Linking shallow and deep crustal processes at volcanic arcs is an important component in evaluating the growth and evolution of the continental crust. Commonly, deep crustal processes and the nature of subarc lithosphere are studied long after the volcanism has ceased in obducted arc terranes. In active arcs, studies of deep crustal processes focus on cumulates derived from middle-lower crustal levels. Although uncommon in the erupted magmas, these cumulates are required by crustal differentiation models of arc magmatism. Quaternary magmas at Baru volcano in Panama contain ubiquitous amphibole-bearing cumulates that provide an opportunity to probe the magma plumbing system of an active arc volcano. We have determined that these cumulates are related to their host magmas by crystal fractionation processes. Pressure and temperature estimates for amphiboles within these cumulates and the host rock are consistent with sampling of mush/magma zones from throughout the arc crust. These mush zones would be localized in deep hot crustal zones where magmatic differentiation of water-saturated arc magmas takes place by crystallization of amphibole-rich cumulates. The identification of middle-lower crustal cumulates is not exclusive to Baru volcano; similar cumulates are common throughout the Panamanian arc and are consistent with a widespread amphibole-rich layer present within the arc crust of Panama. Our results highlight the importance of amphibole fractionation in the differentiation sequence of island arcs effectively driving the residual magma to the average andesitic composition of the continental crust.
The study of magmatic processes is generally constrained to the shallowest parts of the plumbing system of any given arc volcano, with only few studies that address middle to lower crustal processes [e.g., Beard, 1986; DeBari and Coleman, 1989; Cigolini et al., 2008]. Correlating shallow, deep, and upper mantle processes at volcanic arcs has been an important goal for generating a fuller understanding of the growth and evolution of the continental crust [Davidson et al., 2005]. Rudnick and Fountain  and Rudnick and Gao  estimated that the average composition of the continental crust is andesitic to dacitic, emphasizing that the crust is stratified: mafic rocks (or melts) at the base of the crust and andesitic or dacitic-like rocks concentrated in the upper crust. This crustal stratification has been supported by geological evidence observed in arc terranes where mafic rocks are abundant in exposed lower crustal sections [Jagoutz et al., 2006; Dhuime et al., 2007].
The role of a middle to lower crustal amphibole fractionation from hydrous magmas has been recognized by some researchers as an important component in the evolution of magmatism and recycling in volcanic arcs [Boettcher, 1973, 1977; Allen and Boettcher, 1983; Davidson et al., 2007]. Despite being generally absent as a phenocryst phase in arc-related volcanic rocks, the geochemical signature of amphibole fractionation is pervasive in most arcs (decreasing Dy/Yb with fractionation indices [Davidson et al., 2007]). Arc terranes that have been exhumed (e.g., Alaska, Kohistan arc, Bonanza arc) commonly contain amphibole-bearing cumulates and these illustrate the important role of amphibole in the subarc crust. Amphibole in these hydrous cumulates is typically reported as an intercumulus phase that developed as magma conditions evolved toward lower temperature and higher water content [Bachmann et al., 2002]. Most studies in these terranes have concluded that the intercumulus crystallization of amphibole from hydrous magmas has controlled the compositions of derivative liquids until the onset of plagioclase crystallization [Jan and Howie, 1981; DeBari and Coleman, 1989; Greene et al., 2006; Jagoutz, 2010]. In active arcs, amphibole-bearing cumulate xenoliths may be the only available window into middle to lower crustal processes.
The Quaternary magmas at Baru volcano in Panama contain ubiquitous amphibole-bearing cumulates that provide an opportunity to probe the magma plumbing system of an active arc volcano. In this study, we use amphibole compositions from crystals hosted in the most recent products of Baru volcanism along with whole rock major and trace element analyses to evaluate Baru volcano magma differentiation processes. The advantage of using the amphibole chemistry is its sensitivity to both pressure and temperature. Moreover, amphibole is an ideal phase to study magmatic evolution due to its early appearance in the calc-alkaline liquidus in hydrous magmas [Carmichael, 2002; Alonso-Perez et al., 2009]. Our relative pressure and temperature estimates help to elucidate the nature of the Panamanian subarc crust and aid in the study of the magma plumbing system of Baru volcano showing that the magmatic evolution may be closely related to deep differentiation processes in hot zones within the subarc crust. Our detailed study may have broader application, in so far as these amphibole-rich cumulates are not exclusive to Baru volcano and have been identified in other regions of the Panamanian arc (e.g., Quaternary volcanism at El Valle, Oligocene sequence at the Panama Canal) [Hidalgo, 2007; Rooney et al., 2010]. The widespread presence of amphibole-rich cumulates in Panamanian magmas allows for the study of a regional amphibole-rich layer within the arc crust that originated most likely by stalling and fractionating mantle derived water-rich magmas.
2. Study Area and Tectonic Setting
Baru volcano is located in Western Panama, 35 km east of the Costa Rica–Panama border at the terminus of the Talamanca Cordillera. The summit of the volcano at 3,374 m, overlooks populated valleys 2,000 m below (Figure 1). Despite being dormant for the last 400 years, Baru volcano is active, with at least four eruptive episodes in the last 1,600 years and several others in the prior 10,000 years [Sherrod et al., 2007]. The volcanic edifice has been constructed by numerous eruptions. Sherrod et al.  estimated that the last period of abundant lava flows (last 11,500 yr) is presumed to have formed the large andesitic-dacitic dome sequence located at the summit.
Tectonically, Baru volcano is situated at the southern end of the Central American Volcanic Arc (CAVA) (Figure 1). This section of crust was first described by Kellogg and Vega  as the Panama block and is moving northward relative to the Caribbean plate and eastward relative to the South American plate. The northern boundary of the Panama Block is the North Panama Deformed Belt (NPDB) characterized by the southward subduction of the Caribbean Plate [Camacho et al., 2010]. The NPDB crosses Costa Rica on a westward trend to meet the Middle America Trench in a Caribbean Plate–Cocos Plate–Panama Block triple junction [Mann and Kolarsky, 1995; Bird, 2003]. The eastern limit of the Panama Block is the Panama-Colombia suture zone (collision zone), and has been described extensively by Mann and Kolarsky . The southern boundary of the Panama block is the Panamanian Trench (or Southern Panama Deformed Belt). This boundary is complex; it changes character eastward from oblique subduction (between 83°W–80.5°W) of the Nazca plate (V = 5 cm/yr [Jarrard, 1986; Trenkamp et al., 2002]) to a sinistral strike-slip fault (80°W to 78.8°W [Westbrook et al., 1995]). Volcanism at Baru is interpreted to result from the subduction of the Nazca Plate under the Panama Block (Figure 1).
The material entering the Middle America Trench outboard of Panama has been derived from diverse sources including: East Pacific Rise [Meschede et al., 1998; Barckhausen et al., 2001], Cocos Nazca spreading centers [Barckhausen et al., 2001], Sandra Rift [Lonsdale, 2005] and Galapagos hot spot tracks. The Galapagos hot spot tracks are perhaps the most conspicuous features on the subducting Cocos and Nazca plates, best represented by the aseismic Cocos and Coiba ridges. These ridges are overthickened sections of oceanic crust, that stand ∼2 km above the surrounding ocean floor [Walther, 2003] and are interpreted to be colliding with the Middle American Trench [Lonsdale and Klitgord, 1978]. The existing evidence suggests that the interaction of the Cocos and Coiba ridges with the Middle America Trench has a profound impact on the overriding plate, yielding compressive crustal stress conditions quite different from regions of “normal” subduction. The timing of the exact arrival of the Cocos and Coiba Ridges at the Middle America Trench remains highly controversial and ranges from ∼8 Ma to ∼1 Ma [Gardner et al., 1992; Collins et al., 1995; de Boer et al., 1995; Abratis and Worner, 2001; Morell et al., 2008].
3. Field and Petrographic Observations
The regional basement in western Panama is commonly referred to as an over thickened section of the lithosphere composed of oceanic assemblages (21 Ma to 71 Ma [Hoernle et al., 2002]). This local basement is part of the Caribbean Large Igneous Province (CLIP), which is related to Galapagos hot spot; these assemblages are referred to as the Azuero-Soná Complex. The CLIP is exposed in the Azuero and Soná peninsulas, and various islands of the Chiriquí and Montijo gulfs [Denyer et al., 2006]. The thickness of crust is unknown in the mountainous region of western Panama, but is likely thicker than the crust in the Panama Canal region, estimated at ∼25 km [Briceno-Guarupe, 1978].
In other sections of the Central American Arc, Oligocene and Miocene arc sequences lie behind the currently active arc front (i.e., Costa Rican arc) [Carr et al., 2003; Gazel et al., 2009]. At Baru volcano, the Miocene and modern arc products largely overlap. This feature makes Baru volcano a key location to study the temporal evolution of the arc in order to understand the arc initiation and the closure of the Isthmus of Panama. The subdivision of magmatism into two temporal groups simplifies the discussion of the stratigraphy of Baru volcano. The first group constitutes most of the volcanic edifice and is composed of pyroclastic flows, lava flows and lahar deposits [Sherrod et al., 2007]. This unit represents about 60%–70% of the total erupted volume. None of the rocks in the lower part of the sequence have been dated, however these deposits are attributed to the voluminous effusive activity during the late Miocene [de Boer et al., 1991; Drummond et al., 1995] and are the subject of ongoing research. The second group (that is directly addressed here) is composed of the products of Quaternary volcanism at Baru and is represented by at least two subunits. The first is the summit domes unit that is composed mainly of hornblende bearing andesites and dacites [Sherrod et al., 2007; Rausch, 2008]. The second Quaternary unit at the top of the sequence is characterized by fallout tephras and pyroclastic flow deposits. Dated samples from this unit yielded ages of 950 yr B.P. and 855 yr B.P. [Sherrod et al., 2007]. The pumice block chemistry indicates that the composition varies from andesitic to dacitic. The Quaternary units have textural characteristics that suggest magma mingling/mixing at macroscale, but routine microscope work revealed that these textures are produced by fined grained and coarse-grained cumulates hosted within both of the Quaternary units (Figure 2).
The present study focuses on the Quaternary units that bear abundant hornblende gabbro cumulates. Petrographically, the host andesite that contains these cumulates is crystal rich in the summit domes unit (up to 55%), but crystal poor (15%) in the ash flow unit. The mineralogy of the host magmas is dominated by plagioclase and zoned amphiboles with small amounts of Fe-Ti oxides and trace contents of sphene. Some plagioclase crystals have nucleated on resorbed and small quartz crystals (∼0.3 mm) and in some cases crystallites of amphibole are observed in the plagioclase cores. Furthermore, some plagioclase crystals have convoluted and irregular zoning patterns and resorbed cores, while some of the largest amphiboles have orthopyroxene cores.
Texturally, cumulates can be organized in two groups, fine-grained cumulates (Figures 2a–2c and 3a) and coarse-grained cumulates (Figures 2d–2f and 3b–3d). These cumulates occur in both host units (ash flow unit and summit domes unit) but are better preserved in the summit domes unit. None of the cumulate groups present chilled margins at the contact with the andesitic host. Fine-grained cumulates generally form large nodules (5–10 cm) that are composed of variable proportions of amphibole (acicular in some instances) and plagioclase phenocrysts and microlites along with minor interstitial glass. Some of the fine-grained cumulates are amphibole rich (80% amphibole, 5% plagioclase, Fe-Ti oxides and interstitial glass) while some of them are plagioclase rich with little amphibole (90% plagioclase, some amphibole, interstitial glass and Fe-Ti oxides) (Figure 3a). The larger size coarse-grained cumulates are more complex and can be further subdivided in two groups. The most common group forms 2–5 cm nodules of amphibole megacrysts (3–6 mm) with minor (<2%) interstitial glass. In some cases these cumulates contain anhedral plagioclase and apatite phenocrysts (∼0.5 mm) in the interstices (Figure 3b). Less widespread is the second group within the coarse-grained cumulates. This group of cumulates are composed of orthopyroxene and olivine crystals (Fo ∼ 80) that are surrounded by amphibole-dominated reaction rims that in some cases extend through fractures inside the cumulates (Figure 3d). Olivine ghosts and orthopyroxene crystals that have almost completely reacted to form amphibole are also observed.
4. Analytical Techniques
Major and trace element concentrations for the host lavas and the cumulates were determined by X-ray fluorescence (XRF) and Laser Ablation Inductively Coupled Plasma Mass Spectrometry (LA-ICP-MS) analyses at Michigan State University using fused glass disks of previously powdered samples. The procedure followed the outline described by Hannah et al.  and Deering et al.  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®, which uses fundamental parameters [Criss, 1980].
For LA-ICP-MS trace element analyses, a Cetac+ LSX200 laser ablation system coupled with a Micromass Platform ICP-MS was used. The Cetac® LSX-200+ is a Nd:YAG laser with a frequency quadrupled to an UV wavelength of 266nm. The analyses involved continuous ablation (line scan) for approximately three minutes. Strontium, determined by XRF, was used as an internal standard to correct for variations in ablated sample volume and instrument response. Trace element data reduction was done using MassLynx® software. Prior to any calculations, the background signal from the argon plasma was subtracted from each of the standards and samples. Element concentrations in the samples were calculated based on a linear regression method using BHVO, W-2, JB-1, JB-2, JB-3, JA-2, JA-3, BIR, QLO-1, and RGM-1 standards. Only standards with calculated values within 15% of the preferred standard values were used in the final calibration line for samples. Precision and accuracy of both the XRF and LA-ICP-MS chemical analyses have been reported by Vogel et al. . Trace element reproducibility based on standard analyses is typically better than 5%.
Hornblende and plagioclase 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, FeOtot, MgO, CaO, Na2O, K2O, TiO2, MnO) of 98 ± 1 wt % were retained. For glass analyses (intercumulate glass and pumice samples) a 10 μm defocused beam and 5 nA beam current was used. In order to minimize Na migration during glass analyses, the count rate of Na was scanned through time and corrected using a built-in procedure [Devine, 1995].
5. Nature of the Host and Cumulates
Average compositions and standard deviations of major and trace elements are provided in Table 1 for the host and the cumulates. The complete geochemical analyses are presented in the auxiliary material. Samples with totals lower than 96% are excluded from the discussion because it was assumed that these samples were altered. In Figure 5 all the major elements have been normalized to 100%.
Table 1. Average Whole-Rock Chemical Composition for the Quaternary Units at Baru Volcano
Ash Flow Host
Summit Domes Host
XRF (wt %)
The host magmas consist of the ash flow unit and the summit domes unit. These two units are metaluminous calc-alkaline andesites, with a few samples extending into the dacite field. The host units generally present little variation in most of the major and trace elements (Figures 45–6). The ash flow unit is slightly less evolved than the summit domes unit and extends to lower SiO2, K2O, Na2O and higher contents of MgO, CaO, TiO2, Fe2O3T (Table 1 and Figures 4 and 5). Despite these differences, trace element variation between these two units is negligible which is consistent with a similar source for both of these deposits (Table 1 and Figures 6 and 7). The differences observed in major elements between these deposits can be explained by the inability during the sample preparation process to separate hornblende-rich fine-grained cumulates from ash flow unit pumice blocks (host), which overall made this unit more mafic. Detailed microscope work on the most evolved samples of the ash flow unit revealed that they are entirely free of entrained cumulate material, while the most mafic samples have incorporated portions of fine-grained cumulates. In addition, fine-grained cumulates and small coarse-grained cumulates could not be separated from a small group of samples (8 samples) from the summit dome unit. These samples are relatively more enriched in MREE and HREE than the typical host (Figure 7b).
The trace element trends observed in the Quaternary magmatism are typical of island arcs with pronounced depletions in Nb and Ta and enrichment of large ion lithophile elements (LILE) over the more immobile high field strength elements (HFSE). The small variations observed in trace elements between the host units suggest a similar source (Figure 7). Both of the host magmas have depleted heavy rare earth elements (HREE), low Y, high Sr, high Sr/Y, low K2O/Na2O (Figure 7). These are typical of adakite-like volcanism (Figure 8). This type of volcanism has been identified in the Quaternary activity of Panamanian arc (e.g., La Yeguada volcano, El Valle Volcano and Cerro Colorado) [Restrepo, 1987; Roy, 1988; Defant et al., 1991; Hidalgo, 2007] and is a clear indication of active and important regional magmatic processes within the isthmus.
Bulk analyses of cumulates provide detailed geochemical constraints on their origin. However, these analyses are not representative of liquid compositions. We use these bulk analyses simply as a reference to infer relationships to the host magma.
Fine- and coarse-grained cumulates despite having similar mineralogy (dominantly amphibole) have important geochemical differences. Typically, the coarse-grained cumulates have higher MgO, CaO, TiO2, Fe2O3T and lower SiO2, K2O, Na2O than the fine-grained cumulates (Figures 4 and 5). In addition, coarse-grained cumulates present some scatter in Al2O3, TiO2, MgO, CaO, FeO* versus SiO2 (Figure 5). The scatter might represent different source liquids within the cumulates or might just represent variation in the amphibole-plagioclase ratios in this group. Small changes in the amphibole-plagioclase ratios would greatly affect Al2O3 as well as most major element concentrations.
Fine-grained cumulates were more difficult to analyze chemically due to their low abundance, poor preservation, and the difficulties in obtaining a pure sample from the ash flow host. Nonetheless, three samples included in the summit domes unit were analyzed and indicate that these cumulates have SiO2, MgO, Fe2O3T, and CaO intermediate between those of the coarse-grained cumulates and the host units (Table 1 and Figure 5). In addition, the interstitial glass found in both of these cumulates is clearly distinct (Figure 5). The glass contained in the coarse-grained cumulates is more enriched in K2O and more depleted in Na2O than the glass found in the fine-grained cumulates. This suggests that the magma that transported these cumulate nodules was not completely homogeneous, or that slightly different magmas were involved in the transport of these cumulate piles to more shallow levels. The coarse-grained cumulates, despite their more primitive composition, were transported by a slightly more evolved magma than the fine-grained cumulates. Such heterogeneity may reflect the complexity of the volcano plumbing system under Baru and will be evaluated in section 7.
Trace element diagrams effectively describe the differences between the cumulate types and the host magmas (Figures 7 and 8). The coarse-grained unit is more depleted than the fine-grained cumulates in most of the LILE and HFSE except for the HREE that are more enriched in the coarse-grained cumulates (Figure 7). Light rare earth elements (LREE) are commonly more enriched in the fine-grained cumulates while middle rare earth elements (MREE) are slightly more enriched in the coarse-grained cumulates. Compared to the host magmas both cumulate groups are more depleted in LILE and LREE and are more enriched in most HREE and MREE.
5.3. Relationship Between the Cumulates and the Host
The host magmas and the fine-grained cumulates have similar patterns of REE despite the slight enrichment of the later in most trace elements. This implies that these units are consistent with having similar parental magmas despite the contrasting concentrations of major elements (e.g., MgO, SiO2, Fe2O3T, CaO, etc) observed in these units (Figures 4 and 5).
The relationship between the coarse-grained cumulates and the rest of the units is less clear. Both cumulate groups and the evolved host magmas share similar mineralogy except for the presence of minor olivine and enstatite with amphibole reaction rims in the coarse-grained cumulates (Figures 2f and 3d). Nonetheless, occasional olivine ghosts and resorbed and embayed crystals of olivine and enstatite have been also observed in the ash flow unit (host) samples. Plagioclase and amphibole are pervasive in the fine-grained cumulates, the host magmas, and the coarse-grained cumulates (sometimes with little or no plagioclase) which is suggestive of a common origin. Furthermore, amphibole and plagioclase compositions (auxiliary material) from the coarse-grained cumulates and the host magmas partially overlap. This is consistent with these crystal populations being cogenetic. These observations allow us to identify a possible fractionating assemblage that can be used to model the magmatic variation observed at Baru volcano.
Major element mass balance and trace element Rayleigh fractionation schemes were used to constrain the fractionating assemblages as well as the fractionating proportions. Only samples that were free of contamination with other units (i.e., host mixed with cumulate units) were used for the mass balance and Rayleigh fractionation models. We started by selecting one of the most primitive samples found at Baru volcano (CR1100747) and then subtracted the average compositions of the minerals in the cumulate units. To minimize the residuals from ∼54%–64% SiO2, 40% extraction (by mass) of the assemblage found in the cumulates units was needed (Tables 2 and 3). A subtracted assemblage dominated by a high-Al amphibole, effectively reproduces the variation observed in our major element data. Moreover, the cumulates are also successfully reproduced by the modeled solid residue left by the fractionating parental magmas (Table 3 and Figures 8a–8d). Coarse-grained cumulates contain all the phases necessary to explain the major element evolution of the Baru volcano magmatic suites.
Table 2. Mineral Compositions: Averages of Natural Minerals in the Coarse-Grained Cumulatesa
The mineral proportions (wt %) were calculated by least squares optimization schemes.
Trace element Rayleigh fractionation was modeled using the results provided in the major element mass balance (Table 3). Mineral partition coefficients used for the trace element modeling are presented in Table 4. The model mineral proportions, which are in agreement with observed mineral proportions, were used to calculate the trace element composition of a model-derived evolved liquid. If a magma with the composition of one of the most primitive Baru lavas were to crystallize an assemblage dominated by amphibole (∼85%) and minor plagioclase, leaving a melt fraction of 60% of the initial magma, the resulting melt would reproduce much of the trace element variation observed in the host and fine-grained cumulate samples (Table 5). This modeled evolved liquid effectively correlates to most of trace element abundance observed in the most evolved magmas at Baru volcano, except for the HREE (Figure 9).
Table 4. Partition Coefficients for the Mafic to Intermediate Melts Used in the Rayleigh Fractionation Modela
For REE preferred value and range of published partition coefficients are given.
Table 5. Results for the Rayleigh Fractional Crystallization Modela
LL Dacite WBV22100723B
Samples from enclaves and the andesitic-dacitic host are included for comparison. Unit is ppm.
In our modeled magmas HREE are too enriched when compared to the variation observed in the most evolved magmas at Baru volcano (Figure 9). Cryptic garnet fractionation [Rodríguez et al., 2007; Macpherson, 2008; Larocque and Canil, 2010] might be a possible solution for this discrepancy. Garnet is extremely rare as a phenocryst in volcanic rocks, and is not observed in the cumulates from Baru volcano. Nonetheless, its importance in fractionating HREE from parental magmas has been described and detailed in a number of studies [Lee et al., 2006; Hidalgo et al., 2007; Alonso-Perez et al., 2009; Hattori et al., 2010; Kratzmann et al., 2010; Larocque and Canil, 2010]. It is worth noting that even the most primitive of the coarse-grained cumulates already displays a depletion of the HREE that can only be reproduced if a mafic magma undergoes fractional crystallization within the garnet stability field or if it is derived from partial melts of a garnet bearing protolith (Figure 8). The HREE depletion observed in the cumulates would be consistent with early garnet removal or melt extraction from the parental water-rich basalts within the garnet stability field. For our modeling, in addition to the mineral phases that are observed in the cumulates, garnet was subtracted (1.4%) until the concentration of Lu in the model matched the concentration of the average Lu in the host samples, which recreated the HREE depletion pattern observed in the host magmas (Figure 9).
We recognize that the two models (major element mass balance and Rayleigh fractionation) are a simplification of the complex magmatic processes occurring at Baru volcano. Nonetheless, we believe that the evolutionary magmatic trend during the Quaternary can be explained by fractionation processes within the Panamanian crust, without the need for a more rigorous fractional crystallization model derived from experimental data with the specific magma compositions of Baru volcano primitive magmas.
6. Amphibole Compositions and Geothermobarometry
6.1. Amphibole Compositions
Around 81 representative phenocrysts of amphibole from the host units, coarse-grained cumulate and fined grained cumulate units were analyzed for major element composition in core to rim data points. Full geochemical analyses, graphical representation of the chemical variation and calculations related to these amphiboles can be found in the auxiliary material. All of the analyzed amphiboles are calcic amphiboles (CaB ≥ 1.5; Ti < 0.5 atoms per formula unit: apfu) and also classify as magnesian amphiboles (Mg/Mg + Fe2+ 0.5–0.75) (Figure 10). The structural formulae of all the amphiboles used in this study were calculated using cation distribution determined by the procedure of Leake et al.  which estimates ferric iron based on the maximum stoichiometric limits (13 CNK).
In this section we use amphibole compositions from crystals hosted in the most recent eruptives to evaluate the magma plumbing system of Baru volcano. The advantage of using the amphibole chemistry is its sensitivity to both pressure and temperature. In addition, amphibole is an ideal phase to study magmatic evolution due to its early appearance on the calc-alkaline liquidus [Deering, 2009].
6.2.1. Exchange Mechanisms
Several studies have addressed the effects of temperature, pressure and fH2O on the composition of amphiboles [Spear, 1981; Johnson and Rutherford, 1989; Blundy and Holland, 1990; Thomas and Ernst, 1990; Schmidt, 1992; Scaillet and Evans, 1999]. These studies have agreed that the Al-tschermak substitution (SiT + MgM1–M3 = AlT + AlM1–M3) is susceptible to variations in pressure. In contrast, Spear  and later Blundy and Holland  determined that temperature variations are marked by an increase in total Al through the edenite exchange (SiT + A = AlT + (Na + K)A) and an increase in Ti through the Ti-tschermak exchange (2SiT + MnM1–M3 = 2AlT + TiM1–M3). Spear  also described that the Ti content correlates positively with temperature in the presence of a Ti-rich phase. Independently of pressure and temperature, fH2O appears to have an effect on the amphibole Altot: experiments on the Mount Pinatubo magmas have reported an Altot increase with increasing fH2O [Scaillet and Evans, 1999].
Most of the variation observed within the units sampled at Baru volcano can be explained by these exchange mechanisms as a derivative processes of pressure and temperature variation (Figure 11). Figure 11b displays good correlation of [(Na + K)A] with [AlT(Na + K)A] and also correlates with Ti (not shown in Figure 11) which is consistent with edenite exchange and hence a variation in the temperature conditions. This is further supported by Ti-C versus AlIV diagram (Figure 11e). Nonetheless, the influence of pressure is evident (despite some scatter) in the correlation of AlIV with AlVI and the positive correlation of AlVI with AlT (not shown), all of which is consistent with Al-tschermak exchange mechanism (Figure 11a). On the basis of these results, the observed cation variation in the amphiboles from Baru volcano is consistent with amphibole crystallization at polybaric conditions over either a significant temperature range and/or significant fH2O range.
6.2.2. Relative Pressure and Temperature Determinations
The buffering assemblage quartz + alkali feldspar + plagioclase + hornblende + biotite + Fe-Ti oxides + titanite + melt + fluid frequently has been used to calibrate amphibole barometers [Hammarstrom and Zen, 1986; Johnson and Rutherford, 1989; Schmidt, 1992; Anderson and Smith, 1994]. However, this assemblage is not found in any of the sampled units from Quaternary Baru volcanism. Therefore, our pressure estimates should be used with caution and serve only as a relative indicator between the distinct populations of cumulates and the host. Due to the temperature dependence of some of the amphibole barometers [Anderson and Smith, 1994] we have used the formulation of Holland and Blundy  to establish which of the barometers listed in the previous paragraph is more suitable for our suite of samples. The temperature estimates determined by the calibration of Holland and Blundy  were in excess of 800°C (Figure 12) and generally agree (within error) with the calculations of temperature using the Ti solubility method of Femenias et al. . Due to these high temperatures, using the Anderson and Smith  experimental calibration can result in increased corrections that may lead to erroneous pressure estimates. For this reason, we have chosen to determine the pressure estimates using the higher-temperature experimental calibration of Schmidt .
The results derived from the Holland and Blundy  and Schmidt  calibrations have allowed us to characterize our suite of samples and help constrain their origin. The magnesiohornblendes found within the fine-grained cumulates record the lowest pressures and temperatures (Figure 12, P: 0.25–0.4 ± 0.06 GPa and T: 800–850°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 the fine-grained unit are unsystematic.
The amphiboles from coarse-grained cumulates and the amphiboles included within the host andesites-dacites have similar P-T characteristics (Figure 12). The coarse-grained cumulates display significant variability in pressure and temperature (Figure 12, P: 0.3–0.8 ± 0.06 GPa and T: up to 950°C). In a similar fashion, the host amphiboles have recorded extremely variable pressures (0.25–0.8 ± 0.06 GPa) and temperatures (800–900°C) that might be a reflection of the mixed crystal population of this unit. The amphibole compositional variation and the pressure and temperature overlap between the host units and the coarse-grained cumulates is consistent with the amphibole population in both these groups being cogenetic. It is likely that most cumulates suffered disaggregation during transport (forming the host amphibole phenocrysts), while some coarse-grained cumulates survived unperturbed.
7.1. Role of Amphibole During Magma Differentiation
7.1.1. Amphibole-Dominated Crystal “Mush”
We have established that the cumulates and host magmas are likely derived from the same magmatic plumbing system, a statement supported by comparable REE patterns (Figure 7), similar mineralogy (Figures 2 and 3) and overlapping amphibole and plagioclase compositions (Figure 10, see auxiliary material). This determination allowed us to test for crystal fractionation as the main process driving magmatic evolution at Baru volcano (see section 5). Diagnostic trace elements from the Quaternary volcanism at Baru volcano (Dy, Yb, La) are consistent with the fractionating assemblage being dominated by amphibole (Figure 13), as is the ubiquitous presence of amphibole as the sole ferromagnesian phase. Amphibole appears throughout the magmatic differentiation process, as evidenced by the wide compositional ranges observed (Ti magnesiohastingsite-Ti tschermakite-tschermakite-magnesio hornblende) and thus amphibole fractionation is the key component in our modeled crystal fractionation schemes (Figure 9). Specifically we have shown that amphibole-dominated fractionation (∼85% of the fractionating assemblage) can be linked to the magmatic evolution from basalt to dacite, and thus may represent a liquid line of descent.
Recent efforts to understand magma differentiation mechanisms at arcs have been centered on crystal mush models [e.g., Bachmann and Bergantz, 2004, 2008] that are hypothesized to be located in deep crustal “hot zones” [e.g., Annen et al., 2006; Davidson et al., 2007]). In host and cumulate amphiboles from Baru volcano the complex and heterogeneous chemical composition and geothermobarometry results from the sampled crystals are consistent with a crystal mush model of magma evolution that was sampled by ascending magmas. In this model, hydrous magmas stall in the lower crust where after a period of incubation, melt extraction of more evolved magmas (andesite-rhyolite) from the resulting crystal mush (cumulates) is possible [Deering et al., 2008]. This process is long-lived and new replenishments of primitive hot magmas will keep mush generation and melt segregation processes active [Bachmann and Bergantz, 2008]. It is important to remember that local variations in the parental magmas composition or P, T, fO2, fH2O conditions would greatly modify the composition and mode of the crystallizing assemblage.
Our barometric calculations using the amphiboles from the cumulate rocks of Baru volcano have allowed for general pressure estimates that may be consistent with an amphibole-rich accumulation zone (crystal mush) in the Panamanian subarc lithosphere (Figure 12). Amphibole phenocrysts contained within the host units would also be part of this accumulation zone; given the overlapping amphibole compositions and similar P-T variation when compared to the coarse-grained cumulate nodules (Figure 12). This is consistent with the host amphibole phenocrysts being recycled antecrysts from crustal amphibole-rich accumulation zones (coarse-grained cumulate unit). This amphibole-rich accumulation zone has been previously described by Davidson et al.  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 in to the mantle (delamination). Furthermore, it is within this middle-lower crust amphibole-rich layer that water-rich magmas can become stalled during fractionation, driving the interstitial liquids to more evolved compositions.
7.1.2. Pressure and Temperature Conditions
Pressure and temperature estimates derived from amphibole compositions in monomineralic cumulates can give some insight in to the P-T and hydrous conditions of the parental magmas in the region. Experimental studies on water saturated andesitic magmas at moderate to high pressures have concluded that amphibole might be the first and sole phase on the liquidus at depths exceeding ∼7 km [Carmichael, 2002; Alonso-Perez et al., 2009]. Nonetheless, Grove et al. , 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 at the upper mantle/lower crust boundary. Fractionation of olivine and pyroxene in the upper mantle/lower crust would produce a residual liquid which could then follow a fractionation path that would yield a monomineralic amphibole cumulate [Alonso-Perez et al., 2009]. This also agrees with experiments performed by Carmichael  in which he suggested that the andesitic starting compositions could be derived directly from the upper mantle, provided the magma was sufficiently hydrous. In Baru volcano, there is some evidence that fractionation of olivine and pyroxene might be occurring at the mantle/lower crust boundary as the first step in fractionation of water-rich magmas. Coarse-grained cumulates often contain olivine and orthopyroxene (Figure 3). In most samples where olivine crystals are present, they have been reduced to ghosts. However, in some rare occurrences well preserved dunites (∼5 cm diameter) surrounded by orthopyroxene coronas and interstitial amphibole are observed. In addition, orthopyroxenes commonly are present at the cores of amphibole crystals or they have developed amphibole coronas that in some cases extend into the crystal network through fractures.
The mafic association in the cumulates described for Baru volcano supports early fractionation of orthopyroxene and olivine from water-saturated magmas, followed by amphibole-dominated fractionation in the middle-lower crustal region. The variation of Mg number (100*(Mg/(Mg + Fe2+)) with increasing SiO2 is an excellent indicator to test this hypothesis and has been used with great success to access direct mantle equilibration models [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 effect of fractionation processes can be readily detected in the composition of the entrained cumulate material. The Mg number of amphiboles during the fractionation process may record heterogeneities in their parental melts as it is closely related to the Fe-Mg exchange between amphibole and liquid (Kd∑Fe/Mg). Nonetheless, variables such as fO2, fH2O [Grove et al., 2003], and to a lesser extent temperature [Alonso-Perez et al., 2009] may also affect the Mg number of the crystallizing amphiboles. Amphiboles derived as initial fractionating products of a high Mg andesite or a basaltic andesite typically have Mg numbers of ∼79 [Moore and Carmichael, 1998; Grove et al., 2003, 2005]. The monomineralic coarse-grained cumulates from Baru volcano have lower Mg numbers (∼75) than would be expected for the direct fractionation products of a mantle-derived melt. However, the Mg number of amphiboles in the monomineralic cumulates are elevated when compared to initial amphibole fractionation products of a less Mg-rich andesite [Alonso-Perez et al., 2009], and are directly comparable to the early fractionation products of an amphibole-dominated andesite from the Panama Canal Region [Rooney et al., 2010]. Grove et al.  noted that plagioclase was an important phase coexisting at low pressures with amphibole (Mg number ∼75). The absence of plagioclase in equilibrium with the monomineralic cumulates at Baru volcano supports that these cumulates do not represent an evolved portion of the fractionating assemblage described by Grove et al. . The most plausible model for the generation of the lower Mg number (∼75) in the amphiboles from the monomineralic cumulates, is that the parental magmas of Baru volcano fractionated mafic phases in the upper mantle/lower crust boundary [Rooney et al., 2010]. This is supported by the observation of remnants of these phases in the some of the coarse-grained cumulates from Baru. After separation of these mafic phases, the slightly more evolved melt would yield monomineralic amphibole cumulates at middle-lower crustal levels with the appropriate Mg numbers of those observed at Baru.
To achieve rapid and extensive crystallization of amphibole-rich cumulates from hydrous basalts decompression models have been proposed [Tamura and Tatsumi, 2002; Blundy et al., 2006]. Nonetheless, cooling of water-rich magmas would also produce large volumes of amphibole over time scales that are more in accord with U series temporal estimates of magmatic differentiation processes in arcs [Davidson et al., 2007]. In the case of decompression driven amphibole crystallization, the increase of crystallinity in the magmas diminishes the potential of these magmas to reach the surface due to increasing viscosity with decreasing H2O content [Tamura and Tatsumi, 2002; Barclay and Carmichael, 2004]. The crystallizing assemblage preserved within the coarse-grained cumulates at Baru volcano as well as the relative P-T estimates summarized in Figure 12, are consistent with these cumulates being part of crustal “hot zones” where large amounts of amphibole fractionate in response to decompression in water saturated magmas or to extensive cooling. Moreover, the host andesitic magmas at Baru volcano can effectively be correlated to the liquid residue derived from the fractionation of amphibole-rich cumulates in crustal hot zones.
An important consequence of early removal of large quantities of amphibole from water-saturated magmas is that this process would rapidly increase the SiO2 content of the residual magmas [Carmichael, 2002]. In comparison with gabbroic assemblages (olivine + pyroxene + plagioclase), amphibole fractionation will rapidly elevate differentiation indices with relatively less mass removal. 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 lower volume of mass removed from the hydrous magmas [Rooney et al., 2010]. Extensive amphibole fractionation may explain why samples from Baru volcano and other volcanic deposits in the Panamanian arc exhibit some of the most depleted REE patterns in the entire Central American Volcanic Front.
7.2. Role of Garnet Fractionation
In addition to amphibole fractionation, our model required that some garnet (1.4%) needed to be removed from the parental melts to reproduce the HREE and Y depletion observed at Baru magmas (Figure 9). Cryptic garnet fractionation is a viable process to generate HREE and Y depletion due to the high partition coefficients of these elements in garnet [Rodríguez et al., 2007; Macpherson, 2008; Larocque and Canil, 2010]. In our trace element modeling (section 5) bulk partition coefficients for HREE and Y in garnet between 2.45 and 8.2 are needed to fully explain the observed HREE and Y depletions between 54 and 64 wt % SiO2. These values are well within experimental and calculated bulk partition coefficient values on basaltic liquids [McKenzie and O'Nions, 1991; Hauri et al., 1994; Johnson, 1994, 1998]. On the other hand, if amphibole (∼85 wt % of the fractionated assemblage) were the only phase to significantly retain HREE and Y, amphibole melt partition coefficients of ∼4.5–6.8 would be required. Experimentally determined partition coefficients for amphibole in basaltic to basaltic andesitic liquids are below this range (0.3 to 4.3) [Dalpé and Baker, 2000; Bottazzi et al., 1999].
Cryptic garnet fractionation is also supported by the metaluminous character of the Quaternary volcanism at Baru (Figure 4b). High-pressure fractionation of garnet inhibits the evolution toward peraluminous liquids which are produced at restricted pressure range (0.7–1.2 GPa) [Sisson and Grove, 1993; Müntener et al., 2001]. At shallower depths the metaluminous character of the liquid is preserved by the crystallization of An-rich plagioclase [Alonso-Perez et al., 2009]. In addition, the decreasing FeO*/MgO ratio with increasing SiO2 (Figure 13c) is also consistent with garnet fractionation. Garnet has a high Fe/Mg solid/liquid partition coefficient and tends to produce liquids with high Mg contents, therefore during garnet fractionation FeO*/MgO decreases with increasing SiO2. The stability of garnet in experiments on synthetic andesites have concluded that under water saturated conditions 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].
7.3. Magma Ascent Rates at Baru Volcano
The textural evidence preserved within the cumulate groups suggests that the ascent of the Baru volcano magmas may be rapid. Rutherford and Hill  studied the stability of hornblende under experimental conditions and estimated that during adiabatic ascent reaction rims will form around amphibole crystals in only 4 days after crossing the 0.160 GPa pressure boundary. At shallower depths amphibole is no longer stable. The reaction rim would typically have small plagioclase, pyroxene, Fe-Ti oxides and is a result of the decrease of H2O content of coexisting melt during magma ascent [Garcia and Jacobson, 1979]. No reaction rims are evident in any of the amphiboles of the Quaternary products of Baru volcano, supporting a rapid ascent hypothesis in terms of the shallower portion of the magmatic plumbing system.
7.4. Formation of Amphibole Cumulates and Nature of the Panamanian Crust
Amphibole-rich cumulates such as those described from Baru volcano may provide a way to generate the evolved composition of the continental crust. Such cumulates, have been described by Jull and Kelemen  and Dufek and Bergantz  as key components of the continental crust cycle (instability and accretion). The importance of mafic cumulates, cannot be underestimated, as mass balance models predict that their volume should be roughly twice the volume of derivative intermediate to silicic magmas that are either erupted or stall on their path to the surface [Foden and Green, 1992]. In order to generate the SiO2-rich continental crust, these mafic cumulates must be removed from the island arc crust. Delamination and convective erosion, which rely upon density and strength instability in the arc lithosphere, have been invoked as potential mechanisms for the removal of subarc cumulates. In gabbroic cumulates the assemblage would not be dense enough to recycle into the mantle unless a partial melting event depletes the cumulate in the plagioclase component resulting in a higher-density assemblage [Dufek and Bergantz, 2005]. In contrast, amphibole-rich cumulates, which contain little or no plagioclase are ideal candidates to develop a density instability and be returned to the mantle [Müntener et al., 2001; Dufek and Bergantz, 2005; Davidson et al., 2007]. Amphibole also plays a key role in models of convective erosion by promoting destabilization of the arc lithosphere [Arcay et al., 2006]. Specifically, amphibole-rich layers in the subarc lithosphere are ideal sites for shearing to occur due to an abrupt decrease in viscosity within the amphibole layer derived from increased water contents [Arcay et al., 2006]. The ubiquitous presence of amphibole-rich cumulates in the host magmas at Baru volcano enhances the importance of amphibole-dominated fractionation as a dominant process for generating continental crust in mature island arcs.
Amphibole-rich cumulates have been noted elsewhere along the Panamanian arc (El valle volcano [Hidalgo, 2007] and the Panama Canal Zone [Rooney et al., 2010]), indicating that magmatic plumbing systems in the region share some commonalities. The widespread role of amphibole in the Panamanian lithosphere may have its origin in the prevalent tectonic conditions of the arc. The arrival of the Cocos and Coiba ridges to the Middle American Trench generated significant perturbations of the crustal and upper mantle structure of the region. Specifically, enhanced plate coupling, crustal shortening, changes in the stress conditions of the overlying plate, and arc-parallel movement of material have been invoked [Hoernle et al., 2008; LaFemina et al., 2009]. Under these conditions rapid magma ascent would be impeded during the initial stages of the ridge collision due to the sealing of usual crustal transfer structures, causing stalling and differentiation of magmas at lower crustal levels [e.g., Chiaradia et al., 2009; Rooney et al., 2010]. Magma stalling impedes the crystallization of anhydrous phases, which are favored during decompression and degassing [Annen et al., 2006], and cooling promotes the widespread precipitation of amphibole in hydrated magmas [e.g., Carmichael, 2002]. Furthermore, magma stalling and fractionation at deeper crustal levels within the garnet stability field successfully explains the origin of adakite-like volcanism in the Central American arc.
Quaternary magmas erupted at Baru volcano are a product of deep fractionation processes that probe an amphibole-rich hot zone at middle-lower crustal depths. One variety of the ubiquitous cumulates sampled at Baru volcano records high temperatures and pressures while the other variety distinctively shows textures and barometric data consistent with rapid decompression that triggered crystallization processes. The most likely scenario is that an interstitial melt was separated from a mush/magma accumulation zone in the middle-lower crust, carrying with it coarse-grained cumulates that either were disaggregated to become antecrysts within the host or were preserved as cumulates. Soon after, this mixture of melt + antecrysts + cumulates reached shallower levels where previous magma pulses had formed a fined grained cumulate layer by decompression induced crystallization. The fined grained cumulate layer was then eroded and brought to the surface rapidly.
This model for generating the chemical diversity observed in the erupted products of Baru volcano agrees with models that propose that the chemical diversity in arc magmas is largely acquired in the middle-lower crust, whereas textural diversity is related to shallow level processes [e.g., Annen et al., 2006; Davidson et al., 2007]. Moreover, the present study succeeds in linking the chemical variation of large andesitic-dacitic volcanic eruptions to voluminous mafic cumulates. The existence of amphibole-rich cumulates in the subarc lithosphere has been long predicted but proving their influence in the magmatic differentiation process on an active arc has been elusive.
The arrival of the Cocos and Coiba ridges to the Middle American Trench had significant implications for subduction zone magmas in the region. The absence of amphibole-rich cumulates and evolved magma compositions in older western Panamanian volcanics (Neogene) and the recent appearance of amphibole-rich magmas, might be an important indicator of an abrupt transition from to protracted storage in lower crustal levels. Protracted storage allowed for extensive amphibole crystallization that 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. Moreover, 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.
The authors wish to thank Angel Rodriguez and Karla Black for their invaluable help during the sample collection process; without their help and advice it would have been impossible to cover as much area as we did. We like to thank Eduardo Camacho for his contribution to the sample collection logistics and advice. We wish to thank Tom Vogel, Dave Szymanski, Matt Parsons, and Carl Henderson for sample preparation and analytical assistance. We are grateful to Jim Vallance and David Sherrod for their invaluable advice prior to the field season. We wish to acknowledge the very constructive and detailed reviews of two anonymous referees and the careful editorial handling of Joel Baker.