Origin of silicic volcanism in the Panamanian arc: evidence for a two-stage fractionation process at El Valle volcano

Abstract

In the Central American Volcanic Arc, adakite-like volcanism has often been described as volumetrically insignificant. However, extensive silicic adakitic volcanism does occur in the Panamanian arc and provides an opportunity to evaluate the origin of this magma-type as well as to contrast its origin with other Central American silicic magmas. The Quaternary volcanic deposits of El Valle volcano are characterized by pronounced depletions in the heavy rare earth elements, low Y, high Sr, high Sr/Y, relatively high MgO, and low K₂O/Na₂O, when compared with other Quaternary Central American volcanics at similar SiO₂. These chemical features are also diagnostic of adakitic signatures. Our new ⁴⁰Ar/³⁹Ar ages of lava flows and ash flows that compose the volcanic edifice of El Valle volcano illustrate that the eruptive volume of adakitic-like volcanism is substantial during the Quaternary (~120 km³). Adakitic-like magmas dominate the stratigraphic record. Common to all models for the origin of an adakite geochemical signature is the involvement of garnet, as a residual or fractionating phase. The stability of garnet in hydrous magmas has been recently reevaluated with important consequences; garnet is a stable primary igneous phase at pressure and temperature conditions expected for magma differentiation at the roots of a mature island arc. Moreover, adakite-like volcanism erupted at El Valle volcano displays the middle rare earth element depletion observed in other Panamanian volcanic centers that has been attributed to significant amphibole fractionation. Extensive amphibole fractionation may have occurred in two stages. The first stage of fractionation, garnet + amphibole fractionation, occurs from hydrous basaltic–andesitic parental magmas that have ponded at the base of an overthickened crust. The second stage occurs at mid-lower crustal levels where abundant amphibole + plagioclase and minor sphene crystallized from water-rich magmas. These two stages combined may have resulted in an amphibole-rich cumulate layer. This amphibole layer is likely the source of the abundant amphibole-rich cumulate enclaves and blobs found in volcanic products across the Panamanian arc. Stalling of water-rich magmas during this two-stage fractionation process could drive the interstitial liquids to the evolved compositions typical of continental crust, while leaving behind amphibole-rich cumulate rocks that may eventually be returned to the asthenosphere. Differentiation of H₂O-rich magmas under the conditions appropriate for the roots of island arcs may therefore be a key process in developing a better understanding of the generation of continental crust in island arc environments.

Introduction

This research is part of a series of studies in the Central American Volcanic Arc (CAVA) that focus on the production of silicic volcanism in arc settings that lack an evolved continental crust (Hannah et al. 2002; Vogel et al. 2004, 2006; Deering et al. 2007). This region is of importance because it has developed some of the same geochemical characteristics of continental crust (Vogel et al. 2004, 2006). In this study, we explore magma differentiation processes responsible for the volumetric adakitic-like magmas of El Valle volcano.

The Quaternary volcanic deposits of El Valle volcano are characterized by pronounced depletions in the heavy rare earth elements (HREE), low Y, high Sr, high Sr/Y, relatively high MgO, and low K₂O/Na₂O (Tables 1 and 2); when compared to other Quaternary Central American volcanics at similar SiO₂ concentrations. These chemical features are also typical of adakite-like magmas and high magnesium andesites alike. The term high magnesium andesite (HMA) has been used to describe a particular group of rocks with higher Mg# (>50), Ni and Cr than typical adakites (Tatsumi 1982; Kelemen and Dunn 1992; Yogodzinski and Kelemen 1998). The use of the terms adakite and high magnesium andesite is highly controversial as their initial usage had petrogenetic significance. These types of rocks were initially thought to have been derived from direct melting of the eclogitized subducting oceanic crust (not the mantle wedge, as is typical) and are observed only infrequently in subduction settings (Kay 1978; Defant 1990; Yogodzinski and Kelemen 1998; Castillo et al. 1999; Falloon et al. 2008). In this study, we use the term adakite-like to refer to their unusual geochemical characteristics (e.g. HREE depletion) without prejudice as to the mechanism of origin for these magmas.

Table 1 Average major and trace element composition and standard deviations for El Hato ash-flow pumices

Table 2 Average and standard deviations for major and trace element analyses for the dacitic lavas and dacitic domes units

There is a general consensus that the presence of garnet is required at some stage in the petrogenesis of adakite-like signatures, either as a residual phase in the source or as an early crystallizing phenocryst phase, to account for their depleted heavy rare earth element (HREE) concentrations (Hildreth and Moorbath 1988; Defant et al. 1991a, b; de Boer et al. 1995; Castillo 2006; Macpherson 2008; Alonso-Perez et al. 2009). Because garnet-melt equilibrium can occur in a broad spectrum of conditions, several models have been proposed for the origin of adakitic magmas. Experimental studies that involved melting of basaltic material at high pressures (>1.0 GPa) have produced silicic adakite-like material at low degrees of melting (Rapp and Watson 1995; Rapp et al. 2003, 2006). In addition, the role of primary igneous garnet fractionation has been highlighted in the genesis of adakitic magmas (Castillo et al. 1999; Prouteau and Scaillet 2003; Dreher et al. 2005; Rodriguez et al. 2007). Experimental work on realistic hydrous arc magmas compositions suggests that garnet is a stable primary igneous phase at pressure and temperature conditions expected for magma differentiation in the crust or crust–mantle interface of a mature island arc (Müntener et al. 2001; Müntener and Ulmer 2006; Alonso-Perez et al. 2009). Based on these and other experimental results, a range of models have been developed to explain the occurrence of adakite-like volcanism at various locations. Some of these models involve magma produced by melting of the subducted slab in subduction zones (Defant et al. 1991a, b), partial melting of underplated basaltic lower crust (Rapp et al. 2002) or delaminated lower crust (Xu et al. 2002; Macpherson et al. 2006; Macpherson 2008), and melting/crystal fractionation at the base of a tectonically thickened crust (Atherton and Petford 1993; Kay and Kay 2002; Chiaradia et al. 2009).

In the present study, we address the importance of middle to lower crustal differentiation processes and in particular explore the role of garnet and amphibole fractionation in the generation of the adakite signature observed in Quaternary volcanic products at El Valle volcano. These differentiation processes have been described by Annen et al. (2006) to occur in crustal reservoirs or “hot zones” in which mantle-derived magmas may be modified. Recent studies have supported these findings and have illustrated the importance of these lower crustal “hot zones” in the production and evolution of arc magmas (Deering et al. 2008, 2010). These studies have also highlighted the wide spectrum of magmatic products that may be derived from variations in pH₂O, fO₂, and T during dry to wet magma fractionation.

On rare occasions, the processes active within these crustal “hot zones” have been examined either through the study of exhumed arc terranes or magmatic cumulates enclaves extracted from deep crustal levels. In exhumed arc terranes, the occurrence of anhydrous cumulate enclaves that represent a gabbroic assemblage (e.g. olivine, plagioclase, and pyroxene) is widespread (Debari and Coleman 1989; Jagoutz et al. 2006; Larocque and Canil 2010). Nonetheless, hydrous cumulate enclaves have also been identified, where intercumulus amphibole is generated by progressively decreasing temperatures along with elevating water content in evolving magmas (Arculus and Wills 1980; Conrad and Kay 1984; Beard 1986; Costa et al. 2002; Claeson and Meurer 2004). Experimental studies of water-rich magmas have revealed that water-rich magmatic systems will also produce amphibole as a primary cumulus phase, generating amphibole-rich assemblages (Moore and Carmichael 1998; Carmichael 2002; Barclay and Carmichael 2004). These findings have led to the reinterpretation of amphibole-rich sections exposed in arc terranes in Pakistan, China, Antarctica, and Canada (Jagoutz et al. 2006; Sun and Zhou 2008; Tiepolo and Tribuzio 2008; Larocque and Canil 2010). In the Panamanian arc, the importance of an amphibole-rich zone as a site for fractionating middle rare earth element (MREE) is enhanced by the ubiquitous presence of amphibole-rich cumulates as enclaves or blobs in volcanic sequences from the Oligocene to the most recent Quaternary volcanism (Fig. 1) (Hidalgo 2007; Hidalgo and Rooney 2010; Rooney et al. 2011).

Fig. 1

Examples of cumulate textures observed in cumulate enclaves and pods within lavas and pumice fragments representative of the last period of activity at El Valle volcano. a Cross-polarized light micrograph, Iguana unit. Orthopyroxene + olivine cumulate. Typically, these cumulate enclaves are up to 5 cm in diameter and are generally surrounded by an amphibole and minor clinopyroxene crown. b Plain polarized light micrograph. Upper part displays part of a coarse-grained cumulate included in the dacitic lavas of the El Valle volcano summit. Cumulate enclaves in these lavas contain a larger proportion of Fe–Ti oxides (up to 15%) in comparison with the enclaves observed in pumice fragments. (Amph amphibole, Pl plagioclase). Amphibole is dominant, while some plagioclase has crystallized on the interstices of the crystal network as a late-stage product. c Cross-polarized light micrograph, Iguana unit. Olivine crystals surrounded by amphibole + orthopyroxene crown. d Plane polarized light micrograph, domes unit zoned amphibole. Some amphiboles from the Quaternary unit have high Al# cores. These are described in detail in section “Amphibole compositions and geothermobarometry”

In this study for central Panama, we suggest a two-stage process for the origin of the El Valle volcano recent silicic deposits: (1) garnet + amphibole fractionation from parental hydrous magmas that have ponded at the base of the crust (2) amphibole + plagioclase + sphene fractionation at middle to lower crustal levels, resulting in amphibole-rich layers. Our study seeks to improve our understanding of how adakite-like geochemical signals may be produced, in addition to probing the role of amphibole fractionation in arc magmas from oceanic arcs that lack an evolved thick sialic crust. This study may have broad implications in our understanding of the evolution and growth of the continental crust from oceanic arcs (Kelemen 1995; Smithies 2000; Condie 2005).

Tectonic setting and nature of the subducting crust

The tectonic region that encompasses isthmus of Central America is one of the most complex active tectonic areas of the western hemisphere, with the interaction of five tectonic plates: North American, Caribbean, South American, Cocos, and Nazca. The current plate configuration is derived from the ~23 M.a. (Lonsdale 2005) split of the Farallon plate into the Nazca and Cocos plates (online resource 1). This event resulted in important consequences for the tectonic and volcanic regimes in the region including change in direction and acceleration of spreading at the East Pacific Rise (Goff and Cochran 1996; Wilson 1996), and faster, less oblique subduction at the western margins of the Central American trenches (Sempere et al. 1990).

The Cocos plate is subducting underneath the Caribbean plate normal to the southeastward trending Middle America Trench (M. A. Trench) at rates of 76–91 mm/year relative to the Caribbean plate (DeMets 2001; Trenkamp et al. 2002) (online resource 1). This plate is bounded by the Nazca plate to the east. This boundary is well constrained and is best characterized by the north trending and seismically active right-lateral Panama fracture zone that includes the Coiba and Balboa fracture zones. The limit between the Nazca plate and the Southern Panama Deformed Belt (equivalent to the M. A. Trench in this region) is complex; it changes character eastward from oblique subduction (between 83°W and 80.5°W) of the Nazca plate (V = 5 cm/year, Jarrard 1986; Trenkamp et al. 2002) to a sinistral strike-slip fault (from 80°W to 78.8°W; Westbrook et al. 1995). The transition from oblique subduction to strike-slip fault is a direct consequence of the still active southeastern migration (~35 mm/year) of the Panama fracture zone system, and it has resulted in the cessation of subduction along the southern margin of Panama (from 80°W to 78.8°W) in the Late Miocene period (Lonsdale and Klitgord 1978). In northern Panama and Costa Rica, the Caribbean plate is underthrusting (~10 mm/year) Central America along the North Panama Deformed Belt (Kellogg et al. 1995; Mann and Kolarsky 1995).

The material entering the Middle America Trench has a significant impact on the composition of volcanism in the CAVA and has been derived from oceanic spreading centers of various ages on the Cocos and Nazca plates (Barckhausen et al. 2001). The balance of the oceanic crust comprising the Cocos plate originated at Cocos-Nazca spreading centers that had a protracted history with an early spreading center (CN-1; 23 M.a.) and the later CN-2 (19–15 M.a.) (Barckhausen et al. 2001). Built upon this oceanic lithosphere is the aseismic Cocos ridge (13–22 M.a.) that was produced by the Galapagos hot spot. In contrast to the well-studied Cocos plate, details of the age and origin of the western section of the Nazca plate remain less well constrained. This problem is particularly severe for the Nazca crust located between the Panama fracture zone and the Coiba Ridge where the only age constraint is from 15 M.a. sediments (van Andel et al. 1971), though the crust in this region may be as old as 22 M.a. (Lonsdale 2005). Further west, between the 80°W fracture Zone and the South American margin, the Nazca oceanic crust is derived from the now extinct Sandra Rift and is between 9 and 13 M.a. (Lonsdale 2005). Upon this crust lies the 150 km long by 100 km wide Coiba plateau, which rises 1 km above the surrounding crust. The Coiba plateau is interpreted to be the earliest part of the Galapagos hot spot track (Werner et al. 2003).

Stratigraphy and petrography

Stratigraphy

The regional basement in western Panama is composed of oceanic assemblages (21–71 M.a. (Hoernle et al. 2002) that are part of the Caribbean Large Igneous Province (CLIP) and of the Galapagos hot spot track and overlying sediments (Fig. 2). These assemblages are referred to as the Azuero-Soná Complex although they have been mapped under different names (e.g. Sona, Playa Venado, Quebro, Tiribique, Lovaina) (Denyer et al. 2006). The CLIP is observed in Panama within the peninsulas located in the Pacific coast and in various islands outboard southwest Panama (Denyer et al. 2006). The thickness of this CLIP basement unit is unknown, but by correlation to oceanic complexes of the CLIP in the Costa Rican forearc, it can be estimated at ~25 km (MacKenzie et al. 2008; Linkimer et al. 2010).

Fig. 2

Stratigraphy of El Valle Volcano and surrounding areas

El Valle lies at the easternmost extent of the CAVA in Panama (de Boer et al. 1991; Defant et al. 1991a; Sherrod et al. 2007). The volcano includes a 30-km² caldera at its summit where the town of El Valle is located. The caldera floor contains lake deposits and is bounded by steep 200- to 300-m-high walls. A dome complex forms the northern boundary of the caldera, and it was constructed along an E–W trending lineament, which is the highest point of the volcano (1,185 m).

The characterization of the stratigraphic sequence at El Valle volcano was greatly facilitated by our new ⁴⁰Ar/³⁹Ar dating of volcanic deposits (see “Results”), which allowed for the distinction between two periods of volcanic activity. From ~10 to ~5 M.a. (Table 3), volcanic arc andesites were emplaced with similar characteristics to most of the present-day lavas of the Central America Volcanic Front (Hidalgo 2007). In this study, we have calculated an ⁴⁰Ar/³⁹Ar age for these andesites of 5.138 ± 34 k.a. (~5 M.a.). These andesites have also been studied by Defant et al. (1991a) and were referred to by these authors as the “Old group”. In this study, a time of emplacement between ~5 and 10 M.a was determined. These dates are consistent with our ⁴⁰Ar/³⁹Ar age determinations. After ~5 M.a. of quiescence, the volcanic activity on El Valle Volcano resumed with the eruption of dacitic lavas. These dacitic lavas form the peaks on the north flank of the caldera (dome unit) and were erupted at 109 ± 7 k.a. (see “Results”). The silicic volcanism continued with the emplacement of the dacitic Lava flows unit in the Guacamaya and Iguana (85 ± 20 k.a.) volcanic centers. Later, volcanic activity shifted east to India Dormida (56 ± 14 k.a.). Volcanism at the Guacamaya, Iguana and India Dormida volcanic centers has been grouped in Fig. 2 in a single unit, the Iguana unit. Volcanic activity reached its climax with eruption of El Hato unit, which was deposited overlying the India Dormida dacitic lavas. Previous K–Ar studies in the area by Defant et al. (1991a) assign these units older ages (Young group, ~1.55 to ~0.2 M.a.) than our ⁴⁰Ar/³⁹Ar determinations. Given the advantages of the ⁴⁰Ar/³⁹Ar method over the K–Ar method, from this point on we will use only our ⁴⁰Ar/³⁹Ar-determined ages. The El Hato unit is an extensive silicic ignimbrite sheet (approximately 300 km²), which resulted in a caldera collapse event creating a ~30-km² depression (Hidalgo 2007). The thickness of this unit is variable but reaches more than a hundred meters in some locations (Hidalgo 2007).

Table 3 Summary of whole-rock ⁴⁰Ar/³⁶Ar analyses

After the caldera collapsed, more than 90 m of lake sediments (El Valle unit) were deposited between 31,850 years ±1,800 and 2,370 years ±150 (Bush and Colinvaux 1990). Soon after, the lake may have drained through a breach in the southwest caldera wall, where the main outlet of the modern surface water is located. Alternatively, volcanic deposits may have completely filled this local basin (Bush and Colinvaux 1990). These events make it possible to estimate the age of the El Hato ignimbrite to be between ~56 k.a. (age of underlying unit) and ~31.8 k.a. However, it is likely that the caldera-forming event occurred as a consequence of the ignimbrite eruption and hence the age for the El Hato unit might be closer to the age of the oldest lacustrine sediments in the caldera (~31.8 k.a.). The local stratigraphic sequence culminates with emplacement of fan deposits that are observed to overlie the lacustrine sediments in the caldera and extend to the Pacific coast where they typically outcrop in benches and beach terraces. The age for this unit is estimated in <2,370 k.a.

Petrography

The present study focuses on the Quaternary units and their chemical variation; hence, these petrographic descriptions will only include units that were erupted during the Quaternary. All of the Quaternary units contain abundant amphibole cumulates as enclaves or inclusions (blobs) in the lavas or in some pumice fragments (Fig. 1a, b). Table 4 summarizes the petrographic results.

Table 4 Petrographic summary giving averages of crystal proportions and size of crystals for El Hato, dacitic lavas, and Llano Tigre andesites

Lava samples from the domes unit have 21–25% crystallinity and hypocrystalline texture. Primary mineralogy consists of plagioclase, amphibole, quartz, biotite, Fe–Ti oxides, and sphene. Plagioclase crystals display discontinuous zoning (online resource 2), sieve textures, and reaction at the core (online resource 2), which are consistent with crystal–melt disequilibrium. Some plagioclase crystals have nucleated on resorbed small quartz and amphibole crystals (~0.3 mm), and in some cases crystallites of amphibole are observed in the plagioclase cores. Some amphiboles with orthopyroxene cores are present (online resource 2). However, in some occurrences, the orthopyroxenes are entirely consumed and are only present as ghost crystals. Rare olivine crystals and olivine ghosts are observed and are always surrounded by amphibole + pyroxene reaction rims (Fig. 1a, c). In addition, most of the samples contain embayed quartz grains and opacite rims on amphibole crystals. Both of these features support some degree of disequilibrium.

The Iguana unit lavas have similar crystallinity (17–26%) to the domes unit (Table 4). The mineralogy is dominated by plagioclase and zoned amphiboles (Fig. 1d) with small amounts of Fe–Ti oxides and sphene. Olivine crystals that are surrounded by amphibole + pyroxene crowns are also observed in this unit (Fig. 1c). Similar to the other Quaternary units, this unit also contains a considerable amount of amphibole-rich cumulates (amphibole gabbros) as enclaves or blobs. Zoned plagioclase and amphibole crystals are also common (Fig. 1b). The El Hato unit has the same mineralogy and mineral relationships observed in the other units including amphibole-rich cumulates enclaves.

The ubiquitous cumulate enclaves observed in all Quaternary units could be organized in two groups: opx + ol enclaves (Fig. 1a) and amphibole-dominated enclaves (Fig. 1b). The ol + opx cumulates enclaves are less abundant and generally form nodules (~3 cm) that are composed of variable proportions of pyroxene and olivine. The larger-size amphibole-rich cumulate enclaves nodules (5–12 cm) are more complex and include zoned amphibole megacrysts (3–5 mm) with minor (<2%) interstitial glass. In some cases, these enclaves contain anhedral plagioclase (up to 20%) and apatite phenocrysts (~0.5 mm) in the interstices.

Results

Analytical methods

Major and trace element analyses were determined using X-ray fluorescence (XRF) and laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) analyses at Michigan State University on fused glass disks from over a 100 samples. The procedure followed the outline described by Hannah et al. (2002) and Deering et al. (2008) for the preparation of low-dilution fusion glass disks (LDF). The fused disks were analyzed using a Bruker Pioneer S4 X-ray fluorescent spectrograph. XRF element analyses were reduced using Bruker Spectra Plus software. For LA-ICP-MS trace element analyses, a Cetac+ LSX200 laser ablation system coupled with a Micromass Platform ICP-MS was used. Procedures for XRF and LA-ICP-MS analyses followed those outlined by Deering (2009). Precision and accuracy of both the XRF and LA-ICP-MS chemical analyses have been reported in Vogel et al. (2006). Trace element reproducibility based on standard analyses is typically better than 5%.

For age determination, seven samples were handpicked for datable mineral phases (feldspars) and groundmass and were submitted to the Geochronology laboratory at the University of Alaska Fairbanks (UAF) for ⁴⁰Ar/³⁹Ar analysis. The mineral TCR-2 with an age of 27.87 M.a. (Lanphere and Dalrymple 2000) was used to monitor neutron flux and calculate the irradiation parameter, J, for all samples. The samples and standards were wrapped in aluminum foil and loaded into aluminum cans of 2.5 cm diameter and 6 cm height. All samples were irradiated in position 5c of the uranium-enriched research reactor of McMaster University in Hamilton, ON, Canada, for 0.75 megawatt-hours. Although the standard is significantly older than the samples, the irradiation time was chosen to ensure (1) adequate production of ³⁹Ar in the standard (⁴⁰Ar*/³⁹Ar_K of the standard is ~180) to be precisely measured and (2) that the samples were not over irradiated to the point where reactor-induced interferences from calcium would be a problem (⁴⁰Ar*/³⁹Ar_K of the samples is >0.5, and Ca/K < ~1.5). Upon their return from the reactor, the samples and monitors were loaded into 2-mm diameter holes in a copper tray that was then loaded in an ultra-high vacuum extraction line. The monitors were fused, and samples heated, using a 6-watt argon-ion laser following the technique described by York et al. (1981), Layer et al. (1987), and Layer (2000). Argon purification was achieved using a liquid nitrogen cold trap and an SAES Zr–Al getter at 400°C. The samples were analyzed in a VG-3600 mass spectrometer at the Geophysical Institute, University of Alaska Fairbanks. The argon isotopes measured were corrected for system blank and mass discrimination, as well as calcium, potassium, and chlorine interference reactions following procedures outlined in McDougall and Harrison (1999). System blanks generally were 2 × 10⁻¹⁶ mol ⁴⁰Ar and 2 × 10⁻¹⁸ mol ³⁶Ar, which are 10–50 times smaller than fraction volumes. Mass discrimination was monitored by running both calibrated air shots and a zero-age glass sample. These measurements were made on a weekly to monthly basis to check for changes in mass discrimination.

Amphibole compositions were determined using a Cameca SX100 electron microprobe at the University of Michigan equipped with five-wavelength spectrometers, using an accelerating potential of 15 kV, a focused beam with a 10 μm spot size, counting time of ~3 min per mineral, and a 10 nA beam current. Standards used were natural fluor-topaz (FTOP), natural jadeite (JD-1), natural grossular, Quebec (GROS), natural adularia, St. Gothard, Switzerland (GKFS), synthetic apatite (BACL), synthetic Cr₂O₃, and synthetic FeSiO₃ (FESI). Amphiboles typically average ~2 wt% of (H₂O + F + Cl); therefore, only analyses with anhydrous totals (SiO₂, Al₂O₃, FeO_tot, MgO, CaO, Na₂O, K₂O, TiO₂, MnO) of 98 ± 1 wt% were retained.

⁴⁰Ar/³⁹Ar geochronology

A summary of the calculated dates is provided in Table 3 and Fig. 3; and a full data table is provided in the online resource 3. All ages are quoted to the ±1 − σ level and calculated using the constants of Steiger and Jaeger (1977). Only the Llano Tigre andesites (5,138 ± 34 k.a.), dacitic flow unit (85 ± 20 to 56 ± 14 k.a.), and dome (109 ± 7 k.a.) units were processed using whole-rock samples, while age calculations on El Hato unit were done using single feldspar crystals. Geologically meaningful plateau (different shades of gray in Fig. 3a, c, and e indicate individual plateaus; Table 3) and isochron ages were obtained from the dacite and dome whole-rock samples while the andesitic lava yielded a well-defined plateau age. High Ca/K fractions were excluded from plateau calculations, as these were likely due to the degassing of feldspar phenocrysts. Isochron analyses from three samples indicate the presence of slight amounts of excess argon. Isochron ages are slightly different from the plateau ages (which might be biased by excess argon), but the plateau and isochron ages are generally within 2-sigma of one another. For the dome and flow units, composite isochrons were calculated using fractions from each run. Because of the prospect of excess argon, we interpret the isochron ages to better reflect the eruption ages of these samples.

Fig. 3

⁴⁰Ar/³⁹ age spectra (a, c, e) and isochron plots (b, d, f) for dome and Iguana units. Composite isochron ages based on 2 experiments (runs) for sample CG-5-24-06-03 (4A, lava domes from Quaternary group) and 3 and 2 runs for samples of the Iguana unit, ID-5-22-06-04 and sample IG-5-28-06-02 (c, e), respectively. On the isochron plots, “e” denotes fractions excluded from the calculation of the isochron (see also Table 3). g, h ⁴⁰Ar/³⁹ age and Ca/K spectra and isochron plot for the Llano Tigre andesite lava. No isochron age could be determined for this sample. Plateau age calculated for the lowest Ca/K fractions shown in gray (see also Table 3). Full data are available in the online resource 3

The results for the El Hato unit are not included because the calculated ages did not produce meaningful geological ages. Nonetheless, field observations are consistent with El Hato unit overlying the youngest part of the dacitic flow unit (~56 k.a.) and dome units (~109 k.a.). In that way, the age of El Hato Unit would be estimated as younger than 56 ± 14 k.a. (dacitic flow unit) and older than the overlying 31.8 k.a. lacustrine sediments date by Bush and Colinvaux (1990).

Major and trace elements

Chemical variation at El Valle

A summary of the chemical variation observed at El Valle volcano is presented in Tables 1 and 2. Complete analyses for all samples presented in this study are found in the online resource 4. Quaternary samples from El Valle volcano are dacites with little variation in chemical composition (e.g. 66.90–69.97% SiO₂) (online resource 4 and Fig. 4). Both the pumice fragments and lava samples have similar compositions. The lack of compositional variation is striking when compared with other CAVA ignimbrites, which often contain large compositional variations (e.g. 60.11–74.30% SiO₂ for the Costa Rican ignimbrites, Fig. 4). For that reason, the diagrams group the Quaternary magmatism at El Valle volcano in a single unit.

Fig. 4

Bulk rock geochemistry—major element variation versus SiO₂ diagrams shown. Complete composition and standard deviations for major and trace elements for the analysis of glassy pumice fragments and lava samples are provided in the online resource 4. Average compositions are presented in Tables 1 and 2. Samples with totals lower than 96% are excluded because it was assumed that these samples were secondarily hydrated or altered. All major elements have been normalized to 100%. El Valle Quaternary products contain little variation when compared with other Quaternary ignimbrites of the CAVA (Vogel et al. 2006). In addition, TiO₂, Fe₂O₃, and MnO are more depleted; Al₂O₃ and CaO are more enriched, while MgO, Na₂O, and K₂O have similar range than other CAVA ignimbrites

Comparison with other Central American ignimbrites

The chemical signatures of the Quaternary products of El Valle volcano are clearly distinct from other ignimbrites from the CAVA. In comparing the chemical compositions of the El Valle Quaternary volcanic deposits with other ignimbrites within the CAVA, at similar SiO₂ concentrations, El Valle deposits contain elevated concentrations of Al₂O₃, MgO, Na₂O, and CaO and lower concentrations of TiO₂, Fe₂O_3(t) (Fig. 4).

Similar to the major elements, the trace element variations in the Quaternary eruptive products from El Valle volcano are small when compared with other CAVA ignimbrites (online resource 5). Spider and REE diagrams and selected trace element ratios (Figs. 5, 6, 7) could be used to illustrate the differences in chemistry between the CAVA ignimbrites and the El Valle Quaternary units. In Fig. 5, the samples chosen to represent the CAVA dataset were selected to have the similar limited SiO₂ range (66–70%) observed in the El Valle Quaternary suite. Generally, samples from El Valle volcano contain extreme depletions in HFSE (high-field-strength elements), in particular the HREE (Fig. 5). Large-ion lithophile elements (LILE e.g. Ba, Rb, Pb) are also depleted with the exception of Sr. It is worth noting that the Sr content in El Valle Quaternary volcanic rocks is high (>600 ppm) compared with other CAVA ignimbrites.

Fig. 5

Chondrite- and primitive mantle-normalized diagrams (Sun and McDonough 1989) for El Valle Volcano Quaternary deposits, ignimbrites from Nicaragua and Costa Rica (Vogel et al. 2006), and Cerro Patacon near the Panama Canal Zone (data from Rooney et al. 2011), black shaded area represents compositional range for ignimbrites from Costa Rica and Nicaragua. a Chondrite-normalized spider diagrams for El Valle volcano deposits. Samples from El Valle volcano are extremely depleted in all the REE when compared with other CAVA ignimbrites. Cerro Patacon presents the characteristic concave pattern derived from extensive amphibole fractionation. Line 1 represents the Step 1 in Table 5 for Rayleigh fractionation models. Line 2 represents Step 2 in this modeling. b Primitive mantle-normalized spider diagram. Extreme depletion is evident in El Valle suite except in Sr, Zr, and Eu. Decoupling of Nb and Ta is also evident in El Valle samples. In addition, there is a large variation in Ta concentration

Fig. 6

a Chondrite-normalized (Sun and McDonough 1989) Dy_(CN)/Yb_(CN) versus Yb_(CN) diagram. The superimposed adakite-like and typical arc fields are after Jahn et al. (1981). The dacites form El Valle fall within the adakite-like field. b Plot of Sr/Y versus Y. Again, dacites form El Valle fall within the adakite-like field

Fig. 7

Figure depicts a series of diagrams demonstrating the involvement of garnet and amphibole during the differentiation of the Quaternary volcanism at El Valle volcano. a Dy/Yb versus Dy, this diagram uses a MREE/HREE ratio to discriminate garnet fractionation. Samples from El Valle are consistent with garnet fractionation responsible for the depletion observed in the HREE. The line represents Dy/Yb variations from simple Rayleigh fractionation schemes using a garnet (20%) + amphibole (80%) fractionating assemblage. Mineral partition coefficients used for the trace element modeling are presented in Table 6. b Dy versus SiO₂ indicates the effects of amphibole fractionation (note the expanded SiO₂ axis). Dashed line with arrow represents fractionation trend due to amphibole. c Sr/Y versus MgO/FeO* demonstrates the role of garnet fractionation. Fractionation of garnet depletes the residual melt in Y and will increase the MgO/FeO* (see text). Dashed line with arrow represents fractionation trend due to garnet. d Nb/Ta versus Nb. The line represents Nb/Ta variations from simple Rayleigh fractionation schemes using an amphibole (80%) + plagioclase (18%) + sphene (2%) as the fractionating assemblage. Mineral partition coefficients used for the trace element modeling are presented in Table 6. Dashed line with arrow represents trend of CAVA ignimbrites. e SiO₂ versus Mg#. Note the higher Mg# for the El Valle deposits compared with ignimbrites from the CAVA. f K/Rb versus Mg Fractionation of amphibole produces decreasing K/Rb with decreasing Mg# (see text). Dashed line with arrow represents fractionation trend due to amphibole

In addition, the Quaternary products from El Valle exhibit a large variation in Nb/Ta at a restricted and Nb (Fig. 7d) and Mg# (0.24–0.27, not shown in figures) representing the largest variation in Nb/Ta ratios observed in any CAVA ignimbrites. The overall pattern of spider diagrams (including the Nb–Ta anomalies and LILE enrichment) of El Valle volcano is similar to other Central America subduction zone magmas and are consistent with a melt origin by mantle fluxing. However, the differences in trace element abundances and ratios (e.g. online resource 5 and Figs. 5, 6, and 7) suggest to us that the melts generated at El Valle followed a different differentiation process than those generated elsewhere along the CAVA. The most noticeable trace element differences are the high Sr/Y, Dy/Yb and low La/Yb ratios, and low abundance of REE’s (Figs. 5, 6, and 7). These observations may indicate that there were major differences in fractionating products of the El Valle magmas compared with those that generated silicic magmas elsewhere along the CAVA. The differences in chemistry may also indicate that contributions from the different components involved in magma generation processes may have been heterogeneous along the CAVA.

Amphibole compositions and geothermobarometry

In this section, we use amphibole compositions from crystals hosted in the most recent eruptives to evaluate the depths of amphibole crystallization at El Valle volcano. Amphibole is an ideal phase to study magmatic evolution due to its early appearance on the calc-alkaline liquidus (Deering 2009). An additional advantage is that amphibole chemistry is sensitive to both pressure and temperature.

Amphibole compositions

Thirty-three representative phenocrysts of amphibole from the domes, Iguana and El Hato units, and from amphibole-rich cumulates enclaves were analyzed for major element composition in core to rim data points (256 total data points). Full geochemical analyses and calculations related to these amphiboles can be found in the online resource material 6. Mineral formulae for volcanic and low-P experimental amphiboles were calculated following the spreadsheet supplied by Ridolfi et al. (2010), which follows the International Mineralogical Association (IMA) recommendation for calcic amphiboles, where Fe^3+/Fe²⁺ ratio is determined by charge balance after adjusting the tetrahedral (Si, Al, Ti) plus octahedral (Al, Ti, Cr, Fe, Mn, Mg) cations to 13 (Leake et al. 1997).

In Fig. 8, it can be seen that amphibole compositional differences are roughly predictable in terms of Al_T and ^[4]Al contents, and Al#. Data points collected at the rims of amphibole crystals in all units and amphibole-rich cumulates enclaves (48% of the data points collected, 125 data points) are low-Al or Al-free amphiboles, possibly cummingtonites ([]Mg₇Si₈O₂₂(OH)₂, Ca_B < 1.5). Most amphiboles cores and rims (41% of the data points, 106 data points) are calcic amphiboles (magnesiohornblendes and tschermakites pargasites; CaB ≥ 1.5; Ti < 0.5 atoms per formula unit: apfu) and also classify as magnesian amphiboles (Mg/Mg + Fe²⁺ 0.5–0.75). A small portion of the data collected at the cores of amphiboles in amphibole-rich enclaves and in dacitic lavas (9.76%, 23 data points) correspond with compositions atypical of volcanic amphiboles and are assumed to be xenocrystic in agreement with criteria established by Ridolfi et al. (2010) (^[6]Al/Al_T = Al# > 0.21). This is clearly illustrated in Fig. 8 where data from high-P (mantle derived) experimental crystals and deep crustal amphiboles are plotted along with the xenocrysts from El Valle.

Fig. 8

Al_T versus ^[4]Al diagrams for El Valle volcano amphiboles. The figure also shows a shaded field for high-P amphiboles (i.e. crustal amphiboles and high-P mantle experimental amphiboles). This field is derived from data included in Ridolfi et al. (2010). The compositional field of El Valle amphiboles is bounded by the Al# = 0.21 (eq. Al_T = 1.266^[4]Al) and Al# = 0 (i.e. Al_T = ^[4]Al) lines which join at the compositions of ideal cummingtonite (Cmm)

Geothermobarometry

There are several thermobarometric approaches to constraint the crystallization temperature of volcanic rocks. The most reliable and most currently used are the two-phase thermometer titanomagnetite–ilmenite (Andersen and Lindsley 1988; Ghiorso and Evans 2008). Other commonly used thermometers are QUILF (Andersen et al. 1993) and the amphibole–plagioclase (Holland and Blundy 1994). Recent work by Ridolfi et al. (2008) has presented evidence that most of these barometers produce pressure estimates that are extremely inaccurate when compared with experimental results (errors in the vicinity of 280 MPa).

Given that the El Valle volcano volcanics lack suitable ilmenite–magnetite pairs and the uncertainties of other thermobarometric methods, in this study we have used the empirical thermobarometric formulations by Ridolfi et al. (2010). This method is based on the observation that the pre-eruptive crystallization of amphibole is confined to a relatively narrow physical and chemical range closely associated with their dehydration curves. At the stability curves, the variance of the system decreases so that amphibole composition and physical–chemical conditions are strictly linked to each other (Ridolfi et al. 2010). These formulations work independently with different compositional components of amphibole (i.e. Si*, Al_T, Mg*, ^[6]Al), include new calculation schemes for relative oxygen fugacity (∆NNO) and water content in the melt (H₂O_melt) based on amphibole composition, and are easily applicable to all types of calc-alkaline volcanic products.

Relative pressure and temperature determinations

We have excluded the xenocrystic amphiboles from relative pressure and temperature determination, given that most experimental results on calc-alkaline suites have been found to be unsuitable for use in thermobarometric calibrations for high Al# (>0.21) amphiboles due to the high Al₂O₃/SiO₂ ratios of the coexisting melt (Ridolfi et al. 2010) (Fig. 8). However, given the similar chemistry of these amphiboles with deep crust and experimental upper mantle–derived amphiboles, it is extremely likely that these xenocrysts crystallized at temperatures and pressure conditions of the Panamanian lower crust. In addition, low Ca and Al amphiboles (cummingtonites) have also been discarded, as they are unsuitable for pressure and temperature determinations. However, cummingtonite is the last phase to crystallize in equilibrium with high SiO₂ melts (~78 wt%) (Bernard et al. 1996; Pallister et al. 1996; Rutherford and Devine 1996; Shane et al. 2005; Shane et al. 2008; Tappen et al. 2009). This limited stability limits the influence of cummingtonite during magma differentiation processes and constrains their depth of crystallization to shallow levels.

The results derived from the use of the thermobarometric calculations of Ridolfi et al. (2010) (online resource 6) have allowed us to characterize our suite of samples and help constrain their origin. The magnesiohornblendes found within the dacitic hosts (Iguana and domes units) record the lowest pressures and temperatures (Fig. 9, P = 90–200 ± 22 MPa and T = 790–850°C) of our sample suite and display the narrowest range of variation in pressure and temperature. Temperature and pressure variation from core to rim in amphiboles from these units tend to decrease. Tschermakites within this unit record higher temperatures and pressures (Fig. 9, P = 202–330 ± 30 MPa and T up to 930°C). The amphibole compositional variation and the pressure and temperature overlap between the Iguana and domes dacitic units. Amphiboles from both these units plot within the crystal poor zone (i.e. between the maximum stability, black dotted, and 35% crystallinity curve) (Fig. 9). The boundary curve between rhyolite and dacite melts (i.e. SiO₂ 70 wt%) roughly divides the crystallization field of magnesiohornblendes and tschermakitic pargasite, none of which fall in the field of basaltic or andesitic melts.

Fig. 9

a P–T, b log fO₂-T and c T-H₂O_melt for the El Valle amphiboles as determined by amphibole thermobarometry using the calculation scheme by Ridolfi et al. (2010). Error bars represent the expected σ_est (22°C) and maximum log fO₂ errors (0.4 log unit). a, c display representative error bars indicating the variation in accuracy with P and H₂O_melt. In a, the maximum relative P errors range from 11% (at the maximum thermal stability curve; black dotted line) to 25% (at the upper limit of consistent amphiboles; black dashed line). The maximum depth uncertainties, calculated using a crust-specific weight of 2.7 g/cm³, are also reported. The isopleths in a represent the anhydrous SiO₂ content (wt%) of the melt (e.g. SiO₂ 63%), and the P–T stability limits constrain the equilibrium of phases such as biotite (Bt), plagioclase (Pl), orthopyroxene (Opx), clinopyroxene (Cpx), magnetite (Mgn), ilmenite (Ilm), and olivine (Ol) with amphiboles (Mg-Hbl magnesiohornblende, Tsc-Prg tschermakitic pargasite). The dotted dashed curve roughly divides consistent experimental products with different crystallinity (i.e. 35–50 wt% at lower T and 12–35 wt% at higher T). b Shows the NNO and NNO + 2 curves (O’Neill and Pownceby 1993). The maximum thermal stability (black dotted line) and the (lower) limit (black dashed line) of amphiboles are also reported in c where the black error bars indicate the maximum relative error (15%) and σ est (0.4 wt%), respectively

Oxygen fugacity (Fig. 9) increases from the higher P–T tschermakitic pargasite (NNO + 1) to shallow magnesiohornblendes (NNO + 2.5) species. This is consistent with the values that have been inferred for calc-alkaline magmas (∆NNO from −1 to +3, e.g. Gill 1981; Carmichael 1991; Müntener et al. 2001). Figure 9c shows H₂O_melt–T pairs and the stability field of amphiboles, which is roughly constrained by the maximum stability curve and the (lower) limit of amphibole crystallization.

Discussion

Mechanism for the generation of silicic volcanism by garnet and amphibole fractionation

Recent models of magmatic differentiation at arcs have been centered on crystal mush models (e.g. Bachmann and Bergantz 2004, 2008; Dufek and Bachmann 2010) that are hypothesized to be located in deep crustal “hot zones” (e.g. Annen et al. 2006; Davidson et al. 2007). In these models, evolved magmas are generated by liquid extraction from crystal mushes leaving cumulate enclaves as a residue. Under the proper P, T and H₂O conditions, some of these residues would contain considerable amounts of garnet and amphibole. The importance of these mineral phases should not be underestimated. It has been hypothesized that high-pressure fractionation of garnet could generate evolved calc-alkaline liquids (Green and Ringwood 1968; Green 1972). Moreover, Bowen (1928) proposed that amphibole fractionation could be extremely important during the generation of high-silica magmas, and this has been an intensively debated topic ever since (Green and Ringwood 1968; Cawthorn and O’Hara 1976; Allen and Boettcher 1978; Sisson and Grove 1993; Davidson et al. 2007). Fractionation of both garnet and amphibole (due to their low concentration of SiO₂) will efficiently drive the derivative liquids toward high SiO₂, lower TiO₂ (controlled by amphibole) and high MgO/FeO ratios (FeO controlled by garnet), in addition to REE depletion effects. In this section, we explore the generation of silicic magmas in the Panamanian arc through garnet and amphibole fractionation.

Garnet fractionation

Crystal fractionation of garnet at high pressure has special relevance when discussing the generation of silicic magmas with adakitic-like compositions such as in the Panamanian arc. Green and Ringwood (1968) and Green (1972) were the first to propose that garnet crystallization and fractionation in the lower crust will result in HREE-depleted liquids. Their observations were based on the occurrence of garnet phenocrysts in calc-alkaline magmas of evolved composition and on experimental results derived from mineral phase stability at high pressures in a series of calc-alkaline compositions. Later, experimental data by Green (1992), Müntener et al. (2001), and Alonso-Perez et al. (2009) have supported and better constrained the observations of Green and coworkers.

In Panamanian Quaternary volcanism, the discrimination between partial melting and crystal fractionation as the processes responsible for the adakitic-like signature is difficult due to the small chemical variations observed in the differentiation indexes (Fig. 4 and online resources 4 and 5). However, the observation of increasing MgO/FeO* with increasing SiO₂ is consistent with garnet fractionation (Fig. 7c). Further evidence of garnet fractionation is the low Y and high Sr/Y values typical of adakite-like signatures (Figs. 6b, 7c). Garnet fractionation will lead to increasing Sr/Y ratio due to the high affinity of Y in garnet. Simple Rayleigh fractionation models are consistent with garnet crystallization in the Panamanian sub-arc lithosphere (Fig. 7a). Our models use the melt composition of the most primitive Quaternary Panamanian lavas (primitive basalt from the Baru volcano area) as the parental melt, then a crystal assemblage that is dominated by amphibole (80%) and garnet (20%) is modeled, leaving a melt fraction of 60% of the initial magma. The diagrams presented in Figs. 5, 6, and 7 are consistent with involvement of garnet fractionation during the magma differentiation process at El Valle volcano, hence demonstrating deep fractionation processes. This is further supported by our thermobarometric calculations (see “Amphibole compositions and geothermobarometry”) and experimental studies by Alonso-Perez et al. (2009). These experimental studies under water-rich conditions have shown that garnet is stable at pressures as low as ~0.8 GPa, allowing for the garnet to be stable in lower crustal/upper mantle reservoirs (~25 km) in mature oceanic arcs (Alonso-Perez et al. 2009) such as the Panamanian arc.

The melts resulting from our modeling reproduce most of the Dy/Yb variation observed in the El Valle Quaternary magmas (Step 1, Table 5; Fig. 7a). A second fractionation step, however, is needed to fully reproduce the REE systematics of El Valle volcano eruptives (Table 5; Fig. 5).

Table 5 Results for Rayleigh fractional crystallization models

Table 6 Partition coefficients for mafic to intermediate melts used in Rayleigh fractionation models

Amphibole fractionation

The association of almost pure amphibole cumulate enclaves or blobs in the lavas and pumice fragments and the presence of amphibole as the dominant mafic phase in El Valle volcano deposits support the concept of extensive amphibole fractionation in the Panamanian sub-arc crust. Amphibole-dominated cumulate enclaves have been observed along the Panamanian arc in a variety of host compositions and support the occurrence of a region-wide amphibole-rich layer derived from the fractionation of arc magmas (Hidalgo and Rooney 2010; Rooney et al. 2011). The locations within the Panamanian sub-arc crust, where extensive amphibole fractionation could take place, are best constrained by our barometric calculations (see “Amphibole compositions and geothermobarometry”). Our results are consistent with a first step of fractionation (Step 1, Table 5) at lower crustal depths, represented by high Al# (>0.21) amphiboles that have similar compositions as high-pressure experimental amphiboles and high-pressure crustal amphiboles at other locations (Fig. 8). These high Al# amphiboles mainly occur in amphibole cores of amphibole-rich cumulate enclaves and in a limited number of amphiboles that are included in the dacitic hosts (online resource 6). A second step in amphibole fractionation (Step 2, Table 5) is constrained by our barometric calculations and has occurred at relatively shallow depths (3.5–13 km, or ~100 to ~350 MPa). Although given that the Panamanian crust thickness in the vicinity of El Valle volcano may be around 30 km (Briceno-Guarupe 1978), some of the highest calculated pressures for amphibole crystallization (~350 MPa) could be associated with mid-crustal depths. This is consistent with what has been observed at other Panamanian volcanic centers (Hidalgo and Rooney 2010).

Several experimental studies have addressed the role of amphibole fractionation during the differentiation of arc magmas (Helz 1973; Allen and Boettcher 1978, 1983; Carmichael 2002; Grove et al. 2002, 2003, 2006). Recent experimental work by Alonso-Perez et al. (2009) has demonstrated the importance of amphibole fractionation, concluding that while garnet strongly controls some of the trace elements in derivative liquids (e.g. Sr, Y, HREE) and the FeO/MgO ratio, it is amphibole fractionation that controls variation in MREE and most of the major element abundances (e.g. SiO₂, peraluminous trend). The importance of amphibole fractionation in the sub-arc crust to produce the chemical variety that is observed in arcs has also been recently addressed by Davidson et al. (2007), Blundy and Wood (2003) and Tiepolo et al. (2000). These authors have demonstrated the important role of amphibole in fractionating key trace elements such as Nb and Ta.

Extensive amphibole fractionation plays a major role in the depletion observed in MREE in samples from El Valle volcano compared with other CAVA silicic deposits (Fig. 5). Fractionation of amphibole or clinopyroxene can effectively deplete the resulting liquids in MREE (e.g. Cerro Patacon Suite, Fig. 5). The effects of clinopyroxene fractionation are similar to those of amphibole, but for any given composition, Kd_REE clinopyroxene ≪ Kd_REE amphibole (e.g. Fujimaki et al. 1984; Bottazzi et al. 1999). Simple Rayleigh fractionation models are consistent with extensive amphibole crystallization in the Panamanian sub-arc lithosphere (Fig. 7d). Two steps of fractionation are necessary to duplicate the REE variation observed at El Valle lavas. The first step was described in the previous section and involves garnet + amphibole crystallization. The second step is produced when melts derived from crystallization of garnet amphibolites (Step 1 final composition, Table 5) crystallize an assemblage dominated by amphibole (80%), plagioclase (18%), and sphene (2%) (assemblage found in crustal cumulate enclaves), leaving 70% of the initial melt. The resulting melts in step 2 (Table 5) would reproduce most of the REE variation (Fig. 5a line 2) and Nb/Ta ratios (Fig. 7) observed in the El Valle Quaternary magmas.

In this context, the depletion of MREE in the recent volcanic rocks of El Valle volcano is produced by a two-step amphibole-dominated fractional crystallization process. This is supported by the observation of decreasing Dy (MREE) with differentiation (Figs. 6a, 7b) and the dramatic decoupling of Nb/Ta (Fig. 7d). Tiepolo et al. (2000, 2007) have proposed that due to the pressure conditions in the lower portions of the crust (<1.0 GPa), rutile will be absent and thus will not exert any control on Nb and Ta. Under these conditions, fractionation of amphibole will control the Nb/Ta of the derived magmas (Ionov and Hofmann 1995; Foley et al. 2002). In our simple Rayleigh fractionation models, amphibole + plagioclase + ilmenite fractionation are not sufficient to drive the residual liquids to high Nb/Ta ratios. In order to drive the residual liquids to the high Nb/Ta ratios observed at El Valle volcano eruptives, it was necessary to add ~2% of sphene to the crystallizing assemblage (Table 5). However, due to large amount of amphibole crystallization, it is amphibole that is responsible for at least 80% of the variation observed in Nb/Ta.

Due to the higher partition coefficient of K in amphibole in relation to Rb (Dalpe and Baker 2000), K/Rb is a particularly useful indicator of amphibole involvement in a fractionation system. During amphibole fractionation, liquids will trend toward lower K/Rb ratios with progressive differentiation (increasing SiO₂ or decreasing Mg#), which is the variation observed in the El Valle Quaternary suite (Fig. 7f).

The occurrence of almost pure amphibole cumulate enclaves and the presence of amphibole as the dominant mafic phase in El Valle volcano deposits can be used to constrain water content and crystallization history of magmas in the plumbing system of this volcanic center. Experimental studies on andesitic water-rich magmas at moderate to high pressures have concluded that amphibole is the first phase on the liquidus at depths exceeding ~7 km (Moore and Carmichael 1998; Carmichael 2002) until it is joined by garnet at higher pressures (lower crustal/upper mantle depths) (Alonso-Perez et al. 2009). However, Grove et al. (2003), using a more primitive starting composition (basaltic andesite), have demonstrated that the first phases on the liquidus of a water-saturated magma are olivine and orthopyroxene. This incongruity may be overcome by fractionation of these mafic phases in the upper mantle/lower crust boundary (Alonso-Perez et al. 2009). Fractionation of olivine and pyroxene in the upper mantle/lower crust would produce a residual liquid of basaltic–andesitic composition, which could then follow a fractionation path that could yield amphibole-rich cumulate enclaves. In El Valle volcano, there is some evidence that fractionation of olivine and pyroxene might be occurring as the first step in fractionation. Amphibole nodules are in some cases associated with olivine (ghosts) and intensely resorbed orthopyroxenes (see “Stratigraphy and petrography”).

The association of resorbed and ghost crystals of orthopyroxene and olivine with amphibole-rich cumulates enclaves presents the opportunity to evaluate fractionation and mantle equilibration models by observing the variation in Mg-number [100 * (Mg/(Mg + Fe²⁺)] with increasing SiO₂ (e.g. Grove et al. 2003). During fractionation, Mg-numbers commonly decrease with increasing SiO₂; however, andesitic melts in equilibrium with the mantle would have much higher Mg-number at the same value of SiO₂ (Grove et al. 2003). The Mg-number of amphiboles derived from these differing magmas may record such heterogeneities in their parent melts. The absence of plagioclase in equilibrium with the cumulates enclaves of El Valle volcano along with the high Mg-number in amphiboles in this enclaves supports the concept that these cumulates do not represent an evolved portion of the shallow fractionating assemblage described by Grove et al. (2003). An important consideration is that the Fe–Mg exchange between amphibole and liquid (Kd ∑Fe/Mg) is also dependent on fO₂, fH₂O (Grove et al. 2003), and to a lesser extent temperature (Alonso-Perez et al. 2009). With this caveat, the Mg-number of the amphiboles in the cumulates enclaves from El Valle volcano (~80, Hidalgo 2007) is above the value noted for amphiboles derived as initial fractionation products of Mg-rich andesite (~79; Moore and Carmichael 1998) and basaltic andesite (~79; Grove et al. 2002, 2003).

Early and extensive crystallization of amphibole to form an amphibole-rich layer in the Panamanian sub-arc lithosphere would have important chemical consequences. Early removal of large quantities of amphibole in crystal mush zones from water-saturated magmas would rapidly increase the SiO₂ content of the residual magmas (Carmichael 2002). In addition, in comparison with gabbroic assemblages (olivine + pyroxene + plagioclase), amphibole fractionation will rapidly elevate differentiation indices with relative removal of less mass. A geochemical consequence of these contrasting fractionation paths is that trace elements concentrations within a hydrous-magma increase far less than in a dry magma at similar SiO₂ due to the less mass being removed from the hydrous magmas (Carmichael 2002; Rooney et al. 2011). Extensive amphibole fractionation may explain why samples from El Valle volcano and other volcanic centers in the Panamanian arc exhibit some of the most depleted REE patterns in entire CAVA.

Source of Quaternary volcanism

Slab melting

In the Panama region, it has been suggested that melting of a subducting slab is responsible for the recent adakitic signature at the Panamanian volcanic front (Defant et al. 1988, 1991b; Clark 1989; Defant 1990; Drummond and Defant 1990; de Boer et al. 1991). However, the origin of the adakitic signature is equivocal and may be produced through a number of processes that do not require the partial melting of the oceanic crust (Atherton and Petford 1993; Castillo et al. 1999; Kay and Kay 2002; Kessel et al. 2005; Eiler et al. 2007).

The study of adakite-like magmas, since their rediscovery in 1978 (Kay 1978), has yielded important thermal constraints on the generation of melts from subducting oceanic crust. Early workers suggested that oceanic plates that were less than 45 M.a. would melt in a subduction setting, yielding adakitic-like signature magmas (Kay 1978). More recent detailed thermal modeling concluded that 45 M.a. is a substantial overestimation and instead suggested 5 M.a. as a maximum age for oceanic crust to yield a partial melt (Peacock et al. 1994). This presents significant difficulties for models of adakitic signature generation given that the type localities around the world typically have oceanic crust entering the subduction zone that is older than 5 M.a. (Peacock et al. 1994). This problem may be overcome by “flat-subduction” where the down-going slab may reach conditions of partial melting (Gutscher et al. 2000), or by the development of slab tears/window that originate when slight differences in the angular direction of both down-going plates transfer stress into the subducted mid-ocean spreading system. The “feather-tip” edges present along the spreading-center-derived slab windows are more likely to melt in comparison with the more square-ended plate edges found along transform faults (Thorkelson and Breitsprecher 2005), or slab tears caused by differential buoyancy between an oceanic ridge and normal oceanic crust (e.g. Carnegie Ridge; Gutscher et al. 1999).

Thermal models of subduction zones based on experimental results by Kincaid and Griffiths (2003) have produced substantially in higher temperatures in the subducting slab than previous models (e.g. Peacock et al. 1994) and have found that slab melting may be possible over a wider range of plate ages than have previously been assumed. Kincaid and Griffiths (2003) model implies that high slab surface temperature zones will vary in space and time depending on relative strengths of longitudinal and rollback sinking. This model is consistent with the results from high-resolution models of van Keken et al. (2002) that generate significantly higher slab surface temperature over a wide range of subduction parameters. van Keken et al. (2002) models showed that using temperature- and stress-dependent rheology that focuses on the return flow can produce significantly higher temperatures in the corner flow region compared with isoviscous rheology and that in the case of fast subduction of old lithosphere, a strong thermal gradient develops across the subducting oceanic crust that persists to great depth. However, fast subduction is not the case for the Nazca plate outboard of Panama where subduction rates are low (V = 5 cm/year, Jarrard 1986; Trenkamp et al. 2002).

Adakite-like signatures might not be related to slab melting processes. Kessel et al. (2005) have shown that at high pressures (4–6 GPa), a hydrous melt in equilibrium with MORB becomes supercritical. The low-temperature aqueous fluid produced through the dehydration of eclogite is peralkaline and is replaced by a metaluminous hydrous melt that changes from rhyolitic at the solidus toward andesitic at higher temperatures. This process could yield adakite-like trace element patterns. Exploration of this model, however, is beyond the scope of this contribution.

In Central Panama, the “melting of the subducting slab” model is only necessary if garnet stability cannot be achieved under the P–T conditions of the Panamanian sub-arc lithosphere or if the subducting slab was young enough (or hot enough) to melt during the subduction process. Recent experimental, tectonic reinterpretation of the Panama basin and geochemical data are difficult to reconcile with the “melting of the subducting slab” model for Panamanian magmas. The age of the Nazca plate subducting outboard Panama (~13 M.a., online resource 1) is too old to allow partial melting of the subducted lithosphere (Peacock et al. 1994) and subduction is too slow to allow for the application of models that use higher temperatures (van Keken et al. 2002; Kincaid and Griffiths 2003). However, the Nazca plate subducting outboard Panama is characterized by very high heat flow averaging 0.14 W/m² in the Panama Basin (Bowin 1976; Jarrard 1986) that has allowed for “thermal juvenilization” of the subducting plate, complicating the thermal structure of the margin (Hidalgo 2007). Modeling of melting of the subducted lithosphere by Peacock et al. (1994) concluded that the complicated thermal structure of the subducted lithosphere in the Panamanian basin prevents conclusive results to whether slab melting processes are responsible for the recent volcanism at El Valle. Peacock et al. (1994) added that the margin met some of the requirements for slab melting to occur but not all of them.

Additional evidence against a slab melting origin is derived from oxygen isotope data. If the only source of heat that contributes to the slab melting process is the underthrusting of the hot mantle wedge, then the only portion of the slab that should melt would correspond to the upper surface of the slab (1–2 km) (Bindeman et al. 2005). Generally, this upper section consists of ¹⁸O-rich sediments and altered basalts. In this context, ideal slab melts would have values of δ¹⁸O higher than MORB (mid-ocean ridge basalts) and island arc basalts (IAB) and in isotopic equilibrium with mantle peridotites (Bindeman et al. 2005; Eiler et al. 2007). This is not the case for the Panamanian Quaternary magmas where only slight sub-per mil enrichments when compared with MORB can be observed (Bindeman et al. 2005). An array of complicated processes could be proposed to harmonize these subtle enrichments in δ¹⁸O (much less enriched than typical slab melts) with the melting of the slab model. Some of these processes involve mixing of partial melts from different parts of the slab (Eiler et al. 2000; Bindeman et al. 2005) or isotopic exchange with the mantle wedge (Bindeman et al. 2005). Alternatively, and ultimately less complex, low δ¹⁸O observed in the Panamanian magmas can be more easily explained by differentiation/partial melting near the base of the crust, without invoking slab melts.

Other importance evidence against slab melting in Panama could be inferred from geochronological evidence presented in this study. More than half of the El Valle volcanic edifice appears to have evolved in the last ~100 k.a., indicating extremely high magma production rates. Numerical models have shown the difficulty of producing large volumes of magma by partial melting processes (Annen and Sparks 2002; Dufek and Bergantz 2005; Annen et al. 2006). Therefore, the high magma production rates are hard to reconcile with the small volumes of melt that could be produced by partial melting of the slab.

The only evidence presented to support slab melting has come from the observation that garnet must have been a residual phase and that the crust in the Panama region is too thin for garnet to have be a stable fractionating phase (Clark 1989; Defant et al. 1991a, b). Recent experimental work has substantially revised the P–T conditions for garnet stability in water-saturated melts and may now produce and fractionate igneous garnet at pressure and temperature ranges expected in the Panama arc (e.g. Alonso-Perez et al. 2009). These findings argue against the derivation of the garnet signature in Panama from the eclogitized subducting slab.

A model

Given the evidence presented above, we propose a two-stage fractionation process responsible for the chemical properties of El Valle Quaternary magmas: (1) hydrous basaltic magmas are emplaced at the base of the crust at a depth where garnet + amphibole are fractionated to produced HREE/MREE-depleted magmas; (2) extensive amphibole fractionation in the middle-shallow crust results in increased MREE- and TiO₂-depleted high-silica magmas. Models involving crystal fractionation of garnet-bearing assemblages in the lower sections of volcanic arcs have recently become important to explain the adakite-like signature (e.g. Macpherson et al. 2006; Rodriguez et al. 2007; Chiaradia et al. 2009). The occurrence of garnet-bearing cumulate rocks in what has been interpreted to be deep crustal sections of volcanic arcs (Debari and Coleman 1989; Burg et al. 1998) is consistent with crystal fractionation processes rather than partial melting of the subducting slab. Moreover, the occurrence of garnet phenocrysts in calc-alkaline magmas of silicic composition (Fitton 1972; Evans and Vance 1987; Day et al. 1992; Harangi et al. 2001) and experimental results derived from mineral phase stability at high pressures (Green 1992; Müntener et al. 2001; Alonso-Perez et al. 2009) in a series of calc-alkaline compositions support the crystal fractionation processes at the roots of island arcs.

Parental water-rich basalts stored at deep crustal levels would allow for garnet to be a stable crystallizing phase, resulting in the production of adakite-like volcanism during the Quaternary in the Panamanian arc. In addition, the absence of amphibole-rich cumulate enclaves and evolved magma compositions in older western Panamanian volcanics (with the exception of Cerro Patacon Rooney et al. 2011) and the recent appearance of amphibole-rich magmas might be an important indicator of an abrupt transition to protracted storage in lower crustal levels. Protracted storage could allow for extensive amphibole crystallization that may have resulted in an amphibole-rich layer in the Panamanian lithosphere. This process could be used to explain the appearance of more evolved magma sequences (dacites) that are not recorded in western Panama before the Quaternary.

Conclusions

The new ⁴⁰Ar/³⁹Ar geochronological analyses of the volcanism of El Valle volcano presented in this study have unraveled one of the most recent and more productive periods of volcanic activity (younger than ~56 k.a.) in the CAVA and refined previous K/Ar age determinations by Defant et al. (1991a). Moreover, the new age determinations allowed for the categorization of the El Hato pyroclastic flow as one of the largest eruptions during the Quaternary period in the Central America region.

The geochemical homogeneity found in El Valle volcano Quaternary volcanism is unique in the studied silicic ignimbrites of the Central America volcanic front. The depletion of HREE, low Y, high Sr, high Sr/Y, and low K₂O/Na₂O ratios do not occur in the well-studied deposits in Costa Rica and Nicaragua (Gillot et al. 1994; Vogel et al. 2004, 2006). The source of the Quaternary dacites in El Valle are different from the other silicic magmas from the CAVA and can be explained by a two-stage fractionation process in which garnet + amphibole fractionation in the lower crust produces HREE depletion and enrichment in MgO, while amphibole-dominated fractionation in the middle-shallow crust produces increased MREE and TiO₂ depletion as well as SiO₂ enrichment.

Amphibole fractionation within the Panamanian crust is supported by the ubiquitous presence of amphibole-rich cumulate enclaves in the Panamanian volcanic arc from the Oligocene (Cerro Patacon, Canal Zone) to Quaternary age in El Valle units (including El Hato unit) and in Baru volcano in northern Panama. Cumulate enclaves from Cerro Patacon as well from Baru volcano have been interpreted to have crystallized at middle to lower crustal depths (Hidalgo and Rooney 2010; Rooney et al. 2011), hence supporting amphibole fractionation as an early differentiation process in the Panamanian sub-arc crust. This amphibole-rich accumulation zone has been previously described by Davidson et al. (2007) as the “amphibole sponge”, due to its importance as a filter for water in its transition from mantle to crust and as a way to recycle water and incompatible elements back into the mantle (delamination). Amphibole fractionation may also greatly influence the major element concentration in arc lavas. Amphibole contains significantly less SiO₂ and more TiO₂ than basalt, and thus during fractionation, the liquid composition would be efficiently driven to higher SiO₂ and lower TiO₂. Furthermore, it is within this mid-lower crust amphibole-rich layer that water-rich magmas can become stalled during fractionation, driving the interstitial liquids to more evolved compositions typical of continental crust. Differentiation of H₂O-rich magmas under conditions relevant for the roots of island arcs is key for understanding the genesis of dacites and andesites, thus for the generation of continental crust in general.