• Find a Volcano Adagdak Akutan Amak Amukta Aniakchak Augustine Behm Canal-Rudyerd Bay Black Peak Bobrof Bogoslof Buldir Buzzard Creek Carlisle Chagulak Chiginagak Churchill Cleveland Dana Davidof Denison Douglas Duncan Canal Dutton Edgecumbe Emmons Lake Fisher Fourpeaked Frosty Gareloi Gordon Great Sitkin Griggs Hayes Herbert Iliamna Imuruk Lake Ingakslugwat Hills Isanotski Kagamil Kaguyak Kanaga Kasatochi Katmai Kialagvik Kiska Koniuji Kookooligit Mountains Korovin Kukak Kupreanof Little Sitkin Mageik Makushin Martin Moffett Novarupta Nunivak Island Okmok Pavlof Pavlof Sister Recheschnoi Redoubt Roundtop Sanford Seguam Segula Semisopochnoi Sergief Shishaldin Snowy Mountain Spurr St. Michael St. Paul Island Steller Stepovak Bay 2 Table Top-Wide Bay Takawangha Tanaga Tlevak Strait-Suemez Is. Trident Ugashik-Peulik Ukinrek Maars Uliaga Unnamed Veniaminof Vsevidof Westdahl Wrangell Yantarni Yunaska San Francisco Volcanic Field Uinkaret Volcanic Field Brushy Butte Clear Lake Volcanic Field Coso Volcanic Field Eagle Lake Volcanic Field Golden Trout Creek Volcanic Field Lassen Volcanic Center Lavic Lake Volcanic Field Long Valley Caldera Mammoth Mountain Medicine Lake Mono Lake Volcanic Field Mono-Inyo Chain Mount Shasta Salton Buttes Silver Lake Volcanic Field Tumble Buttes Twin Buttes Ubehebe Craters Dotsero Volcanic Center Haleakala Hualalai Kilauea Loihi Mauna Kea Mauna Loa Craters of the Moon Volcanic Field Hell's Half Acre Volcanic Field Shoshone Lava Field Wapi Lava Field Soda Lakes Carrizozo Lava Flow Valles Caldera Zuni-Bandera Volcanic Field Agrigan Ahyi Alamagan Anatahan Asuncion Esmeralda Bank Farallon de Pajaros Guguan Maug Islands Pagan Ruby Sarigan Supply Reef Belknap Blue Lake Crater Cinnamon Butte Crater Lake Davis Lake Volcanic Field Devils Garden Volcanic Field Diamond Craters Volcanic Field Four Craters Lava Field Jordan Craters Volcanic Field Lava Mountain Mount Bachelor Mount Hood Mount Jefferson Newberry North Sister Field Sand Mountain Volcanic Field South Sister Bald Knoll Black Rock Desert Volcanic Field Markagunt Plateau Volcanic Field Santa Clara Volcanic Field Glacier Peak Indian Heaven Volcanic Field Mount Adams Mount Baker Mount Rainier Mount St. Helens West Crater Volcanic Field Find a Wyoming Volcano Yellowstone
• About
• Observatories
• Activity
• Education
• Publications

Changes in Melt Inclusions after Entrapment

• Volumetric Behavior
• Other Origins of Bubbles
• Decrepitation of MI
• Crystallization
• Diffusional Re-equilibration

VOLUMETRIC BEHAVIOR

Melt inclusions can be understood within the same theoretical and experimental framework that has been developed for other types of fluid inclusions (Hollister & Crawford 1982; Roedder 1984). Relative to silicate melt, quartz and most other magmatic phenocrysts are incompressible, so that after entrapment the inclusion volume is nearly constant (Roedder 1984). This means that the MI approximates an isochoric (constant volume) system with dP/dT equal to the ratio of the coefficient of thermal expansion of the melt (a) to its compressibility (b) [~0.5 to 1.5 MPa/°C for typical silicic melt]. Because silicate melts are relatively incompressible, their isochores have relatively steep slopes, and the internal pressure of inclusions will change rapidly with relatively small changes in temperature (Fig 4). As long as the quartz host is strong enough to withstand the pressure differential between the inclusion and external environment, the pressure in the MI will be a function of its temperature.

 

FIGURE 4. Schematic pressure-temperature trajectory for a silicate melt inclusion in quartz. The inclusion traps vapor-undersaturated silicate melt (point A) at temperature Tt, and cools along its isochore until it reaches B. At this point, under equilibrium conditions, the melt should become saturated with a vapor phase and nucleate a bubble (at Tb). The inclusions would then cool along the melt/vapor curve to C. If cooled quickly, however, the inclusion may become supersaturated with respect to the vapor and will continue cooling along the original one-phase (melt) isochore, until the bubble finally nucleates at Tb' (F). The internal pressure in the inclusion will then increase (F to G) until the inclusion returns to the melt/vapor curve. The inclusion follows this curve until Tb/a, when the internal pressure in the inclusion increases due to the 1% volumetric contraction of the host. At Tg, the melt passes through the glass transition, and the bubble ceases to expand. If the inclusion were reheated in the laboratory, it would follow the path, E-D-C-G-B-A. Th, the homogenization temperature, would occur at B. Figure adapted from Lowenstern (1994a).

As a single-phase MI cools below its temperature of entrapment (Tt in Fig 4), it will follow an isochore and depressurize until it becomes saturated with a volatile phase (i.e., a 'shrinkage' bubble nucleates at Tb: Sorby, 1858). Once the compressible bubble has formed, the inclusion will cool along the Melt/Vapor Curve until the supercooled silicate melt passes through the glass transition at Tg. Additionally, as the inclusion cools to room temperature, the internal pressure in the bubble may change as gases condense to their liquid state, and follow their own liquid/vapor curve. At 25 °C, a shrinkage bubble composed of pure H2O should have an internal pressure equivalent to the vapor pressure of H2O (3.2 kPa). A bubble in an inclusion of volatile-free melt should contain only silicate-melt vapor and thus should be a near vacuum. Any bubble with a relatively high concentration of noncondensable gases (e.g., CO2) will retain higher internal pressure at 25°C (up to ~6 MPa, the pressure at which liquid CO2 forms). Water condensing from the cooling shrinkage bubble may react with the inclusion glass, causing hydration at the bubble/glass interface (Anderson 1991).

Fig. 5 shows the evolution of a crystallized melt inclusion as it is heated from room temperature up to 850 °C and then cooled. During heating, the inclusion progressively melts; the vapor bubble contracts until it disappears at Th (850°C). During cooling, a shrinkage bubble forms at Tb' (710 °C) corresponding to point F in Fig. 4. As in most systems, Tb' is less than Th because the inclusion must be significantly underpressured before it reaches saturation with the vapor bubble.

FIG. 5. Transmitted-light photographs of melt inclusion in quartz from Sciuvechi lava at Pantelleria, during heating-stage experiment. (A) During heating, the inclusion remained microcrystalline until temperatures above 700 °C, when (B) melting began around the inclusion periphery. (C) At 800 °C, the crystals had completely melted and all that remained was silicate melt (m) refractory quartz (q) and vapor bubbles (v). (D) At 850 °C, the inclusion had reached Th, at which point the bubble had homogenized into the silicate melt. (E) During cooling of the inclusion, a vapor bubble nucleated at Tb', which, in this example, was ~140 °C below Th. Terminology as in Figure 4. Photographs reprinted from Lowenstern (1994a).

OTHER ORIGINS OF BUBBLES

In principle, MI with similar origins should have similar vapor/melt ratios (see Anderson & Brown 1993). Typically, a single shrinkage bubble will constitute 0.2 to 5 vol% of an inclusion, with the actual value dependent on cooling rate (Lowenstern 1994a), volatile content and melt composition. When nearby MI have widely different vapor/glass ratios, other processes may have caused the bubbles to form.

Rapid decompression and resultant overpressures within the inclusion, or even thermal shock can result in fracturing of a phenocryst, depressurization, and vesiculation. In some circumstances, the fractures that cause this vesiculation may not be evident, as will commonly be the case in polished sections in which part or all of the fracture has been removed during sample preparation. For this reason, workers should consider avoiding inclusions with unusually large vapor/melt ratios, which may have lost volatiles during fracturing events. Wallace & Gerlach (1994) and Pasteris et al. (accepted) concluded that most bubble-rich inclusions from the Pinatubo eruptive products had partly leaked.

Another cause for large vapor/melt ratios, however, is simultaneous entrapment of melt and a separate vapor or fluid phase (Roedder 1965, Belkin et al. 1985, Lowenstern et al. 1991, Frezzotti 1992, De Vivo & Frezzotti 1994). Such inclusions are particularly difficult to distinguish from leaked, bubble-rich inclusions. Belkin et al. (1985) differentiated shrinkage bubbles from those primary vapors trapped as a separate phase by recording their behavior in a crushing stage. Shrinkage bubbles collapsed whereas the primary CO2-bearing bubbles expanded when their host inclusions were opened to atmosphere. Lowenstern et al. (1991) and Lowenstern (1993) used the X-ray microprobe and infrared spectroscopy to analyze the compositions of bubbles within MI. The anomalous Cu and CO2 content of some bubbles was shown to be inconsistent with bubble formation by simple cooling/depressurization of an originally homogenous melt (or vesiculation during leakage). These results are discussed in greater detail below (section on Evidence for Fluid Saturation and Degassing). Another criterion for differentiating primary bubbles from those due to shrinkage is an unusually high homogenization temperature for the bubble (Th) since the inclusion must be taken to pressures greater than that of entrapment to dissolve the extra increment of vapor (Lowenstern, 1994a). However, leaked inclusions also exhibit high Th, and it is most prudent to assume that bubble-rich inclusion have formed due to leakage, unless there is compelling petrographic or chemical evidence to the contrary.

DECREPITATION OF MI

Figure 6 shows a group of vapor-rich fluid inclusions trapped along a fracture that formed by decrepitation of a silicate MI. Fractured and leaked MI are common in volcanic rocks. Some MI may decompress and decrepitate prior to cooling, during unloading of a magma chamber (Bacon et al. 1992; Tait 1992). During near-isothermal decompression, the pressure within an MI remains close to that at entrapment. This means that a large pressure gradient will exist between the inclusion and the crystal exterior, a situation that commonly results in cracking of the inclusion (Tait 1992). If cooling tends to be rapid, however, one might expect a smaller pressure differential between MI and external environment, as the inclusion can decompress along its isochore (Fig. 4; point A to B). In fact, it appears that rapidly erupted rocks (i.e., Plinian deposits) have intact inclusions when compared to more slowly cooled lavas and pyroclastic flow deposits (Skirius et al. 1990; Dunbar & Hervig 1992a,b; Bacon et al. 1992). This correlates with the observation that MI in Plinian tephra tend to lack bubbles, whereas bubbles are common in coerupted pyroclastic flows, which cool more slowly (Clocchiatti 1975; Skirius et al. 1990; Dunbar and Hervig 1992a; Lowenstern 1993). Because most MI become significantly volatile-supersaturated before nucleation of the shrinkage bubble occurs (Figs. 4 and 5; Lowenstern 1994a), inclusions may be underpressured by up to tens of MPa. As shown in Fig. 4 (G to F), nucleation of a shrinkage bubble should cause a sharp increase in the internal pressure of an inclusion, and possibly cause cracking. MI in Plinian tephras may be quenched with sufficient speed that the bubbles are kinetically hindered from forming (Fig. 7; Lowenstern 1994a). In contrast, the lower cooling rates of lavas and pyroclastic flows allow bubble nucleation, resulting in an increase in the internal pressure of the inclusion and decrepitation.

CRYSTALLIZATION

If allowed to cool slowly, MI will crystallize. Crystallized MI are typically described as 'devitrified', though it is not always apparent that the crystals grew from a glass rather than a still-molten liquid (in which case, the inclusions are not technically devitrified). Commonly, crystallization begins with precipitation of host mineral on the inclusion wall. Under such circumstances, the original melt can be calculated from the compositions of MI hosted by different phenocrysts formed from the same magma batch (Watson 1976; chapter 16 of Roedder 1984). Though many workers have concluded that MI compositions change little subsequent to entrapment (Beddoe-Stephens et al. 1983; Dunbar & Hervig 1992a, b; Bacon et al. 1992), other studies find evidence for limited post-entrapment crystallization of MI. Sisson & Layne (1993) reported that olivine-hosted MI from Fuego volcano were affected by post-entrapment crystallization (representing up to 11% of the inclusion mass). Webster & Duffield (1991) estimated that up to 15% of the mass of individual MI was precipitated onto the walls of their host quartz, depleting the glass in SiO2. Presumably, such heterogeneous precipitation should not significantly exceed this value of 15% before the trapped melt becomes saturated with respect to the host phase and crystallization stops. If other crystalline phases begin to form, crystallization of the entire melt can occur.

As with bubble size, the degree of crystallization appears to correlate to cooling rate (Fig. 7). Skirius et al. (1990) found that inclusions from slowly-cooled pyroclastic flows were commonly crystallized whereas those from more rapidly emplaced Plinian units remained glassy. Crystallized MI were remelted in the laboratory and quenched to form glassy inclusions that could be studied by microbeam methods (Skirius 1990; Skirius et al. 1990). In some systems, crystallized MI may be more pristine than glassy MI. Lowenstern & Mahood (1991) analyzed remelted pantellerite MI (e.g., Fig. 1c and 1d and Fig. 5) for their H2O concentrations and found that they were hydrous, whereas all originally glassy inclusions had low H2O contents and seemed to have leaked. Presumably, the higher H2O concentrations in intact inclusions increased ionic diffusivities and allowed crystallization, whereas the glassy inclusions, due to loss of water, were quenched as glass.

When kept at high temperatures, MI may change shape, regardless of whether they crystallize (Beddoe-Stephens et al. 1983). Evidently, the high temperatures permit redistribution of the host mineral, allowing surface energetics to control inclusion morphology. Skirius et al. (1990) noted that during prolonged heating experiments, originally rounded MI became more faceted and the walls took on the shapes of negative crystals. Similar observations have been reported by Lowenstern (1994a) and Clocchiatti (1975) (for both naturally and experimentally heated volcanic rocks) and by Frezzotti (1992) for partly crystallized MI in granites.

DIFFUSIONAL RE-EQUILIBRATION

Fluid, mineral and melt inclusions may not stay as closed systems over long time periods. Roedder (1984) suggested that the compositions of fluid inclusions in granulites may be affected by diffusion of H2 and H2O through the host minerals. This process has been confirmed experimentally and noted in other metamorphic and igneous rocks (Pasteris & Wanamaker 1988; Scowen et al. 1991; Hall et al. 1991; Mavrogenes & Bodnar 1994). Re-equilibration appears to be enhanced by advection along microcracks and mineral defects, whose presence may be difficult to detect. These processes certainly will be important in slowly cooled granites, where melt inclusions, even if glassy, are kept at high temperatures for hundreds to thousands of years.

In molten systems, where thermal and compositional gradients are lower, the effect of diffusional reequilibration will be lessened, but still may occur. Qin et al. (1992) provided analytical solutions to determine the effect of bulk diffusion on the H2O contents of melt and fluid inclusions in quartz. They concluded that small inclusions (<25 µm) within a larger host quartz (1 mm diam.) can reach 95% reequilibration with the external environment in only two years. They suggested that bulk diffusion may have had an effect on the late-erupted Bishop tuff pyroclastic flows, which contain lower H2O concentrations than MI from Plinian units. However, experiments by Qin (1994) did not detect significant (and consistent) loss of H2O from silicate MI during week-long experiments at high pressure and temperature. No other studies have been performed to assess the importance of this process for other (slower diffusing) chemical species. In general, the chemical gradients between inclusion and external melt may be insufficient for diffusion to affect MI compositions in shallow subvolcanic magmas.

Recent work by Bogaard & Schirnick (1994) seems to indicate that diffusional loss of components is negligible, at least for argon. These workers were able to date the time of entrapment of individual MI in quartz from the 0.76 Ma Bishop tuff. The mean apparent age of MI from the basal fallout was 1.89 ± 0.03 Ma, and the group of analyzed inclusions defined an isochron of 1.93 ± 0.06 Ma. Inclusions from later-erupted pyroclastic flows gave an isochron of 2.3 ± 0.4 Ma. Apparently, the MI were trapped over a million years prior to venting of the system 760,000 years ago. The closure temperature for Ar diffusion through quartz is unknown, yet these data would seem to imply that it is greater than the magma temperature of the Bishop tuff (~700 to 770 °C). Plausibly, the inclusions and their host quartz phenocrysts could have cooled to lower temperature within a supercooled glass or granite that was remobilized prior to eruption (though it is difficult to imagine all 500 km³ of magma going through this process). Regardless of the mechanism, very little Ar apparently diffuses through quartz at near-magmatic temperatures.