Myron G. Best. Igneous and metamorphic 2003 Blackwell Science

Подождите немного. Документ загружается.

Multiple lines of independent evidence can

strengthen the mixing interpretation but are never

sufﬁcient to prove this process occurred. Disequilib-

rium textures in a rock that appears to be homoge-

neous but was produced by mixing of felsic and maﬁc

magmas include the following:

1. Complex resorption and overgrowths in phe-

nocrysts in quickly cooled volcanic systems, such

as spongy zones within plagioclase (Figure 7.19)

and rapakivi overgrowth of plagioclase on alkali

feldspar (Figure 7.18): Normal fractional crystal-

lization of ternary feldspars during cooling would

not create rapakivi texture. However, these over-

growth textures can possibly be produced by

changes in P or water fugacity in the magma system.

2. Anhedral, partially dissolved quartz rimmed by an

aggregate of clinopyroxenes (Figure 6.20): The un-

stable quartz grains can be derived from felsic

magma mingled with more maﬁc magma or from

quartz-bearing rock assimilated (discussed later)

into the maﬁc magma in which clinopyroxene is a

stable phase.

Compositional evidence for magma mixing in the

history of a rock or rock suite includes the following:

1. Disequilibrium phases, such as coexisting quartz

and Mg-olivine or calcic and sodic plagioclase.

2. Phenocryst-hosted glass (melt) inclusions that are

compositionally unlike surrounding matrix glass.

3. Distinctive patterns on variation diagrams: Simple

magma mixing produces straight-line trends on

element-element variation diagrams (Figure 12.3

and Special Interest Box 12.1). However, because

of limited thermal energy in most magma systems,

only limited mixing occurs. Consequently, compo-

sitions of the rocks produced by mixing are gener-

ally only displaced slightly away from the parent

magma composition toward the mixing magma

composition.

If mixing proceeds to a homogeneous hybrid

magma, its composition, C

, can be expressed by a sim-

ple mass balance equation of the compositions of the

Differentiation of Magmas

327

01cm

12.11 Mingled rhyolite and basalt lava, Gardiner River, Yellowstone

National Park, Wyoming. In thin section, euhedral plagioclase

and pyroxene phenocrysts from the basalt lava (darker colors)

coexist with corroded quartz and sanidine phenocrysts from

the rhyolite lava. (U.S. Geological Survey photograph courtesy

of R. E. Wilcox and Louise Hendricks.)

Gabbro

Dikes

Pillow

lavas

Fresh magma

from mantle

246

Sea level

Depth (km)

Distance (km)

Older

fractionated

magma

Ultramafic

cumulates

12.12 Schematic cross section through an episodically replenished

oceanic rift magma chamber. Replenishing primitive magma

from the mantle rises buoyantly into the chamber and mixes

with cooler, denser, Fe-rich fractionated magma. Gabbro

made of olivine, pyroxenes, and plagioclase crystallizes on the

chamber walls; olivines and pyroxenes form ultramaﬁc cumu-

lates on the chamber ﬂoor. Evolving basaltic magmas are

episodically injected through ﬁssures in the overlying extend-

ing roof, forming feeder dikes for submarine extrusions,

mostly pillow lavas. (Redrawn from Bryan and Moore, 1977.)

two end-member parent magmas, C

and C

; the mix-

ing proportion, F

, represents the weight fraction of

one of the magmas

12.2 C

 C

 C

(1  F

)

Magma mixing, depending on how disparate the

magmas are, may be accompanied by crystallization of

new phases stabilized by the compositional and ther-

mal properties of the hybrid system.

12.3.2 Assimilation

After leaving its source, a batch of ascending buoyant

primary magma can encounter wall rock of different

composition, especially basaltic magmas from mantle

sources rising into sialic rocks of the continental crust

and any silicate magma encountering Ca-rich limestone

or Al-rich shale or their metamorphic equivalents.

Magmas interact with their surroundings in an attempt

to attain chemical and thermal equilibrium, especially

where they slow or even stop in subterranean storage

chambers. Hot country rocks are by no means inert to

hotter, contrasting magma.

Incorporation of solid rock into a magma of differ-

ent composition is the process of assimilation; it pro-

duces a contaminated magma, which is also hybrid,

like mixed magmas. The contaminant can be country

rock around the magma chamber or xenoliths within

the magma. Assimilation may initially involve simple

physical dispersal of xenoliths and xenocrysts into the

magma, such as Precambrian zircons in Miocene rhyo-

lite. Depending on magma and foreign material com-

positions and temperatures and available time, the for-

eign material chemically equilibrates with the melt to

varying degrees. Minerals may selectively dissolve into

the melt and contaminant ions incorporate into it by

time- and T-dependent diffusion. Commonly, assimila-

tion involves mixing with melts created by melting of

the contaminant rock.

The thermal and chemical principles of assimilation

were enunciated by Bowen (1928) many decades ago.

Assimilation requires thermal energy, the source of

which can only be the magma itself. Heat from the

magma has two sources:

1. That released during cooling to lower T

2. The latent heat of crystallization

As few magmas appear to be superheated above their

liquidus T, the available heat for assimilation is derived

by concurrent crystallization and cooling of the magma

below its liquidus. Section 11.1.1 indicated that the

mass of lower continental crustal rock melted by a mass

of intruded hotter basaltic magma is of the same order

of magnitude, more of this magma would be required

in the cooler shallower crust. Obviously, hotter, more

maﬁc magmas have greater assimilative potential. But

transfer of heat from a magma body into adjacent

cooler rock leads to solidiﬁcation at its contact, build-

ing an armor of solid magmatic rock that inhibits

further assimilation. Pieces of stoped country rock

(Section 9.4.3) within a body of magma afford a signif-

icantly greater surface area over which heat can be

transferred and assimilative processes operate than the

country rock. Volatile ﬂuids liberated from heated

country rock may contaminate volatile-poor magma

with Si, K, Na, and other elements, as the ﬂuid solution

is absorbed into it.

The fate of assimilated crystals depends on their

composition and that of the parent melt. Provided suf-

ﬁcient heat is available, a crystalline phase dissolves if

the silicate melt is not already saturated with respect to

that phase. Thus, quartz xenocrysts can dissolve on a

time scale of days in basaltic melts in which the activity

of silica is 1 (the usual case); melts so contaminated

are enriched in silica. Alkali feldspar, biotite, and horn-

blende in granite assimilated by basalt follow a similar

fate, but the details differ. Assimilation of granite in

basalt magma promotes crystallization of some of the

phases it would have normally precipitated and along

similar lines of liquid descent; however, felsic deriva-

tives are more abundant. Crystals react with a melt

if they would have precipitated from the magma at

higher T. Thus, physically ingested crystals of Mg-rich

olivine into Makaopuhi basalt magma at T  1075°C

and 1 atm (Plate III), where olivine is no longer stable,

would induce precipitation of additional stable pyrox-

ene by a reaction relation with the melt. Xenocrysts

of clinopyroxene—perhaps derived from incorporated

basalt xenoliths—in a granodiorite melt precipitating

stable hornblende but not pyroxene would be ex-

pected to react with the melt, forming, by ionic diffu-

sion, a reaction rim of hornblende surrounding and

possibly eventually replacing the unstable clinopyrox-

ene. Assimilation of quartz xenocrysts, perhaps from

ingested blocks of sandstone, into a quartz-saturated

granitic melt simply adds more modal quartz to the ﬁ-

nal granite. Assimilation of Al-rich minerals into basalt

magma stabilizes calcic plagioclase at the expense of

calcic clinopyroxene, so that leucocratic orthopyrox-

ene gabbro (norite) magmas might form.

Evidence for magma contamination in the history of

a rock is generally only permissive. The presence of

xenocrysts (e.g., quartz in basalt, Figure 6.20) and

xenoliths in a magmatic rock may suggest they are con-

taminants, but xenocrysts can also originate by mixing

of dissimilar magmas, and foreign material can be in-

corporated late into the magma with minimal contami-

nation of the melt. Strained xenocrysts that show un-

dulatory optical extinction under cross-polarized light

in thin section or other solid-state strain effects are es-

pecially useful in distinguishing assimilated solid mate-

rial from phenocrysts in mixed magmas.

328 Igneous and Metamorphic Petrology

Sr and O isotopic signatures (as well as Nd and Pb)

can also provide permissive evidence for assimilation of

continental felsic crust into primitive mantle-derived

basaltic magmas. Precambrian felsic rocks have rela-

tively low Sr but high Rb; over time,

Rb decays and

elevates the

Sr/

Sr ratio (Section 2.6.2) to well

above that of primitive mantle-derived partial melts,

which are typically about 0.703–0.704 (Figures 2.26

and 2.27). If mantle-derived magmas assimilate old

continental crust, their ratio is elevated.

O in mantle

source rocks and their basaltic partial melts is 6%

in contrast to sedimentary rocks, in which 

O 

1032%o (Figure 2.24; Section 2.6.1). Consequently,

mantle-derived magmas that assimilate sedimentary

rocks are enriched in

O. Mantle-derived magmas that

assimilate old felsic rocks and sedimentary rocks (or

their metamorphic equivalents) are enriched in both

Sr and

O (Figure 12.13).

12.4 DIFFERENTIATION IN

BASALTIC INTRUSIONS

Because of their relatively low viscosity and slow cool-

ing, intrusions of basaltic magma potentially provide

an opportunity to evaluate the role of crystal-melt

fractionation in magmatic differentiation. In this sec-

tion, three types of intrusions are examined to deter-

mine the extent to which fractionation does occur

and the nature of the evolved fractionated magmas

and whether intrusions behave as closed systems or

other differentiation processes are also involved. If

they can be accurately interpreted, these intrusions

provide tests of petrologists’ theoretical models and

small-scale, time-restricted laboratory experiments on

crystal-melt systems. Knowledge gained from maﬁc in-

trusions constrains models purporting to account for

compositional variations in extruded lavas that erupted

sequentially from a particular volcanic center and pos-

sibly derived from underlying differentiating storage

chambers.

12.4.1 Palisades Sill

Differentiation in smaller subhorizontal sheets is con-

sidered ﬁrst, using the Palisades sill as an example. The

diabase-granophyre association in such sheets was the

basis for the emphasis by Bowen (1928) that closed-

system crystal-melt fractionation is the most important

process of magmatic differentiation.

The Palisades diabase sill is a subhorizontal sheet in-

trusion exposed for at least 80 km along the Hudson

River opposite New York City. It is one of ennumerable

early Jurassic basaltic dikes and sills intruded along the

east coast of North America during the breakup of the

American and African continental plates (Figure 9.7b;

see also Olsen, 1999). Whether or not it is everywhere

a concordant sheet does not detract from the fact that

it, like many others, displays more or less regular vari-

ations in bulk chemical, modal, and mineral composi-

tion and fabric throughout a vertical extent of hun-

dreds of meters (Figure 12.14). Inward solidiﬁcation of

the sill progressed to a “sandwich horizon” near the

top of the sill where the most evolved residual magma

crystallized to patches of granophyric pegmatite. In

this felsic differentiate, grain size is uneven, ranging

Differentiation of Magmas

329

0.704 0.708 0.712 0.716

Mantle

Altered oceanic

basalt

Deep-ocean

sediment

Continental

sediment

Indonesia

Peru

O (per mil

)

Sr/

12.13 Oxygen and Sr isotopic composition of andesites from the Banda arc, Indonesia, and the Andean arc, southern Peru. Note two Indone-

sia samples in upper right. Range of composition of potential sources shown. (Redrawn from Magaritz et al., 1978.)

from the micrographic quartz-alkali feldspar inter-

growth typical of granophyre (Figure 7.21) to bladelike

Fe-rich clinopyroxenes several centimeters long. An

olivine-rich layer (10–25% of rock) having abun-

dant pyroxene is located 10–25 m above the ﬂoor of

the sill.

The Palisades sill was interpreted for decades to be

a simple product of closed-system differentiation

driven by gravity settling of denser olivines and pyrox-

enes to the intrusion ﬂoor. However, several investiga-

tions in the late twentieth century (see, for example,

Gorring and Naslund, 1995) showed that this hypoth-

esis is ﬂawed. Studies of this and other similar differ-

entiated sills indicate episodic recharge of fresh magma

into the crystallizing fractionating bodies: That is, they

were open magma systems.

The simple gravity settling hypothesis for the

Palisades, wherein precipitating olivine crystals “rain

down” through the crystallizing magma to form the

olivine-rich layer, is rejected for several reasons:

1. The olivine-rich layer pinches and swells along the

exposed extent of the sill and is locally absent.

2. The proportion of olivine increases abruptly into

the layer from essentially zero above it, implying an

unlikely very efﬁcient gravitational settling “sweep”

from above.

3. Olivines in the layer are far more Fe-rich (Fo

55–70

)

than expected from the olivine composition (Fo

)

in the ﬁner grained, chilled margin, presumed to

represent the quenched parent magma. Moreover,

olivines in this chilled margin are embayed and were

presumably unstable when the magma solidiﬁed.

4. Local internal “chilled” contacts within the sill and

signiﬁcant reversals or discontinuities in elemental

trends in vertical sections (Figure 12.15) suggest

multiple injection of additional magma into the

fractionating magma system. One of these fresh

draughts of magma is hypothesized to have carried

a large proportion of suspended olivine that, dur-

ing emplacement into the ﬂat sheet, was smeared

out as the olivine-rich basal layer.

The composite intrusion hypothesis seems intu-

itively probable because during the calculated several-

hundred-year solidiﬁcation time of the sill (Section

8.4.1) new ascending batches of magma feeding the

overlying ﬂood lavas may have intersected the horizon-

tally widespread sill.

Despite the apparent magma replenishment and

mixing, the Palisades magma has an overall vertical

elemental variation that is readily explained by frac-

tionation of plagioclase and clinopyroxene, the major

minerals present throughout the sill. But the exact

physical mechanism of the fractionation is uncertain.

Because of possible restrictions imposed by the plas-

tic yield strength of the melt (Section 8.2.2), gravita-

tional settling of independent crystals of plagioclase

and clinopyroxene through the magma may have been

limited. But if this restriction was valid, how does one

explain the highly asymmetric position of the late felsic

differentiates near the top of the sill (Figure 12.14)?

This seems to imply that gravity acted in some way.

Perhaps large clumps of gravitationally unstable crys-

tals growing near the roof fell to the ﬂoor of the crys-

tallizing sill, augmenting the mat of crystals growing

there. Buoyant residual melt could have escaped from

the interstices of the mat of plagioclase and pyroxene

grains, as a result of compaction under its own weight

(ﬁlter pressing), or convective melt fractionation, or

both. Displaced residual melt apparently accumulated

near the roof of the sill, beneath downward crystallized

magma, in the sandwich horizon, there forming the

most-evolved Fe-rich granophyre.

330 Igneous and Metamorphic Petrology

Country

rock

Country

rock

Chilled margin

Pegmatitic zone

Olivine segregation

layer

Chilled margin

Increasing (Na + K)/Ca in feldspars

Increasing Fe/Mg in mafic minerals

12.14 Schematic cross section through an idealized diabase-

granophyre sill. Lengthening dash lines indicate increasing

grain size into pegmatitic granophyre zone.

Differentiation of Magmas

331

12.4.2 Layered Intrusions

Layered maﬁc intrusions have attracted the attention

of petrologists for many decades (Cawthorn, 1996).

Most of the large intrusions are of Precambrian age

(Table 12.5), and the largest of these is the colossal

Bushveld complex in South Africa which crops out

over an area of about 65,000 km

; the estimated

0.5-million km

of basalt magma that formed this 2-Ga

intrusion was no doubt related to the head of a decom-

pressing mantle plume. The volume of smaller intru-

sions, such as the early Tertiary Skaergaard intrusion in

southeast Greenland (170 km

), is comparable to that

of some ﬂoods of plateau-forming basalt lava (Section

10.2.2).

Layering and Cumulus Fabric. Layering in these intru-

sions (Figures 7.43–7.45) is as pervasive and distinctive

as stratiﬁcation in sedimentary sequences. Single layers

range from millimeters to hundreds of meters in thick-

ness and from meters to tens, and even hundreds, of

kilometers in lateral extent. In the Bushveld complex,

the Merensky Reef—a pyroxenitic layer 1 to 5 m thick

that is the chief global resource of platinum—extends

along strike for nearly 150 km in the eastern part of the

complex and 190 km along strike in the western part,

100

200

300

0.2 0.4 0.6 0.8 0 2 4 6 8

Mg/(Mg + Fe)

Height above base (m)

Th (ppm)

Sandwich

Horizon

Olivine-rich layer

12.15 Chemical variation in rocks of the Palisades sill, New Jersey. Discontinuities in Mg/(Mg  Fe) ratio at arrows is consistent with inﬂux of

new, less differentiated magma into the sill. (Data from Shirley 1987.)

Table 12.5. Large Layered Intrusions

AME

, L

OCATION

EMARKS

Skaergaard, Greenland 55.7 55 km

; 3.5 km thick (see Figure 13.23)

Rum, Scotland 61–58 115 km

; 2 km thick (see Figure 13.23)

Duluth, Minnesota 1100 More than a dozen layered intrusions exposed over 5000 km

Muskox, Northwest Territories, Canada 1270 Canoe-shaped, 11  150 km (Figure 12.17); associated with the Mackenzie dike

swarm (Figure 9.7a)

Sudbury, Ontario 1850 1100 km

Bushveld, South Africa 2050 Largest on Earth: 65,000 km

and 7–9 km thick; exceptional lateral

continuity of individual layers

Jimberlana, Western Australia 2370 End-to-end canoe-shaped complexes averaging 1.5 km wide and 180 km long

underlain by connecting dike

Great Dyke, Zimbabwe 2460 Four end-to-end canoe-shaped layered complexes 4–11 km wide and 550 km long

underlain by a connecting feeder dike

Stillwater, Montana 2700 8  55 km and ~7 km thick; only basal ultramaﬁc cumulates and overlying

gabbroic and anorthositic rocks preserved

Windimurra, Western Australia 2800 35  85 km in area and 5–13 km thick; only slight differentiation

Data from Hatton and von Gruenewaldt (1990) and Cawthorn (1996).

about 300 km distant. Layering is most conspicuously

deﬁned by variations in relative proportions of miner-

als, deﬁning modal layering. Gradational variations

within a single layer, from top to bottom, may be obvi-

ous, forming graded modal layers. Some layers are

graded in grain size. Rhythmic layering, deﬁned as a se-

quence of recurring similar layers, is common.

Another aspect of the fabric of layered intrusions

parallels that of clastic sedimentary deposits. In sand-

stones, a textural distinction between accumulated

sand grains and secondary cement surrounding them

is generally obvious. In cumulate rocks of layered in-

trusions, a similar distinction may be obvious between

collected early formed cumulus grains and intercu-

mulus mineral matter that formed later around them.

This magmatic “cement” ﬁlling space between cumu-

lus grains was precipitated from the intercumulus

melt initially entrapped during accumulation of cumu-

lus grains or from a later, modiﬁed melt percolating

through the cumulate pile (Figure 12.16). On the ﬂoors

of magmatic intrusions, cementation and overgrowth

compete with compaction in “densifying” the cumulus

aggregate, eliminating the intercumulus melt (Hunter,

1996). For a recent discussion of the validity of the cu-

mulate paradigm see Morse (1998).

Cumulates on ﬂoors of layered intrusions do not

necessarily reﬂect gravitational settling, as once widely

believed. For example, the Archean Stillwater complex

of Montana has layers of cumulus chromite grains

(about 0.25-mm diameter, density 4.4 g/cm

) near the

base that grade upward into larger, less dense cumulus

olivine grains (0.7–3.0 mm, 3.3 g/cm

). As settling ve-

locity, according to Stokes’s law (Section 8.3.3), is pro-

portional to the square of the particle size but only to

the ﬁrst power of the density contrast with the melt, the

larger olivines should have settled faster than the

smaller but more dense chromites. The modal grading,

therefore, appears to be upside down. Jackson (1961)

concluded from this and other observations that in

situ bottom crystallization, not gravitational settling,

created this layering. Kinetic factors may also play a

role in development of modal layering (Section 7.9.2;

Figure 7.46).

Evidence for upward migration of melt through the

cumulus ﬂoor pile can be seen in mineral compositions

in the Muskox intrusion (Figure 12.17). This layered

intrusion, like the Stillwater, Duluth, Bushveld, and

Great Dyke intrusions (Table 12.5), formed from

episodic replenishment of basaltic magma during its

crystallization. This open-system recharge is usually ev-

ident in upward reversals in compositional variation

(Figure 12.18) and is logical in view of the multiple

basalt dikes, sills, and lava ﬂows spatially and tempo-

rally associated with the intrusions. Discontinuities in

the chemical composition of cumulus chromite and

olivine occur up section from the modal break marking

332 Igneous and Metamorphic Petrology

12.16 Evolution of cumulus fabric. The three postcumulus processes

illustrated in (b–d) are ideal end members; real cumulate rocks

generally form by some combination of these processes. See

Hunter (1996) and Morse (1998) for additional discussion of

cumulus textures. (a) Original cumulus grains of olivine (bold

relief ), pyroxene (with cleavage), and chromite (small black)

in a melt (stippled). (b) Postcumulus cementation and devel-

opment of poikilitic texture. The intercumulus melt surround-

ing the cumulus grains in (a) crystallized as large plagioclase

grains, ﬁlling the intercumulate pore volume. Extraneous ions

in the melt (Fe, Mg, Ti, etc.) not required by the growing pla-

gioclase must be transferrred out of this space, perhaps by

ionic diffusion. (c) Postcumulus overgrowth and development

of equilibrated texture with approximately 120 triple-grain-

boundary junctions. Intercumulus melt surrounding the cu-

mulus grains (shaded areas enclosed by dashed outlines) in (a)

crystallized as secondary enlargements on the cumulus olivines

and pyroxenes until all pore space was eliminated. Plagioclase

either was unstable or did not nucleate. Monomineralic

dunites, pyroxenites, and anorthosites can originate by sec-

ondary enlargement of accumulations of olivines, pyroxenes,

and plagioclases, respectively. Ions not needed during enlarge-

ment are transferred out of the local interstitial volume by

some means. (d) Postcumulus reaction replacement combined

with overgrowth. Intercumulus melt reacted with cumulus

grains (shaded areas enclosed by dashed outlines) in (a), partly

consuming olivines. Simultaneous secondary enlargement of

the cumulus pyroxenes eliminated all pore spaces.

Differentiation of Magmas

333

the cyclic unit boundary. Irvine (1980) explains this

discrepancy between modal and chemical breaks on

the basis of compaction of the cumulate pile and up-

ward transfer of the intercumulus melt out of it.

Migration of melts having lower Mg/(Mg  Fe) and

Ni/(Ni  Mg  Fe) ratios from the underlying

cumulate unit into the overlying cumulus olivines and

chromites causes them to reequilibrate via reaction re-

lations. The wave of reequilibration for Mg and Fe ex-

tends farther into the overlying cyclic unit than does

that of Ni because the melt contains about a third as

much Mg  Fe as the cumulus crystals but only about

a tenth as much Ni. Accordingly, adjustments in Mg

and Fe are more pronounced.

In large, slowly cooled intrusions, a substantial frac-

tion of the intercumulus melt can be transferred via

ﬁlter pressing or convective melt fractionation from

the compacting ﬂoor cumulates back into the un-

crystallized central part of the intrusion. This is an ef-

fective crystal-melt fractionation that has in the past

been attributed to crystal settling through the whole

chamber.

Dunite

Feeder dike

DDH

Pyroxenite

Granophyre,

roof-rock

breccia

Gabbro

Lake

Marginal zone of gabbro

and peridotite

2 km

12.17 Schematic block diagram of the central segment of the Muskox layered intrusion, Northwest Territories, Canada. Regional tilting to the

north of about 5 exposes most of this 1270-Ma trough-shaped intrusion. Feeder dike 150–500 m wide of picrite and gabbro extends

tens of kilometers southward. The intrusion is as much as 11 km wide and at least 150 km long before disappearing under covering roof

rock. Above the troughlike marginal zone of gabbro and peridotite is a 1800-m-thick series of 42 mappable layers individually ranging

in thickness from 3 to 350 m that are predominantly dunite with a lesser proportion of pyroxenite. This layered series is overlain by gab-

bro and then granophyre formed from the evolved residual magma. DDH, diamond drill hole, whose core has been investigated in de-

tail (Figure 12.18). (Redrawn from Irvine, 1980.)

40 60

85 90 0.2 0.4 0.6 0.8 0 2 4

700

500

600

Depth (m)

100 Mg 100 Mg 100 Ni

Modal %

chromite

in rock

Cyclic unit

Mg + Fe

in chromite

Mg + Fe

in olivine

Ni + Mg + Fe

in olivine

12.18 Analytical data on a part of the core from a drill hole into the Muskox intrusion (DDH in Figure 12.17). Boundaries of a cyclic unit are

deﬁned by discontinuities in modal proportion of cumulus chromite grains in right-hand panel. In the Muskox, each of the 25 cyclic units

represents a magma recharge into the intrusion. Discontinuities in chemical composition of cumulus olivine and chromite occur above

the modal boundaries. The top of unit 3 is chromite-free olivine clinopyroxenite; all other rock is chromite-bearing dunite. (Redrawn

from Irvine, 1980.)

334 Igneous and Metamorphic Petrology

Skaergaard Intrusion. No discussion of layered intru-

sions would be complete without mention of the in-

tensely studied Skaergaard intrusion, located in the re-

mote southeast coast of Greenland just above the

Arctic Circle. This intrusion and related dikes and

overlying thick lava ﬂows formed during the early

Tertiary opening of the north Atlantic (see Figure

13.23). Ironically, the more that has been learned, from

the earliest studies by Wager and Deer (1939) to later

studies by A. R. McBirney and associates (e.g., 1996),

the more controversy seems to develop (Irvine et al.,

1998).

In contrast to many large differentiated sills and

larger intrusions, the exposed part of the Skaergaard

intrusion (Figures 12.19 and 12.20) reveals no un-

equivocal indication of multiple injections of magma

that mixed with previously fractionated magma. For

example, compatible element trends, such as for Ni,

are continuous without major breaks (Figure 12.20).

Therefore, the exposed part of the Skaergaard—above

a hypothetical Hidden Zone—may have formed by

fractionation in a closed magma system. Unfortunately,

no unaltered “chilled” margin rock has been found

that might represent the pristine, parent magma from

which the intrusion fractionated.

Crystallization in the Skaergaard was dominantly

upward from the ﬂoor, producing an uncertain amount

of the Hidden Zone plus the exposed Layered Series of

about 2500 m. This is much greater than the thickness

of the downward crystallized Upper Border Series.

These two series meet at the Sandwich Horizon, which

represents the last-crystallized part of the intrusion,

and are enveloped within the still thinner Marginal

Border Series, the part of the intrusion solidifying in-

ward from the sidewalls. That this is indeed the pattern

of inward crystallization of the intrusion is shown by

variations in whole-rock chemical and mineral compo-

sitions and phase layering. Phase layering is the abrupt

appearance or disappearance of a particular mineral,

such as the vertical disappearance of olivine deﬁning

the lower boundary of the Middle Zone of the Layered

Series and its reappearance in the Upper Border Series.

Cryptic layering reﬂects systematic changes in the

chemical composition of cumulus minerals, shown in

the middle panel of Figure 12.20. Plagioclase becomes

more sodic, to An

in the Sandwich Horizon, from as

much as An

in the uppermost part of the intrusion

and An

in the lowest exposed part. Olivine ranges

from Fo

to pure fayalite Fo

and clinopyroxene to

Mg-free hedenbergite in the Sandwich Horizon. These

cryptic variations are more or less symmetric inward to

the Sandwich Horizon, as are whole-rock chemical

compositions (Figure 12.21).

The extreme enrichment in Fe and limited late en-

richment in silica and alkalies have made the Skaer-

gaard trend a reference standard for fractionation of

tholeiitic basalt magma (Table 12.6; Figures 12.21 and

12.22). However, the physical mechanism of fractiona-

tion that operated in the Skaergaard is still controver-

sial. Two competing hypotheses are brieﬂy summarized

in the following two paragraphs (see also Special Inter-

est Box 8.1). The debate is yet another example that ex-

perienced petrologists can arrive at markedly contrast-

ing opinions about the same rock.

Upper Border Series

Sandwich

Horizon

Upper

Zone

UZc

UZb

UZa

LZc

LZb

LZa

Hidden

zone

gneiss

Precambrian

Cretaceous

sandstone

Eocene

basalt

East West

Layered Series

Middle Zone

Lower

Zone

12.19 Schematic cross section through the Skaergaard intrusion, Greenland. Emplaced at 55.7 Ma, the intrusion measures about km

on the ground and, because of a regional northward tilt of 15–25, the 1200 m of glacially carved topographic relief exposes an almost

continuous 3500-m stratigraphic section through the body. Conﬁguration of the deeper unexposed part of the intrusion, the Hidden

Zone, is based on geophysical measurements. The intrusion can be conveniently divided into four parts (see also Figure 12.20): (1) Mar-

ginal Border Series (MBS) consisting of an inner Layered Series and an outer Tranquil Zone; (2) Thick Layered Series (shaded), consist-

ing, in ascending order, of the Lower Zone, which is divided into three parts (LZa, LZb, LZc); the Middle Zone (MZ); and the Upper

Zone, which is divided into three parts (UZa, UZb, UZc); (3) Upper Border Series (UBS); (4) last crystallizing Sandwich Horizon. (Re-

drawn from McBirney, 1993.)

6  11

Differentiation of Magmas

335

TRACE ELEMENTS

CUMULUS

MINERALS

INTRUSION

+ Olivine

+ Plagioclase

+ Clinopyroxene

+ Fe-Ti oxide

− Olivine

+ Olivine

+ Apatite

Sandwich

Horizon

UZc

UZb

UZa

+ Bustamite

− Clinopyroxene

+ Apatite

Trough

layers

+ Olivine

LZc

LZb

− Olivine

+ Fe-Ti oxide

+ Clinopyroxene

LZa

Hidden

Zone

+ Olivine

+ Plagioclase

± Clinopyroxene

0.5

1.0

1.5

2.0

2.5

3.0

3.5

Tranquil

Zone

Layered

MBS

Upper Border

Series

Plagioclase An

Olivine Fo

30 40

ppm

Ca-rich clinopyroxene En

12.20 Compositional relations and layering in the Skaergaard intrusion. Left-hand panel is a more detailed diagram of Figure 12.19 indicating

phase layering, which delineates the successive layer units (e.g., LZb, LZc, MZ). Also indicated are correlations of crystallization stages

in the Layered, Marginal Border, and Upper Border Series. For example, after early crystallization of the texturally isotropic Tranquil

Zone against the wall rock, an upper layer of the Upper Border Series crystallized concurrently with an outer layer of the Marginal Bor-

der Series (MBS) and the lowermost layer of the Layered Series, LZa. Modal graded layering is represented schematically by dotted lines.

Note cross-bedding in modal layers at the junction of the subhorizontal Layered Series and the steeply inclined Marginal Border Series

and trough layers in the UZa layer. P, schematically indicated patchy and strata-bound masses of coarse-grained maﬁc pegmatite; SA,

schematically indicated masses of secondary anorthosite; T and A, respectively, schematically indicated blocks of troctolite (olivine  pla-

gioclase rock) and anorthosite stoped off the roof of the intrusion. There are many more of SA, T, and A than shown. (Redrawn from

McBirney, 1996; Irvine et al., 1998.)

1.0

2.0

3.0

15 20 25 100 100030/10

Total Fe as FeO (wt. %) Ba (ppm)

Stratigraphic height (km)

in Layered Series

Upper Border Series UBS

Marginal Border Series MBS

Layered Series LS

Sandwich

UBS

MBS

LZa

LZb

LZc

UZa

UZb

UZc

Horizon

MBS UBS

12.21 Chemical variations in the Skaergaard intrusion. Elements in the Upper Border Series and Marginal Border Series are plotted with re-

spect to equivalent unit in Layered Series. (Redrawn from McBirney, 1996.)

336 Igneous and Metamorphic Petrology

In their classic study, Wager and Deer (1939) postu-

lated that convection currents were responsible for the

rhythmic modal layering, trough layering, igneous lam-

ination, and local lenticular and cross bedding and

slump structures. They envisaged the Layered Series as

having formed by gravitational settling and sorting of

crystals from convection currents that descended from

the walls and swept inward across the cumulate ﬂoor.

Depending on the vigor of convection, modally graded

layers would be thin or thick or rhythmic. Irvine et al.

(1998) reafﬁrmed and reﬁned the convection-driven

sedimentation hypothesis, documenting in consider-

able detail what they believe to be supporting evidence.

Beginning in the late 1970s, magmatic sedimenta-

tion via convection in the Skaergaard was rejected by

McBirney and coworkers (1996), largely because pla-

gioclases dominating the cumulate fabric should have

ﬂoated, not sunk, in the increasingly denser Fe-rich

melts. They appealed, instead, to fractional crystalliza-

tion driven by compaction and convective melt frac-

tionation within the pile of cumulate crystals on the in-

trusion ﬂoor. Because this fractionation is driven by

gravity, there ought to be different fractionation effects

in the Layered Series that accumulated on the ﬂoor ver-

sus in the Upper Border Series that was created near

the roof. And both might differ from the Marginal Bor-

der Series. Figure 12.21 shows such differences. Upper

Border Series rocks are depleted in Fe relative to Lay-

ered Series rocks because negatively buoyant Fe-rich

residual melts, perhaps carrying some suspended crys-

tals, drained out of the roof region and ponded in ﬂoor

cumulates. Marginal Border Series rocks are slightly

less Fe-rich than Layered Series rocks, again reﬂecting

draining out of Fe-rich melts, but not as efﬁciently as

from the roof. The greater concentration of incompat-

ible elements, such as Ba, in Upper Border Series rocks

than in Layered Series is thought to reﬂect even later-

stage, postpeak Fe-enrichment inﬁltration of buoyant

incompatible-element-enriched residual melts derived

from underlying magma. These late buoyant melts were

relatively enriched in silica and alkalies and crystallized

largely as scattered patches and dikes of granophyric

rock throughout the intrusion but especially in the Up-

per Border Series. Some late maﬁc granophyres may

have formed from immiscible melts. Other granophyres

may be fused blocks of quartzofeldspathic gneiss coun-

try rock.

Magmatic Replacement in Large Layered Intrusions.

Meter-scale, patchy masses of rock in the Skaergaard

and other intrusions appear to have formed by re-

placement of original rock, facilitated by local con-

centrations of water. These masses include secondary

anorthosite and maﬁc pegmatite that are commonly ad-

jacent to one another (SA and P, respectively, in the left

panel of Figure 12.20). The origin of the anorthosite

(90% plagioclase) by secondary, volume-for-volume

metasomatic replacement is indicated by undisturbed

continuity of layering from adjacent rock through the

anorthosite. Irvine et al. (1998) suggest that locally

higher concentrations of water in the intercumulus

melt shift the olivine-pyroxene-plagioclase cotectic to-

ward plagioclase, locally precipitating more of this

phase and displacing maﬁc minerals into peripheral ar-

eas. Unusually coarse-textured gabbroic rock, or maﬁc

pegmatite, some having additional biotite and amphi-

Table 12.6. Average Compositions of the Units in the Layered Series, Sandwich Horizon (SH), and Granophyre,

Skaergaard Intrusion

MZ UZ

SH G

RANO

SiO

48.12 48.84 41.10 42.79 43.07 41.78 46.00 49.43 60.23

TiO

1.35 1.44 6.92 6.79 5.67 4.06 2.63 2.23 1.18

16.81 12.55 11.02 11.53 11.17 9.51 7.86 7.93 11.29

FeOt 11.13 12.84 21.10 20.00 22.52 26.64 28.67 27.87 14.08

MnO 0.16 0.21 0.26 0.26 0.31 0.41 0.65 0.25 0.24

MgO 9.42 10.13 7.61 6.24 5.62 3.41 0.38 0.09 0.51

CaO 10.11 11.57 9.77 9.87 8.62 9.36 10.14 8.23 5.11

O 2.52 2.13 1.97 2.23 2.55 2.59 2.42 2.72 3.92

O 0.27 0.20 0.20 0.21 0.26 0.36 0.41 0.72 1.94

0.11 0.09 0.05 0.08 0.22 1.88 0.84 0.53 0.27

Sr 285 219 199 218 233 244 263 450

Zr 93 81 72 80 95 97 135 324

La 5.62 4.68 3.84 3.09 3.49 14.57 22.68 57.8

Sm 2.47 3.08 2.34 2.10 2.10 10.00 15.27 34.5

Total oxides are 100.00.

Data from McBirney (1996).