Experimental Constraints on Dacite Pre-eruptive Magma Storage Conditions beneath Uturuncu Volcano

Abstract

Cerro Uturuncu is a long-dormant, compositionally monotonous, effusive dacitic volcano in the Altiplano–Puna Volcanic Complex (APVC) of SW Bolivia. The volcano recently gained attention following the discovery of an ∼70 km diameter Interferometric Synthetic Aperture Radar (InSAR) anomaly roughly centred on its edifice. Uturuncu dacites, erupted over the past ∼1 Myr, invariably have a phase assemblage of plagioclase–orthopyroxene–biotite–ilmenite–magnetite–apatite–zircon and rhyolite glass. To better constrain storage conditions of the dacite magmas and to help understand their relationship with the observed deformation, petrological experiments were performed in cold-seal hydrothermal vessels. Volatile-saturated (P_H2O = P_TOTAL and P_H2O + P_CO2 = P_TOTAL) phase equilibria experiments were run between 50 and 250 MPa and 760 and 900°C at fO₂ ∼ Ni–NiO. Two synthetic starting compositions were investigated based on a typical Uturuncu dacite whole-rock and its rhyolitic groundmass glass. Pre-eruptive magma storage conditions have been estimated by comparing results from the experiments with natural phase assemblages, modes, and mineral and glass compositions. H₂O-saturated experiments constrain storage pressures to 100 ± 50 MPa, equivalent to 1·9–5·7 km below surface. In the dacite, natural phase assemblages are reproduced at 870°C, 100 MPa with both orthopyroxene and biotite stabilized concurrently. Natural glass chemistries are most closely replicated at 50 MPa at 870°C, reflecting the role of decompression crystallization prior to eruption. In H₂O-saturated rhyolite experiments the natural phase assemblage is most closely replicated at 870°C, 50 MPa. Isothermal, mixed volatile dacite experiments at 870°C further constrain storage pressures to 110 ± 10 MPa. Assuming that there has been no dramatic change in the eruptive behaviour of Uturuncu in the last 270 kyr, pre-eruptive storage of dacite magmas at ∼100 MPa precludes their role in producing the large diameter deformation anomaly. If deformation is a result of magmatic intrusion, then intrusion of less evolved magmas into deeper, mid-crustal storage regions is a more probable explanation. Intrusion within the Altiplano–Puna Magma Body (APMB), the extent of which is roughly coincident with the APVC, is most likely. It is proposed that dacite magmas form from andesitic parents, via fractionation and/or assimilation, within the APMB. Dacites then rise buoyantly to shallow storage levels where they stall and crystallize prior to eruption. Microlites form during subsequent ascent from the storage region to the surface.

INTRODUCTION

Arc volcanoes exhibit various styles of eruptive behaviour and generate hazards of different types and magnitude. Erupted magmas can be sourced from a range of depths with erupted volumes ranging from <1 to several thousand km³ (dense rock equivalent). Investigation of physico-chemical attributes of sub-volcanic magma storage systems allows the potentially hazardous effects of volcanic activity to be constrained more accurately.

Magmatism in the Central Andes is a result of eastwards subduction of the Nazca oceanic plate beneath the South American continent, a continuing process since Jurassic times [see summary by de Silva & Gosnold (2007)]. In the Altiplano–Puna Volcanic Complex (APVC, Fig. 1; de Silva, 1989) of the Central Andes voluminous ignimbrites erupted in the Neogene to Recent are widespread. However, currently the most obvious expression of active volcanism in this region are young, largely effusive volcanic centres.

Recent Interferometric Synthetic Aperture Radar (InSAR; Pritchard & Simons, 2002, 2004) and seismic studies (Jay et al., 2012) of Cerro Uturuncu, a Pleistocene dacitic volcano in the APVC in SW Bolivia (Fig. 1), reveal evidence of an ∼70 km diameter region of ground deformation centred approximately at Uturuncu. Deformation has been measured between 1992 and 2006 with a central uplift rate of 1–2 cm a⁻¹ (Sparks et al., 2008). Inverse models of the deformation data suggest a magmatic source with intrusion in the mid- to upper crust (minimum ∼11–17 km below surface; Pritchard & Simons, 2002, 2004; Sparks et al., 2008; Fialko & Pearse, 2012; del Potro et al., 2013; Hickey et al., 2013).

Eruptive products at Uturuncu are dacitic and entirely effusive with no evidence of explosive activity. Existing ⁴⁰Ar/³⁹Ar ages of lavas and domes show that the volcano was active between 890 and 271 ka and has apparently been dormant ever since (Sparks et al., 2008). The edifice itself is constructed on a plain composed mainly of Vilama and Guacha ignimbrite deposits erupted c. 8·41 and 5·65 Ma respectively (Salisbury et al., 2011); its SW flanks directly overlie 8 Ma post-Vilama lavas (S. de Silva, personal communication).

The Altiplano–Puna Magma Body (APMB), a regional low-velocity zone identified by seismic and magnetotelluric data (Schilling et al., 1997; Schmitz et al., 1997; Chmielowski et al., 1999; Brasse et al., 2002), is proposed as the source of much of the magma erupted in the APVC (de Silva et al., 2006). The top of the APMB has been detected around 17–19 km depth below the surface (Chmielowski et al., 1999; Leidig & Zandt, 2003; Zandt et al., 2003) in the region surrounding Uturuncu (Fig. 1). The base is poorly defined but estimates of its thickness range from ∼1 km (Chmielowski et al., 1999; Leidig & Zandt, 2003) to 10–20 km (Yuan et al., 2000).

The InSAR data raise a number of questions about present and future magmatic activity beneath Uturuncu and implications for eruptive hazard. If current ground deformation at Uturuncu is indeed caused by magma intrusion, can we expect to see further effusive eruptions of dacites or are larger-scale explosive eruptions, similar to ignimbrite-forming events, more likely? Alternatively, is magma ponding in the crust, forming plutons with little likelihood of volcanic eruptions? Do multiple reservoirs exist or are eruptions sourced from a single, large source region? What are the likely pre-eruptive volatile contents of these magmas? To better understand past volcanism at Uturuncu and the potential for future activity, knowledge of the magmatic system and how it operates is required. Through detailed experimental study we have attempted to determine the depths at which previously erupted dacite magmas were stored prior to eruption. Integration with inverse modelling of the InSAR data allows us to assess whether intrusion of magmas into this storage region could be responsible for the observed deformation.

Petrology and petrography of Uturuncu dacites

Sparks et al. (2008) and Muir et al. (in press) have presented the petrology of a total of ∼45 Uturuncu lava and dome samples. Only the key features are repeated here. All samples are porphyritic with abundant plagioclase and orthopyroxene both as phenocrysts (>100 μm), microphenocrysts (>30 μm) and microlites (<30 μm). Plagioclase phenocrysts commonly have sieve-textured cores and crystals with normal, reverse, oscillatory and patchy zoning are observed. Orthopyroxene frequently exhibits reverse zoning, but crystals with normal zoning also occur. Ilmenite and magnetite are present in all samples along with minor apatite, zircon, monazite and rhyolitic glass. Biotite is present only as phenocrysts and is commonly reacted with a rim of Fe–Ti oxides, orthopyroxene and apatite. Within a single thin section biotite crystals display varying degrees of reaction from unreacted to completely reacted. Quartz occurs as a minor phase and usually has an embayed form, occasionally with reaction rims of clinopyroxene. Hornblende and clinopyroxene phenocrysts are scarce. Phenocryst contents range between 13 and 34 vol. % (Fig. 2). Microlite contents vary from 3 to ∼40 vol. % and total vesicle-free crystallinities are between 31 and 64 vol. % (Muir et al., in press).

At Uturuncu the shallow storage equilibrium phase assemblage must be distinguished from xenocrystic and antecrystic components incorporated as magmas transited through the thick crust. Isotopic studies of O, Sr, Nd and Pb indicate that crustal contamination plays a major role in the generation of Andean magmas (James, 1982; Hildreth & Moorbath, 1988; Folkes et al. 2013; Michelfelder et al., 2013), further complicating the definition of equilibrium conditions (e.g. McLeod et al., 2012).

Mineral compositions within dacitic Uturuncu lavas and domes span a wide range (Muir et al., in press). Molar anorthite content (An) of plagioclase cores, rims and microlites are An_46–90. Molar magnesium-number [Mg#, 100MgO/(MgO + FeO*)] values of orthopyroxene cores, rims and microlites are Mg#_32–88. Plagioclase phenocryst rims and microlites are skewed towards An₇₄ (Fig. 2). Orthopyroxene phenocryst cores and rims are bimodal with peaks around Mg#₅₂ and Mg#₆₈ (Fig. 2); microlite compositions span the entire compositional range but are commonly around Mg#₆₈.

Orbicules and glomerocrysts of norite (i.e. plagioclase–orthopyroxene-bearing rock) are commonly found in the dacites. Norite plagioclase has some of the highest An contents observed at Uturuncu and spans a relatively limited compositional range of An_75–95 (Muir et al., in press). Some lavas contain andesite enclaves with plagioclase that is some of the most An-rich measured from Uturuncu, but within the range of plagioclase compositions measured in the host dacite lavas. Orthopyroxene core and microlite compositions in andesitic enclaves vary widely and are similar to those analysed in dacite host lavas.

The absence of dominant rim and microlite compositions for plagioclase and orthopyroxene, and the wide range of mineral compositions overall, make equilibrium mineral compositions difficult to define. Although groundmass glass compositions vary between lavas and domes they are all rhyolitic and are homogeneous within a given sample, indicating that equilibrium in the magmatic system is melt-dominated (Pichavant et al., 2007). Uturuncu magmas erupted effusively and so probably ascended slowly enough for melt compositions to evolve as a result of degassing-induced crystallization. Consequently, experimental reproduction of natural glass compositions is useful but not an absolute constraint on pre-eruptive magma storage conditions.

Sample UTDM41B, a multiphase, but aphyric, glassy, relatively young lava sample, was selected as a starting composition for our experiments. This sample is less evolved than most Uturuncu lavas and is microlite-free (Fig. 3), so it is likely to have been less affected by late-stage decompression crystallization. UTDM41B glass compositions have limited chemical variability, providing a clear target for experimental reproduction (Table 1).

Average UTDM41B plagioclase cores are An_{70 ± 9} and rims are An_{72 ± 8}. Orthopyroxene cores have Mg#_{51 ± 3}, and rims have Mg#_{56 ± 2}. Microlites of plagioclase and orthopyroxene are scarce in this sample and have not been analysed. The vesicle-free crystallinity of UTDM41B is 37 vol. % based on point counting using a polarized light microscope. Considering that sieve-textured plagioclase cores—clearly out of equilibrium with the surrounding melt—constitute 5 ± 3 vol. % in UTDM41B (8–30% of the total crystallinity), the equilibrium crystallinity is likely to be between 26 and 34 vol. %.

Temperature, fO₂ and PH₂O constraints from natural samples

Average temperatures and oxygen fugacity (fO₂) calculated from coexisting titanomagnetite–ilmenite pairs in 18 Uturuncu natural samples are 858 ± 35°C (Fig. 2) and fO₂ from nickel–nickel oxide (NNO) + 1 log to NNO – 1 log respectively (Muir et al., in press). For UTDM41B an equilibrium magmatic temperature of 873 ± 13°C and fO₂ ∼NNO (log₁₀ −12·5 ± 0·2) were calculated using ILMAT (Lepage, 2003) with the Fe–Ti exchange thermometer and oxygen barometer of Andersen & Lindsley (1988) and the solution model of Lindsley & Spencer (1982). Using mean compositions of all ilmenite and magnetite compositions for UTDM41B, the Ghiorso & Evans (2008) Fe–Ti exchange model gives a much higher temperature of 931°C with fO₂ of ∼NNO. These data are compiled from oxides that are near, but not in contact with, each other and include only analyses that satisfy the Mg–Mn equilibrium criterion of Bacon & Hirschmann (1988). Orthopyroxene–melt thermometry calculations are in very good agreement with temperatures from coexisting oxides: average temperatures for UTDM41B orthopyroxene that satisfy the equilibrium criteria [K_D(Fe–Mg)^opx–liq = 0·29 ± 0·06] outlined by Putirka (2008), are 861 ± 6°C (Muir et al., in press).

Plagioclase-hosted melt inclusions are all rhyolitic. Volatile compositions from Muir et al. (in press) can be used to calculate minimum trapping pressures, providing constraints on the magmatic conditions under which plagioclase phenocrysts crystallize. Using VolatileCalc (Newman & Lowenstern, 2002) assuming a magma temperature of 870°C, average H₂O of 3·2 ± 0·7 wt % and CO₂ of 160 ppm results in minimum trapping pressures of 50–119 MPa with coexisting equilibrium, molar fluid compositions between xH₂O = 0·87 and 1·00. Melt inclusions are susceptible to various post-entrapment modifications that can vastly reduce the reliability of calculated pressures [see review by Lowenstern (1995)]. Nevertheless, these conditions provide a useful starting point for experiments.

Previous experimental studies

Phase equilibria experimental data exist for several similar silicic arc volcanic systems including San Pedro (Costa et al., 2004) and Chaitén, Chile (Castro & Dingwell, 2009); Pinatubo, Philippines (Scaillet & Evans, 1999); Novarupta (Hammer et al., 2002) and Aniakchuk, Alaska (Larsen, 2006); Santorini, Greece (Cottrell et al., 1999); Fish Canyon, Colorado (Johnson & Rutherford, 1989); Unzen, Japan (Holtz et al., 2005); and Mount St. Helens, Washington (Rutherford et al., 1985). Most of these systems have similar mineral assemblages to Uturuncu but include hornblende ± biotite, apart from Santorini which has neither. Uturuncu is distinct in being dacitic with only biotite as a hydrous mineral phase. Chaitén rhyolites also crystallize only biotite. Orthopyroxene, a major phase at Uturuncu, is absent at Pinatubo and Fish Canyon, but quartz, a trace phase at Uturuncu, is stable in both these systems.

UTDM41B whole-rock SiO₂ (67·0 wt %) is similar to San Pedro, Novarupta and Pinatubo. However, K₂O (4·5 wt %) and TiO₂ (1·0 wt %) are high and Na₂O (2·1 wt %) is markedly low compared with these systems (Fig. 4). Differences in whole-rock chemistry will account for some dissimilarity in phase assemblages. For example, the general absence of hornblende, notable in Uturuncu samples compared with other silicic systems in volcanic arcs, may be due to whole-rock Na₂O being too low to saturate hornblende (Sisson & Grove, 1993). Low Na₂O may also explain the presence of relatively An-rich plagioclase in such evolved magmas.

Estimates of magma storage pressures in these previous experimental studies range from 300 MPa for Unzen dacites (Holtz et al., 2005) to as low as 50 MPa for Santorini rhyodacites (Cottrell et al., 1999) and Novarupta dacites (Hammer et al., 2002). Estimated magmatic temperatures are as low as 760°C for Fish Canyon Tuff quartz-latites and Pinatubo dacites, whereas maximum temperatures of 930°C are estimated for Mount St. Helens dacites.

EXPERIMENTAL AND ANALYTICAL METHODS

Cold-seal pressure vessels

Experiments in this study were designed to constrain Uturuncu magma storage conditions prior to eruption. The majority of experiments were run in cold-seal pressure vessels using H₂O as the pressurizing medium as described by Carroll & Blank (1997) and Humphreys et al. (2008). Capsules were made from 2 or 3 mm external diameter Au tubing cut to lengths of ∼10 mm. For the majority of runs, two capsules containing finely powdered synthetic starting materials were included: one a dry glassy dacite (STXB4) of similar bulk composition to UTDM41B, the other a dry glassy rhyolite (STXG4) with composition based on the matrix glass of UTDM41B (Table 1). These synthetic bulk and glass compositions provide end-members of the Uturuncu chemical system; the reactive magma [in the sense of Pichavant et al. (2007)] will lie somewhere between these two. At equilibrium conditions at the correct magma storage conditions, the melt compositions of the charges containing the bulk and rhyolitic glass starting materials should be the same—the rhyolitic glass with only limited crystallization of the expected microlites and phenocryst rims, the bulk with extensive crystallization of the same phases and a total crystallinity similar to that of the natural sample. Thus the equilibrium storage conditions should be bracketed by the two sets of experiments: at all other conditions the phase assemblage and melt chemistry of rhyolite and dacite charges would not necessarily be expected to match. Equilibrium crystallinity is expected to be ∼26 to 34 vol. % when accounting for antecrysts and xenocrysts as well as any syn-eruptive decompression crystallization.

Synthetic starting materials were made by mixing appropriate amounts of oxides, carbonates and phosphates (SiO₂, TiO₂, Al₂O₃, FeO, MnO and MgO; CaCO₃, Na₂CO₃ and K₂CO₃; and Ca₃PO₄), which were subsequently fused in a platinum crucible at 1 atm for several hours at temperatures between 1350 and 1400°C for the dacite and at 1450°C for the rhyolite. Oxygen fugacity (fO₂) was fixed to NNO using appropriate gas mixes of CO and CO₂. Post-fusion, melts were quenched to glass by dropping the crucible in a H₂O bath. To ensure homogeneity, glasses were subsequently ground in an agate mortar and fused multiple times before being analysed by electron microprobe. For most major oxides, starting compositions are within 2σ of the desired, natural values as determined by X-ray fluorescence (XRF) and electron microprobe (EMP) analyses (Table 1).

H₂O-saturated (P_H2O = P_TOTAL) experiments were run with distilled H₂O added to each capsule by pipette along with 10–15 mg of powdered starting material. The fraction of H₂O in the charges never exceeded 10 wt % (Table 2). Capsules were crimped, welded shut and weighed before and after heating at 200°C to verify weld integrity. Where possible, experiments were run for 7 days to ensure equilibrium conditions were reached. Above 870°C experiments were run for around 72 h. A rapid quench device was used for all experiments in which sample rods are extracted from the hotspot into the cold seal using a magnet under near isobaric conditions.

Cold-seal intrinsic fO₂ was calculated from an experiment run at 870°C, 100 MPa using a gold capsule with a physical mixture of ilmenite and magnetite plus some H₂O. Subsequent analysis of these coexisting oxides by electron microprobe allowed calculations of fO₂ to be made using the ILMAT program of Lepage (2003) with the Fe–Ti exchange thermometer and oxygen barometer of Andersen & Lindsley (1988) and the solution model of Lindsley & Spencer (1982). With temperature fixed to 870°C, calculated log₁₀fO₂ is −13·02 (NNO − 0·3).

Mixed volatile experiments (P_H2O + P_CO2 = P_TOTAL) were performed to assess the effect of CO₂ on phase equilibria. Relative proportions of H₂O and CO₂ required for a given initial fluid xH₂O were calculated using the Papale et al. (2006) solubility model. Capsules were prepared by the same method as H₂O-saturated experiments using powdered synthetic starting material with H₂O added by pipette and CO₂ added as silver oxalate (Ag₂C₂O₄). Au capsules ∼15 mm long were used for the same amount of starting composition—the additional space provided by the longer capsules allowed any volume changes from exsolution of vapour to be accommodated, preventing capsules from rupturing during runs.

Post-experiment, all capsules were re-weighed and checked for the presence or absence of liquid H₂O upon opening. For each capsule, the run product was mounted in epoxy resin and polished before being studied by scanning electron microscope and analysed with the EMP.

Gas-buffered 1 atm furnace experiments

Experiments at 1 atm, buffered at NNO using controlled fluxes of CO and CO₂ gas, were performed between 1000 and 1200°C to constrain the low-pressure liquidus temperature of the system and better define the phase diagram. Crimped Au₈₀Pd₂₀ capsules 1·5 cm long were hung on platinum wire in the furnace hotspot for ∼48 h. Run products were mounted and polished using the same method as described for cold-seal hydrothermal experiments.

Analytical methods

Scanning electron microscope (SEM)

A Hitachi S-3500N SEM with a silicon drift X-ray detector, at the University of Bristol, was used to acquire back-scattered electron (BSE) images of the experimental charges. Energy-dispersive spectrometry (EDS) phase maps were acquired for all dacite experimental run products to allow mineral modes and crystallinities to be measured using the method presented by Muir et al. (2012). Minerals down to ∼1 μm across can be identified using this method and, assuming the 1024 × 1024 pixel area is representative of the whole sample, minimum random errors are less than 1 vol. % (Van der Plas & Tobi, 1965). However, non-systematic, operator errors are not incorporated in this uncertainty estimate.

Electron microprobe (EMP) analysis

Major elements of minerals and glasses in experimental charges were analysed using a Cameca SX100 five-spectrometer microprobe at the University of Bristol, using the methods described by Muir et al. (in press). Glass was analysed using 15 kV accelerating voltage, 2 nA beam current and a 10 μm beam diameter to minimize alkali migration. Crystals in natural samples and all experimentally synthesized plagioclase and orthopyroxene were analysed at 15 kV with a 10 nA beam current and a focused beam.

Fe–Ti oxides synthesized experimentally are typically 1–2 μm in width and were analysed at high spatial resolution using a five-spectrometer JEOL JXA-8530F field-emission gun microprobe at the University of Bristol with a focused beam at 9 kV and 20 nA. Orthopyroxene crystals smaller than 5 μm width were analysed using the same beam setup.

RESULTS

Thirty-seven H₂O-saturated experimental runs were conducted on dacitic and rhyolitic starting compositions between temperatures of 760 and 900°C and at pressures of 50–250 MPa (Table 2). Five H₂O-undersaturated mixed volatile experiments were performed using the dacite starting composition along with three 1 atm, liquidus-defining experiments.

Without exception, natural Uturuncu mineral assemblages include plagioclase, orthopyroxene, ilmenite, Ti-magnetite, biotite, apatite and zircon and a rhyolitic groundmass glass. Trace quartz is present in some lavas and domes and is invariably resorbed. Petrological experiments were conducted with the aim of reproducing this phase assemblage from glassed synthetic starting materials; that is, without any pre-existing natural crystal nuclei. Run products consist of vesicles, glass and mineral phases including plagioclase, orthopyroxene, biotite, ilmenite, Ti-rich magnetite and apatite (Table 3; Fig. 5). Crystal sizes range from <1 μm to 10 μm in width, orthopyroxene being the coarsest and biotite and Fe–Ti oxides the finest. EMP analyses of all minerals large enough to analyse and glasses are provided in Supplementary Data (Supplementary Data is available for downloading at http://www.petrology.oxfordjournals.org).

Evidence that equilibrium conditions were reached in the majority of the experiments includes (1) homogeneous glass compositions, (2) euhedral, non-skeletal equant or tabular crystal habits, and (3) predictable variations in crystallinity with decreasing pressure and temperature. However, mineral and glass compositions do not always vary predictably with pressure and temperature, and reverse zoning in orthopyroxene crystals in STXB4-90-15 at 900°C, 150 MPa is testament to incomplete attainment of equilibrium in some runs.

Dacite H₂O-saturated phase equilibria experiments

Equilibrium phase assemblages for the whole-rock dacite composition (STXB4) are shown in Fig. 6. Fe–Ti oxides and plagioclase are present in all sub-liquidus dacite experiments. Where present, biotite is easily identified on element maps by its high K₂O and MgO content. However, owing to its small size (<1 μm width) biotite could not be analysed by EMP. Most experimental runs have two Fe–Ti oxides: magnetite and ilmenite.

At pressures less than ∼100 MPa orthopyroxene is stable at 870°C but higher temperatures are required to saturate orthopyroxene at greater pressures. At higher pressures and lower temperatures biotite replaces orthopyroxene as the stable mafic phase. Apatite saturates at around 870°C, irrespective of pressure; its resorption in natural samples (Muir et al., in press) is therefore strongly suggestive of heating above this temperature. Quartz is stable only in STXB4-78-15 at 150 MPa and 780°C, a temperature much lower than the orthopyroxene-out phase boundary. Again, quartz resorption in natural samples is suggestive of heating of magmas previously cooled below 780°C.

Crystallinity in the dacite experiments decreases from a maximum of 32 vol. % at 780°C, 150 MPa, to a minimum of 14 vol. % at 870°C, 200 MPa and 100 MPa (Fig. 7); however, various experiments, particularly those at 900°C over a range of pressures and at 870°C at 50 MPa, do not conform to this trend. Crystallinities of all H₂O-saturated runs are lower than observed in the natural rocks—UTDM41B has a total crystallinity of 37 vol. %. This is, however, expected as xenocrysts and antecrysts with sieve-textured cores constitute 8–30% of the total crystallinity of UTDM41B. Although UTDM41B is a particularly glassy sample, microlites and phenocryst rims formed during late-stage decompression from storage regions will also contribute to the total crystallinity in the natural rocks. At 870°C, the magmatic temperature calculated from coexisting oxides in UTDM41B, the natural crystallinity is most closely replicated in experiments performed at 50 MPa (26 vol. %).

Experimental plagioclase compositions vary between An₅₈ and An₇₆ (Fig. 7) and are within the broad range of compositions measured in natural samples (An_46–90; Muir et al., in press). In general, at a given pressure, An content decreases with temperature (Fig. 7) although at 50 MPa this is not the case. Compared with plagioclase cores and rims from UTDM41B (An_{70 ± 9} and An_{72 ± 8} respectively) plagioclase synthesized at 870°C, 50 MPa is closest in composition, with An_{73 ± 8}. Plagioclases in runs at 870°C, 200 MPa and 850°C, 125 MPa are also similar, with An_{69 ± 5} and An_{68 ± 3} respectively.

Experimental orthopyroxene Mg#s range from 45 to 76 and are within the range observed in natural samples (Mg#_32–88; Muir et al. in press). Orthopyroxene synthesized at 900°C, 50 MPa (Mg#₅₈) most closely resembles UTDM41B orthopyroxene rim compositions (Mg#_{56 ± 2}).

Residual melt compositions vary over pressure–temperature (P–T) space as a result of changes in crystallinity and phase assemblage. With increasing P and T, K₂O and SiO₂ in the melt broadly decrease and Al₂O₃, CaO and FeO* increase. Figure 8 shows, however, that changes in glass compositions are largely non-uniform (e.g. anomalously high Al₂O₃ and low K₂O at 840°C, 100 MPa).

Dacite 1 atm experiments

Dacite experiments at 1 atm were used to constrain near-liquidus phase boundaries (Fig. 6). The powdered starting material failed to fuse and equilibrate at 1000°C. At 1100°C, the powder fused and plagioclase crystallized along with Fe–Ti oxides within a rhyolitic glass. At 1200°C, conditions are supra-liquidus.

Rhyolite H₂O-saturated phase equilibria experiments

Identification of phases was possible for all runs, except STXG4-76-25, using BSE images but, owing to very low crystal contents, phase proportions could not be quantified using EDS maps. Phase proportions were quantified only at 810°C, 50 MPa (Fig. 5c); this run has the highest crystallinity (7 vol. %) of all rhyolite experiments. Phase assemblages for rhyolite experiments are indicated in Table 3. Fe–Ti oxides were present in all runs; indeed, most runs crystallized only Fe–Ti oxides. Sanidine, which is not present in the natural samples, crystallized along with plagioclase at 810°C, 50 MPa and at 780°C, 200 MPa, and in the absence of plagioclase at 780°C, 150 MPa. Quartz is present only at 780°C, 200 MPa. Biotite is not present in any of the rhyolite experiments (Table 3). Microprobe analysis of minerals in rhyolite experiments was not practicable as single crystals are too small. Although element maps were acquired for these samples to identify the phases present, given the low crystallinities it is feasible that minor phases could have been overlooked in some experiments. Microprobe analyses of glasses show very little variation in melt composition, in keeping with the low crystallinities.

A typical phase diagram with mineral stability fields is difficult to construct for these experiments because they ‘skim’ the liquidus surface over the P–T range investigated. STXG4-87-5 at 870°C, 50 MPa is the only experiment to crystallize orthopyroxene, a major phenocryst and microlite phase at Uturuncu; plagioclase, Fe–Ti oxides and apatite also crystallized in this charge. In a repeat experiment at 870°C, 50 MPa (STXG4-87-5-1), however, orthopyroxene was not present (Table 3). We suspect the orthopyroxene stability field lies within the experimental P–T error at these conditions. However, given the sluggish diffusion kinetics in these high-viscosity rhyolitic melts, the possibility that equilibrium was not attained cannot be excluded.

Mixed H₂O–CO₂ dacite phase equilibria experiments

Mixed-volatile dacite phase equilibria relations are displayed in an isothermal P–xH₂O (where xH₂O is the initial molar proportion of H₂O to CO₂ in the fluid added to the experimental charges) diagram at NNO in Fig. 9. Experiments were conducted at a range of pressures (50–200 MPa) and xH₂O from ∼0·7 to 0·9 at 870°C (Table 3). Crystallinity is higher in most mixed-volatile charges compared with H₂O-saturated equivalents and is similar to the UTDM41B total crystallinity in STXB4-87-5-xH90 (870°C, 50 MPa, xH₂O = 0·9).

Phase assemblages in mixed-volatile charges are the same as in H₂O-saturated experiments at equivalent temperatures and pressures. STXB4-87-5-xH90 (870°C, 50 MPa, xH₂O = 0·9), STXB4-87-10a-xH80 (870°C, 100 MPa, xH₂O = 0·8) and STXB4-87-10-xH90 (870°C, 100 MPa, xH₂O = 0·9) all crystallized both orthopyroxene and biotite (Fig. 9) along with plagioclase and Fe–Ti oxides.

In addition to temperature, biotite stability is known to be affected by xH₂O, with a maximum thermal stability at xH₂O = 0·5 (Maaløe & Wyllie, 1975; Puziewicz & Johannes, 1990; Scaillet et al., 1995). This was evident only in STXB4-87-10a-xH90, with 4·6 vol. % biotite compared with only 0·4 vol. % in the H₂O-saturated experiment at equivalent temperature and pressure. In STXB4-87-10a-xH80, however, biotite modes are almost the same as in the equivalent H₂O-saturated experiment, but orthopyroxene is much more abundant (2·7 vol. % in the mixed volatile run compared with 1·4 vol. % under H₂O-saturated conditions). Clemens & Wall (1981) observed similar minor effects of varying xH₂O on biotite stability.

DISCUSSION

Experimental phase equilibria constraints on magma storage

H₂O-saturated (P_H2O = P_TOTAL)

Comparison of experimental and natural Uturuncu petrological data allows magma storage conditions to be constrained provided due consideration is given to both the incorporation of xenocrysts or antecrysts and microlite crystallization during pre-eruptive magma ascent. Orthopyroxene and biotite coexist in all natural Uturuncu lavas and domes studied to date [∼45 thin sections in the studies by Sparks et al. (2008) and Muir et al. (in press)] and simultaneous saturation of these mafic phases is fundamental to constraining magma storage conditions.

For the dacite starting composition over most of the H₂O-saturated P–T space studied biotite crystallizes in the absence of orthopyroxene or vice versa. Orthopyroxene is stable at higher temperatures whereas biotite crystallizes at lower temperatures. Only at 870°C and 100 MPa are both phases reproduced concurrently (Fig. 6). A repeat experiment at 870°C and 50 MPa crystallized biotite in the absence of orthopyroxene, whereas the original experiment stabilized orthopyroxene with no biotite, indicating that the orthopyroxene-out and biotite-in phase boundaries are within experimental error at these conditions. The coexistence of orthopyroxene and biotite at 870°C is in good agreement with magmatic temperatures calculated for the natural samples using coexisting oxide and orthopyroxene–melt thermometry. At increasingly higher pressure, orthopyroxene is stable only at temperatures greater than 870°C, thus limiting storage conditions to 100 ± 50 MPa.

Our attempt to constrain equilibrium pre-eruptive storage conditions involved a bracketing approach with two distinct starting compositions: the dacite whole-rock and the rhyolitic groundmass glass. The required Uturuncu phase assemblage should be present in both the dacite and rhyolite experiments at equilibrium pre-eruptive P–T conditions where the saturation surface of the rhyolite system intersects that of the dacite. The low crystal fractions and small crystal sizes within the rhyolite charges create difficulties when trying to establish phase assemblages and modes, but most saturate only Fe–Ti oxides and plagioclase. In the rhyolite system the complete phase assemblage was not reproduced as biotite did not saturate. This is not completely unexpected because in the natural samples biotite is present only as phenocrysts, never as microlites precipitated directly from the groundmass melt. Phenocrysts are typically sub- to anhedral and commonly have reaction rims. Consequently, we would expect the equilibrium mineral assemblage of only microlites and phenocryst rims (i.e. plagioclase, orthopyroxene, ilmenite, magnetite, apatite and zircon) to saturate from the groundmass glass at equilibrium storage conditions. This assemblage (minus zircon) was observed only at 870°C, 50 MPa—the only rhyolite experiment that saturated orthopyroxene. The fact that orthopyroxene saturates concurrently at 870°C, 50 MPa in both the dacite and rhyolite charges at these conditions is remarkable and reinforces findings from dacite experiments that show pre-eruptive magma storage is relatively shallow.

Differences between the phase assemblage of the dacite and rhyolite experiments at 870°C, 100 MPa may be due to decompression crystallization that occurred as the magma ascended from its storage region to the surface. Phenocryst rim growth and microlite crystallization will have a relatively large effect on the groundmass glass composition and almost no effect on the bulk-rock owing to the difficulty of separating crystals from relatively viscous silica-rich melts. Therefore, the groundmass glass composition of the erupted rock is not likely to be representative of the melt at pre-eruptive storage conditions and so will not necessarily crystallize the same phase assemblage as the dacite at these conditions. Instead, we would expect that the required phase assemblage in the rhyolite experiments would be attained only at shallower pressures where the rhyolite melt composition is in equilibrium with phenocryst rims and microlites.

The wide range of phenocryst core, rim and microlite compositions in the natural samples (Fig. 2) precludes the experimental mineral compositions being used to further constrain magma storage conditions; experiments throughout the entire P–T space overlap the natural mineral compositions. Plagioclases synthesized at 870°C, 50 MPa are most similar to the compositions of cores and rims in UTDM41B, and orthopyroxene rim compositions are reproduced most faithfully at 900°C, 50 MPa.

With the addition of components such as sulphur (e.g. Costa et al., 2004) and fluorine (Holloway & Ford, 1975; Muñoz, 1984), biotite stability could potentially be extended to overlap that of orthopyroxene. Costa et al. (2004) showed how sulphur can increase the stability of biotite in silicic systems using Volcán San Pedro, Chile, as an example. Uturuncu melt inclusions (100 ± 200 ppm S, Sparks et al., 2008; 200 ± 320 ppm S, Muir et al., in press) are not particularly sulphur-rich compared with, for example, those of Volcán San Pedro (294 ± 46 ppm S), but are similar to those of Pinatubo (198 ± 26 ppm S, Gerlach et al., 1996). Hence, the increase in biotite stability from the presence of sulphur is expected to be less important at Uturuncu than in more sulphur-rich systems.

Quartz and zircon are present as minor phases (<1 vol. %) in most natural dacites. Our dacite phase equilibria experiments show that quartz is stable ∼120°C cooler than orthopyroxene at a given pressure (Fig. 6). The temperature gap between these two phases could not be bridged even in mixed volatile experiments where, owing to higher crystallinity, quartz would be expected to saturate at higher temperatures. In natural rocks the embayed form of quartz and its tendency to form reaction rims imply that it may be xenocrystic. However, in our experiments quartz saturates at ∼780°C, a temperature slightly higher than the average zircon saturation temperatures of 736 ± 25°C calculated for Uturuncu lavas and domes (Muir et al., in press). If quartz is indeed an equilibrium phase in the natural samples it could form as the magmas cool below ∼780°C; resorption of quartz would imply subsequent heating above this temperature.

Evidence from the natural samples suggests that recharge by hotter magmas is an important process in the genesis and eruption of the Uturuncu magmas (Muir et al., in press). Recharge magmas could have been both hotter, less crystalline dacites and more mafic andesites, for which quenched mafic inclusions provide evidence. Resorption of quartz and formation of reaction rims could therefore occur locally after magmatic recharge events when temperatures are raised above the quartz stability field. A similar argument can be advanced for the observed enrichment of P₂O₅ in groundmass glass from the resorption of apatite in natural samples (Muir et al., in press) as magma is locally heated above ∼870°C. Thus our experimental phase relations, with required phases saturating over a range of temperatures from ∼950°C for orthopyroxene, to ∼870°C for biotite and apatite, 780°C for quartz, and 780°C for zircon, are consistent with a magma storage region experiencing spatio-temporal temperature variations accompanied by incomplete dissolution of lower temperature phases such as quartz, apatite and zircon. Piecemeal construction of a magma body can be envisaged with ingress of new batches of slightly hotter magma, coupled with secular cooling in intervening periods e.g. Annen et al. (2008).

Owing to decompression-driven groundmass crystallization during magma ascent, matrix melt compositions of the natural lava samples are not expected to exactly match experimentally derived residual melts from the dacite experiments at magma storage conditions in the same way that is observed for explosive, rapidly erupted tephra (e.g. Geschwind & Rutherford, 1995). Nevertheless, comparisons of residual melts from experiments using dacite starting material with UTDM41B groundmass glasses provide a further constraint on likely magma storage conditions (Fig. 10). Residual melt in the H₂O-saturated dacite experiment at 870°C, 50 MPa most closely resembles that of UTDM41B groundmass glass. In particular, K₂O, an incompatible component and therefore a proxy for crystallinity, is similar in this experiment whereas it is ∼1 wt % lower in all other runs. At 870°C, 100 MPa and 780°C, 150 MPa natural melt compositions are reproduced for most elements, but K₂O is lower by ∼1·5 wt %, consistent with a slightly lower crystallinity prior to microlite and phenocryst rim crystallization during ascent.

The vesicle-free crystallinity of UTDM41B is 37 vol. % but, accounting for xenocrystic and antecrystic components, the equilibrium crystallinity is likely to be 26–34 vol. %. Of the H₂O-saturated dacite experiments, a crystallinity of 35 vol. % at 780°C, 150 MPa is closest to the total crystallinity of UTDM41B; however, orthopyroxene is not stable at such low temperatures (Fig. 6). At 870°C, 50 MPa, a crystallinity of 28 vol. % is within the estimated range of equilibrium crystallinity for UTDM41B (Fig. 7).

Mixed volatile-saturated (P_{H2O + CO2} = P_TOTAL)

Melt inclusion data from Muir et al. (in press) indicate that Uturuncu melts typically contain 3·2 ± 0·7 wt % H₂O and <160 ppm CO₂. Using VolatileCalc (Newman & Lowenstern, 2002), at 870°C minimum magma storage pressures of 50–119 MPa are calculated with equilibrium coexisting fluids having molar xH₂O = 0·87–1·00. The similarity between magma storage pressures estimated from the volatile contents of melt inclusions and our experiments is remarkable. Our experiments show that between xH₂O = 0·7 and xH₂O = 1·0 the maximum pressure at which biotite and orthopyroxene are stable at 870°C remains constant at 110 ± 10 MPa (Fig. 9). We propose this as the optimum, pre-eruptive storage pressure for the Uturuncu dacites prior to ascent and subsequent decompression-driven crystallization of microlites and phenocryst rims, which served to increase crystallinity slightly, generate more evolved melt and lead to biotite phenocryst breakdown.

Implications of shallow storage of dacites for current deformation

H₂O-saturated and mixed-volatile-saturated experimental phase relations and glass chemistries demonstrate that the Uturuncu dacites are stored shallower than 100 ± 50 MPa at temperatures around 870°C prior to eruption. Assuming an Andean crustal rock density of 2700 kg m⁻³ (Lucassen et al., 2001) this corresponds to a maximum depth of 5·7 km below the surface—similar to hypocentral locations of recent swarms of volcano-tectonic earthquakes at Uturuncu, which were interpreted as being caused by a metastable hydrothermal system (Jay et al., 2012). Melt inclusion trapping pressures (150 ± 50 MPa; Schmitt, 2001) and extant geobarometers (Lindsay et al., 2001) show similarly shallow storage of silicic melts for various ignimbrites in the APVC. Comparably shallow storage of magmas is also a feature of various other arc settings; for example, Novarupta (Coombs & Gardner, 2001; Hammer et al., 2002), Aniakchuk (Larsen, 2006), Santorini (Cottrell et al., 1999) and Montserrat (Barclay et al., 1998).

Pre-eruptive magma storage at 50–150 MPa (depths of 1·9–5·7 km) equates to a maximum of ∼3 km below the base of the edifice. To test these depths as potential causes of the recorded ground inflation we ran a series of mechanical finite-element models (Fig. 11). These numerically solve the subsurface stress and strain that result from the pressurization of a source and then provide displacements at the Earth’s surface. As we are interested only in whether sources at these depths are able to reproduce the observed InSAR spatial deformation anomaly we can represent the crust as an elastic body, thus avoiding more complicated viscoelastic rheologies, which would provide additional temporal information. We use the model setup and approach described by Hickey et al. (2013), which places a uniformly pressurized source cavity within a layered elastic model domain. The crustal heterogeneity is constrained using receiver functions from a nearby seismic study (Wigger et al., 1994; Leidig & Zandt, 2003) and converting seismic velocities to elastic moduli.

As a benchmark, the model results are compared with the ∼7·5 cm maximum displacement recorded by InSAR in the satellite line-of-sight (LOS), between 1996 and 2000 (Pritchard & Simons, 2004). Thus the model output is converted into the LOS for assessment. The total ground uplift anomaly for this period stretches across a 70 km diameter area in a broadly axially symmetric fashion. To fit this uplift footprint, a source that is centred at 1·9, 3·8, or 5·7 km requires a lateral half-width of at least 35 km, therefore being equal in extent to the entire deformation anomaly (Fig. 11). Such a feature has not been observed in recent seismic (Jay et al., 2012) or gravity surveys (del Potro et al., 2013) of the area surrounding Uturuncu. Instead, Hickey et al. (2013) proposed a deeper, vertically elongate source that is mechanically analogous to a diapiric shape protruding upwards from the APMB to fit the observed deformation data (Fig. 11). A similar mechanism was also proposed by Fialko & Pearse (2012), albeit using a different set of magmatic parameters.

Proposed magma genesis and storage model

H₂O-saturated experimental phase equilibria are in agreement with melt inclusion trapping pressures (Muir et al., in press) and indicate that Uturuncu dacites are stored at pressures of 100 ± 50 MPa prior to eruption. Isothermal mixed H₂O–CO₂ experiments at 870°C further constrain storage pressures to 110 ± 10 MPa with xH₂O = 0·7–1·0. The wide range of plagioclase and orthopyroxene compositions in UTDM41B, no doubt partly enhanced by the presence of xenocrysts and antecrysts, and the observed resorption of some crystal phases (quartz, zircon, apatite) imply that crystallization occurred over a somewhat wider range of temperatures, and possibly pressures, in the sub-volcanic storage regions.

Existing geophysical data indicate the existence of the APMB around 17 km depth beneath Uturuncu (Schilling et al., 1997; Schmitz et al., 1997; Chmielowski et al., 1999; Brasse et al., 2002). The hypothesis that multi-level magma systems occur within the crust has existed since the study by Hildreth (1981). In such models, shallower more silicic magmas are thought to evolve from less evolved magmas deeper in the system. At Uturuncu it is likely that these less evolved magmas are stored in the APMB. The composition of these magmas may approximate the andesites found as enclaves in the lavas (e.g. Sparks et al., 2008; de Silva, 1994) and ignimbrites (e.g. de Silva et al., 2006) in the APVC. More evolved, dacitic and rhyodacitic magmas may form by crustal contamination and/or fractional crystallization of these andesites (e.g. Michelfelder et al., 2013) within the crust. If this is the case at Uturuncu, two end-member scenarios for dacite genesis involving andesite magmas can be envisioned, as follows.

1. Hydrous andesite magmas are intruded into the APMB (Fig. 12). Fractional crystallization produces a residual melt of dacitic composition. These dacitic melts rise buoyantly to shallow levels, mixing with crustal melts, crystallizing and fractionating along the way to produce a dacitic bulk composition containing rhyolitic melt that is subsequently erupted (Fig. 12). In this scenario, the least evolved dacites can be related to the most evolved andesites along liquid lines of descent. Differentiation from andesite to dacite may occur within the APMB or during buoyant ascent of magma batches from the APMB as envisaged by del Potro et al. (2013).

2. Alternatively, crustal melting may play a more crucial role in the genesis of the dacitic magmas as evidenced by isotopic data in Michelfelder et al. (2013). The APMB can be envisaged as a mid-crustal ‘hot zone’ (Annen et al., 2006) fuelled by intrusion of andesite magma, which provides an efficient heat source for anatexis of local crustal rocks. Subsequent mixing of these crustal melts with andesitic magmas within the APMB could form melts of dacitic composition that carry a cargo of xenocrystic and antecrystic crystals. These hybrid magmas may rise buoyantly to shallow crustal levels, undergoing decompression crystallization along the way. Significant crustal contamination of the andesitic magmas would mean that the less evolved dacites are unlikely to fall along the liquid lines of descent of the most evolved andesites sampled at Uturuncu.

Distinguishing between these two possibilities remains to be tested experimentally and via isotope geochemistry. In either case, the sub-volcanic magma reservoir beneath Uturuncu would have been constructed primarily by increments of dacitic magma sourced from greater depths. Piecemeal construction of the reservoir is consistent with the experimental data presented here and the petrological data presented by Muir et al. (in press).

The presence of quenched andesite enclaves in the dacite lavas suggests intermittent, direct transfer of andesite magmas to shallow storage levels, in effect bypassing the dacite generation region at depth (Fig. 12). Injection of hotter, less viscous andesite magmas into cooler, more silicic magmas could play an important role in eruption triggering. Rapid thermal cooling and associated quench crystallization of the injected andesitic magma may induce vesiculation. Heating of the more evolved host magmas would reduce volatile solubility, potentially causing renewed or increased bubble generation, although increasing melt fraction during heating would dampen this effect.

CONCLUSIONS

H₂O-saturated phase equilibria experiments using dacite whole-rock and rhyolite groundmass glass starting compositions demonstrate how dacite magma storage conditions beneath Uturuncu volcano prior to previous eruptions were characterized by pressures of 100 ± 10 MPa (equivalent to 5·7 ± 0·6 km below surface) and temperatures of ∼870°C. Natural Uturuncu dacites all have plagioclase, orthopyroxene and biotite as major phases. This assemblage is reproduced in the dacite system only at 870°C, 100 MPa under H₂O-saturated conditions. Melt inclusion trapping pressures in plagioclase phenocrysts from dacite lavas support such shallow storage (Muir et al., in press) and indicate that volatile compositions are between xH₂O = 0·87 and 1·00. Further crystallization of groundmass microlites and phenocryst rims, and partial breakdown of biotite, occurred as the magmas ascended from the storage region.

If the surface deformation at Uturuncu is caused by magma intrusion in a fashion similar to all previous eruptions from this volcano it is unlikely that the evolved magmas in the pre-eruptive storage regions are the cause of inflation. Intrusion of less evolved, possibly andesitic magma at mid-crustal depths (11–17 km below surface; Sparks et al., 2008; Hickey et al., 2013) is required to reproduce the large (70 km diameter) deformation footprint. The APMB is a viable storage region for these andesites from which dacitic liquids may be generated by some combination of crystallization and crustal assimilation.

ACKNOWLEDGEMENTS

Plots were prepared using GMT software (Wessel & Smith, 1991). Mayel Sunagua, SERGEOTECMIN, and SERNAP are acknowledged for their assistance in Bolivia. Thanks go to the PLUTONS working group for critical discussions over the past few years. Comments from Shan de Silva, Jan Lindsay and an anonymous reviewer are gratefully acknowledged.

FUNDING

This work was supported by the Natural and Environmental Research Council (NE/G01843X/1), European Research Council Advanced Grant ‘CRITMAG’ (J.D.B.), a Wolfson Research Merit Award (J.D.B.), European Union Framework Program 7 (J.H.) (grant 282759, ‘VUELCO’). and a Royal Society University Research Fellowship (A.C.R.).

SUPPLEMENTARY DATA

Supplementary Data for this paper are available at Journal of Petrology online.

REFERENCES