Humphreys. 2011. Silicate Liquid Inmmiscibility Within the Crystal Mush ... Etc

Embed Size (px)

Citation preview

  • 7/27/2019 Humphreys. 2011. Silicate Liquid Inmmiscibility Within the Crystal Mush ... Etc

    1/28

    Silicate Liquid Immiscibility within theCrystal Mush: Evidence fromTi in Plagioclasefrom the Skaergaard Intrusion

    MADELEINE C. S. HUMPHREYS*

    DEPARTMENT OF EARTH SCIENCES, UNIVERSITY OF CAMBRIDGE, DOWNING STREET, CAMBRIDGE CB2 3EQ, UK

    RECEIVED JANUARY 8, 2010; ACCEPTED NOVEMBER 2, 2010

    A key target in the study of layered intrusions is to constrain theliquid line of descent of the magma. However, the evolution of the

    interstitial liquid is rarely considered, and its liquid line of descent

    is often assumed to be equivalent to that of the bulk magma.

    Because of extensive sub-solidus and diffusional changes that

    occur in slowly cooled rocks, clues to the composition of the intersti-

    tial liquid can only be obtained using very slowly diffusing trace

    elements and components. This study uses the Ti concentrations

    and anorthite contents of interstitial plagioclase to consider the

    compositional evolution of the interstitial liquid in the Skaergaard

    Intrusion. Ti^XAn zoning of interstitial plagioclase does not

    follow the same cryptic variations that develop in plagioclase pri-

    mocrysts as a function of stratigraphic height, demonstrating that

    the bulk and interstitial liquid lines of descent are not equivalent.After Fe^Ti oxides start to crystallize, Ti concentrations decrease

    in both primocryst and interstitial plagioclase as a result of decreas-

    ing melt Ti. However, in the interstitial plagioclase within a

    single thin section, divergent trends develop adjacent to fine-grained

    interstitial pockets containing diverse mineral assemblages, which

    are interpreted to represent the crystallized products of late-stage

    immiscible liquids. These trends vary systematically as a function

    of stratigraphic height and spatial location within the intrusion.

    The distribution and compositions of these plagioclase zoning

    trends are used to comment on the spatial distribution and differen-

    tial movement of interstitial immiscible liquids within the

    intrusion.

    KEY WORDS: layered igneous rock; immiscibility; plagioclase; zoning;

    cumulate; Skaergaard

    I N T R O D U C T I O NLayered intrusions are of interest for many reasons: for

    example, they reveal information about differentiation

    and fractionation processes that may be operating beneath

    currently active volcanoes, they provide insights into the

    effects of reactive fluids moving within a porous medium

    and they are commonly host to economic precious metal

    deposits. The cumulate rocks that define layered intrusions

    can accumulate by various mechanisms, including the set-

    tling of primocrysts (early formed crystals) onto the crys-

    tal pile, in situ growth of crystals in the cooling boundary

    layers, or deposition from crystal-laden plumes (e.g. Wager

    & Brown, 1968; McBirney & Noyes, 1979; Irvine et al.,

    1998; Marsh, 2006). In most instances, the result is a layerof cumulus crystals surrounded by interstitial liquid.

    Once this crystal mush has formed, the interstitial liquid

    is slowly eliminated through various processes that may in-

    clude overgrowth on primocrysts, nucleation and growth

    of oikocrysts and other interstitial material (including

    new minerals), consolidation and compaction of the mush

    owing to the weight of overlying crystals and composition-

    al convection, which can result in adcumulus textures.

    Although post-depositional processes acting within

    the mush have been discussed in many previous studies

    (e.g. Tait et al ., 1984; Barnes, 1986; Boudreau &

    McCallum, 1992; Haskin & Salpas, 1992; Meurer &

    Boudreau, 1998; Tegner et al., 2009), relatively little work

    has addressed the differentiation of the interstitial liquid

    itself, compared with that of the bulk magma. Studies spe-

    cifically considering the development of intercumulus

    *Corresponding author. Present Address: Department of Earth

    Sciences, University of Oxford, South Parks Road, Oxford, OX13AN.

    Telephone:44 (0)1865 272020. Fax:44 (0)1865 272072.

    E-mail: [email protected]

    The Author 2010. Published by Oxford University Press. All

    rights reserved. For Permissions, please e-mail: journals.permissions@

    oup.com

    JOURNALOF PETROLOGY VOLUME 52 NUMBER1 PAGES147^174 2011 doi:10.1093/petrology/egq076

  • 7/27/2019 Humphreys. 2011. Silicate Liquid Inmmiscibility Within the Crystal Mush ... Etc

    2/28

  • 7/27/2019 Humphreys. 2011. Silicate Liquid Inmmiscibility Within the Crystal Mush ... Etc

    3/28

    equivalent to the LZ, MZ and UZ of the LS, and UBS-t,

    which refers to the outermost part of the UBS and is essen-

    tially equivalent to the HZ (Naslund, 1984; Douglas, 1961).

    In addition to the changes in primocryst assemblage, the

    compositions of the primocrysts change with stratigraphic

    height. For example, plagioclase compositions vary from

    $An70 in HZ to An25 at the Sandwich Horizon (e.g.

    Wager & Brown, 1968; Maale, 1976). This cryptic vari-

    ation is generally accepted to reflect closed-system, frac-

    tional crystallization of essentially a single batch of

    magma. Whereas the LS has been heavily studied since

    early descriptions (Wager & Brown, 1968), the MBS andUBS have been relatively neglected apart from the studies

    of Naslund (1984) and Hoover (1989). The comprehensive

    early petrological study of the MBS (Hoover, 1989)

    showed t hat primocryst olivine compositions are up to

    9 mol % more Fa-rich, clinopyroxene is 5 mol % richer in

    the diopside component, and plagioclase is 5^8 mol %

    more calcic, than primocrysts from equivalent stratigraph-

    ic levels in the LS (Hoover, 1989). Minor element contents

    of plagioclase were not analysed. These observations were

    ascribed to re-equilibration of primocrysts with Fe-rich

    interstitial liquid flowing down the margins of the intru-

    sion (Hoover, 1989). In addition, mineral compositions

    from the upper parts of the MBS (i.e. adjacent to the

    Upper Border Series) were also observed to be more

    Fe-rich, and plagioclase more Na-rich, than elsewhere in

    the intrusion. This was interpreted as the result of either

    more effective fractionation near the roof or modified

    magma composition in the upper parts of the chamber

    (Hoover,1989).

    For Skaergaard, there is still disagreement over changesin oxygen fugacity (e.g. Toplis & Carroll, 1995, 1996; Thy

    et al., 2009), and whether the residual liquid follows the

    Fenner or Bowen compositional trends; that is, whether it

    becomes Fe-rich at low SiO2 or tends towards SiO2-rich

    compositions at relatively lower FeO contents (e.g.

    McBirney, 1975; Hunter & Sparks, 1987; Brooks & Nielsen,

    1990; McBirney & Naslund, 1990; Toplis & Carroll, 1995,

    1996; Thy et al., 2006, 2009). It seems plausible that both

    Fig. 1. Geological map of the Skaergaard Intrusion, after McBirney (1989). The locations of the 1966 Cambridge dri ll core and the Platinovadrill cores 90-22 and 90-10 are shown. Continuous lines (KR02 and SP) indicate the locations of traverses through the Marginal BorderSeries. The star indicates the location of LZc sample 458279 (f romTegner et al., 2009).

    HUMPHREYS INTERSTITIAL LIQUID IMMISCIBILITY

    149

  • 7/27/2019 Humphreys. 2011. Silicate Liquid Inmmiscibility Within the Crystal Mush ... Etc

    4/28

    the Bowen and Fenner trends are represented in the com-

    positions of two immiscible liquids that have been shown

    to coexist by at least UZb (McBirney & Nakamura, 1974;

    Hanghj et al., 1995; Jakobsen et al., 2005). The postulated

    earlier occurrence of immiscibility in the residual liquid

    (LZc, Veksler et al., 2007) is disputed (e.g. Morse, 2008a;

    McBirney, 2008; Philpotts, 2008; Veklser et al ., 2008).Furthermore, the effect, if any, of liquid immiscibility on

    the resulting fractionation trend is unknown, as is the

    stage at which immiscibility might occur within the inter-

    stitial liquid.

    Textural indicators of evolved interstitialliquidRecent work has demonstrated that evolved, late-stage

    interstitial liquid can be present in glass-bearing nodules

    entrained in lava flows as thick films of glass on grain

    boundaries. In cumulate rocks, pyroxene commonly forms

    thick grain boundary films and it has been suggested

    that these represent an original liquid film (e.g. Holness,

    2005; Holness et al., 2007a). At Skaergaard, crystallized

    traces of evolved liquids can also be seen in trails of tiny

    oxide and apatite crystals, clustered particularly along the

    margins of plagioclase chadacrysts within clinopyroxene

    oikocrysts (Fig. 2a). Holness et al . (2011) demonstrate

    that, following immiscibility in the bulk magma, the so-

    lidification of large pockets of interstitial liquid results in

    the formation of paired (spatially associated) intergrowths

    of granophyre and an ilmenite-rich assemblage including

    apatite, skeletal Fe^Ti oxides and clinopyroxene (see

    Holness et al., 2011, figs 10^12). Granophyre-rich pockets

    tend to be associated with the plagioclase-rich regions of a

    sample whereas ilmenite-rich pockets tend to be associated

    with regions rich in olivine and clinopyroxene (Holnesset al., 2011). The paired intergrowths commonly occupy

    planar-sided pockets, which shows that the liquid was not

    reactive, and the variations in their spatial distribution

    are clearest in the MBS. In the Skaergaard Peninsula,

    which represents the most complete chemical sequence,

    they become abundant near the MZ*^UZa* boundary.

    At the same point, evidence of localized reaction between

    oxides and pyroxene disappears (Holness et al., 2011). In

    combination the paired intergrowths represent up to

    44% of the sample by area in the Layered Series, and

    more in the evolved parts of the MBS (Stripp, 2009).

    Samples from the less evolved, outermost parts of the

    Skaergaard Peninsula traverse contain small, isolated,fine-grained, planar-sided pockets of late-stage interstitial

    phases (Fig. 2b and c). These are distinct from the paired

    intergrowths of Holness et al. (2011) in that they are smaller,

    commonly very fine-grained and highly altered; however,

    these fine-grained pockets probably also represent the crys-

    tallized products of late-stage liquids. Where the contents

    of the pockets can be identified, they typically include

    some of quartz, K-feldspar, apatite, non-skeletal Fe^Ti

    oxides, pyroxenes, biotite and zircon, as well as plagioclase

    growing on the margins of the pockets. Their mineralogy

    is highly variable (assuming that the 2D distribution is

    representative of the 3D volume), and the pockets repre-

    sent between 027 and 106% of the samples by area in the

    Skaergaard Peninsula (Table 1). At lower structural levels

    in the MBS, near the base of the intrusion (e.g. northernKraemer Island), the interstitial pockets are more sparsely

    developed, representing at most 020 area % of the sam-

    ples (Table 1). Abundant interstitial low-Ca pyroxene is

    present, commonly surrounding resorbed olivine or adja-

    cent to Fe^Ti oxides (Fig. 2d), and olivine may be partially

    replaced by oxy-symplectites of magnetite and orthopyrox-

    ene (see Holness et al., 2011).

    In the LS, the interstitial pockets are typically present at

    020^035 area % (Table 1), but in the main replacive

    grain-boundary microstructures are seen instead in the

    central parts of the LS (LZc^MZ, Holness et al., 2011; see

    below and Table 1). At the base of the LS low-Ca pyroxene

    is also present, and the interstitial pockets are lessabundant. The coarse-grained, paired granophyric and

    ilmenite-rich intergrowths become abundant around the

    MZ^UZa boundary near the margins of the LS and near

    the UZa^UZb boundary in the centre of the intrusion

    (Holness et al., 2011). Between LZc and the point at which

    the coarse-grained, paired intergrowths dominate, there is

    abundant evidence of reactive melt occupying grain

    boundaries, manifest as irregular (stepped) clinopyrox-

    ene^plagioclase grain boundaries, fish-hook intergrowths

    of pyroxene and plagioclase, olivine rims on oxide grains,

    and mafic symplectites; grains of apatite, amphibole

    and oxides are commonly found within the microstruc-

    tures. This spatial distribution, as well as the microstruc-

    tures involved, is described in detail by Holness et al.

    (2011), who interpret the microstructures as the result

    of chemical reaction following physical separation of

    Fe-rich and Si-rich immiscible liquids within the floor

    cumulates.

    S A M P L E S A N D M E T H O D S

    Samples studiedThis study investigates a suite of samples from the LS and

    MBS of the Skaergaard Intrusion (Table 1). The LS was

    sampled mainly using drill core material, with samples se-

    lected to cover the range of stratigraphy. The lower partsof the LS (HZ^LZb) are covered by the 1966 Cambridge

    Drill Core (Maale, 1976; Holness et al., 2007b), and the

    upper parts (UZa^UZb) by drill core 90-22 (collected by

    Platinova Resources Ltd, 1990; Tegner, 1997), with one

    LZc sample (458279) from a surface reference profile

    (Nielsen et al., 2000; Tegner et al., 2009). Drill core 90-22

    samples the centre of the intrusion (Fig. 1). Four samples

    from MZ^UZa were also examined from drill core 90-10,

    JOURNAL OF PETROLOGY VOLUME 52 NUMBER1 JANUARY 2011

    150

  • 7/27/2019 Humphreys. 2011. Silicate Liquid Inmmiscibility Within the Crystal Mush ... Etc

    5/28

    which is located at the western margin of the intrusion

    (Fig. 1).Plagioclase compositions in the MBS were analysed

    in selected samples from two surface traverses, one on

    Kraemer Island (sample numbers beginning KR02),

    which includes zones HZ*^LZb*, and one on the

    Skaergaard Peninsula (sample numbers beginning SP),

    which samples the base of LZa* to UZb* (the contact is

    unexposed, so it is unclear how much of HZ*^LZa* is

    missing, but the fine-grained sample SP60 is assumed to

    represent the base of LZa*). The figures presented here

    also include plagioclase data for MZ (drill core 90-22)collected by Stripp (2009) as well as data from the

    Cambridge drill core from Humphreys (2009).

    Analytical methodsPlagioclase analyses were obtained using a Cameca SX-100

    electron microprobe at the University of Cambridge. A

    2 mm, 15 kV beam was used, with 10 nA current for major

    elements and 100 nA for minor elements. Typical peak

    pl

    pl

    pl

    pl

    qz

    K-fsp

    200 m

    (b)

    (c)

    pl

    100 m

    pl

    (a)

    0.25 mm

    ox

    ox

    ap

    px

    ap

    ap

    apap

    (d)

    pl

    pl

    pl

    ol

    ol

    ol

    opx opx

    0.25 mm

    Fig. 2. Textural characteristics of interstitial material from the Skaergaard cumulates. (a) Cross-polarized light photomicrograph showingstrings of small, grain-boundary oxides (arrowed) and apatite crystals (labelled) along the margins of plagioclase within clinopyroxeneoikocrysts, marking the location of late interstitial liquids (from HZ*, north Kraemer Island). (b) Back-scattered SEM image of fine-grainedinterstitial pocket containing mafic minerals including apatite, Fe^Ti oxides and pyroxenes, from sample SK46 (LZa*^LZb* boundary) inthe Skaergaard Peninsula. (c) Back-scattered SEM image of fine-grained interstitial pocket containing felsic minerals including quartz andalkali feldspar, as well as unidentifiable, brightly coloured weathered material, from sample SK46 (LZa*^LZb* boundary) in the SkaergaardPeninsula. (d) Cross-polarized photomicrograph showing orthopyroxene rims around olivine crystals, characteristic of the lower parts of theMBS (Kraemer Island) and the Layered Series.

    HUMPHREYS INTERSTITIAL LIQUID IMMISCIBILITY

    151

  • 7/27/2019 Humphreys. 2011. Silicate Liquid Inmmiscibility Within the Crystal Mush ... Etc

    6/28

    Tab

    le1:Locat

    ionsand

    detailsofsamplesstud

    ied

    Sample

    Strat.

    height(m)

    Series

    Zone

    Primocryst

    mineralogy

    Evolved

    pockets?

    Area%

    Pockets

    Replacive

    sympl?

    Non-

    replacive

    sympl?

    Location

    Latitude

    (N)

    Longitude

    (W)

    90-22-421

    1671

    LayeredSeries

    UZb

    plolcpx

    ox

    ap

    Y

    Drillcore90-22

    90-22-461.8

    1630

    LayeredSeries

    UZb

    plolcpx

    ox

    ap

    Y

    026

    T1

    Drillcore90-22

    90-22-481.8

    1610

    LayeredSeries

    UZa

    plolcpx

    ox

    T1

    Y

    Drillcore90-22

    90-22-824.2

    1268

    LayeredSeries

    UZa

    plolcpx

    ox

    (Y)

    0006

    T1

    Drillcore90-22

    458279

    752

    LayeredSeries

    LZc

    plolcpx

    ox

    Y

    0032

    T1

    KraemerIsland

    68811265"

    31842563"

    90-10-212

    LayeredSeries

    UZa

    plolcpx

    ox

    Y

    028

    (Y?)

    Drillcore90-10

    90-10-445

    LayeredSeries

    MZ

    plcpx

    ox

    Y

    035

    (T1)

    Drillcore90-10

    90-10-456

    LayeredSeries

    MZ

    plcpx

    ox

    T1

    Drillcore90-10

    90-10-524.1

    LayeredSeries

    MZ

    plcpx

    ox

    Y

    013

    T1

    Drillcore90-10

    118590

    197

    LayeredSeries

    LZb

    plolcpx

    Y

    033

    Cambridgedrillcore

    118601

    1681

    LayeredSeries

    LZb

    plolcpx

    (Y)

    0061

    Cambridgedrillcore

    118605

    1636

    LayeredSeries

    LZb

    plolcpx

    Y

    030

    (T1)oxy

    Cambridgedrillcore

    118653

    43

    LayeredSeries

    LZa

    plol

    Y

    020

    oxy

    Cambridgedrillcore

    118678

    22

    LayeredSeries

    HZ

    plol

    Y

    011

    Cambridgedrillcore

    Sample

    Dist.from

    margin

    (m)

    Series

    Zone

    Primocryst

    mineralogy

    Evolved

    pockets?

    Area%

    pockets

    Replacive

    sympl?

    Non-

    replacive

    sympl?

    Location

    Latitude

    (N)

    Longitude

    (W)

    SP29

    594

    MBS

    UZb*

    plolcpx

    ox

    ap

    Y

    SkaergaardPeninsula

    688849

    5"

    318453

    2"

    SP20

    4718

    MBS

    MZ*

    plcpx

    ox

    Y

    027

    T1

    SkaergaardPeninsula

    688849

    5"

    31845138"

    SP16

    4273

    MBS

    LZc*

    plolcpx

    ox

    Y

    036

    T1

    SkaergaardPeninsula

    688849

    7"

    31845177"

    SP8

    3042

    MBS

    LZb*

    plolcpx

    Y

    067

    (T1)

    SkaergaardPeninsula

    688850

    2"

    31845286"

    SP46

    133

    MBS

    LZa*-LZb*

    plolcpx

    Y

    106

    SkaergaardPeninsula

    688850

    6"

    31845444"

    SP60

    10

    4

    MBS

    BaseLZa*

    plol

    Y

    047

    SkaergaardPeninsula

    688848

    9"

    31845542"

    KR02-21

    2227

    MBS

    LZb*

    plolcpx

    Y

    014

    KraemerIsland

    68811362"

    31844224"

    KR02-37

    80

    5

    MBS

    LZa*

    plol

    Y

    020

    oxy

    KraemerIsland

    68811378"

    31844340"

    KR02-4

    10

    2

    MBS

    HZ*

    plol

    (Y)

    0015

    oxy

    KraemerIsland

    68811382"

    31844401"

    Stratigraphicheightsof

    90-22andlocationof458279takenfro

    m

    thereferenceprofileofTegneretal.(2009).Y,yesthismicrostructureispresent(parentheses

    indicaterarely);T1,Typ

    e1replacivesymplectites;oxy,oxy-symplectites(Holnesset

    al.,2011).

    JOURNAL OF PETROLOGY VOLUME 52 NUMBER1 JANUARY 2011

    152

  • 7/27/2019 Humphreys. 2011. Silicate Liquid Inmmiscibility Within the Crystal Mush ... Etc

    7/28

    counting times were 20 s for major elements and 30^40 s

    for minor elements. Zoned grain margins, triple junctions,

    the margins of planar-sided interstitial pockets (see

    above) and traverses of optically strongly zoned grain

    edges were preferentially studied, to identify the compos-

    itional variations of interstitial plagioclase from the late

    stages of fractionation. These features were identifiedoptically or by SEM observation using a JEOL JSM-820

    instrument, with a 20 kV beam and c. 15 mm working dis-

    tance. Data from Stripp (2009) from MZ are included for

    completeness; these represent traverses across zoned

    grains but probably do not sample the most extreme

    compositions.

    R E S U L T S

    Primocryst plagioclaseThe cryptic compositional variation of plagioclase primo-

    crysts that occurs through the Layered Series stratigraphyat Skaergaard is summarized here, for later comparison

    with t he interstitial compositions. Anorthite contents of

    primocrysts decrease steadily upwards through the stratig-

    raphy, with a relatively restricted range of compositions

    observed in each zone, from $An63^69 in HZ, through

    $An48^53 at LZc, to $An31^39 at UZb (Fig. 3a). This is

    in agreement with the results of several previous studies

    (e.g. Maale, 1976; McBirney, 1989) and is consistent with

    closed-system fractionation of basaltic magma. TiO2 con-

    tents of plagioclase primocrysts increase to a maximum

    of $013 wt % in the LZb sample, then decrease almost

    linearly with increasing stratigraphic height (decreasing

    XAn: the primocryst trend, Fig. 3a), as a result of satur-ation in Fe^Ti oxides (Humphreys, 2009). Primocryst FeO

    contents decrease with decreasing XAn until approximately

    UZa levels, then start to increase again (Fig. 3b). This

    increase in FeO through the Upper Zone is consistent

    with previous observations (Tegner, 1997) and has been

    interpreted as reflecting the increasing FeO content of

    the fractionating bulk magma (Tegner, 1997; Tegner &

    Cawthorn, 2010). MgO contents are scattered but approxi-

    mately constant until $An45 and then decrease with XAn(Fig. 3c). In general, primocrysts from Layered Series

    rocks near the margins of the intrusion (i.e. drill core

    90-10) contain slightly more FeO and MgO than those

    from the centre of the intrusion (drill core 90-22).

    In the Marginal Border Series, primocryst compositions

    are very similar to those in the LS in terms of XAn,

    but tend to be enriched in MgO and FeO, similar

    to plagioclase from drill core 90-10. This is consistent

    with the observations of Hoover (1989), who interpreted

    the Fe-rich compositions as the result of flow of

    Fe-rich residual liquids down the marginal walls of the

    intrusion.

    Interstitial plagioclase: Marginal BorderSeries, Skaergaard PeninsulaThe MBS as exposed on the Skaergaard Peninsula repre-

    sents a nearly complete transect through the stratigraphy,

    comprising the base of LZa* to UZb*. As expected, inter-

    stitial plagioclase from any given stratigraphic zone typic-

    ally evolves to substantially more sodic compositionsthan the primocrysts from that zone: to $An40 in HZ*,

    through $An35 in LZc*, to $An15 in UZb* (see

    Electronic Appendix 1, available for downloading at

    http://www.petrology.oxfordjournals.org). The compos-

    ition of the most evolved plagioclase analysed from any

    sample is variable. Minor element concentrations show

    complex but systematic variations. In general, many inter-

    stitial plagioclase compositions plot at lowerTiO2 contents

    for a given XAn than the primocrysts, and this will be

    described in detail below. However, one of the most inter-

    esting features of the interstitial compositions is that in

    the lower stratigraphic zones, the interstitial plagioclase

    trend itself splits into two divergent compositional trends,typically adjacent to the planar-sided interstitial pockets

    described above. Strong normal zoning can be observed

    adjacent to pockets containing abundant granophyre or

    quartz, commonly with a clear concave-upward trend

    of TiO2^XAn. In contrast, strong reverse zoning with

    decreasing TiO2 is observed in the vicinity of clearly

    calcic, Fe-rich pockets (typically containing Fe^Ti oxides,

    apatite, clinopyroxenebiotite; Fig. 4). Single spot ana-

    lyses also fall on these trends. There is a wide variation of

    plagioclase compositions even around a single pocket, and

    few plagioclase compositions represent the extreme ends

    of these normal and reverse trends. Many of the interstitial

    pockets are not clearly dominated by either Si-rich orFe-rich minerals, or are strongly altered. As far as can be

    observed, however, reverse or calcic compositions are not

    associated with pockets that contain quartz or granophyre,

    whereas conversely not all pockets that contain Fe-rich

    minerals (apatite, oxides, pyroxenes, etc.) show reverse

    zoning.

    In sample SP60 (base LZa*), the TiO2 and anorthite

    contents of the interstitial plagioclase are initially indistin-

    guishable from primocryst compositions from higher

    stratigraphic zones. With decreasing XAn, TiO2 concentra-

    tions rise to a maximum of c. 012 wt % at $An56, and

    then diverge steeply from the primocryst trend, with inter-

    stitial plagioclase dipping to lower TiO2 contents whereasthe primocrysts show a much shallower decrease (Fig. 5a).

    The interstitial plagioclase trend then itself follows two

    divergent compositional trends after An54 (Fig. 5a) as

    described above. Interstitial pockets represent $047% of

    the thin section by area in this sample (Table 1).

    In sample SP46 (LZa*^LZb* boundary), interstitial

    pockets account for $106% by area (Table 1). Plagioclase

    compositions follow the XAn^TiO2 primocryst trend

    HUMPHREYS INTERSTITIAL LIQUID IMMISCIBILITY

    153

    http://www.petrology.oxfordjournals.org/http://www.petrology.oxfordjournals.org/
  • 7/27/2019 Humphreys. 2011. Silicate Liquid Inmmiscibility Within the Crystal Mush ... Etc

    8/28

    0.0

    0.2

    0.4

    0.6

    0.8

    1.0

    0.00

    0.02

    0.04

    0.06

    0.08

    0.10

    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

    mol % Anorthite

    wt%

    TiO

    2

    wt%MgO

    wt%FeO

    (a)

    (b)

    (c)

    20 30 40 50 60 70 80

    HZ

    LZa

    LZb

    LZc

    MZ

    UZa

    UZb

    20 30 40 50 60 70 80

    Fig. 3. Mole per cent Anorthite vs (a) TiO2, (b) FeO and (c) MgO compositions of primocryst plagioclase from the Layered Series. Data fromHolness et al. (2011) and this study. Approximate zone boundaries are indicated.

    JOURNAL OF PETROLOGY VOLUME 52 NUMBER1 JANUARY 2011

    154

  • 7/27/2019 Humphreys. 2011. Silicate Liquid Inmmiscibility Within the Crystal Mush ... Etc

    9/28

    until $An50, when the interstitial compositions diverge

    to lower TiO2, as for SP60. The two interstitial trends

    observed in SP60 are particularly well defined for sample

    SP46. Adjacent to pockets containing abundant

    granophyre or quartz, there is a very clear, curving trend

    towards very low TiO2 at An28, whereas plagioclase adja-

    cent to pockets containing mafic phases defines a sharply

    reversed trend towards $An75 (Fig. 5b). There is also one

    100 m

    400 m

    0.00

    0.02

    0.04

    0.06

    0.08

    0.10

    0.12

    0.14

    0.16

    10 20 30 40 50 60 70 80 90

    towards Si-richpockets

    towards Fe-richpockets

    X An

    wt%T

    iO2

    0.00

    0.02

    0.04

    0.06

    0.08

    0.10

    0.12

    0.14

    0.16

    0 20 40 60 80 100 120 140 1600

    10

    20

    30

    40

    50

    60

    70X An

    wt%

    TiO2

    Distance from pocket margin (m)

    0.00

    0.02

    0.04

    0.06

    0.08

    0.10

    0.12

    0.14

    0.16

    0 50 100 150 200 250 300 3500

    10

    20

    30

    40

    50

    60

    70

    X An

    wt%T

    iO2

    Distance from pocket margin (m)

    (a)(b)

    (c)

    (d)

    (e)

    ox

    pl

    px + ap

    pl

    qz

    K-fs

    Fig. 4. Continuous zoning profiles towards Fe-rich (a) and Si-rich (d) interstitial pockets from sample SP46 (LZa*^LZb* boundary,

    Skaergaard Peninsula). Zoning at the margins of Fe-rich pockets shows decreasing TiO2 (filled symbols, b) but increasing (reverse) XAn (opensymbols, b) adjacent to the margin (550mm). In c ontrast, zoning at the margins of Si-rich pockets shows a decrease in TiO2 (filled symbols, e)and decreasing (normal) XAn (open symbols, e). This results in two diverging Ti^XAn trends (c; shows four separate profiles; normal zoning pro-files, circles; reverse zoning profiles, diamonds).

    HUMPHREYS INTERSTITIAL LIQUID IMMISCIBILITY

    155

  • 7/27/2019 Humphreys. 2011. Silicate Liquid Inmmiscibility Within the Crystal Mush ... Etc

    10/28

    0.00

    0.02

    0.04

    0.06

    0.08

    0.10

    0.12

    0.14

    0.16

    10 20 30 40 50 60 70 80 90

    0.00

    0.02

    0.04

    0.06

    0.08

    0.10

    0.12

    0.14

    0.16

    10 20 30 40 50 60 70 80 90

    0.00

    0.02

    0.04

    0.06

    0.08

    0.10

    0.12

    0.14

    0.16

    10 20 30 40 50 60 70 80 90

    0.00

    0.02

    0.04

    0.06

    0.08

    0.10

    0.12

    0.14

    0.16

    10 20 30 40 50 60 70 80 90

    0.00

    0.02

    0.04

    0.06

    0.08

    0.10

    0.12

    0.14

    0.16

    10 20 30 40 50 60 70 80 90

    0.00

    0.02

    0.04

    0.06

    0.08

    0.10

    0.12

    0.14

    0.16

    10 20 30 40 50 60 70 80 90

    0.00

    0.02

    0.040.06

    0.08

    0.10

    0.12

    0.14

    0.16

    10 20 30 40 50 60 70 80 90

    HZ* LZa*

    LZb* LZc*

    MZ* UZa*

    UZb*

    X An

    X An

    wt%T

    iO2

    w

    t%T

    iO2

    wt%T

    iO2

    wt%T

    iO2

    (a) (b)

    (c) (d)

    (e)

    (g)

    (f)

    (no samples analysed)

    SP60KR02_4

    SP46KR02_37

    SP16SP8

    KR02_21

    SP20

    SP29

    Fig. 5. TiO2^XAn core and interstitial compositional data for traverses through the Marginal Border Series on the Skaergaard Peninsula(grey circles) and northern Kraemer Island (black triangles), with Layered Series primocryst compositions for comparison (grey crosses, asshown in Fig. 3). Black continuous lines indicate representative continuous zoning profiles from each dataset. No data are available for UZa.Grey arrow in UZb* (g) points to analyses from ilmenite-rich intergrowths (up to 023wt % TiO2). Sample SP60 is from base LZa*,Skaergaard Peninsula, shown in (a) for clarity.

    JOURNAL OF PETROLOGY VOLUME 52 NUMBER1 JANUARY 2011

    156

  • 7/27/2019 Humphreys. 2011. Silicate Liquid Inmmiscibility Within the Crystal Mush ... Etc

    11/28

    point at low TiO2 and intermediate anorthite content

    (An45).

    In LZb* (SP8), interstitial pockets may contain any

    of apatite, quartz, clinopyroxene, biotite/hornblende,

    plagioclase, zircon and opaque minerals, and account for

    $067% by area of the section (Table 1). Interstitial plagio-

    clase compositions lie on the primocryst trend until$An47. Fewer analyses show very evolved compositions

    than in the LZa*^LZb* sample (SP46), but several ana-

    lyses plot on a similar normal zoning trend to that identi-

    fied in SP46, reaching very low TiO2 at $An28. Highly

    calcic, low-TiO2 compositions were not observed, although

    points plotting below the primocryst trend at An49,

    005 wt % TiO2 are arguably part of a steeper trend to-

    wards calcic compositions (Fig. 5c).

    In LZc* (SP16), the few pockets observed were highly

    altered or contained opaque minerals, biotite, plagioclase

    and apatite, and accounted for $036% by area.

    Interstitial plagioclase follows the primocryst trend until

    $An45 ( 3 mol % anorthite), when the interstitial com-positions diverge towards lower TiO2. In this sample a

    reverse interstitial trend is not observed, only steep

    normal zoning (Fig. 5d). Several compositions have anom-

    alously high TiO2 contents; these analyses were from

    traverses along thin interstitial plagioclase cusps between

    clinopyroxene and Fe^Ti oxides, and probably reflect

    diffusive contamination or secondary fluorescence effects.

    These compositions show very slight reverse zoning

    (Fig. 5d).

    For MZ* (SP20), the interstitial plagioclase compos-

    itions lie on the primocryst trend with no divergence to

    lower TiO2; there is no evidence for either of the two dis-

    tinct interstitial zoning trends seen in the lower strati-graphic zones (Fig. 5e). Instead, zoning profiles towards

    interstitial pockets ($027% by area) follow the compos-

    itional trend defined by the primocrysts.

    Interstitial plagioclase from UZb* (SP29) again lies

    on the compositional trend defined by the primocrysts

    (Fig. 5g). Plagioclase adjacent to the granophyric non-

    replacive microstructures described by Holness et al. (2011)

    has very low XAn ($An25), whereas plagioclase from

    within the coexisting ilmenite-rich intergrowths has anom-

    alously high TiO2 (4011wt % and up to 025 wt % TiO2,

    An18^38). These data were excluded on the basis that these

    anomalous compositions were probably caused by second-

    ary fluorescence.

    The other minor elements (Mg and Fe) typically show

    less clear compositional differences thanTi. MgO contents

    of interstitial plagioclase in the Skaergaard Peninsula are

    low and decrease continuously with decreasing XAn, over-

    lapping with the primocryst compositions. Zoning profiles

    towards silicic pockets typically show constant or decreas-

    ing MgO, and zoning profiles towards mafic pockets have

    little change in MgO contents. For FeO, zoning towards

    the felsic pockets or granophyre typically shows constant

    or slightly decreasing FeO contents, and the up-turn

    in FeO seen for cumulus minerals at $An40 is not seen

    (Fig. 6). Zoning towards mafic pockets tends towards

    higher FeO (Fig. 6). For MZ and above, Fe compositions

    are scattered and show no clear trend.

    Interstitial plagioclase: Marginal BorderSeries, Kraemer IslandKraemer Island has a low structural position within the in-

    trusion, so the MBS comprises only HZ*, LZa* and part

    of LZb* at this location. In terms of MgO and FeO con-

    tent, the Kraemer Island plagioclase compositions overlap

    with those from the Skaergaard Peninsula, but FeO con-

    tents are scattered towards high values (Fig. 6; Electronic

    Appendix 2). However, in TiO2 content, interstitial plagio-

    clase from Kraemer Island samples shows some differences

    compared with the Skaergaard Peninsula. The part of the

    Skaergaard Peninsula interstitial trend that matches the

    primocrysts is typically also displayed by interstitial

    plagioclase from Kraemer Island. However, nearly all the

    analyses from Kraemer Island plot on the reverse zoning

    interstitial trend, similar to that defined adjacent to

    Fe-rich pockets in the Skaergaard Peninsula (Fig. 4).

    Many more reversed compositions are found compared

    with the Skaergaard Peninsula, fleshing out the shape of

    this trend and (for LZb*) extending it (Fig. 5). The

    planar-sided pockets found in the Kraemer Island samples

    are much less common than those in the Skaergaard

    Peninsula (0015^020% by area; Table 1); typically the

    identifiable minerals are apatite and pyroxenes as well as

    plagioclase.

    Interstitial plagioclase: Layered Series,Cambridge drill core and drill core 90-22The compositions of plagioclase from the central Layered

    Series are shown in Figs 7 and 8, and extend earlier obser-

    vations (Humphreys, 2009). The results show a complex

    mixture of the different XAn^TiO2 trends observed in the

    Marginal Border Series. In HZ (sample 118678, Fig. 7a),

    the compositions are almost identical to those of sample

    SP60 on Skaergaard Peninsula (see Fig. 5; Electronic

    Appendices 3, HZ-MZ, and 4, UZ). Interstitial pockets

    represent $011% by area (Table 1). LZa plagioclase com-

    positions (sample 118653) are similar to those of HZ, the

    primocryst trend being followed until $An52, with a max-

    imumTiO2 content of 0

    13 wt % observed. The interstitialcompositions separate to low and high XAn contents at

    low TiO2 (Fig. 7b), as for HZ, but in contrast to LZa*

    plagioclase. However, the lowest and highest XAn compos-

    itions observed in these divergent trends are equivalent to

    those seen in LZa* (An35 and An75, respectively).

    Interstitial pockets represent$020% by area (Table 1).

    Plagioclase compositions from LZb (samples 118605,

    118601, 118590) follow the primocryst trend until An49,

    HUMPHREYS INTERSTITIAL LIQUID IMMISCIBILITY

    157

  • 7/27/2019 Humphreys. 2011. Silicate Liquid Inmmiscibility Within the Crystal Mush ... Etc

    12/28

    HZ* LZa*

    LZb* LZc*

    MZ* UZa*

    UZb*

    X An

    X An

    wt%FeO

    (a) (b)

    (c) (d)

    (e)

    (g)

    (f)

    (no samples analysed)

    w

    t%FeO

    w

    t%FeO

    wt%FeO

    0

    0.2

    0.4

    0.6

    0.8

    0

    0.2

    0.4

    0.6

    0.8

    0

    0.2

    0.4

    0.6

    0.8

    0

    0.2

    0.4

    0.6

    0.8

    0

    0.2

    0.4

    0.6

    0.8

    0

    0.2

    0.4

    0.6

    0.8

    0

    0.2

    0.4

    0.6

    0.8

    10 20 30 40 50 60 70 80 90 10 20 30 40 50 60 70 80 90

    10 20 30 40 50 60 70 80 90 10 20 30 40 50 60 70 80 90

    10 20 30 40 50 60 70 80 90 10 20 30 40 50 60 70 80 90

    10 20 30 40 50 60 70 80 90

    SP60KR02_4

    SP46KR02_37

    SP16

    SP8

    KR02_21

    SP20

    SP29

    Fig. 6. FeO XAn core and interstitial compositional data for the Skaergaard Peninsula and northern Kraemer Island traverses through theMarginal Border Series, with Layered Series primocryst compositions for c omparison, as in Fig. 3. Samples and symbols as for Fig. 5.

    JOURNAL OF PETROLOGY VOLUME 52 NUMBER1 JANUARY 2011

    158

  • 7/27/2019 Humphreys. 2011. Silicate Liquid Inmmiscibility Within the Crystal Mush ... Etc

    13/28

    0.00

    0.02

    0.04

    0.06

    0.08

    0.10

    0.12

    0.14

    0.16

    10 20 30 40 50 60 70 80 900.00

    0.02

    0.04

    0.06

    0.08

    0.10

    0.12

    0.14

    0.16

    10 20 30 40 50 60 70 80 90

    0.00

    0.02

    0.04

    0.06

    0.08

    0.10

    0.12

    0.14

    0.16

    10 20 30 40 50 60 70 80 900.00

    0.02

    0.04

    0.06

    0.08

    0.10

    0.12

    0.14

    0.16

    10 20 30 40 50 60 70 80 90

    0.00

    0.02

    0.040.06

    0.08

    0.10

    0.12

    0.14

    0.16

    10 20 30 40 50 60 70 80 900.00

    0.02

    0.040.06

    0.08

    0.10

    0.12

    0.14

    0.16

    10 20 30 40 50 60 70 80 90

    0.00

    0.02

    0.04

    0.06

    0.08

    0.10

    0.12

    0.14

    0.16

    10 20 30 40 50 60 70 80 90

    X An

    X An

    wt%T

    iO2

    w

    t%T

    iO2

    wt%T

    iO2

    wt%T

    iO2

    (a) (b)

    (c) (d)

    (e)

    (g)

    (f)

    HZ LZa

    LZb LZc

    MZ UZa

    UZb

    90-22-461.890-22-421

    90-22-824.2

    90-22-481.890-22-983.97 and 995.09

    90-22-966.16 and 1000.08

    458279

    1966-1186531966-118678

    1966-1185901966-1186011966-118605

    Fig. 7. TiO2^XAn core and interstitial compositional data for the central Layered Series from drill cores 1966 and 90-22 as well as surfacesample (458279), with Layered Series primocryst compositions for comparison (grey crosses; see Fig. 3). Samples are 118678, HZ (a); 118653,LZa (b); 118605, 118601 and 118590, LZb (c); 458279, LZc (d). MZ compositions ( e) include unpublished zoning traverses from samples90-22-966.16 and 90-22-1000.08 (diamonds, Stripp, 2009) and plagioclase zoning towards granophyre veins in samples 90-22-995.09 and90-22-983.97 (crosses, Stripp, 2009). Upper zone samples include 90-22-824.2 (f, lower UZa), 90-22-481.8 (f, UZa^UZb boundary), 90-22-461.8(g, lower UZb) and 90-22-421 (g, mid-UZb). Black continuous lines indicate representative continuous zoning profiles from each dataset.

    HUMPHREYS INTERSTITIAL LIQUID IMMISCIBILITY

    159

  • 7/27/2019 Humphreys. 2011. Silicate Liquid Inmmiscibility Within the Crystal Mush ... Etc

    14/28

    X An

    X An

    wt%FeO

    (a) (b)

    (c) (d)

    (e)

    (g)

    (f)

    w

    t%FeO

    w

    t%FeO

    wt%FeO

    HZ LZa

    LZb LZc

    MZ UZa

    UZb

    90-22-461.890-22-421

    90-22-824.290-22-481.890-22-983.97 and 995.09

    90-22-966.16 and 1000.08

    458279

    1966-1186531966-118678

    1966-1185901966-1186011966-118605

    0

    0.2

    0.4

    0.6

    0.8

    0

    0.2

    0.4

    0.6

    0.8

    0

    0.2

    0.4

    0.6

    0.8

    0

    0.2

    0.4

    0.6

    0.8

    0

    0.2

    0.4

    0.6

    0.8

    0

    0.2

    0.4

    0.6

    0.8

    0

    0.2

    0.4

    0.6

    0.8

    10 20 30 40 50 60 70 80 90 10 20 30 40 50 60 70 80 90

    10 20 30 40 50 60 70 80 90 10 20 30 40 50 60 70 80 90

    10 20 30 40 50 60 70 80 90 10 20 30 40 50 60 70 80 90

    10 20 30 40 50 60 70 80 90

    Fig. 8. FeO XAn core and interstitial compositional data for the central Layered Series. Samples and symbols as for Fig. 7.

    JOURNAL OF PETROLOGY VOLUME 52 NUMBER1 JANUARY 2011

    160

  • 7/27/2019 Humphreys. 2011. Silicate Liquid Inmmiscibility Within the Crystal Mush ... Etc

    15/28

    where the interstitial trend diverges towards low XAn and

    low TiO2. The interstitial trend is similar to that seen for

    LZa, but no highly calcic compositions were observed

    except for one point at An55 (Fig. 7c). The mode of intersti-

    tial pockets is variable in the three LZb samples, represent-

    ing $006 to $033% by area; this could reflect sampling

    issues or inter-sample variations (see below).In LZc (sample 458279) the interstitial trend diverges

    from the primocryst trend at $An48. The interstitial trend

    is similar to that seen for LZb until $An55; most compos-

    itions become more calcic with decreasing TiO2, although

    there is also a hint of the normal zoning trend at around

    An46 (Fig. 7d). One highly calcic interstitial plagioclase

    composition (An78) was recorded at a pl^pl^cpx triple

    junction. Interstitial pockets are very rare in this sample,

    representing only $003% by area (Table 1).

    In MZ (samples from 96616 m and 100008 m depth in

    drill core 90-22; data from Stripp, 2009), the primocryst

    trend is not followed substantially: late-stage interstitial

    plagioclase plots below the primocryst trend at $An45.Few low-TiO2 compositions were observed, making it

    difficult to assess how these data fit into the series. Zoning

    profiles towards isolated granophyre veins in other MZ

    samples (99509 m and 98397 m in drill core 90-22; data

    from Stripp, 2009) lie on the primocryst trend until

    $An38 and then diverge to lower TiO2 at $An32 (Fig. 7e).

    A few analyses plot at anomalously high TiO2 contents

    (above the primocryst trend); these may result from diffu-

    sive contamination or secondary fluorescence effects and

    appear similar to those from LZc* (see Fig. 5d).

    Interstitial plagioclase compositions in UZa (8242 m

    and 4818 m in drill core 90-22) show variable trends.

    The lower sample, from 824

    2 m in mid-UZa, shows twotrends of low-TiO2 interstitial compositions that diverge

    at $An43, one towards $An30 and one towards $An55(Fig. 7f). In contrast, interstitial plagioclase from 4818 m

    (at the UZa^UZb boundary) lies on the primocryst trend

    up to very evolved compositions ($An20), with no evidence

    of highly calcic, low-TiO2 compositions. Interstitial pock-

    ets are very rare (5001% by area in sample from

    8242 m; Table 1).

    Plagioclase compositions from UZb (samples from

    4618 m and 212 m in drill core 90-22), lie on a linear

    trend of decreasing TiO2 and XAn, with compositions

    very similar to those from the UZa^UZb boundary. The

    trend is very slightly steeper than the equivalent trend for

    the Skaergaard Peninsula (Fig. 7g). Again, there is no evi-

    dence of highly calcic, low-TiO2 compositions. Two ana-

    lyses plot at higher TiO2; these are probably similar to

    high-TiO2 compositions from MZ. Interstitial pockets are

    present in the sample from 4618 m ( 026% by area),

    which has no non-replacive symplectites (Table 1).

    As with the MBS (see above) and the lower parts of the

    LS (Humphreys, 2009), MgO and FeO contents of

    interstitial plagioclase also vary with XAn. MgO contents

    are lower than in the MBS, and are highest in HZ^LZ

    (005 wt % MgO), decreasing continuously to $001wt %

    at UZb. FeO contents (Fig. 8) tend to be slightly lower

    than in the MBS, as with the primocryst compositions,

    and decrease from $045wt % in HZ to $015 wt % in

    upper UZb. More evolved compositions tend either to-wards high FeO contents with little difference in FeO, or

    towards lower FeO and lower XAn. As with the MBS sam-

    ples, the most evolved plagioclase compositions (An30 and

    below) do not show the increase in FeO observed in the

    primocryst compositions. The two distinct trends are most

    clear for MZ, UZa and UZb. For the MZ samples, the

    low-Fe trend is defined by plagioclase adjacent to grano-

    phyre veins (Stripp, 2009).

    Interstitial plagioclase: Layered Series,drill core 90-10Four samples in MZ and UZa were examined from drill

    core 90-10, to determine whether spatial position within

    the Layered Series affects interstitial plagioclase compos-

    itions and to look more closely at intra-zone variations

    (Figs 9 and 10; Electronic Appendix 5). For equivalent

    stratigraphic positions, the plagioclase compositions are

    similar to those of drill core 90-22, with core compositions

    in the range An40^48. In general, plagioclase from drill

    core 90-10 is slightly more Fe-rich and more Mg-rich than

    that from drill core 90-22, similar to plagioclase from the

    MBS. In terms of Ti^XAn variations, there is limited evi-

    dence of highly calcic, low-TiO2 compositions, except for

    a few analyses from 90-10-524.1 (meso-gabbro, lower MZ)

    and decreasing TiO2 at approximately constant XAn in

    90-10-456 (oxide-rich layer, mid-MZ; Fig. 9). This behav-

    iour is similar to that seen for drill core 90-22 in LZc orMZ. Samples 90-10-445 (meso-gabbro, top MZ) and

    90-10-212 (mid-UZa) show clear normal zoning with a

    steep decrease in TiO2 between An45 and An35 (Fig. 9),

    similar to the zoning adjacent to veins of granophyre in

    lower MZ (drill core 90-22, Stripp, 2009; see Fig. 7e).

    There is no clear progression of composition with stratig-

    raphy as expected in these MZ samples; the steepness of

    the Ti^XAn trend varies non-systematically with strati-

    graphic height and may depend on the bulk composition

    of the sample, with the steepest trend observed in the

    most oxide-rich sample (90-10-456; Fig. 9). The abundance

    of interstitial pockets varies between these four samples,

    from 0

    13 to 0

    28% by area where present, but the pocketsare absent in 90-10-456 (Table 1).

    SummaryIn general, interstitial plagioclase from a given stratigraph-

    ic zone shows a much greater compositional range than

    primocrysts from that zone. Minor element concentrations,

    in particularTiO2, vary systematically with XAn, but inter-

    stitial plagioclase compositions do not always lie on the

    HUMPHREYS INTERSTITIAL LIQUID IMMISCIBILITY

    161

  • 7/27/2019 Humphreys. 2011. Silicate Liquid Inmmiscibility Within the Crystal Mush ... Etc

    16/28

  • 7/27/2019 Humphreys. 2011. Silicate Liquid Inmmiscibility Within the Crystal Mush ... Etc

    17/28

    the interstitial trend from that of the primocrysts therefore

    indicates that differentiation of the mush liquid does

    not always occur under conditions of near-perfect frac-

    tional crystallizationthe interstitial liquid and the bulkmagma can follow different liquid lines of descent. This de-

    parture from the primocryst trend typically occurs after

    an initial period when interstitial and primocryst com-

    positions coincide (e.g. Fig. 5b). The XAn composition

    at which primocrysts and interstitial plagioclase trends

    diverge decreases with increasing stratigraphic height, sug-

    gesting a porosity or permeability control on crystalliza-

    tion conditions. The gradient of the interstitial trends also

    varies both within samples and with stratigraphic position.

    However, the uppermost parts of the stratigraphy contain

    interstitial plagioclase compositions that do not diverge

    from the primocryst Ti^XAn trend: this suggests that in

    the latest stages of crystallization, the mush undergoesfractional crystallization much as the bulk magma does.

    A key question, therefore, concerns the processes con-

    trolling the interstitial compositional variations from the

    lower stratigraphic zones. Modelling is made difficult

    by continuous solid solution in plagioclase, olivine and

    pyroxenes; peritectic reactions (e.g. pyroxene^olivine); the

    unusual, highly Fe-rich melt compositions reached;

    and the lack of certainty about the original (primocryst)

    compositions of mafic phases owing to sub-solidus diffu-

    sion. However, in a simplified model that assumes negli-

    gible compaction, compositional convection or diffusive

    processes, Ti can be treated simply as a trace element ina liquid that is undergoing progressive fractional crystal-

    lization, with XAn used as a passive marker of both tem-

    perature (T) and F, the fraction of liquid remaining

    (Humphreys, 2009). Using this approach, different bulk

    partition coefficients for Ti (DBTi) can be used to test the ef-

    fects of different crystallizing assemblages, from a starting

    point at moderate XAn and high TiO2 (Fig. 11). In the

    absence of a formula specific to crystallization of the inter-

    stitial liquid, a polynomial function derived from experi-

    mental petrology studies (Toplis & Carroll,1996)

    TK 1295 265F 290F2 1753 1

    is used to determine Tas a function of F, and thus deriveXAn and DTi

    pl (following Humphreys, 2009):

    TC 3:61XAn 899 2

    ln DplTi 0:00535T

    C 9:458: 3

    For simplicity, only three phases are considered: plagio-

    clase, Fe^Ti oxides and clinopyroxene. The DTi values

    for olivine and plagioclase are very similar, so neglecting

    10 20 30 40 50 60 70 80 90

    90-10-456

    90-10-212

    0

    0.2

    0.4

    0.6

    0.8

    10 20 30 40 50 60 70 80 90

    0

    0.2

    0.4

    0.6

    0.8

    10 20 30 40 50 60 70 80 90

    0

    0.2

    0.4

    0.6

    0.8

    90-10-445

    90-10-524.1

    MZ MZ

    UZa

    X An

    w

    t%FeO

    wt%FeO

    (a) (b)

    (c)

    X An

    Fig. 10. FeO XAn core and interstitial compositional data for drill core 90-10, with primocryst compositions for comparison. Samples and sym-bols as for Fig. 9.

    HUMPHREYS INTERSTITIAL LIQUID IMMISCIBILITY

    163

  • 7/27/2019 Humphreys. 2011. Silicate Liquid Inmmiscibility Within the Crystal Mush ... Etc

    18/28

    olivine may overestimate the modal abundance of plagio-

    clase, but plagioclase clinopyroxeneFe^Ti oxides rep-

    resent most of the interstitial material in most rocks.

    Despite its over-simplicity (ignoring crystallization of all

    accessory minerals, for example), the model is useful be-

    cause it demonstrates that, to a first order, significant vari-

    ations in Ti^XAn curvature can be produced simply bydifferent bulk partition coefficients (DBTi), which may cor-

    respond to differing proportions of the interstitial phases

    (Fig. 11). The curvature essentially depends on the propor-

    tion of Fe^Ti oxides crystallizing relative to plagioclase

    and olivine, because of their strong differences in DTi.

    Importantly, the normal zoning trends observed in the

    MBS can be reproduced well, although no unique results

    can be obtained. For example, profiles from sample SP46

    (LZa*^LZb* boundary, Skaergaard Peninsula) can be

    matched by a low bulk DTi that could equate to crystalliza-

    tion of 45% plagioclase, 40% clinopyroxene and 15% Fe^

    Ti oxides (Fig. 11). Therefore, at least in places, interstitial

    crystallization can be explained by relatively simple frac-tional crystallization processes. In principle, using the

    same approach to model equilibrium crystallization will

    result in a slower decrease of Ti concentrations compared

    with fractional crystallization: equilibrium crystallization

    in the mush cannot therefore explain the differences be-

    tween interstitial and primocryst plagioclase compositions.

    However, the interstitial reverse zoning cannot be

    produced by simple fractional crystallization and requires

    a different mechanism. This is of course because XAn is cal-

    culated as a passive marker ofTand F, and therefore de-

    creases monotonically with progressive crystallization.

    The liquid Ca/(CaNa) is not modelled explicitly; the

    simplified dependence of XAn on temperature is taken on

    the basis of the available experimental data (Toplis &

    Carroll, 1995, 1996; Thy et al., 2006). However, processesthat might cause changes in liquid composition must clear-

    ly be considered (see below).

    Given the apparently strong effects of modal mineralogy

    on the Ti^XAn gradient, one might expect the primocryst

    trend to be curved, as predicted by the modelling of Thy

    et al. (2009), or to show significant changes in slope at each

    sub-zone boundary to reflect the changes in mineralogy.

    However, the primocryst trend is almost linear for zones

    LZc and above, suggesting that DBTi changed little during

    differentiation of the upper parts of the intrusion.

    Origin of calcic plagioclase rims and

    reverse zoningAs demonstrated above, the normal interstitial zoningtrends observed are consistent with a relatively simple pro-

    cess of fractional crystallization, but the reverse zoning

    observed in the lower parts of the intrusion is not and

    therefore requires a different mechanism. In principle,

    increased XAn in plagioclase could result from increased

    temperature, increased pH2O, or an increase in the

    anorthite component of the liquid. For the interstitial

    0.00

    0.02

    0.04

    0.06

    0.08

    0.10

    0.12

    0.14

    0.16

    10 20 30 40 50 60 70 80

    wt%T

    iO2

    X An

    1.

    2.

    4.

    3.

    Fig. 11. Numerical modelling results for different bulk Ti partition coefficients, superimposed on interstitial data from sample SP46 (LZa*^

    LZb* boundary, Skaergaard Peninsula) for comparison (symbols as in figure 5b). The continuous curve (1) represents the model fromHumphreys (2009) using DBTi $ 32 after An54 (equivalent to 20% Fe^Ti oxides 40% plagioclase 40% clinopyroxene). The long-dashedline (2) represents DBTi$17 (equivalent to 10% Fe^Ti oxides 50% plagioclase 40% clinopyroxene). The dotted line (3) representsDBTi$ 12 (equivalent to 7% Fe^Ti oxides 70% plagioclase 23% clinopyroxene). The short-dashed line (4) indicates crystallization withDBTi$ 24 ( equivalent to 45% plagioclase40% clinopyroxene15% Fe^Ti oxides), with a starting composition of 42 wt % TiO2 and assum-ing initial Fof 0 28.

    JOURNAL OF PETROLOGY VOLUME 52 NUMBER1 JANUARY 2011

    164

  • 7/27/2019 Humphreys. 2011. Silicate Liquid Inmmiscibility Within the Crystal Mush ... Etc

    19/28

    liquid of a slowly cooled basaltic crystal mush, an increase

    in temperature during solidification seems unlikely, and

    although reversely zoned plagioclase has been produced

    experimentally by metastable crystallization during rapid

    quenching (Lofgren, reported by Grove et al., 1973), this is

    not applicable to intrusive gabbroic suites. If pH2O were

    the cause of the increasing anorthite contents, significantquantities of H2O would need to be present in the liquid

    (Sisson & Grove,1993; Koepke et al., 2005), and the absence

    of hydrous phases associated with the calcic rims argues

    against this. The presence of adjacent calcic and sodic

    rims within the same thin section would require the forma-

    tion of interstitial pockets with widely differing H2O con-

    tents; this also seems unlikely. Furthermore, regions of

    high pH2O should occur preferentially in the outermost

    parts of the intrusion, where external water could most

    easily infiltrate the rocks, or in the most fractionated parts

    (i.e. UZc), where incompatible components are most

    strongly enriched, and not in the centre of the LS. The

    most likely cause of the reverse zoning and calcic rims

    therefore seems to be a change in the composition of the

    interstitial liquid, as argued by Humphreys (2009). There

    are several possible causes of changing liquid composition,

    including convective exchange within the crystal mush,

    dissolution and reprecipitation during compaction, infil-

    tration metasomatism, and silicate liquid immiscibility

    within the mush.

    Compositional convectionCompositional convection within the mush can occur if

    the residual liquid becomes less dense than the adjacent

    crystals or overlying melt (e.g. Sparks et al., 1984; Morse,

    1986). Convection would result in the supply of fresh,

    dense, unevolved melt to the crystallization site, and re-moval of more evolved, less dense melt, and could result

    in adcumulus textures if the convective velocity is greater

    than the growth rate of the crystallization front (e.g.

    Tait et al., 1984; Kerr & Tait, 1986; Tait & Jaupart, 1992).

    For floor cumulates, the melt composition at the magma^

    mush interface will be essentially identical to that in the

    main magma reservoir, with more evolved residual melt

    near the base of the mush (lower Ca content, higher FeO,

    higher alkalis, Thy et al., 2009; although see Boorman

    et al., 2004). Compositional convection within the intersti-

    tial liquid could therefore result in mineral zonation near

    the base of the mush. This would correspond to decreasing

    FeO, increasing Mg-number for mafic phases, and increas-ing XAn (reverse zoning) in plagioclase. However, once

    the interstitial liquid is saturated with Fe^Ti oxides, TiO2concentrations in the interstitial liquid will fall below

    those of the overlying bulk magma, so this reverse zoning

    would also be associated with increasing TiO2 in plagio-

    clase, whereas the observations show increasing FeO and

    decreasing TiO2. Compositional convection would also be

    strongly hindered by the low permeability near the base

    of the crystal mush, as demonstrated by the very small pro-

    portion of total interstitial plagioclase represented by

    calcic compositions. This suggests that convection within

    the crystal mush is not the reason for the observed grain

    boundary zoning.

    Infiltration metasomatismMetasomatism by expulsion of interstitial liquids fromlower parts of the crystal mush during compaction was

    suggested by Irvine (1980) to explain discrepancies be-

    tween observed modal and geochemical zoning patterns

    in the cyclically layered Muskox intrusion, Canada.

    Compaction also occurs at Skaergaard (e.g. McBirney,

    1995; Tegner et al., 2009); however, unlike the Muskox

    Intrusion, the Skaergaard Intrusion represents essentially

    a single pulse of magma that solidified through continuous

    fractional crystallization. Residual liquids deeper in the

    crystal mush are therefore probably more evolved, not

    more primitive, than those at shallow depths in the mush

    or even in the bulk magma (although see Boorman et al.,

    2004), so infiltration metasomatism at Skaergaard would

    result in normal zoning, not reverse zoning. Injection of

    new, primitive magma (e.g. Roelofse et al., 2009) is ruled

    out because of a lack of supporting field evidence, because

    the reversals are seen in samples covering a wide strati-

    graphic range, and because the reversals are seen only in

    the outermost grain margins.

    Dissolution^reprecipitationLocal dissolution of plagioclase can result from com-

    paction during post-cumulus growth, owing to dissolution

    of unfavourably oriented plagioclase and reprecipitation

    of the material in lower stress orientations, as suggested

    for the Stillwater Intrusion (Meurer & Boudreau, 1998)and the Skaergaard Intrusion (Boudreau & McBirney,

    1997). Maale (1976) showed that the dissolving plagioclase

    grain is always more An-rich than that precipitating,

    which should result in enrichment of the anorthite compo-

    nent in the interstitial liquid, and could explain local

    reverse zoning. However, this mechanism cannot account

    for reversed plagioclase within the steeply dipping, and

    relatively rapidly cooled MBS, where compaction should

    be negligible, nor the transition to normal (i.e.

    non-reversed) overgrowth rims in higher stratigraphic

    zones of the MBS. Furthermore, Tegner et al . (2009)

    showed that compaction is unimportant in HZ^LZa,

    where strong reverse zoning is seen.

    Silicate liquid immiscibilityImmiscibility of the interstitial liquid was not previously

    considered (Humphreys, 2009) but could explain the dif-

    fering compositional trends observed in interstitial plagio-

    clase. Liquid immiscibility in silicate magmas has been

    demonstrated unambiguously by petrographic observa-

    tions (e.g. De, 1974; Philpotts, 1979, 1982), as well as

    HUMPHREYS INTERSTITIAL LIQUID IMMISCIBILITY

    165

  • 7/27/2019 Humphreys. 2011. Silicate Liquid Inmmiscibility Within the Crystal Mush ... Etc

    20/28

    experimental studies (e.g. Roedder, 1951; Dixon &

    Rutherford, 1979; Philpotts & Doyle, 1983). Late-stage im-

    miscibility within the Skaergaard magma specifically has

    also been demonstrated, both by the presence of two dis-

    tinct populations of melt inclusions in apatite primocrysts

    (UZb, Jakobsen et al., 2005) and by experimental observa-

    tions (McBirney & Nakamura, 1974; Veksler et al., 2007). Ithas been suggested that in fact the magma intersects the

    two-liquid miscibility gap much earlier in the fractionation

    process, around upper LZ to upper MZ, when the

    magma was $45^60% crystallized (Veksler et al., 2007;

    Holness et al ., 2011), although this is controversial

    (McBirney, 2008; Morse, 2008a; Philpotts, 2008; Veksler

    et al., 2008). Irrespective of the stage of fractionation at

    which immiscibility occurs, the magma at Skaergaard

    exsolves into an Fe-rich liquid that is also rich in Ca and

    P, and a Si-rich liquid that is also rich in alkalis (e.g.

    McBirney & Nakamura, 1974; Jakobsen et al ., 2005;

    Veksler et al., 2007).

    At thermodynamic equilibrium, the chemical potential

    of each component must be the same in every phase in the

    system, including both conjugate liquids. However, this

    does not equate to equal equilibrium concentrations,

    hence the markedly different compositions of conjugate im-

    miscible liquids and their different trace element concen-

    trations (e.g. Veksler, 2009). Furthermore, each conjugate

    liquid may crystallize different modal proportions of the

    minerals present, and partition coefficients can also differ

    (Roedder, 1978; Philpotts, 1979; Veksler et al ., 2006).

    For minerals with complete solid solution, the effects of

    liquid immiscibility on an intermediate composition can

    be thought of in terms of the end-member components.

    For example, in plagioclase feldspar the Si-rich,

    alkali-rich liquid would crystallize dominantly the albite

    (NaAlSi3O8) component whereas crystallization from the

    Fe-rich, Ca-rich liquid would be dominated by the anorth-

    ite component (CaAl2Si2O8). At equilibrium, when both

    conjugate liquids are present in cotectic proportions, there

    is no change in the feldspar composition compared with

    crystallization from a single liquid, but disequilibrium con-

    ditions could result in crystallization of more An-rich

    plagioclase from the Fe-rich liquid and more An-poor

    plagioclase from the silicic liquid. In the Skaergaard cu-

    mulates, the observation of highly discrepant composition-

    al zonation in different parts of the same samples

    indicates that disequilibrium crystallization must be occur-

    ring, resulting in the crystallization of anomalously calcicplagioclase.

    C R YS TA L L I Z A T I O N F R O M

    I M M I S C I B L E L I Q U I D SBecause of experimental difficulties, the composition of

    plagioclase in equilibrium with each of the liquids

    separately is not known independently, but can be esti-

    mated if the anorthite component (XCaAl2Si2O8) and

    albite component (XNaAlSi3O8) of the conjugate liquids

    are known. These components can be calculated (Lange

    et al., 2009) from known or estimated t wo-liquid pairs

    (Table 2). The calculated anorthite components vary from

    0064 to 1

    175 for Fe-rich conjugate liquids and from 0

    076

    to 0247 for Si-rich conjugate liquids, whereas the calcu-

    lated albite components vary from 0047 to 0162 for

    Fe-rich and from 0286 to 0613 for Si-rich conjugate liquids

    (Table 2), based on liquid compositions from

    re-homogenized melt inclusions (Jakobsen et al., 2005), ex-

    perimental compositions (Veksler et al., 2007) and melting

    experiments (McBirney & Nakamura, 1974). The higher

    factors for Si-rich liquids reflect their higher affinity for

    both calcic and sodic feldspars, but in every case, the nom-

    inal XAn for the liquid, calculated as XCaAl2Si2O8/

    (XCaAl2Si2O8XNaAlSi3O8), is substantially lower for

    the Si-rich liquids (0120^0464) than the Fe-rich liquids

    (0

    391^0

    636, Table 2). This demonstrates that, if the liquidsare out of equilibrium with one another, the Fe-rich liquid

    should crystallize more calcic plagioclase, whereas the

    Si-rich liquid should crystallize more evolved (albite-rich)

    plagioclase. Thus the observed reverse zoning and anomal-

    ously high anorthite contents of some interstitial plagio-

    clase are compatible with disequilibrium crystallization

    from an emulsion with greater than cotectic proportions

    of the Fe-rich component, or from the Fe-rich liquid itself.

    It is concluded that strong disequilibrium can occur

    within the mush, and that immiscible melts that exsolve

    from the bulk residual liquid do not necessarily remain in

    equilibrium with each other after exsolution, as suggested

    by Crawford & Hollister (1977).This concept is supported by the results of several previ-

    ous studies. Ripley et al. (1998) reported calcic plagioclase

    (An71^79) in nelsonite bodies from the Duluth Complex,

    Minnesota, which are thought to have formed from

    an immiscible Fe-rich liquid. Similarly, Loferski &

    Arculus (1993) reported highly calcic (An92) rims

    around plagioclase-hosted melt inclusions that had crystal-

    lized to cpx ilmap titanite, rutile and baddeleyite,

    in rocks from the Stillwater Intrusion, Montana. In lunar

    basalts, two distinct plagioclase compositional trends were

    observed adjacent to patches of high-K mesostasis

    (to An64) and low-K mesostasis (to An94^96, Crawford &

    Hollister, 1977). The Loch Ba ring dyke, Scotland, formed

    by mixing of rhyolitic and mafic melts; the melts contain

    plagioclase of An21^32 (rhyolite) and An30^65 (mafic) re-

    spectively (Sparks, 1988).

    Disequilibrium crystallization could occur as a result of

    physical separation, perhaps by gravitational separation of

    two liquids of differing densities, or by diffusion rates fall-

    ing below the threshold required to maintain equilibrium

    (Roedder, 1978). This would require the liquid droplets to

    JOURNAL OF PETROLOGY VOLUME 52 NUMBER1 JANUARY 2011

    166

  • 7/27/2019 Humphreys. 2011. Silicate Liquid Inmmiscibility Within the Crystal Mush ... Etc

    21/28

    be separated by more than the characteristic diffusion dis-

    tance over the relevant crystallization timescale, and

    could be achieved either by complete physical segregation

    of the immiscible liquids, perhaps through gravitationalseparation (because the silicic liquid will have much

    lower density than the Fe-rich liquid), or by partial but

    incomplete separation at different temperatures and

    crystallization rates. A droplet separation of 100mm, for

    example, might permit equilibrium crystallization at slow

    crystallization rates (when there is a longer time available

    for diffusion) or at higher temperatures (when diffusivities

    are higher), yet at more rapid crystallization rates or low

    temperatures the same droplet separation could result in

    disequilibrium and hence changes in mineral composition.

    This means that, under certain conditions, an emulsion

    that contains non-cotectic proportions of two immiscible li-

    quids can undergo disequilibrium crystallization to formcrystals of anomalous composition. A quantitative decou-

    pling of the effects of crystallization rate and diffusivity is

    beyond the scope of this study. However, the compositions

    of crystal rims will depend on the relative rates of droplet

    exsolution, liquid^liquid segregation, diffusion and crystal-

    lization with falling temperature.

    The permeability of the crystal mush may also be im-

    portant in driving disequilibrium, by producing pockets

    of varying composition, in particular in the least evolved

    cumulates. Immiscibility in the interstitial liquid will

    occur at a very late stage of crystallization in the LZ cumu-

    lates, when the residual porosity and permeability arevery low. Local differences in modal mineralogy could

    cause the composition of the residual liquid in each pore

    to vary widely, and residual liquids remaining in the

    pores may not be in chemical equilibrium. The wide vari-

    ation of plagioclase compositions that can occur around a

    single pocket demonstrates that these final stages of solidifi-

    cation can be highly localized. Differing proportions of

    the two liquids in these residual pockets, or perhaps the

    interstitial liquid locally entering the miscibility gap from

    opposite sides of the solvus, could therefore result in the

    disequilibrium trends observed. However, it is not entirely

    clear why the initial interstitial trend is so steep (e.g.

    Fig. 5a). It is possible that the steep gradient reflects highDTi and may also be related to some permeability thresh-

    old, but this cannot be constrained further using the data

    presented here.

    Interstitial liquid immiscibility inthe Skaergaard PeninsulaThe clearest signature of plagioclase crystallization from

    single conjugate liquids is in sample SP46 from the LZa*^

    Table 2: Compositions of conjugate immiscible silicate liquids from Skaergaard

    Jakobsen et al. (2005) McBirney & Nakamura (1974) Philpotts (1982) Dixon & Rutherford (1979)

    Fe-rich Si-rich Fe-rich Si-rich Fe-rich Si-rich Fe-rich Si-rich

    SiO2 4067 6558 514 655 415 733 439 685

    TiO2 186 022 24 12 58 08 461 170

    Al2O3 787 1295 67 96 37 121 70 111

    FeOt 3085 863 266 119 310 32 2345 786

    MnO 051 013 05 00 055 016

    MgO 235 047 04 04 09 00 232 075

    CaO 897 2 67 31 94 18 1018 382

    Na2O 158 433 22 2 08 31 185 286

    K2O 103 368 1 26 07 33 042 12

    P2O5 025 003 17 03 35 007 487 096

    Total 9594 9802 991 966 978 9767 9915 9891

    n 54 17 15 16 7 7

    XCaAl2Si2O8 011 008 010 011 006 008 012 015

    XNaAlSi3O8 008 051 016 037 005 061 009 047

    XAn liquid 058 014 038 023 055 012 057 024

    The compositions are based on the compositions of homogenized melt inclusions in cumulus apatite (Jakobsen et al.,2005), melting experiments (McBirney & Nakamura, 1974; Dixon & Rutherford, 1979) and analysis of exsolvedglasses (Philpotts, 1982). Dataset from Jakobsen et al . (2005). Anorthite components (XCaAl2Si2O8) and albitecomponent (XNaAlSi3O8) of the liquids are calculated following Lange et al . (2009). XAn liquidXCaAl2Si2O8/[XCaAl2Si2O8XNaAlSi3O8].

    HUMPHREYS INTERSTITIAL LIQUID IMMISCIBILITY

    167

  • 7/27/2019 Humphreys. 2011. Silicate Liquid Inmmiscibility Within the Crystal Mush ... Etc

    22/28

    LZb* boundary on the Skaergaard Peninsula. In this

    sample, interstitial plagioclase compositions show clear

    normal zoning in the vicinity of planar-sided interstitial

    pockets containing quartzgranophyre, but reverse

    zoning in the vicinity of a mafic pocket dominated by

    apatitepyroxenesFe^Ti oxides (see above and Fig. 4);

    single spot analyses also fall on these trends. Because ofthe disparate bulk major-element compositions of the

    interstitial pockets, and their similarity to the crystallized

    products of immiscible melt inclusions in apatite

    (Jakobsen et al., 2005), the pockets are interpreted as the

    crystallized products of trapped immiscible residual li-

    quids (see above and Holness et al., 2011). The normal and

    reverse plagioclase zoning trends form by disequilibrium

    crystallization towards pores dominated by the Si-rich

    and Fe-rich liquids, respectively. This is supported by the

    compositional similarity to plagioclase in granophyre

    and mafic replacive symplectites, which represent the crys-

    tallized products of immiscible residual liquids (see

    Holness et al., 2011, fig. 13). Thus any difference between pri-mocryst and interstitial compositions in a given zone is

    related to whether disequilibrium crystallization occurs

    within the mush of that zone, in contrast to the bulk

    magma.

    It is worth noting that calcic plagioclase or reverse

    zoning may be present even in the absence of a nearby

    interstitial pocketreverse-zoned plagioclase commonly

    occurs on grain boundaries or at triple junctions, in con-

    trast to the very An-poor compositions, which are almost

    always associated with quartz-bearing interstitial pockets.

    This could be related to the differing viscosities and/or

    interfacial energies of the two liquids, with the inviscid

    Fe-rich liquid able to form thin films on grain boundaries,and the highly viscous Si-rich liquid tending to sit in more

    equant pore spaces. This is consistent with the morphology

    of inferred residual liquid films as determined from

    glass-bearing nodules and cumulate rocks (Morse &

    Nolan, 1984; Holness, 2005; Holness et al., 2007a).

    It is of interest to estimate the residual porosity when the

    interstitial liquid intersects the miscibility gap, and this

    can be done approximately using the modelling results of

    Thy et al. (2009). According to their modelled liquids, the

    Skaergaard magma reaches UZa (by which time experi-

    ments and the presence of paired granophyric and

    ilmenite-rich intergrowths indicate that immiscibility has

    occurred in the bulk magma; Veksler et al., 2007; Holness

    et al., 2011), after 66^68% fractional crystallization. An

    LZa rock (comprising primocrysts plus LZa liquid and

    following the same liquid line of descent as the bulk

    magma) with an initial porosity (fi) of around 50%,

    would reach immiscibility after 66^68% crystallization of

    the interstitial liquid, leaving a residual porosity (fr) of

    fi(1^066) to fi(1^068), or 17^16%. Using the same ap-

    proach, even if immiscibility did not start until the

    equivalent of UZb (Jakobsen et al., 2005), and allowing a

    further 10% crystallization to account for the HZ (see

    Thy et al., 2009) the interstitial liquid in an HZ rock

    could still intersect the miscibility gap with a residual por-

    osity in the region of 11^75% (this approach assumes

    that porosity is not occluded by other processes such as

    adcumulus growth). Therefore the observation that eventhe most primitive rocks undergo silicate liquid immiscibil-

    ity in residual liquid retained in the last interstitial pores

    is consistent with previous studies demonstrating immisci-

    bility in the bulk liquid (McBirney & Nakamura, 1974;

    Jakobsen et al., 2005). However, the permeability when

    this occurs is probably very low, so the exsolved liquids

    are likely to form in isolated pockets. In more evolved

    rocks, interstitial liquid immiscibility will occur at increas-

    ingly high fr until eventually, at the point where the bulk

    magma itself exsolves into two liquids, even the initial por-

    osity will be filled with an emulsion rather than a single

    liquid.

    S P A T I A L D I S T R I B U T I O N O F

    I M M I S C I B L E L I Q U I D S I N M B SThe diverse compositions of interstitial plagioclase on

    grain boundaries and adjacent to interstitial pockets are

    consistent with disequilibrium crystallization from immis-

    cible liquids during the late stages of solidification. The

    spatial distribution of these diverse plagioclase compos-

    itions can provide insights into the distribution of the two

    liquids within the crystal mush and throughout the intru-

    sion. The simplest sample traverse to consider is that

    through the Marginal Border Series on the Skaergaard

    Peninsula. The MBS crystallized more or less in situ onthe walls of the intrusion, and is therefore free from the ef-

    fects of compaction and current-driven sedimentation,

    unlike the floor cumulates (Layered Series). The

    Skaergaard Peninsula traverse represents the most com-

    plete cross-section through the evolving cumulates.

    The presence of the two distinct interstitial plagioclase

    compositional trends in HZ* indicates that, at the final

    stages of crystallization, the residual interstitial liquid

    reached the miscibility gap and the two liquids were sepa-

    rated, probably into spatially distinct pockets, although

    probably only the most evolved liquids reached immiscibil-

    ity, at a very low residual porosity. In LZa*, the reverse

    Ti^XAn trend appears nearly identical to that in HZ*, butthe two interstitial trends diverge earlier, suggesting that

    the residual liquid formed coarser pockets and more

    evolved compositions.

    Samples from LZb* and LZc* show the same pattern,

    although the most extreme compositions were not

    observed, as reflected in the paucity of very low TiO2 con-

    tents. This may reflect a sampling issue or could reflect a

    genuine low abundance in the rocks (see discussion

    JOURNAL OF PETROLOGY VOLUME 52 NUMBER1 JANUARY 2011

    168

  • 7/27/2019 Humphreys. 2011. Silicate Liquid Inmmiscibility Within the Crystal Mush ... Etc

    23/28

    below). However, despite a lack of definition, two distinct

    trends are still apparent in LZb*, with one clearly heading

    towards $An30 and the other discernible in several ana-

    lyses below the primocryst trend at $An50 (see Fig. 5). In

    LZc* there are two clear trends, one of which defines

    normal zoning whereas the other (a zoning profile towards

    a plagioclase^pyroxene^oxide triple junction) does showslight reverse zoning but is affected by diffusive modifica-

    tion or secondary fluorescence leading to anomalously

    high Ti contents (see Fig. 5). These patterns also indicate

    that two liquids exsolved during disequilibrium crystalliza-

    tion of the interstitial liquid.

    For MZ* and above, interstitial plagioclase compos-

    itions fall on the primocryst trend. The slope is shallower

    than that of the normal zoning trend in stratigraphically

    lower zones, which suggests a lower DBTi and a difference

    in abundance of pyroxene or oxides in the crystallizing as-

    semblage.The appearance of abundant, paired granophyre

    and ilmenite-rich intergrowths within MZ* indicates that

    the onset of immiscibility in the bulk magma also occurs

    at approximately this point (Holness et al., 2011). This is

    earlier than suggested by some researchers (e.g. McBirney

    & Nakamura, 1974) but may be consistent with experi-

    ments (Veksler et al., 2007; although see McBirney, 2008;

    Morse, 2008a; Philpotts, 2008; Veklser et al., 2008) and com-

    positional variations of melt inclusions in plagioclase near

    the LZ^MZ boundary (Jakobsen et al., 2006). The coinci-

    dence of the primocryst and interstitial compositional

    trends at MZ* and above is interpreted as the result of

    crystallization from equilibrium immiscible liquids, as in

    the bulk magma, resulting in no change in feldspar com-

    position and no distinction between interstitial and primo-

    cryst compositions.

    Disappearance of calcic plagioclaseor reverse zoningIt is suggested that the primocryst and interstitial compos-

    itions coincide when the interstitial liquid no longer under-

    goes disequilibrium crystallization, but crystallizes in

    equilibrium as the bulk magma doesin other words,

    when the feldspar composition is buffered by the presence

    of both liquids in equilibrium. In the MBS this occurs

    within upper MZ*, where the appearance of abundant

    paired granophyre-rich and ilmenite-rich intergrowths

    (Holness et al., 2011) also implies that the bulk magma has

    reached the miscibility gap. These intergrowths represent

    large pockets of Si-rich and Fe-rich liquid, respectively,that were present within the crystal mush in the higher

    parts of the stratigraphy (Holness et al., 2011) so the absence

    of two distinct interstitial zoning trends for plagioclase

    there is surprising; one might have expected to see even

    clearer evidence for two distinct trends than in LZ. This

    suggests that chemical communication between the paired

    intergrowths was sufficiently good to preclude disequilib-

    rium crystallization, despite residing in different (albeit

    adjacent) pockets. For the liquids to be in equilibrium,

    they must be separated by less than the characteristic diffu-

    sion distance relative to the crystal interface. The grano-

    phyric and ilmenite-rich intergrowths do clearly occupy

    separate pockets but these are closely associated spatially;

    moreover, the porosity (and permeability) will be high

    when the pockets form, allowing good communication,and cooling rates will be relatively low.

    Alternatively, the paired intergrowths may in fact repre-

    sent pockets of incompletely separated emulsion, with

    varying proportions of the two liquids. Granophyre-rich

    pockets may contain one or more of apatite, oxides, clino-

    pyroxene, biotite or amphibole, and similarly the

    ilmenite-rich intergrowths may be intergrown with grano-

    phyre or K-feldspar as well as biotite, pyroxenes, apatite

    and olivine (Holness et al., 2011). Ilmenite-rich intergrowths

    form from the Fe-rich liquid with a subsidiary component

    of Si-rich liquid, whereas granophyric intergrowths are

    dominated by the Si-rich liquid with a lesser Fe-rich com-

    ponent. Either way, both liquids are still in chemical equi-

    librium, so no change in plagioclase composition resulted.

    TheTi^XAn data presented in this study cannot differenti-

    ate between these explanations; the reality is probably a

    combination of both.

    SummaryPrimitive cumulates from the Skaergaard Peninsula under-

    go silicate liquid immiscibility during crystallization of

    the interstitial liquid. This happens when the residual

    porosity and permeability are low, resulting in isolated

    pockets containing differing proportions of the two

    immiscible liquids. Disequilibrium crystallization can

    occur as a result of rapid crystallization and physical

    separation of the two liquids, and leads to the formationof two distinct interstitial crystallization trends. At

    approximately MZ* and above, slower crystallization and

    inefficient separation of liquids results in equilibrium con-

    ditions and precludes the formation of the discrepant

    plagioclase compositions.

    D I S C U S S I O N

    Differences between Layered Series andMarginal Border SeriesIn comparison with the MBS, interstitial variations in LS

    plagioclase from the 1966 and 90-22 drill cores appear to

    evolve more slowly with increasing stratigraphic height(except for HZ compositions, which are indistinguishable

    from those in HZ* on the Skaergaard Peninsula).

    Specifically, the two distinct interstitial trends tend to split

    at lower TiO2 in the LS, and the presence of two trends

    persists right up to the UZa^UZb boundary, instead of

    dying out in MZ as in the Skaergaard Peninsula.The pres-

    ence of two distinct interstitial trends up to UZa^UZb sug-

    gests that disequilibrium within the mush persisted until

    HUMPHREYS INTERSTITIAL LIQUID IMMISCIBILITY

    169

  • 7/27/2019 Humphreys. 2011. Silicate Li