Journal of Petrology

Journal of Petrology Advance Access originally published online on October 2, 2006
Journal of Petrology 2006 47(12):2303-2334; doi:10.1093/petrology/egl045

This Article Abstract FREE Full Text (PDF) All Versions of this Article: 47/12/2303    most recent egl045v1 Alert me when this article is cited Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Similar articles in ISI Web of Science Alert me to new issues of the journal Add to My Personal Archive Download to citation manager Search for citing articles in: ISI Web of Science (9) Request Permissions Google Scholar Articles by HUMPHREYS, M. C. S. Articles by SPARKS, R. S. J. Search for Related Content GeoRef GeoRef Citation Social Bookmarking          What's this?

© The Author 2006. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: journals.permissions@oxfordjournals.org

Magma Evolution and Open-System Processes at Shiveluch Volcano: Insights from Phenocryst Zoning

MADELEINE C. S. HUMPHREYS^*, JON D. BLUNDY and R. STEPHEN J. SPARKS

DEPARTMENT OF EARTH SCIENCES, UNIVERSITY OF BRISTOL, WILLS MEMORIAL BUILDING QUEEN'S ROAD, BRISTOL BS8 1RJ, UK

RECEIVED DECEMBER 6, 2005; ACCEPTED AUGUST 17, 2006

	ABSTRACT

TOP ABSTRACT INTRODUCTION GEOLOGICAL BACKGROUND ANALYTICAL METHODS ESTIMATION OF INTENSIVE... PHENOCRYST ZONING PATTERNS DISCUSSION CONCLUSIONS REFERENCES

 
Phenocryst zoning patterns are used to identify open-system magmatic processes in the products of the 2001 eruption of Shiveluch Volcano, Kamchatka. The lavas and pumices studied are hornblende–plagioclase andesites with average pre-eruptive temperatures of 840°C and fO₂ of 1·5–2·1 log units above nickel–nickel oxide (NNO). Plagioclase zoning includes oscillatory and patchy zonation and sieve textures. Hornblendes are commonly unzoned, but some show simple, multiple or patchy zoning. Apatite microphenocrysts display normal and reverse zoning of sulphur. The textural similarity of patchy hornblende and plagioclase, together with Ba–Sr systematics in patchy plagioclase, indicate that the cores of these crystals were derived from cumulate material. Plagioclase–liquid equilibria suggest that the patchy texture develops by resorption during H₂O-undersaturated decompression. When H₂O-saturated crystallization recommences at lower pressure, reduced pH₂O results in lower X_An in plagioclase, causing more Al-rich hornblende to crystallize. Plagioclase cores with diffuse oscillatory zoning, and unzoned hornblende crystals, probably represent a population of crystals resident in the magma chamber for long periods of time. In contrast, oscillatory zoning in the rims of plagioclase phenocrysts may reflect eruption dynamics during decompression crystallization. Increasing Fe/Al in oscillatory zoned rims suggests oxidation as a result of degassing of H₂O during decompression. A general lack of textural overlap between phenocryst types suggests that different phenocryst populations were spatially or temporally isolated during crystallization. We present evidence that the host andesite has mixed with both more felsic and more mafic magmas. Olivine and orthopyroxene xenocrysts with reaction or overgrowth rims and strong normal zoning indicate mixing with basalt. Sieve-textured plagioclase resulted from mixing of a more felsic magma with the host andesite. The mineralogy and mineral compositions of a mafic andesite enclave are identical to those of the host magma, which implies efficient thermal quenching, and thus small volumes of intruding magma. Mixing of this magma with the host andesite results in phenocryst zoning because of differences in dissolved volatile contents. We suggest that small magma pulses differentiated at depth and ascended intermittently into the growing magma chamber, producing incremental variations in whole-rock compositions.

KEY WORDS: patchy zoning; magma mixing; Shiveluch

	INTRODUCTION

TOP ABSTRACT INTRODUCTION GEOLOGICAL BACKGROUND ANALYTICAL METHODS ESTIMATION OF INTENSIVE... PHENOCRYST ZONING PATTERNS DISCUSSION CONCLUSIONS REFERENCES

 
Open-system magmatic processes are ubiquitous phenomena at island arc volcanoes. Processes may include segregation and ascent of magmas from depth, accumulation and mixing of magma batches in a shallow chamber, and eruptive processes. The intricacy of the magmatic history is reflected to varying extents in the typically complex petrography of erupted andesites. Eruptive products thus contain clues to the entire magma history, from differentiation in the deep crust to eruption at the surface.

Magma mixing phenomena are particularly widespread (e.g. Anderson, 1976; Eichelberger, 1980; Clynne, 1999). Mixing is envisaged to occur in various scenarios: for example, thermal or chemical convection in a magma chamber (e.g. Huppert et al., 1982; Couch et al., 2001), within the conduit during eruption (Snyder & Tait, 1996; Koyaguchi, 1985) or during sidewall crystallization in mafic magma chambers (McBirney, 1980). Mixing can also occur as a result of recharge of the magma chamber by the arrival of a new magma batch, which may be a trigger for eruptions (e.g. Sparks et al., 1977; Huppert et al., 1982; Murphy et al., 1998). Magma recharge commonly occurs as cool, silicic melt is invaded by hotter, more mafic magma (e.g. Anderson, 1976; Clynne, 1999; Murphy et al., 2000). This process is typically inferred from features such as the presence of mafic enclaves, accompanied by heavily corroded xenocrysts and significant geochemical zonation in phenocrysts (Eichelberger, 1978, 1980). The reverse process (i.e. injection of more silicic magma into basaltic systems) has also been inferred on the basis of field observations (Wiebe, 1987) and geochemistry (Calanchi et al., 1993). Less attention has been paid to the possibility of ‘cryptic’ magma mixing, which we define here as the repeated mixing of small pulses of magma that have similar anhydrous bulk compositions, but may have different fO₂ or volatile contents. Magma recharge would be difficult to detect if the mixing magmas were compositionally similar, because of limited disequilibrium between crystals and melt (D'Lemos, 1996).

Recently, field evidence has shown that plutons can grow incrementally via the coalescence of many much smaller magma bodies (Bateman, 1992; John & Blundy, 1993; Wiebe & Collins, 1998; Atherton, 1999; Glazner et al., 2004). This implies that some subvolcanic magma chambers are built up gradually and over prolonged periods of time from many small batches of magma. The magma batches may have differentiated at depth prior to ascent (Annen et al., 2006) and been injected into a shallow-level magma chamber. The efficiency of magma mixing resulting from such recharge depends on the volume ratio of intruding magma to resident magma, and on the viscosity contrast between the two magmas (e.g. Sparks & Marshall, 1986; Jellinek et al., 1999). This in turn depends on the relative temperatures, dissolved volatile contents and compositions of the magmas (Sparks & Marshall, 1986). If the intruding magma forms enclaves, enclave mineralogy will depend on the rate of quenching as well as the bulk composition of the intruding magma (Blundy & Sparks, 1992). Further fragmentation or disaggregation of enclaves can help to hybridize the mixed magma (Clynne, 1999, and references therein). The resulting mineral textures will depend on how close the enclave mineralogy is to equilibrium with the host magma, and thus on the composition and relative volume of the intruding magma.

As well as recharge and mixing, other processes that contribute to the textural complexity of andesites include the resorption and remobilization of earlier-formed crystals or cumulate materials, as magma ascends from its deep source to a shallow magma storage area (Pyle et al., 1988; Turner et al., 2003; Dungan & Davidson, 2004). In the storage area, the magma may slowly cool and degas, promoting crystal fractionation. Convection in a zoned magma can cause compositional zoning in existing phenocrysts (Singer et al., 1995). Late-stage crystal growth may occur during magma ascent in the conduit, in response to depressurization and degassing of H₂O (e.g. Geschwind & Rutherford, 1995).

In this paper we investigate phenocryst zoning in andesitic dome rocks and pumices from Shiveluch Volcano, Kamchatka, with the aim of identifying different magmatic processes through their effects on the phenocrysts. We identify different populations of phenocrysts based on their textural and geochemical characteristics, and attempt to produce an integrated picture of magmatic processes in their spatial and temporal context. We show that the host andesite has mixed with basalt, producing xenocrysts showing significant disequilibrium textures. The host andesite has also mixed with a mafic andesite, although efficient quenching resulted in cryptic mixing, with identical mineralogy and mineral compositions in both the enclave and the host andesite. Geochemical zonation is produced in crystals from both magmas as a consequence of variations in volatile contents between the magmas. We also discuss the remobilization of cumulates, and the effects of decompression and eruption dynamics on the evolution of the magma at both deep and shallow levels, thus revealing a complex record of open-system processes at this active andesite volcano.

	GEOLOGICAL BACKGROUND

TOP ABSTRACT INTRODUCTION GEOLOGICAL BACKGROUND ANALYTICAL METHODS ESTIMATION OF INTENSIVE... PHENOCRYST ZONING PATTERNS DISCUSSION CONCLUSIONS REFERENCES

 
Shiveluch Volcano is located in the northern Central Depression of the Kamchatka Peninsula (Fig. 1), and has been one of the most active andesitic volcanoes in the Kamchatka–Kurile arc during the Holocene, with at least 60 tephra fall deposits identified, marking explosive eruptions with volumes in excess of 0·1 km³ (Dirksen et al., 2006). It has a compound edifice, comprising sections from two distinct episodes of activity. Old Shiveluch is the remains of a basalt to basaltic andesite stratovolcano (Menyailov, 1955; Melekestsev et al., 1991). The southern section of this edifice was destroyed by a huge landslide during the late Pleistocene (Melekestsev et al., 1991), producing a 9 km wide caldera that exposes NE-trending, basaltic andesite dykes within the cone (Belousov et al., 1999). The younger edifice, Molody Shiveluch, has been growing throughout the Holocene. Eruptive activity at Molody Shiveluch is typified by Plinian eruptions alternating with dome-building and phreatic activity (Belousov et al., 1999). Juvenile material is predominantly amphibole–plagioclase andesite (Melekestsev et al., 1991).

View larger version (67K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 1 Regional map of the Kamchatka Peninsula, showing subduction zones, regions of magmatism and some major active volcanoes. Magmatism is split into the Eastern Volcanic Front, Sredinny Ridge (currently inactive) and the Central Kamchatka Depression. The region north of the Aleutian trench is split by fracture zones and magnetic lineations, indicating an extinct trench trending to the NE (Baranov et al., 1991; Portnyagin et al., 2005). The convergence of the Pacific plate with the Aleutian arc is oblique; the angle of convergence increases eastward (Geist & Scholl, 1994). Shaded topographic base map is taken from NASA Earth Observatory.

 
The most recent phase of volcanic activity started in 1964 with a strong Plinian eruption triggered by edifice failure (Belousov, 1995). This was followed by episodes of dome extrusion in 1980–1981, 1993–1995 and 2001–2004. These periods of active extrusion were separated by less intense activity (e.g. Firstov et al., 1994; Zharinov et al., 1995; Dirksen et al., 2006). In this study we focus on rocks erupted during the most recent extrusive episode, 2001–2004.

Summary of andesite petrology
We studied samples of dome rock, and pumiceous material taken from pyroclastic flows, from early in the 2001–2004 eruption of Shiveluch Volcano (Table 1). The petrology of the andesite has been described by Dirksen et al. (2006), including whole-rock and modal data, and is summarized and extended here. The main phenocrysts (>300 µm) and microphenocrysts (100–300 µm) are plagioclase (56–70% of phenocrysts), hornblende (27–35%), Fe–Ti oxides (1·7–4·8%) and orthopyroxene (0·3–5·2%). Clinopyroxene, apatite and anhydrite are present in minor amounts. Quartz and other silica phases are absent. Xenocrysts of olivine and orthopyroxene are present. The groundmass comprises a rhyolitic glass with microlites of plagioclase, clinopyroxene, orthopyroxene and Fe–Ti oxides. Total phenocryst contents vary from 26% (pumices) to 52% (dome rocks), calculated vesicle-free.

View this table: [in this window] [in a new window]   Table 1 Names and descriptions of samples studied

 
Plagioclase compositions range from An₂₉ to An₇₄; microlite compositions range from An₄₄ to An₅₁. Plagioclase phenocrysts commonly contain inclusions of Fe–Ti oxides, hornblende, apatite and glass. Orthopyroxene and clinopyroxene inclusions occur less commonly; phlogopite, bornite and anhydrite inclusions occur rarely. Hornblende compositions range from Mg-number [100 x Mg/(Mg + Fe_tot)] of 58 to 81, classifying as magnesio-hornblendes (Leake, 1978). Hornblende phenocrysts commonly contain inclusions of plagioclase, Fe–Ti oxides, apatite and glass, with rare pyroxene, anhydrite, phlogopite and bornite inclusions. Small, rounded hornblende is present in the groundmass of some samples. Orthopyroxene phenocrysts and microphenocrysts vary in composition from En₆₆ to En₇₃, and may contain inclusions of Fe–Ti oxides and apatite. Groundmass orthopyroxene compositions are typically En_70–74. Orthopyroxene in reaction rims on olivine xenocrysts, in crystal clots or included in olivine is more Mg-rich (En_74–87). Orthopyroxene xenocrysts, identified by strong normal zoning and a hornblende rim, have high Mg (En₈₉), Cr and Fe³⁺/Fe²⁺ [calculated using Lindsley (1983)] in the core, decreasing towards the rim (En₇₅). Clinopyroxene is occasionally included in the cores of phenocrysts, occurs rarely as large, anhedral crystals, and commonly in the groundmass. Clinopyroxene compositions range from Wo₄₀ to Wo₄₄. Fe–Ti oxides occur as microphenocrysts and microlites, as inclusions in phenocrysts and as symplectites in the breakdown of olivine. Fe–Ti oxides are dominated by spinels (2·7–8·5 wt % TiO₂), with minor ilmenite (28 wt % TiO₂).

Accessory minerals in the andesites studied include apatite and rare anhydrite as primary microphenocrysts, and anhedral Cu–Fe sulphides, cassiterite (SnO₂) and CuO in the groundmass and as inclusions in plagioclase and hornblende. The copper-bearing sulphide is a solid solution of bornite and digenite, ranging from Cu₈FeS₃ to Cu₄₇FeS₁₄. Apatite contains up to 5860 ppm sulphur, and is ubiquitous in the rocks studied.

Melt inclusions are common in both hornblende and plagioclase (Humphreys, 2006). Bubbles are common; partial devitrification or precipitation of daughter phases is common in hornblende-hosted inclusions but rare in plagioclase-hosted inclusions. Analysis of melt inclusions containing daughter minerals was avoided where possible. Because there is little compositional difference between inclusions in different host phenocrysts, or between inclusions with and without bubbles (Humphreys, 2006), we assume that post-entrapment crystallization is limited, and that the inclusions accurately reflect the composition of the melt at the time of trapping. Detailed melt inclusion data have been given by Humphreys (2006) and the results are summarized here. Melt inclusions are consistently rhyolitic in composition, containing 70–80 wt % SiO₂ (normalized to anhydrous) and 0·1–5·1 wt % H₂O as analysed by secondary ionization mass spectrometry (SIMS). Matrix glasses are among the most silica-rich, typically containing 77·5–79·5 wt % SiO₂, and low H₂O (0·10–0·14 wt %). H₂O correlates negatively with SiO₂ content of the inclusions, and the inclusions follow a decompression trend on the Qz–Or–Ab ternary.

The andesites contain dark green, medium- to coarse-grained xenoliths. These are typically several centimetres in diameter with a well-defined, smooth, usually hornblende-rich rim. The modal mineralogy of the xenoliths is highly variable; they contain hornblende, plagioclase, phlogopite and orthopyroxene, with minor clinopyroxene, apatite, Fe–Ti oxides and rounded chromite. No olivine is observed. Small pockets of interstitial glass are present. The xenolith texture varies from poikilitic, with plagioclase oikocrysts in optical continuity, to granoblastic, indicating variable amounts of deformation and recrystallization. Small veins containing calcic plagioclase (e.g. An₉₃) and pyroxenes cut across some xenoliths. The xenoliths are interpreted as cumulates.

One of the samples (shiv01/#3) also contains a small (2–3 cm), elliptical, quench-textured andesitic enclave (Fig. 2a). There is no change in grain size towards the rim of the enclave, which is defined by the edges of crystals protruding into the host lava. The enclave has a high proportion of crystals (85% calculated vesicle free) including randomly oriented, elongate, skeletal hornblende phenocrysts (Fig. 2a and b), phenocrysts of plagioclase, tabular Fe–Ti oxides, acicular apatite and, unusually, abundant anhydrite crystals up to 300 µm in length. Apatite is both more acicular, more abundant, and more S-rich than in the host andesite. Dendritic protrusions extend from the rims of some crystals (Fig. 2c, d). Mineral compositions are indistinguishable from compositions of phenocrysts in the host andesite (see Tables 3 and 4). The matrix comprises a vesicle-rich, microlite-free, rhyolitic glass, with similar compositions to melt inclusions and matrix glasses in the host andesite. The enclave has slightly higher modal amphibole (Dirksen et al., 2006, table 2) in comparison with the phenocryst assemblage in the host andesite, indicating a slightly more mafic bulk composition. Vesicles commonly occur between grain junctions, suggesting that vesiculation was induced by crystallization.

View larger version (95K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 2 Petrographic features of the quenched enclave. Photomicrograph (a) shows elongate, skeletal hornblende (hbl) phenocrysts and plagioclase (pl) in vesicle-rich glass (scale bar represents 1 mm). Back-scattered SEM photograph (b) shows abundant, acicular apatite (ap), anhydrite (anh) and elongate hornblende (scale bar represents 500 µm). Tabular white minerals are titanomagnetite (mt). It should be noted that plagioclase and glass (gl) are indistinguishable in this image. Dendritic extensions (indicated by arrows) on plagioclase and hornblende crystals (c, d) indicate rapid growth. Scale bar in (c) represents 50 µm (c), and in (d) 100 µm.

 

View this table: [in this window] [in a new window]   Table 2 Temperatures estimated for the Shiveluch magma using different methods

 
Whole-rock geochemistry
All samples studied here are pumices and dome rocks of medium-K andesite (Gill, 1981), and plot in the calc-alkaline field of Miyashiro (1974). The rocks contain 61–63 wt % SiO₂. Figure 3 shows selected major and trace element variation diagrams for whole-rock data from these samples (see also Dirksen et al., 2006, table 1). Negative correlations with SiO₂ are seen in MgO, CaO and FeO. Al₂O₃ and K₂O do not vary much; Na₂O correlates positively with SiO₂ (Fig. 3). In the trace elements, weak positive trends with SiO₂ are seen for Ba, Rb, Sr and Zr; Y shows little variation; V defines negative trends. Typical compositions of hornblende and plagioclase phenocrysts from electron probe microanalysis (EPMA) and secondary ion mass spectrometry (SIMS) data are also plotted in Fig. 3. Most major element trends are broadly consistent with fractionation of an assemblage dominated by plagioclase and hornblende. However, the trace element trends suggest the involvement of an additional phase, in which Ba, Rb and V, as well as K₂O, are compatible (Fig. 3). Phlogopite, along with titanomagnetite, is a suitable candidate (Fig. 3). Phlogopite is found with hornblende and plagioclase in cumulate xenoliths, and occasionally as inclusions in phenocrysts.

View larger version (25K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 3 Anhydrous, whole-rock major element (a–c) and trace element (d–f) variation diagrams for the samples studied. Typical hornblende (hbl) and plagioclase (plag) compositions are plotted from EPMA or SIMS analyses, with error bars indicating the range of LA-ICP-MS data where available. Phlogopite compositions are taken from Tischendorf et al. (2001). Trend lines with R2 quoted represent least-squares regression lines; other trend lines are fitted by eye. Grey shaded areas indicate approximate whole-rock compositions of cumulate nodules, reconstructed from modal data (Humphreys, 2006). Star indicates the approximate composition of the quenched enclave, also reconstructed from modal data.

 

	ANALYTICAL METHODS

TOP ABSTRACT INTRODUCTION GEOLOGICAL BACKGROUND ANALYTICAL METHODS ESTIMATION OF INTENSIVE... PHENOCRYST ZONING PATTERNS DISCUSSION CONCLUSIONS REFERENCES

 
Wavelength-dispersive EPMA of minerals and glasses was performed at the University of Bristol using a CAMECA SX-100 five-spectrometer instrument. Minerals were analysed using a 20 kV accelerating voltage, 10 nA beam current and a 5 µm beam diameter. Groundmass glasses and melt inclusions were analysed using a 15 kV accelerating voltage, 2–4 nA beam current and a 15 µm beam diameter, with Na and Si analysed first to reduce the effects of alkali migration, as described in Humphreys et al. (2006). Data reduction was performed online using a stoichiometric PAP correction model (Pouchou & Pichoir, 1984). H₂O contents of hydrous melt inclusions, and adjacent host plagioclase compositions, were estimated as described by Blundy et al. (2006), and were presented in that study.

X-ray fluorescence (XRF) analyses were carried out using a Philips PW1400 X-ray fluoresence spectrometer at Leicester University. Analyses of major elements were performed using fused glass discs made from powders mixed with 80:20 Li metaborate:Li tetraborate. Trace elements were determined on 32 mm diameter, pressed powder briquettes.

Trace element concentrations in some hornblende and plagioclase phenocrysts were determined by inductively coupled plasma mass spectrometry (ICP-MS) using a VG Elemental PlasmaQuad 3 system coupled to a VG LaserProbe II (266 nm frequency-quadrupled Nd–YAG laser) at the University of Bristol. He gas was the ablation medium and a He–Ar mixture carried the sample to the plasma. The laser beam was set to 20 µm diameter, 5 Hz repetition rate and about 25–30 J/cm² energy density. For each analysis, counts were collected on the background for 50 s, before firing the laser and collecting counts for 50 s. Data were collected for ⁴⁵Sc, ⁵¹V, ⁸⁸Sr, ⁹⁰Zr and ¹³⁰Ba. The glass NIST 610 (Pearce et al., 1997) was used as primary standard, with NIST 612 as a secondary standard. Data reduction, using the method of Longerich et al. (1996) and Si as the internal standard, was performed using PlasmaLab software.

	ESTIMATION OF INTENSIVE PARAMETERS

TOP ABSTRACT INTRODUCTION GEOLOGICAL BACKGROUND ANALYTICAL METHODS ESTIMATION OF INTENSIVE... PHENOCRYST ZONING PATTERNS DISCUSSION CONCLUSIONS REFERENCES

 
Temperature
Magma temperatures for the Shiveluch andesite were estimated by Dirksen et al. (2006) using methods including hornblende–plagioclase thermometry and Fe–Ti oxide thermometry. Here, we summarize these temperature estimates and present further information derived from plagioclase–liquid equilibria (Putirka, 2005), which constrains T from information about X_An in plagioclase, liquid composition (from melt inclusions), H₂O content, and pressure, assuming equilibrium between plagioclase and the coexisting liquid. Pressure was calculated from H₂O content using the model of Newman & Lowenstern (2002). This approach is valid because there is good evidence that the magma is H₂O-saturated (see later). Results are given in Table 2.

Hbl–plag temperatures range from 775°C to 906°C ± 35°C, with an average of 841°C ± 35°C (n = 85). A temperature of 808–836°C ± 35°C was obtained for the quenched enclave (n = 3). In comparison, Fe–Ti oxide temperatures range from 834°C to 978°C (Table 2), with an average of 878°C (n = 32). Microlites in contact with the matrix glass return slightly higher temperatures (860–978°C ) than inclusions in phenocrysts (834–904°C). An oxide–oxide temperature of 870°C was estimated for the andesite enclave (n = 2). Plagioclase–liquid temperatures range from 819°C to 975°C, with an average of 867°C (n = 63). These data cover a wide range of temperatures; oxide and plagioclase–liquid temperatures are consistently higher than hbl–plag temperatures. These differences are interpreted to reflect temperature variations throughout the magma's evolution. Hbl–plag data reflect temperatures early in the magmatic history, while hornblende is still crystallizing. Oxide data, which re-equilibrate rapidly (Hammond & Taylor, 1982; Venezky & Rutherford, 1999), are interpreted to reflect higher magma temperatures caused by release of latent heat during late-stage decompression-driven crystallization (Blundy et al., 2006). Plagioclase–liquid temperatures reflect real variation of T and pH₂O during the magma's history. From these data, we estimate that magma temperatures prior to ascent were 840°C.

Pressure
The maximum magma pressure, estimated from the maximum H₂O contents of melt inclusions, is 160 MPa (Humphreys, 2006). The depth of the magma chamber roof at Shiveluch Volcano is thus assumed to be 5–6 km below the summit, a result consistent with seismic data (Dirksen et al., 2006).

Oxygen fugacity
Oxygen fugacity is estimated from Fe–Ti oxide equilibria using the method of Andersen et al. (1993). Values of log(fO₂) obtained are highly oxidizing, and range from –9·0 to –11·2, with an average of –10·3 (Fig. 4). The values obtained correspond to 1·5–2·1 log units above the nickel–nickel oxide (NNO) buffer [quartz–fayalite–magnetite (QFM) + 2·1 to + 2·7], with an average of NNO + 1·9. The andesite enclave gave NNO + 1·8. Uncertainties are approximately ±0·6 log units (Andersen et al., 1993). These oxidizing conditions are consistent with the co-precipitation of sulphides (S^2– oxidation state) with anhydrite (S⁶⁺) and S-bearing apatite (S⁶⁺) (Carroll & Rutherford, 1985, 1988).

View larger version (18K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 4 Temperature vs oxygen fugacity, estimated from Fe–Ti oxide equilibria in Shiveluch samples using the QUILF program (Andersen et al., 1993). Oxides were recalculated using Stormer (1983). Contours represent percent of sulphur existing as oxidized sulphate (S6+), calculated from Carroll & Rutherford (1988). QFM, NNO, MNO and MH buffers are plotted for comparison. Shaded areas represent other silicic, sulphur-rich magmas: Pinatubo (P), Lascar (L), El Chichòn (EC) and the Fish Canyon Tuff (FCT). MNO=manganese oxide; MH=magnetite-haematite.

 

	PHENOCRYST ZONING PATTERNS

TOP ABSTRACT INTRODUCTION GEOLOGICAL BACKGROUND ANALYTICAL METHODS ESTIMATION OF INTENSIVE... PHENOCRYST ZONING PATTERNS DISCUSSION CONCLUSIONS REFERENCES

 
Plagioclase
Three plagioclase phenocryst textures are observed: oscillatory zoning, patchy zoning and sieve texture. Table 3 gives representative compositions of the phenocryst types, and Fig. 5 summarizes the chemical variations between the textures.

View this table: [in this window] [in a new window]   Table 3 Representative compositions and structural formulae of patchy, sieve, oscillatory and other plagioclase phenocrysts and microlites

 

View larger version (8K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 5 XAn and Fe/Al compositions of the plagioclase textural types. Circles indicate rim compositions; core compositions are represented by diamonds, or by diamonds (dark patches) and triangles (bright patches) for patchy plagioclase.

 
Oscillatory zoned phenocrysts (68–75% of plagioclase) typically have rather diffuse, symmetrical zones in their cores, in contrast to the sharper, more asymmetrical boundaries of outer zones (Fig. 6a and b). This may reflect partial re-equilibration of the cores during long residence times in the magma chamber. Compositional variation in oscillatory zoning is typically less than 30 mol % Anorthite. Inner zones show an approximately inverse relationship between Fe/Al and X_An. Outer zones commonly show a sharp boundary to higher X_An, followed by gradual decay to more sodic compositions (Fig. 6a and b). At some outer zone boundaries resorption has taken place, as shown by irregular surfaces (Fig. 7a and b) and rounded corners (Fig. 7c). Melt inclusions are occasionally trapped at these resorption boundaries. Rim compositions are usually consistent across each sample. The rims of oscillatory zoned crystals typically show a gradual increase in the ratio Fe/Al over several zones (Fig. 6), unaccompanied by changes in Mg. Groundmass plagioclase also has higher Fe contents and Fe/Al values (Fig. 5b), as observed in other orogenic andesites (e.g. Nakada & Motomura, 1999; Murphy et al., 2000). Fe in plagioclase is dominated by Fe³⁺ at high fO₂ (Sugawara, 2000), and is therefore mainly incorporated by substituting for Al³⁺ (Longhi et al., 1976; Sugawara, 2000). Fe³⁺ increases with X_An because the number of available Al³⁺ sites increases (Hofmeister & Rossman, 1984). Thus the ratio Fe/Al provides a measure of the proportion of Al sites occupied by Fe.

View larger version (34K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 6 Typical zoning profiles of XAn (diamonds) and Fe/Al (circles) from oscillatory zoned phenocrysts (a, b) and rims of patchy phenocrysts (c). All crystals come from the same sample. White arrows mark the line of each traverse. XAn zoning is characterized by sharp increases in An content, followed by slower decay. Zones are typically more diffuse near the core of the crystal than at the rim. Some zoning patterns near the rims (downward arrows) can be tentatively matched across different phenocrysts. Scale bar represents 100 µm. Crystals also show a gradual, rimward increase in Fe/Al (curved arrows). Error bars represent typical 1 S.D. for XAn and Fe/Al, calculated from EPMA counting statistics.

 

View larger version (70K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 7 Petrographic features of patchy and oscillatory zoned plagioclase. Resorption occurs at some boundaries leaving irregular, wavy zoning surfaces (a, b) and rounded corners (c). Some patchy zoned plagioclase have a well-developed oscillatory-zoned rim (d). Scale bars represent 200 µm (a, d) and 50 µm (b, c).

 
Patchy zoned crystals (20–25% of plagioclase) have an irregular or chequerboard, central corroded core with abundant melt and mineral inclusions, suggesting partial resorption of an older core. In back-scattered scanning electron microscope (SEM) images, melt inclusions are closely associated with more albitic (darker) plagioclase of An_29–54 (Fig. 8a). This suggests that the more An-rich patches (An_42–74, Fig. 8a) formed the original core, and that the more albitic plagioclase crystallized after partial resorption, causing entrapment of melt inclusions. The change in X_An between early, anorthite-rich (brighter) plagioclase and later, albite-rich (darker) plagioclase is typically not accompanied by changes in minor elements (e.g. Fe, Mg). Trace element contents do vary, however: An-rich patches typically contain 120–220 ppm Ba whereas Ab-rich patches contain 210–360 ppm Ba. Sr contents are similar (920–1520 ppm Sr in An-rich patches, 1000–1400 ppm in Ab-rich patches). The rims of patchy phenocrysts (An_37–57) are typically clear and zoned; many show oscillatory zoning (Figs 6c and 7d). In the oscillatory zoned rims, the same range of X_An and pattern of increasing Fe/Al is observed as in oscillatory phenocrysts (Fig. 6c). Outermost individual zones can be correlated between the two phenocryst types (Fig. 6), implying that they had a common magmatic history during the late stages of crystallization.

View larger version (69K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 8 (a) Left: backscattered SEM photograph showing typical patchy texture in plagioclase; right: line sketch highlighting the main features. Melt inclusions (dark grey) are closely associated with the darker, albitic plagioclase. This indicates that the Ab-rich surrounding plagioclase formed, and melt inclusions were trapped, after partial resorption of the original, An-rich core. White circles indicate positions of analyses. (b) Left: backscattered SEM photograph; right: sketch showing typical patchy texture in hornblende. Melt inclusions (MI, dark grey) are associated with stringers of more Al-rich (brighter) hornblende (pairs of black arrows). This implies that the brighter hornblende crystallized from melt, concurrent with entrapment of melt inclusions, after partial dissolution of the more Al-poor (darker) core. Scale bars represent 100 µm.

 
Sieve-textured phenocrysts (3–9% of plagioclase) have a sodic core (An_34–37, Fig. 5a), surrounded by a narrow rim of more An-rich plagioclase (Fig. 5b). The core is commonly anhedral, and in some crystals oscillatory zonation is observed in the core (Fig. 9a). The boundary between the core and rim comprises a network of micrometre-scale melt channels. The channelized zone is usually a narrow ring (Fig. 9b), but sometimes occupies the whole plagioclase interior (Fig. 9c); these variations could be an artefact of sectioning, or related to the extent of disequilibrium encountered by the crystal (Clynne, 1999). Where hornblende inclusions are present in both core and rim of sieved phenocrysts (Fig. 9d), temperatures estimated using hbl–plag geothermometry suggest an increase of about 50°C from the core to the rim. Rims of sieved plagioclase crystals have higher FeO (e.g. 0·43 wt % compared with 0·21 wt %), Fe/Al (e.g. 0·010 compared with 0·005), MgO (0·03 wt % compared with 0·00 wt %) and X_An (An_43–63 compared with An_34–37) than the cores (Table 3). Concentrations of Ba and Sr are very similar in both cores and rims of the sieve plagioclase. These cores are sodic compared with typical oscillatory plagioclase compositions, but similar to the Ab-rich patchy plagioclase (Fig. 5a; Table 3).

View larger version (74K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 9 Petrographic features of sieve-textured plagioclase. Some sieve plagioclase crystals have oscillatory zoned cores (a). The sieve texture comprises a network of micrometre-scale melt channels at the boundary between core and rim. This zone is usually a ring (b) but can exist throughout the interior (c). Backscattered SEM enlargement photograph of area indicated in (b) shows variation of plagioclase compositions (open circles) and temperatures calculated from coexisting hornblende (filled circles) (d). Scale bars represent 200 µm (a), 100 µm (b, d) and 50 µm (c).

 
Hornblende
Many hornblende phenocrysts (65–80%) are compositionally rather uniform, unzoned magnesio-hornblendes. However, the compositions of unzoned hornblendes vary in terms of Al^iv, Al^vi, Mg-number and (Na,K)^A (Fig. 10). Three other textural types are observed: patchy zoning (<10% of phenocrysts), simple zoning (4–20%) and multiple zoning (0–3%) (Fig. 11; Table 4). Overall, hornblendes show a very weak Al^iv–Al^vi (Tschermak) correlation (R² = 0·30; n = 420) (Fig. 12a). No correlation is observed within any single zoned phenocryst (Fig. 12b), but multiple analyses from the same grain plot in a cluster around a given Al^iv–Al^vi. Rims of patchy zoned crystals have lower Al^iv–Al^vi than simple-zoned phenocrysts (Figs 10 and 12b). Qualitatively, this relationship indicates rim crystallization at lower pressure, though quantitative estimates cannot be made because of the lack of K-fsp and quartz in the crystallizing assemblage (Johnson & Rutherford, 1989). Overall there is a good correlation (R² = 0·69) between Al^iv and (Na,K)^A (Fig. 12c), indicating that hornblende crystallization is controlled by the edenite exchange. Multiple analyses from single zoned phenocrysts commonly show a good correlation (e.g. R² = 0·9 or better, Fig. 12d) and can span a wide range of Al^iv and (Na,K)^A (Fig. 10a and d).

View this table: [in this window] [in a new window]   Table 4 Representative compositions and structural formulae of patchy, simple and other hornblende phenocryst types

 

View larger version (21K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 10 Alvi, Alvi, Mg-number and (Na,K)A compositions of hornblende textural types. For simple and patchy zoning, circles represent rim compositions, and core compositions are represented by diamonds, or diamonds (dark) and triangles (bright) for patchy zoning.

 

View larger version (55K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 11 Simple (a), patchy (b) and multiple zoning (c) in hornblende phenocrysts. Mg contents are shown for a profile across each crystal, calculated as atoms per formula unit (pfu). Black, double-ended arrows mark the line of each profile. Temperatures from hornblende–plagioclase or Fe–Ti oxide geothermometry are marked at relevant points. Scale bar represents 100 µm. Patchy phenocrysts comprise a core of dark and light patches, overgrown by a euhedral, oscillatory zoned rim. Simple-zoned phenocrysts have a distinct, Mg-rich rim composition, separated from the core by a sharp boundary. Multiple zoned phenocrysts have distinct, diffuse, Mg-poor (bright) zones.

 

View larger version (23K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 12 Chemical variation in hornblendes. (a) Extent of Tschermak relationship (Aliv vs Alvi) for the whole dataset (n = 420) and (b) for multiple analyses on individual phenocrysts. Rims of patchy crystals tend to show lower Aliv and Alvi than simple or multiple zoned phenocrysts, suggesting crystallization at lower pressure. (c) Edenite exchange (Aliv vs (Na,K)A) for the whole dataset and (d) for multiple analyses on individual phenocrysts. The correlation is stronger than in (a), particularly within a given phenocryst, and is stronger for the rims of patchy phenocrysts than for other zoning types. Grey stars mark analyses from quenched enclave.

 
Patchy zoning in hornblende phenocryst cores (Fig. 11a) comprises irregularly shaped dark and light patches in back-scattered SEM images (Fig. 8b). The crystals often contain abundant mineral and melt inclusions. The texture is strongly reminiscent of patchy zoning in plagioclase (see Fig. 8a). Melt inclusions are commonly associated with brighter patches (Fig. 8b), and mineral inclusions are commonly edged with a sliver of melt, suggesting that they grew from a pre-existing melt pool. Bright regions have higher Al^iv and (Na,K)^A, and lower Mg-number than dark regions (Fig. 10; Table 4). Bright and dark regions are indistinguishable in terms of trace elements, except for Ba, which is slightly more abundant in the Al-rich hornblende (Fig. 13a). This probably results from the higher A-site occupancy of the Al-rich hornblende (Table 4), with Ba substituting for K (Tiepolo et al., 2003). Patchy phenocrysts are sometimes overgrown by a thin, euhedral, oscillatory-zoned rim (Fig. 14a–c), with a good correlation of Al^iv with (Na,K)^A (Fig. 14c). Each new zone is marked by a small but abrupt increase in Mg-number, and decrease in Al^iv and (Na,K)^A (Fig. 14a). Throughout the remainder of each zone the Mg-number, and Al^iv and (Na,K)^A contents return to near original levels (Fig. 14a).

View larger version (13K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 13 Ternary diagram for trace elements Zr–Ba–V in (a) patchy hornblende and (b) simple hornblende. (a) Bright, Al-rich hornblende is slightly more Ba-rich than the dark, Al-poor hornblende, which has more V. (b) Simple cores and rims are virtually indistinguishable in terms of trace elements.

 

View larger version (21K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 14 Chemical zoning information from different hornblende phenocryst types. In each panel, the top figure shows a zoning profile with structural elements reported as cations per formula unit (pfu). Curved arrows highlight zoning patterns. The middle figure shows variation of Mg-number with Aliv. The bottom figure shows variation of (Na + K)A with Aliv. Open symbols represent rim compositions; filled symbols represent cores. Typical errors are given where they are greater than the size of the symbols. (a–c) Rim of patchy hornblende phenocryst. Rims show good correlation of Mg-number vs Aliv, and Aliv vs (Na + K)A, and fine oscillatory zoning. (d–f) Simple hornblende phenocryst. Rims have markedly higher Mg-number than cores. (g–i) Multiple zoned hornblende shows poor correlation between (Na + K)A and Aliv. Zones involve large jumps in Mg-number.

 
Simple-zoned hornblende phenocrysts are characterized by a fairly uniform core, and a sharp change in composition to the rim (Figs 11b and 14d–f), with only occasional evidence of embayment or rounding. Al^iv and (Na,K)^A are well correlated (Fig. 14f). Rims have higher Mg-number (Fig. 14e), Na and Si, higher Fe³⁺/Fe²⁺ (calculated using Holland & Blundy, 1994), and lower Ti and Fe compared with the cores (i.e. reverse zoning, Table 4; Fig. 10). Within a single phenocryst core, Al^iv and (Na,K)^A may show a slight rimward decrease, with a corresponding increase in Mg-number (Fig. 14d). Cores and rims are indistinguishable in terms of trace elements. Normally zoned phenocrysts, with cores similar in composition to the rims of reverse simple zoned crystals, are rare.

Zonation in multiple-zoned hornblendes (Fig. 14g–i) is texturally similar to that in simple-zoned hornblendes, but the zones are less clearly defined (Fig. 11c). Zoning is characterized by broad, diffuse zones that lack correlation of Al^iv with Mg-number (Fig. 14h) or (Na,K)^A (Fig. 14i), and are marked by large jumps in Mg-number (Fig. 14g); for example, from 77 to 58 and back. Dark zones have high Mg-number, Na and Si, and low Ti, Mn, K and Fe compared with bright zones (Fig. 10).

Apatite
Both normal and reverse zoning patterns are observed in apatite. Normally zoned crystals have a rimward increase in S, whereas reversely zoned crystals have a rimward decrease. S correlates with Na₂O (Fig. 15), suggesting a coupled substitution mechanism such as S⁶⁺ + Na⁺ = P⁵⁺ + Ca²⁺ (Liu & Comodi, 1993) or 2S⁶⁺ + 4REE³⁺ + Si⁴⁺ + 2Na⁺ = 4P⁵⁺ + 5Ca²⁺ (Parat et al., 2002). In general, microphenocrysts and microlites from the quenched enclave are more sulphur-rich (550–5860 ppm) than those from the host andesite (235–3300 ppm, Fig. 15).

View larger version (15K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 15 Sulphur and Na contents of apatites from host andesite (filled diamonds) and from the quenched enclave (open circles). Very S-rich apatites are much more common in the quenched andesite. 1:1 line indicates atomic substitution of S6+ for Na+. Error bars indicate typical 1 S.D. from EPMA counting statistics.

 
Olivine
Olivine occurs as large, isolated crystals of composition Fo_80–87, and occasionally contains inclusions of En-rich orthopyroxene and chromite. Strong evidence of disequilibrium is seen, varying from a reaction corona of titanomagnetite, orthopyroxene and amphibole, to complete replacement by sub-grains of orthopyroxene, titanomagnetite, amphibole and sometimes plagioclase. The remnant olivine is commonly strongly zoned, with Mg decreasing rimward. These reaction rims are attributed to reaction of olivine with a silica-rich, oxidizing melt (e.g. van Lamoen, 1979; Arculus & Wills, 1980; Matthews et al., 1994; Dirksen et al., 2006).

	DISCUSSION

TOP ABSTRACT INTRODUCTION GEOLOGICAL BACKGROUND ANALYTICAL METHODS ESTIMATION OF INTENSIVE... PHENOCRYST ZONING PATTERNS DISCUSSION CONCLUSIONS REFERENCES

 
Crystal zoning and dissolution textures are produced by changes in some combination of pressure (P), temperature (T), pH₂O, fO₂ or melt composition. Changes in these parameters reflect physical processes acting at different stages of the magmatic history. For example, prolonged periods of cooling and fractional crystallization in the deep crust result in the segregation of evolved, H₂O-rich residual melts (Annen et al., 2006). Magmas ascending from depth may bring with them old crystals or cumulates (e.g. Turner et al., 2003), which may become resorbed during ascent (Dungan & Davidson, 2004). Once stored in a shallow magma chamber, the magma may experience continued crystal fractionation, cooling, convection, mixing or mingling with new magmas of the same or different compositions, oxidation–reduction processes, decompression during magma ascent and pressure fluctuations related to eruptive processes. In this section we attempt to ascribe individual phenocryst textures to specific events or processes in the magmatic history. Figure 16 summarizes the textural features and geochemical characteristics of each type of zoning discussed.

View larger version (32K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 16 Summary of textures and chemical characteristics of phenocryst zoning.

 
Origin of patchy zoning
Patchy zoning in plagioclase
Patchy plagioclase is characterized by an irregular, partially resorbed, An-rich, Ba-poor core, filled in and overgrown by Ab-rich, Ba-rich plagioclase that trapped abundant melt inclusions (Fig. 16a). Similar textures have previously been attributed to skeletal growth (e.g. Kawamoto, 1992) or partial dissolution (e.g. Vance, 1965; Gerlach & Grove, 1982). Here, we suggest that the patchy texture forms as a result of partial resorption during magma ascent.

Under equilibrium conditions, the concentrations of minor and trace elements in plagioclase are controlled by melt concentrations and partition coefficients (Ginibre et al., 2002). We calculated Ba and Sr concentrations of melts coexisting with patchy plagioclase using the method of Blundy & Wood (1991) for T = 840°C. In general, the calculations indicate that early, An-rich plagioclase was in equilibrium with melts considerably enriched in Ba and Sr compared with the Ab-rich plagioclase (Fig. 17a). The concentrations of Ba and Sr in the melt in equilibrium with early, An-rich plagioclase decrease markedly from approximately 1000 ppm Ba and 700 ppm Sr to approximately 400 ppm Ba and 200 ppm Sr (Fig. 17a). Ba and Sr concentrations in the melt then remain constant at these values while the Ab-rich plagioclase crystallizes. The early, decreasing trend is consistent with evolution of the melt during fractional crystallization of plagioclase (An₆₂) + hornblende + phlogopite ± pyroxenes (Fig. 17a). This assemblage is consistent with the mineralogy of cumulate xenoliths, and with the whole-rock data, which require involvement of a K-rich mineral (see Fig. 3). These early formed plagioclase crystals are then partially resorbed, forming the irregular cores. The residual melt following the early fractional crystallization stage has similar trace element chemistry to the later melt that crystallized the Ab-rich plagioclase infill. This lack of variation in melt chemistry implies that the patchy texture develops in response to an external factor such as P, T or pH₂O.

View larger version (15K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 17 (a) Melt Sr and Ba contents calculated to coexist with early (An-rich), late (Ab-rich) and rim compositions of patchy plagioclase. Melt composition changes strongly during crystallization of An-rich plagioclase, thereafter showing little variation. Arrows indicate melt evolution during crystallization of phenocryst phases. Partition coefficients are taken from Blundy & Wood (1991) for plagioclase (pl), Nash & Crecraft (1985) for phlogopite (phl) and Tiepolo et al. (2003) for hornblende (hbl). We assume that DBa and DSr in pyroxenes (px) are 0·1. The melt starting composition is taken as 700 ppm Sr and 1000 ppm Ba. The calculated melt variation requires fractionation of an assemblage of pl + hbl + phl ± px, in the approximate ratio 25:50:15:10. Tick marks indicate percent fractional crystallization. Typical errors from LA-ICP-MS analysis are the size of the symbols. (b) Melt Sr and Ba contents calculated to coexist with sieve plagioclase cores and rims, using partition coefficients calculated from Blundy & Wood (1991). Rims crystallized from a melt richer in Ba and Sr.

 
In the absence of varying melt composition, variations in X_An can be produced by varying any of temperature, pressure and pH₂O (e.g. Housh & Luhr, 1991). We used the plagioclase–liquid thermometer (Putirka, 2005) to assess the range of T, pH₂O conditions that could produce the observed X_An. In these calculations, the external melt composition (i.e. the major element content of the coexisting liquid, a melt inclusion) is constant. The compositions of the host patchy plagioclase and adjacent melt inclusions were used to calculate T and pH₂O variations. For a given melt inclusion (liquid) composition and H₂O content, X_An increases with temperature, whereas for a given plagioclase–liquid compositional pair, H₂O in the liquid increases as temperature decreases. At constant pH₂O, a small temperature decrease (e.g. T < –30°C) is required to move from early, An-rich to late, albitic plagioclase for a typical plagioclase–liquid pair (Fig. 18). Conversely, at constant T, large changes in pH₂O (e.g. P –40 MPa) are required (Fig. 18), equating to significant changes in depth. A positive T is also possible, but pH₂O must be even larger (Fig. 18). These calculations show that patchy textures in plagioclase develop during conditions of decreasing temperature or decreasing pH₂O.

View larger version (16K): [in this window] [in a new window] [Download PowerPoint slide]   Fig. 18 Anorthite content of plagioclase, calculated as a function of temperature and pH2O using Putirka (2005). The large changes in plagioclase composition (e.g. 20 mol % anorthite) from earlier formed, An-rich plagioclase to later, albitic plagioclase can be produced by small temperature decreases (e.g. –25°C) or larger drops in pH2O (e.g. –40 MPa).

 
These conditions are consistent with textural indications of partial resorption, if the magma is initially undersaturated with respect to H₂O. This is because significant resorption can occur during ascent of H₂O-rich, but undersaturated, magmas (e.g. Holtz & Johannes, 1994; Johannes & Holtz, 1996; Clemens et al, 1997; Holtz et al., 2001; Annen et al., 2006). During adiabatic ascent, the melt fraction of the magma increases as a result of resorption (see Holtz & Johannes, 1994, fig. 7), and temperature changes are small (–70°C/GPa; Annen et al., 2006). As the magma rises, aH₂O will increase until the magma becomes H₂O-saturated. Thereafter, exsolution of H₂O will occur as the magma continues to decompress, causing crystallization of Ab-rich plagioclase once the melt crosses its liquidus (Annen et al., 2006).

Patchy zoning in hornblende
The textural similarity between patchy plagioclase and hornblende (see Fig. 8) suggests a common resorption origin. Bright (late crystallizing) hornblende compositions have higher Al^iv and (Na,K)^A, but lower Mg-number than the earlier (Mg-rich, Al-poor) hornblende (Fig. 16b). Experimental studies indicate that this is consistent with crystallization of the later, Al-rich hornblende from a hotter melt (Scaillet & Evans, 1999; Rutherford & Devine, 2003). However, patchy plagioclase compositions indicate that temperatures changes are small, and that resorption is driven by changing pH₂O (see Fig. 18). An alternative explanation is therefore that the composition of hornblende crystallizing is modified in response to the changing plagioclase composition. For example, decreasing pH₂O in the melt at constant T promotes crystallization of Ab-rich plagioclase (Fig. 18). Hornblende is thus driven to more Al-rich compositions, reflecting the edenite–richterite compositional exchange (Holland & Blundy, 1994):

This explanation requires Al-rich hornblende to coexist with albitic plagioclase (An_29–54), and Al-poor hornblende to coexist with anorthitic plagioclase (An_42–74), without significant temperature changes. Hornblende–plagioclase thermometry shows that pairs of patchy hornblende compositions can coexist with pairs of patchy plagioclase compositions, at temperatures that agree within error. Larger changes in X_An require a greater shift in hornblende Al₂O₃. Changes in plagioclase composition, as a result of changing pH₂O at near-constant T, can therefore cause the observed variations in hornblende composition.

Summary
Textures and compositions of patchy plagioclase indicate resorption as a result of H₂O-undersaturated, adiabatic decompression (Fig. 16a). The textural similarity of patchy hornblende (Fig. 16b) suggests that a similar mechanism causes resorption of hornblende. The crystals originate