- © 2002 American Mineralogist
An extensive electron microprobe survey of amphibole compositions in the Fish Canyon magma (2146 analyses), more than 80% of which are from high-resolution (<10 μm steps) core-to-rim traverses across large euhedral phenocrysts, provides: (1) temporal constraints on the immediately pre-eruptive P-T-fH2O evolution of the magma, and (2) a means of evaluating recent calibrations of the Al-in-hornblende barometer (Anderson and Smith 1995; hereafter AS1995) and thermometers (Blundy and Holland 1990; thermometers A and B of Holland and Blundy 1994; hereafter BH1990, HB1994TA, and HB1994TB).
Hornblende phenocrysts are variable for most major elements (e.g., 5–9 wt% Al2O3 and 44–50 wt% SiO2). This compositional range is controlled by two major temperature-sensitive coupled substitutions. Approximately 50% of the total Al variation (~0.8 atoms per formula unit = apfu) is due to the edenite exchange [TSi + A□ = TAl + A(Na + K)] and another 25–30% is the consequence of a Ti-Tschermak exchange (TSi + M1–M3 Mn = TAl + M1–M3Ti). In contrast, the pressure-sensitive Al-Tschermak substitution (TSi + M1–M3 Mg = TAl + M1–M3Al) did not play a significant role, as M1–M3Al does not correlate with TAl and is always <0.2 apfu.
In order to constrain the ranges of absolute P and T over which these hornblendes crystallized and to assess the sensitivity of the recent thermo-barometric algorithms of BH1990, HB1994TA (requiring silica saturation), HB1994TB (not requiring silica saturation) and AS1995, we have calculated pressures and temperatures for two selected populations of analyses wherein Al2O3 contents are within analytical error (5.95 to 6.05 wt% Al2O3, N = 78 and 7.7 to 7.8 wt% Al2O3, N = 40). The barometric formulation of AS1995 gives a mean pressure of 2.24 ± 0.05 for the high-Al population at 760 °C, which is indistinguishable from the 2.4 ± 0.5 kbar estimate of Johnson and Rutherford (1989a). A high sensitivity to temperature at low P is suggested by the geologically implausibly shallow depths calculated for the low-Al population (<1 kbar at 760 °C). The three thermometric formulations give reasonable results between 706 and 814 °C, but the HB1994TA calibration gives a mean temperature higher by ~50 °C and is more sensitive to small analytical differences (~100 °C spread for each population). HB1994TB is considered the most reliable calibration of the Al-in-hornblende thermometer as it most precisely reproduces T estimates determined by independent methods.
Nine out of 14 traverses across large phenocrysts from the Fish Canyon magma display rimward increases in TAl, A(Na + K), and M1–M3Ti, compensated by decreases in TSi, and M1–M3Mn. Using the HB1994TB algorithm, the low-Al population, typical of near-core compositions, gives a mean temperature of ~715 °C, which is ~35–45 °C above the water-saturated granite solidus at 2–2.5 kbar. The high-Al population, representing the average rim composition, gives a value around 760 °C, which is indistinguishable from independent T determinations using coexisting Fe-Ti oxides and Qtz-Mag oxygen isotope thermometry. These profiles suggest that Fish Canyon hornblendes crystallized during near-isobaric reheating over a temperature range of ~40 °C, which is consistent with our model of rejuvenation and remobilization of a pre-existing near-solidus crystal mush of batholithic dimensions via shallow intrusion of more mafic magma (Bachmann et al. 2002). Crystallization of hornblende from a high-SiO2, low-MgO melt during reheating requires an open system, in which both heat and mass, in particular volatiles, are transferred from the underlying mafic magma.
The Al-in-hornblende geobarometer first proposed by Hammarstrom and Zen (1986) has been widely exploited for estimating the depth of emplacement of calc-alkaline plutons (e.g., Ghent et al. 1991; Vyhnal et al. 1991; Ague and Brandon 1992). However, the simplifying assumptions behind this model and its initial calibration on the basis of P estimates derived from metamorphic mineral assemblages in associated contact aureoles have provoked numerous critical evaluations, including a similar empirical approach on a larger data set (Hollister et al. 1987) and experimental studies designed to calibrate the sensitivity of hornblende composition to other intensive variables, especially T (e.g., Thomas and Ernst 1990; Schmidt 1992). This method also has been applied to volcanic rocks (Johnson and Rutherford 1989b) despite the relatively rare occurrence of the necessary low-variance, near-solidus mineral assemblage in magmas containing substantial melt.
As the role of different intensive variables in controlling amphibole composition is not perfectly resolved, additional investigations of volcanic rocks containing the requisite mineral assemblage offer advantages over studies of plutonic rocks: (1) T and fO2 can be determined with greater certainty, as can the composition of melt (glass) coexisting with rims of hornblende crystals; (2) fewer assumptions about post-crystallization diffusive equilibration are required; and (3) textural evidence pertaining to the timing of hornblende crystallization relative to other phases may be interpreted with less uncertainty. The results of the current investigation are based primarily on high-resolution, core-to-rim electron microprobe traverses (<10 μm point spacing) across large, unbroken euhedral amphibole phenocrysts from the two major units of the Fish Canyon magmatic system, San Juan volcanic field; (1) the basal, glassy portion of the early erupted Pagosa Peak Dacite (Bachmann et al. 2000), and (2) the Fish Canyon Tuff (both outflow and intracaldera facies). We have proposed, on the basis of multiple lines of textural and phase-chemical evidence, that the immediately pre-eruptive evolution of the Fish Canyon magma was characterized by reheating, rejuvenation via new melt production, and remobilization of a pre-existing near-solidus crystal mush of batholithic dimensions via intrusion of more-mafic magma beneath the silicic mush (Bachmann et al. 2002). We infer that the physical mechanism whereby heat, and possibly minor mass (mainly volatile species), were transferred from the underplated magma to the Fish Canyon magma body was similar to that proposed by Couch et al. (2001) for the generation of the current Soufrière Hills eruption, Island of Montserrat. In this context, the purpose of presenting hornblende zoning profiles from Fish Canyon magma is threefold. The first is to use such data from a magma for which intensive parameters have been determined independently in order to evaluate the roles that pressure and/or temperature have played in causing the observed chemical variations. The second is to exploit the large data set to assess the sensitivities of published Al-in-hornblende thermo-barometric calibrations by calculating pressure and temperature for each individual analysis in selected populations having nearly identical Al2O3 content (within 0.1 wt% Al2O3). The third is to use the patterns of compositional profiles in hornblende as proxies for the temporal evolution of pressure, temperature, and fH2O in the Fish Canyon magma shortly prior to eruption.
The voluminous (~5000 km3) Fish Canyon Tuff was erupted from the 100 × 35 km La Garita caldera at ~28 Ma, during a particularly productive period of high-K calc-alkaline magmatism (~4000 km3 of magma per million years) in the San Juan volcanic field, Colorado (Lipman 2000; Lipman et al. 1996; Fig. 1⇓). Shortly prior to and after this climactic eruption, less-explosive volumetrically subsidiary units of the same magma were emplaced (Bachmann et al. 2000), allowing access to a complete suite of samples from early to late-erupted material. In addition to these three units (the precursory Pagosa Peak Dacite, the Fish Canyon Tuff, and the post-collapse Nutras Creek Dacite), samples of Fish Canyon magma are also available as holocrystalline granodioritic xenoliths present in the upper intracaldera Fish Canyon Tuff. These xenoliths have Fish Canyon mineralogy and whole-rock composition, and are interpreted as fragments of the solidified margin of the Fish Canyon magma chamber entrained by the eruption of the Fish Canyon Tuff (Bachmann 2001; Bachmann et al. 2002).
The phenocryst-rich (~45% crystals) Fish Canyon magma was remarkably homogeneous (68 ± 0.5 wt% SiO2; recalculated to 100% anhydrous, as are all major-element rock and glass compositions reported in the text of this paper; Bachmann et al. 2002) for such a large magma body, and is characterized throughout by a near-solidus mineral assemblage of either 12 or 13 phases; in addition to melt, and possibly a gas phase, 11 mineral phases (Pl + Kfs + Qtz + Hbl + Bt + Spn + Mag + Ilm + Ap + Zrn + Po) were present. As phosphate, sulfide, and zircon are among these phases, we suggest that the correct number of components is higher than the 9–11 usually cited for this magma, but regardless of whether minor phases and components are considered, this is a low-variance system wherein the effects of intensive variables can potentially be isolated and quantified. On the basis of the volume estimate of ~5000 km3 for the Fish Canyon Tuff and the area of the La Garita caldera (~2500 km2), the vertical dimension of the magma chamber must have been approximately 1.5 to 2.5 km. Thus, although we could expect some minor variations in the compositions of minerals that are sensitive to P, the total range of P variation in the magma reservoir might have been <1 kbar, within the ±0.5 kbar uncertainty of the recent calibrations of the Al-in-hornblende barometer (Johnson and Rutherford 1989b; Schmidt 1992; Anderson and Smith 1995).
The P, T, fH2O and fO2 of phase equilibration in Fish Canyon magma have been determined by a combination of Fe-Ti oxide thermometry (Whitney and Stormer 1985) and experimental duplication of the mineral assemblage and phase compositions at fH2O = 0.5, 760 °C, P = 2.4 kbar and fO2 = −11.4 (Johnson and Rutherford 1989a). This temperature has been corroborated (760–770 ± 10 °C) by quartz-magnetite oxygen thermometry (Bindeman pers. comm.). The low pressure of phase equilibration obtained by Johnson and Rutherford (1989a; 2.4 ± 5 kbar) is consistent with eruption of the Fish Canyon magma from a caldera, and the extremely evolved high-SiO2 rhyolite matrix glass (>75 wt% SiO2) reflects coexistence of 11 solid phases with a pseudo-eutectic melt just above the solidus temperature.
An important issue with respect to the barometric interpretation of amphibole compositions in such a system is quartz-coexistence with the required near-solidus mineral assemblage: aSiO2 = 1 is considered essential (Johnson and Rutherford 1989b; Rutherford and Johnson 1992). Although the most abundant minerals in the Fish Canyon magma are plagioclase, sanidine, and quartz, the latter is invariably partly resorbed and feldspars show evidence of both resorption and growth (presence of multi-generation Rapakivi textures). As the matrix glass is high-SiO2 rhyolite (>95% normative Qtz + Ab + Or), we believe that the melt was close to saturation in SiO2 and alkali feldspar even though sanidine and quartz were not crystallizing immediately prior to eruption. All the other phases are euhedral and appear to have been in equilibrium with melt immediately prior to eruption (Bachmann et al. 2002). Quartz coexistence is not required by the thermometric algorithms, as they were calibrated using mineral equilibria that do not include quartz (i.e., for HB1994TB, only Pl and Amph are required).
Fish Canyon amphiboles are invariably euhedral and optically display oscillatory compositional zoning. Nearly all crystals contain inclusions of plagioclase, biotite, Fe-Ti oxides, titanite, zircon, and apatite. A large (~5 mm) amphibole phenocryst in a Pagosa Peak Dacite sample contains a resorbed pargasitic core, which we consider to be a mantled relic from a period of magma evolution that predates the immediately pre-eruptive reheating, remelting, and remobilization event postulated by Bachmann et al. (2002).
Plagioclase displays a wide spectrum of textures indicative of both resorption and growth, but is found predominantly as oscillatory zoned overgrowths on volumetrically minor, partially resorbed, and commonly sieve-textured cores. These cores have generally more calcic compositions (An40–An75) than their reverse- and oscillatory-zoned mantles. The baseline zoning trend in plagioclase overgrowths is from an inner An-minimum (An27–28) to more calcic compositions (An32–33) at the rim. These overgrowths also generally display 1–4 internal resorption surfaces, commonly followed by calcic spikes up to An40–42 just outboard of resorption surfaces. Such a pattern is consistent with a dynamic system, in which temperature gradually increased, possibly in conjunction with some mass contribution (mainly H2O?) from a subjacent more-mafic magma (Bachmann et al. 2002). Plagioclase inclusions in hornblende cores have compositions indistinguishable from the oscillatory-zoned overgrowths (~An27–33).
Fish Canyon interstitial melt, well-preserved as non-devitrified glass in multiple vitrophyric samples of the Pagosa Peak Dacite and Fish Canyon Tuff, is a near-eutectic high-SiO2 rhyolite (76–78 wt% SiO2), low in CaO (0.7 wt%), FeO (~0.6 wt%), and MgO (<0.1 wt%), but fairly rich in K2O (5.5–6 wt%). The glass is depleted in Y (5–6 ppm), HREE (<1 ppm Yb), and Zr (~70 ppm), but contains significant amounts of Sr (60–90 ppm), Ba (480–550 ppm), and Eu (0.5 ppm) in comparison to some high-SiO2 rhyolites. These trace-element concentrations are interpreted as evidence for extensive crystallization of hornblende and accessory phases, but limited fractionation of feldspars, in accord with textural evidence. The low MgO content of the coexisting glass indicates that if mass transfer did occur, the high-SiO2 rhyolite matrix liquid served as a transfer medium from which ferromagnesian minerals precipitated in proportion to this added mass.
Electron microprobe analyses (Table 1⇓) were performed at the University of Lausanne using a Cameca SX-50 instrument, equipped with five wavelength-dispersive spectrometers. An accelerating voltage of 15 kV and a beam current of 15 nA were used for amphibole, with a beam diameter of ~2 μm. Special care was taken during the calibration of the major elements; all were regularly checked in the course of the analyses on different standards of known composition (Ort2 = orthoclase (provided by E.Essene, Michigan); Byt1 = bytownite from Crystal Bay (provided by E.Essene, Michigan); Hb1 = hornblende FA86-6, Grenville, Ontario, (Cosca et al., 1991); B6Rhod = rhodonite mineral mount MINM25-53 (provided by Astimex Scientific limited); Andr = andradite (provided by Cameca); Opx1 = orthopyroxene (provided by E.Essene, Michigan); alb4 = albite from Tiburon (provided by E.Essene, Michigan); Mnti = MnTiO3 (provided by Cameca); Fph1 = synthetic F-phlogopite (provided by ETHZ) ) to ensure a relative error of less that 1%. Amphiboles have on average ~2 wt% of (H2O + F + Cl), and only analyses with anhydrous oxide totals (SiO2, Al2O3, FeOtot, MgO, CaO, Na2O, K2O, TiO2, MnO) of 98 ± 1 were retained.
More than 80% of the hornblende analyses (1870 out of 2146) are from high-resolution core-to-rim profiles across 14 large hornblende phenocrysts, with point spacings of 5–10 μm over distances of 300 to 1700 μm (Fig. 2⇓ and Appendix 1a–n). The rest of the data points were acquired by point analysis in ten other individual phenocrysts from all three units of the Fish Canyon system. Analyzed crystals all display crystal faces against fresh glass and contain inclusions of plagioclase, Fe-Ti oxides, biotite, titanite, zircon, and apatite (Fig. 2⇓). Optical zoning is seen as concentric green bands, displaying slight variations in color intensity and generally parallel to the faces of the crystal, indicative of oscillatory zoning (e.g., Fig. 2b⇓ and Appendix 1e). Some crystals display progressively increasing optical absorption toward the rim (Fig. 2b⇓ and Appendix 1h).
Cation abundances were recalculated following the method of Leake et al. (1997), which assumes a formula cation sum of 13, excluding Ca, Na, and K (13eCNK), with all Fe as FeO. The Fe3+/Fe2+ ratios are then calculated assuming a corresponding charge of 46. This recalculation scheme was demonstrated to be the most accurate for site occupancy and Fe3+/Fe2+ determinations in amphiboles (Cosca et al. 1991). However, in order to apply the Holland and Blundy (1994) thermometers, the microprobe analyses used in the assessment of the thermo-barometric algorithms (the low- and high-Al populations: see following sections) were also recalculated using the scheme proposed in Appendix B of Holland and Blundy (1994). In general, both normalization schemes give very similar cation abundances, but the Fe3+/Fe2+ ratio is always higher by a factor of two to three using Leake et al. (1997), as already documented by Anderson (1996). In this case, thermometric calculations appear to be fairly insensitive to the chosen recalculation scheme, as mean temperatures yielded by HB1994TA using either Leake et al. (1997) or Holland and Blundy (1994) are within 5 °C (Table 2⇓).
Fish Canyon Amphibole chemistry
Significant variations in major elements are recorded in Fish Canyon amphiboles, reaching 4 wt% absolute for FeOtot and MgO, 7 wt% for Al2O3, and ~9 wt% for SiO2 (Fig. 3⇓). Part of this spread is due to the presence of a texturally distinct resorbed core of pargasitic composition in one large amphibole in a Pagosa Peak Dacite sample (PCB1). All data points from this core are plotted in open squares in Figures 4⇓ and 5⇓. The vast majority of analyses (>99%) falls in a narrower range (5–9 wt% Al2O3, 44–50 wt% SiO2), which represents compositional variations in prevalent Fish Canyon hornblendes. Analyses of amphiboles from all three major units of the Fish Canyon system (Pagosa Peak Dacite, Fish Canyon Tuff, and Nutras Creek Dacite) overlap perfectly for all major oxides (Fig. 3⇓), indicating the absence of large, stable compositional or thermal gradients in the chamber. Analyses of an amphibole in a granodioritic xenolith also overlap for all major oxides except for slightly higher Na2O and lower FeOtot. Analyses of a natural and an experimental hornblende (Exp 107; 780 °C at 2 kbar and fH2O = 0.5) reported by Johnson and Rutherford (1989a) are shown for comparison.
Nearly all the Al-variations in Fish Canyon amphiboles (~0.8 atoms per formula unit = apfu) are accommodated by the tetrahedral site (TAl). The amount of M1–M3Al is minor (<0.2 apfu) and negative values were calculated for some low-Al analyses (Fig. 4e⇑). Approximately 85–90% of the variations in TAl (~0.8 apfu) are accounted for by: (1) the edenite exchange (TSi + A□ = TAl + A(Na + K); ~50% of the total Al variation); (2) a derivative of the Ti-Tschermak exchange (2TSi + M1–M3Mn = 2TAl + M1–M3Ti; ~25–30%); and (3) the “plagioclase exchange” (TSi + M4Na = TAl + M4Ca; 10–15%) (See Figs. 4a–d⇑; percentages derived from the slopes of regression lines). Fe2+ and Mg also vary by ~0.7 apfu, but are thought to have been exchanged independently of the other cations as a simple substitution (Fig. 5c⇑).
The experimental studies that have focused on the effect of pressure on the Al-content of amphibole (Johnson and Rutherford 1989b; Thomas and Ernst 1990; Schmidt 1992) have demonstrated that the Al-Tschermak substitution (TSi + M1–M3Mg = TAl + M1–M3Al) is favored by increasing pressure. In the near-isothermal (760 ± 20 °C; 0.25 < fH2O < 0.5) experiments of Johnson and Rutherford (1989b), an increase from 2 to 8 kbar induced an increase in both TAl and M1–M3Al of ~0.7 apfu in Fish Canyon hornblende, in equilibrium with liquid, Pl, Kfs, Qtz, Bt, Ttn, and Mag or Ilm. Si and Mg decreased by corresponding amounts. In contrast, temperature variations do not favor the Al-Tschermak substitution, but are marked by an increase in TAl through the edenite exchange (Spear 1981; Blundy and Holland 1990). The Ti-content of amphibole is also known to correlate positively with temperature in the presence of a Ti-rich phase (Spear 1981). Independently of P and T, fH2O appears to have an effect on Altot as well. Experiments on the Mount Pinatubo Dacite (Scaillet and Evans 1999) have reported an Altot (no discrimination between TAl and M1–M3Al) increase with increasing fH2O. On the basis of these experimental results, the observed cation variations in the hornblende in the Fish Canyon magma suggest amphibole crystallization at near-isobaric conditions over either a significant temperature range and/ or significant fH2O range.
Evaluation of Al-in-hornblende thermometers and barometers
In order to compare the Al-in-hornblende thermometers of Blundy and Holland (1990; hereafter BH1990) and Holland and Blundy (1994; hereafter HB1994TA for the edenite-tremolite thermometer, requiring silica saturation, and HB1994TB for the edenite-richterite thermometer, not requiring silica saturation), we have selected two compositionally restricted populations of hornblende analyses (averages in Table 1⇑) defined by those with 5.95–6.05 wt% Al2O3 (low-Al population, N = 78) and 7.7–7.8 wt% Al2O3 (high-Al population, N = 40). These populations are also used as input into the revised barometric formulation of Anderson and Smith (1995). These two populations do not correspond to the extreme limits of Al variation (~5–9 wt% Al2O3, excluding the rare pargasitic compositions; Fig. 3⇑), as ~20% of the total analytical data set has lower Al2O3 and ~10% has higher Al2O3. The low-Al population is representative of Al-troughs that are typically located in the internal part of many Fish Canyon hornblende phenocrysts and the more aluminous compositions are characteristic of phenocryst rims (Fig. 2⇑). We believe that the barometric algorithm, requiring coexistence of multiple phases, can also be applied to the low-Al population, as the large, euhedral amphibole crystals chosen for the analyses contain nearly all the other phenocrystic phases as inclusions (only quartz has never been seen), and the included phases, in particular plagioclase, have compositions indistinguishable from rims of their external counterparts. The implications of zoning profiles in Fish Canyon phenocrysts are discussed subsequently. In this section, we have used these two populations to: (1) compare calculated temperatures and pressures using recent thermometric and barometric calibrations with those obtained independently; (2) assess the sensitivities of the different calibrations to variations in hornblende composition where the variations in Al2O3 are on the order of analytical uncertainties (Table 2⇑, Figs. 6⇓ and 7⇓); and (3) test the temperature calibrations for sensitivity to P and Xab, and the pressure calibration for sensitivity to temperature over the Al-range recorded in most Fish Canyon hornblende phenocrysts.
The calculated mean temperatures and temperature ranges are nearly identical for the BH1990 and HB1994TB thermometers, but significantly different for the HB1994TA (Fig. 6⇑). The HB1994TA calibration gives temperature estimates that are systematically 45–55 °C higher than those from the BH1990 and HB1994TB, and yields temperature ranges of ~100 °C for both populations, vs. ~50–60 °C with BH1990 and HB1994TB, regardless of P and Xab. The consequences of increasing either Xab by 0.05 or P by 0.5 kbar (other parameters held constant) on BH1990 and HB1994TA are to decrease the calculated mean temperatures by ~5–10 °C, which are small shifts relative to the total temperature ranges obtained for each population for any set of parameters, or compared to the differences between calibrations. In contrast, HB1994TB is more sensitive to changes in Xab (at the same pressure, lowering Xab of 0.05 increases the temperature by ~20 °C), but nearly insensitive to small pressure variations (no temperature difference over a 0.5 kbar variation). Neither of these thermometers explicitly includes a term for the temperature-sensitive variations in Ti. The low-Al population is lower in TiO2 than the high-Al population (1.20 ± 0.06 vs. 1.54 ± 0.07 wt% TiO2). Consequently, the temperature differences between the two populations may be underestimated.
On the basis of an assessment of the different Al-in-hornblende thermometric algorithms on data from plutonic rocks, Anderson (1996) concluded that HB1994TB is the most reliable calibration, an assertion confirmed by the data on the Fish Canyon system. The mean temperatures obtained with HB1994TB for the low-Al population (715 ± 10 °C; Xab = 0.66 and P = 2.25 kbar) are consistent with crystallization of the low-Al hornblende compositions slightly above the water-saturated granitic (Qz-Ab-Or-An-H2O) solidus at 2–2.5 kbar (~670–690 °C for an An/Ab ratio of 0.3; Johannes and Holtz 1996), and the calculated temperatures (762 ± 15 °C; Xab = 0.63 and P = 2.25 kbar) for the high-Al population (i.e., hornblende rims) are indistinguishable from those obtained using coexisting Fe-Ti oxides (760 ± 30 °C; Johnson and Rutherford 1989a), Qtz-Mag oxygen isotope thermometry (762 ± 10 °C; Bindeman, unpublished data), and experimental duplication of phases and melt compositions (Johnson and Rutherford 1989a). If we were to assume that the low-Al hornblende population coexisted with plagioclase compositions of An40–45, this combination would yield an average temperature similar to that of the high-Al population (~760 °C). However, the presence of An27–33 inclusions in the cores of most of analyzed hornblendes (Appendix 1h, i, j, k, and m) imposes tight constraints on the coexisting plagioclase compositions (<An35). The HB1994TA formulation is more sensitive to small analytical differences, as noted previously by Anderson and Smith (1995), and it may overestimate (by ~50 °C) the absolute temperature of the system. Although the semi-empirical formulation of BH1990 does not take into account any compositional parameters in hornblende apart from tetrahedral occupancy, the temperature estimates obtained with it closely correspond to the HB1994TB algorithm. On the basis of these thermometric results, we have used temperatures of 700 and 760 °C, respectively, for calculating the apparent pressures recorded by the low-Al and high-Al populations.
Pressure estimates obtained with the Anderson and Smith (1995) calibration are highly sensitive to input parameters (Altot and T), at least for relatively low-P and low-Al hornblendes, such as those in the Fish Canyon magma. Although the mean pressure value calculated for the high-Al population (2.25 ± 0.6 kbar at 760 °C) is indistinguishable from the estimate of 2.4 ± 0.5 kbar favored by Johnson and Rutherford (1989a), the low-Al population yields a significantly lower pressure for any plausible temperatures for this system (1.68 ± 0.6 kbar at 700 °C; 1.52 ± 0.6 kbar at 715 °C; 0.88 ± 0.6 kbar at 760 °C; Fig. 7⇑, Table 2⇑). Even lowering the temperature below the solidus (i.e., 600 °C) still yields pressures barely reaching 2 kbar (Table 2⇑). The observed temperature sensitivity is large compared with the uncertainties associated with estimating temperatures of hornblende crystallization in plutonic systems.
If these pressure estimates were geologically meaningful, the low pressures calculated for near-core compositions in Fish Canyon hornblendes would require a dynamic history wherein the magma first ponded at a very shallow depth (the pressure calculated for the lowest Al-content in the Fish Canyon hornblende database (Altot = 0.83) at 700 °C is 0.97 kbar, ~3 km) and then foundered >3 kilometers immediately before eruption. Thus, calculated pressure values <1.8–2.0 kbar for the low-Al compositions can be dismissed as geologically unreasonable for this magma, and the potential pressure misestimation for magmas which have equilibrated at sub-caldera depths (~2–3 kbar) using Anderson and Smith (1995) calibration is seen to be as much as ~2 kbar, particularly if temperature is not known precisely.
Zoning profiles in Fish Canyon amphiboles
Fourteen core-to-rim microprobe profiles across large, euhedral amphiboles, with lengths ranging from 300 to 1750 μm, are reported in this study (see Fig. 2⇑ and Appendix 1). Eight of them come from five different Pagosa Peak Dacite samples (Bfc PCB1, 68, 91, 171, 196a), which provide the largest phenocrysts. Of the six other profiles, two are from bulk outflow Fish Canyon Tuff samples (Bfc FV and 113), three are from intracaldera Fish Canyon Tuff pumices (one in Bfc 129 and two in 191a), and one is from a granodioritic xenolith with Fish Canyon mineralogy and mineral chemistry (Bfc 187). Traverses in Nutras Creek Dacite amphiboles were not performed due to signs of alteration, but sixteen data points are included in the data set.
Although each profile is different in detail, most (nine out of 14; Appendix 1a to c, g to j, m, and n) are characterized by a progressive rimward increase in Al2O3. Maximum Al contents (~7.5–8.5 wt% Al2O3) typically occur near the rims of phenocrysts, and low-Al troughs (~5.0–6.5 wt% Al2O3) are found near the cores of large crystals. These rimward increases in Al2O3 are accompanied by coupled increases in TiO2, Na2O, K2O, and FeOtot, whereas MgO, MnO, and SiO2 decrease. CaO is the only major element that remains approximately constant (~11.5–12 wt% CaO), even though a slight positive correlation with Al is observed in some traverses. Fine-scale (10–50 μm wide), but large-magnitude (from ± 0.2 to ± 1 wt% in Al2O3) oscillatory zoning is superimposed on this dominant trend. Except for CaO, which remains nearly flat, Al2O3, FeOtot, MgO, SiO2, TiO2, Na2O, and K2O fluctuate coherently. Small peaks in Al2O3 correlate with small peaks in TiO2, Na2O, K2O, and FeOtot, and these are mirrored by troughs in MgO, MnO, and SiO2. This striking interdependence among the different elements and the large amplitude of these variations attests to the importance of coupled substitutions in these chemical variations and indicates that both the long-wavelength trends and fine-scale oscillations were caused by the same factors.
Diverse zoning patterns are observed even for amphiboles with dominant rimward increases in Al. Several profiles (e.g., Fig. 2a⇑, and Appendix 1m) display a relatively gentle rimward rise in Al with low-amplitude oscillatory zoning, whereas others have larger amplitude oscillatory variations (e.g., Fig. 2b⇑), or even prominent Al-spikes (Appendix 1j) superimposed on a progressively rising trend. One example (Appendix 1n) shows an abrupt increase in Al, followed by a relatively flat plateau at ~7.5–8 wt% Al2O3, and several profiles (Appendix 1g, m, and n) begin at relatively high Al content (>7 wt% Al2O3), descend to around 5.5–6 wt% Al2O3, and then rise to 7.5–8 wt% Al2O3.
The remaining five profiles differ from the others in the sense that they do not display a rimward increase in Al. Three of these profiles (two in the PPD: Appendix 1k and l, and one the intracaldera Fish Canyon Tuff; Fig. 2e⇑) show an oscillatory variation of ± 1 wt% Al2O3 around an average similar to the mean of the whole data set (~7 wt% Al2O3). In contrast, the traverse in Figure 2d⇑ (intracaldera Fish Canyon Tuff pumice) shows a higher average (~8 wt% Al2O3) with Al-spikes reaching more than 9 wt% Al2O3. Finally, the profile across the hornblende included in the granodioritic xenolith (Fig. 2f⇑) has a lower average (~6 wt% Al2O3) and it shows a decrease in Al2O3 in the outermost rim of the crystal.
Elemental correlation diagrams (Fig. 4⇑) demonstrate that chemical variations in Fish Canyon amphiboles are dominated by temperature-sensitive substitutions (edenite + Ti-Tschermak) and show minimal evidence for the pressure-sensitive Al-Tschermak exchange. Core-to-rim microprobe profiles can thus be considered as proxies for the temporal thermal evolution of the magma chamber. The TAl, M1–M3Ti and A(Na + K) rimward increases displayed by the majority of profiles in the two major units of the Fish Canyon magmatic system (Pagosa Peak Dacite and Fish Canyon Tuff) suggest a near-isobaric progressive reheating of the Fish Canyon magma shortly prior to eruption, consistent with the conclusion reached in Bachmann et al. (2002) on the basis of multiple independent petrological and geochemical lines of evidence. The ± 0.3–1 wt% Al2O3 oscillatory variations observed in most profiles are interpreted as an indication of phenocryst transport through thermal (± compositional) gradients, rather than reflecting oscillatory zoning produced by near-equilibrium, diffusion-controlled incremental growth, as generally envisaged for fine-scale (1–10 μm) and small amplitude oscillatory zoning in plagioclase (e.g., Pearce 1994; Singer et al. 1995).
Despite the uniform whole-rock and mineral compositions in the various stratigraphic units of the Fish Canyon magmatic system, differences in the zoning profiles imply that: (1) thermal and compositional gradients existed in the Fish Canyon chamber; and (2) individual phenocrysts experienced different growth histories. Phenocrysts with distinct patterns are found next to each other in the same thin sections, requiring that they have been juxtaposed sufficiently close to the time of eruption, such that at the pre-eruption temperature, there was no significant re-equilibration of hornblende compositions. Juxtaposition of phenocrysts with different growth histories shortly prior to eruption has been described by Devine et al. (1998a), Murphy et al. (2000), and Couch et al. (2001) in the andesite of the ongoing Soufrière Hills eruption on the Island of Montserrat, which is also interpreted as a rejuvenated crystal mush following the intrusion of more-mafic magma. To explain this observation, Couch et al. (2001) proposed that the hot mafic magma, intruded at the base of the chamber, triggered partial remelting of the region immediately above the contact. The newly formed hot silicic layer became buoyant and mixed with the overlying cooler interior. A similar model is envisaged for the Fish Canyon magmatic system, but on a much larger scale. The fact that amphiboles from the Montserrat Andesite do not appear to be zoned, but developed reactions rims (Devine et al. 1998b), might be a consequence of the much smaller size of the system, allowing a larger and faster temperature change, driving the magma out of the amphibole stability field.
Profiles that do not show a progressive rimward Al-increase have average compositions close to the mean of the whole data set (Fig. 2e⇑ and Appendix 1k and l). Such patterns can be reconciled with crystallization in regions of the chamber in which temperature did not significantly increase over the crystallization interval. However, the ± 1 wt% Al2O3 oscillatory variations suggest that growing crystals were transported through small thermal/compositional gradients by convective motions. The amphibole profiled in Figure 2d⇑, with its higher average (~8wt% Al2O3) and high-amplitude Al-spikes, may have crystallized in proximity to the base of the silicic magma chamber. As this crystal is present in a sample from the top of the intracaldera tuff facies, considered to represent the last-erupted part of the chamber, growth in a hotter boundary layer close to the mafic-silicic contact may be envisaged. Conversely, the profile across the hornblende of the holocrystalline granodioritic xenolith (Fig. 2f⇑), with its lower (~6 wt% Al2O3) average and rimward Al-decrease, is consistent with crystallization in a colder part of the chamber, which did not record any late reheating.
This interpretation of temperature-induced chemical zoning in hornblende explicitly includes the counter-intuitive deduction that hornblende was crystallizing from a highly evolved, high-SiO2 rhyolite melt as the temperature of the system was gradually increasing. The same conclusion applies to the reversely zoned plagioclase phenocryts. In a closed system, this would be implausible, as the rhyolite melt is nearly devoid of mafic components, and crystallization while a system is being heated appears paradoxical. However, mafic underplating is inferred to be at the origin of the reheating and rejuvenation event that affected the Fish Canyon magma shortly prior to eruption (Bachmann et al. 2002). A voluminous (~200 km3) sequence of volatile-rich andesitic lava flows and associated breccias (the Huerto Andesite; Askren et al. 1991; Lipman 2000; Parat 2001) was emplaced from centers within and adjacent to the La Garita caldera shortly (<0.1 Ma) after the Fish Canyon Tuff eruption (Lipman 2000), and sparse hybrid andesite inclusions are found in the upper intracaldera Fish Canyon Tuff (Bachmann et al. 2002). In this situation, mass additions with higher proportions of H2O, anorthite, and ferromagnesian components, may have been supplied either by the underlying volatile-rich mafic magma body (i.e., Litvinovsky and Podladchikov 1993; Robinson and Miller 1999; Bindeman and Davis 1999), or by extensive melting of Fish Canyon crystal mush (including mafic phases) in the thermal boundary layer immediately above the contact with the mafic magma. Efficient precipitation of the added mafic components might have been promoted by: (1) the presence of pre-existing hornblendes, which provided favorable nucleation sites; and (2) the structure of the high-SiO2 rhyolite matrix liquid, which consisted almost entirely of network-forming components and would have tended to purge itself of most of the added network-modifying elements very quickly. In support of some dilute mass addition from below, independent arguments (high Cl in apatite, progressive inverse zoning in plagioclase, rimward F-increases in biotite) suggest that volatile partial pressures were increasing prior to eruption as the underlying mafic magma was degassing (Bachmann et al. 2002). As the temperature fluctuations recorded by hornblende zoning were limited (~50 °C), and as the highest equilibration temperature reached by the system during rejuvenation was below 800 °C, hornblende and plagioclase would have been stable according to the experimental results of Johnson and Rutherford (1989a), despite increasing temperature.
An alternative scenario to explain the observed zoning patterns without invoking regrowth of hornblende is diffusive reequilibration in the outer part of the hornblende. However, this scenario appears untenable, as fine-scale oscillatory zoning has survived in these hornblendes, and it would require that elements such as Fe and Ti diffused strongly uphill from the melt to the crystal.
Microprobe traverses in euhedral and poikilitic hornblendes of the Fish Canyon magma recorded significant rimward increases in TAl (~0.8 apfu), coupled with increases in Ti, Na, and K that were mainly balanced by decreasing Si and Mn. These chemical variations are mainly a consequence of the edenite substitution, coupled with a Ti-Tschermak exchange, both of which are demonstrated by experimental results to be temperature-sensitive. The pressure-sensitive Al-Tschermak substitution (e.g., Johnson and Rutherford 1989b) does not play a significant role in the Al-zoning variations. The profiles can be reconciled with near-isobaric, gradual reheating of the whole Fish Canyon magma, accompanied by some mass addition (mainly volatiles), and are thus consistent with the model of pre-eruptive rejuvenation and remobilization of a crystal-rich, near-solidus mush upon intrusion of mafic magma (Bachmann et al. 2002). Fe-Mg exchange also occurred in these hornblendes, but probably as an indirect consequence of the temperature-sensitive substitutions. We speculate that the addition of Ti in the M site creates an Mg-avoidance effect. The presence of a large number of phases and the phenocryst-rich nature of the Fish Canyon magma, as well as the differences between mean temperatures for the low- and high-Al populations rendered by the HB1994TB and BH1990 algorithms, suggest that the temperature interval over which these hornblendes crystallized was ≤50 °C. The significant impact that this relatively small temperature variation has on the Al-content of hornblende and the results obtained with the AS1995 barometric formulation with the low- and high-Al population demonstrate that Altot, predominantly composed of TAl, might be too sensitive to temperature to yield accurate pressures at low Al contents. Whereas the results obtained with AS1995, HB1994TB, and BH1990 are encouraging, when both pressure and temperature are known independently, it is clear that calibrations of both barometric and thermometric formulations must carefully consider all the substitutions that are involved. At these low temperatures and pressures, it appears that the Fe-Mg exchange is extremely sensitive to other substitutions in amphibole, as major variations in coexisting liquid seem unlikely to be responsible in this case. This dependence has implications for interpretation of magma evolution trends derived from amphibole zoning profiles, as well as offering additional leverage during future refinements of barometric and thermometric formulations.
Electronic appendix figure captions
Appendix Figure 1. All microprobe profiles in hornblendes from all the different stratigraphic units of the Fish Canyon magmatic system (see Fig. 2⇑ for a subset of six profiles).
Appendix Figure 2. Frequency distribution diagram of pressure calculated for the low- and high-Al populations using the equation from AS1995. For the low-Al population, pressures were calculated for both 700 and 760 °C, considered to be the two end-member temperatures of the range over which the Fish Canyon hornblende crystallized. Also reported is the estimate of mineral equilibration pressure of Johnson and Rutherford (1989a; 2.4 ± 0.5 kbar). 2: TAl vs. A(Na + K) plots for each traverse across hornblende reported in Appendix 1, designed to estimate the contribution of the edenite substitution in the total TAl variation. The slopes of the regression lines consistently around 0.5 illustrate that the edenite exchange accounts for half of the total TAl variation for all analyzed phenocrysts.
Appendix Figure 3. Frequency distribution diagrams of temperature calculated for the low-Al and high-Al populations designed to show the influence of the input parameters P and XAb in the HB1994TB (left), HB1994TA (right) and BH1990 (bottom) calibrations of the Al-in-hornblende thermometer.
Appendix Figure 4. Frequency distribution diagrams of pressure calculated for the low- and high-Al populations at different temperature using the equation of AS1995. These diagrams demonstrate how sensitive the algorithm is to small changes in temperature.
Appendix Figure 5. Major element analyses and structural formulae (13eCNK) of the 2146 electron microprobe analyses of the data set.
To view appendix figures, go to http://www.minsocam.org and find the link in the August/September 2002 table of contents. Or contact the MSA Business Office to purchase copies of deposit number AM-02-014. See inside front cover for contact information.
Note added in proof
Pressure-temperature diagram of barometric reaction 1 proposed by Ague (1997; Tremolite + Phlogopite + 2Anorthite + 2Albite = 2Pargasite + 6Quartz + K-feldspar) using the mineral compositions of Fish Canyon phenocrysts and the TWQ software (Berman 1991). In order to derive a pressure estimate for both the core and rim populations of hornblende in the Fish Canyon magma, the TWQ calculations were run once for the core compositions, and once for the rim compositions. Pressure was then estimated at the temperatures calculated in this paper. For amphibole and plagioclase, rim and core compositions used in this paper were applied (Pl core: XAn = 0.28, XAb = 0.66, XOr = 0.06; Pl rim: XAn = 0.33, XAb = 0.63, XOr = 0.04; Hbl core: average of the low-Al population listed in Table 1⇑; Hbl rim: average of the high-Al population listed in Table 1⇑). As both sanidine and biotite do not show any significant zoning from core to rim, the same representative composition was taken for both core and rim calculations (Bt: average Pagosa Peak Dacite composition listed in Table 3 of Bachmann et al. (2002); Kfs: sanidine composition of point 1 from microprobe traverse illustrated in Fig. 14b of Bachmann et al. (2002), and listed in Table 5). Refer to the TWQ manual (Berman 1991) for the schemes used to recalculate structural formulae.
The pressure estimate using this reaction (2.45 ± 0.05 kbar) is consistent with the value of 2.4 ± 0.5 kbar published by Johnson and Rutherford (1989), based on experimental duplication of the Fish Canyon mineral assemblage and compositions. The pressure value obtained using the Anderson and Smith (1995) calibration of the Al-in-hornblende barometer (this paper) for the rim population at 760 °C is comparable as well (2.24 ± 0.6 kbar), but the core population at 715 °C gives a significantly lower pressure estimate (1.52 ± 0.6 kbar), suggesting that the temperature correction is underestimated.
The nearly identical pressure estimates obtained for the rim and core populations using this reaction strongly support the inference that the aluminum zoning in the Fish Canyon hornblendes is a consequence of near-isobaric temperature (+fH2O) variations.
This contribution is one out of a series of papers on the Fish Canyon magmatic system, and none of them would have been possible without the collaboration of Peter Lipman. We are deeply indebted to his insightful contribution and assistance all along this project, whether in the field, in the lab, or around a Château Margaux. The help of François Bussy (University of Lausanne) in the electron microprobe laboratory is acknowledged. Thanks as well to Ilya Bindeman for doing the oxygen isotope analyses. This project was supported by the Swiss NSF grant #20-49730.96 to Dungan and by a bourse chercheur débutant of the Swiss NSF to Olivier Bachmann. Detailed and constructive reviews from Bernard Evans, Lawford Anderson, Malcom Rutherford, and Jon Blundy are greatly appreciated.
Manuscript handled by Wendy Bohrson
- Manuscript Received September 17, 2001.
- Manuscript Accepted March 28, 2002.