Full text of DOI:10.1007/s00410-002-0351-z

Contrib Mineral Petrol (2002) 143: 300–315
DOI 10.1007/s00410-002-0351-z
Catherine Ginibre Æ Gerhard Worner Æ Andreas Kronz
Minor- and trace-element zoning in plagioclase:
implications for magma chamber processes
at Parinacota volcano, northern Chile
Received: 21 February 2001 / Accepted: 7 January 2002 / Published online: 22 March 2002
ꢀ Springer-Verlag 2002
Abstract Textural and compositional zoning in plagio-
Introduction
clase phenocrysts in a sample from Parinacota volcano
(Chile) was investigated using backscattered electron
Magma chamber processes play an important role in the
images and electron microprobe analysis of major and
trace elements. Large (2 mm) oscillatory zoned crystals
(type I) with resorption surfaces of moderate An dis-
continuities (610% An) and decreasing trace-element
contents (Sr, Mg, Ti) towards the rim reflect melt dif-
ferentiation and turbulent convection in the main mag-
ma body. Early recharge with a low-Sr mafic magma
is seen in the core. Small-scale Sr variations in the
core indicate limited diffusion and thus residence and
differentiation times of the magma shorter than a few
thousand years. Smaller crystals (type II) with low trace-
element/An ratio reflect the influence of an H2O-rich
melt probably from a differentiated boundary layer.
Closed-system in-situ crystallisation, mafic magma re-
charge and the role of a water-rich differentiated
boundary layer can be distinguished from the An–trace
element relationships. Crystals apparently move rela-
tively freely between different parts and regimes in the
magma chamber, evidence for ‘‘convective crystal dis-
persion’’. High-Sr type II crystals indicate an earlier
input of Sr-rich mafic magma. Recharge of two distinct
mafic magma types is thus identified (high-Sr and low-
Sr), which must have been present – at increasing re-
charge rates with time – in the plumbing system
throughout the volcano’s history.
formation of igneous rocks and include crystal growth,
bulk melt differentiation, convection, magma recharge
and other processes depending on dynamic and thermal
regimes. Experimental and numerical models of magma
chamber processes (see reviews and discussion in Marsh
1989a, 1989b, and Huppert and Turner 1991) allow to
constrain the role of physical parameters such as density
and viscosity as a function of melt composition, tem-
perature gradients, geometry of the magma chamber,
etc. However, such models are difficult to apply directly
to natural systems because many of these parameters are
not well constrained.
A complementary approach uses records of the
crystallisation history in volcanic products, such as
zoning in plagioclase. Due to sluggish diffusion of the
Si–Al pair in plagioclase (Grove et al. 1981), major-
element zoning reflecting the An–Ab binary system has
been widely used in studying igneous plagioclase as a
record of magma compositions and crystallisation con-
ditions. Petrographic observation and high-resolution
imaging methods (e.g. laser interference and Nomarsky
interference; Anderson 1983, 1984; Pearce et al. 1987;
Nixon and Pearce 1987) have also been applied. Ex-
perimental studies of crystallisation (Lofgren 1974) and
resorption (Tsuchiyama 1985; Johannes et al. 1994;
Nakamura and Shimakita 1998) or numerical modelling
(Allegre et al. 1981; Loomis 1982; Lasaga 1982;
L’Heureux and Fowler 1996) provided insights into the
kinetics of plagioclase growth and resorption.
C. Ginibre (&) Æ G. Worner Æ A. Kronz
Geowissenschafliches Zentrum Gottingen,
Abt. Geochemie, Goldschmidtstr. 1,
37077 Gottingen, Germany
Although these studies have brought a better under-
standing of the origin of zoning patterns, limitations
remain from insufficient spatial resolution of imaging
and chemical analysis, as well as from the fact that the
effects of several factors (P, T, melt composition, H2O
content) on the An–Ab system cannot be distinguished.
In order to address the latter problem, several recent
studies have combined textural imaging with minor-and
E-mail: catherine.ginibre@chimie.univ-nantes.fr
Fax: +330251125268
Present address: C. Ginibre
Laboratoire de Planetologie et Geodynamique,
Universite de Nantes, Faculte des Sciences et des Techniques,
2 rue de la Houssiniere, BP 92208,
44322 Nantes cedex 3, France
Editorial responsibility: T.L. Grove

301
Table 1. Composition of sample PAR-130. Major element con-
centrations are in wt %: trace element concentrations are in ppm
trace-element analyses by SIMS (Blundy and Shimizu
1991; Singer et al. 1995; Brophy et al. 1996) or isotopic
data (Davidson and Tepley 1997; Tepley et al. 2000).
This combination allowed to better identify the causes of
plagioclase zoning patterns, for example, recharge and
assimilation. However, because of the low spatial reso-
lution of SIMS and microdrilling (a few tens to hundreds
of microns) and the large efforts and costs involved, this
approach does not allow the study of zoning at small
scale for a large number of samples. Despite lower pre-
cision, electron microprobe analysis has the advantage
of a much better spatial resolution (a few lm). Kuritani
PAR-130
SiO2
66.83
0.65
16.05
3.6
0.06
1.26
3.33
4.07
3.89
0.26
100.00
61
TiO2
Al2O3
FeO*
MnO
MgO
CaO
Na2O
K2O
P2O5
(1998), for example, analysed several distinct zones in a Total
V
Cr
Co
Ni
Cu
Zn
Rb
Sr
large number of plagioclase crystals for major and minor
(Fe, Mg) elements using the electron microprobe. He
reconstructed specific locations of plagioclase growth
within the magma chamber of the Rishiri volcano (Ja-
pan). He was able to distinguish plagioclase that origi-
nated from the chamber’s boundary layer at the wall
rock from those of the magma body.
8
30
16
27
83
152
664
15
Y
In the present study we analysed plagioclase pheno-
crysts from a dacite from the Upper Pleistocene to Ho-
locene Parincota volcano of the Central Andean
Volcanic Zone (northern Chile). We used an electron
microprobe for minor and trace elements with a spatial
resolution of a few microns. The trace-element data are
combined with textural observations using BSE images
(Ginibre et al. 2002). We distinguish between the role of
the crystal-chemical influence on the partition coefficients
of minor and trace elements, diffusion in the crystal, bulk
melt composition (X, H2O content, ƒO2), kinetic effects
at the crystal–melt interface, and variation of intensive
parameters (T, P). This approach provides evidence for
several processes: compositional differentiation, magma
recharge of two distinct mafic magmas with and without
chemical mixing, as well as crystallisation in an H2O-
enriched magma chamber boundary layer.
Zr
Nb
Ba
218
11
1,172
the rebuilding of the ‘‘New Cone’’ (<18,000 to about
2,000 years). This new ‘‘healing’’ stage represents a sud-
den change from low eruption rates and large composi-
tional variations to high eruption rates and lavas with a
more restricted compositional spectrum. Two distinct
mafic magmas, the one with low, the other with high Sr
(and Ba) concentrations were erupted in a final stage in
flanc eruptions (Ajata flows). The overall compositional
variation of Parinacota lavas suggests that these two types
of mafic magma may be involved in recharge events and
mixing (Worner et al. 1988; Davidson et al. 1990).
Sample PAR-130 was chosen for its relatively simple
petrography, which should allow us to better identify
individual processes. PAR-130 contains ca. 15% pla-
gioclase (An55–An30), 5% euhedral phenocrysts of or-
tho-and clinopyroxene, below 1% euhedral hornblende
and rounded biotite, two iron-titanium oxides, and a
matrix composed of pyroxene, apatite, oxide, K-feldspar
microcrysts and glass. Most other Old Cone Parinacota
andesitic to dacitic samples contain rounded sanidine
and quartz xenocrysts as well as rhyolitic plagioclase
(An40 to An20) with fine and extensive sieve textures at
the rim, which is typical evidence for late magma mixing.
By contrast, plagioclase phenocrysts from PAR-130
show no resorption at the rim, and the rock contains
neither sanidine nor quartz xenocrysts. PAR-130 has
low Ni and high Rb values, similar to the rhyolites from
Parinacota eruptive stages Ia, i.e. it is close to the dif-
ferentiated end member.
Samples and methods
Samples
Parinacota is an Upper Pleistocene to Holocene volcanic
complex in the Central Volcanic Zone of the Andes.
Sample PAR-130 studied here is a dacite (66.83 wt% SiO2,
Table 1) belonging to the old stratocone stage (eruptive
stage IIa, 53,000–18,000 years; Worner et al. 1988, 2000).
This ‘‘Old Cone’’ stage is characterised by a large com-
positional range (mafic andesites to rather porphyric sil-
icic andesites, dacites and rhyolites). This period had an
extended differentiation history with low eruption rates
and mixing between mafic magmas and older dacitic to
rhyolitic magmas. In an Ni–Rb diagram (Worner et al.
1988), Parinacota Old Cone whole-rock concentrations
fall on mixing lines between mafic and differen- Methods
tiated end members. A catastrophic collapse around
18,150 650 years (Worner et al. 2000) was followed by After petrographical examination, selected crystals were
the rapid eruption of mafic to intermediate andesites and investigated with a JEOL 8900 RL electron microprobe.

302
BSE imaging
analytical uncertainty (70 ppm) and can be neglected.
Iron concentrations were therefore not corrected for
Zoning patterns were documented by back-scattered secondary fluorescence effects at the rims of the crystals.
electron (BSE) images where BSE intensity in plagio-
clase corresponds to An content. BSE intensity profiles
were calibrated for An content using quantitative point
analyses following the approach described by Ginibre
et al. (2002).
Zoning patterns in PAR-130 plagioclase phenocrysts
The diversity of plagioclase zoning patterns in sample
PAR-130 is shown in Fig. 1. We distinguish two types of
plagioclase phenocrysts: type I crystals are 1–4 mm large
and oscillatory zoned with only weak resorptions as
described below. This type comprises a relatively small
proportion of crystals, which represent over half of the
plagioclase population in volume. Type II crystals are
smaller (500 lm to 1.5 mm) and more extensively re-
sorbed. The outer 100 to 200 lm of the larger crystals of
both types I and II show oscillatory zoning with
rounded edges, similar to numerous microphenocrysts.
Quantitative point analyses
Quantitative point analyses for Al, Si, Na, Ca, K, Ba, Sr,
Fe, Ti and Mg were performed at 20-kV acceleration
voltage and 40-nA beam current, with 2- or 5-lm beam
size. Alkali elements were analysed first, and all major
elements were analysed during the first 90 s (16-s
counting time on peak). A study of Na loss shows that
alkalis were not significantly lost during this time period.
Minor (Ba, Sr and Fe) and trace elements (Ti, Mg) were
then analysed over 4-min counting time on the peak.
The influence of alkali loss on the minor and trace ele-
ments is small enough to be negligible because, for these
elements, the intensity/count ratio is constant and in-
dependent of minor variations in major-element com-
position. Table 2 lists the detection limits for Ba, Sr, Fe,
Mg, and Ti, as well as the ranges of concentrations and
statistical errors ( 2r) for the analyses at the concen-
trations measured.
The spatial resolution of wavelength-dispersive
(WDS) analysis is a few microns. In homogeneous
zones, no significant differences were found between
analyses with a 2-and a 5- lm beam. Particular attention
was given to measuring clear, homogeneous zones.
Whenever the size of fine patchy or oscillatory zoning is
below the WDS spatial resolution, the reported analyses
represent an average over these heterogeneities. After
measurement, the points are viewed and precisely lo-
cated on BSE or secondary electron (SE) images, and
can thus be related to the type and width of the An
zoning. Measurement points were subsequently exam-
ined under the optical microscope in order to eliminate
data points too close to inclusions.
Type I crystals
Three representative type I crystals (P1, P3 and P4) were
chosen for detailed investigation. They all have similar
overall zoning patterns in major and minor elements.
BSE An-calibrated and quantitative profiles were per-
formed in P1 and P4, and in more detail in P3. Patchy
growth zones with glass inclusions, which are related to,
and grew subsequently to, distinct resorption surfaces
are here referred to as resorption zones. These were
analysed in detail in P1 and P4.
An content
BSE images and An-calibrated profiles of the analysed
crystals are shown in Fig. 2. The three crystals exhibit an
overall normal zoning pattern (An50 to An30–35), with
oscillations superimposed on this trend. Details of the
oscillatory zoning in plagioclase P3 from sample PAR-
130 using high-resolution BSE images are described
elsewhere (Ginibre et al. 2002). The following types of
zoning are distinguished.
This has been done in particular because secondary
fluorescence of iron-rich phases located near the mea- 1. Saw-tooth patterns with resorption (STR) are 5 to
sured feldspars may influence measured Fe content in
plagioclase (Longhi et al. 1976). None of the measure-
ments presented here is closer than 50 lm to iron-rich
phases (oxide, pyroxene, etc). The influence of the ma-
trix, which contains less than 1 wt% FeO, is below the
20 lm in wavelength, have 5–10 mol% An variations,
and are separated by wavy surfaces (‘‘r’’ in Fig. 3).
These patterns are interpreted to result from growth
interruption and minor resorption of the last grown
layer.
Table 2. Detection limits and
Mg
Ti
Fe
Sr
Ba
ranges of concentration for Mg,
Ti, Fe, Sr and Ba in sample
Detection limita
Concentration range
17
100–250
19
18
100–400
28–31
29
1,000–4,000
60–70
67
1,000–3,000
110–120
50
500–2,000
70–75
PAR-130 (all values are given
in ppm)
Statistical errorb
aCalculated as 2· standard deviation of the intensity at background position (n=97, 95% confidence
level)
bCalculated from counting statistics for each analysis

303
Fig. 1a–h. Crystal types of
plagioclase in PAR-130. a P3
is typical for type I (large
oscillatory zoned) crystals.
Type II crystals show more
resorption or more contrast
between core and rim. b–e
Images of type IIa crystals
P42, P25, P2, and P9. f–h
Images of type IIb crystals
P18, P28, and P32. The
latter are distinguished only
on the basis of their Sr
contents (see Fig. 6)
2. Low-amplitude oscillations (LAO) at 1-lm wave- 3. Major resorption surfaces (R) with discontinuities in
length and 1% An variation are superimposed on
normal zoning of the STR zones and are most likely
due to local kinetic control. These are not considered
further in the present study.
An content up to 10% An crosscut several growth
layers (STR and LAO). Some of these resorption
surfaces form bands containing glass inclusions and
patchy zoning, referred to as resorption zones (Fig. 3).

304
Fig. 2a–f. BSE image of
type I plagioclase crystals
PAR-130-P3, P4, and P1.
Scale bar is 100 lm for all
crystals and profiles. An-
calibrated profiles through
BSE images are shown for
each crystal. For P3, the
profile from high-resolution
accumulated image (Ginibre
et al. 2002) is shown. For P1
and P4, profiles are taken
from the image along the
lines. R0 is a prominent,
important resorption surface
late in the crystallisation
history. It can be correlated
between P4 and P1, and
probably corresponds to the
boundary between zones 7
and 8 in P3 (see Fig. 4). R1 is
a resorption surface with an
An peak near the core, and
probably represents the
same event in the three
crystals (see discussion in
text)
4. High-frequency (3 lm) An variations with high am- therefore interpreted to be simultaneous to further
plitudes (10 mol% An; HFA) at the rapidly grown, growth at the outer rim of the crystal. Individual zones
outer crystal margin are interpreted as an extreme are often too small (65 lm) to be measured separately,
form of STR due to increased dynamic movements of and point analyses are therefore mixed between a sodic
crystals and melt during eruption.
and a calcic end member.
Direct correlation of zoning patterns between P1, P3
and P4 is not possible. There are, however, several
striking similarities between crystals. An An maximum
is observed near the core, followed by a general de-
crease from An50 to minimum of An30 and a slight An
increase towards the rim. The last three STR zones in
crystals P1, P4 and P3 are very similar in size, mor-
phology and compositional range, and may thus be
correlated. The last resorption zone can also be corre-
lated between P1 and P4, having similar composition
and morphology. They also are located directly before
the An minimum in the profile. It will be referred to as
R0 for P1 and P4 in the following. Such an R0 re-
sorption zone is difficult to identify in P3, but might be
represented by the resorption zone between growth
zones 7 and 8.
Special cases are the complex, patchy resorption
zones representing three stages of plagioclase growth.
(1) The last zone grown before resorption (Pl1 on
Fig. 3b) is generally sodic (An30–40). The resorption
surface itself is irregular and penetrates deeply into this
zone. (2) A calcic zone (Pl2, An50–60) is located directly
at the resorption surface and is similar to those
observed in partial melting experiments (Tsuchiyama
1985; Johannes et al. 1994; Nakamura and Shimakita
1998; Hammouda and Pichavant 2000). They therefore
may be interpreted as having grown during or directly
after partial dissolution of the more sodic crystal. (3)
New sodic zones (An30–40) sometimes exhibit oscilla-
tions and occur around the inclusions. These zones are
chemically identical to the next outer growth layer,
which crystallised after the resorption event, and are

305
reproduced in trace-and minor-element zonations. This
may simply be due to the higher analytical uncertainty
for trace elements. Only Fe shows strong oscillations,
and only at the outermost rim (Fig. 2b).
Figure 5 combines the data of P3 with those of the
two other plagioclase crystals PAR-130-P1 and PAR-
130-P4. They exhibit the same general trend from core to
rim as P3, except for the resorption zones, not analysed
in P3. The composition of the core (inwards from R0)
and rim (outwards from R0) of P1 and P4, compared to
those of P3, confirms that this surface separating zones 7
and 8 could be an equivalent of R0 in P1 and P4.
At the scale of the whole crystal, Sr, Fe, Ti, Mg are
roughly positively correlated with An content, with
concentrations high in the core and low in the outer
300 lm (500 to 800 lm from the core). Ba correlates
negatively with An.
In the inner zones (1 to 5), Fe, Ti, and Sr contents do
not correlate well with An content. Fe is fairly constant
at variable An whereas Ti is negatively correlated with
An in zones 1 and 2, as is Sr in zone 2. Thus, the overall
variation of An content with trace elements (Fig. 5) re-
flects igneous differentiation of the host melt(s).
In detail, however, there is more structure to this
variation. Growth zones 6–7 and 8–9 define parallel
linear trends in the An–trace element diagram (Fig. 5).
This is particularly well documented for Sr, Fe, and less
obvious for Ti. From one growth zone to the next, there
are discontinuities in the overall decreases in trace ele-
ments. For Ba, a complex pattern is superimposed on
the strong overall negative correlation. Mg, by contrast,
Fig. 3a–c. Details of zoning patterns in plagioclase. Scale bar is
20 lm. a, b Oscillatory zoning including STR patterns with minor
(r) and major (R) resorption surfaces and superimposed LAO
oscillations, as well as HFA oscillations at the rim (Ginibre et al.
2002). c Patchy zoning at the resorption surface R0 in crystal P1.
Pl0 represents remnant plagioclase grown before resorption. Pl1 is
the plagioclase grown during or immediately after resorption. Pl2 is
the plagioclase grown in the following growth phases, both in the
melt inclusion and outward from the resorption zone
Differences between the crystals are the variable shows a fairly linear positive correlation with An.
amplitude and sharpness of the An peaks as well as the The outermost zone with HFA oscillations in An
number and morphologies of R surfaces and the dis- (zone 10) shows negative correlations of Fe, Ba and Ti
tance of the R1 resorption surface to the rim. This with An content, and an increase in Mg compared to
variable distance may reflect different degrees of re- zone 9 at similar An content. Sr is negatively correlated
sorption and growth rates. In P3, where zones bound- with An.
aries are broadly euhedral, the peak and resorption R1
An content in resorption zone R0 is variable but Fe,
are at 500 lm from the rim. On the contrary in P1, where Ti, Sr and Mg are fairly constant at low levels. The
the resorption surfaces are more rounded between R1 compositions of the resorption zone in fact fall into a
and R0, presumably showing a higher degree of re- triangular-shaped field in the An–trace element plots.
sorption, the same (?) R1 surface is only 300 lm from the Some data may reflect mixed analyses due to the fine
rim.
zonations (An-rich resorption zones and subsequent
sodic growth) whereas other analyses clearly fall below
the differentiation trend, and therefore reflect growth of
a high-An plagioclase with low trace-element contents.
Only Ba does not show compositional differences be-
Minor and trace elements
Based on zoning patterns and composition, we define ten tween growth and resorption zones.
distinct growth zones in crystal P3, which are separated
by major resorption surfaces (R, Fig. 4). These growth
zones are also distinguished by discontinuities in their Type II crystals
anorthite–trace element trends (Fig. 5). High-resolution
(65 lm) trace-element profiles across the crystal PAR- Zoning patterns in type II crystals are more variable
130-P3 show a general decrease of Sr, Fe, Ti, and Mg than in type I, as shown on Fig. 1. Many of them exhibit
from core to rim, and an increase in Ba. In the core (first a resorbed core with a relatively wide, often oscillation-
100 lm, zone 1), local minima for Ti and Sr are ob- free An-rich zone (P25, P2, P31). Oscillatory zoning with
served. In the outer 50 lm (zone 10) Fe, Ti, Mg and Ba frequent rounded resorption (P18), and small An-rich
contents increase whereas Sr content decreases. The cores with microphenocryst-like sodic rim (P32) are also
5–20 lm oscillations in An content (STR) are not observed.

306
Fig. 4. EMP quantitative
point analysis profile of pla-
gioclase PAR-130-P3 from
core to rim for An (in
mol%), and Ba, Sr, Fe, Ti
and Mg (in ppm). The ten
distinct growths zones
shown here are separated by
major resorption surfaces
(see Fig. 3 for details) and
distinguished on Fig. 5 using
the trace-element content
Sr, Fe, and Mg in type II plagioclase are plotted which are believed to record the conditions of crys-
against An in Fig. 6. These crystals all have lower Fe tallisation of the main magma body. We then discuss
and Mg at variable An compared to type I crystals, the additional information provided by the type II
but they fall in the field of the R0 resorption zone. Sr crystals.
concentrations are variable and we distinguish two
subtypes: in type IIa, as for Fe and Mg, Sr is low at
high An content, similar to the R0 zones of type I
crystals. On the contrary, type IIb crystals have high
Sr concentrations in the core, significantly higher than
Control on major elements
Major-element composition of plagioclase is controlled
in the cores of type I crystals (up to 3,800 ppm
by phase relations and thus gives a general idea of the
compared to 2,600 ppm). Types IIa and IIb are not
history of T–P–H2O conditions of the magma in which
distinguishable on the basis of major-element zoning,
the crystals formed. The general An decrease from core
and there also exist intermediate compositions between
to rim can be related to differentiation of the magma.
IIa and IIb, with type IIa being more common than
type IIb.
This is more pronounced in the outer 300 lm of PAR-
130-P3. By contrast, higher average An contents in
zone 3 suggest either a more mafic composition of the
melt or higher water content.
Processes controlling major- and trace-element zoning
Resorption textures very similar to those observed here
have been produced experimentally by heating (Johannes
We first discuss in detail the various factors control- et al. 1994), chemical disequilibrium (Tsuchiyama
ling zoning patterns, applied to the type I crystals, 1985; Nakamura and Shimakita 1998; Hammouda

307
Fig. 4. (Contd)
and Pichavant 2000), and decompression (Nelson 2. Crystals move repeatedly in a gradient towards a
and Montana 1992). The effect of pressure on plagio- hotter, less differentiated magma without any signif-
clase composition is small (20 mol% An for 12 kbar) icant changes in water content.
compared to that of temperature (Housh and Luhr 1991). 3. The magma chamber is periodically replenished with
Thus, repeated and significant decompression is unlikely
to explain observed resorptions. An increase in H2O
content or temperature is a more likely cause. The
most feasible process to produce such rapid changes in
the growth environment is the movement of the crystal in
a melt with a gradient of water content and/or tempera-
ture.
Repeated major resorption surfaces associated with
an abrupt increase followed by a more continuous de-
crease in An at a scale of 100 lm may thus be explained
by three processes.
a hotter, more mafic magma. Only such an event will
be recorded in all crystals present in the magma
chamber at that time.
To distinguish between hypotheses 2 and 3, we need
to find evidence for or against resorptions that are cor-
related between crystals where slightly different patterns
may represent the same event. In our samples the dis-
tance from R1 to the rim, and the degree of resorption in
the crystals are indeed similar. This suggests that the R1
resorption may in fact represent the same event recorded
by all three crystals. If the effect of a water gradient can
1. Crystals move repeatedly in a water concentration be ruled out (see discussion of trace elements), this ob-
gradient towards a more water-rich magma. As water servation favours the hypothesis of magma recharge for
content increases in the melt as a result of crystalli- the R1 resorption. Minor resorption events that cannot
sation, this magma will be also cooler and more dif- be correlated would then suggest individual convective
ferentiated. Such a process can be identified using movement in a thermal and compositional gradient in
minor and trace elements, as discussed below.
between and after the prominent resorption event (R1).

308
Fig. 5a–j. Trace-element varia-
tions vs. An content in plagio-
clase from Parinacota andesite
PAR-130. a–e Plagioclase P3.
Growth-zone numbers corre-
spond to those of Fig. 4a. f–j
Comparison of crystals P1, P3
and P4, including resorption
zones (res). The positive corre-
lation between Fe, Ti and Sr,
Mg from core to rim reflects
magmatic differentiation. By
contrast, the resorption zones
seen in Fig. 3 fall off this trend
to low Fe, Ti and Sr at high An
contents, and probably reflect
changes in water content (see
discussion in text)
Control on minor and trace elements
and the partition coefficients between crystal and
melt. In addition, diffusion in the crystal could
When crystals grow near equilibrium, their concen- homogenise initial variations. The role of these factors
trations in minor and trace elements are controlled in our samples is discussed below for the different
by the concentration of these elements in the melt, elements.

309
Fig. 7. Calculated melt concentration for Ba and Sr based on
element compositions in plagioclase P3 plotted against An content
of plagioclase. Concentrations are calculated using the equations of
Blundy and Wood (1991) and assuming a temperature of 1,000 °C.
Growth-zone numbers as in Fig. 5
creases, whereas Sr decreases with lower An content. This
discrepancy suggests that crystal-chemical control and
the effect of variation in melt concentrations are of dif-
ferent importance for these two elements. Crystallisation
temperatures are estimated on the basis of the Fe–Ti
oxide thermometer (Frost and Lindsley 1991) to range
between 850 and 900 °C. Therefore, a temperature change
from 1,000 to 850 °C is a reasonable assumption for the
crystallisation history. In order to separate crystal-
chemical control from other effects, the Ba and Sr con-
centrations in the melt corresponding to each data point
were calculated following Blundy and Wood (1991) for
these two temperatures. Differences between calculated
melt concentrations at 850 and 1,000 °C are smaller than
analytical uncertainty, so that the effect of temperature
on Ba and Sr partitioning in our sample can be neglected.
The results are shown in Fig. 7, plotted against crystal An
Fig. 6. Sr, Fe and Mg concentrations of all plagioclase crystal
types of PAR-130 plotted against anorthite. At high An content,
types IIa and IIb both have low Fe contents whereas their Sr
contents are low and high respectively
Partition coefficients
Ba and Sr Blundy and Wood (1991) showed that the content, for a temperature of 1,000 °C. Calculated Ba
dominant factors controlling the partition coefficient (K) values (up to 2,200 ppm) are higher than Ba values found
of Sr and Ba in plagioclase are the An content of the in the PAR-130 host rock (Table 2) and in all other Pa-
crystal and temperature, whereas the influence of the rinacota samples (Worner et al. 1988; Davidson et al.
major-element composition of the melt is negligible.
1990). This may be partly explained by the uncertainty in
The empirical equation of Blundy and Wood (1991) the data regression of Blundy and Wood (1991). A data
implies that both Ba and Sr become more compatible as subset containing only their experimental data and in-
the An content of the crystal decreases. As shown in cluding new experimental data of Bindeman et al. (1998)
Fig. 5d, e, Ba concentration in the crystal indeed in- gives a steeper slope and better fit for Ba–temperature

310
correlations. This would result in lower, more realistic Ba Ti Experimental data for Ti show a clear negative cor-
values (=1,600 ppm), consistent with observed bulk-rock relation with An, largely independent of T (Bindeman
compositions. In the following discussion, absolute val- et al. 1998). We approximate this dependence with the
ues must still be considered with caution but general equation ln KTi=–1.75–0.015*An. Calculated Ti con-
trends may be interpreted. Calculated Ba variations in centration in the crystals range between 0.76 0.1% in
the liquid are only slightly larger than the uncertainty, the core and 0.16 0.02% at the rim.
and the trend is qualitatively different from that in the
crystal. Therefore, the overall Ba variations observed in
the crystal reflect crystal-chemical control (An content of
the crystal) on the partition coefficient. Effects of other
Diffusion in the crystal
parameters (melt composition, discussed below) are su-
As trace elements diffuse faster in crystals than major
perimposed and subordinate. By contrast, the calculated
elements, their concentration in plagioclase may not
Sr variations in the liquid are stronger than, and quali-
reflect the melt composition at the time of growth but
tatively similar to the Sr variations observed in the crys-
may have been affected and homogenised by subsequent
tal. The resorption zones are an exception where Sr
diffusion. For andesine compositions at 900 to 1,000 °C,
variations reflect crystal-chemical control.
the diffusivity DSr is in the order of 10–19 to 10–18 m2/s
(Giletti and Casserly 1994; Cherniak and Watson 1994).
Fe Fe partitioning strongly depends on ƒO2 (Longhi et
al. 1976; Phinney 1992; Wilke and Behrens 1999; Su-
gawara 2001). For given PH2O, KFe is a non-linear
The characteristic times for Sr diffusion (T=L2/DSr)
over distances of L=10 lm and L=100 lm are in the
order of 30 and 3,000 years at 900 °C and between 3 and
300 years at 1,000 °C respectively. Estimated growth
rates of plagioclase in natural melts (10–8 to 10–12 cm/s;
Cashman 1990; Davidson and Tepley 1997) imply
growth times between 4 months and 3,000 years for a
1-mm crystal. Sr diffusion time and crystal life time are
thus in the same order of magnitude. As diffusivities and
partition coefficients of Ba sand Sr are known in pla-
gioclase, diffusion in the crystal can be tested for these
two elements. If major-element zoning patterns are
preserved, the equilibrium concentration of Sr and Ba
obtained after diffusion in the crystal should be anti-
correlated with An content, because of the An depen-
dence of the partition coefficients (Zellmer et al. 1999).
Such trends are indeed observed for Sr in the core of P3,
and for Ba over the whole crystal. In order to test
whether this is due to diffusion, we consider Sr concen-
trations in the melt as calculated for Fig. 7 in more de-
tail.
These calculated melt compositions are used here
only as a reference, and represent the concentrations of
hypothetical melts with which each analysed point of the
crystal would have been in equilibrium during growth.
Complete diffusive equilibration of Sr would result in
uniform calculated Sr concentration in the melt. Partial
diffusive equilibration would first affect the small-scale
compositional heterogeneities. The smallest growth lay-
ers would then have constant Ba and Sr equilibrium melt
concentrations at variable An, whereas the composi-
tional variations on a larger scale would remain. Sig-
nificant variation in Sr in the calculated equilibrium melt
concentration occurs at only 50–200 lm scale in zones 1–
function of DƒO2 relatively to FMQ; Wilke and Behrens
1999). Partition coefficients range between 0.03 and 0.36
for basalts at one atmosphere (Phinney 1992), and be-
tween 0.085 and 0.54 for intermediate compositions
under hydrous conditions at 5 kbar (Wilke and Behrens
1999). The data of Bindeman et al. (1998) show a neg-
ative correlation of KFe and An. However, the experi-
ments were done in a narrow range of compositions
where An is likely to be directly correlated to tempera-
ture. In the data set of Sugawara (2001), the correlation
of KFe is better with temperature than with An.
Therefore, DFe probably more strongly depends on
temperature. For ƒO2 between FMQ +0 and +2 and a
P
H2O of more than 5 kbar, KFe is between 0.19 and 0.05.
We adopt an intermediate value of 0.09, which gives a
FeO concentration in the melt of 4.1 wt%. The Fe de-
crease observed in the crystals from core to rim is not
related to changes in partition coefficients, because it
would imply a strong decrease in ƒO2, or an increase in
temperature or An. None of these are likely to occur
during differentiation. Therefore, the measured Fe
content reflects a decrease of Fe concentration in the
melt. At the outermost rim, the situation may be dif-
ferent. Cooling related to eruption and possibly an in-
crease in ƒO2 will increase KFe and could explain the
observed strong increase of Fe. Another possible ex-
planation involving a kinetic effect due to the formation
of a chemical boundary layer at the crystal rim interface
is discussed below. The decrease of KFe predicted for
lower PH2O after ascent is not observed and must be
compensated by the other effects.
Mg Experimental values of KMg in plagioclase vary 5. This indicates that, even in the core, Sr variations at
between 0.025 and 0.05, with only very weak correla- this scale have not been totally erased by diffusion. As
tions with T (Longhi et al. 1976) or An content (Sato Ba diffusion is slower (Cherniak 1999), Ba did not dif-
1989; Bindeman et al. 1998). In the following we con- fuse extensively in the core either. The near-equilibrium
sider a constant partition coefficient of 0.035. Calculated compositions for Sr and Ba at 500-lm scale in the core
MgO concentration in the most mafic melt is are thus not due to diffusive equilibration. Diffusion
1.42 0.5 wt%.
probably smoothed only smallest (=10 lm) oscillations

311
and may have reduced the slope in the Sr–An trends for in Ba with decreasing An content requires the fraction-
the 50–200 lm growth zones.
ation of a K-(and Ba)-rich phase such as biotite or
Following Zellmer et al. (1999), this result can in turn sanidine. Sample PAR-130, however, contains no sani-
be used to infer a maximum residence time of the crystal dine and rare biotites have resorbed, and rounded
in the magma prior to eruption. In the present case, this shapes suggest that they did not crystallise in this melt.
maximum plagioclase residence time in the melt is in the The Ba decrease is also inconsistent with kinetic effects –
order of 3,000 years at 900 °C or only 300 years at Ba, if kinetically controlled, should increase.
1,000 °C.
The change in the degree of differentiation in the host
melt as recorded in the crystals therefore cannot be fully
explained by their crystallisation. A possible scenario to
explain this is progressive mixing with a more differen-
tiated magma. The most plausible origin of such differ-
entiated magma is the boundary layer near the magma
chamber wall where cooling and fractional crystallisa-
tion takes place. The influence of this boundary layer
occurs mostly during the crystallisation of zones 6, 7,
and 8 where the decrease in Ba is most pronounced.
Changes in melt composition Observed trace-element
variations in the plagioclase that cannot be accounted
for by variable partition coefficients or diffusion in the
crystal must reflect variations in the trace-element
composition of the host melt. Possible causes are dif-
ferentiation of the magma due to fractional crystallisa-
tion, movement in a compositional gradient, or recharge
with less differentiated magma. Additionally, kinetic
effects may become important at high growth rates, by
formation of a chemical boundary layer at the crystal–
melt interface. Such a layer would be depleted in pla-
gioclase-compatible elements and enriched in elements
incompatible in plagioclase.
Resorption and mixing with a more mafic magma In
contrast to An content, trace elements allow the
distinction between thermal and chemical causes of
resorption. Hammouda et al. (1996), using Sr isotopes in
partial dissolution of fluorphlogopite-plagioclase pairs,
Differentiation and mixing with a more differentiated showed that the response of the plagioclase crystallising
magma The observed decrease in Sr, Fe, Ti, Mg and An just after resorption to chemical changes in the melt is
content in the plagioclase from core (zone 3) to rim very fast. Thus, in the case of chemical mixing with a less
(zone 9) is qualitatively consistent with melt evolution differentiated magma, a trend reverse to differentiation
expected for an andesite-dacite magma fractionating (i.e. towards higher Fe, Ti, Sr and higher An content) is
plagioclase, pyroxene, Ti–Fe oxides, and small amounts expected in the resorption zone itself as well as in the
of amphibole.
One possible scenario would be to assume that
zones grown subsequently.
Plagioclase growth subsequent to resorption in
changes in host melt trace elements are solely caused by zones 1 to 3, e.g. after R1, is characterised by higher An
crystallisation of the phenocrysts actually present in the and low Ba although the compositional contrast is low
rock. In this case, the resorption surfaces would corre- and the resorptions are weak. High An and increasingly
spond only to physical changes (P, T, water content). lower Ba in zones 1 to 3 (Fig. 4) suggest mixing with
The concentration of Sn, Ti, Fe, and Mg in the less successively more mafic magma. As R1 can be correlated
differentiated melt (in contact with zone 3 of P3 crystal), between crystals, this recharge is common for all type I
calculated using the partition coefficients discussed crystals. Regrowth after resorption in these zones is
above, are consistent with an initial melt composition relatively low in Sr at a given An, which is shown by
equal to or slightly less differentiated than the whole the An–Sr offset (Fig. 5). These observations argue that
rock composition. We thus observe only slightly more recharge is by a relatively Sr-poor magma with an
differentiation of the melt than for closed system, which otherwise low compositional contrast. Therefore, the
requires a small amount of either crystal settling or recharging magma could correspond to a differentiated
influence of a more differentiated magma. However, equivalent of the Sr-poor mafic Ajata andesite (Ginibre
most of the Sr decrease calculated for the melt (700 to et al. 2001).
450 ppm, depending on the temperature) occurs during
The outer growth zones in P3, and in particular
the crystallisation of zones 6 and 7, which only represent zones 7 and 9 immediately outward from the resorp-
less than 10% (in volume) of the crystal. This rapid tion surfaces show high An and trace-element con-
change in Sr is two to three times larger than can be centrations (Fe, Mg, Sr) which then again decrease
explained by in-situ crystallisation of these plagioclase until the next resorption surface. This indicates a
zones. Singer et al. (1995) interpreted similar observa- chemical effect of repeated mixing with mafic magma.
tions as kinetic effects by depletion of a boundary layer The absence of correlation of these resorption surfaces
at the melt–crystal interface due to rapid growth. between crystals (see discussion of An content), and
However, in our case a decrease is also found for Fe, the fact that the compositions lie on the differentiation
which should have the opposite behaviour if kinetically trend suggests that these resorptions, even though they
controlled (see below).
are similar, affected the phenocrysts individually.
The Ba increase with decreasing An content in zones 6 Therefore, they may be caused by convective move-
to 8 is also consistent with the crystallisation of the ments of crystals in a gradient rather than distinct
phases considered above, whereas subsequent decrease recharge events.

312
The resorption zone R0 in P1 and P4, which is large crystallising from a low-Mg and low-Fe differentiated
enough to be analysed, is different. The constant or even melt due to an increase in water content. Because the
lower Fe, Ti, Sr, and Mg concentrations at increasing influences of physical factors on KFe and KMg are dif-
An content indicate no chemical change in the host melt ferent, and KMg is essentially constant, we conclude that
close to the crystal at the time of resorption. The re- the low Fe and Mg concentrations in the crystals should
sorption is thus not due to a bulk chemical change but in fact reflect low concentrations in the melt rather than
rather to an increase in temperature and/or H2O con- variable partition coefficients. By implication, this sug-
tent. On the other hand, when significant dissolution gests that An-rich zones of these crystals grew in a wa-
occurs, a boundary layer may form around the crystal in ter-rich differentiated boundary layer.
which the melt is enriched in the plagioclase component
The variable Sr concentration at high An content
and its compatible elements such as Sr. Regrowth after reflects the influence of different initial magmas. Type IIa
resorption in the calcic rim should thus be also enriched crystals have low Sr concentrations similar to those
in Sr. This is not what we observe, and the low Sr found in the differentiated parts of type I crystals, sug-
content in the resorption zone shows that this effect is gesting the same parent magma. On the contrary, the
limited. To explain the high An, low Sr regrowth after higher Sr concentrations of type IIb crystals at relatively
resorption, two possible explanations remain: (1) a more differentiated compositions inferred from Mg and Fe
mafic melt is present, which may bring the necessary indicate an initial magma with significantly higher Sr
heat but its chemical influence is not recorded directly concentration. Such a magma was in fact erupted from
after resorption, and (2) the resorption is due to an in- Parinacota volcano in the last Ajata stage. Mixing with
crease in water content at otherwise relatively constant this high-Sr magma shortly before eruption is also evi-
melt composition. Because a significantly more mafic denced in Old Cone samples (Ginibre et al. 2001). In
composition is not seen in the zones grown subsequently these samples, Sr concentration in the plagioclase grown
to R0, in P1 and P4 and in the equivalent zone 8 in P3, from the hybrid magma is as high as 4,500 ppm. This is
the latter hypothesis is more likely.
close to what could be expected in plagioclase of An50
crystallising from the high-Sr Ajata andesite (5,000 to
6,000 ppm Sr). Type IIb crystals thus represent re-
maining crystals derived from the differentiation of such
a high-Sr magma.
Crystal margins: kinetic control and changes in partition
coefficients The outermost margin of the crystals corre-
sponds to the final stage of crystallisation during or after
eruption. It is thus expected to have grown faster at
higher cooling rate as well as at lower temperature and
pressure than the rest of the crystal. The change in An
oscillation mode at the outer rim to narrow and high-
amplitude oscillations is typical for this final growth
phase.
Crystallization history of dacite PAR-130
and magma chamber dynamics
The early growth history of type I crystals in andesite
PAR-130 is characterised by successive recharge events of
an only slightly more mafic, Sr-poor magma. This
recharge magma could be a differentiated equivalent of
the Sr-poor mafic andesite end member, which has been
identified as one of the parent magmas to Parinacota
rocks. Following this, the next growth phase documents
increased differentiation and the influence of a water-rich,
differentiated melt, possible representing the magma
chamber’s boundary layer at the cooler wall. Type II
crystals have also partly formed in such an environment.
Superimposed on the differentiation trend, quasi-
cyclic resorption events followed by a shift in major-and
minor-element concentrations toward less differentiated
compositions reflect convective movements in a com-
positional and/or temperature gradient. These move-
ments would be specific to each crystal but zonation and
resorption patterns would become successively more
similar towards the rim. This indicates the convergence
of crystallisation histories. The last resorption R0 rec-
ognised in several type I crystals shows little evidence for
a chemical effect of a new mafic magma after resorption.
Comparison with type IIa crystals suggests that this re-
sorption may also have been caused by an H2O increase
close to the magma boundary layer. This is consistent
with the increased influence of the H2O-rich boundary
A chemical boundary layer may form at the crystal–
melt interface at high growth rate (Bottinga et al. 1966).
Elements which are incompatible in plagioclase (Fe, Ba,
Mg, Ti) will be enriched in this layer whereas compatible
Sr will be depleted. This can qualitatively explain the
increase in Fe, Mg, Ti, and Ba and the decrease in Sr
observed in the outermost margin of crystal P3.
However, the effect is very strong for Fe and weak for
Mg, which differs from the observations of Bottinga et
al. (1966) who reported similar gradients for Mg and Fe
in the melt. This suggests the influence of other factors
for Fe, where a temperature increase and PH2O decrease
cause an increase in Fe partition coefficient. Possibly an
increase in ƒO2 at this stage also plays a role. These
factors have little effect on Ti and Mg, which explains
the difference.
Type II crystals
The smaller and more abundant type II plagioclase
phenocrysts show low Fe and Mg contents at high An.
This can be explained by two processes: (1) lower par-
tition coefficients for Mg and Fe at constant melt com-
position, or (2) increase in An content of the plagioclase

313
layer towards the end of the crystallisation history in
type I crystals.
Type IIa plagioclase crystals derive from a similar
thermal regimes independent of whole-sale mixing
and mafic is a process that we call ‘‘convective
crystal dispersion’’.
parental magma in the boundary layer, whereas type IIb 4. The amount of differentiation recorded in type I
crystals must be older, as represented by cores grown
from the high-Sr end-member magma.
plagioclase is only slightly larger than the closed-
system differentiation and involves mixing from the
differentiated boundary layer. This indicates limited
crystal fractionation by settling, and high crystal re-
tention in the magma. Settling and re-entrainment of
crystals in a convecting magma body depends on
magma viscosity, and high crystal retention is ex-
pected only in very viscous magma (Marsh and
Maxey 1985; Martin and Nokes 1988; Solomatov
et al. 1993; Hawkesworth et al. 2000). However, our
results show that the time scale of crystal settling
must be large compared to that of melt differentiation
by crystallisation.
Magma chamber structure and dynamics
Our study shows four important aspects of the dynamics
of an andesitic/dacitic magma chamber.
1. The zoning patterns in plagioclase record the
existence of gradients (thermal, compositional,
water content) between a main magma body and a
relatively differentiated, crystal-rich boundary layer,
containing older crystals. The water gradient
implies that the main magma body is not water
saturated.
2. Crystallisation occurs both in the main magma body
(type I crystals) and in the crystallising boundary
layer (type II crystals). The chemical influence of
differentiation in the boundary layer is also seen, but
less markedly, in type I crystals. This result may be
compared with that of Kuritani (1998) who found for
a basaltic magma system that the number of crystals
from the boundary layer is small compared to that
from the main magma body, but the thickness and
An contrast of the corresponding growth zones are
higher. In our sample the proportion of crystals in-
fluenced by the boundary layer is larger (over 50%)
but the effect is smaller. This can be explained by the
higher viscosity of andesitic/dacitic magmas: only the
crystals from the interface of a stable crystal mush
and the main magma body may be brought into the
main magma body, whereas deeper layers could be
sampled by a less viscous basaltic magma. The in-
creasing influence of the boundary layer may indicate
an origin near to the roof, and possibly also the in-
fluence of liquid differentiation of the melt at the wall,
as suggested by Turner (1980) and McBirney et al.
(1985).
3. The high frequency of resorption in crystals from
the main magma body indicates the role of con-
vection without evidence of significant mafic re-
charge. The record of compositionally and thermally
variable crystallisation conditions requires not only
movement of crystals in gradients but also sufficient
local mixing. Convection and mixing must also have
produced small-scale heterogeneities which persisted
over a time sufficiently long for the crystal to grow
and to record these variations. This can hardly oc-
cur under conditions of laminar convection, as
proposed by Singer et al. (1995). It requires some
turbulence, even though this is not expected from
the viscosities of a dacitic magma. Repeated move-
Time scale of crystallisation and of magma evolution
The dominant differentiation history recorded in type I
crystals of sample PAR-130 is at contrast with the data
shown by Singer et al. (1995) at San Pedro volcano
(central Chile), which clearly show open-system evolu-
tion. Temporal constraints from Sr diffusion suggest a
differentiation time shorter than a few thousand years.
This is close to the time scale calculated by Hawkes-
worth et al. (2000) for 50% closed-system differentia-
tion of a 10-km3 magma body releasing heat with a
power output of 100 MW. This is consistent with our
result that the evolution of PAR-130 after the initial
recharge is not very far from closed-system crystalli-
sation.
Growth rates of the crystal from in the main magma
body may be calculated from the residence time. The
euhedral zone boundaries of crystal P3 suggest repeated
but restricted resorption. It is reasonable to assume that
the time elapsed between growth episodes is of the order
of magnitude of the growth time itself. In this case, the
growth of a 2-mm crystal in less than a few hundred or
thousand years implies minimum growth rates of 10–10
or 10–11 cm/s. This is consistent with growth rates re-
ported in the literature (Cashman 1990).
High-Sr type IIb crystals and the low-Sr recharge in
type I crystals show the influence, even at low magma
recharge rates, of both high-Sr and low-Sr Parinacota
magmas. Thus, the two parental magmas documented
from volcano evolution and bulk-rock analyses (Worner
et al. 1988) may actually be recorded during plagioclase
growth in a single rock sample.
Acknowledgements We thank Jon Davidson and George Bergantz
for useful discussions. Kurt Knesel and an anonymous reviewer are
acknowledged for their constructive review and thought-provoking
critical comments. This study was part of the DFG-funded
ment of crystals between different compositional and SFB 468.

314
Housh TB, Luhr JF (1991) Plagioclase-melt equilibria in hydrous
system. Am Mineral 76:477–492
Huppert HE, Turner JS (1991) Comment on ‘On convective style
and vigor in sheet like magma chambers’ by Bruce D. Marsh.
J Petrol 32:851–854
Johannes W, Koepke J, Behrens H (1994) Partial melting reactions
of plagioclases and plagioclase bearing systems. In: Parson I (ed)
Feldspars and their reactions. Kluwer, Dordrecht, pp 161–194
Kuritani T (1998) Boundary layer cristallization in a basaltic
magma chamber: Evidence from Rishiri Volcano, northern
Japan. J Petrol 39:1619–1640
Lasaga AC (1982) Toward a master equation in crystal growth. Am
J Sci 282:1264–1288
L’Heureux I, Fowler AD (1996) Isothermal constitutive underco-
oling as a model for oscillatory zoning in plagioclase. Can
Mineral 34:1137–1147
Lofgren GE (1974) An experimental study of plagioclase crystal
morphology: isothermal crystallization. Am J Sci 274:243–273
Longhi J, Walker D, Hays JF (1976) Fe and Mg in plagioclase. In:
Proc 7th Lunar Sci Conf, pp 1281–1300
Loomis TP (1982) Numerical simulations of crystallization pro-
cesses of plagioclase in complex melts: the origin of major and
oscillatory zoning in plagioclase. Contrib Mineral Petrol
81:219–229
Marsh BD (1989a) Magma chambers. Annu Rev Earth Planet Sci
17:439–474
Marsh BD (1989b) On convective style and vigor in sheet like
magma chambers. J Petrol 30:479–530
Marsh BD, Maxey MR (1985) On the distribution and separation
of crystals in convecting magma. J Volcanol Geotherm Res
24:95–150
Martin D, Nokes R (1988) Crystal settling in a vigorously con-
vecting magma chamber. Nature 332:534–536
McBirney AR, Baker BH, Nilson RH (1985) Liquid fractionation.
Part I. Basic principles and experimental simulations. J Volca-
nol Geotherm Res 24:1–24
Nakamura M, Shimakita S (1998) Dissolution origin and syn-en-
trapment compositional changes of melt inclusions in plagio-
clase. Earth Planet Sci Lett 161:119–133
Nelson ST, Montana A (1992) Sieve-textured plagioclases in vol-
canic rocks produced by rapid decompression. Am Mineral
77:1242–1249
Nixon GT, Pearce TH (1987) Laser-interferometry study of oscil-
latory zoning in plagioclase: The record of magma mixing and
phenocryst recycling in calc-alkaline magma chambers, Iz-
taccıhuatl Volcano, Mexico. Am Mineral 72:1144–1162
Pearce TH, Russel, JK, Wolfson I (1987) Laser-interference and
Nomarski interference imaging of zoning profiles in plagioc-
lases phenocrysts from the May 18, 1980, eruption of Mount St.
Helens, Washington. Am Mineral 72:1131–1143
Phinney WC (1992) Partitioning coefficients for iron between pla-
gioclase and basalt as a function of oxygen fugacity: implica-
tions for Archean and lunar anorthosites. Geochim Cosmochim
Acta 56:1885–1895
Sato H (1989) Mg-Fe partitioning between plagioclase and liquids
in basalts of Hole 504B, ODP leg 111: a study of melting at one
atm. College Station, TX, Proc Ocean Drilling Program Sci Res
111:17–26
Singer B, Dungan MA, Layne GD (1995) Textures and Sr, Ba, Mg,
Fe, K, and Ti compositional profiles in volcanic plagioclase:
Clues to the dynamics of calc-alkaline magma chambers. Am
Mineral 80:776–798
Solomatov VS, Olson P, Stevenson DJ (1993) Entrainment from a
bed of particles by thermal convection. Earth Planet Sci Lett
120:387–393
Sugawara T (2001) Ferric iron partitioning between plagioclase and
silicate liquid: thermodynamics and petrological applications.
Contrib Mineral Petrol 141:659–686
Tepley FJ III, Davidson JP, Tilling RI, Arth JG (2000) Magma
mixing, recharge and eruption histories recorded in plagioclase
phenocrysts from El Chichon Volcano, Mexico. J Petrol
41:1397–1411
References
Allegre CJ, Provost A, Jaupart C (1981) Oscillatory zoning: a
pathological case of crystal growth. Nature 294:223–228
Anderson AT Jr (1983) Oscillatory zoning of plagioclases:
Nomarski interference contrast microscopy of etched polished
sections. Am Mineral 68:125–129
Anderson AT (1984) Probable relations between plagioclase zoning
and magma dynamics, Fuego Volcano, Guatemala. Am Min-
eral 69:660–676
Bindeman IN, Davis AM, Drake MJ (1998) Ion microprobe study
of plagioclase – basalt partition experiments at natural con-
centration level of trace elements. Cosmochim Geochim Acta
62:1175–1193
Blundy JD, Shimizu N (1991) Trace element evidence for plagio-
clase recycling in calc-alkaline magmas. Earth Planet Sci Lett
102:178–197
Blundy JD, Wood BJ (1991). Crystal-chemical controls on the
partitioning of Sr and Ba between plagioclase feldspar, silicate
melts, and hydrothermal solutions. Geochim Cosmochim Acta
55:193–209
Bottinga Y, Kudo A, Weill D (1966) Some observation on oscil-
latory zoning and crystallization of magmatic plagioclase. Am
Mineral 51:792–806
Brophy JG, Dorais MJ, Donnelly-Nolan J, Singer B (1996)
Plagioclase zonation in hornblende gabbro inclusions from
Little Glass Mountain, Medicine Lake volcano, California:
implications for fractionation mechanisms and the formation of
composition gaps. Contrib Mineral Petrol 126:121–136
Cashman KV (1990) Textural constraints on the kinetics of crys-
tallization of igneous rocks. In: Nicholls J, Russell JK (eds)
Modern methods of igneous petrology, understanding mag-
matic processes. Mineral Soc Am, Washington, DC. Rev
Mineral 24:259–314
Cherniak DJ (1999) Ba diffusion in Feldspars. In: AGU Fall Meet
1999, EOS Trans Am Geophys Union 80 suppl 46:F1078
Cherniak DJ, Watson B (1994) A study of diffusion in plagioclase
using Rutherford backscattering spectrometry. Geochim Cos-
mochim Acta 58:5179–5190
Davidson JP, Tepley FJ (1997) Recharge in volcanic system: evi-
dence from isotope profiles of phenocrysts. Science 275:826–829
Davidson JP, MacMillan NJ, Moorbath S, Worner G, Harmon RS,
Lopez-Escobar L (1990) The Nevados de Payachata volcanic
region (18°S/69°W, N. Chile) II. Evidence for widespread crustal
involvement in Andean magmatism. Contrib Mineral Petrol
105:412–432
Frost BR, Lindsley DH (1991) Occurrence of iron-titanium oxides
in igneous rocks. In: Lindsley DH (ed) Oxide mineral: petro-
logic and magnetic significance. Rev Mineral 25
Giletti BJ, Casserly JED (1994) Strontium diffusion kinetics in
plagioclase feldspars. Geochim Cosmochim Acta 58:3785–3793
Ginibre C, Kronz A, Worner G (2002) High-resolution quantita-
tive imaging of plagioclase composition using accumulated
back-scattered electron images: new constraints on oscillatory
zoning. Contrib Mineral Petrol 142:436–448
Ginibre C, Worner G, Kronz A (2001) Magma chamber processes
reflected in major, minor and trace element plagioclase zoning
(Parinacota volcano, Chile). J Conf Abstr 6:807
Grove TL, Baker MB, Kinzler RJ (1981) Coupled CaAl-NaSi
diffusion in plagioclase feldspar: Experiments and application
to cooling rate speedometry. Geochim Cosmochim Acta
58:2113–2121
Hammouda T, Pichavant M (2000) Melting of fluorophlogopite-
plagioclase pairs at 1 atmosphere. Eur J Mineral 12:315–328
Hammouda T, Pichavant M, Chaussidon M (1996) Isotopic
equilibration during partial melting: an experimental test of the
behavior of Sr. Earth Planet Sci Lett 144:109–121
Hawkesworth CJ, Blake S, Evans P, Hughes R, Macdonald R,
Thomas LE, Turner SP, Zellmer G (2000) Time scales of crystal
fractionation in magma chambers – integrating physical, iso-
topic and geochemical perspectives. J Petrol 41:991–1006

315
Tsuchiyama A (1985) Dissolution kinetics of plagioclase in the melt
of system diopside-albite-anorthite, and origin of dusty pla-
gioclases in andesites. Contrib Mineral Petrol 89:1–16
Turner JS (1980) A fluid-dynamical model of differentiation and
layering in magma chambers. Nature 285:213–215
Wilke M, Behrends H (1999) The dependence of the partitioning of
iron and europium between plagioclase and hydrous tonalitic
melt on oxygen fugacity. Contrib Mineral Petrol 137:102–114
Worner G, Harmon RS, Davidson J, Moorbath S, Turner DL,
McMillan N, Nye C, Lopez-Escobar L, Moreno H (1988) The
Nevados de Payachata volcanic region (18°S/69°W, N. Chile).
I. Geological, geochemical, and isotopic observations. Bull
Volcanol 50:287–303
Worner G, Hammerschmidt K, Henjes-Kunst F, Lezaun J, Wilke H
(2000) Geochronology (40Ar/39Ar, K-Ar, and He-exposure-)
ages of cenozoic magmatic rocks from Northern Chile(18°-
20°S): Implications for magmatism and tectonic evolution of the
central Andes. Rev Geol Chile 27:205–240
Zellmer GF, Blake S, Vance D, Hawkeswoth C, Turner S (1999)
Plagioclase residence times at two island arc volcanoes (Kameni
Islands, Santorini, and Soufriere, St. Vincent) determined by Sr
diffusion systematics. Contrib Mineral Petrol 136:345–357