Subduction cycling of volatiles and trace elements through the

ORIGINAL PAPER

Subduction cycling of volatiles and trace elements

through the Central American volcanic arc: evidence

from melt inclusions

Seth J. Sadofsky Æ Maxim Portnyagin Æ

Kaj Hoernle Æ Paul van den Bogaard

Received: 8 December 2006 / Accepted: 3 September 2007

 Springer-Verlag 2007

Abstract Compositions of melt inclusions in olivine

(Fo

90-64

) from 11 localities in Guatemala, Nicaragua and

Cost Rica along the Central American Volcanic Arc are

used to constrain combined systematics of major and trace

elements and volatile components (H

O, S, Cl, F) in

parental melts and to estimate volcanic ﬂuxes of volatile

elements. The melt inclusions cover the entire range of

compositions reported for whole rocks from Central

America. They point to large heterogeneity of magma

sources on local and regional scales, related to variable

contributions of diverse crustal (from the subducting and

overriding plates) and mantle (from the wedge and

incoming plate) components involved in magma genesis.

Water in parental melts correlates inversely with Ti, Y and

Na and positively with Ba/La and B/La (with the exception

of Irazu

Volcano), which indicates mantle melting ﬂuxed

by Ba-, B- and H

O-rich, possibly, serpentinite-derived

ﬂuid beneath most parts of the arc. Different components

with melt-like characteristics (high LREE, La/Nb and

probably also Cl, S and F and low Ba/La) control the

geochemical peculiarities of Guatemalan and Costa Rican

magmas. The composition of parental magmas together

with published data on volcanic volumes and total SO

ﬂux

from satellite measurements are used to constrain ﬂuxes of

volatile components and to estimate total magmatic ﬂux in

Central America. We found that volcanic ﬂux accounts for

only 13% of total magmatic and volatile ﬂuxes. The

remaining 87% of magmas remained in the lithosphere to

form cumulates (*39%) and intrusives (*48%). The

intrusive fraction of magmatic ﬂux may be signiﬁcantly

larger beneath Nicaragua compared to Costa Rica. Inter-

estingly, total ﬂuxes of magmas and volatiles in Central

America are quite similar to the global average estimates.

Introduction

Subduction zones provide the return ﬂux of water and other

volatiles from the surface of the earth to the mantle,

making our understanding of global volatile cycling

heavily reliant on understanding subduction zone magma-

tism. Composition of the crustal input, dip of the

subducting slab and volcanic output vary along the Central

American subduction zone (Fig. 1) (Carr et al. 1990, 2003;

Patino et al. 2000; Protti et al. 1995;Ru

pke et al. 2002).

These differences may have a major inﬂuence on elemen-

tal, and especially volatile, cycling through the subduction

system. In particular, it was proposed that the steeply

dipping oceanic plate subducting beneath Nicaragua was

highly hydrated, thus delivering larger amounts of water

and other volatiles into the mantle wedge beneath this

segment of the Central American Volcanic Arc (CAVA),

explaining the distinct geochemistry of the volcanic output

Communicated by J. Hoefs.

Electronic supplementary material The online version of this

article (doi:10.1007/s00410-007-0251-3) contains supplementary

material, which is available to authorized users.

S. J. Sadofsky  K. Hoernle  P. van den Bogaard

SFB 574, University of Kiel, Wischhofstr. 1-3,

Kiel 24148, Germany

M. Portnyagin (&)  K. Hoernle  P. van den Bogaard

Leibniz Institute for Marine Sciences (IFM-GEOMAR),

Wischhofstr. 1-3, Kiel 24148, Germany

e-mail: [email protected]

M. Portnyagin

Vernadsky Institute, Kosigin st. 19, 119991 Moscow, Russia

123

Contrib Mineral Petrol

DOI 10.1007/s00410-007-0251-3

(Abers et al. 2003; Ranero et al. 2003;Ru

pke et al. 2002).

The quantity of volatiles in the Central American magmas,

and the magmatic volatile ﬂuxes and their correlations with

major and trace element geochemistry and magmatic pro-

ductivity are however poorly understood but absolutely

necessary for understanding the magma generation pro-

cesses in the arc system.

Studies of melt inclusions in early-crystallizing pheno-

cryst phases provide direct insights into pre-eruptive water

(and other volatile) contents in magmas (Harris and

Anderson 1984; Portnyagin et al. 2007; Sobolev and

Chaussidon 1996; Wallace 2005). Previous studies of the

volatile content in maﬁc magmas from Central America

found high H

O concentrations in the volcanic front (up to

6% for Cerro Negro in Nicaragua) (Benjamin et al. 2007;

Roggensack et al. 1997, 2001b; Sisson and Layne 1993;

Wade et al. 2006; Walker et al. 2003) and substantially

lower contents in back-arc cinder cones in Guatemala

(Walker et al. 2003). Although the amount of available data

has continuously increased over the past decade, there has

been no attempt to present a regional picture of volatile

abundances in the Central America magmas in conjunction

with major and trace element systematics.

Here, we present new data on the composition of oliv-

ine-hosted melt inclusions from all major types of rocks in

Guatemala, Nicaragua and Costa Rica to address several

questions which are vital to our understanding of the

Central American Subduction Zone: (1) What are the

volatile (H

O, S, Cl, F) concentrations of primitive mag-

mas throughout Central America? (2) To what degree do

trace element contents and their ratios correlate with the

amount of H

O in magmas? (3) What are the volcanic

volatile ﬂuxes for the different arc segments, and how do

they correlate with subduction input?

Geologic setting

The geologic framework of volcanism in Central America

has been extensively discussed in the literature (e.g. Aubo-

uin et al. 1982; Carr et al. 1990, 2003; Leeman et al. 1994;

Protti et al. 1995). In brief, volcanism in Central America

results from the subduction of the Cocos Plate beneath the

Caribbean Plate (Fig. 1a). From Guatemala to northern

Costa Rica, *25 Ma old crust formed at the East Paciﬁc

Rise subducts at a rate of about 80 mm/year and at an angle

varying from 55 in Guatemala and El Salvador to 60–65

in Nicaragua and northern Costa Rica (Protti et al. 1995;

Syracuse and Abers 2006) (Fig. 1). The sediment input is

well-deﬁned from Deep Sea Drilling Project (DSDP) Site

495 (Fig. 1). It can be divided into two layers: an *200-m

thick layer of hemipelagic clay overlies an *250-m-thick

layer of carbonate oozes (Aubouin et al. 1982; Plank and

Langmuir 1998). It has been proposed that the mantle

section of the oceanic plate subducting beneath Nicaragua

contains a substantial amount of serpentinites (Abers et al.

2003; Ranero et al. 2003;Ru

pke et al. 2002). Beneath

central Costa Rica, *15–20 Ma oceanic crust, formed at

100

150

200

250

300

350

400

egdiR

nuomae

niv

Cocos Plate

DSDP

495

HONDURAS

NICARAGUA

GUATEMALA

RODAVLASLE

COSTA

acir

mAelddiM

hcne

Irazú

Cerro-Negro

Masaya

Nejapa

Santa Maria

Atitlán

Fuego

Agua

Pacaya

100

120

140

160

0 250 500 750 1000

Distance alon

arc in Km

Volume of volcanic centers in Km

Crustal thickness in Km

(a)

(b)

(c)

MatnaS

ayacaP

aci

leT

orge

orreC-sal

saL

jeN-euqeyopAa

ayasaM

adanarG-ohcabmo

lane

úzarI

EL SALVADOR

NICARAGUA COSTA RICA

GUATEMALA

Telica

Depth to slab in Km

Carribean Plate

RICA

Mombacho-Granada

Arenal

Fig. 1 Geologic framework of the Central American arc magmatism.

a Schematic map showing the Central American volcanic arc, formed

in response to the subduction of the Cocos Plate beneath the

Caribbean Plate. The Cocos Ridge and Seamount Province belong to

the Galapagos hotspot track subducting beneath southern Costa Rica.

Triangles illustrate locations of frontal volcanoes with localities of

samples selected for this melt inclusion study denoted by ﬁlled

triangles. Thin lines are country boundaries. b Volumes of volcanic

centers along the volcanic front in Central America (Carr et al. 2003).

c Variations in crustal thickness (Carr et al. 2003) and depths to

subducting plate (Syracuse and Abers 2006) along the volcanic front

in the Central America. Here and in the following plots the distance

was measured from the Mexican-Guatemalan border

Contrib Mineral Petrol

123

the Gala

pagos Spreading Center and overprinted by the

Gala

pagos Hotspot (Hoernle et al. 2000; Werner et al. 1999,

2003), subducts at a relatively shallow angle of *50.

Crustal thickness decreases from 50 to \35 km from Gua-

temala to Nicaragua and increases up to *45 km beneath

Costa Rica, with the transition from continental in Guate-

mala to oceanic character of the basement beneath the

southern part of the volcanic belt (Carr et al. 1990).

Combined variations in subduction dip, crustal thickness

and off-set of volcanic front from the trench result in large

variations in the length of mantle columns beneath

frontal volcanoes in Central America, which increases

from *40 km in Guatemala to 85–110 km in Nicaragua

and sharply decreases to 45–75 km beneath Costa Rica

(Fig. 1c).

Methods

Fresh samples of young (primarily historic) maﬁc tephra

were collected from active volcanic centers in Guatemala,

Nicaragua and Costa Rica. Hand-picked olivine pheno-

crysts (0.5–1 mm) were mounted in epoxy, polished and

analyzed optically to locate melt inclusions. Inclusions

selected for chemical analyses were glassy and usually

contained homogeneous glass (± shrinkage bubble). A

small number of inclusions from all regions contained

sulﬁde globules and/or spinel crystals.

Major elements, Cl and S of glass inclusions and major

element composition of olivine were analyzed by a

CAMECA SX50 electron microprobe at IFM-GEOMAR.

Major elements in olivine were analyzed with a 15-kilovolt

(kV), 30-nanoampere (nA) beam 1.2 lm in diameter with a

20-s count time per element. Major elements in glass

inclusions were analyzed with 15-kv, 10-nA beam, 5 lmin

diameter with a 20-s count time per element. Chlorine and

S were analyzed with 15-kV, 30-nA beam, 5 lmin

diameter with a 60-s count time per element. Sulfur was

measured at S

wavelength, and thus can be underesti-

mated up to maximum 30 relative % due to possible

presence of reduced S species in the glasses (Carroll and

Rutherford 1988). Calibration was performed using pure

oxides from CAMECA and natural reference materials

(Jarosewich et al. 1980). San Carlos olivine (USNM

111312/444), basaltic glass VG-2 (USNM 111240/52) and

scapolite (USNM R6600-1) were used as monitors during

routine measurements. Values reported here are averages

of at least three points per glass inclusion. Trace elements,

ﬂuorine and hydrogen concentrations were determined

using a CAMECA ims4f ion microprobe at the Institute of

Microelectronics and Informatics (Yaroslavl, Russia).

Trace elements were determined with a primary beam of

2–

ions 20–30 lm in diameter at 14.5 kV and 15–20 nA

current. Each analysis is based on ﬁve cycles of measure-

ment with count times varying to provide appropriate

counting statistics. Conversion of measured ion intensities

to concentrations followed standard procedures (Portnya-

gin et al. 2002; Sobolev and Chaussidon 1996). Electron

microprobe and ion microprobe data agree within esti-

mated error (10% relative to concentration) as shown by

comparison of Ti concentrations.

Any Fe–Mg disequilibrium between melt inclusions and

host olivines is assumed to be due to post-entrapment

crystallization of olivine, and is corrected by simulating

incremental addition of equilibrium olivine to the melt

inclusion composition until achieving equilibria with the

host (Danyushevsky et al. 2002). The amount of olivine

added was within 3–5 mol% in most cases. The following

results and discussion refer to the compositions corrected

for post-entrapment crystallization of olivine on the

inclusion walls.

Results

Major elements

We present melt inclusion data of olivine phenocrysts from

11 sites in Guatemala, Nicaragua and Costa Rica. These

include Santa Maria, Fuego, Agua and Atitla

n Volcanoes

in Guatemala; Cerro Negro and Masaya Volcanoes, a small

tuff ring near Telica Volcano and cinder cones of the Ne-

japa and Granada lineaments in Nicaragua; and Irazu

and

Arenal Volcanoes in Costa Rica (Fig. 1). Major and trace

element compositions of whole-rock samples selected for

this melt inclusion study are given in the electronic

supplement.

Melt-inclusion-bearing olivine phenocrysts in studied

samples have forsterite content (100 Mg/(Mg + Fe),

mol%) from 64 to 90 (Table 1; Fig. 2). The most magne-

sian olivines were found at Irazu

(up to Fo

) and Granada

(up to Fo

) samples. Typical olivine range in other sam-

ples is Fo

82–70

, but as low as Fo

at Atitla

n. The olivine

phenocryst compositions exhibit a general trend of

decreasing maximum Fo from Costa Rica to Guatemala.

Studied melt inclusions are largely basaltic to basaltic

andesitic in composition, similar to the most maﬁc whole-

rock samples from Central America. Andesitic inclusions

were found in evolved olivines (Fo

70–64

) from Atitla

n and

Irazu

Volcanoes. Major element compositions of melt

inclusions (Table 1, Fig. 2) are generally similar to whole-

rock compositions from the same areas and exhibit clear

regional variations, which are particularly evident from

comparison of a statistically signiﬁcant number of inclu-

sions from Guatemala and Nicaragua trapped in olivines

with similar (moderately magnesian) compositions.

Contrib Mineral Petrol

123

Table 1 Representative compositions of melt inclusions in olivine from the Central American volcanic arc

Volcano Santa

Maria

Fuego Atitla

n Agua Telica Cerro

Negro

Cerro

Negro

Cerro

Negro

Nejapa Nejapa Masaya Granada Granada Granada Arenal Irazu

Irazu

Segment GU GU GU GU NWN NWN NWN NWN SEN SEN SEN SEN SEN SEN CR CR CR CR CR

Rock sample GU-19d GU-3a GU-25b GU-11d P2-16 P2-3a P2-3a P2-3a P2-32d P2-32d P2-47 P2-58 P2-58 P2-58 CR-61C P2-72 P2-72 P2-72 P2-72

Inclusion# s1 1 s1 s7 9 1-10 1-35 2-46a 4a 6a 3a 27a 8 2-69 57 40a 40b 3-1 3-2

Melt inclusions

SiO

(wt%) EMP 53.07 56.78 54.97 50.90 47.00 48.10 49.44 49.53 47.26 48.74 53.71 52.20 47.57 47.98 50.94 60.43 57.62 52.33 52.05

TiO

(wt%) EMP 1.03 1.17 1.57 1.14 0.78 0.77 0.59 0.72 2.63 1.13 1.25 1.16 0.89 1.08 0.79 1.28 1.55 0.89 0.98

(wt%) EMP 19.84 18.00 17.84 19.84 16.33 18.55 18.96 18.06 17.47 17.34 15.72 16.42 18.62 16.68 19.49 16.15 15.90 17.96 18.18

FeO* (wt%) EMP 8.11 7.63 8.69 7.88 11.84 10.73 10.05 9.88 11.28 10.34 11.66 10.55 9.49 10.00 9.60 6.96 8.44 7.13 7.77

MnO (wt%) EMP 0.10 0.15 0.12 0.07 0.25 0.18 0.24 0.17 0.22 0.19 0.23 0.16 0.17 0.18 0.16 0.12 0.11 0.15 0.17

MgO (wt%) EMP 2.76 3.39 3.35 5.08 5.71 6.68 6.09 6.66 6.61 6.88 3.92 6.42 7.59 7.61 4.14 2.33 2.88 5.93 4.77

CaO (wt%) EMP 10.48 7.30 8.38 10.20 14.06 12.74 12.31 12.65 11.19 12.89 8.80 9.84 13.18 14.08 11.65 3.63 4.39 11.65 11.75

O (wt%) EMP 3.50 3.97 3.80 3.63 3.05 1.83 1.89 1.87 2.74 2.21 2.91 2.44 2.05 2.03 2.46 4.05 3.93 2.79 2.81

O (wt%) EMP 0.88 1.40 1.00 0.98 0.71 0.27 0.27 0.30 0.36 0.13 1.49 0.61 0.30 0.19 0.56 4.23 3.78 0.87 1.22

(wt%) EMP 0.23 0.21 0.28 0.28 0.28 0.13 0.17 0.17 0.26 0.15 0.31 0.21 0.15 0.17 0.21 0.82 1.39 0.30 0.29

Total (wt%) 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00

Original total

(wt%)

93.07 93.84 94.37 93.71 96.33 92.77 94.92 92.87 96.53 97.47 98.57 94.62 96.41 96.01 96.03 98.64 98.94 95.59 96.37

O (wt%) SIMS 3.5 3.8 3.3 3.4 3.0 5.2 4.8 4.9 1.9 2.1 1.7 5.0 3.7 2.6 1.6 1.0 1.0 5.2 3.0

S (ppm) EMP 2278 1064 1388 1913 2408 1231 1031 1255 1482 1164 385 953 2071 1253 1625 282 345 1908 1461

Cl (ppm) EMP 896 1017 1017 875 1628 960 689 749 298 264 1242 1111 889 329 1696 1455 2275 766 817

F (ppm) SIMS 519 419 414 393 387 126 182 122 291 100 597 186 203 215 350 2143 3159 611 712

Li (ppm) SIMS 7.9 12.9 10.7 6.5 4.5 3.6 3.7 3.6 3.7 3.3 11.0 5.9 4.1 4.7 5.7 18.9 20.2 5.0 14.2

Be (ppm) SIMS 0.78 0.99 1.04 0.71 0.54 0.27 0.29 0.25 0.50 0.31 0.94 0.36 0.41 0.38 0.79 2.96 3.59 0.76 0.98

B (ppm) SIMS 16.3 23.1 17.2 10.7 21.3 10.2 15.4 16.3 2.3 0.6 25.7 17.0 6.2 1.1 10.0 25.3 10.8 5.1 6.1

K (ppm) SIMS 7332 11979 8603 7402 7172 2205 2248 2408 3817 1069 15274 5581 2632 1476 4482 33811 40620 7670 9393

Ti (ppm) SIMS 5171 6107 8495 6215 4688 4229 3938 3897 17092 8169 8369 6928 5339 6986 4886 6538 9598 5286 5843

V (ppm) SIMS 291 237 232 252 299 307 294 278 591 308 413 275 331 294 234 174 237 274 286

Cr (ppm) SIMS 159 28 36 96 41 67 76 100 108 94 45 33 107 163 27 82 23 91 78

Sr (ppm) SIMS 500 457 411 551 701 397 434 406 476 311 416 374 430 367 870 768 2652 792 826

Zr (ppm) SIMS 76 102 101 95 46 18 21 20 98 47 113 38 31 51 51 317 431 60 71

Y (ppm) SIMS 15 18 19 15 28 12 12 12 25 22 29 19 16 19 18 29 48 14 16

Nb (ppm) SIMS 3.5 4.1 3.8 3.8 1.2 0.8 1.8 1.8 14.6 2.9 4.6 1.9 2.1 5.1 3.8 52 134 6.1 7.6

Ba (ppm) SIMS 410 580 460 457 680 253 239 245 219 52 1140 493 206 79 529 1365 1895 467 531

La (ppm) SIMS 9.0 11.1 10.3 10.9 7.8 2.1 2.6 2.6 6.7 2.8 14.4 5.4 3.4 4.0 14.4 75.9 144 17.7 22.1

Ce (ppm) SIMS 19.5 24.1 23.8 23.1 16.4 4.9 5.8 5.6 17.8 7.8 27 9.9 8.1 11.0 27.5 158 318 36.2 45.4

Nd (ppm) SIMS 11.1 13.0 15.1 13.9 14.3 4.4 4.6 4.4 14.0 7.1 19 7.8 6.6 8.8 15.9 70 125 17.7 22.1

Sm (ppm) SIMS 3.2 3.4 4.0 3.7 4.4 1.5 1.6 1.6 4.0 2.7 4.9 2.3 2.0 2.9 3.4 12.2 21.1 3.9 4.5

Eu (ppm) SIMS 2.1 1.2 1.1 1.0 1.9 0.7 0.5 0.7 1.6 1.1 2.5 1.4 0.8 1.1 1.3 2.8 3.7 1.2 1.5

Gd (ppm) SIMS 1.5 3.3 3.5 2.3 5.1 2.3 2.1 1.6 3.9 3.5 4.8 2.9 2.6 3.1 4.1 9.9 15.1 3.3 4.1

Contrib Mineral Petrol

123

Table 1 continued

Volcano Santa

Maria

Fuego Atitla

n Agua Telica Cerro

Negro

Cerro

Negro

Cerro

Negro

Nejapa Nejapa Masaya Granada Granada Granada Arenal Irazu

Irazu

Segment GU GU GU GU NWN NWN NWN NWN SEN SEN SEN SEN SEN SEN CR CR CR CR CR

Rock sample GU-19d GU-3a GU-25b GU-11d P2-16 P2-3a P2-3a P2-3a P2-32d P2-32d P2-47 P2-58 P2-58 P2-58 CR-61C P2-72 P2-72 P2-72 P2-72

Inclusion# s1 1 s1 s7 9 1-10 1-35 2-46a 4a 6a 3a 27a 8 2-69 57 40a 40b 3-1 3-2

Dy (ppm) SIMS 2.5 2.9 3.3 2.8 5.1 2.0 2.2 2.0 4.1 3.5 4.9 2.9 2.7 3.1 3.4 6.7 10.5 2.5 3.1

Er (ppm) SIMS 1.7 1.9 1.6 1.4 2.9 1.3 1.3 1.3 2.8 2.4 2.9 2.0 1.8 2.1 1.9 3.5 5.5 1.6 1.7

Yb (ppm) SIMS 1.8 1.7 1.9 1.7 2.9 1.2 1.3 1.2 2.6 2.1 2.9 1.8 1.6 2.0 1.8 2.6 4.1 1.4 1.6

Hf (ppm) SIMS 2.0 2.5 2.8 2.6 2.2 0.9 0.9 0.8 2.9 1.7 3.5 1.5 1.1 2.0 2.0 7.5 9.9 1.6 2.0

Th (ppm) SIMS 0.70 1.5 1.1 0.7 0.5 0.10 0.12 0.14 0.32 0.09 1.7 0.39 0.14 0.16 1.06 14.9 10.2 1.6 2.6

U (ppm) SIMS 0.22 0.76 0.5 0.3 0.5 0.13 0.09 0.11 0.20 0.03 1.6 0.39 0.15 0.09 0.43 6.3 3.6 0.63 0.80

Pb (ppm) SIMS 3.0 5.0 4.5 4.4 3.5 0.80 1.14 1.23 1.01 0.32 6.6 2.0 1.2 0.8 3.0 8.5 15.6 1.8 3.1

Host olivine

Fo (mol%) 79.0 75.3 75.4 81.1 80.2 81.1 81.9 82.6 82.3 83.0 72.7 78.8 85.7 86.3 77.0 71.4 71.1 89.0 88.7

SiO

(wt%) EMP 38.86 37.93 37.66 38.38 38.92 38.62 39.62 39.46 39.87 39.88 38.37 38.76 40.18 40.11 37.39 37.94 36.92 40.17 40.34

FeO (wt%) EMP 19.62 22.52 22.39 17.57 18.67 17.76 17.06 16.52 16.81 16.21 24.71 19.65 13.77 13.20 21.22 25.69 25.97 10.71 11.01

MnO (wt%) EMP 0.31 0.38 0.32 0.25 0.34 0.30 0.29 0.28 0.28 0.25 0.49 0.36 0.23 0.22 0.37 0.50 0.48 0.17 0.17

MgO (wt%) EMP 41.29 38.61 38.50 42.27 42.34 42.65 43.37 43.92 43.71 44.26 36.88 40.98 46.44 46.66 39.95 35.99 35.83 48.55 48.65

CaO (wt%) EMP 0.14 0.13 0.14 0.13 0.23 0.18 0.18 0.19 0.22 0.25 0.25 0.18 0.23 0.28 0.11 0.15 0.16 0.13 0.14

NiO (wt%) EMP 0.06 0.07 0.08 0.11 0.07 0.09 0.10 0.11 0.15 0.19 0.08 0.06 0.19 0.21 0.11 0.11 0.11 0.39 0.35

(wt%) EMP 0.03 0.03 0.03 0.03 0.02 0.01 0.03 0.02 0.03 0.05 0.02 0.02 0.05 0.05 0.02 0.02 0.02 0.05 0.05

Total (wt%) 100.31 99.67 99.11 98.74 100.59 99.62 100.64 100.49 101.07 101.08 100.80 100.02 101.09 100.73 99.17 100.41 99.50 100.17 100.72

Notes. Major elements and S and Cl in melt inclusions and host olivines were determined by electron microprobe (EMP), trace elements by secondary ion mass spectrometry (SIMS). Complete

data set of host rock compositions and studied melt inclusions is available on-line

Abbreviations for segments: GU Guatemala, NWN northwestern Nicaragua, SEN southeastern Nicaragua, CR Costa Rica

Contrib Mineral Petrol

123

Inclusions from Guatemala have distinctively higher SiO

O and lower FeO, MgO and CaO than inclusions from

Nicaragua. Nicaragua inclusions generally have the lowest

SiO

and Na

O but highest FeO and CaO. Variations in

TiO

concentrations along the volcanic front are not sys-

tematic. Inclusions in primitive olivines from Costa Rican

volcanoes appear to have intermediate major element

compositions between Nicaragua and Guatemala. Inclu-

sions from Nicaragua and Arenal Volcano in Costa Rica

have the lowest K

O, with exception of evolved inclusions

from Masaya, all below 1 wt%. Melts from Guatemala and

Irazu

volcano are more enriched, up to 3–5 wt% K

Oin

evolved andesitic melts.

Trace elements

Trace element concentrations in the melt inclusions are

highly variable on regional and local scales and generally

fall within the range of reported whole-rock compositions

(Table 1; Figs. 3, 4). In brief, melts from NW Nicaragua,

particularly from Cerro Negro Volcano, have the lowest

light rare earth element (LREE) and Th (\0.5 ppm)

abundances and low La/Yb ratios (\3.5), but the highest

ratios of ﬂuid-mobile elements relative to REE and Th (Ba/

La, Ba/Th, B/La, U/Th, Pb/Ce and Sr/Ce). Melts from

Masaya Volcano in SE Nicaragua and Guatemalan volca-

noes are systematically more enriched in LREE (La/

Yb [ 4), HFSE (Nb, Zr, Hf) and Th and have less pro-

nounced slab-ﬂuid signals (lower Ba/La, U/Th and B/La)

compared to Cerro Negro Volcano. Primitive melts from

Costa Rica, especially from Irazu

Volcano, have distinct

compositions from Guatemalan and Nicaraguan primitive

melts and are distinctively enriched in nearly all highly

incompatible trace elements (La/Yb [ 10, Nb/Y [ 0.4)

and exhibit low Ba/La and B/La, approaching compositions

of oceanic basalts (Figs. 3, 4). Melt inclusions from Arenal

are broadly intermediate in composition between those

from Nicaragua and those from Guatemala and Irazu

Volcano.

OiT

OaC

OeF

OiS

%lo

m,e

iviloo

0.5

1.5

2.5

200

400

600

800

1000

Distance alon

arc in km

GUATEMALA

NICARAGUA

COSTA

RICA

200

400

600

800

1000

Distance along arc in km

GUATEMALA

NICARAGUA

COSTA RICA

GUATEMALA

NICARAGUA

COSTA RICA

MI from this study

MI from literature

Rocks Mg#>0.5

Fig. 2 Along arc variations of

major element composition of

melt inclusions and their host

olivines. Filled symbols denote

data from this study, open

symbols data from olivine-

hosted melt inclusions from the

literature (Benjamin et al. 2007;

Harris and Anderson 1984;

Roggensack 2001a, b;

Roggensack et al. 1997; Sisson

and Layne 1993; Wade et al.

2006; Walker et al. 2003) and

small gray diamonds

compositions of primitive rocks

with Mg# [ 0.5 (Carr et al.

2003). Inclusions were

recalculated to equilibrium with

their host olivine using Petrolog

2.0 software (Danyushevsky

et al. 2002), assuming oxygen

fugacity at Ni–NiO buffer

Contrib Mineral Petrol

123

Melt inclusions from Nicaragua show strong composi-

tional variability over short distances along the volcanic

front and also on the scale of single samples. Cerro Negro

Volcano displays the greatest slab signal (e.g. Ba/La) along

the entire Central American Volcanic Arc. This signal is

however very weak just 60 km away from Cerro Negro in

melts from Nejapa lineament and also in the Granada area

further southeast. These melts have very low B/La \ 1, Ba/

La \ 40 and in this respect are similar to melts from Irazu

Volcano (Fig. 4). Irazu

melts, however, have greater LREE

enrichment than the melts from Nejapa and Granada, which

display smooth EMORB-like patterns of incompatible

elements with the exception of pronounced enrichments in

Ba, Pb and Sr (Fig. 3). A distinctive feature of these melts

is relative enrichment in Nb over other incompatible ele-

ments (Nb/La up to 2). These Nb-rich melts also have

extraordinarily high Nb/U ratios of up to 100 (Fig. 4).

Substantial compositional heterogeneity among Nicara-

guan magmas is also evident from the large scatter of trace

element concentrations in melt inclusions from single

samples. Up to one order of magnitude variations in trace

element concentrations and ratios of highly incompatible

elements are observed for inclusions from Cerro Negro,

Nejapa and Granada (Figs. 3, 4). The Granada melt

inclusions are particularly interesting in that at least two

contrasting groups of melts were sampled. Inclusions of

100

1000

BaTh U Nb B K LaBePbCe SrNd ZrSmEu Ti Dy Li Y Yb

MMD/elpmaS

Agua

Fuego

Santa Maria

100

1000

BaTh U Nb B K LaBePbCe Sr Nd ZrSmEu Ti Dy Li Y Yb

MMD/elpmaS

Telica

Cerro Negro 1971

Cerro Negro 1992

Cerro Negro 1999

100

1000

BaTh UNb B K LaBePbCeSrNdZrSmEu TiDyLi Y Yb

MMD/elpmaS

Nejapa

Mas aya

100

1000

BaTh U Nb B K LaBePbCe Sr Nd ZrSmEu Ti Dy Li Y Yb

/elpmaS

Granada (low-Nb)

Granada (high-Nb)

100

1000

BaTh UNb B K LaBePbCeSrNdZrSmEu TiD

Li Y Yb

MMD/

elpmaS

Arenal

Guatemala

NW Nicaragua

SE Nicaragua

Costa Rica

Atitlan

Irazu (low-Si)

′

Fig. 3 Incompatible trace elements in melt inclusions normalized to the composition of depleted MORB mantle (DMM; Salters and Stracke

2004). Shaded ﬁeld illustrates the entire compositional range of the melt inclusions from Central America

Contrib Mineral Petrol

123

one group (Cerro-Negro-like) in relatively low-Fo olivines

(Fo

) have typical arc-type patterns of trace elements with

low Nb and Zr concentrations and strong enrichment in

ﬂuid-mobile trace elements (K, B, U, Pb and also H

O).

Melts of the other group (Nejapa-like), which occur in

more Fo-rich olivines (up to Fo

), have much lower

concentrations of ﬂuid-mobile trace elements but higher

Nb, Zr and Hf. These two distinctive groups of inclusions

from Granada are henceforth referred to as low-Nb (LNB)

and high-Nb (HNB) groups.

Volatiles

Water concentrations range from \0.5 to *5 wt%, which

is within the range of previously reported data (Benjamin

et al. 2007; Roggensack 2001a, b; Roggensack et al. 1997;

Sisson and Layne 1993; Wade et al. 2006; Walker et al.

2003), but extends the range to more Fo-rich olivines

(Table 1, Fig. 5). Many melt inclusions from olivine

phenocrysts with Fo [ 80 are relatively high in water (2–

5 wt%), whereas most of those from more evolved olivine

phenocrysts (Fo \ 80) have \2 wt% H

O and therefore

have likely experienced degassing before trapping (Fig. 5).

When inclusions in olivine with Fo [ 80 are compared, the

highest H

O concentrations ([4 wt%) were found in the

1971 A.D. eruption of Cerro Negro (similar to those pre-

viously reported at Cerro Negro; Roggensack 2001a;

Roggensack et al. 1997), in low-Nb type inclusions from

Granada and in a single inclusion in high-Fo olivine from

the 1963 A.D. eruption of Irazu

. Inclusions from Guate-

mala and high-Nb inclusions from Granada have

intermediate H

O concentrations (2–4 wt%). Inclusions

from Nejapa, all of which have high Nb, have the lowest

O contents (£2 wt%). Maximum H

O/K

O ratios in the

primitive inclusions are highest in Nicaragua (H

O/K

Oup

to *25, excluding 1 point) and substantially lower in

Guatemala and Costa Rica.

Maximum sulfur concentrations in the melt inclusions

from all segments are 2,000–2,500 ppm. Higher

200

400 600

800

1000

Distance alon

arcinkm

200

400 600

800

1000

aL/aB

hT/U

aL/B

/rS

bY/aL

aL/bN

/bN

MORB+OIB

OIB

MORB

OIB

0.5

1.5

2.5

100

0.01

0.1

MORB

100

120

140

160

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

100

Fig. 4 Characteristic ratios of

incompatible trace elements in

melt inclusions plotted versus

distance along the arc. Filled

symbols denote data from this

study, small gray diamonds

denote compositions of

primitive to moderately evolved

whole rocks with SiO

\ 54

wt%. Horizontal lines illustrate

typical trace element ratios in

mid-ocean-ridge basalts

(MORB) and ocean-island

basalts (OIB) (Chaussidon and

Jambon 1994; Hofmann 1988;

Sun and McDonough 1989)

Contrib Mineral Petrol

123

concentrations (up to 4,000 ppm) were reported for Costa

Rican volcanoes (Benjamin et al. 2007; Wade et al. 2006).

Sulfur concentrations in melt inclusions tend to decrease

with decreasing Fo of olivine and increasing SiO

and K

of the inclusions, which suggests magmatic sulfur degas-

sing occurred in conjunction with crystal fractionation

(Benjamin et al. 2007; Sisson and Layne 1993; Wade et al.

2006). Thus, the measured concentrations should be con-

sidered as minimum estimates for parental magmas, which

may be even more sulfur-rich than the melt inclusions

trapped in variably evolved olivines. Ratios of S/K

exhibit clear regional variations with the highest of up

to *1 in Nicaragua and decreasing down to 0.2–0.4 in

Guatemala and Costa Rica, which is apparently due to

large regional variations in K

O content.

Chlorine concentrations in the melt inclusions range

from 300 ppm (Nicaragua) up to 2,500–3,500 ppm (in low-

Fo olivines from Costa Rican volcanoes) and exhibit a

general increase with decreasing Fo of the olivine (Figs. 5,

6). There are also good positive correlations between Cl

content and incompatible element abundances (e.g. K

in suites of co-genetic inclusions indicating no or only

weak Cl degassing during the early stages of fractionation

of Central American magmas (Benjamin et al. 2007; Sisson

and Layne 1993; Wade et al. 2006). Cl/K

O ratios mea-

sured in inclusions hosted by primitive olivines can

therefore be informative of their parental melts. Distinc-

tively high Cl/K

O (0.2–0.4) is typical for inclusions from

Nicaragua (except degassed inclusions in evolved olivines

from Masaya) and also for inclusions from Arenal Volcano

in Costa Rica (Cl/K

O up to 0.8; Wade et al. 2006).

Potassium-rich inclusions from Guatemala and Irazu

Vol-

cano have relatively low Cl/K

O (0.07–0.15) although their

absolute Cl concentrations are higher than that in

Nicaragua.

The majority of melt inclusions from the Central

America have ﬂuorine concentrations 100–800 ppm with

up to *3,000 ppm in andesitic inclusions in low-Fo oli-

vines from Irazu

(Figs. 5, 6). Fluorine concentrations

increase with decreasing Fo number of the host olivine and

strongly correlate with a number of incompatible elements

(e.g. K

O, Be, La). Melts from Nicaragua generally have

the lowest F concentrations, melts from Guatemala and

Arenal Volcano are moderately enriched, and melts from

Irazu

have the highest F concentrations at a given olivine

composition. The highest F/K

O ratios were found in low-

O melts from Nejapa and Granada.

Parental melts

For comparison of volatile and incompatible element

concentrations of melt inclusions trapped in olivine at

varying degrees of magmatic fractionation and for assess-

ing the possible compositions of parental, mantle-derived

melts, we corrected the data for passive enrichment during

fractional crystallization by adding olivine until equili-

brium was reached with mantle olivine of Fo

(Kelley

et al. 2006; Portnyagin et al. 2007). For this calculation, we

assume that the only changes to the melt compositions

t %wO

t %wS

t %wlC

t %

Host olivin

e Fo mol %

0.00

0.10

0.20

0.30

0.40

0.50

0.0

1.0

2.0

3.0

4.0

5.0

6.0

7.0

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

70 75 80 85 90

Santa Maria Fuego

Atitlan Agua

Χερρο Νεγρο Nejapa

Masaya Granada

Arenal Irazu

′

Fig. 5 Concentrations of volatile components in melt inclusions

plotted versus composition of host olivines. Small gray symbols are

previously published data on volatiles in olivine-hosted inclusions

from Central America (Benjamin et al. 2007; Harris and Anderson

1984; Roggensack 2001a, b; Roggensack et al. 1997; Sisson and

Layne 1993; Wade et al. 2006; Walker et al. 2003). Inclusions in low

Fo olivine (Fo \ 70 mol%) are not shown

Contrib Mineral Petrol

123

result from fractional crystallization of olivine. Since

pyroxene and probably plagioclase were most likely also

liquidus phases of melts in olivines with Fo \ 80, some

major element contents (e.g. Ca, Al, Fe) cannot be accu-

rately estimated in the parental melts with this method.

Incompatible trace elements however are less sensitive to

the modal crystallizing assemblage, and concentrations of

trace elements normalized to Fo

provide a better

assessment of parental melt compositions than uncorrected

melt inclusions or compositions corrected to certain con-

centrations of MgO or SiO

(see Portnyagin et al. 2007 for

details of uncertainties).

The applied normalization to olivine with Fo

straightforward for incompatible elements. Their concen-

trations decrease proportionally to the amount of added

olivine, which does not contain appreciable quantities of

incompatible elements. Concentrations of selected major

and trace elements in the estimated parental melts are

shown in Table 2 (see online supplement for complete data

set). For this calculation, we used the inclusions trapped in

the most magnesian olivines for each sample (indicated in

Table 2). For Granada, where we distinguish two geo-

chemically different groups of melts, we calculated the

parental melts for each group separately.

Behavior of volatile components is more complex dur-

ing magmatic evolution, because a signiﬁcant amount can

be lost through degassing from magmas during fraction-

ation, as demonstrated previously for Fuego, Cerro Negro

and Arenal Volcanoes (Benjamin et al. 2007; Roggensack

et al. 1997; Sisson and Layne 1993; Wade et al. 2006), and

is also evident from our data. Our goal was therefore to

select the least degassed inclusions from the entire data set

in order to get the most reliable estimates for parental

magmas.

The applied approach is illustrated in Fig. 7 using the

extensive data set available for Cerro Negro inclusions

(this study and Roggensack 2001a; Roggensack et al.

1997). First, we assumed that parental melt for the melt

inclusions trapped in olivine from one volcano evolved

exclusively due to crystal fractionation and degassing. If

O wt %

S wt%

Cl wt %

F wt %

O/O

O/lC

O/F

Distance alon

arc in km

200

400

600

800

1000

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.0

0.2

0.4

0.6

0.8

1.0

1.2

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.0

0.2

0.4

0.6

0.8

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.00

0.05

0.10

0.15

200

400

600

800

1000

Fig. 6 Along-arc variations in

concentrations of volatile

components in melt inclusions

and volatile/K

O ratios. Filled

symbols are data of this study,

open symbols denote data from

the literature (see Fig. 5 for

sources)

Contrib Mineral Petrol

123

only crystal fractionation has occurred, ratios of volatile

components to incompatible lithophile elements (e.g. K

should remain constant. If degassing has taken place, these

ratios will decrease due to partitioning of the volatile

component into the ﬂuid phase. In the Cerro Negro melt

inclusions, H

O, S and CO

decrease with increasing K

exhibiting clear degassing trends. Chlorine concentrations

increase with increasing K

O but Cl/K

O decreases, sug-

gesting partitioning of some amount of Cl into ﬂuid phase.

For these volatiles, the maximum measured ratios of vola-

tiles to K

O(H

O/K

O, S/H

O, Cl/K

O and CO

should be closest to the initial ratio in the parental melt.

Fluorine concentrations in melts increase proportionally to

O concentrations suggesting little or no degassing of

ﬂuorine during fractionation. Therefore, mean F/K

O gives

us the best idea of the mean ratios in the parental melt. If

the volatile to K

O ratios in the parental melts is known or

can be assumed, it is possible to estimate the absolute

concentrations of volatiles from independently estimated

O contents in the parental melt.

Trends similar to those in Fig. 7 were also evident for

Fuego Volcano in Guatemala (Sisson and Layne 1993) and

Arenal (Wade et al. 2006) and Irazu

(Benjamin et al. 2007)

Volcanoes in Costa Rica and appear to be typical for

Central America. Therefore, we can extend this approach to

other, more limited data sets of melt inclusion composi-

tions. We have estimated the maximum H

O/K

O, S/K

Cl/K

O and mean F/H

O for all inclusions from every

volcano and then calculated absolute volatile abundances in

parental magmas from the mean K

O concentrations in

them (Table 2). Literature data were also used where

available (Arenal, Fuego, Cerro Negro, Pacaya, Irazu

It should be noted that the applied approach provides no

guarantee that estimated ratios were not affected by earlier

degassing that was not recorded in the melt inclusions.

Therefore, the reported volatile contents and ratios should

Table 2 Estimated compositions of parental melts

Arc segment Guatemala Nicaragua Costa Rica

Volcano Santa Maria Atitla

n Fuego Agua Cerro Negro Nejapa Granada

(low-Nb Group)

Granada

(high-Nb Group)

Arenal Irazu

Olivine Fo range 77–79 75–76 75–77 77–81 79–83 82–83 79–86 86–87 76–79 88–89

SiO

(wt%) 51.2 51.4 49.2 48.9 46.9 46.2 46.9 47.3 49.3 50.2

TiO

(wt%) 0.83 0.99 0.74 0.78 0.57 1.12 0.79 0.94 0.82 0.80

FeO (wt%) 9.0 10.3 10.7 9.4 11.0 11.7 11.2 9.3 10.8 8.0

O (wt%) 2.8 2.8 2.8 2.9 1.4 1.8 1.7 1.8 2.0 2.4

O (wt%) 0.85 0.89 0.61 0.68 0.21 0.15 0.32 0.13 0.38 0.87

O/K

O 4.3 3.5 7 4.2 25 20 13 20 8 6.3

S/K

O 0.3 0.4 0.3 0.3 1.0 1.2 0.7 0.9 0.9 0.23

Cl/K

O 0.09 0.09 0.14 0.11 0.30 0.24 0.24 0.28 0.32 0.08

F/K

O 0.06 0.04 0.04 0.06 0.05 0.11 0.04 0.12 0.07 0.07

O (wt%) 3.7 3.1 4.3 2.9 5.4 3.0 4.2 2.7 3.1 5.5

S (wt%) 0.26 0.36 0.18 0.20 0.21 0.18 0.22 0.12 0.34 0.20

Cl (wt%) 0.08 0.08 0.09 0.07 0.05 0.04 0.08 0.04 0.12 0.07

F (wt%) 0.05 0.04 0.02 0.04 0.01 0.02 0.01 0.02 0.03 0.06

B (ppm) 12.7 13.4 10.0 9.2 9.2 1.2 8.8 1.0 7.6 4.8

Zr (ppm) 59 79 48 71 19 49 26 42 36 56

Y (ppm) 12 15 15 13 10 19 13 16 14 13

Nb (ppm) 2.7 3.0 1.9 2.6 1.2 4.1 1.5 4.1 2.7 5.9

Ba (ppm) 318 358 330 342 190 88 272 59 384 429

La (ppm) 7.0 8.0 6.3 7.5 2.3 3.1 3.5 3.7 10.2 17

Th (ppm) 0.55 0.85 0.71 0.52 0.12 0.16 0.20 0.16 0.74 1.8

U (ppm) 0.17 0.38 0.33 0.23 0.10 0.09 0.19 0.08 0.27 0.6

Notes. Major and trace elements in parental melts represent mean values obtained for a suite of cogenetic inclusions from every sample after

recalculating the compositions to be in equilibrium with olivine Fo

. ‘‘Olivine Fo range’’ indicates the range in Fo of the host olivines for the

inclusions used in determining the parental melt compositions. The volatile/K

O ratios for parental melts were estimated from populations of

variably degassed co-genetic inclusions (see Fig. 7) and generally represent maximum values for H

O/K

O, S/H

O and Cl/K

O and mean

estimates for F/K

O. Absolute concentrations of volatiles in parental melts are calculated from the volatile/K

O ratios and the mean K

O content

in parental melts. Literature data were also used for Arenal, Irazu

, Fuego, Cerro Negro and Pacaya (Benjamin et al. 2007; Roggensack 2001a, b;

Roggensack et al. 1997; Wade et al. 2006; Walker et al. 2003;; Sisson and Layne 1993)

Contrib Mineral Petrol

123

be considered as minimum values for true parental melts.

For this reason, we did not include data for Masaya Volcano

in our compilation, because the inclusions were trapped in

evolved olivines and H

O, S and Cl had already signiﬁ-

cantly degassed, when compared to inclusions in more

forsteritic olivines from Nicaragua or Guatemala (Fig. 5).

The other disadvantage of this approach is that it does not

account for possible variations of volatile contents and

ratios to incompatible elements in parental magmas at a

single locality. In light of large geochemical heterogeneity

of inclusions from some samples, volatile heterogeneity is

also to be expected (Fig. 3). This uncertainty cannot be

quantitatively accounted for at present and requires better

statistics on both volatile components and other incompa-

tible elements in melt inclusions.

Discussion

What are the causes of incompatible trace element

variations along the arc?

Incompatible trace elements measured in melt inclusions

exhibit large regional variations, which agree well with the

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.0

1.0

2.0

3.0

4.0

5.0

6.0

7.0

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0 0.2 0.4 0.6 0.8

0.00

0.01

0.02

0.03

0.04

0.0 0.2 0.4 0.6 0.8

t %wO

t %wlC

t %

t %wS

K O wt %

O wt %

H O/K O=25

+20%

-20%

Cl/K O=0.3

CO /K O=0.6

t %wF

F/K O=0.05

syrC

S/K O=1.0

Degas.

Fig. 7 Water, S, Cl, CO

and F concentrations in Cerro Negro melt

inclusions plotted versus K

O content. Filled symbols are data from

this study, open symbols from Roggensack (2001a), Roggensack et al.

(1997), ‘GM’ labeled box indicates groundmass composition of the

host tephra. Note decreasing H

O, S and CO

concentrations with

increasing K

O, which indicates crystallization accompanied by

extensive magmatic degassing of these volatiles. Chlorine concen-

trations increase with increasing K

O but Cl/K

O decreases slightly

indicating that some Cl partitions into the ﬂuid phase. Maximum

O/K

O, S/H

O, CO

O and Cl/K

O ratios are believed to be

closest to the parental melt composition. Fluorine concentrations

increase nearly proportionally to K

O, which implies that signiﬁcant

F degassing does not occur during crystallization. Therefore F/K

may be informative of parental melt composition. Solid lines illustrate

constant volatile/K

O ratios, which are assumed to be characteristic

for parental melt of this volcano. Dashed lines bracket ±20%

uncertainty of the parental melt estimates. Error bars correspond to

15% relative standard deviation (RSD) for volatiles and 10% RSD for

Contrib Mineral Petrol

123

geochemical zoning observed in the whole rocks (Fig. 4).

Therefore the conclusions drawn from melt inclusion and

whole rock geochemistry are similar. Both whole rock and

melt inclusion compositions change systematically along

the arc. The highest Ba/La, B/La, U/Th and Sr/Ce ratios are

observed in Nicaragua. The ratios decrease southwards into

Costa Rica, reaching very low values similar to oceanic

basalts at the southernmost Irazu

Volcano, and northwards

into Guatemala, reaching intermediate values (Fig. 4). As

pointed out by many authors, the large variation in highly

incompatible-element ratios in the Central American lavas

can be attributed to variable ﬂuid ﬂux from the subducting

Cocos Plate and also to multi-stage and/or multi-compo-

nent processes of mantle enrichment (Carr et al. 1990; Eiler

et al. 2005; Leeman et al. 1994; Patino et al. 2000).

The sedimentary pile on the subducting Cocos Plate

consists of two units: the lower consists of carbonates and

the upper of hemipelagic sediments (see overview in Patino

et al. 2000). Although both sedimentary units have high U/

Th and Ba/La (Figs. 4, 8a), the hemipelagic sediments have

high U concentrations and U/La and low Ba/Th compared

to the carbonates, making it possible to distinguish between

the two sedimentary input components (Fig. 8b). As shown

by Patino et al. (2000), Ba/Th and U/La form negative

arrays for lavas from some volcanoes in Central America

(for example see ﬁeld for Telica in Fig. 8b), indicating that

varying proportions of the two sedimentary units are

involved in the magma generation process. Cerro Negro

and low-Nb Granada melt inclusions extend to high Ba/Th

at low U/La indicating an important role for carbonate

sediments, whereas Masaya melt inclusions trend towards

higher U/La at low Ba/Th, indicative of the presence of a

hemipelagic component in their magma genesis (Fig. 8b).

Most of the melt inclusion data, however, have relatively

low Ba/Th, U/La and U/Th and cluster near depleted

MORB-source mantle (DMM). Melt inclusions with little

contribution of U- and Ba-rich component(s) most likely

derived from subducted sediments include all but one data

point from Guatemala, some data from all volcanoes in

Nicaragua except Masaya and data from both studied Costa

Rican volcanoes.

Additional important questions concerning the genesis

of Central American magmatism that can be addressed

with trace element geochemistry are whether other portions

of the subducting slab also contribute to magma genesis

and whether slab components that are added to the mantle

wedge have ﬂuid- or melt-like properties. Ratios of ele-

ments that are not highly ﬂuid mobile in arc systems, such

as the REE and HFSE, can help evaluate the role of slab

components with melt-like properties. On a diagram of

La/Sm versus Ba/La (Fig. 8c), melt inclusions from central

Nicaragua (Cerro Negro, Granada and Nejapa) form a

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6

La/Y

Y/bN

DMM

3=b

/aL

0=b

EM (?)

)?(.pmocbalS

Telica

(a)

Telica

(b)

100

120

140

160

0246810

La/Sm

aL/aB

DMM

(d)(c)

DMM

500

1000

1500

2000

2500

3000

0.00 0.05 0.10 0.15 0.20

U/La

hT/aB

Santa Maria Fuego

Atitlan Agua

Cerro Negro Nejapa

Masaya Granada

Arenal Irazu

DMM

HS+CS

Slab melts (?)

EM (?)

Fluid

100

120

140

160

0.0 0.5 1.0 1.5

U/Th

aL/aB

′

Fig. 8 Variations of

incompatible trace element

ratios in the melt inclusions.

Small gray symbols illustrate

compositions of primitive to

moderately evolved rocks from

Central America (SiO

\ 54 wt

%). Shadowed ﬁeld encloses

compositions of Telica Volcano

rocks (Patino et al. 2000). CS

and HS labels indicate

compositions of hemipelagic

and carbonate sediments,

respectively, from the DSDP

site 495 (Patino et al. 2000);

DMM is the composition of

depleted MORB mantle source

(Salters and Stracke 2004);

EM–Nb-enriched (OIB-like?)

mantle

Contrib Mineral Petrol

123

nearly vertical trend at low La/Sm. Nejapa and high-Nb

Granada inclusions have the lowest Ba/La ratios, whereas

Cerro Negro and low-Nb Granada inclusions extend to the

highest Ba/La ratios. Taken together with the discussion of

Ba/Th versus U/La systematics above, this trend could be

explained by the addition of variable amounts of Ba-rich

(derived from subducting carbonate sediments) ﬂuid, most

clearly seen in Cerro Negro and low-Nb Granada melt

inclusions, to a relatively depleted (DMM) mantle wedge,

represented most closely by high-Nb Granada and Nejapa

melt inclusions. Oxygen isotope data also indicate sub-

stantial variation in the amount of slab components

involved in the sources of Nicaraguan magmas (Eiler et al.

2005). Cerro Negro magmas, however, have lower d

than those from Nejapa and Granada, opposite of what is

expected for a sediment-derived component with high

O. This observation suggests that the slab-component

(ﬂuid), carrying the strong sediment signature to the Cerro

Negro magma sources, contains only a very small mass

fraction of sedimentary material. The bulk of the slab

component must be largely derived from the lower oceanic

crust or serpentinites (both of which have low d

O; Eiler

et al. 2005), possibly scavenging Ba from overlying sedi-

ments as it migrates upwards.

Melts from Guatemalan Volcanoes and Masaya Volcano

in Nicaragua have intermediate Ba/La and La/Sm and high

O above mantle values compared to similarly Ba-rich

Cerro Negro melts (Fig. 8 a, b). These systematics are

difﬁcult to reconcile with a single slab-component model

(Eiler et al. 2005). A more plausible explanation could be

that another Ba- and LREE-rich and high-d

O crustal

component is involved in magma genesis beneath Guate-

mala. The absolute amount of this crustal component in the

sources of Guatemalan magmas may be even larger than in

Nicaragua, and its geochemical features (high d

O) are

more compatible with derivation of the ﬂuid from altered

oceanic crust and/or subducting sediments rather than

serpentinite (Eiler et al. 2005). The origin and provenance

of this component, which also has relatively low B/Be and

Be/

Be (Leeman et al. 1994; Morris et al. 1990), is

unclear and represents one of the most fundamental prob-

lems of Central American arc volcanism. This component

could be a sediment-derived melt (Eiler et al. 2005), pos-

sibly solidiﬁed and aged in the mantle wedge and melted

later as a result of ﬂuid ﬂuxing from deeper portions of the

subducting slab (Leeman et al. 1994; Patino et al. 2000), or

alternatively melt from the upper continental crust with

high La/Sm and low Ba/La (Carr et al. 1990).

The Costa Rican magmas are characterized by the

highest La/Sm and La/Yb ratios (greatest LREE enrich-

ment) but among the lowest Ba/La, Ba/Th and U/Th ratios,

suggesting either involvement of an enriched mantle wedge

(Abratis and Wo

rner 2001; Carr et al. 1990; Feigenson

et al. 2004) or a melt-like component derived from the

magmatic portion of the downgoing slab (Fig. 8). Although

the Guatemalan melts have intermediate La/Sm, Ba/La, U/

La and Ba/Th compared to Costa Rican and some Nica-

raguan (Cerro Negro and low-Nb Granada) melts, the

lower, mantle-like d

O of the Costa Rican magmas pre-

cludes mixing of the Costa Rican melt-like slab component

with the Nicaragua ﬂuid-like slab component to generate

the composition of the Guatemalan melts. Our conclusion

from these observations is that the along arc geochemical

variations in Central America reﬂect involvement and

mixing of multiple distinct crustal (from subducting and/or

overriding plates) and mantle (wedge and subducting slab,

e.g. serpentinite) components, with slab components hav-

ing both ﬂuid- and melt-like characteristics, and variable

extents of partial melting of the metasomatized mantle

wedge (Eiler et al. 2005).

Two types of Nb enrichment in Central America

The earlier discussion about different slab components

contributing crustal signatures to the Central American

magmas is also relevant for reconciling a dilemma about

mantle versus slab source control on HFSE (Nb, Zr) and

LREE enrichment in lavas (Abratis and Wo

rner 2001;Carr

et al. 1990; Feigenson et al. 2004; Leeman et al. 1994). To

address this problem, we plotted compositions of primitive

melt inclusions from all segments of the arc in coordinates

Nb/Y versus La/Y (Fig. 8d). This diagram clearly dem-

onstrates the existence of two distinctive compositional

trends of coupled enrichment in Nb and La in the Central

American magmas, both converging at very low Nb/Y and

La/Y similar to the depleted MORB source (Salters and

Stracke 2004). Melt inclusions from Nejapa, high-Nb

group from Granada and some Cerro Negro melts (all from

central Nicaragua) form one trend, which points to the

source component with low La/Nb*0.5. Melt inclusions

from Guatemala, Costa Rica and the remaining from Nic-

aragua (Masaya, low-Nb Granada group and some from

Cerro Negro) form the other trend toward a Nb- and La-

rich component with relatively high La/Nb*3.

The low La/Nb observed in some Nicaraguan melts is

unusual for subduction-related magmas and more typical

for incompatible-element-enriched oceanic basalts (Hof-

mann 2003; Sun and McDonough 1989). The simplest

explanation for Nb enrichment in Nicaraguan magmas is

involvement of HFSE-rich peridotite (or pyroxenite;

Sobolev et al. 2007) similar to the source of oceanic island

basalts and possibly imbedded in depleted DMM-like

matrix (Carr et al. 1990; Walker et al. 1990). The Nb-rich

basalts in Nicaragua however have unusually high Nb/U

ratios (up to 100, Fig. 4) compared to most OIB with Nb/U

Contrib Mineral Petrol

123

of 47 ± 10 (Hofmann 2003). Nb/U ratios exceeding 100

have been found in Canary Island basalts, possibly

reﬂecting contamination of these intraplate magmas with

melts from amphibole/phlogopite-rich veins located in the

lithosphere (Lundstrom et al. 2003). Nb-enriched litholo-

gies can also be formed during dehydration and/or partial

melting of slab material (Garrido et al. 2005; Ionov and

Hofmann 1995).

Unlike Nicaragua, Nb-enrichment in Guatemala and

Costa Rica is coupled with strong LREE enrichment

(Fig. 8 c, d). The pronounced negative Nb anomaly in

normalized trace element spectra of these melts (Fig. 3)

points to subduction-related metasomatism. Noteworthy is

also the close coincidence of La/Nb (*3) in Guatemalan

melts and subducting hemipelagic sediments. Sediment

melts can therefore explain some Nb-enrichment in the

magma source beneath Guatemala as proposed recently for

southern Kamchatka (Duggen et al. 2007).

Irazu

melts exhibit the highest La/Y and Nb/Y along the

high-La/Nb trend, which can be in principle explained by

involvement of low-degree melts from subducting sedi-

ments leaving eclogitic residue. This hypothesis however is

not consistent with existing Pb-isotope data indicating

strong HIMU signature in Costa Rican magmas, closely

resembling Gala

pagos-type mantle (Abratis and Wo

rner

2001; Feigenson et al. 2004) or crustal material. On the

basis of this data, a reasonable explanation for the peculiar

geochemical features of the Costa Rican volcanic rocks can

be involvement of partial melts from the subducting slab

(i.e. Gala

pagos hotspot track), leaving a rutile-bearing

eclogitic residue to explain high LREE and La/Nb. Alter-

natively, the Gala

pagos crustal signature could be inherited

from Gala

pagos-related complexes formerly accreted to the

Costa Rican fore-arc, which were eroded and then melted

beneath the arc (Goss and Kay 2006). These two hypoth-

eses can probably be sorted out with the help of detailed

Pb-isotope studies but in essence they both coincide in the

principle conclusion that the enrichment in HFSE and

LREE is derived from the subduction input and not

intrinsic to the mantle in the wedge.

Test of water-ﬂuxed melting beneath Central America

Most previous studies of Central American arc magmatism

assumed that mantle melting is ﬂuxed by water-rich ﬂuids

or melts from the subducting plate (Carr et al. 1990; Eiler

et al. 2005; Patino et al. 2000; Walker et al. 1990),

although this has not yet been directly demonstrated. A

crucial test for the viability of water-ﬂuxed melting is the

existence of an inverse correlation between water content

and incompatible element content in the primary magmas,

which implies open system melting promoted by water

addition to the source (Stolper and Newman 1994). The

best candidates for indices of partial melting are Na, Ti, Y

and HREE. The abundances of these elements in mantle

sources are relatively constant, and their abundances are

either high in the mantle wedge and thus not substantially

affected by the addition of small amounts of ﬂuids (e.g. Na)

or low in slab-derived ﬂuids and melts (e.g. Ti, Y, HREE).

Covariations between H

O, TiO

, Y and Na

O in the

inferred parental Central American magmas are shown in

Fig. 9.

First-order observation from our data is that H

Oin

parental melts does correlate inversely with TiO

, Y and

O. Squared correlation coefﬁcient for linear regressions

range from 0.43 (H

OvsNa

O) to 0.65 (H

O vs TiO

) and

are signiﬁcant at the 99% conﬁdence level. Thus, our

principle conclusion is that melting beneath all parts of the

Central America arc is indeed triggered by addition of

water-bearing components. Similar correlations can also be

produced by mixing of incompatible-element-depleted,

water-rich melts (e.g. Cerro Negro) with incompatible-

element-enriched water-poor melts (e.g. Guatemalan rear-

arc). This however does not explain other geochemical

features of the Central American magmas, which require

several regionally different compositional components as

discussed above (see Fig. 8).

In this work we did not attempt a fully quantitative

modeling of ﬂux melting for the Central America magmas

but made some preliminary estimates of the thermal con-

ditions of melting and possible compositions of mantle

sources, which govern productivity of water-ﬂuxed melting

and are crucial for correct determination of degrees of

partial melting (Kelley et al. 2006; Portnyagin et al. 2007).

We did our modeling following the approach and equations

of water-ﬂuxed melting from Portnyagin et al. (2007),

which assumes simple batch melting with constant parti-

tion coefﬁcients and uses parameterization of hydrous

melting after (Katz et al. 2003).

First, we determined that most compositions of the

Central American parental melts are bracketed by the

model compositions of partial melts generated by water-

ﬂuxed melting of depleted MORB mantle (DMM; Salters

and Stracke 2004) in spinel facies at temperatures from +25

to –50C of the dry peridotite solidus (DPS) (Fig. 9). This

result is in agreement with previous estimates (Portnyagin

et al. 2007) and suggests that (1) mantle wedge beneath the

Central American volcanic belt is relatively cold compared

to mantle source regions beneath mid-ocean-ridges, back-

arcs and possibly intra-oceanic arcs (Portnyagin et al.

2007), (2) nearly all melting is water-ﬂuxed beneath Cen-

tral America, and (3) dry decompression melting, which

occurs at temperatures above the dry peridotite solidus, is

limited to several percentage and is clearly less important

than water-ﬂuxed melting.

Contrib Mineral Petrol

123

Looking at the data in detail, the position of individual

data points and the whole data array relative to the model

melt compositions at different temperatures are not con-

stant in the different diagrams in Fig. 9. For example,

O–Y variations suggest temperatures near DPS, while

O–TiO

and H

O–Na

O suggest temperatures mostly

below DPS. This discrepancy can arise from the very

simpliﬁed modeling approach but also can be informative

of some variability of the mantle source compositions and

changes in partition coefﬁcients during partial melting. For

example, if we assume that all melting in Central America

took place at the DPS temperature, the deviations of melt

compositions towards higher TiO

and Na

O from those

predicted in the model can be explained if the mantle

composition varied from DMM-like (TiO

= 0.133 wt%)

to more Ti- and Na- rich (TiO

up to 0.4 wt%, Na

Oupto

*0.8 wt%) (Fig. 9 a, c). This is in accordance with inde-

pendent evidence of involvement of enriched mantle in

magma genesis in Central America (Carr et al. 1990; Fei-

genson and Carr 1993; Walker et al. 1990; Patino et al.

2000) and with the incompatible-element patterns of high-

Nb melts from Granada and melts from Nejapa which have

low H

O and are likely to most closely reﬂect the mantle

wedge composition (Fig. 3). Deviations of data points

towards lower Y compared to the model predictions at DPS

temperature may suggest that initial mantle concentrations

were lower than the DMM value (Y = 4.07 ppm) or more

likely Y partition coefﬁcient (0.095) chosen for modeling

was too low. We prefer the latter explanation, which is

consistent with the common presence of garnet in the

sources of the Central American magmas (Feigenson and

Carr 1993) and is evident from the higher normalized

contents of intermediate to heavy REE (Fig. 3). According

to our maximum estimate for bulk Y partition coefﬁcient

(*0.2) (Fig. 9b), the quantity of garnet in the sources of

the Central American magmas did not exceed *2.5%.

Reliable estimates of the amount of water introduced

into the sources of magmas require well-justiﬁed source

compositions and partition coefﬁcients for every volcano/

sample (Kelley et al. 2006; Portnyagin et al. 2007). As is

evident from the discussion above, this is not an easy task

for Central America magmatism and requires reconciling

of effects of variable temperature, source composition and

partition coefﬁcients on the composition of mantle melts.

Nevertheless, we made a rough estimate of the amount of

water in the sources in order to check agreement of the

mass balance of mantle and slab components predicted

Fig. 9 Correlations of H

O and a TiO

, b Y and c Na

O content in

inferred parental melts. Compositions of the Guatemala rear-arc

parental melts were calculated using data from Walker et al. (2003)

and are shown for comparison with volcanic front samples. Error

bars correspond to 20% RSD. Solid curves are modeled compositions

of partial melts (from 2 to 30% melting with 2% increments as

indicated in a) produced during water-ﬂuxed melting of depleted

MORB source (DMM; Salters and Stracke 2004) in spinel facies at

different temperatures (C) above and below the Dry Peridotite

Solidus (DPS). Amount of water in mantle source calculated at

DT(DPS) = –25C (or 25C below the DPS) is shown at the top of

diagram (a). Dashed curves illustrate melting trajectories at

DT(DPS) = 0 (or simply at DPS) with slightly different input

parameters: a TiO

= 0.25 wt% in source instead of 0.133 wt% in

DMM; b in presence of 2.5% garnet in the source (bulk D

= 0.2

instead of D

= 0.095 in spinel facies); c Na

O = 0.8 wt% in source

instead of 0.29 wt% in DMM. Modeling was done following

equations and approach by Portnyagin et al. (2007). Partition

coefﬁcients used were the same as in Kelley et al. (2006). Note that

position of data points changes relative to modeled curves in different

coordinates, which indicates that DMM is not an appropriate source

for all magmas in Central America and some melting could take place

in the garnet stability ﬁeld

0.3

0.5

0.7

0.9

1.1

1.3

1.5

1.7

H O wt% at Fo91

oFt

a%twOiT

19oF

tamppY

19o

Fta%tw

(c)

(b)

(a)

H O wt% in source at DPS-25

0.1

0.5

1.0

1.5

2.0

01234567

Santa Maria Atitlan

Fuego Agua

Cerro Negro Nejapa

Granada (LNB) Granada (HNB)

Arenal Irazu

Guat. BVF

-25

-50

+25

T(DPS)=0

=0.25 wt%TiO

-25 -50

+25

T(DPS)=0

2.5% Ga

-25 -50

+25

T(DPS)=0

O =0.8 wt%

T(DPS)

(+)

(-)

0.26

0.22

0.18

0.14

0.10

0.30

0.06

foe

′

Contrib Mineral Petrol

123

from our data and d

O systematics (Eiler et al. 2005).

Overall, the amounts of water in the mantle sources esti-

mated for spinel facies DMM melted at 25C below DPS

vary from *0.1 wt% for rear-arc melts in Guatemala up to

*1.7 wt% for Cerro Negro (Fig. 9a). The latter magmas

also have the lowest d

O in Central America, and Eiler

et al. (2005) estimated up to 4 wt% slab component with

50 wt% water and d

O *0% in their source. We calcu-

late from our data that Cerro Negro sources contained

*3.4 wt% slab component with 50 wt% water, in excel-

lent agreement with the oxygen isotope data.

Trace element proxies of volatiles

In studies of arc magmatism, ﬂuid mobile to less ﬂuid

mobile or ﬂuid immobile trace element ratios (e.g. Ba/La,

Ba/Th, U/Th, Ba/Nb) are often assumed to reﬂect the rela-

tive ﬂuid contribution from the subducting slab to the

mantle wedge. If this can be demonstrated to be the case

for an arc, then these ratios could be used as proxies for

determining water contents of magmas. This correlation

can however be highly obscured due to the diversity of

water-bearing components involved in arc magma gener-

ation. For example, Portnyagin et al (2007) showed that

water in various primitive magmas of Kamchatka inversely

correlates with TiO

and Na

O but does not correlate with

Ba/La or Ba/Nb. The only correlation found was between

relative enrichment in boron and water compared to con-

centrations of ﬂuid-immobile Nb and Zr. On the other

hand, Wade et al. (2006) demonstrated that correlation

between H

O and Ba/La in parental melts may indeed be

signiﬁcant in Central America in contrast to Kamchatka.

Our new data provide further insights into the relationships

between trace element geochemistry and enrichment in

water and other volatile elements in the Central American

magmas.

As illustrated in Fig. 10, water content in the majority of

parental melts of Central America strongly correlate with

Ba/La and B/La ratios. Even better correlation is expected

between these trace element ratios and the amount of water

in the magma sources, because this amount of water is

proportional to H

O/TiO

or H

O/Y, exhibiting even larger

relative range than absolute H

O concentrations (Fig. 8).

Existence of this correlation can be explained if LREE-rich

melt-like components required by magmas in Guatemala,

Nicaragua (e.g. Masaya) and western Costa Rica had low B

and Ba concentrations and were poor in H

O. In this case

elevated Ba/La and B/La ratios in magmas can be ulti-

mately related to the magnitude of ﬂux of Ba-, B- and H

O-

rich ﬂuid, generating the correlations observed in Fig. 10.

The Irazu

volcano, however, falls off the general Central

American arrays in Fig. 10 and exhibits high H

O at very

low B/La and Ba/La. Therefore compositions of slab

components involved in magma genesis in the southern

part of the arc may have lower Ba/La and B/La than in

northern localities, consistent with lower amounts of Ba-

rich sediments on the subducting Cocos Ridge than to the

north. On the other hand, the high H

O contents suggest

that the subducting hotspot track may contribute a large

amount of water to arc volcanism beneath Irazu

or that

much of the water ﬂux from the subducting Cocos Ridge is

concentrated beneath Irazu

Volcano. Clearly the compo-

sition and thermal state of the plate subducting beneath

Irazu

, which has been overprinted by the Gala

pagos hot-

spot, is strikingly different from normal oceanic crust

entering the trench in other parts of the arc (Fig. 1)

(Leeman et al. 1994; Protti et al. 1995). Although we

believe that a plausible explanation for the exceptionally

B/La

aL/a

(a)

(b)

Santa Maria Atitlan

Fuego Agua

Cerro Negro Nejapa

Granada (LNB) Granada (HNB)

Arenal Irazu

Guat. BVF

y = 0.9707x + 1.776

= 0.7317

y = 2.8544x

0.4529

= 0.6095

012345

)%tw(19oFtaOH

y = 0.0481x + 1.1294

= 0.8329

y = 0.4087x

0.5521

= 0.7421

0255075100

)%tw(19oFtaOH

N-MORB

′

Fig. 10 Correlations of water with Ba/La and B/La ratios in the mean

parental melts of the Central American volcanoes. Error bars

correspond to 20% RSD. Data were approximated by power and

linear regression lines (excluding Irazu

Volcano) allowing estimate

of H

O content in parental melt from whole rock Ba/La or B/La

ratios. N-MORB composition is shown for reference (Chaussidon and

Jambon 1994; Sun and McDonough 1989)

Contrib Mineral Petrol

123

high H

O and low Ba/La in Irazu

magmas can be found,

our current conclusion is based on only a few data points,

which need to be veriﬁed through more extensive studies of

volatiles in Costa Rican magmas.

Other volatile components in parental magmas can also

be traced with lithophile element ratios. Chlorine and to

lesser extent sulfur concentrations in parental magmas

correlate with La/Nb ratio (Fig. 11) and only weakly cor-

relate with H

O or Ba/La (not shown), suggesting that both

S and Cl may be delivered to the mantle source via the

melt-like component with high La/Nb and not with a H

O-

rich ﬂuid. Fluorine concentrations in parental melts corre-

late with La/Y ratio suggesting coupled behavior of LREE

and F during magma generation in subduction zones

(Portnyagin et al. 2002). The slightly steeper slope deﬁned

by the data from Costa Rica can reﬂect a more important

role for garnet in the source of these magmas resulting in

elevated La/Y (Feigenson and Carr 1993).

Fluxes of volatile elements

Minimum estimate of volatile ﬂuxes through the Central

American Volcanic Arc can be obtained by combining the

ﬂux of magma to the surface with volatile abundances

determined for parental magmas in this study. Because

direct measurements could only be carried out on a limited

number of samples, we used the previously discussed

correlations between incompatible element ratios (Ba/La,

La/Nb and La/Y) and volatile abundances in parental melts

estimated from melt inclusions in order to constrain the

volatile abundances in parental magmas throughout Central

America from whole rock geochemical data (Fig. 12).

From the mean calculated trace element ratios for each arc

segment as given in the Centam data base of compositions

of whole rocks (Carr et al. 2003), we calculated average

volatile abundances for every arc segment, assuming that

there was no change in the incompatible trace element

ratios as a result of differentiation (Table 3). Volatile ﬂuxes

were then calculated by multiplying the average volatile

content in parental magma for an arc segment with the

average volcanic ﬂux for that segment (Carr et al. 1990,

2007). We note that the volatile contents for central Costa

Rica (Cordillera Central) volcanoes are poorly constrained;

very large variations are predicted from the whole rock

data. In particular, it is not clear how representative the

high H

O concentrations of Irazu

magmas (up to 5 wt%)

are for the neighboring volcanoes. Therefore, we used

average Arenal and Irazu

parental volatile contents for the

other volcanoes in central Costa Rica and assumed an

uncertainty of factor two for all estimates of volatile

abundances and calculated ﬂuxes for the segment-averaged

parental melts of central Costa Rica. This uncertainty,

however, only introduces a small error in the estimates of

y = 0.0239x + 0.0139

= 0.8591

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

0.0 1.0 2.0 3.0 4.0 5.0

Santa Maria Atitlan

Fuego Agua

Cerro Negro Nejapa

Granada (LNB) Granada (HNB)

Arenal Irazu

La/Nb

)%tw(19

oFtalC

(a)

y = 0.042x + 0.0082

= 0.7306

0.00

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0.0 0.5 1.0 1.5

La/Y

tw(19

oFtaF

(c)

y = 0.0723x + 0.0885

= 0.7835

0.00

0.10

0.20

0.30

0.40

0.50

0.0 1.0 2.0 3.0 4.0 5.0

La/Nb

)

%tw(

Excluded from

correlation

(b)

′

Fig. 11 Correlations of Cl, S

and F with trace element ratios

(La/Nb and La/Y) in parental

melts. Error bars correspond to

20% RSD. The data were

approximated by linear

regressions, which allow

assessment of volatile contents

from whole rock trace element

ratios. The regression line in

b does not include data for

Fuego volcano as it could be

affected by S degassing

Contrib Mineral Petrol

123

the total volatile ﬂuxes from the Central American volca-

nism, which are dominated by the Guatemala, El Salvador

and Nicaragua segments

Considering only volcanic output, the largest ﬂuxes of

all volatiles were obtained for Guatemala and El Salvador.

Water ﬂuxes for Nicaragua and Costa Rica are similar.

Sulfur, chlorine and ﬂuorine ﬂuxes tend to be higher in

Costa Rica than in Nicaragua (Fig. 13a). Mean (normalized

to the length of segments) volcanic ﬂuxes of volatiles along

the Central American Arc are 1.1 · 10

kg/m/Ma H

8.3 · 10

kg/m/Ma for S, 2.3 · 10

kg/m/Ma for Cl and

8.9 · 10

kg/m/Ma for F.

Volatile ﬂuxes estimated above represent minimum

magmatic ﬂuxes, because they neither take into account

magmatic fractionation before eruptions nor consider

magmas that completely stagnate and fractionate in the

lithosphere without any portion ever reaching the surface.

The amount of fractionation is relatively difﬁcult to quantify

for all segments of the Central America because parental

melt compositions are not well constrained on a regional

scale. Using Carr et al. (2007) data for volcanic ﬂuxes (1.3–

1.6 · 10

kg/m/Ma), ﬂux of K

O (1.3–1.8 · 10

kg/m/

Ma) in Nicaragua and assuming K

O * 0.2–0.3 wt% in

parental magmas (Table 2), we estimate that volcanic

products in Nicaragua have on average 1–1.2 wt% K

O and

can originate through approximately 75–80% of crystalli-

zation of parental melts. In other words, the ﬂux of parental

magmas feeding volcanism in Nicaragua and, possibly,

along the entire volcanic front in the Central America is at

least 4–5 times higher compared to the amount of erupted

products. Therefore, ﬂuxes of volatiles estimated above

should be at least 4–5 times higher if we take the ﬂux of

primary magmas into account.

The major uncertainty of the ﬂux estimate comes,

however, from the unknown amount of intruded magmas

comprising the hidden magmatic ﬂux, which cannot be

assessed by studying volcanic products. Present-day vola-

tile ﬂuxes obtained by remote-sensing satellites or on-land-

based techniques provide some clues to the magnitude of

the total volatile ﬂuxes (Hilton et al. 2002; Zimmer et al.

2004). Combined with data on concentrations of volatiles

in parental magmas from this study, this allows us to

estimate the total magmatic ﬂux in Central America and

the amount of intruded magmas. This calculation is how-

ever based on a major assumption: that present-day SO

ﬂuxes are representative of long-term SO

ﬂuxes deter-

mined from melt inclusions and volcanic ﬂuxes estimated

for the past *200,000 years.

The total SO

ﬂux from volcanoes in Central America

was estimated as 21.3 · 10

mol/year from COSPEC

measurements (Hilton et al. 2002). Recalculated to mass

units and normalized to 1 m of the arc length, this estimate

gives S ﬂux of 6.45 · 10

kg/m/Ma. This is *8 times more

than our estimate of volcanic S ﬂux (8.3 · 10

kg/m/Ma,

Table 3) and suggests that volcanic rocks in Central

America may account for *13% or less (since not all S will

reach the atmosphere) of the total magmatic ﬂux from the

mantle to the lithosphere if the present-day SO

ﬂux was

constant over the past *200,000 years. By combining

COSPEC SO

ﬂux data (Hilton et al. 2002) with the average

S content in parental magmas (*0.3 wt%, Table 3) and

assuming complete degassing of magmas and that all the

S reaches the atmosphere, we can estimate total magmatic

ﬂux in Central America as *2.2 · 10

kg/m/Ma. Volca-

nic products comprise *13% of this amount, *39% are

0.00

0.02

0.04

0.06

0.08

0.10

0.12

Distance alon

arc in km

@%twO

GLNB

GHNB

Guatemala

Nicaragua

Costa Rica

@%tw

@%twlC

@%twF

H O*= (Ba/La)

Cl*= (La/Nb)f

S*= (La/Nb)f

200

400 600

800

1000

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.00

0.05

0.10

0.15

0.20

F*= (La/Y)f

El Salvador

Fig. 12 Concentrations of volatile components in the Central Amer-

ican parental melts calculated from the compositions of primitive to

moderately evolved rocks (SiO

\ 54 wt%) using trace element

proxies (Ba/La for H

O, La/Nb for Cl and S, and La/Y for F). Black

dots illustrate direct estimates based on this melt inclusion study (see

Table 2 for compositions of parental melts)

Contrib Mineral Petrol

123

Table 3 Fluxes of volatiles in the Central American volcanic arc

Value Unit 1: Guatemala 2: El

Salvador

3: Western

Nicaragua

4: Eastern

Nicaragua

5: Northern

Costa Rica

6: Central

Costa Rica

7: average CA

(volcanic)

8: average CA

(total ﬂux)

9: global

average

10: CA/

global

11:subduction

input

12: output/

input

Segment length km 262 252 166 137 92 150 1,059

Magma ﬂux kg/m/Ma 4.00E + 10 3.75E + 10 1.30E + 10 1.60E + 10 1.50E + 10 2.40E + 10 2.8E + 10 2.2E + 11 1.9E + 11 1.17

Compositions of parental melts

Ba/La (WR) 50 59 94 56 42 31

La/Nb (WR) 3.1 3.1 2.1 1.9 3.5 2.8

La/Y (WR) 0.7 0.36 0.26 0.28 0.6 1.1

O Wt% 3.5 4.0 5.6 3.8 3.1 4* 4.0

Cl Wt% 0.09 0.09 0.06 0.06 0.10 0.08 0.08

S Wt% 0.31 0.31 0.24 0.23 0.34 0.29 0.29

F Wt% 0.04 0.02 0.02 0.02 0.03 0.05 0.03

Estimate from volcanic ﬂux

O ﬂux kg/m/Ma 1.41E + 09 1.48E + 09 7.33E + 08 6.10E + 08 4.71E + 08 9.60E + 08 1.1E + 09 9.1E + 09 0.12 3.51E + 10 0.03

S ﬂux kg/m/Ma 1.25E + 08 1.17E + 08 3.12E + 07 3.61E + 07 5.12E + 07 6.98E + 07 8.3E + 07 5.5E + 08 0.15

Cl ﬂux kg/m/Ma 3.52E + 07 3.30E + 07 8.33E + 06 9.49E + 06 1.46E + 07 1.94E + 07 2.3E + 07 1.5E + 08 0.15 1.00E + 09 0.02

F ﬂux kg/m/Ma 1.50E + 07 8.75E + 06 2.49E + 06 3.19E + 06 5.01E + 06 1.31E + 07 8.9E + 06

Estimate from SO

ﬂux

O ﬂux kg/m/Ma 8.36E + 09 9.1E + 09 0.92 3.51E + 10 0.2

S ﬂux kg/m/Ma

6.45E + 08 5.5E + 08 1.18

Cl ﬂux kg/m/Ma 1.80E + 08 1.5E + 08 1.19 1.00E + 09 0.2

F ﬂux kg/m/Ma 6.93E + 07

Notes. Columns 1–6: Data for single segments along the volcanic front. Magma ﬂuxes correspond to volcanic ﬂuxes after Carr et al. 1990, 2007). Mean ratios Ba/La, La/Nb and La/Y are

calculated for primitive to moderately evolved (SiO

\ 54 wt%) rocks from every segment (Centam data base; Carr et al. 2003). Concentrations of volatiles in parental melts are calculated from

the mean trace element ratios (Figs. 10, 11, 12) except for H

O concentrations in central Costa Rica (marked with ‘*’), which were assumed to be intermediate between Arenal and Irazu

parental melts. Fluxes of volatiles were calculated from volcanic ﬂux and concentrations of volatiles in parental melts; Column 7: Length-normalized average estimate for volcanic ﬂux,

composition of parental melts and volatile ﬂuxes for the Central American volcanic front; Column 8: Total ﬂux of magmas and volatiles in Central America estimated from COSPEC SO

ﬂux

data (Hilton et al. 2002) (underlined value). Fluxes of H

O, Cl and F were calculated using volatile/S ratios in average parental melt from column 7. Total magmatic ﬂux was calculated from the

COSPEC SO

ﬂux and average sulfur content in parental melts; Column 9: Average global arc-length-normalized ﬂuxes of magma (recalculated from (Crisp 1984)) and volatiles (Wallace

2005) at volcanic arcs; Column 10: Ratio of the length-normalized Central America ﬂuxes to the global ﬂuxes; Column 11: Subduction-related input of volatiles to the Central American

subduction zone. An estimate for H

O includes only structurally bound H

O in sediments and oceanic crust (1.8 · 10

kg/m/Ma; Jarrard 2003) and H

O in serpentinites (1.7 · 10

kg/m/Ma,

M. Ivandic, personal communication). Chlorine is stored in sediments and the oceanic crust (6.63 · 10

kg/m/Ma; Jarrard 2003) and serpentinites (3.4 · 10

kg/m/Ma assuming that

serpentinites have sea-water-like H

O/Cl = 50); Column 12: Ratio of the length-normalized volatile ﬂux outputs in Central America to the subduction input

Contrib Mineral Petrol

123

cumulates related to volcanism (assuming average 75%

fractionation of parental melts), and the rest *48% are

intrusives solidiﬁed in the lithosphere (Fig. 13b). The

amount of intrusives and total magmatic ﬂux can certainly

be larger as some S can be ﬁxed in sulphide phases or

released from degassing magmas at depth, thus remaining

in the crust. Sulfur that never reaches the atmosphere, of

course, cannot be accounted for by COSPEC measure-

ments. Nevertheless, we note that the total magmatic ﬂux

for Central America (*2.2 · 10

kg/m/Ma) is surpris-

ingly similar to the global average magma ﬂux

(1.9 · 10

kg/m/Ma, based on the average global magma

output in volcanic arcs of 2.5 km

(Crisp 1984 and

assuming magma density 2.5 g/cm

and the total length of

volcanic arcs world-wide of 33,000 km). The coincidence

of both estimates is very remarkable as they were obtained

in fully independent ways. Volatile ﬂuxes in Central

America are also quite similar to the global arc averages

(Wallace 2005) (Table 3). These amounts represent sig-

niﬁcant mass fraction of volatiles initially stored in the

subducting plate (Table 3). For example, average magmatic

volatile ﬂuxes account for *20% of structurally bound

water and chlorine in the Cocos Plate crust being subducted

beneath Central America (Jarrard 2003).

Water ﬂux versus magmatic productivity

Several authors have proposed that the amount of volatile

elements (particularly H

O) released from the subducting

slab is larger beneath Nicaragua than Costa Rica (Ranero

et al. 2003;Ru

pke et al. 2004), which should result in a

greater magmatic ﬂux for Nicaragua. Our data demon-

strate that mantle sources of magmas in Nicaragua are

indeed more hydrated and generated magmas with

somewhat higher H

O content compared to Guatemala

and Costa Rica (excluding Irazu

Volcano). Calculated

ﬂuxes of H

O are however similar along Central America

or even lower in Nicaragua (Table 3, Fig. 13), reﬂecting

the relatively low volcanic ﬂux estimated for this region

based solely on volcanic ediﬁce volumes (Carr et al.

2007). Due to the many uncertainties in the present

available data, it is difﬁcult to judge if we have an ade-

quate picture of magmatic ﬂuxes along Central America,

which limits our ability to determine accurate volatile

ﬂuxes. The total magmatic ﬂuxes of volatiles in Nicara-

gua are likely to be under-estimated relative to other

Central American arc segments, because (1) tephras not

deposited on the volcanic ediﬁces, which represent a

signiﬁcant proportion of the eruptives, were not included

in the present volcanic ﬂux estimates (A. Freundt and

S. Kutterolf, personal communication) and (2) there may

be a higher proportion of intruded to erupted magmas in

Nicaragua compared to Costa Rica, consistent with evi-

dence for a larger magmatic ﬂux from the *6 times

greater fumarolic SO

ﬂux from the Nicaraguan volcanoes

(Frische et al. 2006) compared to those in Costa Rica

(Zimmer et al. 2004). The amount of eroded volcanic

rocks has also not yet been estimated, serving as another

uncertainty in the volcanic ﬂuxes.

1E+06

1E+07

1E+08

1E+09

1E+10

Guatemala El

Salvador

Western

Nicaragua

Eastern

Nicaragua

Northern

Costa Rica

Central

Costa Rica

Av. CAVA

(volcanic )

Av. CAVA

(total flux)

H2O S Cl F

(a)

aM/m/gknixulfelitaloV

Volcanics ~13%

~ 2.86 x10 kg/m/Ma

(b)

Intrusives ~ 48%

10.6 x10 kg/m/Ma

Cumulates ~ 39%

8.58 x10 kg/m/Ma

Total magmatic flux

22 x 10 kg/m/Ma

Fig. 13 Fluxes of magmas and

volatile components in Central

America. a Volcanic volatile

ﬂuxes estimated from inferred

compositions of parental melts

(this study) and published

volcanic ﬂux data (Carr et al.

1990, 2007), average length-

normalized volcanic ﬂux and

total volatile ﬂux estimated on

the basis COSPEC SO

ﬂux

(Hilton et al. 2002); b mass-

balance between volcanic

output, amount of cumulates

and amount of intrusives

contributing to the total

magmatic ﬂux as estimated

from the total COSPEC SO

ﬂux (Hilton et al. 2002),

volcanic ﬂuxes (Carr et al. 1990,

2007) and compositions of

parental melts (this study)

Contrib Mineral Petrol

123

Further studies aimed at quantifying magmatic ﬂuxes

along Central America clearly have potential to demonstrate

correlation between the amount of volatiles delivered to the

mantle and the magmatic ﬂux. We note however that

magmatic productivity (magmatic ﬂux) and the amount of

water delivered to the mantle do not have to be fully coupled

in subduction zones. The amount of water in the mantle,

which decreases the solidus of mantle peridotite, is certainly

important but not the only factor enhancing melt produc-

tion. Other important parameters affecting magmatic

productivity are (1) mantle temperature, (2) source com-

position and (3) shape and volume of melting region. These

factors are believed to govern magmatic productivity at

ocean ridges and hot spots (Langmuir et al. 1992; McKenzie

and Bickle 1988; Yaxley and Green 1998) and should be

equally important in subduction zones (Carr et al. 1990;

Kelley et al. 2006; Portnyagin et al. 2007). All parameters

affecting the amount and composition of arc magmas

should be taken into consideration in order to better under-

stand the ﬂuxes through the Central American subduction

factory and the origin of island-arc magmatism in general.

Conclusions

We studied melt inclusions in olivine from volcanic rocks

representing all major geochemical types of magmas from

the Central American volcanic front. Inclusions were

characterized for major and trace elements and volatiles

O, S, Cl, S) by electron and ion microprobes. These

results combined with literature data were used to estimate

mean parental magma compositions and elucidate their

origin in different segments of the Central American arc.

The major conclusions from our study follow:

1. Geochemical systematics of primitive melts suggest

that a large diversity of crustal (from subducting plate

and overriding plates) and mantle (incoming plate and

wedge) components are involved in the origin of

geochemical zoning along the Central America volca-

nic arc. Large heterogeneity of melt inclusions from

single rock samples implies that sources of magmas

are highly heterogeneous on the small scale.

2. Two different trends of coupled Nb and LREE

enrichment were recognized. High Nb/La (OIB-like)

component contributes to the composition of some

Nicaraguan magmas. In all other parts of the Central

American Arc, the LREE- and Nb-rich component has

relatively low Nb/La (*0.3) and most likely has a

subduction-related origin through partial melting of

subducting sediments and/or oceanic crust.

3. All parental melts are hydrous and contain *2–5 wt%

O, which correlates inversely with Ti, Y and Na and

directly with Ba/La and B/La. These correlations are

interpreted as indicating variable extents of melting of

heterogeneous mantle sources ﬂuxed by Ba- and

B-rich water-bearing components at conditions close

to the dry peridotite solidus. Cerro Negro Volcano in

central Nicaragua with the highest Ba/La and the

lowest d

O also had the largest amount of water

estimated for the magma source of 1.7 wt% (with the

exception of the anomalous Irazu

Volcano). Central

Nicaragua is also where the down-going plate was

proposed to be particularly hydrated due to serpenti-

nization at the outer rise.

4. Estimated abundances of volatile elements in parental

magmas correlate with some trace element ratios (Ba/

La, B/La, La/Nb, La/Y), which allow the extensive

data base on whole rock compositions to be used to

assess volatile concentrations in magmas along the

entire volcanic arc. We suggest that the best proxy for

O in magmas of Central America (except central

Costa Rica) is Ba/La but that B/La can also be used as

a proxy for water. Chlorine and to lesser extent sulfur

concentrations correlate with La/Nb ratio, whereas

ﬂuorine concentrations correlate with La/Y ratio.

5. We use concentrations of volatiles in parental melts

and published data on volcanic ﬂuxes to estimate

ﬂuxes of volatile elements for different segments of

and for the entire Central American Arc. Flux of water

was found to be similar throughout the Central

American volcanic arc. Fluxes of other volatiles are

somewhat lower in Nicaragua compared to Guatemala,

El Salvador and Costa Rica. The estimates based on

volcanic ﬂux data can be informative of the relative

magnitude of volatile ﬂuxes along the volcanic front, if

degree of crystallization and amount of intruded

magmas are similar for all segments of the arc. The

volcanic ﬂuxes, however, deﬁne the lower limit of the

true magmatic ﬂuxes. The volatile ﬂuxes estimated for

Nicaragua could be underestimated, if there was a

greater quantity of intrusion to extrusion in Nicaragua

than in the rest of the arc.

6. Comparison of COSPEC data of total SO

ﬂux from

volcanoes with our estimates using melt inclusion

volatile contents and the volcanic ﬂux allows assess-

ment of the total volatile and magmatic ﬂuxes in Central

America, which are *8 times higher compared with the

volcanic ﬂux and volatile ﬂuxes using the volcanic ﬂux.

We estimate that volcanic products comprise *13% of

the total magmatic ﬂux, cumulates *39%, and intru-

sives solidiﬁed in the lithosphere the remaining *48%,

which represent the ‘hidden’ fraction of the total

magmatic ﬂux. Total magmatic and volatile ﬂuxes in

Central America are similar to the independently

estimated global averages.

Contrib Mineral Petrol

123

Acknowledgments We would like to thank W. Strauch, G. Rocha,

G. Alvarado, C. Ramirez, O. Mattius and G. Soto for ﬁeld assistance,

M. Tho

ner and S. Simakin for analytical assistance, and J. Phipps

Morgan, L. Ru

pke, T. Hansteen, M. Carr, K. Garofolo, A. Freundt, S.

Kutterolf, H. Whermann and W. Perez for stimulating discussions on

Central American arc magmatism. We are grateful to D. Hilton, L.

Patino, P. Wallace, T. Plank and M. Carr for providing constructive

and helpful comments on earlier versions of this manuscript. This

work was funded by the Deutsche Forschungsgemeinschaft through

Sonderforschungsbereich 574 ‘‘Volatiles and Fluids in Subduction

Zones’’ to the University of Kiel and the Russian Foundation for

Basic Research (grant 06-05-64873-f to M.P.). This is contribution

number 76 to SFB 574.

References

Abers GA, Plank T, Hacker BR (2003) The wet Nicaraguan slab.

Geophys Res Lett 30(2):1098–1101

Abratis M, Wo

rner G (2001) Ridge collision, slab-window formation,

and the ﬂux of Paciﬁc asthenosphere into the Carribean realm.

Geology 29(2):127–130

Aubouin J, Azema J, Carfantan J-C, Demant A, Rangin C, Tardy M,

Tournon J (1982) The Middle-America Trench in the geologic

framework of Central America. In: Initial reports of the ocean

drilling program, vol 67, pp 747–755

Benjamin ER, Plank T, Wade JA, Kelley KA, Hauri EH, Alvarado

GE (2007) High water contents in basaltic magmas from Irazu

volcano, Costa Rica. J Volcanol Geotherm Res (in press)

Carr MJ, Feigenson MD, Bennett EA (1990) Incompatible element

and isotopic evidence for tectonic control of source mixing and

melt extraction along the Central American arc. Contrib Mineral

Petrol 105:369–380

Carr MJ, Feigenson MD, Patino LC, Walker JA (2003) Volcanism and

geochemistry in Central America: progress and problems. In:

Eiler J (ed) Inside the subduction factory, vol.138, pp 153–174

Carr MJ, Saginor I, Alvarado G, Bolge LL, Lindsay FN, Turrin B,

Feigenson MD, Swisher III CC (2007) Element Fluxes from the

Volcanic Front of Nicaragua and Costa Rica. Geochem Geophys

Geosyst 8:Q06001. doi:10.1029/2006GC001396

Carroll MR, Rutherford MJ (1988) Sulfur speciation in hydrous

experimental glasses of varying oxidation states: results from

measured wavelength shifts of sulfur X-ray. Am Mineral

73:845–849

Chaussidon M, Jambon A (1994) Boron content and isotopic

composition of oceanic basalts: geochemical and cosmochemical

implications. Earth Planet Sci Lett 121:277–291

Crisp JA (1984) Rates of magma emplacement and volcanic output.

J Volcanol Geotherm Res 20:177–211

Danyushevsky L, McNeill AW, Sobolev AV (2002) Experimental

and petrological studies of melt inclusions in phenocrysts from

mantle-derived magmas: an overview of techniques, advantages

and complications. Chem Geol 183:5–24

Duggen S, Portnyagin M, Baker J, Ulfbeck D, Hoernle K, Garbe-

Scho

nberg D, Grassineau N (2007) Drastic shift in lava

geochemistry in the volcanic-front to rear-arc region of the

Southern Kamchatkan subduction zone: evidence for the tran-

sition from slab surface dehydration to sediment melting.

Geochim Cosmochim Acta 71:452–480

Eiler JM, Carr MJ, Reagan M, Stolper E (2005) Oxygen isotope

constraints on the sources of Central American arc lavas. Geochem

Geophys Geosyst 6:Q07007. doi:07010.01029/02004GC000804

Feigenson MD, Carr MJ (1993) The source of Central American

lavas: inferences from geochemical inverse modeling. Contrib

Mineral Petrol 113(2):226–235

Feigenson MD, Carr MJ, Maharaj SV, Juliano S, Bolge LL (2004)

Lead isotope composition of Central American volcanoes:

Inﬂuence of the Galapagos plume. Geochem Geophys Geosyst

5:Q06001. doi:10.1029/2003GC000621

Frische M, Garofalo K, Hansteen TH, Borchers R (2006) Fluxes and

origin of halogenated organic trace gases from Momotombo

volcano (Nicaragua). Geochem Geophys Geosyst 7:Q05020. doi:

10.1029/2005GC001162

Garrido CJ, Lopez Sanchez-Vizcay

no V, Go

mez-Pugnaire MT,

Trommsdorff V, Alard O, Bodinier J-L, Godard M (2005)

Enrichment of HFSE in chlorite-harzburgite produced by high-

pressure dehydration of antigorite-serpentinite: implications for

subduction magmatism. Geochem Geophys Geosyst 6:Q01J15.

doi:10.1029/2004GC000791

Goss AR, Kay SM (2006) Steep REE patterns and enriched Pb

isotopes in southern Central American arc magmas: evidence for

forearc subduction erosion?. Geochem Geophys Geosyst

7:Q05016. doi:10.1029/2005GC001163

Harris DM, Anderson AT (1984) Volatiles H

O, CO

and Cl in a

subduction related basalt. Contrib Mineral Petrol 87:120–128

Hilton DR, Fischer TP, Marty B (2002) Noble gases and volatile

recycling in subduction zones. In: Porcelli D, Ballentine C,

Weiler R (eds) Noble gases in geochemistry and cosmochem-

istry, reviews in mineralogy and geochemistry, vol 47.

Mineralogical Society of America, Washington, DC, pp 319–370

Hoernle K, Werner R, Phipps Morgan J, Garbe-Schonberg D, Bryce J,

Mrazek J (2000) Existence of complex spatial zonation in the

Gala

pagos plume. Geology 28(5):435–438

Hofmann AW (1988) Chemical differentiation of the Earth: the

relationship between mantle, continental crust and oceanic crust.

Earth Planet Sci Lett 90:297–314

Hofmann AW (2003) Sampling mantle heterogeneity through oceanic

basalts: isotopes and trace elements. In: Treatise on geochem-

istry, vol 2. Elsevier, Amsterdam, pp 61–101

Ionov DA, Hofmann AW (1995) Nb-Ta-Rich mantle amphiboles and

micas—implications for subduction-related metasomatic trace-

element fractionations. Earth Planet Sci Lett 131(3–4):341–356

Jarosewich EJ, Nelen JA, Norberg JA (1980) Reference samples for

electron microprobe analysis. Geostand Newslett 4:43–47

Jarrard RD (2003) Subduction ﬂuxes of water, carbon dioxide,

chlorine, and potassium. Geochem Geophys Geosyst 4(5):8905

doi:10.1029/2002GC000392

Katz RF, Spiegelman M, Lagmuir CH (2003) A new parameterization

of hydrous mantle melting. Geochem Geophys Geosyst

4(9):1073. doi:1010.1029/2002GC000433

Kelley K, Plank T, Grove TL, Stolper EM, Newman S, Hauri E

(2006) Mantle melting as a function of water content beneath

back-arc basins. J Geophys Res 111:B09208. doi:

10.1029/2005JB003732

Langmuir CH, Klein EM, Plank T (1992) Petrological systematics of

mid-ocean ridge basalts: constraints on melt generation beneath

ocean ridges. In: Phipps Morgan J, Blackman DK, Sinton JM

(eds) Mantle ﬂow and melt generation at mid-ocean ridges, vol

71. AGU, Washington, DC, pp 183–280

Leeman WP, Carr MJ, Morris JD (1994) Boron geochemistry of the

Central American Volcanic Arc: constraints on the genesis of

subduction-related magmas. Geochim Cosmochim Acta

58(1):149–168

Lundstrom CC, Hoernle K, Gill J (2003) U-series disequilibria in

volcanic rocks from the Canary Islands: plume versus litho-

spheric melting. Geochim Cosmochim Acta 67(21):4153–4177

McKenzie D, Bickle MJ (1988) The volume and composition of melt

generated by extension of the lithosphere. J Petrol 29:625–679

Morris JD, Leeman WP, Tera F (1990) The subducted component in

island arc lavas: constraints from Be isotopes and B-Be

systematics. Nature 344:31–36

Contrib Mineral Petrol

123

Patino LC, Carr MJ, Feigenson MD (2000) Local and regional

variations in Central American arc lavas controlled by variations

in subducted sediment input. Contrib Mineral Petrol 138:265–283

Plank T, Langmuir CH (1998) The chemical composition of

subducting sediment and its consequences for the crust and

mantle. Chem Geol 143:325–394

Portnyagin MV, Simakin SG, Sobolev AV (2002) Fluorine in

primitive magmas of the Troodos Ophiolite Complex, Cyprus:

analytical methods and main results. Geochem Int 40(7):625–

632

Portnyagin MV, Hoernle K, Plechov PY, Mironov NL, Khubunaya

SA (2007) Constraints on mantle melting and composition and

nature of slab components in volcanic arcs from volatiles (H

S, Cl, F) and trace elements in melt inclusions from the

Kamchatka Arc. Earth Planet Sci Lett 255(1–2):53–59

Protti M, Gu

ndel F, McNally K (1995) Correlation between the age of

the subducting Cocos plate and the geometry of the Wadati-

Benioff zone under Nicaragua and Costa Rica. In: Mann P (ed)

Geologic and tectonic development of the Caribbean plate

boundary in southern Central America, vol 295, pp 309–326

Ranero C, Phipps Morgan J, McIntosh K, Reichert C (2003) Bending-

related faulting and mantle serpentinization at the Middle

America trench. Nature 425:367–373

Roggensack K (2001a) Sizing up crystals and their melt inclusions: a

new approach to crystallization studies. Earth Planet Sci Lett

187(1–2):221–237

Roggensack K (2001b) Unraveling the 1974 eruption of Fuego

volcano (Guatemala) with small crystals and their young melt

inclusions. Geology 29(10):911–914

Roggensack K, Hervig RL, McKnight SB, Williams SN (1997)

Explosive basalic volcanism from Cerro Negro volcano: inﬂu-

ence of volatiles on eruptive style. Science 277(9):1639–1642

pke L, Phipps Morgan J, Hort M, Connolly JAD (2002) Are the

regional variations in Central American arc lavas due to differing

basaltic versus peridotitic slab sources of ﬂuids? Geology

30(11):1035–1038

pke LH, Phipps Morgan J, Hort M, Connolly JAD (2004)

Serpentine and the subduction zone water cycle. Earth Planet

Sci Lett 223(1–2):17–34

Salters VJM, Stracke A (2004) Composition of the depleted mantle.

Geochem Geophys Geosyst 5(5):Q05004. doi:

05010.01029/02003GC000597

Sisson TW, Layne GD (1993) H

O in basalt and basaltic andesite

glass inclusions from 4 subduction-related volcanoes. Earth

Planet Sci Lett 117(3–4):619–635

Sobolev AV, Chaussidon M (1996) H

O concentrations in primary

melts from island arcs and mid-ocean ridges: implications for

O storage and recycling in the mantle. Earth Planet Sci Lett

137:45–55

Sobolev AV, Hofmann AW, Kuzmin DV, Yaxley GM, Arndt NT,

Chung S-L, Danyushevsky LV, Elliott T, Frey FA, Garcia MO,

Gurenko AA, Kamenetsky VS, Kerr AC, Krivolutskaya NA,

Matvienkov VV, Nikogosian IK, Rocholl A, Sigurdsson IA,

Sushchevskaya NM, Teklay M (2007) The amount of recycled

crust in sources of mantle-derived melts. Science 316:412–417

Stolper E, Newman S (1994) The role of water in the petrogenesis of

Mariana Trough magmas. Earth Planet Sci Lett 121(3–4):293–

325

Sun S-S, McDonough WF (1989) Chemical and isotopic systematics

of oceanic basalts: implications for mantle composition and

processes. In: Saunders AD, Norry MJ (eds) Magmatism in the

ocean basins, vol 42. Geological Society Special Publication,

London, pp 313–345

Syracuse EM, Abers GA (2006) Global compilation of variations in

slab depth beneath arc volcanoes and implications. Geochem

Geophys Geosyst 7:Q05017. doi:10.1029/2005GC001045

Wade JA, Plank T, Melson WG, Soto GJ, Hauri EH (2006) The

volatile content of magmas from Arenal volcano, Costa Rica. J

Volcanol Geotherm Res 157(1–3):94–120

Walker JA, Carr MJ, Feigenson MD, Kalamarides RI (1990) The

petrogenetic signiﬁcance of interstratiﬁed high- and low-Ti

basalts in Central Nicaragua. J Petrol 31(5):1141–1164

Walker JA, Roggensack K, Patino LC, Cameron BI, Matias O (2003)

The water and trace element contents of melt inclusions across

an active subduction zone. Contrib Mineral Petrol 146:62–77

Wallace PJ (2005) Volatiles in subduction zone magmas: concentra-

tions and ﬂuxes based on melt inclusion and volcanic gas data. J

Volcanol Geotherm Res 140(1–3):217–240

Werner R, Hoernle K, van den Bogaard P, Ranero C, von Huene R,

Korich D (1999) Drowned 14-m.y.-old Gala

pagos archipelago

off the coast of Costa Rica: implications for tectonic and

evolutionary models. Geology 27(6):499–502

Werner R, Hoernle K, Barkckhausen U, Hauff F (2003) Geodynamic

evolution of the Gala

pagos hot spot system (Central East Paciﬁc)

over the past 20 m.y.: Constraints from morphology, geochem-

istry, and magnetic anomalies. Geochem Geophys Geosyst

4(12):1108. doi:10.1029/2003GC000576

Yaxley GM, Green DH (1998) Reactions between eclogite and

peridotite: mantle refertilisation by subduction of oceanic crust.

Scweiz Miner Petrogr Mitt 78:243–255

Zimmer MM, Fischer TP, Hilton DR, Alvarado GE, Sharp ZD, Walker

JA (2004) Nitrogen systematics and gas ﬂuxes of subduction

zones: Insights from Costa Rica arc volatiles. Geochem.

Geophys. Geosyst. 5:Q05J11 doi:10.1029/2003GC000651

Contrib Mineral Petrol

123