Full text of DOI:10.1007/s004100000214

Contrib Mineral Petrol 12001) 140: 680±700
DOI 10.1007/s004100000214
A. K. Schmitt á S. L. de Silva á R. B. Trumbull
R. Emmermann
Magma evolution in the Purico ignimbrite complex, northern Chile:
evidence for zoning of a dacitic magma by injection of rhyolitic melts
following ma®c recharge
Received: 4 May 2000 / Accepted: 18 October 2000 / Published online: 16 January 2001
Ó Springer-Verlag 2001
Abstract The 1.3 Ma Purico complex is part of crystal-rich dacitic pumice 164±66 wt% SiO2). Rare
an extensive Neogene±Pleistocene ignimbrite province andesitic and banded pumice 160±61 wt% SiO2) are
in the central Andes. Like most other silicic complexes also present in the uppermost part of the ¯ow unit. The
in the province, Purico is dominated by monotonous di€erent compositional groups of pumice in LPI II ¯ow
intermediate ash-¯ow sheets and has volumetrically unit 1rhyolite, andesite, dacite) have initial Nd and Sr
minor lava domes. The Purico ignimbrites 1total vol- isotopic compositions that are indistinguishable from
ume 80±100 km3) are divided into a Lower Purico Ig- each other and from the dominant dacitic pumice
nimbrite 1LPI) with two extensive ¯ow units, LPI I and 1ꢀNd=)6.7 to )7.2 and 87Sr/86Sr=0.7085±0.7090).
LPI II; and a smaller Upper Purico Ignimbrite 1UPI) However, two lines of evidence show that the andesite,
unit. Crystal-rich dacite is the dominant lithology in all dacite and rhyolite pumices do not represent a simple
the Purico ignimbrites and in the lava domes. It is es- fractionation series. First, melt inclusions trapped in
sentially the only lithology present in the ®rst LPI ¯ow sequential growth zones of zoned plagioclase grains in
unit 1LPI I) and in the Upper Purico Ignimbrite, but the rhyolite record fractionation trends in the melt that
the LPI II ¯ow unit is unusual for its compositional diverge from those shown by dacite samples. Second,
diversity. It constitutes a stratigraphic sequence with a mineral equilibrium geothermometry reveals that da-
basal fall-out deposit containing rhyolitic pumice 168± cites from all ignimbrite ¯ow units and from the domes
74 wt% SiO2) overlain by ignimbrite with dominant had relatively uniform and moderate pre-eruptive
temperatures 1780±800 °C), whereas the rhyolites and
andesites yield consistently higher temperatures 1850±
950 °C). Hornblende geobarometry and pressure con-
straints from H2O and CO2 contents in melt inclusions
indicate upper crustal 14±8 km) magma storage condi-
tions. The petrologic evidence from the LPI II system
Electronic supplementary material to this paper can be obtained
electronically by using the SPRINGER LINK server located at
http://dx.doi.org/10.1007/s004100000214.
A. K. Schmitt 1&)
GeoForschungsZentrum Potsdam, Telegrafenberg,
14473 Potsdam, Germany
thus indicates an anomalously zoned magma chamber
with a rhyolitic cap that was hotter than, and chemi-
cally unrelated to, the underlying dacite. We suggest
that the hotter rhyolite and andesite magmas are both
related to an episode of replenishment in the dacitic
Purico magma chamber. Rapid and e€ective crystal
fractionation of the fresh andesite produced a hot
rhyolitic melt whose low density and viscosity permit-
ted ascent through the chamber without signi®cant
thermal and chemical equilibration with the resident
dacite. Isotopic and compositional variations in the
Purico system are typical of those seen throughout the
Neogene ignimbrite complexes of the Central Andes.
These characteristics were generated at moderate
crustal depths 1<30 km) by crustal melting, mixing and
homogenization involving mantle-derived basalts. For
the Purico system, assimilation of at least 30% mantle-
derived material is required.
S. L. de Silva
Dept. of Geography, Geology, and Anthropology,
Indiana State University, Terre Haute, IN 47809, USA
R. B. Trumbull
GeoForschungsZentrum Potsdam, Telegrafenberg,
14473 Potsdam, Germany
R. Emmermann
Institut fur Geowissenschaften und Lithospharenforschung,
Universitat Giessen, 35390 Giessen, Germany
A. K. Schmitt
Present address:
Dept. of Earth and Space Sciences,
University of California Los Angeles,
Los Angeles, CA 90095, USA
e-mail: axel@argon.ess.ucla.edu
Tel.: +1-310-7945047; Fax: +1-310-8252770
Editorial responsibility: J. Hoefs

681
Introduction
Long-lived episodes of large-scale explosive silicic vol-
canism mark an important evolutionary stage in the
development of continental arcs, e.g., western USA
1Lipman et al. 1978; Gans et al. 1989; Best et al. 1995),
Sierra Madre Occidental in Mexico 1McDowell et al.
1979) and Altiplano-Puna Volcanic Complex, central
Andes 1de Silva 1989a, 1989b). The climactic eruptions
that characterize such episodes typically result in the
formation of large caldera complexes and deposition of
large-volume regionally extensive ignimbrites. Such
events are also viewed as the surface manifestation of
large-scale plutonic activity at depth. The tectono-mag-
matic processes that lead to such so-called ignimbrite
``¯are-ups'' 1Elston 1984; de Silva 1989a) are not well
understood but they are clearly important for heat and
mass transfer within the crust, and modi®cation of
crustal composition and structure.
Although many ignimbrites show evidence of com-
positional zonation or variability, the largest volume ig-
nimbrites rarely do. These are the so-called monotonous
intermediates 1Smith 1979; Hildreth 1981; de Silva 1991).
Hildreth 11981) convincingly argued that the lower re-
gions of crustal silicic magma systems are generally ba-
saltic and that the degree of zonation displayed may be a
function of eruptive draw-up, i.e., large systems appear
homogeneous because the lower, more ma®c layers are
rarely evacuated during eruption. Basic recharge may
therefore be a fundamental process in large silicic mag-
matic systems even if only a few studies have conclusively
demonstrated its role in producing compositional heter-
ogeneity and in¯uencing the evolution of the system.
The late Tertiary Altiplano±Puna Volcanic Complex
1APVC) between 21 and 24°Sin the central Andes is
exceptionally well preserved because of the prevailing
arid climate. The volcanic output of the last 10 million
years is volumetrically dominated by crystal-rich dacitic
ignimbrites and lavas that range in volume from 10s±
100s of km3, to >1000 km3 and form regionally extensive
sheets 1Fig. 1). The largest ignimbrite sheets were erupted
from giant resurgent caldera complexes 1e.g., Sparks
et al. 1985; Gardeweg and Ramirez 1987; de Silva and
Francis 1991; Lindsay et al. 2000). Some moderate vol-
ume dacitic ignimbrites are not associated with typical
caldera morphologies but form ignimbrite ``shields''
1Francis and Baker 1978; de Silva and Francis 1991),
which are poorly understood as a group 1e.g., Lipman
1997). Most APVC ignimbrites are homogeneous; only
rarely have rhyolitic units been found in the erupted se-
quences, and andesitic compositions are likewise scarce
1de Silva 1991; Lindsay et al. 2000). The Purico complex
Fig. 1 Map of the distribution of Neogene and Pleistocene
ignimbrites in the central Andes. Major calderas and Pleistocene-
recent stratovolcanoes of the active arc are also shown 1redrawn
after de Silva and Francis 1991). The extent of the Altiplano±Puna
Volcanic Complex 1APVC, de Silva 1991) is outlined
motivated several petrological and geochemical studies
of the Purico complex. These have debated the relative
roles of crustal anatexis, closed-system and open-system
processes in the origin and evolution of the magmas
1Hawkesworth et al. 1982; Francis et al. 1984; Davidson
et al. 1990). In a reconnaissance study, de Silva 11991)
argued that the compositional zonation at Purico was
largely the result of closed system fractional crystalliza-
tion 1convective fractionation). This paper presents the
results of a geochemical and petrologic re-examination of
the Purico complex based on a stratigraphically well-
constrained suite of samples for which we provide new
whole-rock major and trace element data. Mineral and
glass phases 1matrix and melt inclusions) were analyzed
to determine pre-eruptive temperature, pressure, and
volatile contents. The results suggest that ma®c magma
recharge was instrumental in the development of the
compositional heterogeneity and perhaps also in trig-
gering the eruption. Finally, we examine the petrogenesis
of the Purico magmatic center in the broader context of
large-scale silicic magmatism in the central Andes.
in Northern Chile 1Figs. 1 and 2) is an important The Purico complex
exception wherein rhyolitic and andesitic compositions
were erupted along with the dominant dacite 1Francis Geology
et al. 1984). This, coupled with the fact that the ig-
nimbrites are well exposed and the compositional range The Purico complex 167°45¢W and 22°57¢S) comprises
of the juvenile material spans that of the APVC, has an approximately circular apron of three ignimbrite ¯ow

682
Fig. 2 Geological map, pro®le
and composite stratigraphic
column of the Purico Complex.
Map and pro®le made from
digital elevation model and
Landsat satellite imagery.
Inferred vent area is located
2 km SW of Dome D. Sample
locations for the Purico
ignimbrites and sections in
stratigraphic column: A
Quebrada Corrales Blancos; B
west of Cerro Purico; C Pampa
del Vallecito; D west of Cerro
Chainator; E Quebrada
Vizcachas; F Quebrada Zapar;
G Quebrada Silapeti
units that are presently exposed over an area of ꢀ15± types. Dome D is a ``torta'', a ¯at pancake-shaped ¯ow
25 km around the putative vent region. The units dip dome 1see Fig. 2) that yielded a biotite K±Ar age within
away from the vent region 15,000 m above sea level) at the age range of the ignimbrites 10.98‹0.05 Ma; Francis
ꢀ4°, the most prominent dip slope being to the west into et al. 1984). The Dome D lavas closely resemble the
the Salar de Atacama basin 1Fig. 2). This gives the dacite pumice from the ignimbrite in terms of age,
complex its broad shield-like form. The Purico ignimb- mineralogy and chemical composition, suggesting a
rites overlie a major sequence of older ignimbrites that close genetic relationship for the respective magmas. The
are in part related to the 4.5±4.1 Ma eruptions of the other domes 1Cerros Purico, Negro, Putas, Chascon,
nearby La Pacana caldera 1Gardeweg and Ramirez Aspero, and El Cerillo) have conical shapes typical of
1987; Lindsay et al. 2000). Published K±Ar ages for the Pelean domes in the sense of Blake 11990). Radiometric
Purico ignimbrite range between 1.38‹0.07 and 0.87‹ dating of these domes is lacking but the Chascon, Asp-
0.52 Ma 1Francis et al. 1984; de Silva 1989b), the ages ero, and El Cerillo domes are clearly post-glacial 1Ho-
overlapping for the di€erent ¯ow-units.
locene) and their degree of preservation suggests that all
Several dacitic lava domes overly the Purico ig- are probably <0.5 Ma old 1Francis et al. 1984). Olivine-
nimbrite in the summit region 1Dome D, Cerro Purico bearing ma®c inclusions are common in the dacite lavas,
and El Cerillo), partly surrounding the putative vent for and they have been interpreted as evidence for magma
the ignimbrite eruption. On the basis of age and mor- mingling caused by ma®c recharge 1Francis et al. 1984;
phology, these domes can be characterized as one of two Davidson et al. 1990). Because the young, conical domes

683
are compositionally distinct from the Purico ignimbrites
The total volume of the Purico ignimbrites has been
and the associated Dome D 1see Table 1), they are at- estimated at ꢀ80±100 km3 1de Silva 1991). The rhyolitic
tributed to a separate magmatic event initiated by input fall deposit at the base of LPI II was encountered in both
of new ma®c magma 1Hawkesworth et al. 1982; of the two accessible localities of the ignimbrite on the
Davidson et al. 1990; Tepley and Davidson 1997).
western slope of the Purico shield 1A and B in Fig. 2)
and we assume that it extends throughout this local area.
Such fall deposits are commonly associated with a ¯ow
unit and the fact that rhyolite pumice is found in the
overlying ash ¯ow unit, along with the observation that
Stratigraphy of the Purico ignimbrites
We distinguish an Upper 1UPI) and Lower Purico the fall deposit thickens towards the putative vent, gives
Ignimbrite, and further subdivide the Lower Purico us con®dence that it is part of the LPI II eruption. If the
Ignimbrite 1LPI) into two ¯ow-units designated as LPI I distribution of the fall deposit coincides with the that of
and LPI II for the lower and upper unit, respectively the ignimbrite 11,500 km2), its total volume is estimated
1Fig. 2). The most complete section of the Lower Purico to be 0.4 km3. The volume of dark pumice is hard to
Ignimbrite crops out ꢀ10 km east of the village of San assess given its scarcity in the outcrop, but it is probably
Pedro de Atacama, in quebradas that have unfortu- ꢁ1% of the total volume of LPI II. The volume of the
nately been mined by the Chilean army, making access domes is small compared with the ignimbrite and esti-
to the outcrops interesting.
mates for individual domes range from 0.36 km3 for
LPI I is a massive ¯ow unit whereas LPI II comprises a Aspero to ꢀ 4 km3 for Chascon 1de Silva and Francis
basal surge layer and a normally graded pumice fall de- 1991).
posit, overlain by a pumice-rich ¯ow 1Fig. 2). Coarse-
grained, dacitic crystal-rich pumice 1pl±hbl±bt±qtz‹opx,
accessory apa±zrn±sph) is the dominant juvenile com-
ponent in both LPI ¯ow units. Also present in both ¯ow
units are crystal-rich aggregates with a phenocryst as-
semblage compositionally indistinguishable from the
dacite pumice. These aggregates occur as inclusions in
dacitic pumice or as individual clasts in the ignimbrite.
The LPI II fall deposit is formed by crystal-poor rhyolitic
1sensu lato, see below) pumice 1pl±bt‹hbl‹opx, acces-
sory apa±zrn), which lacks quartz. The base of the
overlying LPI II ash ¯ow also contains clasts of rhyolitic
pumice in addition to the dominant dacitic pumice. Dark
andesitic and banded dacitic±andesitic pumices 1pl±cpx±
opx, accessory apa±zrn) appear in the LPI II ¯ow unit
and increase in abundance towards the upper part, as
does a quartz-poor dacitic variety.
In the Purico summit region close to the putative
vent, LPI II is overlain by the matrix-supported Upper
Purico ignimbrite 1UPI). This moderate to densely
welded ignimbrite contains dacitic clasts similar to the
LPI II dacitic pumice. Details of the stratigraphic dis-
tribution, vesicularity, crystallinity, and modal compo-
sition of all pumice types are given in Table 1 and Fig. 2.
Petrology and geochemistry
Whole-rock geochemistry
The emphasis of this study is on the Purico ignimbrites
and is based on a total of 36 pumice samples from LPI I,
LPI II and UPI 1depending on clast size, the samples are
single or composite). Sampling covers the vertical and
horizontal extent of the deposits 1Fig. 2). An additional 13
samples of lava and ma®c inclusions from selected domes
were included for comparison. We consider the pumice
compositions to be representative of magma composi-
tions with the exception of the banded pumices, which
resulted from magma mingling, and the crystal-rich
inclusions, which are interpreted as cumulates 1de Silva
1989c). Bulk matrix samples were excluded because of the
likelihood of mechanical fractionation or contamination
during eruption and transport. Representative analyses
are presented in Table 2 and the full data set is reported in
Table 1 of the electronic supplementary material.
The Purico pumices and lavas de®ne a high-K calc-
alkaline trend with a total range of SiO2 from 60 to
Table 1 Vesicularity, phenocryst and glass abundances of Purico±Chascon pumice types and lavas 1minimal and maximal values in
brackets). Accessory minerals present: zircon, apatite and sphene, the latter only in dacites. See Appendix for analytical methods
Type
Unit
Dacitic pumice
LPI I and II
Crystal-rich inclusion Rhyolitic pumice Andesitic pumice Dacitic lava Dacitic lava
Dome D
LPI I and II
LPI II
LPI II
Chascon-Aspero
Total of samples
Vesicles 1vol%)
8
65
3
56
4
72
6
39
5
24
4
±
Glass 1wt%)
Pl
Hbl
Qtz
Bt
Opx
Cpx
Others
50 141,62)
70 166,72)
14 110,17)
11 17,14)
5 14,9)
<2
±
<2 Ti-Mag
32 131,33)
68 160,76)
23 120,28)
8 18,8)
6 14,11)
±
80 178,84)
100
<2
±
<2
<2
±
<2 Ti-Mag
50 134,56)
56 149,60)
±
±
±
26 119,30)
16 111,24)
2 Ti-Mag
50 134,58)
75 166,80)
14 110,20)
7 15,11)
5 14,6)
30 123,41)
66 161,71)
11 17,19)
8 15,10)
<2
15 111,22)
<2
±
<2 Ti-Mag
± ±
<2 Ti-Mag <2 Ti-Mag

684
Table 2 Major, trace element, Sr and Nd isotopic composition of given for comparison 1major elements by EPMA. Water and trace
Purico pumice types 1for locations see Fig. 2). Oxides in wt%, element data on glasses determined by SIMS. ±Not analyzed; dp
traces in ppm. Bulk rock analyses by XRF 1Zr, Ba), ICP-MS1Rb, dacitic pumice; dl dacitic lava; ci crystal-rich inclusion; rp rhyolitic
Sr, Y, Nb, Th, U) and ICP-AES 1REE). Matrix glass analyses are pumice; ap andesitic pumice; b.d. below detection)
Sample
Unit
Type
PED962H
LPI II/A
dp
PED979
dome D
dl
PED962D
LPI II/A
ci
PED972
LPI II/A
rp
PED96H2B
LPI II/B
ap
w.r.
glass
74.1
0.16
11.6
0.35
0.04
0.10
0.75
5.31
2.68
0.03
2.68
±
97.80
252
79
9
76
10
348
29
51
±
w.r.
glass
73.1
0.06
11.5
0.57
0.05
0.03
0.48
4.64
3.74
b.d.
2.84
±
97.00
203
57
5
47
5
195
16
23
±
±
0.6
21
8
±
w.r.
glass
73.5
0.10
11.2
0.73
0.02
0.02
0.56
4.66
3.20
0.11
3.30
±
97.45
±
±
±
±
±
±
±
±
±
±
±
±
±
w.r.
glass
71.2
0.19
13.3
1.20
0.06
0.30
1.45
3.42
3.40
0.03
2.64
±
w.r.
glass
67.4
0.63
13.9
2.99
0.05
0.85
2.94
3.42
4.02
0.11
1.26
±
SiO2
TiO2
Al2O3
Fe2O3 total
MnO
MgO
CaO
K2O
Na2O
P2O5
H2O
CO2
Total
Rb
Sr
Y
Zr
Nb
Ba
La
Ce
64.3
0.64
15.2
4.42
0.07
1.94
4.19
3.12
2.94
0.17
1.66
0.07
98.70
172
323
17
140
12
474
30
61
25
66.1
0.58
15.5
4.03
0.07
1.73
3.79
3.48
3.11
0.14
1.17
0.05
99.74
176
285
18
137
14
456
33
64
26
5.0
58.4
0.90
16.7
6.62
0.11
2.97
5.60
2.84
3.26
0.24
1.84
0.06
99.60
156
357
22
168
14
481
34
72
33
66.4
0.46
15.3
3.07
0.08
0.90
3.25
3.18
3.54
0.18
3.21
0.13
99.68
121
362
23
211
13
752
32
60
29
60.6
0.81
16.0
5.78
0.09
2.98
5.46
2.49
3.16
0.18
1.90
0.06
99.57
105
339
22
146
12
431
27
55
25
97.28
109
97.63
144
169
18
102
9
557
25
51
±
195
25
201
9
504
27
59
±
Nd
Sm
Eu
Th
4.8
±
6.8
1.4
18
7.3
0.708545
0.512280
6.0
1.2
9.0
2.6
0.709008
0.512273
±
4.9
1.1
13
4.7
0.708913
0.512294
±
0.98
20
8.5
0.708670
0.512270
0.9
30
13
0.99
24
9.9
±
±
2.3
9
2
1.5
13
7
U
87Sr/86Sr
143Nd/144Nd
±
74 wt% 1Fig. 3). Andesitic pumices from the LPI II unit dacite samples, suggesting that fractionation was gov-
have SiO2 contents between 60 and 62 wt% 1Fig. 3). erned by di€erent bulk mineral-melt distribution coe-
Most dacitic pumices of LPI I and LPI II plot in a cients 1D), i.e., higher DSr in the dacites for a given DRb
.
narrow range between 65 and 67 wt% SiO2, which is This contrast is more prominent in the case of Ba and
similar to the typical large-volume APVC ignimbrites Zr, which show enrichment with increasing SiO2 in the
165±70 wt% SiO2). Rare quartz-free dacitic pumices rhyolites, but depletion in the dacites 1Fig. 3). Depletion
from the upper LPI II unit fall between the ®elds de®ned in the dacites suggest that the melts were saturated in
by the andesitic and dacitic pumices in Fig. 3. The most biotite and/or K-feldspar 1for Ba) and zircon 1for Zr). If
evolved pumices in the Purico units are from the LPI II the rhyolites were derived from continued fractionation
fall deposit and the lowermost part of the overlying ¯ow, of the dacite magma, it is dicult to understand why Ba
and these show a variation in SiO2 between 68 and and Zr depletion did not continue in the rhyolites.
74 wt%. Although not all of these pumices are rhyolites Furthermore, the contents of Th and U, both highly
in the strict sense 1>70 wt% SiO2) we will hereafter refer incompatible trace elements, are exceptionally low in the
to them as rhyolite or rhyolitic for convenience.
rhyolite compared with the dacite and andesite, al-
The variations in major elements with increasing SiO2 though the Th/U ratios remain constant 1ꢀ3) for the
1decreasing total Fe2O3, CaO, TiO2, MgO, MnO, and whole compositional range. The rhyolites cannot there-
increasing K2O) are consistent with progressive frac- fore be simply related to the dacites by fractional crys-
tionation of observed phenocryst phases, in particular tallization.
plagioclase, orthopyroxene, clinopyroxene, and/or
Figure 4 shows the REE abundance patterns in the
hornblende. However, the concept of a continuous pumices, and for comparison, a basaltic andesite 1ma®c
fractionation series from andesite to rhyolite is at odds inclusion) from one of the younger domes. The REE
with the trace element data 1Fig. 3), which show diver- patterns are moderately steep for all pumice types, with
gent compositional trends for the di€erent pumice types. La/Ybn between 8 and 13, and they steepen progres-
For example, the variation trends of Sr and Rb from sively with fractionation from basaltic andesite 1La/
andesite and rhyolite are steeper than those within the Ybn=4.0) to andesitic 1La/Ybn=8.3) and to the dacitic

685
187Sr/86Sr=0.706) and the highest ꢀNd values 1±3.2) of
all Purico±Chascon±Aspero samples, which has been
explained by a greater proportion of mantle-derived
components in the younger domes 1Hawkesworth et al.
1982; Davidson et al. 1990).
Fig. 3 SiO2 and Rb versus selected major and trace elements of
Purico whole rock samples 1left) and LPI glasses 1right). Small dots
represent the compositional range of APVC ignimbrites between 22
and 24°S1Schmitt 1999). Trends for host-mineral crystallization of
melt inclusions are indicated by arrows. Note that major element
variations indicate negligible post-entrapment crystallization
In summary, the whole-rock geochemical and isotope
data show that the Purico magmas cover the composi-
pumices 1La/Ybn=12.3). This steepening correlates with
an increase in the negative Eu anomaly 1decreasing Eu/
Eu* from 0.75 to 0.67). Again, the relatively ¯at pattern
of rhyolitic samples 1La/Ybn=9.3) and relatively low
Eu/Eu* ratio of ꢀ0.7 are inconsistent with the rhyolites
being a continuation of the andesite±dacite trend.
The initial Nd and Sr isotopic compositions 1calcu-
lated for 1.3 Ma) of representative pumice samples
including andesites, dacites, and rhyolites are indistin-
guishable within the range of error 1ꢀNd=±6.7 to ±7.2
and 87Sr/86Sr=0.7085 to 0.7090 resp.; see Table 2 and
Fig. 5). The overlap of isotopic ratios with the range of
basement compositions 1e.g., Lucassen et al. 1999) is the
most widely cited evidence for a crustal source of the
APVC magmas 1Hawkesworth et al. 1982; de Silva 1991)
and is discussed in greater detail in a later section.
Published 87Sr/86Sr ratios for the young Chascon±Asp-
ero domes are between 0.707 and 0.709, slightly lower
than the Purico ignimbrite values. Ma®c inclusions in
Fig. 4 C1 chondrite-normalized 1Anders and Grevesse 1989) REE
patterns for Purico pumices and lavas. Ma®c andesite is an olivine-
bearing inclusion in Chascon dacite lava and is shown for
the dome lavas have the least radiogenic Sr ratios comparison

686
Fig. 5 Initial Sr isotopic compositions versus Sr abundance in
Purico ignimbrites and lavas including Purico, Chascon and
Aspero data from Hawkesworth et al. 11982) and de Silva 11991).
For comparison, values for ignimbrites and post-caldera domes
ranging in age from 4.1±2.6 Ma of the nearby La Pacana system are
plotted 1data from Lindsay 1999). Mixing line 1solid) and
assimilation-fractional crystallization 1AFC) lines 1dashed) were
calculated for dacitic post-caldera dome 1Cerro Bola from La
Pacana) and basaltic±andesite inclusion in Chascon±Aspero lavas
as end-members. +10% steps in fraction of remaining melt. Ratio
of assimilation to crystallization rate in AFC model=0.8. Numbers
indicate mixing proportions in percent
quartz-hosted inclusions of LPI I had to be rehomoge-
nized by heating prior to analysis. Since quartz is lacking
in the rhyolites, only plagioclase-hosted inclusions could
be studied. Mineral compositions were determined to
provide constraints on intensive parameters of crystal-
lization and empirical mineral-melt distribution coe-
cients were obtained by trace element analysis of mineral
and matrix glass pairs from individual pumice clasts 1see
Appendix for analytical details).
tional range of typical large-volume ignimbrites from the
APVC province. The variation in chemical composition
with stratigraphic position in the Purico ignimbrites in-
dicates that the Purico magma chamber was composi-
tionally zoned at the time of LPI II eruption, but the
concept of a regular di€erentiation from the least
evolved andesite through dacite to the most evolved
rhyolite is not supported by the data. Despite its minor
volume, the rhyolite is clearly part of the Purico magma
system and the issue of its origin is essential to under-
stand how this system evolved.
Mineral compositions
Plagioclase is generally the most abundant phenocryst
phase in all pumice and lava types. Photomicrographs,
compositional ranges and pro®les of selected pheno-
crysts from the Purico dacite and rhyolite are presented
in Figs. 6, 7 and 8. The compositional range in the LPI I
dacite 1An38±An48) is more restricted that in the LPI II
dacite pumice 1An38±An68). On average, plagioclase
phenocrysts from rhyolitic pumice 1An40) are slightly
less calcic than those from dacitic pumice as might be
expected, but some cores of normally-zoned plagioclase
from the rhyolite have highly calcic compositions
1maximum An80), which are in the range displayed by
plagioclase phenocrysts in the andesitic pumice 1An70±
An80). The andesite plagioclase commonly displays sieve
textures, with cores intensely in®ltrated by blebs of
matrix glass. Clear rims and groundmass plagioclase
from the andesite are less anorthitic 1An60). The UPI
dacite pumices contain oscillatory zoned plagioclase
with high-An cores 1ꢀAn80) locally preserved, and with
compositions from phenocryst rims and matrix micro-
lites 1An42±An67) in the same range as the LPI dacites.
Plagioclase phenocrysts from Dome D and Chascon-
Aspero dacite lavas show no obvious textural or com-
positional di€erences from the dacite pumices. The ma®c
inclusions in the younger domes typically have high-
An plagioclase laths in the groundmass 1An74±An83).
Mineral chemistry and the composition of matrix
and melt inclusion glasses
Methods
The melt inclusions analyzed are suciently large
1>50 lm in diameter) to represent true liquid compo-
sitions 1Roedder 1984; Lu et al. 1995), and the e€ects of
local compositional gradients in the melt 1e.g., boundary
layer e€ects at the crystal surface) and post-entrapment
changes 1e.g., crystallization of the inclusion host) can be
neglected 1see below). Cracked inclusions and inclusions
with daughter crystals or large bubbles 1>5 vol% of the
inclusion volume) were avoided. Melt inclusions from
LPI II were generally pristine 1undevitri®ed) whereas

687
Oscillatory zoned euhedral plagioclase from the LPI I
and LPI II dacites revealed only minor variation in
anorthite content within the cores and the near-rim areas
in 1generally <10 mol% An). Sr and Ba contents in in-
dividual phenocrysts are uniform within the analytical
error. In contrast, plagioclase phenocrysts from the
rhyolite have more complex zonation patterns and a
wider range in An 135 mol%), Sr 1800±1,100 ppm) and
Ba 1210±340 ppm). Truncations of oscillatory growth
zones correlate with abrupt shifts in composition 1up to
17 mol% in An content, see Figs. 6 and 7) and these
indicate former dissolution surfaces. The rim composi-
tions of the overgrowth zones are typically ꢀAn40. S r
and Ba correlate negatively with An contents, and low-
Ca plagioclase in the rhyolite has the highest Sr abun-
dance 1Sr=1,100 ppm, Ba=340 ppm).
Amphibole and magnesian biotite 1mg#=65) are the
most abundant ma®c minerals in the dacite pumice, la-
vas, and crystal-rich inclusions 1Table 3). Amphibole
compositions range from magnesian hornblende to
magnesian edenite 1mg#=60±70). The rhyolitic pumice
contains rare magnesian pargasite 1mg#=67). Opacitic
rims consisting of ®ne-grained oxides and pyroxene
1analysis see Table 4) were found only in the UPI unit.
Minor orthopyroxene 1hypersthene with average
mg#=64) is present in dacite pumice from the upper
portion of LPI II, in UPI and in dome lavas 1Table 4), and
it occurs both as single phenocrysts and ®ne-grained clots
intergrown with Fe±Ti oxides. Commonly, orthopyrox-
ene is partially mantled by reaction rims of hornblende,
but orthopyroxene has also been observed to mantle
biotite in the quartz-free LPI II dacitic pumice. The
andesites contain abundant orthopyroxene 1mg#=68)
Fig. 6 Photomicrographs 1cross-polarized light, analyzator in) of
plagioclase phenocrysts in rhyolitic pumice from LPI II. Arrows
indicate former dissolution surfaces. Note glassy melt inclusions
aligned on euhedral growth zones. Positions of microprobe and
SIMS traverses from Fig. 8C, D are indicated
Fig. 7 Examples of zonation
pro®les of plagioclase pheno-
crysts from dacitic 1A and B)
and rhyolitic 1C and D) pumice.
Sr concentration in melt were
calculated from mineral-melt
partitioning according to Blun-
dy and Wood 11991) at tem-
peratures of 800 and 850 °C
for the dacite and rhyolite,
respectively

688
Purico glasses, these elements being strongly depleted in
the dacite pumice glasses 1Fig. 3). Matrix glasses from
the crystal-rich inclusions and Dome D lavas have
compositions similar to those of the dacite pumice glass,
which indicates that the variations observed in whole-
rock compositions in this case result from variations in
composition and modal abundance of phenocrysts. The
glass in dark parts of the banded pumices is composi-
tionally identical to the matrix glasses of the andesite,
and glass from the light parts resembles the dacite matrix
glass, indicating that the magma mingling in the banded
pumice involved little or no chemical interaction.
Petrographic details and compositions of quartz- and
plagioclase-hosted melt inclusions are summarized in
Table 6 and Table 2 in the electronic supplementary
material. In general, quartz is the more stable host for
melt inclusions, but plagioclase has the advantage that
the inclusions can be correlated with petrographic and
chemical features of the host 1zoning) and, in any case,
quartz phenocrysts are lacking in the rhyolite.
For the Purico rhyolite, we distinguished two types of
melt inclusions:
1. Irregularly distributed, generally small 1<30 lm)
inclusions are common in high-An cores. These inclu-
sions presumably originated from resorption of the host
mineral and subsequent annealing of melt pockets. Only
minor compositional shifts in the host plagioclase are
found around these inclusions 1Fig. 8).
2. Large inclusions 1up to 75 lm in length) aligned
parallel to crystallographic growth planes that were ap-
parently trapped during periods of rapid crystal growth.
Reversals in An content of the host are commonly as-
sociated with the inclusion-rich zones 1Figs. 7 and 8).
Variations in glass compositions in terms of selected
major and trace elements against SiO2 and Rb were
shown along with whole-rock data in Fig. 3. Major
compositional changes caused by post-entrapment
crystallization of the inclusions can be ruled out because
the observed variations do not correlate with calculated
trends for host-mineral crystallization. A comparison of
glass compositions from the distinct pumice types shows
the following important points:
1. Trace element abundances underscore the whole-
rock evidence for divergent trends between dacitic and
rhyolitic pumice. Matrix and melt inclusion glasses from
the rhyolite have relatively high compatible trace ele-
ment abundances 1e.g., Sr=150±290 ppm), whereas
glasses in the dacites have a lower and more narrow
range in Sr, averaging 70‹20 ppm.
Fig. 8 MgO and Sr in melt inclusions from LPI II rhyolitic pumice
versus An contents of the adjacent host plagioclase. Insert
compares average and range in An composition from microprobe
traverses of plagioclase phenocrysts of LPI II rhyolitic and dacitic
pumice 1ꢀ20 phenocrysts analyzed each)
and augitic clinopyroxene 1mg#=74) as the only ma®c
phases, and both are uniform in composition 1Table 4).
Magnetite and minor ilmenite are present in all
pumice types 1Table 5). The Fe±Ti oxide grains generally
lack exsolution lamellae, which is a prerequisite for ap-
plying two-oxide thermometry 1see below). Sanidine is
absent in the Purico units although it has been found as
a minor constituent in some compositionally similar
APVC dacitic ignimbrites 1e.g., Atana and Tara ig-
nimbrites; Lindsay et al. 2000) and domes 1e.g., Cerro
Chao; de Silva et al. 1994; Chason-Runtu Jarita; Watts
et al. 1999).
2. Ba and Zr in dacite and rhyolite glasses de®ne
distinct populations 1Fig. 3). Compared with whole-rock
compositions, Ba and Zr in glasses are enriched or only
moderately depleted in the rhyolite, whereas both are
Matrix and melt inclusion glasses
Using SiO2 as a fractionation index, the matrix glasses strongly depleted in the dacite.
from the andesitic pumice are the least fractionated
3. Melt inclusions from Ca-rich cores in rhyolite
1SiO2ꢀ70 wt%), followed by rhyolitic pumice 1SiO2= plagioclase have very low incompatible trace element
74±77 wt%) and dacite pumice 1SiO2=77±78 wt%). The concentrations 1e.g., Rb=72 ppm, Th=7 ppm). These
andesite matrix glass has the highest total Fe2O3 abundances are even lower than in the matrix glass of
1ꢀ3.2 wt%), CaO 1ꢀ3 wt%) and MgO 1ꢀ0.9 wt%) of all andesitic pumice 1Rb=144 ppm, Th=13 ppm).

689
Table 3 Selected microprobe analyses of amphibole and biotite. dp average An content of plagioclase rims at a pressure of 200 MPa
Dacitic pumice; dl dacitic lava; ci crystal-rich inclusion; rp rhyolitic 1‹40 °C error). Pressure estimated by Al-in-hornblende barome-
pumice; ap andesitic pumice; mi ma®c inclusion. Oxides in wt%; try, corrected for temperature derived from Fe±Ti-oxide ther-
FeO and Fe2O3 for amphibole based on minimum ferric Fe esti- mometry using equations in Anderson and Smith 11995). Error in
mate for 15 cations excluding 1Na+K)A after Schumacher 11997). Al-in-hornblende barometry calibration ‹50±60 MPa 1Anderson
Temperatures calculated according to the plagioclase±edenite± and Smith 1995), cumulative errors of Al determination and tem-
richterite exchange equilibrium of Holland and Blundy 11994) using perature correction ‹50 MPa
Sample
Unit
Type
ZAP9616 SIL963
dp
PED979
dl
dome D
Mg-ed
PED9728 PED972
ci
LPI II
Mg-hbl
PED9714 CHA974
dl
CHA973
mi
Chascon
Mg-parg
ZAP9616 PED976
rp
dp
LPI I
Mg-ed
rp
LPI II
Mg-parg
dt
dp
LPI I
bio
LPI I
Mg-hbl
UPI
Mg-ed
Chascon
Mg-parg
LPI II
bio
Mineral
SiO2
46.27
1.50
7.41
13.85
10.93
3.25
47.25
1.49
7.33
13.36
10.97
2.66
45.19
1.74
8.86
14.67
12.99
1.86
47.70
1.34
6.83
12.43
10.28
2.39
41.68
2.97
12.67
12.64
10.68
2.18
48.02
1.76
7.73
13.03
12.56
0.18
43.46
1.51
12.05
11.90
10.08
2.01
44.27
1.95
12.32
11.99
8.51
3.88
0.16
15.20
11.66
2.11
0.60
b.d.
b.d.
0.01
)
37.87
4.97
14.11
15.57
±
36.07
3.86
15.03
15.03
±
TiO2
Al2O3
FeO total
FeO
Fe2O3
MnO
MgO
CaO
Na2O
K2O
P2O5
F
Cl
±
±
0.41
0.49
0.42
0.39
0.19
0.31
0.23
0.13
14.56
b.d.
0.11
9.19
b.d.
0.03
0.15
)0.01
)0.03
0.22
14.60
0.21
0.56
8.62
b.d.
0.67
0.07
)0.28
)0.02
14.21
11.46
1.18
0.84
b.d.
0.36
0.10
)0.15
)0.02
14.43
11.95
1.17
0.78
0.05
0.03
0.11
)0.01
)0.02
12.98
11.58
1.29
1.09
0.01
0.09
0.13
)0.04
)0.03
14.48
12.56
1.06
0.64
0.01
0.20
0.07
)0.09
)0.02
13.31
11.47
2.08
0.62
0.05
0.24
0.03
)0.10
)0.01
14.79
11.55
1.31
0.71
0.03
0.18
0.08
)0.08
)0.02
14.82
11.13
2.41
0.72
0.03
0.89
0.03
)0.38
)0.01
±O=F
±O=Cl
)
Total
97.74
An45
98.68
An45
98.17
An45
97.84
An45
98.05
An40
99.12
An45
98.98
An40
100.66
An65
96.63
94.63
T 1°C)
P 1MPa)
795
160
800
150
800
270
730
120
835
±
800
±
830
±
890
±
Table 4 Selected microprobe analyses of pyroxene phenocrysts and Temperatures calculated for cpx and opx pairs using the QUILF
thermometry results. dp Dacitic pumice; dl dacitic lava; dt dacitic program of Andersen et al. 11993) with uncertainties for T of
tu€; rp rhyolitic pumice; ap andesitic pumice. Oxides in wt%; FeO ꢀ‹50 °C
and Fe2O3 calculated by stoichiometric criteria after Droop 11987).
Sample
Unit
Type
SIL963
LPI I
dp
PED979
dome D
dl
PED976
LPI II
rp
PED9721A PED9721A PED962A PED962A PED9714 PED9714 CHA973
LPI II
ap
opx
LPI II
ap
cpx
LPI II
ap
opx
LPI II
ap
cpx
UPI
dt
opx
UPI
dt
cpx
Chascon
dl
opx
Mineral
opx
opx
opx
SiO2
TiO2
Al2O3
FeO total
FeO
Fe2O3
MnO
MgO
CaO
55.28
0.23
2.62
11.31
11.22
0.10
52.17
0.32
1.47
21.07
20.91
0.18
51.90
0.24
2.62
16.95
14.90
2.29
54.03
0.16
0.73
19.84
19.84
±
0.78
23.79
0.87
0.03
51.20
0.40
2.25
9.65
9.65
±
0.29
13.38
20.70
0.38
53.45
0.27
1.03
17.42
17.42
±
0.51
24.99
1.21
0.02
51.24
0.46
1.73
9.28
8.54
0.82
0.30
14.97
20.45
0.30
53.21
0.14
0.82
20.98
20.72
0.29
52.69
0.73
4.43
12.48
12.48
51.84
0.36
3.06
5.53
5.65
±
0.00
15.87
22.37
0.22
±
0.22
0.70
0.36
0.70
0.21
18.28
11.09
0.55
29.67
1.25
0.03
22.19
1.14
0.02
24.48
2.52
0.07
23.17
0.80
0.02
Na2O
Total
T [°C]
100.62
99.10
99.37
100.24
880
98.24
98.89
965
98.82
99.83
825
100.46
99.25
4. Melt inclusion compositions in the rhyolite become Distribution coecients
more evolved 1e.g., decreasing MgO and Sr contents)
with decreasing An content of the adjacent host pla- Empirical mineral-melt partition coecients for plagio-
gioclase, indicating that the melt composition changed clase, amphibole, and biotite in dacitic pumice were
during crystallization and trapping 1Fig. 8). In contrast, derived from trace element compositions in phenocrysts
quartz- and plagioclase-hosted inclusions from the and matrix glasses determined by SIMS and LA-ICP-
dacite are homogeneous and similar to matrix glass MSanalysis. The calculated D values are given in Table 7
composition.
and the data are available in the electronic supplemen-

690
Table 5 Selected microprobe analyses and thermometry results of tures and oxygen fugacities calculated using the QUILF program of
magnetite±ilmenite pairs. dp Dacitic pumice; rp rhyolitic pumice; ap Andersen et al. 11993) with quoted uncertainties for T of ꢀ‹25 °C
andesitic pumice; dl dacitic lava. Oxides in wt%; FeO and Fe2O3 and 0.3 log units for fO2 given an uncertainty of 1 mol% for the
calculated by stoichiometric criteria after Droop 11987). Tempera- ilmenite and magnetite composition by electron microprobe analysis
Sample
Unit
Type
PED971
LPI I
dp
PED962H
LPI II
dp
PED972
LPI II
rp
PED962B
LPI II
ap
PED979
dome D
dl
CHA974
Chascon
dl
Mineral
Ti-Mag Ilm
Ti-Mag Ilm
Ti-Mag Ilm
Ti-Mag Ilm
Ti-Mag Ilm
Ti-Mag Ilm
SiO2
TiO2
0.09
6.29
2.00
0.11
83.00
0.62
0.91
b.d.
0.05
1.72
7.06
0.10
82.42
0.91
1.30
0.03
0.04
5.57
2.68
0.03
82.84
0.67
1.40
0.03
0.07
10.56
2.82
0.04
0.09
6.41
2.10
0.13
83.80
0.64
1.11
b.d.
42.11
0.29
0.04
53.06
0.86
1.67
0.05
4.39
1.94
0.16
86.06
0.38
0.73
0.79
42.39
0.49
0.02
51.29
1.05
1.74
42.30
0.05
0.07
51.96
1.03
2.03
30.32
0.63
0.02
63.01
0.37
1.24
39.96
0.19
0.07
52.37
0.30
3.33
37.22
0.78
0.05
56.74
0.59
1.56
Al2O3
Cr2O3
FeO total
MnO
MgO
0.23
78.11
0.33
2.50
FeO
Fe2O3
34.68
53.78
33.87
19.38
33.68
54.25
33.35
20.72
33.45
54.97
24.68
42.66
36.78
46.01
29.55
25.40
34.91
54.42
33.95
21.27
33.68
58.30
30.90
28.76
Total
T 1°C)
log fO2
98.48
780
)12.7
98.94
98.48
773
)12.7
99.58
98.85
850
)10.6
99.96
99.28
885
)10.8
98.93
99.81
785
)12.4
100.23
99.66
770
)12.1
100.65
tary material 1Table 3). Comparison with literature data in good agreement with Fe±Ti-oxide temperatures from
for similar compositions 1e.g., dacite from Ewart and both groups of samples. The plagioclase±amphibole
Grin 1994) shows generally good agreement with val- temperatures from Chascon lavas and inclusions 1830±
ues determined in our study.
890 °C) are slightly higher than the corresponding Fe±Ti
oxide temperatures.
Temperatures
calculated
from
two-pyroxene
Pre-eruptive magma conditions
thermometry for the andesites 1965 °C) exceed the
corresponding Fe±Ti-oxide temperatures. Similar dis-
crepancies between Fe±Ti-oxide and two-pyroxene
geothermometry were reported for andesite lavas from
Mineral thermometry and barometry
Magmatic temperatures were derived using the Fe±Ti- Lascar volcano 1Matthews et al. 1994) and this problem
oxide thermometer 1Andersen and Lindsley 1988; may re¯ect a higher closure temperature for the pyrox-
Andersen et al. 1993), the plagioclase±amphibole enes compared with magnetite and ilmenite.
thermometer 1Holland and Blundy 1994) and two-py-
A temperature increase of ꢀ50 °C from dacite to
roxene geothermometry 1Lindsley 1983; Andersen et al. rhyolite is documented by various independent geo-
1993).
thermometers and this constitutes further evidence that
Fe±Ti oxide thermometry was based on at least ®ve the rhyolite cannot have originated by fractionation of
magnetite and ilmenite pairs from each sample. All the dacite. Given the precision of ‹25 °C for Fe±Ti-
samples met the criterion for equilibrium based on the oxide temperatures 1Andersen et al. 1993) and ‹50 °C
systematic Mn and Mg partitioning in the oxide phases for pyroxene thermometry 1Lindsley 1983), the temper-
1Bacon and Hirschmann 1988) and no di€erence was ature variations indicate that the incoming andesite was
found between touching mineral pairs and isolated at least ꢀ100±200 °C hotter than the resident dacite.
phenocrysts. Fe±Ti-oxide temperatures for the quartz- Heating of the dacite at the magma interface might ex-
bearing dacite pumices, crystal-rich inclusions, a tu€ plain the presence of opacitic rims in the late- erupted
sample from UPI, two Dome D lavas and one Chascon UPI dacite 1two-pyroxene temperature: 825 °C). Heating
lava range from 750±800 °C. Temperatures obtained of dacitic magma by still later recharge events could
from andesitic pumice are higher, at 860±880 °C. Most account for the high temperatures recorded in Chascon±
oxide pairs in the rhyolite record similarly high tem- Aspero dacite lavas.
peratures ꢀ850 °C and the total range is 800±870 °C.
Oxygen fugacities calculated from the data are generally eter 1Anderson and Smith 1995) was applied to the LPI
1.5 log units above the Ni±NiO-bu€er.
dacites. Application of the barometer in this case is
The temperature-corrected Al-in-hornblende barom-
Plagioclase and amphibole rim compositions from justi®ed despite the lack of sanidine because the melt
the dacite pumice yielded edenite±richterite exchange phase has SiO2>76 wt% and K2O>5 wt%, thus ful-
temperatures between 750 and 810 °C, calculated for ®lling the criteria of Johnson and Rutherford 11989).
pressures between 100 and 300 MPa based on Al- Errors in the calibration, Al-determination, and tem-
in-hornblende barometry 1see below). Coexisting pla- perature correction of this barometer are considerable
gioclase and amphibole in rhyolitic pumice yield higher 1at least ‹60 MPa). but the pressure estimates between
temperatures between 840 and 880 °C. These results are 110 and 220 MPa suggest that the Purico dacites equil-

691

692
Table 7 Mineral/melt partition coecients for Purico dacite composition by SIMS 1see Appendix for analytical procedures).
1PED962H). n Number of replicate analyses. Trace element E&G Range of partition coecients for low-Si rhyolites and dacites
abundance in minerals determined by LA-ICP-MS, glass from Ewart and Grin 11994) for comparison
Mineral
Plagioclase
Hornblende
Biotite
n 1Mineral)
n 1Glass)
5
7
4
1
4
4
Min.
Av.
Max.
E&G
Av.
0.09
Min.
Max.
E&G
E&G
Rb
Sr
Y
0.04
10.38
0.38
0.05
0.06
0.71
1.04
0.80
1.81
0.07
0.03
0.02
9.09
0.03
0.00
0.01
0.54
0.69
0.46
1.61
0.01
0.02
0.09
11.39
1.11
0.22
0.23
0.85
1.49
1.40
2.14
0.20
0.06
0.023±0.24
5.83±10.7
0.083±0.55
0.10±0.55
0.26±0.88
0.28±1.05
±
±
±
±
±
0.04
0.71
14.45
0.72
1.47
0.23
1.73
2.81
5.31
0.08
0.04
0.12
1.34
18.56
1.67
1.96
0.67
4.26
5.47
6.14
0.26
0.12
0.021±0.37
0.25±1.13
4.0±45.2
0.53±0.93
1.5±6.7
0.16±0.92
±
±
±
±
±
2.07
0.78
0.36
0.54
2.26
16.95
0.41
0.49
0.52
0.25
0.28
2.46±5.27
0.25±0.31
0.6±2.3
0.19±0.59
4.6±>7.5
6.4±13.9
±
±
±
±
±
0.94
16.69
1.08
1.72
0.50
2.70
4.03
5.64
0.15
0.08
Zr
Nb
Ba
La
Ce
Eu
Th
U
ibrated in the upper crust 14±8 km, assuming a density of Johannes 11994). Pressure estimates for the dacite from
the lithostatic column of 2.7 g/cm3).
Al-in-hornblende geobarometry are not signi®cantly
higher than the saturation pressure, suggesting that the
melts may have been at saturation before eruption. The
e€ect of CO2 on the calculated saturation pressure,
considering the very low CO2 concentrations, is in the
range of the H2O variability in the melt inclusions.
Water contents
Pre-eruptive water contents in Purico melts have been
constrained by SIMS analysis of melt inclusion glasses
and by the plagioclase-melt geohygrometer of Housh
and Luhr 11991). Detailed interpretation of the volatile Summary of chemical and petrologic constraints
evolution of the Purico system and other APVC felsic
magmas is the subject of a separate paper and the results Before discussing the broader issues of pre-eruptive
are only brie¯y summarized here 1see also Schmitt and zonation in the Purico magma system it is useful to
de Silva 2000).
H2O concentration ranged from 3.5±4.2 wt% in whole-rock and phase chemistry described above:
quartz-hosted melt inclusions from dacite pumice, and 1. Compositionally discrete magma batches were
reiterate our interpretation of the main results from
from 3.2±4.8 wt% in plagioclase-hosted rhyolitic pum- erupted from the Purico system. The isotopic similarity
ice. FTIR measurements showed that CO2 contents were and constant incompatible element ratios 1e.g., Th/U)
low 1<400 ppm). The calculated water contents from over a wide compositional range of more than 10 wt% in
plagioclase-melt equilibrium, adjusted for temperatures SiO2 suggest a common origin for all Purico magmas
determined from Fe±Ti-oxide thermometry, broadly types, but producing the large volumes of dacite by
agreed with the melt inclusion data. However, for the mixing between the rhyolite and the andesite can be
andesite and dacite a 1 wt% di€erence in H2O was ruled out by volume considerations and observed trace
found between the Ab and An calibrations of the geo- element variations.
hygrometer, which exceeds the experimental calibration
2. The dacite magma cannot be the precursor to the
error of 0.4 wt% 1Housh and Luhr 1991). No melt in- rhyolites. The compositions of the dacites are distinct
clusions suitable for water determination were found in from the bulk rhyolite and de®ne di€erent evolutionary
the andesitic pumice. Given the better agreement be- trends. Incompatible trace elements such as Th and U
tween melt inclusion data and the Ab geohygrometer are lower in the rhyolite bulk pumice and matrix glass
calibration in the dacite samples where both methods than in the dacite. Depletion caused by partitioning into
can be compared, the Ab calibration appears preferable an exsolved ¯uid is unlikely given low ¯uid-melt parti-
and this yields an estimated melt water content of tion coecients for Th particularly 1Keppler and Wyllie
ꢀ4 wt% for the andesite, similar to that of the dacite 1990). Furthermore, trends of compatible trace element
and rhyolite magmas.
variation in the andesites and rhyolites diverge from the
These values of melt volatile contents can be used to dacites, and the rhyolites have higher equilibration
calculate saturation pressures if the solubilities of volatile temperatures. Finally, the Sr contents in the An-rich
species in melts are known. For H2O, the melt inclusion cores of plagioclase in the rhyolite require crystallization
data indicate minimum pressures between 100 and from a high-Sr melt 1ꢀ380‹100 ppm, Fig. 7) according
150 MPa for both the dacite and the rhyolite, using H2O to the partitioning model of Blundy and Wood 11991).
solubility data in haplogranitic melts from Holtz and These plagioclase grains are not xenocrystic because the

693
Sr contents of matrix glass 1ꢀ150 ppm) agree with the 6. The homogeneity of quartz- and plagioclase-hosted
melt composition calculated from Sr concentration in inclusions in the dacite and their similarity to matrix
plagioclase rims using the same method. glasses suggest that the inclusions were either trapped at a
3. Melt inclusions in zoned plagioclase from rhyolite late stage of magma evolution or have re-equilibrated
reveal that the trapped melt continuously changed with the surrounding melt by di€usion. The latter would
composition during crystallization of the host and the imply a long residence time of the dacitic magma before
trace element variations in the melt inclusions require a eruption and a slowly cooling chamber.
parental magma less evolved than the dacite.
4. Dissolution surfaces observed in zoned plagioclase
phenocrysts in rhyolite pumice are attributed to tem-
perature variation. Principally, changes in plagioclase Discussion
An content are controlled by both temperature and
water content of the melt 1Housh and Luhr 1991), but in Evolution of the Purico magmatic system and origin
the case of volatile-saturated magmas at constant pres- of compositional variations
sure, as proposed for the pre-eruptive Purico chamber,
water content is bu€ered by H2O solubility in the melt, The Purico ignimbrite consists of compositionally dis-
and thus we believe that the dissolution surfaces indicate tinct juvenile materials that have indistinguishable Nd
re-heating and not variations in H2O 1e.g., Singer et al. and Sr isotopic composition. A convective fractionation
1995). It should be noted that only a fraction of the mechanism 1Wol€ et al. 1990) has been invoked for
phenocrysts in the rhyolite is a€ected. Thus, the tem- zonation in Purico and other central Andean magma
perature variation is likely to be a small-scale phenom- chambers by de Silva 11991). However, a combination of
enon connected with strong temperature gradients in evidence from melt inclusion data and geothermometry
convecting rhyolite, as opposed to external heating by indicates that the rhyolitic magma cannot be viewed
new input of hot ma®c melt 1Singer et al. 1995).
simply as the evolved cap of a dacitic magma chamber as
5. Further evidence for the prevalent thermal varia- previously suggested, and a precursor more ma®c than
tions in the rhyolite magma comes from melt inclusions dacite is indicated.
with high Sr contents 1350 ppm), trapped along a former
Fractional crystallization as a mechanism of gener-
dissolution surface in a zoned plagioclase phenocryst ating the observed compositional variations was tested
1Fig. 6). The zone of entrapment corresponds to an by least-squares mixing models 1LSQ) using reasonable
abrupt shift in An content from An33 to An50. Neither pairs of parent and daughter melts 1Tables 8, 9, and 10).
the H2O contents nor other chemical indices of frac- Trace element variations were then modeled by the
tionation in the inclusions 1e.g., MgO, see Fig. 8) are Rayleigh fractionation equation using D-values mea-
di€erent in the high-Sr inclusions compared with other sured in the Purico dacite 1Table 7) or taken from the
melt inclusions in the plagioclase. Thus, plagioclase dis- literature for andesitic compositions 1Bacon and Druitt
solution most likely resulted from a temperature increase 1988; Blundy and Wood 1991; Dunn and Sen 1994).
alone and not from compositional changes in the magma. Model 1 examines the step from andesite to dacite and
We envisage this process as taking place in a convecting the results indicate that the Purico andesite could pro-
rhyolitic magma with a strong temperature gradient. The duce dacite as a residual melt by removing plagioclase,
high Sr contents in the melt inclusions can be explained orthopyroxene, clinopyroxene, and Fe±Ti-oxide 1frac-
by post-entrapment re-equilibration with host plagio- tion of remaining melt f=0.71). Similar f-values, be-
clase at a higher temperature 1ꢀ60 °C), causing dissolu- tween 0.65 and 0.85, were calculated for selected trace
tion of low-Ca 1high-Sr) and precipitation of high-Ca elements 1Rb, Sr, Ba, and La). Small amounts of zircon
1low-Sr) plagioclase at the inclusion wall.
fractionation 1ꢀ1%) are required to explain the Y data
Table 8 Results of major element mixing models 1LSQ) and coecients for andesite from Bacon and Druitt 11988), Dunn and
trace element Rayleigh fractionation modeling. f Fraction of melt Sen 11994) and Blundy and Wood 11991)
remaining. SR2 Sum of squared residuals. Mineral-melt partition
Model 1
Parent
Daughter
Dacite
% Fractionated mineral phase
Andesite
Pl An70
61.9
Cpx
10.6
Opx
19.9
Mag
7.5
F
0.71
SR2
0.06
co
cl
Mineral/melt partition coecient kD
Bulk-D
Rb
Sr
Ba
La
Y
103
352
423
26
156
312
474
30
0.16
2.2
0.2
0.18
0.066
0.03
0.5
0.1
0.28
2.4
0.01
0.01
0.1
0.26
0.19
0.01
0.01
0.01
0.01
0.2
0.10
1.43
0.16
0.19
2.35a
0.63
0.76
0.52
0.84
0.79
22
16
a Assumes 1% zircon in fractionating assemblage 1k zfor Y in zircon ꢂ200)
D

694
Table 9 Results of major element mixing models 1LSQ) and trace coecients for dacite from Table 7 and from Bacon and Druitt
element Rayleigh fractionation modeling. f Fraction of melt re- 11988), and Dunn and Sen 11994)
maining. SR2 Sum of squared residuals. Mineral-melt partition
Model 2
Parent
Daughter
% Fractionated mineral phase
Quartz-poor Dacite lava Pl An50
Dacite
50.5
Hbl
39.2
Bt
Mag
5.2
F
SR2
5.2
0.80
0.24
co
cl
Mineral/melt partition coecient kD
Bulk-D
Rb
Sr
Ba
Y
126
313
465
18
176
285
456
18
0.04
10.4
0.71
0.38
0.09
0.94
0.5
2.07
0.78
17.0
0.36
0.01
0.01
0.01
0.2
0.16
5.65
1.43
6.76
0.67
0.98
0.96
1.00
16.7
Table 10 Results of major element mixing models 1LSQ) and trace coecients for andesite from Bacon and Druitt 11988), Dunn and
element Rayleigh fractionation modeling. f Fraction of melt re- Sen 11994) and Blundy and Wood 11991)
maining. SR2 Sum of squared residuals. Mineral-melt partition
Model 3
Parent
Daughter
Rhyolite
% Fractionated mineral phase
Andesite
Pl An80
51.8
Cpx
13.3
Opx
25.6
Mag
9.2
F
0.69
SR2
0.15
co
cl
Mineral/melt partition coecient kD
Bulk-D
Rb
Sr
Ba
La
Y
103
352
423
26
117
362
755
33
0.16
2.2
0.14
0.18
0.066
0.03
0.5
0.1
0.28
2.4
0.01
0.01
0.1
0.26
0.19
0.01
0.01
0.01
0.01
0.2
0.09
0.95
0.11
0.20
0.42
0.87
0.56
0.52
0.74
0.79
22
25
and are consistent with the increasing steepness of REE
Therefore, fractional crystallization models broadly
patterns from andesite to dacite. The variation between satisfy the observed chemical variations in the dacite
the least-evolved dacite pumice and evolved dacite lava and the transition from andesite to dacite, and andesite
from Dome D is treated in model 2. The model indicates to rhyolite suites. The inferred changes in the
that less fractionation is required 1f=0.67±0.98) and the fractionating assemblage are mainly temperature
calculated crystallizing assemblage resembles that of the controlled. For example, fractionation within the
crystal-rich aggregates in the ignimbrites, suggesting that cooler dacites involved biotite, zircon, and less calcic
they may be related to this process. Model 3 examines plagioclase, which resulted in depletion of the remain-
the possible derivation of rhyolite from a parental ing melt in Zr and Ba. The absence of these phases
andesite, and illustrates that clinopyroxene and ortho- during higher temperature fractionation of andesitic
pyroxene-dominated fractionation 1f=0.52±0.85) could magma would cause the enrichment of Zr and Ba in
yield the Purico rhyolite. In support of this model, we the residual melt, which is required to explain the
note that the bulk rhyolite composition resembles that of rhyolite composition.
matrix glass from andesitic pumice 1major elements and
The chemical characteristics and geothermometry
especially Zr in Fig. 3) and that the crystallinity of the data from the rhyolite suggest that its precursor magma
andesite 137±56%) is similar to the modeled degree of was chemically more primitive and hotter than the
fractionation. However, the andesite matrix glasses dif- Purico dacite. A mechanism capable of producing re-
fer from bulk rhyolite in their relatively high Rb and Th, sidual rhyolitic melt without prolonged fractionation
and low Ba contents. These di€erences could result from and cooling is phenocryst removal from a crystal-rich
disequilibrium between melt and phenocrysts in the magma by ®lter-pressing 1Sisson and Bacon 1999).
andesite as indicated by, e.g., sieve plagioclase, and/or Rapid, buoyancy-driven ascent of the hot rhyolite
from mixing with the dacite magma. In other words, the magma will minimize interaction with ambient dacite in
composition of interstitial melt in andesite clasts that the chamber, and decompression with concomitant
was quenched during eruption may di€er signi®cantly volatile oversaturation could lead to an overpressur-
from the residual melt composition produced by frac- ization of the magma chamber sucient to trigger
tionation of andesite at greater depths in the chamber. eruption 1e.g., Folch and Marti 1998). Figure 9 illus-
The latter may be better represented by the composition trates schematically our concept of how recharge of
of melt inclusions preserved in high-An plagioclase cores andesitic magma into a homogeneous dacite chamber
from rhyolitic pumice 1Fig. 8).
produced the compositional variations of the Purico

695
Fig. 9 Schematic evolution of the Purico magma chamber 1not to
scale). Step 11) eruption of homogeneous dacitic magma forming
LPI I, perhaps triggered by andesite recharge. Step 12) Cooling and
crystallization of injected andesite releases hot, volatile rich
rhyolitic melt. Volatile exsolution during decompression leads to
explosive eruption and a high temperature/high volatile magma
sustains a plinian column. Step 13). Following the plinian eruption,
the dacite and underlying andesite are subsequently tapped to form
LPI II. Slowly degassed dacite magma remaining in the chamber
forms late dacite-e€usive domes. Magma chamber depth estimated
from Al-in-hornblende barometry and minimum H2O saturation
pressures of hydrous melt 1see text)
not unique to the Purico system; the LPI II eruptive
sequence presents a variant of the concept of silicic re-
plenishment of intermediate magma chambers discussed
by Eichelberger and Izbekov 12000) and Eichelberger
et al. 12000).
Fluid dynamical experiments have shown that a low-
density silicic melt injected into a resident magma tends
to remain separate if viscosity contrasts are high
1Campbell and Turner 1986; Huppert et al. 1986). Hot,
volatile-rich and therefore less dense and less viscous
rhyolitic melt released at the base of a denser, more
viscous dacitic magma could rise through the latter and
remain una€ected, especially if the dacite layer is thin
1Huppert et al. 1984). This is all the more likely for the
Purico system since the dacite magma was already par-
tially erupted during the LPI I event. Table 11 shows
calculated densities 1Bottinga et al. 1982; Lange 1994;
Ahrens 1995) and viscosities 1Shaw 1972; Pinkerton and
Stevenson 1992) of all Purico magma types. Like other
studies on the viscosity of silicic magmas 1Scaillet et al.
1998), we do not consider the e€ects of a separate vol-
atile phase. The calculated melt viscosities di€er by more
than one order of magnitude, the di€erences mainly
re¯ecting the temperature contrast 1at least ꢀ100 °C
between andesite and dacite) rather than the di€erent
melt H2O contents 1range of ꢀ1 wt% between dacite
and rhyolite using the maximum values from melt in-
clusions). The e€ect of crystallinity on magma viscosity
is also important, and Table 11 shows that the di€erence
in magma viscosities between the crystal-rich dacite and
crystal-poor rhyolite are signi®cantly greater than the
contrast in melt viscosity alone.
system. Following eruption of the LPI I ignimbrite,
andesitic recharge led to production of a hot, volatile-
rich and low-viscosity rhyolitic magma, which erupted
as a sustained plinian column. Subsequent eruption of
the LPI II ignimbrite progressed when entrainment of
denser, volatile-poorer and more viscous magma from
the main dacitic portion of the magma chamber caused
the plinian eruption column to collapse. Minor
volumes of reheated dacitic and andesitic magma were
tapped during the waning stages of the eruption.
Subsequently, degassed dacite magma was passively
squeezed out by ongoing andesite ascent to form the
younger dacite domes. Evidence that andesite recharge
was important for the dacite dome e€usion is the
presence of ma®c inclusions in the younger dome lavas
1Hawkesworth et al. 1982; Davidson et al. 1990; Tepley
and Davidson 1997).
Physical aspects and implications for styles
of zonation in ignimbrites
Experimental models of magma chamber evacuation
allow an estimation of the eruption draw-up heights,
which control whether or not a lower magma layer is
tapped during the eruption process 1Blake and Ivey
1986a, 1986b). Draw-up heights depend mainly on dis-
An important question remaining is how a volumetri-
cally minor rhyolitic melt can e€ectively separate from
its parental magma and rise through much larger vol-
umes of dacitic magma without mixing? This problem is
Table 11 Calculated pre-eruptive magma densities 1crystals plus Lange 11994) and Ahrens 11995) and viscosities g after Shaw 11972)
interstitial melt) and viscosities of melts and magmas of the Purico and Pinkerton and Stevenson 11992)
complex. Density values q calculated after Bottinga et al. 11982),
Type
Crystallinity
1vol%)
H2O melt
1wt%)
T
1°C)
q magma
1kg/m3)
Logg melt
1Pa s)
Logg magma
1Pa s)
Andesite
Dacite
Rhyolite
40
40
20
4
4
5
900
800
850
2750
2520
2440
5.1
6.7
5.5
5.9
7.6
5.8

696
charge rate and the viscosity of the upper layer. The Sr abundances, indicating a dominance of fractional
discharge rate of the Purico ignimbrite is not well con- crystallization versus mixing. This contrasts with the
strained, but two independent estimates can be made for larger La Pacana complex, which displays a signi®cant
eruption of the basal fall deposit 1following methods variation in Sr isotope composition, in part correlated
outlined in Carey and Sparks 1986; Wilson and Walker with the degree of di€erentiation 1Fig. 5). Principally,
1987; Carey and Sigurdsson 1989). These estimates this isotopic variation in the La Pacana rocks could
range from 3 ´ 107 to 5 ´ 107 kg/s or 6 ´ 104 to result from wall-rock assimilation and simultaneous
1 ´ 105 m3/s, assuming a density of 500 kg/m3, typical crystallization of the andesite±dacite magma 1AFC) or it
for most plinian eruptions 1Carey and Sigurdsson 1989). could result from mixing between dacite and an iso-
Given these values and the viscosity of the dacite mag- topically more primitive ma®c magma. For La Pacana,
ma, the critical thickness of dacite which would be we favor the mixing hypothesis because the chemically
needed to prevent the underlying andesite magma from most evolved rhyolites 1Toconao ignimbrite) are
being tapped is between 430 and 490 m. There is no isotopically similar to the associated dacites 1Atana
evidence that the LPI I eruption tapped andesitic magma ignimbrite) and can be related to them by fractional
but the andesitic clasts in LPI II ignimbrite indicate that crystallization 1Lindsay et al. 2000). An AFC process
the second eruption clearly did. This implies that the would produce more assimilation as fractionation pro-
thickness of the dacite layer had decreased to a few gresses. In a study of the nearby Lascar andesite±dacite
hundred meters following the LPI I event, and therefore stratovolcano, Matthews et al. 11994) suggested that
it seems likely that the rhyolitic melt could ascend only the relatively hot andesitic magmas 1890±970 °C)
through the dacite rapidly and with only minor cooling. evolved by an AFC process whereas a ¯attening of the
De Silva 11991) noted that compositional variation in 87Sr/86Sr versus Sr trend was observed for compositions
APVC ignimbrites is greatest in the smaller deposits, at ꢀ60 wt% SiO2. A dominance of fractionation over
whereas the large volume deposits are homogeneous. assimilation is expected to be even greater for the cooler
Based on these arguments of draw-up height, we can La Pacana dacites 1ꢀ760 °C; Lindsay 1999).
speculate that compositional zonation by ma®c recharge
The La Pacana, Purico, and Chascon±Aspero dome
may be a common process in the APVC magma systems magmas clearly have crustal isotopic signatures, but the
generally, but it could go undetected in the larger isotopic variations among them suggest that the crustal
chambers if the viscosity and thickness of the dacite contribution decreased with time between 4.1 1La Pa-
layer are sucient to prevent tapping of lower layers.
cana) and <0.5 Ma 1Chascon±Aspero). Source hetero-
geneities are unlikely to explain this variation because all
centers are in close proximity, and particularly so for the
Purico and Chascon-Aspero domes 1Fig. 2). Mixing
calculations indicate that the Purico compositions could
be produced by adding ꢀ30% of basaltic andesite, as
represented by the ma®c inclusions in Chascon±Aspero
lavas, to the most ``crustal'' high-87Sr/86Sr magma from
the La Pacana complex 1Cerro Bola dacite dome;
Lindsay et al. 2000). Using the same end-members, the
proportion of basaltic andesite in the Chascon dacites
would be ꢀ60% whereas even the most primitive Sr
isotope ratio in the La Pacana dacites requires <20%
basaltic andesite. These values are minimum estimates
for the proportion of ma®c components in the di€erent
magmas because the ``crustal'' end-member, Cerro Bola
dacite, may itself be a hybrid magma and not a pure
crustal melt.
A rough upper limit of the depth of crustal melting
beneath the APVC region can be derived from the REE
characteristics of the Purico and other APVC ignimb-
rites. These show no evidence for the heavy REE
depletion that would result from equilibration of melts
with residual garnet 1i.e., La/Ybn>20). Experimental
melting studies show that garnet is produced by incon-
gruent melting of biotite in quartzo-feldspathic compo-
Source constraints and implications for the evolution
of APVC ignimbrite volcanism
There is consensus that a large crustal contribution is
present in the central Andean ignimbrite magmas, but it
is unclear if the large-volume ignimbrites represent pure
crustal melts 1Hawkesworth et al. 1982) or products of
mixing between mantle-derived ma®c magmas and
crustal melts 1Francis et al. 1989; Ort et al. 1996). There
has been very little discussion in previous studies about
the depth of crustal melting or of the postulated mixing
and homogenization zone. The comprehensive suite of
chemical and isotopic data from the Purico complex
allows us to place some constraints on the deeper roots
of this magmatic system and its relationship with other,
compositionally similar APVC ignimbrites. Purico is
spatially associated with the La Pacana caldera complex,
the largest in the APVC, and we focus on a comparison
with that complex. It can be expected that the compo-
sition of the deeper sources for both will be similar
despite the di€erences in shallow-level magma chamber
evolution.
The isotopic variations within the La Pacana and
Purico complexes, if considered together, show the
e€ects of both fractional crystallization and an overall
mixing or assimilation trend between a low-87Sr/86Sr and
a high-87Sr/86Sr component 1Fig. 5). The Purico suite is
Ä
sitions at pressures >1.25 GPa 1Patino-Douce and
Beard 1995), and the lack of a garnet signature in the
REE data thus suggests that melting took place at
depths of <ꢀ30 km. A further constraint on the upper
generally homogeneous in 87Sr/86Sr over a large range in limit of magma generation comes from considering the

697
data trend in the Sr isotopic diagram 1Fig. 5). The tra-
jectory de®ned by the Purico and La Pacana data in this
diagram is close to a simple mixing trend, although an
AFC process would seem far more likely because inter-
action between ma®c magmas and cooler crustal melts
must involve simultaneous crystallization of the former.
If AFC is envoked, the data trend in Fig. 5 requires that
the bulk distribution coecient for Sr during fraction-
ation of the ma®c magma, DSr, is close to 1. This value
varies with the relative proportion of clinopyroxene and
plagioclase in the fractionating assemblage, which is
pressure sensitive. To demonstrate this, we calculated
the proportions of plagioclase and clinopyroxene pro-
duced by crystallization of a typical arc basaltic andesite
1Leon Muerto in Trumbull et al. 1999) at di€erent
pressures, using the MELTSsoftware of Ghiorso and
Sack 11995) and assuming 2 wt% H2O at the Ni±NiO
bu€er. At pressures <500 MPa 1ꢀ15 km) plagioclase
fractionation is important 150% of solid phases), Sr
behaves as a compatible element 1DSr=1.35 using par-
tition coecients in Tables 8, 9, and 10), and the AFC
model predicts a strong depletion in Sr with increasing
crustal assimilation. For higher pressures, >700 MPa
1ꢀ20 km), clinopyroxene dominates during the ®rst 50%
fractionation 1forming ꢀ75% of solid phases) and this
causes a bulk distribution coecient near 1 1DSr=0.93),
similar to a mixing trend. The latter is consistent with
the isotopic trend for the most primitive compositions
1lowest 1/Sr in Fig. 5), suggesting that magma interac-
tion took place at depths >ꢀ20 km, but we note that
this trend is also a€ected by secondary fractionation and
magma mixing.
Thus the isotopic and trace element characteristics of
the ignimbrites suggest that crustal melting and AFC
interaction with ma®c arc magma occurred in mid-
crustal levels between 35 and 15 km. Adopting the model
of de Silva 11989a) and Laube and Springer 11998), we
envisage a mid-crustal ponding of arc-derived basalt±
andesite magmas as the dominant mechanism for
generating the ignimbrite magmas. Geophysical obser-
vations 1low s-wave velocities and high electrical
conductivity; Schmitz et al. 1997; Chmielowski et al.
1999) provide good evidence that a partially molten zone
exists today beneath the Altiplano-Puna plateau at
ꢀ20 km depth. This may be the present manifestation of
the mid-crustal source for the APVC ignimbrites.
Crustal melting and hybridization with andesites must
have been active since at least 10 Ma, the onset of ig-
nimbrite activity in the APVC. The peak activity of
large-volume ignimbrite volcanism in the APVC ceased
with the ꢀ4 Ma La Pacana and Guacha caldera-forming
events, and younger activity is more sporadic and much
smaller in volume. As described above, the proportion
of mantle-derived magma in the La Pacana, Purico, and
Aspero±Chascon systems appears to increase with time,
and therefore the Purico system may represent the
waning stages of a major crustal melting event and a
return to the dominance of more basic magmas from the
active arc.
Conclusions
The Pleistocene ignimbrites and post-ignimbrite lava
domes of the Purico complex in northern Chile represent
one of the few silicic ignimbrite centers in the central
Andes where compositional zoning can be linked to
ma®c recharge. The initial ignimbrite eruption at Purico
1LPI I) tapped homogeneous, crystal-rich dacite. During
the following LPI II event, a relatively hot crystal-poor
rhyolite was deposited ®rst in a highly explosive erup-
tion, followed by dacite, which was tapped along with
minor amounts of compositionally distinct andesites.
Post-ignimbrite lava domes are dacitic and also contain
andesitic clasts. The andesite, dacite, and rhyolite pum-
ices from the zoned LPI II units have indistinguishable
Sr and Nd isotopic compositions 1ꢀNd=)6.7 to )7.2
and 87Sr/86Sr=0.7085±0.7090), but petrological and
geochemical evidence indicate that they do not represent
a simple fractionation sequence. The dacites seem to
have equilibrated at 4±8 km depth and moderate tem-
peratures 1780 °C) under water-saturated conditions
1ꢀ4 wt% H2O). The pre-eruptive temperature of the
rhyolite is hotter than the dacite and trace element data
from melt inclusions and matrix glass indicate that the
dacite magma is too evolved to be the source of the
rhyolite. Compositional patterns of zoned plagioclase in
the rhyolite indicate strong local temperature gradients
in the magma.
To explain these features we envisage injection of
fresh andesite into the dacitic magma chamber between
the LPI I and LPI II eruptions. Thermal and chemical
disequilibrium with dacite promoted rapid crystalliza-
tion and e€ective fractionation of andesite to rhyolite.
The hot, low density and low viscosity rhyolitic melt was
able to ascend without signi®cant equilibration with the
resident dacite. The ascent and concomitant decom-
pression of this volatile-rich rhyolite is a likely trigger for
the plinian eruption, which marks the beginning of the
LPI II ignimbrite.
The isotopic composition of Purico units suggest that
all magmas are hybrid with varying degrees of crustal
and arc-derived mantle components. Mixing calcula-
tions based on Sr isotopic ratios indicate a minimum
proportion of 30% arc-derived basaltic andesite in the
ignimbrites, and a decreasing crustal contribution with
time.
Acknowledgements Funding for this study was provided by the
Deutsche Forschungsgemeinschaft in the framework of the
collaborative research center SFB 267 ``Deformation processes in
the Central Andes''. We thank Moyra Gardeweg, SERNAGEO-
MIN Chile and colleagues from the Universidad Catolica del Norte
in Antofagsta for logistical help and advice during ®eld work. For
their help with analytical work we are grateful to our colleagues
from the GeoForschungsZentrum Potsdam, in particular Rudi
Naumann 1XRF), Erika Kramer 1REE), Wolfgang Siebel 1Sr- and
Nd-isotopes), Dieter Rhede and Oona Appelt 1EPMA), and Peter
Dulski 1LA-ICP-MS). Renate Karstadt skillfully drafted the geo-
logical map of Purico on the basis of ®eld and remote sensing data.
Some microprobe analyses were done at the University of Giessen
with the help and advice of Rudiger Borchardt. Rick Hervig, Kurt

698
Roggensack and other members of the geology department at
Arizona State University 1ASU) are thanked for their hospitality References
during a 6-month stay of A.K.S. and invaluable advice in doing
SIMS analysis on melt inclusions. The visit to ASU was made
Ahrens TA 1ed) 11995) A handbook of physical constants: Mineral
possible by a grant to A.K.S. from the Deutscher Akademischer
Austauschdienst. Reviews by Charles Bacon and an anonymous
journal reviewer helped clarify this presentation.
physics and crystallography, vol 2. American Geophysical
Union, Washington, DC, pp 1±354
Anders E, Grevesse N 11989) Abundances of the elements; mete-
oritic and solar. Geochim Cosmochim Acta 53:197±214
Andersen DJ, Lindsley DH 11988) Internally consistent solution
models for Fe±Mg±Mn±Ti oxides; Fe±Ti oxides. Am Mineral
73:714±726
Analytical procedures
Andersen DJ, Lindsley DH, Davidson PM 11993) QUILF: a Pascal
program to assess equilibria among Fe±Mg±Mn±Ti oxides,
pyroxenes, olivine, and quartz. Comp Geosci 19:1333±1350
Anderson JL, Smith DR 11995) The e€ects of temperature and
fO2 on the Al-in-hornblende barometer. Am Mineral 80:549±
559
Bacon CR, Druitt TH 11988). Compositional evolution of the
zoned calcalkaline magma chamber of Mount Mazama Crater
Lake Oregon. Contrib Mineral Petrol 98:224±256
Bacon CR, Hirschmann MM 11988) Mg/Mn partitioning as a test
for equilibrium between coexisting Fe±Ti oxides. Am Mineral
73:57±61
Best MG, Christiansen EH, Deino AL, Gromme CS, Tingey DG
11995) Correlation and emplacement of a large, zoned, dis-
continuously exposed ash ¯ow sheet; the 40Ar/39Ar chronology,
paleomagnetism, and petrology of the Pahranagat Formation,
Nevada. J Geophys Res B 100:24593±24609
Blake S11990) Viscoplastic models of lava domes. In: Fink JH 1ed)
Lava ¯ows and domes; emplacement mechanisms and haz-
ard implications. Springer, Berlin Heidelberg New York,
pp 88±126
Blake S, Ivey GN 11986a) Magma-mixing and the dynamics of
withdrawal from strati®ed reservoirs. J Volcanol Geotherm Res
27:153±178
Blake S, Ivey GN 11986b) Density and viscosity gradients in zoned
magma chambers and their in¯uence on withdrawal dynamics.
J Volcanol Geotherm Res 30:201±230
Blundy JD, Wood BJ 11991) 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, Weill D, Richet P 11982) Density calculations for sil-
icate liquids. I Revised method for aluminosilicate composi-
tions. Geochim Cosmochim Acta:46:909±919
Campbell IH, Turner JS11986) The in¯uence of viscosity in
fountains in magma chambers. J Petrol 27:1±30
Carey S, Sigurdsson H 11989) The intensity of plinian eruptions.
Bull Volcanol 51:28±40
Carey S, Sparks RSJ 11986) Quantitative models of the fallout
and dispersal of tephra from volcanic eruption columns.
Bull Volcanol 48:109±125
Chmielowski J, Zandt G, Haberland C 11999) The central Andean
Altiplano±Puna magma body. Geophys Res Lett 26:783±786
Davidson JP, de Silva SL, Holden P, Halliday AN 11990) Small-
scale disequilibrium in a magmatic inclusion and its more silicic
host. J Geophys Res B 95:17661±17675
de Silva SL 11989a) Altiplano±Puna volcanic complex of the central
Andes. Geology 17:1102±1106
de Silva SL 11989b) Geochronology and stratigraphy of the
ignimbrites from the 21°30¢ Sto 23 °30¢ Sportion of the
central Andes of northern Chile. J Volcanol Geotherm Res 37:
93±131
de Silva SL 11989c) The origin and signi®cance of crystal rich in-
clusions in pumices from two Chilean ignimbrites. Geol Mag
126:159±175
de Silva SL 11991) Styles of zoning in central Andean ignimbrites;
insights into magma chamber processes. In: Harmon RS, Ra-
pela CW 1eds) Andean magmatism and its tectonic setting, Geol
Soc Am Spec Pap 265:233±243
For determining pumice and lava crystallinity, X-ray
di€raction 1XRD) spectra of the samples and of refer-
ence materials of known crystal-content were compared.
Reference materials were prepared by mixing powders of
natural minerals plagioclase, quartz and hornblende,
and natural aphyric rhyolitic glass. Mineral phases were
quantitatively determined after Emmermann and La-
uterjung 11990). Relative 2r errors on major mineral
phase and glass abundance are generally between 5 and
10%. Densities of selected pumice and lava samples were
determined by weighing clasts in air and water following
the Houghton and Wilson 11989). Dense rock equivalent
density was determined by He-pycnometry on ®ne-
grained rock powders 1<63 lm) of compositionally
similar samples. The vesicularity of each clast was then
calculated from the measured density di€erences. Phe-
nocryst, microlite, and glass compositions of all pumice
and lava types were determined using electron microp-
robe analysis 1EPMA), secondary ionization mass
spectrometry 1SIMS), laser ablation mass spectrometry
with inductively coupled plasma 1LA-ICP-MS) and
infrared spectroscopy 1FTIR). For EPMA of glasses, a
10-nA beam at 15 kV accelerating voltage was rastered
over a 20-lm2 area to minimize Na loss. 1r errors 1%rel)
of EPMA analysis based on 32 replicate analyses on
natural rhyolitic glass: SiO2=0.3, TiO2=10.0, Al2O3=
6.1, Fe2O3=5.6, MnO=21.4, MgO=8.3, CaO=3.3,
K2O=1.1, Na2O=1.4, P2O5=50.0. SIMS spot analyses
on glasses and step-scans on plagioclase were made on a
modi®ed Cameca 3f ion microprobe at Arizona State
University and on a Cameca 6f ion microprobe at the
GeoForschungsZentrum Potsdam with an instrumental
set-up described in Ihinger et al. 11994). Standard ma-
terials used were NBS610 and NBS612 glasses for trace
elements and experimentally synthesized hydrous rhyo-
lite glasses with water determined by Karl±Fischer ti-
tration 1Westrich 1987). 1r errors 1% rel) of SIMS
analysis based on counting statistics and instrumental
background 1for H2O): H2O=0.15 wt% 1abs.), Rb=4.7,
Sr=6.2, Y=6.0, Nb=11.0, Ba=2.5, La=14.0, Ce=8.7,
Th=17.0, U=18.0. LA-ICP-MSmineral analyses were
carried out at the GeoForschungsZentrum Potsdam
using an UV-Microprobe attached to a PQS-S ICP-
mass-spectrometer by VG-Elemental 1see Schmitt 1999
for analytical details). For details of X-ray ¯uorescence
1XRF), ICP-MS, emission spectroscopy 1ICP-OES) and
conventional mass spectrometry 1TIMS) procedures the
reader is referred to Trumbull et al. 11999) and Schmitt
et al. 12000).
de Silva SL, Francis PW 11991) Volcanoes of the central Andes.
Springer, Berlin Heidelberg New York, 216 pp

699
de Silva SL, Self S, Francis PW, Drake RE, Ramirez RC 11994) Huppert HE, Sparks RSJ, Whitehead JA, Hallworth MA 11986)
E€usive silicic volcanism in the central Andes; the Chao Dacite
and other young lavas of the Altiplano±Puna volcanic complex.
J Geophys Res B 99:17805±17825
Replenishment of magma chambers by light inputs. J Geophys
Res B 91:6113±6122
Ihinger PD, Hervig RL, McMillan PF 11994) Analytical methods
for volatiles in glasses. In: Carroll MR, Holloway JR 1eds)
Volatiles in Magmas Rev Mineral 30:67±121
Johnson MC, Rutherford, MJ 11989) Experimentally determined
conditions in the Fish Canyon Tu€, Colorado, magma cham-
ber. J Petrol 30:711±737
Droop GTR 11987) A general equation for estimating Fe3+
concentrations in ferromagnesian silicates and oxides from
microprobe. Mineral Mag 51:431±437
Dunn T, Sen C 11994) Mineral/ matrix partition coecients for
orthopyroxene plagioclase and olivine in basaltic to andesitic
systems;
Geochim Cosmochim Acta 58:717±733
a combined analytical and experimental study. Keppler H, Wyllie PJ 11990) Role of ¯uids in transport and frac-
tionation of uranium and thorium in magmatic processes.
Nature 348:531±533
by dike injection: contrasting cases at Karymsky Volcano, Lange RA 11994) The e€ect of H2O, CO2 and F on the density and
Eichelberger JC, Izbekov PE 12000) Eruption of andesite triggered
Kamchatka and Mt Katmai, Alaska. Philos Trans R Soc Lond
358:1465±1485
Eichelberger JC, Chertko€ DG, Dreher ST, Chrisopher JN 12000) Laube N, Springer J 11998) Crustal melting by ponding of ma®c
viscosity of silicate melts. In: Carroll MR, Holloway JR 1eds)
Volatiles in magmas. Rev Mineral 30:331±369
Magmas in collision: rethinking chemical zonation in silicic
magmas. Geology 28:603±606
Elston WE 11984) Mid-Tertiary ash ¯ow tu€ cauldrons, south- Lindsay JM 11999) Geology and magmatic evolution of large
magmas;
81:19±35
a numerical model. J Volcanol Geotherm Res
western New Mexico. J Geophys Res B 89:8733±8750
Emmermann R, Lauterjung J 11990) Double X-ray analysis of
cuttings and rock ¯our: a powerful tool of rapid and reliable
volume felsic ignimbrites of the La Pacana caldera central
Andes Chile. PhD Thesis, University of Giessen. Sci Tech
Rep 99/16, GeoForschungsZentrum Potsdam
determination of borehole lithostragraphy. Sci Drill 1:269±282 Lindsay JM, Schmitt AK, Trumbull RB, de Silva SL, Siebel W,
Ewart A, Grin WL 11994) Application of proton-microprobe
data to trace-element partitioning in volcanic rocks. Chem Geol
117:251±284
Emmermann R 12000) Magmatic evolution of the La Pacana
Caldera system, central Andes, Chile: compositional variation
of two cogenetic, large-volume felsic ignimbrites and implica-
tions for contrasting eruption mechanisms. J Petrol 1in press)
Folch A, Marti J 11998) The generation of overpressure in felsic
magma chambers by replenishment. Earth Planet Sci Lett Lindsley DH 11983) Pyroxene thermometry. Am Mineral
163:301±314 68:477±493
Francis PW, Baker MCW 11978) Sources of two large ignimbrites Lipman PW 11997) Subsidence of ash-¯ow calderas: relation to
in the central Andes; some Landsat evidence. J Volcanol caldera size and magma-chamber geometry. Bull Volcanol
Geotherm Res 4:81±87 59:198±218
Francis PW, McDonough WF, Hammill M, O'Callaghan LJ, Lipman PW, Doe BR, Hedge CE, Steven TA 11978) Petrologic
Thorpe RS11984) The Cerro Purico shield complex north evolution of the San Juan volcanic ®eld. Southwestern Colo-
Chile. In: Harmon RS, Barreiro BA 1eds) Andean magmatism; rado; Pb and Sr isotope evidence. Geol Soc Am Bull 89:59±82
chemical and isotopic constraints. Shiva Publications, Kent, Lu F, Anderson AT, Davis AM 11995) Di€usional gradients at the
pp 106±123
Francis PW, Sparks RSJ, Hawkesworth CJ, Thorpe RS, Pyle DM,
crystal/melt interface and their e€ect on the composition of melt
inclusions. J Geol 103:591±597
Tait SR, Mantovani MSM, McDermott F 11989) Petrology and Lucassen F, Franz G, Thirlwall M, Metzger K 11999) Crustal re-
geochemistry of volcanic rocks of the Cerro Galan caldera
northwest Argentina. Geol Mag 126:515±547
Gans PB, Mahood GA, Schermer ER 11989) Synextensional
cycling of metamorphic basement: late Paleozoic granites of
northern Chile 1ꢀ22°S) Implications for the composition of the
Andean crust. J Petrol 40:1527±1551
magmatism in the Basin and Range Province; a case study from Matthews SJ, Jones AP, Gardeweg MC 11994) Lascar Volcano
the eastern Great Basin. Geol Soc Am Spec Pap 233:1±53
Gardeweg M, Ramirez CF 11987) La Pacana Caldera and the
northern Chile; evidence for steady-state disequilibrium. J Pet-
rol 35:401±432
Atana ignimbrite; a major ash-¯ow and resurgent caldera McDowell FW, Clabaugh SE, Elston WE 11979) Ignimbrites of the
complex in the Andes of northern Chile. Bull Volcanol
49:547±566
Ghiorso MS, Sack RO 11995) Chemical mass transfer in magmatic
Sierra Madre Occidental and their relation to the tectonic
history of western Mexico. In: Chapin CE, Elston WE 1eds)
Ash-¯ow tu€s. Geol Soc Am Spec Pap 180:113±124
processes. IV, A revised and internally consistent thermody- Ort MH, Coira BL, Mazzoni MM 11996) Generation of a crust±
namic model for the interpolation and extrapolation of liquid±
solid equilibria in magmatic systems at elevated temperatures
and pressures. Contrib Mineral Petrol 119:197±212
mantle magma mixture; magma sources and contamination at
Cerro Panizos central Andes. Contrib Mineral Petrol
123:308±322
Hawkesworth CJ, Hammill M, Gledhill AR, van Calsteren P, Patin
Rogers G 11982) Isotope and trace element evidence for late-
stage intra-crustal melting in the High Andes. Earth Planet Sci
Ä
o-Douce AE, Beard JS11995) Dehydration-melting of biotite
gneiss and quartz amphibolite from 3 to 15 kbar. J Petrol
36:707±738
Lett 58:240±254
Pinkerton H, Stevenson RJ 11992) Methods of determining the
rheological properties of magmas at sub-liquidus temperatures.
J Volcanol Geotherm Res 53:47±66
Hildreth W 11981) Gradients in silicic magma chambers; implica-
tions for lithospheric magmatism. J Geophys Res B 86:10153±
10192
Roedder E 11984) Fluid inclusions. Rev Mineral 30:646
Holland JD, Blundy TJB 11994) Calcic amphibole equilibria and a Scaillet B, Holtz F, Pichavant M 11998) Phase equilibrium con-
new amphibole±plagioclase geothermometer. Contrib Mineral straints on the viscosity of silicic magmas. 1. Volcanic±plutonic
Petrol 104:208±224 comparison. J Geophys Res B 103:27257±27266
Holtz F, Johannes W 11994) Maximum and minimum water con- Schmitt AK 11999) Melt generation and magma chamber processes
tents of granitic melts: implications for chemical and physical
properties of ascending magmas. Lithos 32:149±159
Houghton BF, Wilson CJN 11989) A vesicularity index for
pyroclastic deposits. Bull Volcanol 51:451±462
Housh TB, Luhr JF 11991) Plagioclase-melt equilibria in hydrous
systems. Am Mineral 76:477±492
in the Purico complex and implications for ignimbrite forma-
tion in the Central Andes. PhD Thesis, University of Giessen.
Sci Tech Rep 99/18 GeoForschungsZentrum Potsdam
Schmitt AK, de Silva SL 12000) The Merzbacher and Eggler 11984)
geohygrometer: a cautionary note on its suitability for high-K
suites. J Petrol 41:357±362
Huppert HE, Sparks RSJ, Turner JS 11984) E€ects of volatiles on Schmitt AK, Emmermann R, Trumbull RB, Buhn B, Henjes-Kunst
mixing in calc-alkaline magma systems. Nature 297:554±557
F 12000) Petrogenesis and 40Ar/39Ar geochronology of the

700
Brandberg complex Namibia: evidence for a major mantle Tepley FJ III, Davidson JP 11997) Recharge in volcanic systems;
contribution in metaluminous and peralkaline granites. J Petrol evidence from isotope pro®les of phenocrysts. Science
41:1207±1239 275:826±829
Schmitz M, Heinsohn WD, Schilling FR 11997) Seismic, gravity Trumbull RB, Wittenbrink R, Hahne K, Emmermann R, Busch.
and petrological evidence for partial melt beneath the thick-
ened central Andean crust 121±23°S). Tectonophysics
270:313±326
W, Gerstenberger H, Siebel W 11999) Evidence for Late Mio-
cene to Recent contamination of arc andesites by crustal melts
in the Chilean Andes 25±26°Sand its geodynamic implications.
J SAm Earth Sci 12:135±155.
Schumacher JC 11997) The estimation of ferric iron in electron
microprobe analysis of amphiboles. Mineral Mag 61:312±321
Shaw HR 11972) Viscosities of magmatic silicate liquids; an
empirical method of prediction. Am J Sci 272:870±893
Singer BS, Dungan MA, Layne GD 11995) Textures and Sr, Ba,
Watts RB, de Silva SL, Jimenez de Rios G, Croudace I 11999)
E€usive eruption of viscous silicic magma triggered and driven
by recharge: a case study of the Cerro Chascon±Runtu Jarita
Dome complex in southwest Bolivia. Bull Volcanol 60:241±264.
Mg, Fe, K and Ti compositional pro®les in volcanic plagioclase Westrich HR 11987) Determination of water in volcanic glasses by
clues: to the dynamics of calc-alkaline magma chambers. Am
Mineral 80:776±798
Sisson TW, Bacon CR 11999) Gas-driven ®lter pressing in magmas.
Geology 27:613±616
Smith RL 11979) Ash-¯ow magmatism. In: Chapin CE, Elston WE
1eds) Ash-¯ow tu€s. Geol Soc Am Spec Pap 180:5±27
Sparks RSJ, Francis PW, Hamer RD, Pankhurst RJ, O'Callaghan
LO, Thorpe RS, Page R 11985) Ignimbrites of the Cerro Galan
Caldera, NW Argentina. J Volcanol Geotherm Res 24:205±248
Karl±Fischer titration. Chem Geol 63:335±340
Wilson L, Walker GPL 11987) Explosive volcanic eruptions VI.
Ejecta dispersal in plinian eruptions: the control of eruption
conditions and atmospheric properties. Geophys J Royal
Astrom Soc 89:657±679
Wol€ JA, Worner G, Blake S11990) Gradients in physical
parameters in zoned felsic magma bodies; implications for
evolution and eruptive withdrawal. J Volcanol Geotherm Res
43:37±55