Rejuvenation of previously intruded silicic magma is an important process leading to effusive rhyolite, which is the most common product of volcanism at calderas with protracted histories of eruption and unrest such as Yellowstone caldera (Wyoming), Long Valley caldera (California), and Valles caldera (New Mexico) in the United States. Although orders of magnitude smaller in volume than rare caldera-forming supereruptions, these relatively frequent effusions of rhyolite are comparable to the largest eruptions of the 20th century, and pose a considerable volcanic hazard. However, the physical pathway from rejuvenation to eruption of silicic magma is unclear, particularly because the time between reheating of a subvolcanic intrusion and eruption is poorly quantified. This study uses nanometer-scale trace element diffusion in sanidine crystals to reveal that rejuvenation of a near-solidus or subsolidus silicic intrusion occurred in ∼10 mo or less following a protracted period (220 k.y.) of volcanic repose, and resulted in effusion of ∼3 km3 of high-silica rhyolite lava at the onset of Yellowstone’s last volcanic interval. The future renewal of effusive silicic volcanism at Yellowstone will likely require a comparable energetic intrusion of magma that remelts the shallow subvolcanic reservoir and generates eruptible rhyolite on month to annual time scales.
REJUVENATION AND ERUPTION OF RHYOLITE LAVAS AT YELLOWSTONE CALDERA
Thermal rejuvenation of near-solidus magma reservoirs as a mechanism leading to silicic volcanic eruptions is implicated by cargoes of recycled crystals with inherited isotopic compositions (Bindeman and Valley, 2001, 2003) and multimodal age populations (Charlier et al., 2005; Bacon and Lowenstern, 2005). Eruptible silicic magma may be produced by wholesale remelting of subsolidus intrusions (Simakin and Bindeman, 2012) or periodic expunging of crystal-poor rhyolite from near-solidus but molten bodies of crystal-rich magma (Charlier et al., 2005; Bachmann et al., 2014) that may persist for as much as hundreds of thousands of years (Bachmann and Bergantz, 2004; Wotzlaw et al., 2013). Common to both silicic magma production paths is the trigger of thermal rejuvenation by intrusion of new hot magma. Despite numerous studies investigating the longevity of silicic magma bodies, the time scales associated with rejuvenating near-solidus or subsolidus intrusions at calderas with protracted histories of episodic volcanism and the subsequent thermochemical evolution leading to eruption remain poorly resolved.
Effusion of rhyolite is the most frequent type of eruption within Yellowstone caldera (Wyoming, USA) and the most probable type of future volcanic eruption (Christiansen et al., 2007; Girard and Stix, 2012). At least 23 distinct eruptions of rhyolite lava have occurred within Yellowstone since the caldera-forming eruption at 640 ka, largely distributed into three intervals of volcanism separated by tens to hundreds of thousands of years of repose (Christiansen, 2001; Christiansen et al., 2007; Watts et al., 2012). The onset of these major intervals of intracaldera volcanism are marked by the eruption of rhyolite lavas (Christiansen, 2001) with disequilibrium mineral assemblages and/or low 18O/16O compositions indicating thermal rejuvenation and cannibalization of subvolcanic intrusions that were altered by the caldera’s hydrothermal system during protracted volcanic repose (Bindeman and Valley, 2001; Pritchard and Larson, 2012). The youngest interval of post-caldera volcanism ended an ∼220 k.y. period of volcanic quiescence by erupting ∼2–3 km3 of compositionally identical rhyolite to form the Scaup Lake (SCL) and South Biscuit Basin rhyolite lava flows at ca. 260 ka (Christiansen et al., 2007; Girard and Stix, 2009). This study targets the SCL rhyolite because its mineralogic and petrologic characteristics reflect the most recent major thermal rejuvenation that led to the growth of a voluminous reservoir responsible for episodic eruptions of rhyolite lavas from ca. 170 to 75 ka (Watts et al., 2012) (Fig. DR1 in the GSA Data Repository1).
The SCL rhyolite contains ∼30% phenocrysts of reversely zoned quartz, clinopyroxene, orthopyroxene, plagioclase, and sanidine with accessory zircon and Fe-Ti oxides. The rims of the phenocrysts indicate crystallization from hotter and less-evolved rhyolite melt (Fig. 1; see the Data Repository), consistent with thermal rejuvenation of near-solidus or subsolidus intrusions followed by cooling and renewed crystallization (Girard and Stix, 2009). For example, SCL plagioclase and sanidine phenocrysts have ubiquitous rims with sharp boundaries that truncate multiple interior domains (Fig. 2). The rims are enriched in Ba, Sr, Ca, and in some cases Mg and Ti, relative to the adjacent grain interiors, consistent with final stage growth from a hotter rhyolite melt formed by injection of a less evolved magma or remelting of feldspar and mafic phases (Bachmann et al., 2002). The ubiquity of this rim on SCL sanidine suggests that it represents a system-wide magmatic event. Two-feldspar thermometry using these plagioclase and sanidine rims yields temperatures of 819 ± 20 °C assuming a pressure of 3 kbar based on geophysical observations of the depth of the current magma reservoir (Farrell et al., 2014), and quartz geobarometry (Girard and Stix, 2012) (see the Data Repository).
DIFFUSION OF ELEMENTAL ZONING IN SANIDINE
We employ diffusion modeling of fine-scale Ba, Sr, and Mg zoning across the inner boundaries of the high-temperature rims on SCL sanidine to constrain the timing of magma remobilization and rim growth relative to eruption for the most recent major rejuvenation of the Yellowstone reservoir. To effectively resolve short trace element profiles, we used a NanoSIMS 50L ion microprobe to measure concentration profiles with 0.3 μm spacing over ∼7 µm. The crystallographic orientations of the concentration profiles were measured via electron backscattered diffraction and used to determine the effective diffusivity for elements with anisotropic diffusion in sanidine. The measured concentration profiles were modeled with a misfit-minimization forward-modeling approach that utilizes an analytical solution to the diffusion equation appropriate for an abrupt change in composition far from the rim and/or where the diffusive distance is small, and a systematic brute-force method for model space searching (see the Data Repository).
If a step function represents the initial concentration profile prior to diffusion, then Ba self-diffusion across the inner rim boundary indicates an elapsed time between crystallization of the rim and eruption of (n = 4 grains, 95% confidence) at 819 °C, whereas the Sr profiles indicate an elapsed time of yr (n = 4 grains, 95% confidence) (Fig. 3). Three of four Mg profiles are flat, consistent with faster diffusivity of Mg, with the other sanidine profile (grain 04-43) yielding an elapsed time of yr at 819 °C (see the Data Repository). A critical observation is that the Ba, Sr, and Mg profiles have equivalent widths (4–7 μm) (Fig. 2; see the Data Repository), a characteristic that rules out simple diffusion from an initial step function given their contrasting experimentally determined diffusivities (Table DR1; see footnote 1). If changes in the melt composition due to the resorption of feldspar and mafic minerals during remelting with or without mixing were not instantaneous compared to the time scale of rim growth, then the initial concentration profile would not be a step function. Instead, the initial concentration profile would resemble a relaxed diffusion profile but in fact represent a changing concentration of trace elements in the host melt during crystal growth (Fig. 2). Given the mismatch between the calculated time scales for Ba, Sr, and Mg and the observation that their profile shapes are identical when scaled to the same relative concentration difference, we interpret that the chemical zoning recorded in the sanidine rims nearly entirely represents crystal growth from an evolving melt composition in the aftermath of rejuvenation with little to no subsequent diffusion. Therefore, the time scales calculated assuming a stepped initial condition for the individual Ba, Sr, and Mg profiles (Fig. 3) significantly overestimate the time scale between crystal growth and eruption. Instead, the time interval between sanidine rim growth and eruption for all of the sanidine phenocrysts examined is constrained to <40 yr, which is the time required for the Sr concentration profile to diffusively diverge in shape and length from the Ba profile at the resolution of the NanoSIMS analyses. A decade is necessary for Mg to diffuse from the shape of the Ba concentration curve to the flat profiles recorded in 3 of the sanidine analyzed, so together these observations place an upper limit of ∼10–40 yr on the time scale between the start of rim growth and eruption for the SCL magma as a whole. For grain 04-43, which has a resolvable step in Mg concentration across the rim-core boundary, the diffusive time scale recorded by fast diffusing Mg refines the time interval to <0.8 yr or 10 mo, suggesting that this phenocryst records the final period of remobilization before eruption (Fig. 3). These short durations between rejuvenation and eruption are consistent with crystal growth rates for feldspars in silicic magma based on experiments and natural samples (e.g., Swanson et al., 1989; Zellmer and Clavero, 2006). At appropriate degrees of undercooling, experimental crystallization rates (∼10−10 cm/s) predict ∼15 yr for growth of the entire sanidine rim (∼500 μm) and 1.5–3.5 mo for the growth of the 4–7 μm concentration gradients (see the Data Repository). Any time elapsed between sanidine dissolution in response to rejuvenation and the onset of renewed crystallization must be <∼1200 yr due to preservation of 18O/16O disequilibrium in the resorbed cores of SCL zircons (Bindeman et al., 2008), but is likely much shorter because the entire population of SCL sanidine crystals would be consumed in only ∼30–45 yr within heated rhyolite given their rate of dissolution (see the Data Repository). Diffusion during SCL lava emplacement is likely to have been insignificant because silicic lavas are typically emplaced as a series of quickly quenched lobes (e.g., Tuffen et al., 2013).
The renewed precipitation of sanidine at higher temperatures may reflect oversaturation triggered by magma ascent and concomitant exsolution of dissolved H2O in the SCL magma (Almeev et al., 2012), addition of CO2 by new magma (Wark et al., 2007), and/or relatively limited addition of K-Na–enriched melt derived from remelting of sanidine-rich cumulates (Bachmann and Dungan, 2002). Reverse zoning in sanidine and other phases reflecting new crystallization after remobilization is recognized for other silicic magmas at calderas, such as the voluminous Fish Canyon Tuff (Bachmann et al., 2002).
The results of this study illustrate that the frequently made assumption in geospeedometry of a step-function initial condition can be inaccurate despite petrographic evidence for resorption, and can be addressed by interrogating diffusion time scale concordance between multiple trace elements that are geochemically similar. Other studies recognize the importance of alternate initial conditions for diffusion modeling (Costa et al., 2008), including deconvolution of the effects of simultaneous crystal growth and diffusion of Ba and Sr in sanidine (Chamberlain et al., 2014). By employing a third faster diffusing element (Mg), this study reveals that Ba and Sr profiles across the sanidine core-rim boundary reflect changes in melt composition rather than diffusive relaxation across a sharp boundary.
MONTHS FOR REJUVENATION TO ERUPTION
The high-spatial-resolution analyses of sanidine phenocrysts in this study indicate a time scale of only several months to dozens of months for rejuvenating silicic intrusive material at Yellowstone to produce its most frequent type of eruption after a protracted period of volcanic repose, during which the subvolcanic system existed as near-solidus or subsolidus state waiting for remobilization (e.g., Cooper and Kent, 2014). These brief time scales are remarkably similar to others calculated for eruption triggering, but in those cases due to magma mixing in established, liquid-dominated silicic magma chambers (10–60 yr, Matthews et al., 2012; ≤100 yr, Druitt et al., 2012). The recurring eruptions of rhyolite lava at Yellowstone caldera are individually comparable in size (∼1–10 km3) to recent silicic eruptions at Chaiten (Chile) or Pinatubo (Philippines), and would likely lead to considerable social and economic disruption and possible interference with North American air travel. If the ∼3 km3 volume for the SCL and South Biscuit Basin rhyolites is derived from a subvolcanic area outlined by their vent zones, then a ≥40–60-m-thick region of near-solidus or subsolidus rhyolite may have been remobilized. Recent physical modeling indicates that a near-solidus body of this size could be rejuvenated by an intrusion of hot, near-liquidus rhyolite within several years or less (Simakin and Bindeman, 2012; Huber et al., 2012), consistent with our results. Today, Yellowstone’s vigorous hydrothermal system circulates through a several-kilometers-thick cap of subsolidus, fractured rhyolite above a shallow crustal reservoir (∼10,000 km3) of near-solidus magma mush containing 5%–32% or ∼200–600 km3 of rhyolitic melt (Farrell et al., 2014) and a near-solidus lower crustal reservoir (∼46,000 km3) containing ∼2% or ∼900 km3 of basaltic partial melt (Huang et al., 2015); both are ultimately sustained by intrusion of mafic magma at a rate similar to the active Hawaiian hotspot (Lowenstern and Hurwitz, 2008). The results of this study reveal that a sufficiently energetic rejuvenation of Yellowstone’s shallow crystal-melt mush and/or hydrothermally altered wall rock could lead to an effusive eruption within months. Fortunately, any significant rejuvenation of the reservoir is likely to be associated with deformation or seismicity and identifiable by geophysical monitoring (e.g., Fialko et al., 2001; Wicks et al., 2006).
This research benefitted from discussions with J. Lowenstern, C. Bacon, T. Sisson, R. Christiansen, H. Wright, N. Matthews, and S. Hurwitz at the U.S. Geological Survey (USGS), and manuscript reviews by O. Bachmann, C. Barnes, K. Cooper, C. Deering, J. Lowenstern, L. Mastin, and two anonymous reviewers. This work was supported by a USGS Mendenhall Postdoctoral Fellowship and Arizona State University funding to Till. We thank C. Hitzman for expert help with the Stanford University NanoSIMS and E. Miranda for electron backscatter diffraction analyses at California State University–Northridge.