A key parameter in the study of magma evolution is the time scale on which magmatic processes occur. Using nanoscale secondary ion mass spectrometry (NanoSIMS), SIMS, and cathodoluminescence (CL) analyses, we have measured titanium (Ti) diffusion profiles in quartz phenocrysts from a Jurassic rhyolite of the El Quemado Complex (Patagonia, Argentina), providing new insights into the time scales of the associated volcanic processes. CL imaging of quartz phenocrysts reveals oscillatory magmatic zoning. We determined Ti concentrations with SIMS and acquired multiple NanoSIMS profiles across growth zones from core to rim. All transects show sharp changes in the 48Ti/29Si ratio, which correlate reasonably well with changes in CL intensity. Diffusion modeling of Ti in quartz yields a surprisingly short time scale for quartz crystallization of 5.6 ± 2.2 yr and a rapid crystal growth rate of 2.3 × 10−12 m/s. Based on the observed quartz textures, we suggest that the rhyolite erupted shortly after initial onset of crystallization, followed by decompression-driven quartz dissolution during fast magma ascent. We further argue that the observed oscillatory zoning and the variation of the Ti concentration of the quartz phenocryst does not reflect temperature, pressure, or titanium activity (aTi) changes of the magmatic system, but rather is the result of growth kinetics, which has important implications for the Ti-in-quartz thermometry.
Knowledge about volcanic systems has increased substantially in the last decades. Nevertheless, understanding the “trigger” for large eruptions is still a major scientific challenge, and its resolution would help in risk evaluations. A key parameter in the study of magma evolution is the time scale on which thermodynamic variations occur in magmas. Combined radiometric dating and diffusion chronology can now uncover processes taking place on time scales from a few hours to millions of years (Turner and Costa, 2007; Till et al., 2015). The basis for diffusion chronology is chemical zoning of minerals. To capitalize on the information contained in mineral zoning, one needs to be able to measure the (diffusion) profiles with a spatial resolution that is significantly better than the characteristic diffusion length scale (Saunders et al., 2014). Until now, most diffusion chronology studies have been based on grayscale profiles obtained from backscattered secondary-electron or cathodoluminescence (CL) images, or element profiles measured with electron microprobe. Such methods allow multiple diffusion profiles to be obtained rapidly, but the spatial resolution commonly limits the sharpness of diffusion profiles.
Nanoscale secondary ion mass spectrometry (NanoSIMS) offers the exciting possibility to measure compositional variations of trace elements with a sub-micrometer spatial resolution (Hellebrand et al., 2005; Charlier et al., 2012; Lloyd et al., 2014; Saunders et al., 2014; Hofmann et al., 2014; Till et al., 2015). Here we present unique data on titanium trace-element geochemistry in quartz. Quartz is a common mineral in igneous rocks. It is stable under a wide range of compositions and pressure-temperature conditions. It incorporates a significant number of different trace elements, some of which have been used as thermometers. Quartz can contain Al, Ti, as well as alkalis, and commonly records a detailed growth history with small-scale zoning patterns. In their pioneering work, Wark and Spear (2005) linked the variations in CL intensity to variation in Ti, using it for diffusion modeling to gain insight into magma evolution.
We present high-resolution NanoSIMS Ti profiles in well-preserved quartz phenocrysts from the Jurassic El Quemado Complex in Patagonia (Argentina). These profiles permit us to calculate magma residence times, which are among the shortest documented for a large volcanic system.
The volcanic El Quemado Complex is part of a large silicic igneous province, the Chon Aike province, which is associated with the Gondwana breakup. It covers large parts of Patagonia and the Antarctic Peninsula, with an estimated volume of ∼235,000 km3 (Pankhurst et al., 1998). For comparison, the volume erupted during formation of the Long Valley caldera (California, USA) is estimated at 600–650 km3 (Hildreth and Wilson, 2007). Volcanism in Patagonia occurred in three main episodes during the Jurassic. The El Quemado Complex formed during the youngest event, between 157 Ma and 153 Ma (Pankhurst et al., 2000).
SAMPLES, ANALYTICAL METHODS, AND MODELING PARAMETERS
For this study, we chose quartz phenocrysts from a kilometer-sized rhyolitic lava dome preserving magmatic flow banding and vesicles. The rocks are weakly deformed and experienced a regional metamorphic overprint at anchizone conditions (<250 °C; for details of location and regional geology see the GSA Data Repository1). The sample contains 4% quartz phenocrysts in a very fine-grained matrix of quartz and feldspar, interpreted to have been glass, with accessories of biotite, zircon, and secondary ilmenite. Quartz phenocrysts are mostly subhedral with some embayments. Some of the quartz phenocrysts contain recrystallized melt inclusions, and most display oscillatory zoning in CL images. We used microtomography (see Skora et al., 2006) to ensure that the selected phenocrysts were cut perpendicular to the c-axis. The Ti concentration was measured with the 1280HR SwissSIMS (Lausanne, Switzerland) against a standard of known composition. High-resolution Ti profiles in the quartz phenocrysts were obtained either as a linear chain of point analyses for longer traverses or using the line scan mode for higher spatial resolution with a NanoSIMS 50L instrument (CASA, Lausanne, Switzerland). Analytical details are provided in the Data Repository, but briefly, by bombardment of the Au-coated sample with an O– primary beam focused to a spot size of ∼700 nm, 29Si+ and 48Ti+ secondary ions were extracted and collected simultaneously by electron multipliers at a mass resolution of ∼6000, enough to eliminate any potential interference. The data are reported as 48Ti/29Si ratios. Before fitting error functions to the profiles, they were normalized to the maximum 48Ti/29Si ratio measured in each profile. Uncertainties on the reported diffusion times represent uncertainties from the fitting procedure, not including uncertainties of experimentally determined diffusion coefficients for Ti in quartz (Cherniak et al., 2007). For details of the fit parameters, see the Data Repository.
Determination of the crystallization temperature is notoriously difficult. Here we used an estimate based on the Ti content of the rhyolitic melt in equilibrium with rutile (Hayden and Watson, 2007). Because no rutile is present in the rhyolite, we assumed ideal mixing to calculate Ti activity as a function of temperature for the measured Ti whole-rock composition of the rhyolite. This is a good approximation for the Ti content of the melt because only minor amounts of phenocrysts are present. Simultaneous solution of the average Ti in quartz (Wark and Watson, 2006) and activity of Ti in the melt yields temperatures between 780 and 848 °C. The Ti activity in the melt clusters between 0.13 and 0.26. We use a temperature of 820 °C in our diffusion calculations; this is the mean of the different calibrations. The zircon saturation temperature (Watson and Harrison, 1983) of 770 °C obtained from whole-rock Zr content is in agreement with this estimate, because no inherited zircon cores were found and all zircons are very small (<125 µm).
CL imaging suggests magmatic oscillatory zoning (Fig. 1). The grayscale values (GSV) of the CL images correlate reasonably well with variations in Ti concentration in the quartz crystal, as has been proposed by, e.g., Wark and Watson (2006): the higher the GSV, the higher the Ti concentration. In total, the point profile (Figs. 1B and 2A) highlights three bright (higher 48Ti/29Si ratio) and three dark zones (lower 48Ti/29Si ratio). Nevertheless, it is essential to establish the diffusing element through chemical analysis because other trace elements, such as Al (Götze et al., 2001), can produce variations in the CL intensity. Diffusion rates are element specific, and hence the diffusing element needs to be established. Using the NanoSIMS allows us to use the actual chemical data for diffusion modeling.
A 200 µm traverse was first measured with individual points (Fig. 2A) to localize the Ti gradients. Comparing the sharpness of the interface between two growth zones in the point profile (∼10 µm) with those of the line scans (∼5 µm) highlights the spatial resolution gained in the line scans (Figs. 2B–2D). The width of each interface is 5–10 times that of the physical resolution of the NanoSIMS beam of ∼700 nm; hence, only limited smoothing of the profile is due to the analytical method. These line scans were used for diffusion modeling. Obtained diffusion times at the average temperature of 820 °C, from core to rim, are: 5.6 ± 2.2 yr, 4.7 ± 2.0 yr, and 4.1 ± 2.2 yr.
The use of microtomography permits cutting through the morphological center of individual phenocrysts. Assuming that crystals grow isometrically, as suggested by their overall morphology, this reduces the probability of oblique cuts. Hence, we minimize uncertainties in the geometry of the Ti-in-quartz zoning. Similarly, the spatial resolution of the analytical method used is crucial for diffusion chronometry (see, e.g., Saunders et al., 2014; Till et al., 2015). To study magmatic processes at an annual time scale, a spatial resolution in the sub-micrometer range is essential. We measured profiles with a spatial resolution of ∼700 nm. This corresponds to a time interval of 0.6 yr (or 7.2 mo) at 820 °C for Ti diffusion in quartz, using Cherniak et al. (2007).
One of the largest error sources in diffusion chronology is the temperature estimation at which diffusion took place (Spear, 2014). The maximum temperature for quartz growth of 848 °C would result in a diffusion time for the quartz of 2.5 ± 1.0 yr, while the minimum of 780 °C would result in 17.9 ± 6.0 yr for the innermost profile (Fig. 2B). Slow cooling of the rhyolite after eruption could further flatten the profiles. This would further decrease the residence time in the magma before eruption. We expect this effect to be small because we are looking at approximately kilometer-long, relatively thin dome-like eruptions around a fissure.
An additional difficulty arises from the estimation of the initial shape of the Ti profile prior to diffusion. While we believe (see below) that initial profiles were most likely already somewhat smooth (see also Till et al., 2015), we nevertheless assume a step function for the initial profile because this yields maximum diffusion times and represents a sensible end member.
Based on the discussion above, we propose that the maximum residence time for this quartz phenocryst is on the order of 5.6 ± 2.2 yr (assuming 820 °C). The residence times calculated are the same within error for all core-to-rim profiles measured. Nevertheless, it is interesting to note that the obtained residence times decrease from the center to the rim, resulting in a minimum growth rate of the quartz crystals of 2.3 × 10−12 m/s.
Our data reveal a surprisingly short time scale for crystallization of the quartz crystals. Note that the phenocrysts show rhythmic zoning but no dissolution events inside the quartz. Dissolution did occur in many crystals at the rim. Both dissolution and new phenocryst rim growth are commonly interpreted to reflect changes in temperature, pressure, or water activity of the melt (aH2O) in response to heating, decompression, and/or addition of volatiles to the system (Bachmann et al., 2002; Cashman and Blundy, 2013). The absence of internal quartz resorption textures demonstrates that quartz was stable throughout its growth history. Nonetheless, the embayment features in the quartz crystal indicate quartz dissolution shortly prior to or during eruption. Consequently, only one major temperature and/or pressure change occurred—just prior to or during eruption. The subtle changes in Ti concentration between the different growth zones are on the order of only a few parts per million (27.6 ± 1.6 to 33.8 ± 2.0 ppm). Assuming Ti activity to be buffered by the whole-rock composition at 0.18, this results in temperature variations of ∼30 °C using the Ti-in-quartz thermometer of Wark and Watson (2006). Similarly, using the calibrations of Thomas et al. (2010) and Huang and Audetat (2012), a pressure change of ∼1.1 kbar and ∼0.8 kbar, respectively, would also result in the observed Ti variations in quartz. It is, however likely that these temperature and pressure changes would result in quartz phenocryst resorption, as quartz stability is very sensitive to changes in temperature and pressure. Consequently, we advocate that the observed zonation is the result of disequilibrium growth of the crystal, analogous to the suggestion by Fowler et al. (2002) for zircon. A growth rate–dependent incorporation of trace elements during mineral crystallization from a silicic melt has been proposed (e.g., Watson and Liang, 1995; Ginibre et al., 2002) and experimentally observed by Huang and Audétat (2012) for quartz in a fluid. The fast quartz growth rates determined here approach laboratory growth rates. We suggest that crystal growth rates are fast relative to diffusion of Ti in rhyolitic melts. This would lead to the formation of an enriched boundary layer in front of the growing crystal. An oscillatory mode of growth would then result in varying degrees of enrichment. Such a mechanism would cause exponential changes in elemental concentration, where the initial slope depends on the relative growth rates and Ti diffusion in the melt. Using an exponential profile instead of a step profile for diffusion modeling would result in even shorter diffusion times. It underscores the fact that the growth time of a few years determined here are maximum values—indeed, much shorter times of a few days or months are likely. Our results further imply that one must use Ti-in-quartz thermometry carefully, even if buffered by TiO2, when quartz crystals show oscillatory zoning. This could also influence our estimate of temperatures presented above. Nevertheless, based on the fact that Ti concentrations vary only slightly, from 27.6 ppm to 33.8 ppm, we conclude that the temperatures obtained from estimation of Ti activity in the melt and the concentration in the crystal are robust.
The time scale of magma accumulation prior to large eruptions has been the subject of several recent studies making use of diffusion chronometry (e.g., review by Cashman and Giordano, 2014). In their recent work, Gualda et al. (2012) presented different lines of evidence to constrain the time scales of quartz crystallization within the Bishop Tuff (California, USA), one of the most studied magma bodies. On the basis of Ti diffusion in quartz, kinetics of melt inclusion faceting, crystal size distributions, and modeling, they suggested quartz crystallization within 500–3000 yr before eruption. They also derived growth rates for quartz phenocrysts that are typically two orders of magnitude slower (1.0 × 10−14 m/s) than those obtained in this study. Chamberlain et al. (2014), in their work on the Bishop Tuff, included Ti-in-quartz and Fe-Mg–in–pyroxene profiles from which they calculated time scales of <150 yr, similar to the estimates given by Wark et al. (2007). Studies from the Taupo Volcanic Zone (New Zealand) by Matthews et al. (2012), based on Ti profiles within quartz, suggest that quartz rims grew 10–85 yr before eruption. Overall, studies on the Bishop Tuff and other magmatic systems report decadal to centennial time scales. Finally, the high-spatial-resolution analyses of sanidine phenocrysts from Yellowstone (Wyoming, USA) in the recent study by Till et al. (2015) indicate a time scale of only several months to tens of months for rejuvenation of the silicic magma. Using the spatial resolution of the NanoSIMS, our data provide compelling evidence for sharp internal boundaries providing robust, albeit very short, time scales. We constrain quartz crystallization times on the order of years for a large magmatic system—in agreement with recent works using feldspar data (Druitt et al., 2012; Till et al., 2015).
The single growth event of quartz phenocrysts, the small phenocryst amount in the rhyolite, as well as the short quartz residence time strongly suggest that the rhyolite erupted shortly after initial onset of crystallization from a lower-crustal magma chamber. Decompression-driven quartz dissolution occurred during magma ascent, resulting in the observed embayments on the rims of the crystals. We propose that the small- to medium-sized eruptions that built up the El Quemado Complex were generated by ascending isolated magma batches from lower- to mid-crustal magma chambers, similar to recent models of fast magma transfer through a dike-like feeder system (Scandone et al., 2007; Castro et al., 2013). Our data thus add to a growing body of evidence that consider batch-like magma supply to be the normal mode in many silicic systems.
We thank the authorities of the Parque Nacional de los Glaciares (Argentina) for their collaboration. We thank A. Kosmal (El Chaltén), our field assistants, and P. Vonlanthen for their help. Excellent reviews by F. Spear, D.J. Cherniak, and B. Welsch greatly helped to improve the manuscript. We acknowledge funding by the Swiss National Science Foundation and the Center for Advanced Surface Analysis (funded by the Swiss University Conference and École polytechnique fédérale de Lausanne).