JOURNAL OF PETROLOGY
VOLUME 37
NUMBER4
EAGES 837-858
1996
M. J. TOPLIS* AND M. R. CARROLL
DEPARTMENT OF GEOLOGY, UNIVERSITY OF BRISTOL, BRISTOL BS8 1RJ, UK
Differentiation of Ferro-Basaltic Magmas
under Conditions Open and Closed to
Oxygen: Implications for the Skaergaard
Intrusion and Other Natural Systems
Because processes such as fractional crystallization and crys- as Iceland and Hawaii evolve under conditions open to oxygen,
tallization under conditions closed to oxygen are difficult to whereas evidence from plutonic environments (e.g. Skaergaard
simulate in the laboratory there is a need for quantitative and Kiglapait layered intrusions) suggests that they evolved
models of magma crystallization behaviour. Comparison of under conditions more closed to oxygen. The compositional
experimental data on an iron-rich basaltic composition with evolution of the melt phase in volcanic and plutonic systems may
predictions of the MELTS free energy minimization algorithm therefore be different, although the results of this study suggest
shows that although liquidus temperatures and silicate mineral that magnetite saturation will limit Fe enrichment in all
equilibria are predicted relatively well, the saturation of Fe-Ti environments to <20wt% FeO*, consistent with enrichments
oxides is not We have used the same experimental data to con- reportedfor volcanic glasses.
struct an alternative crystallization model based on known
equilibrium phase relations, mineral-melt partitioning of major KEY WORDS: Skaergaard; ferro-basalt; iron enrichment; oxygen
elements, and mass balance constraints. The model is used to jugaaty
explore the consequences of equilibrium and fractional crystallization in systems open and closed to oxygen. Liquid lines of
descent for perfect equilibrium and perfect fractional crystal- INTRODUCTION
lization are predicted to be very similar. In a system open to Iron enrichment during the early stages of differoxygen the model predicts that magnetite saturation leads to entiation of dry subalkaline basalts at low pressure is
strongly decreasing iron and increasing silica contents of resid-well established, but the extent of this enrichment is
ual liquids, whereas systems closed to oxygen crystallize less still debated (Hunter & Sparks, 1987, 1990; Morse,
abundant magnetite, leading to a less pronounced iron depletion 1990; Brooks & Nielsen, 1990; McBirney & Naslund,
in the liquid. Predicted bulk solid compositions and variations 1990). Tholeiitic lavas typically show enrichments
offo, with falling temperature agree well with those observed orup to ~ 15 wt% FeO* (where FeO* indicates total
inferred from the cumulates of the Skaergaard intrusion, but Fe as FeO), and more rarely as high as ~19 wt %
none of the predicted liquid lines of descent are consistent with FeO* (e.g. Brooks et al., 1991). In all cases, iron
the extreme iron enrichment proposedfor this intrusion based onenrichment of the melt phase is followed at lower
mass balance calculations. Compositionalfactors such as water temperature by iron depletion and silica enrichment
and phosphorus are not thought to be the source of the dis- which may be related to the precipitation of Fe-Ti
crepancy as the cumulates of the Basistoppen sill (which closely oxides. However, evidence from plutonic rocks (e.g.
resemble those of Skaergaard) may be used to calculate a liquid the Skaergaard intrusion: Wager, 1960; Wager &
line of descent in agreement with that predicted by the modelfor Brown, 1967; and the Kiglapait intrusion: Morse,
fractional crystallization closed to oxygen. A comparison of the 1981) has been used to postulate residual melt compredicted T-io, paths and liquid lines of descent with those positions reaching 30 wt% FeO*, without appreinferredfrom natural systems suggests that volcanic centres suchciable increase of melt silica content (McBirney &
* Corresponding author.
Present address: Bayerisches Geoinstitut, UniversitSt Bayreuth,
D-954+0 Bayreuth, Germany. Telephone: + 49-93-5537ia
Fax + 49-921-553769. e-mail; [email protected]
© Oxford University Pros 1996
JOURNAL OF PETROLOGY
VOLUME 37
NUMBER 4
AUGUST 1996
Naslund, 1990). Furthermore, iron enrichment is therefore important to assess whether MELTS satisproposed to have continued in these plutonic envir- factorily predicts ferro-basaltic mineral-melt equionments, even after the appearance of' cumulus libria and liquid lines of descent where experimental
magnetite. Hunter & Sparks (1987, 1990) suggested data exist (e.g. Toplis & Carroll, 1995).
that iron enrichment to levels >20 wt% FeO* did
Toplis & Carroll (1995) described the experinot occur and is an artefact of the assumptions and mental equilibrium crystallization of a synthetic,
uncertainties inherent in the calculation of melt nine-component ferro-basalt (sample SCI, composicompositions from plutonic rocks. They proposed tion shown in Table Al, in the Appendix) to ~70
that magnetite crystallization marked the end of wt % crystallized, in conditions open to oxygen, over
absolute iron enrichment in the Skaergaard magma, a range of_/o, from one log unit above to two log
and that crystallization led to the formation of silicic units below the fayalite-magnetite-quartz (FMQ,)
differentiates as seen in many other tholeiitic vol- buffer. The composition SCI was based on that of
canic provinces; this proposal has been contested by dyke C from the base of the Skaergaard intrusion
many Skaergaard workers (Morse, 1990; Brooks & reported by Brooks & Nielsen (1978), who proposed
Nielsen, 1990; McBirney & Naslund, 1990).
it as a possible parental composition to the exposed
The lack of correspondence between the observed portion of the Skaergaard. The experimentally
magmatic liquid compositions from volcanic envir- determined phase relations are consistent with the
onments and those inferred for plutonic environ- cumulate mineralogy of the lower and middle zones
ments may be due to differences in the factors and of the intrusion (Ol + Plag, LZa; Ol + Plag + Cpx,
processes controlling differentiation (Morse, 1990). LZb; Ol + Plag + Cpx + Fe-Ti oxides, LZc; and
Indeed, experimental studies of iron-bearing haplo- Plag + Cpx + Fe-Ti oxides-Ol, MZ), making SCI
basaltic compositions by Osborn (1959) and Presnall particularly relevant to discussion of the evolution of
(1966) demonstrated significant differences between the Skaergaard intrusion, around which much recent
systems evolving closed to oxygen (at constant total controversy concerning the low-pressure differcomposition) and those evolving open to oxygen. For entiation of basalts has centred (Hunter & Sparks,
crystallization closed to oxygen, the residual melts 1987, 1990; Morse, 1990; Brooks & Nielsen, 1990;
showed continued iron enrichment after the McBirney & Naslund, 1990). The results are not,
appearance of magnetite, whereas experiments open however, limited to the case of the Skaergaard
to oxygen did not produce such iron-rich liquids. intrusion, as SCI is typical of many ferro-basalts.
However, the studied systems did not contain sodium The experimentally determined values of crystalor titanium, both of which are important con- lizing proportions of liquidus phases, and mineralstituents of natural basaltic melts and crystalline melt partitioning of major elements have been
phases, with Ti being particularly important because extensively compared with available literature data
of its role in forming Fe—Ti oxides. Phase equi- to put them in a broader context in the study by
librium experiments provide valuable constraints on Toplis & Carroll (1995), to which the reader is
the petrogcnesis of natural basaltic systems, but pro- referred for further information. These experimental
cesses such as crystallization under conditions closed results extend our knowledge of mineral-melt equito oxygen, or perfect fractional crystallization are libria in iron-rich systems, and have been used to
difficult or impossible to simulate in the laboratory, construct a crystallization model for ferro-basalt SCI
and thus understanding such processes requires based on known phase relations, mineral—melt parquantitative models of magma crystallization beha- titioning of major elements and mass balance conviour. Previous models have been based on experi- straints, which is used to explore the consequences of
mentally determined partition coefficients (e.g. equilibrium and fractional crystallization under
Langmuir & Hanson, 1981; Nielsen, 1988, 1990; conditions open and closed to oxygen.
Weaver & Langmuir, 1990), or Gibbs free energy
minimization (e.g. MELTS; Ghiorso et al., 1983;
Ghiorso & Carmichael, 1985; Ghiorso & Sack,
1995). The most recent version of MELTS (Ghiorso PREDICTION OF FERRO& Sack, 1995) is constrained by data from >1500 BASALTIC DIFFERENTIATION
experiments, and permits the calculation of crystalPRODUCTS USING MELTS
lization paths for a wide variety of crystallization
processes, including fractional crystallization, and The latest version of the Gibbs free energy minisystems closed to oxygen. However, few of the mization program MELTS (Ghiorso & Sack, 1995)
experimental constraints used in the construction of offers the possibility to model a wide range of
the model are from ferro-basaltic systems, and it is igneous processes. However, ferro-basaltic systems
are poorly represented in the experimental data base
838
TOPLIS AND CARROLL
FERRO-BASALTIC MAGMA DIFFERENTIATION
stabilities of the Fe-Ti oxides. Magnetite is predicted
to be the stable Fe-Ti oxide at low _/o,» a n d ilmenite
the stable phase at high _/o,> whereas the experimental results of both Toplis & Carroll (1995), and
Snyder et al. (1993) show ilmenite to be stabilized by
decreasing/ Oj , and magnetite by increasing/ Ot .
Large discrepancies are abo apparent between
observed and predicted liquid lines of descent along
the FMQ, buffer (Fig. 2). MELTS predicts much
greater variation in liquid composition than is
experimentally observed, particularly for the components FeO* and SiO2. Melt compositions are predicted to reach 28 wt % FcO* and 38 wt % SiO2 just
before magnetite saturation, compared with 17 wt %
FeO* and 50 wt % SiO2 observed experimentally. In
addition, MELTS predicts a very large discontinuous increase in the percent crystallized per °C
at magnetite saturation (resulting in the drastic
changes in melt composition observed in Fig. 2),
which is not supported by the experimental results.
Although the MELTS program may be used successfully to estimate liquidus temperatures and
silicate phase equilibria, it is clear that the saturation
of Fe-Ti oxides is not well constrained, and that
erroneous liquid lines of descent are predicted.
Therefore, it would appear that for ferro-basaltic
crystallization processes, where the saturation of FeTi oxides plays an important role, an alternative
approach to the MELTS model may, for the time
being, give more instructive insights. We do not
propose that the model we present in this paper is
universally applicable to the crystallization of
basaltic magmas, but we do believe that the general
behaviour observed for composition SCI may be
used to explore the consequences of crystallization
under conditions not easily duplicated in the
laboratory, or modelled by currently available calculation methods, especially those conditions where
crystallization of Fe-Ti oxide minerals exert considerable influence on basaltic differentiation paths
and crystalline products.
for the MELTS model, and care must be taken when
applying it outside of its calibration range.
MELTS has been used to calculate the equilibrium crystallization path of composition SCI
(Toplis & Carroll, 1995) over the experimentally
studied range of oxygen fugacities, to compare the
predictions with the experimental results (see Figs 1
and 2). MELTS predicts the liquidus temperature of
SCI to within 25°C, with plagioclase and olivine
correctly predicted as the liquidus phases. However,
significant discrepancies are apparent. A low-Ca
pyroxene is predicted to be stable, but was not
observed in any of the experiments (which included
reversals within the predicted pigeonite stability
field). However, low-Ca pyroxene was observed in
experiments carried out on a phosphorus-bearing
daughter liquid of SCI (Toplis et ai., 1994), suggesting that the phosphorus-free liquids modelled
here are indeed close to being saturated in low-Ca
pyroxene. The greatest discrepancy between the
modelled and experimental results is the predicted
1040
1060
1100
1120
1140
1160
1180
1200 1220
Temperature C O
b) is
llm,LoCaCpx
1.0
._
11
0.5
lag
Mt
\
a
f 0.0
J -0J
/f\
\
< -13
MODELLING OF FERROBASALTIC DIFFERENTIATION
Ol
HiCa Cpx /
•2.0
•Zi
1040
1060
1080
1100
1120
1140
1160
I ISO
1200 1220
Temperature C O
Fig. 1. (a) Experimentally determined phase diagram of composition SCI (Toplii & Carroll, 1995) a» a function of/ O f . (b)
Predicted phase diagram of composition SCI using MELTS
(Ghiorso & Sack, 1995).
An alternative to the Gibbs free energy minimization
approach of the MELTS model is to use experimentally constrained distribution coefficients, such as the
models of Nielsen (1988, 1990) and Langmuir
(Langmuir & Hanson, 1981; Weaver & Langmuir,
1990). A potential problem in using such existing
models for ferro-basaltic systems is that mineralliquid exchange coefficients may be different in ironrich systems, and the saturation of the iron—titanium
oxides as a function of/ 0 , is poorly constrained.
839
JOURNAL OF PETROLOGY
a)
VOLUME 37
NUMBER 4
b)
30
'
*
1
1
20 -
%
FeO*
\
SiO2
."2 60
\
3
V
\
•^55
s
>
•
\
15
•
1
!„
1080
1100
1140
1160
35
1080
1180
1100
Temperature (°C)
c) 12
I
I
'
M.
,
'
'
'
1120
1140 1160
1180
d)
1
MgO
t
i
9
O
8
7
•
Temperature (°C)
CaO
11
•3 10
cr
u
"
•
40
GhiorsoA Sack, 1995
Toplu & Carroll. 1995
1120
•
45
" " " • •
'l
•
9
•
•
•
•a
AUGUST 1996
00
3 - *
6
1080
I
1100
1120
.
I
.
I
2
1080
,
1140 1160
1180
Temperature (°C)
1100
1120
1140
1160
1180
Temperature (°C)
Fig. 2. Experimentally determined liquid evolution as a function of temperature ((ymboli; Toplij & Carroll, 1995), and predictions of
MELTS (dashed lines; Ghiorso 4 Sack, 1995) for equilibrium crystallixation of SCI along the FMQ. buffer.
Exchange coefficients between olivine, plagioclasc,
pyroxene and iron-rich liquids, cotectic proportions
of stable phase assemblages and simple criteria for
predicting Fe-Ti oxide saturation for the composition SCI have been provided by Toplis & Carroll
(1995). Our objective is to parameterize the equilibrium, open to oxygen crystallization behaviour of
composition SCI, and then to use the parameterization to explore the consequences of crystallization conditions which are not amenable to
experimental study (e.g. systems closed to oxygen,
perfect fractional crystallization). First we discuss
the factors affecting magmatic differentiation, and
the differences between the end-member modes of
crystal lization.
E q u i l i b r i u m vs fractional crystallization
We consider the case of a simplified magmatic
system which, at a given temperature, may consist
of: (1) a silicate liquid; (2) 'coexisting' crystalline
solids which may react with the liquid; (3) 'isolated'
solids which may not react with the liquid; in this
analysis we do not consider a vapour or fluid, or
sulphide liquid. During perfect equilibrium crystallization the solids (crystals) are never isolated from
reaction with the liquid, whereas for perfect fractional crystallization all the solids are isolated from
further reaction with liquid as soon as they are
formed (e.g. by crystal settling, zoning, overgrowths). For this simple system the 'bulk composition' (sum of melt and solid phases) remains constant
with falling temperature, and the coexisting solids
(crystalline phases Pi_», e.g. olivine, plagioclase) are
in chemical equilibrium with the coexisting liquid.
The concentration of component t (e.g. SiC>2, FeO,
MgO) in the coexisting bulk solid (CBS) may be
calculated from the composition and relative modal
abundances of the coexisting phases by mass balance
using
840
TOPLIS AND CARROLL
FERRO-BASALnC MAGMA DIFFERENTIATION
Systems open and closed to oxygen
The work of Osborn (1959), and Presnall (1966)
where W) is the weight percent of component i in demonstrates how magma oxidation state can vary
phase _;', and rK is the relative modal proportion of with falling temperature. In a system closed to
phase PH (i.e. pnjY.BpK, where pH is the absolute modal oxygen, the ferric and ferrous iron contents of the
bulk system (melt + total solids) are fixed. The
proportion of phase PH).
Over a given temperature interval (7"° to T*), the ferric—ferrous ratio of the melt increases during crysliquid proportion falls from /° to /* wt%, corre- taUization dominated by pyroxenes, olivine and
sponding to a crystaUization interval of (/° —/*)% plagioclase because ferric iron is incompatible in
crystallized. The absolute modal abundance of an these phases whereas ferrous iron is strongly compaindividual crystalline phase may either increase tible in pyroxene and olivine. Crystallization of
(implying crystallization), or decrease (implying ferric-iron-bearing Fe-Ti oxides will moderate, and
resorption), but the total abundance of the solid may reverse this trend. The changing ferric-ferrous
phases must also change by (/"-/•) wt%. If no ratio of the melt phase implies a change in the oxisolids are removed during crystallization then the dation state of the magma relative to a given solidtotal mass of component t in the system is the same gas buffer (e.g. AFMOj Kilinc et al., 1983; AFMQ,
at temperature T^ and T*, and the mass of i trans- represents the oxidation state relative to the fayaliteferred from the liquid to the coexisting bulk solid magnetite-quartz buffer), and thus crystallization
closed to oxygen results in a uniquely defined T-fot
may be calculated from mass balance constraints:
path which does not parallel the FMO_ buffer. In a
system open to oxygen the total amounts of ferric
and ferrous iron in the bulk system may vary with
(100-/°)
falling temperature (although total iron stays constant). The variation of/o, with falling temperature
Mass of i at temperature 7
• (2) is not uniquely defined, and may increase or decrease
j (loodepending on the quantity of oxygen exchanged. We
will only consider the case of systems open to oxygen
Mass of t at temperature T*
which follow T-fot paths parallel to the FMQ,
Thus, the composition of the residual liquid {W\*) buffer, as this behaviour has been inferred by
may be calculated if the crystallization interval, the numerous studies of coexisting oxides in natural volcomposition of the initial liquid (Wzf), and both sets canic rocks (e.g. Carmichael, 1967). Under these
of coexisting bulk compositions (WCBSO> M^CBS*) a r c conditions the ferric—ferrous ratio of the melt phase
(but not the bulk system) remains approximately
known.
constant
(Sato, 1978; Carmichael & Ghiorso, 1986).
The transfer of chemical components from the
liquid to the bulk solid may be considered to take
place by two distinct mechanisms: (a) through the
formation of (/ — /*) wt% of 'new' crystalline Details of the calculation procedure
material which has crystallized during the specified The model considers sequential crystallization incretemperature interval, and (b) owing to re-equi- ments of 1 wt%. The stable phase assemblage and
libration of pre-existing crystalline solid solutions to relative crystallizing proportions of phases are estiachieve the required new equilibrium composition at mated for each crystallization interval. The
lower temperature (e.g. changes in Xpo of olivine, appearance of plagioclase, olivine and clinoXAB °f plagioclase). If many crystals are present then pyroxene, and their cotectic proportions before Fephase re-equilibration may significantly influence the Ti oxide saturation are those determined expericompositional evolution of the liquid. In the case of mentally for SCI (Toplis & Carroll, 1995; see Tables
perfect fractional crystallization the compositional A2 and A3, below). Magnetite saturation was estievolution of the liquid over any given crystallization mated using the ferric-iron content of the melt, and
interval is controlled only by the composition of the ilmenite saturation estimated using the TiOj content
new solids formed, whereas in the case of equilibrium of the melt [criteria shown in Table A5, below, and
crystallization, both formation of new crystalline described by Toplis & Carroll (1995)]. The promaterial and re-equilibration of existing solids must portion of magnetite is varied so that the ferric-iron
occur, and liquid lines of descent may therefore be content of the coexisting liquid lies on the 'magnetite
expected to diverge with falling temperature, even if saturation' curve described by Toplis & Carroll
the identity and proportions of the phases remain (1995). Compositions of the stable minerals are then
iteratively adjusted to satisfy (a) mass balance
identical.
(1)
841
JOURNAL OF PETROLOGY
VOLUME 37
NUMBER 4
AUGUST 1996
oxide
requirements of the system [equation (2)], and (b)
1
FMQ
crystal—liquid exchange coefficients presented in
n
16 -'
~rr-Feo*
Table A4, below, and discussed by Toplis & Carroll
"
"•"•--_
(1995).
14
Systems evolving parallel to the FMQ, buffer
AI2O3
12
(open to oxygen) have been modelled by fixing the
0
10 :------o"
ferric—ferrous ratio of the melt phase to a constant
o
CaO "
o
value. For systems closed to oxygen the total abund8
MgO"
ance of ferric iron was fixed, and the final oxidation %
6
state of each crystallization interval iteratively
adjusted until the compositions and the proportions
TiO2 4 r + *
•---«--of the crystalline phases satisfy the mass balance and
1
1
1
1
1
7
element partitioning constraints of the model.
1
0
9
0
M
O
O
1
1
1
0
1
1
2
0
1
1
3
0
1
1
40 1150 1160 1170
The model has been used to predict four endTemperature (°C)
member crystallization paths of haplobasalt SCI: (1)
equilibrium crystallization parallel to the FMQ, Fig. 3. Companion of the predicted liquid compositions for equibuffer (open to oxygen); (2) fractional crystallization librium crystallisation of the composition SCI along the FMQ,
buffer (dashed lines) with experimentally observed values.
parallel to the FMQ, buffer; (3) equilibrium crystallization closed to oxygen; (4) fractional crystallization closed to oxygen. Calculations were explained if the modal proportion of clinopyroxene is
performed at 05 logio/o, intervals for oxygen fuga- not constant, as imposed by the model, but increases
cities at the liquidus in the range FMQ,+1 to slightly with decreasing temperature.
FMQ,-2.
Calculated ferric-iron contents of the total system
(crystals + liquid) do not remain constant as crystallization proceeds parallel to the FMQ, buffer (Fig.
RESULTS
4), a result of the system being open to oxygen. For
Evolution parallel to the FMQ buffer:
example, before the crystallization of magnetite the
equilibrium crystallization
ferric-iron content of the bulk system falls with
decreasing temperature because none of the crysComparison with experimental results
talline phases contain abundant ferric iron and the
The first test of the model is to reproduce the liquid, with constant Fe s+ /Fe 2+ , is decreasing in
experimentally determined equilibrium crystal- proportion; this requires a net removal of oxygen.
lization paths under conditions open to oxygen The crystallization of magnetite reverses this trend,
(Toplis & Carroll, 1995). Although this is the data
set used to constrain our calculations it should be
Open to O2
noted that only silicate phase appearance and modal
proportions are taken directly from the experimental
results. The appearance of Fe-Ti oxides, and the
compositional evolution of silicate minerals and
liquid with decreasing temperature are not fixed by B
to
the input data to the model. The model reproduces
experimental olivine compositions to within 1 mol %
Fo, the plagioclases to within 2 mol % An, and Cpx 3
to within 1 mol % En which are all typical of the
errors on the experimentally determined values. The
saturation temperatures of both the magnetiteulv6spinel and ilmenite-haematite solid solutions are
predicted to within ±5°C of the experimentally
determined values. Calculated and experimental
liquid compositions as a function of temperature are
1080
1100
1120
1140
1160
also in good agreement, as shown in Fig. 3. Model
Temperature (°C)
melt compositions lie within the range of the standard deviation of the experimental results, with the Fig. 4. The ferric iron content of the bulk lyitem (liquid + total
for equilibrium crystallization parallel to the FMQ buffer
exception of CaO which is slightly overestimated at solids)
as a {unction of temperature, illustrating that cryitalliration
the lowest temperatures. This may be most simply
under these conditions involves the exchange of oxygen.
842
TOPLIS AND CARROLL
FERRO-BASALTIC MAGMA DIFFERENTIATION
and the increased ferric-iron content of the bulk
system requires oxygen gain.
a)
Evolution parallel to the FMQ_ buffer:
fractional crystallization
•S
Crystal compositions
The modelled evolution of solid-phase compositions
with decreasing temperature shows important differences between equilibrium and fractional crystallization (Fig. 5). While olivine and plagioclase are
alone on the liquidus the difference between equilibrium and fractional crystallization is small, but
once clinopyroxene is stable the compositions of the
solid phases diverge, with mineral compositions produced by fractional crystallization being more
evolved than those produced by perfect equilibrium
crystallization at a given temperature.
Eqm xstll n
Frac xstll n
1080
1100
1120 1140 1160
1180
Temperature (°C)
b)
0.75
0.70
0.65
0.60
Liquid line of descent
a, 0.55
Fractional and equilibrium crystallization along the
FMQ, buffer result in very similar liquid lines of X 0.50
descent (Fig. 6), although fractional crystallization
0.45
yields a slightly greater iron enrichment and lower
magnesium contents at a given temperature. The
0.40
resulting change in the Fe/Mg ratio accounts for the
1080
divergence of the ferromagnesian mineral compositions with falling temperature even though liquid
compositions are generally similar.
1100 1120 1140
1160 1180
Temperature (°C)
Evolution closed to oxygen: equilibrium
crystallization
T-fOi paths
Crystallization in a system closed to oxygen can
yield large changes in the Fe 3+ /Fe 2+ of residual
melts and thus T~—fot paths will in general not
remain parallel to the FMQ, buffer curve, which
corresponds to approximately constant Fe 3+ /Fe 2+ in
the melt (Sato, 1978; Carmichael & Ghiorso, 1986).
1090
1100 1110 1120
1130
During the crystallization of silicates and ilmenite,
Temperature (°C)
ferric iron is enriched in the melt with respect to
ferrous iron, and the AFMQ_ increases, as shown in
Fig. 5. Comparison of the predicted tilicmte phase compositions
Fig. 7. Once magnetite begins to crystallize the T— (a) Xfo, (b) Xju, and (c) mg-numba^CpI^ for equilibrium (dashed
/ o , path changes to one of falling/Ol as also observed line), and fractional (continuous line with symbols every 5%
crystallized) crystallization at FMQ,- 2.
by Osborn (1959) and Presnall (1966) in their work
on haplobasaltic crystallization.
Crystal compositions, and liquid lines ofdescent
For closed system crystallization the compositions of
the silicate minerals are almost identical to those
produced by crystallization parallel to the FMO_
buffer. Liquid compositions produced by equilibrium
crystallization in systems open and closed to oxygen
are also similar (Fig. 8), although in the closed to
oxygen case the peak in FeO* concentration is
shifted to higher temperature and FeO* decreases
less strongly after magnetite saturation. This is due
to the earlier precipitation of magnetite caused by
843
JOURNAL OF PETROLOGY
NUMBER 4
VOLUME 37
AUGUST 1996
Equilibrium crystallization, FMQ
Open to O2, FMQ
20
16 -
I
?:
I
12
u
8 -
8 -
4
4 -
-
1060
1080
1100
1120
1140
1160
1080
1180
1100
1120
1140
1160
1180
Temperature (°C)
Temperature (°C)
Fig. 8. Predicted liquid compositions for perfect equilibrium crystallization under conditions closed to oxygen (initial ferric-ferrous
ratio denned by the FMQ_buffer). Dashed lines show the variation
of liquid composition for perfect equilibrium crystallization along
the FMQ,buffer (as Fig. 3).
Fig. 6. Predicted liquid compositions for perfect fractional crystallization along the FMQ, buffer (shown as continuous lines). The
predicted compositions for liquids produced by perfect equilibrium
crystallization (as Fig. 3) are also shown for reference (dashed
lines).
1.5
Mineral compositions
1
Minerals produced by fractional crystallization have
similar compositions regardless of whether they form
under conditions open or closed to oxygen. These
model compositions contrast with those produced by
equilibrium crystallization, with olivines and pyroxenes being more iron rich, and plagioclase more
sodium rich (see Fig. 5). Plagioclase and olivine
compositions are relatively insensitive tofot> whereas
clinopyroxenes develop a minimum in wollastonite
content at a l l / O j (Fig. 9) that occurs at lower mgnumber at lower/ O j . This feature is also commonly
observed in natural tholeiitic rock series (Campbell
J? -0.5
-1.5
-2
1080
1100
1120
1140
1160
Temperature (°C)
Fig. 7. Calculated T-^/o, paths for perfect fractional crystallization
of composition SCI under conditions closed to oxygen (continuous
lines), and equilibrium crystallization closed to oxygen (dashed
lines).
Fractional crystallization; closed to O2
0.46
the increase in / o , during the early stages of differentiation, and lower abundance of magnetite in later
stages of closed system crystallization.
§
0.44
-
0.42
.
0.40
-
Evolution closed to oxygen: fractional
crystallization
0.38
.
T-tOl paths
036
-
The T-fot paths for fractional and equilibrium
crystallization under conditions closed to oxygen are
almost identical (Fig. 7), the deviation being caused
by the different total iron contents (FeO*) of the
melts, even though the ferric-iron content is
identical.
0.75 0.70
0.65 0.60 0.55
0.50 0.45
0.40
mg-no. (Cpx)
Fig. 9. The predicted variation of wollastonite content of the Cpx
as a function of m£-number(CpI) in a portion of the pyroxene quadrilateral.
844
TOPLIS AND CARROLL
FERRO-BASALTIC MAGMA DIFFERENTIATION
& Nolan, 1974), and it may mark the onset of iron
depletion in the liquid.
proportion of magnetite over each crystallization
interval. The model predicts that for fractional crystallization parallel to the FMQ, buffer the initial
The liquid line ofdescent
cotectic proportion of magnetite is on the order of
40-45
w t % of the crystallizing solids at a l l / O l (Fig.
Only small differences are predicted in the liquid
lines of descent for fractional and equilibrium crys- l l a ) . With decreasing temperature the cotectic protallization under conditions closed to oxygen (Fig. portion of magnetite falls, with the sharpest decrease
10), similar to the case of evolution parallel to FMQ, observed at lower fot- A large initial crystallizing
Shi & Ldbourel (1991) have shown that equilibrium proportion of magnetite was inferred from the
and fractional crystallization paths in the system experimental results of Toplis & Carroll (1995), and
CMAS + FeO are also almost identical (their fig. is also supported by a number of other lines of evid11), and Hess et al. (1975) observed that the degree ence. Hill & Roeder (1974) carried out crystalof fractional crystallization had very Little effect on lization experiments on two basaltic compositions
the experimentally determined liquid line of descent over a large range of/ 0 l , and noted that 'there is
of an iron-rich lunar basalt composition. These always a rapid increase in the modal amount (of
results suggest that ferro-basaltic liquids follow spinel) over a consistent narrow zone of f(o,)~
similar paths regardless of the degree of fractional temperature conditions . . . in the vicinity of the
crystallization, which may explain the relative spinel liquidus'. Furthermore, Carmichael (1967)
success of experimental studies in reproducing natur- observed large numbers of Fe-Ti oxide phenocrysts
ally observed liquid lines of descent, even though after the saturation of magnetite in the rocks of the
many experiments approach equilibrium crystal- Thingmuli volcanic series, a feature which was also
lization conditions, whereas natural systems are predicted for the Thingmuli tholciites by the Gibbs
likely to evolve by some imperfect fractional crystallization.
a)
50
Open toO 2
D
"
I
'
I
"
30
20
•
if
.*
c
o
o
10 . ° i
° A*
Key
D FMQ+I
* FMQ
O FMQ-I
p
1
1
0 ' *_rV
1050 1060 1070 1080 1090 1100 1110 1120
Temperature (°C)
i
'
i
'
i
b)
F e O * ^ ^ .
16 -
o
o
o
o
o
o
a
Closed to O 2 , FMQ
20
A
40
Modal proportion of Fe—Ti oxides
The saturation temperature and crystallizing proportion of the Fe—Ti oxides have large effects on iron
and silica enrichment in crystallizing magmas. The
experimental observation that the ferric-iron content
of magnetite-saturated melts is an approximately
/o.-independent function of temperature (Toplis &
Carroll, 1995) is used in the model to calculate the
" ^ ^ ^ ^ ^
AI2O3
12 -
CaO •
8 -
MgO '
•a
4 i
1060
1080
.
i
Eqm xstll"
Frac xstll"
.
1100
1120
1140
1160
1060 1070 1080 1090 1100^1110 1120
Temperature (°C)
1130
Temperature (°C)
Fig. 10. Predicted liquid compositions for perfect fractional cryitallization of SCI under conditions dosed to oxygen, and an
initial ferric-ferrous ratio defined by the FMQ buffer. Dashed
lines show the variation of liquid compositions predicted for
perfect equilibrium crystallization under the same conditions.
845
Fig. 11. (a) Predicted wt% crystallizing proportion of magnetite
as a function of temperature for fractional crystallization parallel
to the FMQ buffer, (b) Predicted wt % crystallizing proportion of
magnetite as a function of temperature for fractional crystallization under conditions dosed to oxygen.
JOURNAL OF PETROLOGY
NUMBER 4
VOLUME 37
AUGUST 1996
0* (melt)
energy minimization model of Ghiorso & Carmi- a)
Open to O2
chael (1985).
20
1
l
1
Our model predicts the relative modal proportion
FMQ
of magnetite in a system closed to oxygen to be
18
greatest at magnetite saturation (Fig. lib), subsequently falling with decreasing temperature. In
fractional
this respect, it is similar to evolution parallel to the
16
' crystallization
FMQ, buffer (see Fig. lla). However, the proportion
of magnetite produced in a system closed to oxygen
14 f equilibrium s^*&
is predicted to be less than that produced during
crystallization
evolution parallel to the FMQ, buffer. Furthermore,
12
at a given temperature the proportion of magnetite
is of a similar magnitude at all _/o,> in constrast to
i
i
i
i
i
systems open to oxygen. With decreasing initial _/o,
in
magnetite saturates at lower temperature, and has
60
48
50
52
54
56
58
has a lower initial crystallizing proportion. Ilmenite
is predicted to stop crystallizing when magnetite
appears on the liquidus in a system evolving parallel
Wt% SiO2 (melt)
to FMQ, In contrast, ilmenite is predicted to constitute a relatively constant proportion of the
mineral mode in systems closed to oxygen once b)
Equilibrium
i
i
i
i
icrystallization
i
magnetite is on the liquidus (e.g. Fig. 13a, below).
FMQ*
18
closed to oxygen
I
\
16
14
\f aximum. iron enrichment and
covariation of iron and silica in the melt
A
12
A comparison of the covariation of FeO* and SiOj
produced during fractional and equilibrium crystallization open to oxygen shows that iron enrichments
are ~ 1 wt% FeO* greater for fractional crystallization, but that the trends of FeO* vs SiO2 are
parallel once magnetite is a stable phase (Fig. 12a).
On the other hand, systems evolving open and closed
to oxygen show contrasting covariations of iron and
silica (Fig. 12b). Before the saturation of magnetite
the crystallization paths are very similar, but after
this iron depletion is much less pronounced for the
case of a system closed to oxygen.
The sharp depletion in iron and enrichment in
silica of the melt phase in systems open to oxygen
may be related to the large initial crystallizing proportion of magnetite, which is independent of / o ,
(Fig. lla). Systems closed to oxygen crystallize a
much smaller proportion of magnetite, especially at
lower _/o, (Fig- lib). Maximum iron enrichments of
the melt phase may therefore be expected for fractional crystallization in a system closed to oxygen at
low _/o,- The greatest level of iron enrichment predicted by this model is 20 wt% FeO*, for fractional
crystallization at FMO_— 2, a conclusion consistent
with the observation that the most iron-rich terrestrial basaltic glasses contain <20 wt% FeO*
(Brooks et al., 1991).
\
-
10
open to oxygen
ft
48
i
i
i
i
i
i
50 52 54 56 58 60 62 64
Wt% SiO2 (melt)
Fig. 12. (a) Companion of the covariation of iron and silica in the
melt predicted for equilibrium (open lymboli) and fractional
(rilled symbols) cryitallization along the FMQ_ buffer. A symbol
is shown every 5% crystallized. Silica enrichment begins at
~ 5 5 % crystallized for both equilibrium crystallization and fractional crystallization, (b) The calculated covariation of FeO* and
SiOj in the melt products of SCI produced by perfect fractional
crystallization under conditions open to oxygen (small symbols,
dashed line) and closed to oxygen (large symbols, continuous
line). The system closed to oxygen started at the FMQ, buffer
and the system open to oxygen evolved at FMO.+ 0-5. These two
paths saturate magnetite at approximately the same wt % crystallized.
Modal proportions, and compositions of
liquids and coexisting bulk solids
Equilibrium and fractional crystallization lead to
very different relative modal proportions of the
crystalline phases. For example, when a new phase
appears on the liquidus during perfect fractional
crystallization it does so in its cotectic proportion
(Fig. 13a). In contrast, when a new phase appears
846
TOPLIS AND CARROLL
FERRO-BASALTIC MAGMA DIFFERENTIATION
Fractional crystallization; closed to O2
a)
1060
1080
1100
1120
1140
1160
Temperature (°C)
Equilibrium crystallization; closed to O 2
b)
100
UOI
t:
S
80
60
a.
ati
f>
40
20
1080
1100
1120
1140
1160
Temperature (°C)
Fig. 13. (a) Predicted relative wt % proportions of the solid phases
produced by perfect fractional crystallization of SCI under condition! dosed to oxygen, and an initial ferric-ferroui ratio corresponding to the FMQ. buffer, (b) Predicted relative wt%
proportions of the solid phases as a function of temperature for
fractional crystallization in a system closed to oxygen. The initial
ferric—ferrous ratio of the starting composition was set to that at
the FMQ, buffer.
on the liquidus during equilibrium crystallization its
relative proportion is much lower than its cotectic
proportion because the new phase coexists with
earlier formed minerals (Fig. 13b). The composition
of the bulk solid controls the compositional evolution
of the liquid [equation (2)], but the relation between
the composition of the liquid and that of the coexisting bulk solid is very different for perfect fractional
and perfect equilibrium crystallization.
Owing to the variation of relative modal proportions of crystalline phases during perfect fractional
crystallization (Fig. 13a), a given component in the
coexisting bulk solid varies with falling temperature
(Figs 14a—d) by either (1) a continuous variation
caused by changes in mineral chemistry for a fixed
phase assemblage (e.g. the variation of FeO* for the
assemblages Ol + Plag or Ol + Plag+Cpx), or (2) a
discontinuous variation caused by a change in the
stable phase assemblage (e.g. the variation of CaO
when Cpx becomes a liquidus phase). During perfect
fractional crystallization the compositional evolution
of the liquid is controlled only by the composition of
the coexisting solids. For example, the appearance of
magnetite causes a large discontinuous increase in
the FeO* content of the bulk solid (which was
poorer, but becomes richer in FeO* than the coexisting liquid), leading to a change from iron
enrichment to iron depletion in the liquid. Similarly,
the onset of magnetite crystallization greatly reduces
the silica content of the bulk solid, causing a rapid
increase in the SiO2 content of the liquid. It should
be noted that only in the case of perfect fractional
crystallization may the composition of a single bulk
solid be used to calculate die compositional evolution of the coexisting liquid.
During perfect equilibrium crystallization the bulk
solid contains crystals which were not formed during
limited crystallization intervals (Fig. 13b), thus no
discontinuous breaks in the composition of the coexisting bulk solid are observed, and the compositional
variation of the liquid with falling temperature may
no longer be simply related to the compositional
variation of the bulk solid (Figs 14e-h). For
example, the appearance of magnetite causes a peak
in the FeO* content of the liquid, but the FeO*
content of the bulk solid does not become greater
than that of the liquid at that point (compare the
case for fractional crystallization above). In the case
of equilibrium crystallization, both the initial and
final bulk solid compositions over a defined crystallization interval must be known before the compositional evolution of the liquid can be calculated. It is
clear that if bulk solids which are the products of an
imperfect fractional crystallization are used to
estimate the compositional evolution of the liquid by
assuming perfect fractional crystallization then erroneous liquid lines of descent will be calculated. This
is important because mass balance calculations used
to estimate the liquid lines of descent for layered
intrusions (e.g. Skaergaard, Kiglapait) rely heavily
on the assumption that cumulates arc the products of
perfect fractional crystallization.
IMPLICATIONS FOR FERROBASALTS AND GABBROS
T-fot paths of natural magmas: open or
closed to oxygen?
The T-fo7 paths of natural magmas have been estimated using a number of techniques, the most
847
JOURNAL OF PETROLOGY
VOLUME 37
a)
NUMBER 4
AUGUST 1996
e)
Eqm xstlln
16
12
8
4
n
1060
1080
1100
1120 1140 1160
^s
I
1080
Temperature (°C)
.
I
.
1100
I
1120
-
. . . ,v
1140 1160
Temperature (°C)
b)
1060
1080
1100
1120 1140 1160
1080
Temperature (°C)
16
c)
•
1
1
-
1
14
8"
t£
I0
^
8
—«—'•
•
1
1100
1120
1140
1160
Temperature (°C)
1
g)
J^°
6
1060
1080
1100
1120
1140 1160
1080
Temperature (°C)
1100
1120
1140 1160
Temperature (°C)
h)
d)
1060
1080
MOO
1120 1140
1160
1080
1100
1120
1140
1160
Temperature (°C)
Temperature (°C)
Fig. 14. (a-d) Comparison of the wt % oxide component in the bulk solid (filled symbols) and the coexisting liquid (open symbols) for
perfect fractional crystallization in a system closed to oxygen starting at FMQ, (e-h) Comparison of the wt % oxide component in the
bulk solid (filled symbols) and the coexisting liquid (open symbols) for perfect equilibrium crystallization in a system closed to oxygen
starting at FMQ,
848
FERRO-BASALTIC MAGMA DIFFERENTIATION
reliable of which are (1) the use of mineral equilibria
such as magnetite-ilmenite pairs (e.g. Buddington &
Lindsley, 1964), and (2) the use of ferric-ferrous
ratios of melt quenched to glass (e.g. Carmichael &
Ghiorso, 1986). Most estimates based on mineral
equilibria are restricted to relatively evolved, multiply saturated magmas, thus the early stages of differentiation are often poorly constrained. The use of
ferric—ferrous ratios has traditionally been applied
only to fresh rapidly quenched glasses, as plutonic
rocks may experience complex low-temperature
redox re-equilibration (Frost & Lindsley, 1992).
Two-oxide thermometry (Carmichael, 1967), and
Ca-QUIIF equilibria (Frost & Lindsley, 1992)
indicate that the Thingmuli volcanic suite from
Iceland evolved along a T-fo, path parallel to and
slightly below the FMQ, buffer curve, implying evolution open to oxygen. A similar conclusion of evolution parallel to the F M Q buffer has been inferred
for Hawaiian volcanics based on two-oxide thermometry (Anderson & Wright, 1972). The available
evidence thus suggests that the large-volume tholeiitdc volcanism on both Iceland and Hawaii may
involve magma evolution under conditions open to
oxygen. The T-fo, paths of mid-ocean ridge basalt
(MORB) glasses, estimated from measured ferricferrous ratios (Byers it al., 1984; Christie it al., 1986)
show that the most primitive MORBs may be relatively reduced ( ~ F M Q - 2 ) but oxidation accompanying differentiation yields oxidation states of
FMQ to FMQ+ 1 at Fe-Ti oxide saturation. This
behaviour is consistent with crystallization of a
system closed to oxygen, although as discussed by
Juster et al. (1989), assimilation of country rocks
could produce a similar trend.
1
its)
TOPLIS AND CARROLL
i
•
1
0.5
MZ
c
L *_
'
1
LZb L Z a "
4
»
_
0
o
a
-0.5
:
-1
<
KEY
^ ^ ^ ^
WUHBUI(I»7I)
Mcflincy t Nvh
Thn vnrit
i
<
1
.
1
.
1
.
i
i
•
i
1020 1040 1060 1080 1100 1120 1140 1160
Temperature (°C)
Fig. 15. Companion of predicted T-fc^ path for the fractional
cryitallixation of the composition SCI under condition! closed to
oxygen, itarting at FMQ,-0-5, with that estimated for the Skaergaard intrusion by Williams (1971), and McBimey & Nailund
(1990) (LZ, lower xone; MZ, middle rone; UZ, upper *one).
the predicted and observed values of _/o, show
approximately parallel decreases with falling temperature, suggesting that the modelled path is not
unreasonable. The contrasting trends during the
early stages of differentiation may be due to open
system behaviour, or alternatively simply due to the
difficulty of estimating / O f conditions for cumulates
formed before the appearance of Fe-Ti oxides. The
estimated T-fo, path of the Kiglapait intrusion
(Morse, 1981) shows the same general trend of
decreasing / o , once Fe-Ti oxides appear on the
liquidus, although the decrease with falling temperature is not as great as that inferred for Skaergaard. The T-fo, path of the Newark Island layered
intrusion, Labrador, has also been estimated, by
Snyder it al. (1993). Those workers used a qualiEstimated T-fo,
paths for the Skaergaard
tative comparison of the ratio of magnetite to
intrusion calculated from mineral equilibria are all
ilmenite observed in the intrusion (cumulus +
in broad agreement (Williams, 1971; Morse it al.,
intercumulus) to that produced by the equilibrium
1980; Frost & Lindsley, 1992). These data suggest
crystallization of a suitable starting composition over
that magnetite saturation occurred when magma/o,
an appropriate range of/o,, and concluded that
was slightly above the FMQ buffer, but further
precipitation of cumulus magnetite leads to a
cooling and crystallization was accompanied by a
decrease of AFMQ with falling temperature.
decrease i n / O j relative to the FMQ buffer, reaching
Although the conclusion is reasonable, it is inap~ F M Q - 2 in the most evolved rocks from the
propriate to use phase proportions estimated from
Sandwich Horizon (Frost it al., 1988). Intrinsic/ o ,
equilibrium crystallization experiments to make
measurements suggesting much lower oxygen fugainferences about complex cumulus processes which
cities (Sato & Valenza, 1980) are inconsistent with
were probably dominated by fractional crystalthe observed stable mineral assemblage (Frost &
lization along unknown T-fo, paths. Nevertheless,
Lindsley, 1992), and the reliability of the method
the results of our modelling indicate that Fe—Ti
has been questioned. Comparison of estimates of the
oxide saturation during fractional crystallization
7"-/ o , path of the Skaergaard intrusion (Williams,
under conditions closed to oxygen yields a relatively
1971;*McBirncy & Naslund, 1990) with that which
constant proportion of ilmenite and a decreasing
we calculate for fractional crystallization of SCI
proportion of magnetite as temperature decreases
closed to oxygen (Fig. 15) shows that after the
(Fig. 13a), producing a variation in the ratio
appearance of Fe-Ti oxides (LZc of Skaergaard),
849
JOURNAL OF PETROLOGY
VOLUME 37
magnetite/ilmenite consistent with that observed at
Newark Island. These observations suggest that
many plutonic magmas may evolve under conditions
approaching the closed to oxygen case, in contrast to
the more open behaviour indicated by volcanic
paragcnescs showing T-fOt paths subparallel to the
FMQ, buffer.
NUMBER*
AUGUST 19%
at a given stratigraphic level, because, in the case of
the Kiglapait intrusion, plagiodase will not be able
to re-equilibrate on the timescale of cooling of the
intrusion (Morse, 1984). At 80 PCS plagiodase
crystals have a range of ~ 5 mol % anorthite around
an average value of ~An 55 (Morse, 1979, 1981). By
analogy with the experimental results of Toplis &
Carroll (1995), such a range of values is consistent
with plagioclases which crystallized over a 20°C
temperature interval. Imperfect fractional crystallization of magma containing minerals which crystallized over a 20°C range of temperature would
explain the low initial proportion and gradual
appearance of both pyroxene and Fe-Ti oxides
(compare Fig. 13b). The decrease in olivine proportion during the interval 80-89 PCS would suggest
that olivine was not a crystallizing liquidus phase at
this time, and the subsequent increase in its abundance would signal its reappearance on the liquidus
(analogous to the Middle Zone of the Skaergaard
intrusion). The fact that the decrease in pyroxene
abundance at 89 PCS coincides with the return of
olivine as a liquidus phase is consistent with a
decrease in the cotectic proportion of pyroxene when
olivine reappears on the liquidus. Similarly, the
declining abundance of Fe-Ti oxides after 91 PCS
may be due to a progressive decrease in their cotectic
proportion with falling temperature, as suggested by
the modelling presented in this paper (Fig. 13a). At
95 PCS the appearance of apatite does not show a
gradual rise in abundance, suggesting a more efficient crystal fractionation at this level.
Minei
Efficiency of crystal removal in
solidifying plutons
Plutonic rocks with bulk compositions far removed
from those of erupted materials provide compelling
evidence that crystal fractionation occurs in many
magmas. However, the phenocrysts found in nearly
all volcanics show that melts are not completely
cleared of crystals as cumulates form. The contrasting crystal proportions formed during perfect
equilibrium and perfect fractional crystallization
(Figs 13a and b) can be used as indices of the efficiency of crystal removal. Morse (1979, 1981) gave a
detailed account of the variation of mineral proportions as a function of weight percent solidified (PCS)
for the troctolite-gabbro-fcrrogabbro cumulates of
the Kiglapait intrusion, shown in Fig. 16. The variations of crystal proportions with stratigraphic
height (in particular, pyroxene and the Fe—Ti
oxides) are not what one would expect for perfect
fractional crystallization (compare Fig. 13a). Morse
(1979) attributed these variations to suppressed
nucleation followed by supersaturation and overproduction, but an alternative interpretation is that
they simply reflect inefficient crystal fractionation.
At the Skaergaard intrusion, Wager & Brown
Evidence for a degree of equilibrium crystallization
(1967) reported that 'there are certain abrupt
is provided by variations in plagiodase composition
entrances and exits of minerals', suggesting a close
approach to perfect fractional crystallization.
However, McBirney (1989) noted that the dis3b
appearance of olivine at the base of the MZ, and its
| Kiglapait; Morse (1981)
reappearance at the base of the UZ cannot be
30 I
Cpx/\
sharply defined, and that 'scattered large crystals of
25
olivine persist well into the interior of the zone', a
feature
which suggests less than perfect fractionation
•a 20
at Skaergaard.
15
The effiriency of crystal removal during cumulate
rd
formation has been considered by Sparks et al.
10
"(3
'•
• *^
V^» . \r y \
(1993), who proposed that the formation of modal
5
layering in basic intrusions is caused by crystals
rr M> >
-o- ^ 1 Apatite
reaching a critical concentration in the melt, at
n
100
95
90
85
80
75
which point convection is no longer able to keep
them suspended, leaving the magma clear of crystals.
Wt% Solidified (PCS)
Estimated values of the critical concentration are
Fig. 16. The variation of wt% abundance of minerals for the
upper portion of the Kiglapait, taken from CIFW norms of the ~001 wt%, implying that fractionation should be
smoothed rock model of Morse (1981). Plagiodase is not reported.
very effident. The poor development of modal
At lower levels of the intrusion, olivine and plagiodase are the layering at the Kiglapait intrusion (Morse, 1979)
only cumulus minerals, and have relatively constant relative
may imply that cumulates formed via in situ
proportions of 70 wt % plagiodase:30 wt % olivine.
850
TOPLIS AND CARROLL
FERRO-BASALTIC MAGMA DIFFERENTIATION
boundary-layer crystallization rather than crystal
settling.
A comparison of predicted bulk solids
and natural cumulates
An important comparison to make is between the
compositions of the predicted bulk solids from our
model for SCI crystallization and those of natural
gabbroic and ferro-gabbroic cumulates. In light of
the evidence from inferred T-/Oi paths that at least
some plutonic systems crystallize under conditions
approaching the closed to oxygen case, predicted
bulk solids for systems closed to oxygen will be compared with data from gabbroic intrusions. Although
the compositions of the crystalline products of perfect
equilibrium crystallization cannot be compared with
those of real rocks (other than at the solidus), and
perfect fractional crystallization is a mathematical
construction which is rarely, if ever, achieved in
nature, we consider that natural magmas are
bracketed by these two end-member cases.
The phase relations of SCI are consistent with
those of the Skaergaard intrusion as discussed above,
with which the model will be compared. (It should
be noted that McBirney (1989) and McBirney &
Naslund (1990) clearly stated that pigeonite is not a
cumulus phase at the Skaergaard intrusion, so
absence of pigeonite in our model does not compromise the interpretation. As shown by Toplis et al.
(1994), stabilization of pigeonite as an intercumulus
phase may be caused by enrichment of intercumulus
liquids in phosphorus.) The prevalence of modally
layered rocks at the Skaergaard intrusion makes
assessment of 'average' compositions somewhat difficult, but McBirney (1989) reported values of the
average compositions of each subzone based on an
analysed sample set large enough to be representative of the average rock on this scale. Comparison of these average Skaergaard compositions
with those predicted for fractional crystallization in a
system closed to oxygen are in broad agreement
(Fig. 17), showing that realistic cumulate compositions can be calculated from our experimental data
and parameterization.
and perfect equilibrium crystallization, supporting
the contention that these rocks formed by an
imperfect fractional crystallization process.
Differences between natural and modelled compositions may be expected owing to a component of
trapped interstitial melt as well as post-cumulus
redistribution of certain elements in the natural
environment. However, it is clear that the modelled
cumulate compositions can help to provide insights
into the processes associated with the formation of
natural cumulates.
The liquid lines of descent of natural
systems
Suites of glasses
The results of the modelling presented above show
that the variation in concentration of many of the
melt components with falling temperature is very
similar for the four end-member modes of crystallization considered. An important exception is the
covariation of iron and silica in the melt, which is
characteristically different for melts evolving under
conditions open and closed to oxygen (Fig. 12b). A
comparison of melt compositions for suites of volcanic glasses from Iceland (Carmichael, 1964), and
the Loch Ba ring dyke, Mull (Sparks, 1988) with the
results of the model presented above shows that the
trends for these natural suites closely parallel those
predicted for evolution open to oxygen (Fig. 19a),
which is also consistent with the inferred T—fot paths
of these magmas. The higher FeO* contents of the
lavas from the Loch Ba ring dyke compared with
Icelandic rocks may be due to crystallization at
lower/ Ot .
The covariation of iron and silica in the glasses
from Iceland or Loch Ba are typical of large-volume
extrusive magmatism, and natural volcanic series
which show alternative variations of these components are rare. The cone sheets from Centre 3 in
Mull (Thomson, 1986) provide one such example
(Fig. 19b), with a trend of compositional evolution
suggestive of crystallization under conditions closed
to oxygen. The cone sheets are small-volume intrusions, and may represent liquids expelled from a
Compositions of the cumulates of the Kiglapait coexisting magma chamber where cumulates were
intrusion (Morse, 1981) provide an additional forming. If this interpretation is correct, it provides
example with which to compare modelled bulk solid further evidence that plutonic magma bodies are
compositions. The variation of SiC>2 of the Kiglapait more likely to evolve under conditions relatively
rocks (Fig. 18a) clearly defines a crystallization path closed to oxygen. Furthermore, it is interesting to
intermediate between that predicted for perfect note that the Centre 3 cone sheets from Mull were
fractional and perfect equilibrium crystallization. formed during the same period of igneous activity
The variation of the FeO* content of the bulk solid which terminated in the formation of the Loch Ba
(Fig. 18b) is also broadly consistent with an evo- ring dyke. Thus is may be inferred that the early
lution intermediate between that of perfect fractional evolution of Centre 3 was characterized by formation
851
JOURNAL OF PETROLOGY
VOLUME 37
AUGUST 1996
NUMBER*
'
1
1
1
1
50
s
X)
LZb
V UZa MZ
45
«—
LZc
-
40
,
1080
1100
1120
1
1060
1140
1080
d)
c)
i
22
~o
i
•
i
i
i
-
-
18
-
-
16
-
n
UZa
"3 8
MZ
LZc
«
LZb
UZa
• __
: _ - — • — * — _/—
*
-
LZb
LZc
•
•
1080
</>
MZ
10
1060
1140
73
-
12
R
1120
12
i
20
14
<
i
1100
Temperature (°C)
Temperature (°Q
1100
i
i
1120
4 <:
•
i
1140
1060
Temperature (°C)
i
1080
1
1100
1120
1140
Temperature (°C)
Fig. 17. a comparison of the compositions of bulk solids predicted by the model for fractional crystallization in a system closed to oxygen,
starting at the FMQ buffer, with the average bulk-rock compositions for the subzones of the Skaergaard intrusion reported by McBimey
(1989). The compositional variation within each zone is typically ~20% relative for each oxide component The temperatures for the
Skaergaard intrusion are arbitrary, but chosen to compare phase assemblages, if possible, with those of the predicted model. LZb, olivine
plagiodase, dinopyroxene; LZc, olivine, plagiodasc, clinopyroxene, magnetite; MZ, plagioclase, dinopyroxene, magnetite; UZa, as
LZc.
of low-volume intrusive magmas (and presumably
cumulates at depth) evolving under conditions closed
to oxygen, whereas the latest stage was characterized
by a much larger volume of volcanic products produced by evolution in a system more open to oxygen.
Systems open and closed to oxygen may thus be
closely associated, both spatially and temporally.
Inferred by mass balance
In the absence of preserved magmatic liquids from
plutonic environments, estimations of liquid lines of
descent for these environments have relied on mass
balance arguments involving incremental subtraction of observed cumulate compositions from a
proposed initial bulk composition. The liquid lines of
descent of the Kiglapait intrusion (Morse, 1981), the
Skaergaard intrusion (Wager, 1960; Wager &
Brown, 1967) and the Basistoppen sill, East
Greenland (Naslund, 1989) have been calculated in
this way. These intrusions all consist of a continuous
series of cumulates which range in composition to
magnesium-free end-member mafic phases, and very
sodic plagioclases (~An2o~An3o) at the highest
stratigraphic levels.
The Kiglapait intrusion is the largest of the three
examples, with a volume of ~3500 km3. The compositions of the calculated liquids and observed bulk
solids for this intrusion (Morse, 1981) and the predictions of the model presented in this study for
fractional crystallization of a system closed to oxygen
are shown in Fig. 20a. The calculations of Morse
(1981) imply that magnetite crystallization does not
terminate iron enrichment of the liquid phase, and
that the maximum iron enrichment is ~22 wt%
FeO* at ~43 wt% SiO2; this behaviour is at odds
with all of the liquid lines of descent predicted by the
model presented here. If the observed cumulates
were produced by in situ crystallization, it is possible
that the boundary layer contained very evolved
liquids, as suggested and quantitatively modelled by
Langmuir (1989). However, the need for evolved
boundary-layer melts at the Kiglapait intrusion has
been questioned by recent Rayleigh fractionation
852
FERRO-BASALTIC MAGMA DIFFERENTIATION
TOPLIS AND CARROLL
a)
%crystallized (Kiglapait)
a)
100
95
90
85
80
75
70
20
1
*.
16 -
"3
a.
50
1
'
1
1
*
9
14
'
Open to o 2
18
55
•3
•
••-.°
*!»
•" ••
12 -
FMQ-l
8
40
6
-
FMQ • .
10
45
.
-
'••-..
Key
•
Thtapnufi (CimiichKl. 1964)
Loch Biifai dyke (Spskt. I9IS)
I
.
I
.
I
-O
o
°
d
i
48 50 52 54 56 58 60 62 64
I
35
I
Wt% SiO2 (melt)
30
FeO*'(bul ksol id)
I
1
1
i
1
1
1
1
—•— Kigbptil Mora (1981)
25
b)
• Fyrfed fractkral
A
-- - IYJICCI ffpnhhriwn
i
i
i
1
i
Closed to 0 2 "
18
-&•
•&'
20
15
i * * *„
10
16
K
\
\
V
*
o
V.
5
0
0.2
•y.
i
0.3
i
0.4
1
0.5
i
0.6
i
0.7
i
0.8
i
0.9
•
. . FMQ-l
14
12
o
:
oo
FMQ
•
•
•
.
_
Key
8
1
Fraction of liquid remaining (model)
modelling (Morse, 1996). This latest study also
shows that the composition of the summation liquid
(Morse, 1981) is inconsistent with known constraints
on olivine-melt equilibria, and the validity of this
summation liquid has been questioned. If the cumulates are not the products of perfect fractional crystallization (e.g. see Fig. 16), this may in part explain
the failure of rock summation to reproduce the real
liquid line of descent.
The Skaergaard intrusion, with an estimated
volume of ~300 km 3 , is considerably smaller than
the Kiglapait intrusion. The calculated liquid evolution for Skaergaard (Wager & Brown, 1967) also
shows continued iron enrichment after the
appearance of magnetite, again in contradiction to
any of the liquid lines of descent calculated using the
Y
10
v-
Fig. 18. Comparison of the observed bulk-rock compositions from
the Kiglapait intrusion (Morse, 1981) with bulk-solid compositions predicted from the model presented in this study, for perfect
equilibrium and perfect fractional crystallization of a system
closed to oxygen for (a) SiO^ and (b) FeO*. The upper scale
refers to the estimated wt % crystallized of the Kiglapait intrusion
and the lower scale is the fraction of liquid remaining in the
model. The two scales have been combined so that Fe-Ti oxide
saturation occurs at approximately the same point.
;
O
~ O Centre 3 cone ihecu Lock Bi (Thomson. 1986)
1
1
1
1
1
1
48
50
52
54
56
58
1
i
60
62
64
Wt% SiO2 (melt)
Fig. 19. (a) The covariation of iron and silica for naturally occurring lavas from Iceland and Loch Ba ring dyke, Mull. The dashed
lines show the predicted variations for fractional crystallization of
a system open to oxygen at FMQ, and FMQ.+ 1. (b) The covariation of iron and silica for naturally occurring lavas from Centre 3
cone iheeti of Loch Ba, Mull. The dashed lines show the predicted
variations for fractional crystallization of a system closed to
oxygen at FMQ, and FMQ.+ 1.
model presented here (Fig. 20b). This discrepancy
occurs despite: (a) the modelled composition being
proposed as a possible parental composition to the
exposed portion of the Skaergaard (SCI =dyke C;
Brooks & Nielsen, 1978); (b) the model predicting a
realistic T~fox path ( Fi g- 1 5 ) ; ( c ) ^ e model predicting realistic cumulate mineralogy and compositions (Fig. 17). This inconsistency supports
suggestions (e.g. Hunter & Sparks, 1987, 1990) that
the calculations for the liquid evolution of Skaergaard may be flawed by (1) an incorrect choice of
starting composition, (2) erroneous relative volumes
of the zones used in the mass balance calculations, or
(3) venting of a significant volume of silicic differentiates after the formation of the highest exposed
levels of the intrusion.
853
JOURNAL OF PETROLOGY
VOLUME 37
The Basistoppen sill, described by Naslund (1989),
is much smaller ( ~ 6 km3) and much better constrained than either the Skaergaard or the Kiglapait
intrusion. Naslund (1989) noted: 'The excellent
exposure, uncomplicated structure, good chilled
margin, and lack of strong modal layering facilitate
a calculation of a differentiation trend.' In contrast
to the Skaergaard and the Kiglapait intrusions, the
a)
I
6
40 42 44 46 48 50 52 54 56 58 60 62
Wt% SiO2 (melt)
b)
26
24
Skaergaard (Wager & Brown, 1967)
20
9
PU
14
10
8
40 42 44 46 48 50 52 54 56 58 60 62
Wt% SiO2 (melt)
c)
Basistoppen sill (Naslund, 1989)
20 _
(Mt+ at 68%)
18 16
14
calculated liquid line of descent of the Basistoppen
sill (Naslund, 1989) is entirely consistent with the
modelled liquid evolution for fractional crystallization in a system closed to oxygen (Fig. 20c).
If the calculated liquid trend of Wager & Brown
(1967) for the Skaergaard intrusion is real, then
some factor or process which has not been taken into
account in the modelling must have affected the
compositional evolution of the residual liquids. The
results of our modelling are strictly valid only for the
composition SCI crystallizing under anhydrous conditions at atmospheric pressure. In addition, SCI
does not contain any of the trace elements (most
notably sulphur), or the minor elements manganese
or phosphorus which may have an effect on the
liquid line of descent (Toplis et al., 1994). If the
inferred liquid line of descent is real, a compositional
difference may be the source of the discrepancy with
the model. However, the good agreement of the
liquid line of descent modelled here with that
inferred from the cumulates of the Basistoppen sill,
and the compositional similarity of the cumulates of
the Basistoppen sill with those of the Skaergaard
intrusion suggests that compositional factors are
unlikely to be responsible for the large deviation
between our modelled liquid lines of descent and
that inferred from the cumulates of the Skacrgaard
intrusion.
The evolution of plutonic vs volcanic
environments
18
16
12
*
o
AUGUST 1996
UZb
22
£
*
NUMBER. 4
9 T ^: ? " - ^ . . ,
87*
-K
FMQ^
•\ Initial liq
12
i
48
i
50
52
54
56
58
Wt% SiO2 (melt)
60
62
The results of this study suggest that the magmas in
many volcanic environments evolve open to oxygen,
whereas those which have yielded some of the best
studied layered intrusions evolved under conditions
closer to the closed to oxygen case. Volcanic environments such as those at Hawaii or mid-ocean
Fig. 20. (a) The calculated wt*/. FeO* and SiO 2 for the liquids
and coexisting cumulates of the Kiglapait intruiion. The wt%
crystallized (from Mone, 1981) U shown next to each liquid
symbol. The predictions from the model for fractional crystallization in a system dosed to oxygen at FMQ and F M Q - 1 are also
shown for reference (dashed lines), (b) The calculated wt% FeO*
and SiOj for the liquids and coexisting cumulates of the Skaergaard intrusion. The rone or subzone reported by Wager &
Brown (1967) is shown next to the relevant symbol. HZ, Hidden
zone; LZ, lower zone; MZ, middle zone; UZ, upper zone. The
predictions from the model for fractional crystallization in a
system closed to oxygen at FMQ. and F M Q - 1 are also shown
for reference (dashed lines), (c) The calculated wt% FeO* and
SiO2 for the liquids of the Baiistoppen sill (Naslund, 1989).
Numbers show the estimated wt % crystallized at each point. No
estimates of liquid composition are made between 58 and 87%
crystallized, but magnetite is reported to appear as a cumulus
phase at ~ 6 8 % crystallized. The predictions from the model for
fractional crystallization in a system closed to oxygen at FMQ
and FMQ— 1 are also shown for reference (dashed lines).
854
TOPLIS AND CARROLL
FERRO-BASALTIC MAGMA DIFFERENTIATION
ridges, are characterized by large magma fluxes
through them, with frequent replenishment and
mixing between more evolved and more primitive
compositions; these factors make it difficult to
envisage magmatic conditions which could be characterized as being closed to oxygen. Plutonic bodies
which represent a single magma injection, followed
by cooling and crystallization of this magma at
depth, provide much better candidates for systems
capable of evolving under closed conditions. The
lack of replenishment in the Skaergaard case may
explain its similarity to our predictions for a system
closed to oxygen. Similar closed to oxygen magma
bodies may develop in active volcanic areas where,
owing to the vagaries of magma plumbing systems, a
volume of magma becomes 'trapped' in a chamber
and remains there, cooling and crystallizing without
new magma input or mass loss owing to eruption.
Such bodies are only uncovered long after active
volcanism has ceased and erosion has exposed them.
some plutonic environments (e.g. Skaergaard,
Kiglapait and Newark Island layered intrusion)
suggests that these environments evolved under conditions more closed to oxygen.
Crystal modal abundances and bulk-rock compositions from the Kiglapait intrusion are consistent
with inefficient crystal fractionation. In contrast,
evidence from the Skaergaard intrusion is consistent
with a close approach to perfect fractional crystallization. The results of the modelling are particularly
relevant to the crystallization of the Skaergaard
intrusion, because the modelled composition has
been proposed as a possible parental composition.
The modelling predicts a variation of oxygen
fugacity with falling temperature which is in good
agreement with that inferred from the natural
cumulates. Furthermore, predicted bulk solid
(cumulate) compositions agree well with bulk-rock
compositions from Skaergaard. However, none of the
predicted liquid paths are consistent with the liquid
line of descent calculated for the Skaergaard
intrusion (Wager & Brown, 1976). Several compositional factors were not taken into account in the
model (i.e. water and phosphorus) which may
explain the discrepancy. However, the cumulates of
the smaller and better constrained Basistoppen sill
(which closely resemble those of the Skaergaard
intrusion) may be used to calculate a liquid line of
descent in agreement with that predicted by the
model for fractional crystallization closed to oxygen.
CONCLUSIONS
Experimental constraints from Toplis & Carroll
(1995) have been used to model equilibrium and
fractional crystallization of a single ferro-basaltic
composition in a range of fot from FMQ_+ 1 to
FMQ,— 2, under conditions open and closed to
oxygen. Modelled liquid lines of descent for perfect
equilibrium and perfect fractional crystallization are
very similar. For all the modelled conditions magnetite saturation terminates iron enrichment and
initiates silica enrichment of the melt phase,
although important differences exist between the
evolution of systems open and closed to oxygen.
More magnetite crystallizes in a system open to
oxygen, and this leads to a large decrease in the iron
content of the liquid and a large increase in the silica
concentration with decreasing temperature. In comparison, magnetite saturation in systems closed to
oxygen leads to less iron depletion but a similar silica
enrichment. This difference is due to the dominance
of magnetite as an Fe-Ti oxide phase during open
system processes, and the increased importance of
ilmenite in closed system processes.
The maximum iron enrichment of the melt phase
predicted by this modelling is ~20 wt% FeO*,
consistent with maximum enrichments reported for
volcanic glasses (Brooks et al., 1991). A comparison"
of the predicted T-fOt paths and liquid lines' of
descent with those inferred from natural systems
suggests that large-scale volcanic centres, such as
Iceland, Hawaii and the Loch Ba ring dyke evolved
under conditions open to oxygen. The case for
MORB is less clear. In contrast, the evidence from
ACKNOWLEDGEMENTS
This work has benefited from numerous discussions
with Jon Blundy, Guy Libourel, Henri Soulard and
Steve Sparks. Alastair Davies is particularly thanked
for help in installing a working version of the
MELTS program, as well as technical and moral
support during modelling aspects of the paper. P.
Baker, K. Cox, R. Hunter and particularly T. Sisson
are acknowledged for critical reviews which helped
improve the clarity of the manuscript. M.J.T.
acknowledges receipt of an NERC studentship and
EEC fellowship during completion of this work.
REFERENCES
Andenon, A. T. ft Wright, T. L., 1972. Phenocryiu and glass
inclusions and their bearing on the oxidation and mixing of
basaltic magmas, Kilauea Volcano, Hawaii. Amtrican
MUtralegisl 57, 188-216.
Brooks, C. K. ft Nielsen, T. F. D., 1978. Early stages in the differentiation of the Skaergaard magma as revealed by a closely
related suite of dike rocks. Lilhos 11, 1-14.
Brooks, C. K. & Nielsen, T. F. D., 1990. The differentiation of the
Skaergaard Intrusion. A discussion of Hunter and Sparks.
Contributimt to Mvurakgj and Ptbolofj 104, 244-247.
855
JOURNAL OF PETROLOGY
VOLUME 37
NUMBER 4
AUGUST 1996
Brooks, C. K., Larsen, L. M. & Nielien, T. F. D., 1991.
Hunter, R. H. & Sparks, R. S. J., 1990. The differentiation of the
Importance of iron-rich tholciitic magmu at divergent plate
Skaergaard Intrusion. Reply to McBimey and Naslund.
margins: a reappraisal. Geology 19, 269-272.
Contributions to Mineralogy and Petrology 106, 248-254.
Buddington, A. F. & Lindjley, D. H., 1964. Iron—titanium oxide Juster, T. C , Grove, T. L. & Perfit, M. R., 1989. Experimental
minerals and synthetic equivalents. Journal of Petrology 5, 310—
constraints on the generation of Fe-Ti basalts, andesites, and
357.
rhyodacites at the Galapagos spreading centre, 85°W and
95°W. Journal of Geophysical Research 94, 9251-9274.
Byers, C. D., Christie, D. M., Muenow, D. W. & Sinton, J. M.,
1984. Volatile contents and ferric-ferrous ratios of basalt,
Kilinc, A., Carmichael, I. S. E., Rivers, M. L. & Sack, R. O.,
ferrobasalt, andesite and rhyodacite glasses from the Galapagos
1983. The ferric-ferrous ratio of natural silicate liquids equili955°W propagating rift. Geochimica tt Cosmochimica Acta 48, brated in air. Contributions to Mintralogy and Petrology 83, 136—140.
2239-2245.
Langmuir, C. H., 1989. Geochemical consequences of in situ
Campbell, I. H. & Nolan, J., 1974. Factors affecting the stability
crystallization. Nature 340, 199-205.
field of Ca-poor pyroxene and the origin of the Ca-poor miniLangmuir, C. H. &. Hanson, G. N., 1981. Calculating mineralmum in Ca-rich pyroxenes from tholeiitic intrusions.
melt equilibria with stoichiometry, mass balance, and singleContributions to Mineralogy and Petrology 48, 205-218.
component distribution coefficients. In: Newton, R. C ,
Navrotsky, A. & Wood, B. J. (eds) Advances in Physical
Carmichael, I. S. E., 1964. The petrology of Thingmuli, a
Geochemistry, Vol. 1: Thtrmodynamics of Minerals and Melts. Berlin:
Tertiary volcano in eastern Iceland. Journal of Petrology 5, 435Springer-Verlag.
460.
Carmichael, I. S. E., 1967. The iron—titanium oxides of salic volMcBimey, A. R., 1989. The Skaergaard layered series: I.
canic rocks and their associated ferromagnesian silicates.
Structure and average compositions. Journal of Petrology 30, 363Contributions to Mintralogy and Petrology 14, 36—64.
397.
Carmichael, I. S. E. &. Ghiorso, M. S., 1986. Oxidation-reduction McBirney, A. R. & Naslund, H. R., 1990. The differentiation of
relations in basic magma: a case for homogeneous equilibria.
the Skaergaard Intrusion. A discussion of Hunter and Sparks.
Earth and Planetary Scitna Letters 78, 200-210.
Contributions to Mintralogy and Petrology 104, 235-240.
Christie, D. M., Carmichael, I. S. E. & Langmuir, C. H., 1986.
Morse, S. A., 1979. Kiglapait geochemistry: II. Petrography.
Oxidation states of mid-ocean ridge basalt glasses. Earth and
Journal of Petrology 20, 591-624.
Planetary Science Letters 79, 397-411.
Morse, S. A., 1981. Kiglapait geochemistry IV: The major
elements. Geochimica et Cosmochimica Acta 45, 461—479.
Frost, B. R. & Iindsley, D. H., 1992. Equilibria among Fe-Ti
oxides, pyroxenes, olivine and quartz: Part II. Application.
Morse, S. A., 1984. Cation diffusion in plagiodase feldspar. Sdtnce
American Mineralogist 77, 1004-1020.
225, 504-505.
Frost, B. R., Lindsley, D. H. & Anderson, D. J., 1988. Fe-Ti
Morse, S. A., 1990. The differentiation of the Skaergaard
oxide—silicate equilibria: assemblages with fayalitic olivine.
Intrusion. A discussion of Hunter and Sparks. Contributions to
American Mintralogist 73, 727-740.
Mineralogy and Petrology 104, 240-244.
Ghiorso, M. S. & Carmichael, I. S. E., 1985. Chemical mass
Morse, S. A., 1996. Kiglapait mineralogy III: Olivines and the
transfer in magmatic processes. II. Applications in equilibrium
principles of igneous fractionation. Journal of Petrology, in press.
crystallization, fractionation and assimilation. Contributions to Morse, S. A., Lindsley, D. H. & Williams, R. J., 1980. Concerning
Mintralogy and Petrology 90, 121-141.
intensive parameters in the Skaergaard Intrusion. American
Journal of Science 280A, 159-170.
Ghiorso, M. S. & Sack, R. O., 1995. Chemical man transfer in
magmatic processes. IV. A revised and internally consistent
Naslund, H. R., 1989. Petrology of the Basistoppen sill, East
thermodynamic model for the interpolation and extrapolation
Greenland: a calculated magma differentiation trend. Journal of
of liquid—solid equilibria in magmatic systems at elevated temPetrology 30, 299-319.
peratures and pressures. Contributions to Mintralogy and Petrology Nielsen, R. L., 1988. A model for the simulation of combined
119, 197-212.
major and trace element liquid lines of descent. Geochimica et
Cosmochimica Acta 52, 27-38.
Ghiorso, M. S., Carmichael, I. S. E., Rivers, M. L. & Sack, R.
O., 1983. The Gibbs free energy of mixing of natural silicate
Nielsen, R. L., 1990. Simulation of igneous differentiation proliquids; an expanded regular solution approximation for the
cesses. In: Nicholls, J. & Russell, J. K. (eds) Modem Methods of
calculation of magmatic intensive variables. Contributions to
Igneous Petrology: Understanding Magmatic Processes. Mintralogical
Mintralogy and Petrology 84, 107-145.
Society of America, Rtuitms in Mineralogy 24, 65—105.
Helz, R. T., 1987. Differentiation behavior of Kilauea Dri lava
Osborn, E. F., 1959. Role of oxygen pressure in the crystallization
lake, Kilauea Volcano, Hawaii: an overview of past and current
and differentiation of basaltic magma. American Journal of Sdtnce
work. In: Mysen, B. O. (ed.) Magmatic Processes: Physiochtmical 257, 609-647.
Principles. Geochemical Society Special Publication 1, 241-258.
Presnall, D. C , 1966. The join forsterite-diopside-iron oxide and
its bearing on the crystallization of basaltic and ultramafic
Hess, P. C , Rutherford, M. J., Guillemette, R. N., Ryerson, F. J.
magmas. American Journal of Sdtnce 264, 753—809.
& Tuchfeld, H. A., 1975. Residual products of fractional cryjtalh'zation of lunar magmas: an experimental study. Proceedings Sato, M., 1978. Oxygen fugacity of basaltic magmas and the role
of the 6th Lunar Science Conference. Houston, TX: Lunar Science of gas-forming elements. Geophysical Research Letters 5, 447—449.
Institute, pp. 895-909.
Sato, M. & Valenza, M., 1980. Oxygen fugacities of the layered
series of the Skaergaard intrusion, East Greenland. American
Hill, R. & Roeder, P., 1974. The crystallization of spinel from
Journal of Science 280A, 134-158.
basaltic liquid as a function of oxygen fugacity. Journal of Geology
82, 709-729.
Shi, P. & Libourel, G., 1991. The effects of FeO on the system
CMAS at low pressure and implications for basalt crystalHunter, R. H. & Sparks, R. S. J., 1987. The differentiation of the
Skaergaard Intrusion. Contributions to Mineralogy and Petrology 95, lization processes. Contributions to Mintralogy and Petrology 108,
129-145.
451^161.
856
TOPLIS AND CARROLL
FERRO-BASALTIC MAGMA DIFFERENTIATION
Snyder, D., Carmichael, I. S. E. * Wiebe, R. A., 1993.
Table A2: Temperature,
Experimental itudy of liquid evolution in an Fe-rich, layered
percent crystallized andphase
mafic intrusion: constraints of Fe-Ti oxide precipitation on the
T—fo, anc^ T—p paths of tholeiitic magmas. Contributions to
appearance
Mintralogy and Pttrology 113, 73-86.
Sparks, R. S. J., 1988. Petrology and geochemsitry of the Loch Ba
ring-dyke, Mull (N.W. Scotland): an example of the extreme
r(K) = 1295 + 266 x (FUq) - 2 9 0 x (/TJq) 2 +175 x (FUq)3
differentiation of tholeiitic magmas. Contributions to Mintralogy
where 7*(K) ii the temperature in Kelvin, and FUq Is the weight fraction of
and Pttrology 100, 446-461.
liquid
Sparks, R. S. J., Huppert, H. E., Koyaguchi, T. * Hallworth, M.
A., 1993. Origin of modal and rhythmic igneous layering by
Ptaglodasa appears when
sedimentation in a convecting magma chamber. Naturt 361,
246-248.
Olivine appears when
Thomson, B. A., 1986. The petrology and geochemistry of the
Clinopyroxeno appears when
/TJq = 0-74 + (0-5 x AFMQ)
Tertiary cone-sheet complex, Island of Mull, Scotland. Ph.D.
Magnetite appears when
wt%FejO 3 (liq)
Thesis, King's College, London.
= exp(24-22 - {3-19 x [10OO0/r(K)]})
Toplis, M. J. & Carroll, M. R., 1995. An experimental study of
llmentte
appears
when
wt%H0a(llq)
the influence of oxygen fugacity on Fe—Ti oxide stability, phase
relations, and mineral—melt equilibria in ferro-basaltic systems.
= [0-0409X7TQ] -40-30
Journal of Petrology 36, 1137-1170.
Toplis, M. J., Libourel, G. & Carroll, M. R., 1994. The role of
phosphorus in the crystallization processes of basalt: an experiTables A2 and A3, respectively. A reaction relation of olivine
mental study. Gcochimua it Cosmockimua Ada 58, 797-810.
with the liquid was inferred experimentally, and is a common
Wager, L. R., 1960. The major element variation of the layered
series of the Skaergaard intrusion and a re-estimation of the feature of many tholeiitic rock series (Helz, 1987; Juster it al.,
1989). The onset of resorption was taken from the experimental
average composition of the hidden layered series and of the
results, as was the magnitude of this resorption, which was fixed
successive residual magmas. Journal ofPitrology 1, 364—398.
at 6 or 10% depending on the other phases present (i.e. 1 g melt
Wager, L. R. & Brown, G. M., 1967. Laytrtd Igntous Rods.
plus 006 or 01 g olivine reacting to form 1-06 or 11 g of proEdinburgh: Oliver and Boyd, 558 pp.
ducts). Mineral-melt equilibria imposed for the solid phases and
Weaver, J. S. & Langmuir, C. H., 1990. Calculation of phase
equilibria in mineral—melt systems. Computers and Gioscitncis coexisting liquids are shown in Table A4. Olivines were considered to be a binary solution of fonterite and fayalite, and do
16(1), 1-19.
Williams, R. J., 1971. Reaction constants in the system FeO- not contain any calcium. Plagiodases were considered to be a
mixture of albite and anorthite, with a fixed FeO* content of ~ 1
MgO—SiOj—O2; intensive parameters in the Skaergaard intruwt%. The dinopyroxene composition was calculated from the
sion, East Greenland. Amtrican Journal of Scitna 271, 132—146.
ternary En—Fs—Wo, to which 3 mol % of the titanium Tchermak
molecule (MTiAljOj, where M is a divalent cation) was added.
For melts not containing an Fe-Ti oxide an additional 2 mol %
RECEIVED SEPTEMBER 4, 1995
of the Al Tchermak molecule (MAljSiOg) was added, to closely
reproduce the compositions of experimentally produced clinopyrREVISED TYPESCRIPT ACCEPTED MARCH 11, 1996
oxenes. Magnesium and aluminium are added to the magnetiteulvdspinel solid solution as MgFejO 4 and FeAl2O+, respectively.
The molar ratio of Al to Mg is fixed to be three, and the ratio of
MgO in the magnetite to that in the liquid is fixed to be 0 5 .
APPENDIX
Magnesium is added to the ilmenite—haematite solid solution as
The model considers crystallization of composition SCI (Table
MgTiOj, such that the molar ratio of MgO in the ilmenite to
Al) in fixed increments of \% which are converted to temperathat in the liquid is also 0-5. Ferric and ferrous iron contents of
tures using the third-order polynomial shown in Table A2, obthe liquid and, above the FMO_ buffer, pyroxene are partitioned
tained from the experimentally determined wt % crystallized as a
as a function o f / o , •* shown in Table A5. The compositions of
function of temperature (Toplis & Carroll, 1995), which was
the Fe-Ti oxides are a function of foj their major element
found to be independent (±5°C) of/o,. The constraints used for
chemistries calibrated from the data of Toplis & Carroll (1995)
phase saturation and crystallizing proportions, based on the exare also shown in Table A5.
perimental results of Toplis & Carroll (1995), arc shown in
Table Al: Modelled composition (in wt %),SC1
SiO2
TIO2
AljOj
FeO1
MgO
CaO
NajO
K2O
48-7
2-9
14-9
13-1
65
10-9
2-7
0-3
Composition taken from Toplis & Carroll (1995). FeO*, all iron as FeO.
857
JOURNAL OF PETROLOGY
VOLUME 37
NUMBER 4
AUGUST 1996
Table A3: Relative crystallizing proportions
(1) Parallel to the FMQ buffer, olivine stable:
Plag + OI
7ft30
Plag + OI + Cpx
48:12:40
Ptofl + OI + Cpx + llm
37:11:435
(2) Parallel to the FMQ buffer, olMne unstable:
Plag + C p x - O r
60:50:-10
(Pl»g + Cpxt
55:45)
Plag + C p x + l l m - O r
4&49:9:-6
(Plag + Cpx + llmt
46:46:8)
Plag + C p x + M t - O I "
60:66:Mtt:-6
(Plag + Cpx + Mtt
47:51Mtt)
'Values apply to equilibrium crystallization where crystals of olivine are present but resorbing.
tValues apply to fractional crystallization where olivine is not present
CThe proportion of magnetite is allowed to vary until the ferric iron content of the residual melt lies on
the magnetite 'saturation curve'. Proportions are then normalized to 100.
Table A4: Mineral-melt constraints
OlivlrxMnelt
Table A5: The effect qffOi
Ifrnp-numbef (liq) >30then/faSiJj = 0-32
AFMQ Is defined as the /a, relative to the FMQ buffer
If/TV-number ( l i q ) < 3 0 t h e n X d S ^
Liquid*
% FeO' as FejOj
Pyroxene
wt% Fe20,°0-2x AFMQ (for AFMQ > 0)
= 0 0 1 x [86-85-(5-88x AFMQ)-0-984X AFMQ2)]
= 0-5 - 0-06 x [/np-number (liq)]
Pyroxene-mett
/QfJiJ3 = O-17 + O-OO18x [/ngr-number (Bq)] •
ffjf-117
Plagloclase-melt
Magnetrte-ulvospinel XM, = 0-65-(0-133 x AFMQ)
j t f = 0-51xxS?'/XtjJ0+0-25x(xS°/Xt3?0)2
llmenlte-heematite
0-919-(0-028x AFMQ)
This equation represents the ferric-ferrous ratio of SC1 as a
function of fo, at 1120°C calculated using Kilinc etal. (1983). This
equation was used for all subsequent melts. The compositional
dependence of the ferric-ferrous ratio of the modelled melts
[estimated using Kilinc etal. (1983)] is < 10% (relative).
858