chemical evolution of intercumulus liquid, as recorded in...

Chemical Evolution of Intercumulus Liquid, asRecorded in Plagioclase Overgrowth Rims fromthe Skaergaard Intrusion

MADELEINE C. S. HUMPHREYS*DEPARTMENT OF EARTH SCIENCES, UNIVERSITY OF CAMBRIDGE, DOWNING STREET, CAMBRIDGE CB2 3EQ, UK

RECEIVED FEBRUARY 14, 2008; ACCEPTED DECEMBER 9, 2008ADVANCE ACCESS PUBLICATION JANUARY 7, 2009

The intercumulus liquid of a crystal mush fills pore spaces, and typ-

ically solidifies to form overgrowths on cumulus grains and poikilitic

post-cumulus minerals. If the liquid is immobile, solidification pro-

duces zoned intercumulus minerals, as a result of progressive fractio-

nation of the residual liquid. Convection within the mush results in

buffering of the liquid composition, and thus limits mineral zonation.

For fully solidified cumulates, ‘fossil’changes in liquid composition or

porosity are difficult to identify. However, detailed study of immobile

minor components of plagioclase overgrowth rims can provide infor-

mation about the progressive solidification of intercumulus material.

Ti contents of plagioclase overgrowths, in samples from the lower-

most parts of the Skaergaard Intrusion, show strong variations with

anorthite content.With decreasing XAn,Ti concentrations first rise

and then fall, consistent with changing TiO2 contents of

the intercumulus liquid during solidification. TiO2 in plagioclase

decreases sharply at An55, reflecting local saturation of Fe^Ti

oxides.Ti in clinopyroxene oikocrysts also falls rimward, but zoning

in faster diffusing species (Fe, Mg) is limited. Other than slight

reverse zones that may occur on the plagioclase margins, XAn falls

continuously during crystallization.The reverse zoning is interpreted

as the result of compaction-driven dissolution and reprecipitation

of plagioclase. The continual decrease in XAn is exploited, together

with back-scattered electron images of the cumulates, to produce

calibrated images showing regions of progressive crystallization.

This allows the regions crystallizing at each stage of solidification

to be visualized. These images show that the final remnants of

interstitial melt were present in triangular pockets and as thin

grain-boundary melt films.This approach can provide information

about the progressive reduction of porosity during cumulate

solidification.

KEY WORDS: residual liquid; cumulate; plagioclase; porosity;

Skaergaard

I NTRODUCTIONThe formation of cumulate rocks begins with capture ofcrystals into a mushy boundary layer on the margins ofthe intrusion. The mechanisms of accumulation may varygreatly, from in situ crystallization (e.g. McBirney &Noyes, 1979), to deposition from currents (e.g. Irvine et al.,1998), or crystal settling (or flotation, e.g.Wager et al.,1960).However, the result is a crystal framework of grains withinterstitial liquid. As the mush layer solidifies, the mushand interstitial liquid experience strong changes in tem-perature, and there is compositional evolution as crystalli-zation proceeds.The nature of intercumulus crystallizationdepends on the mobility of the interstitial liquid and thepermeability of the mush. For example, adcumulus-stylecrystallization should occur where there is high permeabil-ity, such that the evolving intercumulus liquid can main-tain communication with liquid from the main magmareservoir by diffusion and/or compositional convection(e.g. Kerr & Tait, 1986; Tait & Jaupart, 1992). Perfect adcu-mulates would contain only primocryst minerals withessentially unzoned overgrowths, and no evolved phases(e.g. Morse, 1998). Conversely, in a system without circula-tion of the interstitial liquid, orthocumulus-style crystalli-zation will occur. Pure orthocumulates would containprimocrysts with zoned overgrowth rims (on plagioclaseprimocrysts) and evolved intercumulus phases that

*Telephone: þ44 (0)1223 333433. Fax: þ44 (0)1223 333450.E-mail: [email protected]

� The Author 2009. Published by Oxford University Press. Allrights reserved. For Permissions, please e-mail: [email protected]

JOURNALOFPETROLOGY VOLUME 50 NUMBER1 PAGES127^145 2009 doi:10.1093/petrology/egn076

correspond to crystallization from a trapped liquid (e.g.Wager et al., 1960; Morse, 1998).Pure adcumulates and orthocumulates are idealized

end-members of cumulus crystallization, and may occuronly rarely, if at all, in nature (e.g. Wager et al., 1960;McBirney & Hunter, 1995; Grant & Chalokwu, 1998;Morse, 1998). First, the likelihood of convection within themush will depend on the physical properties (density andviscosity) of the evolving interstitial liquid as well as themush permeability. Therefore the style of crystallizationmay change between orthocumulus and adcumulus (orvice versa) during the course of solidification. For example,if the interstitial liquid passes through a density maximumduring differentiation, this may allow compositional con-vection to be initiated (e.g. Sparks et al., 1984; Morse, 1988;Toplis et al., 2008). There is still disagreement overthe changing density of the Skaergaard liquid, but somestudies indicate that it may pass through a density maxi-mum, although at differing stages of fractionation (e.g.Wager & Brown, 1968; Hunter & Sparks, 1987; Toplis &Carroll, 1995; Tegner, 1997). Differences in mush geometrymay also result in variations in the occurrence or extent ofmovement of the interstitial melt (Be¤ dard et al., 1992).Compaction of the cumulate pile will reduce the volumeof the interstitial material crystallized but will not alterthe compositional path taken by the liquid (Meurer &Meurer, 2006). Second, the preservation of zoned over-growth rims will depend on species diffusivities relative tothe cooling rate. Diffusion rates for a given element willvary between minerals, as well as with temperature. Forvery slow cooling rates, which are relevant for theSkaergaard Intrusion, only the most slowly diffusing spe-cies will retain a zoned profile caused by continuous differ-entiation of immobile interstitial liquid. Suitable tracerspecies can therefore be used to assess how the compositionof the interstitial liquid changes with differentiation, andthus gain insights into the processes occurring in themush during solidification.For the Skaergaard Intrusion, the liquid line of descent

of the crystallizing bulk magma has been estimated usingeither experimental petrology (e.g. Toplis & Carroll, 1995;Thy et al., 2006) or reconstructions based on bulk analysesof whole-rocks or mineral separates (Wager & Brown,1968; McBirney, 1989; Jang & Naslund, 2001; Nielsen,2004). However, relatively little attention has been paid tothe compositional evolution of the interstitial liquid (e.g.Toplis et al., 2008). This study investigates the compositionof the interstitial liquid by examining the concentrations ofminor components and trace elements in plagioclase over-growths and clinopyroxene oikocrysts, in rocks from thelower parts of the Layered Series of the SkaergaardIntrusion. Slowly diffusing components (e.g. CaAl^NaSiin plagioclase) do not re-equilibrate with the interstitialliquid on the timescales of cooling for the intrusion, and

can therefore be used to distinguish periods of orthocumu-late-style crystallization, which result in zoned over-growths, from adcumulus-style crystallization withsignificant compositional convection, which results in abuffered liquid composition and unzoned overgrowths.Trace element compositional variations in the interstitialmaterial are used to constrain further the compositionalevolution of the residual liquid. XAn decreases more orless continuously during intercumulus crystallization,allowing back-scattered electron (BSE) images to be usedto visualize the spatial distribution of the liquid at eachstage of solidification. These results are discussed in thecontext of convection, compaction and compaction-drivendissolution^reprecipitation of plagioclase.

GEOLOGICAL SETT INGThe summary given here is drawn from several previousstudies that have described the Skaergaard Intrusion indetail (e.g. Wager & Brown, 1968; McBirney, 1989; Irvineet al., 1998). The intrusion is one of a series of Tertiary plu-tonic complexes that crop out on the east coast ofGreenland (Nielsen, 1987). Roughly oval in plan view, itcovers an area of �80 km2 and has an estimated volumeof 280 km3 (Nielsen, 2004). It cuts through Archaeangneiss basement, Cretaceous sediments and Tertiarybasalts. The layered rocks were originally flat-lying in thecentre of the intrusion, and have been tilted 10^208 to theSSE by regional post-solidification subsidence (Nielsen,2004), so that43�5 km of stratigraphy is currently exposed,with stratigraphically higher rocks exposed in the south.The layered rocks are divided into three series (Fig. 1):

the Marginal Border Series (MBS), which crystallizedinwards from the walls of the intrusion; the LayeredSeries (LS), which crystallized upward from the floor; andthe Upper Border Series (UBS), which crystallized down-wards from the chamber roof. Each series is subdividedfurther on the basis of cumulus mineral assemblage(Wager & Deer, 1939). In the Layered Series, cumulus oli-vine is present throughout the Lower Zone (LZ), absent inthe Middle Zone (MZ), and present again (though moreferric in composition) in the Upper Zone (UZ). The com-positions of the cumulus minerals show gradual crypticvariation towards more evolved compositions with strati-graphic height, ascribed to closed-system fractional crystal-lization of the magma body (Wager & Deer, 1939). TheMBS and UBS show equivalent fractionation trends interms of bulk composition, mineral assemblage and crypticvariations. The Hidden Zone (HZ) belongs to the LS andis not exposed, but was sampled in part by the 1966Cambridge Drill Core I and contains cumulus olivine andplagioclase. The core is �700m long and extends fromLZb (as defined by McBirney, 1989) �150m into theHidden Zone (Fig. 1).

JOURNAL OF PETROLOGY VOLUME 50 NUMBER 1 JANUARY 2009

128

SAMPLES AND METHODS USEDThe five samples in this study are taken from the HiddenZone, LZa and the lower parts of LZb, from the 1966Cambridge Drill Core I (Table 1). The bottom of the coreis thought to be close to the base of the intrusion (Maal�e,1976). The rocks are primarily olivine^plagioclase cumu-lates, with intercumulus plagioclase, augite, invertedpigeonite, Fe^Ti oxides and minor biotite and apatite(Fig. 2). In the LZb samples, it is difficult to determine

whether clinopyroxene is cumulus or intercumulus,because it initially has a poikilitic habit even when growingas a cumulus phase (Holness et al., 2007b, 2007c). Mostsamples contain at least one symplectite of Fe^Ti oxideswith a silicate phase (e.g. Haselton & Nash, 1975; Stripp& Holness, 2006).Samples were examined optically and using a JEOL

JSM-820 scanning electron microscope at the Universityof Cambridge. Plagioclase and pyroxene compositionswere analysed using a Cameca SX-100 electron

1966 core

Gneiss LZb

LZa

MZ

UZa

UZbUZc

Basalt

MBS

LZc

UTTENTALSUND

SKAERGAARD PENINSULA

IVNARMIUT

2 km

Forbindelsgletscher

LZb

SHMBS

LZc

MZ

UZa

UZb

Vandfaldsdalenmacrodyke

NUUK

THULE

TASIILAQ

SKAERGAARDINTRUSION

400 km Marginal Border Series (MBS)Lower Zone a (LZa)Lower Zone b (LZb)

Lower Zone c (LZc)Middle Zone (MZ)

Upper Zone a (UZa)Upper Zone b (UZb)Upper Zone c (UZc)Sandwich Horizon (SH)

GlacierGneissBasalts and sediments

Basistoppen SillVandfaldsdalenmacrodyke

Fig. 1. Map showing the location of the Skaergaard Intrusion and its main lithological subdivisions, after McBirney (1989). Inset shows regio-nal location map. The location of the Cambridge 1966 Drill Core is also marked.

HUMPHREYS SKAERGAARD INTERCUMULUS LIQUID EVOLUTION

129

microprobe at the University of Cambridge. Majorelements were analysed using a 15 kV, 10 nA beam; minorelements were analysed with a 100^200 nA beam. Thebeam was focused to a 2 mm spot, with peak countingtimes of 20 s for major elements and typically 40 s forminor elements. Typical analytical errors are given inTables 2 and 3. To obtain representative compositionsfrom all stages of crystallization, plagioclase grainboundaries and triple junctions were analysed as wellas primocryst cores and post-cumulus overgrowths. Forclinopyroxene, thin cusp-shaped protrusions at grain

boundaries (Holness et al., 2007a) and oikocryst marginswere analysed in addition to the interiors of oikocrysts.

PREV IOUS STUDIES OFSKAERGAARD PLAGIOCLASEPlagioclase textures and zoning patterns have been stud-ied in detail by Carr (1954), Maal�e (1976) and Toplis et al.(2008). Cores are typically euhedral and have compositionsthat gradually become more albite-rich with increasingstratigraphic height (e.g. Maal�e, 1976; Tegner, 1997; Toplis

Fig. 2. Textural features of the samples studied. (a, b) Cumulus plagioclase (pl) and olivine (ol) with interstitial augite (aug); samples 118653and 118605; cross-polarized light. (c) Cumulus augite, plagioclase and olivine with intercumulus inverted pigeonite (pig); sample 118678; plane-polarized light. (d) Cumulus olivine and plagioclase with intercumulus oxides (ox). Biotite (bi) is seen on grain boundaries (sample 118601;plane-polarized light). Field of view is �2�3mm for all photographs.

Table 1: Mineralogy and stratigraphic position of samples studied from the 1966 Cambridge Drill Core I

Sample Stratigraphic Location Cumulus Intercumulus

height (m)

118678 þ57�8 LZb Plagioclase þ olivineþ augite Plagioclaseþ augiteþ inverted pigeoniteþ Fe–Ti oxidesþ biotite

118653 þ25�6 LZb Plagioclaseþ olivine� augite Plagioclaseþ augiteþ inverted pigeoniteþ Fe–Ti oxidesþ biotite

118605 þ25�3 LZb Olivineþ plagioclase Plagioclaseþ augiteþ inverted pigeoniteþ Fe–Ti oxidesþ biotite

118601 �91�1 LZa Olivineþ plagioclase Plagioclaseþ augiteþ inverted pigeoniteþ Fe–Ti oxidesþ biotite

118590 �160�2 HZ Olivineþ plagioclase Plagioclaseþ augiteþ inverted pigeoniteþ Fe–Ti oxidesþ biotite


130

Table 2: Representative compositions of core, overgrowth and evolved rim plagioclase, for each sample

118590 (HZ) 118601 (LZa)

Core Core Core Core Over- Over- Over- Evolved Core Core Core Over- Over- Over- Evolved

growth growth growth rim growth growth growth rim

SiO2 49�93 50�36 52�18 51�74 54�93 55�42 56�93 60�98 52�48 52�09 52�39 54�74 54�09 53�41 54�05

TiO2 0�07 0�09 0�08 0�08 0�06 0�03 0�01 0�01 0�09 0�09 0�09 0�08 0�12 0�03 0�03

Al2O3 31�11 30�80 28�76 29�24 27�96 27�79 26�11 24�24 28�97 29�09 30�35 28�24 28�42 28�44 28�20

FeO 0�41 0�39 0�39 0�37 0�24 0�30 0�31 0�22 0�40 0�35 0�38 0�42 0�28 0�39 0�38

MgO 0�03 0�03 0�04 0�04 0�02 0�01 0�02 0�01 0�04 0�04 0�04 0�03 0�03 0�04 0�04

CaO 14�69 14�12 12�16 12�76 10�65 10�47 8�67 6�42 12�42 12�72 12�88 10�45 11�00 11�61 11�26

Na2O 2�95 3�24 4�28 4�07 5�13 5�22 6�34 7�50 4�29 4�11 3�89 5�14 4�79 4�86 4�79

K2O 0�17 0�21 0�27 0�23 0�34 0�34 0�41 0�60 0�27 0�31 0�28 0�34 0�37 0�31 0�39

Total 99�44 99�33 98�23 98�62 99�43 99�63 98�88 100�02 99�02 98�87 100�35 99�52 99�18 99�20 99�24

XAn 72�6 69�8 60�2 62�5 52�4 51�6 42�0 31�0 60�6 62�0 63�6 51�9 54�7 55�9 55�2

118605 (LZb) 118653 (LZb)

Core Core Core Core Over- Over- Evolved Core Core Core Core Over- Over- Evolved

growth growth rim growth growth rim

SiO2 54�87 49�69 52�08 54�08 56�89 54�42 58�66 54�57 52�61 49�16 50�24 57�00 54�40 58�49

TiO2 0�08 0�08 0�09 0�13 0�01 0�10 0�01 0�11 0�08 0�08 0�08 0�03 0�07 0�02

Al2O3 27�73 31�10 29�23 28�60 26�43 27�74 25�37 28�00 29�92 32�27 30�62 26�68 28�06 25�48

FeO 0�32 0�39 0�42 0�36 0�28 0�38 0�21 0�31 0�38 0�40 0�37 0�31 0�36 0�35

MgO 0�03 0�03 0�04 0�04 0�02 0�04 0�02 0�02 0�03 0�03 0�03 0�02 0�04 0�02

CaO 10�69 14�79 12�76 11�04 9�10 10�66 7�83 10�98 12�52 14�73 14�42 9�50 11�06 7�59

Na2O 5�19 3�04 4�09 4�99 6�23 5�18 6�93 5�04 4�13 2�79 3�32 6�03 5�02 7�02

K2O 0�41 0�17 0�31 0�38 0�36 0�39 0�29 0�30 0�22 0�15 0�14 0�39 0�39 0�46

Total 99�43 99�36 99�13 99�68 99�43 99�02 99�42 99�42 99�94 99�69 99�29 100�04 99�51 99�47

XAn 52�0 72�2 62�2 53�8 43�7 52�0 37�8 53�7 61�8 73�8 70�0 45�5 53�7 36�4

118678 (LZb)

Core Core Core Core Over- Over- Typical 2�

Growth growth errors

SiO2 50�51 50�83 50�43 53�82 53�82 53�87 1�14

TiO2 0�06 0�08 0�08 0�09 0�05 0�09 0�018

Al2O3 30�55 29�99 30�33 28�51 28�95 28�14 1�61

FeO 0�42 0�42 0�43 0�34 0�33 0�39 0�06

MgO 0�04 0�04 0�03 0�03 0�04 0�04 0�015

CaO 14�13 13�58 13�98 11�25 11�22 11�26 0�42

Na2O 3�31 3�59 3�35 4�85 4�96 4�85 0�46

K2O 0�22 0�23 0�22 0�35 0�40 0�42 0�05

Total 99�36 98�82 98�93 99�32 99�85 99�16

XAn 69�3 66�7 68�9 55�0 54�3 54�8

The full dataset is given in Electronic Appendix 1.


131

et al., 2008). Many grains are strongly twinned.Crystallized melt inclusions are common, manifest asnegative-crystal shaped patches of albitic plagioclase, occa-sionally associated with other minerals (Hangh�j et al.,1995; Fig. 3a). Cores may be unzoned or show oscillatoryvariations of amplitude �3mol% An or less (Wager &

Deer, 1939; Carr, 1954; Maal�e, 1976). Maal�e (1976) recog-nized two types of oscillatory textures: simple plagioclaseare most common from the LZ upwards and typicallyhave euhedral oscillatory zonation (Fig. 3b), whereas com-plex plagioclase are most common in the HZ and showstrong, irregular internal resorption surfaces cross-cutting

Table 3: Representative clinopyroxene compositions from each sample

Analysis SiO2 TiO2 Al2O3 FeO MnO MgO CaO Na2O Total XEn XWo Mg-no.

118590_cpx1g 52�15 0�36 1�16 9�93 0�23 13�88 20�89 0�26 98�86 43�6 40�3 71�4

118590_cpx1f 51�04 0�98 2�62 9�46 0�25 13�44 20�44 0�43 98�64 43�9 40�2 71�7

118590_cpx2a 51�14 0�80 1�88 10�23 0�26 13�43 20�39 0�33 98�46 43�3 39�7 70�1

118590_cpx2e 52�20 0�37 1�20 10�81 0�26 13�55 20�59 0�29 99�27 43�0 39�4 69�1

118590_cpx2f 51�72 0�43 1�30 9�54 0�27 13�45 21�18 0�27 98�16 44�8 39�5 71�5

118601_cpx3o 51�16 0�95 2�08 9�28 0�23 14�19 20�37 0�34 98�62 43�0 41�7 73�2

118601_cpx1f 51�66 0�77 2�10 8�51 0�29 14�71 19�66 0�54 98�28 42�0 43�8 75�5

118601_cpx4a 51�63 0�71 1�87 9�03 0�24 14�58 20�41 0�31 98�77 42�8 42�5 74�2

118605_cpx2b 51�40 0�63 1�79 9�84 0�26 14�02 20�64 0�31 98�88 43�1 40�8 71�8

118605_cpx4a 52�51 0�17 0�74 9�36 0�25 13�84 21�97 0�20 99�04 45�3 39�7 72�5

118653_cpx2h 51�90 0�70 1�85 8�85 0�24 14�51 20�63 0�30 99�01 43�2 42�3 74�5

118653_cpx5j 50�97 0�74 1�98 9�28 0�25 14�22 20�82 0�30 98�56 43�5 41�3 73�2

118653_cpx1b 51�04 0�84 2�06 8�82 0�27 13�86 20�85 0�35 98�11 44�4 41�0 73�7

118653_cpx3h 51�86 0�89 2�10 9�29 0�26 14�01 20�38 0�30 99�13 43�2 41�4 72�9

118678_cpx1h 50�78 0�72 2�07 9�76 0�25 14�50 20�07 0�35 98�51 41�9 42�1 72�6

118678_cpx1e 51�32 0�65 2�00 9�21 0�25 14�44 20�44 0�28 98�59 42�8 42�1 73�6

118678_cpx1b 51�17 0�97 2�37 9�04 0�24 14�17 20�84 0�36 99�16 43�8 41�4 73�6

Typical 2� errors (wt %) 1�10 0�03 0�1 1�20 0�07 0�62 0�54 0�16

The full dataset is given in Electronic Appendix 2.

pl

pl

px

MI

r

r

cpx

pl

(a) (b)

Grain margin

Post-cumulusovergrowth

Primocrystcore

Fig. 3. Typical plagioclase textures in the Lower Zone of the Layered Series. (a) Euhedral plagioclase core with two narrow reverse zones (r)leading to the normally zoned rim. A euhedral, crystallized melt inclusion (MI) is seen in the centre of the core. A resorption horizon (whitedashes) cuts across parts of the rim of the grain. Field of view �2�3mm. (b) Subtle, euhedral oscillatory zoning (white arrowheads) is seen inthe core of some plagioclase grains. Scale bar: 200 mm.


132

the oscillatory variations. The complex textures disappearabove þ30m in the stratigraphy, and may be related totemperature variations experienced during crystal settlingprior to the onset of convection (Maal�e, 1976) or tomagma injection in the early life of the intrusion (Holnesset al., 2007b). The oscillatory variations have been ascribedto variations in temperature (Wager & Deer,1939; Maal�e,1976) or pH2O (Carr, 1954) during magma convection.Maal�e (1976) also described ‘divided zoning’, whichoccurs in the MZ and UZ and comprises crystallographi-cally controlled blebs of compositionally distinct plagio-clase in the primocryst cores. This texture was ascribed toskeletal growth during supercooled crystallization(Maal�e, 1976).The margins of the primocrysts are marked by one or

more reverse zones, which are succeeded by normallyzoned post-cumulus overgrowths that may be locallyembayed (Maal�e, 1976; Fig. 3a). The composition of theovergrowth rims also becomes more albitic with increasingheight in the intrusion (Toplis et al., 2008).

MINERALOGICALCOMPOSIT IONSPlagioclaseMajor elements

Plagioclase compositions from the samples studied span awide range of compositions (Table 2; Electronic Appendix1, available at http://www.petrology.oxfordjournals.org),from An74 to An30. The different textural features corre-spond to distinct major element compositions. The coresof cumulus grains are An-rich, typically varying fromAn60 to An70. The average core composition becomesslightly more albitic with increasing stratigraphic height,in agreement with previous studies. Oscillatory zonedcores show muted compositional variations of 3^6mol%anorthite, in agreement with the measurements of Carr(1954) and Maal�e (1976). Post-cumulus overgrowth rimsare more sodic than the cores, with compositions typicallyAn57^50. Occasionally, reverse zoning of �5mol%anorthite is observed (An50^55).The most evolved composi-tions (An56^31) are found at at grain boundaries, or atthree-grain triple junctions. There is no systematic differ-ence in overgrowth composition between samples from dif-ferent stratigraphic horizons.

Minor elements

Minor elements in plagioclase include Fe, Ti, Mg and K.Plagioclase compositions from Skaergaard show strongvariations in minor element concentrations.

(1) K2O contents correlate negatively with anorthite con-tent. The upper part of the trend is sharp and welldefined; however, the compositions are scattered tolower K2O values for a given XAn (Fig. 4a). In

particular, this is true for the post-cumulus over-growths (An55^50) and cores (An65^60).

(2) Ti contents vary strongly, from 0�01 to 0�133wt %TiO2. TiO2 concentrations increase from An75 toAn60, then increase more rapidly to a maximum of0�133wt % at �An55 (Fig. 4b). Post-cumulus over-growths record a sharp drop in TiO2 from 0�133wt% to �0�03wt % over a small range in XAn

(�An55^50). At �An50, TiO2 continues to decreasebut the gradient shallows, and the most evolved rimsshow consistently lowTiO2 contents.

(3) MgO concentrations are low, typically 50�05wt %(Fig. 4d). Plagioclase cores record a wide variation ofMgO contents, between 0�019 and �0�045wt %; themost An-rich cores have slightly lower Mg concentra-tions. The post-cumulus overgrowths show decreasingMgO from 0�045wt % to �0�02wt % as anorthitecontent decreases from An60 to An50. The mostevolved rims contain very low MgO concentrations,typically50�02wt %.

(4) FeO compositions are scattered but decrease steadilywith falling XAn (Fig. 4c). There is no clear pattern ofrising and falling concentrations.

Plagioclase traverses

Core to rim traverses through individual plagioclase crys-tals show similar compositional variations in major andminor elements (Fig. 5), but can provide further informa-tion about the progressive changes in composition. XAn

decreases rimwards from the core (�An60^70) to an over-growth rim of �An45^55. In some traverses, there is a rimplateau at �An50, consistent with the observations of Topliset al. (2008). K2O concentrations increase rimward to�0�35wt %, and FeO typically shows a gradual, continu-ous decline towards the rim. TiO2 concentrations gradu-ally increase rimwards, then strongly decrease just insidethe rim, starting at �An55. In detail, the profiles showthat the drop in TiO2 may coincide with a slight increasein XAn. MgO concentrations show a clear decreaserimward.In some traverses, the outer margins of the grains show

compositional reversals. XAn and FeO reach a minimumand increase once more just inside the rim, whereas K2Ocontents reach a maximum and then decrease. TiO2 con-centrations continue to fall, whereas there is no discerniblechange in MgO.

Clinopyroxene compositional variationsClinopyroxene compositions are more Ca-rich than thecrystallization trend defined by Brown et al. (1957) andBrown & Vincent (1963) but are consistent with theobserved subsolidus trend defined by Nwe (1976) for rocksfrom the Lower Zone. Pyroxenes from each sample plotwithin a restricted range of En contents (Wo44^54En46^55;Table 3; Electronic Appendix 2), but do not show


133

http://www.petrology.oxfordjournals.org

systematic compositional variation with increasing strati-graphic height. TiO2 correlates weakly with Al2O3

(Fig. 6a) and Na2O (Fig. 6b), but not with other elements.MnO decreases slightly with increasing Mg-number(Fig. 6c). Mg and Fe contents do not vary spatially, butcompositional profiles across towards the outermost mar-gins of clinopyroxene oikocrysts (Fig. 7) show that Al andTi contents typically decrease slightly towards oikocrystrims and towards cpx^plag^plag triple junctions, consis-tent with previous observations of oikocryst growth(Claeson et al., 2007).

MINOR COMPONENTS RECORDCHANGES IN INTERCUMULUSL IQU ID COMPOSIT IONIn agreement with many previous studies (e.g.Wager et al.,1968; Maal�e, 1976; Naslund, 1984), the major element

compositions of plagioclase overgrowth rims atSkaergaard show a continuous, gradual decrease of XAn

towards the margins.This reflects the decrease of anorthitecontent with decreasing temperature and increasing differ-entiation of the melt. In practice, major-element diffusionin plagioclase is too slow to allow equilibrium to be main-tained between crystal and liquid, so decreasing XAn is areliable indicator of increasing differentiation. Slowly dif-fusing trace elements should also record changes to thecomposition of the intercumulus liquid during differentia-tion. The TiO2 content of plagioclase overgrowthsincreases to a maximum and then declines as XAn

decreases (Fig. 8), suggesting that the intercumulus liquidpasses through a maximumTiO2 content, probably relatedto the onset of Fe^Ti oxide crystallization in the mush porespace. This is consistent with the sequence of experimentalliquid compositions of Thy et al. (2006), which showed aTimaximum at the onset of Fe^Ti oxide crystallization, and

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

20 30 40 50 60 70 80X An

Post-cumulus overgrowthsPrimocryst cores

Grain margins/ triple junctions0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

20 30 40 50 60 70 80

0.0

0.2

0.4

0.6

0.8

1.0

20 30 40 50 60 70 800.00

0.02

0.04

0.06

0.08

0.10

20 30 40 50 60 70 80

wt%

K2O

Wt%

TiO

2

X An

X An

Wt%

FeO

Wt%

MgO

X An

(a) (b)

(c) (d)

GDG

D

GD G

D

Fig. 4. Minor element compositional variation in plagioclase (point analyses), sorted on the basis of textural observations. Bars giverepresentative�1� errors. Inset shows theoretical growth (continuous lines, G) and diffused (dashed lines, D) compositional profiles. Thegrowth profiles are estimated on the basis of likely changes in temperature, liquid composition and XAn during fractionation, with Dpl fromBindeman et al. (1998) or equation (1). Diffused profiles are estimated in the same way, assuming constant temperature and liquid composition(Costa et al., 2003).


134

118605-6pA

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0 50 100 150 200 250 300 350 400 450 500 55030

35

40

45

50

55

60

65

70

FeO

TiO2

TiO2

TiO2

K2O

K2O

K2O

MgO

X An

118678-4pA

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0 40 80

0 40 80

120 160 200 240 280 32030

35

40

45

50

55

60

65

70

75

118590-5pD

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

120 160 200 240 28030

35

40

45

50

55

60

65

70

wt%

oxi

de

XA

n

wt%

oxi

de

XA

n

wt%

oxi

de

XA

n

CORE

CORE

FeO

MgO

X An

FeO

MgO

X An

REVERSE RIM OVERGROWTH

REVERSERIM

OVERGROWTH

COREOVERGROWTH

Fig. 5. Representative plagioclase traverses from core to rim. Typically, TiO2 contents rise slightly and then fall rimward. XAn

generally decreases rimward. Reversed rims occur in some grains, manifest as increasing XAn and decreasing TiO2, and coincide withfalling K2O.


135

with studies of lunar plagioclase (Steele et al., 1980; Smith& Brown, 1988) in which Ti concentrations in plagioclaseare inferred to give an indication of the Ti content of theliquid. In contrast, the MgO content of plagioclase over-growths decreases continuously, perhaps reflecting contin-ual crystallization of intercumulus clinopyroxenecrystallization (or depletion of MgO as a result of fractio-nation in the bulk magma).The geometry of clinopyroxene oikocryst growth can be

highly irregular (Claeson et al., 2007). However, composi-tional profiles show that, in general, the margins are

depleted in Ti and Al relative to the centres of the oiko-crysts (Fig. 7). This is consistent with removal of Ti andAl components from the liquid during crystallizationof Fe^Ti oxides and plagioclase, respectively.

0.10

0.15

0.20

0.25

0.30

0.35

64 66 68 70 72 74 76 78Mg-number

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

0.0 0.5 1.0 1.5 2.0 2.5 3.0wt% Al2O3

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

0.0 0.1 0.2 0.3 0.4 0.5 0.6 wt% Na2O

wt%

MnO

(a)

(b)

(c)

wt%

TiO

2w

t% T

iO2

Fig. 6. Compositional variation of intercumulus clinopyroxene. Barsgive�1� errors.

Traverses118678118605118601118590118653

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

20 30 40 50 60 70 80

wt%

TiO

2

X An

(Pre-cumulus)Stage IStage II

0.851.0

0.68

0.1

0.2

0.3

0.4

0.5

0.6

Fig. 8. CalculatedTiO2 content of plagioclase during fractional crys-tallization (model curve). All plagioclase analyses (traverses andpoint analyses) are shown (grey circles). (See Table 4 for details ofparameters used for the calculation.) Numbers denote the fraction ofintercumulus liquid remaining (tick marks at higher XAn representthe ‘pre-cumulus’ stage; see text). The composition of the initial inter-cumulus liquid is calculated assuming fractional crystallization of oli-vine and plagioclase from the Skaergaard parental melt (pre-cumulusstage). The intercumulus crystallization calculations comprise twostages (horizontal bars): Stage 1 is intercumulus crystallization priorto Fe^Ti oxide saturation; Stage 2 is intercumulus crystallizationafter Fe^Ti oxide saturation. Plagioclase with composition4An60 isclassified as core material on the basis of textural observations.

0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

0.90

118678_cpx1a118653_cpx5118601_31b118605_10cpxa

wt%

TiO

2

0.00

0.50

1.00

1.50

2.00

2.50

0 100 200 300 400 500 600Distance from margin (µm)

wt%

Al 2

O3

(a)

(b)

Fig. 7. Representative clinopyroxene compositional traverses, withdistance from rim of interstitial material. (a) TiO2 and (b) Al2O3typically decrease towards the margins.


136

Fe, Mg compositional variations inplagioclaseTiO2 concentrations in plagioclase show strong variationsthat are inferred above to result from changes in the liquidTi content. MgO contents of post-cumulus overgrowthsshow decreasing MgO and XAn, consistent with Mg deple-tion of the liquid as a result of clinopyroxene oikocrystgrowth. Only limited variation of FeO is seen in plagio-clase (Fig. 4). The observed Fe variation in the overgrowthrims typically comprises a steady decrease with XAn.However, a rise and then steeper decrease, similar to thetrend forTiO2, would be expected as a result of local satu-ration in Fe^Ti oxides. The muted FeO variations cannotbe the result of compositional convection within the crystalmush, because theTiO2 contents are not buffered (see sub-sequent discussion). Fe partitioning into plagioclase hasbeen shown to increase with oxygen fugacity (e.g.Phinney, 1992; Wilke & Behrens, 1999). However, althoughthe fO2 of the Skaergaard Intrusion is thought to havedecreased during differentiation (e.g. Frost & Lindsley,1992; Toplis & Carroll, 1996), the change of DFe

pl with fO2

is very small at the quartz^fayalite^magnetite buffer(QFM) and below (Phinney, 1992). Therefore changes infO2 cannot explain the limited variation in Fe. Despitethe paucity of Fe partitioning data, Fe is known to occurin plagioclase as Fe3þ and minor Fe2þ, substitutingfor Al3þ or Ca2þ (Longhi et al., 1976; Smith & Brown,1988; Lundgard & Tegner, 2004). The observed negativecorrelation of Fe with XAn is consistent with publisheddatasets, which show that the Fe content of plagioclasedecreases with anorthite content and with crystallizationtemperature (e.g. Smith, 1983; Smith & Brown, 1988;Tegner, 1997).A likely explanation for the relatively insignificant vari-

ation of Fe and Mg in plagioclase compared withTi is thedifference in cation diffusivity. Diffusion will act to reducevariations in minor components within the crystal, at arate dependent on temperature and species diffusivity, andonly very slowly diffusing species will closely track thechanging melt composition. In plagioclase, the coupledsubstitution CaAl^NaSi is very slow (5�4�10^22m2/s at12008C, Grove et al., 1984) and in practical terms, diffusionof the initial anorthite profile will be minimal. Ti4þ diffu-sion is likely also to be very slow, because the activationenergy for diffusion increases with ion charge (e.g.Hofmann & Magaritz, 1977; Jambon, 1982), so composi-tional variations in TiO2 can easily be preserved.Experimental data for Mg and Fe diffusion in plagioclaseare limited, but the diffusivity of Mg in An95 has beenmeasured at �10^19m2/s at 12008C (LaTourrette &Wasserburg, 1998), a little faster than that of Sr at thesame temperature (Cherniak & Watson, 1994; Giletti &Casserly, 1994), and much faster than CaAl^NaSi. Thecharacteristic diffusion lengthscales (L¼ˇ4Dt) for Mg

and CaAl^NaSi support this argument. On the basis ofthe estimated cooling time of the Skaergaard Intrusion(�34 kyr, Gettings, 1976), at a constant temperature of11008C the diffusion lengthscale for CaAl^NaSi interdiffu-sion is �10 mm whereas that for Mg is �510 mm (diffusioncoefficients from Grove et al., 1984; LaTourrette &Wasserburg, 1998). Plagioclase overgrowths typically haveapparent thicknesses of �200^300 mm, so assuming thatthe diffusivity of Fe is similar to that of Mg, any initial var-iations in Mg and Fe will be significantly dampened by dif-fusion. Eventually, the compositions will tend towards theequilibrium compositional profiles for Mg and Fe (Fig. 4),which will reflect only changes in partition coefficientrelated to the smooth normal zoning of the underlyingXAn profile (Zellmer et al., 1999; Costa et al., 2003).

Compositional variations in clinopyroxeneFractional crystallization of intercumulus clinopyroxeneshould produce significant Mg, Fe zoning in the oikocrysts,but no variation is observed. This is probably due to rapidMg^Fe diffusion during cooling, or partial re-equilibrationwith the interstitial liquid during growth (e.g. Barnes,1986). The available diffusion data show that Mg^Fe diffu-sion is approximately two orders of magnitude faster thanthat of Al (Sautter et al., 1988; Anovitz, 1991; Dimanov &Sautter, 2000).This is consistent with the observed decreasein diffusivity for cations with small ionic radius (vanOrman et al., 2001); Ti diffusion should therefore also beslower than that of Fe^Mg. The lack of Mg^Fe zonationand presence of Al^Ti zoning is therefore consistent withvariations in species diffusivity. Al in clinopyroxenedecreases with increasing melt fractionation (LeBas, 1962;Loucks, 1990), and the correlation between Ti and Alreflects charge balancing (Brown et al., 1957; Kushiro,1960). The decrease in Al and Ti observed towards themargins of oikocrysts and towards cpx^plag^plag triplejunctions is therefore consistent with decreasing Ti (and/or Al) in the intercumulus melt during its solidification.

MODELL ING TIO2

CONCENTRAT IONS IN THEINTERCUMULUS L IQU IDVariations in the TiO2 contents of plagioclase are inter-preted as the result of changingTiO2 in the residual inter-cumulus liquid.Therefore, theTiO2 contents of plagioclasecan be used to reconstruct the liquid composition andhence investigate crystallization of the interstitial liquid.Changes in mineral chemistry during intercumulus crys-tallization are caused by changes in temperature and meltcomposition. Further complicating factors include the pos-sibility of large-scale compositional convection within thecrystal mush (e.g. Morse, 1986) or melt loss through com-paction (Meurer & Boudreau, 1998a). It is therefore


137

assumed initially that there is no compaction, no composi-tional convection within the mush, and only minimal dif-fusive exchange within the interstitial liquid. The validityof these assumptions is discussed below. Important para-meters for the model include the initial TiO2 content ofthe interstitial liquid, the initial porosity of the mush, thepartition coefficient for Ti in plagioclase (DTi

pl) and thebulk distribution coefficient for TiO2 (which depends onthe modal proportions of the crystallizing intercumulusmaterial as a function of temperature). The values of theseparameters are discussed below.

Initial TiO2 content of the interstitial liquidThe bulk composition of the parental liquid that crystal-lized to form the Skaergaard Intrusion has been debatedin many previous studies. The parental liquid compositionhas been estimated from chilled marginal rocks (Wager &Brown, 1968; Hoover, 1989), from the chilled margins ofassociated dyke swarms (Brooks & Nielsen, 1990) andfrom mass-balance models (Nielsen, 2004) based on bulk-rock analyses (McBirney, 1989). Experimental studieshave recently used samples of the dykes (Thy et al., 2006)or synthetic equivalents of them (Toplis & Carroll, 1995).Estimates have TiO2 contents in the range 2�05^2�92wt% (Brooks & Nielsen, 1990), 2�35^2�72wt % (Hoover,1989), and 3�09wt % (Nielsen, 2004). The modelling pre-sented here uses phase relationships derived from experi-mental studies (Toplis & Carroll, 1995; Thy et al., 2006).The TiO2 content of the parental liquid is assumed tohave an initial value of 2�05wt %, equivalent to the com-position of the experimental starting material used in thesestudies.

Ti (plagioclase^melt) partitioncoefficient, DTi

pl

To reconstruct theTiO2 content of the interstitial melt (m),the partition coefficient forTi in plagioclase (DTi

pl¼CTi

pl/CTi

m, where C is the concentration) must be known. Dataon DTi

pl from the literature are few, in part because of poorcounting statistics for typical electron microprobe analysesof Ti. Bindeman et al. (1998) reported dependence on bothtemperature and XAn, with an Arrhenius relationship.However, the experiments reported by Bindeman et al.(1998) were performed in air and are therefore not appro-priate for Skaergaard magmas. Be¤ dard (2006) alsoreported similar Arrhenius equations for regressions ofdata compiled from previous studies, but with very lowR2. In this study, DTi

pl is calculated by linear regression ofthe experimental glass and plagioclase compositions ofThy et al. (2006), such that

ln DplTi ¼ 0 � 00535Tð8CÞ � 9 � 458 ð1Þ

Using the same data, there is no correlation of ln DTipl

with XAn. However, the co-variation of XAn with T is

taken into account when calculating the fractional crystal-lization trend (see below).

Initial porosity of the mushThe initial porosity pi of the mush is defined as the liquid-filled pore space present immediately following crystaldeposition. Crystallization will reduce the porosity overtime, to the residual porosity pr (Morse, 1986). In theabsence of compaction, pi will be the same as the propor-tion of trapped liquid, such that pr¼ pi F, where F is thefraction of liquid remaining during progressive fractionalcrystallization. The initial porosity of a crystal mush hasbeen estimated from experiments (e.g. Jackson, 1961;Finney, 1970; Philpotts et al., 1999) and numerical models(e.g. Jerram et al., 1996; Saar et al., 2001; Rudge et al.,2008), whereas the trapped liquid fraction has been calcu-lated from geochemical arguments (e.g. Irvine, 1980;Tegner et al., in preparation).Numerical modelling shows that the initial porosity of

the crystal framework will be influenced by the rate ofaccumulation (Blumenfeld et al., 2005), crystal shape andaspect ratio (e.g. Williams & Philipse, 2003; Rudge et al.,2008) and preferential alignment of crystals (Saar et al.,2001). Interconnected frameworks of plagioclase crystalsform experimentally at very high porosities (�75 vol.%,Philpotts et al., 1999). Random packings of monodispersespheres give a maximum theoretical packing density of�0�64 (equivalent to a minimum porosity of 36%).Packings of more elongate particles have higher porosity(Rudge et al., 2008); however, polydisperse grain packingswill have lower porosity relative to monodisperse packingsof the same grain shapes (e.g. Bezrukov et al., 2001).Several researchers have estimated the trapped liquid

fraction from the concentrations of incompatible elementsin cumulate rocks. Irvine (1980) estimated 50^58%trapped liquid in the marginal picrites of the MuskoxIntrusion. For Skaergaard specifically, Henderson (1970)determined a trapped liquid fraction of 15^24% in samplesfrom LZa from bulk P contents. Similarly, Tegner et al.(in preparation) proposed trapped liquid fractions thatdecrease from 28^47% in LZa to �4^5% in MZ andUZa. Compaction is minimal in the lower parts of theLower Zone (Tegner et al., in preparation), which suggestsan initial porosity equivalent to the trapped liquid fraction,�15^47%. The initial porosity of a crystal mush can alsobe estimated from the textures of glassy cumulate nodules.Studies by Tait (1988) described cumulate nodules fromLaacher See, Germany with �15^35 vol.% glass. Holness& Bunbury (2006) reported glass contents of 25^40% foramphibole-bearing nodules from Kula,Turkey.Based on these previous studies, the initial porosity of

the polydisperse olivineþplagioclase� clinopyroxenemush at Skaergaard, prior to compaction and intercumu-lus crystallization, is taken to be 35%.


138

Crystallization of the intercumulusmaterialBecause there is no systematic difference in plagioclaseovergrowth composition between samples, previouslypublished data from HZ, LZa and lower LZb plagioclaseare pooled and considered together. Thus, the modellingwill produce an ‘average’ result for the lowermost partsof the Skaergaard Intrusion. The fractional crystallization(FC) model presented here has three stages. The firststage is effectively a period of ‘primocryst’ growth, whichtakes into account that the primocrysts are depositedinto an ‘LZ’ liquid that is differentiated from theSkaergaard parental liquid. Olivine and plagioclaseare assumed to be crystallizing in constant proportions(70% plagioclase and 30% olivine; Toplis & Carroll, 1996;Table 4) from the parental liquid, which had 2�05wt %TiO2. Textural observations show that the most evolvedprimocryst cores have composition �An60; therefore thispre-cumulus stage of the calculations ends whenthe plagioclase composition reaches An60 (see below).When this occurs, TiO2(m) is 3�30wt % (Table 4).The second stage describes intercumulus crystallization

prior to saturation of Fe^Ti oxides. This stage begins with3�3wt%TiO2 in the interstitial liquid, at An60, andassumes55 vol.% overgrowth of plagioclaseþ 45 vol.% crystalliza-tion of clinopyroxene oikocrysts (Toplis & Carroll,1996).Tiis strongly partitioned into the liquid, and is thereforedescribed by a low bulk partition coefficient (DTi

B). Thepoint at which the melt becomes saturated in Fe^Ti oxides(the beginning of phase 3) is determined from the experi-mentally derived constraint ofToplis &Carroll (1996):

wt %TiO2ðmÞ ¼ 0 � 0409Tð�CÞ � 40 � 3 ð2Þ

The third stage describes intercumulus crystallizationfollowing saturation of the liquid with Fe^Ti oxides.It assumes overgrowth of plagioclaseþ crystallization ofclinopyroxene oikocrystsþ crystallization of intercumulusFe^Ti oxides in the ratio 40:40:20 (Toplis & Carroll, 1996;Toplis et al., 2008). Ti is compatible in the bulk crystallizingassemblage, and has a high DTi

B. For all phases, DTi iscalculated using partition coefficients from Be¤ dard (2005)for olivine and Hart & Dunn (1993) for clinopyroxene,with DTi

pl from Thy et al. (2006) as previously described(Table 4).For all three stages of the calculation, the proportion of

liquid remaining (F) is quantified as a function of temper-ature using an empirical calibration (Toplis & Carroll,1996):

TðKÞ ¼ 1295þ 265F � 290F2 þ 175F3 ð3Þ

This implicitly assumes that the liquidus of the inter-cumulus liquid is the same as that of the bulk magma.The variation of XAn with temperature is estimated fromThy et al. (in preparation):

Tð�CÞ ¼ 3 � 61XAn þ 899 ð4Þ

The TiO2^XAn plagioclase compositions are calculatedas follows:

(1) For each value of F, calculate temperature using equa-tion (3).

(2) Calculate XAn using equation (4), and hence calculateDTi

pl.(3) Calculate TiO2(m) in the liquid using the

Rayleigh fractional crystallization (FC) equation CL/C0¼ f (D ^ 1), and the bulk distribution coefficient for

Table 4: Details of the calculations used to estimateTiO2 in plagioclase as a function of XAn

DTiB DTi

pl C0 (wt % CL (wt % fi fL Fi FL Ti (8C) Tf (8C) XAn at Residual

TiO2) TiO2) end of phase porosity (%)�

‘Pre-cumulus’ stage (fractional

crystallization of 70% plagþ 30%

ol from Skaergaard parental liquid)

0�037 0�035 2�05 3�30 1�0 0�61 1�0 0�61 1172 1115 An60 100

Intercumulus crystallization, Stage 1

(fractional crystallization of 55%

plagþ 45% cpx)

0�20 0�025 3�30 4�48 1�0 0�685 0�61 0�42 1115 1095 An54 23

Intercumulus crystallization, Stage 2

(fractional crystallization of 40% plagþ

40% cpxþ 20% Fe–Ti oxides)

3�17 0�023 4�48 0�07 1�0 0�017 0�42 0�06 1095 1038 An38 3�5

Fractional crystallization is assumed, CL/C0¼ F(D – 1), where CL is concentration of element in the liquid, C0 is initial liquidconcentration, F is fraction of interstitial liquid remaining [used to estimate the liquidus temperature, equation (2)], f isfraction of liquid remaining [within each stage of the calculation; used to calculate TiO2(m)]. Subscripts i and L indicate‘initial’ and ‘final’ values within each stage. DTi

B is bulk partition coefficient for Ti, calculated using partition coefficientsDTi

ol¼ 0�04 (Bedard, 2005), DTi

cpx¼ 0�4 (Hart & Dunn, 1993) and DTi

pl calculated from XAn (see text for details).�Calculated residual porosity assumes negligible compaction and an initial porosity of 35% (see text).


139

Ti. It should be noted that f is distinct from F in equa-tion (3) and represents the fraction of liquid remain-ing within each stage of the calculation, such that f isreset to zero at each new stage. DTi

B changes at eachstage (e.g. when Fe^Ti oxide saturation is reached).

(4) CalculateTiO2 in plagioclase from steps (2) and (3).

Using a combination of equations (3) and (4) indicates arange of intercumulus crystallization conditions fromAn60 at �11158C, to An38 at �10358C. Table 4 summarizesthe calculation process and the values of the coefficientsused.Figure 8 shows the results of the fractional crystalliza-

tion (FC) calculations described above. Despite the sim-plicity of the model, the calculated plagioclase Ti trendmatches the shape and composition of the analytical datawell. At the point of oxide saturation, the liquid reaches amaximum calculated TiO2 content of �4�5wt %, in goodagreement with the experimental liquid compositions ofThy et al. (2006). The onset of Fe^Ti oxide crystallizationcauses a strong decrease in calculated plagioclase Ti con-tent. These features support the interpretation that the Ticompositional variations in plagioclase can be explainedby changes to the composition of the interstitial liquid.TiO2(m) can also be estimated using the Ti contents ofintercumulus clinopyroxene, which crystallizes at thesame time as the plagioclase overgrowth rims. Assuminga constant DTi

px of 0�4 (Table 4), TiO2(m) is estimated at0�2^2�9wt %. Although the highest TiO2 contents are notrecorded, the range is consistent with the range of TiO2(m)

calculated from plagioclase compositions.

DISCUSS IONAlthough the simplified model produces a trend that issimilar to the data, in detail there are aspects of the datathat are not matched by the calculated trend, as follows.

(1) The calculations show that the rock is fully solidified(zero residual porosity) at An34, whereas the mostevolved plagioclase compositions analysed reachAn30. The modelled final plagioclase has no TiO2 atAn35, whereas the data show 0�01^0�02wt % TiO2.

(2) The drop inTiO2 between An50 and An55 is accom-panied by a slight increase in XAn (Fig. 4b). The cal-culated FC trend does not match this, with continualdecrease in XAn.

The first point is related to the range of calculated XAn,which would be altered by changes to the liquidus equation(3) and/or equation (4). The liquid line of descent of theSkaergaard magma has been contested for decades and isstill unresolved (e.g. Hunter & Sparks, 1987; McBirney &Naslund,1990; Morse,1990; Toplis & Carroll,1996; Ariskin,2003; Thy et al., 2006), and the compositional evolution ofthe intercumulus liquid is even less well understood.

Furthermore, very few experimental data are available atevolved plagioclase compositions (An50 and below; Thyet al., in preparation). Despite these limitations, experi-mental calibrations are still probably the best means ofestimating crystallization conditions. The observation ofslight reverse zoning associated with decreasing TiO2

(Fig. 4b) is interpreted as equivalent to the reversalsdescribed by Maal�e (1976), whose analyses also showeddecreasing TiO2. Similar features have also been observedby Shimizu (1978) for Skaergaard and other mafic intru-sions (e.g. Kiglapait troctolites, Morse & Nolan, 1984;Harp Lake, Emslie, 1980). At Harp Lake, pyroxene tra-verses also showed decreasing Al contents towards thegrain margins (Emslie, 1980), as observed at Skaergaard.The reversals in XAn cannot be explained by interactionwith the overlying magma reservoir, because this wouldresult in a buffering of the TiO2 content to higher values,instead of the continual decrease that is observed. Instead,the reverse zoning is interpreted as the record of compac-tion-driven resorption and reprecipitation of plagioclase.Resorption of unfavourably oriented plagioclase has beendescribed by several workers (e.g. Maal�e, 1976; Nicolas &Ildefonse, 1996; Meurer & Boudreau, 1998b). The resorbedmaterial, which has higher XAn than the crystallizinggrains, is redeposited in more favourable orientations(Maal�e, 1976) and results in reverse zoning. The TiO2

content of the crystallizing plagioclase continues to fallduring reverse zoning, which demonstrates that intercumu-lus crystallization of Fe^Ti oxides continues during resorp-tion of plagioclase.

Compositional convection within thecrystal mushPlagioclase profiles from throughout the Lower Zone inSkaergaard were reported byToplis et al. (2008). They alsoobserved outer rims with constant or slightly reversed pla-gioclase compositions at An50^55, and interpreted these asevidence of compositional convection within the mush.They argued that the residual liquid passed through a den-sity maximum following crystallization of Fe^Ti oxides,resulting in gravitational instability of the residual liquid.In other words, intercumulus crystallization initiallyoccurs in situ, and is followed by a period of crystallizationwhere the liquid is mobile, allowing buffering of the plagi-oclase composition (Toplis et al., 2008). The decrease ofTiO2 in reversely zoned plagioclase confirms that thereversals occur after Fe^Ti oxide saturation. However, thestrong decrease of TiO2 in plagioclase during formation ofthese intermediate rims means that crystallization must bein situ. Any overlying melt circulating by convection mustbe less evolved than the interstitial liquid; therefore con-vection would buffer the interstitial liquid to more calcic,but also more Ti-rich compositions. The plagioclase com-positional traverses (Fig. 5) show that reverse zoning isalways accompanied by decreasing TiO2, and therefore


140

that there can only have been minimal chemical commu-nication with the main magma reservoir.Fe^Ti oxides make up only �2 vol.% of the samples stu-

died, and hence about 5 vol.% of the intercumulus mate-rial (given an initial porosity of 35%). The oxide-richregions are heterogeneously distributed, leaving parts ofeach sample that are rich in oxides whereas other areascontain none. The decreasing plagioclase TiO2 concentra-tions show that there can have been only minimal commu-nication with the main magma reservoir. However, theremust have been at least some local millimetre- to centi-metre-scale diffusive or convective exchange within theresidual liquid, for the plagioclase overgrowths to recorddecreasing TiO2 even in oxide-free areas. Similar conclu-sions have been drawn for granitic systems from the distri-bution of cuneiform alkali feldspar pockets (Bryon et al.,1996). Further constraints on the lengthscale for suchexchange are not possible from these data, given the two-dimensional nature of the sections.

Spatial distribution of residual liquidduring crystallizationAlthough the fractional crystallization calculationsdescribed above are dependent on the choice of partitioncoefficient and liquidus calibration, the results can beused to visualize the evolving spatial distribution of resid-ual liquid in the crystallizing cumulates. The key observa-tion is that, other than the slight reverse zoning observedbetween An55 and An50, intercumulus crystallizationresults in a more or less continuous decrease in theanorthite content of plagioclase. For plagioclase, the grey-scale intensity of a back-scattered SEM image correlateslinearly with anorthite content (Ginibre et al., 2002),because the intensity is related to mean atomic number ofthe sample. BSE SEM images, which contain electronmicroprobe spot analyses, were therefore calibrated foranorthite content by correlating the greyscale with knownXAn, using the public domain image processing pro-gramme ImageJ (Rasband, 1997^2008). Good linear corre-lations were produced, commonly with R240�98 [for adiscussion of the use of BSE images for studying zoningprofiles see Ginibre et al. (2002)]. A series of thresholds (cor-responding to4An60,An60^55,An55^50, etc.)wasthenappliedto each calibrated photograph, generating a set ofblack and white images showing only the regions at therequired anorthite content (e.g. An50^55). Finally, eachimagewas‘despeckled’to remove noise and increase clarity.Figure 9 gives an example set of such images. As

expected, the overgrowth rims initially form parallel tothe euhedral growth faces of the primocryst cores, withcrystallization ceasing locally when two growing marginsmeet (Fig. 9). Any slight reverse zoning between An50 andAn55 is not a problem for this method because this materialis all included within one threshold bracket. The thre-sholded images also provide information about the very

last stages of solidification. In particular, the most evolvedplagioclase compositions (5An45) commonly occur at themargins of triangular pockets and along plagioclase^plagi-oclase grain junctions (Fig. 9). These are heterogeneouslydistributed even within a single thin section, and typicallyrepresent a small or insignificant volume proportion ofeach image. The distribution of these evolved margins sug-gests that in the final stages of solidification, the residualliquid occupies pore corners and thin films along grainboundaries (Fig. 9). This is consistent with the suggestionof Morse & Nolan (1984) and observations of grain bound-ary melt films in partially crystalline cumulate nodules(Holness et al., 2007a).

Effects of compactionDuring solidification, the crystal mush porosity can bereduced by mechanical compaction (adjustment of porespaces), or by ‘chemical compaction’ (dissolution andreprecipitation), as well as by crystallization. The plagio-clase textures and reverse zoning described above provideevidence for ‘chemical compaction’. Dissolution of moreAn-rich plagioclase causes the residual liquid to becomemore Ca-rich, and therefore results in reverse zoning inareas of reprecipitation. Chemical compaction should notaffect the volume of interstitial material because the sum ofthe grain areas remains constant (Meurer & Boudreau,1998b), but would affect its spatial distribution.Mechanical compaction should not result in a modifica-

tion of the compositional trend of the residual melt unlessthere is significant interaction with more evolved liquidsthat may be squeezed out of the underlying layers. Suchinteraction would result in lower concentrations of bothCa and Ti in the melt, and hence plagioclase with lowerTiO2 and lower XAn. This is in contrast to convective cir-culation with the overlying magma reservoir (or overlyingcumulates in the case of a very thick mush), which wouldbuffer the liquid to higherTiO2 and higher XAn. The maineffect of mechanical compaction will be to reduce thevolume of interstitial material relative to a region of mushthat had not lost any residual liquid.A major difficulty in quantifying changes in porosity is

in determining when ‘closed-system’ intercumulus crystalli-zation beginsçwhere is the boundary between essentiallyadcumulus and essentially orthocumulus behaviour? At themagma^mush interface, (open-system) exchange betweenthe intercumulus liquid and the magma reservoir is possi-ble. Compositional convection is unlikely to occur at thisstage unless the density of the interstitial liquid is less thanthat of the overlying magma body (Toplis et al., 2008).However, diffusive exchange is possible, and could beenhanced by convection in the main magma body. Aftersome time, dependent in part on the crystal accumulationrate, the interstitial liquid in the mush may become iso-lated from the magma reservoir. At this point, no chemicalexchange between intercumulus liquid and external


141

(c)

(d)

(e)

(f)

MI

An65-An60

An60-An65

An55-An50

An50-An45

An45-An40

(b)

Primocryst cores

c

a

b

e

d

e

(a)

ol

ol

pl

pl

pl

bi

bi

Fig. 9. Example set of plagioclase composition maps, constructed from calibrated BSE images (see text for details). Shaded regions are compo-sitions within the XAn range stated. (a) Original BSE image with plagioclase grain boundaries traced.The traced boundaries are also includedon the composition maps (b^f), as a guide for the eye. Scale bar in (a):100 mm. (b) An65^60. Only primocryst cores have high anorthite contents.Traces of oscillatory zoning can be seen. (c) An60^55. Crystallization comprises initial overgrowth on the primocryst cores. (d) An55^50.Overgrowth starts to be concentrated along grain boundaries (pairs of arrowheads a, b and c). (e) An50^45. Crystallization is now mainly con-centrated on films at grain boundaries (pairs of arrowheads) and in triangular pores or channels (arrow). Growth has now ceased at grainboundaries a and b; at c crystallization continues to either side (arrowheads d and e). (f) An45^40. In this part of the sample, crystallizationhas almost completely finished byAn40. The last vestiges of melt are found at the grain boundary e.


142

reservoir is possible (although convective circulationwithin the mush may still occur) and closed-system (ortho-cumulus-style) crystallization can occur. Possible mechan-isms for sealing the mush could include deposition ofmodally distinct layers with a strong contrast in permeabil-ity produced by finer grain size or variations in crystalshape, or the formation of an adcumulate ‘hardground’(Morse, 1986; Petersen, 1987; Mathez et al., 1997; Holnesset al., 2007c). Alternatively, at lower temperatures, diffusionrates in the interstitial liquid may be effectively too slow toalter the compositions of phases crystallizing some dis-tance below the mush interface. This study implicitlyassumes that the transition between adcumulus- andorthocumulus-style crystallization happens early in theintercumulus crystallization. However, the match betweenthe FC calculations and the observed trend suggests thatthis assumption is reasonable for the lower parts of theLayered Series of the Skaergaard Intrusion.

CONCLUSIONSThe minor element concentrations of intercumulus plagio-clase overgrowths show strong variations that can beexplained by changes in the composition of the residualliquid during fractional crystallization of the interstitialliquid. In particular, the onset of intercumulus Fe^Tioxide crystallization causes melt Ti to decrease, resultingin falling TiO2 contents in plagioclase. Decreased Al andTi at the margins of clinopyroxene oikocrysts are also con-sistent with expected changes in melt composition.However, faster diffusing species (e.g. Mg, Fe) show littlevariation because of the prolonged cooling history. Thereis no systematic difference in plagioclase compositionbetween samples from HZ to LZb, suggesting that theyfollow a similar crystallization path. Fractional crystalliza-tion calculations indicate that the Fe^Ti oxide saturationoccurs after �30% intercumulus crystallization. Constantor slightly reverse XAn in plagioclase rims is explained bycompaction-driven resorption of unfavourably orientedplagioclase grains, during continued in situ intercumuluscrystallization.The continual decrease of XAn during solidification

means that BSE images can be used to visualize the spatialdistribution of melt during solidification.The most evolvedplagioclase compositions correspond to the last to crystal-lize, which are triangular pockets and films along grainboundaries. This approach should be useful in understand-ing the manner in which porosity is reduced during mushsolidification.

ACKNOWLEDGEMENTSThe author was supported by aJunior Research Fellowshipfrom Trinity College, University of Cambridge. ChrisHayward is thanked for his assistance with electron

microprobe analyses. The manuscript was improved byconstructive reviews from Troels Nielsen, Tony Morse,Jean Be¤ dard and an anonymous reviewer, and by editorialhandling by Marjorie Wilson. Marian Holness and JohnMaclennan are also thanked for helpful discussions andcomments on an earlier version of the manuscript.

SUPPLEMENTARY DATASupplementary data for this paper are available at Journalof Petrology online.

REFERENCESAnovitz, L. M. (1991). Al zoning in pyroxene and plagioclase:Window

on late prograde to early retrograde P^T paths in granulite ter-ranes. American Mineralogist 76, 1328^1343.

Ariskin, A. A. (2003). The compositional evolution of differentiatedliquids from the Skaergaard Layered Series as determined by geo-chemical thermometry. RussianJournal of Earth Sciences 5, 1^29.

Barnes, S. J. (1986). The effect of trapped liquid crystallization oncumulus mineral compositions in layered intrusions. Contributionsto Mineralogy and Petrology 93, 524^531.

Be¤ dard, J. H. (2005). Partitioning coefficients between olivine andsilicate melts. Lithos 83, 394^419.

Be¤ dard, J. H. (2006). Trace element partitioning in plagioclasefeldspar. Geochimica et Cosmochimica Acta 70, 3717^3742.

Be¤ dard, J. H., Kerr, R. C. & Hallworth, M. A. (1992). Porous sidewalland sloping floor crystallization experiments using a reactive mush:Implications for the self-channelization of residual melts incumulates. Earth and Planetary Science Letters 111, 319^329.

Bezrukov, A., Stoyan, D. & Bargiel, M. (2001). Spatial statistics forsimulated packing of spheres. Image Analysis and Stereology 20,203^206.

Bindeman, I. N., Davis, A. M. & Drake, M. H. (1998). Ion microprobestudy of plagioclase-basalt partition experiments at natural con-centration levels of trace elements. Geochimica et Cosmochimica Acta

62, 1175^1193.Blumenfeld, R., Edwards, S. F. & Ball, R. C. (2005). Granular matter

and the marginal rigidity state. Journal of Physics: Condensed Matter

17, S2481^S2487.Brooks, C. K. & Nielsen, T. F. D. (1990). A discussion of Hunter &

Sparks. Contributions to Mineralogy and Petrology 104, 244^247.Brown, G. M. & Vincent, E. A. (1963). Pyroxenes from the late stages

of fractionation of the Skaergaard Intrusion, East Greenland.Journal of Petrology 4, 175^197.

Brown, G. M., Vincent, E. A. & Brown, P. E. (1957). Pyroxenes fromthe early and middle stages of fractionation of the SkaergaardIntrusion, East Greenland. Mineralogical Magazine 31, 511^543.

Bryon, D. N., Atherton, M. P., Cheadle, M. J. & Hunter, R. H. (1996).Melt movement and the occlusion of porosity in crystallizinggranitic systems. Mineralogical Magazine 60, 163^171.

Carr, J. M. (1954). Zoned plagioclases in layered gabbros of theSkaergaard intrusion, east Greenland. Mineralogical Magazine 30,367^375.

Cherniak, D. J. & Watson, E. B. (1994). A study of strontium diffusionin plagioclase using Rutherford backscattering spectroscopy.Geochimica et Cosmochimica Acta 58, 5179^5190.

Claeson, D.T., Meurer,W. P., Hogmalm, K. J. & Larson, S.-—. (2007).Using LA-ICPMS mapping and sector zonation to understandgrowth and trace elements partitioning in sector-zoned


143

clinopyroxene oikocrysts from the Norra Ulvo« gabbro, Sweden.Journal of Petrology 48, 711^728.

Costa, F., Chakraborty, S. & Dohmen, R. (2003). Diffusion couplingbetween trace and major elements and a model for calculation ofmagma residence times using plagioclase. Geochimica

et Cosmochimica Acta 67, 2189^2200.Dimanov, A. & Sautter, V. (2000). ‘Average’ interdiffusion of

(Fe,Mn)^Mg in natural diopside. EuropeanJournal of Mineralogy 12,749^760.

Emslie, R. F. (1980). Geology and petrology of the Harp LakeComplex, central Labrador: an example of Elsonian magmatism.Geological Survey of Canada Bulletin 293, 136 pp.

Finney, J. L. (1970). Random packings and the structure of simpleliquids 1. The geometry of random close packing. Proceedings of theRoyal Society of London, Series A 319, 479^493.

Frost, B. R. & Lindsley, D. H. (1992). Equilibria among Fe^Ti oxides,pyroxenes, olivine and quartz: Part II. Application. American

Mineralogist 73, 1004^1020.Gettings, M. E. (1976). Some thermal models of the Skaergaard

Intrusion. PhD thesis, University of Oregon, Eugene.Giletti, B. J. & Casserly, J. E. D. (1994). Strontium diffusion kinetics in

plagioclase feldspars. Geochimica et Cosmochimica Acta 58, 3785^3797.Ginibre, C., Kronz, A. & Woerner, G. (2002). High-resolution quanti-

tative imaging of plagioclase composition using accumulated back-scattered electron images: new constraints on oscillatory zoning.Contributions to Mineralogy and Petrology 142, 436^448.

Grant, N. K. & Chalokwu, C. I. (1998). The cumulate paradigmaffirmed. Journal of Geology 106, 641^644.

Grove, T. L., Baker, M. B. & Kinzler, R. J. (1984). Coupled CaAl^NaSi diffusion in plagioclase feldspar: Experiments and applica-tions to cooling rate speedometry. Geochimica et Cosmochimica Acta

48, 2113^2121.Hangh�j, K., Rosing, M. & Brooks, C. K. (1995). Evolution of the

Skaergaard magma: Evidence from crystallized melt inclusions.Contributions to Mineralogy and Petrology 120, 265^269.

Hart, S. R. & Dunn, T. (1993). Experimental cpx/melt partitioning of24 trace elements. Contributions to Mineralogy and Petrology 113, 1^8.

Haselton, J. D. & Nash, W. P. (1975). Ilmenite^orthopyroxene inter-growths from the Moon and the Skaergaard Intrusion. Earth and

Planetary Science Letters 26, 287^291.Henderson, P. (1970). The significance of the mesostasis of basic

layered igneous rocks. Journal of Petrology 11, 463^473.Hofmann, A. W. & Magaritz, M. (1977). Diffusion of Ca, Sr, Ba and

Co in a basaltic meltçimplications for geochemistry of mantle.Journal of Geophysical Research 82, 5432^5440.

Holness, M. B. & Bunbury, J. M. (2006). Insights into continental rift-related magma chambers: Cognate nodules from the KulaVolcanicProvince, Western Turkey. Journal of Volcanology and Geothermal

Research 153, 241^261.Holness, M. B., Anderson, A. T., Martin, V. M., Maclennan, J.,

Passmore, E. & Schwindinger, K. (2007a).Textures in partially soli-dified crystalline nodules: a window into the pore structure ofslowly cooled mafic intrusions. Journal of Petrology 48, 1243^1264.

Holness, M. B., Nielsen,T. F. D. & Tegner, C. (2007b).Textural matur-ity of cumulates: A record of chamber filling, liquidus assemblage,cooling rate and large-scale convection in mafic layered intrusions.Journal of Petrology 48, 141^157.

Holness, M. B.,Tegner, C., Nielsen,T. F. D., Stripp, G. & Morse, S. A.(2007c). A textural record of solidification and cooling in theSkaergaard Intrusion, East Greenland. Journal of Petrology 48,2359^2377.

Hoover, J. D. (1989). The chilled marginal gabbro and other contactrocks of the Skaergaard intrusion. Journal of Petrology 30, 441^476.

Hunter, R. H. & Sparks, R. S. J. (1987). The differentiation of theSkaergaard Intrusion. Contributions to Mineralogy and Petrology 94,451^461.

Irvine, T. N. (1980). Magmatic infiltration metasomatism, double dif-fusive fractional crystallisation and adcumulus growth in theMuskox Intrusion and other layered intrusions. In:Hargraves, R. B. (ed.) Physics of Magmatic Processes. Princetion, NJ:University Press, pp. 325^383.

Irvine, T. N., Andersen, J. C. �. & Brooks, C. K. (1998). Includedblocks (and blocks within blocks) in the Skaergaard intrusion:Geologic relations and the origin of rhythmic modally gradedlayers. Geological Society of America Bulletin 110, 1398^1447.

Jackson, E. D. (1961). Primary textures and mineral associations in theUltramafic Zone of the Stillwater Complex, Montana. USGeologicalSurvey, Professional Papers 358, 1^106.

Jambon, A. (1982). Tracer diffusion in granitic meltsçexperimentalresults for Na, K, Rb, Cs, Ca, Ba, Ce, Eu to 13008C and a modelof calculation. Journal of Geophysical Research 87, 797^810.

Jang,Y. D. & Naslund, H. R. (2001). Major and trace element compo-sition of Skaergaard plagioclase; geochemical evidence for changesin magma dynamics during the final stage of crystallization of theSkaergaard intrusion. Contributions to Mineralogy and Petrology 140,441^457.

Jerram, D. A., Cheadle, M. J., Hunter, R. H. & Elliott, M. T. (1996).The spatial distribution of grains and crystals in rocks. Contributionsto Mineralogy and Petrology 125, 60^74.

Kerr, R. C. & Tait, S. R. (1986). Crystallization and compositionalconvection in a porous medium with application to layered igneousintrusions. Journal of Geophysical Research 91, 3591^3608.

Kushiro, I. (1960). Si^Al relations in clinopyroxenes from igneousrocks. AmericanJournal of Science 258, 548^554.

LaTourrette, T. & Wasserburg, G. J. (1998). Mg diffusion in anorthite:implications for the formation of early solar system planetesimals.Earth and Planetary Science Letters 158, 91^108.

LeBas, M. J. (1962). The role of aluminium in igneous clinopyroxeneswith relation to their parentage. American Journal of Science 260,267^288.

Longhi, J.,Walker, D. & Hays, J. F. (1976). Fe and Mg in plagioclase.Proceedings of the 7th Lunar Science Conference, Vol. 2. Geochimica et

Cosmochimica Acta Supplement 1281^1300.Loucks, R. R. (1990). Discrimination of ophiolitic from nonophiolitic

ultramafic^mafic allochthons in orogenic belts by the Al/Ti ratio inclinopyroxene. Geology 18, 346^349.

Lundgaard, K. L. & Tegner, C. (2004). Partitioning of ferric and fer-rous iron between plagioclase and silicate melt. Contributions to

Mineralogy and Petrology 147, 470^483.Maal�e, S. (1976). The zoned plagioclase of the Skaergaard Intrusion,

East Greenland. Journal of Petrology 17, 398^419.Mathez, E. A., Hunter, R. H. & Kinzler, R. (1997). Petrologic evolu-

tion of partially molten cumulate: the Atok section of the BushveldComplex. Contributions to Mineralogy and Petrology 129, 20^34.

McBirney, A. R. (1989). The Skaergaard Layered Series: I. Structureand average compositions. Journal of Petrology 30, 363^397.

McBirney, A. R. & Hunter, R. H. (1995). The cumulate paradigmreconsidered. Journal of Geology 103, 114^122.

McBirney, A. R. & Naslund, H. R. (1990). The differentiation of theSkaergaard intrusion. Contributions to Mineralogy and Petrology 104,235^240.

McBirney, A. R. & Noyes, R. M. (1979). Crystallization and layeringof the Skaergaard Intrusion. Journal of Petrology 20, 487^554.

Meurer, W. P. & Boudreau, A. E. (1998a). Compaction of igneouscumulates, Part I: Geochemical consequences for cumulates andliquid fractionation trends. Journal of Geology 106, 281^292.


144

Meurer, W. P. & Boudreau, A. E. (1998b). Compaction of igneouscumulates, Part II: Compaction and the development of igneousfoliations. Journal of Geology 106, 293^304.

Meurer, W. P. & Meurer, M. E. S. (2006). Using apatite to dispel the‘trapped liquid’ concept and to understand the loss of interstitialliquid by compaction in mafic cumulates: an example from theStillwater Complex, Montana. Contributions to Mineralogy and

Petrology 151, 187^201.Morse, S. A. (1986). Convection in aid of adcumulus growth. Journal of

Petrology 27, 1183^1214.Morse, S. A. (1988). Motion of crystals, solute, and heat in layered

intrusions. Canadian Mineralogist 26, 209^224.Morse, S. A. (1990).The differentiation of the Skaergaard Intrusion. A

discussion of Hunter & Sparks. Contributions to Mineralogy and

Petrology 104, 240^244.Morse, S. A. (1998). Is the cumulate paradigm at risk? An extended

discussion of the cumulate paradigm reconsidered. Journal of

Geology 106, 367^370.Morse, S. A. & Nolan, K. M. (1984). Origin of strongly reversed rims

on plagioclase in cumulates. Earth and Planetary Science Letters 68,485^498.

Naslund, H. R. (1984). Petrology of the Upper Border Series of theSkaergaard Intrusion. Journal of Petrology 25, 185^212.

Nicolas, A. & Ildefonse, B. (1996). Flow mechanism and viscosity inbasaltic magma chambers. Geophysical Research Letters 23, 2013^2016.

Nielsen, T. F. D. (1987). Tertiary alkaline magmatism in EastGreenland: a review. In: Fitton, J. G. & Upton, B. G. J. (eds)Alkaline Igneous Rocks. Geological Society, London, Special Publications

30, 489^515.Nielsen, T. F. D. (2004). The shape and volume of the Skaergaard

Intrusion, Greenland: Implications for mass balance and bulk com-position. Journal of Petrology 45, 507^530.

Nwe, Y. Y. (1976). Electron-probe studies of the earlier pyroxenes andolivines from the Skaergaard Intrusion, East Greenland.Contributions to Mineralogy and Petrology 55, 105^126.

Petersen, J. S. (1987). Solidification contraction: Another approach tocumulus processes and the origin of igneous layering. In: Parsons, I(ed.) Origins of Igneous Layering. NATOASI Series C: Mathematical and

Physical Sciences 196, 185^245.Philpotts, A. R., Brustman, C. M., Shi, J., Carlson, W. D. &

Denison, C. (1999). Plagioclase-chain networks in slowly cooledbasaltic magma. American Mineralogist 84, 1819^1829.

Phinney, W. C. (1992). Partition coefficients for iron between plagio-clase and basalt as a function of oxygen fugacity: Implications forArchaean and lunar anorthosites. Geochimica et Cosmochimica Acta 56,1885^1895.

Rasband, W. S. (1997^2008). ImageJ. Bethesda, MD: US NationalInstitutes of Health. http://rsb.info.nih.gov/ij/.

Rudge, J. F., Holness, M. B. & Smith, G. C. (2008). Quantitative tex-tural analysis of packings of elongate crystals. Contributions to

Mineralogy and Petrology 156, 413^429.Saar, M. O., Manga, M., Cashman, K. V. & Fremouw, S. (2001).

Numerical models of the onset of yield strength in crystal^meltsuspensions. Earth and Planetary Science Letters 187, 367^379.

Sautter, V., Jaoul, O. & Abel, F. (1988). Aluminum diffusion indiopside using the 21Al(p, �) 28Si nuclear reaction:preliminary results. Earth and Planetary Science Letters 89, 109^114.

Shimizu, N. (1978). Analysis of the zoned plagioclase of different mag-matic environments: A preliminary ion-microprobe study. Earth

and Planetary Science Letters 39, 398^406.Smith, J.V. (1983). Some chemical properties of feldspars. Mineralogical

Society of America, Reviews in Mineralogy 2, 281^296.

Smith, J. V. & Brown,W. L. (1988). Feldspar Minerals: Crystal Structures,

Physical, Chemical, and Microtextural Properties. NewYork: Springer.Sparks, R. S. J., Huppert, H. E., Turner, J. S., Sakuyama, M. &

O’Hara, M. J. (1984).The fluid dynamics of evolving magma cham-bers. Philosophical Transactions of the Royal Society of London 310(1514),511^534.

Steele, I. M., Hutcheon, D. & Smith, J. V. (1980). Ion microprobeanalysis and petrogenetic interpretations of Li, Mg, Ti, K, Sr, Bain lunar plagioclase. Proceedings of the 11th Lunar and Planetary ScienceConference. Geochimica et Cosmochimica Acta Supplement 571^590.

Stripp, G. R. & Holness, M. (2006). Enigmatic late-stage textures inmafic cumulates: Skaergaard Intrusion, East Greenland.Transactions of the American Geophysical Union 87(Supplement),V51B^1672.

Tait, S. R. (1988). Samples from the crystallising boundary layer of azoned magma chamber. Contributions to Mineralogy and Petrology 100,470^483.

Tait, S. R. & Jaupart, C. (1992). Compositional convection in a reac-tive crystalline mush and melt differentiation. Journal of GeophysicalResearch 97, 6735^6756.

Tegner, C. (1997). Iron in plagioclase as a monitor of the differentia-tion of the Skaergaard intrusion. Contributions to Mineralogy and

Petrology 128, 45^51.Thy, P., Lesher, C. E., Nielsen, T. F. D. & Brooks, C. K. (2006).

Experimental constraints on the Skaergaard liquid line of descent.Lithos 92, 154^180.

Toplis, M. J. & Carroll, M. R. (1995). An experimental study on theinfluence of oxygen fugacity on Fe^Ti oxide stability, phase rela-tions, and mineral^melt equilibria in ferro-basaltic systems.Journal of Petrology 36, 1137^1170.

Toplis, M. J. & Carroll, M. R. (1996). Differentiation of ferro-basalticmagmas under conditions open and closed to oxygen: Implicationsfor the Skaergaard Intrusion and other natural systems. Journal ofPetrology 37, 837^858.

Toplis, M. J., Brown, W. L. & Pupier, E. (2008). Plagioclase in theSkaergaard intrusion. Part 1: Core and rim compositions in thelayered series. Contributions to Mineralogy and Petrology 155, 329^340.

Van Orman, J. A., Grove, T. L. & Shimizu, N. (2001). Rare earthelement diffusion in diopside: influence of temperature,pressure, and ionic radius, and an elastic model for diffusion insilicates. Contributions to Mineralogy and Petrology 141, 687^703.

Wager, L. R. & Brown, G. M. (1968). Layered Igneous Rocks. London:Oliver & Boyd, 588 pp.

Wager, L. R. & Deer, W. A. (1939). Geological investigations in eastGreenland, Part III. Petrology of the Skaergaard intrusion,Kangerdlugssuaq, east Greenland. Meddelelser om Gr�nland 105,1^352.

Wager, L. R., Brown, G. M. & Wadsworth, W. J. (1960). Types ofigneous cumulates. Journal of Petrology 1, 73^85.

Wilke, M. & Behrens, H. (1999).The dependence of the partitioning ofiron and europium between plagioclase and hydrous tonalitic melton oxygen fugacity. Contributions to Mineralogy and Petrology 137,102^114.

Williams, S. R. & Philipse, A. P. (2003). Random packings of spheresand spherocylinders simulated by mechanical contraction. PhysicalReview, E 67, 051301.

Zellmer, G. F., Blake, S., Vance, D., Hawkesworth, C. & Turner, S.(1999). Plagioclase residence times at two island arc volcanoes(Kameni Islands, Santorini, and Soufrie' re, StVincent) determinedby Sr diffusion systematics. Contributions to Mineralogy and Petrology

136, 345^357.


145

http://rsb.info.nih.gov/ij/

chemical evolution of intercumulus liquid, as recorded in...

Documents