FIRE and MUD Contents

Relative Timing of Fluid and Anhydrite Saturation: Another Consideration in the Sulfur Budget of the Mount Pinatubo Eruption

Jill Dill Pasteris,¹ Brigitte Wopenka,¹ Alian Wang,¹ and Teresa N. Harris¹

¹ Department of Earth and Planetary Sciences, Washington University, Campus Box 1169, St. Louis, MO 63130-4899.

ABSTRACT

Raman microsampling spectroscopy and petrographic analysis were done on four major types of inclusions in quartz and plagioclase phenocrysts in phenocryst-rich dacite from the June 15, 1991, eruption of Mount Pinatubo, Philippines: (1) glass inclusions, typically containing one or more bubbles, (2) vapor-dominated inclusions, (3) aqueous inclusions consisting of liquid and vapor ± solids, and (4) solid inclusions. In quartz-hosted glass inclusions with bubbles, Raman analysis revealed only CO₂ in some enclosed bubbles, but not SO₂ or H₂S (below detection limit of 1 bar partial pressure); dissolved H₂O, but not CO₂ (less than the detection limit of 900 parts per million) was detected in the glass phase. The observed bubble:glass ratios in such inclusions, the Raman-inferred CO₂ pressures in some bubbles, the documentation by Raman spectroscopy of gas leakage from bubbles over time, and the petrographic documentation of physical breaching of many glass inclusions suggest that these inclusions may represent the two-phase entrapment at depth of coexisting melt and supercritical fluid but that the fluid was not retained. No volatile species were detected in the vapor-dominated inclusions, so very low gas pressures are indicated. Raman spectroscopic analysis revealed anhydrite among the solid phases in fracture-lining aqueous inclusions that are found only in quartz, implying an elevated sulfate content in late-stage aqueous fluids. Solid inclusions of apatite, zircon, and amphibole were identified spectroscopically. Of note is the lack of anhydrite inclusions in our samples. This observation, coupled with reports of high SO₃ in apatite inclusions, lack of sulfur isotopic zonation in anhydrite phenocrysts, and relatively low bulk-sulfur contents in Mount Pinatubo dacites (only 25% of those in El Chichón eruptive rocks), suggests that the Mount Pinatubo dacitic magma may have reached saturation with respect to a CO₂-SO₂-H₂O fluid before anhydrite saturation occurred. Since sulfur strongly partitions into a fluid over a melt phase, early saturation of the magma with a supercritical aqueous fluid could have effectively stripped sulfur from the melt. In this case, neither the bulk-sulfur content nor the abundance of anhydrite crystals in eruptive rocks could be used as an indicator of the original sulfur content of the system. If fluid saturation indeed occurred early in Mount Pinatubo's magmatic history, then the (H₂O+CO₂)/SO₂ ratio in the melt may have controlled the timing of the removal of sulfur from the melt via fluid exsolution and the proportion of the available sulfur that ultimately reached the atmosphere.

Note to readers: Figures open in separate windows. To return to the text, close the figure's window or bring the text window to the front.

INTRODUCTION

Mount Pinatubo is one of the active members of the Luzon volcanic chain, which is part of the Bataan frontal arc. It lies about 120 km east of the Manila trench, above the eastward subducting Eurasian plate. Mount Pinatubo is also within the southern part of the areally extensive Zambales ophiolite. The volcano's eruptions in June 1991 ended a 500-year hiatus in activity. The climactic June 15th eruption produced an estimated 4 to 5 km³ dense-rock equivalent of dominantly dacitic pumice and tephra (W.E. Scott and others, this volume) and released about 20 Mt SO₂ into the atmosphere (Bluth and others, 1992). One of the working hypotheses for explaining the large amount of sulfur released into the atmosphere is that a separate, sulfur-bearing volatile phase coexisted with the dacitic magma at depth prior to eruption (see Westrich and Gerlach, 1992; Gerlach and others, this volume). Various sources of the sulfur have been considered, including an underlying basalt reservoir and a hydrothermal system (Gerlach and others, this volume; Pallister and others, this volume; Rutherford and Devine, this volume).

The goals of our study of fluid and solid inclusions in quartz and plagioclase phenocrysts are to provide more data by which to evaluate the hypothesis of a coexisting fluid phase during eruption and to determine what kinds of information can be obtained on the available samples by Raman microsampling spectroscopy. Raman analysis of volatile-bearing and solid inclusions can help to track the movement of sulfur within the volcanic system and provide information on the state of gas saturation of the magma during the time of phenocryst precipitation. Combined Raman, petrographic, and petrologic data are used to infer the mechanism by which gas saturation was reached.

ANALYTICAL TECHNIQUE

Raman spectroscopy is an optical technique that monitors the inelastic scattering of monochromatic visible light as it interacts with covalently bonded molecules in solids, liquids, or gases. The position of the peaks and the number of peaks in the Raman spectrum are controlled by the energy and symmetry, respectively, of the molecular vibrations. In Raman microsampling spectroscopy, the sample is both viewed and excited in a research-grade optical microscope. The same high-magnification, high-numerical-aperture objective is used to optically image the sample (via a video camera and a TV monitor), focus the laser (to about 1 m spot diameter), and transmit the scattered radiation to a monochromator. Nondestructive analysis is performed at ambient atmosphere and temperature and can be done on materials as small as 1 m in diameter on or below the optically transparent surface of a sample. Geological samples require minimal or no sample preparation in that (1) unmounted polished wafers, polished thin sections, loose grains on a glass slide, and irregular masses of material all can be analyzed and (2) no carbon coating or other sample preparation is required; unopened fluid inclusions can be analyzed nondestructively. The inelastically scattered Raman radiation is monitored by a photon detector and recorded in terms of intensity (number of photons per second) as a function of Stokes Raman shift (relative wavenumbers, cm^-1). The Raman shift is the difference in frequency between the exciting laser radiation and the Raman-scattered radiation.

Raman microsampling spectroscopy was done with either a single-channel Jobin-Yvon RAMANOR U-1000 or a multichannel Jobin-Yvon S-3000 laser Raman microprobe, both marketed by Instruments SA. In both of our instruments, the 514.5-nm green line of an argon-ion laser is focused into the sample by a modified research-grade microscope. The beam can be focused into an inclusion or portion of an inclusion, typically below the surface of the polished sample. The configuration of both of our instruments is such that 180° backscattered radiation is collected. The higher the numerical aperture (N.A.) of the objective, the better is the optical throughput. Given the fact that the Raman effect is an extremely weak phenomenon, it is desirable to use the objective with the highest numerical aperture possible. However, the size of an inclusion and its depth below the surface (necessitating a specific free-working distance) also dictate the choice of an objective. We used the following objectives from several manufacturers: 80x Nachet (N.A. 0.90), 80x ultra-long-working-distance Olympus (N.A. 0.75), 100x Olympus (N.A. 0.95), and 160x Leitz (N.A. 0.95). The laser power at the surface of the sample usually was fixed at 15 mW, but lower power was used when the inclusion showed signs of laser heating (see below).

The U-1000 consists of a 1-m double monochromator with two plane holographic gratings (1,800 grooves/mm). The detector is a thermoelectrically cooled RCA C-31034-2 photomultiplier tube. Spectra are obtained in the dispersive scanning mode with variable step size and counting time. Typical analytical conditions for analysis of gases were 0.5 cm^-1 stepping interval and 10 s dwell time per point. The volatiles that typically were analyzed for are CO₂, CO, N₂, H₂S, SO₂, CH₄, and H₂O. For analysis of solids, the stepping interval was increased to 1 cm^-1, and dwell times varied between 1 and 10 s, depending on the strength of the signal.

The S-3000 is a triple-monochromator system that consists of a 320-mm double-monochromator with two 600 grooves/mm gratings (acting as a two-stage foremonochromator in subtractive mode) and a 1-m third-stage monochromator with interchangeable plane holographic gratings. For the analyses performed in this study, we used a grating with 600 grooves/mm in the third-stage monochromator. The detector is a 1-in, 1,024-element, proximity-focused, intensified optical diode array that is cooled by a Peltier-effect thermoelectric photocathode cooler. To cover the total spectral range of interest (from ~100 cm^-1 to ~4,000 cm^-1), data were acquired in four different spectral windows centered at 830, 1,600, 2,600, and 3,500 cm^-1. The acquisition time was chosen such that a satisfactory signal-to-noise ratio was achieved but such that the detector was not oversaturated in the wavenumber region of interest. We typically took 20 acquisitions at 10 s for each window.

The choice of which of these instruments to use is governed by the type of spectral information required and the amount of time available. As a scanning monochromator with gratings of high groove density, the U-1000 is the instrument of choice when higher spectral resolution and exact knowledge of band position are required. It is particularly useful when the Raman bands for gases are used to infer gas pressure and when exact peak positions for solids (such as zircons in this study) are required. Analyses on the U-1000 are time consuming, however, requiring about 10 min for each gas to be analyzed. The S-3000 is the instrument of choice for reconnaissance analysis of a wide range of gas species and rapid (1-2 min) identification of solid phases. A total of only about 25 min, including the time necessary for background measurements, is required for a full-spectrum analysis under the conditions listed in the preceding paragraph.

SAMPLES

The pumices that we studied (obtained from H.R. Westrich, Sandia National Laboratory, and T.M. Gerlach, U.S. Geological Survey) were collected from the surface of the pyroclastic-flow deposits in the Sacobia River valley, northeast of Mount Pinatubo. The pumices are of white, phenocryst-rich dacite, which comprises about 80 vol% of the June 15th tephra. Grain separates of quartz phenocrysts (provided by H.R. Westrich and T.M. Gerlach) were made into doubly polished thick wafers, which were cemented onto a glass plate with UV-curing epoxy, whereas the porous pumice samples were vacuum impregnated with blue epoxy before they were made into doubly polished thick rock wafers. The rock wafers were cemented onto glass plates by use of acetone-soluble superglue, which permitted them to be released as free-standing wafers. (Despite their epoxy impregnation, the rock wafers were still too friable to be analyzed by microthermometry on a gas-flow stage.) All samples were appropriate for both petrography and in-situ Raman microprobe analysis.

The typical phenocryst phases in these rocks, as reported in petrologic studies documenting a large number of samples, are plagioclase (An_34-66), hornblende, cummingtonite, biotite, quartz, magnetite, anhydrite, ilmenite, and apatite (Rutherford and Devine, this volume; Pallister and others, this volume). Petrography of our samples indicates that plagioclase phenocrysts are about 50 times more abundant than quartz and that the quartz grains typically show resorption features. We did not find any anhydrite phenocrysts in our samples, which could be due to the fact that water was used in some of the sample preparation.

Our Raman study concentrated on fluid and solid inclusions in quartz and plagioclase phenocrysts. These two mineral hosts were chosen for their optical clarity, for the abundance of plagioclase phenocrysts in pumice samples, and for the availability of grain separates of quartz phenocrysts. Inclusions were inventoried and analyzed in four thin sections and doubly polished wafers and in 35 quartz grains in grain mounts (table 1). Approximately 30 solid and 30 fluid inclusions were analyzed by Raman spectroscopy.

Table 1. Description of inclusions within quartz and plagioclase phenocrysts from the June 15, 1991, crystal-rich dacite.

THE INCLUSIONS

Four major types of inclusions were recognized optically in the quartz, plagioclase, and hornblende phenocrysts: (1) glass inclusions, usually containing one or more bubbles, (2) vapor-dominated inclusions, (3) aqueous inclusions, typically consisting of a large bubble, a small liquid phase, and multiple solid phases, and (4) solid inclusions. Only in quartz were all four types found. Raman microprobe analysis was done on examples of each of these types of inclusions in quartz and plagioclase phenocrysts.

GLASS INCLUSIONS

The glass inclusions are by far the dominant type in both the quartz and plagioclase phenocrysts, as well as in the unanalyzed hornblende grains. In quartz, glass inclusions tend to be square with rounded corners, reflecting the symmetry of the ß-quartz phase in which they were trapped initially. In hornblende, the glass inclusions are irregularly to rectangularly rounded. In plagioclase, they are ovoid to rectangular. Inclusion diameters typically range from 5 to 60 m, but a few inclusions exceed 100 m. Most of the glass inclusions are colorless and contain at least one bubble. Much less common are glass inclusions that are light brown (± small bubbles) and inclusions whose glass phase appears to be flecked with minute dark particles (glass undergoing devitrification?). The number and size of the bubbles vary greatly among glass inclusions, as does the total volume percent of bubbles, even among inclusions within the same phenocryst grain (fig. 1A). Even in the smallest glass inclusions (<20 m), there is a variable bubble:glass ratio.

Figure. 1. Photomicrographs of inclusions within phenocrysts of Mount Pinatubo dacite. A, Glass inclusions in quartz showing different numbers and sizes of bubbles and different bubble:glass ratios. B, Growth zones in plagioclase lined by glass inclusions; two apatite inclusions on left. C, Core of plagioclase phenocryst densely populated by single-bubble glass inclusions with constant bubble:glass ratio. D, Examples of isolated inclusions in plagioclase: glass inclusion with a single gas bubble in the middle, intersected by narrow fracture (arrow); three small one-phase, approximately square, high-relief vapor inclusions to the left. E,F, Multiple bubbles in glass inclusions in quartz; note bimodal size distribution in F and high bubble:glass ratio in both inclusions. G, Glass inclusions in quartz: both inclusions show light halos that are interpreted as partial healing of a decrepitation zone; left inclusion is crosscut by a long fracture. H, Glass-crystal inclusion in quartz; tablet-shaped birefringent daughter mineral is plagioclase. I, One-phase euhedral vapor inclusion (bottom left) coexisting with multiphase aqueous inclusion (top; smaller, similar inclusion on right) in quartz (see text and Raman spectra of larger aqueous inclusion in figures 3B and 4A). J, One-phase rounded to elliptical vapor inclusions in quartz; cloudiness induced by density of inclusion population. K, Zircon inclusions (needles) and apatite inclusions (blades) in plagioclase phenocryst. L, Large apatite needle isolated in plagioclase phenocryst; two tubular structures in core give no Raman signature and are inferred to be voids created during rapid growth. M, Amphibole inclusion (blade) in quartz phenocryst. N, Zircon inclusion in quartz.
Figure1A-H
Figure 1I-N

There are three major differences in the occurrence of glass inclusions hosted by quartz and by plagioclase phenocrysts. First, only in plagioclase do arrays of glass inclusions outline growth zones, which arrangement clearly identifies them as primary inclusions (fig. 1B). Such inclusions usually are smaller than the average for plagioclase. Furthermore, the cores of some plagioclase phenocrysts are densely populated with what appear to be primary glass inclusions (fig. 1C). The rest of the glass inclusions in plagioclase, as well as those in quartz, appear to be isolated individuals (fig. 1D), randomly arranged in clusters, or aligned along fractures (assumed or readily visible). Second, single bubbles are common in glass inclusions in hornblende and plagioclase (fig. 1C,D), whereas multiple bubbles (typically 4-8) are common in glass inclusions in quartz (fig. 1A,E,F). In the latter, bubbles can comprise 50+ vol% of the inclusion, in many cases showing a bimodal size distribution; the larger bubbles are on the order of 15 m. Third, the association of glass inclusions with fractures is particularly strong in quartz. Many glass inclusions reside within brownish ribbonlike features inclined to the quartz sample's surface, which are interpreted as incompletely healed fractures (fig. 1G). Short, dark fractures in many cases crosscut the glass inclusions. There is no apparent relation between the volume percent of bubbles in glass and the presence of a fracture. Only in quartz phenocrysts are "halos" seen around glass inclusions (fig. 1G). These halos are recognized by an optical interface exactly concentric with the square inclusion and located 5 to 10 m outward from the edge of the glass. The halo zone is perfectly transparent and optically continuous with the rest of the host quartz. It is recognized as a separate entity because it has even less color than the immediately surrounding quartz and because its outer interface in many cases marks the transition into a fluid-filled decrepitation zone. The Raman spectra of the halos around those inclusions are indistinguishable from the Raman spectrum of the quartz host. It is tempting to label these fracture-associated glass inclusions in quartz as secondary, that is, trapped after the phenocryst already had formed and subsequently had undergone brittle fracture. However, it is possible that the difference in the thermal coefficient of expansion between quartz and the glass/melt inclusions induced fracturing during cooling of the phenocryst.

Both the obvious sharp, dark fractures that intersect many glass inclusions and the halos around inclusions are evidence of breaching of the glass and release of fluids in quartz phenocrysts. At low magnification, the halo-rimmed inclusions are seen to reside within brownish zones indicating major fractures. We infer that each halo, with its enhanced optical clarity (fig. 1G), represents a zone of localized healing within the fracture. The healing was induced by fluid released from the glass inclusion. Such healing could have occurred only at elevated temperature, not during sample preparation. The small dark fractures emanating from many glass inclusions are common in quartz phenocrysts from volcanic rocks and may represent the approximately 1% decrease in volume that occurs in the transition of ß to quartz (see Roedder, 1984, p. 69). In our rock wafers that were vacuum impregnated with blue epoxy, there is also evidence of breaching of glass inclusions in plagioclase: Veils of blue within phenocrysts indicate throughgoing fractures; spherical blue blebs commonly occur within glass inclusions, usually with no optical indication of any fracture connection to the sample's surface.

Most of the above-listed breaches are believed to have occurred at elevated temperatures, as previously explained. However, some fracturing and release of fluid apparently occurred in response to sample preparation: Raman analyses for CO₂ on the same bubbles in several glass inclusions over a 3-month period showed a noticeable upshift in the Raman peak position and a marked decrease in the peak's intensity, both indicating a drop in CO₂ pressure (fig. 2).

Figure 2. Raman spectra of CO₂ in same bubble of glass inclusion in quartz, providing evidence for leakage over time. A, Original analysis. B, Repeat-ed analysis 3 months later, in which same analytical settings were used. cps, counts per second.

Most of the Raman microprobe analyses of bubbles in glass inclusions in quartz revealed no gases. It should be noted that the Raman spectrum of water vapor in the bubbles cannot be distinguished from the spectrum of water dissolved in the enclosing glass matrix of the bubbles (fig. 3A). Thus, we cannot comment on the presence of water within the bubbles in glass. Some bubbles, however, show the Raman spectrum of CO₂ (fig. 2). Moreover, in some of the gas-bearing bubbles, the specific peak position for CO₂ (based on calibrations of pure CO₂ done in our laboratory by J.C. Seitz, 1993) indicates internal gas pressures on the order of a couple of tens of bars partial pressure of CO₂ (assuming that the CO₂ peak positions are minimally affected by the presence of low-density water vapor). No SO₂ (dominant sulfur gas species under preeruption conditions) or H₂S (dominant sulfur gas species at room temperature, in presence of water) was detected in any of the bubbles. If the latter species are present in the bubbles, they are below our detection limits of approximately 1 bar partial pressure. Because no CO₂ was detected in the glass, its concentration is below our detection limit, which is estimated to be on the order of 900 ppm. The natural low-level fluorescence of plagioclase increased the spectroscopic background and thus precluded the Raman detection of any gas species in glass inclusions hosted by plagioclase.

Figure 3. Raman spectral sensitivity to different forms of H₂O in inclusions in Mount Pinatubo phenocrysts. Raman spectra of (A) water dissolved in glass inclusion in quartz and (B) liquid water in multiphase aqueous inclusion in quartz (shown in fig. 1I). cps, counts per second.

In rare cases, birefringent phases occur in the glass-dominated inclusions. The individual tablet-shaped solids enclosed in some glass inclusions in quartz (fig. 1H) have been identified as plagioclase by Raman spectroscopy. In several plagioclase phenocrysts, there are glass inclusions that contain single bubbles that are defined by a low-birefringent thin shell that is not quite round in cross section. Raman analysis of the shell produces no bands other than those of the plagioclase host. Thus, the shell is believed to represent plagioclase that has nucleated and grown from the glass.

VAPOR-DOMINATED INCLUSIONS

One-phase inclusions that appear to be totally vapor occur in low abundance in both plagioclase and quartz phenocrysts. In plagioclase, approximately square, dark (high-relief) inclusions of 3 to 12 m in diameter occur as isolated individuals or in arrays of two or three inclusions (fig. 1D). In quartz, one-phase inclusions occasionally occur in small numbers associated exclusively with glass inclusions. More typically in quartz, one-phase inclusions occur closely associated with vapor-dominated (>90 vol%) and other less vapor-rich aqueous inclusions that also may have daughter minerals (see below). In such cases, the 5- to15-m one-phase vapor inclusions are subhedral to euhedral, approaching the negative crystal shape of ß-quartz dipyramids (fig. 1I). In these cases, the aqueous and vapor inclusions appear to line secondary fractures. No CO₂, H₂O, SO₂, or H₂S was detected in the one-phase vapor inclusions. If present, these gases occur at levels below the detection limits of about 1 bar partial pressure for each species listed, except H₂O, whose detection limit is higher but unquantified. These inclusions show no physical evidence of leakage, and we infer that they represent the trapping of very low density vapor.

Several quartz grains are almost homogeneously translucent, due to their incorporation of a myriad of one-phase, rounded to elliptical high-relief inclusions of a wide range of sizes (fig. 1J). In several grains, not all the inclusions have one phase; some inclusions have a few volume percent liquid (probably water) at one or both extremities (similar to the aqueous inclusions described below), and some of these show visible fractures projecting from them. It is difficult to characterize such inclusions as primary or secondary. The optical properties of such grains are so poor as to preclude Raman analysis of the inclusions. Our interpretation is that these quartz grains precipitated in the presence of a vapor-saturated fluid.

AQUEOUS INCLUSIONS

A small proportion of inclusions is recognized by the presence of a colorless liquid phase (identified as water by Raman analysis) along the walls or in the extremities of the inclusion. Such aqueous inclusions typically are less than 30 m in diameter and are ovoid to euhedral (doubly terminated prismatic; some dipyramids) in shape. They were observed only in quartz phenocrysts (in grain separates and in situ in rock wafers) and only in about one out of every seven crystals. The quartz grains containing aqueous inclusions also have a small number of square glass inclusions elsewhere in the grain. Where aqueous inclusions occur, they usually dominate the inclusion population of that phenocryst grain or particular volume within the phenocryst by densely populating what appear to be secondary fractures. The aqueous inclusions display various amounts of liquid water or brine, typically less than 30 vol%. Most aqueous inclusions are dominated by a vapor bubble, which can comprise more than 90 vol% in liquid-vapor inclusions and up to 70 vol% in liquid-vapor-solid inclusions (fig. 1I). In the bubble phase of such inclusions, all of the gases analyzed for, including H₂O, were below our Raman detection limits. The small amounts of visible liquid gave a strong Raman band for the O-H stretching vibration of liquid water in the spectral region 3,000 to 3,600 cm^-1 (fig. 3B). However, there was no obvious Raman band in the 950 to 1,150 cm^-1 region for dissolved sulfate.

Solid, apparent daughter phases were investigated in about 10 different aqueous inclusions, which can be subdivided into two groups on the basis of whether any of the solids showed birefringence. The inclusions with birefringent solids tend to have the negative-crystal shape of doubly terminated prisms and to be smaller (<15 m) than those with only isotropic solids. In some cases, only a birefringent solid is present, whereas in other cases, it appears that there are isotropic solids at one end and birefringent ones at the other end of the inclusion. It is difficult to optically distinguish a liquid phase in such inclusions, but they have a bubble whose size ranges from 10 to 80 vol%. There are population clusters whose proportions of solid phases all appear to be very similar. Raman spectroscopy was done on five of the small birefringence phases, but no bands other than those for quartz were detected. Thus, the daughter phases either are ionic crystals (poor Raman scatterers), are weak Raman-scatterers, and (or) have Raman bands in spectral positions masked by those for quartz. A simple ionic phase, such as NaCl, is unlikely because the crystals are birefringent.

In one quartz phenocryst, Raman analyses were done on five different aqueous inclusions showing only isotropic phases. These particular inclusions were chosen for their size and good optics and appear representative of that subgroup of aqueous inclusions: euhedral but not prismatic, vapor dominated, and exhibiting multiple solids and a liquid phase. The most diverse suite of solids was analyzed in the inclusion shown in figure 1I. The largest solid (top of inclusion, immediately left of center) is isotropic, colorless, square, and of low relief. It yields no Raman bands except those of the host quartz and is inferred to be halite. The small, dark daughter crystal (near upper apex of inclusion) appears to heat under the laser beam and to move around in the inclusion. Laser power was reduced to minimize this effect, but no additional Raman bands were detected; its identity remains unknown. The other colorless, isotropic crystal (top, immediately right of center) that has higher relief than the inferred halite has Raman bands at about 170 and 240 cm^-1, in addition to any bands that might be masked by those of the quartz host. Its identity also has not yet been established. Of greatest interest in the present study are the two smallest colorless daughter crystals (top apex). These gave Raman bands at 1,018 and 1,130 cm^-1, which are diagnostic of anhydrite (fig. 4). Their identification as anhydrite was unexpected because the crystals show no birefringence. However, the crystals do appear to demonstrate the retrograde solubility of anhydrite, because additional precipitation seems to occur when the laser beam irradiates the area occupied by these crystals and presumably causes localized heating. The lack of detection of the Raman band for dissolved sulfate in the associated liquid is explained by the fact that anhydrite reaches saturation at very low levels of sulfate concentration. The other four aqueous inclusions also showed the Raman bands of anhydrite but no birefringence. It is inferred that the anhydrite exists as very fine-grained material that does not exhibit strong birefringence.

Figure 4. Raman spectra of (A) colorless daughter mineral (anhydrite) in multiphase aqueous inclusion in quartz (shown in fig. 1I), (B) quartz matrix near the same inclusion, and (C) anhydrite standard.

SOLID INCLUSIONS

Inclusions of several types of transparent and opaque phases are common within plagioclase, quartz, and hornblende phenocrysts. Raman analysis was done on several types of transparent solids. Apatite is a common inclusion phase; it is markedly more abundant in plagioclase than in quartz. Apatite inclusions occur as slightly yellow-green, low-birefringent, stubby prisms and needles on the order of 5 to 50 m long, with an aspect ratio of about 1:3 to 1:15, and they typically do not display crystallographic orientation within their host. Isolated apatite grains are the more common occurrence, but low-density aggregates of apatite prisms also occur (fig. 1K). In some cases, apatite prisms are intergrown with opaque phases. Some of the wider apatite prisms (several micrometers in diameter) have tubular core structures (fig. 1L). Raman analyses of the tubules show no bands other than those for the host apatite, so it is suggested that the tubules are structures that developed during rapid growth (see Wyllie and others, 1962; Gardner, 1972) and are not elongated mineral or fluid inclusions. Of considerably lesser abundance are inclusions identified by Raman spectroscopy as amphibole (fig. 1M). Despite intensive petrographic and spectroscopic analysis, we found no inclusions of anhydrite in either plagioclase or quartz phenocrysts.

The solid inclusions that proved most interesting petrologically are zircons. In contrast to other reports of no or only rare zircon inclusions (Matthews and others, 1992), in our samples, zircon is abundant as inclusions within quartz phenocrysts; some quartz phenocrysts contain 10 or more zircon grains of various shapes. The grains range from stubby ovoids of a few micrometers length to elongated prisms (10+ m long) that, in many cases, show good terminations (fig. 1N). The prisms show parallel extinction but only moderate birefringence, due to their thinness. In contrast, zircon inclusions are rare in plagioclase. Raman analysis indicates that where needlelike and bladelike solid inclusions occur together in a phenocryst (fig. 1K), the needles typically are zircon and the blades apatite.

Three spectral features of the zircon inclusions investigated in quartz and plagioclase are of note: (1) the Raman peak positions of several bands characteristic of zircon (for example, Knittle and Williams, 1993) are upshifted, by as much as 10 cm^-1, in essentially all zircon inclusions in quartz, (2) in several cases, the Raman peak positions are different from one zircon inclusion to the next within the same quartz phenocryst host, and (3) only the zircon inclusions (rare) in plagioclase have a "normal" Raman spectrum. We believe that the spectral upshifting is due to chemical substitution in the zircons. It was difficult to obtain electron microprobe analyses of the zircon inclusions because it is difficult to grind and polish the sample so as to intersect (but not eliminate!) such small individual grains. The electron microprobe analysis (by K. Bartels, Washington University) of a zircon inclusion whose main Raman band lies at 1,114 cm^-1 revealed about 1.5 wt% HfO₂, which is not compositionally unusual. We infer that the spectroscopically important substitution may be by a light element, and are doing further studies to document the chemistry of the zircons.

DISCUSSION

The goal of this paper is to integrate our data on fluid and solid inclusions with other available petrologic information in order to better understand the evolution of the magmatic system at Mount Pinatubo. The andesitic to dacitic eruptive sequence in mid-June was reversed from the more usual volcanic compositional progression from more felsic to more mafic. Data from coexisting iron-titanium-oxide phases indicate that, at eruption, the dacitic magma had a temperature of 780 ± 10°C and an oxygen fugacity (f_O2) of 3 log units above that of the nickel-nickel oxide buffer (f_O2 = NNO + 3; Pallister and others, this volume; Rutherford and Devine, this volume); these factors indicate that the dacites from Mount Pinatubo are more oxidized than the trachyandesites from El Chichón. The preeruption pressure at Mount Pinatubo is estimated to be about 2,200 bar (Rutherford and Devine, this volume). Gerlach and others (this volume) calculate that the fluid phase that evolved from the magma was approximately 85 mol% H₂O, 10 mol% CO₂, and 5 mol% SO₂. These data provide important constraints on the timing of volatile saturation and the sulfur budget at Mount Pinatubo, which can be evaluated further on the basis of information from solid and fluid inclusions.

EVIDENCE FROM THE SOLID PHASES

It is difficult to determine the relative timing of phenocryst precipitation at Mount Pinatubo on the basis of textural relations alone. However, it seems significant that plagioclase is so much more abundant than quartz and that quartz grains typically show resorption. A difference in the timing of the initial precipitation of plagioclase and quartz also is evidenced by the differences in the types of their inclusions--for instance, the higher bubble:glass ratio in glass inclusions in quartz and the occurrence of late-stage aqueous inclusions exclusively in quartz. These observations suggest that quartz phenocrysts postdate most of the plagioclase phenocrysts. However, plagioclase (and hornblende) almost certainly continued to precipitate after quartz reached saturation.

Spectral differences between zircon inclusions in plagioclase and quartz and spectral variations among zircon inclusions within the same quartz grain are evidence that zircons from different petrologic environments were encapsulated during the growth of these phenocrysts. In the context of what is known about the petrology and geophysics of the volcanic system at Mount Pinatubo, at least the following possibilities should be considered to explain the different spectral signatures of the zircon inclusions: (1) zircons reached saturation early in two different melts that presumably mixed before the major phenocryst phases crystallized in the dacite (there is no additional petrologic evidence to substantiate such mixing; J. Pallister, USGS, oral commun., 1993), (2) zircons reached saturation at different levels in a compositionally zoned magma chamber and thereby produced different minor- and trace-element signatures in grains that subsequently were encapsulated in growing phenocrysts (no additional petrologic evidence for this), (3) some of the zircons crystallized directly from a silicate melt, whereas others precipitated from a supercritical aqueous fluid that coexisted with the melt, (4) the zircons and their complex chemistry reflect only local equilibria at the interface between siliceous melt and phenocrysts precipitating from it (Michael, 1988; Bacon, 1989), and (5) some of the zircons are inherited from wallrocks that were mechanically and (or) chemically incorporated into the magma. The latter may be true for the equidimensional smallest zircons but not for the abundant needlelike zircons with excellent terminations (fig. 1N).

It appears significant that, although anhydrite is recognized as a phenocryst phase that precipitated from the magma (Bernard and others, 1991; Pallister and others, 1992; Imai and others, 1993; Rutherford and Devine, this volume; Pallister and others, this volume; Fournelle and others, this volume), it does not occur as inclusions within plagioclase or quartz phenocrysts in the samples we studied and indeed is relatively rare as inclusions in other Mount Pinatubo rocks (Pallister and others, this volume; Rutherford and Devine, this volume; Fournelle and others, this volume; J. Fournelle, University of Wisconsin, written commun., 1994). It thus seems possible that, in most parts of the magmatic system, anhydrite reached saturation only after quartz phenocrysts had begun to precipitate, meaning that anhydrite was a late phenocryst phase. From our Raman data, we know that anhydrite definitely had reached saturation by the time of the development of late-stage aqueous fluids. Our hypothesis of late saturation with anhydrite, which addresses the overall petrologic character of the dacite, does not preclude the possibility of early, local saturation of the magma with small amounts of anhydrite (see point 4 above concerning zircon). Nor is our hypothesis in conflict with the late entrapment of anhydrite phenocrysts in early-saturating phases such as hornblende and plagioclase (Fournelle and others, this volume), which continued to precipitate throughout much or all of the crystallization history of the dacite.

STATE OF FLUID SATURATION

Several petrologic studies (Gerlach and others, this volume; Pallister and others, this volume; Rutherford and Devine, this volume) have provided chemical evidence that the Mount Pinatubo dacitic magma became saturated with an aqueous fluid during its evolution, but the timing of the saturation still remains unconstrained. On the basis of mass balance, Gerlach and others (this volume) argue convincingly that the preeruptive melt coexisted with a separate fluid phase. The best candidate inclusions from which to evaluate the state of fluid saturation of the dacitic magma are the glass inclusions with bubbles, which are enclosed within quartz and plagioclase phenocrysts. The large bubble-to-glass ratio (bubbles representing up to 50+ vol% of the inclusion) clearly indicates that the bubbles cannot simply be due to contraction during cooling of the glass. The spherical shape of the bubbles reflects the molten state of the host when the bubbles were formed or trapped. The common occurrence of multiple bubbles probably reflects the high viscosity of the rhyolitic melt, which hindered coalescence of the gas packets. (In Lowenstern's (1994) heating experiments on hydrous peralkaline rhyolitic melt inclusions, nucleation of multiple bubbles commonly occurred during cooling.) The common bimodal distribution of bubble sizes in Mount Pinatubo glass inclusions further suggests that there were two events of bubble formation or entrainment. If the bubbles in glass were formed in a closed system, the two most reasonable mechanisms by which they formed are (1) by the degassing of volatiles from the melt inclusions as a result of the decreases in pressure (and temperature) that occurred during the phenocrysts' rapid rise to the surface (exsolution model) and (or) (2) by the trapping of a magmatic volatile phase that coexisted with the melt at depth (magmatic entrapment model).

The exsolution model can be evaluated from data on the solubility of CO₂ in rhyolitic melts under the pressure-temperature conditions pertinent to Mount Pinatubo (Fogel and Rutherford, 1990) and infrared data on the concentration of CO₂ dissolved in the glass inclusions as they exist now (Wallace, 1993; Rutherford and Devine, this volume). An upper bound on the pressure of exsolved gas in the glass inclusions can be calculated by overestimating the amount of CO₂ expelled during decompression and underestimating the relative volume of bubbles. The rhyolitic melt that was trapped in the phenocrysts has an estimated density of 2.4 g/cm³ (Westrich and Gerlach, 1992). If the inclusions represent CO₂-saturated melt entrained at 800°C and 2.2 kbar pressure, then the initial concentration of dissolved CO₂ was about 1,800 ppm by weight (Fogel and Rutherford, 1990). This value is appropriate for a very low-H₂O rhyolite, rather than an H₂O-saturated melt like that at Mount Pinatubo. From Fourier-transform infrared (FTIR) analyses, Wallace (1993) reports a range of 280 to 415 ppm CO₂ in Mount Pinatubo glass inclusions in quartz that were chosen because they show no visible cracks and have either no bubbles or only very small bubbles (glass presumably not degassed). In contrast, Rutherford and Devine's (this volume) FTIR analyses of Mount Pinatubo glass inclusions in quartz and plagioclase indicated dissolved CO₂ concentrations of less than 20 ppm (glass presumably degassed). For the purpose of modeling, we assume an upper-bound loss of 900 ppm by weight of CO₂ from the time of trapping, and the retention of that CO₂ in bubbles in the glass. If the bubble(s) represent 20 vol% of the glass inclusion, then the partial pressure of CO₂ in each bubble should be about 6 bar--if no leakage has occurred. This pressure is significantly lower than that indicated by Raman spectroscopy of some bubbles in glass inclusions within quartz, which appear to contain up to a couple of tens of bars partial pressure of CO₂.

On the other hand, if the bubbles represent a magmatic fluid coexisting with the melt at 2 kbar pressure and 800°C, then isochoric cooling would produce an inclusion that now should consist of a vapor phase that is enriched in CO₂ compared to the surrounding phase of liquid water, which should comprise several tens of vol% of the total "bubble" (Roedder, 1984, p. 226). In fact, no liquid is visible in the bubbles, nor did the Raman spectra that were obtained for H₂O in several inclusions show a band structure indicative of liquid water (see Wopenka and others, 1990). Therefore, the Raman data indicate much lower pressures and densities than those expected for the magmatic entrapment of fluids.

Thus, because neither of the endmember models presented above explains the spectroscopic observations, a more complex mechanism or open-system behavior is suggested. One possibility is that the fluid species diffusively reequilibrated with melt surrounding individual phenocrysts (Qin and others, 1992). However, the rapidity of the cooling process and the short time since eruption do not support such a model. On the other hand, there is overwhelming evidence of physical leakage from the bubble-bearing glass inclusions: abundant fractures transecting glass inclusions, partially healed decrepitation halos around glass inclusions (fig. 1G), decreases in CO₂ pressure in individual bubbles over time as documented by repeated Raman analyses (monitoring shifts in band position and decreases in intensity; fig. 2), and observations of the infiltration of colored epoxy into glass inclusions induced by vacuum impregnation of the rock before it was sectioned. Thus, the gas pressures currently monitored in the bubbles in glass inclusions cannot be used to constrain the pressure at which trapping occurred. Leakage of volatiles from glass inclusions is a common phenomenon (Anderson, 1991; Lowenstern and others, 1991; Tait, 1992). Anderson (1991) and Tait (1992) present detailed discussions of leakage mechanisms in phenocrysts, and Anderson and Brown (1993) discuss the possible significance of coexisting bubble-free and bubble-bearing glass inclusions.

In summary, there is petrographic and Raman spectroscopic evidence that a separate fluid phase coexisted with the dacitic melt early in its crystallization history. In addition, there is good petrographic evidence for fluid saturation during quartz crystallization, and the Raman data support this interpretation. In quartz, we infer that the bimodal distribution in the size of bubbles in glass inclusions reflects the entrapment of magmatic fluid together with melt, followed by the exsolution of dissolved volatiles from the melt during rise and depressurization (large and small bubbles, respectively). For plagioclase, the petrographic evidence for fluid saturation is not so compelling as for quartz, but the water contents of the glass inclusions (Gerlach and others, this volume; Pallister and others, this volume; Rutherford and Devine, this volume) indicate that the magma already was saturated with an aqueous fluid during the precipitation of plagioclase.

Obviously, the loss of fluid from the bubbles in glass makes it difficult to prove that magmatic volatiles were trapped together with melt. However, the experimental results and photographic documentation of Lowenstern (1994) on hydrous peralkaline melts support our interpretation of coentrapment of volatiles and melt. In his review paper on evidence from fluid inclusions of immiscibility in magmatic systems, Roedder (1992) cites several studies in which liquid water was documented in the "shrinkage bubbles" within glass inclusions; such high-density fluid indicates entrapment at elevated pressure and depth.

FLUID EVOLUTION

One of the most obvious, and yet critical, observations about the dacitic magmatic system at Mount Pinatubo is that fluid saturation was reached with respect to an aqueous fluid. Despite the attention that has been focused on the number of million tons of SO₂ that were released into the atmosphere, volatile saturation may have been initiated by CO₂ and ultimately became dominated by water (Gerlach and others, this volume). Furthermore, Gerlach and others (this volume) estimate a vapor:melt distribution coefficient of almost 800 for sulfur. The strong preferential partitioning of sulfur into a supercritical aqueous fluid phase and the vast quantities of magmatic water released into the atmosphere imply that the timing of water saturation and the evolution of this aqueous fluid had a large impact on the sulfur budget. We infer that it was this dense, sulfur-and CO₂-bearing aqueous fluid that was so poorly retained in the bubbles within glass inclusions in the early phenocrysts such as plagioclase and quartz.

Those fluid inclusions that are preserved in the phenocrysts, however, provide a means of tracking the evolution of this aqueous fluid during the course of crystallization and rise of the magma. For instance, the attainment of two-phase equilibria in the water-rich fluid of the Mount Pinatubo magmatic system is recorded by the liquid-bearing aqueous inclusions in quartz. Regardless of the mechanism that induced phase separation into a liquidlike and a vaporlike fluid, chloride, sulfate, and other dissolved nonvolatile components would have partitioned selectively into the liquid phase, whereas the gases (CO₂, SO₂) would have partitioned selectively into the vapor phase. This partitioning is reflected in the large volumetric proportion of solids in some of the liquid-bearing inclusions (fig. 1I).

One means of generating coexisting liquid-rich and vapor-rich inclusions is to decrease the pressure and temperature of an originally supercritical aqueous fluid, which could still be present even below the dacite solidus. Such a decrease, particularly in pressure, would bring the initially one-phase supercritical fluid into the two-phase field of liquid + vapor. The localization of liquid- and vapor-rich inclusions along fractures in some quartz grains indicates that those host crystals already had crystallized before two-phase separation occurred in the aqueous fluid coexisting with the magma. The low density of the aqueous inclusions, as indicated by their large volumetric proportion of vapor, does not necessarily mean that the quartz hosts were near the surface during trapping or that they have leaked (no physical evidence of leakage in most cases). The low pressure required for the separation of an initial supercritical fluid into vapor and liquid could have been induced by decompression associated with the ascent and eruption of the magma. In contrast, those quartz phenocrysts that are totally clouded by one-phase vapor or very vapor-rich inclusions (fig. 1J) probably crystallized during the stage of vapor effervescence, when they effectively trapped the low-density vapor phase. Some of these inclusions have obvious fractures, as mentioned above.

Another explanation for the high-salinity aqueous inclusions is that they represent hydrous saline melt, documenting high-temperature immiscibility that arose in the course of differentiation. Such immiscibility is common in porphyry copper systems (Roedder, 1992). Among the necessary criteria for saline melt inclusions are very high temperatures of homogenization (greater than 500°C) and the coexistence of silicate melt inclusions (Roedder, 1992; Lowenstern, 1994). The inclusions that we studied have not been subjected to microthermometric analysis. Although there are melt inclusions in the same quartz grains that contain abundant aqueous inclusions, the spatial and genetic relations between the two types of inclusions cannot be ascertained. In some quartz grains, the rare round, low-relief, isotropic inclusions (no Raman signature) and other optically complex inclusions containing bleblike phases may represent the entrapment of a saline melt. We cannot rule out the possibility that some of the quartz grains that were investigated represent xenocrysts derived from the wallrocks. However, such contamination is believed to be minimal, given that the volcano already had "cleared its throat" during earlier eruptions in mid-June.

One-phase vapor inclusions in plagioclase and quartz appear to have monitored two distinct processes. As discussed above, some appear to represent the vapor member of an aqueous brine that has segregated into liquid + vapor (fig. 1I). Others probably represent the capture of volatiles that leaked from glass inclusions and were subsequently trapped along fractures (fig. 1D). One-phase vapor inclusions in plagioclase phenocrysts are restricted to the latter type, whereas one-phase inclusions in quartz are mostly of the former type.

AQUEOUS FLUID VERSUS ANHYDRITE SATURATION: A DIFFERENT MODEL FOR THE BEHAVIOR OF SULFUR IN MELTS

In the experiments of Carroll and Rutherford (1987, 1988) and Luhr (1990) on the solubility of sulfur in oxidized calc-alkaline melts, the silicate melts are saturated with respect to both anhydrite and an aqueous fluid phase. The typical application of these experiments to anhydrite-bearing calc-alkaline systems involves the assumption that the magmas first reached saturation with respect to anhydrite and subsequently with respect to water (Carroll and Rutherford, 1987; Luhr, 1990). In the El Chichón eruptive rocks, that saturation sequence is evidenced by the abundance of anhydrite phenocrysts and, especially, anhydrite inclusions in silicate phenocrysts (Luhr and others, 1984). However, an unexpected lack of anhydrite has been noted in evolved volcanic rocks from various localities whose magmas were sufficiently oxidized to have stabilized a sulfate phase. Carroll and Rutherford (1987) have suggested that the latter magmas may have had bulk-sulfur concentrations that were too low to support much precipitation of anhydrite.

The petrographic and Raman evidence in the Mount Pinatubo samples studied by our group and others suggests the consideration of another model, namely, that H₂O-CO₂-SO₂ fluid saturation was reached before anhydrite saturation. This model would explain several otherwise anomalous observations that have been made on the Mount Pinatubo dacites, some of them already highlighted in the literature.

1. Late saturation with anhydrite would explain the dearth of anhydrite inclusions in early phenocryst phases such as plagioclase.
2. Late precipitation of anhydrite also would explain the high concentrations of sulfur detected in many apatite grains from Mount Pinatubo. Pallister and others (1993) report SO₃ concentrations of up to 1.1 wt% in unspecified apatites from the dacite. It seems particularly significant that early apatite grains, trapped within other phases, consistently have higher SO₃ contents (0.39 wt%) than do discrete apatites (0.13 wt%; Imai and others, this volume). As soon as anhydrite had reached fluid saturation, sulfur would have partitioned preferentially into that phase over apatite.
3. The Mount Pinatubo magmas are estimated to have equilibrated at several log f_O2 units higher than those at El Chichón (Carroll and Rutherford, 1987; Luhr, 1990; Westrich and Gerlach, 1992; Imai and others, 1993; Gerlach and others, this volume; Pallister and others, this volume; Rutherford and Devine, this volume; Imai and others, this volume), and sulfur solubility in silicate melt has been shown to increase with f_O2, although not greatly at the pertinent temperature of 800°C (Carroll and Rutherford, 1987, 1988; Luhr, 1990). This high oxidation state makes it reasonable to assume that the Mount Pinatubo magma could have accommodated at least as much, and perhaps somewhat more, dissolved sulfur as the El Chichón magma. It therefore is possible that the Mount Pinatubo magma could have reached anhydrite saturation at a later fractionation stage than the El Chichón magma. There is no conclusive evidence, however, about the actual original sulfur contents of the two magmatic systems.
4. McKibben and others (this volume) have interpreted the lack of sulfur isotopic zoning in anhydrite crystals from Mount Pinatubo (June 12th and 15th eruptions: McKibben and others, this volume; McKibben and Eldridge, 1993) as evidence that extensive SO₂ degassing, if it occurred, must have occurred before much anhydrite had precipitated.
5. The reported 0.14 wt% bulk-sulfur content in phenocryst-rich dacite from Mount Pinatubo (Pallister and others, this volume) compared to the 0.6 wt% sulfur in the El Chichón eruptive rocks (Luhr and others, 1984; Carroll and Rutherford, 1987; Bernard and others, 1991) also supports a model in which much of the initial sulfur content of the Mount Pinatubo magmatic system was extracted early.
6. The relatively constant sulfur concentration among glass inclusions within a variety of phenocrysts is interpreted by Rutherford and Devine (this volume) as evidence of anhydrite saturation of the magma. However, the sulfur partitioning imposed on the system by the coexistence of a fluid phase also could have buffered the sulfur concentration in the melt.

The above hypothesis of early saturation of the magma with respect to a water-dominated fluid would have important petrologic consequences for water-rich, highly oxidized, calc-alkaline magmas: Magmatic systems that had high concentrations of volatiles early in their evolution, due to their initial melt chemistry or to external introduction of volatiles, could segregate a separate fluid phase before much crystallization had occurred. The specific composition of the fluid phase would reflect both the relative concentrations and solubilities of volatiles in the magma; in most magmas, the fluid phase rapidly would become dominated by water. As soon as a separate fluid developed, sulfur species would be among those that were selectively extracted into it from the magma. One possible consequence of the temporal segregation of sulfur into an early fluid phase (subsequently released into the atmosphere) and later anhydrite and matrix glass phases (retained in the rock) is that the abundance, size, and timing of precipitation of anhydrite crystals in recent or historic hydrous calc-alkaline volcanics are not necessarily good indicators of the sulfur or sulfate concentration in the original melt.

A corollary to the above reasoning is that the ratio of other volatiles (for example, H₂O and CO₂) to SO₂ in the early melt of a given magmatic system could control the timing of sulfur removal from the magma via fluid exsolution, as well as the proportion of the available sulfur that reaches the atmosphere (compared to the proportion retained in the rocks as quenched glass and anhydrite). Due to their abundance and solubility, certain volatile species would induce fluid segregation from a magma. Magmas that reached saturation early with respect to water or a more chemically complex fluid would be stripped of a greater proportion of their total sulfur and thereby retain a lower bulk-rock sulfur content in their extrusive products than would magmas that reached saturation with respect to anhydrite earlier than to a fluid.

In this regard, it is important to consider the role of CO₂ in the initiation of fluid immiscibility in the Mount Pinatubo magmatic system. Gerlach and others (this volume) estimate that up to 10% of the bulk volatile fluid was CO₂, and CO₂ is much less soluble in siliceous melts than is H₂O. Thus, the initiation of fluid saturation could have been induced by CO₂. In bubbles within glass inclusions hosted by quartz, the Raman detection of CO₂, together with the lack of Raman detection or optical observation of H₂O, suggests that CO₂ was a major component of the early magmatic fluid phase at Mount Pinatubo. Finally, if crustally assimilated sulfur was an important component in this magma, then the timing of sulfur assimilation was important because of its effect on the (H₂O+CO₂)/SO₂ ratio. Our model of early saturation with respect to a fluid phase, specifically before anhydrite saturation, is consistent with the calculations of Gerlach and others (this volume) indicating that the dacitic magma reached fluid saturation before eruption.

Although our model explains many features of the Mount Pinatubo phenocryst-rich dacites, it is not applicable to all oxidized, calc-alkaline systems. For instance, this model seems inappropriate for the El Chichón eruptive rocks. A difference in the relative timing of saturation with respect to water and anhydrite might account for two petrographic differences between the samples we studied from those two eruptions. Our Mount Pinatubo thin sections had less anhydrite and a higher volumetric proportion of bubbles in the glass inclusions (possibly an effect of cooling history) than did the two samples from El Chichón (obtained from J.F. Luhr, Smithsonian Institution). Our conclusion is that the relative, as well as absolute, concentrations in a melt of such volatile species as H₂O, CO₂, and SO₂ will control the timing of saturation with respect to a fluid phase, anhydrite, and other volatile-containing phases. Further experimental studies of oxidized, calc-alkaline systems may reveal how small differences in initial composition can strongly control their eventual fractionation paths and their products of crystallization.

SUMMARY AND CONCLUSIONS

Detailed petrographic and Raman spectroscopic analyses of inclusions suggest that much of the plagioclase crystallized before quartz, that anhydrite was a very late phenocryst phase, and that the dacitic magma reached aqueous fluid saturation relatively early in its crystallization. In glass inclusions in quartz, the high bubble-to-glass ratio and, particularly, the Raman-determined gas contents and pressures in some bubbles suggest that these inclusions represent the simultaneous trapping of a melt and a supercritical fluid phase at depth. This trapping was followed rapidly by exsolution of volatiles during the rise and depressurization of the magma; thus, some glass inclusions in quartz show a bimodal distribution of bubble sizes. It is unfortunate, but typical, that quartz and its glass inclusions were unreliable containers for fluid, as evidenced by petrographic observation of inclusion rupture and by Raman documentation of low to below-detectable gas contents in the bubbles in glass. It therefore is not possible to use the measured gas contents and pressures to infer the pressure-temperature conditions of initial fluid trapping. For inclusions in plagioclase, the physical evidence of fluid trapping is not so compelling as for quartz, but the high water contents of the glass inclusions (Gerlach and others, this volume; Pallister and others, this volume; Rutherford and Devine, this volume) are in accord with water saturation of the magma at the time of plagioclase precipitation.

In our model, the evolving supercritical, CO₂- and sulfur-bearing aqueous fluid that coexisted with the melt subsequently underwent two-phase separation into liquid and vapor due to an increase in fluid salinity resulting from fractionation (not probable), the rise and depressurization of the magma (more probable), or both. This episode in the development of the fluid is recorded in the late-stage coexistence of all-vapor and liquid-vapor fluid inclusions in quartz. The continued high concentration of chlorine and sulfur in the fluid phase is documented by the abundance of chloride and sulfate daughter minerals in the liquid-rich aqueous inclusions.

Our Raman spectroscopic identifications and petrographic observations, coupled with the observations of other petrologists, indicate a very low abundance of anhydrite inclusions within phenocrysts; however, the latter contain abundant inclusions of hornblende, plagioclase, and especially apatite and zircon. These observations led us to suggest that anhydrite might have been among the last phenocryst phases to reach saturation. Raman evidence of the entrapment of CO₂-rich fluid in phenocrysts, especially quartz, suggests consideration of the possibility that saturation with a CO₂-bearing aqueous fluid was reached early in the evolution of the dacitic magma at Mount Pinatubo (before the eruption began) and that partitioning of sulfur into this fluid delayed the precipitation of anhydrite. Other mineral-chemical and sulfur isotopic data lend support to this hypothesis. We further postulate that the early release of an H₂O-CO₂-SO₂ fluid and its effective extraction of SO₂ from the coexisting melt in part might account for the large volume of SO₂ released into the atmosphere during the June 15th eruption. The importance of the timing of volatile saturation, as a means of extracting specific elements, has been recognized in porphyry copper systems (for example, Cline and Bodnar, 1991; Lowenstern, 1993). Furthermore, as Imai and others (1993) discuss, there are several important links between the Mount Pinatubo volcano and nearby porphyry copper deposits in the Philippines.

The observations and interpretations presented above could be explained better and tested if the results of several additional types of studies were available: (1) experimental monitoring of oxidized, sulfur-rich melts that are saturated with respect to an aqueous fluid but not anhydrite could provide important information on sulfur partitioning, especially if the volatile species in equilibrium with the melts were analyzed; (2) as suggested by the work of Gerlach and others (this volume), the partitioning of sulfur gas species should be investigated in systems consisting of a silicate melt and an aqueous fluid, in which the CO₂/H₂O ratio in the fluid is varied; and (3) the Raman spectral sensitivity of inclusion phases to minor compositional differences should be investigated further.

ACKNOWLEDGMENTS

This research was supported by National Science Foundation grant GER-9023520 to J.D. Pasteris. The authors thank Terrence Gerlach, Henry Westrich, James Luhr, Akira Imai, and Paul Pohwat for providing samples for this study and Karen Bartels for electron microprobe analysis. The authors also benefited from discussions with Alfred Anderson, Charles Bacon, Harvey Belkin, John Fournelle, Terrence Gerlach, Michael McKibben, James Luhr, John Pallister, Paul Wallace, and Henry Westrich. Reviews on an earlier version of this manuscript by Terrence Gerlach, Jacob Lowenstern, James Luhr, Gwendolyn Miner, Christopher Newhall, and Malcolm Rutherford are gratefully acknowledged. The authors retain full responsibility, however, for all data and interpretations.

REFERENCES CITED

Anderson, A.T., 1991, Hourglass inclusions: Theory and application to the Bishop rhyolitic tuff: American Mineralogist, v. 76, p. 530-547.

Anderson, A.T., and Brown, G.G., 1993, CO₂ contents and formation pressures of some Kilauean melt inclusions: American Mineralogist, v. 78, p. 794-803.

Bacon, C.R., 1989, Crystallization of accessory phases in magmas by local saturation adjacent to phenocrysts: Geochimica et Cosmochimica Acta, v. 53, p. 1055-1066.

Bernard, A., Demaiffe, D., Mattielli, and Punongbayan, R.S., 1991, Anhydrite-bearing pumices from Mount Pinatubo: Further evidence for the existence of sulfur-rich silicic magmas: Nature, v. 354, p. 139-140.

Bluth, G.J.S., Doiron, S.D., Schnetzler, C.C., Krueger, A.J., and Walter, L.S., 1992, Global tracking of the SO₂ clouds from the June, 1991 Mount Pinatubo eruptions: Geophysical Research Letters, v. 19, no. 2, p. 151-154.

Carroll, M.R., and Rutherford, M.J., 1987, The stability of igneous anhydrite: Experimental results and implications for sulfur behavior in the 1982 El Chichón trachyandesite and other evolved magmas: Journal of Petrology, v. 28, pt. 5, p. 781-801.

------1988, Sulfur speciation in hydrous experimental glasses of varying oxidation state: Results from measured wavelength shifts of sulfur X-rays: American Mineralogist, v. 73, p. 845-849.

Cline, J.S., and Bodnar, R.J., 1991, Can economic porphyry copper mineralization be generated by a typical calc-alkaline melt?: Journal of Geophysical Research, v. 96, p. 8113-8126.

Fogel, R.A., and Rutherford, M.J., 1990, The solubility of carbon dioxide in rhyolitic melts: A quantitative FTIR study: American Mineralogist, v. 75, p. 1311-1326.

Fournelle, J, Carmody, R., and Daag, A.S., this volume, Anhydrite-bearing pumices from the June 15, 1991, eruption of Mount Pinatubo: Geochemistry, mineralogy, and petrology.

Gardner, P.M., 1972, Hollow apatites in a layered basic intrusion, Norway: Geological Magazine, v. 109, p. 385-391.

Gerlach, T.M., Westrich, H.R., and Symonds, R.B., this volume, Preeruption vapor in magma of the climactic Mount Pinatubo eruption: Source of the giant stratospheric sulfur dioxide cloud.

Imai, A., Listanco, E.L., and Fujii, T., 1993, Petrologic and sulfur isotopic significance of highly oxidized and sulfur-rich magma of Mt. Pinatubo, Philippines: Geology, v. 21, p. 699-702.

------this volume, Highly oxidized and sulfur-rich dacitic magma of Mount Pinatubo: Implication for metallogenesis of porphyry copper mineralization in the Western Luzon arc.

Knittle, E., and Williams, Q., 1993, High-pressure Raman spectroscopy of ZrSiO₄: Observation of the zircon to scheelite transition at 300 K: American Mineralogist, v. 78, p. 245-252.

Lowenstern, J.B., 1993, Evidence for a copper-bearing fluid in magma erupted at the Valley of Ten Thousand Smokes, Alaska: Contributions to Mineralogy and Petrology, v. 114, p. 409-421.

------1994, Chlorine, fluid immiscibility and degassing in peralkaline magmas from Pantelleria, Italy: American Mineralogist, v. 79, p. 353-369.

Lowenstern, J.B., Mahood, G.A., Rivers, M.L., and Sutton, S.R., 1991, Evidence for extreme partitioning of copper into a magmatic vapor phase: Science, v. 252, p. 1405-1409.

Luhr, J.F., 1990, Experimental phase relations of water- and sulfur-saturated arc magmas and the 1982 eruptions of El Chichon volcano: Journal of Petrology, v. 31, pt. 5, p. 1071-1114.

Luhr, J.F., Carmichael, I.S.E., and Varekamp, J.C., 1984, The 1982 eruptions of El Chichón volcano, Chiapas, Mexico: Mineralogy and petrology of the anhydrite-bearing pumices: Journal of Volcanology and Geothermal Research, v. 23, p. 69-108.

Matthews, S.J., Jones, A.P., and Bristow, C.S., 1992, A simple magma-mixing model for sulfur behaviour in calc-alkaline volcanic rocks: Mineralogical evidence from Mount Pinatubo 1991 eruption: Journal of the Geological Society, v. 149, p. 863-866.

McKibben, M.A., and Eldridge, C.S., 1993, Sulfur isotopic systematics of the June 1991 eruptions of Mount Pinatubo: A SHRIMP ion microprobe study [abs.]: Eos, Transactions, American Geophysical Union, v. 74, no. 43, p. 668.

McKibben, M.A., Eldridge, C.S., and Reyes, A.G., this volume, Sulfur isotopic systematics of the June 1991 Mount Pinatubo eruptions: A SHRIMP ion microprobe study.

Michael, P.J., 1988, Partition coefficients for rare earth elements in mafic minerals of high silica rhyolites: The importance of accessory mineral inclusions: Geochimica et Cosmochimica Acta, v. 52, p. 275-282.

Pallister, J.S., Hoblitt, R.P., Meeker, G.P., Knight, R.J., and Siems, D.F., this volume, Magma mixing at Mount Pinatubo: Petrographic and chemical evidence from the 1991 deposits.

Pallister, J.S., Hoblitt, R.P., and Reyes, A.G., 1992, A basalt trigger for the 1991 eruptions of Pinatubo volcano?: Nature, v. 356, p. 426-428.

Pallister, J.S., Meeker, G.P., Newhall, C.G., Hoblitt, R.P., and Martinez, M., 1993, 30,000 years of the "same old stuff" at Pinatubo [abs.]: Eos, Transactions, American Geophysical Union, v. 74, no. 43, p. 667.

Qin, Z., Lu, F., Anderson, A.T., Jr., 1992, Diffusive reequilibration of melt and fluid inclusions: American Mineralogist, v. 77, p. 565-576.

Roedder, Edwin, 1984, Fluid Inclusions, in Ribbe, P.H., ed., Reviews in mineralogy, v. 12: Washington, D.C., Mineralogical Society of America, 644 p.

------1992, Fluid inclusion evidence for immiscibility in magmatic differentiation: Geochimica et Cosmochimica Acta, v. 56, p. 5-20.

Rutherford, M.J., and Devine, J.D., this volume, Preeruption pressure-temperature conditions and volatiles in the 1991 dacitic magma of Mount Pinatubo.

Scott, W.E., Hoblitt, R.P., Torres, R.C., Self, S, Martinez, M.L., and Nillos, T., Jr., this volume, Pyroclastic flows of the June 15, 1991, climactic eruption of Mount Pinatubo.

Tait, S., 1992, Selective preservation of melt inclusions in igneous phenocrysts: American Mineralogist, v. 77, p. 146-155.

Wallace, P., 1993, Pre-eruptive gas saturation in the June 15, 1991, Mount Pinatubo dacite: New evidence from CO₂ contents of melt inclusions [abs.]: Eos, Transactions, American Geophysical Union, v. 74, no. 43, p. 668.

Westrich, H.R., and Gerlach, T.M., 1992, Magmatic gas source for the stratospheric SO₂ cloud from the June 15, 1991, eruption of Mount Pinatubo: Geology, v. 20, p. 876-870.

Wopenka, B., Pasteris, J.D., and Freeman, J.J., 1990, Analysis of individual fluid inclusions by Fourier transform and Raman microspectroscopy: Geochimica et Cosmochimica Acta, v. 54, p. 519-533.

Wyllie, P.J., Cox, K.G., and Biggar, G.M., 1962, The habit of apatite in synthetic systems and igneous rocks: Journal of Petrology, v. 3, p. 238-243.

FIRE and MUD Contents

PHIVOLCS | University of Washington Press | U.S.Geological Survey

This page is <https://pubs.usgs.gov/pinatubo/pasteris/>
Contact: Chris Newhall
Last updated 06.11.99