Earth and Planetary Science Letters 154 Ž1998.

153–166

Subduction controls on the compositions of lavas from the

Ecuadorian Andes
a,1 a,)
Roberto Barragan , Dennis Geist , Minard Hall b, Peter Larson c , Mark Kurz d

a
Department of Geology and Geological Engineering, UniÕersity of Idaho, Moscow, ID 83844, USA
b
Instituto Geofisico, Escuela Politecnica Nacional, Quito, Ecuador
c
Geology Department, Washington State UniÕersity, Pullman, WA 99164, USA
d
Department of Marine Chemistry and Geochemistry, Woods Hole Oceanographic Institution, Woods Hole, MA 02543, USA
Received 30 September 1996; revised 7 July 1997; accepted 1 August 1997

Abstract

Three volcanoes of the Ecuadorian Andes, Atacazo, Antisana, and Sumaco, lie in a transect perpendicular to the trench
and the main trend of the Andean arc. Each of the volcanoes lies on crust of substantially different age, composition, and
thickness. Few compositional or isotopic features correspond in a straightforward way to the type of the crust through which
the magmas have passed. Isotopic data limit assimilation to - 15% at each of the volcanoes. Instead, a systematic
relationship exists between the compositions of the lavas and the depth to the Benioff zone, suggesting that subduction
imparts the principal control on the compositions of the magmas. Atacazo’s lavas have low concentrations of the
incompatible trace elements and very large LILrHFS ratios. Sumaco’s lavas are strongly enriched in the incompatible trace
elements and have small LILrHFS ratios. Antisana’s lavas are intermediate in almost every respect. These features are
consistent with devolatilization of the subducted slab controlling the extent of partial melting of a depleted mantle source. A
mixing and melting model suggests the volcanic front magmas are made by large extent of partial melting Ž; 15%. and
include a large slab input Ž1.1% added to the depleted mantle.. The magmas of the middle belt of volcanoes are made by
smaller extent of partial melting Ž3%., induced by moderate amounts of slab-derived fluid Ž0.06%.. The back arc magmas
result from small degrees of melting Ž2%. and small slab input. q 1998 Elsevier Science B.V.

Keywords: Ecuador; Andes; subduction; andesites; assimilation

1. Introduction of magmas at an active continental margin w1–3x.

The Chilean Andes provide an excellent setting to
A series of studies conducted on rocks from the test for assimilation, because the crust thickens from
southern and central Andes have demonstrated the about 30 to 60 km along the strike of the arc. In the
important role of crustal assimilation in the genesis areas underlain by the thinner crustal sections the
geochemical characteristics of the lavas are thought
to be mostly imparted by the subducted lithosphere
)
Corresponding author. E-mail: [email protected]
and overlying mantle wedgew4,5x, but compositional
1
Present address: Oryx Ecuador Energy Company, P.O. Box features that correspond to changing crustal thick-
17-11-6545, Quito, Ecuador. ness have been attributed to assimilation w1–3x. These

0012-821Xr98r$19.00 q 1998 Elsevier Science B.V. All rights reserved.

PII S 0 0 1 2 - 8 2 1 X Ž 9 7 . 0 0 1 4 1 - 6
154 R. Barragan et al.r Earth and Planetary Science Letters 154 (1998) 153–166

Fig. 1. Terrane map of northern Ecuador Žadapted from w6,8x.. Closed triangles are volcanoes. Toothed line indicates trench.

findings suggest that the crust at Andean-type mar- predictable way to the composition or thickness of
gins may be dominated by recycling of older crustal the crust on which they sit. Instead, most of the
material, thus only small amounts of juvenile crust geochemical features are simply related to the depth
are produced. The other important implication is that of the Benioff zone. We conclude that the composi-
continental-arc magmas may be so strongly modified tion of these lavas is mostly determined by processes
by crustal contamination that little can be discerned occurring in the slab and mantle, and in fact neither
about their primary source in the subducted slab and crustal nor lithospheric-mantle interaction plays an
upper mantle. important role.
This paper reports the results of a study of three
volcanoes, Atacazo, Antisana, and Sumaco, that lie
perpendicular to the strike of the Ecuadorian Andes,
representing a distance of ; 100 km across an active 2. Analytical techniques
volcanic arc ŽFig. 1.. The Ecuadorian Andes are
fundamentally different from the southern Andes, Major-element and Ni, Cr, Sc, V, Ba, Rb, Sr, Zr,
especially in the geologic history of the upper plate. Y, and Nb analyses were by XRF in the Geoanalyti-
Specifically, several allochthonous terranes were ac- cal Laboratory at Washington State University ŽPeter
creted to the northwest part of the South American Hooper, pers. commun., 1997.. Precision was tested
continent between the Middle Jurassic and Late by multiple analyses of a single specimen. Major-
Eocene w6–8x. Consequently, magmas of the Ecuado- element precision is - 2% of the absolute abun-
rian Andes have traversed crust of vastly different dance; trace-element precision is - 5% except for
age, composition, and thickness. We have found that Rb and Nb, which are 10% in low-abundance sam-
the compositions of these lavas are not related in any ples. Rare-earth elements ŽREE., Hf, Ta, Th, and U
 R. Barragan et al.r Earth and Planetary Science Letters 154 (1998) 153–166 155

were analyzed by INAA at Oregon State University. 3. Geologic setting

Uncertainties are 7% for U and Ce, and - 5% for
the other elements ŽM. Streck, pers. commun., 1994.. The Ecuadorian Andes result from the subduction
Oxygen isotope analyses were performed using tech- of the Nazca plate beneath the South American plate.
niques reported in w9x; uncertainty is "0.1‰. Sr and The active volcanic arc lies north of 28S ŽFig. 1.. To
Nd isotopic analyses were performed by techniques the south of 28S, the Nazca plate subducts at an
reported in w10x. Internal precision is typically angle of ; 158, creating an amagmatic convergent
"0.000020 for 86 Srr 87 Sr and "0.000010 for boundary. To the north, in the study area, the dip of
143
Ndr 144 Nd. Ratios are reported relative to the Benioff zone is between 258 and 308 w11x. The
0.710240 for SRM987 for 86 Srr 87 Sr and relative to active composite volcanoes of Ecuador lie in two
0.511852 for the La Jolla standard. parallel chains called the Cordillera Occidental Žwest.

Fig. 2. Major-element Harker diagrams of lavas from Atacazo, Antisana, and Sumaco volcanoes. Fields on K 2 O diagram from w16x.
156 R. Barragan et al.r Earth and Planetary Science Letters 154 (1998) 153–166

and the Cordillera Real Žeast.. The two chains lie is made up of the Pinon˜ terrane, an allochthonous
40–50 km apart and are separated by the inter- terrane of basaltic basement of Cretaceous age over-
Andean depression w12x. A third volcanic zone, con- lain by younger arc rocks w13x. It is thought to have
sisting of back-arc volcanoes, lies in the Amazonian become sutured to South America in the Early Eocene
basin, east of the Andes. w6x. Bouguer anomalies of up to q162 mGal indicate
A notable feature of these three volcanic zones is that the entire 30 km crustal section is mafic w14x.
that each lies on a different crustal province, marked Paleocene to Early Eocene calc-alkaline arc rocks
by differences in crustal age, composition, and thick- Žcalled the ‘‘Macuchi’’ formation. are tectonically
ness and separated by steeply-dipping shear zones emplaced on the east side of the Pinon˜ block. The
w6–8x. The western part of Ecuador Žnorth of 2.58S. ˜ terrane is the Cala-
eastern boundary of the Pinon

Fig. 3. Selected trace-element variation diagrams of lavas from Atacazo, Antisana, and Sumaco volcanoes.
 R. Barragan et al.r Earth and Planetary Science Letters 154 (1998) 153–166 157

cali–Pallatanga–Palenque Žor ‘‘Romeral’’. fault, this basement, and is overlain in turn by a sedimen-
which bounds the west side of the inter-Andean tary fan of Late Jurassic to modern age whose source
depression in this region ŽFig. 1.. has been the Andes w7x. Bouguer gravity anomalies
The volcanoes of the Cordillera Real are under- indicate that the crustal thickness in this area is a
lain by metamorphosed Paleozoic and Mesozoic S- typical 35–40 km w14x. Abundant seismic reflection
type granites and metasedimentary rocks of continen- profiles reveal many thrust faults, of Late Cretaceous
tal affinity w6,8x. The east margin of this terrane is to Quaternary age, which are apparently related to
defined by large-scale, steeply west-dipping faults transpression driven by plate convergence w7x. No
that have thrust the metamorphic rocks of the young extensional structures have been identified
Cordillera Real over unmetamorphosed continental- that could conceivably be related to the back-arc
platform sedimentary rocks to the east. The Cordillera setting and the alkaline volcanism.
Real is characterized by negative Bouguer anomalies
as great as y292 mGal, indicating low-density crust 3.1. Atacazo Õolcano
to depths of nearly 60 km w14x.
The South American craton of the Amazon basin Atacazo volcano is located in the Cordillera Occi-
lies to the east of the Andes. Deep oil wells have dental ; 25 km southwest of Quito and rises to
penetrated granulites thought to be from the western 4455 m. An older, mid-Pliocene to Pleistocene
part of the Guyana shield w6x. A thick sequence of Ž410,000 " 10,000 to 300,000 " 10,000 y. andesitic
Paleozoic and Mesozoic sedimentary rocks rests upon edifice is truncated by a large late Pleistocene–Holo-

Fig. 4. MORB-normalized incompatible-element diagram of selected lavas from the three volcanoes. Normalization factors and element
order from w29x.
158 R. Barragan et al.r Earth and Planetary Science Letters 154 (1998) 153–166

Fig. 5. Variation in element ratios with respect to the depth to the Benioff zone w11x for lavas from Atacazo, Antisana, and Sumaco.
Numerators are elements that are abundant in slab-derived fluid and denominators are elements substantially less mobile w30–32x.
Continental crust Ž CC . w27x and mantle Ž M . w29x values are also shown.

cene summit caldera that opens to the southwest w15x. pronounced negative Nb and Ta anomalies on
The older lavas are dominantly porphyritic, two-py- MORB-normalized diagrams ŽFig. 4..
roxene andesites. Two younger andesitic to dacitic
3.2. Antisana Õolcano
domes are located on the outer southeastern flank of
the caldera, and three dacitic domes lie within the 6 Antisana volcano, one of the highest mountains in
km wide, 900 m deep caldera. The youngest dome, Ecuador, rises to an altitude of 5735 m and is located
Ninahuilca, erupted only 2370 " 70 years ago w15x. in the Cordillera Real, 50 km southeast of Quito. It is
The dome lavas are two-pyroxene andesites, horn- surrounded by older volcanic rocks, but these in turn
blende andesites, and hornblende dacites. lie upon S-type granites and metasedimentary rocks
The lavas of Atacazo belong to the medium-K of the Loja division w8x. An older edifice makes up
calc-alkaline clan w16x ŽFig. 2.. Atacazo’s lavas are the eastern half of the volcano and is truncated by
notably poor in the incompatible trace elements, and several large amphitheaters that open to the inacces-
the compatible elements ŽNi, Sc, and V. decrease sible northeast, south, and east slopes. A younger
with increasing SiO 2 concentrations ŽFig. 3.. Ata- composite cone grew upon the northwest part of the
cazo’s lavas are all light-REE enriched but have old edifice and constitutes the bulk of the volcano. In
lower LarYb than lavas from the other two volca- addition, two lava flows erupted from satellite fis-
noes ŽFigs. 4 and 5.. All of Atacazo’s lavas have sures 15 km west of the base of the cone in the 18th
 R. Barragan et al.r Earth and Planetary Science Letters 154 (1998) 153–166 159

century w17x. All Antisana lavas examined in this Dataset 2 ., the differences among the volcanoes is
study contain 30–40% plagioclase ŽAn 34 – 46 ., 10% overwhelmingly greater than the variation within the
hypersthene, 2% augite, and sparse olivine ŽFo 62 – 70 . individual suites. Thus, the focus of this work is a
phenocrysts. The satellite flows contain sparse re- comparison of the three volcanoes.
sorbed quartz.
Antisana’s lavas are of the high-K calc-alkaline 4.1. Isotopic tests of assimilation
type ŽFig. 2.. Most of the samples plot along simple
linear trends on Harker diagrams. The incompatible The 87 Srr 86 Sr ratios of lavas from these volca-
elements are richer in Antisana lavas than those of noes range from 0.70416 to 0.70428, only slightly
Atacazo, especially the highly-incompatible trace el- higher than typical Žor ‘‘isotopically coherent’’ w20x.
ements ŽFigs. 3 and 4.. Compatible-element trends island arc lavas, but well within the global range of
are indiscernible from those of Atacazo ŽFig. 3.. island arcs ŽFig. 7.. Likewise, the 143 Ndr 144 Nd
ratios of the Ecuadorian lavas range from 0.51279 to
3.3. Sumaco Õolcano 0.51289, slightly lower than typical island arc lavas
Ž0.51297 to 0.51305. w20x. d 18 O values of the
Sumaco volcano rises from the Amazonian basin, Ecuadorian lavas range from q6.1 to q8.1‰, with
to the east of the Andes and 105 km east of Quito the highest values coming from Atacazo; d 18 O val-
w18x. The volcano is constructed upon a Phanerozoic ues from Antisana and Sumaco volcano fall within
sedimentary sequence that overlies cratonic South the island arc range of q6.0 " 0.3‰ w21x.
America w7x. Sumaco has a symmetrical composite Measurable assimilation of continental crust may
morphology, and there are reports of three explosive be indicated by the higher d 18 O values of the Ata-
eruptions, the most recent in 1933 w19x. The lavas of cazo lavas and the lower 143 Ndr 144 Nd of the Anti-
Sumaco are distinctly alkaline and are feldspathoid- sana lavas, but mass balance indicates that the amount
bearing tephrites, basanites, and phonolites. The lavas of assimilation must be small. The high d 18 O values
are porphyritic and contain phenocrysts of hauyne measured at Atacazo may be accounted for by ; 15%
and nosean Ž5–8%., leucite, sodic augite, and plagio- assimilation of crustal rocks with a d 18 O value of
clase ŽAn 33 – 53 .. Olivine, apatite, and magnetite oc- q20‰, assuming a parental magma with a typical
cur in most samples as well. Small crystals of island-arc value of q6.0‰ Žcf. w22x.. If assimilation
nepheline occur in the groundmass. The alkaline is responsible for increasing the d 18 O values of the
mineral assemblage is reflected in the major-element Atacazo magmas, the likely assimilant would be
compositions of the lavas, which shows their altered Cretaceous mafic rocks of the Pinon˜ terrane.
shoshonitic affinity ŽFig. 2.. Sumaco’s lavas are rich Other considerations suggest that 15% assimilation is
in all of the incompatible elements and light-REE an overestimation. The very low Nb and Ta abun-
enriched ŽFigs. 3 and 4.. Ta and Nb anomalies are dances in Atacazo lavas Žsome with Nb - 1.5 ppm
much smaller than in the lavas of Antisana and and Ta - 0.20 ppm. preclude significant assimila-
Atacazo ŽFig. 4.. Despite their low silica contents tion of crustal rocks. Also, the reported measure-
and mafic appearance, Sumaco’s lavas are strongly ments are on whole-rock powders; recent work in
evolved, with lower MgO, Cr, and Ni abundances our laboratories indicates that incipient weathering in
than the andesites of Atacazo and Antisana ŽFig. 3.. optically-pristine oceanic basalts may raise the d 18 O
value by ; 0.5–1.0‰ w23x. It is also possible that
the parental magmas of the Atacazo andesites have
4. Geochemical comparisons of the volcanoes unusually high d 18 O, but this cannot be tested until
a thorough study of intra-suite variation of Atacazo
The outstanding feature of this transect across the is completed.
Ecuadorian Andes is the systematic change in com-
position of the lavas away from the trench ŽFigs.
2–6.. For nearly all of the analyzed elements ŽTable 2
http:r r w w w .elsevier.nlr locater epsl, m irror site:
1; for complete data see EPSL Online Background http:rrwww.elsevier.comrlocaterepsl
160 R. Barragan et al.r Earth and Planetary Science Letters 154 (1998) 153–166

Fig. 6. Variation in the abundances of incompatible trace elements in lavas from the Ecuadorian Andes with distance above the Benioff
zone.

The 143 Ndr 144 Nd ratios of Antisana lavas are tion is direct comparison of Atacazo, which lies on
consistently lower than those of the other two volca- young, thin crust of oceanic affinity, to Sumaco,
noes ŽFig. 7.. This probably relates to the fact that which lies on cratonic crust. Atacazo’s lavas are
these magmas have ascended through an exceedingly much richer in SiO 2 than Sumaco’s, a relation oppo-
thick section of continental crust and assimilated site that expected for assimilation of crustal material.
some of it. Nonetheless, the extent of assimilation is Moreover, the 87 Srr 86 Sr and 143 Ndr 144 Nd ratios of
probably small. For example, simple assimilation of lavas from the two volcanoes are virtually identical
8% continental crust with 143 Ndr 144 Nd of 0.5120 is ŽFig. 7.. This is in striking contrast to the volcanoes
required to lower a magma’s 143 Ndr 144 Nd ratio of Chile, where the 87 Srr 86 Sr ratios correlate quite
from 0.51288 to 0.51279, assuming similar Nd con- strongly with crustal thickness, increasing from
centrations in the basaltic to andesitic magma and 0.7036 to 0.7057 as the crust thickens from 30 to 60
the continental crustal materials. km w1x.
Other than these two marginally-significant ex- Sumaco’s lavas are fundamentally different from
ceptions, none of the elements, elemental ratios, and other back-arc shoshonitic provinces of the Andes
isotopic ratios for which we have data change in a whose lavas have large contributions from the crust
logical way in relation to the type or thickness of the and lithospheric mantle. For example, lavas from the
crust on which the volcanoes sit, as would be the Pocho province in Argentina are believed to have
case if the magmas result from extensive assimila- assimilated large amounts of lithospheric mantle or
tion of the lower crust w1x. The most obvious illustra- lower-crustal granulites w24x, because they are no-
 R. Barragan et al.r Earth and Planetary Science Letters 154 (1998) 153–166 161

Table 1
Representative analyses
Atacazo volcano Antisana volcano
At 01 8557 At 3.2 An HHJ-An GS-3 3D2
SiO 2 59.00 59.43 58.28 62.68 54.12 47.56
Al 2 O 3 17.05 15.85 15.70 16.02 19.74 18.89
TiO 2 0.65 0.67 0.88 0.73 0.76 1.29
FeO ) 7.18 7.83 6.92 5.81 6.09 8.99
MnO 0.11 0.12 0.11 0.09 0.22 0.32
CaO 7.05 6.71 6.42 4.91 6.01 8.63
MgO 4.25 4.90 5.46 2.69 1.67 2.49
K 2O 0.89 0.98 2.30 2.84 4.31 4.43
Na 2 O 3.68 3.38 3.72 4.01 6.61 6.77
P2 O5 0.12 0.12 0.23 0.22 0.45 0.64
Zr 91 90 171 183 403 441
Ni 38 60 91 31 9 4
Sc 24 20 20 10 4 1
V 166 136 167 129 132 292
Ba 384 406 661 827 2806 2563
Rb 19 20 84 111 106 125
Sr 362 351 580 583 2542 4059
Zr 91 90 171 183 403 441
Y 16 13 18 17 35 49
Nb 2.9 3.4 9.8 11.8 54.0 82.0
87
Srr 86 Sr 0.704187 0.704304 0.704213 0.704194 0.704195 0.704243
143
Ndr 144 Nd 0.512850 0.512890 0.512796 0.512799 0.512894 0.512893
d 18 O q8.1 q8.0 q6.7 q6.1 q6.2 q6.2
Sc 20.4 6.4 16.8 9.9 3.4 2.5
Cs 0.44 0.35 2.41 5.22 2.57 2.21
La 6.9 7.4 24.9 29.8 135.0 173.0
Ce 12.5 14.5 59.3 64.4 237.0 302.0
Nd 8.9 7.4 23.5 26.2 87.6 121.0
Sm 2.41 2.36 4.82 5.07 14.00 18.50
Eu 0.78 0.73 1.13 1.06 3.65 5.03
Tb 0.44 0.25 0.49 0.46 1.08 1.59
Yb 1.50 0.48 1.60 1.52 3.38 4.67
Lu 0.22 0.07 0.25 0.24 0.53 0.63
Hf 2.36 2.85 4.39 4.73 6.73 5.65
Ta 0.22 0.24 0.70 0.94 2.68 3.54
Th 1.2 1.3 12.2 16.1 29.1 27.8
U 0.73 5.19 6.39 10.70 7.26

Techniques are discussed in the text. Complete data set available from EPSL Online Žhttp:rrwww.elsevier.nlrlocaterepsl, mirror site:
http:rrwww.elsevier.comrlocaterepsl..

tably poor in Rb, U, and Th and have low LarYb mantle. Sumaco’s shoshonitic lavas are isotopically
and 143 Ndr 144 Nd ratios. Sumaco’s lavas lack these identical to the arc lavas, thus are likely related to
features. Shoshonitic lavas from the Puna plateau similar sources without substantial contributions by
w25x and Bolivian Altiplano w26x have even higher the ancient lithospheric mantle that underlies the
87
Srr 86 Sr and lower 143 Ndr 144 Nd ratios ratios, re- South American craton.
flecting substantial Ž) 20%. upper-crustal contribu- Much of the discussion that follows focuses on
tions and perhaps an origin in ancient lithospheric the trace-element ratios BarNb and LarYb. Assimi-
162 R. Barragan et al.r Earth and Planetary Science Letters 154 (1998) 153–166

Fig. 7. Variation of Sr and Nd isotopes in lavas from the Atacazo, Antisana, and Sumaco volcanoes, showing fields from Andes Southern,
Central, and Northern Volcanic Zones ŽSVZ, CVZ, NVZ; from w2x for comparison. d 18 O values indicated in italics.

lation of typical continental crustal material w27x in lated, the - 15% of assimilation that is indicated by
any proportion is unlikely to produce the systematic the Sr, Nd, and O isotopic ratios is unlikely to
changes observed in the Ecuadorian lavas ŽFig. 8.. change markedly the LarSm or BarNb ratios of the
Even if atypical crustal materials are being assimi- magmas. Because the silica contents, incompatible-

Fig. 8. Sr and Nd isotopic ratios compared to LarYb and BarNb ratios. Rectangles indicate extreme limits of mixing between typical
mantle material Ž M . w29x and typical crustal materials Ž C . w27x. Most of the Ecuadorian lavas lie outside this limit.
 R. Barragan et al.r Earth and Planetary Science Letters 154 (1998) 153–166 163

element concentrations and ratios, and Sr, Nd and O that the magmas are the result of progressively lower
isotopic ratios do not relate in a logical way to degrees of partial melting of the mantle wedge Že.g.,
crustal thickness or the type of crust through which w4,5x..
the magmas ascend, we conclude that assimilation is
not important in bringing about the first-order differ- 4.4. Mixing calculations
ences in the compositions of these lavas.
The first-order compositional differences between
4.2. Different slab contributions the magmas from the three volcanoes are thought to
be controlled by two parameters: the amount of
It is well established that ratios of the large-ion slab-derived fluid and the extent of melting. These
lithophile elements ŽLIL; e.g., K, Rb, Cs, Ba. to two parameters are directly related in that large fluid
high-field strength metals ŽHFS; e.g., Zr, Nb, Ta, Ti. contributions correspond to large degrees of melting
and the REE are higher in island-arc lavas than they Žcf. Figs. 5 and 6.. All of the lavas considered here
are in mid-ocean ridge and ocean-island basalts w28x. are evolved andesites, thus the absolute concentra-
This fundamental difference has been attributed to tions of the incompatible trace elements in the pri-
mass transfer of a fluid Žmelt or volatile-rich phase. mary magmas are almost unconstrained. Therefore,
from the subducted slab to the overlying mantle we focus our modeling effort on ratios of incompati-
wedge. ble trace elements that are unlikely to be strongly
Lavas from Atacazo volcano have LILrHFS and modified by crystal fractionation or small amounts of
LILrREE ratios similar to some island arc andesites. assimilation.
In stark contrast, lavas from Sumaco have much We have generated a melting and mixing model
lower LILrHFS and LILrREE ratios, nearly as low that mimics diagnostic trace-element ratios of the
as those typical of ocean-island lavas. Lavas from three suites ŽFig. 9.. The model assumes that the
Antisana volcano have intermediate ratios. Overall, compositional differences in these lavas are due to
there is a clear relation between the depth to the two principal factors: Ž1. different degrees of partial
Benioff zone and ratios of elements thought to be melting of a depleted mantle source Žas indicated by
mobile in slab-derived fluids to those thought to be the Nd isotopic ratios.; and Ž2. different contribu-
immobile ŽFig. 5.. This systematic change is not tions of a fluid or magma derived from the sub-
likely due to assimilation brought about by greater ducted lithosphere. These models are obviously not
crustal thickness or in the type of crust, as the ratios unique, but show that this explanation is workable
are intermediate at Antisana Žthickest crust. and low- with reasonable parameters. Batch modal melting of
est at Sumaco Žmature continental crust.. Instead, the depleted MORB-producing mantle is assumed
systematic change in these ratios is likely due to a Žsource composition calculated from the parameters
decreasing influence of slab-derived fluid into the given in w29x., and bulk melt–solid distribution coef-
mantle wedge as the Nazca plate descends further ficients are assigned as La s 0.01, Yb s 0.8, Ba s
beneath South America Žcf. w4,5x.. 0.01, and Nb s 0.3. The relatively high distribution
coefficient for Yb presumes a garnet-bearing source.
4.3. Different extents of melting The relatively high Nb distribution coefficient pre-
sumes a phase such as amphibole or rutile remains in
The concentrations of incompatible trace elements the residuum. Nb concentrations as low as 1.1 ppm
in the lavas steadily increase away from the trench as in Atacazo andesites Žcf. nMORB concentrations of
does the LarYb ratio ŽFigs. 4 and 6.. As noted 3.5 ppm w29x. virtually require that a Nb-retentive
above, this steady increase is not likely due to assim- phase was present during melting, no matter the
ilation of crustal materials, as it is not accompanied source. The concentration of La and Ba in the slab-
by an systematic changes in 87 Srr 86 Sr, d 18 O, derived fluid is that calculated for Marianas back-arc
143
Ndr 144 Nd, and SiO 2 . The most likely explana- basin basalts w30x: Ba s 1370 ppm and La s 99 ppm.
tion for the increase in incompatible-element concen- Nb and Yb concentrations in the fluid are assigned to
trations and LarYb ratios away from the trench is be 0 and 10 ppm, respectively Žestimated from w31x..
164 R. Barragan et al.r Earth and Planetary Science Letters 154 (1998) 153–166

Fig. 9. Partial melting–slab fluid mixing model for BarNb and LarYb ratios in comparison to observed ratios in lavas; model parameters
are discussed in the text. Dashed curÕes connect melts produced by equal degrees of partial melting; solid curÕes connect melts with equal
proportions of slab-derived fluid.

Reasonably well-fitting models using the parame- species, undergoes different devolatilization reac-
ters above indicate that Atacazo’s magmas may re- tions, and different residual phases are stable w32x.
sult from an average of 15% melting, Antisana’s Nonetheless, the model demonstrates that these mag-
magmas from 3% melting and Sumaco’s magmas mas may be produced by decreasing inputs of slab-
from 2% melting ŽFig. 9.. Atacazo’s magmas have derived fluids, which in turn cause decreasing ex-
an average of 7% fluid component Ždiscounting the tents of melting.
single sample with extremely high BarNb., Anti- NbrTa ratios are in accord with the interpretation
sana’s magmas 2%, and Sumaco’s nil. These fluid that the principal control on the compositions of the
proportions are for the amount of slab-derived fluid Ecuadorian andesites are imparted by a slab compo-
in the melt; normalizing them by the extent of nent and different extents of melting of a depleted
melting gives the proportion of fluid introduced into mantle source. Plank and White w33x have demon-
the mantle wedge, which results in 1.1%, 0.06%, and strated that mid-ocean ridge basalts and depleted arc
nil for the three volcanoes. We emphasize that the lavas Žthose with low Nb concentrations. have
magnitude of these results depend strongly on the NbrTa ratios ranging from ; 6 to the chondritic
source composition, the bulk distribution coeffi- ratio of 17 " 1. The low ratios are attributed to
cients, and especially the fluid composition, which derivation of both arc and mid-ocean ridge magmas
are poorly constrained. For example, the composition from a source that had been previously depleted by
of the introduced fluid no doubt changes as the slab partial melting. In contrast, arc lavas from the Bis-
descends, becomes progressively depleted in mobile mark and Tongan arcs have NbrTa ratios ranging
 R. Barragan et al.r Earth and Planetary Science Letters 154 (1998) 153–166 165

from 17 in lavas relatively depleted in incompatible Žwhich varies from 308 to 808. along strike of the arc
elements to 33 in potassic lavas w34x. These higher w35x. The reasons why some continental arc magmas
ratios are attributed to enrichment of the mantle undergo large amounts of assimilation and others do
wedge by silicic melts that have left rutile in the not remain elusive, but clearly crustal thickness is
residuum, as opposed to enrichment by a volatile-rich not the only control.
phase. The lavas from the three Ecuadorian volca-
noes of this study have consistently increasing
NbrTa ratios with increasing Nb concentrations, Acknowledgements
which in turn relates to the distance from the trench.
NbrTa ranges from 6 to 15 at Atacazo, indicating Field work was funded by a seed grant from the
derivation from a depleted, MORB-like source w33x. University of Idaho Research Office. INAA analyses
In contrast, NbrTa ratios at Sumaco are all higher were conducted under the U.S. D.O.E.’s Reactor
than the chondritic ratio, ranging from 19 to 23, Sharing Program to Oregon State University. We
similar to other potassic arc lavas w34x. thank Scott Linneman for an informal review and
Sue Kay, Mark Feigenson, and anonymous for jour-
nal reviews. Thanks to Terry Plank for insight into
5. Conclusions: melt genesis and evolution the NbrTa issue and Karen Harpp for review.
Apologies to Jorge for the near destruction of his
Trace-element concentrations and ratios and iso- jeep. [CL]
topic ratios in lavas from a transect across the
Ecuadorian Andes indicate that assimilation does not References
play a fundamentally significant role in their genesis.
Instead, the characteristic aspects of the magmas are w1x W. Hildreth, S. Moorbath, Crustal contributions to arc mag-
principally controlled by the amount of fluid intro- matism in the Andes of central Chile, Contrib. Mineral.
duced into the mantle wedge by the subducted slab. Petrol. 98 Ž1988. 455–489.
w2x J.P. Davidson, N.J. McMillan, S. Moorbath, G. Worner, R.S.
The amount of fluid introduced in turn controls the Harmon, L. Lopez-Escobar, The Nevados de Payachata vol-
extent of melting. At the arc front, a relatively large canic region Ž188Sr698W, N. Chile., II. Evidence for
input of fluid causes large extents of melting. With widespread crustal involvement in Andean magmatism, Con-
successive devolatilization, decreasing amounts of trib. Mineral. Petrol. 105 Ž1990. 412–432.
w3x D.R. Tormey, F.A. Frey, L. Lopez-Escobar, Geochemistry of
fluid or magma are introduced into the mantle wedge
the active Azufre–Planchon–Peteroa volcanic complex, Chile
from the subducted lithosphere, and the extent of Ž35815X S.: evidence for multiple sources and processes in a
melting becomes smaller. Cordilleran arc magmatic system, J. Petrol. 36 Ž1995. 265–
These geochemical variations believed to link the 298.
amount of slab fluid to the extent of partial melting w4x R.L. Hickey, F.A. Frey, D.C. Gerlach, L. Lopez-Escobar,
are common elsewhere in the Andes and in other Multiple sources for basaltic arc rocks from the southern
volcanic zone of the Andes Ž348–418 S.: trace element and
continental arcs. Hickey and coworkers have de- isotopic evidence for contributions from subducted oceanic
scribed similar features in lavas in lavas of the crust, mantle and continental crust, J. Geophys. Res. 91
Southern Volcanic Zone of the Andes w4,5x. A simi- Ž1986. 5963–5983.
lar trend exists in a transect across the central Andes, w5x R. Hickey-Vargas, H.M. Roa, L.L. Escobar, F.A. Frey, Geo-
although those lavas have a much larger crustal chemical variations in Andean basaltic and silicic lavas from
the Villarrica–Lanin volcanic chain Ž39.58S.: an evaluation
contribution w26x. Ryan and others w31x find almost of source heterogeneity, fractional crystallization and crustal
exactly the same trends in a transect perpendicular to assimilation, Contrib. Mineral. Petrol. 103 Ž1989. 361–386.
the Kurile island arc, which are also explained by a w6x T. Feininger, Allochthonous terranes in the Andes of Ecuador
direct relation between the amount of slab-derived and northwestern Peru, Can. J. Earth Sci. 24 Ž1987. 266–278.
w7x H.R. Balkwill, G. Rodrigue, F.I. Paredes, J.P. Almeida.
fluid and the extent of partial melting. A similar
Northern part of Oriente basin, Ecuador: reflection seismic
relation between slab input and degree of melting expression of structures, in: A.J. Tankard, R. Suarez, H.J.
exists in the Central American arc, but the principal Welsink ŽEds.., Petroleum Basins of South America, Am.
control there is changing dip of the subduction zone Assoc. Pet. Geol. Mem. 62 Ž1995. 559–571.
166 R. Barragan et al.r Earth and Planetary Science Letters 154 (1998) 153–166

w8x J.A. Aspden, M. Litherland, The geology and Mesozoic w23x D. Geist, T. Naumann, P. Larson, Evolution of Galapagos ´
collisional history of the Cordillera Real, Ecuador, Tecton- magmas: mantle and crustal fractionation without assimila-
physics 205 Ž1992. 187–204. tion, J. Petrol. Ž1997, submitted..
w9x P.L. Larson, H.P. Taylor, 18 Or 16 O ratios in ash-flow tuffs w24x S.M. Kay, C.E. Gordillo, Pocho volcanic rocks and the
and lavas erupted from the central Nevada caldera complex melting of depleted continental lithosphere above a shallowly
and the central San Juan caldera complex, Colorado, Contrib. dipping subduction zone in the central Andes, Contrib. Min-
Mineral. Petrol. 92 Ž1986. 146–156. eral. Petrol. 117 Ž1994. 25–44.
w10x M.D. Kurz, T.C. Kenna, D.P. Kammer, J.M. Rhodes, M.O. w25x S.M. Kay, B. Coira, B. Viramonte, Young mafic back arc
Garcia, Isotopic evolution of Mauna Loa volcano: a view volcanic rocks as indicators of continental lithospheric de-
from the submarine southwest rift zone, in: J.M. Rhodes, J.P. lamination beneath the Argentina Puna plateau, central An-
Lockwood ŽEds.., Mauna Loa Revealed, Geophys. Monogr. des, J. Geophys. Res. 99 Ž1994. 24323–24339.
92 Ž1995. 289–306. w26x J.P. Davidson, S.L. de Silva, Late Cenozoic magmatism of
w11x M. Barazangi, B. Isacks, Spatial distribution of earthquakes the Bolivian altiplano, Contrib. Mineral. Petrol. 119 Ž1995.
and subduction of the Nazca plate beneath South America, 387–408.
Geology 4 Ž1976. 686–692. w27x S.R. Taylor, S.M. McLennan, The Continental Crust: Its
w12x M.L. Hall, C.A. Wood, Volcano-tectonic segmentation of the Composition and Evolution, Blackwell, Oxford, 1985, 312
northern Andes, Geology 13 Ž1985. 203–207. pp.
w13x P.J. Gossens, W.I. Rose, Chemical composition and age w28x M.R. Perfit, D.A. Gust, A.R. Bence, R.J. Arculus, S.R.
determination of tholeiitic rocks in the Basic igneous com- Taylor, Chemical characteristics of island arc basalts: impli-
plex, Ecuador, Geol. Soc. Am. Bull. 84 Ž1973. 1043–1052. cations for mantle sources, Chem. Geol. 30 Ž1980. 227–256.
w14x T. Feininger, M.K. Seguin, Simple Bouguer gravity anomaly w29x A.W. Hofmann, Chemical differentiation of the Earth: the
field and the inferred crustal structure of continental Ecuador, relationship between mantle, continental crust and oceanic
Geology 11 Ž1983. 40–44. crust, Earth Planet. Sci. Lett. 90 Ž1988. 297–314.
w15x R. Maruri, Estudio vulcanologico del complejo volcanico w30x E. Stolper, S. Newman, The role of water in the petrogenesis
Ninahuilca, Thesis, Escuela Politecnica Nacional, Quito, of Mariana trough magmas, Earth Planet. Sci. Lett. 121
1993, 140 pp. Žunpublished.. Ž1994. 293–325.
w16x A. Peccerillo, S.R. Taylor, Geochemistry of Eocenc calc-al- w31x Y. Tatsumi, D.L. Hamilton, R.W. Nesbitt, Chemical charac-
kaline volcanic rocks from the Kastamonu area, northern teristics of fluid phase released from a subducted lithosphere
Turkey, Contrib. Mineral. Petrol. 58 Ž1976. 63–81. and origin of arc magmas: evidence from high-pressure
w17x T. Wolf, Geografia y Geologia del Ecuador. Brockhaus, experiments and natural rocks, J. Volcanol. Geotherm. Res.
Leipzig, 1892. 29 Ž1986. 293–309.
w18x R.J. Colony, J.H. Sinclair, The lavas of the volcano Sumaco, w32x J.G. Ryan, J. Morris, F. Tera, W.P. Leeman, A. Tsvetkov,
eastern Ecuador, Am. J. Sci. 216 Ž1928. 299–312. Cross-arc geochemical variations in the Kurile arc as a
w19x T. Simkin, L. Siebert, Volcanoes of the World, 2nd ed., function of slab depth, Science 270 Ž1995. 625–627.
Geoscience Press, Tucson, AZ, 1994, 349 pp. w33x T. Plank, W.M. White, Nb and Ta in arc and mid-ocean
w20x W.M. White, P.J. Patchet, Hf–Nd–Sr isotopes and incompat- ridge basalts, Eos ŽTrans. Am. Geophys. Union. 76 Ž1995.
ible element abundances in island arcs: implications for F655.
magma origins and crust–mantle evolution, Earth Planet. Sci. w34x A.J. Stolz, K.P. Jochum, B. Spettel, A.W. Hofmann, Fluid-
Lett. 67 Ž1984. 167–185. and melt-related enrichment in the subarc mantle: evidence
w21x R.S. Harmon, J. Hoefs, Oxygen isotope heterogeneity of the from NbrTa variations in island-arc basalts, Geology 24
mantle deduced from global 18 O systematics of basalts from Ž1996. 587–590.
different geotectonic settings, Contrib. Mineral. Petrol. 120 w35x M.J. Carr, M.D. Feigenson, E.A. Bennett, Incompatible ele-
Ž1995. 95–114. ment and isotopic evidence for tectonic control of source
w22x D.E. James, L.A. Murcia, Crustal contamination in northern mixing and melt extraction along the Central American arc,
Andean volcanics, J. Geol. Soc. London 141 Ž1984. 823–830. Contrib. Mineral. Petrol. 105 Ž1990. 369–380.