Journal of Petrology Advance Access originally published online on August 18, 2007
Journal of Petrology 2007 48(10):1955-1972; doi:10.1093/petrology/egm045
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Equilibration Scales in Silicic to Intermediate Magmas—Implications for Experimental Studies
1Institut des Sciences de la Terre D’orléans (ISTO), UMR 6113 CNRS-UO, 1A Rue de la Férollerie, 45071, Orléans Cedex 02, France
2CSIC, Institut de Ciènces de la Terra ‘Jaume Almera’, Lluis Solé I Sabaris S/N, 08028 Barcelona, Spain
RECEIVED JANUARY 19, 2007; ACCEPTED JULY 10, 2007
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONSTATES OF EQUILIBRIUM IN...EQUILIBRATION SCALESTHE LOCAL EQUILIBRIUM MODELIMPLICATIONS FOR PHASE...DISCUSSIONCONCLUDING COMMENTSAPPENDIXREFERENCES |
|---|
Experimental phase equilibrium studies are increasingly being used for the determination of intensive variables (P, T, fH2O, fO2 ) in silicic to intermediate magmas. In contrast, silicic igneous bodies are now perceived as open, periodically recharged, systems involving only limited chemical equilibration. Thus, the use of laboratory-determined crystal–liquid equilibrium data needs clarification. Here we review the field, petrological and geochemical evidence concerning states and scales of chemical equilibrium in silicic magma bodies. It is concluded that total chemical equilibrium is generally not the rule. However, a subsystem in local equilibrium (the reactive magma) can be identified. Equilibration scales in silicic magmas are rate-limited either by diffusive flux in crystals (DICL regime) or by diffusive flux in the melt (MD regime). The recognition that equilibrium in magmas is limited to a reactive subsystem requires phase equilibrium studies to be chemically scaled. Experiments, either of total or partial equilibrium type, should aim at a close reproduction of equilibrium states specific to natural systems. The laboratory reconstruction of the natural equilibrium states guarantees a precise determination of the pre-eruptive parameters and a reliable application of the experimental data to active volcanic systems.
KEY WORDS: silicic magmas; chemical equilibrium; timescales; experimental studies
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONSTATES OF EQUILIBRIUM IN...EQUILIBRATION SCALESTHE LOCAL EQUILIBRIUM MODELIMPLICATIONS FOR PHASE...DISCUSSIONCONCLUDING COMMENTSAPPENDIXREFERENCES |
|---|
For almost a century, experimental phase equilibrium studies have provided a quantitative basis for understanding magmatic processes. Originally performed on simple synthetic systems (e.g. Mysen, 1989
), experimental phase equilibrium studies have shifted to chemically complex compositions containing volatiles and multivalent elements and, thus, have the potential to closely simulate the evolution of silicic to intermediate magmas (e.g. Eggler & Burnham, 1973; Maaloe & Wyllie, 1975; Clemens & Wall, 1981; Naney, 1983; Scaillet et al., 1995). Initially aimed at the formulation of general principles of silicate crystallization and magma differentiation (Bowen, 1913, 1914), phase equilibrium studies are now increasingly being used to determine intensive parameters (P, T, fH2O, fO2) in igneous systems (e.g. Rutherford et al., 1985; Martel et al., 1998; Scaillet & Evans, 1999; Costa et al., 2004). Knowledge of these pre-eruptive parameters is of important practical application for petrological monitoring of active volcanoes, physical modelling of eruptions (Dobran et al., 1994) and for computing compositions of magmatic gases released to the atmosphere (Scaillet & Pichavant, 2003).
The present-day development of phase equilibrium studies has benefited from continuous experimental and analytical improvements. For example, specific methods have been introduced to promote attainment of crystal–liquid equilibrium in hydrous, SiO2-rich magmas at relatively low temperatures (e.g. Pichavant, 1987; Holtz et al., 1992; Scaillet et al., 1995). Therefore, it has now become technically possible to chemically equilibrate a wide range of silicate compositions in the laboratory. However, igneous systems in nature are increasingly appreciated to be out-of-equilibrium and, therefore, the use of laboratory-determined crystal–liquid equilibrium data needs to be rigorously justified.
Experimental petrologists very early on recognized that total equilibrium is not the rule in natural magmas and that, theoretically, cooling rates and reaction kinetics should affect crystallization paths (e.g. Bowen, 1914; Young, 1998). Nevertheless, the application of phase equilibrium studies to igneous rocks has been generally justified (e.g. Schairer, 1957). Similarly, the use of mineral–mineral or mineral–melt geothermobarometers to retrieve intensive variables (e.g. Buddington & Lindsley, 1964; Carmichael et al., 1970) and of crystal–liquid partitioning to model magmatic processes (e.g. Gast, 1968; Allègre & Minster, 1978) both rely on the assumption that chemical equilibration occurs, at least locally, in igneous rocks. In contrast, various studies have stressed the importance of magma mixing and mingling, hence of disequilibrium, in igneous systems (e.g. Kuno, 1950; Anderson, 1976; Eichelberger, 1978). The view that igneous bodies are open, periodically recharged systems involving limited equilibration has progressively gained acceptance. More recently, the development of analytical techniques has extended to small scales the documentation of trace element and isotopic disequilibrium in igneous processes (e.g. Bea, 1996a, 1996b; Hammouda et al., 1996; Davidson & Tepley, 1997; Knesel et al., 1999). Therefore, there is an heightened appreciation that chemical equilibrium is not necessarily established in igneous systems.
The trend toward the application of phase equilibria to natural magmas, on one hand, and the recognition that magmas are not necessarily near-equilibrium systems, on the other, call for a methodological clarification. In this paper, we contend that the specifics of natural equilibrium states should be better taken into consideration, and that phase equilibrium experiments should aim at the reproduction of equilibrium states specific to magmatic systems. With this approach, the intensive parameters (P, T, fH2O, fO2) to be experimentally constrained can be viewed simply as the variables controlling natural pre-eruptive equilibration. Their reliable determination requires matching between experimental data (phase assemblages, mineral and glass compositions) and natural eruption products (phenocryst assemblages and phenocryst and glass compositions) or, in other words, matching between experimental and natural equilibrium states. To reach this goal, both a precise evaluation of natural equilibrium states and methods that allow these natural equilibrium states to be experimentally adjusted are necessary.
This paper consists of two complementary parts. In the first part, the field, petrological and geochemical evidence concerning states and scales of chemical equilibrium in silicic to intermediate (i.e. andesitic to rhyolitic) magmas is analysed. We argue that total equilibrium is generally not the rule. However, a subsystem in local equilibrium can be identified. A general approach to evaluate natural equilibration scales is outlined and illustrated by several natural examples. In the second part, the implications for experimental studies are considered. Methods allowing natural equilibrium states to be experimentally adjusted are presented and discussed.
|
STATES OF EQUILIBRIUM IN SILICIC TO INTERMEDIATE MAGMAS |
|---|
|
TOPABSTRACTINTRODUCTIONSTATES OF EQUILIBRIUM IN...EQUILIBRATION SCALESTHE LOCAL EQUILIBRIUM MODELIMPLICATIONS FOR PHASE...DISCUSSIONCONCLUDING COMMENTSAPPENDIXREFERENCES |
|---|
Total chemical equilibrium?
Most silicic to intermediate igneous bodies, especially from convergent plate boundaries, can be viewed as hybrid, periodically recharged open systems that preserve evidence for chemical heterogeneities at various scales (e.g. Eichelberger et al., 2000
; Fig. 1). At large to medium scales (from 105 down to 101 cm), chemically zoned ash-flow sheets (Smith, 1979) and plutons (e.g. Pons et al., 2006), mafic enclaves (Bacon, 1986) and banded rocks (Smith, 1979; Poussineau, 2005), illustrate the existence of compositional heterogeneities associated with magma chamber recharge and evolution (Fig. 1). The origin of these compositional gradients is still debated (e.g. Eichelberger et al., 2000, 2006). Although one should bear in mind that chemically heterogeneous whole-rock compositions do not necessarily imply chemical disequilibrium, the evidence from large to medium scales of observation consistently indicates at best limited (or incomplete) chemical equilibration in magma bodies (Eichelberger et al., 2000, 2006). Recent models favouring magma reservoirs as the product of amalgamation of small discrete magma batches (Glazner et al., 2004; Annen et al., 2006) imply scales of chemical equilibration necessarily smaller than in the case of a once molten single large body.
|
1 µm revealed by back-scattered electron imaging of plagioclase phenocryst from the P1 eruption of Mt. Pelée, Martinique (Martel et al., 1998
|
Probably the best evidence for the lack of total chemical equilibration (Table 1) in magmas comes from small-scale (100 down to 10–4 cm) petrographical, mineralogical and geochemical observations (Fig. 1b and c). As a rule, crystal populations in silicic to intermediate magmas, especially from convergent plate boundaries, are heterogeneous and may include cognate phenocrysts, phenocrysts inherited from a previous crystallization episode (inherited phenocrysts) and xenocrysts. Isotopic fingerprinting techniques now unambiguously demonstrate the mixed origin of individual crystals (e.g. Cioni et al., 1995
; Davidson & Tepley, 1997; Knesel et al., 1999). Disequilibrium phenocryst assemblages comprising, for example, calcic plagioclase, Al-rich amphibole, Mg-rich olivine and Cr-spinel coexisting with sodic plagioclase, Al-poor amphibole, quartz and a rhyolitic interstitial glass frequently occur, sometimes in the same rock (e.g. Feeley et al., 1996; Clynne, 1999; Costa & Singer, 2002). Mantling and/or resorption textures on amphibole or biotite (Rutherford & Hill, 1993; Fougnot et al., 1996) indicate reaction relationships, hence disequilibrium, between phenocrysts and melt. Fe–Mg exchange equilibrium may not be established, as shown by the occurrence of homogeneous (i.e. constant Fe/Mg ratio) cummingtonite and biotite coexisting with grossly heterogeneous orthopyroxene and hornblende phenocrysts (e.g. Schmitz & Smith, 2004). Compositionally zoned crystals imply growth under variable compositions and pre-eruptive P–T–fH2O–fO2 conditions without subsequent re-equilibration. For plagioclase phenocrysts (e.g. Singer et al., 1995; Pichavant et al., 2002; Fig. 1c), the kinetics of CaAl–NaSi atomic diffusion is extremely slow (Tsuchiyama, 1985), and diffusive chemical equilibration for crystals 10–1 to 100 cm in size may require durations longer than typical lifetimes of magma chambers (see below). Therefore, total equilibrium is generally not the rule in silicic to intermediate magmas.
|
Evidence for local equilibrium
Like any natural system governed by the laws of thermodynamics, magmas spontaneously tend to evolve toward chemical equilibrium, to an extent that depends on timescales of magmatic processes and on reaction kinetics. As for metamorphic systems (Thompson, 1959
), it may be postulated that local equilibrium (Table 1), that is, the existence of a subsystem of limited spatial extent that achieves chemical equilibrium, applies to silicic to intermediate magmas. Although commonly assumed, but rarely explicitly justified, local equilibrium can be tested by a variety of approaches, some of which are outlined below.
The most straightforward method is to compare natural phase assemblages and compositions with data from phase equilibrium studies. For example, compositions along electron microprobe traverses in phenocrysts from Mt. Pelée and Volcán San Pedro eruptive products are compared with results from phase equilibrium experiments performed on samples representative of each magma (Fig. 2). Assemblages and rim compositions of the main phenocrysts phases are simultaneously reproduced or bracketed by experimental charges (Fig. 2). Al2O3 and TiO2 (not shown) concentrations in amphibole phenocrysts from the San Pedro dacite are also experimentally bracketed. SiO2 contents of natural glasses and of experimental glasses in these critical charges overlap (Fig. 2). Thus, prior to eruption, coexisting mineral and melt phases in these two magmas were at chemical equilibrium. This is the condition necessary for the pre-eruptive parameters of each reservoir to be determined from the P–T–fH2O–fO2 of the respective critical charges (for details, see Martel et al., 1999; Costa et al., 2004). It should be noted that plagioclase core compositions are not reproduced by the experiments (Fig. 2). This implies that chemical equilibrium with the liquid is restricted to plagioclase rims. Consequently, in the two examples considered, total equilibrium is not established at the scale of the entire magma, yet a chemically equilibrated subsystem can be identified.
The preceding approach is the most general to test for chemical equilibrium between coexisting magmatic phases, as it allows the assemblage and composition of all phases present to be checked. However, equilibrium data for the magma of interest must be available, either from experiments or from thermodynamic models (see below). Other, less general methods involve testing the compositions of a subset of magmatic phases, such as glasses or selected phenocrysts. For example, natural glasses from Mt. St. Helens eruption products follow major element compositional trends similar to those of experimental liquids from phase equilibrium studies (Blundy & Cashman, 2001), suggesting that their compositions are controlled by crystal–liquid equilibrium relations. Models of element partitioning between coexisting phases also provide a test. The demonstration that Fe–Ti oxide and feldspar phenocryst pairs satisfy element partitioning models in a large number of magmatic systems (Bacon & Hirschmann, 1988; Fuhrman & Lindsley, 1988) constitutes positive evidence for local equilibrium, at least between the respective phenocrysts. Finally, thermobarometric calculations on assemblages of coexisting phenocrysts or phenocrysts plus melt can also be used to check for chemical equilibration in magmas. The fact that several independent thermometers and barometers yield a narrow range of P–T conditions (e.g. Manley & Bacon, 2000) is strong evidence for pre-eruptive equilibration between the phases used in the calculations. In the same way, the Al-in-hornblende barometer for estimating pressures of crystallization of intermediate calc-alkaline plutons is based on the chemical equilibration, either totally or partially along their rims, of at least nine coexisting phases. The validity of the barometer has been confirmed by comparing pressures derived from experimental calibration and from contact-aureole assemblages (e.g. Schmidt, 1992), thus strongly suggesting that hornblende, melt and the other mineral phases equilibrate until near-solidus conditions. Therefore, we conclude that local equilibrium is the prevailing situation in silicic to intermediate magmas.
|
EQUILIBRATION SCALES |
|---|
|
TOPABSTRACTINTRODUCTIONSTATES OF EQUILIBRIUM IN...EQUILIBRATION SCALESTHE LOCAL EQUILIBRIUM MODELIMPLICATIONS FOR PHASE...DISCUSSIONCONCLUDING COMMENTSAPPENDIXREFERENCES |
|---|
Equilibration mechanisms
In the context of local equilibrium, the spatial extent of the equilibrated subsystem, or equilibration scale (Table 1), must be defined. For silicic magmas, equilibration scales may conceivably range from 10–4 (interface between phases) to 105 cm (magma body). A first-order appreciation of equilibration scales may be obtained from the analysis of the dispersion of whole-rock compositions. However, chemical homogeneity at the hand specimen scale (or larger) may mask heterogeneities at smaller scales, as emphasized above. Therefore, investigations at scales as low as 10–4 cm are usually necessary.
Magmatic systems may evolve toward chemical equilibrium through a variety of reaction mechanisms, including nucleation and growth of new crystals, dissolution of pre-existing crystals, coupled crystal dissolution and reprecipitation and diffusive equilibration within crystal and melt phases. Vesiculation of gas bubbles also needs to be considered in volatile-bearing systems subjected to pressure gradients. Thus, equilibration scales depend on the kinetics of these individual reaction mechanisms. Rates of crystal growth are essentially controlled by reactions at the crystal–melt interface and by chemical diffusion in the melt (e.g. Kirkpatrick, 1981; Cashman, 1990). Crystal dissolution may be either reactive (i.e. rate-controlled by interface processes) or diffusive (Liang, 1999). In the latter case, it is essentially rate-controlled by diffusion in the melt (Kuo & Kirkpatrick, 1985; Zhang et al., 1989; Liang, 1999), and diffusion in crystals is unimportant (Liang, 2000) unless coupled dissolution–reprecipitation in a finite geometry is specifically considered (Liang, 2003). Rates of bubble growth essentially depend on diffusion in the melt and also on the melt viscosity (e.g. Navon et al., 1998; Martel & Bureau, 2001; Castro et al., 2005). It is worth stressing that the rates of some of the reaction mechanisms above are strongly influenced by the flow regime. Convective crystal dissolution (e.g. Kerr, 1995), which is probably the dominant dissolution mechanism in natural magmas (Liang, 2000), is faster than diffusive dissolution. In the same way, nucleation rates for plagioclase and clinopyroxene are markedly higher in dynamic (e.g. stirred) than in static crystallization experiments (Kouchi et al., 1986). Conversely, processes such as crystal growth or dissolution may promote convection; for example, through the production of boundary melts with densities different from that of the far-field melt. Therefore, for a rigorous evaluation of equilibration rates, a coupled chemical reaction–transport approach (e.g. Lasaga, 1998) is probably necessary.
Despite these complexities, all that is needed for the present discussion is a comparison between the rates of the individual equilibration mechanisms mentioned above. Timescales of chemical diffusion in representative crystals and in hydrous melt, normalized to a diffusion distance of 1 cm, are shown in Fig. 3. Most calculations are for major elements because these are the elements governing phase equilibria. The components selected for the calculations in Fig. 3 (e.g. CaAl–NaSi in plagioclase, Ca in diopside, Fe–Ti in magnetite, SiO2 and H2O in hydrous melt) are assumed to be rate-limiting for their respective phase. In other words, equilibration for the other components (either major, minor or trace elements) would take place at rates equal or faster than equilibration for the elements considered in Fig. 3. As an illustration, Mg diffusion is faster than NaSi–CaAl in plagioclase by at least two orders of magnitude (Fig. 3; La Tourette & Wasserburg, 1998; Costa et al., 2003). MgO concentrations in the plagioclase phenocryst from San Pedro (Fig. 4) closely follow the variations in An content (Fig. 3), whereas in the Mt. Pelée phenocryst MgO concentrations are homogeneously distributed in most of the crystal despite strong variations in An content (Fig. 4). Although we recognize that a component-specific treatment would be more rigorous, the approach followed here has the advantage of relative simplicity and is sufficient for our purpose.(opx)Al 2 O 3
|
|
In Fig. 3, crystalline phases have by far the longest diffusion timescales, implying that diffusion in crystals and diffusion-in-crystal-controlled reprecipitation (Liang, 2003
) are the slowest equilibration mechanisms. A large difference in equilibration timescales is apparent between crystalline phases (nearly 10 orders of magnitude between magnetite and zircon), despite the fact that the results for zircon concern a trace element (Pb) that diffuses faster than tetravalent cations such as Hf, Th and U (Cherniak & Watson, 2003). This stresses the necessity of a phase-by-phase approach, each crystalline phase having a characteristic equilibration scale. In contrast, mechanisms involving diffusive flux in the melt have more grouped timescales, within 3–4 orders of magnitude of each other (Fig. 3). Diffusive equilibration in the hydrous melt, which can also be taken as an order-of-magnitude estimate for the rate of diffusive crystal dissolution (Kuo & Kirkpatrick, 1985; Zhang et al., 1989; Liang, 1999), is the third slowest mechanism. There is a large difference between diffusion timescales in crystals and in the hydrous melt: at 900°C, diffusive equilibration (normalized to 1 cm) in plagioclase requires durations around six orders of magnitude longer than in hydrous melt. Compared with diffusive crystal dissolution, convective crystal dissolution is two orders of magnitude faster (Fig. 3). Timescales for bubble growth in low-viscosity melts, estimated from the H2O diffusion data (Navon et al., 1998; Martel & Bureau, 2001; Castro et al., 2005), overlap with crystal growth in hydrous melts (Cashman, 1990) and are in the same range as convective crystal dissolution (Fig. 3). These last mechanisms (crystal and bubble growth, convective crystal dissolution) appear to be the fastest.
Equilibration regimes
The analysis above stresses the marked contrast in equilibration rates between mechanisms controlled by diffusive flux in crystals and those controlled by diffusive flux in the melt. Therefore, it is proposed to distinguish between two end-member equilibration regimes in silicic to intermediate magmas: (1) diffusion-in-crystals-limited (DICL); (2) melt-dominated (MD) equilibration regimes.
(1) The common coexistence of a homogeneous interstitial melt with chemically zoned phenocrysts demonstrates that diffusion in crystals and/or diffusion-controlled reprecipitation rate-limit the approach toward equilibrium in many silicic to intermediate magmas. A homogeneous interstitial melt is a sign that there is no diffusive flux in the melt. Because diffusion in crystals is slow (Fig. 3), the DICL regime is necessarily associated with relatively long magmatic timescales, longer than the characteristic timescales for convection in magmas (from days to months from evidence in natural systems). Because diffusive equilibration in crystals is directly linked to magmatic timescales (e.g. Costa et al., 2003; Hawkesworth et al., 2004), equilibration scales can be calculated from diffusivities if residence times of crystals are known. As an illustration of the approach, equilibration scales are calculated and shown in Fig. 5 for an equilibration regime rate-controlled by diffusion in crystals. Timescales ranging from a few decades, through a few hundred thousand years to 1 Ma, typical of residence times of crystals in intermediate to silicic magma chambers (Costa & Chakraborty, 2004; Hawkesworth et al., 2004; Costa & Dungan, 2005), are considered. Magnetite phenocrysts require durations much shorter than 5000 years to homogenize by diffusion (i.e. of the order of 30 years for an equilibration distance of 200 µm; Fig. 5). In contrast, more than 1 Ma is necessary to homogenize plagioclase phenocrysts over distances >1 mm (Fig. 5). Zircon is much slower than plagioclase to homogenize (<10 µm after 1 Ma; Fig. 5). The calculations show that, for a magma chamber having evolved under closed-system and steady-state pre-eruption conditions (900°C) for 5000 years, only plagioclase (and zircon, if present) can preserve out-of-equilibrium compositions. The equilibration scale would thus be essentially limited by the presence of plagioclase phenocrysts (Fig. 6). Examples of such a situation are provided by the Mt. Pelée andesite and the San Pedro dacite, which both have strongly zoned plagioclase phenocrysts coexisting with homogeneous Fe–Ti oxides, orthopyroxene, amphibole and interstitial melt (Fig. 2).
|
|
This approach cannot be extrapolated to very short timescales because other faster equilibration mechanisms will become involved. However, for practical purposes, it is worth emphasizing that certain crystalline phases such as magnetite are able to record thermal perturbations of relatively short durations. For example, calculations as in Fig. 5 show that magnetite should re-equilibrate over distances of the order of 10 µm (i.e. detectable with the electron microprobe) after 100 days at 900°C (Nakamura, 1995
; Hammer et al., 2002; Devine et al., 2003). (2) The occurrence in some magmatic systems of chemically heterogeneous interstitial glasses (see below for illustration and examples) shows that equilibration scales in magmas are not always limited by diffusion in crystals. Heterogeneous interstitial glasses imply the existence of chemical gradients in the melt phase quenched by the eruption. These gradients are the sign that equilibration mechanisms such as convective mixing between melts, chemical diffusion in the melt, crystallization and/or crystal dissolution are taking place, most probably in response to compositional and thermal perturbations of the system (Fig. 6). From Fig. 3, these mechanisms all have relatively short timescales and, therefore, they will not last for very long unless the system is periodically recharged. Because the rates of these reaction mechanisms are strongly influenced by the flow regime (Fig. 3), we view convective motion as being intimately linked with the MD regime. Eventually, once convection has ceased, the chemical gradients in the melt phase will progressively smooth out and the system will enter the DICL equilibration regime until the next recharge of the magma reservoir.
Koyaguchi & Kaneko (2000) have distinguished two stages in the thermal evolution of silicic magma chambers after basalt replenishment, an initial short-duration convective-melting followed by a long-duration conductive cooling stage. During the first stage, melting and crystallization take place rapidly in the presence of a convecting liquid; the second stage corresponds to slow conductive cooling at constant melt fraction (Koyaguchi & Kaneko, 2000). Chemical equilibration during the first stage would thus proceed by reaction mechanisms such as convective mixing between liquids, crystal dissolution and crystallization (i.e. by MD processes), and during the second stage by intracrystalline diffusion at slowly decreasing temperature (i.e. by DICL processes) (Fig. 6). Therefore, the thermal evolution of periodically recharged silicic magma bodies induces mechanisms of chemical equilibration that are in close correspondence to the two end-member regimes identified in this paper. This provides a physical framework for the development of the MD and DICL equilibration regimes in magma chambers.
Natural illustrations and examples
In practice, the evaluation of equilibration scales in natural magmas, which must be based on detailed field studies, requires petrographical, mineralogical and geochemical data at different scales, from the whole-rock hand specimen down to the micrometre. One factor that needs to be assessed first is the major element composition of interstitial glasses. Homogeneous glasses imply crystal-limited equilibration scales, making detailed studies of phenocryst assemblages and zonations in crystals appropriate. Heterogeneous interstitial glasses are the signs of chemical gradients frozen in the melt phase, and it is necessary first to constrain the scale of chemical variability in glasses before attention is turned to crystals.
Homogeneous interstitial glasses are commonly encountered in andesitic–dacitic lavas erupted from central-vent volcanoes. At the whole-rock scale (centimetres to decimetres), these eruption products show restricted chemical variability, apart from the common presence of mafic enclaves. Examples include Mt. Pinatubo 1991 (Scaillet & Evans, 1999), Mt. Pelée (Martel et al., 1998, 2000), Soufrière Hills (Harford et al., 2003) and Volcán San Pedro (Costa & Singer, 2002), among others. In this group, the chemical dispersion of interstitial glass composition is typically small (e.g. < 5 wt % SiO2). No general and systematic variation in glass chemistry, for example with position in the eruptive sequence, is observed. Another important group with homogeneous interstitial glasses comprises evolved compositions (rhyodacites and rhyolites) from zoned magma chambers. Examples include Krakatau 1883 (Mandeville et al., 1996), Mt. Mazama 6845 BP (Bacon & Druitt, 1988) and Novarupta 1912 (Coombs & Gardner, 2001; Hammer et al., 2002). In all these examples, the equilibration scale is controlled by diffusive equilibration in phenocrysts and by diffusion-in-crystal-controlled reprecipitation (DICL equilibration regime). An extreme case of reduction of the equilibrium scale occurs when phenocrysts are all out of equilibrium with the interstitial melt, as proposed for the recent dacitic eruption of Mt. Unzen (Nakada & Motomura, 1999). Conversely, nearly aphyric and chemically homogeneous magma bodies (e.g. Novarupta 1912 rhyolite, Coombs & Gardner, 2001) probably approach total equilibrium. For higher, more frequent, phenocryst contents, plagioclase limits equilibration scales in many intermediate arc magmas (the situation discussed above and in Figs 2 and 5). The persistence of quartz xenocrysts in andesite (e.g. Hammer et al., 2002) may be attributed to the exceptionally sluggish dissolution kinetics of quartz. Quartz dissolution is dominated by interfacial kinetics (Hammouda & Pichavant, 1999) rather than by chemical diffusion in the melt as for most other mineral phases (Kuo & Kirkpatrick, 1985; Zhang et al., 1989; Liang, 1999).
Examples of chemically variable interstitial glasses are provided by large-volume caldera-forming zoned eruptions, such as Krakatau 1883, Mt. Mazama 6845 BP, Novarupta 1912 and Laacher See Tephra, among others. The Laacher See Tephra (Wörner & Schmincke, 1984) illustrates the case of interstitial glass compositions varying regularly along with the location in the eruptive sequence. The compositional variations concern the major, but also the minor and the trace elements. For example, CaO decreases and Th increases from around 4 wt % and 10 ppm at the top to 0·4 wt % and 120 ppm at the base of the sequence (Wörner & Schmincke, 1984). Thus, a continuous and progressive chemical zonation of interstitial glasses is established at the scale of the entire magma reservoir (104–105 cm). At smaller scales (below 102–103 cm?), the melt phase may be chemically homogeneous although this remains to be confirmed by appropriate studies. In general, the smaller the scale of the chemical homogeneity in glasses, the smaller the equilibration scale.
Glasses in mafic products of the Mt. Mazama 6845 BP, Krakatau 1883 and Novarupta 1912 eruptions systematically exhibit compositional gradients (Bacon & Druitt, 1988; Mandeville et al., 1996; Hammer et al., 2002). The range of matrix glass SiO2 contents is 67–74 wt % in the Krakatau 1883 grey dacite (Mandeville et al., 1996), 63–72 wt % in the Mt. Mazama 6845 BP andesites and cumulate scoriae (Bacon & Druitt, 1988) and 67–79 wt % in the Novarupta 1912 andesitic to dacitic products (Hammer et al., 2002). These glasses are chemically heterogeneous at various scales. The Krakatau 1883 products document chemical zonation in the melt phase at scales as small as <100 cm (heterogeneity in glass composition at the thin-section scale, Mandeville et al., 1996); a compositional gap in melt composition is present (Mandeville et al., 1996). These features necessarily imply that compositionally distinct magmatic liquids coexisted over very short timescales in the magma chamber before the eruption. In the Mt. Mazama 6845 BP products, interstitial glass MgO concentrations correlate with Fe–Ti oxide temperatures (Druitt & Bacon, 1989). This implies that Fe–Ti oxides had sufficient time to equilibrate to the locally prevailing temperature and, thus, that the melt phase may be chemically homogeneous over distances >100–101 cm (Fig. 3). In the same way, the variations in glass SiO2 concentrations at the thin-section scale in the Krakatau 1883 products are superimposed on a longer-scale chemical variation, marked by the correlation between glass SiO2 concentrations and Fe–Ti oxide temperatures (Mandeville et al., 1996). Therefore, different scales of chemical variability in the melt phase may be juxtaposed. Both the Krakatau and Mazama systems stress the link between MD equilibration mechanisms, compositional zonations in the melt phase and thermal gradients. Systems with zoned interstitial glasses can be viewed as the sum of small individual subsystems, each equilibrated under local conditions of temperature. This stresses the fact that, for such systems, any determined set of pre-eruptive parameters is of local significance only.
In conclusion, the large dispersion of equilibration scales in nature, from <100 to 104–105 cm, must be emphasized. It is consistent with the timescales of physical processes that affect silicic magma chambers (Hawkesworth et al., 2004).
|
THE LOCAL EQUILIBRIUM MODEL |
|---|
|
TOPABSTRACTINTRODUCTIONSTATES OF EQUILIBRIUM IN...EQUILIBRATION SCALESTHE LOCAL EQUILIBRIUM MODELIMPLICATIONS FOR PHASE...DISCUSSIONCONCLUDING COMMENTSAPPENDIXREFERENCES |
|---|
The main conclusions of the analysis above concerning chemical equilibrium in silicic to intermediate magmas can be summarized as follows.
(1) Total chemical equilibrium is generally not the rule, although exceptions approaching this situation probably exist. However, a subsystem in local equilibrium can be typically identified in all natural magmas. The recognition that local equilibrium is the prevailing case validates the use of principles of chemical equilibrium to interpret compositional diversity and heterogeneities at various scales in magmas.
(2) Equilibration scales in magmas are rate-limited either by diffusive flux in crystals (DICL regime) or by diffusive flux in the melt (MD regime). In the former case, equilibration mechanisms are relatively slow and associated with relatively long timescales. In the latter case, relatively short mechanisms, intimately linked with convective motion, are involved. Equilibration scales in the DICL regime can be estimated from data on magmatic timescales and diffusivities in crystals. For the MD regime, the evaluation of equilibration scales requires first that the convective state of the system is specified.
It is proposed to make a formal distinction between two discrete parts in silicic magmas. The first part corresponds to the subsystem of defined volume that is at chemical equilibrium, and whose spatial extension can be evaluated from the equilibration scales above. One example of an equilibrated subsystem is illustrated by the case of a homogeneous interstitial melt in equilibrium with phenocryst rims. The second part comprises all phases (or fractions of phases) that are out of equilibrium with the first; for example, phenocryst cores, but also xenocrysts or even foreign magmatic liquids, as discussed in the examples above. In the following, the subsystem at chemical equilibrium will be designated as the reactive magma. Systems with strong chemical gradients (zoned interstitial glasses present; see above) can be treated as the sum of small equilibrated subsystems; that is, as the sum of small reactive magmas.
One important aspect of the concept is that the reactive magma is uniquely sensitive to physical processes or perturbations that affect the system and to their individual timescales. For a given system, there is no unique reactive magma. For example, let us consider an andesitic magma body evolving under closed-system steady-state pre-eruption conditions for 500 and 5000 years, and the same magma body decompressed from 200 MPa to the surface in 1 day and 1 month (Fig. 6). These four situations will generate different equilibration scales and reactive magmas despite the fact that they concern the same body. The proposal that discrete episodes (each characterized by specific states of local equilibrium, hence specific reactive magma volumes and compositions) need to be distinguished in the lifetime of a magma chamber is now receiving support from geochronology. Zircon age data allow discrete stages of magma crystallization to be deconvoluted in long-lived silicic magmatic systems (e.g. Charlier et al., 2005). Therefore, the reactive magma model appears consistent with temporal constraints on the evolution and dynamics of silicic magma bodies.
|
IMPLICATIONS FOR PHASE EQUILIBRIUM STUDIES |
|---|
|
TOPABSTRACTINTRODUCTIONSTATES OF EQUILIBRIUM IN...EQUILIBRATION SCALESTHE LOCAL EQUILIBRIUM MODELIMPLICATIONS FOR PHASE...DISCUSSIONCONCLUDING COMMENTSAPPENDIXREFERENCES |
|---|
Chemical scaling of experiments
The recognition that chemical equilibrium in natural magmas is limited to a reactive subsystem requires phase equilibrium studies to be adapted. It is essential to approach closely the equilibrium state of the magma studied for a reliable determination of the pre-eruptive parameters. By analogy with the scaling of experiments in physics, phase equilibrium experiments need to be chemically scaled so that experimental and natural equilibrium states can be matched. To achieve this goal, specific experimental strategies and techniques need to be implemented. The choice of starting compositions is an obviously important point, although not the only one. In this respect, the use of synthetic analogues, rather than of natural rock starting materials, needs to be seriously considered. Synthetic oxide mixtures allow major and trace element compositions of starting products to be adjusted, as required for the chemical scaling of the experiments. With such mixtures, the textural and chemical complexities typical of bulk magmatic rocks would be absent. However, the use of natural rock starting materials has the decisive advantage of allowing direct comparison between experimental charges and eruption products. This comparison step is necessary for the precise determination of the pre-eruptive parameters (e.g. Martel et al., 1999
; Scaillet & Evans, 1999; Costa et al., 2004). Therefore, natural rock starting materials of suitable composition, scaled to the local natural equilibrium state, have to be preferred. Currently, there are essentially two groups of methods being developed to approach the equilibrium states of silicic magmas; that is, total and partial equilibrium experiments.
Total equilibrium experiments
Bulk-rock sample(s) chosen to be as close as possible to the composition of the reactive magma appropriate for pre-eruptive storage are used as starting materials (e.g. Martel et al., 1999; Scaillet & Evans, 1999; Costa et al., 2004). Thus, the chemical scaling of the experiments is entirely based on the rock selection. In practice, the sample is either fused to a glass or ground to a fine powder, to facilitate approach toward equilibrium. It should be noted that the use of glass does not necessarily mean acceptance of the idea that the studied magma body was once totally molten and that crystallization followed equilibrium crystallization. Glass is simply a convenient starting material to work with from the perspective of the attainment of equilibrium. In particular, the use of initially anhydrous starting glasses (e.g. Pichavant, 1987; Holtz et al., 1992; Scaillet et al., 1995; Martel et al., 1999) greatly minimizes nucleation difficulties, especially for feldspars. This is because H2O undersaturation and high degrees of undercooling initially prevail in the experiment, both conditions favouring crystal nucleation (Fenn, 1977). Crystallization experiments of this type generally yield homogeneous crystalline phases (Fig. 7a; Table 2), as crystal growth occurs under fixed conditions. Therefore, this method ensures a close approach to equilibrium. Reversals can be performed (two-stage crystallization–melting experiments, Holtz et al., 1992) to check for the attainment of equilibrium. As the starting rock is assumed to correspond to the composition of the reactive magma, the experimental and natural equilibrium states are identical (total equilibrium). However, the selection of the starting rock represents the main practical difficulty.
|
|
The critical importance of the selection of the starting composition may be illustrated on a ternary Qz–Ab–An phase diagram. Let us consider a simplified magma consisting of an interstitial melt (l) coexisting at equilibrium with plagioclase phenocrysts of composition Pl (Fig. 8a). The proportion of plagioclase phenocrysts is 17%, corresponding to the bulk composition b, which is also the composition of the reactive magma, as l and Pl are in equilibrium. Experiments performed on a starting composition identical to b at constant temperature (800°C), pressure and for two different melt H2O contents (4 and 6 wt %) define two equilibrium liquid + plagioclase pairs (Fig. 8a), which bracket the composition of the interstitial melt and plagioclase phenocrysts. Experimental crystallinities (10 and 23%) also bracket the proportion of plagioclase phenocrysts in the magma. These results confirm that the natural l + Pl assemblage is an equilibrium assemblage, indicate that the natural equilibrium state is closely approached in the experiments and, consequently, allow the pre-eruptive melt H2O concentration to be bracketed (H2O melt = 4–6 wt %). It should be noted that, although a pre-eruptive temperature of 800°C was assumed, temperature (and also pressure) can be constrained in a way identical to that above for the melt H2O content.
|
At the same temperature (800°C), for the same melt H2O contents (4 and 6 wt % in melt), pressure and fO2, experiments starting from composition b1 yield two new equilibrium liquid + plagioclase pairs that differ from those obtained when starting from composition b (Fig. 8a and b). Composition b1 is chosen so as to be enriched in plagioclase Pl relative to composition b (48% vs 17%), as would happen for a rock with cumulus plagioclase. As a result, plagioclase is more abundant in charges starting from the b1 than from the b composition. Yet, the two new experimental pairs bracket the natural equilibrium l + Pl pair because the accumulated plagioclase has the equilibrium composition Pl (i.e. l, b, b1 and Pl are collinear; Fig. 8b). Therefore, by using the starting composition b1, the same bracket on pre-eruptive melt H2O content is obtained (i.e. 4–6 wt % H2O). However, the fact that plagioclase proportions in experimental charges (33 and 53%) are higher than in the reactive magma b (17%) is an indication that b1 grossly differs from the reactive magma composition. If now the experiments (at 800°C, for the same pressure and fO2) start from a rock with xenocrystic plagioclase present (i.e. b2, obtained by adding 16% plagioclase Plx to b, Fig. 8c), the two experimental liquid + plagioclase pairs fail to bracket the natural l + Pl equilibrium pair. To approach l, melt H2O concentrations >6 wt% would be needed, and, to approach Pl, < 4 wt% would be needed (Fig. 8c). In other words, the pre-eruptive melt H2O content cannot be defined if b2 is used as an experimental starting material. The definition of the pre-eruptive parameters therefore requires the use of starting compositions along the l–Pl tie-line or, more generally, of compositions lying within the multicomponent space defined by phases coexisting at equilibrium in the magma chamber.
Tests developed to check total equilibrium experiments involve comparison between phase proportions in addition to phase assemblages and compositions, in experimental and natural products. These tests can only be performed a posteriori, once the experimental programme has been completed. If the starting composition perfectly approximates the reactive magma, phase assemblages, compositions and proportions should be identical in both types of products (Appendix Fig. A1). As an example, the three experiments that are used to bracket the pre-eruptive conditions of the San Pedro dacite reservoir in Fig. 2 have crystal contents ranging between 24 and 37 wt %, whereas the erupted dacite contains 29 wt % phenocrysts (Costa et al., 2004). In the case of the Mt. Pelée reservoir, the three experiments that best define the pre-eruptive conditions (Fig. 2) have total proportions of crystals between 48 and 50 vol. %, identical to the proportion of phenocrysts in the starting materials (48 vol. %, Fig. A1). However, crystallinities in these experiments are higher than in the reactive magma by around 5 vol. % (Appendix, Fig. A1), indicating that the experimental starting materials (Martel et al., 1998, 1999) depart a little but significantly from the reactive magma composition (Appendix, Fig. A1; Table A1). Consequently, either the pre-eruptive melt H2O concentration determined from these experiments is slightly underestimated (by <0·5 wt %) or, alternatively, the pre-eruptive temperature is slightly overestimated (by <25°C). In the same way, Blundy & Cashman (2001) have identified a mismatch between natural and experimental phase proportions, on one hand, and phase compositions, on the other hand, in experiments on the Mt. St. Helens dacite (Rutherford et al., 1985; Rutherford & Devine, 1988). One possible explanation would be the use, in these experiments, of a starting material less chemically evolved (i.e. poorer in SiO2, richer in CaO) than the reactive magma appropriate for the 1980 Mt. St. Helens magma chamber.
Partial equilibrium experiments
As in the case of total equilibrium experiments, bulk-rock sample(s) representative of the magma studied are used. However, instead of glass or fine crystalline powder, coarse-grained starting materials obtained by light crushing are used as starting materials (Cottrell et al., 1999; Coombs et al., 2000; Hammer & Rutherford, 2002). In so doing, exposure of phenocryst cores to the melt is minimized and the reactive experimental system reduces (because of kinetic factors) to interstitial glass and phenocryst rims, phenocryst cores being excluded. The chemical scaling of the experiments is thus based on reaction kinetics, rather than on the selection of the appropriate rock composition as in the case of the total equilibrium approach.
The main characteristics of such experiments are that most crystalline phases are chemically heterogeneous (Fig. 7b), this heterogeneity being inherited from the starting material. Thus, total crystal–liquid equilibrium is not attained in the charge. In detail, the experimental results reflect the influence of non-intensive parameters such as the grain size of the starting powder, the experimental duration, and the heating and cooling rate. These latter parameters are imposed by practical laboratory constraints on equipment and run durations. Consequently, they cannot be scaled exactly to values in natural systems and, therefore, natural equilibrium states cannot be exactly reproduced. However, experimental melts reach steady-state compositions relatively rapidly. Partial equilibrium between melt and phenocryst rims is attained typically after durations of a few days (e.g. Cottrell et al., 1999; Coombs et al., 2000), although the definition of rim compositions may not be always straightforward (Fig. 7b). To allow a close approach to the natural equilibrium state, experiments of this type are usually kept within a narrow range of P–T–fH2O–fO2 conditions. Therefore, some a priori knowledge of these parameters (e.g. temperature from Fe–Ti oxides and melt H2O content from the analysis of glass inclusions) is usually needed. Variations of melt and plagioclase rim compositions with experimental conditions provide a check of initial estimates of temperature and melt H2O content, and allow the other parameters (e.g. pressure) to be determined (Cottrell et al., 1999; Coombs et al., 2000).
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONSTATES OF EQUILIBRIUM IN...EQUILIBRATION SCALESTHE LOCAL EQUILIBRIUM MODELIMPLICATIONS FOR PHASE...DISCUSSIONCONCLUDING COMMENTSAPPENDIXREFERENCES |
|---|
This overview illustrates the complementarity of experimental strategies and methods currently developed to approach equilibrium states in natural magmas. Total equilibrium experiments are to be preferred for systems with relatively large reactive magma volumes (i.e. DICL systems). Partial equilibrium experiments are effective especially when information is already available on the conditions of natural equilibration, and they might be ideal to refine results from total equilibrium experiments. Combining the essential aspects of the two strategies above appears conceivably possible, and future studies in this direction are encouraged. Magmatic systems with small reactive magma volumes are especially challenging to handle (the reactive magma composition being difficult to define), and this should stimulate future developments. Some recent experimental studies illustrate the difficulty in reconstructing natural equilibrium states for systems with small reactive magma volumes such as, but not limited to, MD systems. For example, the fact that phenocryst assemblages are not reproduced in experimental charges despite variations and adjustments in P–T–fH2O–fO2 conditions (Berndt et al., 2001
; Gardner et al., 2002; Harms et al., 2004; Holtz et al., 2005) is a strong indi