Review of Santa María geochemistry

The major and trace element compositions and Sr, Nd and Pb isotope ratios from 27 basaltic andesite lava flows that comprise the Santa María composite cone were presented by Singer et al. (2011). In order to emphasize the profound compositional and temporal bimodality of the entire Santa María–Santiaguito complex, Singer et al. (2011, figs 5 & 6) compared these 27 mafic cone-forming lavas to the historically-erupted dacite tephra and lavas in variation diagrams. The major and trace element compositions from whole rock samples of the historically-erupted materials described in the previous section, including pumice, scoria and lithic fragments collected from the 1902 airfall deposit, and from each of the major phases of the Santiaguito dome complex, plus quenched mafic inclusions in two of the dacite lava domes, are presented in this paper. From a subset of these whole rock samples we also present new Sr, Nd and Pb isotope ratios and δ¹⁸O(WR) values. Jicha et al. (2010) present whole rock ²³⁸U–²³⁰Th isotope compositions from Santa María cone-forming lavas, the 1902 dacite pumice and basaltic andesite scoria, and several Santiaguito dacite dome lavas. Here we report new ²³⁸U–²³⁰Th isotope compositions from minerals and glass separated from the 1902 dacite pumice and from a dacitic lava that erupted in 1972 to form part of the El Monje dome in the Santiaguito dome complex.

Whole rock major and trace element analyses

Twenty-nine whole rock samples (including three glass separates from 1902 pumice) of 250–500 g each were crushed (tungsten-carbide piston crusher), powdered (aluminum oxide ceramic shatterbox and puck), and analysed for major (X-ray fluorescence) and trace elements (inductively-coupled plasma mass spectrometry, ICP-MS) at the Open University, UK following methods of Rogers et al. (2006). Details on procedures including precision and accuracy are provided in Singer et al. (2011).

Pb, Sr and Nd isotopes

A representative subset of the whole rock samples were analysed for ⁸⁷Sr/⁸⁶Sr (18 samples), ²⁰⁸Pb/²⁰⁴Pb, ²⁰⁷Pb/²⁰⁴Pb, ²⁰⁶Pb/²⁰⁴Pb (14 samples) and ¹⁴³Nd/¹⁴⁴Nd (17 samples) at the UW-Madison Radiogenic Isotope Laboratory via thermal ionization mass spectrometry (TIMS, Micromass Sector 54 instrument). Analytical procedures follow those described in Singer et al. (2011). For Pb isotope analyses, instrumental mass bias was corrected empirically based on analyses of NIST SRM-981 (n=10) and SRM-982 (n=11) run under similar conditions as samples. The pooled average mass fractionation correction based on the measured ²⁰⁷Pb/²⁰⁶Pb analysis of SRM-981 and the measured ²⁰⁸Pb/²⁰⁶Pb ratio of SRM-982 was 0.00148±0.00045 per amu. Strontium and neodymium isotope measurements were exponentially corrected for mass fractionation using ⁸⁶Sr/⁸⁸Sr=0.1194 and ¹⁴⁶Nd/¹⁴⁴Nd=0.7219, respectively. Repeat analysis of NIST SRM-981 yielded an average ⁸⁷Sr/⁸⁶Sr ratio of 0.710267±0.000018 (2-SD; n=30). Duplicate samples of the USGS rock standard BCR-1 were processed through the entire analytical procedure and the measured ⁸⁷Sr/⁸⁶Sr ratios were 0.705021 and 0.705028. The measured ¹⁴³Nd/¹⁴⁴Nd for two in-house Nd standards (Ames I and Ames II) are 0.512149±0.000014 (n=15; 2-SD) and 0.511974±0.000016 (n=15; 2-SD), respectively. The measured ¹⁴³Nd/¹⁴⁴Nd for La Jolla Nd was 0.511844±0.000015 (n=6; 2-SD). Duplicate samples of the USGS rock standard BCR-1 measured ¹⁴³Nd/¹⁴⁴Nd ratios were 0.512920±0.000010 and 0.512908±0.000008 (errors are 2-SE in-run statistics).

U and Th isotopes

In addition to two whole rock samples of dioritic lithic fragments in the 1902 airfall deposit, minerals and glass separated from 1902 dacite pumice sample SM-07-09B and dacite lava sample SG-05-01 from the Mitad dome, were analysed for U–Th isotope composition. Solution-based U and Th isotope measurements were obtained using a Micromass Isoprobe multi-collector ICP-MS at UW-Madison’s Radiogenic Isotope Laboratory using procedures of Jicha et al. (2009). Rock standards including AThO, BCR-1, and AGV-1 were analysed along with variable concentration IRMM-035 and IRMM-036 solutions to monitor accuracy, reproducibility and external precision (details in Singer et al. 2011). The mean ²³²Th/²³⁰Th values for IRMM 035 (87 799±1006; 2-SD; n=15) and IRMM 036 (326 300±2126; 2-SD; n=8) are indistinguishable from consensus values (Sims et al. 2008).

O isotopes

Whole rock powders were analysed at the University of Wisconsin-Madison by laser fluorination (Valley et al. 1995) using an air-lock sample chamber (Spicuzza et al. 1998) with bromine pentafluoride as the reagent and a 25W CO₂ laser. Isotope ratios were measured using a Finnigan MAT 251 dual inlet mass spectrometer. Whole rock δ¹⁸O values were standardized using standard UWG-2 garnet (accepted δ¹⁸O=5.80‰ VSMOW; Valley et al. 1995). Eight standards run on each day of analysis (three days total) yield a reproducibility of <±0.16‰ (2 std. dev.). The average difference in δ¹⁸O value between duplicate analyses of whole rock powders was <0.08‰.

Electron probe microanalysis of phenocryst minerals

Epoxy mounts were prepared with separates of plagioclase, iron–titanium oxides, pyroxenes and amphibole phenocrysts from 1902 dacite pumice sample SM-07-09B and scoria sample SM-7-09E. Phenocrysts and their melt inclusions were imaged in back-scattered electron (BSE) mode on the Hitachi S-3400N Variable Pressure Scanning Electron Microscope (VP-SEM) at the University of Wisconsin-Madison. Energy Dispersive Spectrometry (EDS) located grains of specific minerals (e.g. orthopyroxene v. clinopyroxene or ilmenite v. titanomagnetite) which then were analysed by Wavelength Dispersive Spectrometry (WDS) on the Cameca SX51 5-spectrometer electron microprobe at UW-Madison. Major element compositions of minerals were obtained using a 1 μm diameter beam with an accelerating potential of 15 keV. Natural and synthetic standards of olivine, pyroxene, plagioclase, amphibole and iron–titanium oxides were used for calibration. Automation and data reductions used Probe for Windows software with the matrix correction of Armstrong (1988). In anticipation of element migration under a focused electron beam (Morgan & London 1996), analyses of plagioclase-hosted melt inclusions employed time-integrated intensities collected at 2 s intervals for 20 s, allowing Na, K and Si concentrations in the glass to be extrapolated to a time-zero value. To obtain a large number of anorthite content profiles of plagioclase phenocrysts, we used greyscale values from the BSE images as a proxy for An content generally following the procedures outlined by Ginibre et al. (2002) and Triebold et al. (2006). The An profiles were calibrated via WDS analyses of 10 points along each profile.

Secondary Ion Mass Spectrometry (SIMS) analysis of trace elements in plagioclase

Polished gold-coated epoxy mounts of plagioclase crystals were analysed for K, Mg, Ti, Fe, Sr and Ba using the Cameca ims-1280 ion microprobe at the WiscSIMS Laboratory, UW-Madison. The analysis conditions are after Kita et al. (2004) using high mass resolution mode (mass resolving power of 5000) without energy filtering. The primary beam of O⁻ ions with 1 nA intensity was focused to a spot of c. 10 µm in diameter at the sample surface. For each element, secondary ion intensity was normalized to the intensity of the ³⁰Si⁺ ions. A natural plagioclase standard with known trace element concentrations (Lab1, Kita et al. 2004) was analysed multiple times in the same analytical session, from which relative sensitivity factors of each element were determined. Reproducibility of ratios between respective ions and ³⁰Si⁺ of repeated analyses of plagioclase standard was better than 3% (1σ).

SIMS analysis of plagioclase-hosted melt inclusions

In polished, gold-coated epoxy mounts 2 mm thick, melt inclusions ranging from 30–100 μm across in plagioclase from 1902 dacite pumice sample SM-07-09B were analysed for hydrogen using the Cameca IMS-1280 ion microprobe at the WiscSIMS Laboratory, UW-Madison. A primary beam of Cs⁺ (20 keV impact energy and 0.6 nA beam current) was focused to a spot of ∼8 µm at the sample surface and a normal incidence electron gun was used for charge compensation. General instrument parameters are similar to that of oxygen isotope analysis (Kita et al. 2004). The negative secondary ions of H⁻ and ³⁰Si⁻ were measured using the axial electron multiplier in single collection mode. In order to precisely aim at these melt inclusions, we obtained secondary H⁻ ion images of the sample surface by rastering the primary beam over a 50×50 µm areas for three minutes and identified the location of melt inclusions from a hot spot of high H⁻ count rates. Subsequently, we used a stationary primary beam centred on the melt inclusion and detected secondary H⁻ and ³⁰Si⁻ ions that were emitted from only the central 2 µm squares of the sputtered area by using a field aperture with opening of 400×400 µm (magnification of transfer optics was ×200 from sample to the field aperture). Mass resolving power was set to 3000. The secondary H⁻ ion intensities were normalized to that of the ³⁰Si⁻ ion and the H⁻/³⁰Si⁻ ratios were obtained after correction for hydrogen background as measured several times on the host plagioclase crystals between measurement of the melt inclusions.

The seven rhyolite glasses used as standards for the hydrogen measurements were produced experimentally from a sample of a natural obsidian flow, and contain between 0.1 and 6.1 wt% H₂O (Tenner et al. 2010). Electron probe major element compositions, and SIMS data from these glasses can be found in the Supplementary material. The glass standards were analysed multiple times during the analytical session to generate a calibration curve for H⁻/³⁰Si⁻ ratios (Supplementary material). Following SIMS analysis, the SiO₂ content of each analysed inclusion was measured by electron probe and the calibration curve was used to estimate the wt% H₂O with a precision of about ±3% (1σ; Supplementary material).

Major and trace elements and Sr–Nd–Pb–O isotopes

The whole rock major and trace element concentrations and Sr, Nd, and Pb isotope ratios are summarized in Table 1. The O isotope results from the whole rock samples of this study, as well as samples of the cone-forming basaltic andesite lavas measured for elemental concentrations and Sr–Nd–Pb isotope ratios by Singer et al. (2011), are summarized in Table 2. Representative plots of K₂O, MgO, Th, Sr, ⁸⁷Sr/⁸⁶Sr, ¹⁴³Nd/¹⁴⁴Nd, ²⁰⁶Pb/²⁰⁴Pb and δ¹⁸O(WR) v. SiO₂ serve to illustrate the major geochemical and isotopic features of the 1902 eruptive products and to contrast these with the preceding Santa María cone-forming lavas and the subsequently erupted Santiaguito dome lavas (Fig. 3). In the 1902 airfall deposit, dacite pumice contains 64.6 to 67.2 wt% SiO₂, whereas the basaltic andesite scoria has 53.9 to 55.9 wt% SiO₂, thus revealing the strongly bimodal nature of the 1902 magma body distinguished by a >10% gap in SiO₂. Rhyodacitic glass separated from three dacite pumice blocks contains 67.8 to 70.4 wt% SiO₂ and is richer in K₂O and Th (and other incompatible elements) and poorer in MgO and Sr on average (and other compatible elements) than the bulk dacite pumice (Fig. 3). The 1902 scoria is similar in major and trace element composition to the more SiO₂-rich of the basaltic andesite lava flows that erupted during the latest cone-forming phase of Santa María volcano between 35 and 25 ka. Lavas comprising the Santiaguito dome complex contain less SiO₂, 62.4 to 64.7 wt%, compared to the 1902 dacite pumice and the quenched inclusions we have measured from two of these flows are remarkably similar to the 1902 scoria with respect to major and trace element composition (Fig. 3). The six fine- to medium-grained diorite to granodiorite lithic fragments we chose to analyse from the 1902 deposit range from 48 to 60% SiO₂ and, with the exception of MgO, are not collinear with the basaltic andesitic cone-lavas of Santa María volcano, instead being scattered with no simple relationship to the volcanic rocks. These results are consistent with, and reinforce, the observations of Rose (1987a) regarding the bimodality of the volcano, and the similarity among mafic and silicic magmas over time. However, owing to vast improvements in analytical precision during the last three decades, and a larger number of measurements than obtained by Rose et al. (1977), the Sr, Nd and Pb isotope results we have obtained reveal several previously unappreciated features of this system.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Variation of selected major elements K₂O and MgO (in wt%), trace elements (in ppm) and Sr, Nd and Pb isotope ratios, and δ¹⁸O(WR) values with wt% SiO₂. Diorite and granodiorite fragments from the 1902 airfall deposit are denoted as lithic fragments; quenched mafic inclusions found in Santiaguito dome dacite lavas are denoted as QMI. The major and trace element and radiogenic isotope ratios for the early and late mafic cone lavas (51–55% SiO₂) are from Singer et al. (2011).

The ⁸⁷Sr/⁸⁶Sr ratios of 0.703848 and 0.703880 obtained from two blocks of 1902 dacite pumice are similar to that of 0.703837 in the glass separated from a third block; however, these ratios are significantly lower than ratios that range from 0.703897 to 0.703950 obtained from four scoria lapilli from the same eruption (Fig. 3). The Santiaguito dacite lavas (and quenched mafic inclusions) erupted in 1939 from the Mitad dome, 1972 from El Brujo, 1972 and 2002 from Caliente, have ⁸⁷Sr/⁸⁶Sr ratios similar to those of the 1902 dacite pumice; however, lavas from the eruptions in 1950 of El Monje, and 1973 of El Brujo are significantly higher at 0.703905 and 0.703977, respectively (Fig. 3). Notably, the entire range of ⁸⁷Sr/⁸⁶Sr ratios found in the rhyodacite glass, dacite pumice and dacite lava flows erupted in 1902 and afterward is comparable to that of the basaltic andesite lavas that erupted 35–25 ka during the late phase of Santa María cone growth (Fig. 3). With one exception, the diorite–granodiorite lithic fragments span this same range. For the most part ¹⁴³Nd/¹⁴⁴Nd ratios mirror those of ⁸⁷Sr/⁸⁶Sr. The 1902 rhyodacite glass has a higher ¹⁴³Nd/¹⁴⁴Nd ratio than the dacite pumice, whereas pumice has a similar ratio to three of the four scoria samples. The 1902 scoria and quenched basaltic andesite inclusion from the 1939 La Mitad dome lava span a range of ¹⁴³Nd/¹⁴⁴Nd ratios comparable to those of the basaltic andesitic lavas that erupted 35–25 ka during the late phase of Santa María cone growth. The lithic fragments have higher ¹⁴³Nd/¹⁴⁴Nd ratios than any of the volcanic rocks (Fig. 3). The ²⁰⁶Pb/²⁰⁴Pb ratios of 18.70–18.72 for the 1902 dacite pumice and rhyodacite glass separated from it are slightly higher on average than those of the two scoria samples, 18.68–18.70, and as seen for the other isotopes, all the historically-erupted materials fall within the range spanned by the latest cone-forming basaltic andesite lavas erupted 35–25 ka on Volcán de Santa María (Fig. 3). The lithic fragments have ²⁰⁶Pb/²⁰⁴Pb ratios that fall at the higher end of the range defined by the volcanic rocks.

The δ¹⁸O(WR) values of 6.25 to 6.42‰ for most of the cone-forming basaltic andesite lavas correlate positively with SiO₂, whereas the youngest of these lavas from the summit of the cone that is ⁴⁰Ar/³⁹Ar-dated at ∼35 ka (Escobar-Wolf et al. 2010; Singer et al. 2011) has a much higher value of 6.95‰ (Fig. 3). In the 1902 deposit the basaltic andesite scoria has a δ¹⁸O of 6.54‰ and the dacite pumice is 6.95‰; the latter is identical to that of the youngest cone lava. The Santiaguito dacite dome lavas have δ¹⁸O between 6.55 and 6.82‰, whereas the quenched basaltic andesite inclusions in two of these lavas are 6.50 and 6.60‰ (Fig. 3), similar to the 1902 scoria. The δ¹⁸O values are inversely correlated with ⁸⁷Sr/⁸⁶Sr such that the early cone lavas have lower δ¹⁸O and higher ⁸⁷Sr/⁸⁶Sr than the youngest cone lavas and the dacites that erupted long after cone growth ceased (Fig. 4).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Plot of δ¹⁸O(WR) v.⁸⁷Sr/⁸⁶Sr for the Santa María cone lavas, 1902 dacite pumice and basaltic andesite scoria and Santiaguito dacite dome lavas and their quenched mafic inclusions (QMI). The lower panel shows these data relative to the range in mid-ocean ridge basalt (MORB; from data in Eiler 2001; Hofmann & Hart 1978; Cohen & O’Nions 1982).

U–Th isotopes

Whole rock (²³⁸U/²³²Th) and (²³⁰Th/²³²Th) activity ratios of nine historically erupted dacite pumice, scoria, dacite dome lavas and their inclusions obtained by Jicha et al. (2010) are summarized in Table 3 to provide context for mineral data discussed below. These activity ratios, plus those obtained from 13 basaltic andesite lavas erupted on Santa María volcano between 72 and 25 ka (each corrected for time since eruption using the ⁴⁰Ar/³⁹Ar ages; Singer et al. 2011), are all in U-excess by between 3 and 26% (Fig. 5). The older cone lavas are on average further from the equiline, lower in the (²³⁰Th/²³²Th) activity ratio, and thus exhibit a greater degree of U-excess than most of the younger cone lavas and the historically-erupted materials. The 1902 dacite pumice and five Santiaguito dacite dome lavas have remarkably similar (²³⁸U/²³²Th) and (²³⁰Th/²³²Th) activity ratios of about 1.26 and 1.16, respectively, whereas the 1902 scoria and basaltic andesite inclusions from two of the dome lavas have slightly lower (²³⁰Th/²³²Th) activity ratios of 1.14. Notably the two magmas that plot nearest to the equiline with the lowest U-excesses include those represented by the 1902 scoria, and one of the youngest basaltic andesite lavas of the Santa María cone (sample SM-07-01 of Singer et al. 2011) that erupted from a vent near the summit of the volcano at ∼35 ka (Fig. 5).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

Plot of (²³⁸U/²³²Th) and (²³⁰Th/²³²Th) activity ratios for basaltic andesite cone forming lavas of Santa María volcano, 1902 dacite pumice and basaltic andesite scoria, and Santiaguito dacite dome lavas and their quenched mafic inclusions (QMI). The dashed arrows indicate 10, 25 and 50 kyr of closed-system ingrowth of ²³⁰Th via radioactive decay.

The (²³⁸U/²³²Th) and (²³⁰Th/²³²Th) activity ratios of glass, magnetite, orthopyroxene, and amphibole from the 1902 dacite pumice and from the 1972 Caliente dacite dome lava are reported in Table 3. The activity ratios of glass, magnetite and orthopyroxene in the 1902 dacite pumice yield an isochron of 1.9±1.2 ka (Fig. 6a), whereas glass, whole rock, magnetite and amphibole from the 1972 Caliente dome lava define an isochron of 9.5±2.5 ka (Fig. 6b). Note that the whole rock samples of the 1902 dacite pumice and mafic scoria plot below the 1.9 ka isochron in such a way as to be consistent with mixing of the mafic and silicic melts as discussed below. We interpret these internal isochrons as estimates of the time since crystallization began in the two dacite magmas (e.g. Condomines et al. 2003; Jicha et al. 2005).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 6.

Plots of (²³⁸U/²³²Th) and (²³⁰Th/²³²Th) activity ratios for: (a) phases from the 1902 airfall deposit including whole rock (wr) dacite pumice and glass, magnetite (mag) and orthopyroxene (opx) separated from the pumice, and whole rock scoria. The glass, magnetite and orthopyroxene yield an internal isochron of 1.9±1.2 ka; (b) phases from the 1972 AD Caliente dacite flow, including whole rock, glass, magnetite and amphibole. These phases define an internal isochron of 9.5±2.5 ka.

When acquiring data from basaltic andesite lavas of the Santa María cone, we discovered that ⁸⁷Sr/⁸⁶Sr ratios are inversely correlated with the activity ratio (²³⁸U/²³⁰Th) which is a measure of U-excess in the magmas prior to eruption (Jicha et al. 2010; Singer et al. 2011). Moreover, we observe that over the last ∼75 kyr ⁸⁷Sr/⁸⁶Sr ratios have declined, and δ¹⁸O values increased as the magmas evolved in composition such that they appear to be closer to a state of secular equilibrium in the U–Th isotope system. With the exception of the dacite lava from the 1973 El Brujo flow, the historically-erupted 1902 dacite pumice (and glass) and other Santiaguito dacite lavas erupted between 1939 and 1972 plot in the group with the smallest U-excesses, lowest ⁸⁷Sr/⁸⁶Sr ratios, and highest δ¹⁸O values (Fig. 6). Basaltic andesite magmas that also fall in this group are represented by the 1902 scoria and one of the youngest cone-forming lavas on Santa María that erupted between 35 and 25 ka.

Oxide and pyroxene composition, oxygen fugacity and thermometry

Titanomagnetite and rare ilmenite in the 1902 dacite pumice and basaltic andesite scoria typically occur as small (<30 μm) inclusions within phenocrysts of amphibole and plagioclase. Like Fe–Ti-oxides from some other well-studied volcanoes (Santorini, Mt St Helens, Mt Pelée), ilmenite and titanomagnetite in the 1902 ejecta are relatively homogenous and show none of the zoning observed at the Soufrière Hills Volcano of Montserrat or Mt Unzen of Japan (Nakada & Motomura 1999; Venezky & Rutherford 1999; Pichavant et al. 2002; Devine et al. 2003). Titanomagnetite–ilmenite pairs in plagioclase or amphibole host crystals that have log(Mg/Mn) satisfying the equilibrium criteria of Bacon & Hirschmann (1988) were used for estimating temperature and f O₂ (Table 4). Hematite contents in ilmenite above 30 wt% led us to use the algorithm of Ghiorso & Sack (1991) to estimate equilibrium temperature and f O₂. Nine oxide pairs from the dacite pumice yield tightly clustered temperatures between 860 and 885 °C and log f O₂ values between −10.2 and −10.6, whereas four pairs in the basaltic andesite scoria reveal a wider range of temperatures between 925 and 1040 °C with log f O₂ values of −8.0 and −9.5. Together, these T–f O₂ estimates form an array comparable to the Ni–Ni–O+2 buffer curve and are thus similar to dacite from Shiveluch and Pinatubo volcanoes, but significantly more oxidized than dacite from Mt St Helens (Hattori 1993; Evans & Scaillet 1997; Blundy et al. 2006) and the Valley of 10 000 Smokes (Hildreth 1983) (Fig. 7).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 7.

Calculated equilibration temperature and oxygen fugacity (f O₂) for Fe–Ti oxide pairs. Each pair of grains is located within a single host phenocryst in the 1902 dacite pumice or scoria. The range of temperature estimates for co-existing pyroxenes in the 1902 dacite and scoria are indicated (representative data from Table 5). For comparison, Fe–Ti results from Pinatubo (Hattori 1993), Mt St Helens, Shiveluch (Blundy et al. 2006) and the Valley of Ten Thousand Smokes (Hildreth 1983) are illustrated.

Pyroxene phenocrysts in 1902 dacite pumice include augite (En_34–44 Fs_13–20 Wo_37–48) and hypersthene (En_34–44 Fs_27–37 Wo_0.2–2.3), whereas the basaltic andesite scoria contains augite (En_38–46 Fs_10–19 Wo_39–46) and bronzite–hypersthene (En_59–80 Fs_20–39 Wo_0.25–4.0); zoning is subtle, <2 mol%, in both minerals (representative analyses in Table 5). As a check on the T–f O₂ estimates from the oxides, temperatures of clinopyroxene and orthopyroxene pairs were estimated using two calibrations of experimental data–one that includes clinopyroxenes with Mg# >0.75 and one that excludes them (Putirka 2008, equations 36 and 37, respectively). Compositions from pyroxene pairs that did not have K_D (Fe–Mg) values within three standard deviations of the equilibrium value of 1.09±0.014 were not used to calculate temperatures, and if the two calibrations were within the ±30 °C uncertainty of one another, the average temperature was taken as representative (Table 5). The two-pyroxene temperatures are consistent with those estimated from the two-oxide pairs, suggesting that pyroxene crystallized between 750 and 870 °C in the dacite and between 890 and 1000 °C in the basaltic andesite (Fig. 7). The range of two-pyroxene temperatures to values lower than those recorded by the oxides likely reflects Fe–Mg diffusion and partial re-equilibration during magma ascent that did not affect the iron–titanium oxide pairs because they were encased within larger host phenocrysts. Using the two-oxide data, we take the temperature of the dacite, 880 °C, and 1020 °C of the scoria (using the three highest temperature pairs in Table 4) as our best estimates of pre-eruptive storage conditions for the compositionally zoned magma.

Plagioclase zoning and melt inclusions in 1902 dacite and scoria

The 450 plagioclase crystals imaged and analysed by EMPA display several patterns of zoning, but by far the most abundant are andesine (An_44–36) that include one or more irregularly sculpted resorption surfaces characterized by an abrupt rimward increase of 1020 mol% anorthite. About 80% of crystals in the dacite pumice and 70% in the basaltic andesite scoria contain more than one resorption surface, several contain as many as five of these ‘saw-tooth’ zones which are common in mixed or mingled calc-alkaline magmas (e.g. Singer et al. 1995; Humphreys et al. 2006). Trace element profiles show that about half of the anorthite ‘spikes’ are accompanied by significant changes in FeO, TiO₂, K₂O, MgO, MnO, Ba, and Sr (Fig. 8). Mantles of these phenocrysts – the outermost 10–30% of the radial distance – are not disturbed by resorption surfaces, and commonly contain low-frequency, low-amplitude oscillatory zoning, or are homogenous. About 15% of phenocrysts in the scoria are very calcic, weakly zoned (An_93–82), and free of resorption surfaces.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 8.

Representative illustrations of zoning and melt inclusions in plagioclase revealed by backscattered electron images. Core-to-rim profiles of measured anorthite content determined by wavelength dispersive electron probe microanalysis, and FeO, MgO, Ba and Sr contents determined by secondary ion mass spectrometry. The box in the lower left panel with a backscattered electron image is enlarged above to show four melt inclusions typical of those measured for H₂O contents. See text for discussion.

Approximately one-third of plagioclase crystals in the 1902 dacite pumice and one-fifth of those in the scoria contain optically clear melt inclusions. The anorthite variations and oscillatory zoning in the crystals that host these melt inclusions are similar, regardless of whether they are found in the dacite or the scoria. The melt inclusions are typically ovoid, up to 50 μm across, are commonly located along resorption surfaces, but can also be found in the cores and, less commonly, the mantles of crystals (Fig. 8). The inclusions range from dacite to high silica rhyolite with between 69 and 79 wt%, SiO₂ (anhydrous) and 1.5–3.5% K₂O (Fig. 9a). Interestingly, the melt inclusions in the dacite pumice overlap the composition of the rhyodacitic glass separated from the bulk pumice, whereas those in plagioclase from the scoria extend to significantly higher SiO₂ and K₂O contents (Fig. 9a).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 9.

Compositions of plagioclase-hosted melt inclusions. (a) K₂O v. SiO₂ plot shows that inclusions in 1902 dacite pumice overlap the composition of bulk glass separated from pumice blocks and that together inclusions in the pumice and scoria span a range of composition similar to that of phenocryst-hosted melt inclusions from Mt St Helens and Shiveluch dacite and andesite (data from Blundy et al. 2006). Inclusion compositions obtained by electron probe microanalysis, dacite pumice and glass by X-ray fluorescence. (b) H₂O (wt%) v. K₂O plot for 20 melt inclusions in 1902 dacite pumice. The vast majority of inclusions show a decline in H₂O with increasing SiO₂ comparable to the inclusions in Mt St Helens dacite (data from Blundy et al. 2006). (c) Interpretation of the variation in melt inclusion compositions (see text for details).

The H₂O contents of 20 plagioclase-hosted melt inclusions in the dacite pumice determined by SIMS are between 1.4 and 6.9 wt% (Table 6) and in general H₂O declines as K₂O increases from about 2 to 3 wt% (Fig. 9b). Given that typical intermediate composition arc magmas contain very low amounts of dissolved CO₂ (e.g. below 400 ppm; Bacon et al. 1992; Wallace & Gerlach 1994; Saito et al. 2001; Roman et al. 2006), we assume that the volatile budget in the plagioclase-hosted melt inclusions in the 1902 dacite pumice is 100% H₂O. Accordingly, the pressure at which the melt inclusions were trapped was calculated using the program VolatileCalc (Newman & Lowenstern 2002) indicating minimum values for crystallization of the plagioclase between 170 and 15 MPa, corresponding to depths of about 5–1 km, or greater (Fig. 9c).

Deep origin of dacite during volcanic repose

The decline in ⁸⁷Sr/⁸⁶Sr, rise in ¹⁴³Nd/¹⁴⁴Nd, and drop in the degree of U-excess during c. 75 kyr of mafic cone-building volcanism (Figs 3 & 5), led Jicha et al. (2010) and Singer et al. (2011) to infer that during this period several batches of mantle-derived basalt cooled, crystallized, partially melted lower crustal rocks that include a MORB-like component, and that blending of these partial melts with the basalt created hybrid basaltic andesite magmas. In contrast to Heydolph et al. (2012), who propose assimilation of mantle pyroxenite, Singer et al. (2011) interpret the ‘old’ ²⁰⁷Pb/²⁰⁴Pb ratios of the cone-forming mafic lavas to reflect fluid derived from sediment subducted into the mantle wedge. A large fluid flux beneath Santa María helps to explain the exceptionally high ²³⁸U-excesses (Fig. 5) and the strongly oxidized state of these magmas (Fig. 7). The Sr, Nd, Pb, O and U–Th isotope ratios of the historically-erupted dacite pumice and lava and basaltic andesite scoria and quenched mafic inclusions are similar to the youngest cone-forming basaltic andesite lavas that, according to Jicha et al. (2010) and Singer et al. (2011), reflect a greater extent of blending basalt with lower crustal partial melts than do earlier erupted cone lavas owing to the sustained and repeated flux of basalt that progressively heated the lower crust. Development of a MASH-like zone at the base of the crust beneath Santa María was long ago proposed by Rose (1987a) on the basis of major element and sparse trace element data that suggested replenishment of a basaltic reservoir system in which magmas fractionally crystallized to a greater extent during the last phase of cone-growth. The similarity between the major, trace element and isotopic composition of the 1902 scoria and the last-erupted cone lavas dated between 35 and 25 ka (e.g. Figs 3 & 10) strongly argues that the historically-erupted basaltic andesite underwent similar processing in the deep crust. Rose (1987a) hypothesized that as the cone grew larger, the crustal column increased in thickness, thereby retarding the ability of the hybrid basaltic andesite magma to ascend to the surface.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 10.

Trace element variation in Santa María dacites. In upper panel chondrite-normalized rare earth element (REE) contents indicate that dacites are depleted in middle to heavy REE relative to cone-forming and 1902 basaltic andesite compositions (normalizing values from Sun & McDonough 1989). The process diagrams of Dy/Yb and K/Rb v. Rb in the middle and lower panels illustrate fractional crystallization models discussed in the text.

The ⁴⁰Ar/³⁹Ar chronology of cone-building, in parallel with new trace element and isotopic data from both the composite cone (Escobar-Wolf et al. 2010; Singer et al. 2011) and now the historically-erupted dacites, enable us to further evaluate the Rose (1987a) hypothesis and document a major shift in processes that took place below the volcano. The ⁴⁰Ar/³⁹Ar geochronology indicates that the final, rapid pulse of cone-growth, fuelled by eruptions of hybrid basaltic andesite magma, ceased by 25 ka. We propose that at 25 ka the basaltic andesite magma began to pond deep within the crust, and having either exhausted the supply of readily fusible phases in the crustal wall rocks, or by becoming insulated from these crustal rocks as rinds or envelopes of partly solidified crystal-rich mush formed along the periphery of the MASH zone, or both, it cooled slowly such that fractional crystallization occurred without significant assimilation.

Modelling of four trace elements: K, Rb, Dy and Yb, using the EC-RAFC algorithm of Bohrson & Spera (2007), helps quantify key aspects of how basaltic andesite fractionally crystallized to form the 1902 dacite and illustrates that this behaviour contrasts sharply with that of the earlier period of open-system crystallization, melting and assimilation that affected cone-forming lavas. We chose one of the youngest cone-forming basaltic andesite lavas, sample SM-07-01 from the 3772 m summit of Santa María volcano (Singer et al. 2011) as a parental composition with 9681 ppm K, 22.3 ppm Rb, 3.62 ppm Dy and 2.11 ppm Yb. The liquidus temperature of this parent magma was set at 1020 °C based on two-oxide thermometry of the remarkably similar scoria in the 1902 airfall deposit (Fig. 7). The equilibration temperature of the modelled daughter magma was set at 880 °C, based also on two-oxide thermometry of the 1902 dacite pumice. A solidus temperature for hydrous hornblende dacite at lower crustal pressure was estimated at 725 °C based on experiments of Conrad et al. (1988). Heat capacities of the magma and wallrock are identical to those used by Singer et al. (2011) and recommended by Bohrson & Spera (2007) and the initial temperature of the wallrock was assumed to be 775 °C. Given these boundary conditions, the 140 °C drop in temperature from 1020 to 880 °C results in 43 wt% of the basaltic andesite parent magma crystallizing to form cumulate solids. The bulk distribution coefficients that best predict the concentrations of K, Rb, Dy and Yb in the residual melt are 0.09, 0.005, 1.75 and 1.35, respectively, consistent with fractionation of a plagioclase+hornblende+pyroxene+magnetite cumulate (Fig. 10; Davidson et al. 2007).

The O isotope data are consistent with the modelling presented by Singer et al. (2011) wherein the cone-forming basaltic andesite lavas reflect evolution of a deep crustal reservoir through both recharge with basalt and increasing extents of assimilation of a MORB-like component that, relative to the fluid-modified mantle wedge below Santa María, is characterized by lower ⁸⁷Sr/⁸⁶Sr ratios and higher δ¹⁸O values (Fig. 4).

The O isotope data are also consistent with a closed-system crystallization origin of the dacites from basaltic andesite magma left-over from the cone-building period. It is well known that extensive high temperature crystallization of calc-alkaline magmas has little effect, perhaps increasing by a few tenths permil, on δ¹⁸O of derivative magmas (Matsuhisa 1979; Singer et al. 1992). Specifically, crystal fractionation of parent magmas similar to the cone-forming lavas at Santa María with between 50 and 54% SiO₂ and δ¹⁸O values of between 6.50 and 6.95‰ could produce the historically erupted dacites. Conspicuously, the youngest cone-forming basaltic andesite lava that we have measured (sample SM-07-01 that was used as a parent magma in the fractional crystallization model) has a δ¹⁸O value identical to the 1902 dacite and thus represents a potential parent magma for this dacite (Fig. 3). If assimilation has also contributed to the origin of the dacites, the assimilant must have a δ¹⁸O(WR) value similar to that of the late cone-forming lavas.

The ²³⁸U/²³⁰Th isochron of 1.9±1.2 ka defined by glass, magnetite and orthopyroxene in the 1902 dacite pumice (Fig. 6a) implies that crystals carried to the surface in this magma began to form very shortly before the eruption. That the whole rock sample of 1902 dacite pumice falls below the isochron may reflect mixing between the basaltic andesite, which plots well below the isochron, and the bulk dacite, that lies between the basaltic andesite and the isochron (Fig. 6a). Mixing of the basaltic andesite with the dacite is recorded in detail in the zoning of plagioclase phenocrysts in both of these magmas (Fig. 9) that will be discussed below. More surprising is the 9.5±2.5 ka ²³⁸U/²³⁰Th isochron defined by glass, magnetite, amphibole and bulk material from the 1972 lava flow on the north flank of the Caliente dacite dome (Figs 2 & 6b). From this isochron, we infer that portions of the dacite magma which now comprise the Santiaguito dome complex that grew via effusive and mildly explosive activity in the decades following the explosive onset of silicic volcanism at Santa María in 1902, began to crystallize several thousand years prior to the 1902 eruption. In this scenario, deeper, occluded reaches of the dacite magma body, perhaps including crystal-rich zones near the cooler, more viscous, periphery of the silicic reservoir, had begun to solidify no later than 10 kyr prior to the Plinian eruption of 1902, but were unable to erupt until decades after the eruption in 1902 of the more silica- and H₂O-rich melt layer that capped the system.

The volume of dacite erupted explosively in 1902 is estimated at 8.5 km³ dense rock equivalent (Williams & Self 1982). If our model is correct, the minimum volume of basaltic andesite parent magma required to create the 1902 dacite is about 20 km³. Notably, this is well within the 40–67 km³ of magma estimated to have fuelled the preceding growth of the 8 km³ cone of basaltic andesite that comprises Santa María volcano, based on EC-RAFC modelling that accounts for recharging of the system with mantle-derived magma, assimilation of partially melted crust, and fractional crystallization (Fig. 10c; Singer et al. 2011). Using the thermal model of Hawkesworth et al. (2000) and a cooling time of 25 kyr suggests that a 20 km³ mass of basaltic andesite which drops 140 °C and crystallizes 43% would generate a power output of about 25 MW, which is perhaps an order of magnitude lower than power generated during the preceding period of open-system behaviour during the later period of cone growth. The correspondingly modest and protracted release of heat during 25 kyr of cooling in the deep crust may also explain why evidence that partial melting and assimilation were important in creating a large volume of dacite is absent. If the 9.5±2.5ka ²³⁸U/²³⁰Th isochron from the 1972 Caliente flow is representative of when a large portion of the dacite began to crystallize, the 25 MW estimate of power output is a maximum. Alternatively, it seems likely that portions of the dacite magma may have crystallized at various times during the past 25 kyr, depending on where–within the plexus of mushy-to-molten environments of the deep crustal MASH zone–they began to cool.

An additional clue that may bear on the duration of the closed-system fractional crystallization process is the U–Th isotope composition of the quenched basaltic andesite inclusions found in the 1972 El Brujo dacite dome lava and the 1939 La Mitad dacite dome lava (Samples SM-07-04i and -06i; Tables 1 and 3). The two inclusions have nearly identical (²³⁰Th/²³²Th) and (²³⁸U/²³²Th) activity ratios of 1.14 and 1.25, respectively, despite being erupted 33 years apart and from nearly opposite ends, about 2 km apart, of the Santiaguito dome complex (Fig. 2). Their (²³⁸U/²³²Th) activity ratio is slightly lower than that of the 1902 dacite pumice and Santiaguito dacite dome lavas, consistent with fractionation of amphibole+clinopyoxene in which D_Th>D_U and generation of a slight ²³⁸U excess is to be expected (Brenan et al. 1995). Moreover, the (²³⁰Th/²³²Th) ratio is significantly lower than the dacites by an amount equivalent to 25 kyr of closed-system ingrowth of ²³⁰Th (Fig. 5). These two inclusions are remarkably similar, not only in U–Th isotope composition, but also in major and most trace element concentrations, to the basaltic andesite lava, SM-07-01, used as a parent magma in the fractional crystallization model illustrated in Figure 10.

New fuel to the fire: the 1902 basaltic andesite and shallow processes

The Sr, Nd and Pb isotope ratios of the 1902 basaltic andesite scoria are, on average, slightly different than the average ratio of the 1902 dacite pumice or glass (Fig. 3). However, the scoria is lower in δ¹⁸O than the dacite. Moreover, the scoria has a U–Th isotope composition that is slightly lower in (²³⁰Th/²³²Th) than the 1902 dacite pumice, but significantly lower in (²³⁸U/²³²Th) than either the 1902 pumice or the basaltic andesite inclusions in the 1939 and 1972 Santiaguito dacite dome lavas (Figs 5 & 6a). Collectively, these observations indicate that the basaltic andesite scoria produced mainly at the end of the 1902 Plinian eruption is not directly related to the much larger mass of dacite erupted at the onset. Instead, we propose that this basaltic andesite represents an influx of new magma into the MASH zone that, shortly prior to the 1902 eruption, spread out beneath the dacite magma, but was prohibited from mixing across the interface by the contrasts in both density and viscosity of the two magmas. The basaltic andesite supplied heat and perhaps volatiles via diffusion across the interface to the dacite, but, as two-oxide thermometry shows, thermal equilibration, and thus extensive chemical mixing of the two magmas, did not take place (Sparks & Marshall 1986). However, the resulting increase in buoyancy, even if small, may have rapidly promoted convection within the dacite and exsolution of H₂O across the upper part of the dacite magma cap (Snyder 2000). In turn, exsolution and vesiculation could have initiated the ascent of magma that led to the Plinian eruption.

Plagioclase zoning and the compositions of melt inclusions provide additional clues about processes that attended the latest stages of evolution of the two magmas that erupted in 1902. The abrupt, large amplitude reversals in plagioclase zoning that occur rimward of resorption surfaces and are commonly accompanied by increases in FeO, MgO, and Sr (Fig. 8) in phenocrysts contained in both the 1902 dacite pumice and scoria, argue that these two magmas mingled physically and that some limited chemical mixing had also begun to take place (e.g. Singer et al. 1995; Humphreys et al. 2006). The remarkable similarity in zoning–including multiple reversals of composition–within the majority of phenocrysts in the two magmas, together with the overlapping composition of melt inclusions trapped in these phenocrysts (Fig. 9a), imply that many plagioclase crystals were exchanged back-and-forth between rhyolitic melt and basaltic andesite magma.

The co-variation of K₂O and SiO₂ in the plagioclase-hosted melt inclusions (Fig. 9a) is consistent with fractional crystallization of the 1902 dacite magma to produce rhyodacite to high-silica rhyolite melt compositions. Moreover, the drop in H₂O with increasing K₂O recorded by the vast majority of these melt inclusions is remarkably similar to the behaviour recorded by plagioclase and amphibole-hosted melt inclusions in the dacite pumice and lava erupted during May to October, 1980 at Mt St Helens (Fig. 9b; Blundy & Cashman 2005). The drop in H₂O with increasing K₂O is also paralleled by andesite pumice erupted between 2001 and 2004 at Shiveluch volcano, Kamchatka (Blundy et al. 2006). Blundy & Cashman (2005) and Blundy et al. (2006) have proposed that this co-variation reflects decompression-driven crystallization of the melt under water-saturated conditions at both Mt St Helens and Shiveluch volcanoes (Fig. 9c). The co-variation of H₂O and K₂O in the 1902 Santa María dacite strongly suggests that rapid decompression-driven crystallization of plagioclase from pressures of at least 180 MPa or >5 km depth occurred en route to the surface such that few inclusions experienced syneruptive degassing and most record evolution of residual high silica rhyolite melt (Fig. 9c). If correct, this conclusion suggests that the model of Blundy & Cashman (2005) may apply widely to dacitic magma that erupts explosively in subduction zones around the world.

The exchange of plagioclase crystals between the thermally and chemically contrasting melts led to the repeated reversals in zoning during mingling and mixing of the two magmas that occurred mainly during magma ascent. It is remarkable that the interaction of these two contrasting magmas en route to the surface – a complication that did not affect the Mt St Helens dacite (Blundy & Cashman 2005; Blundy et al. 2006) – led to complex zoning in plagioclase, but did not obliterate the signature of H₂O-saturated decompression of the silicic melts. Perhaps, more importantly, the plagioclase crystals that dominate the phenocryst modes of the 1902 dacite and scoria record mainly late-stage, shallow-level, fractional crystallization that occurred after the large body of volatile-rich dacite had accumulated during an earlier, and likely far more protracted, period of fractional crystallization deep within the crust. Thus, andesitic to dacitic magma erupted explosively at frontal arc composite volcanoes including Mt St Helens (Berlo et al. 2007), Shiveluch (Blundy et al. 2006; Humphreys et al. 2006), and Santa María appears to share a common two-stage crystallization history. At Santa María volcano, a ‘snap-shot’ recording of the early stage of fractional crystallization – corresponding to slow cooling in the deep crust – is reflected in the 9.5±2.5 ka ²³⁸U–²³⁰Th isochron of the 1972 dome lava and the U–Th isotope composition of quenched mafic inclusions in the dome lavas (Figs 5 & 6b), whereas the later and more rapid decompression-driven crystallization in the conduit during magma ascent to the surface is consistent with the 1.9±1.2 ka isochron from the 1902 dacite pumice (Fig. 6a).