmostprobably
have occurredbetween approximately1,025ø and 1,000 m of the layered marie suite) (Molyneux,
1,100øC dependingon pressure,fo2, and composi- 1970b; Von Gruenewaldt,1973; Reynolds,1983a).
tion (e.g., Evans and Moore, 1968; Eggler and Olivine is also locally abundantin the underlying
Burnham,1973). Similarly,the temperatureat which subzone C and below the Main Magnetite layer.
primary Ti magnetitecrystallizationceasedafter the The decreasein fo2 with progressivefractionalcrys-
precipitationof some2,000 m of cumulusrocksmay tallization appearsto reflect a marked depletion in
have been in the order of 50 ø to 100øC lower. The Fe•O3 from the melt as a result of the precipitation
marked changein Ti magnetite compositionduring of copiousquantitiesof magnetite. This Fe203 de-
thisintervalindicatesthat thefo• of the fractionating pletionwouldhavesignificantlyloweredthe overall
liquid would have followed a path similar to that oxygencontentof the magmaand would have con-
indicated in Figure 3. The absolute temperature tributed to a lowering of the ambientfo•.
limits are not known, but higher or lower values The relative abundanceof Ti magnetite(typically
will only serveto alter the slopeof thefo• trajectory,5-10 vol %) in the upper zone silicate rocks is
e.g., a magnetite crystallizationinterval between variable over short distances,but the overall content
1,100 ø to 950øC will result in a decrease in the remains constant with increasingheight until ap-
fo• of approximately 4.5 logunitswhereasa probably proximately200 m below the roof where its abun-
more realistic crystallizationinterval of 1,075ø to dance decreasesabruptly (Molyneux, 1970b; Reyn-
975øC will resultin a decreasein thefo• of approx- olds, 1983a). The increasingTi contentof the mag-
imately 3.5 log units.The importantfeature of this netite with progressivefractionalcrystallizationand
diagram is that the fo• appears to decreaseat an decreasingfo• suggeststhat significantTi depletion
appreciablygreaterrate with decreasingtemperature of the melt did not take place during the earlier
than the 1.5 log unitsper 100øC fall in temperature stagesof magnetitefractionation.This is in marked
in systemsthat are crystallizingat a constantoxygen contrast to the behavior of V which exhibits a
content (Eggler and Burnham,1973; Hill and Roe- relativelyhigh distributioncoefficientfor magnetite
der, 1974). relative to silicateliquid (D = 24-63, Ewart et al.,
The marked decreasein the fo• toward the later 1973). A relativelyrapid overalldepletionof V from
stagesof crystallizationis alsoreflectedin the pres- the residualliquidwouldthereforebe expected.The
enceof abundantcumulusfayaliticolivinethrough- V205 content of the Ti magnetite decreasespro-
out subzoneD of the upper zone (the uppermost gressivelyfrom approximately2.1 percentin one of
 REVIEW OF MAGNETITE-RICH LAYERS 1097

the lowermostlayersto approximately1.5 percent crystallizationfrom limited volumes of magma in

toward the top of the Main Magnetite layer, a convectivecells at the baseof the magmacolumn.
verticaldistanceof approximately
135 m (Willemse, Late-stageiron enrichmentof fractionating
basaltic
1969; Molyneux, 1970a and b). Thereafter the V205 magmasis also accompaniedby an increasein the
contentdecreasessteadilyuntil a low value of 0.2 densityof their residualliquids(StolperandWalker,
percent is attained in the uppermostlayer (layer 1980; Sparks and Huppert, 1984). This density
21), a vertical distanceof approximately865 m increasebecomesparticularlymarkedwhen plagio-
(Molyneux,1970b).Thissteadyoveralldeclinein V claseis an importantcrystallizingphase(Sparksand
content rather than a sudden decrease as reflected Huppert, 1984). Plagioclaseis an importantcumulus
in Figure 2 for the Karoo rockscombinedwith the mineral (approximately55-60 vol %) throughout
behaviorof TiO2 suggeststhat either the residual the entire underlyingmainzone aswell asthe upper
liquid may have been augmentedby additionalin- zone (Molyneux, 1970b; Von Gruenewaldt, 1973)
fluxesof unfractionatedmagmaor thatothercomplex and it is probablethat the densitiesof the residual
processes were operativeduringcrystallization.Un- liquidsfromwhichthe upperzonecrystallized would
fortunately,the detailed mineralogicaland geo- have exceeded those of an unfractionated basaltic
chemicaldata that are necessaryto resolvethese magma.The in situbottomcrystallizationof plagio-
possibilitiesare not yet available. clase would have resulted in the generation of
quantitiesof dense,Fe-enrichedliquid in the im-
Developmentof layering mediatevicinity of the crystallizationsites.Sucha
liquid will not be able to mix freely with the over-
The developmentof monomineralic layerswithin lying magma and will tend to accumulateon the
stratiformbasicintrusionshasbeen the subjectof floor of the magmachamberwhere it will form a
considerabledebate. The general consensus,until stagnantlayer (Sparks et al., 1980; Sparks and
recently,wasthat theserocksformedby magmatic Huppert, 1984; Jaupartet al., 1984). The formation
sedimentationprocessesthat resulted in the accu- of such a stagnantlayer would depend largely on
mulationof precipitatedcrystalsin layers on the the relative volumesof any Fe-rich phasessuchas
floor of the magmachamber(Wager and Brown, ilmenite, Ti magnetite,and/or fayaliticolivine that
1968). Theselayersmay have formedby a simple were coprecipitatingwith the plagioclase.Crystal-
processof gravitativecrystalsettling(Wageret al., lization of these denserphaseswould contributeto
1960; Irvine, 1979) or by a variety of other mech- a decreasein the densityof the residualmelt (Stolper
anisms,includingconvectiveoverturn(Hess,1960), and Walker, 1980; Sparksand Huppert, 1984) and
crystallizationfrom a stagnantsheetof magmanear would opposethe formationof a denselayer. The
the floorof the intrusion(Jackson, 1961), convection thicknessof such a stagnantlayer might also be
and turbidity currents(Wager, 1967), and contam- increasedby the downwardflow and accumulation
inationby sialicmelt derivedfrom meltingof roof of denserresidualliquid during the crystallization
rocks(Irvine, 1975) or by the influx of additional of plagioclase alongthe marginsof an intrusionwith
magma(Irvine, 1977). Recent work has, however, inward-dippingsides(Sparksand Huppert, 1984).
shownthat crystalsettlingmaynot be an important StagnantFe-enriched layers might conceivably
factorin the solidification of large basicintrusions have formed from time to time in responseto
and that in situ bottomcrystallization mightbe the plagioclase-dominated fractionationduring crystal-
dominantmechanism(e.g., Campbell,1977, 1978; lizationof the upper zone. Suchlayerswouldhave
McBirney and Noyes, 1979; Irvine, 1980a and b, provided the ideal bulk compositionsfrom which
1982). copiousquantitiesof Ti magnetitemay have crystal-
The conceptsof density-stratifiedmagmacham- lized aloneor togetherwith only minor amountsof
bers,crystallizationfrom double-diffusiveconvection silicates. Crystallization of Ti magnetite would
cells, and the influx of new pulsesof magmaof thereforehave taken place from a relatively re-
differingtemperatureand densityinto the magma stricted volume of liquid and the expected trace
chamberhave also been proposedto accountfor element distribution would be similar to that re-
the developmentof layering(e.g.,ChenandTurner, ported by Cawthorn and McCarthy (1980, 1981),
1980; Huppert and Sparks, 1980; Rice, 1981; McCarthy et al. (1983), and Cawthornet al. (1983).
Campbellet al., 1983; Sparksand Huppert, 1984). Opaque oxide-dominatedfractionationwould result
In particularCawthornandMcCarthy(1980, 1981) in a decreasein the densityof the stagnantlayer
and Cawthornet al. (1983) reporteda rapid deple- until it reachedthe densityof the overlyingmagma.
tion of Cr in Ti magnetitewith increasingheightin This would then result in mixingof the two liquids
the Main Magnetitelayer of the BushveldComplex. and disappearanceof the stagnantlayer (Huppert
They cited this as evidencethat individual ore-rich and Sparks,1980; Sparksand Huppert, 1984). The
layers developed by diffusion-controlledbottom operationof sucha mechanismmight alsoaccount
1098 IYAN M. REYNOLDS

for the relatively gradualdepletionin V from the upper zoneof the BushveldComplexwascausedby
residual melt and also account for the behavior in situ bottom crystallization,then it becomesnec-
of TiO2. essaryto invoke some mechanismwhereby foz is
The overall dense nature of the Fe-enriched substantiallyand simultaneouslyincreasedat the
magma from which the upper zone crystallized base of the magmacolumn. This would therefore
would alsohave been importantwith respectto any rule out processesof a more localizednature. The
additionalinfluxesof unfractionated basalticmagma. relativelylargenumberof individualore-richlayers,
The unfractionatedmagmawouldbe lessdensethan and in somecases,their closespacing,may alsorule
that already present in the magma chamber and out new magmainfluxesasa controllingmechanism.
would rise as an entrainingturbulentplume (Sparks This led Klemm et al. (1982) and Reynolds(1985)
et al., 1980; Campbell et al., 1983; Sparks and to postulate that the required oxygen may have
Huppert, 1984). This would have resulted in a been introduced from below into the base of the
rather thorough mixing of the two magmaswith magmacolumn. It may, however, be possiblethat
importantimplicationsfor the developmentof den- such influxesof oxygen are neither required nor
sity stratificationor double-diffusive
convectioncells. probable.
Unfortunately, no data are currently available on The relative effectsof water contentandfo, on
possiblemagmainfluxesin the upper zone and this the crystallizationof basalticmagmasare well doc-
topic is not pursuedfurther. umented (e.g., Hamilton et al., 1964; Hamilton and
Anderson, 1967; Carmichael and Nicholls, 1967;
The plagioclaseproblem
Eggler and Burnham, 1973). These studiesindicate
Plagioclaseis a commonconstituentof ore-rich that thefo= is controlledlargelyby the initial Fe2Oa/
layersand the footwallsuccessions of many(but not FeO ratio of the magma,the oxidationof Fe+2 to
all) layersconsistof anorthositeor other plagioclase- Fe+a,the temperature,andthefn•o/fn•ratio.Simple
rich rocktypes(Molyneux,1970b;Von Gruenewaldt, silicatefractionationin a closedsystemwill result
1973). Experimental studies by Campbell et al. in an increasein the Fe2Oa/FeOratio of the residual
(1978) have shown that plagioclasewill float in melt and may causea rise in fo•. Similarly,anhydrous
relativelydenseiron-enrichedbasalticliquidsof the silicate fractionationin a closed systemwill also
type expectedto be presentduring the later stages result in an increase in the water content of the
of fractionalcrystallization.It is therefore difficult, residual melt. The effects of water content on the
if not impossible,to reconcile any mechanismin- fo• in water-undersaturatedbasaltic magmas are
volvingthe gravitationalaccumulationof plagioclase complexand have been discussed by Hamilton and
with these facts. Irvine (1980b) has shown that Anderson(1967) and Eggler and Burnham(1973).
intermittent currents might be able to transport Water will dissociateat high temperatureaccording
plagioclasecrystalsand depositthem on the floor of to the equation 2H20 = 2H2 + O2 and Osborn
the magma chamber, but the applicability of this (1959) suggestedthat diffusion-controlled hydrogen
model is not known.The presenceof plagioclaseat lossfrom sucha systemmight result in an increase
this level in the complexis mosteasilyexplainedby in foz. Water can thereforeact as an oxygenbuffer,
in situ bottom crystallization,in which new crystals but its buffering capacityis limited by the actual
of plagioclase (andothersilicates)nucleateandgrow quantitiesthat are present (Hamilton and Anderson,
on the upper surfacesof earlier formed crystalsat 1967). These authorssuggestthat between i and 2
the top of the crystal pile. A proportion of these wt percent water might be sufficientto act as a
crystalswill therefore become firmly attachedto buffer since2 percent water would theoreticallybe
the underlying pile and will not be able to float able to oxidizeslightlymore than 15 percent FeO
upward.The preferredorientationexhibitedby many in the melt to Fe2Oa. The primary water contents
plagioclasecrystalsmight also reflect this mode of of basalticmagmasare not accuratelyknown, but
nucleation and growth; this would result in each such melts are probably water undersaturatedat
batch of crystalshaving the same crystallographic pressuresgreater than 1 kb (e.g., Hamilton et al.,
orientationas thoseon which they nucleated. 1964; Hamilton and Anderson, 1967; Carmichael et
al., 1974). The water content of tholeiitic basalt
Oxygen,fugacity, water, and the Fe20a/FeOratio rangesbetween0.06 and 0.10 wt percent(Friedman,
Most recent theoriescall for episodicincreasesin 1967), whereas that of submarinebasaltsranges
fo= as a mechanismfor the precipitationof copious between approximately0.25 and 0.5 wt percent
quantitiesof Ti magnetite.Ulmer (1969) hasoutlined (Moore, 1970). Extensivefractionalcrystallization
severalmechanisms by whichthismightbe achieved, mightconceivablyincreasethe water contentof the
but of these, only changesin pressuremight be residualmelt to between 0.5 and 1.0 wt percent or
expected to influence the whole magma chamber possiblyeven in excessof theseamounts.The oxygen
simultaneously.If it is assumedthat layering in the that is required for the precipitation of abundant
 REVIEW OF MAGNETITE-RICH LAYERS 1099

magnetite might therefore be present in the frae- layers with anorthosite or other plagioclase-rich
tionating magma, thus obviating the necessityto rock types.
introduce it from external sources. The in situ bottom crystallization of abundant
Progressivefractionalcrystallizationmay lead to plagioclasewould not only increase the total Fe
an increasein Pi-i=oin the residualmelt, but this is content of the stagnantlayer but might also result
not necessarilyaccompaniedby an increasein fo= in an increase in the water content because of the
because the dissociation constant of water decreases anhydrousnature of the crystallizingsilicates.The
markedly with decreasingtemperature (Hamilton coprecipitationof small amountsof ilmenite and
andAnderson,1967). BothfI-i•oandfI-i2will therefore Fe+2-bearingsilicates (clinopyroxene,pigeonite,
changewith temperature and total pressureso that fayalitic olivine) would probably not significantly
the calculatedfo• trajectory of the residualmelt can decreasethe densityof the stagnantlayer but would
be shownto be nearly parallel to the QFM buffer contributeto a rise in the FesOa/FeOratio through
curve (Egglet and Burnham, 1973). In addition, a removalof Fe+s from the system.At the same
these authors have demonstrated that mineral as- time, crystallizationof plagioclasewould result in a
semblages in the wall rockswill alsoexerta buffering decreasein the alkali content of the residual liquid
effect on fo2 in magmaticsystemsthat are open to layer and this might resultin a concomitantdecrease
H2 and note that the fo• of the magmamight well in Fe+a solubility.The complexinterplayof these
be lowered relative to a buffer curve in cases where factors could result in a marked increase in the rate
H20 candiffuseintothe magma.EggletandBurnham of Ti magnetiteprecipitationand under favorable
(1973) concludedthat the fo• of a basaltic melt conditionsmight give rise to the growth of distinct
would not be moved by more than 1.3 log units ore-rich layers.
from the originalbufferduringdifferentiationor by The markedvariationin cumulusmineralogysuch
exchangeof fluidswith the wall rocksexceptnear as the presenceof stringersand layers of silicate
the roofsof shallowmagmachambers. mineralsand variationsin the volumeof Ti magnetite
Carmichael and Nicholls (1967) demonstrated reflectsdifferencesin the rate of Ti magnetitepre-
that the Fe•Oa/FeO ratiosof magmasincreasewith cipitation relative to that of the associatedsilicate
an increasein alkali content(or normativefeldspathic minerals and is evidently related to the diffusion
components)and termed this the alkali-ferric iron ratesof the required ionsfrom undepletedliquid to
effect.They notedthat the Fe•Oa/FeO ratio would the growthsiteson the floorof the magmachamber.
therefore tend to rise in the residualliquid during Another possibilityis that gravitationalsettling of
fractionalcrystallizationand that this effect would thesecrystalsover a shortvertical distance(several
be enhancedby fallingtemperature.The solubility meters)may alsohave contributedto the formation
of Fe+• in the melt maythereforealsobe relatedto of the ore-rich layers and associatedvariations in
the amount of alkalis that are present and may oxide/silicateratios.
represent an important mechanismby which the The above mechanismwould probably be most
Fe•Oa/FeOratioin a fractionating magmais allowed effectiveat higher liquidustemperaturesthat permit
to increase. the existenceof higherfo• conditionsand a greater
degreeof dissociation of water. It may thereforebe
Formationof Ti magnetite-richlayers significantthat the greatestnumber of magnetite
The precipitationof smallamounts(5-10 vol %) layers (the lower layers, the Main Magnetite layer,
of opaqueoxidesthroughoutthe bulk of the upper and layers 1-7) are all developedwithin a vertical
zone appears to reflect the normal crystallization distanceof approximately350 m near the base of
courseof the upper zoneliquid. The markedoverall the upper zone. Many of theselayers,notablylower
decreasein fo• that accompaniedthe cumulative layers 1 and 2, the Main Magnetite layer, and layers
crystallization
of suchlargequantitiesof Ti magnetite 1, 2, 3, and 7, showthe developmentof anorthositic
(Fig. 3) appearsto reflect a markeddecreasein the footwall successions. They also show the develop-
Fe•Oa/FeOratio of the successive residualliquids. ment of the greatest amounts of fairly massive,
The formationof a dense,stagnantlayerof highly silicate-poor,Ti magnetite ore.
Fe-Ti-(V)-enrichedliquid on the floorof the magma Magnetite layers are absent from the overlying
chamber will create a favorable environment for the 520 m of the successionand then reappear as a
precipitationof copiousquantitiesof Ti magnetite. groupof relativelythin layers(layers8-14) that are
The mechanismwhereby a layer of this nature may developedwithin a verticaldistanceof approximately
be formedhasbeendiscussed aboveandis dependent 350 m. A single very narrow layer (15) is present
on the relative proportionsof crystallizingsilicate within the overlying approximately700 m and the
phases.This situationis mostlikely to occurwhen remaining layers (16-21) are developed within a
plagioclasefractionationbecomesdominantandmay vertical distanceof some 125 m. These layers are
accountfor the commonassociation of Ti magnetite alsoassociatedwith plagioclase-richrocks,but fay-
1100 IVAN M. REYNOLDS

alitic olivine becomesan increasinglyimportant co-

precipitating phase with increasing stratigraphic
height. The presenceof olivine together with in-
creasingbulk TiO2 contentsof magnetitereflectsa
gradualoverall decreasein fo2 with fractionalcrys-
tallization.The crystallizationof fayaliticolivinewill
opposethe formationof a denseFe-enrichedstagnant
liquid layer, but it is conceivablethat suitablecom-
binations of events would allow this to occur at
infrequentintervals.The fractionationof olivine will
alsocausethe Fe203/FeO ratio of the residualliquid
to rise and might thereforecontributeto the copre-
cipitation of relatively small volumes(5-10%) of Ti
magnetite. The formation of the stratigraphically
higherTi magnetitelayersis alsoenvisagedto have
taken placefrom denseFe-enrichedstagnantlayers
as outlined above. The less robust and thinner nature
of many of these layers might be ascribedto the FIG. 4. Camera lucida drawing of euhedraland subhedral
precipitationof smalleramountsof magnetiterelative Ti magnetitecrystalsenclosedby large clinopyroxenegrains.
to silicatesbecauseof a lack of availableoxygenin Note the presence of chainlike oxide aggregates that may
the system.The formation of the final 10-m-thick representa primary crystallizationfeature that has been pre-
compositelayer 21 is more difficult to explain,but served by the early precipitation of pyroxene. Ti magnetite
= white, pyroxene = shaded.
it mightinvolvethe introductionof additionaloxygen
from external sourcesas it is located very near to
the roof of the intrusion. + liquid into essentiallymonomineraliclayers also
The Ti magnetitelayers are consideredto have touches on one of the major problem areas of
developedlargely by in situ bottom crystallization, layeredigneousrocks.Hess(1939, 1960) postulated
but the possibilityof gravitationalsettlingof oxide that growth of primary crystalswould continue as
crystals over a short distance is not entirely dis- long as diffusional exchange with the overlying
counted. Experimental evidence (Campbell et al., magmawasmaintained.Under favorableconditions,
1978) suggeststhat spinel crystalsoften nucleate a layer consistingof a single mineral plus liquid
and grow in chains,but evidence for the existence might therefore be convertedinto a dense,mono-
of thesechainsin Bushveldrocksis rarely preserved mineralic rock. Wager et al. (1960) introducedthe
due to subsequentsubsolidusmodification.Occa- term adcumulusgrowthto describethis processand
sionalexamplescan,however,be foundwhere chains concludedthat it would only occurin the uppermost
of Ti magnetitecrystalsare enclosedby later crys- layer of the crystal pile. Irvine (1978, 1980a) con-
tallizingpyroxenegrains.This is particularlyevident cluded that these rocks formed as a result of contin-
at the base of layer 13 which is characterizedby a ued diffusion-controlled growth of existingmineral
gradationalcontact. More convincingevidence for phasesfrom intercumulusliquids.
the existence of these chainlike structures is afforded In contrast,Voll (1960) notedthat grainboundary
by chromite crystals in the UG-2 chromite layer relationships andtriple junctionsin manyadcumulus
which have undergonelimited subsolidus modifica- rocks, particularlychromitites,were indicative of
tion. Thesecrystalsare alignedin chainlikecolumns solid state readjustmentprocesses.He concluded
that are oriented perpendicularto the lithological that these textures were formed, at least in part, by
layering (Eales and Reynolds, 1983). Growth of the annealingof crystalsunder the influence of
these chainlike bodies would result in a loosely interfacial free energy. Similar conclusionswere
packed framework of crystalscontaininga variable reached by Weedon (1965) and Vernon (1970).
proportionof interstitialliquid (Fig. 4). The individ- Vernon (1970) noted that single phase adcumulus
ual crystalsmay vary in size and the chainlike rocks consist largely of polygonal grains with
aggregateswill probablycollapseduringcompaction smoothlycurvedinterfacesand concludedthat they
to form closerpackedaggregates(Fig. 5). had formed by sintering. He also noted that the
interprecipitateliquid would be expelledgradually
Densificationof Ti magnetitelayers during annealingand that the grainswould grow
Primary Ti magnetite layers will consistof a togetherunder the influenceof interfacialfree en-
looselypacked aggregateof Ti magnetitecrystals ergyto producea space-filling aggregate of polygonal
and a certain proportion of interprecipitate liquid. grainshavinginterfacialanglesof 120ø.
The conversionof these assemblages of crystals Naturalpolycrystallinesulfideaggregates andan-
 REVIEW OF MAGNETITE-RICH LAYERS 1101

same ore sample in mutually perpendicular direc-

tions, but no significantdifferenceswere noted. All
samples of both massive and disseminatedores
yielded constanttrue interfacialanglesof 120ø + 1o
betweenTi magnetitecrystals.
The natureof grainboundaries in the Ti magnetite
ores,togetherwith their distinctivepolygonaltexture
and well defined triple junctions,clearly indicates
that these mineralshave undergonepostcrystalliza-
tion modification. Adcumulus growth processes
whereby individual grains continue to enlarge by
growth into open spaceswould result in the forma-
tion of impingementtexturesthat wouldbe expected
to display a wide range of true interfacial angles.
The constancyof the 120ø interfacial angles at
mutualcontactsbetweenTi magnetitecrystalsclearly
demonstratesan equilibrium relationship that has
FIG. 5. Compaction might cause the collapse of chainlike beenwell documented in the metallurgicalliterature.
aggregates(Fig. 4) and form more denselypackedassemblages. It is therefore concludedthat annealingplays an
Gravitationalaccumulationwould alsoproducea similarprimary
texture. importantrole in the densificationof compactedTi
magnetitelayers and the formation of essentially
monomineralic ores.
healed metals also commonly display polygonal
foamliketextures.Stanton(1964) duplicatedthese Annealing
sulfidetexturesby annealingloosegrainsof sulfides Annealingor sinteringrefersto a processwhereby
in vacuoat temperatureswell below their melting loose grains and other aggregatesare consolidated
points.The formationof similartexturesby annealing
processesis well documentedin the metallurgical
literature(e.g.,Smith,1948, 1964; CobleandBurke, lOO

1963; Thfmmler and Thomma, 1967). Stanton

(1965) noted that the Fe-Ti oxide aggregatesof
the Skaergaardintrusionconsistof approximately
straight-sidedpolyhedrathat meet in well-defined
triple points;he concludedthat their behaviorshould
90
n'105
8O
....
conformto the normal principlesgoverningthe
70
shapesof polycrystalline
aggregates.
Interfacialanglesbetween coexistingphasesin
any randomsectionthrougha polycrystallineaggre- 6O
gatewill not normallyrepresenttrue anglesbecause
the grain boundariesare not all oriented normal to z 50
the section plane (Smith, 1948, 1964; Stanton,
1964).The true interfacialanglesbetweencoexisting
Ti magnetitecrystalsin a wide range of Bushveld IJ

ores and silicate-rich rocks have been determined

by the authorusingthe methodof RieggerandVan 3O
Vlack (1960). These authorsdemonstratedthat the
true interfacialanglein a polycrystalline
aggregate 2O
canbe obtainedfrom the medianpoint on a cumu-
lative frequencycurve basedon a minimumof 30 10
measurements. The margin of error is lessthan lø
One hundredmeasurements of triple junctionsbe-
tween Ti magnetite crystalswere made on polished 90 100 110 120 130 lt, 0 150
slabsof each sampleusing a stereomicroscope 0BSERVEO INTERFA[IAL
equippedwith a rotatingstageandthe resultsplotted ANGLE
as cumulativefrequencycurves(Fig. 6). Possible FIG.6. An exampleshowingthe useof a cumulativefrequency
orientationeffectswere investigatedby measuring curveto determinethe true interfacialanglein a polycrystalline
triple junction angleson three slabscut from the aggregate.
1102 I¾,4N M. REYNOLDS

into dense polycrystallinematerialsby heating at they are eventuallyeliminated.Thisultimatelyresults

temperaturesbelow their melting points.This may in a coarse-grainedpolycrystallineaggregatewith
occurin the presenceof a fluid phaseand resultsin equilibriumgrainboundaryrelationships.
the formation of foamlike microstructures in which
grain boundariesmeet in well-defined triple junc- The presenceof impuritiesor inclusionshinders
tions having interfacial anglesof 120ø (Coble and migrationof grain boundariesbecausethey require
Burke, 1963; Thiimmler and Thomma, 1967). Three an energeticallyunfavorableincreasein surfacearea
stagesin this processare normally recognizedand in order to migrate past suchimpurities.The pres-
their analoguesare ubiquitousfeaturesof Bushveld ence of large numbersof inclusionswill therefore
Ti magnetite ores (Fig. 1): tend to inhibit grain growth.The rangeof textures
displayedby BushveldTi magnetiteores (Fig. 1)
1. Initial stageof neck growth: This is diffusion can be readily interpreted in terms of the above
controlled and results in the formation of a very annealingmodel.Essentiallypure Ti magnetite-rich
short grain boundary or neck between grains at areasdevelopa markedlycoarsegrainsize, whereas
their pointsof contact(Fig. 6A and B). Once initi- the presenceof silicateor ilmeniteinclusionsresults
ated,neckgrowthwill continueuntil the acuteangle in a markedly finer grain size. The location of
between grainshasbeen largely eliminated.This is entrappedinterstitialliquid and externallyexsolved
accompanied by intergrainshrinkageand a marked ilmenite granulesat triple junction points and as
densificationof the sinteringmaterialtakesplace. films along grain boundariesis entirely consistent
2. Intermediatestage:This is initiated after ex- with this model. The mechanism is also clearly
tensiveneck growth when individualgrainsbegin applicableto other essentiallymonomineraliclayers
to losetheir identity. Grain growth proceedsunder such as chromitite, anorthosite, bronzitite, and du-
the driving force of interfacial free energy and nitc. The progressiveannealingof an initial Ti mag-
resultsin the developmentof equilibriuminterfacial netiteaggregateis illustratedschematically
in Figures
anglesbetweencoexistingphases.The intergranular 7 and 8. Note that this processdoesnot require the
poresremaininterconnected,but their volumesde- introduction of additional material.
creasedue to expulsionof the interprecipitateliquid Annealingprocesseswould be initiated immedi-
during annealing.Densification(i.e., the formation atelyafterinitialprecipitationof Ti magnetitecrystals
of a closelypacked,space-fillingaggregate)is there- as soonas point contactshad been established.This
fore achievedby modificationof originalgrain mor- processwould be promotedby the prevalenceof
phologieswhich, in turn, resultsin the expulsionof relativelyhigh temperaturesfor a longperiodduring
interstitialliquid due to a reductionin pore space. initial slowsubsoliduscooling.The rate of annealing
The sinteringprocessdoesnot require the introduc- would decreaseprogressivelywith falling tempera-
tion of additionalmaterialsand the sinteredproduct ture and the developmentof equilibrium textures.
therefore occupiesa significantlysmaller volume The grain boundaryrelationships between Ti mag-
than the original unsinteredaggregate. netite and externallyexsolvedilmenitegranules(that
3. Finalstage:Thisisreachedwhenporesbecome exhibitall the featuresexpectedfrom a minor phase
pinchedoff by grain growth and their liquid crys- in a polycrystalline
aggregate)indicatethat annealing
tallizesas entrappedphases.This resultsin a dense processes remainedactiveuntil at leastthis stageof
polycrystallineaggregatethat is characterizedby microtexturaldevelopmenthad been reached.This
equilibriumgrainboundaryrelationships. Solidstate stageof microstructuraldevelopmentis believed to
growthof numbersof largepolygonalcrystalsat the ceaseat approximately800øC (Anderson,1968).
expenseof smallercrystalsmay also occur (Smith, The dense Fe-enrichedliquid that resultsfrom
1948, 1964; Coble and Burke, 1963). This is accom- extensiveplagioclase fractionationwill collectin the
plished by grain boundarymigrationwhich takes intersticesbetween the cumuluscrystalsbefore it
place under the driving force of interfacialfree becomesabundantenoughto form a distinctlayer
energy. The lowestenergy state is representedby at the baseof the magmacolumn.The subsequent
straight-sided polyhedrathat exhibit six sideswhen behavior of this liquid will also be important in
viewed in random sections.The requirementsof controllingthe texturesthat will be formed(Wager
spacefilling dictatethat thosegrainswith lessthan et al., 1960).Limitedpostcumulus growthof cumulus
six sides exhibit concavegrain boundaries,while silicatephaseswill occuruntil the liquidis depleted
those with more than six sides exhibit convex in suitablecomponentsand the residuewill crystal-
boundaries.These grain boundariesmigratetoward lize largelyas opaqueoxidesto form classicortho-
their centers of curvature in order to decrease their cumulustextures.In other cases,muchof this liquid
total surface area with the result that grains with maybe displacedupward(infiltrationmetasomatism)
more than six sideswill grow larger, whereasthose by compaction,adcumulusgrowth, and annealing
with less than six sides will decrease in size until processes. The movementof this liquid pastcumulus
 REVIEW OF MAGNETITE-RICHLAYERS 1103

FIG. 7. A schematicdiagramillustratingsuccessivestagesin the annealingof a looselypacked

aggregate of Ti magnetitecrystalsasshownin Figure5. A. Neckgrowthcommences at pointsof
mutualcontact.B. Neck growthis almostcompleteand the volumeof interstitialliquid hasbeen
considerably reduced.C. Commencement of the intermediatestagein which equilibriumgrain
boundaryrelationships
arebeginningto develop.Noteformationof triplejunctions.D. The beginning
of graingrowthwhereseverallargegrainsgrowat the expenseof numbersof smallergrains.Note
that additional material need not be introduced during annealing.

Ti magnetiteand ilmenitegrainswouldalsoresult
in their continuedpostcumulusgrowth and may
contributemateriallyto the densification
of ore-rich
layersthatareundergoing annealing.
Thismechanism
may thereforeaccountfor the commonoccurrence
of a zoneof dense,coarse-grained,
andwell-annealed
Ti magnetiteat the bottomof individuallayers.A
similar situationmay exist in silicate rocks that
contain some cumulusoxide phasesthat become
significantlyenlargedduringthis processto yield
products thatappearto represent extremeexamples
of orthocumulus rocks.
The sharplower contactsof Ti magnetitelayers
that overlieanorthositic
rocksare not easilyexplained
in termsof the accumulation of a stagnantlayer of
FIG. 8. Final annealingproductconsisting
of coarsepolygonal
dense, Fe-enriched liquid. These anorthositesmay
grains of Ti magnetite showingequilibrium grain boundary be virtuallyfree of cumulusoxidephases,and al-
relationships. thoughthey do containpatchesof intercumulus
1104 IVAN M. REYNOLDS

oxides, the volumes are not as great as might be Late-stagealteration

expected. Microscopicexaminationof these essen- In addition to extensive late-stage alteration of
tially monomineralicrocksrevealsthe presenceof footwallrocks,the BushveldTi magnetiteoresshow
textures that are indicative of a high degree of considerable evidence of late-stage reaction and
annealing.It is debatablewhether annealingalone alteration.This is reflected in the developmentof a
would have been responsiblefor the very effective wide range of finer scaleilmenite intergrowthsin
expulsionof intercumulusliquids and the effects the Ti magnetite and the oxidation of previously
other processeswill also have to be investigated, exsolvedulv6spinel to ilmenite + magnetite. These
e.g., the manner in which anorthositelayersform. features are commonly restricted to the marginal
areasof individualgrainsand alongfractures(Reyn-
Associatedsulfides olds, 1983a, 1985). Symplectitelikeintergrowthsof
olivine or pyroxene and plagioclasemay be devel-
Disseminatedsulfidesare commonlyfound asso-
oped alongcontactsbetween Ti magnetiteand pla-
ciated with the thicker Ti magnetitelayers in the
gioclase(Willemse, 1969; Molyneux, 1970a and b;
BushveldComplex (Liebenberg, 1970; Buchanan,
Reynolds 1983a, 1985). Biotite rims between Ti
1976; Von Gruenewaldt, 1976, 1979). These are
magnetite and plagioclaseare normally ascribedto
usuallyFe rich (pyrrhotiteand/orpyrite)andcontain reactionbetweenTi magnetiteandlate-stageresidual
smallamountsof Cu-bearingphasessuchas chalco-
fluids(Morse, 1979; Parsons,1980, 1981; Reynolds,
pyrite and cubanitc.The pyrrhotite that is present 1983a and b, 1985).
below the Main Magnetitelayer is Ni bearing and
containsexsolvedpentlanditc.The sulfidesin the Summaryof ProposedGenetic Model
ore-richlayersare generallypresentassmallgrains
that are located interstitially between the larger Any proposedmodel for the genesisof Ti mag-
oxide crystals.In addition,small roundedsulfide netite layers is constrainedby their petrological,
inclusionsare alsocommonlypresentin the oxides, mineralogical,and geochemicalcharacteristicsas
particularly
withinilmenite.The highestsulfidecon- well as those of their silicate host rocks. It must
centrationsare located within the uppermostmeter explainnot only the mechanisms that are reponsible
of the footwall successionsof thicker Ti magnetite for the precipitationof large quantitiesof Ti mag-
layersand are locatedinterstitiallybetweensilicate netite and the developmentof ore-rich layers,but
crystalsand alonglate-stagefractures.They are not alsothe mechanismsby which primary precipitates
normallypresentin sufficientquantitiesto be of are convertedinto essentiallymonomineralicpoly-
economic interest. crystallineores. An attempt hasbeen made to de-
Solubilityof sulfur in mafic magmasis largely velop a geneticmodel that not only explainsmany
dependenton Fe+2content(Maclean,1969; Haugh- of the characteristicfeatures of Ti magnetite layers
ton et al., 1974). It is believed that precipitationof but is also consistent with current views on the
large quantitiesof magnetiteresultsin a decrease crystallizationof layered mafic intrusion.The pro-
in sulfursolubilityand the separationof dropletsof posedmodelmaybe summarized asfollows:
immiscible sulfide liquid (Buchanan, 1976; Von
Gruenewaldt,1976, 1979). This mechanismaccounts 1. A lengthy period of fractional crystallization
for sulfidesthat are intimately associatedwith Ti of basaltic magma, probably involving several in-
magnetitelayersand may alsoaccount,in part, for fluxes,finallyresultsin the concentration of substan-
some of the sulfides that are associated with inter- tial amounts of Fe, Ti, and V in the late-stage
cumulus oxides. residual magma.This creates favorable conditions
The operationof annealingprocesses will probably for large-scaleTi magnetiteprecipitationwhich will
result in the densificationof oxide-rich layers at an be initiated oncethe temperatureof the magmahas
earlystageduringsubsolidus cooling,thusproducing decreased sufficiently
to intersectthe spinelstability
relativelyimpermeablebodies.These impermeable field at ambientfo•.
barriers would impede the movementof late-stage 2. Ti magnetiteprecipitationdependslargely on
deuteric fluidsand other late-stagevolatile compo- the Fe203/FeOratio of the liquid whichis, in turn,
nentsthat might become concentratedor dammed a functionof thefo•, temperature,andwater content
up in the footwall areas of thicker Ti magnetite of the fractionatingmagma.The Fe203/FeO ratio
layers. These late-stagefluids are responsiblefor may alsobe increasedby the crystallizationof il-
extensive alteration of silicate assemblagesin the menitc and Fe+•-bearingsilicates(clinopyroxene,
immediatefootwall areasand it is probablethat part pigeonitc, fayalitic olivine).
of the Cu-bearing sulfide was precipitated from 3. The crystallizationof abundant plagioclase
these fluids. leads to an increase in the total Fe content and a
 REVIEW OF MAGNETITE-RICHLAYERS ] ] 05

concomitant increasein densityof the residualliquid The model above is basedon the direct precipi-
which may accumulateto form a stagnantlayer on tation of Ti magnetitecrystalsbut can equallybe
the floor of the magmachamber. appliedto the separationandconcentrationof phos-
4. This denselayer will not mix with the overlying phorus-poor,Fe-Ti oxide liquids in the manner
magmaandthusprovidesa suitableenvironmentfor envisagedfor the formationof apatite-oxideores
the crystallizationof copiousquantitiesof Ti mag- (Philpotts,1967; Kolker, 1982; Reynolds,1983a).
netite. Magnetiteprecipitationis controlledby the This alternativemechanismcan, unfortunately,not
Fe203/FeO ratio of the liquid whichis bufferedby be effectivelyevaluatedin termsof currentlyavail-
the presenceof small amountsof "water" in the ableexperimentaldataon Fe-Ti-V-enrichedresidual
residualliquid. liquidsof basalticparentageand clearly warrants
5. Variationsin the relativeproportionsof oxide further investigation.
andsilicatephasesmayreflectdifferences in diffusion
Acknowledgments
rates for the different ionic speciesin the liquid
and, consequently, crystallizationrates.The crystal- The ideas embodied in this article have evolved
lization of Ti magnetitedecreasesthe densityof the over a period of someten yearsduringwhich time
residualliquid of the stagnantlayer until it is the manypeoplehavecontributedto my understanding
sameasthat of the overlyingmagmawhereuponthe of theseenigmaticrocks.In particular,I would like
two liquidswill mix, thusterminatingthe cycle.The to acknowledgethe adviceand guidanceof H. V.
gradationalupper contactsmay reflect the gradual Ealesof RhodesUniversitywho initiatedmy studies
loweringof the total Fe contentof the liquid or of in this field and has encouragedthis interest over
the Fe203/FeO ratio to levels where magnetite the past decade.S. A. Hiemstra and staff at the
precipitationis retarded. National Institute for Metallurgy (now MINTEK)
6. The ore-rich layersdeveloplargely by in situ are thankedfor the opportunityto initiate studies
bottom growth of Ti magnetitecrystals.These nu- into various South African titaniferous iron ores
cleate and grow in chainswhich may later collapse while I wasin their employ.Financialsupportfrom
to form a more closely packed arrangement.The the NationalInstitute of Mines, the CSIR (University
initial ore-rich layers will therefore consistof an ResearchGrantsDivisionand Co-operativeScientific
aggregate of T magnetite crystals together with Program),and RhodesUniversityis gratefully ac-
variable amountsof coprecipitatedsilicateminerals knowledged.Critical reviewsby H. V. Eales,G. Von
andilmenite.A high proportionof interstitialliquid, Gruenewaldt,and two anonymousreferees are ac-
possiblyon the order of 50 vol percent,is initially knowledged,but the final responsibilityfor the
present. contents rests with the author. Pat Shuttleworth is
7. Densificationand subsequentconversionof thankedfor typing the manuscript.
ore-richlayersinto essentiallymonomineralic
rocks December21, 1983; September19, 1984
is accomplishedlargely by annealingat elevated
subsolidustemperaturesand may be augmentedby REFERENCES

the introduction of intercumulusliquid that is ex- Andersen,D. J., andLindsley,D. H., 1979, The olivine-ilmenite
pelled from the underlyingcrystalpile duringinfil- thermometer: Lunar Planet. Sci. Conf., 10th, Houston, Texas,
tration metasomatism.This leads to expulsion of 19-23 March, 1979, Proc., p. 493-507.
Anderson, A. T., 1968, Oxidation of the LaBlache Lake titanif-
high proportionsof interstitialliquid and formation erous magnetite deposit, Quebec: Jour. Geology, v. 76, p.
of essentiallymonomineralicTi magnetite layers 528-547.
that exhibitpolygonalgrainboundarieswhichmeet Bateman, A.M., 1951, The formation of late magmatic oxide
in triple junctionswith interfacialanglesof 120ø. ores: ECON. GEOL., v. 46, p. 404-426.
Bishop,F. C., 1980, The distributionof Fe•+ and Mg between
8. Subsequent subsolidus coolingresultsin the coexistingilmenite andpyroxenewith applicationsto geother-
developmentof a wide rangeof ilmenite,ulv6spinel, toomerry: Am Jour. Sci., v. 280, p. 46-77.
and transparentspinelintergrowthsin the Ti mag- Bowles,J. F. W., 1976, Distinct cooling historiesof troctolites
netite. The nature and relative degree of develop- from the Freetown layered gabbro: Mineralog. Mag., v. 40,
ment of these intergrowthsreflectsboth the com- p. 703-714.
positionsof their hostphasesand the fo• conditions • fugacity
1977, A method of tracing the temperatureand oxygen
historiesof complexmagnetite-ilmenitegrains:Min-
that existedduring subsolidus cooling. eralog. Mag., v. 41, p. 103-109.
9. The thicker and more massiveTi magnetite Buchanan,D. L., 1976, The sulfide and oxide assemblagesin
layersformedrelativelyimpermeablebarriersto the the BushveldComplex rocksof the Bethel area: Geol. Soc.
upward movementsof volatile phasesduring later SouthAfrica Trans., v. 79, p. 76-80.
Buddington,A. F., and Lindsley, D. H., 1964, Iron-titanium
stagesof cooling.Thesebarriersled to a concentra- oxide mineralsand syntheticequivalents:Jour. Petrology,v.
tion of volatilesin the footwall areasof theselayers, 5, p. 310-357.
causingextensivedeutericalterationof the silicates. Cameron,E. N., and Desborough,G. A., 1969, Occurrence and
1106 IVAN M. REYNOLDS

characteristicsof chromite deposits--eastern BushveldCom- Evans, B. W., and Moore, J. G., 1968, Mineralogy as a function
plex: ECON. GEOL. MON. 4, p. 23-40. of depth in the prehistoric Makaopuhi tholeiitic lava lake:
Campbell, I. H., 1977, A study of cumulate processesand Contr. Mineralogy Petrology, v. 14, p. 27-35.
macrorhythmiclayering in the Jimberlana layered intrusion Ewart, A., Bryan, W. B., and Gill, J. B., 1973, Mineralogy and
of Western Australia, Part 1: The upper layered series:Jour. geochemistryof the youngervolcanicislandsof Tonga, S. W.
Petrology,v. 18, p. 183-215. Pacific:Jour. Petrology, v. 14, p. 429-465.
-- 1978, Someproblemswith the cumulustheory: Lithos, v. Friedman, I., 1967, Water and deuterium in pumice from the
11, p. 311-323. 1959-1960 eruption of Kilauea Volcano, Hawaii: U.S. Geol.
Campbell, I. H., Roeder, P. L., and Dixon, J. M., 1978, Crystal Survey Prof. Paper 575-B, p. 120-127.
buoyancy in basaltic liquids and other experimentswith a Hamilton, D. L., and Anderson, G. M., 1967, Effects of water
centrifuge furnace: Contr. Mineralogy Petrology, v. 67, p. and oxygenpressureon the crystallizationof basalticmagma,
369-377. in Hess, H. H., and Poldervaart, A., eds., Basalts:New York,
Campbell, I. H., Naldrett, A. J., and Barnes, S. J., 1983, A Interscience, v. 1, p. 445-482.
model for the origin of platinum-richsulfidehorizonsin the Hamilton, D. L., Burham, W. C., and Osborn, E. F., 1964, The
Bushveldand Stillwater complexes:Jour. Petrology, v. 24, p. solubility of water and effectsof oxygen fugacity and water
133-165. content ou crystallizationin marie magmas:Jour. Petrology,
Carmichael,I. S. E., and Nicholls, J., 1967, Iron-titanium oxides v. 5, p. 21-39.
and oxygen fugacitiesin volcanicrocks:Jour. Geophys.Re- Haggerty, S. E., 1976, Opaque mineral oxides in terrestrial
search, v. 72, p. 4665-4687. igneousrocks,in Rumble,D., ed., Oxide minerals:Mineralog.
Carmichael, I. S. E., Turner, F. J., and Verhoogen, J., 1974, Soc. America Short Course Notes, v. 3, p. Hg 101-Hg 300.
Igneouspetrology:New York, McGraw-Hill, 739 p. Hall, A. L., 1932, The BushveldIgneousComplex of the central
Cawthorn, R. G., and McCarthy, T. S., 1980, Variatious in Cr Transvaal:South Africa Geol. Survey Mem. 28, 510 p.
content of magnetite from the upper zone of the Bushveld Haughton, D. R., Roeder, P. L., and Skinner, B. J., 1974,
Complex--evidence for heterogeneityand convectioncurrents Solubility of sulfur in marie magmas:ECON. GEOL., v. 69, p.
in magmachambers:Earth Planet. Sci. Letters, v. 46, p. 335- 451-467.
343. Hess, H. H., 1939, Extreme fractionalcrytallization of a basaltic
1981, Bottom crystallization and diffusion control in magma:The Stillwaterigneouscomplexlabs.I: Am. Geophys.
layered complexes:Evidencefrom Cr distributionin magnetite Union Trans., Pt. 3, p. 430-432.
from the Bushveld Complex: Geol. Soc. South Africa Trans., -- 1960, Stillwater igneouscomplex, Montana: A quantitative
v. 84, p. 41-50. mineralogicalstudy: Geol. Soc. America Mem. 80, 225 p.
Cawthorn, R. G., McCarthy, T. S., and Davies, G., 1983, Hill, R., and Roeder, P. L., 1974, The crystallizationof spinel
Vertical gradients in a single grain of magnetite from the from basaltic liquid as a function of oxygen fugacity: Jour.
BushveldComplex, South Africa: Mineralog. Mag., v. 47, p. Geology, v. 82, p. 709-729.
27-34. Himmelberg, G. R., and Ford, A. B., 1977, Iron-titanium oxides
Chen, C. F., andTurner, J. S., 1980, Crystallizationin a double- of the Dufek intrusion, Antarctica: Am. Mineralogist, v. 62,
diffusive system:Jour. Geophys.Research,v. 85, p. 2573- p. 623-633.
2593. Huppert, H. E., and Sparks,R. S. J., 1980, The fluid dynamics
Coble, R. L., and Burke, J. E., 1963, Sintering in ceramics: of a basalticmagma chamber replenishedby influx of hot,
Prog. Ceramic Sci., v. 3, p. 197-251. denseultrabasicmagma:Contr. MineralogyPetrology,v. 75,
Coertze, F. J., 1966, The genesisand geologicalenvironment p. 279-289.
of the BushveldComplex in the area southwestof the Leolo Irvine, T. N., 1967, Chromium spinelas a petrogeneticindicator,
Mountains: South Africa Geol. Survey Bull. 47, 57 p. Part 2, petrologic application: Canadian Jour. Earth Sci., v.
Duchesne, J.-C., 1970, Microtextures of Fe-Ti oxide minerals 4, p. 71-103.
in the South Rogalandanorthositiccomplex (Norway): Soc. -- 1975, Crystallizationsequencesin the Muskox intrusion
Geol. BelgiqueAnn., v. 93, p. 527-544. and other layered intrusions--II. Origin of chromitite layers
Dunn, J. A., andDey, A. K., 1937, Vanadium-bearing titaniferous and similar depositsof other magmatic ores: Geochim. et
iron oresin Singhbhumand Mayurbhanji,India: Mining Geol. Cosmochim. Acta, v. 39, p. 991-1020.
Metall. Inst. India Trans., v. 31, p. 117-184. -- 1977, Origin of chromitite layersin the Muskoxintrusion
Du Toit, A. L., 1918, Plumasite(corundum-aplite)and titanif- and other layered intrusions:Geology, v. 5, p. 273-277.
erous magnetiterocks from Natal: Geol. Soc. South Africa -- 1978, Infiltration metasomatism,adcumulusgrowth, and
Trans., v. 21, p. 53-73. secondary differentiation in the Muskox intrusion: Carnegie
Eales,H. V., 1979, AnomalousKaroo spineIsalongthe chromite- Inst. Washington Year Book 77, p. 743-751.
titanomagnetitejoin: South African Jour. Sci., v. 75, p. 24- -- 1979, Rockswhose compositionis determined by crystal
29.
accumulationand sorting, in Yoder, H. S., ed., The evolution
Eales, H. V., and Reynolds, I. M., 1983, The nature of the of the igneous rocks: Princeton, Princeton Univ. Press, p.
chromititelayerswithin the UGl-Bastardunit interval, RPM 245-306.
Union section labs.I: Symposiumon the BushveldComplex, -- 1980a, Magmaticinfiltrationmetasomatism, double-diffu-
Univ. of Pretoria,Nov. 1983, Programmeand Abstracts,p. sive fractional crystallizatiouand adcumulusgrowth in the
22-24.
Muskox intrusion and other layered intrusions, in Hargraves,
Eales, H. V., Reynolds, I. M., and Gouws, D. A., 1980, The R. B., ed., Physicsof magmaticprocesses: Princeton,Princeton
spinel-groupmineralsof the centralKarootholeiitic province: Univ. Press, p. 325-383.
Geol. Soc. South Africa Trans., v. 83, p. 243-253.
-- 1980b, Magmaticdensitycurrentsand cumulusprocesses:
Eales, H. V., and Robey, J. van A., 1976, Differentiation of
Am. Jour. Sci., v. 280A, p. 1-58.
tholeiitic Karoo magma at Birds River, South Africa: Contr.
Mineralogy Petrology, v. 56, p. 101-117. -- 1982, Terminologyfor layered intrusions:Jour.Petrology,
Eggler,D. H., and Burnham,C. W., 1973, Crystallizationand v. 23, p. 127-162.
fraetionationtrends in the system andesite-H20-CO2-O2at Jackson,E. D., 1961, Primary textures and mineral associations
pressuresto 10 kb: Geol. Soc.America Bull., v. 81, p. 2,517- in the ultramarie zone of the Stillwater Complex, Montana:
2,532. U.S. Geol. Survey Prof. Paper 358, 106 p.
 REVIEW OF MAGNETITE-RICH LAYERS 1107

Jaupart, C., Brandeis, G., and Allegre, C. J., 1984, Stagnant and their origin:ECON.GEOL.,v. 23, p. 724-761 and p. 895-
layers at the bottom of convectingmagmachambers:Nature, 922.
v. 308, p. 535-538. Parsons, I., 1980, Alkali-feldspar and Fe-Ti oxide exsolution
Klemm, D. D., Snethlage, R., Dehm, R. M., Henckel, J., and textures as indicators of the distribution and subsolidus effects
Schmidt-Thome, R., 1982, The formation of chromite and of magmatic"water" in the Klokkenlayered syeniteintrusion,
titanomagnetite depositswithin the BushveldIgneousComplex, southGreenland: Royal Soc. Edinburgh Trans., Earth Sci., v.
in Amstutz, G. C., et al., eds., Ore genesis--the state of the 71, p. 1-12.
art: Berlin, Springer Verlag, p. 351-370. -- 1981, The Kiokken gabbro-syenitecomplex,southGreen-
Kolker, A., 1982, Mineralogy and geochemistryof Fe-Ti oxide land: quantitative interpretation of mineral chemistry: Jour.
and apatite (nelsonite) depositsand evaluation of the liquid Petrology, v. 22, p. 233-260.
immiscibilityhypothesis:ECON.GEOL.,v. 77, p. 1146-1158. Peck,D. L., Wright, T. L., and Moore, J. G., 1966, Crystallization
Liebenberg,L., 1970. The sulfidesin the layered sequenceof of tholeiitic basalt in Alae lava lake, Hawaii: Bull. Volcanol.,
the BushveldComplex: Geol. Soc. South Africa Spec. Pub. 1, v. 29, p. 629-656.
p. 108-207. Philpotts,A. R., 1967, Origin of certain iron-titaniumoxide and
Lindsley, D. H., 1976, Experimental studiesof oxide minerals, apatite rocks:ECON. GEOL., v. 62, p. 303-315.
in Rumble, D., ed., Oxide minerals:Mineralog. Soc. America Reynolds,I. M., 1978, Mineralogical studiesof South African
Short Course Notes, v. 3, p. L61-L88. titaniferousiron ores: Their applicationto extractive metal-
Lister, G. F., 1966, The compositionand origin of selected lurgy: Geol. Soc.SouthAfrica Trans., v. 81, p. 233-240.
iron-titanium deposits:ECON. GEOL.,v. 61, p. 275-310. 1979a, Grain boundary relationshipsin some Bushveld
Luyt, J. F. M., 1976, Vanadium: South Africa Geol. Survey titaniferous magnetite cumulates labs.I: Cong. Geol. Soc.
Handb. 7, p. 241-245. South Africa, 18th, Port Elizabeth, Sept. 1979, Abstracts,p.
MacLean, W. H., 1969, Liquidusphaserelationshipsin the FeS- 229-301.
FeO-SiO2 system, and their application in geology: ECON. -- 1979b, A mineralogicalinvestigationof coexistingiron-
GEOL., v. 64, p. 865-884. titanium oxidesfrom variousigneousrockswith specialref-
McBirney, A. R., and Noyes, R. M., 1979, Crystallizatio• and erence to some South African titaniferous iron ores: Unpub.
layering of the Skaergaardintrusion: Jour. Petrology, v. 20, Ph.D. dissert., Rhodes Univ., 632 p.
p. 487-554. -- 1980, Ore petrographyand mineralogyof the vanadium-
McCarthy, T. S., Cawthorn, R. G., Wright, C. J., and McIver, bearing titaniferous magnetite layer of the Kaffirskraalintru-
J. R., 1983, Mineral layering in the Bushveld Complex-- sion, Heidelberg district,Transvaal:Geol. Soc. SouthAfrica
implicationsof Cr, Ni and Cu abundancesin magnetite,from Trans., v. 83, p. 221-230.
closely spacedmagnetite and intervening silicate-richlayers -- 1981, The mineralogyandpetrographyof sometitaniferous
labs.I:Symposiumon the BushveldComplex, Univ. Pretoria, iron ores from the UshushwanaComplex: Geol. Soc. South
Nov. 1983, Programmewith Abstracts,p. 65. Africa Trans., v. 84, p. 261-269.
Molengraaff, G. A. F., 1904, Geology of the Transvaal:Jo- -- 1983a, Contrasted mineralogy and textural relationships
hannesburg,T. and A. Constable,90 p. in the uppermosttitaniferousmagnetitelayersof the Bushveld
Molyneux, T. G., 1970a, The geologyof the area in the vicinity complex in the Bierkraal area, north of Rustenburg labs.]:
of Magnet Heights, easternTransvaal,with specialreference Symposiumon the Bushveld Complex, Univ. Pretoria, Nov.
to the magneticiron ore: Geol. Soc. South Africa Spec. Pub. 1983, Programmewith Abstracts,p. 72-74.
1, p. 228-241. 1983b, The iron-titanium oxide mineralogy of Karoo
-- 1970b, A geologicalinvestigationof the BushveldComplex dolerite in the easternCape and southernOrange Free State:
in Sekhukhunelandand part of the Steelpoortvalley, eastern Geol. Soc.SouthAfrica Trans., v. 86, p. 211-220.
Transvaal, with particular reference to the oxide minerals: -- 1985, The mineralogyand ore petrographyof the Bushveld
Unpub. D.Sc. dissertation, Pretoria Univ., South Africa, titaniferous magnetite-richlayers, in Maske, S., ed., Geology
126 p. of the mineral deposits of southern Africa: Johannesburg,
-- 1972, X-ray data and chemical analysesof some titano- Geol. Soc. South Africa (in press).
magnetiteand ilmenite samplesfrom the BushveldComplex, Rice, A., 1981, Convectivefractionation:A mechanismto provide
SouthAfrica: Mineralog. Mag., v. 38, p. 863-871. cryptic zoning (macrosegregation)layering, crescumulates,
Moore, J. G., 1970, Water content of basalt erupted on the banded tuffs and explosivevolcanismin igneousprocesses:
oceanfloor: Contr. MineralogyPetrology,v. 28, p. 272-279. Jour. Geophys. Research, v. 86, p. 405-417.
Morse, S. A., 1979. Kiglapait geochemistry II: Petrography: Riegger, O. K., and Van Vlack, L. H., 1960, Dihedral angle
Jour. Petrology, v. 20, p. 591-624. measurement: AIME Trans., v. 218, p. 933-935.
Muan, A., and Osborn,E. F., 1956, Phaseequilibria at liquidus Roedder, E., 1978, Silicate liquid immiscibilityin magmasand
temperatures in the system MgO-FeO-Fe203-SiO2: Am. Ce- in the systemK•O-FeO-Al•O3-SiO•:An exampleof serendipity:
ramic Soc. Jour., v. 39, p. 121-140. Geochim. et Cosmochim.Acta, v. 42, p. 1597-1617.
Naslund, H. R., 1976, Liquid immiscibility in the system -- 1979, Silicate liquid immiscibility in magmas,in Yoder,
KAISi3Os-NaA1Si•Os-FeO-Fe20•-SiO2 and its applicationto H. S., ed., The evolution of the igneous rocks: Princeton,
natural magmas:CarnegieInst. WashingtonYear Book 75, p. Princeton Univ. Press, p. 15-57.
592-597. Roeder, P. L., and Osborn, E. F., 1966, Experimental data for
Neumann, E. R., 1974, The distribution of Mn 2+ and Fe 2+
the systemMgO-FeO-Fe•O3-CaAI2Si•O8and their petrologic
implications:Am. Jour. Sci., v. 264, p. 428-480.
betweenilmenitesandmagnetites in igneousrocks:Am. Jour. Schwellnus,C. M., and Willemse, J., 1943, Titanium and vana-
Sci., v. 274, p. 1074-1088.
dium in the magneticiron ores of the BushveldComplex:
Osborn, E. F., 1959, The role of oxygen pressurein the Geol. Soc. South Africa Trans., v. 46, p. 23-38.
crystallization anddifferentiationof basalticmagma:Am. Jour. Singewald,J. T., 1912, The iron ore depositsof the Cebolla
Sci., v. 257, p. 609-647. district, GunnisonCounty, Colorado:ECON. GEOL., v. 7, p.
• 1962, Reactionseriesfor subalkalineigneousrocksbased 560-573.
on differentoxygenpressureconditions:Am. Mineralogist,v. Smith, C. S., 1948, Grains, phasesand interfaces:An interpre-
47, p. 211-226. tation of microstructures:Am. Inst. Mining Metall. Engineers
Osborne,F. F., 1928, Certain magmatictitaniferousiron ores Trans., v. 175, p. 15-51.
1108 IVAN M. REYNOLDS

• 1964, Some elementaryprinciples of polycrystallinemi- Greenland:Neus Jahrb. Mineralogie Abh., v. 94, p. 993-
crostructure: Metall. Rev., v. 9, p. 1-48. 1016.
Sparks,R. S. J., and Huppert, H. E., 1984, Density changes Vincent, E. A., and Phillips, R., 1954, Iron-titanium mineralsin
during the fractionalcrystallizationof basalticmagmas:Fluid layeredgabbrosof the Skaergaardintrusion,eastGreenland:
dynamicimplications:Contr. MineralogyPetrology,v. 85, p. Geochim. et Cosmochim.Acta, v. 6, p. 1-26.
300-309. Voll, G., 1960, New work on petrofabrics:LiverpoolManchester
Sparks,R. S. J., Meyer, P., and Sigurdson,H., 1980, Density Geol. Jour., v. 2, p. 503-567.
variation amongstmid-ocean ridge basalts:Implications for Von Gruenewaldt,G., 1973, The main and upper zonesof the
magma mixing and the scarcity of primitive lavas: Earth BushveldComplexin the Roossenkal area, easternTransvaal:
Planet. Sci. Letters, v. 46, p. 419-430. Geol. Soc. SouthAfrica Trans., v. 6, p. 207-227.
Speidel,D. H., 1970, Effect of magnesiumon the iron-titanium -- 1976, Sulfidesin the upper zone of the easternBushveld
oxides:Am. Jour. Sci., v. 268, p. 341-353. Complex:ECON.GEOL.,v. 71, p. 1324-1335.
Spencer,K. J., and Lindsley,D. H., 1981, A solutionmodelfor -- 1979, A review of somerecent conceptsof the Bushveld
coexistingiron-titaniumoxides:Am. Mineralogist,v. 66, p. Complex with particular reference to sulfide mineralization:
1189-1201. Canadian Mineralogist, v. 17, p. 233-256.
Stanton, R. L., 1964, Mineral interfaces in stratiform ores: Inst. Wager, R. L., 1960, The major elementvariationof the layered
Mining Metallurgy Trans., v. 74, p. 45-79. series of the Skaergaardintrusion: Jour. Petrology, v. 1, p.
1965, Mineral interfaces in stratiform ores--author's 364-398.
reply to discussion:
Inst. Mining Metallurgy Trans., v. 74, p. --
1967, Rhythmicandcrypticlayeringin maficandultramarie
8O7-811.
plutons, in Hess, H. H., and Poldervaart, A., eds., Basalts:
Stevens,R. E., 1944, Compositionof some chromitesof the New York, Interscience,p. 573-622.
westernhemisphere:Am. Mineralogist,v. 29, p. 1-34.
Wager, R. L., and Brown, G. M., 1968, Layered igneousrocks:
Stolper,E., andWalker, D., 1980, Melt densityandthe average London, Oliver and Boyd, 588 p.
compositionof basalt:Contr. Mineralogy Petrology,v. 74, p.
7-12. Wager, R. L., Brown, G. M., and Wadsworth,W. J., 1960,
Taylor, H. P., and Forester, R. W., 1979, An oxygen and Typesof igneouscumulates:Jour.Petrology,v. 1, p. 73-85.
hydrogenisotopestudyof the Skaergaardintrusiveand its Wagner,P. A., 1928, The iron depositsof the Unionof South
country rocks:A descriptionof a 55-m.y. old fossilhydro- Africa: South Africa Geol. Survey Mem. 26, p. 21-41.
thermal system:Jour. Petrology,v. 20, p. 355-419. Walker, F., and Poldervaart,A., 1949, Karroo doleritesof the
Thiimmler, F., and Thomma, W., 1967, The sintering process: Union of South Africa: Geol. Soc. America Bull., v. 60, p.
591-706.
Metall. Rev. v., 12, p. 69-108.
Ulmer, G. C., 1969, Experimentalinvestigationof chromite Walsh,K. L., Cawthorn,R. G., andMcCarthy,T. S., 1983, The
spineIs:ECON. GEOL.MON. 4, p. 114-131. distributionof vanadiumin magnetitein the BushveldComplex
Van der Merwe, M. J., 1976, The layered sequenceof the [abs.]:Symposiumon the BushveldComplex,Univ. Pretoria,
Potgietersrustlimb of the BushveldComplex: ECON.GEOL., Nov. 1983, Programmewith Abstracts,p. 101.
v. 71, p. 1337-1351. Weedon, D. S., 1965, The layered.ultrabasicrocks of Sgurr
Vernon, R. H., 1970, Comparativegrain boundarystudiesof Dubh, Isle of Skye:ScottishJour.Geology,v. 1, p. 41-68.
somebasicand ultrabasicgranulites,nodules,and cumulates: Willemse,J., 1969, The vanadiferous magneticiron ore of the
ScottishJour. Geology, v. 6, p. 337-351. BushveldIgneous Complex: ECON. GEOL.MON. 4, p. 137-
Vincent, E. A., 1960, Ulv6spinel in the Skaergaardintrusion, 208.